Evaluation of Ion Mobility-Mass Spectrometry for ... · Evaluation of Ion Mobility-Mass Spectrometry for Comparative Analysis of Monoclonal Antibodies Carly N. Ferguson,1,2 Ashley

DOI: 10.1007/s13361-016-1369-1J. Am. Soc. Mass Spectrom. (2016) 27:822Y833

Evaluation of Ion Mobility-Mass Spectrometryfor Comparative Analysis of Monoclonal Antibodies

Carly N. Ferguson,1,2 Ashley C. Gucinski-Ruth1

1U.S. Food and Drug Administration, Center for Drug Evaluation and Research, Office of Pharmaceutical Quality, Office ofTesting and Research, Division of Pharmaceutical Analysis, 645 S. Newstead Ave., St. Louis, MO 63110, USA2Present Address: Pfizer Inc., Chesterfield, MO, USA

Abstract. Analytical techniques capable of detecting changes in structure are nec-essary tomonitor the quality of monoclonal antibody drug products. Ionmobility massspectrometry offers an advanced mode of characterization of protein higher orderstructure. In this work, we evaluated the reproducibility of ion mobility mass spec-trometry measurements andmobiligrams, as well as the suitability of this approach todifferentiate between and/or characterize different monoclonal antibody drug prod-ucts. Four mobiligram-derived metrics were identified to be reproducible across amulti-day window of analysis. These metrics were further applied to comparativestudies of monoclonal antibody drug products representing different IgG subclasses,manufacturers, and lots. These comparisons resulted in some differences, based on

the four metrics derived from ion mobility mass spectrometry mobiligrams. The use of collision-induced unfoldingresulted in more observed differences. Use of summed charge state datasets and the analysis of metrics beyonddrift time allowed for a more comprehensive comparative study between different monoclonal antibody drugproducts. Ion mobility mass spectrometry enabled detection of differences between monoclonal antibodies withthe same target protein but different production techniques, as well as products with different targets. Thesedifferences were not always detectable by traditional collision cross section studies. Ion mobility mass spectrom-etry, and the added separation capability of collision-induced unfolding, was highly reproducible and remains apromising technique for advanced analytical characterization of protein therapeutics.Keywords: Ion mobility-mass spectrometry, Monoclonal antibodies, Higher order structure

Received: 27 October 2015/Revised: 12 February 2016/Accepted: 19 February 2016/Published Online: 17 March 2016

Introduction

Over 40 monoclonal antibody (mAb) drug products haveobtained FDA approval since the 1980s, when the first

mAb was approved [1]. MAbs are used to treat a variety ofconditions, including cancer, autoimmune disorders, inflam-mation, and infection. These mAb products are complex innature as they are large (~150 kDa) molecules composed ofmultiple 25–30 kDa subunits. Subunits can be fused in avariety of combinations resulting in chimeric, human, human-ized, and murine antibody types [2]. With the periods of patent

exclusivity for several mAb products scheduled to end withinthe next few years, several follow-on biologics license appli-cations (BLAs) are expected [3]. Several follow-on productshave already been approved in Canadian and Europeanmarkets[4–8]. The introduction of follow-on biologics has large finan-cial implications as biologics account for 1% of total prescrip-tions, yet 28% of total prescription expenditures [9, 10].Additionally, BLAs will continue to be submitted for newmAb drug products, including antibody-drug conjugates(ADCs) and mAb fragment products.

In order to ensure the safety and efficacy of these biologicdrug products, analytical methods sensitive to the compositionand structure of these biopharmaceutical products are neces-sary. Primary, secondary, tertiary, and quaternary structure, aswell as other critical quality attributes (CQAs) may be affectedby a wide array of factors, including changes in manufacturingprocesses and improper storage practices [11–13]. Severalestablished techniques exist to characterize primary structure

Electronic supplementary material The online version of this article (doi:10.1007/s13361-016-1369-1) contains supplementary material, which is availableto authorized users.

Correspondence to: Ashley C. Gucinski–Ruth;e-mail: [email protected]

B American Society for Mass Spectrometry (outside the USA), 2016

RESEARCH ARTICLE

of protein therapeutics, such as peptide mapping using HPLCand top-down or bottom-up mass spectrometry (MS) [14]. Anumber of U.S. Pharmacopeia Convention (USP) monographsutilize HPLC for peptide mapping to confirm primary se-quence. Primary structure determination can identify modifica-tions such as glycosylation, oxidation, deamidation, and reduc-tion of disulfide bonds. Although primary structure is useful incharacterization of protein therapeutics, these modifications toprimary structure may not reflect changes in higher levels ofstructure accurately, and thus other techniques are necessaryfor complete protein characterization [15–18].

Secondary, tertiary, and quaternary structures are oftengrouped under the term higher order structure (HOS). X-raycrystallography, nuclear magnetic resonance (NMR), circulardichroism (CD), and a number of spectroscopic techniques areroutinely used for determination of secondary structure [19–21]. A recent study utilized deep ultraviolet resonance Raman(DUVRR) to sensitively characterize mAb secondary structure[22]. The study related changes in HOS to changes in drugefficacy.

Characterization of tertiary and quaternary HOS has provento be analytically challenging. X-ray crystallography and NMRcan be used; however, these techniques are limited by largesample amount requirements andmolecular weight restrictions.Native MS has been employed for some protein therapeuticstudies after significant advancements in instrumentation,namely high resolution mass spectrometers that allow for theiraccurate mass determination. These native MS studies oftenfocus on glycoform identification and subunit stoichiometry[23]. Two MS-based methods have emerged for protein HOScharacterization: hydrogen/deuterium exchange MS (HDX-MS) and ion mobility-MS (IM-MS) [24]. These MS-basedmethods are promising as they can be performed using rela-tively small amounts of sample, in contrast to traditional NMRand X-ray crystallographic approaches. HDX-MS enables de-tection of changes in conformation of a protein based on levelsof deuterium incorporation. Regions of proteins that are distinctor exhibit conformational changes can be identified throughdifferential experiments [25]. Because the technique incorpo-rates a digestion step and coupling with HPLC, sequenceinformation and post-translation modifications (PTMs) can beidentified along with conformational changes. Significant ad-vances have beenmade inHDX-MS, though challenges remainin obtaining reproducible results due to back-exchange, lowtemperatures, and decreased HPLC resolution. Automationadvancements in the form of robotics exist; however, thissystem is expensive and requires a large footprint in lab space.

