Production and Characterization of High-Affinity Human Monoclonal Antibodies to Human Immunodeficiency Virus Type 1 Envelope Glycoproteins in a Mouse Model Expressing Human Immunoglobulins

ABSTRACT Human (Hu) monoclonal antibodies (MAbs) against the human immunodeficiency virus type 1 (HIV-1) envelope glycoproteins (Env) are useful tools in the structural and functional analysis of Env, are under development both as potential prophylaxis and as therapy for established HIV-1 infection, and have crucial roles in guiding the design of preventative vaccines. Despite representing more than 50% of infections globally, no MAbs have been generated in any species against C clade HIV-1 Env. To generate HuMAbs to a novel Chinese C clade Env vaccine candidate (primary isolate strain HIV-197CN54), we used BAB5 mice that express a human immunoglobulin (Ig) M antibody repertoire in place of endogenous murine immunoglobulins. When immunized with HIV-197CN54 Env, these mice developed antigen-specific IgM antibodies. Hybridoma fusions using splenocytes from these mice enabled the isolation of two Env-specific IgM HuMAbs: N3C5 and N03B11. N3C5 bound to HIV-1 Env from clades A and C, whereas N03B11 bound two geographically distant clade C isolates but not Env from other clades. These HuMAbs bind conformational epitopes within the immunodominant region of the gp41 ectodomain. N3C5 weakly neutralized the autologous isolate in the absence of complement and weakly enhanced infection in the presence of complement. N03B11 has no effect on infectivity in either the presence or the absence of complement. These novel HuMAbs are useful reagents for the study of HIV-1 Env relevant to the global pandemic, and mice producing human immunoglobulin present a tool for the production of such antibodies.

Human (Hu) monoclonal antibodies (MAbs) against the human immunodeficiency virus type 1 (HIV-1) envelope glycoproteins (Env) are useful tools in the structural and functional analysis of Env, are under development both as potential prophylaxis and as therapy for established HIV-1 infection, and have crucial roles in guiding the design of preventative vaccines. Despite representing more than 50% of infections globally, no MAbs have been generated in any species against C clade HIV-1 Env. To generate HuMAbs to a novel Chinese C clade Env vaccine candidate (primary isolate strain HIV-1 97CN54 ), we used BAB5 mice that express a human immunoglobulin (Ig) M antibody repertoire in place of endogenous murine immunoglobulins. When immunized with HIV-1 97CN54 Env, these mice developed antigen-specific IgM antibodies. Hybridoma fusions using splenocytes from these mice enabled the isolation of two Env-specific IgM HuMAbs: N3C5 and N03B11. N3C5 bound to HIV-1 Env from clades A and C, whereas N03B11 bound two geographically distant clade C isolates but not Env from other clades. These HuMAbs bind conformational epitopes within the immunodominant region of the gp41 ectodomain. N3C5 weakly neutralized the autologous isolate in the absence of complement and weakly enhanced infection in the presence of complement. N03B11 has no effect on infectivity in either the presence or the absence of complement. These novel HuMAbs are useful reagents for the study of HIV-1 Env relevant to the global pandemic, and mice producing human immunoglobulin present a tool for the production of such antibodies.
Monoclonal antibodies (MAbs) are essential reagents in a wide variety of research, diagnostic, and clinical settings. MAbs against the human immunodeficiency virus type 1 (HIV-1) envelope glycoproteins (Env) have enabled extensive studies into Env structure and antigenicity (25,58,60), and neutralizing MAbs (NMAbs) of human origin have been used therapeutically against HIV-1 in clinical trials (53) and as experimental microbicides in the simian-HIV macaque challenge model (54). Moreover, NMAbs are especially useful in the design of vaccine strategies that aim to recapitulate such neutralizing antibody (NAb) responses in vivo. Since infused human NMAbs provide sterilizing immunity against HIV-simian immunodeficiency virus chimeric viruses (SHIV) in macaque challenge models (13,29,45), such NAb responses elicited via active immunization might provide protective immunity against HIV-1 (7). Although an extensive catalogue of MAbs to HIV-1 Env currently exists (see http://hiv-web.lanl.gov /content/immunology/index.html for details), the majority of these were derived from patients infected with the B clade of HIV-1, which predominates in Europe and the Americas, or from animals immunized with Env from such isolates. Although many of these MAbs bind non-clade B Env, there are doubtless many Env epitopes that remain uncharacterized due to the focus on the B clade, which represents a minority of the circulating pandemic strains (21a). The C clade of HIV-1 accounts for 50% of infections globally and is particularly prevalent in sub-Saharan Africa and South Asia (21a). The relative "success" of C clade viruses justifies an intensified effort to understand their biology. To this end, the generation of MAbs should focus on clade C and other non-clade B Env isolates in order that the library of MAbs against HIV-1 Env can be expanded.
