Influence of Electrosorption, Solvent, Temperature, and Ion Polarity on

Sep 6, 2008 - Therefore, we studied the factors influencing detection and elution of charged oligosaccharides from PGC columns coupled to an ESI sourc...
7 downloads 13 Views 977KB Size
Download PDF
Anal. Chem. 2008, 80, 7534–7542

Influence of Electrosorption, Solvent, Temperature, and Ion Polarity on the Performance of LC-ESI-MS Using Graphitic Carbon for Acidic Oligosaccharides Martin Pabst and Friedrich Altmann* Department of Chemistry, University of Natural Resources and Applied Life Sciences (BOKU), 1190 Vienna, Austria Porous graphitic carbon (PGC) emerges as an ideal stationary phase for LC-ESI-MS of complex oligosaccharides. Therefore, we studied the factors influencing detection and elution of charged oligosaccharides from PGC columns coupled to an ESI source. Electrosorption by the carbon surface leads to total retention of very acidic glycans on instruments where voltage is applied to the spray needle. This problem can be eliminated by thorough electrical grounding. A point of general importance is the influence of ionic strength on the elution and peak shape of glycans containing several carboxylic acid groups in the form of sialic acids or uronic acids. Solvent pH had a marginal effect on the ionization efficiency in both ion polarities, but the content of organic solvent strongly influenced signal intensity of acidic glycans in the negative mode. As a consequence, detection in the positive ion mode appears preferable when neutral and charged glycans shall be quantitated in the same sample. While retention of neutral glycans is not affected by pH, sialylated species are retained somewhat stronger at acidic pH resulting in a larger spread of the entire elution range of N-glycans. Remarkably, retention of glycans on PGC increased at higher temperatures. The biological effects of protein glycosylation slowly but irresistibly move into the focus of biochemical research. This calls for a capable methodology allowing fast, sensitive, and yet detailed analysis of the complex mixtures of glycans occurring on glycoproteins or tissues. The combination of porous graphitic carbon chromatography (PGCC) with ESI-MS emerges as an especially powerful strategy, which has been successfully applied by several research groups.1-16 Some of these groups demonstrated the suitability of PGC-ESI-MS for highly sialylated and even sulfated * To whom correspondence should be adressed. Tel.: +43-1-36006 6062. Fax: +43-1-36006 6059. E-mail: [email protected]. (1) Davies, M. J.; Smith, K. D.; Carruthers, R. A.; Chai, W.; Lawson, A. M.; Hounsell, E. F. J. Chromatogr. 1993, 646, 317–326. (2) Barroso, B.; Didraga, M.; Bischoff, R. J. Chromatogr., A 2005, 1080, 43– 48. (3) Barroso, B.; Dijkstra, R.; Geerts, M.; Lagerwerf, F.; van Veelen, P.; de Ru, A. Rapid Commun. Mass Spectrom. 2002, 16, 1320–1329. (4) Hashii, N.; Kawasaki, N.; Itoh, S.; Hyuga, M.; Kawanishi, T.; Hayakawa, T. Proteomics 2005, 5, 4665–4672. (5) Itoh, S.; Kawasaki, N.; Hashii, N.; Harazono, A.; Matsuishi, Y.; Hayakawa, T.; Kawanishi, T. J. Chromatogr., A 2006, 1103, 296–306.

