INAT UBLlSHI'NG COMPANY
`
`
`
`
`
`JANUARY 1992
`"
`
`
`n',Ii1:¢HT\T()T n(;
`
`R4,;
`
`A
`
`THE INTERNATIONAL
`MONTHLY FOR
`INDUSTRIAL BIOLOGY
`
`
`‘[0911NNISNOOSIMA zennr.|.flflZ6L£bIID000#
`
`
`VNun..;1AVMISHEAINH
`
`
`SOIG
`
`
`
` ICOMMERCIIALIZING
`AGBIOTECH
`
`IN THE1990S
`
`
`
`T:
`
`-
`
`
`
`‘
`:
`
`
`HIV VACCINE TRIALS 8: PATENT BATTLES
`THE CASE FOR HIGH-DENSITY ANIMAL CELL CULTURE
`.CYSTIC FIBROSIS REGULATOR FROM MOUSE MILK $0
`FASTER DNA SEQUENCING IN ULTRA-THIN GELS
`ANTIBODIES HELP REFOLD PROTEINS
`9\O~\
`\$\<Z~
`
`¢
`Q
`
`
`
`I
`
`.
`
`BEQ1026
`
`
`
`BEQ 1026
`Page 1
`
`

`
`OWGY
`l FEBRUARY 1992
`
`VOLUME 10
`NUMBER 2
`
`Hypertexting through the
`Regulations • John Hodgson
`
`176
`
`Growth Enhancement in
`Transgenic Atlantic Salmon by
`the Use of an "All Fish"
`Chimeric Growth Hormone
`Gene Construct • Shao Jun Du
`eta/.
`
`High Resolution Removal of
`Virus from Protein Solutions
`Using a Membrane of Unique
`Structure • Anthony J. DiLeo
`eta/.
`
`Expression of Cloned
`Homologous Fermentative
`Genes in Clostridium
`acetolmtylicum ATCC 824
`• Lee D. Mermelstein eta/.
`
`Autolytic Lactococcus /actis
`Express ing a Lactococcal
`Bacteriophage Lysin Gene
`• Claire A. Shearman eta/.
`
`182
`
`190
`
`196
`
`READER SERVICE
`200
`
`Reader Service Card
`
`~01
`
`PCR and Other Amplification
`Techniques
`
`New Products
`
`Announcements
`
`204
`)(207
`CLASSIFIED
`209
`
`Recruitment • Edu-Guide
`• Venture Capital and
`Other Services
`
`THE LAST WORD
`216
`
`The Reality of Controversy
`• Jay D. Hair
`
`THE COVER, courtesy of Takeda Chemical
`Industries (Osaka) , shows the first crystal of
`interleukin-2, obtained in 1986. See page
`157 for an assessment of IL-2's potential in
`treating infectious diseases.
`
`146
`
`147
`
`148
`
`152
`
`Putting the bST Human-Health
`Controversy to Rest
`• Henry I. Miller
`
`bST & the EEC: Politics vs .
`Science • William Vandaele
`
`Signal Transduction:
`Untangling its Golden Promise
`• Stephen M. Edgington
`
`REVIEW
`157
`
`Rational Immunotherapy with
`lnterleukin 2
`• Gilla Kaplan eta/.
`
`RESEARCH PAPERS
`163
`
`High Level Escilericilia coli
`Expression and Production of a
`Bivalent Humani zed Antibody
`Fragment • Paul Carteret a/.
`
`169
`
`High-Leve l Expression of a
`Recombinant Antibody from
`Myeloma Cells Using a
`Glutamine Synthetase Gene as
`an Amplifiable Selectable
`Marker • C.R. Bebbington
`eta/.
`
`The Limiting Case
`
`Surprise is Research's Repri se
`
`THE FIRST WORD
`J 109
`COMMENTARY
`' 100
`IN THE NEWS
`x 112
`DATELINE
`\ 118
`'x 120
`
`Slew of Biotech Alliances
`
`Sandoz Buys into Systemix
`
`NIH Panel Against
`Gene Patents
`
`FDA Backlash Over Proposals
`
`Obstacles to Canadian Biotech
`
`Small Step for Biopesticides
`
`Companies Target Anxiety
`
`Jdentifying Microbes
`
`ATS: Devices and Desires
`
`121
`'<: 122
`123
`124
`124
`126
`K 127
`Products Emphasized at H & Q
`128 Gravity Doesn 't Slow Growth
`129
`ARTICLES
`x 132
`
`Biosensors Galore
`
`Genes in a Bottle
`• Mimi Bluestone
`
`137
`~
`141
`
`PCR: Catching the Next Wave
`• Stephen M. Edgington
`
`Selectable Markers Gene: Safe
`for Plants? • Richard B .. Flavell
`• Ed Dart • Roy L. Fuchs
`• Robert T. Fraley
`
`llio( l frhtwlog)· (ISSN 0 7~i :l-222X) is publish ed 111 0 111hl )' by
`N:uurc Jlublish ing Co .. hcadcpmrtcrcd a 1 65 lllccckcr S1..
`New Yo rk, NY 100 12-2467, which is owned br Nature
`America, Inc .. a subsidim")' of ~lac mill a n ~ l ag:uin cs of
`Lo ndo n.
