| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Department of Cell Research and Immunology, Tel-Aviv University, Tel-Aviv, Israel
1Correspondence: Department of Cell Research and Immunology, Tel Aviv University, Ramat Aviv, Aviv 69978, Israel. E-mail: gershoni{at}tauex.tau.ac.il
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Key Words: Mapitope algorithm • combinatorial phage display peptide library • epitope mapping • vaccine design
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
, scores of vaccines have been developed, many of which are quite effective, as reflected by the fact that small pox has been declared eradicated, and polio and measles just about so (1
2
3)
. New approaches to vaccines can replace the need for whole killed or attenuated viruses, which can be costly and sometimes dangerous. The advent of recombinant technologies has enabled the production of subunit vaccines [e.g., hepatitis A vaccine by Aventis Pasteur (4)
, hepatitis B vaccine by Merck or SmithKline Beecham (5)
, and influenza vaccine by Chiron Behring or Berna Biotech (6)
]. The demonstration of naked DNA as an effective vaccine vector (7)
has provided hope for a new generation of novel, effective, and commercially viable vaccines. Despite this sense of optimism and accomplishment, however, there are still devastating diseases that simply have not complied with state-of-the-art vaccine paradigms and continue to leave havoc in their wake. HIV (HIV-1) is such a pathogen and, for the moment, seems vaccine resistant.
Nevertheless, reports have been published illustrating that sterilizing immunity can be achieved against this virus. R. M. Ruprecht et al. (8)
, for example, have shown that administration of a defined mixture of four HIV-neutralizing monoclonal antibodies (mAbs 2F5, b12, 2G12, and 4E10) to Rhesus monkeys can protect them against a SHIV challenge. It was also shown that HIVIG compiled from the serum of HIV-infected individuals often shows an effective neutralizing activity against an impressive broad range of HIV-1 primary isolates representing a diversity of clades (9)
. Moreover, HIV-infected, long-term asymptomatic individuals tend to have potent serum antibodies that neutralize HIV isolates well (10
, 11)
. All this shows that, in nature, highly effective cross-reactive neutralizing antibodies against HIV are being generated. The goal therefore is to produce immunogens that are able to elicit such neutralizing activity. One way to reach this goal could be to backtrack from the neutralizing antibody (Ab) to its corresponding epitope and use the latter, or its parts, as an immunogen.
Recently we proposed such a "reversed immunological" approach (12)
. Screening a monoclonal antibody (mAb) of interest against a combinatorial phage display library of random peptides leads to the production of a panel of different peptides that have the common attribute of specifically binding that mAb. Each peptide is postulated to represent only a fragment of the mAb’s epitope. However, collectively the peptides contain information that allows a prediction of the epitope on the surface of its antigen. The prediction is made using "Mapitope" (E. M. Bublil et al., unpublished results), a computer algorithm that uses the peptide panel as input and predicts epitope candidates on the surface of the antigen as output. The algorithm was developed using mAb:antigen cocrystals as reference models. In each case the model mAb was screened against random peptide libraries to generate mAb-defining peptide panels. These were used to predict epitopes on the corresponding antigens, and the predictions were compared with the genuine epitopes defined in the cocrystals (13
14
15
16)
. The algorithm was found to be accurate in predicting the conformational discontinuous epitopes of the 13b5 mAb that binds HIV-1 p24, the 17b mAb that binds HIV-1 gp120 (12)
, and, more recently, the Trastuzumab (Herceptin®) Ab that binds the Her2 receptor and the anti-FVIII mAb, Bo2C11 (15
, 16)
.
In this report we describe a comprehensive analysis of the epitope of the HIV-1 neutralizing mAb b12 (11
, 17)
. We believe that the predictions are enlightening with respect to the b12 epitope in particular and a comprehensive example for epitope mapping in general.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Vectors
The pCMV-Tag-tpaJRFLgp120 was kindly provided by Dr. Dennis Burton (Scripps Institute, La Jolla, CA, USA). This plasmid encodes gp120JRFL from the JRFL primary isolate of HIV-1, and contains the tissue plasminogen activator (TPA) leader, which allows secretion of gp120JRFL into the medium of transfected cells (18)
.
