Measurements of Monoclonal Antibody Self-Association Are

Apr 5, 2016 - ... Rensselaer Polytechnic Institute, Troy, New York 12180, United States. ‡ Contrafect Corporation, Yonkers, New York 10701, United S...
0 downloads 0 Views 2MB Size
Subscriber access provided by UNIV OF PITTSBURGH

Article

Measurements of monoclonal antibody self-association are correlated with complex biophysical properties Steven B Geng, Michael Wittekind, Adam Vigil, and Peter M. Tessier Mol. Pharmaceutics, Just Accepted Manuscript • DOI: 10.1021/acs.molpharmaceut.6b00071 • Publication Date (Web): 05 Apr 2016 Downloaded from http://pubs.acs.org on April 10, 2016

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Molecular Pharmaceutics is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 29

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Molecular Pharmaceutics

Measurements of monoclonal antibody self-association are correlated with complex biophysical properties Steven B. Geng,1 Michael Wittekind,2 Adam Vigil,2 Peter M. Tessier1,* 1

Center for Biotechnology & Interdisciplinary Studies, Isermann Dept. of Chemical & Biological Engineering, Rensselaer Polytechnic Institute, Troy, NY 12180 2 Contrafect Corporation, Yonkers, NY 10701

Successful development of monoclonal antibodies (mAbs) for therapeutic applications requires identification of mAbs with favorable biophysical properties (high solubility and low viscosity) in addition to potent bioactivities. Nevertheless, mAbs can also display complex, non-conventional biophysical properties that impede their development, such as formation of soluble aggregates and subvisible particles as well as non-specific interactions with various types of surfaces such as nonadsorptive chromatography columns. Here we have investigated the potential of using antibody selfinteraction measurements obtained via affinity-capture self-interaction nanoparticle spectroscopy (ACSINS) at dilute concentrations (0.01 mg/mL) for ranking a panel of 12 mAbs in terms of their expected biophysical properties at higher concentrations (1-30 mg/mL). Several mAb properties (solubility, % monomer, size-exclusion elution time and % recovery) displayed modest correlation with each other, as some mAbs with deficiencies in one or more properties (e.g., solubility) failed to show such deficiencies in other properties (e.g., % monomer). The ranking of mAbs in terms of their level of self-association was correlated with their solubility ranking. However, the correlation was even stronger between the average ranking of the four biophysical properties and the AC-SINS measurements. This finding suggests that weak self-interactions detected via AC-SINS can manifest themselves in different ways and lead to complex biophysical properties. Our findings highlight the potential for using highthroughput self-interaction measurements to improve the identification of mAbs that possess a collection of excellent biophysical properties without the need for cumbersome analysis of each individual property during early candidate selection. Keywords: mAb; solubility; aggregation; viscosity; self-association; size-exclusion chromatography. Running title: Screening complex biophysical properties of mAbs *Corresponding author: [email protected]

1

ACS Paragon Plus Environment

Molecular Pharmaceutics

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 2 of 29

INTRODUCTION The physical properties of drugs have long played a key role in their success as therapeutics. Lipinski's rule of five specifies that (among other things) small molecules should not be too hydrophobic, as defined by the log of the octanol-water partition coefficient not being greater than five.1 The physical properties of biomolecules, which are harder to define and require alternative measures, also play a key role of their success as therapeutics.2-5 Conventional biophysical measures of proteins include solubility, aggregation, viscosity and self-association.6-11 The fact that self-association is a key determinant of protein solubility and viscosity has led to many attempts to correlate self-interaction measurements at dilute protein concentrations to solubility and viscosity measurements at elevated concentrations.11-22 However, proteins such as monoclonal antibodies (mAbs) also display a range of non-conventional biophysical properties such as their propensity to interact non-specifically with off-target molecules and various surfaces (polyspecificity)23-26 as well as to stick to non-adsorptive chromatography columns (i.e., size-exclusion columns) that leads to poor recoveries and abnormally long elution times.25,