`North American Advertising: !Jio/Tu lwolo[Jy. 65 Bl ccckn
`St. , New York, NY 1001 2-2467; pho ne (2 12) •177-9600. fax
`(2 12) 50!'r l 36•1. European Adverlisin9: Bio( Trdwofog)·. 4
`Liulc Essex St., Lo ndo n WC2 R 3 LF, U. K.: pho ne (07 1) 872-
`0 I 03. fax (07 I ) 2·10-2408, tel ex 262024. New Subscriptions,
`renewals, changes of address, back issues, and all customc.(cid:173)
`scrvicc questions sho uld be addressed w : /Jio/ Trrluwlom·.
`Suhsc ription Dc pl., P.O. llo x 1721. Rivcnon , N.J OH077-
`
`7:tll , o r call to ll free (800) !l24-0~28. Outside North
`America: /Jio/Fu lmolofO·· Circulation Ma nager. Mac millan
`~ l ag:at.i n es Ltd ., llruncl Rcmd. 13asittg:stokc. l-l ants RG21
`2XS. U. l{. Ann ual Subscription Rates: U.S. / C.ut;ula: U.S.
`S l98 (in stituti o nal/corpora te). U.S. S59 (i ndi,i du<~lmak­
`ing pe rso nal paym ent ); U. K.: £70:A ustr;llia: A 1;10; Rest of
`Wo rld: £70. Japan: Coni<~ C t K)"o iku~ha l.td .. Huura i 1\uild(cid:173)
`ing. 20. Nishishinj uku-1 cho mc . Shinjuku-ku. Tokyo 160 .
`.J;tpa n ; pho ne (03) ~\42- 104 ·1 ( for llio/ lf rluwloJO' With.Japa(cid:173)
`•wsc tt<tnsbtio n ) o r.J a pa n Puhlic:ations Trading Co. Ltd., 2-
`1 San ~ga ku-cl~ ~ ,} ;:h o_rn c, C hi>:~~l a-k~ t , To?·o. J? l)apa n:
`S!~;:~O: W~l, ~i~,,~~~~~i: ~~~~~~~~~~- ~i~-. ~;:.~~~~~;~i ~:~:
`
`$22.50. Annual Buyer-s' G ui de: U.S./Canada: U.S. 25; Rest
`of Wo rld: Surface
`.S. 27.50, Air ~ l ai l
`.S. 35. Reader
`Inquiry Sen.-icc: llio/ Trdmolog:y. N:nure ,\mnica, Inc..
`Reader Scr..-i ce Dep t.. P.O . Box 8902, Bo ulder, CQ 80328-
`~~~~;~:\~~:k~~~{~o{ci(~~t2J.~(~·: Re prints Dep t., 65 Bleecker
`POST~Ii\STER : SC IJ(I acldt·t~ss cl1a nges to /Jio/ f rrhuololQ'.
`Subsuiption Dept.. P.O. Box 172 1. Ri\"crton. NJ OBOi I·
`732 1. Second class post;tge paid at New York, NY I 001 2-
`2467 ;mel ;ul(li tion:ll mai ling olliceo;. Executive Officers o f
`Na ture Ame ri ca, Inc.: Nichol as 13)'01111 Shaw, Chainn;m o f
`the Board; And )' Suth e rland. Prt:sidcnt : Edward Valis.
`Secreta ry-Treasurer. Co p)•right © Natu re America, Inc.
`
`BEQ 1026
`Page 2
`
`

`
`RESEARCH
`IGH LEVEL ESCHERICHIA COLI EXPRESSION AND
`PRODUOION OF A BIVALENT HUMANIZED
`TIBODY FRAGMENT
`Paul Carter 1•*, Robert F. Kelley\ Maria L. Rodrigues\ Brad Snedecor2, Michael
`Covarrubias2
`3
`, Mark D. Velligan2 •
`, Wai Lee T. Wong\ Ann M. Rowland\ Claire E.
`Kotts5
`, Monique E. Carver5
`, Maria Yang6,James H. Bourell7 , H. Michael Shepard8
`and Dennis Henner6
`De partments of 1Protein Eng!n.eering, 2Cell Culture and Fermentation Research and Development, 4 lmmunological Research
`and Assay ~echnology, 5 MediCmal and Analytical Chemistry, 6Cell Genetics, 7Protein Chemistry and 8Cell Biology, Genentech
`Inc, 460 Pomt San Bruno Boulevard, South San Francisco, CA 94080. 3Current address: Gene Labs, 505 Penobscot Drive,
`Redwood City •. CA 94063. *Corresponding author.
`Many clinical uses of antibodies will re(cid:173)
`quire large quantities of fragments which
`are bivalent and humanized. We therefore
`attempted to generate humanized F(ab') 2
`fragments by secretion from E. coli. Titers
`of 1-2 g r• of soluble and functional Fab'
`fragments have been routinely achieved as
`judged by antigen-binding ELISA. Sur(cid:173)
`prisingly, this high expression level of
`Fab' in the periplasmic space of E. coli
`does not drive dimerization. However, we
`have developed a protocol to directly and
`efficiently recover Fab' with the single
`hinge cysteine in the free thiol state, al(cid:173)
`lowing F(ab') 2 formation by chemically(cid:173)
`directed coupling in vitro. The E. coli
`derived humanized F(ab') 2 fragment is in(cid:173)
`distinguishable from F(ab') 2 derived from
`limited proteolysis of intact antibody in its
`binding affinity for the antigen, pl85HER2,
`and anti-proliferative activity against the
`human breast tumor cell line, SK-BR-3,
`which over-expresses pl85HER2• This sys(cid:173)
`tem makes E. coli expression of bivalent
`antibody fragments for human therapy (or
`other uses) practical.
