Cardiolipin Polyspecific Autoreactivity in Two Broadly Neutralizing HIV-1 Antibodies

See allHide authors and affiliations

Science  24 Jun 2005:
Vol. 308, Issue 5730, pp. 1906-1908
DOI: 10.1126/science.1111781


The design of a human immunodeficiency virus–1 (HIV-1) immunogen that can induce broadly reactive neutralizing antibodies is a major goal of HIV-1 vaccine development. Although rare human monoclonal antibodies (mAbs) exist that broadly neutralize HIV-1, HIV-1 envelope immunogens do not induce these antibody specificities. Here we demonstrate that the two most broadly reactive HIV-1 envelope gp41 human mAbs, 2F5 and 4E10, are polyspecific autoantibodies reactive with the phospholipid cardiolipin. Thus, current HIV-1 vaccines may not induce these types of antibodies because of autoantigen mimicry of the conserved membrane-proximal epitopes of the virus. These results may have important implications for generating effective neutralizing antibody responses by using HIV-1 vaccines.

A major obstacle to generating a successful human immunodeficiency virus–1 (HIV-1) vaccine is the inability to induce broadly reactive neutralizing antibodies after immunization with HIV-1 envelope proteins (Env) (1). HIV-1 infection and vaccination induce multiple types of antibodies, including antibodies to envelope variable loops, the CD4 binding site, and the chemokine receptor binding site. However, these antibodies do not control HIV-1 and are readily escaped by the virus (1). The human monoclonal antibodies (mAbs) 2F5 and 4E10 represent rare antibodies with broadly neutralizing activity made from B cells of HIV-1–infected humans, which react with conserved membrane-proximal amino acids in HIV-1 gp41 (24). Both antibodies have relatively long hydrophobic CDR3 regions (3, 4). A major conundrum has been that HIV-1 envelopes that express membrane-proximal epitopes fail to induce equivalent antibodies in animal models or humans (58). Long hydrophobic CDR3 regions are generally typical of poly-reactive autoantibodies (9), and HIV-1–infected patient B lymphocytes are driven to make polyclonal antibodies with reactivity to the endogenous phospholipid, cardiolipin (10). In light of these observations, we assayed 2F5 and 4E10 mAbs, along with other mAbs to HIV-1, for cardiolipin and other autoantigen reactivities.

The mAbs 2F5 and 4E10, two additional rare broadly reactive neutralizing mAbs (2G12 and IgG1b12) (1113), and 31 common human mAbs to HIV-1 Env were tested for reactivity with cardiolipin (14) (Table 1). Both 2F5 and 4E10 reacted with cardiolipin, whereas all 33 of the other human mAbs did not react. mAb 2F5 also reacted with histones and the centromere B autoantigen, whereas mAb 4E10 reacted with the systemic lupus erythematosus (SLE) autoantigen SS-A/Ro (Table 1). We also tested all human mAbs against the human epithelial HEp-2 cell line that is used in clinical antinuclear antibody assays (15). Both 2F5 and 4E10 also reacted with HEp-2 human epithelial cells in a diffuse cytoplasmic and nuclear pattern (14) (Fig. 1, A and B), further revealing polyspecific autoreactivity of both antibodies.

Fig. 1.

Reactivity of 2F5, 4E10, and IgG1b12 mAbs with human HEp-2 epithelial cells. (A) mAb 2F5 reacting with HEp-2 cells in a diffuse cytoplasmic and nuclear pattern. (B) mAb 4E10 reacting with HEp-2 cells in a pattern similar to 2F5. (C) mAb IgG1b12 reacting with HEp-2 cells in a diffuse cytoplasmic pattern, with nucleoli reactive in the nucleus. (Inset) Higher magnification of cells showing the nucleolar reactivity of IgG1b12 (arrows). (D) Negative reactivity of mAb 1.9F on HEp-2 cells. Antibody amounts per slide assayed in [(A) to (D)] were 3.75 μg per slide (150 μg/ml) of mAb. mAb 2F5 was positive on HEp-2 cells at 0.125 μg per slide (5 μg/ml). mAb 4E10 was positive on HEp-2 at 0.125 μg per slide (5 μg/ml), and IgG1b12 was positive at 1.25 μg per slide (50 μg/ml). All panels are shown at magnification ×200; the inset in (C) is ×400. Images shown are from an experiment representative of three performed.

Table 1.

Reactivity of HIV-1 Env human mAbs with autoantigens and human cells. All mAbs were negative in assays for reactivity with ribonucleoprotein, La (SSB), Sm, Scl-70, and Jo-1, except for Ku32 mAb, which reacted with Sm, and IgG1b12, which reacted with ribonucleoprotein. Ro (SSA), dsDNA centromere B, histone, and cardiolipin antibody values are in relative units based on a standard curve. –, negative.