7534

Analytical Chemistry, Vol. 80, No. 19, October 1, 2008

glycans.7-9 They had chosen a slightly alkaline buffer (ammonium bicarbonate or ammonium acetate) and detection in the negative ion mode. Good results were, however, also obtained in the positive mode with an acidic solvent.17 These conditions allowed the separation of a large number of isomers of neutral and monoand disialylated glycans as they occur, e.g., on antibodies.18 The separation of isobaric structures in combination with single-stage MS or, in the near future, tandem MS or MSn has the potential to allow a fast but nevertheless highly reliable identification of glycan species to the very detail. At the same time, this approach does provide relative quantitation of isobaric variants, which is also achieved by MSn-based methodology but only with some difficulty.19 Researchers trying to implement PGC-ESI-MS will want to base the choice of experimental conditions on knowledge of the influence of the relevant factors. Moreover, they will want to avoid pitfalls such as failure to elute all components of their sample. The existence of such problems may be indicated by the observation that about nine years after the first very promising (6) Karlsson, N. G.; Schulz, B. L.; Packer, N. H.; Whitelock, J. M. J. Chromatogr B Anal. Technol. Biomed. Life Sci. 2005, 824, 139–147. (7) Karlsson, N. G.; Wilson, N. L.; Wirth, H. J.; Dawes, P.; Joshi, H.; Packer, N. H. Rapid Commun. Mass Spectrom. 2004, 18, 2282–2292. (8) Kawasaki, N.; Haishima, Y.; Ohta, M.; Itoh, S.; Hyuga, M.; Hyuga, S.; Hayakawa, T. Glycobiology 2001, 11, 1043–1049. (9) Kawasaki, N.; Ohta, M.; Hyuga, S.; Hyuga, M.; Hayakawa, T. Anal. Biochem. 2000, 285, 82–91. (10) Kawasaki, N.; Ohta, M.; Itoh, S.; Hyuga, M.; Hyuga, S.; Hayakawa, T. Biologicals 2002, 30, 113–123. (11) Ninonuevo, M.; An, H.; Yin, H.; Killeen, K.; Grimm, R.; Ward, R.; German, B.; Lebrilla, C. Electrophoresis 2005, 26, 3641–3649. (12) Schulz, B. L.; Sloane, A. J.; Robinson, L. J.; Prasad, S. S.; Lindner, R. A.; Robinson, M.; Bye, P. T.; Nielson, D. W.; Harry, J. L.; Packer, N. H.; Karlsson, N. G. Glycobiology 2007, 17, 698–712. (13) Wilson, N. L.; Schulz, B. L.; Karlsson, N. G.; Packer, N. H. J. Proteome Res. 2002, 1, 521–529. (14) Thomsson, K. A.; Karlsson, H.; Hansson, G. C. Anal. Chem. 2000, 72, 4543–4549. (15) Thomsson, K. A.; Karlsson, N. G.; Hansson, G. C. J. Chromatogr., A 1999, 854, 131–139. (16) Schulz, B. L.; Packer, N. H.; Karlsson, N. G. Anal. Chem. 2002, 74, 6088– 6097. (17) Pabst, M.; Bondili, J. S.; Stadlmann, J.; Mach, L.; Altmann, F. Anal. Chem. 2007, 79, 5051–5057. (18) Stadlmann, J.; Pabst, M.; Kolarich, D.; Kunert, R.; Altmann, F. Proteomics 2008, 8, 2858–2871. (19) Takegawa, Y.; Deguchi, K.; Ito, S.; Yoshioka, S.; Sano, A.; Yoshinari, K.; Kobayashi, K.; Nakagawa, H.; Monde, K.; Nishimura, S. Anal. Chem. 2004, 76, 7294–7303. 10.1021/ac801024r CCC: $40.75  2008 American Chemical Society Published on Web 09/06/2008

reports on the use of PGC in combination with ESI-MS,9,13-15,20 a forward citation analysis does not hint at a distribution of this technique much beyond the walls of the initially involved laboratories. The reproducibility of separations on PGC has recently been shown to be affected by the electrospray voltage.21 Retention times of the 4-hydroxy-3-methoxy-L-phenylalanine (3-methyl-DOPA) and other catecholamines were increased by column polarization, which could be overcome by electrical grounding.21 In this work, we show that polarization of a PGC column by the ESI voltage drastically affects the elution of acidic oligosaccharides, where it not only increases the retention of monosialylated glycans but even fully prevents the elution of glycans with several sialic acids. We further examined the effects of pH, ionic strength, and temperature on the performance of carbon columns in the analysis of different types of glycans. Positive and negative ion detection mode were compared with regard to their suitability for quantitative evaluation of LC-MS data. Finally, the influence of these parameters on glycosaminoglycan fragments was investigated. MATERIALS AND METHODS N-Glycans. Bovine thyroid stimulating hormone (bTSH), bovine fibrin, and bovine fetuin were purchased from SigmaAldrich (St. Louis, MO). Recombinant erythropoietin was obtained from Janssen-Cilag Pharma, (Vienna, Austria). The monoclonal antibody 4E10, produced in CHO cells, was kindly provided by Dr. Renate Kunert (Department of Biotechnology, BOKU, Vienna).18 N-Glycans were released from 2 mg of the glycoprotein by incubation with 1 U of peptide N-glycosidase F (Roche, Mannheim, Germany) in ammonium acetate buffer of pH 8.5 containing 0.2% Igepal CA-630 (Sigma-Aldrich). The glycans were purified by successive passage over a 50-mg reversed-phase SPE cartridge (Strata C18, Thermo Scientific, Waltham, MA) and Sephadex G15 (1 × 50 cm, GE Healthcare, Vienna, Austria). The PNGAse F released N-glycans of bTSH were fractionated on a PGC column (150 × 3 mm, 5 µm, Thermo Scientific) using 0.1% formic acid buffered to pH 9.0 with ammonia and 95% acetonitrile in water as solvents. A gradient was developed from 10-25% acetonitrile at a flow rate of 0.6 mL/min, and glycans were detected by UV absorption at 210 nm. This yieldedsbeside otherssa monosulfated hybrid-type (reference glycan 2) and a diantennary disulfated glycan (reference glycan 3) (Figure 1). These glycans both contain GalNAc instead of galactose, according to ref 22. Similarly, core-R1,6-fucosylated di-, tri-, and tetraantennary glycans with R2,3-linked sialic acids (glycans 4, 5, and 6, respectively) were isolated from recombinant erythropoietin. The neutral diantennary glycan 1 was prepared from fibrin and purified by chromatography on carbon.17 Fucosylated IgG-style N-glycans with zero, one, and two sialic acid residues were available from a previous study.18 Glycosaminoglycans. Hyaluronic acid from human umbilical cords (Sigma-Aldrich) was dissolved in 0.1 M citrate-phosphate (20) Itoh, S.; Kawasaki, N.; Ohta, M.; Hyuga, M.; Hyuga, S.; Hayakawa, T. J. Chromatogr., A 2002, 968, 89–100. (21) Tornkvist, A.; Nilsson, S.; Amirkhani, A.; Nyholm, L. M.; Nyholm, L. J. Mass Spectrom. 2004, 39, 216–222. (22) Green, E. D.; Baenziger, J. U. J. Biol. Chem. 1988, 263, 25–35.

