Carbon Position-Specific Isotope Analysis of Alanine and

Feb 10, 2005 - zene derived entirely from C(ring), and toluene was pro- .... bling is known to occur and, furthermore, is unique to each IRMS system...
0 downloads 0 Views 175KB Size
Download PDF
Anal. Chem. 2005, 77, 1746-1752

Carbon Position-Specific Isotope Analysis of Alanine and Phenylalanine Analogues Exhibiting Nonideal Pyrolytic Fragmentation Christopher J. Wolyniak, Gavin L. Sacks, Bruce S. Pan, and J. Thomas Brenna*

Division of Nutritional Sciences, Savage Hall, Cornell University, Ithaca, New York 14853

Recent advances in gas chromatography combustionisotope ratio mass spectrometry (GCC-IRMS) has made compound-specific isotope analysis routine, but reports on position-specific isotopic analysis are still scarce. Online GC-pyrolysis (Py) coupled to GCC-IRMS is reported here for isolation and isotopic characterization of alaninol and phenethylamine, analogues of alanine and phenylalanine, respectively. Ideally, pyrolytic fragments will originate from unique sites within the parent molecule, and isotope ratios for each position within the parent can either be measured directly or calculated from fragment isotope ratios without substantially degrading the analytical precision. Alaninol pyrolysis yielded several fragments, of which CO and CH4 were used for isotope ratio calculations. Isotope labeling experiments showed that CO derived entirely from the C(1) position, while all three positions of alaninol contributed to CH4 (29.0 ( 0.3% from C(1), 3.6 ( 0.2% from C(2), and 66.9 ( 1.1% from C(3)). We demonstrate iterative use of mass balance to calculate isotope ratios from all positions despite the nonideal positional fidelity of CH4. Pyrolysis of phenethylamine generated benzene and toluene fragments. Benzene derived entirely from C(ring), and toluene was proportionately formed from C(3) and C(ring). Relative intramolecular isotope ratios (∆δ13C) were calculated directly from δ13C of fragments or indirectly by mass balance. Though the C(3) isotope ratio was calculated from the benzene and toluene fragments, propagation of errors showed that the final precision of the determination was degraded due to the small contribution that C(3) makes to toluene. Samples of each amino acid from four different vendors showed natural variability between sources, especially at the C(1) position of alaninol (range of ∆δ13C ∼ 50‰). The average precision was SD(∆δ13C) < 0.20‰ for directly measured positions of alaninol and phenethylamine. The precision of indirectly measured positions was poorer (SD(∆δ13C) ) 0.94‰ for alaninol, 6.54‰ for phenethylamine) due to propagation of errors. These data demonstrate that GC-Py-GCC-IRMS data can be used to extract high-precision isotope ratios from amino acids despite nonideal positional fidelity in fragments and that * To whom correspondence should be addressed. E-mail: . Phone: (607) 255-9182. Fax: (607) 255-1033.

1746 Analytical Chemistry, Vol. 77, No. 6, March 15, 2005

natural intramolecular variability in δ13C can be used to distinguish different sources of amino acids. High-precision compound-specific isotope analysis (CSIA) is now a well-established technique that employs on-line separations prior to isotope analysis of whole molecules. It has proven useful for studies of sources and pathways by which a sample is formed and is usually implemented with gas chromatography-combustion isotope ratio mass spectrometry (GCC-IRMS) systems.1-5 Recent applications include authenticating natural products such as cinnamaldehyde6 and caffeine,7 investigating amino acid metabolism in pollen-feeding butterflies,8 and tracing the origin of hydrocarbons in natural gas deposits.9 The majority of these studies involve empirical correlations between samples and putative sources, in part because extrapolation of isotope ratios at the whole-molecule level are not easily referred to the elementary mechanism(s) by which analytes are created. Isotopic fractionation as a result of (bio)chemical processes occurs primarily at reactive sites, as a result of different rates of bond breaking and bond formation among the isotopomers involved.10 Natural variation in stable isotope ratios studied at the intramolecular, or position-specific, level thus yields the most specific information about chemical origins.11,12 However, the development of facile techniques for high-precision positionspecific isotope analysis (PSIA) is at an early stage. While NMR spectroscopy is widely used for determining site-specific D/H ratios at natural abundance,13 it is rarely used for other isotopes (1) Brenna, J. T.; Corso, T. N.; Tobias, H. J.; Caimi, R. J. Mass Spectrom. Rev. 1997, 16, 227-258. (2) Asche, S.; Michaud, A. L.; Brenna, J. T. Curr. Org. Chem. 2003, 7, 15271543. (3) Lichtfouse, E. Rapid Commun. Mass Spectrom. 2000, 14, 1337-1344. (4) Schmidt, T. C., Zwank, L., Elsner, M., Berg, M., Meckenstock, R. U., Haderlein, S. B. Anal. Bioanal. Chem. 2004, 378, 283-300. (5) Meier-Augenstein, W. J Chromatogr., A 1999, 842, 351-371. (6) Sewenig, S.; Hener, U.; Mosandl, A. Eur. Food Res. Technol. 2003, 217, 444-448. (7) Richling, E.; Hohn, C.; Weckerle, B.; Heckel, F.; Schreier, P. Eur. Food Res. Technol. 2003, 216, 544-548. (8) O’Brien, D. M.; Boggs, C. L.; Fogel, M. L. Proc. R. Soc., London Ser. B 2003, 270, 2631-2636. (9) Prinzhofer, A.; Battani, A. Oil Gas Sci. Technol.-Rev. Inst. Fr. Petrole 2003, 58, 299-311. (10) Hayes, J. M. In Stable isotope geochemistry; Cole, D. R., Ed.; Mineralogical Society of America: Washington, DC, 2001; pp 225-277. (11) Schmidt, H. L. Naturwissenschaften 2003, 90, 537-552. (12) Brenna, J. T. Rapid Commun. Mass Spectrom. 2001, 15, 1252-1262. (13) Martin, G. J. Isot. Environ. Health Stud. 1998, 34, 233-243. 10.1021/ac048524v CCC: $30.25

