Antiretroviral Therapy and Central Nervous System HIV-1 Infection

Prevention of ADC

As for major opportunistic infections, the widespread use of HAART has markedly reduced the incidence of overt ADC (stages 1–4). In the EuroSIDA cohort, ADC was the most common severe CNS disease before HAART and showed the greatest reduction in incidence between 1994 and 2002 [17]. This decline conforms to common clinical experience in the developed world. In our practice, patients presenting with subacute or progressive cognitive-motor decline similar to that encountered before HAART are uncommon, and confined almost exclusively to untreated patients or those failing treatment because of drug resistance or nonadherence. Both the disease presentation and the contextual background of immunosuppression are anachronistic in this setting. As commented earlier, these patients may present additional diagnostic problems related to concurrent diseases and risks that obscure or confound diagnosis. The continued prevalence of cognitive-motor impairment in treated patients who have not fully recovered neurologically and the importance of milder, more indolent brain injury may become increasingly important issues.

Treatment of ADC

While it is clear that HAART can arrest ADC and reverse its neurological disability, the general magnitude of this effect is variable and not precisely defined. Since the early ACTG study (Protocol 005) showing a therapeutic benefit of zidovudine monotherapy for ADC [20], there have been few controlled treatment studies [21–24]. Most reports document only anecdotal or observational experience, or describe neurological or neuropsychological test results as secondary outcomes. In these studies, treatment regimens were usually not chosen to directly address specific neurologic issues. Nonetheless, the aggregate experience that antiretroviral therapy ameliorates ADC appears compelling and indicates that neurological dysfunction can be reversed. While at times the magnitude of recovery is dramatic, more frequently residual signs or symptoms remain. Figure 1 provides anecdotal evidence of the effects of HAART on biologic and neurologic parameters in two ADC patients in our practice [12, 25, 26]. These observations are similar to those reported by others [27].

Pathogenesis of ADC

These therapeutic effects are consistent with current views of ADC pathogenesis. While fundamental questions regarding mechanistic details remain, the principal pathological substrate of severe ADC is HIVE characterized by foci of infected macrophages and microglial, along with multinucleated-cells derived from these cells by virus-related cell fusion [8, 10, 11, 28]. Other changes include widespread white matter pallor, accompanied by microglial and astrocytic proliferation. HIV-1 brain infection manifests a predilection for the basal ganglia and deep white matter, providing a substrate for the ‘subcortical’ type of dementia found in these patients [7]. However, specialized histological techniques and electron microscopy also show neuronal abnormalities [29–31].

These pathological changes indicate that HIV-1 is the pathogenetic ‘driver’ of ADC/HIVE, and reversal of clinical disease by HAART supports this. Indeed, the extent of reversal suggests that a substantial share of neurological dysfunction relates to active, reversible toxic processes, and that the disease might be looked upon as a multifocal microscopic metabolic encephalopathy. An important factor in this is that HIV-1 infects ‘extrinsic’ brain cells, particularly macrophages and microglia of bone marrow lineage, rather than inducing direct cytolytic infection of neurons, oligodendrocytes or astrocytes, the major ‘functional elements’ of the brain. For this reason, major theories of brain dysfunction and damage have focused on ‘indirect’ mechanisms of injury involving both virus- and cell-coded signals and pathways [8, 31]. The role of macrophages is at the center of these theories, both as the major substrate for HIV-1 replication and as the source of important toxins [32, 33]. A plethora of putative signals and toxins have been implicated, and it is likely that brain injury and clinical dysfunction involve several related or overlapping pathways [34–37].

Diagnosis and Evaluation of ADC and Treatment Outcomes

Despite its basic characterization more than two decades ago [6], the nosology and diagnosis of ADC are difficult both in clinical practice and clinical trials [38–40] [24, 26]. In large part this is because ADC remains a phenotypic diagnosis that relies on recognition of the clinical syndrome and exclusion of alternative diagnoses, rather than on specific laboratory-based findings. Though slowed movements, spastic gait, and hyperactive deep tendon reflexes may be particularly suggestive of ADC in the context of an HIV-infected patient with slowed cognitive processing [6], these abnormalities are not diagnostically specific. A recently-suggested revision in research terminology, using the overall designation HIV-associated neurocognitive disorders (HAND), that relies heavily on neuropsychological testing, adds quantitative standardization to evaluation, but does not add greater specificity to diagnosis [41]. HAND also encompasses individuals with milder or absent complaints or functional impairment identified by lower test performance than ‘normals’ (see below), likely lowering the specificity of identifying HIV-1-related brain injury. Because of this reliance on neuropsychological testing to define and categorize HAND, we continue to use the ADC terminology and staging in clinical practice [42].