`
`F unctional monovalent rodent antibody Fv and
`
`chimeric Fab fragments have been obtained by
`co-secretion of their corresponding light and
`heavy chain fragments into the peri plasmic space
`of E. coli where chain assembly and disulfide bond forma(cid:173)
`tion is favored 1
`2
`. We have extended these pioneering
`•
`studies (reviewed in ref. 3) in three ways in an attempt to
`generate antibody fragments from E. coli suitable for
`human therapy. Firstly, our antibody fragment was pre(cid:173)
`viously humanized 4 using the strategy originally devel(cid:173)
`oped by Winter5
`7 in an attempt to improve its potential
`-
`clinical efficacy by reducing immunogenicity and perhaps
`9
`increasing serum half-life 8
`. Secondly, we have increased
`·
`the expression titers of functional fragments about 1000-
`fo ld from the milligram to gram per liter range. Thirdly,
`
`we have constructed bivalent F(ab') 2 fragments since this
`will be required for many clinical applications. For exam(cid:173)
`ple, cross-linking of cellular receptors may be required for
`biological activity (as in this study). Bivalent fragments
`may also be preferred over monovalent fragments be(cid:173)
`cause they have a significantly longer serum half-life 10
`and they can bind Rolyvalent or surface-bound antigens
`with higher avidity 1
`•
`Our chosen antibody, HuMAb4D5-8 (ref. 4) is a human(cid:173)
`ized version of MuMAb4D5 (refs. 12,13). HuMAb4D5-8
`binds the extracellular domain (ECD) of p 185"ER2 3-fold
`more tightly than the murine parent antibody and is
`almost as potent in specifically inhibiting the growth of
`human tumor cell lines over-expressing p185"ER2
`• Over(cid:173)
`expression of p185"ER2 is associated with multiple human
`malignancies and appears to be integrally involved in
`progression of 25-30% of primary human breast and
`15
`ovarian cancers 14
`•
`•
`Fab' fragments differ from Fab fragments by the addi(cid:173)
`tion of a few hinge region residues at the carboxy termi(cid:173)
`nus of the heavy chain CH 1 domain, including one or
`more cysteines. Fab' dimerization requires chance en(cid:173)
`counters of hinge cysteinyl thiols to form disulfide bonds
`without the assistance of extensive interactions between
`CH3 domains possible in the case of intact antibodies.
`Thus we attempted to express a Fab' fragment to very
`high titers in the periplasmic space of E. coli in the hope
`that this would drive formation of F(ab') 2 •
`RESULTS
`Vector design for Fab' expression. The plasmid,
`pAK19 (Fig. 1) is designed to co-express the light chain
`and heavy chain Fd' fragment of HuMAb4D5-8 from a
`synthetic dicistronic operon (Fig. 2) which is cloned be(cid:173)
`tween the EcoRI and Hindlll sites of pBR322 (ref. 16).
`The operon is under the transcriptional control of the E.
`coli alkaline phosphatase (phoA) promoter 17 which is in(cid:173)
`ducible by phosphate starvation. Each antibody chain is
`preceded by the E. coli heat-stable enterotoxin II (stll)
`signal sequence 18 to direct secretion to the periplasmic
`space of E . coli. Th~ humanized variable domains (V Land
`V H) 4 are precisely fused on their 3' side to human K 1 CL
`(ref. 19) and IgG 1 CH 1 (ref. 20) constant domains respec(cid:173)
`In order
`to express
`the Fab'
`fragment. of
`tively.
`HuMAb4D5-8 it was necessary to extend the CHI gene
`segment to encode part of the cysteine-containing anti(cid:173)
`body hinge region. A hinge sequence containing a single
`cysteine was chosen (CysAlaAla) to avoid intra-hinge
`disulfide bonding which may occur with hinges containing
`multiple cysteines21
`. The Sal! and Sphl sites within the
`
`810/TECHNOLOGY VOL 10 FEBRUARY 1992
`
`163
`
`BEQ 1026
`Page 3
`
`

`
`EcoRl
`
`Amp'
`
`pAK19
`(6420 bp)
`
`Hindll1
`
`ori
`FIGURE I Schematic representation of plasmid pAK19 containing
`the discistronic operon for expression of HuMAb4D5-8 Fab'.
`See Results and Figure 2 for additional details.
`
`tetracycline resistance gene (Tel') were removed by site(cid:173)
`directed mutagenesis without altering the amino acid
`sequence of the corresponding gene product (see Exper(cid:173)
`imental Protocol). Construction of different Fab' variants
`by cassette mutagensis22 of pAK 19 was facilitated by
`installing unique Sali and Sphi sites flanking the end of
`the CH 1 gene. The CH 1 gene is immediately followed by
`the strong bacteriophage 'A t0 transcriptional terminator
`(ter) 23 in an attempt to improve mRNA stability and to
`prevent transcriptional read-through into the Tetr gene.
`However this is not an important feature of pAK19 since
`installation of the terminator had no significant effect
`upon the expression titers of HuMAb4D5-8 Fab' (not
`shown).