© 2005 American Chemical Society Published on Web 02/10/2005

due to poor sensitivity and long analysis times. Therefore, most reported methods of PSIA of 13C/12C, 15N/14N, and 18O/16O employ IRMS. PSIA requires the off-line or on-line fragmentation of the parent molecule, followed by IRMS analysis for isotopic characterization of the fragments. Off-line carbon (C-)PSIA studies relied upon chemical degradation, isolation, and combustion of the fragments to CO2, followed by isotopic analysis using dual-inlet IRMS.14,15 Although these reports yielded important insights into biological processes, these methods are tedious and require large sample sizes. GCC-IRMS for on-line separation and combustion of the fragmentation products improves throughput and sample requirements. For example, a method was recently described in which vanillin is converted quantitatively to guaiacol off-line. The two compounds are analyzed by GCC-IRMS to yield site-specific 18O/16O and 13C/12C ratios of the methoxy group of vanillin versus the rest of the molecule.16 Unfortunately, in this method and others, not all positions are accessible by routine chemical methods, especially skeletal carbon sites. Furthermore, collecting volatile fragments is cumbersome, making off-line PSIA of small molecules difficult. In-source dissociation provides an alternative to off-line degradation and has been successfully used for studies of N2O isotopomers.17 Rapid site-specific 15N/14N measurements are possible in conjunction with a GCC-IRMS system.18 A complication of this methodology is that some degree of isotopic scrambling is known to occur and, furthermore, is unique to each IRMS system. Most groups rely upon carefully measured correction factors, although a recent report of a well-characterized N2O standard may make comparisons of data between laboratories easier.19 In general, in-source fragmentation has not provided a framework for PSIA measurement of molecules other than N2O. We have reported several times on a general strategy for PSIA, based on coupling of an on-line pyrolysis (Py) furnace to a GCCIRMS system following an initial GC separation step.20 Pyrolytic fragmentation can usually be generated throughout the molecule, including along the carbon backbone, which increases the number of sites accessible to PSIA, while on-line collection and separation of fragments improves throughput and reduces the possibility of inadvertent contamination. C-PSIA by (GC-)Py-GCC-IRMS systems of fatty acid methyl esters,20 alkanes,21 fatty alcohols,22 and short-chain organic acids23,24 have been reported by our group and others. Since pyrolysis is typically nonquantitative, the fragmentation step is a potential source of isotopic fractionation. In (14) Abelson, P. H.; Hoering, T. C. Proc. Natl. Acad. Sci. U.S.A. 1961, 47, 623632. (15) Monson, K. D.; Hayes, J. M. J Biol Chem 1982, 257, 5568-5575. (16) Bensaid, F. F.; Wietzerbin, K.; Martin, G. J. J. Agric. Food Chem. 2002, 50, 6271-6275. (17) Toyoda, S.; Yoshida, N. Anal. Chem. 1999, 71, 4711-4718. (18) Rockmann, T.; Kaiser, J.; Brenninkmeijer, C. A. M.; Brand, W. A. Rapid Commun. Mass Spectrom. 2003, 17, 1897-1908. (19) Kaiser, J.; Park, S.; Boering, K. A.; Brenninkmeijer, C. A. M.; Hilkert, A.; Rockmann, T. Anal. Bioanal. Chem. 2004, 378, 256-269. (20) Corso, T. N., Brenna, J. T. Proc. Natl. Acad. Sci. U.S.A. 1997, 94, 10491053. (21) Corso, T. N., Brenna, J. T. Anal. Chim. Acta 1999, 397, 217-224. (22) Corso, T. N.; Lewis, P. A.; Brenna, J. T. Anal. Chem. 1998, 70, 3752-3756. (23) Yamada, K.; Tanaka, M.; Nakagawa, F.; Yoshida, N. Rapid Commun. Mass Spectrom. 2002, 16, 1059-1064. (24) Dias, R. F.; Freeman, K. H.; Franks, S. G. Org. Geochem. 2002, 33, 161168.

