HIV-associated neurocognitive disorders (HAND), consisting of asymptomatic neurocognitive impairment, minor cognitive–motor disorder, and HIV-associated dementia (HAD), affect 20%–50% of HIV-infected individuals despite widespread use of combination antiretroviral therapy (cART).1–4 HIV enters the central nervous system (CNS) via trafficking of infected monocytes and lymphocytes, and activation of monocytes/macrophages is associated with severe forms of HAND in patients with evidence of ongoing viral replication.5–10 However, little is known about mechanisms underlying milder forms of HAND in patients with cART-mediated virological suppression.
In the pre-cART era, cerebrospinal fluid (CSF) markers associated with more severe forms of HAND included CSF HIV RNA11–14 and markers indicating intrathecal immune activation (CCL2,15–19 neopterin,20–22 β-2 microglobulin,12,23 tumor necrosis factor,23–25 interleukin (IL)-6,23,24 and quinolinic acid23,24). Since widespread use of cART, the prevalence of HAD has decreased, but prevalence of milder forms of HAND has remained similar or increased.1–3,26 Furthermore, several studies show evidence of ongoing intrathecal immune activation despite cART. High CSF neopterin and β2-microglobulin, both markers of immune activation, have been found in subjects with suppressed plasma viral loads (VLs).27–30 Elevated soluble CD14 (sCD14), a marker of monocyte activation, was detected in plasma6,31 and CSF32 of HIV+ patients with neurocognitive impairment. High CSF levels of CCL2, a monocyte chemoattractant, have also been associated with cognitive impairment and altered metabolites in brain in cART-treated HIV+ patients.33,34 These studies provide evidence that immune activation continues to contribute to HAND pathophysiology in the cART era.
Recently, we found that plasma sCD14 was associated with impaired neurocognitive testing, particularly in attention and learning domains, in a cART era cohort of HIV+ subjects with advanced disease.31 Here, we examined the relationships between CSF monocyte activation markers (sCD14, CCL2, and IL-6) and neurocognitive test scores in HIV+ subjects from the same study cohort and used a multiplex assay to examine inflammatory cytokines/chemokines in plasma and CSF from subjects on suppressive cART.
MATERIALS AND METHODS
Sixty-seven HIV+ subjects (98% with nadir CD4 counts <300) with samples and data collected between 1999 and 2009 were from 4 sites (Manhattan HIV Brain Bank, National Neurological AIDS Bank, California NeuroAIDS Tissue Network, and Texas NeuroAIDS Research Center) within the National NeuroAIDS Tissue Consortium (NNTC) (n = 57) and from CNS Highly Antiretroviral Therapy Effects Research (CHARTER) (n = 10), a 6-center observational cohort study. Eighty-two percent were on cART, with 31% having undetectable plasma HIV RNA. Sixty-six subjects were examined in our previous study,31 with available CSF samples from the same time as plasma samples and neurocognitive testing. All subjects provided written informed consent under local institutional institutional review board approval. HAND clinical diagnoses were determined using established criteria35 based on formal neurocognitive testing and neurological evaluation. Neuropsychological impairment due to other causes (NPI-O) was diagnosed when factors in addition to primary HIV could contribute to neurocognitive impairment (Table 1). Current substance abuse was determined by Psychiatric Research Interview for Substance and Mental Disorders or Composite International Diagnostic Interview and urine toxicology. Subjects with severe psychiatric disorders, a confounding neurological disorder, or active systemic infection were excluded. Plasma and CSF HIV RNA were log10 transformed for statistical analysis. Undetectable plasma and CSF VL values were assigned log10 values of 2.6 or 1.7, reflecting sensitivity cutoffs of the assay; values below these cutoffs reflect lower assay sensitivity for some sites. Fifteen plasma samples from healthy donors testing HIV/hepatitis C virus (HCV) seronegative and 20 CSF samples from nondiseased controls were from Bioreclamation LLC, Westbury, NY. Nineteen HIV/HCV seronegative plasma samples were from healthy donors at Research Blood Components, Brighton, MA, with written informed consent and institutional review board approval. Because there was no clinical information available for CSF samples from Bioreclamation, CSF samples were prescreened for sCD14 and CCL2 levels by enzyme-linked immunosorbent assay to identify samples with values outside normal ranges reported in the literature (<0.25 μg/mL and <1000 pg/mL, respectively), which excluded 6 male and 10 female CSF samples with high CCL2 or sCD14 levels, respectively.
