|Home | About | Journals | Submit | Contact Us | Français|
Broadly HIV-1–neutralizing antibodies (BnAbs) display one or more unusual traits, including a long heavy chain complementarity-determining region 3 (HCDR3), polyreactivity, and high levels of somatic mutations. These shared characteristics suggest that BnAb development might be limited by immune tolerance controls. It has been postulated that HIV-1–infected individuals with autoimmune disease and defective immune tolerance mechanisms may produce BnAbs more readily than those without autoimmune diseases. In this study, we identified an HIV-1–infected individual with SLE who exhibited controlled viral load (<5,000 copies/ml) in the absence of controlling HLA phenotypes and developed plasma HIV-1 neutralization breadth. We collected memory B cells from this individual and isolated a BnAb, CH98, that targets the CD4 binding site (CD4bs) of HIV-1 envelope glycoprotein 120 (gp120). CH98 bound to human antigens including dsDNA, which is specifically associated with SLE. Anti-dsDNA reactivity was also present in the patient’s plasma. CH98 had a mutation frequency of 25% and 15% nt somatic mutations in the heavy and light chain variable domains, respectively, a long HCDR3, and a deletion in the light chain CDR1. The occurrence of anti-dsDNA reactivity by a HIV-1 CD4bs BnAb in an individual with SLE raises the possibility that some BnAbs and SLE-associated autoantibodies arise from similar pools of B cells.
Broadly HIV-1–neutralizing antibodies (BnAbs) have been isolated that bind to multiple epitopes on the envelope glycoproteins gp120 and gp41 (reviewed in ref. 1). 2G12 recognizes a posttranslational glycan epitope on gp120 (2, 3). CH103–CH106 (4), b12 (5, 6), VRC01 (7), and numerous other mAbs with VRC01-like characteristics, such as VRC03, VRC-PG04, CH30–CH34, and the HAAD motif antibodies (7–9), recognize the CD4 binding site (CD4bs). HJ16 binds to an epitope near the CD4bs (10). PG9 and PG16, CH01–CH04, and PGT141–PGT145 recognize conformational epitopes in the gp120 V1/V2 region with binding dependence on V2 asparagine residue 160 (11–13). PGT121–PGT123, PGT135–PGT137, PGT125–PGT128, PGT130, and PGT131 recognize a diverse set of carbohydrate or carbohydrate-dependent epitopes in the gp120 V3 region (13). 3BC176 and 3BC315 recognize a conformational HIV-1 spike epitope in the proximity of the V3 loop and the CD4i site that is exposed in part by CD4 binding (14). Finally, 2F5 (2, 15, 16), 4E10 (2, 17), 10E8 (18), and Z13 (19) bind to the membrane proximal external region (MPER) of gp41.
Each of these antibodies displays one or more unusual characteristics, such as polyreactivity with human and/or nonhuman antigens, long heavy chain complementarity-determining region 3 (HCDR3) loops, and high levels of somatic mutations (1, 13, 20–23). These traits suggest that the development of these types of BnAbs may be limited by immune tolerance controls that impede the required tortuous or polyreactive BnAb maturation pathways (20, 21, 24). This notion is supported by a number of observations in 2F5 VH×VL knockin mice: targeted expression of the 2F5 VHDJH/VLJL rearrangements triggered a near-complete B cell developmental blockade at the pre-B to immature B cell stage (25); even when clonal deletion mechanisms were circumvented, 2F5 VHDJH/VLJL-expressing broadly neutralizing B cell rescue was limited by κ chain editing and an anergic phenotype (26); and MPER-specific serum neutralizing IgG responses were elicited in 2F5 knockin mice immunized with MPER-lipid complexes by rescuing the anergic self-reactive 2F5-expressing B cells that survived the B cell developmental blockade (26).
We have previously suggested that the paucity of subjects developing BnAbs may be due to the frequent deletion of B cell precursors that acquire autoreactivity during their maturation process (at either early or late stages) and hypothesized that HIV-1–infected subjects with autoimmune diseases might be capable of developing BnAbs in the context of their autoreactive humoral response (24, 26–28). The observations that HIV-1 infection is reported with a disproportionately low frequency among subjects with SLE support this hypothesis (29–34).
To date, there have been no reports of studies of the HIV-1 antibody repertoire of SLE subjects from whom BnAbs have been isolated. Here, we describe CH98, an HIV-1 CD4bs BnAb isolated from an HIV-1–infected individual with SLE (subject CH5329; see Methods). CH98 displayed a number of unusual antibody traits shared by autoimmune disease autoantibodies.