recent work on C-PSIA of methionine and leucine analogues, we demonstrated that pyrolysis-induced fractionation can be taken into account reporting intramolecular isotope ratios relative to a chemically identical standard.25 In addition, we showed that isotopically labeled parent molecules can be used to measure the “structural fidelity” of fragments, that is, the degree to which fragments correspond to specific positions in the parent. In this paper, we extend PSIA by GC-Py-GCC-IRMS to alaninol, the amino alcohol analogue of alanine, and phenethylamine, the decarboxylated analogue of phenylalanine and the first aromatic compound that we have considered in detail. Though we report below that alaninol does not yield fragments of sufficient structural fidelity to permit direct assignment of fragment positions to parent positions, we demonstrate calculation of high-precision PSIA values from the alaninol fragments of nonideal fidelity, which is an extended version of the method used to correct for scrambling in N2O isotopomer measurements. Finally, we applied our methods to commercial sources of alanine and phenylalanine to determine the extent to which intramolecular δ13C variability is present among a set of samples. EXPERIMENTAL SECTION Synthesis of Amino Acid Analogues. Amino acids were chemically modified to allow GC analysis. The decarboxylation reaction applied to Phe can be used for several different amino acids, allowing the observation of several amino acids in a single reaction mixture.26 Decarboxylation of Ala yields ethylamine, which readily evaporates from solution and cannot be conveniently handled. Alaninol was thus chosen as an appropriate derivative due to its boiling point (170 °C).The compounds for analysis were either synthesized from their respective amino acid or purchased commercially and used without further derivatization. For brevity, we refer to vendors by three-letter abbreviations. Alaninol was acquired from Acros (Geel, Belgium), “Acr”, Aldrich (St. Louis, MO), “Ald”, and Fluka (St. Louis, MO), “Flu”, with underivatized Ala from Sigma (St. Louis, MO), “Sig”. Phenylalanine was obtained from Acros, Aldrich, Alfa Aesar (Ward Hill, MA), “Aae”, and Sigma. 13C-Labeled alanine and phenylalanine were obtained from Cambridge Isotope Laboratories (Andover, MA). Samples obtained as free amino acid were derivatized for analysis. Ala reduction was accomplished using methods previously described.27 Briefly, 20 mg of Ala was added to 21 mg of NaBH4 in 50 mL of THF. With the solution in an ice bath, I2 (170 mg in 300 µL of THF) was added dropwise. After evolution of H2 gas was complete, the mixture was refluxed for 15 h. Methanol was added to remove excess NaBH4, solvent was evaporated, and 20% (w/w) KOH was added. Methylene chloride was used to extract the product. The methylene chloride was evaporated, and product was dissolved in methanol at a concentration of 2.5 µg/µL (82 nmol/µL). Phenethylamine was prepared using the method of Wallbaum et al.28 Phenylalanine (25 mg) was added to tetraethylene glycol dimethyl ether (500 µL), and 6 µL of 2-cyclohexen-1-one was added (25) Sacks, G. L., Brenna, J. T. Anal. Chem. 2003, 75, 5495-5503. (26) Hashimoto, M., Yutaka, E. D. A., Osani, Y., Iwai, T., Aoki, S. Chem. Lett. 1986, 893-896. (27) Zaideh, B. I., Saad, N. M. R., Lewis, B. A., Brenna, J. T. Anal. Chem. 2001, 73, 799-802. (28) Wallbaum, S., Mahler, T., Martens, J. Synth. Commun. 1994, 24, 13811387.

