Ion Composition Elucidation (ICE)
Identification of Ions Produced from Components in a Complex Mixture Using Mass Peak Profiling
Andrew H. Grange and William C. Brumley
U.S. Environmental Protection Agency, National Exposure Research Laboratory, Characterization Research Division, P.O. Box 93478, Las Vegas, Nevada 89193-3478
The characterization of complex mixtures is a goal of environmental, pharmaceutical, synthetic, and biochemists. When performing gas chromatography-mass spectrometry analyses, total ion chromatograms for extracts of complex samples can provide numerous chromatographic peaks. Many of the compounds responsible for the peaks cannot be identified because the background-subtracted mass spectra are degraded by coelution of multiple components. In addition, library mass spectra often are unavailable for all compounds in a mixture. The authors used a high-resolution technique, called mass peak profiling from selected-ion recording data, to characterize a complex sample from a U.S. Environmental Protection Agency Superfund site. The technique was adapted to determine compositions of molecular ions and other prominent ions corresponding to 47 chromatographic peaks.
Several techniques resolve complex mixtures into individual components. Gas and liquid chromatography (GC and LC) and capillary electrophoresis (CE) are the most frequently used separation techniques in environmental analysis. Researchers are studying multidimensional separations to explore n2 and n3 peak capacities resulting from two- (1,2) and three-dimensional (3) separations. In most discussions, peak capacities and other measures of separation efficiency are not assigned to the detector. This omission ignores the fact that separations can be performed not only in the time domain but also in domains such as fluorescence lifetimes, wavelength of absorption-emission, and mass-to-charge ratio (m/z). Separations in these domains are made as chromatographic peaks separated in the time domain enter the detector. Sophisticated devices for transferring effluents from one column to another are unnecessary.
With mass spectrometry (MS) detection, hundreds of components can be resolved based on mass-to-charge ratio. The mass resolution is defined as M/deltaM, where M is the average mass of two mass peak profiles with equal amplitudes resolved with a 10% valley, and deltaM is the mass difference between them. In Figure la, several mass peaks are resolved.
Figure 1. Resolution of mass peaks. Shown are (a) four calculated mass peak profiles separated by 10% valleys with mass scales corresponding to 100, 1,000, and 10,000 mass resolution and calculated mass peak profiles for two ions at (b) 10,000 and (c) 20,000 resolution. The shaded rectangles mark the error limit for single exact mass determinations.
The x-axis labels illustrate that this number of peaks is resolvable throughout smaller mass ranges as the mass resolution increases. At 10,000 resolution, 100 times as many peaks can be resolved as at 100 resolution.
In Figures 1b and 1c, the same mass range is shown at resolutions of 10,000 and 20,000. In Figure 1c, the width of the two mass peaks and the error limits for an exact mass determination, depicted as shaded rectangles, are half as large as those in Figure 1b. At 20,000 resolution, the error limits do not overlap, and the two elemental compositions corresponding to the profiles are distinguishable, assuming only one ion is present. Higher mass resolution generally separates analyte and interference ions and provides greater specificity.
A common application of high-resolution MS is to confirm elemental compositions based on exact mass determinations for synthesized compounds introduced by probe to provide analyte ions for at least 10-100 s (4). A high-resolution MS technique, mass peak profiling from selected-ion recording data (originally used for quality assurance.[5,6]) recently was adapted to determine elemental compositions of ions from compounds that exit capillary GC or LC columns with peak widths of less than 10 s.(7,8) In this technique, mass-to-charge ratios are monitored across mass peak profiles, and the profiles are plotted from the peaks observed in the ion chromatograms.
The characterization of the major components in mixtures from barrels, underground storage tanks, and U.S. Environmental Protection Agency (EPA) Superfund sites is an important analytical goal of the Characterization Research Division of the EPA's National Exposure Research Laboratory. In the study described in this article, we applied mass peak profiling from selected-ion recording data to this mission. Using this technique, the characterization of pharmaceutical, biological, and synthetic mixtures also is feasible.
