Rapid Assessment of Molecular Similarity between a Candidate Biosimilar and an Innovator Monoclonal Antibody Using Complementary LC–MS Methods

August 2, 2010

Intact protein LC–MS detected a mass variance of 62 Da and peptide mapping located a difference of two amino acids.

ABSTRACT

There is an emerging interest in developing biosimilar monoclonal antibodies (MAbs). To avoid expensive clinical trials and shorten time to market, the biosimilar industry must establish that a developing product is, as much as possible, similar to a marketed innovator product through comprehensive analysis. Here, we demonstrate that ultra high pressure liquid chromatography (UHPLC) and mass spectrometry (MS) can be used routinely to characterize minor differences between a candidate biosimilar and an innovator IgG1 MAb. A two amino acid residue variance in the heavy chain sequence was detected by LC–MS intact protein mass measurement and located by tryptic peptide mapping with data independent acquisition LC–MS. Microheterogeneities due to N-linked glycosylation and chemical degradation were comprehensively catalogued and compared. The results show that complementary LC–MS methods can be used as a set of routine tools for rapid comparison of molecular similarity between a candidate biosimilar and a commercially available MAb.

(WATERS CORPORATION)

Recombinant monoclonal antibodies (MAbs) represent a class of efficient, but expensive, biotherapeutics. Developing less costly generic "biosimilar" MAbs is of great interest to both drug companies and consumers. Biosimilar drugs can be defined as compounds that are as similar as possible, structurally and functionally, to an innovator drug. The driving force for the interest in biosimilars for generic drug manufacturers is the upcoming patent expiration for marketed products.1 However, developing biosimilar products is challenging. Currently, there are no registered biosimilars in the US, and only recently has a mechanism for their approval been created.2 Biosimilar manufacturers will be under pressure to ensure that their products conform as closely as possible to existing products if they wish to avoid repeating expensive clinical trials and thus reach the market faster. Therefore they have a vested interest in comprehensive product analysis at all stages of development and manufacturing.

The key criteria for approval of biosimilars include quality, efficacy, and safety. Therefore, the generic biopharmaceuticals industry must try to demonstrate consistency of a biosimilar to the innovator reference product in every aspect. A number of physicochemical and biological tests are required by regulatory authorities for the characterization of MAbs.3,4 European Union regulations specify that state-of-the art characterization studies should be applied to the "similar biological product" (Section 5.2)5 and that physicochemical characterization should include a determination of the primary structure (section 5.2.2) while post-translational modifications (PTMs) such as glycosylation exhibit only minor levels of microheterogeneity.

In the work presented here, we demonstrate the application of state-of-the-art liquid chromatography (LC) and mass spectrometry (MS) to perform a comprehensive comparison of innovator and biosimilar MAbs. LC–MS analysis was carried out at the intact protein level, providing the information about molecular weight and glycan heterogeneity. Middle-down analysis of the MAb after reduction provided mass information on the antibody light chain and heavy chain separately, whereas the bottom-up approach (analysis of the MAb tryptic digest) allowed us to locate and identify the differences in primary MAb sequence and PTMs.

For peptide mapping, we used a novel, data-independent acquisition method that allows for simultaneous collection of both MS and fragment ion data for every peptide in the data set independent of their mass, intensity, and retention time.6,7 The data-independent MS acquisition method permits reliable acquisition of both quantitative (peptide MS signal) and qualitative (peptide sequence) data in a single LC–MS analysis.

The goal of this manuscript is to demonstrate the utility of a comprehensive set of advanced UHPLC–MS methods and their suitability for comprehensive comparison of innovator products and biosimilar MAb drug candidates, using a biosimilar of trastuzumab as an example.