Neuroimaging should be obtained in the process of diagnostic evaluation in all patients with AIDS and cognitive impairment. Certain findings on computed tomography (CT) or magnetic resonance imaging (MRI) are consistent with or even suggestive of ADC, including diffuse cerebral atrophy and subcortical or periventricular white matter changes which appear hypodense on CT or bright on T2 sequences on MRI. However, neuroimaging in this setting is most useful as a means to exclude other common neurological conditions in individuals with AIDS, since neither atrophy nor white matter changes are sensitive or specific for ADC [43–45]. Among other techniques, single photon emission computed tomography (SPECT), positron emission tomography (PET), and magnetic resonance spectroscopy (MRS) have been investigated as tools for both diagnosis and pathogenetic studies of ADC [43, 45–47]. Although typical patterns of abnormalities have been documented in ADC using each method, none of these has entered into routine clinical practice. It is hoped that rapid advances in functional and anatomical neuroimaging will eventually prove more useful in diagnosis [48].

A second, related clinical shortcoming centers on the problem of distinguishing ongoing disease activity from static injury due to prior insults, caused either by HIV-1 or by other conditions. This has presented particular difficulty in structuring and evaluating clinical trials of ADC [24, 26]. Improvement in neurological function using quantitative performance testing has been the mainstay of assessing antiretroviral and other treatments [38], but more objective measures of disease activity might improve treatment assessment, just as plasma HIV-1 RNA and blood CD4+ T cell counts have made the evaluation of systemic therapies more precise [26].

These considerations have rekindled interest in the use of CSF (or blood) biomarkers and biological measures obtained by magnetic resonance methodologies [48]. These methods have the potential to advance the study of CNS HIV-1 infection and ADC, both at the clinically overt and more subtle, subclinical levels (see below). Among the promising CSF indicators are markers of neural injury and local immune activation. The axonal neurofilament light chain (NFL) protein and the neuronal protein tau are frequently elevated in CSF in patients with ADC/HIVE, may be detected before the development of ADC, and normalize with treatment [25, 49–51], while CSF immunoactivation markers such as neopterin and MCP-1 (CCL2) and CSF HIV-1 RNA can be used to measure the effects of treatment on CNS intrathecal immunoactivation and virus replication [52–56]. As these markers may be elevated in the setting of opportunistic infections of the nervous system or alternate neurodegenerative disorders besides ADC, none is specific enough to be used individually for diagnosis of ADC. However, levels of markers of neuronal injury and intrathecal inflammation may be useful in monitoring response to therapy within the nervous system once the diagnosis of ADC has been established. A combination approach using these different classes of biomarkers might override some of the issues of sensitivity and specificity that have limited previous approaches [26]. Further studies are needed to establish the applications and limitations of biomarkers for disease classification, diagnosis, and treatment assessment [48].

General character of CNS infection

Infection of the CNS is a nearly constant component of the ecology of HIV-1 infection. This has been most clearly characterized by studies analyzing the CSF. Thus, chronic HIV-1 infection is detected within the CSF of nearly all those infected – from the period of initial viremia after primary exposure, through the course of neuorologically asymptomatic infection, and, in those developing ADC [2, 4, 5, 57]. A number of studies, chiefly involving convenience cohorts, have provided a coherent picture of this infection, showing that it is nearly ubiquitous but variable in its magnitude and in its relation to plasma viral load [5, 27, 56, 58]. Figure 2 illustrates cross-sectional data derived from our own experience with 104 HIV-1 infected, untreated, neurologically asymptomatic subjects who underwent lumbar punctures in the context of research rather than for diagnosis; data from 8 untreated ADC patients are also shown [5, 56]. In neuroasymptomatic patients, the median plasma HIV-1 RNA was 4.56 log₁₀ copies/mL (intraquartile range, IQR 4.02 – 5.16), CSF HIV-1 RNA was 3.50 log₁₀ copies/mL (IQR 2.44 – 4.18), and the plasma:CSF difference was 1.07 log₁₀ copies/mL (IQR 0.33 – 1.86). Thus, in untreated individuals, the CSF HIV-1 RNA is generally about ten-fold below plasma levels, but the difference between the viral concentrations in the two fluids varies considerably. Notably, CSF HIV-1 RNA concentrations in the ADC patients were similar to those of the asymptomatic patients, particularly when considered in relation to their blood CD4+ counts. This indicates that CSF HIV-1 RNA levels cannot be used to distinguish HIV-1 infected patients who are neurologically asymptomatic from ADC patients [4, 5, 57].

