Ion Composition Elucidation (ICE)
Identifying Endocrine Disruptors by High-Resolution Mass Spectrometry
Andrew H. Grange and G. Wayne Sovocool
Environmental Sciences Division, NERL, U.S. EPA, PO Box 93478, Las Vegas, NV 89193-3478
Phone: (702) 798-2137
Fax: (702) 798-2142
Identification of trace amounts of compounds found in the environment, including endocrine disruptors, is difficult. Two analytical tools developed by the U.S. EPA's Environmental Sciences Division, Mass Peak Profiling from Selected Ion Recording Data (MPPSIRD) and a Profile Generation Model (PGM), utilize high resolution mass spectrometry to determine elemental compositions of molecular and fragment ions. These tools provide structural details of molecules and limit the identity of compounds to a manageable number of isomers. Compounds tentatively identified through literature searches can be confirmed by comparison of analyte mass spectra and retention times with those of standards. The enhanced speed and sensitivity provided by MPPSIRD allow elemental compositions to be determined for ions from compounds entering the mass spectrometer as chromatographic peaks in amounts too small for FTIR and NMR to be used. Application of MPPSIRD and the PGM for solving real world environmental contamination problems is reviewed, and a detection limit for ethinylestradiol with probe introduction is determined.
The EPA is currently interested in human and ecosystem exposure to endocrine disruptors (1) - compounds that interfere with endogenous hormone systems. Possible endocrine disruptors in the environment include certain pesticides, industrial by-products, and pharmaceuticals. Such chemicals could be found in surface water or wells as a result of agricultural run off (2), leaching from contaminated sites (3,4), or in treated wastewater discharged from urban areas (5).
Biologically based assays are often used to screen for the presence of endocrine disruptors. While these tests are sensitive, they are non-specific and not inclusive. Cross-reactivity is common but can be an advantage when searching for unidentified endocrine disruptors. Once a positive result is obtained from a biologically based assay, it is important to identify the compound or compounds responsible for the reactivity. Toxicological studies can then be performed for the identified chemicals. Also, when screening for a specific compound, possible cross-reactivity suggests that a more specific confirmatory technique should be applied to some fraction of the samples showing a positive result for a target analyte.
Limitations of Low Resolution Mass Spectrometry (LRMS)
When possible, compounds to be identified are introduced into the ion source of a mass spectrometer from a gas chromatography (GC) column or from a direct insertion probe and are ionized by electron impact (EI). The electron energy most commonly used, 70 ev, is sufficient to induce fragmentation of the molecular ion. Any remaining molecular ions and the resulting fragment ions provide the mass spectrum of a compound, which is a histogram of ion abundances as a function of mass-to-charge (m/z) ratio. A compound can be identified by matching its mass spectrum with one in a reference library of mass spectra. The low mass resolution provided by quadrupole or ion trap mass spectrometers is not a limitation when only one good library match is found.
If several plausible matches are found, however, low mass resolution is insufficient to distinguish between matches for compounds with different elemental compositions. In addition, such libraries have a limited number of entries and often no matches are found. Missing from mass spectral libraries are most waste products from industrial syntheses. Only half of the 100 drugs most prescribed in the US in 1997 (6) were found in the NIST library (7). To be included in the library, gas chromatographic or direct probe introduction of a drug into the ion source must provide a mass spectrum induced by electron impact ionization. Many drugs and environmental contaminants, however, are too polar, ionic, thermolabile, or involatile to be introduced by GC or probe. Liquid sample introduction using electrospray ionization (ESI) or atmospheric pressure chemical ionization (APCI) are most often used to analyze such compounds (8). These low-energy ionization processes provide little or no fragmentation unless additional voltages are applied to components of the sources. No commercial mass spectral libraries for ESI or APCI are available. Clearly, LRMS is insufficient for identifying many compounds found in the environment.
