Antiretroviral-Drug Concentrations in Semen: Implications for Sexual Transmission of Human Immunodeficiency Virus Type 1

The majority of human immunodeficiency virus type 1 (HIV-1) infections result from seminal transmission (17, 104). The size of the viral innoculum in seminal fluid is likely the major determinant of transmissibility. Current HIV prevention strategies include comprehensive population interventions promoting condom use, early diagnosis and treatment of sexually transmitted diseases, and education programs to decrease the rate of sexual-partner change and other high-risk sexual behavior (17, 18, 40, 76, 115). However, antiretroviral therapy targeted to reduce viral shedding in genital secretions may also assist in preventing transmission (17, 86).

Although the HIV-1 pandemic continues to be driven by transmission of virus in genital fluids, little is known about the pharmacokinetics and pharmacodynamics of antiretroviral drugs in the genital tract (46). Similarly to the effect of antiretroviral therapy on viral burden in blood and lymphoid tissue (6, 14, 41, 42, 120, 132), potent combinations of HIV-1 reverse transcriptase and protease inhibitors which concentrate in semen may lead to marked viral suppression in this compartment (2, 21, 30, 37, 38, 41–43, 58–60, 84, 117, 118, 124, 138). This article reviews the biology of HIV-1 in the male genital tract, the impact of antiretroviral therapy on seminal shedding, and the emergence of drug-resistant variants. Major determinants of drug distribution into semen are discussed, and strategies for using this information to predict antiretroviral concentrations in this compartment are suggested. Finally, a discussion of animal models critical to investigating correlations between seminal antiretroviral concentrations and viral transmission is included.

BIOLOGY OF HIV-1 IN THE MALE GENITAL TRACT

A separate compartment of HIV replication.

The concentration of HIV-1 in blood offers the best prediction of the concentration in semen (19, 126). However, generally HIV-1 is detected in semen less frequently, and in lower concentration, than in blood (19, 26, 38, 43, 123–125). In addition, comparisons of viral genotypes and phenotypes in semen and blood indicate that the male genital tract constitutes a relatively distinct compartment for viral replication (12, 31, 96, 126, 128, 140).

The existence of distinct viral populations in systemic lymphoid and genital tissues may be due to differences between the major cell types sustaining viral replication or due to separate immunologic controls within the compartments. Accordingly, the dynamics and magnitude of specific antiretroviral-drug effects on HIV-1 shedding in semen are not likely to be equivalent to those in blood and lymphoid tissue, even if comparable local drug concentrations are achieved. Comparisons of blood and brain-derived viral pol gene sequences provide evidence that HIV-1 evolution with or without pharmacological selection pressure can proceed at different rates in separate anatomic compartments (131).

A possible reservoir of long-lived virus.

As with the central nervous system and other tissues rich in cells derived from monocyte/macrophage lineage, long-lived infrequently replicating cells in the male genital tract could constitute a viral sanctuary for HIV-1 to escape both host immunity and antiretroviral therapies. Such tissue reservoirs are especially important if antiviral-drug regimens achieve only partially suppressive concentrations in infected genital cells which are capable of productive infection.

Cellular origins of HIV in the genital tract and role of cell-associated versus cell-free virus in mucosal transmission.

In the majority of cases, envelope sequences of viral quasispecies in blood from patients with primary HIV-1 infection suggest that sexual transmission selects for the macrophage-tropic, or R5, virus (10), which preferentially uses the CCR-5 coreceptor (139, 140). Preferential transmission of R5 virus may be partly explained by a predominance of such variants in semen in infected men (96, 141). However, the limited data are inconsistent (23, 126).

Analysis of the cellular fraction of semen (101, 130) and immunocytochemical and in situ molecular investigations of human and nonhuman primate tissues suggest that cells of the macrophage lineage may be an important source of virus in the male genital tract (79). Recent studies confirm that macrophage and dendritic cells are able to support high-level HIV-1 replication under certain conditions, especially in the presence of immune stimulation (93, 98). If HIV-1 in the genital tract can originate from relatively long-lived macrophages, the dynamics of local viral production and clearance will differ significantly from those in systemic lymphoid tissues. This may influence both short- and long-term effects of antiviral regimens in this compartment.

HIV-1 is present in semen as both host cell-associated virus and cell-free virus. The relative importance of each in mucosal transmission remains unclear. While cell-associated virus can be cultured in vitro in up to 55% of semen samples by optimized peripheral blood mononuclear cell (PBMC) coculture assays (127), cell-free virus in seminal plasma is isolated in only 10% of cases or fewer by the same assays (85). In vitro systems, however, differ significantly from the mucosal milieu where sexual transmission occurs, and seminal plasma components may be toxic to donor PBMCs (127). Some in vitro models suggest a critical role for cell-to-cell transmission of HIV-1 (94, 117). Conversely, simian immunodeficiency virus (SIV) or chimeric simian-human immunodeficiency virus (SHIV) is most readily transmitted by vaginal inoculation of cell-free virus (69, 77, 81). The efficiency of in vivo SIV transmission (77, 81) as well as in vitro HIV-1 infection of both dividing and nondividing cells (137), is enhanced by human seminal plasma. Comparison of viral envelope sequences amplified from blood compartments of recently infected gay male patients and from the semen samples of their transmitting sexual partners suggests that both cell-associated and cell-free virus can be transmitted by anal intercourse (141).

