Comprehensive Analysis of Human Immunodeficiency Virus Type 1-Specific CD4 Responses Reveals Marked Immunodominance of gag and nef and the Presence of Broadly Recognized Peptides

Abstract

Increasing evidence suggests that human immunodeficiency virus type 1 (HIV-1)-specific CD4 T-cell responses contribute to effective immune control of HIV-1 infection. However, the breadths and specificities of these responses have not been defined. We screened fresh CD8-depleted peripheral blood mononuclear cells (PBMC) from 36 subjects at different stages of HIV-1 infection for virus-specific CD4 responses by gamma interferon enzyme-linked immunospot assay, using 410 overlapping peptides spanning all HIV-1 proteins (based on the clade B consensus sequence). HIV-1-specific CD4 responses were identified in 30 of the 36 individuals studied, with the strongest and broadest responses detected in persons treated in acute infection who underwent treatment interruption. In individuals with identified responses, the total number of recognized HIV-1 peptides ranged from 1 to 36 (median, 7) and the total magnitude of responses ranged from 80 to >14,600 (median, 990) spot-forming cells/10⁶ CD8-depleted PBMC. Neither the total magnitude nor the number of responses correlated with viremia. The most frequent and robust responses were directed against epitopes within the Gag and Nef proteins. Peptides targeted by ≥25% of individuals were then tested for binding to a panel of common HLA-DR molecules. All bound broadly to at least four of the eight alleles tested, and two bound to all of the HLA-DR molecules studied. Fine mapping and HLA restriction of the responses against four of these peptides showed a combination of clustering of epitopes and promiscuous presentation of the same epitopes by different HLA class II alleles. These findings have implications for the design of immunotherapeutic strategies and for testing candidate HIV vaccines.

Increasing evidence indicates that cellular immune responses play a major role in containing human immunodeficiency virus type 1 (HIV-1) replication (41). In the acute phase of HIV infection, the development of strong virus-specific CD8 T-cell responses correlates with the initial decline of viremia after its peak (3, 11, 33, 69) and precedes the production of any neutralizing antibody. In a monkey model of simian immunodeficiency virus infection, early control of the virus fails if CD8 T cells are experimentally depleted prior to infection (56). If CD8 T cells are depleted during the chronic phase of infection, the virus level rises until CD8 T cells are restored (30). Additional arguments for a determinant role of cytotoxic-T-lymphocyte (CTL) responses are the association of slow progression to AIDS with defined HLA class I alleles (43), the association of loss of viremia control with viral immune escape from dominant CTL responses in the simian immunodeficiency virus model (4), and the report of a sudden increase of viremia after HIV-1 superinfection associated with declining CTL responses directed against the first virus (2).

However, the generation of strong HIV-specific CTLs in infected individuals (64) fails to completely control the infection in the majority of persons. One defect in the immune response against HIV-1 is an impairment of CD4 T-cell proliferation to HIV antigens (7, 34, 37, 42, 49, 54, 55, 63, 67). Animal models indicate that virus-specific CD4 T cells are critical for the maintenance of effective immune control, and this is in part due to effects on CTL function (5, 29, 31, 40, 57, 60, 62, 70). CD4 help is also necessary for efficient neutralizing-antibody-mediated viral clearance (52). CD4 T-cell responses are also important in chronic human viral infections. For example, loss of virus-specific CD4 cells in human hepatitis C virus (HCV) infection leads to persistent viremia and progressive liver dysfunction, and broadly directed CD4 responses are associated with resolution of viremia (16). Recently, the importance of CD4 T cells for effective immune control of a virus causing chronic disease in humans has been shown for HCV infection of chimpanzees (24). Depletion of CD4 T cells before reinfection of two immune apes resulted in persistent, low-level viremia despite functional intrahepatic memory CD8⁺-T-cell responses. Incomplete control of HCV replication by memory CD8⁺ T cells in the absence of adequate CD4 T-cell help was associated with the emergence of viral escape mutations in class I major histocompatibility complex (MHC)-restricted epitopes and failure to resolve HCV infection.

In HIV infection, CD4 T helper cell responses may also be critical for long-term antiviral control. HIV-specific CD4 T-cell responses are thought to be impaired early in the course of the disease, and preferential infection of these cells by HIV contributes to their depletion (18). Numerous studies of HIV-1-specific CD4 T-cell responses have demonstrated the ability of peripheral blood mononuclear cells (PBMC) to proliferate in response to stimulation with viral antigens (7, 34, 37, 42, 49, 54, 55, 63, 67). HIV-1-specific lymphocyte proliferative responses are usually weak or absent in untreated HIV-1-infected individuals but are more frequent in long-term nonprogressors (LTNP) than in subjects with progressive disease (34, 37, 55, 63, 67). These responses may be enhanced in subjects who receive prolonged highly active antiretroviral therapy (HAART) with suppressed viral replication (37, 42), especially when they are treated during acute HIV infection (54). In addition, these responses have also been detected in individuals receiving therapeutic immunization (53). Effector gamma interferon (IFN-γ)-secreting HIV-1-specific CD4 T cells can be detected in freshly isolated PBMC by intracellular cytokine staining (45, 51, 67) or by IFN-γ enzyme-linked immunospot (ELISPOT) assays (48), even though lymphocyte proliferative responses may be absent (6, 42, 49, 67), and HIV-1-specific CD4 T cells are detectable in most HIV-1-infected individuals with normal or moderately decreased CD4 counts (6). However, the frequency of HIV-specific IFN-γ-secreting cells is much lower than that of cytomegalovirus (CMV)-specific CD4 T cells in the same individuals (51) or in HIV-seronegative subjects (9). Although a comprehensive analysis of CD4 T-cell responses directed against the whole expressed HIV-1 genome has been done by flow cytometry with peptide pools (6), the breadth and specificity of the response at the single-peptide level has not been determined, and few regions containing virus-specific CD4 T-cell epitopes have been defined (10, 19, 39).

This study was designed to determine the breadths and specificities of HIV-1-specific CD4 T-cell responses detected by IFN-γ ELISPOT and to identify immunodominant regions of the clade B protein sequences across the entire expressed HIV-1 genome. HIV-1-specific CD4 T-cell responses were then compared among subjects categorized by virus load and treatment status. The peptides targeted by a large proportion of the individuals studied were assessed for binding to various common HLA-DR alleles, and for a subset of these responses, the precise epitopes targeted were defined and the HLA class II-presenting allele was determined. Finally, HIV-1-specific CD4 T-cell responses were contrasted with CD8 T-cell responses in a subset of individuals.