Compartmentalization of CNS Infection

One of the important features of CSF infection (and CNS infection more broadly) is that its component viral populations can diverge from those of plasma. Genetic compartmentalization of CSF HIV-1 infection has been nicely demonstrated in a series of studies by Swanstrom and colleagues using the heteroduplex tracking assay (HTA) to compare blood and CSF viral populations cross-sectionally and longitudinally [3, 59, 60]. These, and companion studies of simian immunodeficiency virus (SIV) infection [61] have shown that during acute infection HIV-1 populations in CSF and blood may be identical, even as they change together over time. Subsequently, during chronic infection the populations diverge, with the greatest divergence in patients with ADC/HIVE. This general picture of progressive compartmentalization of CSF and brain HIV-1 in relation to plasma virus has also been dissected by detailed clonal analysis of viral populations [62, 63].

Functional compartmentalization has also been shown with respect to drug resistance, chemokine receptor utilization, and cell tropism. Differences in drug susceptibility between CSF and blood HIV-1 populations have been reported [64–67], and though CSF replication within an environment containing subtherapeutic drug concentrations might enhance selection of resistant mutants [68–70], this is only rarely the cause of isolated CSF viral escape [71]. While utilization of chemokine receptors by CSF and blood HIV-1 populations is most commonly concordant with a predominance of CCR5-using (R5) viruses, dual-tropic viruses can be found in the CSF, either concordant or discordant with plasma virus populations [72]. Clonal analysis of receptor utilization suggests viruses are exchanged between the two fluids in chronic asymptomatic patients [72]. Perhaps the most important question in the pathogenesis of HIV-1 associated neurological disease is whether compartmentalized viruses are selected with respect to neurotropism or neuropathogeneticity. A number of studies have suggested that certain envelope sequences may be important in ADC/HIVE, although there is conflicting data on specific genetic ‘signatures’ [73–75], and analysis for specific neuropathic sequences has not yet been used clinically.

Inflammation and immunoactivation

CSF analysis also indicates that HIV-1 infection is associated with chronic intrathecal immunoactivation-inflammation, as indicated by frequent CSF pleocytosis and elevated levels of several soluble immunological markers [5, 58, 76]. As noted in Figure 2, mild elevation of CSF white blood cells (WBCs) is common in asymptomatic infection. Eighty-five to 95 percent of these WBCs are lymphocytes, with the great majority of these being T cells and the remainder composed of monocytes. Our studies show a median CSF WBC count of 4 WBCs per μL (IQR 1 – 12) across a wide range of plasma CD4+ cells, with the exception of markedly lower WBC counts in subjects with blood CD4+ counts below 50 cells per μL. This common ‘incidental’ CSF pleocytosis is important to consider when interpreting CSF findings in HIV-1-infected individuals undergoing diagnostic lumbar puncture for various reasons, for example in the diagnosis of neurosyphilis. From our studies, we have derived some tentative ‘rules’ for interpreting CSF cell counts (Table 1). Our subjects with pleocytosis did not have headaches or any other neurological symptoms. A similar absence of symptoms was particularly striking in a series of subjects who developed pleocytosis after treatment interruption [77, 78]. These observations also raise the question of whether to use the term asymptomatic pleocytosis rather than aseptic meningitis in this context.

