A planar DMS-MS system for biomarker detection can provide performance advantages in selectivity and in sensitivity, as well as ease of use. These advantages include the following.
- Separation of isobars. Because DMS separation depends on ion properties that are largely independent of mass to charge ratio, DMS is able to separate isobaric compounds in many cases.
- Chemical noise reduction. Electrospray ionization sources generate heterogeneous ion populations comprised of the ion of interest as well as a multitude of different ion species and clusters, as well as fragment ions generated in the source or in the atmosphere-to-vacuum interface. These ions are be dispersed across the m/z range, contributing to chemical background. This chemical noise is usually greatly suppressed when DMS is tuned to select the target ion species for introduction into the MS.
- Separation of charge states. Increasing charge both increases ion mobility, and, because collision energy is increased, changes the shape of the dependence of the curve describing mobility on field. As a result, ion species occurring at the same m/z value but with different charge have much different DMS properties.
- Greater orthogonality than time-of-flight ion mobility spectrometry (IMS). DMS parameters are less correlated with mass than the IMS drift time [24, 25] because DMS depends on additional non-geometric molecular properties such as charge distribution and polarizability, and because DMS accesses a range of effective temperatures during the DMS field cycle.
- Short residence time. Residence times in planar DMS configurations are typically in the range of 1 to 4 ms. This short residence time reduces diffusion losses and minimizes unwanted reactions.
- Continuous operation. DMS filtration, unlike time-of-flight IMS configurations, operates continuously without the ion losses caused by a shutter.
- Filtration of intact molecular ion. MS/MS techniques are used to provide selectivity for many applications, but each fragmentation step in this method results in ion losses and the uncertainty in relative fragmentation efficiency requires careful calibration. DMS-MS operates on the ion prior to fragmentation while providing selectivity that can be equivalent to MS/MS. In addition, MS/MS techniques are time-consuming to apply in a mixture containing unknowns, because tentative identifications must be made and then verified in separate experiments. When necessary for confirmation of results, the DMS-CID-MS technique described in the section on isobar separation can provide verification of DMS ion selection based on the fragmentation pattern of a pre-selected ion population.
- Ease of Use
- Transparent Mode. DMS in the planar configuration used here can be operated in transparent mode, which allows all ions to pass through the filter with minimum attenuation. This mode is used to optimize ion source operation. Differential mobility devices with cylindrical geometry do not have this capability.
- Constant Intensity. Planar DMS resolution is nearly independent of operating parameters separation voltage (SV) and compensation voltage (CV). The cylindrical or FAIMS configuration focuses or defocuses ions in a way that varies with separation voltage (SV) making intensities and resolution dependent on SV.
- Polarity independent. Planar DMS operates to filter ions of both polarities, so no electrical changes are necessary when MS polarity is changed.
The following sections describe the DMS-MS experimental tests that have been performed to verify these expectations in the context of biomarker detection. We analyze DMS-MS performance for the separation of isobaric compounds, describing how DMS-MS with fragmentation at the MS inlet following DMS ion selection can operate in a mode like MS/MS of a triple-quadrupole instrument. We then examine separation of charge states, separation of biomarkers that are structural isomers, and give an example of DMS-MS applied to one of the urine samples used in biomarker discovery.
DMS-MS Separation of Isobaric Compounds
The need to make quantitative measurements of multiple species occurring at nearly the same m/z or in the presence of isobaric interferents arising from chemical noise is one of the factors that drive the usage of very high resolution, and high cost, mass spectrometers. As a result, many diagnostic or analytical measurements cannot be made in field settings. If compounds close in m/z can be separated by DMS pre-filtration, high mass resolution is less important. In addition, quantification with triple-quadrupole MS/MS methods requires calibration based on production scans of each target compound, while MS/MS sensitivity is reduced by loss of signal in the fragmentation step that is highly compound dependent.
As a reference point for isobaric separations, we have analyzed two different mixtures, one of five isobaric compounds of mass 315 Da, and a second group of six nearly isobaric compounds of mass 308–309 Da. The first group of five compounds all appear in positive mode ESI-MS at the same unit m/z. and provide the chemical identity and structure of the five compounds, each of which has been measured in DMS-MS under low fragmentation conditions at m/z 316 (as the protonated molecules, [MH]+1). also gives the exact mass for the compounds and protonated molecules, as well as the m/z difference between ions, and the resolution required to resolve them.
