Anal Bioanal Chem (2011) 400:2427–2438 DOI 10.1007/s00216-011-4923-5
ORIGINAL PAPER
Analysis of recombinant human follicle-stimulating hormone (FSH) by mass spectrometric approaches Josephine Grass & Martin Pabst & Martina Chang & Manfred Wozny & Friedrich Altmann
Received: 29 October 2010 / Revised: 15 March 2011 / Accepted: 16 March 2011 / Published online: 3 April 2011 # Springer-Verlag 2011
Abstract Recombinant human follicle stimulating hormone is an important drug in reproductive medicine. Thorough analysis of the heterodimeric heavily glycosylated protein is a prerequisite for the evaluation of production batches as well as for the determination of “essential similarity” of new biosimilars. The concerted application of different liquid chromatography-mass spectrometry methods enabled the complete depiction of the primary structure of this pituitary hormone. Sequence coverage of 100% for the α- as well as the β-chain was achieved with tryptic peptides. Most of these peptides could be verified by tandem mass spectrometry. Sitespecific analysis of all four glycosylation sites was, however, not possible with tryptic but with chymotryptic peptides. Quantification of the glycoforms of each glycopeptide was accomplished with the software MassMap®. Both protein Published in the special issue Analytical Sciences in Austria with Guest Editors G. Allmaier, W. Buchberger, and K. Francesconi. Electronic supplementary material The online version of this article (doi:10.1007/s00216-011-4923-5) contains supplementary material, which is available to authorized users. J. Grass : M. Pabst : F. Altmann (*) Department of Chemistry, University of Natural Resources and Life Sciences (BOKU), Muthgasse 18, 1190 Vienna, Austria e-mail:
[email protected] M. Chang Polymun Scientific, 1190 Vienna, Austria M. Wozny MassMap GmbH & Co. KG, 82515 Wolfratshausen, Germany
subunits gave interpretable mass spectra upon S-alkylation and separation on a C5 reversed-phase column. Glycan isomer patterns were depicted by separation on porous graphitic carbon, using mass spectrometric detection for the evaluation of the glycopeptide liquid chromatography-electrospray ionization data. The currently marketed product Gonal-f™ and a potential biosimilar were compared with the help of these procedures. Keywords Follicle-stimulating hormone . Follitropin . Glycoprotein . Biosimilar
Introduction Follicle-stimulating hormone (FSH), also known as follitropin, exerts important roles in the reproductive organs of females and males. Recombinant human FSH (r-hFSH) is used for hormonal therapy in order to stimulate ovulation— mostly in the context of in vitro fertilization [1]. R-hFSH is also used to treat male patients with certain defects of spermatogenesis [2]. FSH consists of two subunits, α and β, which are both glycosylated [3]. Each subunit contains two glycosylation sites carrying Asn-linked glycans of the complex type with sialic acids and, in the natural FSH, also sulfates [4–7]. The same alpha-subunit is present in all gonadotropic hormones, i.e., FSH, luteinizing hormone, and human chorionic gonadotropin. These hormones differ only in their beta chains [8]. The currently approved and marketed product is Gonal-f™ (Follitropin alpha, Merck Serono S.A.). This drug is produced in Chinese hamster ovary (CHO) cells. The advent of biosimilars calls for reproducible in-depth analysis of this highly heterogeneous
2428
glycoprotein. In a pioneering work on the use of mass spectrometry for the characterization of FSH and related hormones, FSH peptides and glycans were analyzed by fast atom bombardment mass spectrometry, and spectra of the intact glycosylated alpha-chain were obtained by electrospray ionization mass spectrometry (ESI-MS) [9]. While no tandem MS was used for peptide identification in this early work, spectra assignments were confirmed by measuring the mixture before and after carboxyl esterification, thus determining the number of acidic residues in the peptides generated by Glu-C/CNBr double digestion. Thus, complete sequence coverage was obtained for the α-chain [9]. In order to conduct site-specific glycosylation analysis, Dalpathado et al. subjected proteinase K-derived glycopeptides from natural human and equine FSH to high-resolution ESI-MS [7]. A mixture of sialylated and sulphated complex-type oligosaccharides was found in human as well as equine FSH. One glycosylation site of the human α-chain could not be identified by this approach. For a comparison of the glycosylation status of several recombinant pituitary hormones, Gervais and coworkers applied high-pH anion exchange chromatography and reversed-phase high-performance liquid chromatography (HPLC) of 2-aminobenzamide labeled glycans for separation, as well as ESI and matrix-assisted laser desorption/ionization/ ionization–time-of-flight mass spectrometry (MALDI-TOF MS) for peak identification [10]. MALDI-TOF MS spectra were generated from both α- and β-chains of various pituitary hormones, but the resolution of MALDI-TOF was too low to resolve glycoforms [11], even though in another report, the α-chain gave several distinct signals [10]. The present report applies contemporary strategies of glycoprotein analysis to obtain comprehensive data on the protein and the carbohydrate portions of r-hFSH. The essential core of data was obtained by capillary liquid chromatographyelectrospray ionization-tandem mass spectrometry (LC-ESIMS/MS) of chymotryptic peptides/glycopeptides and LCESI-MS/MS of deglycosylated tryptic peptides. ESI-MS spectra of both subunits in glycosylated and deglycosylated states were acquired. Finally, the isomer signature of FSH glycans was determined by carbon-LC-ESI-MS.