Biopanning and amplification
The panning procedures were carried out as described in ref. 19
. Briefly, wells of 6-well tissue culture plates (Corning Inc., Corning, NY, USA) were coated with 70 µg/ml of protein G (Sigma Chemical Co., St. Louis, MO, USA) in 50 mM Tris-HCl buffer, pH 7.4, 150 mM NaCl (Tris-buffered saline, TBS). The plate was blocked with 0.25% gelatin in TBS (w/v) (TBSG) for 2 h at room temperature, washed briefly with TBS, then incubated for 4 h with 35–70 µg/ml of the screened mAb. Unbound Ab was washed out with TBS, and the plate was incubated overnight at 4°C with 1011 phages from the random peptide library in TBSG. The plate was extensively washed with TBS and bound phages were eluted using elution buffer (0.1M HCl adjusted to pH 2.2 with glycine, 1 mg/ml BSA) for 10 min at room temperature with gentle agitation, and the eluate was neutralized with 1M Tris-HCl pH 9.1. Eluted phages were amplified: a 5 ml culture of E. coli DH5
was grown at 37°C, 225 rpm until 
1.5 optical density (OD)600. Eluted phages were added to the bacteria and incubation was continued for an additional 30 min. The culture was then added to a flask containing 100 ml of TYx2 medium, and 1 h later tetracycline was added (50 µg/ml). Incubation was continued overnight at 37°C. Phages were precipitated overnight at 4°C using 33% (w/v) polyethylene glycol with 3.3M NaCl (PEG/NaCl).
Immuno-screening
Phages from the PEG precipitate (see above) were resuspended in 110 µl of TBS and applied onto nitrocellulose membrane filters using a vacuum manifold. The membranes were blocked with 5% milk (w/v) in TBS (m-TBS) for 1 h, then gently rocked for 2 h at room temperature, or overnight at 4°C, with 2–3 µg/ml of the mAb of interest dissolved in m-TBS. The membranes were washed six times in TBS and reacted with HRP-conjugated donkey anti-human IgG Ab (Jackson, West Grove, PA, USA) diluted 1:5000 in m-TBS, washed, and signals were developed using the Enhanced Chemo-Luminescence (ECL) reaction.
DNA preparations and DNA sequencing
Double-stranded DNA was obtained using the QIAprep® Spin Miniprep Kit (QIAGEN GmbH, Germany). Sequences of the foreign inserts in the affinity-selected phages were obtained using the primer 5'-GGTCAGACGATTGGCCTTG-3', an antisense primer that anneals to a sequence 238 bp downstream from the insert.
Mutagenesis
Mutations in gp120 were generated using PCR in which the PFU turbo DNA polymerase (Stratagene, La Jolla, CA, USA) was used to extend a pair of mutated primers using the pCMV-Tag-tpaJRFLgp120 plasmid as the template. The PCR product was then treated with Dpn-I enzyme (NEB, Beverly, MA, USA), which is specific for methylated DNA, digesting only the parental DNA template. The DNA containing the desired mutations was then used to transform E. coli DH5 followed by plating on selective medium plates (LB ampicillin 100 µg/ml). The DNA containing the desired mutations, confirmed by sequencing, was used to transfect 293T cells (5–10 µg of mutated or wild-type (WT) pCMV-TagtpaJRFLgp120 DNA). The media containing mutated or WT gp120JRFL were collected for further use.
Modified phage displayed peptides were produced using two complementary mutated oligonucleotides coding for 12 amino acids flanked by two cysteine residues. Oligos were designed with Sfi-I sites at either end compatible with the fth-1 vector. The oligonucleotides were annealed, then digested with Sfi-I (NEB, Beverly, MA, USA) and introduced into the fth-1 vector as described previously (20)
.
Binding of antibodies to mutated gp120
Nunc-immunoTMModules ELISA wells (Nunc, Roskilde, Denmark) were coated overnight with 100 µl of 5 µg/ml of the anti-gp120 Ab D7324 (International Enzymes Inc., Fallbrook, CA, USA) in TBS, washed twice with TBS, blocked for 1 h with m-TBS, and incubated for at least 4 h with 100 µl of TBS with 0.5% milk (w/v) containing equal amounts of WT or mutated gp120JRFL. Wells were washed five times with TBS, then reacted with 0.1 µg/ml of b12, m14 mAbs, or CD4 in TBS containing 0.5% milk (w/v) for overnight incubation. Wells incubated with CD4 were washed five times with TBS, then reacted for 1 h with 3 µg/ml mouse anti-CD4 mAb (CG9; see ref. 21
) in TBS containing 0.5% milk (w/v). Wells were washed again and incubated with HRP-conjugated goat anti-human Ab (for b12 and m14) (Jackson) and with HRP-conjugated rabbit anti-mouse Ab (for CG9) (Jackson) in TBS containing 0.5% milk (w/v). Wells were washed five times with TBS and reacted with the TMB ELISA substrate (Chemicon International, Temecula, CA, USA). Absorbance was measured at 650 nm.