The production of MAbs often involves the use of rodents, as they are amenable to such studies in terms of costs and numbers, and their B cells are easily immortalized. However, the most useful MAbs in a clinical setting are human MAbs (HuMAbs), since they can be used clinically in humans without rejection by the host immune system. For this reason, many researchers have produced rodent MAbs and then made them suitable for use in humans by replacing portions of the rodent immunoglobulin (Ig) chain with the equivalent human structures. Such methodologies have allowed rodent MAbs to be used successfully in humans without induction of anti-rodent Ig Abs that neutralize the transferred rodent MAb (16). Obviously, making HuMAbs directly in rodents would obviate this genetic manipulation of the desired MAb clone. Moreover, all of the broadly reactive NMAbs against HIV-1 that have been derived to date are of human origin, and many, but not all, have long complementarity-determining region heavy chain 3 (CDR-H3) loops that mice and some other rodents are unable to produce (8,10,37,41,57). It is thought that these extended loops allow some NMAbs to access conserved, often recessed surfaces of HIV-1 Env and are thus responsible in part for their breadth of activity (7). Therefore, it has been recommended that Env-based vaccine antigens be tested in animal models, such as rabbits, that can elaborate long CDR-H3 loops and MAbs derived from nonhuman primates and humans infected with simian immunodeficiency virus and HIV-1, respectively (7). However, the generation of transgenic mice producing human immunoglobulin provides a unique opportunity to isolate human MAbs from a rodent model, combining the convenience of rodents with the CDR-H3 loop flexibility of a human Ig repertoire. The BAB5 mice used in this study express a human IgM repertoire and are capable of producing affinitymatured Ab responses (35).

Reagents
Mice. Ten-to twelve-week-old BAB5 mice were obtained from the joint specific-pathogen-free animal breeding facility of the MRC Laboratory of Molecular Biology and the Babraham Institute (Cambridge, United Kingdom). BAB5 mice carry the human Ig heavy and both and light chain loci and have inactive endogenous Ig heavy and loci (35). All experiments were performed under appropriate licenses in accordance with the UK Animals (Scientific Procedures) Act 1986.
Immunizations and preparation of hybridomas. The mice were immunized by subcutaneous injection of 25 g of gp140 97CN54 formulated with 15 l of the CpG-oligodeoxynucleotide based ImmunEasy adjuvant (QIAGEN Ltd., Crawley, United Kingdom). In all, four booster immunizations were given, and blood samples were taken from the tail for serological analysis immediately before each immunization. After the final booster immunization the mice were transferred to a licensed animal facility at the University of Oxford, where they were euthanized 3 days after the final booster and the splenocytes were harvested for the preparation of hybridomas. Hybridomas were prepared by the fusion of the immune splenocytes isolated from immunized BAB5 mice with NS-1 murine myeloma cells using polyethylene glycol 1500 (Roche Diagnostics, Lewes, United Kingdom) as previously described (19,65). The supernatants from the hybridomas were screened for reactivity with gp140 97CN54 and a second HIV-1 clade C Env antigen, gp140 96ZM651.8 , by enzyme-linked immunosorbent assay (ELISA). A second gp140 97CN54 capture ELISA was also used to confirm antigen-specific MAb production. Selected colonies underwent four rounds of dilution cloning before a final round of colony selection based on MAb production levels.
Production and concentration of IgM MAbs. Clarified hybridoma TCSNs were filtered using a 0.22-m-pore-size Stericup filter (Millipore Ltd., Watford, United Kingdom). Sterile ammonium sulfate solution [76% wt/vol (NH 4 ) 2 SO 4 ; Merck Chemicals Ltd., Poole, United Kingdom] was added in the ratio 1:1 (vol/vol) to the filtered TCSN, the mixture was incubated at 4°C for 2 h and centrifuged at 13,000 ϫ g for 1 h at 4°C, and the pellet was dissolved in sterile distilled H 2 O. The precipitation process was repeated, and the solution was then dialyzed in SnakeSkin tubing (Perbio Science UK Ltd., Cramlington, United Kingdom) with a 10-kDa molecular weight cutoff against sterile phosphate buffered saline for 48 h at 4°C. Dialyzed IgM MAbs were filter sterilized using 0.22-m-pore-size Spin-X filters (Fischer Scientific UK, Loughborough, United Kingdom). The IgM MAb concentration was determined by ELISA with known concentrations of human IgM() as a standard.
