oma lines (two d epitope-specific lines, two a lines, and two y lines) were ..... Huse, W. D., Sastry, L., Iverson, S. A., Kang, A. S., Alting-Mees, M.,. Burton, D. R.
Proc. Nati. Acad. Sci. USA Vol. 89, pp. 3175-3179, April 1992 Biochemistry
Human combinatorial antibody libraries to hepatitis B surface antigen SUZANNE L. ZEBEDEE*, CARLOS F. BARBAS I1t, YAO-LING HOM*, ROGER H. CAOTHIEN*, RICHARD GRAFF*, JULI DEGRAW*, JAYASHREE PYATI*, ROBERT LAPOLLA*, DENNIS R. BURTONt, RICHARD A. LERNERt, AND GEORGE B. THORNTON* *The R. W. Johnson Pharmaceutical Research Institute, 3535 General Atomics Court, Suite 100, San Diego, CA 92121; and tThe Scripps Research Institute, Departments of Molecular Biology, Immunology and Chemistry, 10666 North Torrey Pines Road, La Jolla, CA 92037
Contributed by Richard A. Lerner, December 26, 1991
murine antibodies for epitope mapping makes the hepatitis target an ideal model system to study the immune response to a viral antigen.
Human antibody Fab fragments that bind to ABSTRACT hepatitis B surface antigen (HBsAg) were generated by using a recombinant phage surface-display expression system. Characterization of HBsAg-specific Fab fragments isolated from two vaccinated individuals reveals diversity in specificity of antigen binding and in the sequences of the complementaritydetermining region. The sequence results show examples of human light-chain promiscuity that result in fine specificity changes and a strong relationship to a human germ-line gene. This application illustrates further that this technique is a powerful tool to isolate distinct human antibodies against immunogenic viral targets.
MATERIALS AND METHODS Immunization and RNA Preparation. A human volunteer was immunized with the hepatitis B virus recombinant vaccine Recombivax (Merck). On day 7 following the second booster injection (individual BO), whole blood was collected. A second volunteer, who received Heptavax (Merck) injections in 1986, followed by the Recombivax immunization schedule in 1990 (individual JM), donated lymphocytes by leukophoresis on day 14 following the second booster injection. The cells from both individuals were pelleted, and leukocytes were separated on Histopaque gradients (Sigma). RNA was prepared by sequential extractions with 6 M and 7.5 M guanidinium chloride (19). Serum titers to HBsAg were checked with the AUSAB test (Abbott). Library Construction. RNA isolated from JM and BO was subjected to reverse transcription (Perkin-Elmer) using the 3'-specific heavy-chain (yl, Fd region) and K and A light-chain primers previously described (4, 20). The antibody DNAs were then amplified in the presence of the corresponding 5'-specific primers for heavy- and light-chain genes by polymerase chain reaction (PCR) techniques. Heavy-chain PCR products were combined and digested with Xho I and Spe I, and K and A light-chain DNAs were separately combined and digested with Xba I and Sac I to yield three populations of digested antibody DNAs (per individual) for gel purification and cloning as described (21). These fragments were cloned into the pComb3 vector (7, 21, 22). Following bacterial transformation, the four combinatorial libraries were treated as described by Barbas and Lerner (21) to prepare phage. Solid-Phase Selection. Microtiter wells were coated with 100 Ag of HBsAg (adw and ayw subtypes; Boston Biomedica) overnight at 40C. The wells were blocked, incubated with one of the four phage libraries (typically >101" colony-forming phage per well), washed, and eluted (7, 22). The selected phage were then allowed to infect Escherichia coli XL1-Blue cultures (Stratagene), used to prepare new panned phage stocks, and reincubated with antigen in microtiter wells. From each of three incubations with HBsAg, the output phage titer was quantitated as described (7, 21). Nitrocellulose Filter Assay. Bacterial clones from each round of antigen panning were streaked on agar plates, induced, transferred to nitrocellulose, and treated with chloroform (21, 23). The filters were blocked and incubated with
The expression of diverse human monoclonal antibodies through the use of combinatorial systems has potential for the identification and development of these reagents as therapeutics. Use of a modified bacteriophage A genome for bacterial expression of both human and mouse antibodies has been described (1-6). The essence of this approach is that a repertoire of antibodies can be re-created from random pairings of cloned heavy- and light-chain genes. Recently, improvements to the expression of combinatorial antibody libraries were developed to allow for antigenspecific screening of larger libraries (7, 8). These improvements are based on reports demonstrating the expression of peptides and proteins on the surface of filamentous phage (9-18). The combinatorial antibody approach directs the expression of recombinant antibody Fab fragments to the surface of filamentous phage through the selection of a phagemid that coexpresses the combinations of heavy- and light-chain genes. Heavy-chain expression is linked to the phage coat protein gene III membrane anchorage domain, and the resulting fusion protein becomes anchored into the periplasmic space. Expression of the light chain from the plasmid DNA as a secreted protein permits Fab assembly within the periplasmic space. After helper phage infection, this antibody Fab then becomes incorporated into a phage by replacement of the wild-type gene III protein, and the Fab is displayed on the phage surface. The main advantage of the phage surface technology is the ability to sort large combinatorial libraries by a powerful enrichment and selection process. The selection process makes the combinatorial antibody technology attractive for the isolation of specific high-affinity human antibodies and for antibody isolation from individuals with low serum titers to the desired antigen. To examine the application of the phage surface-display technology to a viral system, we chose to generate and characterize human antibodiest to hepatitis B surface antigen (HBsAg). The availability of vaccines, purified antigen, and
Abbreviations: HBsAg, hepatitis B surface antigen; CDR, complementarity-determining region; FR, framework; HRP, horseradish
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.