Competition ELISA
Nunc-immunoTMModules ELISA wells were coated with 5 µg/ml D7324 mAb and blocked as described above, then incubated with 5 µg/ml of gp120451 in TBS containing 0.5% milk (w/v) for 2 h. Wells were washed five times with TBS. The captured gp120 was then incubated with varying concentrations of b12, m14, or b6 mAbs, in TBS with 0.5% milk (w/v) for 1 h, followed by addition of 5 µg/ml of biotinylated 17b mAb for an additional hour. Wells were washed five times and 1:2000 HRP-conjugated streptavidin (Jackson) in TBS with 0.5% milk (w/v) was added for 1 h. The wells were washed with TBS and reacted with the TMB ELISA substrate. Absorbance was measured at 650 nm.
The Mapitope algorithm
Mapitope is an updated user-friendly version of the algorithm published by D. Enshell-Seijffers et al. (12)
and predicts epitopes based on the assumption that the panel of peptides derived from a random peptide library collectively represents the epitope of the mAb, which they bind. The computer program that implements the Mapitope algorithm was written in C++ and is freely available for academic use.
The underlying principle of Mapitope is that the simplest meaningful fragment of an epitope is an "amino acid pair" (AAP) of residues that lies within the footprint of the epitope. These AAPs can be related to one another on the surface of the antigen such that a cluster of pairs is defined that constitutes the majority of the epitope footprint (i.e., the epitope is in essence a cluster of connected AAPs). The AAPs of the epitope need not be consecutive tandem residues of the antigen, but often are the result of a juxtaposition of distant residues brought together through folding of the polypeptide chain. Accordingly, we defined the distance parameter, D. A legitimate AAP can be considered as such when the distance between the two corresponding carbon alphas is less than D. AAPs of the epitope are simulated by tandem residues of the peptide’s affinity selected from the random library. Thus, each peptide is assumed to contain one or more epitope-relevant AAPs that are the basis for mAb recognition of that peptide. To identify the most meaningful AAPs present in the panel of peptides, the peptides are deconvoluted into pairs. For example, a peptide of the sequence ABCDE ... would be written as the series of pairs: AB BC CD DE, etc. All AAPs derived from the panel of peptides are then pooled, and the frequency of each type is calculated and it is determined whether its representation in the pool is higher than the random expectation. Thus, a second parameter of the algorithm (the first being D) is the number of SD above randomness for a given pair, which is defined as the statistical threshold (ST). Once the most significant AAPs of the pool are identified, the algorithm seeks the pairs for a selected D value on the surface of the antigen and attempts to link them into clusters. In Mapitope, only one member of a pair need be exposed; thus, a third parameter (E) is the surface accessibility threshold.
Depending on the size and quality of the peptide panel, one can miss specific residues of the epitope contained within segments of the predicted clusters. Consider, for example, a prediction that contains residues #200 and #208 as well as all the residues between them except for #203 and #204. In such a case, two short peptides, 200–202 and 205–208, would be expected to be separated by two unpredicted residues, 203 and 204. Mapitope would include the entire segment 200–208 (9 amino acids) in its prediction, "filling in the gap" that stems from the missed residues. Therefore, a fourth parameter,—I, defines the maximum gap (i.e., the number of unpredicted amino acids) between two residues to be connected.
As contacts between the mAb and antigen are often through functional moieties of the R groups, conserved residues were consolidated into 13 functional subgroups of amino acids and given single-letter notations:
B=R,K; J=E,D; O=S,T; U=L,V,I; X=Q,N; Z=W,F; A=A; C=C; G=G; H=H; M=M; P=P; Y=Y.
In summary, a mAb is used to screen a random peptide library to generate a panel of peptides recognized by the mAb. These peptides are deconvoluted into AAPs and the most statistically significant pairs (SSPs) are identified. These are then mapped in the crystalline structure of the antigen, and the most elaborate and diverse clusters on the surface of the antigen are identified. These are regarded as the predicted epitope candidates.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
, 17)
.
