Isolation and characterization of nucleic acid-binding antibody fragments from autoimmune mice-derived bacteriophage display libraries

Gene, 137 (1993) 77-83 0 1993 Elsevier Science Publishers

B.V. All rights reserved.

77

0378-1119/93/$06.00

GENE 07550

Isolation and characterization of nucleic acid-binding antibody fragments from autoimmune mice-derived bacteriophage display libraries (Filamentous phage surface display; immunoglobulins; MRL/lpr mice; recombinant DNA; expression in Escherichia cd)

Michael J. Calcutt, Marie T. Kremer, Michael F. Giblin, Thomas P. Quinn and Susan L. Deutscher Department ofBiochemistry, University of Missouri, Columbia, MO 65212, USA Received by J.W. Larrick:

9 August

1993; Accepted:

10 August

1993; Received at publishers:

27 August

1993

SUMMARY

The display of antibody fragments (Fab) on the surface of filamentous bacteriophage and selection of phage that bind to a particular antigen has enabled the isolation of Fab with numerous specificities, including haptens, proteins and viral particles. We have examined the possibility of isolating nucleic acid-binding Fab by constructing a combinatorial library of phage displaying Fab derived from autoimmune (MRL/lpr) mice. Autoimmune mice were chosen because they contain antibodies (Ab) reactive against nuclear components, including DNA, RNA and protein complexes. The library was panned against single-stranded (ss) calf thymus (CT) DNA and the selected Fabs were analyzed further. Characterization of the nucleic acid-binding phage led to the identification of two kinds of Fab with quite different properties. One Fab bound with high affinity a variety of ssDNA molecules, as well as several model RNA substrates. This Fab has been affinity purified to greater than 95% and competition studies revealed a marked preference for binding to poly(dT). The second Fab showed a reduced binding to RNA ligands and a restricted number of ssDNA molecules. Analysis of the deduced amino acid (aa) sequences of the Fab variable (V) regions revealed that the heavy (H) chain V region from the strong nucleic acid-binding Fab was derived from a V, gene that is used recurrently in autoantibodies. This Vn domain was most similar to an anti-ssDNA autoimmune monoclonal antibody (mAb) suggesting that antigen-binding specificities present in an autoimmune repertoire may be directly accessed by this approach. A light (L) chain shuffling experiment was performed to investigate how indiscriminate a H chain fragment from a nucleic acid-binding Fab could be and still retain antigen binding. In contrast to previous reports employing haptens or proteins, only a few L chains were isolated that enabled DNA and RNA binding. Furthermore, the deduced aa sequences of these L chains were almost identical to that of the original Fab. The few aa changes present resulted in a shift in antigen specificity, suggesting that fine tuning of substrate recognition may be possible. This report demonstrates that phage Fab display libraries will be a powerful method for the isolation of nucleic acid-binding Ab in vitro and may be applicable to the study of immune recognition, especially as it occurs in autoimmune disease.

Correspondence to: Dr. S.L. Deutscher, Department of Biochemistry, Ml21 Medical Sciences Building, University of Missouri, Columbia,

region of Ab; H chain, heavy chain; Fd, V, and Cm; FR, framework Ig, immunoglobulin; IP, immunoprecipitation; IPTG, isopropyl-@-

MO 65212, USA. Tel. (1-314) 882-2454;

thiogalactopyranoside; monoclonal mAb,

Fax (1-314) 884-4597.

amino acid(s); Ab, antibody(ies); Abbreviations: aa, Ag, antigen(s); bp, base pair(s); BSA, bovine serum albumin; Cb, carbenicillin; Cm, constant H chain region 1; cDNA, DNA complementary to RNA; CDR, complementarity-determining region; cfu, colony-forming unit(s); CT, calf thymus; DNA-l, initial nucleicacid-binding Fab (product of the initial clone); DNase I, deoxyribonuclease I; ds, double strand(ed); ELISA, enzyme-linked immunosorbent assay; Fab, Ab fragments having the Ag-binding site;

kb, Ab;

o/n, overnight; p, plasmid; phoresis; PCR, polymerase

kilobase( L chain, NC, nitrocellulose; nt,

light chain; nucleotide(s);

PA, polyacrylamide; PAGE, PA-gel electrochain reaction; PolIk, Klenow (large) frag-

ment of E. coli DNA polymerase I; pSK+, pBluescript SK+; R, resistance/resistant; RT, reverse transcription; rt, room temperature; SB, super broth (medium); SDS, sodium dodecyl sulfate; SLE, systemic lupus erythematosus; ss, single strand(ed); TBS, 50 mM Tris/lSO mM NaCl pH 7.5; V, variable region.