MATERIALS AND METHODS

Study subjects.

Thirty-six HIV-1-infected individuals at different stages of infection were recruited from the Massachusetts General Hospital (MGH) and the Lemuel Shattuck Hospital in Boston. Documentation of HIV-1 infection at least 1 year prior to analysis was required. Eight HIV-1-negative subjects were also included as controls. All participants gave informed consent prior to entry into the study. Patient characteristics are summarized in Table 1. The subjects were stable with regard to antiviral treatment, having been either on or off therapy for at least 3 months at the time of analysis. Most of the subjects in this study were at an early stage of HIV-1 infection (median CD4 T-cell count, 732; range, 28 to 1,801 cells/mm³). The virus load measured by reverse transcription-PCR (Amplicor; Roche Molecular Systems, Pleasanton, Calif.) was below the limit of detection (50 RNA copies/ml) in all treated subjects; the median virus load was 2,913 RNA copies/ml in untreated persons (range, <50 to 257,000 RNA copies/ml).

TABLE 1.

Subject characteristics at time of analysis

Subjecta	Age (yr)	Duration of HIV infection (yr)b	Current therapyc	Virus load (copies/ml)	CD4 count (cells/μl)	Total HIV-specific CD4 response (SFC/106 cells)d	No. of recognized peptidese
Acute-Rx
AC20	4	4	HAART	<50	1,223	195	1
AC21	40	4	HAART	<50	489	1,105	7
AC32	38	3	HAART	<50	571	310	2
AC34	34	3	HAART	<50	443	205	0
AC63	38	2	HAART	<50	732	140	0
AC82	44	1	HAART	<50	1,082	270	3
AC87	39	4	HAART	<50	917	815	11
Acute-STI
AC02	48	5	STI-no HAART	5,790	714	560	3
AC03	36	5	STI-no HAART	2,850	805	1,610	7
AC05	44	5	STI-no HAART	12,900	361	890	24
AC06	42	5	STI-no HAART	23,000	427	2,750	18
AC10	36	5	STI-no HAART	267	755	2,030	19
AC14	50	4	STI-no HAART	2,300	503	1,875	18
AC15	46	5	STI-no HAART	17,800	535	3,120	29
AC26	49	4	STI-no HAART	82,600	660	1,480	12
AC46	52	2	STI-no HAART	2,975	888	>14,610	36
Chronic-Rx
CRT1	33	1	HAART	<50	172	80	1
CRT2	36	13	HAART	<50	579	120	0
CRT3	40	1	HAART	<50	609	580	2
CRT4	52	3	HAART	<50	146	60	0
CRT5	34	3	HAART	<50	1,163	180	1
Chronic-no Rx
CRU1	51	1	None	16,300	899	1,360	2
CRU2	55	1	None	32,800	1,050	1,540	9
CRU3	29	1	None	52,700	371	295	4
CRU4	34	2	None	23,500	1,051	230	3
CRU5	38	6	None	257,000	28	0	0
CRU6	38	2	None	10,791	239	1,365	8
Controller-LTNP
CO1	34	3	None	212	1,082	1,320	9
CO2	44	4	None	1,620	798	130	0
CO5	37	6	None	<50	732	250	3
CO6	50	7	None	1,140	672	745	3
CO7	54	4	None	390	742	230	6
LT4	50	10	None	<50	832	4,540	34
LT6	37	15	None	754	832	1,175	22
LT9	40	14	None	<50	1,801	640	7
LT14	48	7	None	<50	1,340	880	6

^aSubjects AC20, CRT3, CRU5, CRU6, and LT14 are females. All other individuals are males.

^bIndicates years of documented HIV-1 infection.

^cHAART defined as regimen containing at least three drugs.

^dSum of responses to the positive peptide pools in IFN-γ ELISPOT in SFC per 10⁶ CD8-depleted PBMC. As each peptide is in two pools, only the responses to the pools with contiguous overlapping peptides were used for the calculation of the total response.

^eNumber of confirmed responses to individual peptides in IFN-γ ELISPOT.

The study included 36 subjects in different disease stages. Individuals were divided into five groups according to their immunological and virological status and treatment history: individuals treated during acute HIV-1 infection and on continuous therapy since diagnosis (acute-Rx; seven subjects), subjects treated during acute HIV-1 infection who subsequently discontinued treatment in the context of a supervised treatment interruption (STI) trial (acute-STI; nine subjects), individuals treated during chronic HIV-1 infection (chronic-Rx; five subjects), untreated persons with chronic HIV-1 infection (chronic-no Rx; six subjects), and individuals with low viremia (<2,000) and high CD4 counts (>500 cells/mm³) in the absence of antiviral treatment (controllers-LTNP; nine subjects).

HLA typing.

DNA for typing was extracted using the Puregene DNA isolation kit for blood (Gentra Systems, Minneapolis, Minn.) according to the manufacturer's instructions. HLA class II typing was performed at the MGH Tissue Typing Laboratory using sequence-specific primer PCR, as previously described (14). High-resolution HLA class II typing was performed by Dynal Biotech HLA Diagnostics (Bromborough, United Kingdom) using sequence-specific primer PCR (14).

Isolation of PBMC and CD8 depletion.

Fresh blood was processed within 4 h of phlebotomy. PBMC were isolated on a Ficoll-Hypaque density gradient (Sigma, St. Louis, Mo.), and CD8 depletion (for studies of CD4 T-cell responses) or CD4 depletion (for studies of CD8 T-cell responses) was performed at the time of PBMC isolation with the RosetteSep depletion reagents (StemCell Technologies, Vancouver, Canada). Flow cytometry was performed as described previously (28) to ensure the purity of cell populations after CD8 T-cell depletion. Contamination by CD8 T cells was always <1% in 34 assessed samples (median, 0.02% residual CD8 T cells; range, 0.005 to 0.7%). In a subgroup of four subjects, CD4 depletion was performed for assessment of CD8 T-cell responses. These subjects were selected based solely on the availability of samples to perform these assays.

Synthetic HIV-1 peptides.

A set of 410 peptides (15- to 20-mers overlapping by 10 amino acids) spanning the entire expressed HIV-1 genome was synthesized at the MGH Peptide Core Facility on an automated peptide synthesizer (Advanced Chemtech, Louisville, Ky.). These peptides were based on the HIV clade B consensus sequence 2001, available from the Los Alamos National Laboratory HIV immunology database (http://hiv-web.lanl.gov/content/hiv-db/CONSENSUS/M_GROUP/Consensus.html).

