Chronic brain damage in HIV-infected individuals under antiretroviral therapy is associated with viral reservoirs, sulfatide release, and compromised cell-to-cell communication

Abstract

HIV infection has become a chronic and manageable disease due to the effective use of antiretroviral therapies (ART); however, several chronic aging-related comorbidities, including cognitive impairment, remain a major public health issue. However, these mechanisms are unknown. Here, we identified that glial and myeloid viral reservoirs are associated with local myelin damage and the release of several myelin components, including the lipid sulfatide. Soluble sulfatide compromised gap junctional communication and calcium wave coordination, essential for proper cognition. We propose that soluble sulfatide could be a potential biomarker and contributor to white matter compromise observed in HIV-infected individuals even in the current ART era.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00018-023-04757-0.

Introduction

Human Immunodeficiency Virus (HIV) has infected more than 85 million since the start of the epidemic (https://www.unaids.org/en/resources/fact-sheet). 37.7 million people live globally with HIV, and almost 70% have access to antiretroviral therapy (ART) (https://www.who.int/gho/hiv/epidemic_response/ART/en/). Although ART is clinically successful, HIV remains incurable due to the early generation of circulating and tissue-associated viral reservoirs that can repopulate the body with the virus upon ART interruption [1]. ART has prolonged the lifespan of HIV-infected individuals, accelerating several comorbidities, including HIV-associated neurocognitive disorders (HAND) [2, 3].

Today, HIV-infected individuals under effective ART do not present neuronal loss or apparent neuropathology indicating subtle or localized damage as suggested for many groups [4, 5]. Brain imaging of patients with HAND showed localized white matter and less pronounced gray matter damage in correlation with lower nadir CD4 [6] and cognitive impairment in suppressed individuals [7, 8]. The main mechanism of the formation of the white matter abnormalities is unclear, but several groups propose demyelination, inflammation, synaptodendritic injury, and vascular compromise [9–12]. In the same direction, several groups identified potential sera and CSF biomarkers that could predict chronic/localized CNS damage, including neurofilament light chain (NFL), β-Amyloid_1-42, calcium-binding protein B (S100B), neopterin, Cathepsin B, kynurenine to tryptophan ratio, monocyte chemoattractant protein-1 (MCP-1 or CCL2), ATP, and several growth factors (see review [13]). Despite the long list of potential biomarkers of HIV CNS disease, most are associated with late events of tissue destruction and cannot detect early signs of CNS compromise. Also, some types of ART have been described to have toxicity in oligodendrocyte maturation, suggesting that HIV and ART are toxic to the white matter [14, 15]. However, overall, the mechanisms of spatial and highly localized damage and associated inflammation are unknown.

Here, we demonstrate that HIV reservoirs, myeloid and glial, locally compromise the myelin sheath resulting in the release of sulfatide species. Sulfatide is a glycosphingolipid class that exhibits a galactosyl-3-O-sulfate polar head group and a ceramide moiety composed of fatty acid and a long-chain base [16]. Sulfatide is highly expressed in brain tissue [17] and is mainly localized in the myelin sheath [18]. Sulfatide and the enzymes involved in its synthesis and degradation have been associated with several diseases, including brain cancer, subcortical dementia [19], multiple sclerosis [20], endothelial damage [21], and Alzheimer’s disease [22, 23], but their role in HIV and NeuroHIV is unknown.

Our manuscript demonstrates that HIV reservoirs are associated with damaged myelin tracks resulting in sulfatide release. Soluble sulfatide compromised gap junctional communication and calcium waves, promoting the amplification of chronic CNS damage observed in the HIV-infected population under ART.