Figure 1. Equimolar standard mixture. The six glycans are referred to by numbers in this article. In the proglycan nomenclature (www. proglycan.com), their abbreviations are A4A4, Man5su4-An4, su4An4su4-An4F6, Na3-4Na3-4F6, [Na3-4Na3-4]Na3-4F6, and [Na3-4Na3-4][Na3-4Na3-4]F6, respectively. The concentrations of the six compounds of the glycan mixture were evaluated by normal-phase HPLC of 2-aminobenzamide-labeled glycans. To circumvent problems arising from partial loss of R2,3-linked Neu5Ac, the glycans were desialylated prior to labeling. Peak areas differed by less than 5%.

buffer of pH 5.0 and degraded at 37 °C overnight with 0.2 mg/ mL hyaluronidase purified from bee venom.23 PGC-LC-ESI-MS (PGCC-MS). Oligosaccharide samples were reduced with borohydride as described recently.17 Aliquots containing ∼2.5 pmol of each reduced glycan were subjected to LC-MS on a 0.32 × 50 mm Hypercarb column (Thermo Scientific). Solvent A consisted of 65 mM formic acid buffered to pH 3.0 with ammonia if not otherwise stated. Solvent B was acetonitrile. The influence of pH was investigated with ammonium formate buffers of different pH values prepared from 65 mM formic acid in each case. To compare different published procedures, aqueous solvents A consisting of 10 or 20 mM ammonium bicarbonate 6,7,12,13 or of 5 mM ammonium acetate of pH 9.65 were tested. For all experiments, a flow of 8 µL/min was delivered by an Ultimate 3000 capillary LC system (Dionex, Sunnyvale, CA). After a loading time of 8 min, a rather steep gradient from 4.5 to 20% B over 30 min followed by a ramp to 80% in 12 min was applied. Analytes were detected by ESI-MS on a Q-TOF Ultima Global (Waters Micromass, Manchester, UK) operated in either the positive or negative ion mode. The applied ES voltage for negative ion detection was -2.9 kV and for positive mode +3.4 kV. For experiments with electrical grounding, grounded metal clamps were applied at both ends of the sleeve of the capillary column, whereby on the effluent side, the clamp was actually applied to a short steel capillary inserted between column and peek tubing. For runs with polarized carbon, at least one mock run was performed without grounding clamps before sample injection. (23) Kolarich, D.; Altmann, F. Anal. Biochem. 2000, 285, 64–75.

Analytical Chemistry, Vol. 80, No. 19, October 1, 2008

7535

Figure 2. Influence of the electrospray voltage on retention of N-glycans. The standard mixture (Figure 1) was subjected to chromatography on a porous graphitic carbon column with mass spectrometric detection in the positive (A and B) or negative mode (C and D). Runs A and C were conducted with grounded column, whereas chromatograms B and D were made one run (∼1 h) after removal of the grounding clamps. The mock run after regrounding of the column immediately after run D is shown in panel E. The compounds missing in run D elute when the carbon column was depolarized. The chromatograms shows the intensity (SIM) of the exact masses of the most abundant ion of each sugar (see also Supporting Information Figure S1). The gradient used here was much steeper as would be applied for isomer analysis.17,18

Column Regeneration. The column was cleaned regularly with 20 column volumes each of an acidic and a basic solvent, both in 50% acetonitrile. The acid was 0.3% TFA. The basic solution was 0.3 formic acid with 0.1 triethylamine buffered with ammonia to pH 10. Alternatively, the column can be manually filled with 4 M trifluoroacetic acid and heated to 115 °C for several hours followed by thorough washing with acetonitrile containing solvent. HPLC of Fluorescence-Labeled Glycans. HPLC-purified glycans were desalted with minicolumns containing 10 mg of carbon (Thermo Scientific),24 and the effluents were freeze-dried. Six pure glycans were mixed in an equimolar ratio (Figure 1). To evaluate the mixture, a part was desialylated by incubation with 1% formic acid for 20 min at 95 °C. After drying, the glycans were labeled with 2-aminobenzamide and subjected to normal-phase HPLC as described.25,26 RESULTS Influence of Electrospray Voltage on Retention of NGlycans on Carbon. A panel of oligosaccharides of known structures was prepared by fractionation of glycans with PGC (24) Packer, N. H.; Lawson, M. A.; Jardine, D. R.; Redmond, J. W. Glycoconjugate J. 1998, 15, 737–747. (25) Kolarich, D.; Turecek, P. L.; Weber, A.; Mitterer, A.; Graninger, M.; Matthiessen, P.; Nicolaes, G. A.; Altmann, F.; Schwarz, H. P. Transfusion 2006, 46, 1959–1977. (26) Bigge, J. C.; Patel, T. P.; Bruce, J. A.; Goulding, P. N.; Charles, S. M.; Parekh, R. B. Anal. Biochem. 1995, 230, 229–238.

