Proc. Natl. Acad. Sci. USA
`Vol. 92, pp. 11907-11911, December 1995
`Biochemistry
`
`Expression studies of catalytic antibodies
`HELLE D. ULRICH, PHILIP A. PATIEN, PRISCILLAL. YANG, FLOYD E. ROMESBERG, AND PETER G. SCHULTZ*
`Howard Hughes Medical Institute, Department of Chemistry, University of California, Berkeley, CA 94720
`
`Contributed by Peter G. Schultz, August 31, 1995
`
`ABSTRACT We have examined the positive influence of
`human constant regions on the folding and bacterial expres(cid:173)
`sion of active soluble mouse immunoglobulin variable do·
`mains derived from a number of catalytic antibodies. Expres·
`sion yields of eight hybridoma· and myeloma-derived chimeric
`Fab fragments are compared in both shake flasks and high(cid:173)
`density fermentations. In addition the usefulness of this
`system for the generation of in vivo expression libraries is
`examined by constructing and expressing combinations of
`heavy and light chain variable regions that were not selected
`as a pair during an immune response. A mutagenesis study of
`one of the recombinant catalytic Fab fragments reveals that
`single amino acid substitutions can have dramatic effects on
`the expression yield. This system should be generally appli·
`cable to the production ofFab fragments of catalytic and other
`hybridoma-derived antibodies for crystallographic and struc·
`tore-function studies.
`
`Many examples of antibody-catalyzed reactions have been
`reported, ranging from efficient acyl transfer and redox reac(cid:173)
`tions to stereoelectronically disfavored pericyclic and cycliza(cid:173)
`tion reactions (1, 2). In addition to providing a strategy for
`generating catalysts for reactions difficult to achieve by exist(cid:173)
`ing enzymatic or chemical methods, antibody catalysis pro(cid:173)
`vides a tool to gain increased insight into the mechanisms of
`biological catalysis and the evolution of catalytic function.
`These latter aims are facilitated by the availability of three(cid:173)
`dimensional structures of antibody active sites, coupled with
`mutagenesis experiments to confirm hypotheses about reac(cid:173)
`tion mechanisms. Unfortunately, such studies have been ham(cid:173)
`pered by difficulties in expressing many of the antibodies in
`recombinant form in quantities sufficient for mechanistic and
`crystallographic studies. In addition, the ability to express
`antibodies or antibody fragments in a bacterial system in
`soluble form, either periplasmically or cytoplasmically, is a
`prerequisite for attempts to improve their catalytic efficiency
`using random or directed mutagenesis, coupled with in vivo
`selection techniques (3, 4).
`A number of systems for the production of antibody frag(cid:173)
`ments have been described, most of which use bacterial signal
`sequences to direct the immunoglobulin chains to the
`periplasm or culture medium, where correct folding and
`disulfide bond formation are possible (for review, see ref. 5).
`However, reports of expression systems rarely describe their
`general applicability. For example, we have examined a variety
`of published systems by using a number of hybridoma- and
`myeloma-derived antibodies, including Fab and single-chain
`Fv fragments, various promoters [T7, lac, tac, and alkaline
`phosphatase (PhoA)], signal sequences (pelB and stll), fusions
`to maltose binding protein and ThiA ( 6), exchange of the
`constant (C) regions, and expression as a single-chain Fv
`fragment in the yeast Pichia pastoris (7) (ref. 8 and H.D.U. and
`P.G.S., unpublished results). In each case, the majority of the
`protein was found in insoluble aggregates, suggesting that the
`primary structure of the antibody greatly influences its correct
`
`folding in the periplasm and remains the limiting factor for its
`performance in a given expression system. The extensive study
`of the antibody McPC603 by Knappik and Pliickthun (9)
`supports this notion. A number of antibodies have been
`derived from phage display methods that express well and in
`soluble form (10). However, the well-behaved nature of these
`antibodies is likely a consequence of concurrent selection for
`correctly folded proteins and may not apply to hybridoma(cid:173)
`derived antibodies.
`One of the earliest reports of the expression of soluble
`antibody fragments in Escherichia coli was based on a chimeric
`Fab fragment (11). Carter and coworkers (12-14) have ob(cid:173)
`served large increases in expression yields of three humanized
`Fab fragments over the corresponding murine proteins; where
`reported, the murine-human chimeric Fab fragment was
`intermediate (12). Based on these observations, we have
`examined the expression of a panel of catalytic antibodies in a
`system that exploits this beneficial influence of the human
`sequences on the folding of the Fab fragment in the bacterial
`periplasm. Herein we report the expression studies of both
`hybridoma-derived antibodies and artificially constructed
`combinations of heavy and light chains and report the effects
`of point mutations and growth conditions on yields. This
`system appears to be generally applicable for the facile pro(cid:173)
`duction of useful quantities of active soluble mouse-human
`chimeric Fab fragments.