ELISPOT assays.

Screening for CD4 T-cell responses was performed by IFN-γ ELISPOT using a strategy based on a matrix of peptide pools as previously described (1). The 410 peptides spanning HIV-1 were divided into 72 pools of 10 to 12 peptides each. Each peptide was included in two different pools. The pattern of responses to the pools indicated specific candidate peptides, which were then individually retested. For analysis of CD4 T-cell responses, fresh CD8-depleted PBMC were plated in 96-well polyvinylidene plates (Millipore, Bedford, Mass.) that had been precoated with the anti-IFN-γ antibody M700A (Endogen, Woburn, Mass.) at a concentration of 0.5 μg/ml in phosphate-buffered saline. Cells were added at a concentration of 60,000 to 100,000 per well in 200 μl of RPMI medium (Sigma) supplemented with 10% human AB serum, 10 mM HEPES buffer, 2 mM l-glutamine, and 50 U of penicillin-streptomycin (all from Mediatech, Herndon, Va.). The final concentration of each peptide was 10 μg/ml. Negative control wells (cells and medium only) were run in quadruplicate on each plate, along with positive control wells (phytohemagglutinin at 1 μg/ml) to ensure that the cells were responsive. The plate contents were incubated for 20 h at 37°C and 5% CO₂ and then washed six times with phosphate-buffered saline, after which they were incubated at room temperature for 90 min with the corresponding biotinylated secondary antibody, M701B (Endogen), followed by washes and a 45-min incubation with a streptavidin-alkaline phosphatase conjugate (Bio-Rad, Hercules, Calif.). After the plates were washed, 5-bromo-4-chloro-3-indolylphosphate and nitroblue tetrazolium substrates (Bio-Rad) were added. The matrix was analyzed, and candidate peptides were tested individually either with cells kept in an incubator overnight (30 of the 36 study subjects) or with freshly isolated PBMC from a second blood draw performed 1 week to 1 month (median, 2 weeks) after the first visit (6 of the 36 subjects). The number of spots per well was determined using an automated ELISPOT plate reader (AID EliSpot reader system; Autoimmun Diagnostika GmbH, Strasburg, Germany). The results were expressed as spot-forming cells (SFC) per million CD8-depleted PBMC after the subtraction of the background. Analysis of samples was performed only if the background was <30 SFC per million input cells. A response was considered positive if >50 SFC/10⁶ CD8-depleted PBMC were detected and they were at least 3 standard deviations above background.

In order to compare the breadths of CD4 responses to those observed in CD8 cells, a comprehensive analysis of CD8 responses was performed in a subgroup of four individuals with a similar strategy using CD4-depleted PBMC.

CD4 T-cell lines.

CD8-depleted PBMC were stimulated with the peptide of interest at 10 μg/ml in R10 HAB medium supplemented with recombinant interleukin-2 (IL-2) (50 U/ml) at a concentration of 2 million cells/ml on 24-well plates. Indinavir (Merck) was added at a concentration of 0.4 μM, and zidovudine (GlaxoSmithKline) was added at a concentration of 0.5 μM. The plate contents were incubated at 37°C in 5% CO₂ and fed two or three times weekly with medium exchanges. After 10 to 12 days, the cells were assessed for specificity by intracellular cytokine staining (ICS) for IFN-γ. If the specificity was ≥5%, the cells were used directly in fine-mapping and restriction experiments. If the specificity was <5%, peptide-specific cells were selected by an IFN-γ capture assay (Miltenyi Biotech, Auburn, Calif.), as previously described (13). Briefly, the cells were stimulated with 10 μg of peptide/ml and costimulatory CD28 and CD49d monoclonal antibodies (1 μg/ml; BD Biosciences, San Jose, Calif.). After incubation for 5 h, IFN-γ was processed according to the manufacturer's protocol and isolated on a magnetic cell sorting column. Selected cells were expanded with IL-2 and irradiated allogeneic feeder cells for an additional 10 to 14 days and then used in fine-mapping and restriction experiments, as described above.

ICS for IFN-γ and flow cytometry.

ICS for IFN-γ was performed as previously described (51). For fine mapping of epitopes, 3 × 10⁵ to 5 × 10⁵ expanded PBMC were incubated with autologous B cells at a ratio of 5 to 1 and incubated with 2 μg of serial truncations (one or two amino acids) of the corresponding peptide/ml with anti-CD28 and anti-CD49 antibodies (Becton Dickinson) at 37°C and 5% CO₂ for 1 h before the addition of brefeldin A (10 μg/ml; Becton Dickinson). The cells were incubated for an additional 5 h at 37°C and 5% CO₂ and then washed and stained with anti-CD4-phycoerythrin antibody and 7-AAD (Becton Dickinson) at 4°C for 20 min. The cells were fixed with solution A (Caltag), permeated with solution B (Caltag), and then stained with fluorescein isothiocyanate-conjugated anti-IFN-γ and allophycocyanin-conjugated CD69 (Becton Dickinson).

For analysis of HLA class II restriction, autologous and partially HLA-matched Epstein-Barr virus-transformed B-lymphoblastoid cell lines (BCL) and L cell lines (LCL) stably transfected with HLA class II molecules were pulsed with 1 μg of peptide/ml for 90 min at 37°C and 5% CO₂. HLA-DR-transfected cell lines were a kind gift of Epimmune (Foster City, Calif.). The peptide-pulsed BCL or LCL were then washed three times, incubated with the same numbers of peptide-specific cells, and stained as described above.

Six-parameter flow cytometric analysis was performed on a two-laser FACSCalibur instrument (Becton Dickinson). At least 5,000, and usually >10,000, events were collected in the live (7AAD_low) CD4⁺ lymphocyte gate. Data were analyzed using FlowJo software (Tree Star, San Carlos, Calif.).

HLA-DR binding studies.

Frequently targeted peptides were tested for binding in vitro to eight common HLA-DR molecules. DR molecules were purified, and binding assays were performed essentially as previously described (66). Purified human HLA-DR molecules were incubated with unlabeled HIV-1 peptides and 1 to 10 nM ¹²⁵I-radiolabeled probe peptides for 48 h. MHC binding of the radiolabeled peptide was determined by capturing MHC-peptide complexes on LB3.1 (anti-HLA-DRA) antibody-coated Lumitrac 600 plates (Greiner Bio-one, Frickenhausen, Germany) and measuring bound counts per minute using the TopCount (Packard Instrument Co., Meriden, Conn.) microscintillation counter. The binding data were analyzed, and 50% inhibitory concentrations (IC₅₀s) (nanomolar) were derived, as previously described (59, 66).