Table 1 Five components of DMS-MS isobaric mixture. The five protonated molecules all were observed in mass spectrometry at 316 Da as shown in . The DMS ion filter resolves the mixture into 4 distinct peaks, with fendiline and chlorprothixene remaining (more ...)
2D structures of the five compounds of .
We recorded DMS-MS spectra of each of the compounds separately at several separation voltages in order to identify compensation voltage peak positions for each compound. In order to visualize the challenge presented by the mixture, presents the mixture mass spectrum under conditions that minimize fragmentation with no DMS separation (DMS-transparent mode, SV=0, CV=0 which passes all ions), obtained on the single-quadrupole Waters Micromass ZQ. The mass spectrum consists of a single peak group with isotopic satellite peaks. If the observations are examined in detail (zoomed inset to the figure), we see that the mass spectrum could be incorrectly interpreted as arising from a single ion, not five chemically distinct ions, with a isotope distribution typical of organic compounds in this mass range. shows the DMS compensation voltage (CV) tuning characteristics at SV 1400 V of MS ion count for the base peak, m/z 316 [MH]+, and for the next two higher m/z values, indicating the distribution of isotopic masses. The carrier of each peak is identified based on the DMS CV characteristics of standard samples. Four separate peaks are observed for the five compounds, with fendiline and chlorprothixene remaining unresolved. Of the five compounds only chlorprothixene contains chlorine. The presence of chlorprothixene (C18H18ClNS) overlapped with fendiline in the CV = +2V peak is evident from the high relative intensity of m/z 318 [MH + 2]+ due to 37Cl. Absence of chlorine in the other 3 peaks is demonstrated by low m/z 318 intensity compared to the m/z 316 base peak and m/z 317 intensities for the other ions.
Figure 2 (A) A single group of peaks appears at m/z 316 in nanoESI-MS of a mixture of the five compounds of on Waters ZQ MS. DMS was operated in transparent mode (SV=0, CV=0) and the MS under non-fragmenting (low cone voltage) inlet conditions. Observed (more ...)
To activate the molecular ions, approximating MS/MS conditions but without initial m/z selectivity, we increased the inlet cone voltage to induce fragmentation at the atmospheric-pressure to vacuum interface of the mass spectrometer after the DMS filter. We refer to this method as DMS-CID-MS (DMS collision-induced dissociation mass spectrometry). Both DMS-CID-MS and DMS-MS (no fragmentation) are two-dimensional separation techniques because a full or selected mass spectrum is recorded at each setting of DMS SV and CV tuning parameters. DMS-CID-MS fragment identifications are obtained by first DMS-selecting specific ion species and then passing them into the mass spectrometer with additional energy from the inlet cone voltage difference. This provides a level of specificity similar to a triple-quadrupole instrument. Fragments obtained in this way originate from the DMS-selected ion only, by may require some extra scrutiny because of possible ion reactions in the MS inlet area and DMS selectivity can be limited by overlapping features in DMS CV tuning.
presents the DMS-CID-MS results for each of the five compounds in the five-component mixture obtained at SV 1200 V by using CV values of −6.15 V, −1.65 V, −0.08 V, +2.22 V, and 3.8 V for DMS selection of oxefendazole, flusilazole, fendiline, chlorprothixene, and pamaquine, respectively. To minimize effects from the overlap of fendiline and chlorprothixene DMS peak shapes, CV values at the edges of the feature were used. The “All Ions” trace was computed by summing mass spectra at all CV values recorded across the CV scan. Dominant fragments selected from these results allow us to use fragment MS spectra to evaluate in another way the selectivity achieved by DMS. shows peak positions observed for separation voltages of 1200 V (106 Td at 1 atm, 50°C), and 1400 V (123 Td). DMS compensation voltages for these compounds with SV 1200 V span approximately 10 V, with peak widths of 1.2 to 1.5 V. In , using the characteristic fragment ions in the table, each of the five ions is separately detected. Fendiline and chlorprothixene are overlapped, as expected from the previous figure. All ions are resolved, but signal intensities detected through the CID fragment spectrum for these five compounds are typically 30% of the signals observed using DMS-MS for the same species.