`
`MATERIALS AND METHODS
`Strains and Plasmids. E. coli JM109 (15) was used for
`cloning, and the strain 25F2 (12), obtained from D. Henner
`(Genentech), was used for expression. Plasmids pMY61,
`pMY60, and pMY55 were gifts from M. Yang and D. Henner
`(Genentech). pMY61 contains an expression cassette for the
`humanized Fab fragment of antibody D1.3 (16), similar to that
`ofpAK19 (12), as anEcoRI-Sph I fragment in pBR322. Minor
`differences exist upstream of the ribosome binding sites.
`pMY60 and pMY55 contain the isolated heavy and light chains
`in the same context.
`Vector Construction and Cloning of Antibody Genes. Prim(cid:173)
`ers H1-H12 and Ll-L9 have been described by Huse et al. (17).
`The antibody 48G7 had been cloned (8). The 2E11 (18) and
`S107 (19) murine Fab fragments were cloned into the same
`vector from their cell lines analogously: total RNA was isolated
`from = 108 cells by standard methods (20), mRNA was purified
`by oligo(dT) affinity chromatography (15) with a kit from
`Pharmacia, eDNA was generated with primers L9 and H12 for
`2E11 or L9.and IgAH3 for S107 (the latter is designed to match
`the 3' end of the IgA heavy chain C region C8 1 sequence:
`5'-GTAATAGGACTAGTAGGAGTAGGACCAGA-3'),
`followed by PCR amplification with primers HZ, H12, L5, and
`L9 for 2E11 and IgAH3, L9, and primers designed to match the
`published variable [ K chain and heavy chain variable region
`(respectively, V K and V 8 )] sequences for S107. The VH and V K
`
`The publication costs of this article were defrayed in part by page charge
`payment. This article must therefore be hereby marked "advertisement" in
`accordance with 18 U.S.C. §1734 solely to indicate this fact.
`
`Abbreviations: PhoA, alkaline phosphatase; V, variable; C, constant;
`J, joining; H (as subscript), heavy chain; K (as subscript), K chain;
`CDR, complementarity-determining-region.
`*To whom reprint requests should be addressed.
`
`11907
`
`BEQ 1042
`Page 1
`
`

`
`11908 Biochemistry: Ulrich et a/.
`
`Proc. Nat/. Acad. Sci. USA 92 (1995)
`
`regions were then moved into pMY60 and pMY55, respec(cid:173)
`tively, by PCR amplification with primers that produced Mlu
`1-BstEII (for 48G7: Mlu 1-Apa I) fragments of VH and
`Eco R V -Kpn I fragments of V K· Heavy and light chains were
`combined in an EcoRI-Sac 1-Mlu I triple ligation; a short
`Sac 1-M/u I fragment of pMY61 provided the intergenic
`region. The plasmid containing the 48G7 sequences was
`designated pDEI434. It was further modified by subcloning
`the entire expression cassette into pGC-1 (8), yielding
`pDEI440, and making the following substitutions by site(cid:173)
`directed mutagenesis (21): A Sac I site in the C" region was
`removed by introduction of a silent mutation. Two Hindlll
`sites upstream of the promoter were deleted by Hindlll
`digestion, blunting with the Klenow fragment of DNA
`polymerase I, and religation. The EcoRV site was replaced
`with a Sac I site, and unique Hindlll, Xho I, and BstEII sites
`were introduced as shown in Fig. 1. The expression cassette
`was then inserted back into pMY61 to yield p4xH. The
`following primers were designed to match the murine joining
`( J) regions and incorporate a Hind III into J" and a BstEII site
`into JH for cloning: JH1, TGAGGAGACGGTGACCGTG(cid:173)
`GTCCCTGCGCCCCA; JH2, TGAGGAGACGGTGACC(cid:173)
`GTGGTGCCTTGGCCCCA; JH3, TGCAGAGACGGTG(cid:173)
`ACCAGAGTCCCTTGGCCCCA; J~, TGAGGAGACGGT(cid:173)
`GACCGAGGTTCCTTGACCCCA; J"1' TTTGATTTCAAG(cid:173)
`CTTGGTGCCTCCACCGAACGT; 1"2, TTTTATTTCAAG(cid:173)
`CTTGGTCCCCCCTCCGAACGT; 1"3, TTTTATTTCAAG(cid:173)
`CTTGGTCCCATCACTGAACGT; JA, TTTTATTTCA(cid:173)
`AGCTTTGTCCCCGAGCCGAACGT; 1"5, TTTCAGCTC(cid:173)
`AAGCTTGGTCCCAGCACCGAACGT. They were used as
`an equimolar mixture (JH1-JH4 or J"1-J"5) during reverse
`transcription and PCR instead of H12 and L9 for cloning the
`VH and V" regions of the hybridomas 18R.136.1 (22), 28B4.2
`(23), AZ-28 (24), and 39,A11.1 (25) from their cell lines
`directly into p4xH. The V regions of 7G12-A10 (26) were
`cloned into p4x (p4xH lacking the Hindiii site) by using
`JH1-JH4 and a set of J" primers that incorporated a Kpn I site
