Coupling Capillary Electrophoresis and High-Field Asymmetric

High-field asymmetric waveform ion mobility spectrom- etry (FAIMS) is a new technology for atmospheric pres- sure, room temperature separation of gas-...
0 downloads 0 Views 147KB Size
Anal. Chem. 2004, 76, 4676-4683

Coupling Capillary Electrophoresis and High-Field Asymmetric Waveform Ion Mobility Spectrometry Mass Spectrometry for the Analysis of Complex Lipopolysaccharides Jianjun Li,*,† Randy W. Purves,‡ and James C. Richards†

Institute for Biological Sciences, National Research Council of Canada, 100 Sussex Drive, Ottawa, Ontario, Canada, K1A 0R6, and Ionalytics Corporation, 1200 Montreal Road, Ottawa, Ontario, Canada, K1A 0R6

High-field asymmetric waveform ion mobility spectrometry (FAIMS) is a new technology for atmospheric pressure, room temperature separation of gas-phase ions. The FAIMS system acts as an ion filter that can continuously transmit one type of ion, independent of mass-to-charge ratio (m/z). Capillary electrophoresis-electrospray mass spectrometry (CE-MS) has been extensively used for the analysis of complex bacterial lipopolysaccharides (LPS). The coupling of FAIMS to CE-MS provides a sensitive technique for the characterization of these complex glycolipids, permitting the separation of trace-level LPS oligosaccharide glycoforms for subsequent structural characterization using tandem mass spectrometry. This was demonstrated for LPS from nontypeable Haemophilus influenzae strain 375 following O-deacylation with anhydrous hydrazine. This strain of H. influenzae can express a triheptosyl-containing glycoform to which four hexose residues are linked forming the outer-core region of the molecule. This has been referred to as the Hex4 glycoform. Glycoforms have been identified which differ in the number of phosphoethanolamine substituents in the inner-core. With the use of CE-FAIMS, isomeric Hex4 glycoforms containing two PEtn groups were separated and characterized by MS/MS. FAIMS provided a significant reduction in mass spectral noise, leading to improved detection limits (∼70 amol of the major glycoform). The extracted mass spectrum showed that the apparent noise was virtually eliminated. In addition to the reduction of chemical background, the ion current was increased by as much as 7.5 times as a result of the atmospheric pressure ion-focusing effect provided by the FAIMS system. The linearity of response of the CE-FAIMS-MS system was also studied. The calibration curve is linear for ∼3 orders of magnitude, over a range of 40 pg/µL to 10 ng/µL. Capillary electrophoresis (CE) is a high-resolution technique for the separation of complex biological mixtures and has been * Corresponding author. E-mail: [email protected]. Phone: (613) 9980326. Fax: (613) 952-9092. † Institute for Biological Sciences. ‡ Ionalytics Corporation.

4676 Analytical Chemistry, Vol. 76, No. 16, August 15, 2004

widely applied to biological analysis.1-2 The application of capillary electrophoresis electrospray ionization mass spectrometry (CE-ESI-MS) techniques to the characterization of carbohydrates provides unparalleled resolution and serves as a structural tool in the identification of oligosaccharide populations and substituted phosphorylated functionalities present in lipopolysaccharide (LPS) molecules.3-7 LPS molecules are a group of biopolymers that are found on the outer membrane of gram-negative bacteria. Such structural tools are playing an increasing role in understanding the molecular interactions between host cells and foreign pathogens (cell adhesion, epitope recognition, etc.) as well as in the comprehension of many intricacies of cellular molecular biology. Recently we have employed CE-ESI-MS to investigate the role of LPS oligosaccharide expression by Haemophilus influenzae in virulence and disease pathogenesis.8 Nontypeable H. influenzae is a leading cause of otitis media in children. The carbohydrate regions of LPS molecules provide targets for recognition by host immune responses, and the expression of certain oligosaccharide epitopes contributes to the pathogenesis of H. influenzae infections.9 H. influenzae LPS is composed of a membrane-bound lipid A moiety to which the oligosaccharide portion is attached by a single 3-deoxy-D-mannooctulosonic acid (KDO) residue. Molecular structural studies of LPS from mutant and wild-type strains of H. influenzae have resulted in a structural model consisting of a conserved triheptosyl inner-core moiety in which each of the heptose residues can (1) Krylov, S. N.; Dovichi, N. J. Anal. Chem. 2000, 72, 111R-128R. (2) Hu, S.; Dovichi, N. J. Anal. Chem. 2002, 74, 2833-2850. (3) Cox, A. D.; Li, J.; Brisson, J.-R.; Moxon, E. R.; Richards, J. C. Carbohydr. Res. 2002, 337, 1435-1444. (4) Thibault, P.; Li, J.; Martin, A.; Richards, J. C.; Hood, D. W.; Moxon, E. R. In Mass Spectrometry in Biology and Medicine; Burlingame, A. L., Carr, S. A., Baldwin, M. A., Eds.; Humana Press: Totowa, NJ, 1999; pp 439-462. (5) Kelly, J.; Masoud, H.; Perry, M. B.; Richards, J. C.; Thibault, P. Anal. Biochem. 1996, 233, 15-30. (6) Li, J.; Thibault, P.; Martin, A.; Richards, J. C.; Wakarchuk, W. W.; van der Wilp, W. J. Chromatogr., A 1998, 817, 325-336. (7) Auriola, S.; Thibault, P.; Sadovskaya, I.; Altman, E. Electrophoresis 1998, 19, 2665-2676. (8) Bouchet, V.; Hood, D. W.; Li, J.; Brisson, J. R.; Randle, G. A.; Martin, A.; Li, Z.; Goldstein, R.; Schweda, E. K. H.; Pelton, S. I.; Richards, J. C.; Moxon, E. R. Proc. Natl. Acad. Sci. U.S.A. 2003, 100, 8898-8903. (9) Moxon, E. R.; Maskell, D. In Molecular Biology of Bacterial Infection: Current Status and Future Perspectives; Harmaeche, C. E., Penn, C. W., Smyth, C. J., Eds.; Cambridge University Press: Cambridge, UK, 1992; pp 75-96. 10.1021/ac049850d CCC: $27.50

