Correlates of Nontransmission in US Women at High Risk of Human Immunodeficiency Virus Type 1 Infection through Sexual Exposure

Abstract

Seventeen women who were persistently uninfected by human immunodeficiency virus type 1 (HIV-1), despite repeated sexual exposure, and 12 of their HIV-positive male partners were studied for antiviral correlates of nontransmission. Thirteen women had ≥1 immune response in the form of CD8 cell noncytotoxic HIV-1 suppressive activity, proliferative CD4 cell response to HIV antigens, CD8 cell production of macrophage inflammatory protein–1β, or ELISPOT assay for HIV-1–specific interferon-γ secretion. The male HIV-positive partners without AIDS had extremely high CD8 cell counts. All 8 male partners evaluated showed CD8 cell–related cytotoxic HIV suppressive activity. Reduced CD4 cell susceptibility to infection, neutralizing antibody, single-cell cytokine production, and local antibody in the women played no apparent protective role. These observations suggest that the primary protective factor is CD8 cell activity in both the HIV-positive donor and the HIV-negative partner. These findings have substantial implications for vaccine development.

Heterosexual transmission of human immunodeficiency virus type 1 (HIV-1) has dominated the spread of the HIV pandemic. Recent research has focused on the apparent resistance of some highly exposed but still HIV-negative persons [1–5]. Understanding effective resistance mechanisms is crucial for designing preventive measures and determining which vaccine candidates warrant intensive testing.

Longitudinal cohort studies have implicated cofactors—symptomatic sexually transmitted diseases (STDs) [6–8] and virus load of the donor [9–11], in particular—in the transmission of HIV by sexual contact. It remains unclear why some persons remain uninfected despite repeated exposure. In various studies, resistance has been attributed, in part, to noncytotoxic CD8 cell responses [12–14], β-chemokines [15], intrinsic CD4 cell resistance [16], cytotoxic T lymphocyte (CTL) activity [1, 5, 17–19], neutralizing antibody to coreceptor molecules [20], viral phenotype [21], or mucosal antibody [22, 23]. Thus far, studies have focused mainly on one or two putative mechanisms at a time. As evidence mounts that a combination of factors may be involved in nontransmission [24], a more comprehensive picture of resistance potential is needed for vaccine testing.

Previously, we studied 212 discordant couples (where one partner was HIV positive and the other was HIV negative) in stable heterosexual relationships and compared them with 68 concordant couples (where both were HIV positive) for whom the direction of spread of the virus could be inferred. The following variables correlated significantly with discordance (nontransmission): higher CD8 cell counts and, to a lesser extent, lower plasma virus levels in the HIV-positive partner [25], less anal intercourse [26], and less frequent evidence (by serum antibody) of infection with herpes simplex virus type 2 (HSV-2) and Mycoplasma genitalium [27].

This report focuses on a small group of highly exposed uninfected (EU) women and, when available, their HIV-positive male partners. The cohort was studied intensively by sharing specimens with multiple laboratories to evaluate the potential role of the following factors: CD8 cell noncytotoxic activity, CD8 cell chemokine production, mucosal antibody, CD4 cell coreceptor mutations, CD4 cell proliferative response, neutralizing antibody, and CTL responses. CD8 cell anti-HIV activity of the male partners also was evaluated. The goal was to establish resistance profiles and to determine whether there appeared to be more than one profile. On the basis of findings in our prior studies, we hypothesized that there would be a single resistance pattern of CD8 cell functional activities in both the HIV-positive men and their HIV-negative female partners.

Methods

Study cohort

The study cohort consisted of 17 women who remained uninfected, despite a history of heavy exposure to HIV through repeated, unprotected sexual contact with an infected partner, and 12 of their regular, male HIV-positive partners. Criteria for inclusion were longstanding sexual partnership up to the time of the male partner's first positive HIV test and/or continued unprotected intercourse after the male partner was infected and no other identified risk for HIV infection for the women. The HIV-negative status of the women was determined by HIV-1 antibody status, qualitative plasma DNA polymerase chain reaction, and cocultivation. HIV antibody–positive status was confirmed by repeat ELISA and Western blot tests.

Participants completed a structured questionnaire administered by a trained interviewer regarding frequency and type of sexual behaviors. Both partners were examined for evidence of STDs, as described by Perez et al. [27].

Control subjects

Healthy seronegative control subjects who were at low risk of HIV exposure were recruited locally for laboratory assays by the participating laboratories. Seven women who were infected with HIV-1 by sexual exposure to a male partner and 9 HIV-positive men who had transmitted HIV-1 to their partners constituted HIV-positive control subjects. These HIV-positive control subjects were members of HIV-positive concordant couples who were recruited from the same northern New Jersey clinical centers and met the same eligibility criteria for sexual exposure as the HIV-discordant couples. Because of scheduling constraints and specimen availability, not all participants underwent the same studies.

Flow cytometric immunophenotyping

Lymphocyte subsets were determined by standard flow cytometry techniques for simultaneous direct 2-color immunofluorescence staining of whole blood. CD4 and CD8 cell counts and CD38 and CD45RO/RA phenotyping were derived from total and differential leukocyte counts obtained by using a Sysmex E-2500 electronic cell counter (TOA Medical Electronics). Flow cytometry and plasma RNA assays are quality assured within the National Institutes of Health (NIH)/Division of AIDS program.

Virus load

We used a nucleic acid sequence–based amplification assay (Organon Teknika) to quantitate HIV RNA extracted from 100 μL of plasma.