Native IM-MS allows ions to be separated by mass as wellas conformation to potentially allow distinction between iso-baric compounds if sufficient differences in their structures(and corresponding mobilities) are present. In IM-MS, ionstravel through a drift tube and are separated based on theirmobilities on amillisecond timescale. There are twomain typesof ion mobility mass spectrometers: drift tube (DTIMS) andtraveling wave (TWIMS) [26, 27]. Both instruments enable thedetermination of a collision cross section (CCS), a

characteristic of the ion that reflects its size and shape. InDTIMS, ions travel through a neutral drift gas, where smallerions experience fewer collisions and travel faster down the drifttube than larger ions. DTIMS allows a direct calculation ofCCS based on the measured drift time (tD) [28]. TWIMSincorporates a traveling wave, which ions travel over for thelength of the mobility region. Larger ions travel over the wavemore frequently and thus take longer to reach the end of theflight tube. Smaller ions ‘surf’ the wave and reach the end ofthe flight tube faster [27]. Unlike DTIMS, tD measurementsfrom TWIMS experiments are not commonly used to directlycalculate CCS, though studies have shown it is possible [29].Instead, a complex series of calculations are used to convert tDto CCS [30, 31].

Larger proteins and drug products, such as mAbs, are diffi-cult to resolve based on standard IM-MS experiments becauseof high molecular weights and complex structures [32]. Earlyexperiments by Jarrold and Honea suggested that an annealingtechnique could aid in the separation of larger ions. In theirstudy of silicon clusters, heating ions via a collision energyapplied immediately prior to entrance into the drift tube result-ed in better separation based on differences in structure causedby the annealing [33]. Zhong et al. have employed this conceptin collision-induced unfolding (CIU) IM-MS experiments ofproteins, applying some collision energy in the trap region(TCE) of a standard TWIMS QTOF instrument. Similar tothe annealing experiments, different proteins may unfold inunique, specific patterns [32].

Several studies of large proteins, protein complexes, andprotein therapeutics have incorporated IM-MS or CIU IM-MSto determine CCS values [23, 30–32, 34–38]; however, thesuitability of this technique to compare and differentiate proteintherapeutics has not yet been evaluated. In this work, weevaluated the suitability of IM-MS and CIU IM-MS measure-ments for analysis of protein therapeutic systems. Severalmetrics were reproducible across technical replicates. Methodreproducibility was evaluated in a time-course experiment, inwhich several metrics were identified as reproducible in bothIM-MS and CIU IM-MS experiments within a giventimeframe. Multiple lots from individual drug products werealso analyzed to identify any lot-to-lot variation. Further IM-MS and CIU IM-MS comparisons were made between differ-ent IgG subclasses, mAbs from the same subclass, mAbs withdifferent targets, mAbs produced via different production tech-niques, and mAbs from different manufacturers.

ExperimentalMaterials

MS-grade acetonitrile, MS-grade water, 7 kDa desalting col-umns, and 10 kDa molecular weight cutoff (MWCO) filterswere purchased from Fisher Scientific (Pittsburgh, PA, USA).Alcohol dehydrogenase (ADH), ammonium acetate, bovineserum albumin (BSA), concanavalin A (ConA), andtransthyretin (TTR) were purchased from Sigma-Aldrich (St.

C. N. Ferguson and A. C. Gucinski-Ruth: IM-MS Evaluation for mAb Comparative Analysis 823

Louis, MO, USA). Adalimumab (Abbott, North Chicago, IL,USA), denosumab (Amgen, Thousand Oaks, CA, USA),infliximab (Johnson and Johnson, New Brunswick, NJ,USA), panitumumab (Amgen, Thousand Oaks, CA, USA),and rituximab (Genentech, South San Francisco, CA, USA)were purchased from the US marketplace. Rituximab (Dr.Reddy’s Laboratories, Ltd., Hyderabad, India) was purchasedfrom a non-U.S. marketplace. Sodium cesium iodide (NaCsI)was purchased from Waters (Manchester, UK). EconoTipsemitters were purchased from New Objective (Woburn, MA,USA).

IM-MS Sample Preparation

MAbs were buffer exchanged into water using 10 kDaMWCOfilters and concentrated to 100 μM. MAbs were further buffer-exchanged and diluted to a working concentration of 2 μM in150 mM ammonium acetate. Protein calibrant stocks for CCSwere prepared at the following concentrations: 26 μM ADH,15 μM BSA, 30 μM ConA, 13 μM PK, and 100 μM TTR inwater. CCS calibrants were further buffer-exchanged and di-luted to a working concentration of 5 μM in 150 mM ammo-nium acetate using 10 kDa MWCO filters.

IM-MS and CIU IM-MS Data Acquisition

Data were acquired on a Waters Synapt G2Si HDMS (Milford,MA, USA) equipped with an 8 kDa quadrupole. Settings wereoptimized as follows: source temperature 80 oC, capillaryvoltage 1.4 kV, sampling cone and source offset 100 V, trapbias 75V, trap DC –6.5 V, trap entrance 0 V, trap exit –1 V, IMDC entrance 20, transfer DC exit 15, transfer collision energy5 V, IMS wave height 40 V, transfer wave velocity 150 m/s,and transfer wave height 3 V. A wave ramp was used from 800to 50 m/s for IMS wave velocity. Trap collision energy (TCE)was applied at 0, 50, 75, or 100 V. Cone gas was set to 40 L/hand no purge gas was used. Nano flow gas was set to 0.3 barswhen needed. Data were acquired for 5 or 10 min.

IM-MS and CIU IM-MS Data Analysis

Chromatograms, extracted mobiligrams, andmass spectra wereanalyzed in MassLynx v4.1 (Waters). Mobility data were ana-lyzed in DriftScope v2.5 (Waters).

Collision Cross Section Data Acquisition

CCS data were acquired on a Waters Synapt G2Si HDMS(Milford, MA, USA) equipped with an 8 K quad. Settings wereoptimized for native-like ions according to Ruotolo et al., andparameters were set as follows: source temperature 80 oC,capillary voltage 1.0 kV, sampling cone and source offset20 V, trap bias 45 V, trap DC –6.5 V, trap entrance 0 V, trapexit –1 V, IM DC entrance 10, transfer DC exit 9, trap andtransfer collision energy 0 V, IMS wave velocity 250 m/s, IMSwave height 30 V, and transfer wave velocity 200 m/s [31].Cone gas was set to 40 L/h and no purge gas was used. Nano

flow gas was set to 0.3 bar when needed. Data were acquiredfor 5 min at transfer wave heights of 8, 9, and 10 V.

