Mass Spectrometry of the Humanized Monoclonal Antibody Campath 1H

Patent and Trademark Office) is a humanized monoclonal antibody directed against the CDw52 antigen. This antigen is an abundant molecule present on mo...
2 downloads 0 Views 828KB Size
Anal. Chem. 1995, 67, 835-842

Mass Snectrometrv of the Humanized Monoclonal Antibody CAMPATH IH David S. Ashton, Christopher R. Beddell,* David J. Cooper, Sarah J. Craig, Anne C. Lines, Ronald W. A. Oliver, and Marjorie A. Smith The Wellcome Research Laboratories, Beckenham, Kent BR3 3BS, U.K.

TWOmass spectrometric techniques, electrospray ionization (ESI) and matrix-assisted laser desorption ionization (MALDI) have been used to study the intact humanized monoclonal antibody CAMPATH l H , its fully and partially deglycosylated species, and 13 fragments prepared from it. The transformed ESI mass spectra of the glycosylated species gave complex patterns of molecular masses (il4:s). These have been substantially assigned to the presence of a mixture of glycoforms, each resulting from the combination of a single protein species with specitic glycans of four distinct masses. The MALDI mass spectra of the glycosylated species, with the exception of that of the smallest fragment Fc/2, which indicated the presence of three of the glycans, gave single Mrvalues comparable to the mean Mr calculated from the ESI results. The Mr values for the 10 prepared nonglycosylated species support the validity of the published amino acid sequence for the antibody and define the cleavage sites for the enzymic fragmentations. It is concluded that mass measurement of the Fc/2 fragment using ESI techniques provides a convenient means of preliminary assessment of the major glycosylated entities. The analyte forming the subject of this paper (CAMPATH 1H; trade mark of Wellcome group companies, registered in the US. Patent and Trademark Office) is a humanized monoclonal antibody directed against the CDw52 antigen. This antigen is an abundant molecule present on most, if not all, human lymphocytes.' CAMPATH 1H has been shown to eliminate large numbers of tumor cells, resulting in remission for patients with non-Hodgkin lymphoma without antiglobulin response,2 and it is undergoing clinical trials. It is produced for this purpose by gene expression in cultured Chinese hamster ovary cells and purified using a multistage chromatographic and ultrafiltration p r o ~ e s s .CAM~ PATH 1H is, like other antibodies, an immunoglobulin (IgG) whose molecular structure may be summarized by reference to the diagram shown in Figure 1. It consists of two -24 kDa light polypeptide chains (LC) and two -49 kDa heavy polypeptide chains (H-C) linked together by two interdisulfide (LC) - (H-C) bridges and two interdisulfide (H-C)- (H-C) bridges to form a Y-shaped molecule (Figure 1). Each molecule also contains a total of 12 intrachain disuK.de bridges and an asparagine residue (301) in each heavy chain Pigure 1) that is amenable to glycosylation. (1) Hale, G.; Xia, M.-Q.; Tighe, H. P.; Dyer M. J. S.;Waldmann,H. TissueAntigen 1990,35,118-127. (2) Hale, G.; Dyer, M. J. S.; Clark, M. R; Phillips, J. M.; Marcus, R; Reichmann, L;Winter, G.; Waldmann, H. Lancet 1988,2,1394-1399. (3) Page, M. J.; Sydenham, M. A.BioTechnologv 1991,9,64-68.

0003-2700/95/0367-0835$9.00/0 0 1995 American Chemical Society

PvGlu

C i s - - - Cis233 Glycan

I

Glycan

H-C

H-C

Component

Description

H-C(1); L-C(1)

Reduced and alkylated (withiodoacetamide) heavy and light chains Reduced heavy and light chains Complete L-C plus fragment of H-C,specifically Pyroglutamate (1) - Histidine (228) Dimer of [L-C plus fragment of H-C,specifically Pyroglutamate (1) -Leucine (238)J Monomeric fragment of H-C,specifically Threonine (229) Glycine (450) Dimer of Fc/2 fragments, joined by two interchain disulphide bridges

H-C(I1); L-C(II) Fab F(ab')2 Fc/2 Fc

-

Figure 1. Schematic diagram of intact CAMPATH 1H IgG and tabulated description of the component species analyzed. Disulfide bridges are shown by dashed lines.

Now as is evident from the numbering of these chains for CAMPATH 1H in Figure 1, the complete amino acid sequences of both chains have been ~pecified,~ and further, an overall carbohydrate analysis has been perf~rmed.~The remaining unsolved analytical problems posed by this glycoprotein are those of determining the number of distinct oligosaccharide component glycans, their exact molecular structure, and the percentages of each in a particular batch. These problems are common to all glycoproteins and they are not easily solved, especiallyin the case of the immunoglobulins where they are compounded by the (4) Crowe, J. S.; Hall, V. S.; Smith, M. A; Cooper, H. J.; Tite, J. P. Clin. Ex@.

