Compartmentalized Human Immunodeficiency Virus Type 1 Present in Cerebrospinal Fluid Is Produced by Short-Lived Cells

Abstract

Human immunodeficiency virus type 1 (HIV-1) invades the central nervous system (CNS) during primary infection and persists in this compartment by unknown mechanisms over the course of infection. In this study, we examined viral population dynamics in four asymptomatic subjects commencing antiretroviral therapy to characterize cellular sources of HIV-1 in the CNS. The inability to monitor viruses directly in the brain poses a major challenge in studying HIV-1 dynamics in the CNS. Studies of HIV-1 in cerebrospinal fluid (CSF) provide a useful surrogate for the sampling of virus in the CNS, but they are complicated by the fact that infected cells in local CNS tissues and in the periphery contribute to the population pool of HIV-1 in CSF. We utilized heteroduplex tracking assays to differentiate CSF HIV-1 variants that were shared with peripheral blood plasma from those that were compartmentalized in CSF and therefore presumably derived from local CNS tissues. We then tracked the relative decline of individual viral variants during the initial days of antiretroviral therapy. We found that HIV-1 variants compartmentalized in CSF declined rapidly during therapy, with maximum half-lives of approximately 1 to 3 days. These kinetics emulate the decline in HIV-1 produced from short-lived CD4⁺ T cells in the periphery, suggesting that a similarly short-lived, HIV-infected cell population exists within the CNS. We propose that short-lived CD4⁺ T cells trafficking between the CNS and the periphery play an important role in amplifying and maintaining HIV-1 populations in the CNS during the asymptomatic phase of infection.

Human immunodeficiency virus type 1 (HIV-1) infection of the central nervous system (CNS) can cause severe neurological disease and may also result in the establishment of a unique viral reservoir that is relatively inaccessible to some antiretroviral therapies (11, 21, 29, 40). During primary infection, HIV-1 likely enters the CNS via infected monocytes and macrophages (20, 22, 39), although the virus might also cross the blood-brain barrier as cell-free virions, via infected CD4⁺ T lymphocytes, or by direct infection of brain microvascular endothelial cells (1). Several studies have suggested that macrophage-tropic forms of HIV-1 preferentially invade the brain (2, 12), and most HIV-1 detected in the brain at autopsy is found within macrophages and microglia (16, 18, 47).

The mechanisms by which HIV-1 persists in the CNS over the entire course of infection are largely unknown. Studies of HIV-1 population dynamics in the periphery during the initiation of antiretroviral therapy have provided a detailed understanding of the kinetics of virus infection in vivo and have revealed the contribution of specific cell types to the HIV-1 population in this compartment (15, 23, 31, 43, 50). Effective antiretroviral therapies block the infectious cycle by various means, depending on the drug(s) used, while cells already infected with HIV-1 can continue to produce viral RNA. As a result, the decline in HIV-1 RNA during antiretroviral therapy reflects the life span of virus-producing cells. There are at least two phases of HIV-1 RNA decay in peripheral blood plasma during antiretroviral therapy (15, 23, 50). An initial rapid decline reflects the clearance of free virus and the turnover of short-lived, productively infected CD4⁺ T cells. A subsequent, more gradual decay is thought to reflect the turnover of long-lived infected cells such as macrophages and resting CD4⁺ T cells (15, 42, 50).

