Review of Orthogonal Methods to SEC for Quantitation and Characterization of Protein Aggregates

Peer-Reviewed
Article submitted: Aug. 25, 2014.
Article accepted: Oct. 16, 2014.

Fotosearch/Getty Images

Abstract

Determination of the soluble aggregate levels in protein pharmaceuticals has historically relied upon size-exclusion chromatography (SEC). In recent years, however, there has been a realization that SEC can provide incorrect estimates of the actual aggregate content for a variety of reasons. As a result, orthogonal methods have been evaluated as alternatives to SEC for estimating the amounts and sizes of aggregates in protein products. Of these orthogonal methods, two have been preferred, both in the literature and by regulatory agencies: sedimentation velocity analytical ultracentrifugation (SV-AUC) and field-flow fractionation (FFF), especially asymmetrical-flow field-flow fractionation (AF4). This review describes the relative disadvantages and advantages of these various methods. In doing so, the use of orthogonal methods to SEC is discussed and examples are presented showing how analytical ultracentrifugation (AUC), AF4, and SEC compare in aggregate quantitation.

Quantitative assessment of aggregate levels is required by regulatory agencies for any protein therapeutic. Historically, the primary method for measuring the size and amount of soluble aggregates in protein samples has been size-exclusion chromatography (SEC). While SEC can be an effective tool in terms of sensitivity, precision, and throughput, there are some significant technical limitations of SEC that can lead to erroneous estimates of aggregate levels in protein products (1). These shortcomings have been described and discussed in detail in the literature (1--4). A list of the advantages and disadvantages of SEC is given in Table I.

As a result of these shortcomings, there has been an emphasis on identifying analytical techniques that can provide information on protein aggregate content using alternative separation approaches. Given the concern over the potential connection between aggregation and immunogenicity, regulatory authorities, such as FDA, have been encouraging applicants to develop analytical methods that provide an estimate of aggregate content obtained with orthogonal methodologies and verify the results of SEC measurements (5, 6). While not routinely used as methods to release products to market, regulatory agencies realize that alternative orthogonal methods might be informative. It is expected that  orthogonal methods will be used more often in the future as quality control methods for biopharmaceuticals.

The two most common orthogonal methods are analytical ultracentrifugation (AUC) and asymmetrical-flow field-flow fractionation (AF4). These are termed orthogonal methods in that they separate species of various sizes by different mechanisms and are used in addition to primary methods of separation. Orthogonal methods also allow the separation of aggregates of widely differing sizes. For example, the resolution of large soluble aggregates with SEC reaches its size limit at the void volume. While this size resolution is column-dependent, it rarely exceeds 10⁷ Daltons, corresponding to an upper size limit near 100 nm (7). By comparison, the size range for AUC and AF4 can easily extend into the micron-size range. Given the different separation methodologies that are available, estimates of aggregate levels may, and do, vary. This review describes what has been reported about orthogonal methods as they apply to protein aggregation, not only in terms of quantitation but also in terms of characterization and size determinations.

Size-Exclusion Chromatography
Currently, there is a concern that underestimation of aggregate content in protein drug products might lead to deleterious health effects when these products are used. As a result, there have been efforts to develop improved SEC techniques and approaches. Some fundamental limitations to SEC remain, however, namely that SEC tends to underestimate aggregate content in protein samples. Consequently, a recent review states, “SEC is also an accurate method if confirmed with an orthogonal method…” (4). While the accuracy of aggregate quantitation by SEC can be improved by changes to the mobile phase composition and detection methodology, there is no doubt that SEC exhibits better precision than most analytical methods, including the orthogonal methods described herein (8--10).

One limitation of SEC is the potential for adsorption of the protein to the stationary phase of the column. More specifically, there have been reports of the aggregate preferentially adsorbing to the column (3, 11, 12), leading to an underestimation of the aggregate content. As a result, it has been proposed that adding compounds, such as arginine (Arg) or nonaqueous solvents, could help enhance solubility and reduce adsorption (13--16).

