A Comparison Between Emerging and Current Biophysical Methods for the Assessment of Higher-Order Structure of Biopharmaceuticals

Journal Pre-proof A comparison between emerging and current biophysical methods for the assessment of higher order structure of biopharmaceuticals Jie Wen, Dipanwita Batabyal, Nicholas Knutson, Harrison Lord, Mats Wikström PII:

S0022-3549(19)30663-X

DOI:

https://doi.org/10.1016/j.xphs.2019.10.026

Reference:

XPHS 1753

To appear in:

Journal of Pharmaceutical Sciences

Received Date: 8 August 2019 Revised Date:

14 October 2019

Accepted Date: 15 October 2019

Please cite this article as: Wen J, Batabyal D, Knutson N, Lord H, Wikström M, A comparison between emerging and current biophysical methods for the assessment of higher order structure of biopharmaceuticals, Journal of Pharmaceutical Sciences (2019), doi: https://doi.org/10.1016/ j.xphs.2019.10.026. This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2019 Published by Elsevier Inc. on behalf of the American Pharmacists Association.

A comparison between emerging and current biophysical methods for the assessment of higher order structure of biopharmaceuticals

Jie Wen*, Dipanwita Batabyal, Nicholas Knutson, Harrison Lord, and Mats Wikström*

Amgen Inc, Higher Order Structure, Attribute Sciences, Thousand Oaks, CA 91320, United States.

Running Title: Biophysical methods for the assessment of HOS

*Corresponding authors: Amgen, Inc. Thousand Oaks, CA 91320, United States. Email address: [email protected] (Mats Wikström) and [email protected] (Jie Wen)

1

ABSTRACT:

The higher order structure (HOS) of protein therapeutics is a critical quality

attribute (CQA) directly related to their function1. Traditionally, the HOS of protein therapeutics has been characterized by methods with low to medium structural resolution such as Fourier transform infrared, circular dichroism, and intrinsic fluorescence spectroscopy, and differential scanning calorimetry2. Recently, high-resolution nuclear magnetic resonance (NMR) methods have emerged as powerful tools for HOS characterization3-5. NMR is a multi-attribute method with unique capabilities to provide information about all the structural levels of proteins in solution. We have in this study, compared 1 D 1H Profile NMR with the established biophysical methods for HOS assessments using a set of blended samples of the monoclonal antibodies belonging to the subclasses IgG1 and IgG2. The study shows that Profile NMR can distinguish between most sample combinations (93%), DSC can differentiate 61% of the sample combinations, and NUV CD can differentiate 52% of the sample combinations, whereas no significant distinction could be made between any samples using FTIR or intrinsic fluorescence. Our data therefore shows that NMR has superior ability to address differences in HOS, a feature that could be directly applicable in comparability and similarity assessments.

Keywords: monoclonal antibody; higher order structure; protein structure; secondary structure; tertiary

structure;

biophysical

methods;

comparability;

characterization.

2

biosimilarity;

biopharmaceutical

INTRODUCTION The higher order structure (HOS) of proteins encompasses secondary, tertiary, and quaternary structure and signifies a critical quality attribute directly related to the structural integrity and function of therapeutic proteins. The characterization of HOS represents a significant challenge for biopharmaceuticals and has traditionally been performed using low to medium-resolution biophysical methods including Fourier Transform Infrared spectroscopy (FTIR), circular dichroism (CD) spectroscopy, intrinsic fluorescence spectroscopy, and differential scanning calorimetry

(DSC)2.

With

the

increasing

interest

in

different

protein

modalities

in

biopharmaceutical development and the rapidly expanding area of biosimilar development, there is a growing need for new analytical methods with better specificity than the methods currently available.

During the development of protein therapeutics, the innovator product will go through multiple change processes, in which it is required to show that any process related drug product variations are within the acceptable criteria, and therefore considered comparable. In a similar fashion, it is required to show similarity between the biopharmaceutical reference product and biosimilar’s. Recently, the application of NMR for the assessment of HOS has been suggested as a technology with the potential to more accurately assess differences in HOS as compared to established methods.

