Best Paper on Size Exclusion Chromatography

8/10/2019 Best Paper on Size Exclusion Chromatography

1/13
http://localhost/var/www/apps/conversion/tmp/scratch_6/dx.doi.org/10.1016/j.jpba.2014.04.011mailto:[email protected]://www.elsevier.com/locate/jpbahttp://www.sciencedirect.com/science/journal/07317085http://localhost/var/www/apps/conversion/tmp/scratch_6/dx.doi.org/10.1016/j.jpba.2014.04.011http://localhost/var/www/apps/conversion/tmp/scratch_6/dx.doi.org/10.1016/j.jpba.2014.04.011


2/13

Please cite this article in press as: S. Fekete, et al., Theory and practice of size exclusion chromatography for the analysis of proteinaggregates, J. Pharm. Biomed. Anal. (2014), http://dx.doi.org/10.1016/j.jpba.2014.04.011

ARTICLE IN PRESSG Model

PBA-9537; No.of Pages13

2 S. Fekete et al. / Journal of Pharmaceutical andBiomedical Analysis xxx (2014) xxxxxx

1. Introduction

The clinical use of therapeutic proteins has enabled the treat-ment of a wide range of life-threatening diseases. Many of thesediseases were considered incurable or untreatable only a fewdecades ago. Almost half of the new drugs recently approved bythe United States Food and Drug Administration (FDA) are ther-apeutic proteins, and the drug-pipeline landscape is now shiftingeven more toward this drug class of drugs [1]. There are severalhundred potential biopharmaceuticals that have been evaluatedin clinical trials for the treatment of genetic diseases, cancer andinfectious diseases [24]. Because the development of biopharma-ceuticals and biosimilars is complex, regulatory bodies such as theFDA and European Medicines Agency (EMA) require comprehen-sivedrugsubstance characterization, lot-to-lotandbatch-to-batchcomparisons, stability studies, impurity profiling, glycoprofilingand related protein and excipient determination, such as polysor-bate or protein aggregate elucidation [5,6].Due to the unique physical and chemical properties of ther-

apeutic proteins, they are prone to a number of changes duringtheirpreparation,formulationorstorage [7]. Thesechanges includeseveral possiblemodifications, suchasoxidation,deamination,gly-cosylation, aggregation, misfolding, or adsorption [811]. These

modifications couldlead tothepotentialloss of therapeutic efficacyor unwanted immune reactions [7]. Aggregate levels in protein-based drugs are a critical quality attribute due to their potentialimmunogenicity [12,13]. In recent studies, cytotoxic effects havealso been observedwith several biotherapeutic proteinsdueto sig-nificant denaturation or chemical alterations of the native protein[8,9,14]. In this context, the characterization of protein modifica-tions requires a number of analytical methods because one singletechnique does not allow for the assessment of all the requiredparameters [6].Protein aggregates are generally characterized using sodium

dodecylsulfate polyacrylamide gel electrophoresis (SDS-PAGE)[15], size-exclusion liquid chromatography (SEC-HPLC) [1618],asymmetricalfieldflowfractionation(AF4)[19], fluorescencespec-

troscopy[20], circulardichroism(CD)[21]or light-scattering-basedmethods, such asmulti-angle laser light scattering (MALLS) [22]. Inaddition, native mass spectrometry appears to be an alternative,emerging method to identify and characterize small oligomericprotein aggregates [2325].Among these techniques, SEC-HPLC is employed today as

the standard separation technique for the quantification of pro-tein dimers, trimers and oligomers. The main advantage of thisapproach is themildelutionconditions that allowfor thecharacter-ization of the protein withminimal impact on the conformationalstructureand the localenvironment.Recently, somedevelopmentshave been made in SEC-HPLC via the introduction of columnspacked with sub-2m particles or the tandem and parallel inter-laced approaches that enable a new level of chromatographic

performance or throughput [26,27]. Conversely, several detectiontechniques are now available, ensuring improved sensitivity. Theaim of this study is to review the possibilities and new develop-ments in SEC-HPLC applied to the analysis of protein aggregates.

2. Theory of SEC separation

SEC separates biomolecules according to their hydrodynamicradius. The stationary phase consists of spherical porous par-ticles with a carefully controlled pore size, through which thebiomolecules diffuse basedon theirmolecular sizedifferenceusingan aqueous buffer as the mobile phase.Basically, SEC is an entropically controlled separation process

in which molecules are separated on the basis of molecular size

differences(filtering)rather thanby theirchemical properties [28].In SEC, the partitioning is ideally driven by entropic processeswithout anyadsorption, thus thegenerally used Gibbs free-energyequation becomes

ln KD = S0R

(1)

where KDis the thermodynamic retentionfactor,S0 is thechangein the system entropy and R is the gas constant. However, it is

important to notice that in practice, temperature can indirectlyimpact the elution time to a small degree by altering the confor-mations of the proteins and affecting mobile phase viscosity andsolute diffusivity. The thermodynamic retention factor in SEC isdifferent from other chromatographic modes. Here, the thermo-dynamic retention factor is the fraction of the intraparticle porevolume that is accessible to the analyte [29].

Conversely, an analogous retention factor, k*, was suggested byEngelhardt as the ratio of the probabilities of the sample stayingin the stagnant mobile phase inside the pores and in the movingmobile phase between the particles (interstitial volume) [30]. InSEC, this k* value is limited, and its maximum value is given by theratio of the pore volume (Vp) and the interstitial volume (Vz) o f acolumn or similarly by the pore porosity (p) and the interstitial

porosity (z):

k =Velu Vz

Vz(2)

kmax =Vp

Vz=

p

z(3)

Therefore, the ratio of the pore porosity to the interstitialporosity determines the achievable retention range for a givenseparation. Usingcurrently availablestate-of-the-art columns, thismaximum retention factor is around k*1.61.8. These parame-ters and the associated selectivity can be altered by packinga column more densely (or decreasing particle size) or by usingparticleswith larger pore volumes.

