IGHV1-69 B Cell Chronic Lymphocytic Leukemia Antibodies Cross-React with HIV-1 and Hepatitis C Virus Antigens as Well as Intestinal Commensal Bacteria

B-cell chronic lymphocytic leukemia (B-CLL) patients expressing unmutated immunoglobulin heavy variable regions (IGHVs) use the IGHV1-69 B cell receptor (BCR) in 25% of cases. Since HIV-1 envelope gp41 antibodies also frequently use IGHV1-69 gene segments, we hypothesized that IGHV1-69 B-CLL precursors may contribute to the gp41 B cell response during HIV-1 infection. To test this hypothesis, we rescued 5 IGHV1-69 unmutated antibodies as heterohybridoma IgM paraproteins and as recombinant IgG1 antibodies from B-CLL patients, determined their antigenic specificities and analyzed BCR sequences. IGHV1-69 B-CLL antibodies were enriched for reactivity with HIV-1 envelope gp41, influenza, hepatitis C virus E2 protein and intestinal commensal bacteria. These IGHV1-69 B-CLL antibodies preferentially used IGHD3 and IGHJ6 gene segments and had long heavy chain complementary determining region 3s (HCDR3s) (≥21 aa). IGHV1-69 B-CLL BCRs exhibited a phenylalanine at position 54 (F54) of the HCDR2 as do rare HIV-1 gp41 and influenza hemagglutinin stem neutralizing antibodies, while IGHV1-69 gp41 antibodies induced by HIV-1 infection predominantly used leucine (L54) allelic variants. These results demonstrate that the B-CLL cell population is an expansion of members of the innate polyreactive B cell repertoire with reactivity to a number of infectious agent antigens including intestinal commensal bacteria. The B-CLL IGHV1-69 B cell usage of F54 allelic variants strongly suggests that IGHV1-69 B-CLL gp41 antibodies derive from a restricted B cell pool that also produces rare HIV-1 gp41 and influenza hemagglutinin stem antibodies.


Introduction
The initial B cell responses to HIV-1 envelope (Env) gp41 are non-neutralizing [1] and are polyreactive with human intestinal commensal bacterial antigens [2]. Env gp41 antibodies that arise following HIV-1 transmission do not select virus escape mutants and therefore exert no anti-viral immune pressure [1]. We have recently demonstrated that gp41-reactive B cells can be isolated prior to infection in HIV-1-uninfected humans and that HIV-1 activates preexisting B cells that are cross-reactive with gp41 and non-HIV-1 antigens including microbial antigens [2]. However, the pool of B cells from which the initial HIV-1 Env B cell response is derived is not known.
B chronic lymphocytic leukemia (B-CLL) is a clonal expansion of CD5 + B lymphocytes frequently associated with unmutated B cell receptors (BCRs) [3]. B-CLL cells with unmutated immunoglobulin heavy variable regions (IGHVs) (unmutated CLL, U-CLL) show a preferential usage of IGHV1-69 gene segment (,25%) and frequently have BCRs that are polyreactive and autoreactive despite dramatic structural restrictions [4][5][6][7][8][9][10]. The cellular origin of B-CLL cells has been an area of considerable debate. For example, it has been proposed that B-CLL cells derive from human B-1-like cells, marginal zone (MZ) innate B cells, or transitional B cells, based on cell surface phenotype and molecular and functional characteristics [11]. In this regard, recent studies identified a human equivalent of murine B-1 cells (CD20 + , CD27 + , CD43 + , CD70 2 ) [12] and circulating CD5 + human B cells [13] as the precursors of CLL B cells. It has also been proposed that B-CLL cells with BCR stereotypy could derive from B-1-like progenitor cells adapted to particular antigenic challenges while B-CLL cells with heterogeneous BCRs could derive from conventional B cells [14]. In addition, anti-viral innate antibodies have been reported to be derived from B-1/MZ B cells [15][16][17].