Mononuclear cell susceptibility

Genotyping of the chemokine receptor (CR) genes CCR5 and CCR2 was done as described by Kostrikis et al. [28], to determine the presence of CCR5-Δ32 or CCR2-64I mutation alleles.

CD4 cell infectivity

Purified CD4 T lymphocytes from uninfected women were exposed in vitro to their partner's HIV isolate or to laboratory strains SF2 or SF33, as described by Stranford et al. [12].

Neutralizing antibody

Serum was diluted 10- and 100-fold, after which neutralizing antibody to HIV-1₂₀₄₄ was measured as TCID₅₀/mL, as described by Wu et al. [29].

Mucosal antibodies

HIV-specific IgA and IgG levels were measured in serum, vaginal washes, and saliva of highly exposed women and HIV-positive control subjects by indirect ELISA [30] and Western blot chemiluminescent (WBC) detection [31]. The cutoff for antigen-specific IgA and IgG positivity was 100 ng/mL. Vaginal washings were obtained after 72 h of abstinence from vaginal intercourse or douching, as described elsewhere [32]. The fluid was centrifuged, and the supernatant and sediment were frozen at −70°C. Saliva was collected in an intraoral Schaffer cup placed over Stensen's duct to stimulate secretion from the parotid gland [32]. Aliquots were stored at −70°C. Positive controls for IgG and IgA HIV antibodies included serum, saliva, seminal fluid, and vaginal washes from 50 HIV-positive clinic patients at the University of Alabama at Birmingham. Although all serum samples and 69%–100% of external secretions were positive for IgG anti-HIV antibodies, the percentage of persons who were weakly positive for IgA anti-HIV antibodies varied from 68% in serum to 24% in vaginal washings. As detailed elsewhere [30], naturally occurring IgA but not IgG non-HIV antibodies in the gp120 or gp160 preparations used may cause false positivity of ELISA for IgA anti-HIV antibodies. This can be avoided by use of a more sensitive and specific WBC assay.

CD8 cell–nonspecific stimulation studies: chemokine production

Peripheral blood mononuclear cells (PBMC) were isolated from heparinized peripheral blood by density centrifugation. In total, 2 × 10⁶ cells were cultured in the presence and absence of 25 ng/mL PMA (Sigma), 1 μg/mL ionomycin (Sigma), and 1.3 μL Golgi Stop (PharMingen) containing monensin for 4 h at 37°C. Surface staining was done using peridinin chlorophyll protein anti-CD8 and allophycocyanin anti-CD4 (Becton Dickinson Immunocytometry Systems [BDIS]). Intracellular staining was performed with fluorescein isothiocyanate anti–macrophage inflammatory protein (MIP)–1β (R&D Systems) in combination with phycoerythrin anti–interferon (IFN)–γ (PharMingen) after permeabilization by use of a Cytofix/Cytoperm kit (PharMingen). Cell samples were analyzed subsequently by flow cytometry (FACSCalibur; BDIS).

Subset-specific single-cell cytokine production

Whole blood aliquots were incubated with PMA (20 ng/mL) and ionomycin (1 μg/mL) in the presence of brefeldin A (10 mg/mL) as an inhibitor of intracellular transport. The cells then were labeled with fluorescent conjugated monoclonal antibodies (MAbs) to specific CD4 or CD8 cell markers. The erythrocytes were lysed with NH₄Cl and were washed, and a membrane permeabilization buffer was added. Intracellular cytokine expression was identified by intracellular labeling by using fluorescent tagged MAbs to the specific cytokines: interleukin (IL)–2, IL-4, IL-10, tumor necrosis factor (TNF)–α, TNF-β, and IFN-γ. Production of MIP-1β chemokine was quantified as a percentage of the total CD4 or CD8 cell subset.

CD4 cell proliferation assays

Lymphocyte proliferative responses to recall microbial antigens and to HIV antigens were measured by culturing 10⁵ unstimulated or antigen-stimulated PBMC for 6 days and measuring the incorporation of tritiated thymidine [33]. Recall antigens used were candida, tuberculin purified protein derivative, streptokinase, and tetanus toxoid. HIV antigens used were p24, p25, p66, gpl60-LAI, gp160-MN (baculovirus and Vero cell expression systems), and gp120-LAI. Results were expressed as a stimulation index (SI) for each antigen (counts per minute stimulated/counts per minute unstimulated).

CD8 cell noncytotoxic anti-HIV activity

PBMC were isolated from whole blood by ficoll-hypaque (Sigma) gradient separation [34], and the CD4 and CD8 cellular fractions were purified by using anti-CD4 or anti-CD8 antibody–coated immunomagnetic beads (Dynal). Acute suppression assays were conducted as described elsewhere [35]. In brief, CD4 cells from unexposed control subjects were stimulated with phytohemagglutinin (PHA) and were acutely infected with 10⁴ TCID₅₀ of SF33 virus, a β-chemokine–resistant syncytium-inducing strain. These infected CD4 cells were cultured alone and in the presence of CD8 cells at CD8:CD4 cell ratios of 0.5:1 and 2:1 in growth medium containing 100 U/mL human recombinant IL-2 (provided by Glaxo-Wellcome), were placed in 96-well plates at a volume of 2000 μL/well, and were incubated for 7–14 days. Collected supernatants were monitored for reverse transcriptase (RT) activity [36]. Percent suppression was calculated by comparing the mean RT value from 6 control wells containing only CD4 cells with mean RT activity in duplicate wells containing the same ratio of CD8 and CD4 cells together. CD8 cells from an unexposed control subject were included in each set of experiments. Suppression of the SF2 strain of HIV was assayed by following the same protocol. SDs between replicate wells were routinely <10%.