Figure 3 (A) Setting a high inlet cone voltage allows characteristic fragment ions for each of the five molecular ions to be identified, at a cost of reduced sensitivity due to the multiple fragmentation paths for each of the ions. SV 1200 V was used with CV values (more ...)
Table 3 DMS-MS characteristics of the isobaric mixture of five compounds. Results are given for DMS separation voltages (SV) of 1200 V (106 Td) and 1400 V (123 Td), positive MS ion mode. MS/MS information was determined by inducing CID in the inlet cone. The (more ...)
We have also tested DMS-MS performance on a set of six compounds of approximately the same mass. These six compounds, benoxinate, bestatin, nifenazone, phenylbutazone, quinoxyfen, and warfarin have molecular weights of 308 Da, except for quinoxyfen at 307 Da. All formulas, structure, and computed isotope ratios for the ion species related to this mixture are given in the supplemental material
for this paper. shows the DMS-MS analysis at characteristic m/z values for the base peaks and isotopes, similar to , of nanoESI of the 6-component mixture. Conditions minimizing fragmentation at the DMS-MS transition were used so that the MS contained only [MH]+ and [MNa]+ base peaks, maximizing sensitivity. The [MH]+ base peak of quinoxyfen is m/z 308 and [MH]+ is m/z 309 for bestatin, benoxinate, nifenazone, and phenylbutazone. The mass spectrum near m/z 309 without DMS separation is shown in the inset to the figure. [Warfarin·Na]+ and [phenylbutazone·Na]+ were also observed and appear at m/z 331, as labeled. All but one pair of ions (benoxinate, nifenazone) are resolved by the combined DMS-MS performance (DMS alone at m/z 331 for warfarin·Na+ and phenylbutazone·Na+, and at m/z 308-9 for protonated quinoxyfen, bestatin and phenylbutazone). Benoxinate and nifenazone remain unresolved at 1400V. They are partially separated by 1 V CV at SV 1200V (peak FWHM ~1.6 V) but separation of the other ions is reduced at the lower separation voltage.
Figure 4 NanoESI-DMS-MS 2D spectrum of the 6 component mixture is shown for SV 1400 V. At each DMS compensation voltage, a full mass spectrum was obtained, but MS ion counts are shown only for characteristic masses. All but one pair of ions (benoxinate, nifenazone) (more ...)
For the six compound mixture, characteristic fragment ions could not easily be identified by DMS-CID-MS to distinguish each compound from all others. Because DMS separation is based on the ion before fragmentation, it is not ambiguous in this way, but peak positions must be identified by the use of standards, as was done to determine the labels in . In this mixture, quinoxyfen contains two chlorine atoms, resulting in a very different pattern of intensities of isotopic peaks from other components of the mixture. This makes it possible to test the quality of the quinoxyfen separation by measuring isotope ratios. The observed and predicted isotopic peak intensities for [quinoxyfen·H]+ and for [phenylbutazone·H]+ are shown in , and the agreement with predictions is good.
Figure 5 (A) Quinoxyfen (C15H8Cl2FNO) is here selected as [MH]+ by CV −6 V for SV 1400 V. The observed DMS-selected mass spectrum is shown and compared with predicted isotopic ratios. (B) Phenylbutazone (C19H20N2O2) is selected as [MH]+ by CV +7 V for (more ...)