In SEC, the size-based separation allows the construction of a

calibration curve based on a set of known analytes, which can beused toestimatethemolecularweightof anunknownanalyte.Typ-ical calibration curves arebased on proteins or polymers of knownmolecularweights. By plotting logMvs. the retentionvolume, onetypicallyobtains a third order polynomial with a linear regionpro-viding the highest resolution andmolecularweight accuracy [29].The largest proteins,which areexcluded from thepores, elutefirst.Subsequent proteins elute in order of decreasing size. As proteinsshapes could also vary (e.g., globular, rod-like or flexible chains),their Stokes radii do not correlate exactly with molecular weight.Another source of error in the calibration curve is that non-idealadsorption may alter the retention volume [3133].

Theslopeof theline in thelinear portion of thecalibrationcurveis a measure of the selectivity of the stationaryphase,which can be

defined by the relationshiplog M=m KD + b (4)

wherem and b are the slope and intercept of the line, respectively.As the pore size distribution of the particle becomes narrower, theslope becomes less steeper,which results in a greater selectivity todiscriminateanalytesof similar size.A typical SECcalibrationcurveis shown in Fig. 1.

Similarly to other separation techniques, the main interest ofmodern SEC is to improve the analysis throughput by reducing theanalysis time. Theanalysis time in SECismostlydetermined by themobilephase flowrate with a given columnconfigurationbecauseall of the analytes of interest are eluted before the total void time(volume) of the column (there is no partition and retention) [34].

To shorten the separation time in SEC, the ratio of the column void
http://localhost/var/www/apps/conversion/tmp/scratch_6/dx.doi.org/10.1016/j.jpba.2014.04.011http://localhost/var/www/apps/conversion/tmp/scratch_6/dx.doi.org/10.1016/j.jpba.2014.04.011


3/13




S. Fekete et al. / Journal of Pharmaceutical andBiomedical Analysis xxx(2014) xxxxxx 3

Fig. 1. Typical calibration curve in SEC.(Reprinted fromRef[29]).

volume to the flow rate needs to be decreased. Reducing the col-

umn sizeand increasing theflowrate arethestraightforwardwaysforperformingfastSECanalysis [34]. Shorteningthecolumn lengthhas been applied in fast SEC analysis to keep the backpressure ata reasonable level even at high flow rate when separating syn-thetic polymers [35,36]. However, decreasing the column lengthproportionally reduces the number of theoretical plates.Popovici and Schoenmakers reported several considerations

for fast separations using different commercial SEC columns atroom temperature [36]. The effects of particle size, column length,and mobile phase flow rate on SEC separation were studied andreported in the past [36,37]. The main difficulty in achieving bothhigh-speed andhigh-resolution separation in SEC is the slowmasstransfer of the large solutes between the interstitial space and thepore space [37]. To increase the mass transfer rate, the tempera-

ture could be a valuable parameter. As the temperature increases,the mobile phase viscosity decreases, and the analytes diffusiv-ity improves. Furthermore, this approach allows the use of higherflow rates for given column dimensions. As an example, Park et al.showed some very fast high temperature SEC separations, per-formed at 110 C [34]. Therefore, and according to theory, theanalysis time in SEC can be shortened by applying small parti-cles, shortcolumns andelevated column temperatures. Engelhardt

proposed the followingsemi-empirical relationship todescribe thecolumnperformance in SEC [30]:

H= 3.5 dp + 1.3(1+ k)DM

u + 0.6

k

(1+ k)2d2p

DMu (5)

where His the plate height, dp is the particle diameter, DMis themolecular diffusion coefficient and u is the mobile phase linearvelocity.Based on this model, some predictions have beenmade to

estimate the impact of the particle size and mobile phase temper-ature (DMdepends on the temperature) on the Hu plot as shownin Fig. 2.

Today, various particles of 320m are routinely applied withdifferent pore-sizes [28]. The most common column dimension inanalytical SECis 30cmcolumnlengthand 4.68mm internaldiam-eter. With these conventional SEC columns, the analysis time isgenerally between 15 and 50min.

When high-resolution separation is required, the separationpower (resolution) of an SEC column increases in direct propor-tion to the square root of the column length; thus, the separationof complex samples requires long columns that canbe obtained byjoining multiple columns in a series.

3. Occurrence of protein aggregation in pharmaceuticalformulations

Thecontrolof protein aggregationis alwaysa concernduringthepurification, formulation andmanufacturingof therapeutic proteinproducts [3740]. The kinetics or rates of aggregate formation areof primary interest because they determine the aggregate levelsover a given period of time in different physical environments andat different protein concentrations and temperatures experiencedby a protein during its processing, shipping, or storage.

At a minimum, aggregates represent a process related impurityand/or degradation product that must be controlled at relativelylowlevelsof concentrationthroughoutthemanufacturinganddur-ing theproduct storage [41]. Recentconcernshave also been raised

regarding thepotential immunogenicity of aggregates,particularlythose that are composed of multiple folded or partially foldedmonomers, although the precise mechanisms that make a par-ticular aggregate size, morphology, and/or structure more or lessimmunogenic remain uncertain [42]. Therefore, it is also of inter-est to differentiate between the rates of formation for aggregatesranging from small oligomers to soluble high molecular-weight(HMW) aggregates and larger, effectively insoluble particles.

Fig. 2. Theoretically expected impact of theparticle sizeand mobile phase temperature on columnperformance. (For the calculations, a 50kDa protein was assumed.).


4/13





Fig.3. Schematicrepresentationof someof thekey stepsin non-nativeaggregation.

(Reprinted fromRef[48]).

In this context, theterm soluble denotes species that aremolec-ularly dispersed in solution, while insoluble is used to denoteaggregates that have coalesced into a macro- or micro-phase sep-arated domains or particles. It has also been shown recently thatprotein aggregates can reversibly transition between soluble andinsoluble or condensed states [43,44]. Alternatively, one candifferentiate aggregates based on size measurement using otheranalytical techniques [38,45].

