INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The coding sequences of the human immunodeficiency virus type 1 (HIV-1) envelope have been used extensively to classify these viruses genetically into at least 11 clades, designated A to K, in group M (major) and into groups O (outlier) and N (new) (18, 19, 22, 26, 30, 33, 44). However, several studies have shown that the immune response to HIV infection is neither clade specific nor clade restricted (14, 23, 36, 37, 48, 50, 51). Sera and monoclonal antibodies (MAbs) derived from HIV-1-infected persons cross-react with HIV-1 isolates or proteins from different clades, suggesting that despite the genetic diversity that distinguishes these viruses, they have common antigenic epitopes (37, 51). The results of these studies indicate that antigenic regions that are conserved across different HIV clades and those that are distinct could be important for designing a polyvalent vaccine potent against a broad spectrum of HIV isolates. However, in order to understand how to construct a polyvalent HIV-1 vaccine, immunologically defined groups of HIV-1 isolates need to be identified along with the shared and distinguishing immunogenic epitopes that induce protective immunity.

Immunologic classifications of HIV-1 have been published. Those based on studies with sera from HIV-1-infected subjects defined HIV-1 "serotypes," while classifications based on studies with MAbs defined HIV-1 "immunotypes." For example, studies comparing genetic and immunologic classifications of HIV-1 by examining isolates of group M and group O with homologous and heterologous sera revealed that HIV-1 could be grouped immunologically and that genetic subtypes did not correlate with neutralization serotypes (39, 47). Similarly, multivariate analysis of a neutralization matrix comprised of 14 sera and 16 primary isolates identified eight "neutralization serotypes" (37). However, since these cluster analyses were based on data generated with polyclonal sera, the conserved epitopes characterizing each serotype could not be deciphered.

In order to identify conserved epitopes specific for each immunologically defined group of viruses, MAbs that are directed at defined epitopes would be needed. Using MAbs, such HIV-1 immunotypes were determined by multivariate analysis of data from the immunochemical reactivities of 1,176 combinations of human anti-V3 MAbs and V3 peptides which were representative of 56 viruses from eight clades (A to H) obtained from patients around the world (51). The analysis revealed seven immunologically defined groups of peptides, each containing peptides from more than one clade. Inspection of the amino acid sequences of the peptides in each of the peptide clusters revealed unique "signature sequences" that suggested structural motifs characteristic of each V3-based immunotype.

In recent studies using anti-HIV-1 gp120 and gp41 MAbs derived from patients infected with HIV-1 clade B or E, we studied the antigenic conservation of epitopes expressed on the surface of intact, native primary HIV-1 isolates of group M, clades A to H, obtained from patients around the world (34, 36). In these studies we observed that epitopes located in the V3 and C5 regions of gp120 and in the cluster I region of gp41 are shared and well exposed compared to the CD4-binding domain (CD4bd), V2, and C2 regions of gp120 and the cluster II region of gp41. This work confirmed previous studies showing that HIV-1 genetic clades do not correspond with serotypes or immunotypes and suggested that HIV-1 isolates could be grouped according to immunotypes that could enlighten our approach to designing a polyvalent HIV vaccine.

In the present study, we again applied a multivariate method of data analysis to the matrix of data generated with anti-HIV-1 human MAbs binding with intact, native HIV-1 primary isolates of group M, clades A to H. The analysis revealed three immunotypes of HIV-1 and the antigenic epitopes that are specific for each.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

HIV-1 isolates. The 26 HIV-1 group M isolates of clades A to H studied have been described previously (36). The clade to which each isolate belongs is shown in Fig. 1. These HIV-1 isolates were obtained from patients in Cameroon (CA1, CA4, CA5, CA13, and CA20), Belgium (VI191), Uganda (92UG021 and 92UG001), France (BX08), the United States (MNp, IIIB, and JR-FL), Senegal (SG2728), Zimbabwe (2036), Rwanda (92RW021), Zambia (ZB18), the Ivory Coast (CI13), Zaire (MAL), Brazil (93BR019 and 93BR029), Thailand (92THA009, 92THA011, BK131, and CM235), and Gabon (VI525 and VI526). HIV-1 viral stocks were prepared as previously described (38). Briefly, 1 ml of p24-positive HIV-1 culture supernatants was used to infect 3-day phytohemagglutinin-stimulated HIV-negative donor peripheral blood mononuclear cells (PBMCs) (38, 39). After 2 to 3 weeks of culture, the culture supernatants from infected PBMCs were aliquoted (1 ml/tube) and stored in liquid nitrogen until use. The p24 concentration in each virus stock was quantitated using a noncommercial p24 enzyme-linked immunosorbent assay (ELISA) (34).

