harge heterogeneity profiling of monoclonal antibodies using lowonic strength ion-exchange chromatography and well-controlled pHradients on monolithic columns
ohammad Talebia, Anna Nordborga, Andras Gaspara, Nathan A. Lacherb, Qian Wangb,iaoping Z. Heb, Paul R. Haddada, Emily F. Hildera,∗
Pfizer Analytical Research Centre (PARC) and Australian Centre for Research on Separation Science (ACROSS), School of Chemistry, University of Tasmania,obart, Tasmania, AustraliaAnalytical R&D, Pfizer BioTherapeutics Pharmaceutical Sciences, Chesterfield, MO, USA
r t i c l e i n f o
rticle history:eceived 13 May 2013eceived in revised form 13 August 2013ccepted 16 August 2013vailable online 21 August 2013
In this work, the suitability of employing shallow pH gradients generated using single componentbuffer systems as eluents through cation-exchange (CEX) monolithic columns is demonstrated for thehigh-resolution separation of monoclonal antibody (mAb) charge variants in three different biopharma-ceuticals. A useful selection of small molecule buffer species is described that can be used within verynarrow pH ranges (typically 1 pH unit) defined by their buffer capacity for producing controlled andsmooth pH profiles when used together with porous polymer monoliths. Using very low ionic strengtheluents also enabled direct coupling with electrospray ionisation mass spectrometry. The results obtainedby the developed pH gradient approach for the separation of closely related antibody species appear tobe consistent with those obtained by imaged capillary isoelectric focusing (iCE) in terms of both resolu-tion and separation profile. Both determinants of resolution, i.e., peak compression and peak separationcontribute to the gains in resolution, evidently through the Donnan potential effect, which is increasedby decreasing the eluent concentration, and also through the way electrostatic charges are distributed
on the protein surface. Retention mechanisms based on the trends observed in retention of proteins atpH values higher than the electrophoretic pI are also discussed using applicable theories. Employingmonolithic ion-exchangers is shown to enable fast method development, short analysis time, and highsample throughput owing to the accelerated mass transport of the monolithic media. The possibility ofshort analysis time, typically less than 15 min, and high sample throughput is extremely useful in theassessment of charge-based changes to the mAb products, such as during manufacturing or storage.
. Introduction
The advances in biotechnology in the last quarter of the 20thentury have led to the development of new technologies for theroduction of complex biomolecules which could potentially besed in human health care in the areas of diagnostics, preventionnd treatment of diseases. Qualities, such as high (target) selec-ivity, the ability to initiate immune recognition of the target, andong circulation half lives, have made the development of human-sed mAbs the fastest growing segment of therapeutic drugs [1,2]. In
he production of mAbs the final product often exhibits a number ofariations from the expected or desired structure. These alterationsay result from either known or novel types of posttranslational
modifications or from spontaneous, non-enzymatic protein degra-dation which bring about charge and size heterogeneity. Commonmodifications of the primary sequence include N-glycosylation,methionine oxidation, proteolytic fragmentation, and deamidation[3]. It has been shown that charge variants of therapeutic proteinscan have significantly different bioactivity. For example, Harris et al.[4] showed that deamidated variants of recombinant human mAbshad reduced potency in a bioactivity assay. As protein charge het-erogeneity is an important factor in quality assessment of proteintherapeutics, regulatory authorities such as the International Con-ference on Harmonisation (ICH) have set criteria for monitoringand characterising the degree and profile of variations to ensurelot-to-lot consistency and product stability [5].
Considering the large size of antibodies and the minor struc-tural diversity between the variants, the existence of these variantsimposes a great challenge for their separation. Ion-exchange (IEX)chromatography is a non-denaturing technique used widely to
eparate and isolate protein charge variants for subsequent charac-erisation. However, when operating under a salt gradient approachclassical mode), IEX chromatography has been shown to exhibitimited selectivity when complex proteins with the same numberf effective charges are to be separated [6] and lack of robustnesshen carboxypeptidase B (CPB)-treated mAbs are to be analysed
7].Capillary isoelectric focusing (CIEF) is another separation tech-
ique used frequently to assess charge heterogeneity of proteins inhich a complex mixture of ampholytes (polyionic organic elec-
rolytes) is used to establish a pH gradient into a capillary with theid of an electric field. The electric field causes protein isoforms toocus along the capillary according to their isoelectric point wherehey have zero net charge and then mobilise towards an on-columnetector located at one end of the capillary. Due to the distortionf the pH gradient, which affects reproducibility in migration timend peak area, the mobilisation step often requires optimisation [8].he introduction of imaged capillary electrophoresis (iCE), wheremaging is performed of an entire capillary, has overcome this issuey eliminating the need for the mobilisation step through singleoint detection. While CIEF is perhaps the most powerful of thenown separation technologies for charge variants, the difficultyf collecting fractions when compared to IEX chromatography hasonfined the method to be suitable for monitoring of variants butot for their preparative separation or isolation (peak identifica-ion) [2,6]. Also, some authors believe that while the separationsre consistent between the two methods, CIEF is not as precises IEX chromatography and therefore cannot be considered as auitable replacement [9]. To the contrary, however, some have con-luded that CE techniques could be superior to IEX chromatographyn terms of both separation speed and obtainable high resolutionnd therefore could constitute a routine tool for assessing chargeeterogeneity of proteins [8,10].