High Resolution Mass Spectrometry (HRMS)
For double focusing mass spectrometers, mass resolution is defined as R = M/deltaM for a 10% valley between partially overlapping profiles of equal height, where M is the average center mass of the two profiles and deltaM is the mass difference between the maxima of the two profiles (9). At mass resolutions of 10,000 or more, the exact mass of the molecular ion formed from a compound can be determined to within a few ppm. The error in the measurement expressed as ppm is (Me - Mt)/Mt x 106, where Me is the experimentally determined exact mass and Mt is the theoretical exact mass. If the ion's mass is less than 150 amu and contains only C, H, N, O, P, or S atoms, its exact mass usually corresponds to a unique elemental composition. Structural features of the molecule can be deduced from the elemental compositions of the molecular and fragment ions.
Historically, use of HRMS for characterizing or identifying environmental contaminants has been limited. The number of elemental compositions possible for an exact mass increases rapidly with an ion' s mass, the number of elements considered, and the error limits of the exact mass determination. Most environmental contaminants have molecular weights greater than 150 amu, and for a given exact mass, multiple compositions are possible.
Using double focusing mass spectrometers, exact masses were determined for analyte ions using peak matching techniques or electric full scans with high mass resolution. Both techniques were too slow to acquire the necessary data for compounds that eluted into a mass spectrometer from a GC over an interval of about 10 sec.
Endocrine disruptors present at very low concentrations in lakes, streams, or drinking water could impact biota or influence human fetal development (10). Thus, low detection limits are desirable. Unfortunately, there is an inverse relationship between mass resolution and sensitivity; as the resolution is increased by narrowing the entrance and exit slits to provide accurate exact masses and to increase selectivity, fewer ions reach the detector.
To overcome the scan speed and sensitivity limitations, the EPA's Environmental Chemistry Branch (ORD/NERL, Las Vegas) developed a HRMS data acquisition technique, Mass Peak Profiling from Selected Ion Recording Data (MPPSIRD) (11,12). For full scans, all rn/z ratios across a wide mass range are monitored sequentially, starting at one end of the mass range and finishing at the other. With selected ion recording (SIR), a limited number of rn/z ratios are monitored for a longer time, and jumps are made between the rn/z ratios of interest. A mass resolution of 10,000 and SIR are used in EPA Methods 8290 (13) and 1613 (14) for quantitation of polychloro-dibenzofurans and polychloro-p-dibenzodioxins to achieve low detection limits while discriminating against interferences. The rn/z ratios at the maxima in the mass peak profiles of analyte and calibrant ions are monitored. Two profiles are monitored for each analyte, and the ratio of the signals indicates a lack of major interferences when it falls within 15% of the expected value. Beyond these methods, researchers use abbreviated full scans across narrow mass ranges that include the full mass peak profile of each analyte ion to reveal interferences (15,16). An incorrect exact mass or deviation of the profile from a Gaussian shape results from interferences. Increased sensitivity is realized when multiple rn/z ratios are used with SIR to plot the profiles of analytes for the same purpose (17). These data acquisition techniques provided a quality assurance tool for dioxin analyses. At the ESD, automated procedures were developed to prepare SIR descriptors (which contained a list of the rn/z ratios to be monitored), acquire SIR data, plot profiles, and calculate exact masses and relative abundances. From these data, elemental compositions of ions were determined.
Compared with full scanning, MPPSIRD provided a 170-fold lower detection limit (6 fg) for a tetrachlorobiphenyl eluting from a GC with 10,000 resolution (11). At 20,000 resolution, MPPSIRD provided a six-fold faster cycle time (0.8-1.0 sec/cycle) compared with conventional mass spectrometric scans. Consequently, data can be acquired for ions from compounds that enter the ion source as chromatographic peaks. In addition, routine use of up to 20,000 resolution provides excellent discrimination against interferences having the same nominal (integer) masses, but different elemental compositions than the ion under study (18).
