./. Mol. Hiol. (1992) 226. 889-896
Selection of Phage Antibodies Mimicking
by Binding Affinity
Affinity Maturation
Robert E. Hawkins’?, Stephen J. Russell’ and Greg Winter’*2t ‘M RC Laboratory of Molecular Biology, and 2MRC Centre for Protein, Engineering Hills Rocxd. Cam,hridgr (‘RZ 2&H. (‘.h’. (Received 21 .Jcwwary
1.992; accepted 10 April
1992)
We describe a process, based on display of antibodies on the surface of filamentous bacteriophage, for selecting antibodies either by their affinity for antigen or by their kinetics of dissociation (off-rate) from ant,igen. For affinity selection, phage are mixed with small amounts of soluble biotinylated antigen (< 1 pg) such t’hat t’he antigen is in excess over phage but with the concentration of antigen lower t,han the dissociation constant (Kd) of t,he antibody. Those phage bound to antigen are then selected using streptavidin-coated paramagnetic beads. The process can dist,inguish between antibodies with closelv related affinities. For off-rate selection. antibodies are preloaded with biotinylated antigen and diluted into excess unlabelled ant,igen for variable times prior to capture on streptaridin coated paramagnetic beads. To mimic the afinit’y maturation process of the immune system. we introduced random mutations int,o the a&body genes in vitro using an errorprone polymerase, and used affinity select)ion to iso1at.e mutants with improved afhnitv. Start,ing with a small library (40.000 clones) of mutants (average 1.7 base changes per Vu gene) of thrb mouse antibody B1.8, a,nd using several rounds of affinity selection. we isolatcad a mut)ant wit’h a fourfold improved affinity to the hapten 4-hydroxy-Fi-iodo-3-nitrophrnac~etyl-( NH’-caproic acid (mut,ant h’, = 9.4( 5 0.3) nM compared wit,h B1.8 h’, = 419( k 1%) n;M). The relat’ive increase in affinity of the mutant is comparable to the increase seen in the anti-4-hydroxy-3 nitrophenylacetyl/NIP-caproic acid murine secondary immune response. h’rywords: filamrnt,ous phagt,: antibody: affinity: maturation
1. Introduction The display of antibody fragments on the surface of filamentous phage that encode the antibody genes (MctJafferty et aZ., 1990: Hoogenboom et al.. 1991: Kang et nl.; 1991; Kreitling et al., 1991; Barbas et al.. 1991), and the selection of phage with antigen binding a&vi&s by panning, offer a powerful way of making antibodres from V-gene libraries from immunized animals or humans (Clackson et al.. 1991; Burt)on it al., 1991). Very rare phage can be isolat,ed. because in contrast to filter screening of soluble antibodies (Huse of al., 1989), the number of phage t,hat can be screened is limited only by the size of the init,ial library. Using the tnRNA of spleen cells taken from hyperimmunized animals (Huse et al., 19X9), or the 1)NA of antigen-selected memory cells, the V-genes encoding antigen binding actrvrties are greatly t Authors to lvhom all correspondence should be addrt~ssrd
enriched (Hawkins & Winter. 1991). However, in small random combinatorial libraries derived from a large number of heavy and light chain V-gene elrment)s (Huse rt ul.. 1989), t,he original pairings between ant’ibody heavy and light, clhain of the B-cell are likely to be lost (Winter & Milstein, 1991). Indeed, pairings from such libraries have been shown to be promiscuous. For example. different light chains were found associated with a single heavy chain. and vice uer~u (Clackson et rcl.. 1991). Which (if any) of these combinations correspond to the original pairing in the B-cell is not clear. Since the original B-cell pairings are selected by the immune system according to their affinity of binding or their kinetics of binding t,o antigen (Foote bz Milstein, 1991), many of the pairings in a combinat’orial library are likely to have a lower affinity. However. in a very large library (> 10’ members), some of the original pairings will be present (Winter & Milstein. 1991). and a few of the new pairings may even have an improved affinity. Tt
triutatil random It is also drsirahle to selec3 the highest. &nit! ittltji hod& from “secondary” phagr ;~tltiho(l~ lihrwt,ies iti \vhich t.hch selrct,rd V-gene pairings from it primary library art’ further divrrsified, for c~xatnple. hy keeping one chain fixed and shuffling the other ((‘lac*ksotr f,t rrl.. I!I!II ). Kew pairings can he c*rc~ated, k~adittp to atttihotli~s with irrrprovc4 affinities (Marks rt cl!.. 1992). Altt~rrratjivrly, random F)oint mutation ~ultl hc used. as in th(t immune system (Milstjein, 1990; Winter Kr Milstein. 1991). Tnd4. the isolation of human antihodirs wit bout imtntttrizafion. using a single large anti (Ii\-erse \7-get~c library displayed on phage (Marks it n/.. 19!)1 )- and t,htbir improvemc~nt should canable evt’n t h(> highest affinity antibodies to he tnadr t~titircly itt rifrft (\l:inter $ LMilstIeirt. 1991). Howrvpr, it’ is difficult to select hrtwren phages acc*ordirtg to their attinities. especially het,wclcbtt t,hose with similar afinities. Phages hearing high affinit.y antibodies caan he selecGd from those with low affinities using antigen immobilized on solid phase. but finr affinity disc>rimination is dificult t)tac.aruse the binding is dictated hy both the afIinit>. and the avidity of the phage (Cllackson et al., 1991 ). Alternatively. phage antibodies could he selected by hitiding to biotinylated antigen in solution, followed by csapture of the phagt on a strept~tvidin-~oat~,(l as shown for phage displaying small surface. peptides (Parmlry & Smith. 1988). The high affinity, phagr antibodies could then hr enricahed hy compe’ tit ion for limit,ing amounts of antigen, as prnposPt1 for peptide phagc (Scott & Smith. 1990). But for affnity maturation of phage antibodies, the mart! low affinity (parent) phage could caornpete ant tht, hindiny of rare high aftinity variant’s, \.VP have thereforr adapted t hc hiotin-c+aptttrr tnrthod (Parmlry & Smith. 19t18) for the affinit? rnat.urat,ion of phage antibodies. Inst.ead of limiting antigen caoncentration such that competitiott for thtb antigen occurs (Scott & Smith. 1990), WP used the antigen in excess over phage to drive t,he binding. To favour the binding of the higher afinity phapcl ant,ihodies. we srt thr concentration of antigen (Ag] lower than t)he desired dissocGtion chonstant (h’,). The phage antibodies in complex wit,h biotinylated ent,igrn were then captured on streptavidin-c:oatetl paramagnet.ic heads. The discrimination M wept1 different ant,ibody mutants is related to the ratio of their affinities. This process has also been adaptcd to select for prolonged off-rat,e: after binding of biot~inylated antigen the phage are diluted into ex~ss unlahelled a,ntigen. Riotinylated antigen then dissociates frctm t,he phagt: at, a rate proportional to koff and the exposed antibody heads are hound b; unlahelled antigen. The longer this off-reaction IS allowed to proceed before capture, the greater the selection for prolonged off-rate. We have used the method of affinity selection to achieve affinity maturation of’ a primary response antibody (anti-NP/NTPt hpbridoma) by selecting a
\I.il ti irnprovt‘d
affinity
2. Materials Bovittr
trottt
ii i~i,t,;~t? rot
ttrirl writ s.
and Methods
(il)
.-Inti!qrtt
set’ttm albumin
(BSA:
Kot~hrirt#rt,~ILl;tttttltt,ittt
(;rrma,ny)
was c,onjugated with ,h;TP-c~aproat~~-(~~stt(.(,ittt rnitlr ((‘ambridge Research Kiochrmicals. SorthM ic,lr. I!.K.) as tlc~scrihrd l)y Browttstonr c/ rrl. (I!Mi). Tht. act.ivatrtl haptett was tlissolvecl in tlittrt~th~lf;~rtrtar~~itl~~ attd added rithtbr in a molar ratio of IO: I 01’ 0.7: 1 STP-c~a~~roatr-l~-su~~~~inimitlr;I~S1\ (;“;W,“HSA and in 0.2 hi-soclirrtrt hytlrog:c~rr Nl P0.71;-‘-EDT.!. In t.he c~xllerimrttts using motlt~l antihodirs the s&ction methods were trstrd to examine the rffecats of antigen ~o~ic~rntration. “double” rounds of selection. and time after floodittg ott the number of phagr bound. For this purpoar the srlr?tiotts were performed in clif%retit tubes ~~rowsst~i itt parallrl and the titrrs of phagt. c>lrrted \vt’re c~c~triliai~f~~l.
