Blood Biomarkers of Neuronal/Axonal and Glial Injury in Human Immunodeficiency Virus-Associated Neurocognitive Disorders

As of 2019, approximately 38 million people were living with the human immunodeficiency virus (HIV)/acquired immunodeficiency syndrome worldwide [1]. The availability of combination antiretroviral therapy (cART) has successfully improved the prognosis of people living with HIV (PLWH) and transformed HIV infection into a chronically manageable disease [2, 3]. Among the significant advances in HIV prognosis since the introduction of cART, there has been a dramatic decrease in the incidence of HIV encephalitis and HIV-associated dementia (HAD), which is the most severe form of HIV-associated neurocognitive disorders (HAND) [2]. Nevertheless, less severe forms of cognitive impairment remain prevalent and are a significant cause of morbidity in PLWH. It is estimated that 15–55% of PLWH experience some form of HAND [2, 4]. Of note, aging is one of the main risk factors for HAND, with PLWH older than 50 having a 2-fold higher risk than younger PLWH [5]. This demographic shift is relevant as over half of people with HIV are aged 50 years and older [6].

There are no effective treatments for HAND, and the mechanisms underlying this complex spectrum of neurocognitive symptoms are yet to be determined. The neuropathology associated with HIV has shifted in the cART era. In the pre-cART era, the neuropathological changes observed in PLWH included encephalitis and neuronal loss, which were related to severe forms of dementia. However, neuronal loss is not commonly observed in cART-treated patients with HAND, suggesting that subtle neuronal changes drive the pathology. These include simplification of neuronal networks and loss of dendritic spines in brain regions key to the cognitive processes affected in HAND [4]. The HIV-related synaptodendritic injury results from both direct (viral proteins) and indirect (e.g., inflammatory and vascular factors) mechanisms. Accordingly, studies investigating surrogate biomarkers for cognitive impairment in HIV have primarily focused on inflammatory/immune, oxidative stress, vascular, and glial markers [2, 7]. A few studies have evaluated cerebrospinal fluid (CSF) levels of neurodegeneration (i.e., neuronal and axonal injury) markers, such as neurofilament light chain (NfL) and tau proteins in PLWH [8‒12]. The levels of neurodegeneration markers seem to increase with HIV progression [11‒14] and decrease after cART initiation [8, 15, 16].

While CSF-derived markers are an excellent source of information about central nervous system (CNS) injury in HIV infection, CSF sampling requires an invasive procedure and can cause significant distress in some patients. Conversely, blood samples are much easier to obtain and widely available. A growing literature on neurodegenerative diseases has shown that blood-based molecules, especially NfL, can be reliable biomarkers of the underlying processes, with great potential to inform research and clinical practice [17, 18]. However, only a few studies have investigated blood-based neurodegeneration markers in PLWH, demonstrating a high correlation between blood and CSF levels of NfL [19] and a relationship between increased blood NfL levels and worse scores on neuropsychological assessments [20]. Of note, a recent study in adolescents with perinatally acquired HIV found no differences in plasma NfL levels compared to controls, with both groups showing similar changes in NfL concentrations in a 4.6-year follow-up assessment. In addition, van der Post et al. [21] found no significant associations between plasma NfL levels and neuroimaging and neurocognitive outcomes. The discrepancy with previous reports may be explained by the differences in the studied populations (i.e., adult vs. pediatric) and due to a potential difference in the pathophysiological mechanisms associated with neurocognitive impairment in perinatally acquired HIV in comparison with adult infection [21]. Whether changes in blood-based neurodegeneration markers are associated with HAND, specifically in older adults, is yet to be determined. In the current study, we sought to investigate whether blood-based markers of neuronal and glial injury are altered in older adults with HAND compared to PLWH without HAND and HIV-seronegative persons without cognitive impairment. We assessed plasma levels of NfL as a marker of neurodegeneration [18], phosphorylated Tau 181 (pTau181) as a marker of tau pathology [22], and glial fibrillary acidic protein (GFAP) as a marker of glial degeneration [23]. We hypothesize that individuals with HAND will present increased blood levels of NfL, pTau181, and GFAP compared to PLWH without HAND and seronegative individuals.

This study included 132 participants aged 50 and older recruited from greater San Diego County, as previously described [24]. The participants were divided into three groups: PLWH with HAND (HAND+, n = 29), PLWH without HAND (HAND−, n = 73), and HIV-seronegative persons without cognitive impairment (HIV−, n = 30). HIV serostatus was confirmed with MedMira Reveal Rapid HIV-1 Antibody Tests (MedMira Laboratories, Halifax, NS, Canada). The Reveal Rapid HIV-1 Antibody Test is a single-use, qualitative immunoassay to detect antibodies to HIV-1 in human serum or plasma specimens in a clinical laboratory setting. The overall sensitivity is 99.8% for both plasma and serum samples, and the specificity is 98.6% and 99.1% for plasma and serum samples, respectively.

