Abstract
The effect of injection-drug use on human immunodeficiency virus type 1 (HIV-1) env genetic evolution was examined in 15 seroconverting injection-drug users followed up for 4 years. After adjustment for non-drug-related independent variables significantly associated with genetic diversity (time since seroconversion and progressor status), injection frequency was positively and highly significantly associated with HIV-1 env genetic diversity (P = .003). The mutation rate in those who had injected at least once a day during the previous 6 months was estimated to be 62% greater than the rate in those who had not injected at all. If the positive effect of drug-injection frequency on env genetic diversity extends to the HIV-1 pol gene, the risk of emergence of resistance to antiretroviral drugs may be enhanced by increased drug-injection frequency, especially under the selection pressure of antiretroviral therapy.
Both in vivo animal studies and in vitro experiments have suggested that use of illicit drugs may influence the course of human immunodeficiency virus type 1 (HIV-1) infection. Illicit drug use can affect parameters of immune function such as immunoglobulin concentration, lymphocyte number and proliferation, and cell surface marker expression [1–5]. Heroin and cocaine have also been reported to enhance HIV replication in vitro [4].
Despite these findings, epidemiologic studies evaluating the in vivo effect of illicit drugs on HIV-1 infection have been contradictory. An early study conducted in injection-drug users (IDUs) suggested that the use of drugs might be associated with accelerated CD4+ cell loss [6]. However, studies comparing IDUs with homosexual men [7–9] or with homosexual men and persons infected through heterosexual contact [10] did not confirm that illicit drug use affects HIV-1 disease progression. In these studies, patterns of CD4+ T cell decline and progression to AIDS were not appreciably different between HIV-positive IDUs and other HIV-positive groups. Longitudinal analyses of IDUs in seroprevalent cohorts showed that rates of CD4+ T cell decline were not significantly associated with different injection frequencies or injection intensity of specific drug types [11–13].
Although progression to AIDS and CD4+ T cell decline are important clinical parameters of the status of HIV-1 infection, they represent the culmination of whatever competing host and viral factors may be interacting. As such, they may not be the most sensitive indicators of drug effects on host and/or virus biology during the course of infection.
One feature that distinguishes HIV-1 infection from that of other microbial pathogens is the genetic heterogeneity that may exist among viral clones isolated from the same individual [14–18]. This heterogeneity arises from mutations appearing during reverse transcription and is amplified by the high replication rate of HIV-1, its persistence over many years in the same host, and the high viral population density. The ability of the virus to maintain such a high mutation frequency is responsible for the drug resistance and escape from antiviral immune responses that characterize this disease [19–22].
Which of the mutations ultimately persists in the infected individual is determined by viral replication properties and host selection forces, including such factors as the immune response to the virus, the activation state of potentially infectable cells, and the variable expression on those cells of coreceptors for the virus [23, 24]. Multiple studies have explored the association between patterns of HIV-1 evolution at the nucleotide level and disease progression [14, 16, 25, 26]. Most of these studies have focused on patterns of mutation in the env gene, which encodes a protein that tolerates a high mutation frequency, strongly influences cell tropism, and is an important target for neutralizing antibody [27–29].
In a study of HIV-1 envelope evolution in a cohort of 15 IDUs followed up for 4 years from the point of seroconversion, we determined that viruses isolated from those IDUs with more rapidly progressive disease showed greater genetic diversity in the env region. In those with more rapidly progressive disease, that diversity was also associated with a greater frequency of nonsynonymous nucleotide changes, that is, changes at the nucleotide level that resulted in changes at the amino acid or protein level [14]. In addition, when the original level of the CD4+ count was controlled for, the extent of viral env genetic diversity at a given point in the disease course correlated with the decline in CD4+ T cell counts over the subsequent year.
This established association between these parameters of viral evolution within an individual and disease progression provided another criterion against which the impact of drug use on disease progression could be measured. We hypothesized that different drug-use patterns might be a cofactor for patterns of genetic evolution of HIV-1. The purpose of the present study was to examine prospectively whether different patterns of drug use were associated with patterns of HIV-1 env evolution in IDUs.
