JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1991 VOL. 6 1 I5 Determination of Low Concentrations of Lithium in Biological Samples Using Electrothermal Atomic Absorption Spectrometry* Barry Sampson Department of Chemical Pathology Charing Cross Hospital Fulham Palace Road London W6 8RF UK A method is described for the determination of endogenous lithium concentrations in serum and urine using electrothermal atomic absorption spectrometry. The method has also been applied to the analysis of nanolitre samples of micro-puncture fluid from rat kidney tubules containing pharmacological concentrations of lithium. The graphite tubes are coated in situ with tantalum to give improved sensitivity and increased pre-atomization stability. A mini-flow of gas at a flow-rate of 30 ml min-l is used during atomization to eliminate background interference.Background correction is not necessary. The characteristic mass is 0.98 pg of lithium. Serum samples are de- proteinized with an equal volume of 10% nitric acid before analysis. Urine samples are diluted 5-fold with 5% nitric acid. The normal range of serum lithium is up to 0.39 pmol I-' and normal urinary excretion is up to 9 pmol over a period of 24 h. Keywords Lithium; electrothermal atomic absorption spectrometry; tantalum coated tubes; serum; urine Lithium is used clinically to treat manic depression the thera- peutic concentration in plasma being 0.4-0.8 mmol 1-I.\ At this level the determination of lithium presents no problem and most clinical laboratories measure lithium routinely using air- propane flame emission spectrometry or air-acetylene flame atomic absorption spectrometry (FAAS).Ion-selective elec- trodes for lithium are also available. Naturally occurring concentrations of lithium are far lower than those used therapeutically (given above). The concensus from recent publications is that the normal serum concentra- tion is less than 1 pmol l-1.24 The flame techniques referred to are not sufficiently sensitive at this concentration. Standard flame photometers incorporating a fixed ratio pre-dilution of the sample with a reference solution of potassium or caesium can achieve a detection limit of 3&50 pmol 1-I. Flame atomic absorption spectrometry can give a detection limit for lithium in aqueous solutions of about 2 pmol 1-I but for real samples matrix effects raise this limit considerably.Flame emission spectrometry using a dinitrogen oxide-acetylene flame is much more sensitive2 and can rival electrothermal atomization in detection limits but is also subject to matrix effects with real samples. Flame assays also require comparatively large volumes of sample. Laboratories involved in clinical studies may be reluctant to install a dinitrogen oxide flame system for which there is little other application in clinical analysis. In- ductively coupled plasma mass spectrometry has also been used for low level lithium assays in clinical samples.6 The determination of lithium at these low concentrations is needed for several areas of study in medicine pharmacology and physiology. Examples from this laboratory include studies of the physiology of renal clearance of lithium in animals and humans and clinical studies.Animal studies include the deter- mination of lithium in nanolitre samples of kidney tubular fluid obtained by micro-puncture in experiments to study the renal handling of lithium in rats. This technique involves the collec- tion of tubular fluid \*;a a sharpened micropipette inserted into a superficial nephron in the exposed kidney of an anaesthetized animal. Analysis of the fluid obtained can provide valuable in- formation about renal tubular fun~tion.~ Renal clearance of lithium is used as a marker for proximal sodium and water re- sorptionx.' in the kidney. In clinical studies clearance of exoge- nous lithium at near pharmacological doses (serum concentration of up to 0.3 mmol 1-I) has been used however there is evidence that lithium may have dose related effects on * Presented at the Fifth Biennial National Atomic Spectroscopy Sympo- sium (BNASS) Loughborough.UK 18th-20th July. 1990. sodium clearance. It has been suggested that measurement of the clearance of endogenous lithium may give equally valid results without the need for the administration of a pharmaco- logically active susbtance.' A further application is the study of the absorption of lithium from a topically applied ointment containing lithium succinate that is used to treat skin condi- tions;IoJ1 such treatment has been applied to the control of uraemia induced pruritis.I2 Measurements of low lithium con- centrations have also been used to monitor industrial exposure to lithium resulting from the use of lithium alloy^.