Integration of Continuous-Flow Accelerator Mass Spectrometry with

Anal. Chem. 2008, 80, 5079–5085

Integration of Continuous-Flow Accelerator Mass Spectrometry with Chromatography and Mass-Selective Detection Jimmy Flarakos,† Rosa G. Liberman, Steven R. Tannenbaum, and Paul L. Skipper* Biological Engineering Department Massachusetts Institute of Technology Cambridge, Massachusetts 02139 Physical combination of an accelerator mass spectrometry (AMS) instrument with a conventional gas chromatographmass spectrometer (GC/MS) is described. The resulting hybrid instrument (GC/MS/AMS) was used to monitor mass chromatograms and radiochromatograms simultaneously when14C-labeled compounds were injected into the gas chromatograph. Combination of the two instruments was achieved by splitting the column effluent and directing half to the mass spectrometer and half to a flowthrough CuO reactor in line with the gas-accepting AMS ion source. The reactor converts compounds in the GC effluent to CO2 as required for function of the ion source. With cholesterol as test compound, the limits of quantitation were 175 pg and 0.00175 dpm injected. The accuracy achieved in analysis of five nonzero calibration standards and three quality control standards, using cholesterol-2,2,3,4,4,6-d6 as injection standard, was 100 ( 11.8% with selected ion monitoring and 100 ( 16% for radiochromatography. Respective values for interday precision were 1.0-3.2 and 22-32%. Application of GC/ MS/AMS to a current topic of interest was demonstrated in a model metabolomic study in which cultured primary hepatocytes were given [14C]glucose and organic acids excreted into the culture medium were analyzed. Accelerator mass spectrometry (AMS) has attained considerable importance for the analysis of low-level radioisotopes in research fields such as archeology, oceanography, astrophysics, and the geological sciences.1,2 Because it is a mass spectrometrybased ion-counting technique, AMS is typically many orders of magnitude more sensitive than decay counting.3 Originally developed for chronological studies in the 1970s, radiocarbon AMS has subsequently gained popularity in the biological sciences due to its great accuracy and precision and unmatched sensitivity.4–6 * To whom correspondence should be addressed. E-mail: [email protected]. Fax: 617 252 1787. † Current address: Novartis Pharmaceuticals Corp,, One Health Plaza, 4363219, East Hanover, NJ 07936-1080. (1) Elmore, D.; Phillips, F. M. Science 1987, 236, 543–550. (2) Vogel, J. S.; Turteltaub, K. W.; Finkel, R.; Nelson, D. E. Anal. Chem. 1995, 67, 353A–359A. (3) Litherland, A. E. Annu. Rev. Nucl. Part. Sci. 1980, 30, 437–473. (4) Vogel, J. S.; Turteltaub, K. W.; Felton, J. S.; Gledhill, B. L.; Nelson, D. E.; Southon, J. R.; Proctor, I. D.; Davis, J. C. Nucl. Instrum. Methods B 1990, 52, 524–530. (5) Vogel, J. S.; Turteltaub, K. W. Nucl. Instrum. Methods B 1994, 92, 445– 453. 10.1021/ac800286g CCC: $40.75  2008 American Chemical Society Published on Web 05/22/2008

