Detection of HIV-1-Specific CTLs in the Semen of HIV-Infected Individuals¹

The World Health Organization estimates that >70% of HIV infections are established as a result of sexual contact (1). HIV in semen is found in both cell-associated and cell-free forms (2), and HIV variants differ from those found in blood, which suggests viral compartmentalization (3). The highest viral loads in semen are associated with peripheral blood CD4 counts of <200/μl (4, 5), concomitant sexually transmitted infections (6), and asymptomatic genital tract inflammation (4, 5). Conversely, semen virus levels decrease with antiviral therapy and with treatment of sexually transmitted diseases (4, 6, 7).³ Since intracellular virus is most readily eliminated by CTLs, the presence of anti-HIV CTLs in urogenital mucosa would be predicted to decrease viral load in genital secretions, contain virus locally, and decrease the chance of transmission to sexual partners.

Vigorous memory CTL responses against all the major HIV proteins have been demonstrated in the blood of infected individuals (8, 9, 10, 11). HIV-specific CTLs have also been isolated from various sites including the lymph node (12), spleen (12, 13), cerebrospinal fluid (14), and lung (15). Recently, SIV-specific CTLs were isolated from the vaginal mucosa of chronically infected macaques inoculated with SIV by the vaginal route (16), and HIV-specific CTLs were isolated from the cervical mucosa of infected women (17) indicating that, at least in the female, an antiviral response can be generated and maintained locally in the genital tract mucosa. We have recently demonstrated that the semen of HIV-infected men contains functional T lymphocytes, and that the majority of these cells are CD8⁺ and express a marker of cytolytic activity (18). In this study, we sought to determine whether T lymphocytes from the semen of seropositive men also had anti-HIV cytotoxic activity.

Five seropositive men were recruited from Fenway Community Health Center and Massachusetts General Hospital to provide semen and blood samples. Informed consent was obtained from all individuals. Eligibility criteria included a previous leukocytic semen specimen (>5 × 10⁴ viable leukocytes per ejaculate) and a CD4 count of >500/μl. The clinical data for each individual are summarized in Table I.

Semen was obtained by masturbation into sterile specimen cups containing 10 ml of RPMI 1640. Blood was collected in EDTA-treated vacutainers. Viable semen round cells (SRCs)⁴ and PBMCs were isolated on density gradient separation medium (Ficoll-Hypaque, Pharmacia, Piscataway, NJ and washed three times in HBSS containing 100 μg/ml gentamicin and 1 μg/ml amphotericin B before resuspension in RPMI supplemented with 2 mM l-glutamine, 50 μg/ml gentamicin, 0.25 μg/ml amphotericin B, and 10% FCS (all obtained from Life Technologies, Grand Island, NY) (cRPMI). Since the isolated SRC fraction contains a large proportion of immature germ cells, SRCs (10% of sample) and PBMCs were applied to 8-spot slides, allowed to air dry, fixed in acetone, and later immunostained with Abs to CD3 and CD8 to enumerate exactly the number of T cells (18). Using published data obtained from leukocytic HIV⁺ individuals, we ascribed arbitrary values to the proportion of CD3 cells in each SRC population for immediate cloning purposes and adjusted the values after staining (18). Arbitrary values were 1% for SRCs and 70% for PBMCs. Cells were seeded in 96-well plates at numbers shown previously to approximate limiting dilution (10–300 cells/well) in 100-μl volumes containing 1.5 × 10⁶ irradiated (5000 rad) allogeneic PBMCs/ml, 1 μg/ml PHA, and 50 U/ml IL-2 (Boehringer Mannheim, Indianapolis, IN) in cRPMI. After 2 to 3 wk, wells exhibiting growth were restimulated with irradiated PBMCs, PHA, and IL-2.

