Charge Variant Analysis of Monoclonal Antibodies Using DirectCoupled pH Gradient Cation Exchange Chromatography to High-Resolution Native Mass SpectrometryFlorian Füssl,† Ken Cook,‡ Kai Scheffler,§ Amy Farrell,† Stefan Mittermayr,† and Jonathan Bones*,†,∥

†NIBRT−The National Institute for Bioprocessing Research and Training, Foster Avenue, Mount Merrion, Blackrock, Co. Dublin,A94 X099, Ireland‡Thermo Fisher Scientific, Stafford House, 1 Boundary Park, Hemel Hempstead, HP2 7GE, United Kingdom§Thermo Fisher Scientific, Dornierstrasse 4, 82110 Germering, Germany∥School of Chemical and Bioprocess Engineering, University College Dublin, Belfield, Dublin 4, D04 V1W8, Ireland

*S Supporting Information

ABSTRACT: Charge variant analysis (CVA) of monoclonal antibodies (mAbs) using cation exchange chromatography isroutinely used as a fingerprint of the distribution of posttranslational modifications present on the molecule. Traditional salt orpH based eluents are not suited for direct coupling to mass spectrometry due to nonvolatility or high ionic strength. This makesfurther analysis complicated when an alteration in the charge variant profile or the emergence of an additional peak isencountered. Here, the use of pH gradient elution using volatile, low ionic strength buffers is reported with direct coupling tohigh-resolution Orbitrap mass spectrometry. The development of a universal method based on pH elution was explored using anumber of mAb drug products. Optimized methods facilitated the separation and identification of charge variants includingindividual glycoforms of the mAbs investigated using the same buffer system but with tailored gradient slopes. The developedmethod represents an exciting advance for the characterization of biopharmaceuticals as intact entities through the combinationof native charge variant separations with high-resolution native mass spectrometry.

Monoclonal antibody (mAb) drug products, derived fromindustrial scale bioprocessing of mammalian cells, existin a variety of different isoforms due to posttranslationalmodification (PTM) events. These occur both enzymaticallyand nonenzymatically following biosynthesis of the nascentprotein within the cell and subsequent secretion into theconditioned media.1 Regulatory guidelines require character-ization of the primary sequence and PTMs present on mAbs toensure that the recombinant therapeutic protein meetspredefined quality specifications.2 Certain primary sequencevariations and PTMs present on mAbs, such as C-terminallysine truncation, N-terminal pyroglutamate formation, sialyla-tion present on N-glycosylation, and deamidation of asparagineor glutamine residues alter the isoelectric point (pI) of themolecule, resulting in the generation of a series of acidic andbasic charge variants of the mAb.3 Charge variant patterns,

traditionally monitored using cation exchange chromatographyor capillary isoelectric focusing, act as a fingerprint of themanufacturing process and deviations in peak areas or theappearance of new peaks require in depth characterization toensure that no additional or undesired PTM has beenintroduced.4−7 Such alterations in the molecular surface chargedistribution can affect antigen or receptor binding or may causeconformational changes that may increase the risk of aggregateformation and potential induction of immunogenicity.8−10

Current strategies to identify alterations in the charge variantpatterns involve fractionation of the peak of interest, multilevelcharacterization using peptide mapping, intact mass analysis,

Received: December 15, 2017Accepted: March 1, 2018Published: March 1, 2018

Article

pubs.acs.org/acCite This: Anal. Chem. 2018, 90, 4669−4676

© 2018 American Chemical Society 4669 DOI: 10.1021/acs.analchem.7b05241Anal. Chem. 2018, 90, 4669−4676

Dow

nloa

ded

via

UN

IV C

OL

G D

UB

LIN

on

Nov

embe

r 19

, 202

0 at

20:

47:2

5 (U

TC

).Se

e ht

tps:

//pub

s.ac

s.or

g/sh

arin

ggui

delin

es f

or o

ptio

ns o

n ho

w to

legi

timat

ely

shar

e pu

blis

hed

artic

les.

pubs.acs.org/achttp://pubs.acs.org/action/showCitFormats?doi=10.1021/acs.analchem.7b05241http://dx.doi.org/10.1021/acs.analchem.7b05241

and N-glycan profiling followed by functional assessment usingreceptor binding or other bioassays, depending on the locationof the modification within the primary sequence.11,12 Theability to hyphenate charge variant analysis (CVA) directly tohigh-resolution mass spectrometry (MS) is highly desired sothat potential modifications in peaks of interest can becharacterized during the analysis based on a change in theproteins average mass and retention time.Charge variant separations are routinely performed using

