Peripheral blood samples from chronically HIV-infected individuals were obtained from the Specialty Care Center at the University of Nebraska Medical Center (UNMC), Omaha, NE, USA. All study subjects were free from other ongoing infections and were chronically infected (>2 years) with stable viral load and CD4⁺ T-cell count after ART. The controls were all healthy people excluded from ongoing infections and chronic diseases, and were matched with the HIV-infected individuals by age, gender, and race. CD4⁺ T-cell counts and frequency of circulating plasmablasts in CD19⁺ B cells in all study subjects were analyzed. A summary of the demographics of all study subjects is shown in Table S1 in Supplementary Material.

All human samples were collected with written informed consent under protocols approved by the UNMC Institutional Review Board (IRB).

Mononuclear cells (PBMC) were purified by Ficoll gradient centrifugation from 20 mL of peripheral blood samples. PBMCs were stained with FITC-conjugated anti-human CD3 Abs, BV421-conjugated anti-human CD19 Abs, BV510-conjugated anti-human CD20 Abs, PE-conjugated anti-human CD27 Abs, and APC-conjugated anti-human CD38 Abs (BD Biosciences, San Jose, CA, USA). After staining, HIV-positive samples were fixed with 4% formaldehyde (Sigma-Aldrich, St. Louis, MO, USA) for 30 min at room temperature (RT) to eliminate live virus. Samples were analyzed at the UNMC FACS core facility. Circulating plasmablasts were identified as CD3⁻CD19^lowCD20^lowCD27^hiCD38^hi cells by FACS Aria (BD Biosciences) and sorted into 96-well PCR plates at one cell/well with 10 µL of lysis buffer (10 mM Tris–HCl pH8.0, 10 U RNasin (Promega, Madison, WI, USA). The sorted single cells were directly subjected to RT-PCR amplification or stored at −80°C for future analysis. Gating and analysis of frequencies of plasmablasts were performed using by BD FACSDIVA™ software (Figure S1 in Supplementary Material).

Single-cell PCR analysis was applied in the case of seven healthy controls and eight HIV-infected individuals (Table S2 in Supplementary Material). The 96-well PCR plate containing sorted cells was incubated at 65°C for 4 h to reverse the crosslinking. RT-PCR was performed using the QIAGEN OneStep RT-PCR kit following the manufacturer’s instructions. Two rounds of PCR reactions were used to amplify IgH, Igκ, or Igλ genes from single cells as previously described (24). To facilitate subcloning, a third round of PCR was performed. Third-round PCR products with unique restriction enzyme digestion sites were subcloned into corresponding Igγ1, Igκ, or Igλ expression vectors. The resulting plasmids were sequenced, and paired IgH and IgL genes from the same well were cotransfected into 293T cells [American Type Culture Collection (ATCC) CRL-3216] to express recombinant mAbs. Culture supernatant was collected 7 days after transfection and Abs were purified using Protein A Agarose beads (ThermoFisher Scientific, Madison, WI, USA).

The Ig heavy (IgH) and light (IgL) variable (V) region sequences were analyzed using the ImMunoGeneTics information system (IMGT)/V-Quest program (http://www.imgt.org/IMGT_vquest/share/textes/) (25) to assign VH, DH, and JH genes. B cells were considered to belong to the same clone on the basis of identical V, D, and J gene segment usage and CDR3 length for both heavy- and light-chain Ig genes. Unique sequences (GenBank accession MG385193–MG385264) were analyzed, whereas identical sequences were eliminated. Potential VH replacement products were identified using VH replacement analyzer-1 (VHRFA1) (26).