© 2004 American Chemical Society Published on Web 07/10/2004

provide a point for elongation by hexose-containing oligosaccharide chains or for attachment of non-carbohydrate substituents.10-16 The analysis of LPS glycoform populations expressed by bacteria isolated from an in vivo source is very challenging due to the small sample amounts and the natural heterogeneity of glycan distribution, as well as the presence of other impurities. Additionally, background noise6-8 in CE-MS analysis is a problem when analyzing trace-level O-deacylated LPS oligosacchrides. Employing precursor ion scanning in the MS analysis reduces the background;8 however, the sensitivity is also reduced. In an effort to reduce chemical noise and enhance detection limits in carbohydrate analysis, the application of high-field asymmetric waveform ion mobility spectrometry (FAIMS) in combination with capillary electrophoresis-mass spectrometry was investigated. FAIMS operates at atmospheric pressure and room temperature and provides ion separation based on compounddependent changes in ion mobility that occur under the influence of high electric field. The response to changes in ion mobility distinguishes FAIMS from ion mobility spectrometry (IMS) where separation is based on absolute ion mobility. Detailed descriptions of the principles of FAIMS have been presented elsewhere.17-19 Briefly, in FAIMS, a high-voltage asymmetric waveform is applied to a set of electrodes. The asymmetric waveform is composed of a high-voltage component, called the dispersion voltage (DV), and an opposite polarity low-voltage component. Ions introduced between two concentric electrodes oscillate as a result of the applied waveform. The distance an ion travels during the two opposite polarity phases of the waveform are not equal due to differences in the ion’s mobility under the high- and low-field conditions. Over time, the ion will migrate toward one of the two electrodes. A low dc voltage, called the compensation voltage (CV), is applied to one of the electrodes to compensate for this migration. The magnitude of the CV required to transmit a particular ion through the electrodes is dependent on the ion’s ratio of high- to low-field mobility. Additionally, a unique ionfocusing mechanism resulting from the cylindrical design of the FAIMS electrodes ensures that this selectivity is accompanied by high analyte transmission efficiency. The coupling of FAIMS to electrospray mass spectrometry provides a unique tool to facilitate and accelerate the analysis requirements. The FAIMS system acts as an ion filter that can (10) Masoud, H.; Moxon, E. R.; Martin, A.; Krajcarski, D.; Richards, J. C. Biochem. 1997, 36, 2091-2103. (11) Plested, J. S.; Makepeace, K.; Jennings, M. P.; Gidney, M. A. J.; Lacelle, S.; Brisson, J.-R.; Cox, A. D.; Martin, A.; Bird, A. G.; Tang, C. M.; Mackinnon, F. M.; Richards, J. C.; Moxon, E. R. Infect. Immun. 1999, 67, 5417-5426. (12) Hood, D. W.; Cox, A. D.; Gilbert, M.; Makepeace, K.; Walsh, S.; Deadman, M. E.; Cody, A.; Martin, A.; Månsson, M.; Schweda, E. K. H.; Brisson, J. R.; Richards, J. C.; Moxon, E. R.; Wakarchuk, W. W. Mol. Microbiol. 2001, 39, 341-350. (13) Hood, D. W.; Makepeace, K.; Deadman, M. E.; Rest, R. F.; Thibault, P.; Martin, A.; Richards, J. C.; Moxon, E. R. Mol. Microbiol. 1999, 33, 679692. (14) Månsson, M.; Hood, D.; Li, J.; Richards, J. C.; Moxon, E. R.; Schweda, E. K. H. Eur. J. Biochem. 2002, 269, 808-818. (15) Månsson, M.; Bauer, S. H. J.; Hood, D. W.; Richards, J. C.; Moxon, E. R.; Schweda, E. K. H. Eur. J. Biochem. 2001, 268, 2148-2159. (16) Risberg, A.; Masoud, H.; Martin, A.; Richards, J. C.; Moxon, E. R.; Schweda, E. K. H. Eur. J. Biochem. 1999, 261, 171-180. (17) Buryakov, I. A.; Krylov, E. V.; Nazarov, E. G.; Rasulev, U. K. Int. J. Mass Spectrom. Ion Processes 1993, 128, 143-8. (18) Purves, R. W.; Guevremont, R. Anal. Chem. 1999, 71, 2346-2357. (19) Purves, R. W.; Guevremont, R.; Day, S.; Pipich, C. W.; Matyjaszczyk, M. S. Rev. Sci. Instrum. 1998, 69, 4094-4105.