7536

Analytical Chemistry, Vol. 80, No. 19, October 1, 2008

HPLC. The standard cocktail consisted of one neutral, two sulfated, and three sialylated glycans with different numbers of charged groups (Figure 1). The structures were assigned based on mass and literature data.22,27 Purity and complete sialylation of the glycans were assessed by PGCC-MS (data not shown). A mixture of these six compounds was prepared, and their equimolarity was verified by NP-HPLC of 2-AB labeled, desialylated glycans (Figure 1). This quantitation was performed with the desialylated compounds as complete labeling of the R2,3-sialylated glycans was accompanied by substantial desialylation. Sialic acids in R2,3-linkage are prone to lactone formation and therefore especially labile28ssomething that may escape the attention of the reader of literature on labeling by reductive amination. To save precious instrument time, a very fast gradient unsuitable for isomer analysis was used throughout this work. Essential for providing elution of all components was grounding of the column to prevent any electrical polarization of the stationary phase, which occurs although ESI source and column are connected by nonconductive fused-silica or peek tubing only (Figure 2). To give an idea of the dimensions, the potential at the column outlet at a flow rate of 8 µL/min was a bit over 0.5 V relative to the HPLC apparatus, not half the voltage of a standard (27) Hokke, C. H.; Bergwerff, A. A.; Van Dedem, G. W.; Kamerling, J. P.; Vliegenthart, J. F. Eur. J. Biochem. 1995, 228, 981–1008. (28) Perreault, H.; Costello, C. E. J. Mass Spectrom. 1999, 34, 184–197.

Figure 3. Influence of salt concentration. The neutral standard A4A4 (glycan 1) and tetrasialylated oligosaccharide (glycan 6) were chromatographed on graphitic carbon with 0.1% formic acid (pH 2.3) or formate buffers of increasing concentration. The buffers were made of 0.1-0.5% (v/v) formic acid (∼13-65 mM) adjusted to pH 3.0 with ammonia. Only the sialylated glycans’elution time and peak shape depended on the ionic strength of the eluent. Figures on the right side of the panels give the intensities of the peak of glycan 1. The elution positions of other, especially of sulfated, glycans are shown by arrows in some chromatograms. Glycans 5 and 6 were not eluted with the unbuffered solvent. Arrows designated hya4 to hya8 indicate the elution positions of hyaluronan oligosaccharides consisting of the indicated number of monosaccharides.

AA battery. The column was clamped on both sides to suppress the back-voltage from the ESI source. Optimal results were only obtained when the first grounding point was attached to a short steel capillary inserted between column and the peek tubing line connecting to the ESI sourcesa measure that goes well beyond usual safety precaution in LC-ESI. Only a single mock run after removal of the clamps, the same sample produced a totally different chromatogram (Figure 2B and D). While the neutral glycans were hardly affected by the grounding status, retention times of sulfated glycans strongly increased and sialylated glycans were not (positive ion mode, Figure 2B) or only partially (negative ion mode, Figure 2D) eluted. The missing tri- and tetraantennary glycans from the run shown in Figure 2D eluted in a consecutivesnow groundedsmock run (Figure 2E). Obviously, uneluted analytes did not fall prey to electrolytic degradation but were trapped on the carbon surface if the ES voltage was allowed to take effect on the column. The observations described so far were obtained with the standard conditions at pH 3.0, but strong electrosorption also occurred at higher pH values (data not shown).

Influence of Ionic Strength. With the grounded column, we compared the elution behavior of the two outpost references, i.e., the diantennary asialo 1 and the tetrasialylated glycan 6 (see Figure 1) at different ionic strengths of formate buffer of pH 3.0. Very evidently, only the sialylated oligosaccharide is affected by buffer concentration (Figure 3). Buffered 0.1% (26 mM) formic acid of pH 3.0 eluted glycan 6 only late and as a broad hump. Unbuffered formic acid, having an ever lower ionic strength, was unable to elute the tetrasialylated structure 6 anyway (Figure 3). Sulfated glycans are similarly affected by ionic strength (Figure 3). Only glycans with up to two sialic acids or sulfate groups could be eluted with formic acid alone, albeit as rather broad peaks. A negative aspect of increasing ionic strength is the decreasing ionization efficiency (Figure 3), even though the ESI source tolerated salt concentrations of up to 150 mM without abounding reduction of ionization efficiency (factor 4-5 in terms of ion count but less in signal-to-noise ratio). Mass spectrometrists are apprehensive toward the use of salt-containing solvents. However, Analytical Chemistry, Vol. 80, No. 19, October 1, 2008