The polyreactivity, anti-gp140 reactivity, and DWEYS reactivity of the obtained recombinant mAbs were analyzed by ELISA. ELISA plates were coated with double-stranded DNA (dsDNA) and single-stranded DNA (ssDNA) in PBS, lipopolysaccharide (LPS), insulin (Sigma-Aldrich, St. Louis, MO, USA), DWEYS in coating buffer (pH = 7.6) at 10 µg/mL, and YU2 gp140 and gp140 trimer (HIV-1 clade B) (Immune Tech, Suzhou, Jiangsu, China) in PBS at 1 µg/mL at 4°C overnight. Coated plates were blocked with 100 µL PBST/BSA buffer (150 mM NaCl, 50 mM Tris–HCl, 1 mM EDTA, 2.0% BSA, and 0.05% Tween-20) for 2 h at RT, followed by incubation with 100 µL of recombinant mAbs (1:4 serially diluted in PBS with 2% BSA) at RT for 1 h. After washing four times with PBST, biotin-conjugated goat anti-human IgG Fc Abs (Jackson ImmunoResearch Laboratories Inc., West Grove, PA, USA) were added (100 µL, 1:4,000 dilution) to each well and incubated at RT for 1 h. Bound Abs were detected with horseradish peroxidase (HRP)-conjugated streptavidin (Jackson ImmunoResearch Laboratories Inc., West Grove, PA, USA, 1:8,000 dilution). For color development, 3,3′,5,5′-tetramethylbenzidine (TMB, Kirkegaard & Perry Laboratories, Gaithersburg, MD, USA) was added and incubated for 10 min. The reaction was stopped with 0.5 M H₂SO₄, and the absorbance at 450 nm (OD450) was measured on a POLARstar Omega plate reader (BMG LABTECH, Cary, NC, USA). The PG9, PG16, 2F5, and 4E10 Abs (gift from Dr. Yongjun Guan) were used as standards for gp140 reactivity and polyreactivity testing. For anti-gp140 reactivity, the cut-off value was set as 3 SEM above the average reactivity of control Abs. For reactivity against ds-DNA, ss-DNA, insulin, LPS, and DWEYS, the cut-off value was set as 3 SEM above the average reactivity of control Abs.

Indirect immunofluorescent ANA assay (IIFA) was performed using the HEp-2 ANA kit (MBL International, Woburn, MA, USA) according to the manufacturer’s instructions. Briefly, HEp-2 cell-coated slides were incubated with purified Ab (25 µg/mL) for 30 min at RT. After washing with PBS for 4 times, slides were incubated with FITC-labeled anti-human IgG Abs (Kirkegaard & Perry Laboratories, Gaithersburg, MD, USA) for 30 min at RT. Slides were washed with PBS and visualized under an Olympus 1 × 81 inverted fluorescence microscope (Olympus, Tokyo, Japan). Images were captured with the same exposure setting and analyzed using the Slidebook5 program (Intelligent-Imaging). Specific staining patterns, such as nuclear and cytoplasmic, were documented.

Neutralization of a selected panel of representative virus isolates by mAbs was measured using pseudovirus infection of TZM-bl cells as previously described (27).

