Sirtuin-2, NAD-Dependent Deacetylase, Is a New Potential Therapeutic Target for HIV-1 Infection and HIV-Related Neurological Dysfunction

To identify peripheral blood biomarkers associated with the outcome of HIV infection and the development of HIV-related neurological disease, we employed an antibody array previously used successfully in the study of Alzheimer’s disease (9). We measured levels of 185 plasma factors in samples from untreated, chronically HIV-infected individuals classified as “HIV-high” (n = 30) with a plasma viral load (pVL) of >10,000 HIV RNA copies/mL (range, 15,000 to 1,200,000; median, 147,500) and “HIV-low” with a plasma viral load of <10,000 HIV RNA copies/mL (range, <50 to 10,000; median, 975) (n = 20) (see Table S1 in the supplemental material). Data analyses identified 20 factors as most differentially detected between the two groups (see Table S2 in the supplemental material), of which SIRT2 emerged as the most significant candidate, considering fold change differences of plasma levels and the P value of pairwise comparisons (Fig. 1A and B). SIRT2 was elevated in the HIV-high group (Fig. 1C) and the relative plasma levels of SIRT2 correlated positively with pVL and with HIV proviral levels across all HIV-infected individuals studied (Table 1). Correspondingly, the SIRT2 gene was significantly more highly expressed in peripheral blood mononuclear cells (PBMCs) from the HIV-high (Fig. 1D) than from the HIV-low group. This difference was even more pronounced when comparing the HIV-high individuals to an additional cohort of HIV elite controllers (EC) (HIV-infected individuals with undetectable viral load in the absence of antiretroviral treatment) (Fig. 1D). The same associations with viral parameters observed for SIRT2 plasma levels were also detected in PBMCs, as SIRT2 gene expression correlated with pVL and HIV proviral levels across all of the groups (Fig. 1E and F).

In order to study the relationship between SIRT2 plasma levels and the other molecules measured in the array, we performed correlation analysis across all measured markers. Figure 2A shows the molecules in plasma that are significantly correlated with SIRT2 levels (P < 0.05, Spearman's rank test). The functional relationship between these statistically related factors was further explored using the STRING application (Fig. 2B). Several functions reported in open-access databases (Gene Ontology, UniProt, and Reactome) were identified, including Regulation of Cell Death (GO, 0010941; false-discovery rate (FDR) = 0.00030), Innate Immunity (KW0399; FDR = 3.57 × 10⁷), and Complement Cascade (HSA-166658; FDR = 3.44 × 10⁶). Interestingly, SIRT2 was functionally connected with brain-derived neurotrophic factor (BDNF), while BDNF was at the same time connected with alpha-synuclein (SNCA) and microtubule-associated protein tau (MAPT) (Fig. 2B), which are both substrates of SIRT2 in brain tissue (11). In addition, SNCA and MAPT are, together with BDNF, well-established biomarkers of neurological damage (12) and correlated positively with SIRT2 levels in plasma (Fig. 2C to E). Although CNS injury was not determined in the two groups of HIV-high/low, trends toward increased levels of specific markers of neurological damage (SNCA, MAPT, and BDNF) in individuals with poor virus control were observed (Fig. 3A to C), with the comparison of SNCA reaching statistical significance. These data identify SIRT2 as a soluble factor in plasma that discriminates between controlled and uncontrolled HIV infection and also showed a relationship with the levels of several biomarkers of neurological damage (13).

To further explore the relationship between SIRT2 levels in plasma and CNS damage in HIV-infected individuals, we determined the levels of SIRT2 in matched cerebrospinal fluid (CSF) and plasma samples from treatment-naive HIV-infected as well as from seronegative individuals. While this small set of samples did not show a significant difference in SIRT2 levels in either group for CSF or plasma (Fig. 3D), the CSF SIRT2 levels correlated strongly with MAPT (P < 0.001; rho = 0.7569) (Fig. 3E) and neurofilament light protein (NFL) (P < 0.001; rho = 0.7131) (Fig. 3F), a well-known biomarker of neuronal damage in HIV infection (14).