Developed by Sluyterman et al. [11–15] in the late 1970s,hromatofocusing (internal pH gradient) is recognised as thehromatographic analogy to IEF [9], mitigating many of the short-omings of classical IEX chromatography and combining somenique features of both methods. Chromatofocusing has beenemonstrated to be useful for separating protein isoforms due to
ts high resolving power and ability to retain the protein nativetate [7,16]. There are however some limitations to this techniqueuch as the cost of polyampholyte buffers employed, the necessityf column regeneration after each separation, and the inflexibilityn controlling pH gradient slope [7,17,18]. Alternatively, pH gra-ient can be conducted externally by pre-column mixing of twoluting buffers at different pH values consisting of common bufferpecies. As the slope and profile of the pH gradient can be easilyontrolled by changing the elution program with less dependencen the buffer composition and column chemistry, this manner ofntroducing pH gradients should allow for more convenient methodevelopment and optimisation [17,18]. The externally induced pHradient has been applied for separation of deamidated variants of
mAb [3], resolving C-terminal lysine isoforms of a mAb after treat-ng with carboxypeptidase B [7] and also for the analysis of chargeariants of full-length mAbs [9].
Currently, particle-packed columns represent the most commontationary phases for high performance liquid chromatography.espite immense popularity, their application for rapid and effi-ient separation of macromolecules is not as convenient as formall molecules. This is mostly because of slow diffusional massransfer of large solutes and also the large void volume existingetween the packed particles [19]. Additionally, biocompatibility
f stationary phases has become a new challenge when analysingiomolecules (including peptide and proteins). As defined by Lit al. [20], a biocompatible stationary phase material should be ableo resist non-specific adsorption of biomolecules and preserve the
A 1317 (2013) 148– 154 149
bioactivity of the target biomolecules. These challenges are wellmet by employing monolithic media. Mass transfer in monolithicsorbents is mostly dominated by convection, rather than diffusion,and is therefore fast, even for large biomolecules. On the otherhand, the expected biocompatibility of the most frequently usedpolymers in making porous monoliths, i.e., poly(meth)acrylate andpolyacrylamide, make these stationary phases highly suited for usein protein separation applications. We recently reviewed advancesin polymer monoliths for IEX chromatography of biomolecules andaddressed the importance of reducing non-specific interactionsbetween analyte and stationary phase [21]. While IEX chromatog-raphy of proteins using monolithic columns is frequently seen inthe literature [22–24], very little effort has been directed towardsemploying this technique for separation of large proteins, such asmAbs.
In continuing our recent efforts to resolve charge variants ofmAbs with the aid of IEX monolithic columns [25]; the maximumachievable resolution for mAb isoforms was pursued in this workusing CEX columns in combination with simple, yet efficient, buffersystems. Unlike previous reports [6,9,18,26], we operated IEX chro-matography employing shallow pH profiles over a limited pH range(typically 1 pH unit) generated by single component buffer sys-tems at very low ionic strength. The suitability of the proposedbuffer system in direct coupling of IEX chromatography to MSwas also demonstrated. Due to their size and complexity, mAbsare typically characterised by two or more orthogonal separationmethods [9]. Therefore, the performance of the developed methodwas also assessed by comparing the results with those obtained byiCE. It was hoped that similar charge heterogeneity profiles couldbe achieved for mAbs analysed under two different separationmechanisms.
2. Experimental
2.1. Reagents and chemicals
The buffering species used in this work, includingimidazole, piperazine dihydrochloride hydrate (PDH), andtris(hydroxymethyl)aminomethane (Tris), diethanolamine (DEA)and ammonium hydroxide (AMH), 28% (v/v) were all obtainedfrom Sigma–Aldrich (Sydney, Australia) and triethanolamine (TEA)was from BDH (Poole, England). Sodium chloride, hydrochloric acidand sodium hydroxide (98.8%), methanol (LC–MS grade) were alsofrom Sigma. All chemicals were of analytical grade unless specifiedotherwise. For iCE experiments, pharmalyte pH 3–10, sucrose andurea were obtained from Sigma–Aldrich, while methyl cellulose(1%) and the Chemical Test Kit were from ProteinSimple (formerlyConvergent Bioscience, Toronto, ON, Canada). The pI markersincluding pIs 5.13, 6.14, 7.2 and 9.5 were also obtained fromProteinSimple. Samples of three different IgG2 mAb formulations,which are referred to as mAb1, mAb2 and mAb3, were preparedby recombinant DNA technology at Pfizer Inc.