Affinity-selected b12 mimotopes
At the onset of our investigation, two reports had already been published describing peptides that had been isolated from phage display libraries that were specific for b12 (22
, 23)
. It has, however, been our experience in using the Mapitope algorithm to prefer disulfide constrained looped peptides, as these tend to more efficiently display the R groups of tandem residues, making them more accessible for simultaneous binding by the Ab being screened. Therefore, b12 was used to repeatedly screen phage display random peptide libraries produced at Tel-Aviv University (TAU) based on the fth-1 expression vector in which random peptides are recombinantly fused to the NH2 terminus of a recombinant protein VIII of the fd filamentous bacteriophage (20)
. The libraries used displayed 12mer random peptides flanked by constant cysteine residues, and a total of 19 cys-looped peptides was ultimately isolated. As illustrated in Fig. 1
the TAU peptides (Fig. 1C
) had little similarity to the panel of 32 linear peptides published by L. J. Boots et al. (23)
(Fig. 1B
). The two peptides isolated by L. L. Bonnycastle and published by M. B. Zwick (22
, 24)
(Fig. 1A
) did resemble the TAU peptides, as they all contained a short motif in which the amino acid pair SD was consistently preceded by an aromatic residue (usually W), preceded by an aliphatic methyl (L, V, I, or A). Furthermore, the motif "methyl, aromatic, SD" was often followed by E/D or an aliphatic methyl. Nowhere within the linear sequence of gp120 is there such a homologous motif; however, M. B. Zwick et al. (24)
had proposed that this motif mimic a part of the D-loop of gp120 (residues R273-I285, RSVNFTDNAKTII).
|
Validation of TAU peptides
Since different binding assays are used in different studies, it was important to confirm that the TAU peptides bound b12 at least as well as the peptides reported by others. As is shown in Fig. 2
a standard dot blot assay was performed comparing seven representatives of the TAU peptides with the ED1 peptide published by Zwick and the fth-1 vector as a negative control. An ED1 peptide was constructed using two complementary oligonucleotides that corresponded to the ED1 amino acid sequence that were cloned into the Sfi-I cloning sites of the fth-1 vector (see Materials and Methods). As expected, ED1 bound the b12 mAb very well. Of the TAU peptides, most had binding properties similar to those found for ED1. The E51 peptide is an example of a poor binder whereas others (e.g., F88 and A11) bound slightly better than ED1.
|
Role of SSPs in binding of b12 to the TAU peptides
Before conducting a Mapitope analysis on the TAU peptide panel, we wanted to convince ourselves that the fundamental concept of the algorithm [i.e., the role of statistically significant amino acid pairs (SSPs)] in mAb recognition (see Materials and Methods and E. M. Bublil et al., unpublished results)] holds for the peptides isolated in this case. Therefore, a sample peptide was sought that contained two serine residues: one that is a member of SSPs and the other that is not. Such a peptide is C10 (CLWSDLLSQYTKPC). Note that serine 4 is a member of the pairs WS and SD; both are SSPs (ST values of 7.13 and 7.00, respectively). Serine 8, however, participates in the LS and SQ pairs; neither are enriched in the peptide panel (ST values of –1.24 and 0.00, respectively). Therefore, two mutant peptides were constructed in which serines 4 and 8 were changed to alanine, respectively (Fig. 3
A). As illustrated in Fig. 3C
, the switch to alanine for serine 8 had no effect on binding at all. On the other hand, the single mutation of serine 4 knocked out b12 binding completely (anti-M13 mAb was used to confirm that equal amounts of phages were loaded on the filter, Fig. 3B
). This illustrates the differential effect of the serine mutations and the importance of an SSP for mAb recognition.
|
Mapitope prediction of the b12 mAb epitope
Next we conducted a Mapitope analysis of the TAU peptides, using as initial parameters: ST = 3; D = 9; E = 5; I = 0. Four clusters were identified and designated A, B, C, and D. Clusters A and B fall in the periphery of the CD4 binding site on the neutralizing surface of gp120. Clusters C and D occupy areas of gp120 that, in part, would be inaccessible to mAb binding as it would be occluded in the trimeric quaternary configuration of the HIV-1 envelope. Moreover, Cluster C contained residues that are part of the C1 helix of gp120 (residues H105-W112). Numerous antibodies have been reported to bind this helix (such as GV1A8; see ref. 25
) that do not interfere with b12 binding at all (data not shown). Therefore, we proceeded in our analysis to specifically focus on Clusters A and B (see
Fig. 6B
).