weak or strong cation exchange (CEX) stationary phases withcarboxylate or sulfonate functionalities, respectively.4 Elutioncan be performed by displacement using either a gradient ofincreasing salt concentration or by the generation of a pHgradient that results in analyte desorption when the mobilephase pH equals the apparent pI of each variant. The workingprinciple is similar to chromatofocusing with the difference thatproteins are not focused along an immobilized pH-gradient butelute as the buffer pH changes over time.13−17 A limitation toboth these traditional approaches is the use of non-MScompatible salts and components in the mobile phase whichprevents direct coupling of CVA to MS, often requiring the useof a desalting intermediary step such as reversed-phasechromatography.18 An additional limitation of this approachis denaturation of the protein, resulting in the generation ofcomplex mass spectra containing highly charged mAb ions dueto the loss of tertiary structure and exposure of chargedresidues. The potential for direct coupling of CVA with nativemass spectrometry, wherein tertiary structure is maintained andthe charge states observed by MS are reflective of the surfacecharge of the mAb in solution, is highly desirable.19,20 Theability to separate and analyze large and complicated moleculessuch as mAbs in the native state is attractive as alterations in thesurface charge distribution may provide an insight intopotential changes in molecular conformation which may affectfunction, thereby providing the analyst with the tools topotentially link sequence, structural, and functional informa-tion.21

Limited reports of the direct hyphenation of CVA to MSexist in the literature. Salt gradient elution using an increasingconcentration of ammonium acetate at pH 7 was used for theCVA-MS analysis of lysozyme and β-interferon allowing for thedetection of modifications such as deamidation inducedfollowing molecular stress.22,23 CVA-MS analysis of IgG2mAbs was described using a single pH unit gradient change ona weak cation exchanger using ammonium hydroxide eluentscontaining 20% methanol to promote desolvation.24 Low-resolution native spectra of the intact mAbs were obtained onthe time-of-flight (TOF) instrument used; however, the lowinstrument resolution and a high degree of ammoniumadduction present on the ions prevented deconvolution andspectral annotation.24 CVA-MS using ammonium acetate saltgradient elution applying solutions of different pH and saltconcentration was also recently reported using a middle-upstrategy. In this respect, normal and aged IgG samples weredigested with IdeS protease and the resulting (Fc/2)2 andF(ab′)2 fragments were separated on a strong cation exchangerhyphenated to QTOF-MS.25 The native structures, including apreserved interaction of the (Fc/2)2 domain as dimer postdigestion, were determined with good mass accuracy. Spectralquality however was poor, presumably as a result of insufficientdesolvation or high levels of adduct formation in the highammonium containing eluent. Acidic species were predom-inantly on the F(ab′)2 fragment and consisted of oxidation and

deamidation, which were observed to be significantly increasedin the aged sample.25 In 2015 Stoll et al. developed a 2D-LCapproach for the intact and middle-up characterization ofrituximab charge variants employing CEX chromatographybefore RPLC−TOF-MS, which in a follow-up study wasutilized for the middle-up characterization of several mAbs andtheir biosimilars by Sorensen et al.26,27 Whereas this methodhas successfully been proven applicable to the identification ofcharge variants such as different C-terminal lysine forms, itrequires an elaborate instrumental setup due to the 2D-LCapproach employed.In the current report, we investigate the use of low ionic

strength pH gradient elution of mAbs from a strong cationexchanger coupled to high-resolution native Orbitrap massspectrometry for high-resolution, high-definition character-ization of mAb charge variants. pH linearity across the gradientrange was first investigated, demonstrating the applicability ofthe proposed approach for mAbs of different pI using the sameLC−MS setup. The utility of this method for the assignment ofantibody variants such as different glycosylation and C-terminallysine forms is exemplarily shown using a number of mAb drugproducts. The proposed CVA-MS approach represents a simpleyet powerful method for the characterization of complexbiopharmaceuticals on the intact level without the need forsample preparation.

■ EXPERIMENTAL SECTIONChemicals and Reagents. Deionized water was provided

by an Arium pro UV ultrapure water system from SartoriusStedim Biotech (Goettingen, Germany). Acetic acid (ACSreagent grade, ≥99.7%), ammonium bicarbonate (BioUltra,≥99.5%), ammonium hydroxide solution (BioUltra, 1 M inH2O), and DL-dithiothreitol (≥98%) were purchased fromSigma-Aldrich (Wicklow, Ireland). Water with 0.1% formic acid(Optima, LC−MS grade) and acetonitrile with 0.1% formicacid (Optima, LC−MS grade) were obtained from FisherScientific (Dublin, Ireland). All reagents and materials requiredfor KingFisher SMART digestion were obtained from ThermoFisher Scientific (Sunnyvale, CA). The monoclonal antibodiesused were kindly provided by the Hospital Pharmacy Unit ofthe University Hospital of San Cecilio in Granada, Spain.