I W Z W Z U1992,87, ~O~. 105-110. (5) Lines, A. C.; Lines, D. IC, unpublished.

Analytical Chemistty, Vol. 67, No. 5, March 1, 1995 835

loo

,~ z

T

J

2300

:

: I :

2500

~ 2700

: : 2900

:

3100

;

=

:

3300

50

:

:

3500

4

:

MIZ

MIZ

9

i

m

IS

50

23b0

'

2500

'

2

20000

MIZ

60000

100000

140000

I80000

MIZ

Figure 2. ESI-MS (left) and MALDI-MS (right) of (a, top) CAMPATH 1H antibody and (b, bottom) deglycosylated CAMPATH 1H antibody. The inset shows expanded a portion of the ESI-MS for intact antibody.

combination of high molecular mass (-150 kDa) of the protein component with a high degree of glycosylation variability. Since particular glycoform patterns may determine the biological activity of these molecules, the bioanalyst is required to identify and monitor them. In order to simplify this analytical problem, our strategy has been to prepare the smaller glycoprotein fragments listed in F i e 1by specific methods in order to retain the original IgG glycosylation pattern and to subject these smaller moieties to mass spectrometric analysis. The corresponding polypeptide fragments listed in Figure 1were also prepared and mass analyzed in order to ascertain whether their measured molecular masses are consistent with the values calculated from the published amino acid sequences. 836 Analytical Chemistry, Vol. 67,No. 5, March 7, 7995

Until recently, physicochemical analysis of IgGs could only be performed by a combination of several methods. Thus, SDSPAGE would be employed to yield an approximate average Mr and ion exchange chromatography or isoelectric focusing techniques to ascertain the total number of glycoforms present in the molecule, Le., the number of sets of otherwise identical polypep tide chains with unique attached oligosaccharides. Finally, the exact composition of the individual glycans would be determined by physicochemical analysis of the carbohydrates released by chemical or enzymatic reaction. Overall, such detailed analyses are time consuming and hence are not performed routinely. However, the analytical situation has changed since 1989because mass spectrometers have become available for the rapid measure-

5i

47

I

2

MJZ Figure 3. ESI-MS of a mixture of fully (major species) and partially deglycosylated (Le. monoglycosylated)CAMPATH 1H antibody, with ion charge states for each shown.

knowledge, no paper reporting the combined use of the two ionization techniques has yet been published. The ESI-MS technique has been shown to yield (i) the M,of an intact IgG but no glycosylation pattem,11J4 (ii) the Mr of the nonglycosylated LC of an IgG produced by a collisionally activated decomposition (CAD) of one of its multiply charged molecular ions,I2 (iii) the M, of a chromatographically separated, nonglycosylated Fab component of a monoclonal antibody (MoAb)8, (iv) the M,'s of genetically engineered Fab and F(ab?z fragments of a humanized MoAb.I5 This ionization technique has also been employed to measure the M,'s (25 kDa) of the various glycoforms of differing molecular masses of the reduced and alkylated monomeric Fc components of two MoAbs,'O confirming that the ESI-MS technique should give distinctive glycosylation patterns. This is also supported by the recent report of a study of the humanized monoclonal antibody (H~u-anti-TAC)'~ in which the authors o b tained a glycosylation pattern from ESI-MS measurements on the reduced heavy chain H-COD. This pattern involved three components of differing masses, two of which were assigned to specific glycoprotein structures on the basis of the coincidences of measured and calculated M,'s. The single MALDI-MS paperg contains a detailed account of the application of this technique to measure the chemical average M,'s of seven intact MoAbs and one aglycosylated MoAb, together with the measurement of the M,'s of their reduced light chains and chemical average M,'s of the glycosylated heavy chains. No glycosylation patterns were observed for either the intact MoAbs or the component heavy chains. However, by utilizing the DNA sequence data to calculate the M,'s of the polypeptide chains, the authors were able to estimate the average total carbohydrate mass for four of the MoAbs and, in one case, were able to confirm the MS result by comparing it to a previously determined chemical carbohydrate analysis. In this paper we present and discuss the results of our mass spectrometric study of the intact IgG CAMPATH 1H and of its various glycoprotein and protein components obtained by chemical and by enzymatic procedures. The use of both the ESI and MALDI techniques was aimed at assessing their applicability to the problem of analyzing this important class of biomolecule.