T cells were enumerated by a standard indirect immunohistology technique that has been described previously in detail (18). Primary Abs recognizing CD3 and CD8 Ags (Dako, Carpinteria, CA) were used, and Ab-positive cells were visualized using an alkaline phosphatase/anti-alkaline phosphatase kit (Dako). A total of 200 cells were counted per Ab, and the number of positive cells in the semen sample was calculated from the SRC count. Clones were phenotyped using similar methodology, with Abs recognizing the following Ags: CD3, CD8, CD4, CD57 (Dako), TCRαβ (BF1, a kind gift of Dr. Michael Brenner, Department of Rheumatology, Brigham and Women’s Hospital), and TIA-1 (a kind gift of Dr. P. Johnson, Dana-Farber Cancer Institute, Boston, MA) (19).

Before cloning, BLCLs were established by transformation of peripheral blood B cells by EBV obtained from the supernatant of the B95-8 cell line (8).

Recombinant vaccinia viruses constructed from the HIV IIIB isolate and expressing the following proteins were used in this study: env, gag, and pol (Vabt 408); gag (Vabt 401); p17 (Vabt 228); p24 (Vabt 286) (kind gifts of Therion Biologics, Cambridge, MA); env (PE16); and pol (CF21) (National Institutes of Health AIDS Repository, Rockville, MD). Wild-type vaccinia virus was used as a control (NYCBH, Therion Biologics). Synthetic p24 HIV peptides (12–21 aa) were synthesized as described previously (20). Lyophilized peptides were resuspended at 2 mg/ml in 10% DMSO.

Cloned T cells were tested for HIV-specific cytolytic activity in a standard chromium release assay (8). Target cells were autologous BLCLs pulsed with HIV peptides or infected with vaccinia constructs expressing HIV proteins (14). Partially MHC-mismatched allogeneic BLCLs were also used in MHC restriction experiments. In brief, BLCLs to be vaccinia infected were incubated with 3 to 4 multiplicities of infection per cell of recombinant virus for 16 to 18 h at 37°C. Cells were washed twice, labeled with 150 μCi Na₂⁵¹Cr for 2 to 3 h, and washed four times. The sensitization of BLCLs with peptides was achieved by incubating cells with 10 μg/ml of peptide with the chromium. Clones and labeled BLCLs were incubated together in various ratios in 150-μl volumes in 96-well plates at 37°C for 4 to 6 h. Supernatants (100 μl) were subsequently harvested and counted in a Cobra gamma counter (Packard Instrument, Meriden, CT). The percentage of specific lysis was calculated from the following formula: 100 × ([experimental release − spontaneous release]/[maximum release − spontaneous release]). Maximum release values were obtained by the lysis of targets with 5% Triton X-100. Assays were only evaluated if spontaneous release was <30% of maximum release. Clones were considered positive if lysis of the targets was at least three times greater than the lysis of the control vaccinia target and if the HIV-1-specific lysis was >10% (14, 21).

Our previous studies have shown that viable CD3⁺ cell counts in HIV⁺ men range from undetectable to >2.2 × 10⁵ per ejaculate, with a median value of 9.0 × 10³ in men taking antiretroviral medication and 2.5 × 10⁴ in men who are not on therapy (18). Since the number of cells recovered from semen is relatively small and the estimated frequency of HIV-specific T cells in asymptomatic individuals is estimated to range from 1/10² to 10⁴ in PBMCs (11, 22, 23), we optimized the conditions of this study by recruiting men who had previously had a leukocytic semen sample (defined as >5 × 10⁴ viable mononuclear leukocytes per ejaculate). A previous study from our laboratory indicates that this is not a major selection step, since 83% of HIV⁺ men with peripheral blood CD4 counts of >200/μl that are not on antiretroviral therapy are included in this category (18). As shown in Table I, the median CD4 count of the patients was 932, and the median length of infection was 12 yr. None of the individuals were on antiretroviral therapy, and all were asymptomatic, with the exception of one man who had recently developed Kaposi’s lesions.