We extended these analyses for the present study to include the CNS compartment to characterize cellular sources of HIV-1 and to reveal potential mechanisms of HIV-1 persistence in the CNS. A major limitation of evaluating HIV-1 population dynamics in the CNS is the inability to track viruses directly in the brains of infected subjects. Thus, it has been necessary to rely on studies of HIV-1 populations in cerebrospinal fluid (CSF) to investigate the dynamics of HIV-1 in the CNS. Several studies have validated the use of CSF as a surrogate source of virus from the brain, and CSF viral loads often predict the neurological outcome of HIV-1 infection (8, 9, 24, 46, 49). However, although CSF is an integral component of the CNS, in the context of HIV-1 infection it more accurately serves as an intermediate compartment between the brain and the periphery (7, 10, 13). Unique HIV-1 lineages evolve over time in the brain and peripheral blood, which is most likely a reflection of different selective pressures on viral replication in each compartment, whereas viral populations in the CSF appear to be genetically related to those in both the brain and the periphery (3, 48, 51). Antiretroviral therapy typically reduces the amount of bulk HIV-1 RNA in the CSF, and potent antiretroviral therapies have reduced the incidence of HIV-associated neurological disease (6, 7, 14, 19, 29). However, it is uncertain which cellular sources of HIV-1 are affected during therapy when total viral RNA levels decline in the CSF, since infected cells in both local CNS tissues and the periphery may contribute to the population pool of HIV-1 in CSF. Although it is well established that HIV-1 populations are often compartmentalized in CSF, a bulk HIV-1 RNA reduction in CSF does not necessarily indicate control of the subset of virus produced from locally infected cells in the CNS.

We utilized a sensitive gel-based system, the heteroduplex tracking assay (HTA), to distinguish between HIV-1 genetic variants in CSF arising from peripheral blood from those arising from local CNS sources. We then monitored by HTA and phosphorimaging the decline of individual HIV-1 variants during the initial days of antiretroviral therapy to characterize the infected cell types contributing to the viral populations in CSF. The HTA, a variation of the heteroduplex mobility assay first described by Delwart et al. (4, 5), is performed by first amplifying a region of the HIV-1 genome by reverse transcription-PCR (RT-PCR) and then annealing a single-stranded, radioactively labeled DNA probe corresponding to the region amplified by PCR to the PCR product. Clustered mismatches, insertions, or deletions between the probe and variants in the PCR product cause kinks and bends to form in the heteroduplexes, slowing their migration through a nondenaturing polyacrylamide gel and resulting in the display of a mixture of genotypic variants as a series of distinct heteroduplex bands. An analysis of highly variable regions of the HIV-1 genome by HTA reveals the viral population as a complex mixture of coexisting genotypes, which allows a stringent comparison of multiple populations within a single patient (17, 35, 39). Unlike conventional cloning and sequencing methods, the reproducibility of population sampling is readily validated by HTA. Therefore, HTA can be used to accurately quantify the relative contribution of individual genotypic variants to the total viral population in a biological sample, while detecting variants that comprise as little as 3% of the total population (38), making it a useful method for identifying compartmentalized HIV-1 genetic variants in CSF and tracking their decline during antiretroviral therapy.

MATERIALS AND METHODS

Study subjects, sampling, and study design.

The study subjects, plasma/CSF sampling, and medication for this study have been previously described (13). Briefly, blood plasma and CSF samples were obtained at the Vanderbilt Clinical Research Center (Nashville, Tenn.) from four asymptomatic, treatment-naïve, HIV-1-infected patients with >200 CD4⁺ T cells/mm³. Plasma samples were collected at 3-h intervals, and CSF samples were collected continuously for 48 h through lumbar intrathecal catheters. Two rounds of sampling were performed, with the first round occurring 120 to 72 h prior to therapy and the second round occurring 72 to 120 h after the initiation of therapy. Patients commenced therapy with stavudine, lamivudine, and nelfinavir at time zero. HIV-1 RNA quantification was determined with the Nuclisens nucleic acid sequence-based assay (Organon Teknika, Durham, N.C.).

RT-PCR and HTA.