ment of molecular masses of large molecules (Mr 5 200 kDa) that incorporate the new soft ionization techniques of electrospray (ESI) and matrix-assisted laser desorption ionization (MALDI). As we have discussed elsewhere,'jESI-MS can measure the M,'s of purified, desalted proteins to an accuracy of 0.01%and better. MALDI time-of-flight MS (MALDI-TOF-MS) can measure M,'s with a typical accuracy of 0.5%; with internal standards, the accuracy is -0.05% and 0.01%has been ~laimed.~ Theoretically, therefore, the ESI-MS technique should provide a rapid method of performing analysis of those glycoforms in an intact IgG possessing substantially different M,'s. Both techniques should detect glycosylation patterns for the glycosylated components Fc, Fc/2 and the reduced heavy chain H-C(ID, and the reduced and alkylated heavy chain H-CO provided the instrumental mass resolution is sufficient to separate the various glycoforms. EXPERIMENTAL SECTION To date we have located a total of eight papers in the Materials. Chemicals. DL-Cysteine, dithiothreitol @TI'), literature8-'5 in which these new mass spectrometric techniques ethylenediaminetetraacetic acid disodium salt (EDTA), formic acid, have been applied to the analysis of intact antibodies and of their iodoacetamide, AnalaR sodium hydroxide, disodium tetraborate, protein and glycoprotein components; seven of the p a p e r ~ ~ J ~ - ~ ~ acid, mercuripapain crystallized suspension in 70%ethanol, acetic involve ESI-MS and the remaining one MALDI-'lBF-MS.g To our pepsin lyophilized powder, horse heart myoglobin, bovine trypsi(6) Ashton, D. S.; Beddell, C. R; Green, B. N.; Oliver, R W. A FEBS Left, 1994,342, 1-6. (7) Beavis, R C.; Chait, B. T. Anal. Chem. 1990,62, 1836-1840. (8) Feng, R; Konishi, Y.; Bell, A. W. J. Am. SOC.Mass Spectrom. 1991,2, 387401. (9) Siegel, M. M.; Hollander, I. J.; Hamann P. R; James, J. P.; Hinman, L;Smith, B. J.; Farnsworth,A P. H.; Phipps, A,; King, D. J.; Karas, M.; Ingendoh, A; Hillenkamp, F. Anal. Chem. 1991,63, 2470-2481. (10) Jiskoot, W.; van de Werken, G.; Coco Martin, J. M.; Green, B. N.; Beuvery, E. C.; Crommelin, D. J. A. Phamac. Res. 1992,9, 945-951. (11) Feng, R; Konishi, Y. Anal. Chem. 1992,64, 2090-2095. (12) Feng, R; Konishi, Y. A n d . Chem. 1993,65,645-649. (13) Lewis, D. A; Guzzetta, A. W.; Hancock, W. S.; Costello, M. Anal. Chem. 1994,66, 585-595. (14) Verentchikov, A. N.; Ens, W.; Standing, K. G. Anal. Chem. 1994,66, 126133. (15) Bourell, J. H.; Clauser, K. P.; Kelley, R; Carter, P.; Stults, J. T. Anal. Chem. 1994,66, 2088-2095.

nogen, and bovine serum albumin @SA) and its dimer were from Sigma, Poole, Dorset, UK. A variant of bacterial subtilisin and a bacterial thiol oxidoreductase were kindly provided to VG Analytical by Dr. T. Halkier, Novo Nordisk, Denmark. Recombinant N-glycanase was obtained from Genzyme, Cambridge, UK. Immobilized papain gel was purchased from Pierce and Waniner, Chester, UK A solution of CAMPATH 1H antibody in phosphatebuffered saline at -16 mg/mL was obtained in-house. Acetonitrile super purity grade was from Romil Chemicals, hughborough, Leicestershire, UK. The Afti-Gel protein A MAPS I1 kit employed was purchased from Bic-Rad Laboratories, Heme1 Hempstead, Hertfordshire, UK. Ultrafree-MC polysulfone membrane filters, of appropriate molecular weight cutoff, were obtained from Millipore Corp., Watford, Hertfordshire, UK. Analytical Chemistry, Vol. 67,No. 5, March 1, 1995

837

50934

148942

a

h

I

50400

50800

148500

51200

Mr

149000

149500

Mr 529 15

26543

I

26382

26705

I

26000

26400

26800

Mr

52500

53000

53500

Mr

Flgure 4. Mass transforms from ESI mass spectra of CAMPATH 1H antibody and glycan-containing fragments: (a, top left) H-C(ll), monoglycosylated; (b, bottom left) Fc/2, monoglycosylated; (c, top right) CAMPATH 1H antibody, diglycosylated; (d, bottom right) Fc, diglycosylated.

