Ion Composition Elucidation (ICE)
Identification of Unanticipated Compounds 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
Localized outbreaks of acute illness could result from deliberate addition of unanticipated compounds into water, air, or food.  Cancer clusters resulting from long-term exposure to trace-levels of compounds are more difficult to detect, but can be revealed by epidemiological studies. To find the causative agents in either case, the compounds in sample extracts must be identified before any contribution they make toward the observed malady can be assessed.  Most trace-level compounds will pose little or no risk.  However, all compounds unique to the locality should be identified, quantified, and classified by their toxicity to ensure that no toxic compounds have been overlooked.
Sample extracts are often complex mixtures.  Separation of components in time by gas chromatography followed by mass selection of ions using mass spectrometry (GC/MS) provides distinctive mass spectra for many organic compounds.  Mass spectrometer data systems permit automated searches of mass spectral libraries for each compound's mass spectrum.  When a single good match is found, a standard can be purchased, and its mass spectrum and retention time on the column can be compared with those of the analyte to confirm its identity.  But two problems with this strategy often occur: multiple library matches may provide several candidate compounds,1 or worse yet, no matches may be found, since mass spectra for most compounds are absent from the libraries.
In these cases, determining the numbers of atoms of each element that comprise the molecular ion (a molecule missing one electron) greatly reduces the number of compounds that could produce the mass spectrum.  Determination of the elemental compositions of the fragment ions produced by electron impact ionization further constrains the number of possible compounds so that searches of the commercial and chemical literature for each compound become feasible. Converging toward a compound's identity based on this knowledge is illustrated by Figure 1. (ICE will be described later.)
|Figure 1.  Target diagram illustrating the dependence of the number of possible compounds on measurements of exact masses and relative abundances.  Once the molecular ion composition is known, determination of the fragment ion compositions limits the number of possible isomers.|
Quadrupole and ion trap mass spectrometers are most commonly used for GC/MS.  These instruments provide insufficient resolving power to separate mass peaks from different combinations of atoms that provide the same nominal (or integer) mass.  For example, in Figure 2 are shown three mass peak profiles with the same nominal mass (m/z 281) that were acquired with a double focusing mass spectrometer using a resolving power of 5000 (10% valley definition).2*
*Assume two mass peak profiles with equal abundance partially overlap with a 10% valley (10% of the maxima) between them.  The resolving power is the average mass of the two profiles divided by the mass difference between the maxima (Mave/delta m).
Figure 2.  Mass peak profiles plotted from selected ion recording data acquired with 5000 resolving power for three ions with a nominal mass of m/z 281.  The three ion chromatograms display (top) a steady signal from a calibrant ion, (middle) a slowly increasing signal from a column bleed ion as the oven temperature increased, and (bottom) a chromatographic peak due to an eluting analyte ion.
The profiles were plotted from selected ion recording data, as described later.  They arose from the C6F11+ ion of the calibrant (perfluorokerosene), the C7H21O4Si4+ ion from column bleed, and the C18H19NS+ ion produced from a compound in a sample extract.  The measured exact masses of the three ions are provided above each profile.  The calculated exact mass for the analyte ion was 281.1238 atomic mass units (amu), 2.4 millimass units (mmu) (8.5 parts per million of the exact mass [ppm]) less than the measured value, and 71 mmu (253 ppm) and 138 mmu (490 ppm) greater than the measured masses for the other two ions.
Based on this exact mass
measurement, can we conclude C18H19NS+
is the analyte ion composition? Not yet, because many
other combinations of atoms are possible.  For an error limit
of 3.4 mmu (12 ppm), the error limit for a single exact mass determination
made with 5000 resolving power 3,
104 compositions are possible based on combinations of C, H, N,
O, F, P, and S atoms.  The measured exact mass of the molecular
ion reduced the number of possible compositions, but fell far short
of providing a unique composition.  Additional mass spectrometric
measurements were required before this ion was unequivocally determined
to be C18H19NS+.
|Higher-mass isotopes of elements|
of atoms of the higher-mass isotopes (13C, 2H,
15N, and 34S) provide mass peak profiles +1
and +2 amu heavier than the monoisotopic profile for which all atoms
are the most abundant isotopes (12C, 1H, 14N,
and 32S).  Measured exact masses of the +1 and +2
profiles and their abundances relative to the monoisotopic profile
provide four additional values to distinguish among compositions.
