BIOLOGICAL MASS SPECTROMETRY, VOL. 20, 724-730 (1991)

Rapid, Automated Analysis of 13Cand '*O of C 0 2 in Gas Samples by Continuous-flow, Isotope Ratio Mass Spectrometry S. J. Presser,? S. T. Brookes and A. Linton Europa Scientific Ltd, Europa House, Electra Way, Crewe, Cheshire, CW1 lZA, UK

T. Preston Scottish Universities Research and Reactor Centre, NEL Estate, East Kilbridge, Glasgow, G75 OQU, UK

A rapid, automated method for isotopic analysis of I3C and "0 in CO, has been developed. A variety of gas samples containing CO, can be swept from serological tubes into a helium carrier flow; impurities are separated on a GC column so that a pure pulse of CO, in He flows into the mass spectrometer. Isotopic ratio determinations are carried out as the pulse passes through the mass spectrometer, allowing a sample to be measured every 4 min. A double, concentric needle-probe is used to flush the sample from the tube so that 100% sample recovery is achieved, maximizing sensitivity and preventing the possibility of fractionation. The precision of the technique, u(,,-,), is better than 0.2% (0.0002 atom per cent excess) for "C and 0.4% (0.83 p.p.m.) for I8O for 10 pmol of CO, at natural abundance. Samples containing only atmospheric concentrations of CO, can also be analysed.

-

INTRODUCTION Stable isotopes are widely used as tracers in many fields of research and increasingly for routine diagnostic tests in some clinical areas. Of particular importance to biomedical research is the isotopic analysis of CO, because of the large number of processes which can be traced using I3C and "'0 and the easy collection of CO, in the breath.' Historically these types of analyses have always been carried out using either gas chromatography/mass spectrometry (GC/MS) or dualinlet isotope ratio mass spectrometry. The former has the advantage of high sensitivity (only nanomoles of sample are required) but unfortunately the poor precision of the technique (- 100-500%0 or -0.1-0.5 APE)' requires that the quantities of isotope needed for subsequent detection are prohibitively large for many experiments. The dual-inlet isotope ratio mass spectrometer has good precision (some systems are capable of 0.005% or 0.O00 006 APE for samples > 3 pmol) but is very slow-it takes at least 15 min to analyse CO, from each breath The analytical precision of the latter technique is far greater than required by metabolic studies, the minimum acceptable precision being set by the 'biological noise' of "C in breath, which has been estimated as having a variation, '), of -0.4-0.7% for a typical 2-6 h experimental

-

-

Because of the large numbers of samples that tracer experiments generate there is a need for fast, automated analytical systems. Automated systems have been available for some time and have been used in a number of s t ~ d i e s ~ ,but ~ , 'these have always been associated with

t Author to whom correspondence should be addressed. 1052-9306/91/110724-07 $05.00

0 1991 by John Wiley L Sons, Ltd.

conventional, dual-inlet isotope ratio mass spectrometry and have required the use of cryogenic trapping to separate the CO, from non-condensable gases and from water. Apart from being time consuming and costly this technique cannot be used for certain groups of subjects whose breath may be contaminated with gases that would distil with the CO, and interfere at the masses of interest in the mass spectrometer.' Examples of these groups are patients anaesthetized with N,O (m/z 44) and subjects who have a number of organic compounds in their breath (e.g. from ketone bodies or alcohol) that would break down in the mass spectrometer to provide isobaric interferences. An alternative approach is that of gas chromatographic separation of the interfering gases followed by continuous-flow, isotope ratio mass spectrometry (CF-IRMS). Purpose-built systems employing CF techniques for stable isotope tracer studies began to be investigated a few years ago. Initially these were for solid and liquid samples and involved interfacing an elemental analyserg*'* or gas chromatograph" to a lowresolution, multiple-collector, isotope ratio mass spectrometer. More recently a similar technique was used for analysis of I3C in CO, where breath samples were manually injected into the carrier flow of an elemental analyser linked to a mass spectrometer;I2 in this study it was again stressed how desirable full automation would be because of the large number of samples. EXPERIMENTAL Instrumentation Breath samples are collected and loaded into preevacuated, septum-capped, serological tubes using a method similar to that used by Schoeller and Klein.13 Received I3 M a y 1991 Revised 8 August 1991