Participants were excluded from the parent study if they presented a medical history of severe psychiatric disorders, neurological disease not due to HIV infection, or major head injury; or if they reported color-blindness, current substance use disorder, or had a positive breathalyzer for alcohol or urine screens for illicit substances. In addition, participants were excluded from the present study if they did not complete the neurocognitive assessments or did not have stored samples available for the biomarker assays. The study exclusions were based on a two-tiered process. The study recruiters excluded participants prior to enrollment in consultation with the principal investigator. The study research assistant also performed the urine drug screening and semi-structured interview to confirm the study exclusion criteria.

The diagnosis of HAND followed the Frascati criteria [25] and was based on age-adjusted scores from either the Cogstate [26] or the Cognition Battery from the NIH Toolbox [27]. The different batteries resulted from an administrative shift during the parent project, described in Matchanova et al. [28]. The neuropsychological test data were summarized with the Global Deficit Score (GDS) [29], which converts individual T-scores to deficit scores that range from 0 (T-scores >39) to 5 (T-scores <20). These individual deficit scores are then averaged to derive the GDS for which higher scores reflect more significant impairment. Consistent with prior research in neuroHIV, GDS values ≥0.5 were used to define HAND [29]. Within the HIV group with HAND, 11 individuals met the criteria for mild neurocognitive disorder, and 18 met the criteria for asymptomatic neurocognitive impairment based on the entire assessment, which included the Heaton et al. [30] revision of the Lawton and Brody Activities of Daily Living Scale and the Karnofsky Performance Status Scale [31].

All participants provided written informed consent before completing the study procedures, and the University of California, San Diego, IRB approved this study. Diagnoses of current and lifetime substance use, major depressive, and anxiety disorders were obtained from the Composite International Diagnostic Interview (CIDI) [32], which is a structured interview that derives diagnoses consistent with the Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV) [33].

Sociodemographic factors including age, education, sex, and race/ethnicity were obtained by self-report during a semi-structured interview. Diagnoses of lifetime major depression, generalized anxiety, and substance use disorder were obtained via the CIDI (as noted above). Common medical comorbidities (e.g., cardiovascular disease, arthritis, cancer, liver disease, kidney disease) were obtained via self-report during a semi-structured interview with a research nurse.

Peripheral blood was collected by venipuncture on the same day of the clinical assessment. Plasma levels of NfL, pTau181 (MyBioSource, San Diego, CA, USA), and GFAP (Millipore, Temecula, CA, USA) were measured by enzyme-linked immunosorbent assay (ELISA) according to the procedures supplied by the manufacturers.

NfL (catalog number: MBS705750) and GFAP (catalog number: NS830) assays employed the quantitative sandwich ELISA technique, while the pTau181 (catalog number: MBS744973) assay was based on the competitive ELISA detection method. The manufacturer reported the assay’s sensitivity as <7.8 pg/mL for NfL, 1.5 ng/mL for GFAP, and 1.0 pg/mL for pTau181. All samples were assayed in duplicate, and the technician was blind to the group allocation of the samples. The inter- and intra-assay coefficients of variation were below 10%.

The NfL and GFAP variables were normally distributed per Shapiro-Wilk (p > 0.05), but the pTau181 variable was non-normal and had a mild positive skew (p < 0.0001) (online suppl. Fig. 1; see www.karger.com/doi/10.1159/000527659 for all online suppl. material). As detailed below, the statistical analyses were tested using independent sample t tests, χ², or Pearson correlations (or their non-parametric counterpart for non-normal variables). We used a confound approach to identifying covariates [34]. Specifically, any descriptive variable from Table 1 that differed significantly between the 3 neuroHIV study groups and was significantly associated with the biomarker was included as a covariate. Our primary study hypotheses were tested with linear multiple regression, one-way ANOVA, or a Wilcoxon ranked sum test. Where relevant, effect sizes were estimated with R² and Cohen’s d. The effect size was considered small if Cohen’s d = 0.2, medium if d = 0.5, and large if d = 0.8. The power to detect a medium effect size given our sample of 132 and up to 5 covariates was strong (power, 1-beta = 0.94) using a critical alpha of 0.05. Analyses were conducted in JMP (SAS, version 16.0.0).