The phage sc~Fvl)l.3 and huscFvl)l.3 wer(’ incaubated srlbarately with a low conclentration of Iliot.in>~lated HEI, (IO n>f) at room t~rmprrature for I h in KYHS. The off rrac,tion was started by diluting 50 /LI pctrtions of this binding reaction into ex(bess unlabeilrd H El, (950 ~1 of a 20 pt stocak in tSPl%). and leaving for variable times. I’hage rrtaitting hound biotinylat,rd H E:l. were then paracapturt~d by adding 6 x IO’ strrftta~iditt-c,oatrd tnagnrtir beads and incubating for I ntin with c.otrtinuous mixing. The heads were washed 7 times with BPIIS. I’hag~e LStar? elutrd and infrc.tetl as desc,ribrtl above.
The binding of the sc*FvKl.8 phagt> to the 9ll’,,I~SA (.ott,jugatta for EI,ISA was ltt:rf~~rmrtl as tfesc-ribed by &l(.(‘afferty ef nl. (1990). and thr expression and binding of soluble sc~F~Bl.8 for ELlSA as tl). Hawkins Xr \Vinter (1!)!)2). Sequencing template was prepared from phagr as dt~s~ribed by (‘lac*kson rf al. (1991 ) and srqtrc~ncrd (Sanget uf ml.. 1977) \vith phagr T7 l)iYA l)olyttteras~~ (Srquenasr. I’sIS. (‘leveland. I!.S.A.). The expression of t,hr soluble sc,Fv KI .X f’ragtnrnts was intlucrd using 1 1 of culture grown in shaker flasks (see Hawkins & M’intrr. 1992). Thtb bac+ria were pelleted. taken up in 2.5 tnl of PBS. I mw-F:l)T?r and the fragments harvested from the periplastn (Skrrra rf ol. l!Nl) 3 h after itttluciion. Tht, frac’tion was applircl to it I ml XT’Spharosr column. washed with IO ml of’ PISS and elut,ed with .Aml of’ OlwHCl, glycinr (pH 2.5). Thti rluted scFv w;l;, c.ollf~c~t?c!in I ml fractions into 0.5 ml of ic,e-(*old I at-Tris.H(‘I (ltH 7.4) and then the pcbak 1)rotcsin fractions dialysrd into I’M. hfYinit:- n~t~asurrmrttts wart’ lirrformed by fluort~s~rnc~c~quench tttration usinlr tht> hapten XII’(‘.\P-OH as drsc>ribrrl (.Jonrs rt CI!.. IHXR). ‘I’h(, anti-lysozyme scFv fragmrttts \vvr(’ purified as dt~~cribed by ivard et ~1. (1!%9) and affinities tnrasurrd by flttortxs(.enc*r qut.n(lh titratiott (Foote R- \2’inttLr, 19!)2).
3. Results (a)
S&ction
hy czfinity
or o$rlrtp
Flnorescrrtc~~ yuenc+ titration tneasurements of dissociation caonstants. for I)roleir~-t~ritiliotlv inter-
~.-___-
___
_--Table 1
actions, give values that, are wnsistent) wit,h ~Jswtdoequilibrium relaxation methods hut tend to give high standard errors (Marks rt al.. 1991: Foot,e 8 Winter, 1992). We thus determined the K, value fot scFvD1.3 as 15 (+X) n>t and sc:FvDl .3 (TyrlOl --+Phe) as 250 (+_80) IIM. This shows t’hat the mutant scFv fragment has a dissociatiott wnstant approximat,ely 17.fold higher t hatt t hr parent scFvDl.3. The HuLysl 1 rwtta,prd antihod! h as been shown to have a dissociation constant fourfold higher than the parent Dl.3 antihotly (Foote & Winter. 1992), and presumably the differfwce is retained in the scFv fragment,s. i‘he ratios of the dissociation constants fiJ1 wFvl)l .3. huwFvDl.3, scFvl)l .3 (Tyr101 +Phr) are thrwfore HkJOUt
1 : 4 : 17.
The
titration of scFvI)l.3 and sc:Fvl)l.3 -+Phe) phage with hiotinylated HF: L (‘l’abtc 1) shows that’ t,he phape con ~JP wlrc~ted acwrding to their afinities, rspwiwll,v at t,hr Iowt‘r c.otrc.f,ntratiort of H It: I,. as expected frotn our tnodcl (see hrlow). Here. t.hr 17.fold higher affinity of the wl’v I)1 .3 over the wFvDl.3 (TyrlOl -+F’he) heads allows it 14fold greater pi&~ of the former phagc’ f’rom the heads. At high conc.rntratiorts of HEI. (100 rt>T) the titre of’ phage c~aptured falls. pwsunnhly due to c:otnpetitioti hJ: exc’f’ss free antigetr (not (TyrlOl
Table
3
selection. Thtb swottd round ap~~~ws to IF tttow efficient, than the first, as shown hy t,hrk rlutiott of wFvD1.3 and huscFvl)l.3 phagc (Table 2): the yield is 0.3?, from t,he first round. and 1.75”,, from the second. Furthermore. the fourfold higher af& tlit,y of the wFvI)l.3 over the hus~Fvl)l.3 givrs a 2.5.fold greatet yield of the former phagc frOll1 the tuo-foltl 1~~x1s in the titst round. and a furthw factor in the swond round. The fraction of’ l)hagtl c~apturrd in the ahsenw of hiot,inylated antigen was very low in both rounds (5 x IO- ’ ()(, of input phap in round 1 and 04)3(~,, in round 2). ‘I’hC af%tritirs Of ttlr I)1 .:j itIltiiJOt1~ iJ.tld thfl hlttTl& nized I)1 .S itrttihod>~ (Hul,ysl 1) diffw k)x, ti~ttrfold. ;Lrld this is largely attrihtttahlr to ttlrl dtf?errtrc~c~ in off-rates (Foote & Il’inter. 1992). For off-r:tt,r sele(. tion we used the phagr displaying scFv frwgrrtt~nts &rived f’rotn t,hwr antihodirs. ‘l’nl~lrx 3 hhows that ()ti-rat,(~ selection u it h hiofitlylate(1 H EI, pvt. I~IOW t ha,rt 8. ZO-fold gWittf>t’ yield of scaFVI) I .3 ])hag,lt,0vf.r huwl+l)l .3 lJhagt> af’t)tsr six minrrtt~s. Howt~v~~r. 1~). I:! mitiutw. ttrc* t1iffVrc~nc.c~iti ?,ielcl IlitCl fallow to 7..5ffjtd.
?J,S Itl?
Of
tlllttll~f~l~
Yll7t~Vl
[)tl2Lgt’
f’iYJtIt t ttC’ I)t’ittlS
approwhwl its t~ac~kgrottn~l \.;tlut~ (al)pwttt ai 60 rnin). lndwtl. hackground hitttlitry is a rnaljor I’A(*tc>t~ lirnitittg Itic, f)o\vtlr of ofi-ratcb st~lw~tion.
stlon~tl). Two cwnsewt~ivc~ rounds of wlectiott c-art tw used (\vit bout, growth of thtb phage in hact.eria hrt~wren ra,cdtt round) IJ~ eluting the phage from the heads with t.riethylamine. neutralizing t hc eluat,e and suhjwt’ing the phap to w second round of affrtit?.
11 clorirs wt’w sec~uc~ricwl. Thtk ovwall ttiutatioti freqitt~nc~y \vas 4 .“]kh (1.7 pet’ \‘,, getw). XII t,ypw ot tnutat~iott \v(Lt’c‘ stbert t~xwp~ (i/t ’ or (‘,‘(i c*hartgw I~ut tht~rr wits it l)iilS iri favorer of (‘I‘,‘(‘..&/(:) tnutations ilnd (A4:rr.7;.4j triut;rtiotts (‘r:(-. A(: (9): ‘r#A. ,\;‘I’
(5): r/c;. .I:(’ (2): (~3’. (::‘:i (3): ((:.-I. (::7’ (3); (7:. (i/t’
Tit ws of wFvl)l.:3 ptl”g”
(0)).
‘I’tlf~
tlltttl~jrt~
grrw
(arid
J- [~f’t’
1’”
gt~trw
t~rlrriittirtrtl.