`into the J region. Twenty-five PCR cycles (94°C for 1 min;
`50-55°C or 42°C during the first 5 cycles for VH for 2 min; and
`72°C for 1.5 min) were sufficient to amplify the desired
`sequences. The following primers were found to be optimal:
`primers H6 and L6 for 18R.136.1, H2 and L5 for 28B4.2, H2
`and L3 for AZ-28, H2 and L5 for 39,A11.1, and H7 and L5 for
`7G 12-AlO. Several clones of each chain were sequenced (27)
`to exclude PCR errors. We found that in some cases multiple
`J region primers had served to amplify the same V region,
`making one position in the J region ambiguous. The authentic
`sequence was identified by producing and PCR-amplifying
`eDNA encoding the whole Fab fragment and sequencing
`directly over the J region using a 32P-labeled primer and a Vent
`polymerase PCR sequencing kit (New England Biolabs). The
`combinations ofthe AZ-28, 18R.136.1, 28B4.2, 2Ell, and S107
`
`Mlul
`X hoi
`
`p4xH
`(5677 bp)
`
`FIG. 1. Map of the vector p4xH, indicating the restriction sites used
`for cloning. PphoA, PhoA promoter; stll, bacterial leader sequence;
`bla, 13-lactamase gene; Ori, coiEl origin of replication.
`
`chains with 48G7 were constructed by swapping EcoRI(cid:173)
`BamHI fragments containing the light chain sequences be(cid:173)
`tween pDEI434 and the respective expression vectors. The
`48G7 mutants were generated by site-directed mutagenesis in
`pDEI440 and were expressed directly in this vector.
`Expression and Purification of Chimeric Fab Fragments.
`For small-scale expression, 25 ml of Mops medium [30 mM
`Mops, pH 7.4/70 mM NaCI/10 mM KCl/1.6 mM MgS04/20
`mM N~Cl/0. 15% glucose/thiamine hydrochloride (1 ,.,_g/
`ml)/ampicillin (100 J.tg/ml)) supplemented with yeast extract
`and Casamino acids (Difco) in baffled shake flasks were
`inoculated with 0.5 ml of an overnight culture of 25F2 freshly
`transformed with the appropriate plasmid and incubated with
`vigorous shaking (300 rpm) at 30°C for 24 h unless otherwise
`noted. Optimal induction was found with yeast extract at 0.15
`g/liter and Casamino acids at 0.55 g/liter. Periplasmic lysates
`were prepared by resuspension of the cells on ice at 200 OD580
`units/ml in lysis buffer [20% (wt/vol) sucrose/30 mM
`Tris·HCl, pH 8.0/1 mM EDTA), transfer to microcentrifuge
`tubes, incubation with lysozyme at 1 mg/ml (added from a
`stock solution at 10 mg/ml in lysis buffer) at room temperature
`for 30 min, and a 3-min centrifugation in a microcentrifuge to
`pellet the spheroplasts. PhoA activity was measured by diluting
`the crude extract into DEA buffer [9.7% (vol/vol) diethanol(cid:173)
`amine, pH 9.8/0.5 mM MgCh/0.02% NaN3] with p(cid:173)
`nitrophenyl phosphate at 1 mg/ml and monitoringA405 • For a
`crude estimation, serial dilutions were set up in microliter
`plates, and the level of induction was judged visually by the
`highest dilution that still produced a signal. Larger cultures
`(1.5 liters) were treated as described above, scaling up all
`volumes accordingly and pelleting the spheroplasts in an SS34
`rotor (12,000 rpm, 15 min). High-density fermentations were
`performed in a 2.5-liter Bioflow III fermentor essentially as
`described by Carteret a/. (12), but with 3.3 mM NaH2P04 and
`6 mM KzHP04. Within 24 h, the cells were fully induced at an
`ODsso of 70-100 units, depending on the clone. Periplasmic
`lysates were prepared in 200-300 ml of lysis buffer with
`incubation for 1 h. The spheroplasts were pelleted (GS3 rotor
`at 8500 rpm for 30 min, if necessary followed by a centrifu(cid:173)
`gation at 16,000 rpm for 10 min in an SS34 rotor) and subjected
`to another incubation in lysis buffer as before. The superna(cid:173)
`tants were either subjected to an ammonium sulfate precipi(cid:173)
`tation (80% saturation) or concentrated to =100 ml in an
`Amicon ultrafiltration cell. Purification on protein G affinity
`columns (Sepharose CL-6B, Pierce) was performed at room
`temperature in 50 mM Mes, pH 5.5/100 mM NaCI. Protein was
`eluted with 100 mM glycine (pH 2.8); and the eluate was
`neutralized immediately with 0.1 vol of 1M Tris·HCl (pH 9.0).
