326 GOODALL AND DAVIES : AUTOMATIC DETERMINATION OF PENICILLIN [Vol. 86 Automatic Determination of Penicillin in Fermentation Broth An Improved Iodimetric Assay BY R. R. GOODALL AND 130SEDA DAVIES (Imperial Chemical Industries Ltd. , Pharmaceuticals Division, A lderley Park, Macclesfeld, Cheshive) The automatic analytical method described by Ferrari, Russo-Alesi and Kelly for assaying penicillin fermentation broths has been modified in order to overcome certain experimental difficulties and to improve the precision. The method described in this paper requires preliminary manual dilution to bring the concentrations of the reactants (penicilloic acid and iodine) into the linear range of the recorder - calorimeter unit, in which the excess of iodine is measured as its blue complex with starch.It is convenient to include penicillinase in the diluent, so that the automatic stage is confined to iodination, mixing with starch solution and recording the optical density of the residual iodine as its blue complex. Four of these are stan- dards, and, as the assay of each broth requxes both a “blank” to determine iodine-absorbing impurities and a “test” to determine absorption of iodine after conversion by enzyme to penicilloic acid, the net average output of assays is 18 per hour. With the accompanying standards, preparations, etc., the number of samples analysed in a normal working day is 100. The standard deviation of the assay (including both “test” and “blank” iodinations) is from 2 to 2.5 per cent. ; this is similar to that of the manual titrimetric method, in which each operator can carry out 25 tests daily.The method has been in routine use for several months. Forty iodine absorptions are recorded hourly. THREE colorimetric reactions of the penicillin nucleus have been adapted for mechanisation, These are (a) the formation of a ferric - hydroxamic $acid complex, (b) conversion to penicilloic acid, which is used to reduce arsenomolybdate to a blue colour and (c) conversion to penicilloic acid and then reaction with iodine. An automatic adaptation of method (a) has recently been described by Niedennayer, Russo-Alesi, Lendzian and Kelly.l A possible disadvantage of this method is its low intrinsic sensitivity, so that the effect of turbidity present in the sample of broth or arising during the reaction cannot be minimised by dilution ; further, under certain conditions of fennenta- tion the “blanks” for broths are of the same order as the “potencies,” with consequent loss of precision.This criticism is less likely to apply to the iodimetric assay, as the consumption of iodine by penicilloic acid is exceptionally high (about 9 equivalents per mole). Method (b), first described by Pan,2 has been adapted by Green and Monk3 to a semi-automatic procedureMay, 19611 I N FERMENTATION BROTH: AN IMPROVED IODIMETRIC ASSAY 327 involving manually operated units for rapid dispensing, extraction, evaporation, development of colour and measurement with a recording spectrophotometer. Method (c) is the most generally accepted procedure as to specificity and precision.Penicillin broth is converted by penicillinase to penicilloic acid, which is then allowed to react with iodine under acid conditions ; allowance is made for the iodine absorption of the broth without penicillinase treatment. This method is routinely used in these laboratories without any preliminary treatment of the fermentation broth.4 A visit to the Technicon Instrument Corporation’s laboratories in New York provided the opportunity for one of us (R.R.G.) to try out the fully automatic adaptation of method (c) described by Ferrari, Russo-Alesi and Kelly5 in 1959. Subsequently, our object has been to improve the sensitivity and precision of the method to the point at which a standard deviation of from 2 to 3 per cent., together with high speed of analysis, can be expected for fermentation broths of potencies greater than 2000 units per ml.Such broths have to be tested in large numbers during a mutant-screening programme. The iodination of penicilloic acid is not stoicheiometric. Consequently, the problem of achieving the desired accuracy and precision with an automatic method (this applies equally to “blank” and “test,” as the results are calculated from