HIV-specific IFN-γ secretion of PBMC (ELISPOT assay)

Frozen PBMC were processed by ELISPOT assay, according to the protocol of Larsson et al. [37], which detects primarily CD8 T cell responses. PBMC were assayed directly after thawing in a culture medium without prior in vitro stimulation, cultivation, or expansion. PBMC (5 × 10⁵ viable cells) were infected with the recombinant vaccinia virus (rVV)–expressing HIV-1 antigens pol, nef, gag, and env or control vaccinia (NIH Research and Reference Reagent Program). As positive controls for IFN-γ production, PBMC were polyclonally stimulated with PHA. Spot-forming cells in replicate wells were enumerated by ELISPOT series 4 analyzer (Cell Technology). Background counts for PBMC infected with control rVV were subtracted from the counts obtained for HIV antigens. Results are reported as sfc/10⁶ PBMC. When the viable cell count was <2 × 10⁶, ELISPOT was done by using as many HIV antigens as possible in the following order: pol, nef, gag, and env.

Cocultivation

HIV-1 isolations were performed by cocultivation of a study participant's PBMC or plasma with a mixture of PBMC from 3 uninfected donors. We added 10⁶ of the participant's PBMC to 9 × 10⁶ normal PBMC in 10 mL of complete medium (RPMI 1640; Gibco Life Technologies) and 20% fetal bovine serum (Gibco Life Technologies) containing 2.0 μg/mL PHA-L (Roche Molecular Biochemicals) and incubated the combination at 37°C for 2 days. The cells were centrifuged (500 g) from the culture medium and were resuspended in fresh complete medium containing 20 IU of IL-2/mL (Roche Molecular Biochemicals). Ten percent of the conditioned culture medium was retained. We mixed 1 mL of a participant's plasma with 10⁷ normal cells and incubated these at 37°C for 1 h. The cells were washed twice with Dulbecco's PBS (Gibco/Life Technologies) and were cultured as described above. Cell supernatant was monitored every 48 h for HIV-1 p24 core antigen (ELISA; Coulter). At each time point, the cells were counted, and the culture was adjusted to a density of 2 × 10⁶ cells/mL. The cultures were maintained for 14 days or until a positive result was obtained. Cultures were considered to be positive for HIV-1 when the p24 concentration in the cell supernatant reached 100 pg/mL (∼4 times the background) on 2 consecutive time points [38].

Statistical analysis

Summaries are reported as mean ± SD unless noted. Comparisons among groups were done by using t tests, analysis of variance, or rank sum tests. Associations among immune responses and subject characteristics were tested by Fisher's exact test. We used the Spearman rank correlation coefficient to measure the strength of association among numeric variables. SAS version 6.12 (SAS Institute) and StatXact (Cytel) software programs were used to conduct the analyses. The criterion for statistical significance was P < .05; all P values are 2-tailed.

Results

Demographics and Risk Profiles

The cohort consisted of 17 EU women who remained seronegative after 1–7 years of follow-up (median, 4 years; table 1). There were 12 persistently HIV-serodiscordant couples and 5 additional EU women whose HIV-positive male partners were no longer available for study because of death, unwillingness to participate, or travel distance from residence. Table 1 lists the duration and frequency of exposure and time since last unprotected intercourse. All couples had either frequent unprotected vaginal intercourse up to the time of the HIV-positive diagnosis or frequent unprotected intercourse after the man became seropositive. Five women had practiced anal intercourse with their partner during a time of risk for infection. Five couples maintained unprotected sexual contact throughout the study follow-up. The women were 20–51 years old (38 ± 8 years): 11 (65%) identified themselves as white, 5 (29%) as Hispanic, and 1 (6%) as black. The men were of the same ethnicity as their partners: 8 (67%) white, 3 (25%) Hispanic, and 1 (8%) black.

Table 1.

Risk characteristics of human immunodeficiency virus type 1 (HIV-1)–exposed uninfected women.

Subject	Age, years	Ethnic group	Duration of exposure, years	Last unprotected exposure, month/year	Frequency of vaginal intercourse during exposure	In vitro HIV-1 infectivity of CD4 cellsa	STD profileb
A	44	Hispanic	9	Ongoing	>10 times/month	NA	HSV-2
B	25	White	6	Ongoing	20 times/month	+P, +	HSV-2
C	45	White	8	Ongoing	9 times/month	+P, +	HSV-2, M. genitalium, Hx
D	30	White	9	Ongoing	4–6 times/month	+	No Hx
E	39	Hispanic	11	Ongoing	0.5 time/month	NA	HSV-2
F	48	Hispanic	3	6/1995	Daily	+	Hx
G	34	White	2	8/1994	Daily	NA	M. genitalium, Hx
H	20	White	8	11/1995	Daily	+	HSV-2, M. genitalium, Hx
I	38	Hispanic	1	8/1995	>10 times/month	+P	HSV-2
J	37	White	4	4/1996	>10 times/month	+P	HSV-2, M. genitalium
K	32	White	6	1989	>10 times/month	+P, +	M. genitalium
L	36	White	7	3/1996	>10 times/month	NA	HSV-2, M. genitalium, Hx
M	39	White	2	1989	>10 times/month	+	No Hx
N	32	White	5	1990c	6–9 times/month	NA	HSV-2, Hxd
O	38	White	3	1985	>10 times/month	NA	M. genitaliume
P	31	Black	3	1998	4–5 times/month	NA	Hxd,e
Q	51	White	1	1999	4–5 times/month	NA	No Hx d,e

NOTE. HSV-2, herpes simplex virus type 2; Hx, history of genital herpes, trichomonas, chlamydia, gonorrhea, or syphilis; NA, not available; STD, sexually transmitted disease.