These data were validated by analyzing the open access GEO data set GSE28160, which contains data on postmortem brain tissue from patients that died with diagnosed HAND pathology with a severe symptomatology (HIV-associated dementia [HAD]) (12). We evaluated SIRT2 expression levels among HIV-infected individuals with cART-treated HAD (n = 6) or not receiving cART (cART-untreated HAD, n = 8) and seronegative individuals (n = 6) (Table 2). As expected from our experimental data, brain SIRT2 levels tended to be higher in cART-untreated individuals with HAD compared to those of seronegative individuals (Fig. 3G), suggesting increased levels of SIRT2 in uncontrolled infection. Interestingly, SIRT2 expression levels in brain tissue from HIV-infected individuals (cART treated and cART untreated with HAD) correlated with MAPT expression levels (Fig. 3H), reinforcing the association in our data between SIRT2 and markers of CNS injury.

Overall, these results show that higher levels of SIRT2 in the peripheral blood are detected in uncontrolled HIV infection and that these levels are also correlated with viral parameters as well as with biomarkers of neurological dysfunction.

Several studies have suggested that HIV-related neuropathogenesis may be triggered by initial viral entry into the CNS, followed by first pathological processes and the establishment of the viral reservoir in the brain (15). Rapid initiation of cART is currently recommended for all newly diagnosed HIV infections, with expected beneficial effects on brain reservoir and reduced longitudinal neurological dysfunction (16–18). To examine whether the time to cART initiation impacts neurologic status and if SIRT2 plasma levels can be informative of this process, a cohort of HIV-infected individuals who underwent a prospective neurologic evaluation and started cART either (i) <3 months after (“early-cART”) or (ii) >6 months after (“later-cART”) the estimated date of acquisition (Fig. 4A) was studied (8). Both groups were longitudinally evaluated, including baseline (day of cART initiation), 4 weeks, and 1 year, on cART treatment time points (Fig. 4A and Table 3). Plasma viral loads and the estimated days since infection to cART initiation were different between groups at baseline (Table 3). While neuropsychological battery tests (NPZ12 score) and brain global gray matter volume measurements showed no differences between the two groups and over time (Fig. 4B and C and Table 4), an involution of the medial orbitofrontal volumetry was detected in the later-cART after 1 year on treatment (Fig. 4D and Table 4). In parallel, four markers of neurological damage (NFL, glial fibrillary acidic protein [GFAP], ubiquitin C-terminal hydrolase L1 [UCHL-1], and MAPT) were measured in plasma samples (Fig. 4E to H). Only MAPT was increased in the plasma after 1-year on treatment in the later-treated individuals (Fig. 4H and I). Nevertheless, the medial orbitofrontal volumetry and MAPT plasma levels were not significantly associated (Fig. 4J).

In addition, SIRT2 plasma levels were similarly elevated after 1 year in the later-cART group compared with those of individuals that initiated cART within less than 3 months from infection (Fig. 4K and L and Table 4). Moreover, SIRT2 plasma levels were negatively correlated with medial orbitofrontal volumetry (Spearman rho = −0.5357; P value = 0.0422) (Fig. 4M), indicating that early treatment is crucial for limiting the progressive neurological dysfunction in HIV infection and suggesting that SIRT2 could serve as plasma biomarker for these degenerative processes.

In order to investigate more directly a potential effect of SIRT2 on HIV replication and virus reactivation, we conducted in vitro experiments using specific SIRT2 inhibitors to modulate HIV growth in different target cell types. First, the specific SIRT2 inhibitor AK-1, which is actively being evaluated in Alzheimer’s disease models, was used to inhibit HIV replication in PHA-activated T cells and monocyte-derived macrophages (MDMs) infected in vitro with different HIV strains. In the presence of SIRT2 inhibitor, significantly reduced p24 levels were detected in HIV_NL4-3-infected PHA blasts (at day 3 and after 1 week postinfection) (Fig. 5A and B) and in HIV_BaL-infected MDMs (day 4 postinfection) (Fig. 5C). The reduction in p24 levels in both cell types was not due to cytotoxic effects, as cell viability was not affected in AK-1-treated compared to uninfected conditions and when comparing PHA-activated T cells to MDMs (Fig. 5D to F). The observed differences of AZT in different cell types are in line with previously published data (19).

As SIRT2 is predominantly expressed in the brain and given its association with markers of neurological damage in the participants of our study (Fig. 2 and 3), we tested its effect on viral replication in primary microglial cells, which are known target cells for HIV and potential sites for brain reservoir (20). Similar to the effects observed in peripheral blood cells, the results from microglial cells showed a significant reduction of HIV replication upon inhibition of SIRT2 (Fig. 5G and H).

