Lrge Scle Process Development for Hydrophoic Interction Chromtogrphy, Prt 2: Controlling Process Vrition Pete Ggnon, 1 Eric Grund, 2 nd Torgny Lindäck 2 Astrct The scle-up of hydrophoic interction chromtogrphy (HIC) relies on users successfully nticipting the nture nd mgnitude of vritions tht ffect the finl process. Prt 2 of this 4-prt rticle exmines oth direct nd indirect sources of vrition in HIC processes nd suggests wys to control them. These concepts cn e used to improve scle-up nd on-line mnufcturing performnce of HIC nd other chromtogrphic methods. Introduction Scle-up filures, or filures of on-line mnufcturing processes, re occsionlly cused y externl circumstnces such s mechnicl or electricl filure. However, most prolems result from unchrcterized vritions in process chemistry, rw mterils, equipment, nd the mnufcturing environment. These prolems re generlly predictle nd controllle. Controlling vrition egins with identifying its underlying cuses. Process developers cn then ssess the mgnitude of vrition. Some sources cn e eliminted; others cn e controlled. Remining vritions cn usully e ccommodted y ltering process specifictions. Adopting this proctive strtegy significntly enhnces scle-up success nd helps to ensure dependility nd reproduciility of online mnufcturing This rticle originlly ppered in BioPhrm 8(4) 36 41 (1995) nd is dpted with permission. Reprints of this nd other BioPhrm rticles cn e ordered through the Advnstr Reprint Office y clling Mry Clrk t (541) 984 5226 1 Vlidted Biosystems, Inc., 5800 North Kol Rod, Suite 5127, Tucson, AZ USA 85750 2 Phrmci Biotech AB, S-751 82 Uppsl, Sweden processes. In ddition to supporting good process economics, this strtegy provides n invlule foundtion for process vlidtion. Mterils nd Methods As noted in Prt 1, we otined Source 15ETH, 15ISO, nd 15PHE hydrophoic interction prepcked columns nd ulk medi from Phrmci Biotech AB (Uppsl, Sweden) (1). All three medi re sed on 15-mm d p monodisperse spheres with pore-size distriution suitle for lrge proteins. The se mtrices re composed of poly(styrene divinylenzene) coted with hydrophilic polymer. We otined different monoclonl ntiodies from Becton Dickinson Immunocytometry Systems (Sn Jose, CA) for this study nd used them s process models. We purchsed uffers nd slts from Sigm Chemicl Compny (St. Louis, MO). All uffer components were Americn Chemicl Society (ACS) grde or etter. Process wter ws prepred using reverse osmosis nd deioniztion. We filtered uffers through 0.22-mm filter immeditely fter formultion nd ssigned five-dy expirtions. The experimentl methods re descried in the figure legends. Results nd Discussion Process-relted vritions. The most importnt fctors influencing the roustness of HIC scleup re gel selection, inding conditions, nd smple-ppliction methods. These issues were discussed in Prt 1 of this rticle (1). Temperture is nother process-relted source of vrition. For most proteins, retention on HIC medi increses with temperture (2 4). This ppers to e compound effect, resulting oth from protein conformtionl chnges nd from modifiction of the degree to which slts lter wter structure (2,5,6). As result, HIC processes re enerlly more sensitive to tem-
2 0 AU 280nm 3.0 0 AU 280nm 3.0 2.0 (min) 5.0 Figure 1. Mintining selectivity y incresing the inding slt concentrtion fter reducing temperture from () 23 C to () 4 C. Column: 30 mm x 6.4 mm Source 15ISO; uffer A (): 1.55 M mmonium sulfte, 0.1 M sodium phosphte (ph 7.0); uffer A (): 1.95 M mmonium sulfte, 0.1 M sodium phosphte (ph 7.0); uffer B: 0.1 M sodium phosphte (ph 7.0); flow rte: 940 cm/hour (5 ml/min); smple: 1 ml mouse IgG 1 scites. Conditions: equilirte column with 5 ml (5 column volumes) 80% A; lod smple y on-line dilution, 20% smple, 80% A; wsh with 2 ml 100% A; elute ntiody with 10-mL liner grdient to 100% B; strip column with 10 ml 100% B. Htched res indicte elution position of the MA. perture vritions thn other methods. Figure 1 shows tht mintining the selectivity of n ntiody frctiontion cross temperture reduction from 23 C to 4 C required incresing