This technology, referred to as Profile NMR, is based on a one-

dimensional NMR method in which the strong signals from excipients are efficiently suppressed leaving a spectrum of the protein product only3-4. In addition to the 1D NMR method applied in this study, a 2D 1H-13C method was recently introduced5, that shows great promise for the higher order structure assessment of monoclonal antibodies6. Finally, mass spectrometric methods, such as hydrogen-deuterium exchange methods, has also gained considerable interest for the assessment of biopharmaceuticals7-8.

3

In this study we have compared the established methods for the assessment of HOS for biopharmaceuticals with 1D Profile NMR. The results show that Profile NMR can distinguish between most sample combinations, DSC and NUV CD can differentiate between approximately half of the sample combinations, whereas no significant distinction could be made between any samples using FTIR, or intrinsic fluorescence. Our findings therefore exemplify the superiority of NMR in the assessment of higher order structural attributes.

MATERIAL AND METHODS Sample preparation. The test solutions were prepared from 100 mg/mL monoclonal antibodies belonging to the IgG1 and IgG2 subclasses in the formulation buffer 10 mM sodium acetate buffer, 9% sucrose at pH 5.2. The sequence identity between the IgG1 and IgG2 molecules used in this study is 95%9. The IgG1 molecule has been shown to contain glycosylation’s on N302, whereas the IgG2 molecule harbor glycosylation modifications on N298 (data not shown). Stock solutions of IgG1 and IgG2 were prepared at 50 mg/mL in the same formulation buffer. All blended samples were prepared at a total protein concentration of 50 mg/mL in the same formulation conditions as above. The sample set is shown in Table 1.

Fourier-Transformed Infrared Spectroscopy (FTIR) Method The FTIR spectra of the protein solutions were recorded at room temperature using a Bruker Vertex 70. The experiments were performed at a total protein concentration of 50 mg/mL Fourier transform infrared spectrometer, equipped with an MCT detector. The samples were measured directly in the Aqua Spec Cell that employs CaF2 windows separated by a 6-um spacer. For each spectrum, a 256-scan interferogram was collected in a single beam mode, with a 4 cm-1 resolution. Reference spectra were recorded under identical conditions with the formulation buffer in the cell. The reference spectra for buffer blank were subtracted from the

4

protein spectra, according to previously established criteria10. The second-derivative spectrum was calculated with Bruker OPUS software. The final spectrum was smoothed with a 9-point function to remove white noise.

Analysis of FTIR data Extensive work on mAbs have been done previously to establish universal acceptance criteria for similarity and comparability analysis. Currently we have used two methods for the statistical analysis of the comparability data namely spectral similarity and weighted spectral difference1011

. Spectral similarity was quantitatively determined using the Thermo OMNIC software QC

compare function. Briefly, the similarity between any two spectra/profiles is obtained using the Thermo OMNIC software QC Compare function. The spectral similarity of the second-derivative FTIR spectra in the amide I region (1700 - 1600 cm-1) was calculated by comparing each second-derivative spectrum of each samples to the spectrum of a control sample (e.g. reference standard or other native control collected within the same experimental run) at 4 cm-1 resolution. The result is a value between 0 and 100%, which indicates how well the sample spectrum matches the reference spectrum. A value of 100% indicates the spectra are identical. From our internal studies, Lower tolerance limit with 95% confidence and 95% coverage is 95%11. WSD (weighted spectral difference) is a Euclidean distance weighted by relative signal magnitude. It is an enhanced version of SD (spectral difference), with an additional factor to magnify differences in spectral regions with higher signal intensity (both negative and positive signals). WSD is calculated using the absolute normalized absorbance value of the Amide I band (1700 to 1600 cm-1 region). Comparability acceptance criteria using the WSD with upper tolerance interval with 95% confidence and >99% coverage is 0.04812.