Assuming aggregation is the fastest degradation route, productshelf life and the effective rate coefficient of monomer loss (kobs)are equivalent quantities when one is concerned with relativelylow conversion percentages of monomer to aggregate (100nm) as the loss of total peak area [49]. However,there is no clear definition for aggregate acceptance limits in bio-pharmaceuticalproducts.The FDAandInternationalConferenceonHarmonization (ICH) consider aggregates to be process or productrelated impurities [50,51]. In the particular case of a biosimilarity

study, the FDA guidelines recommend the following: If compara-tive physicochemical analysis reveals comparable product-relatedimpurities at similar levels between the twoproducts (originatorsand biosimilar), pharmacological/toxicological studies to charac-terize potential biological effects of specific impurities may not benecessary. However, if themanufacturing processusedtoproducetheproposed biosimilar product introduces different impurities orhigherlevels of impurities than thosepresent inthereferenceprod-uct, additional pharmacological/toxicological or other studiesmaybe required [50].Themain challenge in analyzing the degraded/aggregated pro-

tein formulations lies in thecomplex, dynamic, heterogeneous andtransient nature of degraded/aggregated species. The number ofchemicalandphysicaldegradationpathwaysis veryhigh andmanyof them are intrinsically related.

4. Separation of aggregates in SEC

4.1. Possible interactions with the stationary phase

Proteins tend to interact with the charged surface sites of thestationary phase [29,3133,5254]. Non-binding interactions aredominated by two types of interactions: (1) electrostatic interac-tions and (2) hydrophobic interactions [32].

Due to the ionic interactions, protein adsorption can occur,inducing elution time shifting, band tailing or asymmetrical bandelution. The three-dimensional conformation of the proteins canalso be changed [31,55]. Electrostatic interactions can be classi-fied into two groups. (1) If the protein and the stationary phasesurface are identically charged, ion-exclusion can occur due tothe electrostatic repulsion. In this way, the protein is preventedfrom entering the pores of the particle and, therefore, elutes faster(decrease in elution time). (2) If the protein and the particle areoppositely charged, the adsorption of protein on the stationaryphase surface may result in increased elution time.

Possible hydrophobic interactions between the protein andhydrophobic sites of thestationary phase canalso lead to increasesin elution time.

The last generation of SEC materials is much more inert than

the older versions, minimizing the possible electrostatic andhydrophobic interactions between the proteins and stationaryphase surface. This behavior could be attributed to the regularimprovement of silica materials (type B silica or hybrid silica) andthe use of diol bonding. Except the nature of the stationary phasebonding and the interactions between the proteins and the sta-tionary phase can also be minimized by proper adjustment of themobile phase composition.

4.2. Mobile phase composition

To reduceelectrostatic interactions, a commonapproach in SECis to increase the ionic strength or salt concentration of the mobile

phase. In this way, the unwanted secondary interactions can bereduced, and the peak shapes can be improved and quantifica-tion can becomemore precise. The recovery of aggregates can alsobe increased by the addition of salts (e.g., 100mM sodium chlo-ride) [54,56]. It wasdemonstrated that byvarying themobile phaseionic strength, the retention time of some monoclonal antibodies(mAbs) was shifted, and poor peak shapes were observed at lowsodium chloride concentration [54]. The aggregate recovery wasalso improvedwhen increasing thesalt concentrationin themobilephase [56]. In addition to the ionic strength increase, another valu-able approach consists of increasing theconcentration of a counterion in the mobile phase. Indeed, very high concentrations of thesame ions can increase the hydrophobic or ion-exclusion effects[57]. Tomitigate theundesired secondaryionic interactions in SEC,

it is also recommended to use some additives in the mobile phase.Arginine is commonlyused as anSECadditive [58,59]. The additionof arginine reduces the possible secondary interactions betweenthe protein and the stationary phase as it can bind to the pro-tein, leading to an improvement in protein aggregates quantitationand peak shape [29,58,59]. As an example, Fig. 4 shows the SECchromatographic profiles of recombinant human basic fibroblastgrowthfactor (bFGF) in thepresenceandabsenceofarginine.How-ever, arginine shows significant UVabsorbance at lowwavelengths(below220nm) therefore can drastically decrease sensitivity.

Some organic modifiers (e.g., methanol and ethanol) could beused toreduce thestrengthof hydrophobicinteractionsdue totheirhighereluentstrength compared towater [60]. Therefore, thepeakshapes can be improved in some cases by the addition of 510% of

an organic modifier.


5/13





Fig. 4. SEC chromatographic profiles of recombinant human basic fibroblast growth factor (bFGF) in 0.2M NaCl (A) and 0.2M arginine (B). There is little resolution in theseparation of bFGF (arrow) from thesalt in (A), while baseline resolution is observed in (B).


The pH of the mobile phase can also affect the interactionsbetween proteins and the stationary phase. In an optimal case,the pH should be close to the isoelectric point (pI) of the proteinto limit secondary interactions. Indeed, there is a risk for ion-exchange interactions when the pH is lower than the pI, whileion-exclusion effects canoccurat pHvaluesabovethepI. Thiseffectbecomesespecially importantwhenworking at lowionicstrengths[33].