View larger version (131K): [in this window] [in a new window]   FIG. 1.   Binding profiles of human MAbs and HIV-1 isolates (a) grouped according to MAb specificities in the various envelope regions (anti-V3, anti-C5, and anti-gp41 cluster I) and virus clades (A to H) and (b) classified according to MAb groups (I to V) and virus immunotypes (1 to 3). The binding capacity of each MAb to each virus was measured by ELISA by assaying the amount of p24 released from MAb-bound virus after lysis. The p24 levels detected are represented as log-transformed data and were computed from the data presented in Table 1. Means shown were computed by adding a value of 1 to the amount of p24 captured by each MAb-virus combination (shown in Table 1). Subsequently, the values were transformed logarithmically and the means were calculated. For clarity, these logarithmic values have been color coded. Color ranges between 0.0 and 1.04 log (yellow-gray) correspond to lack of binding. Color ranges from >1.04 log (blue-dark blue-red) correspond to weak to strong binding. In both panels, the identity of the isolate is followed by a suffix indicating its coreceptor usage. In panel b, the prefix of the virus designation denotes the virus clade, and the prefix of the MAb designation denotes the MAb specificity.

Human anti-HIV MAbs. Twenty-eight human anti-HIV-1 MAbs were used to elucidate the existence of immunotypes among the 26 HIV-1 isolates. These MAbs included 19 specific for the V3 region of gp120 (MAbs 447-52D, 419-D, 694/98-D, 838-D, 412-D, 1006-15D, 504-D, 257-D, 537-D, 311-11D, 386-D, 418-D, 1334, 782-D, 453-D, 908-D, 1027-15D, 268-D, and 1324E), 4 specific for the C5 region of gp120 (MAbs 670-D, 1331A, 858-D, and 989-D), and 5 specific for the cluster I region of gp41 (MAbs 50-69, 246-D, 240-D, 1367, and 181-D). The immunochemical specificity of each of these MAbs has previously been described (3, 14-16, 34, 49, 51, 52). All MAbs except 1324E were produced using PBMCs from HIV-1 clade B-infected persons. MAb 1324E was made from the cells of a clade E-infected individual (14). Human MAbs 860-55D to parvovirus B19 (11) and 246-D to gp41 (49) were used as negative and positive controls, respectively, as explained below. Human anti-p24 MAb 91-5 (13) was used in a noncommercial p24 ELISA to measure virus capture.

Virus-binding assay. The virus-binding assay has been described in detail in previous studies (34, 36). Briefly, ELISA wells were coated at 4°C overnight with 100 µl of MAb at 10 µg/ml. After washing with phosphate-buffered saline (PBS) and blocking with 0.2 ml of 3% bovine serum albumin (BSA) in PBS (BSA-PBS), virus was added (100 µl/well at 100 ng of p24 per ml). After incubation at 37°C for 1 h, each well was washed with RPMI 1640 to remove unbound virus, and the bound virus was lysed with 250 µl of 1% Triton X-100. Positive and negative controls were performed with each isolate in each experiment. The amount of p24 captured in each MAb-virus test combination was quantified using human anti-p24 MAb 91-5 (13) in a noncommercial p24 ELISA (4).

Cluster analysis. The multivariate method of analysis has been used extensively and described in studies analyzing neutralization using HIV-1 isolates of different clades and sera from HIV-1-infected persons (37), in studies of the dynamics of neutralization escape mutants in a chimpanzee naturally infected with SIV_cpz (35), and in analyses of the immunoreactivity of V3 peptides representing different HIV-1 clades with human anti-HIV-1 MAbs (51). Independent studies have employed this method to study the interaction between drugs and receptors and between viruses and antiviral compounds (2, 25). In the context of the study below, the analysis is based on the principle of grouping MAbs and viruses with similar specificities and profiles of reactivity.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Binding patterns of anti-gp120 and anti-gp41 MAbs to intact, native primary HIV-1 isolates of group M, clades A to H. In our previous work, we examined the ability of MAbs directed at epitopes in the V2, C2, V3, CD4bd, and C5 regions of gp120 and at clusters I and II of gp41 to bind to intact, native HIV-1 primary isolates of group M, clades A to H (36). The studies revealed that MAbs directed at epitopes in the V3 and C5 regions of gp120 and in the cluster I region of gp41 bound better than MAbs to epitopes in the V2, C2, and CD4bd regions of gp120 and in the cluster II region of gp41. These data demonstrated that the former regions are better exposed on HIV-1 primary isolates, and therefore 28 MAbs to V3 and C5 of gp120 and to cluster I of gp41 were used here to identify HIV-1 immunotypes. Table 1 shows the binding values for the 28 MAbs directed to these regions on 26 intact, native primary HIV-1 viruses of group M. The various HIV-1 isolates tested are listed according to their clades, and the MAbs are listed according to their specificities. Only binding levels of >11 pg of p24 per ml are considered positive (see above). Twenty-four of the 26 isolates (92%) gave mean binding levels with the 28 MAbs of >11 pg of p24/ml (shown in the right-hand column of Table 1), and 23 of the 28 MAbs (82%) gave mean binding levels with the 26 virus isolates of >11 pg of p24/ml (shown in the bottom row of Table 1).