Far Ultraviolet Circular Dichroism Spectroscopy (FUV CD) Method

5

For modalities with prominently helical secondary structure, and/or other modalities with protein concentrations less than/equal to 10 mg/mL, FUV circular dichroism spectroscopy represents an alternative to FTIR for assessing the secondary structure. For the assessment of secondary structure, the spectra were collected in the spectral region (190-250 nm) where the peptide bond, alpha helix, and beta sheets give rise to circular polarized light signals indicating a secondary structure.

FUV CD spectra were measured using an Applied Photophysics

Chirascan spectropolarimeter (Applied Photophysics Ltd, Leatherhead, UK) at ambient temperature using cuvettes with a path length of 0.01cm. Spectra are analyzed at a step size of 0.5 nm, a bandwidth of 1 nm, a response time of 2 seconds, with a 4-scan average. All mAb samples were diluted to approximately 0.6 mg/mL with product buffer prior to measurements. Each sample was measured in triplicate and buffer blanks were subtracted prior to data analysis. The resulting spectra were corrected for concentration and contributions from buffer and reported as mean residue molar ellipticity, using an extinction coefficient of 1.51 and a mean residue weight of 109 Da. Spectral similarity was quantitatively determined using the Thermo OMNIC software QC compare function (Waltham, Massachusetts). Analysis of FUV CD data Extensive work on mAbs have been done to establish universal acceptance criteria for similarity and comparability analysis for FUV CD. Currently we have used two methods for the statistical analysis of the comparability data namely spectral similarity and weighted spectral difference (as described in detail in the previous section for FTIR)11-12. After obtaining average sample spectra, they were then corrected for concentration, mean residue weight, and cell pathlength to normalize the spectra. One run was randomly chosen and designated as the reference sample, and its spectrum as the reference spectrum. All spectra, including the reference spectrum, were compared against this to obtain the similarity values. The wavelength range used for the analysis was from 240 - 200 nm.

6

The result is a value between 0 and 100%, which indicates how well the sample spectrum matches the reference spectrum. A value of 100% indicates the spectra are identical. From our internal studies, Lower tolerance limit with 95% confidence and 95% coverage is 95%. Weighted spectral difference (WSD) is used to compare a sample FUV CD spectrum to an appropriate reference. Calculations can be performed with Microsoft Excel software version 2003 or newer, using x-y data from the MRME spectra in Pro-Data Viewer software or MRE spectra in Spectra Manager software. Wavelength range to calculate WSD for FUV CD is from 240-200 nm and the acceptance criteria based on our extensive studies for WSD is 1600 deg.cm2.decimol-1.

Near Ultraviolet Circular Dichroism Spectroscopy (NUV CD) Method Tertiary structure was assessed by NUV CD spectroscopy. For the assessment of tertiary structure, the spectra were collected in the spectral region (250-340 nm) where the disulfide bonds and aromatic side chains: tryptophan, tyrosine and phenylalanine give rise to circular polarized light signals indicating tertiary structure. NUV CD measurements were made on an Applied Photophysics Chirascan spectropolarimeter (Applied Photophysics Ltd, Leatherhead, UK) ambient temperature using cuvettes with a path length of 1 cm. Spectra are analyzed at a step size of 0.5 nm, a bandwidth of 1 nm, a response time of 2 seconds, with a 4-scan average. All mAb samples were diluted to approximately 0.6 mg/mL with product buffer prior to measurements. Each sample was measured in triplicate and buffer blanks were subtracted prior to data analysis. The resulting spectra were corrected for concentration and contributions from buffer and reported as mean residue molar ellipticity, using an extinction coefficient of 1.51 and a mean residue weight of 109 Da. Spectral similarity was quantitatively determined using the Thermos OMNIC software QC compare function (Waltham, Massachusetts).