Materials and Methods
Study population
The 15 individuals to be studied were a subset of subjects from the ALIVE (AIDS Linked to the Intravenous Experience) Study, a cohort of IDUs established to characterize the natural history of HIV infection in injection-drug-using populations. The rationale, organization, and data collection have been described elsewhere [30]. Briefly, subjects were recruited through extensive community outreach efforts between February 1988 and March 1989 from a variety of community-based agencies. Eligibility requirements for enrollment included being aged ⩾18 years, reporting a history of injecting drugs within the prior 10 years, and being AIDS-free at study entry.
Data collection
All volunteers who agreed to participate in the study were followed up semiannually. At each visit, participants had a venipuncture for HIV serologic testing and were interviewed about illness history, drug use, and sexual practices. Serum samples were assayed for HIV-1 antibodies by ELISA (Genetic Systems, Seattle) and Western blot (DuPont, Wilmington, DE).
Questions about drug use covered duration of drug use at baseline and, during the previous 6 months, frequency of overall injection and drug type. In addition, for each drug type, the participants were asked to report the number of days during which that drug was injected. Participants were also questioned about whether they used their own drug-preparation equipment (“works”), whether they used works previously used by another IDU, the number of people with whom they shared a needle, and whether they attended shooting galleries where multiple users were injecting drugs.
Although no objective data are available to confirm the reported drug-use patterns, systematic study of the reliability of such information has shown consistency of reported data when subjects are queried at different time points [31]. Furthermore, studies of the ALIVE cohort have indicated no significant positive association between socially desirable responses and reported drug-use frequency [32].
Six of the individuals followed up in this study were given zidovudine at some point during the follow-up period. It is not known whether these individuals were adherent to the therapeutic regimen. Three of these individuals were using zidovudine at only ⩽2 visits. There was no significant correlation between zidovudine use and the different parameters of genetic evolution or between zidovudine use and frequency of injection of drugs of abuse (data not shown).
Genetic characterization of HIV-1 evolution
Fifteen individuals were selected for study on the basis of the availability of specimens from the time of HIV-1 seroconversion and of differences in disease course, as defined by patterns of CD4+ T cell decline. They were classified as rapid progressors, moderate progressors, and nonprogressors on the basis of the CD4+ T cell level achieved during a period of up to 4 years after seroconversion. Rapid progressors were defined as having attained a level of <200 CD4+ T cells/mL within 2 years of seroconversion; moderate progressors' CD4+ T cell levels declined to 200–650/mL during the 4-year period of observation, and nonprogressors maintained CD4+ T cell levels >650/mL throughout the observation period.
Methods for analysis of viral genetic evolution have been described elsewhere [14]. Briefly, the genetic characterization focused on a 285-nucleotide sequence, extending from the second constant region domain through the third hypervariable domain of the env region of proviral DNA, obtained from the subjects' frozen peripheral blood mononuclear cells (PBMC). Between 6 and 21 env sequences for each visit were amplified by polymerase chain reaction from each individual, ensuring that the viral copy number was high enough to prevent oversampling of the same viral clone [14]. The amplified products were then cloned and sequenced. Changes in HIV-1 sequences over time were quantified by 3 parameters: (1) the genetic diversity at each visit, defined as the mean number of nucleotide differences between pairs of clones from that visit; (2) divergence, quantified as the median percentage of nucleotides per clone at a given visit that differed from those present in the consensus env sequence from the first visit of that subject; and (3) the rate of synonymous (amino acid-preserving) nucleotide mutation per potential site of synonymous mutations (dS) and the rate of nonsynonymous (amino acid-changing) nucleotide mutation per potential site of nonsynonymous mutations (dN) [14]. All of the sequences analyzed can be obtained from Genbank (accession numbers AF089110–AF089708 and AF016760–AF016825).
Definition of drug-use variables
For this longitudinal analysis, the drug-use information obtained at each visit was analyzed for the previous 6 months. A participant was classified at the baseline date as a current user if he or she had injected at least once during the prior 6 months. For each visit, the overall injection frequency within a 6-month period was analyzed as a categorical variable according to 3 different categories corresponding to the daily averages: none, <1 injection per day, and ⩾1 injection per day.