7537

Figure 4. Influence of buffer ions and pH. The left panel shows the separation of the standard mixture with different buffers as aqueous solvent: 5 mM acetic acid buffered to pH 9.0 with ammonia (A), 10 mM ammonium bicarbonate of pH 8.0 (B), and 65 mM ammonium formiate of pH 3.0 (C). The figures on the right side of each trace give the maximal signal intensity of the base peak of compound 1. While the lower peak intensity of C compared to B can be explained by the salt concentration, the value of A points at the role of the buffer salt. The arrows in panels A-C give the elution position of hyaluronan oligosaccharides consisting of the indicated number of monosaccharides (i.e., hya4 consists of two repeating units). The right panel depicts the influence of pH using the same buffer, i.e., 65 mM formic acid brought to pH 10, 9.0, 4.5, and 3.0 with ammonia. (traces D-G, respectively). The pH value affected retention of the sialylated standard but barely the peak intensity. Circles mark peaks resulting from fragments or adducts of peaks of different mass, notably A4A4, which was the most abundant glycan in the mixture.

in the case of acidic oligosaccharides, the gain in chromatographic performance justifies the loss of ionization efficiency. Influence of pH on Retention of N-Glycans on a PGC Column. Investigations with more alkaline solvents such as 5 mM ammonium acetate at pH 9.6 as chosen by Itoh et al.5,20,29 and 10 mM ammonium carbonate (pH ∼8.0) as used by Karlsson et al.,6,7,12,30 revealed a decrease of retention of sialylated sugars with increasing pH (Figure 4). However, conclusions drawn from these data (Figure 4) are uncertain as the different buffer substances may themselves influence the result. An additional set of experiments using one buffer system (65 mM ammonium formate) at various pH values was performed and corroborated this notion (Figure 4). Solvent pH as well as ionic strength markedly affected sialylated but hardly at all neutral glycans (Figure 4). The retention difference between the diantennary asialo and the tetrasialylated glycan was much smaller at higher pH values (Figure 4). Thus, at lower pH values, carbon columns have a higher peak capacity. After having compared the resolution of glycans of different charge states at different pH values, we examined the separation of isobaric glycans. The peak pattern of neutral glycans remained essentially unimpressed by pH and even buffer (Figure 5). To our surprise, this also applied to mono- and disialylated glycans (Figure 5). The distances of the three disialylated species were essentially the same at the three pH values tested, i.e., ammonium (29) Hashii, N.; Kawasaki, N.; Itoh, S.; Harazono, A.; Matsuishi, Y.; Hayakawa, T.; Kawanishi, T. Rapid Commun. Mass Spectrom. 2005, 19, 3315–3321. (30) Estrella, R. P.; Whitelock, J. M.; Packer, N. H.; Karlsson, N. G. Anal. Chem. 2007, 79, 3597–3606.

7538

Analytical Chemistry, Vol. 80, No. 19, October 1, 2008

formate of pH 3.0 and 9.0 and ammonium carbonate of pH 7.9-8.0. However, while at the higher pH value the elution region of the neutral and mono- and disialylated glycans overlapped, they were shifted at pH 3.0 (Figure 5). It must be emphasized that these results were obtained with a very fast gradient with a slope of 0.667 versus 0.153%/min for analytical separations,17 where a higher gain in separation of differently sialylated species can be expected. Ion Types. In positive mode, the major pseudomolecular ion was either the [M + 2H]2+ or the [M + 3H]3+ ion whereby in the case of the tetrasialylated glycan 6 and the disulfated standard 3 the [M + NH4 + 2H]3+ ions predominated (see Supporting Information Figure S1). In negative mode, the acidic glycans yielded very clean [M - xH]x signals, but the neutral standard partly yielded unidentified adducts. It shall be noted here that only the most abundant ion species of each glycan was considered for quantitations and SIM traces in this work, regardless of the number and intensity of other ions. “In source” fragmentation was generally very low. The highest degree of fragmentation was observed in positive mode at pH 8 (see Supporting Information Figure S2). In addition to desialylation of sialic acid, loss of a complete trisaccharide antenna could be observed. The sulfated glycans 2 and 3 underwent some desulfation in positive mode. Spectra made at different times, and hence somewhat different settings of collision gas (argon) pressure, insinuated that much of the “in source” fragmentation actually occurs in the collision cell. However, this just corroborates the importance of chromatographic separation before MS analysis.

Figure 6. Influence of column temperature on elution time. The six compounds were mixed in roughly similar amounts and chromatographed at increasing temperature. The upper panel shows the PGC chromatograms at lowest and highest temperature tested.The lower panel shows the Van’t Hoff plot. The negative slope of the lines implies a positive entropy of adsorption to the stationary phase. Figure 5. Effect of buffer on the separation of isobaric compounds. A mixture of glycans with either neutral, monosialylated, or disialylated glycans was subjected to chromatography on porous graphitic carbon with 65 mM formate buffer at pH 3.0 or 20 mM ammonium carbonate at pH of ∼8.0. The SIM traces representing neutral digalactosylated with or without core fucose are shown as panels 3-A and 8-A depending on the pH. Panels B depict the results for monosialylated glycans and panels C for the three disialylated isomers.