Mathematical analysis of the data matrix. In previous studies, analysis of the immunoreactivity of various human anti-V3 MAbs with V3 peptides derived from 56 HIV-1 isolates was used to test the validity of the mathematical methods (51). The mathematical analyses classified the MAbs studied into groups which correlated well with their immunochemical properties. Given the concordance of the mathematically and immunochemically derived groupings of the MAbs, we used the same mathematical techniques to analyze for the presence of immunologically defined clusters among the 28 anti-HIV-1 MAbs on the basis of the profiles of their reactivities with the 26 intact, native HIV-1 isolates, i.e., on the pattern of specific binding of each MAb to each of the 26 isolates. Similarly, analyses were done to group the 26 viruses into immunotypes on the basis of their profiles of reactivities with the 28 MAbs. The results of the mathematical clustering of the 28 MAbs and the 26 viruses are displayed in Fig. 1b and as dendograms in Fig. 2 and 3. The 28 MAbs were classified into five Ab groups, and the 26 viruses were grouped into three immunotypes. The prefix of each MAb listed in Fig. 1b and 2 identifies the region on the viral envelope to which the MAb is directed, and the prefix of each isolate in Fig. 1b and 3 identifies the clade to which the virus belongs.

View larger version (21K): [in this window] [in a new window]   FIG. 2.   Dendrogram of 28 MAbs showing their immunologic relationships defined by binding with 26 intact, primary HIV-1 isolates. The dark line denotes the point in the tree at which model-based clustering delineates the most probabilistic number of clustersfive in this case. Each MAb group (cluster) is numbered I to V. The distance between merged groups can be determined on the scale shown on the y axis.

View larger version (22K): [in this window] [in a new window]   FIG. 3.   Dendogram of 26 intact, primary HIV-1 isolates showing their immunologic relationships defined by binding with 28 MAbs. The dark line denotes the point in the tree at which model-based clustering delineates the most probabilistic number of clustersthree in this case. Each viral immunotype is numbered 1 to 3. The distance between merged immunotypes can be determined on the scale shown on the y axis.

Cluster analysis of the MAbs. The five groups of MAbs and their relationships to one another, as revealed by the multivariate analysis, are shown in the dendogram in Fig. 2. The analysis defined Ab groups which were either homogeneous or heterogeneous with respect to the specificities of the MAbs they contained. Three of five Ab groups (I, II, and III) contained anti-V3 MAbs (with the one exception of MAb 50-69 to gp41 in Ab group II). The other two Ab groups (IV and V) contained mixtures of V3, C5, and gp41 MAbs.

Cluster analysis of the isolates. In order to define virus immunotypes, the dendogram displaying the results of the analysis of the grouping of isolates on the basis of their reactivities with the MAbs was cut at the same height as the MAb dendogram (Fig. 2). Only three HIV-1 immunotypes were identified (Fig. 3). The clustering of the isolates did not correlate with the viruses' genetic subtypes (denoted by the prefixes of the isolates in Fig. 3), with coreceptor usage (denoted by the suffixes of the isolates, Fig. 3), or with the geographic area from which the isolates were derived. Thus, none of the three HIV-1 immunotypes contained isolates from a single country or continent. For example, HIV-1 immunotype 1 comprises isolates from patients in Africa, the United States, Asia, and South America. These results are consistent with earlier work showing that HIV-1 serotypes identified on the basis of neutralization of HIV-1 isolates by sera and HIV immunotypes identified on the basis of V3 peptide reactivity with MAbs do not correlate with genetically defined clades (37, 51). The results reported here, like the previously published conclusions, suggest that isolates of different clades that belong to the same immunotype share distinguishing epitopes. These results also suggest that despite the genetic variability that marks the HIV-1 family of viruses, there are certain immunologic characteristics that are widely shared among virus isolates and others that divide the viruses into immunologically identifiable groups.