ELISAs. ELISA plates (Greiner Bio-One Ltd., United Kingdom) were coated with 50 l/well of gp140 at 1 g · ml Ϫ1 or any of the nested gp140 97CN54 peptides at a concentration of 10 g · ml Ϫ1 , in 100 mM NaHCO 3 (pH 8.5) overnight at 4°C. The plates were washed three times in phosphate-buffered saline (PBS) supplemented with 0.05% Tween 20 and blocked for 1 h at room temperature with 200 l/well of 2% (wt/vol) nonfat milk (Marvel) dissolved in PBS supplemented with 0.05% Tween 20. Plates were washed as before, and a dilution series of either IgM MAb or TCSN in PBS containing 1% (wt/vol) bovine serum albumin (BSA) (as sample buffer [SB]) was added directly to the ELISA plate for 1 h. For the gp140 97CN54 capture ELISAs, plates were coated with 1:2 to 1:10 (vol/vol) diluted IgM MAb containing TCSN. After blocking, a gp140 97CN54 dilution series in SB was added for 2 h. MAb IgG1b12 or antiserum ARP422 was used to detect captured gp140 97CN54 . For the MAb competition ELISAs, antigen was incubated with a saturating concentration of the first MAb for 1 h before the addition of a 50% binding concentration of the second MAb and incubation for 1 h more. After washing, 50 l/well of the appropriate HRP-conjugated antispecies IgG or IgM Ab at a concentration of 0.8 g · ml Ϫ1 diluted in SB was added for 1 h. ELISAs were then developed using TMB reagent (Jencons, United Kingdom), and the absorbance was measured at 450 nm. Assay cutoffs were calculated as the mean absorbance plus two standard deviations of wells that lacked primary Abs but were otherwise treated identically. The binding of the MAb in the presence of an excess amount of competing MAb was compared to the binding of the MAb without competing MAb.
Surface plasmon resonance (SPR). Twelve thousand response units of gp140 97CN54 and an irrelevant control protein, BSA (Sigma-Aldrich), were immobilized onto sensor chip CM5 flow cells (Biacore) as described previously (2) using the BIAcore 2000 (Biacore, Uppsala, Sweden). MAbs and sCD4 at a concentration of 50 g · ml Ϫ1 (except 447-52D, which was used at 10 g · ml Ϫ1 ) were injected into the flow cells. For competition analysis MAbs were used at saturating concentrations. The CD4i-specific MAbs and A32 were tested alone or in combination with sCD4. The sCD4 alone was injected over the chip surface three times, and the mean response was subtracted from the response observed with the combinations of CD4i surface MAbs and sCD4. All experiments were carried out at 25°C. Any signal detected in the BSA flow cell was subtracted from that in the gp140 97CN54 flow cell to give the antigen-specific signal.
Determination of relative avidity or true affinity by ELISA. To determine the relative avidity of the concentrated IgM MAbs, binding to gp140 was then measured in a standard ELISA. The relative avidity is reported as the concentration of MAb that gave half-maximal binding. The dissociation constant (K D ) was calculated for the monomeric fractions of the MAbs by the methods described elsewhere (14,44,48). Each MAb was tested in triplicate, the experiment was repeated twice, and the average K D was then determined from the pooled data.
Neutralization and Ab-dependent complement-mediated virus inactivation assays. The IgM MAbs were tested for neutralizing activity and complementmediated inactivation (CMI) against primary isolate HIV-1 97CN54 (a gift from C. Moog, INSERM, Strasbourg, France) (50) in an infectivity reduction assay using phytohemagglutinin-IL-2-activated human peripheral blood mononuclear cells. For the CMI assay a 10% final concentration of fresh normal human serum (NHS) was included as a source of complement. As a control, serum from the same donor that had been heat inactivated (HI) at 56°C for 30 min was used. These assays were adapted from protocols described previously (1,5). Briefly, HIV-1 97CN54 was diluted in IL-2 medium containing a final concentration of 100 g · ml Ϫ1 IgM MAb or control Abs with or without 10% fresh or HI NHS, in six wells per dilution of virus. MAb and virus were incubated together at 37°C for 1 h before addition of 2 ϫ 10 5 peripheral blood mononuclear cells. The cells were washed at 24 h, and the medium was exchanged again on days 3 and 7. On day 10, 100 l of TCSN was taken for the p24 assay. The p24 assay was based on that described previously (31). Wells were considered positive for HIV-1 infection if they had an absorbance reading in the p24 assay greater or equal to the mean background reading plus three standard deviations. Scores were entered into the program ID-50 v5.0 (46) to calculate viral titers.
Statistics. Analyses were performed using GraphPad Prism version 4.01 (18a). For normally distributed data, comparisons of more than two groups were carried out using the one-way analysis of variance test. Comparison of two data sets used unpaired t tests. For data that were not normally distributed, Kruskal-Wallis analysis was performed to compare more than two data sets. The Mann-Whitney test was used to compare two data sets. Bonferroni's correction was applied when multiple two-group comparisons were made. Error bars represent the 95% confidence interval (CI) unless stated otherwise.

RESULTS
The antigenic profile of gp140 97CN54 . Native HIV-1 Env has a complex quaternary structure that should be retained in the soluble form of the glycoprotein if NMAbs to conserved, discontinuous epitopes are to be obtained after immunization.