Collision Cross Section Data Analysis

Calibrants used were alcohol dehydrogenase (ADH), bovineserum albumin (BSA), concanavalin A (ConA), andtransthyretin (TTR). CCS was calculated using methods de-scribed by Ruotolo et al. and Bush et al. [30, 31]. Briefly,mobiligrams were extracted for the top 3 or 4 most abundantcharge states for each calibrant at each transfer wave height.Plots of Ω’ (CCS in nm2) versus tD’ (ms) were fitted with apower regression model in Microsoft Excel. From mAb acqui-sitions, tD’ values were used to calculate Ω’ and finally Ω.

Results and DiscussionMetric Selection and Evaluation of Metric Repro-ducibility Across Technical Replicates

Five metrics were initially evaluated across three technicalreplicates of the same rituximab (U.S.) sample for reproduc-ibility and suitability as potential comparative values. Thefollowing mobiligram components were monitored to deter-mine their suitability as a potential metric: drift time of the mostintense peak (A) in the mobiligram (td-A), drift time of thesecond most intense peak (B) in the mobiligram (td-B), peakwidth at 50%maximum intensity (FWHM), peak width at 10%maximum intensity (FWTM), and the intensity ratio betweentd-A and td-B (peak intensity ratio, PIR). A mobiligram is shownin Figure 1 with each of the five metrics labeled.

PIR was observed to be inconsistent and dependent uponthe width of the m/z selection window used in DriftScope (datanot shown). Reproducible values between technical replicatesof a single rituximab (U.S.) preparation were observed for thefollowing four metrics: td-A, td-B, FWHM, and FWTM(Table 1). The values of these metrics were also reproducibleacross multiple (three) preparations of a rituximab (U.S.) sam-ple (Supplementary Table S1).

As multiple charge states were observed for each mAb drugproduct analyzed, mobiligrams were constructed in a variety ofways in order to probe any changes between charge state andstructure. Each of the top four charge states were independentlyextracted and mobiligrams analyzed; additionally, summedcharge state mobiligrams were extracted and analyzed fromthe sums of the top two, three, or four most intense chargestates. Metrics were reproducible in mobiligrams across all fourcharge states and the three summed charge state datasets.Average, standard deviation, and coefficient of variation(%RSD) values are shown in Table 1 for each of the top fourmost intense charge states individually as well as the summedtop two, three, or four most intense charge states. Standarddeviations of 0.00 reflected variation lower than reported pre-cision. In IM-MS experiments (no applied TCE), the td-Ameasurements (%RSD ≤ 1.44%, standard deviation ≤0.08 ms) were observed to be the most reproducible metrics,

824 C. N. Ferguson and A. C. Gucinski-Ruth: IM-MS Evaluation for mAb Comparative Analysis

although the variability observed in the FWHM and FWTMmeasurements was also acceptable (FWHM: %RSD ≤ 7.02%,standard deviation ≤ 0.04 ms; FWTM: %RSD ≤ 9.02%, stan-dard deviation ≤ 0.07 ms). In CIU IM-MS experiments (50, 75or 100 V TCE), the td-A and td-B measurements were the mostreproducible (td-A: %RSD ≤ 2.26%, standard deviation ≤0.16 ms; td-B: %RSD ≤ 1.97%, standard deviation ≤0.16 ms); similar to IM-MS experiments without TCE applied,FWHM and FWTMwere also reproducible (FWHM:%RSD ≤6.66%, standard deviation ≤ 0.11ms; FWTM:%RSD ≤ 4.99%,standard deviation ≤ 0.25 ms). Mobiligrams extracted from thesums of the top two or four most intense charge states yieldedthe most reproducible metric values in all IM-MS and CIU IM-MS experiments.

Samples were also subjected to analysis using TCE valuesof 25, 125, 150, 175, and 200 V. CIU IM-MS experiments at25 V TCE were the most variable (data not shown). Although25 V TCE may be sufficient to induce some unfolding, thisprocess may occur via a variety of pathways, translating to highvariability in mobiligram profiles. If differences in CIU IM-MSenergy profiles are not observed at 0, 50, 75, or 100 V TCE, asecond round of experiments utilizing TCE settings of 125–200 V may be used as an additional potentially discriminatorymeasurement as these instrument settings also yielded repro-ducible results (data not shown).

While data extracted from individual charge states werehighly reproducible, the order of the most intense charge states

was not when TCEwas applied. Figure 2 shows a series of CIUmobiligrams of the 24+ charge state of a rituximab (U.S.)sample at 0, 50, 75, and 100 V TCE with corresponding massspectra of each acquisition. When TCE was applied, the chargestate distributions (CSDs) were inconsistent. The top two andtop four most intense charge states remained the same, sug-gesting the sums of these charge states were better suited forcomparisons. This observation was consistent with abovediscussed results that indicated these two summed charge statedatasets were the most reproducible across metrics in both IM-MS and CIU IM-MS experiments.

Metric Reproducibility of Technical Replicates Between Prep-arations of Same Lot of Product Triplicate preparations ofthe same lot of rituximab (U.S.) were prepared independentlyand analyzed on d 1 to determine the amount of variation in thefour metrics from IM-MS and CIU IM-MS measurementsbetween different preparations of the same sample.Mobiligrams were extracted for the sums of the top two ortop four most intense charge states, and data are shown inSupplementary Table S1. FWTM and td-A measurements werethe most reproducible (td-A: %RSD ≤ 1.44%, standard devia-tion ≤ 0.08 ms; FWTM: %RSD ≤ 3.38%, standard deviation ≤0.19 ms), while FWHM measurements were the least repro-ducible (%RSD ≤ 11.35%, standard deviation ≤ 0.15 ms).Although FWHM was less reproducible than td-A and

Fig. 1. Metrics evaluated based on suitability to describe IM-MS mobiligrams

C. N. Ferguson and A. C. Gucinski-Ruth: IM-MS Evaluation for mAb Comparative Analysis 825

FWTM, FWHM measurements were still acceptable based onstandard deviation and %RSD (<15%) values [39]. Secondarypeaks were not observed in d 1 experiments; thus no td-Bmeasurements were possible.

Time-Course Metric Monitoring

The same dilutions of rituximab (2 μM in 150 mM ammoniumacetate prepared on d 1) were also analyzed on days 2, 3, 4, and 5post-preparation. By d 5, the CSD shifted to higher charge states(Supplementary Figure S1A), suggesting a distinguishablechange in structure, indicating samples must be analyzed befored 5 in order to obtain meaningful measurements. Metrics frommobiligrams extracted from d 5 samples were not evaluated forreproducibility because of the apparent structural change.