Of’ tIltttittiOtlS with
no
“hot
\arif~cl SfJols”)
f’tY)tll \vit
0 h
t0
I): I4
f)tlii#‘S wwf~ sf~ltY*trfl \cit’ti X tint-hiot ittylatwl NTPo.,l~SA (rounds 1 and 2). and gr;ldually rt~tluwd to 1 1111 STI’,.,KSA OVPI sitcwssivc~ routtds. Incarrasittg ttum brrs of phaps \verc c>lutetl afttlr f’iL(‘h round ant1 t~tw tiutnhrrs rxptdfd brtwwtr fetch wlwt iott I)\. rcbittftLc*lion and growth (Jlatc~rials and .‘sitiglc round”). I’hagts itil)ttt \VitS always .\lethods. The
3x x0 x O~IT,X 1.4x lHl2.i x
10’0 106 106 ttlh IOh
IO0
pl
2 x
1o’O
of
a
100
fJtlil#‘);
i\t taac*h round
X IJh+y j .?\g]
fYJtlf?tltt’iltt~ ilrltl
\v(‘re: rottttcl
tllltltlJ~t5
(a~J~Jroxirllat~rl? Of
[JhLLgt’
f’llltfY1
I d rt>l. 0.7 x IO’: round
In Vitro
Antibody
Table 4 acid dijjewnces and afJinity measurements mutant Bl .X antibodiw isolated from library
Amino
of
Aginity
893
Maturation
t,itration and shown t,o range from 48 nM t)o 9 nM (Table 4). 4. Discussion
Number of active sites per scFv fragment
34 63 8% !)3 104 106 111 Kd$ (nM) (‘lOW IS1.X Ies, multivalent binding of phage to the para magnetic brads will help prevent dissociation during thtb washing steps. However, mult,ivdent, hintliny (‘an arise directly or indirectly. Direct multivalent capt,ure c~mld arisr when at, least t)wo heads arc simult~aneously complexed with hintin?-lated antigen in solution. Assuming three ant,ihody hratls per phage (Gray rt WI.. 1981; (ilaser-Wutt)ke it nl.. 1989) and a proport~ion (p) of the heads t~onnti to antigrn. then thtl fradon of phage with two or more heads bound is: p3 + 3p2( 1 -p). a,nd the frac*t)ion with one head bound (rnonovalent~) is: 3p( I -P)~. When [ Ag] is signific~ant,ly lower than h’,. t.htL majority of phage will he monovalent, retluc*ing the likelihood of tlirrct multivalent (*Rptllrt~. Howrvrr. rnonovalent~ Capt,ure cY)uttl hc cY)rivert~ed indirrdly to multivalent (sapture if thr phag~~ I~ountl to (~x(*ess antigen on the head surface with their other heads. Finally the antigrn-c:oatetl heads c~mltl capt,ure low affinity phage that have not hound antigen in solution. In light of the different possihlc modes of phagr capture. it is dific~ult to caal(*ulat.r t,hr power of st+c:t ion. Howevflr, for capturr of monovalent phages only. we would expect a capture ratio (phayr I : phage 2) of (K, + IAgJ)/(K, + 1AR]) x 1pAhI ]i 1pAt,%], whereas for dired capture of bivalrnt phagc only. the srlrc+ion is rnorr powerfill. with iin (*apt urp of approximat,rl> t*xprc+ecl ratin (K2 +lAgj)2,‘(KI + [A4g])‘x [pAbl]/[pAbZ], hut at t trc, tlxpense of frwrr phage csapturtd. Whichevtar t)ro(*(bss dominates. tht, power of the selection will I)tl reduced by unwant’rtl capture of low aflinity phagr. Nrvertheless the selrrtion of phagrs at differt~nt c,onc,t~ntrat’ions of antigen (Table I). with [HEI,] < K, of ttot’h antihotlies. shows that when thr affin-t itas of each head for antigen differ 1)~. 1 ‘i-fold. t hr ?;itJlds 01’ ptiage hiding to bea,& (*an differ h>14~fold. Thus, t.he phagrs (*an he seltdtd ac.c.ortiiny to their affinities. Sincat, the antigen c,onc,rtrtr’at.iotl is g:reat,ly in c~xc~ss of thth phagcl c~ollc~erltratiotr. if t hc, two phagcs had been mixd toget’her in the hamt’ sollltiotr with antigt,n and thrtr chapturtd \vtL ~voul~l t~xpt~c~t to SN thts same result, The rtalativr yields. caorresponti to the rxl)ec+ecl c~nric~hmrnt t htkreforr. f’;1cator for the two phages. ( )III. tntbt hod stioul(l give c*ornparat)tt~ tlisc~rirriitra1 ioli hc+wctAn aftinit its ibs that of S(*ott & Smith (l!I!K)). t)ui it clot+ not rely on thr c*ompetition of phages for liniitillg itlltigPl1. l IO” M ‘: R, < 0.1 nM). our rrquirrmrrits for antigen ex(‘ess over pha,ge anti [Ag 1 < K, mean t,hat the stattd ion should he c*arried out in a larger volume th>Ln usrtl here but is dherwise still appliddr. Although the capture of ;*rrt,igrrl-t)oun(l f)hage allows them to he selec+td ac*c~orcling t)o their afEnties, wt. rrasonrcl that a c.apt,urr procarss t);tsr)tl on the ofCrate might) t)p rnor(’ po\verfbl in vichw of the exponential nat’ure of the (aurvcs for dissocaiat ion of c*omplextd ant,ibody (Fig. 3). \Vr thrreforch devised a process in which phagtx wt’rt’ loaded with t)iot)inytatecl antigen. diluted trn~fi)ld int.0 a large t’x(‘t’ss of OII st~rt~pt~avitliliunlahelld antigen. and (2pt urfd (*oated t)ratls itftf>r a variabtt~ period of titncb. This sdrds for phages at~c.ortlitlg t’o their oft’-rates (Table :3). suggesting ill1 eriric~timent of 2fiLfoltl would I)(> t~xp~d,etl for t tit> t LVO phag,rr anti hoditas. att)hougtI t hry 1~rot)at)t~ tlifl+r t,>’ no tnor(’ t tlittl fi)urfoltl in affinit)?; or oft’-rates,
In the irnmunr: system. t,he affinit.irs of i~ntit)otiirs a,rr mat,urtd by il pro(‘ess of somatic mutation ancl antigerl-drivchll setect)ion (Derek 61 Milstc~in. 1!)87; We&art Pt 01.. 1970). aIlti t tic, itfKnity built uf, in iL st’t>p-wise manner (Sharon. 1990: Kocks 8: Rajewsky. 1988). /jr P~PO aSnit \- mat urat ioti appears to have both iill affinity ktnci a kintktica c*omponrnt (Foote & Milstein. t !-Ml). t\t t tlcl t~xtrrme. ori-rat,c:s may t)o limited by diffusio~i of’ antigen. itncl off-rates t)?; t hr rat,fl of int~rrnalizal,ion of antigen. Here we havt, ustd t.hr method of’ afftiit,y srldion to mimic. afinit,y rnaturat ion t)J, st~lt~c%iny a I)hage antibody wit.h hlght,r affinity from it tihrar\of rnutant phagta antibotlirs; it should also k~e poss~blt to select for niutarlts atxording to offrates (as described ahovcs) or prrsumahly ori-rates.
In Vitro
Antibody
Stageof immuneresponse Figure 4. C:omparison of in vivo and in vitro affinity maturation. The relative affinities (Kd) of primary and secondary response hybridomas (0) are taken from (:umano & Rajewsky (1986) and include R1.8 (the antibody used here). The affinities (Kd) of our initial B1.8 scFv and final R4 mutant are shown (w).
We used the polymerase chain reaction (Leung et ccl., 1989) to introduce random mutations into the B1.8 antibody gene. Unlike primer or synthesisbased techniques such as “spiked” oligonucleotides (Derbyshire et crl., 1986), the polymerase chain reaction can be used to mutate entire populations of diverse antibody genes. Sequencing of the B1.8 variants revealed an average of 1.7 mutations per V-gene, with a distinct bias in favour of (T/C, A/G) mutations and (A/T, T/A) mutations. We then used the affinity selection process to isolat’e a higher affinity antibody from a mutant repertoire of the B1.8 antibody using 11 rounds of select’ion. After five rounds of selection sequencing of random clones suggested that the population was oligoclonal, and one mutant (B3) was found to have a twofold improved affinity. At the end of the process (11 rounds), a single mutant (B4), with a fourfold higher affinity dominated the population (Table 4). The in vitro affinity maturation process gives an afflnity improvement comparable to that seen in anti-NP/NTP secondary response hybridomas (Fig. 4). The modal affinity of anti-NIP secondary response hybridomas improves fourfold over the primary response and the best observed increase is sixfold (Cumano & Rajewsky, 1986). although it has been shown experimentally that a single mutation in CDRI (Trp33-+Leu) can increase t)he affinit)y up to tenfold (Allen et al., 1988). Other characteristic changes involve the cent’ral residues of the DFL 16.1 II-segment (Cumano & Rajewsky. 1986). The Serl04-+Gly mutation in mutants B3 and K4 (and responsible for the affinity increase in 1%) is located in the II-segment and appears (with other mutat)ions) in 2/10 secondary response hybrifrom tlomas (Allen et nl., 1988) and 4/23 hybridomas transf&red mrmory cells (Siekevit,z et al., 1987). To mimic. t)ertiary and later immune responses and t,o obtain the highest affinity antibodies, larger and more ext’rnsively mutated libraries could be used. or rounds of mutations could be introduced between the cycles of growth and selection. A large tlumber of cycles may be required. but it is possible
895
A.finit?y Maturation
to interpose two rounds of selection between each round of growth (Table 2), and to process several antibodies in parallel. It may also be advantageous to combine independent single mutations from different clones in a response” “secondary response” into one “tertiary clone, for example by taking advantage of PCR crossover between the highly related templates. The affinities of the single point mutants can prove additive (Kocks & Rajewsky, 1988; Sharon. 1990; Foote & Winter, 1992). If multiple mutations are needed to improve antibody affinities by large factors, the selective mutation of the hypervariable loops with synthetic “spiked” oligonucleotide primers may prove more attractive than random mutation of the entire gene. Such primers could be designed for each hypervariable loop of each of the germ-line V-genes. Indeed by “cassette mutaof regions identified as important by genesis” alanine scanning, a range of variants of human growth hormone were selected by panning from phage libraries: their affinities for the receptor ranged from threefold worse to eightfold greater than the wild-type (Lowman et al., i991). In summary, we have demonstrated methods of affinity and off-rate selection for antibodies displayed on the surface of bacteriophage and used the method of affinity selection to increase the affinity of a primary response antibody. The selection method allowed us to mimic, and may in the future allow us to improve on the natural process of antihody affinity maturation. However, the select,ion method is by no means restricted to libraries of random point mutants as the V-genes could be diversified by many other strategies, for example by gene conversion or chain shuffling. N’e thank C. Milstein. A. Fersht and a referee for critically reading the manuscript, J. Foote for advice roncernmg affinity measurements and the humanized III.3 c*lone, P. Holliger, T. Simon. T. Bonnert and M. Hxier for advice and gifts of reagent,s. R.E.H. is supported by an MRC Recombinant DPU’A Training Fellowship. S..J.R is supported by an MRC (‘linician Srientist F~~llowship.