Materials and methods

Human tissue sections and myelin structure analysis

Human brain tissues were obtained from the NeuroBioBank (https://neurobiobank.nih.gov/) and the National NeuroAIDS Tissue Consortium (NNTC; www.NNTC.org). The patient information is summarized in Table 1. All tissues analyzed had a similar postmortem interval < 24–48 h. Frozen fresh tissues were cut in 25 μm thick serial sections using a Leica CM1850 cryostat (Buffalo Grove, IL) and thaw-mounted onto frosted glass microscope slides. Serial sections were fixed with PFA 4% (Cat# 15,710-S, Electron Microscopy Science, Hatfield, PA) for 20 min and stored at 4 °C. The first two serial sections of human frontal cortex fixed tissues were stained for Luxol fast blue and Hematoxylin and Eosin, respectively, by the UTMB anatomical pathology core facility and scanned with the NanoZoomer.2ORS (Hamamatsu Photonics, Japan). Subsequent serial sections were incubated with proteinase K (Cat# AM2546, Thermo Fisher Scientific, Waltham, MA) for 10 min at room temperature, followed by the HIV DNA probe hybridization (Nef-PNA Alexa 488, Alexa488-GCAGCTTCCTCATTGATGG, PNA Bio, Thousand Oaks, CA). The probe was washed in preheated stringent wash working solution (PNA ISH kit, Cat# K520111-2, Dako products-Agilent Technologies, Santa Clara, CA) diluted at 1:60 at 55 °C. To perform antigen retrieval, the sections were incubated in antigen retrieval solution (Cat# S1700, Dako products-Agilent Technologies) at 80 °C for 30 min and permeabilized with Triton X-100 (Cat# X-100, Sigma-Aldrich, St. Luis, MO) 0.1% in TBS for 10 min. To block unspecific Fc receptors and other unspecific binding sites, our blocking solution contained: 1 ml 0.5 M EDTA pH 8.0 (Cat# 15,575–038, Thermo Fisher Scientific), 100 μl gelatin from cold-water fish skin (Cat# G7765, Sigma-Aldrich), 0.1 g bovine serum albumin (BSA immunoglobulin-free, Cat# A2058, Sigma-Aldrich, or BSA fraction V, Cat# BP1605, Thermo Fisher Scientific), 100 μl horse serum (Cat# H1138, Sigma-Aldrich), 5% human serum (Cat# MT35060CI, Thermo Fisher Scientific), and 9 ml ddH₂O was used to incubate the sections for at least 1 h at room temperature or overnight at 4 °C. Then, the primary antibodies (MBP, antimouse, diluted 1:500, Cat# 5M199, BioLegend, San Diego, CA; PLP, antimouse, diluted 1:200, Cat# MAB388; RRID: AB_177623, Millipore) were used to treat the sections overnight at 4 °C. The slides were washed several times with TBS at room temperature and incubated with the appropriate secondary antibodies conjugated for 2 h, followed by several washes in TBS. Then, tissues were mounted using Prolong Diamond Anti-Fade Mount Medium (Cat# P36930, Thermo Fisher Scientific). Tissues were examined by confocal microscopy using an A1 Nikon with a spectral detection system (Tokyo, Japan). Analysis and quantifications of 3D reconstructions were performed using NIS Elements and ImageJ software.

Table 1.

Patient Information

Sample	HIV-CS	Age	Sex	Plasma VL (log copies/ml)	CD4 (cell/mm3)	CSF VL (log copies/ml)	cART	Years with HIV
1	NA	51	M	NA	NP	NA	NA	NA
2	NA	46	M	NA	NP	NA	NA	NA
3	NA	40	F	NA	NP	NA	NA	NA
4	NA	43	M	NA	NP	NA	NA	NA
5	HIV + MND	54	M	5.88	18	3.27	3TC; TFV; SQV; RTV	13
6	HIV + MND	54	F	2.6	211	1.7	3TC; D4T; NVP; FTV	10
7	HIV + MND	41	F	4.09	566	1.89	DLV; D4T; 3TC	5
8	HIV + MND	49	M	3.32	1	3.36	KTA; CBV; EFV; ABC	18
9	HIV + HAD	58	M	5.7	5	3.14	D4T; ZDV; 3TC; KTA	23
10	HIV + HAD	47	F	2.6	98	4.34	RTV; ZDV; 3TC; KTA	2
11	HIV + HAD	44	F	3.54	78	4.88	RTV; IDV; D4T; 3TC	4
12	HIV + HAD	55	F	1.3	8	1.3	TMC; TFV; RGV; FTC	7
13	NA—AD	54	M	NA	NP	NA	NA	NA
14	NA—AD	45	F	NA	NP	NA	NA	NA
15	NA—AD	71	F	NA	NP	NA	NA	NA
16	NA—AD	85	F	NA	NP	NA	NA	NA

Sample HIV-CS Age Sex Plasma VL (log copies/ml) CD4 (cell/mm3) CSF VL (log copies/ml) cART Years with HIV

NA: not applicable; NP: not present; CS: cognitive status; AD: Alzheimer’s Disease; 3TC: epivir; KTA: anile heads; TFV: tenofovir; SQV: saquinavir; RTV: ritonavir; DLV: delavirdine; D4T: stavudine; NVP: Nevirapi ne; FTV: Fortovase; CBV: combivir; IDV: indi navir; EVF: efavirenz; ABC: abacavir; ZDV: zidovudine; TMC: etravirine; RGV: raltegravir; FTC: emtricitabine.

