摘要:

Tutorial Review Getting more from your laboratory control charts Eamonn Mullins Industrial Statistics Unit, Department of Statistics, Trinity College, Dublin 2, Ireland Received 9th November 1998, Accepted 8th February 1999 1 Introduction 2 Case study 2.1 The X-bar chart 2.2 The range chart 2.3 Chart analysis 2.4 Appropriate measures of variability 3 Control chart performance 3.1 Average run length analysis 3.2 ARL analysis: general application 4 Improving precision by replication 4.1 Effects of averaging 4.2 Components of analytical variability 4.3 Estimating the components of analytical variability 4.4 Effects of within-run replicates on test result precision 4.5 An alternative approach: between-run replicates 5 Measures of repeatability and intermediate precision 5.1 Repeatability 5.2 Reproducibility 5.3 Intermediate measures of precision 6 Conclusion 7 Acknowledgements 8 Appendix 8.1 Range charts 8.1.1 Action and warning limits 8.1.2 Probability limits 8.2 Converting from average range to standard deviation 8.3 Standard deviation charts 9 References 1 Introduction Statistical quality control (SQC) charts are used routinely in the analytical laboratory to monitor the stability of the measurement process.They are an excellent tool in assuring both the working analyst and laboratory accreditation assessors that the laboratory’s analytical systems remain in control, i.e., stable. The purpose of this paper is to show that, in addition to this primary function, SQC charts can also provide valuable information which allows both criticism of the performance of the charts themselves and also useful analysis of the quality of measurement of test samples.By careful analysis of control charts data, the analyst can determine if the measurements are fit for purpose, decide whether the number of replicates may be reduced or investigate possible assay improvement strategies. The discussion begins with a case study from a pharmaceutical QC laboratory.The case study serves several purposes. For those not familiar with SQC charts, it provides an introduction to their construction and use, and warns against using inappropriate measures of assay variability. It will, I hope, be of interest in itself, as there are few case histories in the analytical literature which illustrate the kinds of problems which may arise in setting up and using control charts. Finally, the data from the case study are used to illustrate and make concrete the ideas discussed in the later sections of the paper. 2 Case study The data in Table 1 are from a potency assay of a control material which is to be used in monitoring the routine analysis of a pharmaceutical product to ensure stability of the analytical process. The assay protocol is as follows. Ten tablets are weighed and transferred to an extraction vessel and 500 ml of methanol are added. The vessel is vibrated for 12 min. The material is then filtered.Dilution takes place in two stages, 5 + 45 and then 5 + 95. An internal standard is added at the second dilution and a vial is filled for HPLC analysis. Three independent replications of this procedure are carried out. Potency is determined in milligrams per tablet. The data were collected over several weeks, with varying numbers of measurements per day, as a preliminary data collection exercise with a view to setting up control charts for the assay.There are 29 sets of three replicates, each set being the work of a single analyst. A total of eight analysts were involved in generating the data. 2.1 The X-bar chart The control chart most commonly used in analytical laboratories is the X-bar chart.1–4 The X-bar chart is a plot, in time order, of the averages of the replicate measurements on the control material. Three lines are placed on the chart to aid interpretation, a centre line and upper and lower control limits. Where reference materials are in use for control purposes and the assay is known to be unbiased, the centre line is set at the assigned value.Where there is no assigned value, as in the current case, the overall average of the data is used as the centre line. The control limits (also known as action limits) are placed three standard errors above and below the centre line. The standard error is the standard deviation of the averages; this is discussed further below. When the plotted data vary at random between the control limits, the analytical system is said to be ‘in Eamonn Mullins is a Senior Lecturer in the Department of Statistics, Trinity College, Dublin and is a member of the Irish National Accreditation Board, which is responsible for laboratory accreditation in Ireland.He has wide industrial consultancy experience and has run many training courses in the use of statistics in both laboratory and manufacturing environments. Analyst, 1999, 124, 433–442 433statistical control’.Points outside the control limits, long runs of points on one side of the centre line, or runs upwards or downwards indicate analytical system instability—the system is said to be ‘out-of-control’. Fig. 1(a) shows an X-bar chart based on the data in Table 1. As noted above, the control limits are calculated as CL ± 3SX-bar (1) where the centre line, CL, is the overall mean of the data. The standard error SX-bar can be calculated directly from the 29 replicate means using the usual equation for standard deviation. Alternatively, SX-bar can be calculated as S R X-bar = 1 1 128 .(2) where the constant in the denominator can be found in Table A1 (see Appendix) and øR 1 is the average of the magnitudes of the differences between successive plotted points, i.e., |x2 2 x1|, |x3 2x2|, etc. These ‘moving ranges’ are a reflection of the random analytical variability when the system is stable. The two methods of estimation should give similar results for in-control data, but the moving range method is preferable in establishing preliminary control limits, as it is less sensitive to large outliers.Thus eqn. (2) gives SX-bar = 18.80 for the 29 means, while the direct estimate gives 23.13, owing to a large outlier. The X-bar chart of Fig. 1(a) is useful primarily for detecting bias in the analytical system; this could, for example, be a shortterm shift (spike), a persistent shift or a drift in the response of the system.A different chart is required for monitoring analytical precision. 2.2 The range chart Analytical precision is concerned with the variability between repeated measurements of the same analyte, irrespective of the presence or absence of bias. When several replicate measurements are made within one analytical run, their range, i.e., the difference between the largest and smallest values, can be used to monitor the stability of analytical precision.In the case of Table 1, three replicate measurements on the control material were made in each analytical run and Fig. 1(b) is based on the 29 ranges. The range chart has the same general features and is used in the same way as the X-bar chart: it has a centre line and upper and lower control limits and similar rules for detecting out-ofcontrol conditions are applied to the plotted points. The basis for calculating the control limits is different, since sample ranges do not follow a normal distribution. This technical difference presents no difficulties in practice as tables of constants have been prepared which simplify the calculation of the control limits. The centre line of the chart is the average, øR 2, of the within-run ranges; the control limits are obtained by multiplying the average range by multipliers shown in Table A1 (Appendix).The multipliers used may correspond to three standard error limits, or probability limits may be preferred.Fig. 1(b) Table 1 Twenty-nine sets of three replicate potency measurements (mg) Run Potency Run Potency Run Potency Run Potency 1 499.17 492.52 503.44 9 487.21 485.35 479.31 17 484.17 490.72 493.45 25 491.27 488.90 500.77 2 484.03 494.50 486.88 10 493.48 496.37 498.30 18 493.61 488.20 503.90 26 489.85 488.42 487.00 3 495.85 493.48 487.33 11 553.72 554.68 500.71 19 482.25 475.75 488.74 27 492.45 484.96 490.58 4 502.01 496.80 499.64 12 495.99 499.36 482.03 20 459.61 465.03 465.57 28 198.92 479.95 492.15 5 463.99 457.61 469.45 13 511.13 504.37 501.00 21 509.11 510.18 506.46 29 488.68 476.01 484.92 6 482.78 484.65 524.30 14 510.16 498.59 501.48 22 489.67 487.77 497.26 7 492.11 485.58 490.24 15 479.57 462.64 479.57 23 487.82 489.23 493.45 8 500.04 499.11 493.98 16 494.54 493.99 495.08 24 489.23 491.11 484.07 Fig. 1 Full dataset: (a) X-bar chart and (b) range chart. 434 Analyst, 1999, 124, 433–442shows a range chart for the pharmaceutical data, using three standard error limits. 2.3 Chart analysis The control limits in Fig. 1 are required to act as yardsticks by which the future stability of the analytical process will be monitored. It seems clear that they will provide useful signals of future instability only if they themselves are based on data that reflect in-control conditions. It is important, therefore, before accepting the control limits and centre lines as aids for routine monitoring of future assays, to assess the stability of the analytical process that generated the data on which they are based.Examination of Fig. 1 shows that run 28 is a very clear outlier in both charts. Inspection of the data in Table 1 shows one extremely low point (198.92) in run 28. Investigation of this point led the laboratory manager to believe that the analyst had omitted to vibrate the extraction vessel as required by the protocol. This would account for the very low value as, without vibration, the active ingredient would not be extracted fully from the tablets.At this stage set number 28 was excluded from the dataset, as it was considered to reflect unstable conditions, and the two charts were redrawn as Fig. 2. Note that the average range has almost halved from 21.5 in Fig. 1(b) to 11.8 in Fig. 2(b), after exclusion of the large outlier. Also, in Fig. 2(b) the upper control limit for the range chart has dropped from 55.3 to 30.3. This is a dramatic illustration of the impact that outliers can have on a statistical analysis and emphasizes the need for careful data scrutiny when setting up control charts.Exclusion of set 28 has also had an important impact on the X-bar chart. The centre line has shifted from 488.5 units in Fig. 1(a) to 492 units in Fig. 2(a). The control limits are ±40.2 units from the centre line in Fig. 2(a), whereas they are ±56.4 units from the centre line in Fig. 1(a). Fig. 2(b) shows runs numbers 6 and 11 as outliers. Fig. 2(a), the corresponding X-bar chart, also shows run 11 as an outlier.Investigation of the two sets of results showed that they were produced by the same analyst. These results, together with other results on routine test samples, suggested problems with this analyst’s technique. For example, when analysing a batch of test samples this analyst tended to filter all samples in the batch before completing the dilution stages. This meant that open beakers of filtered material were left lying on the bench for various amounts of time.This would allow the methanol to evaporate and thus concentrate the remaining material, leading to high results. Modification of the analyst’s technique (including ensuring that the analysis of one test sample is completed before filtration of a second) gave results in line with those of the other analysts. The investigation also called into question the ruggedness of the assay and led to further investigation of the analytical operating procedure.At this stage sets 6 and 11 were also excluded from the dataset and the charts redrawn as Fig. 3. The average range has decreased substantially again and is now 9.0 in Fig. 3(b). Both charts appear to reflect a measurement process that is in control: there are no points outside the control limits and no extensive runs in the data. Despite this, the laboratory manager had doubts about runs 5 and 20 in the X-bar chart (because of the exclusion of sets 6 and 11, set 20 is the 18th value in Fig. 3). Excluding these low points from the calculations of the control limits would give narrower limits, which would signal the two points as being out of control. However, there were no obvious explanations for either of these points representing out-ofcontrol conditions. Also, Fig. 3(a) indicates that the deviations of these points from the centre line are consistent with chance variation, as judged by its limits. The laboratory manager felt that it was important not to start with control limits that would be too narrow, since this would lead to disruptive false alarm signals.Accordingly, she decided not to exclude them from the dataset. If indeed these points do reflect unusual behaviour then, as more data accumulate, the control limits can be redrawn to reflect the better analytical performance that is being achieved. 2.4 Appropriate measures of variability This preliminary data collection exercise gave initial control limits (those of Fig. 3) for use in the routine monitoring of the analytical system. Note that different estimates of variability were used in constructing the X-bar and range charts. In the range chart the variability that is being monitored is that which Fig. 2 After exclusion of set 28: (a) X-bar chart and (b) range chart. Fig. 3 After exclusion of sets 6, 11 and 28: (a) X-bar chart and (b) range chart. Analyst, 1999, 124, 433–442 435arises under essentially repeatability conditions, i.e., one analyst makes three independent measurements on the same material within a short time interval under the same conditions. In the Xbar chart, the control limits are based on the variation between the averages of the three replicates.This variation contains two components: it includes the repeatability variation, but also contains the chance run-to-run variation which arises from many sources, including day-to-day, analyst-to-analyst and HPLC system-to-system variation.In discrete manufacturing applications of control charts, the usual practice is to select from the process at regular time intervals a small number of products or components and to measure some critical parameter of the selected items. The range chart is based on the within-sample ranges, just as the range chart in analytical applications is based on the within-run ranges. In the manufacturing context the within-sample range is, in most cases, also used as a basis for drawing the X-bar chart.The assumption behind this practice is that all the random variation is captured in the within-sample variation and that differences between sample averages either reflect this variability, or else are due to ‘assignable causes’, i.e., disturbances which reflect out-of-control production conditions. This assumption works well in very many cases, but not universally, and needs to be investigated.5,6 In the analytical context, however, the experience is that, superimposed on the within-run chance variation, there is also further random variation, which is associated with chance perturbations that affect separate analytical runs differently.2 Some of the possible causes are listed above.This second component of chance variation is referred to as between-run variation and, as will be seen later, can be more important in determining the precision of analytical results than is the within-run variation. Fig. 4 shows the results of incorrectly using the within-run variability as a basis for control limits for the X-bar chart.The correct limits, shown in Fig. 3(a), are given by the centre line ±33.1 units, while those in Fig. 4 are ±9.2 units. The limits shown in Fig. 4 are too narrow and, if used routinely, would lead to many false alarm signals which would cause severe disruption to the analytical process. This shows the dangers of following the recipes for control chart construction which appear in most engineering quality control textbooks (probably the major source of such information) and which are inappropriate in the analytical context. 3 Control chart performance The ability of X-bar charts to detect changes in an analytical system depends on various factors, including the inherent variability of the analytical process, the number of replicate measurements made and the rules used to signal an out-ofcontrol state. In this section the effects of these factors on the time to detection of a shift in the analytical system are examined and illustrated using the case study data. 3.1 Average run length analysis Various performance measures are used to characterize the ability of control charts to detect problems with analytical systems. Perhaps the most intuitively appealing of these is the average run length (ARL), i.e., the average number of points plotted until an out-of-control signal is found. Here the word ‘average’ carries the same long-run interpretation as, for example, the statement that if a coin is tossed 10 times, then, on average, we expect five heads. We do not interpret the coin tossing statement to mean that if it is done once, the result will be five heads and five tails.So also, in any particular use of a control chart, the actual run length, i.e., the number of points plotted before an out-of-control signal occurs, may be either less than or greater than the ARL. Because the distribution of run lengths is skewed, the individual run lengths will be less than the ARL more than 50% of the time, but occasionally they will greatly exceed the ARL.Table 2 is based on the pharmaceutical case study data and shows ARLs for different shifts in the mean level (i.e., biases) under three rules for signalling out-of-control conditions for the X-bar chart: (i) rule (a) a single point outside the action limits (±3 standard errors); (ii) rule (b) includes rule (a) but also recognizes as an out-ofcontrol signal a run of eight points above or below the centre line; (iii) rule (c) includes rule (a) but also recognizes two points in a row between the warning limits (±2 standard errors) and the action limits as an out-of-control signal. For a shift in the mean level, the signal will be triggered by two points between the warning and action limits on one side of the centre line.The ARLs are tabulated in bias steps from zero up to a bias of 33.1 mg. Inspection of Table 2 shows that the average run length is 370 while the system remains in control and action limits only are used, i.e., on average, not until 370 analytical runs are carried out will an out-of-control signal be observed.This is one of the main reasons for choosing three standard error action limits—a very low false alarm rate is experienced. However, the consequence of ensuring that the false alarm rate is low is that the chart is not sensitive to small to moderate biases.Thus, the second column of Table 2 shows long ARLs for small biases: for a bias of 11 mg the ARL is 44, meaning that, on average, not Fig. 4 X-bar chart with control limits incorrectly based on within-run variability. Table 2 Average run lengths for three rules Bias/mg Action limits Action limits + runs rule Action limits + warning limits Bias (standard error) 0.0 370 150 280 0.0 2.2 310 110 220 0.2 4.4 200 60 130 0.4 6.6 120 34 75 0.6 8.8 72 21 43 0.8 11.0 44 15 26 1.0 13.3 28 11 16 1.2 15.5 18 8.6 11 1.4 17.7 12 7.0 7.4 1.6 19.9 8.7 5.9 5.4 1.8 22.1 6.3 4.9 4.1 2.0 24.3 4.7 4.1 3.2 2.2 26.5 3.6 3.4 2.6 2.4 28.7 2.9 2.8 2.2 2.6 30.9 2.4 2.4 1.9 2.8 33.1 2.0 2.0 1.7 3.0 436 Analyst, 1999, 124, 433–442until 44 analytical runs have been carried out will the analyst be given clear evidence of a shift in the analytical system.The effects of adding supplementary rules to the basic action limits rule can be seen in the ARLs displayed in columns three and four of the table.Using the rule of a run of eight in a row above or below the centre line, in addition to the action limits rule, has a dramatic effect on the ARL: at a bias of 11 mg the ARL is reduced to about one third of its previous value, i.e., from 44 to 15. However, when the system is in control (i.e., no bias), the average run length between false alarms is now 150, as opposed to 370 when using action limits alone. The disadvantage of this increase in the false alarm rate must be set against the reduction in the ARL when biases occur.Comparison of columns two and three shows very substantial improvements in detection rates for small biases. Addition of the runs rule has little impact for large biases, as these are quickly detected by the action limits rule on its own. Column four shows the effect of supplementing action limits with a rule that considers two points in a row between the warning and action limits as a signal of bias.This column shows ARLs intermediate between those of columns two and three, for small to moderate biases. It is clear that choosing a set of rules involves a trade-off between having long ARLs for zero biases, together with correspondingly long ARLs for moderate biases, and having short ARLs when there are moderate biases in the analytical system, together with the inconvenience of more frequent false alarm signals. The more rules in the set the shorter the ARL for zero bias, i.e., the more frequent are the false alarms, but the detection of biases will be correspondingly faster.The implications of the possible biases and how quickly they are likely to be detected may now be judged in the light of the specification limits for the product, the analysis of which is to be monitored by the control chart. If product potency is well within the specification limits, then an assay which is subject to relatively large biases can still be fit for purpose.Accordingly, potentially long ARLs will be acceptable when assessing the performance of the control chart. If, on the other hand, the manufacturing process is producing material with a potency close to a specification limit, then it will be important that measurement error does not carry the measured value over the limit, in either direction. In such circumstances, the measurement process requires tight control. It may even be necessary to reassess the assay protocol to achieve both good precision for test results and good performance characteristics for the control chart.Some ideas which are relevant to such an assessment are discussed below. One general lesson that can be learned from the ARL analysis is that out-of-control signals do not necessarily appear immediately an assignable cause begins to affect the analytical system. For large biases, the chart may be expected to produce a signal quickly, but small to moderate biases can take some time to be detected.This means that while the search for assignable causes should focus initially on the immediate neighbourhood of the signal, it should be extended backwards in time if no obvious assignable cause is found. 3.2 ARL analysis: general application To allow the ARL analysis to be applied in different practical contexts, the last column of Table 2 shows the ARLs tabulated in bias steps that are a multiple of the standard error of the plotted points. To adapt Table 2 to any given analytical system all that is required is to multiply column 5 by the appropriate standard error.Thus, for our case study, column 1 of Table 2 was obtained by multiplying column 5 by 11.046, which was calculated from the average of the moving ranges of successive points plotted in Fig. 3(a), using eqn. (2). Champ and Woodall7 presented a table of ARLs for many different combinations of rules. 4 Improving precision by replication Given time and resources, analysts would always prefer to make several measurements of the analyte of interest and report the average of the set of measurements rather than make a single determination and report the single result.It is intuitively obvious that if a measurement is subject to random error and the measurement is repeated a number of times, then positive errors will tend to cancel negative errors and the average value will be subject to less random error than the individual measurements.However, the precision of final test results depends not only on the number of replicates, but also on whether these are placed in the same or different analytical runs. The implications for assay precision are discussed in this section and illustrated using the case study data. 4.1 Effects of averaging The extent of the averaging effect for random error is illustrated in Fig. 5, which represents idealized histograms of both individual measurements and of averages of four independent measurements of a measurand.The narrower curve is called ‘the sampling distribution of the mean’, as it describes the behaviour of averages (means), based on repeated sampling, and its standard deviation is called ‘the standard error’. Mathematical theory shows that, if the standard deviation of the distribution of individual measurements is s, the standard error of the distribution of sample means is sAn, where n is the number of independent measurements on which each mean is based.Suppose the individual measurements represent purity values, in percentage point units, and that they vary about a long-run average value of 90 with a standard deviation of 0.5. It follows that averages of four measurements will also vary about a long-run value of 90, but their standard error will be 0.25. Recall that approximately 95% of the area under any normal curve lies within two standard deviations (or standard errors) of the mean. This implies that while 95% of individual purity measurements would be between 89 and 91, 95% of averages of four measurements would be between 89.5 and 90.5.The effect of the averaging process is that the result is likely to be closer to the long-run average value, i.e., the true value, on the assumption of an unbiased analytical system. Fig. 5 Idealized histograms of the distributions of individual measurements (a) and of the averages of four measurements (b). Analyst, 1999, 124, 433–442 4374.2 Components of analytical variability The simple statement of the effect of averaging given above needs to be adjusted somewhat to take account of the realities of analytical procedures. Because of the manner in which analytical workload is typically organized, the full benefits of the averaging process will not usually be realized.As will be seen below, requiring, for example, three analysts each to make a single determination of a measurand, as opposed to the usual practice of one analyst making all three determinations, carries serious implications for the precision of the final test result.The reason for this is that the variability of the final determination has two components, viz., within-run and between-run variability, and the way averaging affects the final result is determined by which way the measurements are made. When combining variability from different sources, it is necessary to work in terms of variances (i.e., squares of standard deviations), as the components of the variation will be additive in this scale. Each of the averages of the three replicates plotted in the control chart is subject to a variance component due to the within-run variability; this is sw 2/n, where n is the number of replicates on which X-bar is based (here n = 3) and sw is the within-run standard deviation.Each average is also subject to a variance component due to between-run variability (sB 2), so that the combined variance for the plotted averages is sX-bar 2 = sw 2/n + sB 2 (3) Because the three replicates are separately subject to withinrun random influences, the averaging process works, as described above, to reduce the effects of within-run variability on the test result.However, since all three replicates experience the same between-run influences, i.e., they each experience those same influences that affect their particular run, averaging has no effect on the between-run component of analytical variability.The implications of this are illustrated and discussed below. The use of control chart data to estimate these components of analytical variability will first be discussed. 4.3 Estimating the components of analytical variability The two variance components can be estimated using the summary statistics calculated for setting up the range and X-bar charts. The within-run standard deviation (sw) may be estimated directly from the range chart, which is based on the average of the within-run ranges. Factors for converting from ranges to standard deviations are given in Table A1 (Appendix).Using the average range from the final range chart [Fig. 3(b)], an estimate of the within-run standard deviation is � . . . . sw = = = R2 1 693 9 00 1 693 5 316 (4) The between-run standard deviation, sB, is estimated indirectly, using eqn. (3) re-expressed as � � � / s s s B X-bar 2 w 2 2 = - n (5) Recall that the standard error of the sample means (sX-bar) was estimated using eqn.(2). For the data on which Fig. 3(a) is based, this gives S R X-bar X-bar = = = = � . . . . s 1 1 128 12 46 1 128 11 046 (6) Using this value and the sw value calculated above, the between-run standard deviation sB is estimated as follows: � � � / ( . ) ( . ) / . . . s s s B X-bar 2 w 2 2 2 2 11 046 5 316 3 122 014 9 420 112 594 = - = - = - = n so that �s B = 10.611 We are now in a position to determine the impact of the number of replicates (n) on the standard error of the averages plotted in the X-bar chart.This quantity (sX-bar) determines the precision of test sample results. We have already seen its role in determining the performance characteristics of the X-bar chart, as described by Table 2. Hence it is a key parameter in determining both the precision of the analytical system and also our ability to monitor system stability. 4.4 Effects of within-run replicates on test result precision In very many cases the number of replicate measurements made will be decided by traditional practice rather than by analysis of the characteristics of the analytical system. Thus, in the case study described above, the test results had been routinely based on three replicate measurements and so, when control charts were initiated, three control samples were included in each analytical run.Using the estimates of within-run and betweenrun variances, the implications for the magnitude of �s X-bar of different numbers of within-run replicates can now be calculated as follows: � � / � ( .) / ( . ) . / . s s s X-bar w 2 B 2 = + = + = + n n n 5 316 10 611 28 260 112 593 2 2 (7) The results are given in Table 3. To examine the implications for the ARL performance of the X-bar chart, based on any given number of within-run replicates, the last column of Table 2 is multiplied by the corresponding standard error from Table 3. However, it is obvious from Table 3 that reducing the number of replicates from three to two or even one would have only a marginal impact on the performance characteristics of the X-bar chart in this case.The reason for this is, of course, that the calculation is dominated by the between-run variance: �s 2 B = 112.593 is much larger than �s 2 w/n = 28.2601/n, for all values of n. Reducing the number of control sample replicates to two per analytical run would increase the standard error of the result ( �s X-bar) by approximately 2%.Thus, the analytical work involved in maintaining an X-bar chart for monitoring bias could be reduced by one third with little impact on the quality of the chart. Measuring the control material only once would reduce the work even further, but would mean an increase of approximately 7.5% in the standard error and also that a range chart to monitor precision could no longer be based on the control material measurements. This, however, could be overcome by using test sample replicates as a basis for monitoring assay precision (see discussion of standard deviation charts in the Appendix).Going in the opposite direction in the table shows that there is little advantage to be gained from increasing the number of within-run replicates to five. In addition to giving information on the average run length performance characteristics of the X-bar chart as the number of replicates changes, Table 3 gives the precision with which test samples are measured, again for different numbers of withinrun replicates.If a disimprovement of 2% in the precision of test results has little impact on their fitness for purpose, then the number of replicate analyses on test samples can also be reduced. This will have a major impact on laboratory workload. Table 3replicates (n) on �s X-bar n 1 2 3 4 5 �s X-bar 11.87 11.26 11.05 10.94 10.87 438 Analyst, 1999, 124, 433–442While this is highly desirable, a laboratory manager would need strong evidence of assay stability, particularly in a regulated environment, before reducing the number of replicates.Control charts provide this evidence. Thus, control charts have a dual role to play. Used on a day-to-day basis, a control chart is a tool for monitoring the on-going stability of the analytical system and for signalling when problems occur. Analysed retrospectively, it provides documentary evidence of system stability and provides measures of the quality of routine analyses.This is discussed further in a later section. Note that while the discussion has been illustrated by the case study data, which were collected as a training set to establish control limits, in practice it would be desirable for such calculations to be based on a long series of routine control chart data. 4.5 An alternative approach: between-run replicates If either the precision of test results or the ARL characteristics of the control chart are considered unsatisfactory, then action is required.The natural response is to carry out a ruggedness study to try to identify and then reduce the sources of variability affecting the assay. A detailed ruggedness study will take time to carry out (and there is no guarantee of success in reducing variability). In the meantime, the analysis set out below may provide a short-term improvement or, where the variability cannot be reduced satisfactorily, an alternative longer term solution.The structure of the formula for the standard error of the plotted values carries implications for the way analysis might be carried out and suggests a second approach to reducing assay variability. The current protocol involves making three replicate determinations of potency within one analytical run. This applies to both test samples and the control material. The resulting standard error is � � / � s s s X-bar w 2 B 2 = + n where n = 3.If, instead, the protocol required a single determination of potency in each of three analytical runs, then the standard error of the mean of three such replicates would be � � / � / s s s X-bar w 2 B 2 = + n n (8) where n = 3. Under the current protocol only the within-run variation is averaged, since all measurements are within the one run, while under the alternative protocol averaging would take place over the between-run variation also. The numerical implications for the case study data can be seen in Table 4.In the first row of Table 4 only the within-row variance, �s 2 w = 28.260, is divided by n. The second row shows smaller standard errors, because here the between-run variance, �s 2 B = 112.593, is also divided by n. Since �s 2 B is much larger than �s 2 w, the difference is substantial. It is clear from Table 4 that the nature of the replicates has a major impact on the precision of the final measured value. The second protocol gives better precision for only two replicates than does the first for three replicates.There is nearly a 40% reduction in the standard error in switching from the first to the second protocol while keeping n at 3, and slightly more than a 50% reduction when n is 5. From a practical point of view, although the same number of measurements may be involved in both cases, there are obvious logistical considerations involved in choosing between the two protocols. Nevertheless, the alternative approach is worth considering in situations where there is large run-to-run variation in results, as it may be the only way to achieve acceptable precision. 5 Measures of repeatability and intermediate precision A common requirement in analytical method validation is that the repeatability and either the reproducibility or intermediate precision of the method be estimated; for example, see the International Conference on Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use (ICH) guidelines.8 Repeatability and reproducibility are descriptions of the precision of test results under different specifications of the conditions under which measurements are made.Repeatability allows for as little as possible variation in analytical conditions, whereas reproducibility allows for maximum variation in conditions, while still measuring the same analyte. Intermediate precision lies somewhere between these extremes, depending on the number of factors allowed to vary. These measures are usually obtained from specially designed studies but, as will be seen, repeatability and intermediate precision may also be estimated from control chart data. 5.1 Repeatability The International Standards Organization9 defines repeatability conditions as ‘conditions where independent test results are obtained with the same method on identical test items in the same laboratory by the same operator using the same equipment within short intervals of time’.To provide a quantitative measure of repeatability, the ‘repeatability limit’ is defined as ‘the value less than or equal to which the absolute difference between two test results obtained under repeatability conditions may be expected to be with a probability of 95%’. On the assumption that test results follow a normal distribution, this leads to an expression for the repeatability limit (r): r = 1.96A2sr 2 (9) where sr is the repeatability standard deviation, i.e., the standard deviation of test results obtained under repeatability conditions.In interpreting this definition in terms of our case study, the repeatability conditions may be taken to correspond to those under which the three within-run replicate measurements were made. Accordingly, the repeatability limit may be calculated as r = 1.96A2 �s 2 w (10) where �s w is the within-run standard deviation for single determinations and which, as we have seen, may be calculated directly from the range chart data.While the definition of repeatability conditions refers to a single analyst, in any practical context the definition will be understood to embody the assumption that all the analysts who carry out the analysis are fully trained, to the extent of achieving a uniform level of precision. Without such an assumption, the method requires to be separately validated for every analyst. With the assumption, a combined standard deviation based on the data from a range chart [see eqn.(4)] provides a realistic estimate of the level of repeatability precision that is being achieved routinely in the laboratory. Table 4 Estimated standard errors for means of replicates under different protocols Protocol 1 2 3 4 5 n replicates within a single run 11.87 11.26 11.05 10.94 10.87 Single measurement in each of n runs 11.87 8.39 6.85 5.93 5.31 Analyst, 1999, 124, 433–442 439Arguably, such an estimate is better than that which will result from a once-off special method validation study.Firstly, it encompasses the work of all the analysts who will be involved routinely in using the method. Once-off studies will frequently represent the work of a single analyst, often one who has more than average experience and who carries out special duties such as method validation studies. Second, it is based on a record of assay stability, without which any estimate is meaningless.Consequently, it provides a measure of precision based on a large number of replicate analyses, carried out over a relatively long time period which makes it more broadly representative, while still fulfilling the requirement of replicates being close together in time. In very many if not most cases, the reported result will be an average of several measurements; thus, in our case study, the reported result is the average of three within-run replicates. Where a measure of the repeatability of the reported value (the final value) is required, the Standards define the ‘repeatability critical difference’ to be ‘the value less than or equal to which the absolute difference between two final values each of them representing a series of test results obtained under repeatad to be with a specified probability’.Again under the assumption of a normal distribution and a specified probability of 0.95, this definition results in a slightly modified version of eqn.(10): repeatability critical difference = 1.96A2 �s 2 w/3 (11) 5.2 Reproducibility Repeatability is a very narrowly conceived measure of analytical precision, describing the performance capability of the analytical system under idealized conditions. Reproducibility, on the other hand, lies at the core of the question of measurement quality in situations where disputes are likely to arise about a test result, e.g., when vendor and customer, or manufacturer and regulatory authority, disagree on the value of some measured quantity.The ISO9 defines reproducibility conditions as ‘conditions where test results are obtained with the same method on identical test items in different laboratories with different operators using different equipment’. The ‘reproducibility limit’ is defined as ‘the value less than or equal to which the absolute difference between two test results obtained under reproducibility conditions is expected to be with a probability of 95%’.Again on the assumption of normally distributed measurement errors, both within and between laboratories, this definition leads to an expression for the reproducibility limit: reproducibility limit = 1.96A2(sr 2 + sL 2) (12) where sr is the repeatability standard deviation, sL is a measure of the laboratory-to-laboratory variation and Asr 2 + sL 2 is the reproducibility standard deviation. An inter-laboratory study, involving multiple measurements of the same analyte in each of a large number of laboratories, is required to measure the reproducibility of a method. 5.3 Intermediate measures of precision Where the assay is product dependent, as will be the case for many pharmaceutical products, the use of an inter-laboratory proficiency style study will not, for commercial reasons, usually be an option, and hence estimates of reproducibility will not be available. As an alternative, measures of precision, which are intermediate between repeatability and reproducibility, are often recommended.8 Measures of intermediate precision are similar to reproducibility in that the various factors which are maintained constant for the definition of repeatability may be allowed to vary, but all measurements take place in only one laboratory.The principal factors which may vary are time, operator, instrument and calibration.9 Replacing sL 2 in the reproducibility standard deviation equation by a variance component associated with varying conditions within one laboratory gives an intermediate measure of precision. Different measures of intermediate precision may be defined, depending on which combination of factors is allowed to vary over the course of the validation study; see ISO 5725-3 for a detailed discussion of such measures and ISO 5725-6 for a discussion of the uses of control charts for monitoring intermediate precision.The measure of between-run variability sB, which can be derived from the X-bar chart, is a global measure of the effects of all the factors that contribute to intermediate precision.As such it could be combined directly with the repeatability standard deviation to give a pragmatic measure of intermediate precision, resulting in an intermediate precision standard deviation of Asr 2 + sB 2 . Alternatively, if the control chart data are dominated by one or two analyst/instrument combinations, it may be appropriate to delete (randomly) some of these data points to achieve a more balanced representation of different analytical conditions.Where a very long record of analyses is available it may even be possible to select a completely balanced dataset, i.e., one which includes, for example, equal representation of all analyst–instrument combinations. This would mimic the conditions for a designed experiment, except that it would not involve the randomization procedures recommended in experimental studies.On the other hand, it would involve analyses carried out over a much longer time frame than would be normal for method validation studies. Method validation studies provide initial evidence that a method is capable of performing the required analytical task and give estimates of the analytical precision that is achievable, as measured by repeatability, intermediate precision measures or reproducibility. Control charts are complementary, in that they provide evidence of stability, i.e., assurance that the analytical system continues to be capable of generating data fit for purpose.As discussed above, control chart data can also provide updated estimates of two of these method validation parameters, viz., repeatability and intermediate precision standard deviation, based on current data. In order that such calculations can be carried out without time-consuming searches of laboratory records, it is desirable that ancillary information, such as identity of analyst and instruments used, be recorded together with the test results in a readily accessible form.A final word of caution is in order. In using historical data, care must be taken not to be over-zealous in deleting what appear to be unusual observations. Recall that in the case study only those extreme observations for which there were reasonable explanations for their being extreme were deleted. Wholesale trimming of the data runs the risk of producing estimates of analytical variability that are biased downwards, resulting in optimistic measures of precision, which could be misleading. 6 Conclusion This review has tried to show that, apart from their routine use in monitoring analytical stability, control charts are a rich source of information on analytical performance over time. Analysis of the data gathered for control charting purposes can not only allow critical appraisal of the performance of the charts themselves, but also give important insights into the quality of the test sample results.Control charts are sometimes seen as an unacceptable addition to laboratory workload. Such a view ignores their impact on the quality of laboratory output. However, apart from this, control charts can actually reduce the total analytical effort in some circumstances, paradoxical as this may seem at first. Often, decisions on the numbers of replicate 440 Analyst, 1999, 124, 433–442analyses of test samples are made arbitrarily (perhaps following laboratory tradition), without reference to the properties of the analytical system.Where the assay is shown by the control charts to be stable over time and to have good precision, careful analysis of the properties of the assay may show that it is possible to reduce the number of replicates, without serious loss of precision. Any such reduction will have a far more important impact on laboratory workload than will the routine use of control charts.Where the assay precision involves substantial between-run variability, immediate improvements in assay quality, as measured by test result precision, can be achieved by including replicate analyses of the same test sample in different analytical runs rather than, as is usual, including all replicates in the same analytical run. This may be a short-term solution while ruggedness studies are being carried out with a view to improving the analytical system, or it may be necessary even in the longer term to achieve fitness for purpose.Finally, the possibility of using control chart data for estimating repeatability and intermediate measures of precision was discussed. This does not necessarily replace an initial study of assay precision, when the method is first validated, but it can provide continuing assurance of assay quality and up-to-date information on these method validation parameters, without need for special studies. 7 Acknowledgements I thank Fintan Keegan and Colette Ryan for discussions that stimulated me to think about the issues addressed in this paper, for comments on an earlier draft and for permission to use the data on which the case study is based.I am grateful to Tony Bent, Jack Doyle and Sarah Tait r helpful comments on an earlier draft. Michael Stuart provided detailed criticism and constructive suggestions which have greatly improved the structure of the paper. For this, and for the many convivial discussions which have helped shape my approach to statistics, I am deeply grateful. 8 Appendix Multipliers for constructing control charts are given in Table A1. 8.1 Range charts 8.1.1 Action and warning limits.The multipliers D3 and D4 give action limits (±3 standard errors) for the range chart when multiplied by the average range øR 2. Corresponding warning limits (±2 standard errors) can be obtained by drawing lines above and below øR 2, two-thirds of the distance between øR 2 and the action limits, for non-zero multipliers. 8.1.2 Probability limits. The three standard error action limits are a direct carryover of the convention used for X-bar charts, which is based on the assumption of a normal distribution. However, for sample sizes less than seven this convention gives a lower control limit which would be negative if the proper three standard error multiplier were used, hence D3 is normally set to zero. The reason for this is that ranges do not follow a normal curve: their sampling distribution is skewed.Probability limits are an alternative method for defining control limits which take into account the skewness of the sampling distribution curve for the sample range. Multiplying the average range (øR 2) by DA3 and DA4 gives limits with probabilities of 0.001 of observing values of R below the lower control limit or above the upper control limit when the analytical system is in control, i.e. when the precision remains constant.Similarly, DB3 and DB4 give warning limits corresponding to false alarm probabilities of 0.025. 8.2 Converting from average range to standard deviation An estimate of the standard deviation (or the standard error if the data are themselves averages) may be obtained from an average range øR , based on samples of size n by using the d2 factors in Table A1 as follows: �s = R d2 Thus, the standard error of the plotted points for Fig. 1(a) was calculated as S R X-bar = = = 1 1 128 21 21 1 128 18 80 .. . . where the differences between successive pairs of points (n = 2), the moving ranges, were averaged to give øR 1. 8.3 Standard deviation charts Range charts are traditionally used to monitor process variability in preference to an alternative chart based on the standard deviation. This tradition survives from a time, before calculators were widely available, when the calculation of standard deviations was considered difficult and, therefore, prone to error. The range, on the other hand, is easy to calculate.If, however, the sample sizes are large (more than 10 is often cited) the range uses the information in the data less efficiently than the standard deviation, and in such cases the standard deviation chart should be used in preference (see Howarth3 or Montgomery10 for further details on these charts). In an analytical context, there is a second reason why standard deviation charts may be of interest. Very often some, if not all, test samples will be measured at least in duplicate.2 In such circumstances, it will often be possible to monitor the variability of the analytical process using test sample replicate results. In situations such as those of our case study, the test samples will vary somewhat in their average results, owing to batch-to-batch variability. However, the variations are unlikely to be so large as to affect the analytical precision. Accordingly, the within-run test sample replicate ranges can be averaged and then converted to a standard deviation for each run, which can form the basis for a standard deviation chart. Where there are varying numbers of test samples in different analytical runs, then a fixed number of sets of replicates (say 5–10) might be decided upon for use in drawing the standard deviation chart. This chart will, in fact, be more powerful for detecting shifts in analytical precision than one based on only a small number of control material replicates. It will, however, require more computational effort, although this is of little concern where the charts are to be computer generated. Table A1 Multipliersa for constructing control charts Sample size D3 D4 DA3 DA4 DB3 DB4 d2 2 0 3.27 0 4.12 0.04 2.81 1.128 3 0 2.57 0.04 2.98 0.18 2.17 1.693 a Extensive tables of these multipliers can be found in most quality control textbooks. Analyst, 1999, 124, 433–442 4419 References 1 E. Mullins, Analyst, 1994, 119, 369. 2 International Union of Pure and Applied Chemistry, Harmonized Guidelines for Internal Quality Control in Analytical Laboratories, Pure Appl. Chem., 1995, 67 (4), 649. 3 R. J. Howarth, Analyst, 1995, 120, 1851. 4 Analytical Methods Committee, Analyst, 1989, 114, 1497. 5 L. C. Alwan and H. V. Roberts, Appl. Stat., 1995, 3, 269. 6 R. Caulcutt, Appl. Stat., 1995, 3, 279. 7 C. Champ and W. H. Woodall, Technometrics, 1987, 29 (4), 393. 8 International Conference on Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use (ICH), Guidelines for Industry: Validation of Analytical Procedures, ICHQ2A, ICH, Geneva, 1995. 9 International Organization for Standardization, Accuracy (Trueness and Precision) of Measurement Methods and Results, Parts 1–4, 6, ISO, Geneva, 1994. 10 D. C. Montgomery, Introduction to Statistical Quality Control, Wiley, New York, 3rd edn., 1997. Paper 8/08742C 442 Analyst, 1999, 124, 433–4

