Investigating the Influence of Glycosylation on Protein Conformation and Dynamics

Structure and structural dynamics are important attributes that enable proteins to perform many different activities in vivo. Although most of the unique structural and functional properties of a protein are dictated by its amino-acid sequence, there are many changes, called post–transitional modifications (PTMs) that can alter a protein as it is expressed, folds, and ages (1). These alterations can significantly change and control the specificity and strength of protein interactions, as well as influence the physico–chemical properties of the protein (e.g., stability, solubility, immunogenicity) (2). Probably one of the most common and important of these PTMs is glycosylation—the attachment of oligosaccharides to proteins.

During recent years, it has become increasingly apparent that glycosylation is important in controlling the function and solution behavior of proteins in solution (3, 4). More than half of all eukaryotic proteins are glycoproteins (5). The oligosaccharides, also referred to as glycans or carbohydrates, commonly found on a protein exist as either N–linked or O–linked oligosaccharides and typically consist of 2 to 14 monosaccharides chemically linked in a linear or branched configuration. N–linked oligosaccharides are chemically linked to asparagines (and to a lesser extent arginine) within the consensus sequence –Asn–Xaa–Ser/Thr (6), while O–linked oligosaccharides are chemically linked to any serine or threonine (7).

To date, most of the progress in characterizing the carbohydrates on glycoproteins has focused on studying their structure and composition. These results have shown that a diversity of oligosaccharide structures can exist at any given glycosylation site on a protein. This diversity is one of the major causes of the microheterogeneity observed in proteins and is one of the major challenges biopharmaceutical companies face in demonstrating lot–to–lot comparability in manufacturing a glycosylated protein biopharmaceutical. Differential glycosylation on proteins will also be an important factor in determining similarity between innovator products and their follow–ons.

Progress has been made in developing analytical tools capable of characterizing oligosaccharides in great detail (8, 9) and in associating their chemical differences with functional activities (10, 11). However, the physico–chemical basis for understanding just how oligosaccharides alter the function and structural properties of a protein is significantly lacking. A major challenge exists in understanding the details, at a molecular level, as to how oligosaccharides influence the structure and structural dynamics of a protein. Do the oligosaccharides float freely about the protein linkage site, do they interact directly with the protein, or are the oligosaccharides interacting with and influencing the protein surface or interior (see Figure 1)? Although nuclear magnetic resonance (NMR) and X–ray crystallography have contributed greatly in providing some of the answers to these and other questions, more orthogonal tools are needed. One tool that offers significant promise in helping to understand carbohydrate–protein interactions is hydrogen/deuterium–exchange mass spectrometry (H/DX–MS). Although the application of H/DX–MS to protein conformational studies is not new, recent developments now make H/DX–MS more capable, attractive, and informative (12). The authors briefly highlight how H/DX–MS can enhance our knowledge of the molecular details driving carbohydrate–protein interactions. In addition, we speculate on how these glycans may actually manipulate and control the biological function of a protein. Such scientific understanding should facilitate the development of new and more useful biopharmaceuticals.

Figure 1: A representative glycosylated protein (PDB: 1AUC) with its glycan (shown as cyan sticks) in different hypothetical orientations. A. Glycan not interacting with protein. B. Glycan interacting with the protein surface. C. Glycan interacting with the protein interior.

Brief history of H/DX–MS

Hydrogen exchange (HX) is a phenomenon that was identified in the 1950s by Hvidt and Linderstrom–Lang (13) and was used in developing the technique of hydrogen/deuterium exchange (H/DX). Later developments in H/DX coupled this technology with various spectroscopic techniques, which included NMR and FTIR to help better study and understand protein structure (14, 15). The eventual coupling of H/DX to MS (H/DX–MS) came in 1991 through the work of Katta et al. (16) and was further enhanced by the work of Zhongqi Zhang and David L. Smith while they were at Purdue University. They paired H/DX–MS experiments with proteolytic digestion (17). This latter union provides localized information concerning the relative amount and rate of H/DX in different parts of a protein. Such H/DX information can achieve peptide resolution as low as 5–10 amino–acid residues with nearly complete sequence coverage. H/DX–MS also has low sample requirements (down to a few picomoles) and can be performed in native or formulated buffer conditions. In addition, more recent advances in terms of both hardware and software now enable H/DX–MS experiments to be completed in a few days or less (12). As a result of these developments H/DX–MS has steadily gained in popularity as a technique for studying the higher–order structure of proteins.