Preparation of Protein and Glycoprotein Components of

CAMPATH 1H. (i) Preparation of Glycosylated Fragments: H-C(I); H-CUI); Fc and Fc/2. H-CO was prepared from CAMPATH 1H by reduction with DTT and alkylation with iodoacetamide following the published method.16 The reaction mixture containing the derivatized H-CO was separated from the L C O using gel filtration chromatography. H-C(II) (with light chain) was prepared by incubating CAMPATH 1H with DTT (0.1M) in Tris buffer for 2 h at 37 "C and then purified by ultratiltration. Fc (16) Crestfield,k M.; Moore, S.; Stein, W. 627.

H.J. Bid. Chem. 1963,238,622-

838 Analytical Chemistry, Vol. 67, No. 5, March 1, 7995

was formed, together with Fab, by classically subjecting a solution of the antibody to proteolytic cleavage using papain. The h a 1 purification of Fc (and Fab) was achieved by afsnity chromatography using the protein A column. Fc/2 was prepared from the purified solution of Fc by reduction with DTT for 2 h at 37 "C, and the product was purified by ultrafitration prior to analysis. (ii) Preparation of Deglycosylated Entities: CAMPATH lH, H-C(I), H - C W ; Fc and Fc/2. These substanceswere all prepared from the corresponding glycoproteins using the enzyme Nglycanase as follcws: Sufficienteme was added to a of the chosen glycoprotein in phosphate-buffered solution, con-

26558

4

26395

r 26000

1

.

*

27000

.

-

1

28000

-

.

.

-

~

29000

MIZ Figure 5. MALDI-MSof a glycan-containingfragment of CAMPATH 1H antibody (Fc/2, monoglycosylated).

centration 5-16 mg/mL, until a ratio of 20 pg of glycoprotein to 0.14 unit of enzyme was achieved. The resulting solution was then maintained at a temperature of 26 "C for 60 h in order to ensure that the enzymatic deglycosylation proceeded to completion, after which time the product was purified by ultrafiltration prior to analysis. Partial deglycosylation of CAMPATH 1H was effected by reducing the enzymatic digestion time to 15 h at 34 "C. (iii) Preparation of the NonglycosylutedEntities: LC(I); LC(II); Fab; Fab(H-C) and F(ab')z. The first two fragments, L C O and L C O ,were prepared and purified together with the corresponding H-C fragments as indicated in (i). The Fab fragment was prepared simultaneously with Fc (see (i)). The Fab(I4-C) fragment was prepared from Fab by reduction with DTT and then purified by ultraiiltration. The F(ab')z fragment was prepared from CAMPATH 1H by the action of pepsin. A solution of the antibody -10 mg/mL was prepared in 0.1 M sodium acetate and adjusted to pH 4.3 with acetic acid. To this solution was added pepsin to give an enzyme to substrate ratio of 1:lOO (w/w) and the digest incubated at 37 "C for 24 h. The reaction was then terminated by raising the pH to 9 with 2 M Tris-HC1 and 1 M NaOH. The solution containing the F(ab')z fragment was then purified by ultra!iltration prior to analysis. Purity of Samples. Solutions of all of the prepared samples gave the requisite number of UV-absorbing peaks when they were subjected to capillary zone electrophoresis prior to analysis. Further, all samples were purified by ultrafiltration to remove potentially interfering salts or buffers prior to ESI-MS analysis. Mass Spectrometry. (1) Electrospray Mass Spectrometty. The electrospray mass spectra were measured with either a VG BioQ or a VG Quattro I1 quadrupole mass spectrometer, both having an m/z range of 4000, equipped with an electrospray interface. Resolution of the mass spectrometers was 2000 at m/z 1000 (based on the width at half-height of a singly charged