Carbon-14 and tritium are also of interest as alternatives to the stable isotopes 13C and 2H in biological isotope tracer studies because their low natural abundance provides vastly improved signal-to-noise ratios and because the small chemical concentrations required are unlikely to perturb significantly any biochemical pathways under investigation. There are two principal methods for introducing carbon samples into the ion source of an AMS instrument. Most commonly, the sample is first converted to graphite. The graphite sample is then packed into an aluminum holder that is mechanically transported through a vacuum lock to the appropriate location in the ion source where the graphite is at the focal point of an ionizing beam of Cs+. The graphitization process has been described in detail.7 Ongoing refinements have led to improved efficiency and throughput, but it remains a process that requires considerable effort and care to execute properly. Graphitization is also a process that remains essential for chronological studies, because it is only with graphite samples that it is possible to obtain the high ion currents and extended counting times necessary to achieve high statistical precision. The second method is associated with gas-accepting ion sources and involves injection of CO2, either as pure gas or conveyed in a continuous flow of helium. Most of the development to date has involved modification of conventional solid-target cesium sputter ion sources to permit their operation as gasaccepting sources.8,9 A solid target is still used as cathode, but in place of graphite there is a piece of titanium onto which the CO2 is directed and adsorbed and from which C- ions are extracted. Another, fundamentally different, approach involves formation of C+ ions in a mirowave-induced plasma followed by extraction of ions through a charge-exchange canal to give negative ions.10 Regardless of the nature of the gas-accepting ion source, each utilizes an input of gaseous CO2 produced from the sample to be analyzed. In contrast to the process of graphitization, production of CO2 can be accomplished in a continuous manner using a flow reactor. (6) Brown, K.; Tompkins, E. M.; White, I. N. H. Mass Spectrom. Rev. 2006, 25, 127–145. (7) Ognibene, T. J.; Bench, G.; Vogel, J. S.; Peaslee, G. F.; Murov, S. Anal. Chem. 2003, 75, 2192–2196. (8) Bronk, C. R.; Hedges, R. E. M. Nucl. Instrum. Methods B 1990, 52, 322– 326. (9) Hughey, B. J.; Skipper, P. L.; Klinkowstein, R. E.; Shefer, R. E.; Wishnok, J. S.; Tannenbaum, S. R. Nucl. Instrum. Methods B 2000, 172, 40–46. (10) Roberts, M. L.; Schneider, R. J.; von Reden, K. F.; Wills, J. S. C.; Han, B. X.; Hayes, J. M.; Rosenheim, B. E.; Jenkins, W. J. Nucl. Instrum. Methods B 2007, 259, 83–87.

Analytical Chemistry, Vol. 80, No. 13, July 1, 2008

5079

Such a reactor is typically a tube containing CuO in powder or wire form, and the complete systems generally incorporate one or more scrubbers to remove, at a minimum, the H2O of combustion. Technology development for combining gas chromatography with continuous flow combustion was initiated several decades ago11,12 and has led to commercial instruments specifically designed for gas chromatography-isotope ratio mass spectrometry (GC/IRMS). Direct coupling of a GC with an AMS ion source through an in-line combustion reactor was first reported by us in 2000.9 Subsequently, this laboratory13 and one other14 have described operation of a complete AMS instrument with CO2 input from a GC-combustion reactor combination, now referred to as GC/AMS. In this report, we describe a further development of GC/AMS achieved by insertion of a tee at the end of the GC column and direction of the split flow into both a mass spectrometer for simultaneous mass spectral analysis or ion-selective detection and an AMS instrument. With its combination of detectors, one that has subattomole sensitivity for a specific isotope and one that is informative regarding molecular structure, this combination instrument is expected to significantly expand the range of complex biological studies to which AMS can be applied. EXPERIMENTAL SECTION Chemicals and Supplies. Cholesterol and cholesterol2,2,3,4,4,6-d6 werepurchasedfromSigma(St.Louis,MO).[4-14C]Cholesterol (55 mCi/mmol) was obtained from American Radiochemicals (St. Louis, MO). HPLC grade acetonitrile, chloroform, methanol, and toluene were obtained from EMD (Charlotte, NC) while anhydrous sodium sulfate and copper oxide (CuO) (no. 33307) were purchased from Alfa Aesar (Ward Hill, MA). Samples were stored and derivatized in 1.5-mL Eppendorf siliconized microcentrifuge tubes (West Chester, PA). Derivatizing reagent POWER Sil-Prep (N-trimethylsilylimidazole, N,O-bis(trimethylsilyl)acetamide (BSA) + trimethylchlorosilane) was purchased from Alltech (Deerfield, IL). A mixture of toluene and CH3CN (1:1) used for reconstituting samples for derivatization was stored in glass over anhydrous sodium sulfate. Calibration and Quality Control Standards (QCs). Cholesterol and the internal standard (IS) cholesterol-d6 were dissolved separately in methanol/chloroform mixture (9:1) to make stock solutions A and IS. Stock solution R of radiolabeled cholesterol (5 nCi/mL) was prepared by dilution of the purchased solution with methanol. Aliquots of stock solution A were further diluted in methanol to yield solutions with concentrations of 0.05, 1.25, 2.5, 5.0, and 7.5 mg/mL cholesterol (spiking solutions A). Similarly, aliquots of stock solution R (5 nCi/mL ) 11 100 dpm/ mL) were diluted in methanol to yield solutions with concentrations of 250, 1250, 2500, 5000, and 7500 dpm/mL (spiking solutions R). Binary calibration standards were prepared by adding equal volumes (10 µL) of spiking solutions A and R to 0.98 mL of methanol. They were designated std B with 0.5 (2.5), std C 12.5 (11) Matthews, D. E.; Hayes, J. M. Anal. Chem. 1978, 50, 1465–1473. (12) Merritt, D. A.; Freeman, K. H.; Ricci, M. P.; Studley, S. A.; Hayes, J. M. Anal. Chem. 1995, 67, 2461–2473. (13) Skipper, P. L.; Hughey, B. J.; Liberman, R. G.; Choi, M. H.; Wishnok, J. S.; Klinkowstein, R. E.; Shefer, R. E.; Tannenbaum, S. R. Nucl. Instrum. Methods B 2004, 223-224, 740–744. (14) Ramsey, C. B.; Ditchfield, P.; Humm, M. Radiocarbon 2004, 46, 25–32.