`Characterization of the Recombinant Fab Fragments. Pro(cid:173)
`teins were analyzed by SDS/PAGE followed for purified
`protein by silver staining or for crude lysates by Western blot
`analysis (15) using an anti-human K chain-PhoA conjugate
`(Southern Biotechnology Associates) for detection and nitro
`blue tetrazolium chloride and 5-bromo-4-chloro-3-indolyl
`phosphate as substrate. The same antibody with p-nitrophenyl
`phosphate as substrate was used to detect hapten binding by
`ELISA. Wells were coated with bovine serum albumin (5
`J.tg/ml) conjugates of the appropriate haptens (8, 18, 22-26).
`Catalytic actjvity was measured as described (8, 22-24, 26).
`Concentrations of purified antibody were determined by mea(cid:173)
`suring Azso and A260 (28). Extinction coefficients varied be(cid:173)
`tween 61,500 and 74,200 M- 1·cm-t.
`
`RESULTS AND DISCUSSION
`Vector Construction. To investigate the beneficial influence
`of the human sequences on the folding of the Fab fragment in
`the bacterial periplasm, we chose to pursue chimeric constructs
`as opposed to humanized versions .of our catalytic antibodies.
`Humanization is an involved process that almost inevitably
`
`BEQ 1042
`Page 2
`
`

`
`Biochemistry: Ulrich et a/.
`
`Proc. Nat/. Acad. Sci. USA 92 (1995)
`
`11909
`
`results in some change in affinity for the hapten (29) that would
`likely be detrimental to catalysis. Due to the independence of
`V and C domains (30), no change in the structure of the V
`domains is observed when the murine C regions are exchanged
`for human versions (31).
`The expression system is based on that described by Carter
`et a/. (12) and this group (8). The vector from which it is
`derived, pMY61, is a pBR322 derivative very similar to pAK19
`(12); it uses the same PhoA promoter, inducible by phosphate
`starvation (32), stll signal sequences for secretion into the
`periplasm (33), and human C regions (34, 35), excluding the
`Cys-Ala-Ala hinge at the C terminus of the CH1 region. The
`V regions of the p-nitrophenyl phosphonate-specific antibody
`48G7 were subcloned into this vector from pDE166 (8). The
`genes of the phosphocholine-specific myeloma S107 (19) and
`those of the hybridoma 2E11 (18) were derived from analogous
`constructs, into which they had been previously cloned from
`the corresponding cell lines (H.D.U., P.A.P., and P.G.S.,
`unpublished results) .. The vector bearing the 48G7 Fab frag(cid:173)
`ment was then modified to allow the use of the V region PCR
`primers H1-H10 and Ll-L7, described by Huse eta/. (17), to
`clone antibody genes of unknown sequences: a Sac I site in the
`CK sequence and a Hindiii site upstream of the promoter were
`removed, Sac I and Xho I sites were incorporated at the N
`terminus of VK and VH, respectively, and Hindiii and BstEII
`sites were introduced into the J region sequences, resulting in
`the vector p4xH (Fig. 1).
`PCR primers incorporating Hindiii and BstEII sites into JK
`and JH, respectively, were designed to complement the V
`region primers. Combinations of these primers were used to
`amplify the V regions of the antibodies 18R.136.1 (22), 28B4.2
`(23), AZ-28 (24), 39,A11.1 (25), and 7G12-A10 (26) from
`eDNA obtained from their respective cell lines. The sequences
`of the light and heavy chain V regions of the cloned hybrido(cid:173)
`mas along with 48G7 and S107 are listed in Fig. 2 according to
`their subgroups as classified by Kabat et al. (36).
`Expression Studies. Small-scale shake-flask expression in
`Mops medium allowed for a quick estimation of relative yields,
`since the antibodies could easily be visualized on Western blots
`and their hapten-binding activity could be confirmed by
`ELISA assays of crude periplasmic fractions. Induction of the
`PhoA promoter was followed by measuring PhoA activity in
`
`the lysate. As shown in Fig. 3, accumulation of the antibody
`lagged 1-2 h behind the induction of PhoA.