78 INTRODUCTION

The biological importance of nucleic acid-protein complexes is exemplified by their critical roles in gene regulation (Maniatis and Reed, 1987). Elucidation of the composition and function of many nucleic acid-protein complexes has been facilitated by Ab specific for certain of these structures, present in autoimmune mice including MRL/lpr strains (Brick et al., 1990) and humans with autoimmune diseases including SLE, and Sjogren’s syndrome (Lerner et al., 1981). These autoimmune diseases are characterized by the presence of a variety of antinuclear Ab directed against DNA, RNA and protein complexes. Of all known antinuclear Ab, anti-DNA are the best characterized (Schwartz and Stollar, 1985) and are known to be valuable disease markers (Pisetsky et al., 1990). While less is known about RNA Ab, they are of great interest due to their ability to recognize specific RNA molecules including Ul RNA (Deutscher and Keene, 1988) which is involved in pre-mRNA splicing (Maniatis and Reed, 1987). Ag-Ab complexes, especially those associated with autoimmunity, have traditionally been studied using mouse mAb or polyclonal antisera. Recently, new approaches have been devised to produce Ab using recombinant DNA methodologies, thus minimizing the need for mAb production in vivo. The application of PCR allows amplification of the Ig gene repertoire (Larrick et al., 1989), enabling the cloning and expression of functional Fub genes in E. coli (Skerra et al., 1988). Furthermore, methods exist to assemble Fab on phage surfaces so that those with a particular binding specificity can be enriched from a library by their affinity for Ag (Ward et al., 1989). The combinatorial Ab library method, in particular, has been useful for generating large numbers of Ab fragments because H and L chains are randomly combined, and has been used successfully to isolate Fab which bind to haptens (Huse et al., 1989) proteins (Barbas and Lerner, 1991) and viruses (Barbas et al., 1992). The goal of this study was to isolate and characterize nucleic acid-binding Fab from phage combinatorial libraries derived from autoimmune mice and compare these Fab to those Ab produced in autoimmune mice in vivo. The libraries were constructed from autoimmune mice since they provide an enriched source of antinuclear Ab. These experiments will lay the foundation for future structural studies of Fab-nucleic acid complexes.

RESULTS

AND DISCUSSION

(a) Identification and isolation of nucleic acid-binding Fab

A combinatorial library of DNAs encoding Fd (Vu plus Cm) and L chain DNA IgG fragments from auto-

immune MRL/lpr mice (Brick et al., 1990) has been constructed in the phage Fab display vector pComb3 (Barbas et al., 1991) as outlined in Fig. 1. Approximately 10’ transformants were obtained after electroporation of each of the Fd and L chain ligations. The product of the first clone isolated, DNA-l, bound 32P-labeled ss CT DNA strongly on NC membrane colony lifts, whereas control colonies of either random clones from the library or cells harboring pComb3 did not. Similar results were obtained with HindHI-digested denatured h phage DNA, although when the ds form of the same DNA was used it was impossible to differentiate between DNA-l and controlFab binding (data not shown). A second clone, MK5, was identified as a weaker DNA binder by ELISA (Schwartz and Stollar, 1985) using ss CT DNA as ligand. Colony lifts of bacterial cells producing MK5 also resulted in a reproducible stronger signal than control strains (data not shown). To establish whether the Fd and L chains of DNA-l and MK5 were similar or identical to each other or to any previously determined Ab sequences, the nt sequences of the Fab genes were determined. The deduced aa sequences were then compared to each other (Fig. 2) and to entries in the Protein Identification Resource database using the FASTA program (Pearson and Lipman, 1988). Strong similarities between DNA-l and MK5 were restricted to the FR of these Fab chains. The Vu sequence from DNA-l was very similar to a number of sequence members of the Vu558 family that is used recurrently in autoimmune mice. DNA-l VH was most similar (95% identity) to an anti-ssDNA mAb (202.~38) from NZB x NZW Fl autoimmune mice (Marion et al., 1992) with only three aa changes in the FR and CDR3. Interestingly, the Vu region from an autoimmune antiRNA mAb (D444; Eilat et al., 1988) was also very similar to DNA-l Vu, having a shorter CDR3 and a single aa change in CDR 1 (Fig. 2). The Vi. sequence from DNA-l was also very similar, although not identical to entries in the database. The Vu and V, sequences from MK5 were less similar to sequences in the database than the DNA-l counterparts. Although there were entries with up to 82% identity, no Vu or V, sequences with extensive similarities in the CDRs were found. (b) Fab production and purification DNA-l Fab was purified by affinity chromatography

to obtain material for further binding analyses. As shown in Fig. 3a, the Fab was purified to greater than 95%, a substantial enrichment over the Fab levels produced in the cell lysates (Fig. 3a, lane 3). It was important to evaluate the relative levels of expression of not only DNA-I but the other nucleic acid-binding Fab clones as well,