When chromatographers analyze complex environmental samples by GC-MS with low mass resolution, they often find several library matches for compounds with different elemental compositions for background-subtracted mass spectra. The partial coelution of other components can provide extraneous mass peaks, and mass peaks associated with the analyte and another component can be removed from the background-subtracted mass spectrum that is compared with the library. When the analyte is a minor component, the resulting mass spectrum often is unrecognizable.
For many good quality mass spectra, the analyte spectrum is not in the library, so a correct match is impossible. In these cases, the characterization of components is incomplete. The determination of the elemental composition of the molecular ion or other large-mass ion provides supplemental information for determining the likely identity and source of contaminants found in Superfund sites. Hence, low-resolution mass spectra provide an inventory of the components present, and mass peak profiling from selected-ion recording data can further resolve and characterize these components.
The compositions of ions are determined from data acquired at mass resolutions of 10,000 to 20,000.(7,8) The composition of molecular ions eliminates library matches for incorrect compositions and limits analyses' identities to a finite number of isomers. When analyses target particular analytes, their absence can be documented to a low detection limit, or their presence can be confirmed when other information indicates they might be present.
We studied a complex environmental extract using GC-high-resolution MS using mass peak profiling from selected-ion recording data to determine the compositions of important ions formed from components in the extract that provided chromatographic peaks in the total ion chromatogram. Probe introduction also was used to test mass peak profiling from selected-ion recording data for rapid confirmation of target compounds in the extract and to determine if several ions not corresponding to a GC peak in the total ion chromatogram could be identified.
The sample from a Superfund site in West Virginia was a black, highly viscous liquid at room temperature. We dissolved 34 mg of the sample in 1 mL of methylene chloride and injected 0.1-2 mL of the extract onto the GC column or into the probe tip capillary, depending on the relative abundance of the compound under study and the mass resolution.
We installed a 30 m x 0.25 mm, 0.25-mm df RTx-5 column (Restek, Bellefonte, Pennsylvania) in a Hewlett-Packard model HP 5890 gas chromatograph (Wilmington, Delaware) interfaced with a Fisons model 70-250SE double-focusing mass spectrometer (Danvers, Massachusetts). Our temperature program started at 40 oC for 3 min, then we increased the temperature at a rate of 8 oC/min to 320 oC and held that temperature for 15 min. We heated the probe ballistically to 400 oC to volatilize components as narrow peaks.
We acquired low-resolution (approximately 1000) mass spectra from 45 to 800 Da as components were eluted from the GC column to provide background-subtracted mass spectra for comparison to the 42,000-entry National Bureau of Standards library included on the data system. Because one spectrum was acquired every 5 s, only three or four points defined the narrowest chromatographic peaks. The source temperature was 250 oC, the transfer line was maintained at 280 oC, the filament current was 500 mA, and the photomultiplier tube voltage was 300 V.
With scanning modes, analysts observe a mass range, most of which is baseline. To enhance sensitivity by a factor of 100 for targeted analytes, analysts can use the selected-ion recording mode. In this technique, users monitor a single mass-to-charge ratio that they presume corresponds to the apex of the mass peak profile for each of several ions, and they ignore the baseline regions. To plot a mass peak profile, we monitored multiple mass-to-charge ratios across a single profile and plotted the chromatographic peak area from each ion chromatogram, which provided the profile in Figure 2.
Figure 2. The mass peak profile in (a) was plotted using the chromatographic peak areas determined for (b-f), the five ion chromatograms recorded at different mass-to-charge ratios across the mass peak profile.
