lable at ScienceDirect

Journal of Pharmaceutical Sciences 107 (2018) 1806-1819

Contents lists avai

Journal of Pharmaceutical Sciences

journal homepage: www.jpharmsci .org

Pharmaceutical Biotechnology

Analytical Comparability Assessments of 5 Recombinant CRM197Proteins From Different Manufacturers and Expression Systems

John M. Hickey 1, Vishal M. Toprani 1, Kawaljit Kaur 1, Ravi P.N. Mishra 2, Akshay Goel 2,Natalia Oganesyan 3, Andrew Lees 3, Robert Sitrin 4, Sangeeta B. Joshi 1, David B. Volkin 1, *

1 Department of Pharmaceutical Chemistry, Macromolecule and Vaccine Stabilization Center, University of Kansas, Lawrence, Kansas 660472 Biological E. Ltd., Hyderabad, Telangana 500078, India3 Fina Biosolutions LLC, Rockville, Maryland 208504 PATH, Washington, Dist. of Columbia 20001

a r t i c l e i n f o

Article history:Received 21 November 2017Revised 1 February 2018Accepted 1 March 2018Available online 8 March 2018

Keywords:CRM197conjugate vaccinecarrier proteinbiophysicalstabilitycomparabilitymass spectrometryimmunoassays

Abbreviations used: TT, tetanus toxoid; DT, diphtreacting material 197; cIEF, capillary isoelectric focchromatography; SV-AUC, sedimentation velocity aSEC, size exclusion chromatography; RMM, resonanmicroflow imaging; EPD, empirical phase diagram.The authors John M. Hickey and Vishal M. Toprani coThis article contains supplementary material availableor via the Internet at https://doi.org/10.1016/j.xphs.20* Correspondence to: David B. Volkin (Telephone: 7

5736).E-mail address: volkin@ku.edu (D.B. Volkin).

https://doi.org/10.1016/j.xphs.2018.03.0020022-3549/© 2018 American Pharmacists Association

a b s t r a c t

Cross-reacting material 197 (CRM197), a single amino acid mutant of diphtheria toxoid, is a commonlyused carrier protein in commercial polysaccharide protein conjugate vaccines. In this study, CRM197

proteins from 3 different expression systems and 5 different manufacturers were obtained for ananalytical comparability assessment using a wide variety of physicochemical and in vitro antigenicbinding assays. A comprehensive analysis of the 5 CRM197 molecules demonstrate that recombinantCRM197's expressed in heterologous systems (Escherichia coli and Pseudomonas fluorescens) are overallhighly similar (if not better in some cases) to those expressed in the traditional system (Corynebacteriumdiphtheriae) in terms of primary sequence/post-translational modifications, higher order structuralintegrity, apparent solubility, physical stability profile (vs. pH and temperature), and in vitro antigenicity.These results are an encouraging step to demonstrate that recombinant CRM197 expressed in alternativesources have the potential to replace CRM197 expressed in C diphtheriae as a source of immunogeniccarrier protein for lower cost polysaccharide conjugate vaccines. The physicochemical assays establishedin this work to monitor the key structural attributes of CRM197 should also prove useful as comple-mentary characterization methods (to routine quality control assays) to support future process andformulation development of lower cost CRM197 carrier proteins for use in various conjugate vaccines.

© 2018 American Pharmacists Association®. Published by Elsevier Inc. All rights reserved.

Introduction

Polysaccharide vaccines have been used effectively since the1970s to mitigate bacterially associated diseases, such as pneu-monia and meningitis, in adults.1 The transient immune responseelicited by polysaccharide vaccines limit efficacy in individualswith an immature (<2 years) immune system.2 Subsequently,

heria toxoid; CRM197, cross-using; AEX, anion exchangenalytical ultracentrifugation;t mass measurement; MFI,

ntributed equally.from the authors by request18.03.002.85-864-6262; Fax: 785-864-

®. Published by Elsevier Inc. All rig

conjugate vaccines, composed of bacterial polysaccharideschemically linked to a carrier protein, were developed to stimu-late a more robust and sustained (T-cell-dependent) immuneresponse. While commercially available polysaccharide proteinconjugate vaccines (e.g., Prevnar®, Menveo®, VaxemHib®) havesubstantially decreased incident rates of bacterially associateddiseases,3,4 their complex nature leads to costly production whichcurrently is one major hurdle limiting vaccine coverage indeveloping countries.5,6

Three carrier proteins have been used in a majority of com-mercial conjugate vaccines.7 These immunogens consist of tetanustoxoid (TT), diphtheria toxoid (DT), and a single amino acidsubstitution of diphtheria toxin (G52E), termed as cross-reactingmaterial 197 or CRM197. Conjugate vaccines using these 3 proteinshave proven to be safe and effective, and from the limited literatureavailable, induce comparable immune responses to similar poly-saccharide antigens.7 While the clinical efficacy of these proteinsappears similar, inherent manufacturing requirements and chal-lenges differentiate these 3 immunogens. TT and DT require

hts reserved.

J.M. Hickey et al. / Journal of Pharmaceutical Sciences 107 (2018) 1806-1819 1807

detoxification before polysaccharide conjugation using formalde-hyde treatment. Conversely, the amino acid substitution in CRM197renders the protein nontoxic and therefore is more advantageous tomanufacture than TT or DT because CRM197 does not requireformaldehyde treatment (resulting in more lysine groups availablein the carrier protein as potential polysaccharide conjugation sites).Production of CRM197 is hindered, however, by the strict growthrequirements of the traditional expression system (C diphtheriae)and generally lower yields compared to other expression systems.8

Given these challenges, the evaluation of recombinant CRM197expressed in multiple heterologous organisms (e.g., E coli, P fluo-rescens, Bacillus subtilis) is being pursued as an alternative source ofcarrier proteins.9

Given the complex process of manufacturing conjugatevaccines,10 delineating the physicochemical properties of recom-binant CRM197, expressed in C diphtheriae or a heterologous system,is essential to mitigate risk during manufacturing, includingsubsequent polysaccharide conjugation, to ensure the immunoge-nicity of the final drug product. CRM197 has a similar overallconformation to diphtheria toxin; however, the crystal structure ofCRM197 (PDB ID: 4AE0) elucidated subtle but functionally impor-tant (i.e., lack of toxicity in CRM197) differences between the 2proteins.11 CRM197 is translated as a 58 kDa polypeptide inC diphtheriae, which can be cleaved by a trypsin-like protease into 2subunits (fragments A and B) linked through a disulfide bond.During manufacturing of CRM197, cleavage of the polypeptide, alsoknown as “nicking,” is generally limited to <5% of total CRM197.12

Following purification of CRM197, polysaccharides are generallyconjugated to primary amines on the carrier protein through eitherdirect or indirect (linker molecule) methods.13 In addition tomultiple factors that can affect the immunogenicity of a vaccineduring conjugation (e.g., polysaccharide composition, linker length,conjugation chemistry), the accessibility of specific side chains inthe carrier protein can affect polysaccharide conjugation.14 There-fore, ensuring the carrier protein is consistently manufactured andwell characterized before conjugation is critical.

In this study, 5 different recombinant CRM197 molecules fromthe traditional expression system (C diptheriae) and 2 heterolo-gous systems (E coli and P fluorescens) were analytically charac-terized and comprehensively compared in terms of structuralintegrity, solubility, and conformational stability profiles using awide variety of biochemical, biophysical, and in vitro antigenbinding assays. The results from this study provide baseline datasets of the key structural attributes of CRM197 to enable futurecomparisons of the physicochemical and immune-reactivityproperties of recombinant CRM197 irrespective of expressionsystem and production process.

Table 1Summary of Sample Information for 5 Different CRM197 Molecules Used in This Study

CRM197 SampleName

Source System for RecombinantCRM197 Expression

Lot Number (Date)

EcoCRM™ Fina Biosolutions LLC E coli NO13p46 (November

rCRM Biological E. Ltd. E coli P/003/16 (February, 2Pfenex CRM Reagent Proteins P fluorescens 162M4001 (May, 201

Vaxform CRM VaxForm LLC C diphtheriae Not provided (2006)

C7 CRM List BiologicalLaboratories, Inc.

C diphtheriae 14936A1 (February, 2

Material and Methods

Materials

The sample information for the 5 different CRM197 moleculesexamined in this study is summarized in Table 1. EcoCRM™ (FinaBiosolutions LLC, Rockville, MD), rCRM (Biological E. Ltd., Hyder-abad, Telangana, India), Pfenex CRM (Reagent Proteins, San Diego,CA), C7 CRM (List Biological Laboratories Inc., Campbell, CA) andVaxForm CRM (VaxForm LLC, Lehigh Valley, PA) were diluted to2.0 mg/mL with their respective formulation buffers and thendialyzed against 10 mM sodium phosphate, 150 mM NaCl, pH 7.2(phosphate-buffered saline [PBS] buffer) using 10 kDa molecularweight (MW) cutoff Slide-A-Lyzer Mini Dialysis devices (ThermoScientific, Rockford, IL). The samples were dialyzed overnight at 4�Cwith 3 buffer changes at 3 h intervals, and the concentration of eachCRM197 sample was determined both pre and post centrifugation(13,000 rpm for 5 min). Each of the CRM197 samples was furtherdiluted to 1.0mg/mLwith PBS buffer for all subsequent analyses. Allother chemical and reagents were purchased from commercialsources.