References Allen, I).. Simon. T., Sablitzky. F., Rajrwxky, K. & (‘umano. A. (1988). Antibody engineering for the analysis of affinity maturation of an anti-hapten response. 14:MRO J. 7, 1995-2001. JSarbas. C. F.. Kang. A. S.. Lerner, R. :I. & Benkovic. S .I. (I991 ). Assembly of rombinatorial antihody libraries on phage surfaces: t’he gene ITT site. Proc.
Sat. Acad. Sci., f7.S.A. 88, 7978-7982. Bass. S.. (:rrene. R. & Wells. .J. A. (1990). Hormone phage: an enrichment method for variant proteins with altered binding properties. f’rotths. 8. 309-314. Ktarek. (‘. 8r Milstein, (‘. (1987). Mutation drift and repert,oirca shift in the maturation of the immune rr’sponse. Immunol. Kec. 96. Z-41 Bhat. ‘I’. N.. I+ntlry. (:. A., Fischmann. T. 0.. Boulot, C:. & I’oljak. R. .I. (1990). Small rearrangements in SIruc.tures of Fv and Fah fragmrnts of antibody D1.3 OII ant,igPn binding. X’crfurP (Loudon). 347. W-485.
ISrritling. F.. Diibrl. S.. Srehaus. T.. Klrwinghaus. I, B Lit,tlr, M. (1991 ), A surface expression ve(.tor 1’01 antibody scrrrning. (&r/e. 104, M-153. Krownstone. A.. Mitchison. Xv. =\. & Pitt~Rivers. 1~. (lQ66). C’hemical and serological studies with an iodine-containing svnthetic immunological determinant ~-hSdroxy-S-iodo-;i-nitroph~ttyla~~eti,, acid (XII’) and related compounds. I77arrc,rnolo!/y. 10. 465-48 I. Eurton. I). It.. Barbas. (‘. F.. l’rrsson. M. A. .I.. Koenig. S.. (‘lrarrwk. R. $1. & Lerner. R. ~1. (1QQl). .L\ large array of human rnonoclonal antihodirs to typca I huma,n itnmunodeficien~~~~ virus from combinatorial librarirs of asymptomati(. srropositiv? individuals. I’roc. Z’at, Acnrl. A??., C’.S.,d. 88. lOlR1-10137. Clac*kson. T., Hoogenboom. H. R.. Grifiths. A. I). KWinter. C:. (1991). Making antibody fragments using phagr display libraries. Snt/rrr I London). 352. 624-628. (‘urnatto. ~1. & Rajrusk? K. (1986). (‘lorral rr~ruitment generation Wld somatic, mutation it1 the of irrlmclnologic~al memory to the hapten XI’. I?,VNO .J. 5. 2459-2468. I)rrhyshirr. K. $1.. Salvo, .I. ,I. & (:rirtdlry. Iu. 1). F. (I 986). :I simple and efic%nt proc*edurr fot saturation mutagenesis using mixrtl trligorlu~lrvtitlrs. c:rnr. 46. 145-l . .5%. I)owrr, 11’. -1.. Miller. J. F. & Ragsdale. C’. \f:. (1988). High c~ffic~ienc~y tzransformation of E. colt’ by high voltage t,lrc~t,roporation. Xtr~l. Acids RPS. 16. 6l%i-6145. Foot,e. J. & Milsteirr. (‘. (1991). Kinetic maturation of an irn m une response. Sature (London J 352. T,30-:T,R?. Foote, .I. & \Vint,rr. (:. (1992). Antibody framework confbrmation wsidurs aflerting t,hr of the hyprrvariablr loops. d. A’ol. Biol. 224. 48G4QQ (:ibsort. ‘I’. (1984). Studies on the Epstein-Barr virus genome. Ph.D. thrsis. University of (:ambridgch. (ilasrr-Wuttkr. C;.. Keppner ,J. Br Rasched I. (1989). Pore-forming properties of the adsorpt,ion protein of filamrntous phagr fd. Riochinr. Riophys. Acta. 985. 239-247. (iray. C’.. Brown. II. 8r. Marvin. I). (l!+Xl). Adsorption c~~mplex of filament,ous fd virus. J. A’ol. Hiol. 146. 621-627. Hawkins. Ii.. E. Br Winter. G. (1992). (‘rll selection strategies for making antibodies from variable gene libraries: tapping the memory pool. Eur. J. In~n~unol. 22. 867--X70. Hrmsley. A.. Arnheim. N.. Tone\, M. I)., (‘ortopassi, (:. Hr (ialas. I>. .I. (I 989). A simple method for site-directed mutagenesis using the polymerasp chain reac%ion. ~Vucl. =1cids Res. 17, 65456551. Hoogenboom, H. R.. Griffiths, A. D., ,Johnson, K. S.. (‘hiswell. D. ,J., Hudson. P. & Winter. G. (1991 ). Multi-subunit proteins on the surface of filamentous phagr: rnethodologies for displaying antibody (Fat)) hoaq and light chains. Mucl. Acids RP.s. 19. 4133-4137. Hose. W. I).. Sastry. I,.. Tverson, S., Kang, A. S., Alt,ingMers. M., Burton. D. R., Benkovic, S. .J. & Lerner, R. A. (1989). (:eneration of a large combinatorial library of the immunoglobulin library in phage lambda. Science, 246, 1275-1281. .Iones. I’. T.. Dear, P. H., Foote. tJ., Neuberger, M. S. & Winter. (:. (1986). Replacing the complementaritydetermining regions in a human antibody with those from a mo&e. Nature (London), 321. 522-525. Kang, A. S.. Rarbas, (‘. F., Janda, K. D., Renkovic. S. .J. & Lcrner. R. A. (1991). Linkage of recognition and replicat~ion functions by assembling combinat,orial antibody Fab libraries along phage surfaces. Proc. Sol. =1cad. Sri.. I’.S.A. 88. 4363-4366.
Lrung. I). LIP.. (‘hrrr. b:. & (~c~rdtlel. I). V. (I%%)) A method for random mutagenesis of’ a dt+trrtl lJS.1 segment using a modified polyrnc~rasr c.ttain rc~;l,c~ti~,tr Trrhniqur. 1. 1 I -15. Lowmarr. H. I. .I, .\. (1991). Scalecating high afl?nit>, binding J)rotcains I)> Niochcnti.utry 30 monovalent phagr ctisplay. 10832~
108RX.