In a multiclade, multitier panel of HIV-1 strains, consisting of 4 tier-1A, 4 tier-1B, and 34 tier-2 isolates, including 9 transmitted/founder viruses, plasma from subject CH5329 neutralized 41 of 42 HIV-1 strains (97.6%), with a mean ID50 titer of 925 (range, 21–9,204; Figure Figure1A).1A). The only HIV-1 strain that was not neutralized, G.P0402, was a clade G. CH5329 plasma was screened in 2 different but complementary assays for CD4bs activity. CH5329 plasma bound to the HIV-1 Env resurfaced core 3 (RSC3) protein with a mean endpoint titer 5-fold higher than that of the RSC3Δ371I/P363N protein (Figure (Figure1B1B and Supplemental Figure 1; supplemental material available online with this article; doi: 10.1172/JCI73441DS1), indicative of the presence of CD4bs mAbs (7). The RSC3 protein preferentially blocked the neutralization ability of CD4bs mAbs such as VRC01, but not mAbs directed against other sites, such as the gp41-binding 2F5 BnAb. However, the ability of RSC3 protein to block neutralizing activity in HIV-infected sera is rare (35) and indicative of the presence of CD4bs neutralizing antibodies (7, 8). Therefore, the 30%–55% reduction we observed in total serum neutralization against 2 different HIV-1 strains, HxB2 and JRFL (Figure (Figure1C),1C), indicated that CD4bs antibodies markedly contribute to plasma neutralization.
mAb CH98 VHDJH and VLJL gene segments were isolated from a clonal memory B cell culture containing 29 ng/ml IgG at the end of stimulation with culture supernatant that bound to RSC3 but not RSC3Δ371I, and were expressed as a full recombinant IgG1 mAb (11, 36, 37). CH98 neutralized 27 of 43 HIV-1 isolates (62.8%), with a median IC50 neutralization level of 4.2 μg/ml (range, <0.02–49 μg/ml), accounting for 65.9% of the observed plasma neutralization breadth (Figure (Figure1A).1A). No other antibodies were retrieved that bound to RSC3 or neutralized the A.Q23.17 HIV-1 strain, which was selected to screen culture supernatants for heterologous neutralizing activity (see Methods). Thus, the frequency of peripheral blood circulating CD4bs neutralizing antibodies in this subject was estimated to be 0.003%.
CH98 uses the VH3–30 gene segment, notably not the VH1–2 or VH1–46 segments commonly associated with VRC01-like CD4bs BnAbs (8, 9), paired with the Vλ2–23 light chain gene segment. CH98 is highly mutated, with VHDJH and VLJL mutation frequencies of 25.3% and 15.8%, respectively, and displays a long HCDR3 of 21 aa (Supplemental Table 1). From 454 pyrosequencing, we identified 4 additional VHDJH rearrangements in the CH98 clonal lineage, and all were highly mutated and shared numerous mutations with the CH98 VHDJH (Supplemental Figures 2 and 3).
The CH98 light chain CDR 1 (LCDR1) is 27 nt long and underwent a 15-nt deletion preceded by an activation-induced cytidine deaminase (AID) hotspot (AGT; Supplemental Figure 4). TGT motifs are also present near the start and the end of this deletion (Supplemental Figure 4), raising the hypothesis that slipped-strand synthesis may have occurred at this position. It is also worth noting that the CH98 VL/JL gene segment rearrangement had a very long n-region insertion of 24 nt in the LCDR3 that were not encoded by either of the VL2–23 or JL3 germline sequences (Supplemental Figure 4).
CH98 bound to RSC3 protein (IC50, 0.51 μg/ml), but not to RSC3Δ371I/P363N; in contrast to the CD4bs BnAb VRC01, CH98 did not bind to RSC3 G367R mutant or to the gp120 outer domain stabilized protein OD4.0 (Supplemental Table 2). Binding to WT YU2 gp120 Env (IC50, 0.03 μg/ml) was abrogated by the D368R mutation, which disrupts a key contact residue within the CD4bs, but not by the I420R mutation (IC50, 0.11 μg/ml), which disrupts the CCR5 coreceptor binding site (Supplemental Table 2). Also, CH98 potently neutralized the HIV-1 YU2 strain (IC50, 0.9 μg/ml). Therefore, we epitope mapped CH98 on a panel of YU2 gp120 core protein mutants previously used to map other mAbs binding within and outside the CD4bs (38, 39). We compared the binding profile of CD4bs BnAb CH98 with those of other CD4bs BnAbs — CH103 (4), VRC01 (7), CH31–CH34 (8), and b12 (5, 6) — as well as non-neutralizing CD4bs b6 mAb and CD4-Fc to 31 YU2 gp120 core single-point mutants (Figure (Figure2).2). Single-point mutations outside the CD4bs did not affect CH98 binding. Within the CD4bs, mutations in position 365 and 368 abrogated binding of all CD4bs BnAbs tested, but not mAb b6, and mutations in positions 282, 461, and 473 reduced CH98 binding to 55%, 64%, and 31%, respectively (Figure (Figure2).2). Overall, CH98 was the BnAb least sensitive to the mutations represented in the panel, which indicates that these positions are less relevant for CH98 binding to its cognate epitope than for the other CD4bs BnAbs. The only antibodies sensitive to all the mutations that also affected CH98 binding were the BnAbs CH103 and VRC01.