Materials and Methods

Separations were performed on an Acquity UHPLC system (Waters Corporation) equipped with tunable ultraviolet (TUV; for peptide mapping) or fluorescence (FLR, for released glycan profiling) detectors. All MS measurements were made on a Synapt HDMS system (Waters). The systems were controlled by MassLynx 4.1 software (Waters); data processing was performed with BiopharmaLynx v. 1.2 (Waters). Simglycan (v. 2.75) (Premier Biosoft International) was used for glycan identification from a MALDI MS–MS experiment. The conditions of intact protein analysis and peptide mapping were the same as described in recently published papers.8,9 The conditions for hydrophilic interaction chromatography-fluorescence (HILIC–FLR) analysis were described elsewhere.10

Results and Discussion

Intact Protein LC–MS Analysis

The intact MS analysis of large proteins such as MAbs requires appropriate preparation but can be performed routinely. The samples must be thoroughly desalted before MS analysis to obtain a sufficient signal. In our experience, the best results are obtained with on-line desalting.11

The reversed-phase desalting column was used in an on-line setup to obtain UHPLC–MS data shown in Figure 1. Deconvoluted data for the innovator and biosimilar MAbs are presented as a mirror plots; the glycoform annotation is based on the deconvoluted MS signals. One can clearly observe protein mass heterogeneity resulting from the presence of several main protein glycoforms. Several dominant peaks can be distinguished for each MAb sample; the difference in mass of ~162 Da corresponds most likely to loss of galactose, G, and ~146 to fucose, F, from the N-linked glycan core structure. The molecular weight of the innovator MAb is consistent with its expected primary sequence, i.e., its mass corresponds to the sum of the mass of two G0F, G0F/G1F, two G1F, G1F/G2F, or two G2F glycans (Figure 1, upper trace). Interestingly, the biosimilar candidate shows different glycoform heterogeneity and all main glycoforms show a mass shift of ~64 Da compared with the innovator drug. This shift is putatively caused by the presence of unknown PTMs or by differences in the primary amino acid sequence.

Figure 1. Intact MAb MS analysis. Multiply charged spectra were deconvoluted into protein molecular mass and compared using BiopharmaLynx software. Heterogeneity is caused by the MAb's glycoforms. Notice the mass shift of biosimilar drug candidate compared to the reference innovator sample.

Further investigation of intact protein mass data suggests the presence of minor glycoforms such as G0/G0F in each MAb (a mass difference of 146 Da, due to incomplete occupancy of a fucosylation site on one of the two G0F core glycan structures). In addition, a low-level Man5/Man5 form was detected in both MAbs.

Although the intact protein analysis is fast and useful, it is difficult to quantitatively measure the relative glycoform ratios, especially for minor glycoform species (Figure 1). Therefore, we performed a partial reduction of MAb samples, followed by on-line desalting and MS analysis. The deconvoluted masses of light and heavy chains are shown in Figures 2A and 2B, respectively. Whereas the light chains are identical, the heavy chain data for the innovator and biosimilar MAbs show a distinct mass shift for all glycoforms. Analyzing the 50-kDa heavy chain alone allows for better assessment of glycosylation heterogeneity (whereas the information about heterogeneity on the intact MAb molecular level is now lost). Man5, G0, G0F, G1, G1F, G2, and G2F glycoforms were clearly distinguishable in deconvoluted MS spectra (Figure 2B).

Figure 2. UHPLC–MS analysis of reduced MAb samples. Data were processed by BiopharmaLynx software. A) Light chain of innovator and biosimilar MAb are identical. B) Heavy chain of biosimilar candidate MAb shows consistent mass shift of ~32 Da for all protein glycoforms compared to the reference sample. Glycosylation pattern differences were detected.

When comparing the innovator to the biosimilar MAb, the consistent mass shift of ~32 Da is observed for all glycoforms, suggesting a possible modification of the biosimilar MAb candidate heavy chain. This is consistent with the previously detected ~64 Da difference in the intact protein mass (i.e., the combined contribution of two heavy chains). In the next section, we describe a peptide mapping experiment used to pinpoint the origins of the mass shift.

Application of Peptide Mapping to Locate Sequence Variants

Tryptic digests of both samples were analyzed using the data-independent acquisition UHPLC–MS approach ("UPLC–MSE "). It has been shown that data-independent acquisition LC–MS with alternate low and elevated collision energy scanning provides significant benefits in terms of completeness and speed of peptide mapping.8 In contrast to data-dependent acquisition (DDA) LC–MS–MS, which requires a list of precursors (targeted DDA) or relies on intensity-biased precursor selection on the fly (often resulting in omission of minor peptides), data-independent acquisition MS allows for sequencing of all peptides above the limit of detection and for accurate quantification by MS signals in a single analysis.

Figure 3. UHPLC–MS tryptic peptide maps of innovator and biosimilar MAbs presented as total ion chromatograms.