Analytical Chemistry, Vol. 77, No. 6, March 15, 2005

1747

as a catalyst. The suspension was heated at 170 °C for 4 h. The resulting solution was diluted 10:1 with methanol to yield a concentration of ∼4 µg/µL (82 nmol/µL) for analysis. GC-Py-GCC-IRMS Instrumentation. The GC-Py-GCCIRMS system used has been previously described20 and is discussed here briefly. One microliter of solution (2.5 mg/mL alaninol, 4 mg/mL phenethylamine) was injected splitless, using an autosampler (Varian 8200, Varian, Inc. Walnut Creek, CA) into GC1 (HP 5890, Palo Alto, CA). Components were separated using a DB-1 capillary column (60 m × 0.32 mm × 0.5 µm, J&W Scientific, Palo Alto, CA). For alaninol, the GC1 oven temperature was initially 50 °C and ramped to 200 °C at 10 °C/min. For phenethylamine, GC1 was initially 100 °C, ramped to 200 °C at 20 °C/min, ramped to 220 °C at 2 °C/min, ramped to 250 °C at 20 °C/min, and held at 250 °C for 13 min. Head pressure was set at 33 psi with a flow rate of 25 cm/s through GC-1 at 100 °C. An electronically controlled rotary valve (Valco, Houston, TX) controlled the output of GC1, sending the analyte to either a flame ionization detector (FID), or to pyrolysis-GCC-IRMS analysis. The FID was used to determine retention times for the individual components, which were then used to set valve-switching times to direct a specific component to IRMS analysis. A single length of 0.32-mm fused-silica capillary tubing passed from the rotary valve of GC1 through the pyrolysis furnace to the input of GC2. After exiting the GC1 rotary valve, the effluent traveled a continuous length of heated transfer line to the pyrolysis furnace. A resistively heated Fibercraft furnace (Thermcraft, Winston-Salem, NC), created a pyrolysis zone of 20 cm. The capillary passed through the furnace secured with a ceramic tube, similar to a design described elsewhere.20 Pyrolysis temperature was maintained within (1 °C and controlled by a CN9000A series temperature controller (Omega Engineering, Stamford, CT). Following pyrolysis, the resulting fragments were directed to GC2 via a heated transfer line. Fragments were separated in GC2 (Varian 3400, Walnut Creek, CA) on a CarbonPLOT column (30 m × 0.32 mm × 1.5 µm, J&W Scientific, Folsom, CA). For alaninol, the oven was held at 30 °C for 20 min, ramped to 160 °C at 20 °C/min, ramped to 200 °C at 10 °C/min, and held at 200 °C for 5 min. For phenethylamine, the oven was held at 30 °C for 21.5 min, ramped to 235 °C at 20 °C/min, and held at 235 °C for 30 min. During GCC analysis, the pyrolysis furnace was held at 300 °C to act as a heated transfer line, and a 1-m length of 0.32mm fused-silica capillary tubing replaced the GC2 column, with GC2 held at 250 °C. In both cases, the eluent stream passed through a rotary valve where the sample could be directed to either molecular or isotope ratio analysis. Molecular analysis was accomplished with a Varian Saturn III QISMS ion trap, operating in positive ion electron impact mode. The Wiley mass spectral database (Palisades, Newfield, NY) was used to aid the identification of spectra collected. For isotopic analysis, the eluent stream was directed to a combustion furnace. The furnace was resistively heated to 940 °C by a second Fibercraft furnace and was composed of a 30-cm, 0.5-mm-i.d. ceramic tube loosely packed with oxidized Cu. The combustion products passed through a Nafion water trap and an open split (10:1 split ratio) into a Finnigan-MAT 252 (Bremen, Germany). IRMS data were collected via high-precision NI435x data acquisition boards (National Instruments, Austin, TX). Operation 1748

Analytical Chemistry, Vol. 77, No. 6, March 15, 2005

of the system was controlled and data collected using SAXICAB,29 a home-written Labview 6i30 based program. The program allows for control of the GC ovens, the GC1 rotary valve, pyrolysis furnace, and CO2 standard pulses. SAXICAB was also used to calculate δ13Cpdb values. Previous work has shown that identical results are obtained using either SAXICAB or ISODAT, the proprietary vendor software.29 The summation method was used to integrate peaks, using the “dynamic” background correction algorithm.31,32 The final δ13Cpdb value was determined with an adjustment to the R45 signal to account for the presence of 17O.33 Isotope Ratio Reporting. In IRMS, the isotope ratio is reported relative to a standard using delta notation,

δ13Cpdb )

[

]

RSPL - Rpdb × 1000 Rpdb

(1)

where Rx ) 13C/12C, SPL refers to the sample and pdb refers to PeeDee Belemnite, the international standard with Rpdb ) 0.011 237 2. Reporting isotope ratios as δ13Cpdb is analytically correct for BSIA and CSIA, where both the analyte and standard gas have undergone the same degree of fractionation in steps that are nonquantitative, primarily within the IRMS itself. In pyrolytic PSIA, the pyrolysis step results in additional fractionation in the analyte to which the standard gas is not subject. To account for the difference, relative delta values, ∆δ13Cstd, are reported instead of absolute delta values, with the isotope ratio reported with respect to a selected standard. When 13C is present at near natural abundance levels, the relative isotope ratio is the difference between standard and sample δ values.

∆δ13Cstd ) δ13Cpdb(spl) - δ13Cpdb(std)

(2)