`Under optimized expression conditions, the cells reached an
`OD580 of 2.0-2.5 units and remained stably induced for > 36 h
`without detectable lysis. Replacement of the PhoA promoter
`by the lac or tac promoter resulted in significant reductions of
`expression levels in some cases. Yields were more rigorously
`determined after purification of the protein from 1.5-liter
`cultures. As noted by Carter et a/. (12), the chimeric Fab
`fragments can be easily purified by protein G affinity chro(cid:173)
`matography. This method yielded the protein at >95% purity.
`Expression yields are given in Table 1.
`With the exception of the myeloma-derived S107, the yields
`are in the range of 1-3 mg/liter for four antibodies and 0.1-0.3
`mg/liter for the remaining three, which compares well with
`reported values (11). No obvious correlation between se(cid:173)
`quence homology and expression level was observed. The Fab
`fragments retained their hapten-binding and catalytic activity
`where tested (Table 1). Importantly, in all antibodies tested,
`the majority of the protein was produced in soluble form,
`whereas only a small fraction of Fab was soluble in the systems
`that we had examined previously. Even the small amounts of
`S107 Fab fragments were soluble, and only for the 2E11 clone
`was > 10% of total antibody associated with the cytoplasmic
`fraction (data not shown). Large quantities of recombinant
`protein could be produced in high-density (70-100 ODsso
`units) fermentations in a 2.5-liter Bioflow III fermentor by
`using a modification of the protocol described by Carter et a/.
`(12). Again, the antibodies were produced in soluble form,
`associated with the periplasm, and yields (10-150 mg/liter)
`were sufficient for crystallization and extensive kinetic studies
`(Table 1).
`On Western blots of crude periplasmic fractions of AZ-28,
`an additional band below the functional Fab fragment was
`detected. Deleting the heavy chain sequences from the plasmid
`and comparing the pattern of expression indicated that the
`additional band was a disulfide-linked light chain dimer.
`Deletion of the 48G7 heavy chain sequences resulted in light
`chains that migrated as monomers on a nonreducing gel (data
`not shown). The expression levels of these light chains were
`strongly reduced compared to the complete Fab fragment,
`indicating that isolated light chains are fairly unstable in the E.
`
`A
`
`FR4
`FR3
`FRI
`CDR3
`CDR2
`FR2
`CDRI
`1 DIVMTQSPTFLAVTASKKVTISC TASESLYSSKHKVHYLA WYQKKPEQSPKLLIY GASNRYI GVPDRFTGSGSGTDFTLTISSVQVEDLTHYYC AQFYSYPLT FGAGTKVEIK
`2 DIVMTQTPLSLPVSFGDQVSISC RSSQSLANSYGNTYLS WYLIIXPGQSPQLLIY GISNRFS GVPDRFSGSGSGTDFTLKISTVKPEDLGMYYC LQGTRQPPMYT FGGGTKVEIK
`3 ELVMTQTPLSLPVSLGDQASISC RFSQSIVHSNGNTYLE WYLQKSGQSPKLLIY KVSNRFS GVPDRFSGSGSGTDFTLKISRVEAEDLGVYYC FQGSHVPRT FGGGTKLEIK
`4 ELVM'l'QTPLSLPVSLGDQASISC RSSQSLLHSNGNTYLH WYLQKPGQSPKLLIY KVSNRFS GVPDRFSGSGSGTDFTLKISRVEAEDLGVYFC SQVTHVPPT FGGGTKLEIK
`5 DIQMTQSPSSLSASLGERVSLTC
`RASQEINGYLG
`WLQQKPDGTIKRLIY AASTLHS GVPKRFSGSRSGSDYSLTISSLESEDFADYYC LQYASYPRT FGGGTKVEIK
`6 ELVMTQSPSSLSASLGERVSLTC
`RASQDISNYLI
`WVQQKPDGTIKRLIY STSTLDS GVPKRFSGSRSGSDYSLTISSLESEDFADYYC LQYASSPFT FGSGTKLEIK
`7 ELVLTQSPSSMYASLGERVTITC
`KASQDINSYLN