Influence of Temperature. Increasing temperature is expected to reduce the affinity of analytes to the stationary phase. Indeed, this was also found for CPG as a phase for supercritical fluid chromatography using phenolic compounds as model analytes.31,32 Retention of oligosaccharides by PGC in LC, however, clearly increases with higher temperatures resulting in a negative slope in a Van’t Hoff diagram (Figure 6). In other words, the negative free energy for binding originates from a strongly positive entropy value, which compensates the positive enthalpy of binding. A simple calculation ingnoring possible influences of phase ratio or temperature dependence of enthalpy33 yielded a ∆H value of ∼4 kJ/mol and a ∆S of ∼15 J/(mol · K) for nonsulfated glycans. To the authors’ knowledge, this has hitherto not been reported for PGCC. It remains to be shown, whether this phenomenon can be exploited in work with smaller, weakly retained oligo- or monosaccharides.34 (31) (32) (33) (34)

Cui, Y.; Olesik, S. V. Anal. Chem. 1991, 63, 1812–1819. Engel, T. M.; Olesik, S. V. Anal. Chem. 1990, 62, 1554–1560. Chester, T. L.; Coym, J. W. J. Chromatogr., A 2003, 1003, 101–111. Antonio, C.; Larson, T.; Gilday, A.; Graham, I.; Bergstrom, E.; ThomasOates, J. J. Chromatogr., A 2007, 1172, 170–178.

Ion Polarity and Quantitation. The ESI-MS can be operated in any polarity, when the column is electrically decoupled. It seems to be logical that acidic glycans are measured in the negative ion mode, which in fact allows the most sensitive detection. In PGCC, the late elution of sialooligosaccharides and the strong effect of the percentage of organic solvent on their ionization in negative mode further boost this favorable fact (Figure 7). However, neutral and also monosialylated glycans cannot keep pace with their highly charged cousins (Figure 7), and therefore, negative mode detection causes a strong bias toward acidic glycans. As a consequence, positive ion mode is preferable for samples with mixtures of neutral and acidic glycans as it gives a much more proportionate picture (Figure 2). The profile in Figure 2 shows the SIM trace for the most intense signal of each compounds. This simple approach yielded results in good agreement with those obtained by HPLC of fluorescent derivatives even without the application of any correction factors (Figure 2). Sulfated glycans gave a mixed picture: both were underrepresented in positive mode and only the disulfated glycan did better in negative mode (Figure 2). By the way, the general limit of detection in positive mode with the pH 3.0 system and a column of 0.32-mm inner diameter was well below 100 fmol, which comfortably allows the analysis of PAGE bands or spots with a decent Coomassie staining. Analysis of glycans from a recombinant antibody underpinned the advantage of positive mode detection for PGC-LC-ESI (Figure 8). This sample has recently been analyzed by LC-ESI-MS of glycopeptides as well as by normal-phase HPLC of 2-aminobenAnalytical Chemistry, Vol. 80, No. 19, October 1, 2008

7539

Figure 7. Influence of ion polarity and acetonitrile concentration on detection sensitivity. A mixture of neutral diantennary (open circles) and tetrasialylated (black squares) standard glycans was added via a splitter to the solvent flow from the gradient pump delivering increasing concentrations of acetonitrile. The experiment was performed in the positive (left panel) and negative ionization modes (right side). The internal rectangles frame the solvent percentage range relevant for chromatography. From carbon columns, highly sialylated glycans elute at higher acetonitrile concentrations than neutral glycans, which amplifies their overrepresentation in the negative mode.

zamide-labeled glycans.18 The result obtained in positive ion mode in this work shows a quantitative profile similar to that of the normal-phase HPLCsboth at pH 3.0 and at 7.9 (Figure 8). The relative amounts of disialylated N-glycan was ∼10% higher when measured by PGC-LC-ESI while the monogalactosylated glycan was underrepresented. The ESI-Q-TOF system may introduce bias at the stage of ionization or transfer efficiency of ions of differing mass. On the other hand, loss of 3-linked sialic acid may lead to underrepresentation of sialylated glycans by NP-HPLC.18 To conclude, positive ion detection provides quantitative data that are a priori quantitatively representative. As with all other methods, the quest for the true value to the very percent can only be answered with the help of validated reference material, which, however, is not yet available. PGC-ESI-MS of Uronic Acid-Containing Oligosaccharides. PGC has been applied to glycosaminoglycan-derived oligosaccharides with either ammonium carbonate or 0.1% formic acid.2,6 We used hyaluronate oligomers to assess how far the observations described above applied to compounds containing uronic acids. Indeed, polarization of the carbon surface completely prevented the elution of these oligomers (see Supporting Information Figure S3). The pH value likewise had a pronounced effect. Generally, retention of all oligomers (tetra to octa) was much stronger at acidic pH. Under both alkaline conditions, tetra- to octasaccharides eluted before or at the position of the neutral diantennary N-glycan 1, while at pH 3, the octasaccharide eluted long after the tetrasialylated standard 6. With low ionic strength (0.1% formic acid), the hexasaccharide was the largest oligomer to be eluted within the actual gradient before the washing step. Thus, the need for ionic strength for hyaluronate glycans was similar to that for sialylated N-glycans albeit possibly not that stringent. The impressive previous demonstrations of the suitability of PGCC for sulfated glycosaminoglycan fragments2,6,30,35 obviates any further experiments in this direction. We want to point out that only in studies using buffered eluents were oligomers of two or more repeating units described.6,30,35 (35) Antonopoulos, A.; Favetta, P.; Helbert, W.; Lafosse, M. J. Chromatogr., A 2007, 1147, 37–41.