4.3. Columns

There are mostly two types of SEC packing materials: (1) sil-

ica, with or without surface modification, and (2) cross-linkedpolymeric packings, which possess non-polar (hydrophobic),hydrophilic, or ionic character [28]. The most common silicapackings consist of chemically bonded 1,2-propanediol functionalgroups that provide a hydrophilic surface. This stationary phaseblocksor reactswithmany of theacidicsilanol groupsandneutral-izing the surface. Therefore it makes it ideal for SEC separation ofbiopolymers [28]. Bare silica is also a useful packing material fornon-aqueous polar or non-polar organic mobile phases; however,it is not recommended with aqueous mobile phases because of thepresenceofactivesilanolsites.Thelatesttypeofsilica-relatedpack-ing isanethylene-bridgedhybrid inorganicorganic (BEH)materialthat iscurrentlyavailable atsizesof1.7mthefirstsub-2mSECpacking and 2.5m [26,61]. Compared to regular silica packings,

BEHparticles have improved chemical stability as well as reducedsilanol activity [61].There have been a number of different hydrophilic cross-

linked packings developed for the SEC of biopolymers. Most ofthese packings are proprietary hydroxylated derivatives of cross-linked polymethacrylates [28]. Unusual polymeric packings foraqueous SEC include sulfonated cross-linked polystyrene, polydi-vinylbenzenederivatizedwithglucoseoranion-exchangegroups, apolyamidepolymer,andhigh-performance,crossed-linkedagarose[28]. However, when using organic or polar SEC packings foraqueous SEC, one must prevent the adsorption that is possibleif there are extensive hydrophobic regions on the packing orpolymer.For interested readers, the most important SEC column

providers and products were recently reviewed in Ref. [28].

4.4. Method development

In any SEC method, a variety of parameters can beadjusted to improve the resolution between the native pro-tein and aggregates and make the method more robust andreliable.

First, an appropriate pore size has to be selected. The pore sizeselection depends on the size (or molecular weight) of the proteinanditsaggregatestobeseparated.Moleculeswithadiameterlargerthan the largest pores in the stationary phase are unable to enterthe porous volume of the particles. Therefore, they pass throughthecolumnandelute first with the interstitialvolume(Vz). Smallermolecules enter the pores within the particles and depending ontheir apparent size they elute later in order of decreasing size. For

protein characterization, typical pore sizes between 150and500Aareapplied.Forcommontherapeuticproteins(MW1580kDa),apore size of 150200A is well suited, while a 200300- pore sizeis usually applied for mAbs (MW150kDa). For very large pro-teins (MW>200kDa, e.g., Pegloticase or PEGylated proteins), the5001000A phases offer the best selectivity.

Resolution between the aggregates and native protein peaksis enhanced when decreasing particle size. The achievable platenumbers are inversely proportional to the particle size, while theresolution is inversely proportional to the square root of particlesize. Historically, 510m particles are commonly used in routineSECapplications, butseveral providers offer now3mparticles oreven sub-2m packings. By using these state-of-the-art columnspacked with very fine particles, the resolution can be significantly

improvedor theanalysistimecanbe shortenedby applyingshortercolumns. The use of very fine particles in SEC has shown clearadvantages in resolution [30,62].

Because SEC is an entropically controlled separation where noretentionoccurs, largepore volumes (high porosity)are required toensure appropriate selectivity. Generally, this large pore volume isprovided by long- andwide-bore columns. In routine SEC applica-tions,a 30cmcolumnlengthwith internal diameters (I.D.)of 6, 7.8,8 or 10mm is generally employed. These SEC columns are referredto as standard bore columns. Now, several vendors offer narrowbore columns with 4.6-mm I.D. and 15cm lengths that are packedwith very efficient, small particles. By using these columns, similarseparation power can be attained as with 5m particles in 30cmstandard bore columns, but the analysis time can be reduced by a

factor of 34 [26].


6/13





After selecting the appropriate column, themobile phase com-position has to be optimized. As mentioned in Section 4.2, anappropriate pH and salt concentration are required to ensure cor-rect peak shape and resolution. Generally, the pH should be closeto the isoelectric point of the protein, and the protein should besoluble and stable in the buffer. It is goodpractice touse a bufferedsolutionwith an ionic strengthof 50200mM. This conditionhelpsto reduce non-specific interactions between the protein and thestationary phase. The most commonly used buffer is phosphate at10200mM concentration and pH between 6 and 7.2. When theisoelectric point of the protein is unknown, a pH of 6.5 may be agood starting point. To improve the peak shape, sodium chloride,potassiumchlorideor sodiumsulfate isoftenaddedintothemobilephase at a concentration of 10250mM. However, when workingat low ionic strength, non-ideal conditions can improve selectiv-ity in some cases. Enhancement in selectivity and resolution wasreportedformodelproteinsinthesorptivemodewithmobilephasepH below the pI[63]. In some cases, the addition of organic mod-ifiers, such as methanol or ethanol, can also help to improve theresolution andpeakshapedueto a reductionofpossiblehydropho-bic interactionsbetween proteins and the surface of the stationaryphase. Please note that buffers with high salt concentration andorganic modifiers can be hazardous for the pump and system. Salt

precipitation canblock the tubing, seals or plungers. Therefore it issuggested to flush the salts from the system regularly with wateror with a high aqueous (e.g. 5% acetonitrile or methanol) mobilephase.In terms of mobile phase flow rate, it is common to work at

0.21.0ml/minwith standardbore columns. Inmost cases, theres-olution can be improved by decreasing the flow rate, albeit thethroughput is reduced.As in other chromatographic modes, sample loading (both

volume and mass) affects SEC chromatographic resolution andsensitivity [64,65]. In the case of column overloading, resolutionbetween analytes can be decreased. The ideal volume loads corre-spondto samplevolumes lower than15%of theaccessiblecolumnvolume. If the sample volume increases beyond this cut-off value,

the resolution may be reduced, and peak tailing may be observed.To conclude the discussion of SEC method development, it

remains partly empirical because chromatographic behavior andpeak shape are hardly predictable. Therefore, the optimization ofan SEC separation is sometimes time-consuming.