The td-A, FWHM, and FWTM values obtained displayedhigh reproducibility, on average, across d, 2, 3, and 4 fordilutions initially made on d, 1. Metrics td-A, FWHM, andFWTM were averaged over d 1 through 4. IM-MS and CIUIM-MSmeasurements of td-A, FWHM, and FWTM from d 1–4were all within one standard deviation of the 4-d average,except one FWHM measurement on d 4 (SupplementaryTable S2, red). These measurements were made frommobiligrams extracted from the sums of the top two or top fourmost intense charge states. When individual charge states wereanalyzed across d 1 – 4, some td-A, td-B, FWHM, and FWTM

measurements fell outside of one standard deviation of the 4-daverage (Supplementary Table S3, red).

Additionally, fresh dilutions from stock (100 μM in waterprepared on d 1) were made and analyzed on d 2, 3, 4, and 5. Ashift in CSD by d 5 was observed (Supplementary Figure S1B).Reproducibility of analyzed metrics remained high across d 2, 3,and 4. Metrics were averaged over d 1 through 4. On day 2, IM-MS measurements of td-A from the sum of the top two mostintense charge states fell outside one standard deviation of the 4-daverage by 0.04 ms (Supplementary Table S4, red). This resultwas likely due to a small decrease in signal-to-noise observed ond 2 (Supplementary Figure S1B) and not indicative of a signifi-cant change in conformation. A Savitsky-Golay smoothing algo-rithm eliminated this difference, suggesting smoothing wouldlead to evenmore reproducible results. This analysis of smoothedversus unsmoothed data was outside the scope of this manuscript;thus, all data analyzed were in the raw format. All other IM-MSand CIU IM-MS measurements across the 4 d fell within onestandard deviation of the 4-d average. When individual chargestates were analyzed across d 1 through 4, some td-B and FWTMmeasurements fell outside of one standard deviation of the 4-daverage (Supplementary Table S5, red). These data suggest thatIM-MS and CIU IM-MS comparisons should be made betweensamples run within 4 d of initial stock preparation when compar-ing themobiligrams extracted from the sums of the top two or topfour most intense charge states. If individual charge states are tobe compared, further considerations are necessary.

Table 1. Average ± Standard Deviation of Metrics Extracted from Mobiligrams of the Top Four Individual Charge States (C.S.) and Summed Top Two, Three, orFour Charge States for Technical Replicates of One Rituximab Sample. Trap Collision Energy (TCE) Values are Denoted in Column Two

C.S. (+) TCE (V) Technical replicates of a rituximab Sample %RSD

td-A (ms) td-B (ms) FWHM (ms) FWTM (ms) td-A td-B FWHM FWTM

23 0 5.94 ± 0.00 – 0.39 ± 0.03 0.75 ± 0.04 0.00 – 7.02 5.4523 50 7.33 ± 0.00 6.31 ± 0.08 2.35 ± 0.04 5.27 ± 0.04 0.00 1.28 1.60 0.7123 75 7.60 ± 0.00 – 1.16 ± 0.00 5.05 ± 0.25 0.00 – 0.33 4.9923 100 7.69 ± 0.08 8.11 ± 0.08 1.48 ± 0.02 5.07 ± 0.14 1.05 1.00 1.49 2.7524 0 5.67 ± 0.00 – 0.39 ± 0.02 0.72 ± 0.05 0.00 – 6.03 6.9524 50 7.24 ± 0.08 – 1.47 ± 0.06 5.54 ± 0.02 1.12 – 3.77 0.3724 75 7.88 ± 0.00 – 1.60 ± 0.03 5.25 ± 0.12 0.00 – 1.59 2.2924 100 7.88 ± 0.00 – 1.12 ± 0.01 4.96 ± 0.04 0.00 – 0.60 0.8925 0 5.39 ± 0.00 – 0.39 ± 0.03 0.76 ± 0.02 0.00 – 7.60 3.2025 50 7.19 ± 0.00 – 1.72 ± 0.08 5.76 ± 0.09 0.00 – 4.88 1.6225 75 7.60 ± 0.00 8.21 ± 0.16 1.43 ± 0.02 5.39 ± 0.19 0.00 1.97 1.43 3.4725 100 7.74 ± 0.00 8.16 ± 0.00 1.18 ± 0.04 4.89 ± 0.04 0.00 0.00 3.49 0.8026 0 5.25 ± 0.00 – 0.41 ± 0.03 0.81 ± 0.07 0.00 – 6.60 9.0226 50 7.14 ± 0.16 7.33 ± 0.00 1.90 ± 0.09 5.65 ± 0.17 2.26 0.00 4.58 2.9326 75 7.47 ± 0.00 8.02 ± 0.00 1.46 ± 0.05 5.63 ± 0.03 0.00 0.00 3.49 0.5926 100 7.51 ± 0.08 8.02 ± 0.00 1.31 ± 0.10 5.21 ± 0.12 1.00 0.00 7.87 2.23Top 2 0 5.62 ± 0.08 – 0.50 ± 0.02 0.91 ± 0.03 1.44 – 3.79 2.87Top 2 50 7.19 ± 0.00 – 1.63 ± 0.10 5.63 ± 0.03 0.00 – 6.07 0.57Top 2 75 7.65 ± 0.08 – 1.55 ± 0.00 5.30 ± 0.06 1.06 – 0.00 1.06Top 2 100 7.88 ± 0.00 – 1.35 ± 0.02 5.18 ± 0.10 0.00 – 1.39 2.00Top 3 0 5.62 ± 0.08 – 0.62 ± 0.04 1.08 ± 0.04 1.44 – 6.94 3.44Top 3 50 7.24 ± 0.08 – 1.68 ± 0.11 5.58 ± 0.05 1.12 – 6.66 0.88Top 3 75 7.60 ± 0.00 – 1.48 ± 0.02 5.29 ± 0.05 0.00 – 1.27 0.93Top 3 100 7.88 ± 0.00 – 1.39 ± 0.00 5.25 ± 0.02 0.00 – 0.00 0.36Top 4 0 5.62 ± 0.08 – 0.70 ± 0.00 1.19 ± 0.01 1.44 – 0.58 1.23Top 4 50 7.24 ± 0.08 – 1.75 ± 0.09 5.57 ± 0.02 1.12 – 4.90 0.33Top 4 75 7.60 ± 0.00 – 1.47 ± 0.02 5.27 ± 0.03 0.00 – 1.28 0.61Top 4 100 7.88 ± 0.00 – 1.41 ± 0.02 5.30 ± 0.06 0.00 – 1.32 1.22

826 C. N. Ferguson and A. C. Gucinski-Ruth: IM-MS Evaluation for mAb Comparative Analysis

Comparison of Multiple Lots of a Single DrugProduct

Three lots of rituximab (U.S.) were prepared independently andanalyzed by IM-MS and CIU IM-MS. Mobiligrams extractedfrom the sums of the top two or top four most intense chargestates were analyzed and compared. Metrics td-A and FWTMwere consistent across all three lots at each TCE value (0, 50,75, and 100 V), as shown in Figure 3a and c. FWHM was notconsistent across all three lots when 50 or 75 V TCE wereapplied, as shown in Figure 3b. Intact mass analysis by high-resolution MS did not indicate any differences between thethree lots (data not shown). Although td-A and FWTM wereconsistent across multiple lots of rituximab (U.S.), other com-parison experiments in this study used averages of multiple lotswhen multiple lots were available to account for the differencesobserved in FWHM. In comparative studies where FWHMwillnot be utilized, multiple lots may not be necessary.