Since SIRT2 plasma and expression levels were also correlated with HIV proviral levels (Table 1 and Fig. 1F), we tested the effects of SIRT2 inhibition on HIV reactivation. Inhibition of SIRT2 with AK-1 significantly reduced reactivation of HIV in the phorbol myristate acetate (PMA)-activated J-LAT A2 cell line (Fig. 6A to C) without affecting cell viability (Fig. 6D). Together, these data suggest that SIRT2 targeting reduces HIV replication and virus reactivation from latency, indicating that SIRT2 is required for effective viral infection in peripheral blood cells and likely in the CNS, given its sites of expression and associations with markers of neurological damage.

Chronic untreated HIV-infected individuals (n = 60) enrolled at the IMPACTA clinics (Peru), Hospital Germans Trias i Pujol (Spain), Sahlgrenska University Hospital in Gothenburg (Sweden), and University of California San Francisco (USA) were classified according to their degree of control of viral replication (see Table S1 in the supplemental material). HIV-infected participants from the ARBRE study (ClinicalTrials registration no. NCT03835546) recruited at Fundació Lluita per la Sida, Hospital Universitari Germans Trias i Pujol, (Spain) who underwent longitudinal neuropsychological and neuroimaging assessments (Tables 2 and 3), were also included. These participants were divided into 2 arms according to the time from estimated date of HIV acquisition to initiation of cART. The “early-cART” arm (n = 9) started cART within less than 90 days and the “later-cART” arm (n = 10) longer than 6 months since estimated time of HIV acquisition (Tables 2 and 3) (8). Available plasma, CSF, and dry-pellet PBMC were stored until use. Blood samples from non-HIV-infected donors from the Banc de Sang i Teixits in Barcelona for in vitro studies were used. The study was approved by the Comité Ètic d’Investigació Clínica of Hospital Germans Trias i Pujol (CEIC EO-12-042 and PI-18-183), and all participants provided their written informed consent. All of the research involving human research participants was performed in accordance with the Declaration of Helsinki.

A custom-designed chip previously used in a study of Alzheimer’s disease (9) was used to detect and quantify 185 proteins in plasma samples. After normalization and clustering analyses, differences between groups of patients were analyzed using the t test, and molecules with a significance level (P value < 0.05 and FDR < 0.1) were included in further analyses. The functional analysis was performed using the application STRING: functional protein association networks (https://string-db.org).

CSF samples were used for proximity extension analysis with Olink (https://www.olink.com/data-you-can-trust/technology/) for evaluation of neurology, neuroexploratory, and inflammation panels. Relative expression levels are expressed as normalized protein expression (NPX).

Plasma samples were used for ultrasensitive single-molecule array (SIMOA)-based detection of tau, NFL, GFAP, and UCHL1 on an SR-X instrument (Quanterix). We used the commercially available Neuro 4-plex B kit for absolute quantification of tau, NFL, GFAP, and UCHL1. In brief, samples were thawed and centrifuged and then plated and diluted 4× with the sample dilution buffer to start the establish protocol for SIMOA. Antibody-attached beads designed to bind to specific targets were incubated with the samples, before secondary fluorescent antibodies were added. The plates were loaded into SIMOA array discs in which each well holds one bead and the enzymatic signal can be read.

Human SIRT2 enzyme-linked immunosorbent assay (ELISA) kit (Aviva Systems Biology) was used to measure the SIRT2 levels in plasma, and the absolute levels were quantified by applying a 4-parameter logistic curve analysis.

Cognitive evaluation covered 6 cognitive domains to provide a global composite score (2 measures per domain, global NPZ-12). This included a digit test of the Wechsler adult intelligence scale (WAIS-IV); the trail making test (TMT-A) and the symbol digit modalities test (SDMT); grooved pegboard test (GPT); California verbal learning test (CVLT-II); the initial letter “p” and the animals test; and the trail making test (TMT-B) and the Tower of London test (TOL). The vocabulary test of the WAIS-IV was used to estimate premorbid intelligence.