RAPID, AUTOMATED ANALYSIS BY CONTINUOUS FLOW IRMS HI

sc

Figure 1. Operation of the concentric, double needle-probe. EX, 13 ml phial, Exetainer; HI, helium carrier gas in; HO,helium carrier and sample gas out; NP, concentric, double needle-probe; SC, screw cap; SP, septum.

Two types of collection tube have been used routinely; the 20 ml Vacutainer (Becton and Dickson, Rutherford, New Jersey, USA) and the 13 ml Exetainer (Labco, High Wycombe, UK). We found during this work that the Exetainer is a more suitable container for our new technique for a number of reasons (to be published elsewhere) but the Vacutainer was also used as it has become something of an 'industry standard' for collection of gases for isotopic analysis. The containers are loaded into a 220-place, modified autosampler (Sample

725

changer 222, Gilson Medical Electronics SA, Villiers-leBel, France), where the sample gas is transferred to a carrier flow of helium gas using a double, concentric needle (see Fig. 1). A full schematic of the system is shown in Fig. 2. Prior to inserting the needle into the container, any air, or traces of the previous sample, are flushed out with helium for a few seconds by opening valves V1 and V2 (see Fig. 2). The needle is then driven through the septum and valve V3 is closed and valve V2 opened, allowing the carrier flow of helium (-90 ml min-') to flow through the container and back into the carrier line. In this way sample gas is flushed out of the container; it then passes through a drying tube (MgClO,) to remove water vapour and to a packed column G C (18 inches long, $ inch o.d., Carbosieve-G, 60-120 "C depending on application) to separate the CO, from other gases. Sample N, and O2 elute from the column first and are bled to waste via V5, whilst a standby flow of pure helium (-20 ml min-') is bled into the mass spectrometer inlet via V4. As the CO, elutes valves V4 and V5 are opened so that the standby flow goes to waste and the carrier flow goes to the mass spectrometer. The purified pulse of C 0 2 is carried past the mass spectrometer inlet, where a small proportion is bled into the ion source of the mass spectrometer. If any N 2 0 or organic gases are present in the sample they elute later and may also be bled to waste, though as they are well resolved from the CO, they may be monitored by the mass spectrometer without interfering with the isotopic measurement. The standby flow of helium is necessary to maintain steady ion-source conditions whilst the carrier flow is switched away to prevent unwanted gases from entering the mass spectrometer.

Figure 2. Schematic of CF equipment. EX, 13 ml phial, Exetainer; FC, flow controller; FS, flow switch; FM, flow meter; GC, gas chromatograph column; NP, concentric, double needle-probe; NV, needle valve; PG, pressure gauge; RG, pressure regulator; Vl-V5, solenoid valves; W, bleed to waste; WT, water scrubber.

726

S. J. PROSSER, S. T. BROOKES, A. LINTON AND T. PRESTON

Flow switches (FS, Fig. 2) are provided as a quick means of switching off the helium flows when the instrument is not being used. The flow controllers and needle valve could be used for this purpose but this would require fiddly resetting of the flow rates when next using the instrument. The flow switches allow the helium flows to be switched immediately to their previous rates. Isotopic analyses are performed on an 11 an radius, standard geometry, triple-collector mass spectrometer (Tracermass, Europa Scientific, Crewe, UK). The triple collection facility is necessary in order to measure m/z 44, 45 and 46 simultaneously so that both the 13C/'2C and 180/'60 ratios can be calculated for the COZ;l4 peak jumping is not possible for C F IRMS because of the necessity of real-time measurement as the peak elutes. During sample analyses, reference gases are analysed at the beginning of each batch run and every five to ten samples thereafter; in this way any instrumental drift is corrected for and a check on instrument performance is maintained. Reference gas is analysed in exactly the same way as sample gases; an amount of CO, of known isotopic composition and partial pressure is injected into the same type of containers as used for the samples and these are placed, along with the samples, in the autosampler in the correct sequence. The advantages of this system are that: (1) the reference gas is treated in exactly the same way as a sample so that instrumental effects should then cancel out; and (2) it is simple to choose, or make up, a reference gas that is suitable for a particular experiment, i.e. amount and isotopic enrichment of CO,; for instance 5% CO, in N, is used for a reference gas against breath samples.