5. Detection modes in SEC

In SECseparationsof proteins,UV is still thepredominantdetec-tion mode [66,67]. UV wavelengths of 270, 275 and 280nm givespecific response for aromatic amino acids (tryptophan, tyrosine,phenylalanine) and are commonly used to assess protein aggre-gation [68]. At lower UV wavelengths (210, 214 or 220nm), the

amide peptide bond has a relatively high absorbance. Therefore,these lower UV wavelengths offer enhanced sensitivity but alsolower specificity. Finally, higherwavelengths provide a greater lin-ear dynamic range. Dual wavelength detection including both lowand high detection wavelengths can be useful for purity profilingin SEC [22,69]. The lower wavelength provides the sensitivity forthe low abundance species (aggregates), while the higher wave-lengthprovides a greater linear range for themajor species (nativemonomer).Using a fluorescence detector, both sensitivity and selectiv-

ity can be significantly improved [7073]. Fluorescence detectionenables themeasurement of lowlevelsof aggregates. Intrinsicpro-tein fluorescence is generallyobservedat anexcitationwavelengthof 280nm and up to 295nm. The emission is mostly due to exci-

tation of tryptophan residues, and scarcely attributed to tyrosine

and phenylalanine residues [74]. The maximum emission occursbetween 300 and 370nm, depending on the folding, conforma-tion andmobilephase properties. It is also important tonotice thatproteins can be covalently labeled with various fluorophores, thusproducingfluorescentprotein conjugates.Theemission fromtheseattachedtags iscalled extrinsicfluorescence. Taggingaproteinwithfluorescent labels is an important and valuable tool for studyingstructure andmicroenvironment [74].

Modern analytical SEC platforms are capable of more thanjust sample detection with a single concentration detector. Thecombination of multiple detectors including refractive index(RI), ultraviolet (UV), light scattering (LS) or multi-angle lightscattering (MALS), multi-angle laser light scattering (MALLS) and viscometer (IV) allows extensive characterization of proteinsamples [75]. RI and UV both allow accurate concentration mea-surements. Light scattering detectors allow the measurement ofthemolecularweightwithout theneed forcolumncalibration[76].Intrinsicviscosity isameasure of themolecular density andenablesstructural changes to be assessed. A combination of light scatter-ing and intrinsic viscosity allows the determination of the size ofthemolecules. This set of detection is called tetra-detection [75]. Atypical tetra-detector analytical SEC system is shown in Fig. 5.

MALS detectors are generally more sensitive to the high MW

aggregates and enable their confirmation and analysis.Mass spectrometry (MS) is another possibility for obtaining

molecular weight information. However, there are challenges tointerfacing SECwith MS due to the inherent incompatibility of themobilephases containing high concentrations of non-volatile salts[77]. SECmobile phases aretypicallynon-denaturingaqueoussolu-tions close to thephysiological pH range.Suchmobilephasesmadewith phosphate buffer lead to ion suppression and contamina-tion of themass spectrometer. Ammonium formate or ammoniumacetate at reasonable ionic strengths could be used as an alter-native to phosphate buffer, but physiological pH conditions maybe difficult to achieve because these buffers are adapted for acidicconditions (pKaof 3.75 and 4.75 for formic and acetic acid, respec-tively) [78]. Toovercome thisdifficulty, SEC-MSmethodshavebeen

developedusing denaturing mobilephases containing organic sol-vents and ion-pairing reagents instead of large amounts of nonvolatile salts [7982]. Acetonitrile and trifluoroacetic acid (TFA)in the mobile phase allows for successful SEC-MS experiments.Another possible method for avoiding the problems encounteredwith the mobile phase is touse off-lineMSdetection. As an exam-ple,Matrix-assistedLaserDesorption Ionization(MALDI) combinedwith Time of Flight (TOF) analyzer has been used in the past foroff-line SEC [83].

Comprehensive two dimensional LC (2D-LC) is most likely agood alternative to make SEC separations compatible with MSdetection and to improve the separation power. Using SEC in thefirst dimension and reversed-phase LCmode as thesecond dimen-sion would be ideal because this strategy provides a high peak

capacity, even for fast gradient separations, in addition to straight-forward compatibility with MS.

6. Possible pitfalls of modern SEC

Today,a wide variety ofporouspackingmaterialswith differentparticle sizes are available for the separation and characterizationof macromolecules [84]. The pore volume and the pore size dis-tribution determine the molar mass range that can be evaluated,whereas the particle size determines the diffusion processes ofproteins through the column [84]. Smaller particles lead to nar-rower peaks due to better column efficiency. In terms of efficiencyand achievable analysis time, porous materials with a small par-

ticle sizes seem to be the most favorable columnmaterial for SEC


7/13





Fig. 5. Schematic view of a multiple-detector SEC systemequippedwith ultraviolet UV/PDA), light scattering (LS), refractive index (RI) andviscometric detectors.(Adapted from Ref[73]).

[26]. However, with very fine particles (sub-2m particles), theseparation quality is improved at the cost of pressure. It has beenestablished in reversed phase liquid chromatography (RPLC) thathigh pressure (i.e.,>500bar)might cause importantfrictional heat-ing effects inside the column, and this heating is also likely truefor SEC [85]. Therefore, temperature sensitive proteins might suf-fer fromon-columnaggregation or denaturation [86]. In summary,reducing the particle size to the sub-2m range increases the riskof shear degradation [87].

In general, SEC separations are performed at low to moderatepressures (typically P


8/13





Fig. 6. Representative chromatograms on theeffect of column temperature (A) andpressure (B) on theobservedamountof antibody aggregates.(Reprinted fromRef[26]).