We monitored each mass-to-charge ratio listed in the selected-ion recording descriptor that controls data acquisition for 30 ms during each cycle. The selected-ion recording mode increased sensitivity by more than 100-fold relative to normal scanning. A 0.8-s cycle enabled us to take 10 or more samplings across the narrowest chromatographic peaks to better delineate their shapes and resolve closely eluting components. The increased sensitivity permitted use of higher mass resolution, which provided greater selectivity. Grange and coworkers (7) monitored 10 mass-to-charge ratios across the M profile and 5 points across the top portion of a calibration ion profile with 10,000 or 20,000 resolution.
In a different study, Grange and colleagues (8) monitored six points across 60% of the mass range of the M, M+1, and M+2 profiles when exact masses and relative abundances of the M+1 and M+2 partial profiles were required to determine elemental compositions based on five criteria. The five criteria were:
- agreement to within 2.5 ppm of the experimentally determined and calculated exact masses of the M profile,
- agreement to within 2.5 ppm of the experimentally determined and calculated exact masses of the M+1 partial profile of the composition identified,
- agreement to within 2.5 ppm of the experimentally determined and calculated exact masses of the M+2 partial profile of the composition identified,
- agreement to within 5.5% of the relative abundance of the M+1 partial profile relative to the M partial profile, and
- agreement to within 5.5% of the relative abundance of the M+2 partial profile relative to the M partial profile.
They determined the exact masses as the weighted average of several points near the maximum of each profile. The abundances of M+1 and M+2 partial profiles relative to the M partial profile were determined as ratios of the sum of the six points across each partial profile. The photomultiplier tube voltage was 390 V when 10,000 resolution was used and 500 V with 20,000 resolution. See references (7-9) for greater detail.
In our study, we obtained both full and partial profiles. We used the automated procedures developed for the previous studies to prepare multiple selected-ion recording descriptors for each injection, process data, and plot mass peak profiles. We acquired data for as many as 11 ions per injection. With probe introduction, we used single selected-ion recording groups to acquire data.
RESULTS AND DISCUSSION
Figure 3 displays the total ion chromatogram for a 0.5-m injection of the extract acquired using low mass resolution.
Figure 3. A total ion chromatogram (top trace) acquired with low mass resolution for a Superfund sample extract and the m/z chromatogram (bottom trace) acquired with 20,000 resolution for the same extract.
We obtained background-subtracted mass spectra for 47 of the chromatographic peaks. For most compounds, we found either multiple library matches or no matches. Hence, low-resolution MS demonstrated that the sample was a complex mixture, but elemental compositions of molecular or other important ions could not be determined.
Figure 4. The background-subtracted mass spectrum for (a) an unidentified compound and library matches for (b) C10H11N3OS, (c) C19H3O, and (d) C8H8N2S.
Two of the matches lack the m/z 246 ion, but the analyte molecule could contain one of these compositions as part of its structure. Determining the composition of the m/z 246 ion would exclude at least two of the matches and limit the compound's identity to a number of isomers, if the m/z 246 ion were the molecular ion.
Determination of ionic compositions
Because the masses of elements are different, each elemental composition has a unique exact mass. If the error limit in the exact mass determination is small enough, the exact mass corresponds to a single composition.
We determined the compositions for the largest mass ion containing only the most abundant isotopes that corresponded to each chromatographic peak in Figure 3. In most cases, this ion probably was the molecular ion. To ensure that they studied the most abundant analyte ion for each nominal mass, Grange and co-workers (7,8) surveyed a 1600-ppm mass range using 3000 resolution. Figure 5a shows an example.
Figure 5. Mass peak profiles for (a) a perfluorokerosene ion and an analyte ion acquired with 3000 resolution over a mass range of 1600 ppm, (b) the analyte ion acquired with 10,000 resolution over a mass range of 160 ppm, and (c) M, M+1 and M+2 partial profiles acquired with 20,000 resolution and a mass increment of 5 ppm. Also shown are partial profiles for the calibration ion.
