Delayed central nervous system virus suppression during highly active antiretroviral therapy is associated with HIV encephalopathy, but not with viral drug resistance or poor central nervous system drug penetration
AIDS 2003; 17(13):1897-1906
Christian Eggers; Kurt Hertogs a; Hans-Jörg Stürenburg; Jan van Lunzen b; Hans-Jürgen Stellbrink. From the Neurological Department, University Hospital Hamburg, Germany, aVirco N.V., Central Virological Laboratory, Mechelen, Belgium, and the bDepartment of Medicine, University Hospital Hamburg, Germany
Objective: HIV-1 encephalopathy (HIVE) is associated with high levels of viral RNA in the central nervous system (CNS). Highly active antiretroviral therapy (HAART) effectively reduces HIV replication in both plasma and cerebrospinal fluid (CSF). Some individuals, however, exhibit delayed CSF HIV RNA suppression in the presence of rapid plasma responses. We investigated the reasons for this discrepancy.
Design: CSF and plasma were collected prospectively in paired samples before and once or several times during HAART in 40 HIV-positive subjects. Ten had HIVE and 30 patients were neurologically asymptomatic or had non-HIVE neurological manifestations.
Methods: The slopes of viral RNA decay during HAART were compared between the compartments. The presence of HIVE was defined by clinical standards and its severity categorized according to the Memorial Sloan Kettering score. CSF and plasma levels of antiretroviral drugs were measured. Viral drug resistance during HAART in CSF and plasma was analysed both genotypically and phenotypically.
Results: Slow CSF viral decay and a high degree of compartmental discordance (slopeCSF/slopeplasma) were both significantly correlated with HIVE (P < 0.00002). There was no correlation of a rapid CSF response with Centers for Disease Control and Prevention stage, CD4 cell count, or with the number of antiretroviral compounds and their known CSF penetration. Slow CSF viral decay was associated with neither low levels of antiretroviral drugs in the CSF or plasma, nor with viral drug resistance.
Conclusions: None of the treatment-associated variables, but only the presence of HIVE, was associated with delayed virus elimination during HAART in the CSF. This suggests a distinct pattern of viral replication in the CNS in HIVE.
HIV-1 encephalopathy (HIVE) is regarded as a direct manifestation of HIV infection of the central nervous system (CNS. It is associated with high levels of virus in the cerebrospinal fluid (CSF) and in the brain parenchyma, and with ongoing replication of virus in the CNS. Strain differences between plasma virus and CNS virus as well as the lack of association between CSF viral load and disturbances of the blood-CSF barrier imply that replication of HIV in the CNS is at least partly independent from the haematolymphatic system.
The blood-brain barrier represents an obstacle to the penetration of antiretroviral drugs. As a consequence, viral replication could persist in the CNS and cause CNS disease even in the absence of detectable levels of HIV replication in the peripheral blood. Ongoing replication in the CNS (e.g., due to insufficient drug penetration) could lead to drug resistance with subsequent redissemination to the peripheral blood.
Studies investigating viral suppression in the CSF during highly active antiretroviral therapy (HAART) demonstrated that with effective suppression of plasma viraemia, CSF viral load is also well suppressed in most patients. Although the kinetics of viral decay are similar in the two compartments in most patients, some individuals show a slower elimination of HIV RNA from the CSF.
We asked the question if slower elimination of viral RNA from the CSF is due to drug resistance, insufficient drug levels, or different patterns of HIV replication in the CNS. In this study we addressed this issue by analysing viral RNA, antiviral drug resistance, and drug levels in parallel CSF and peripheral blood samples obtained longitudinally during HAART in patients with and without clinical signs and symptoms of HIVE.
Forty HIV-infected patients were studied. Twenty-seven of them were naive to antiretroviral compounds, and 13 were pretreated. Subjects were included in the latter group only if HAART had been unchanged for at least 12 weeks. Among the 10 subjects with HIVE eight were antiretroviral naive.