CH103 and VRC01 bind to the CD4bs with 2 distinct modes of epitope recognition: CH103 uses a loop dominated mode of interaction including all VHDJH and VLJL CDRs (4), whereas VRC01 contacts the CD4bs by mimicking CD4 with a large footprint that is dominated by HCDR2 and encompasses both heavy and light chains (40). To determine whether CH98 falls into one of these categories, we threaded the CH98 heavy and light chains onto the gp120-complexed structure coordinates of CH103 (4) and VRC01 (40).
CH98 threading on the VRC01 gp120-complexed structure resulted in steric clashes between the LCDR3 of CH98 and the D loop of gp120 (Figure (Figure3A).3A). Relaxation of the model to alleviate the clashes resulted in a complex with a binding energy score of –31.96 Rosetta energy units (REU; see Methods), which was higher (i.e., less favorable) than those of both the solved VRC01-gp120 complex and the CH98 complex modeled in the CH103 mode of recognition (–40.84, and –37.86 REU, respectively). Conversely, CH98 threading onto the gp120-complexed CH103 crystal structure resulted in a model compatible with a CH103-like mode of antigen recognition (Figure (Figure3B).3B). Between the 2 mAbs, aa sequence homology was 36% and 56% for the VHDJH and VLJL, respectively (Supplemental Figure 5 and Supplemental Table 3). This model supported the observed binding data (Figure (Figure2)2) and predicted that all the heavy and light chain CDRs of CH98 make direct contact with the 5 residues affecting CH98 binding to the YU2 gp120 core (Supplemental Table 4).
In particular, D461 was predicted to contact LCDR1 (Supplemental Figure 6A), which is 9 aa long and underwent a 5-aa deletion. Therefore, we asked whether a LCDR1 deletion-reverted CH98 would be compatible with the gp120-complexed CH103 mode of recognition. Adoption of the canonical conformation for a LCDR1 of length 14, as defined by North and colleagues (41), allowed the deletion-reverted LCDR1 to avoid steric clashes at the gp120 interface; however, the binding energy score of the complex was less favorable than that of the mature CH98 complex (–35.09 versus –37.86 REU; Supplemental Figure 6B). Our interpretation is that the 15-nt deletion in the CH98 LCDR1 may have played a favorable role during the process of affinity maturation of CH98.
In a panel of 9 autoantigens associated with presence of autoimmune diseases (Sjogren’s syndrome antigens A and B [SSA and SSB, respectively], Smith antigen [Sm], ribonucleoprotein [RNP], centromere B [Cent B], histone, scleroderma 70 [Scl 70] and Jo-1 proteins, and dsDNA), CH98 reacted with dsDNA (Figure (Figure4A).4A). CH98 binding to dsDNA was comparable to that previously observed for the CD4bs BnAb CH103 (Supplemental Figure 7). In the same assay, CH5329 plasma from the same specimen from which CH98 BnAb was isolated was found to be positive for SSA and dsDNA antibodies (mean ± SD, 268 ± 29 and 163 ± 22 response units [RU], respectively; Figure Figure44B).
In HEp-2 cell IFA staining, CH98 displayed diffuse fine granular cytoplasmic staining, exhibited stronger binding to dividing cells, and bound to intracellular vesicles in a subset of HEp-2 cells (Figure (Figure4,4, C–E). CH98 reactivity with 9,400 additional human proteins was screened using microarray technology (28). CH98 showed polyreactivity with numerous human antigens, as measured by 1 log stronger binding to more than 90% of the test proteins than the non-polyreactive control antibodies 151K and palivizumab. CH98 also bound with higher affinity, measured as a >2 log increase in binding compared with antibody controls, to STUB1 (also known as CHIP), an E3 ubiquitin-protein ligase (Figure (Figure4,4, F and G). Binding of CH98 and CH5329 plasma to STUB1 were confirmed in direct-binding ELISA with an endpoint titer of 3.3 μg/ml for CH98 and plasma titer of 270 (Supplemental Figure 8, A and B). We previously demonstrated that members of the CD4bs BnAb CH103 clonal lineage displayed reactivity with dsDNA and with UBE3A, another E3 ubiquitin-protein ligase (4). Such reactivity was also shared by the CD4bs BnAb VRC01 (4). However, these BnAbs did not bind to STUB1, and CH98 did not bind to UBE3A (data not shown).
Taken together, these findings suggested that the immune system dysregulation that allows for production of autoreactive antibodies during SLE may have played a permissive role in the development and maturation of the BnAb CH98. Moreover, the similarities between CH98 and the SLE autoantibodies of long HCDR3 and dsDNA reactivity are suggestive of similar pools of B cells and mechanisms of their origin.
Here, we describe an individual who, when studied 14 years after HIV-1 transmission and 6 years after onset of SLE, controlled her viral load off antiretroviral therapy (ART) and developed plasma BnAb activity.