Materials and methods Material and general pretreatment Three lots of the reference product Gonal-f were subjected to buffer exchange in 50 mM ammonium acetate pH 7.0 and concentration using centrifugal filters with a cut-off of 3 kDa (Millipore, Vienna). For comparison, a potential biosimilar FSH was produced in a stable CHO cell line (established by co-transfection with cDNAs coding for the
J. Grass et al. Fig. 1 ESI-MS of intact subunits of r-hFSH. S-alkylated FSH subunits were separated by RP-HPLC. Panels A, B depict the RP-HPLC separation of glycosylated and deglycosylated r-hFSH (Gonal-f), respectively. In C, D, the deconvoluted spectra of deglycosylated αand β-chains are shown. Panels E–H are the deconvoluted spectra of the glycosylated α- and β-subunits of Gonal-f (E, G) and the biosimilar (F, H). Spectrum interpretation was based on the average mass of Scarbamidomethylated subunits. Addition of two di-antennary, disialylated N-glycans to the mass of the α-chain (10,776 Da) results in a mass of the M+H+ ion of 15,189 Da, which matches well with the major peak observed in panels E, F. Attachment of one di-antennary and one tri-antennary N-glycan, both fully sialylated, to the S-alkylated β-chain (13,170 Da) gives a mass of 18,531 Da, which corresponds to a major glycoform as seen in G, H. Sugar symbols are as suggested in www.functionalglycomics.org
FSH subunits) using a continuous-perfusion fermentation process. This biosimilar FSH was treated similarly and adjusted to the same protein concentration. The proteins were desalted by passage over C4 solid-phase extraction cartridge (Supelclean LC-4, wide pore SPE, 500 mg, Sigma-Aldrich, Vienna) using 60% acetonitrile in 0.1% trifluoroacetic acid for elution. After freeze-drying, the samples were S-alkylated with iodoacetamide in the presence of 15 mM dithiothreitol and 8 M urea [12]. Protein was subsequently precipitated with a fivefold volume of cold acetone. After 30 min, the precipitate was collected, and the samples were re-dissolved in a small volume of water. Protein mass spectrometry The protein mass of S-alkylated r-hFSH was determined both for glycosylated and deglycosylated forms. To this end, 4 μg of protein was digested with 0.4 U of PNGase F (Roche, Mannheim) in 20 μL 50 mM ammonium acetate buffer of pH 8.4 at 37 °C overnight. One microgram of protein was injected into a wide-pore C5 column (Discovery BIO wide pore C5, 50×0.32 mm, 3 μm particles; Supelco at Sigma-Aldrich, Vienna). The aqueous solvent was 0.1% formic acid in water, and a gradient from 10% to 50 % acetonitrile was developed over 20 min at a flow rate of 4 μL/min. Detection was performed with a Q-TOF mass spectrometer (Waters Micromass Q-TOF Ultima Global) in the positive-ion mode. The MS profile was set to 1450, 1650, and 1850 with standard weighting and the RF lens 1 to 100. The instrument was calibrated with caesium iodide. Data were deconvoluted using the MaxEnt1 function of MassLynx (Waters). Glycopeptide and peptide analysis Five micrograms of each S-alkylated and desalted r-hFSH sample was dissolved in 20 μL of 25 mM ammonium bicarbonate and digested overnight at 37 °C with either 0.15 μg of chymotrypsin (α-chymotrypsin, bovine pancre-
Analysis of r-hFSH by mass spectrometric approaches
BPI
19.26
B
* 22.70
21.61
* * 14
16
18
20
22
24
14
18
20
22
24
13167
D
13183
10790
10606
16
10774
C
Intensity
18.95
A
2429
13118
10805
10716
13198
13086
10600 10700 10800 10900
13000 13100 13200 13300
Mass
Time (min)
alpha-chain 15183
E 14892 15257
15839
15548
14601
16205
14437 14400
14600
14800
15000
15200
15600
15800
16000
16200
16400
15187
14896
100
15400
16497
F 15553
15262
%
15203
1491 1
14605
15844 15918 15860
15568
15276
14970
15627
14620
14441
16209 16226 1628416385 16502
0 14400
14600
14800
15000
15200
15400
15600
15800
16000
16200
16400
betachain 18525
G 1827818379
18234
18644
18200
18400
18600
%
17873
18239 18385 18284
18093
6A4S1F
H
18800
18604
19000
18896
18648 18546 18749
18664
18457
19547
19838
19036
18911
19200
19187
19400
19261
19042 18969 19114
7A6S1F
18000
18531
100
18934
18088
6A5S1F
17800
17987
6A6S1F
18744
6A4S2F
17868
19181
6A5S1F 18890
7A5S2F
18567
7A5S1F
18394
18219 18187
6A6S1F
18293
5A4S1F
19600
19800
7A6S2F 19552 19917 19843
19405
19202 19303
18838
17992
0 17800
18000
18200
18400
18600
18800
19000
19200
19400
19600
19800
Mass
2430
as, sequencing grade, Sigma-Aldrich) or 0.04 μg of trypsin (sequencing grade, Roche). Aliquots of each digest were loaded on a BioBasic C18 column (BioBasic-18, 150× 0.32 mm, 5 μm, Thermo Scientific) using 65 mM ammonium formate of pH 3.0 as the aqueous solvent. A gradient from 10% to 55 % acetonitrile was developed over 20 min at a flow rate of 6 μl/min. Detection was performed with a Q-TOF mass spectrometer in the positive-ion, plain MS mode. The MS profile was set to 450, 750, and 950. For in-depth sequence analysis, a sample of the tryptic peptides was dried and redissolved in 50 μL 0.1 M citratephosphate buffer of pH 5.0 with peptide: N-glycosidase A (Proglycan, Vienna). The samples were first analyzed using a “data-dependent acquisition” mode. Peptides that had not yielded useful fragment spectra in this mode were dealt with in a second LC-ESI-MS/MS run with preselected parent ions and optimized fragmentation energies.