Patient samples with viral RNA levels of <10,000 copies/ml were first concentrated by centrifugation at 25,000 × g for 1.5 h prior to RNA extraction and RT-PCR to allow for sufficient RNA template sampling by HTA. On therapy, CSF samples from subjects 1 and 4 were pooled and concentrated prior to RNA extraction to allow for sufficient template sampling. All RT-PCR and HTA procedures for all data described were performed independently at least twice to validate sampling by ensuring reproducibility in both the number of HTA variants detected and their relative abundance (as reported by standard deviation calculations). An RT-PCR blank product was also analyzed by HTA for each gel to identify single-stranded probe and background bands. RNA extraction, RT-PCR, and HTA methods (V1/V2 and V3) were performed as previously described (17, 27, 28). A V4/V5 HTA probe based on the NL4-3 HIV-1 clone was produced and used for HTA by similar methods to those used for the V1/V2 probe (17). The primers used to clone the NL4-3 V4/V5 region were HIVenvV4 (HXB2 7349-7378 [5′-TTTTAATTGTGGAGGGGAATTTTTCTACTG-3′]) and HIVenvV5 (HXB2 7676-7647 [5′-ATATAATTCACTTCTCCAATTGTCCCTCAT-3′]). Heteroduplexes were resolved by electrophoresis in 6% native polyacrylamide gels for V1/V2 and V4/V5 HTA and in 12% gels for V3 HTA. V3 RT-PCR variants were cloned, screened by V3 HTA to identify sequences corresponding to individual heteroduplex bands, and sequenced to predict CCR5 versus CXCR4 coreceptor usage as previously described (25-28, 39).

Phosphorimager analysis and calculations.

Phosphorimager screens were exposed to dried HTA gels, and the relative abundance of each variant detected in all samples was determined by using ImageQuant software (Molecular Dynamics, Sunnyvale, Calif.). Variant RNA concentration data were determined by multiplying the relative abundance of individual variants by the total HIV-1 RNA concentrations for those samples. Unpaired, two-tailed Student's t tests were used for statistical calculations of CNS compartmentalization and viral RNA half-lives. Half-life calculations are presented as maximum possible values because CSF samples were not collected between 0 and 72 h of therapy, a time when the half-life may have been less (13).

RESULTS

Compartmentalization of HIV-1 variants in CSF.

Blood plasma and CSF samples were obtained prior to the initiation of three-drug therapy with stavudine, lamivudine, and nelfinavir from four asymptomatic, treatment-naïve, HIV-infected subjects without AIDS. Details regarding these subjects and the serial CSF and plasma sampling technique are presented elsewhere (13). HTAs targeting the V1/V2 and V4/V5 hypervariable regions of the HIV-1 env gene were used to identify specific viral RNA variants in plasma and CSF and also to distinguish CSF variants arising from local CNS tissues from those shared with the blood. In all four subjects, the pretreatment HIV-1 populations in CSF and plasma clearly differed, although to varying degrees for different individuals (Fig. 1A). In particular, multiple variants in the CSF from each subject were either unique to CSF or were selectively enriched in CSF relative to plasma, consistent with the conception that productive viral replication in both peripheral tissues and local CNS tissues contributes to the HIV-1 population in CSF (3, 7, 13, 48).

FIG. 1.

Compartmentalization of HIV-1 in the CNS. (A) Blood plasma (P) and CSF (C) samples obtained prior to the initiation of therapy were subjected to RNA extraction, RT-PCR, and an HTA targeting the V1/V2 or V4/V5 hypervariable region of env to resolve distinct HIV-1 genetic variants. Single-stranded probe and background bands are indicated by black hexagons for the V1/V2 region for subjects 1 and 2 and the V4/V5 region for subject 3. HTAs with no single-stranded probe band shown had no detectable variant bands that ran near or above the probe; all bands shown ran below the single-stranded probe. V1/V2 probes produced from the standard HIV-1 molecular clones Ba-L (subjects 1 and 4) and JR-FL (subjects 2 and 3) were chosen based on the resolution of maximum numbers of variants. The V4/V5 probe used for all four subjects was produced from the NL4-3 HIV-1 molecular clone. (B) Phosphorimager analysis of V1/V2 and V4/V5 HTA variants detected in panel A. The relative abundance of each variant detected by HTA was determined and related to the total HIV-1 RNA in all plasma or CSF samples collected prior to therapy to calculate a value representing the relative copy number of each individual variant. Error bars represent the standard deviations from a minimum of two independent RT-PCR and HTA analyses of all pretherapy samples. HTA variants that were either unique (*) to the CSF versus plasma (and significantly above the minimum level of detection for plasma variants) or had a significantly higher copy number in the CSF than in plasma were considered to be compartmentalized in the CNS and are indicated by arrows in panel A.