An open question regarding the development of BnAbs in HIV-1–positive subjects is whether acquisition of autoreactivity is a common prerequisite for the development of neutralization breadth. We have previously described the CH103 CD4bs BnAb isolated from an SLE-negative, chronically HIV-1–infected individual and demonstrated that antibodies in the CH103 clonal lineage acquired autoreactivity with a number of autoantigens, including dsDNA, concurrently with acquisition of HIV-1 neutralization breadth (4). We have also described an individual in whom the onset of gp41 MPER 2F5-like BnAb plasma activity was temporally associated with the development of plasma dsDNA and Jo-1 autoantigen reactivity, without appearance of any clinical manifestation of autoimmune disease (42). Thus, in some individuals, acquisition of BnAb activity is clearly associated with the acquisition of autoreactivity.
Antibodies to dsDNA are a hallmark of SLE and are present in approximately 70% of subjects with SLE, but only 0.5% of SLE-negative individuals (32). Conversely, dsDNA antibodies are not a typical finding during HIV-1 infection, but have been commonly reported in the plasma of the few SLE subjects with HIV-1 infection (30, 33). The presence of plasma dsDNA antibodies has been documented in subjects diagnosed with SLE both after and before HIV-1 diagnosis (91.7% and 64.3%, respectively) (30). The incidence of plasma BnAbs in these reported cases, however, is unknown.
We report herein a monoclonal HIV-1 BnAb, CH98, isolated from a chronically HIV-1–infected individual who subsequently developed SLE. CH98 neutralized HIV-1 through binding to the CD4bs on HIV-1 gp120 and reacted with dsDNA at levels comparable to those previously observed for the CD4bs BnAb CH103. CH98 accounted for only a portion of the plasma neutralization breadth (65.9%). This observation, in conjunction with the finding that RSC3 plasma absorption only partially abrogated plasma neutralization (30%–55%), implies that other neutralizing antibody specificities contributed to the overall plasma neutralization breadth in this individual. These data are in line with previous observations indicating that plasma neutralization breadth can be mediated by a polyclonal response comprising antibodies with distinct specificities and variable degrees of breadth (43–45). Using a combination of epitope mapping and threading of the CH98 heavy and light chains onto the gp120-complexed structure coordinates of CD4bs BnAbs, we determined that CH98 likely uses a CH103-like Env loop mode of epitope recognition that involves all the heavy and light chain CDRs. Conversely, CH98 threading onto the gp120-complexed structure of VRC01 resulted in steric clashes between the LCDR3 of CH98 and the D loop of gp120. This result was not unexpected: VRC01 and all VRC01-like antibodies had unusually short LCDR3s (5 residues), and it has been suggested that the VRC01-like orientation of gp120 binding does not leave enough space between the antibody and the gp120 D loop to accommodate longer LCDR3s (46). Indeed, CH98 has a 12-residue-long LCDR3, and therefore our model is compatible with this hypothesis.
CH98 was also characterized by a short LCDR1 (9 aa) that underwent a 5-aa deletion during the process of affinity maturation. This 15-nt deletion in CH98 LCDR1 was preceded by an AID hotspot. AID hotspots regulate secondary antibody diversification and are directly involved in the process of somatic hypermutation (47, 48). Since we predicted a less favorable binding energy score of the deletion-reverted CH98 complex compared with the mature CH98 complex, our interpretation is that, similar to our prior observation for CH103 (4), a single macromutation may have improved binding to the cognate epitope. In addition, LCDR1s shorter than 11 aa are rare (46), but commonly found among CD4bs BnAbs (4, 46), and both CH103 and VRC01 also have deletions in their LCDR1s (4, 40). We interpret the disproportionate frequency of short LCDR1s among CD4bs BnAbs as being indicative of a common selection process of CD4bs BnAbs that favors antibodies with shortened LCDR1s.
This is the first chronically infected SLE individual with concomitant HIV-1 infection that we have been able to study, and it was striking that this individual also had robust plasma BnAb activity, given that the incidence of chronically HIV-1–infected subjects with this level of plasma neutralization breadth is only approximately 20% (49–52).
In a literature review from 1981 to 2012 by Carugati and colleagues, of 53 HIV-1/SLE patients, 54.7% were diagnosed with HIV-1 before SLE (33). The diagnosis of HIV-1 in our subject preceded that of SLE, and we cannot conclude that the CH98 maturation was initiated as part of the SLE-induced autoreactive response, nor can we exclude the possibility that silent SLE was present before the time of HIV-1 diagnosis.
When generated, BnAbs appear late (2–4 years) after HIV-1 transmission, and serum neutralization breadth correlates with viral load levels (49). This has been interpreted as an indication that persistent antigen stimulation is required to develop neutralization breadth (53–55). In stark contrast, our study subject developed BnAbs while controlling viremia, which suggests that acquisition of HIV-1 neutralization breadth may not always require chronic stimulation with high levels of HIV-1 viremia. A possible explanation is that the maturation of CH98 was not hindered in this subject by immune tolerance control mechanisms, thus facilitating the acquisition of mutations required to achieve neutralization breadth, possibly driven by exposure to both HIV-1 and autoantigens (21, 24, 27). The low viral load and normal CD4 count of this subject while off ART could not be explained by HIV-1–controlling HLA types (56) or by HIV-1 infection with a nef-defective strain (57, 58); studies are in progress to determine whether the presence of SLE contributed to virus control.