Comparison of Two mAb Subclasses: IgG1and IgG2

Several IgG1 products and two IgG2 products were comparedby using IM-MS and CIU IM-MS (Supplementary Table S4).Metrics td-A, td-B, FWHM, and FWTM measured frommobiligrams extracted from the sum of the top two mostintense charge states were averaged across multiple lots, whenavailable (Supplementary Table S6); within each lot, threetechnical replicates were averaged. Metrics were first analyzedfor values common to only IgG1 or only IgG2 compounds. Nodistinct groupings were observed in IM-MS studies; however,CIU IM-MS experiments at 75 and 100 V TCE resulted inseparation of IgG1 and IgG2 compounds when td-A measure-ments were compared (Figure 4a and b). Drift times of IgG2

products were lower than those of IgG1 products. IgG2s con-tain more disulfide bonds and are more thermally stable. Thesedata suggest IgG2 products were less unfolded at 75 and 100 VTCE, which may correspond to a more stable higher orderstructure or reflect the higher number of disulfide linkages forIgG2s relative to IgG1 molecules. When comparisons weremade individually between specific IgG1 and IgG2 products,several significant differences were observed in metrics atsome TCE voltages. Table 2 shows a panitumumab (IgG2)versus rituximab (U.S., IgG1) comparison. Red-filled cellsindicate TCE conditions that resulted in significant differencesbetween the specified metrics. These products were significant-ly different across all four metrics in both IM-MS and CIU IM-MS experiments. Other IgG1 and IgG2 comparisons are shownin Supplementary Table S7. These data enabled direct compar-isons between multiple IgG1 and IgG2 products by IM-MS andCIU IM-MS. Additionally, these data indicated the utility ofCIU IM-MS experiments as an additional mode of separationwhen traditional IM-MS experiments were not able to distin-guish between the two subclasses.

Comparison of Two IgG1mAbs with Different Tar-gets: Rituximab (CD20) and Infliximab (TNF-α)

Within each IgG subclass, two additional factors were consid-ered: target and production technique. Rituximab andinfliximab are both produced as chimeric IgG1 mAbs; howev-er, rituximab targets CD20 protein and infliximab targets tumornecrosis factor alpha (TNF-α). Three lots of rituximab (U.S.)and two lots of infliximab were averaged and metrics wereanalyzed from mobiligrams extracted from the sum of the toptwo most intense charge states. A difference in td-A measure-ments was observed in IM-MS experiments, whereas a

Figure 2. CIU IM-MS diagram of extracted 24+ mobiligrams. Mass spectra are shown to the right with top one through four chargestates marked in order of intensity: first (triangle), second (circle), third (diamond), and fourth (square) most intense charge states

C. N. Ferguson and A. C. Gucinski-Ruth: IM-MS Evaluation for mAb Comparative Analysis 827

Figure 3. Measurements of (a) td-A, (b) FWHM, and (c) FWTM frommobiligrams extracted from the sumof the top twomost intensecharge states for three lots of rituximab (U.S.): lot 1 (blue), lot 2 (red) and lot 3 (green). All TCE values (0, 50, 75 and 100 V) are shown.Error bars indicate average + standard deviation (Supplementary Table S6)

828 C. N. Ferguson and A. C. Gucinski-Ruth: IM-MS Evaluation for mAb Comparative Analysis

difference in td-B measurements was observed in CIU IM-MSexperiments at 50 V TCE (Table 2, red). Intact mass analysisindicated a mass difference of approximately 1000 Da betweenthe two products (data not shown). Further differences wereobserved in FWHM and FWTMmeasurements at various TCEvalues. Figure 5a shows overlaid IM-MS mobiligrams frominfliximab and rituximab (U.S.), with the inset highlighting thedifference between td-A. Although the difference in td-A at 0 VTCEwas significantly different, it was less obvious upon visualinspection. When 50 V of TCE were applied, the visual differ-ence was more apparent, as shown in Figure 5b. Infliximabwasobserved as a single drift peak, whereas rituximab (U.S.)

mobiligrams contained a second peak. These data further indi-cate the complementary utility of IM-MS and CIU IM-MS todistinguish between two mAbs from the same subclass.

Comparison of Two IgG1 mAbs with Same Targetand Different Production Techniques: Infliximab(Chimeric) and Adalimumab (Human)

Both adalimumab and infliximab target TNF-α; however,adalimumab is a human mAb while infliximab is a chimericmAb. One lot of adalimumab (due to lack of availability ofadditional lots) and two lots of infliximab were averaged, and

7.25

7.30

7.35

7.40

7.45

7.50

Dri

ft T

ime

(ms)

75 V TCE td-A

IgG2IgG1

IgG2IgG17.25

7.30

7.35

7.40

7.45

7.50

7.55

7.60

7.65

7.70

7.75

7.80

Dri

ft T

ime

(ms)

100 V TCE td-A

(a)

(b)

Figure 4. Separation of IgG1 and IgG2 products are shown at (a) 75 V or (b) 100 V TCE. Metrics were measured from mobiligramsextracted from the sum of the two most intense charge states for the following mAb drug products: adalimumab (red), denosumab(orange), infliximab (green), panitumumab (teal), rituximab (Non-U.S. - purple), and rituximab (U.S. - blue). Plots indicate average ±standard deviation for a given number of lots (Supplementary Table S6)

C. N. Ferguson and A. C. Gucinski-Ruth: IM-MS Evaluation for mAb Comparative Analysis 829

Table 2. Four Key mAb Comparisons for All Four Metrics at Every TCE Value. A Filled Red Cell Indicates Significant Differences Between the CorrespondingMetric from the Two mAbs Being Compared at that Particular TCE Setting