The basics of H/DX–MS

The mechanisms of HX have been reviewed before in detail (18, 19). Briefly, HX occurs in a protein when labile hydrogen atoms change places with hydrogen atoms in solution. If an H₂O solution is replaced with D₂O (that is with deuterium instead of hydrogen), the labile hydrogen atoms in the protein will switch with deuterium atoms. In proteins, there are several different types of hydrogen atoms (e.g., aliphatic, hydroxyl, etc.), however, only the backbone amide hydrogen atoms exchange within a timeframe that can be monitored in a typical H/DX–MS experiment. Because every amino acid has a backbone amide (except for proline), these hydrogens act as sensors for monitoring the local conformational environment of the protein backbone (18). H/DX information can provide significant indirect insight into the conformational status of the protein polypeptide backbone. This information is especially informative in comparison studies of protein variants (19, 20).

Many factors influence H/DX such as temperature, pH, salt, and buffer composition (18). As a result, in a typical H/DX–MS experiment, the exchange reaction is carried out under conditions where these and other factors are well controlled (see Figure 2). The most common type of H/DX experiment is a continuous–labeling experiment where deuterium is added in 10– to 20–fold excess, and the reaction is allowed to incubate. As the reaction incubates, aliquots are removed over time (usually ranging from 10 s to more than 8 h) from the reaction and quenched to slow the labeling reaction. To quench the labeling, the pH is reduced to 2.5–2.6, where the hydrogen–exchange rate for a typical backbone amide is at a minimum (18). For the average backbone amide hydrogen, dropping the pH from 7.0 to 2.5–2.6 reduces the exchange by 4 orders of magnitude. In addition, lowering the temperature from 25 °C to 0 °C reduces the rate of exchange by another order of magnitude (18).

Figure 2: A schematic representation of the workflow for a continuous labeling H/DX–MS experiment (arrows indicate the order in which each step in an H/DX-MS experiment is carried out). In the initial step, a protein solution is incubated at ambient temperature in its formulated buffer. The protein is then diluted ~20 fold with deuterated formulation buffer. The protein is then incubated and at predetermined time points, a sample is taken from the reaction and quenched by dropping the pH to ~2.5-2.6 and the temperature to 0 °C. This quenched protein is then digested with an acidic protease (pepsin), and the resulting peptides are separated at 0 °C under acidic conditions chromatographically before they are introduced into the mass spectrometer. The mass of each peptide is determined for each deuterium time point, and the deuterium incorporation is plotted versus time. If the structure of the protein is known, it can be plotted onto this structure, otherwise different data analysis and display formats can be used to interpret the H/DX–MS data.

The insertion of a digestion step just after quenching (17), but prior to LC–MS (see Figure 2) cuts the protein into smaller pieces (peptides). During the digestion step, quench conditions must be maintained to preserve the deuterium label; therefore an acid protease (usually pepsin) is used for proteolytic digestion. After digestion, the peptic peptides are separated by reversed phase LC, and the mass of each peptide is measured by the mass spectrometer. Both digestion and chromatographic separation steps are carried out on–line. The reversed phase chromatography step not only functions to separate peptides, it also serves the important function of removing buffer salts and other excipients not compatible with MS.

Following the chromatographic separation, the peptides are analyzed by mass spectrometry, and the amount and location of the deuterium are determined (electrospray is most commonly used, although MALDI has also been used (21)). Typically, the protein and amino-acid sequence are known, and the resulting pepsin peptides are unambiguously identified by tandem MS. The solvent used in the digestion and chromatography steps are pure (100%) H² O. As a result, deuterium that was incorporated in the protein starts to exchange back to hydrogen during these steps. To minimize this, and to supply enough time for analysis, the temperature of the quench, digestion, and separation steps is dropped, and the chromatography is done as quickly as possible under acidic pH conditions. At 0 °C and pH 2.5–2.6, in a properly controlled experiment, only about 20%–30% of the deuterium is lost during analysis. If desired, control experiments can be performed to correct for this loss (17).