To estimate the exposure to different drugs, we determined whether the participant had used cocaine alone, heroin alone, or both heroin and cocaine during the prior 6 months. We also determined whether the participant had shared needles, injected with his or her own drug paraphernalia, attended a shooting gallery, or used a needle after its use by another IDU. These variables were categorized as either “never” or “ever” during the prior 6 months.
Analysis
The analysis was performed considering 4 outcomes of HIV-1 genetic variation: mutational divergence from the first visit, genetic diversity at a single visit, the (dN) rate of nonsynonymous mutations at potential sites of nonsynonymous mutation, and the (dS) rate of synonymous mutations at potential sites of synonymous mutation.
The independent variable of interest, frequency of injection-drug use, was categorized as follows: no drug use in the past 6 months, <1 injection per day in the past 6 months, and ⩾1 injection per day in the past 6 months. The effect of frequency of injection-drug use on each of the genetic variation outcomes was analyzed as a categorical variable with no drug use as the reference category.
Analyses for repeated measures to assess the effect of various drug-injection patterns on the genetic variation over time were performed by use of the generalized estimating equations approach to linear models [33]. This method takes into account the correlations of the within-person repeated observations by not weighting those repeated measurements as strongly as if they were independent measurements from different people. This approach was considered because of the robustness of the method with respect to specifying correlation structures.
Univariate analyses were performed first. Then, for the variables that reached statistical significance in unadjusted models (P < .05), multivariate models were used. For the adjustment we considered the following variables: sex (female preference], male), age at se-roconversion (as a continuous variable), time since seroconversion (as a continuous variable), duration of drug use at baseline (<10 years [=reference], >10 years), current drug use (no [=reference], yes), CD4+ T cell count at each visit categorized into quartiles (⩾749 [=reference], 522–748, 352–521, ⩽351), and disease progression (rapid [=reference], moderate, stable). AIDS status was not considered, because only 1 subject had AIDS, and that subject had it at only 1 time point.
Results are reported as the linear regression parameter estimates, along with 95% confidence intervals (CIs), for the effect of injection-drug use frequency on the genetic variable of interest. The linear regression parameter (β) corresponds to the change of the mean genetic variation measure when comparing no injection-drug use with either <1 injection per day or ⩾1 injection per day, whichever is specified.
Role of superinfection in genetic evolution parameters
Because greater frequency of injection increases the risk of exposure to new HIV-1 variants, it was necessary to exclude the possibility that sequence diversity or divergence was attributable to superinfection with new viruses. For this purpose, a phylogenetic tree of clones from the different subjects (not shown) was constructed that showed the independent segregation of those clones, with the exception of clones from 2 subjects who were known to be epidemiologically related (subjects 1 and 2) [34]. Examination of these trees revealed that the segregation of clones from individual subjects was monophyletic, showing no evidence of superinfection.
Results
Baseline demographics
Table 1 presents the demographic and baseline characteristics of the 15 HIV-seroconverting IDUs included in this analysis. They were predominantly men (73%) and black (100%), with a median age of 32 years. Fourteen participants were current IDUs who reported injection-drug use at a median of 25 days prior to the baseline visit. A total of 78 visits were observed by the 15 participants in study follow-up, with a median of 5 visits per participant. The median time between visits was 6.2 months, and the maximum follow-up time after seroconversion was 4 years and 5 months (median, 21.9 months).
Demographics and baseline characteristics of 15 participants.
Demographics and baseline characteristics of 15 participants.
Genetic variation and drug-use characteristics
The patterns of drug use, viral genetic evolution, and CD4+ T cell decline for each study subject are characterized in table 2. The subjects were placed into 3 progressor groups on the basis of the level to which their CD4+ T cells declined during the period of observation. At the initial seropositive visit, the CD4+ T cell counts ranged from 405 to 1072 cells/mL. Among all subjects during the study, the CD4+ T cell count changed at a median annual rate ranging from −593 cells/mL per year to +53 cells/mL per year. At the first seropositive visit, the mean number of nucleotides within the 285-bp envelope sequence that differed among the clones within a single subject (the intravisit genetic diversity) ranged from 0.9 to 15.16. For all visits, the median for this genetic diversity among clones within a single subject ranged from 1.40 to 12.76 nucleotides per clone. Within an individual subject, at a given visit, the median percentage of nucleotides per clone that diverged from those present in the consensus sequence from the first seropositive visit ranged from 0.46% to 4.14%. The median dN rate ranged among the different subjects from 0.0% to 4.2%, and the median dS rate ranged from 0.0% to 4.9%. Median drug-injection frequency in a given subject ranged, over the course of the study, from 0 to 5.2 injections per day in the previous 6 months.
