Anal. Chem. 1993, 65,1034-1042
1834
Peptide Mapping and Evaluation of Glycopeptide Microheterogeneity Derived from Endoproteinase Digestion of Erythropoietin by Affinity High-Performance Capillary Electrophoresis Robert S. Rush,*>+ Patricia L. Derby,t Thomas W.Strickland,$and Michael F. Rohdet Department of Protein Structure, Mail Stop 14-2-A-229, and Department of Protein Chemistry, Mail Stop 14-2-A-223, AMGEN Znc., Amgen Center, Thousand Oaks, California 91320-1789
High-performance capillary electrophoresis (HPCE) has been employed to characterize the peptide map of recombinant human erythropoietin (rHuEPO)expressed from Chinese hamster ovary (CHO) cells. The methodology employs an ion pairing agent, 100 mM heptanesulfonic acid in 40 mM sodium phosphate buffer, pH 2.5, to increase peptide resolution, to decrease analyte wall interactions, and to evaluate glycopeptide microheterogeneity. The total tryptic map is segregated into two regions, nonglycosylated and glycosylated peptides. Reproducibility of the peptide map is excellent; the map results in baseline separation of 16 tryptic peptides and one doublet peak composed of two peptides (resolution 0.22). The map furthermore allows for the evaluation of the microheterogeneity associated with the three rHuEPO glycopeptides. At least 12 glycopeptide forms were separated in the initial peptide map. Peptides were identified by Edman sequencing, and the glycopeptides were further subjected to Dionex anion-exchange chromatography. To simplify the level of complexity associated with the glycopeptides, much of the characterization employed asialoglycopeptides and employed several endoproteolytic digestions. The relative percent distribution for each purified asialoglycopeptide was calculated to define the level of complexity and to tentatively assign a known structure to the HPCE peak. The level of structural complexity oft he asialoglycopeptides appears to increase from the simplest 0-linked form to the more complex N83, N38,and N24 glycosylation positions, respectively. HPCE evaluation of glycopeptide microheterogeneity appears to be simpler, faster, and just as sensitive as other more frequently employed methods for glycopeptide characterizations.
INTRODUCTION High-performance capillary electrophoresis (HPCE) as an analytical methodology has been extensively reviewed,1-3 and specific applications directed toward proteins and/or peptide
* Address correspondence to this author.
Department of Protein Structure. Department of Protein Chemistry. (1) Kuhr, W. G.; Monnig, C. A. Anal. Chem. 1992,64,389R-407R. (2)Hjerten, S.Electrophoresis 1990 11, 665-690. (3)Jorgenson, J. W. Methods (San Diego) 1992,4 (3),179-188.
+ 8
0003-2700/93/0365-1834$04.00/0
separations have been d i ~ c u s s e d . ~HPCE , ~ methodology is finding increased usefulness in the characterization of complex biochemicalseparations, Le., peptide mapping of recombinant Frequently,these proteins are posttranslationally modified by glycosylations a t either/or both the N or 0 sites9Jo by mammalian expression systems. To date, the normal analytical approach has been to characterize either the protein or the carbohydrate portion of the glycoprotein separately and then combine the results. Our approach has been to develop an analytical methodology that can address both the translationally expressed protein and the posttranslationally modified protein simultaneously by affinity HPCE peptide mapping.sJ1 The model protein system employed to evaulate this methodology was recombinant human erythropoietin (rHuEP0). rHuEPO is an ideal choice: the protein is small and very well-characterized;12-16the extent of glycosylation positions and the structures of the carbohydrates have been extensively c h a r a c t e r i ~ e d ; land ~ - ~the ~ effect of carbohydrate on the structure and stability of rHuEPO have also been (4)Novotny, M. V.;Cobb, K. A.; Liu, J. Electrophoresis 1990,11,735749. (5)Rush, R. S.In Applications of Enzyme Biotechnology;Kelly, T. W., Baldwin, T. O., Eds.; Academic Press: New York, 1992;pp 233-250. (6)Grossman,P.D.;Colburn,J.C.;Lauer,H.H.;Nielson,R.M.;Riggin, R. M.; Sittampalam, G. S.; Rickard, E. C. Anal. Chem. 1989,61,11861194. (7)Wu, S.