Figure 5. Mobiligrams of infliximab (blue) and rituximab (U.S., black). (a) 0 V TCE. Inset shows a zoomed region, with significantlydifferent td-A measurements listed as average ± standard deviation. (b) 50 V TCE

830 C. N. Ferguson and A. C. Gucinski-Ruth: IM-MS Evaluation for mAb Comparative Analysis

metrics were analyzed from mobiligrams extracted from thesum of the top two most intense charge states. Differences wereobserved in both td-A and td-B measurements from CIU IM-MSexperiments at 50 V (Table 2 and Figure 6). Intact massanalysis indicated a mass difference of approximately 500 Dabetween the two products (data not shown). Comparisonsbetween FWHM or FWTM measurements yielded differencesin IM-MS experiments (both) and CIU IM-MS experiments at75 V (FWTM), shown in Table 2. The addition of CIU IM-MSexperiments enabled distinction of these two similar mAbproducts, which were previously indistinguishable by IM-MSbased on drift peak comparisons.

Comparison of an IgG1 Product (rituximab)from Different Manufacturers: U.S. and Non-U.S.

In analysis of follow-on BLAs, compared products will havethe same target and production technique. Both U.S. and Non-U.S. rituximab products are chimeric mAbs that targeted CD20protein. Only a difference in td-A was observed in CIU IM-MSexperiments at 50 V TCE. No differences in td-B, FWHM, orFWTM were observed in either IM-MS or CIU IM-MS exper-iments. Supplementary Figure S3 shows overlaid IM-MSmobiligrams from U.S. and Non-U.S. rituximab products, withthe inset highlighting the difference between td-A measure-ments. Though td-A measurements were significantly differentstatistically between the two species at a 50 V TCE, applicationof a Savitzky-Golay smoothing algorithm (window 3, order 2)resulted in the td-A values overlapping perfectly. Even withadditional smoothing (up to window 3, polynomial order 5),differences in the mobility profile between the two structuresremain. Sufficient data is not available to determine if thesedifferences are real or artifacts of the limited number of binsand bin sizes of IM-MS experiments on the Synapt platform

that incorporate m/z values inclusive of intact monoclonalantibodies. The results shown here are not strong enough tobe able to conclusively distinguish between the two rituximabdrug products from two manufacturers. This may be attribut-able to a limitation of the platform (bin size and/or resolvingpower) or it may provide some evidence of the consistency ofthe product between manufacturers. What is most clear fromthis data is that while IM-MSmay be used to clearly distinguishbetween mAb drug products with larger, more global differ-ences, orthogonal methods are necessary to determine thesimilarity of difference between these two rituximab products.Although intact mass analysis did not indicate any differencesbetween the two products (data not shown), a more detailedmass spectrometric approach (e.g., middle down or bottom up)may provide additional insights in the compositional differ-ences, if any, between the two products. While there is no wayto circumvent the bin size limitations on the Synapt G2siplatform, future studies will attempt to explore alternativeIM-MS platforms that may provide smaller bin sizes and yieldmore data points per peak in the mobiligram.

Collision Cross Sections

Typical protein comparisons by IM-MS use td-A measurementsto calculate a collision cross section (CCS, Ω). In TWIMS,these calculations are based on calibration curves constructedfrom protein standards of known CCS. Protein standards areselected to bracket the molecular weight of the sample ofinterest; however, options are limited to those available inonline databases [30, 40–42]. Additionally, calibration curvesare obtained under mild source conditions and no collisionenergy to preserve native state conformations. These mildconditions indicate that CCS calculations from CIU IM-MSexperiments extrapolated as CIU are not employed during

Figure 6. CIU IM-MS mobiligrams of adalimumab (blue) and infliximab (black) at 50 V TCE

C. N. Ferguson and A. C. Gucinski-Ruth: IM-MS Evaluation for mAb Comparative Analysis 831

calibration curve construction. Finally, because calculation ofCCS requires a td-A and an m/z, only individual charge statescan be compared.

Previously discussed comparisons were repeated underCCS-compatible source conditions. The top three most intensecharge states were compared. No grouping of IgG1 or IgG2products was observed based on CCS (SupplementaryFigure S2A–C). No significant differences in CCS were ob-served in adalimumab-infliximab comparisons or U.S.-Non-U.S. rituximab comparisons. When infliximab and rituximabwere compared, CCS measurements of all three charge stateswere significantly different.When panitumumab and rituximabwere compared, CCS measurements of the 23+ charge stateswere significantly different; however, CCS measurements ofthe remaining two charge states were not significantly different.Previously discussed IM-MS and CIU IM-MS experimentsresulted in td-A differences of 1% to 4%, which correspondedto differences in CCS of 0.5 to 1.8 nm2. Standard deviations ofCCS measurements ranged from 0.00 to 0.61 nm2, thus 2% to4% td-A differences were well outside this range whereas 1%differences were not. Direct comparison of mAbs utilizing thefour metrics appeared to be more sensitive in distinguishingbetween previously discussed products.

ConclusionsAs the number of protein therapeutics entering the marketcontinues to grow, highly sensitive analytical techniques willbe imperative to ensure product quality. IM-MS was found tobe an adequate technique for the characterization of a specificclass of protein therapeutics, monoclonal antibodies. The IM-MS measurements displayed low variability when technicalreplicates or multiple preparations of a mAb drug product wereanalyzed. Reproducibility was evaluated based on four metricsselected to describe the mobiligram of a mAb product: td-A, td-B, FWHM, and FWTM. An additional mode of separationcould be obtained when necessary by employing CIU IM-MSexperiments, which was found to provide similar reproducibil-ity of all four metrics across technical replicates and multiplepreparations of a mAb drug product. This reproducibility con-tinued across a 4-d window after initial sample preparation,widening experimental timeframes within which meaningfulcomparisons can be made.

Although individual charge states are traditionally used inIM-MS studies, mobiligrams comprised of the sums of the twoor four most intense charge states were also analyzed as a wayto obtain a more global measurement profile of the mAb drugproduct. These summed charge state datasets exhibited higherreproducibility and incorporated multiple charge states for amore comprehensive analysis. The utilization of summedcharge state datasets may enable a standardized analysis versuschoosing one or more individual charge states to make com-parisons. Multiple lots of one product were reproducible acrossall metrics when the sum of the top two charge states was used.Specific comparisons between IgG1 and IgG2 products

resulted in significant differences across all metrics. The con-sistency of these results was not observed across all individualcharge states, further indicating the utility of summed chargestate datasets. As a mAb is present in multiple charge states andstructures simultaneously, the use of multiple charge states togenerate mobiligrams will enable comparisons to more accu-rately reflect the entire protein population instead of only thestructure for a single charge state.