27

The

propensity of antibodies to interact non-specifically with different types of molecules and surfaces has been correlated with low solubility23, 26, 27 and fast antibody clearance.24, 25 Thus, methods for identifying antibodies with favorable biophysical properties early in antibody discovery that address both standard and non-standard biophysical attributes would be helpful for improving the selection of mAb variants that can be developed into effective antibody therapeutics. We have developed a high-throughput method for measuring antibody self-association referred to as affinity-capture self-interaction nanoparticle spectroscopy (AC-SINS).13 This method involves coating gold nanoparticles with polyclonal antibodies specific for human antibodies and using these conjugates to immobilize human mAbs. The optical properties of the conjugates, including the wavelength of maximum absorbance (plasmon wavelength), are related to the interparticle separation distances. mAbs with high propensity to self-associate show reduced interparticle separation distances and red-shifted plasmon wavelengths.12,

13, 25, 28-32

We find that this approach can be used to identify highly soluble

mAbs during early antibody discovery using extremely dilute (0.001-0.01 mg/mL) and unpurified antibody samples.12 In this work, we have evaluated the self-association behavior for a panel of 12 mAbs using ACSINS. Notably, we identified multiple mAbs with significant positive plasmon shifts (indicative of attractive self-interactions) that were soluble at moderate antibody concentrations (20-35 mg/mL). We posited that the attractive self-interactions detected by AC-SINS may manifest themselves in other ways 2

ACS Paragon Plus Environment

Page 3 of 29

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Molecular Pharmaceutics

in addition to reduced solubility, including deficiencies in other conventional biophysical properties such as aggregation (low % monomer) or non-conventional ones such as non-specific interactions with various surfaces (e.g., non-adsorptive chromatography columns). This is based on the possibility that the molecular forces (e.g., hydrophobic and electrostatic interactions) mediating antibody self-association can in some cases also mediate non-specific antibody interactions with off-target molecules and surfaces. Here we have tested this hypothesis by comparing the ranking of mAbs with the lowest selfassociation propensity (as judged by AC-SINS) with the rankings obtained by evaluating multiple conventional and non-conventional biophysical properties.

RESULTS AC-SINS measurements of mAb self-association in extremely dilute solutions

We first sought to evaluate the self-association behavior for a panel of 11 mAbs (human IgG1s) that have not been characterized previously with the goal of identifying the mAbs with the most favorable biophysical properties. As a control, we also evaluated a mAb (CNTO607, herein referred to as mAb L) that displays significant self-association and low solubility.14,

33

Due to limiting quantities (a few

milligrams) of each mAb with which to perform these and related biophysical studies (solubility and size-exclusion chromatography measurements), we chose to use AC-SINS to evaluate antibody selfassociation given the low sample requirements. To maximize the sensitivity of these experiments, we used gold particles (20 nm) coated with polyclonal goat anti-human Fc antibody to capture human mAbs at a saturating mAb concentration (0.01 mg/mL; Figure 1).13, 28 The absorbance spectra of the resulting gold conjugates revealed a wide range of behaviors (Figure 2). The plasmon wavelength was 533 nm for the control conjugates in the absence of mAb, which is consistent with previous reports.12, 13, 28, 29 The 12 mAbs displayed a range of plasmon wavelengths from 532 nm to 543 nm. The mAbs with the lowest plasmon wavelengths were mAbs E (532 nm) and A (533 nm). mAb L (CNTO607) displayed the highest plasmon wavelength (543 nm), which is consistent with its high propensity to self-associate.14, 33 Multiple other mAbs also displayed high plasmon wavelengths (541 nm for mAbs F and J). To better differentiate between the behaviors of the 12 mAbs, we measured the plasmon wavelengths at different densities of immobilized mAb (Figure 3A). To vary the amount of immobilized mAb, we co-adsorbed each mAb with different amounts of human polyclonal antibody (pAb) at a constant total human antibody concentration (0.01 mg/mL). We reasoned that the sensitivity for discriminating between different mAbs would be enhanced at moderate loadings due to the saturation of the plasmon wavelength at high loadings. Indeed, we find that mAb L displays much higher plasmon 3