-L.; Teshima, G.; Cacia, J.; Hancock, W. S. J . Chromatogr. 1990,516,115-122. (8)Rush, R. S.;McGinley, M. D.; Stoney, K. S.; Rohde, M. F.Methods (San Diego) 1992,4 (3),189-202. (9)Kornfeld, R.; Kornfeld, S. Annu. Reo. Biochem. 1985,54,631-664. (10)Lai, P.-H.; Everett, R.; Wang, F.-F.;Arakawa, T.; Goldwasser, E. J . Biol. Chem. 1986,261,3116-3121. (11)Guttman, A.; Cooke, N. Anal. Chem. 1991,63,2038-2042, J.F.;Smalling,R.;Egrie, (12)Lin,F.-K.;Suggs,S.;Lin,C.-H.;Browne, J. C.; Chen, K. K.; Fox, G. M.; Martin, F.; Stabinsky, Z.; Badrawi, S. M.; Lai, P.-H.; Goldwasser, E. h o c . Natl. Acad. Sci. 1985,82,7580-7584. (13)Jacobs, K.; Shoemaker, C.; Rudersdorf, R.; Neill, S. D.; Kaufman, R. J.; Mufson, A.; Seehra, J.; Jones, S. S.; Hewick, R.; Fritsch, E. F.; Kawakita, M.; Shimizu, T.; Miyake, T. Nature 1985,313,806-810. (14)Yamaguchi, K.; Akai, K.; Kawanishi, G.; Ueda, M.; Masuda, S.; Sasaki, R. J . Biol. Chem. 1991,266,20434-20439. (15)Davis, J. M.; Arakawa, T.; Strickland, T. W.; Yphantis, D. A. Biochemistry 1987,26,2633-2638. (16)Recny, M.A.; Scoble, H. A.; Kim, Y. J . Biol. Chem. 1987,262, 17156-17163. (17)Sasaki, H.; Bothner, B.; Fukuda, M. J . Biol. Chem. 1987,262, 12059-12076. (18)Takeuchi, M.; Takasaki, S.; Miyazaki, H.; Kato, T.; Hoshi, S.; Kochibe, N.; Kobata, A. J . Biol. Chem. 1988,263,3657-3663. (19)Higuchi, M.; Oh-eda, M.;Kuboniwa, H.; Tomanoh, K.; Shimonaka, Y.; Ochi, N. J . Biol. Chem. 1992,267,7703-7709. (20)Tsuda, E.; Goto, M.; Kunihisa, A.; Ueda, M.; Kawanishi, G.; Takahashi, N.; Sasaki, R.; Chiba, H.; Ishihara, H.; Mori, M.; Tejima, H.; Endo, S.; Tejima, Y. Biochemistry 1988,27,5646-5654. (21)Takeuchi, M.; Inoue, N.; Strickland, T. W.; Kubota, M.; Wada, M.; Shimizu, R.; Hoshi, S.; Kozutsumi, H.; Takasaki, S.; Kobata, A. R o c . Natl. Acad. Sci. U.S.A.1989,86,7819-7822. 0 1993 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 65, NO. 14, JULY 15, 1993
studied.22 Although our detection system is several orders of magnitude less sensitive than either fluorescent methods2S26 or laser-induced fluorescent detection28927 of derivatized sugars, we estimate that our mass limit of detection is on the order of 10-25 fmol. This level agrees with the literature for limits of mass detection by UV/vis adsorption for capillary electrophoresis.%The protein requires little sample handling, other than endoproteinase digestion, and no derivatization reactions are required. The rationale for employing heptanesulfonic acid as an ion pairing agent for peptide mapping is as follows: (1) the negatively charged acid ion pairs with basic amino acid residues, thus reducing the net charge on the analyte such that the electrophoretic mobility is reduced; (2) the negatively charged acid is electrostatically repelled from the wall of the fused silica capillary, thus decreasinganalytewall interactions; and (3)glycopeptidesrespond similarly to the nonglycosylated peptides, thus enhancingthe separation range of glycopeptides based on the contribution of the carbohydrate structure on the effective charge of the glycopeptide. Quantitative aspects of HPCE have been reviewed,29 and it has been clearly shown that analyte wall interactions led to decreased selectivity.30 Many different approaches have been taken to deactivate the silica wall interactions.13 Blocking these interactions, either by coating or by addition of buffer complexation reagents, should enhance the ability to observe sharp well-defined peaks. In this paper, we investigate reproducibility of migration times, peak area, plate counts per meter, and percent distribution of various glycopeptide forms associated with specific N-linked or 0-linked glycosylation positions of rHuEPO endoproteinase-derived peptides. Wherever possible, we compare HPCE results to those obtained by classical analytical procedures. The goals are to further develop HPCE analysis of peptide mapping and to extend this strategy to glycopeptide mapping.