peroxidase. MThe sequences reported in this paper have been deposited in the GenBank data base (accession nos. M88309-M88319). 3175
3176
Biochemistry: Zebedee et al.
HBsAg (ad/ay) particles (0.5 Ag/ml) in phosphate-buffered saline containing 1% bovine serum albumin, washed in phosphate-buffered saline containing 0.05% Tween-20, incubated with horseradish peroxidase (HRP)-conjugated mouse antiHBsAg antibody (16D; Ortho Diagnostics), and developed with 600 ,ug of 3,3'-diaminobenzidine (Sigma) per ml and 0.03% hydrogen peroxide. Conversion to Soluble Fab Fragments and Periplasmic Extract Preparation. Purified phage DNA, isolated from the third round of infected bacterial cells, was digested as described (21) to remove gene III sequences. Single clones were picked from the resulting transformation, grown in 5 ml of LB broth with carbenicillin (50 ,ug/ml) to an OD6w of 1.0, and induced with 2 mM isopropyl P-D-thiogalactopyranoside (IPTG) overnight at 370C. Culture supernatants were taken after the bacteria were pelleted. Extracts were prepared, essentially as described (24), from the pellet after resuspension in residual medium, incubation with 50 ttl of chloroform at 250C for 30 min, addition of400 1.d of 10 mM Tris, pH 8.0/1 mM phenymethylsulfonyl fluoride, and centrifugation at 7000 rpm in a Sorval SS34 rotor to remove bacterial debris. Supernatants and extracts were stored at 40C. Fab Concentration Determination and Indirect ELISAs. To determine the concentration of human Fab in each periplasmic extract, a sandwich ELISA was performed as follows: sheep anti-human Fab (Cappel Laboratories) was coated in Costar 96-well plates overnight at 4°C, periplasmic extract was added in serial 3-fold dilutions for seven rows, followed by HRP-conjugated anti-human Fab and substrates for color development. Supernatant and extract Fab amounts were plotted and quantitated according to known concentrations of human IgG (Cappel) as a standard. For indirect ELISAs, plates were coated with either adw or ayw subtype HBsAg (0.5 ,ug/ml) and then incubated with a standardized amount of protein extract in serial 5-fold dilutions for eight rows, followed by HRP-conjugated anti-human Fab for detection. Competition ELISAs. Mouse ascites fluids from six hybridoma lines (two d epitope-specific lines, two a lines, and two y lines) were obtained from Dave Milich (The Scripps Research Institute). HBsAg (adw or ayw; Boston Biomedica) was coated onto microtiter plates at 0.5 ,g/ml overnight at 4°C. Wells were blocked, incubated with human Fab at 4 times maximal binding levels (OD490 of 2.5, determined by indirect ELISA) for 1 hr at 37°C, and then incubated with mouse antibody for 1 hr at 37°C at a concentration to yield an OD of 2.5. Wells were washed and developed with HRPconjugated anti-mouse (Fab)2. Percent inhibition was calculated as 100 {[experimental OD/control OD (without initial competitor)] x 100}. For the reciprocal assay, mouse antibody was incubated with HBsAg-coated plates at a 4-fold excess, and human Fab to yield an OD of 2.5 was added and detected with goat anti-human Fab, essentially as described above. Nucleotide Sequence Analysis. Plasmid DNA was prepared as described (19) and nucleotide sequence was determined by using Sequenase 2.0 (United States Biochemical). To obtain the heavy-chain sequences, primers from the 5' vector sequence (5'-GGCCGCAAATTCTATTTCAAGG-3') and the CH1 constant region (5'-CGCTGTGCCCCCAGAGGT-3') were used. To obtain the light-chain sequence, the 5' vector sequence (5'-CTAAACTAGCTAGTCGCC-3') and CA primer (5'-GAGACACACCAGTGTGGC-3'), or CK primer (5'-CACAACAGAGGCAGTTCC-3') were used. The sequences were assembled by using MacVector analysis programs (IBI). Electron Microscopy. Phage that contained membranebound Fab fragments were isolated from individual JM clones after the initial (negative) or third (positive) panning round to HBsAg and confirmed for binding by using the filter assay as described above. The phage were diluted to 104 colony-