Summary data on CD4+ T cell decline, genetic variation, and drug-use frequency in 15 seroconverters.
Summary data on CD4+ T cell decline, genetic variation, and drug-use frequency in 15 seroconverters.
Drug-use behavior for the cohort as a whole is further defined in table 3. The median injection frequency during the previous 6 months was 49.5 injections (range, 1.0–360 injections). During 21.8% of the visits, subjects reported no injections in the previous 6 months. The analysis for type of drug used showed that both heroin and cocaine were used during the previous 6 months in 70.4% of the visits, cocaine exclusively in 21.4%, and heroin exclusively in 8.2% of the visits.
Overall drug-use patterns in cohort.
Overall drug-use patterns in cohort.
Unadjusted linear regression parameters
Table 4 describes the unadjusted linear coefficient parameters (β) with the 95% CIs estimated for the effect of injection-drug use frequency on genetic diversity, mutational divergence, the dS rate, and the dN rate. The mean number of env nucleotides that differed among the different clones isolated from the same subject at the same visit (the genetic diversity) increased by an estimated 0.24 mutations (95% CI, −0.91 to 1.40; P = .673) for those who had injected less than once daily in the previous 6 months, compared with those who had no injections, and by 1.83 mutations (95% CI, 0.25–3.40; P = .023) for those who had injected at least once daily in the previous 6 months, compared with those who had no injections. An analysis of the linear dose response between the injection frequency category and mean genetic diversity showed that if the injection-drug use frequency increased by 1 category (none, <1/day, ⩾1/day), the mean genetic diversity increased by 0.97 mutations per clone (β = 0.97, P < .005; data not shown).
Unadjusted linear regression estimates for effect of drug-use frequency on genetic variation during study period.
Unadjusted linear regression estimates for effect of drug-use frequency on genetic variation during study period.
Because both heroin and cocaine had been used in the previous 6 months in 70.4% of drug-use observations, it is impossible to decipher the independent effects of the individual drugs on genetic diversity. The rates of genetic divergence, synonymous mutation, and dN mutation were not significantly associated with the frequency of drug use in this unadjusted model.
The other drug-use behaviors analyzed included whether subjects used their own works, whether they used works previously used by another IDU, the number of different people with whom they shared a needle, and shooting gallery attendance. These parameters were evaluated because of their potential to increase the exposure of an individual subject to different sources of virus, which could increase genetic variation. None of these behaviors had a significant effect on any of the genetic variation measures (data not shown), consistent with the mono-phyletic nature of the observed viruses. Time since serocon-version and disease progression status were the only two variables identified as important confounders in the univariate models (data not shown) and, thus, controlled for in the multivariate models.
Adjusted linear regression parameters
Table 5 shows the final adjusted models that describe the effect of injection-drug use frequency on genetic variation after adjusting for time since seroconversion and disease progression status. When time since seroconversion (β = 0.13, P < .001) and disease progression status (β = −2.05, P = .021) were included in the model, diversity increased by an estimated 1.21 mutations per clone (95% CI, 0.04–2.38; P = .042) for those who injected less than once daily and 2.33 mutations per clone (95% CI, 0.87–3.78; P = .002) for those who injected at least once daily, both compared with those who did not inject at all in the last 6 months. Again, the linear dose response between injection frequency category and mean genetic diversity was significant (β = +1.16; 95% CI, 0.39–1.93; P = .003; data not shown). Of the other genetic variation measures, the dS rate also showed an association with injection frequency, the significance of which depended on the specific comparisons among injection frequency categories being made (table 5 [P = .13, .08, or .02]). However, the dN rate and percentage of divergence were not significantly associated with injection-drug use frequency (table 5).