While UV absorbance (whether at a single wavelength or with a diode array detector) is by far the most widely employed detection scheme in SEC, there can be distinct advantages to using other types of detectors or even multiple detectors. The use of a fluorescence detector has been reported to aid in quantitation of aggregates for interferon-∝ (17), monoclonal antibodies (mAbs) (18), and erythropoietin (19). In addition, use of fluorescence detection can be beneficial if polysorbate is present, as the interference from this surfactant can be minimized using this technique. Comparison of the relative UV-absorbance at two wavelengths (214 nm and 280 nm) is another strategy that has been shown to improve sensitivity (20).

Analytical Ultracentrifugation
Analytical ultracentrifugation (AUC) employs a high-speed centrifuge equipped with a rotor with various cell compartments. AUC can be operated in two different modes: sedimentation velocity and sedimentation equilibrium. In terms of using AUC as an orthogonal method to SEC, one is primarily concerned with sedimentation velocity measurements, even though it will be denoted as AUC in this review. In sedimentation velocity mode of operation, samples are spun at high rates of speed, with larger species sedimenting to a greater degree than the smaller ones (21). The velocity and shape of the moving boundary is used to estimate the sedimentation coefficient, which can then be related to molecular weight once shape information is obtained or assumed (8, 22). Even though the resulting distribution of various size species looks somewhat like a chromatogram, these profiles are the outcome of mathematical modeling and thus are not a direct measure of content, as is the case with chromatographic methods. The ability to analyze and interpret sedimentation velocity analytical ultracentrifugation (SV-AUC) data has no doubt been facilitated by newer software algorithms, such as SEDFIT (23, 24). More details about AUC methodology can be found in numerous review articles (9, 21, 25).

As with any analytical method, there are advantages and disadvantages to AUC (Table II) (9). The resolution of oligomeric species, such as dimers and trimers from the monomer by AUC, appears to be better than it is for SEC or AF4. Consequently, this can be considered to be a significant advantage of the SV-AUC technique. At the same time, AUC can measure species as large as several hundred million Daltons in molecular weight (26). The ability to make measurements without the dilution effects characteristic of SEC is worth noting (Table II). It should also be pointed out that one can often make accurate determinations of aggregate content by AUC directly in the formulation buffer.

One disadvantage of AUC compared to SEC is its higher limit of detection (LOD). The LOD issue was examined by Pekar and Sukumar (27), where they found the LOD of AUC was about three times higher than that of SEC. Another disadvantage of AUC is that excipients such as sugars and polyols can create density gradients, which can increase the limit of quantitation (22). These factors may cause one to underestimate aggregate levels, but there are modeling approaches that may help account for these effects (22).

The greatest overall disadvantage of AUC is its low throughput, as samples must be spun for many hours to reach equilibrium. Additionally, proper analysis of AUC data requires a significant amount of time and experience, which limits throughput as well. While newer software algorithms have improved the ease of use, the long data collection and analysis times required for SV-AUC result in relatively few samples being analyzed compared with AF4 or SEC.

SEC/SV-AUC Orthogonal Studies
There have been some noteworthy direct comparisons of AUC and SEC for aggregate quantitation. For example, the two methods, when used for measuring aggregates of ∝-glucosidase, are correlated (28). In this particular study, however, the authors note that the detection limit for SEC is much better than AUC, with a LOQ that was near 3% aggregate for AUC and 0.2% for SEC. In another study, SV-AUC and SEC were comparable when measuring dimer levels in three different protein samples which had dimer contents ranging from 1% to 12% (29). In this same study, the two methods produced analogous results for a highly aggregated sample; the aggregate content was 17.2% for AUC and 17.9% for SEC (29). When aggregate contents are low, it appears that the orthogonal methods tend not to agree, likely because the techniques separate aggregates by different principles and across different size ranges. However, as the aggregate content increases, the numbers become more similar, although some exceptions to this trend have been reported (8).

Asymmetrical-Flow Field-Flow Fractionation
The second, predominant orthogonal method to SEC is asymmetrical-flow field-flow fractionation (AF4). A number of articles on AF4 have now been published (30-33).