McCafferty. .I.. (~riffit~hs. A I).. Wint,rr. (;. & (‘hiswc~ll. I). ,I. (1990). I’hage ant)ibodirs: filamentous phage displaying aritibod> variablt domaitrh. ,v’nturr, (Londo,~). 348. 552-5.X .I. I).. Hoogenboom. H. I\‘.. Honnrrt, 7‘ I’.. Marks, M(~(‘af&rty. ,I. & (:rifIitha. A. I). (1991). By-passing itntntcnizwtion: human atttibodies from \--g,rrtlt. librarirh displayed on bac+rriophag,rr. .I .‘il
Selection of Phage Antibodies Mimicking
by Binding Affinity
Affinity Maturation
Robert E. Hawkins’?, Stephen J. Russell’ and Greg Winter’*2t ‘M RC Laboratory of Molecular Biology, and 2MRC Centre for Protein, Engineering Hills Rocxd. Cam,hridgr (‘RZ 2&H. (‘.h’. (Received 21 .Jcwwary
1.992; accepted 10 April
1992)
We describe a process, based on display of antibodies on the surface of filamentous bacteriophage, for selecting antibodies either by their affinity for antigen or by their kinetics of dissociation (off-rate) from ant,igen. For affinity selection, phage are mixed with small amounts of soluble biotinylated antigen (< 1 pg) such t’hat t’he antigen is in excess over phage but with the concentration of antigen lower t,han the dissociation constant (Kd) of t,he antibody. Those phage bound to antigen are then selected using streptavidin-coated paramagnetic beads. The process can dist,inguish between antibodies with closelv related affinities. For off-rate selection. antibodies are preloaded with biotinylated antigen and diluted into excess unlabelled ant,igen for variable times prior to capture on streptaridin coated paramagnetic beads. To mimic the afinit’y maturation process of the immune system. we introduced random mutations int,o the a&body genes in vitro using an errorprone polymerase, and used affinity select)ion to iso1at.e mutants with improved afhnitv. Start,ing with a small library (40.000 clones) of mutants (average 1.7 base changes per Vu gene) of thrb mouse antibody B1.8, a,nd using several rounds of affinity selection. we isolatcad a mut)ant wit’h a fourfold improved affinity to the hapten 4-hydroxy-Fi-iodo-3-nitrophrnac~etyl-( NH’-caproic acid (mut,ant h’, = 9.4( 5 0.3) nM compared wit,h B1.8 h’, = 419( k 1%) n;M). The relat’ive increase in affinity of the mutant is comparable to the increase seen in the anti-4-hydroxy-3 nitrophenylacetyl/NIP-caproic acid murine secondary immune response. h’rywords: filamrnt,ous phagt,: antibody: affinity: maturation
1. Introduction The display of antibody fragments on the surface of filamentous phage that encode the antibody genes (MctJafferty et aZ., 1990: Hoogenboom et al.. 1991: Kang et nl.; 1991; Kreitling et al., 1991; Barbas et al.. 1991), and the selection of phage with antigen binding a&vi&s by panning, offer a powerful way of making antibodres from V-gene libraries from immunized animals or humans (Clackson et al.. 1991; Burt)on it al., 1991). Very rare phage can be isolat,ed. because in contrast to filter screening of soluble antibodies (Huse of al., 1989), the number of phage t,hat can be screened is limited only by the size of the init,ial library. Using the tnRNA of spleen cells taken from hyperimmunized animals (Huse et al., 19X9), or the 1)NA of antigen-selected memory cells, the V-genes encoding antigen binding actrvrties are greatly t Authors to lvhom all correspondence should be addrt~ssrd
enriched (Hawkins & Winter. 1991). However, in small random combinatorial libraries derived from a large number of heavy and light chain V-gene elrment)s (Huse rt ul.. 1989), t,he original pairings between ant’ibody heavy and light, clhain of the B-cell are likely to be lost (Winter & Milstein, 1991). Indeed, pairings from such libraries have been shown to be promiscuous. For example. different light chains were found associated with a single heavy chain. and vice uer~u (Clackson et rcl.. 1991). Which (if any) of these combinations correspond to the original pairing in the B-cell is not clear. Since the original B-cell pairings are selected by the immune system according to their affinity of binding or their kinetics of binding t,o antigen (Foote bz Milstein, 1991), many of the pairings in a combinat’orial library are likely to have a lower affinity. However. in a very large library (> 10’ members), some of the original pairings will be present (Winter & Milstein. 1991). and a few of the new pairings may even have an improved affinity. Tt
triutatil random It is also drsirahle to selec3 the highest. &nit! ittltji hod& from “secondary” phagr ;~tltiho(l~ lihrwt,ies iti \vhich t.hch selrct,rd V-gene pairings from it primary library art’ further divrrsified, for c~xatnple. hy keeping one chain fixed and shuffling the other ((‘lac*ksotr f,t rrl.. I!I!II ). Kew pairings can he c*rc~ated, k~adittp to atttihotli~s with irrrprovc4 affinities (Marks rt cl!.. 1992). Altt~rrratjivrly, random F)oint mutation ~ultl hc used. as in th(t immune system (Milstjein, 1990; Winter Kr Milstein. 1991). Tnd4. the isolation of human antihodirs wit bout imtntttrizafion. using a single large anti (Ii\-erse \7-get~c library displayed on phage (Marks it n/.. 19!)1 )- and t,htbir improvemc~nt should canable evt’n t h(> highest affinity antibodies to he tnadr t~titircly itt rifrft (\l:inter $ LMilstIeirt. 1991). Howrvpr, it’ is difficult to select hrtwren phages acc*ordirtg to their attinities. especially het,wclcbtt t,hose with similar afinities. Phages hearing high affinit.y antibodies caan he selecGd from those with low affinities using antigen immobilized on solid phase. but finr affinity disc>rimination is dificult t)tac.aruse the binding is dictated hy both the afIinit>. and the avidity of the phage (Cllackson et al., 1991 ). Alternatively. phage antibodies could he selected by hitiding to biotinylated antigen in solution, followed by csapture of the phagt on a strept~tvidin-~oat~,(l as shown for phage displaying small surface. peptides (Parmlry & Smith. 1988). The high affinity, phagr antibodies could then hr enricahed hy compe’ tit ion for limit,ing amounts of antigen, as prnposPt1 for peptide phagc (Scott & Smith. 1990). But for affnity maturation of phage antibodies, the mart! low affinity (parent) phage could caornpete ant tht, hindiny of rare high aftinity variant’s, \.VP have thereforr adapted t hc hiotin-c+aptttrr tnrthod (Parmlry & Smith. 19t18) for the affinit? rnat.urat,ion of phage antibodies. Inst.ead of limiting antigen caoncentration such that competitiott for thtb antigen occurs (Scott & Smith. 1990), WP used the antigen in excess over phage to drive t,he binding. To favour the binding of the higher afinity phapcl ant,ihodies. we srt thr concentration of antigen (Ag] lower than t)he desired dissocGtion chonstant (h’,). The phage antibodies in complex wit,h biotinylated ent,igrn were then captured on streptavidin-c:oatetl paramagnet.ic heads. The discrimination M wept1 different ant,ibody mutants is related to the ratio of their affinities. This process has also been adaptcd to select for prolonged off-rat,e: after binding of biot~inylated antigen the phage are diluted into ex~ss unlahelled a,ntigen. Riotinylated antigen then dissociates frctm t,he phagt: at, a rate proportional to koff and the exposed antibody heads are hound b; unlahelled antigen. The longer this off-reaction IS allowed to proceed before capture, the greater the selection for prolonged off-rate. We have used the method of affinity selection to achieve affinity maturation of’ a primary response antibody (anti-NP/NTPt hpbridoma) by selecting a
\I.il ti irnprovt‘d
affinity
2. Materials Bovittr
trottt
ii i~i,t,;~t? rot
ttrirl writ s.
and Methods
(il)
.-Inti!qrtt
set’ttm albumin
(BSA:
Kot~hrirt#rt,~ILl;tttttltt,ittt
(;rrma,ny)
was c,onjugated with ,h;TP-c~aproat~~-(~~stt(.(,ittt rnitlr ((‘ambridge Research Kiochrmicals. SorthM ic,lr. I!.K.) as tlc~scrihrd l)y Browttstonr c/ rrl. (I!Mi). Tht. act.ivatrtl haptett was tlissolvecl in tlittrt~th~lf;~rtrtar~~itl~~ attd added rithtbr in a molar ratio of IO: I 01’ 0.7: 1 STP-c~a~~roatr-l~-su~~~~inimitlr;I~S1\ (;“;W,“HSA and in 0.2 hi-soclirrtrt hytlrog:c~rr Nl P0.71;-‘-EDT.!. In t.he c~xllerimrttts using motlt~l antihodirs the s&ction methods were trstrd to examine the rffecats of antigen ~o~ic~rntration. “double” rounds of selection. and time after floodittg ott the number of phagr bound. For this purpoar the srlr?tiotts were performed in clif%retit tubes ~~rowsst~i itt parallrl and the titrrs of phagt. c>lrrted \vt’re c~c~triliai~f~~l.