79

I

MIWlpf

Spleen and lymph node RNA

RT w

cDNA PCR

mouse

/I L chains

PCR

H chains

XhoIiSpef

Ligate transform E. co& Fab display phagemids 1

Amplify library H + L chain combinatorial library

* Inf& with helper phaga

4 rounds of panning versus ssDNA 1 Transduce E. coli 1 SuW?n eoIonies fbr ?PJDNA binding -t”

Colony purify Prepare plasmid ) Retransform ealla Rescreen for DNA binding Inset 1

Inset 2

Fig. 1. Isolation of clones expressing DNA-binding Fu6 from a combinatorial phage display library. A combinatorial library of antibody Fd and K L chain DNA, derived from two female MRLifpr mice, was constructed in the pbage display vector pComb3, using previously described protocols (Barbas and Lerner, 1991). Briefly, cDNA synthesized from lymph node and spleen total RNA was used as template for PCR of IgG Fd and K L chain DNA. A library of L chains was constructed first, into which Fd DNA was ligated. Following ele~tropora~~n of E. cafi XL-I-Blue, helper phage VCSM13 (Stratagene, La Jolla, CA, USA) were added to superinfect the culture resulting in packaged phagemid particles with surface displayed Fab fragments. The library of phagemid particles was enriched for those displaying anti-DNA Fab by four rounds of panning against ss CT DNA. After the fourth round of ampli~~atio~ a portion of the phagemid particles was used to transduce E. cdi XL-I Blue. The resulting CbR colonies were screened for the synthesis of anti-DNA Fab by transfer and lysis on NC filters followed by incubation with 32P-labeled ss CT RNA. Inset 1 shows an autoradiograph from the initial identification of such a clone (designated DNA-l). Following colony purification, plasmid DNA was extracted and used to demonstrate that the DNA-binding phenotype was plasmid-borne. Inset 2 shows an autoradiograph demonstrating the C3’P]DNAbinding abiiity of E. coli transformed with either plasmid from DNA-l (+) or pComb3 f-). Methods: The library of phage particles was panned against ss CT DNA by a modification of a previously described procedure (Parmley and Smith, 1988). Three wells of a microtiter plate were coated o/n with 30 pg of polylysine in TBS. The wells were washed and coated with 5 ng denatured CT DNA in 50 pl of 0.1 M sodium bicarbonate, pH 8,6 for 30 min at rt. The wells were blocked with a T&S/3% (w/v) BSA solution for 30 min at rt, blocking solution removed, and 50 ul of the phage library (approximately IO” cfu) was added to each well for 45 min at rt. The wells were washed, phage eluted and amplified for subsequent rounds of panning. Clones producing Fab were screened for DNA binding by performing colony lifts essentially as described (Barbas and Lerner, 1991) except &hat DNase I was omitted from the lysozyme buffer. The membranes were incubated in TBS containing non-fat dried milk and 20 ng of denatured sZP-labeled random primed DNA (US biochemical, Cleveland, OH, USA) for 4 h at rt. Unbound DNA was removed by three 15 min washes in TBS after which the membranes were subjected to autoradiography.