7.1. Application of SEC for the characterization ofmAbs

Recently, Latypov et al. showedan impressive high-throughputmethod that is able to differentiate candidate recombinant humanmonoclonal IgG1 and IgG2 antibodies based on their propensityto form aggregates when subjected to agitation (vortexing) stress[95]. IntactmAbswere separatedfromsoluble andinsoluble aggre-gatesusing size exclusion chromatography, undernon-denaturingconditions, and the individual components of the mixture wereidentified with a UHPLC-TOF-MS platform and quantified vs. anunstressed control. An internal standard was added to the mix-ture after stress, and used to correct for ionization differencesbetween samples. Treatmentof thesampleswith theIdeS (Fabrica-tor) enzyme significantly reduces sample complexity, and allowedfora large numberof candidatemolecules tobeassessed in a single

analysis. The IdeS enzyme generated F(ab)2 (fragment of antigenbinding) and Fc (fragment crystallizable) domains of the antibody.A volatileammoniumacetate saltwasused togenerateanMScom-patible mobile phase matching the pH formulation. The mobilephasewas 25mM ammoniumacetate and5% acetonitrile inwater,adjusted to pH 5.2. Separation was performed using an AcquityUPLC BEH200 SEC 1.7m4.6mm150mm column at ambienttemperature. A mixture of 8 intact mAbs at t= 0h, 1 h and 4 h of agitationand thefragments generatedby digestionwith IdeS,wereanalyzed at the same time points. The growth of HMW species didnot account for theoverall monomer loss because of the formationof insoluble aggregates [95]. Fig. 7 shows the corresponding SECchromatograms obtained at different agitation times.Another study reported an online fluorescent dye detection

method suitable for SEC and asymmetrical flow field flow frac-tionation (AF4) [96]. The noncovalent, extrinsic fluorescent dye(Bis-ANS) was added to the mobile phase or the sample, and thefluorescence emission at 488nm was recorded at an excitationwavelength of 385nm. By combining SEC and AF4 with online dyedetection, it was possible to simultaneously detect heat-inducedaggregation and structural changes of monomeric and aggregatedIgG [96].Wtzig et al. showed a 15min SEC separation of IgG1 antibody

aggregates using a conventional 30cm long column packed with5m particles and demonstrated the precision and repeatabilityof monomer and aggregate quantitation [97]. However, by usingstate-of-the-art column technologyandUHPLC instrumentation, itwas demonstrated that the analysis time can be reduced to less

than 3min [26].

A highlysensitivecapillarySECmethodologywasdevelopedfor

Fab aggregate analysis in human vitreous humor [98]. The capil-lary SEC method enabled picogram sensitivity with an RSD of lessthan 8% for the relative peak area of HMW of the Fab fragments.Fig. 8 shows representative chromatograms obtained with capil-lary SEC by injecting antibody Fab fragments. Another study also

Fig. 7. UV chromatograms at 280nm of a mixture of (A) 8 intact mAbs at t=0h(black), 1h (blue), and 4h (red) of agitation, and (B) the fragments generated bydigestion with Fabricator, at the same time points. The growth of HMW did notaccount for the overall monomer loss because of the formation of insoluble aggre-gates. (For interpretation of the references to color in this figure legend, thereaderis referred to thewebversion of thearticle.)



9/13


10/13





Fig.9. SECanalysis of thermallyaggregated interferonalpha-2bsample. Separationofinterferon alpha-2b (0.1mg/ml)containingmonomer (peak 1, 53%), dimer (peak2, 6%)andHMWaggregate (peak 3, 38%). Thesample wasincubatedin a water bathat70 C for30min. Fluorescencedetection was performed.


The isolated excipients were switched online, to a mixed-modecolumn for separation, and detected by an evaporative light scat-tering detector (ELSD). A volatile bufferwas used for both SEC andmixed-mode separation. This method facilitated the detection ofdifferentexcipientsby ELSDandprovidedpotentialfor online char-acterizationof theproteinbymass spectrometry(MS). Themethodwas finally applied to quantify protein and excipients in differentbiopharmaceutical products, including antibody drug conjugates(ADC) and vaccines [108].SEC methods coupled with intrinsic fluorescence detection

were developed for evaluating the stability and degradation pro-files of interferon alpha-2 drug substances and drug products [7].The method allowed baseline resolution of the active ingredi-ent from the excipients contained in the final product, including

large amounts of albumin. HMWaggregateswith apparentmolec-ular weight of650kDa, as well as dimers and denatured andreduced variants, were successfully identified and separated fromthe native proteins. This chromatographic method, which quan-titatively measured physical and chemical changes taking place insolution formulations,was found tobecapableofmonitoring inter-feronalpha-2stability[7]. Fig.9 shows theSECprofiles of thermallyaggregated interferon samples.

8. Some future trends in SEC

8.1. Interlaced andparallel interlaced SEC

Using two SEC columns in parallel can significantly increase

the analysis throughput. A further reduction of analysis time canbe obtained by combining interlaced sample injections with par-allelization of two narrow bore columns packed with sub-2mparticles. With this strategy, both lag times before and after thepeaks of interest can successfully be eliminated, resulting in anassay time below 2min [27].Farnanet al.[109]havedescribedthe methodologyof interlaced

SEC in detail. In summary, the methodology is basedon injecting anew sample before the ongoing analysis of a previous sample hasended.Thesubsequentinformationphasebeginsimmediatelyafterthesalt fraction of theprecedingsamplehaseluted. Fig. 10A showsthe schematic view of an interlaced SEC injection. The total timerequired for the analysis of n samples can be expressed as:

ttotal = tlag+ n (tinf+ thold + tsalt) (6)

Fig.10. Schematicview of interlaced (A)and parallel interlaced (B)SEC separation.


where ttotal isthetimerequiredforthewholeanalysis,tlagisthefirstpart of the elution window after the injection (without any com-pound of interest), tinfis the informationphase that corresponds tothe time window in which aggregates and monomer elute, tholdisthe third part of thechromatogrambetween themonomer andsaltpeaks and tsaltis the elution region of salt species (low molecularweight compounds eluting at t0).

A further increase in throughput can be achieved when apply-ing interlaced injections on two columns operating in parallel(Fig. 10.B). Compared to interlacedchromatography, theassaytimein parallel interlaced SEC is further reduced by thold. Two switch-ing valves are used to direct the flow alternately between the

autosampler, the two columns and the detector, thus enablingthe elimination oftlag, thold and tsalt. The use of two columns andswitching valves require two distinct programs on which pumps,autosampler and column compartment, including the switchingvalves, are controlled.