7540

Analytical Chemistry, Vol. 80, No. 19, October 1, 2008

PGC-ESI-MS of Nucleotide Phosphates. A mixture of UMP, UDP-glucuronic acid, GTP, and ATP was subjected to PGC-ESIMS using 65 mM formic acid as such or buffered to pH 3.0 or 9.0. In keeping with a recent study on nucleoside phosphates,36 unbuffered acid could only elute UMP. The buffered solvents of acidic and alkaline pH gave roughly the same elution pattern with the trinucleotides eluting in the region of the neutral N-glycan 1 (see Supporting Information Figure S3). Thorough grounding of the PGC column was an absolute prerequisite for the elution of triphosphates. Notably, much harsher conditions were needed to elute sugar phosphates in a recent studyspossibly to surmount electrosorption.34 Column Aging and Regeneration. The strong adsorption capability of graphitic carbon unfortunately leads to rapid “aging” of a column, which leads to decreasing retention of analytes. While this “fouling” may sometimes facilitate elution of strongly retained substances with solvents of low ionic strength or in the absence of proper electrical grounding, this phenomenon is opposed to the want for a reproducible chromatographic behavior. The outstanding stability of PGC fortunately allows the use of strong acidic and basic eluents. Cleaning by the procedure described in this paper restores the retention capabilities to that of a new a column. DISCUSSION Graphitic (or graphitized) carbon has been acclaimed as a most useful stationary phase for LC-MS for many years now.9,37,38 Compared against this promise, only very few groups, have, however, credibly reported on the detection of highly sialylated glycans.3-5,7,9,10,20 These groups probably have used or use either a sufficiently aged column, a column mounting system that confered electrical grounding, or an electrospray system, where (36) Xing, J.; Apedo, A.; Tymiak, A.; Zhao, N. Rapid Commun. Mass Spectrom. 2004, 18, 1599–1606. (37) Kawasaki, N.; Ohta, M.; Hyuga, S.; Hashimoto, O.; Hayakawa, T. Anal. Biochem. 1999, 269, 297–303. (38) Wuhrer, M.; Deelder, A. M.; Hokke, C. H. J. Chromatogr., B: Anal. Technol. Biomed. Life Sci. 2005, 825, 124–133.

Figure 8. PGCC-ESI-MS analysis of a recombinant antibody in positive and negative mode. Glycans from the monoclonal antibody 4E10 produced in CHO cells18 were analyzed with the fast gradient in both ionization modes. The chart at the bottom shows the comparison of the MS results with the quantitation obtained by normal-phase HPLC of 2-aminobenzamide-labeled glycans (white bars) with the results of carbon LC with MS detection using either formate of pH 3 (black bar) or ammonium carbonate of pH 8 (gray bars).

the ionization voltage is not directly applied to the spray needle (e.g., an Agilent or Bruker ESI source). All these groups have used eluents of some ionic strength. Other, less lucky researchers may have given up work with carbon columns after a few frustrating trials. A prime aim in HPLC analysis is the ability to elute all analytes applied to the column. When PGC-LC-ESI-MS shall be exploited for the analysis of isomers of complex glycans, reproducibility of elution times becomes crucial. Because of initial troubles to provide these two features on our capillary-flow ESI-Q-TOF system, we investigated various parameters and found several points that are essential for success with PGC columns and that