2.2. Chromatography
The IEX chromatography was performed on a Dionex DX-500 liquid chromatograph (Thermo Fisher Scientific, Lane Cove,Australia) consisting of a GP50 Gradient Pump, AD25 UV/VisAbsorbance Detector, AS50 Thermal Compartment and AS50Autosampler. Detection was performed at 280 nm. Flow-rate was1 mL/min, the injection volume was 10 �L and the column com-
partment temperature was set at 30 ◦C. Instrument control anddata acquisition were performed using Dionex Chromeleon soft-ware, version 6.80 SR5. Chromatograms were transferred to ASCIIfiles and redrawn using Origin 8.1 (Northampton, MA).
1 atogr. A 1317 (2013) 148– 154
PPcai
bwBsblcecFtBweam
wtlpta
2
1bawow
soA(rcflplma
2
ww(flw1Tpt
Fig. 1. pH gradient profiles obtained for mAb1 and mAb2. The mobile phase com-position was 12.5 mM DEA and 12.5 mM TEA to either pH 7.75 (eluent A) or pH10 (eluent B). Gradient: 0–100% B in 10 min, 100% B for 3 min. Column: ProSwift
50 M. Talebi et al. / J. Chrom
The monolithic IEX columns used were ProSwiftTM SCX-1S androSwiftTM WCX-1S (4.6 mm × 50 mm) and the packed column wasroPac WCX-10, 4 mm × 250 mm, all from Dionex. The monolithicolumns are methacrylates-based with sulfonic acid and carboxyliccid functionality for SCX and WCX, respectively. The ProPac WCXs a tentacle type ion-exchanger bearing carboxylate groups.
Unless otherwise stated, mobile phases were generally preparedy dissolving appropriate amounts of the buffer components inater prior to splitting into two aliquots denoted as eluent A and. The pH of each portion was then adjusted with concentratedodium hydroxide or hydrochloric acid. The elution was performedy a linearly ascending pH gradient from 0% to 100% eluent B fol-
owed by isocratic elution for 3 min before returning the eluentomposition to the starting condition (100% eluent A). The gradi-nt volumes were 10 and 30 mL for monolithic and packed columns,orresponding to about 14 and 10 column volumes, respectively.or each elution, the column was pre-equilibrated with at leasthree column volumes of eluent A prior to sample introduction.efore measurement of peak areas, each sample chromatogramas subtracted from the relevant blank injection prepared from
luent A. Fractions of the column eluent were collected every 1 minnd the offline pH measurement was carried out using a pH metreodel labCHEM-CP from TPS (Springwood, QLD, Australia).All eluents were prepared using water purified via a Milli-Q
ater purification system (Millipore, Bedford, MA) and filteredhrough a 0.2 �m nylon filter prior to use. mAb samples were ana-ysed as received without buffer exchange or any other sampleretreatment process. After dilution in eluent A to a concentra-ion of approximately 0.3 mg/mL, samples were stored at 5 ◦C untilnalysed.
CEX chromatography was carried out using a ProSwiftTMWCX-S (4.6 mm × 50 mm) column under pH gradient mode. 5 mM AMHuffer containing 20% (v/v) methanol at pH 9.5 was used as eluent And at pH 10.5 as eluent B. pH of eluents was adjusted before mixingith methanol. Elution was performed by running a linear gradient
f eluent A to eluent B in 20 min at a flow-rate of 0.4 mL/min, whichas split (1:100) before introducing into MS.
Hyphenated with the CEX chromatography, electrospray ioni-ation time of flight (ESI-TOF) mass spectrometry was performedn a micrOTOF-Q mass spectrometer (Bruker Daltonics, Melbourne,ustralia) equipped with an Agilent G1385A microflow nebuliser
Agilent technologies, Melbourne, Australia). The instrument wasun in a positive ion mode with m/z range of 500–10,000 and aapillary voltage of 4500 V (−500 V end plate offset). Drying gasow of 5 L/min at 300 ◦C was used with a 20.3 psi nebuliser gasressure. The instrument was tuned and calibrated using an Agi-
ent ES Tuning Mix (catalogue no. G2431A) in enhanced quadraticode. The deconvolution of ESI mass spectra was performed using
maximum entropy algorithm (Bruker Daltonics).