The study groups included 102 PLWH (29 HAND+, 73 HAND−) and 30 HIV– individuals whose sociodemographic and clinical characteristics are shown in Table 1. A series of ANOVAs (or their non-parametric counterpart, Wilcoxon ranked sum tests) showed that the study groups did not differ on any of the continuous variables listed in Table 1 (p < 0.05), with the exception of age and education (p < 0.05). Likewise, a series of χ² analyses showed that the study groups did not differ on any of the categorical variables in Table 1 (p < 0.05), with the exception of sex and the frequency of lifetime diagnoses of major depression (p < 0.05). These five variables were therefore evaluated as potential covariates in each of the three biomarker analyses described below. The blood levels of NfL, GFAP, and pTau181 were detectable in all participants. Outliers were Windsorized, and the data were transformed to log 10.

Among the five potential covariates, only sex met our a priori criteria for inclusion in the model. Specifically, a t test showed that women (mean = 3.30, SD = 0.05) had higher NfL levels than men (mean = 3.10, SD = 0.02) in the full sample (p = 0.0001). All other analyses of covariates were null (p < 0.05). In a regression model with sex and study group predicting NfL, the overall model was significant (adjR² = 0.14, F(3,128) = 7.90, p < 0.0001), and both sex (p < 0.001) and neuroHIV group (p = 0.020) were independent contributors. At the univariable level, we see that the HAND + group had numerically higher NfL values at a medium effect size as compared to the HAND– group (p = 0.039, Cohen’s d = 0.45). A comparable medium effect size difference was also evident between the HAND+ and HIV− groups (Cohen’s d = 0.49), but the difference fell at the level of a trend (p = 0.059) (Fig. 1).

No variable met criteria for inclusion as a covariate in the analyses related to GFAP. As such, we conducted a one-way ANOVA, which showed no significant association between neuroHIV group and GFAP (F(2,117) = 0.88, p = 0.416) (Fig. 1).

No variable met criteria for inclusion as a covariate in the analyses related to pTau181. As such, we conducted a Wilcoxon ranked sum test due to the non-normal distribution of the dependent variable. Results revealed no significant association between neuroHIV group and pTau181 (χ² (2, 129) = 1.97, p = 0.373; Fig. 1).

Herein, we determined blood levels of neuronal and glial markers in relation to cognitive status in PLWH. We hypothesized that HAND is associated with neuronal and glial degeneration. Therefore, individuals with HAND would present increased blood levels of NfL, pTau181, and GFAP compared to PLWH without HAND and seronegative individuals. We found that plasma levels of NfL (but not pTau181 and GFAP) were significantly associated with HAND. Noteworthy, higher levels of NfL were associated with HAND diagnosis even after adjusting for sex. Our data show that neuronal degeneration (as evidenced by increased levels of NfL), but not tau pathology or glial degeneration, is related to cognitive status in PLWH.

A few studies have evaluated CSF levels of neuroaxonal injury markers in PLWH [10‒12]. These markers seem to correlate with HIV progression [11, 13, 14], being elevated in PLWH with chronic HIV infection [12], and significantly associated with neuroimaging markers of neuronal damage [10, 12]. NfL is a sensitive marker of neuronal injury in HIV, and it is markedly increased in patients with HAD [8, 11, 14, 35]. Among patients with HAD, NfL CSF levels correlated with the disease stage and decreased after initiating effective antiretroviral treatment [8]. Of note, elevated levels of NfL in CSF were predictive of HAD in a 2-year longitudinal study [9]. Conversely, cART resulted in decreased CSF levels of NfL [15, 16] in association with clinical improvement of patients with HAD [15].

Our results corroborate the data from CSF samples, showing that higher levels of NfL in plasma are associated with poorer cognitive status in PLWH. A few studies have investigated blood NfL levels in PLWH. Two studies found moderate-to-high correlations between blood and CSF NfL levels in PLWH [19, 36]. Higher plasma NfL levels were associated with worse scores on neuropsychological assessments [20] and with a diagnosis of dementia [19] in PLWH. In addition, a study using plasma neuronal extracellular vesicle proteins reported that NfL was among the three best variables that combined were good predictors of cognitive impairment in PLWH (including HAND) [37]. However, these findings are contradictory as studies found no association between plasma levels of NfL and cognitive performance on neuropsychological assessment [38] or diagnosis of cognitive impairment [36]. The most significant association between NfL and cognitive performance occurred among PLWH with dementia or untreated PLWH [19, 20]. In contrast, the lack of association was described in treated cognitively asymptomatic PLWH [38]. The inconsistent relationship between cognition and plasma NfL levels may be driven by cART, as plasma levels of NfL decrease after cART initiation [20].