J. Grass et al.
column (100×0.32 mm, Thermo Scientific, Vienna) with 65 mM ammonium formate buffer as the aqueous solvent on a Q-TOF Global Ultima (Micromass) using ESI. A very steep gradient from 2% to 42% acetonitrile was developed over 20 min at a flow rate of 8 μL/min. For selected samples, the standard gradient for isomer separation was applied as described [13, 15].
Results The reference r-hFSH was obtained as a ready-to-use formulation containing various additives that interfered with the subsequent analysis. In order to analyze the reference and biosimilar in parallel, all r-hFSH samples were purified, desalted, and S-alkylated before MS analysis. Mass spectrometry of intact subunits
Glycopeptide quantification Quantification of peaks in the full-scan MS data was performed by means of the expert software MassMap®, version 2010-06-13 (MassMap GmbH & Co. KG, Wolfratshausen, Germany). At first, low-intensity noise was removed to increase the speed of processing. The parameters employed for identification of the glycopeptides and their adducts were chosen such that no real peak was missed. The noise-corrected data sets were subjected to so-called pepmap-analyses. Then, the data were automatically tested for the presence of the molecules of interest. For the identified and accepted molecules, the software automatically generated the following numbers: location of the chromatographic peak and deviation of the molecular mass from the given value. In questionable cases, peak identification was evaluated by the user. Last but not least, an integral of the peak volume including the relevant peaks of the isotope pattern (i.e., the first five peaks in the present case) of the diverse charged ions (two to four) and of alternative ion species, i.e., ammonium adducts for Gp3 and Gp4 was calculated. The MassMap settings and results were counterchecked by “manual” processing with MassLynx. Here, we measured the intensity of the base peak of the summed spectrum covering the entire analyte peak in the extracted ion chromatogram. Oligosaccharide analysis Glycan analysis was performed as described [13]. Briefly, 2 μg of each sample was digested with 0.2 U of PNGase F overnight at 37 °C in 100 μL 50 mM ammonium acetate of pH 8.4. Glycans were subsequently reduced in 200 μL 1% sodium borohydride at room temperature for 4 h. Excess salt was removed using a 10-mg HyperSep Hypercarb SPE cartridge (Thermo Scientific, Vienna) [14]. The glycans were then analyzed by LC-ESI-MS on a porous graphitic carbon
FSH is dissociated by the acidic conditions prevailing during LC-ESI-MS analysis [16]. Attempts to measure r-hFSH (with or without reducing agent) without permanent derivatization were unsuccessful. However, when S-carbamidomethylated r-hFSH was subjected to LC-ESI-MS, two peaks with high mass ions appeared (Fig. 1). The first larger peak comprised the α-chain. The second peak contained the β-chain. Due to the noisy spectrum, this second peak was difficult to deconvolute. Nevertheless, the major glycoforms of the βchain could be recognized as doubly fucosylated and highly sialylated (Fig. 1). Interestingly, the order of elution of αand β-chains was exactly opposite to that reported in a previous work that used neutral phosphate buffer as the eluent [11]. The two N-glycans on the β-chain contained two to four antennae. Considering the N-glycan structures of CHO-cell produced erythropoietin [17], “tetra-antennary” N-glycans may in fact be tri-antennary glycans with lactosaminerepeats. The bottom-up analyses of glycopeptides and oligosaccharides indeed corroborated this assumption (see below). The Gonal-f β-chain appeared to harbour a series of peaks 247 Da smaller than the readily interpretable major signals. However, the spectrum of the deglycosylated drug protein (Fig. 1) showed no sign of satellite compounds, nor were these apparent in the glycopeptides from glycan analyses. Another probably misleading detail is the presence of peaks 16 Da larger than the major compounds. Their considerable height would point at a high degree of methionine oxidation. FSH contains two Met residues in the α- and one in the β-chain, which did not appear to be oxidized to a degree explaining this picture (see chapter “Peptides”). A plausible explanation for this discrepancy could be in-source oxidation during intact protein analysis (Michael Graninger, Vienna, personal communication).