Lumbar puncture was performed prior to initiation or change of therapy and at variable intervals thereafter with a minimum of one follow-up lumbar puncture during HAART. Patients underwent lumbar puncture for the evaluation of neurologic manifestations of HIV infection or as part of an observational study for CNS manifestations of HIV infection approved by the local ethics committee. All subjects declared their informed consent prior to the analysis. Peripheral blood samples were obtained in parallel with lumbar puncture.
Of the 10 patients with HIVE, eight fulfilled the criteria for HIV-1-associated dementia complex. The other two had a neuro-psychiatric syndrome attributable to late stage HIV infection of the CNS (new onset psychosis with cognitive impairment and symmetric leukoencephalopathy in one; cognitive impairment with mild sensory hemisyndrome and symmetric leukoencephalopathy in the other). The severity of HIVE was quantified by the Memorial Sloan Kettering (MSK) scale. Three patients with inconclusive diagnoses and some degree of cognitive impairment insufficient to warrant the diagnosis of HIVE were given an MSK score of 0.5 (equivocal).
The patients received three to five antiviral compounds: two nucleoside analogues plus a non-nucleoside reverse transcriptase (RT) inhibitor, a protease (PR) inhibitor (with or without ritonavir boosting) or both. Treatment remained unchanged during the observation period.
Viral load in plasma and CSF was determined using the Roche Amplicor HIV Monitor quantitative PCR assay (Roche Molecular Systems, Mannheim, Germany) with a lower limit of detection of 20-50 HIV RNA copies/ml according to the manufacturer's instructions. CSF cytology, CSF white and red cell count, total protein, albumin quotient ([Alb]CSF / [Alb]Ser ? 1000), and intrathecal production of immunoglobulins G, A and M were analysed in every subject. A red cell count >20/[mu]l in the CSF samples was an exclusion criterion.
Analysis of viral kinetics
The median interval between lumbar puncture and initiation or change of HAART was 6 days (range 0-34 days). The last viral load values prior to initiation or change of HAART were carried forward as day 0.
Decay kinetics were analysed by obtaining the slope of the viral load decrease between baseline and the first value during HAART. In addition, the ratio (`discordance') of the slope in the CSF and that in the plasma (discordance = slopeCSF/slopeplasma) was calculated in order to correct for a possible bias in favour of slower kinetics due to longer time intervals between the initiation of HAART and the second lumbar puncture.
Viral elimination kinetics
HIV RNA was detectable in the plasma of all patients and in the CSF of all but two patients at baseline. While the mean slope of virus decay in the plasma and CSF was similar (-0.26 and -0.25, respectively) the variation was higher in the CSF [standard deviation (SD) plasma 0.16; SD CSF 0.24]. Plasma viral load decreased in all patients during HAART, while CSF viral load decreased in all but three patients in whom there was even an increase. The median discordance of the decay slope between the two compartments was 0.92 [95% confidence interval (CI), 0.71-1.54), indicating that the overall viral decay in the CSF compartment was slightly slower than in the peripheral blood.
The decay of CSF viral load was significantly slower in patients with HIVE (median slope, -0.003; range -0.11 to 0.03) than in patients without HIVE (median slope, -0,27; range -1.2 to -0.05; P < 0.00002, Mann-Whitney U test). The same result was obtained when patients were stratified according to the MSK score (Spearman's r, 0.63; P < 0.00002). In contrast, the plasma HIV RNA decay slope did not differ between patients with and without HIVE (median, -0.21 versus -0.26; P = 0.1); or the MSK score (r, 0.16; P = 0.33).
In order to assess if CSF HIV RNA decays exponentially as it does in the plasma, a more detailed analysis was performed in two of our patients by obtaining multiple CSF samples from a ventricular catheter. CSF virus decayed exponentially as did plasma virus.
