Objective. To characterize the effect of partially suppressive combination antiretroviral therapy on cerebrospinal fluid (CSF) human immunodeficiency virus (HIV)–1 RNA levels and CSF inflammation.
Design. The study was a cross‐sectional analysis of 139 HIV‐1–infected subjects without active neurological disease, categorized as having successful therapy (plasma HIV‐1 RNA level ⩽500 copies/mL), having failure of therapy (plasma HIV‐1 RNA level >500 copies/mL), or not receiving therapy. The control group consisted of 48 HIV‐negative subjects. CSF and plasma HIV‐1 RNA assays had a lower limit of quantification of 2.5 copies/mL. Genotypic resistance testing was performed on a subset of subjects.
Results. Of the 47 subjects with successful therapy, CSF HIV‐1 RNA levels were <2.5 copies/mL in 34 (72%). Only 1 had an HIV‐1 RNA level >500 copies/mL. Although plasma HIV‐1 RNA levels were similar in 35 subjects with failed therapy and 57 of those not receiving therapy (P=.84), CSF HIV‐1 RNA levels were at least 10‐fold lower in subjects with failed therapy (P<.0001). This disproportionate effect of treatment on CSF HIV‐1 RNA levels was found across the range of plasma HIV‐1 RNA levels and was not explained by differences in levels of drug resistance in plasma or CSF. Therapy reduced CSF inflammation in both treated groups.
Conclusions. In our cohort, antiretroviral therapy had a greater effect on HIV‐1 RNA levels in CSF than in plasma and reduced intrathecal inflammation, even in the presence of drug resistance.
In 20% of individuals with advanced AIDS, chronic central nervous system (CNS) exposure to HIV is eventually associated with the development of a syndrome of cognitive and motor dysfunction that was initially characterized as the AIDS dementia complex . HIV is detected in the cerebrospinal fluid (CSF) of nearly all persons infected by HIV, including early after primary infection and throughout the course of neurologically asymptomatic infection [2–5 ]. CSF HIV infection is frequently “compartmentalized,” so that viral dynamics and populations in the CSF and blood are related but differ to varying degrees, depending on the phase of systemic infection and the presence of AIDS dementia complex [6, 7].
Viral compartmentalization may lead to functional differences between virus populations in CSF and plasma that have bearing on immune recognition, coreceptor usage, antiviral drug susceptibility, and, perhaps, neuropathogenicity [8–12 ]. Furthermore, many antiviral drugs penetrate poorly through blood‐brain and blood‐CSF barriers and thus achieve low concentrations in the CNS [13–15 ]. Compartmentalized replication in the presence of low drug concentrations theoretically provides an environment that favors the further selection of resistant mutants [16–18 ]. On the basis of these issues, it has been argued that CNS infection provides a reservoir for HIV infection and that our present arsenal of antiretroviral medications may be poorly effective in treating HIV in the nervous system [19, 20].
The assumption that neurological HIV may be more difficult to treat than systemic HIV has not been borne out in practice. Observational studies have indicated a marked decrease in the incidence of AIDS dementia complex in the developed world since the introduction of combination antiretroviral therapy (ART) [21, 22]. Studies of the HIV RNA burden in CSF and plasma have indicated that HIV is generally suppressed in CSF in the setting of effective systemic ART [5, 23–25 ]. However, these studies did not address the issue of CSF responses in the setting of incompletely suppressive systemic treatment, which is widely prevalent in some populations of treatment‐experienced patients, mainly in association with resistance to antiretroviral drugs [26–28 ]. Of particular interest is the critical question of whether the CSF compartment “escapes” therapy in subjects receiving ART who do not achieve or maintain undetectable plasma HIV RNA levels.