continuously transmit one type of ion, independent of m/z. Previous studies have shown improved tryptic peptide detection and identification20-22 using FAIMS, as singly charged ions that make up a significant fraction of the background noise are transmitted though the FAIMS electrodes under different conditions than multiply charged peptides. In this paper, the application of CE-FAIMS-MS to the analysis of an O-deacylated LPS oligosaccharide mixture from H. influenzae is presented. The analytical potential of CE-FAIMS-MS to separate closely related glycoform families in O-deacylated LPS from H. influenzae strain 375 is demonstrated, and on-line tandem mass spectrometry is shown to provide a powerful technique for structural characterization of isomeric glycoforms. MATERIALS AND METHODS Chemicals and Materials. Fused-silica capillaries with 192 µm o.d. × 50 µm i.d. were obtained from Polymicro Technologies (Phoenix, AZ). Methanol and 2-propanol (IPA) were from EM Science (Gibbstown, NJ). Anhydrous hydrazine and ammonium acetate were obtained from Fisher Scientific (Fair Lawn, NJ) and formic acid from BDH (Toronto, Canada). The enzymes proteinase K, deoxyribonuclease I (DNase), and ribonuclease (RNase) were obtained from Sigma (St. Louis, MO). All aqueous solutions were filtered through a 0.45-µm filter (Millipore, Bedford, MA) before use. Bacterial Strains and Growth Conditions. Nontypeable H. influenzae (NTHi) strain 375 was selected from a collection of 107 OM isolates of NTHi obtained by tympanocentesis.13 The strain was cultured at 37 °C in brain-heart infusion (BHI) broth supplemented with 10 µg/mL of haemin, 2 µg/mL of NAD in 10 L batches.16 Preparation and Extraction of Lipopolysaccharides. LPS from H. influenzae strain 375 was extracted by the hot phenol/ water method as described previously16 and purified from the dialyzed aqueous phase by ultracentrifugation (45 000 rpm, 4 °C, 5 h) after treatment with DNase, RNase, and proteinase K.11 Preparation of O-Deacylated LPS. Purified LPS was dissolved in anhydrous hydrazine (∼2 mg/mL) and incubated at 37 °C for 50 min with constant stirring to release O-linked fatty acids from the lipid A region of the molecules. This also resulted in release of ester-linked acetates and glycine groups from the innercore region.23 The reaction mixtures were cooled (0 °C), excess hydrazine was destroyed by addition of cold acetone (5 vols), and the final product was obtained by centrifugation. The pellets were washed with acetone, centrifuged, and then lyophilized from water. CE-FAIMS-MS. A Prince CE system (Prince Technologies, The Netherlands) was coupled to an API 3000 mass spectrometer (Applied Biosystems/Sciex, Concord, Canada) via a microspray interface. The Ionalytics Selectra (FAIMS) electrodes (Ionalytics Corporation, Ottawa, Canada) were installed between the microspray source and the mass spectrometer and were operated (20) Barnett, D. A.; Ells, B.; Guevremont, R.; Purves, R. W. J. Am. Soc. Mass Spectrom. 2002, 13, 1282-1291. (21) Barnett, D. A.; Ding, L.; Ells, B.; Purves, R. W.; Guevremont, R. Rapid Commun. Mass Spectrom. 2002, 16, 676-680. (22) Venne, K.; Bonneil, E.; Barnett, D. A.; Thibault, P. In 51st ASMS Conference on Mass Spectrometry and Allied Topics, Montreal, Quebec, Canada, June 8-12, 2003; American Society for Mass Spectrometry. (23) Li, J.; Bauer, S. H. J.; Månsson, M.; Moxon, E. R.; Richards, J. C.; Schweda, E. K. H. Glycobiology 2001, 11, 1009-1015.