challenge the view that they are just reversed-phase columns with a stronger retention for hydrophilic analytes.39 The most dangerous pitfall on all instruments where voltage is applied to the ESI source is the small voltage leaking back to the column through the LC solvent in the peek or fused-silica capillary. The resulting polarization of the carbon surface21 causes abnormal or total retention of charged oligosaccharides. Notably, the nonappearance of tetrasialylated N-glycans was due to trapping on the column and not to electrolytic degradation (Figure 2). The surface polarization problem, which reminds one of the multiple uses of graphite as an electrode material, can easily be circumvented by electrical grounding. An alternative strategy toward reproducible elution of acidic compounds is column aging by repeated use. However, both the exact conditions and a stable end point for column depolarization by this route are difficult to define. Therefore, we propose the use of new or thoroughly cleaned PGC columns together with suitable solvents. Using a mixture of six reference glycans comprising neutral, sulfated, and multiply sialylated species, we investigated the ability of various solvent systems to elute all components. Only eluents of a certain ionic strength were able to elute all six glycans. The minimal salt concentration promoting elution of sialooligosaccharides was lower in the alkaline range. An acidic eluent (i.e., formate buffer of pH 3.0) requires a higher buffer strength and thus decreases ionization yield. This, however, is compensated by a stronger retention of sialylated glycans (Figures 4 and 5). This effect diminishes the incidental overlapping of peaks of different degrees of sialylation and minimizes problems arising from fragmentation. Although fragment peaks are almost negligible except for rather small glycans,17 they turn into a problem when very different amounts of glycans are present in the same peak. In the analysis of recombinant therapeutics, for example, the reliable determination of small amounts of incompletely sialylated glycans is crucial and requires separation from the major glycans. For the analysis of sialylated N-glycan isomers, we have recently chosen a 65 mM formate buffer (from 0.25% formic acid) of pH 3.0.17 From the present results, it can be concluded that a slightly lower ionic strenght would do to elute all components of an N-glycan sample as sharp peaks. Together with a slightly lower pH, this may give an even better selectivity and at the same time a higher detection sensitivity. In summary, the two most useful solvent systemssin our handssare 10 mM ammonium carbonate (pH 7.8-8.0), which provides highest sensitivity and maximum elution power for various analytes, and ∼30-65 mM ammonium formate of pH 2.6-3.0, because of the higher selectivity. Alkaline ammonium acetate caused a 20% higher pressure than other buffers without providing a compensatory advantage. Another result was that all the components of the N-glycan mixture (neutral, sialylated, and sulfated glycans) could be detected in both ion modes. Sialylated glycans performed well in both detection modes. They could be measured with highest sensitivity in the negative mode, but positive ion detection provided a much more proportionate response of neutral and sialylated glycans. Notably, higher signals in negative mode were only obtained for glycans with two or more sialic acid residues, (39) Chin, E. T.; Papac, D. I. Anal. Biochem. 1999, 273, 179–185.

Analytical Chemistry, Vol. 80, No. 19, October 1, 2008

7541

whereas neutral and monosialylated species were drastically underrepresented as seen in the experiments with monoclonal antibody glycans (Figure 8). A recent multimethod study showed comparable results for IgG glycans measured in negative or positive ion mode, but the small amounts of sialylated glycans occurring there had been outside the focus of this work.40 Figure 2 demonstrates that mono- and disulfated glycans are likewise differently displayed in different detection polarities. Taken together, positive mode detection has the advantage of delivering roughly realistic proportions without application of correction factors for simultaneous quantification of neutral and sialylated glycans. As the results somewhat depend on the settings of the mass spectrometer, the precise true value can nevertheless only be deduced from the control experiments with precharacterized samples, e.g., of glycan mixtures, where the individual components have been quantitated by validated methodology. Negative mode detection clearly has its merits when only multiply charged glycans shall be analyzed, especially as a substantial degree of the highly informative cross-ring cleavages occurs in MS/MS.41,42 Glycosaminoglycan fragments will likewise (40) Wada, Y.; Azadi, P.; Costello, C. E.; Dell, A.; Dwek, R. A.; Geyer, H.; Geyer, R.; Kakehi, K.; Karlsson, N. G.; Kato, K.; Kawasaki, N.; Khoo, K. H.; Kim, S.; Kondo, A.; Lattova, E.; Mechref, Y.; Miyoshi, E.; Nakamura, K.; Narimatsu, H.; Novotny, M. V.; Packer, N. H.; Perreault, H.; Peter-Katalinic, J.; Pohlentz, G.; Reinhold, V. N.; Rudd, P. M.; Suzuki, A.; Taniguchi, N. Glycobiology 2007, 17, 411–422. (41) Sagi, D.; Peter-Katalinic, J.; Conradt, H. S.; Nimtz, M. J. Am. Soc. Mass Spectrom. 2002, 13, 1138–1148. (42) Vakhrushev, S. Y.; Zamfir, A.; Peter-Katalinic, J. J. Am. Soc. Mass Spectrom. 2004, 15, 1863–1868.

7542

Analytical Chemistry, Vol. 80, No. 19, October 1, 2008

be measured in negative mode2,6 as long as monocharged saccharides are not an issue. A curious detail is the unexpected increase of retention at higher column temperatures. Unlike the situation with many other stationary phases, analytes bound to carbon appear to have a higher entropy than in the free state. This anomaly constitutes an aspect that must be taken into account by theories on the mechanism of analyte binding by PGC. To conclude, this work points out a number of factors relevant for the successful application of PGC columns for LC-ESI-MS of charged oligosaccharides. These factors may also turn out useful for the analysis of other substance classes. ACKNOWLEDGMENT The Q-TOF Ultima Global was financed by the Austrian Council for Research and Technology Development. The work was not supported by the Austrian Science Fund. We thank Renate Kunert for the 4E10 sample, Johannes Stadlmann for his patiently waiting for us to finish measurements, and, as so often, Thomas Dalik for valuable technical assistance. SUPPORTING INFORMATION AVAILABLE Additional information as noted in test. This material is available free of charge via the Internet at http://pubs.acs.org.

Received for review May 20, 2008. Accepted August 8, 2008. AC801024R