|
|
|
Characterization of Clusters A and B
To try and discriminate between the two clusters, we conducted analyses at ever increasing ST values. Using SSPs of higher ST values tends to eliminate irrelevant clusters. Analysis of the TAU peptides using a range of ST values (3 to 7), however, did not discriminate between Cluster A and Cluster B. Both continued to persist at ST = 7 and both were lost above this value. The next step in defining the predicted clusters is to conduct analysis as a function of D (Fig. 4
). As can be seen for both Cluster A and Cluster B, there is a point in which the analysis breaks down with a sharp rise in the number of amino acid residues predicted to be associated with the cluster. This sharp rise is the Q-point and predictions should be conducted at the Q-point (D=9 and D=8.5 for Cluster A and Cluster B, respectively). We tend not to include in the final prediction outlier strands of the antigen, which are predicted only at Q-point without prior indication at lower D values. The predicted residues for each cluster are given in Table 1
.
|
The amino acids that are recognized by the Ab in binding the antigen obviously should be surface accessible. However, this condition may not be required for both residues of a given SSP, the logic being that there could possibly be a situation in which one residue would function to position the other of a pair rather than to directly contact the Ab. In such an instance, the contact residue alone needs to be surface accessible, and the other residue having a more structural capacity could actually be just below the surface. Mapitope analysis allows for such unique situations by conducting the analysis without restricting surface accessibility ("–N" analysis). For such –N analyses, we use a more restrictive D value (D=7) in order to focus on only those residues most closely associated with the antigen’s surface. As shown in Table 1
, relaxing the analysis, requiring only one residue to be surface accessible, introduces 4 residues to Cluster A and 6 residues to Cluster B.
Finally, Mapitope defines predicted strands of each cluster in which the first and last residue of a given strand must be predicted. However, short gaps of unpredicted residues (up to 3 contiguous residues) are permissible and can be included in the prediction. Thus, Cluster A comprises three strands: T358-F361; T455-F468; F391-F396, and Cluster B four strands: V254-T257; D368-F376; E381-Y384; I420-I424.
In view of the fact that the two published sets of b12-specific peptides exist, we conducted Mapitope analyses on them as well. The Boots peptides are numerous but linear, and there are only two Bonnycastle peptides, which is statistically too few. Nonetheless, as is shown in Table 2
, the predictions using these alternative peptide panels are remarkably similar to those obtained with the TAU peptides. Again, the two major clusters for each peptide panel coincide well with Clusters A and B described above. Here and there residues are added or missed for each prediction, but for the most part all three data sets of peptides produce similar results. This demonstrates that the common means for selecting the peptides (i.e., b12 binding) in three different laboratories using different random peptide libraries is enough to drive the Mapitope algorithm to the same prediction.
|
Mutations in gp120
Predictive algorithms cannot be expected to provide one single correct solution, but rather are intended to reduce enormously complex problems into experimentally manageable ones. This is the case for the Mapitope prediction of the b12 epitope. We are confronted with two epitope candidates that appear to be of equal credibility. A standard method of validation is site-directed mutagenesis of residues proposed to be critical for epitope integrity and Ab binding. Extensive mutagenesis of HIV-1 gp120 has been done and reported over the last decade, and much has been accomplished with the specific goal of identifying key residues relevant to b12 binding. Indeed, of the 18 residues predicted for Cluster B based on the TAU peptides, 8 specific amino acids have been site-mutated by other investigators (18
, 26
, 27)
, and all have been shown to dramatically reduce b12 binding (at least by 70%) to the mutated gp120s. Six of these predicted residues (S256, D368, E370, V372, T373, Y384) as well as residue N386 (predicted based on the Bonnycastle peptides), when mutated to alanine, almost knocked out b12 binding completely (
85%). In the case of Cluster A, the picture is less clear. Based on the TAU peptides, 17 residues are predicted, of which 6 have previously been mutated by others (18
, 26
, 27)
. We extended the mutational analysis of Cluster A by repeating five of the mutations as reference points and adding an additional seven novel mutations (F361A, F391A, N392A, S393A, T394A, E466A, and F468A). The results are shown in Fig. 5
. The strongest effect of all but three mutations (F361A, F391A, and F468A) ranged between 40% to as much as 65% inhibition of b12 binding (e.g., I467A and D457A, respectively). The most striking effect, however, was for the three phenylalanine residues mentioned above, which knocked out b12 binding almost completely. However, as illustrated, these mutations had similar deleterious effects on CD4 binding or binding of the antigp120 mAb m14, which has been shown to compete with b12 (18
, 26
, 28
, 57)
. It would thus appear that these residues, which are not surface accessible (Table 1)
, have a more structural role in maintaining gp120 conformation; by converting these amino acids to alanine, one disrupts the native configuration of the protein. Therefore, we produced three more mutations converting the phenylalanines into tyrosines to test whether maintaining an aromatic residue at these positions would reconstitute gp120 native structure, as monitored by CD4 binding or m14, and still have an effect on b12 binding. Whereas the degree of recovery was somewhat variable for b12 binding compared with that of CD4 or m14, all three reagents regained the capacity to bind the mutated gp120s concomitantly. Thus, it would appear that mutagenesis of residues in both clusters affected b12 binding to variable degrees, yet Cluster B seemed more sensitive to point mutations than Cluster A.