DMS is especially useful in combination with mass spectrometry because its separation mechanism is complementary to m/z. The presented examples show how useful DMS can be in resolving mixtures that are difficult for mass spectrometry, in this case without the use of clustering modifiers in the DMS transport gas. The use of polar modifiers in the drift gas results in 8 to 10 fold improvement in DMS resolution, as discussed in more detail below for the citrate-isocitrate separation, and in Schneider et. al. [26
DMS Separation of Charge States
In complex samples ionized in atmospheric pressure sources, ions may be created with more than one charge state. Heavier ions that are multiply charged by the ion source appear in the same m/z range as lighter species. For small molecule biomarkers, this aliasing of heavier species into the m/z range of the biomarker causes errors in intensities, and adds complexity to the mass spectrum. This resulting degradation of the mass spectrum is similar in effect to the interference from chemical noise that is generated when heavy ions fragment into lighter ions in the biomarker range. Both effects can be reduced by a DMS prefilter. Multiply-charged ions typically have compensation voltages that are more positive than singly-charged ions with the same m/z. Two effects are involved in this result: (1) multiply-charged ions are field-accelerated more strongly, so the hard-collision limit is reached more rapidly, and (2) larger ions often have internal van der Waals interactions that reduce the strength of interactions with the drift gas, making ion-neutral interactions more hard-sphere-like [22
]. In addition, ions within chemical series also tend to move to positive CV as mass increases, but for similar m/z values the separation between charge states is large.
With DMS, we frequently observe a reduction in chemical noise, in which low intensity mass peaks are suppressed and correct isotope ratios are restored for a particular target species. For this particular example, we have observed DMS separation of two charge states that occur at similar m/z values. The sample was a mixture of the tetrapeptide methionine-arginine-phenylalanine-alanine (MRFA) and polyethylene glycols (PEG) from 400 to 600 Da, in 50/50 methanol/water. shows DMS resolution of [MRFA·H3
at m/z 524.2 from a doubly charged species at m/z 522.7 and from a doubly-charged ion of lower m/z, on a JEOL Accutof TOF. The complete mass spectrum is not shown, but is complicated with many near coincidences and contains a variety of MRFA/PEG clusters and fragments. The two species that are nearly coincident in mass-to-charge ratio (z=+1, m/z =524.2 [MRFA·H3
, and z=+2, m/z 522.7 (PEG related)) are separated by +6.2 V in DMS compensation voltage. It is of less practical importance for DMS-MS applications because the species are separated by low-resolution mass spectrometry alone, but also shows DMS-MS characteristics for the doubly-charged species [MRFA·H3
, which is separated by an additional 8.5 V from the singly-charged species. For the nearly-coincident peaks near m/z 524, DMS is able to select either species, completely suppressing the other. In , DMS filtered mass spectra in the range from m/z 520 to 530 for the MRFA/PEG mixture are shown at the compensation voltages that select each peak (1200 V SV): +5.6 V in the upper panel, selecting [MRFA·H3
, and +11.8 V in the lower panel selecting the doubly-charged species. Both doubly-charged and singly-charged ions, within chemical series tend to move to positive CV as mass increases, but for similar m/z values the separation between charge states is large, with the higher charge state to more positive CV. These specific CV values depend on the concentration of neutral solvent modifier molecules in the transport gas stream, but are stable under standardized conditions, as we have discussed in other publications [22
Figure 6 DMS-MS spectra of MRFA/PEG mixture for selected m/z values showing DMS separation by charge state. The tetrapeptide MRFA was mixed with polyethylene glycols (200 to 800 Da) and recorded on a JEOL Accutof TOF in positive ion mode. Ions of nearly the same (more ...)
Figure 7 DMS filtered mass spectrum from m/z 520 to 530 for the MRFA / PEG mixture. The compensation voltage for the doubly-charged species is +6.2 V higher than that of the singly-charged [MRFA·H3O]+1 ion of similar m/z. Within a charge state, higher (more ...)
DMS-MS Separation of Candidate Biomarkers Citrate / Isocitrate
Citrate and / or isocitrate have been identified as potential biomarkers both by the LC/MS studies [1
] and the GC/MS study[3
] using classification techniques such as orthogonal partial least squares and random forests, and are associated with some of the largest metabolomic signals. Citric acid and isocitric acid are important in the citric acid cycle, in glycolysis, and in cellular respiration.. Citric acid is achiral, but isocitric acid can have both D and L forms. Our sample of citric acid was the citric acid monohydrate while isocitrate was obtained as the tri-sodium salt of DL-isocitric acid.