The relative and absolute abundances of each HIV-1 RNA variant detected in blood plasma and CSF were then quantified (Fig. 1B). While accurately reflecting the relative abundances of detected variants, absolute variant copy numbers may be somewhat overestimated because additional minor variants may be present below the limits of HTA detection. Viral variants detected in CSF were considered to arise from infected cells compartmentalized within the CNS if the variants were either unique to the CSF or present at significantly higher copy numbers in the CSF versus plasma, because at least a portion of such variants must be produced from a local source rather than through equilibration with peripheral blood plasma. Additional HIV-1 variants in CSF were selectively enriched more than twofold in relative abundance but did not have higher copy numbers than those in plasma (subject 1, V1/V2 variant C6 and V4/V5 variants C1 and C2) (Table 1). Therefore, we cannot rule out the possibility that these variants arose entirely from a peripheral source. Sources of compartmentalized variants in CSF may include macrophages, microglia, astrocytes, and/or trafficking CD4⁺ T cells within the CNS. Viral variants identified as being compartmentalized within the CNS based on these criteria are shown in Fig. 1A. In all but one case (subject 4, V4/V5 variant C8), unique variants detected in the CSF were present at levels higher than the minimum level of detection for plasma variants (Fig. 1B). Variants that were present at higher copy numbers in the CSF versus the periphery were also selectively enriched in CSF an average of 4.3-fold in relative abundance, providing additional evidence that these variants did not arise exclusively from peripheral tissues. Remarkably, 11 to 85% of the HIV-1 in CSF was compartmentalized in the CNS based on these criteria (Table 1), suggesting that a large portion of the CSF HIV-1 population in asymptomatic subjects may arise from localized viral replication in the CNS.

TABLE 1.

Relative decay of HIV-1 variants in CSF after initiation of therapy^a

Subject	BL cell count (cells/μl)		BL HIV-1 RNA (log10 copies/ml)		% HIV-1 RNA decrease (h posttherapy)		Data for CSF variants shared with blood			Data for compartmentalized CSF variants			t1/2P valuec
	Blood CD4+	CSF WBC	Plasma	CSF	Plasma	CSF	% VL in CSFb	% Decrease in VL	t1/2 max (h)	% VL in CSFb	% Decrease in VL	t1/2 max (h)
1d,e	236	0	4.9	3.3	96 (72)	65 (85)	89	59	89	11	69	66	0.637
2	230	13	4.7	4.7	82 (72)	92 (73)	15	88	24	85	92	20	0.060
3	770	3	4.7	4.5	92 (72)	64 (73)	70	53	42	30	77	35	0.003
4d	384	9	4.6	4.1	83 (72)	81 (82)	76	81	35	24	84	31	0.397

^aAbbreviations: BL, baseline; WBC, white blood cells; VL, viral load; t_1/2 max, average maximum half-life for HIV-1 variants.

^bBased on env region that reflected the highest level of compartmentalization in CSF (V1/V2 for subjects 1 and 4 and V4/V5 for subjects 2 and 3).

^cStatistical significance of t_1/2 max between shared and compartmental variants. Based on two-tailed, unpaired Student's t test. P values of <0.050 were considered statistically significant.

^dDue to low HIV-1 RNA concentrations, on-therapy samples from subject 1 (for V4/V5) and subject 4 (for both V1/V2 and V4/V5) were pooled and concentrated to ensure reproducible sampling by HTA. The time points analyzed for viral RNA decline represent the last sample time points used for the pool.

^eNo CSF viral variants from subject 1 were compartmentalized based on the criteria described in the legend to Fig. 1. However, V1/V2 variant C6 and V4/V5 variants C1 and C2 were considered compartmentalized for the calculations shown because they were selectively enriched more than twofold in relative abundance in CSF versus plasma.

Viral populations in plasma and CSF were also analyzed by HTA targeting the V3 hypervariable region of env, followed by cloning and sequencing to predict coreceptor usage based on the V3 coding sequence. Unlike the V1/V2 and V4/V5 HTAs, HTA targeting V3 did not reveal considerable levels of compartmentalization in the CSF (Fig. 2). Although we did not detect any V3 variants unique to CSF, relatively minor differences in the relative abundances of variants in plasma and CSF were apparent, notably for subject 2. Cloning and sequence analysis of the V3 variants identified by HTA revealed no coding sequences associated with CXCR4 usage (25, 26), suggesting that all detectable variants in plasma and CSF for all four subjects were likely R5-tropic (data not shown).