the mmonium sulfte concentrtion y pproximtely 25%. Other proteins my e ffected to greter or lesser extents nd should e evluted on cse-y-cse sis. The most common cuse of temperturerelted vritions in HIC processes is insufficient equilirtion of the smple. When using smples s smll s few microliters during erly process development, even cold smples will equilirte rpidly enough tht temperture effects re negligile. However, s process development proceeds into evlutions of cpcity nd other prmeters tht require lrger smple volumes, temperture effects re mplified in prllel. At full process scle, too-frequent cuse of process filure occurs when smple is tken directly from cold storge for room-temperture process (Figure 2). Filures of this sort pose the worst ostcle to susequent investigtions ecuse the evidence is lost s soon s the smple temperture reches equilirium. Avoiding this type of prolem is lrgely mtter of eduction for oth process developers nd mnufcturing stff. It is importnt tht temperture specifictions e included in mnufcturing stndrd operting procedures (SOPs). SOPs should emphsize tht ll process mterils uffers nd rw product must e t specified temperture efore eginning process. Minor temperture vritions etween development nd mnufcturing lortories re unlikely to cuse serious process devitions, ut they should not e dismissed. Consult fcility records to determine the rnge nd verge tempertures of mnufcturing res then develop processes to those specifictions. Vritions from rw mterils. Rw mterils for mnufcturing include rw product, uffer components, process uffers, nd gel medi. Routine vritions in these mterils seldom result in process filure, ut they do hve n impct, so it is importnt to chrcterize nd ccommodte them. Rw product. Typiclly, the rw product is the lest controlled component in purifiction process. Even under the est circumstnces, significnt vritions in product concentrtion nd the proportion of contminnts re likely. Very often, users conduct erly process development with mterils tht re inconsistent with mterils to e used in the finl production process. The worst possile sitution is when process developers use filed production lot such s one with micro-
3 0 AU 280nm 3.0 0 AU 280nm 3.0 2.0 (min) 5.0 0 AU 280nm 3.0 0 AU 280nm 3.0 2.0 (min) 6.0 Figure 2. HIC process filure cused y nonequilirtion of smple temperture. Shown re profiles generted () with oth the smple nd system t 23 C nd () with the smple t 4 C nd the system t 23 C. Other conditions were the sme s in Figure 1. Htched res indicte elution position of the ntiody. Figure 3. Process vrition cused y vrying the ntiody concentrtion of the cell culture superntnt. Shown re profiles generted with () 90 nd () 20µg/mL levels of ntiody. Buffer A: 1.5 M mmonium sulfte, 0.1 M sodium phosphte (ph 7.0); uffer B: 0.1 M sodium phosphte (ph 7.0); smple: 2 ml mouse IgG 2 cell culture superntnt. Other conditions were the sme s in Figure 1. Htched res indicte elution position of the ntiody. il contmintion for method development. Such medi often contin elevted levels of nucleotides, endotoxins, nd proteses. These contminnts cn foul chromtogrphy medi or cuse other interference prolems tht would not e encountered normlly, The product itself my even e ltered. Before process proceeds to scle-up, process developers should finlize the production SOP nd chrcterize multiple lots of product for routine vritions. It must e demonstrted tht the purifiction method functions dequtely t the extremes of the rnge (see Figure 3). If production medi representing these extremes re unville, process developers cn simulte them y spiking smples with prtilly purified product or with product-free growth medi. To otin n indiction of the mount of vrition for prticulr cell line in the sence of productionehvior dt, users cn consult dt from production histories of similr estlished products. Chromtogrphy medi. Users often tke for grnted tht commercil chromtogrphy medi offer identicl performnce chrcteristics from lot to lot. In fct, chromtogrphy medi re mnufctured to meet specified rnges not fixed vlues nd lot-to-lot differ-