To increase the largest rn/z ratio for which unique elemental compositions could be determined from 150 to 600 amu, a Profile Generation Model (PGM) (19) was written in QuickBasic to consider mass peak profiles with rn/z ratios 1 and 2 amu larger than that of the ion of interest. Associated with the molecular ion (M), are M+1 and M+2 profiles that arise from the presence of heavier isotopes. If M contains C, H, N, O, P, or S, then 13C, 2H, 15N, 17O, 18O, 33S, or 34S are considered. Ions containing one atom of an isotope heavier by 1 amu than the most common isotope contribute to the M+l profile, while ions having one atom of an isotope heavier by 2 amu or two atoms of isotopes heavier by 1 amu contribute to the M+2 profile. In addition to the exact mass of the M profile, the model calculates the exact masses of the M+1 and M+2 profiles and the abundances of the M+1 and M+2 profiles relative to the M profile. Comparison of calculated and experimental values provides four additional criteria for distinguishing between possible elemental compositions of an ion.
MPPSIRD was developed using a VG 70-250SE double focusing mass spectrometer. Recently, a new Finnigan MAT 900-trap hybrid mass spectrometer was installed in our laboratory. The VG manual uses the term "selected ion recording" to describe acquiring data only at discrete m/z ratios, while Finnigan MAT's term for the same technique is "multiple ion detection" (MID). To maintain continuity in the published literature, we continue to call the technique MPPSIRD even when data are acquired with the Finnigan instrument.
In Figure 1a are shown ion chromatograms corresponding to the maxima in the molecular (M), M+1, and M+2 mass peak profiles for an analyte and for the profiles of two calibration ions from the calibrant, perfluorokerosene (PFK).
Figure 1. Ion chromatograms (a) from which the areas under the chromatographic peaks correspond to the maxima in the five partial mass peak profiles in (b).
The areas under the chromatographic peaks and those for m/z ratios acquired at 10-ppm intervals on either side of the maxima were plotted to provide the partial profiles in Figure 1b. The analyte was volatilized from a heated probe into the ion source, while a constant level of PFK entered the ion source from a heated reservoir. The baseline excursions in the ion chromatogram for the calibrant occurred when the shield voltage in the ion source was set to 0 V for 5 sec. Automated procedures determined the area between the baseline excursions, which defined a simulated chromatographic peak for the ever-present calibration ion. Integration of the areas under the wide peaks in Figure 1a obtained as the analyte was volatilized from a heated probe improved the signal-to-noise ratio and contributed to the enhanced sensitivity of MPPSIRD. Before data were acquired for each cycle, the data system software locked onto the center mass of the profile of the first calibration ion to compensate for calibration drift, thereby avoiding loss of resolution in the plotted mass peak profiles despite integration of the peak areas over a long interval.
The five partial profiles in Figure 1b were acquired with a mass resolution of 10,000. The two smaller displays show partial profiles for the calibration ions, which were each plotted from five m/z ratios. The three partial profiles for the analyte were each plotted from seven m/z ratios that span 70% of the mass range of each profile. Monitoring partial rather than full profiles permits study of five ions simultaneously using all 31 m/z ratios available in a MID descriptor. Exact masses of the M+1 and M+2 profiles and their abundances relative to the M profile were determined from the partial profiles. The weighted average of the areas defining each partial profile provided its exact mass, and ratios of the sums of the areas yielded the relative abundances.
Files to be executed by the MAT 900 data system were prepared by a Lotus 123 (v2.2) program on the personal computer (PC) that contains the ion trap data system after the user specified the center masses of the analyte profiles, the mass resolution used to set the mass increments, and the data file name. These files were transferred to the mini-computer that controls the Finnigan MAT 900 using Ftp, a DOS-based file transfer program supplied by Finnigan, where their execution prepared the MID descriptors, induced the baseline excursions, and, after data acquisition, prepared an ASCII file with the m/z ratios and chromatographic peak areas. After transfer of the ASCII file back to the PC, Lotus procedures prepared and plotted the profiles using WordPerfect 5.1. DOS batch files provided automated transition between Lotus, Ftp, and WordPerfect on the PC.