7

Analysis of NUV CD Extensive work on mAbs have been done to establish universal acceptance criteria for similarity and comparability analysis for NUV CD. Currently we have used two methods for the statistical analysis of the comparability data namely spectral similarity and weighted spectral difference (as described in detail in the previous section for FTIR)11-12. For NUV CD after obtaining sample spectra, samples were then corrected for concentration, mean residue weight, and cell pathlength to normalize the spectra. One run was randomly chosen and designated as the reference sample, and its spectrum as the reference spectrum. All spectra, including the reference spectrum, were compared against this to obtain the similarity values. The wavelength range used for the analysis was from 320 – 250 nm. The result is a value between 0 and 100%, which indicates how well the sample spectrum matches the reference spectrum. A value of 100% indicates the spectra are identical. From our internal studies, Lower tolerance limit with 95% confidence and 95% coverage is 95%. Weighted spectral difference (WSD) is used to compare a sample NUV CD spectrum to an appropriate reference. Calculations can be performed with Microsoft Excel software version 2003 or newer, using x-y data from the MRME spectra in Pro-Data Viewer software or MRE spectra in Spectra Manager software. Wavelength range to calculate WSD for NUV CD is from 300-250 nm and the acceptance criteria based on our extensive studies for WSD is 4.3 deg.cm2.decimol-1

Intrinsic Tryptophan Fluorescence Spectroscopy (FLD) Method Intrinsic tryptophan fluorescence spectra (FLD) were obtained using an Applied Photophysics Chirascan fluorimeter (Applied Photophysics Ltd, Leatherhead, UK) at ambient temperature using cuvettes with a path length of 1 cm. Tryptophan residues were excited at 280nm and the

8

emission spectra were acquired from 285 to 450 nm using a step size of 1 nm and a 1 second integration time. All mAb samples were diluted to approximately 1 mg/mL with product buffer prior to measurements with a 5 nm bandwidth and a 10-scan average. Each sample was measured in triplicate and buffer blanks were subtracted prior to data analysis. The spectra were overlaid with each other and the similarity compared by calculating the variability of the maximum fluorescence intensity and the wavelength at the maximum fluorescence. Analysis of FLD data Extensive work on mAbs have been done to establish universal acceptance criteria for intrinsic fluorescence13-14. Currently we have used fulfillment of the two criterions for the analysis of the comparability data for intrinsic fluorescence. The wavelength of maximum emission intensity of the sample runs should be within (λmax): ±2 nm. The variation in intensity of the maximum emission (Imax) should be: <10%. Differential Scanning Calorimetry (DSC) Method Differential scanning calorimetry (DSC) was performed using MicroCal VP-capillary differential scanning calorimeter (Microcal/Malvern Instruments, Worcestershire, United Kingdom). Thermal stability was assessed by DSC measurements that were made in 6 replicates with scans completed from 10°C to 100°C using a scan rate of 60 °C/h. All mAb samples were diluted to approximately 1 mg/mL with product buffer prior to measurements.

Plotting heat

capacity as a function of temperature yields what is referred to as a DSC thermogram. An endothermic thermal transition in the sample cell will cause positive deflection of the signal. Post-run data analysis yields the thermal transition temperature (Tm). The data analysis was performed using MicroCal Origin 7 software and the statistical analysis was done using JMP 11.2 software.

9

Analysis of DSC data We have established the DSC method precision (repeatability) for mAbs by studying many mAb molecules. The repeatability of the DSC method for the CH2 and CH3 domains of the IgG1 and IgG2 molecules are ± 0.3°C. The repeatability of the DSC method for the Fab domain of the IgG1 and IgG2 molecules are ± 0.1°C. Equivalence acceptance criterion (EAC) is defined as three times the repeatability of the Tm for each domain. For example, for the Fab domain the repeatability of the Tm measurements is ± 0.1°C, and the EAC is ± 0.3°C.

Nuclear Magnetic resonance (NMR) Method All NMR experiments were performed on a Bruker Avance III spectrometer operating at 600.13 MHz equipped with a TCI cryoprobe.