Methods

Ultraviolet-Visible Absorbance SpectroscopyThe ultraviolet (UV)-visible absorption spectra of each CRM197

sample at 1.0 mg/mL was recorded using an HP-8453 photodiodearray detector (Agilent Technologies, Santa Clara, CA) equippedwithdeuterium (D2) and tungsten (W) lamps. In addition, samples werecentrifuged at 13,000 � g for 5 min, and the UV-visible absorptionspectra of the resulting supernatant were recorded. The Beer-Lambert law was used to calculate the actual concentration givenan extinction coefficient of a 0.1% solution of CRM (0.932 mg/mL�1

cm�1; calculated using http://web.expasy.org/protparam/). Spectrawere collected from 190 nm to 1100 nm using a 0.5 s integrationtime and 1 cm path length quartz cuvettes. The instrument was firstblankedusing thebuffer for each samplebeforemeasuring solutionscontaining protein. All UV-visible absorbance spectra were mathe-matically corrected for light scattering using a technique included inthe manufacturer's data analysis software (Chemstation UV-Visanalysis software; Agilent Technologies).

SDS-PAGEFor the reduced samples, 2.5 mg of each CRM197 sample was

mixed with 4� LDS loading dye (Life Technologies, Grand Island,NY) containing 50 mM dithiothreitol (Invitrogen, Carlsbad, CA) and100 mM iodoacetamide (Life Technologies) and incubated at room

StorageTemperature (�C)

Formulation Buffer Reported StockConcentration

, 2016) �80 (frozen liquid) 25 mM of HEPES, 10% glycerol,pH 7.2

5.0 mg/mL

016) �80 (frozen liquid) PBS, 5% sucrose, pH 7.35 2.2 mg/mL4) 4 (Lyophilized) 250 mg of sucrose, 1.1 mg of

sodium phosphate monobasic,15 mg of sodium phosphatedibasic, 0.275 mg ofpolysorbate 80

20 mg/vial

�20 (frozen liquid) 10 mM of sodium phosphate,5% sucrose, pH 7.0

4.0 mg/mL

015) 4 (lyophilized) 10 mM of sodium phosphate,5% lactose (when reconstitutedwith 0.25 mL of water)

0.5 mg/vial

J.M. Hickey et al. / Journal of Pharmaceutical Sciences 107 (2018) 1806-18191808

temperature for 30 min. The reduced CRM197 samples were thenseparated by sodium dodecyl sulfate polyacrylamide gel electro-phoresis using NuPAGE 4%-12% Bis-Tris (Life Technologies) gels andMES running buffer (Life Technologies). A similar procedure wasfollowed for nonreduced CRM197 samples except the dithiothreitolwas omitted during the incubation at room temperature for a30-min step. The NuPAGE gels were run for 60min at 120 V. Proteinbands were visualized by staining with Coomassie Blue R-250(Teknova, Hollister, CA) and destained with a mixture of 40%methanol, 10% acetic acid, and 50% ultrapure water. Gels weredigitized using an AlphaImager (ProteinSimple, Santa Clara, CA) gelimaging system.

Intact Mass Spectrometry AnalysisEach CRM197 sample in PBS (50 pmol) was injected into an 1200

series LC system (Agilent Technologies), bound to a C8 micro-trap(Michrom Bioresources Inc., Auburn, CA), desalted, and then sub-jected to electrospray ionization time-of-flight mass spectrometry(model 6220; Agilent Technologies). Mass spectra were acquiredfrom 400 to 2000 m/z at a scan rate of one spectrum per second.Protein mass spectrometry (MS) spectra were deconvoluted usingMassHunter Quantitative Analysis software (v B.07.00; AgilentTechnologies).

Liquid ChromatographyeMS Peptide MappingEach CRM197 sample was incubated overnight at 37�C with 2 mg

of trypsin or GluC. Trifluoroacetic acid (0.05%) was added to quenchproteolysis, and ~20 mg of each digested sample was subjected toliquid chromatography (LC)eMS. The peptides from the digestedprotein solutionwere separated by a liquid chromatography system(Thermo Scientific, Waltham, MA) before analysis. Peptides wereinjected onto a C18 column (1.7 mm, 2.1 � 150 mm; Waters) and a55 min 5%-50% B gradient (A: H2O and 0.05% trifluoroacetic acid; B:acetonitrile and 0.05% trifluoroacetic acid; 200 mL/min flow rate) forseparation. MS was performed using a LTQ-XL ion trap (ThermoScientific) and the Xcalibur 2.0 software (Thermo Scientific). Theinstrument was also tuned using a standard calibration peptide(angiotensin II; Sigma, St. Louis, MO) for maximal sensitivity beforerunning any experiments. The mass spectra were acquired in theLTQ over a mass range of m/z 350-1900. The ion selection thresholdwas 10,000 counts, and the dynamic exclusion duration was 5 s.

Raw experimental files were initially evaluated manually todetermine if the ion counts and fragmentation of each peptideweresufficient for further analysis. The rawdatafileswere thenprocessedusing PepFinder (v 2.0) software (Thermo Scientific). The databaseused for this experiment consisted of the primary sequences ofCRM197 (PDB ID: 4AE0)11 and trypsin or GluC. Potential commonpost-translational modifications (e.g., Asn deamidation, Metoxidation, mixed disulfide bond formation) were included duringthe analysis. Peptide assignments of MS/MS spectra were validatedusing a confidence score of �95% and disulfide-bonded peptideassignments were also manually verified.

Capillary Isoelectric FocusingElectrophoretic separation by isoelectric point was performed

using an ICE280 system (Protein Simple, Santa Carla, CA) equippedwith aD2 lamp (280nm), a PrinCEmicroinjector, and an autosampler.The fluorocarbon-coated cartridge was rinsed with ultrapure waterbefore and after daily experiments. Separation efficiency and transfertime were determined by separation of a hemoglobin analyticalstandard.CRMsampleswereanalyzedunder the followingconditionsadapted from Rustandi et al.15 and Loughney et al.16: ampholytes: amixture of pH 4-6.5 and pH 3-10 in a 2:1 ratio (GE Healthcare, Pitts-burgh, PA), IEFmarkers (4.65, 7.05), additives (20% glycerol), and 1.0%methylcellulose (Protein Simple). The sample was resolved using the

prefocusing conditions of 1min at 1500 V and a focusing condition of8min at 3000 V. Data analysis and peak integrationswere performedusing Chromperfect software (Protein Simple).

Anion Exchange ChromatographyAnion exchange chromatography (AEX) was performed on

a Shimadzu Prominence ultra fast liquid chromatography high-performance liquid chromatography (HPLC) systemequippedwith adiode array detector using a TSKgel BioAssist Q (4.6 mm � 5 cm, 10mm) column (Tosoh Biosciences, King of Prussia, PA). Twentymicrograms of each CRM197 sample was injected onto the columnfor each run, and the experiment was performed in triplicate. Thecolumn and autosampler temperatures were set at 30�C and 4�C,respectively. The mobile phases consisted of (A) 20 mM sodiumphosphate, pH 7.4 and (B) 20 mM sodium phosphate, 1 M sodiumchloride, pH 7.4. The flow rate was 0.7 mL/min, and the gradientconsisted of 0% B (5min), 0%-100% B (10min),100% B (3min),100%-0% B (2 min), and 0% B (5 min). Protein peaks were monitored usingthe absorbance signal at 214 nm. In addition, the samples weresubjected toAEXwithout the columnattached tobetter determine ifany fraction of the sample binds or cannot pass through the column(i.e., sample recovery). Data analysis was performed usingLC-Solution software (Shimadzu, Kyoto, Japan).

Far-UV Circular DichroismFar-UV circular dichroism (CD) spectroscopy was performed

using a Chirascan-plus CD spectrometer (Applied Photophysics Ltd.,Leatherhead, UK) equipped with a 6-cuvette position Peltiertemperature controller (Quantum Northwest, Liberty Lake, WA)and a high-performance solid-state detector. The lamp (150 W air-cooled Xe arc) housing, monochromator and sample compartmentwere continuously purged with N2 gas. The CD spectra of theprotein samples at 0.2 mg/mL were collected in the range of 260-200 nm using 1-nm step size and a 0.5 s sampling time. Quartzcuvettes (0.1-cm path length) sealed with a Teflon stopper (StarnaCells Inc., Atascadero, CA) were used. The CD signal at 222 nm wasmonitored as a function of temperature from 10�C to 90�C in 1.25�Cintervals. The heating rate was 1�C/min, and the equilibration timeat each temperature was 2min. All datawere subjected to a 5-pointSavitzky-Golay smoothing filter using the Chirascan software(Applied Photophysics), and PBS buffer alone was subtracted fromall measurements.

Intrinsic Tryptophan Fluorescence Spectroscopy and Static LightScattering

The intrinsic tryptophan fluorescence spectra of each CRM197sample was measured in triplicate using a dual emission PTI QM-40Spectrofluorometer (Photon Technology International, Inc., Bir-mingham, NJ) equipped with a 4-position cell holder Peltier tem-perature control device, a high-power continuous 75 W short-arcXe lamp (Ushio), and a Hamamatsu R1527 photomultiplier tube.Data were collected using FelixGX software (Photon TechnologyInternational, Inc.). Fluorescence emission spectra of 0.2 mg/mLCRMwere recorded as a function of temperature (10�C-90�C) usingan excitation wavelength of 295 nm (>95% tryptophan [Trp]).Emission spectrawere collected from 305 nm to 405 nmwith a stepsize of 1 nm and an integration time of 1 s. Static light scatteringdata were acquired concurrently with the fluorescence spectra byusing a second detector (90� to the incident light and 180� to thefluorescence detector) that collected light scattered signal at theexcitation wavelength (295 nm) as a function of increasing tem-perature. The excitation and emission slits were both set such thatthe initial signal at 10�C had an emission maximum of ~800,000counts per second for fluorescence spectra and an emissionmaximum of ~20,000 counts per second for light-scattering

J.M. Hickey et al. / Journal of Pharmaceutical Sciences 107 (2018) 1806-1819 1809

spectra. The spectra were collected at 1.25�C intervals with a 2-minequilibration time at each temperature with samples in quartzcuvettes (1-cm path length). The position of the emission wave-length maximum was determined using a mean spectral center ofmass method executed using in-house software (MiddaughSuite)after formulation buffer subtraction. This analysis algorithm in-creases the signal-to-noise ratio, but the peak positions aregenerally red shifted by 5-10 nm from their experimental positions.