Analytical Chemistry, Vol. 76, No. 16, August 15, 2004

4677

Chart 1

at atmospheric pressure. The dispersion voltage was set to 4000 V. A curtain/carrier gas of 50/50 nitrogen/helium was introduced in the region between the curtain plate and outer electrode of the FAIMS electrodes, at a combined flow rate of 2.5 L/min. A fraction of this gas exited the curtain plate to desolvate the arriving electrospray ions, while the remainder carried the ions through the electrodes and into the mass spectrometer. A sheath solution (2-propanol/methanol, 2:1) for the electrospray interface was delivered at a flow rate of 0.75 µL/min to a low dead volume tee (250 µm i.d., Chromatographic Specialities, Brockville, Canada). Separations were obtained on a 90-cm length of bare fused-silica capillary using 30 mM of morpholine in deionized water, pH 9.0, containing 5% methanol. The outlet of the capillary was tapered to ca. 15 µm i.d. using a laser puller (Sutter Instruments, Novato, CA). Mass spectra were acquired with dwell times of 3.0 ms per step of 1 m/z unit in full-mass scan mode or 25 ms per channel for selected ion monitoring (SIM) experiments. In all of the CE-MS and CE-FAIMS-MS experiments, 20 nL of sample was typically injected by using 200 mbar for a duration of 0.1 min. RESULTS AND DISCUSSION ESI-MS of O-Deacylated LPS from H. influenzae Strain 375. The cell wall LPS of H. influenzae is comprised of a heterogeneous mixture of oligosaccharide glycoforms and a membrane-anchoring lipid A component. A structural model for the LPS oligosaccharide portion of nontypeable H. influenzae strain 375 has been proposed.8,13 It contains the conserved L-glycero-Dmanno-heptopyranosyl trisaccharide inner-core region (HepIHepIII) that is observed in every H. influenzae stain studied to date (ref 15 and references therein; see structure 1 in Chart 1). HepI is linked to the lipid A portion of the molecule via a KDO residue that carries phosphate (P) (R1 ) H) or pyrophosphoethanolamine (PPEtn) (R1 ) PEtn) substituents at the O-4 position.10 The lipid A is comprised of a bis-1,4′-phosphorylated β-1,6-linked D-glucosamine disaccharide which is substituted by 3-hydroxytetradecanoic and tetradecanoic acid groups as amide 4678 Analytical Chemistry, Vol. 76, No. 16, August 15, 2004

Figure 1. ESI-MS spectrum (m/z 400-1600) for 10 ng/µL of O-deacylated LPS from H. influenzae strain 375. O-deacylated LPS was dissolved in 30 mM morpholine buffer (pH 9.0) containing 5% methanol, and the sample was delivered at a rate of 0.5 µL/min using a Prince CE system. The sheath solution of 2-propanol/methanol (2: 1) was delivered using a Harvard syringe pump at a flow rate of 0.75 µL/min. An electrospray voltage of -5000 V was typically used as the ionization voltage.