5080


(12.5), std D 25 (25), std E 50 (50), and std F 75 (75) ng/µL (dpm/ mL). In the same manner, binary quality control samples were prepared and designated QC A with 7.5 (7.5), QC B 40 (40), and QC C 70 (70) ng/µL (dpm/mL). The IS working solution was prepared by diluting an aliquot of stock IS with methanol to a final concentration of 60 ng/µL. Cell Cultures. Six-well plastic plates (BD Biosciences) were coated with collagen by addition of 600 µL/well of a solution containing 3 mg/mL of rat tail collagen type I (Cohesion Technologies Inc., Palo Alto, CA) in H2O. Prior to its introduction into the wells, the collagen solution was mixed with 10× PBS (10% by volume) containing glucose (20 g/L) and NaHCO3 (37 g/L), and the pH was adjusted to 7.4 with HCl. The collagen was permitted to gel 1 h in the incubator prior to seeding with hepatocytes. Hepatocytes were isolated from male Fisher 344 inbred rats (Taconic, Germantown, NY) following a previously described15 protocol and immediately plated on the collagen-coated well. The cell viability, measured with trypan blue exclusion, was typically 87-91%. The cells were incubated in Hepatocyte Growth Medium (BD Biosciencies) modified as follows: purified bovine serum albumin was omitted, the concentration of niacinamide was 0.305 g/L, and the concentrations of ZnCl2, ZnSO4, CuSO4, and MnSO4 were 0.0544, 0.075, 0.02, and 0.0025 mg/L, respectively. The medium was supplemented further with 20 ng/mL epidermal growth factor. Incubator oxygen tension was set at 20% and carbon dioxide at 8.5%. Cells were seeded at a density of 500 000/well with 1.5 mL of culture medium, which was exchanged after 2 days. [14C]Glucose was added on day 4. Aliquots of medium were removed at 3, 6, and 24 h after addition of glucose. Sample Preparation. A 25-µL aliquot of culture medium was placed in 2 mL of 2% ammonium hydroxide in water (v/v) and loaded onto a Waters Accell Plus QMA Sep-Pak 3-cm3, 500-mg cartridge. The cartridge was washed sequentially with 1 volume each of 2% ammonium hydroxide in water (v/v), water, and methanol. The acidic fraction was eluted from the cartridge with 2 mL of 5% formic acid in methanol (v/v). The eluate was evaporated in vacuo in a Savant Speedvac SC110 (Thermo Electron, Waltham, MA). The dry residue was reconstituted according to the procedure implemented to derivatize cholesterol. Injection Solution Preparation. Standards B-F were mixed with IS (50 µL each) in 1.5-mL microcentrifuge tubes and evaporated to dryness in a Speedvac. A sample corresponding to standard A (i.e., no cholesterol or [14C]cholesterol) was prepared from IS only. Samples were reconstituted in toluene/CH3CN (25 µL) and derivatizing reagent (25 µL), vortexed for 30 s, and incubated at 70 °C for 15 min. Samples were cooled to room temperature prior to injecting 0.7 µL, in splitless mode, into the GC. Final sample volume was the same as the volume of the standards. A14CO2/12CO2 gas standard was prepared as follows: Carrierfree 14CO 2 was produced by treatment of NaHCO3 (Amersham Biosciences) with concentrated H2SO4. This was diluted with both CO2 and helium to achieve a carbon isotope ratio of 1.1 × 10-7 and a CO2 concentration in He of 3300 ppm. The isotope ratio was estimated from the CO2 in He concentration, determined by (15) Powers, M. J.; Janigian, D. M.; Wack, K. E.; Baker, C. S.; Beer Stolz, D.; Griffith, L. G. Tissue Eng. 2002, 8, 499–513.