As expected, the systematic error due to longer intervals between the initiation of HAART and the first sample during HAART led to slower estimates of viral decay. This was true for both the blood (r, 0.47) and CSF (r, 0.51) compartment. Yet, the correction for this error by calculating the median discordance showed a significant difference between the HIVE group and the non-HIVE group (0.012 ± SD, 0.28 versus 1.01 ± SD 1.22; P < 0.00002, Mann-Whitney U test). A strong correlation was also observed between the discordance and the MSK score for HIV encephalopathy (Sperman's rank: r, 0.70; P = 0.000001).
The pre-treatment CSF white cell count (± SD) was higher in the non-HIVE group (median, 10.0 ± 42.6/[mu]l) than in the HIVE group (median, 2.7 ± SD 3.9/[mu]l). The CSF white cell count and the magnitude of its change (mostly reduction) during treatment were not significantly associated with the CSF slope of virus decay (Spearman's r, -0.16 and -0.08, respectively).
The CSF slope of viral decay was the same for antiretroviral naive and pretreated patients (median slope, -0.23 for both groups).
The CD4 cell count at the start of therapy was neither significantly correlated with the slope of CSF virus decay nor with the compartmental discordance (Spearman's r, 0.01 and 0.09; respectively). Accordingly, the Centers for Disease Control and Prevention (CDC) stage had no impact on CSF viral decay kinetics.
Antiviral compounds: pharmacological properties and CSF levels
To investigate the impact of CNS drug penetration on virus elimination in the CSF we categorized the antiretroviral compounds into two groups according to their CNS penetration (and manufacturer's information) and their impact on neurological end points demonstrated in other studies. Zidovudine, stavudine, abacavir, nevirapine, efavirenz and indinavir were assigned to the group with good CNS penetration, whereas the group with poor CNS penetration comprised didanosine, zalcitabine, lamivudine, and all protease inhibitors except indinavir. There was no correlation between the number of drugs with 'good' CNS penetration and the slope of CSF viral decay or the compartmental discordance (r, -0.03; P = 0.84 for discordance).
In addition, the number of antiretroviral drugs administered - irrespective of their CNS penetration - did not correlate with the slope of viral decay in either compartment or with the compartmental discordance (r, -0.05; P = 0.77 for discordance).
To exclude individual differences in CNS penetration, the serum and CSF levels of antiviral compounds were determined in a subgroup of nine patients with slow CSF HIV RNA kinetics and in a control group of 10 patients with fast CSF HIV RNA kinetics. Groups were matched for CSF viral load, CD4 cell count, age, and albumin quotient as a measure of blood-brain barrier permeability.
Because plasma levels of antiviral compounds are known to be highly variable with different time intervals between drug intake and sampling of blood, the comparison of plasma drug levels between groups requires identical sampling protocols. However, during steady state the concentration-time curve in the CSF is much flatter than in the plasma, making the timing of sampling less important. In order to mimimize the impact of the nature of these clinical samples and the impact of differences in the concentration-time curve between CSF and blood, we regarded a calculation of the CSF : plasma ratio as inappropriate. Instead, we determined the number of the compounds administered which reached CSF concentrations above the respective IC50 levels. There was no significant correlation between the number of these drugs and the CSF slope (r, -0.22; P = 0.41) or the compartmental discordance of virus decay (r, 0.10; P = 0.73). In another approach, we calculated the ratio of the measured CSF levels to the published IC50 concentrations for the individual substances, and calculated their sum. Again, there was no significant correlation between the sum of ratios and the CSF slope (r, -0.15; P = 0.58) or the compartmental discordance (r, -0.09; P = 0,74).
Plasma virus decay was not slower in HIVE patients than in non-HIVE patients, arguing against non-adherence as a reason for the slower CSF viral decay. Furthermore, the serum drug concentrations were not lower in the HIVE patients (data not shown).