80 CLOne

FRI

WRI

FR2

WR2

FR3

WR3

FR4

LESGPELAKPGASVKRSCKASGYTFl LESGPELVKPGASVKMSCKASGIfFf PQSGPELVKPGASVKl6CKASGYllT

KYUnH SYVMN GYVMH

WKPRPGPGLEUIG M/KOKPGOGLEUIG UVKQRPGQGLAUIG

YINPSSGYTDYNPKFKG YINPYNDGTKYNEKFKG YlWPYNDGTKYNEKFKG

KATLTVDKSSSTAYblELNSLTSEDSAWYCAR KATLTSDKSSSTAYldELSSLTSEDSAWYCVR KATLTSDKSSSTAYTELSSLASEDSAAYYCAR

SAYYRSFDY GGYRPWAlDY GGFDY

UCPGTTLTVS UGPGTSVTVS UGQGTTLTVD

ELVLTPSPAlMSASPGP~lIlC ELQMl9SPASLSASVGETVTITC

SASSSVSSRFLN RASENIYSYLA

WWKSCASPKLYIY UYQOKPGKSWLLW WY GKSWLLW NY

DTSKLAP NAKTLAE NAKTL

GVPARFSGSGSGTSYSLTISSMEAEDAASYFC GVPSRFSGSGSGTPFSLKINSLPPEDFGSYYC GVPSRFSGSGSGT

HQUSSYPLT PHHYGTPLT claim

FGAGTKLELK FGAGTKLELK wAm~

VH sequences MK5 DNA-l D444

VL

sequences

HK5 DNA- 1 DNAl-J7 DNAl-J4D

Fig. 2. Deduced aa sequences of the variable domains of nucleic acid-binding Fab. DNA fragments encompassing Fd and L chain regions of Fab isolated in this study were subcloned into pSK+ and the nt sequence determined using T3 and T7 sequencing primers with a Sequenase 2.0 kit (US Biochemical). The isolation of DNAI-J7 and DNAI-J40 is described in section d. The highlighted boxes show the differences between DNAI-J7, DNAl-J40 and DNA-l. The Vn region of D444, an anti-RNA autoantibody is shown for comparison with DNA-I (see section c). The nt sequences corresponding to the Fab fragments have been deposited in GenBank (DNA-l H, No. UOO927;DNA-l L, No. UO0929; MK5 H, No. UOO940;MK 5 L, No. UOO941)

before we could perform a comparative analysis of their different binding specificities. The levels of Fub gene expression in E. cofi XL-l Blue cell lysates were examined by Western blot as shown in Fig. 3b. Similar amounts of Fab were produced from the different clones expressed in E. c&i, all of which were recognized by the secondary Ab used in the blotting procedure (Fig. 3b, lanes l-3), although it was impossible to determine the exact quantity of functional Fab in the lysates. (c) Characterization of Fab-nucleic acid interactions (1)

Binding afJinity of purified DNA-l

DNA-l was capable of binding more defined ligands than ss CT DNA including linearized pSK+, HindIIIdigested h DNA, and DNA homopolymers (see section ~2). The binding of DNA-l to a homogeneous DNA ligand (linearized pSK+) was determined to assess the relative affinity of the Fab-nucleic acid complex. Solution binding assays were performed using 0.1 pg of purified DNA-l with varying amounts of 32P-labeled linearized, denatured pSK* (Deutscher and Keene, 1988). Binding of DNA-l to pSK+ was a hyperbolic function of the free DNA concentration. The data were fitted to equation (1) using non-linear least-squares methods. r = ~(~)/~~(~)

(1)

where r = moles ligand bound/moles protein; R = moles ligand bound at saturation; A = free ligand concentration, and K, = dissociation constant for the protein-ligand complex. The K, was determined to be (4.4rt2.4) x lo-* M and the stoichiometry of binding (R) appeared to be close to 1:l. The affinity of DNA-l Fab for this DNA ligand was strong but similar to that reported for haptenbinding Fab (Hawkins et al., 1992) and a dsDNA-binding murine mAb (Ballard et al., 1984). It would be difficult to compare the affinity of DNA-l to other nucleic acid-

binding Fab obtained from phage libraries since this is the first reported such isolate. ( 2) Specijicity of nucleic acid-binding Fub The specificity of DNA-l and MK5 Fab for various nucleic acids was determined by a combination of solution binding and competition studies. Denatured restriction fragments of pSK+ were used as Ag since they represent a more defined ligand than CT DNA. When both the 67-nt and 2960-nt XhoI-XbaI ss fragments of pSK’ were used, purified DNA-l Fab or extracts containing DNA-1 bound the larger fragment preferentially (Fig. 4A, lanes 7 and 9), whereas DNA was not immunoprecipitated in a mock reaction lacking Fab or containing a control Fab (Fig. 4A, lanes 5 and 6). However, when DNA-l or MRL/lpr antiserum was incubated with only the 67-nt fragment, significant IP resulted (Fig. 4A, lanes l-4). Subsequent studies have shown that the 26-nt HindHI-XhoI fragment of the pSK+ polylinker can also be bound by DNA-l, albeit less strongly (data not shown). It was of interest to examine whether DNA-1 Fab could bind RNA, since it bound ssDNA strongly and because two of the Vu CDRs were strikingly similar to those of an anti-RNA mAb, D444 (Fig. 2; Eilat et al., 1988). A 32P-labeled Ul RNA transcript was used as ligand and was bound by DNA-l Fab (Fig. 4B, lane 4). Previous studies have shown the importance of H chain CDRs in determining Ag specificity, most notably through contacts between Ag and CDR3 (Chothia and Lesk, 1987). Since D444 has been reported to bind RNA exclusively (Eiiat et al., 1988), we wished to determine whether changing CDR3 of DNA-l Fd to that of D444 (H chain) and pairing it with DNA-l L chain would result in a Fab that only bound RNA. Accordingly, splice overlap PCR (Higuchi, 1990) was used to shorten CDR3 of

81 b.

a.