and ester linkages. The triheptosyl inner-core unit is substituted by a β-D-glucopyranose residue at the O-4 position of HepI and by a phosphoethanolamine residue (PEtn) at the O-6 position of HepII. The β-D-glucopyranose attached to HepI is substituted at the O-6 position by a phosphocholine residue (PCho). HepIII is elongated through O-2 by oligosaccharides of varying length (R2) through sequential addition of sugar units. LPS glycoforms containing β-D-glucopyranose, lactose (β-D-Galp-(1-4)-β-D-Glcp), globotriose (R-D-Galp-(1-4)-β-D-Galp-(1-4)-β-D-Glcp), and sialyllactose (R-Neu5Ac-(2-3)-β-D-Galp-(1-4)-β-D-Galp) extensions have been identified by direct infusion ESI-MS.12,13,15,16 We have found that when H. influenzae strain 375 is grown under laboratory conditions, the major LPS glycoform populations carry globotriose at HepIII.13 Populations of glycoforms containing an additional PEtn substituent at O-4 of HepIII (R3) are also observed.8 Figure 1 shows the mass spectrum (m/z 400-1600) obtained from the negative ion ESI-MS of O-deacylated LPS of broth-grown H. influenzae strain 375. The ESI mass spectrum is dominated by molecular peaks corresponding to doubly [M - 2H]2-, triply [M - 3H]3-, quadruply [M - 4H]4-, and quintuply [M - 5H]5deprotonated ions. Molecular ions correspond to related glycoforms differing in the substitution pattern of the inner-core triheptose unit attached to O-deacylated lipid A.15 Triply charged ions at m/z 921.3, 962.3, and 1003.3 represent globotriosecontaining Hex4 glycoform populations (R2 ) globotriose) containing one, two, and three PEtn groups, respectively. The ion at m/z 962.3 arises from a mixture of isomeric glycoforms that carry

Figure 2. ESI-FAIMS-MS analysis of 10 ng/µL of O-deacylated LPS from H. influenzae strain 375: (a) compensation voltage spectra (DV 4000 V; CV 0-40 V; m/z 400-1600), (b) extracted mass spectrum of the CV from 12 to 17 V, (c) extracted mass spectrum of the CV from 22 to 29 V, (d) reconstructed CV spectrum for ions m/z 1003.5, (e) reconstructed CV spectrum for ions m/z 962.5, and (f) reconstructed CV spectrum for ions m/z 921.5. All other conditions were the same as those in Figure 1.

a second PEtn group at either KDO (R1 ) PEtn, R3 ) H) or HepIII (R1 ) H, R3 ) PEtn), while the ion at m/z 1003.3 contains PEtn at both positions. Related doubly charged ions at m/z 1382.3, 1444.3, 1505.8 and their sodium adducts are also observed in the spectrum. Sialylated glycoform populations are also expressed by this bacterium when an exogenous source of sialic acid is available for growth.8,12 Therefore, it is expected that the ions due to the sialyllactose-containing LPS glycoforms, such as m/z 753.3 ([M - 4H]4-), 1005.3 ([M - 3H]3-), and 1508.3 ([M - 2H]2-) would also be observed at low levels in this spectrum. However, these ions were not observed without front-end separation, because they are likely buried by the predominant ions at m/z 752.3, 1003.3, and 1505.8, since the degree of sialylation is limited to a few percentages of molecules under the growth conditions employed.12,13 ESI-FAIMS-MS. Under particular conditions of DV and CV, FAIMS acts as an ion filter by transmitting selected ions. To separate a mixture of ions, the CV can be scanned to obtain a compensation voltage spectrum (CV spectrum). In FAIMS-MS experiments, a conventional API 3000 microspray interface was employed, except that the distance from sprayer to orifice was increased from ca. 1 cm to ca. 2 cm. CV spectra were obtained

by scanning the compensation voltage applied to the FAIMS, while monitoring a given m/z range. For analysis of H. influenzae strain 375 O-deacylated LPS, the dispersion voltage was fixed at 4000 V and the CV was scanned from 0 to 40 V over the range of m/z 400-1600 (Figure 2). Under these conditions, the appropriate compensation voltages for all glycoforms could be grouped according to the charge state. Doubly charged ions corresponding to Hex4 glycoforms having one (m/z 1382.5), two (m/z 1444.5), or three (m/z 1505.5) PEtn groups were observed at low CV (1217 V, Figure 2b), whereas the corresponding triply charged ions having one (m/z 921.5), two (m/z 962.5), or three (m/z 1003.5) PEtn groups were observed at high CV (22-29 V, Figure 2c). The CV map for all the major glycoforms, differing in charge states and mass-to-charge ratios, shows that a majority of LPS glycoforms can be detected by operating FAIMS at a few discrete values of CV rather than scanning the CV across a wide range. This is illustrated in parts d and f of Figure 2 for the reconstructed CV spectra for ions m/z 1003.5, 962.5, and 921.5, respectively, which correspond to the triply charged major glycoform ions expressed by H. influenzae strain 375. The ability to reduce scanning requirements is very important for coupling FAIMS with CE-MS, due to the narrow peak widths. Although FAIMS can be used Analytical Chemistry, Vol. 76, No. 16, August 15, 2004