`WFQQKPGKSPKTLIY RTNRLVD GVPSRFSGSGSGQDYSLTISSLEYEDMGIYYC LQYDEFPYT FGSGTKLEIK
`8 ELVMTQTPKFMSTTVGDRVSITC
`KASQNVGTPVA
`WYQQKPGQSPKLLIY SASNRYT GVPDRFTGSGSGTDFTLTISNMQSEDLADYPC QQYSSYPLT FGGGTKVEIK
`B
`
`CDR3
`FRI
`FR3
`CDR2
`FR2
`CDRI
`1 QVQLLESGAEIVRPGASVKLSCKASGYIFT DYWMN WVKQRPGQGLENIG GINPSDSSTNYNQRFKD KVTLTVDTSSNSAYIQLSSLTSEDSAVYFCAT
`SGRFRGGY
`2 QVQLLESGAELVKPGASVKLSCKASGYTFT
`RDMDY
`SYWMII WVKQRPGRGLENIG MIDPNSGG'l'KYNEKFKS KATLTVDKPSNTAYMQLSSLTSEDSAVYYCTR
`3 QVQLLESGAELMKPGASVKISCKATGYTFS SFWIE WVKQRPGHGLENIG EILPGSGG'l'HYNEKFKG KATPTADKSSNTAYMQLSSLTSEDSAVYYCAR GHSYYFYDGDY
`4 QVQLQQSGAELVKPGASVKLSCTASGFNIK DTYMH WVKQRPKQGLBWIG RIDPANVDTKYDPRFQD KATITADTSSKTTYLQLSGLTSEDTAVYYCAS
`YYGIY
`5 QVQLLESGGGLGQPGGSLRLSCATSGFTFT DYYFN WARQPPGKALBWLG FIRNKAKGYTTEYSASVKG RFTISRDNSQGILYLQMNTLRAEDSATYYCAR WGSYAMDY
`6 QVKLLESGGGLVQPGGSLRLSCATSGFTFS DFYME WVRQPPGKRLBWIA ASRNKANDYTTEYSASVKG RFIVSRDTSQSILYLQMNALRAEDTAIYYCAR DYYGSSYWYFDV
`7 QVQLLESGPELKKPGETVKISCKASGYTFT NYGMN WVKQAPGKGLKWMG WINTY'I'GEPTYADDFKG RFAFSLETSASTAYLQINNLKNEDTATYFCVQ AERLRRTFDY
`8 QVQLLESGPGIFQPSQTLSLTCSFSGFSLS TDGIGVG WIRQPSGKGLBWLA HI~IVLKS RLTISKDTSNNQVFLKIASVDTADTGTYYCVR
`IRDYYGYAMDY
`
`FR4
`1 WGAGTTVTVSS
`2 WGAGTTVTVSS
`3 WGQGTSVTVSS
`4 WGQGTTLTVSS
`5 WGQGTSVTVSS
`6 WGAGTTVTVSS
`7 WGAGTTVTVSS
`8 WGQGTSVTVSS
`
`FIG. 2. Amino acid sequences of the cloned V regions. Subgroup numbers are indicated in parentheses. (A) Light chains. Sequences: I, Sl07
`(I); 2, 2Ell (II); 3, 2884.2 (II); 4, 39, All.l (II); 5, 48G7 (V); 6, 18R.136.1 (V); 7, AZ-28 (V); 8, 7G12-A10 (V). (B) Heavy chains. Sequences:
`1, 18R.136.1 (IIA); 2, 7G12-A10 (IIA); 3, AZ-28 (liB); 4, 48G7 (II C); 5, 2884.2 (IliA); 6, S107 (IliA); 7, 39,All.l {misc.); 8, 2Ell (misc.). FRl-4,
`framework regions.
`
`BEQ 1042
`Page 3
`
`

`
`11910 Biochemistry: Ulrich et al.
`
`Proc. Nat/. Acad. Sci. USA 92 (1995)
`
`time, h
`
`time, h
`
`c
`+P04
`
`3
`
`5
`
`7
`
`9
`
`11
`
`13
`
`24
`
`40
`
`FIG. 3. Time course of antibody expression in shake flasks . .l,
`48G7; e, 18R.136.1. (A) ODsso of the cultures after inoculation at a
`1:200 dilution of an overnight culture. (B) Induction of the PhoA
`promoter, as measured by specific PhoA activity of the periplasmic
`lysates prepared . at the indicated times. The activity [log(OD4os
`units/min)] of a lysate prepared from a culture grown in the presence
`of2.0mM sodium phosphate was -1.64. (C) Western blot of 18R.136.1
`in the periplasmic fractions at the indicated times (20 JLl per lane).
`+ P04, lysate from a culture grown in the presence of 2.0 mM sodium
`phosphate.
`
`coli periplasm. Since the dimers do not bind to the protein G
`affinity column, they do not present a problem for purification.
`Characterization of Chain Combinations. Part of the diver(cid:173)
`sity of the immune response is generated by the random
`combination of heavy and light chains during B-cell develop(cid:173)
`ment. Due to this combinatorial diversity, antibody chains are
`well adapted to associate randomly with each other to form
`stable pairs. The association derives mainly from the interac(cid:173)
`tion between the C regions (37), but the V regions are
`promiscuous enough to allow for random pairings. This prop(cid:173)
`erty is exploited in phage and in vivo expression libraries.