Disease stage and seminal shedding of HIV-1.

HIV-1 can be detected in semen during and soon after the primary seroconversion illness (27, 121), indicating that the male genital tract is an early site of viral dissemination and replication after infection by sexual exposure. During the time when HIV-1 replication is unchecked by host immune responses, newly infected patients may transmit a viral population which is well adapted for sexual transfer. Mathematical modeling of the North American epidemic suggests that this mechanism may have been responsible for the initial exponential increase in new cases of HIV-1 infection (50). Therefore, antiretroviral therapies initiated during primary infection (13, 48, 58, 72) may reduce viral shedding in genital secretions, reduce the likelihood of further transmission, and potentially modify epidemic spread within high-risk communities.

Seminal virus is frequently detectable during the period of clinical latency, when CD4⁺ cells are well preserved (19, 26, 68, 123, 126, 133). Recent studies have also shown larger amounts of HIV-1 shed in semen in patients with advanced disease and low CD4⁺ T cell counts (19, 123, 133), likely reflecting higher blood viral burdens. Therefore, if antiviral drugs are to have a beneficial impact on the HIV epidemic, potent and tolerable regimens that penetrate the genital tract and inhibit virus production by genital cells should be delivered throughout the course of disease.

Innoculum size and transmission of HIV-1.

As with other sexually transmitted diseases, the probability of HIV-1 transmission may depend largely on the level of inoculum at the mucosal surface. This evidence comes from both nonhuman primate models of retrovirus infection (81) and clinical investigations correlating levels of HIV-1 in blood with the probability of viral transmission by parenteral (11, 64), vertical (35, 113), and sexual routes (33). It is not yet known whether a critical level of HIV-1 in genital secretions is required for sexual transmission: such information may be obtained through animal models, or by careful prospective study of HIV-1-discordant partners, and would assist in clarifying the likely effects of antiviral drugs and vaccines on viral transmission. Similarly to vertical transmission (82, 113) there may be no finite viral-load threshold below which sexual transmission does not occur.

ANTIRETROVIRAL DRUGS AND THE GENITAL TRACT

Composition of human ejaculate and mechanisms of drug penetration.

Human ejaculate is primarily composed of secretions from testes (10% of ejaculate volume), seminal vesicles (50 to 70% of ejaculate volume), and prostate (20 to 30% of ejaculate volume). Secretions from the urethral and bulbourethral glands, epididymis, and ampullae contribute an additional 10% (70). Although much literature has been devoted to drug effects on spermatozoal physiology and morphology, data on the distribution of drugs into the seminal compartment are scarce and incomplete (3, 95). Exact mechanisms of distribution are currently unknown, and conflicting data exist (129).

Although genital tract transporters such as P glycoprotein (an ATP-dependent, efflux membrane transporter located throughout the body) cannot be discounted as influencing seminal concentrations of antiretroviral agents (114), their specific effects are currently unknown. However, if it is assumed that the boundary between the systemic circulation and the seminal compartment behaves as a lipid barrier, most drugs will enter by passive diffusion (67, 71). Three physicochemical properties are likely to regulate the rate at which compounds passively diffuse from blood into semen: the dissociation constant, the partition coefficient, and plasma protein binding.

The dissociation constant (pK_a) defines the pH at which equimolar concentrations of un-ionized (neutral) and ionized forms of a compound exist. Most pharmacological compounds are weak acids or weak bases that are ionized at specific pH values. Weak acids become ionized with increasing pH, and weak bases become ionized with decreasing pH. As un-ionized compounds diffuse more readily across lipid barriers, weaker acids are more likely to cross these barriers when the pH of the fluid (i.e., plasma) is less than the pK_a of the compound. Conversely, weak bases are more likely to cross lipid barriers when present in an environment with a pH greater than its pK_a. Additionally, ion-trapping mechanisms will cause acidic compounds (pK_a < 6) to accumulate in alkaline compartments and basic compounds (pK_a > 8) to accumulate in acidic compartments.

In order to reach seminal compartments, compounds presumably must either fully penetrate lipid barriers or penetrate the lipid membrane of secretory cells for packaging and secretion. The partition coefficient is a measure of the ability of an un-ionized compound to passively diffuse across cell or compartment barriers. The more lipid soluble the compound, the greater the partition coefficient and the greater the compound’s ability to cross these barriers.