Analysis of r-hFSH by mass spectrometric approaches
2431
Gp1 (alpha-chain),
4165.3
VQKNVTSESTCCVAKSY
3874.2
4530.3 4821.5
3941.2 3730.2
3419.1
3000
3200
3570.1
3400
4181.3
3600
3800
4000
5164.6
5552.4
4306.3
3701.2
5253.5
4200
4400
4600
4800
5000
5200
5400
5600
5800
Gp2 (alpha-chain), 3279.0
KVENHTACHCSTCY
3971.1
3570.1
3680.0 3587.1
3279.2 2823.9
3296.0 3952.2 4243.2 4226.3 3325.9
2987.9
2647.9
4627.2
4336.2
3466.8 4045.06
4608.4
2600
2800
3000
3200
3400
3600
3800
4000
4200
4400
4800
5000
5200
5400
5A4SF
5A3SF
4A3SF
4600
Gp3 (beta-chain), NSCELTNITIAIEKEECRF
5698.8
5204.7
5042.7 4896.7
4677.6
4895.3
5187.2
5569.8
6080.2
6005.8 5860.8 5989.9 6007.1
5698.2
5424.8
5716.4
5789.7
5934.8
6354.9
5552.2
6226.2
6354.4 6446.1
7102.1
6736.0
6591.16720.4
5041.4
7110.9 6751.0
4751.6
4600
4800
5000
5200
5400
5600
5800
6000
7A4SF
6737.0
6224.8 6208.9
5934.3
5261.8
5A4S
4A4S
5350.8
5407.8
6A4SF
6371.4
5552.8 5187.7
6371.9
6080.8
5716.0
4A3S
5333.8
6200
6400
6600
6915.96
6800
7000
Gp4 (beta-chain), CISINTTW 2926.0
3362.1
2908.9
3345.0
2765.9
3053.9
3710.1
3071.0
3627.1
2628.9
3727.1
2611.8
4018.2
4001.2
4324.3
2200
2400
2600
2800
3000
3200
3400
3600
3800
4000
4200
4383.3
4400
Mass Fig. 2 Glycopeptides. The deconvoluted spectra of the four glycopeptides derived from the r-hFSH biosimilar are shown with cartoons of possible N-glycan structures. Of note is that the glycan isomers and
arrangements are arbitrary. The clubs sign labels ammonia adducts, which dominate in glycopeptides 3 and 4 even though all spectra shown here were generated in one and the same LC-ESI-MS experiment
(+)
16.3
11.6
12.6
17.7
6.6
20.4
20.6
20.7
11.3
11.4
36–44
45–51
46–51
46–63
64–67
68–75
68–92
68–91
76–92
MSMSa
15.9
10.0
23.9
18.9
20.4
9.8
21.9
21.9
20.8
50–54
50–62
55–62
63–86
87–97
98–110
98–111
98–111
15.5
22.5
21.6
23.4
13–17
18–29
19–29
19–33
(+)
+
++
+
++ MSMS
24
1–18
SQPGAPILQCMGCCF
SQPGAPILQCM
FSQPGAPILQCM
QENPF
APDVQDCPECTLQENPFF
APDVQDCPECTLQENPF
21.7
1–17
++
Chymotryptic peptides APDVQDCPECTL
GLGPSYCSFGEMoxKE
GLGPSYCSFGEMKE
GLGPSYCSFGEMK
CDSDSTDCTVR
VPGCAHHADSLYTYPVATQ CHCGK
ELVYETVR
TCTFKELVYETVR
TCTFK
DPARPKIQKTCTFK
DPARPK
DLVYKDPARPK
DLVYK
FCISINTTWCAGYCYTR
18.7
MSMS
MSMS
MSMS
MSMS
MSMS
MSMS
MSMS
MSMS
MSMS
MSMS
1–12
+
(+)
++
(+)
++
++
++
(+)
++
+
(+)
(+)
Alpha-chain
Part II, chymotryptic peptides
15.1
41–54
13.1
15.1
36–40
41–46
+ +
27.8
NSCELTNITIAIEKEECR
19–35
++
MSMSa
23.0
1–18
NSCELTNITIAIEKEECR
MSMS
NSCELTNITIAIEK
+
23.7
1–18
(+)
23.8
1–14
VENHTACHCSTCYYHK
VENHTACHCSTCYYHKS
VTVMGGFKVENHTACH CSTCYYHK
VTVMGGFKVENHTACH CSTCYYHKS
VTVMGGFK
Tryptic peptides
(+)
MSMS
++b
SYNR
TMLVQKNVTSESTCCVAK
MSMSa MSMS
TMLVQK
KTMLVQK
AYPTPLRSK
AYPTPLR
APDVQDCPECTLQENPF FSQPGAPILQCMGCCFSR
Tryptic peptides
MSMS
MSMS
MSMS
MSMS
+
+
++
++
(+)
++
++
Mode
Beta-chain
76–91
(+)
16.3
36–42
++
25.6
1–35
Intensity
Time (min)