1

kDa

2

3

Fig. 3. Fab production tion. SDS-PAGE purified

DNA-l

left including:

4

of E. co[i extracts

Fab. Lane 1, protein 97.4 (phosphorylase

B

from E. coli. (a.) Fab purifica-

and purification

analysis

producing

size standards

DNA-l

nt -

and

(in kDa) shown at

B), 68 (BSA), 43 (ovalbumin),

and

29 (carbonic anhydrase); lane 2, E. coli extract without plasmid DNA; lane 3, E. coli extract expressing DNA-l Fab gene; lane 4, purified DNA-l

Fab. Total

protein

E. coli extracts

(10 Kg) from

or purified 165-

protein (0.2 ug) was separated on a 0.1% SDS-lo% PA gel in the absence of B-mercaptoethanol and stained with Coomassie Blue according to published

procedures

B-mercaptoethanol fide bond between A single colony

(Hoeffer

Scientific,

San Francisco,

CA, USA).

was eliminated from the samples to retain the disulthe H and L chains in the produced Fab. Methods:

was used to inoculate

5 ml SB (for total extracts)

or 1 1

SB supplemented with Cb (50 ug/ml) and 20 mM MgCI,. Cultures were grown at 37°C and induced with 1 mM IPTG after 6 h and incubated o/n as described

(Barbas

lysed by freeze-thaw,

and Lerner,

and centrifuged

1991). Cells were spun down,

at 13 000 x g. The resulting

super-

natant was used for SDS-PAGE or applied to a 3 ml goat IgG antiFab affinity column for purification essentially as described by Barbas et al. (1992).

The

DNA-l

Fab

was

eluted

with

10.8 mM

sodium

phosphate/44.5 mM citric acid/O.5 M NaCl, and the first three 0.5 ml fractions containing Fab were neutralized with 1 M Tris pH 9, concentrated

in a lo-kDa

cut-off

spin column,

and dialyzed

against

0.1 M

sodium phosphate/O.25 M NaCl pH 7.6 for subsequent use. (b.) Western blot analysis of Fab production. Total protein (2.0 ug) from E. co/i extracts was separated on a 0.1% SDS-lo% PA gel and transferred to a NC membrane in an electroblotter. A goat anti-mouse Fab antibody conjugated to alkaline phosphatase was used to detect bacterial Fab following standard protocols (BioRad, Hercules, CA, USA). Lanes: 1, MK5; 2, DNA-ld; including:

3, DNA-l

200 (myosin

Fab; 4, protein

size standards

(in kDa)

heavy chain), and those shown in panel a. Cell

lysates were prepared as described in panel a. p-mercaptoethanol was eliminated from samples to maintain an intact Fab molecule, which was critical

since the anti-Fab

Ab recognized

only the L chain.

DNA-l from GGYRPYYAMDY to GGFDY (CDR3 of D444) and the resulting Fab, designated DNA-Id, was tested for DNA and RNA binding by IP. Extracts containing DNA-ld failed to bind pSK+ DNA (Fig. 4A, lane 8), had marginal binding to Ul RNA (Fig. 4B, lane 3) and failed to bind 16s rRNA (data not shown). Thus, the differences in FR, the one change in Vu CDRl, or the L chain appear to have a significant contribution to D444 specificity, although it is possible that mAb D444 does not bind RNA as a Fab molecule. Nevertheless, the

Fig. 4. Binding of Fab to DNA and RNA. (A) DNA binding. Binding reactions were performed by IP and contained denatured 32P-labeled DNA derived from pSK+. Lanes l-4 contained

gel-purified

67-nt XhoI-

XbaI DNA; lanes 5-13 contained both XhoI-XbaI fragments of pSK+. The Ab or Fab -containing samples used were: lane 1, normal mouse antiserum (1 ~1); lane 2, MRL/lpr mouse antiserum (1 ~1); lane 3, control mouse Fab (4 pg); lane 4, purified DNA-l Fab (0.5 u(p); lane 5, no Fab added;

lane 6, control

mouse

Fab (4 ug); lane 7, purified

DNA-l

Fab (0.5 ug). E. co/i extracts (10 pg of total protein) used for IP contained the following Fab: lane 8, DNA-ld; lane 9, DNA-l; lane 10, MK5; lane 11, DNA-J4O;

lane 12, DNA-l

Fd plus MK5 L chain;