.4. Imaged capillary electrophoresis (iCE)
iCE profiles of mAbs were obtained using an iCE280 analyserith operational software from Convergent Bioscience, equippedith an Alcott 719 AL autosampler. A transparent capillary column
50 mm, 100 �m i.d.) was used with its inner surface coated withuorocarbon to minimise electroosmotic flow. The test solutionsere prepared using various amounts of pI markers, pharmalyte,
% methyl cellulose, 5 M urea, 20% sucrose, and mAb samples.hroughout the analysis, the capillary was kept at ambient tem-erature while the autosampler was set at 8 or 15 ◦C, depending onhe mAb sample analysed. The injection volume was 35 �L and the
SCX-1S (4.6 mm × 50 mm); Detection: UV at 280 nm; Flow-rate: 1 mL/min; Columncompartment temperature: 30 ◦C.
analysis was performed by applying a sample transfer time of 100 s,pre-focusing at 1500 V for duration of 1 min followed by focus-ing for 5 min at 3 kV. Detection was performed at 280 nm. Furtherdetails of the iCE conditions used are provided in the SupportingInformation.
3. Results and discussion
With the aim of improving the resolution, a series of new buffersystems based on both organic and inorganic buffer species weredesigned and applied using monolithic columns. To obtain suffi-cient binding of the proteins to the cation-exchanger, the lower pHof the gradient was chosen to be at least 1 pH unit below the elec-trophoretic pI values of mAbs, that is 8.8 for mAb1, 8.5 for mAb2and 8.4 for mAb3.
3.1. TEA-DEA buffer system
The first successful buffer system in eluting two of the mAbsof interest was prepared by mixing equimolar amounts of TEA(pKa 7.76) and DEA (pKa 8.88) resulting in a system buffering thepH range of approximately 7.5–10. Fig. 1 shows the separationachieved for mAb1 and mAb2 on a ProSwift SCX-1S column usingthis buffer system in the pH range of 7.75–10 with each buffer com-ponent at a concentration of 12.5 mM. A somewhat linear pH profilefor this system over the studied pH range was achieved (Fig. 1). Noelution was observed for mAb3. Acidic isoforms (pI lower than themain component) are observed for mAb1, while basic isoforms aremore pronounced for mAb2. Indications of additional isoforms arealso present, but as barely discernible shoulders of the main peaks.
The effect of flattening the pH gradient profile on chromato-graphic resolution was of special interest in this study. As the pHgradient slope is reduced there is more time for differential move-ment of the isoforms through the column, which could lead tobetter resolution [6]. In chromatofocusing, it is possible to generateshallow gradient slopes by limiting the pH range of the gradient orreducing the concentration of the mobile phase buffer components
[12,15,27]. Data presented later in this study show that these twostrategies in obtaining higher resolution are also applicable to theexternal pH gradient approach.
M. Talebi et al. / J. Chromatogr. A 1317 (2013) 148– 154 151
F25
3
oAtucs2Frtpas4
aIont
FP
effect of eluent pH range and gradient slope on resolving mAb1isoforms. By maintaining the gradient slope at 0.1 pH units/min,
ig. 2. The effect of eluent concentration (DEA) on the elution profile of mAb2. (A)0 mM, pH 9–10; (B) 10 mM, pH 9–10; (C) 5 mM, pH 9.2–10.2. For elution to occur at
mM concentration, more basic pH range is required. Other conditions as in Fig. 1.
.2. DEA buffer system
As seen in Fig. 1, elution of mAbs in the TEA-DEA buffer systemccurred around the end of the pH range applied. The pH of eluent
was therefore increased from 7.5 to 9. A simultaneous reduc-ion in gradient slope was achieved as the gradient time remainednchanged at 10 min. In addition, because of its negligible bufferapacity in the new pH range, TEA was removed from the bufferystem. The influence of buffer concentration within the range0–5 mM on separation efficiency of mAb2 isoforms is shown inig. 2. A decrease in buffer concentration at the same gradient slopeesults in an increase in the resolution of the charge variants fromhe main peak. For elution at 5 mM, a further increase in workingH range from 9–10 to 9.2–10.2 is required. These findings are ingreement with Farnan and Moreno [9], who achieved improvedeparation efficiency and higher resolution for mAb isoforms by a-fold decrease in the concentration of buffer composition.
The impact of column chemistry on separation efficiency waslso evaluated for mAb1 (Fig. 3) and mAb2 (Fig. S1 in the Supportingnformation). As can be seen, a trivial impact of column chemistry
n the selectivity is recognisable. However, there are more promi-ent fluctuations in the pH profile and a longer titration time forhe weak cation exchanger (see pH profiles). As the working pH
ig. 3. Comparison of separation of variants for mAb1 on ProSwift WCX-1S androSwift SCX-1S columns. Eluent: 5 mM DEA, pH 9.2–10.2.
Fig. 4. Interrelationship between eluent concentration and pH range on separationefficiency of mAb1. The gradient slope was 0.1 pH units/min. Eluent: 5 mM AMH,pH 9.2–10.2 (A); 2.5 mM AMH, pH 9.5–10.5 (B).
range is high enough to ensure full ionisation of the carboxylicgroup of the weak cation exchanger (pKa ∼ 5), the reason for dif-ferences in the pH profile might be due to the different chemistriesof the stationary phases [25].