All subjects were administered an identical comprehensive test battery designed to assess 7 categories of neurocognitive function.36 Demographically corrected T global T scores were generated from individual T scores as described.31,36 T scores correlate negatively with severity of neurocognitive impairment, with values below 40 signifying impairment (40 corresponds to 1 SD of 10 from a normalized mean of 50).
sCD14, IL-6, and CCL2 were quantified by enzyme-linked immunosorbent assay (R&D Systems, Minneapolis, MN). A multiplex immunoassay (Bio-source 25-plex Human Cytokine Assay; Life Technologies, Invitrogen, Grand Island, NY) was used to measure IL-1β, IL-2, IL-4, IL-5, IL-6, IL-7, IL-8, IL-10, IL-12, IL-13, IL-17, interferon (IFN) alfa, IFNγ, tumor necrosis factor, GM-CSF, CCL2/MCP-1, CCL3/MIP-1α, CCL4/MIP-1β, CXCL10/IP-10, CXCL9/MIG, CCL11/eotaxin, and CCL5/RANTES using a Bio-Plex Workstation (BioRad, Hercules, CA).
Data were analyzed using the Mann–Whitney test and Spearman correlation. Receiver operating characteristic curves were generated to measure the ability of markers to predict neurocognitive impairment (global T score < 40) in univariate and multivariate models. Hierarchical clustering was performed with dChip software37,38 using Euclidean distance and average linkage. Comparison criteria required the fold change (FC) between group means to exceed a specific threshold, with mean difference significant by unpaired t test (P < 0.05). Concentration values were log2 transformed, and cytokines with >50% missing data were excluded from further analysis; 17 plasma and 7 CSF biomarkers met these criteria and were included for further analysis. The limit of detection (LOD) was provided by the manufacturer. In the lowest value mode,39 the lowest detected value corresponded to the mode among detectable values below the LOD. Missing values were imputed using the LOD unless >15% of samples had detectable values below the LOD; in this case, missing values were imputed using the lowest detected value. Principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA) were performed on the Metaboanalyst web portal40 using normalized and autoscaled expression values. The web portal uses the prcomp function of the stats package and plsr function of the pls package of R. Class labels were permuted 2000 times to test whether differences between groups were significant.41
Higher Plasma and CSF sCD14 Levels Are Associated With Impaired Neurocognitive Test Performance
The study cohort consisted of 67 HIV+ subjects with advanced disease (median current and nadir CD4 counts of 85 and 52 cells/μL, respectively) and high frequency of drug abuse (49%) and HCV coinfection (36%) (Table 1). The majority (82%) were treated on cART, with 31% and 46% having undetectable plasma or CSF HIV RNA, respectively (<400 and <50 copies/mL, respectively). First, we compared monocyte activation markers between HIV+ and uninfected subjects in plasma and CSF. Plasma and CSF levels of sCD14, CCL2, and IL-6 were higher in HIV+ compared with uninfected subjects (P < 0.05; see Figure, Supplemental Digital Content 1, http://links.lww.com/QAI/A316). In HIV+ subjects, HIV RNA and sCD14 were 2- to 3-fold higher in plasma compared with CSF, whereas CCL2 levels were 2-fold higher in CSF than plasma (P < 0.0001; see Figure, Supplemental Digital Content 2, http://links.lww.com/QAI/A317). CD4 count correlated inversely with plasma and CSF VL, sCD14, and plasma CCL2, whereas plasma sCD14 correlated positively with IL-6 (r = 0.248, P = 0.04; see Table, Supplemental Digital Content 3, http://links.lww.com/QAI/A318). Plasma and CSF sCD14 levels were strongly correlated (r = 0.487, P < 0.0001), whereas plasma and CSF CCL2 levels trended toward correlation (r = 0.234, P = 0.06). These results show that sCD14 levels are higher in plasma than in CSF, whereas CCL2 levels are higher in CSF than in plasma, and sCD14 levels in plasma and CSF are strongly correlated in HIV+ subjects.
To examine relationships between monocyte activation markers and HAND, we compared plasma and CSF levels of HIV RNA, sCD14, CCL2, and IL-6 with global T scores. Plasma and CSF sCD14 levels were higher in subjects with global T scores <40 (impaired) versus ≥40 (unimpaired) (P = 0.02 and P = 0.022, respectively) and correlated inversely with global T scores (r = −0.263, P = 0.03 and r = −0.282, P = 0.02, respectively; Fig. 1). In contrast, plasma and CSF CCL2 or IL-6 levels were similar between groups. Receiver operating characteristic analysis for single markers demonstrated that sCD14 levels in plasma and CSF yielded higher area under the receiver operating characteristic curve (AUROC) values for classification of global T scores <40 (0.657 and 0.659, respectively) compared to 4 conventional markers (plasma and CSF VL and current and nadir CD4 count; AUROC = 0.541, 0.566, 0.607, and 0.598, respectively). In multivariate models for predicting T score <40, plasma and CSF sCD14 added incremental value to logistic regression models using age and cART and improved AUROC values (by 0.12 and 0.14, respectively) when added to age and cART, whereas the 4 conventional markers did not improve the predictive ability of these models. Thus, elevated CSF sCD14 levels are associated with global T scores, indicating neurocognitive impairment, and yield higher AUROC values for predicting impaired global T scores than 4 conventional markers in univariate and multivariate models.