Since AF4 separates protein species in an open channel, concerns over analyte adsorption to packing material (such as that which is present in SEC columns) are largely eliminated (Table I). However, adsorption of analyte to the membrane used in an AF4 system is still a possibility (34, 35). Separation in an AF4 system is achieved by applying a cross-flow perpendicular to the direction of sample/analyte flow in the AF4 channel. Upon application of the cross flow, analytes accumlate near the semipermeable membrane and the lagging edge of the channel laminar flow. Because low molecular-weight species have larger diffusional coefficients than high molecular-weight species, they diffuse more readily into the faster moving region of the laminar flow (farther above the membrane), and thus, have faster elution times than larger species. As with HPLC methods, the cross flow can be programmed to be constant or applied in a gradient fashion, which allows one to fine-tune resolving power for a given set of analytes. The relative effects of various operational parameters in AF4 on peak shape and position have been described in some detail by the existing literature (32). There has been at least one report on the effect of mobile-phase composition on the separation of proteins (36), but this aspect has been largely overlooked to date. AF4 can be run with a wide variety of mobile phases, and an AF4 separation in the actual formulation buffer can also be conducted.

Fotosearch/Getty Images

SEC/AF4/SV-AUC Orthogonal Studies
For a long time, the lack of commercial instrumentation hindered further development of AF4 as an orthogonal method. With the advent of vendors providing AF4 instruments (e.g., Wyatt Instruments, Postnova), the field has developed rapidly over the past decade. During this time, studies using AF4 to characterize higher-molecular weight species, such as nanoparticles (40, 41) and protein aggregates (40), began to appear in the literature.

Some of the first comparisons of SEC and AF4 were quite striking in their differences, in that aggregates levels measured by the two techniques varied quite significantly (42). One example of this from the literature involves the measurement of aggregates in a granulocyte colony-stimulating factor (G-CSF) sample (43). While SEC determined the G-CSF samples contained <3% aggregate, AF4 found as much as 49% aggregate. Different instrumentation or protocols could have the potential to change these estimates of aggregate content, but this is purely speculative and remains to be demonstrated.

The earliest example of a comparison of SEC and AF4 was published more than 20 years ago (44), before AF4 was commercially available. In this work, a mAb was evaluated by both techniques. The authors showed that the precision of AF4 was not as good as SEC. As for quantitation of aggregates, SEC measured an aggregate content approximately half of that determined by AF4 (11.2% by SEC compared with 23.3% by AF4 for one mAb, 7.1% by SEC compared with 15.6% by AF4 for another mAb). The reasons for these differences were not discussed, but they could have arisen from variations in mobile phase composition, or they may reflect the differences in the mechanism of separation between SEC and AF4. In another study, where a mAb was exposed to UV light, both AF4 and SEC were used to measure loss of monomer and appearance of fragments and aggregates (45). In this study, it was determined that the two methods were comparable, but AF4 was less effective than SEC at quantifying fragments (46).

A comparison of SEC and AF4 on stressed IgG formulations found that AF4 detected fewer dimeric species than SEC when samples were unstressed or heated to 70º C. When the temperatures were increased to 80º C, however, where aggregation was significant, the SEC method underreported high molecular aggregates relative to AF4 (18).

In another study examining severely heat-stressed mAbs, AF4 was found in every case to provide better separation and more complete recovery of the sample than SEC (38), however, AF4 underestimated the level of fragments relative to SEC. In a final aggregation study comparing SEC and AF4, it was found that a highly aggregated mAb sample had similar, but not identical, aggregate levels with the two techniques (29.8 % by SEC and 26.0 % by AF4) (32).

A limited number of studies have evaluated all three methods (SEC, SV-AUC, and AF4) simultaneously. In one such study, all three methods were used to examine aggregate contents of mAb samples stressed under a variety of different conditions (47). For unstressed material, the aggregate content by SEC was found to be 0.4%, while it was determined to be 4.2% by AUC, a difference of more than an order of magnitude (47). The aggregate content found using AF4 was 2.4%, a value in between those found with the other two methods. However, when the samples were subjected to heating, the aggregate content levels became much more similar across all three methods (values ranged from 15% to 17% aggregate). It is possible that some of the differences seen between the various methods could be attributed to the shape of the aggregates, because shape, as with end-to-end assembly, affects the mechanism of separation and how the aggregates are sized by each method. Finally, the lower precision of AF4 compared with SEC was another likely cause for the difference in aggregate levels measured by the two methods.