Kobie and colleagues recently identified an autoreactive plasma antibody population (9G4 anti-idiotype antibodies) that is commonly found in SLE subjects and is responsible for serum HIV-1 neutralization breadth in ART-naive, SLE-negative, chronically HIV-1–infected subjects; the authors proposed that this kind of antibody may derive from the same pool of B cells of SLE-induced antibodies (34). The BnAb CH98 differs from 9G4-positive antibodies in that it uses a VH3–30 gene segment, not VH4–34, which defines 9G4 antibodies. Therefore, our findings demonstrated that autoreactive antibody populations beyond 9G4-positive antibodies have the potential to broadly neutralize HIV-1 strains in SLE subjects.
In conclusion, we demonstrated that a virus-controller SLE patient with chronic HIV-1 infection developed the highly mutated CD4bs BnAb CH98, and this antibody displayed characteristics typical of SLE-induced autoantibodies. Our findings support the hypothesis that lax mechanisms of tolerance control associated with SLE may have favored the maturation of this BnAb by favoring its evasion from deletion during the process of somatic maturation and the acquisition of reactivity with self-antigens. Moreover, our data support the concept that BnAbs develop from a similar pool of B cells as do autoimmune B cells in SLE.
Subject CH5329 is a female patient, aged 33 years at the initiation of the study. At 19 years of age, the patient was diagnosed with HIV-1; at 27 years of age (8 years after HIV-1 diagnosis), the subject was diagnosed with SLE. The patient sought health care at the Duke University Medical Center Clinics in 2009 (age 31 years). At admission, her general health conditions were overall good with no fever, chills, night sweats, nausea, vomiting, or diarrhea. CD4 count was 460 cells/mm3. She had not received any ART. SLE diagnosis was confirmed at admission based on American College of Rheumatology criteria for SLE (59): malar rash, discoid rash, photosensitivity, arthritis, and presence of plasma speckled anti-nuclear antibody titer of 160. The subject also reported a history of seizures that preceded the diagnosis of both SLE and HIV-1 infection, as well as a history of asthma, which was controlled by daily triamcinolone and albuterol. In 2009, her plasma was negative for SSA, SSB, RNP, Sm, and dsDNA antibody. Serum complement levels were normal. Her SLE was treated with plaquenil from 2005 to January 2008 and 7.5 mg methotrexate once weekly from 2006 to December 2012, when plaquenil was restarted and continued up to date.
The patient was enrolled in 2011 in the CHAVI008A study of chronic HIV-1–infected individuals. Plasma, PBMCs, and leukapheresis samples were collected and studied. At enrollment, the subject had received no ART, and her CD4 count and viral load were 568 cells/mm3 and 4,150 HIV RNA copies/ml, respectively. Her MHC class I phenotype was HLA-B7 and HLA-B40. Analysis of 64 HIV-1 3′-half genome sequences generated by the single genome amplification method showed the intact nef gene in all sequences (Supplemental Figure 9). In 2011, her plasma was positive for dsDNA antibodies; negative for SSA, SSB, Sm, RNP, Scl 70, Jo-1, Cent B, and histone antibodies; and showed an ANA-positive pattern in HEp-2 cell intracellular fluorescence staining with an endpoint titer of 320.