Cell culture
Epstein-Barr virus (EBV)-stimulation of patient peripheral blood mononuclear cells (PBMCs) and generation of B-CLL heterohybridoma cell lines have been described previously [28]. We stimulated PBMCs from 58 B-CLL patients (33 IGHV1-69 and 25 IGHV2/IGHV3) with EBV in the presence of a Toll-like receptor 9 agonist ODN2006 (12.5 mg/ml; Invivogen) and cyclosporin A (0.5 mg/ml), and cultured the cells in the presence of feeder cells, J774A.1 (50,000 cells per well; American Type Culture Collection, TIB-67) that had been exposed to c-irradiation (40 Gy) from a Shepherd irradiator. Three weeks after stimulation, culture supernatant was collected from each well, and levels of total IgM were measured using a previously described method [28]. We obtained 39 patient cultures (22 IGHV1-69 and 17 IGHV2/IGHV3) that produced similar levels of IgM. Of the 22 IGHV1-69 samples, 21 were U-CLL and 1 mutated CLL (M-CLL) while of the 17 IGHV2/IGHV3 samples, 9 were U-CLL and 8 M-CLL ( Table  S1). As negative controls, EBV-stimulated B cell cultures from PBMCs of 20 normal subjects were studied.
Briefly, ELISA plates (Costar, Cambridge, MA) were coated with 1-5 mg/ml of test antigens in 0.1 M sodium bicarbonate buffer. After incubating overnight at 4uC, plates were blocked with PBS containing 15% goat serum, 4% whey protein, 0.5% Tween-20, and 0.05% NaN 3 . Then test supernatants or mAbs diluted in the blocking buffer were distributed to wells and incubated for 2 hours at room temperature. After washing with PBS-0.5% Tween-20, bound human IgM or IgG was detected with horseradish peroxidase-conjugated goat anti-human IgM or IgG (m-chain or c-chain specific; Jackson ImmunoResearch Laboratories, West Grove, PA) and peroxidase substrate tetramethylbenzidine (Kirkegaard and Perry Laboratories, Gaithersburg, MD) using a SpectraMax Plus384 plate reader (Molecular Devices, Sunnyvale, CA). The detection limit of IgM in each well was 60 ng/ml; negative wells with undetectable levels of IgM were assigned 10 ng/ml to permit logarithmic transformation of the data.
Reactivity of mAbs to aerobic and anaerobic bacteria whole cell lysates was tested by binding antibody multiplex (Luminex) assays as previously described [1,2]. Bacterial whole cell lysates were prepared using previously described methods [2,35]. In addition, surface plasmon resonance analysis of mAb reactivity to MN gp41 and HCV E2 proteins was performed on a BIAcore 3000 (BIAcore Inc.) using the methods as previously described [2].

Expression of recombinant IgG 1 mAbs
Live EBV-stimulated B cells from selected wells were sorted as single cells using a BD FACS Aria (BD Biosciences, San Jose, CA), and the isolated VH and VL gene pairs were assembled by PCR into the linear full-length immunoglobulin heavy-and light-chain gene expression cassettes for production of recombinant IgG 1 mAbs by transfection in the human embryonic kidney cell line, 293F (American Type Culture Collection) using the methods as previously described [29].
We next expressed the 5 B-CLL mAb V H DJ H and V L J L genes as full-length IgG 1 recombinant mAbs [29]. All 5 B-CLL recombinant IgGs bound to MN gp41 ( Figure 2B). Of these, CLL698 and CLL821 IgGs bound to the immunodominant region of HIV-1 clade B BAL gp41 (RVLAVERYLRDQQLL-GIWGCSGKLICTTAVPWNASWSNKSLNKI) ( Figure 2B). However, CLL246 and CLL698 IgGs did not bind to any other linear peptides tested including DP107, MPR.03, MEPR656, and overlapping 15-mer MN gp41 linear peptides (data not shown). These results indicated that multivalent IgM antibodies with high avidity interactions could enhance low affinity interactions between the unmutated IgG antibodies and the linear peptides tested.
Gp41 antibodies that arise in HIV-1 infection frequently crossreact with intestinal commensal bacterial antigens and indeed have been postulated to derive from pre-transmission environmental antigen-reactive antibodies from memory B cells [2]. Therefore, we tested reactivity of B-CLL mAbs with aerobic and anaerobic intestinal commensal bacterial whole-cell lysates using binding antibody multiplex assays [2]. We found all 5 IGHV1-69 unmutated IgMs reacted with aerobic and/or anaerobic intestinal commensal bacterial whole-cell lysates (Figure 1). The recombinant IgGs of CLL526 and CLL1324 also reacted with aerobic and/or anaerobic intestinal commensal bacterial whole-cell lysates ( Figure 2B). Similarly, all 5 IGHV1-69 unmutated IgMs and their recombinant IgGs also reacted with HCV E2 protein (Figure 1 and Figure 2B). Two mAbs were chosen for cross-competition studies with HCV E2; recombinant E2 competitively inhibited the binding of CLL821 and CLL1324 IgGs to gp41 (Figure 3).