ISSN:0003-2654    

DOI:10.1039/a808742c

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Studies of organic residues from ancient Egyptian mummies using high temperature-gas chromatography-mass spectrometry and sequential thermal desorption-gas chromatography-mass spectrometry and pyrolysis-gas chromatography-mass spectrometry Stephen A. Buckley, Andrew W. Stott and Richard P. Evershed* Organic Geochemistry Unit, School of Chemistry, University of Bristol, Cantock’s Close, Bristol, UK BS8 1TS Received 18th November 1998, Accepted 1st February 1999 The techniques of gas chromatography-mass spectrometry (GC-MS) and sequential thermal desorption-gas chromatography-mass spectrometry (TD-GC-MS) and pyrolysis-gas chromatography-mass spectrometry (Py-GC-MS) have been utilised to characterise the constituents of tissue-derived or applied organic material from two Pharaonic Egyptian mummies with a view to identifying embalming practices/substances. The results obtained using TD-GC-MS revealed a series of monocarboxylic acids with the C16:0, C18:1 and C18:0 components dominating in both mummies. The thermal desorption products related to cholesterol, i.e., cholesta-3,5,7-triene and cholesta-3,5-diene (only in Khnum Nakht), were detected in both mummies.Khnum Nakht also contained a number of straight chain alkyl amides (C16–C18) and an alkyl nitrile (C18). Other products included the 2,5-diketopiperazine derivative (DKP) of proline–glycine (pro–gly) which was a major component (7.9%) in Khnum Nakht but only a very minor component in Horemkenesi.Py-GC-MS of samples of both specimens yielded a series of alkene/alkane doublets (Horemkenesi C6–C18, Khnum Nakht C6-C24) which dominated their chromatograms. Series of methyl ketones in the C9–C19 chain length range were also present, with C5–C7 cyclic ketones occurring in Horemkenesi only. These ketones are indicative of covalent bond cleavage, probably of polymerised acyl lipids. Nitrogenous products included nitriles (C9–C18) which were significant in both samples, and amides which were only detected in Khnum Nakht.Also present amongst the pyrolysis products were three steroidal hydrocarbons, cholest-(?)-ene, cholesta-3,5,7-triene and cholesta-3,5-diene. High temperature-GC-MS of trimethylsilylated lipid extracts yielded similar monocarboxylic acids to that obtained using TD-GC-MS, while a series of a,w-dicarboxylic acids and a number of mono- and di-hydroxy carboxylic acids not seen in the thermal desorption or pyrolysis GC-MS analyses were significant constituents in both mummy samples.Overall, the use of GC-MS and sequential TD-GC-MS and Py-GC-MS has demonstrated in both mummies the presence of a complex suite of lipids and proteinaceous components whose compositions indicates extensive alteration via oxidative and hydrolytic processes during long-term interment. None of the classical embalming resins was detected but an exogenous origin for at least a proportion of these components cannot be discounted since fats, oils and gelatin have been proposed as embalming agents in mummification.The combined approach of sequential TD- and Py-GC-MS has potential for application to the characterisation of embalming materials in mummies. Most importantly these techniques virtually eliminate any destruction of the mummified bodies thereby allowing the scope of investigations of ancient Egyptian funerary practices to be significantly extended. Introduction People have long been fascinated by the ancient Egyptian ‘art’ of mummification, a practice carried out in Egypt from ca. 2600 BC.1 Despite this interest however, surprisingly little is known about the practice, particularly the use of organic preservatives or ‘embalming resins’. Though the ancient Egyptians left no written record of embalming technology, direct evidence can be derived from the types of preservatives that have survived in association with the bodies. Proposed organic preservatives include bitumen, beeswax, true resins [e.g., coniferous (diterpenoid) and non-coniferous (Pistacia)], and gum resins (e.g., myrrh and frankincense).2 The few chemical studies performed on Egyptian mummies have used a variety of analytical techniques to determine the nature and origin of ‘embalming resins’ based on the presence of specific biological marker compounds.Techniques employed to date include gas chromatography-mass spectrometry (GC-MS),3–6 pyrolysis-mass spectrometry (Py-MS),7 high performance liquid chromatography (HPLC)8 and fast atom bombardment tandem mass spectrometry.9 However, many questions still remain unanswered and a great deal more research is needed before we can claim to have a reasonable understanding of the technology of embalming during the 3000 years in which mummification was practised in Egypt.When deciding on an appropriate analytical approach it is necessary to consider both the valuable nature of the specimens from which the samples are to be taken, and the specific nature of those samples (i.e., aged organic materials of uncertain origin).Due to the irreplaceable nature of the samples (i.e., mummies) the ability to accommodate very small sample sizes is an important consideration, particularly since a large number of mummies need to be studied to provide a comprehensive and meaningful picture of the embalming materials employed. In addition, the approach adopted must recognise the nature of the samples analysed, taking account of the fact that the compositions of the organic materials utilised in the mummification process are likely to have changed substantially over time as a Analyst, 1999, 124, 443–452 443result of natural degradative processes.These processes will include oxidation, reduction, hydrolysis, aromatisation and polymerisation. However, organic components which are resistant to chemical and biological degradation, and are characteristic of the original ‘embalming resins’, can be expected to survive for very long periods of time and may be recognisable in a relatively unchanged state.These specific compounds represent the biomarker components that will be used to identify the ancient ‘resins’. The possibility of encountering both free and polymerised biomarker compounds must be considered when deciding upon the analytical approach. Solvent extraction followed by GC-MS is often utilised as an appropriate approach for the characterisation and identification of a wide variety of organic components of biological tissues.This approach can provide a great deal of valuable information about the nature of aged organic residues.3–6,10,11 After suitable derivatisation even polar polyfunctional compounds, commonly present, are amenable to GC-MS.4,11–14 Sample preparation is however relatively time consuming and sample losses, especially of volatile components which may be trapped within the sample matrix,15 can be problematical. The technique of thermal desorption coupled with gas chromatography-mass spectrometry has proven to be a rapid and direct method for the identification of free biomarkers in a broad range of organic materials.16–21 Thermal desorption is time efficient since the extraction step is effectively instantaneous. Furthermore the technique involves minimal sample manipulation which reduces the problems of contamination, sample loss and other experimental errors inherent with ‘wet chemical’ procedures.The very small sample sizes necessary ( < 0.1 mg) allow the virtually non-destructive analysis of precious specimens, such as mummies. Importantly, TD can be conveniently combined sequentially with Py-GC-MS. Presented herein are the results of a combined sequential TDGC- MS and Py-GC-MS approach to study both ‘free’ and ‘bound’ biomarkers thought to be present in heterogeneous organic matter from two ancient Egyptian mummies, at least one of which had been described as embalming resin.22 The samples were also analysed by solvent extraction then GC-MS following separation of the acidic and neutral components.The TD and solvent extraction approaches were assessed and their advantages and disadvantages discussed. Experimental Samples description and preparation Samples from two Egyptian mummies were investigated. Resin-like material was taken from the left hip/spine region of Horemkenesi (ca. 1000 BC), Ha 7386/948 (Bristol Museum), and bandage/resin/tissue from Khnum Nakht (ca. 2000 BC) (Manchester Museum). Samples were ground (under liquid nitrogen) using a mortar and pestle prior to chemical analysis. Thermal desorption-gas chromatography-mass spectrometry (TD-GC-MS) and pyrolysis-gas chromatography-mass spectrometry (Py-GC-MS) A CDS 1000 Pyroprobe (Chemical Data System, Oxford, PA, USA) unit fitted with a platinum coil probe was used for the thermal extraction of free biomarker compounds and the pyrolysis of bound/polymerised components from the ground samples ( < 0.1 mg).The samples were loaded into quartz tubes plugged with solvent extracted glass wool. The quartz tubes were then inserted into the platinum coil of the probe which was inserted into the heated injector of a GC interfaced to the mass spectrometer. The quartz tubes were pre-cleaned by heating them in a furnace at 600 °C for 24 h. They were also heated to 1000 °C in the platinum coil probe prior to use.The TD/Py temperature was held for 10 s. TD-GC-MS and Py-GC-MS were carried out on a Carlo Erba (Milan, Italy) 4130 gas chromatograph fitted with a fused silica capillary column (50 m 3 0.32 mm id) coated with a dimethyl polysiloxane bonded stationary phase (CP Sil-5 CB, 0.4 mm film thickness; oven temperature programme, 35 °C (5 min) to 320 °C (15 min) at a rate of 4 °C min21) interfaced to a Finnigan (Sunnyvale, CA, USA) 4500 mass spectrometer operated in full scan mode (40–650 Da, 1 scan sec21; electron energy, 70 eV; filament current, 350 mA; source temperature, 170 °C).The Pyroprobe interface temperature was 280 °C and the transfer line temperature from the GC to the mass spectrometer was 290 °C. Helium was used as carrier gas. Peaks were identified based on both their mass spectra (NIST/EPA/NIH Mass Spectral Database) and retention times. TD- and Py-GC-MS The optimum TD temperature of the Pyroprobe coil, was determined using the following sequential TD-GC-MS analysis of the ground mummy resin-like material.An aliquot of each of the two samples was heated in a stepwise fashion at 290, 310, 330, 350, 360 and 400 °C. Pyrolysis of the ‘bound’ components of the two thermally desorbed samples was then carried out using a temperature of 610 °C. Sample preparation for GC-MS The samples were ultrasonically extracted with chloroform– methanol (2 + 1 v/v; 3 3 60 min). The extracts were combined and the solvent reduced by rotary evaporation.Following transfer to a screw-capped vial samples were then evaporated under nitrogen at 40 °C. Aliquots of each of the combined extracts [dispersed in dichloromethane–propan-2-ol (DCM/ IPA)] were separated into acid and neutral fractions by solid phase extraction (SPE) using a Bond Elut NH2 aminopropyl column (500 mg per 2.8 ml) (Phenomenex Ltd, Macclesfield, Cheshire, UK). The neutral components were eluted with dichloromethane–propan-2-ol (2 + 1 v/v; 20 ml), prior to the elution of the acids with 5% acetic acid in ether (20 ml).The majority of the solvent for each fraction was then removed (rotary evaporator), and evaporated under nitrogen at 40 °C. Both fractions were trimethylsilylated at 70 °C for 1h with N,Obis( trimethylsilyl)trifluoroacetamide containing 1% of trimethylchlorosilane (Sigma Chemical Co., St Louis, MO, USA) and analysed by GC and GC-MS. High temperature (HT)-GC-MS GC-MS was carried out on a Carlo Erba (Milan, Italy) 5160 gas chromatograph fitted with a fused silica capillary column (15 m 3 0.32 mm id) coated with a dimethylpolysiloxane bonded stationary phase (DB1, 0.1 mm film thickness; oven temperature programme, 50 °C (2 min) to 350 °C (10 min) at a rate of 10 °C min21) interfaced to a Finnigan (Sunnyvale, CA, USA) 4500 mass spectrometer operated in full scan mode (40–650 Da, 1 scan s21; electron energy, 70 eV; filament current, 350 mA; source temperature 170 °C).Helium was used as carrier gas. Peaks were identified based on both their mass spectra (NIST/ EPA/NIH Mass Spectral Database) and retention times. 444 Analyst, 1999, 124, 443–452Results Analytical approach The nature of the proposed embalming materials suggests that both free and polymerised compounds are likely to be present. Sequential TD-GC-MS and Py-GC-MS offers a particularly appropriate and convenient approach to the study of such materials for the following reasons: (1) small sample sizes are required ( < 0.1 mg) thereby making it a virtually nondestructive means of studying historically valuable museum specimens such as mummies, (2) it is a rapid technique requiring minimal sample preparation, thus allowing high sample throughput, and (3) it minimises problems of contamination and sample losses by reducing sample handling and the use of ‘wet’ chemical treatments.The utility of this approach is demonstrated below through the analysis of two Pharaonic Egyptian mummies.The results of thermal desorption and pyrolysis will be presented first, followed by the results obtained from more conventional analyses using solvent extraction, fractionation and GC-MS. A discussion will then be given of the relative merits of the two approaches. Optimum thermal desorption temperature for TD-GC-MS In order to determine the most suitable thermal desorption temperature for the ‘free’ biomarkers (as opposed to pyrolysis which involves the cleavage of covalent bonds) the samples were heated, each one sequentially, at increasingly higher probe temperatures. The components released from each thermal desorption step were monitored. At 290 °C (see Fig. 1) a range of compounds were observed, including the pyrolysis product proline–glycine (DKP). At 310, 330, 350 and 360 °C there was no significant release of volatile components (see Fig. 1). At 400 °C GC small peaks were seen in both samples at short retention time corresponding to the generation of carbon dioxide, however, no other volatile species were observed. A probe temperature of 610 °C produced a complex series of pyrolysis products from both specimens, although these were far less abundant in Horemkenesi than in Khnum Nakht.These latter pyrolysis products are presumed to result from covalent bond cleavage in the ‘bound’ material, i.e., polymeric and/or functionalised components.Thus, 290 °C was chosen for TD since it was as equally effective as somewhat higher temperatures, i.e., up to 400 °C, whilst having the additional advantage of limiting the extent of molecular rearrangements, e.g., dehydration, dehydrogenation, transmethylation, etc., which could take place at more elevated TD temperatures. The temperature of 610 °C was chosen for pyrolysis studies aimed at characterising polymerised components. Thermal desorption-gas chromatography-mass spectrometry (TD-GC-MS) The results of the TD-GC-MS analysis of resin-like material taken from the left hip/spine of Horemkenesi are shown in Fig. 1a. A range of compounds were seen which were identified by their mass spectra (see Table 1). These included a series of free fatty acids (C6 to C18). The major fatty acid was C16:0 (M+. 256), with C18:0 (M+. 284), C18:1 (M+. 282) and C14:0 (M+. 228) also present in significant amounts. These four fatty acids were the dominant components in the chromatogram.Eluting significantly later than the fatty acids, the steroidal compound cholesta-3,5,7-triene (M+. 366) was detected albeit as a very minor component. In addition, the aldehyde n-nonanal ([M- 18]+. 124) was also found to be present. No other significant components were observed. At a thermal desorption temperature of 290 °C none of the characteristic alkene/alkane doublets was seen which were detected in pyrolysates of acyl lipids at higher temperatures. This indicates that the free fatty acids are present in, and desorbed from, the matrix during the thermal evaporation stage and probably do not arise as a result of covalent bond cleavage.The presence of n-nonanal is presumed to arise through the oxidation of the double bond in oleic acid which is present at high abundance in human tissues as well as many other animal and plant sources.23,24 The results of the TD-GC-MS analysis of bandage/resin/ tissue from Khnum Nakht are shown in Fig. 1b. The compounds observed were again identified by their mass spectra (see Table 1). The fatty acid profile was similar to that of the Horemkenesi sample (carbon number range C7 to C18), with the major components being C16:0, C18:1, C18:0 and C14:0 in decreasing order of abundance. However, in contrast to Horemkenesi, two steroidal compounds were observed as major components in Khnum Nakht, eluting at later retention times (72.4 and 72.9 min). The earliest eluting of these was characterised by the presence of m/z 141, 143, 247, 253 and M+. 366 identifying it as cholesta-3,5,7-diene (see Fig. 2a). The later eluting peak was characterised by the presence of m/z 145, 147, 247, 255, 260 and M+. 368, identified as cholesta-3,5-diene (see Fig. 2b). Also present in significant amounts (again unlike Horemkenesi) were a number of compounds of proteinaceous origin. The 2,5-diketopiperazine derivative, pro–gly (ions m/z 70, 83, 111, 154) was identified as a major component eluting just prior to the C14:0 fatty acid.Two 2,5-diketopiperazine derivatives of proline– alanine (pro–ala) (ions m/z 70, 97, 125, 168) were also identified, as were other diketopiperazine derivatives at lower abundances. Other compounds noted were octadecanenitrile ([M-15]+. 250) characterised by ions of m/z 110 and 124, and C16:0 (M+. 255), C17:0 (M+. 269), C18:1 (M+. 281) and C18:0 (M+. 283) amides characterised by ions of m/z 59 and 72. n- Fig. 1 Reconstructed total ion chromatograms of the thermal desorption profiles (290 °C for 10 s) of (a) Horemkenesi, resin-like materia and (b) Khnum Nakht, bandage/resin/tissue.Pro–gly (DKP) = 2,5-diketopiperazine of proline and glycine; pro–ala (DKP) = 2,5-diketopiperazine of proline and alanine. Cholesta-3,5,7-triene in Horemkenesi is not shown, only being present as a trace component. Analyst, 1999, 124, 443–452 445Nonanal was again present as a minor component. As before, typical lipid pyrolysis products (e.g., alkenes, alkanes, ketones, etc.) were absent using a TD temperature of 290 °C.Interestingly, however, products normally associated with the pyrolysis of proteins25–27 were seen, suggesting that these may be formed at relatively low temperatures during the TD at 290 °C. Their absence from solvent extracts suggests that they are probably not present in their free form in the ancient tissues. Pyrolysis-gas chromatography-mass spectrometry (Py-GC-MS) The pyrogram obtained for Horemkenesi (Fig. 3a) is much more complex and of very different character compared with the thermal desorption profile. The pyrolysis products identified by their mass spectra are listed in Table 2. Dominating the chromatogram were a series of alkene/alkane doublets (m/z 55/57) (C6 to C18 maximising at C6), characteristic of covalent bond cleavage of the bound polymeric constituents of the mummy tissue induced by the higher pyrolysis temperature. Methyl ketones (C9 to C11 , C15 to C17 , C19:0 and C19:1; see Table 2), characterised by mass chromatograms m/z 58 and 71, were also present in significant abundance.More unusually the 6-methyl-3,5-dien-2-ones (m/z 81, 109 and 124; C17 and C18; see Table 2) were observed. In addition, saturated (m/z 110 and 124) and unsaturated (m/z 108 and 122) nitriles [C9:0 and C10:0, C16:0, C16:1, C16:2, C18:0, C18:1 (two isomers)] were present in appreciable amounts. The aromatic components benzene, toluene, styrene, ethyl benzene and o-xylene, the cyclic ketones cyclopentanone, cyclohexanone and cycloheptanone/methylcyclohexanone, and the alicyclic hydrocarbons dimethylcyclopentane and methylcyclopenta-1,3-diene (two isomers) were also observed (see Table 2).Although Py-GC-MS provided some interesting and valuable data the pyrolysate represented a minor proportion of the whole sample (i.e., c.f. TD) which may suggest that it was less extensively polymerised compared with Khnum Nakht (see below).The Py-GC-MS results obtained for Khnum Nakht are shown in Fig. 3b. The pyrogram is dominated by a wide range of compounds, identified by their mass spectra (see Table 2). Alkene/alkane doublets (C6 to C24) dominated the chromatogram, although the profile is quite different, with the n-C15 homologue the most abundant. Methyl ketones (C10, C11 and C16 to C19) and, more unusually, ethyl (3-) ketones (m/z 57 and 72; C18 and C19) were identified.In addition, saturated (C14:0 to C18:0) and unsaturated (C18:1) nitriles were present as major components. The C16:0 and C18:0 alkyl amides were present as minor components (these were absent from Horemkenesi). The profile of aromatic and alicyclic components was similar to that obtained from Horemkenesi. Cholest-(?)-ene (M+. 370), cholesta- 3,5,7-triene and cholesta-3,5-diene were also observed, as was a pyrrole. In contrast to the sample from Horemkenesi, the pyrolysate constitutes a significant portion of the sample (i.e., cf.TD), indicating the presence of abundant bound biomarkers and pointing to a greater extent of polymerisation. Table 1 Composition of thermal extracts (290 °C) from thermal desorption-gas chromatography-mass spectrometry (TD-GC-MS). Values in bold indicate base peaks. Masses underlined indicate molecular ions (M+.). Other masses are characteristic fragment ions Relative abundance (%)a Compound name Mass spectral data m/z Horemkenesi, resinlike material Khnum Nakht, bandage/resin/tissue Carboxylic acids— Monocarboxylic acids Octanoic acidb,d Nonanoic acidb,d Dodecanoic acidd Tetradecanoic acidd Pentadecanoic acidd Hexadecenoic acidd Hexadecanoic acidd i+ai-Heptadecanoic acidd Heptadecanoic acidd Octadecenoic acid (32)d Octadecanoic acidd Aldehydes— Nonanalb,d Amides— Hexadecanamidec Heptadecanamidec Octadecenamidec Octadecanamidec Nitriles— Octadecanenitrilec Steroids— Cholesta-3,5,7-trienec Cholesta-3,5-dienec 2,5-Diketo-piperazines— Pro–glyc Pro–ala (32)c Saturated Unsaturated 43,57,73,129 41,55,69,123 — 158 200 228 242 254 256 270 270 282 284 57, 70, 98, 114, 124 255, 59, 72, 86, 128 269, 59, 72, 86, 128 281, 59, 72, 81, 126 283, 59, 72, 86, 128 43, 41, 97, 110, 124 366, 43, 135, 141, 143, 247, 351 368, 43, 145, 147, 247, 353 154, 83, 70, 98, 111 168, 70, 97, 125 0.2 0.2 0.1 5.0 0.6 1.8 71.6 Tr ( < 0.1) 0.3 12.2 6.6 0.8 — — — — — Tr ( < 0.1) — 0.4 — 1.1 3.0 — 2.8 0.4 0.4 36.8 0.3 0.6 8.0 12.8 0.2 8.5 0.3 1.0 7.2 0.4 4.6 1.1 7.9 0.8 a Relative abundances of compounds are determined from areas under peaks in TIC profiles.b Highly volatile components which would be expected to be reduced in abundance during sampling handling in ‘wet’ chemical analyses. c Components formed thermolytically during TD. d Components in their original form. 446 Analyst, 1999, 124, 443–452High temperature-gas chromatography-mass spectrometry (HT-GC-MS) The results for the acid fraction of Horemkenesi obtained by solvent extraction and SPE (see Experimental) are shown in Fig. 4a. A range of carboxylic acids were detected and identified by their mass spectra (see Table 3). The chromatogram was dominated by a series of monocarboxylic acids, including nalkanoic acids in the C9 to C24 carbon number range (with the exception of C11 , C21 and C23), the saturated iso- and anteisomethyl branched C15 and C17 and C16:1 , C18:1 and C24:1 unsaturated components. The distribution of monocarboxylic acids was similar to that obtained using TD (see Figs. 1 and 4), with C14:0, C16:0, C18:0 and C18:1 as the major components. Also present in significant abundance were a series of a,w- dicarboxylic acids from C6 to C12, with the C8 and C9 components predominating. In addition, relatively large amounts of monohydroxy and dihydroxy carboxylic acids were present with the erythro and threo isomers of both 9,10-dihydroxyhexadecanoic acid and 9,10-dihydroxyoctadecanoic acid predominating.These polar hydroxy (mono- and di-) and diacids were not seen using TD presumably since these more highly functionalised compounds are not amenable to TD-GCMS in their underivatised form. These polar acids are not unimportant since they may well indicate post-mortem transformations of the fatty acids originally present and may give clues relating to the method of preservation and the general environmental conditions to which the body was exposed.4,28 The results for the neutral fraction of Horemkenesi obtained after solvent extraction and SPE are shown in Fig. 5a.The components present included methyl esters, monoalkylglyceryl ethers, monoacylglycerols and small amounts of steroidal components (see Table 4). The major neutral compound was the bis-TMS ether C18:1 1-monoacylglycerol, with the 1-monoacylglycerols containing C14:0, C16:0 and C16:1 fatty acyl moieties present as major components.Interestingly, the C18:0 1-monoacylglycerol was only a relatively minor component. Observed in almost equal abundance to the major 1-monoacylglycerols were the C16 and C18 1-O-monoalkylglyceryl ethers (as their bis-TMS ethers). Also dominant in the neutral fraction were the methyl esters of the carboxylic acids C16:0, C18:0, C18:1 and the C18:0 9,10-dihydroxy compound. Cholesterol and cholesta-3,5-dien-7-one were present in low abundance. The neutral fraction of Horemkenesi represents an almost insignificant proportion of the total extractable lipid.However, the relatively high abundance of the 1-monoacylglycerols compared to the relatively small amounts of cholesterol and the cholesta-3,5-dien-7-one agrees well with the results of the TD analyses, confirming the virtual absence of steroidal compounds in the sample. The GC profile for the acid fraction of Khnum Nakht obtained by solvent extraction and SPE is shown in Fig. 4b. A range of carboxylic acids was present and identified by their mass spectra (see Table 3).The chromatogram was dominated by a series of monocarboxylic acids. This included saturated straight chain acids in the C12 to C24 carbon number range (with the exception of C21), the saturated iso- and anteiso-methyl branched acids in the C15 to C17 range, and unsaturated acids including C16:1, C18:1, C20:1 and C24:1. The profile of the monocarboxylic acids, dominated by the C16:0 , C18:0 and C18:1 components, was comparable with the TD profile obtained for this sample.A series of a,w-dicarboxylic acids in the C7 to C11 carbon number range were present as relatively minor components together with moderate amounts of monohydroxy and dihydroxy carboxylic acids. Although a number of the polar acids found in Horemkenesi were also detected in the Khnum Nakht acid fraction they were not nearly as abundant, suggesting that the sample from Khnum Nakht is in a far less advanced state of oxidisation.Fig. 2 Mass spectra of the two steroidal compounds detected in Khnum Nakht, bandage/resin/tissue after thermal desorption at 290 °C for 10 s; (a) cholesta-3,5,7-triene and (b) cholesta-3,5-diene. Fig. 3 Reconstructed total ion chromatogram of the pyrolysis profiles (610 °C for 10 s) of (a) Horemkenesi, resin-like material and (b) Khnum Nakht, bandage/resin/tissue, after thermal desorption (290 °C for 10 s). Note: - = alkenes; 5 = alkanes; È = alicyclic hydrocarbons; : = aromatic hydrocarbons; 2 = 2-alkanones; 8 = 3-alkanones; ½ = cyclic ketones; Ó= nitriles; ¦ = amides; * = steroids.C10 and C15 refer to the carbon numbers of the alkenes and alkanes within the retention time windows. Analyst, 1999, 124, 443–452 447Table 2 Pyrolysis products (610 °C) from pyrolysis-gas chromatography-mass spectrometry (Py-GC-MS). Values in bold indicate base peaks. Masses underlined indicate molecular ions (M+.). Other masses are characteristic fragment ions Compounds present (]) Compound namea Mass spectral data m/z Horemkenesi, resin-like material Khnum Nakht, bandage/resin/tissue n-Alkenes/n-Alkanes— C5 (alkane) C6 �C 18 C19 �C 24 Alicyclics— Dimethylcyclopentane Methylcyclopenta-1,3-diene (32) 2-Alkanones— C9 C10 C11 C15 i-C16 n-C16 C17 C18 C19 2-Alkenones— C19 (32) 2-Alka-3,5-dienones— 6-Methylhexadeca-3,5-dien-2-one 6-Methylheptadeca-3,5-dien-2-one 3-Alkanones— C18 Cyclic ketones— C5 C6 C7 Aromatics— Benzene Toluene Ethylbenzene m/p-Xylenes o-Xylene Styrene Trimethylbenzene Nitriles (saturated)— Nonanenitrile Decanenitrile Tetradecanenitrile Pentadecanenitrile Hexadecanenitrile Heptadecanenitrile Octadecanenitrile Nitriles (unsaturated)— Hexadecenenitrile Octadecenenitrile (32) Nitriles (diunsaturated)— Hexadecadienenitrile Amides— Hexadecanamide Octadecanamide Steroids— Cholest-(?)