Alpha-chain
Part I, tryptic peptides
Table 1 Peptides and glycopeptide found by LC-ESI-MS in the biosimilar sample
863.30
601.29
674.82
634.28
1083.96
1010.42
702.79
789.34
781.30
716.82
658.25
910.40
1008.54
823.40
656.30
845.50
683.38
651.36
637.36
1086.97
1297.52
1090.52
803.41
1068.66
1017.60
1200.80
1295.31
838.47
539.26
1420.93
719.41
847.51
1032.58
817.46
1029.94
Found
2
2
2
1
2
2
2
2
2
2
2
3
1
2
1
1
1
2
1
2
4
2
2
4
4
4
4
1
1
3
1
1
1
4
H+
GP4 non-glycc
GP3 3A3SF
GP3 non-glyc
GP3 non-glyc
GP2*** 2A2S
GP2** 2A1S
GP2* 2A2S
GP2 2A2S
GP1 2A2S
2432 J. Grass et al.
21.1
19.5
14.2
16
14.7
14.3
17.4
19
20.2
15.9
14.7
20–27
32–39
32–37
34–39
40–53
54–58
59–73
59–74
75–103
104–111
107–111
+
+
+
++
+
++
++
++
+
++
++
++
MSMS
GEMKE
CSFGEMKE
TYPVATQCHCGKCDSDS TDCTVRGLGPSY
ETVRVPGCAHHADSLY
ETVRVPGCAHHADSL
KELVY
KDPARPKIQKTCTF
TRDLVY
CYTRDL
CYTRDLVY
CISINTTW
CISINTTW
TNITIAIEKEECRF
NSCELTNITIAIEKEECRF
593.30
987.39
1098.13
906.43
824.90
651.37
845.46
766.41
827.37
1089.50
1115.70
994.47
862.44
1502.70
1164.05
622.25
1378.85
1324.50
719.38
1389.23
863.51
904.49
1
1
3
2
2
1
2
1
1
1
3
1
2
4
2
1
2
2
2
2
1
1
And 12.1
And at 13.6
Also 2+
GP4 3A3SF
GP4 non-glyc
GP3 3A3SF
GP3 non-glyc
GP2* 2A2S
GP2 2A2S
GP1 2A2S
c
b
a
Smeared over several minutes
Strong signal only as glycopeptide
MS/MS of deglycosylated peptide in separate run
MS spectra of tryptic and chymotryptic peptides. Sequences were taken from Uniprot (P01215 and P01225)
The elution time, approximate intensity from (+) to ++, and the actually observed major ion with its mass and charge state are given. Parts I and II list the identified peptides. Masses are theoretical masses of the M+H+ ions and the charge state in which they occurred in real samples. Random mass error was +/− 20 ppm. Parts III and IV show the sequence tags as deduced from fragment masses of MS/
NSCELTNITI AIEKEECRFC ISINTTWCAG YCYTRDLVYK DPARPKIQKT CTFKELVYET VRVPGCAHHA DSLYTYPVAT QCHCGKCDSD STDCTVRGLG PSYCSFGEMKE
APDVQDCPEC TLQENPFFSQ PGAPILQCMG CCFSRAYPTP LRSKKTMLVQ KNVTSESTCC VAKSYNRVTV MGGFKVENHT ACHCSTCYYH KS
Part IV, sequence tags obtained with tryptic plus chymotryptic peptides
NSCELTNITI AIEKEECRFC ISINTTWCAG YCYTRDLVYK DPARPKIQKT CTFKELVYET VRVPGCAHHA DSLYTYPVAT QCHCGKCDSD STDCTVRGLG PSYCSFGEMKE
APDVQDCPEC TLQENPFFSQ PGAPILQCMG CCFSRAYPTP LRSKKTMLVQ KNVTSESTCC VAKSYNRVTV MGGFKVENHT ACHCSTCYYH KS
Part III, sequence tags obtained with tryptic peptides
22.2
20–27
+
++
MSMS
22
6–19
MSMS
23.8
NSCELTNITIAIEKEECRF
1–19
KVENHTACHCSTCYY
KVENHTACHCSTCY
NSCEL
+
+
+
++
24.5
11.8
75–89
NRVTVM
VQKNVTSESTCCVAKSY
9
8.8
75–88
++
MSMS
RSKKTML
SRAYPTPL
1–19
12.7
66–71
++
MSMS
1–5
13.5
49–65
+
++
Chymotryptic peptides
9.7
42–48
Beta-chain
18.7
34–41
Analysis of r-hFSH by mass spectrometric approaches 2433
2434
J. Grass et al.
Table 2 Glycopeptide patterns of Gonal-f samples and the biosimilar Glycan
Y13B6742
Y14B3486
Y14B8282
biosimilar
Glycopeptide 1 Alpha-chain Asn52
Unglyc 2A0S 2A1S 2A2S 3A1S 3A2S 3A3S 4A2S
0.0 11.5 63.2 100.0 6.8 20.4 14.5 0.0
0.0 15.5 68.5 100.0 5.1 20.4 15.9 0.0
0.0 13.9 70.3 100.0 0.0 15.3 16.6 0.0
0.0 14.3 73.9 100.0 22.2 38.6 30.7 4.2
Glycopeptide 2 Alpha-chain Asn78
Unglyc 2A1S 2A2S 3A1S 3A2S 3A3S 4A2S Unglyc 3A1SF 3A2S 3A2SF 3A3S 3A3SF 4A1SF 4A2S 4A2SF 4A3S 4A3SF
0.0 44.4 100.0 4.9 12.0 0.0 0.0 7.5 3.9 9.8 33.0 36.0 100.0 8.2 4.4 15.2 15.3 41.9
0.0 46.7 100.0 4.7 9.6 5.1 0.0 6.3 2.9 9.7 31.5 35.1 100.0 4.8 5.1 15.0 13.7 40.4
0.0 45.7 100.0 4.4 11.5 6.0 0.0 4.5 3.6 8.4 33.6 33.1 100.0 3.9 4.2 12.1 11.6 35.5
0.0 56.7 100.0 15.5 28.6 18.0 6.2 19.7 3.5 15.2 32.3 38.2 96.8 10.1 21.2 38.4 50.5 99.9
4A4S 4A4SF 5A2S 5A2SF 5A3S 5A3SF 5A4S 5A4SF 6A4S 6A4SF 7A4SF Unglyc 2A1S 2A1SF 2A2SF 3A1S 3A1SF 3A2SF 3A3SF
23.0 70.7 0.0 1.8 2.0 19.5 8.4 19.7 0.0 0.0 0.0 15.8 100.0 28.2 92.8 2.0 1.2 5.9 17.1
22.0 66.7 0.0 5.7 3.7 10.7 7.3 17.8 0.0 0.0 0.0 15.8 100.0 48.9 85.6 2.0 0.0 6.8 20.7
21.5 71.4 0.0 5.2 3.2 15.4 6.5 16.2 0.0 0.0 0.0 15.8 100.0 40.6 87.7 1.8 0.0 6.6 21.2
48.0 100.0 10.1 28.5 33.4 86.0 40.8 97.2 18.2 59.7 12.5 15.3 100.0 23.2 73.8 1.9 1.2 10.4 22.1
Glycopeptide 3 Beta-chain Asn7
Glycopeptide 4 Beta-chain Asn24
A glycan structure 5A4SF consist of five Gal-GlcNAc units either in the form of antennae or of Gal-GlcNAc lactosamine repeats. Within each glycopeptide, quantities are expressed relative to the most abundant form S stands for sialic, F for fucose which essentially will be core α1,6-fucose
Analysis of r-hFSH by mass spectrometric approaches
2435
one cleavage version (Table 1). Both glycosylation sites of the α-chain were fully occupied, whereas Asn7 of the β-chain site appeared as the free peptide to 1.2% to 2% and Asn24 to 5.6% to 6.2 % (Table 2). In r-hFSH, three sites were occupied by mainly di-antennary N-glycans, whereas Gp3 from the β-chain carried larger Nglycans with lactosamine repeats, especially in the case of the biosimilar (Fig. 2).