4 ences cn hve significnt effects on process reproduciility. This is especilly true of HIC medi ecuse of the dependence of selectivity upon oth lignd density nd hydrophoicity. (7 11). It is lso true tht the QC tests used y gel mnufcturers cn never revel ll of the performnce chrcteristics relevnt to the rnge of poteintil user pplictions. For ny given chromtogrphy product, users should evlute medi from three or more lots to chrcterize vrition. Cpcity, selectivity, nd resolution re the key vriles. Process developers cn test mixtures of commercilly ville model proteins ut should lso include the product of interest in crude or purified form. The min point of this testing is to otin n estimte of mtrix vriility tht hs mening in the context of your specific requirements nd operting conditions (see Figure 4). Acceptle rnges of mtrix vrition re process dependent, nd some processes re more tolernt thn others. However, s generl guideline, lot-to-lot vrition greter thn 5% is cuse for concern. Rnges greter thn10% will lmost certinly result in significnt process vrition. Such vritions my require lot-specific method djustments to ensure dequte reproduciility. Conducting process development on one prticle size nd scling up to nother on the sme medium increses the likelihood of gelssocited process vrition. It is importnt to chrcterize the vrition of the different medi individully nd then compre rnges nd verges. If the differences etween prticle size rnges re of the sme pproximte mgnitude s the differences within ech medium, the prolems should e no greter thn those encountered when developing nd scling up using the sme medium. If the differences re significnt, users should exercise cution. Users should lso e wre tht medi with different mtrix nd lignd chemistries cn yield significntly different results for endotoxin, virus, nd nucleotide clernce, even if their protein frctiontion cpilities re similr. Buffer components. The qulity nd consistency of uffer components cn ffect the selectivity of HIC seprtions significntly. Users should e wre of hevy metl contmintion. Hevy metl inding y proteins generlly increses their hydrophoicity (12,13). 0.0 AU 280nm 0.02 2.0 time (min) 4.0 Figure 4. Reproduciility of selectivity using three lots of HIC medi. Columns: 30 mm 3 6.4 mm Source 15ISO, lot numers 225234/5, 27227/2, nd P34Ep(A)12PG7; uffer A: 1.5 M mmonium sulfte, 0.1 M sodium phosphte (ph 7.0); uffer B: 0.1 M sodium phosphte (ph 7.0); smple: purified mouse IgG 1 (0.5 mg/ml). Conditions: equilirte column with 5 ml (5 column volumes) 100% A; inject 20 µl of smple; wsh column with 2 ml 100% A; elute ntiody with 10-mL liner grdient to 100% B; strip column with 10 ml 100% B. Buffer slts tht contin high levels of hevy metls nd vry from lot to lot cn cuse process vrition. Not every seprtion will exhiit detectle sensitivity to this prmeter, so process developers should test for it with known high nd low metl controls. Lortory nd ACS grde slts cn serve this purpose. Using ethylenediminetetrcetic cid (EDTA) in process uffer formultions cn suppress hevy metl induced vriility, ut the est prctice is to purchse slts tht re controlled for metl content. Users should otin certifictes of nlysis for criticl uffer components s mtter of routine. Buffer preprtion. Differences in uffer formultion conventions e udully not prolem within either Development or Mnufct-uring, ut prolems occsionlly occur when processes re trnsferred from one to the other. Development stff must rememer tht formulting hundreds or thousnds of liters of uffer t time my impose constrints on process uffer formultion. The methods used y the mnufcturing stff should e downscled nd