It has been proposed that B-CLL cells derive from autoreactive B cell precursors [6,37]. In this regard, 2 of 5 recombinant IgG mAbs (CLL698 and CLL1324) bound to double-stranded DNA but not to the other test autoantigens including SSA, SSB, Sm, RNP, Scl-70, Jo-1, centromere B, and histone (data not shown). In our indirect immunofluorescence staining assay, however, none of the IgM paraproteins or the recombinant IgG mAbs reacted with HEp-2 epithelial cells, and none showed rheumatoid factor activity (data not shown). In functional assays, none of the IgM or IgG B-CLL mAbs neutralized HIV-1 strains, SF162 (clade B), BG1168 (clade B), or MN (clade B) ( Table S3) [2]. Similarly, none of the IgM mAbs inhibited syncytium formation by HIV-1 ADA (clade B) and MN nor captured HIV-1 virions, SF162 or BG1168 (Table S4 and Table S5). Moreover, none of the IgMs neutralized a HCV subtype 1a strain, HCVpp-H77 (Table S3) [38].
The HCDR3 sequences are the principal determinants of antibody-binding specificity in most antibodies [41]. Thus, we compared HCDR3 sequences of the 5 gp41-reactive IGHV1-69 B-CLLs with those of 47 gp41-reactive IGHV1-69 antibodies isolated from HIV-1-infected patients. The analysis revealed similar HCDR3 sequences due to common usages of IGHJ6 and IGHD3 gene segments that were preferentially used by gp41-reactive B-CLL mAbs. For example, the long HCDR3 sequences of mAbs Ab2757 (25 aa) and Ab6064 (23 aa) were remarkably similar (60% and 52% aa identity, respectively) to that of CLL1324 ( Figure 4). However, IGHJ4 was the most frequently used gene segment (32%) in the HIV-1 infection-derived IGHV1-69 gp41 antibodies in contrast to the infrequent use of IGHJ4 by IGHV1-69 B-CLL (,4%) [18]. In addition, IGHD3-3, the most frequent D gene segment used by the gp41-reactive B-CLL mAbs was found in only 4% (2/47) of the HIV-1 infection-derived IGHV1-69 gp41 antibodies. The mean HCDR3 length of the HIV-1 infectionderived IGHV1-69 gp41 antibodies was significantly shorter than that of the gp41-reactive IGHV1-69 B-CLL antibodies (16.1 aa vs. 22 aa; Mann-Whitney test, p = 0.0041). Moreover, the sequence pattern cluster analysis of HCDR3s indicated that none of the HIV-1 infection-derived IGHV1-69 gp41 antibodies belonged to the known major B-CLL stereotype subsets [14]. These results indicate that the gp41-reactive IGHV1-69 CLL B cells have molecular features distinct from those found in most IGHV1-69 gp41 B cells during HIV-1 infection.

Virus binding activity of B-CLL and clinical outcomes
When we divided the B-CLL samples based on their binding activity to the test viral antigen preparations ( Figure S1), we found that virus antigen-binding reactivity of B-CLL cultures correlated with B-CLL clinical course. The Kaplan-Meier plots of the analyses revealed that B-CLL cases with anti-viral reactivity correlated with poor clinical outcomes measured as time to first treatment (TFT) and overall survival of the patients ( Figure 5). The median TFTs for virus-binding and non-virus binding groups were 37 mo and 86 mo, respectively (p = 0.011, Mantel-Cox test), and the median overall survival for virus binding and non-virus binding groups were 131 mo and 177 mo, respectively (p,0.0001, Mantel-Cox test). This was especially impressive when restricting the analysis to IGHV1-69 samples ( Figure 5B and Figure 5D). The median overall survival for virus binding and non-virus binding groups were 117 mo and indefinite, respectively (p = 0.012, Mantel-Cox test). Of note, all but one (CLL1011) IGHV1-69 samples were U-CLL and would therefore be expected to have poor clinical outcome [3]. However, the U-CLL IGHV1-69 samples could be segregated by virus binding activity, with the non-binders to viral antigens having good clinical outcome. These findings suggested that certain BCRs with innate anti-viral reactivity may be important factors in determining the outcome of the B-CLL clinical course.