Conclusion
The precision and accuracy of AUC and AF4 have evolved to the point where they can be used as orthogonal methods to SEC to determine the aggregate content of protein pharmaceuticals. Other orthogonal methods have been explored, but their use is not as routine or the methodologies not as robust as AUC and AF4. As the case studies in this review illustrate, AUC and AF4 may provide different estimates of the extent of aggregation relative to SEC. While the reasons for the differences may not always be obvious, they do allow one to obtain additional insight into the stability of a product as well as serve as added methods to assess purity and quality. As regulatory agencies continue to request information above and beyond SEC data, there will be an increasing need for these orthogonal methods to be refined and employed. Finally, as more case studies appear in the literature, the context for proper interpretation of the results will become more clear.

About the Authors
Ryan R. Manning is president, Great Lakes Bio Design LLC; Ryan E. Holcomb is scientist, Legacy BioDesign LLC and research associate, department of chemistry, Colorado State University; Glenn A. Wilson is president, West Coast BioDesign LLC; Charles S. Henry is professor of chemistry, department of chemistry, Colorado State University; and Mark Cornell Manning is chief scientific officer, Legacy BioDesign LLC and affiliate faculty, department of chemistry, Colorado State University (corresponding author). Email: manning@legacybiodesign.com; Phone: 001-970-231-9744.