Proc. NatL Acad Sci. USA 89 (1992)
forming units per jul and incubated with adw/ayw HBsAg (International Enzymes, Fallbrook, CA) (0.5 ,ug/ml) for 30 min at 250C. This mixture was incubated with 300-mesh grids previously bound with rabbit anti-HBsAg for 10 min and then was fixed and treated with uranyl acetate for staining. The fields were viewed with a Hitachi HU12a at x 30,000 and x60,000 magnification.
Western Blot Analysis. Human Fab fragments from IPTGinduced 5-ml culture supernatants were concentrated (Centricon 10; Amicon) and separated in a nonreducing 10%o acrylamide gel (19). The proteins were transferred to nitrocellulose (25), blocked, incubated with HRP-conjugated goat anti-human Fab, and developed.
RESULTS To prepare human monoclonal antibodies that recognize hepatitis B virus, the antibody repertoires from two individuals vaccinated with recombinant HBsAg were cloned and expressed by a phage surface-display technique (7). Each individual was immunized with recombinant HBsAg vaccine, and lymphocytes were collected on day 7 (individual BO) or day 14 (individual JM) following the second Recombivax boost. Serum titers to HBsAg were examined and found to be 1:200 (BO) and 1:10,000 (JM). RNA was extracted from each of the cell samples and used to prepare PCR-amplified antibody-specific DNAs corresponding to heavy-chain (yl, Fd region) and K and A light-chain families for cloning into the pComb3 vector system. Four libraries were generated and expressed on the surface of phage: BO heavy chains with BO K or A light chains, and JM heavy chains with JM K or A light chains. To select the population of human Fab fragments that bind to HBsAg, the phage libraries were incubated with HBsAg (adw and ayw subtypes) in a microtiter well. The phage that bound to the well were eluted, reamplified in bacteria, and reincubated (panned) with HBsAg for three rounds of antigen selection. This enrichment process was monitored by titering the phage colony-forming units eluted from the antigencoated well at each stage (Table 1). Two of the four panned libraries, BOA and JMK, were enriched for phage that bind to HBsAg as demonstrated by the increasing number of colonyforming units eluted as the panning proceeded. Two of the libraries did not show any specific enrichment for HBsAgpositive phage (BOK, JMA). To confirm the enrichment numbers, a nitrocellulose filter assay for the detection of HBsAg antibodies was devised. This assay involves incubating the membrane-bound Fab fragments with HBsAg particles and detection with a mouse monoclonal antibody to HBsAg (HRP-conjugated). The number of clones that expressed HBsAg-specific Fab fragments for the BOA and JMK libraries increased as the panning stages progressed (Fig. 1). This result parallels the enrichment numbers and shows that the third-pan phage stocks were 95-100%o positive for anti-HBsAg. Interestingly, two positive JMA clones were also detected by this assay on the third-pan filter, whereas the enrichment data did not identify the presence of positive binders in this library. The expression of anti-HBsAg Fabs on the surface of phage allows electron microscopic examination of the phage Table 1. Enrichment of repertoire libraries to HBsAg Colony-forming units First pan Third pan Second pan Library BOK 2.3 x 10 4.0 x 1O 3.79 X 104 BOA 6.2 x 103 4.99 x 104 5.12 x 106 9.5 x 105 JMK 1.6 x 10 3.6 x 106 9.6 x 104 3.1 x 104 JMA 3.5 x 104
Xw ~ v .
Biochemistry: Zebedee et al. Initial
BO i.