mAb type and antibody name Cardiolipin HEp-2 cell reactivity Ro (SSA) dsDNA Centromere B Histones
Membrane-proximal external region (2F5) 47 +Cytoplasmic nuclear 290 - 1776 1011
Membrane-proximal external region (4E10) 15,434 +Cytoplasmic nuclear 221 - - -
CD4 binding site (IgG1b12) - +Cytoplasmic nucleolar - 513 479 185
CD4 binding site (F1, 5E, 25G) - - - - - -
Adjacent CD4 binding site (A32) - - - - 1131 -
Adjacent CD4 binding site (1.4G) - - - 768 1422 539
Adjacent CD4 binding site (1.4C, 4.6H, 4.11C) - - - - - -
Third variable loop (CO11, F2A3, F3.9F, LA21, 447-52D) - - - - - -
gp41 immunodeficient region (7B2, KU32) - - - - - -
gp41 immunodeficient region (2.2B) - +Intermediate filament - - 314 -
C1-C4 gp120 (8.2A, 2.3B) - - - - - -
C1-C4 gp120 (EH21, C11) - - - - - -
Glycan-dependent (2G12) - - - - - -
CCR5 binding site (1.7B, 2.1C, LF17, E51, 1.9F, LA15, 4.8E, LA28, 1.9E, E047, 2.5E, ED10) - - - - - -
Positive control serum 34 +Homogeneous nuclear 1365 228 624 281
Negative control <16 - <120 <120 <120 <120

Of the two other rare neutralizing mAbs, one mAb, 2G12 (11, 13), was not autoreactive, whereas another mAb against the CD4 binding site, immunoglobulin G 1b12 (IgG1b12) (12), reacted with ribonucleoprotein, double-stranded DNA (dsDNA), centromere B, and histones, as well as with HEp-2 cells in a cytoplasmic and nucleolar pattern (Table 1 and Fig. 1). Of the 31 more common mAbs to HIV-1 studied, only two mAbs with specificity for binding near the CD4 binding site (A32 and 1.4G) and two mAbs to a non-neutralizing gp41 epitope (2.2B and KU32) showed any evidence of autoantigen reactivity (Table 1).

To determine whether 2F5 and 4E10 were similar to anticardiolipin antibodies found in SLE or the antiphospholipid antibody syndrome (APS) (1620), both mAbs were tested for lupus anti-coagulant activity (18) and for the ability to bind to prothrombin (19), β2 glycoprotein-1 (17), phosphatidylserine (PS) (16), phosphatidylcholine (PC) (16), phosphatidylethanolamine (PE), and sphingomyelin (SM) (14). Whereas 2F5 was negative for all of these reactivities, 4E10 had lupus anticoagulant reactivity (14); reacted strongly with PS, PC, and PE; reacted weakly with SM (Fig. 2A) and prothrombin (14); and did not react with β2 glycoprotein-1 (14).

Fig. 2.

Assay of mAbs 2F5 and 4E10 against lipids and specificity of 2F5 and 4E10 mAb binding to cardiolipin. Microtiter plates were coated with 0.2 to 10 μg of lipid, blocked with 3% phosphate-buffered saline (PBS)–bovine serum albumin (BSA), and mAbs were diluted in PBS–Tween 20 (0.05%) containing 3% BSA and 2% normal goat serum. (A) Enzyme-linked immunosorbent assay (ELISA) reactivity of mAbs 4E10 (solid bars) and 2F5 (open bars) to CL, PS, PC, PE, and SM. Whereas both 4E10 and 2F5 reacted with cardiolipin, only 4E10 reacted with the other lipids tested. The reactivity of control human anti-CCR5 binding site mAb 1.7b was negative (33). Both mAbs 4E10 and 2F5 bound directly to cardiolipin, because cardiolipin binding in ELISA did not require the presence of bovine serum as a source of β2 glycoprotein-1; as an additional negative control, the reactivity of mAbs against empty methanol-coated plates was negative for 2F5 and was seen only at concentrations of 4E10≥2.5 μg per milliliter (33). (B) To show specificity of binding of mAb 2F5 to cardiolipin, we used 150 μg of 2F5 per milliliter and 500 μg of murine mAb to 2F5 idiotype 3H6 per milliliter, which blocks the neutralization of HIV-1 by mAb 2F5 (8). The 2F5 anti-idiotype significantly blocked the binding of mAb 2F5 to cardiolipin by a mean of 70% in three separate experiments (P < 0.03). In separate ELISAs, mAb 2F5 bound to cardiolipin with EC50 responses that ranged from 0.66 to 2.2 μM (33). (C) The dose response curve of 4E10 mAb binding to cardiolipin. The EC50 response of 4E10 binding (80 nM) was calculated from a four-parametric sigmoidal-curve fitting analysis. Binding data were acquired from an ELISA of 4E10 mAb binding (0.5 to 1000 nM) to cardiolipin coated on an ELISA plate (1.35 μg per well). (D) Soluble HIV-1 Env gp140 oligomers (CON-S) expressing the 4E10 epitope inhibit binding of 4E10 mAb to cardiolipin. The IC50 of inhibition of 4E10 binding to cardiolipin was calculated to be 145 nM. The inhibition assay was carried out with varying concentrations of gp140 (19.25 to 1230 nM) mixed with 10 μg of 4E10 mAb per milliliter, which were then added to wells containing 1.35 μg of cardiolipin. mAb 3H6 (1 mg/ml) (but not control mAb) also blocked the binding of mAb 2F5 to Sjögren's syndrome antigen A (SSA)/Ro, centromere B, and histones (33) as well as blocked to background the binding of MAB 2F5 to HEp-2 cells in indirect immunofluorescence assay (33). (E) The nominal epitope of the HIV-1 Env gp41 inhibits the binding of mAb 2F5 to cardiolipin. In this experiment, two gp41 2F5 epitope peptides, 2F5 amino acids 652 to 671 (QQEKNEQELLELDKWASLWN), and 2F5 amino acids 656 to 670 (NEQELLELKDWASLW-biotinylated) were used at concentrations ranging from 0 to 50 μM to block the binding of mAb 2F5 to cardiolipin (10 μg per well) (34). First, the apparent binding affinity of mAb 2F5 to each peptide was determined and found to be 0.34 nM for 2F5 652–671, and 0.47 nM for 2F5 656–670. (E) demonstrates that both nominal 2F5 epitope peptides inhibited 2F5 mAb binding to cardiolipin, with IC50 of 2F5 652 to 671 equal to 0.9 μM, and IC50 of 2F5 656 to 670 equal to 2.1 μM. In contrast, a scrambled version of 2F5 656 to 670 did not significantly inhibit mAb 2F5 binding at all concentrations (0 to 50 μM) tested. This is represented by the single diamond at the 50 μM concentration of scrambled 2F5 656 to 670 peptide. All data in [(A) to (E)] are representative of at least two experiments performed.