The following reagents were obtained from the NIH AIDS Reagent Program: the TZM-bl cell line; the plasmid pSG3^ΔEnv to produce HIV pseudoviruses; and plasmids containing HIV envelopes of SF162, BaL.26, and WITO4160.33. Briefly, 5 µg of plasmids carrying the functional Env clone from selected virus isolates and 10 µg of the backbone pSG3^ΔEnv recombinant plasmids were transfected into HEK 293T/17 cells (5 × 10⁶ cells/mL). Virus supernatants were harvested after 48 h of incubation at 37°C with 5% CO₂. p24 protein in each virus sample was quantified by using the AlphaLISA HIV p24 Biotin-Free detection kit (Perkin Elmer, Waltham, MA, USA), and input virus was normalized to 5–10 ng/mL for the following TZM-bl cell assay. Pseudotyped viruses containing the Env of amphotropic murine leukemia virus (aMLV) were used as the negative control. Then, 10 µL of mAbs fivefold serially diluted from a starting concentration of 50 µg/mL were incubated with 40 µL of replication-competent virus samples in duplicate for 30 min at 37°C in 96-well clear flat-bottom black culture plates (Greiner Bio-One, Monroe, NC, USA). TZM-bl cells were added at a concentration of 10,000 cells per 20 µL to each well in DMEM containing 75 µg/mL DEAE-dextran and 1 µM indinavir. Cell-only and virus-only controls were included on each plate. Plates were incubated for 24 h at 37°C in a 5% CO₂ incubator, after which the volume of culture medium was adjusted to 200 µL by adding complete DMEM containing indinavir. 48 h post-infection, 100 µL was removed from each well and 100 µL of SpectraMax Glo Steady-Luc reporter assay (Molecular Devices, LLC., CA, USA) reagent was added to the cells. To determine luciferase activity, cells were lysed using M-PER Mammalian Protein Extraction Reagent (Pierce, Rockford, IL, USA) and approximately 20 µL of each lysed sample was transferred to a microplate. Reporter Lysis Buffer (Promega, Madison, WI, USA) was added to each sample in the microplate and the light intensity was measured using a SpectraMax i3x multi-mode detection platform following the manufacturers’ instructions. Non-infected cells in wells were used to determine background luciferase activity, which was subtracted from the activity measured for all other samples. Relative light units (RLU) per milliliter were calculated by dividing the luciferase values by their corresponding dilutions. The amount of infectious virus was determined as RLU per nanogram of p24. The 50% inhibitory concentrations IC50 was defined as the Ab dilution that caused a 50% reduction in neutralization. IC50s of known bnAbs P9, PG16, 2F5, and 4E10 were also determined against the cross-clade panel of pseudoviruses (Table S3 in Supplementary Material).

SH-SY5Y cells (ATCC 2266) were cultured in DMEM medium with 10% heat-inactivated fetal bovine serum (FBS), 2 mM l-glutamine, 100 U/mL penicillin, 100 µg/mL streptomycin, and 25 mM HEPES. For immunofluorescence microscopy analyses, cells were subcultured in glass coverslips pre-coated with l-lysine. HIV201B2 and U78Ab3 were labeled with Dylight 488 (ThermoFisher Scientific, Bremen, Germany) according to the manufacturer’s instructions. For flow cytometry analyses, cells were fixed with 4% paraformaldehyde and stained with Dylight 488-labeled mAb (10 µg/mL) at RT, and recorded using an Olympus 1 × 81 inverted fluorescence microscope (Olympus, Tokyo, Japan). For SH-SY5Y cell competition-binding assessment, cells were coincubated with Dylight 488-labeled HIV201B2 and DWEYS with and increasing molar ratio, and recorded using a BD Accuri C6 system (BD Biosciences, San Jose, CA, USA).

SH-SY5Y cells cultured in DMEM medium were treated with HIV201B2 and U78Ab3 with increasing concentrations (2, 10, 50, and 250 µg/mL) for 24 h. Cell viability was measured using the MTT formazan (Sigma Aldrich, St. Louis, MO, USA) assays following the manufacturer’s instructions and reading the absorbance at 570 nm using a standard spectrophotometer.

Statistical analyses were performed using GraphPad Prism 5.0 (GraphPad Software, Inc., San Diego, CA, USA). For Ig gene repertoire, Ab reactivity, and amino acid (aa) usage, statistical analyses were performed using Fisher’s exact test. For the CD4⁺ T-cell count, the percentage of plasmablasts in CD19⁺ B cells, and the frequency of VH replacement, the statistical analyses were performed using the Mann–Whitney test. For IgH CDR3 length and VH mutation rate, the statistical analyses were performed using Student’s t test on condition that the F-test for equal variance was executed first and two-sided P > 0.05. The correlations of CD4⁺ T-cell count, viral load and plasmablast frequency, anti-dsDNA reactivity, and reactivity against other antigens were determined by Spearman’s ranked correlation test. Two-sided P ≤ 0.05 was considered statistically significant, and P < 0.0001 was considered extremely statistically significant.