-ene Cholesta-3,5,7-triene (?) Cholesta-3,5-diene Alkanes: 57, 71, 85, 99 Alkenes: 55, 69,83, 97 98, 41, 56, 70, 79 43, 58, 71 43, 58, 71 43, 58, 71 43, 58,71 43, 57, 71 58, 43, 71 254, 43, 58, 71 43, 58, 71 282, 43, 58, 71 43, 55, 58, 71 109, 43, 81, 124 109, 43, 81, 124 57, 43, 72, 239 84, 55, 42 98, 55, 42, 69 112, 55, 43, 68 78 91, 65 106, 91, 65 106, 91, 65 106, 91, 65 104, 78, 63 120, 105, 91, 65 41, 43, 97,110, 124 55, 41, 43, 97, 110, 124 41, 43, 97, 110, 124 41, 43, 97, 110, 124 43, 41, 97, 110, 124 43, 41, 97, 110, 124 43, 41, 97, 110, 124 41, 43, 97, 108, 122, 136 41, 43, 97, 108, 122, 136 40, 44, 105, 120, 134 59, 72, 86, 128 283, 59, 72, 86, 128 370, 355, 43, 108, 135,147 366, 43, 135, 141, 143 368, 43, 145, 147, 247, 353 ] ] ] — — ] ] ] ] ] ] ] — ] — ] ] ] ] — ] ] ] ] ] ] — ] ] — ] ] — — ] — ] ] ] ] — — — — — ] ] ] ] ] ] ] — ] ] — — ] ] ] ] ] — — ] — — — ] ] ] ] — ] ] — — ] ] ] ] ] — ] — ] ] ] ] ] a None of the components of the pyrograms is detectable in the solvent extracts and thus are presumed to have derived through thermolysis of the largely heteropolymeric resin-like material. 448 Analyst, 1999, 124, 443–452The GC profile for the neutral fraction of Khnum Nakht after solvent extraction and SPE is shown in Fig. 5b. The neutral fraction was (unlike the Horemkenesi neutral fraction) dominated by cholesteryl esters and several related steroidal compounds, with monoglyceryl ethers making a significant, albeit smaller, contribution (see Table 4). The major free steroid identified was cholesta-3,5-dien-7-one, with cholesterol and 3-hydroxycholest-5-en-7-one (7-ketocholesterol) also abundant.At longer retention times (26.5 and 27.4 min) two cholesteryl esters were present, again as major components; these would have been too involatile to successfully elute from the column utilised in the TD-GC-MS, though it is anticipated that they would undergo a 1,2-elimination upon TD at 290 °C to give the fatty acid and cholesta-3,5-diene (with a lesser amount of cholesta-3,5,7-triene, via dehydrogenation).Also significant were the C16 and C18 1-O-monoalkylglyceryl ethers, although in contrast to Horemkenesi, C18:0 and C18:1 1-monoacylglycerols were detected only as minor components and other 1-monoacylglycerols were absent, indicating almost complete hydrolysis of the original acyl lipids that would have dominated the tissues at the time of death (if indeed they do originate from the body). Unlike Horemkenesi, the neutral fraction constitutes a much more significant proportion of the total lipid extracts of the sample.The steroids and sterol esters are present in similar abundance to the C15:0 and C17:0 fatty acids, although the C16:0, C18:1 and C18:0 components constituted the bulk of the extractable lipid present in the sample. Notably, the C16 and C18 1-O-monoalkylglyceryl ethers were not observed in the analysis of the same sample with TD, thus demonstrating the problems associated with the analysis of polyfunctional compounds using this latter approach.Discussion When analysing amorphous organic materials from aged human remains it is important to consider the chemical composition of original tissues in order to bee to distinguish between human body lipids, proteins, etc., and components which may originate from any applied embalming materials. Adipose tissue, skin and muscle compose the three major sources of lipids in humans. Although the composition of human adipose tissue can be influenced by dietary input29–32 it is usually reasonably consistent.29,33,34 The tissue is largely composed of triacylglycerols ( > 90% of the lipids), phospholipids (0.2–4%), glycerol ethers ( Å 1%), non-saponifiable fat (0.2–1.3%), squalene (trace amounts), proteins (2–3%) and water (5–30%).Cholesterol constitutes 0.03–0.3% and cholesterol fatty acyl esters Å 0.1%. The fatty acid composition of the adipose tissue is dominated ( > 90%) by myristic, palmitic, palmitoleic, stearic, oleic and linoleic acids, typically 2–5, 18–25, 5–9, 3–7, 40–62 and 6–16%, respectively, of total fatty acids (free fatty acids are only minor components in adipose tissue).Human skin lipids differ markedly in composition to those found in adipose tissue.35–40 They are dominated by wax esters (20–22%), triacylglycerols (29–32%) and free fatty acids (30–33%). Notably, the acid chain length of the wax esters is C16–C24 (C24 dominant). Squalene is a major component (11–13%), with smaller amounts of free sterols (1–2%) and sterol esters (2–3%).Collagen, an important constituent of skin, is composed of polypeptide chains, the major amino acids being glycine ( Å 33%), alanine ( Å 11%), proline ( Å 13%) and hydroxyproline ( Å 10%).41 Muscle is composed predominantly of proteins (e.g., myosin and actin) consisting of amino acids, of which glutamic acid ( Å 11%), leucine ( Å 10%), cysteine ( Å 10%), lysine ( Å 8%), alanine ( Å 7%), isoleucine ( Å 6%), arginine ( Å 6%), threonine ( Å 6%) and aspartic acid ( Å 5%) are the major components.42 Of the lipids present in muscle, phospholipids dominate ( > 60% of total lipids).The triacylglycerols ( > 80%) and cholesterol ( > 15%) constitute the majority of the neutral lipids.10 The oxidised lipids found in the total lipid extracts of the mummies studied herein are not usually present in mammalian tissues. However, previous studies carried out on aged human remains which have been naturally mummified have revealed significant amounts of oxidised lipids.4 The relative proportion of saturated to unsaturated fatty acids (particularly C16:0 to C18:1) has generally been found to increase over time, partially due to the oxidation of the double bond in the unsaturated acids resulting in short chain a,w-C6-C12 diacids.These a,w-diacids were observed by G�ulaçar et al.4 and their presence here is not inconsistent with an aged fat from a relatively dry aerobic burial environment.Notably, however, significant amounts of unsaturated fatty acids, including C18:1 hydroxy fatty acids, were found in both samples. Gulaçar et al. observed an absence of unsaturated monocarboxylic acids and detected only saturated hydroxy fatty acids. The hydroxy and dihydroxy carboxylic acids can derive from varying degrees of oxidation of the original unsaturated C16:1, C18:1, C18:2 fatty acids. However, it should be noted that since these mummies (certainly Horemkenesi) come from periods when embalming was extensively practised this cannot be assumed.Unsaturated and saturated monohydroxy carboxylic acids are found in plant waxes23 which could have been utilised. They could also originate from the oxidation of plant or animal derived oleic/linoleic acids applied during mummification. The branched chain (iso and anteiso) C15:0 and C17:0, fatty acids which are present in significant amounts in bacteria, also occur in living human adipose tissue as minor components (typically 0.3% and 1.0%, respectively29).The quantities found in the samples are consistent with this and so do not suggest they are of bacterial origin. The longer chain C22 and C24 fatty acids are known to occur in skin lipids in significant amounts, hence, their presence in the samples in low abundance could be endogenous, rather than being microbially derived. Although phospholipids are susceptible to degradation and as such would not be expected to survive intact, the presence of the amides and proportion of the Fig. 4 Reconstructed total ion chromatograms from the GC-MS of the acid fraction (after solvent extraction and fractionation) of (a) Horemkenesi, resin-like material and (b) Khnum Nakht, bandage/resin/tissue. (Acids are present as their TMS derivatives.) Analyst, 1999, 124, 443–452 449fatty acids seen in the aged tissues of Khnum Nakht may reflect the presence of phospholipids. Since these amides were not seen in the total lipid extracts they are presumed to derive thermolytically.The steroidal compounds in Khnum Nakht indicate an animal fat, possibly human origin. The results obtained above show how the combined approach of conventional GC-MS and sequential TD-GC-MS and Py- GC-MS has allowed the characterisation of the two mummy samples. They were both identified as being composed of degraded acyl lipids having undergone almost complete hydrolysis, resulting in the dominance of the fatty acids in both samples.However, the techniques employed revealed notable differences in the nature of the samples. The Horemkenesi sample had undergone a significant degree of oxidation and was essentially a mixture of free fatty acids having very little polymerised material present. In contrast, Khnum Nakht although hydrolysed was far less oxidised and contained appreciable amounts of steroidal compounds including cholesterol (effectively absent in Horemkenesi). Khnum Nakht was also highly polymerised in nature as indicated by the characteristic alkene/alkane doublets seen in the Py-GC-MS analyses.The large pro–gly peak present in the TD of Khnum Nakht points to a proteinaceous origin, though its formation at Å 300 °C is interesting since it is generally observed as a pyrolysis product at higher temperatures (e.g. 610 °C/ 750 °C25–27). Clearly it is being formed in the probe at relatively low temperatures. It is significant that the samples do not appear to have been treated with any di- or triterpenoid resins traditionally associated with the embalming process in ancient Egypt.2 Diterpenoids have been identified in a later mummy currently being investigated in our laboratory.43 The absence of terpenoid resins in Horemkenesi is particularly surprising since he was mummified when the embalming process was at its peak.The sample analysed was considered, and in fact identified, in previous studies22 to be a material applied extensively to the body as part of the embalming process and not endogenous to the body itself.This may serve to illustrate the need for a fairly comprehensive Table 3 Composition of acid fraction from gas chromatography-mass spectrometry (GC-MS) following solvent extraction procedures (all compounds determined as their trimethylsilyl derivatives). Values in bold indicate base peaks. Masses underlined indicate molecular ions (M+.). Other masses are characteristic fragment ions Relative abundance (%)a Compound name Mass spectral data m/z Horemkenesi, resinlike material Khnum Nakht, bandage/resin/tissue Monocarboxylic acids— Dodecanoic acid Tridecanoic acid Tetradecanoic acid i-Pentadecanoic acid ai-Pentadecanoic acid Pentadecanoic acid i-Hexadecanoic acid Hexadecenoic acid Hexadecanoic acid i-Heptadecanoic acid ai-Heptadecanoic acid Heptadecanoic acid Octadecenoic acid (32) Octadecanoic acid Nonadecanoic acid Eicosenoic acid Eicosanoic acid Docosanoic acid Tetracosenoic acid Tetracosanoic acid a,w-Dicarboxylic acids— Hexanedioic acid Heptanedioic acid Octanedioic acid Nonanedioic acid Decanedioic acid Undecanedioic acid Dodecanedioic acid Monohydroxy-carboxylic acids— 8-Hydroxy- and 11-Hydroxy-octadecenoic acids 8-Hydroxy-, 9-Hydroxy- and 10-Hydroxy-octadecenoic acids 10-Hydroxy-octadecanoic acid Dihydroxycarboxylic acids— 90-Dihydroxyhexadecanoic acid (t)b 9,10-Dihydroxyhexadecanoic acid (e)c 9,10-Dihydroxyheptadecanoic acid (t)b 9,10-Dihydroxyoctadecanoic acid (t)b 9,10-Dihydroxyoctadecanoic acid (e)c Saturated 73, 117, 132, 145 Unsaturated 73, 117, 129, 145 257 (M+. 2 15) 271 285 299 299 299 313 311 313 327 327 327 339 341 355 367 369 397 423 425 73, 117, 129, 149 159, 275 (M+.-15) 173, 289 187, 303 201, 317 215, 331 229, 345 243, 359 73, 117, 129, 149 241 343 241 227 329 215, 331 73, 117, 129, 147 187, 317 187, 317 201, 317 215, 317 215, 317 Tr ( < 0.1) Tr ( < 0.1) 3.4 Tr ( < 0.1) Tr ( < 0.1) 0.4 Tr ( < 0.1) 1.0 43.9 0.5 0.8 0.8 11.7 9.3 Tr ( < 0.1) — 0.2 Tr ( < 0.1) Tr ( < 0.1) Tr ( < 0.1) 1.1 2.1 4.4 5.0 0.6 0.2 Tr ( < 0.1) 0.3 0.2 Tr ( < 0.1) 1.3 0.1 Tr ( < 0.1) 9.7 2.7 0.1 0.2 2.4 0.3 0.2 1.0 0.4 0.6 40.5 0.5 0.7 0.8 22.8 20.2 0.1 0.5 0.8 0.2 0.2 0.3 — 0.1 0.3 0.6 Tr ( < 0.1) Tr ( < 0.1) — 1.6 2.1 0.3 — — — 1.3 0.6 a Relative abundances of compounds determined from areas under peaks in TIC profiles.b Threo isomers.c Erythro isomers. 450 Analyst, 1999, 124, 443–452study before we can claim to have anything like a thorough understanding of the process of mummification. It should be noted here that although derivatisation of the polar oxidised fatty acids, e.g., methylation with tetramethylammonium hydroxide (TMAH) may provide a way forward, by increasing their volatility and thus allowing them to elute satisfactorily from the column, there are significant problems with this approach.44 Hydroxy and dihydroxy, as well as keto acids can undergo undesirable side reactions with the methylating reagent.This leads to the formation of artifact compounds, several nitrogen containing species being produced for each oxidised acid.44 The elucidation of these species is therefore problematic and would necessitate further work to identify the original oxidised fatty acids from which they derive. Such artifacts could also mask other components present which would otherwise be detected.With this in mind, and with convenience and ease of analysis being a major consideration, such a derivatisation step was omitted. Conclusions In summary, the use of TD-GC-MS, Py-GC-MS and GC-MS has demonstrated in both mummies the presence of a complex suite of lipids and proteinaceous components whose compositions indicates extensive alteration via oxidative and hydrolytic processes during long-term interment. The findings demonstrate that though highly polar compounds are lost TD is useful as a ‘fingerprint’ technique for identification of the major classes of lipids in mummies.In addition, it has been shown that sequential TD-GC-MS and Py-GC-MS facilitates the very rapid screening of complex organic materials with sub-milligram sample sizes to reveal ‘free’ and ‘bound’ (possibly polymerised) components. However, polyfunctional polar compounds revealed in conventional GC-MS analyses are not seen when employing the TD- or Py-GC-MS approaches since they are either structurally altered or prevented from eluting from the GC column in their underivatised form due to their high polarities.The results obtained show that those compounds containing two carboxyl groups, a carboxyl group and one or more hydroxyl groups, and two hydroxyl groups are insufficiently volatile to be amenable to TD-GC-MS. In particular the series of aliphatic products revealed by Py-GC-MS (following TD-GC-MS) indicates the likely presence of polymerised lipid.Moreover, the use of sequential TD- and Py-GCFig. 5 Reconstructed total ion chromatogram from the GC-MS of the neutral fraction (after solvent extraction and fractionation) of (a) Horemkenesi, resin-like material and (b) Khnum Nakht, bandage/resin/tissue. MAG, monoacylglycerols; MAlG, monoalkylglycerols. (Neutrals are present as the free compounds or their TMS derivatives.) Table 4 Composition of neutral fraction from gas chromatography/mass spectrometry (GC-MS) following solvent extraction procedures (all compounds determined as the free compounds or as their trimethylsilyl derivatives).Values in bold indicate base peaks. Masses underlined indicate molecular ions (M+.). Other masses are characteristic fragment ions Relative abundance (%)a Compound name Mass spectral data m/z Horemkenesi, resin-like material Khnum Nakht, bandage/resin/tissue Methyl esters— Methyl hexadecanoate Methyl octadecenoate Methyl octadecanoate Methyl 9,10-dihydroxyoctadecanoate 1-Monoacylglycerols— C14:0 C16:1 C16:0 C18:1 C18:0 1-O-Alkylglycerols— C16 C18 Steroidal compounds— Cholesta-3,5-dien-7-one Cholesterol 3-Hydroxycholest-5-en-7-one (7-ketocholesterol) A cholesteryl ester A cholesteryl ester 74, 129, 143; 55, 69, 123; 73, 129, 147 270 296 298 215, 259 73, 103, 129, 147 211, 285, 343, 431 237, 311, 369, 382, 457 239, 313, 371, 459 265, 339, 397,410, 485 267, 341, 399 205, 73, 103, 117, 133, 147 313, 357, 370, 445 341, 385, 398, 473 382, 174, 161, 187, 269, 367 458, 129, 247, 329, 353, 368, 443 472, 73, 129, 174, 367, 382, 457 368, 145, 147, 247, 260, 353 368, 145, 147, 247, 260, 353 9.2 9.6 5.7 4.0 4.0 7.3 7.2 24.7 1.2 14.5 9.3 1.2 2.0 — — — — — — — — — — 0.8 0.3 8.2 5.2 20.7 13.8 9.1 21.5 20.3 a Relative abundances determined from areas under peaks in TIC profiles.Analyst, 1999, 124, 443–452 451MS: (i) serves to minimise contamination and sample losses; (ii) has the advantage of requiring very small sample sizes, essentially eliminating any destruction of the mummified bodies, and (iii) potentially allows the rapid screening of large numbers of samples.Together these advantages serve to aid the chemical investigation of historically valuable museum specimens of this nature. Most importantly, the use of these techniques virtually eliminates any destruction of the mummified bodies thereby allowing us to extend our studies of ancient Egyptian funerary practices.However, ultimately a combination of conventional GC-MS (of extractable components), TDand Py-GC-MS approaches is essential in order ensure that the polyfunctionalised compounds are identified since these may provide valuable compositional information which would otherwise be overlooked, particularly important when studying aged organic residues of uncertain origin. Turning to the embalming agents themselves, none of the classical embalming ‘resins’ was detected, however, an exogenous origin for at least a proportion of the lipids and proteinaceous components cannot be discounted since fats, oils and gelatin have been proposed as embalming agents in mummification.45,46 Further investigations in this laboratory of another nine samples from various parts of the body of Horemkenesi have confirmed the presence of only lipid and proteinaceous constituents.In addition, analyses of similarly small samples of resin-like organic matter taken from several mummies of varying age have shown that these techniques have the capacity to reveal the presence of the more classical embalming substances, e.g., beeswax and tree resins.43 Acknowledgements We thank Rosalie David of Manchester Museum and Sue Giles of Bristol Museum for kindly making the samples available to us.Thanks also goes to Jim Carter and Andy Gledhill for their valued technical assistance and support. Artur Stankiewicz is thanked for his help and advice on Py-GC-MS.NERC provided financial support for mass spectrometry facilities (F14/6/13). References 1 C. Andrews, Egyptian Mummies, British Museum, 6th Impression, 1990, p. 7. 2 A. Lucas, Ancient Egyptian Materials and Industries, Histories and Mysteries of Man Ltd, London, 4th ed., 1989, p. 270. 3 J. Rullkotter and A. Nissenbaum, Naturwissenshaften, 1988, 75, 618. 4 F. O. Gülaçar, A. Buchs and A. Susini, J. Chromatogr., 1989, 479, 61. 5 J. Koller, U. Baumer, Y.Kaup, H. Etspüler and U. Weser, Nature London, 1998, 391, 343. 6 U. Weser, Y. Kaup, H. Etspüler, J. Koller and U. Baumer, Anal. Chem., 1998, 70, 511A. 7 M. M. Wright and B. B Wheals, J. Anal. Appl. Pyrolysis, 1987, 11, 195. 8 A. Nissenbaum, J. Archaeological Sci., 1992, 19, 1. 9 M. L. Proefke, K. L. Rinehart, M. Raheel, S. H. Ambrose and S. U. Wisseman, Anal. Chem., 1992, 64, 105A. 10 R. P. Evershed and R. C. Connolly, Naturwissenschaften, 1988, 75, 143. 11 R. P. Evershed, Archaeometry, 1990, 32, 139. 12 J. R. Cronin, S. Pizzarello, S. Epstein and R. V. Krishnamurthy, Geochim. Cosmochim. Acta, 1993, 57, 4745. 13 K. Kawamura and R. B. Gagosian, J. Chromatogr., 1988, 438, 309. 14 G. Eglinton, D. H. Hunneman and A. McCormick, Org. Mass Spectrom., 1968, 1, 593. 15 R. P. Evershed, H. A. Bland, P. F. van Bergen, J. F. Carter, M. C. Horton and P. A. Rowley-Conwy, Science, 1997, 278, 432. 16 M. Li and R. B. Johns, J. Anal. Appl. Pyrolysis, 1991, 20, 161. 17 M.Li and R. B. Johns, J. Anal. Appl. Pyrolysis, 1990, 18, 41. 18 M. Li, PhD. Thesis, University of Melbourne, Australia, 1991. 19 A. W. Stott and G. D. Abbott, J. Anal. Appl. Pyrolysis, 1995, 31, 227. 20 M. Bjorøy, K. Hall, P. B. Hall, P. Leplat and R. Loberg, J. Chem. Geol., 1991, 93, 1. 21 W. Püttmann, C. B. Eckhardt and R. G. Schaefer, Chromatographia, 1988, 25, 279. 22 J. H. Taylor, Unwrapping a Mummy, British Museum Press, London, 1995, p. 92. 23 F. D. Gunstone, J. L. Harwood and F. B. Padley, The Lipid Handbook, Chapman and Hall Ltd, London, 1986, p. 19. 24 J. Hirsch, Handbook of Physiology, Section 5: Adipose Tissue, American Physiology Society, Washington, DC, 1965, p. 182. 25 B. A. Stankiewicz, P. F. van Bergen, I. J. Duncan, J. F. Carter, D. E. G. Briggs and R. P. Evershed, Rapid Commun. Mass Spectrom., 1996, 10, 1747. 26 B. A. Stankiewicz, J. C. Hutchins, R. Thomson, D. E. G. Briggs and R. P. Evershed, Rapid Commun. Mass Spectrom., 1997, 11, 1884. 27 T. O. Munson and D. D. Fetterolf, J. Anal. Appl. Pyrolysis, 1987, 11, 15. 28 F. O. Gülaçar, A. Susini and M. Klohn, J. Archaeological Sci., 1990, 17, 691. 29 J. Hirsch, Handbook of Physiology, Section 5: Adipose Tissue, American Physiology Society, Washington, DC, 1965, p. 183. 30 S. Dayton, S. Hashimoto, W. Dixon and M. L. Pearce, J. Lipid Res., 1966, 7, 103. 31 A. C. Beynen, R.J.J. Hermus and J. G. A. J. Hautvast, Am. J. Clin. Nutr., 1980, 33, 81. 32 D. A. Leaf, W. E. Connor, L. Barstad and G. Sexton, Am. J. Clin. Nutr., 1995, 62, 68. 33 A. Gellhorn and W. Benjamin, Handbook of Physiology, Section 5: Adipose Tissue, American Physiology Society, 1965, p. 661. 34 J. Hirsch, J. W. Farquhar, E. H. Ahrens, Jr., M. L. Peterson and W. Stoffel, Am. J. Clin. Nutr., 1960, 8, 499. 35 N. Nicolaides, Science, 1974, 186, 19. 36 L. A. Goldsmith, Physiology, Biochemistry, and Molecular Biology of the Skin, Oxford University Press, New York, 2nd edn., 1991, vol. 1, p. 205. 37 G. D. Fasman, Handbook of Biochemistry and Molecular Biology, Lipids, Carbohydrates, Steroids, CRC Press, Boca Raton, FL, 3rd edn., 1975, p. 507. 38 N. Nicolaides, Lipids, 1967, 2, 266. 39 N. Goetz, G. Kaba and P. Boré, J. Chromatogr., 1982, 233, 19. 40 S. Hada, Acta Med. Biol., 1968, 16, 17. 41 R. M. Schultz and M. N. Liebman, in Textbook of Biochemistry with Clinical Correlations, ed. T. M. Devlin, Wiley-Liss, New York, 4th edn., 1997, p. 51. 42 M. O. Dayhoff, Atlas of Protein Sequence and Structure, The National Biomedical Research Foundation, Silver Spring, MD, 1969, vol. 4, p. D-179. 43 S. A. Buckley and R. P. Eveshed, to be published. 44 K. B. Anderson and R. E. Winans, Anal. Chem., 1991, 63, 2901. 45 A. Lucas, Ancient Egyptian Materials and Industries, Histories and Mysteries of Man Ltd, London, 4th edn., 1989, p. 327. 46 A. Lucas, Ancient Egyptian Materiels and Industries, Histories and Mysteries of Man Ltd, London, 4th edn., 1989, p. 3. Paper 8/09022J 452 Analyst, 1999, 124, 443–452