Plasma protein binding determines the amount of drug available to cross cell membranes. Generally, the driving force for diffusion of a compound into tissue or body fluids is its concentration in plasma or serum: higher concentrations in plasma yield higher peripheral-compartment concentrations. However, when a significant fraction of a compound is tightly bound to serum or plasma proteins, its concentration in peripheral compartments is determined more accurately by the concentration of unbound drug.

For a lipid-soluble compound, the ratio of drug concentration in any given fluid to that in plasma can be predicted by a modified version of the Henderson-Hasselbach (HH) equation. Winningham et al. (129) used this method to calculate the theoretical steady-state ratios of prostatic fluid (PF) to plasma (P) drug concentrations for a variety of antibacterial agents. Stronger acids were predicted to not accumulate in prostatic fluid (PF:P ∼ 0.2), weaker acids (pK_a > 9) were predicted to approach plasma drug concentrations (PF:P ∼ 1), and stronger bases (pK_a > 9) were predicted to concentrate in prostatic fluid (PF:P ∼ 6) by ion-trapping mechanisms.

Although this equation provides a practical approach to predicting drug concentrations in genital-tract compartments, measured prostatic concentrations of drugs have often, but inconsistently, been lower than predicted (95). Factors contributing to these observations include suboptimal lipid solubility, high degree of ionization at plasma pH, shifts in the compound’s apparent pK_a in complex biological fluids, and incomplete equilibration between plasma and semen before sampling. In addition, investigations into canine prostatic-fluid secretion processes reveal that, during ejaculation, the composition of the fluid formed by each cell can vary during the course of secretion, can vary from cell to cell at any given time during secretion, and can be modified as the fluid passes through glandular tubules and ducts (102, 111). Thus, although drugs may achieve high concentrations in resting secretions, evaluation of ejaculate can confound the measurements of true concentrations in the various compartments of the male genital tract through dilution and other processes.

Selected physicochemical and pharmacokinetic characteristics that impact seminal concentrations of available antiretroviral drugs are shown in Table 1. Weakly basic, lipophilic compounds that have minimal protein binding would be expected to achieve higher concentrations in seminal fluid. Table 2 summarizes the predicted concentrations of representative drugs from the major classes of antiretroviral agents currently approved by the Food and Drug Administration, calculated by the modified HH equation. The predicted overall concentration ratio in seminal plasma is calculated as the sum of each fluid/plasma ratio multiplied by its respective percent contribution to total seminal volume. On the assumption of a lipid barrier, the semen/plasma drug concentration ratio for the majority of compounds is approximately 1, predicting that most will likely achieve concentrations in semen similar to those in plasma. Zidovudine, didanosine (ddI), efavirenz, and nelfinavir are possible exceptions, as they are predicted to achieve threefold higher concentrations in semen than in plasma. However, this equation yields only an estimate of the concentration ratio occurring at a single point in time. The extent of antiretroviral-drug accumulation in the genital tract and the influence of newly formed fluid on antiretroviral concentrations in resting secretions have yet to be fully elucidated.

TABLE 1.

Physicochemical and pharmacokinetic parameters of representative antiretroviral agents potentially influencing seminal-compartment distribution

Drug	pKa	Lipid solubility (partition coefficient)	Time to Cmaxa (h)	Elimination half-life in serum (h)	Protein binding (%)
Nucleoside reverse transcriptase inhibitors
Zidovudine	9.7	Slightly lipophilic (1.15)	0.6–0.8	0.6–1.7	20–38
Lamivudine	4.3	Hydrophilic (NAb)	1–1.5	3–5	10–50
Didanosine	9.1	Slightly lipophilic (NA)	0.5–4.6	0.6–2.9	<5
Zalcitabine	4.4	Hydrophilic (0.04)	0.5–2	1–3	<4
Stavudine	10.0	Hydrophilic (NA)	3.8	1–1.5	Negligible
Abacavir	0.4, 5.1	Lipophilic (NA)	0.7–1.7	0.9–1.7	50
Nonnucleoside reverse transcriptase inhibitors
Nevirapine	2.8	Slightly lipophilic (1.8)	4	25–30	60
Delavirdine	4.3–4.6	Slightly lipophilic (2.98)	1	4.4–11	98
Efavirenz	10.2	Lipophilic (NA)	2–5	40–52c	99.5
Protease inhibitors
Ritonavir	2.8	Lipophilic (4)	2–4	3–5	98
Saquinavir	1.1, 7.1	Lipophilic (NA)	NA	6	>98
Indinavir	6.2	Hydrophilic (NA)	0.5–1.1	1–3	60
Nelfinavir	6.0, 11.1	Lipophilic (NA)	2–4	3.5–5	99
Amprenavir	1.9	Lipophilic (NA)	1.2	7–10	95

^aC_max, maximum concentration of drug in serum. 

^bNA, not available. 

^cAt steady state. 

TABLE 2.

