Persistence of HIV-associated cognitive impairment, inflammation, and neuronal injury in era of highly active antiretroviral treatment -
13 March 2011
"A significant HIV-by-age interaction effect (P value = 0.0026) was also observed for Glx/Cr in the FWM across all HIV-positive groups, independent of cognitive status. Further analysis showed that an HIV-positive individual at age 30 years would have the same value as HIV-negative individual at age 56 years"
Harezlak, Jareka; Buchthal, Stevenb; Taylor, Michaelc; Schifitto, Giovannid; Zhong, Jianhuid; Daar, Erice; Alger, Jeffreyf; Singer, Elysef; Campbell, Thomasg; Yiannoutsos, Constantina; Cohen, Ronaldh; Navia, Bradfordi; the HIV Neuroimaging Consortium
aIndiana University School of Medicine, Indianapolis, Indiana, USA
bUniversity of Hawaii, Honolulu, Hawaii, USA
cUniversity of California-San Diego Medicine Center, La Jolla, California, USA
dUniversity of Rochester School of Medicine, Rochester, New York, USA
eLos Angeles Biomedical Research Institute at Harbor-UCLA Medical Center, University of California, USA
fDavid Geffen School of Medicine at UCLA, Los Angeles, California, USA
gUniversity of Colorado Medical Center, Denver, Colorado, USA
hThe Miriam Hospital and Brown University, Providence, Rhode Island, USA
iTufts University School of Medicine, Boston, Massachusetts, USA., USA
Objective: To determine whether cognitive impairment and brain injury as measured by proton magnetic resonance spectroscopy (MRS) persist in the setting of HAART.
Design: This study is an observational cohort study.
Methods: MRS was performed in 268 patients: HIV-negative controls (N = 28), HIV-positive neuroasymptomatic individuals (N = 124), and individuals with AIDS dementia complex (ADC; N = 50) on stable antiretroviral therapy (ART) with a mean duration of infection of 12 years and CD4 cell count of 309 cells/μl. Four metabolites were measured over creatine: N-acetyl aspartate (NAA), marker of neuronal integrity; choline (Cho), myoinositol, markers of inflammation, and glutamate and glutamine (Glx) in the basal ganglia, frontal white matter (FWM), and mid-frontal cortex. Analyses included analysis of variance, analysis of covariance, linear, and nonparametric regression models.
Results: Cognitive impairment was found in 48% of HIV-infected individuals. Both HIV-positive groups showed significant increases in myoinositol/creatine or Cho/creatine in all brain regions when compared to controls; a significant decrease in Glx/creatine in the FWM was observed in the neuroasymptomatic group; and only individuals with ADC showed a significant reduction in NAA/creatine, although a significant trend for decreasing NAA/creatine in the basal ganglia was found across the groups. Effects related to aging and duration of infection, but not central nervous system penetration effectiveness were observed.
Conclusion: Brain inflammatory changes remain ubiquitous among HIV-infected individuals, whereas neuronal injury occurs predominantly in those with cognitive impairment. Together these findings indicate that despite the widespread use of HAART, HIV-associated cognitive impairment and brain injury persist in the setting of chronic and stable disease.
HIV infection is well known to cause neurological damage resulting in cognitive and behavioral impairments that constitute the AIDS dementia complex (ADC), also referred to as HIV-associated neurocognitive disorder (HAND) [1-3]. The introduction of HAART or combined antiretroviral therapy (cART) has resulted in marked improvement in survival with a substantial increase in the number of asymptomatic infected patients with improved immunological status [3,4]. Despite the reported systemic and cognitive benefits, the effects of HAART on neurological function have remained uncertain, and specifically, whether HIV-infected patients who are otherwise asymptomatic can develop brain pathology and cognitive impairment in the setting of chronic and stable disease [5-8]. Preliminary data suggest that HIV may continue to affect the brain even in the presence of HAART [9-14]. Coupled with increased survival and an aging patient population, the potential persistence of brain injury associated with HIV infection could result in an increase in the prevalence of impairment in the chronically infected and treated population.