To directly and prospectively examine how failure of ART influences the relationship between the HIV RNA burden in CSF and plasma, we established a large sentinel neurological cohort (SNC) in which HIV‐1–infected subjects were categorized into groups at entry with respect to not only whether they were receiving ART but also the effect of treatment on plasma viremia. We report the baseline analysis of the relationships between HIV‐1 RNA levels in CSF and plasma, measures of drug resistance in each compartment, and indices of CSF inflammation in this cohort.
Subjects, Materials, and Methods
Study design and participants. Study participants were enrolled in the SNC protocol between 2001 and 2005. Subjects had no active neurological diseases at entry and were assigned to 1 of 3 groups according to treatment status and baseline plasma HIV‐1 RNA levels: (1) successful therapy, subjects receiving stable combination ART for at least 3 months with baseline plasma HIV‐1 RNA levels <500 RNA copies/mL; (2) failed therapy, subjects receiving stable combination ART for at least 3 months with baseline plasma HIV‐1 RNA levels >500 RNA copies/mL; and (3) subjects who had not received ART for at least 16 weeks. All treatments and other aspects of clinical care were managed by the subjects’ health care providers and were independent of the present observational study. For analysis, we used the baseline cross‐sectional paired samples of plasma and CSF obtained at study entry. We enrolled a group of HIV‐negative participants in the study as control subjects and obtained CSF for background laboratory determinations. Protocols were approved by the University of California, San Francisco, Committee on Human Research, and informed consent was obtained from all participants.
Study procedures. All CSF was obtained for study purposes rather than for clinical diagnosis and was processed as described elsewhere [5, 29]. Subjects also underwent phlebotomy for concurrent blood sampling, along with general medical and standardized neurological assessments at the baseline visit, as described elsewhere .
Virological methods. HIV‐1 RNA levels were measured in cell‐free CSF and plasma using the ultrasensitive (measurement range, 50–75,000 copies/mL) Roche Amplicor HIV‐1 Monitor assay (versions 1.0 and 1.5; Roche Diagnostic Systems). When values were above the linear range of the assay, the standard version (measurement range, 400–750,000 copies/mL) was used for undiluted or diluted samples; when values were below the range of the assay, an ultrasensitive modification of the assay was used that has a quantitation limit of 2.5 copies/mL . HIV‐1 RNA levels were transformed to log10 values for all analyses, and the relationship between levels in plasma and CSF was analyzed as the log10 difference of these values (log10 plasma HIV‐1 HIV-1 RNS − log10 CSF HIV-1 RNA)
Drug resistance analysis. Sequencing of the HIV reverse transcriptase (RT) and protease (PR) sites from a subset of the plasma and CSF samples was performed using either the PhenoSense GT HIV assay (Monogram) or the TRUGENE HIV‐1 drug resistance assay (Bayer Healthcare). CSF and plasma‐derived RT and PR sequences were analyzed using the Stanford University HIVdb Drug Resistance Algorithm sequence analysis program (available at: http://hivdb.stanford.edu/) . For each sequenced sample, each antiretroviral drug in a patient’s regimen was weighted as 1 (high‐level resistance), 0.5 (intermediate or possible resistance), or 0 (no resistance). A total genotype sensitivity score (GSS) was calculated for each sample from treated subjects as the sum of these weighted scores, in accordance with methods used by others .
CNS‐penetration scores for antiretroviral medications. We used a drug‐penetration scoring system for each treated subject based on the number of CNS‐penetrating drugs in each participant’s regimen, similar to that used by others . Stavudine, lamivudine, abacavir, zidovudine, emtricitabine, nevirapine, efavirenz, indinavir, and PR inhibitors (except saquinavir) boosted with ritonavir were considered to be penetrating. Tenofovir, didanosine, zalcitabine, delavirdine, atazanavir, amprenavir, ritonavir, and enfuvirtide were considered to be nonpenetrating.