Fluid distribution characteristics of selected weakly acidic or basic antiretroviral agents^a

Drug	pKa	CPR/CP	CSV/CP	Cother/CP	CS/CP
Weak acids
Lamivudine	4.3	0.1	2.5	1.3	1.7
Zalcitabine	4.4	0.1	2.5	1.3	1.7
Nevirapine	2.8	0.1	2.5	1.3	1.7
Delavirdine	4.6	0.1	2.5	1.3	1.7
Ritonavir	2.8	0.1	2.5	1.3	1.7
Weak bases
Zidovudine	9.7	7.9	0.4	0.8	2.7
Didanosine	9.1	7.8	0.4	0.8	2.7
Abacavir	5.1	1.0	1.0	1.0	1.0
Efavirenz	10.2	7.9	0.4	0.8	2.7
Indinavir	6.2	1.4	1.0	1.0	1.1
Saquinavir	1.1/7.1	1.0–3.3	∼1.0	∼1.0	∼1.5
Nelfinavir	6.0/11.1	1.3/7.9	1.0/0.4	1.0/0.8	1.1/2.7

^aBody fluid (C_F) and blood plasma (C_P) drug concentration ratios were derived from modified HH equations predicting weakly acidic and weakly basic drug concentrations, where pH_F is the pH of body fluid and 7.4 is the pH of plasma. For weak acids, C_F = 1 + 10^{pH_F − pK_a} and C_P = 1 + 10^{7.4 − pK_a}. For weak bases, C_F = 1 + 10^{pK_a − pH_F} and C_P = 1 + 10^{pKa − 7.4}. The assmptions are that (i) the pH of prostate fluid is 6.5, the pH of seminal vesicle fluid is 7.8, and the pHs of all other fluids are 7.5 and (ii) the relative contributions of fluids to total seminal volume are as follows: prostate, 30%; seminal vesicles, 60%; other, 10%. C_PR, prostatic-fluid drug concentration; C_SV, seminal-vesicle drug concentration; C_other, bulbourethral- and urethral-gland, epididymis, and ampulla drug concentration; C_S, semen drug concentration. 

Pharmacological studies of antiretroviral agents in semen.

In an attempt to corroborate some of the above theoretical considerations with published experimental data (excluding data in abstract form), a literature search of Medline, IOWA, International Pharmaceutical Abstracts, and Current Contents yielded a single original article investigating the concentration of zidovudine in the serum and seminal plasma in six patients (46).

After 200 mg of zidovudine was administered, semen-to-semen drug concentration ratios ranged from 1.3 to 20.4 in samples obtained 0.75 to 1.25 h after administration (coinciding with peak serum drug concentrations) and from 2.3 to 16.8 in samples obtained 3 to 4.5 h after administration (coinciding with trough serum drug concentrations). This suggested sequestration or selective transport of the compound into the male reproductive tract. Explanations provided by the authors for the wide range of drug concentrations included noncompliance, gastrointestinal malabsorption, and delayed absorption. Comparisons of seminal 3′-azido-3′-deoxy-5′-O-β-d-glucopyranuronosylthymidine (GAZT) and zidovudine concentrations excluded local conversion of GAZT (zidovudine’s major metabolite) back to zidovudine by β-glucuronidase. This example demonstrates that theoretical calculations (Table 2) may be suboptimal in predicting the passage of antiretrovirals into the seminal compartment.

More recently, we measured blood and seminal plasma zidovudine (300 mg twice daily or 200 mg three times daily) and lamivudine (150 mg twice daily) concentrations in nine HIV-positive, antiviral-therapy-naïve men (93). A total of 71 semen and plasma samples were collected during the first 200 days of antiretroviral therapy. Similarly to the observations by Henry et al. (46), the median semen/blood plasma zidovudine ratio in our investigation was 5.9 (25th to 75th percentile, 0.95 to 13.5). This is the first report of lamivudine concentrations in the genital tract. As predicted by the zidovudine data, lamivudine also concentrated in the seminal plasma, with a median semen/blood plasma ratio of 9.1 (25th to 75th percentiles, 2.3 to 16.1).

Recently, an investigation of ritonavir and saquinavir seminal concentrations in seven subjects was published in abstract form (119). The concentrations of both protease inhibitors were generally less than 5% of the concurrent concentrations in plasma. The physiologic mechanisms underlying this observation remain to be determined.

Future studies must use systematic sampling strategies to accurately determine the pharmacokinetics of antiretroviral drugs in semen. Measurement of pharmacokinetics in semen of other classes of antiretrovirals such as nonnucleoside analogues and protease inhibitors is imperative. Additionally, since concentrations of nucleoside analogue reverse transcriptase inhibitors (which are essentially prodrugs) in semen do not necessarily reflect intracellular concentrations of the active triphosphorylated metabolites (7, 103), determining the relationship between drug concentrations in cell-free seminal plasma and concentrations of active metabolites in CD4⁺ cells is an important area of investigation.

Effects of antiviral therapy on HIV-1 shedding.