Statistical analysis.

Statistical analysis and graphical presentation were performed using GraphPad Prism version 3.0 software. Nonparametric statistical tests were used to avoid the assumption of normally distributed data sets, and the results are given as medians with ranges. Statistical analysis of the difference between two groups for a given variable was done with the two-tailed Mann-Whitney test. Values for P of <0.05 were considered significant.

RESULTS

Magnitude of HIV-specific CD4 T-cell responses shows dominance of Gag and Nef.

In order to assess the relative contributions of CD4 T-cell responses to different viral proteins, we first investigated the ability of peptide pools representing expressed gene products to induce IFN-γ production in CD8-depleted PBMC. There was a wide range in the magnitudes of responses observed, and responses to Gag and Nef pools were immunodominant (Fig. 1). Peptides contained within p17, p24, and p15 were frequently targeted, as were multiple regions of Nef. The only other peptide pool recognized by >25% of individuals was located in integrase. In persons with positive responses, the total magnitudes of responses against the entire expressed genome ranged from 80 to >14,600 SFC/10⁶ CD8-depleted cells (median, 990 SFC/10⁶ CD8-depleted cells) (Table 1). These data using peptide pools indicate that multiple regions of Gag and Nef are targeted by HIV-specific CD4 T cells and that other regions of the virus are only infrequently targeted, if at all.

FIG. 1.

Magnitudes of responses to peptide pools spanning the entire expressed HIV-1 genome. The horizontal axis represents the 36 peptide pools and the corresponding HIV proteins, and the vertical axis represents the magnitudes of responses expressed as the number of IFN-γ-producing cells/10⁶ depleted PBMC in IFN-γ ELISPOT.

The highest HIV-1-specific CD4 T-cell epitope density is in Gag and Nef, with limited epitopes identified within other proteins.

In order to characterize the responses at the single-peptide level, IFN-γ ELISPOT assays were performed using mixtures of 9 to 12 peptides in a matrix format in which each peptide was present in two wells. The pattern of positive wells allowed the prediction of individual positive peptides, which were then retested and confirmed at the single-peptide level. Overall, 30 (83%) subjects recognized at least one peptide, and in persons with detectable responses, there was a broad range in the number of responses to individual peptides targeted, from 1 to 36 (median, 7 per individual). The dominant targets in terms of breadth of responses were Gag and Nef (Fig. 2). Only 112 of the 410 peptides (27%) were recognized by CD4 T cells in this study, and of these 112 peptides, 52 were in Gag and 16 were in Nef. The percentages of peptides targeted within the individual proteins were highly variable (Table 2), with 88% of p15-Gag peptides recognized by at least one subject and no recognition of Vif and Vpu peptides. Recognition of >50% of overlapping peptides was observed only for the proteins Gag-p17, Gag-p24, Gag-p15, and Nef. These four proteins accounted for 82% (271 out of 329) of the total number of responses identified and for 61% of all positive peptides.

FIG. 2.

Individual peptide recognition by CD4 cells across the entire expressed HIV-1 genome. The horizontal axis represents the 410 overlapping peptides and the corresponding HIV proteins, and the vertical axis represents the percentages of study subjects with a response to each individual peptide.

TABLE 2.

Numbers and percentages of peptides recognized within individual HIV-1 proteins

Protein	No. of overlapping peptides recognized/total	% Of overlapping peptides recognized
p17 (Gag)	14/17	82
p24 (Gag)	23/32	72
p15 (Gag)	15/17	88
Protease (Pol)	2/20	10
Reverse transcriptase (Pol)	10/76	13
Integrase (Pol)	9/37	24
Vif	0/24	0
Vpr	5/11	45
Vpu	0/10	0
Tat	3/12	25
Rev	4/15	27
gp120 (Env)	6/67	9
gp41 (Env)	5/46	11
Nef	16/27	59

In order to confirm the specificity of the responses, contemporaneous assays were performed in a group of eight HIV-1-seronegative controls using CD8-depleted PBMC. At the time of testing, investigators were blinded to the HIV serostatus of four of the eight control subjects. No response against any of the 410 peptides could be confirmed in this cohort, indicating that the responses were virus specific.

Taken together, these data indicate that HIV-1 Gag and Nef are frequent targets of HIV-specific CD4 cells and that not only is the magnitude of responses to these proteins the strongest, they also contain the highest epitope density.

Eight peptides were targeted by at least 25% of individuals (Table 3). These were located in p17-Gag (ASRELERFAVNPGLL, ERFAVNPGLLETSEGCR, and SLYNTVATLYCVHQRIEV), p24-Gag (WIILGLNKIVRMYSPTSI and YVDRFYKTLRAEQASQEV), p15-Gag (RQANFLGKIWPSHKGR), and Nef (PEKEVLVWKFDSRLAFHH and KFDSRLAFHHMARELH). Ninety-three percent of the subjects who had HIV-specific CD4 T-cell responses recognized at least one peptide within this small subset of eight. The most frequently targeted peptide, p24-Gag (YVDRFYKTLRAEQASQEV), was recognized by 58% of the individuals studied.

TABLE 3.