4679

Figure 3. CE-MS analysis of 1 ng/µL of O-deacylated LPS from H. influenzae strain 375: (a) TIE (m/z 600-1200) and RIE for m/z 962.5, (b) extracted mass spectrum from 6.86 to 7.16 min, and (c) extracted mass spectrum from 7.16 to 7.41 min. Separation conditions: 20 nL injection of O-deacylated LPS, bare fused-silica (90 cm × 50 µm i.d., 190 µm o.d.), 5% methanol in 30 mM morpholine, pH 9.0, +30 kV. The other conditions were the same as those in Figure 1.

to separate different ions based on their structural features, the reduction of chemical background noise rather than the separation of individual LPS glycoforms, which offered improved detection of the analytes in the biological sample, was the focus of this study. Detection Limits of CE-MS and CE-FAIMS-MS. Normally, O-deacylated LPS oligosaccharide glycoforms have an overall net negative charge above pH 4.0 and can be separated under anionic conditions with capillary zone electrophoresis.4,6,24 In this application, 30 mM morpholine buffer, containing 5% methanol with the pH adjusted to 9.0 with formic acid, was employed for all of the CE-MS and CE-FAIMS-MS experiments. The CE-MS analysis of a 1 ng/µL sample of O-deacylated LPS from H. influenzae strain 375 obtained using negative ion detection is presented in Figure 3. In our experience, the peak at 3.31 min in the total ion electropherogram (TIE) (m/z 600-1200) corresponds to positively charged impurities in the sample mixture. The peak at 4.85 min is related to the neutral solvents, a marker for electroosmotic flow (EOF) of the electrophoresis. However, no peaks corresponding to LPS glycoforms are visible in this electropherogram. Indeed the migration times for LPS ions are around 7 min, as indicated by the reconstructed ion electropherogram (RIE) for m/z 962.5 (Figure 3a). The extracted mass spectra were taken from the combination of ions over the TIE peaks (6.86 to 7.16 min and 7.16 to 7.41 min, respectively) and are shown in parts b and c of Figure 3. As expected, triply charged ions corresponding to Hex4 glycoforms having one, two, and three PEtn groups are the major ions observed. In this extracted spectrum (Figure 3b), the signalto-noise ratio for the ions at m/z 921.5, 962.5, and 1003.5 are (24) Thibault, P.; Richards, J. C. In Methods of Molecular Biology (Bacterial Toxins: Methods and Protocols); Holst, O., Ed.; Humana Press: Totowa, NJ, 2000; Vol. 45, pp 327-344.

4680 Analytical Chemistry, Vol. 76, No. 16, August 15, 2004

approximately 1.0, 4.3, and 3.7, respectively. These results indicate that the detection limit of CE-MS analysis for the major LPS glycoforms from H. influenzae is on the order of 1 ng/µL. Peaks resulting from the quadruply charged ions of the major LPS glycoforms were observed at m/z 721.5 and 752.5, although the peaks were not significantly above the background noise. Moreover, ions related to the sialylated LPS glycoforms were still very weak as shown in Figure 3c, perhaps due to the high background noise. In this example, the baseline of the TIE was around 3.2 × 106 counts per second (cps), resulting in background noise in the extracted mass spectra of between 6000 and 12 000 cps for the m/z range of 600-900, and 3000 to 6000 cps for the m/z range of 900-1200. The net signal intensities for ions at m/z 962.5 and 1003.5 are estimated as 6000 and 5000 cps, respectively. It was expected that the detection limit for CE-MS could be significantly improved if the chemical background could be reduced. On the basis of the initial results of the ESI-FAIMS-MS experiment (Figure 2b,c), fixed compensation voltages for transmitting the triply charged and doubly charged O-deacylated LPS ions were identified as 25 and 14.5 V, respectively. Negative ion CE-FAIMS-MS analysis of 1 ng/µL of O-deacylated LPS extracted from H. influenzae strain 375 is presented in Figure 4, in which the values of DV and CV were fixed at 4000 and 25 V, respectively. Under these conditions, a peak at 6.67 min and a shoulder peak with a higher migration time (6.82 min) are clearly observed above background. The average background for the TIE was found to be approximately 450 times lower than that in conventional CE-MS (i.e., 7000 cps vs 3.2 × 106 cps). Concomitantly, the signal intensities for the ions of interest were significantly increased. These observations are the result of atmospheric pressure ion

Figure 4. CE-FAIMS-MS analysis of 1 ng/µL of O-deacylated LPS from H. influenzae strain 375: (a) TIE (m/z 600-1200), (b) extracted mass spectrum from 6.56 to 6.86 min, and (c) extracted mass spectrum from 6.91 to 7.11 min. FAIMS conditions: DV 4000 V, CV 25 V. The other conditions were the same as those in Figure 3.