Extrinsic 8-Anilino-1-Naphthalene Sulfonate FluorescenceSpectroscopy

8-Anilino-1-naphthalene sulfonate (ANS) was used as anextrinsic fluorescence probe in the presence of CRM197 with thesame instrument as described previously. A dye to protein molarratio of 25:1 was used for sample preparation. ANS was excited at372 nm, and emission spectra of ANS were collected from 400 nmto 600 nm every 2 nm as a function of temperature from 10�C to90�C. The corresponding ANS in buffer spectrum was subtractedfrom ANS in the presence of protein spectrum before data analysis.The emission peak intensity was determined using a mean spectralcenter of mass method executed in the in-house software. Theonset melting temperature (Tonset) values were determined byidentifying the point at which the baseline deviated from linearityusing Origin 8.0 software.

Differential Scanning CalorimetryCRM197 samples were loaded in the autosampler tray held at

4�C, and differential scanning calorimetry (DSC) was performed intriplicate using an Auto-VP capillary differential scanningcalorimeter (MicroCal/GE Health Sciences, Pittsburgh, PA) equippedwith Tantalum sample and reference cells. Scans were completedfrom 10�C to 90�C using a scanning rate of 60�C/h. Referencesubtraction and concentration normalization were performedusing Origin (OriginLab, Northampton, MA). Data analysis wasperformed using the MicroCal LLC DSC plug-in for the Origin 7.0software package. The results were fitted to a multistate modelwith 2 transitions to calculate the melting temperature (Tm) values.The Tonset was determined by identifying the point where the heatcapacity (Cp) value for the first thermal transition reached�500 calmol�1 �C�1.

Sedimentation Velocity Analytical UltracentrifugationSedimentation velocity analytical ultracentrifugation (SV-AUC)

experiments were performed on a Proteome Lab XL-I (BeckmanCoulter, Fullerton, CA) analytical ultracentrifuge equipped with ascanning UV-visible optical system. All experiments were con-ducted at 20�C after �1 h of equilibration after the rotor reachedthe set temperature, at a rotor speed of 40,000 rpm and withdetection at 280 nm. CRM197 samples and PBS alone were eachloaded into Beckman charcoal-epon 2 sector cells with a 12-mmcenterpiece and either sapphire or quartz windows.

The data were analyzed using Sedfit (Dr. Peter Schuck, NIH). Apartial specific volume of 0.73 mL/g was calculated using Sednterp(Professor Thomas Laue, University of New Hampshire and BITC)based on amino acid sequence and used in the analysis. The bufferdensity and viscosity used in the analysis, 1.0058 g/mL and0.010195 Poise respectively, were calculated using Sednterp basedon buffer composition. A continuous c(s) distribution was usedwith 200 scans. A range of 0-15 svedbergs was used, with a reso-lution of 300 points per distribution and a confidence level of 0.95.Baseline, radial-independent noise, and time-independent noisewere fit, while the meniscus and bottom positions were setmanually. Distributions were imported into Origin (OriginLab,Northampton, MA) before reporting.

Size Exclusion ChromatographyA Shimadzu Prominence ultra fast liquid chromatography HPLC

system equipped with a diode array detector (with absorbancedetection at 214 nm) was used. Twenty micrograms of each CRM197sample was injected onto a TSKgel BioAssist G2SWxl column (7.8 �300 mm; Tosoh Biosciences) and the corresponding guard column(Tosoh Biosciences). The columns were operated at 30�C and equili-brated with at least 10 column volumes of mobile phase (10 mMsodium phosphate, 150 mM NaCl, pH 7.2) before sample injection. Aflow rate of 0.7 mL/min was used with a 30-min run time. A gelfiltration standard (Bio-Rad, Hercules, CA) was subjected to sizeexclusion chromatography (SEC) before and after the CRM sample toensure integrity of the column and HPLC system. In addition, thesampleswere subjected to SECwithout the columnattached to betterdetermine if insoluble aggregates are present in the samples or if thesample binds to the column (i.e., sample recovery). LC-Solutionsoftware (Shimadzu) was used for data analysis.

Resonant Mass MeasurementAn Archimedes Particle Metrology System (Malvern

Instruments Ltd., Malvern, UK) was used to quantify submicron-sized particles (0.2-2 mm). A Hi-Q micro-H sensor, calibrated us-ing 1-mm polystyrene beads (Thermo Scientific), and a referencesolution of 1:20 D2O:H2O were used for all measurements. Beforedaily measurements, the accuracy of the sensor was assessed byanalysis of 1-mm polystyrene beads (Thermo Scientific). To achievea clean baseline, the sensor and the tubing were rinsed with 20%Contrad-20 followed by ultrapure water. The sensor was thenloaded with particle-free water, and 2 “sneeze” operations wereperformed. Each CRM197 sample at 1.0 mg/mL was loaded for 30 s,and a stop trigger of 300 particles was used. The total number andsize distribution of submicron particles were analyzed in triplicate.The limit of detection was determined empirically (0.030 Hz) andused throughout the study. Data analysis was performed usingParticleLab software 1.9.30 (Malvern, UK).

Microflow ImagingThe concentration and size range of subvisible particles

(2-100 mm) were measured and quantified using a DPA-4200 flowmicroscope (Protein Simple) that was previously calibrated using10-mm polystyrene particle standards (Thermo Scientific). The in-strument was flushed with particle-free water until a clean baselinewas achieved. Each CRM197 sample at 0.2 mg/mL (1:5 dilution) wasanalyzed in triplicate experiments. The samples were carefullydrawn up in a low protein binding, filter-tip pipette (Neptune Sci-entific, San Diego, CA) and analyzed using a flow rate of 0.2 mL/min.Data analysis was performed using MVSS software (Protein Simple).

Polyethylene Glycol Precipitation AssayThe protocol for the high-throughput version of the poly-

ethylene glycol (PEG) assay for assessing apparent solubility of CRMmolecules was adopted from the study by Gibson et al.17 and Top-rani et al.18 Stock solutions of PBS pH 7.2 and PBS containing 40% w/v PEG-10,000 at pH 7.2 were mixed to prepare various concentra-tions of PEG solutions ranging from 0% to 40% w/v PEG. A volume of200 mL of the various PEG-10,000 solutions was added to wells of a96-well polystyrene filter plate (Corning Life Sciences, Corning,NY). Fifty microliters of each CRM sample in PBS pH 7.2 (1 mg/mL)was then added to each well to a final protein concentration of 0.2mg/mL. The plates were incubated overnight at room temperatureand then centrifuged at 3500 rpm (1233 rcf) for 15 min, and thefiltrate was collected in a clear 96-well collection plate (GreinerBio-One North America Inc., Monroe, NC). Subsequently, 200 mL offiltrate was transferred into a 96-well UV Star microplate, and theprotein concentration in each filtrate was determined using a

J.M. Hickey et al. / Journal of Pharmaceutical Sciences 107 (2018) 1806-18191810

SpectraMax M5 UV-Visible plate reader. The % PEGmidpt andapparent solubility values (thermodynamic activity) were calcu-lated as described previously.17

Empirical Phase Diagrams/Radar ChartsEach of the 5 CRM197 samples were dialyzed against 10-mM

sodium phosphate buffer containing sodium chloride at an ionicstrength of 0.15 at pH 5.8, 6.3, 6.8, 7.2, 7.6, and 8.0 using Slide-A-Lyzer dialysis devices (10 kDa MW cutoff; Thermo Scientific).After 4 rounds of buffer exchange overnight at 4�C, each CRM197sample was collected from the dialysis device and centrifuged at13,000 � g at 4�C for 5 min. The biophysical stability of CRM197molecules was determined using far-UV CD spectroscopy, intrinsictryptophan fluorescence spectroscopy, static light scattering, andDSC (as described previously) for each protein sample at each pH asa function of temperature. Empirical phase diagrams (EPDs) andradar charts were constructed as data visualization tools to sum-marize the effect of pH and temperature on the physical stabilityprofile of the CRM197 samples (using the biophysical stability datasets) as described in detail previously.19,20