Proton magnetic resonance spectroscopy (1H-MRS) provides a sensitive and noninvasive in-vivo method to detect inflammatory and neuronal changes in the brain [15-29]. Such changes are reflected in levels of specific cerebral metabolites, including N-acetyl aspartate (NAA), choline (Cho), myoinositol (MI), glutamate and glutamine (Glx). Before the routine use of cART, a number of MRS-derived abnormalities were described including reduced levels in the ratio of NAA to total creatine (NAA/creatine [Cr]), a marker of neuronal metabolism, particularly among patients with moderate-to-severe ADC [15-17,21-25], whereas elevations in the Cho compounds (and Cho/Cr), considered markers of cell membrane damage and MI/Cr, a glial cell marker, were reported at all stages of HIV [14-25,29]. Earlier work by these authors identified increases in MI in the frontal white matter as the primary abnormality among neuroasymptomatic individuals, whereas those with cognitive impairment showed more diffuse inflammatory changes with evidence of neuronal injury . In a subsequent study based on a factor analysis, basal ganglia and neuronal factors were identified as critical determinants of ADC [25,27], consistent with the ex-vivo findings [30-32].
To date, there have been no studies that have demonstrated the pattern or extent of brain injury in chronically infected patients on cART, particularly among neuroasymptomatic individuals and more specifically whether the previously described patterns of brain disorder have changed in response to such treatment. The HIV Neuroimaging Consortium (HIVNC) was formed to examine this issue in a prospective multicenter study of a large cohort of 300 chronically HIV-infected individuals on stable cART. Participants underwent comprehensive neurocognitive, biomarker and imaging assessments that include both volumetric imaging and MRS across seven centers. We hypothesized that HIV-infected patients with a history of chronic and stable disease on cART would continue to show patterns of brain injury similar to those observed in the pre-HAART era. Such findings would suggest that HIV infection either through direct or indirect mechanisms remains active in the brain in the setting of chronic and stable disease and would pose a continued challenge in the management of these patients. In this report, we described the findings from the multicenter MRS study. We have recently described evidence of volumetric loss based on structural imaging analysis in a subset of these patients .
This study aimed to determine whether patterns of brain injury as measured by MRS have changed in response to cART in a chronically infected and aging cohort of 268 individuals. The cohort was unique and different from those previously described due to its size (N = 240), age (40% older than 50 years), and duration of infection (60% were infected more than 10 years and nearly 10% more than 20 years), reflecting the established benefits of ART on survival. Several studies have reported that despite these benefits, individuals are facing a number of complications related to the effects of aging, chronic immune activation or treatment, including lipodystrophies, metabolic disorder, and cardiovascular disease [35,36]. More recently, there is some evidence suggesting that chronically infected yet stable individuals may be developing cognitive impairment [37,38]. Therefore, it is noteworthy that nearly 50% of our individuals showed some degree of cognitive impairment ranging from subclinical impairment in 30% to frank impairment in the remaining individuals. Further prominent increases in Cho/Cr and MI/Cr, markers of inflammation, were found in all HIV-positive groups, including neurologically asymptomatic individuals, when compared to HIV-negative controls, whereas a significant decrease in NAA/Cr, indicative of neuronal injury, was detected only among ADC patients. Combined, the current results indicate that despite effective and lasting antiretroviral treatments, cognitive and metabolite impairments, indicative of inflammation and neuronal injury, persist in the setting of chronic and stable HIV infection. Furthermore, the pattern of MRS findings has remained remarkably consistent with those observed prior to the widespread use of cART [23-25]. The results, therefore, provide compelling evidence that the early treatment benefits following the introduction of cART either do not persist or do not affect brain injury to the extent anticipated and should renew concern about the possibility of resurgence in HIV-associated brain disorder and cognitive impairment in individuals who are otherwise stable on antiretroviral treatment.
Brain inflammation in HIV
Neuropathological studies have consistently demonstrated that HIV is associated with gliosis and mononuclear cell infiltration with a particular predilection for the basal ganglia and white matter pathways [30-32]. Significant elevations in Cho and MI reflective of these inflammatory responses have been previously observed in the same regions, but less often in the cortical gray matter [16-20]. The ubiquitous presence of an inflammatory response as shown in this study continues to support prevailing models of HIV neuropathogenesis, suggesting that chronic inflammation, possibly as a result of chronic immune activation, plays a critical role in HIV-associated brain injury [31,32]. The lack of further increases in inflammation with worsening cognitive impairment suggests that a threshold level of inflammation is reached during early stages of infection. Alternatively, this finding could in part reflect the effects of decreases in inflammation with either duration of infection or aging among ADC individuals (see below). Prominent involvement of the mid-frontal cortex suggests that the pattern of brain injury may be evolving from a predominantly subcortical disorder as previously described to one which, in the setting of chronic disease and cART, now prominently includes the cortex, consistent with our own recent findings of volumetric loss and ventricular enlargement in these patients . Future studies will need to address the effects of this evolving pattern of cortical injury on cognitive performance.