Cell culture methods

Cortical human fetal tissue was obtained as part of a research protocol approved by the Albert Einstein College of Medicine and collected between 2011 and 2018. The tissue collection protocols were approved by the Albert Einstein College of Medicine, Rutgers University, and the University of Texas Medical Branch Institutional Review Board (Protocol Numbers, Pro20140000794, Pro2012001303, 18–0136, 18–0135, 18–0134 to E.A.E). Human astrocyte cultures were prepared as previously described [24]. High glucose Dulbecco’s modified Eagle’s medium (DMEM, Cat# 11,995–065, Thermo Fisher Scientific) supplemented with 10% fetal bovine serum (FBS, Cat# S12450H, Atlanta Biologicals, Flowery Branch, GA), penicillin, and streptomycin (Cat# 15,070,063, Thermo Fisher Scientific) were used to grow the cells at 37 °C in a humidified atmosphere with 5% CO₂. The sulfatide used was a commercial mix derived from a bovine brain (Cat# 131,305, Avanti Polar Lipids), and the concentration selected was based on its lower pathological action of abnormally released lipid components at sites of injured and/or inflamed brain lesions [25].

HIV infection

Confluent cultures of human astrocytes were infected with HIV_ADA (20–50 ng/ml HIV-p24) using a described protocol [26]. Briefly, astrocyte cultures were exposed to the virus for 24 h, the medium was removed, and astrocytes were washed extensively to eliminate the unbound virus before adding a fresh medium. The media was collected on day 7 postinoculation.

Total mRNA isolation

Human primary astrocytes were treated with sulfatide 10 μg/ml (Cat# 24,323, Cayman Chemical, Ann Arbor, MI) for 6, 12, and 24 h. Untreated and treated cells were harvested in TRI reagent (Cat# 93,289, Sigma-Aldrich) according to the manufacturer’s instructions. Cells were scratched in TRI reagent (1 ml for 60 mm well plate), and 0.2 ml of chloroform (Cat# 288,306, Sigma-Aldrich) was added to the samples. After 3 min, the mixtures were transferred to 1 ml Eppendorf containing 0.5 ml of phase-lock gel (Cat# 2,302,830, QuantaBio, Beverly, MA) and centrifuged at 12,000 × g for 15 min at 4 °C to separate the RNA (aqueous phase) from proteins (red organic phase) and DNA (interphase). The aqueous phase was transferred to a fresh tube containing 0.5 ml of isopropanol (Cat# 278,475, Sigma-Aldrich). After 10 min, the samples were centrifuged at 12,000 × g at 4 °C for 15 min. The supernatant was removed, and 1 ml of 75% ethanol (Cat# 51,976, Sigma-Aldrich) was added to the RNA pellet at the bottom of the tube. The mixture was centrifuged at 7500 × g at 4 °C for 5 min. The RNA was eluted from the filter in 100 μl of DEPC-treated water (Cat# AM9938, Thermo Fisher Scientific) and warmed for 10–15 min at 60 °C to dissolve fully. The RNA extract obtained was stored at -20 °C until use. The concentration of each sample was calculated by spectrophotometric analysis (OD 260/280) using NanoDrop 2000 UV − Vis Spectrophotometer (Thermo Fisher Scientific).

Reverse transcription PCR

Reverse transcription for first-strand cDNA synthesis was performed using the SuperScript III First-Strand (Cat# 18,080–051, Thermo Fisher Scientific) according to the manufacturer’s instructions. 50 µM oligo (dT) was complexed with 2 μg of total RNA in a final volume of 10 μl. Samples were incubated at 65 °C for 5 min and then placed on ice for at least 1 min. 10 μl of cDNA synthesis mix (containing 10 × R.T. buffer, 2 mM MgCl₂, 0.1 M DTT, 40 U/μl RNaseOUT, 200 U/ μl SuperScript III RT) was added before the use of the Thermocycler (50 °C for 50 min, 85 °C for 5 min) (Cat# 170–9703, BioRad, Hercules, CA). The cDNA obtained was stored at − 20 °C until use.