Plasma and CSF sCD14, CCL2, and IL-6 Levels Do Not Distinguish Between Groups Defined by HAND Clinical Diagnoses
We compared global T scores between subjects stratified into 2 groups: no neurocognitive impairment and HAND or NPI-O clinical diagnosis grouped together. As expected, global T scores were higher in subjects with no neurocognitive impairment compared to those with an HAND or NPI-O clinical diagnosis (P < 0.0001; see Figure, Supplemental Digital Content 4, http://links.lww.com/QAI/A319). We compared levels of plasma and CSF sCD14, CCL2, and IL-6 between the 2 groups but found no differences (see Figure, Supplemental Digital Content 5, http://links.lww.com/QAI/A320). Thus, plasma and CSF sCD14, CCL2, and IL-6 did not distinguish between subjects defined by HAND clinical diagnoses.
Plasma and CSF sCD14 Levels Are Associated With Global T Scores Indicating Neurocognitive Impairment in Viremic but Not Aviremic HIV+ Subjects
Next, we compared relationships between monocyte activation markers and global T scores in plasma and CSF for subjects grouped according to detectable (≥400 HIV RNA copies/mL) or undetectable (<400 HIV RNA copies/mL) plasma VL. These groups had similar median age, but median current and nadir CD4 counts were lower in viremic compared with aviremic subjects (54 and 41 versus 249 and 71 cells/μL, P < 0.0001 and P < 0.05, respectively). sCD14, CCL2, and IL-6 remained elevated in plasma and CSF in both viremic and aviremic subjects compared with controls (Fig. 2A). Conversely, CCL2 and IL-6 were higher in CSF than plasma of aviremic subjects. Plasma sCD14 correlated with CSF sCD14 (r = 0.574, P < 0.0001), whereas plasma and CSF sCD14 correlated inversely with global T scores (r = −0.354, P = 0.017 and r = −0.297, P = 0.047, respectively) in viremic but not aviremic subjects (Figs. 2B, C). In aviremic subjects, CD4 count correlated inversely with CSF CCL2 levels (r = −0.598, P = 0.003), whereas plasma CCL2 correlated with CSF CCL2 (r = 0.520, P = 0.018). Unexpectedly, aviremic subjects with global T scores <40 had lower CSF CCL2 compared to those with global T scores ≥40 (P = 0.005). Because illicit drug use and HCV coinfection are comorbidities of HIV infection with immunomodulatory effects that may influence risk of HAND,42–44 we examined performed subgroup analysis. Neither CSF nor plasma biomarkers levels showed significant differences when subjects were classified according to patterns of substance abuse (within 6 months) (opiate or cocaine users compared with nonusers, n = 19, 21, and 27, respectively) or HCV coinfection (data not shown). Thus, plasma and CSF sCD14 levels correlated inversely with global T scores in subjects with detectable but not undetectable plasma VL.
Central Nervous System Penetration Effectiveness Score Is Not Associated With Differences in CSF Biomarker Levels
To examine relationships between CNS penetration of cART regimens and biomarker levels, we compared CSF sCD14, CCL2, and IL-6 levels with central nervous system penetration effectiveness (CPE) score, which assigns each drug a value of 0 (low), 0.5 (intermediate), or 1.0 (good penetration); scores are summed to determine overall CPE rank of a regimen.45 The median value for 2008 CPE rank (2.0) was the cutoff for comparing biomarker levels in subjects with good (≥2) versus poor (<2) CPE; scores were then evaluated as continuous variables compared with CSF sCD14, CCL2, and IL-6 levels. Additionally, data were analyzed using the revised CPE 2010 scoring system.46 No associations were found between CPE rank and CSF sCD14, CCL2, and IL-6 levels or global T scores, using either CPE scoring system. Subjects with suppressed plasma VL (<400 copies/mL) also showed no significant association between CPE rank and CSF biomarker levels or global T scores. Thus, higher CPE scores did not seem to influence CSF monocyte activation markers or global T scores in this cohort.