The number of possible compositions increases exponentially as an
ion's mass increases4, doubles when
the error limit doubles <sup>(4)</sup>, and increases
each time a new element is added to those deemed possible.
Comparison of calculated and measured values of three exact masses
and two relative abundances, rather than of a single exact mass,
increases 4-fold the upper mass limit of ions for which unique ion
compositions can be determined.3
As illustrated in Figure 1, knowing the composition
of the molecular ion limits a compound's identity to a number of
possible isomers.  Determining the compositions of fragment
ions then reduces this number.
To measure all five
quantities for analytes that elute from a GC with a residence time
of less than 10 s in the ion source of the mass spectrometer, an
instrument must provide at least 1 scan each second, excellent mass
accuracy (since the number of possible compositions doubles when
the error limit is doubled), a linear dynamic range of at least
103 (since the relative abundance of the +2 profile can
be 1% or less), and high resolving power to discriminate against
mass interferences, if complex mixtures are analyzed.  When
a recently developed scanning procedure is used, double focusing
mass spectrometers provide the best combination of these instrumental
Mass Peak Profiling from Selected Ion Recording Data (MPPSIRD)
In Figure 3 are shown full and partial mass peak profiles plotted by connecting discrete points.  Each point is the area of a chromatographic peak observed in the ion chromatogram acquired at the m/z ratio of the point.
Figure 3.  (a) Two full mass peak profiles, each delineated by three points, for ions from an analyte and an ever-present compound.  The resolving power was 3000 (10% valley definition) and the mass increment was 100 ppm.  (b) The full mass peak profile for the analyte ion delineated by 10 points out of the 21 monitored.  The resolving power was 10,000 and the mass increment was 10 ppm.  (c) Partial profiles plotted from 7 points for the m/z 182 ion and its +1 and +2 profiles acquired using a resolving power of 10,000 and a 10 ppm mass increment.  For all three experiments the partial profiles for the lock mass and calibration mass were each monitored by 5 m/z ratios.
For example, the maximum point in Figure 3a is the area under the chromatographic peak to the left of the point.  Another point corresponding to the smaller maximum is the area under the simulated chromatographic peak above it.  The compound that produced this ion is ever-present.  The baselines in the ion chromatogram were induced by reversing the polarity of a voltage in the ion source for 5 s both before and after the chromatographic peak of the analyte responsible for the larger mass peak profile eluted.  Using a Finnigan MAT 900S double focusing mass spectrometer, each of 31 m/z ratios was monitored for 20 msec during each 1-s selected ion recording cycle.  Partial profiles for the lock mass and calibration mass, which always bracket the analyte mass, were each plotted from five m/z ratios.  The remaining 21 m/z ratios were used to monitor a mass range, a single profile, or three partial profiles using 7 m/z ratios to delineate the top portions of each profile.
The three exact masses and two relative abundances were measured in the last experiment of a set of three experiments that employs successive approximation.5  The data displayed in Figure 3a determined how many mass peak profiles were present across a relatively wide mass range (2000 ppm or 364 mmu in this example) centered about the mass observed for an analyte ion in a low resolution mass spectrum.  The data were acquired with a resolving power of 3000 and exact masses were monitored every 100 ppm (18.2 mmu).  Two mass peak profiles, each delineated by 3 points, revealed the presence of an analyte ion (chromatographic peak) and an ever-present calibrant (steady signal), perfluorokerosene.  The estimated exact mass of the analyte profile, calculated as the weighted average of its three points, was used as the center mass in the next experiment.
For Figure 3b, exact masses were monitored every 10 ppm (1.82 mmu) to provide a well-defined profile delineated by 10 points.  The resolving power was 10,400 as measured from the 5% of maximum levels of the full profile.  The exact mass, accurate to within 6 ppm, was then entered into a Profile Generation Model (PGM)3, which provided a list of compositions consistent with the measured exact mass within its error limits.  The user then chose a hypothetical composition from the list.