the difference between them) to ensure substantial agreement with the well established manual method is probably one of the most difficult applications demanded of the AutoAnalyzer. The nature of these problems is discussed below. EXPERIMENTAL DEVELOPMENT The absorption spectrum of iodine in an excess of potassium iodide solution exhibits a maximum a t 360mp.The relationship between optical density at 360mp and concen- tration over the range 0 to 1 micro-equivalent of iodine per ml is linear and can be observed as a change in optical density (0.2 cm) of from 0 to 1.8 if a suitable spectrophotometer is used. At this wavelength, however, a Technicon colorimeter fitted with suitable filters does not respond well to changes in concentration of iodine, presumably because of the low photo- activity of selenium cells in the near-ultra-violet region. Ferrari, Russo-Alesi and Kellp measured optical densities at 420mp, at the “heel” of the absorption curve, but their pub- lished calibration graph does not cover all the available scale and also has an unfavourable slope below a concentration of 2000 units of penicillin per ml, at which the optical density approaches 0.5 (equivalent to 32 per cent.transmission), corresponding to concentrations of about 1 micro-equivalent of iodine per ml. To attain the requisite precision in an absorptiometric adaptation for penicillin broths, it was necessary to span a large part of the available scale in the determination of total iodine absorption (“test”). About one-third of this must then be deducted to allow for the iodine- absorbing impurities (“blank”). To attain optical densities of the desired order over a range of low concentrations, the colorimetric reaction must be highly sensitive. Iodine absorbs only slightly in the visible spectrum, but for this work an absorption band in the visible region is required so that the AutoAnalyzer can be used to good advantage.Of the alternative methods for determining iodine, one based on the starch - iodine complex seemed to be suitable, as the absorption spectrum of this complex is broad and has a maximum at 600mp; further, Beer’s law is obeyed for concentrations less than 1 micro-equivalent of iodine per ml. The chosen concentration of soluble starch was just sufficient to develop maximum optical density with 1 micro-equivalent of iodine per ml. Although the peak absorption of this starch - iodine complex is a t 600 mp, our available interference filters had maximum trans- mission at 526 mp; our measurements were therefore made below the maximum, and some sensitivity was lost. The absorption peak, however, is not sharp, so that the difference is not critical.The blue complex is stable, provided that high concentrations of electrolyte or acid are avoided. The level of absorption is not affected by changes in room temperature. The scale of an AutoAnalyzer recorder normally records percentage transmission ( T ) , but a “linearising” device that derives 100 x the mantissa of log(T, per cent.) and displays the result on a linear scale is also available. The chart is calibrated in equal divisions from 0 to 100, and, within the limits of linear response (about 20 to 99), the transposition to optical density (d) is (100 - %)/loo, where x is the observed reading; for our purpose d is not required. This scale is about the same length as that of a normal spectrophotometer having logarithmic calibration, but the “linearised” scale can be read more precisely in the region where a328 GOODALL AND DAVIES : AUTOMATIC DIETERMINATION OF PENICILLIN [VOl.86 logarithmic form is cramped. This facilitates the precise graphical interpolation of results with the chart reader (see “Apparatus,” p. 33111. Linearity cannot be maintained below a reading of about 20, so that, for example, the curvature shown in Fig. 1 is due to a decrease in electronic response and not to deviation from Beer’s law. Other spectrophotometric measurements on these solutions show that the linear relationship continues up to 0-25 micro- equivalent of iodine per ml. I O o o 0 5 0.10 0.15 I 0.20 0.21 Concentration of iodine (as starch complex), micro-equivalents per ml Fig. 1. Relationship between concentration of iodine and chart reading when a 626-12-27 filter and a 10-mm flow