The numbers of viable round cells recovered from the semen samples used in this study ranged from 2.5 to 34.0 × 10⁵ cells per ejaculate (Table II). Immunostaining revealed that between 1.1% and 14.3% of round cells (median 2.5%) were CD3⁺, with CD8⁺ lymphocytes contributing from 33 to 95% of the T cell population. Cells were successfully cloned from semen at a limiting dilution with PHA and IL-2, but cloning efficiencies were considerably lower than those derived from blood (Fig. 1). Fungal contamination was also observed in some semen-derived cultures, but these wells were not taken into account when calculating cloning efficiencies. Wells exhibiting proliferation were restimulated, and those which were successfully expanded were screened for anti-HIV cytolytic activity using autologous BLCLs infected with a trivalent vaccinia vector expressing env, gag, and pol proteins. HIV-specific cytolytic activity was found in all five donors, with multiple lines isolated from each donor (Table II).

To precisely define the HIV cytolytic responses in semen, lines from three individuals were expanded and subsequently tested for activity against the individual gag, pol, and env proteins (Table III). Activity against all three proteins was noted in each donor, indicating that CTLs derived from the urogenital tract have a broad-based activity to HIV. With the exception of two lines, all lines had activity against a single protein. In donor 9320, the predominant protein recognized was pol (43% of all specific activity), followed by gag (36%) and envelope (21%). In donors POGO and 161J, the predominant protein recognized by CTLs was gag (67 and 57% of specific activities, respectively). In one individual (9320), we also performed a parallel analysis on a small number of lines derived from the peripheral blood. Although pol-specific activity dominated in the semen of this individual, only one of four blood lines was pol-specific.

Two of the gag-specific lines from donor POGO were examined in greater depth for fine epitope specificity, phenotype, and MHC restriction. We first used vaccinia constructs expressing the p17 and p24 regions of gag to demonstrate that both gag-specific lines recognized epitopes within the p24 protein (Fig. 2). When a series of nested 21-mer peptides covering the p24 sequence in pools of four (data not shown) were used followed by individual peptides, data indicated that both lines recognized the gag peptide representing aa 303 to 324. Analysis with shorter (12-mer) peptides indicated that both recognized the aa sequence 307 to 318. Using BLCLs partially matched at MHC class I loci, we were also able to determine that both of these lines were restricted by HLA B44 (Fig. 3). Phenotyping studies indicated that 100% of the cells in both lines stained positively for TCRαβ, CD8, and TIA-1.

In this study, we describe a novel method for investigating CTL activity in the male urogenital tract using semen as a source of locally derived T cells. We reported previously that the semen of asymptomatic, seropositive men not taking antiretroviral therapy contains viable, functional T cells that are predominantly of the CD8⁺ phenotype and express markers of mucosal derivation and cytolytic activity (18). Seminal T cells may originate from many sites within the male genital tract including the rete testis, epididymis, prostate, vas deferens, and urethra (24). Since the potential antiviral cytolytic activity of male urogenital tract cells has not been investigated, we optimized study conditions by recruiting individuals who had peripheral blood CD4 counts of >500/μl and a previous leukocytic ejaculate. We found substantial cytolytic activity in all five individuals examined, although cloning efficiencies were lower than in blood. This may reflect a lower functional capacity of T cells recovered from semen due to their exposure to highly immunosuppressive seminal plasma (25) or may reflect innate differences between T cells derived from a mucosal vs a peripheral location. For example, intraepithelial lymphocytes (IELs) derived from normal gastrointestinal mucosa demonstrate a considerably lower proliferative capacity than their blood counterparts (26), a phenotypically distinct profile (27), and an oligoclonal T cell repertoire (28, 29), and may include unique regulatory subsets of T cells (30).