ACS Paragon Plus Environment

Molecular Pharmaceutics

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 4 of 29

wavelengths than the other mAbs at low mAb loadings (10-50% mAb). At high mAb loadings (50-100% mAb), the plasmon response of mAb L saturates, while that of the other mAbs is generally linear. To enable comparison between the plasmon wavelength measurements for different mAbs, we calculated the plasmon shifts for each measurement (relative to the control value for 0% mAb) and reported the average plasmon shift for each mAb (Figure 3B). These results reveal that mAbs L, J, C, F, D and I are the most associative antibodies in PBS at dilute concentrations (0.01 mg/mL) and are expected to display suboptimal biophysical properties when evaluated at higher concentrations (1-35 mg/mL). mAb solubility and size-exclusion chromatography analysis

We next sought to evaluate the relationship between self-association and multiple biophysical properties for the 12 mAbs. In particular, size-exclusion chromatography was used to measure the tendency of the mAbs to bind to the inert porous matrix of the column, which results in low recoveries and/or abnormally long elution times. In the absence of mAb-column interactions, the mAbs are expected to elute at similar times because they have similar molecular weights and structures. However, we find that the 12 mAbs display a wide range of elution times and profiles (Figure 4). Several mAbs showed delayed elution times (19-32 min) compared to the elution time for a human polyclonal IgG control (18 min). Notably, mAbs C, F, and J elute the latest (27-32 min) and at similar times as a small control analyte (acetone, 28 min), revealing that these mAbs interact with the chromatography matrix. mAbs B, G, H and I displayed modestly delayed elution times (20-25 min). mAbs C and D displayed significant aggregation, as evidenced by multiple peaks in the chromatograms and low % monomer (63% and 41%, respectively). Interestingly, mAb L (which has the lowest solubility) showed a comparable elution time to the polyclonal antibody control and was mostly monomeric. We also evaluated the fractional recovery of each mAb that was injected into the size-exclusion column (Table 1). mAb L displayed the poorest recovery (30%). Three other mAbs also displayed recoveries less than 50%, which were mAbs C (49%), F (46%) and D (38%). Interestingly, the reduced recovery of these mAbs is not well correlated with their elution times or fractional monomer content. The mAbs with the best recoveries were mAb A (99.5%), G (93%) and B (92%). Each of these mAbs was also primarily monomeric and displayed modestly delayed elution times. We also evaluated the relationship between self-association and solubility for these mAbs given the expected correlation between these two properties (Figure 5). The limited amount of each mAb only allowed us to concentrate them from 1 mg/mL (0.5 mg/mL for mAb L) to a maximum theoretical concentration of ~33 mg/mL (~17 for mAb L) using centrifugal concentrators. Most mAbs were soluble