Neuronal injury in HIV
Significant neuronal loss, as well as damage to the synaptic dendritic tree, is a well described feature in brains of HIV-infected patients [39-42]. It is of interest then that in the current study, reduced levels in NAA/Cr, reflecting these events, were found among patients with cognitive impairment (ADC stage ≥1). Further, the trend analysis showed decreasing levels, predominantly in the basal ganglia, suggesting that neuronal injury may also be present at subclinical stages, possibly in individuals who may be at risk for cognitive impairment. When combined with the results from the Cho/Cr and MI/Cr analysis, these findings support a two-stage model of brain injury, which suggests that inflammation during the subclinical stages of HIV infection is followed by decreases in neuronal function, which eventually lead to cognitive impairment in susceptible individuals. These studies also support the notion that a decrease in NAA/Cr may represent a critical event in HIV neuropathogenesis and provide a sensitive and useful in-vivo biomarker for cognitive impairment in the setting of chronic and stable disease .
Neurotoxicity in HIV
Glutamate and glutamine levels were assessed via the Glx peak on MRS. Glial cells and neurons are believed to be the primary sources of this metabolite; thus, disturbances could reflect dysfunction in one or both elements . Past studies have suggested that increases in glutamate resulting from glial cell dysfunction may lead to neuronal injury [43,44]. The neurotoxic hypothesis of HIV-associated brain injury would predict that increases in the Glx peak would parallel decreases in NAA and a decline in cognitive function. However, in contrast to other metabolites, no significant differences in this index were found other than a decrease in the frontal white matter in the neuroasymptomatic group. The basis for this finding is unclear, but could reflect damage to neuroglial elements at earlier stages of infection that antedate decreases in NAA as previously described in the smian immunodeficiency virus (SIV)-infected macaque .
Persistence and risk factors for HIV-associated brain injury
The reasons for the persistence of brain injury in the context of stable disease and treatment are unclear. Increases in MI and decreases in NAA have been described in antiretroviral-naive individuals during the early stages of primary HIV infection . Whether such patterns of injury persist in certain individuals or recur following treatment and whether such changes could result in chronic neurological damage over time will need to be addressed in prospective future studies such as those underway in the context of the HIVNC.
There have been no prior studies describing factors that may contribute to in-vivo patterns of HIV-associated brain injury. We, therefore, conducted a preliminary analysis of this issue. Age had a profound effect on certain metabolites, notably MI/Cr and Glx/Cr. Interestingly, Glx/Cr levels approached those observed in older healthy individuals suggesting that HIV infection may be accelerating age-related processes in the brain similar to its effects on the cardiovascular system [35,36]. The observed decreases in MI/Cr in the FWM with age and in the basal ganglia with increasing duration of disease may reflect age-related disturbances in glial cell metabolism or the effects of chronic 'burnt out' brain disease. The effects of chronic antiretroviral treatment, chronic immune activation and comorbidities as well as their interactions also need to be considered. It is noteworthy no relationship between CPE rank and metabolite levels was found suggesting that the use of CNS-penetrating agents may have little effect on brain injury associated with HIV infection and may not be as important as effective control of plasma HIV RNA levels. It remains to be determined whether the long-term administration of certain antiretroviral agents can result in neurological injury [45,46].
The current findings indicate that despite the systemic benefits of cART, the brains of HIV-infected individuals continue to show damage in the setting of chronic and stable disease. Ongoing monitoring of neurologic function, even additional therapeutic intervention, may thus be warranted in this population.