The phage sc~Fvl)l.3 and huscFvl)l.3 wer(’ incaubated srlbarately with a low conclentration of Iliot.in>~lated HEI, (IO n>f) at room t~rmprrature for I h in KYHS. The off rrac,tion was started by diluting 50 /LI pctrtions of this binding reaction into ex(bess unlabeilrd H El, (950 ~1 of a 20 pt stocak in tSPl%). and leaving for variable times. I’hage rrtaitting hound biotinylat,rd H E:l. were then paracapturt~d by adding 6 x IO’ strrftta~iditt-c,oatrd tnagnrtir beads and incubating for I ntin with c.otrtinuous mixing. The heads were washed 7 times with BPIIS. I’hag~e LStar? elutrd and infrc.tetl as desc,ribrtl above.
The binding of the sc*FvKl.8 phagt> to the 9ll’,,I~SA (.ott,jugatta for EI,ISA was ltt:rf~~rmrtl as tfesc-ribed by &l(.(‘afferty ef nl. (1990). and thr expression and binding of soluble sc~F~Bl.8 for ELlSA as tl). Hawkins Xr \Vinter (1!)!)2). Sequencing template was prepared from phagr as dt~s~ribed by (‘lac*kson rf al. (1991 ) and srqtrc~ncrd (Sanget uf ml.. 1977) \vith phagr T7 l)iYA l)olyttteras~~ (Srquenasr. I’sIS. (‘leveland. I!.S.A.). The expression of t,hr soluble sc,Fv KI .X f’ragtnrnts was intlucrd using 1 1 of culture grown in shaker flasks (see Hawkins & M’intrr. 1992). Thtb bac+ria were pelleted. taken up in 2.5 tnl of PBS. I mw-F:l)T?r and the fragments harvested from the periplastn (Skrrra rf ol. l!Nl) 3 h after itttluciion. Tht, frac’tion was applircl to it I ml XT’Spharosr column. washed with IO ml of’ PISS and elut,ed with .Aml of’ OlwHCl, glycinr (pH 2.5). Thti rluted scFv w;l;, c.ollf~c~t?c!in I ml fractions into 0.5 ml of ic,e-(*old I at-Tris.H(‘I (ltH 7.4) and then the pcbak 1)rotcsin fractions dialysrd into I’M. hfYinit:- n~t~asurrmrttts wart’ lirrformed by fluort~s~rnc~c~quench tttration usinlr tht> hapten XII’(‘.\P-OH as drsc>ribrrl (.Jonrs rt CI!.. IHXR). ‘I’h(, anti-lysozyme scFv fragmrttts \vvr(’ purified as dt~~cribed by ivard et ~1. (1!%9) and affinities tnrasurrd by flttortxs(.enc*r qut.n(lh titratiott (Foote R- \2’inttLr, 19!)2).
3. Results (a)
S&ction
hy czfinity
or o$rlrtp
Flnorescrrtc~~ yuenc+ titration tneasurements of dissociation caonstants. for I)roleir~-t~ritiliotlv inter-
~.-___-
___
_--Table 1
actions, give values that, are wnsistent) wit,h ~Jswtdoequilibrium relaxation methods hut tend to give high standard errors (Marks rt al.. 1991: Foot,e 8 Winter, 1992). We thus determined the K, value fot scFvD1.3 as 15 (+X) n>t and sc:FvDl .3 (TyrlOl --+Phe) as 250 (+_80) IIM. This shows t’hat the mutant scFv fragment has a dissociatiott wnstant approximat,ely 17.fold higher t hatt t hr parent scFvDl.3. The HuLysl 1 rwtta,prd antihod! h as been shown to have a dissociation constant fourfold higher than the parent Dl.3 antihotly (Foote & Winter. 1992), and presumably the differfwce is retained in the scFv fragment,s. i‘he ratios of the dissociation constants fiJ1 wFvl)l .3. huwFvDl.3, scFvl)l .3 (Tyr101 +Phr) are thrwfore HkJOUt
1 : 4 : 17.
The
titration of scFvI)l.3 and sc:Fvl)l.3 -+Phe) phage with hiotinylated HF: L (‘l’abtc 1) shows that’ t,he phape con ~JP wlrc~ted acwrding to their afinities, rspwiwll,v at t,hr Iowt‘r c.otrc.f,ntratiort of H It: I,. as expected frotn our tnodcl (see hrlow). Here. t.hr 17.fold higher affinity of the wl’v I)1 .3 over the wFvDl.3 (TyrlOl -+F’he) heads allows it 14fold greater pi&~ of the former phagc’ f’rom the heads. At high conc.rntratiorts of HEI. (100 rt>T) the titre of’ phage c~aptured falls. pwsunnhly due to c:otnpetitioti hJ: exc’f’ss free antigetr (not (TyrlOl
Table
3
selection. Thtb swottd round ap~~~ws to IF tttow efficient, than the first, as shown hy t,hrk rlutiott of wFvD1.3 and huscFvl)l.3 phagc (Table 2): the yield is 0.3?, from t,he first round. and 1.75”,, from the second. Furthermore. the fourfold higher af& tlit,y of the wFvI)l.3 over the hus~Fvl)l.3 givrs a 2.5.fold greatet yield of the former phagc frOll1 the tuo-foltl 1~~x1s in the titst round. and a furthw factor in the swond round. The fraction of’ l)hagtl c~apturrd in the ahsenw of hiot,inylated antigen was very low in both rounds (5 x IO- ’ ()(, of input phap in round 1 and 04)3(~,, in round 2). ‘I’hC af%tritirs Of ttlr I)1 .:j itIltiiJOt1~ iJ.tld thfl hlttTl& nized I)1 .S itrttihod>~ (Hul,ysl 1) diffw k)x, ti~ttrfold. ;Lrld this is largely attrihtttahlr to ttlrl dtf?errtrc~c~ in off-rates (Foote & Il’inter. 1992). For off-r:tt,r sele(. tion we used the phagr displaying scFv frwgrrtt~nts &rived f’rotn t,hwr antihodirs. ‘l’nl~lrx 3 hhows that ()ti-rat,(~ selection u it h hiofitlylate(1 H EI, pvt. I~IOW t ha,rt 8. ZO-fold gWittf>t’ yield of scaFVI) I .3 ])hag,lt,0vf.r huwl+l)l .3 lJhagt> af’t)tsr six minrrtt~s. Howt~v~~r. 1~). I:! mitiutw. ttrc* t1iffVrc~nc.c~iti ?,ielcl IlitCl fallow to 7..5ffjtd.
?J,S Itl?
Of
tlllttll~f~l~
Yll7t~Vl
[)tl2Lgt’
f’iYJtIt t ttC’ I)t’ittlS
approwhwl its t~ac~kgrottn~l \.;tlut~ (al)pwttt ai 60 rnin). lndwtl. hackground hitttlitry is a rnaljor I’A(*tc>t~ lirnitittg Itic, f)o\vtlr of ofi-ratcb st~lw~tion.
stlon~tl). Two cwnsewt~ivc~ rounds of wlectiott c-art tw used (\vit bout, growth of thtb phage in hact.eria hrt~wren ra,cdtt round) IJ~ eluting the phage from the heads with t.riethylamine. neutralizing t hc eluat,e and suhjwt’ing the phap to w second round of affrtit?.
11 clorirs wt’w sec~uc~ricwl. Thtk ovwall ttiutatioti freqitt~nc~y \vas 4 .“]kh (1.7 pet’ \‘,, getw). XII t,ypw ot tnutat~iott \v(Lt’c‘ stbert t~xwp~ (i/t ’ or (‘,‘(i c*hartgw I~ut tht~rr wits it l)iilS iri favorer of (‘I‘,‘(‘..&/(:) tnutations ilnd (A4:rr.7;.4j triut;rtiotts (‘r:(-. A(: (9): ‘r#A. ,\;‘I’
(5): r/c;. .I:(’ (2): (~3’. (::‘:i (3): ((:.-I. (::7’ (3); (7:. (i/t’
Tit ws of wFvl)l.:3 ptl”g”
(0)).
‘I’tlf~
tlltttl~jrt~
grrw
(arid
J- [~f’t’
1’”
gt~trw
t~rlrriittirtrtl.