Discussion
In this paper, we have demonstrated that one third of IGHV1-69 B-CLL BCRs are polyreactive for infectious agent or commensal bacterial antigens ( Figure S1 and Figure 1). B-CLL IgM reactivity with infectious agent antigens was significantly correlated with poor clinical outcomes ( Figure 5). Moreover, there was a striking difference in IGHV1-69 allelic use by B-CLL versus HIV-1 IGHV1-69 antibodies. While IGHV1-69 B-CLL BCRs predominantly used F 54 allelic variants, IGHV1-69 HIV-1 Env gp41 antibodies from HIV-1 infected patients predominantly used L 54 ( Table 1).
Liao et al. [2] have demonstrated that the initial blood plasma cell response in acute HIV-1 infection to gp41 is highly mutated and comprised of polyreactive gp41 antibodies that cross-react with intestinal commensal bacterial antigens. This work led to the hypothesis that the initial gp41 response to HIV-1 may be in part derived from commensal bacteria-activated memory B cells with BCRs that cross-react with Env gp41 and not from naïve B cells [2]. Thus, HIV-1 Env in the context of HIV-1 infection induces a dominant Env gp41 antibody response that is polyreactive with host and intestinal commensal bacterial antigens [2]. The observation that IGHV1-69 B-CLL BCRs are similarly polyreactive and cross-react with intestinal commensal bacteria (Figure 1) raises the hypothesis that the B-CLL cell population is an expansion of members of the innate polyreactive B cell repertoire with reactivity to a number of infectious agent antigens including intestinal commensal bacteria. Hence, our results suggested that the initial response to gp41 in HIV-1 may derive from the same pool of B cells as B-CLL. However, it is striking that B-CLL B cells predominantly utilize F 54 IGHV1-69 allelic variants while HIV-1 Env gp41 B cell BCRs from HIV-1 infection utilize L 54 allelic variants ( Table 1). Therefore, the B-CLL IGHV1-69 B cell usage of F 54 allelic variants demonstrate that the initial response to gp41 in HIV-1 may not derive from the same pool of B cells as B-CLL. In fact, the B-CLL IGHV1-69 B cells may drive from an F 54 allelic variant B cell pool that produces rare gp41 and hemagglutinin stem antibodies. It has been demonstrated that the F 54 IGHV1-69 allelic variant B cells arise during early human fetal liver development [42]. They were found in a high proportion of B cells in the primary follicles of fetal spleen [43] and in the mantle zones of adult tonsil [44]. Thus, B-CLL B cells may derive from this mantle zone pool of polyreactive B cell precursors [18,45,46].
The 5 gp41-reactive unmutated B-CLL mAb clones had similar HCDR3 sequences due to common IGHV-D-J rearrangements, and as well, had long HCDR3s (21-23 aa) (Figure 2A). Three clones (CLL246, CLL526, and CLL698) belong to subset 7  according to the major stereotyped BCR subset numbering based on a sequence pattern cluster analysis of B-CLL HCDR3s ( Figure 2A) [14]. Unmutated B-CLL B cells with stereotypy give rise to the hypothesis that they are derived from a subset of B cells selected for ability to bind to bacterial and viral antigens, characteristics of B-1, transitional and MZ B cells [11]. It has been proposed that a small population of CD20 + CD27 + CD43 + CD70cells present in human umbilical cord and adult peripheral blood represent a B cell subset analogous to the murine B-1 subset [12], and human transitional and MZ B cells share traits that are similar to murine B-1 B cells, and collectively produce pre-formed antibodies to pathogens [47]. For both HIV-1 and HCV, we found no neutralizing antibodies among any of the B-CLL gp41 or HCV E2-reactive antibodies. Similarly, acute HIV-1 infection gp41 antibodies are nonneutralizing [1,2]. In contrast, the influenza-reactive non-mutated IGHV1-69 antibodies F10 and CR6260 neutralized a broad spectrum of influenza strains [21,24]. If IgM antibodies can coat infectious agent virions, they may impede virus migration across mucosal surfaces [48,49]. However, virus capture assays showed that none of gp41-reactive B-CLL mAbs captured test HIV-1 virions. Moreover, acute HIV-1 infection gp41 antibodies do not exert immune pressure via selecting escape mutants [1].