IgG+ memory cells were isolated from PBMCs as previously described (11). Briefly, IgG+ memory B cells were isolated from frozen PBMCs by 2 rounds of separation using magnetic beads (11). First, PBMCs were incubated for 30 minutes at 4°C with PE-conjugated anti-CD2 (catalog no. 555327), anti-CD14 (catalog no. 555398), anti-CD16 (catalog no. 555407), anti-CD235a (catalog no. 555570), and anti-IgD (catalog no. 555779) antibodies (all from BD Biosciences — Pharmingen), followed by a 15-minute incubation with anti-PE microbeads (Miltenyi Biotec) at 4°C. After washing, cells were passed through magnetic columns using the Deplete_S program of the AutoMACs separator (Miltenyi Biotec). The negative fraction was then incubated for 15 minutes with anti-IgG microbeads (catalog no. 130-047-501; Miltenyi Biotec). After washing, IgG+ memory B cells were selected using the Possel_S program of the AutoMACs separator (11). All steps of the separation procedure were performed using cold PBS with 0.1% BSA buffer (PBSB). IgG+ memory B cells were stimulated for 14 days in vitro to induce proliferation and differentiation into antibody-secreting cells for functional studies on the culture supernatants, using a previously described stimulation protocol (11). Briefly, cells were resuspended in complete medium containing 2.5 μg/ml CpG ODN2006 (tlrl-2006; InvivoGen), 5 μM CHK2 kinase inhibitor (Calbiochem/EMD Chemicals), and EBV (200 μl supernatant of B95-8 cells/104 memory B cells) and incubated in bulk overnight at 37°C in 5% CO2 (11). After overnight incubation, 28,800 viable IgG+ memory B cells were manually transferred at an average density of 3 cells/well in 96-well round-bottomed tissue culture plates containing ODN2006, CHK2 kinase inhibitor, and 5,000 irradiated (7,500 cGy) CD40 ligand–expressing L cells per well. The cell density of 3 cells per well was chosen to achieve the expansion of a single cell in the majority of wells, according to Poisson distribution and previous experimental observations using this method. Medium was refreshed after 7 days, and supernatants were harvested at day 14 (11). Cell culture supernatants were screened for IgG levels, differential binding to RSC3 and RSC3Δ371I, and neutralization of the clade A.Q23.17 HIV-1 strain. This screening strategy was selected to isolate BnAbs based on the observed plasma differential binding to RSC3 and RSC3Δ371I and the high plasma ID50 A.Q23.17 HIV-1 strain neutralization. The RSC3 gp120 core antigen is a protein designed to retain the antigenic structure of the neutralizing surface of the CD4bs while eliminating other antigenic regions of the HIV-1 envelope glycoprotein, and the deletion of a single aa in position 371 (RSC3Δ371I; originally described as ΔRSC3) abrogates or heavily reduces the binding of CD4bs BnAbs (7). Since CD4bs BnAbs such as VRC01, VRC01-like, and CH103 have been previously isolated from memory B cells that bound to RSC3 using antigen-specific cell sorting (4, 7, 44, 60), we decided to use this protein to screen cultured memory B cells cultured to retrieve CD4bs BnAbs using our culture system.
RNA from positive cultures was extracted using standard procedures (RNeasy minikit; Qiagen), and the genes encoding Ig VHDJH and VLJL rearrangements were amplified by RT and nested PCR without cloning, using a previously reported method (11, 36). Briefly, reverse transcription was performed at 55°C for 1 hour after the addition of 50 U/reaction of Superscript III RT (Life Technologies) and 0.5 μM human IgG, IgM, IgD, IgA1, IgA2, Igκ, and Igλ constant-region primers. Separate reactions were used to amplify individual families of VH, Vκ, and Vλ genes from the cDNA template; this was performed using 2 rounds of PCR: first round, 5 μl RT reaction product, 5 U HotStar Taq Plus (Qiagen), 0.2 mM deoxynucleoside triphosphates (dNTPs), 0.5 μM nested constant-region primers (IgH consisting of IgM, IgD, IgG, IgA1 and IgA2; Igκ or Igλ), and matched variable-region primers; second round, 2.5 μl first-round reaction product, 5 U HotStar Taq Plus, 0.2 mM dNTPs, 0.5 μM nested constant-region and nested variable-region primers. First-round PCR was cycled as follows: 95°C for 5 minutes; 35 cycles at 95°C for 30 seconds, 55°C (VH and Vκ) or 50°C (Vλ) for 60 seconds, and 72°C for 90 seconds; and 1 cycle at 72°C for 7 minutes. Second-round PCR was similar, except that the extension step was performed at 58°C (VH), 60°C (Vκ), or 64°C (Vλ). Samples of VH, Vκ, and Vλ chain PCR products were analyzed on 2% agarose gels (Qiagen). Products were sequenced in the forward and reverse directions using a BigDye sequencing kit on an ABI 3700 instrument (Applied Biosystems). Sequence base calling was performed by using Phred. V, D, and J region genes and mutations were analyzed using the SoDA information system (61). All primers used for the second round of PCR included tag sequences at the 5′ end of each primer to permit the assembly of the VH and VL genes into functional linear Ig gene expression cassettes by overlapping PCR, as reported elsewhere (36).
Isolated Ig VHDJH/VLJL gene pairs were assembled by PCR into linear full-length Ig heavy- and light-chain gene expression cassettes as previously described (36). The human embryonic kidney cell line 293T (ATCC) was grown to near confluence in 6-well tissue culture plates (BD) and transfected with 2 μg/well purified PCR-produced IgH and IgL linear Ig gene expression cassettes using Effectene (Qiagen). Supernatants were harvested from the transfected 293T cells after 3 days of incubation at 37°C in 5% CO2.
Neutralizing antibody assays with TZM-bl cells were performed as previously described (62). Purified recombinant antibodies were tested starting at a 50-μg/ml final concentration and titrated using serial 3-fold dilutions. Plasma antibody was tested starting at a 1:20 dilution. Pseudotyped lentiviruses were added to the antibody dilutions at a predetermined titer to produce measurable infection and incubated for 1 hour. TZM-bl cells were added and incubated for 48 hours before lysis, after which the supernatant was measured for firefly luciferase (Britelite Plus; Perkin-Elmer) activity using a luminometer. Data were calculated as the reduction in luminescence compared with control wells and reported as IC50 (in μg/ml) for purified recombinant antibodies or ID50 titer (as 1/dilution) for plasma (62). Plasma neutralization resistance was defined as an ID50 titer <20. For further information on the viruses used for neutralization screening, see Supplemental Table 5.