Glycopeptides Analysis of the consistency of protein glycosylation at the glycopeptide level has the advantage of having available standard proteomic methods, which utilize the favourable ionization properties of the peptide moiety that outstrip the lower proton affinity of sugars. Thus, the glycopeptide spectra approximate the true glycosylation pattern at a given site. Only three glycopeptides of r-hFSH could be found in tryptic digests. The C-terminal glycopeptide of the β-chain remained undetected, even though its unglycosylated form could be found as smearing peak at the end of the chromatographic gradient. Fortunately, however, all four glycopeptides could be identified in chymotryptic digests (Fig. 2, Table 1). Some sites occurred in more than
Quantitation of glycoforms Glycopeptides are detected in several charge states depending on their size and composition. Within one glycopeptide, the ratio between, e.g., quadruply, triply, and doubly charged ions changes with increasing glycan size. Thus, a
2371.8
A 2225.9
1934.8
3028.1
2080.8
2882.0
2737.0
2590.9
4A3S 4A3SF
1643.6 1400
1600
1800
2000
2200
2800
3000
2225.8
2A1S
B
2600
2371.9
1934.8
2A1SF 3A3SF
3A2S
2080.8
2591.0
2737.0
2882.0
3028.1
1643.6
1400
1600
C
1800
2000
4A3SF 3391.1
5A3S
3410.1
4A3S 3247.0 3264.0
2200
4A4SF 3684.2 3701.23775.2 3538.1 3612.1 3758.2 3556.4 3629.1
6A3SF 5A4S
3700
3200
3400
4431.4* 4448.4
4140.3* 4285.3
3500
3000
6A4SF
4049.3
3484.03
3300
2800
5A4SF
3920.3
3337.2
2600
4066.3*
5A3SF
4A4S
3427.10
2400
3900
4100
4565.3
4347.1
4300
4759.5 4742.7
4500
4700
4813.7
4900
Mass
Fig. 3 Glycan spectra of Gonal-f™ (A) and the potential biosimilar (B). Reduced oligosaccharides were infused to an ESI-MS via a graphitic carbon column. The entire elution zone of oligosaccharides was summed and deconvoluted/deisotoped. Panel C shows an extension of the mass range of panel C to allow viewing the high
mass N-glycans. The cartoons show one of several possible isomeric structures. The masses of M + H+ ions are labelled, but with increasing size, glycans tend to form ammonia adducts, which are marked with black circles or, in case of the mass values, with an asterisk
2436
J. Grass et al.
worthwhile quantification of glycoforms must include all charge states. This can be accomplished by “deconvolution” of the spectrum as recently described [18, 19]. Deconvolution and MaxEnt3-deisotoping convert the different charge states and isotope peaks into one signal. However, for larger glycopeptides, this often leads to split signals, which complicate quantification. The present report evaluated the data with MassMap®. This software determines the intensity of an entire mass peak including the relevant isotopes and adducts. Ammonium adducts were prominent for glycopeptides 3 and 4 (Fig. 2). Massmap® allowed the correct assessment and identification of signals leading to the results listed in Table 2. A reproducible quantification of glycoforms in complex data sets can thus be achieved, which appears particularly relevant in the context of batch control. Peptides In the case of recombinant proteins intended for therapeutic use, a complete sequence coverage is obligatory. For the FSH α-chain, this could be achieved with tryptic peptides.