Figure 3 shows a mirror plot of UHPLC–MS peptide maps of the biosimilar and innovator MAbs generated by BiopharmaLynx software. The detected tryptic peptide masses were matched against the theoretical values using a published trastuzumab sequence.12 Although the total ion chromatogram traces do not show significant differences between samples (Figure 3), the compare function in the software visualized chromatogram regions for each of the MAbs that differed. Besides differences in glycosylation, the software indicated additional differences, made apparent by displaying charge-reduced and isotope-deconvoluted MS "stick chromatogram" plots, as shown in Figure 4. The mirror plot of the selected region of the chromatogram highlights that the innovator and biosimilar MAbs each contain a unique peptide without a corresponding signal in the other sample. The peptide HT35 (heavy chain, tryptic peak T35) with the amino acid sequence EEMTK found in the innovator product was not present in the biosimilar candidate MAb. Instead, a peak with a mass of 605.32 Da was observed. The calculated difference in mass between this unknown peptide and EEMTK is ~32 Da, which correlates with our previous results for the differences in mass in the heavy chain (~32 Da) and intact MAb (~64 Da).

Figure 4. UHPLC–MS data of peptide maps shown as deisotoped and charge reduced chromatograms (all charge states and isotopes were deconvoluted into singly charged ion "stick"). Selected part of chromatogram between 3.5 and 7.5 min shows one unique peptide detected in innovator and biosimilar MAb digests.

The MS data suggest that the innovator and biosimilar MAbs have a local inconsistency in the primary amino acid sequence located in the HT35 peptide. Because the UHPLC–MS data contain fragmentation data for all peptides (including the unknown and unidentified ones) it was possible to investigate the unknown peak in BiopharmaLynx, as well as in the PepSeq programs, with the latter capable of assisting de novo sequencing. This task was simplified by the availability of information about alternative MAb allotypes in the DrugBank.13,14 An alternative sequence matched the HT35 peptide mass and the fragmentation pattern matched the DELTK sequence, with a difference of two amino acids from the innovator MAb EEMTK peptide. Both sequences were confirmed using data-independent acquisition MS (Figure 5).

Figure 5. Spectra of unique peptides, obtained using data-independent acquisition MS, confirm the primary sequence of HCT35 peptides for innovator antibody. As expected, it is EEMTK. Biosimilar drug candidate sequence of the HCT35 peptide is DELTK, which is a known MAb allotype.

Analysis of Glycans

Intact protein UHPLC–MS and peptide mapping provided information about the heterogeneity of MAb glycosylation and indicated that there are differences in the glycosylation patterns of the investigated samples. To gather more detailed and quantitative information about the glycosylation patterns, we performed LC analysis using a hydrophilic interaction chromatography (HILIC) column of released glycans using fluorescent detection (FLR). Glycans were cleaved from MAbs by PNGase F enzyme, labeled with 2-AB fluorescent dye, and analyzed using an Acquity UHPLC HILIC glycan column. Chromatograms featuring well resolved glycans including isomers of G1 a/b and G1F a/b are shown in Figure 6.

Figure 6. LC-FLR profiling of released and 2-AB labeled glycans. Peak assignment was verified by reference standards and by MALDI–MS analysis. Relative quantitation values shown are based on fluorescence signal.

The HILIC method quantified differences in the glycan profiles of the innovator and biosimilar MAbs. Whereas G1F was the most abundant glycan structure in the innovator product, G0F was the most abundant in the biosimilar sample. The sensitivity of fluorescent detection permits the relative quantitation of minor glycans. Glycan identity was confirmed by LC–MS–MS.

Conclusions

Comprehensive investigation of an innovator MAb (trastuzumab) and a biosimilar candidate MAb with LC–MS and LC–FLR techniques revealed unexpected differences in their primary sequence. Intact protein analysis showed mass differences and heterogeneity in the glycosylation patterns. To further elucidate whether the differences result from sequence variations or post-translation modifications, a peptide mapping experiment was performed, revealing an unexpected peptide sequence. The UHPLC–MS data analysis of the peptides in BiopharmaLynx and PepSeq software, in conjunction with information about possible allotype sequences from DrugBank, allowed the identification of the biosimilar candidate drug sequence in a single UHPLC–MS experiment. Clear differences in the levels of glycosylation observed in LC–MS experiment of intact and denatured MAbs were confirmed by an LC–FLR experiment with released labeled glycans detected by fluorescence. The proposed set of experiments is robust and suitable for characterization and quality monitoring of biopharmaceutical proteins at all stages of drug development.