Increased CSF Inflammatory Biomarkers Distinguish HIV+ Subjects on Suppressive cART From Uninfected Controls Regardless of Cognitive Status
To examine inflammatory biomarker patterns in plasma and CSF of HIV+ subjects with cART-mediated virological suppression, we used a multiplex assay to measure 18 plasma and 7 CSF biomarkers in 14 aviremic HIV+ subjects and 14 healthy uninfected controls. Univariate analysis identified 8 cytokines/chemokines (sCD14, IL-6, IL-8, CCL2, CXCL9, CXCL10, IL-2R, CCL11, and CCL3) with higher levels in plasma from aviremic HIV+ subjects compared with healthy controls (FC 1.37–4.38, P < 0.05; see Table, Supplemental Digital Content 6, http://links.lww.com/QAI/A321), whereas CCL4 showed a trend toward significance (FC 2.6, P = 0.054). Similar analysis of matched CSF samples detected 7 cytokines/chemokines (CXCL10, CCL3, CCL2, IL-8, IFNγ, IL-6, and sCD14) at higher levels in aviremic HIV+ subjects compared with healthy controls (FC 1.82–13.09, P < 0.05; see Table, Supplemental Digital Content 6, http://links.lww.com/QAI/A321). IL-8, IFNγ, and CCL2 were the top-ranked CSF biomarkers distinguishing aviremic HIV+ subjects from healthy controls (FC 2.89–13.09, P < 0.0001).
Next, we examined whether aviremic HIV+ patients clustered according to clinical subgroups and covariates defined by neurocognitive diagnosis, global T score (impaired or unimpaired, corresponding to <40 and ≥40, respectively), and HIV disease markers. Supervised hierarchical clustering for 18 plasma and 7 CSF biomarkers across aviremic HIV+ subjects and healthy controls showed that none of these biomarkers in plasma and CSF clustered aviremic HIV+ subjects according to neurocognitive diagnosis, global T score, or current or nadir CD4 counts. When analysis was applied to only 7 plasma biomarkers detected in CSF, we identified a major cluster of 11 of 14 controls with low rate of misclassification (14.2%), second cluster of 6 aviremic HIV+ subjects, and third cluster with 6 aviremic and 3 control subjects (Fig. 3A). Unsupervised hierarchical clustering of 7 CSF biomarkers revealed that aviremic HIV+ subjects and healthy controls segregated with 100% accuracy into 2 major clusters (Fig. 3B). The aviremic HIV+ subject cluster separated into 2 subclusters based on differential CSF CCL3 levels, each subcluster consisting of 7 aviremic subjects including various HAND diagnoses. IL-8 and IFNγ levels were higher (FC 3.58, P = 0.001 and FC 1.45, P = 0.023, respectively), whereas CCL2 trended toward significance (FC 1.99, P = 0.07), in the subgroup with higher versus lower CSF CCL3 levels. These findings show that sCD14, IL-6, CCL2, CCL3, CXCL10, IL-8, and IFNγ are increased in CSF from aviremic HIV+ subjects compared with controls in a small study regardless of neurocognitive status and identify IL-8, CCL2, and IFNγ as a CSF biomarker cluster.
Next, we performed PCA and PLS-DA to examine variance in the CSF biomarker data. Unsupervised analysis of 7 CSF biomarkers by PCA revealed that the first 3 components explain 88.7% of the variance and discriminate aviremic HIV+ subjects from controls with 100% accuracy (Fig. 3C). Supervised analysis by PLS-DA revealed that aviremic HIV+ subjects and controls could be separated along the axis defined by 3 components explaining 83.9% of the variance (Fig. 3D); permutation tests confirmed significant separation (P < 0.0005). Variable importance in projection scores plot ranked IL-8, IFNγ, and CCL2 as the top 3 biomarkers explaining variance between the 2 groups (Fig. 3D). For plasma biomarkers, PCA did not reveal significant discrimination between aviremic HIV+ subjects and controls, whereas PLS-DA predicted 56% of the variance along the axis defined by 3 components with significant separation (P = 0.019). We further examined IL-8, IFNγ, and CCL2 by Spearman correlation; in aviremic HIV+ subjects, IL-8 correlated positively with IFNα and CCL2 in plasma and with IFNγ and CCL2 in CSF (Fig. 4). Additionally, CCL3 levels correlated with IL-8 and IFNγ (r = 0.815, P = 0.0003 and r = 0.728, P = 0.003, respectively) in CSF. These findings suggest that increased CSF sCD14, IL-6, CCL2, CCL3, CXCL10, IL-8, and IFNγ distinguish HIV+ patients on suppressive cART from uninfected controls.