cell are used I- , 0 “ O 5 0 2 4 6 s I 10 I 1 ’ 12 ‘ 1 14 Concentration of potassium phenoxymethylpenicillin standard, pg per ml Fig.2. Relationship between concentration of penicillinase-treated potassium phenoxymeth ylpenicillin standard and chart reading when a 526-12-27 filter is used The response curve obtained from a suitable range of potassium phenoxymethylpenicillin standards treated by the method described later is shown in Fig. 2. An increase in the concentration of the penicillin diminishes the concentration of the blue complex and so produces an increase in the chart reading. The “blank” for the broth is determined by usingMay, 19611 IN FERMENTATION BROTH : AN IMPROVED IODIMETRIC ASSAY 329 that part of the curve below a concentration of 5 pg per ml and the “test” on the linear part above this level.The curvature in the 0 to 5 pg per ml region arises in the recorder for the reason already stated. The development of an assay under these conditions required that the penicillin broth be treated with penicillinase and then diluted to a level at which the iodine consumed was less than the limit shown in Fig. 1 (0-15 micro-equivalent per ml). The use of the dialyser unit for rapid automatic dilution was not as promising as the more orthodox approach finally adopted. After an unsuccessful attempt to operate an automatic three-stage serial dilution of the sample drawn from the sampler plate into a network of pumping circuits, the system suggested by Dr.Holms of these laboratories was put into practice as a preliminary manual operation. This system consists essentially of a pair of self-levelling dilution pipettes (see “Apparatus,” p. 331) arranged in series. 0.073 inch bore Sample solution (0 t o 15 pg per ml) “ Cactus ” joint , b A C G in at 2.0 ml per minute >Air in at 2.5 ml per minute -Iodine solution (0.64 micro-equivalent per ml) in a t 2-0 mi per minute 0.073 inch bore silicone-rubber tubing 0.073 inch bore Starch - potassium iodide solution h j o i n t - t i 4 2 mg per rnl) in a t 2.0 ml per minute To colorimeter - and recorder a Mixing coils (3) Fig. 3. Flow diagram for iodimetric assay of penicillin Suitable standard solutions of phenoxymethylpenicillin were prepared and inactivated in 100-ml portions, so that a set of these could be incorporated on each sampling plate.As alkaline or enzymatic inactivation of the standards produced similar results, the former procedure was ultimately adopted to avoid enzymatic contamination of subsequent “blanks.” If the standards were inactivated with enzyme, the excess would have to be destroyed; obvious methods, such as boiling or precipitation by protein reagents, were thought to be impracticable. Inactivation of the samples by penicillinase was retained on the grounds of specificity. The mid-point of the acetate buffering range, pH 4.6, was chosen as the final pH for storing solutions before analysis; this was in order to avoid decomposition of the penicillin in the “blank,” a complication that could be expected at a lower pH and would result in spuriously high blank values.The necessity for high dilution outside the available AutoAnalyzer units prompted us to incorporate the treatment with enzyme in the dilution stage. The automatic part of the analysis was therefore confined to the iodination and complex-forming stages, There was no dialysis, so that theoretically there should not be much opportunity for background “noise” caused by inter-stage effects. As can be seen from the flow diagram (Fig. 3), 2 ml per minute of diluted sample are mixed with 2 ml per minute of iodine solution (0.64 micro-equivalent per ml), segmented‘ with 5 ml per minute of air and allowed to react for 2 minutes; no increase in absorption of iodine is obtained by prolonging the reaction. Subsequently, the injection of 2 ml per minute of starch -potassium iodide solution converts all the residual iodine to the blue complex.The fixed parts of the reaction circuit are in glass, with Tygon butt-joints between coils. In the feed- and pump-lines for sample and starch - potassium iodide solutions, Tygon tubing is used. Because of the high solubilityof iodine in Tygon, the dilute iodine solution enters via a polythene line into a silicone-rubber pump-tube; the starch - iodine