Neurological image data were obtained by magnetic resonance imaging (MRI) (3 Tesla Magnetic Resonance Imaging Siemens Verio scanner). A high resolution T1-weighted three-dimensional (3-D) structural image using a 3-T scanner (Siemens Verio; Siemens Healthcare Sector, Germany) with a 32-phased-array head coil (192 slices in the axial plane; repetition time = 1,900 ms; echo time = 2.72 ms; flip angle = 9°; field of view = 260 × 260 mm; matrix size = 256 × 256 pixels; in-plane resolution = 0.96 × 0.96 mm²; slice thickness = 0.9 mm) was measured. After preprocessing and inspection for the presence of artifacts, all imaging time points were processed following a standard VBM-DARTEL pipeline to obtain MNI normalized and modulated images. Images were spatially smoothed with an 8-mm full width at half maximum (FWHM) isotropic Gaussian kernel. Differences at the whole-brain and voxel-wise level with a P < 0.05 significance threshold were explored. These analyses were controlled for age and total gray matter volume. Voxel values from significant regions were extracted to perform further statistical comparisons.

The data set from the NCBI Gene Expression Omnibus (GEO) (http://www.ncbi.nlm.nih.gov/geo/) databank (accession number GSE28160) was used for evaluation of SIRT2 expression levels in postmortem brain tissues (47).

RNA samples from PBMC dry pellets were retrotranscribed and TaqMan gene expression assay (Applied Biosystems) was used for detection of SIRT2 (Hs01560289_m1) and TBP (Hs99999910_m1). Gene amplification was performed in an Applied Biosystems 7500 Fast real-time PCR system thermocycler, and the relative expression was calculated as 2^−ΔCT (where C_T is the median threshold cycle from 3 replicates).

HIV proviral DNA was quantified in PBMCs by droplet digital PCR (ddPCR) in duplicates as previously described (48). Briefly, two different primer/probe sets annealing to the 5′ long terminal repeat and Gag regions, respectively, were used to circumvent sequence mismatch in the patient proviruses, and the RPP30 housekeeping gene was quantified in parallel to normalize sample input. Raw ddPCR data were analyzed using the QX100 droplet reader and QuantaSoft v.1.6 software (Bio-Rad).

Isolated PBMCs from non-HIV-infected donors were stimulated with PHA (5 μg/mL) and interleukin-2 (IL-2) (10 U/mL). After 3 days, PHA blasts were infected with the HIV_NL4-3 strain (multiplicity of Infection [MOI], 0.01). For monocyte-derived macrophages (MDMs), PBMCs were depleted using the EasySep human monocyte enrichment kit (Stem Cell). Monocytes were then incubated with macrophage colony-stimulating factor (M-CSF) at 1 μg/mL for 4 days before infection with the HIV_BaL strain (MOI, 0.01). HIV replication in PHA blasts and MDMs was evaluated under the following conditions: zidovudine (AZT) (200 μg/mL), dimethyl sulfoxide (DMSO), 10 μM AK-1 (3-[(hexahydro-1H-azepin-1-yl) sulfonyl]-N-(3-nitrophenyl)-benzamide). After 3 days or 1 week in PHA-blast and 4 days in MDM, p24 in supernatant was quantified by ELISA (INNOTEST HIV p24 antigen MAb).

Primary microglial cells (iCell Microglia; FUJIFILM Cellular Dynamics) were thawed and cultured for 3 days in a 96-well plate at 20,000 glial cells/well following the manufacturer’s recommendations. After 3 days, microglial cells were infected with the HIV_NLAD8 strain (MOI, 0.01), and after 16 h, cells were washed. Infected microglial cells were then incubated in the absence or presence of 10 μM AK-1 (Sigma-Aldrich) or AZT (200 μg/mL). After 3 days, p24 in the culture supernatant was quantified by ELISA (INNOTEST HIV p24 antigen MAb).

The J-LAT A2 cells, which comprises transfected Jurkat cells with a green fluorescent protein (GFP)-encoding HIV minigenome (49), were stimulated with PMA (1 ng/mL) and cultured in the presence or absence of AK-1 (10 μM). After 24 h, GFP expression was evaluated on a fluorescence-activated cell sorter (FACS) Canto flow cytometer (Becton, Dickinson), and the data were analyzed using FlowJo version 10 software.

Mann-Whitney U test, Wilcoxon matched pairs tests and paired and unpaired t tests, and analysis of variance (ANOVA) test for multiple comparisons corrected by the original FDR method of Benjamini and Hochberg test were applied using GraphPad Prism, version 8. The Spearman's rank test was applied for correlation analyses. For all analyses, P values of <0.05 were considered statistically significant.