Real-time SPR analysis of intermolecular interactions allows detection of weak interactions and those with fast dissociate rate constants that can go undetected by ELISA. We therefore characterized the antigenicity of gp140 97CN54 by SPR using 15 MAbs, 2 PAbs, 1 oligoclonal Ab (raised against a gp120 peptide), and sCD4 ( Fig. 1A and B) to evaluate whether the molecule is appropriately folded. gp140 97CN54 bound sCD4 and the antibody 412d, whose gp120 binding characteristics closely mimic those of the N terminus of the coreceptor CCR5 (61), suggesting that the molecule is capable of binding both its receptors. Of the five CD4i-specific MAbs tested, two bound to gp140 97CN54 in the absence of sCD4 while all five bound when sCD4 was coinjected (Fig. 1B). MAbs E51 and 412d demonstrated a similar binding profile, showing some binding in the absence of sCD4, which increased four-to fivefold upon coinjection of sCD4 (P was equal to 0.0286 in each case; one-tailed Mann-Whitney test). Binding of MAbs 48d, 17b, and X5 was only observed in the presence of sCD4, consistent with the phenomenon of coreceptor binding site exposure after engagement of sCD4 and indicative of a functional Env molecule. C1-C4 region-specific MAb A32 bound in the absence of sCD4 as expected (59), and binding increased by 40% when sCD4 was coinjected (P ϭ 0.0286; one-tailed Mann-Whitney test). Overall, gp140 97CN54 was recognized by 13 of 15 MAbs and 2 of 3 PAbs despite being of clade C origin. The panel of Abs used here were derived from humans infected with clade B virus or animals immunized with clade B Env, except for CA13, which was derived from a mouse vaccinated with a clade A construct. The recognition of conformational epitopes involved in CD4 binding (IgG1b12) and of the CD4i surface of gp140 97CN54 suggests that this recombinant antigen has a native structure. Of the moderately to broadly reactive NMAbs, IgG1b12 and 447-52D bound gp140 97CN54 but 2G12 did not, as the glycan epitope is absent. HIV-1 97CN54 Env does not possess the epitope for the broadly reactive NMAb 2F5, and the gp140 97CN54 construct terminates just before the epitope for the broadly active NMAb 4E10, which is otherwise present on the HIV-1 97CN54 virus. The degree and stability of binding varied greatly depending on the Ab and the nature of the epitope involved ( Fig. 1C and D). IgG1b12 and 447-52D bound with relatively high avidity but dissociated substantially within 420 s of the end of MAb injection. Similarly, sCD4 dissociated noticeably from the oligomer (Fig. 1C). By contrast, MAb 5F3 and PAb HIVIg both bound with high avidity and did not dissociate appreciably in the 420 s after injection (Fig. 1D).
Immune response to gp140 97CN54 in BAB5 mice. The immunogenicity of gp140 97CN54 in BALB/c mice has been investigated and shown to elicit reciprocal endpoint IgG titers on the order of 10 5 when injected subcutaneously in the presence of CpG-oligodeoxynucleotide-containing adjuvants (data not shown). Because BAB5 mice generally produce lower titers of Ab in response to immunization than nontransgenic animals (35), we gave the mice booster injections four times to achieve maximal responses and to encourage affinity maturation. Three BAB5 mice were immunized a total of five times with 25-g doses of gp140 97CN54 (Fig. 2), and their antigen-specific IgM titers were established after one, two, four, and five immunizations. IgM titers were detected after a single immunization of gp140 97CN54 , and these increased by approximately fourfold after a booster immunization. On the day of the fusion the VOL. 14,2007 HUMAN MONOCLONAL ANTIBODIES TO HIV-1 IN A MOUSE MODEL 159 reciprocal titers were ϳ300. Titers of this magnitude are in keeping with those seen after immunization of BAB5 mice with other antigens (38), and moreover this final bleed was taken 3 days after the fifth booster was given, too early for detection of peak serum Ab titers. Other studies have suggested that MAbs can be readily isolated despite relatively weak serum Ab responses (55). Due to the low IgM titers on the day of the fusion and the limiting quantities of serum, the sera were not tested for neutralization of HIV-1 97CN54 . Isolation and characterization of the multimeric nature of IgM MAbs after hybridoma fusion. Of the three BAB5 mice that were immunized, the splenocytes from one mouse were fused with NS-1 myeloma cells on the day of culling while the others were cryopreserved. In the initial fusion, 77 primary hybridomas were obtained, of which 14 showed significant reactivity with gp140 97CN54 and 2 reacted with gp140 96ZM651.8 . Thirteen of the hybridomas either failed to grow or grew but lost antigen reactivity during dilution cloning. Several attempts to produce hybridomas from the cryopreserved splenocytes were unsuccessful. Eventually a second successful fusion from a different mouse produced 42 primary hybridomas, of which 11 showed reactivity towards gp140 97CN54 and 1 showed reac- tivity towards gp140 96ZM651.8 . Again, many hybridomas