Our findings suggest that pTau181 and GFAP were not associated with cognitive status in PLWH. Former studies also failed to show any association between pTau181 and GFAP and cognitive performance in PLWH [37, 39, 40]. PTau181 is a marker of tau pathology known to be elevated in patients with Alzheimer’s disease. Plasma levels of pTau181 correlate with CSF levels and are associated with pathological markers and clinical diagnosis of Alzheimer’s disease [41, 42]. GFAP is a protein found in astrocytes, hypothesized to support cell migration and protect brain tissue from stress and changes in the context of disease [43]. Evidence indicates that levels of GFAP can track subtle CNS injury in many neurological and systemic disorders. Blood GFAP is used in the clinical setting to evaluate mild traumatic brain injury, and GFAP levels correlate with clinical severity and extent of intracranial pathology. In addition, GFAP was described as a predictor of cognitive decline and conversion to dementia in older individuals [44]. The lack of association between pTau181 and GFAP levels with HAND indicates that the cognitive impairment in HIV is not driven by tau pathology or glial degeneration, as in classical neurodegenerative diseases. Instead, our data show that the pathophysiological process underlying cognitive impairment in PLWH may be caused by neuronal/axonal degeneration. In addition, mechanisms outside the CNS, such as peripheral inflammation, may play an essential role in cognition in PLWH. A comprehensive, recently published review of soluble biomarkers of cognition in PLWH brought NfL as the only CNS marker detected in the blood associated with cognition in PLWH. In addition to NfL, markers of HIV persistence (HIV DNA), co-infection (Toxoplasma IgG status), inflammation (chemokines, cytokines, c-reactive protein levels), and immune activation were significantly associated with cognitive status in PLWH [45].

The current findings should be interpreted considering the limitations of our study. First, the relatively small sample size does not allow us to exclude the participants with detectable virus and control for inflammatory parameters or match the groups on age and sex, and virus detectability in plasma. In addition, only a limited set of blood-based biomarkers was measured. Neuroimaging was unavailable for this cohort, hindering the investigation of potential associations between biomarkers and structural changes. The cross-sectional design does not allow a better understanding of the dynamic nature of blood-based biomarkers in PLWH over time. Finally, we did not ascertain the effects of potential confounders such as comorbidities and inflammation, and these variables may influence NfL levels. The comprehensive clinical assessment of both HIV– controls and PLWH, including a careful neuropsychological battery to define HAND, can be regarded as a strength of our study.

Our results corroborate the view that blood NfL seems to be a promising biomarker of cognitive impairment in PLWH. Blood NfL may be used in combination with the neuropsychological assessment in the screening of cognitive problems, to monitor disease progression, or serve as a prognostic marker of cognitive impairment in PLWH. Future studies should focus on longitudinal analysis of blood NfL in PLWH, which in combination with neuroimaging (e.g., atrophy in MRI studies) and neuropathological (e.g., histopathological analysis of neurodegeneration) studies would provide a better understanding of how blood NfL can predict brain damage and cognitive impairment in PLWH.

The Neuropsychiatry Program is supported by the UT Health Department of Psychiatry and Behavioral Sciences. The views expressed in this article are those of the authors and do not reflect the official policy or position of the Department of the Navy, Department of Defense, nor the US government. The authors are grateful for the considerable efforts of Dr. Erin Morgan and Marizela Verduzco for coordination of the parent project, Dr. J. Hampton Atkinson and Jennifer Marquie Beck for participant recruitment, grant co-investigators Drs. Mark Bondi and Elizabeth Twamley, Drs. Scott Letendre, and Sara Gianella Weibel for overseeing the neuromedical aspects of the parent project, and Donald R. Franklin, Stephanie Corkran, Jessica Beltran, Anastasia Matchanova, and Javier Villalobos for data processing.

All participants provided written informed consent, and the University of California, San Diego, IRB approved this study (# 130257).

The authors have no conflicts of interest to declare.

This research was supported by National Institutes of Health grants R01-MH073419 and P30-MH062512.

Natalia P. Rocha worked on the design and conceptualization of the study, executed the biochemical analyses, and wrote the first draft of the manuscript. Antonio L. Teixeira worked on the design and conceptualization of the study and wrote the first draft of the manuscript. Gabriela D. Colpo worked on the design and conceptualization of the study, executed the biochemical analyses, and reviewed the manuscript. Michelle A. Babicz and Jennifer L. Thompson assisted with data coding and quality assurance and revised the manuscript. Steven Paul Woods worked on the design and conceptualization of the study, oversaw the recruitment and evaluation of the study participants, designed and executed the data analysis and interpretation, and revised the manuscript.

The data supporting this study’s findings will be available upon reasonable request and its online supplementary material files.