They monitored two or three points across each profile, and these provided an estimate of the exact mass. Then they obtained a full mass peak profile at 10,000 resolution across a mass range of 160 ppm, as illustrated in Figure 5b. For five ions with masses of less than 140 Da, only one composition containing carbon, hydrogen, oxygen, nitrogen, phosphorus, or sulfur atoms corresponded to the observed exact mass within an error limit estimated to be 6 ppm for a single determination made at 10,000 resolution.(9)
For larger mass ions, more compositions are possible for a given error limit, so additional criteria are necessary to select the correct composition from a list of possible compositions. The exact masses of M+1 ions containing one +1 isotope or M+2 ions containing a +2 isotope or two +1 isotopes also are unique. Because the isotopic abundances of +1 and +2 isotopes also differ among elements, the abundances of M+1 and M+2 profiles relative to the M profile differ among compositions. These differences are large enough that one or more criteria based on the exact masses of M, M+1, and M+2 profiles and the relative abundances of the M+1 and M+2 profiles obtained using mass peak profiling from selected-ion recording data at 20,000 resolution can generally eliminate all but the correct composition for ions with masses up to 600 Da as described in Reference 9.
Using a single selected-ion recording descriptor, which is limited to 24 mass-to-charge ratios, we acquired data at 20,000 resolution to construct M, M+1, and M+2 partial profiles like those in Figure 5c, and we used these profiles to determine exact masses and relative abundances. These values were compared automatically to those calculated by a profile generation model (8,9) for all possible compositions containing carbon, hydrogen, oxygen, nitrogen, phosphorus, and sulfur atoms to eliminate all but the correct composition. Generally, the maximum error limits associated with one determination at 20,000 resolution (3 ppm for exact masses and 11% of relative abundances) provided exclusion of all other possible compositions. The smaller error limits estimated for triplicate determinations (2.5 ppm and 5.5%) were necessary to identify two ions. Reference 9 fully describes the model.
To prepare a selected-ion recording descriptor to monitor M, M+1, and M+2 profiles, Grange and Brumley (9) chose a hypothetical composition from the list of possible compositions of M based on its exact mass. When they used the correct composition, the three partial profiles were well centered in the observation windows, and the composition was confirmed as correct based on all five criteria unless interferences distorted the shape or abundances of the M+1 or M+2 partial profiles. The correct hypothetical composition was chosen routinely by assuming larger mass ions were produced from compounds compositionally similar to the smaller mass ions already identified. Most ions studied contained NS, N2, or N2S2 and none contained a phosphorus atom. In general, three injections were necessary for each ion to provide the required profiles, but we observed as many as 11 ions for a single injection. Using in-house automated procedures and a profile generation model, we were able to prepare selected-ion recording descriptors, process data, plot profiles, and compare calculated values in a straightforward and routine manner. In a few cases, however, the original hypothetical composition was incorrect, and we performed additional injections.
The identification of low-mass ions provided information helpful for characterizing higher-mass ions from other components. The composition determined for m/z 246 (C14H18N2S) was consistent with the third library match in Figure 4d. Figure 6a shows the background-subtracted mass spectrum, and Figure 6b illustrates an excellent library match for a compound with a composition of C13H16N2S.
Figure 6. The background-subtracted mass spectrum for (a) an unidentified compound and (b) the library match for N-cyclohexyl-2-benzothiazole.
The m/z 150 ion appears to result from the loss of C6H10 (cyclohexyl group). The m/z 164 ion in the unknown analyte's spectrum could then be the result of loss of the same atoms from a similar compound with a methyl group attached to the aromatic ring or to the amine nitrogen.
Interferences with the M+1 and M+2 profiles distorted exact masses or relative abundances used for criteria testing with GC introduction for 9 of the 42 ions for which partial profiles were plotted. Despite these interferences, we determined the compositions based on the remaining unaffected exact masses and relative abundances. Table 1 lists the tentative identifications of compounds we found when we obtained a single good library match with the background-subtracted mass spectrum that was consistent with the composition of the ion studied. The retention times for standards could provide confirmation for the tentatively identified compounds.