Methods A number of samples from different experiments have been analysed on the C F system to test its suitability for biomedical studies. In order to test the accuracy of the system, experiments that showed a wide range of isotopic variation were selected. "C breath C02 samples To investigate the performance breath samples from a (13C)urea breath test designed to test for the presence of Helicobacter pylori were analysed. This test was first used by Graham et al." and has more recently been standardized by Logan et ~ 1 . 'After ~ a high-fat meal a patient was given 100 mg of (13C)urea (99% 13C, Tracer Technologies Inc., Sommerville, Massachusetts, USA). A baseline breath sample was taken before the administration of the dose and subsequent samples taken every 10 min thereafter. Subjects blow into a 750 ml alveolar breath collection bag (Quintron, Milwaukee, Wisconsin, USA) through a non-return valve. Samples are withdrawn from the collection bag using a plastic 20 ml syringe and transferred to the sample container.

water samples. These were prepared by diluting 10% '80-labelled H,O with tap water and were calibrated against the international water standards 302-A and 304-A (IAEA, Vienna, Austria). "0 in aqueous samples is measured by equilibrating the fluid with CO, ,

using a technique similar to that of McMillan et al.," 1 ml of fluid is equilibrated with 12 ml of 5% CO, in N, in an Exetainer. The equilibration, carried out at a steady, fixed temperature, can be performed either by leaving the tubes on their sides overnight or by agitating for four hours in a shaker bath. From isotopic analysis of the CO, it is possible to deduce the original "0 enrichment of the fluid using a simple correction procedure,14 to take account of the fractionation between the vapour and liquid phases, modified slightly so that any convenient equilibration temperature may be used." "0 breath CO, samples contaminated with N,O. The system was also investigated for measuring breath CO, when contaminated with N,O, which forms ions of the same mass as CO,. Natural abundance ratios of m/z 45/44 and 46/44 are very different for CO, and N,O, so only a small amount of N,O contamination has a large effect on the measured ratio; it is necessary to completely remove or separate it from the CO, before measurement. To illustrate this effect a patient who had recently undergone surgery was given a drink of 10% 180-labelled H,O. Breath samples were subsequently collected and transferred to duplicate containers. One batch of samples was analysed with the gas chromatograph set at 120°C and a second set with the gas chromatograph set to 60 "C.

Studies of protein turnover using ~-(l-'~C)leucine as a tracer induce only a small increase in 13C enrichment of the total leucine in a tissue protein. Increases are typically in the range 10-40% (613CpDB).Hence, quantitative determinations of 13C increases are best achieved by analysing only the labelled carboxyl group by using the method of Read et al." Three micromoles of L-leucine (natural abundance and I3C-enriched) were dried into Exetainers. The sample containers were put on ice and 1 ml of buffer (9.4 g trilithium citrate (4H,O), 7.4 g citric acid (H,O), 5 ml hydrochloric acid (37%) in 1 litre of water) containing 50 mg ninhydrin (Sigma Chemical Co., Dorset, UK). Sample containers were subsequently evacuated to 1 x lop3 mbar and heated at 100°C for 30 min to liberate CO, from the carboxyl group of the amino acid. Containers were allowed to cool to room temperature and helium (99.998%)added to bring the containers to atmospheric pressure prior to isotopic analysis. Preparation of CO, from amino acid carboxyl groups.