For an unidentified ion, an exact mass is determined for the full M profile at 10,000 resolution. This exact mass and the possible elements comprising the ion are entered into the PGM, and a table listing all possible compositions of the ion within the error limits of the mass determination is provided. (To shorten the list, at least 1/3 of the mass is usually assumed to be from C atoms.) Based on any available knowledge of the sample, the user selects from the list a hypothetical composition and enters the corresponding exact masses of the M, M+1, and M+2 profiles into the automated procedure that prepares the MID descriptor to acquire the data pictured in Figure 1.
The five experimental values are then entered into the PGM to provide a table in which ranges of acceptable values that account for errors associated with the use of partial profiles (19) are calculated and compared to the observed values.
Table 1. Comparison of Experimental and Theoretical Values*
*Possible compositions based on the exact mass of the molecular ion and the error in its determination; calculated mass defects for the M, M+1, and M+2 partial profiles; and calculated relative abundances for the M+1 and M+2 partial profiles assuming the profiles' center masses for each composition are used in the MID descriptor, or in parentheses, the range expected if the center masses of the hypothetical (last) composition are used.
The mass defects (fractional masses) are listed rather than the full 7-digit exact masses to allow a larger font. The last line in Table 1 lists average experimental values from triplicate determinations. An "X" next to an entry indicates disagreement with the predicted value, and the composition is rejected. In this example, as is generally the case, only one composition, C20H2402, was consistent with the five measured values. Automatic interpretation by the PGM of the data acquired using MPPSIRD provides compositions of ions with m/z ratios up to 600 amu. This capability expands the utility of HRMS to a much larger fraction of environmentally significant compounds.
Municipal Well Pollutants
In a recent example, MPPSIRD and the PGM were used to determine the elemental composition of several isomers found in a municipal well that serviced Toms River, NJ, where an increased incidence of childhood cancer had been observed (3). No mass spectra similar to the one in Figure 2a were found in the NIST library.
Figure 2. Mass spectra for an isomer in a well water extract (a) and for a standard with the same retention time (b).
To illustrate the reduction in possible compounds that results from determining the elemental composition of the molecular ion, the library was searched for m/z 210 and for its composition, C14H14N2. For this mass, 342 entries were found, but for the composition, the possibilities were reduced to 13. Determination of compositions of fragment ions and neutral losses based on their exact masses and the elemental limits established by the molecular ion revealed whether fragment ions contained 0, 1, or 2 N atoms. This information and review of library mass spectra for smaller molecules with hypothesized sub-structures suggested that the isomers contained a tetralin core with a cyano group and an alpha-cyanoethyl or beta-cyanoethyl group attached to it. A literature search for the limited number of possible isomers that remained revealed that polymer synthesis of styrene and acrylonitrile probably produced the isomers. Three of the five isomers in the well water provided the same mass spectra and GC retention times as three isomers in a sample from a similar industrial synthesis. The concentration of the isomers in the well water extract was too low for study by FTIR or NMR.
This example illustrates that information in addition to fragment and molecular ion compositions is generally required to identify compounds. However, the elemental compositions greatly limit the number of isomers that are possible and the time required to search the literature for such isomers.
Characterization of a Superfund site sample
A small amount of a tar-like sample from a Superfund site in W. Virginia was dissolved in methylene chloride and injections were made onto a 30-m, 0.25-mm i.d., 0.25-mm film thickness, RTx-5 column (18). The GC temperature program started at 40 oC for 3 min and increased by 8 oC/min to 320 oC, where it remained for 15 min. Compounds were investigated that were responsible for 47 peaks in the total ion chromatogram (TIC), 35 of which are shown in Figure 3.
Figure 3. Portion of a total ion chromatogram (upper trace) acquired with 10,000 resolution and an ion chromatogram for the C8H7NS+ ion (m/z 149.0299) from alkylated benzothiazoles (lower trace) acquired with 20,000 resolution.