Formally, an additional cross-term appears, but it is neglected because its value is essentially zero at natural abundance levels.25 For these analytical studies, one sample is designated arbitrarily as the standard. Determination of Fidelity and Fractional Contribution. The relative contribution of a carbon to a fragment formed during pyrolysis was determined using labeled Ala and Phe according to methods previously established.25 Separate samples were used for each position with Ala labeled in the 1, 2, or 3 position and Phe labeled in the 2 or 3 position, or ring moiety. Figure 1 shows structures and position labels, assigned using standard IUPAC naming rules. Each sample was labeled in only one position or moiety. Labeled compounds were derivatized and added to unlabeled compound solution to make total 13C label between 0 and 300‰. For each position, three solutions were used with different 13C concentrations, one with no label added, and two labeled solutions at different 13C concentrations. We define fidelity to be ideal when >95% of a fragment’s carbon can be assigned to a specific position or moiety within the parent analyte. (29) Sacks, G. L., Brenna, J. T., Sepp, J. T. Presented at the 49th ASMS Conference on Mass Spectrometry, Chicago, IL, 2001. (30) 6i ed.; National Instruments: Austin, TX, 2000. (31) Ricci, M. P., Merritt, D. A., Freeman, K. H., Hayes, J. M. Org. Geochem. 1994, 21, 561-571. (32) Sacks, G. L., Wolyniak, C. J., Brenna, J. T. J. Chromatogr., A 2003, 1020, 273-282. (33) Santrock, J., Studley, S. A., Hayes, J. M. Anal. Chem. 1985, 57, 14441448.

Figure 1. Derivatives and position labels of analytes: (A) alaninol, derivative of Ala; (B) phenethylamine, derivative of Phe. During derivatization, carbon 1 of Phe is removed.

The observed isotope ratio of a fragment or molecule (13Robs) is a weighted sum of each carbon position:

Robs )

∑R [X ] i

i

(3)

i

where Ri is the isotope ratio of carbon i and [Xi] is the mole fraction of carbon i contributing to the fragment. To determine fidelity, the expression is expanded to separate labeled and unlabeled carbons:

Robs )

∑R [X ] + R i

i

lab[Xlab]

Figure 2. (a) Pyrogram of alaninol at 900 °C. Fragments formed were A ) CO, B ) CH4, C ) ethylene, D ) ethane, E ) propylene, and F ) CH3CN. (b)Fidelity plot for fragments used in isotope ratio calculations. C(1), C(2), and C(3) refer to carbon positions. CO fragment C(1) slope ) 0.989 ( 0.007, CH4 fragment C(1) slope ) 0.290 ( 0.003, C(2) slope ) 0.036 ( 0.002, and C(3) slope ) 0.669 ( 0.011. - - represents ideal fidelity, slope ) 1. Table 1. Fractional Contribution of Carbon Positions from Alaninol Pyrolysis Fragments at 900 °C

(4)

fragment

i

where Rlab is the isotope ratio of the labeled carbon position, and [Xlab] is the mole fraction contribution of the labeled carbon to the fragment. The term containing all nonlabeled carbons is constant, and Rlab varies for each solution. Rlab is calculated using the isotope ratio for the whole compound determined by GCC analysis and the weighted sum equation for the observed isotope ratio, assuming natural abundance at remaining carbon positions. Equation 4 is plotted as Robs versus Rlab; the resulting slope, [Xlab], is structural fidelity, the fraction that labeled carbon contributes to a particular fragment. Errors of slopes were determined using the regression data analysis tool in Microsoft Excel 2000 for Windows XP (Redmond, WA). RESULTS AND DISCUSSION Fragmentation of Analytes. Both compounds require 900 °C to provide sufficient signal for analysis with desired precision. At temperatures above 950 °C, CH4 partially decomposes to C and H2.21 Fragmentation was observed at lower temperatures with smaller signal and correspondingly poorer precision. Alaninol. Figure 2a shows a representative pyrogram at 900 °C, with peaks corresponding to CO, CH4, ethylene, ethane, propylene, and acetonitrile. Figure 2b shows fidelity plots for the CO and CH4 fragments, and Table 1 presents the calculated fidelities for all fragments. The slopes are the fractional contribution of the labeled position to the fragment; for example, 29.0 ( 0.6% of the CH4 fragment originates from C(1) of alaninol. The fragments formed and their corresponding fidelities were consistent with expectations based upon the molecular structure. Ideal fidelity for the CO fragment was expected as C(1) is the only carbon bonded to oxygen in the parent molecule. CH4 came from all three carbons in the molecule; the most favored carbon was

CO naa na CH4 na na ethane na na propylene na na acetonitrile

a

fractional contribution C(1) C(2) C(3) C(1) C(2) C(3) C(1) C(2) C(3) C(1) C(2) C(3) C(1) C(2) C(3)

98.9 ( 0.2% na na 29.0 ( 0.6% 3.6 ( 0.5% 66.9 ( 2.5% 25.4 ( 0.8% 4.4 ( 0.4% 59.7 ( 2.7% 31.1 ( 2.3% 29.8 ( 3.8% 37.9 ( 3.7% 19.4 ( 1.6% 55.2 ( 0.2% 38.0 ( 1.2%

Not available.