5 pumps pump proportioning vlve mixer column mixer column Figure 5. Digrms illustrting () post- nd () prepump solvent-proportioning formts. The lck re in ech digrm represents the internl system volume from the point of solvent proportioning to the column. used consistently during process development. Development stff should lso use mnufcturing SOPs for uffer formultion, dhering explicitly to the sme chemicl formultions (counterion, hydrtion numer) nd uffer component grdes. Development uffer storge nd expirtion protocols should strictly dhere to Mnufcturing conventions. Development instrumenttion lnces, ph nd conductivity meters should e clirted nd mintined ccording to the progrms in plce for Mnufcturing equipment. Both deprtments should keep uffer logs tht note uffer ph nd conductivity vlues on lot-y-lot sis. Those logs serve s useful tools to distinguish routine vriility from loss of process control. Vritions from process equipment. Vritions due to differences in equipment etween Development nd Mnufct-uring often necessitte process djustments fter scle-up. Chrcterizing these differences in dvnce my not eliminte lst-minute refinements, ut it cn help void the lrge devitions tht send processes ck to development. Users should check the following primry equipment fetures: composition of wetted prts, mixer efficiency, ccurcy of solvent proportioning, nd system internl volumes. The composition of wetted prts is prticulr concern for two resons, oth of which cuse more prolems with HIC thn with other chromtogrphy methods. Becuse HIC routinely uses high slt concentrtions, the corrosion of stinless steel surfces nd the susequent leching of metl ions re perpetul concerns. The other prolem stems from the hydrophoicity of ruy check vlves nd spphire push rods in some lortory-scle equipment. In HIC methods in which smples re pplied through the pump, hydrophoic proteins nd lipids often foul the outer surfce of these components, which cn led to vritions in flow precision, nd in turn, to grdient errtions or clogs. Periodic clening with sodium hydroxide nd methnol helps to minimize this prolem, ut performing regulr dignostic procedures to detect flow normlities is importnt. Mixer efficiency is more importnt for HIC thn for most other methods ecuse of the high differentil viscosity etween high- nd low-slt solutions (14). If smples re loded using on-line dilution (see Prt 1), mixer efficiency ecomes especilly importnt ecuse poor mixing of smples nd inding uffer will prolong the durtion of the high slt protein interfce nd encourge precipittion of protein in the lines. Mixer efficiency must e t lest s good on mnufcturing chromtogrphs s on process development systems. Vritions in solvent-proportioning ccurcy etween chromtogrphs re frequent sources of scle-up prolems. When methods re developed nd scled up on postpump solventproportioning systems usully no prolems occur; however, methods tht re developed nd scled up on prepump solvent-proportioning systems or trnsferred from one formt to nother frequently require process corrections. Figure 5 is digrmmtic comprison of prend postpump solvent-proportioning systems. The internl fluid volume etween the point of solvent proportioning nd the column is the key vrint. With postpump proportioning, this volume is typiclly smll, ut it cn e very high with prepump proportioning. Figure 6 shows vritions in grdient ccurcy using pre- nd postpump solvent-propor-
6 0 % uffer A 100 0 % uffer A 100 c d Figure 6. Comprison of grdient ccurcy for (,c) post- nd (,d) prepump solvent-proportioning systems. Column volumes: (,) 1 ml nd (c,d) 25 ml; totl run volumes: (,) 30 ml nd (c,d) 750 ml. The dshed nd solid lines represent the ctul nd progrmmed grdients, respectively. tioning systems t different process scles. We generted these profiles y progrmming grdients etween wter (uffer A) nd 1% cetone in wter (uffer B) nd compred the opticl profiles t 280 nm with the progrmmed grdients. We corrected the position of the opticl grdients reltive to the progrmmed grdients for the volume etween the column nd the monitor on oth systems. Actul profiles therey reflect the grdient t the column. We used the sme proportionl grdient configurtion for ech profile: five column volumes 100% A; five column volumes 65% A; 10 column volume liner grdient to 100% B; nd 10 column volumes 100% B. Note tht conformnce of the grdient to progrmmed specifictions improves with incresing reltive process volume on oth solvent-proportioning