failed to grow or progressively lost antigen reactivity. Finally, two IgM MAbs, N3C5 and N03B11, were isolated from two of the three mice. These clones stably secreted IgM MAbs and were shown by ELISA to capture gp140 97CN54 , which could then be detected using MAb IgG1b12. Both novel MAbs were shown to use light chains (data not shown). N3C5 and N03B11 were concentrated by ammonium sulfate precipitation to yield stocks of approximately 200 g · ml Ϫ1 (N3C5) and 4 g · ml Ϫ1 (N03B11). To ascertain whether the IgM MAbs were pentameric or monomeric IgM, the two forms were separated by size exclusion chromatography (Fig. 3), and the eluate fractions were tested by ELISA to determine which contained IgM. Both IgM MAbs consisted of a mixture of pentamers and monomers as shown by the spread of human IgM in the fractions between 900 kDa and 180 kDa. Considerable running through of one species into subsequent fractions occurred, as is typical with this technique. The monomers outnumbered pentamers by approximately 10:1 in pooled fractions covering the separate peaks (data not shown). The monomer and pentamer fractions were tested for binding to gp140 97CN54 (Fig. 4); both forms bound gp140 97CN54 when applied at equal concentrations. Under these conditions an equal number of binding sites were present, but there were fivefold fewer pentameric than monomeric molecules. Monomeric N3C5 bound to gp140 97CN54 to a greater extent than the same MAb in pentameric form (P ϭ 0.0001; unpaired two-tailed t test), suggesting that steric factors may render the binding of bulky pentamers less favorable than the binding of the monomeric form. N03B11 monomers and pentamers appeared to bind equally in this ELISA. Breadth of recognition, relative avidity, and true affinity of N3C5 and N03B11. We estimated the relative binding avidity of the ammonium sulfate concentrated MAbs for a panel of Env by ELISA, to establish the conservation of the epitopes for these MAbs among isolates from clades A, B, C, D, and F (Fig.  5). MAb N3C5 bound the homologous gp140 97CN54 with high avidity and bound heterologous gp140s from clades A and C with similar or slightly lower avidity, whereas N03B11 bound to the homologous gp140 and another gp140 from a geographically distant clade C isolate. Neither of the MAbs bound to gp120 97CN54 nor to a range of other gp120s from clade B (data not shown), suggesting that the epitope was either trimer dependent or located on the gp41 ectodomain. The avidity of N3C5 for the homologous antigen was significantly higher than that of N03B11 (median avidity of 0.27 nM versus 1.5 nM; P ϭ 0.0012; two-tailed Mann-Whitney test with Bonferroni's correction). However, the binding avidity did not differ significantly when the two MAbs were compared on gp140 96ZM651.8 . N3C5 bound significantly better to gp140 97CN54 than to gp140 92UG037 (0.27 nM versus 1.0 nM; P ϭ 0.0008; two-tailed Mann-Whitney test with Bonferroni's correction), but the binding of N3C5 and N03B11 did not vary significantly between gp140 97CN54 and gp140 96ZM651. 8 .
The K D of the monomeric fraction was tested at 22 and 37°C at two solid-phase antigen concentrations (Fig. 6). Insufficient quantities of pentamers were obtained to determine the K D for this fraction. The results show that an increase in temperature from 22°C to 37°C had no significant effect on the K D at either concentration of solid-phase antigen. The 50-fold change in solid-phase antigen concentration had no significant effect on the K D of either MAb on gp140 97CN54 at either temperature. However, at 37°C the K D of N3C5 on gp140 96ZM651.8 differed 2.2-fold for a 50-fold difference in coating antigen concentration (P ϭ 0.0076; two-tailed unpaired t test with Bonferroni's correction). The binding of N3C5 to gp140 92UG037 , known to be weaker than its binding to the clade C Env molecules from the earlier experiment (Fig. 5), could not be detected at a solid-phase concentration of 0.1 g · ml Ϫ1 . Overall, at 22°C, coating antigen concentration had no significant effect on the calculated K D for either MAb on either clade C gp140. Unlike the relative avidity assay (Fig. 5) which used the crude MAb preparations, this assay showed that at 22°C and with a solidphase antigen concentration of 5 g · ml Ϫ1 , N3C5 had a higher affinity for gp140 96ZM651.8 than gp140 97CN54 or gp140 92UG037 , with 96-fold (P ϭ 0.0066) and 60-fold geometric mean differences (P ϭ 0.0285), respectively (two-tailed Mann-Whitney test with Bonferroni's correction). In addition, N3C5 had a higher affinity for gp140 96ZM651.8 than N03B11 did (140-fold geometric mean difference; P ϭ 0.0066; two-tailed Mann-Whitney test with Bonferroni's correction). The affinities of N3C5 and N03B11 for the various gp140 molecules are listed in Table 1. The first assay is useful in that it reflects the probable outcome when the MAbs are used in a standard ELISA, but it cannot be compared directly to the affinity assay due to important differences between the reagents, assays, and calculations.