The Profile spectra were recorded with the PGSTE

experiments15 were the duration and the strength of the gradients were optimized to eliminate resonances from the formulation buffer3 the acquisition time and relaxation delay were 1.36 and 2.0 s, respectively. The gradient pulses were 63 G/cm at 100% power and calibrated using the calibgs pulse sequence and 90%H2O/10%D2O sample with a samples height of 1 cm in a Shigemi tube (Shigemi Inc.). The Profile spectra (32k data points) were recorded at 313K with 1024 scans per experiment. The NMR samples were prepared as 200 mL solutions in 4 mm Shigemi tubes at 50 mg/mL (Shigemi Inc.). The raw data were first Fourier transformed in Topspin 3.5 and the spectra were further processed in Matlab (MathWorks Inc.) as described previously3. Analysis of Profile NMR data Each sample was run in 5 replicates and the signal correlations (MATLAB R2018b, ‘xcorr’ function), for each sample (auto - and cross-correlations), were determined from the ensemble average of each set of processed spectra. The standard deviation for the combined autocorrelations were then used to set a threshold for distinguishability between samples. Since the

10

auto-correlations reports on the variability between replicates of a sample, two samples are considered distinguishable if their cross-correlation score is below the auto-correlations.

RESULTS The higher order structure (HOS) was characterized using several biophysical methods. The secondary structure was assessed by FTIR and FUV CD, the tertiary structure by NUV CD and intrinsic fluorescence, and the thermal stability/tertiary structure by differential scanning calorimetry. In addition to these established methods, 1D Profile NMR was used to characterize and compare the overall higher order structure of the samples in the controlled sample set.

The assessment of secondary structure The two major methods for the assessment of the secondary structure of biopharmaceuticals are FTIR and FUV CD. FTIR requires higher product concentrations, normally above 10 mg/mL for mAbs, whereas FUV CD is used for analysis for product with lower concentration down to 0.1 mg/mL. For the analysis of the secondary structure by FTIR all samples were measured without dilution at a protein concentration of 50 mg/mL. For samples with lower concentration (< 10 mg/mL) the sensitivity of the FTIR method can become an issue. For samples of lower concentration, the alternative method to assess secondary structure of protein is FUV CD. The CD analysis is normally performed at a concentration of 0.5 mg/mL. The comparative analysis of both FTIR and FUV CD is based on the QC compare function in which a reference spectrum is selected, and all other spectra is thereafter compared to the reference. In this study, the sample corresponding to 100% IgG1 was used as the reference. Spectral similarity between any two spectra obtained using OMNIC QC software were analyzed with a 7-point smooth and a 4 nm resolution. The result is a value between 0 and 100%, where 0% indicates no similarity and 100% indicates the spectra are identical. The comparative data for the FTIR data is shown in Figure 1, and the FUV CD data is shown in Figure 3. Each

11

sample was run in triplicates and the spectral similarity of 100% corresponds to the IgG1 sample. According to the comparability criteria for FTIR and FUV CD, a spectral similarity greater that 95% indicates that the samples compared are indistinguishable. Based on spectral similarity FTIR can distinguish between the 100% IgG1 and IgG2 only in some of the runs but not consistently in all the replicates (Figure 2). However, all the other mixtures are indistinguishable using either spectral similarity or WSD for analysis. FTIR is not sensitive enough to conclude a distinguishable difference. FUV CD can distinguish between 100% IgG1 and IgG2 using spectral similarity as analysis (Figure 4), however the other mixtures are indistinguishable, and it runs into the same inconsistencies as FTIR leading to the same conclusion.

The assessment of tertiary structure The primary method for assessing tertiary structure is NUV CD. In situations in which the products concentration as low as 0.3 mg/mL. The intrinsic fluorescence of tryptophan residues in the protein can be used as an indicator of the integrity of the tertiary structure for samples that are below the limit of detection of NUV CD.