Comparison of IgG1 products with the same target butdifferent production techniques highlighted the use of CIUIM-MS, as TCE resulted in significantly differentmobiligram profiles for adalimumab and infliximab.Comparative studies between IgG1 products with differenttargets yielded differences in all metrics, whereas a differ-ence in td-A was observed in comparative studies betweentwo rituximab products from different manufacturers. Thisdifference in td-A observed between the two products (U.S.and Non-U.S.) was statistically significant, but small (1%).Upon smoothing of the data, the td-A became identical,although the mobility profiles continued to display someslight differences, even with additional smoothing.Although it has been demonstrated that IM-MS can be usedto make comparisons between many different systems, thisresult highlights the need for orthogonal methods to also beused when making comparisons between mAb drug prod-ucts. Additionally, the differences obtained without smooth-ing may be indicative of the limited digital resolution of theplatform instead of real chemical differences. Future studiesmay employ the use of stressors to further distinguish be-tween products from different manufacturers. Additionally,future studies will endeavor to use alternative IM-MS plat-forms that may offer increased digital resolution and, thus,more points per peak in the mobiligram. In this work, wedemonstrated the suitability of IM-MS and CIU IM-MS tocompare protein therapeutics, specifically mAb drug prod-ucts. This reproducible technique may be used to furthercharacterize structural changes in a variety of protein ther-apeutics, including newer molecules such as antibody-drugconjugates and fusion proteins. We suggest this proposedframework for characterization and comparative profiling ofmAbs using IM-MS and CIU IM-MS.

AcknowledgmentsThis project was supported in part by an appointment to theResearch Participation Program at the Center for Drug Evalu-ation and Research administered by the Oak Ridge Institute forScience and Engineering and the U.S. Food and Drug Admin-istration for C.N.F. This work was funded by the FDA Centerfor Drug Evaluation and Research Critical Path Programawarded to A.C.G. The authors thank Michaella Levy fromUS FDA–DPA for intact mass analysis, and Abby Gelb fromthe University of Nebraska–Lincoln for useful discussions andsuggestions on CCS calculations.

832 C. N. Ferguson and A. C. Gucinski-Ruth: IM-MS Evaluation for mAb Comparative Analysis

This publication reflects the views of the authors and shouldnot be construed to represent FDA’s views or policies.

References1. Aggarwal, R.S.: What’s fueling the biotech engine—2012 to 2013. Nat.

Biotechnol. 32(1), 32–39 (2014)2. An, Z.: Therapeutic monoclonal antibodies: from bench to clinic. John

Wiley and Sons, Inc., Hoboken (2009)3. Gaffney, A.: FDA receives first-ever biosimilar application for monoclonal

antibody. Available at: http://www.raps.org/Regulatory-Focus/News/2014/08/11/20001/FDA-Receives-First-Ever-Biosimilar-Application-for-Monoclonal-Antibody/. Accessed 14 Jan 2015

4. European Public Assessment Report: Omnitrope. Available at: http://www.ema.europa.eu/docs/en_GB/document_library/EPAR_-_Summary_for_the_public/human/000607/WC500043689.pdf. Accessed 18 Sept 2015

5. Health Canada Summary Basis of Decision for Inflectra. Available at:http://www.hc-sc.gc.ca/dhp-mps/prodpharma/sbd-smd/drug-med/sbd_smd_2009_omnitrope_113380-eng.php. Accessed 18 Sept 2015

6. European Medicines Agency: European medicines agency recommendsapproval of first two monoclonal antibody biosimilars [Press Release].(2013)

7. Health Canada Summary Basis of Decision for Remsima. Available at:http://www.hc-sc.gc.ca/dhp-mps/prodpharma/sbd-smd/drug-med/sbd_smd_2014_remsima_160195-eng.php. Accessed 18 Sept 2015

8. Health Canada Summary Basis of Decision for Inflectra. Available atLhttp://www.hc-sc.gc.ca/dhp-mps/prodpharma/sbd-smd/drug-med/sbd_smd_2014_inflectra_159493-eng.php. Accessed 18 Sept 2015

9. Biosimilar Implementation, a Progress Report from FDA: Hearing beforethe Committee on Health, Education, Labor, and Pensions, Senate, 114thCongress (2015)

10. Aitken, M.L.: Medicine use and shifting costs of healthcare. IMS institutefor healthcare informatics, Parsippany (2014)

11. Boven, K., Knight, J., Bader, F., Rossert, J., Eckardt, K.U., Casadevall, N.:Epoetin-associated pure red cell aplasia in patients with chronic kidneydisease: solving the mystery. Nephrol. Dial. Transplant. 20(Suppl 3), iii33–iii40 (2005)

12. The difference with biologics: the scientific, legal, and regulatory chal-lenges of any follow-on biologics scheme [White Paper]. BiotechnologyIndustry Organization (2007)

13. Woodcock, J., Griffin, J., Behrman, R., Cherney, B., Crescenzi, T., Fraser,B., Hixon, D., Joneckis, C., Kozlowski, S., Rosenberg, A., Schrager, L.,Shacter, E., Temple, R., Webber, K., Winkle, H.: The FDA’s assessment offollow-on protein products: a historical perspective. Nat. Rev. DrugDiscov.6(6), 437–442 (2007)

14. Beck, A., Sanglier-Cianferani, S., VanDorsselaer, A.: Biosimilar, biobetter,and next generation antibody characterization by mass spectrometry. Anal.Chem. 84(11), 4637–4646 (2012)

15. Chumsae, C., Gaza-Bulseco, G., Sun, J., Liu, H.: Comparison of methio-nine oxidation in thermal stability and chemically stressed samples of afully human monoclonal antibody. J. Chromatogr. B Anal. Technol.Biomed. Life Sci. 850(1/2), 285–294 (2007)

16. Burkitt, W., Domann, P., O’Connor, G.: Conformational changes in oxi-datively stressed monoclonal antibodies studied by hydrogen exchangemass spectrometry. Protein Sci. 19(4), 826–835 (2010)

17. Kroon, D.J., Baldwin-Ferro, A., Lalan, P.: Identification of sites of degra-dation in a therapeutic monoclonal antibody by peptide mapping. Pharm.Res. 9(11), 1386–1393 (1992)

18. Jefferis, R.: Recombinant antibody therapeutics: the impact of glycosyla-tion on mechanisms of action. Trends Pharmacol. Sci. 30(7), 356–362(2009)