8.2. Sub-3m particles, ultrahigh-pressure SEC

Since the introduction of ultra-high-performance/pressure liq-uidchromatography(UHPLC) systemsandcolumns,whichenabledhigher resolution, sensitivity and peak capacity, there has been avariety of new analytical advancements in LC and LCMS basedon the use of low dispersion systems [110]. Because column peakdispersion and therefore the apparent efficiency depends on

the column volume, the achievable plate numbers and soluteretention factor (k), the importance of system dispersion is worthmentioning. In the SEC mode, there is no retention; thus, rela-tively low columnpeak variance is expected. However, achievableplate numbers with large molecules are relatively low and more-over the volume of 150mm4.6mm columns is reasonably large(1700l). Therefore, these two contributions compensate forthe low retention. Finally, the expected column peak variance isapproximately 200400l2. Using state-of-the-art UHPLC instru-ments (possessing extra-column peak variance between 2 and30l2), the system dispersion is negligible and does not impactthe column efficiency.

TheUHPLC technology wasoriginallydeveloped for RPLCappli-cations, but it is now also available for SEC operation due to the

availabilityof 1.7and2.5mBEHsilicaparticleswith varying pore


11/13





Fig. 11. A very fast SEC separation of panitumumab aggregates by using narrow-bore column packed with 1.7m particles. Peaks 1 and 2 are aggregated forms ofthe native antibody.


sizes. Today, columns for aqueous and non-aqueous SEC applica-tionswithporesizesof125900Aarecommerciallyavailable[110].Very fast separations of peptide, myoglobulin and insulin aggre-gates were demonstrated with 1.7m SEC columns [111]. Thesevery efficient columns were also applied for the characterizationof recombinant mAbs [26]. Fig. 11 presents a 4min separation ofpanitumumab aggregates.It appears that applying 1.7 and 2.5m particles in SEC opened

a new level ofSECseparations, but ithas tobe keptin mindthat onvery fineparticles, theseparation quality is improved at thecost ofpressure (and frictional temperature gradients). Therefore, thereis a risk of creating on-column aggregates when analyzing sensi-tive proteins under high pressure (i.e., >200bar) conditions [26]

as discussed in Section 6. The other disadvantage of sub-2m SECseparations is that currently there are only a very limited numberof commercially available stationary phases.

8.3. Capillary SEC

To improve the sensitivity of aggregate determination or han-dle very small amounts of available samples, the use of capillarycolumns in SEC appears to be a promising approach. However,several keymodifications to a commercially available liquid chro-matography systemare required to reduce thesystemvolumeandassociated extra-column band broadening, which could be criticalfor capillary SEC operation. Until now, the number of applicationsin this field is rather limited, but 300mm300m I.D. SEC capil-

lary columns were successfully applied for the separation of mAbsfragments [98,99].

9. Conclusion

To conclude, SECremains an importantstrategy for thedetailedcharacterization of protein aggregation. Indeed, due to theirpoten-tialimmunogenicity, theamountofaggregateshas tobethoroughlycontrolled during the production and storage of therapeutic pro-teins.During recent years, there have been a number of advances in

SEC that improve the quantity of information that can be gainedfrom a single injection. Among them, we can cite the use of shorter and narrower columns packed with smaller-sized parti-

cle to improve the throughput and resolution, but care should be

taken to avoid (or at least limit) the risk of shear degradation. Tofurther decrease analysis time, parallel interlaced SEC can also beimplemented.

Various detectorsmay bemultiplexed in SEC, including refrac-tive index (RI), ultraviolet (UV), multi-angle laser light scattering(MALLS) andviscometer (IV), forextensivecharacterization of pro-tein samples. It is also important to note that mass spectrometry(MS) plays a pivotal role in the structural elucidation; for this rea-son, some efforts were recentlymade to successfully couple SEC toMS by changing the nature of mobile phase constituents.

The sale of biologics grew at an incredible rate during thelast three years. The new drugs currently in preclinical and clin-ical phases are very difficult to characterize; for example, thecase of therapeutic mAbs or ADCs as well as bispecific antibodies[112114]. Thus,therewillbe anincreasingneedforhighlyefficientSECmethods. In this context, the use of a 2D-LCplatform includingSEC in the first dimension and RPLC in the second dimension maybeof interest to (i)further improve theresolving powerof thechro-matographic step and(ii) improve thecompatibilityofSECwithMSdue to the intermediate RPLC step.

References

[1] S. Fekete, J.L. Veuthey, D. Guillarme, New trends in reversed-phase liquidchromatographic separations of therapeutic peptides and proteins: theoryand applications, J. Pharm. Biomed. Anal. 69 (2012) 927.

[2] G.Walsh,Pharmaceuticals,biologicsand biopharmaceuticals,in: Pharmaceu-ticalBiotechnology: ConceptsandApplications,Wiley,WestSussex, England,2007, pp. 111.

[3] www.imshealth.com.(consulted in November 2011).[4] http://www.portfoliomanagementsolutions.com/the-organization-of-

pharmaceutical-rd/small-molecule-drugs-versus-biomolecular-drugs-biologics/ (consulted in November 2011).

[5] A. Beck,J. Reichert,Approvalof thefirst biosimilarantibodies inEurope,mAbs5 (2013) 621623.

[6] I.S.Krull, A. Rathore, T.E.Wheat, Current applications ofUHPLCin biotechnol-ogy. Part I. Peptide mapping andamino acid analysis, LCGC N. Am.29 (2011)838848.

[7] A. Diress, B. Lorbetskie, L. Larocque, X. Li, M. Alteen, R. Isbrucker, M. Girard,Study of aggregation, denaturation and reduction of interferon alpha-2

products by size-exclusion high performance liquid chromatography withfluorescence detection and biological assays, J. Chromatogr. A 1217 (2010)32973306.

[8] M. Bucciantini, E. Giannoni, F. Chiti, F. Baroni, L. Formigli, J. Zurdo,N. Taddei,G. Ramponi, C.M. Dobson, M. Stefani, Inherent toxicity of aggregates impliesa common mechanism for protein misfolding diseases, Nature 416 (2002)507511.