complex also leaves the circuit via a polythene line.330 GOODALL AND DAVIES : AUTOMATIC DETERMINATION OF PENICILLIN [Vol. 86 In a normal AutoAnalyzer, a constant flow of the recipient stream passes from the dialyser to the colorimeter; the concentration of solute in this stream alters by diffusion through the cellophane membrane.In our apparatus, there is no separate recipient stream, and the act of sampling increases the throughput per minute of the liquid stream from 4 ml (2 ml each of iodine and starch - potassium iodide solutions) to 6 ml (2 ml each of iodine, starch - potassium iodide and sample solutions). Thus, there is a cyclical increase in flow and in dilution of the reagents by 50 per cent. during the time of sampling, which lasts for 1 minute, with 30-second intervals. The small “blank” peak when buffer is used as sample is a measure of this dilution effect, and the sharp rise at the end of the peaks (see Fig. 4) is attributed to the change in flow rate; this rise can be ignored. 80 70 60 50 40 30 20 10- - - - - - - - - - - - - - n i r I I/ 40 30 20 10 Fig.4. Typical recordings for penicillin broths: (a) “tests”; (b) “blanks.” Reading from right to left, the peaks in (a) are for buffer solution, thirty-three broths, buffer solution, three standards and buffer solution; those in (b) are for buffer solution, thirty-four “blanks,” buffer solution, five standards and buffer solution. (The “blanks” do not correspond to the “tests”) Several aspects of mixing become apparent when the starch - iodine complex is formed. For example, if the liquid injected into the main segmented stream is of different density, complete mixing requires passage of the segments, through approximately three loops of the mixing coil. Removal of air from the stream and passage down a capillary tube cause “drag,” so that small inter-sample boundaries become lost. Unwanted mixing and “dead” space occur at the entry to the cuvette.Precision is dependent on the absence of drift, both in the electrical components and in the concentration of iodine. To avoid drift from the components, the output of the photo- cells has to be matched and must remain so at the working temperature; this is checked by connecting the output from each photo-cell to a potentiometer and tapping off a suitable fraction to a spot galvanometer; one galvanometer is used for each photo-cell.May, 19611 IN FERMENTATION BROTH: AN IMPROVED IODIMETRIC ASSAY 331 Loss of iodine into the silicone-rubber pump-tube is greatest during the first 30 minutes; thereafter, slight loss occurs by diffusion through the wall of the tube.The effect on the assay is minimised by interpolating standards at 1-hour intervals or more frequently. METHOD APPARATUS- Compound pipettes for serial dilution-Three pipettes of the type shown in Fig. 5 are required. In operation, capillary C (bore approximately 1-3 mm; volume 0.3366 ml; length approximately 16 cm, depending on volume of joint to stopcock) is filled with sample by suction, and the outside of the jet is wiped dry. Bulb B is filled with diluent, and its contents are then used to displace and dilute the sample in C by turning the stopcocks to deliver as shown. If the volumes of bulb B, capillary C and length D are B, C and D, respectively, then the dilution factor is given by the expression [B + (C - D)]/C. After drainage, a short length of liquid, D, remains in the capillary.t Sample in Fig. 6. Compound dilution pipette To achieve a dilution of 17-5 to 1, bulb B is blown to a capacity of 5.8 & 0.1 ml and the capillary C (0-016 to 0.017 ml per cm) is made 1 cm longer than the calculated length. The operative dilution factor is determined by diluting N hydrochloric acid. A mark is made 1 cm from the tip of C, and the volume from the mark to the stopcock is filled with the acid by suction. An aliquot of the acid, after dilution as described above, is titrated with 0.1 N sodium hydroxide, nitrazine yellow being used as indicator. The ratio between the norrnalities of the acid before and after dilution is the effective dilution factor. When the desired value is attained the surplus length of capillary is removed.It is necessary to use a high-quality stopcock, such as DOVC/B/B, obtainable from G. Springham & Co., Harlow New