ISSN:0003-2654    

DOI:10.1039/a809022j

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

The determination of polychlorinated biphenyls in municipal sewage sludges using microwave-assisted extraction and gas chromatography-mass spectrometry Guillaume Dupont, Corine Delteil, Valérie Camel* and Alain Bermond Institut National Agronomique Paris-Grignon, GER Chimie Analytique, 16 Rue Claude Bernard, 75231 Paris cedex 05, France. E-mail: camel@inapg.inra.fr Fax: +33 1 44 08 16 53; Tel: +33 1 44 08 17 25 Received 21st December 1998, Accepted 3rd March 1999 The presence of organic micropollutants [such as poly(chlorobiphenyls)] in municipal sewage sludges is a major problem on account of risks associated with the agricultural valorisation of the sludges.In France, since January 1998, maximum values for trace organic pollutants are imposed [0.8 mg kg21 dry matter for the sum of seven poly(chlorobiphenyls)]. The aim of this study was to develop a reliable, accurate and fast analytical procedure (extraction, clean-up, quantification) in order to determine polychlorinated biphenyls in municipal sewage sludges.Such pollutants could be efficiently extracted from dried sewage sludge samples using microwave-assisted extraction. This technique affords several advantages as compared to the classical Soxhlet extraction, mainly rapidity and reduction in solvent consumption. Extractions (10 min under 30 W with 30 ml hexane–acetone 1 : 1) were conducted in the presence of copper to avoid sulfur interferences in the extracts. The latter were further concentrated and purified onto disposable silica cartridges.After final concentration and addition of the internal standard, they were analysed by gas chromatography coupled to mass spectrometry (in the selected ion monitoring mode). Two types of sludges (from Achères and Valenton sewage treatment plants near Paris, France) were analysed, whose polychlorobiphenyl concentrations differed largely (Achères sludge being the most contaminated). The microwave extraction compared favourably with the classical Soxhlet extraction.In addition, the sludge concentrations found with our experimental procedure were close to the analyses made by another laboratory, even though some discrepancy was noted. Aim of investigation Polychlorinated biphenyls (PCBs) have been an important environmental problem for many years, because of the large quantities released into the environment, their persistence, and their potential toxicity to a broad spectrum of organisms, including humans (their lipophilic nature contributes to their high bioaccumulation potential).PCBs were marketed worldwide in large quantities for many years as transformer and capacitor oils, cutting oils, hydraulic oils, heat transfer fluids, and metalcasting release oils. They were also used in carbonless copy paper, paints and pesticides. There are 209 PCB isomers and congeners.1 Several studies reported their gas chromatographic analysis with either electron capture detection or mass spectrometry (GC-MS).2–4 However, the separation of all PCBs remains quite a challenge, even with high resolution capillary GC.2,5 For that reason, only seven congeners are recommended by the Community Bureau of Reference; these are PCB 28, 52, 101, 118, 153, 138 and 180. These congeners have been selected as indicators on the basis of their wide range of chlorination and of their relatively high concentrations in technical PCB mixtures and in the environment.PCBs have been sought in different environmental matrices. Their extraction from sediments and sludges generally includes extraction with an organic solvent, removal of sulfur and cleanup. 2,6,7 Extraction is usually achieved using classical techniques, mainly Soxhlet extraction,2,6,8–11 which is time and solvent consuming. So, recent techniques have been developed in the past few years to enable their rapid extraction from environmental matrices. In particular, microwave-assisted extraction (MAE) gave efficient recoveries of PCBs in reasonable times from water12 as well as soils and sediments,13–16 mostly using closed systems.The concern about the presence of PCBs in sewage sludges is rather recent,17 even though sewage sludges have been used for the fertilization of cultivated lands for several years. In France, the new regulation of 8 January 1998 imposes a maximum acceptable limit for the sum of the seven PCBs (0.8 mg kg21 dry weight).A few studies have been conducted to determine PCB concentrations in sludge samples, giving several values for the sum of PCBs depending on the sludges: 3.3 mg kg21 dry weight,18 1.3 mg kg21 dry weight6 and from 0.106 to 0.712 mg kg21 dry weight.11 In addition, PCBs were found persistent in sewage-sludge amended soils.6,9,10,19 However, severe discrepancy has been observed between results obtained by different laboratories.20 For that reason, there is a need for a fast analytical method that enables accurate determination of PCBs in sewage sludges.Thus, this study was undertaken to develop a fast and efficient analytical procedure to accurately determine PCBs in sewage sludges. Pollutants were extracted from the matrix under microwave energy, purified onto silica, and further analysed using GC-MS. Extraction efficiencies were compared to results obtained using classical Soxhlet extraction (precognised in the US Environmental Protection Agency (EPA) Method 3540 for the determination of PCBs in solid matrices).8 Finally, our results were compared to the values determined by an independent laboratory in order to validate our method.Analyst, 1999, 124, 453–458 453Experimental Sewage sludge samples Sewage sludge samples were obtained from two municipal wastewater stations near Paris, Achères and Valenton. Sludge samples from Achères came from only one sampling (January 1997). In contrast, samples from Valenton were received every two weeks (between July and September 1998), and were supposed to be composite samples over that period.Upon their reception in the laboratory, sludges were kept frozen in order to avoid any sludge modification. Before their analysis, large samples (200 g) were taken and dried in an oven (60 °C, 18 h); they were further homogeneized with a mortar. This treatment was reported to minimize possible losses. They were bottled in a polypropylene box (the absence of contamination from the polypropylene was checked) and stored in the dark at room temperature (to keep them dry).The absence of water should make the matrix more accessible to the organic extractant solvent.2 However, as drying at moderate temperature may result in the presence of residual water, dried sludge samples (10 g) were also kept at 105 °C in an oven for 24 h in order to correctly determine their dry weight. In a few experiments, dried sludge samples were spiked with the PCB congeners before the extraction (at 0.15 mg kg21 dry weight), in order to estimate the extraction recoveries. Reagents and chemicals Reagents and chemicals were supplied as follows: a standard solution of the seven PCB congeners (10 mg l21) in isooctane (PCB 28, 52, 101, 118, 138, 153 and 180), and individual standard solutions, 1,2,3,4-tetrachloronaphthalene (TCN), octachloronaphthalene (OCN) or PCB 29 (each at 10 mg l21) in isooctane by CIL Cluzeau (Paris, France); analytical-reagent grade copper metal and nitric acid solution 68% by Prolabo (Briare, France); HPLC-grade isooctane by Merck (Nogent-sur- Marne, France) and HPLC-grade acetone and n-hexane by Prolabo (Briare, France).PCBs purities were guaranteed between 97 and 99.7%. Other purities were stated to be higher than 99%. Blank experiments were conducted in order to check the absence of PCB contamination in concentrated solvent volumes (i.e., 50 ml concentrated to around 0.6 ml).Five stock standard solutions (at the following concentrations: 20, 50, 80, 150 and 200 mg l21) were prepared by diluting the PCB solution in isooctane. A stock standard solution (1 mg ml21) of TCN (internal standard) was prepared by dilution in isooctane. The internal standard was added and diluted (50 mg l21) in all calibration standards and sample extracts before injection. All solutions were stored at 5 °C in the dark. All glassware was washed first with a water-detergent solution, next cleaned for 24 h in a nitric acid solution (20%) and then dried at 250 °C for 8 h.Activated copper bars (0.5 cm long) were cut and immersed in 30% nitric acid for 30 s. The bars were cleaned sequentially with acetone and hexane. They were further added to the samples before extraction in order to remove sulfur interferences. Estimation of PCBs in ambient air of the laboratory The air concentration was estimated, as possible contamination could occur if paints and rubbers contain PCBs.2 This was done by placing Petri dishes containing 2 g of octadecyl-bonded silica (Lichroprep RP-18, 40–63 mm, from Merck, Darmstadt, Germany) in several rooms of the laboratory for two weeks.After this period, the silica was transferred to a glass column and eluted with diethyl ether–hexane 1 : 9 (10 ml). The extract was further concentrated to around 0.6 ml and analysed by GCMS (after the internal standard addition). Microwave-assisted extraction MAE experiments were performed with a Soxwave 100 open microwave solvent extraction system (Prolabo, Briare, France) whose maximum power is 300 W.A 1 g aliquot of dried sewage sludge was weighted in the extraction cell. The mixture hexane– acetone 1 : 1 (30 ml) was then added to the sample and the solution stirred; this solvent was used as it gave efficient extractions under microwave energy.14,21 Activated copper bars (1 g) were added to each sample just before extraction to remove sulfur by sulfide formation.2,8 Extractions were performed at 30 W during 10 min (unless other conditions specified in the text), as such conditions enabled satisfactory extraction of several pollutants from soils and sediments.22 After cooling to room temperature, the extracts were filtered (using ashless filter papers, 110 mm, Prolabo) to remove the copper and the matrix (no contamination from the filter papers was observed).The glassware was rinsed with 5 ml of hexane–acetone 1 : 1.Next, the extracts were concentrated to approximately 2 ml using a rotary evaporator at room temperature and reduced pressure. The flask was then rinsed two times with 5 ml of n-hexane; this volume was added to the extracts, which were finally concentrated under a gentle stream of nitrogen to approximately 2 ml before clean-up. Soxhlet extraction Extractions were performed with a Soxhlet apparatus as suggested by the US EPA method 3540B.8 Initially, a dried sludge sample (1 g) was extracted with a mixture of hexane– acetone 1 : 1 (250 ml).Activated copper bars (1 g) were added to the sample before extraction in order to remove sulfur compounds. Extractions were performed during 6 h (4–5 cycles per hour) as already performed in another study.6 After cooling to room temperature, the extracts were concentrated to approximately 2 ml using a rotary evaporator at room temperature and reduced pressure. The flask was then rinsed two times with 5 ml of n-hexane; this volume was added to the extracts, which were finally concentrated under a stream of nitrogen to approximately 2 ml before clean-up. Clean-up method The determination of trace pollutants in sewage sludges requires a clean-up step to remove lipids and fats.This is usually achieved on an adsorbent, either silica or Florisil.6 Thus, purification was performed onto disposable solid-phase extraction silica cartridges (Supelclean LC-Si, 1 g, 6 ml, supplied by Supelco, Saint-Quentin Fallavier, France).A Visiprep vacuum manifold system (Supelco) was used. Cartridges were conditioned with 4 ml of n-hexane (the solvent was allowed to soak the entire cartridge for 5 min before passing through the cartridge). Care was taken to ensure that cartridges never dried before sample application. The extracts (2 ml) were transferred on top of the cartridge and allowed to pass through (at approximately 2 ml min21). When the entire extracts were passed through, the sample vials were rinsed with an additional 0.5 ml of solvent.This volume was added to the cartridges. The PCB congeners were finally eluted with 5 ml of hexane (the solvent was allowed to soak the cartridge before elution). Then, the extracts were concentrated under a stream of nitrogen to an appropriate volume (0.6 ml). The internal standard was added and diluted (50 mg l21) in the extracts. Blank experiments were conducted in order to check the absence of PCB contamination from the solid-phase extraction silica cartridges. 454 Analyst, 1999, 124, 453–458Gas chromatography-mass spectrometry Extracts were analysed on a Hewlett-Packard (Avondale, PA, USA) Model 5890 Series II gas chromatograph interfaced to a Hewlett-Packard 5971A MS Engine mass spectrometer MS/ DOS ChemStation and equipped for some extracts with a Hewlett-Packard 6890 Series autosampler. The acquisition was performed with the G1034C© program (M03.65.06 Version by Hewlett-Packard 1989-94).The samples were analysed on a 50 m 3 0.22 mm id 3 0.25 mm film thickness HT-8 (or 1,7-dicarba-closo-dodecarborane phenylmethyl siloxane) silica capillary open-tubular column, as this column gave enhanced selectivity in the analysis of PCBs and supported elevated temperatures.5 The column was protected with a 1 m 3 0.22 mm id deactivated non-polar fused-silica capillary column. Unless specified, the column temperature was held at 50 °C for 1 min and then increased at 30 °C min21 to 180 °C, subsequently programmed at 6 °C min21 to 300 °C and finally increased at 30 °C min21 to 360 °C where it was held for 5 min.The total analysis time was 32.3 min. The carrier gas was hydrogen at a linear velocity of 35–40 cm s21 for 250 °C (column head pressure: 18 psi). The injection volume was 1.5 ml and the injection temperature 290 °C. The injector was set in the splitless mode with split vent closed during 1 min after injection.The interface temperature was 300 °C. The electron energy was set at 70 eV and spectral data were acquired at a rate of 4.1 scan s21. The MS detector was operating in the selected ion monitoring (SIM) mode and the m/z values monitored were 256 and 258 for dichlorobiphenyls (PCB 28 and 29), 292 and 294 for tetrachlorobiphenyl (PCB 52), 266 and 268 for TCN (internal standard), 326 and 328 for pentachlorobiphenyls (PCBs 101 and 118), 360 and 362 for hexachlorobiphenyls (PCB 138 and 153), and 324 and 326 for heptachlorobiphenyl (PCB 180).When OCN was used as an internal standard, monitored m/z values for this compound were 402 and 404. The instrument was tuned weekly with perfluorotributylamine using the Automatic Tune internal program. In addition, to increase sensitivity, a 200 V overpotential was applied to the electronic multiplier. Results and discussion Choice of the chromatographic conditions Separation of the seven congeners and the internal standard was easily achieved using a low ramp temperature as indicated in the experimental part of this paper.The final column temperature was rather high (i.e., 360 °C) in order to remove contaminants from the column, and thereby to increase the column life. In our preliminary experiments, OCN was used as the internal standard, as suggested by several previous results.1 However, this compound had a retention time (37.4 min) very different from the seven congeners, due to its much lower volatility.So, we found it preferable to use an internal standard whose volatility is closer to the seven PCB volatility, to avoid discrimination inside the injector as well as to reduce the total analysis time. So, TCN was chosen as precognised by several studies;2,6 in that way, complete separation could be achieved within 32.3 min. In the splitless injection mode, the residence time of the needle inside the injector chamber is of prime importance to achieve satisfactory and repeatable injections.As illustrated in Table 1, results from a Soxhlet extract showed a higher overall precision with slow injections (i.e., 3 s). Besides, for the less volatile compounds (PCB 153, 138 and 180) the concentrations determined increased upon slow injection (from 20 to 45%); this clearly shows that these compounds were insufficiently volatilised during rapid injection (i.e., 1 s). As this injection time effect was not observed with standard solutions, it was due to matrix effects.We assumed that the presence of the matrix hindered PCB volatilisation (due to preferential volatilisation of matrix interferences), leading to better results with slow injections. Successive injections of sludge extracts resulted in frequent clogging of the mass spectrometer, leading to severe sensitivity loss. As an illustration, Table 2 presents the variations of the congener response coefficients (relative to the internal standard TCN) over a period of about six months.Due to elevated values of the relative standard deviations (RSDs), the calibration standard solutions were injected alternately with sludge extracts, in order to redraw the internal calibration curves daily. It can also be noted that the mass spectrometer sensitivity decreased as the number of chlorine atoms in the molecule increased. As an example, the response coefficient of PCB 180 was six times lower than that of PCB 28.Efficiency of the clean-up step The efficiency of the silica clean-up step was first investigated with standard solutions of the seven congeners. Solutions at the 100 mg l21 level (for each PCB) in hexane–acetone 1 : 1 (2 ml) were applied on top of silica cartridges, already conditioned with hexane–acetone 1 : 1. PCBs were analysed in the hexane– acetone 1 : 1 fraction (5 ml). Recoveries were quite low (68–81%) due to incomplete retention of the PCBs as well as insufficient elution of the retained compounds.To improve this step, similar experiments were performed, replacing hexane– acetone by hexane in each step (conditioning, sample solvent, and elution solvent). This led to improved efficiencies (mean recoveries between 85 and 109%). Quantitative recoveries were obtained for most of the congeners; the slightly lower values for PCB 28 and 52 are probably the result of volatilisation upon the extract concentration under a gentle stream of nitrogen.In addition, during the purification of sludge extracts, we noted that hexane–acetone was a less selective eluting solvent, as polar interferences were eluted by acetone. So, hexane was used Table 1 Influence of the injection time on the PCB concentrations found in a sludge extract PCB congeners 28 52 101 118 153 138 180 Injection time: 1 s Concentration in the extract/mg l21 20.1 15.9 26.0 19.4 54.0 37.3 39.3 RSD (%) 13.7 17.3 4.4 8.7 10.3 9.4 10 Injection time: 3 s Concentration in the extract/mg l21 23.1 19.9 25.2 21.5 63.7 45.5 56.7 RSD (%) 10.5 12.9 5.8 6.6 4.4 4.1 3.6 Table 2 Mean values of the response coefficient (relative to TCN as the internal standard) for the seven congeners, along with variation over timea PCB congeners 28 52 101 118 153 138 180 Relative response coefficient 1.33 0.84 0.54 0.52 0.29 0.28 0.22 RSD (%) 15.37 20.19 26.29 23.37 31.41 27.27 26.69 a This study was conducted over nearly six months, using 31 calibration curves.Analyst, 1999, 124, 453–458 455as the extract solvent before the clean-up step, and as the elution solvent. Due to the high selectivity of the SIM detection mode, only minor modifications of the chromatogram were obtained upon purification. Indeed, a large peak around 12 min was eliminated. However, the purification step was required to minimize the mass spectrometer clogging and column contamination. Estimation of PCBs in ambient air of the laboratory The presence of the more volatile PCBs (congeners 28, 52, 101 and 118) was found in two rooms, where either extraction (and further evaporatory concentration) or clean-up were performed. The corresponding concentrations in the C18 bonded silica extract are given in Table 3.The observed concentrations were over 0.5 ng g21, thereby indicating a significant contamination in the air.2 This ambient air contamination was assumed to be due to partial losses of these volatile compounds during the experimental procedure, especially during the evaporatory concentration step. Blank experiments (with solvent extraction) were conducted to check that the presence of PCBs in ambient air did not result in extracts contamination. Efficiency of the MAE extraction Evaluation of the extraction efficiency was performed with spiked sludge samples (from Valenton).All extractions were performed in triplicate, and each extract was injected three times in the gas chromatograph. Sludge samples were preextracted under similar conditions to remove the initial PCBs; then, once the residual solvent had been completely evaporated, these samples were spiked with 150 ml of a standard solution containing the seven congeners at the 1 mg l21 level, leading to a sludge spiking of 150 mg kg21 dry weight (i.e., close to the initial sludge concentrations).Extraction efficiencies were estimated based upon final analysis of the purified extracts. Satisfactory results could be obtained, the overall recoveries ranging from 88 to 105% (mean recovery: 94.6%).As these values take into account several steps additional to the extraction (concentration, purification, and final concentration), we considered that microwave-assisted extractions were satisfactory. Comparison of MAE and Soxhlet extractions Extractions of Achères sludge samples carried out by MAE and Soxhlet apparatus were compared. They were performed in triplicate, and each extract was injected three times in the gas chromatograph. Results are presented in Table 4.Even though overall better extractions were achieved by Soxhlet, efficiencies of MAE were satisfactory (between 81 and 116%) with regard to the time required (10 min for MAE versus 6 h for Soxhlet). The lower precision for MAE results was partly attributed to the filtration step required after MAE, as it may lead to partial losses. The PCB 28 was slightly better recovered using MAE than Soxhlet extraction; this could be related to the stronger concentration of the extract in the latter case (because of a larger solvent volume), resulting in higher losses of this volatile compound. This was confirmed by the presence of this compound (as well as traces of PCB 52 and 101) in the laboratory ambient air.Effects of MAE parameters MAE recoveries may be influenced by the extraction time, the microwave power supplied, and the solvent volume used. So, their effect has been investigated. Results presented in Fig. 1 show that increasing the extraction time (from 10 to 30 min) resulted in lower mean recoveries, possibly due to losses upon volatilization. Slightly better results were obtained with 10 min extractions under 90 W instead of 30 W. Yet, as this power resulted in strong heating of the sample (experiments conducted under 120 W showed excessive heating, with possible explosions), a power of 30 W was preferable. Finally, the solvent Table 3 Concentrations of the PCB congeners found in octadecyl-bonded silica (2 g) to estimate contamination in an ambient air laboratory PCB congeners 28 52 101 118 Concentration/mg kg21 Extraction room 3.7 47.5 33.3 9.2 Clean-up room 1.6 4.6 3.2 Traces Table 4 Comparison of Achères sludge concentrations determined with either MAEa or Soxhlet extractionb PCB congeners 28 52 101 118 153 138 180 Sum MAE Sludge concentration/mg kg21 85.9 48.8 82.4 60.0 174.2 117.2 128.8 697.3 RSD (%) 21.0 39.0 31.7 27.9 32.8 30.6 29.9 27.4 Soxhlet Sludge concentration/mg kg21 73.8 48.9 89.3 73.0 211.5 141.9 159.2 797.5 RSD (%) 17.9 32.2 13.5 17.8 10.4 12.2 12.5 10.0 MAE/Soxhlet (%) 116.4 99.8 92.3 82.3 82.4 82.6 80.8 87.4 a Dried sludge (1 g) + activated copper (1 g) extracted with 30 ml hexane–acetone 1 : 1 during 10 min under 30 W (injection time for GC: 1 s).b Dried sludge (1 g) + activated copper (1 g) extracted with 250 ml hexane–acetone 1 : 1 during 6 h (injection time for GC: 1 s).Fig. 1 Influence of MAE conditions on the PCB concentrations found in Achères sludge samples. GC conditions: autosampler injection (3 s); column temperature: 50 °C (1 min)–180 °C at 30 °C min21, to 300 °C at 6 °C min21, to 360 °C (5 min) at 30 °C min21 (total analysis time: 32.3 min). 456 Analyst, 1999, 124, 453–458volume (50 ml instead of 30 ml) had no significant influence on the recoveries. So, the following conditions were chosen: 30 ml solvent, 10 min extraction under 30 W.Additional experiments were conducted in order to estimate losses of volatile PCBs throughout the experimental procedure, as evidence of such losses has been given (i.e., recovery of spiked PCB 28 around 80%, and presence of volatile congeners in laboratory ambient air). So Valenton sludge samples were spiked with PCB 29 (at the 50 mg kg21 level) as this non-natural congener is often suggested as a recovery standard (because its volatility allows the detection of evaporation losses).2 The overall recoveries were around 65–70% for this compound.As the congeners 28, 52 and 101 are less volatile than PCB 29, their recoveries should be higher as already estimated for sludge samples spiked with the native PCBs. Finally, our complete experimental procedure is detailed in Fig. 2. Comparison of our results with those of an independent laboratory Sludge samples were regularly taken in the Valenton plant in order to make composite samples over two weeks.These samples were further divided in two parts. The first one was analysed by our laboratory, and the second one by an independent laboratory (Institut Pasteur, Lille, France). The experimental procedure used in this laboratory was as follows: Soxhlet extraction of dried sludge (4–5 g) with 100 ml hexane– acetone 1 : 1 during 6 h; then the extract was concentrated up to 1 ml and cleaned-up onto Florisil (1 g). PCBs were further eluted with hexane (10 ml). The final extract was again concentrated up to 1 ml and analysed by GC coupled to electron capture detection (after addition of an internal standard).As shown in Figs. 3 and 4, our results compared favourably with the concentrations determined by the other laboratory. This was quite satisfactory, as previous results from interlaboratory studies on the determination of PCBs in sediment samples showed severe discrepancy between results (differences be- Fig. 2 Experimental procedure for the determination of PCB in sludge samples.Fig. 3 Comparison of individual PCB concentrations determined in several Valenton sludge samples by our laboratory (INA-PG) and an independent laboratory (Institut Pasteur). Analyst, 1999, 124, 453–458 457tween laboratories might be nearly one order of magnitude).1 However, depending on the sample to be analysed, differences occurred between their results and our values. In fact, it must be pointed out that the samples received by both laboratories might have differed somewhat, as sludge samples had been taken from a sludge tank, and sent immediately to the laboratories without any treatment.In particular, they had not been homogeneized, so that the samples received in both laboratories were not exactly the same. So, future experiments will be conducted on composite samples that will be homogeneized in the wastewater treatment plant, before being divided in sub-samples sent to the laboratories involved in the study.This will avoid variability due to the sample heterogeneity (which is crucial for sewage sludges). Acknowledgements The authors thank the SIAAP (Syndicat Interdépartemental d’Assainissement de l’Agglomération Parisienne) and the SEDE company for their contribution to this study and their financial support. They also thank the Prolabo company for the loan of the microwave unit used in this study. Finally, they express their gratitude to Drs L. Eveleigh and S.Bourgeois for their scientific contributions and useful advices. References 1 V. Lang, J. Chromatogr., 1992, 595, 1. 2 F. Smedes and J. de Boer, Trends Anal. Chem., 1997, 16, 503. 3 G. Frame, Anal. Chem., 1997, 69, 468A. 4 M. D. Mullin, C. M. Pochini, S. McCrindle, M. Romkes, S. Safe and L. Safe, Environ. Sci. Technol., 1984, 18, 468. 5 B. Larsen, M. Cont, L. Montanarella and N. Platzner, J. Chromatogr., 1995, 708, 115. 6 I. Folch, M. T. Vaquero, L. Comellas and F. Broto-Puig, J.Chromatogr. A., 1996, 719, 121. 7 P. Hess, J. de Boer, W. P. Cofino, P. E. G. Leonards and D. E. Wells, J. Chromatogr. A., 1995, 703, 417. 8 U.S. EPA Methods, Test Methods for Evaluating Solid Waste, Physical/Chemical Methods, SW-846, U.S. Environmental Protection Agency, 3rd edn., proposed update II, November1992. 9 E. Eljarrat, J. Caixach and J. Rivera, Environ. Sci. Technol., 1997, 31, 2765. 10 A. Marcomini, P. D. Capel, Th. Lichtensteiger, P. H. Brunner and W. Giger, J. Environ. Qual., 1989, 18, 523. 11 R. E. Alcock and K. C. Jones, Chemosphere, 1993, 26, 2199. 12 F. I. Onuska and K. A. Terry, J. High Resol. Chromatogr., 1995, 18, 417. 13 V. Lopez-Avila, J. Benedicto, C. Charan, R. Young and W. F. Beckert, Environ. Sci. Technol., 1995, 29, 2709. 14 V. Lopez-Avila, R. Young, J. Benedicto, P. Ho, R. Kim and W. F. Beckert, Anal. Chem., 1995, 67, 2096. 15 V. Lopez-Avila and J. Benedicto, Trends Anal. Chem., 1996, 15, 334. 16 A. Pastor, E. Vazquez, R. Ciscar and M. de la Guardia, Anal. Chim. Acta, 1997, 344, 241. 17 R. L. Chaney, J. A. Ryan and G. A. O’Connor, Sci. Total Environ., 1996, 185, 187. 18 T. Davis, J. Pyle, J. Skillings and N. Danielson, Bull. Environ. Contam. Toxicol., 1981, 27, 689. 19 B. C. Fairbanks, A. O’Connor and S. E. Smith, J. Environ. Qual., 1987, 16, 18. 20 F. Acobas, V. Beluze, D. Benanou, C. Bonnin, L. Herremans and P. Sztajnbok, ADEME Journées techniques des 5 et 6 Juin 1997 - Aspects sanitaires et environnementaux de l’épandage des boues d’épuration urbaines, Paris, 1997, 111. 21 V. Lopez-Avila, R. Young and W. F. Beckert, Anal. Chem., 1994, 66, 1097. 22 H. Budzinski, A. Papineau, P. Baumard and P. Garrigues, C. R. Acad. Sci. Paris, 1995, 321, 69. Paper 8/09893J Fig. 4 Comparison of the concentrations determined for the sum of the seven congeners in several Valenton sludge samples by our laboratory (INA-PG) and an independent laboratory (Institut Pasteur). 458 Analyst, 1999, 124, 453–458