The specificity of mAb binding was demonstrated by the inhibition of 2F5 binding to cardiolipin by a 2F5 anti-idiotype mAb, as well as by gp41 peptides, and by the inhibition of 4E10 binding to cardiolipin by gp140 envelope protein (Fig. 2, B to E) (21). A murine mAb to 2F5 idiotype, 3H6, which blocks 2F5 neutralization of HIV-1 (21), inhibited binding of mAb 2F5 to cardiolipin by 70% (Fig. 2B). The amino acid sequence to which mAb 2F5 binds on HIV-1 gp41 includes amino acids 662 to 668 (Glu-Leu-Asp-Lys-Trp-Ala) (3, 22). Two synthetic peptides containing this sequence, but not a scrambled peptide, also blocked the binding of 2F5 to cardiolipin by 70% with an inhibitory concentration 50% (IC50) of 0.9 and 2.1 μM(Fig. 2E). mAb 4E10 bound to cardiolipin with a half-maximal molar concentration (effective concentration 50%, EC50) of 80 nM (Fig. 2C), and recombinant oligomeric HIV-1 envelope inhibited the binding of mAb 4E10 to cardiolipin with an IC50 of 145 nM. Thus, the Fab regions of 4E10 and 2F5 mAbs that are involved in binding to gp140 and neutralizing HIV-1 are also involved in binding to cardiolipin. However, the apparent affinity of each with cardiolipin was distinct, with mAb 4E10 nanomolar, whereas the apparent affinity of mAb 2F5 was micromolar.

Antibodies to cardiolipin can be found in patients with disordered immunoregulation caused by autoimmune disease or infection (10, 2325). Antibodies to cardiolipin are induced by syphilis, leprosy, leishmaniasis, Epstein-Barr virus, and HIV-1 (10, 2325). Unlike antibodies to cardiolipin found in SLE, “infectious” antibodies to cardiolipin, such as those found in HIV-1 infection, are rarely pathogenic and are transient (2325). Consistent with this, mAbs 2F5, 2G12, and 4E10 have been administered to HIV-1–infected patients with no observed side effects (2).

Neutralizing antibodies against 2F5 and 4E10 epitopes are rarely made in HIV-1–infected humans (5, 6) or in the setting of immunization (7, 8). The observation that human mAbs against these epitopes are polyspecific autoantibodies may explain why these antibodies are rarely made in humans. Autoreactive B cell clones with long CDR3 lengths are normally deleted from the repertoire or made tolerant to self antigens, thereby preventing antibody production (9). Thus, it is conceivable that HIV-1 may have evolved to escape membrane-proximal antibody responses by having conserved neutralizing epitopes as mimics of autoantibody epitopes. Similarly, these data also raise the hypothesis that current HIV-1 vaccines do not routinely induce robust membrane-proximal antienvelope neutralizing antibodies, because antibodies targeting these epitopes are derived from autoreactive B cell clones that are normally deleted or made tolerant upon antigenic stimulation by HIV-1 Env.

Finally, these observations may also explain the rare occurrence of HIV-1 in SLE patients who may be unable to delete these self-reactive clones (26, 27). If broadly neutralizing antibodies to HIV-1 are made in the context of disordered B cell immunoregulation in autoimmune disease, then autoimmune patients may be fully or partially protected on this basis (27, 28). Polyspecific autoantibodies made by innate B1 and marginal zone B cells are the first line of defense against infections (29, 30). The study of the regulation of polyspecific autoantibody production and the B cells from which they are derived may provide insights into how to induce broadly neutralizing antibodies in the setting of HIV-1 vaccination.

Supporting Online Material

Materials and Methods

Fig. S1

Table S1


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article