First Pan
A e
0
IN%,
/.. 4.
3177
Third Pan
Second Pan
*
Proc. Natl. Acad. Sci. USA 89 (1992)
,... .
Ol C:,'4
BO
jm! 1.21
il"k. vtowr .:
JM x
ok
A114A
B
%i
i.
1*Ame..'.I .
JM x
FIG. 1. Nitrocellulose filter assay to detect HBsAg antibodies. Agar plates were streaked with the bacterial colonies from the initial (pre-pan), first, second, or third panning stage and transferred to nitrocellulose. After chloroform treatment, the filters were incubated with HBsAg (Boston Biomedica) (1 ug/ml), washed, and then incubated with HRP-conjugated mouse anti-HBsAg (Ortho Diagnostics) prior to substrate development. Panning stages and libraries are as indicated.
interaction with the HBsAg particle. Freshly prepared phage from several JMK clones (chosen from pre-pan or third-pan filter) were incubated with adw/ayw subtype HBsAg particles, secured to the microscopy grid by rabbit anti-HBsAg, and negatively stained (Fig. 2). Electron micrographs from positive (third pan; Fig. 2B) phage show an association of the filamentous phage with HBsAg particles on one end of the phage. Interactions of single phage with a particle, two phage per particle, and single phage with two particles bound were observed (data not shown). To characterize further the antibody Fab fragments, each of the third-pan phage populations from the positive libraries (BOA, JMK) was converted from membrane-bound to soluble Fab fragments. This manipulation requires removal of the gene III sequences from the pComb3 phagemid and subsequent library transformation into bacteria for the production of soluble Fab. To verify Fab expression in the periplasmic space and in the medium, proteins from individual converted clones were separated by gel electrophoresis and examined for Fab production by Western blot analysis using goat anti-human Fab secondary antibody. Under nonreducing conditions, a single positive band at 50 kDa (or 25 kDa for reducing conditions), equivalent to the human Fab control was seen for each of the clones examined (data not shown). To examine the population of converted soluble clones for reactivity to HBsAg, Fab concentration was quantitated to normalize for the Fab production of each clone. The extract Fab concentrations ranged from 0.1 to 16.0 pug/ml (data not shown). The clones were tested for HBsAg binding by an indirect ELISA with equivalent amounts of Fab from 24 clones from each of the two positive libraries (BOA, JMK). Thirty-two clones that bound to HBsAg in a concentrationdependent manner were identified (see Fig. 3 for examples). Sixteen clones that bound HBsAg were selected (8 from each individual) for further analysis and comparison.
FIG. 2. Electron microscopy of HBsAg particles and filamentous phage. Fresh phage were prepared from the pre-pan stage (A) or third pan stage (B) of the JMK library and incubated with HBsAg particles for 30 min. These complexes were added to rabbit anti-HBsAg antibody attached to a grid and then were washed, fixed, and negatively stained. The grids were viewed with a Hitachi electron microscope. (x 35,000.)
Competition ELISAs using mouse monoclonal antibodies with defined specificities to the group (a) or subtype (d, w) epitopes of HBsAg were performed on selected Fab clones from both libraries (Table 2). The assays were performed using the human Fab fragments as the competitor and showed competition of mouse antibody binding. The results were
3-
2-
0
-
0
1-
1
0.02
0.05
0.16
0.47
l
1.4
4.22
Fab, ng/ml FIG. 3. Specificity of binding of human soluble Fab to HBsAg. Equivalent concentrations of clone periplasmic extracts were diluted and incubated with bound HBsAg in a 96-well plate. After the plate was washed, anti-human Fab coupled to HRP was added, followed by reagents for color development. BO clone 08; o, BO clone 10; m,
*,
BO clone 12.
3178
Biochemistry: Zebedee et al.
Proc. Natl. Acad. Sci. USA 89 (1992)
Table 2. Percent inhibition of mouse monoclonal antibody binding to HBsAg with human Fab fragments Human antibody Fab Polyclonal BO JM serum Mouse 08 10 12 antibody 09 L2 01 10 15 HIV* BO JM 5D 14.3 0 0 0 0 45.9 62.4 22.4 84.5 0 43.3 15D 33.4 24.3 27.2 44.5 0 26.0 18.3 0 0 66.0 70.0 10A 25.9 11.3 13.9 24.3 2.1 2.4 0 4.0 0 47.3 57.0 14A 0.5 0 0.7 0 12.6 7.6 27.2 4.2 0 70.0 29.2 2Y 0 0 2.0 0.1 0 0 0 3.0 0 12.4 25.2 7Y 0 0 14.5 0 12.2 0 0 8.0 0 29.0 31.0 Assays were performed with adw subtype (5D, 15D, 10A, 14A) or ady subtype (2Y, 7Y). *Human immunodeficiency virus human Fab clone 14 (ref. 22).