EXPERIMENTAL SECTION Protein Purification. rHuEPO was expressed from either Chinese hamster ovary (CHO)cella or fromEscherichia coli and purified to homogeneityby sequentialcolumn chr0matography.B The primary source of rHuEPO was CHO-expressed material, unless otherwise designated. Endoproteinase Digestion Conditions. Endoproteinase digestions were conducted employing standard procedures at 1:50 (w/w) trypsin (Boehringer Mannheim, Indianapolis, IN) to nonreduced rHuEPO at 37 O C for 16 h. An aliquot of the sample was then dried by vacuum evaporation (Speed Vac) and reconstituted in 20 p L of a 10% glacial acetic acid solution for HPCE. The finalpeptide concentration was 3.75 mg/mL unless otherwise stated. Other sample pretreatments involved Arthrobacter ureafaciens neuraminidase (Calbiochem, La Jolla, CA) or Nglycanase (Genzyme, Cambridge, MA) to remove sialic acid residues or N-linked carbohydrate prior to trypsin digestions, respectively. Glu-C (ProMega, Madison, WI) digestions were also employed aa described.31 HPCE Conditions. An Applied Biosystems (Foster City,CA) Model 270 HT HPCE was employed. HPCE conditions and (22) Narhi, L. 0.;Arakawa, T.; Aoki, K. H.; Elmore, R.; Rohde, M. F.; Boone, T.; Strickland, T. W. J. Biol. Chem. 1991,266,23022-23026. (23) Yeung, E. S.; Kuhr, W. G. Anal. Chem. 1991,63,275A-282A. (24) Albin, M.; Weinberger, R.; Sapp, E.; Moring, S.Anal. Chem. 1991, 63,417-422. (25) Honda, S.; Makino, A.; Suzuki, 5.; Kakehi, K. Anal. Biochem. 1990,191, 228-234. (26) Liu, J.; Shirota, 0.; Wieeler, D.; Novotny, M. h o c . Natl. Acad. Sci. U.S.A. 1991,88,2302-2306. (27) Liu, J.; Shirota, 0.; Novotny, M. Anal. Chem. 1991,63,413-417. (28) Yeung, E. D. LC GC 1989, 7 and 118-128. (29) Goodall, D. M.; Williame, S.J.; Lloyd. D. K. Trends Anal. Chem. 1991,10, 272-279. (30) Cobb, K. A.: Dolnik. V.: Novotnv. - . M. Anal. Chem. 1990.62.2478. . 2489. (31) Derby, P. L.; Strickland, T. W.; Rohde, M. F. In Techniques in Protein Chemistry W,Angeletti, R. H., Ed.; Academic: San Diego, 1993.
1895
solvent optimization was as previously described for affinity HPCE peptide mappine using 100 mM heptanesulfonic acid (Regis Chemical Co., Morton Grove, IL) ion pairing reagent in 40mM sodium phosphate buffer, pH 2.5. The electropherograms were monitored at 200 nm, 0.025 AUFS at a collection rate of 15 Hz at a temperature of 30 "C. Data was collected and analyzed with Beckman Instruments, Inc., CALS system PeakPro (Waldwick, NJ) run on a Micro VAX 3300 computer from Digital Corp. (Maynard, MA). HPLC and Dionex Conditions. HPLC reverse-phase fractionation of tryptic fragments and anion-exchange chromatography of the carbohydrate moieties were as described.31 Edman Sequencing. Peptide identity was confirmed by sequentialEdman degradationon a Hewlett Packard (PaloAlto, CA) Model GlOOOA protein sequencer according to the manufacturer's instructions.
RESULTS AND DISCUSSION Peptide Mapping. A representative rHuEPO affinity HPCE tryptic map is shown in Figure 1. Most notably, the map is segregated into two portions; the first is the nonglycosylated peptides which migrate past the detector within 32 min, and the second part is the tryptic glycopeptides which exhibit depressed migration and clear the detector in 80 min. Eighteen tryptic peptides are identified within the first 30 min from a theoretical possible of 21 peptides, assuming complete digestion. The other three peptides are associated with three N-glycosylation positions; N24 plus N38 are associatedwith the same tryptic peptide, N83 and the 0-linked site (serine 126) reside on separate tryptic fragments. Obviously, considerable microheterogeneity is associated with the carbohydrate structure(s) as indicated by the number of peaks observed for the glycopeptides. At least 12 glycopeptide forms were partially or totally separated as indicated in Figure 1(discussed in detail below). It is possible that not all of the sialic acid-containingglycopeptideswere observed under these conditions; however, it is believed that within experimental limits of determination,all the peptides were observed. Direct comparison of total integrated area associated with the rHuEPO tryptic map to that observed for a sialidase-treated rHuEPO sample indicated comparable areas for comparable loadings (discussed below, see Figure 4). The injection or buffer spike was characteristically observed a t 7.8 min. Figure 2A,B represents expanded views of the nonglycosylated and glycosylated segments of the rHuEPO HPCE map for duplicate back-to-back injections from a common tryptic digest. The majority of peptides are baseline separated the exception is the peptide pair 14 and 15. The resolution factor for these two peptides is approximately 0.22. Asymmetry factors are near one, indicating optimal loading and near-Gaussian-distributedpeaks. Positive identification of the specific peak@)required collection of the reverse-phase HPLC peaks followed by identification of peak migration times and purity by HPCE and by N-terminal amino acid sequencing on each isolated peak. The late migrating glycopeptides still exhibit considerable wall interaction even with this solvent system as evidenced by the shallower and much broader peaks (discussed below). Table I presents the composite statistics associated with HPCE mapping of rHuEPO with this system over a 3-month period of time. The statistical sample was composed of both CHO and E. coli rHuEPO. Five separate trypsin digestions were used, two of which were pretreated with sialidase and/ or N-glycanase. The sialidase removes the sialic acid resides from both the N- and 0-linked oligosaccharides while the N-glycanase cleaves the N-linked oligosaccharides leaving an aspartic acid residue in place of the original asparagine residue. Two capillaries were used for this study. The migration times and area counts are compared for the nonglycosylated peptides (Table I, peaks 1-18) and the major glycopeptides (Table I, peaks 19-30). Mean and standard