CSF processing and background laboratory studies. Background CSF determinations included cell counts, differentials, and albumin concentrations and were performed by the San Francisco General Hospital Clinical Laboratories. Similarly, blood CD4+ and CD8+ T lymphocyte counts were determined by flow cytometry, and blood albumin concentrations were measured at each visit. CSF neopterin levels, a measure of intrathecal macrophage activation and inflammation , and plasma neopterin levels were measured by ELISA in accordance with the manufacturer’s instructions (American Laboratory Products).
Statistical analysis. Descriptive statistics were performed using SSPS PC (version 12.0; SPSS) or Prism (version 4; GraphPad) software. When only 2 groups were being compared, comparisons were done using the Mann‐Whitney U test. Differences among ⩾3 groups were detected using 1‐way analysis of variance, with multiple‐group comparisons done using the Tukey post hoc test. Associations between HIV‐1 RNA levels in the 2 compartments were assessed using linear regression and tests of covariance.
Subject characteristics. One‐hundred thirty‐nine HIV‐1–infected participants were included (47 with successful therapy, 35 with failed therapy, and 57 not receiving therapy). Forty‐eight HIV‐negative participants served as a comparison group (table 1).
The subjects in the 4 groups were similar with respect to age and sex. Subjects not receiving therapy were more recently infected than were subjects with failed therapy, and they had higher nadir blood CD4+ T cell counts than did the other 2 HIV‐1–positive groups. Thirty‐five percent of the subjects not receiving therapy were naive for ART, and the remainder had not been receiving ART for a median of 14.5 months (interquartile range [IQR], 7.0–41.7 months). The current CD4+ T cell counts were similar in all 3 infected groups. As expected, the HIV‐negative control subjects differed from all 3 HIV‐1–infected groups in their T cell profiles. The ongoing ART regimens of the 2 treatment groups were not different with respect to the number and class of ART drugs, except for greater use of nucleoside RT inhibitors in subjects with successful therapy. Likewise, the 2 treated groups did not differ with respect to the number of CNS‐penetrating drugs (results were similar when a penetration score system similar to that in the CNS HIV Anti‐Retroviral Effects Therapy Research study was used) . The 2 treated groups also had similar durations of therapy.
Virological findings. The HIV‐1 RNA profiles in plasma and CSF and the differences between the 2 fluids differed among the infected groups (figure 1). The subjects with successful therapy differed from the other 2 groups. Their plasma virus level was, by definition, <500 copies of HIV‐1 RNA/mL (figure 1A). When ultrasensitive testing was used, CSF HIV‐1 RNA levels were consistently lower than those in plasma (figure 1B) (P<.001, paired t test, CSF vs. plasma). Although 12 (25.5%) subjects had a plasma HIV‐1 RNA level below the limit of detection when the ultrasensitive method (<2.5 copies/mL) was used, all of these, and an additional 22 of 47 subjects with successful therapy (total, 72%), had <2.5 HIV‐1 RNA copies/mL in the CSF (P<.001, χ2 test). Only 4 subjects in this group had CSF HIV‐1 RNA levels >50 copies/mL, and all of them had low but measurable HIV‐1 RNA levels in the plasma. Only 1 subject had a CSF HIV‐1 RNA level >500 copies/mL. These results indicate that virological “escape” in the CSF is unusual in the setting of systemic viral suppression.