Using the deuterium uptake information, a comparison study can be conducted on a protein and its variant forms (i.e., a protein with and without different oligosaccharides). The main feature monitored in such comparisons is the difference in the locations and extent of deuterium incorporation as a function of time. These results provide indirect information concerning conformational changes in the polypeptide backbone of a protein as a result of a PTM. We have noted that protein deglycosylation can result in greater deuterium incorporation in some areas, and less in others.

Effect of deglycosylation on the conformation of an IgG1 antibody

Recombinant monoclonal antibodies (rmAb) are an important class of biopharmaceutical proteins. Antibodies are used to treat a wide array of medical conditions from autoimmune diseases to cancer to various other diseases (22). Today, more antibodies are in biopharmaceutical development pipelines and in clinical programs then are any other class of protein (23). The most common therapeutic rmAb is the immunoglobulin gamma 1 (IgG1). The IgG1 is made up of two identical subunits, a so–called homodimer. It is also a glycosylated protein with an apparent molecular weight approaching 150 kDa. An IgG1 is composed of two identical heavy chains and two identical light chains that are covalently connected by disulfide bonds. On each heavy chain (typically on residue Asn298), there is a single N–linked biantennary type oligosaccharide (6, 24). For IgG1 antibodies, glycosylation is a major source of heterogeneity, as most antibody expression systems (such as Chinese Hamster Ovary cells (25)) are incapable of producing an antibody containing a glycan with a uniform structural composition. Changes in IgG1 glycosylation (i.e., different levels of sialic acid, galactose, fucose, high mannose structures) have been shown to induce thermal destabilization, influence effector function, drug clearance, and may also provoke an immunogenic response (3, 26). Therefore, understanding the impact that glycans have on the conformation of an IgG1 (or any other glycosylated protein) is very important.

Recently, H/DX–MS was used to study the effect of glycans on the conformation of an intact IgG1 (24–27). It was demonstrated that complete removal of the IgG1 glycan resulted in differences in deuterium incorporation throughout the IgG1 Fc region (see Figure 3). Some regions were shown to incorporate more deuterium as a result of glycan removal, while other regions incorporated less deuterium. These data correlated nicely with previous structural studies on deglycosylated IgG1–Fc (26, 28). Since IgG1 glycans were shown by X–ray crystallography to make several contacts with many residues within the IgG1 Fc region (29), it is not surprising that removing the glycan results in a change in the IgG1 Fc conformation. This change in conformation ultimately resulted in thermal instability of the IgG1 Fc (30) and complete loss of effector function (27), relative to the native IgG1. Because glycans are thought to anchor the IgG1 Fc in a stable conformation, which promotes favorable interactions with effector function proteins. This and other results (31) suggest that removal of the glycans weakens the structural rigidity of the IgG1 Fc. As a result, glycan removal leads to a significant change in the conformation of the IgG1 Fc. The change observed by H/DX clearly indicates that structural perturbation may extend well beyond the protein regions directly in contact with the oligosaccharide. This type of change is indicative of allosteric effects, where perturbations in one region influence conformation at a distance.

Figure 3: A summary of the H/DX-MS results (24, 27) obtained when comparing different IgG1 glycovariants. (Top) Representative structure of an IgG1 antibody [PDB: 1HZH] (36) and common glycans. (Bottom) The left-hand panel shows the structural changes observed by H/DX-MS when the oligosaccharides were removed from the IgG1, while the right-hand panel shows the structural changes observed by H/DX-MS when only galactose was removed from the IgG1 oligosaccharide.