1836
ANALYTICAL CHEMISTRY, VOL. 65, NO. 14, JULY 15, 1993
i
10 9
20 19
I
I
26
I
I
28 11
Non-glycosylated
Clycosylated
I
I
I
I
I
I
I
10
20
30
40
50
60
70
I
BO
I
90
Migration Time ( m i n ) profile of trypsindigested rHuEPO (3.75 mg/mL), vacuum injected for 2 s into a 50-pm i.d., 360-pm 0.d. capillary of total length (L)to effective length (I) of L / / 7 5 / 5 0cm, respectively. The sample was dissolved in 10% glacial acetic acid for electrophoresis. Electrophoresls was conducted at 16 000 V at 30 O C (electric field 213 V/cm) in 40 mM sodium phosphate buffer, pH 2.5, containing 100 mM heptanesulfonic acid ion pairing agent; the profile was monitored online at 200 nm, 0.025 AUFS at a data collection rate of 15 Hz for 95 min. The current level was about 110 pA for a power load of about 2.35 Wlm. The nonglycosylated peptide and the glycopeptide sections of the map are designated and the corresponding HPCE peaks are numbered according to migration time. Flgure 1. HPCE
deviations are reported with 6' of freedom. The peptide number represents the order of HPCE migration and can easily be compared to Figures 1 or 2 for identification. The percent error in migration times ranged from a low of 0.19 to a high of 1.29 for the nonglycosylated peptides and from 1.13 to 2.03 for the glycopeptides. These errors can be greatly reduced by comparing back-to-back injections from a single digestion. Under these conditions, the percent error in migration times of the nonglycosylated peptides for triplicate analysis yields an error range from 0.04 to 0.47 (data not shown). Integration of the peak areas generates an error range of 2.0-14 % for the nonglycosylated peptides. Three peptides (numbers 5, 11, and 16) were outside this range with errors of 41, 38, and 1 7 % , respectively, for the statistical sample. The large errors associated with peptides 5 and 11are likely due to difficulty in tryptic cleavage. Peptide 5 (residues 1114) is adjacent to and contiguous with peptide 11 (residues 5-10 and 155-162) which contains an intrachain disulfide linkage involving cys 7 and cys 161. Errors of about 4.8 and 1.1%were noted for peptides 5 and 11, respectively, when examined from a common digest, in agreement with the overall range reported for the larger statistical sample. Peptide 16 (residues 117-131) contains the 0-linked carbohydrate site, and the error associated with this peptide agrees in magnitude with the other glycopeptides which exhibit peak area errors in the range of 2.5-36.7 % , Glycopeptides numbered 22-25 and 30 present the largest errors, represent the smallest amount present, and are pushing the level of detectability. The microheterogeneity of the carbohydrate structures will be discussed further.
Evaluation of the number of theoretical plates associated with each peptide can give important information.32 The plates per meter of capillary are also presented in Table I and clearly show improved separations and elevated plate counts compared to available open-tube literature values; more typical efficiencies are in the range of 30 000-125 000 plates/ m.33 Operating under our conditions provides separations approaching 25-40 % of the theoretical maximum of a million plate~/m.3~>3~ In general, the percent plate count error varied from a low of 2 to a high of 27 with 4O of freedom for the peaks reported. Plate counts were calculated for the nonglycosylated peptides and the three largest glycopeptides peaks, numbers 19, 20, and 26, respectively. The error associated with these calculations could be reduced considerably by computer calculation of plate counts rather than by manual measurements. Separation efficiencies of the other smaller glycopeptide peaks were not calculated because of the difficulty in accurately measuring the width a t half-height manually. The reduction in plates of these glycopeptides, compared to the nonglycosylated peptides, appears to be at least a factor of 10 for those glycopeptides not reported in Table I but numbered in Figure 1. The solvent system employed here provides separations for the nonglycosylated peptides in the open-tube mode equivalent to those reported for acrylamide-coated capillaries in terms of plate counts;30 however, the full potential of HPCE separation of glycopeptides has yet to be achieved. The method is extremely easy (32) Giddings, J. C.J. Chromatogr. 1989, 480, 21-33. (33) Liu, J.; Dolnik, V.; Hsieh, Y.-Z.; Novotny, M. Anal. Chern. 1992, 64, 1328-1336. (34) Jorgenson, J. W.;Lukacs, K. D. Science 1983,222, 266-272.