We examined the CD4⁺ T-cell count and percentage of plasmablast in each study subject by flow cytometry. The CD4⁺ T-cell counts were significantly reduced in chronically HIV-infected individuals compared with those in healthy controls (Figure S2A in Supplementary Material; P < 0.0001 by Mann–Whitney test). The percentage of plasmablasts among CD19⁺ B cells (mean ± SD, 2.20 ± 0.18%) was significantly higher than that in healthy controls (mean ± SD, 0.47 ± 0.09%) (Figure S2B in Supplementary Material; P < 0.0001 by Mann–Whitney test). These results indicate ongoing immune responses to HIV infection. Plasmablast frequency did not correlate with CD4⁺ T-cell count (Figure S2C in Supplementary Material; r = 0.2964, P = 0.2834 by Spearman’s ranked correlation test) or viral load (Figure S2D in Supplementary Material; r = −0.1574, P = 0.5753 by Spearman’s ranked correlation test).

We expressed 64 mAbs and 72 mAbs from the circulating plasmablasts of four healthy controls and four chronically HIV-infected individuals, respectively (Table S2 in Supplementary Material). To determine the antigenic specificity of each mAb, we performed ELISAs against both HIV gp140 monomer and trimer (HIV-1 clade B, YU2). Well-known bnAbs 2F5, 4E10, PG9, and PG16 (28–30) were analyzed as standards to optimize the assay. PG9 and PG16 are the type of Ab that do not bind to monomeric gp120 but are capable of neutralizing a large proportion of viruses at low titers (30). Consistent with previous findings (30, 31), 2F5 and 4E10 were reactive with both gp140 monomer and trimer; PG9 was weakly reactive with gp140 trimer, whereas PG16 was not reactive with gp140 in either form (Figure 1A). Therefore, PG16 and 4E10 were used as the negative and positive control, respectively, in gp140-ELISA. The four HIV-infected individuals exhibited varying frequencies of Abs reactive against gp140, with a total of 27.8% of the anti-gp140 Abs (Figure 1B). This frequency was significantly higher than that in the control group (Figure 1B). However, the gp140-affinities of these Abs were extremely low. We included two of the most sensitive tier-1 clade B virus isolates (SF162 and BaL.26), and one moderately sensitive tier-2 clade B virus isolate (WITO4160.33) (32–34) (Table S3 in Supplementary Material). We tested the mAbs for neutralization via generation of pseudoviruses followed by TZM-bl cell infection assay. While none of them was capable of neutralize SF162, the gp140-binding and non-gp140-binding mAbs displayed no significant difference in neutralization potency (Figure 1C; P = 0.0866 by unpaired t-test). No Ab was identified to be able to neutralize other viruses tested (data not shown). Therefore, none of these mAbs seemed to be the PG9 and PG16 type, which bind to trimeric gp120 but not monomeric gp120, and have broad and potent neutralization activity. Taken together, these results suggest that mAbs cloned from randomly sorted plasmablasts of chronic HIV-infected individuals with ART experience were weak gp140-binding and non-neutralizing Abs.