Town, Essex. Wire racks-To contain thirty-six 6-inch x 1-inch test-tubes. Test-tubes, 6-iutch x 1-inch-Heated to 150" C, after washing, to destroy residual enzyme. Technicon A utoA nalyzer-This instrument comprises- (a) A sampling plate.332 GOODALL AND DAVIES : AUTOMATIC DETERMINATION OF PENICILLIN [Vol. 86 (b) A pump and a manifold carrying the sample line (0.073 inch i.d.), two air lines (0.081 inch id.) in parallel, an iodine line (approximately 0.073 inch id.) of silicone rubber and a starch line (0.073 inch id.) arranged as shown in Fig. 3. (c) A glass reaction circuit constructed as shown in Fig. 3. (d) A colorimeter unit having a 10-mm flow cell of modified construction such that the volume of the cuvette is decreased to 1.2 nil.The cuvette is a piece of flattened tube and does not require plane polished windows. A pair of interference filters trans- mitting in the range 520 to 600 mp is required. We use 526-12-27 filters; this code signifies maximum transmission (27 per cent.) at 526mp and a band width of 12mp at half the peak height. A recorder with a “linearised” scale. (e) Polystyrene sample cups. Chart reader-A 16-inch x 22-inch Simplon drawing-board having a movable transparent horizontal rule and two clips (obtainable from Messrs. T. Holden, Albert Square, Manchester 2). The rule is engraved, after purchase, in 1- and 0-1-inch divisions over 20 inches. REAGENTS- Iodine solution, apj5roximately 0.64 micro-equivalent per ml-Place 1200 ml of distilled water in an amber-glass bottle, add 7.7 ml of a 01.1 N solution of iodine in 2 per cent.w/v potassium iodide solution, and shake. Starch - potassium iodide solution-Add a suspension of 26.4 g of soluble starch in 200 ml of water to 1800ml of boiling water; stir during; the addition. Dispense 200-ml portions of the solution into ten 20-02 bottles, and sterilise with steam at 5 lb per sq. inch for 30 minutes. Each day, dilute 200 ml of this solution with 1 litre of distilled water, add 1.44 g of potassium iodide, and keep in an amber-glass bottle. Phosphate bufer solution, 0.5 M (PH 6.5)-Disslolve 39 g of anhydrous sodium dihydrogen orthophosphate and 35.5 g of anhydrous disodiuni hydrogen orthophosphate in water, and dilute to 1 litre.Phosphate bufler solution, 0-01 M (PH 66)-1Mute 20 ml of the 0-5 M phosphate buffer solution to 1 litre with distilled water. Penicillinase solutiort-Dissolve the contents of one vial of penicillinase (inactivation rate in 20ml is 20,000 units of penicillin G per minute; obtainable from the Distillers Co. (Biochemicals) Ltd., Speke, Liverpool) in 250 ml of the 0-01 M phosphate buffer solution. Acetate bufer solution, 0.5 M (PH 4.6)-To 500 ml of N acetic acid add 227 ml of N sodium hydroxide, and mix. Acetate bufer solution, 0.05 M (PH 4*6)-Dilute 100 ml of the 0.5 M acetate buffer solution to 1 litre. Standard phenoxymet~yl~enicillin solutions-Prepare a standard solution containing 50 mg of potassium penicillin V in 500ml of distilled water; this solution can be kept for 1 week at 5” C.Each day, prepare standards containing 2, 3, 4, 5, 10 and 15 pg of penicilloic acid per ml as follows. By pipette, place accurately measured 2-, 3-, 4-, 5-, 10- and 15-ml portions of the solution in separate 100-ml calibrated flasks, hydrolyse by adding 3.0 ml of N sodium hydroxide to the contents of each flask, and, after 10 minutes, adjust the pH of each solution to approximately 4.6 by adding 6.0ml of N acetic acid. Dilute the contents of each flask to the mark. Mix equal volumes of the solutions containing 10 and 15 pg per ml to provide a standard containing 12.5 pg per ml. These solutions comprise the penicilloic acid standards. ESTABLISHMENT OF BASE LINE- Begin to pump the iodine and starch solutions, switch on the colorimeter, recorder and chart-drive, and record for about 30 minutes with the sample line empty before beginning the determinations.