The comparison of subjects with failed therapy and those not receiving therapy was particularly noteworthy. Although plasma HIV‐1 RNA levels were generally lower in the subjects with failed therapy than in those not receiving therapy (median, 3.92 log10 RNA copies/mL in subjects with failed therapy and 4.42 log10 RNA copies/mL in those not receiving therapy), this difference was not statistically significant (P=.84, Tukey test) (figure 1A). By contrast, the CSF HIV‐1 RNA level was much lower in subjects with failed therapy (median, 1.78 [IQR, 1.44–2.55] log10 RNA copies/mL) than in those not receiving therapy (median, 3.48 [IQR, 2.35‐4.14] log10 RNA copies/mL) (P<.0001) (figure 1B). The median within‐subject difference between plasma and CSF HIV‐1 RNA levels (i.e., log10 plasma RNA copies/ml−log10 CSF RNA copies/mL) was 0.88 in subjects not receiving therapy (IQR, 0.28–1.58) and 2.14 in those with failed therapy (IQR, 1.64–2.73) (P<.0001) (figure 1C). This relatively greater treatment‐associated reduction in CSF, compared with plasma, was maintained throughout the range of plasma HIV‐1 RNA levels (figure 1D), which indicates a fundamental difference in the relationship of CSF to plasma HIV‐1 RNA levels in subjects receiving incompletely suppressive therapy, compared with those not receiving therapy. Preliminary analysis of our longitudinal cohort showed that, in subjects with systemic virological failure, the median difference between plasma and CSF HIV‐1 RNA levels over the course of all visits remained 2.14 (IQR, 1.60–2.83), which suggests that the relationship between HIV‐1 RNA levels in these compartments in persons with failed therapy remains stable over time (data not shown).
Drug resistance. To assess the relationship between drug resistance and response to treatment, we performed genotypic resistance testing on plasma samples from a subset of 47 subjects. Genotypic resistance testing was performed on paired CSF samples from 34 of these subjects who had CSF HIV‐1 RNA levels adequate for amplification in the polymerase chain reaction (PCR) assay. Table 2 lists the major resistance mutations and GSSs in those 34 subjects for whom both fluids were tested. The proportion of subjects with at least 1 major plasma mutation was much higher in subjects with failed therapy (21/23) than in those not receiving therapy (4/21; P<.001 for the comparison, χ2 test), which suggests that virological failure in the setting of treatment was associated with drug resistance. To investigate the susceptibility of virus in each compartment to the subjects’ treatment regimens, we compared GSSs in the CSF and plasma in the 13 subjects not receiving therapy, using resistance testing, in both fluids. GSSs in the CSF and plasma were highly correlated (Spearman’s r=0.884; P<.0001), which indicates that the disproportionate impact of failed therapy on the CSF HIV‐1 RNA level was not due to increased susceptibility of CSF HIV species. Finally, we found no evidence of differential frequency of pol mutations in the 2 compartments that might affect viral fitness  and, thus, influence the relationship between HIV‐1 RNA levels in CSF and plasma in subjects with failed therapy.
Intrathecal inflammation. Both fully and incompletely suppressive therapy also had an effect on intrathecal inflammatory responses (figure 2). Whereas the subjects not receiving therapy exhibited a variable but common CSF pleocytosis (present in >50% of subjects), neither subjects with successful therapy nor those with failed therapy differed from HIV‐negative control subjects (figure 2A).
CSF neopterin levels were also reduced by treatment (figure 2B); median levels of this macrophage activation marker were lower in both treatment groups, compared with those in subjects not receiving therapy. Levels in subjects with failed therapy remained higher than those of HIV‐negative control subjects. CSF neopterin levels in subjects with successful therapy were also higher than those of HIV‐negative control subjects, although this difference was not significant, nor was it significantly different from that of subjects with failed therapy. Reductions in plasma neopterin levels were similar (data not shown). There were no differences in the CSF&rcolon;blood albumin ratios among the 4 groups (data not shown).
Our results, which demonstrate a high degree of CSF viral suppression in treated subjects who exhibit good systemic responses to therapy, agree with those of a number of reports documenting the salutary effects of combination ART on HIV‐1 levels in CSF [23–25 ]. They indicate that CSF responses are generally more favorable than those detected in plasma and do not reflect an exaggeration of systemic drug failure related to limited CNS drug penetration and increased local resistance, as has been predicted [38, 39]. This conclusion is confirmed by a comparison of CSF HIV‐1 RNA levels in subjects with failed therapy and those not receiving therapy, which revealed lower CSF HIV‐1 RNA levels in the former group, despite similar plasma HIV‐1 RNA levels in both groups. The disproportionate suppression of CSF versus plasma HIV‐1 RNA levels was further emphasized by the examination of the relationship of the CSF HIV‐1 RNA level to that of plasma, which indicated a 10‐fold greater difference in plasma and CSF HIV‐1 RNA levels in subjects with failed therapy, compared with those not receiving therapy.