A number of prospective and cross-sectional studies have examined the relationship between current antiretroviral therapy and the ability to isolate virus from semen (Table 3). Results from a number of older cross-sectional studies have been varied (2, 44, 60, 61, 68). However, many of these are limited by the use of insensitive, nonquantitative viral-culture techniques and the inclusion of heterogeneous patient populations with various antiretroviral histories and different likelihoods of harboring resistant virus. Additionally, these investigations used zidovudine monotherapy, which has limited antiviral efficacy of short duration (30).

TABLE 3.

Summary of published studies evaluating the effects of antiviral therapy on HIV-1 shedding in semen^a

Reference [yr]	n	Study type	Drug therapy and no. of subjects
Anderson et al. (2) [1992]	95	Cross-sectional	None, 64; AZT, 31
	14	Longitudinal (5–8 mo)	None, 9; AZT, 5
Delwart et al. (23) [1998]	5	Cross-sectional	None, 2; AZT, 2; AZT + ddI, 1
Eron et al. (31) [1998]	11 (5 ARV naïve, 6 NA experienced)	Longitudinal (8–58 wk)	Naïve patients, single or dual NA or NA PI; experienced patients, 2 NAs + 1 PI
Gilliam et al. (38) [1997]	11 on newly initiated therapy, 11 on stable therapy	Longitudinal (up to 90 wk in the new TX group, up to 26 wk in the stable TX group)	New TX; AZT or ddI ± DLV; stable TX;no TX, 6; mono-TX, 1; combination TX, 4
Gupta et al. (43) [1997]	6	Longitudinal (0–28 wk)	IND ± efavirenz
Hamed et al. (44)	36	Cross-sectional	ART naïve, 9; AZT, 19; ddl, 4; AZT + ddl, 4
	6	Longitudinal (8 wk after TX initiation)	AZT or ddl, AZT + ddl, or AZT + ddC
Krieger et al. (60) [1991]	34	Cross-sectional	None, 30; AZT, 22
Krieger et al. (59) [1995]	56	Mixed (single samples from 22 subjects; multiple samples from 34 subjects)	None, 8; AZT or ddI, 30; AZT + ddC 2, unknown, 16
Kroodsma et al. (61) [1994]	16	Cross-sectional	AZT, 7; ddI, 6; AZT + ddI, 3
Musicco et al. (86) [1994]	436 couples (HIV-1-infected men and their monogamous seronegative female sexual partners)	Cohort (740 person-year follow-up)	AZT, 64
Vernazza et al. (123) [1997]	101	Cross-sectional	ARV-naïve, 53; 1 or 2 NAs, 29; no therapy, 19 discontinued
Vernazza et al. (124) [1997]	4 (19 TX naïve, 25 experienced)	Longitudinal (0–10 wk)	AZT, 1; IND, 1; AZT + 3TC, 5; 2 NAs + PI, 16; 1 or 2 NAs ± SQV or DLV, 21
Vernazza et al. (125) [1999]	114 cases, 55 historical controls	Case control	Cases: 2 NAs PI, 97; ≥2 NAs, 16; ddI + HU, 1; controls; drug naïve
Zhang et al. (138) [1998]	7	Cross-sectional	2 NAs + PI for 5–41 mo prior to study