^"^^ Electrothermal atomization techniques should provide a sen- sitive assay but there are severe matrix interferences notably from inorganic components of the sample.s The addition of ammonium nitratels and/or potassium phosphateI6 have been used to minimize these intereferences and to stabilize the lithium during the pre-atomization step; however in most in- stances matrix-matched standards have been required. Lithium carbide formation is also a problem in the furnace.Use of tan- talum foil a tantalum boat or a tantalum carbide coated tube have been proposed as possible solutions.s Published methods for using coated tubes have normally used tubes pre-soaked in a solution of the coating metal often dissolved in hydrofluoric acid at reduced p r e ~ s u r e .~ . ~ ~ A recently published method for the determination of lithium in micro-puncture samples in- volved direct deposition of up to 6 nl of sample onto a tanta- lum platform which was then inserted into the furnace.'* Background correction has also been required in most assays. The method presented here involves coating of the tube in situ with a water soluble salt of tantalum eliminating the need for hydrofluoric acid solutions. Background correction is not required. Experimental Apparatus All work was performed with a Perkin-Elmer 3030 atomic ab- sorption spectrometer with an HGA 600 furnace and AS-60 autosampler. For background absorption studies a hollow cathode lamp with neon fill gas was used.Perkin-Elmer pyro- lytic graphite coated graphite tubes were used for all experi- ments. Argon was used as the purge gas except where oxygen was used as indicated. Results and peak plots were recorded on a Perkin-Elmer PR 100 printer. Atomic emission spectrometry of samples containing high concentrations of lithium was performed using an IL 943 flame photometer with a diluent containing caesium.116 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1991. VOL. 6 Reagents A lithium standard solution 1 mmol I-I (IL flame photometer standard with I40 mmol 1-1 of Na and 5 mmol I-' of K) was diluted to a stock standard of 40 pmol I-' in 1% v/v nitric acid (BDH Poole Dorset UK Aristar grade). Working standards were prepared as required by dilution with de-ionized water.Ammonium nitrate and ammonium phosphate modifier solu- tions were prepared by mixing equimolar amounts of ammonia solution and nitric acid or ammonia solution and phosphoric acid ( 2 + 1). All reagents were BDH Aristar grade. Triton X- 100 (BDH scintillation grade) was added to the modifiers as indicated. In the initial experiments the tantalum modifier solution was tantalum AAS standard ( 1 g I-' as ammonium hexafluorotantalate in water Aldrich Gillingham Dorset UK). A revised method was later adopted using a saturated solution of ammonium heptafluorotantalate (about 50 g I-' of tantalum) (Aldrich) was used. Other reagents used were of Aristar grade or equivalent where possible otherwise the purest grade available was used. Sample Collection Blood samples were collected in cleaned 10 ml polystyrene tubes or 10 ml glass Vacutainer tubes (Becton-Dickinson Toronto Canada) and allowed to clot.The serum was separat- ed into 3 ml polystyrene tubes and stored frozen (-20 "C) until analysed. No difference was observed in the results from the different tube types and no lithium contamination was detect- ed. Twenty-four hour urine samples were collected in 4.5 1 polyethylene bottles without any added preservative. Aliquots were transferred into 25 ml polystyrene universal containers and stored frozen. The final step in the preparation of the serum samples was to precipitate the proteins with an equal volume of 10% v/v nitric acid. Urine samples were diluted 5- fold with 5% v/v nitric acid. Micro-puncture samples from rat kidneys were collected with glass constriction pipettes of 3&70 nl capacity and diluted with 50 p1 of de-ionized