`Fab' expression titers. HuMAb4D5-8 Fab' was ex(cid:173)
`pressed in E. coli strain 25F2 transformed with plasmid
`pAK19 grown to high cell density (120 to 150 OD55 0 ) in a
`10 I fermentor (see Experimental Protocol). The titer of
`cell-associated soluble and functional Fab' is routinely 1 to
`2 g 1- 1 as judged by antigen binding ELISA 4 of crude
`samples obtained by fully disrupting cells by sonication .
`Only very modest quantities of Fab' (s 100 mg 1- 1
`) are
`found in the culture media under these fermentation
`conditions which are optimized for high titers of func(cid:173)
`tional cell-associated Fab'. The H uMAb4D5-8 Fab' frag(cid:173)
`ment was found to bind tightly to both Staphylococcal
`protein A and to Streptococcal protein G allowing their
`use for affinity purification . Very similar estimates of titers
`were obtained for culture media or cell paste samples
`after affinity purification on Staphylococcal protein A
`prior to antigen-binding ELISA.
`Preparation of F(ab') 2 • Despite these extraordinarily
`high Fab' titers only trace quantities of antibody fragment
`were recovered as the disulfide-linked F(ab') 2 from either
`the cell pastes or culture supernatants. In contrast, the
`relatively weak disulfide bond between light and heavy
`chains24 is apparently quantitatively formed as judged by
`
`TOLE I Analysis of HuMAb4D5-8 F(ab')2 fragment by p185" •• 2
`ECD binding affinity and anti-proliferative activity with breast
`carcinoma, SK-BR-3 cells.
`
`Source
`
`E. coli
`293 cells<
`
`290
`300
`
`Relative cell
`proliferation b
`
`53
`50
`
`•Kd values for the p185"••• ECD were determined as previously
`described 4 and the standard error of the estimates are :S ± 10%.
`bProliferation of SK-BR-3 cells
`incubated for 96 h with
`HuMAb4D5-SH F(ab') 2 shown as a percentage of the untreated
`control as described•. Data represent the maximal anti-prolifer(cid:173)
`ative effect calculated as the mean of triplicate determinations at
`a F(ab')2 concentration of 10 f.Lg ml- 1. Data are taken from the
`same experiment in which the maximal anti-proliferative effect
`of the murine parent antibody, MuMAb4D5, was 47% and the
`standard error is
`:S ± 15 %. <Derived from limited pepsin
`digestion 45 of intact antibody 4
`.
`•
`
`SDS-PAGE (Fig. 3). We therefore recovered Fab' mole(cid:173)
`cules from fermentation pastes with the hinge cysteine
`present as the free thiol and used the directed chemical
`coupling method of Brennan et al. 21 to form the F(ab') 2 in
`vitro. This was achieved by affinity purification ofFab'-SH
`on Streptococcal protein Gat pH 5 to maintain the thiol in
`the less reactive protonated form and in the presence of
`EDT A to chelate metal ions capable of catalyzing disulfide
`bond formation. Simple osmotic schock (see Experimental
`Protocol) was inefficient in releasing Fab' molecules fro m
`cell pastes. However virtually quantitative release of solu(cid:173)
`ble Fab' from cell pastes (as judged by Western blotting)
`was achieved by partial digestion of the bacterial cell wall
`without extensive cell lysis using chicken egg white lyso(cid:173)
`zyme. The yield of Fab'-SH after protein G purification
`was 2: 50% of that detected by antigen-binding ELISA 4 of
`cells fully disrupted by sonication. The Fab' preparations
`are routinely 75 to 90% in the reactive free thiol form
`(Fab'-SH) as determined by reaction with 5,5'-dithiobis(2-
`nitrobenzoic acid) (DTNB 25
`) to form the thionitroben(cid:173)
`zoate derivative (Fab'-TNB). In contrast, no free thiol was
`found associated with Fab' purified from culture super(cid:173)
`natants suggesting that the hinge cysteine had been either
`covalently modified (eg by oxidation or formation of
`disulfide-linked adducts with small molecules) or removed
`by proteolytic degradation.
`Equimolar quantities of Fab'-SH and Fab'-TNB were
`coupled together at pH 7.5 in the presence of. EDT A to
`form the F(ab') 2 fragment by a disulfide exchange reac(cid:173)
`tion (see Experimental Protocol). The coupling reaction
`was followed by monitoring the increase in absorbance at
`412 nm upon release of the thionitrobenzoate anion and
`found to reach completion after 30 to 60 minutes at 37°C.
`No free thiol was detected in the reaction mix after
`coupling. F(ab') 2 was separated from Fab' by size exclu(cid:173)
`sion chromatography. The coupling efficiency was esti(cid:173)
`mated as 82% based upon the integrated chromatograph
`peak areas for Fab' and F(ab') 2 (not shown) plus the free
`thiol content of Fab'-SH (84%) and TNB content of
`Fab'-TNB (77%). Only trace quantities of HuMAb4D5 -8
`F(ab') 2 were formed in mock coupling reactions contain(cid:173)
`ing either Fab'-SH or Fab'-TNB alone as anticipated 21
`(Fig. 3).