^a+P, infectivity with partner's viral strain; +, infectivity with strain other than partner's; +P, +, infectivity with both strains.

^bHSV-2, antibody positive for HSV-2; M. genitalium, antibody positive for Mycoplasma genitalium.

^cSexual exposure continued with occasional lapses in condom use.

^dM. genitalium serologic test results not available.

^eHSV-2 serologic test results not available.

STDs

HSV-2 and M. genitalium serologic test results were available for 15 EU women. Ten women (67%) were antibody-positive for HSV-2, a slightly greater prevalence than that reported among uninfected women in discordant couples (41%) and among HIV-positive women in concordant couples (60%) [18]. Eight women (53%) were antibody-positive for M. genitalium, a prevalence that was not significantly lower than that reported among 33 women whose partners had transmitted HIV (67%) [18]. None of the women had clinical signs of current STD infection. However, 8 women had a clinical history of genital herpes, trichomonas, chlamydia, gonorrhea, or syphilis (table 1).

HIV-Positive Male Partners

CD4 and CD8 cell counts

At enrollment, CD4 cell counts for the 12 HIV-positive male partners were 9–1903 cells/mm³ (median, 460 cells/mm³; table 2); 4 had cell counts <200 cells/mm³, and 6 had cell counts ≥500 cells/mm³. At enrollment, CD8 cell counts were 345–3466 cells/mm³ (median, 1134 cells/mm³), in contrast with the laboratory mean of 514 ± 183 cells/mm³ (maximum, 1605 cells/mm³) in 1864 uninfected healthy persons. In men, these high values were relatively stable during the observation period, as illustrated by the medians over time (table 2) and by figure 1, which shows the range of CD8 cell counts for 7 men who had not progressed to AIDS. Six of these nontransmitting men had very high CD8 cell counts (>1600 cells/mm³, which is >5 SD beyond the normal mean) at least once during follow-up, and 3 had CD8 cell counts >2000 cells/ mm³. At the time of or within 1 year of the women's most recent sexual exposure, the partners' CD8 cell counts were 404–3160 cells/mm³ (median, 1414 cells/mm³). In a previous study of a larger cohort of HIV-discordant male partners, the men had higher CD8 cell counts than HIV-positive male partners who transmitted HIV to their partners [17]. The median counts of the men in this cohort were compared with those of 54 transmitting partners from the previous study and were significantly higher (mean 1201 ± 819 cells/mm³ vs. 766 ± 465 cells/mm³; P = .035, t test of square-root transformed counts).

Table 2.

Immunologic and virologic characteristics of human immunodeficiency virus (HIV)–positive male partners.

Female partner	First positive HIV test	Antiretroviral therapya	Virus load in plasma, HIV RNA copies/mLa	Virus load in semen, HIV RNA copies/mLb	CD4 cell count, cells/mm3a	Median CD8 cell count, cells/mm3c	Neutralizing antibody	Percentage CD8 cell suppression of HIV-1SF33 replicationa	IFN-γ in PBMC ELISPOTd
A	None	None	<1000	17,000	739	3199	NA	98	pol
B	None	None	<1000	1100	189	806	+	70	NA
C	1991	Zdv, acyclovir	20,000	32,000	377	1446	NA	78	—
D	1990	None	<1000	700	981	1219	NA	96	pol, gag
E	1987	None	<1000	NA	1903	1640	+	68	pol
I	1987	Zdv, 3TC	10,000	7100	90	296	NA	NA	NA
K	1989	None	23,000	486,000	655	740	+	86	—
L	1995	Zdv, ddI	<1000	600	315	1602	NA	61	—
N	1990	None	<1000	600	1180	1718	NA	99	pol
O	1985	Zdv	690,000	119,000	152	345	NA	NA	—
P	1982	3TC, Abc, Apv	460,000	NA	9	318	NA	NA	NA
Q	1999	3TC, Zdv, Abc	410	NA	542	1084	NA	NA	NA

NOTE. 3TC, lamivudine; Abc, abacavir; Apv, amprenavir; ddI, didanosine; IFN-γ, interferon-γ; NA, not available; PBMC, peripheral blood mononuclear cells; Zdv, zidovudine.

^aAt the time of enrollment.

^bDuring 1995–1996 (the virus load for 2 transmitting HIV-positive partners was 23,000 and 45,000 HIV RNA copies/mL, respectively).

^cMedian of ≤13 observations over 1–5 years.

^dHIV antigens with response ≥3× that in control subjects.

Figure 1.

CD8 cells/10⁶ cells in human immunodeficiency virus–infected, regular male partners of highly exposed seronegative women. Plot displays all cell counts for men without AIDS at the time of study enrollment and for up to 5 years. Control subjects are 1864 healthy unexposed adults whose CD8 cells were measured at the same laboratory (median CD8 cell count, 484 CD8 cells/10⁶ cells; 99th percentile, 1075 CD8 cells/10⁶ cells; maximum, 1605 CD8 cells/10⁶ cells). Letters at the bottom are designations of female partners.