formts, especilly on the prepump formt. When the column volume is smll in the prepump system, ctul grdient ccurcy (compred with the grdient progrm) is extremely poor. Figure 6 shows time lg nd gross errtion of grdient shpe. Both re functions of the rtio of system internl volume to process volume. Reproduciility of the lg nd the errtion re excellent when this rtio is kept constnt. Prolems rise when the rtio chnges (for exmple, during scle-up). If product is eluted ner one of the set points in liner grdient, scled-up process will likely vry significntly from specifictions. The effects my rnge from simple shift of the product s elution position to chnge in reltive purity. If process employs nrrow-intervl step grdient, scle-up is likely to fil outright. The rtio of internl system volume to column volume cn lso ffect the efficiency of smple equilirtion using on-line dilution. Figure 7 contrsts the clernce profile of protein solution introduced through the proportioning vlve in prepump solvent-proportioning system with the profile of solution introduced through dedicted line in postpump proportioning system. Clernce time is importnt ecuse it prllels the precolumn residence time during which proteins loded through pump will e exposed to inding concentrtions of slt. The longer the precolumn residence time, the higher the risk of precolumn ggregtion or precipittion. As Figure
7 7 shows, n pproximtely 7-fold differentil etween the two systems precolumn residence times occurs t 5 ml/min. To void these prolems, minimize the rtio of system internl volume to column volume nd keep tht rtio constnt cross process scles, or s close to it s possile. If its not possile to mintin constnt rtio, the next est solution is to chrcterize the rtios. Knowing wht to expect my mke it possile to develop generic rules of thum for pplying process djustments etween nonmtching systems. Developing sfety mrgins. In spite of the fct tht different sources of vrition ct on HIC methods y different mechnisms, most vritions hve the net effect of either incresing or reducing product retention. This fct provides simplistic ut prcticl sis for modeling the effects of vritions nd determining pproprite sfety mrgins. Regrdless of their source or mechnisms of ction, most vritions cn e stndrdized y expressing their effects in terms of the chnge of slt molrity required to crete similr vrition. Users cn determine this reltionship y mesuring the left- or right-hnd shift of the product pek in liner-grdient elution profile. For exmple, decrese of 5 C my produce pproximtely the sme effect s reducing the mmonium sulfte concentrtion y 0.075 M (Figure 1). If the sources of vrition hve een identified nd chrcterized, this mesurement provides resonle method for expressing the cumultive vrition tht cn ffect process. Tle 1 shows how this informtion cn e used to estimte process extremes, ginst which the process cn e modeled efore scle-up. The mgnitude of the cumultive vrition from center in ech direction represents the minimum sfety mrgins tht must e dded to insulte the process from this vrition. Figure 8 compres set of profiles tht lck dequte sfety mrgins with set of profiles from protected process. Liner grdients re especilly effective for uffering process vrition ecuse sfety mrgins cn e dded simply y extending the set points. The originl grdient selectivity cn e preserved y proportiontely incresing grdient volume. The worst errtion tht cn occur under ctul run conditions will e vrition in grdient 0 time (min 6 Figure 7. Precolumn residence time profiles for post- (solid line) nd prepump (dshed line) solvent-proportioning systems. The postpump system profile ws generted y introducing pulse of out 100 ml of 1% cetone through pump B t 1 ml/min. Flow ws stopped riefly while the system ws progrmmed to 100% pump A (wter). Flow ws strted t 5 ml/min, nd the signl ws llowed to return to seline. The prepump system profile ws generted using the sme solvents. The pump B line ws first purged to ring the cetone solution up to the proportioning vlve. The system ws switched to 100% A (wter) nd rinsed until the signl ws restored to seline. A pulse of pproximtely 100 ml of 1% cetone ws introduced through line B t 1 ml/min. Flow ws stopped riefly, the system ws switched to 100% A, flow ws strted t 5 ml/min, nd the signl ws llowed to return to seline. slope. Assuming tht the product is eluted ner the middle of the grdient, the reltionship etween the product nd the contminnts will remin fundmentlly unchnged. Step-grdient processes cn e prtilly insulted from process vrition, ut dding sfety mrgins involves compromise: roder step intervls increse surnce tht the product will elute fully within the oundries of the pproprite step, ut the price my e lower purity. Conclusions The scle-up processes tht provide the est results nd the mnufcturing processes with the highest consistency of conformnce to specifictions re the ones in which process developers hve fully nticipted the rnge of
8 Tle 1. Estimting cumultive process vrition Vrile Rnge Men Process equivlent Process temperture 21 25 C 23 C -0.015 M +0.015 Smple temperture 18 25 C 21 C -0.03 M +0.03 M Buffer ph 6.8 7.2 7.0 Binding slt molrity 1.45 1.55 1.50-0.05 M +0.05 M Grdient precision (%) 62.0-0.03 M +0.03 M Gel vrition (%) 62.5-0.04 M +0.04 M Totl -0.165 M +0.165 Expressed s the chnge in mmonium sulfte concentrtion required to produce comprle effect. c d e f Figure 8. HIC profiles generted y unprotected (top) nd protected (ottom) processes. Column: 30 mm 3 6.4 mm Source 15ISO. Buffer A: () 1.35 M mmonium sulfte, 0.1 M sodium phosphte (ph 7.0); () 1.50 M mmonium sulfte, 0.1 M sodium phosphte (ph 7.0); (c) 1.65 M mmonium sulfte, 0.1 M sodium phosphte (ph 7.0). Buffer B: 0.1 M sodium phosphte (ph 7.0). Flow rte: 940 cm/h (5 ml/min); smple: 1 ml mouse IgG 1 scites; detection: UV sornce t 280 nm (3.0 AUFS). Unprotected conditions: Equilirte column with 5 ml 65% A; lod smple y on-line dilution: 35% smple, 65% A; wsh with 2 ml 65% A; elute ntiody with 5-mL liner grdient to 27% B; strip column with 10 ml 100% B; totl run time: 5 min. Protected conditions: Equilirte column with 5 ml 75% A; lod smple y on-line dilution: 25% smple, 75% A; wsh with 2 ml 75% A; elute ntiody with 10-mL liner grdient vritions the process will fce. Some sources of vrition re unique to given method, such s the sensitivity of HIC to temperture vritions. Others re more generic, ut ll sources of vrition cn e identified nd evluted. Some cn e eliminted or reduced. The rest cn e modeled to define sfety mrgins tht will protect the purifiction process.
9 This pproch expnds the role of process development to include tsks such s chrcterizing lot-to-lot vriility of chromtogrphy medi. It lso constrins process development to following mnufcturing SOPs nd conventions for uffer preprtion. But ultimtely, it provides the gretest ssurnce tht scle-up nd mnufcturing will perform predictly nd consistently. References (1) P. Ggnon, E. Grund, nd T. Lindäck, BioPhrm 8 (3), 21 27 (1995). (2) S.-L. Wu, A. Figuero, nd B. Krger, J. Chromtogr. 371, 3 (1986). (3) S. Goheen nd S. Englehorn, J. Chromtogr. 317, 55 (1984). (4) R. Ingrhm, S. Lu, A. Tnej, nd R. Hodges, J. Chromtogr. 327, 77 (1985). (5) F. Regnier, Science 238, 319 (1987). (6) M. Dixon nd E. We, Adv. Protein Chem. 16, 197 (1961). (7) S. Hjertén, K. Yo, K.-O. Eriksson, nd B. Johnsson, J. Chromtogr. 359, 99 (1986). (8) S. Hjertén, Biochim. Biophys. Act 412, 51(1975). (9) J. Rosengren, S. Påhlmn, nd S. Hjertén, J. Chromtogr. 101, 281 (1974). (10) D. Gooding, M. Schmuck, nd K. Gooding, J. Chromtogr. 107, 114 (1984). (11) D. Nu, Biochromtogrphy 5, 62 (1990). (12) T. Arkw nd S. Timsheff, Biochemistry 21, 6545 (1982). (13) T. Arkw nd S. Timsheff, Biochemistry 23, 5912 (1984). (14) R. Chicz nd F. Regnier, Met. Enzymol. 182, 392 421 (1990). This rticle ws downloded from Vlidted Biosystems Qurterly Resource Guide for Downstrem processing. The entire Newsletter cn e ccessed free on the internet. Vlidted Biosystems, Inc. 5800 North Kol Rod, Suite 5127 Tucson, AZ USA 85750-0912 e-mil: info@vlidted.com http://www.vlidted.com phone: (520) 529-1095 fx: (520) 529-1021