Of’ tIltttittiOtlS with
no
“hot
\arif~cl SfJols”)
f’tY)tll \vit
0 h
t0
I): I4
f)tlii#‘S wwf~ sf~ltY*trfl \cit’ti X tint-hiot ittylatwl NTPo.,l~SA (rounds 1 and 2). and gr;ldually rt~tluwd to 1 1111 STI’,.,KSA OVPI sitcwssivc~ routtds. Incarrasittg ttum brrs of phaps \verc c>lutetl afttlr f’iL(‘h round ant1 t~tw tiutnhrrs rxptdfd brtwwtr fetch wlwt iott I)\. rcbittftLc*lion and growth (Jlatc~rials and .‘sitiglc round”). I’hagts itil)ttt \VitS always .\lethods. The
3x x0 x O~IT,X 1.4x lHl2.i x
10’0 106 106 ttlh IOh
IO0
pl
2 x
1o’O
of
a
100
fJtlil#‘);
i\t taac*h round
X IJh+y j .?\g]
fYJtlf?tltt’iltt~ ilrltl
\v(‘re: rottttcl
tllltltlJ~t5
(a~J~Jroxirllat~rl? Of
[JhLLgt’
f’llltfY1
I d rt>l. 0.7 x IO’: round
In Vitro
Antibody
Table 4 acid dijjewnces and afJinity measurements mutant Bl .X antibodiw isolated from library
Amino
of
Aginity
893
Maturation
t,itration and shown t,o range from 48 nM t)o 9 nM (Table 4). 4. Discussion
Number of active sites per scFv fragment
34 63 8% !)3 104 106 111 Kd$ (nM) (‘lOW IS1.X Ies, multivalent binding of phage to the para magnetic brads will help prevent dissociation during thtb washing steps. However, mult,ivdent, hintliny (‘an arise directly or indirectly. Direct multivalent capt,ure c~mld arisr when at, least t)wo heads arc simult~aneously complexed with hintin?-lated antigen in solution. Assuming three ant,ihody hratls per phage (Gray rt WI.. 1981; (ilaser-Wutt)ke it nl.. 1989) and a proport~ion (p) of the heads t~onnti to antigrn. then thtl fradon of phage with two or more heads bound is: p3 + 3p2( 1 -p). a,nd the frac*t)ion with one head bound (rnonovalent~) is: 3p( I -P)~. When [ Ag] is signific~ant,ly lower than h’,. t.htL majority of phage will he monovalent, retluc*ing the likelihood of tlirrct multivalent (*Rptllrt~. Howrvrr. rnonovalent~ Capt,ure cY)uttl hc cY)rivert~ed indirrdly to multivalent (sapture if thr phag~~ I~ountl to (~x(*ess antigen on the head surface with their other heads. Finally the antigrn-c:oatetl heads c~mltl capt,ure low affinity phage that have not hound antigen in solution. In light of the different possihlc modes of phagr capture. it is dific~ult to caal(*ulat.r t,hr power of st+c:t ion. Howevflr, for capturr of monovalent phages only. we would expect a capture ratio (phayr I : phage 2) of (K, + IAgJ)/(K, + 1AR]) x 1pAhI ]i 1pAt,%], whereas for dired capture of bivalrnt phagc only. the srlrc+ion is rnorr powerfill. with iin (*apt urp of approximat,rl> t*xprc+ecl ratin (K2 +lAgj)2,‘(KI + [A4g])‘x [pAbl]/[pAbZ], hut at t trc, tlxpense of frwrr phage csapturtd. Whichevtar t)ro(*(bss dominates. tht, power of the selection will I)tl reduced by unwant’rtl capture of low aflinity phagr. Nrvertheless the selrrtion of phagrs at differt~nt c,onc,t~ntrat’ions of antigen (Table I). with [HEI,] < K, of ttot’h antihotlies. shows that when thr affin-t itas of each head for antigen differ 1)~. 1 ‘i-fold. t hr ?;itJlds 01’ ptiage hiding to bea,& (*an differ h>14~fold. Thus, t.he phagrs (*an he seltdtd ac.c.ortiiny to their affinities. Sincat, the antigen c,onc,rtrtr’at.iotl is g:reat,ly in c~xc~ss of thth phagcl c~ollc~erltratiotr. if t hc, two phagcs had been mixd toget’her in the hamt’ sollltiotr with antigt,n and thrtr chapturtd \vtL ~voul~l t~xpt~c~t to SN thts same result, The rtalativr yields. caorresponti to the rxl)ec+ecl c~nric~hmrnt t htkreforr. f’;1cator for the two phages. ( )III. tntbt hod stioul(l give c*ornparat)tt~ tlisc~rirriitra1 ioli hc+wctAn aftinit its ibs that of S(*ott & Smith (l!I!K)). t)ui it clot+ not rely on thr c*ompetition of phages for liniitillg itlltigPl1. l IO” M ‘: R, < 0.1 nM). our rrquirrmrrits for antigen ex(‘ess over pha,ge anti [Ag 1 < K, mean t,hat the stattd ion should he c*arried out in a larger volume th>Ln usrtl here but is dherwise still appliddr. Although the capture of ;*rrt,igrrl-t)oun(l f)hage allows them to he selec+td ac*c~orcling t)o their afEnties, wt. rrasonrcl that a c.apt,urr procarss t);tsr)tl on the ofCrate might) t)p rnor(’ po\verfbl in vichw of the exponential nat’ure of the (aurvcs for dissocaiat ion of c*omplextd ant,ibody (Fig. 3). \Vr thrreforch devised a process in which phagtx wt’rt’ loaded with t)iot)inytatecl antigen. diluted trn~fi)ld int.0 a large t’x(‘t’ss of OII st~rt~pt~avitliliunlahelld antigen. and (2pt urfd (*oated t)ratls itftf>r a variabtt~ period of titncb. This sdrds for phages at~c.ortlitlg t’o their oft’-rates (Table :3). suggesting ill1 eriric~timent of 2fiLfoltl would I)(> t~xp~d,etl for t tit> t LVO phag,rr anti hoditas. att)hougtI t hry 1~rot)at)t~ tlifl+r t,>’ no tnor(’ t tlittl fi)urfoltl in affinit)?; or oft’-rates,
In the irnmunr: system. t,he affinit.irs of i~ntit)otiirs a,rr mat,urtd by il pro(‘ess of somatic mutation ancl antigerl-drivchll setect)ion (Derek 61 Milstc~in. 1!)87; We&art Pt 01.. 1970). aIlti t tic, itfKnity built uf, in iL st’t>p-wise manner (Sharon. 1990: Kocks 8: Rajewsky. 1988). /jr P~PO aSnit \- mat urat ioti appears to have both iill affinity ktnci a kintktica c*omponrnt (Foote & Milstein. t !-Ml). t\t t tlcl t~xtrrme. ori-rat,c:s may t)o limited by diffusio~i of’ antigen. itncl off-rates t)?; t hr rat,fl of int~rrnalizal,ion of antigen. Here we havt, ustd t.hr method of’ afftiit,y srldion to mimic. afinit,y rnaturat ion t)J, st~lt~c%iny a I)hage antibody wit.h hlght,r affinity from it tihrar\of rnutant phagta antibotlirs; it should also k~e poss~blt to select for niutarlts atxording to offrates (as described ahovcs) or prrsumahly ori-rates.
In Vitro
Antibody
Stageof immuneresponse Figure 4. C:omparison of in vivo and in vitro affinity maturation. The relative affinities (Kd) of primary and secondary response hybridomas (0) are taken from (:umano & Rajewsky (1986) and include R1.8 (the antibody used here). The affinities (Kd) of our initial B1.8 scFv and final R4 mutant are shown (w).
We used the polymerase chain reaction (Leung et ccl., 1989) to introduce random mutations into the B1.8 antibody gene. Unlike primer or synthesisbased techniques such as “spiked” oligonucleotides (Derbyshire et crl., 1986), the polymerase chain reaction can be used to mutate entire populations of diverse antibody genes. Sequencing of the B1.8 variants revealed an average of 1.7 mutations per V-gene, with a distinct bias in favour of (T/C, A/G) mutations and (A/T, T/A) mutations. We then used the affinity selection process to isolat’e a higher affinity antibody from a mutant repertoire of the B1.8 antibody using 11 rounds of select’ion. After five rounds of selection sequencing of random clones suggested that the population was oligoclonal, and one mutant (B3) was found to have a twofold improved affinity. At the end of the process (11 rounds), a single mutant (B4), with a fourfold higher affinity dominated the population (Table 4). The in vitro affinity maturation process gives an afflnity improvement comparable to that seen in anti-NP/NTP secondary response hybridomas (Fig. 4). The modal affinity of anti-NIP secondary response hybridomas improves fourfold over the primary response and the best observed increase is sixfold (Cumano & Rajewsky, 1986). although it has been shown experimentally that a single mutation in CDRI (Trp33-+Leu) can increase t)he affinit)y up to tenfold (Allen et al., 1988). Other characteristic changes involve the cent’ral residues of the DFL 16.1 II-segment (Cumano & Rajewsky. 1986). The Serl04-+Gly mutation in mutants B3 and K4 (and responsible for the affinity increase in 1%) is located in the II-segment and appears (with other mutat)ions) in 2/10 secondary response hybrifrom tlomas (Allen et nl., 1988) and 4/23 hybridomas transf&red mrmory cells (Siekevit,z et al., 1987). To mimic. t)ertiary and later immune responses and t,o obtain the highest affinity antibodies, larger and more ext’rnsively mutated libraries could be used. or rounds of mutations could be introduced between the cycles of growth and selection. A large tlumber of cycles may be required. but it is possible
895
A.finit?y Maturation
to interpose two rounds of selection between each round of growth (Table 2), and to process several antibodies in parallel. It may also be advantageous to combine independent single mutations from different clones in a response” “secondary response” into one “tertiary clone, for example by taking advantage of PCR crossover between the highly related templates. The affinities of the single point mutants can prove additive (Kocks & Rajewsky, 1988; Sharon. 1990; Foote & Winter, 1992). If multiple mutations are needed to improve antibody affinities by large factors, the selective mutation of the hypervariable loops with synthetic “spiked” oligonucleotide primers may prove more attractive than random mutation of the entire gene. Such primers could be designed for each hypervariable loop of each of the germ-line V-genes. Indeed by “cassette mutaof regions identified as important by genesis” alanine scanning, a range of variants of human growth hormone were selected by panning from phage libraries: their affinities for the receptor ranged from threefold worse to eightfold greater than the wild-type (Lowman et al., i991). In summary, we have demonstrated methods of affinity and off-rate selection for antibodies displayed on the surface of bacteriophage and used the method of affinity selection to increase the affinity of a primary response antibody. The selection method allowed us to mimic, and may in the future allow us to improve on the natural process of antihody affinity maturation. However, the select,ion method is by no means restricted to libraries of random point mutants as the V-genes could be diversified by many other strategies, for example by gene conversion or chain shuffling. N’e thank C. Milstein. A. Fersht and a referee for critically reading the manuscript, J. Foote for advice roncernmg affinity measurements and the humanized III.3 c*lone, P. Holliger, T. Simon. T. Bonnert and M. Hxier for advice and gifts of reagent,s. R.E.H. is supported by an MRC Recombinant DPU’A Training Fellowship. S..J.R is supported by an MRC (‘linician Srientist F~~llowship.