Summary statistics are provided for participants having at least one out of nine metabolite ratios available and at least one neurological evaluation. As sex and age were significantly different between HIV-negative controls and HIV-infected individuals (Table 1), all subsequent analyses were adjusted for these two factors in the linear regression models. As a group, HIV-infected individuals showed a median age of 47 years with a median duration of HIV infection of 12 years, including 59% infected for more than 10 years and 9% for over 20 years (Table 2). The median nadir CD4 count was 34 cells/μl, but the baseline count was 309 cells/μl, and the median CPE score was 1.5, comparable to scores previously reported . Nearly 52% of individuals were cognitively normal and 48% were cognitive impaired, including 28% with subclinical impairment and 20% with definite cognitive impairment as defined by ADC stages 1-3 
Comparison of magnetic resonance spectroscopy metabolite ratios between HIV-infected individuals and HIV-negative controls
Phantom evaluation was done to remove possible variability related to site differences in imaging acquisition. As the differences between the HIV-negative and HIV-infected individuals remained qualitatively the same, results from the unadjusted analyses are presented here. In the majority of cases, metabolite ratios from each of the three HIV-infected groups, including neuroasymptomatic individuals, were significantly different from the HIV-negative controls (Fig. 1, Tables 3 and 4). Notably, myoinositol/Cr in all three regions of the brain was significantly higher in all HIV-infected groups compared to the HIV-negative group, whereas Cho/Cr was significantly higher in the basal ganglia and mid-frontal cortex. Participants with subclinical impairment (ADC stage 0.5) generally showed the same pattern as the neuroasymptomatic group. In contrast, NAA/Cr in the basal ganglia and frontal white matter was significantly reduced only in individuals with cognitive impairment (ADC stage ≥1) when compared to HIV-negative controls. Glx/Cr ratio in the frontal white matter was lower in all HIV groups, but reached statistical significance only in the neuroasymptomatic group.
Comparison among HIV-infected groups
We then compared the metabolite ratios among the HIV-infected individuals with varying levels of cognitive impairment in the three regions of interest. Levels for Cho/Cr and MI/Cr did not differ among the three groups. In contrast, NAA/Cr levels in the basal ganglia were significantly lower in the advanced ADC group compared to the neuroasymptomatic group.
To further assess how metabolite levels change with respect to cognitive impairment, a trend analysis was performed in the HIV-positive group (Table 3). A prominent negative trend in the NAA/Cr was found in the basal ganglia (P = 0.001) with the neuroasymptomatic group showing the highest NAA/Cr levels, whereas levels progressively decreased with ADC stage. Weaker negative trends were observed in the frontal white matter and mid-frontal cortex, but did not reach statistical significance after correction for multiple comparisons. Consistent with the pairwise comparisons, no trends were observed with Cho/Cr or MI/Cr.
Risk factors associated with magnetic resonance spectroscopy abnormalities
Given the characteristics of the cohort and recent reports suggesting that age and treatment may play a role in certain cognitive and systemic complications , we postulated that age, duration of HIV infection, and the lack of CNS-penetrating antiretroviral agents would correlate with greater inflammation (higher Cho/Cr and MI/Cr) and neuronal injury (lower NAA/Cr). Significant age-related effects were noted for Glx/Cr, MI/Cr, and NAA/ Cr, but these results varied depending on the metabolite (Fig. 2). All groups, including HIV-negative controls, showed similar declines in NAA/Cr in the mid-frontal cortex (MFC) with increasing age (P value <0.0001). In contrast, a significant HIV-by-age interaction (P value = 0.0004) was observed with respect to MI/Cr in the FWM, which increased with age in the neuroasymptomatic and ADC stage 0.5 groups, but unexpectedly decreased with age among individuals with ADC at least 1, whereas these levels remained relatively unchanged in controls. A significant HIV-by-age interaction effect (P value = 0.0026) was also observed for Glx/Cr in the FWM across all HIV-positive groups, independent of cognitive status. Further analysis showed that an HIV-positive individual at age 30 years would have the same value as HIV-negative individual at age 56 years. A significant interaction with duration of HIV infection was found for MI/Cr levels in the basal ganglia such that individuals with subclinical neurological disease (ADC 0.5) displayed decreased MI/Cr levels with duration; those with ADC stage 1 or greater displayed an inverted 'U' relationship, that is, increased levels, which then decreased with duration, whereas neuroasymptomatic individuals showed an increase in MI/Cr (P value = 0.002). Duration was also associated with decreasing NAA/Cr levels in the MFC regardless of cognitive status (P value = 0.0065). In contrast, there was no relationship between the CPE rank and any of the metabolite ratios.