Figure 1. Schematic diagram of the pertinent gas connections between chromatograph, MSD, and AMS. The dashed line rectangle represents the GC oven. A high-temperature externally actuated valve is used to select either vacuum (position A) or the AMS (position B) as destination for the part of the gas flow not going to the MSD.

comparison with a commercially supplied CO2/He mixture (5000 ppm) using the AMS Faraday cup measurements. The14C concentration was determined by liquid scintillation counting of NaOH-extractable radioactivity from measured volumes of the gas mixture. Instrumentation. The system used in the studies described here was assembled from the component parts of a HewlettPackard 5972MSD and includes a capillary tubing connection through a combustion reactor to the AMS instrument. The gas chromatograph and the mass spectrometer were physically separated to facilitate configuring the overall system. A hightemperature four-port switching valve (Valco, Houston, TX) was installed inside the GC oven as shown schematically in Figure 1, which illustrates the relevant flow paths. The GC column was connected to a tee upstream of the valve so that one part of the flow was directed to the MSD and the other part to the combustion reactor. The valve, which is located between the tee and the reactor, serves two essential functions. Diverting flow to vacuum (position A) permits the system to be operated without connection to the AMS. With the AMS in operation, the valve is switched to vacuum as needed to divert the solvent front produced by a sample injection. The two capillaries leading to the MSD and the reactor were brought out together through the oven wall via a heated zone to their respective destinations. Helium is provided to the switching valve through a capillary leak sized so that flow through the reactor remains nearly constant when flow from the GC column is diverted. The combustion reactor has been described previously.9,13 It is composed of a quartz tube (6 mm × 2 mm × 12 in.) filled in the

center 6 in. with CuO powder held in place with quartz wool. Two 2 mm × 1 mm quartz tubes are inserted in the ends to reduce dead volume. In operation, it is heated to 700-750 °C. The CuO packing requires periodic conditioning with O2. This operation was typically performed overnight after a day of operation using a flow of ∼20 mL/min O2 through the reactor maintained at 600-700 °C. Downstream of the reactor is a Nafion water separator (Perma-Pure, Toms River, NJ), which in turn is connected to the ion source through a transfer line composed of ∼ca. 2 m of 0.03-in.-i.d. PEEK tubing. Because conductance through this tubing is high relative to the restricted flow entering it, the reactor is maintained at reduced pressure. Actual pressure has not been determined, but it is sufficiently low that peak broadening relative to that observed with the MSD is acceptable for many, if not most, applications. Valves between the reactor and transfer line are provided to permit direct introduction of gas standards into the AMS system. The AMS instrument and ion source have also been described previously.9,16 A gas ionization chamber is used as a particle counter for detection of 14C as doubly charged positive ions; its design and performance are described elsewhere.17 An off-axis Faraday cup immediately before the tandem accelerator is used to monitor macroscopic stable-isotope negative ion currents. Typically, the stable-isotope ion current in an AMS instrument is (16) Liberman, R. G.; Hughey, B. J.; Skipper, P. L.; Wishnok, J. S.; Klinkowstein, R. E.; Shefer, R. E.; Tannenbaum, S. R. Nucl. Instrum. Methods B 2004, 223-224, 82–86. (17) Liberman, R. G.; Becker, U. J.; Skipper, P. L. Nucl. Instrum. Methods A 2006, 565, 686–690.