Resistance analysis was carried out prior to HAART in CSF and plasma samples in seven patients with HIVE and 10 matched patients without HIVE and rapid CSF virus decay. In two patients with slow (13 and 33) and three patients with rapid CSF viral decay (6, 18 and 40) there were mutations at positions 10, 20, 36, 71 or 77 of the pol gene possibly associated with partial resistance to ritonavir and indinavir in the CSF. Phenotypically the virus was sensitive to all protease inhibitors. Among those three subjects, in whom the CSF and plasma virus concentration was high enough to allow for sequence analysis, the mutations were identical in plasma and CSF in two patients, but in one patient resistance mutations were found in the plasma only. Patient 9 with rapid CSF viral decay had mutations associated with phenotypic lamivudine resistance identical in both compartments. In the remaining six HIVE and the five non-HIVE patients no resistance mutations were detected.
Virus load in long-term observation
Continued sampling for viral load in both compartments confirmed an incomplete virological response in the CSF in five patients with HIVE (Fig. 1). For example, in patients 35 and 31 the plasma viral load was below the limit of detection (20 copies/ml) after 16 and 58 weeks of HAART, while CSF viral load remained detectable at 25 000 and 1100 copies/ml, respectively. In patient 35 there was even an increase in CSF virus load after 3 weeks of HAART, while plasma virus load decreased during that time. Clinically, all patients improved during HAART.
In this study we analysed the reasons for inter-individual differences in the elimination of HIV RNA from the CSF during HAART, taking into account patho-physiological explanations as well as differences between virus strains in plasma and CSF and insufficient drug exposure.
In accordance with previous reports we observed parallel CSF and plasma HIV RNA decay kinetics in most patients. The variability, however, was much higher in the CSF than in the plasma. In some subjects HIV RNA increased in the CSF despite a typical decline in the plasma. Our results indicate that slow virus elimination from the CSF and a high extent of compartmental discordance of viral decay kinetics between plasma and CSF is associated with the presence and the severity of symptomatic CNS involvement. This interpretation is further supported by the persistence of low levels of CSF HIV RNA despite undetectable levels in the plasma during long-term follow-up in a subset of patients.
However, a number of potentially confounding variables have to be considered: Estimates of viral decay kinetics may be influenced by different time intervals from the beginning of HAART until the first on-treatment measurement. As a consequence, absolute numbers of virions cleared per day cannot be calculated reliably based on these values. Therefore we did not attempt to analyse the magnitude but only the discordance between viral turnover in the two compartments. We calculated a factor that we called 'compartmental discordance' (slope CSF / slope plasma). This allowed us to obtain a quantitative expression of discordant responses. Using this approach we again observed the high extent of association of discordant kinetics with both the presence of HIVE and the MSK score. With this type of analysis, different sampling intervals per se are unlikely to explain delayed viral elimination in the CSF and its association with HIVE.
In cross-sectional studies some, but not all authors found a positive correlation between CSF white cell count and CSF viral load. CSF viral decay might therefore be expected to be more rapid in subjects with higher baseline CSF cell counts. This phenomenon was described by Ellis et al. but not by Cinque et al. In our patients, the slope of the CSF viral decay was associated neither with baseline CSF cell count nor with the magnitude of its reduction during HAART, arguing against the impact of CSF cell count.
Antiviral drugs differ largely in their ability to penetrate the blood-CSF barrier. Among the different classes of compounds, protease inhibitors appear to have the lowest CSF levels. Compatible with this view, Gisolf et al. described the failure of a pure protease inhibitor-containing regimen to suppress CSF viral load sufficiently. Our observations, however, suggest that the clinical relevance of CSF penetration may be overestimated at least in the non-encephalopathic patient. Neither the number of drugs categorized as sufficiently penetrating into the CSF nor their CSF levels nor the relationship of CSF levels to IC50 values had an impact on decay kinetics. It has to be kept in mind, however, that an exposure-response relationship as established for protease inhibitor plasma concentrations and viral RNA suppression remains to be defined for the CSF.