ANALYTICAL CHEMISTRY, VOL. 85, NO. 14, JULY 15, 1993
1837
11
15
I16
A
-L
-
E
-
::
10
9
1
50
1
60
Migration
70
Time
so
90
(min)
Flgurr 2. (Panel A): Expanded vlew of duplicate back-to-back Injections of the nonglycosylatedpeptide section of the rHuEPO HPCE tryptlc map. (Panel B): Expanded vlew of back-to-back injections of the tryptic glycopeptide section of the rHuEPO map. Condltlons and peak numberlng as In Figure 1.
to use and can be employed to differentiate and identify microheterogeneity within the tryptic peptide map of rHuEPO. Figure 3 presents a comparison of CHO-expressedrHuEPO that has been treated with N-glycanase to that of E. coli
expressed rHuEP022 (nonglycosylated) by HPCE peptide mapping. Six peptides were observed that migrated differently in the E. coli preparation compared to the CHOexpressed material: they are labeled a-f, and the numerical designation correspondswith CHO-expressed tryptic peptides
1838
ANALYTICAL CHEMISTRY, VOL. 65, NO. 14, JULY 15, 1993 11
-
1 3
2-
-
-.
12
2-
d
U
L
0-
~
~
T
I
12
'
:3
I
'
14
I
'
15
I
16
'
I
17
'
I
'
19
18
I
'
20
I
21
\ligration
'
22
Timc
I
'
23
I
24
'
~I r ' - I r - - r - - r - r - r + 25 26 27 26 29 30 31 32
33
34
(niin)
Figure 3. Comparison of Mglycanase-treated CHO cell-expressed (lower trace) versus E. coll expressed (upper trace) rHuEPO tryptic maps by HPCE. The numerlcally labeled peaks are those reported In Figure 1. Observed differencesbetween the maps are deslgnated by an alphanumeric label. Other conditions are as in Figure 1.
Table I. Evaluation of Reproducibility of HPCE Erythropoietin Tryptic Map Peak Parameters. Deak Darameters peak no.
migration time
area counts
platesim
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
10.67 i 0.04 11.18 f 0.04 12.92 f 0.05 14.71 f 0.03 15.47 f 0.06 15.73 f 0.07 16.65 f 0.07 18.20 f 0.09 18.94 f 0.12 19.50 f 0.11 20.79 f 0.08 22.42 f 0.14 23.14 f 0.14 24.52 i 0.17 24.60 f 0.16 25.20 f 0.15 31.13 f 0.30 32.09 f 0.42 43.68 f 0.49 50.26 f 0.70 53.49 f 0.67 54.96 f 0.75 62.75 f 0.93 64.21 f 0.85 66.60 f 0.86 71.70 f 1.29 73.42 f 1.35 75.70 f 1.52 78.13 f 1.17 80.00 f 1.62
4.52 f 0.51 5.54 f 0.50 5.51 f 0.69 13.80 f 0.36 4.82 f 1.97 38.02 f 0.77 31.27 f 1.10 35.16 f 4.57 50.53 f 2.47 66.50 1.42 69.11 f 6.39 143.62 f 4.74 84.52 f 2.25 121.88 f 4.94 172.50 f 1.35 142.65 f 4.52 19.64 f 0.92 14.00 f 1.93 111.12 f 3.44 182.72 f 4.61 9.82 f 0.57 10.03 f 2.27 24.74 f 9.08 20.98 f 4.51 22.61 f 4.08 156.09 f 0.51 36.23 f 1.99 67.06 f 3.66 8.08 f 0.97 16.54 f 3.49
258 000 f 27 000 250 000 f 37 000 275 OOO f 22 300 259 800 f 55 OOO
*
ND 198 000 f 49 000 264 OOO f 35 000 327 OOO f 70 000 283 000 f 76 OOO 245 000 f 24 000 366 000 f 87 000 267 000 f 32 000 264 000 f 38 000 239 000 f 1000 164 000 f 22 000 282 000 f 29 000 166 000 f 13 000 193 000 f 38 000 191 000 f 5 OOO 163 000 f 2 OOO
ND ND ND ND ND 117 000 f 6 000
ND ND ND ND
Data presented represent the mean f SD for N equal to 7. ND = not determined.