Chromatographic Gradient Optimization. The buffersystem used for all charge variant separations consisted of 25mM ammonium bicarbonate and 30 mM acetic acid as buffer A(pH 5.3) and 10 mM ammonium hydroxide in 2 mM aceticacid as buffer B (pH 10.18). After mixing, buffers were allowedto rest for 24 h at room temperature. Optimization ofchromatographic separations was performed on an Ultimate3000 HPLC system equipped with a variable wavelengthdetector and an Ultimate 3000 PCM-3000 pH and conductivitymonitor (Thermo Fisher Scientific, Germering, Germany). Thecolumn used for all separations was a MAbPac SCX-10 RS 2.1mm × 50 mm column, 5 μm particles (Thermo FisherScientific, Sunnyvale, CA). A flow rate of 400 μL min−1 wasapplied, the column compartment was held at 25 °C, and dataacquisition was performed utilizing UV absorption at 280 nm.Injection amount for method optimization was 50 μg ofantibody per run. The optimized gradients for the separation ofall 5 mAbs are shown in Table 1.

CVA-MS Method Parameters. Chromatographic separa-tions for LC−MS experiments were performed on a ThermoScientific Vanquish Flex Quaternary UHPLC system (ThermoFisher Scientific, Germering, Germany). The chromatographic

Analytical Chemistry Article

DOI: 10.1021/acs.analchem.7b05241Anal. Chem. 2018, 90, 4669−4676

4670

http://dx.doi.org/10.1021/acs.analchem.7b05241

column, buffers, gradients, and all run parameters applied werethe same as described for gradient optimization. In total, 100 μgof mAb were injected per LC−MS run. For mass spectrometricdata acquisition, a Thermo Scientific Q Exactive Plusquadrupole-Orbitrap mass spectrometer, enabled with theBioPharma option allowing for mass detection up to m/z8 000, equipped with a heated electrospray ionization-II (HESI-II) probe in a standard Ion Max ion source (Thermo FisherScientific, Bremen, Germany) was used. MS tune and methodparameters can be obtained from Table S1 of the SupportingInformation.Collection of Bevacizumab Charge Variant Peaks.

Bevacizumab separation for peak collection was performed withidentical chromatographic run parameters as described forchromatographic gradient optimization using a ThermoScientific Vanquish Flex Quaternary UHPLC system.Sample Preparation for Peptide Mapping of Bev-

acizumab Charge Variant Peaks. In order to obtainsufficient amounts of protein for peptide mapping, peaks of 4CVA runs of 100 μg, respectively, were collected and merged.Samples were buffer exchanged to water using VIVASPIN spinfilters from Sartorius (Goettingen, Germany) with a cutoff sizeof 10 000 Da. Digestion was carried out via a Thermo ScientificKingFisher Duo Prime Purification System under the control ofBindit software (version 4.0) (Thermo Fisher Scientific, Vantaa,Finland). Samples were diluted with SMART Digest buffer in aKingFisher Deepwell 96 well plate. A volume of 15 μL ofmagnetic bead solution was mixed with 100 μL of SMARTDigest buffer. Magnetic beads were picked up with KingFisherDuo 12-tip combs and were washed for 1 min in SMARTDigest buffer diluted with water in a ratio of 1:4 (v/v).Digestion was performed at 70 °C for 40 min. Digested sampleswere reduced by addition of DTT to yield a 10 mMconcentration, incubated for 40 min at 37 °C and immediatelyanalyzed by LC−MS.

Peptide Mapping. Peptide mapping was performed on aThermo Scientific Vanquish Flex Binary UHPLC systemcoupled online to a Q Exactive Plus quadrupole-Orbitrapmass spectrometer. For chromatographic separation, anAcclaim VANQUISH C18 column 250 mm × 2.1 mm, 2.2μm particles (Thermo Fisher Scientific, Sunnyvale, CA) wasused. Separations were performed at 25 °C, and a flow rate of0.3 mL min−1 with 0.1% formic acid in water as buffer A and0.1% formic acid in acetonitrile as buffer B using a gradient of2−40% buffer B over 45 min was used. Details regarding MStune and method settings can be obtained from Table S2 in theSupporting Information.

Data Processing. Data acquisition and visualization ofLC−UV measurements was performed in Thermo ScientificChromeleon CDS 7.2. LC−MS measurements were performedunder Thermo Scientific Xcalibur 4.0. CVA-MS raw files wereexported to Thermo Scientific BioPharma Finder 2.0 softwareand analyzed using the ReSpect algorithm via either manualspectra averaging following deconvolution or the SlidingWindow deconvolution feature. The time windows for manualpeak averaging were set to comprise a maximal portion of thepeak area of the peaks of interest while excluding possiblypresent heterogeneities at the front and back ends. All data fileswere processed with target mass matching utilizing antibodysequences together with modifications such as disulfide bonds,glycosylation, and in the case of infliximab also lysinetruncation. All annotations based on the so generatedtheoretical masses can be observed in Tables S7−S10 of theSupporting Information. Details on the parameters used forBioPharma Finder analysis can be found in Table S3 of theSupporting Information. Peptide mapping data files were alsoanalyzed via the BioPharma Finder 2.0 software. For peptideidentification bevacizumab light and heavy chain sequenceswere used and the following variable modifications werespecified: lysine loss, deamidation of Gln and Asn, oxidation ofMet, glycation of Lys, and isomerization of Asp. Search resultswere filtered for a confidence score ≥0.8 and a recovery ≥1.Final results are based on average values of triplicate injections.