3.3. AMH buffer system
Although suitable for resolving the isoforms of given mAbs, thelow volatility of DEA might limit its application for mass spectro-metric detection. In order to address this issue, we explored theuse of AMH which is a volatile buffer species with pKa 9.25. For thisbuffer, acceptable chromatographic resolution of protein isoformswas obtained for even lower concentrations than 5 mM (Fig. 4). Thisindicates that the focusing effect of the buffer system increases bydecreasing the concentration. Based on earlier results, the optimumpH range had to be adjusted when decreasing the eluent concen-tration to allow maximum separation efficiency. Fig. 5 displays the
it was found that although the fine structure of the acidic region
Fig. 5. Influence of operational pH range and gradient slope on resolution of mAb1variants. Eluent: 2.5 mM AMH. pH range and gradient slope: 9.3–10.3 and 0.1 (A);9.5–10.5 and 0.1 (B); 9.7–10.5 and 0.08 pH units/min (C).
1 atogr.
rhw9siaa
e(9atwtmpba5f
muadiiwttttttbfioceetbt
3
maaHSee
rw
(
wtDp
52 M. Talebi et al. / J. Chrom
emains unaltered (Fig. 5, traces A and B), basic variants previouslyidden within the threshold of the major peak were clearly resolvedhen the pH range was raised 0.2 pH units further from 9.3–10.3 to
.5–10.5. This step-wise optimisation protocol illustrates the pos-ibilities offered when using a pH gradient over a narrow pH range,n that it enables not only formation of controlled pH profile, butlso permits the fine tuning of pH within the range defined by thepplied buffer system to obtain the desired separation efficiency.
Interestingly, it was found that low ionic strength eluents gen-rated a significant back-pressure with the ProPac WCX-10 columnpressure upper limit = 120 bar). Once eluted with 5 mM AMH pH.5 at 0.5 mL/min, the initial back-pressure of 94 bar was monitorednd found to increase gradually. This behaviour is most likely dueo the osmotic pressure generated from the difference between theater content of the very dilute eluent and the IEX sorbent. Unlike
he packed column, the higher permeability and rigid structure ofonolithic ion-exchangers resulting from their porous properties
ermits fast generation of pH gradients at moderate and stableack-pressure (<70 bar) even at very low buffer concentrations,s well as minimising column titration times (typically less than
min). These merits offer a rapid analysis time that is applicableor high-throughput process development.
While quite successful in resolving charge heterogeneity ofAb1 and mAb2, the simplified buffer systems failed to elute mAb3
nless the eluent ionic strength was increased through addition of salt. Rozhkova [7] has previously reported the suitability of con-ucting pH gradient separation of mAb variants by adding NaCl
nto eluents. Accordingly, 2.5 mM AMH eluents, pH 9–10 contain-ng different concentrations of NaCl ranging from 20 to 40 mM
ere used for eluting mAb3. Results indicate partial resolving ofhe main component from part of the acidic species (Fig. S2 inhe SI). Basic variants, however, remained entirely hidden underhe wide shoulder of the major peak. One possible explanation forhis strong retention might be the differences in modification site,ype of modification, and/or degree of modification occurring inhe protein [2], all of which influence the strength of interactionsetween the protein molecule and the ion-exchanger. These modi-cations vary from those that change the number of charge residuesn the surface of the protein to those being less connected to theharge but can change antibody conformation. Deamidation, forxample, is one possible modification which is likely to have anffect on retention of a protein by affecting the number of posi-ively charge groups over the surface of a protein and hence itsinding to a cation-exchanger [3]. Further investigation is requiredo confidently determine the characteristics of the mAb variants.
.4. Effects of eluent concentration and pH on resolution
The overriding consideration in this work was towards maxi-um achievable resolution for mAb isoforms. pH and ionic strength
re two major characteristics of the eluent governing the elutionnd separation of proteins in pH gradient IEX chromatography.ere, we take advantage of the general expressions proposed byluyterman and Elgersma [14] for the pH gradient approach toxplain the interplay between these two parameters and theirffects on separation efficiency.