Because these two compounds are structural isomers, with identical chemical compositions, both appear in the anion mass spectrum at m/z 191, as [M-H]−. Nonetheless, at high separation voltages, we find that the citrate and isocitrate ions appear to be weakly separated by DMS under the conditions used here. This separation cannot be achieved by mass spectrometry, or by low field ion-mobility techniques. This result is shown in . The citrate and isocitrate samples were prepared identically and run sequentially. Nonetheless, this weak separation would be expected to have some dependence on spray conditions and this small separation would not be reliable in production analytical applications. A modification to the technique which would increase the separation is necessary for applications, and we find that use of a transport gas modifier is effective.
Figure 8 (A) DMS-MS of citrate and isocitrate anions. These two isobaric ions are weakly separated by DMS with no modifier in the transport gas at 1400 V SV under these conditions. (B) DMS-MS of citrate and isocitrate anions with drift gas modifier. Citrate and (more ...)
DMS resolution can be improved by additions known as modifiers to the composition of the transport gas used to carry ions through the DMS into the mass spectrometer. This is a technique that is applicable to essentially all systems, but is especially useful when chemical and structural differences are small [22
]. The choice of modifier depends on ion polarity and, to a lesser degree, on the chemical structure of the target ions. These modifiers participate in field-dependent clustering and declustering in the DMS separation field, increasing the difference between high and low field mobilities, thereby increasing the compensation voltage. For negative ions, chloride-based modifiers are frequently effective. By adding a small partially-chlorinated aliphatic hydrocarbon (~0.1% 1,2,3-trichloropropane in this case) to the desolvation gas, we have found that citrate can be detected separately from isocitrate, as shown in . Chlorine-containing modifiers have been used for some time in a number of contexts such as explosives detection [27
], to enhance DMS separation of anions and are believed to be mediated by chloride anion.
The use of drift gas modifiers to enhance separation for cations and for anions in planar DMS instrumentation is an active area of investigation and has recently been shown to lead to an order of magnitude increase in selectivity when evaluated over a wide spectrum of target ions in for complex mixtures[26
]. The enhanced citrate-isocitrate separation shown here is another example of the general utility of the technique.
DMS-MS Separation of Dicarboxylic Acids
The most recent radiation biomarker paper[3
] identifies several aliphatic dicarboxylic acids as down-regulated biomarkers (adipic, pimeric, suberic, azelaic acids (HO2
H, for j=4,5,6,7). The dicarboxylic acids appear as anions with negative compensation voltages, near −10 V CV for 1000 V SV (88 Td). This characteristic distinguishes them from many of the other biomarkers, and from typical interferences. In addition, these ions are detected with high sensitivity under low-CID conditions. In analyzing a standard, we found that the azelaic acid sample contained traces of other dicarboxylic acids, with j=5 to 10, and provided an interesting demonstration of DMS separation of a related series of compounds. shows DMS separation of dicarboxylic acids appearing in azelaic acid sample obtained for 1000 V SV. The acids are HO2
- for j=5 to 10, (j=5 (pimelic acid), j=6 (suberic acid), j=7 (azelaic acid), j=8 (sebacic acid), j=9 (undecanedioic acid) and j=10 (dodecanedioic acid)) appearing as the deprotonated molecules. Each of the DMS peaks is separated by 1 V or more in compensation voltage. The curve peaking at −10 V under the conditions of is azelaic acid. Shorter chain acids appear at more negative compensation voltages, longer chains at more positive values. In , the peaks are labeled by relative intensities of the separate ion signals in the DMS-transparent and DMS-filtered mass spectra, and show a dynamic range of approximately 500.
Figure 9 DMS separation of dicarboxylic acids appearing in azelaic acid sample. In the legend, from top to bottom, and left to right in the figure, are HO2C(CH2)jCO2- for j=5 to 10 (j=5 (pimelic acid), j=6 (suberic acid), j=7 (azelaic acid), j=8 (sebacic acid), (more ...)