■ RESULTS AND DISCUSSIONThe principle of CVA using pH gradient elution relies on achange in buffer pH per unit time. As soon as the pH reaches avalue equal to a proteins apparent isoelectric point (pI), theprotein is no longer retained and elutes from the column. Thisfocusing effect concentrates the protein variants at the pH oftheir apparent pI. Ideally, buffers with considerable bufferingcapacity are applied to guarantee a linear pH change matchingthe chromatographic gradient programed. Premade buffercocktails, very well suiting this requirement, have beencommercialized and are available today.28,29 Online hyphen-ation of CVA to MS does, thereby, not only require the controlof the pH slope but also additionally introduces the need foroptimization of ion transfer for MS sensitivity. A delicatebalance needs to be obtained between low salt concentrationsfor MS sensitivity while maintaining control of the pH tofacilitate reproducible chromatography. As eluents comprised ofvolatile salts have much less buffering capacity, reduced controlof pH makes column choice more important, as the columnitself will buffer against pH changes occurring during gradientelution. Hence, low capacity columns, for example packed withpellicular or solid particles, exclusively allowing surfaceinteraction and thus exhibiting fast mass transfer of largemolecules for high-resolution separations, can be used to

Table 1. Optimized Gradients for the CVA of FiveMonoclonal Antibodies Using the MS Friendly BufferSystem

mAb time (min) % buffer B curve

trastuzumab0 400.5 55 510 100 5

adalimumab0 400.5 50 510 100 7

infliximab0 400.5 45 510 55 5

bevacizumab

0 400.5 45 56 50 710 100 5

cetuximab

0 400.5 45 57 65 710 100 5

Analytical Chemistry Article

DOI: 10.1021/acs.analchem.7b05241Anal. Chem. 2018, 90, 4669−4676

4671

http://pubs.acs.org/doi/suppl/10.1021/acs.analchem.7b05241/suppl_file/ac7b05241_si_001.pdfhttp://pubs.acs.org/doi/suppl/10.1021/acs.analchem.7b05241/suppl_file/ac7b05241_si_001.pdfhttp://pubs.acs.org/doi/suppl/10.1021/acs.analchem.7b05241/suppl_file/ac7b05241_si_001.pdfhttp://pubs.acs.org/doi/suppl/10.1021/acs.analchem.7b05241/suppl_file/ac7b05241_si_001.pdfhttp://dx.doi.org/10.1021/acs.analchem.7b05241

minimize such buffering effects. Our column of choice was aMAbPac SCX-10 RS column with 5 μm particle size and smalldimensions which very well suited these requirements. Theshort column length contributes to minimizing bufferingcapacity which allowed more precise control over the pHgradients applied as well as faster re-equilibration. In addition,column length has a minimal effect on chromatographicresolution using pH gradient elution due to the contributionof chromatofocusing to the separation mechanism.29 Mobilephase compositions with low concentrations of volatile salts(≤60 mM) were investigated, which enabled highly repeatablechromatography (RSD < 2% relative peak area and

pH region. In the case of adalimumab, utilization of a curvedgradient resulted in superior chromatographic selectivity foracidic variants whereas the later eluting basic species ofcomparatively low signal intensity are focused by the increasingsteepness of the gradient (Figure 2B). Infliximab andbevacizumab apparent pI values differ by 0.7 pH units. Bothantibodies, however, seem to elute at a similar buffercomposition (Figure 2C,D). This can likely be contributed tothe weak buffering capacity of the buffers especially across thepH range close to both antibody pI values.The compilation of the optimized separations for all five

mAbs tested, shown in Figure 2, demonstrates that this novelmethod is applicable to mAbs with various pI values using thesame column and buffer system. It also suggests ease ofgradient optimization for new candidates as similar gradientsare likely suitable to mAbs with a similar pI and a comparablecharge variant complexity. Furthermore, the resolution of thecharge variant peaks using the volatile buffers in this study inseveral cases compares favorably with previous reports usingsalt- and pH gradients.7,15,24,29,31 A comparison of a chromato-graphic separation of cetuximab using the volatile buffer systemto a separation that can be achieved using the CX-1 pH-gradient buffers can be obtained from Figure S1c,d.Hyphenation with high-resolution Orbitrap mass spectrom-

etry was then investigated for each of the developed gradientsusing the volatile pH buffer system. There are several criticalparameters for the acquisition of mass spectra of large proteinsunder native conditions that relate to source settings since theelectrospray process is performed with purely aqueous mobilephase: the probe heater and transfer capillary temperatures, in-source fragmentation (in-source CID) for declustering anddesolvation, as well as the S-Lens setting for transporting thelarge protein species through the front end of the ion optics. Inaddition, the resolution setting plays an important role as