~

~~

~

~~

~

RESULTS AND DISCUSSION

The sampling method employed ensures that the container is kept at a positive pressure during analysis; thus if the septum leaks it is only possible for sample gas to leak out of the phial and not for air to leak in, small amounts of back diffusion being insignificant. A small amount of gas leaking out of the sample container will not change the isotopic signature of the remaining gas to any great degree; however, if air containing CO, at the natural abundance isotopic value were to leak into the container it would dilute isotopically enriched

RAPID, AUTOMATED ANALYSIS BY CONTINUOUS FLOW IRMS

gas and could lead to a substantial error; this effect becomes more damaging at low concentrations of CO, . For this reason, sampling methods that required a partial vacuum in the phial were avoided. The technique of flushing the contents of the container into the carrier flow also has the benefit of ensuring that 100% of the sample gas is recovered. As there is no residual gas there can be no fractionation-the sensitivity of the system is also maximized. To minimize any mixing between the carrier and sample gases it is necessary for the inlet and outlet of the needle-probe to be positioned close to the ends of the sample container (see Fig. 1). The importance of the positioning of these holes is demonstrated in Fig. 3. Figure 3(a) shows the CO, eluting when the inlet and outlet are both placed halfway down the container. It is obvious that a great deal of mixing is taking place in the sample tube and the time required to flush all the sample from the container is very great. Figure 3(b) shows the same sample being

721

flushed through with the inlet and outlet 1 mm from either end of the container (it is identical for flow up or down the tube); it is indistinguishable from Fig. 3(c), which shows a similar amount of C O , being injected manually through a septum purge head into the carrier flow. The tailing apparent at the foot of the peaks of Fig. 3(b) and (c) is due to chromatographic effects and dead volumes in other parts of the system rather than mixing as it appears on both traces-it can be deduced that there is no substantial mixing caused by the method of flushing the container. Any possible fractionation effect during flushing was further investigated by flushing the containers for varying times and recording the yield and isotopic composition-this is shown in Fig. 4. Full recovery seems to take place after only 20 s-at 90 ml min-' this is a little more than one container volume-but there is an isotope effect that continues past this. However, as long as the container is flushed for >60 s (around four

m/z 44

1E-9A

111111111111 44 1E-9A

d Z

b

Illlrlllllll 44 1E-9A

m/Z

C

1 ,1 1 1 1 1 1 1 1 1 1 1 200

100

I

Time (s)

Figure 3. Shape of the CO, peak eluting from the GC column, monitored at m/z 44 at the mass spectrometer, for different sampling methods: (a) shows the peak shape with both inlet and outlet of needle in the centre of the serological tube; (b) shows the peak shape with the inlet at the bottom and the outlet at the top of the tube; (c) shows the peak shape for manual injection of the sample gas through a septum inlet into the carrier flow. All traces are for a sample of 10 ml of 5% CO, in nitrogen.

S. J. PROSSER, S. T. BROOKES, A. LINTON AND T. PRESTON

728

-39.0

-40.0

-41.0

-42.0

-43.0

-44.0

I-

I

I

0

100

200

300

Flushing Time (s) Figure 4. The effect of the length of time of flushing the sample phial on the measured isotopic enrichment and sample yield.

to five container volumes) no fractionation takes place. From this a standard of 100 s flushing period was established. Precision. The general performance of the instrument was established by repeated analysis of reference gases. The results shown in Table 1 are isolated batch runs representative of many such experiments that have been performed on the instrument. With 10 pmol of CO, a better than 0.2% for 13C and 0.4% for precision, CT(""0 is routinely achieved. With 1 pmol of CO, precision falls to -0.4% and 1.5%, respectively. It is pos-

Table 1. Precision of measurement for reference gas Sample

10 p o l CO, (1.72% in 13 ml)

Mean

1 pmol CO,

(0.17% in 13 ml)

Mean

0.2 pmol CO, a (350p.p.m. in 13 ml)