the free methyl group at position 3. C(1) contributes to methane to a lesser extent, as its formation requires the split of a C-C and C-O bonds. C(2) contributes little since it requires the breaking of two C-C bonds and a C-N bond. The relatively high pyrolysis temperature compared to previous work25 caused CO and CH4 fragments to be most prevalent. Among the larger fragments, three carbon fragment propylene was observed, though fidelity was nonideal, with C(3) slightly favored (37.9% fidelity) over C(2) (29.8% fidelity) and C(1) (31.1% fidelity). Fidelities were slightly outside the 95% confidence limit for equal contribution from each position. Ideal fidelity was observed for only the C(1) position, indicating that the CO fragment isotope ratio can be used as a measure of the C(1) isotope ratio after appropriate calibration. There are no fragments with ideal fidelity for the C(2) and C(3) positions; however, the isotope ratios for these positions can be calculated from the measured fidelities using mass balance equations. CO and CH4 were used in calculating isotope ratios as these fragments were the most abundant and gave the best precision. Analytical Chemistry, Vol. 77, No. 6, March 15, 2005

1749

Table 2. Fractional Contribution of Carbon Positions from Phenethylamine Pyrolysis Fragments at 900 °C fragment

fractional contribution C(2)

benzene na na toluene na na a

Figure 3. Phenethylamine (a) pyrogram of phenethylamine at 900 °C. Fragments formed were A ) ethylene, B ) HCN, C ) benzene, and D ) toluene. (b) Fidelity plot for fragments used in isotope ratio calculations. C(3) and C(ring) refer to carbon position/moiety. Benzene fragment C(ring) slope ) 0.966 ( 0.015, toluene fragment C(3) slope ) 0.140 ( 0.004, and C(ring) slope ) 0.864 ( 0.022. - - - represents ideal fidelity, slope ) 1.

Average precisions for CO and CH4 were SD(δ13C) < 0.45‰ and SD(δ13C) < 0.40‰, respectively. Equation 5 represents the equivalence of ∆δ13C for the CO fragment and C(1). Equation 6 derives from the safe assumption that each carbon contributes equally to the isotope ratio in CSIA analysis. Coefficients for eq 7 are the fidelity for the specific carbon determined in the labeled compound study.

∆δ13C(CO) ) ∆δ13C(1) (5) 1 13 (1) 1 13 (2) 1 13 (3) 13 ∆δ C(Total) ) ∆δ C + ∆δ C + ∆δ C (6) 3 3 3

naa na 96.6 ( 3.3% na 14.0 ( 0.8% 86.4 ( 4.9%

C(3) C(ring) C(2) C(3) C(ring)

Not available.

ratios and fidelity could not be determined for the ethylene and HCN peaks. The benzene and toluene peaks were larger as these peaks contained six and seven carbons, respectively. Figure 3b shows the plots used to calculate fidelity and fractional contributions are shown in Table 2. In contrast to alaninol, fidelities of the benzene and toluene fragments were ideal, with each coming from a single moiety in the parent compound. In the case of toluene, the ring carbon contributes six out of seven carbons and has a fractional contribution of 86.4 ( 4.9%, ∼6/7. C(3) contributes the remaining carbon with fractional contribution of 14.0 ( 0.8%. Benzene peaks were measured with a precision of SD(δ13C) < 0.47‰ while the toluene precision was SD(δ13C) < 0.35‰. Relative isotope ratios were calculated for C(2), C(3), and C(ring). Variations among the six carbon atoms in the ring were not calculated, as the benzene ring does not produce observable fragments during pyrolysis.21 Isotope ratios for each position were calculated using equations for benzene and toluene fragments and the total compound, represented by eqs 10-12, respectively. Coefficients in eq 12 were weighted to account for six carbons in the ring.

∆δ13C(Bz) ) ∆δ13C(ring) 13

13

(3)

(10) 13

(ring)

(7)

∆δ C(Tol) ) 0.14∆δ C + 0.86∆δ C (11) 1 1 6 ∆δ13C(Total) ) ∆δ13C(2) + ∆δ13C(3) + ∆δ13C(ring) (12) 8 8 8

Using eq 5, eqs 6 and 7 have two unknowns and can be solved simultaneously for C(2) and C(3), using measured values for ∆δ13C(Total) and ∆δ13C(CH4). Equations 8 and 9 show the rearranged forms of the equations used to solve for C(2) and C(3) isotope ratios.

The relative isotope ratio for the ring position, ∆δ13Cring, was determined directly from the benzene fragment as shown in eq 10. C(3), ∆δ13C3, was determined via eq 11 for the toluene fragment, rearranged as follows.

∆δ13C(CH4) ) 0.29∆δ13C(1) + 0.04∆δ13C(2) + 0.67∆δ13C(3)

∆δ13C(2) )

∆δ13C(3) )

1 0.29 1 ∆δ13C(1) ∆δ13C(CH4)- ∆δ C(Total)3×0.67 3 3×0.67 0.04 1 3 3×0.67 (8)

(

13

∆δ13C(3) )

)

∆δ13C(CH4) - 0.29∆δ13C(1) - 0.04∆δ13C(2) 0.67 (9)

Propagated errors were calculated based on eqs 8 and 9, with 1/3 being the only term without measurement error. Phenylalanine. Pyrolysis was conducted at 900 °C and formed ethylene, HCN, benzene, and toluene fragments. Figure 3a shows the pyrogram for this molecule. Due to low signal levels, isotope 1750

Analytical Chemistry, Vol. 77, No. 6, March 15, 2005

∆δ13C(Tol) - 0.86∆δ13C(ring) 0.14

(13)

where ∆δ13C(toluene) and ∆δ13Cring are measured. ∆δ13C(2), was found using mass balance for the whole phenethylamine molecule (eq 14).