4

ACS Paragon Plus Environment

Page 5 of 29

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Molecular Pharmaceutics

at these intermediate concentrations. The exceptions were mAb D and mAb L, which displayed reduced solubility (16 mg/mL for mAb D and 6 mg/mL for mAb L). The self-association and biophysical measurements are summarized in Figure 6 and Table 1. The strongest correlations were observed for solubility versus plasmon shift (Figure 6A, R2=0.68) and % recovery versus plasmon shift (Figure 6B, R2=0.64), while elution time (Figure 6C, R2=0.039) and % monomer (Figure 6D, R2=0.044) displayed little correlation with the plasmon shift measurements. Nevertheless, mAbs with the smallest plasmon shifts typically possess the best biophysical properties, while those with the largest plasmon shifts display one or more suboptimal biophysical properties (data is summarized in Table 1). This is also highlighted in Figure 7, which demonstrates that most of the mAbs with the best biophysical rankings for each property (ranking of 1 is better than average, 2 is average, 3 is worse than average and 4 is more than one standard deviation worse than average) display the best plasmon shift rankings (ranking of 1 is better than one standard deviation of the average, 2 is better than average, 3 is average, 4 is worse than average and 5 is worse than one standard deviation of the mean), although some exceptions are observed. To evaluate these rankings more quantitatively, we evaluated their rank correlation with each other as well as with the plasmon shift rankings (Table 2). Spearman’s rank correlation coefficients (1 or -1 for perfect direct and inverse correlations respectively, 0 for no correlation) and their corresponding pvalues were used to judge these correlations. The strongest rank correlations were for % recovery and plasmon shift (Spearman’s coefficient of 0.89, p-value of 0.0001), solubility and % recovery (Spearman’s coefficient of 0.75, p-value of 0.005), and solubility and plasmon shift (Spearman’s coefficient of 0.64, p-value of 0.026). In contrast, the weakest correlations were observed for elution time and % monomer (Spearman’s coefficient of -0.04, p-value of 0.89), and elution time and % recovery (Spearman’s coefficient of 0.14, p-value of 0.66). Nevertheless, we noticed that mAbs with high self-association propensity did not display defects in all four biophysical properties. For example, mAbs F, C and J display good solubility at the concentrations evaluated (24-27 mg/mL) despite the fact that they display significant plasmon shifts (3 of the 4 most associative mAbs) and deficiencies in other biophysical properties (Table 1). Thus, we wondered if mAb self-association would be better correlated with the average ranking of mAbs based on all four of their biophysical properties (elution time, solubility, % monomer and % recovery; Figure 7 and Table 2). Indeed, the mAbs with the best average rankings based on the four biophysical properties (mAbs A, G, E, B, H, K and I) show the lowest plasmon shift rankings (Figure 8). The Spearman’s coefficient for the correlation between mAb self-association and the average biophysical ranking is 0.84 5

ACS Paragon Plus Environment

Molecular Pharmaceutics

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 29

(p-value of 0.0006), which is similar to the best correlation between self-association and any single biophysical property (Spearman’s coefficient of 0.89 for the correlation between % recovery and selfassociation). Collectively, these results suggest that the degree of mAb self-association detected by ACSINS may manifest itself in different ways in terms of conventional and non-conventional biophysical properties. The goal of our study was to test the ability of AC-SINS to identify mAbs with the best biophysical properties. Thus, we wondered how the ranking of mAbs is influenced by the solution conditions (PBS, pH 7.4) that we used to perform the AC-SINS assay. To evaluate this question, we also performed ACSINS at a different solution condition at lower pH and without salt (pH 6, 10 mM citrate; Figure 9). Notably, there is a strong correlation between mAbs that display high self-association in PBS and those that display high self-association in the citrate solution (Spearman’s coefficient of 0.97, p-value of 1.29·10-7). This finding suggests that intrinsic properties of the 12 mAbs (i.e., their sequence and structure) are the primary determinants of their level of self-association.

DISCUSSION Multiple previous studies have established links between antibody self-association and biophysical properties such as solubility and viscosity.11-22 For example, we reported that AC-SINS measurements of mAb self-association for extremely dilute antibody solutions (µg/mL) are correlated with light scattering and solubility measurements at three to five orders of magnitude higher antibody concentrations.12 These and other findings13,

14, 21, 22, 25, 33

are generally consistent with the findings

presented here demonstrating that self-association and solubility are inversely correlated (Figure 6A). Additionally, measurements of dilute antibody self-association (mg/mL) have also been correlated with high-concentration viscosity measurements (hundreds of mg/mL).11, 16, 18, 20 14 We found that the level of mAb self-association detected by AC-SINS is useful for identifying antibodies at risk for displaying deficiencies in one or more conventional and/or non-conventional biophysical properties. It is well known that antibodies (and other proteins) can interact with sizeexclusion chromatography matrices and elute at times that are poorly correlated with their size.25, 27, 34, 35 The origin of this behavior may be due in part to attractive electrostatic interactions between antibodies and chromatographic particles, as it is common practice to use elevated salt concentrations to reduce such non-specific interactions.36 However, it appears that non-electrostatic interactions also contribute to abnormally long retention times, as some antibodies also interact with such columns even at elevated salt concentrations.27, 34 6