Spectral similarity between any two spectra

obtained using OMNIC QC software were analyzed with a 7-point smooth and a 4 nm resolution. The result is a value between 0 and 100%, where 0% indicates no similarity and 100% indicates the spectra are identical. In this study, the sample corresponding to 100% IgG1 was used as the reference spectrum. The comparative data for the NUV CD data is shown in Figure 5. Each sample was run in triplicates and the spectral similarity of 100% corresponds to the IgG1 sample. NUV CD data shows an interesting trend. Using spectral similarity to analyze the comparability data we see that near UV CD in unable to distinguish between 100% IgG1 and IgG2 and all the mixtures. However, using WSD for analysis we see that NUV CD can not only distinguish between the 100% IgG1 and IgG2 but it was also able to distinguish between some of the mixtures as indicated in Figure 6.

12

Intrinsic fluorescence is based on the detection of the fluorescence of tryptophan residues in the proteins that are sensitive to differences in the packing of the hydrophobic core. Maximum emission intensity shifts to longer wavelengths when tryptophan residues are exposed to more polar environments whereas maximum emission intensity shifts to shorter wavelengths when tryptophan residues are less exposed to polar environments. Thus, the emission wavelength of tryptophan fluorescence reflects the local tryptophan environment, which reflects the tertiary structure of the protein16-17.

Similarity was established using internal universal acceptance

criteria of ±2 nm maximum wavelength emission intensity deviation.

As in the other

comparisons, 100% IgG1 was used as the reference. The comparative data for the Intrinsic fluorescence data is shown in Figure 7. The comparisons of the intrinsic fluorescence data between the different samples shows that intrinsic fluorescence is not even able to distinguish between the IgG1 and IgG2 samples due to similar fluorescence properties (Figure 8).

The assessment of thermal stability/tertiary structure In addition to the spectroscopic methods for tertiary structure analysis by NUV CD and intrinsic tryptophan fluorescence, differential scanning calorimetry (DSC) is used for assessing the thermal stability. DSC can also provide information on the stability of each domain of the molecules. The DSC thermograms of the molecules are shown in Figure 9. From the DSC profiles in the figure, we can clearly see the differences between the IgG1 and IgG2 molecules. There are three endothermic thermal transitions for both IgG1 and IgG2 molecules.

The

unfolding thermal transition temperatures (Tm) of the CH2 and CH3 domains of both molecules are quite similar. However, the unfolding thermal transition temperatures of the Fab domain of the IgG1 and IgG2 molecules are significantly different. The Tm values were analyzed using the EAC defined in Material and Methods. The results are summarized in Figure 10. The green

13

colors indicate that the samples are distinguishable, and the red colors indicate that the samples are indistinguishable. The DSC method can differentiate 61% of the sample combinations.

The assessment of Higher order Structure by Profile NMR analysis The Profile NMR method exploits differences in the diffusion of large antibodies and formulation components to generate a highly resolved one-dimensional

1

H spectrum of the entire

monoclonal antibody (mAb). The subsequent subtraction of the featureless component of this spectrum yields a detailed fingerprint spectrum, suitable for spectral similarity calculations3-4. Figure 11 shows the 1H profile overlay of the eight various IgG blends. Since the frequencies emitted are proportional to the static magnetic field of the instrument, they are reported as a ratio to the static field in parts per million (ppm). Even without further mathematical analysis, visually one can see clear differences between the sample blends, particularly demarcated between the 20% and 80% IgG2 samples. The results obtained from the auto- and crosscorrelation analysis of the spectra are shown in Figure 12.