The center masses of the hypothetical ion and its +1 and +2 partial profiles were then used as the center masses of the three partial profiles monitored in Figure 3c, again with 10 ppm mass increments and a resolving power of 10,000.  The exact masses were calculated as the weighted average of the points from each profile and the relative abundances as the ratio of the summed points for the +1 and +2 partial profiles divided by the summed points for the m/z 182 profile multiplied by 100%.  The ion investigated was determined unequivocally to be C3H7AsS2+.  The measured values in Figure 3 were consistent with the calculated values for the three partial profiles: 181.92051, 182.92272, 183.91636, 4.71%, and 8.89%.
In Figure 4 are shown calculated (a) and measured (b) mass peak profiles for a C23H28O2Br+ ion. The theoretical exact masses and relative abundances of the composite +1 and +2 profiles are listed under the profiles in Figure 4a.
Figure 4.  (a) Calculated profiles for the C23H28O2Br+ ion, the three +1 ions and three most abundant +2 ions, and the composite +1 and +2 profiles.  (b) Partial m/z 415, +1, and +2 profiles plotted from chromatographic peak areas under ion chromatograms for 7 m/z ratios across each profile.  (c) Ion chromatograms for the m/z ratios at the maxima of the partial profiles.
In Figure 4b, these values were obtained from the top portions of the mass peak profiles, which were plotted from areas under selected ion recording (SIR) chromatograms acquired as the two isomers evident in Figure 4c were eluted into the mass spectrometer.  The summed chromatographic peak areas in each panel of Figure 4c provided the maxima of the partial profiles.
A Profile Generation
Model (PGM) automatically determines the correct ion composition
by rejecting all compositions with calculated values of the three
exact masses and two relative abundances that are inconsistent with
the measured values.  Use of MPPSIRD and the PGM in concert
is Ion Composition Elucidation (ICE).
Mass spectral interpretation
of the compositions of the molecular ion (if observed) and the fragment
ions in a mass spectrum using ICE greatly reduces the number of
possible compounds that could produce the spectrum.  Even then,
an expert in mass spectral interpretation is required to hypothesize
possible isomers based on common fragmentations from ions formed
from numerous classes of compounds and to search the chemical and
commercial literature for candidates.
Example 1 - A well pollutant
Table 1 is a partial output from the PGM based on the experimental values from Figure 4b.
The seven compositions listed are possible based on the measured exact mass of the monoisotopic ion within the error limits of its measurement (3 ppm for one determination at 20,000 resolving power).  The mass defects are the non-integer part of the exact masses of the monoisotopic ion and the +1 and +2 profiles.  Only the last composition provides calculated exact masses and relative abundances consistent with the five measured values.
In Figure 5, the m/z 415 ion was observed in the (a) raw and (b) background subtracted mass spectra.
Figure 5.  (a) raw mass spectrum of a compound found in a well and (b) its background subtracted mass spectrum.
Because 79Br and 81Br have nearly equal isotopic abundances, the two mass peaks at m/z 415 and 417 visible above the chemical noise suggested a mono-brominated compound might be present in an extract of chlorine-disinfected, well-water.  M/z 430 and 432 ions became apparent above the chemical noise in the background subtracted mass spectrum.  ICE was also used to determine that the m/z 430 ion was C24H31O2Br+.  No compounds with this composition were found in the NIST or Wiley mass spectral libraries.
With a single composition
to consider for the apparent molecular ion, chemical reasoning and
searches of the chemical and commercial literature can lead to compound
identification.  Chlorination of the well water, which contained
bromide ions, could brominate organic compounds.  The structure
of Quinbolone, an anabolic steroid, is shown in Figure
5b and has three possible allylic bromination sites, more than
enough to account for the two isomers observed in the ion chromatograms
in Figure 4c. Substitution of a Br atom for
an H atom would provide the observed composition.  A feed lot
was located near the well and anabolic steroids are often used to
stimulate growth.  Purchase of Quinbolone, its chlorination
in the presence of bromide ions, and examination of the mass spectra
and retention times of the products would be logical next steps
in the identification process for this compound.
Example 2 - Pollutants in Toms River, NJ
When chemicals are found in a drinking water supply servicing a locality suffering an elevated occurrence of illness, it is important to identify those chemicals in order to facilitate toxicological studies and to ascertain the source of the chemicals.  Such was the case for water from a municipal well near Toms River, NJ, where an increased incidence of childhood cancer was observed.  Several isomers in a well-water extract provided similar mass spectra, which were not found in the NIST or Wiley mass spectral libraries.  One example is shown in Figure 6.