13,

MK5 Fd plus DNA-l L chain. Methods: s*P-labeled DNA was prepared using PolIk to fill in XhoI+XbaI-cut pSK+ with [“P]dCTP and [a’P]dGTP. The resulting from the polylinker) and precipitation,

DNA contained fragments of 67 bp (derived 2960 bp. Following phenol extraction and

the DNA was resuspended

in 10 mM Tris pH 8.0 and

boiled for 5 min prior to use. Ab- or Fab-containing samples were incubated with either 50 ng of the 67-nt fragment or 0.5 ug of both fragments

in TBS buffer for 10 min on ice, after which 4 pg goat anti-

Fab IgG was added. Samples were incubated on ice for 10min and bound to 70 ul of Staphylococcus aureus cells for a further 10 min on ice. The immune complexes were isolated, washed, and DNA eluted as described (Deutscher and Keene, 1988). DNA eluted from the complexes was analyzed on 6% PA/S M urea gels and subjected to autoradiography. (B) RNA binding. IP reactions contained SP6 RNA 3ZP-labeled Ul RNA (165 nt) (0.3 pg) and polymerase-transcribed E. coli extracts

containing

the following

Fab, or antisera:

lane 1, normal

mouse antiserum ( 1 ~1); lane 2, MRL/lpr mouse antiserum (1 ~1); lane 3, DNA-ld; lane 4, DNA-l; lane 5, MK5; lane 6, DNAl-J40. The reactions were performed and RNA analyzed essentially as described in A with the addition of RNase inhibitor (25 units RNasin) to the IP buffer.

DNA- Id protein product furnished a useful negative control Fab in IP reactions. The nucleic acid-binding specificity of MK5 Fab was

82 also investigated by solution binding assays. Despite the original identification of the MK5 clone by a positive DNA ELISA and its detectable binding of the same Ag on NC filters, it was not possible to detect binding of denatured pSK+ DNA by IP (Fig. 4A, lane 10). Whether this is a reflection of a more stringent specificity than DNA-l Fab or a reflection of differences in Ag binding assays (ELISA versus IP) is currently being investigated. Nevertheless, MK5 was able to bind Ul RNA (Fig. 4B, lane 5) and 16s rRNA in solution (data not shown), albeit at reduced levels compared to DNA-1 Fab. The specificity of the DNA-l Fab-nucleic acid interaction was also examined by competitive binding assays with various deoxyribo- and ribopolymers. Table I shows that 10 ng of dT polymer inhibits binding of DNA-l to denatured pSK+ by greater than 50%, whereas 10 ug of a dC polymer was required to achieve a similar inhibition. The ribopolymers tested and a dA polymer did not, however, effectively compete for binding. The fact that none of the ribopolymers competed for binding of denatured pSK+ DNA to DNA-l Fab suggested that the affinity of DNA-l for RNA may be less than that for DNA. Future binding experiments with purified DNA-l and RNA ligands will allow one to test this idea directly. (d) L chain shuffling experiments Due to the very nature of the combinatorial approach, the Fd and L chains present in DNA-l were unlikely to have been originally paired in the mouse repertoire. The probable low number of Fab present in the library that contain DNA-l Fd makes it quite possible that additional ssDNA-binding Fab could be generated with alternative L chains. In fact, the ability of different L chains in combination with a specific II chain to bind a hapten TABLE I Inhibition of DNA-l Fab-[32P]pSK+ Amount of polymer used