MARTIN GILAR, PHD, is a principal researcher, HONGWEI XIE, PHD, is a senior research scientist, ASISH CHAKRABORTY, PHD, is a senior chemist, JOOMI AHN is a senior research chemist, YING QING YU, PHD, is a principal chemist, WEIBIN CHEN, PHD, is a principal chemist, ST. JOHN SKILTON, PHD, is the senior marketing manager, and JEFFERYR. MAZZEO, PHD, is the biopharmaceutical business director, all at Waters Corporation, Milford, MA, 508.482.3119, martin_gilar@waters.comDEEPALAKSHMI P. DAKSHINAMOORTHY is a senior applications specialist at Waters India Pvt Ltd, Bangalore, India.

REFERENCES

1. Hughes B. Gearing up for follow-on biologics. Nat Rev Drug Discov. 2009;8:181.

2. The Patient Protection and Affordable Care Act, Pub.L. 111–148, 124 Stat. 119 (Mar 23, 2010).

3. Beck A, Bussat MC, Zorn N, Robillard V, Klinguer-Hamour C, Chenu S, et al. Characterization by liquid chromatography combined with mass spectrometry of monoclonal anti-IGF-1 receptor antibodies produced in CHO and NS0 cells. J Chromatogr B Analyt Technol Biomed Life Sci. 2005;819:203-18.

4. Reichert JM, Beck A, Iyer, H. European Medicines Agency workshop on biosimilar monoclonal antibodies, 2 July 2009, London, UK. MAbs. 2009;1:394–416.

5. European Commission. Guideline on Similar Biological Medicinal Products containing Biotechnology-derived proteins as active substance: Quality Issues. Brussels, Belgium; 2006 June. Available from: http://www.ema.europa.eu/pdfs/human/biosimilar/4934805en.pd.

6. Silva JC, Denny R, Dorschel C, Gorenstein MV, Li GZ, Richardson K, et al. Simultaneous qualitative and quantitative analysis of the Escherichia coli proteome: a sweet tale. Mol Cell Proteomics. 2006;5:589–607.

7. Silva JC, Gorenstein MV, Li GZ, Vissers JP, Geromanos SJ. Absolute quantification of proteins by LCMSE: A virtue of parallel MS acquisition. Mol Cell Proteomics. 2006;5:144–56.

8. Xie H, Gilar M, Gebler JC. Characterization of protein impurities and site-specific modifications using peptide mapping with liquid chromatography and data independent acquisition mass spectrometry. Anal Chem. 2009;81:5699–708.

9. Xie H, Chakraborty A, Ahn J, Yu YQ, Dakshinamoorthy DP, Gilar M, et al. Rapid comparison of a candidate biosimilar to an innovator monoclonal antibody with advanced liquid chromatography and mass spectrometry technologies. MAbs. 2010;2:4.

10. Ahn J, Bones J, Yu YQ, Rudd PM, Gilar M. Separation of 2-aminobenzamide labeled glycans using hydrophilic interaction chromatography columns packed with 1.7 microm sorbent. J. Chromatogr B Analyt Technol Biomed Life Sci. 2009;878:403–8.

11. Chakraborty AB, Berger SJ, Gebler JC; Waters Corporation. Biopharmaceutical Applications Notebook: A Focus on Protein Therapeutics. Waters application note 2007;720002439en.

12. Harris RJ, Kabakoff B, Macchi FD, Shen FJ, Kwong M, Andya JD, Shire SJ, Bjork N, Totpal K, Chen AB. Identification of multiple sources of charge heterogeneity in a recombinant antibody. J Chromatogr B Biomed Sci Appl. 2001;752:233–45.

13. DrugBank.ca [web site]. Alberta, Canada: David Wishart, Departments of Computing Science & Biological Sciences, University of Alberta. [cited 2010 July 9]. Available from http://www.drugbank.ca/drugs/DB00072.

14. Jefferis R, Lefranc MP. Human immunoglobulin allotypes: possible implications for immunogenicity. MAbs. 2009;1:332–8.