shown in Figures 1 and 2. These differences are likely due to some of the known differences between the E. coli preparation and that observed from the CHO-expressed material. The differencesare as follows: N-terminal sequence analysis has shown that E. coli expressed rHuEPO exists in
a des-1 and des-2 form in addition to the fully processed species, which is the sole product of CHO cell expression;22 two peptides derived from N-glycanase conversion of asparagine residues to aspartic acids (one from N24 + N38 and the other from N83 in the CHO sample); the peptide which contains the 0-linked carbohydrate in CHO rHuEPO cells but is not glycosylated in E. coli;and the C-terminal peptide which in E. coli expressed rHuEPO is arg 166, a residue that is posttranslationally removed in CHO cell-expressedrHuEPO. Conclusive identification of these peptide differences has not yet been completed, which illustrates the importance of developing a definitive method such as HPCE MS for structural evaluation. The fact that baseline separation of most of these peptides can be obtained illustrates the importance of high plate counts and resolution in HPCE peptide mapping. The electrophoretic mobility is linear with respect to charge ( Q ) to mass (n is the number of amino acid in the peptide) ratio as previouslydescribed).8ss A plot of the electrophoretic mobility versus In (Q + l)/n0.43for selected rHuEPO peptides gave a slope of 3.82 X 1 V , an intercept of 2.54 X l t 5 ,and the linear correlation coefficient of 0.95 which is in close agreement with that reported for platelet-derivedgrowth factor peptides.6 The peptides were selected based on knowledge of the N-terminal sequence analysis, and those peptides that exhibited putative charge suppression were not included in the mobility versus effective charge plot. Additionally, modeling the rHuEPO map with this approach for known sequences allows for the calculation of the expected migration time based on known amino acid sequence data. Comparison of the calculated versus the observed migration times would then serve the dual function of confirming that all peaks were observed in the peptide map and that the peaks corresponded with known sequences. Should the calculated migration time not agree with the observed, then the reasons for the discrepancy can be investigated. For example, the calculated migration times for des-1, des-2, and the C-terminal (35) Grossman, P. D.; Colburn, J. C.; Lauer,
1989,179, 28-33.
H.H. Anal. Biochem.
VOL. 85, NO. 14, JULY 15, is03
CHEMISTRY,
ANALYTICAL
iaae
---->
c c
4 c z
4 z
%
-L
3 4
-
c 4
20
30
40
Migration
50
60
Time
70
80
90
(min)
Flgurr 4. Comparison of the effect of SlaHdase and Mglycanase treatments on the total CHO-expressed rHuEPO HPCE tryptlc peptide maps. (Trace A): control rHuEPO map, (trace B): slalidase-treated rHuEPO, and (trace C): Nglycanase-treated rHuEPO. Other condltlons are as In Figure 1 and as given In the text.
Ih
B
A
20
I
I
I
I
I
I
I
30
40
50
60
70
a0
90
3ligration
Time
(rnin)
Flgurr 5. Evaluationof the aslalo mlcroheterogeneltyassoclated wlth each glycosylation positionsof rHuEPO by HPCE mapplng the reverse-phase collected fractions. (Trace A): O-llnked site, (trace B): N83 slte, (trace C): N24 N38 slte, and (trace D) aslalo rHuEPO total peptlde control map. Other conditions are as In Figure 1.
+
peptide derived from E. coli rHuEPO are 17.2,14.7, and 19.2 min, respectively. The observed times were 16.6, 14.5, and 25.6 min, respectively. The major discrepancy was noted
between the calculated and the observed migration time for the suspected C-terminalpeptide. This discrepancy is likely due to charge suppression,which is sequence-dependent,and
1840
ANALYTICAL CHEMISTRY, VOL. 65, NO. 14, JULY 15, 1993
Table 11. Asialo Oligosaccharide Distribution on C H O Call-Expressed rHuEPO. % distribution HPCE Ref 19
b
Ref 31
carbohydrate structure
N24
N38
N83
N24
N38
1N83
N24
N38
N83
biatennary triantennarp tetraantennary tetraantennary-Lac-l tetraantennary-Lac-2 tetraantennary-Lac-3
15.5 11.4 34.9 3.6 19.1 4.3
3.9 6.5 31.3 26.1 8.4 9.9
1.4 10.0 31.8 32.1 16.5 4.7
16.5 26.7 35.3 17.0 4.3 0.3
0.10 5.4 55.3 30.3 8.3 0.5
NDb 10.4 56.4 25.8 6.1 1.0
15.6 14.2 24.8 22.0 11.9 1.4
3.6 13.2 39.0 30.5 12.2 2.5
2.5 3.3 46.1 30.8 14.0 3.4
a Percentages for HPCE based on integrated peak areas, carbohydrate mapping by re~erse-phase,'~ and anion-exchange ~hromatography.3~ ND = not detected or assiened. There are two Dossible structures: here we report the sum.
-
I
1
1 w 1
N
A &
+-,
, - r - F - - T T r . - , ~ - - ~ - - 7
50
1
.