This effect, which has now been demonstrated prospectively in a study designed to assess HIV‐1 RNA levels in CSF in the face of incompletely suppressive, or “failed,” therapy, is supported by our earlier observations in a convenience sample and in subjects stopping or starting treatment [5, 29]. The HIV‐1 RNA level cutoff of 500 copies/mL between the treated groups was set to allow an assessment of the difference in plasma and CSF HIV‐1 RNA levels without a “floor” effect of the detection limit on CSF HIV‐1 RNA obscuring differential effects of treatment in this compartment. The advent, after the initiation of the study, of the ultrasensitive modification of the PCR assay for HIV‐1 RNA that has a detection limit of 2.5 copies/mL partly mitigated this problem and allowed the detection of a greater effect of treatment on CSF, even in subjects with successful therapy.
These results are consistent with prior observations from our group and others. For example, a greater effect of treatment on HIV‐1 RNA levels in CSF than in plasma was evident in patients with drug‐resistant HIV who interrupted therapy. The increase in CSF HIV‐1 RNA levels was proportionally greater than the increase in plasma HIV‐1 RNA levels . We have also reported on 5 subjects with ongoing or previous treatment failure who started a new drug regimen and had either only minor or transient reductions in plasma HIV‐1 RNA levels but continued reductions in CSF HIV‐1 RNA levels . The systemic failure was due to drug resistance in plasma that either was present at baseline or emerged during treatment, yet CSF still responded.
The “disproportionate” treatment‐mediated effect on HIV‐1 RNA levels in CSF, compared with plasma, among persons with incomplete viral suppression was seen in nearly all subjects and was apparent across the range of plasma HIV‐1 RNA levels. This response is the opposite of that predicted on the basis of the limitations of treating infection within an isolated tissue compartment, including limited tissue penetration of many antiretroviral drugs and the consequent establishment of a tissue reservoir in which not only could replication continue in the face of systemic suppression but there might be an enhanced opportunity for selection of resistant mutants [16–18 ]. CSF infection is frequently compartmentalized and exhibits genetic variations in HIV populations, compared with those in plasma, that include differences in drug susceptibility [8–11 , 40]. Although these differences may occasionally lead to local treatment failure , our results suggest that it is far more common that these differences do not interfere with treatment effects in CSF.
Only a few subjects in each treatment group exhibited higher HIV‐1 RNA levels in CSF than in plasma. Among subjects with successful therapy, there were only 2 with CSF HIV‐1 RNA levels >0.5 log10 copies/mL more than levels in plasma. One of these subjects had a HIV‐1 RNA levels of 68 copies/mL in CSF and 19 copies/mL in plasma; he had a history of multidrug resistance and, at the time, was receiving zidovudine, lamivudine, efavirenz, lopinavir, and enfuvirtide. The dissociation might be related to the poor penetration of enfuvirtide , which appears to have affected his viral suppression, in association with resistance to the other drugs. The other, with a CSF HIV‐1 RNA level of 2440 copies/mL and a plasma HIV‐1 RNA level of 180 copies/mL, was receiving the relatively nonpenetrating regimen of didanosine, delavirdine, and lopinavir. One of subjects with failed therapy had a higher HIV‐1 RNA level in CSF (6910 copies/mL) than in plasma (1510 copies/mL); he was receiving only zidovudine and lamivudine and was resistant to both. These results indicate that the isolated escape of CSF HIV in the setting of treatment is unusual and generally insubstantial, although we have previously encountered subjects in whom such escape was noted . In subjects not receiving therapy, only 1 had a CSF HIV‐1 RNA level >0.5 log10 copies/mL above the level in plasma (CSF, 35,700 copies/mL; plasma, 5200 copies/mL). In noting the infrequency of CSF HIV‐1 RNA levels greater than plasma HIV‐1 RNA levels in these subjects, it is important to emphasize that the present study excluded subjects with active AIDS dementia complex, in whom an inverted difference in plasma and CSF HIV‐1 RNA levels is much more common .