Direct binding assays were performed in 384-well plates as previously described (11). Briefly, plates were coated overnight at 4°C with 15 μl purified proteins at optimized concentrations and blocked with assay diluent (PBS containing 4% [w/v] whey protein, 15% normal goat serum, 0.5% Tween 20, and 0.05% sodium azide) for 2 hours at room temperature. Primary culture supernatants were screened at a 1:3 dilution in assay diluent, whereas purified recombinant antibodies were tested using 3-fold serial dilutions starting at 100 μg/ml. 10 μl of primary antibodies were added to each well and incubated for 2 hours at room temperature. Plates were developed using 15 μl/well horseradish peroxidase–conjugated goat anti-human IgG Fc (catalog no. 109-035-098; Jackson ImmunoResearch Laboratories) at a 1:9,000 dilution in 5% goat serum and PBS for 1 hour, followed by 30 μl/well of SureBlue Reserve TMB One Component microwell peroxidase substrate (catalog no. 53-00-03; KPL). Development was stopped with 15 μl/well of 0.1 M HCl, and plates were read at OD450 with a VersaMax microplate reader (Molecular Devices).
Indirect immunofluorescence binding of mAbs or plasma to HEp-2 cells (Zeuss Scientific) was performed as previously described (24). Briefly, 20 μl of antibody at 50 μg/ml or plasma in serial dilutions was aliquoted onto a predetermined spot on the surface of a slide (ANA HEp-2 kit). After incubation for 25 minutes at room temperature and washes, 20 μl of secondary antibody (goat anti-human Ig FITC at 20 μg/ml; Southern Biotech) was added to each spot and incubated in a humid chamber for 25 minutes in the dark. After washing and drying, a drop of 33% glycerol was added to each spot, and the slide was covered with a 24- by 60-mm coverslip. Images were taken on an Olympus AX70 instrument with a SpotFlex FX1520 CCD and with a UPlanFL 40× 0.75-NA objective at 25°C in the FITC channel using SPOT software. All images were acquired for 8 or 10 seconds. Image layout and scaling were performed with Adobe Photoshop without image manipulation.
Binding of CH98 to human host cellular antigens was performed using a ProtoArray 5 microchip (Life Technologies) according to the manufacturer’s instructions. In brief, ProtoArray 5 microchips were blocked and exposed to 2 μg/ml CH98 or a human myeloma protein (151 K; Southern Biotech), or palivizumab for 90 minutes at 4°C. Protein-antibody interactions were detected by 1 μg/ml Alexa Fluor 647–conjugated anti-human IgG. The arrays were scanned at 635 nm with 10-μm resolution, using 100% power and 600 gain (GenePix 4000B scanner; Molecular Devices). Fluorescence intensities were quantified using GenePix Pro 5.0 (Molecular Devices). Lot-specific protein spot definitions were provided by the microchip manufacturer and aligned to the image.
A panel of 32 HIV-1 YU2 gp120 core mutants displayed on S. cerevisiae (38) was probed for binding to CD4bs mAbs and CD4-Fc (Sino Biological). For each core mutant, 600,000 induced yeast cells were incubated with anti-HA Ab clone 16B12 (Covance) to detect expression and CD4bs mAbs or CD4-Fc in 50 μl PBSB in black 96-well plates (Greiner Bio One) for 1 hour at room temperature while shaking. Yeast cells were washed twice with 200 μl cold PBSB and subsequently incubated with goat anti-mouse and anti-human detection antibodies conjugated to Alexa Fluor 488 or Alexa Fluor 647 (Life Technologies) in 50 μl PBSB for 20 minutes shaking at room temperature and covered from light. Yeast cells were washed once as above, suspended in 200 μl PBSB, and analyzed for fluorescence using a MACSQuant (Miltenyi Biotec) flow cytometer. For each mutant, 10,000 events were gated on HA signal, and median fluorescence intensities (MFIs) were determined and normalized for expression. This method has been previously used to test approximately 2 dozen mAbs (binding within and outside the CD4bs), and the observed results were consistent both with crystal structures whenever they were available and with epitope maps generated by other means, whether or not the cocrystals/other maps were generated with YU2 (38, 39). Additionally, fine specificity differences in the residues identified as involved in binding for multiple antibodies have been found to correlate with either broad or narrow neutralization activity (S.K. Grimm and M.E. Ackerman, unpublished observations), further supporting the validity of the method.