References
1. J.F. Carpenter, T.W. Randolph, and W. Jiskoot et al., J. Pharm. Sci. 99 2200-2208 (2010).
2. K. Štulik, V. Pacáková, and M. Tichá, J. Biochem. Biophys. Methods. 56 1-13 (2003).
3. T. Arakawa, D. Ejima, and T. Li et al., J. Pharm. Sci. 99 1674-1692 (2010).
4. P. Hong, S. Koza, and E.S.P. Bouvier, J. Liq. Chrom. Rel. Technol. 35 2923-2950 (2012).
5. A.S. Rosenberg, AAPS Journal, 8 E501-E507 (2006).
6. J. Carpenter, B. Cherney, and A. Lubinecki, et al., Biologicals 38 602-611 (2010).
7. H.C. Mahler, W. Friess, and U. Grauschopf et al., J. Pharm. Sci. 98 2909-2934 (2009).
8. J.S. Philo, AAPS Journal 8 E564-E571 (2006).
9. J.S. Philo, Curr. Pharm. Biotechnol. 10 358-372 (2009).
10. S. Grotefend, L. Kaminski, and S. Wroblewitz et al., J. Pharm. and Biomed. Anal. 71:127-138 (2012).
11. E. Watson and W.C. Kenney, J. of Chromatogr. 436 289-298 (1988).
12. K. Tsumoto, D. Ejima, and A.M. Senczuk et al., J. Pharm. Sci. 96 1677-1690 (2007).
13. D. Ejima, R. Yumioka, and T. Arakawa et al., J. Chromatogr. 1094 49-55 (2005).
14. R. Yumioka, H. Sato, and H. Tomizawa et al., J. Pharm. Sci. 99 618-620 (2010).
15. T. Arakawa, Y. Kita, and D. Ejima et al., Protein Pept. Lett. 15 544-555 (2008).
16. V.E. Klyushnichenko and A.N. Wulfson, Pure Appl. Chem. 65 2265-2272 (1993).
17. A. Diress, B. Lorbetskie, and L. Larocque et al., J. Chromatogr. 1217 3297-3306 (2010).
18. A. Hawe, W. Friess, and M. Sutter et al., Anal. Biochem. 378 115-122 (2008).
19. S.R. Gunturi, I. Ghobrial, and B. Sharma, J. Pharm. and Biomed. Anal. 43 213-221 (2007).
20. M.D. Bond, M.E. Panek, and Z. Zhang et al., J. of Pharm. Sci. 99 2582-2597 (2010).
21. J. Liu, J.D. Andya, and S.J. Shire, AAPS Journal 8 E580-E589 (2006).
22. J.P. Gabrielson, K.K. Arthur, and B.S. Kendrick et al., J. Pharm. Sci. 98 50-62 (2009).
23. P. Schuck, Biophys. J. 78, 1606-1619 (2000).
24. P. Schuck, M.A. Perugini, and N.R. Gonzales et al., Biophys. J. 82 1096-1111 (2002).
25. J.P. Gabrielson and K.K. Arthur. Methods 54 83-91 (2011).
26. J.L. Cole, J.W. Lary, and T.P. Moody et al., Methods Cell Biol. 84 143-179 (2008).
27. A. Pekar and M. Sukumar, Anal. Biochem. 367 225-237 (2007).
28. H. Hughes, C. Morgan, and E. Brunyak et al., AAPS Journal 11 335-341 (2009).
29. S.A. Berkowitz, AAPS Journal 8 E590-E605 (2006).
30. J.C. Giddings, Science, 260 1456-1465 (1993).
31. K.-G. Wahlund, J. of Chromatogr. 1287 97-112 (2013).
32. S. Cao, J. Pollastrini, and Y. Jiang, Curr. Pharm. Biotechnol. 10 382-390 (2009).
33. J. Pollastrini, L.O. Narhi, and Y. Jiang et al., “Field-flow fractionation in therapeutic protein development,” in Field-Flow Fractionation in Biopolymer Analysis, K.D. Caldwell and S.K.R. Williams, Eds., (Spring-Verlag, Vienna, 2012), pp. 73-88.
34. B. Demeule, C. Palais, and G. Machaidze et al., mAbs 1 142-150 (2009).
35. G.E. Kassalainen and S.K.R. Williams, “Assessing protein-ultrafiltration membrane interactions using flow field-flow fractionation,” in Field-Flow Fractionation in Biopolymer Analysis, K.D. Caldwell and S.K.R. Williams, Eds., (Spring-Verlag, Vienna, 2012), pp. 23-26.
36. J.E.G.J. Wijnhoven, J.-P. Koorn, H. Poppe, and W.T. Kok, J. of Chromatogr. 732 307-315 (1996).
37. G. Yohannes, M. Jussila, and K. Hartonen et al., J. of Chromatogr. 1218 4104-4116 (2011).
38. A. Hawe, S. Romeijn, and V. Filipe at al., J. of Pharm. Sci. 101 4129-4139 (2012).
39. R. Lang, G. Winter, and L. Vogt et al., Drug Dev. Ind. Pharm. 35 83-97 (2009).
40. C. Arfvidsson and K.-G. Wahlund, J. Chromatogr. 1011 99-109 (2003).
41. L. Zhao, L. Li, and G.Q. Liu et al., Food Res. Int. 53 409-416 (2013).
42. W. Fraunhofer, G. Winter, and C. Coester, Anal. Chem. 76 1909-1920 (2004).
43. R. Pavišić, I. Dodig, and A. Horvatić et al., Eur. J. Pharm. Biopharm. 76 357-365 (2010).
44. A. Litzen, J.K. Walter, and H. Krischollek et al., Anal. Biochem. 212 469-480 (1993).
45. A.J. Freitag, K. Wittmann, and G. Winter et al., LC-GC Europe 24, 134+ (2011).
46. Y. Jiang, M.E. Miller, and P. Li et al., Amer. Lab. 32 98-108 (2000).
47. J.P. Gabrielson, M.L. Brader, and A.H. Pekar et al., J. Pharm. Sci. 96 268-279 (2007).

Article DetailsBioPharm International
Vol. 27 No. 12
Pages: 32-39

Citation: When referring to this article, please cite it as R. Manning et al., "Review of Orthogonal Methods to SEC for Quantitation and Characterization of Protein Aggregates," BioPharm International27 (12) 2014.