A number of immunological markers are elevated in the CSF in HIV-1 infection, including neopterin [79], beta-2-microglobulin (β₂M) [80], quinolinic acid [81], IP-10 (CXCL10) [82], and MCP-1 (CCL2) [52, 54], along with several others that have received less attention [83]. Frequent elevation of neopterin and IP-10 concentrations in asymptomatic subjects confirms the presence of intrathecal immunoactivation as a relatively consistent finding in asymptomatic HIV-1 infection [53, 82]. Even higher levels of some markers, including neopterin and MCP-1, in ADC/HIVE patients suggest that accelerated immunoactivation is involved in the genesis of brain injury and that these markers may be used diagnostically to identify ADC/HIVE or predict its development [52, 84, 85].

Brain injury in chronic ‘asymptomatic’ infection

The constant presence of HIV-1 and the associated immunoactivation and inflammation in the CSF raise the important question of whether chronic asymptomatic infection is accompanied by ongoing, low-grade brain injury despite the lack of overt symptoms and signs. If this chronic inflammatory state leads to brain injury, will survivors show delayed neurological impairment years later? Will it provide a foundation for greater vulnerability to other neuropathologies that accumulate with age? Importantly, would earlier treatment ameliorate subclinical neurological injury and prevent later deterioration?

Parallel studies of neuropsychological test performance in HIV-1-infected populations have raised these same questions. Indeed, diminished group performance has led to inclusion in the new HAND classification a designation: HIV-associated asymptomatic neurological impairment (ANI) [41]. In this same classification, mild neurocognitive disorder (MND) is used when similar cognitive dysfunction is accompanied by neurological symptoms, while HIV dementia (HAD) is reserved for more severe impairment, similar to the ADC designation. While in one sense this terminology and the criteria for each of these subtypes represent an advancement in formal diagnosis, they remain dissociated from pathobiologically-based disease characteristics [26, 48]. The new designations do not objectively separate any impairment unrelated to HIV-1 from that caused by the virus, and do not clearly distinguish active from static brain injury.

Further studies will be needed to determine the long-term implications of subclinical brain injury, and how to differentiate it from ‘classical’ subacute ADC [86, 87]. Most importantly, studies will need to define whether injury will occur at blood CD4+ cell levels greater than those defined by guidelines for initiating treatment, and whether the damage will continue during treatment in the presence of low-level or undetectable CSF HIV-1 RNA [88]. With respect to the role of biomarkers, it will be critical to determine whether these markers can be used to detect ongoing injury and predict future neurological deterioration.

Treatment effects on chronic CNS HIV-1 infection

In general, CSF HIV-1 infection responds very well to HAART [56, 89–91], so that when plasma HIV-1 RNA levels become undetectable, so do CSF levels. However, the relative rates of viral decay in the two compartments may differ in some patients, with HIV-1 RNA concentrations falling more slowly in the CSF than in plasma. Slower decay has been noted in subjects with ADC and lower blood CD4+ T cell counts but without CSF pleocytosis (Figure 3) [5, 27, 92, 93]. These observations can be interpreted as being consistent with a simple model of compartmentalized CSF infection, with the lag in CSF viral response due either to slow cell turnover and consequent prolonged viral release by macrophages or to less potent drug concentrations within the CSF as a result of reduced drug entry (see below). Indeed, this simple model emphasizes the potential importance of CNS drug penetration [69, 94].

Several observations, however, suggest that this simple model may not fully account for treatment effects in all settings. For example, it may not explain the overall effectiveness of a wide variety of drug regimens in suppressing CSF HIV-1 RNA [90] or why it is rare to find cases of high levels of CSF virus in the presence of suppressed plasma virus levels. The very rapid decay of CSF HIV-1, equivalent to that of plasma virus, in some subjects is also perhaps difficult to explain when component drugs penetrate the CSF so poorly that local drug concentrations are lower than that found in systemic tissues. We have noted that in the presence of systemic treatment failure related to drug resistance, CSF HIV-1 RNA levels were about 100-fold lower than those in plasma compared to untreated subjects who had a 10-fold difference between plasma and CSF levels [56]. Rather than the CSF serving as an isolated, treatment-refractory reservoir, CSF HIV-1 RNA is usually more effectively treated than plasma virus.