Homology models of CH98 in complex with gp120 were built using Rosetta protein modeling software (63). The sequence of the heavy and light chains of CH98 were threaded independently onto the template’s (CH103, ref. 4; VRC01, ref. 40) heavy and light chain structures (PDB codes 4JAN and 3NGB, respectively). The best-scoring heavy and light chain models from the independent threading runs were then superposed onto the paired heavy and light chain template, and this model was allowed to relax using the Rosetta FastRelax protocol. After relaxation, the model was minimized with AMBER (64). This homology model of the unliganded antibody was then superposed into the solved antibody-gp120 complex structure. The Rosetta loop refinement protocol (65) was used to allow for loop flexibility in the V5 loop of gp120 (a superposition of solved gp120 structures from the PDB indicated the V5 loop to be the most flexible loop in the CD4 binding site) and the HCDR3 loop (the most flexible of the 6 CDR loops; ref. 66). The top-scoring complex model after loop refinement then underwent a last round of relaxation with Rosetta to further refine the complex. A model of the LCDR1 deletion-reverted CH98 complex was built using the above homology modeling procedure using the CH98 mature light chain sequence with the 5-aa GSYNL insertion in LCDR1. One additional step was added to the procedure that placed the LCDR1 in its canonical conformation based on length and sequence using the CDR structural definitions of North and colleagues (41). Predicted binding energy scores for all modeled complexes were computed in Rosetta by subtracting the energy score of the antibody and gp120 alone from the energy score of the antibody-gp120 complex and expressed as REU.
V, D, and J segments of the Ig heavy chain were amplified and sequenced using the 454 pyrosequencing platform as previously described (8). Briefly, a cDNA library was constructed by isolation of mRNA from 22 × 106 PBMCs using the Oligotex kit (Qiagen) and concentrated to 26 μl in elution buffer. Reverse transcription was performed in 2 35-μl reactions (13 μl mRNA, 3 μl oligo(dT)12–18 at 0.5 μg/μl [Invitrogen], 7 μl 5× first-strand buffer [Invitrogen], 3 μl RNase Out [Invitrogen], 3 μl 0.1M DTT, 3 μl dNTP mix, each at 10 mM, and 3 μl SuperScript II [Invitrogen]) and incubated at 42°C for 2 hours. The cDNA was combined, purified, and eluted in 20 μl elution buffer (NucleoSpin Extract II kit; Clontech). Heavy and light chain variable region rearrangements were amplified by PCR from 5 μl cDNA as template (the rough equivalent of transcripts from 5.5 × 106 PBMCs) using the Phusion High Fidelity enzyme (New England Biolabs) according to the manufacturer’s instructions. The 50 μl reaction mix was composed of water, 10 μl 5× buffer, 1 μl dNTP mix, each at 10 mM, 1 μl of each primer at 25 μM, and 1 μl enzyme. Gene-specific forward primers for VH3 gene amplification were VH3 Leader A, C, D, E, and F, and the reverse primers were 3′Cγ-CH1 and 3′Cμ-CH1 (8, 67). The forward primer for VL2 amplification was 5′L-VL1/2-5′-GCACAGGGTCCTGGGCCCAGTCTG-3′, and the reverse primer was 3′CL-5′-CACCAGTGTGGCCTTGTTGGCTTG-3′. The PCR cycling program was run according to Phusion enzyme instructions, using 55°C as the annealing temperature for the heavy chain reaction and 66°C for the lambda. The PCR products at the expected size (450–500 bp) were gel purified (Qiagen), followed by phenol/chloroform extraction. 454 pyrosequencing of the PCR products was performed on a GS FLX sequencing instrument (Roche-454; Life Sciences) using the manufacturer’s suggested methods and reagents.
Values were expressed as mean ± SEM of 2 independent experiments where indicated. Statistical analysis was performed using GraphPad Prism 5 (GraphPad Software).
This study was approved by the Duke IRB and the NIH NIAID DAIDS PSRC committee. Informed consent was received from the participant prior to inclusion in the study according to U.S. federal, local, and Duke University Medical Center regulations.
We thank Krissey Lloyd, Christina Scholarchuk, Julie Blinn, Andrew Foulger, Fangping Cai, Miroslawa Bilska, Morgan Gilman, and Rekha Bhaskarabhatla for excellent technical support; and Kelly A. Soderberg, Megan McCormick, Jennifer Kirchherr, Celia C. LaBranche, and Janet Mueller for project and/or regulatory management and coordination. We thank the NISC Comparative Sequencing Program of the NIH Intramural Sequencing Center, National Human Genome Research Institute, for 454 pyrosequencing. This study was supported by grants from NIH, NIAID, Center for HIV/AIDS Vaccine Immunology (AI067854), and Center for HIV/AIDS Vaccine Immunology–Immunogen Discovery (AI100645).
Conflict of interest: The authors have declared that no conflict of interest exists.
Note regarding evaluation of this manuscript: Manuscripts authored by scientists associated with Duke University, The University of North Carolina at Chapel Hill, Duke-NUS, and the Sanford-Burnham Medical Research Institute are handled not by members of the editorial board but rather by the science editors, who consult with selected external editors and reviewers.
Citation for this article:J Clin Invest. 2014;124(4):1835–1843. doi:10.1172/JCI73441.