∆δ13C(2) ) 8∆δ13C(Total) - ∆δ13C(3) - 6∆δ13C(ring) (14) Multiple Source Studies. The method was applied to samples containing natural abundance 13C from four different commercially available sources for each molecule to determine whether real intramolecular isotopic variability could be detected. Alaninol. At the molecular level, the δ13C of all four sources are within 3.2‰, with Flu and Sig being indistinguishable (Flu

Figure 4. Variation among sources for Ala and Phe. Results are given as relative isotope ratios, with the selected standard being set to 0‰ for all carbon positions. The line across each graph is the standard, ∆δ13C ) 0‰. (a) Variation among alaninol sources, with Fluka used as the standard. (b) Variation among phenethylamine sources, with Aae used as the standard.

∆δ13CFlu ) 0.00‰, Sig ∆δ13CFlu ) 0.01 ( 0.23‰). Figure 4a shows relative isotope ratios for the Ala sources at both the positionspecific and compound-specific levels. Significant variations in the amount of 13C are apparent throughout the samples, especially in the C(1) and C(2) positions. In the C(1) position, enrichment of up to ∆δ13C1,Flu ) 50.4 ( 0.2‰ is observed in Acr. In the C(2) position, depletion of up to ∆δ13C(2)Flu ) -33.1 ( 1.5‰ is observed in Ald. The C(3) position exhibits less variation, with ∆δ13C(3)Flu ranging only from -11.7 ( 0.7 to 3.3 ( 0.4‰ among the samples. C(1) (measured directly) had an average standard deviation of SD(∆δ13C(1)) ) 0.19‰, while C(2) and C(3) (computed indirectly) had an average standard deviation of SD(∆δ13C) ) 0.94‰. Phenethylamine. At the molecular level, Acr, Ald, and Sig differed in 13C concentration by less than 0.1‰. Aae was depleted by ∼16‰ with respect to the other samples. Calculated relative isotope ratios for Phe analysis are plotted in Figure 4b. Acr, Ald, and Sig samples had about the same enrichment in the ring position with ∆δ13CAae(Rg) ∼ 17‰ (17.1, 17.2, and 16.5‰, respectively) and an enrichment of ∼1‰ with respect to parent compounds. Acr, Ald, and Sig had similar enrichment in the C3 position, ∆δ13CAae(3) ∼ 28‰ (28.4, 28.2, and 28.0‰, respectively). Sig has the same amount of 13C as the other samples with distinctive enrichment at the C(2) position. It is depleted in 13C at the six ring positions of Sig. Depletion is spread among the six ring positions, diluting the effect on ∆δ13CAae, resulting in a ∆δ13C(ring) reduction by only 0.6‰. The ring moiety (measured directly) had an average standard deviation of SD(∆δ13C(ring)) ) 0.18‰, while positions C(2) and C(3) (computed indirectly) had an average standard deviation of SD(∆δ13C) ) 6.04‰. While this error is expectedly large, sources were nevertheless distinguishable because the individual positions have larger isotope ratio differences and the ring moiety was determined with higher precision, thus providing a good target for sourcing.