Finally, several studies have shown that unmutated B-CLL B cells, similar to natural or innate IgM antibodies, frequently express polyreactive antibodies that bind to autoantigens associated with apoptosis and oxidation as well as to components of the outer membrane of bacteria [37,50]. Of note, it has been demonstrated that human B-1-like cells (CD20 + CD27 + CD43 + CD70 2 ) displayed a skewed BCR repertoire as indicated by preferential expression of anti-phosphorylcholine and anti-DNA specificities [12]. Our findings that unmutated B-CLL cell gp41 reactivity is selective for the F 54 IGHV1-69 gene segment and has characteristics of B-1-like, transitional and MZ B cell derived antibodies strongly suggest that B-CLL IGHV1-69 gp41 antibodies derive from a restricted B cell pool that also produces rare HIV-1 gp41 and influenza hemagglutinin stem antibodies. Figure S1 Binding characteristics of B-CLL B cell cultures. To compare binding activities of B-CLL IgMs expressing IGHV1-69 vs. IGHV2/IGHV3 gene families, we stimulated PBMCs from B-CLL patients with EBV using the methods as previously described [28], and the cells were plated at 5,000 cells per well in total of 20 wells per patient sample. To profile binding characteristics of IgMs, we screened the culture supernatants in ELISA. HIV-1 antigens included aldrithol-2 (AT-2)-inactivated HIV-1 virions ADA (Clade B); HIV-1 group M consensus Env, ConS gp140; and deglycosylated JRFL gp140. HIV-1 Env gp41 linear epitope peptides included HR-1 region peptide, DP107 (NNLLRAIEAQQHLLQLTVWGIKQLQARI-LAVERYLKDQ); Env clade B HR-2 region peptide, MPER656 (NEQELLELDKWASLWNWFNITNWLW); and Env clade C HR-2 region peptide, MPR.03 (KKKNEQELLELDK-WASLWNWFDITNWLWYIRKKK). As an initial approach to ensure reactivity of IgMs were of B-CLL origin, rather than IgMs from contaminating B cells, we defined positive samples as they produced 10 or more wells ($50%) reactive with each test antigen. Of 440 IGHV1-69 B-CLL cultures from 22 patients, 67 wells reacted with DP107, 20 reacted with the MPER656, and 37 reacted with MPR.03. The reactivities of 340 IGHV2/IGHV3 B-CLL cultures (17 patients) for these epitopes were 3, 2, and 1 well, respectively (p,0.0001, p = 0.0007, and p,0.0001; Fisher's exact test vs. the IGHV1-69 group). Data are expressed in number of wells positive for each test antigen. NA, not applicable. ''-'' denotes no binding. 1 IGHV and IGKV/IGLV mutation frequencies (%) were compared with germline according to IMGT. 2 Two B-CLL mAbs were isolated from separate experiments (Hwang et al., 2012), and the results for binding activity were obtained from the purified IgM paraproteins. 3 HCDR3 subset numbers were assigned using previously described methods [14]. (TIF) Figure S2 Binding characteristics of healthy control B cell cultures. We stimulated PBMCs from 20 healthy control subjects with EBV using the methods as previously described [28], and the cells were plated at 5,000 cells per well in total of 20 wells per sample. To profile binding characteristics of IgMs, we screened the culture supernatants in ELISA. HIV-1 antigens included aldrithol-2 (AT-2)-inactivated HIV-1 virions ADA (Clade B); HIV-1 group M consensus Env, ConS gp140; and deglycosylated JRFL gp140. HIV-1 Env gp41 linear epitope peptides included HR-1 region peptide, DP107 (NNLLRAIEAQQHLLQLTVWGIKQLQARILAVER-YLKDQ); Env clade B HR-2 region peptide, MPER656 (NEQELLELDKWASLWNWFNITNWLW); and Env clade C HR-2 region peptide, MPR.03 (KKKNEQELLELDK-WASLWNWFDITNWLWYIRKKK). The reactivities of 400 cultures from 20 non-CLL control subjects for DP107, MPER656, and MPR.03 were 2, 10, and 4 wells, respectively (p,0.0001, p = 0.14, and p,0.0001; Fisher's exact test vs. the IGHV1-69 group). Data are expressed in number of wells positive for each test antigen. NA, not applicable. (TIF)