5081

one component of an isotope ratio measurement. In the system described here, the current is used principally as a diagnostic for ionization efficiency, while quantitative analysis is a function of peak area determinations in the14C versus time profiles as described below. Chromatography. Gas chromatographic analysis was carried out using a Hewlett-Packard (HP) model 5890 GC fitted with a split/ splitless injector and 5972 mass spectrometer (MS) detector. Helium was used as the carrier gas with a flow rate of 1 mL/min. The components were separated on a 30 m × 0.25 mm i.d., 0.25µm-film thickness HP-5-MS fused-silica capillary column from Agilent Technologies (Palo Alto, CA). The injector (temperature) was set to 250 °C and in splitless mode; the injection port was equipped with a Supelco (Bellefonte, PA) wool-packed, glass deactivated liner to minimize peak broadening. The column was initially maintained at 100 °C for 1 min; subsequently, the temperature was raised to 320 °C at a rate of 25 °C/min, which was then held for an additional 3.70 min (total program time, 13.50 min). Ultrahigh-purity helium was used as carrier gas. The carrier gas was set to 60 psi and column head pressure to 12 psi. Data acquisition in selected ion mode used 100 ms dwell time. Electron impact ionization energy was 70 eV. Compounds were identified by matching mass spectra with the NBS library of standard compounds and the use of authentic reference standards. Quantitative Analysis. Data were acquired by the MSD using selected-ion monitoring at m/z 458 for isotopically normal cholesterol and at m/z 464 for the internal standard cholesterol-d6. The resulting chromatograms were integrated using Agilent Chemstation 3.1 to determine peak areas. Calibration curves were constructed by linear regression of the peak area ratio versus nominal concentration of cholesterol. Data acquired by AMS were treated similarly. The particle detector software generates data in the form of 14C count rate as a function of time. A custom macro for Chemstation permits that software to import the data. Integration to produce peak areas was then performed using regular Chemstation functions. Normalization was performed to compensate for variation in ionization efficiency and transmission. Several injections of 14CO2 gas standard were made immediately before and immediately after each GC run. The number of 14C detector counts produced by each injection was determined by peak integration, and averages for the pre- and postrun injections were calculated. Linear interpolation was then used to determine the expected number of counts from the gas standard at the time during the GC run when cholesterol eluted. This number was used to normalize the area of the [14C]cholesterol peak. Typically, detector response was higher postrun as compared to prerun and the increase was in the range 10-30%. This increase is in line with previous observations16 that the target cathodes of the ion source produce increasing ion currents during the first half of their usable lifetime. [14C]Calibration curves were based on the ratio of normalized 14 [ C]cholesterol to cholesterol-d6 peak area measured by the MSD. This ratio was multiplied by a factor of 106 MSD counts to generate unitless values falling within a convenient range that were plotted as a function of the nominal concentration of [14C]cholesterol and subjected to linear regression analysis. QC concentrations of cholesterol and [14C]cholesterol were backcalculated from the calibration curves. 5082


RESULTS Quantitative Analysis of Cholesterol. The primary objective of this research was to create an analytical instrument with exceptional sensitivity for tracking the isotope 14C while simultaneously providing structure information by mass spectrometry. The hardware is described above in the Experimental Section and partially illustrated in Figure 1. Performance of the instrumentation was initially evaluated by applying it to quantitative analysis of an isotopically normal compound, cholesterol, and its 14C-labeled isotopomer. Isotope dilution mass spectrometry (IDMS) with cholesterol-d6 serving as IS was used to quantitate normal cholesterol. The equivalent of IDMS with cholesterol-d6 again as IS was used to quantitate [14C]cholesterol. Three-part mixtures of a constant concentration of IS and varying amounts of the other two isotopomers were prepared and analyzed with simultaneous MSD and AMS detection. Figure 2 illustrates representative data for a mixture having the lowest concentrations of normal cholesterol and [14C]cholesterol. The top and middle panels depict the extracted ion chromatograms (EIC) produced by normal and cholesterol-d6, respectively. Though not apparent in the figure, there is a 0.03-min difference in retention times such that the IS elutes earlier than normal cholesterol. [14C]Cholesterol elutes later than both normal cholesterol and IS. In this case, the difference is largely a function of the greater path length to the detector as well as residence time in the reactor. The reactor also causes a noticeable degree of peak broadening in the14C trace (bottom panel) relative to the peak width exhibited by the EICs. A certain amount of peak broadening is inherent to the reactor because its volume and cross section are large relative to the capillary tubing used. The inherent broadening caused by dimensional changes is largely offset by having the reactor at reduced pressure. Pressure in the reactor varies throughout its length, but the actual values have not been determined. In any event, they are clearly dependent on the actual conductances that characterize a particular CuO packing. Further illustration of peak width can be found in Figure 3. Because the AMS response can be calibrated with [14C]CO2, it is possible to compare the amount of isotope reaching the instrument with the amount injected into the chromatograph. The peak shown in the bottom panel of Figure 2 was calculated to represent 0.0011 dpm. The amount injected was nominally 0.0035 dpm. Assuming the split was 1:1, as designed, these data indicate that 63% of the injected sample was delivered to the detectors. Precision and accuracy were assessed by preparing, on each of three different days, the required calibration curves and then quantitating three QC samples. Results are collected in Table 1, which also includes the relevant statistics. The relative standard deviation (RSD) for [14C]cholesterol ranged from 22 to 32% and accuracy was within ±16% of the nominal concentration. Greater precision and accuracy were observed for cholesterol-d6: the RSD ranged from 1.0 to 3.2% and the accuracy was ±12%. Accuracy of the two analyses was similar (16 and 12%), but the precision differed greatly. High precision for cholesterol-d6 is to be expected since analyte and IS were both monitored with the same detector. The lower precision observed for analyses of [14C]cholesterol undoubtedly results in part from monitoring analyte with one detector, the AMS instrument, while the IS was