In this study of HIV+ subjects with advanced disease, CSF sCD14 levels were higher in subjects with impaired neurocognitive test performance and correlated inversely with global T scores. CSF CCL2 and IL-6 did not show these associations. AUROC values for predicting impaired global T score <40 were higher for plasma and CSF sCD14 than for current or nadir CD4 count or CSF or plasma VL and remained significant after adjusting for age and cART. The predictive ability of CSF sCD14 in single-marker and multivariate models was at least as good as plasma sCD14 for predicting impaired global T scores. These findings extend our previous analysis of plasma sCD14 in the study cohort31 and suggest that CSF sCD14 is a useful biomarker to monitor intrathecal inflammation during HAND progression and therapeutic responses.
Despite cART-mediated virological suppression, monocyte/macrophage activation markers (eg, neopterin, sCD14, CCL2, CCL3, CCL4, CCL5, and CXCL10) are frequently detected in CSF.6,27,29,30,33,34,47–50 In the present study, plasma and CSF sCD14 levels were associated with global T scores in subjects with detectable but not undetectable VL. Furthermore, increased CSF sCD14, IL-6, CCL2, CCL3, CXCL10, IL-8, and IFNγ robustly distinguished aviremic HIV+ subjects from controls regardless of cognitive status in a small study of 28 subjects. In unsupervised analyses, these CSF biomarkers discriminated aviremic HIV+ subjects from controls with 100% accuracy, with IL-8, CCL2, and IFNγ explaining most of the variance between groups. IL-8, produced mainly by activated monocytes and natural killer cells, is a chemoattractant for neutrophils, T cells, and natural killer cells expressing CXCR1 or CXCR2. In a recent study, increased CSF IL-8, CCL2, and CXCL10 were strongly associated with cerebral metabolites related to neuronal injury and neuroinflammation in subjects with HAND.34 Subgroup analysis of untreated subjects or those failing cART indicated no correlation of IL-8, CCL2, and CXCL10 with inflammatory pattern scores related to neuronal injury, suggesting that cART alters relationships between these chemokines and cerebral metabolites. Consistent with these findings, we found clustering of CSF IL-8, CCL2, and IFNγ in heat maps and positive correlations of CSF IL-8 levels with IFNγ and CCL2 in subjects on suppressive cART. These findings are consistent with a model in which IFNγ-driven pathways contribute to ongoing intrathecal immune activation in HIV+ patients despite cART-mediated virological suppression.
Another interesting finding was the identification of 2 clusters of aviremic HIV+ subjects distinguished by differential CSF CCL3 levels: one with higher CCL3 levels consisting of 5 of 7 subjects with impaired global T scores along with higher CSF IL-8/IFN-γ levels and an another cluster with lower CCL3 levels consisting of 5 of 7 subjects with unimpaired T scores. Although the small sample size limits conclusions that can be drawn, this preliminary finding, together with previous studies implicating increased CSF CCL3 in HAND pathogenesis,51,52 suggests that CCL3 as a potential CSF biomarker that may distinguish clinical subgroups warrants further investigation. Further studies of larger cohorts of aviremic subjects followed longitudinally are needed to determine relationships of CSF IL-8, CCL2, and IFN-γ (and CCL3) to risk of developing HAND in HIV+ patients on suppressive cART and their utility as biomarkers to monitor intrathecal inflammation and therapeutic responses.
Although we found an association between CSF sCD14 levels and impaired neurocognitive test performance, we found no differences in plasma or CSF sCD14 levels between HIV+ subjects classified by HAND diagnosis. A previous study of CD14+/CD16+ monocytes in patients with HAD32 found higher levels of sCD14 in CSF from HIV+ individuals with HAD compared to controls, but differences were significant only for those with a diagnosis of moderate to severe dementia and sample sizes were small. We cannot exclude the possibility that our inability to detect differences in plasma or CSF biomarkers between HAND subtypes reflected methodological problems related to patient selection or criteria used for assigning a clinical diagnosis. Nonetheless, our findings suggest that T scores are more sensitive indicators of neurocognitive impairment than clinical diagnoses in the cART era, and using continuous descriptors rather than categorical diagnoses is important for demonstrating associations between biomarker levels and impaired neurocognitive function.
The relationship of CPE score to improvement in clinical outcome is not linear and might be explained by a delicate balance between viral suppression in the CNS and potential neurotoxicity of certain antiretroviral drugs.46 We found no association between CPE score and CSF sCD14, CCL2, or IL-6 levels for the total cohort or aviremic subgroup. Similarly, a recent magnetic resonance spectroscopy study did not find an association between CPE score and commonly measured cerebral metabolites.53 Although interpretation of our findings is limited by the small sample size, they are indicative of ongoing intrathecal inflammation in HIV+ patients regardless of CPE score.