Peak width and peak separation are the two determinants ofesolution. The width of a protein band in terms of pH units can beritten as:
�pH)2 ≈ D(dpH/dV)ϕ(dZ/dpH)
(1)
here D denotes the diffusion coefficient of a protein, dpH/dVhe pH gradient slope and ϕ is equivalent to the dimensionlessonnan potential [14]. This equation implies that an increase ineak focusing is consistent with the lower ionic strength (buffer
A 1317 (2013) 148– 154
concentration) used, which increases the absolute value of ϕ. Evi-dence of this inference can be seen in Fig. 2, in which the resolutiongain for the mAb2 main isoform can be related to the focusing effectobtained by decreasing the ionic strength. In fact, the capabilityof focusing eluent bands is known as one inherent advantage ofpH gradient IEX chromatography over conventional salt gradientat a fixed pH [18], in which the absence of a focusing effect canbe partly related to the lack of the Donnan potential, as a resultof the high salt concentration involved. Trace C in Fig. 2 indicatesthat while employing the same pH range is likely to maintain thedZ/dpH unchanged, the positive effect of this kinetic factor on peakwidth can be highlighted by shifting up the pH range further, whichalong with more decrease in ionic strength leads to an even greaterincrease in resolution. The dominating effect of dZ/dpH on peakfocusing can also be seen by comparing traces A and B in Fig. 5where there is likely no significant difference between the Don-nan potentials due to the constant eluent concentration (2.5 mM).As should be expected, the peaks became broader when the pHgradient slope (dpH/dV), as another determinant of peak width inEq. (1), decreased further from 0.1 (trace B) to 0.08 pH unit mL−1
(trace C) by keeping the other conditions unchanged, probably dueto the domination of another kinetic determinant, i.e., diffusioncoefficient of protein (D). This therefore suggests that the rate oftitrating the ion-exchanger with pH has become lower than theequilibrium state of protein molecules, which could compromisethe peak focusing gains from shallower gradients.
The contribution of the other factor governing resolution, i.e.,peak separation, appears to be the main influence on resolutiongains for isoforms in Fig. 4, where the peak focussing for main iso-forms seems to be compromised, despite the expected focussingeffects as the eluent concentration decreases and the pH rangeshifts up further. In fact, almost all of the posttranslational mod-ifications and degradations can change surface charge propertiesof an antibody, either directly by changing the number of chargedgroups or indirectly by introducing conformational alterations [2].According to the electrostatic model developed by Tsonev and Hirsh[6] there is a relationship between the magnitude of a shift inelectrophoretic pI and the relative charge distribution in a givenprotein. This, in turn, implies that isoforms can be resolved basedon their apparent isoelectric point (pIapp, being the pH at which theprotein is eluted from the column) [12] when titrating by a gra-dient of pH, relating the resolution achieved in Fig. 4 to a greaterseparation of the peaks (for more discussion on pIapp and retentionmechanism see SI). Similar arguments based on the distribution ofcharges on the surface of a protein have also been used by otherworkers to explain the trends observed in resolution for chromato-focusing of �-lactoglobulin A and B [12], and haemoglobin variants[12,28].
3.5. Loading capacity
The loading capacity of the proposed approach for the separa-tion of mAb charge variants was also assessed. While some minorloss of resolution occurred when a sample load of about 118 �gmAb1 was injected onto the column, the overall separation pat-tern and the fine structure of the acidic region remained unaltered(Fig. S3 in the SI). By considering the low ionic strength of thebuffer system employed, a significant potential of this approachfor scale up can be seen, enabling it to be used along with classicalIEX chromatography for preparative purposes.
3.6. Profiling charge heterogeneity of mAbs by iCE
To assess the resolving power offered by the developed proce-dure, analysis of mAbs by iCE was also included in the study. Thedifference in separation mechanism of each technique can offer
M. Talebi et al. / J. Chromatogr. A 1317 (2013) 148– 154 153
Table 1Reproducibility and percentage of isoforms resolved by CEX chromatography and iCE.
RSDs % of the measurements are given in parentheses (n = 5 for CEX and 3 for iCE).a 5 mM DEA, pH 9.2–10.2.b 2.5 mM AMH, pH 9.5–10.5.
Fd
oufiwbes
(ammonium adducts) from 24+ to 30+. Further magnification of
ig. 6. iCE profiles for mAb1 (A), mAb2 (B) and mAb3 (C) (analysis conditions areescribed in the Supporting Information).
rthogonal and complementary information required to obtain annambiguous assignment of protein variants. The obtained pro-les (see Fig. 6) clearly demonstrate the distribution of isoformsithin acidic, main and basic species of the mAbs. A comparison
etween iCE profiles and IEX chromatograms reveals some inter-sting similarities in separation efficiency. For example, there is atriking similarity in resolving mAb1 acidic isoforms between the
two separation methods, although iCE still offers more resolutionin the basic region (compare Figs. 5 and 6). However, the resolutionbetween the main isoform and variants appears to be superior tothat obtained using iCE when the DEA buffer system is employed(see Fig. 3). Some similarities in resolving isoforms between pHgradient IEX and iCE can also be seen for mAb2 (see trace B inFig. 6 and trace C in Fig. 2). As an indication of peak purity, the peakarea percent for the sum of the acidic species, the major peak andthe sum of the basic species was also compared in both separationmethods (Table 1). It was not possible to achieve equivalent sepa-ration efficiency for all of the different antibodies analysed by IEXchromatography under identical separation conditions. The DEAbuffer system for mAb1 and AMH for mAb2 provide closer com-parison between iCE and IEX peak area percent data. For mAb3,while increasing the ionic strength of the buffer system through theaddition of an inert salt provided better separation efficiency thanthat obtained in our previous work [25], this strategy still appearsinadequate in providing resolutions comparable to iCE.