DMS-MS Testing on Biofluid Samples
Radiation-exposure biomarker discovery by high-resolution ultra-performance liquid chromatography time-of-flight mass spectrometry (UPLC –TOFMS) described in the first two of the recent group of papers is based on a comparison of urine samples from exposed and sham-treated mice, with doses ranging from 0 to 8 Gy [28
Both of these studies identified N
-hexanoylglycine as a relevant biomarker, along with a number of other candidate or validated biomarkers. The most recent study[3
], based on a rat model, uses chemical derivatization followed by GC-MS, did not identify this compound, which may be unstable under these conditions, but does have a number of other biomarker identifications in common with the LC-MS work. As an example of the performance of DMS-MS on biological samples, we examine the DMS-MS negative ion spectrum using the single-quadrupole Waters ZQ of a urine sample from the treated group in the mass range that includes both N
-hexanoylglycine, and suberic acid. These molecules are identified as biomarkers by LC-MS and GC-MS studies, respectively.
Suberic acid DMS-MS properties in negative ion mode are shown in along with those of several other dicarboxylic acids, for nanospray from methanol/water solution. N
-hexanoylglycine also appears most strongly in negative ion mode. In the medical and scientific literature, N
-hexanoylglycine is associated with medium-chain acyl-coA dehydrogenase deficiency (MCAD, Reyes syndrome), with dicarboxylic aciduria[29
], and with a few other conditions[30
]. This compound has also been the subject of previous GC-MS studies in a medical context.[31
The DMS-MS results for the urine sample from the treated group are shown in under three different conditions: no DMS selectivity (inset), DMS selecting N-hexanoylglycine, and DMS selecting suberic acid. [M-H]− signals for N-hexanoylglycine and suberic acid appear at m/z 172 and 173, respectively.
Figure 10 DMS-MS mass spectrum of mouse urine from treated group at SV 1000 V, resolving 2 biomarkers from each other, and suppressing a background of chemical noise. The peak at m/z 172 (CV −4.6 V) was identified as N-hexanoylglycine by comparing fragmentation (more ...)
Without DMS selectivity, each peak intensity is due to some intensity from the desired peaks, but also includes additional intensity from main and isotope peaks of compounds nearby or identical in mass and fragments and clusters originating from other compounds in the complex matrix, which are transformed into the biomarker mass range either in the ionization process or in the mass spectrometer interface. Quantitative measurement is difficult under these conditions, especially for low concentrations.
With DMS, these interferents are suppressed and the spectra at m/z 172 and 173 are observed separately, at distinct DMS compensation voltages. The identity of hexanoylglycine in the DMS-filtered mass-spectrum was verified by observing identical fragmentation patterns of the DMS-selected mass peak in the biosample in comparison with that of a pure standard, and by comparison of the DMS compensation voltage from a standard. Suberic acid was not verified by MS/MS in this series of tests, but the DMS compensation voltage is in agreement with the value from a standard. In addition, the intensity of the [M+1] isotope peak for the separated compounds is in agreement with the predicted [M+1]/[M] ratio within 1%. Because it is possible to verify that selected intensity observed from the urine sample is due to the target biomarker, the measurement accuracy is improved, especially for biomarkers that appear at low concentrations where they may be completely lost in chemical noise. To perform the same analysis without DMS necessitates very high resolution, and/or MS/MS, and some other preseparation techniques like LC or derivitization/GC.
Sensitivity and selectivity of DMS-MS is clearly adequate for these two biomarkers. For other biomarkers, the sensitivity depends on concentrations and is reduced by ion suppression effects that occur in a complex matrix [32
]. Concentrations of the biomarker components in urine samples have been given as part of the biomarker discovery reports for only a few of the identified species. Tyburski et. al.[1
] list the concentrations relative to creatinine (typ. 3 mM) of N
-hexanoylglycine as 400 μM/mM, of taurine as 8 mM/mM, and of β-thymidine as 6–10 μM/mM. Even considering suppression effects, the concentrations of N
-hexanoylglycine and taurine are readily detectable, and are quantifiable by controlled addition. β-thymidine has not yet been detected in our DMS-MS testing of biofluids. Thymidine is subject to significant suppression effects, occurs at a low concentration, and has not always been detected in animal model testing. Quantitative measurements of concentrations for other biomarkers would be useful, but is not currently available.