binding

Binding inhibition by: Dcoxypolymcrsa

Ri~polymersa

(Kang et al., 1991) and tetanus toxoid (Collet et al., 1992) has been demonstrated. To determine whether DNA-l Fd could form a functional anti-nucleic acid Fab with alternative L chains, DNA-l Fd DNA was (i) exchanged with MKS Fd so that DNA-l/MKS hybrids could be evaluated, and (ii) ligated to the L chain library DNA generated during library construction. When tested by IP, neither DNA-l/MK5 hybrid Fab were able to bind denatured pSK+ DNA (Fig. 4A, lanes 12 and 13), or RNA (data not shown). Several Fab clones were isolated from random pairing of DNA-l Fd and library L chains that bound denatured CT DNA on colony lifts. One clone was identical to DNA-l VL in aa sequence although different codons were used in two places. Two isolates, DNAl-J7 and DNAl-J40, were studied further and were found to bind the 2960-nt pSKf fragment but not the gel purified 67-nt fragment (data not shown). Both DNAL-J7 and DNAl-J40 also bound Ul RNA. These two isolates bound the same ligands indistinguishably and IP results for DNAl-J40 are shown in Fig. 4. The fact that DNAl-J7 and -540 had diminished binding to the 67-nt polylinker fragment is striking when the deduced aa sequences of the V, from these clones are compared to DNA-l. Fig. 2 shows that DNAL-J7 and DNAl-J40 are 92% and 93% identical to DNA-l, respectively. With respect to DNAl-J40 V, the backbone aa differences are very conservative changes. These changes are present in other Ab sequences, indicating that they are acceptable substitutions and do not destabilize the molecule’s structure. Therefore, the Ala to Thr change in CDRl, and/or the Gly to Ser and Leu to Pro changes in CDR3 appear to account for the difference in Ag recognition between DNA-l and DNAl-J40. Based on the crystal structure of the mouse Fab R19.9 (Lascombe et al., 1989), the substituted aa residues in CDR3 are more likely to be in close contact with the Ag than those in CDRl. Mutagenesis experiments are underway to examine how the aa substitutions in DNAl-540 alter Ab-nucleic acid recognition.

(ug) dA 0.01 0.1 1 10 50

5.2 13.5 29.8

dC

dT

A

C

G

-

0 0 _

-

3.5 69.1 79.0

61.6 98.8 99.2 100 --

0 10.3 _

0 26.9 _

U

18.2 9.8

“Binding of linearized, denatured [32P]pSK+ (2690 bp) to DNA-l Fab (as described in Fig. 4 legend) was inhibited by a 10 min preincubation of Fab with varying concentrations (0.01-50 pg) of deoxypolymers or ribopolymers. Binding inhibition represents and was calculated as % inhibition =( binding in presence of competitor) x loo/binding in absence of competitor. Dashes represent no data collected at these concentrations.

(e) Conclusions (I) A phage Fab display library derived from autoimmune mice has been constructed, from which nucleic acid-binding Fab have been isolated. This is the first report of the isolation of nucleic acid-binding Fab from phage combinatorial Ab libraries. (2) One of the nucleic acid-binding Fab has been highly purified and shown to bind a ssDNA ligand with high affinity and a stoichiometry of approximately 1:l. (3) Fab DNA-1 contained a Vn domain that is widely used in autoantibodies and is 95% identical to the Vn domain of an anti-ssDNA autoantibody, although CDR3

83 is unique to DNA-l. The strong homology suggests that the nucleic acid specificity of DNA-l is due to VH being derived from an anti-DNA Ab in the original mouse rather than being a specificity created by the combinatorial approach. (4) L chains that can support nucleic acid-binding with DNA-l Fd appear to be restricted in sequence and number. This suggests that in contrast to Fab for haptens and proteins, greater stringency is required to form a functional nucleic acid-binding Fab. (5) Subtle changes in nucleic acid specificity can result from a very limited number of aa replacements in a V, domain.

ACKNOWLEDGEMENTS

We thank Frank O’Sullivan for providing the MRL/lpr mice, Kirstie Calcutt for assistance with the manuscript, James Cullison for isolating DNAI-J40 and DNAl-J7, Carlos Barbas III and Dennis Burton for their technical advice, and Peter Tipton for intellectual input. This research was supported by grants from the Life and Health Insurance Medical Research Fund (No. 4225) and the National Institutes of Health (No. 5R29 GM47979) to S.L.D.

REFERENCES Ballard, D.W., Lynn, S.P., Gardner, J.F. and Voss, E.W.: Specificity and kinetics defining the interaction between a murine monoclonal autoantibody and DNA. J. Biol. Chem. 259 (1984) 3492-3498. Barbas III, C.F. and Lerner, R.A.: Combinatorial immunoglobulin libraries on the surface of phage: rapid selection of antigen specific Fabs. Methods: A Companion to Methods Enzymol. 2 (1991) 119-124. Barbas III, C.F., Kang, AS., Lerner, R.A. and Benkovic, S.J.: Assembly of combinatorial libraries on phage surfaces: the gene III site. Proc. Natl. Acad. Sci. USA 88 (1991) 7978-7982. Barbas III, C.F., Bjorling, E., Chiodi, F., Dunlop, N., Cababa, D., Jones, T.M., Zebedee, S.L., Persson, M.A., Nara, P.L., Norrby, E. and Burton, D.R.: Recombinant human Fab fragments neutralize human type 1 immunod~ficiency virus in vitro. Proc. Natl. Acad. Sci. USA 89 (1992) 9339-9343. Brick, J.E., Ong, S.H., Bathon, J.M., Walker, S.E. and O’Sullivan, F.X.: Anti-histone antibodies in the serum of autoimmune MRL and NZB/NZWl Fl mice. Clin. Immunol. Immunopathol. 54 (1990) 372-381.