ISSN:0003-2654    

DOI:10.1039/a809893j

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Headspace solid-phase microextraction for the determination of trace levels of taste and odor compounds in water samples Mingliang Bao,a Osvaldo Griffini,*b Daniela Burrini,b Daniela Santianni,b Katia Barbierib and Marco Mascinia a Department of Public Health, Epidemiology and Environmental Analytical Chemistry, University of Florence, Via G. Capponi 9, 50121 Florence, Italy b Water Supply of Florence, Via Villamagna 39, 50126 Florence, Italy Received 5th October 1998, Accepted 27th January 1999 The use of the headspace solid-phase microextraction (SPME) technique, combined with gas chromatography-ion-trap detection mass spectrometry (GC-ITDMS), for the determination of 34 taste- and odor-causing organic compounds in water is presented.The compounds studied include aliphatic hydrocarbons, aldehydes, ketones and alcohols. The factors affecting the headspace SPME process, such as fiber type, salt addition, stirring, headspace volume and sampling time, were examined.The polydimethylsiloxanedivinylbenzene- coated fiber was found to be effective for the extraction of the compounds studied. The precision of the method was evaluated with spiked bidistilled water and river water samples. The RSDs obtained were similar for both water samples and in the range 4.3–17.1%. Using the standard addition calibration method, the problem of matrix effects observed for river water samples can be reduced. The method showed good linearity over two orders of magnitude of concentration in river water.With 40 ml of water sample, the detection limits were lower than 1 ng l21 for 2-methylisoborneol and geosmin, and 0.8–50 ng l21 for the other compounds. Introduction The control of taste and odor problems in drinking water is of great importance to water utilities because the taste and odor of the drinking water are the primary criteria that consumers use for judging the quality and acceptability of their water supply.Numerous surveys of customer attitudes and opinions about the quality of drinking water, both in Europe and in the USA, have shown that the public generally is more concerned about how their water tastes and smells than about other issues regarding their water supply, and taste and odor complaints from consumers are frequently the major problems received by the drinking water suppliers.1 Worldwide, most of the taste and odor problems in drinking water supplies appear to be caused by the presence of certain metabolites, such as aliphatic hydrocarbons, aldehydes, ketones, alicyclic alcohols and sulfur-containing compounds, produced by algae and microbiological processes in raw water or in finished water storage facilities and piping.2–4 For example, most of the earthy–musty odor problems have been reported to be caused by two biogenic alicyclic alcohols: 2-methylisoborneol (MIB) and geosmin.3 Because of the varying characteristics of the organic compounds present in raw water over a range of concentrations, the control and treatment of the taste and odor problem are difficult if the specific causes of a particular off-flavor are unknown.The analytical methods used for the detection and quantification of taste- and odorcausing compounds in water samples include sensory and instrumental analysis. Sensory analysis is based on the use of trained human noses (panelists), common descriptive terms and reference standards. Results obtained with sensory analysis, such as the flavor-profile analysis (FPA) method,5 include the description of each flavor and its respective intensity.However, sensory analysis is not always reliable since there can be marked differences in the response, not only between individuals to specific compounds,6 but also of individuals from day to day.7 Because a variety of organic compounds are present in water, and many known taste- and odor-causing compounds, such as geosmin and MIB, can cause off-flavor problems at concentrations as low as a few ng l21, the instrumental analytical methods require very low detection limits and a high discriminatory capability.The most suitable method for the determination of taste- and odor-causing compounds at trace levels present in water involves a preconcentration step in combination with capillary gas chromatography (GC) followed by mass spectrometry (MS) for compound identification. A variety of methods, including closed-loop stripping analysis (CLSA),8,9 open stripping analysis (OSA),10 liquid–liquid extraction (LLE),11 solid-phase extraction (SPE) with different adsorbents, 12,13 hollow fiber stripping analysis (HFSA)14 and liquid–liquid microextraction (LLME),15 have been reported to date for the preconcentration of taste- and odor-causing compounds from water samples.CLSA, OSA and HFSA in combination with GC–MS can detect low ng l21 concentrations of taste- and odor-causing compounds, such as MIB and geosmin, in water, but these methods are time consuming and expensive because of the specialized equipment.SPE and LLME are rapid and inexpensive, but to achieve the required limits of detection, a concentration step (solvent evaporation) is required, which increases the sample preparation step and may also cause the loss of volatile analytes during the evaporation. Recently, solid-phase microextraction (SPME), developed by Pawliszyn and co-workers,16,17 has become popular in environmental analysis.SPME is a convenient and solvent-free extraction method that involves the use of a thin polymer-coated silica fiber to adsorb analytes of interest from a sample matrix. This method combines extraction, concentration and sample introduction in one step, and has been shown to be efficient for the extraction of organic compounds with different volatility and polarity from different environmental samples, such as water,18–23 air24,25 and soil,26,27 flavors in beverages and food,28–31 and drugs in biological matrices, such as human blood and urine.32,33 The purpose of this paper was to develop and optimize an SPME procedure for the determination of trace levels of tasteand odor-causing compounds belonging to different classes in Analyst, 1999, 124, 459–466 459water.The method is based on the extraction of the analytes of interest from the headspace over the water sample with SPME followed by gas chromatography-ion-trap detection mass spectrometry (GC-ITDMS) analysis.Factors affecting the SPME process, such as the extraction mode, fiber type, effects of salt addition and stirring, headspace volume over the water sample, precision, linear range and detection limits, were examined. Experimental Reagents Table 1 lists the 34 taste- and odor-causing compounds studied. The compounds selected represent four groups: aliphatic hydrocarbons (a-pinene, b-pinene, camphene, 2-carene, 3-carene, a-terpinene, g-terpinene, limonene, 2,6-dimethyl- 2,4,6-octatriene), aldehydes (C6–C10 linear aldehydes, benzaldehyde, citronella, b-cyclocitral, citral, 2,4-decadienal), ketones (2-methyl-3-heptanone, 6-methyl-5-hepten-2-one, fenchone, camphor, pulegone, geranylacetone, b-ionone) and alcohols (dihydromyrcenol, linalool, borneol, menthol, MIB, a-terpineol, geosmin).The standards of all these compounds were purchased from Aldrich (Milwaukee, WI, USA) with the exception of MIB (99.9%) and geosmin ( > 98%) which were obtained from Wako Pure Chemicals, Ltd.(Osaka, Japan). Stock standard solutions of each compound at 1 mg ml21 were prepared with pure analyte dissolved in methanol, and then diluted with methanol to prepare mixed working standard solutions. Stock standard solutions were kept at 220 °C. Sodium chloride (NaCl) of analytical grade (Merck, Darmstadt, Germany) was previously heated at 550 °C for 8 h. Apparatus The SPME holder for manual sampling and SPME fibers were obtained from Supelco (Bellefonte, PA, USA).Four commercially available SPME fibers differing in sorbent phase coating [100 mm polydimethylsiloxane (PDMS), 65 mm polydimethylsiloxane- divinylbenzene (PDMS-DVB), 65 mm carbowaxpolydimethylsiloxane (CW-DVB) and 75 mm carboxen-polydimethylsiloxane (CX-PDMS)] were tested and compared in this study. The fibers were conditioned in the GC injector port according to the manufacturer’s instructions. A magnetic stirrer from VELP Scientifica (Milan, Italy) was used for stirring the water samples during the SPME procedure.Analyses were carried out in a Varian (Walnut Creek, CA, USA) 3400 GC system coupled to a Finnigan (San Jose, CA, USA) Mat 800 ion-trap detection mass spectrometer. A 30 m 3 0.25 mm id (0.25 mm film thickness) DB-5 coating fused-silica capillary column (J & W Scientific, Folson, CA, USA) was used. The GC oven was held at 40 °C for 1 min, increased to 130 °C at 4 °C min21 and then from 130 to 280 °C at 10 °C min21. The carrier gas was helium at 9.65 3 104 Pa.The injection port was kept at 240 °C for PDMS, PDMS-DVB and CW-DVB fibers, and at 280 °C for the CX-PDMS fiber. Injections (fiber desorption) were carried out in the splitless mode and the split valve was closed for 3 min. Preliminary experiments showed that complete desorption was achieved for all the extracted analytes after 3 min of desorption at a temperature of 240 °C for PDMS, PDMS-DVB and CW-DVB Table 1 Nomenclature of taste- and odor-causing compounds investigated in this study Common name IUPAC name Mr a CAS NRb Hexanal Hexaldehyde 100.2 66-25-1 Heptanal Heptaldehyde 114.2 111-71-7 a-Pinene 2,6,6-Trimethylbicyclo(3.1.1)hept-2-ene 136.2 80-56-8 2-Methyl-3-heptanone Butyl isopropyl ketone 128.2 13019-20-0 (+)-Camphene 2,2-Dimethyl-3-methylene-[1r]-bicyclo(2.2.1)heptane 136.2 5794-03-6 Benzaldehyde Benzaldehyde 106.1 100-52-7 b-Pinene 6,6-Dimethyl-2-methylenebicyclo(3,1,1)heptane 136.2 18172-67-3 6-Methyl-5-hepten-2-one 6-Methyl-5-hepten-2-one 126.2 110-93-0 2-Carene 3,7,7-Trimethylbicyclo(4.1.0)hept-2-ene 136.2 554-61-0 Octanal Octylaldehyde 128.2 124-13-0 3-Carene 3,7,7-Trimethylbicyclo(4.1.0)hept-3-ene 136.2 13466-78-9 a-Terpinene 1-Isopropyl-4-methyl-1,3-cyclohexadiene 136.2 99-86-5 (+)-Limonene (R)-4-Isopropenyl-1-methyl-1-cyclohexene 136.2 5989-27-5 g-Terpinene 1-Isopropyl-4-methyl-1,4-cyclohexadiene 136.2 99-85-4 Dihydromyrcenol 2,6-Dimethyl-7-octen-2-ol 156.3 18479-58-8 (2)-Fenchone (2)-1,3,3-Trimethyl-2-norbornanone 152.2 7787-20-4 (±)-Linalool dl-3,7-Dimethyl-3-hydroxy-1,6-octadiene 154.3 78-70-6 Nonanal Nonylaldehyde 142.4 124-19-6 2,6-Dimethyl-2,4,6-octatriene 2,6-Dimethyl-2,4,6-octatriene 136.2 673-84-7 (+)-Camphor 1,7,7-Trimethylbicyclo(2.2.1)heptan-2-one 152.2 464-49-3 (R)-(+)-Citronella 3,7-Dimethyl-6-octenal 154.3 2385-77-5 (2)-Borneol (1s)-endo-1,7,7-Trimethylbicyclo(2.2.1)heptan-ol 154.3 464-45-9 (-)-Menthol 5-Methyl-2-(1-methylethyl)cyclohexanol 156.3 2216-51-5 2-Methylisoborneol (MIB) 1,2,7,7-Tetramethyl-exo-bicyclo(2.2.1)heptan-2-ol 168.3 237-42-8 a-Terpineol (s)-2,2,4-Trimethyl-3-cyclohexene-1-methanol 154.3 10482-56-1 Decanal Decylaldehyde 156.3 112-31-2 b-Cyclocitral 2,6,6-Trimethyl-1-cyclohexene-1-carboxaldehyde 152.2 432-25-7 (+)-Pulegone (R)-5-Methyl-2-(1-methylethylidene)-cyclohexanone 152.2 89-82-7 (±)-Citral 3,7-Dimethyl-2,6-octadienal 152.2 5392-40-5 (2)-Bornyl acetate [1s]-1,7,7-Trimethylbicyclo(2.2.1)heptan-ol, acetate 196.3 5655-61-8 2,4-Decadienal trans,trans-2,4-Decadienal 152.2 25152-84-57 Geosmin trans-1,10-Dimethyl-trans-9-decalol 182.3 19700-21-1 Geranylacetone trans-6,10-Dimethyl-5,9-undecadien-2-one 194.3 3796-70-1 b-Ionone 4-(2,6,6-Trimethyl-1-cyclohexen-1-yl)-3-buten-2-one 192.3 79-77-6 a Mr, relative molecular mass.b CAS RN, Chemical Abstracts Service Registry Numbers. 460 Analyst, 1999, 124, 459–466 Finally, 30% of NaCl was added to all samples in further experiments and sampling was performed with magnetic stirring. Effects of headspace volume The effects of the headspace volume on the extraction of the compounds studied by headspace SPME were investigated as follows. One set of experiments was performed using a constant vial (62 ml), but with different water volumes (10, 20, 30 and 40 ml). In this case, the percentage headspace decreased from 81.9 to 27.4% when the water volume was increased from 10 to 40 ml.Typical results are shown in Fig. 4. For non-polar compounds, such as a-pinene, camphene and 2-carene, a nearly linear increase in response was observed when the percentage headspace decreased from 81.9 to 27.4. These results indicate again that the rate-controlling step in the headspace SPME process for compounds with high volatility is the diffusion of the analyte into the SPME fiber, since the analytes diffuse quickly to the fiber coating when the headspace volume is smaller.For the polar compounds studied, the response significantly increases when the percentage headspace decreases from 81.9 to 63.7; a decrease in the percentage headspace from 63.7 to 27.4 only produces a slight increase in the response, especially for compounds such as pulegone, MIB, geosmin and b-ionone. As mentioned before, the mass transfer of analytes from the liquid phase to the headspace is often the limiting factor in the headspace SPME process for polar compounds.Thus, when the water volume increases, the polar analytes will take more time to transfer from the liquid to the headspace phase. Another set of experiments was performed using a constant percentage headspace (27.4%), but with different vials: 62 ml vial, 36 ml vial (6.5 3 2.7 cm) and 17 ml vial (5.5 3 2.0 cm). In this case, the water volume was 40, 26 and 11 ml, respectively. Fig. 5 shows the typical results. For the non-polar compounds studied, the responses obtained with the 62 ml and 36 ml vials were similar, but significantly higher than that obtained with the 17 ml vial.On the other hand, the response obtained for the polar compounds studied decreased with the decrease in vial size from 62 ml to 17 ml. Additionally, triplicate determinations showed that the relative standard deviations (RSDs) obtained decreased as the vial size increased; the mean RSD value for the 34 compounds studied was 6.8% with 62 ml vials, 8.2% with 36 ml vials and 11.7% with 17 ml vials.As a Fig. 6 Extraction–time profiles for the taste- and odor-causing compounds studied in water by headspace SPME using the PDMS-DVB fiber. Table 2 Precision (RSD) of the proposed headspace SPME-GC-ITDMS method for taste- and odor-causing compounds spiked in different water matricesa Bidistilled River water water Relative Spiking RSD recoveryb RSD Compound level/ng l21 (%) (%) (%) Hexanal 50, 200, 500 14.1 67.0 17.2 Heptanal 50, 200, 500 8.2 79.3 7.9 a-Pinene 50, 200, 500 6.8 67.9 6.9 2-Methyl-3-heptanone 50, 200, 500 5.9 82.0 4.5 (+)-Camphene 50, 200, 500 5.5 74.5 5.9 Benzaldehyde 50, 200, 500 14.8 89.0 15.9 b-Pinene 50, 200, 500 4.8 73.7 4.5 6-Methyl-5-hepten-2-one 50, 200, 500 7.9 76.8 9.6 2-Carene 50, 200, 500 5.1 66.2 6.4 Octanal 50, 200, 500 6.7 65.6 9.2 3-Carene 50, 200, 500 4.6 64.0 5.7 a-Terpinene 50, 200, 500 5.9 65.2 6.3 (+)-Limonene 50, 200, 500 7.1 65.9 7.9 g-Terpinene 50, 200, 500 4.6 64.6 4.9 Dihydromyrcenol 50, 200, 500 7.1 65.9 5.7 (2)-Fenchone 50, 200, 500 4.5 79.0 4.7 (±)-Linalool 50, 200, 500 8.6 61.1 8.1 Nonanal 50, 200, 500 5.8 61.6 10.1 2,6-Dimethyl-2,4,6-octatriene 50, 200, 500 4.7 60.3 4.3 (+)-Camphor 50, 200, 500 7.1 82.5 7.6 (R)-(+)-Citronella 50, 200, 500 6.0 71.4 5.6 (2)-Borneol 50, 200, 500 6.1 64.3 7.9 (2)-Menthol 50, 200, 500 6.1 65.4 5.6 MIB 5, 20, 50 6.4 78.8 6.8 a-Terpineol 50, 200, 500 5.8 61.5 5.3 Decanal 50, 200, 500 9.6 62.0 12.3 b-Cyclocitral 50, 200, 500 5.9 76.4 4.3 (+)-Pulegone 50, 200, 500 7.1 68.8 7.9 (±)-Citral 50, 200, 500 10.9 61.2 10.2 (2)-Bornyl acetate 50, 200, 500 5.4 76.8 9.9 2,4-Decadienal 50, 200, 500 11.5 62.1 14.3 Geosmin 5, 20, 50 6.2 73.5 5.3 Geranylacetone 50, 200, 500 5.7 69.6 8.7 b-Ionone 10, 40, 100 7.1 74.1 7.1 a Water sample volume was 40 ml, containing 30% of NaCl; sample vial volume was 62 ml; sampling time was 40 min with stirring; four determinations were performed for each spiking level.b Relative recoveries for spiked river water were calculated relative to the spiked bidistilled water after blank correction. Analyst, 1999, 124, 459–466 463result of these data, the sample volume selected for further experiments was 40 ml in a 62 ml vial. Extraction–time profile We studied the extraction–time profile between 5 and 120 min. Fig. 6 shows the results obtained. It is evident that the time needed to reach equilibrium depends on the polarity and the relative molecular mass of the analyte.For the non-polar compounds studied, extraction equilibrium was reached in 10 min, while for the polar compounds studied, the equilibration times ranged from 10 min to more than 120 min, and generally increased with increasing relative molecular mass of the analyte. For example, hexanal and heptanal reached extraction equilibrium in 10 and 20 min, respectively, while for nonanal and decanal, equilibrium was not reached even after 120 min.Based on the curves shown in Fig. 6, an extraction time of 40 min was selected for further experiments, because this provides sufficient extraction (most analytes reaching more than 80% of their final equilibrium value by 40 min) and allows the headspace SPME procedure to be performed approximately in the same time as that required for GC analysis. Precision, linearity and detection limits The precision of the proposed headspace SPME method in optimized conditions was assessed by analyzing spiked samples of bidistilled Milli-Q water and river water.The spiked levels were 5, 20 and 50 ng l21 for MIB and geosmin, 10, 40 and 100 ng l21 for b-ionone and 50, 200 and 500 ng l21 for the other compounds studied. For each level and each type of aqueous sample, four extractions were performed. The results are reported in Table 2. A comparison of the data shows that the RSD values obtained from spiked river water samples were similar to those obtained from spiked bidistilled water samples and ranged from 4.3 to 17.2%.The data on the relative recovery (%) listed in Table 2 from spiked river water samples were calculated by normalizing to the results obtained from spiked bidistilled water samples after correcting for the data obtained from non-spiked river water samples. The relative recoveries from spiked river water samples were between 58.1 and 89.0%.Based on these data, the water matrix seems to have an appreciable effect on the headspace SPME procedure for the compounds studied. Thus, the method of external standard calibration would lead to an inaccurate quantification in this case. The problem of matrix effects on the reliability of headspace SPME quantification can be reduced by using a standard addition calibration method or isotopically labeled internal standards. In this study, the method of standard addition calibration was used to evaluate the linearity of the proposed headspace SPME method and to quantify the real sample.A series of river water samples spiked with seven different concentrations of the analytes studied was analyzed by the headspace SPME procedure described above. The spiking levels were in the range 2–300 ng l21 for MIB and geosmin, 4–600 ng l21 for b-ionone and 20–3000 ng l21 for the other compounds studied. For each level, three or four replicates were performed.Table 3 shows the linear ranges, slopes, Table 3 Linearity range, slopes, correlation coefficients (R2), quantification ions and limits of detection (LODs) for the analysis of taste- and odor-causing compounds in river water with headspace SPME-GC-ITDMSa No. Compound Linear range/ng l21 Slope area/ counts ng 121 R2 Quantification ionsb LOD/ ng l21 1 Hexanal 50–3000 55.22 0.989 T 50 2 Heptanal 20–3000 204.2 0.975 T 18 3 a–Pinene 20–3000 134.4 0.973 93 1.0 4 2–Methyl–3–heptanone 20–3000 43.58 0.997 128 3.0 5 (+)–Camphene 20–3000 88.86 0.964 93 1.3 6 Benzaldehyde 50–3000 11.07 0.985 77 50 7 b–Pinene 20–3000 120.6 0.981 93 0.9 8 6–Methyl–5–hepten–2–one 20–3000 40.66 0.991 108 3.5 9 2–Carene 20–3000 99.26 0.985 121 2.5 10 Octanal 20–3000 713.6 0.992 T 20 11 3–Carene 20–3000 286.6 0.985 93 1.5 12 a–Terpinene 20–3000 243.1 0.999 121 1.4 13 (+)–Limonene 20–3000 25.86 0.991 67 9 14 g–Terpinene 20–3000 220.2 0.998 93 0.9 15 Dihydromyrcenol 20–3000 268.9 0.997 59 1.2 16 (2)–Fenchone 20–3000 186.9 0.991 81 1.0 17 (±)–Linalool 20–3000 74.69 0.997 71 4.0 18 Nonanal 20–3000 1086 0.991 T 8.0 19 2,6–Dimethyl–2,4,6–octatriene 20–3000 210.7 0.998 121 1.5 20 (+)–Camphor 20–3000 49.14 0.990 95 4.5 21 (R)–(+)–Citronella 20–3000 85.64 0.997 95 13 22 (2)–Borneol 20–3000 115.7 0.999 95 3.2 23 (2)–Menthol 20–3000 71.16 1.000 81 3.0 24 MIB 2–300 299.2 0.995 95 0.7 25 a–Terpineol 20–3000 23.06 0.999 59 8 26 Decanal 20–3000 1288 0.987 T 8 27 b–Cyclocitral 20–3000 121.6 0.997 137 2.0 28 (+)–Pulegone 20–3000 97.36 1.000 81 6.0 29 (±)–Citral 50–3000 145.3 0.995 69 25 30 (2)–Bornyl acetate 20–3000 452.1 0.995 95 0.8 31 2,4–Decadienal 50–3000 264.8 0.997 81 20 32 Geosmin 2–300 487.9 0.999 112 0.5 33 Geranylacetone 20–3000 261.9 0.997 69 1.4 34 b–Ionone 4–600 192.9 0.999 177 1.5 a Water sample volume was 40 ml, containing 30% of NaCl; sample vial volume was 62 ml; sampling time was 40 min with stirring; seven plots with different concentrations (2–300 ng l–1 for MIB and geosmin, 4–600 ng l–1 for b–ionone and 20–3000 ng l–1 for the other compounds) were used.b T, total ion used for quantification. 464 Analyst, 1999, 124, 459–466correlation coefficients (R2) and limits of detection (LODs). For most of the compounds studied, the resulting calibration curves, obtained by plotting the GC–ITDMS response (area counts) vs. analyte concentration, were found to have good linearity in the tested concentration range, with R2 values ranging between 0.983 and 1.000.The LODs were calculated by comparing the signal-to-noise ratio (S/N) obtained by extraction of a river water sample with the lowest spiking level (2 ng l21 for MIB and geosmin, 4 ng l21 for b-ionone and 20 ng l21 for the other compounds) to S/N = 5. The LODs for MIB and geosmin were 0.7 and 0.5 ng l21, respectively. For the other compounds studied, the LODs were between 0.8 and 50 ng l21. These LODs were achieved using only 40 ml of water sample with 40 min of extraction time and are comparable to those obtained by methods such as CLSA-GC-MS (2 l of water sample and 2 h of extraction time),9 HFSA-GC-MS (3.8 l of water sample and 2 h Fig. 7 Typical GC–ITDMS chromatograms obtained by headspace SPME for spiked samples of bidistilled water (A) and river water (B) and non-spiked river water sample (C). Spiking level was 20 ng l–1 for MIB and geosmin, 40 ng l–1 for b-ionone and 200 ng l-1 for the other compounds studied.For peak numbers, see Table 3. Analyst, 1999, 124, 459–466 465of extraction time),14 and LLME-GC-ITDMS (1 l of water sample and > 1 h of extraction and concentration time).15 Fig. 7 shows the ITD chromatograms obtained after extraction of spiked samples of bidistilled water (A) and river water (B) and non-spiked river water samples by the proposed headspace SPME procedure. The chromatograms shown in Fig. 7 indicate that the GC resolution and peak shapes are perfectly acceptable, and the chromatogram of the spiked river water sample shows minimal background interferences when compared to that of the spiked bidistilled water sample.As shown in Fig. 7, in the non-spiked river water sample, compounds including 6-methyl-5-hepten-2-one, nonanal, borneol, MIB, decanal, 2,4-decadienal, geosmin and geranylacetone were detected. Table 4 shows the results obtained by triplicate analysis. The concentrations of MIB and geosmin determined were 5.9 ± 0.8 and 4.1 ± 0.6 ng l21, respectively. In addition, the same river water samples were also analyzed using an LLME method as described in the experimental section, and the results are also shown in Table 4.The data in Table 4 show that the concentrations obtained with headspace SPME were comparable to those obtained by LLME. Finally, to check the uniformity of response of different fibers, four fibers (one of which had been used more than 200 times) from two lots were compared.The extraction efficiency and RSD were found to be similar. Conclusions A method for the determination of trace levels of 34 taste and odor-causing compounds belonging to four major classes has been developed. By using a PDMS-DVB-coated fiber, the headspace SPME method, in conjunction with GC-ITDMS analysis, reveals a high degree of precision, good linearity over a wide range of concentration and high sensitivity. Using only 40 ml of water sample, detection limits obtained in river water are in the low ng l21 range for all the compounds examined in this study. Compared to other methods currently in use for the determination of taste- and odor-causing compounds present in trace levels in water, this method offers a number of practical advantages: smaller sample volume, shorter extraction time, simplicity of extraction and low cost.References 1 M. J. McGuire, Wat. Sci. Technol., 1995, 31, 1. 2 F. Jüttner, Wat. Sci. Technol., 1983, 15, 247. 3 J. Mallevialle and I. H. Suffet, Identification and Treatment of Tastes and Odors in Drinking Water, Cooperative Research Report of the AWWA Research Foundation and Lyonnaise des Eaux, Denver, USA, 1987. 4 S. L. Kenefick, S. E. Hrudey, E. E. Prepas, N. Motkosky and H. G. Peterson, Wat. Sci. Technol., 1992, 25, 147. 5 S. W. Krasner, M. J. McGuire and V. B. Ferguson, J. AWWA, 1985, 77, 34. 6 J. E. Amoore, J. AWWA, 1986, 78, 70. 7 P.-E. Persson, Wat. Res., 1980, 14, 1113. 8 K. Grob, J. Chromatogr., 1973, 84, 225. 9 S. W. Krasner, C. J. Hwang and M. J. McGuire, Wat. Sci. Technol., 1983, 15, 127. 10 R. Sävenhed, H. Borén, A. Grimvall and A. Tjeder, Wat. Sci. Technol., 1983, 15, 139. 11 P. B. Johnsen and J.-C. W. Kuan, J. Chromatogr., 1987, 409, 337. 12 V. C. Blok, G. P. Slater and E. M. Giblin, Wat. Sci. Technol., 1983, 15, 149. 13 E. D. Conte, S. C. Conway, D. W. Miller and P. W. Perschbacher, Wat. Res., 1996, 30, 2125. 14 A. K. Zander and P.Pingert, Wat. Res., 1997, 31, 301. 15 M. L. Bao, K. Barbieri, D. Burrini, O. Griffini and F. Pantani, Wat. Res., 1997, 31, 1719. 16 C. L. Arthur and J. Pawliszyn, Anal. Chem., 1990, 62, 2145. 17 D. Louch, S. Motlagh and J. Pawliszyn, Anal. Chem., 1992, 64, 1187. 18 C. L. Arthur, K. Pratt, S. Motlagh, J. Pawliszyn and R. P. Belardi, J. High Resolut. Chromatogr., 1992, 15, 741. 19 K. D. Buchholz and J. Pawliszyn, Environ. Sci. Technol., 1993, 27, 2844. 20 D. W. Potter and J. Pawliszyn, Environ. Sci. Technol., 1994, 28, 298. 21 T. K. Choudhury, K. O. Gerhardt and T. P. Mawhinney, Environ. Sci. Technol., 1996, 30, 3259. 22 A. A. Boyd-Boland, S. Magdic and J. Pawliszyn, Analyst, 1996, 121, 929. 23 L. Pan and J. Pawliszyn, Anal. Chem., 1997, 69, 196. 24 M. Chai, C. L. Arthur, J. Pawliszyn, R. P. Belardi and K. F. Pratt, Analyst, 1993, 118, 1501. 25 P. Martos and J. Pawliszyn, Anal. Chem., 1997, 69, 206. 26 A. Fromberg, T. Nilsson, B. R. Larsen, L. Montanarella, S. Facchetti and J. O. Madsen, J. Chromatogr. A, 1996, 746, 71. 27 K. J. James and M. A. Stack, J. High. Resolut. Chromatogr., 1996, 19, 515. 28 X. Yang and T. Pepard, J. Agric. Food Chem., 1994, 42, 1925. 29 A. Steffen and J. Pawlizyn, J. Agric. Food Chem., 1996, 44, 2187. 30 J. Song, B. D. Gardner, J. F. Holland and R. M. Beaudry, J. Agric. Food Chem., 1997, 45, 1801. 31 Z. Zhang, M. J. Yang and J. Pawliszyn, Anal. Chem., 1994, 66, 847. 32 V. P. Lee, T. Kumazawa, K. Sato and O. Suzuki, Chromatographia, 1996, 42, 135. 33 H. L. Lord and J. Pawliszyn, Anal. Chem., 1997, 69, 3899. 34 T. Górecki and J. Pawliszyn, Analyst, 1997, 122, 1079. Paper 8/07714B Table 4 Taste- and odor-causing compounds determined in river water by headspace SPME-GC-ITDMS and LLME-GC-ITDMS No. Compound Headspace SPME/ng l21 LLME/ ng l21 8 6-Methyl-5-hepten-2-one 34 ± 5 39 ± 6 18 Nonanal 58 ± 7 67 ± 8 22 Borneol 38 ± 4 33 ± 6 24 MIB 5.9 ± 0.8 4.9 ± 0.7 26 Decanal 92 ± 11 121 ± 14 31 2,4-Decadienal 37 ± 6 46 ± 6 32 Geosmin 4.1 ± 0.6 5.4 ± 0.8 33 Geranylacetone 51 ± 7 62 ± 7 Mean ± s (n = 3). 466 Analyst, 1999, 124, 459–466

ISSN:0003-2654    

DOI:10.1039/a807714b

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Methodology for assessing the properties of molecular imprinted polymers for solid phase extraction Jørgen Olsen,a† Paul Martin,*a Ian D. Wilsona and Graeme R. Jonesb a Department of Safety of Medicines, Zeneca Pharmaceuticals, Mereside, Alderley Park, Macclesfield, Cheshire, UK SK10 4TG b Department of Chemistry, Keele University, Staffordshire, UK ST5 5BG Received 4th January 1999, Accepted 1st March 1999 Four molecular imprinted polymers (MIPs) were prepared using propranolol as template molecule.The retention and recovery of propranolol and eleven structurally related compounds were assessed on all four MIPs and a blank polymer in solid phase extraction (SPE) experiments using methanol–water–triethylamine (TEA) (1%) elution solvents. Cumulative elution curves were produced and these were compared between blank and imprinted polymer for each compound. Only one polymer demonstrated significantly different retention properties compared to the blank indicating that only this polymer was sufficiently selective to be considered for SPE.In the case of the selective MIP, two classes of compounds were observed, with one group demonstrating selective retention whilst the second group showed similar retention on both MIP and blank. The compounds showing selective retention on the MIP were closely structurally related to the template propranolol, differing in having methoxy substituents on the naphthyl or a modified alkyl side chain whilst those that were equally well retained on the blank and MIP all contained amide groups close to the secondary amino group on the side chain.The use of eleven structurally related compounds in this way enabled the key structural features important to specific binding to the molecular imprints to be explored. They also provide a means to rapidly evaluate the polymers in order to select those suitable for SPE from aqueous solutions. Introduction Molecular imprinting is an attractive approach in areas where selective recognition of compounds is required since molecular imprinted polymers (MIPs) offer recognition sites with the potential to specifically bind molecules upon which the material is templated.The synthesis of MIPs involves the assembly of monomers around a template molecule followed by polymerisation in the presence of a cross-linker. Subsequent removal of the template molecule provides a material with ‘imprints’, or cavities, allowing specific rebinding of the template molecule.The binding affinities for the interaction between compound and polymer have been compared with antibody-like recognition. However, compared to antibodies, MIPs offer several advantages in being stable at high temperatures and in organic solvents. In addition, MIPs may be prepared rapidly in contrast to the long development times needed for antibody production in animals. However it should be appreciated that successful sample preparation has been achieved using immuno-affinity based SPE using antibodies.1 To date, MIPs have been employed as ‘artificial’ antibodies in ‘immunoassay’2 and in liquid chromatography as chiral stationary phases.3 Recently, MIPs have been investigated as highly selective sorbents for solid phase extraction (SPE) in order to concentrate and clean up samples prior to analysis. Currently the materials used routinely in SPE with aqueous samples are usually based on mechanisms employing lipophilic or ionic interactions with little discrimination between similar structures.MIPs, however, offer the possibility of achieving selective extractions, similar to those achieved by immunobased extraction systems, and thus may represent an advance on conventional SPE materials. Recent developments in the use of MIPs for SPE have been reviewed.4 However, despite the obvious potential of this approach there are problems associated with the use of MIPs for SPE which remain to be solved before the widespread use of the technique can be contemplated.For example, difficulties in achieving complete removal of template molecule have been observed5,6 and the leaching of template from the sorbent during the extraction can subsequently interfere with the analysis.7 In addition, the polymers themselves can extract compounds from the sample matrix via a range of non-specific interactions in addition to the desired ‘template’-mediated specific extraction.In the case of this latter problem methods are required that enable the rapid characterisation of a MIP to distinguish between specific and non-specific binding. Only when this is established will it be possible to know whether the retention due to the imprinting process offers a real advantage over methods based on non-specific binding. In addition, where the extraction is due to imprinting a means of determining the extent of cross-reactivity of similar molecules with the template may be an important consideration.Current approaches to assess MIPs for use in SPE compare extraction on the MIP to that on a reference polymer which may either be non-imprinted5,6 or made using a different template molecule.7 These comparisons have been done by assessing the two polymers under the same conditions as SPE sorbents5,6 or by employing different techniques (HPLC).8,9 Various measures to estimate the differences between the MIP and reference polymer have been used including superior retention on the MIP compared to the reference polymer.6–9 The superior retention was estimated as capacity factors in HPLC or in SPE by the amount of solute retained at the application6,10 or wash step.6 Furthermore, selectivity and enrichment factors have been applied to the SPE-based evaluation of the sorbents.8 In addition, a more indirect approach has been described where superior retention of the template molecule to structurally † Current address: The Royal Danish School of Pharmacy, Department of Analytical and Pharmaceutical Chemistry, Universitetsparken 2, DK-2100 Copenhagen, Denmark.Analyst, 1999, 124, 467–471 467related compounds was used as evidence of the existence of specific recognition.11 The aim of the work reported here was to develop a rapid and simple means for the assessment of the suitability of propranolol- derived molecular imprints for SPE from aqueous solutions. The methodology was developed in order to be able to distinguish between non-specific binding and that due to the imprinting process with the intention that it could be used as a quality control measure for MIP production.Experimental Results Racemic propranolol used as the template was supplied by Sigma (Dorset, UK). The propranolol analogues were supplied by Zeneca Pharmaceuticals (see structures in Fig. 1) and included M51932, M52487, M47070, M115716, M109056, M81509, M49666, M109055, M65318, M45655 and M87086.Radiolabelled 14C-propranolol was prepared in Zeneca Pharmaceutical’s isotope chemistry laboratory. HPLC grade methanol, ammonium acetate (AR grade), analytical grade trifluoroacetic acid (TFA) and triethylamine (TEA) were purchased from Fisons (Loughborough, UK). The water was de-ionised. Materials for SPE Four polymers were prepared by the method described by Andersson2 using propranolol as the template molecule. Methacrylic acid (Aldrich) was used as functional monomer and ethylene glycol dimethacrylate (EGDMA) (Aldrich, for polymers 2, 3, 4) or butanediol dimethacrylate (polymer 1) as crosslinker.Toluene was used as the porogenic solvent. The composition of the MIPs is presented in Table 1. Polymers 2 and 4, though nominally the same, were synthesised in two different laboratories. A blank reference polymer containing no template Fig. 1 Structures of the compounds used in the study. 468 Analyst, 1999, 124, 467–471was prepared with the same monomer to cross-linker ratio as polymer nos. 2 and 4. The bulk polymers were ground in a mortar and sieved through a 40 mm sieve. Fines were removed by repeated sedimentation in ethanol. Finally, the particles were washed using the solvents used by Andersson.2 After drying in a vacuum oven (60 °C), 30 mg of the powdered material were packed into 1 ml cartridges with frits to secure the phase in place. Cartridges and frits were supplied by IST (Hengoed, UK).Methods Solid phase extraction. The cartridges were conditioned with 3 3 1 ml of methanol–TFA (100 : 1, v/v) to remove potential interfering template and to wet the phase, 2 3 1 ml methanol to remove TFA and 1 ml of water to condition the phase prior to the application of aqueous samples. After conditioning the MIP, 0.5 ml of the sample for SPE was applied to the cartridge. The samples consisted of a mixture of M51932, M52487, M47070, M115716, M109056 (batch A) or M81509, M49666, M109055, M65318, M45644, M87086 (batch B), respectively. The concentration of each analyte in solution was 20 mg ml21.For elution, aliquots (0.5 ml) of 0–100% methanol in water solutions containing TEA (1%) were used in 10% increments of methanol to recover the compounds. Finally, the MIP was washed three times with 1 ml of a solution consisting of methanol–TFA (100 : 1, v/v) to secure full recovery. The passage of solvents through the cartridge in each of the steps was achieved by vacuum of 15 psi (‘Bakerbox’ SPE device, J.T. Baker, Philipsburgh, NJ, USA). HPLC Analysis. Prior to HPLC analysis, the eluents from the SPE cartridges were blown to dryness under nitrogen (Turbo Vap LV Evaporator, Zymark) and re-dissolved in 0.3 ml of mobile phase. For batch A compounds the mobile phase consisted of methanol–water–TFA–ammonium acetate (600 : 400 : 1 : 7.7, v/v/v/w). For batch B the mobile phase was methanol–water–TFA–ammonium acetate (550 : 450 : 1 : 7.7, v/v/v/w).HPLC was performed using a 25 cm 3 4.6 mm id HiRBP column (Hichrom) with a flow rate of 1 ml min21. The pump was an LDC Analytical Constametric 3200 (Stoke, Staffs, UK). The detection was UV at 254 nm (Perkin Elmer, UV–Vis, LC 290) and the injection volume was 40 ml and was achieved using a Perkin Elmer, ISS 200 auto sampler. Data collection and analysis was performed by Beckman Peakpro. Cumulative elution curves for propranolol on all polymers were obtained using radiolabelled propranolol according to the procedure described by Martin et al.11 The elution curves for propranolol had to be performed using radiolabelled compound because leaching of the template precluded chromatography.Application and elution solvents were collected in plastic scintillation vials (Packard, Downers Grove, IL, USA) to which were added 10 ml of Ready Value scintillation cocktail (Beckman, Fullerton, CA, USA) and counted on a Beckman LS 1801 scintillation counter.This approach was used to overcome template leaching during analysis which might interfere with subsequent quantification. Indeed, it has been shown in this laboratory that only 80% of the template could readily be removed from the MIP whereas the last 20% leached slowly from the polymer.7 Results and discussion Assessment of the four polymers in SPE Cumulative elution curves were used to evaluate elution from the MIPs as they have previously been used successfully to describe elution characteristics in SPE.12 In this study the amount of organic content was varied from 0–100% (in 10% steps) with a fixed amount of ionic modifier (1% TEA) as previous work showed this combination to provide the greatest selectivity.11 Elution curves provide an easy means to view the total extraction performance of an SPE phase, but obviously they represent the combined contributions of both specific and nonspecific interactions.However, comparison of the results obtained with an imprinted polymer and one prepared in the absence of the template molecule can provide some insight into the contribution of specific and non-specific mechanisms.This is illustrated in Fig. 2 for M52487 on all four polymers and the blank. As this result shows, only polymer 2 appears to show any selectivity relative to the blank. However, whilst such elution curves are useful when trying to distinguish between specific and non-specific interactions for a few compounds they become very cumbersome when faced with larger numbers simply due to the large volume of data generated by such studies.In order to evaluate all four MIPs for specific ‘imprint-based’ interactions we examined their performance as measured by the solvent composition required to give a recovery of either 20 or 50% of the extracted material (ES20 or ES50 values). This gives a single figure which can be readily compared with that of the blank polymer.In this case the ES20 values proved to be more useful than the ES50 as with polymer 2 the recovery of 50% of analytes with methanol–TEA-based eluents was not achieved in all cases due to strong specific interactions. The ES20 values obtained for each MIP were then plotted against those obtained for the blank polymer thus giving a measure of the relative retention on the MIP to the molecular imprinting process. This comparison of retention on the MIPs and blank polymer is illustrated in Fig. 3A–D for polymers 1 to 4, respectively. The important feature to note in Fig. 3A, 3C and 3D is that for polymers 1, 3 and 4 the ES20 values for all of the compounds were similar to those observed for the blank reference polymer. Table 1 Composition of the four polymers assessed Polymer Monomer Cross-linker Ratio of monomer to template 1 Methacrylic acid Butanediol dimethacrylate 2 : 1 2 Methacrylic acid EGDMA 2 : 1 3 Methacrylic acid EGDMA 8 : 1 4 Methacrylic acid EGDMA 2 : 1 Fig. 2 Cumulative elution curves for M52487 on MIPs and blank polymer. Key: polymer 1 –––/–––; polymer 2 –––-–––; polymer 3 –––:–––; polymer –––3–––; blank ~~~5~~~ . Analyst, 1999, 124, 467–471 469This result indicated that under the test conditions employed here there was no apparent difference in the strength of retention between MIP and blank for all analytes. These findings demonstrate that either these polymers were not successfully imprinted or that interactions due to molecular imprinting were minor compared to the non-specific effects with the result that imprinting did not affect the position of the curves.In contrast, polymer 2 showed marked differences in the ES20 data for some of the analytes compared to the blank polymer (Fig. 3B). Thus within the test compounds there appeared to be two classes of compounds. One group of compounds showed essentially little or no difference in the ES20 values between MIP and blank, whilst the second class were much more strongly retained on the MIP.In addition to the graphical representation shown in Fig. 3A–D the data can also be represented as a ratio of the ES20 values for the blank and imprinted polymer as an ES ‘index’ (ES20 blank : ES20 MIP). The ES index results for the test compounds on polymer 2 are presented in Table 2 with high values (approximately 1) indicative of similar retention on MIP and blank and low values (approximately 0.6–0.8) indicative of better retention on the MIP compared to the blank.Comparing the MIP selectivity (ES20 index) to the structure of the compounds it was possible to comment on the structural features that reduced the specific binding compared to the template molecule propranolol. The best retained compound was M52487 which differs from propranolol in that it possesses a methoxy group on the naphthyl ring. Similarly, M49666 differs only in having a methoxy group in a different position on the naphthyl ring and again this compound is only marginally less well retained than propranolol.These two results indicate that small changes in the naphthyl ring do not reduce the strength of binding. Other selectively retained compounds included M45655 (minor change to the alkyl side chain) and M47070 (increased distance between naphthyl ring and hydroxy) indicating that minor changes in these regions only reduce specific binding marginally. Five of the compounds (M87086, M115716, M109056, M109055 and M81509) contained amide groups close to the amino-alcohol and these were the worst retained.It is conceivable that the amide (which is a strong hydrogen bonding group) disrupted the binding of the basic secondary amino group on the b-blocker side chain to the acid functions of the polymer. Given the probable importance of this structural feature for interacting with the MIP binding site, the reduction in the strength of binding of the amide-containing compounds relative to propranolol is readily explicable.The two remaining compounds (M51932 and M65318) appeared to show greater selective binding than the amides but were less well retained than those compounds with only minor modifications to the basic propranolol structure. M51932 possessed 2 methoxy groups on the naphthyl ring and this increased bulk evidently reduced the strength of selective binding. The compound M65318 did not contain a hydroxy group on the side chain and there is a shorter distance between the naphthyl and the secondary amine.Assuming that there are binding sites for both the lipophilic naphthyl and the amine, reducing the distance between the two groups may disrupt the binding of one/both groups in their respective binding sites. The poor performance of polymers 1, 3 and 4 under the extraction and elution conditions used in this evaluation showed that imprint binding did not play a significant role in the retention of the test compounds.As such, these polymers would be unsuitable for selective extraction of propranolol-like Fig. 3 Plot of ES20 values on the MIP compared to the blank polymer for polymers 1 (A), 2 (B), 3 (C), 4 (D). The line has the slope 1 illustrating the same ES20 for the polymer and the reference polymer. Key: M49666 +; M45655 /; M115716 3; M51932 8; M109055 -; M87086 ½; M109056 3| ; M52487 -; M65318 --; propranolol 2; M81509 5; M47070 ¦. Table 2 ES20 index values of all compounds on polymer 2 Compound ES20 index (ES20 blank : ES20 polymer 2) M51932 0.76 M52487 0.66 M47070 0.72 M115716 0.87 M109056 0.88 M81509 1.09 M49666 0.71 M109055 0.82 M65318 0.77 M45655 0.69 M87086 0.84 Propranolol 0.67 470 Analyst, 1999, 124, 467–471compounds from aqueous solution.Despite these results polymers 1, 3 and 4 may still contain imprints but the experimental conditions may not have been appropriate for selective binding. Thus, the best conditions for exploiting the imprinted recognition sites are often claimed to be achieved in the solvent in which the MIP was prepared (in this case toluene).Clearly different types of interaction between solute and polymer can be expected to dominate in non-polar organic media compared to highly polar aqueous solvents from which these model compounds were extracted. Nevertheless, it is clear that it is possible to obtain MIPs such as polymer 2 that can be exploited to extract anaytes directly from aqueous solution with specificity due to the molecular imprinting process.Given that sample preparation is applied to aqueous samples it is encouraging that MIPs show specificity in aqueous-based systems. However, the mechanism of extraction from aqueous solution onto these MIPs is a combination of a number of different processes, and even when imprinting is providing a significant contribution, as in the case of polymer 2, a high degree of nonspecific binding is involved in the retention of the compounds (probably including hydrophobic reversed-phased type interactions and ionic interactions of the secondary amine with the free carboxyl groups on the bulk polymer).By choosing 4 polymers to the same template molecule but made in different laboratories or with slightly different compositions, we attempted to generate phase diversity. As the results shown in Fig. 3A–D demonstrate, the evaluation procedure described for the polymers rapidly enabled the identification of MIPs with suitable extraction/elution characteristics for use with aqueous samples.Having identified polymer 2 on the basis of such a test scheme it should then be possible to devise methods for the selective extraction of the template molecule (or close structural analogues such as M47070) from aqueous matrix, followed by separation from unrelated contaminants extracted non-specifically together with related substances (e.g., M81509) that are not ‘recognised’ by the MIP.Reproducible manufacturing of MIPs with the required characteristics is obviously an issue that will need to be addressed before MIPs can become established as a viable SPE material. The sort of variation in properties shown above clearly demonstrates the need for suitable tests to evaluate different batches in order to assure the precision and accuracy of an assay. The use of cumulative elution curves and ES20 indexes may provide the basis for such quality control procedures.Conclusions A method for evaluating propranolol-derived MIPs was developed which indicated the relative importance of specific and non-specific binding that resulted in compound retention. Evaluation was based on estimation of ES20 indexes which quantified retention on a MIP relative to a blank. Only one of the four MIPs demonstrated markedly better retention compared to the blank which indicated that there was batch-to-batch variation in the performance of MIPs. By evaluating the retention of compounds structurally related to the template propranolol on the best MIP it was possible to determine which structural features contributed to imprint binding. Acknowledgements Steve Rimmer (University of Lancaster), Joe Irvine, Ken Jones (Affinity Chromatography), John Clewes (University of Keele) for providing polymers for the study. References 1 B. A. Rashid, P. Kwasowski and D. Stevenson, Pharm. Sci., 1996, 2, 115. 2 L. I. Andersson, Anal. Chem., 1996, 68, 111. 3 M. Kempe and K. Mosbach, J. Chromatogr. A, 1995, 691, 317. 4 J. Olsen, P. Martin and I. D. Wilson, Anal. Comm., 1998, 35, 13H. 5 L. I. Andersson, A. Paprica and T. Arvidsson, Chromatographia, 1997, 46, 57. 6 B. A. Rashid, R. J. Briggs, J. N. Hay and D. Stevenson, Anal. Commun., 1997, 34, 303. 7 P. Martin, I. D. Wilson, G. R. Jones and K. Jones, in Drug development assay approaches; including molecular imprinting and biomarkers, ed. E. Reid, H. H. Hill and I. D. Wilson, Royal Society of Chemistry, Cambridge, 1998, p. 21. 8 B. Sellergren, Anal. Chem., 1994, 66, 1578. 9 J. Matsui, M. Okada, M. Tsuruoka and T. Takeuchi, Anal. Commun., 1997, 34, 85. 10 M. T. Muldoon and L. H. Stanker, Anal. Chem., 1997, 69, 803. 11 P. Martin, I. D. Wilson, D. E. Morgan, G. R. Jones and K. Jones, Anal. Commun., 1997, 34, 45. 12 B. Law, S. Weir and N. A. Ward, J. Pharm. Biomed. Anal., 1992, 10, 167. Paper 9/00040B Analyst, 1999, 124, 467–471 471