`Physical and chemical characterization of F(ab') 2•
`Purified HuMAb4D5-8 F(ab') 2 analyzed by SDS-PAGE
`under non-reducing conditions (Fig. 3) gave a single
`major band of the expected mobility (Mr -96 kD). After
`reduction with dithiothreitol and SDS-PAGE a doublet of
`bands of similar intensities was observed as expected from
`release of a stoichiometric quantity of free light chain (Mr
`-23 kD) and heavy chain Fd' fragment (Mr -25 kD) (not
`shown). The molecular masses of F(ab') 2 and reduced
`light and heavy chain Fd' fragments were precisely deter(cid:173)
`mined by high resolution electrospray mass spectrometry
`(see Experimental Protocol) as 95961±10, 23442±3 and
`24556±6. These data are in excellent agreement with the
`expected masses of 95965, 23443 and 24551 D, respec(cid:173)
`tively. Amino terminal sequence analysis of F(ab') 2 (8
`cycles) after electroblotting on to polyvinylidene difluo(cid:173)
`ride membrane 26 gave the expected mixed sequence fro m
`a stoichiometric 1: 1 mixture of light and heavy chains
`(VdVH:DIE, IIV, Q/Q, MIL, TIV, QIE, SIS and PIG) with
`no evidence of additional sequences. Correct folding of
`the F(ab') 2 fragment is strongly suggested by the far UV
`circular dichroism spectrum which is characteristic of the
`immunoglobulin fold 27 plus the high thermal stability
`(T m=82oC) determined by differential scanning calorim(cid:173)
`etry28 (R.K., unpublished result) and was confirmed by
`X-ray
`crystallographic
`structure determination of
`HuMAb4D5 Fab and Fv fragments (C. Eigenbrot, unpub-
`
`BEQ 1026
`Page 4
`
`

`
`lished data).
`Functional characterization of F(ab') 2 • The function of
`HuMAb4D5-8 F(ab'h was shown by measurement of
`antigen binding and anti-proliferative activity against the
`p185tiER2 over-expressing human breast carcinoma line,
`SK-BR-3 (Table 1). The binding affinity of the F(ab') 2
`fragment for p185HER2 ECD is identical to that of the
`corresponding fragment derived from limited proteolysis
`of whole antibody expressed in mammalian cells. Further(cid:173)
`more the F(ab')2 fragment binds 1.98±0.05 molecules of
`p185 HER2 ECD as judged by titration calorimetry (not
`shown) as anticipated for a bivalent molecule. The F(ab') 2
`fragment derived from E. coli has identical anti-prolifer(cid:173)
`ative activity with SK-BR-3 cells to both the intact bivalent
`Hu MAb4D5-8 parent antibody derived from 293 cells4
`and F(ab') 2 fragment derived from limited pepsin diges(cid:173)
`tion of intact antibody (Table 1). Furthermore the human(cid:173)
`ized F(ab')2 fragment is virtually as potent as the murine
`parent antibody, MuMAb4D5, in its anti-proliferative
`activity and significantly more potent than eight other
`murine anti-g185HER2 monoclonal antibodies that we have
`3
`examined 12
`• In contrast, the monovalent Fab fragment
`•
`does not significantly affect the growth of SK-BR-3 cells
`but binds p185HER2 ECD with similar affinity to the F(ab') 2
`fragment (not shown). This suggests that crosslinking of
`p185HER2 on the surface of cells may be required for
`in hibiting their proliferation, and underscores the impor(cid:173)
`tance for a bivalent form of the antibody.
`
`DISCUSSION
`T he gram per liter titers of functional HuMAb4D5-8
`Fab' that we have obtained represents a 1000-fold in(cid:173)
`crease over levels reported for the simpler rodent Fv and
`chimeric Fab fragments 1
`2
`. This higher titer is due largely
`•
`to higher cell densities (> 100-fold higher than in ref. 1)
`resulting from the more precisely controlled environment
`of the fermentor than the simple shake flask. Very tight
`control of expression prior to induction also plays a
`crucial role in achieving high cell densities and thus high
`expression titers since E. coli constitutively expressing the
`corresponding humanized Fab fragment grew very poorly
`and could not be stably maintained in liquid cultures
`(P. C., unpublished observation). The humanized variable
`domains contribute significantly to the high titers since
`replacement with corresponding domains from the mu(cid:173)
`rine parent antibody reduced expression titers by 100-
`fold under the same fermentation conditions and by
`10-fold under conditions optimized for expression of the
`chimeric fragment (R.K., B.S. and M.C., unpublished
`data). These dramatic differences in expression titers
`between humanized and chimeric molecules probably
`reflect many factors, which may include differences in the
`relative expression levels of light and heavy chains, kinet(cid:173)
`ics of membrane transport, folding and association of
`light and heavy chains plus competing light chain dimer(cid:173)
`ization.
`T he Fab' fragment of HuMAb4D5-8 may be a gener(cid:173)
`ically useful template for high level E. coli expression of
`humanized Fab' molecules since high expression titers
`to 600 mg 1- 1
`(400
`) were obtained by using
`the
`Hu MAb4D5-8 Fab' framework to display specificities
`against CD3 (ref. 29) or CD18 (P.C., B.S., M.C., L. Presta,
`7
`unpublished data) by CDR grafting5
`. A similar reduc(cid:173)
`•
`tion in expression titers was also observed between the
`humanized and chimeric versions of the anti-CD3 Fab' 29
`•
`H uMAb4D5-8 Fab' shows very little tendency to form
`F(ab') 2 in vivo despite the very high expression titers and
`quantitative formation of the weak disulfide between light
`chain and heavy chain 2 4 as judged by SDS-PAGE (Fig. 3).