Quantitative real-time PCR (qRT-PCR)

The amplified cDNA was used to amplify and quantify GAPDH, Cx43, and ZO-1 mRNA expression by qRT-PCR using ABsolute Blue qPCR SYBR low ROX mix (Cat# AB4323A, Thermo Fisher Scientific) in a StepOnePlus Real-Time PCR system (Cat# 4,376,600, Thermo Fisher Scientific). The primers used correspond to human GAPDH forward: 5′-CTTGACGGTGCCATGGAATTTG-3′, GAPDH reverse: 5′-GGGTTAAGGGAAAGAGCGACC-3′; Cx43 forward: 5′-CCCCATTCGATTTTGTTCTGC-3′, Cx43 reverse: 5′-GGGTTAAGGGAAAGAGCGACC-3′; ZO-1 forward: 5′-TTTTAGGATCACCCGACGAC-3′; ZO-1 reverse: 5′-CGCCTTTGGACAAAGAGAAG-3′. The program used was denaturation for 15 min at 95 °C and 40 cycles of denaturation, 30 s at 95 °C; anneal, 30 s at 56 °C; and amplification, 30 s at 72 °C. According to the C.T. values, expression was determined using the ΔΔCT method (Applied Biosystems, Life Technologies).

Western blotting

Human primary astrocytes untreated and treated with sulfatide for 6, 12, and 24 h were harvested in RIPA buffer (Cat# 9806, Cell Signaling, Danvers, MA) containing protease and phosphatase inhibitors (20 mM; pyrophosphate, 20 mM; NaF, 100 mM; NaVO₃, 200 μM; leupeptin, 500 μg/ml; aprotinin, 40 μg/ml; soybean trypsin inhibitor, 2 mg/ml; benzamidine, 1 mg/ml; ω-aminocaproic acid, 1 mg/ml; PMSF, 3 mM; and EDTA, 20 mM) and lysed. The protein content of each cell lysate was determined using Bradford’s method (Bio-Rad Labs). Samples containing 10 μg of protein were digested with 30 U of bovine intestinal alkaline phosphatase (Cat# P0114-10KU, Sigma-Aldrich) for 20 h at 4 °C to eliminate the Cx43 phosphorylation and collapse all the isoforms into the nonphosphorylated isoform [27–29]. Other undigested samples were used to analyze human Cx43, ZO-1, and GAPDH total protein levels. Proteins were separated in 7.5% SDS-PAGE, transferred to a nitrocellulose membrane, and incubated sequentially with a blocking solution (5% nonfat milk in Tris-buffered saline). Antibodies anti-Cx43 (antirabbit, dilution 1:1000, Cat# C6219, RRID:AB_476857, Sigma-Aldrich, St. Luis, MO), anti-ZO-1 (antirabbit, dilution 1:1000, Cat# 40–2200, RRID:AB_2533456, Thermo Fisher Scientific), anti-GAPDH (antirabbit, dilution 1:1000, Cat# 2118, RRID:AB_561053, Cell Signaling), and antirabbit IgG conjugated to HRP (dilution 1:1000, Cat# 7074, RRID:AB_2099233, Cell Signaling) were used. Antigen − antibody complexes were detected by ECL (Cat# NEL103E001EA, Perkin Elmer, Boston, MA). The resulting immunoblot signals were scanned, and densitometric analysis was performed using Image Studio Lite Ver 5.2 software. All results were normalized to the values obtained for control conditions.