Characterizing the epitopes of N3C5 and N03B11. The studies with the panel of Env, including gp140 and gp120 molecules, suggested that both MAbs bound to the gp41 region. To explore the epitope further, both MAbs were tested for binding to a set of linear 15-mer peptides representing the entire gp140 97CN54 sequence. Whereas MAbs 447-52D and 4E10 recognized their specific peptides in the V3 loop and membraneproximal extracellular region (MPER), respectively, the IgM MAbs did not bind to the linear peptides (data not shown). To determine whether the epitopes of N3C5 and N03B11 were linear or conformational in nature, the binding of both MAbs to native and denatured Env was examined by ELISA (Fig.  7A). Binding of the MAbs to both native and denatured gp140 97CN54 was detectable, but binding to the denatured mol- Binding of N3C5 to gp140 92UG037 could not be detected by solid-phase antigen at a concentration of 0.1 g · ml Ϫ1 ; thus, panel C refers only to the two different temperatures tested with a solid-phase gp140 concentration of 5 g · ml Ϫ1 . The boxes represent the median and 25th and 75th percentiles, the outlier values for K D (six replicates). ‫,ءء‬ P Ͻ 0.01 (unpaired two-tailed t test with Bonferroni's correction).
ecule was dramatically reduced, indicating that the epitopes involved are largely conformation dependent, as anticipated from the failure of the MAbs to bind the linear peptides. A control gp41 MAb, 5F3, showed a similar reduction in binding to denatured gp140 97CN54 , while a polyclonal serum, ARP422, bound equally well to both native and denatured gp140 97CN54 , demonstrating that both forms bound equally to the ELISA plate (data not shown).
To narrow down the region containing the N3C5 and N03B11 epitopes, both MAbs were studied in competition analyses against gp41 MAb 5F3 (3) and 50-69 (18) and against each other ( Fig. 7B; see Fig. S1 in the supplemental material for the positions of the 5F3 and 50-69 epitopes in relation to the structural features of the gp41 ectodomain). These studies showed a high degree of concordance between results from SPR assays and ELISA. MAb 5F3 inhibited the binding of N3C5 by 80 to 90% and of N03B11 by 98 to 100%. When the reciprocal competition experiments were performed, N3C5 inhibited the binding of 5F3 by 21 to 42%, whereas N03B11 inhibited 5F3 binding by only 8 to 15%. MAb 50-69 inhibited the binding of N3C5 in the SPR assay by 25% but did not significantly inhibit N3C5 when measured in the ELISA assay. The inhibition of N03B11 by 50-69 was more consistent in both assays, at 43% and 61% in ELISA and the SPR assay, respectively. In the reciprocal competition experiments, N3C5 inhibited the binding of 50-69 by SPR but not in the ELISA assay, whereas N03B11 did not inhibit 50-69 to a significant degree in either assay. The two IgM MAbs were tested against each a Values were determined with a solid-phase antigen concentration of 5 g · ml Ϫ1 (six replicates). ND, not determined (does not bind). other by SPR only, and it was found that N3C5 inhibited the binding of N03B11 by 61%, while N03B11 did not significantly inhibit the binding of N3C5. The binding of controls, the MAb IgG1b12 (to a conformational epitope in the CD4 binding site), and the polyclonal antibodies ARP422 and HIV Ig was not affected significantly by any of the gp41 ectodomain regionspecific MAbs (data not shown). Taken together, these results suggest that N3C5 and N03B11 both have conformational epitopes in the gp41 ectodomain proximal to the 5F3 and 50-69 epitopes, overlapping the gp41 immunodominant region. N3C5 and N03B11 are unlikely to share the same epitope due to differences in their ability to block the binding of one another and MAb 50-59, as well as differences in avidity of binding to gp140 92UG037 . A multiple sequence alignment of the gp41 ectodomains of isolates 97CN54, 97ZM651.8, 92UG037, 93BR029, IIIB, HAN2, and 92UG021 (see Fig. S1 in the supplemental material) showed that there were 24 amino acid substitutions between the clade C isolates that bound N03B11 and the clade A isolate that did not. These included 8 nonconservative, 6 semiconservative, and 10 conservative substitutions. These were scattered though-out the gp41 ectodomain but were more frequent on the C-terminal side of the 50-69 epitope. N3C5 bound to gp140 92UG037 despite these 24 amino acid substitutions. There were seven amino acid substitutions that differentiated the ability of N3C5 to bind gp140 from isolates 97CN54, 96ZM651.8, and 92UG037 from those isolates tested that it could not bind. These included only one nonconservative substitution (a T or K in isolates that bound N3C5 for an E or L at amino acid position 637 [position 130 in gp41 ectodomain]) in those that did not, one semiconservative substitution, and five conservative substitutions. Again, these changes were distributed throughout the gp41 ectodomain but were more fre-quent on the C-terminal side of the 50-69 epitope (five of seven substitutions).