The presence of HIV-specific CTLs in a mucosal site was first demonstrated with T cells isolated from bronchoalveolar lavages (15). The first isolation of HIV- or SIV-specific CTLs from the genital tract was by Lohman et al. (16). Macaques were vaginally infected with SIV, acutely and chronically infected animals were sacrificed, and vaginal IELs were extracted, expanded, and tested for CTL activity. IELs were characteristically CD8⁺, TCRαβ⁺, and CD2⁺; precursor frequencies were similar to that in macaque blood; and chronically infected monkeys had greater numbers of virus-specific cells than those that were acutely infected. More recently, HIV-specific CTLs were cloned from cervical mononuclear cells that had been extracted from HIV-infected women by cervical cytobrush collection (17). Although cytobrush sampling can induce local bleeding, the IEL-like phenotype of these cells was suggestive of their mucosal derivation. The authors noted HIV-specific activity in the cervix of 63% of the subjects sampled; the highest anti-HIV cervical CTL activity was associated with a peripheral blood CD4 count of >500 cells/μl. A comparison between small numbers of blood and cervix-derived gag-specific lines in one donor indicated that CTLs derived from the two sites recognized common epitopes. We did not undertake a comparison of epitope specificities in the paired lines that we derived from one of our donors, but we did note a predominance of pol specificity in semen that was not seen in the blood of this individual. Clearly, it still remains to be established whether blood and mucosal CTLs exhibit different CTL repertoires, whether these will reflect differences in TCR repertoires, where and how Ag is presented to these cells, or whether distinct HIV variants are being recognized (3).

The isolation of HIV-specific cytotoxic T cells from the female and male urogenital tracts indicates that virus-specific lymphocytes are recruited to the urogenital mucosa of HIV-infected individuals. The presence of these effector cells in the genital tract may be critical in controlling local HIV infection, thus reducing the viral load in mucosal secretions and the chance of transmission to partners during sexual intercourse. A study looking at the correlation of CTL activity with viral load in semen is currently underway. Indications that a local immune response may prevent dissemination of virus come from a study which found that some seronegative women with seropositive partners and a history of multiple unprotected sexual exposures had HIV-specific secretory IgA but not IgG in their vaginal secretions (31). We presently hypothesize that an appropriately immunized individual with a strong, local, antiviral mucosal immune response should be able to eradicate or to locally contain HIV following genital transmission. If partners are concordant for MHC class I molecules, an immunized individual may also be able to specifically kill donor HIV-infected cells derived from the genital secretions of their partner (32, 33). In contrast, MHC discordance may result in a host vs donor immune response, and this allogeneic recognition may result in an alternative mechanism of eradicating virally infected cells. Whether MHC discordance between sexual partners would decrease the chance of virus transmission or concordance would increase sexual transmission, as has been recently shown in a perinatal transmission study, has yet to be investigated (34).

Clearly it would be advantageous to develop an anti-HIV vaccine that would elicit HIV-specific mucosal CTLs. Although some studies indicate that immunization at a mucosal site is not a prerequisite for induction of a mucosal immune response (35, 36), there is a substantial amount of evidence to indicate that mucosal immunization strategies are more effective at generating a local response (37). In addition, long-lived antiviral CTLs can be detected in mucosal tissues after mucosal but not systemic immunization, suggesting that the maintenance of memory cells is also dependent upon the chosen route of immunization (38). Of further note is the observation that inoculation at proximal mucosal sites appears to be more effective at inducing measurable immune responses than inoculation at distant mucosal sites. For example, recent studies have indicated the effectiveness of a vaginal or rectal immunization protocol in inducing a genital tract immune response (39, 40, 41). Consequently, we plan to address in our future studies whether antiviral CTL responses can be induced in the male urogenital tract by vaccination and, if so, which routes of immunization are the most effective at inducing and maintaining a local immune response.

We thank the individuals who donated samples for this study. We also thank Dr. Eric Rosenberg at Massachusetts General Hospital and Robert Goldstein and Judy Erdman at Fenway Community Health Center for recruiting patients, Lidiya Shver and Lynne Tucker for excellent technical assistance, and Dr. Gail Mazzara at Therion Biologics for providing recombinant vaccinia viruses.