Previous analyses of mAbs derived from HIV gp140-specific memory B cells showed that the majority (75%) were poly/autoreactive (20). To determine polyreactivity, we analyzed the plasmablast-derived mAbs from our study by ELISA using the same panel of antigens, i.e., dsDNA, ssDNA, insulin, and LPS. Abs reacting against at least two of these antigens were considered to be polyreactive (20). Previously identified bnAbs 2F5, 4E10, PG9, and PG16 (28–30) were analyzed as standards to optimize the assay. The bnAbs 2F5 and 4E10 were polyreactive, but PG9 and PG16 were not (Figure 2A). Notably, all mAbs in the present study were expressed from the accurately gated single plasmablasts of HIV-positive individuals without any preselection of antigenic specificity. Although the frequencies of polyreactive Abs varied among the four HIV-infected individuals (from 33 to 88%), they were all significantly elevated compared with those in controls (Figure 2B). We found that 56.9% of mAbs (41/72) from the HIV group were polyreactive (Figure 2B). Further analyses showed that the anti-dsDNA reactivity of the Abs derived from the HIV-infected individuals were strongly correlated with their reactivities against ssDNA (Figure 2C), insulin (Figure 2D), and LPS (Figure 2E) but were only weakly correlated with their reactivity against HIV gp140 (Figure 2F). When we focused on the gp140-binding Abs, their anti-dsDNA reactivity showed a moderate correlation with their gp140 reactivity (Figure 2G). Moreover, there was no significant difference in the percentages of polyreactive Abs between anti-gp140 and non-anti-gp140 mAbs (Figure 2H). Taken together, these results indicate that circulating plasmablasts in chronically HIV-infected individuals with ART experience predominantly produce polyreactive Abs regardless of their gp140 reactivity; among these polyreactive Abs, some of them bind weakly to gp140.

To identify molecular differences that might contribute to polyreactivity, we compared the 72 IgH genes expressed as mAbs from chronic HIV-infected individuals (Table S4 in Supplementary Material). The distributions of individual VH, DH, and JH genes of polyreactive and non-polyreactive Abs were examined. Unexpectedly, there were no preferential VH, DH, or JH gene segments (Figures 3A–C). The HCDR3 region is the canonical antigen-binding site with the most variability. Previous studies have shown that long and charged HCDR3 loops are associated with polyreactivity (35). We examined the variation in the HCDR3 length and aa content. For polyreactive Ab genes, the average HCDR3 is composed of ~15 aa and is slightly longer than the HCDR3 of genes encoding non-polyreactive Abs (Figure 3D; P = 0.2768 by unpaired t-test). To examine the contribution of HCDR3 length in more detail, we compared the OD values reflecting the strength of reactivity against dsDNA in groups of Abs derived from chronic HIV-infected individuals according to their HCDR3 length. Notably, Abs with longer HCDR3 were more reactive to dsDNA. A declining trend of OD values was seen as the HCDR3 length decreased (Figure 3E). The relative proportional distribution of aa was similar in the Ab repertoire of both groups (Figure 3F). To gain further insight into the mechanisms that might contribute to polyreactivity, we also analyzed the VH mutation rate. The non-polyreactive Abs and polyreactive Abs showed a similar rate of mutation in VH regions (Figure 3G). Overall, HCDR3 length is a prominent contributor to the polyreactivity characteristic of mAbs from chronic HIV-infected individuals.

Next, we performed HEp-2 cell-based IIFA to determine whether the recombinant mAbs derived from the plasmablasts of HIV-infected individuals were autoantibodies (ANAs). Among the 72 recombinant mAbs derived from the HIV-infected individuals, 40 (55.6%) reacted with fixed HEp-2 cells. This frequency was significantly higher than that in the control donors (Figure 4B). Notably, among these 40 ANAs, 95% (38/40) showed a homogeneous nuclear staining pattern, whereas only 5% (2/40) exhibited cytoplasmic antigen staining (Figures 4A,C), which reflected their dsDNA-binding abilities. Moreover, 32.5% (13/40) showed positive gp140-binding results (Figure 4D). Additionally, these 40 ANAs exhibited significantly stronger dsDNA-binding affinities than the 32 non-autoreactive ANAs (Figure 4E; P = 0.0038 by unpaired t-test). Thus, the mAbs derived from the circulating plasmablasts of chronically HIV-infected individuals prominently exhibit poly/autoreactive and non-neutralizing characteristics.