Reactivity and specificity profiles of HIV-1 immunotypes with Ab groups. In order to identify epitopes (i) that are conserved on isolates belonging to the same HIV-1 immunotypes, (ii) that differentiate the members of the three immunotypes, and (iii) that are shared among all immunotypes, the average reactivity of viruses in each HIV-1 immunotype with each of the individual MAbs was examined. The results are shown in Fig. 4. The data show that MAbs belonging to Ab group I bound preferentially to HIV-1 isolates belonging to HIV-1 immunotype 3 rather than to isolates of HIV-1 immunotypes 1 and 2. This can also be observed by examining the overall pattern of binding in Fig. 1b. In contrast, MAbs of Ab group II bound better to isolates of both HIV-1 immunotypes 1 and 3 than to viruses of immunotype 2 (Fig. 4 and 1b), and MAbs of Ab group III bound preferentially to isolates of HIV-1 immunotype 1 (though overlap with some isolates in immunotype 2 was observed). While the MAbs of Ab group IV generally bound poorly to the HIV-1 isolates and the MAbs of Ab group V generally bound strongly, the MAbs in these two Ab groups did not distinguish among the three immunotypes.

View larger version (29K): [in this window] [in a new window]   FIG. 4.   Average p24 capture profiles by each MAb for the isolates in the three viral immunotypes. The MAbs are arranged according to their respective groups (I to V) on the x axis, and their reactivity profiles with the various viral immunotypes (1 to 3) are shown in the graph. The reactivity of each individual MAb with all isolates belonging to each viral immunotype is plotted in the graph. The broken line represents the cut-off (1.04 logs), as described in the text. Only values above the broken line represent significant positive binding.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

In order to protect against organisms that exhibit significant genetic variation, polyvalent vaccines are needed. Genetic variations among microorganisms such as Streptococcus pneumoniae and poliovirus have been documented, and the antigenic relationships of each have been studied (8, 17, 21, 29), resulting in definition of immunologic characteristics that distinguish each into serotypes. In the case of poliovirus, for example, three serotypes have been documented and proven critical to polio vaccine development (9, 20, 42).

Given the extreme variability of HIV-1, it is probable that a polyvalent vaccine will be needed for protection against the large number of strains. However, in order to understand how to construct such a polyvalent vaccine that will be useful against a maximum number of strains, serotypes or immunotypes of HIV-1 must be identified.

It is clear from several independent studies that HIV-1 genotypes do not generally correlate with immunotypes, although clades B and E appear to represent a special case (28). The poor correlation between HIV-1 genotypes and immunotypes is a consistent finding whether HIV peptides or recombinant proteins were analyzed with sera or MAbs. Even sera from individuals immunized with vaccine constructs based solely on clade B CXCR4-tropic strains have been found to cross-react with V3 peptides from diverse clades and to neutralize HIV-1 primary isolates that are from clades B and C which are CXCR4-tropic, CCR5-tropic, or dual-tropic (46), and cross-clade reactive cell-mediated immunity has also been induced with vaccines (10). These studies suggest that selection of vaccine components should not be based entirely on the choice of immunogens representing distinct HIV-1 genotypes, but rather should rely on selecting representatives of HIV immunotypes. However, studies aimed at defining HIV-1 serotypes or immunotypes, and the epitopes that distinguish them, have been sparse. Those studies that have been performed were based on the use of polyclonal sera (which made the definition of critical epitopes an impossible task [37]) or on peptides or recombinant envelope proteins that do not mimic the native structure of the virus (51).

The data presented above show that the patterns of binding of MAbs to intact virions do not correspond with the country or continent of origin of the virus or with the clade or coreceptor which characterizes the isolate. Yet many of these viruses have common antigenic epitopes that could be useful for selecting the minimum number of HIV-1 strains to be combined into a polyvalent vaccine in order to represent and protect against the maximum number of HIV-1 isolates.