ELISA Using Polyclonal AntibodiesOne hundred microliters of 0.5 mg/mL heparin-binding

epidermal growth factorelike growth factor (Sigma) in PBS wasadded to each well of a 96-well plate. The plate was then incubatedovernight at 4�C. The following day, the plate was washed 3 timeswith PBST (PBS with 0.05% Tween-20) and 200 mL of blocking bufferwas added (PBSþ 5%w/v fat-freemilk). The platewas incubated for1 h at room temperature, and then, the blocking solution wasremoved, and the platewaswashed 3 timeswith PBST. Each CRM197sample in PBS pH 7.2 (at 2 mg/mL) was then diluted with PBSTþ 1%fat-freemilk to 100 mg/mL. A 1:2 dilution serieswas performedwitheach CRMmolecule in triplicate and 100 mL of each CRM197 solutionwas added to eachwell in the 96-well plate. The platewas incubatedfor 1 h at 37�C, and then, the CRM197 solutions were removed, andthe platewaswashed 3 timeswith PBST. One hundredmicroliters ofa rabbit CRM197 polyclonal antibody (at 0.5 mg/mL in PBST, AIC LLC,Rockville, MD) was added to each well, and the plate was incubatedfor 1 h at 37�C. The primary antibody solutionwas removed, and theplate was washed 3 times with PBST. One hundred microliters of agoat anti-rabbit antibody (at 0.25 mg/mL in PBST; Jackson Immu-noResearch, West Grove, PA) was added to each well, and the platewas incubated for 1 h at 37�C. The secondary antibody solutionwasremoved, and the platewaswashed 3 timeswith PBST. One hundredmicroliters of SureBlue TMB peroxidase (SeraCare Life, Milford, MA)was added to eachwell, and the platewas incubated for 1 h at roomtemperature in the dark. One hundred microliters of 1 N HCl wasadded to quench the reaction, and the absorbance at 450 nmof eachwell wasmeasured using a SpectralMaxM5 plate reader (MolecularDevices, Sunnyvale, CA). Statistical analysis of the ELISA results wasperformed using Prism software (GraphPad Software Inc., La Jolla,CA). A asymmetrical (5 parameter) dose-response (Eq. 1) curve wasfitted to each data set.21

logXb ¼ logEC50 þ ð1=HillSlopeÞ*log��

2Yð1=SÞ�� 1

�(1)

ELISA Using mAbsOne hundred microliters of C7 CRM in PBS (1:1 serial dilutions

from 10 to 0.005 mg/mL) was added to each well of a 96-well plate.The plate was then incubated overnight at 4�C. The following day,the plate was washed 3 times with PBST, and 200 mL of blockingbuffer was added (PBS þ 5% w/v fat-free milk). The plate was incu-bated for 1 h at room temperature, and then, the blocking solution

was removed, and the plate was washed 3 times with PBST. The 5monoclonal mouse CRM197 antibodies (AB8306, AB8307, AB8308,AB8310, andAB53827;Abcam, Cambridge,MA)werediluted 1:1000with PBST þ 0.1% w/v bovine serum albumin (BSA), and 100 mL ofeach antibody was added (individually) to the CRM197-containing96-well plate. The platewas incubated for 2 h at room temperature,and then, the antibody solutions were removed, and the plate waswashed 3 times with PBST. The goat anti-mouse antibody (ThermoFisher) was diluted 1:4000 with PBSTþ 0.1% w/v BSA, and 100 mL ofthe antibodywas added to eachwell. The platewas incubated for 1 hat roomtemperature, and then, the secondaryantibody solutionwasremoved, and the platewaswashed 5 timeswith PBST. One hundredmicroliters of p-nitrophenyl phosphate-liquid substrate (Sigma)was added to each well, and the plate was incubated for 1 h at roomtemperature in the dark. One hundred microliters of 2 N NaOHwasadded to quench the reaction, and the absorbance at 405 and 490nm of each well was measured using a SpectraMax M5 plate reader(Molecular Devices). The absorbance value at 490 nm was thensubtracted from the 405 nm value.

Measuring Binding Kinetics by OctetThe binding kinetics of CRM with mAbs was evaluated using

Octet Red 96 system (Pall ForteBio LLC, Fremont, CA). Anti-mousebiosensors (Pall ForteBio LLC) were hydrated for ~15-20 min inkinetics buffer (PBS pH 7.2 þ 0.1% BSA), followed by precondition-ing comprising 3 rounds of 20 s exposure each to regenerationbuffer (10mMglycine, pH 1.7) alternatingwith kinetics buffer. Afterthe biosensor preconditioning, the kinetic assay comprised 5 steps:(1) 300 s equilibration with kinetics buffer; (2) 450 s of antibodyloading to capture CRM197 mAb onto the biosensor; (3) 200 sequilibrationwith kinetics buffer to obtain a stable baseline after FcmAb immobilization; (4) 600 s association with each of the 5CRM197 samples individually; and (5) 600 s dissociation with ki-netics buffer. The biosensors were regenerated for up to 5 times bysubjecting to alternating cycles of regeneration and kinetics bufferfor 5 s each. Kinetics assay were performed using 8 nM of eachmAband a range of CRM197 concentrations (0-200 nM). Data analysiswas performed using Octet Data Analysis (v 8.2) software. Afterdata processing, including reference subtraction using the 0 nMCRM197 concentration (buffer blank) trace, baseline alignment, andinterstep correction, the association and dissociation traces of the 5CRM197 samples with each of the 5 mAbs individually were fit to a1:1 binding model. KD, ka, and kdis values were extracted from thecurve-fitting analysis of the kinetic data.

Results

Protein Purity, Primary Structure, and Post-TranslationalModification Analysis of 5 CRM197 Molecules

Under nonreducing or reducing SDS-PAGE conditions, 1 bandwas primarily observed in each CRM197 sample (Fig. 1a). Themigration of the main band (between the 49 and 62 kDa MWprotein markers) was consistent with the theoretical MW of aCRM197 monomer (58.4 kDa). A small amount of lower MW specieswas observed in rCRM and C7 CRM. Independent analysis of LC-MSpeptide mapping confirmed these lower MW species were the A(~21 kDa) and B (~38 kDa) subunits of CRM197 (data not shown).

Intactmass spectrometrywas used to elucidate and compare themolecular composition of the 5 CRM197 proteins. As shown inFigure 1b, a single species was observed in the deconvoluted massspectra of rCRM, Pfenex CRM, and Vaxform CRM (58,541 ± 1 Da),whichwas consistent with the expectedmass of an unmodified andnonreduced CRM197 monomer (Table 2). EcoCRM™ contained 2species that differed in mass by ~131 Da. LC-MS peptide mapping

Figure 1. Primary structure analysis of 5 CRM197 molecules. (a) SDS-PAGE analysis of CRM197 molecules under nonreducing and reducing conditions. (b) Representative intact massspectrometry analysis of the nonreduced CRM197 molecules. (c) Representative LC peptide mapping chromatograms of nonreduced trypsin-digested CRM197 samples. The red andblue arrows indicate the 2 peaks containing Cys461-Cys273 and Cys186-Cys200 bonded peptides, respectively. The black and green arrows indicate an N-terminal peptide with orwithout a Met residue, respectively.

J.M. Hickey et al. / Journal of Pharmaceutical Sciences 107 (2018) 1806-1819 1811

confirmed these 2 species were the unmodified CRM197 monomerwithout (58,410± 1 Da) andwith (58,541± 1 Da) an N-terminalMetresidue (Fig.1c). The relative abundance of these 2 CRM197 species inEcoCRM™ was approximately 60% (with an N-terminal Met) and40% (without an N-terminal Met) using the corresponding peptideUV214nm peak areas. In addition to the unmodified CRM197monomerin C7 CRM, 2 additional higher MW species were observed (CRM197MW þ ~324 Da and CRM197 MW þ ~648 Da), which may representpost-translational modifications involving disaccharides (e.g.,potentially lactose glycations in the lyophilized samples).

The 5 different CRM197 molecules were also subjected to pep-tide mapping analysis to confirm their primary amino acid

sequence. As shown in Figure 1c, the trypsin-digested reversedphase ultra high performance liquid chromatography chromato-grams of the 5 CRM197 samples were similar. MS/MS analysis ofindividual trypsin or GluC digests confirmed a similar amino acidcomposition of the 5 CRM197 proteins (Supplemental Fig. S1). The2 native disulfide bonds in CRM197 (Cys187-Cys201 and Cys462-Cys472) were identified in each CRM197 molecule, and no non-native disulfide bonds or free Cys residues were observed in anysample. Multiple smaller peaks were observed in C7 CRM, andwhile the identities of these peaks are unknown, they couldpotentially represent residual protein contaminants notcompletely removed by the purification process.

Table 2Summary of the Key Structural Attributes (Physicochemical and In Vitro Antigenicity) of Recombinant CRM197 Proteins (in PBS pH 7.2) Produced From 3 Different ExpressionSystems and Obtained From 5 Different Manufacturers

Structural Attribute Analytical Method Measurement EcoCRM™ rCRM Pfenex CRM Vaxform CRM C7 CRM

Primary structure Mass spectrometry Intact protein mass (Da) 58,410 ± 1 58,410 ± 1 58,410 ± 1 58,410 ± 1 58,410 ± 1

58,541 ± 1a58,734 ± 1a

59,058 ± 1a

Post-translational modificationa N-terminal Met(þ131 Da)

e e e Glycation(þ324, þ648 Da)

Protein chargeheterogeneity

cIEF Main peak (pI 5.6-5.7) (%) 92 ± 0 99 ± 0 99 ± 0 95 ± 0 79 ± 1Acidic species (pI 5.5-5.6) (%) 7 ± 0 0 ± 0 0 ± 0 4 ± 0 21 ± 1Basic species (pI 5.7-5.8) (%) 1 ± 0 1 ± 0 1 ± 0 1 ± 0 0 ± 1

AEX Main peak (%) 92 ± 0 100 ± 0 100 ± 0 83 ± 0 75 ± 1Acidic species (%) 8 ± 0 0 ± 0 0 ± 0 17 ± 1 24 ± 1

Higher order structure CD Spectral minima at 10�C (nm) 208 & 222 208 & 222 208 & 222 208 & 222 208 & 222Tm (�C) 58.2 ± 1.0 57.9 ± 0.7 57.9 ± 0.3 59.6 ± 0.6 NAb

Intrinsic Trp fluorescence Peak emission maximumat 10�C (nm)