Virus load

The dates of first positive HIV test results for the men were from 1982 to 1999. At the time of enrollment, plasma levels of HIV-1 RNA were undetectable or were <1000 copies/mL in 7 men, 10,000–50,000 copies/mL in 3 men, and ≥400,000 copies/mL in 2 men (table 2). Six of the 12 men were receiving antiretroviral therapy at the time of enrollment; only 1 was receiving a protease inhibitor at enrollment. Semen from 9 of 9 partners of EU women had a detectable virus load. The semen virus loads of these 9 men did not differ significantly from virus loads in the semen of 2 HIV-positive men who had infected their female partners (HIV RNA range, 500–486,000 copies/mL vs. 23,000 copies/mL and 45,000 copies/mL; P = .35).

CD8 cell noncytotoxic suppression of viral replication in CD4 cells

CD8 cells from the PBMC of 8 HIV-positive partners of uninfected women (A–E, K, L, and N) were tested for CD8 cell suppression of HIV-1 replication in acutely infected CD4 cells, using the syncytium-forming strain SF33. CD8 cells from these HIV-positive partners of EU women exhibited strong noncytotoxic anti-HIV activity. The level of inhibition of viral replication (RT activity) of strain SF33 at a CD8:CD4 ratio of 2:1 was 61%–99% (mean, 82%; median, 82%). These levels did not differ from percent suppression by cells from 7 HIV-positive men who had transmitted HIV to their partners (median, 93%; range, 46%–99%; and mean, 83%; median, 93%; P = .7, rank sum test).

EU Women

Lymphocyte subpopulations

Total CD4 and CD8 cell counts and subpopulation counts in the 17 EU women were normal (mean CD4 cell count, 1065 ± 341 cells/mm³; mean CD8 cell count, 564 ± 223 cells/mm³), except in 1 (subject G) who had a CD4 cell count of 1777 cells/mm³ and a CD8 cell count of 1166 cells/mm³. The EU women had a lower mean proportion of CD45⁺CD4RA (naive) cells than 28 low-risk laboratory control subjects (41% vs. 46% among control subjects, P = .02), whereas the mean proportion of CD45⁺CD4RO (memory) cells was higher among EU women but not significantly so (77% vs. 72%, P = .13). EU women had a lower mean proportion of CD38⁺ cells than control subjects (42% vs. 48%, P = .03). The mean proportion of cells expressing HLA-DR receptor was only slightly lower in EU women than in control subjects (20% vs. 25%, P = .11).

CD4 cell studies: CR mutation patterns

CCR5-Δ32 genotyping of 14 EU women indicated that only 2 were heterozygous for the Δ32 deletion (subjects B and D); the remaining 12 were homozygous wild type. Four women (D, I, J, and O) among 12 were heterozygous for CCR2-64I; none was homozygous. These EU women did not exhibit genetic resistance conferred by either CR mutation.

CD4 cell infectivity

CD4 cells from 5 EU women were exposed in vitro to strains of their partner's virus. Cells from 3 of these women and 4 additional EU women were exposed in vitro to the SF2 strain of HIV-1. All samples were easily infected and showed high levels of viral replication (table 1). Thus, CD4 cells from all 9 women tested demonstrated in vitro susceptibility to HIV infection. Results for 8 of these women (A–E, K, L, and N) are included in the study by Stranford et al. [12].

Neutralizing antibody

Serum samples from the 17 EU women were tested for the presence of neutralizing antibody. Two women tested positive, 1 with a moderate response (1:20 titer) and 1 with a low-level response (1:5 titer); the remaining 15 women tested negative (table 3). Three HIV-positive male partners who were tested were all strongly positive for neutralizing antibody.

Table 3.

Immune responses of exposed uninfected women.

Subject	Neutralizing antibody TCID50/mL to HIV-12044	Percentage CD8 cell suppression of HIV-1 replicationa		Percentage CD8 cell production of MIP-1 βb	CD4 cell proliferative response to HIV antigensc	HIV-specific IFN-γ in PBMC: ELISPOT responses to HIV antigensd
		SF33	SF2			Year 1	Year 2
A	0	0	NA	39	Neg	–	–
B	0	53	94	12	Neg	gag	–
C	0	0	NA	9	Neg	gag	(–)
D	1:5	71	NA	16	3 Pos	pol, env	–
E	0	0	86	3	Neg	–	NA
F	0	73	82	7	1 Pos	(–)	pol
G	0	59	NA	13	NA	–	NA
H	0	40	NA	17	3 Pos + p24	(–)	pol
I	0	35	NA	NA	Neg	–	–
J	0	13	NA	1	Neg	–	(–)
K	0	60	62	5	Neg	env	NA
L	0	0	NA	2	Neg	–	–
M	0	0	NA	7	Neg	–	(–)
N	1:20	60	91	5	Neg	–	NA
O	0	0	NA	5	3 Pos + p25	(–)	NA
P	0	60	NA	NA	Neg	nef	NA
Q	0	0	NA	36	Neg	–	NA
Overall, no. positive/no. tested	2/17	7/17	4/5	4/15	4/16	7/17	7/17

NOTE. Bold type indicates a positive response. IFN-γ, interferon-γ; MIP-1β, macrophage inflammatory protein–1β; NA, not available; neg, negative; PBMC, peripheral blood mononuclear cells; pos, positive.

^aData in bold italic type indicate percentage suppression of replication above cutoff of 45% for strain SF33 and 80% for strain SF2.

^bMedian of subject's values; bold italics indicate median exceeds 15%, maximum in control subjects.

^cHIV antigens used: p14, p25, p66, gp120, gp160-LAI, and gp160MN. Data are no. of positive antigens.

^dAntigens for which response ≥3 × maximum in time-matched control subjects. Underline indicates repeat assay confirmed positive result. Parentheses indicate inconclusive repeat assays (1 negative and 1 positive).