Mean

6'3C (%)

6"O (%)

-41.68 -41.78 -41.45 -41.52 -41.73 -41.63f 0.13

-20.14 -20.51 -1 9.99 -1 9.75 -20.32 -20.14* 0.26

-41.15 -41.a8 -41.43 -41.65 -41.97 -41.62* 0.33

-18.38 -20.15 -1 6.67 -1 8.81 -20.33 -1 8.87* i .49

-41.22 -42.30 -41.98 -41.10 -41.a8 -41.70f 0.52

-22.26 -20.82 -20.28 -21.29 -21.86 -21.30 0.79

"These results were obtained using a modified Tracermass mass spectrometer with differential pumping, allowing greater sensitivity by increasing the flow of gas into the mass spectrometer without a deleterious effect on the abundance sensitivity and hence accuracy of measurement.

sible to measure down to atmospheric concentrations of CO, (350 p.p.m. giving 0.20 pmol in a 13 ml Exetainer) with a routine precision of -1.0% for 13C and -2% for l*O. I3C breath CO, results. Breath samples from a (13C)urea breath test were taken in duplicate and analysed on the system described here and a dual-inlet system using cryogenic trapping (Sira 10, VG Isotech, Middlewich, UK). The results are shown in Table 2. Results from a positive test are used to demonstrate the range of enrichment over which the two systems are in good agreement. There is no evidence of systematic variation between the two systems, nor any worsening in the comparison at higher enrichment. The variation that does exist can be explained by the storage of samples in Vacutainers prior to analysis. The results were obtained analysing the samples against a reference gas with 613C = -41.63960, so the maximum difference between sample and reference gas is 35%.

-

'*O from CO, in equilibrium with water. To test the accuracy for l80 determinations, water standards with a range of "0 enrichment were analysed. The results shown in Table 3 are the 6 l 8 0 values for the water standards, corrected from the value obtained for the

Table 2. Comparison of CF and dual-inlet tecbniques for carbon isotopic composition of breath CO, from ('3C)urea breath test Sample

Baseline 10 min 20 rnin 30 min 40 min 50 min 60 min

613C by CF (%)

-23.38 -1 6.71 -1 0.13 -6.87 -5.88 -1 0.25 -1 2.81

613C by dual-inlet (%)

-23.44 -1 6.81 -1 0.33 -7.60 -6.1a -10.34 -1 2.99

Difference

0.06 0.10 0.20 0.73 0.30 0.09 0.18

729

RAPID, AUTOMATED ANALYSIS BY CONTINUOUS FLOW IRMS

Here the isotope of interest is again l80; an increase in "0 enrichment occurs as body water equilibrates rapidly with alveolar water so that a plateau is reached (time increasing from sample A0 to sample A80). This effect is observed for the experiment where the N,O is separated from CO,; the 13C values are steady at a baseline value, as would be expected. With NzO interference (poor COz/N,O separation on the gas chromatograph at high temperature) the results are meaningless-a reverse effect is apparent for the "0 and the 13C values are now also decreasing with time instead of showing a steady baseline value. These effects are due to the gradual flushing of N 2 0 from the subject's body after cessation of N,O administration. Breath CO, analysis of "0 should be measured as soon as possible after collection. This is facilitated by the rapid throughput of the technique described. Loss of enrichment has been noted after prolonged storage of similar samples.21This can be minimized by storage at low temperature in a vessel that ensures samples only contact polytetrafluoroethene (PTFE)and glass.

Table 3. Comparison of CF and dual-inlet techniques for oxygen isotopic composition of CO, in equilibrium with water Sample

EO E l 00 E200 E500

P O by CF

1%)

6% by dual-inlet (%)

Difference

-29.95 65.44 206.34 479.02

0.01 -0.1 0 0.14 -0.05

-29.94 65.34 206.48 478.97

CO, in equilibrium with it. Again duplicate samples were analysed on the CF and dual-inlet systems and again there is good agreement between the two systems, with no evidence of systematic variation or increasing error with increasing enrichment over a range of 500%. Samples were run against a reference gas of = -22.14% on the CF system and against the international l 8 0 water standard 302-A on the dualinlet system.