1
r- - ' " ' ' " ' I " " '
60
hligration
70
Time
so
(min)
Evaluation of the microheterogenelty associated with N24 and N38 asiaioglycopeptides generated from Glu-C dlgestion of rHuEPO. The peptides were isolated by reverse-phase HPLC fractionation, concentrated to 3 kL, and analyzed in the second dimension by HPCE: sample matrix blank (trace A), N24 asiaiogiycopeptide (trace B), and N38 asialoglycopeptide (trace C). Other conditions are as in Figure 1 and as given in the text. is related to the primary structure of this peptide, TGDR. for the observed mobility deviations; similar to that observed The migration times of peptides d-f appear consistent with for the E. coli carboxy terminal peptide mentioned above the expected migrationtimes for the glycopeptidesof rHuEPO and to that reported for a number of proteins.36 Nevertheless, which would be nonglycosylated in this map of E. coli modeling can provide useful information. Utilizing HPCE mapping techniques to study carbohydrate microheterogeexpressed material. The lack of direct correspondence neity will be discussed below for the rHuEPO glycopeptide between the CHO material and the E. coli is likely due to the structures. aspartic acid generated from asparagine during the N-glycanase reaction and to the presence of 0-linked carbohydrateGlycopeptide Mapping. Figure 4 compares the tryptic containing peptides on the intact CHO sample (Figure 3). maps of intact (trace A), to sialidase-treated (trace B),and Modeling the separation has its limits, as noted above for N-glycanase-treated (trace C) rHuEPO. Clearly, removing the C-terminal peptide, but it may help with peak identifithe sialic acid residues results in an increased mobility due cation if the sequence is known. Furthermore, not all peptides to the removal of negative charges and to the reduction in separate according to this empirical mode1,35 most notably, mass, and appears to result in three groups or regions of t h e glycopeptides did not, which is not surprising because of glycopeptides. The first region exhibits a migration time of the changes in the effective charge due to the carbohydrate. 35 min and is composed of a single peak; the second region Removing the sialic acid residues with sialidase not only is composed predominately of a triplet that migrates at 45reduces the charge associated with the glycopeptide, but it 50 min, and the last region is composed of a multiplet of at also reduces the mass; the result is increased mobility least seven peaks migrating from 55 to 68 min. Removal of compared to the normal glycopeptide (see Figure 4, glycothe N-linked sugar simplifies the profile by eliminating the peptide mapping below). Other interactions between the second and third group of glycopeptides. These two groups carbohydrate side chains, such as increased wall interactions contain the three N-linked carbohydrate structures whose or changes in the hydrodynamic volume of the glycopeptides, mobilities have been altered by N-glycanase treatment could also explain mobility deviations from the model systems. (compare Figure 4, panels A and C). The two large peaks Some of the other nonglycosylated peptides also did not fit (43.7 and 50.3 min, respectively), in either the control map this model well. These peptides exhibited amino acid or from the N-glycanase-treated sample, correspond to the sequences that were consistent with a charge suppression 0-linked peptide as shown by their collapse to a single peak mechanism, thereby altering the effective charge of the respective peptide. These charge changes may be responsible (36) Compton, B. J. J. Chronatogr. 1991,559, 357-366.
ANALYTICAL CHEMISTRY, VOL. 65, NO. 14, JULY 15, 1993
1841
&ss+Ea+m' Triantennary
Figure 7. Major carbohydrate structures assoclated wlth CHO-expressed ~ H U E P O . ~ ~ - * ~ O ~ ~
(35 min) after sialidase treatment (compare Figure 4, panels
A and B). Microheterogeneity associated with this peptide is likely to focus on the degree of sialylation; the peak a t 35 min in trace B is surmised to lack sialic acid, whereas the two peaks at 43.7 and 50.3 min in trace A are thought to contain one and two sialic acid residues, respectively. Further evaluation of glycopeptide microheterogeneity by W C E utilized sialidase-treated samples to reduce the level of complexity introduced by the presence of sialic acid residues. Identification of the specific asialoglycopeptide associated with the tryptic digests involved collecting the reverse-phase HPLC peaks and subjecting them to individual second dimensional HPCE analysis. Three glycopeptides were collected and subjected to HPCE as shown in Figure 5. The O-linked site (trace A), N83 (trace B), and N24 +N38 site (trace C) from the asialopeptide map were compared to the complete map (trace D). The amino acid sequence of each HPLC glycopeptide pool was confirmed by N-terminal sequencing. There appears to be a single asialo form associated with the O-linked carbohydrate. Multiple forms