Limitations of our study include its cross-sectional design and small sample size, which may have limited the power to detect significant associations. Also, narrow selection criteria used to define the study cohort (CD4 nadir < 300) limit our findings to those with advanced HIV disease. We included NPI-O subjects, as many likely exhibit neurocognitive deficits attributable to HIV, and there is site-to-site variation in assigning a diagnosis of NPI-O. Subgroup analysis by type of illicit drug use or HCV coinfection did not show significant differences in plasma or CSF sCD14, CCL2, or IL-6, consistent with other studies that failed to demonstrate associations between these comorbidities and elevated sCD14 or CCL2.5,54 Thus, these comorbidities are unlikely to account for associations between elevated sCD14 and impaired global T scores. The study cohort was from NNTC, which specifically recruits individuals with advanced disease, and CHARTER, which includes a large population of well-controlled HIV+ subjects, to represent a diverse population of HIV-infected individuals with a broad range of VLs. As such, the study cohort reflects a bias of urban cohorts with large populations of nonsuppressed patients and intravenous drug users and results cannot be generalized to all populations.
In conclusion, plasma and CSF sCD14 levels are associated with impaired neurocognitive test scores in HIV+ patients on nonsuppressive cART, providing evidence that monocyte activation continues to contribute to HAND pathogenesis in the cART era. CSF IL-8, CCL2, CCL3, CXCL10, IFN-γ, and IL-6 levels remain elevated in patients on suppressive cART, even in subjects without cognitive impairment. Plasma and CSF sCD14 may be particularly useful as biomarkers to monitor systemic and intrathecal monocyte activation, HAND progression, and therapeutic responses in HIV+ patients on cART.
The authors thank NNTC and CHARTER sites for providing plasma samples and clinical data for patients with AIDS. The authors also acknowledge support from the Harvard Center for AIDS Research Biostatistics Core and the Mount Sinai Institute for NeuroAIDS Disparities.
1. McArthur JC, Brew BJ. HIV-associated neurocognitive disorders: is there a hidden epidemic? AIDS. 2010;24:1367–1370.
2. McArthur JC, Steiner J, Sacktor N, et al.. Human immunodeficiency virus-associated neurocognitive disorders: mind the gap. Ann Neurol. 2010;67:699–714.
3. Sacktor N, McDermott MP, Marder K, et al.. HIV-associated cognitive impairment before and after the advent of combination therapy. J Neurovirol. 2002;8:136–142.
4. Tozzi V, Balestra P, Lorenzini P, et al.. Prevalence and risk factors for human immunodeficiency virus-associated neurocognitive impairment, 1996 to 2002: results from an urban observational cohort. J Neurovirol. 2005;11:265–273.
5. Ancuta P, Kamat A, Kunstman KJ, et al.. Microbial translocation is associated with increased monocyte activation and dementia in AIDS patients. PLoS One. 2008;3:e2516.
6. Ryan LA, Zheng J, Brester M, et al.. Plasma levels of soluble CD14 and tumor necrosis factor-alpha type II receptor correlate with cognitive dysfunction during human immunodeficiency virus type 1 infection. J Infect Dis. 2001;184:699–706.
7. Gartner S, Liu Y. Insights into the role of immune activation in HIV neuropathogenesis. J Neurovirol. 2002;8:69–75.
8. Gonzalez-Scarano F, Martin-Garcia J. The neuropathogenesis of AIDS. Nat Rev Immunol. 2005;5:69–81.
9. Gras G, Kaul M. Molecular mechanisms of neuroinvasion by monocytes-macrophages in HIV-1 infection. Retrovirology. 2010;7:30.
10. Kaul M, Zheng J, Okamoto S, et al.. HIV-1 infection and AIDS: consequences for the central nervous system. Cell Death Differ. 2005;12(suppl 1):878–892.
11. Ellis RJ, Hsia K, Spector SA, et al.. Cerebrospinal fluid human immunodeficiency virus type 1 RNA levels are elevated in neurocognitively impaired individuals with acquired immunodeficiency syndrome. HIV Neurobehavioral Research Center Group. Ann Neurol. 1997;42:679–688.
12. Enting RH, Foudraine NA, Lange JM, et al.. Cerebrospinal fluid beta2-microglobulin, monocyte chemotactic protein-1, and soluble tumour necrosis factor alpha receptors before and after treatment with lamivudine plus zidovudine or stavudine. J Neuroimmunol. 2000;102:216–221.