Final adjusted linear regression parameters estimates for effect of injection-drug use frequency on intravisit genetic diversity, genetic divergence, and rate of synonymous (dS) and nonsynonymous (dN) mutations during study period, reported as estimated β (95% confidence interval) for each comparison.
Final adjusted linear regression parameters estimates for effect of injection-drug use frequency on intravisit genetic diversity, genetic divergence, and rate of synonymous (dS) and nonsynonymous (dN) mutations during study period, reported as estimated β (95% confidence interval) for each comparison.
Effect of injection-drug use frequency on genetic diversity relative to diversity observed in non-ID Us
Further analyses were undertaken to define more clearly the magnitude of the effect of injection frequency on genetic diversity. These analyses were designed to express the increase in genetic diversity associated with increased injection-drug use as a proportion of the underlying mutation rate among those who did not inject in the previous 6 months. These analyses considered that the underlying mutation rate itself differed according to changes in the confounding variables, time since seroconversion, and progressor group.
First, it is estimated (table 5) that, for any progressor type and at any time since seroconversion, the mean diversity is 1.21 mutations per clone greater among those injecting <1 time daily and is 2.33 mutations per clone greater among those injecting ⩾1 times daily, each compared with those not injecting in the last 6 months. The final multivariate regression model also indicates that the mean genetic diversity is significantly greater for those in the faster progression groups (i.e., rapid progressor > moderate progressor > nonprogressor; data not shown). Because the absolute difference in diversity between injection categories is the same for any progressor type, the percentage of change would be the largest among the nonprogressors, who had the least diversity, and the smallest among the rapid progressors, who had the most diversity. Also, as time after seroconversion increases, these changes in percentage would be slightly lower because of the mean diversity significantly increasing over time.
Considering, for example, the group of rapid progressors only, the multivariate model indicates that 21.6 months after seroconversion (the median time from seroconversion for this group), the mean diversity of those not injecting drugs in the last 6 months is 5.51 mutations per clone, of those injecting<l time daily is 6.73 mutations per clone, and of those injecting ⩾1 times daily is 7.84 mutations per clone (table 6). This implies a percentage increase of 22% in the injecting <1 time daily IDUs and of 42% in the injecting ⩾1 times daily IDUs, relative to noninjectors. Again, these estimated percentage increases would be slightly higher among moderate progressors or nonprogressors.
Percentage of change in mutations per clone by injection frequency category.
Percentage of change in mutations per clone by injection frequency category.
Our study consists of 6 rapid progressors (40%), 6 moderate progressors (40%), and 3 nonprogressors (20%). The total observations (n = 78) in our study are composed of 32% rapid progressors, 51% moderate progressors, and 17% nonprogressors. Assuming this composition of subjects and observations, for the entire study population, at 21.6 months after seroconversion, the estimated mean diversity for those not injecting drugs, injecting <1 time daily, and injecting ⩾1 times daily is 3.77, 4.98, and 6.09 mutations per clone, respectively. This implies a percentage increase of 32% among those injecting <1 time daily and 62% among those injecting ⩾1 times daily, compared with those not injecting.
For a random sample of seroconverters, a more likely proportional representation of nonprogressing, moderately progressing, and rapidly progressing subjects would be 5%, 90%, and 5%, respectively [35]. Under this scenario, again at 21.6 months after seroconversion, the entire population would have an estimated mean diversity for those not injecting, injecting <1 time daily, and injecting ⩾1 times daily of 3.46, 4.68, and 5.79, respectively, reflecting a percentage increase of 35% among those injecting <1 time daily and 62% among those injecting ⩾1 times daily, compared with those not injecting.
In summary, these data indicate an estimated percentage increase in mean genetic diversity of roughly 35% among those injecting <1 time daily and 67% among those injecting ⩾ 1 times daily, relative to those not injecting in the last 6 months. These estimates would be slightly lower if rapid progressors only were considered and slightly higher if the population under study consisted of fewer rapid progressors and more nonprogressors than the hypothetical composition mentioned above. Furthermore, at a later time after seroconversion, these percentage changes would be slightly higher (data not shown).