Further comparison between the obtained profiles highlightsanother interesting correlation between the two techniques. Thefour major variants are eluted within approximately 0.6 pH unitfrom IEX (see the pH profile in Fig. 3) and within 0.5 pH unitfrom iCE, indicating the similarity of the separation efficiency ofthe two approaches, despite the differences in their separationmechanisms. A similar observation was reported by Wang et al.in comparing the performance of IEX and CIEF for the separation ofmAb variants [29].
3.7. LC–MS analysis
Unlike RPLC, IEX chromatography is not readily hyphenatedwith MS due to the high content of salt involved [30]. Therefore,the integrated MS-based strategies are traditionally performed inmulti-dimensional approaches relying on the separation efficiencyof CEX chromatography for protein fractionation in the first dimen-sion and MS compatibility of RPLC in the second dimension. Thefeasibility of the direct coupling of the developed low ionic strengthpH gradient with MS was investigated by employing AMH buffersystem with the ProSwift WCX-1S column and mAb1 as analyte.Methanol was added to both eluents in order to enhance the ioni-sation efficiency. While this mixture was found to be stable enoughfor at least a day, precaution was taken to avoid irreproducibilityof day-to-day analyses by refreshing the eluents each day. Fig. 7shows the averaged mass spectrum of the intact mAb1 after pHgradient elution. The averaged mass spectrum was acquired in therange from 12 to 22 min from the corresponding base peak chro-matogram (Fig. S4 in the SI) and features multiply charged ions
the spectrum (see insets) revealed more charge states which can beattributed to the existence of different species (isoforms). To gainmore insight into the possible modifications, deconvolution of the
154 M. Talebi et al. / J. Chromatogr.
Fo
iarw1bortmro
4
ctfaamdbiRsdtiaouh
otta
aEra
[
[[[[[[[[[[[[[[[
[
[
ig. 7. ESI mass spectrum of intact mAb1 with insets showing the expanded viewf charge states of the antibody.
ntact mass spectrum was performed using a maximum entropylgorithm (Fig. S5 in the SI). Unfortunately due to the insufficientesolution, unambiguous recognition of the expected modificationsas not readily possible (typical mass resolution was 11,000 at m/z
522). Nevertheless, the observed mass difference of 1445 mighte attributed to one unit of glycan (G0F) attached to the N-terminalf the mAb, which is possible to be verified via a deglycosylationeaction. As a proof of concept, the results can suggest viability ofhis approach for the study of the existence of different isoforms of
Abs, especially in combination with MS detection having higheresolving power. Further investigation on this topic is the focus ofur future studies.
. Conclusions
The application of a shallow pH gradient generated by simpleomponent buffer systems within a narrow pH range in combina-ion with (CEX) monolithic columns provided remarkable resultsor the separation of mAbs in this study. The separation of basicnd acidic isoforms with qualities comparable to iCE was achievedt very low ionic strength. The expressions developed by Sluyter-an and Elgersma were used for interpreting the results, with the
ominant factors influencing both peak width and peak separationeing the Donnan potential, which is directly proportional to the
onic strength, and the rate of protein charge alteration with pH.esolution improvement through peak separation was hypothe-ised to occur via increasing the electrophoretic pI of variants to aifferent extent determined by the relative charge distribution onhe protein surface. This potentially confers separation capabilityn pH gradient IEX chromatography which is not applicable with
conventional salt gradient at a fixed pH. By keeping in mind thebserved trends of increasing pI, this concept might potentially besed for other mAbs by choosing a suitable buffer component thatas buffering capacity that can cover the protein pIapp.
While packed columns may not be readily used with the devel-ped dilute buffer systems due to the high back-pressure involved,he possibility of high-throughput analysis and fast re-equilibrationime can be considered as additional advantages of the proposedpproach when employing monolithic columns.
Employing volatile buffer species at very low concentration
lso enables direct interfacing of the ion-exchange separation toSI-MS. As inferred from the results, despite the elution occur-ing at pH values where both strong and weak CEX columnsre expected to be fully ionised, there are differences in elution
[[
[
A 1317 (2013) 148– 154
profiles, which most likely result from the type of ion-exchangers.Further studies of the pH gradient elution of mAbs might there-fore benefit from the inclusion of other ion-exchange chemistries.We expect that more improvements in separation efficiency shouldbe achieved through designing monolithic columns with highercapacities, offering higher Donnan potential, as well as higherbiocompatibility and porosity (for employing higher flow-rates),which is the focus of our future investigations.
Acknowledgements
The authors gratefully thank Dr. Kelly Flook from ThermoFisherScientific for providing the monolithic columns used in thiswork. E.F.H. acknowledges the Australian Research Council foran ARC Future Fellowship (FT0990521) and P.R.H. acknowledgesthe Australian Research Council for an ARC Federation Fellowship(FF0668673).