13. McArthur JC, McClernon DR, Cronin MF, et al.. Relationship between human immunodeficiency virus-associated dementia and viral load in cerebrospinal fluid and brain. Ann Neurol. 1997;42:689–698.
14. Brew BJ, Pemberton L, Cunningham P, et al.. Levels of human immunodeficiency virus type 1 RNA in cerebrospinal fluid correlate with AIDS dementia stage. J Infect Dis. 1997;175:963–966.
15. Cinque P, Vago L, Mengozzi M, et al.. Elevated cerebrospinal fluid levels of monocyte chemotactic protein-1 correlate with HIV-1 encephalitis and local viral replication. AIDS. 1998;12:1327–1332.
16. Conant K, Garzino-Demo A, Nath A, et al.. Induction of monocyte chemoattractant protein-1 in HIV-1 Tat-stimulated astrocytes and elevation in AIDS dementia. Proc Natl Acad Sci U S A. 1998;95:3117–3121.
17. Kelder W, McArthur JC, Nance-Sproson T, et al.. Beta-chemokines MCP-1 and RANTES are selectively increased in cerebrospinal fluid of patients with human immunodeficiency virus-associated dementia. Ann Neurol. 1998;44:831–835.
18. Sanders VJ, Pittman CA, White MG, et al.. Chemokines and receptors in HIV encephalitis. AIDS. 1998;12:1021–1026.
19. Sevigny JJ, Albert SM, McDermott MP, et al.. Evaluation of HIV RNA and markers of immune activation as predictors of HIV-associated dementia. Neurology. 2004;63:2084–2090.
20. Brew BJ, Bhalla RB, Paul M, et al.. Cerebrospinal fluid neopterin in human immunodeficiency virus type 1 infection. Ann Neurol. 1990;28:556–560.
21. Griffin DE, McArthur JC, Cornblath DR. Neopterin and interferon-gamma in serum and cerebrospinal fluid of patients with HIV-associated neurologic disease. Neurology. 1991;41:69–74.
22. Shaskan EG, Brew BJ, Rosenblum M, et al.. Increased neopterin levels in brains of patients with human immunodeficiency virus type 1 infection. J Neurochem. 1992;59:1541–1546.
23. Wiley CA, Achim CL, Schrier RD, et al.. Relationship of cerebrospinal fluid immune activation associated factors to HIV encephalitis. AIDS. 1992;6:1299–1307.
24. Achim CL, Heyes MP, Wiley CA. Quantitation of human immunodeficiency virus, immune activation factors, and quinolinic acid in AIDS brains. J Clin Invest. 1993;91:2769–2775.
25. Mastroianni CM, Paoletti F, Valenti C, et al.. Tumour necrosis factor (TNF-alpha) and neurological disorders in HIV infection. J Neurol Neurosurg Psychiatry. 1992;55:219–221.
26. Bhaskaran K, Mussini C, Antinori A, et al.. Changes in the incidence and predictors of human immunodeficiency virus-associated dementia in the era of highly active antiretroviral therapy. Ann Neurol. 2008;63:213–221.
27. Eden A, Price RW, Spudich S, et al.. Immune activation of the central nervous system is still present after >4 years of effective highly active antiretroviral therapy. J Infect Dis. 2007;196:1779–1783.
28. Cysique LA, Brew BJ, Halman M, et al.. Undetectable cerebrospinal fluid HIV RNA and beta-2 microglobulin do not indicate inactive AIDS dementia complex in highly active antiretroviral therapy-treated patients. J Acquir Immune Defic Syndr. 2005;39:426–429.
29. Hagberg L, Cinque P, Gisslen M, et al.. Cerebrospinal fluid neopterin: an informative biomarker of central nervous system immune activation in HIV-1 infection. AIDS Res Ther. 2010;7:15.
30. Yilmaz A, Price RW, Spudich S, et al.. Persistent intrathecal immune activation in HIV-1-infected individuals on antiretroviral therapy. J Acquir Immune Defic Syndr. 2008;47:168–173.
31. Lyons JL, Uno H, Ancuta P, et al.. Plasma sCD14 is a biomarker associated with impaired neurocognitive test performance in attention and learning domains in HIV infection. J Acquir Immune Defic Syndr. 2011;57:371–379.
32. Fischer-Smith T, Croul S, Sverstiuk AE, et al.. CNS invasion by CD14+/CD16+ peripheral blood-derived monocytes in HIV dementia: perivascular accumulation and reservoir of HIV infection. J Neurovirol. 2001;7:528–541.