Discussion
In a cohort of IDUs followed for up to 4 years from seroconversion, there was a highly significant correlation between frequency of injection during the previous 6 months and the genetic diversity among the HIV-1 env genes in different viral clones amplified from a subject at a given time point. After adjusting for confounding variables, the number of mutations per clone increased by an estimated 62% when individuals who had not injected in the previous 6 months were compared with individuals who injected ⩾1 per day. Despite this marked increase in genetic diversity, there was no significant association between genetic divergence from the seroconversion sequences and injection frequency. In some comparisons, increased injection frequency was significantly associated with the synonymous mutations rate, but injection frequency was never significantly associated with nonsynonymous nucleotide changes.
To understand why increased genetic diversity at a given time point is not associated with increased genetic divergence from the viruses present at seroconversion, one must consider that the pattern of evolution is influenced both by host factors, such as the immune response, and by replicative properties of the virus. A previous analysis of this cohort showed significant differences in patterns of evolution between individuals with rapid and slow disease progression [34]. Those patterns suggested that viruses were evolving under different host selection pressures in different individuals, and those differences were manifested primarily as differences in the host's tolerance for viruses with structural changes at the protein level.
This analysis of genetic evolution in IDUs examined 4 parameters: genetic diversity among clones present at a single visit, genetic divergence from the consensus sequence present at the time of seroconversion, the rate of synonymous nucleotide changes, and the rate of nonsynonymous nucleotide changes. All of these parameters, except diversity, are related to changes from the original seroconversion sequences. The parameters other than diversity are, therefore, reflective of the long-term selection pattern determined by host-virus interactions. Diversity can reflect more recent viral mutation patterns.
The significant association in this study between injection frequency and diversity suggests that increased injection frequency results in increased generation of mutations. Although this increase may result in an increase in the rate of synonymous mutations (table 5), it clearly does not alter the rate of appearance of nonsynonymous changes, which would result in changes in protein structure. Thus, injection-drug use frequency does not appear to alter the underlying host-virus dynamic that would define divergence and patterns of nonsynonymous mutation. Such a finding is consistent with studies showing that disease course is similar in IDUs and nonusers [10–13].
How might increased drug use result in increased genetic diversity? Genetic diversity in viruses sampled at a given time point can be a measure of how actively the virus is replicating or how large a virus population actually exists in the host. Rapid replication results in an increase in the number of mutations that occur as the virus undergoes more rounds of reverse transcription. Many of those mutations will give rise to replication-incompetent viruses or viruses that are eliminated by the host immune system. Thus, many of the mutations that arise from increased viral replication do not persist to contribute to increased divergence from the original infecting strain. That the increased diversity associated with increased injection frequency does not result in greater divergence simply indicates that host selection forces influencing viral evolution from the time of seroconversion are not altered by the effects of increased injection frequency.
Drugs of abuse have been shown to enhance HIV-1 replication [36–40], a process that is mediated, in vitro, by the effects of cytokines such as transforming growth factor β [40]. When injected in vivo, drugs of abuse may also contain adulterants that could stimulate immune reactivity in a manner that would enhance HIV-1 replication, as has been observed with other antigenic challenges [41, 42].
That frequency of drug use is not associated with markers of a sustained impact on viral evolution does not mean that the increased diversity associated with increased drug use is necessarily clinically unimportant. For those injecting ⩾1 times daily, the mutation frequency increased by an estimated 62%, compared with those not injecting. If this level of increase in env genetic diversity reflects changes in other genes, specifically the gene encoding viral protease and reverse transcriptase, then more frequent injection-drug use might be associated with increased appearance of mutations associated with resistance to antiretroviral therapy. For example, Killam et al. noted rapid appearance of zidovudine-resistant variants in morphine-dependent rhesus macaques, infected with simian immunodeficiency virus (SIVmac239), that were receiving no antiretroviral therapy [43]. Although the current study suggests that in clinical HIV-1 infection the underlying virus-host dynamic might select against persistence of therapy-resistant mutants, additional selective pressure introduced by antiretroviral therapy could favor outgrowth of the resistant variants. Thus, active IDUs could be at greater risk for the emergence of therapy-resistant HIV-1 variants than non-IDUs. At present, data on whether such a potential risk is actually realized in the clinical setting are not available.