`To test whether the chimeric expression system would
`accommodate pairs of heavy and light chains that were not
`selected to associate with each other during an immunization
`process, chain combinations between 48G7 and the other
`clones were constructed and expressed. In no case was the
`expression of a hybrid clone worse than that of the poor
`partner, indicating that the yield depends to a large degree on
`
`Table 1. Yields and activities of the cloned Fab fragments
`
`Yield, mg/liter
`
`Activity present
`
`Catalysis
`ELISA
`Fermentor
`Shake flask
`Antibody
`+
`+
`2.7 ± 29%
`48G7
`156
`+
`+
`1.6 ± 22%
`54
`AZ-28
`+
`+
`0.34 ± 20%
`20
`18R.136.1
`+
`ND
`0.25 ± 41%
`13
`2Ell
`+
`+
`0.11 ± 17%
`ND
`2884.2
`ND
`ND
`ND
`<0.02
`S107
`+
`+
`7G12-A10*
`56
`1.6
`+
`ND
`39,All.l*
`ND
`1.3
`For shake flasks, the yield after purification is shown, based on four
`cultures grown under identical conditions (mean ± SD as a percentage
`of the mean). For the fermentor, the yield after purification is shown,
`based on one experiment. ND, not determined.
`*Preliminary results based on one culture obtained independently
`(F.E.R., H.D.U., P.G.S., unpublished results).
`
`the expression level of the individual chains and not their
`association. The combinations also revealed which of the
`chains was responsible for the varying expression levels ob(cid:173)
`served. For example, the combination of the AZ-28 heavy
`chain and the 48G7 light chain expressed like AZ-28 itself,
`whereas the reverse combination showed levels comparable to
`48G7, and light chain dimers were no longer observed, indi(cid:173)
`cating that for AZ-28, the heavy chain limits expression. In
`contrast, the poor expression of S107 could not be attributed
`to one of the chains in particular, since neither combination
`with 48G7 resulted in any appreciable increase in expression
`compared to S107 itself. In all other cases, the combinations
`showed expression levels somewhere in between those of the
`two partners. Based on these results, it is likely that the
`chimeric expression system will be suitable for the generation
`of expression libraries by cloning random combinations of V
`regions from the spleen of an immunized mouse without
`significant losses in diversity resulting from insufficient ex(cid:173)
`pression levels.
`Effects of Point Mutations on the Expression Level. Protein
`produced in this system has been used to solve the crystal
`structure of the recombinant Fab fragment of 48G7 to 2.0 A
`(P.A.P. and P.G.S., unpublished results). To better understand
`the mechanism of this antibody a series of amino acid substi(cid:173)
`tutions around the active site was generated. Interestingly,
`although the use of rare codons was avoided in all cases, some
`of the mutations had large effects on the protein yield (Fig. 4 ).
`According to their expression levels, the mutants fell into three
`classes. The Tyr33HHis, His35HAla, His35HGln, and His35HGlu
`mutants, Tyr ~ Phe mutants at positions 99H and 100H, and the
`triple mutation GL48G7H (Asn30LSer /Gly34LSer I
`V K
`HisSSLAsp) all behaved like the wild-type 48G7. The Tyr91LLys,
`Tyr94LAla, Arg96LAla, Arg96LLys, and Tyr100HHis mutants
`showed 2- to 10-fold reductions in expression yields. In contrast
`to the Tyr99HPhe mutant, expression of the other three Tyr99H
`mutants was estimated to be at least 100-fold lower than that of
`wild-type 48G7 and comparable to the level of S107.
`All mutations lie in the complementarity-determining re(cid:173)
`gion (CDR) loops or the {3-strands immediately behind a loop,
`and none of them affect buried residues. Changes from a Tyr
`to a hydrophilic or charged amino acid like His, Arg, or Lys
`occur in all three classes. Most notably, Tyr99H seems to
`accommodate no other change than the conservative replace(cid:173)
`ment with Phe, even though it resides in a highly variable
`solvent-exposed position at the tip of the heavy chain CDR3
`loop. Similar behavior is observed for the Phe and His mutants
`of the adjacent residue, TyriOOH. Tyr91 L, Tyr94L, and Arg96L
`lie in the light chain CDR3 loop, and in the latter case, even
`the conservative change of Arg (which makes a hydrogen bond
`
`... ----
`
`-
`
`FIG. 4. Effects of point mutations on the expression level. Western
`blots were prepared from 25-ml cultures of 48G7 and the indicated
`mutants in pDEI440 (20 JLl per lane) grown in parallel under identical
`conditions. wt, Wild type; GL48G7H = AsnJOLSer/ Gly34LSer/
`His55LAsp.
`
`BEQ 1042
`Page 4
`
`

`
`Biochemistry: Ulrich et a/.