Sourcing of Multiple Samples. Ala samples from all sources had measured isotope ratios within the narrow range of ∆δ13C ) 3.16‰, and thus only the sources at the extremes could be distinguished reliably at a (CSIA) precision level of SD(δ13C) ) 0.3‰ with sufficient numbers of replicate analyses. However, the intramolecular values show substantial variability. Flu and Sig had similar intramolecular isotopes, but are still distinguishable from one another since they differ by +3.00 ( 0.15, -6.28 ( 0.96, and 3.31 ( 0.37‰ for the C1, C2, and C3 positions, respectively. Likewise, Ald and Acr had values for all three positions within ∼2‰, with Ald slightly lower in 13C. Overall, PSIA enables Flu and Sig to be distinguished from one another, and both can be distinguished from Ald and Acr. Ald and Acr are also distinguishable from one another, but the correspondence of the PSIA results with the CSIA results suggests that they may derive from a common source that was slightly altered isotopically as, for instance, in fractional crystallization. CSIA for Phe samples indicates that one sample, Aae, clearly differs from the other three, which are in a range of ∆δ13C < 0.1‰. PSIA analysis strongly supports the hypothesis that Acr, Ald, and Sig originate from the same common source. The isotope ratios associated with the C2 and C3 positions are not significantly different. As discussed below, the errors are relatively large due to the way isotope ratios for these positions are calculated; thus, the sensitivity to detect differences is modest. Limits on Sample Size. A useful guideline for routine CSIA analysis is that the analyte mass should be ∼10 ng of C on-column to yield a precision of SD(δ13C of direct determination) < 0.3‰. This minimum level is ∼1 order of magnitude in excess of the statistical limit,25 indicating that other sources of noise limit precision. Two main factors raise minimum sample size requirement in PSIA compared to CSIA. In PSIA, the 10-ng minimum sample amount applies to each position in the target molecule. In other words, the minimum sample size refers directly to C’s at specific positions, which are the “analytes” in PSIA. Assuming GC-Py-GCC-IRMS analysis does not introduce any sources of noise greater than those in CSIA analysis, this factor introduces scaling for molecular loading on column that is linear in the number of (moles of C)/(mole of target molecule). For instance, the minimum amount of Phe (9 mol of C/mol) that must be introduced on column to yield the amount of CO2 similar to 10 ng of Phe is 90 ng. This scale-up of 1 order of magnitude should be considered when assessing minimum sample requirements. The second factor further limiting sensitivity in PSIA analysis is fragmentation efficiency, defined as the molar yield of useful fragment per mole of target molecule on column. In alaninol, both CO and CH4 had an efficiency of ∼40%, among the highest for alaninol. Phenethylamine fragments were obtained at a lower efficiency, with benzene at ∼10% and toluene at ∼30%. In the current study, solution concentrations were set to convenient values that were greater than that required for analysis. From pyrolysis products obtained in this and other studies, we estimate that minimum loads on column are ∼1 µg of Ala with the present configuration, while the lower pyrolysis efficiency for Phe dictates a minimum of ∼3 µg. Error Analysis. Errors for positions requiring calculated isotope ratios were computed using conventional propagation of error equations.34 Phe positions C(2) and C(3) are large by highAnalytical Chemistry, Vol. 77, No. 6, March 15, 2005

1751

precision isotope ratio standards because of the nature of the fragments used to compute these positions. The propagation of error equation for the C(3) position is as follows:

SD(∆δ13C(3)) )

x[

[

] ] [

∆δ13C(Tol) - F(ring)∆δ13C(ring) F(3)

SD(∆δ13C(Tol) - F(ring)∆δ13C(ring)) ∆δ13C(Tol) - F(ring)∆δ13C(ring)

2

+

]

SD(F(3)) F(3)

2

(15)

where F(3) ) 0.14 ( 0.01 and F(ring) ) 0.86 ( 0.05, the relative contributions of the C(3) position and C(ring) moiety to the total benzyl moiety of the Phe molecule. The final calculated error depends on the degree of mismatch between these two factors, as is more easily demonstrated by rearrangement of the bracketed term of eq 15,

SD(∆δ13C(3)) )

x[

[

∆δ13C(Tol) - ∆δ13C(ring) 1-F

(ring)

+ ∆δ13C(ring)

] [

SD(∆δ13C(Tol) - F(ring)∆δ13C(ring)) ∆δ13C(Tol) - F(ring)∆δ13C(ring)

2

+

]

]

SD(F(3)) F(3)

2

(34) Meyer, S. L. Data Analysis for Scientists and Engineers; Wiley: New York, 1975.

Analytical Chemistry, Vol. 77, No. 6, March 15, 2005

CONCLUSIONS Relative intramolecular isotope ratios were calculated for two amino acid analogues; alaninol (from Ala) and phenethylamine (from Phe). In each case, ∆δ13C values were calculated for each carbon position in the compounds using fragments formed via pyrolysis and the isotope ratio for the whole compound. Alaninol samples from Flu and Sig were shown to have the same compound-specific isotope ratio with different ratios at the positionspecific level. Full isotope ratio analysis for a compound with n carbons requires at minimum the compound isotope ratio and n - 1 fragments from pyrolysis, which includes at least n - 1 of the carbons in the compound. The position-specific isotope ratios can be determined using these fragments with mass balance equations and fidelity information. We have demonstrated that GC-Py-GCC-IRMS system for PSIA can be utilized even when the fragments formed are not of ideal fidelity, expanding the number of samples that can be analyzed using the system.

(16)

The denominator of the first bracketed term, “1 - F(ring)”, is the fractional contribution of C(3) carbon to the total benzyl moiety. Because that single position is a small fraction of the whole, the small denominator amplifies the calculated SD to higher values despite the precision of the measured values. For this reason, it is ideal to calculate isotope ratios of a position between small

1752

fragments. The large C(2) error is driven by the C(3) error, which figures into the ∆δ13C(2) computation; additional errors specifically associated with the C(2) measurements do not appreciably contribute to the overall error in ∆δ13C(2).

ACKNOWLEDGMENT This work was funded by NIH grant GM49209. G.L.S. and B.S.P. acknowledge support from NIH training grant DK07158. We thank Sarah Chiang and Sara Metzger for assistance in derivatization reactions and sample preparation.

Received for review October 5, 2004. Accepted December 20, 2004. AC048524V