19. Harris, L.J., Skaletsky, E., McPherson, A.: Crystallographic structure of anintact IgG1 monoclonal antibody. J. Mol. Biol. 275(5), 861–872 (1998)

20. Arbogast, L.W., Brinson, R.G., Marino, J.P.: Mapping monoclonal anti-body structure by 2D 13C NMR at natural abundance. Anal. Chem. 87(7),3556–3561 (2015)

21. Harn, N., Allan, C., Oliver, C., Middaugh, C.R.: Highly concentratedmonoclonal antibody solutions: direct analysis of physical structure andthermal stability. J. Pharm. Sci. 96(3), 532–546 (2007)

22. Bueno, J., Long, D., Kauffman, J.F., Arzhantsev, S.: Deep-ultravioletresonance Raman (DUVRR) spectroscopy of therapeutic monoclonal anti-bodies subjected to thermal stress. Anal. Chem. 87(15), 7880–7886 (2015)

23. Marcoux, J., Champion, T., Colas, O., Wagner-Rousset, E., Corvaia, N.,VanDorsselaer, A., Beck, A., Cianferani, S.: Native mass spectrometry andion mobility characterization of trastuzumab emtansine, a lysine-linkedantibody drug conjugate. Protein Sci. 24(8), 1210–1223 (2015)

24. Chen, G., Warrack, B.M., Goodenough, A.K., Wei, H., Wang-Iverson,D.B., Tymiak, A.A.: Characterization of protein therapeutics by massspectrometry: recent developments and future directions. Drug Discov.Today 16(1/2), 58–64 (2011)

25. Engen, J.R.: Analysis of protein conformation and dynamics by hydrogen/deuterium exchange MS. Anal. Chem. 81(19), 7870–7875 (2009)

26. Cohen, M.J., Karasek, F.W.: Plasma chromatography—a new dimensionfor gas chromatography and mass spectrometry. J. Chromatogr. Sci. 8(6),330–337 (1970)

27. Giles, K., Pringle, S.D., Worthington, K.R., Little, D., Wildgoose, J.L.,Bateman, R.H.: Applications of a traveling wave-based radio-frequency-only stacked ring ion guide. Rapid Commun. Mass Spectrom. 18(20),2401–2414 (2004)

28. Clemmer, D.E., Hudgins, R.R., Jarrold, M.F.: Naked protein conforma-tions: cytochrome c in the gas phase. J. Am. Chem. Soc. 117(40), 10141–10142 (1995)

29. Giles, K., Wildgoose, J.L., Langridge, D.J., Campuzano, I.: A method fordirect measurement of ion mobilities using a travelling wave ion guide. Int.J. Mass Spectrom. 298(1/3), 10–16 (2010)

30. Bush, M.F., Hall, Z., Giles, K., Hoyes, J., Robinson, C.V., Ruotolo, B.T.:Collision cross sections of proteins and their complexes: a calibrationframework and database for gas-phase structural biology. Anal. Chem.82(22), 9557–9565 (2010)

31. Ruotolo, B.T., Benesch, J.L., Sandercock, A.M., Hyung, S.J., Robinson,C.V.: Ion mobility-mass spectrometry analysis of large protein complexes.Nat. Protoc. 3(7), 1139–1152 (2008)

32. Zhong, Y., Han, L., Ruotolo, B.T.: Collisional and Coulombic unfolding ofgas-phase proteins: high correlation to their domain structures in solution.Angew. Chem. Int. Ed. Engl. 53(35), 9209–9312 (2014)

33. Jarrold, M.F., Honea, E.C.: Annealing of silicon clusters. J. Am. Chem.Soc. 114(2), 459–464 (1992)

34. Debaene, F., Wagner-Rousset, E., Colas, O., Ayoub, D., Corvaïa, N., VanDorsselaer, A., Beck, A., Cianférani, S.: Time resolved native ion-mobilitymass spectrometry to monitor dynamics of IgG4 Fab arm exchange andBbispecific^ monoclonal antibody formation. Anal. Chem. 85(20), 9785–9792 (2013)

35. Damen, C.W.N., Chen, W., Chakraborty, A.B., van Oosterhout, M.,Mazzeo, J.R., Gebler, J.C., Schellens, J.H.M., Rosing, H., Beijnen, J.H.:Electrospray ionization quadrupole ion-mobility time-of-flight mass spec-trometry as a tool to distinguish the lot-to-lot heterogeneity in N-glycosylation profile of the therapeutic monoclonal antibody trastuzumab.J. Am. Soc. Mass Spectrom. 20(11), 2021–2033 (2009)

36. Bagal, D., Valliere-Douglass, J.F., Balland, A., Schnier, P.D.: Resolvingdisulfide structural isoforms of IgG2monoclonal antibodies by ionmobilitymass spectrometry. Anal. Chem. 82(16), 6751–6755 (2010)

37. Pacholarz, K.J., Porrini, M., Garlish, R.A., Burnley, R.J., Taylor, R.J.,Henry, A.J., Barran, P.E.: Dynamics of intact immunoglobulin G exploredby drift-tube ion-mobility mass spectrometry and molecular modeling.Angew. Chem. Int. Ed. Engl. 53(30), 7765–7769 (2014)

38. Ruotolo, B.T., Hyung, S.J., Robinson, P.M., Giles, K., Bateman, R.H.,Robinson, C.V.: Ion mobility-mass spectrometry reveals long-lived, un-folded intermediates in the dissociation of protein complexes. Angew.Chem. Int. Ed. Engl. 46(42), 8001–8004 (2007)

39. Guidance for Industry: Bioanalytical Method Validation. Washington, DC:US Department of Health and Human Services, FDA, CDER and CVM(2013)

40. Shelimov, K.B., Clemmer, D.E., Hudgins, R.R., Jarrold, M.F.: Proteinstructure in vacuo: gas-phase confirmations of BPTI and cytochrome c. J.Am. Chem. Soc. 119(9), 2240–2248 (1997)

41. Valentine, S.J., Anderson, J.G., Ellington, A.D., Clemmer, D.E.: Disulfide-intact and -reduced lysozyme in the gas phase: conformations and pathwaysof folding and unfolding. J. Phys. Chem. B 101(19), 3891–3900 (1997)

42. Valentine, S.J., Counterman, A.E., Clemmer, D.E.: Conformer-dependentproton-transfer reactions of ubiquitin ions. J. Am. Soc. Mass Spectrom.8(9), 954–961 (1997)

C. N. Ferguson and A. C. Gucinski-Ruth: IM-MS Evaluation for mAb Comparative Analysis 833