Analysis of the deglycosylated sample by MS/MS identified peptides that covered the entire length of the α-chain. However, the C-terminal peptide of the α-chain was found only in an incompletely cleaved form that gave an intense signal as glycopeptide but was difficult to find in the deglycosylated sample. In case of the β-chain, an additional chymotryptic digest was necessary to obtain strong peptide signals for the entire protein sequence (Table 1). Peptides are identified by their mass and by their sequence tags, which are the best denominators of amino acid sequence that can be obtained by MS/MS. Counting only the amino acid residues defined by sequence tags, in-depth sequence coverage of 63% and 79% for the α- and β-chains were obtained with a combination of 298 tryptic and chymotryptic peptides and of 55% and 69% if only trypsin was used. Tryptic peptide peaks were accompanied by earlier eluting satellite peaks larger by 16 Da. These may be interpreted as arising from oxidation, which amounted to about 8% for the β-chain Met and much less for the α-chain. However, chymotryptic digests did not show any sign of such peptides. So, while this result is inconsistent, it nevertheless demon25.87
100
4A4SF %
0
6
8
10
12
14
16
18
20
22
24
26
28
100
30 29.57
32
34
SIC intensity
4A3SF 26.22 %
23.81 25.36 24.58
27.16 28.71
0
6
8
10
12
14
16
18
20
22
24
26
28
30
32
34
22
24
26
28
30
32
34
17.71
100
18.57 15.47 16.59
4AF
%
19.94 20.46
0
6
8
10
12
14
16
18
20
Elution time
Fig. 4 Separation of isobaric N-glycans with or without N-acetyllactosamine (LacNAc)-repeats. The upper two traces show the elution profile of tetra-antennary glycans with four sialic acids (4A4SF) or three sialic acids (4A3SF) in the same LC-ESI-MS run of free, reduced glycans from the potential biosimilar. For the 4A4SF glycan, essentially, just one structure is conceivable and in fact, only one peak
occurs. The 4A3SF glycan may be the result of undersialylation or of LacNAc extension of either of the two possible tri-antennary Nglycans. Eight peaks of as yet unknown structure can be seen. Upon desialylation, six peaks still occur. The peak at 17.7 min represents the true tetra-antennary N-glycan, while others arise from LacNAc extension of tri-antennary glycans
Analysis of r-hFSH by mass spectrometric approaches
strates that the huge oxidized peaks appearing in intact protein analysis were artifacts. Glycans Released and reduced N-glycans were analyzed by LC- ESIMS on porous graphitic carbon. A very steep gradient was applied, and spectra from the entire elution zone of Nglycans were combined to give a simple picture of the glycan sizes present (Fig. 3). Selected samples were analyzed with a long gradient that allowed isomer resolution. An interesting aspect was the discrimination of a true tetra-antennary Nglycan with three sialic acids from tri-antennary structures with a lactosamine repeat (Fig. 4).
Conclusion A complex glycoprotein such as r-hFSH requires orthogonal analytical methods for its characterization. Whole protein ESI-MS, especially in the deglycosylated version, gave a valuable global view of the size of the FSH chains and of potential degradation products. A more detailed picture is provided by peptide and glycopeptide analysis using standard proteomics LC-ESI-MS techniques. Complete sequence coverage of both FSH chains was obtained both with tryptic and with chymotryptic digestion, provided the glycopeptides were taken into account. A complete site-specific glycosylation analysis was only possible with chymotryptic glycopeptides. Accurate quantification of all oligosaccharides present was obtained by quick passage over porous graphitic carbon with mass spectrometric detection. Positive-mode ESI yields a uniform detection sensitivity for the major ion species: neutral, mono-, and multiply sialylated glycans [20]. Although the positive-mode ESI results (Fig. 3) were similar to a MALDI spectrum, they constituted a more truthful profile than that obtained by reflectron-mode MALDI-TOF MS, which tends to skew the higher mass range (data not shown). The use of a long gradient allows ascribing structural details such as the linkage of sialic acids [15]. In the case of CHOcell produced glycoproteins, α2,3-linkages preponderate. The possibility to measure the ratio of under-sialylation versus elongation by lactosamine repeats appears quite interesting in the present case of r-hFSH but also for other recombinant glycoproteins, such as erythropoietin. For glycan profiling, other methods, such as MALDI-TOF MS or normal-phase HPLC of fluorescently labelled glycans [21–23], could alternatively be applied, albeit with the inherent problems of quantification in the case of MALDI (see above) or peak identification in the case of fluorescence HPLC [24]. A comparison of the glycan profiles of free glycans and of glycopeptides reveals that the very large structures easily seen in the spectrum of GP3 got somewhat
2437
lost in the glycan analysis, as here, the oligosaccharides from all four sites are pooled thereby diluting structures present on only one site. The spectrum thus resembles that shown in a previous publication, where tetra-antennary N-glycans were the largest structures identified [10]. However, zooming in on the elution and mass range of larger N-glycans reveals the presence of up to four “antennae” also in the free N-glycan pool (Fig. 3). The procedures when applied to two r-hFSH preparations revealed them to have apparently identical polypeptide chains but a somewhat different glycosylation pattern. In particular, for the biosimilar, the N-terminal glycosylation site of the β-chain contained a higher percentage of tri- and tetra-antennary glycans and of Nacetyllactosamine (LacNAc) repeats compared with the reference medicine. LC-ESI-MS analysis revealed that glycans with apparently one missing sialic acid are mainly rather fully sialylated structures with LacNAc repeats. This discrimination may matter as undersialylation and increased numbers of antennae have an opposite effect on the biospecific activity of r-hFSH [10]. Both USA and European regulatory guidelines allow for diversity in the glycosylation profile for biosimilar medicines (CHMP/437/04 Guideline on Similar Biological Medicinal Products; Biologics Price Competition and Innovation Act, section § 7002(k)(2)) provided that the protein structure is adequately defined (by comparability exercise) and the safety/efficacy profile of the biosimilar versus the reference medicine can be clinically demonstrated. To conclude, LC-ESI-MS analysis of tryptic and/or chymotryptic (glyco-)peptides turned out as the most informative approach. LC-ESI-MS of free glycans may, however, add details of relevance for the biological activity of the recombinant glycoprotein. Acknowledgments The authors gratefully acknowledge Dr. Richard Peck (Basel, Switzerland) for carefully reading the manuscript and Dr. Thomas Hemetsberger for valuable advice.