Mucosal antibodies

Quantitative ELISA for IgA and for IgG antibodies from vaginal lavage specimens were available for 7 HIV-positive control women and 14 EU women, and 3 HIV-positive control subjects and 2 EU women had detectable antigen-specific IgA. Saliva samples from 3 of 7 HIV-positive control subjects and 2 of 14 EU women were positive for HIV-specific IgA. The saliva from 1 EU woman had a moderately high level of IgA antibodies by ELISA, but this was not confirmed by WBC. None of the women had a strong HIV-specific IgA response in saliva or vaginal washings, regardless of HIV status. Antigen-specific IgG levels were detectable in vaginal lavage specimens from 4 of 7 HIV-positive control subjects and from none of the 14 EU women. Antigen-specific IgG was detectable in saliva from 7 HIV-positive control subjects but was absent in saliva from all 14 EU women tested.

CD8 cell noncytotoxic suppression of viral replication in CD4 cells

CD8 cells from 17 EU women and 20 unexposed control subjects were tested for CD8 cell suppression of HIV-1 replication in acutely infected CD4 cells. The percent suppression of replication of strain SF33 at a CD8:CD4 cell ratio of 2:1 is shown in table 3. Suppression from 7 EU women exceeded a cutoff value of 45%, the maximum among control subjects. Percentages of suppression among the EU women were significantly higher than percentages from unexposed control subjects (P = .014, rank sum test). CD8 cells from 5 EU women also were tested for suppression of the SF2 strain; 4 showed significant suppression (≥80%), including 1 woman who did not suppress the SF33 strain. In all, 8 (47%) of the 17 EU women had significant CD8 cell noncytotoxic suppressive activity to ≥1 virus strain (table 3). Some suppressive activity results from women A–O, presented here individually, were included in aggregate with other cohorts in a report by Stranford et al. [12].

Chemokine production

CD8 cells from 15 EU women, 16 unexposed control subjects, and 9 HIV-positive male partners were analyzed for the production of MIP-1β chemokine after stimulation by PMA/ionomycin. By using the median value of replicates for 11 subjects with repeat samples, the percentage of cells producing MIP-1β was higher in cells from HIV-exposed persons than from control subjects (median, 7.21% [range, 1.10%–39.23%] vs. 3.36% [range, 0.02%–15.17%]; P = .014, rank sum test) and was similar to the percentage of MIP-1β–producing cells from HIV-positive partners (median, 14.6%; range, 0.96%–34.2%). Four EU women and 3 partners had median percentages that exceeded the maximum value observed among unexposed control subjects (table 3).

Single-cell cytokine production

The distributions of IL-2, IL-4, IL-10, TNF-α, TNF-β, and IFN-γ levels among 12 EU women were no different than those among 28 unexposed uninfected control subjects (data not shown) and provided no evidence of increased Th1 or Th2 type cytokine production in response to nonspecific stimuli.

CD4 lymphocyte proliferative response

Lymphocytes from all 13 EU women tested responded strongly to ≥1 recall microbial antigens with a median SI of 205 (mean, 234; range, 28–612). Lymphocytes from 4 women responded to ≥1 HIV antigen with an SI >5 (table 3): 1 subject responded to 3 env antigens and p24, 1 to 3 env antigens, and 1 to a single env antigen. The fourth woman responded strongly to p25 and 3 other antigens, including rgp160 (IIIB and MN), a response that was confirmed by retest 10 weeks later: the mean SI to gp160 was 93 at the first measurement and 35 at the second. The median SI for the responses to HIV antigens in these 4 women was 24 (range, 8.7–159). Responses to recall antigens always were larger than responses to HIV antigens, but the magnitude of responses to microbial antigens was not correlated with responses to HIV antigens. Lymphocytes from unexposed control subjects responded to HIV antigens with an SI <3.

PBMC production of HIV-specific IFN-γ: ELISPOT assay

PBMC from 17 EU women, 8 HIV-positive male partners, and 11 low-risk control subjects were assayed by ELISPOT for response to 4 HIV-1 antigens (pol, nef, gag, and env) determined by numbers of sfc/10⁶ PBMC. The range of responses was 0–610 sfc in the EU women, 0–875 sfc in HIV-positive partners, and 0–50 sfc in low-risk control subjects. A result was considered to be positive if it was ≥3 times the maximum observed among control subjects for the same antigen. When possible, positive results were confirmed by repeat testing on a replicate sample from the same date. Samples drawn in 2 consecutive years were available from 10 EU women. Seven EU women (41%) had positive ELISPOT results for ≥1 HIV antigen at one time point (table 3). The positive results of 4 women were confirmed by repeat assay. Three other EU women for whom no replicate sample was available also had positive results; 1 (subject P) had a subsequent positive result 5 months later.

Four of the 8 HIV-positive male partners each had a positive result (table 2), and PBMC from all 4 recognized HIV pol. One HIV-positive man had a particularly strong ELISPOT response (>800 sfc/10⁶ PBMC) directed toward HIV pol.

In summary, 13 of 17 EU women exhibited ≥1 strong immune response among CD8 cell suppression of HIV-1, CD8 cell production of MIP-1β, CD4 cell proliferative response to HIV antigens, or HIV-specific ELISPOT response (table 3). Six women had ≥2 immune responses. No response was related to time since the woman's last exposure. CD8 cell suppression of HIV-1 and ELISPOT responses each occurred in nearly half the women.