water in polystyrene microvi- als (400 p1 capacity) for direct use in the furnace autosampler. Samples were stored frozen until analysed.The dilution of these samples was at least 1000-fold and they were treated as pure aqueous solutions for analysis. Table 1 flow of 300 ml min-' was used throughout the programme Furnace programme for tantalum coating of tubes. An argon gas Hold/s Step TemperatureTC Ramp/s I * 2* 3* 4 80 120 600 2500 20 I 99 10 5 10 I 5 * 75 or 99 pl sample volume injected and programme steps 1-3 repeat- ed up to ten times before step 4. Table 2 analysis Instrumental parameters and furnace programme for sample Instrumentul puiumeters- Wavelength 670.8 nm Slit width 0.7 mm Measurement Integrated absorbance Integration time 5 s Sample volume 25 pl Replicates 2 Furnace progiumme- Step TemperatureTC Ramp/s Hold/s rate/ml min-' Read Internal gas flow- - 1 90 1 1 300 2 120 20*/30t 5 300 - 300 1*/5t 5*/10f 3 1250 4 20 1 10 300 5 2100 0 5 30 On 6 2400 I 5 300 - - - * Water and micro-puncture samples. t Serum and urine samples.Table 3 pmol I-' lithium standard Effect of tantalum coating on the signal from 20 p1 of 1 Amount of tantalum/mg Peak area/A s 0 0.1 I .0 5.0 0.035 f 0.002 0.280 f 0.004 0.290 f 0.004 0.300 k 0.003 Results and Discussion Method Development It was found that careful alignment of the furnace and selec- tion of atomization temperature was needed to minimize inter- ference from emission from the hot graphite tube during atomization.It is also important to use the lowest possible atomization temperature in order to minimize emission. The atomization characteristics of lithium were studied in aqueous solutions with and without the addition of chemical modifiers. Without any additions it was found that the maximum pre-atomization temperature possible was less than 750 "C. An atomization temperature of 2100 "C was used as this gave the optimum compromise between analytical sensi- tivity and low emission noise. While the addition of phosphate results in increased stability of the lithium and can allow the use of higher pre-atomization temperatures it can also contrib- ute to non-specific absorbance. The peak shape was poor; con- siderable tailing was observed on numerous occasions as the absorbance had not returned to the base line within 10 s of atomization.Tubes were coated with tantalum by replicate injections of 99 p1 of the 1 g I-' modifer solution with a furnace pro- gramme drying the solution between each injection at the tem- peratures indicated in Table 1. About 1 mg of tantalum was deposited in the tube. The tube was taken through several com- plete programmes (Table 2 ) until no further signal was record- ed. Both the phosphate modifier and tantalum coating allow the use of higher pre-atomization temperatures; the phosphate modifier gives a similar sensitivity compared with untreated tubes and the tantalum coating provides a higher sensitivity. The benefit of tantalum coating is shown in Table 3.Uncoated graphite tubes were used in some early experiments but showed a low sensitivity. Tantalum coated tubes were used for all subsequent work. By using the saturated solution of tanta- lum slightly improved results for sensitivity were obtained and the sensitivity remained constant for longer. The tube coating process was also much faster. Two replicate injections of 75 1.11 were used. By using this technique the characteristic mass (0.0044 A s) is 0.98 k 0.14 pg (ten determinations). Despite the improve- ment shown with tantalum tailing can remain a minor problem and with high-concentration samples at least 5 s are required for the signal to return to the base line. Micro-puncture Samples Micro-puncture samples can be analysed by use of the method developed up to this point.As stated above it was considered possible to treat the samples as pure aqueous solutions and they were assayed against aqueous standards. Errors may ariseJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY. MARCH 1991 VOL. 6 117 Table 4 Accuracy of micro-puncture sample analysis Concentration calculated/ Concentration found/ mmol I-' mmol I-' 0.42 0.36* 0.72 0.62 0.9 1 0.84 1.37 1.41 1.87 I .86 * Mean of two samples. Table 5 Effect of internal gas flow-rate during atomization on lithium absorbance measured as peak area and background