focusing that occurs in the cylindrical FAIMS analyzer as described.18 The extracted mass spectra across the TIE peaks in Figure 4 (from 6.56 to 6.86 min and 6.91 to 7.11 min, respectively) are similar to that shown in Figure 3, except that the apparent noise was virtually eliminated. It must be pointed out that these extracted mass spectra represent the average signal intensities over the electropherogram peaks, providing a qualitative comparison of signal-to-noise. In an effort to quantitatively evaluate the advantage of using FAIMS, a concentration of 1 ng/µL of LPS was employed for replicated injections in both CE-MS and CE-FAIMS-MS analyses with SIM scan mode. In the CE-MS experiments, the average peak area of six injections was calculated to be 8.4 × 104 counts for glycoform at m/z 962.5, with a relative standard deviation value of 15.1% (n ) 6). However, in the CE-FAIMS-MS experiments, the calculated average peak area of six injections was 6.4 × 105 counts for the same ions (m/z 962.5), with a relative standard deviation value of 3.2% (n ) 6). These data suggested that the signal of interest was increased as much as 7.5-fold when the FAIMS device was used. It is noteworthy that triply charged ions at m/z 1005.5 from a sialyllactosecontaining Hex3 glycoform (Hex2 Neu5Ac1 PEtn2) are observed at levels significantly above background under the FAIMS conditions (Figure 4c). Interestingly, the RIE at m/z 962.5 contains peaks at two migration times (data not shown), which indicated the existence of isomeric glycoforms. A quantitative study of ESI-FAIMS-MS has shown that the FAIMS transmission provides linear calibration for 10 nM to 10 mM leucine enkephalin.18 Since coupling the technique to CE-MS may be expected to be more difficult than conventional ESI-FAIMS-MS, the linearity of response of the CE-FAIMS-MS system was studied here using the O-deacylated LPS as analytes. The dependence of peak area on sample concentration was

investigated for replicate injections (six runs for each concentration) of serial dilutions of this O-deacylated LPS mixture over a range of 20 pg/µL to 100 ng/µL using SIM acquisition mode, and the plot of peak area versus concentration is shown in Figure 5. The calibration curve is linear for ∼3 orders (40 pg/µL to 10 ng/ µL) of magnitude with a correlation coefficient (r2) of 0.995 for the major glycoform. Excellent reproducibility of migration times was obtained over the six runs, with relative standard deviation (RSD) values less than 1.5%, whereas RSD values of 3.2% to 10.2% were achieved for peak areas. It is expected that the slightly higher variation of peak area, compared to conventional ESI-FAIMS-MS, may relate to the variations in the injection volume of capillary electrophoresis. If the content of glycoform Hex4 Hep3 PCho1 PEtn2 P1 KDO1 lipid A-OH (Mr 2888 Da) is assumed to be 50% of the total O-deacylated LPS mixture, the linearity corresponds to a molarity range of 7.0 fmol/µL to 1.7 pmol/µL. As indicated in Figure 6, a detection limit of ∼70 amol for the glycoform with triply charged ions at m/z 962.5 was reached for a 20 nL injection. However, the intensity for the ion at m/z 1003.5 was much weaker than expected for the same concentration. This may result from the presence of an extra PEtn in this glycoform (structure 1, R1, R3 ) PEtn) resulting in stronger interaction between this LPS glycoform and the capillary surface. Further improvement in sensitivity would be expected from employing preconcentration techniques and/or reducing the adsorption of O-deacylated LPS to the capillary column. CE-FAIMS-MS/MS. Although FAIMS resulted in separation of glycoforms containing one, two, and three PEtn groups, no evidence of isomeric LPS glycoforms was observed in the ESIFAIMS-MS experiment (Figure 2). However, the glycoforms of Hex4 Hep3 PCho1 PEtn2 P1 KDO1 lipid A-OH were revealed by the CE separation. On the basis of previous studies, glycoforms Analytical Chemistry, Vol. 76, No. 16, August 15, 2004

4681

Figure 5. Plot of peak areas vs concentration for O-deacylated LPS from H. influenzae strain 375 for the LPS glycoforms with m/z 962.5. Mass spectra were acquired using SIM scan mode with 25 ms per channel. The other conditions were the same as those in Figure 4.

Figure 6. CE-FAIMS-MS analysis of 20 pg/µL of O-deacylated LPS from H. influenzae strain 375: (a) TIE (m/z 600-1200) and (b) extracted mass spectrum from 6.50 to 6.91 min. The other conditions were the same as those in Figure 4.