328 ± 1 329 ± 1 329 ± 1 330 ± 1 331 ± 1

Tm (�C) 41.2 ± 0.3 42.3 ± 0.3 42.3 ± 0.3 42.2 ± 0.6 43.2 ± 0.2Static light scattering Tonset (�C) 43.0 ± 0.6 43.4 ± 0.2 43.6 ± 0.3 44.6 ± 0.5 45.2 ± 0.7Extrinsic ANS fluorescence Peak emission intensity

(105 counts/s)441 ± 7 441 ± 9 437 ± 8 443 ± 11 411 ± 16

Tm (�C) 40.8 ± 0.5 41.5 ± 0.3 41.4 ± 0.5 41.3 ± 0.6 42.5 ± 0.2DSC Tonset (�C) 32.7 ± 0.3 35.0 ± 0.2 34.8 ± 0.2 33.5 ± 0.2 35.2 ± 0.6

Tm1 (�C) 42.0 ± 0.0 42.9 ± 0.0 42.8 ± 0.0 42.8 ± 0.2 44.0 ± 0.1Tm2 (�C) 51.3 ± 0.2 51.3 ± 0.1 51.1 ± 0.2 51.1 ± 0.3 51.7 ± 0.1

Aggregate/particleanalysis

SEC Monomer (%) 98 ± 0 99 ± 0 100 ± 0 71 ± 0 73 ± 1Aggregates (%) 1 ± 0 0 ± 0 0 ± 0 28 ± 0 13 ± 1Fragment (%) 2 ± 0 1 ± 0 0 ± 0 1 ± 0 14 ± 0

SV-AUC Monomer (%) 99 ± 0 99 ± 0 100 ± 0 70 ± 0 68 ± 1Aggregates (%) 1 ± 0 0 ± 0 0 ± 0 30 ± 0 11 ± 1Fragment (%) 0 ± 0 0 ± 0 0 ± 0 0 ± 0 21 ± 0

RMM (0.2-2 mm) Total particles after dilution(number/mL � 105)

1.7 ± 0.4 2.3 ± 0.1 1.6 ± 1.2 4.6 ± 0.7 5.1 ± 0.4

MFI (2-100 mm) Total particles after dilution(number/mL)

85 ± 56 133 ± 19 84 ± 9 109 ± 16 453 ± 163

Apparent solubility PEG precipitation assay PEGmidpt (%) 23.3 ± 0.2 23.0 ± 0.2 23.5 ± 0.1 21.5 ± 0.2 24.0 ± 0.2Apparent solubility in PBS pH 7.2

(mg/mL)30 ± 12 27 ± 9 31 ± 11 16 ± 10 34 ± 12

In vitro antigenicity ELISA log EC50 (mg/mL) 1.6 ± 0.9 2.3 ± 0.9 1.8 ± 0.3 1.3 ± 0.3 1.3 ± 0.3BLI Antibody ab8306 ka (105 M�1 s�1) 4.2 ± 1.9 4.4 ± 0.8 5.0 ± 2.0 4.0 ± 0.9 2.9 ± 0.9

kdis (10�3 s�1) 1.3 ± 0.3 1.4 ± 0.5 1.5 ± 0.4 1.6 ± 0.9 1.7 ± 0.4KD (nM) 3.4 ± 2.0 3.2 ± 1.3 3.3 ± 1.3 4.2 ± 2.6 6.1 ± 3.0

Antibody ab8307 ka (105 M�1 s�1) 4.4 ± 0.7 4.8 ± 0.8 5.4 ± 1.7 4.9 ± 1.0 3.5 ± 0.7kdis (10�3 s�1) 1.4 ± 0.4 1.3 ± 0.6 1.4 ± 0.4 1.6 ± 0.4 1.6 ± 1.0KD (nM) 3.3 ± 1.5 2.9 ± 1.7 2.8 ± 1.5 3.4 ± 1.0 4.9 ± 2.9

See Table 1 for sample information for each protein.a Post-translational modification.b Broad and slightly biphasic nature of the transition in C7 CRM prevented accurate calculation.

J.M. Hickey et al. / Journal of Pharmaceutical Sciences 107 (2018) 1806-18191812

Charge Heterogeneity Analysis of 5 CRM197 Molecules

Capillary isoelectric focusing (cIEF) andAEXwereutilized toassessthe overall charge heterogeneity of each CRM197 protein. As shown inFigure 2a, the pI of themajor peak in each CRM197 samplewas ~5.7 asmeasured by cIEF, which was consistent with previously reported pIof CRM197.15 The EcoCRM™ and Vaxform CRM samples contained asmall amount (4%-7%) of a more acidic species (pI 5.5-5.6) relative tothe main peak (Table 2), while the relative area of this species wasmore prominent in C7CRM sample (~21%). The elutionprofile of eachCRM197 protein through AEX was overall similar to their cIEF elec-tropherograms profiles inwhich the retention time of themajor peakwas consistent across all 5 samples. In addition, EcoCRM™ and C7CRMcontainedmeasurableacidic species,8%and24%, respectively, asmeasured by AEX. For Vaxform CRM sample, ~4% of the relativelymore acidic species were observed through cIEF, but this species wasfound to be more prominent through AEX analysis (~17%).

Higher Order Structure Analysis of 5 CRM197 Molecules

Far-UV CD spectroscopy was used to assess the overall secondarystructure of each CRM197 protein. As shown in Figure 3a, double

minima at 208 and 222 nm were observed at 10�C indicating a pri-marilya-helical secondary structure, consistentwith theX-raycrystalstructure of CRM197 (PDB ID: 4AE0).11 Second-derivative UV spec-troscopy, intrinsic Trp fluorescence spectroscopy, and extrinsic ANSfluorescence spectroscopy analyses were performed to evaluate andcompare the overall tertiary structures of the 5 CRM197 molecules at10�C. Six major peaks were observed for each CRM197 sample bysecond-derivative UV spectroscopy, which correspond to differentlocal polarity environments of the 3 aromatic amino acids (phenyl-alanine [250-270 nm], tyrosine [250-290 nm], and tryptophan[250-300 nm]) in each protein (Fig. 3b). The similar positions of thesepeaksbetween the5CRM197proteins suggest a similaroverall tertiarystructure via a similar environment around the average aromatic acidresidues. The intrinsic Trp and extrinsic ANS fluorescence spectrafurther support an overall similar tertiary structure across the 5samples. As shown in Figure 3c, the intrinsic Trp fluorescencewavelengthofmaximumintensity (lmax)was~330nmat10�C,whichindicated a similar hydrophobic environment for the average Trpresidue in each CRM197 molecule. Finally, the extrinsic ANS fluores-cence emission peak maximum and peak intensity at 10�C was notnotably different between the 5 proteins (Fig. 3d), which suggested acomparable overall surface hydrophobicity across the 5 proteins.

Figure 2. Charge heterogeneity analysis of 5 CRM197 molecules. The overall charge heterogeneity of the CRM197 samples was compared through both (a) cIEF and (b) AEX. Themigration of 2 pI markers (4.7 and 7.1) is indicated in the cIEF electropherogram for C7 CRM.

J.M. Hickey et al. / Journal of Pharmaceutical Sciences 107 (2018) 1806-1819 1813

Size and Aggregate/Particle Analysis of 5 CRM197 Molecules

The size analysis of monomer, aggregate, and fragment spe-cies in each CRM197 sample was assessed under nondenaturingsolution conditions by SV-AUC and SEC. The sedimentationcoefficient value of the major species in each CRM197 sample was3.9 ± 0.1 S, consistent with a CRM197 monomer (Fig. 4a). Therelative area of this monomeric peak ranged from 68% to 99% ofeach sample's total peak area (Table 2). Furthermore, some largerspecies (e.g., dimer, trimer) with higher sedimentation values(~5-8 S) were observed in EcoCRM™, Vaxform CRM, and C7 CRMwhile smaller species (e.g., fragments), at ~1-3 S sedimentationvalues, were present in rCRM and C7 CRM samples. SEC analysisshowed comparable size distributions of monomer, aggregate,and fragment species in which EcoCRM™, rCRM, and Pfenex CRMwere primarily monomeric (~97%-100%), while Vaxform CRMand C7 CRM contained more abundant smaller fragment (1% and14%, respectively) and larger aggregate (28% and 13%, respec-tively) species (Fig. 4b).

In addition to size analysis by SV-AUC and SEC, the presence oflarger aggregates/particles was assessed through resonant massmeasurements (RMMs) and microflow imaging (MFI). RMM mea-sures the concentration and size distribution of submicron (0.2-2mm) particles, while MFI measures the same parameters for sub-visible (2-100 mm) particles. As reported in Table 2, a slightly higher

number of submicron particles were present in Vaxform CRM andC7 CRM (~5 � 105 particles/mL) compared to EcoCRM™, rCRM, andPfenex CRM (~2 � 105 particles/mL). The majority of submicronparticles were in the size range of 0.2-1.0 mm.MFI analysis indicateda low number of subvisible particles in EcoCRM™, rCRM, PfenexCRM, and Vaxform CRM (~100 particles/mL), while a slightly highernumber of subvisible particles were measured in C7 CRM (~450particles/mL). The majority of submicron and subvisible particles inall the 5 CRM197 samples were 0.2-1.0 mm and 2-5 mm, respectively.

Conformational Stability, Aggregation Propensity, and RelativeSolubility of 5 CRM197 Molecules

The conformational stability of each CRM197 proteins' overallsecondary structure was assessed through monitoring their molarellipticity values at 222 nm as a function of temperature (10�C-90�C) in PBS buffer, pH 7.2. As shown in Figure 5a, each of theCRM197 molecules (except C7 CRM) showed a sharp transition. TheC7 CRM sample showed a biphasic transition between 55�C and75�C, which may be due to heterogeneity in the protein sample (asobserved by SDS-PAGE and intact mass analysis; see Figs. 1 and 2).The calculated apparent thermal melting temperature values (Tm)of EcoCRM™, rCRM, Pfenex CRM, and Vaxform CRM were ~58�C(Table 2). For C7 CRM, the broad and slightly biphasic nature of thetransition prevented accurate calculation of an accurate value of Tm.