(The chart reading should be within the limits 6 to 8; if it is not, suitably adjust the concentration of the iodine solution.) PREPARATION OF “TESTS” AND “BLANKS”- “BZanks”-Use one of the compound dilution pipettes to take a sample of the broth and to dilute it with 0.01 M phosphate buffer solution; shake the tube used to receive theMay, 19611 I N FERMENTATION BROTH: AN IMPROVED IODIMETRIC ASSAY 333 diluted solution. Use another compound dilution pipette (A) to sample this solution and to dilute it with 0.05 M acetate buffer solution. Each “blank” so prepared by two serial dilutions is now ready for loading on to the sample plate.Fill the plate with “blanks” as follows. Rinse a polystyrene cup with the diluted sample, fill the cup, insert it in the plate, and suitably mark the aluminium surface beside each sample with washable coloured pencil or ink. Load the plate in this order: buffer solution, standard solutions containing 2, 3, 4 and 5 pg of penicilloic acid per ml, thirty “blanks,” buffer solution and, finally, the four standard solutions. Remove the feed tube from the cuvette, flush the cuvette with water, and check that the recorder reads 100; if it does not, adjust the “100 per cent. T” potentiometer. Replace the feed tube, insert the polythene sampling tube through the crook, and throw the sampling switch to ON.^' Prepare as many plates of “blanks” as are required for the day, and record all of them before the “tests” (see below).While the plates of “blanks” are running, at the rate of forty per hour, prepare the “tests” as described below. “Tests”-Use the third compound dilution pipette to take a sample of broth and to dilute it with the penicillinase solution, shake the receiving tube, and set it aside for a t least 30 minutes. Then use pipette A to sample this solution and to dilute it with 0.05 M acetate buffer solution as before. Load the samples and standards on a plate in this order: buffer solution, standard solutions containing 5, 10, 12.5 and 15 pg of penicilloic acid per ml, “tests,” buffer solution and, finally, the four standard solutions. (Omit the first set of standards on subsequent plates.) When the “blank” recordings are complete, remove the plate, and replace it with the plate carrying the “tests.” Check the “100 per cent.T” setting as before, and begin to sample the “tests” a t the rate of forty per hour. CLEANING THE SYSTEM- The object of running all the “blanks” first is to avoid interaction with traces of residual enzyme from the “tests.” At the end of the day, or next morning before further “blanks” are recorded, clean the system by pumping water for 5 minutes, N hydrochloric acid for 10 minutes and then water again for 5 minutes through all the liquid-conveying lines. CALCULATION OF RESULTS- Clip the chart of the “blank” records on the Simplon drawing-board, align it horizontally, and place the transparent plastic scale over the peaks for the standard solutions.With use of a calibration incorporating the dilution factor and the factor 1.53 (to convert micrograms of potassium phenoxymethylpeniciln to units), construct a graph on the chart paper relating potencies of the standards (in units per millitre) to mean chart readings. Maintain the scale in horizontal alignment, slide it over each “blank” record until the leading edge covers the top of the peak, and read the appropriate potency at the point where the scale cuts the graph. When the “tests” have been recorded, use the standards recorded on this chart to construct a graph from which the “test” potencies can be read in a similar manner. Deduct the “blank” from the “test” potency, and record the result as the “net potency” of the sample.(Omit the first set of standards on subsequent plates.) RESULTS AND DISCUSSION OF THE METHOD Typical test and blank recordings are shown in Fig. 4, the appropriate standards being at the left-hand side of each record; the calibration graphs plotted resemble that in Fig. 2. Results for samples of experimental shaken-flask fermentations for phenoxymethylpenicillin are shown in Tables I and 11. Similar results have been obtained for the assay of benzyl- penicillin fermentations. Results by the manual and automatic methods (turntable rate, forty samples per hour) are compared in Table I, which covers a potency range of approximately 2500 to 5000 units per ml. Variation in the amount of iodine absorbed by the “blank” clearly has a significant effect on the net potency; compare, for example, the results for samples Nos. 11 and 18.To calculate the standard deviation of the method, including dilution, and plate-to-plate variations in standards, “blanks” and “tests,” replicate determinations on twenty experi- mental broths were made, a fresh plate being used for each set. (All these samples were highly turbid, because of dispersed oil; the dilution, however, overcame interference from334 GOODALL AND DAVIES : AUTOMATIC DETERMINATION OF PENICILLIN [Vd. 86 TABLE I: COMPARISON OF RESULTS BY AUTOMATIC AND MANUAL Sample No. 1 2 3 4 6 6 7 8 9 10 11 12 13 14 16 16 17 18 19 20 21 22 23 24 26 “Test” potency, ‘ units per ml 4430 3550 4720 4920 4270 6380 6020 4680 6020 4260 6660 3430 4600 4220 6260 4490 4660 6660 6740 3840 4640 6640 6370 4620 4160 Mean 4746 ‘Blank” potency, units per ml 1280 1080 1260 1420 1270 1220 1040 1420 1200 1470 1280 1160 1360 980 1300 1080 1370 970 1270 960 1710 1160 1300 1490 1260 1251 Net potency, units per ml 3150 2470 3470 3500 3000 4160 3980 3160 3820 2790 4270 2280 3140 3240 4960 3410 3190 4690 4470 2890 2830 4490 4070 3130 2890 3494 METHODS Net potency by manual method, units per ml 3120 2205 3510 3450 2795 4055 4085 3075 3665 2686 4226 21 60 3090 3370 4705 3420 3000 4600 4395 2706 2890 4410 4180 2965 2796 3422 this source.) Aliquots from each sample were diluted, and the “blanks” were determined as described above.Fresh aliquots of each sample were diluted with enzyme solution, and, after 1 hour at room temperature, the “test” determinations were made.The total sequence was repeated, and the second plate of “blanks” was analysed before the first plate of “tests” to avoid reaction with any residual enzyme introduced into the lines at the “test” stage. Table I1 shows the results and Table I11 the statistical analysis. The standard deviation of 75 units can be expressed as 2.1 per cent. of the mean net potency of 3558 units per ml. TABLE 11: REPLICATE RESULTS FOR EXPERIMENTAL FERMENTATION BROTHS Net potency found in- Sample No. 1 2 3 4 6 6 7 8 9 10 11 12 13 14 16 16 17 18 19 20 fist series of assays ( A ) , units per ml 3690 2860 3360 3806 3866 3686 3636 3220 3536 3666 3120 3760 3996 3936 3656 3570 2950 3890 3555 3216 second seriis of assays (B), unils per ml 3615 2986 3480 3946 3905 3760 3690 3096 3630 3695 3260 3850 4060 3946 3630 3780 2766 3870 3700 3216 Difference umts per ml 26 126 120 140 60 76 (B - A ) , - 46 - 126 96 30 140 90 66 10 75 210 - 186 - 20 146 0May, 19611 IN FERMENTATION BROTH: AN IMPROVED IODIMETRIC ASSAY 335 This deviation includes a small bias (51 units between means) on the two sets of repeat determinations, which is the sum of the following deviations.The means of blanks 1 to 20 were (A), 1891 and (B), 1870 and of tests 1 to 20 were (A), 5424 and (B), 5453. This leads to a net positive bias in the (B) series of 21 + 29 = 50 units over the (A) series. Our working experience is that a standard deviation in the range 2 to 2.5 per cent. can be expected in the routine analysis of the type of broth already described. Tests with pure crystalline penicillinv have given results within 1 per cent. of the theoretical. TABLE 111 STATISTICAL ANALYSIS OF RESULTS IN TABLE I1 Sum of net potencies found- First series of assays, units per ml . . .. .. .. .. Second series of assays, units per ml . . .. .. .. Sum of differences (B - A ) , units per ml . . .. .. .. First series of assays, units per ml . . .. .. .. * . Second series of assays, units per ml . . .. .. .. Mean for both series. units per ml . . .. .. .. .. Sum of squares of differences, I: (B - A)' . . .. .. .. Mean net potency found- .. .. .. .. .. .. .. 1 x 223,350 40 Variance, 70,655 71,675 1020 3533 3584 3558 223,350 5583.75 Standard deviation, units .. .. .. .. * . .. 74.7 Standard deviation, as percentage of mean net potency for both series 2.1 Preliminary experiments in the same system at the higher rate of sixty per hour also We thank Mr. M. Gent, now of the Bradford Institute of Technology, for carrying out show promise. the statistical analyses and for helpful discussions. REFERENCES 1. 2. 3. 4. 5. Niedermayer, A. O., Russo-Alesi, F. M., Lendzian, C. A,, and Kelly, J. M., A n d . Chew., 1960, Pan, S. C., Ibid., 1954, 26, 1438. Green, N. C., and Monk, P. R., Chem. G. Ind., 1959, 1210. Bethel, M. M., and Bond, C. R., Analyst, 1961, 86, in the press. Ferrari, A., Russo-Alesi, F. M., and Kelly, J. M., Anal. Chem., 1969, 31, 1710. 32, 664. Received Octobev llth, 1960