signal. Sample 30 pl of undiluted normal urine; and lithium concentration approximately 1.5 pmol 1-I Internal gas flow-rate/ml min-' LithiumlA s BackgroundtA s 300 200 100 50 30 10 0 0.14 - 0.I9 - 0.24 - 0.28 0.004 0.30 0.007 0.31 0.018 0.3 1 0.027 in these experiments owing to uncertainty in the exact volume of the constriction pipettes which were made in the labora- tory. Although accuracy could not be assessed from recovery experiments because of the small volumes involved it was as- sessed in two alternative ways. One method was to prepare di- lutions of standard solutions in the same manner as samples using the same constriction pipettes. The results are shown in Table 4. Accuracy was also determined by comparing results obtained by analysing serum samples containing pharmacolog- ical amounts of lithium after a 1000-fold dilution (Fig. 1). Within-batch precision of the micro-puncture fluid assay was assessed by repeated injection of a single sample into the furnace.At all concentrations the within-batch precision found was 1-2%. The detection limit (3 x standard deviation of the blank) is 0.04 Fmol 1-1. Serum Samples As the deuterium arc background corrector cannot function at the long wavelength (670.8 nm) of lithium the presence of background absorption was assessed by measuring absorbance at the non-resonance line of 671.6 nm obtained from a neon filled hollow cathode lamp. For this study urine diluted with an equal volume of 10% nitric acid was used. Using gas-stop during atomization was found to give a rapid but significant smoke peak. Introducing a mini-flow of argon gas during atomization reduced this to negligible amounts (Fig. 2) but had a smaller effect on the lithium atomization signal (Table 5).A gas flow-rate of 30 ml min-I was found to be adequate. For the determination of lithium in serum attempts were first made to assay samples after aqueous dilution. A sample dilution of 1 + 1 with 0.1% v/v Triton X-100 and a 30 pl sample volume gave adequate sensitivity. Oxygen ashing was introduced to minimize the build-up of carbon residues in the tube however this was found to cause a decrease in the life- time of the tube and marked changes in sensitivity were ob- served during a batch of 30 samples. The oxygen ashing step was felt to be the major contributor to the problems encountered. A modification which involved de-proteinizing the samples with an equal volume of 10% nitric acid containing 100 mg 1-I tantalum was introduced. The tantalum was added to replenish the tantalum coating on the graphite tube surface otherwise it was found that the sensi- tivity decreased during a run although the tube life was greater than 400 firings.In subsequent work when a higher density of tantalum coating (up to 5 mg per tube) was used the c '- 2.0 3 0 0.4 0.8 1.2 1.6 2.0 Li concentration (AESVmmol I-' Fig. 1 Comparison of assays by atomic emission spectrometry (AES) and electrothermal atomic absorption spectrometry (AAS) for samples with pharmacological concentrations of lithium. Samples for AAS were diluted 1 + lo00 with water before assay. The equation of the line is given by v = 1.03s - 2.85 x correlation coefficient ( R 2 ) = 0.94 0.05 (a) 0.025 - 0 5.0 Time/s Fig. 2 Comparison of background absorbance (671.6 nm) for urine with (a) internal gas flow-rate of 30 ml min-' or (h) gas stop during atomiza- tion.Sample 30 pl of undiluted normal urine; and Li concentration ap- proximately 1.5 pmol I-' loss of sensitivity was found to be reduced and tantalum was no longer added to the de-proteinizing solution. The furnace programme used is shown in Table 2. The slopes of standard additions lines for de-proteinized serum were similar to aqueous standards and aqueous calibra- tion was used. Recovery experiments using aqueous standards gave good recoveries (Table 6). Within-batch precision for normal serum was 7.8% at a concentration of 0.15 pmol I-'. Between batch precision using the same serum was 19% ( n = 7). Urine Samples For the initial studies on urine samples the same methods as for serum samples were used however at a 1 + 1 dilution there was a marked variation in the standard additions slopes.By using a higher dilution and/or smaller sample volume this variation was reduced but remained significant for some urine samples. The reason for the