Most frequently recognized peptides

% Of subjectsa	Protein	Amino acid positions	Sequence	% Of subjects with response
≥25	p17	37-51	ASRELERFAVNPGLL	36
	p17	42-58	ERFAVNPGLLETSEGCR	28
	p17	77-94	SLYNTVATLYCVHQRIEV	25
	p24	133-150	WIILGLNKIVRMYSPTSI	25
	p24	164-181	YVDRFYKTLRAEQASQEV	58
	p15	66-81	RQANFLGKIWPSHKGR	28
	Nef	176-193	PEKEVLVWKFDSRLAFHH	36
	Nef	184-199	KFDSRLAFHHMARELH	25
>10-<25	p17	1-18	MGARASVLSGGELDRWEK	14
	p17	9-26	SGGELDRWEKIRLRPGGK	22
	p17	17-34	EKIRLRPGGKKKYKLKHI	14
	p17	32-46	KHIVWASRELERFAV	11
	p17	70-86	TGSEELRSLYNTVATLY	11
	p24	9-26	QMVHQAISPRTLNAWVKV	11
	p24	23-40	WVKVVEEKAFSPEVIPMF	11
	p24	31-47	AFSPEVIPMFSALSEGA	11
	p24	141-158	IVRMYSPTSILDIRQGPK	17
	p24	156-173	GPKEPFRDYVDRFYKTLR	11
	p24	185-202	MTETLLVQNANPDCKTIL	11
	p15	37-52	HIAKNCRAPRKKGCWK	14
	p15	72-89	GKIWPSHKGRPGNFLQSR	17
	p15	93-112	TAPPEESFRFGEETTTPSQK	14
	p15	111-127	QKQEPIDKELYPLASLR	17
	p15	118-137	KELYPLASLRSLFGNDPSSQ	17
	Int	250-267	VIQDNSDIKVVPRRKAKI	11
	Rev	14-30	KTVRLIKFLYQSNPPPS	11
	Nef	81-97	YKAAVDLSHFLKEKGGL	11
	Nef	104-121	QKRQDILDLWVYHTQGYF	11
	Nef	112-127	LWVYHTQGYFPDWQNY	14
	Nef	162-178	NSLLHPMSLHGMDDPEK	11
	Nef	190-206	AFHHMARELHPEYYKDC	14

^aPercentage of studied individuals in whom peptide was recognized.

The high frequency of recognition of some peptides corresponds to a combination of overlapping but distinct epitopes and promiscuous epitopes presented by different HLA-DR alleles.

In order to further characterize the fine specificity of responses directed against this subset of frequently recognized peptides, we analyzed IFN-γ responses to serial truncations of the peptides. The truncations differed in length by two amino acids for the initial location of the epitope and then by one amino acid for finer mapping. The effects of both serial truncations and variations in peptide concentration on magnitude of the IFN-γ responses are illustrated in Fig. 3. It shows two examples for which different truncations of the same peptide, p24 _133-150, were tested at 10-fold dose intervals ranging from 20 μg/ml to 2 pg/ml. The results demonstrate that a higher concentration of each truncation results in a reproducible increase in the magnitude of the response. Optimal responses were obtained with the full 18-amino-acid peptide and did not significantly differ from results obtained with the 14-mer truncation, p24_133-146. Additional shortening of the peptide resulted in incremental decrease of the response, and therefore, a clear-cut minimal epitope was difficult to define. For all our fine-mapping analyses, we defined the “core epitope” as the smallest peptide triggering a response at least 50% of that observed with the full-length peptide at a concentration of 2 μg/ml. This definition is consistent with criteria that have been used in a study of CMV-specific CD4 T-cell responses (8). In the case of the peptide p24_133-150-specific CD4 cell line in subject LT04, these criteria defined the 11-mer ILGLNKIVRMY as the core epitope. However, smaller responses were still detected for a 10-mer and a 9-mer.

FIG. 3.

Dose-response analysis of truncations of peptide p24_133-150. Cell lines of subject LT04 specific for peptide p24_133-150 were stimulated with serial dilutions of different truncations of the full-length peptide. The dotted line labeled A corresponds to the response to the full-length peptide at 2 μg/ml, and the dotted line B corresponds to 50% of this response. The minimal peptide triggering a response corresponding to ≥50% of the response to the full-length peptide at 2 μg/ml was defined as the core epitope in the fine-mapping experiments.

The generation of CD4 T-cell lines also provided a means of assessing the HLA restriction of these peptide-specific CD4 cells. Cells from the cell line of interest were stimulated with peptide-pulsed autologous B cells, partially HLA-matched BCL, and if available for the specific HLA-DR alleles, with the corresponding HLA-DR-transfected LCL. IFN-γ production was assessed by ICS. Representative data demonstrating HLA class II restriction of this same peptide p24_133-150-specific cell line in subject LT04 are shown in Fig. 4. Peptide-pulsed DRB1*0801-matched BCL were unable to stimulate IFN-γ production by the cell line, whereas DRB1*1301-matched BCL induced IFN-γ production comparable to that of autologous peptide-pulsed BCL. Furthermore, the DRB1*1301- transfected LCL was able to efficiently present the peptide. These results indicate that p24_133-150 is DRB1*1301 restricted in this subject.

FIG. 4.

Peptide p24_133-150-specific CD T cells are restricted by HLA DRB1*1301 in subject LT04. Peptide-pulsed autologous and partially HLA-matched B cells, as well as HLA DR-transfected L cells, were incubated with the p24_133-150-specific cell line from subject LT04 for 6 h; stained for CD4, 7AAD, CD69, and IFN-γ; and analyzed by flow cytometry. The percentage of CD4 + 7AAD_low CD69⁺ IFN-γ⁺ cells is indicated in the upper right quadrant of each plot.

These methods were used to characterize the fine specificity of responses against the frequently recognized peptides p17_37-51, p17_42-58, p24 _133-150, and p24_164-181 and to examine their HLA restriction in selected individuals. These data are summarized in Fig. 5 and show that the core epitopes according to the criteria defined above are 9 to 12 amino acids long.

FIG. 5.

Fine mapping and restriction of epitopes located in peptides p17_37-51, p17_42-58, p24_133-150, and p24_164-181. Peptide-specific CD4 cells lines were used to determine the core epitope and the restricting HLA allele of these frequently recognized peptides. For the two epitopes NKIVRMYSPTSI located in peptide p24_133-150, the significance of the N-terminal asparagine and C-terminal isoleucine (shaded) was not determined, because the required truncations were not available at the time of the assays. We could not differentiate presentation by DRB1*1501 or DRB5*0101. Because of the linkage disequilibrium existing between these two DR molecules, no BCL expressing only one of the two molecules was available. LCL transfected with DR1*1501 and DRB5*0101 were not available at the time that the experiments were performed.

The two overlapping peptides p17_37-51 and p17_42-58 were frequently recognized simultaneously in our cohort. Nine of the 13 subjects targeting peptide p17_37-51 also targeted peptide p17_42-58, and 9 of the 10 subjects targeting peptide p17_42-58 also targeted peptide p17_37-51, suggesting that a common epitope may be located in the overlap. Fine-mapping experiments confirmed that in subject AC87, the cell lines specific for these two peptides targeted the same core epitope, ERFAVNPGLL, located in the overlap and presented by DRB3*0202. However, in subject AC05, the p17_37-51-specific cell line targeted a different epitope, RELERFAVN, presented by DRB1*1302, whereas the p17_42-58-specific cell line also targeted ERFAVNPGLL and was restricted by another DRB3 allele, DRB3*0301. As this core epitope is located in the overlap of the two peptides used to grow the cell lines, these data also suggest that a peptide longer than the core epitope may be necessary to efficiently trigger peptide-specific CD4 T-cell proliferation.