References 1. Messinis IE, Messini CI, Dafopoulos K (2010) The role of gonadotropins in the follicular phase. Ann NY Acad Sci 1205 (1):5–11 2. Check JH (2007) Treatment of male infertility. Clin Exp Obstet Gynecol 34(4):201–206 3. Pierce JG, Parsons TF (1981) Glycoprotein hormones: structure and function. Annu Rev Biochem 50:465–495 4. Green ED, Baenziger JU (1988) Asparagine-linked oligosaccharides on lutropin, follitropin, and thyrotropin. II. Distributions of sulfated and sialylated oligosaccharides on bovine, ovine, and human pituitary glycoprotein hormones. J Biol Chem 263(1):36–44 5. Green ED, Baenziger JU (1988) Asparagine-linked oligosaccharides on lutropin, follitropin, and thyrotropin. I. Structural elucidation of the sulfated and sialylated oligosaccharides on bovine, ovine, and human pituitary glycoprotein hormones. J Biol Chem 263(1):25–35
2438 6. Hard K, Mekking A, Damm JB, Kamerling JP, de Boer W, Wijnands RA et al (1990) Isolation and structure determination of the intact sialylated N-linked carbohydrate chains of recombinant human follitropin expressed in Chinese hamster ovary cells. Eur J Biochem 193(1):263–271 7. Dalpathado DS, Irungu J, Go EP, Butnev VY, Norton K, Bousfield GR et al (2006) Comparative glycomics of the glycoprotein follicle stimulating hormone: glycopeptide analysis of isolates from two mammalian species. Biochemistry 45(28):8665–8673 8. Stockell Hartree A, Renwick AG (1992) Molecular structures of glycoprotein hormones and functions of their carbohydrate components. Biochem J 287(Pt 3):665–679 9. Amoresano A, Siciliano R, Orru S, Napoleoni R, Altarocca V, De Luca E et al (1996) Structural characterisation of human recombinant glycohormones follitropin, lutropin and choriogonadotropin expressed in Chinese hamster ovary cells. Eur J Biochem 242(3):608–618 10. Gervais A, Hammel YA, Pelloux S, Lepage P, Baer G, Carte N et al (2003) Glycosylation of human recombinant gonadotrophins: characterization and batch-to-batch consistency. Glycobiology 13 (3):179–189 11. Loureiro RF, de Oliveira JE, Torjesen PA, Bartolini P, Ribela MT (2006) Analysis of intact human follicle-stimulating hormone preparations by reversed-phase high-performance liquid chromatography. J Chromatogr A 1136(1):10–18 12. Simpson RJ (2002) Proteins and proteomics: a laboratory manual. CSHL Press, Woodbury, NY 13. Pabst M, Bondili JS, Stadlmann J, Mach L, Altmann F (2007) Mass+retention time=structure: a strategy for the analysis of Nglycans by carbon LC-ESI-MS and its application to fibrin Nglycans. Anal Chem 79(13):5051–5057 14. Packer NH, Lawson MA, Jardine DR, Redmond JW (1998) A general approach to desalting oligosaccharides released from glycoproteins. Glycoconj J 15(8):737–747 15. Toegel S, Pabst M, Wu SQ, Grass J, Goldring MB, Chiari C et al (2010) Phenotype-related differential alpha-2,6- or alpha-2, 3-
J. Grass et al.
16.
17.
18.
19.
20.
21.
22.
23.
24.
sialylation of glycoprotein N-glycans in human chondrocytes. Osteoarthritis Cartilage 18(2):240–248 Carvalho CM, Oliveira JE, Almeida BE, Ueda EK, Torjesen PA, Bartolini P et al (2009) Efficient isolation of the subunits of recombinant and pituitary glycoprotein hormones. J Chromatogr A 1216(9):1431–1438 Hokke CH, Bergwerff AA, Van Dedem GW, Kamerling JP, Vliegenthart JF (1995) Structural analysis of the sialylated N- and O-linked carbohydrate chains of recombinant human erythropoietin expressed in Chinese hamster ovary cells. Sialylation patterns and branch location of dimeric N-acetyllactosamine units. Eur J Biochem 228(3):981–1008 Stadlmann J, Pabst M, Kolarich D, Kunert R, Altmann F (2008) Analysis of immunoglobulin glycosylation by LC-ESI-MS of glycopeptides and oligosaccharides. Proteomics 8(14):2858–2871 Kolarich D, Weber A, Turecek PL, Schwarz HP, Altmann F (2006) Comprehensive glyco-proteomic analysis of human alpha1antitrypsin and its charge isoforms. Proteomics 6(11):3369–3380 Pabst M, Altmann F (2008) Influence of electrosorption, solvent, temperature, and ion polarity on the performance of LC-ESI-MS using graphitic carbon for acidic oligosaccharides. Anal Chem 80 (19):7534–7542 Melmer M, Stangler T, Schiefermeier M, Brunner W, Toll H, Rupprechter A et al (2010) HILIC analysis of fluorescencelabeled N-glycans from recombinant biopharmaceuticals. Anal Bioanal Chem 398(2):905–914 Ruhaak LR, Huhn C, Waterreus WJ, de Boer AR, Neususs C, Hokke CH et al (2008) Hydrophilic interaction chromatographybased high-throughput sample preparation method for N-glycan analysis from total human plasma glycoproteins. Anal Chem 80 (15):6119–6126 Morelle W, Faid V, Chirat F, Michalski JC (2009) Analysis of Nand O-linked glycans from glycoproteins using MALDI-TOF mass spectrometry. Methods Mol Biol 534:5–21 Pabst M, Altmann F (2010) Instrumental strategies towards glycan analysis. Proteomics 11(4):631–643