`Nevertheless it has been possible to directly recover func-
`
`tiona) Fab' fragments secreted into the periplasmic space
`of E. coli with the unpaired hinge cysteine predominantly
`(75 to 90%) in the free thiol form . These data suggest that
`the redox potential of the periplasmic space of E. coli is
`sufficiently oxidizing to favor formation of the disulfide
`between light and heavy chains (plus the stronger intra(cid:173)
`domain disulfides) but is insufficiently oxidizing to drive
`formation of the weaker inter-heavy chain disulfide
`bond 24
`. The high concentration of Fab' in the peri plasmic
`space may result in the free hinge cysteinyl thiol itself
`perturbing the redox potential by titrating the redox
`components of the peri plasmic space.
`Generating Fab'-SH fragments by direct expression in
`E. coli without need for a reduction step obviates the
`inherent problems in obtaining them from intact antibod(cid:173)
`ies: differences in susceptibility to proteolysis and non(cid:173)
`specific cleavage resulting in heterogeneity, low yield, as
`
`FIGURE 2 Dicistronic operon for expression of HuMAb4D5-8
`Fab' in plasmid pAKI9 . TheflwB binding box (PB)42 and -10
`regions of the phoA promoter 1 are underlined as is the ribosomal
`bind ing site (SD) preceding each chain which is required for
`efficient initiation of translation, and the bacteriophage A t0
`transcriptional terminator (ter)23
`. The humanized variable do(cid:173)
`mains (V Land V H) 4 are precisely fused on their 5' ends to a gene
`segment encoding the heat stable enterotoxin II (stii) signal
`sequence 18 shown in lower case and on their 3' side to human K 1
`CL (ref. 19) and IgG I CHI (ref. 20) constant domains respec(cid:173)
`tively. The complementarity determining region (CDR) residues
`according to the hypervariable sequence definition43 and a struc(cid:173)
`tural definition 44 are underlined and overlined, respectively.
`EcoRI
`Hindiii
`1 ~TCAACTTCTCCATAC.TM'GGATAAGGAAATACAGACATGA.AAAATCTCA'T'TGCIGAGTTGM'A'M'TA:i\GCT
`HindU I
`7 6 TGCCCAAAAAGAAGAAGAGTCGAATGAA.CTGI'GTGCGCAGGT'A~TITGGAGATTATCGTCACT'GCAATGCT
`
`151 TCGCAATATGGCGCAAAATGACCAACAGCGGM'GA'I'TGATCAGGTAGAGGGGGCGCTG'l"ACGAGGTAAAGCCCGA
`
`2 2 6 TGCCAGCATTCCTGACGACGATACGGAGCTGC'rGCGCGATTACGT'AAAGAAGrTA'I"''GAA.GCATCCTCGT'CAGI'A
`
`r-- """"'
`
`301 AAAAG'M'AATC'T'TTI'CAACAGCTGTCATAAAG'I'TGTCACGGCCGAGACTTATAGTCGCTTTG'TTT'TTA'l""!"l"rlTA
`--=10
`PB
`EcoRI
`3 7 6 ATG'l'A'T'TTG'l'AACTAGAATICGAGCTCGG'I'ACCCGGGGATCCTCTAGAGGTTGAGG1"GA.TTTTATGAAAAAGAAT
`SO
`~t.I~ +-k
`n
`-23
`
`4 51 A TCGC.\T1'"I'CTT'C'M'GTCTATG1'"1'CGTT'MTT-c'I'ATTGCTACAAACGCGT'ACGCTGATATCCAGATGACCCAG
`i
`i
`f
`f
`f
`a
`a
`s m
`v
`s
`a
`n
`a
`y
`a D
`-19
`1
`1
`I
`11 T Q
`t
`Q
`Vc +
`
`526 TCCCCG.\GCTCCC1'G'I'CCGCC'I'CI'GI'GGGGATAGGGTCACCATCACCTGCCGTGCCAGTCAGGATGTGAATACT
`7 S P S S L S A S V G
`0 R V T
`I
`T C R A
`S 9
`0 V N T
`CDR- Ll
`601 s.i£IGTAGCCTG...I'ATCAACAGAAACCAGGAAAAGCTCCGAAACTACTGATTTA~CT'TCCTCTAC'TC'T'
`P G K A
`P
`K
`L
`L
`I
`Y S A
`S
`F
`L
`Y
`S
`32 A V A W Y Q Q K