confirmed to be similar in a reciprocal assay with mouse monoclonal HBsAg antibody as the competitor (data not shown). The general binding trends of the human Fab fragments show differences between BO and JM clones, but all of the selected Fabs from each individual appear to recognize a similar epitope (except the JMA clone). Clones isolated from the BOA library inhibit binding of the 15D and 10A mouse antibodies, whereas the JMK clones inhibit the binding of the 5D mouse line. Data obtained for the polyclonal serum from BO and JM (Table 2) demonstrate that these antibody specificities were present in the sera of both individuals. The nucleotide sequences encoding the variable domains of heavy and light chains from each of the selected clones were determined. Fig. 4 compares the predicted amino acid sequences for the HBsAg binders. Of the eight selected BOA clones, five share identical heavy- and light-chain sequences (as shown for clone 08 light chain and BO heavy chain, Fig. 4), and the others differ only in the light chain (clones 09, 10, 12). A comparison of the four different A chain sequences shows mostly identical FR structure (26) with amino acid substitutions in CDR1, CDR2, and CDR3 domains among the clones. The FR4 region of the light-chain J domain has also contributed to the diversity of these clones. The nucleotide sequence results for the JM clones examined predict four distinct light-chain proteins and two heavychain proteins (Fig. 4). The two positive A clones identified from the HBsAg filter assay (see Fig. 1) were sequenced and found to be identical (Fig. 4, L2 and JMa). The same heavy chain (JMa) that pairs with this A light chain was also found to associate with two other K clones (Fig. 4, clones 01 and 15). The other sequenced K clones (five separate clones) from JM
were found to use a different pairing ofheavy chain (3Mb) and light chain (Fig. 4, clone 10). The K sequences of JM share a similar FR (26) but differ in the length and composition of the CDR domains. The two isolated JM heavy chains have similar FR sequences except for FR2 length variability; the CDR sequences contain amino acid substitutions, with the most variation observed in CDR3.
DISCUSSION Combinatorial antibody libraries generated from human samples have been used to isolate high-affinity human antibody Fab fragments to tetanus toxoid (4, 7), human immunodeficiency virus type 1 gpl20 (22), and HBsAg (this work). We have used the phage surface-display technique to isolate HBsAg-specific binders from the lymphocytes of vaccinated individuals with both high and low titers to HBsAg and have shown that high-affinity antibodies with sequence variation can be selected. The data further demonstrate the power and selection capabilities ofthe combinatorial antibody technique for the cloning of human antibodies. HBsAg is a well-characterized molecule that can readily