Peptide p24_133-150 contains a DRB1*1301-restricted epitope that partially overlaps with a distinct epitope presented by at least two different DR alleles. The dose-response curves for the serial truncations of the two DRB1*1301-restricted responses differed slightly, resulting in the definition of core epitopes differing by one amino acid.

Similarly, the most frequently recognized peptide in our study, peptide p24_164-181, contains at least two distinct, partially overlapping epitopes. The 12-mer RFYKTLRAEQAS was shown to be presented by at least three different HLA-DR alleles, although the core epitope was demonstrated to be shorter by one amino acid in subject AC04 and two amino acids in subject AC87. These data also show that the same DRB1*1301 can present two distinct overlapping epitopes in the same peptide.

Taken together, these results demonstrate that the presence of both overlapping but distinct epitopes and promiscuous epitopes presented by different HLA-DR molecules contributes to the high frequency of recognition of some HIV-1 peptides.

The most frequently recognized peptides bind broadly to multiple HLA-DR molecules.

We next assessed whether the eight peptides that were shown to be targeted by ≥25% of the studied individuals showed cross-reactive HLA-DR binding capacity (Table 4). Some of the core epitopes shown in Fig. 5 were also investigated in these experiments. Assays of binding to eight frequent HLA-DR molecules were performed. All eight peptides bound to at least four of the eight HLA-DR alleles tested, and two peptides (the most frequently recognized, p24_164-181 and p24_133-150) bound to all eight tested HLA-DR molecules. The core epitopes RFYKTLAEQAS and NKIVRMYSPTSI bound to seven and eight of the eight HLA-DR molecules tested, respectively. Interestingly, testing of truncations of peptide p24 _164-181 showed that differences of one amino acid can result in marked differences in binding to some HLA-DR molecules, whereas binding to others is unaffected.

TABLE 4.

Peptide-binding capacities of common HLA-DR molecules

Protein	Amino acid positions	Sequence	IC50a								No. of molecules bound (out of 8)
			DRB1*0101	DRB1*0401	DRB1*0405	DRB1*0701	DRB1*1101	DRB1*1302	DRB1*1501	DRB5*0101
p17	37-51	ASRELERFAVNPGLL	285.39	90	431	116	1,834	10	56	3,673	6
p17	41-51	LERFAVNPGLL	3,185	19,263	23,880	1,407	7,115	59	6,822	25,832	1
p17	42-58	ERFAVNPGLLETSEGCR	96	1,222	113	3,095	463	16	16,298	7,393	4
p17	77-94	SLYNTVATLYCVHQRIEV	279	61	110	78	3,841	92	1,701	781	6
p24	133-150	WIILGLNKIVRMYSPTSI	0.62	1.4	3.3	108	19	1.8	0.31	102	8
p24	133-144	WIILGLNKIVRM	12	5,353	8,920	13,634	107	0.71	17	171	5
p24	135-145	ILGLNKIVRMY	3,292	653	1,233	5,536	2,074	5.4	6.5	1,615	3
p24	139-150	NKIVRMYSPTSI	0.48	4.4	6.6	482	46	156	0.84	760	8
p24	164-181	YVDRFYKTLRAEQASQEV	6.0	15	35	231	101	97	176	32	8
p24	167-178	RFYKTLRAEQAS	3.8	700	770	900	790	>50,000	217	803	7
p24	168-180	FYKTLRAEQASQE	0.37	13	89	1,887	65	42,686	651	88	6
p24	168-179	FYKTLRAEQASQ	0.42	36	3,429	3,944	24	>50,000	1,352	76	4
p24	169-177	YKTLRAEQA	71	33,147	>50,000	30,666	26,791	>50,000	>50,000	24,712	1
p15	66-81	RQANFLGKIWPSHKGR	758	446	732	7,187	629	454	1.3	48	7
Nef	176-193	PEKEVLVWKFDSRLAFHH	3.4	738	1,050	97	144	512	0.080	177	7
Nef	184-199	KFDSRLAFHHMARELH	269	3,087	3,864	70	47	28,600	124	102	5

^aAn IC₅₀ of ≤1,000 is considered significant (underlined). Boldface indicates peptides tested in comprehensive screening experiments and recognized by at least 25% of individuals.

Strongest HIV-specific CD4 T-cell responses are seen in individuals who discontinued therapy after treatment of primary infection.

We next investigated the relationship between the magnitude of HIV-specific CD4 responses, the number of peptides recognized, the virus load, and the treatment status (Fig. 6). Among the subjects treated during acute infection, the magnitudes of responses were significantly higher in individuals who were off therapy after one to three treatment interruption cycles (acute-STI) than in those individuals continuously on therapy (acute-Rx) (medians, 1,890 [range, 560 to 14,600] and 270 [range, 140 to 1,105] SFC/10⁶ CD8-depleted PBMC, respectively; P = 0.0007). Similarly, the number of responses was significantly higher in individuals in the acute-STI group than in the subjects in the acute-Rx group (medians, 18 [range 3 to 36] and 2 [range, 0 to 11], respectively; P = 0.002). These differences remained significant when a Bonferroni correction was applied for the three comparisons made. The median time off therapy for the subjects who had discontinued treatment was 497 days (range, 110 to 1,126 days). In chronic HIV-1 infection, there was a trend toward increased numbers of responses in untreated subjects compared to continuously treated individuals (P = 0.05), although the numbers studied were small. Some individuals with nonprogressive infections had broad and strong responses; the number of responses detected by IFN-γ production in this group was variable and not statistically different from that in untreated subjects with chronic viremia (>10,000 RNA copies/ml). In spite of the differences existing among the groups, our results did not show a direct correlation between the virus load on one hand and the magnitude or breadth of responses on the other hand when the whole cohort of 36 subjects was considered. Based on a linear-regression model, no correlation was found between the plasma virus load (in log₁₀ RNA copies/ml) and either the total magnitude of the HIV-specific IFN-γ response (P = 0.33) or the number of individual peptides targeted (P = 0.18). However, there was a highly significant correlation between the magnitudes and breadths of responses (P < 0.0001).

FIG. 6.

Comparison of magnitudes and breadths of responses among the different study groups. (A) Total magnitude of HIV-specific CD4 T-cell responses per individual for the different groups described in Materials and Methods. (B) Numbers of confirmed CD4 T-cell responses per individual for the different groups. The median values are indicated by horizontal bars. P values were calculated using the two-tailed Mann-Whitney test.