STD screening	HIV-1 detection	Results
Yes	Microculture	Median AZT concn 0.004 (0–0.019) ng/ml; AZT associated with a decrease in detection of seminal HIV-1 (adjusted OR, 0.04; 95% CI, 0–0.63)
Yes	Microculture	HIV-1 detected in 43% of men without AZT TX and in 0% of men with AZT TX
Unknown	HIV-1 RNA sequence analyses by nested PCR	Distinct variant populations in BP and SP in 3/5 subjects; no semen-specific signature amino acid sequence detected
Yes	HIV-1 RNA by NASBA, HIV-1 RNA sequence analysis by Affymetrix	Median BP and SP HIV-1 RNA levels before TX, 5.1 and 5.7 log10 copies/ml; median maximal change in BP and SP HIV-1 RNA levels with TX, −0.95 and −1.41 log10 copies/ml; baseline SP RT resistance mutations in 3/6 NA-experienced subjects (vs 1/5 for TX-naïve subjects); 8/11 subjects developed resistance mutations during TX (to NAs and PIs in BP; to NAs in SP)
Yes	Microculture of seminal cells, HIV-1 RNA by NASBA	Before TX, 50% seminal-cell culture positive; after TX, 0.1% seminal-cell culture positive; median SP HIV-1 RNA reduction, 1.01 log10 (78% BLQ); in 55%, decrease in SP HIV RNA was > BP HIV RNA decrease; in stable TX group, median SP HIV-1 RNA, 1.45 log10 lower than that in patients on TX (P = 0.05), with less within-subject variability
Unknown	HIV-1 RNA by NASBA	SP RNA decreased 5- to 10-fold after 2 wk (n = 3), SP RNA decreased 4- to 150-fold after 4 wk (n = 6), HIV-1 RNA remained undetectable in BP and SP for duration of study
Yes	RT-PCR	HIV-1 DNA detected in semen in 67% of untreated patients and 78% of treated patients, HIV-1 RNA detected in semen in 44% of untreated patients and 48% of treated patients
Yes	RT-PCR	Before TX, HIV-1 DNA detected in blood in 6/6 patients and in semen in 4/6 and remained stable over 8 wk of TX; before TX, HIV-1 RNA detected in BP 4/6 in patients and in SP of 2/6 patients and decreased to BLQ during 8 wk of TX
Unknown	Mixed-lymphocyte culture method, with HIV p24 antigen detection	16/55 semen samples HIV positive (10 in cellular fraction only, 3 in plasma fraction only, 3 in both fractions); no difference in HIV isolation between TX and no TX (41 vs 23%) or between symptomatic and asymptomatic (28 vs 32%)
Yes	Mixed-lymphocyte culture method, with HIV p24 anigen detection by ELISA	36/215 semen samples HIV positive (24 in cellular fraction only, 5 in plasma fraction only, 7 in both fractions); CD8+ in BP predictive of HIV-1 in semen (OR, 5.0; 95% CI, 1.0–23.9); TX not associated with HIV-1 in semen (OR, 1.7; 95% CI, 0.6–5.1)
Unknown	RT-PCR	Recovery of HIV-1 RNA, 68% in BP and 44% in SP; 25% of paired semen and blood samples had discordant genotypes (codon 215 or 74)
Unknown	Seroconversion measured by ELISA with Western blot confirmation	Incidence rates of seroconversion with and without AZT TX, 4.4 and 3.8, respectively, per 100 person-years; relative risk of sexual transmission of HIV with AZT TX, 0.5 (95% CI, 0.1–0.9)
Yes	NASBA	Seminal HIV lower in treated vs untreated patients (P = 0.03); treatment was an independent inverse predictor of HIV RNA detection (OR, 0.38; P = 0.054)
Yes	RT-PCR (blood), NASBA (semen), and microculture	At baseline, 68% had HIV-1 RNA in SP and 37% had HIV-1 in coculture; at follow-up, 27% had HIV-1 RNA in SP and 12% had HIV-1 in coculture; the 27% with undetectable BP RNA also had undetectable SP RNA and culture
Yes	HIV-1 RNA by Nuclisens (114 cases, 55 controls) HIV-1 DNA by nested PCR (67 cases, 55 controls)	67% detection frequency of HIV RNA in controls vs 2% in cases (OR, 0.01; 95% CI, 0.0–0.3); 38% detection frequency of HIV DNA in controls vs 16% in cases (OR, 0.32; 95% CI, 0.12–0.80)
Unknown	HIV-1 RNA by RT-PCR, p24 antigen ELISA of mixed-lymphocyte coculture, DNA sequence analysis	HIV-1 RNA in BP and SP BLQ; HIV-1 DNA in PBMC of 7/7 and in seminal cells of 4/7; replication-competent HIV-1 in PBMC of 3/7 and in seminal cells of 2/3 (1 macrophage tropic and 1 dual tropism)

^aOnly Anderson et al. 2 in the cross-sectional study measured antiretroviral-drug concentrations, in 19 of 31 patients. BP, blood plasma; SP, seminal plasma; RT, reverse transcriptase; TX, antiretroviral therapy; NA, nucleoside analogue; PI, protease inhibitor; AZT, zidovudine; ddI; didanosine; ddC, zalcitabine; IND, indinavir; SQV, saquinavir; DLV, delavirdine; HU, hydroxyurea; BLQ, below limit of quantitation; STD, sexually transmitted disease; ELISA, enzyme-linked immunosorbent assay; OR, odds ratio; CI, confidence interval; ARV, antiretroviral; NASBA, nucleic acid sequence-based amplification. 

Recent studies of a variety of antiretroviral regimens have shown decreases in seminal plasma HIV-1 RNA concentrations commensurate with those in blood plasma in most patients (38, 43, 124). Potent and sustained suppression of HIV-1 shedding in semen can be attained with powerful antiviral combinations (38, 43, 72, 124, 125). However, longer follow-up is required to assess the duration of these effects and to determine the relative rates of emergence of drug-resistant virus within the compartments. Recently Zhang et al. (138) determined that although HIV was eliminated from the seminal plasma in seven men on potent combination antiretroviral therapy, proviral DNA could still be detected in seminal cells, and HIV could be recovered in culture.

Delwart and coworkers reported that plasma viral rebound after suppression by protease inhibitors was associated with the initial appearance of viral populations (present before the initiation of drug treatment) containing envelope sequences distinct from those in the wild-type population (22). This finding suggests the possibility of repopulation from a viral sanctuary (such as the genital tract) where drug concentrations and viral suppression may have been inadequate. The possibility of long-term disease control increases the necessity to investigate the effects of antiretroviral agents in HIV sanctuary sites such as the male genital tract.