33. Chang L, Ernst T, St Hillaire C, et al.. Antiretroviral treatment alters relationship between MCP-1 and neurometabolites in HIV patients. Antivir Ther. 2004;9:431–440.
34. Letendre SL, Zheng JC, Kaul M, et al.. Chemokines in cerebrospinal fluid correlate with cerebral metabolite patterns in HIV-infected individuals. J Neurovirol. 2011;17:63–69.
35. Antinori A, Arendt G, Becker JT, et al.. Updated research nosology for HIV-associated neurocognitive disorders. Neurology. 2007;69:1789–1799.
36. Woods SP, Rippeth JD, Frol AB, et al.. Interrater reliability of clinical ratings and neurocognitive diagnoses in HIV. J Clin Exp Neuropsychol. 2004;26:759–778.
37. Li C, Hung Wong W. Model-based analysis of oligonucleotide arrays: model validation, design issues and standard error application. Genome Biol. 2001;2:RESEARCH0032.
38. Li C, Wong WH. Model-based analysis of oligonucleotide arrays: expression index computation and outlier detection. Proc Natl Acad Sci U S A. 2001;98:31–36.
39. Domenici E, Wille DR, Tozzi F, et al.. Plasma protein biomarkers for depression and schizophrenia by multi analyte profiling of case-control collections. PLoS One. 2010;5:e9166.
40. Xia J, Psychogios N, Young N, et al.. MetaboAnalyst: a web server for metabolomic data analysis and interpretation. Nucleic Acids Res. 2009;37:W652–W660.
41. Bijlsma S, Bobeldijk I, Verheij ER, et al.. Large-scale human metabolomics studies: a strategy for data (pre-) processing and validation. Anal Chem. 2006;78:567–574.
42. Burdo TH, Katner SN, Taffe MA, et al.. Neuroimmunity, drugs of abuse, and neuroAIDS. J Neuroimmune Pharmacol. 2006;1:41–49.
43. Martin-Thormeyer EM, Paul RH. Drug abuse and hepatitis C infection as comorbid features of HIV associated neurocognitive disorder: neurocognitive and neuroimaging features. Neuropsychol Rev. 2009;19:215–231.
44. Nath A, Maragos WF, Avison MJ, et al.. Acceleration of HIV dementia with methamphetamine and cocaine. J Neurovirol. 2001;7:66–71.
45. Letendre S, Marquie-Beck J, Capparelli E, et al.. Validation of the CNS penetration-effectiveness rank for quantifying antiretroviral penetration into the central nervous system. Arch Neurol. 2008;65:65–70.
46. Letendre SL, Ellis RJ, Ances BM, et al.. Neurologic complications of HIV disease and their treatment. Top HIV Med. 2010;18:45–55.
47. Gisolf EH, van Praag RM, Jurriaans S, et al.. Increasing cerebrospinal fluid chemokine concentrations despite undetectable cerebrospinal fluid HIV RNA in HIV-1-infected patients receiving antiretroviral therapy. J Acquir Immune Defic Syndr. 2000;25:426–433.
48. Cinque P, Bestetti A, Marenzi R, et al.. Cerebrospinal fluid interferon-gamma-inducible protein 10 (IP-10, CXCL10) in HIV-1 infection. J Neuroimmunol. 2005;168:154–163.
49. Brew BJ, Letendre SL. Biomarkers of HIV related central nervous system disease. Int Rev Psychiatry. 2008;20:73–88.
50. Sinclair E, Ronquillo R, Lollo N, et al.. Antiretroviral treatment effect on immune activation reduces cerebrospinal fluid HIV-1 infection. J Acquir Immune Defic Syndr. 2008;47:544–552.
51. Schmidtmayerova H, Nottet HS, Nuovo G, et al.. Human immunodeficiency virus type 1 infection alters chemokine beta peptide expression in human monocytes: implications for recruitment of leukocytes into brain and lymph nodes. Proc Natl Acad Sci U S A. 1996;93:700–704.
52. Letendre SL, Lanier ER, McCutchan JA. Cerebrospinal fluid beta chemokine concentrations in neurocognitively impaired individuals infected with human immunodeficiency virus type 1. J Infect Dis. 1999;180:310–319.
53. Harezlak J, Buchthal S, Taylor M, et al.. Persistence of HIV-associated cognitive impairment, inflammation, and neuronal injury in era of highly active antiretroviral treatment. AIDS. 2011;25:625–633.
54. Farag NH, Rashed HA, Hassan M, et al.. Hepatitis C infection, cognition, and inflammation in an Egyptian sample. J Med Virol. 2011;83:261–266.