Correlation of Responses

The immune responses in this study tended to occur independently. However, ELISPOT positivity corresponded to higher percentages of noncytotoxic suppression of replication of the SF33 strain: median suppression of 60% (range, 0%–73%) among those with a positive ELISPOT versus median suppression of 0% (range, 0%–60%) among those without a positive ELISPOT (rank sum test, P = .02). The proportion of women who were ELISPOT positive was greater among those classified as positive for suppression of SF33 (5 [71%] of 7) than among those who were not SF33 suppressors (2/10), although this difference did not reach statistical significance (P = .058).

Correlations between factors in HIV-positive partners and immune responses in the EU women also were explored. Pairwise associations of the woman's positivity on neutralizing antibody, CD8 cell suppression, production of MIP-1β, CD4 proliferative response, and positive ELISPOT responses with the male partner's antiretroviral therapy at enrollment, HIV RNA load in plasma and semen (low [≤1500 copies/mL] vs. high [>1500 copies/mL]), median CD8 cell count (<1000 cells/mm³ vs. ≥1000 cells/mm³), CD8 cell suppression, and positive ELISPOT response were evaluated. None had significant associations by Fisher's exact test. These tests were limited in scope and had low statistical power because of the few couples with complete data. Semen virus loads >1500 HIV RNA copies/mL were observed in partners of 1 of 4 women with cell suppressive activity, compared with 4 of 5 women who showed no suppressive activity. The partners of the 3 women with positive MIP-1β production all had median CD8 cell counts of >1000 cells/mm³, compared with 4 (57%) of 7 partners of women with no such CD8 cell production. Although such differences might suggest a trend, they could readily occur by chance alone (P = .20 and P = .48, respectively). Time gaps between partners' tests also undermine the interpretation of apparent associations.

Discussion

In this comprehensive study of EU heterosexual partners of HIV-positive men, CD8 cell activity appeared to be the dominant factor in nontransmission. The results indicate that characteristics of the donor men and the potential recipient women both contribute to nontransmission. We had 3 major findings: first, the probable contribution of the potential HIV-positive donor to nontransmission, as suggested by their very high CD8 cell counts combined with the evidence of functional (noncytotoxic) HIV suppression by the CD8 cell population; second, the demonstration in 11 of 17 seronegative women of CD8 cell–mediated HIV suppression, either as noncytotoxic suppression or β-chemokine inhibition; and third, the presence of HIV-1–specific IFN-γ production, as exhibited by positive ELISPOT assay of PBMC, in nearly half the women.

Although we studied few men for CD8 cell counts, our previous studies of a large cohort showed that higher CD8 cell counts characterized the nontransmitting infected partners in HIV-discordant couples [29]. Among the men without AIDS, their CD8 cell counts were extraordinarily high. As expected, CD8 cell counts fell as AIDS developed; the men with AIDS could have had high CD8 cell counts during the exposure period during which unprotected sex occurred frequently. The robust CD8 cell response was accompanied by clear evidence of HIV suppression both by noncytotoxic activity and β-chemokines. Since the noncytotoxic CD8 cell activity was determined at a fixed CD8:CD4 ratio, it seems reasonable to assume that HIV-positive men with extremely high CD8 cell counts would manifest greater overall viral suppressive activity than HIV-positive men with fewer CD8 cells. Our previous study [25] showed that the vigorous CD8 cell response, the major variable characterizing nontransmission, appeared to act only in part by reduction of circulating virus load. The very high CD8 cell counts could reduce transmission potential in several ways—by reduction of virus titers in semen, by alterations in the virus that modify its infectivity (making it more immunogenic), or by induction of a substance or substances that, together with the virus, are transferred during sexual intercourse and induce host defenses.

In all, 13 women exhibited an immune response that could at least partly explain their persistent seronegativity. The predominant induced defenses in the women were CD8 cell–mediated noncytotoxic suppression, β-chemokine (MIP-1β) production, and ELISPOT activity. High CD8 cell noncytotoxic activity is associated with reduced susceptibility of PBMC to infection [12]. Despite more than a decade of study, the nature of noncytotoxic, non–β-chemokine suppression remains undefined. The CD8 cell suppressive activity of the HIV-negative women was not accompanied by increased numbers of CD8 cells, except in one woman. However, ELISPOT responses showed a tendency to correlate with suppressive activity, at least of one viral strain. Because of the small sample size and lack of opportunity to confirm some of the positive ELISPOT responses, a relationship between the 2 responses must be viewed as tentative. A statistical correlation may reflect general strength of CD8 cell response underlying both the suppression activity and the IFN-γ response detected by ELISPOT, whereas the 2 activities nevertheless function separately. The responses otherwise appeared to occur independently of one another. Of note, the time since the last exposure was not related to any of the women's responses.

The positive ELISPOT assay results on antigen-reactive T cells in EU women are intriguing. CTL activity was not found in 15 women by standard bulk in vitro stimulation assay [39] against the same 4 HIV antigens nor in 14 women by limiting dilution assay (LDA) at one highly experienced laboratory [12]. However, nearly half the HIV-negative women showed HIV-1–specific IFN-γ–producing cells in the ELISPOT assay, a presumably direct and more sensitive measure of antigen-reactive T cells than traditional LDA of CTLs [38, 40, 41].