Immunofluorescence

Human primary astrocytes were grown on coverslips and treated with sulfatide (resulting in a final concentration of 10 μg/ml) for 6, 12, and 24 h. Untreated and sulfatide-treated cells were fixed with 4% PFA (Cat# 15,710-S, Electron Microscopy Science, Hatfield, PA) and permeabilized in 0.1% Triton (Cat# X-100, Sigma-Aldrich) for 2 min at room temperature. Cells were incubated in blocking solution (0.5 M EDTA pH 8.0; Cat# 15,575–038, Thermo Fisher Scientific), 1% fish gelatin from cold water (Cat# G7765, Sigma-Aldrich), 0.1 g albumin from bovine serum-immunoglobulin free (Thermo Fisher Scientific), 1% Horse Serum (Cat# H1138, Sigma-Aldrich), 5% human serum (Cat# 31,876, Thermo Fisher Scientific), 9 ml ddH₂0) overnight at 4 °C and then in diluted primary antibodies (anti-Cx43, antirabbit, dilution 1:1000, Cat# C6219, RRID: AB_476857, Sigma-Aldrich, St. Luis, MO; anti-ZO-1, antimouse, dilution 1:200, Cat# MABT11, RRID: AB_10616098, Millipore) overnight at 4 °C. Cells were washed several times with PBS (Cat# BP665-1, Thermo Fisher Scientific) at room temperature and incubated with the appropriate secondary antibodies for at least 2 h at room temperature followed by several washes in PBS. Cells were then mounted using Prolong Gold Anti-Fade Reagent (Cat# P36930, Thermo Fisher Scientific) and examined using an A1 confocal microscope with a spectral detection device (Nikon, Japan). Antibody specificity was confirmed by replacing the primary antibodies with the appropriate isotype-matched control reagent, anti-IgG₂A, or the IgG fraction of normal mouse/rabbit serum (Cat# A2179, RRID: AB_257981/ Cat# A0418, RRID: AB_257885, Sigma-Aldrich).

MALDI-MSI analysis

Serial sections of the unfixed human brain tissues were cut at 12 µm using a Leica CM 1860 cryostat and thaw-mounted onto Indium Tin Oxide (ITO) glass slides (Delta Technologies, CO). After sectioning, tissues were placed in a desiccator for 15 min and then transferred to a − 80 °C freezer for storage. Prior to MALDI-MSI analysis, the thaw-mounted tissue sections were removed from the − 80 °C freezer and allowed to reach room temperature for 15 min. 1,5-Diaminonaphthalene (D21200, Sigma-Aldrich) dissolved in 70% Acetone (34,850, Sigma-Aldrich) was used as a matrix for negative ion mode lipid imaging. This matrix was applied to the tissue by automated air spray deposition using the TM-sprayer (HTX Technologies LLC, Chapel Hill, NC). The nozzle temperature was set to 60 °C, the flow rate was 50 µL/min, and 20 passes over the tissue were performed. MALDI-MSI analysis was performed using Q Exactive HF Hybrid Quadrupole-Orbitrap Mass Spectrometer (Thermo Fisher Scientific) equipped with an Elevated Pressure Matrix-Assisted Laser/Desorption Ionization (EP MALDI) source integrating an Nd: YAG laser (Spectroglyph LLC, Kennewick, WA). Data were acquired at 40 µm² lateral resolution in negative ion mode at a laser power of 7 µJ. External mass calibration was performed using calibration solutions (A39239, Thermo Fisher Scientific). The accurate mass-measured lipid peaks were identified by matching reference lipids in the LIPID MAPS and Human Metabolome Database within a ± 0.002 Da mass tolerance window. Thermo RAW format data files were converted to imzML using ImageInsight software (Spectroglyph LLC, Kennewick, WA). Data visualization was performed using SciLs software (SCiLS GmbH, Bremen, Germany).

Scrape loading assay

The functionality of G.J. channels was evaluated by scrape loading using a previously published protocol [30]. In brief, 0.5% Lucifer yellow (a fixable dye permeable to G.Js, 5% w/v in 150 mM LiCl; 520-Da dye, Cat# L0144, Sigma-Aldrich), and Dextran-Alexa Fluor 594 (dye not permeable to G.J.; 10,000 k.Da., Cat# D22913, Thermo Fisher Scientific) dissolved in PBS were added to confluent primary astrocytes treated with sulfatide for 6, 12 and 24 h, and a scrape using a blade was performed. The dye solution was left on the cells for 5 min and discarded, and the plate was rinsed with PBS. Gap junctional permeability was also inhibited by adding the blocker 18-α-glycyrrhetinic acid (AGA, 50 µM in DMSO, Cat# G8503, Sigma-Aldrich) or 1-Octanol (0.5 mM, Cat# 297,887, Sigma-Aldrich) for 5 min in the media before dye permeabilization.

To examine the unspecific uptake of L.Y. from the media, plates were prepared by exposing the cells under similar conditions to the dye mixture, but without scraping. The marker dye, dextran-Alexa Fluor 594, remained entrapped, thus labeling the primary loaded cells at the edge of the scrape areas. The damaged cells could pick up the dye mixture and transfer L.Y. into the neighboring cells through functional G.J. The distance or L.Y. diffusion from the damaged area was normalized by the distance or diffusion of the impermeable dextran-Alexa Fluor 594 and measured using the confocal microscope Nikon A1. The percentage of coupling was calculated by direct comparison to untreated cultures.