Figure 3. HOS analysis of 5 CRM197 molecules. (a) The overall secondary structures were evaluated by far-UV CD spectra, (b) the overall tertiary structures were monitored by both(b) second-derivative UV absorbance spectra and (c) intrinsic tryptophan fluorescence emission spectra. (d) The overall surface hydrophobicity of the samples was determined byextrinsic ANS fluorescence emission spectra. The 5 CRM197 samples were compared at 10�C in PBS pH 7.2.

J.M. Hickey et al. / Journal of Pharmaceutical Sciences 107 (2018) 1806-18191814

The conformational stability of each CRM197 proteins' overalltertiary structure versus temperature was evaluated through bothintrinsic and extrinsic fluorescence spectroscopy in PBS buffer, pH7.2. A single transition (red shift) was observed in the intrinsictryptophan fluorescence spectra peak position versus temperaturein all 5 samples with Tm values at ~42�C. The Tm values calculatedfrom changes in ANS fluorescence emission peak intensity versustemperature in the presence of the protein were also similar be-tween the 5 CRM197 molecules (~41�C). These results indicated arelative overall similar tertiary structure stability in terms of theaverage Trp residue and average surface hydrophobicity ofthe proteins, respectively. Furthermore, DSC was used to assess theoverall conformational stability of each of the 5 CRM197 samples asa function of temperature in PBS buffer, pH 7.2. As shown inFigure 5d, one major and one minor endothermic peaks wereobserved for all CRM197 molecules at ~40�C and 50�C, respectively.Overall, the DSC thermograms of rCRM, Pfenex CRM, and VaxformCRM samples appeared similar, but subtle differences wereobserved in EcoCRM™ (slightly lower stability of Tm1) and C7 CRM(slightly higher stability of Tm1) (Table 2).

The aggregation propensity and relative solubility of the 5CRM197 samples was also studied by static light scattering versustemperature and PEG precipitation curves, respectively. The lightscattered from each CRM197 sample was measured as a function oftemperature (10�C-65�C) in PBS buffer. As shown in Figure 5e, onesharp transition was observed in all samples that began (Tonset) at~45�C. The relative apparent solubility (thermodynamic activity)profiles of the 5 CRM197 molecules were assessed using PEG pre-cipitation assay as described previously.18 As shown in Figure 5f, theshape of the PEG curves (concentration of protein, [mg/mL] versus

percentage [w/v] PEG-6000) of Vaxform CRM appeared to have ashifted profile compared to the 4 other CRM197 samples (C7 CRM,rCRM, EcoCRM™, and Pfenex CRM). A possible reason for the morerapid precipitation of Vaxform CRM as a function of PEG concen-tration is the relatively higher levels of aggregates present in thesample as observed by SEC and SV-AUC. From the PEG versusprotein concentration curves, 2 different relative solubilityparameters (PEGmidpt and apparent solubility) were calculated.PEGmidpt value is a parameter to quickly and reliably compare datasets while the apparent solubility value is extrapolated from eachprecipitation curve and assumes no interaction between the PEGand protein, which is not necessarily the case for all proteins.17 ThePEGmidpt value was lowest for Vaxform CRM (~21.5%), while rCRM,EcoCRM™, Pfenex CRM, and C7 CRM had similar PEGmidpt values(~23%; Table 2). A comparison of the apparent solubility values ofeach of the proteins showed a similar trend as seen in the PEGmidptvalues, in which Vaxform appeared to have a lower apparent sol-ubility (~16 mg/mL) compared to C7 CRM, rCRM, EcoCRM™, andPfenex CRM (~27-34 mg/mL) in PBS buffer, pH 7.2.

Physical Stability Profiles of 5 CRM197 Molecules During pH/Temperature Stresses

The physical stability profiles of the 5 CRM197 molecules wereevaluated in PBS buffer under a wide range of pH (pH 5.8-8.0) andtemperature (10�C-90�C) conditions. For each pH and temperature,the overall secondary and tertiary structural integrity, overallconformational stability, and aggregation propensity of eachCRM197 sample were measured using CD, intrinsic tryptophanfluorescence spectroscopy, DSC, and static light scattering,

Figure 4. Size analysis of 5 CRM197 molecules. The size and distribution of monomer,aggregate, and fragment species in each of the CRM197 samples were comparedthrough (a) SV-AUC and (b) SEC. The elution times and MW values of gel filtrationstandard proteins are shown above the SE chromatograms.

J.M. Hickey et al. / Journal of Pharmaceutical Sciences 107 (2018) 1806-1819 1815

respectively. Representative data sets for C7 CRM in PBS buffer atdifferent pH values from these different biophysical techniques areshown in Figure 6a (see Supplemental Figs. S2-S5 for the corre-sponding biophysical data sets for each of the 5 CRM197 proteins).The molar ellipticity at 222 nm versus temperature of C7 CRMshowed a dramatic pH dependence with a single thermal transitionevent at pH values of s 5.8-7.2 (with Tm values ranging between~37�C and 62�C). At pH 7.6 and 8.0, however, a distinct thermaltransition was not observed, which prevented accurate calculationof Tm. Clear differences in the stability of the overall secondarystructurewere observed in the CRM197 samples (i.e., more stable pH� 7.2 and less stable at pH 5.8-6.8). An increase in Trp fluorescencepeak position of C7 CRM spectra was observed with increases intemperature for each pH, indicating a pH-dependent alteration ofthe protein's overall tertiary structure. The observed Tmvalues werehighest at pH 6.8-7.2 (42�C-44�C), slightly lower at pH 6.3, 7.6, or8.0 (41�C-42�C), and distinctly lower at pH 5.8 (34�C; Table 2). DSCresults indicated the highest overall conformational stability of C7CRM occurred at pH 6.8-7.2 and was slightly lower at the other

solution pH conditions tested. Finally, the temperature at which C7CRM began (Tonset) to scatter more light (i.e., aggregate) was lowestat pH 5.8 (29�C) and increased with an increasing solution pH(Tonset was 47�C at pH 8.0) as measured by static light scattering.Similar pH-dependent physical stability profiles were observedwith EcoCRM™, rCRM, Pfenex CRM, and Vaxform CRM (seeSupplementary Figs. S2-S5 and Supplemental Table S1).

Using the large biophysical data sets from the aforementioned 4techniques, 2 data visualization tools were utilized, an EPD and aradar chart to better compare the overall physical stability profilesof each CRM197 protein as a function of pH and temperature.19,20

The EPD displayed changes in the structural characteristics ofeach CRM as measured by the 4 different biophysical techniques(CD, intrinsic tryptophan fluorescence, static light scattering, andDSC), as a function of temperature and pH. The EPD plots for all 5CRM197 molecules indicated the presence of 6 distinct coloredphases (designated as regions I-VI) (Figure 6b and SupplementalFig. S6). Each color represented regions with similar structuralcharacteristics, and changes in color reflected structural alterationsin each CRM197 protein as measured by the biophysical measure-ments. In addition to the native-like region at low temperatures(region I), 5 structurally altered protein states were observed bothwith EPD. Region II denotes an overall altered conformation ofCRM197 molecule, region III represents a structurally altered statebeginning to aggregate, region IV corresponds to an aggregatedstate of protein, region V represents more extensively aggregatedand structurally altered state, and region VI shows substantialsecondary structure alterations.

Radar plots were also constructed to visualize the same physicalstability data sets for each of the 5 CRM197 proteins as a function ofsolution pH and temperature. A radar plot displays the results of thedifferent biophysical analyses in a polygon, in which each vertex ismapped to a particular biophysical technique and concentric circlesrepresent the signal variation/strength. As shown in Figure 6c, thesignals for each of the biophysical techniques (CD molar ellipticity,intrinsic tryptophan fluorescence, DSC, and static light scattering)were minimal in region I, representing the native-like state for theCRM197 molecule. Region II represents CRM197 in an overall struc-turally altered conformation as indicated by an increase in the DSCsignal and alterations in the intrinsic tryptophan fluorescence peakposition. Region III was assigned based on the loss of tertiarystructure as monitored by intrinsic tryptophan fluorescence.Region IV represents CRM197 in aggregated state as indicated by anincrease in the light scattering signal. Region V corresponds to amore aggregated and structurally altered state as seen from notablechanges in the signals from light scattering and CD, and finally,region VI reflects loss of secondary structure as measured by CD.Overall, a comparison of EPD and radar charts of multiple CRM197molecules tested here indicated similar biophysical stability pro-files versus pH and temperature. In addition, the total area for thenative-like state as a function of pH and temperature, region I, wassimilar (35%-37%) for all 5 CRM197 molecules.

In Vitro Antigenicity Analysis of 5 CRM197 Molecules

For comparing the in vitro antigenicity of each CRM197 molecule,the steady-state and preesteady-state kinetics were measuredthrough an ELISA (using polyclonal CRM197 antibodies) and biolayerinterferometry (BLI; using individual measurements with 5different monoclonal CRM197 antibodies), respectively. While a vi-sual comparison of the binding curves of the CRM197 molecules tocommercially available polyclonal CRM197 antibodies suggestedsubtle differences in antigenicity between the 5 CRM197 molecules(Fig. 7a), a statistical analysis indicated, however, that the steady-state interactions of each CRM197 molecule with the polyclonal

Figure 5. Conformational stability, aggregation propensity, and relative solubility analysis of the 5 CRM197 molecules. The conformational stability profiles of the CRM197 samples inPBS buffer (pH 7.2) were compared as a function of temperature (10�C-90�C) through (a) far-UV CD, (b) intrinsic tryptophan fluorescence spectroscopy (MSM peak position),(c) extrinsic ANS peak intensity in the presence of each protein, and (d) DSC. (e) The aggregation propensity of the CRM197 samples were compared through static light scattering asa function of temperature (10�C-90�C). (f) The relative apparent solubility of the CRM197 samples was compared as a function of an increasing concentration (0%-33% w/v) of amolecular crowding agent (PEG-6000). MSM, mean spectral center of mass.