`- - -
`COR-L2
`
`67 6 GGAGTCCCT'TCTCGC'T'I'C'I'CTGGATCCAG.\TCTGGGACGGATTTCAC7CTGACCATCAGCAGTC'TGCAGCCGGAA
`57 G V
`P
`S R
`F S G S R S G
`T
`0
`F
`T
`L
`T
`I
`S S
`L
`0
`P E
`
`7 51 GACTTCGCAACTTA'M'ACTGTCAGCAACATTATACTACTCCTCCCACGTTCOOACAGGGTACCAAGGTGGAGATC
`82
`0
`F A T Y Y C
`0
`9 H Y T
`T
`P
`P
`T
`F G Q G T K V
`E
`I
`CDR-L3
`826 AAACGAACTGTGGCTGCACCATCTGTCTTCATCTI'CCCGCCATCI'GATGAGCAGTTGAAATCTGGAACTGCCTC'I'
`107 K R T
`V A A
`P S V
`F
`I
`F
`P P
`S
`0
`E Q
`L
`K
`S G
`T
`A
`S
`CL -~
`
`901 GTI'GTGTOCC'I"GCTGAATAAC7'TCTATCCCAGAGAGGCCAAAcrfACAGTGGAAGGTGGATAACGCCCTCCAATCG
`132 V V C
`L
`L N N
`F Y
`P
`R
`E A K V Q W K V
`0 N A
`L Q
`S
`
`97 6 GGTAACTCCCAGGAGAGTGTCACAGAGCAGGACAGCAAGGACAGCACCTACAGCC"TCAGCAGCACCCTGACGCTG
`T
`E Q
`0
`S K
`0
`S
`T
`Y
`S
`L
`S S
`T
`L
`T
`L
`157 G N S Q E S V
`
`1051 AGC.AAAGC.AGAC7'ACGAGAAACACAAA.GfCTACGCC'TGCGAAGTCACCCATCAGGGCC7GAGCTCGCCCGTCACA
`182 S K A D Y
`E K H K V Y A C
`E V
`L
`S S P V
`T
`T H Q G
`
`112 6 AAGAGCM'CAACAGGGGAGAGTGT'I'AAGCTGATCCTCTACGCCGG.ACGCATCGTGGCCCTAGTACGCAAGTTCAC
`207 K
`S
`F N R G
`E C Oc
`
`1:~; GTAAAAAGGGTATCTAGAGG'i'TGA~'M"M'A~G~TAl_'C~A~A~A~~C~
`stii _._
`
`127 6 'IT"'TCTA'M'GCTACAAACGCGTACGCTGAGG'ITCAGCt'GG"''GGA.GTCTGGCGGTGGCCTGGTGCAGCCAGGGGGC
`s
`a
`n
`a
`a
`i
`f
`y
`-9
`E V
`0
`L
`V
`E S G G G
`L
`V Q
`P G G
`t
`VH +
`
`1 351 TCACTCCGT'T'TGTCCTGTGCAGCT'l'CTG§CTI'CMCATfMbG,A,CACCTATATACACTGGGTGCGTCAGGCCCCG
`K p T X
`17 S
`L R
`L
`S C A A
`S G
`N
`I
`I
`H
`\.>1
`P
`V R Q A
`f
`COR- Hl
`14 2 6 GGTAAGGGCCTGGA.\TGGGT'TGCAAGGA'T'M'ATCCTACGAATGG'M'ATACTAGATATGCCGATAGCGTCAAGGGC
`T N G X T
`42 G
`K G
`L
`E H V A
`B
`I
`X P
`R Y
`A
`0
`S V
`K G
`CDR -H2
`1501 CGTTTCACTATMGCGCAGACACA'J'CCAAA.AACACAGCCTACCI'GCAGATGAACA.GCCTGCGTGC1'GAGGACACT
`67 R
`F
`T
`I
`S A
`0
`T
`S K N
`T
`A
`Y
`L
`0 M N
`S
`L R A
`E
`0
`T
`
`157 6 GCCGTC'TATIA'ITGTTCT AGATGGOOAGG'"'..GACGGCTTCTATOC'fATGGACTAC'l'GGGGTC.A.AGGAACCCTGGTC
`92 A V
`Y Y C
`S R H G G
`0 G
`F
`Y A N
`0
`Y
`T
`L
`V
`\'./ G Q G
`CDR-H3
`1651 ACCGTCTCCTCGGCCTCCACCAAGGGCCCATCGGI'C'M'CCCCCT'GGCACCCTCCI'CCAAGAGCACCTCTGGGGGC
`11 7
`T
`V
`S S A S
`T
`K G
`P S V
`F
`P
`L
`A
`P
`S S K
`S
`T
`S G G
`CHl • ._
`
`172 6 ACAGCGGCCC1"GGGCTOCCTGGI'CAAGGACTACTTCCCCGAACCGG1'GACGGTGI'CG'l'GGAACTCAGGCGCCCTG
`142 T
`A A
`L G C
`L
`V
`K D Y
`F P E P V
`T
`V
`S H N
`S G A
`L
`
`1801 ;..ccAGCGGCGTGCACACCTICCCGG....tGTCCTACAGTCCTCAGGACTCTACTCCCTCAGCAGCGTGGTGACCGTG
`L Q S S G
`167 T
`T
`F
`L
`Y
`S
`L
`T
`V
`S G V H
`P A V
`S S V V
`Sdli
`187 6 CCCTCCAGCAOC'M'GGGCACCCAGACCTACATCTGCAACGTGAATCA CAAGCCCAGCAACACCAAGGTCGACAAG
`192 P
`S S
`S
`L G
`T
`Y
`I
`C N V N H
`K
`P S N
`T
`K V D K
`T Q
`
`5bh!
`19 51 AAAGT'TGAGCCCAAATC'M'GTGACAAAA.c:rCACACA TGCOCCGCGTGACGCOOCATOCGACGGCCCTAGAGTCCC
`2 1 7
`K V
`E P K
`S C
`0
`K
`T H
`T C
`A A Op
`
`Hindiii
`202 6 TAACGCTCGG'IT-GCCGCCGGGCGT1"T'M'TA T'TGTTAACTCATGT'TTGACAGCIT ATCATCGATAA'GCTT
`
`BEQ 1026
`Page 5
`
`

`
`k0123456769
`
`---
`------
`
`-+f(ab\
`
`-+ Fab', Fab
`
`92
`66
`45
`31
`21
`14
`
`FIGURE 3 Purification of F(ab')2 fragment of HuMAb4D5-8.