The cross-reactivity of anti-dsDNA ANAs with the DWEYS peptide sequence present in NMDAR subunits is a recognized molecular mechanism of neuropsychiatric lupus (NPSLE) (36–38). To determine whether the poly/autoreactive mAbs share this cross-reactivity with NMDAR, we measured their DWEYS reactivity by ELISA. We found significantly higher frequencies of DWEYS-reactive mAbs in the four chronically HIV-infected individuals than in the controls (Figure 5A). We used HIV201B2 as a model NMDAR-reactive ANA and U78Ab3 as an isotype-matched control Ab. HIV201B2 showed high affinity to both dsDNA and DWEYS (Figure 5B) and was autoreactive with a homogeneous nuclear staining pattern (Figure 4A). We first determined by immunofluorescence staining that HIV201B2 bound robustly to SH-SY5Y cells, with staining on the outer membrane where NMDARs are supposedly expressed, which was not observed with U78Ab3 (Figure 5C, top left). We simultaneously recorded staining via flow cytometry; of note, as increasing amounts of DWEYS were applied, HIV201B2 showed a gradual loss of binding to SH-SY5Y cells (Figure 5C, bottom). This result provides solid evidence that HIV201B2 bound to SH-SY5Y cells via DWEYS. We continued to explore how this binding would affect the cells. Remarkably, HIV201B2 induced evident cell death as its concentration in the culture medium increased, whereas U78Ab3-treated cells exhibited a more normal cell viability after 24 h (Figure 5D). When the DWEYS peptide with increasing concentrations was applied together with the cross-reactive Ab, the cell toxicity was gradually blocked (Figure 5E). Several other mAbs were also tested, and similar effects were seen (Figure S3 in Supplementary Material). We therefore suspect that non-neutralizing anti-HIV mAbs with strong polyreactivity and autoreactivity, such as HIV201B2, may be pathological due to their cross-reactivity with autoantigens.

We obtained 248 unique IgH gene sequences from eight chronically HIV-infected individuals and 228 unique IgH gene sequences from seven control donors (Table S2 in Supplementary Material). Analyses of the potential VH, DH, and JH germline gene usages among IgH genes derived from both groups showed diversified repertoires (Figures 6A–C). The VH3-23 gene, the most often utilized VH gene in naïve and memory B-cell repertoires (39), was also the most frequently used in the plasmablast repertoires of both groups (Figure 6A). Notably, the frequency of VH4-34 utilization in IgH genes derived from HIV-infected individuals was significantly elevated (Figure 6A). VH1-2 gene utilization was marginally more frequent in the HIV group (Figure 6A). The heavy-chain CDR3 (HCDR3) is the antigen-binding region with the highest variability and often plays a decisive role in antigen binding. We did not find a significant difference in HCDR3 region length between the two groups (Figure 6D). Similarly, there was no significant difference in the VH mutation rate between repertoires of control and HIV-infected individuals (Figure 6E).

Previous evidence has demonstrated that VH replacement contributes to the neutralization of B-cell autoreactivity and diversification of the Ab repertoire (40–42). VH replacement utilizes evolutionarily conserved cryptic recombination signal sequences (cRSSs) to generate novel VH rearrangements, with a short stretch of nucleotides retained as a VH “replacement footprint” at the VH–DH junction (N1) (43). We used the VHRFA program (26) to identify VH replacement footprints in the N1 regions (Table S5). In IgH genes derived from the plasmablasts of chronically HIV-infected individuals, VH replacement products were significantly enriched compared with the frequency in the control group (Figure 7A). These identified VH replacement products exhibited significantly longer CDR3 regions than the non-VH replacement products (Figure 7B) and preferentially contributed charged aa, such as Glu (D) and Asp (E), to the CDR3 regions (Figure 7C).

Among the 35 VH replacement products derived from HIV-infected individuals, seven were expressed as recombinant mAbs. Interestingly, 57.1% (4/7) of the Abs encoded by VH replacement products were polyreactive (Figure 7D) and 85.7% (6/7) were autoreactive (Figure 7E); 28.6% (2/7) reacted with gp140 (Figure 7F). These results indicate that the enriched VH replacement products in the plasmablast repertoires of chronically HIV-infected individuals may contribute to the generation of poly/autoreactive Abs.