Some recent studies have systematically compared LDA and ELISPOT in viral systems [31, 38, 40, 42]. In these reports, the estimates of antigen-specific CD8 T cell frequency obtained by ELISPOT were greater than those obtained by LDA by 2–20-fold or more. This was shown both for “effector” cells during acute infection [40] and for “memory” cells persisting months or years after infection [40, 42]. Thus, LDA may underestimate the true frequency of antigen-specific cells by 10-fold or greater. In addition, it is believed that these 2 assays detect overlapping, but not identical, T cell populations. LDA measures only CTL precursors with the capacity to divide and proliferate after antigenic stimulation for 1–2 weeks, whereas ELISPOT measures populations that are capable of secreting cytokine after short-term antigenic stimulation but that may not be capable of surviving and proliferating in the context of an LDA assay. With these considerations, it is tenable and not contradictory that some EU women have antigen-specific T cells detectable by ELISPOT but not by LDA.

Several recent reports identified HIV-specific CTLs in repeatedly exposed uninfected persons [5, 43]. In general, these works relied on in vitro restimulation of PBMC in the presence of HIV protein antigens or peptides. Under these conditions, it is difficult to rule out the possibility of in vitro priming. In contrast, direct demonstration of IFN-γ production in ELISPOT assay suggests that antigen-specific CD8 cells are present in the peripheral blood of these exposed uninfected persons. A report [5] of peptide-specific CTL assays of PBMC from commercial sex workers in Nairobi noted that parallel ELISPOT assays detected activity in previously negative or equivocal CTL cultures. Our results suggest that the ELISPOT assay is a more sensitive measurement of antigen-reactive CD8 cells than traditional bulk ⁵¹Cr release assays and should be included in vaccine evaluation.

The anti-HIV responses in women were not limited to CD8 cell suppression of viral replication and presumed CTL activities, but, overall, the other responses were seen in fewer of the women. A prompt CD4 cell proliferative response is important in control of viremia [44]. Four women showed such a response to ≥1 HIV antigens, including 1 who showed no CD8 cell suppressive activity, but responses in 3 of the women were weak. Weak CD4 cell proliferative responses in highly exposed negative persons were reported in one study [45] but not in another [44]. One woman repeatedly had a very strong CD4 cell proliferative response to multiple HIV antigens. Her pattern is unique in this small cohort. Her ELISPOT assay result was inconclusive, and she exhibited no other potential defense mechanisms. The remote possibility remains that she is infected despite being negative by cocultivation and antibody negative for the 12 years since her last exposure. In that case, as seen in some long-term HIV-positive nonprogressors, her strong CD4 cell proliferative responses could result presumably from constant HIV stimulation by small numbers of virus in some sequestered site.

Only one EU woman had a moderate titer of neutralizing antibody in the absence of any evidence of infection. It is uncertain whether relatively small amounts of neutralizing antibody would represent an effective defense mechanism [46]; however, we found no evidence that neutralizing antibody played a significant protective role in this cohort.

One woman, who was originally the 18th member of the study cohort, seroconverted after 9 years, during which she continued to have frequent unprotected sex with her HIV-positive partner. She was 1 of 5 women with no potentially protective immune responses. The estimated date of seroconversion was between August 1997 and June 1998, a time when a strong neutralizing antibody response (1:80) appeared, and subsequent cocultivation studies were positive. In December 1998, she had a strong ELISPOT response, although 10 months later her ELISPOT assay result was negative. It is possible that the earlier ELISPOT response represents an acute reaction to viral infection that was proceeding at that time.

In this study, mucosal antibodies did not appear to be a factor in nontransmission. Because 2 independent techniques (ELISA and WBC) and a large collection of external secretions from control subjects positive for IgG and IgA HIV antibodies were used in the present study, we conclude that the vaginal washes from these EU women did not contain HIV-specific IgG or IgA antibodies. As illustrated in the current study, especially in parotid saliva, results generated by ELISA may require further verification and should be interpreted cautiously [30, 47].

Dorrell et al. [48] also found no mucosal anti-HIV antibodies, but others [3, 22, 30] detected antibodies to HIV antigens in external secretions, including saliva, semen, urine, and vaginal washings. Mucosal IgA with the capacity to neutralize HIV-1 primary isolates was found in apparently HIV-1–resistant cohorts of commercial sex workers in Nairobi and Italian discordant couples [3, 49, 50]. The discrepancy between these studies and ours could be related to differences in methodologies [22, 51] or in the populations studied. CD8 T cells from the cervical mucosa of a subgroup of resistant commercial sex workers have exhibited HIV-specific responses measured by ELISPOT as IFN-γ release [22]. These responses were comparable to blood responses in the same person.

The lack of transmission cannot be ascribed to reduction in CD4 cell infectivity. The CD4 cells of 9 women all were readily infected, 5 by their partner's virus. None of the women was homozygous for CCR5Δ32 or CCR2 promoter region mutations that disable receptors for HIV.

Nelson et al. [21] did not find a difference in CD8 cell counts between transmitting and nontransmitting men in a study of heterosexual transmission in Thailand, in contrast to our finding that nontransmitting men had very high CD8 cell counts with HIV suppressive activity. No data on functional CD8 cell behavior in either the male or female partners were reported for the Thai study. Our study shows that such functional analyses are warranted; nearly half the HIV-negative women showed CD8 cell virus-suppressive activities, but only one had a high CD8 cell count.

We believe that this study, supported by other data, makes a strong case for the primacy of the CD8 cell response in reducing the risk of HIV transmission. Vaccines are being assessed primarily by induction of CTLs measured indirectly and of neutralizing antibody. Our data suggest the need for broader immunologic measurements in vaccine evaluation, particularly CD8 cell noncytotoxic HIV suppression, β-chemokine production, and ELISPOT assay, in addition to traditional CTL analysis. Clearly, EU persons have multiple potentially effective defense mechanisms. It seems likely that vaccines that stimulate as many of those immune defenses as possible will be more effective in halting this ferocious pandemic.