high variability has not been in- vestigated but is thought to be due to the calcium and/or phos- phate content. Urine samples were assayed at a 5-fold dilution with 5% nitric acid and a recovery sample was included with118 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 199 I VOL. 6 Table 6 Recovery of lithium from serum and urine Lithium added/ Lithium found/ Sample pmol I-' pmol I-' Recovery* (96) Serum 0.5 0.47-0.58 101 f8 1 .o 0.9 1-1.16 101 f 10 2.0 I .76-2.05 9 5 f I I Urine 1 .o 0.90-1.07 98 f 8 *Recovery f standard deviation n = 7.Table 7 Results of assays on clinical samples Lithium/pmol I-' Sample Median Range Serum- Normal subjects 0.17 0.05-0.39 (n = 19) Chronic renal failure 0.53 0.39-1.28 (n = 10) Normal subjects 1 S O 0.264.9 (n = 13) Urine- 3.8 0.5-9. I (pmol per 24 h) each urine. If the recovery was low the assay was repeated using standard additions. The furnace programme used was the same as for serum samples. Recovery data for the urine assay are given in Table 6. Within-batch precision for normal urine at a concentration of 2 pmol 1-i was 2.5% and between-batch precision was 9.8%. Clinical Samples Serum samples from normal subjects and from patients with chronic renal failure treated by haemodialysis or peritoneal dialysis and urine samples from normal subjects (Table 7) were analysed.Most normal subjects have serum lithium levels not much above the detection limit of the assay. Patients with chronic renal failure have higher concentrations but there is no clinical significance associated with this. The urine lithium ex- cretion in normal subjects is variable and may be related to dietary intake." Conclusions The use of tantalum coated tubes is shown to give an en- hanced sensitivity for lithium determination in the electrother- mal atomizer. In situ coating of the tubes with the water soluble ammonium heptafluorotantalate is rapid and avoids the need for use of hydrofluoric acid solutions. Tantalum coating allows the use of higher pre-atomization temperatures. Use of a mini-flow of argon at a flow-rate of 30 ml min-1 in the furnace during atomization gives a substantial reduction in the residual background absorbance with a minimal effect on assay sensitivity.The assay has been applied to lithium deter- mination in nanolitre volumes obtained from rat kidney micro-puncture experiments and to serum and urine. Serum is de-proteinized with nitric acid before being assayed; urine samples are diluted with nitric acid. This assay may be suit- able for the determination of lithium in samples other than those used here. I wish to thank Drs. S. Walter and D. Shirley Department of Physiology Charing Cross and Westminster Medical School for the micro-puncture samples Prof. G. MacGregor and Dr. D. Singer Department of Medicine St. Georges Hospital Medical School for some of the patient samples and Dr.J. R. Curtis Department of Medicine Charing Cross and Westmin- ster Medical School for samples from patients with chronic renal failure. 1 2 3 4 5 6 7 8 9 10 1 1 12 13 14 15 16 17 18 References Srinavasen D. P. and Hullin R. P. Br. J. Hosp. Med. 1980,24,466. Matusiewicz H. Anal. Chim. Actu 1982 136,215. Miller N. L. Durr J. A. and Alfrey A. C. Anal. Biochem. 1989 182,245. Fridrich L. Zazgomik J. Kopsa H. Schmidt P. Hinterberger W. and Maly K. J. Clin. Chem. Clin. Biochem. 198 I 19,672. Shen L. Shan X.-q.. and Ni Z.-m. J. Anal. At. Spectrom. 1988 3 989. Abou-Shakra F. R. Havercroft J. M. and Ward N. I. Truce Elem. Med. 1989,6 142. Quamme G. A. and Dirks J. H. Kidney Int. 1986,30 152. Thomsen K. Nephron 1984,37,2 17. Thomsen K. Kidney Inr. 1990,37 (suppl. 281 S10. Skinner G. B. Lancet 1983,2,288. Boyle J. and Burton J. L. BI-. Med. J. 1986,292,28. Sampson B. Curtis J. R. Stewart J. C. M. and Cream J. J. Truce Elem. Med.. in the press. Barber D. Braithwaite R. A. and Brown S. S. Proceedings of the National Meeting of the Association of Clinical Biochemists Birmingham UK 15-19 May 1989 A61. Bencze K. Pellikan Ch. and Kronseder A. Arfzl. Luh. 1989 35 102. Ehrlich B. E. and Diamond J. M. Biochem. Biophvs. A m 1978 532,264. Trapp G. A. Anal. Biochem. 1985,148 127. Zatka V. J. Anal. Chem. 1978,50,538. Willis. L. R. Broughton M. C. and Foster R. Kidney Int. 1990,37 575. Paper- Ol03313H Received July 24th I990 Accepted October- 5th 1990