Appendix A. Supplementary data
Supplementary data associated with this article can befound, in the online version, at http://dx.doi.org/10.1016/j.chroma.2013.08.061.
References
[1] M.R. Etzel, W.T. Riordan, J. Chromatogr. A 1216 (2009) 2621.[2] J. Vlasak, R. Ionescu, Curr. Pharm. Biotechnol. 9 (2008) 468.[3] M. Perkins, R. Theiler, S. Lunte, M. Jeschke, Pharm. Res. 17 (2000) 1110.[4] R.J. Harris, B. Kabakoff, F.D. Macchi, F.J. Shen, M. Kwong, J.D. Andya, S.J. Shire,
N. Bjork, K. Totpal, A.B. Chen, J. Chromatogr. B 752 (2001) 233.[5] Q6B: Specifications Test Procedures and Acceptance Criteria for Biotechnology,
63, Federal Register, 1996, pp. 31506.[6] L.I. Tsonev, A.G. Hirsh, J. Chromatogr. A 1200 (2008) 166.[7] A. Rozhkova, J. Chromatogr. A 1216 (2009) 5989.[8] Z. Sosic, D. Houde, A. Blum, T. Carlage, Y. Lyubarskaya, Electrophoresis 29 (2008)
4368.[9] D. Farnan, G.T. Moreno, Anal. Chem. 81 (2009) 8846.10] D.M. Good, P. Zurbig, A. Argiles, H.W. Bauer, G. Behrens, J.J. Coon, M. Dakna,
S. Decramer, C. Delles, A.F. Dominiczak, J.H. Ehrich, F. Eitner, D. Fliser, M.Frommberger, A. Ganser, M.A. Girolami, I. Golovko, W. Gwinner, M. Haubitz,S. Herget-Rosenthal, J. Jankowski, H. Jahn, G. Jerums, B.A. Julian, M. Kellmann,V. Kliem, W. Kolch, A.S. Krolewski, M. Luppi, Z. Massy, M. Melter, C. Neususs,J. Novak, K. Peter, K. Rossing, H. Rupprecht, J.P. Schanstra, E. Schiffer, J.U.Stolzenburg, L. Tarnow, D. Theodorescu, V. Thongboonkerd, R. Vanholder, E.M.Weissinger, H. Mischak, P. Schmitt-Kopplin, Mol. Cell. Proteomics 9 (2010)2424.
11] L.A.Æ. Sluyterman, C. Kooistra, J. Chromatogr. 470 (1989) 317.12] L.A.Æ. Sluyterman, J. Wijdenes, J. Chromatogr. 206 (1981) 429.13] L.A.Æ. Sluyterman, J. Wijdenes, J. Chromatogr. 206 (1981) 441.14] L.A.Æ. Sluyterman, O. Elgersma, J. Chromatogr. 150 (1978) 17.15] L.A.Æ. Sluyterman, J. Wijdenes, J. Chromatogr. 150 (1978) 31.16] X. Kang, D. Frey, J. Chromatogr. A 991 (2003) 117.17] L. Shan, D.J. Anderson, J. Chromatogr. A 909 (2001) 191.18] L. Shan, D.J. Anderson, Anal. Chem. 74 (2002) 5641.19] S. Xie, R.W. Allington, F. Svec, J.M.J. Frechet, J. Chromatogr. A 865 (1999) 169.20] Y. Li, M.L. Lee, J. Sep. Sci. 32 (2009) 3369.21] A. Nordborg, E.F. Hilder, Anal. Bioanal. Chem. 394 (2009) 71.22] B. Gu, Z. Chen, C.D. Thulin, M.L. Lee, Anal. Chem. 78 (2006) 3509.23] B. Gu, Y. Li, M.L. Lee, Anal. Chem. 79 (2007) 5848.24] E. Hilder, F. Svec, J. Frechet, J. Chromatogr. A 1053 (2004) 101.25] A. Nordborg, B. Zhang, X.Z. He, E.F. Hilder, P.R. Haddad, J. Sep. Sci. 32 (2009)
2668.26] T. Ahamed, S. Chilamkurthi, B.K. Nfor, P.D. Verhaert, G.W. van Dedem, L.A. van
der Wielen, M.H. Eppink, E.J. van de Sandt, M. Ottens, J. Chromatogr. A 1194(2008) 22.
27] Y. Liu, D.J. Anderson, J. Chromatogr. A 762 (1997) 47.
28] X. Kang, D.D. Frey, Anal. Chem. 74 (2002) 1038.29] F. Wang, B. Peklansky, C. Anderson, Y. Wang, R.R. Rustandi, J. Liq. Chromatogr.
& Rel. Technol. 35 (2012) 1259.30] D.S. Rehder, T.M. Dillon, G.D. Pipes, P.V. Bondarenko, J. Chromatogr. A 1102