be used to study the gene III expression system and antigen immunogenicity in humans. Our electron microscopy results confirm both that antibody Fab fragments are expressed on the phage surface (7) and that the Fab-gene III fusion is properly located at the phage end (27, 28). The rare observation of a single Fab-expressing phage bound to two HBsAg particles suggests that some of the phage may express more than one Fab on their surface with this system. However, since the majority of the observed particle-phage associa-
LIGHT CHAIN SEQUENCE FR1
CDR1
a
08 AELTQPPSASGTPGQRVTISC 09 AELTQPPSASGTPGQRVTISC 10 AELTQPPSASGTPGQRVTISC 12 AELTQPPSASGTPGQRVTISC
SGSSSNIGTNTVN SGSSSNIGTNTVN SGSTSNIGTNTVN
JMX
L2 AELTQPPSVSGAPGQRVTISC
JMK
01 AELTQSPGTLSLSPGERATLSC 10 AELTQSPATLSLSPGERATLSC 15 AELTQSPGTLSLSPGERATLSC
FR2
CDR2
FR3
CDR3
FR4
CL
SGSSSNIGSNTVN
WYQQLPGRAPKLLIY WYQQLPGAAPKLLIY WYQQLPGTAPKLLIY WYQQLPGTAPKLLIY
SNNERPS SNNERPS SNSERPS SNNERPS
GVPDRFSGSKAGTSASLAISGLOSEDEADYYC GVPDRFSGSKSGTSASLAISGLOSEDEADYYC GVPDRFSGSKSGTSASLAISGLQSEDEAEYYC GVPDRFSASKSGTSASLAISGLQSEDEADYYC
EAWDDNLHGPV EAWDDSLQGPL EAWDDSLQGPV AAWDDSLHGPV
FGGGTRLTVLR FGGGTKLTVLG FGGGTKVTVLG FGGGTKLTVLR
QP QP QP
TGSSSNIGADSEVT
WYTRLAGAAPKLLIF
DNSFRPS
QSFDKEPEESN
FRRGTRVTVLR
QP
RASQSVFSNYLA RASQSVSSYLA RASQSVSSSYLA
WYQQKPGQAPRLLIY
GASSRAT DASNRAT GASSRAT
GVPDRFSVSRSGTSASLAITGLQAEDEGDYYC GIPDRFSGGGSGTDFTLTINRLEPEpFAVYYC GIPARFSGSGSGTDFTLTISSLEPEDFAVYYC GIPDRFSGSGSGTDFTLTISRLEPEDFAVYYC
QQYGSSPST
FGQGTKVELK FGGGTKVEIK FGGGTKVELK
RT RT RT
WYQQKPGQAPRLLIY WYQQKPGQAPRLLIY
QQRSNWPPS QQYGSSPT
QP
HEAVY CHAIN SEQUENCE
CDR1
FR2
BO
AQVKLLESGAEVRKPGASVKVSCKASGYTFT
GYYLH
WVRQAPGQGLEWMG
WIHPNNGGTFHPPKFRG
RVTMTRDTSISTAYMELSSLRSDDTAVYYCAI
EYGDTDCGGDCYSI
JMa
AQVKLLESGPGLVKPSETLSLTCTITGGSIS
SIYSSGTTYYNPSLES
AQVKLLESGPGQVKPSETLSLTCTVSGGPLS
SSSYFWG SSSNYWG
WIRQPREGAEWIG
JMb
WIRQTPGKRLEWIG
SIYNSGTAYYNPSLET
RVTMSVDTSKNQFSLNLDSVTAADTAVYYCAR RVTISVDTSKNQVSLNLKSVTAADTAMYYCAR
PLTHDDFLTAYYPGGGYFDL
FRI
CDR
FR3
CDR3 PLRLNYNSRSYYPGIGPFDM
FR4 CHI WGQGTLVTVSS AS WGQGTMVTVSS WGRGTLVTVSS
AS AS
FIG. 4. Predicted amino acid sequences for the selected BO and JM anti-HBsAg clones. The nucleotide sequences of eight BOA, two JMA, and seven JMK clones were determined and translated into protein sequence' (single-letter code). The sequences are segregated into complementarity-determining regions (CDRs) and framework (FR) regions based on homology to known sequences (26). Only the clones with unique sequences are presented here for the light chain and the heavy chain. Sequences are as follows, from top to bottom: light chains BOA08, BOA09, BOA10, BOA12, JMA2, JMKO1, JMK10, and JMK15 and heavy chains BO (assembles with all BO light chains), JMa (assembles with JMA, JMKO1, and JMK15), and JMb (assembles with JMK10). Five clones of the BOA library and of the JMK library were sequenced and found to be identical to the BOA08 and JMK10 sequences.