Figure 6.  A background subtracted mass spectrum for the most abundant of five isomers found in a municipal well water extract.  Compositions were determined for 11 ions.
The compositions of the molecular ion and 10 fragment ions listed in Table 2 were determined using ICE.
The fragment ion chromatograms tracked those of the molecular ion and provided the same retention times.  The exact mass alone of each fragment ion provided the correct composition, because the strict elemental limits provided by the molecular ion (0-14 C, 0-14 H, and 0-2 N atoms) limited the number of possible compositions for each fragment ion to no more than three. Agreement to within 0.5 mmu between the calculated and measured exact masses identified the correct compositions shown in red in Table 2.
Consideration of these
fragment ions and the corresponding composite neutral losses by
an expert in mass spectral interpretation indicated the compound
contained a tetralin core with external C3H4N
and CN groups as depicted by the structures in Figure 6.  Where these external groups were attached to the
rings could not be determined from the table.  Even so, only
a small number of possible isomers remained.  A search of the
chemical literature soon located 1:2 styrene:acrylonitrile byproducts
of an industrial polymerization process as the candidate compounds.
A standard from an industrial process currently used was obtained.
Three isomers were observed in the standard with the same retention
times and mass spectra as three of the five isomers observed in
the well water.  These confirmations established the industrial
source of the pollutants.
Example 3 - Superfund site samples
byproducts of production dumped into Superfund sites can yield very
complex mixtures of compounds.  For one such site, ion compositions
were determined for 51 compounds, most of which contained NS, N2S,
or N2S2, including the analyte from Figure
Most of the summed chromatographic peak areas was due to alkyl benzothiazoles
as indicated by the presence of the characteristic C8H7NS+
ion in many of the mass spectra.  These compounds are used in
the dye and rubber industries.  Characterization of these compounds
provided confirmation that the dumped material was from a nearby
examples demonstrated that ICE is a powerful analytical tool when
applied to analytes introduced into the mass spectrometer in the
gas-phase.  But most compounds are too polar, ionic, thermolabile,
or too high in mass to traverse a GC column or to volatilize from
a heated probe.  Such compounds require liquid sample introduction
using electrospray ionization (ESI) or atmospheric pressure chemical
ionization (APCI).  These low-energy ionization techniques yield
adduct ions (e.g., the molecular ion plus Na or K) and few fragment
ions.  Additional factors that make application of ICE to such
compounds difficult are poorer chromatographic resolution, much
lower sensitivity than electron impact ionization, competition between
calibrants and analytes for ionization which depletes the analyte
signal, and the lack of commercial ESI or APCI mass spectral libraries.
Despite these limitations, we will soon investigate the use ICE
to characterize and identify compounds separated by high performance
ICE + Containment
one laboratory in Germany6 has used a modified
form of MPPSIRD to identify a naturally occurring, bioaccumulative,
organohalogen compound, only our lab now performs ICE as described
herein.  This lab is not equipped to work with unanticipated
compounds that could be extremely toxic.  ECB is ready and willing
to transfer ICE technology to containment labs within secure facilities.
If necessary, we will adapt the ICE code for the data systems of
other models of double focusing mass spectrometers.  Labs expected
to identify potentially toxic compounds intentionally added to water,
air, or food supplies would then have a powerful new analytical
tool for doing so.  In addition, we hope to transfer this technology
to labs with double focusing mass spectrometers that desire to investigate
the "mystery" compounds observed in total ion chromatograms for
more common sample extracts.
occur where a single chemical or a complex mixture of chemicals
not found in mass spectral libraries is added intentionally or unintentionally
to water, air, or food.  These compounds must be identified
before their toxicological properties can be investigated.
Determination of the elemental compositions of the ions in mass
spectra sufficiently limits the number of possible isomers for each
compound to make searches of commercial and chemical literature
practical.  If a compound were not found in the literature,
isolation and concentration of the compound would be necessary to
perform other analyses, such as NMR, FTIR, or x-ray crystallography
to identify each compound.
|The authors are grateful to Floyd A.Genicola for providing the As-containing compound for which the data used to construct Figure 3 was obtained.  This data helped confirm his identification of the compounds as 2-methyl-1,3,2-dithiarsolane.|