DISCUSSION The assessment of HOS attributes is critical in order to show product comparability due to manufacturing changes, and to support the demonstration of biosimilarity between a reference product and a biosimilar product. There have been limited studies evaluating the differentiating capabilities of different HOS characterization methods. In this study we have compared established biophysical methods for HOS assessments with one newly introduced NMR-based method for biopharmaceutical studies to evaluates their differentiating ability for the HOS of biopharmaceuticals. Monoclonal antibodies are currently the most commonly used protein structural family in protein therapeutics. The two subclasses of a mAb, IgG1 and IgG2, selected for this HOS study, are highly relevant as model proteins for the assessment of HOS attributes for biopharmaceuticals

14

in comparability and similarity studies. The samples were blended into mixtures containing various percentages of IgG1 and IgG2 to test the differentiating ability of each method for these samples. Since IgG1 and IgG2 have slightly different structures, this set of samples is expected to be present a significant challenge for product differentiation. As concluded from our studies, these subtle differences are not completely distinguishable using current methods including the application of NMR in the HOS characterization, suggesting that this sample set is appropriate for evaluating the differentiating ability of these methods. The primary structural similarity between the IgG1 and IgG2 molecules used in this study is very high (95% sequence identity)9. This means that the mixtures containing, for instance 90% of IgG1 and 10% of IgG2, represent a minor difference in overall structural properties. However, we were able to show that the 1D Profile NMR method could readily distinguished even such subtle structural differences. Our results show that FTIR, FUV CD, and intrinsic fluorescence spectroscopy are unable to differentiate between most sample combinations indicating that these three methods are not even sensitive enough to differentiate the subtle differences between IgG1 and IgG2 subclasses. In contrast, both DSC can differentiate between approximately 61% and NUV CD can differentiate between approximately 52% of the sample combinations, suggesting that these two methods have moderate differentiating ability for distinguishing the subtle differences between the IgG1 and IgG2 structures. Finally, the Profile NMR data show that this new method is able to distinguish between most sample combinations (93%) implying that NMR has superior ability to address subtle differences in HOS. The NMR sensitivity is a feature that could be directly applicable in high resolution structural analysis applied to product characterizations in comparability and similarity assessments.

Acknowledgements We would like to express our gratitude for important input from Linda Narhi and Jette Wypych.

15

References (1)

Berkowitz SA, Engen JR, Mazzeo JR, Jones GB. Analytical tools for characterizing biopharmaceuticals and the implications for biosimilars. Nat Rev Drug Discov. 2012;11:527–540.

(2)

Gabrielson JP, Weiss WF. Technical decision-making with higher order structure data: starting a new dialogue. J Pharm Sci. 2015;104:1240–1245.

(3)

Poppe, L, Jordan, JB, Lawson, K, Jerums, M., Apostol, I., and Schnier, PD. Profiling Formulated Monoclonal Antibodies by 1H NMR Spectroscopy. Anal. Chem. 2013;85, 9623−9629.

(4)

Poppe, L., Jordan, JB., Rogers, G., and Schnier, PD. On the Analytical Superiority of 1D NMR for Fingerprinting the Higher Order Structure of Protein Therapeutics Compared to Multidimensional NMR Methods. Anal. Chem. 2015;87, 5539−5545.

(5) Arbogast, LW., Brinson, RG., and Marino., JP. Mapping Monoclonal Antibody Structure by 2D 13C NMR at Natural Abundance. Anal. Chem. 2015;87, 3556-3561. (6) Brinson RG., et al., Enabling adoption of 2D-NMR for the higher order structure assessment of monoclonal antibody therapeutics. MABS 2019;11, 94-105. (7) Zhang, Z., and Smith, DL. Determination of amide hydrogen exchange by mass spectrometry: A new tool for protein structure elucidation. Protein Science 1993;2, 522-531. (8) Goswani, D., Zhang, J., Bondarenko, PV., and Zhang, Z. MS-based conformation analysis of recombinant proteins in design, optimization and development of biopharmaceuticals. Methods 2018;144, 134-151. (9) Dillon, TS., Ricci, MS., Vezina, C., Flynn, GC., Liu, YD., Rehder, DS., Plant, M., Henkle, B., Li, Y., Deechongkit, S., Varnum, B., Wypych, J., Balland, A., and Bondarenko, PV. Structural