The reason for this ‘disproportionate CSF response’ is unclear. It may relate to enhanced intracellular drug effects on CNS cells [95]. Because failed treatment reduces generalized systemic immune activation compared to patients not receiving therapy [96], we speculate that immune activation may importantly contribute to both CSF HIV-1 RNA levels and their response to treatment [97]. Whatever the contributing mechanisms, these observations suggest the need for more complex models of infection to better characterize CSF infection and the salutary effects of HAART.

Given these uncertainties, how should treatment be tailored to CNS infection? In the neurologically asymptomatic patient, i.e., most of those initiating therapy, it is likely that the CNS warrants no special consideration. In patients presenting with symptomatic ADC or more advanced immunosuppression, the basis for recommendations are less clear. In the absence of better evidence, we generally recommend that primary consideration be given to prescribing a potent regimen that suits the individual ADC patient, using guidelines for systemic treatment, and to consider using four rather than three drugs to optimize treatment potency and accelerate viral suppression. After this, secondary consideration can be given to CNS drug penetration, with efforts to include 2 or more drugs with better than average penetration. Letendre and colleagues have proposed a simple scheme for selecting drugs for these properties, rating them as 0 (low), 0.5 (intermediate) and 1 (high) [69]. Based on a literature review, they rate: tenofovir, didanosine, zalcitibine, nelfinavir, ritonavir-boosted and unboosted saquinavir, ritonavir, boosted tiprenavir and enfuvirtide as 0; stavudine, lamivudine, emtricitabine, efavirenz, unnboosted aprenavir and fosamprenavir, and unboosted atazanavir and indinavir as 0.5; and zidovudine, abacavir, delavirdine, neveripine, and boosted aprenavir (and fosamprenavir), atazanavir, indinavir and lopinavir as 1. This list will likely be modified as more information about CNS drug penetration and effect becomes available. While representing a simple schematic, it must also be recognized that it is based largely on pharmacokinetic rather than pharmacodynamic data, and that other drug properties, such as intracellular metabolism, may importantly contribute to CNS efficacy. Hence, clinicians must recognize that the use of particular ‘CNS regimens’ or ‘neuro-HAART’ is not yet supported by clear empirical evidence.

Treatment effects on CNS immunoactivation

HAART clearly suppresses CSF immunoactivation, as indicated by reduced CSF WBC counts and immunological markers. In our series, pleocytosis was largely eliminated, not only by suppressive treatment but also by failed treatment related to drug resistance [5, 56]. This underlies the recommendation that any pleocytosis in a treated patient should be considered clinically suspicious (Table 1). CSF neopterin concentrations were also reduced by treatment, with suppressive therapy having a greater effect than failed treatment. However, these levels do not always return to normal with suppressive treatment, indicating that low-level CNS immunoactivation raising can persists despite undetectable blood and CSF HIV-1 RNA [53, 98]. Whether this might be due to continued viral replication within the CNS below the levels of detection or to residual immune activation (an immunological ‘scar’) is not yet known.

Adjuvant treatment strategies

Alternative treatment modalities have been suggested for ADC and milder cognitive impairments associated with HIV-1 [99–102]. These approaches, grouped together as adjuvant therapies, follow several rationales, including neuroprotection, immunomodulation, and symptom control. Results of a number of trials have been reviewed by Turchan and colleagues [103]. Overall, none has shown clear and substantial clinical effects. In part, this may relate to the great difficulty in structuring and implementing clinical trials to test adjuvant therapies while maintaining best available antiviral treatment [24]. However, when considered in relation to the sometimes dramatic and profound effect of HAART on untreated ADC, it is possible that these approaches are skirting the main treatment target, CNS HIV-1 infection, and are unlikely to compare favorably with direct antiretroviral therapy. On the other hand, these strategies may help illuminate the mechanisms underlying disease pathophysiology and may eventually prove to have an accessory role, either to accelerate neurological improvement when HAART is initiated or to dampen the persistent immunopathological processes that follow viral clearance. They should not be regarded as alternatives to HAART, but as supplementary therapies.