Biochemistry: Zebedee et al. tions were one particle per phage, the monovalent Fab display model for this technique is supported. As expected, the individual with low serum titer to HBsAg (BO) showed a more limited immune response to the vaccine than the high-titer individual (JM). Of the characterized BO anti-HBsAg binders, only limited light-chain sequence diversity was observed: the light chains were found in association with the same heavy chain, and the Fab fragments all recognize a similar epitope of HBsAg. However, among the clones analyzed from the high-serum-titer (JM) library, significantly more diversity was observed. Two heavy-chain proteins were identified and several light chains showed differences in CDR length and composition, especially in the CDR3 and FR4 regions. The hypervariability of the CDR3 domain has been noted previously (29, 30) and can be contributed by various mechanisms of D-region diversity. A comparison of the CDR3 heavy-chain sequences from the human anti-HBsAg clones with the antibody data base did not identify exact matches with other known antibody sequences. However, light-chain comparisons revealed that the K chain from JM clone 15 is identical in FR, CDR1, and CDR2 to the germ-line HumKv325 gene (31) and homologous in all the CDRs to the autoantibody TayKv322 (32). Other antibody sequences that are related to the HumKv325 gene include antibodies to rheumatoid factor, chronic lymphocytic leukemia, Sjogren syndrome, and cytomegalovirus (32-35). The significance of this K-chain similarity between the HBsAg antibodies and the autoimmune antibodies is unclear; but some viruses may play a role in inflammation (34-36), or alternatively, the human germline light-chain antibody repertoire may be limited. Antibody characterization data for the HBsAg Fab fragments suggest that much of the antigen recognition is dependent upon the human heavy chain, and the light chains show more promiscuity. For example, the same BO heavy chain associates with several light-chain sequences to recognize the same epitopes on HBsAg. Also, the specificity of the germline-related JM15 clone for HBsAg is strictly dependent on the heavy chain, as an identical light chain has been documented to assemble with distinct heavy chains to bind different antigens (31-35). Human heavy-chain promiscuity was also observed among the JM clones; one heavy chain (JMa) was able to assemble with both K and A light chains, although the two classes of antibodies recognize different epitopes (see Table 2). These results demonstrate that chain promiscuity can be utilized to affect specificity changes and that chain shuffling experiments can be utilized to re-engineer antibodies via the combinatorial approach (37). Eight different variable-region clones with sequences that vary in heavy- and light-chain combinations, framework subgroups, and CDR length and composition have been characterized. It cannot be determined from the combinatorial approach used to generate the positive Fab fragments whether these exact heavy/light chain pairings occur in the antibody repertoire of the vaccinated individuals, but only that the separate heavy and light chains exist. The importance of each isolated heavy/light chain combination is the identification of a functional Fab that has high affinity and specificity for the target antigen. Changes in the antibody sequences from each of the original individual repertoires, such as potential mutations from PCR error and cloning construction (at the amino terminus of each mature protein), have been tolerated by the binding of the expressed antibody to HBsAg. The isolation of human antibody Fabs that bind to hepatitis B virus will allow their examination as passive immunization agents for the prevention and therapy of viral hepatitis B. Antibodies to HBsAg are known to protect against infection (38). We have described human antibodies that possess the protective group (a) reactivity. These antibodies, generated from HBsAg-selected combinatorial libraries, represent prototype passive immunotherapy candidates.
Proc. Natl. Acad. Sci. USA 89 (1992)
3179