ISSN:0003-2654    

DOI:10.1039/a900040b

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Determination of organochlorine pesticide residues in honey, applying solid phase extraction with RP-C18 material† D. Tsipi,a M. Triantafylloub and A. Hiskia*c a General Chemical State Laboratory, Pesticide Residues Laboratory, Division C, 16 An. Tsoha, 11521, Athens, Greece b Agricultural University of Athens, 75 Iera Str., 11855, Athens, Greece c Institute of Physical Chemistry, NCSR Demokritos,15310, Athens, Greece Received 14th December 1998, Accepted 23rd February 1999 In this study, a new clean up method was developed for the routine multiresidue determination of organochlorine pesticide residues in honey.The analytical procedure requires sample extraction with methanol, followed by a clean up step through a C18 Sep-Pak cartridge. Finally, pesticides are eluted with hexane. The determination of organochlorine pesticide residues was performed by capillary gas chromatography with electron capture detection. The mean recoveries of 18 organochlorine pesticides were estimated at various concentrations and found very efficient in most cases.The detection limits were found to be between 0.05 and 0.20 mg kg21. The occurrence of organochlorine compounds in the food chain has already been reported in several studies.1–4 This class of organic compounds consists of one of the most important groups of dangerous organic contaminants. Honey is an exported product of Greece with great economic importance. According to EEC regulations, honey as a natural product, must be free of any chemical contaminants and safe for human consumption.5 Many methods have been reported for the determination of pesticides in honey, used against the Varroa mite diseases (acaricides and organophosphorous pesticides) 6–11 or in agriculture for insect control on numerous field crops.12,13 However, only a few are concerned with organochlorine pesticides although their occurrence has been reported in several studies.14–17 These last methods, following the classical analytical procedures for the determination of pesticides in non-fatty foods, employ time consuming clean-up steps that make them impractical for routine analysis.It is therefore necessary for monitoring purposes to develop a specific and rapid method for the determination of organochlorine pesticide residues in this substrate. Fernandez Muino and Simal Lozano18 proposed a multiresidue method for determination of organochlorine pesticides in honey, which uses a Florisil clean up step for the isolation of pesticides followed by gas chromatography with electron capture detection (GC-ECD).Good recoveries of eight organochlorine pesticides were obtained together with a minimized matrix interference. However this method involves a complicated liquid–liquid extraction step in which there is a possibility of formation of a whitish gel which obscures the separation and gives recoveries below 60%. Furthermore, the proposed method was applied only to a small group of organochlorine pesticides. In the present study a quick and simple alternative method, by drastically reducing the liquid–liquid extraction step, for the determination of 18 organochlorine pesticides in honey, is presented.This method involves sample extraction with methanol, followed by solid phase extraction on C18 cartridges and elution with hexane. The target compounds studied, namely a-HCH, b-HCH, lindane, d-HCH, heptachlor, aldrin, heptachlor epoxide, a- endosulfan, 4,4-DDE, dieldrin, endrin, b-endosulfan, 4,4-DDD, endrin aldehyde, endosulfan sulfate, 4,4-DDT, methoxychlor and endrin ketone, were determined by capillary gas chromatography with electron capture detection (GC-ECD).Confirmation was achieved using two GC columns of different polarity. Experimental Materials The solvents used (methanol and hexane) were pesticide residue free (Pestiscan, Lab Scan, Dublin, Ireland). Water was the product of Reidel-de Haen (Pestanal), Seelze, Germany.a- HCH and endrin were obtained as solid materials from Reidelde Haen, with a purity of 98–99%. Lindane and aldrin were obtained as solid materials from Alltech, Chicago, IL, USA, with purities of 99%. The other pesticides were obtained from Polyscience, Niles, IL, USA, as solutions in methanol. Stock solutions of each pesticide were prepared in methanol at 1000 mg ml21. The mixture of the 18 organochlorine pesticides were purchased from Polyscience, as a solution of 2000 mg ml21 in methanol.Working solutions were prepared by diluting the stock solutions as required. Solid phase extraction was carried out using bonded-phase silica C18 0.85 ml filled cartridges, containing 360 mg of C18 octadecyl sorbent, Sep-Pak ‘classic’, product of Waters, Milford, MA, USA. Procedure Sample extraction and clean up. Honey (10 g) was dissolved in 50 ml of methanol and the mixture was stirred for an hour. Then 25 ml of the above solution, after filtration, was diluted in 2 l of distilled water, at pH 2.The mixture was passed through a C18 cartridge, which had previously been conditioned with 10 ml of methanol and then with 5 ml of water. The C18 cartridge was fitted to a glass column (25 cm 3 1 cm) which was connected to a 2 l flask reservoir. Head pressure was applied with extra pure nitrogen to increase the flow, to about 10 ml min21. After the sample volume had passed through the column, the cartridge was dried for 1 h under a stream of † Presented in part at the International Conference on Quality and Safety Aspects of Food and Nutrition in Europe ’95, Helsinki, Finland, August 22–25, 1995.Analyst, 1999, 124, 473–475 473nitrogen and the organochlorine pesticides were eluted with 10 ml of hexane. The extract was rotary evaporated (bath temperature, 30 °C) to Å 1 ml and the residue transferred quantitatively with hexane into a 5 ml volumetric flask for the GC-ECD analysis.Gas chromatographic analysis. The analysis of the 18 organochlorine pesticides was carried out by capillary gas chromatography using the following instruments: (1) Varian (Paolo Alto, CA, USA) Model 3400 gas chromatograph equipped with ECD, split/splitless injection port, a DB-1 fused silica capillary column by J&W Scientific, Rancho Cordova, CA, USA (30 m 3 0.32 mm id, 0.25 mm film thickness) and autosampler Model 8200 cx, with a program for the evaluation of GC runs (DAPA Scientific, Kalamunda, Australia); and (2) Carlo Erba (Milan, Italy) Model Mega 2 gas chromatograph equipped with ECD, split/splitless injection port, a DB-5 fused silica capillary column by J&W Scientific (30 m 3 0.25 mm id, 0.25 mm film thickness) and autosampler Model A200S, with a program for the evaluation of GC runs (Chrom-Card, Fisons Instruments, Rodano, Milan, Italy).The temperature program applied was as follows: 80 °C for 1 min, 80–218 °C at 8 °C min21, 218 °C for 18 min, 218–250 °C at 4 °C min21 and 250 °C for 10 min.The injection was carried out splitless at 250 °C and the injection volume was 1 ml. Standard solutions of each target compound were analysed under the mentioned conditions on DB-1 and DB-5 columns for the determination of their retention times. The linearity of the ECD system was tested by analysing standard solutions of the studied pesticides in the range 0.2 to 40 mg l21. Five point external standard calibration was used for the quantitative measurements.Recovery experiments and detection limits. Recovery experiments, concerning the 18 organochlorine pesticides, were carried out, in triplicate, at various fortification levels, by adding known volumes of pesticide standards in hexane, to homogenized honey samples. After solvent evaporation the samples were analysed according to the proposed method. The recovery values were calculated from calibration graphs that were constructed from the concentration and peak area of the chromatograms obtained with standards of the organochlorine pesticides.Blank analyses were performed in order to check interference from the sample. The detection and quantification limits of the target compounds were determined after spiking honey samples at lower concentration levels. Their values were calculated considering a signal-to-noise ratio of 3 or 10, respectively. Results Retention times (tR) of 18 organochlorine pesticides were determined individually on DB-1 and are presented in Table 1.The GC-ECD chromatogram of a honey sample, spiked to 0.4 mg l21 for each organochlorine pesticide is presented in Fig. 1. The resolution of lindane/b-HCH and 4,4 DDE/dieldrin pairs was poor under the conditions employed in this study. Therefore, the identification of peak identity was performed on DB-5 column (tR = 17.73/17.97 and 26.63/26.87, respectively). The matrix interference during analysis of honey samples in the GC-ECD system was limited.Gas chromatograms of spiked honey samples were quite similar to those obtained with the standard solution of pure pesticides. For that reason, the preparation of standard solutions in control sample extracts was not necessary. The gas chromatogram of honey extract, presented in Fig. 2, shows good baseline stability with a few interfering peaks, indicating that the proposed clean up is suitable for the determination of the target analytes.The detector response for all target compounds was linear in the concentration range 0.2 to 40 mg l21 and the correlation Table 1 Retention times (tR), detection limits, mean percentage recoveries and relative standard deviations (RSD) of 18 organochlorine pesticides at three different fortification levels in honey (n = 3) on a DB-1 column (GC-ECD) Mean percentage recovery (RSD) Organochlorine tR/ Detection limit/ No. pesticides min mg kg21 20 mg kg21 10 mg kg21 4 mg kg21 1 a-HCH 12.67 0.08 66 (6) 71 (4) 80 (12) 2 b-HCH 13.57 0.06a 72 (5) 84 (6) 83 (8) 3 Lindane 13.57 0.08a 65 (8) 93 (8) 86 (10) 4 d-HCH 14.22 0.08 120 (6) 85 (4) 107 (13) 5 Heptachlor 15.54 0.08 66 (3) 85 (10) 77 (3) 6 Aldrin 16.43 0.10 61 (5) 87 (7) 72 (2) 7 Heptachlor epoxide 17.44 0.09 75 (7) 82 (4) 81 (4) 8 a-Endosulfan 18.42 0.12 71 (4) 73 (1) 56 (6) 9 4,4-DDE 19.20 0.10a 80 (6) 77 (7) 79 (2) 10 Dieldrin 19.25 0.10a 83 (5) 103 (4) 80 (6) 11 Endrin 19.88 0.16 108 (3) 94 (3) 88 (5) 12 b-Endosulfan 20.18 0.05 72 (4) 74 (6) 87 (8) 13 4,4-DDD 20.70 0.16 116 (7) 101 (7) 77 (7) 14 Endrin aldehyde 20.87 0.20 80 (3) 104 (1) 48 (8) 15 Endosulfan sulfate 21.88 0.18 74 (2) 92 (4) 73 (4) 16 4,4-DDT 22.27 0.10 79 (4) 102 (2) 125 (2) 17 Methoxychlor 24.01 0.08 81 (3) 73 (3) 53 (3) 18 Endrin ketone 25.61 0.20 74 (2) 83 (4) 78 (4) a Refers to the DB-5 column Fig. 1 Gas chromatogram of a honey sample extract spiked at 0.4 mg kg21, on a DB-1 column with ECD. The numbers refer to pesticides, according to Table 1; a, b and c are interfering peaks. 474 Analyst, 1999, 124, 473–475coefficients were better than 0.999. The calculation of the amount of the organochlorine pesticides present was carried out using the DB-1 column. The results were confirmed with the DB-5 column. In case one or more of the pesticide pairs that were not resolved on the DB-1 column were present, identification was performed with the DB-5 column. Quantification was carried out by DB-5 only when both components of each pair were identified, otherwise it was based on DB-1.Recovery experiments, concerning the 18 organochlorine pesticides, were performed in honey samples, at three fortification levels of 4, 10 and 20 mg kg21. The results of a series of threefold experiments for each fortification level are presented in Table 1. The mean recoveries, at the three fortification levels, approach successful recovery in most cases.The mean recoveries of honey samples fortified at the 20 mg kg21 level were between 61 and 120%. The recoveries of the same pesticides at the 10 mg kg21 level, ranged from 71 to 104%. The recoveries at the lower fortification level (4 mg kg21) were between 72 and 125% except for endrin aldehyde, methoxychlor and a-endosulfan which was only recovered with 48, 53 and 56%, respectively. It seems that the recovery values were not related to the spiking level. The precision of the method expressed by the relative standard deviation (RSD) of the mean recovery values, when triplicate spiked honey samples were analysed, was better than 13%.The detection limits of the target compounds were in the range of 0.05 to 0.20 mg kg21 and are shown in Table 1. The corresponding quantification limits, always 3.3 times the detection limits, varied between 0.16 and 0.66 mg kg21 and were approximately four times lower than those reported in the literature.18 Conclusions In this paper, a routine multiresidue method, for the determination of the 18 most important organochlorine pesticide residues in honey, is reported.This method applying solid phase extraction, followed by gas capillary chromatography with electron capture detection, is effective for the analysis of the target analytes and at the same time is quick and of low cost. Solid phase extraction with RP-C18 material, without a further clean up step, yields high recovery rates for almost all compounds investigated.The main advantages of the method described, compared to previously reported analytical procedures, are: (a) sample treatment is easier and faster; and (b) a larger number of organochlorine pesticides can be simultaneously determined. Furthermore, with the analytical method presented, trace level determination of organochlorine pesticides at sub-ppb levels is possible and gives reliable results. The lack of interferences due to the complex matrix, the high recovery values, and the sensitivity of this method offer a valuable tool for the determination of organochlorine pesticides in honey samples.References 1 H. G. Gorchev and C. F. Jelinek, Bull. W. H. O., 1985, 63, 945. 2 R. Moilanen, J. Kumpulainen and H. Pyysalo, Ann. Agric. Fenn., 1986, 25, 177. 3 K. Kannan, S. Tanabe, A. Ramesh, A. Subramanian and R. Tatsukawa, J. Agric. Food Chem., 1992, 40, 518. 4 V. Hietaniemi and J. Kumpulainen, Food Addit. Contam., 1994, 11, 685. 5 EEC Directive, 74/409, 1974. 6 G. Formica, J. Assoc. Off. Anal. Chem., 1984, 67, 896. 7 W. Ebing, Fresenius’ Z. Anal. Chem., 1987, 327, 539. 8 O. Stricker, K. Clerschner and G. Vorwohl, Dtsch. Lebensm.- Rundsch., 1989, 85, 72. 9 A. T. Thrasyvoulou and N. Pappas, J. Apiary Res., 1988, 27, 55. 10 M. Barbina Taccheo, M. De Paoli and C. Spessoto, Pestic. Sci., 1988, 23, 59. 11 M. Barbina Taccheo, M. De Paoli and C. Spessoto, Pestic. Sci., 1989, 25, 11. 12 M. A. Carcia, M. I. Fernandez and M. J. Melgar, Bull. Environ. Contam. Toxicol., 1995, 34, 825. 13 D. Tsipi, A. Hiskia and M. Triantafyllou, in Natural Antioxidants and Food Quality in Atherosclerosis and Cancer Prevention, ed. J. Kumpulainen and J. Salonen, Royal Society of Chemistry, Cambridge, 1996, pp. 439–444. 14 J. Ogata and A. Bevenue, Bull. Environ. Contam. Toxicol., 1973, 9, 143. 15 C. B. Estep, N. Menon, H. E. Williams and A. C. Cole, Bull. Environ. Contam. Toxicol., 1977, 17, 168. 16 J. Serra Bonvehi, Alimentaria, 1985, 166, 55. 17 G. R. Trevisani, F. Michelini and M. Baldi, Boll. Chim. Unione Ital. Lab. Prov., Parte Sci., 1982, 33, 69. 18 A. M. Fernandez Muino and J. Simal Lozano, Analyst, 1991, 116, 269. Paper 8/09724K Fig. 2 Gas chromatogram of an unspiked honey sample extract, on a DB-1 column with ECD; a, b and c are interfering peaks. Analyst, 1999, 124, 473–475 475

ISSN:0003-2654    

DOI:10.1039/a809724k

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Urine polyamines determination using dansyl chloride derivatization in solid-phase extraction cartridges and HPLC C. Molins-Legua, P. Campíns-Falcó,* A. Sevillano-Cabeza and M. Pedrón-Pons Departament de Química Analítica, Facultad de Química, Universitat de Valencia, C/ Dr. Moliner 50, 46100- Burjassot, Valencia, Spain Received 9th November 1998, Accepted 9th February 1999 The derivatization of biogenic amines such as putrescine, cadaverine, spermidine and spermine with dansyl chloride in solid phase extraction cartridges is described. Different types of filling materials were tested in order to have the highest retention of the different analytes.The best results were obtained by using C18 cartridges. The optimal conditions were: amine solution buffered at pH 12, 2 mM dansyl chloride (acetone–bicarbonate solution 20 mM (pH 9–9.5), 2 + 3 v/v) as reagent concentration, room temperature and 30 min reaction time. The developed procedure was applied to the determination of these polyamines in urine samples from healthy controls and cancer patients using HPLC with 1,7-diaminoheptane as internal standard.The concentrations ranged from 0.5 to 5 mg mL21 and the detection limits were 10 ng mL21 for all polyamines. By concentrating the urine extracts, the detection limits were improved down to 2 ng mL21. The accuracy and the precision of the method were tested. The proposed dansylation method is advantageous with respect to solution dansylation.It improves the total analysis time, avoids high temperatures that can affect the thermal stability of the derivatives and could make possible the automation of the procedure. Introduction Polyamines (such as putrescine, cadaverine, spermine or spermidine) are essential for normal growth and cellular differentiation.1–3 They are commonly present in significant amounts both in prokaryotic and eukaryotic cells. Malignant cell proliferation is associated with increased cellular polyamine metabolism.4 Several diagnostic assays based on polyamine detection have been developed to screen for cancer, to evaluate efficacy of therapy and to detect relapse.5 Researchers have demonstrated that tumour cells contain a much higher concentration of polyamines, and patients with many types of cancers have probably an enhanced urinary excretion of polyamines.Several methods have been described for their determination in biological samples, high performance liquid chromatography (HPLC) being the one preferred.The direct detection of polyamines is difficult because they do not absorb in the ultraviolet region and, consequently, they do not have native fluorescence, required in most of these procedures of derivatization. However, most of them require an extraction procedure to remove interfering derivatives or reagent excess. These features prevent automatization of the derivatization procedure and decrease reproducibility. Pre-column or post-column derivatization, followed by spectrophotometric and fluorimetric detection, are, therefore, commonly applied.Fluorimetric reagents such as dansyl chloride,6 fluorescamine,7 o-phthaldialdehyde( OPA)–2-mercaptoethanol8,9 or o-phthaldialdehyde– ethanethiol10 have been proposed for determination of polyamines. A number of pre-column derivatization techniques have been developed for HPLC with UV/VIS spectrophotometric detection.11 They include reactions with benzoyl chloride, 12 p-toluenesulfonic chloride (tosyl chloride), 2,4-dinitrofluorobenzene, 4-fluoro-3-nitrobenzotrifluoride, quinoline- 8-sulfonyl chloride and 4-dimethylaminoazobenzene-4A-sulfonyl chloride (dansyl chloride).The detection limits reached with the fluorimetric reagents are generally better than those obtained with the spectrophotometric reagents. The most commonly employed reagents for converting polyamines into fluorescent products are dansyl chloride and OPA–thiol.The stability of the dansyl derivatives is better than that achieved with the isoindoles formed with the OPA–thiol reagents. One advantage of using dansyl derivatives is the possibility of forming chemiluminiscence with oxalic acid bis(2,4,6-trichlorophenyl ester)–H2O2, which can increase sensitivity and selectivity. During the last five years, this research group has been studying derivatization procedures off- or on-line on solid phase supports. A commercial Si–OH-modified packing material is used instead of polymeric reagent specially prepared for solidphase derivatization. The cartridges are used to purify the sample and concentrate the analytes and, next, the derivatization agent is passed through and the derivatized analytes are retained.Some applications, based on this methodology, have been described for amphetamine and related compound determination in urine samples, by using different reagents such 9-fluorenylmethyl chloroformate (FMOC), OPA, 1,2-napthoquinone 4-sulfonate (NQS)13–15 or dansyl chloride (Dns- Cl).16 This paper extends our methodology to the analysis of biogenic polyamines.A simple and rather quick off-line HPLC procedure for both clean up and derivatization of polyamines on solid phase extraction is proposed. The optimal conditions have been established and the final procedure has been applied to the determination of these amines in urine samples. Since the determination of polyamines is of great interest in cancer urinary profiles (due to the concentration levels being not completely clear) or in metabolic studies, the potential of the described approach has been tested by analysing these biogenic amines in the urine of healthy volunteers or cancer patients.Experimetal Apparatus The chromatographic system that was used consisted of a quaternary pump equipped with an automatic injector (1050 series) (Hewlett-Packard, Palo Alto, CA, USA) with a sample Analyst, 1999, 124, 477–482 477loop injector of 100 mL, and a high-pressure six-port valve (Rheodyne model 7000; Cotati, CA, USA).A fluorescence detector (Hewlett-Packard, 1050 Series) (flow cell, 5 mL) was coupled in series and linked to a data system (Hewlett-Packard HPLC Chem Station) which was used for data acquisition and storage. The chromatographic signal was excitation at 252 nm and emission at 500 nm for fluorescence. A Vac Master –10 sample processing station (International Sorbent Technology, Hengoed, UK) at a flow-rate of about 5 mL min21 was used.A rotary evaporator was also used. All the assays were carried out at room temperature. Reagents All the reagents were of analytical grade. Acetonitrile, methanol and acetone (Scharlau, Barcelona, Spain) were of HPLC grade. Putrescine dihydrochloride, cadaverine dihydrochloride, spermine tetrahydrochloride, spermidine trihydrochloride, dansyl chloride and 1,7-diaminoheptane were obtained from Sigma (St.Louis, MO, USA). Sodium hydrogencarbonate (Probus, Badalona, Spain), sodium hydroxide (Panreac, Barcelona, Spain), imidazol (99%) from Sigma and nitric acid was also used. Columns and mobile phases Bond Elut C18 200 mg and Bond Elut Certify 130 mg from Varian (Harbor City, CA, USA), high performance extraction disk cartridges C18 from 3M Empore (St. Paul, MN, USA) and LiChrolut EN (200 mg) and LiChrolut TSC (300 mg) from Merck (Darmstadt, Germany) were used to retain the analytes and later to perform an off-line derivatization.A C18 Lichrospher (125 3 4 mm id, 5 mm particle size) (Merck), column was used as an analytical column for separation of the amine derivatives. An acetonitrile–imidazol solution (1 mM, pH 7.0) (70 + 30 v/v) mixture in gradient elution mode was used as the eluent at a flow rate of 1 mL min21. The gradient used was 70% of acetonitrile at zero time, 90% at 5 min, and 70% at 9 min. After 9 min the percentage of acetonitrile was kept constant.All the solvents were filtered through a 0.45 mm nylon membrane (Teknokroma, Barcelona, Spain) and degassed with helium before use. Preparation of solutions Standard solutions of 2.0 g L21 Dns-Cl were prepared by dissolving the pure compound in acetone. Standard solutions of the amine compounds were prepared by dissolving the pure compounds in water (1000 mg mL21). Working amine solutions were prepared by diluting the standard solutions in water.The pH was adjusted by adding the corresponding amount of hydrogencarbonate buffer which was prepared by dissolving in water the appropiate amount of sodium hydrogen carbonate and then adjusting the pH with 10% NaOH (w/v). All solutions were stored in the dark at 4 ºC. Solution derivatization The amines were derivatized according to the method described by Marcé et al.17 To 0.1 mL of amine, 1 mL of 10 mM hydrogen carbonate buffer (pH 9.0) and 0.9 mL of acetone containing 1.0 mM dansyl chloride were added successively, and the mixture was incubated for 10 min at 70 ºC.An aliquot (20 mL) of the solution was injected into the HPLC system. Fractions I and II obtained in the pH experiment were derivatized by following this procedure. Fraction I corresponds to the fraction collected when the amine solution is passed through the column and fraction II is collected when the reagent solution is passed through the column. Derivatization on solid-phase supports Solid-phase extraction cartridges were conditioned by drawing 1.0 mL of methanol through followed by 1.0 mL of hydrogen carbonate buffer (pH 12).Aliquots (1 mL) of the samples were then transferred to the cartridges and 0.5 mL of Dns-Cl reagent prepared in acetone–hydrogencarbonate buffer solution (20 mM, pH 9.5), (2 + 3 v/v) was flushed through the cartridges. After a given reaction time, the cartridges were dried under vacuum with air at a flow of 5 mL min21.The derivatives formed were desorbed from the cartridges with 1 mL of acetonitrile. A 20 mL aliquot of the resulting mixture was finally injected into the chromatographic system. Urine samples Untreated urine samples were spiked with the analytes (putrescine, cadaverine, spermidine, and spermine) at a concentration range of 0.5 and 5 mg mL21. 1,7-Diaminoheptane was included in the sample as internal standard (IS) at a concentration level of ca 2.5 mg mL21. These samples were alkalised with NaOH and carbonate buffer to pH 12.Volumes of 1 mL of these samples were placed into conditioned C18 cartridges. Then the analytes were derivatized in the solid-phase extraction cartridges as described above. The percentage of analyte recovered after clean-up plus derivatization was calculated by comparing the peak area obtained for a particular assay with those obtained for standard solutions containing an equivalent amount of analyte. Each sample was assayed in triplicate.Real samples Five urine samples from cancer patients and ten corresponding to healthy volunteers were analysed. A portion (1 mL) of untreated urine sample was placed in the column together with the IS and the alkalised medium. Then, the analytes were derivatized as described above. Concentration procedure: 1 mL of the organic phase (acetonitrile) was removed and dried under a N2 stream while the tube was standing at 40 ºC in a water bath. The residue was dissolved in 250 mL of acetonitrile.Finally, a 20 mL sample was injected onto the HPLC system. Two of the samples from healthy volunteers were also processed by a conventional procedure:18 to 2.0 mL of urine in a test-tube, 0.2 mL of sodium hydroxide solution (10% in water) and 2.0 mL of diethyl ether were added successively. The tube was capped and shaken vigorously for 2 min. After centrifuging at 1000g for 10 min, the ether layer was collected. The extraction procedure was repeated twice.The ether extract was dried in a rotary evaporator system after the addition of a drop of diethyl ether containing 0.1 M HCl. The derivatization procedure was performed according to the description given above for solution derivatization. Results and discussion In order to have a reference for the derivatization reaction, polyamines were derivatizated in solution according to the conditions described by Marcé et al.17 (Fig. 1, Procedure I). The steps related to the sample clean-up and removal of the reagent excess were avoided (Fig. 1, Procedure II).No differences were found between both procedures and no interference of reagent 478 Analyst, 1999, 124, 477–482was obtained at the retention time of the analytes. These results were taken as 100% of derivatized product. Dansylation on solid-phase supports The retention conditions of the amines on the solid support and their reaction conditions with dansyl chloride were studied. C18 cartridges were initially used to retain the analytes and to perform the derivatization procedure.Amine solutions at different pH values ranging from 10 to 13 were tested. Three different fractions were analysed. Fraction I corresponded to the portion collected when the amine solution was passed through the column. Fraction II was collected when the reagent solution was passed through the column and fraction III was the elution of the reaction products formed in the cartridge.In order to show the retention of the biogenic amines or the loss of those species when the reagent was flushed, fractions I and II, respectively, were derivatized according to the dansylation in solution. 1,7-Diaminoheptane, spermine and spermidine did not give analytical signals in fractions I and II. Thus, those polyamines were totally retained in the cartridge. Fig. 2 shows the percentage of the total analytical signal obtained in the different fractions for cadaverine and putrescine.The higher the solution pH the lower was the amount of analyte eluted in fractions I and II because the retention of both analytes in the cartridge increased with the pH. Then, the formation of the derivatives in the cartridge was higher as the solution pH was also higher (see fraction III of Fig. 2) The results at pH 13 are not shown in the figure because in these conditions the reagent was destroyed and no analytical response was obtained. The analytical signals obtained for the derivatives formed in the cartridge (fraction III) for all polyamines as a function of the pH of the solution flushed through, are given in Fig. 3. We selected amine solutions buffered at pH 12 from Fig. 3 in order to improve spermine determination, which is the most retained species and the last eluted in reversed liquid chromatography. To establish the optimal pH for polyamine retention, the reaction conditions in the column were investigated, and parameters such as pH, reagent concentration and time were studied.The reaction was carried out at a basic pH, and the influence of this parameter was studied in the range 8–10. No big differences were observed by modifying the pH of the flushed solution in this range, so, in accordance with our paper16 concerning dansylation in C18 supports of amphetamines, a reagent mixture of hydrogen carbonate solution at pH 9.5 and Dns-Cl in water–acetone (3 + 2 v/v) was used. The effect of the Dns-Cl concentration was evaluated in the range from 1.85 3 1024 to 1.85 3 1023 M.As can be seen in Fig. 4 the analytical signal increased with the reagent concentration. Higher amounts of reagent could not be used in order to avoid analyte elution and the formation of undesirable reagent compounds. Three different reaction times were studied (5, 15 and 30 min). Analytes such as putrescine, cadaverine or 1,7-diaminoheptane (SI) did not show any change with time, however, as the response of spermine and spermidine increased with time, 30 min was chosen as the most suitable reaction time.The results obtained using the selected conditions were compared with those obtained by performing solution derivatization after working the solid-phase extraction procedure without dansyl chloride. The efficiency of the derivatization step in the cartridge was tested. As for putrescine, cadaverine and 1,7-diaminoheptane, similar amounts of reaction products Fig. 1 Different derivatization procedures.I, solution derivatization according to Marcé et al.,17 II, solution derivatization procedure modified; III, derivatization in solid cartridges. Fig. 2 Effect of the pH of the polyamine solution flushed through the cartridge. Reaction conditions: room temperature (25 °C), sodium hydrogen carbonate solution (20 mM) pH 9.5, putrescine 1.1 mg mL21 and cadaverine 1.20 mg mL21. For the meaning of I, II and III fractions see text. Analyst, 1999, 124, 477–482 479were obtained, being nearly 100%.However, for spermine and spermidine the amounts of reaction products were ca. 60–70%. In order to see whether the reaction between the amines and Dns-Cl was dependent on the solid support, different materials were tested. In Fig. 5 the recoveries with respect to dansylation in solution for each analyte by using different types of cartridges and performing the derivatization at optimal conditions are given. The best results of retention and derivatization were obtained by using the C18 packing, and it was the cartridge selected.Analytical properties According to Won Suh et al.19 for quality control of urine samples, the concentration range chosen was 500 ng mL21 up to 5 mg mL21. We selected urine with the lowest amounts of biogenic amines as a control in order to study the calibration graph in urine and the accuracy of the method. Fig. 6 shows the chromatograms corresponding to unspiked and spiked urine sample, respectively.The equations of the calibration graphs were: y = 20.0113 + 0.6258C (r = 0.999) for putrescine; y = 0.0239 + 0.7438C (r = 0.999) for cadaverine; y = 0.0031 + 0.4025C (r = 0.998) for spermidine and y = 20.0189 + 0.2722 C (r = 0.999) for spermine. The linearity was good in this range. The limit of detection (LOD) was calculated as the amount of analyte giving a peak height three times the maximum noise peak height of a blank biological sample observed at the retention time of each analyte.The LOD was 10 ng mL21. The recoveries obtained when spiked urine samples were processed are shown in Table 1. In all cases, 1,7-diaminoheptane was used as IS. This table also gives the repeatability and reproducibility achieved between days. Fig. 3 Effect of the pH of the polyamine solution flushed through the cartridge. Reaction conditions: room temperature (25 °C), sodium hydrogen carbonate solution (20 mM) pH 9.5, dansyl chloride 2 mM. 1,7-Diaminoheptane 2.26 mg mL21 (1), cadaverine 1.20 mg mL21 (2), putrescine 1.10 mg mL21 (3), spermidine 1.19 mg mL21 (4) and spermine 1.25 mg mL21 (5). Fig. 4 Effect of the reagent concentration (Dns-Cl) on the reaction rate. Conditions: room temperature (25 ºC), sodium hydrogen carbonate solution (20 mM) pH 9–9.5. 1,7-Diaminoheptane 3 mg mL21 (1), cadaverine 1.49 mg mL21 (2), putrescine 1.18 mg mL21 (3), spermidine 0.95 mg mL21 (4), and spermine 0.47 mg mL21 (5). Dns-Cl concentration was 0.1–15 mM.Fig. 5 Solid-phase extraction dansylation percentage recovery with respect to solution dansylation vs. solid support: 1, C18; 2, TSC; 3, Certify; 4, Disk; and 5, EN. Conditions: putrescine 1.1 mg mL21, cadaverine 1.20 mg mL21, spermidine 1.19 mg mL21, spermine 1.25 mg mL21 and 1,7-diaminoheptane 2.26 mg mL21. For experimental details see Experimental. Fig. 6 Chromatograms obtained for blank urine (–) and spiked with polyamines with putrescine (Put) 0.83 mg mL21, cadaverine (Cad) 0.88 mg mL21, spermidine (Spd) 0.88 mg mL21 and spermine (Sp) 0.93 mg mL21, (–).For experimental details see Experimental. Table 1 Analytical data for the determination of free polyamines in urine samples. Recoveries take into account all the concentrations tested Analyte Recovery (%) (n = 8)a Reproducibility between day (%) (n = 3) Repeatability precision (%) (n = 8) Putrescine 95 ± 4 3 4 Cadaverine 90 ± 8 6 8 Spermidine 80 ± 7 3 7 Spermine 82 ± 6 6 6 a Recoveries considering the calibration graph with standards. 1,7-Diaminoheptane was used as internal standard. 480 Analyst, 1999, 124, 477–482Table 2 gives the accuracy obtained for the urine samples spiked with different concentrations. As can be seen in this table the accuracy is good independently of the polyamine concentration. Determination of polyamines in urine samples We analyzed ten different urine samples of healthy subjects and we found that the biogenic amine concentrations were near to the detection or determination limits. We also analyzed two of these samples by using liquid–liquid extraction as sample clean up and solution dansylation and similar results were obtained. In order to improve the detection limits of the method, a concentration procedure was proposed.The elution volume (containing the reaction products) was removed and dried under a N2 stream while the tube was standing at 40 ºC in a water bath. The proposed procedure improved the detection limits to 2 ng mL21. Fig. 7 shows the chromatograms of a healthy volunteer (sample 6) and the five urine samples (samples 1 to 5) from cancer patients. As can be seen variation in the shape of the chromatogram is observed. The putrescine concentration found was 0.33, 0.44, 0.51 and 0.69 mg mL21 for urine samples 1, 2, 4 and 5, respectively. The cadaverine concentration was 0.26, 0.36, 0.18 and 0.4 mg mL21 for samples 1, 2, 4 and 6, respectively. The spermidine concentration was 0.20, 0.08, 0.13, 0.05, 0.03 and 0.18 mg mL21 for samples 1, 2, 3, 4, 5 and 6, respectively and the spermine concentration was 0.12, 0.03, 0.03 and 0.12 mg mL21 for samples 1, 2, 4 and 6, respectively.All analytes were found in urine samples 1, 2 and 4. However in urine sample 3 only spermidine was found. Spermine was found in the six urine samples. Won Suh et al.19 found that the ratio putrescine/ spermidine in urine was significantly greater in cancer patients than in normal subjects.This significant difference of ratio values supports the well known fact that extracellular putrescine concentrations may reflect either rapid tumor growth or tumor cell loss.20 Although a putrescine/spermidine ratio cannot be calculated for the healthy person because putrescine is below the detection limit, the results obtained could be in agreement with those of Won Suh et al.19 According to other authors19,21,22 greater differences can be expected in the acethylated form and can correspond to other differences observed in the chromatograms.More work is needed in order to determine definite profiles of these biogenic amines in cancer patients and healthy volunteers. The proposed method can serve in this sense as it is demonstrated. Conclusions This paper shows the possibility of using Dns-Cl reagent to perform polyamine derivatization in urine samples. The derivatization is performed in C18 solid phase cartridges, which permits the clean-up and derivatization of the analytes on the solid support, decreasing sample handing and time of analysis.The time of analysis is less than 1 h, in which 30 min corresponds to the derivatization reaction in the cartridges; during this time other work can be done. However, the conventional procedure takes more than 1.5 h working all the time using tedious operations (double liquid–liquid extraction, evaporation of the solvent). The proposed procedure also avoids heating the mixture and allows the use of more polar solvents such as acetonitrile than those employed in liquid–liquid extraction.The volume of solvent employed is smaller than that required in the conventional method. The procedure has been applied to spiked and real urine samples from healthy volunteers and cancer patients, respectively. The results obtained from spiked urine samples are accurate and precise, which allows validation of the procedure. The detection limits have been improved to 2 ng mL21 by concentration of the urine extracts.By using this procedure, the determination of polyamines (putrescine, cadaverine, spermine and spermidine) in urine of unknown cancer patient samples can be performed with satisfactory sensitivity, reproducibility and a minimum time. Acknowledgement The authors are grateful to the CICYT for financial support received for the realisation of the Project SAF 95-0586. References 1 A. E. Pegg, Biochem.J., 1986, 234, 249. 2 A. E. Pegg and P. P. MacCann, ISI Atlas of Science: Biochemistry, Institute for Scientific Information, Philadelphia, PA, USA, 1988, p. 2. 3 A. E. Pegg, Cancer Res., 1988, 48, 759. 4 D. H. Russell, C. C. Levy, S. C. Schimpff and I. A. Hawk, Cancer Res., 1971, 31, 1555. Table 2 The accuracy in the determination of free polyamines in spiked urine samples. Analyte Added conc./ mg mL21 Found conc./ mg mL21 (n = 3) Error (%) Putrescine 0.83 0.79 ± 0.12 +4.89 1.65 1.74 ± 0.06 25.43 2.20 2.17 ± 0.27 +1.39 3.30 3.26 ± 0.30 +1.15 Cadaverine 0.88 0.83 ± 0.66 +5.35 1.77 1.85 ± 0.13 24.69 2.35 2.33 ± 0.14 +0.99 3.53 3.53 ± 0.21 20.09 Spermidine 0.88 0.90 ± 0.06 22.91 1.76 1.73 ± 0.00 +1.21 2.34 2.26 ± 0.11 +3.50 3.51 3.65 ± 0.06 -3.90 Spermine 0.93 0.85 ± 0.03 +8.31 1.85 1.96 ± 0.00 -5.90 2.47 2.46 ± 0.08 +0.28 3.7 3.69 ± 0.24 0.32 Fig. 7 Chromatograms corresponding to concentrated urine from a healthy volunteer (--) and urine from cancer patients (-). For experimental details see Experimental. Analyst, 1999, 124, 477–482 4815 J. Janne, E. Holtta, A. Kallio and K. Kapyaho, Special Top. Endocrinol. Metab., 1983, 5, 227. 6 N. Seiler, B. Knödgen and F. Eisenbeiss, J. Chromatogr., 1978, 145, 29. 7 M. Kai, T. Ogata, K. Haraguchi and Y. Ohkura, J. Chromatogr., 1979, 163, 151. 8 R. L. Heideman, K. B. Fickling and L. J. Walker, Clin. Chem., 1984, 30, 1243. 9 J. R. Shipe and J. Savory, Ann. Clin. Lab. Sci., 1980, 10, 41. 10 T. Skaaden and T. Greibrokk, J. Chromatogr., 1982, 247, 1243. 11 J. Seiler, J. Chromatogr., 1986, 379, 157. 12 C. F. Verkoelen, J. C. Romijn, F. H. Schroeder, W. P. van Schalkwijk and T. A. W. Splinter, J. Chromatogr., 1988, 426, 41. 13 P. Campíns-Falcó, A. Sevillano-Cabeza, C. Molins-Legua and M. Kohlman, J. Chromagr.,1996, 687, 239. 14 R. Herráez-Hernández, P. Campíns-Falcó, A. Sevillano-Cabeza and I. Trümpler, Anal. Chim. Acta, 1997, 334, 125. 15 R. Herráez-Hernández, P. Campíns-Falcó and A. Sevillano-Cabeza, Anal. Chem., 1996, 68, 734. 16 C. Molins-Legua, P. Campíns-Falcó and A. Sevillano-Cabeza, Anal. Chim. Acta, 1998, 378, 85. 17 M. Marcé, S. D. Brown, T. Capell, X. Figueras and F. A. Tiburcio, J. Chromatogr. B., 1995, 666, 329. 18 K. Hayakawa, N. Imaizumi, H. Ishikura, E. Minogawa and N. Kobayashi, J. Chromatogr., 1990, 515, 459. 19 J. Won Suh, S. Hwa Lee, B. Chul Chung and J. Park, J. Chromatogr., B., 1997, 688, 179. 20 M. M. Abdel-Monem and K. Ohno, J. Pharm. Sci., 1977, 66, 1089. 21 M. M. Abdel-Monem and J. L. Merdink., J. Chromatogr., 1981, 222, 363. 22 C. Löser, U. Wunderlich and U. Fölsch, J. Chromatogr., 1988, 430, 249. Paper 8/08736I 482 Analyst, 1999, 124, 477–482