Effects of complement on MAb neutralization of HIV-1 97CN54 . To determine whether the IgM MAbs possess antiviral activity, we tested their ability to reduce the infectivity of the autologous primary isolate HIV-1 97CN54 alone, or in the presence of fresh NHS as a source of complement. At a concentration of 100 g · ml Ϫ1 N3C5 demonstrated weak but statistically significant neutralizing activity, reducing the viral titer by 79% (P Ͻ 0.05; unpaired one-tailed t test), while N03B11 reduced viral titer by 59%, but this was not statistically significant (Fig. 8). By comparison, the positive control MAb 447-52D achieved 92% neutralization at a concentration of 10 g · ml Ϫ1 . Interestingly, when N3C5 was combined with fresh NHS, the degree of neutralization was half that seen with the combination with HI NHS, (P Ͻ 0.05; unpaired one-tailed t test). N03B11 appeared to inhibit infection in the presence of complement, but this result was not significant. The fact that neither MAb is potently neutralizing is not surprising given the paucity of NAb responses against primary HIV-1 isolates generated to vaccine candidates tested to date and given the fact that with the exception of the MPER, the gp41 ectodomain is largely obscured by gp120 on virion-associated Env.

DISCUSSION
Here we describe the use of mice producing human Ig to derive human IgM MAbs to a novel clade C HIV-1 Env vaccine candidate. This study demonstrates proof of principle for the use of such transgenic mice for the direct preparation of HuMAbs to HIV-1 Env-based vaccine antigens. While these mice might be used to produce novel HuMAbs, their utility in preclinical HIV-1 vaccine studies for screening of vaccine antigens may be restricted. The chief reason for this is the low titers of serum Abs (35) and the low volumes of serum that can be obtained for neutralization studies. Since neither IgM MAb reported here binds to representative clade B isolates, this suggests that these unique MAbs may be useful probes in further studies of the antigenicity of non-clade B Env and may also have utility in a clade-specific diagnostic setting. Since the human Ig-producing BAB5 mouse strain used here produces high-affinity IgM MAbs as apposed to IgG MAbs, these MAbs would be suitable as capture Abs when assessing human IgGbinding Ab responses. Other potential applications include use in Ab competition studies and as affinity chromatography reagents to purify Env.
The predominantly monomeric nature of N3C5 and N03B11 is likely to be due to the inability of the murine J chain to substitute adequately for the human J chain. Despite this and the lack of class switching from IgM to IgG in these mice, the IgM response underwent efficient affinity maturation as demonstrated by the derivation of the high-affinity MAbs N3C5 and N03B11. Prior studies using human Ig-producing transgenic mice have demonstrated their suitability for deriving MAbs against antigens such as human blood cells, tumor cell lines, haptens, the human acetylcholine receptor, and HIV-1 Env antigens (20,21,28,38,55). Interestingly, the immunization of the XMG2 XenoMouse strain with gp120 SF162 allowed the isolation of IgG2() MAbs that displayed neutralizing activity against the autologous primary isolate HIV-1 SF162 , FIG. 8. The neutralization of HIV-1 97CN54 by IgM MAbs N3C5 and N03B11. Bars represent percent neutralization compared to an equal concentration of an irrelevant Ab (D7324). The MAbs were also combined with heat-inactivated (HI) or fresh (CЈ) serum (NHS) to assess CMI. N3C5 and N03B11 were used at 100 g · ml Ϫ1 while the positive control, MAb 447-52D, was used at 10 g · ml Ϫ1 . For the CMI assay the activity in the presence of CЈ NHS was compared to that in the presence of HI NHS. Error bars represent the 95% CI. The results show pooled data from two independent experiments. * , P Ͻ 0.05 (unpaired two-tailed t test).