Submaximal HIV-1 suppression and emergence of HIV-1 drug-resistant variants.

Exposure of productive cells in the genital tract to submaximally suppressive antiviral-drug concentrations could allow continuing HIV-1 production and infectiousness, despite systemic efficacy (31). Equally important, partial suppression of viral replication in the genital tract by antiviral drugs may lead to the selection of drug-resistant mutants. Cross-sectional comparisons of reverse transcriptase sequences in various tissues of previously treated patients indicate that viral selection in response to antiretroviral pressure can indeed be compartmentalized (131). A number of cases of primary infection with sexually transmitted drug-resistant virus have now been reported (29, 45, 51, 112). The prevalence of codon 215 zidovudine resistance reverse transcriptase mutations appears to be increasing in some populations (105), indicating the potential for widespread transmission of drug-resistant virus.

The development of drug resistance to all classes of antiretrovirals occurs with therapy that is not completely suppressive (84), and multidrug-resistant variants have been observed (84, 106, 109). An inverse relationship between trough concentrations of ritonavir and both viral load and the emergence of drug-resistant virus has been demonstrated (84). Thus, with the widespread administration of combination antiretroviral regimens, it will be critical to deliver adequate drug concentrations to all compartments in which the virus can replicate. The likelihood of partial adherence to therapy (75), as well as drug interactions and altered drug absorption and metabolism related to HIV complications (21, 74, 97), increases the probability that drug-resistant mutants will be generated in the systemic compartment and in the genital tract, resulting in transmission of resistant virus to new hosts.

Three investigations have examined the evolution of drug-resistant virus in semen (12, 31, 61). Following treatment with ddI, Kroodsma et al. (61) demonstrated that the rate of appearance of ddI resistance-conferring mutations differed between HIV-1 populations in blood and semen. Byrn et al. (12) examined protease gene sequences of virus recovered from the cells of paired specimens of blood and semen from two men infected with HIV-1 for more than 8 years (12). One patient had received antiretroviral therapy for a number of years (including a protease inhibitor for 4 months), while the other was naïve to antiretroviral therapy. Phylogenetic analyses of the sequences from each patient sample revealed distinct viral populations in the blood and semen, ranging from 4 to 8 amino acid substitutions. In the patient receiving the protease inhibitor, only the blood clones contained mutations characteristic of emerging protease inhibitor resistance. Most recently, Eron et al. (31) compared polymerase and protease gene sequences between blood and seminal plasma in HIV-infected men on newly initiated antiretroviral therapy. Resistance to antiretroviral agents was documented for both blood and seminal plasma. The rates and patterns of emergence of resistance in the two compartments were frequently different. A phylogenetic analysis of reverse transcriptase and protease sequences demonstrated that the majority variant in seminal plasma was almost always distinct from that of the majority variant in blood, and frequently there was substantial genetic distance between the variants.

ANIMAL MODELS FOR EVALUATING SEMINAL DRUG EXCRETION

Animal models provide a method of screening chemically and pharmacologically active investigational compounds that cannot yet be administered to humans and allow for intensive experimentation and dissection of drug penetration and viral replication in various genital-tract tissues and glands. Although no animal studied to date has reproductive and metabolic characteristics comparable to those of humans (5, 56), at least six different animal models have been used to examine drug effects in semen (1, 5, 24, 55, 95). The two most widely investigated, nonrodent species used in toxicology, pharmacokinetic, and pharmacodynamic studies are the canine and the nonhuman primate (24), whereas the two commonly used nonrodent animal models of immunodeficiency virus infection are felids and nonhuman primates (37).

Canine model.

The canine model has been used primarily for investigations of prostatic penetration of antibacterial agents. In this area of study, prostatic-secretion results are generally comparable to those for humans (102, 129). However, there are a number of limitations to using this model for investigations of the pharmacokinetics and pharmacodynamics of antiretroviral drugs in semen.

The pharmacokinetic parameters of absorption, metabolism, and excretion impact the concentrations of drugs and metabolites in plasma and in turn, influence total drug exposure and the concentrations of drugs achieved in peripheral compartments. Generally, the canine model does not closely reflect the pharmacokinetic parameters seen for humans. The canine model has altered gastrointestinal absorption (89), decreased hepatic blood flow and relative liver size, and altered phase I (cytochrome P-450-mediated oxidation-reduction reactions) and phase II (conjugation reactions) drug-metabolizing enzyme activity (5, 16, 87, 108). These differences can particularly impact the pharmacokinetics of antiretroviral compounds relying primarily on phase I (nonnucleoside reverse transcriptase inhibitors and protease inhibitors) or phase II (nucleoside analogues) metabolic processes.

In addition, differences in the type and sequence of gland contribution to seminal plasma are important for predicting drug concentrations and activities in specific components of the reproductive tract. Canine reproductive-tract anatomy, glandular contribution to secretions, and seminal-fluid composition are markedly different from those of humans (25, 115). Seminal-fluid composition also varies between the resting state and the ejaculate state, and this variation can confound measures of drug concentrations (102, 111, 129).