J.M. Hickey et al. / Journal of Pharmaceutical Sciences 107 (2018) 1806-18191816

antibodies were not significantly (p >0.05) different from eachother (Table 2).

The preesteady-state kinetics of between each CRM197 and 5mAbs were measured using BLI. Before the BLI assay, the concen-trations of each of the 5 mAbs required to monitor binding to C7CRM was optimized with an ELISA. As shown in Figure 7b, affinitydifferences were observed between the 5 antibodies and C7 CRM asmeasured by the ELISA. The 5 mAbs were categorized into 3 groups,one higher affinity mAb (AB53827), 2 intermediate affinity mAbs(AB8308 and AB8310), and 2 lower affinity mAbs (AB8307 andAB8306). The preesteady-state kinetics (ka, kdis, and KD) were thenmeasured between each CRM197 molecule and mAb using BLI. Arepresentative association and dissociation binding curve for C7CRM and a lower affinity mAb (AB8306) as measured by BLI isshown in Figure 7c. Similar to the ELISA results with C7 CRM andthe 2 lower affinity mAbs, the rate of association (3-5 � 105

M�1 s�1), dissociation (1-2 � 10�3 s�1), and binding affinity (3-6nM) values measured by BLI with either mAb was comparable be-tween all the 5 CRM197 molecules (Table 2). For the 3 other mAbs(AB8308, AB8310, and AB53827), preesteady-state kinetics couldnot be accurately measured by this technique owing to their strongaffinity (KD < nM) for each of the 5 CRM197 molecules.

Discussion

In this study, the physicochemical and immunological bindingproperties of recombinant CRM197 from the traditional expressionsystem (C diptheriae) and 2 heterologous systems (E coli and P flu-orescens) were evaluated and compared as part of an analyticalcomparability assessment. A comprehensive suite of analytical

techniques assessed numerous structural attributes including pri-mary structure and post-translational modifications, higher orderstructure (HOS), overall surface hydrophobicity and charge het-erogeneity, and aggregate/particle formation from small multimersto larger subvisible particles. Overall, the 5 CRM197 moleculesexhibited overall highly similar physicochemical characteristics(with some exceptions in terms of subtle physicochemical differ-ences as described in the following paragraph) in a common buffer(PBS, pH 7.2). These similarities were also conserved in terms ofconformational stability under different stress conditions (pH/temperature changes) and for relative solubility (in the presence ofmolecular crowding agent, PEG). Finally, each CRM197 displayedessentially equivalent in vitro antigenicity in terms of binding to aseries of antibodies (polyclonal and mAbs specific for CRM197).Together, these observations indicate that the various expressionsystems produced generally comparable CRM197 molecules.

The largest differences noted between the CRM197 proteins werethe presence of product-related variants/impurities. The nature ofthese species in some cases was determined, and in other cases,will require additional analysis. For example, N-terminal hetero-geneity with a Met residue observed in EcoCRM™ was quantita-tively described (~60% vs. 40% for CRM197 with vs. withoutN-terminal Met) and was not necessarily unexpected, given a lowercatalytic efficiency of methionyl aminopeptidase in some strains ofE coli.22,23 Two low-abundant species in C7 CRM contained post-translational modifications with þ324 and þ648 Da, which werepossibly caused by nonenzymatic glycation between a reducingsugar (lactose) within the formulation buffer and the protein,especially upon lyophilization, as has been observed with otherfreeze-dried proteins.24-26 The C7 CRM sample also contained

Figure 6. Physical stability analysis of the 5 CRM197 molecules as a function of pH and temperature. (a) The molar ellipticity at 222 nm, intrinsic Trp fluorescence MSM peakposition, DSC, and static light scattering at 295 nm for C7 CRM in PBS buffer at 6 different pHs (5.8, 6.3, 6.8, 7.2, 7.6, and 8.0) were measured as a function of temperature(10�C-90�C). (b) An EPD and radar chart for C7 CRM were generated from the data in panel (a). Six distinct biophysical states of C7 CRM (regions I-VI) were observed as a function ofpH and temperature (see Results section for more details). (c) Summary of calculated radar charts of each of the 5 CRM197 molecules. The contribution of each biophysical techniquetoward the different structural states of each CRM197 is indicated on each vertex in the key. MSM, mean spectral center of mass.

J.M. Hickey et al. / Journal of Pharmaceutical Sciences 107 (2018) 1806-1819 1817

Figure 7. In vitro antigenicity analysis of the 5 CRM197 molecules. (a) ELISA reactivity of each CRM197 protein as a function of protein concentration (0.005-100 mg/mL) to a polyclonalCRM197 antibody. (b) ELISA reactivity of C7 CRM as a function of protein concentration (0.005-10 mg/mL) to 5 different monoclonal CRM197 antibodies. (c) Representative associationand dissociation binding curves of 6 concentrations (4-200 nM) of C7 CRM to a monoclonal CRM197 antibody (AB8306) as measured by octet analysis.

J.M. Hickey et al. / Journal of Pharmaceutical Sciences 107 (2018) 1806-18191818

lower and higher MW species that were observed by 2 differentorthogonal sizing techniques (SV-AUC and SEC). The presence ofthese fragment and soluble aggregate species may have arisenduring the bulk manufacturing process, lyophilization, and subse-quent storage. Analysis of material from each downstream pro-cessing step would be needed to identify the mechanism(s) thatinduced the formation of these species. In addition, the VaxformCRM sample contained relatively elevated levels (~29% by SV-AUC)of higher MW species, which could potentially be a consequence ofthe material's age and long-term storage (~11 years). Furthermore,although speculative, the higher MW species in this sample maypossibly be composed of the observed acidic peak through AEXbecause the cIEF results indicated that the levels of the acidicspecies in Vaxform CRM were greatly reduced. Unlike AEX, cIEFassess the overall charge (pI) of a generally unfolded moleculefollowing migration through a pH gradient; therefore, the larger(noncovalently associated) species in Vaxform CRM would likelydissociate during analysis. The more acidic species in C7 CRM,however, were observed at a similar amount (21%-24%) by bothAEX and cIEF. This acidic peak in C7 CRM may represent CRM197containing the lactose glycan(s) variants that were observed byintact mass analysis. Finally, the composition of the low-abundantacidic species (7%-8%) observed in EcoCRM™ through AEX andcIEF could represent chemically modified CRM197 (i.e., Asn deami-dation), which has been suggested previously,15 and determinationof these species is suggested for future work. In addition, potentialdifferences in process-related impurities (e.g., residual host cellproteins, residual DNA) between CRM197 from different manufac-turers should also be examined as part of future comparabilitywork.

In this work, we have implemented state-of-the-art analyticalmethods to comprehensively characterize the structural integrity,conformational stability, and relative solubility of CRM197 bulkmaterials. Mass spectrometry analysis can rigorously define theprimary structure and post-translational modifications of proteins,ensuring the structural integrity and defining the presence ofstructural variants. A current technical challenge is the use ofanalytical tools to rigorously characterize the HOS of protein. Thiswork did not utilize higher resolution methods such as X-raycrystallography or hydrogen deuterium exchange mass spectrom-etry because such methods are not commonly available, expensive,and are relatively time consuming. Instead, the key structuralattributes of the CRM197moleculeweremonitored using commonlyavailable analytical instruments. In addition, the biophysical tech-niques utilized in this work, combinedwith data visualization tools,

can monitor the conformational stability of the CRM197 in acomprehensive manner as a function of stress conditions such aspH and temperature. The stability of protein structure can be asensitive probe for the overall higher order structural integrity ofproteins, and it has been suggested such an approach is animportant part of protein comparability studies and biosimilarityanalyses.27-29 In this work, the resulting EPDs were similar overallfor the 5 CRM197 molecules, indicating a similar structural integrityand conformational stability. Not surprisingly, all the CRM197 mol-ecules showed a notable decrease in the overall conformationalstability profile as the solution pH was lowered from 6.8 to 5.8, aresult which is consistent with the known conformational changeof diphtheria toxin at pH 5.0 related to insertion of the toxin intothe endosomal membrane.30

Interestingly, despite the noted structural heterogeneity differ-ences noted between the 5 CRM197 samples, their binding reactivityto polyclonal ormultiple individualmAbswas equivalent. This resultcould be due to the low levels of these variants in the preparationsand an associated lack of sensitivity to detect small differences inthese types of binding assays. Future work would need to focus onthe isolation of the variants and then a more direct comparison oftheir in vitro binding properties. Furthermore, the specific epitopesrecognized by the 5 anti-CRM197 mAbs described in this study arecurrently unknown, but in general, in vitro immunological bindingassays are a critical probe of the overall structural integrity of bi-ologicals. Owing to the challenges in rigorously defining all aspectsof HOS of proteins with current analytical technology (as describedpreviously), functional or potency assays are required to ensure tothe overall structure/function of protein vaccine or drug candidatesare maintained during development. To this end, biological potencyassays (either in vitro assays or in vivo animal studies depending onthe nature of the vaccine antigen) remain a cornerstone of theoverall analytical strategy to ensure comparability and bio-similarity31 In the case of the carrier protein CRM197 for use in aspolysaccharide conjugate vaccines, in vitro binding assays are auseful tool for probing the overall conformational integrity of themolecule as part of comparability assessments (especially usingmAbs with known binding to specific epitopes to a reference withknown in vivo activity as part of future work). Nonetheless, todirectly compare the ability of the different CRM197 molecules togenerate immune responses, an in vivo biological potency assaywould bemore suited for the final polysaccharide conjugate vaccineand not the carrier protein alone.