Calcium imaging

Human primary astrocytes are seeded in 35 mm Glass Bottom MatTek Dishes (Cat# P35G-1.5–14-C, MatTek Corporation, Ashland, MA). The confluent cells were transferred to a temperature-regulated chamber at 37 °C of the confocal microscope Nikon A1 and incubated with 10 μM fluo-4 diluted in DMSO (Cat# F14201, Thermo Fisher Scientific); photometric data for [Ca²⁺] were generated by exciting cells at 488 nm and measuring emission at 516 nm every 5 s for 20 min. An intracellular calibration was performed with each experiment by determining the fluorescence in the absence of a Ca²⁺ indicator (fluorescence_min) for 5 min and the presence of 3.6 μM ATP (Cat# A1852, Sigma-Aldrich) and 2 μM ionomycin (fluorescence_max) (Cat# I3909, Sigma-Aldrich). The mean [Ca²⁺] was determined from three independent plate areas and analyzed 20 cells per position.

Statistical analysis

All data were expressed as mean ± standard deviation (S.D.). Differences among groups were analyzed using a t test with Welch’s Correction or the Holm − Sidak method (alpha = 0.05). In addition, we used a 2-way ANOVA multiple-compassion test for multiple comparisons. The level of significance was accepted at p ≤ 0.05. GraphPad Prism 8 software was used for the statistical analyses performed.

Results

Demographics of the tissue samples analysed

Table 1 summarizes the demographics of the individuals analyzed. The population analyzed did not show differences among the uninfected and the HIV-positive groups with MND (HIV-MND) and HAD in age, except for the uninfected with Alzheimer’s disease group that includes two young and two old subjects (HIV-negative, mean = 45 ± 4.7 years; HIV-positive with MND, mean = 50 ± 6.1 years; HIV-positive with HAD, mean = 51 ± 6.6 years; HIV-negative-Alzheimer’s disease = 63 ± 17.8 years; Table 1). Sex was equally distributed (HIV-positive = 37.5% female and 62.5% male; HIV-negative = 50% female and 50% male; Table 1). The HIV-positive cohort had an average of 11.5 ± 5.4 years living with HIV for the MND group and 9 ± 9.6 for the HAD group. HIV-MND cohort had a mean of HIV RNA of 4 ± 1.4 log copies/ml and 2.6 ± 0.9 log copies/ml in the plasma and CSF, respectively, and an average CD4 count of 199 ± 262.6 cells/mm³. The HAD cohort had a mean HIV RNA of 3.3 ± 1.9 log copies/ml and 3.4 ± 1.6 log copies/ml in the plasma and CSF, respectively, and an average CD4 count of 47.3 ± 47.8 cells/mm³ (see Table 1).

Myelin structure is compromised in the brain areas containing HIV reservoirs.

Several laboratories using different imaging and histological techniques demonstrated that HIV-infected individuals under ART show localized damage or hyper-densities in the white matter (W.M.) [31–33]. To determine whether HIV brain tissues obtained from individuals with and without cognitive impairment had W.M. compromise (HIV-MND and HAD), we analyzed the status of the myelin tracks compared with uninfected (control) and Alzheimer’s brain tissues (A.D.). Alzheimer’s brain tissues were included as a positive control for demyelination due to the severity of the selected cases.

Serial sections of the prefrontal cortex were selected due to documented synaptic − dendritic damage in HIV-infected individuals [34]. Gross histological analysis was performed using the Luxol Blue and H&E staining (Fig. 1A–H). Luxol Blue stains myelin lipoproteins in blue and the neurons in violet (Fig. 1A–D). A remarkable separation between the grey and white matter was observed in brain tissues obtained from control-uninfected and HIV-MND/HAD individuals (Fig. 1A, B, C). In the brain tissues obtained from individuals with severe Alzheimer’s disease, W.M. and G.M. could not be delineated based on the Luxol blue or H&E-stained images due to the severe demyelination process (AD, Fig. 1D,H).

Fig. 1.