Effect of fucosylation and galactosylation on the conformation of an IgG1 antibody

In another experiment, a subtle change to the IgG1 glycan was made, and the conformation and dynamics of the IgG1 were investigated using H/DX–MS. IgG1 samples were compared in various combinations with and without galactose and fucose (27). Previous published work reported that decreased levels of fucose and increased levels of galactose increase effector function responses, and that these effects appear to be additive (27–28, 32). It was also shown that neither fucose nor galactose is directly involved in Fc receptor interactions (29, 33). If this is indeed the case, how is it that the addition of galactose or subtraction of fucose affects the effector functionality? Experiments using H/DX–MS showed that the deuterium incorporation into the IgG1 backbone with or without fucose was essentially the same, indicating no difference in the conformation of the IgG1 backbone. The fact that no change in H/DX–MS was seen when comparing the IgG1 with and without fucose indicates that the change in protein conformation may involve the repositioning of amino-acid side chains, which cannot be detected by H/DX–MS. It is also possible that segments of the protein's polypeptide backbone are repositioned to an environment that does not directly affect the solvent exposure of the backbone amide. In H/DX–MS experiments comparing IgG1 with and without galactose, the only difference that could be detected involved the reduction in deuterium incorporation in the IgG1 Fc heavy chain at residues 240–250 when galactose was present. The direct involvement of these IgG1 Fc residues with Fc receptors has been shown to be unimportant (27). These findings coupled with X–ray crystallography data that indicate that the Fc glycans form several contacts with a number of amino acids on the IgG1 Fc (29) are puzzling.

To better understand these data, H/DX MS binding experiments were performed to investigate the interaction between an Fc receptor, FcγRIIIa, and two IgG1 variants (27). In this work, an IgG1 with no fucose and complete galactosylation was found to interact differently with FcγRIIIa relative to the native IgG1, which contained fucose and very little galactose (~15%). While no direct evidence emerged to explain the molecular details on how galactose and fucose influence effector function, these data suggest that a change to the IgG1 backbone conformation important for Fc functionality may not have occurred. In the case of fucose, its presence may sterically interfere with optimal conformational changes required for the IgG1 Fc to bind the Fc receptor with maximum interactions, leading to significantly weaker binding. Conversely, the presence of galactose may interact with various amino–acid side chains on the IgG1 Fc to better facilitate a necessary conformational change that maximize IgG1 Fc interaction with Fc effector proteins. As a result, the absence of fucose or the presence of galactose may promote conformational flexibility within specific regions of the IgG1, which are favorable for effector function interactions. Such information should be of interest to those involved in engineering Fc–specific effector activity, as this information provides insights important for developing detailed models in explaining how carbohydrate–protein structure facilitates IgG1–effector function.

Conclusion

Early glycoprotein research has pointed to a general reduction in dynamic motion and increased thermal stability of a protein with its coupling to a carbohydrate (34). However, details at the molecular level on how protein dynamics and stability arise as a result of carbohydrate–protein interaction are not clear. X–ray crystallography is not always possible when a protein is flexible and/or modified by an oligosaccharide. In addition, the view of the unique conformation captured in a crystal may not be representative of those found in solution. In the case of NMR, the generally low sensitivity and potentially large sources of overlapping signals, put restrictions on the size of the protein that can reasonably be studied. In addition, the need to label proteins (usually using ¹⁵ N or ¹³ C) significantly limits routine sample analysis and throughput, which limits the widespread application of NMR.

H/DX–MS is a sensitive solution–based technique that offers spatial resolution to a few amino-acid residues, typically 5–10. In fact, recent developments using electron transfer disassociation (ETD) offers an opportunity for H/DX–MS to reach single–residue resolution with nearly complete sequence coverage (35). Given these attributes and those mentioned earlier in this report, H/DX–MS is capable of providing useful information to answer many questions in the biopharmaceutical industry related to the higher order structure of biopharmaceuticals. This capability was briefly demonstrated in this report by the ability to conduct an array of comparability studies to reveal the impact of different oligosaccharides on the protein structure of an IgG1. Such capabilities are useful to researchers responsible for designing new protein biopharmaceuticals, as well as those in biopharmaceutical process development responsible for delivering a stable and consistent commercial drug product. As a result, H/DX–MS should see a growing interest from a wide spectrum of biopharmaceutical scientists.

Acknowledgments

The authors would like to thank Dr. Helena Madden, Dr. Rohin Mhatre, and Dr. John R. Engen for helpful discussions and for critical reading of the manuscript.

Damian Houde* is a scientist, and Steven A. Berkowitz, is a principal investigator, both at Biogen Idec, Inc., 14 Cambridge Center, Cambridge, MA 02142, tel: 617.914.5841, fax: 617.679.3476.