-

I8O breath CO, samples contaminated with N,O. Figure 5 shows the separation of N,O from CO, on the gas chromatograph when run at 60°C. By restricting the measurement window to the appropriate times it is simple to measure only the CO, isotope ratios and allow the N 2 0 to elute before analysing the next sample. The necessity of this is shown in Table 4, where data from duplicate breath samples from an anaesthetized subject are presented, first with a column temperature of 120°C and no separation between the CO, and N,O and then with a column temperature of 60°C so that the N,O does not interfere with the analysis.

CO, from ~-(l-'~C)leuciw.The measurement of C 0 2 produced from the decarboxylation of L-( l-13C)leucine requires the ability to measure small samples of CO, . This reaction has already been performed in serological tubes in order to allow automated measurement on a dual-inlet type mass ~pectrometer.~ Because of the small amount of CO, available from this reaction, typically 0.25-5 pmol, it was found to be necessary to use the cold finger arrangement on such a system to enable measurement and special corrections made to overcome problems in measuring small ion-beams. Because of the small sample capability of the CF system it is possible to measure these samples without any special precautions. Results for this reaction using 3 pmol of CO, are shown in Table 5 for baseline and enriched samples. The precision of 0.15% for the baseline and 0.38% for the enriched samples compares well with the results of Read et a1.'' and Scrimgeour et aL7

Table 4. Isotopic composition of CO, in breath contaminated with N,O Sample

A0 A5 A1 0 A20 A40 A80

Good ssparation (GCat 60°C)

Poor separation (GCat 120°C) 6'"C (%) 6'00 (%)

150.43 77.45 61.71 45.03 11.23 -7.57

P C

61'0

(%)

-25.46 -25.26 -25.55 -26.24 -26.03 -24.58

262.98 170.14 151.63 143.25 100.50 66.63

(%)

0.75 14.39 16.74 28.57 42.20 40.45

Summary

The precision of the system is adequate for many biomedical studies whilst the accuracy is demonstrated by

d z 44 4E-9A

I

l

l

l

l

l

l

l

l

l

l

l

I

100

200

300

Time (s) Figure 5. Separation of CO, and N,O from a breath sample from an anaesthetized subject, monitored at m/z 44 at the mass spectrometer. Total baseline separation between the two peaks is -20 s with GC conditions of 60°C and 90 ml min-' carrier flow.

730

RAPID AUTOMATED ANALYSIS BY CONTINUOUS FLOW IRMS

of samples to be undertaken simply and quickly. Using this apparatus 220 samples and standards can be analysed automatically in 18 h. The system is also capable of measuring sample gases other than C0,-nitrogen and its oxides (NO and N,O) have been successfully measured.

Table 5. Isotopic composition of 3 pmol of C 0 2 from decarboxylation of t(113C)Ieucine Sample

Baseline 1 Baseline 2 Baseline 3 Average Enriched 1 Enriched 2 Enriched 3 Average

-

6 3 C (6)

-29.60 -29.62 -29.87 -29.70

5

f 0.1

Acknowledgements

9.01 9.77 9.35 9.38 0.38

*

comparative analyses using conventional dual-inlet mass spectrometry. The full automation and large batch-handling ability, combined with the speed of analysis, allows studies which generate a large number

S. J. Prosser i s a part-time student with the Planetary Sciences Unit, Open University, Milton Keynes, UK. The research was partly funded by the D T I under the SMART (1989) scheme. We are indebted to P. Johnson (BSIA Brentford, Middlesex, U K ) for analysing the duplicate samples on his dual-inlet mass spectrometer. Thanks also to I. Mecrow (Booth Hall Children's Hospital, Manchester, UK) for providing '*O breath samples, and E. Milne (Rowett Research Institute, Aberdeen, UK) for his valuable advice on preparing CO, from ~ 4 1 "C)leucine. Many thanks to G. J. Atkins for the atmospheric CO, sample analyses and A. J. Park for writing the control software for the system (both of Europa Scientific, Crewe, UK).