known to be relatively susceptible to neutralization, in a sensitive pseudovirus assay (21). The same mouse strain was also used to derive 39 MAbs to the full-length single-chain analogue of the gp120-CD4 receptor complex, composed of gp120 Ba-L and domains 1 and 2 of CD4. However, none of this panel of MAbs possessed neutralizing activity against HIV-1 Ba-L or HIV-1 SF162 , potentially underlining the insufficiencies of the immunogen as an HIV-1 vaccine candidate (20). We have taken the derivation of human MAbs in transgenic mice further in our model system by the use of a more sophisticated trimeric gp140 immunogen from a highly relevant primary clade C isolate that we demonstrated to be functional and appropriately folded. As far as we are aware, this is the first publication regarding MAbs of any species induced to clade C Env immunogens. While neither of our novel MAbs are potently neutralizing we feel it is too early to make inferences as to the suitability of gp140 97CN54 as a vaccine candidate. Such judgments would be better informed by studies in rabbits or macaques. The fact that both IgM MAbs bind to somewhat similar epitopes might reflect a bias in the screening ELISAs and suggests that further optimization of the screening and selection assays would be beneficial. MAb N3C5 bound soluble recombinant uncleaved gp140 97CN54 with high affinity but only weakly neutralized the autologous isolate. This finding suggests that N3C5 cannot bind the functional virion-associated Env with a similarly high affinity. It is known that the ability of MAbs to bind virion-associated Env is necessary but not sufficient for neutralization (4,32,36), while the ability to bind any epitope on a functional spike is a dominant determinant of neutralization (64). Indeed, many nonneutralizing MAbs appear to bind nonfunctional forms of virion-associated Env; the most important of these have been identified as gp120-gp41 monomers (32) and gp120-depleted gp41 stumps (4). However, other forms, including the uncleaved gp160 precursor, alternative trimeric isoforms that expose the nonneutralizing face of gp120, or gp120-gp41 dimers or tetramers, may also serve as decoys (32). Since N3C5 was raised specifically against a trimeric uncleaved gp140, the most likely explanation for its failure to neutralize the virus potently is that the epitope is well-exposed on the trimeric recombinant uncleaved form but is either partially occluded or that it is conformationally altered on the cleaved, virion-associated form. Structural alterations in the epitope may arise from cleavage at the gp120-gp41 cleavage site, interactions between the MPER and the viral lipid envelope, or may depend on an intact intraviral gp41 tail, the presence of which has been demonstrated to modulate epitope exposure and viral neutralization sensitivity (22,24). These potential differences in epitope conformation between the immunogen and the virionassociated Env could result in unfavorable binding kinetics between N3C5 and the virus. These could include a slow on rate or fast off rate, thus reducing neutralization potency (47). Moreover, other MAbs to this region are generally not neutralizing (15,63,66) or, similar to N3C5, are weakly neutralizing (11,12), suggesting that this may be a general phenomenon associated with this region of Env. If steric occlusion of MAb binding is responsible for the weak neutralization activity of N3C5, then Fab fragments may be more potent inhibitors, in a manner analogous to that observed with CD4-induced epitope-binding MAbs such as 17b and X5 (26). The fact that N3C5 was weakly neutralizing but enhanced HIV-1 infection in the presence of human complement reinforces the concept that this MAb binds virion-associated Env, although gp41 stumps present a potential target for this MAb on the virion that would allow it to bind and activate complement without neutralizing virus infectivity. Moreover, MAbs to epitopes defined as gp41 cluster I and cluster II enhanced infection in the presence of complement in a manner similar to that observed with N3C5 (39). The inability of N3C5 to mediate CMI may be due to its monomeric nature. Pentameric IgM is a potent activator of the lytic complement cascade, and if more of the MAb had been in the pentameric form and this was not sterically occluded from the epitope, we might have observed CMI. Instead, it is likely that binding of monomeric IgM led only to complement-mediated opsonization followed by increased infection of target cells expressing complement receptors, such as macrophages (40).
Our attempts to define the epitope of these MAbs lead us to conclude that both MAbs bind distinct but related conformational epitopes in the immunodominant region of the gp41 ectodomain. Interestingly, neither MAb we isolated here bound to the two clade B gp140 molecules we tested, indicating a lack of epitope conservation in B clade viruses. Moreover, MAb N3C5 bound to the geographically distant heterologous clade C strain gp140 96ZM651.8 (a Zambian isolate) with an affinity almost 1 log 10 -fold higher than that of binding to the autologous isolate, suggesting that gp140 96ZM651.8 presents the epitope in a more favorable context. A related study comparing the binding of a panel of MAbs to eight clade C gp140s has indicated that gp140 96ZM651.8 is recognized by the widest range of MAbs with the strongest affinities, while gp140 97CN54 shows a less favorable antigenic topology (S. A. Jeffs, S. Larsen, and S. M. Vieira, unpublished results). In all, 24 amino acids varied between isolates that bound N03B11 and those that did not. Only seven amino acids varied between strains that bound N3C5 and those that did not. Although we cannot rule out the possibility that different substitutions affected the binding in different ways, it is likely that the critical contact residues are among those identified.
In conclusion, we describe the isolation from transgenic mice immunized with clade C HIV-1 gp140 of two human MAbs, one of which is weakly neutralizing for the autologous virus. We propose that such mice might be useful for the future production of HuMAbs against antigens of interest and that the novel MAbs obtained will be of use as probes for Env antigenicity and function.

ACKNOWLEDGMENTS
We acknowledge the support of the EU Consortium Eurovac and the UK Medical Research Council (grant G0000635). S.L.D. was supported by a grant from the AICR to M. S. Neuberger.
We thank M. Brüggemann and Michael S. Neuberger for provision of the BAB5 mice; J. E. Robinson, D. R. Burton, D. Katinger, and S. Zolla-Pazner for their generous gifts of MAbs; G. Pantaleo and Eurovac for CN54 gp120 sequence-derived synthetic peptides; and the NIBSC CFAR and NIH Reagent Program for reagent supply. We also thank C. Moog for providing HIV-1 97CN54 , M. Puklavec for providing the NS-1 murine myeloma cells, and A. Akoulitchev for assistance with the liquid chromatography.
We have no conflicting commercial interests relevant to the work published here.