HIV-1-specific CD4 responses can be more narrowly directed than CD8 responses in the same person.

In order to compare the HIV-specific CD4 and CD8 T-cell responses in the same individuals, we performed a comprehensive analysis of CD8 T-cell responses in a subgroup of four individuals using CD4-depleted PBMC in the same matrix-based strategy and compared these data to CD4 T-cell responses. In all four individuals, we observed stronger CD8 responses in terms of the magnitude and number of peptides targeted, as well as a broader targeting by CD8 T cells with regard to the number of viral proteins recognized (Fig. 7), contrasting with the weaker and less numerous CD4 T-cell responses, which were more narrowly directed against only some of the HIV proteins. Therefore, many highly conserved peptides outside Gag and Nef proteins were not targeted by CD4 T cells, although they generated strong CD8 T-cell responses. Moreover, the immunogenic peptides were infrequently the same for CD4 and CD8 T-cell responses.

FIG. 7.

Comparative patterns of HIV-1-specific CD4 and CD8 T-cell responses. Subjects AC46, AC14, CRU6, and CO6 were comprehensively screened for HIV-1-specific CD4 and CD8 T-cell responses using CD8- and CD4-depleted PBMC, respectively The horizontal axis represents the 410 overlapping peptides and the corresponding HIV proteins, and the vertical axis represents the magnitudes of responses in IFN-γ ELISPOT, expressed as SFC. The red bars crossing the gap between the CD4 and CD8 graphs are peptides targeted by both cell types in the same individual.

DISCUSSION

In this study, we performed a comprehensive analysis of the breadths and specificities of HIV-1-specific CD4 T-cell responses in order to determine the regions of the virus that are most immunogenic. In persons at various stages of infection, HIV-1-specific CD4 responses were identified in the majority of individuals, with consistent immunodominance of Gag and Nef proteins. We showed that the focused recognition of overlapping but distinct epitopes and epitopes promiscuously presented by different HLA class II molecules both contribute to the high frequency of responses to a limited subset of peptides that was observed. Moreover, these frequently recognized peptides broadly bound to different common HLA-DR alleles. Despite the extensive HLA diversity in the study population, there was a single peptide in Gag that was recognized by >50% of the subjects. This peptide lies in a well-conserved region of the C-terminal dimerization domain of p24 (capsid antigen), which is required for capsid dimerization (23, 68) and Gag oligomerization (21). Interestingly, this peptide partially overlaps with the major homology region, a 20-amino-acid-long stretch essential for virus assembly, maturation, and infectivity that is conserved across all known oncoviruses and lentiviruses (22, 65; R. Patarca and W. A. Haseltine, Letter, Nature 312:496, 1984) and in the yeast retrotransposon Ty-3 (44). Ninety-three percent of the subjects with positive responses targeted at least one peptide within a small subset of eight. These results suggest that a limited number of peptides may induce CD4 T-cell immune responses in a genetically diverse population. The paucity of CD4 T-cell responses to other HIV-1 proteins, including well-conserved regions of the virus, contrasted with the pattern of CD8 T-cell responses in the same subjects in whom there was broad targeting of other proteins.

The immunodominance of Gag and Nef for CD8 T-cell responses has been well documented, but there have been limited data on the precise locations of highly targeted regions for CD4 T-cell responses. These results indicate that HIV-1-specific CD4 T-cell responses were predominantly located in p17, p24, and p15 and throughout Nef in all groups of individuals studied, irrespective of treatment or disease status. Specific CD4 T cells were readily detectable in the majority of individuals, although with a low magnitude in most of them, in agreement with previously published results (6). Whereas the majority of the Gag and Nef peptides were recognized by one or more individuals, no response could be detected against some of the accessory proteins (Vpu and Vif), and large parts of the proteins encoded by the Pol and Env genes were also poorly targeted. This contrasts with the pattern of responses observed for CD8 T cells. Whereas Gag and Nef are often immunodominant proteins for CD8 T-cell responses (1, 20), other proteins are frequently targeted as well. We confirmed these different patterns of responses in a subgroup of four individuals for whom both CD4 and CD8 T-cell responses had been comprehensively studied. These data show that strong CD8 T-cell responses can be detected against proteins not targeted by CD4 T cells; however, the significance of this observation is unclear.

Such a preferential targeting of a limited number of viral proteins by CD4 T cells, although striking, does not appear to be unique to HIV-1. A marked hierarchy of immunodominance has been described for CD4 T-cell responses to Epstein-Barr virus proteins (35). The mechanisms of such preferential targeting are not known. Sequence variation of the autologous viral sequence compared to the consensus sequence used for the synthetic peptides may contribute to the pattern of immunodominance observed. The conserved structure of Gag, for example, may contribute to the high frequency of responses observed against p24 peptides. However, very well conserved regions of the genome, particularly in Pol, were rarely targeted by CD4 T cells and were simultaneously well recognized by CD8 T cells. In contrast, the more variable Nef protein was frequently recognized. Taken together, these data suggest that mechanisms other than sequence variation contribute to the hierarchy of responses observed. Definitive proof would require sequencing of the autologous virus and synthesis of autologous peptides to allow the assessment of responses to autologous virus.

The variable recognition of HIV-1 proteins by CD4 T cells might also be affected by a differential ability of HIV-1 proteins to be processed and then presented through the exogenous pathway. In an animal model using mice immunized with HIV-1 Env, CD4 T-cell epitopes were clustered in slightly glycosylated, exposed nonhelical strands of the protein (61). Additionally, anti-Env antibodies can alter the uptake or processing of gp120 (26). However, data for subjects with acute HIV-1 infection showed that relatively broad and strong CD4 T-cell responses can occasionally be generated against proteins infrequently targeted in the chronic phase of HIV-1 infection, for example, against the components of Env, gp120, and gp41 (39, 47). These studies showed that these Env-specific responses quickly wane over time, regardless of treatment status, and that Gag-specific responses predominate later on (39), in agreement with our study, which focused on subjects for whom HIV-1 diagnosis had been established at least 1 year prior to analysis. Therefore, the paucity of responses observed against at least some of the HIV-1 proteins may not be due to an intrinsic inability to be processed by the exogenous pathway and presented by HLA class II molecules. Of note, the reported disappearance of responses observed during acute HIV-1 infection can occur in the absence of immune escape mutations, and these HIV-1-specific CD4 T cells can persist at very low frequencies, undetectable by standard assays, like IFN-γ ELISPOTs or lymphocyte proliferation assays, but identified by analysis of T-cell receptor transcripts (39).