ISSN:0003-2654    

DOI:10.1039/a808736i

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Multi-beam circular dichroism detector for HPLC Atsushi Yamamoto,a Shuji Kodama,a Akinobu Matsunaga,a Kazuichi Hayakawab and Mitsuo Kitaokac a Toyama Institute of Health, 17-1 Nakataikoyama, Kosugi-machi, Toyama 939-0363, Japan b Faculty of Pharmaceutical Sciences, Kanazawa University, 13-1 Takara-machi, Kanazawa 920-8640, Japan c Shimadzu Corporation, 1 Nishinokyo-kuwabara-cho, Nakagyo-ku, Kyoto 604-8442, Japan Received 5th January 1999, Accepted 26th February 1999 A novel multi-beam circular dichroism (CD) detector for HPLC is described. A unique feature of this detector is the use of a retardation plate to give many quarter-waves in the wavelength region of interest.When a polarizing prism and a thin quartz plate as the retarder are placed in a conventional photodiode array (PDA) detector, the kinetic spectra of an optically active analyte recorded by the instrument contain sinusoidal CD waves in the wavelength axial direction. Extraction of these CD waves superimposed on the absorption spectra was performed with MS Excel.Both the instrumental conditions for the PDA data acquisition and the polynomial data processing were optimized by monitoring the chromatographic elution of camphor. The limit of detection for ajimaline, a chiral alkaloid, was 26 ng injected amount. Circular dichroism (CD) is defined as the difference in absorbance between left- and right-circularly polarized beams in a chiral medium. In a widely used method, alternating beams with opposite rotational senses are transmitted through the sample cell.Although the sensitivity of a CD spectrometer can be improved by alternating the rotational sense more frequently, the smaller the value of De/e (the anisotropy factor, AF1) of a general chiral compound, the longer the CD measurement time will be. This causes a problem when a CD spectrometer is used as an HPLC detector. This is because a low volume flow-cell is needed for high resolution, but if the volume is too low, the time that the analyte is in it may be too short for an accurate CD measurement.Only a few studies have used this type of CD detector for the HPLC analysis of microgram quantities of chiral compounds.2–4 Recently, in order to supplement this defect of the conventional CD spectrometer as an HPLC detector, a double-beam CD detector has been developed. In one such instrument developed by Rosenzweig and Yeung5, left- and right-circularly polarized beams created from the double refraction of calcite intersect precisely in the sample cell.Another instrument built in our laboratory is equipped with a split-type flow-cell.6 In this type of instrument, two photodiodes are used to offset the noise of the light source so that CD can be measured directly with high sensitivity. In this system, however, individual differences among diodes cause an increase in the background absorption signal, which makes it difficult to maintain the baseline.As a result, it is difficult to measure the CD of analytes with high absorptivities. This problem with the double-beam detection system was largely attributed to the use of relative difference in absorbance between the two diodes as the CD value. We thought that we could overcome this problem by using more diodes. In this paper the development of a multi-beam HPLC-CD detector, which is excellent in both sensitivity and operability, is presented.This optical system can be easily constructed by placing a polarizing prism and a quartz plate in front of a flowcell of a conventional photodiode array (PDA) detector. Experimental The LC system consisted of a Shimadzu (Kyoto, Japan) LC- 10AD pump, a Rheodyne (Cotati, CA, USA) 7125 injector with a 20 mL loop, a Tosoh (Tokyo, Japan) CO-8011 column oven and a Shimadzu SPD-M10Avp PDA detector. A Gran-Taylor prism (1 cm square, Sigma Koki, Hidaka, Saitama, Japan) was placed in the compartment for the wavelength calibrating filter in the PDA detector.A quartz plate (5 mm wide, 0.455 mm thick) purchased from Five Lab (Kawasaki, Japan) was placed on the incident light side of the flow-cell, whose principal axis was inclined by 45° from the polarizing axis. The threedimensional data obtained were converted into ASCII format with the aid of the detector’s bundled software, and were processed with a spreadsheet program (Excel, Microsoft Corp., Redmond, WA, USA).The CD spectra were obtained with a Jasco (Hachiouji, Tokyo, Japan) J-500C CD spectrometer and the UV spectra with a Shimadzu UV-2200 spectrophotometer. Analysis of camphor was performed on a 15 cm 3 4.6 mm id reversed-phase column (Supelco TPR-100, Sigma Aldrich Japan, Tokyo, Japan) maintained at 40 °C. Acetonitrile–water (3 + 1) was used as the mobile phase at a flow rate of 0.6 mL min21. Analysis of ajimaline, which is an alkaloid contained in Rauwolfia serpentina, was performed on a 25 cm 3 4.6 mm id reversed-phase column (Capcell Pak UG120 Å, Shiseido, Tokyo, Japan) maintained at 40 °C. 0.02 M phosphoric acid– acetonitrile (5 + 3) was used as the mobile phase at a flow rate of 0.6 mL min21. Single enantiomers of camphor were purchased from Sigma Aldrich Japan, and ajimaline was from Nacalai tesque (Kyoto, Japan). HPLC-grade acetonitrile and other reagents were purchased from Wako Pure Chemicals (Osaka, Japan). Results and discussion Principle of multi-beam CD detection Holzwarth7 suggested the possibility of measuring the CD spectrum by placing a polarizing prism and a multiple-order retardation plate in the spectrophotometer.The principle of this method lies in the opposite rotational senses of the circularly polarized beams created from the adjacent quarter-waves, i.e., the polarization in the incident beam alternates as the wavelength is varied. As a result, the absorption spectrum of an optically active analyte oscillates with wavelength.Although Analyst, 1999, 124, 483–485 483the sample beam in the spectrophotometer is monochromatic, we applied this principle to the PDA detector where the incident beam is not spectrally diffracted in advance. We assumed that the spectral diffraction does not destroy the CD data which is included in such circularly polarized beams, because CD is derived from light absorption. While in the conventional CD spectrometer the monochromatic circularly polarized beam alternates with time, in the proposed method the circularly polarized beam alternates with wavelength.The experimental design of this multi-beam CD detector is depicted in Fig. 1. A thin plate of quartz as a uniaxial, birefringent crystal, which was vertically cut along the principal crystallographic axis, was used as a multiple-order retarder because of its low cost. It is easy to construct this detector if there is enough space to set a polarizing prism in a conventional PDA detector.Effect of PDA parameters In the conventional CD spectrometer, high-frequency modulation of the circularly polarized beam reduces the noise and improves its sensitivity. The same applies to the multi-beam CD detector. The shorter the 1/4 wavelength interval produced from the thin plate is, the better is its detection sensitivity. Since the photodiodes in the general PDA detector are arranged about every 1 nm, there is a limit to the period of the CD wave that is detectable. A quartz plate with 0.455 mm thickness was used to make the 1/4 wavelength interval less than 5 nm.The interval caused by this plate at around 250 nm was 4 nm. The spectral bandwidth is another factor that influences the resolution of the spectrum. A wide bandwidth reduces the noise but sacrifices the resolution. In the SPD-M10Avp PDA detector, the spectral bandwidth is adjusted by varying the number of diodes used for data acquisition at a specific wavelength. Although the noise levels in the absorption spectrum depended on the stability of the detector, they were usually in the ranges 6 31025–1024 absorbance units (AU) when using 1 diode, 2 3 1025–3 3 1025 AU when using 2 diodes, and 1025–3 3 1025 AU when using 3 diodes, and further increases in the number of diodes did not bring about any improvement.The absorption spectra of (+)-camphor measured by altering the spectral bandwidth are shown in Fig. 2. The spectra were differentially processed to easily find the CD wave.The amplitude of the CD wave decreased as the number of diodes increased. Two diodes, corresponding to a spectral bandwidth of 2 nm, were used for subsequent experiments by considering the ratio of the noise level and the CD amplitude. Extraction of the CD wave from the absorption spectrum All data processing was performed by the Savitzky–Golay polynomial algorithm.8 In this procedure, the degree of the polynomial used should be decreased and the number of data points used should be increased to reduce the effect of the noise.In this experiment, however, the period of the CD wave obtained was about 10 nm and practicable data was limited to this range. As shown in Fig. 2, high-order differential processing reduces the influence of the original spectrum of the analyte, but leads to an increase in noise by using a high-degree of polynomial with a limited number of data points. It should be mentioned that the absorbance values at half-waves on this spectrum are independent of the chiral purity of the analyte, and the line passing through these points is indeed the baseline that we were looking for.The CD wave was extracted by subtracting the spectrum of the racemate which was normalized to pass these zero-points. As mentioned above, the higher the degree of the polynomial, the easier is the normalization. The CD chromatograms of camphor with various concentrations and various chiral purities which were drawn up in this manner from their 2nd differential spectrum are shown in Fig. 3. The vertical axes of these chromatograms show the aggregate of the CD values in the range 260–308 nm in order to gain high sensitivity. Differential processing from the 0th to the 3rd gave almost the same chromatograms. One difference between the analysis of a standard and the analysis of an actual sample is that the latter might have some impurities that overlap with the peak of interest. Since differential processing has been very effective in such a case,9 2nd differential spectra were used for all subsequent procedures.Limits of detection of the multi-beam detector Although the noise level of this PDA detector is 2 3 1025– 3 3 1025 AU with a spectral bandwidth of 2 nm as mentioned Fig. 1 Schematic diagram of the multi-beam HPLC-CD detector. The part within the dotted line represents a conventional PDA detector. Fig. 2 Second differential spectra of (+)-camphor in a chromatographic cell with various numbers of diodes corresponding to spectral bandwidth: amount injected, 10 mg; spectral bandwidth marked on figure; for chromatographic conditions see Experimental.Fig. 3 CD chromatograms of camphor: upper trace, 10 mg of (+)-isomer; second trace, 1 mg of (+)-isomer; third trace, 10 mg of (+)/(2) = 10/9 isomer; bottom trace, 10 mg of (+)/(2) = 9/10 isomer. 484 Analyst, 1999, 124, 483–485above, subsequent data processing reduces it to below 1025 AU.The proposed detection system can theoretically measure the amplitude of the CD wave within about 2 3 1025 AU. The ratio between this amplitude and the absorption of the analyte is identical to the AF. Generally, the AF of a chiral compound is in the order of 1023 or less. However, the CD wave of such a compound can be detected with data processing if its peak height is more than 0.02 AU. This value can always be obtained with HPLC, so this system is thought to be a universal CD detector for chiral compounds.The limit of detection of this system depends on the e of the analyte. The CD and UV spectra of ajimaline in acidic medium are shown in Fig. 4. This alkaloid shows both maxima around 250 nm with an AF of about 1/800. Consequently, a CD wave can be detected at this wavelength when the amount that is injected gives a peak height of 16 mAU. Under the present LC conditions, an injection of 0.15 nmol corresponded to this peak height.CD chromatograms of ajimaline obtained by using a spectrum without the quartz plate as a baseline are shown in Fig. 5. We were able to detect 0.08 nmol (26 ng) of ajimaline, which is about half the calculated amount. Moreover, an injection of 0.25 nmol ajimaline resulted in the appearance of a negative peak around 290 nm, where the AF of ajimaline is about 21/2600. This suggested that a CD amplitude below 1025 AU could be measured. In this experiment, the quartz plate used was restricted in thickness for the resolution of the PDA detector.As the measurement accuracy of a conventional CD spectrometer depends on the frequency of the beam modulation, the same may be said of the relationship between the sensitivity of the proposed system and the quarter-wave interval produced by the quartz plate. The application of a thinner quartz plate to the PDA detector with higher resolution will improve the sensitivity of this detection system.Conclusion A multi-beam CD detector, which can detect a CD wave on the wavelength axial direction, was constructed by inserting a prism and a thin quartz plate into a PDA detector. The CD wave was successfully extracted by differential processing of the absorption spectrum. Although this processing is also useful for removing the influence of impurities on a chromatogram, a high-order differential causes an increase in baseline noise, and further investigations revealed that 2nd differential processing gave the best results.The proposed detection system under the established conditions had not only a sensitivity that was equal to or better than that of a double-beam CD detector, but also a facility comparable to a conventional CD spectrometer to detect the absolute De. This system was able to detect as little as 26 ng of ajimaline, which is CD-active around 250 nm (e is about 8000 and De is about 10). References 1 P. Salvadori, C. Bertucci and C. Rosini, Chirality, 1991, 3, 376. 2 C. Bertucci, P. Salvadori and L. F. L. Guimaraes, J. Chromatogr. A, 1994, 666, 535. 3 O. Zerbinati, R. Aigotti and P. G. Daniele, J. Chromatogr. A, 1994, 671, 281. 4 K. Takatori, S. Toyama, S. Fujii and M. Kajiwara, Chem. Pharm. Bull., 1995, 43, 1797. 5 Z. Rosenzweig and E. S. Yeung, Appl. Spectrosc., 1993, 47, 2017. 6 A. Yamamoto, A. Matsunaga, K. Hayakawa and M. Nishimura, Biomed. Chromatogr., 1997, 11, 362. 7 G. Holzwarth, Rev. Sci. Instrum., 1965, 36, 59. 8 A. Savitzky and M. J. E. Golay, Anal. Chem., 1964, 36, 1627. 9 A. Yamamoto, A. Matsunaga, M. Ohto, E. Mizukami, K. Hayakawa and M. Miyazaki, Analyst, 1995, 120, 337. Paper 9/00152B Fig. 4 UV (— ) and CD (- - - - -) spectra of ajimaline in 0.02 M H3PO4– MeCN (2 : 1). Fig. 5 CD chromatograms of ajimaline at the detection limit. The amounts of analytes are 26 (·····) and 86 ng (––––) injected. Analyst, 1999, 124, 483–485 485

ISSN:0003-2654    

DOI:10.1039/a900152b

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

Application of cloud-point methodology to the determination of polychlorinated dibenzofurans in sea water by high-performance liquid chromatography A. Eiguren Fernández, Z. Sosa Ferrera and J. J. Santana Rodríguez* Department of Chemistry, Faculty of Marine Sciences, University of Las Palmas de G.C., 35017 Las Palmas de G.C., Spain Received 3rd December 1998, Accepted 26th January 1999 Polychlorinated dibenzofurans (PCDF) are organic compounds with very toxic effects for humans and the environment.As they are present in low concentrations, an extraction technique is necessary prior to their determination by high-performance liquid chromatography (HPLC). In this work, the methodology of cloud-point extraction, using two non-ionic surfactants oligoethylene glycol monoalkyl ether (Genapol X-080) and polyoxyethylene-10-cetyl ether (Brij 56), is applied to the extraction and preconcentration of PCDF in sea water samples prior to their determination by HPLC with fluorescence detection.Introduction Interest in determining trace levels of contaminants in environmental samples has increased with the recognition that even trace levels of pollutants can pose risks for humans. The polychlorinated dibenzofurans (PCDF) are considered one of the most hazardous pollutants of the environment.1 It is known that isomers of PCDF having substituents at positions 2,3,7 and 8 are among the strongest synthetic poisons possessing carcinogenic and mutagenic effects.2–4 Various combustion processes represent major sources of these compounds in the environment. The PCDF formed are emitted to the atmosphere, and due to their hydrophobic and lipophilic properties they are dispersed onto land or into water and will adsorb strongly to organic material, both in the particulate and dissolved phases.The analysis of PCDF is complicated due their extremely low levels of concentration in the samples. Tedious, costly and time consuming methods for extraction with an organic solvent or clean-up steps are required prior to the analysis of these compounds by gas or liquid chromatography.5–7 Micelles and other organized amphiphilic assembles are increasingly utilized in analytical methods.8 Their unique microheterogeneous structures capable of selective interaction with different solute molecules, can strongly modify solubility, chemical equilibria, kinetics and the spectroscopic properties of analytes and reagents. Separation procedures based on the peculiar properties of aqueous non-ionic and zwitterionic surfactant solutions have been also proposed as an alternative to the use of traditional organic solvents.Efficient preconcentrations of organic solutes have been obtained using the cloud-point phase separation phenomena.9–13 The analytical potential of the cloud-point phenomenon mediated phase separations, cloud-point extraction (CPE), has been discussed by several authors.12–17 The small volume of the surfactant-rich phase allows us to preconcentrate and extract the analytes in one step, prior to gas or liquid chromatographic analysis.Moreover, the surfactantrich phase is compatible with the micellar and aqueous–organic mobile phase in liquid chromatography, which facilitates the application of this analytical method, with the obvious benefits. In this work, we report the results of a study of the experimental parameters which affect the extraction efficiency and preconcentration factor of the CPE process of a series of PCDF previous to their determination by high-performance liquid chromatography (HPLC), using two non-ionic surfactants, oligoethylene glycol monoalkyl ether (Genapol X-080) and polyoxyethylene-10-cetyl ether (Brij 56).This CPE methodology has been applied to the analysis of a mixture of PCDF in sea water samples. Experimental Reagents Stock solutions of 1.0 31024 M of PCDFs (AccuStandard, Inc., New Haven, CT, USA) were prepared by dissolving appropriate amounts of each PCDF in ethanol (Merck, Darmstadt, Germany).Stock solutions of the surfactants, oligoethylene glycol monoalkyl ether (Genapol X-080) (Sigma, St. Louis, MO, USA), and polyoxyethylene-10-cetyl ether (Brij 56, Sigma) were prepared in deionized water. HPLC-grade methanol and KNO3 were obtained from Panreac Quimica, S.A. (Barcelona, Spain). All solvents and analytes were filtered through a 0.45 mm nylon membrane filter, and ultra-high-quality water, obtained using a water purification system, was used throughout.Apparatus The HPLC system consisted of a Waters pump (model 510) fitted with a Rheodyne injector valve (model 7725i) with a 20 ml sample loop and a Waters 474 scanning fluorescence detector. The system and the data management are controlled by Millennium software from Waters (Waters Cromatografia S.A., Barcelona, Spain) The stationary-phase column was a Waters Nova-Pack C18, 3.9 3150 mm id, 4 mm particle diameter.The maxima lex and lem of each PCDF in the presence of surfactant were determined using a Perkin-Elmer Hispania, S.A. (Madrid, Spain) luminescence spectrophotometer, model LS- 50. Excitation and emission slits of 5 nm were used. Analyst, 1999, 124, 487–491 487A thermostatted bath (model Tectron 200) and a centrifuge (model Mixtasel), from Selecta (Barcelona, Spain), were also used. Procedure Preconcentration. The preconcentration step was carried out keeping aliquots of 10 ml, which contain the analytes and surfactant (2% w/v), in a thermostatically controlled bath for 15 min, at a temperature of 95 °C for both surfactants.After this time, the aliquots were centrifuged for 5 min (3500 rpm) to achieve the separation of the two phases. Previous to the centrifugation, the tube holder was heated for 15 min at a temperature close to 95 °C in an electric system. Liquid chromatography analysis with fluorescence detection.The study of the chromatographic conditions established a methanol–water (85 + 15) mobile phase and 1 ml min21 flow rate as the most adequate conditions in order to obtain an efficient separation of the six PCDF. As the extracted surfactant-rich phase is compatible with the mobile phase, 20 ml of the extracted volume was injected directly into the liquid chromatograph and the detection of the analytes was carried out by monitoring the relative intensity of fluorescence at the maxima lem and lex of each PCDF under study.Two different ranges were studied: 3.4–27.2 mg ml21 and 168–3405 ng ml21. The fluorescent conditions ( lem and lex ) were previously determined using solutions containing a known concentration of each PCDF and a surfactant concentration above the critical micellar concentration ( Table 1). Determination of PCDF in sea water samples. Prior to the analysis, sea water was passed successively through filters of different porosity (0.45 and 0.22 mm) and ultraviolet radiation to avoid the posible interference of marine microorganisms.Solutions of 10 ml containing 5 ml sea water and 2% (w/v) surfactant concentration were spiked with suitable amounts of PCDF and analysed according to the established CPE method. Results and discussion The cloud-point temperature of these surfactants, necessary for the extraction step, was studied in a previous work.12 In that study, the consolution curve of each surfactant was obtained when the temperature at which the turbidity appeared was represented against the concentration of surfactant in solution.The phase diagrams indicate a cloud-point temperature of 75–80 °C for Genapol X-080 and 85–90 °C for Brij 56. At the same time, the ratio between the concentration and volume of the added surfactant and the surfactant-rich phase volume obtained after the cloud-point process was studied for both surfactants.The results showed an increase in the extracted volume of surfactant-rich phase with the surfactant concentration in solution, but the volume of added surfactant had no influence on the volume of surfactant-rich phase.12 A 2% (w/v) surfactant concentration and a volume of 4 ml were chosen for the study carried out in this work, obtaining in these conditions a preconcentration factor of 10.Optimization of the preconcentration factor There are different factors that can alter the extraction process and it is very important to optimize them in order to obtain good recovery factors.Some of these variables are concentration and volume of added surfactant, pH, ionic strength, pressure, etc.18,19 In the case of PCDF, some of the factors have no influence on the recovery percentages, thus pH is not important because these compounds do not present ionic forms; therefore, only the equilibration time, initial concentration of analyte in solution, ionic strength and surfactant concentration are the parameters which have been optimized.Effect of the equilibration time. The recovery percentage, and because of this the efficiency of the cloud-point extraction process, depends on the time that the analytes have to interact with micelles and get into their core. For this reason, to optimise the extraction process, it is necessary to study the behavior of each surfactant when the equilibration time varies. If we represent recovery percentages against equilibration time, two different behaviors can be observed: the behavior of the analytes with an even number of chlorine atoms and those with an odd number of chlorine atoms.This different behavior would be due to the structure of the molecules (different position of chlorine atoms) which generate different interactions between the analyte and the surfactant micelles, and therefore the solubilization effect would also be different. For Genapol X-080, we observed that MonoPCDF presents a similar behavior to the PCDF with a even number of chlorine atoms.The recovery percentages of these analytes decrease with time, to 13 min, increasing to 15 min ( maximum recovery percentage) and going down again for longer times. In the case of PCDF with odd numbers of chlorine atoms, recovery percentages increase with time to 13 min, decreasing for 15 min and increasing for longer periods ( Fig. 1a). When we used Brij 56, the recovery percentages decreased sharply for all the PCDF with time up to 13 min, but with longer periods two behaviors could be observed: for PCDF with even Table 1 Fluorescent characteristics of PCDF in 2% (w/v) aqueous solutions of Genapol X-080 and Brij 56 Genapol Brij 56 Abbrevia- lex/ lem/ lex/ lem/ No. PCDF tion nm nm nm nm 1 Dibenzofuran DBF 278 316 278 316 2 4-Chlorodibenzofuran MonoCDF 265 333 280 321 3 2,8-Dichlorodibenzofuran DiCDF 255 334 291 334 4 2,4,8-Trichlorodibenzofuran TriCDF 256 335 291 337 5 2,3,7,8-Tetrachlorodibenzofuran TetraCDF 258 335 302 337 6 1,2,3,4,8-Pentachlorodibenzofuran PentaCDF 258 335 294 344 Fig. 1 Influence of equilibration time on recovery percentages of PCDF using: (a) Genapol X-080, (b) Brij 56; (1) DBF; (2) MonoCDF; (3) DiCDF; (4) TriCDF; (5) TetraCDF; (6) PentaCDF. 488 Analyst, 1999, 124, 487–491numbers of chlorine atoms, the recovery percentage, in all the range of concentration, increased up to 20 min, decreasing gradually over longer periods; the rest of the PCDF presented low recovery percentages for 20 min, increasing up to 25 min (Fig. 1b). An equilibration time of 15 min was chosen as adequate for all the analytes in solutions. Effect of the ionic strength. The presence of salt in the solution can be important for the extraction process and in the extracted volume of surfactant-rich phase. The study of the influence of this parameter was carried out by adding different percentages of KNO3 (1–10% w/v) to the solution.The results indicated that the addition of salts does not affect the extracted volume of surfactant-rich phase and as the recovery percentages obtained are similar we can conclude that the ionic strength does not have notable influence on the extraction process. But the addition of the inert salt increases the density of the bulk aqueous phase and facilitates the separation process of the two phases. These results are in accord with those obtained by other authors.20 Effect of the analyte concentration.The initial concentration of analyte is another factor that can affect the recovery percentages. To determine this effect, solutions containing different concentrations of PCDF were subjected to the CPE procedure. As the analyte concentration varied between 50.4–2043 ng ml21 the recovery percentages remained practically constant for all the analytes studied in both surfactants, so we can conclude that the initial concentration of PCDF in solution had no influence on the recovery percentages when we used Genapol X-080 and Brij 56 as extractants (Fig. 2). Effect of the added surfactant percentage. When the influence of the added surfactant percentage (w/v) was studied, the results showed that this parameter had no influence on the recovery percentages for the analytes under study (Fig. 3). Cloud-point preconcentration and liquid chromatographic analysis Another important step in this study is the optimization of the chromatographic conditions. Once the conditions of CPE were optimized to obtain good recovery percentages, it was necessary to optimize the conditions for the separation and determination of PCDF using HPLC with fluorescence detection.When chromatography is used as a separation technique, it is necessary to obtain a good relationship between analysis time Fig. 2 Effect of the initial analyte concentration in solution on the recovery percentages: (a) Genapol X-080, (b) Brij 56; (1) DBF; (2) MonoCDF; (3) DiCDF; (4) TriCDF; (5) TetraCDF.Fig. 3 Influence of added surfactant percentage (w/v) on the recovery percentages: (a) Genapol X-080, (b) Brij 56; (1) DBF; (2) MonoCDF; (3) DiCDF; (4) TriCDF; (5) TetraCDF. Table 2 Analytical parameters for the determination of PCDF with CPE Calibration range Parameter DBF MonoCDF DiCDF TriCDF TetraCDF PentaCDF 168–3405 ng ml21 a (31026)a 1.2 0.1 1.1 0.6 2.5 0.1 b (31025)b 4.0 0.1 1.0 0.4 1.1 0.2 3.4–27.2 µg ml21 a (31024)a 12.2 0.6 10.3 58.0 23.6 0.7 b (31024)b 14.8 1.5 13.9 36.6 1.1 0.7 a a = Slope.b = Intercept. Fig. 4 Elution of a mixture of seven PCDF using 2% (w/v) Genapol X-080 as preconcentrant: (1) DBF; (2) MonoCDF; (3) DiCDF; (4) TriCDF; (5) TetraCDF; (6) PentaCDF. Concentration of each PCDF, 500 ng ml21; eluent, methanol–water (85 + 15); flow rate, 1 ml min21. Analyst, 1999, 124, 487–491 489and analyte separation. An efficient separation of the six PCDF in study, in a relatively short period, was obtained when we used a methanol–water (85 + 15) mobile phase and a flow rate of 1 ml min21 (Fig. 4). The study of the analytical characteristics of the determination of PCDF using Genapol X-080 and Brij-56 showed a similar behavior for both surfactants. The calibration curves were obtained by duplicate injection of the sample containing 2% (w/v) of surfactant and the corresponding PCDF concentration. Two different ranges were studied: 3.4–27.2 mg ml21 and 168–3405 ng ml21.A linear relationship between fluorescence signal and concentration of analyte was observed and high correlation coefficents were obtained for the different calibration curves (0.999–0.998). Table 2 lists the calibration characteristics of the method. Once the methodology under study was optimized, the recovery percentages for each PCDF were determined. Different mixtures of PCDF were prepared in 10 ml of 2% (w/v) aqueous solutions of Genapol X-080 and Brij 56. After the CPE step, 20 ml of the extracted surfactant-rich phase were injected directly into the chromatographic system.The data obtained, listed in Table 3, show excellent recovery percentages for the different analytes studied in both surfactants. The limits of detection21 calculated for each PCDF in Genapol X-080 and Brij 56, using CPE methodology, are listed in Table 4, together with the relative standard deviation for six samples to which the complete procedure (cloud-point preconcentration and extraction and chromatographic separation) was applied for all the compounds studied in 2% (w/v) Genapol X-080 and Brij 56 solutions.The results obtained for both parameters were similar when Genapol X-080 and Brij 56 were used. Analytical applications The proposed method was applied to the determination of PCBF in sea water samples from different Spanish areas (Arinaga and Agaete, Canary Islands and Elantxobe, Vizcaya). Sea water samples were previously spiked with suitable amounts of each PCBF.The results indicate satisfactory recovery percentages. (Table 5). Acknowledgements This work was supported by funds provided by D.G.I.C.Y.T. (Spain), research Project No. PB94-0431. Table 3 Recovery percentages obtained for each PCDF after CPE and chromatographic determination* 168–3405 ng ml21 3.45–27.2 mg ml21 Genapol Brij 56 Genapol Brij 56 PCDF Recovery % Recovery % Recovery % Recovery % DBF 100 96 96 92 MonoCDF 98 94 97 92 DiCDF 99 93 98 90 TriCDF 88 95 93 95 TetraCDF 98 92 88 99 PentaCDF 97 68 99 89 * Referred to the initial concentration of PCDF in solution.Table 4 Relative standard deviation (in range 168–3405 ng ml21) and limits of detection for PCDF PCDF DBF MonoCDF DiCDF TriCDF TetraCDF PentaCDF RSD (n = 6) 0.009 0.204 0.016 0.131 0.043 0.087 LOD/ng ml21 0.5 17.6 13.3 4.9 0.7 27.5 Table 5 Recovery percentages for PCDF in sea water samples Genapol Brij 56 Origin Added*/ Found*/ Recovery Added*/ Found*/ Recovery Added*/ Found*/ Recovery Added*/ Found*/ Recovery of sample PCDF ng ml21 ng ml21 (%) ng ml21 ng ml21 (%) ng ml21 ng ml21 (%) ng ml21 ng ml21 (%) Agaete DBF 84.0 83.2 99 840 789.6 94 84.0 77.3 92 840 789.6 94 MonoCDF 101.2 91.1 90 1012 931.1 92 101.2 95.1 94 1012 991.7 98 DiCDF 118.5 104.3 88 1185 1054.6 89 118.5 107.8 91 1185 1102.0 93 TriCDF 135.7 127.6 94 1357 1302.7 96 135.7 119.4 88 1357 1262.0 93 TetraCDF 153.0 142.3 93 1530 1468.8 96 153.0 137.7 90 1530 1514.7 99 PentaCDF 170.2 163.4 96 1702 1615.0 95 170.2 115.7 68 1702 1565.8 92 Aringa DBF 84.0 84.8 101 840 764.4 91 84.0 79.8 95 840 756.0 90 MonoCDF 101.2 105.2 104 1012 921.0 91 101.2 93.1 92 1012 961.4 95 DiCDF 118.5 117.3 99 1185 1149.4 97 118.5 107.8 91 1185 1208.7 102 TriCDF 135.7 133.0 98 1357 1262.0 93 135.7 123.5 91 1357 1411.3 104 TetraCDF 153.0 148.4 97 1530 1591.2 104 153.0 146.9 96 1530 1407.6 92 PentaCDF 170.2 161.7 95 1702 148.9 91 170.2 127.6 75 1702 1582.9 93 Elentrobe DBF 84.0 87.3 104 840 806.4 96 84.0 83.2 99 840 756.0 90 MonoCDF 101.2 89.0 88 1012 971.5 96 101.2 99.2 98 1012 961.4 95 DiCDF 118.5 111.4 94 1185 1161.3 98 118.5 106.6 90 1185 1196.8 101 TriCDF 135.7 124.8 92 1357 1357.0 100 135.7 119.4 88 1357 1411.3 104 TetraCDF 153.0 145.3 95 1530 1606.5 105 153.0 139.2 91 1530 1560.6 102 PentaCDF 170.2 161.7 95 1702 1719.0 101 170.2 119.1 70 1702 1688.0 98 * Referred to the initial concentration of PCDF in solution. 490 Analyst, 1999, 124, 487–491References 1 C.Rappe, Pure Appl. Chem., 1996, 68(9), 1781. 2 C. L. Fletcher and W. A. McKay, Chemosphere, 1993, 26, 1041. 3 D. Broman, C. N�af, C. Rolff and Y. Zeb�uhr, Environ. Sci. Technol., 1991, 25, 1850. 4 R. E. Clement, Anal. Chem., 1991, 63, 1130. 5 B. Larsen and S. Facchetti, Fresenius’ J. Anal. Chem., 1994, 348, 159. 6 M. Oehme, A. Bartonova and J. Knutzen, Environ. Sci. Technol., 1990, 24, 1836. 7 E. Maier, B. Griepink and U. Fortuniti, Fresenius’ J. Anal. Chem., 1994, 348, 23. 8 E. Pramauro, Ann. Chim. (Rome), 1990, 80, 101. 9 R. G. Laughrin, in Advances in Liquid Crystals, ed. G. H. Brown, Academic Press, New York, 1978, vol. 3, pp. 41, 76, 106. 10 H. Watanabe, in Solution Behavior of Surfactants, Plenum Press, New York, 1982, pp. 1305–1313. 11 W. L. Hinze and E. Pramauro, Crit. Rev. Anal. Chem., 1993, 24, 113. 12 A. Eiguren Fern�andez, Z. Sosa Ferrera and J. J. Santana Rodr�ýguez, Quim. Anal., 1997, 16, (2), 283. 13 A. Eiguren Fern�andez, Z. Sosa Ferrera and J. J. Santana Rodr�ýguez, Anal. Chim. Acta, 1998, 358, 145. 14 W. L. Hinze, Ann. Chim. (Rome), 1987, 77, 167. 15 N. D. Gullickson, J. F. Scamehom and J. H. Harwell, in Surfactant- Based Separation Processes, eds. J. F. Schamehorn and J. H. Harwell, Marcel Dekker, New York, 1989, pp. 139–151. 16 H. Hoshino, T. Saitoh, H. Taketoni, T. Yotsuyanagi, H. Watanabe and K. Tachikawa, Anal. Chim. Acta, 1983, 147, 339. 17 T. Saitoh and W. L. Hinze, Talanta, 1995, 42(1), 119. 18 R. P. Frankewich and W. Hinze, Anal. Chem., 1994, 66, 944. 19 S. P. Moulik, Curr. Sci., 1996, 71(5), 368. 20 E. Minati and D. Zanette, Colloids Surf. A, 1996, 113, 237. 21 S. Lindsay, in High Performance Liquid Chromatography, Wiley, New York, 1992, pp. 71, 72. Paper 8/09457H Analyst, 1999, 124, 487&n