These findings lead to question if some or any of these subtlephysiochemical differences/impurities observed in the bulk CRM197

J.M. Hickey et al. / Journal of Pharmaceutical Sciences 107 (2018) 1806-1819 1819

have an effect (if any) on the preparation of the final drug productof conjugate vaccine. From the limited literature available, conju-gate vaccines containing CRM197 from different expression systemselicit similar immune responses in animal models.5,32 It is currentlyunknownwhich analytical readout(s) performed in this work couldpredict detrimental effects on subsequent conjugation of the pro-tein to various polysaccharides and its effects on the safety andefficacy of the conjugate vaccine drug product. Furthermore, forindustrial use of CRM197 as a carrier protein in conjugate vaccines,the critical-to-quality parameters have to be established withprospective acceptance criteria to ensure that the carrier proteinmeets the quality requirements for robust and sustainablemanufacturing. Typically, this informationwould be available to themanufacturer in a pharmacopeia/theWHO Technical Review Seriesmonograph. Unfortunately, such a monograph does not currentlyexist for CRM197 carrier protein. It is also important for the industryto have this critical-to-quality parameters quantified, so that thesecould be used for in-depth characterization, lot release, and sta-bility analysis. While additional studies are needed to determine ifthe observed subtle differences between different sources ofcommercially available CRM197 influence the development of con-jugate vaccines, the presented results provide a basis to developand establish specifications for this important carrier protein in thefuture.

Acknowledgments

This work was supported by PATH. The authors would like tothank Dr. Ronald Toth at the University of Kansas for performing theSV-AUC analysis. The authors would also like to thank Dr. GarryMorefield for donating the Vaxform CRM protein. The ELISA assayusing polyclonal antibodies was developed by Natalia Oganesyan atFina BioSolutions LLC.

References

1. Grabenstein JD, Klugman KP. A century of pneumococcal vaccination researchin humans. Clin Microbiol Infect. 2012;18 Suppl 5:15-24.

2. Strugnell R, Zepp F, Cunningham AL, Tantawhichien T. Vaccine antigens. In:Garçon N, Stern PL, Cunningham AL, Stanberry LR, eds. Understanding ModernVaccines: Perspectives in Vaccinology. Amsterdam, The Netherlands: Elsevier BV;2011:61-88.

3. Shinefield HR. Overview of the development and current use of CRM(197)conjugate vaccines for pediatric use. Vaccine. 2010;28(27):4335-4339.

4. Hickey JM, Sahni N, Toth RT, et al. Challenges and opportunities of using liquidchromatography and mass spectrometry methods to develop complex vaccineantigens as pharmaceutical dosage forms. J Chromatogr B Analyt Technol BiomedLife Sci. 2016;1032:23-38.

5. Brady C, Killeen K, Taylor W, Patkar A, Lees A. Carrier protein outsourcing: anew paradigm for rapidly translating novel conjugate vaccines into the clinic.BioProcess Int. 2012;10(10):50-55.

6. Levine OS, Bloom DE, Cherian T, et al. The future of immunisation policy,implementation, and financing. Lancet. 2011;378(9789):439-448.

7. Knuf M, Kowalzik F, Kieninger D. Comparative effects of carrier proteins onvaccine-induced immune response. Vaccine. 2011;29(31):4881-4890.

8. Goffin P, Dewerchin M, De Rop P, Blais N, Dehottay P. High-yield production ofrecombinant CRM197, a non-toxic mutant of diphtheria toxin, in the periplasmof Escherichia coli. Biotechnol J. [Epub ahead of print] 2017.

9. Zhou J, Petracca R. Secretory expression of recombinant diphtheria toxinmutants in B. subtilis. J Tongji Med Univ. 1999;19(4):253-256.

10. Broker M, Costantino P, DeTora L, McIntosh ED, Rappuoli R. Biochemicaland biological characteristics of cross-reacting material 197 CRM197, anon-toxic mutant of diphtheria toxin: use as a conjugation protein in

vaccines and other potential clinical applications. Biologicals. 2011;39(4):195-204.

11. Malito E, Bursulaya B, Chen C, et al. Structural basis for lack of toxicity ofthe diphtheria toxin mutant CRM197. Proc Natl Acad Sci U S A.2012;109(14):5229-5234.

12. Ravenscroft N, Costantino P, Talaga P, Rodriguez R, Egan W. Glycoconjugatevaccines. In: Nunnally BK, Turula VE, Sitrin RD, eds. Vaccine Analysis: Strategies,Principles, and Control. Berlin, Heidelberg: Springer Berlin Heidelberg; 2015:301-381.

13. Broker M, Berti F, Costantino P. Factors contributing to the immunogenicityof meningococcal conjugate vaccines. Hum Vaccin Immunother. 2016;12(7):1808-1824.

14. Lockyer K, Gao F, Derrick JP, Bolgiano B. Structural correlates of carrier proteinrecognition in tetanus toxoid-conjugated bacterial polysaccharide vaccines.Vaccine. 2015;33(11):1345-1352.

15. Rustandi RR, Peklansky B, Anderson CL. Use of imaged capillary isoelectricfocusing technique in development of diphtheria toxin mutant CRM197.Electrophoresis. 2014;35(7):1065-1071.

16. Loughney JW, Ha S, Rustandi RR. Quantitation of CRM197 using imagedcapillary isoelectric focusing with fluorescence detection and capillary West-ern. Anal Biochem. 2017;534:19-23.

17. Gibson TJ, McCarty K, McFadyen IJ, et al. Application of a high-throughputscreening procedure with PEG-induced precipitation to compare relativeprotein solubility during formulation development with IgG1 monoclonalantibodies. J Pharm Sci. 2011;100(3):1009-1021.

18. Toprani VM, Joshi SB, Kueltzo LA, Schwartz RM, Middaugh CR, Volkin DB.A micro-polyethylene glycol precipitation assay as a relative solubilityscreening tool for monoclonal antibody design and formulation development.J Pharm Sci. 2016;105(8):2319-2327.

19. Maddux NR, Joshi SB, Volkin DB, Ralston JP, Middaugh CR. Multidimensionalmethods for the formulation of biopharmaceuticals and vaccines. J Pharm Sci.2011;100(10):4171-4197.

20. Kim JH, Iyer V, Joshi SB, Volkin DB, Middaugh CR. Improved data visualizationtechniques for analyzing macromolecule structural changes. Protein Sci.2012;21(10):1540-1553.

21. Giraldo J, Vivas NM, Vila E, Badia A. Assessing the (a)symmetry ofconcentration-effect curves: empirical versus mechanistic models. PharmacolTher. 2002;95(1):21-45.

22. Natarajan C, Jiang X, Fago A, Weber RE, Moriyama H, Storz JF. Expression andpurification of recombinant hemoglobin in Escherichia coli. PLoS One.2011;6(5):e20176.

23. Liao YD, Jeng JC, Wang CF, Wang SC, Chang ST. Removal of N-terminalmethionine from recombinant proteins by engineered E. coli methionineaminopeptidase. Protein Sci. 2004;13(7):1802-1810.

24. Costantino HR, Carrasquillo KG, Cordero RA, Mumenthaler M, Hsu CC,Griebenow K. Effect of excipients on the stability and structure of lyophi-lized recombinant human growth hormone. J Pharm Sci. 1998;87(11):1412-1420.

25. Zhang Q, Ames JM, Smith RD, Baynes JW, Metz TO. A perspective on theMaillard reaction and the analysis of protein glycation by mass spectrom-etry: probing the pathogenesis of chronic disease. J Proteome Res. 2009;8(2):754-769.

26. Toprani VM, Hickey JM, Sahni N, et al. Structural characterization and physi-cochemical stability profile of a double mutant heat labile toxin protein basedadjuvant. J Pharm Sci. 2017;106(12):3474-3485.

27. Alsenaidy MA, Jain NK, Kim JH, Middaugh CR, Volkin DB. Protein comparabilityassessments and potential applicability of high throughput biophysicalmethods and data visualization tools to compare physical stability profiles.Front Pharmacol. 2014;5:39.

28. More AS, Toprani VM, Okbazghi SZ, et al. Correlating the impact of well-definedoligosaccharide structures on physical stability profiles of IgG1-Fc glycoforms.J Pharm Sci. 2016;105(2):588-601.

29. Kim JH, Joshi SB, Tolbert TJ, Middaugh CR, Volkin DB, Smalter Hall A.Biosimilarity assessments of model IgG1-Fc glycoforms using a machinelearning approach. J Pharm Sci. 2016;105(2):602-612.

30. Leka O, Vallese F, Pirazzini M, Berto P, Montecucco C, Zanotti G. Diphtheriatoxin conformational switching at acidic pH. FEBS J. 2014;281(9):2115-2122.

31. Federici M, Lubiniecki A, Manikwar P, Volkin DB. Analytical lessons learnedfrom selected therapeutic protein drug comparability studies. Biologicals.2013;41(3):131-147.

32. Lees A, Oganesyan N, Krauss I, Nguyen D. EcoCRMTM: affordable CRM197 vac-cine Carrier protein. Poster Presented at: Protein Engineering Summit, Boston,MA, May 1 2017.