ACS Paragon Plus Environment

Page 7 of 29

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Molecular Pharmaceutics

Our finding that chromatographic properties such as % recovery from size-exclusion chromatography columns are correlated with antibody self-association is generally consistent with a recent study that also used AC-SINS to evaluate antibody self-association.25 Interestingly, these investigators observed much stronger correlations between AC-SINS measurements and mAb elution time (Spearman’s coefficient of 0.76, p-value of 0.001) than we observed (Spearman’s coefficient of 0.42, p-value of 0.17). Nevertheless, our observed correlations between AC-SINS ranking and the combined ranking of multiple chromatographic properties (elution time and % monomer, Spearman’s coefficient of 0.51, p-value of 0.09) are generally consistent with those previously reported when correlating AC-SINS with elution time (Spearman’s coefficient of 0.76, p-value of 0.001), % monomer (absolute value of Spearman’s coefficient of 0.43, p-value of 0.095) or the combination of the two properties (Spearman’s coefficient of 0.60, p-value of 0.04; calculated by ranking each property from best to worst).25 Although the % mAb recovery after size-exclusion analysis was not reported for the previous study, we find even stronger correlations between the self-association ranking and the combined ranking of three chromatographic properties including % recovery in addition to retention time and % monomer (Spearman’s coefficient of 0.84, p-value of 0.0006). This correlation is similar to the best correlations we observed that include solubility measurements in addition to various chromatographic measurements (Spearman’s coefficients of 0.82-0.85; Table 2). Our finding that antibody self-interaction measurements are correlated with size-exclusion chromatography measurements raises the possibility that they may also be correlated with other types of non-conventional chromatographic measurements. Although we did not explore this possibility in this study, others have evaluated correlations between AC-SINS and cross-interaction chromatography previously.25, 29 Cross-interaction chromatography involves measuring the elution time of mAbs from a column with particles coated with immobilized human polyclonal antibody.26 Cross-interaction chromatography measurements have been correlated with antibody solubility, which may be due to the fact that mAb-polyclonal antibody cross-interactions sample some of the same types of interactions that occur for mAb self-association. As expected, the correlation between AC-SINS and cross-interaction chromatography is imperfect (Spearman’s coefficient of 0.41, p-value of 0.11),25 which is likely due to the specific nature of some types of mAb self-interactions that are missed by the cross-interaction analysis. It is also notable that chromatographic measurements obtained with non-conventional size-exclusion columns have been used for ranking antibodies in terms of their biophysical properties (e.g., solubility).25-27,

33

A notable example is the use of standup monolayer adsorption chromatography, a 7

ACS Paragon Plus Environment

Molecular Pharmaceutics

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 8 of 29

size-exclusion method that uses a Zenix column (Sepax Technologies) presenting a surface coated with a monolayer of hydrophobic molecules with hydrophilic head groups (the specific chemistry is proprietary).27 The investigators demonstrated that antibodies with low solubility display abnormally long elution times for the Zenix column relative to a more traditional size-exclusion column (such as the TSKgel column that we used). This suggests that chromatography columns presenting mixed modes of interaction may be useful for detecting non-specific mAb interactions that are correlated with antibody biophysical properties. Collectively our findings demonstrate the utility of the AC-SINS methodology for identifying mAbs with favorable and suboptimal biophysical properties. One of the main strengths of AC-SINS is its efficiency in terms of time and protein consumption. AC-SINS measurements can be performed in 384well plates in a parallel manner (i.e., tens to hundreds of measurements can be performed at the same time), they require minuscule amounts of antibody (