Chothia, C. and Lesk, A.M.: Canonical structures for the hypervariable regions of immunoglobulins. J. Mol. Biol. 196 (1987) 901-917. Collet, T.A., Roben, P., O’Kennedy, R.O., Barbas III, CF., Burton, D.R. and Lerner, R.A.: A binary plasmid system for shuffling combinatorial antibody libraries. Proc. Natl. Acad. Sci. USA 89 (1992) 10026- 10030. Deutscher, S.L. and Keene, J.D.: A sequence-s~cific confo~ation~ epitope on Ul RNA is recognized by a unique autoantibody. Proc. Natl. Acad. Sci. USA 85 (1988) 3299-3303. Eilat, D., Webster, D.M. and Rees, A.R.: V region sequences of antiDNA and anti-RNA autoantibodies from NZB/NZW Fl mice. J. Immunol. 141 (1988) 1745-1753. Hawkins, R.E., Russell, S.R. and Winter, G.: Selection of phage antibodies by binding affinity: mimicking affinity maturation. J. Mol. Biol. 227 (1992) 889.-896. Higuchi, R.: PCR Protocols: A Guide to Methods and Applications. Academic Press, New York, 1990. Huse, W.D., Sastry, L., Iverson, S.A., Kang, A.S., Alting-Mees, M., Burton, D.R., Benkovic, S.J. and Lerner, R.A.: Generation of a large combinatorial library of the immunoglobulin repertoire in phage lambda. Science 246 (1989) 1275-1281. Kang, A.S., Jones, T.M. and Burton, D.R.: Antibody redesign by chain shuffling from random combinatorial libraries. Proc. Natl. Acad. Sci. USA 88 (1991) 11120-11123. Larrick, J.W., Danielsson, L., Brenner, C.A., Abrahamson, M., Fry, K.E. and Borrebaeck, C.: Rapid cioning of rearranged immunoglobulin genes from human hybridoma cells using mixed primers and polymerase chain reaction. Biochem. Biophys. Res. Commun. 160 (1989) 1250-1256. Lascombe, M.B., Alzari, P.M., Boulot, G., Saludjian, P., Tougard, P., Berek, C., Haba, S., Rosen, E.M., Nisonoff, A. and Poljak, R.J.: Three-dimensional structure of Fab R19.9, a monoclonal murine antibody specific for the p-azo~nzenearsonate group. Proc. Natl. Acad. Sci. USA 86 (1989) 607-611. Lerner, M.R., Boyle, J.A., Hardin, J. and Steitz, J.A.: Two novel classes of small ribonucleoproteins detected by antibodies associated with lupus erythematosus. Science 211 (1981) 400-402. Maniatis, T. and Reed, R.: The role of small nuclear ribonucleoprotein particles in pre-mRNA splicing. Nature 325 (1987) 673-678. Marion, T.N.. Tillman, D., Jou, N. and Hill, R.J.: Selection of immunoglobulin variable regions in autoimmunity to DNA. Immunol. Rev. 128 (1992) 123-149. Parmley, S.F. and Smith, G.P.: Antibody-selectable filamentous fd phage vectors: affinity purification of target genes. Gene 73 (1988) 305-318. Pearson,’ W.R. and Lipman, D.J.: Improved tools for biological sequence comparison. Proc. Natl. Acad. Sci. USA 85 (1988) 2444-2448. Pisetsky, D.S., Grudier, J.P. and Gilkeson, G.S.: A role for immunogenic DNA in the pathogenesis of systemic lupus erythematosus. Arth. Rheum. 33 (1990) 153-159. Schwartz, R.S. and Stollar, B.D.: Origins of anti-DNA autoantibodies. J. Clin. Invest. 75 (1985) 321-332. Skerra, A. and Pluckthun, A.: Assembly of a functional immunoglobulin fragment in E. co/i. Science 240 (1988) 1038-1040. Ward, E.S., Gussow, D., Griffiths, A.D., Jones, P.T. and Winter, G.: Binding activities of a repertoire of single immunoglobulin variable domains secreted from E. cob. Nature 341 (1989) 544-546.