[9] W. Wang, Protein aggregation and its inhibition in biopharmaceutics, J.Pharm.289 (2005) 130.

[10] D.R. Booth,M. Sunde, V. Bellotti, C.V. Robinson,W.L. Hutchinson, P.E. Fraser,P.N. Hawkins, C.M. Dobson, S.E. Radford, C.C.F. Blake,M.B. Pepys, Instability,unfolding and aggregation of human lysozyme variants underlying amyloidfibrillogenesis, Nature 385 (1997) 787793.

[11] A. Beck, E.W. Rousset, D. Ayoub, A. VanDorsselaer,S.S. Cianferani, Character-ization of therapeutic antibodiesand related products, Anal.Chem. 85 (2013)715736.

[12] M. Vazquez-Rey, D.A. Lang, Aggregates in monoclonal antibodymanufactur-ingprocesses, Biotechnol. Bioeng. 108 (7)(2011)14941508.

[13] A.S. Rosenberg, Effects of protein aggregates: an immunologic perspective,

AAPS J. 8 (3) (2006) E501E507.[14] A. Cudd, T. Arvinte, R.E.G. Das, C. Chinni, A. Macintyre, Enhanced potency

of human calcitonin when fibrillation is avoided, J. Pharm. Sci. 84 (1995)717719.

[15] W.Wang,Instability, stabilization, and formulation of liquid protein pharma-ceuticals,J. Pharm.185 (1999) 129188.

[16] V. Sluzky, Z. Shahrokh, P. Stratton, G. Eberlein, Y.J. Wang, Chromatographicmethods forquantitation of native, denaturated andaggregatedbFGF in solu-tion formulations, Pharm.Res. 11 (1994) 485490.

[17] V.N. Uversky, Use of fast protein size-exclusion liquid chromatography tostudy theunfolding of proteins which denature through themolten globule,Biochemistry 32 (1993)1328813298.

[18] W.K. Hartmann, N. Saptharishi, X.Y. Yang, G. Mitra, G. Soman, Characteri-zation and analysis of thermal denaturation of antibodies by size exclusionhigh-performance liquid chromatography with quadruple detection, Anal.Biochem. 325 (2004) 227239.

[19] W. Fraunhofer, G. Winter, The use of asymmetrical flowfield-flow fraction-ation in pharmaceutics and biopharmaceutics, Eur. J. Pharm. Biopharm. 58(2004) 369383.
http://localhost/var/www/apps/conversion/tmp/scratch_6/dx.doi.org/10.1016/j.jpba.2014.04.011http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0010http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0010http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0010http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0010http://www.portfoliomanagementsolutions.com/the-organization-of-pharmaceutical-rd/small-molecule-drugs-versus-biomolecular-drugs-biologics/http://www.portfoliomanagementsolutions.com/the-organization-of-pharmaceutical-rd/small-molecule-drugs-versus-biomolecular-drugs-biologics/http://www.portfoliomanagementsolutions.com/the-organization-of-pharmaceutical-rd/small-molecule-drugs-versus-biomolecular-drugs-biologics/http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0025http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0025http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0025http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0025http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0045http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0045http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0045http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0045http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0060http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0060http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0060http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0060http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0065http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0065http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0065http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0075http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0075http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0075http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0075http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0095http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0095http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0095http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0095http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0095http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0095http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0095http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0095http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0095http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0095http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0095http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0095http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0095http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0095http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0095http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0095http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0095http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0095http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0095http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0095http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0095http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0095http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0095http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0095http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0095http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0095http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0095http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0095http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0095http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0090http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0085http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0080http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0075http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0075http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0075http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0075http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0075http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0075http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0075http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0075http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0075http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0075http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0075http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0075http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0075http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0075http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0075http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0075http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0075http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0070http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0065http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0065http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0065http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0065http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0065http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0065http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0065http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0065http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0065http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0065http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0065http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0065http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0065http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0065http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0065http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0065http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0065http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0060http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0060http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0060http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0060http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0060http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0060http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0060http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0060http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0060http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0060http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0060http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0060http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0060http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0060http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0060http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0060http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0060http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0060http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0060http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0055http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0050http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0045http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0045http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0045http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0045http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0045http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0045http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0045http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0045http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0045http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0045http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0045http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0045http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0045http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0045http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0045http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0040http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0035http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0030http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0025http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0025http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0025http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0025http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0025http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0025http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0025http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0025http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0025http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0025http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0025http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0025http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0025http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0025http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0025http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0025http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0025http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0025http://www.portfoliomanagementsolutions.com/the-organization-of-pharmaceutical-rd/small-molecule-drugs-versus-biomolecular-drugs-biologics/http://www.portfoliomanagementsolutions.com/the-organization-of-pharmaceutical-rd/small-molecule-drugs-versus-biomolecular-drugs-biologics/http://www.portfoliomanagementsolutions.com/the-organization-of-pharmaceutical-rd/small-molecule-drugs-versus-biomolecular-drugs-biologics/http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0010http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0010http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0010http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0010http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0010http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0010http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0010http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0010http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0010http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0010http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0010http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0010http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0010http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0010http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0010http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0010http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0010http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0010http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0010http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0010http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0010http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0010http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0005http://localhost/var/www/apps/conversion/tmp/scratch_6/dx.doi.org/10.1016/j.jpba.2014.04.011


12/13
http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0385http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0380http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0375http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0375http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0375http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0375http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0375http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0375http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0375http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0375http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0375http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0375http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0375http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0375http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0375http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0375http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0375http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0375http://refhub.elsevier.com/S0731-7085(14)00202-7/sbref0375http://refhub.elsevier.com/S0731-7085(14)00202-7

Best Paper on Size Exclusion Chromatography

Documents

Transcript of Best Paper on Size Exclusion Chromatography