The present study identified numerous peptides that were recognized in >10% of the studied individuals (Table 3). Some of them overlap or contain described CD4 T-cell epitopes, like p17-Gag_37-51 (36), p24-Gag_133-150 (10), and p24-Gag_164-181 (38), but most have not been described as targets for CD4 responses. Moreover, the target epitopes for CD4 responses differ from those for CD8 responses. For example, in a large study using the same set of overlapping peptides, the eight peptides most frequently targeted by CD8 cells were clustered in the middle region of the Nef protein (amino acids 65 to 148) (20), whereas our results showed that CD4 T-cell responses more frequently targeted the C-terminal end of this protein

The number of responses to overlapping peptides may overestimate the number of epitopic regions, because an epitope can be located in the overlap of two consecutive peptides. Positive responses to two adjacent overlapping peptides occurred for 30% of the responses, and responses to three adjacent peptides occurred in 13% of the responses, which could correspond in some cases to one and two epitopic regions, respectively. Fine mapping of epitopes located in the two frequently recognized adjacent peptides p17_37-51 and p17_42-58 confirmed that the 10-amino-acid overlap of our peptide set can correspond to the core epitope triggering a positive response. However, we also demonstrated that simultaneous responses to these two peptides in an individual can also correspond to two distinct epitopes. Although a frequent association of positive responses is suggestive of an epitope located in the overlap, it is difficult to make general conclusions.

Different factors likely contribute to the broad recognition of a limited subset of HIV peptides among individuals with various HLA types. Promiscuous presentation of peptides by several class II alleles has been reported in different settings (17, 58, 59). In order to investigate this possibility, studies of binding to a panel of eight common HLA-DR molecules were performed and showed that the eight most frequently recognized peptides bound broadly to a minimum of four of the eight alleles tested, and two, including the most frequently targeted peptide, p24_164-181, bound to all of the DR molecules tested. Of note, all peptides in this subset contain at least one sequence consistent with the previously described HLA-DR supermotif (59, 66). However, this characteristic is not unique to this subset of peptides, and several HIV-1 peptides harboring this HLA-DR supermotif that have been shown to bind to many HLA-DR molecules (66) were rarely targeted in our study. This is particularly true for those peptides located in the reverse transcriptase.

Fine mapping and HLA restriction of the precise epitopes targeted showed that the presence of both overlapping but distinct epitopes and epitopes that were identical or that differed by only one or two amino acids and were presented by different HLA-DR alleles contributed to the high frequency of recognition. Additional amino acids on one or both sides of these core epitopes lead to an increase in the IFN-γ response of the specific cell line, as recently described for CMV-specific CD4 T-cell responses (8). The data illustrate a previously noted dichotomy between class I and class II binding and T-cell recognition. Class I molecules bind peptides of an exact size, and therefore the minimal ligand is also optimal. By contrast, class II molecules bind peptides in multiple registers, where a core region engages the binding grove with additional flanking residues contributing a small part of the overall binding energy. Therefore, the core binding region is usually not the optimal ligand. Conversely, the optimal ligand is not the minimal ligand but rather corresponds to the core binding region plus a few flanking residues on either or both sides.

It was noted long ago that recognition of a given epitope in the context of multiple T cells is usually due to the same binding core, with the flanking regions being variably crucial for some T-cell receptors and not others. Similarly, it has been demonstrated (46, 50, 58) that recognition of promiscuous peptides in the context of different HLA class II molecules is due to the peptide binding MHC in the same register but with each interaction and T-cell recognition being variably influenced by the regions flanking the core. This has been demonstrated in the cases of influenza virus (HA 307 to 319), tetanus toxoid (TT 830 to 843), and Plasmodium falciparum CSP (378 to 398). The data shown here are consistent with this general mode of recognition.

Comparison of the magnitudes and breadths of responses among the different groups showed that the strongest and broadest responses were detected in subjects treated in acute infection who underwent treatment interruption and in individuals who spontaneously controlled HIV replication without therapy. Some subjects in the latter group, however, had weak or undetectable HIV-specific CD4 T-cell responses in IFN-γ ELISPOT assays, suggesting that spontaneous control of viremia does not always require strong HIV-specific IFN-γ-secreting CD4 responses. Overall, there was no correlation between the virus load at the time of the study and either the magnitudes or breadths of IFN-γ responses, in agreement with previously published data (6). By contrast, recent studies suggest an inverse correlation between plasma viremia and the number of IL-2-secreting HIV-specific CD4 T cells (25, 27). One limitation of our measurements of the breadths of responses may be due to the sensitivity of our assays, and this may contribute to the correlation observed between the magnitudes and breadths of responses in the entire cohort. Most of the individuals treated during acute HIV infection who subsequently stopped therapy in the context of an STI trial showed strong and broad responses in contrast to subjects continuously on therapy. Previous studies have shown an increase of HIV-specific IFN-γ-secreting CD4 T cells after treatment interruption and rebound of viremia (15, 42). This increase, however, was short-lived in subjects treated during chronic HIV infection, lasting <1 month, with responses going back to a low or undetectable level (15). This contrasts with our data for subjects treated during acute HIV infection, in whom strong responses were detected at a median of >16 months after the last treatment interruption. However, the extent to which the magnitude or breadth of the responses contributes to maintenance of viremia control is unknown and will require further longitudinal studies. The importance of CD4 T-cell help for adequate generation of efficient, protective memory CD8 T-cell responses during priming has recently been shown, but an absence of CD4 help during rechallenge with antigen did not appear to impair the proliferative capacity of memory CD8 T cells in some models (12, 29, 57, 60). However, the loss of CD4 T cells during chronic viral infections can adversely affect the function of antigen-specific CTLs (40, 70), and other models suggest that memory CD8 T-cell differentiation continues for several weeks after the resolution of an acute viral infection (32). Likewise, whether the quality and quantity of the HIV-specific CTLs and B cells is modified by the prolonged persistence of these HIV-specific CD4 T-cell responses remains to be shown.

In conclusion, HIV-1-specific CD4 T-cell responses are markedly dominated by epitopes in Gag and Nef in established HIV-1 infection. Although the contributions of these responses are unknown, identification of peptides frequently targeted across a diverse population with regard to HLA class II types may prove useful for the design of new immunotherapies and vaccines.