FIG. 2.

Analysis of the V3 hypervariable region of env by HTA. Pretherapy RNA samples from plasma (P) and CSF (C) were subjected to RT-PCR and an HTA targeting the V3 region of env, as previously described, to reveal distinct V3 genetic populations (27, 28). All bands shown ran below the single-stranded probe band and represent distinct V3 genetic variants in the biological samples.

Dynamics of HIV-1 decline in CSF during antiretroviral therapy.

To examine the decline of compartmentalized HIV-1 RNA variants detected in the CSF, we performed V1/V2 and V4/V5 HTAs on longitudinal plasma and CSF samples obtained at baseline and during the first 72 to 120 h of antiretroviral therapy. By this analysis, slowly decaying variants will comprise an increasingly larger percentage of the total population over time. Therefore, changes in relative abundance between viral variants after the initiation of therapy reflect the relative decline of variants and the turnover rates of the infected cells from which they arise.

The concentrations of HIV-1 RNA in blood plasma or CSF typically must exceed 5,000 copies/ml for reproducible sampling by these HTAs, which is crucial for an accurate quantification of viral variants. Only subjects 2 and 3 had HIV-1 RNA concentrations above this threshold in CSF samples taken while they were on therapy (13). We first cloned RT-PCR products from the CSF of these two subjects and then screened the clones by HTA and sequencing to confirm that the identified CNS-compartmentalized variants had distinct HIV-1 env sequences (data not shown). We then tracked the relative abundances of all variants in serial plasma and CSF samples by V1/V2 and V4/V5 HTAs (Fig. 3). Surprisingly, the relative abundances of the different variants in CSF did not change appreciably after the initiation of therapy, indicating that all CSF variants declined at similar rates.

FIG. 3.

Longitudinal HTA analysis of HIV-1 in CSF and peripheral blood plasma obtained during the initial days of antiretroviral therapy. Plasma and CSF samples obtained just prior to and immediately after the initiation of antiretroviral therapy were analyzed by an HTA targeting V1/V2 or V4/V5. Sample time points are shown, with time zero representing the initiation of antiretroviral therapy. CSF variants indicated by arrows were compartmentalized relative to plasma, with asterisks noting variants unique to CSF versus peripheral blood plasma. The black hexagons denote single-stranded probe and background bands. Band intensities were maintained in samples taken after the initiation of therapy by including RNAs extracted from larger volumes of sample.

Heteroduplexes were then quantified by phosphorimager analysis, and the data were related to the total HIV-1 RNA concentration in each sample to quantify the absolute change in each CSF variant during the first 72 to 120 h of therapy (Fig. 4). As suggested by the data shown in Fig. 3, CNS-compartmentalized variants in CSF declined as rapidly as CSF variants shared with the blood. In fact, the CNS-compartmentalized variants from subject 3 appeared to decline more rapidly than variants shared with the blood.

FIG. 4.

Relative decline of individual HIV-1 genetic variants in CSF during the initial days of antiretroviral therapy. Viral genetic variants in the CSF detected by longitudinal V1/V2 and V4/V5 HTAs were quantified by phosphorimaging to monitor their relative decline during therapy. The absolute RNA concentration of each HIV-1 variant was determined by multiplying the percent abundance of the variant by the total viral RNA concentration in the sample and dividing the result by 100. Each data point represents the average copy number of an individual HIV-1 variant detected for the indicated sample time point. The 0-h time point shows the mean RNA concentrations for variants detected in all pretreatment (120 to 72 h prior to the initiation of therapy) CSF samples tested. The line segments illustrate the decline of variants from the averages at the 0-h time point to those at the first CSF sample time point after the initiation of therapy. Dashed line segments indicate CNS-compartmentalized variants. Error bars represent standard deviations for reproducibility from a minimum of two independent RT-PCR and HTA analyses conducted for all sample time points shown. Error bars at the 0-h time point represent standard deviations for all replicates of all pretreatment samples analyzed.