A few low-mass compounds were identified from library matches of mass spectra alone, but these compounds accounted for only a small fraction of the total signal from the sample. Using a mass resolution of 20,000, MPPSIRD and the PGM were used to determine the elemental compositions of the presumed molecular ion in each mass spectrum for the remaining chromatographic peaks. The composition labels in Figure 3 demonstrate that most of the ions contained at least one N atom and one S atom. In addition, the ion chromatogram in Figure 3 (lower trace) was obtained using 20,000 resolution for the C8H7NS+ ion (149.0299 amu) characteristic of alkylbenzothiazoles. This mass resolution discriminated against the common C8H5O3+ ion (m/z 149.0239) from phthalates, although no phthalates were present. The same retention time for peaks in the TIC and the ion chromatogram indicated that many of the compounds contained the benzothiazole moiety. Benzothiazoles are used by the dye and rubber industries. These results were consistent with the known source of this contamination. This study provided elemental compositions for significant ions formed from compounds that accounted for most of the sample's mass and thereby adequately characterized the sample for determining the origin of the waste.
This sample demonstrated the importance of the wide linear dynamic range afforded by double focusing mass spectrometers. The areas under the peaks in the TIC in Figure 3 differed by more than a factor of 10, and the area under the M+2 profile was as small as 1% of the area under the M profile for ions not containing S atoms.
For the same sample (18), elemental compositions were determined for several highermass compounds that did not elute from the GC column by introducing the sample extract into the ion source with a heated direct insertion probe. High mass resolution was used to separate ions in the mass domain, where separation of neutral compounds prior to ionization by GC had failed. This work suggested that using a heated probe for rapid screening of target analytes, including endocrine disruptors such as ethinylestradiol (EED), could be feasible. EED is a potent endocrine disruptor, routinely used for birth control. Treated wastewater entering Lake Mead, the drinking water supply for Las Vegas, could possibly contain EED (20).
In Figure 1b are shown partial profiles obtained for 10 ng of EED volatilized from the probe. The average values from three such determinations were entered into the PGM to provide Table 1. The elements considered were C, H, N, O, P, S, and F. If such results were obtained for a sample, investigation of characteristic fragment ions would provide additional evidence for the presence of EED. Because data from both the M+1 and M+2 profiles are needed to reject all but the correct composition, the detection level is about 50 times higher than one based on the molecular ion alone. However, the correct exact mass for C20H2402 would still provide a more specific screen for EED than a biologically based assay and would be susceptible to fewer, and different, interferences.
To determine a detection limit based on the molecular ion alone, the detector sensitivity was increased and 1, 10, and 100 pg amounts of the analyte were inserted on the probe. MPPSIRD was used to plot the full M profile. As the concentration of analyte introduced into the source was reduced, the importance of interferences with the same nominal mass increased. For the full profile in Figure 4a acquired with 10,000 resolution, the profile of a lower-mass interference was observed to partially overlap that of the analyte.
Figure 4. Full mass peak profiles obtained with (a) 10,000 resolution and (b) 20,000 resolution for ethinylestradiol.
For lower concentrations, the signal due to the analyte would be obscured by the interference, and an accurate mass for the analyte could no longer be obtained. For the profile in Figure 4b acquired with 20,000 resolution, this interference was no longer important. However, a 3-fold loss in ion abundance accompanied the 2-fold increase in resolution and increased the detection limit. Although 10 pg was readily detected with 20,000 resolution, a laboratory contamination level of the same magnitude was discovered. A molecular or fragment ion with the same elemental composition as the analyte ion was produced from the contaminant. EED had been weighed out a few feet from the instrument a few weeks earlier and was probably responsible for the contamination. Even so, detection of 10 pg above the contamination level corresponds to a detection limit of 1 ppt if the EED in 1mL of water were concentrated into a 0.1 mL extract followed by injection of 1 æ of the extract into the probe tip capillary for analysis.