Clusters A and B compared with other published predictions and to the CD4 binding site
Figure 6
describes the two clusters (Fig. 6B
) and compares them with respect to the CD4 binding site on gp120 (Fig. 6C
) and a prediction by R. Pantophlet et al. (18)
for the b12 epitope based on docking in silico of b12 onto the gp120 structure (Fig. 6A
). As illustrated, both predicted clusters reside on the periphery of the CD4 binding site and slightly overlap with it (Fig. 6E
, indicated in cyan). The same can be said for the Pantophlet prediction, which encompasses the CD4 binding site, yet extends over a larger surface (Fig. 6D
). For comparison, the residues of TAU-predicted Clusters A and B as well as the CD4 site and the Pantophlet prediction have been color highlighted in the linear sequence of gp120 (Fig. 6F
). The strand compositions of clusters A and B are also given (Fig. 6G, H
).
Mapitope prediction of the m14 and b6 epitopes
At this point we decided to map the epitopes of two additional mAbs (m14 and b6) known to compete with b12 against gp120 (18
, 26
, 28
, 57)
, the rationale being that the epitopes of all three competing antibodies (b12, b6, and m14) should map to the same region of gp120. Thus, we screened both m14 and b6 against our random peptide libraries and isolated peptide panels for each (Fig. 7
A). We then conducted Mapitope analyses for m14 and b6 mAbs, which generated single best clusters for each mAb (Fig. 7B
). The predicted epitopes of mAbs m14 and b6 mapped very close to one another and with some overlap with the predicted b12 Cluster B. Moreover, these predictions focused our attention on their proximity to yet a fourth epitope—namely, that of mAb 17b, for which the epitope has been definitively determined by cocrystallization (14)
.
|
Competition between b12, m14, and b6 with 17b for gp120 binding
If indeed Cluster B of the b12 predictions is correct, one should be able to demonstrate some competition between b12 and 17b for gp120 binding. Furthermore, m14 and b6 should also be able to compete against 17b, with m14 being anticipated to be the most effective inhibitor and b12 the least.
To test this hypothesis, we conducted competition experiments using ever increasing concentrations of m14, b6, or b12 to inhibit the binding of biotinlylated 17b mAb to gp120. As depicted in Fig. 7C
, all three mAbs inhibited 17b binding, with m14 being the most efficient, b12 the least, and b6 with intermediate efficacy. Less than 1 µg/ml b12 inhibited the binding of 5 µg/ml of 17b by 50%, and this inhibition could be increased to 80% using 10 µg/ml b12 (J. P. Moore et al. previously reported b12 competition with mAb 17b; see ref. 29
).
Thus, we would conclude that Cluster B is most likely to coincide with the genuine epitope of mAb b12.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
, 31)
. We find ourselves debating what constitutes protective immunity against HIV—the pendulum of enthusiasm for neutralizing antibodies as opposed to cellular immunity, mucosal or other, continues to sway back and forth (32)
; in the meantime the number of individuals infected by HIV continues to rise (33)
.