*To whom all correspondence should be addressed.

References

1. C.T. Walsh, S. Garneau–Tsodikova, and G. J. Gatto Jr., Angew Che m.Int. Ed. Engl. 44 (45), 7342–7372 (2005).

2. G. Walsh and R. Jefferis, Nat. Biotechnol. 24 (10) ,1241–1252 (2006).

3. J.N. Arnold et al.,. Annu. Rev. Immunol. 25, 21–50 (2007).

4. N.Mitra et al., Trends Biochem. Sci. 31 (3), 156–163 (2006).

5. R.J. Sola, J.A. Rodriguez–Martinez, and K. Griebenow, Cell. Mol. Life Sci. 64 (16) 2133–2152 (2007).

6. T.W. Rademacher, R.B. Parekh, and R.A. Dwek, Annu. Rev, Biochem, 57, 785–838 (1988).

7. P. Van den Steen et al., Crit. Rev. Biochem. Mol. Biol. 33 (3), 151–208 (1998).

8. N.J. Agard and C.R. Bertozzi, Acc. Chem. Res. 42 (6), 788–797 (2009).

9. J. Zaia, Chem. Biol. 15 (9) 881–892 (2008).

10. M.R. Wormald and R.A. Dwek, Structure 7 (7), R155–160 (1999).

11. J.J. Caramelo and A.J. Parodi, Semin. Cell. Dev. Biol. 18 (6), 732–742 (2007).

12. J.R. Engen, Anal. Chem. 81 (19), 7870–7875 (2009).

13. A. Hvidt and K. Linderstrom–Lang, Biochim. Biophys. Acta 14 (4) 574–575 (1954).

14. P.I. Haris and D. Chapman, Biopolymers 37 (4), 251–263 (1995).

15. S.W. Englander and L. Mayne, Annu. Rev. Biophys. Biomol. Struct. 21, 243–265 (1992).

16. V. Katta and B.T. Chait, Rapid Comm. Mass Spec. 5 (4), 214–217 (1991).

17. Z. Zhang and D.L. Smith, Protein Sci. 2 (4), 522–531 (1993).

18. S.W. Englander and N.R. Kallenbach, Q. Rev. Biophys. 16 (4), 134 (1983).

19. T.E. Wales and J.R. Engen, Mass Spectrom. Rev. 25 (1), 158–170 (2006).

20. A.N. Hoofnagle, K.A. Resing, and N.G. Ahn, Annu. Rev. Biophys. Biomol. Struct. 32, 1–25 (2003).

21. J.G. Mandell, A.M. Falick, and E.A. Komives EA, Anal. Chem. 70 (19), 3987–3995 (1998).

22. O.H. Brekke and I. Sandlie, Nat. Rev. Drug Discov. 2 (1), 52–62 (2003).

23. O. Leav, Nat. Rev. Immunol. 10 (5), 297–297 (2010).

24. D. Houde et al., Anal. Chem. 81 (7), 2644–2651 (2009).

25. H.E. Chadd and S.M. Chamow, Curr. Opin. Biotechnol. 12 (2), 188–194 (2001).

26. Y. Mimura et al., Mol. Immunol. 37 (12–13), 697–706 (2000).

27. D. Houde , Mol. Cell. Proteomics 9 (8), 1716–1728 (2010).

28. Y. Yamaguchi et al., Biochem. Biophys. Acta 1760 (4), 693–700 (2006).

29. P. Sondermann et al., Nature 406 (6793), 267–273 (2000).

30. H. Liu and C.G. Bulseco, Immunol. Lett. 106 (2), 144–153 (2006).

31. S. Krapp et al., .J Mol. Biol. 325 (5), 979–989 (2003).

32. A. Okazaki et al., J. Mol. Biol. 336 (5), 1239–1249 (2004).

33. S. Radaev et al., J. Biol. Chem. 276 (19), 16469–16477 (2001).

34. D. Shental–Bechor and Y. Levy, Curr. Opin. Struct. Biol. 19 (5), 524–533 (2009).

35. K.D. Rand et al., Anal. Chem. 81 (14), 5577–5584 (2009).

36. E.O. Saphire et al., Science 293 (5532), 1155–1159 (2001).