Direct drug toxicity

Peripheral nervous system and skeletal muscle toxicities of nucleoside reverse transcriptase inhibitors were well recognized from early experience. Neuropathy related to the deoxynucleosides was appreciated as a dose-limiting toxicity during initial clinical trials [104] and proved to be a common toxicity, particularly when two drugs were used in combination [105]. Similarly, toxic myopathy complicated zidovudine therapy at doses higher than those now used [106]. These direct toxicities are attributed to drug effects on mitochondrial DNA polymerase [107–109]. Fortunately, there is little to suggest that this class of drugs has similar toxic effects on the CNS. The same is true for the protease inhibitors, which do not alter CNS function. The only drug that clearly and commonly affects the CNS is the non-nucleoside reverse transcriptase inhibitor, efavirenz [110]. This most frequently manifests as vivid and, at times, dysphoric dreams, but an array of other neuropsychiatric symptoms have been reported [111]. A prospective ACTG study of 200 HIV-infected patients taking efavirenz compared to approximately 100 controls noted no difference in neuropsychological test performance or formal measures of depression or anxiety between the two arms [112]. The efavirenz group had more neurological symptoms, including bad dreams, during the first but not subsequent weeks of treatment, and 6% of patients stopped efavirenz because of CNS symptoms. While the CNS symptoms characteristically resolve spontaneously after the initial weeks of treatment, in some patients these effects either persist or are sufficiently unpleasant to necessitate changing therapy. The mechanism of toxicity is uncertain; some studies suggest a relationship to drug exposure as measured by plasma efavirenz levels [110, 113]. The ACTG study described above suggested an association between neurological symptoms in the first week of treatment and a genetically determined reduction in efavirenz metabolism [114]. One report also suggested an interactive effect of efavirenz and tenofovir on neuropsychiatric side effects [115]. Fortunately, these effects appear fully reversible with drug discontinuation. As emphasized by Cespedes and Aberg in a recent review, severe neuropsychiatric symptoms from efavirenz should be managed by drug substitution and not dose adjustment. Lowering the dose may place patients at risk for virological failure and resistance [110].

A recent report by Robertson and colleagues of an ACTG study of neuropsychological test performance in a group of subjects following treatment interruption has raised the question of whether HAART impairs neurological function [116]. The study was designed to test whether the resurgence of viremia following treatment interruption is accompanied by deterioration in neuropsychological performance. Unexpectedly, the opposite was observed – test performance actually improved after stopping therapy. One interpretation of this observation is that the subjects had, in fact, been subclinically impaired by their treatment regimens and were able to return to normal when treatment was stopped. An alternative is that the improvement was related to a ‘practice effect’ from repetition of the neuropsychological tests. To us, this is the more plausible explanation. However, because there was no simultaneous control group included in the study, final interpretation remains uncertain.

CNS HIV-1-related IRIS

Recently, there has been interest in the issue of whether IRIS provoked by CNS HIV-1 infection rather than another opportunistic pathogen may develop in patients beginning HAART [117, 118]. Context for this speculation is provided by the clinically important model of IRIS in progressive multifocal leukoencephalopathy (PML). PML is a brain opportunistic infection caused by the otherwise benign JC virus (JCV) [119]. While there is no specific treatment, clinically observations indicated that PML remitted rarely in a few HIV-1-infected patients before the advent of HAART [120], but with HAART, the disease is arrested in 50% or more of cases [121, 122]. This presumably relates to the restoration of anti-JCV immunity as part of the general immunological recovery induced by HAART. Thus, immune reconstitution is the objective of treating PML. However, an unexpected number of PML cases develop in the context of starting HAART [123, 124]. Moreover, these may have atypical features, including local edema, inflammation, and contrast enhancement on neuroimaging. Because these reactions develop in foci of PML, it is likely that JCV is the inciting antigen and that restored immunity leads to a local inflammatory response that can be extreme in some patients, to the point of causing accelerated immunopathological injury [125]. While corticosteroids are often given, no formal trial has tested this or other treatment strategies. The outcome appears to be similar in PML/IRIS to that of ‘classical’ non-inflammatory PML [126].

This intense immunologic reaction provides a possible precedent for cases of encephalitis with atypical features that develop in patients beginning HAART, in which only HIV-1 can be implicated as the inciting antigen [117, 118]. These cases may show atypical robust perivascular inflammation and leukoencephalopathy [31, 117]. Given the large number of patients who begin HAART with low CD4 counts, this is clearly a rare disorder. However, if indeed this represents CNS HIVE/IRIS, it is of considerable pathogenetic and clinical interest, and warrants special treatment approaches.