ISSN:0003-2654    

DOI:10.1039/a809457h

出版商:RSC    

年代:1999

数据来源: RSC

摘要:

HPLC with fluorescence detection of methamphetamine and amphetamine in segmentally analyzed human hair Osama Al-Dirbashi,a Naotaka Kuroda,a Shigeko Inuduka,b Francesco Menichinic and Kenichiro Nakashima*a a School of Pharmaceutical Sciences, Nagasaki University, 1–14 Bunkyo-machi, Nagasaki 852-8521, Japan. E-mail: naka-ken@net.nagasaki-u.ac.jp b Kinki Regional Narcotics Control Office, 4–1–76 Otemae, Chuo-ku, Osaka, Japan c Department of Pharmaceutical Sciences, University of Calabria, 87036 Arcavacata di Rende, Italy Received 16th November 1998, Accepted 11th February 1999 A sensitive high-performance liquid chromatographic method with fluorescence detection for determining methamphetamine and its major metabolite, amphetamine, in abusers’ hair segments was developed. Methamphetamine and amphetamine in hair samples collected from addicts were extracted into acidified methanol, derivatized with 4-(4,5-diphenyl-1H-imidazol-2-yl)benzoyl chloride, separated isocratically on an ODS column using TRIS–HCl buffer (0.1 mol dm23, pH 7.0)–methanol (30 + 70 v/v) as the mobile phase and the derivatives were detected fluorimetrically at 440 nm (lex 330 nm).Calibration curves obtained by using control human hair spiked with standard solutions were linear (r � 0.999) up to at least 676.1 ng mg21 for amphetamine and 746.1 ng mg21 for methamphetamine. The detection limits at a signal-to-noise ratio of 3 were 51.4 and 74.6 pg mg21 hair for amphetamine and methamphetamine, respectively.Using control hair spiked with standard solutions, the intra- and inter-day relative standard deviations (n = 5) were � 8.6% for both the target compounds. The method was successfully applied to the segmental analyses of methamphetamine abusers’ hair samples. 1. Introduction Problems related to ever-increasing substance abuse have been attracting more public attention and research efforts. Drugs of abuse including methamphetamine (MP) are known to cause serious social and law enforcement problems in addition to pathological changes in many organs and systems in the body.In Japan, MP and amphetamine (AP) are controlled by the Stimulants Drug Control Law, the former being abused by most Japanese addicts whereas the latter is detected as the main metabolite in biological samples.1,2 For forensic science laboratories and criminal justice agencies, hair might become the indispensable matrix for measuring exposure to xenobiotics, particularly drugs of abuse.Owing to its biological stability and physical state, hair is easy to collect and store until analysis can be carried out.3 Moreover, the slow growth rate and the absence of drug metabolism in hair allow investigations over a lengthy period in comparison with other biosamples.4 Hair analyses might disclose drug abuse history by providing long-term information on an individual’s illegal drug consumption, in contrast to short-term information that blood or urine analysis provides.The detection of several drugs in human hair has been reported, including cannabis,5 phencyclidine,6 cocaine,7–9 opiates,8–11 tricyclic antidepressants12 and nicotine, 10,12 in addition to MP and related compounds, referred to hereafter as methamphetamines (MPs).1,4,8,12,13–18 Although gas chromatography–mass spectrometry (GC-MS) is the most frequently utilized method to determine drugs in hair,3–6,8–17 other techniques such as radioimmunoassay (RIA)7 and highperformance liquid chromatography with peroxyoxalate chemiluminescence detection (HPLC-POCL) after derivatization with a fluorescent reagent have also been reported.1,18 A lophine (2,4,5-triphenylimidazole) derivative, namely 4-(4,5-diphenyl-1H-imidazol-2-yl)benzoyl chloride (DIB-Cl), was recently synthesized in our laboratory19 and proved to be a superior fluorescent labeling reagent for MPs owing to its reactivity with primary and secondary amines in different solvents under mild conditions, stability in acetonitrile and excellent detectability in comparison with some other commercially available fluorescent reagents.20,21 In this paper, we report a simple and highly sensitive HPLC method with fluorescence detection for determining the concentrations of MP and AP in abusers’ hair samples. Segmental analyses to reveal the drug abuse time course were performed on a 1 cm length basis. 2. Experimental 2.1. Materials and reagents DIB-Cl was synthesized according to our reported method.19 MP hydrochloride was obtained from Dainippon Pharmacy (Osaka, Japan).AP sulfate was synthesized in our laboratory according to known procedures.22,23 Acetonitrile and methanol (HPLC grade) were purchased from Wako (Osaka, Japan). Analytical-reagent grade tris(hydroxymethyl)aminomethane (TRIS) was obtained from Sigma (St. Louis, MO, USA). Water was de-ionized and passed through a pure line WL21P system (Yamato Kagaku, Tokyo, Japan).All other chemicals were of analytical-reagent grade and used as received. Stock standard solutions of MPs were prepared by dissolving suitable amounts of these compounds in water to give a final concentration of 1.0 mmol dm23 per compound. These solutions were kept in the dark at 4 °C and were stable for at least 6 months. Working standard solutions were prepared by dilution in NaHCO3–Na2CO3 buffer (10.0 mmol dm23, pH 9.0), Analyst, 1999, 124, 493–497 493hereafter referred to as carbonate buffer, and in 5.0 mol dm23 HCl–methanol (1 + 20 v/v) for standard sample derivatization and to spike the hair samples, respectively.Spiked hair samples were prepared by adding aliquots of MP standard solutions to hair samples collected from healthy subjects in our laboratory who had not taken any medication over the past few months. Hair samples belonging to MP addicts were obtained from Kinki Regional Narcotics Control Office (Osaka, Japan).These samples were kept at 220 °C until needed. TRIS–HCl buffer was prepared by adjusting the pH of TRIS aqueous solution to 7.0 with concentrated HCl and then adjusting the volume to give a final concentration of 0.1 mol dm23. 2.2. Hair sample pre-treatment and fluorescence derivatization After marking their roots and tips, 10 strands of hair were stuck in parallel on a commercially available double-faced sticker and cut into 1 cm long segments. Each group (10 hair segments of 1 cm length each) was further cut to give 20 pieces of approximately 0.4–0.6 cm.Individually, these groups were washed by successive sonication in 10 ml of 3.5 mmol dm23 aqueous sodium dodecyl sulfate (SDS) (0.1%) and 5 ml of ethanol for 1 h and 10 min, respectively. After decanting the ethanol, the hair was rinsed with 10 ml of distilled water, filtered off, thoroughly flushed with about 30 ml of distilled water and left to dry either at room temperature or in a dish dryer.Individually, the hair groups were accurately weighed, placed in clean test-tubes and portions of a mixture of 5.0 mol dm23 HCl– methanol (1 + 20 v/v) were added to give a final hair concentration of 1 mg ml21. After sonication for 1 h, the hair was left to stand in the extraction solvent for 24 h at room temperature. After the hair had been filtered off, 100 ml of the filtrate were transferred into a screw-capped, brown-glass reaction vial and dried under a gentle stream of N2 at room temperature.To the residue, 10 ml of carbonate buffer (10.0 mmol dm23, pH 9.0) and 180 ml of 0.1 mmol dm23 DIB-Cl in acetonitrile were successively added, vortex mixed and kept at room temperature (ca. 24–27 °C). After 10 min, 10 ml of 8.0 mol dm23 ammonia solution (28%) were added, vortex mixed for a few seconds and 5 ml of the resultant mixture were injected into the HPLC system. Fig. 1 shows the derivatization reaction scheme. 2.3. Efficiency of external decontamination procedure Approximately 0.5 cm long control hair samples were soaked overnight in aqueous MP standard solutions at concentrations of 0.01, 0.5 and 10.0 mmol dm23, which are equivalent to 1.4, 67.6 and 1352.1 ng mg21, respectively, of AP and 1.5, 74.6 and 1492.1 ng mg21 of hair, respectively, of MP. After filtering off and drying hair naturally, it was extracted into 5.0 mol dm23 HCl–methanol (1 + 20 v/v) with or without previous washing with SDS, ethanol and distilled water as described in Section 2.2.These samples were measured in triplicate. 2.4. Extraction efficiency The extraction efficiency was assessed by sonicating 5 mg of abuser’s hair sample D in 5 ml of 5.0 mol dm23 HCl–methanol (1 + 20 v/v) for 1 h followed by incubation at room temperature for 48 h. Portions of 100 ml were taken at 0, 6, 12, 24 and 48 h after the sonication. These samples were measured in triplicate. 2.5. HPLC system and operating conditions The HPLC system consisted of an LC10AS HPLC pump (Shimadzu, Kyoto, Japan), a Model 7125 injector with a 5 ml sample loop (Rheodyne, Cotati, CA, USA), a 250 3 4.6 mm id, 5 mm Daisopak SP-120-5-ODS analytical column (Daiso, Osaka, Japan), a CTO-6AS column oven set at 35 °C (Shimadzu), an RF-550 spectrofluorimeter (Shimadzu), set at lex and lem of 330 and 440 nm, respectively, and a U-228-2P- 500 recorder (Nippon Denshi Kagaku, Tokyo, Japan); 0.5 mm id stainless-steel tubing was used in all flow lines.The HPLC separation was carried out isocratically by using TRIS–HCl buffer (0.1 mol dm23, pH 7.0)–methanol (3 + 7 v/v) as the mobile phase at a flow rate of 1.0 ml min21. The eluent was pre-mixed and de-gassed prior to use. 2.6. Calibration curves and reproducibility Control human hair was treated as described in Section 2.2 except for the addition of MP standards to give final concentrations in the range 135.2 pg mg21–676.1 ng mg21 for AP and 149.2 pg mg21–746.1 ng mg21 of hair for MP prior to the extraction step.These samples were sonicated for 1 h, left to stand at room temperature for 24 h, then analyzed as described. Calibration curves were constructed by plotting the fluorescence intensity (FI) as peak height in arbitrary units against the concentration in ng mg21. Intra- and inter-day reproducibilities of the proposed method were assessed by using control hair spiked with 27.0 and 29.8 ng mg21 of AP and MP, respectively.These samples were analyzed as described over a period of 9 d. 3. Results and discussion The primary purpose of this study was to establish a simple analytical method that provides information about the state of chronic stimulants abuse. For this purpose, segmental analysis of abusers’ hair samples seems to be sufficiently informative if a highly selective and sensitive analytical method is employed. Several methods for the determination of MPs in hair have been reported but, to the best of our knowledge, none of them claimed a detection limit of sub-picogram level per injection.Recently, although MPs as dansyl chloride (Dns-Cl) derivatives could be determined successfully in a single human hair by HPLC with POCL detection, the detection limit was about 2 pg on-column.1 Fig. 1 Reaction scheme for labeling MP with DIB-Cl. 494 Analyst, 1999, 124, 493–497Using stable isotope dilution GC-MS, a detection limit of 0.5 ng mg21 hair, which is equivalent to 50 pg per injection, was also reported.15 More recently, we reported that MPs in human urine could be detected as DIB derivatives in the sub-picogram level per injection by using a simple HPLC–fluorescence system;20,21 hence in this study we optimized a new analytical method using DIB-Cl as the fluorescent label for the determination of these drugs in abusers’ hair segments to reveal the time course of substance abuse. 3.1. Pre-treatment of hair and fluorescence derivatization It has long been known that drugs and elements found in the body are incorporated into hair via a mechanism or mechanisms that are not completely understood.Generally, drugs present in blood might diffuse passively into the rapidly growing hair follicles and become stably embedded in the hairshaft. In addition, the secretions of sweat, apocrine and sebaceous glands are possible vehicles for drug transfer into hair. For smokable drugs such as MPs, the analyst should consider the external exposure (i.e., from air) as another potential route of contamination. 24 Hair does not grow continuously, so it should not be treated as a uniform structure. Following an active growth and a short transitional phase, hair follicles enter a resting (i.e., telogen) phase in which the hairshaft stops growing and can be removed easily. On the scalp of an adult human, approximately 15% of the hairs are in the resting phase.25 Therefore, in our experiments we used 10 hair strands from each abuser that had the same physical properties (color, length) to avoid misleading results due to telogenic hair.After cutting into segments of 1 cm, each group of hairs was washed by sonication in 3.5 mmol dm23 SDS, which has been reported to remove the external contamination on the sample,14,15 followed by sonication in ethanol. However, when control hair samples were soaked in standard aqueous solutions of MPs overnight, significant adsorption of these compounds onto or into the hair strands took place at concentrations of � 0.5 mmol dm23. As shown in Table 1, the externally adsorbed amounts of MPs could not be totally removed, which suggests that part of the adsorbed amounts might have permeated into the hairshafts rather than being adsorbed on the surface.Therefore, the external contamination should not be underestimated. However, if the metabolites together with the parent drug were simultaneously detectable in a suspect’s hair, this suggests that they were incorporated through the blood or the secretions of the sweat, apocrine and sebaceous glands rather than being from an external source such as air.Sonication of hair in 5.0 mol dm23 HCl–methanol (1 + 20 v/v) for 1 h followed by overnight incubation at room temperature was reported to be an efficient extraction method of MP and AP from human hair.4,15 As shown in Fig. 2, there was a proportional relationship between the incubation time and the relative recovery; hence we employed an incubation period of 24 h.The derivatization reaction was performed according to our previously reported conditions21 except that acetone as the reagent’s solvent was changed to acetonitrile, since DIB-Cl is stable for a longer time in the latter, and the addition of 10 ml of 8.0 mol dm23 ammonia solution, that reacts with the excess DIB-Cl to stop the reaction and results in faster elution of the reagent blank peaks. The derivative’s stability was monitored over 24 h when kept at room temperature in the dark.DIB-AP and -MP were sufficiently stable and no significant change in the peak heights was observed after 24 h. This suggests the possibility of the simultaneous preparation of many samples to be analyzed later or to be injected through an automated sample injector. 3.2. Chromatographic separation Fig. 3 illustrates typical chromatograms of control human hair, control hair spiked with 27.0 and 29.8 ng mg21 of AP and MP, respectively, and the first centimeter from the scalp surface of hair sample D collected from a drug abuser.As shown in Fig. 3(A), no interference due to endogenous compounds in human hair was observed and the two target compounds were well separated from the reagent blank. The retention times of DIBAP and -MP were 35.4 and 40.5 min, respectively. The use of 5.0 mol dm23 HCl–methanol (1 + 20 v/v) resulted in a more selective extraction in comparison with liquid–liquid extraction using hexane after treating the hair with 2.5 mol dm23 NaOH solution.However, the peaks corresponded to compounds coextracted with MPs and those due to the reagent blank, preventing attempts at reducing the retention time. 3.3. Calibration curves, detection limits and reproducibility Calibration curves of AP and MP using spiked hair samples were obtained by plotting the peak height against the corresponding concentration in ng mg21 hair. Linear relationships were obtained in the concentration range 135.2 pg mg21–676.1 ng mg21 for AP and 149.2 pg mg212746.1 ng mg21 of hair for MP.The regression equations and correlation coefficients for the two compounds were as follows: AP, y = 0.26x + 1.04 (r = 1.000); and , y = 0.11x + 0.54 (r = 0.999); y is the fluorescence intensity as peak height in arbitrary units and x is the concentration in ng mg21 hair. The detection limits at a signal-to-noise ratio of 3 were 51.4 (0.13) and 74.6 (0.19) pg mg21 hair (pg per 5 ml injection) for AP and MP, respectively. Although it is known that spiking hair with drug substances will not result in a situation similar to those incorporated in vivo, unlike other biosamples such as urine, Table 1 Effect of washinga on the external contamination of control hair after being soaked in MP aqueous solution overnight Washed before extraction/ Extracted without washing/ Concentration ng mg21 hair ± SD ng mg21 hair ± SD of AP and MP (n = 3) (n = 3) in water/ mmol dm23 AP MP AP MP 0.01 NDb ND ND ND 0.5 0.8 ± 0.1 1.1 ± 0.3 2.1 ± 0.1 2.4 ± 0.3 10.0 12.2 ± 0.4 12.2 ± 0.5 29.5 ± 0.8 28.6 ± 0.5 a Washing by sonication with SDS and ethanol followed by rinsing with distilled water.For details, see Section 2.2. b Not detectable. Fig. 2 Effect of incubation time of hair at room temperature after 1 h of sonication in 5.0 mol dm23 HCl–methanol (1 + 20 v/v) on the efficiency of extraction of MPs from abuser hair sample D.Curves: 1, AP; and 2, MP. Analyst, 1999, 124, 493–497 495adding these substances to a control hair gives information about other compounds found in normal hair that might be coextracted and interfere with the determination of the target compound. Intra- and inter-day variation data are summarized in Table 2. The reproducibility of the present method was investigated over 9 d using control hair samples spiked with AP and MP at concentrations of 27.0 and 29.8 ng mg21, respectively.Satisfactory intra- and inter-day reproducibility data (n = 5) with relative standard deviations (RSD) of � 8.6% for both compounds were obtained. In general, for the determination of MPs by GC-MS, derivatization is required to increase the volatility and it has been performed using trifluoroacetic anhydride at 55 °C for 20 min4,15,16,26 or pentafluoropropionyl anhydride for 30 min at 60 °C8 or 40 min at 80 °C.17 To be sensitively detected by POCL after HPLC separation, Takayama et al.derivatized AP and MP with Dns-Cl at 45 °C for 1 h.1,18 However, in our study, the derivatization of MPs with DIB-Cl for 10 min at room temperature was not only found to be simple, but also led to > 10 times higher sensitivity in comparison with the HPLCPOCL method.1 Compared with GC-MS methods,8,15,17 the present method is still more than two orders of magnitude more sensitive in terms of the injected amount. To carry out hair analysis for drugs of abuse, Moeller27 recently reported that 20–50 mg of hair material are sufficient for standard procedures with electron-impact detection and single ion monitoring. An amount of hair in the range 10–50 mg was used in GC-MS methods.8,15,17 Owing to its high sensitivity, sub-milligram amounts of hair were used in our study, which are comparable to those reported by Takayama et al., who performed the analysis on a single hair.1,18 In our procedure, however, the 1 cm-based segmental analysis of 10 hair strands from each addict is obviously more informative than a whole single hair analysis. 3.4. Segmental analysis of abusers’ hair samples Human scalp hair grows at an average rate of 1 cm per month.8,14 Therefore, segmental analysis of hair cut into sections 1 cm long provides a monthly-based drug abuse time course as this information is stored in the keratinized hair. In this study, six hair samples from known abusers were segmentally analyzed and the concentrations of MP determined in different segments were in the range of 1.8–170.7 ng mg21.AP, the main metabolite of MP, was also detected in many of the segments and its concentrations ranged from 0.5 to 11.4 ng mg21. In addition to segment 3–4 of sample C, MP was also not detectable in segments which were more than 4 cm from the scalp surface of the 14 cm long hair sample B. This suggests that sample B corresponds to a relatively recent rather than a chronic abuse case.Interestingly, sample C contained both black and white hair in which MP was determined at the level of 5.5–10.6 ng mg21 in the former whereas it was not detectable in the latter. This suggests that the presence of melanin, the principal pigment of the hair, is a determining factor for the incorporation of these compounds into hair according to their affinity to this pigment. The chemical similarity between the amino acid tyrosine, a precursor of melanin,24,25 and MPs might significantly participate in the incorporation mechanism of these substances in pigmented hair.Similarly, Takayama et al. reported that MP was detected at a concentration of 10.2 ng mg21 in black hair whereas it was not detectable in white hair from the same subject.18 In a detailed study, Nakahara et al. also reported a high correlation coefficient (r2 = 0.979) between the product melanin affinity 3 lipophilicity of 19 neutral and basic drugs, including MPs, and their incorporation rates into hair.28 Results of the segmental analysis of six hair samples of MP abusers are summarized in Table 3.Except for sample D, which contained a maximum MP concentration of 170.7 ng mg21 in segment 2–3, our results regarding the concentrations of MPs in abusers hair are, however, comparable to those reported in the literature,1,4,12,13,18 with a maximum of 125.9 ng mg21.12 In conclusion, hair, as a biological sample, provides the toxicological or forensic analyst with answers corresponding to questions related to drugs of abuse that could not be clarified through conventional blood or urine analysis.Owing to its availability, ease of collection, storage, stability and stability of the drugs embedded in its keratinized strands for lengthy periods, hair is an excellent sample when chronic exposure to Fig. 3 Typical chromatograms with fluorescence detection corresponding to (A) control human hair, (B) control human hair spiked with 27.0 ng mg21 of AP and 29.8 ng mg21 of MP and (C) segment 0–1 of abuser hair sample D.Peaks: 1, DIB-AP; and 2, DIB-MP. Arrows in (A) indicate the expected retention times of DIB-AP and -MP. For other experimental conditions, see text. Table 2 Intra- and inter-day reproducibilities of the determination of AP and MP in spiked human hair Concen- Intra-day (n = 5) Inter-day (n = 5) tration added/ Found/ RSD Found/ RSD Compound ng mg21 ng mg21 (%) ng mg21 (%) AP 27.0 25.8 7.7 26.9 8.0 MP 29.8 28.2 6.6 29..9 8.6 496 Analyst, 1999, 124, 493–497drugs of abuse is in question.In this paper, we have presented a new method for the determination of MP and AP in 1 cm long hair segments of abusers. Not only is the proposed method simple and highly sensitive, but also a sub-milligram amount of hair material is sufficient to carry out the analysis. The fluorescence derivatization reaction proceeds under mild conditions and the resultant derivatives are stable for at least 24 h when kept at room temperature in the dark.The present method was successfully applied to hair samples collected from MP abusers, and hence might be useful for those interested in forensic and toxicological investigations. References 1 N. Takayama, S. Tanaka and K. Hayakawa, Biomed. Chromatogr., 1997, 11, 25. 2 T. Inoue, K. Tanaka, T. Ohmori, Y. Togawa and S. Seta, Forensic Sci. Int., 1994, 69, 97. 3 A. Pollettini, A. Groppi and M. Montagna, Forensic Sci.Int., 1993, 63, 217. 4 Y. Nakahara, K. Takahashi, Y. Takeda, K. Konuma, S. Fukui and T. Tokui, Forensic Sci. Int., 1990, 46, 243. 5 V. Cirimele, P. Kintz and P. Mangin, Forensic Sci. Int., 1995, 70, 175. 6 M. H. Slawson, D. G. Wilkins, R. L. Foltz and D. E. Rollins, J. Anal. Toxicol., 1996, 20, 350. 7 G. Koren, J. Klein, R. Forman and K. Graham, J. Clin. Pharmacol., 1992, 32, 671. 8 M. R. Moeller, P. Fey and R. Wennig, Forensic Sci. Int., 1993, 63, 185. 9 A. M. Bermejo Barrere and S.Trano Rossi, Forensic Sci. Int., 1995, 70, 203. 10 P. Kintz and P. Mangin, Forensic Sci. Int., 1993, 63, 99. 11 A. Tracqui, P. Kintz, B. LudesC. Jamey and P. Mangin, J. Forensic Sci., 1995, 40, 263. 12 I. Ishiyama, T. Nagai and S. Toshida, J. Forensic Sci., 1983, 28, 380. 13 O. Suzuki, H. Hattori and M. Asano, J. Forensic Sci., 1984, 29, 611. 14 Y. Nakahara, K. Takahashi and K. Konuma, Forensic Sci. Int., 1993, 63, 109. 15 Y. Nakahara, K. Takahashi, M. Shimamine and Y.Takeda, J. Forensic Sci., 1991, 36, 70. 16 S. Suzuki and T. Inoue, J. Anal. Toxicol., 1989, 13, 176. 17 M. Rothe, F. Pragst, K. Spiegel, T. Harrach, K. Fischer and J. Kunkel, Forensic Sci. Int., 1997, 89, 11. 18 N. Takayama, S. Tanaka, R. Kizu and K. Hayakawa, Jpn. J. Toxicol. Environ. Health, 1998, 44, 116. 19 K. Nakashima, H. Yamasaki, N. Kuroda and S. Akiyama, Anal. Chim. Acta, 1995, 303, 103. 20 O. Al-Dirbashi, J. Qvarnstrom, K. Irgum and K. Nakashima, J. Chromatogr. B, 1998, 712, 105. 21 O. Al-Dirbashi, N. Kuroda, F. Menichini, S. Noda, M. Minemoto and K. Nakashima, Analyst, 1998, 123, 2333. 22 N. Nagai, Yakugaku Zashi, 1893, 13, 901. 23 A. Buzas and C. Dufour, Bull. Soc. Chim. Fr., 1950, 139. 24 G. L. Henderson, Forensic Sci. Int., 1993, 63, 19. 25 M. R. Harkey, Forensic Sci. Int., 1993, 63, 9. 26 T. Niwaguchi, S. Suzuki and T. Inoue, Arch. Toxicol., 1983, 52, 157. 27 M. R. Moeller, Ther. Drug Monit., 1996, 18, 444. 28 Y. Nakahara, K. Takahashi and R. Kikura, Biol. Pharm. Bull., 1995, 18, 1223. Paper 8/08912D Table 3 Concentration of MP and AP in abusers’ hair samples determined by the present method Concentrationa of MP/AP in hair segmentsb/ng mg21 Sample Color Length/cm 0–1 1–2 2–3 3–4 4–5 5–14 A Black 5 34.5/1.5 12.4/0.7 20.7/1.1 14.7/0.6 11.0/0.5 B Brownc 14 22.9/1.1 14.2/NDe 26.6/1.1 5.1/ND ND/ND ND/ND C Grayd 4 10.6/1.0 6.4/1.1 5.5/0.8 ND/ND D Black 3 106.9/8.4 119.3/8.0 170.7/11.8 E Black 5 1.8/ND 1.8/ND 2.8/2.5 11.5/ND 3.7/ND F Black 3 2.7/0.8 2.5/0.5 2.3/ND a Average of duplicate measurements. b Segments are numbered according to their distance from the scalp surface (cm). c Brown dye applied to a black hair. d A mixture of black and white hair. MPs were not detectable in white hair and these results correspond to the black portion. e Not detectable. Analyst, 1999, 124, 493–497 497

ISSN:0003-2654    

DOI:10.1039/a808912d

出版商:RSC    

年代:1999

数据来源: RSC