isotopically resolved ion peak). At an observed mass of 1000 Da (calibration with myoglobin) a resolution of 665 was calculated. Samples (10 fiL) of the material under investigationwere injected directly into the electrospray source via a loop injector (Rheadyne 5717) after a 50%dilution with acetonitrile and the addition of formic acid to 1%.The concentration of these samples varied depending on the M,of the species under investigation (1-5 mg/ mL). The mobile phase used throughout was water/acetonitrile (5050 v/v) and formic acid (l%), at a flow rate of 3 pL/min, using a Pharmacia pump (Model 2248). The mass spectrometer was scanned over an m/z range appropriate for the material being examined and calibrated either with horse heart myoglobin or bovine trypsinogen. The source conditions were optimized for each sample; typical values were capillary potential 3.75 kV, counter electrode 900 V, and cone voltages 160 and 260 V. The source was maintained at a temperature of 65 "C with a bath gas flow of 120 L/h and a nebulizing gas flow of 80 L/h. (2) Matrix-Assisted Laser Dao@ion M a Spectrometty. The laser desorption spectra were all obtained using a VG TofSpec mass spectrometer. A 2 p L volume of the sample solution, concentration 10 pmol/pL in distilled water, was mixed with 2 pL of a saturated solution of sinapinic acid (SA) with or without appropriate internal standards @SA, BSA dimer, subtilisin, thiol oxidoreductase). The mixed sample, 2 yL, was then spotted onto a metal sample holder and allowed to dry before loading into the mass spectrometer. The sample was then irradiated using a 337 nm Nz laser at minimum laser power to produce ionization without fragmentation. Spectra were then acquired and averaged, normally from some 30 shots, until a representative trace was obtained. All data were then acquired in a linear mode, 0.65 m path length, and a source voltage of 25 kV. The MALDI-MS data were smoothed. Capillary Zone Electrophds. This high-resolution separation technique was employed to check the purity of all of the samples prior to mass spectrometric analysis. The CZE was performed using an Isco Model 3850 electropherograph. Aqueous solutions of the samples of -1 mg/mL concentration were manually injected, loading 4 nL via a split-teejunction device onto an uncoated fused silica column, 50 pm id., 80 cm long, with 50 cm effective length, and monitored at 200 nm. Separations were carried out using 100 mM disodium tetraborate buffer, pH 9.3, at 175 V/cm. RESULTS AND DISCUSSION Figure 2 presents the ESI and MALDI mass spectra of (a) CAMPATH 1H and (b) fully deglycosylated CAMPATH 1H in parallel as an aid to discussion. The previously published ESIMS spectra of intact IgGs by Feng and Konishill consisted of the charge state distribution 62+ to 72+, rather than the fuller series shown in Figure 2, due to the limited mass range of their instrument (2400 Da). Glycoforms were not reported, and in this mass range, the resolution of their instrument would probably have been insufficientto resolve any present. The complete ESIMS spectrum of an intact IgG published recently1*was obtained using a reflecting time-of-flight mass spectrometer. Comparison of the two ESI-MS spectra presented in Figure 2 shows that all of the molecular ion peaks for the glycosylated CAMPATH lH, Figure 2a, show considerable heterogeneity, which may at least in part be assigned to glycan heterogeneity. By contrast, the molecular ion peaks of the deglycosylated CAMPATH lH, Figure Analytical Chemistty, Vol. 67, No. 5,March 1, 1995

839

Table 1. Summary of Mass Spectrometric Results.

(1) ESI-MS Results for Glycosylated and Deglycosylated Species

+

(1 3; 2+2)e

no.6

lA lA lA

glycosylated deglycosylated carbohydrate polypeptide(seq) carbohydrate

1B 1C

CAMPATH lH(0) (1 4; (2 4; 2+3)' 3+3)e

148 668' 145 766' 2 904 145 764 2 904

CAMPATH 1H (ID

Fc

+

+

(1

+ 3, 2 + 2)e

(1 4; 2 + 3)'

+

(2 4; 3 + 3)e

(1 + 3,

2

+ 2)e

(1 + 4; 2 + 3)e

(2 + 4; 3 3)e

148 810 145 766f 3 046 145 764 3 046

148 942 145 766' 3 178 145 764 3 178

148 669 145 772 2 899 145764 2 905

148 819 145 772 3 049 145764 3 055

148 990 145 772 3 220 145764 3 226

52 765' 49 868' 2 899 49 865 2 900

52915 49868' 3 049 49 865 3 050

53086 49868' 3 220 49865 3 221

+

+

no.b

lA lA

glycosylated deglycosylated carbohydrate polypeptide(sea) carbohydrate

1A 1B 1C

no.b 2A 2A 2A 2B 2C

glycan no. glycosylated deglycosylated carbohydrate polypeptide (sed carbohydrate

50 618 49 329 1290 49 327 1291

50 778 49 329 1450 49 327 1451

50 934 49 329 1 606 49 327 1607

51 108 49 329 1780 49 327 1781

26 231 24 940 1292 24 938 1293

26 382 24 940 1443 24 938 1444

26 543 24 940 1 604 24 938 1 605

(2) MALDI-TOFdResults for Glycosylated and Deglycosylated Species CAMPATH CAMPATH CAMPTH IH(I1) H-C(I) H-C(1I) FC IH (0) IH 0 nf nf nf nP nf nf (2)e 148 365(3+) 145 708(2+) 2 659 145 764 2 601

speciesC

MJsed

F(ab'h Fab Fab (H-C) LC (1) L-c (ID LC (11)