Samples from subjects 1 and 4 were also analyzed by RT-PCR and HTA to evaluate the relative decline of all detected variants, although in most cases reproducibility was obtained only if the samples were pooled and concentrated to increase the number of RNA templates (Table 1). As with subjects 2 and 3, all HIV-1 CSF variants from subjects 1 and 4 declined rapidly during the initial days of therapy. Unexpectedly, CSF variants that likely arose from local CNS sources declined somewhat more rapidly than CSF variants shared with peripheral blood plasma in all four subjects, although this difference was statistically significant only for subject 3 (Table 1). These findings strongly suggest that a population of short-lived HIV-infected cells within the CNS contributes the vast majority of compartmentalized virus present in CSF.

DISCUSSION

The mechanisms of HIV-1 persistence in the CNS are poorly understood, and only a few studies to date have evaluated HIV-1 population dynamics in this compartment during the initial days of antiretroviral therapy. Previous studies have consistently found that bulk CSF viral loads decline rapidly in asymptomatic subjects commencing therapy (6, 7, 13, 14), indicating that short-lived cells produce the vast majority of HIV-1 in CSF. There are at least two potential interpretations of these earlier findings. First, the vast majority of HIV-1 present in CSF during the asymptomatic stage of infection is produced by short-lived CD4⁺ T cells infected in the periphery, with the treatment of CNS-derived HIV-1 contributing little to the bulk viral RNA decline in CSF. In this case, a rapid decline of periphery-derived HIV-1 in the CSF obscures the slower decline of any CNS-derived variants. Alternatively, short-lived cells infected in both the periphery and the CNS contribute to the HIV-1 population pool in the CSF of asymptomatic subjects. Unfortunately, a measure of the overall decline of HIV-1 in CSF during therapy cannot distinguish between these two possibilities. We therefore determined whether distinct HIV-1 genetic variants are compartmentalized in CSF, which would be indicative of local production of virus in the CNS, and then measured the relative decline of individual variants during the initial days of antiretroviral therapy in four asymptomatic subjects. We found that as much as 85% of the HIV-1 population in CSF was enriched and/or unique to the CSF versus peripheral blood plasma, suggesting that a significant portion of the HIV-1 population in the CSF of asymptomatic subjects can be produced from locally infected cells in the CNS. Furthermore, these compartmentalized variants declined rapidly during the initial days of therapy, some more rapidly than CSF variants shared with the peripheral blood, suggesting that the vast majority of compartmentalized HIV-1 in CSF is produced by short-lived cells.

It is not certain which infected cell types produce the compartmentalized HIV-1 variants detected in CSF. Based primarily on autopsy studies of patients who had HIV-associated dementia, most HIV-1 in the CNS is thought to reside in relatively long-lived macrophages and microglia (18, 47). The viruses produced by such cells of the monocyte lineage would likely decay with a half-life of several weeks or more (15, 50), rather than the 1 to 3 days we observed for variants compartmentalized in the CNS. Although it is conceivable that HIV-1-infected macrophages and microglia in the CNS turn over much more rapidly than originally thought, we are not aware of any evidence in support of this speculation. Similarly, any HIV-1 that might be produced by astrocytes or neurons would decline even more slowly during the initiation of therapy since these cell types are very long-lived.

In peripheral blood, infected, activated CD4⁺ T cells rapidly turn over, and their clearance corresponds with the rapid decline in HIV-1 RNA in plasma during the initial days of therapy (15, 23, 31, 43, 50). Very few CD4⁺ T cells reside in the CNS, reflecting its relatively immunity-privileged nature. However, small numbers of CD4⁺ T cells normally migrate between the CNS and the periphery, which is important for immune surveillance in the CNS (37), and some studies have suggested a relationship between lymphocyte pleocytosis and the HIV-1 load in CSF (30, 36). Infected CD4⁺ T cells in the periphery are a likely source of HIV-1 variants shared between the CSF and plasma (Fig. 5). It is also feasible that CD4⁺ T cells trafficking into the CSF, but originally infected in the periphery, provide a source of shared or enriched variants in CSF versus plasma. This cannot explain, however, why subpopulations of CD4⁺ T cells appear to be productively infected with HIV-1 variants unique to CSF.