In addition, mechanisms such as membrane-associated transporter molecules might act differently in blood and CNS, thereby reducing the value of measuring extracellular concentrations.
As resistance mutations may be observed in CSF virus and may be restricted to this compartment, we explored whether the emergence of drug-resistant virus was associated with insufficient virologic response to HAART in the CSF. Neither significant genotypic nor significant phenotypic resistance was detected in plasma and CSF prior to HAART in these mostly antiretroviral-naive patients. The failure to detect any significant resistance during HAART also strongly argues against drug resistance as the primary cause of delayed virus decay from the CSF.
As a consequence, of all the possible explanations for delayed virus decay from the CSF, symptomatic CNS HIV disease appears to be the most likely one.
What is the significance of viral decay kinetics for the biological mechanisms of HIV infection and replication in a compartment? As HAART blocks the viral life cycle by preventing the de novo infection of hitherto uninfected cells, virus detected during HAART is produced mainly by previously infected cells.
The rapid initial phase of plasma virus decline during potent HAART is attributed to the death rate of high-level virus-producing CD4 T-cells. The second phase, which is considerably slower, reflects production by longer-lived cells, e.g., monocytes and macrophages. As a consequence, in subjects with rapid CSF response to HAART, cells producing the CSF virus are probably as short-lived as those in the peripheral blood. CD4-positive T lymphocytes are the main producers of plasma virus in all stages of infection. However, they are rare in the CNS. The preferential cell types replicating HIV in the brain are macrophages and perivascular microglia. Both belong to the monocyte-macrophage lineage and are recruited to the CNS from the bone marrow. These perivascular cells undergo a turnover with a half-life in the range of several weeks to months. Theoretically, HIV infection may shorten the survival of these cells, but long-term survival and chronic virion production has been demonstrated in vitro in macrophages and microglia.
In view of the slow turnover of macrophages/microglia it appears unlikely that CSF virus is produced mainly by these cells in the setting of a rapid CSF virus decay. Rapid kinetics would rather be compatible with virus production by shorter-lived lymphocytes. According to their role in immunosurveillance of the CNS, lymphocytes are known to migrate from the peripheral blood to the CNS. Rapid kinetics would be compatible with a release of virions within the brain from migrating infected CD4 T cells. Alternatively, migrating cells could become infected within the CNS.
In contrast, slow viral decay kinetics in the CSF, as observed in our patients with HIVE, are reconcilable with a release of CSF virions from productively infected, long-lived cells such as macrophages/microglia. Virus production by these cells remains largely unaffected by antiviral drugs. The fact that in our HIVE patients the slope of initial CSF virus decay was similar to that of the second phase of plasma virus decay after the first weeks of HAART fits well into the hypothesis of virion production by long-lived cells. In view of their ability to produce HIV in the brain, macrophages/microglia are plausible candidates for the production of CSF virus of patients with delayed virus decay. This is compatible with the established features of HIVE in which high numbers of brain macrophages/microglia correlate with clinical dementia.
Neurological manifestations may also occur with acute primary HIV infection, as in two of our patients, one of whom had encephalitis. Interestingly, their CSF virus decay was as rapid as in the plasma. Slow CSF virus decay might therefore be characteristic only of chronic CNS infection.
In summary, delayed virus decay from the CSF during HAART is associated with late-stage symptomatic HIV infection of the CNS. It appears not to be associated with drug resistance or the pharmacokinetic properties of the antiretroviral compounds used. This phenomenon probably reflects host characteristics such as the cell type in which virus is produced preferentially and might reflect compartment-specific viral quasispecies distribution. Although we did not observe novel resistance mutations during HAART in slow CSF responders in this study, incomplete suppression in the CNS could facilitate the development of resistance despite profound suppression in the peripheral blood. The value of repeated CSF analysis in patients with symptomatic HIV disease of the CNS during HAART for treatment monitoring and the detection of drug resistance warrants further investigation.