`
`Proc. Nat/. Acad. Sci. USA 92 (1995)
`
`11911
`
`to the hapten) to Lys leads to a significant reduction in
`expression level. In contrast, significant changes can be made
`to Tyr-33H and His-3SH in CDRl of the heavy chain without
`significant effects (both residues contact the hapten). One
`possibility is that the E. coli translation machinery might be
`affected by small alterations in the mRNA, as observed by
`Duenas et a/. (38). Alternatively, amino acid substitutions in
`the loops could make the protein thermodynamically unstable,
`thereby rendering it susceptible to proteolytic attack. This
`phenomenon has also been reported by other groups (39, 40).
`Alfthan et al. ( 41) have observed similar correlations of
`expression levels and protease sensitivity with temperature
`stability of Fab fragments that use different murine CHl
`regions. On the other hand, Knappik and Pliickthun (9) found
`positions in the heavy chain loops of the single-chain antibody
`McPC603 that influenced expression efficiency mainly
`through solubility effects without affecting the in vitro stability
`of the protein. The mutations in 48G7 seem to be of a different
`nature, however, since increased amounts of insoluble aggre(cid:173)
`gates, as found by Knappik and Pliickthun (9), were not
`observed in our poorly expressing clones. It is possible that the
`human C regions, which are absent in Pliickthun's system, keep
`even unstable or partly unfolded V regions in solution, thus
`preventing aggregation and allowing proteases to remove the
`unfolded chains. It should be interesting to investigate whether
`the changes in expression titer in our system can indeed be
`correlated with protein stability.
`Conclusion. We have demonstrated that a range of different
`hybridoma-derived antibodies and combinations of heavy and
`light chains not selected as pairs in vivo can be expressed as
`murine-human chimeric Fab fragments in E. coli. This ex(cid:173)
`pression system provides good yields of seven of eight anti(cid:173)
`bodies assayed, and in each case the protein remained in
`soluble form. This chimeric system seems to be a viable
`compromise between the high expression levels of humanized
`antibodies and the low levels observed for murine Fab frag(cid:173)
`ments, while avoiding the risk of losing catalytic activity by
`changing the V regions. Crystallography should be facilitated
`due to the greater homogeneity of the recombinant antibody
`when compared with murine Fab fragments generated by
`papain digestion, and site-directed mutagenesis becomes much
`less complicated, with the caveat of mutation effects on
`expression levels. Most importantly, however, attempts to
`improve the catalytic efficiency of existing antibodies by
`genetic approaches or the selection of catalytic clones from in
`vivo eDNA libraries (3, 4) should now become feasible.
`
`We thank D. Henner, B. Snedecor, M. Yang, and P. Carter
`(Genentech) for valuable discussions. We also thank L. Hsieh and E.
`Driggers for their help with 28B4.2 and AZ-28. P.L.Y. and P.A.P.
`contributed equally to this work. This work was supported by the
`Assistant Secretary for Conservation and Renewable Energy, Ad(cid:173)
`vanced Industrial Concepts Division of the U.S. Department of Energy
`under Contract DE-AC03-76F00098. P.G.S. is a Howard Hughes
`Medical Institute Investigator; H.D.U. is supported by a National
`Science Foundation Predoctoral Fellowship, P.A.P. was supported by
`a Damon Runyon-Walter Winchell Cancer Research Fund Fellow(cid:173)
`ship, P.L.Y. is supported by a Howard Hughes Medical Institute
`Predoctoral Fellowship in Biological Sciences, and F.E.R. is supported
`by National Institutes of Health Postdoctoral Fellowship F32AI09136.
`
`1. Schultz, P. G. & Lerner, R. A. (1993) Ace. Chern. Res. 26,
`391-395.
`2. Lerner, R. A., Benkovic, S. J. & Schultz, P. G. (1991) Science 252,
`659-667.
`3. Smiley, J. A. & Benkovic, S. J. (1994) Proc. Nat/. Acad. Sci USA
`91, 8319-8323.
`4. Tang, Y., Hicks, J. B. & Hilvert, D. (1991) Proc. Nat/. Acad. Sci.
`USA 88, 8784-8786.
`5. Pliickthun, A. (1991) Bio/Technology 9, 545-551.
`6. La Vallie, E. R., DiBlasio, E. A., Kovacic, S., Grant, K. L., Schen(cid:173)
`del, P. F. & McCoy, J. M. (1993) Bio/Technology 11, 187-193.
`
`7. Buckholz, R. G. & Gleeson, M.A. G. (1991) Bio/Technology 9,
`1067-1072.
`8. Lesley, S. A., Patten, P. A. & Schultz, P. G. (1993) Proc. Nat/.
`Acad. Sci USA 90, 1160-1165.
`9. Knappik, A. & Pliickthun,