97 950 47 968 24 407 23 857 23 571 23 571

148 570 145 698 2 874 145764 2 806

148 706 145 628 3 080 145764 2942

51 430(1+) 49 994(1+) 1437 49954 1476

50 802* 52 904(1+) 49 366* 49 826(1+) 1437 3 080 49327 49 865 1475 3 039

Fc/2 (3)e

+14 f 2

0 0 +1 -1

0.014 0.004 0 0 0.004 0.004

(4)e

26 394(1+) 26 557(1+) 26 729(1+) 24 930(l+)g 24 930(1+)g 24 930(1+)g 1465 1 628 1 800 24 938 24 938 24 938 1456 1619 1791

(3) ESI and MALDI-TOFd Results for Nonglycosylated Species Mr(ESI) AM % MI (MALDI) 97 964 47 970 24 407 23 857 23 572* 23 570

26 705 24 940 1 766 24 938 1 767

97 466(1+) 47 919(1+) 24 424(1+) 23 700(1+) 23 499(1+)g nm

AM

96

-484 -49 +17 -157 -72

0.49 0.10 0.07 0.66 0.31 nc

+

a Key: CAMPATH 1H (O), direct measurement of M,of CAMPATH 1H; CAMPATH 1H (I), indirect measurement of M,= BM,[LC(II)] 2Mr[H-C(II)1- 32 Da; CAMPATH 1H (II), indirect measurement of M,= 2MJFabI + M,[Fcl - 36 Da; H-C(I), measured M,of reduced and iodoacetamide alkylated heavy chain; H-C(II), corresponding measured Mr or calculated MI* of reduced heavy chain where M,*= M,[H-C(I)] 627.57 Da (11-SH groups alkylated); L C 0 , measured M,of reduced and iodoacetamide alkylated light chain; LC(II), corresponding measured M, or calculated M,*of reduced light chain, where M,*= M,[LC(I)] - 285.25 Da (5SH groups alkylated); Fab(H-C), H-C part of Fab, namely, PyroGlu(1) His(228). (A) carbohydrate, M,(glycosylated) - M,(deglycosylated) n; (C) carbohydrate, MJglycosylated) - M,(polypeptide(sed); nm, not measured; nc, not calculable; nr, not resolved; AM = M,(meas) - MI(sef; MI(seq), molecular mass calculated from amino acid sequence; n, number of Asn to Asp conversions upon deglycosylation. * (A) Direct and in irect measurement of the M,'s (Da) of native and deglycosylated CAMPATH 1H and of its various components; (B) calculated M,'s of the corresponding polypeptides; (C) derived M,of the carbohydrate($. Calculated and measured M,'s and corresponding M,discrepancies (AM) of polypeptide components. Listed M,values as Mr(z+), where z is the charge state of ion employed to calculate M,.e Glycan number defined in Table 2. f Resolved major component of lowest mass. 8 The component adducts with sinapinic acid.

-

2b, show reduced heterogeneity and that which is present can be attributed to the presence of adducts. The ESI-MS spectrum presented in Figure 3 of the CAMPATH 1H antibody subjected to limited deglycosylation shows two distinct series of multiply charged molecular ion peaks, the most abundant of which is assigned to the deglycosylated antibody, and the minor species is assigned to a semideglycosylated moiety. The next set of graphical data to be presented and discussed is that obtained from CAMPATH 1H and three of the four glycosylated fragments prepared from it, namely, Fc, H-CUI), and Fc/2. Figure 4 shows the transformed ESI-MS data obtained for these substances, and Figure 5 shows the singly charged molecular ion of Fc/2 obtained by MALDI-MS. The triplet structure of the latter, with the ion of maximum abundance at m/z 26 558 and two adjacent partially resolved ion peaks of lower abundance d ~ e r i n gin mean m/z value by k167.5, may be 840 Analytical Chemistry, Vol. 67, No. 5, March 1, 1995

+

*.

Table s ~ ~ c ~ and u rResidue ~ s Yasses of Some l,,,munoglobulin ~l~~~~~

no. 1

2 3 4

chemical structure

calc residue M, (Da)

(Man)3 (G1cNac)d (Man)3(GlcNac)4(Fuc)1 (Gal)1(Man)3 (GlcNAc)4 (Fuc)1

1299.21 1445.35 1607.49 1769.63

(Gal)z(Man)3(GlcNAc)4(Fuc)l

assigned to the presence in the samples of three glycoprotein molecular species differing from the adjacent by f l hexose residue. It may thus be concluded that MALDI spectra of glycoproteins with M, 26 kDa can provide a monitor for their glycosylation patterns whereas MALDI spectra of the other three glycoproteins of higher M, (150 kDa) do not since the spectra obtained (not shown) consist of broad structureless molecular ion