These studies suggest that MPPSIRD and the PGM will be powerful tools for identifying endocrine disruptors. Initial attempts to identify such compounds will focus on analysis of tertiary treated wastewater that flows into Lake Mead, the drinking supply for Las Vegas. Concentrations of consumer product wastes, pharmaceuticals, lawn and garden pesticides, and their metabolic or waste treatment products should be much greater before dilution in the lake. In another chapter of this book, Snyder et al. report separating numerous unidentified compounds in the treated water by HPLC. As the examples above illustrated, using MPPSIRD and the PGM with separation by GC to determine elemental compositions of ions is routine in our laboratory. Before trying to identify the compounds observed by Snyder, we must first interface HPLC to our Finnigan MAT 900S-trap mass spectrometer through the ESI and APCI sources. Exact masses are determined by calibrating between two known masses from calibrants that bracket the mass of the analyte ion. A problem with liquid sample introduction is finding appropriate concentrations of mass calibrants that provide ions under the same conditions as the analytes. No nearly universal calibrant such as PFK, which provides ions every 12-14 amu when gas phase sample introduction is used, is available for liquid sample introduction. To avoid this problem, we have developed a methodology to determine elemental compositions of ions without using mass calibrants. Elemental compositions are determined from relative abundances of M+1, M+2, F+1, and F+2 ions, exact mass differences between ions, and the appearance of the M+1 or F+1 profiles (21). This methodology requires infusion of previously separated analytes. We anticipate that using parts of this new approach and a single mass calibrant to achieve lock-on by the data system will allow determination of elemental compositions of ions from compounds entering the ESI or APCI source as HPLC peaks. At that time, we will be prepared to investigate compounds separable by either GC or HPLC after extraction from waste streams, lakes, soils, and other sample matrices.
Government Performance and Results Act (U.S. EPA), Goal 8, Objective 3, Subobjective 1, Endocrine Disruptors, #8.
Halling-Sorensen, B.; Nielsen, S.N.; Lanzky, P.F.; Ingerslev, F.; Lutzhoft, H.C.H.; Jorgensen, S.E.; Chemosphere 1998, 36,357-393.
Ferguson, P.L.; Grange, A.H.; Brumley, W.C.; Donnelly, J.R.; Farley, J.W.; Electrophoresis , 1998, 19, 2252-2256.
Buser, H.R.; Muller, M.D.; Theobald, N; Environ. Sci. & Technol., 1998, 32, 188-192.
NIST Standard Reference Database IA, NIST Mass Spectral Search Program and NIST/EPA/NIH Mass Spectral Library, Ver. 1.0, For Use with Microsoft Windows, U.S. Dept. of Commerce, Technology Administration, National Institute of Standards and Technology, Standards Reference Data Program, The NIST Mass Spectrometry Data Center, Gaithersburg, MD 20899, January 1995.
Cole, R.B., Ed.; Electrospray Ionization Mass Spectrometry, Wiley: New York, 1997, 9, 324.
White, E.A.; Wood, G.M.; Mass Spectrometry: Applications in Science and Engineering , Wiley: New York, 1986, 163-164.
Method 8290, Test Methods for Evaluating Solid Wastes, 18 #2, SW -846, US EPA, Office of Solid Waste and Emergency Response, Washington, DC, 1986.
Method 1613, National Center for Environmental Publications and Information, Cincinnati, OH, 1994.
Tong, H.Y.; Giblin, D.E.; Lapp, R.L.; Monson, S.J.; Gross, M.L. Anal. Chem. 1991, 63, 1772-1780.
Tondeur, Y.; Hass, J.R.; Harvan, D.J.; Albro, P.W. Anal. Chem. 1984, 56, 373-376.
Allan, A.R.; Roboz, J. Rapid Commun. Mass Spectrom. 1988, 2, 246-249.
Raloff, J.; Sci. News 1998, 153, 187.