signals from large proteins decay rather quickly in the Orbitrapmass analyzer.32 Thus, typically the lowest possible resolutionsetting is recommended for detection with highest sensitivitysince a low resolution setting translates directly into shortdetection times required to capture the protein signal with thehighest sensitivity. However, when chromatography isperformed in the presence of salts, protein species are proneto form adducts resulting in additional peaks in the spectra.With the lowest resolution setting, adducted species are notresolved resulting in broader and asymmetrical spectral peakscausing decreased mass accuracy following deconvolution.Figure 3 shows the benefit of the higher resolution settingfor the analysis of the major charge variant of adalimumab. Thesignal intensity in the chromatogram acquired at a resolutionsetting of 35 000 is decreased by a factor of 2 but with thebenefit of partially resolving the peaks with and without adductsresulting in separated, distinct peaks after deconvolution. Theability to resolve adducts is in these particular cases ofimportance, as it enables isoform annotation with muchimproved mass accuracies.Deconvolution and closer investigation of the adducted

forms revealed constant mass difference of around 40−42 Dafrom the main forms, indicating the occurrence of CO2adduction. That proteins are prone for adduction in thepresence of ammonium bicarbonate in solution has previouslybeen demonstrated.33 The fact that the mass differenceobserved does not fully correspond to the expected theoreticalmass of CO2, namely, 44 Da (Figure S3), can be explained bythe merely partial resolution of the adducted species,consequently leading to slight shifts toward the lower m/zregion. Another important feature of native mass spectrometryis the reduced number of charge states and the increase inspatial spectral resolution. This allows superior distinction

Figure 3. Adalimumab acquired at two different resolution settings, namely, 17 500 (blue) and 35 000 (magenta). The base peak chromatograms, themass spectra of the main peak, as well as magnifications of charge state 26 are shown.

Analytical Chemistry Article

DOI: 10.1021/acs.analchem.7b05241Anal. Chem. 2018, 90, 4669−4676

4673

http://pubs.acs.org/doi/suppl/10.1021/acs.analchem.7b05241/suppl_file/ac7b05241_si_001.pdfhttp://pubs.acs.org/doi/suppl/10.1021/acs.analchem.7b05241/suppl_file/ac7b05241_si_001.pdfhttp://dx.doi.org/10.1021/acs.analchem.7b05241

between isoforms of adjacent charged states which couldoverlap in a denatured charge envelope.An important feature of therapeutic proteins is glycosylation

as it has an impact on protein activity,34 stability,34,35 andlifetime.36,37 mAbs share a highly conserved site for N-glycosylation at Asn297 in their heavy chains;37 however, thenumber and relative abundance of different glycans can varyconsiderably from one to the other. The glycosylation patternserves a crucial quality criterion for therapeutic mAbs and ismonitored rigorously. During intact mass analysis of mAbs,annotation of glycoforms can be compromised due to thecoelution and presence of near isobaric proteoforms of themolecule. Under conditions of higher peak purity such as thoseattained herein, where charge variants elute separately, CVA-MS enables superior annotation of glycoforms for each chargevariant peak. Figure 4 shows the glycoform assignment of thefive most abundant glycoforms of the main peak oftrastuzumab, all with a mass deviation of less than 10 ppmfollowing spectral deconvolution.

Infliximab exhibits three major charge variants which havebeen identified to be lysine truncated forms, caused byincomplete carboxypeptidase (CpB) activity.38 Lysine variantsshow a mass difference of ∼128 Da which can be challenging toobserve using intact mass analysis without efficient separation.The isoform with an additional lysine present at the C-terminusmay overlap with the glycoform carrying one additionalgalactose which may compromise deconvolution and annota-tion of both species. The presence of a C-terminal lysine on amAb heavy chain increases the net number of positive chargesand consequently increases retention on the cation exchangecolumn which makes it well resolvable in CVA. The traditionalroute to identify the three major lysine peaks requires eitherCVA before and after CpB digestion or preparative IEC forfractionation followed by desalting using reversed phase LC−MS to obtain the intact mass data from each of the threefractions. It is demonstrated herein that the required data for aconfident identification of lysine variants can be obtained fromonly one analytical CVA-MS experiment as shown in Figure 5.

Figure 4. Glycoform assignment of the trastuzumab main peak; mass analysis was performed with a resolution setting of 35 000: (A) base peakchromatogram, (B) raw spectrum of the main peak, and (C) deconvoluted spectrum with glycoform annotations and mass deviations.

Figure 5. Infliximab glycoform and lysine variant assignment; mass analysis was performed with a resolution setting of 35 000: (A) base peakchromatogram of infliximab and schematic of the principle of Sliding Window deconvolution which was applied, (B) combined deconvolutedspectrum, and (C) glycoform assignment and relative quantification.