We thank Didier Leturcq and Ann Moriarity for help in assay development and Cheng-Ming Chang for electron microscopy work. We thank Dave Milich and Raju Koduri for discussions, T. T. Wu for help with sequence comparisons, and John L. Gerin, Jack Geltosky, and Raju Koduri for reviewing the manuscript. 1. Sastry, L., Alting-Mees, M., Huse, W. H., Short, J. M., Hay, B. N., Janda, K. D., Benkovic, S. J. & Lerner, R. A. (1989) Proc. Natl. Acad. Sci. USA 86, 5728-5732. 2. Huse, W. D., Sastry, L., Iverson, S. A., Kang, A. S., Alting-Mees, M., Burton, D. R., Benkovic, S. J. & Lerner, R. A. (1989) Science 246, 1275-1281. 3. Caton, A. J. & Koprowski, H. (1990) Proc. Nat!. Acad. Sci. USA 87, 6450-6454, and correction (1991) 88, 1590. 4. Persson, M. A. A., Caothien, R. H. & Burton, D. R. (1991) Proc. Natl. Acad. Sci. USA 88, 2432-2436. 5. Mullinax, R. L., Gross, E. A., Amberg, J. R., Hay, B. N., Hogrefe, H. H., Kubitz, M. M., Greener, A., Alting-Mees, M., Ardourel, D., Short, J. M., Sorge, J. A. & Shopes, B. (1990) Proc. Natl. Acad. Sci. USA 87, 8095-8099. 6. Williamson, R. A., Persson, M. A. A. & Burton, D. R. (1991) Biochem. J. 277, 561-563. 7. Barbas, C. F., Kang, A. S., Lerner, R. A. & Benkovic, S. J. (1991) Proc. Natl. Acad. Sci. USA 88, 7978-7982. 8. Kang, A. S., Barbas, C. F., Janda, K. D., Benkovic, S. J. & Lern.r, R. A. (1991) Proc. Natl. Acad. Sci. USA 88, 4363-4366. 9. Smith, G. P. (1985) Science 228, 1315-1317. 10. Parmley, S. F. & Smith, G. P. (1988) Gene 74, 305-318. 11. Bass, S., Greene, R. & Wells, J. A. (1990) Proteins: Struc. Funct. Genet. 8, 309-314. 12. Cwirla, S. E., Peters, E. A., Barrett, R. W. & Dower, W. J. (1990) Proc. Natl. Acad. Sci. USA 87, 6378-6382. 13. Devlin, J. J., Panganiban, L. C. & Devlin, P. E. (1990) Science 249, 404-406. 14. McCafferty, J., Griffiths, A. D., Winter, G. & Chiswell, D. J. (1990) Nature (London) 348, 652-654. 15. Scott, J. K. & Smith, G. P. (1990) Science 249, 387-390. 16. Clackson, T., Hoogenboom, H. R., Griffiths, A. D. & Winter, G. (1991) Nature (London) 352, 624-628. 17. Hoogenboom, H. R., Griffiths, A. D., Johnson, K. S., Chiswell, D. J., Hudson, P. & Winter, G. (1991) Nucleic Acids Res. 19, 4133-4137. 18. Lowman, H. B., Bass, S. H., Simpson, N. & Wells, J. A. (1991) Biochemistry 30, 10832-10838. 19. Maniatis, T., Fritsch, E. F. & Sambrook, J. (1982) Molecular Cloning: A Laboratory Manual (Cold Spring Harbor Lab., Cold Spring Harbor, NY). 20. Kang, A. S., Burton, D. R. & Lerner, R. A. (1991) Methods: Companion Methods Enzymol. 2, 111-118. 21. Barbas, C. F. & Lerner, R. A. (1991) Methods: Companion Methods Enzymol. 2, 119-124. 22. Burton, D. R., Barbas, C. F., III, Persson, M. A. A., Koening, S., Chanock, R. M. & Lerner, R. A. (1991) Proc. Natl. Acad. Sci. USA 88, 10134-10137. 23. Helfman, D. M., Feramisco, J. R., Fiddes, J. C., Thomas, G. P. & Hughes, S. H. (1983) Proc. Natl. Acad. Sci. USA 80, 31-35. 24. Ferro-Luzzi Ames, G., Prody, C. & Kustu, S. (1984) J. Bacteriol. 160, 1181-1183. 25. Zebedee, S. L. & Lamb, R. A. (1988) J. Virol. 62, 2672-2772. 26. Kabat, E. A., Wu, T. T., Perry, H. M., Gottesman, K. S. & Foeller, C. (1991) Sequences of Proteins of Immunological Interest (U.S. Dept. Health Human Ser.), Publ. No. 91-3242. 27. Chang, C. N., Model, P. & Blobel, G. (1979) Proc. Natl. Acad. Sci. USA 76, 1251-1255. 28. Crissman, J. W. & Smith, G. P. (1984) Virology 132, 445-455. 29. Kabat, E. A. & Wu, T. T. (1991) J. Immunol. 147, 1709-1719. 30. Sanz, I. (1991) J. Immunol. 147, 1720-1729. 31. Radoux, V., Chen, P. P., Sorge, J. A. & Carson, D. A. (1986) J. Exp. Med. 164, 2119-2125. 32. Kipps, T. J., Tomhave, E., Chen, P. P. & Fox, R. I. (1989) J. Immunol. 142, 4262-4268. 33. Kipps, T. J., Tomhave, E., Chen, P. P. & Carson, D. A. (1988) J. Exp. Med. 167, 840-852. 34. Newkirk, M. M., Gram, H., Heinrich, G. F., Ostberg, L., Capra, J. D. & Wasserman, R. L. (1988) J. Clin. Invest. 81, 1511-1518. 35. Hay, F. C., Soltys, A. J., Tribbick, G. & Geysen, H. M. (1991) Eur. J. Immunol. 21, 1837-1841. 36. Smith, L. H. & Wilkes, R. M. (1978) Calif. Med. 119, 38-44. 37. Kang, A. S., Jones, T. M. & Burton, D. R. (1991) Proc. Natl. Acad. Sci. USA 88, 11120-11123. 38. Seeff, L. B., Wright, E. C., Zimmerman, H. J., Alter, H. J., Dietz, A. A., Felsher, B. F., Finkelstein, J. D., Garcia-Pont, P., Gerin, J. L., Greenlee, H. B., Hamilton, J., Holland, P. V., Kaplan, P. M., Kjernan, T., Koff, R. S., Leevy, C. M., McAuliffe, V. J., Nath, N., Purcell, R. H., Schiff, E. R., Schwartz, C. C., Tamburro, C. H., Vlahcevic, Z., Zemel, R. & Zimmon, D. S. (1978) Ann. Intern. Med. 88, 285-293.