Table 1. Nominal Masses, Compositions, and Tentative Identifications of Ions Corresponding to Chromatographic Peaks in the Total Ion Chromatogram
aCriteria tested for m/z 162 ion due to low signal.
bInsufficient to exclude all other compositions not containing phosphorus atoms.
cFor this ion, an absence of phosphorus was assumed.
We observed interferences with both calibration and analyte ions from the calibrant perfluorokerosene, column bleed at high column temperatures, and other compounds in the extract. Figure 7a shows the profiles of three ions with a nominal mass of 281 Da that were resolved at a resolution of 5000.
Figure 7. Mass peak profiles (a) acquired with 5000 resolution over a mass range of 800 ppm for an ion from the calibrant perfluorokerosene, an ion due to column bleed, and an ion from an analyte. The ion chromatograms (b-d) correspond to the maxima in the profiles.
Figures 7b-7d show the ion chromatograms corresponding to the maxima in the profiles. The source of each profile is indicated by the appearance of the ion chromatograms. The traces show a constant signal arising from the C6F11+ ion of perfluorokerosene, a signal from the column bleed ion determined to be C7H21O4Si4+ that increased on the temperature ramp, and a chromatographic peak from a C18H19NS+ ion produced from a compound in the extract. We blocked the ion beam with the valve between the source and analyzer sections of the instrument to induce the baseline level and provide a simulated chromatographic peak for the perfluorokerosene ion. This practice enabled automated area integration under the perfluorokerosene traces.
At 3000 resolution, the adjacent profiles in Figure 7a would overlap and interfere with each other. Accordingly, interferences were more common at 3000 resolution that at 20,000 resolution. These interferences were indicated by partial profiles lacking a maximum or by relative abundances greater than those of all possible compositions. When the overlap between the normalized ion chromatograms used to construct partial profiles was incomplete, the presence of an interference was confirmed.
Ions identified using probe introduction
Three factors made the identification of large molecular weight compounds in the extract difficult with GC introduction: their abundance was low, the chromatographic peaks became broad for late-eluted compounds, and the column bleed level was high on the 320 oC plateau. Probe introduction eliminated column bleed and provided narrower peaks for these compounds. We observed fractionation, and heavier components volatilized later. However, the separation was very incomplete, and interference occurred more often. In addition, more compositions were possible for the larger mass ions we observed with probe introduction, and we made fewer unequivocal identifications when interferences distorted one or more of the values used for criteria testing. Despite these difficulties, we were able to identify four additional ions related to those identified with GC introduction.
Screening for a class of compounds
Several of the mass spectra displayed a prominent C8H7NS+ (m/z 149.0299) ion. The first compound that provided the ion, according to a good library match, was 2-methylbenzothiazole. Thus, this fragment ion is associated with compounds containing the benzothiazole group. Benzothiazole is toxic (10), and its derivatives are used as rubber accelerants (11). In Figure 3, the bottom trace acquired with 20,000 resolution corresponds to the maximum in the partial profile of the C8H7NS+ ion. At this resolution, the interference from the common C8H5O3+ ion (m/z 149.0239) formed from phthalates would be excluded. At least 25 chromatographic peaks in the trace correspond to peaks in the total ion chromatogram for compounds containing both a nitrogen and a sulfur atom. We determined the exact masses and relative abundances throughout 28 time domains, some of which contained multiple peaks. Three or more criteria not subject to interference permitted rejection of four other possible compositions based on carbon, hydrogen, oxygen, nitrogen, phosphorus, sulfur, and silicon atoms for 20 domains. At 20,000 resolution, mass peak profiling from selected-ion recording data confirmed the presence of compounds containing the benzothiazole group.