Table 3. List of Predicted and Measured M,'s for the Diglycosylated Species Mr (Da) Fc glycans

calc

1+3 2+2 1+4 2 f 3 2+4 3+3

52 771 52 755 52 933 52 917 53 080 53 080

145000

CAMPATH 1H

calc

meas 52 765 52 915 53 086

148 671 148 655 148 833 148 817 148 979 148 979

146000

meas 148 668 148 810 148 942

147000

Mr Figure 6. Mass transform of ESI-MS for deglycosylatedCAMPATH 1H antibody. The expected mass corresponds to the major component of lowest mass (labeled); adduct species at higher mass values are evident. Table 4. Summary of Atr% Measured by MALDI Techniques and Calculated by Peak Height-Weighted Averaging from Glycan Assigned ESI Atr%

species CAMPATH 1H [ O ] CAMPATH 1H [I11 Fc H-C[III

Mr (Da) MALDI (meas) 148 365 148 706 52 904 50 802

ESI (av) 148 827 148 830 52 926 50 868

peaks only. This conclusion is consistent with the observations reported by othersg and is a direct result of the limited resolution of the TOF in~trumentation.~~ By contrast, all of the mass transforms of the ESI mass spectra for the glycosylated species shown in Figure 4 show well-resolved features and these will now be discussed in detail. A comparative study of parts a and b of Figure 4 for the two monoglycosylated fragments H-CUI) and Fc/2 (17) Overberg, A; Hassenbuerger, A; Hillenkamp, F. In Mass Spectromety in the Biological Sciences:A Tutorial; Gross, M. L., Ed.: Kluwer: Dordrecht, The Netherlands, 1992; pp 181-197.

reveals corresponding patterns composed of four well-resolved glycoforms of differing molecular mass. Further, plausible assignments as to the overall chemical composition of the glycans forming these glycoforms may be made by calculating the glycan residue masses, by subtracting the M, of the nonglycosylated species from that of the corresponding glycosylated species, and by comparing them with the residue masses of glycans of known structure. The results of these subtractions are listed in the overall numerical summary in Table 1, (A) and (C) for the ESI, and section 2, (A) and (C) for the MALDI measurements. A study of these sections of Table 1reveals that the following glycan residue masses (with ranges) were obtained for the four most abundant glycoforms from the more accurate ESI data for these two fragments: glycan (l), 1292 f 2 Da; glycan (2) 1447 f 4 Da; glycan (3) 1606 f 2 Da; glycan (4) 1774 f 8 Da. A literature study of the reported immunoglobulinglycans led to the four structures listed in Table 2, whose calculated residue masses coincide, within reasonable limits, with those found, a finding that accounts for glycan assignments given in the last two columns of Table 1, section 1. These 4 are among 12 glycan components detected, after enzymic release from CAMPATH 1H and derivatization, by HPLC and IvWL,DI-TOF-MS.~* Considerationof the transformed ESI-MS of the diglycosylated species CAMPATH 1H and Fc shown in Figure 4c,d reveals a more complicated glycosylation pattern than that found for the monoglycosylated fragments. Such a complex pattern would be expected since there are 10 possible pairwise combinations of 4 glycans. Simple calculation of these combinations of the Mr of the above residue masses of the assigned glycans leads to the prediction that seven of them would probably be resolvable at an Mr of -50 kDa (Fc) and -150 kDa (CAMPATH 1H). Further, if the observed rehtive abundances of the four resolved glycans (Figure 4a,b) are taken into consideration, then it is predicted that the diglycosylated species should consist of three major M, peaks, each composed of two glycan combinations, and these are listed, together with the major M,'s measured for Fc (see Figure 4d) and numerically labeled M,'s for CAMPATH 1H (see Figure 4c), in Table 3. A comparative study of columns 2 and 3 in Table 3 shows close agreement between this set of Mr values for Fc, (AMr),, 18 Da. There is also good agreement for CAMPATH 1H (AMr),, 37 Da; see Table 3, columns 4 and 5. However, it will be noticed from Figure 4c that the labeled peaks to which glycans have been assigned do not fully represent the major peaks present. Our measurements on the deglycosylated CAMPATH 1H by ESI-MS (see Figure 6) clearly indicated the presence of adducts of unknown origin. The peak mass value is -100 Da higher in Figure 6 than the nonadducted component mass labeled; another sample (data not shown) showed even more adducting, with the peak higher by -175 Da than the nonadducted mass. If a comparable degree of adduction occurs in CAMPATH lH, the glycan assignments for CAMPATH 1H in Table 3 will be incorrect. If it were assumed that the peaks seen in Figure 4c are, for example, 140 Da higher in mass than the nonadducted species they represent, adducted glycoform assignment becomes glycans (1 3, 2 2), 148 810 Da; glycans (1 4, 2 3), 148 942 Da; glycans (2 4,3 3), 149 115 Da (see peak A in Figure 4c) and the major features in the mass transform are better accounted

+

+ +

+

+

+

(18) Ashton, D. S.; Beddell, C. R: Cooper, D. J.; Lines, A C. Anal. Chim.Acta, in press.