FIG. 5.

Model of HIV-1 dynamics in CSF of asymptomatic subjects. Our findings suggest that a population of short-lived infected cells within the CNS contributes to the vast majority of locally produced HIV-1 in CSF. We propose that short-lived CD4⁺ T cells trafficking into the CNS become infected with HIV-1 variants produced by longer-lived infected cells (i.e., macrophages and microglia) residing in the CNS, or perhaps by other locally infected CD4⁺ T cells. These newly infected T cells then serve to amplify the HIV-1 population in the CNS to high concentrations that are readily detected in the CSF (blue, CSF-compartmentalized HIV-1 variants; red, CSF HIV-1 variants shared with peripheral blood).

The most plausible model is that in asymptomatic subjects, uninfected CD4⁺ T cells migrate from peripheral tissues into the CNS, where they become infected with HIV-1 produced by long-lived macrophages and microglia residing in this compartment (Fig. 5). These newly infected CD4⁺ T cells then amplify monocyte lineage-derived, CNS-compartmentalized variants to the concentrations found in the CSF. In contrast, CNS macrophages and resident microglia do not directly contribute substantial amounts of HIV-1 RNA to the CSF pool of variants, either because these cells cumulatively produce fewer virions than CD4⁺ T cells or because viruses cannot access the CSF from deep brain tissue, where infected macrophages and microglia may reside. Unique CSF variants are unlikely to arise from CD4⁺ T cells that became infected in the periphery, since this would almost certainly make these variants also detectable in peripheral blood plasma. It is possible that CD4⁺ T cells alone could sustain the chronic replication of unique variants entirely restricted to the CNS, although this seems unlikely given the relatively small and variable number of CD4⁺ T cells in the CNS. Although coreceptor use does not necessarily predict HIV-1 neurotropism (12, 44), all viral variants detected in the plasma and CSF appeared to be R5-tropic based on their V3 coding sequences, suggesting that these variants probably have the capacity to replicate in macrophages.

The region of the CNS where CD4⁺ T cells could become productively infected with HIV-1 is not known. There was no apparent relationship between CSF leukocyte counts, HIV-1 RNA concentrations in CSF, and CNS-compartmentalized variant decay (Table 1). However, there are multiple potential routes of leukocyte entry into the CNS (i.e., not always through CSF) and several proposed mechanisms of HIV-1 neuroinvasion (1, 37), with the added complexity that HIV-1 neuroinvasion and leukocyte migration patterns may vary depending on the disease stage (32-34, 36, 41).

Consistent with this proposed model of HIV-1 compartmentalization in the CNS, a slower bulk CSF HIV-1 RNA during therapy has been observed in patients with HIV-associated dementia and/or advanced AIDS, in contrast with the case for asymptomatic patients (6, 7, 45), suggesting a greater direct contribution by infected macrophages and microglia to the HIV-1 CNS population. These cells are the key mediators of HIV-1 neuropathogenesis (21), which typically manifests clinically later during infection with the onset of immunodeficiency. Thus, it may be possible to reveal viruses produced from these long-lived cells by carrying out similar analyses with samples collected from subjects who have progressed to a more advanced disease state.

Our findings suggest that trafficking CD4⁺ T cells may play an active role in the persistence of compartmentalized HIV-1 in the CNS, further illustrating the dynamic nature of HIV-1 populations in this compartment. Although additional studies are needed to directly observe and characterize the biological consequences of the infection of trafficking CD4⁺ T cells, our findings raise the possibility that these cells may contribute to HIV-1 neuropathogenesis. Our observations also provide new insights into potential mechanisms of persistence in the CNS for other neurotropic viruses and suggest that uninfected cells migrating into various tissue compartments may play an active role in the maintenance and amplification of a compartmentalized viral population.