References Allen, I).. Simon. T., Sablitzky. F., Rajrwxky, K. & (‘umano. A. (1988). Antibody engineering for the analysis of affinity maturation of an anti-hapten response. 14:MRO J. 7, 1995-2001. JSarbas. C. F.. Kang. A. S.. Lerner, R. :I. & Benkovic. S .I. (I991 ). Assembly of rombinatorial antihody libraries on phage surfaces: t’he gene ITT site. Proc.
Sat. Acad. Sci., f7.S.A. 88, 7978-7982. Bass. S.. (:rrene. R. & Wells. .J. A. (1990). Hormone phage: an enrichment method for variant proteins with altered binding properties. f’rotths. 8. 309-314. Ktarek. (‘. 8r Milstein, (‘. (1987). Mutation drift and repert,oirca shift in the maturation of the immune rr’sponse. Immunol. Kec. 96. Z-41 Bhat. ‘I’. N.. I+ntlry. (:. A., Fischmann. T. 0.. Boulot, C:. & I’oljak. R. .I. (1990). Small rearrangements in SIruc.tures of Fv and Fah fragmrnts of antibody D1.3 OII ant,igPn binding. X’crfurP (Loudon). 347. W-485.
ISrritling. F.. Diibrl. S.. Srehaus. T.. Klrwinghaus. I, B Lit,tlr, M. (1991 ), A surface expression ve(.tor 1’01 antibody scrrrning. (&r/e. 104, M-153. Krownstone. A.. Mitchison. Xv. =\. & Pitt~Rivers. 1~. (lQ66). C’hemical and serological studies with an iodine-containing svnthetic immunological determinant ~-hSdroxy-S-iodo-;i-nitroph~ttyla~~eti,, acid (XII’) and related compounds. I77arrc,rnolo!/y. 10. 465-48 I. Eurton. I). It.. Barbas. (‘. F.. l’rrsson. M. A. .I.. Koenig. S.. (‘lrarrwk. R. $1. & Lerner. R. ~1. (1QQl). .L\ large array of human rnonoclonal antihodirs to typca I huma,n itnmunodeficien~~~~ virus from combinatorial librarirs of asymptomati(. srropositiv? individuals. I’roc. Z’at, Acnrl. A??., C’.S.,d. 88. lOlR1-10137. Clac*kson. T., Hoogenboom. H. R.. Grifiths. A. I). KWinter. C:. (1991). Making antibody fragments using phagr display libraries. Snt/rrr I London). 352. 624-628. (‘urnatto. ~1. & Rajrusk? K. (1986). (‘lorral rr~ruitment generation Wld somatic, mutation it1 the of irrlmclnologic~al memory to the hapten XI’. I?,VNO .J. 5. 2459-2468. I)rrhyshirr. K. $1.. Salvo, .I. ,I. & (:rirtdlry. Iu. 1). F. (I 986). :I simple and efic%nt proc*edurr fot saturation mutagenesis using mixrtl trligorlu~lrvtitlrs. c:rnr. 46. 145-l . .5%. I)owrr, 11’. -1.. Miller. J. F. & Ragsdale. C’. \f:. (1988). High c~ffic~ienc~y tzransformation of E. colt’ by high voltage t,lrc~t,roporation. Xtr~l. Acids RPS. 16. 6l%i-6145. Foot,e. J. & Milsteirr. (‘. (1991). Kinetic maturation of an irn m une response. Sature (London J 352. T,30-:T,R?. Foote, .I. & \Vint,rr. (:. (1992). Antibody framework confbrmation wsidurs aflerting t,hr of the hyprrvariablr loops. d. A’ol. Biol. 224. 48G4QQ (:ibsort. ‘I’. (1984). Studies on the Epstein-Barr virus genome. Ph.D. thrsis. University of (:ambridgch. (ilasrr-Wuttkr. C;.. Keppner ,J. Br Rasched I. (1989). Pore-forming properties of the adsorpt,ion protein of filamrntous phagr fd. Riochinr. Riophys. Acta. 985. 239-247. (iray. C’.. Brown. II. 8r. Marvin. I). (l!+Xl). Adsorption c~~mplex of filament,ous fd virus. J. A’ol. Hiol. 146. 621-627. Hawkins. Ii.. E. Br Winter. G. (1992). (‘rll selection strategies for making antibodies from variable gene libraries: tapping the memory pool. Eur. J. In~n~unol. 22. 867--X70. Hrmsley. A.. Arnheim. N.. Tone\, M. I)., (‘ortopassi, (:. Hr (ialas. I>. .I. (I 989). A simple method for site-directed mutagenesis using the polymerasp chain reac%ion. ~Vucl. =1cids Res. 17, 65456551. Hoogenboom, H. R.. Griffiths, A. D., ,Johnson, K. S.. (‘hiswell. D. ,J., Hudson. P. & Winter. G. (1991 ). Multi-subunit proteins on the surface of filamentous phagr: rnethodologies for displaying antibody (Fat)) hoaq and light chains. Mucl. Acids RP.s. 19. 4133-4137. Hose. W. I).. Sastry. I,.. Tverson, S., Kang, A. S., Alt,ingMers. M., Burton. D. R., Benkovic, S. .J. & Lerner, R. A. (1989). (:eneration of a large combinatorial library of the immunoglobulin library in phage lambda. Science, 246, 1275-1281. .Iones. I’. T.. Dear, P. H., Foote. tJ., Neuberger, M. S. & Winter. (:. (1986). Replacing the complementaritydetermining regions in a human antibody with those from a mo&e. Nature (London), 321. 522-525. Kang, A. S.. Rarbas, (‘. F., Janda, K. D., Renkovic. S. .J. & Lcrner. R. A. (1991). Linkage of recognition and replicat~ion functions by assembling combinat,orial antibody Fab libraries along phage surfaces. Proc. Sol. =1cad. Sri.. I’.S.A. 88. 4363-4366.
Lrung. I). LIP.. (‘hrrr. b:. & (~c~rdtlel. I). V. (I%%)) A method for random mutagenesis of’ a dt+trrtl lJS.1 segment using a modified polyrnc~rasr c.ttain rc~;l,c~ti~,tr Trrhniqur. 1. 1 I -15. Lowmarr. H. I. .I, .\. (1991). Scalecating high afl?nit>, binding J)rotcains I)> Niochcnti.utry 30 monovalent phagr ctisplay. 10832~
108RX.
McCafferty. .I.. (~riffit~hs. A I).. Wint,rr. (;. & (‘hiswc~ll. I). ,I. (1990). I’hage ant)ibodirs: filamentous phage displaying aritibod> variablt domaitrh. ,v’nturr, (Londo,~). 348. 552-5.X .I. I).. Hoogenboom. H. I\‘.. Honnrrt, 7‘ I’.. Marks, M(~(‘af&rty. ,I. & (:rifIitha. A. I). (1991). By-passing itntntcnizwtion: human atttibodies from \--g,rrtlt. librarirh displayed on bac+rriophag,rr. .I .‘il