Analytical Chemistry, Vol. 67, No. 5, March 1, 1995

841

for. Additionally the value of 31 Da, is comparable. These values of the maximum error in the measurement of the Mt of the major glycoforms of the intact CAMPATH 1H are not in excess of those reported by others. At the recent meeting of the ASMSI'~the ESI-MS measurement for another humanized monoclonal antibody was given as M,(meas) 149 265 f 100 Da. These workers also commented on the difficulty they had experienced in assigning the correct charge states to the broad heterogeneous ion peaks observed in the raw ESI-MS (Figure 2a, inset) and that they therefore used manual peak centroid assignments. In this study, a similar process was used initially in conjunction with the computerized normal transformation algorithm. They also commented that part of the ion peak broadening observed in the intact ESI-MS was due to gas-phase adduction. In order to facilitate comparison of the M, values for glycosylated components obtained by MALDI and ESI-MS, where glycoforms are resolved only in the latter, the peak height-weighted average for the glycofonns has been calculated (Table 4). A study of the corresponding values given in Table 4 shows that they lie within or near 0.1%of one another provided they are based on component M,'s below or near 50 kDa. However, at -150 kDa (CAMPATH 1H [O]), the discrepancy is over 0.3%. Consideration will now be given to the last set of mass spectrometric measurements presented in this study, namely, those obtained for the nonglycosylated species listed in section 3 of Table 1. The maximum M, discrepancy (AM) from ESI-MS is found for the largest component species, F(ab'h, but even this value of 14 Da is within 0.02% of the expected value. These measurements provide good supporting evidence for the correctness of the listed (Figure 1) molecular structures of the various components and, by implication,for the validity of the published4 amino acid sequences of the light chain and of the heavy chain of CAMPATH 1H if it is assumed that the N-terminal residue of the heavy chain is pyroglutamic acid and the C-terminalresidue is Gly 450 (Figure 1). The M, discrepancyvalues (Ah!) calculated from the MALDI mass spectrometric measurements listed in (19) Johnson, W. P.; Roberts, G. D.; Burman, S.; Cam,S. A. Presented to the American Society for Mass Spectrometry, 1 June 1994; Abstr 78, p 669.

842 Analytical Chemisty, Vol. 67,No. 5, March 1, 1995

section 3 of Table 1 are much larger than those calculated from the ESI results. The Ahf magnitudes range from 0.07% to 0.7%. Both mass spectrometric techniques may therefore be employed to determine the M, of these protein fragments of CAMPATH lH, but generally the ESI technique is to be preferred over the MALDI-TOF method since there is no possibility of adducting with the acid matrices employed in the latter process and the accuracy of the h a 1 M, value obtained is higher. To conclude,this study has shown that MALDI-TOF-MSyields average M, values of the glycofonns present in CAMPATH 1H and its glycosylated fragments, with the exception of the Fc/2 fragment, for which three glycoform peaks were seen, partially resolved. The ESI-MS of the smaller monoglycosylated species can be interpreted in terms of distinct glycan structures, which in turn facilitates the assignment of the diglycosylated species where resolved. Support for the proposed amino acid sequence is provided primarily by the close correspondence between calculated masses and measured masses for deglycosylated and aglycosylated species, especially those at masses of -50 kDa and below, but also secondarily from the peptide masses deduced from individual glycoform masses after the glycan has been assigned. The principal problem of direct accurate analysis at high mass in this study appears to be extensive adducting of unknown origin. This problem is not entirely confined to high mass-for example it was observed for deglycosylated Fc, but at high mass it confuses glycoform assignment. ACKNOWLEDGMENT We thank Professor R Jefferis, Mr. M. M. Payne, and Mr. P. L. Francis for helpful discussions and Mr. G. Critchley, Dr. C. Porter, and Dr. A. Maisey of Fisons Instruments, VG Analytical, and Dr. B. N. Green of Fisons Instruments, VG Biotech, for technical assistance. Received for review July 19, 1994. Accepted October 27, 1994.8

AC940721E @

Abstract published in Advance ACS Abstracts, January 1, 1995.