Nonhuman primate model.

Nonhuman primates demonstrate marked similarities to humans in almost all aspects of anatomy (32) and physiology (20, 24, 55, 92, 100, 108, 116, 136), which underlie the value of these animals for experimental investigations (80). Although the male reproductive system of the nonhuman primate is structurally and biochemically similar to that of man (4, 8, 53, 66, 68, 99), an analytical limitation exists in using this model for pharmacokinetic investigations. Upon ejaculation, a reaction between the secretions of the prostatic cranial lobe and seminal vesicles causes the ejaculate to coagulate within seconds. The large fraction (55 to 68%) of solid, rubbery coagulum can be digested only with high concentrations (i.e., 5% alpha-chymotrypsin) of proteolytic enzymes (34). This poses an analytical challenge for assessing seminal antiretroviral concentrations, since only a scant amount of liquid fraction is available for assay (62).

An advantage of the nonhuman primate model is the ability to be infected with SIV, a pathogen closely related both morphologically and genetically to HIV (36, 54, 77, 78, 81). Reproductive-tract pathology in chronically SIV-infected male rhesus macaques resembles that of AIDS patients, and SIV-infected cells have been found at all levels of the reproductive tract (78, 79).

Additionally, intraviral recombinants of SIV (SHIV) have been developed which carry the genes of HIV (e.g., env, vpu, tat, and rev). This genetic manipulation provides particular advantages for specific immunologic studies of HIV-1. Whereas inoculation of nonhuman primates with HIV-1 has rarely resulted in an AIDS-like syndrome, animals can be successfully infected with SHIV (65). Therefore, the nonhuman primate model can be used not only for the pharmacokinetic study of antiretroviral-drug concentrations in semen but also for preliminary pharmacodynamic study of the effectiveness of pharmacological agents in the prevention of sexual transmission of SIV or SHIV (81, 122). However, the major limitations of cost, specialized care, restricted supply, and analytical challenges make this model less desirable for investigating the pharmacokinetics and pharmacodynamics of antiretroviral agents in semen (57).

Feline model.

Advantages of the feline model for pharmacological study of antiretroviral agents include greater availability, less expense, and less specialized care compared to nonhuman primates. Although a structurally dissimilar genital tract (39) and divergent absorption, metabolism, and excretion characteristics (15, 52) may make this model less suited for predicting antiretroviral compartmentalization in humans, there have been a limited number of pharmacological investigations with this model to date, and the effects of these characteristics are largely unknown (88, 118).

Feline immunodeficiency virus (FIV) infection is similar to HIV-1 infection in causing an acute flu-like illness, followed by a clinically latent period with progressive immune dysfunction (9). FIV has been isolated from the blood (134), saliva (73), semen (52), cerebrospinal fluid (134), and milk (107) of domestic cats and has demonstrated transmission by parenteral (28), oral (83, 107), and rectal (83) exposures, as well as vertically both in utero (91) and through nursing (107). Venereal transmission is naturally uncommon (134, 135); however, the ability to induce FIV infection through artificial insemination has recently been demonstrated (51), making the cat a promising model for the study of viral transmission.

One constraint to using the FIV model for pharmacodynamic investigations of antiretroviral therapies targeted against HIV is the limited sensitivity of the virus to current therapies. Although many nucleoside analogues inhibit FIV replication, nonnucleoside reverse transcriptase inhibitors do not significantly inhibit the FIV reverse transcriptase (88). Additionally, currently available HIV protease inhibitors are not effective against FIV, although preliminary reports suggest that specific structural features can confer inhibitory activity to both FIV and HIV-1 (63). In vivo antiretroviral activity has yet to be demonstrated.

CONCLUSIONS

HIV is spread primarily from men to their sexual partners with modest efficiency. It has been well established that HIV replicates in the genital tract, a separate compartment subject to unique selective pressures. The phenotype of the organism and the concentration of HIV in semen almost certainly determine the efficiency of transmission. The ability of antiretroviral drugs to penetrate or concentrate in seminal plasma and the intracellular phosphorylation of nucleoside analogues will help to determine the concentration of HIV detected in seminal plasma and seminal cells. The study of the pharmacology of antiretroviral drugs in semen is rudimentary. Data exist for zidovudine, lamivudine, ritonavir, and saquinavir. Unexpectedly high concentrations of the nucleoside analogues in seminal plasma and low concentrations of protease inhibitors deserve emergent study, both because such information will lead to development of drugs specifically designed toward the reduction of HIV in semen (for the purposes of prevention of HIV transmission) and to help avoid the development of resistant HIV isolates. Correlation of animal model data to human data is a critically important and essential goal for drug development. Based on the available evidence, it seems likely that antiretroviral drugs can be used to reduce the concentration of HIV in semen and, therefore, will play a role in prevention of transmission.