Analytical Chemistry Article

DOI: 10.1021/acs.analchem.7b05241Anal. Chem. 2018, 90, 4669−4676

4674

http://dx.doi.org/10.1021/acs.analchem.7b05241

In addition to isoform identification also quantitativeinformation can be gained. In this regard, the Sliding Windowdeconvolution option of Biopharma Finder 2.0 facilitates thecalculation of individual peak areas by tracking MS intensitiesover time. This information can in consequence be used toperform relative quantification of protein isoforms as it isshown in the diagram in Figure 5 for the three most abundantglycoforms of the three infliximab C-terminal lysine variants.Bevacizumab exhibits a simple glycosylation pattern and low

charge heterogeneity. The charge variant separation shows aminor portion of acidic variants, with the most abundantone eluting directly in front of the main peak. One verycommon modification leading to acidic variants is deamidationof asparagine or glutamine residues, which has been shown tonaturally occur during antibody downstream processing,formulation, and storage.1,39,40 The mass shift induced bydeamidation, which is only ∼1 Da, makes the modified andnonmodified variant undistinguishable by mass spectrometricmeans, at least if both are entering the MS system at the sametime. The fact, however, that deamidation results in thegeneration of one additional acidic residue on the protein,changes its apparent pI in solution, which in consequencereduces the retention on the cation exchange resin.3,39,40 Figure6 shows the CVA-MS analysis of bevacizumab and a

comparison of the masses of the three most abundantglycoforms of both the main peak and the most abundantacidic species. All three of them show an increase in averagemass resembling the mass shift caused by a single deamidationevent. The presence of a singly deamidated form is alsosuggested by means of retention time as deamidation haspreviously been shown to be a chromatographically distinguish-able acidic modification on mAbs3 as well as by a peakcollection and peptide mapping conducted in the course of thisstudy (Figure S4). These results suggest that this novel methodis not only capable of distinguishing different glycoforms andlysine variants but also low abundant, near isobaric species suchas deamidation products.

■ CONCLUSIONSA method for online hyphenation of charge variant analysis tohigh-resolution native Orbitrap mass spectrometry using lowionic strength eluents on a strong cation exchange phase hasbeen described. This new approach offers a simple and

universal strategy for mAb characterization on the intact levelwithout any need for sample preparation. Chromatographicresolution in the CVA analysis for several investigated mAbscompared favorably to previous approaches employing non-MSfriendly conditions and facilitated the separation of multiplecharge variant species even within a narrow pI range. MS datagenerated from even minor variants was sufficient in quality toachieve component identification with high mass accuracyallowing for the identification of critical quality attributesincluding accurate intact mass, lysine truncation, glycosylation,and deamidation from a single LC−MS injection. The methodenables the observation of the glycoform distribution for eachseparated charge variant allowing for potential interplaybetween modifications to be investigated. Conveniently, themethod bypasses the elaborate sample preparation required byalternative techniques to obtain the same level of information,which saves time and prevents the appearance of unintention-ally induced modifications or analytical bias. The developedCVA-MS method offers potential for improvement, forexample, by further enhancing sensitivity or by avoidingadduction, but represents a significant advance for all fields ofintact protein mass spectrometry including biopharmaceuticalcharacterization, proteomics, and structural biology.

■ ASSOCIATED CONTENT*S Supporting InformationThe Supporting Information is available free of charge on theACS Publications website at DOI: 10.1021/acs.anal-chem.7b05241.

Comparison of cetuximab charge variant profilesacquired with different gradients and buffers ; threeconsecutive runs of 50 μg of bevacizumab used formethod repeatability determination; adalimumab adduc-tion annotation; comparison of portion of deamidationof bevacizumab main and most dominant acidic peaks;MS tune and method parameters for CVA-MS runs andfor peptide mapping experiments; BioPharma Finder 2.0data analysis parameters for CVA-MS runs; pI values ofthe mAbs analyzed; calculation of RSDs for evaluation ofmethod repeatability and for relative peak area for 4peaks of 8 runs over 24 hours of 50 μg of bevacizumab;deconvolution result output table from BioPharmaFinder Software comparing results obtained from MSresolution setting of 17 500 and 35 000 of theadalimumab main peak; BioPharma Finder 2.0 outputsof the glycoform assignment of the trastuzumab mainpeak, lysine variant- and glycoform assignment ofinfliximab, and main- and acidic variant comparison ofbevacizumab (PDF)

■ AUTHOR INFORMATIONCorresponding Author*Phone: +353 1215 8100. Fax: +353 1215 8116.

ORCID

Stefan Mittermayr: 0000-0001-8950-0156Jonathan Bones: 0000-0002-8978-2592NotesThe authors declare no competing financial interest.