It would appear that some governing concepts have gained consensual agreement. An effective vaccine will have to stimulate both arms of the immune system and provide a protective barrier at the mucosa (32
, 34
35
36
37
38)
. Neutralizing antibodies need to be broadly cross-reactive and able to knock out primary isolates spanning at least HIV-1 Clades A, B, and C (39
40
41)
.
In this study we have focused on what could ultimately be a component of an AIDS vaccine, the concept being that starting with a proven cross-reactive neutralizing mAb one should be able to identify its epitope and, based on it, design a corresponding immunogen that would stimulate protective immunity in the naive vaccinee (42)
. Therefore, one can propose vaccines based on selected defined epitopes. Epitope-based vaccines would avoid deleterious consequences of using intact pathogens or subunits that could lead to enhanced infections or autoimmunity (42)
. Recent concerns surrounding the vaccine against Borrelia burgdorferi (the agent of Lyme disease) is a case in point (43
44
45
46)
. The outer surface protein A (OspA) of the bacterium apparently contains an epitope that has homology with the human leukocyte function-associated antigen-1 (LFA-1). As a result, the vaccine has been purported to lead to an autoimmune reaction to the autoantigen ultimately causing arthritis. Modifying native OspA by deleting the homologous epitope has generated a novel vaccine that should be safer (47)
.
The use of isolated epitopes has been demonstrated as potentially effective in producing protective immunity. One of the first illustrations of this is the study of the 447–52D anti-V3 loop human mAb that shows broad cross-neutralization (48
49
50)
. This mAb was used to screen phage display peptide libraries, and an isolated mimotope 15mer peptide was used to immunize rabbits. The polyclonal sera cross reacted and neutralized HIV-1 isolates SF2 and AL-1 despite the fact that there is little linear sequence homology between the V3 loops of these isolates and the mimotope (51)
. The phage display approach for mimotope isolation and epitope based vaccination was also shown to be effective for measles virus (MV). M. W. Steward et al. (52)
illustrate that an 8mer mimetic of the F protein of MV can elicit protective immunity to otherwise fatal encephalitis induced by either MV or the related canine distemper virus in mice. The recent case of an anti-CCR5 mimetic derived by phage analysis by Golding and colleagues (53)
show yet an additional advantage of this approach. The murine mAb 2D7 is extremely effective in blocking R5 HIV infection (54)
. However, so far an effective humanized version of the mAb has not been attainable. The identification of mimotopes corresponding to the conformational epitope of this mAb has led to the production of an epitope-based immunogen that, when used to vaccinate rabbits, produces antibodies that neutralize infection of human peripheral blood mononuclear cells with R5 HIV isolates (53)
.
Thus, the task of elucidating the conformational epitope of the b12 mAb has been undertaken with the idea that its epitope should be effective for generating a component of an AIDS vaccine. We are not the first to have been engaged in identifying the b12 epitope. An initial attempt to map the epitope of the b12 mAb was performed by M. B. Zwick et al. (24)
. By random mutagenesis at fixed positions using sublibraries of the two peptides obtained by Bonnycastle (22)
, the B2.1 phage was isolated, which binds b12 with high affinity. The B2.1 sequence (HERSYMFSDLDNRCI) was then correlated to the D-loop of gp120 (amino acids 273–285, RSVNFTDNAKTII), although the homology between the loop and the mimetic is not striking and use of the B2.1 peptide as an immunogen has not produced the desired neutralizing response (24)
. In 2001, the crystal structure of the b12 mAb was resolved by E. O. Saphire et al. (27)
. The availability of the crystals of the antigen and the Ab enabled the group to conduct in silico docking of the two. One hundred computational docking experiments were performed in parallel using "AutoDock" (55)
. They conclude that the b12 epitope can fit onto gp120 by binding an epitope extending from the stem of the V1/V2 loops across the neutralizing face and penetrating the CD4 binding site (see also Fig. 6A
). The docking model was further analyzed by R. Pantophlet et al. (18)
using saturating mutagenesis of amino acids selected primarily from a list of likely contact residues based on the docking model. The combined docking/mutagenesis prediction encompasses a broad, rather extended surface of the antigen. Moreover, 10 of the mutations performed in the predicted epitope (K97A, L125A, S199A, N256A, S365A, V430A, T455A, G458A, G473A, and M475A) do not inhibit b12 binding (18)
. In fact, the majority enhance b12 binding by at least 2-fold compared to WT. Mutation G473A is an extreme example that enhances b12 binding by 10-fold compared to WT.