Why does failed treatment have this disproportional effect on HIV‐1 RNA levels in CSF, compared with plasma? We hypothesize at least 3 contributing mechanisms. First, it is possible that CSF viruses were less resistant to treatment than were those in plasma. On the basis of our limited analysis, this seems unlikely. Although drug resistance in plasma virus explained treatment failure, we did not find higher drug susceptibility of HIV in CSF than in plasma. It should be noted, however, that our comparisons of resistance mutations was selective, because, when CSF HIV‐1 RNA levels were <1000 copies/mL, they could not be amplified for analysis. It is possible that resistant viruses are less fit to replicate in the local CSF/CNS environment than in systemic tissues because of differences in immunologic pressures or cell sources.
Second, the CSF treatment effect might be mediated through alterations in immune activation and CSF cell traffic. Previous studies have shown that failed treatment reduces T cell activation, which suggests that this reduction may be responsible for the lower rate of CD4+ T cell decreases in subjects with treatment failure [43, 44]. This might also plausibly have an effect on CSF HIV‐1 RNA levels, because T cell traffic is increased by immune activation [45, 46]. Using the heteroduplex tracking assay to compare the kinetics of plasma and CSF viral population responses to therapy, Harrington et al.  recently presented evidence that CSF HIV‐1 RNA levels in subjects with preserved CD4+ T cell levels is produced by short‐lived CSF cells that presumably traffic CD4+ T cells. These observations suggest that the local amplification of HIV contributes importantly to the CSF HIV‐1 RNA level. Thus, it is possible that the reduced immune activation that accompanies incomplete viral suppression in subjects with failed therapy might also cause reduced cell traffic, and, as a result, a lower CSF HIV‐1 RNA level. The reduction in CSF pleocytosis provides some evidence of reduced cell traffic in this group.
Third, despite lower, often “subtherapeutic” drug levels in CSF, many antiretroviral drugs may still be effective in suppressing virus production in the nervous system. This might be related to differences in intracellular drug metabolism and substrate competition that impair the replication of resistant virus in this compartment .
Our observations may have clinical implications for the long‐term prognosis of patients with partial viral suppression. Although one must still bear in mind the general caveat that CSF infection is not always indicative of the more‐important HIV encephalitis that underlies AIDS dementia complex, to the extent that CSF findings do reflect viral events in the brain, these observations provide some optimism regarding the treatment and prevention of AIDS dementia complex, even in the face of systemic treatment failure. The effects of treatment on intrathecal inflammation (CSF white blood cells) and macrophage activation (CSF neopterin) also support this therapeutic value, given that the brain injury of AIDS dementia complex is thought to result from “indirect” mechanisms involving macrophage activation and the production of endogenous neurotoxins . Indeed, these treatment effects may explain the reduction in the incidence of AIDS dementia complex since the introduction of combination ART and the anecdotal rarity of new cases of more‐severe (stage ⩾2) AIDS dementia complex in patients receiving treatment, despite the high prevalence of drug resistance and continued viremia . Importantly, these results may also have implications for the prevention of AIDS dementia complex in resource‐poor areas of the world, where treatment options after the development of drug resistance are more limited than those in the developed world.
We thank the subjects who volunteered for these studies and the staffs of the San Francisco General Hospital/University of California, San Francisco (UCSF), General Clinical Research Center and the AIDS Research Institute–UCSF Laboratory of Clinical Virology for their invaluable help.