Figure 6. Comparison of the main- and an acidic species in thebevacizumab base peak chromatogram. Mass analysis was performedwith a resolution setting of 35 000. The mass differences for the threemost abundant glycans are indicating the presence of a singlydeamidated species and are shown in tabular form.

Analytical Chemistry Article

DOI: 10.1021/acs.analchem.7b05241Anal. Chem. 2018, 90, 4669−4676

4675

http://pubs.acs.org/doi/suppl/10.1021/acs.analchem.7b05241/suppl_file/ac7b05241_si_001.pdfhttp://pubs.acs.orghttp://pubs.acs.org/doi/abs/10.1021/acs.analchem.7b05241http://pubs.acs.org/doi/abs/10.1021/acs.analchem.7b05241http://pubs.acs.org/doi/suppl/10.1021/acs.analchem.7b05241/suppl_file/ac7b05241_si_001.pdfhttp://orcid.org/0000-0001-8950-0156http://orcid.org/0000-0002-8978-2592http://dx.doi.org/10.1021/acs.analchem.7b05241

■ ACKNOWLEDGMENTSThe authors gratefully acknowledge funding from ScienceFoundation Ireland under Grant Number 13/CDA/2196 andcollaborators in Thermo Fisher Scientific for instrument accessand support.

■ REFERENCES(1) Du, Y.; Walsh, A.; Ehrick, R.; Xu, W.; May, K.; Liu, H. MAbs2012, 4, 578−585.(2) Berkowitz, S. A.; Engen, J. R.; Mazzeo, J. R.; Jones, G. B. Nat. Rev.Drug Discovery 2012, 11, 527−540.(3) Harris, R. J.; Kabakoff, B.; Macchi, F. D.; Shen, F. J.; Kwong, M.;Andya, J. D.; Shire, S. J.; Bjork, N.; Totpal, K.; Chen, A. B. J.Chromatogr., Biomed. Appl. 2001, 752, 233−245.(4) Fekete, S.; Beck, A.; Veuthey, J. L.; Guillarme, D. J. Pharm.Biomed. Anal. 2015, 113, 43−55.(5) Fekete, S.; Beck, A.; Guillarme, D. J. Pharm. Biomed. Anal. 2015,111, 169−176.(6) Fekete, S.; Beck, A.; Fekete, J.; Guillarme, D. J. Pharm. Biomed.Anal. 2015, 102, 33−44.(7) Fekete, S.; Beck, A.; Fekete, J.; Guillarme, D. J. Pharm. Biomed.Anal. 2015, 102, 282−289.(8) Hintersteiner, B.; Lingg, N.; Janzek, E.; Mutschlechner, O.;Loibner, H.; Jungbauer, A. Biotechnol. J. 2016, 11, 1617−1627.(9) Hintersteiner, B.; Lingg, N.; Zhang, P.; Woen, S.; Hoi, K. M.;Stranner, S.; Wiederkum, S.; Mutschlechner, O.; Schuster, M.;Loibner, H.; Jungbauer, A. MAbs 2016, 8, 1548−1560.(10) Khawli, L. A.; Goswami, S.; Hutchinson, R.; Kwong, Z. W.;Yang, J.; Wang, X.; Yao, Z.; Sreedhara, A.; Cano, T.; Tesar, D.; Nijem,I.; Allison, D. E.; Wong, P. Y.; Kao, Y. H.; Quan, C.; Joshi, A.; Harris,R. J.; Motchnik, P. MAbs 2010, 2, 613−624.(11) Tang, L.; Sundaram, S.; Zhang, J.; Carlson, P.; Matathia, A.;Parekh, B.; Zhou, Q.; Hsieh, M. C. MAbs 2013, 5, 114−125.(12) Sandra, K.; Sandra, P. Bioanalysis 2015, 7, 2843−2847.(13) Lingg, N.; Berndtsson, M.; Hintersteiner, B.; Schuster, M.;Bardor, M.; Jungbauer, A. J. Chromatogr A 2014, 1373, 124−130.(14) Talebi, M.; Shellie, R. A.; Hilder, E. F.; Lacher, N. A.; Haddad,P. R. Anal. Chem. 2014, 86, 9794−9799.(15) Zhang, L.; Patapoff, T.; Farnan, D.; Zhang, B. J. Chromatogr A2013, 1272, 56−64.(16) Kang, X.; Kutzko, J. P.; Hayes, M. L.; Frey, D. D. Journal ofchromatography. A 2013, 1283, 89−97.(17) Schmidt, M.; Hafner, M.; Frech, C. J. Sep Sci. 2014, 37, 5−13.(18) Griaud, F.; Denefeld, B.; Lang, M.; Hensinger, H.; Haberl, P.;Berg, M. MAbs 2017, 9, 820−830.(19) Rose, R. J.; Damoc, E.; Denisov, E.; Makarov, A.; Heck, A. J.Nat. Methods 2012, 9, 1084−1086.(20) Rosati, S.; Rose, R. J.; Thompson, N. J.; van Duijn, E.; Damoc,E.; Denisov, E.; Makarov, A.; Heck, A. J. Angew. Chem., Int. Ed. 2012,51, 12992−12996.(21) Thompson, N. J.; Rosati, S.; Rose, R. J.; Heck, A. J. Chem.Commun. (Cambridge, U. K.) 2013, 49, 538−548.(22) Muneeruddin, K.; Bobst, C. E.; Frenkel, R.; Houde, D.; Turyan,I.; Sosic, Z.; Kaltashov, I. A. Analyst 2017, 142, 336−344.(23) Muneeruddin, K.; Nazzaro, M.; Kaltashov, I. A. Anal. Chem.2015, 87, 10138−10145.(24) Talebi, M.; Nordborg, A.; Gaspar, A.; Lacher, N. A.; Wang, Q.;He, X. Z.; Haddad, P. R.; Hilder, E. F. Journal of chromatography. A2013, 1317, 148−154.(25) Leblanc, Y.; Ramon, C.; Bihoreau, N.; Chevreux, G. J.Chromatogr. B: Anal. Technol. Biomed. Life Sci. 2017, 1048, 130−139.(26) Sorensen, M.; Harmes, D. C.; Stoll, D. R.; Staples, G. O.;Fekete, S.; Guillarme, D.; Beck, A. mAbs 2016, 8, 1224−1234.(27) Stoll, D. R.; Harmes, D. C.; Danforth, J.; Wagner, E.; Guillarme,D.; Fekete, S.; Beck, A. Anal. Chem. 2015, 87, 8307−8315.(28) Lin, S.; Baek, J.; Pohl, C. Thermo Fisher Scientific ApplicationNote 20946, Available online at https://tools.thermofisher.com/