`Proteins (2 fLg per sample) were reacted with 4 mM iodoaceta(cid:173)
`mide, analyzed by SDS-PAGE on a 4 to 20% gel (BioRad,
`Mini-PROTEAN II ready gel) under non-reducing conditions
`and stained with Coomassie brilliant blue '(R250) . Fab'-SH (lane
`3) was coupled with Fab'-TNB (lane 5) to form F(ab')2 (lane 7)
`which was then purified by size exclusion chromatography (lane
`8) and compared with F(ab')2 derived from limited pepsin
`digestion44 of full length HuMAb4D5-8 expressed in mamma(cid:173)
`lian cells4 (lane 9). Also shown are molecular weight standards
`(lane 1), HuMAb4D5-8 Fab (lane 2), and mock coupling reac(cid:173)
`tions containing either Fab'-SH (lane 4) or Fab'-TNB (lane 6)
`alone.
`
`well as partial reduction which is not completely selective
`for the hinge disulfide bonds24 . Furthermore, by engi(cid:173)
`neering the hinge region to leave a single cysteine residue
`we prevent intra-hinge disulfide bonding without resort(cid:173)
`ing to the use of highly toxic arsenite to chelate vicinal
`thiols 21
`•
`E. coli-derived HuMAb4D5-8 F(ab'h has been gener(cid:173)
`ated by a variety of different strategies. Here we have
`focused upon coupling of Fab'-SH using the DTNB(cid:173)
`directed method of Brennan et al. 21 This approach was
`chosen because our next step was to generate bispecific
`F(ab') 2 fragments between HuMAb4D5-8 Fab'-SH and a
`humanized anti-CD3 antibody29 where a directed-cou(cid:173)
`is essential. We have also generated
`pling method
`thioether-linked F(ab') 2 by directed coupling of Fab'-SH
`using N,N'-1,2-phenylenedimalemide (o-PDM) as devel(cid:173)
`oped by Glennie et al. 30 The thioether linkage is chemi(cid:173)
`cally more stable than a disulfide and may prolong half(cid:173)
`life of the F(ab')2 in vivo31
`• Monospecific HuMAb4D5-8
`F(ab') 2 is most conveniently generated by coupling the
`Fab'-SH to itself in the presence of half a molar equiva(cid:173)
`lent of either DTNB or o-PDM (not shown) . Yet another
`efficient route to HuMAb4D5-8 F(ab') 2 is by oxidation of
`Fab'-SH in vitTO at pH 7.5 in the absence of EDTA
`presumably assisted by dissolved oxygen and adventitious
`metal ions. Coupling occurs at modest concentrations of
`Fab' (:s 1 mg ml- 1) that are 10 to 100-fold lower than
`those found in vivo (not shown). In contrast to Fab'
`containing a single hinge cysteine, a HuMAb4D5-8 Fab'
`variant containing the hinge sequence CysProProCys
`found in human IgGI formed up to 25% (by mass) of the
`bivalent F(ab') 2 fragment in vivo. This was estimated by
`analytical size exclusion chromatography after affinity
`purification on protein G and blocking any free thiol by
`reaction with iodoacetic acid. Thus expression of Fab'
`CysProProCys provides a direct route to the secretion of
`functional bivalent F(ab')2 fragments in E. coli, albeit a
`relatively inefficient one.
`Another potential application of Fab'-SH fragments is
`in the field of protein purification. The possibility of using
`antibodies for one step purification has long been recog(cid:173)
`nized (reviewed in ref. 32). Unfortunately immunoaffinity
`purification has only rather rarely been useful because it is
`arduous, time-consuming and expensive to obtain large
`amounts of suitable antibodies and immobilize them on
`appropriate matrices. Our system for high level E . coli
`expression of Fab'-SH provides a solution to most of
`these problems. For example, we have exploited the free
`hinge cysteinyl thiol for site-specific immobilization of
`HuMAb4D5-8 Fab'-SH on an activated thiol support as
`we have _previously described for an engineered subtilisin
`
`variant3 3
`, enabling the p185HER:l ECD to be affinity puri(cid:173)
`fied from solution (M.R., P.C., unpublished data).
`Fab'-SH fragments are also likely to be useful for
`site-directed attachment of radionuclides, (pro)drugs, tox(cid:173)
`ins or other reactive moieties for imaging or therapy. For
`example, site-specific of attachment of phospholipid to
`Fab'-SH has been used for immunospecific targeting of
`liposomes to cells34
`. This has clear advantages over ran(cid:173)
`dom attachment which generates heterogeneous popula(cid:173)
`tions of molecules with different physicochemical and
`biological properties and often results in reduced specific
`binding to antigen.
`EXPERIMENTAL PROTOCOL
`Vector construction. Plasmid pAK 19 (Fig. I) was constructed
`by subcloning35
`the dicistromc operon for expression of
`HuMAb4D5-8 Fab' (Fig. 2, ref. 29) into P.BR322 (ref. 16)
`restricted with Hindiii plus EcoRI and is available for research
`use upon request to the authors. The Sphl and Sail sites in the
`Tetr gene of pBR322 were removed by site-direct