Figure 2. GC/MS/AMS analysis of cholesterol. (Top) isotopically normal cholesterol, 0.5 ng/µL; (middle) cholesterol-d6, 60 ng/µL; (bottom) [14C]cholesterol, 0.005 dpm/mL.

monitored with the MSD. Normalization of the AMS instrument response was performed using 14C[CO2], but this normalization is unlikely to be completely effective in eliminating instrument variability. A second potential source of variability is the combustion reactor. The contribution from this source is unknown. In principle, the performance of the reactor could be normalized by alternating injections of a 14C-labeled test compound and 14C[CO2] directly into the reactor, but the apparatus was not configured to permit this mode of operation. An alternative approach to improving precision of the AMS analysis would be to incorporate a 14C-labeled compound as IS 14 for C-labeled analytes. With such an approach, even though stable-isotope labeling would be sufficient to render the IS distinguishable from analyte by mass spectrometry, it would be insufficient with respect to AMS detection. A compound different and chromatographically resolvable from the analyte would be required as IS. AMS precision would likely be improved with this approach, but MS precision would be degraded. Application to Metabolite Profiling Glucose. Metabolism of [14C]glucose in a cell culture system was selected as a means to provide a complex mixture of 14C-labeled metabolites that could

be used to evaluate the dual functioning of the GC/MS/AMS instrument as a radiotracer and provider of structure information. Primary rat hepatocytes were maintained on a layer of collagen and, after being established, were exposed to [14C]glucose. It was expected that most of the normal products of the Krebs cycle would appear 14C-labeled in the medium and, since they are all carboxylic acids, would be amenable to a common procedure for isolation, derivatization, and GC analysis. Aliquots of medium were taken at several time points after exposure in the anticipation that different metabolite profiles would be observed as a function of time. Results from a representative time-course experiment are shown in Figure 3, which depicts the GC/AMS analyses of samples drawn at 3, 6, and 24 h. Although the chromatograms are intended to be directly comparable because they have the same y-axis scaling and identical volume, aliquots of culture medium were drawn and processed identically; it remains an assumption that recovery was constant. With that caveat, there are still several notable features that can be described with confidence. First, the spectrum of compounds present in the medium, as determined by relative retention time, is largely the Analytical Chemistry, Vol. 80, No. 13, July 1, 2008

5083

Figure 3. GC/AMS analysis of [14C]glucose metabolites present in cell culture medium at various times after introduction of the sugar into the medium. Quantitative analysis was performed on the 12 peaks indicated. Three of the peaks (1, 3, 10) were identified by mass spectrometry as discussed in the text. Table 1. Quantitative Analysis of Isotopically Normal and [14C]Cholesterol Using Cholesterol-d6 As Internal Standard [14C]cholesterol

cholesterol QC sample

A

B

C

A

B

C

mean SDa RSD (%) accuracy (%)

(7.5) 6.6 0.21 3.2 88

(40) 38 0.38 1.0 96

(70) 68 2.2 3.2 97

(7.5) 8.3 2.6 32 111

(40) 47 10 22 116

(70) 61 18 30 87

a QC sample concentrations and calculated values are given in units of ng/µL (cholesterol) or dpm/mL ([14C]cholesterol).