Chart 2 having masses of 2766, 2889, and 3012 Da could be assigned to the Hex4 Hep3 PCho1 PEtn1 P1 KDO1 lipid A-OH, Hex4 Hep3 PCho1 PEtn2 P1 KDO1 lipid A-OH, and Hex4 Hep3 PCho1 PEtn3 P1 KDO1 lipid A-OH, respectively.8,13 The tandem mass spectra for glycoforms having masses of 2766 and 3012 Da, giving rise to ions at m/z 921.5 and 1003.5, correspond to the molecular structures 2 and 3, respectively (see Chart 2).8 These two LPS glycoforms contain one or three PEtn groups and give rise to only a single migration peak in the electropherogram. Two isomeric LPS glycoforms can give a molecular weight of 2889 Da, in which a second PEtn is added, either at HepIII or to the phosphate group at the KDO. To characterize the structures of these two isomeric glycoforms, CE-FAIMS-MS/MS experiments were conducted. The CE-MS/MS electropherogram for the precursor ion at m/z 962.5 is presented in Figure 7a where the two isomers are 4682

Analytical Chemistry, Vol. 76, No. 16, August 15, 2004

partially separated by CE. The MS/MS data obtained on ions extracted from 6.62 to 6.72 min and 6.98 to 7.05 min are shown in parts b and c of Figure 7, respectively. Both of the precursor ions generated a singly charged fragment ion at m/z 952.5, corresponding to the lipid A-OH. It is well established that O-

Figure 7. CE-FAIMS-MS/MS analysis of O-deacylated LPS from H. influenzae strain 375: (a) TIE (m/z 200-1200), (b) product ion spectrum of m/z 962.5 from 6.62 to 6.72 min, and (c) product ion spectrum of m/z 962.5 from 6.98 to 7.05 min. The separation conditions were the same as those in Figure 4. Elab: 120 eV (laboratory frame reference).

deacylated LPS samples cleave between the KDO-lipid A-OH bond under negative ion collisional activation dissociation.24 After losing the lipid A-OH, both of the isomers give rise to the doubly charged fragment ion at m/z 967.5. As previously observed,4,10 the fragment ion at m/z 219.5 may be assigned to the deprotonated PPEtn, which leads to the conclusion that a pyrophosphoethanolamine is attached to the KDO residue instead of a phosphate group in the ions at 6.7 min. This is the major isoform in the peak of lower electrophoretic mobility. In contrast, only a very weak fragment at m/z 219.5 was observed in the product spectrum of the LPS glycoform with higher electrophoretic mobility (∼7.0 min). Instead, this glycoform loses a phosphate group and a phosphoethanolamine to give the doubly charged product ions at m/z 919.5 and 897.5, respectively (Figure 7c). These data indicate that the isomer with lower electrophoretic mobility (shorter migration time, ∼ 6.7 min in Figure 7a) contains a KDO-PPEtn moiety, whereas the isomer with higher electrophoretic mobility contains KDO-P. The relative mobilities of these isomeric glycoforms can be related to the ionizations of phosphate versus pyrophosphoethanolamine at pH 9.0 (separation buffer). CONCLUSIONS FAIMS provides electronically controlled separation for gasphase ions at atmospheric pressure, based on the changes in mobility that occur at a high electric field strength. The ability to act as an ion filter and continuously transmit one type of ion, independent of mass-to-charge ratio, makes FAIMS ideal for CE-MS.

With the use of LPS from H. influenzae strain 375 as an example, the CV map for all of the major glycoforms that differ in charge states and mass-to-charge ratios shows that a majority of LPS glycoforms can be detected by operating FAIMS at a few discrete values of CV rather than scanning the CV across a wide range. The ability to reduce scanning requirements is very important for coupling FAIMS with CE-MS, due to the narrow CE peak widths. Although FAIMS can be used to separate different ions based on their structural features, the reduction of chemical background noise rather than the separation of individual LPS glycoforms was utilized in this study. The use of FAIMS to reduce mass spectral noise significantly improved detection limits. The linearity of response of the CE-FAIMS-MS system was studied using the LPS as analytes. The calibration curve was linear for ∼3 orders of magnitude, over a range of 40 pg/µL to 10 ng/µL. ACKNOWLEDGMENT The authors thank E. Richard Moxon and Derek W. Hood for the provision of Haemophilus influenzae strain 375 used in this study. Ade`le Martin is acknowledged for preparation of Odeacylated LPS and Kenneth Chan for technical assistance.

Received for review January 26, 2004. Accepted May 21, 2004. AC049850D

Analytical Chemistry, Vol. 76, No. 16, August 15, 2004

4683