REFERENCES 1. D. Halliday and M. J. Rennie, Clin. Sci. 63,485 (1 982). 2. %o is delta (6)units expressed in parts per thousand where d = ((Rs/Rr) - 1 x 1000. APE represents atom per cent excess, where APE = (Rs - Rr)/(l - Rs/Rr) x 1 0 0 . Rs is the 13C/12C

("'0/''0)ratio for the sample and Rr is the same ratio for the reference gas. Throughout this work the 6 notation is predominantly used, normalized to the international standard PDB, which has a '%/' 2C ratio of 0.01 1 237 2 and an '80/'60ratio of 2.079f - 3. For quick reference: around natural abundance values a 1 % 613C change is equivalent to -0.001 APE and 1 % 6j8O change is equivalent to a ~ 2 . 0 p.p.m. 8 change from natural abundance (2074 p.p.rn.). 3. D. A. Schoeller and P. D. Klein, Biomed. Mass Spectrom. 6, 350 (1979). 4. C. M. Scrimgeour and M. J. Rennie, Biomed. Mass Spectrom. 15,365 (1988). 5. R. H. Eggers, A. Kulp, R. Tegeler, F. E. Ludtke, G. Lepsein, B. Meyer and F. E. Bauer, fur. J. Gastroenterol. Hepatol. 2, 437 (1 990). 6. C. J. Gregg, J. Y. Hutson, J. R. Prine, D. G. Ott and J. E. Furchner, Life Sci. 13, 775 (1973). 7. C. M. Scrimgeour, K. Smith and M. J. Rennie, Biomed. Mass Spectrom. 15,369 (1988). 8. J. M. Barua, C. M. Scrimgeour, I. M. Mackay and M. J. Rennie, Roc. Notr. SOC.49,201A (1990). 9. T. Preston and N. J. P. 0wens.Analyst 108,971 (1983).

10. T. Preston and

N. J. P. Owens, Biomed. Mass Spectrom. 12,

510 (1985). 1 1 . A. Barrie, J. Bricout and J. Koziet, Biomed. Mass Spectrom. 11,583 (1984). 12. T. Preston and D. C. McMillan, Biomed. Mass Spectrom. 16, 229 (1988). 13. D. A: Schoeller and P. D. Klein, Biomed. Mass Spectrom. 5, 29 (1 978). 14. H. Craig, Geochim. Cosmochim.Acta 12,133 (1 957). 15. D. Y. Graham, P. D. Klein, D. J. Evans Jr, D. G. Evans, L. C. Alpert, A. R. Opekun and T. W. Boutton, Lancet ii, 1174 (1987). 16. R. P. H. Logan, S. Dill, M. M. Walker, F. Bauer, P. A. Gummet, A. Hirschl, B. Rathbone, P. Johnson, J. H. Baron and J. J. Misiewicz, Rev. ESP. fnferm. Dig. 78, (Suppl. 1 ) . 14 (1990). 17. S. Epstein and T. Mayeda, Geochim. Cosmochim.Acta 4, 21 3

(1953). 18. D. C. McMillan, T. Preston and D. P. Taggart, Biomed. Mass Spectrom. 18. 543 (1 989). 19. C. A. M. Brenninkmeijer, P. Kraft and W. G. Mook, Isotope Geosci. 1, 1 81 (1 983). 20. W. W. Read, M. A. Read, M. J. Rennie, R. C. Griggs and D. Halliday, Biomed. Mass Spectrom. 1 f , 348 (1 984). 21. D. A. Schoeller, E. van Santen, D. W. Peterson, W. Dietz, J. Jaspan and P. D. Klein, Am. J. Clin. Nutr. 33. 2686 (1 980).