same at all three time points. The greatest change in absolute abundance occurred between 6 and 24 h. The greatest changes in relative abundance also occurred during this interval. Peak 12, for example, was undetectable at the first two times but is one of the most abundant at 24 h. Other changes in relative abundance are also notable. Specifically, peaks 2, 6, and 8 exhibit little change with time, while peak 4 appears to increase from 3 to 6 h and then disappear altogether. The variation in the metabolite profile 5084


is intriguing, but it should be noted that these results provide an insufficient basis from which to draw any conclusions regarding cellular function. They are presented primarily to illustrate the quality of data that could be obtained with an approach such as the one described here. In further evaluation of data quality, replicate analyses (N ) 3) of culture medium drawn at one time point were performed. The RSD for each of the peaks ranged from 7.7 to 92%, with a median value of 24%. The greatest variance occurred where peaks were at or below the limit of detection. Metabolite Identification. The MSD was operated in full scan mode during the GC/MS/AMS analyses of [14C]glucose metabolites in order to characterize compounds identified as being 14Clabeled from the GC/AMS profile. Total ion chromatograms (TIC) were first compared with their associated GC/AMS traces. Mostly, peaks well-defined in the TIC did not match the peaks detected by AMS nor did they exhibit mass spectra consistent with the expected Krebs cycle products. A different search strategy was thus implemented. Extracted ion chromatograms were generated based on molecular ions or prominent fragment ions (e.g., [M -

72]+) of expected metabolites. These EICs were then compared with the 14C traces. While not exhaustive, this strategy did succeed in producing evidence of compound identity for three of the peaks shown in Figure 3, peaks 1, 4, and 9. These three metabolites were apparently lactic acid, succinic acid, and glutamic acid. Further support for the presumed structures was obtained by demonstrating that EICs for prominent ions in mass spectra of the authentic compounds exhibited the same profile with good fidelity. Lastly, authentic compounds were derivatized and subjected to GC/MS analysis and demonstrated to have the same retention time as the compounds originally detected by GC/MS/ AMS analysis. DISCUSSION Continuous-flow14C-AMS continues to be a major objective of instrumentation development for accelerator mass spectrometry. An obvious requirement for any continuous-flow instrument is a gas-accepting ion source. As detailed in the introduction, development of gas ion sources for AMS is proceeding along at least two distinct paths: modification of existing cesium sputter sources and coupling of a microwave-plasma sourceswhich produces a C+ ion beamswith a charge-exchange canal to yield the required Cbeam. Both approaches make use of CO2 as the form of carbon accepted. The essential link required to combine gas chromatography with AMS is, then, some apparatus that converts the organic compounds eluting off the GC column into CO2. Technology for (18) Eglinton, T. I.; Aluwihare, L. I.; Bauer, J. E.; Druffel, E. R. M.; McNichol, A. P. Anal. Chem. 1996, 68, 904–912. (19) Stott, A. W.; Berstan, R.; Evershed, R. P. Anal. Chem. 2003, 75, 5037– 5045. (20) Buchholz, B. A.; Fultz, E.; Haack, K. W.; Vogel, J. S.; Gilman, S. D.; Gee, S. J.; Hammock, B. D.; Hui, X.; Wester, R. C.; Maibach, H. I. Anal. Chem. 1999, 71, 3519–3525.

oxidizing organic compounds present in the effluent of a gas chromatograph is well developed, with commercial instruments now available for conducting GC/IRMS. Home-built apparatus such as we have described is not difficult to construct. Thus, it is natural that the first efforts to integrate separation technology with AMS would involve gas chromatography. The presence of helium as the only other component of the sample stream is also an attractive feature of gas chromatography for combining it with analyte combustion. Combination of HPLC, the other mainstay of separation technology, with AMS entails the significant challenge of separating analyte, prior to combustion, from a vastly greater amount of material than the small amount of carrier gas used for GC. Gas chromatography is also more readily joined with conventional mass spectrometry because carrier gas constitutes a much smaller mass than HPLC mobile phase that needs to be separated from analyte. Conventional MS adds a significant dimension to analysis by AMS, which is limited to determination of isotope concentration. Evidence for compound identity in addition to isotope concentration can be achieved by chromatographic resolution prior to AMS analysis, e.g.,18–20 but such evidence has limited value. In contrast, acquisition of mass spectral data simultaneously with AMS analysis during a chromatographic separation should have substantial value, and that is the principal motivation for development of the instrumentation described here. ACKNOWLEDGMENT This work was supported by the National Institutes of Health through grants R43CA084688 and P30-ES002109. Received for review February 11, 2008. Accepted April 22, 2008. AC800286G


5085

Integration of Continuous-Flow Accelerator Mass Spectrometry with

Recommend Documents