of asialoglycopeptides were observed for the N83 and for the N24 + N38 sites, respectively. It was possible to compare and tentatively assign a known carbohydrate structure to the associated HPCE peak areas based on the percent area distribution for site N83. To make these assignments, knowledge of the distribution of carbohydrate forms determined from anion-exchangechromatography was req~ired.'~~sl From the relative percentages present and the predicted variation in asialoglycopeptidemobility with size of oligosaccharide, the three major peaks in N83 (Figure 5, trace B)are consistent with the following known structures: tetraantenary, 47 min; tetraantennary-Lac-l,49 min; and tetraantennary-Lac-2, 51 min. These three peaks represent approximately 94% of the total area associated with this peptide. Smaller amounts of three other carbohydrate structures are also present and reported in Table I1 for the major asialo forms of carbohydrate structures known to be present in rHuEPO (see also Figure 7). The asialo N24 + N38 tryptic glycopeptide forms (Figure 5, trace C) are still too complex to make structural assignments
1842
ANALYTICAL CHEMISTRY, VOL. 65, NO. 14, JULY 15, 1993
based on HPCE peak area distributions. In an attempt to reduce the level of complexity, N24 and N38 were separated by reverse-phase chromatography after Glu-C digestion (cleavage after glutamic acid residues) of rHuEPO followed by sialidase treatment of the collected peptides. The respective peaks were identified by N-terminal sequence analysis, concentrated, and subjected to HPCE analysis as shown in Figure 6. The degree of microheterogeneity associated with each of these peptides is apparent. Comparison of the integrated HPCE areas to known relative amounts of asialo carbohydrate structures is shown in Table 11. Based on the relative percent distribution, it would appear that the N24 tetraantenary and tetraantenary-Lac-l glycopeptide forms may be comigrating. This conclusion is based on the relative distribution of forms presented in Table 11; the HPCE distribution results appear to be inflated by approximately 10% for the tetraantennary form and about 10-15 % lower than expected for the tetraantennary-Lac-1 form. N38 also exhibits a complex pattern and, when compared with known relative distributions, appears remarkably similar to N24. The three largest peaks (migration time 66,69, and 71.8 min, respectively) represent about 66% of the total integrated area. The other six peaks can be divided among the other carbohydrate structures (TableI1and Figure 7). Both N24 and N38 asialoglycopeptides exhibited more peaks than identified in Table 11, indicating more than six different forms of carbohydrate associated with each glycosylation site. Incomplete sialidase treatment could generate a multiplicity of forms, but in our case, the reduction in migration time is consistent with the removal of sialic acid residues. In general, HPCE analysis appears to be consistent with the conclusions reached by others17-19,21931 concerning carbohydrate microheterogeneity. It is encouraging that the strategy employed for peptide mapping can also be extended to asialoglycopeptide separations as well. Utilizing this approach, it appears possible (37)Towns, T.K.; Regnier, F. E. Anal. Chem. 1991, 63, 1126-1132. (38)Hoffetetter-Kuhn, S.;Paulas, A.; Gassmann, E.;Widmer, H. M. Anal. Chem. 1991,63, 1541-1547. (39)Honda, S.;Suzuki, S.; Nose, A.; Yamamoto, K.; Kakehi, K. Carbohydrate Res. 1991, 215, 193-198. (40)Rudd,P. M.; Scragg, I. G.; Coghill, E.;Dwek, R. A. Glycoconjugat e J. 1992, 9,86-91. (41) Landers, J. P.; Oda, R. P.; Madden, B. J.; Spelsberg, T. C. Anal. Biochern. 1992,205, 115-124.
to evaluate both the translational (peptide map) and posttranslational activities (glycopeptide map) associated with protein expression systems in a single HPCE map. The level of carbohydrate complexity in our system, as in other systems, is a measure of posttranslational processing and can be rankordered as follows: 0-linked, N83, N38, and N24 in increasing order of structural complexity. This ordering is also consistent with the literature.19J1 Considerable work remains to definitively identify all carbohydrate structures with the corresponding HPCE migration time. Figure 7 pictorally represents the major carbohydrate structures associated with rHuEPO glycosylation positions. It is important to note that the lactose group may be placed on any branch of the carbohydrate structures,resulting in different conformations and potentiallyexhibiting different electrophoretic mobilities as a result. The ability to accurately quantitate is dependent on reproducible injections and on reducing disruptive wall interaction^.^^-^^,^^ Utilization of the buffer additive heptanesulfonic acid represents a dynamic approach to reducing wall interactions and to improving separations and quantitation. The simplicity and reproducibility of the present system is attractive; however, chemically modified capillary surfaces represent yet another approach to solve the same problems with high e f f i c i e n ~ y .Several ~ ~ ~ ~ fluorometric methodologies have been published focusing on detectability and quantitation of carbohydrates as individual sugars or glycopeptide conjugate~.2”27~3~~ Finally, it is anticipated that the next major analytical improvements for the characterization of glycopeptides and oligosaccharide structures by HPCE may well be in the area of coupled HPCE mass spectrometry.
ACKNOWLEDGMENT The assistanceand collaboration of Mr. Charles Branhnam in supplying reverse-phase purified Glu-C-generated N24 and N38 rHuEPO glycopeptides is acknowledged. Dr. Alan Herman’s comments and helpful discussions were also appreciated.
RECEIVEDfor review January 8, 1993. Accepted March 31, 1993.