content/sfs/brochures/AN-20946-HPLC-mAbPac-pH-Gradient-mAb-AN20946-EN.pdf (Accessed December 15, 2017).(29) Farnan, D.; Moreno, G. T. Anal. Chem. 2009, 81, 8846−8857.(30) Goyon, A.; Excoffier, M.; Janin-Bussat, M. C.; Bobaly, B.;Fekete, S.; Guillarme, D.; Beck, A. J. Chromatogr. B: Anal. Technol.Biomed. Life Sci. 2017, 1065−1066, 119−128.(31) Wagner-Rousset, E.; Fekete, S.; Morel-Chevillet, L.; Colas, O.;Corvaia, N.; Cianferani, S.; Guillarme, D.; Beck, A. Journal ofchromatography. A 2017, 1498, 147−154.(32) Makarov, A.; Denisov, E. J. Am. Soc. Mass Spectrom. 2009, 20,1486−1495.(33) Terrier, P.; Douglas, D. J. J. Am. Soc. Mass Spectrom. 2010, 21,1500−1505.(34) Jefferis, R. Curr. Pharm. Biotechnol. 2016, 17, 1333−1347.(35) Gavrilov, Y.; Shental-Bechor, D.; Greenblatt, H. M.; Levy, Y. J.Phys. Chem. Lett. 2015, 6, 3572−3577.(36) Debeljak, N.; Sytkowski, A. J. Drug Test. Anal. 2012, 4, 805−812.(37) Arnold, J. N.; Wormald, M. R.; Sim, R. B.; Rudd, P. M.; Dwek,R. A. Annu. Rev. Immunol. 2007, 25, 21−50.(38) Hong, J.; Lee, Y.; Lee, C.; Eo, S.; Kim, S.; Lee, N.; Park, J.; Park,S.; Seo, D.; Jeong, M.; Lee, Y.; Yeon, S.; Bou-Assaf, G.; Sosic, Z.;Zhang, W.; Jaquez, O. MAbs 2017, 9, 365−383.(39) Chelius, D.; Rehder, D. S.; Bondarenko, P. V. Anal. Chem. 2005,77, 6004−6011.(40) Timm, V.; Gruber, P.; Wasiliu, M.; Lindhofer, H.; Chelius, D. J.Chromatogr. B: Anal. Technol. Biomed. Life Sci. 2010, 878, 777−784.

Analytical Chemistry Article

DOI: 10.1021/acs.analchem.7b05241Anal. Chem. 2018, 90, 4669−4676

4676

https://tools.thermofisher.com/content/sfs/brochures/AN-20946-HPLC-mAbPac-pH-Gradient-mAb-AN20946-EN.pdfhttps://tools.thermofisher.com/content/sfs/brochures/AN-20946-HPLC-mAbPac-pH-Gradient-mAb-AN20946-EN.pdfhttps://tools.thermofisher.com/content/sfs/brochures/AN-20946-HPLC-mAbPac-pH-Gradient-mAb-AN20946-EN.pdfhttp://dx.doi.org/10.1021/acs.analchem.7b05241