Here we approach the mapping of the b12 epitope using a novel method that combines combinatorial phage display peptide analyses with a computational algorithm. The Mapitope algorithm was first developed using two model cocrystals: mAb 17b:HIV gp120 and mAb 13B5:HIV p24 (12)
. Since the original article, Mapitope predictions have been validated for two additional cocrystals: Herceptin:Her-2/neu and mAb Bo2C11:Factor VIII (E. M. Bublil et al., unpublished results). Moreover, we recently applied Mapitope to the prediction of the 80R neutralizing human Ab against the receptor binding domain of the SARS coronavirus spike protein (56)
. This prediction has since been confirmed by cocrystallization (W. A. Marasco, personal communication). Thus, there is ample evidence that Mapitope can be an effective means of mapping conformational discontinuous epitopes.
The particular case for the b12 mAb has combined Mapitope with multiple analyses and empirical experiments. Mapitope predicts two strong epitope candidates on the gp120 surface based on the TAU peptides described in this report as well as the published peptides of Bonnycastle and Boots. A major concern in these predictions is that there is only a limited amount of data regarding gp120 structure. The crystal structure used in this study is of truncated core gp120 (gp120 missing both N and C termini, V1, V2, and V3 loops and glycomoieties). Moreover, the structure is that of CD4 complexed gp120 (14)
. In view of the fact that CD4 competes for b12 binding (26)
, it would be preferable to have conducted the analyses using free and intact gp120. The lack of a more extensive understanding of intact gp120, monomeric or trimeric, is definitely a handicap in the attempt to map the correct b12 epitope (for us and for all the other studies dealing with this problem), and so our prediction can be no better than the gp120 structures available.
We have therefore complemented the Mapitope analysis of b12 with additional experiments, namely, comparisons with the predictions of two additional competing mAbs, m14 and b6 (26
, 57)
, conducting additional mutational analyses and ultimately competitive ELISA tests to evaluate our conclusions regarding Cluster A vs. Cluster B.
The fact that the b12 mAb neutralizes a broad range of HIV isolates from different clades is quite remarkable. Therefore, one would expect the b12 epitope to include elements that are conserved among the varied repertoire of HIV-1 isolates. Even further support of Cluster B can be found in ConSurf analysis of both predicted clusters. The ConSurf algorithm (http://consurf.tau.ac.il) can be used to determine the level of conservation of each amino acid comprising the predicted clusters. ConSurf uses evolutionary data in the form of multiple sequence alignment for a protein family and deduces the importance of residues from their level of conservation in families of homologous proteins (58)
. Each amino acid is assigned a number and a color according to a scale representing the level of conservation. Thus, the less conserved amino acids acquire the value of 1 (blue) and the most conserved are assigned a value of 9 (red).
Figure 8
A shows the level of conservation of Cluster A as calculated using the ConSurf algorithm. Of the 17 amino acids comprising the cluster (not including interconnecting amino acids), only one amino acid is ranked 9 (I359), and 8 amino acids are assigned with the level of conservation of 5 or less. In contrast to Cluster A, Cluster B seems much more conserved (Fig. 8B
). Of the 18 amino acids comprising Cluster B, nine are assigned the value of 9, the highest possible score for conservation. An additional six amino acids are also highly conserved (score 7 or 8). Only one amino acid in Cluster B has a conservation level of lower than 6 (T373).
|
Thus there are multiple lines of evidence presented in this study that support the prediction that Cluster B coincides with the b12 epitope:
1. Cluster B ranks as one of the two strongest clusters in the Mapitope analyses using the TAU peptide panel.
2. This prediction is independently supported by two additional peptide panels derived from different libraries from two different laboratories.
3. Cluster B maps close to the CD4 binding site.
4. Cluster B appears to be the most sensitive to mutations of key residues.
5. The predicted Cluster B overlaps with the predictions of two competing antibodies m14 and b6.
6. All three antibodies, m14, b6, and b12 compete against mAb 17b binding to gp120.
7. Cluster B is remarkably conserved as compared to Cluster A.
We therefore propose that Cluster B or any of its four component peptide segments be used in the production of epitope based immunogens in an AIDS vaccine.
|
ACKNOWLEDGMENTS |
|---|
Received for publication March 9, 2006. Accepted for publication April 10, 2006.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