16

and Functional Characterization of Disulfide Isoforms of the Human IgG2 Subclass. J. Biol. Chem. 2008;283, 16206-16215. (10) Dong, A., and Caughey, WS. Infrared methods for study of hemoglobin reactions and structures. Methods Enzymol. 1994; 232, 139-175 (11) Li, CH, Nguyen, X, Narhi, L., Chemmalil, L, Towers, E, Muzammil, S, Gabrielson, J, Jiang, Y. Applications of circular dichroism (CD) for structural analysis of proteins. J Pharm Sci 2011;100, 4642-4654 (12) Teska, BM., Li, C., Winn, BC., Arthur, KK., Jiang, Y., Gabrielson, JP. Comparison of quantitative spectral similarity analysis methods for protein higher-order structure confirmation. Analytical Biochemistry. 2013;434, 153-165. (13) Dinh, NN., Winn, BC., Arthur, KK., and Gabrielson, JP. Quantitative spectral comparison by weighted spectral difference for protein higher order structure confirmation. Analytical Biochemistry. 2014;464, 60-62. (14) Narhi, LO., Wypych, J., Langley, KE., Arakawa, T. Changes in conformation and stability upon SCF/sKit complex formation. J. Protein Chem 1998;17, 387-396. (15) Cotts, RM, Hoch, MJR, Sun, T, Markert, JT, Pulsed field gradient stimulated echo methods for improved NMR diffusion measurements in heterogeneous systems. J. of Magnetic Resonance. 1969;83 (2), 252-266. (16) Ramachander, R., Jiang, Y., Li, C., Eris, T., Young, M., Dimitrova, M., and Narhi, L. Solid state fluorescence of lyophilized proteins. 2008;376, 173-182. (17) Lakowicz, Joseph R. Principles of Fluorescence Spectroscopy 2nd edition. Kluwer Academic/Plenum Publishers, New York; 1999:1-23, 63, 445-481

Figure legends

17

Figure 1. FTIR second derivative spectra (Amide I region) of IgG1, IgG2, and their different blends. Figure 2. WSD (panel on left) and spectral similarity scores (panel on right) showing which samples are distinguishable (green) and indistinguishable (red) when compared against a reference IgG1 replicate 1 (grey). Figure 3. FUV CD spectra of IgG1, IgG2 and their different blends.

Figure 4. WSD (panel on left) and spectral similarity scores (panel on right) showing which samples are distinguishable (green) and indistinguishable (red) when compared against a reference IgG1 replicate 1 (grey).

Figure 5. NUV CD spectra of IgG1, IgG2 and their different blends.

Figure 6. WSD (panel on left) and spectral similarity scores (panel on right) showing which samples are distinguishable (green) and indistinguishable (red) when compared against a reference IgG1 replicate 1 (grey).

Figure 7. Intrinsic Fluorescence spectra of IgG1, IgG2 and their different blends.

Figure 8. Maximum wavelength emission intensity deviation of FLD data showing which samples are distinguishable (green) and indistinguishable (red) when compared against a reference IgG1 replicate 1 (grey). Figure 9. DSC thermograms of IgG1, IgG2, and their different blends.

Figure 10. DSC results showing which samples are distinguishable (green) and indistinguishable (red). The diagonal elements represent the comparison of the same sample pairs are therefore not used for comparison (grey).

Figure 11. NMR Profile Spectra of IgG1, IgG2, and their different blends.

18

Figure 12. Profile NMR results showing which samples are distinguishable (green) and indistinguishable (red). The diagonal elements represent the comparison of the same sample pairs are therefore not used for comparison (grey).

19

Table 1. List of IgG1 and IgG2 sample blends. All samples are prepared in the same formulation buffer.

Sample IgG2 95% IgG2 90% IgG2 80% IgG2 20% IgG2 5% IgG2 IgG1

IgG2 (mg/mL) 50 47.5 45 40 10 2.5 0

IgG1 (mg/mL) 0 2.5 5 10 40 47.5 50