Searching for a trace compound
Table 1 lists only those ions produced from major components in the mixture. For trace components, we obtained poorer background-subtracted mass spectra, and interferences with M+1 and M+2 partial profiles were more common. For one trace component, we found a library match for strychnine, which had m/z 334 as the most abundant ion. Because strychnine is highly toxic, it piqued our curiosity. We further investigated the m/z 334 ion and determined its composition to be C24H18N2. Each of the three exact masses ruled out strychnine (C21H22N2O2) as the source of the m/z 334 ion. In this example, we observed no interferences, and the composition met all five criteria for C24H18N2.
Confirming or refuting the presence of compounds using probe introduction
We used probe introduction with 20,000 resolution to monitor M, M+1, and M+2 partial profiles to screen nine compositions listed in Table 1. Each composition differed in the number of oxygen, nitrogen, or sulfur atoms and represented a range of relative concentrations; some corresponded to small and others to large chromatographic peaks. The fifth column in Table 1 illustrates that fewer criteria were met with probe introduction. A unique composition for the ion with a mass of 326 Da was unidentified; for larger ions, more compositions were possible and all of the criteria often are needed to exclude all but the correct composition. We determined the unique compositions for eight ions with masses of less than 300 Da based on the criteria for which interferences were unimportant.
We performed a probe introduction of the extract while monitoring the M, M+1, and M+2 partial profiles of strychnine. Using 20,000 resolution discriminated against the M, M+1, and M+2 ions produced from the compound responsible for the m/z 334 ion. Only a low noise level was recorded as the probe tip was heated. We found no strychnine in the sample.
The high specificity and sensitivity of high-resolution MS using mass peak profiling from selected-ion recording data allowed the characterization of the major components in a complex mixture. We determined the unique, elemental compositions for ions from 47 components that provided chromatographic peaks and for several other compounds that did not. We tentatively identified a few of the low-mass compounds using library spectra, and for these, the determination of elemental compositions of ions provided confirmatory evidence. This example suggests that mass peak profiling from selected-ion recording data could determine elemental compositions of molecular or fragment molecular ions to help identify components in other complex mixtures, including drug candidates in combinational libraries and metabolic products in biological fluids.
Our monitoring of the partial profiles of a characteristic fragment ion indicated the presence of many molecules containing the benzothiazole moiety. We also observed the absence of strychnine at a low detection limit, which refuted a poor library match for strychnine with a low-resolution mass spectrum.
We could compare the exact masses, relative abundances, and retention times of identified ions with those from compounds in chemical inventories and purchasing records of industrial facilities. This information would provide strong evidence for determining the source of compounds dumped on EPA Superfund sites.
A.W. Moore Jr. and J.W. Jorgenson; Anal. Chem. 67, 3448-3455 (1995).
Z. Liu, D.G. Patterson Jr., and M.L. Lee; Anal. Chem. 67, 3840-3845 (1995).
A.W. Moore Jr. and J.W. Jorgenson; Anal. Chem. 67, 3456-3463 (1995).
M.L. Gross; J. Am. Soc. Mass Spectrom. 5, 57 (1994).
J. Roboz, A. Allan, and M. Chu; Proceedings of the 34th ASMS Conference on Mass Spectrometry and AlLied Topics American Society for Mass Spectrometry, Santa Fe, New Mexico, (1986), pp. 212-213.
A.R. Allan and J. Roboz; Rapid Commun. Mass Spectrom. 2, 246-249 (1988).
D.V. Sweet, Ed., Registry of Toxic Effects of Chemical Substances (U .S. Department of Health and Services, Cincinnati, Ohio, NIOSH publication number 93-101-3, 1993), p. 25513.
J.L. Lewis Jr., Ed., Hawley's Condensed Chemical Dictionary (Van Nostrand Reinbold Co.), New York, 12th ed., (1993) p. 132.
Acknowledgment: This work was performed under Contract 68-C5-0091 while Andrew Grange was employed by Lockheed Environmental Sciences and Technologies, 980 Kelly Johnson Drive, Las Vegas, NV 89119.