The best-exposed epitopes that occur on the surface of intact virions are found in the V3 and C5 regions of gp120 and in antigenic cluster I of gp41 (spanning amino acids 579 to 613 of the envelope) (36). Therefore, the presence of these epitopes on the surface of virions of 26 isolates was examined immunologically and then analyzed using a mathematical approach, previously described, for identifying how to best classify diverse groups of HIV-1 isolates and MAbs (51). This method, applied to the interaction of MAbs and intact virions, revealed that these 26 HIV-1 isolates, from diverse clades and geographic origins, could be divided into a small number of immunotypes. This is not dissimilar to results derived from other studies and mathematical analyses based on data matrices generated (i) by neutralization assays using polyclonal sera from HIV-1-infected subjects with isolates from HIV-1 groups M and O (37), (ii) with immunochemical data matrices generated with human anti-HIV-1 MAbs and V3 peptides (51), or (iii) by cluster analysis of V3 peptides and polyclonal sera from HIV-infected patients from around the world (41). In each case, these independent studies revealed that there are fewer HIV-1 immunotypes than genotypes and that these immunotypes do not correlate with clades or coreceptor usage (37, 41, 51).

While studies of HIV-1 neutralization with polyclonal sera revealed the presence of neutralization serotypes (37), the epitopes involved could not be identified, although the complex nature of the humoral immune response would suggest that Abs of multiples specificities might contribute to the findings. Interestingly, the study presented above demonstrates that virus immunotypes were distinguished primarily on the basis of anti-V3 MAbs. Thus, of the MAbs that best distinguished among the three virus immunotypes, 17 of 18 were directed against V3 (Fig. 4). The role of the anti-V3 response in protective immunity against primary HIV-1 isolates is still highly controversial: only one human MAb, 447-52D, has been shown to have significant neutralizing activity for primary isolates (7). However, there is growing evidence that Abs to the conformational epitopes in V3 may be quite potent (40, 43; P. F. Zhang, X. Chen, J. B. Margolick, J. E. Robinson, S. Zolla-Pazner, M. N. Flora, and G. V. Quinnan, Jr., submitted for publication). While none of the anti-V3 MAbs identified here as MAbs that can distinguish among the three HIV-1 immunotypes have been shown to have broad or potent neutralizing activity when tested separately, these Abs could have biological effects when present in combination with each other, with Abs in the other two Ab groups, or with Abs not tested in this study. Indeed, many examples of the additive and synergistic interaction of Abs against HIV-1 exist in the literature (24, 27, 45), and the data presented here provide an indication as to which of the myriad of possible combinations of MAbs might be useful to test in synergy experiments to determine which combinations might most potently enhance neutralization of primary isolates and which combinations might give the broadest neutralization among diverse isolates.

To date, HIV-1 vaccine constructs have used immunogens derived from, at most, two HIV-1 clades, B and E (1). While the results of these trials are not yet available, it is unlikely that such a vaccine will protect against the majority of all HIV-1 viruses. It is equally unlikely that the preparation of a polyvalent vaccine including immunogens derived from all the many HIV-1 clades will be feasible. Thus, it is imperative to generate data which will help to estimate how many isolates, and which ones, need to be included in a vaccine that will give the maximum breadth of protection. For this, identification of HIV-1 immunotypes is a critical step.

The results of the study presented above clearly demonstrate that immunotypes of HIV-1 can be defined, as they have been for other microorganisms, and that each HIV-1 immunotype comprises isolates of different genetic subtypes from different geographic origins and of strains with different coreceptor tropisms. These findings give rise to a testable hypothesis, that polyvalent vaccines comprising representatives of a few HIV-1 immunotypes may give rise to broader immunity than polyvalent vaccines comprising representatives of the many HIV-1 genotypes.

Meanwhile, the different HIV-1 immunotypes, as defined here, can be defined on the basis of both common and distinguishing epitopes. Shared epitopes are recognized on most virions by MAbs that belong to Ab group V that are specific for antigenic determinants at the tip of the V3 loop, in the C-terminal portion of gp120, and at the apex of the immunodominant loop of gp41. In contrast, immunotypes can be distinguished primarily on the basis of the recognition by MAbs of what appear to be distinct shapes in the V3 regions of the three viral immunotypes. These shapes are dictated, only in part, by the V3 amino acid sequence; as shown previously, these shapes are also profoundly influenced by conformational parameters to which the rest of the gp120 molecule contributes (12, 51). Thus, immunoreactivity to V3 peptides is only of limited value. The ability to probe the antigenic nature of the intact virus particle, both immunochemically and functionally, thus promises to yield critical information pertinent to the human antiviral immune response and to the generation of broad protective immunity.