Human prefrontal cortex tissue samples obtained from HIV-infected individuals with MND and HAD showed a decreased expression of myelin structural proteins compared to control uninfected and Alzheimer’s disease tissues. Human prefrontal cortex obtained from control uninfected, HIV-infected with MND (HIV-MND), HAD, or Alzheimer’s Disease (A.D., used as a positive control) patients were analyzed by histologic and immunofluorescence staining for MBP and PLP. A–D Representative images of Luxol Blue stained tissues for all the conditions analyzed, the inset corresponds to the amplification shows the interface area between the W.M. and G.M. E–H Representative images of H&E stained tissues for all the conditions analyzed, the inset corresponded to the amplification shows the interface area between the W.M. and G.M. I–L Distribution of MBP at the WM/GM interface (the white line indicates the interface). M–P Distribution of PLP at the interface (the white curve indicates the interface). Q Quantification of the total pixels per area for MBP staining in control (black curve), HIV-MND (red curve), HAD (blue curve), and A.D. (green curve) human brain tissues for MBP in the closer W.M. and G.M. to the interface (200 µm); data points were analyzed by 2-way ANOVA multiple compassion test (n = 3). R Quantification of the total pixels per area for PLP in control (black curve), HIV-MND (red curve), HAD (blue curve), and A.D. (green curve) human brain tissues for PLP in the closer W.M. and G.M. to the interface (200 µm); data points were analyzed by 2-way ANOVA multiple compassion test (n = 3). For graph Q, all data points for AD and HAD were significant, except point -30 for HIV MND and points after 180 µm in the gray matter in HAD conditions. For graph R, all data points significantly differed from the control conditions. Overall, our data indicate that HIV infection compromises the WM/GM interface and expresses key myelin-related proteins, MBP and PLP

Confocal and image analysis for MBP (Fig. 1I–L, WM/GM interface is indicated in white lines) and PLP (Fig. 1M–P, WM/GM interface is indicated in white lines) indicates that both proteins were concentrated within the W.M. in uninfected brains (Fig. 1, I and M, control). Quantification of the positive pixel versus distance from the W.M./G.M. interface confirmed that MBP and PLP were concentrated in the W.M. in uninfected brains (Fig. 1Q, R, respectively). In brains obtained from HIV-infected individuals with MND and HAD, a reduction in MBP (Fig. 1J, K, quantification in Q) and PLP (Fig. 1N, O, quantification in R) staining were detected in the W.M., interface, and the G.M. The reduction in MBP and PLP staining was not associated with alterations in the numbers of cell bodies; instead, a reduction in cell processes was detected (see Fig. 1 J–K and N–O, respectively). Analysis of brains obtained from individuals with severe A.D. indicates that MBP and PLP distribution (Fig. 1L, P, respectively) and expression (Fig. 1Q, R, respectively) were reduced at the W.M., interface, and G.M. compared to uninfected-control and HIV conditions (Fig. 1Q, R). Overall, our data indicate that HIV infection and associated cognitive impairment compromised the distribution and expression levels of MBP and PLP.

HIV infection compromises myelin structure

To further examine the characteristic of the myelin compromise induced by HIV, we determined in more detail the structure and distribution of the myelin sheath (thickness of the myelin around the axons and the length between the Ranvier’s nodes, see cartoon in Fig. 2A) using higher resolution confocal microscopy, 3D reconstruction, and deconvolution. To perform these analyses, tissues in two orientations were used, sagittal and coronal, to observe cross and longitudinal axons. First, cross-sections of axons are represented as “donut” shaped MBP structures with the axon in the center (Fig. 2B–E). Second, longitudinal axons enable us to determine the length of the myelin sheath, including Ranvier’s node (Fig. 2F–I).

Fig. 2.

Human prefrontal cortex tissues obtained from HIV-infected individuals with HIV-MND and HAD show a decreased myelin thickness and compromised organization. Human prefrontal cortex isolated from control uninfected, HIV-infected with MND (HIV-MND) and HAD, and Alzheimer’s Disease (A.D.) patients, used as a positive control, were analyzed by immunofluorescence staining for MBP. A The cartoon shows the structure of the myelin sheath and the distribution of our interesting proteins, MBP and PLP. B–E Distribution of MBP in the G.M. of human brain tissues to analyze myelin thickness (star indicates magnification at the right corner). F–I Distribution of MBP in the W.M. of human brain tissues to analyze myelin length (star indicates magnification at the right corner). J Quantification of the myelin thickness in µm in the W.M. and G.M. of control (black bars), HIV-MND (red bars), HAD (blue bars), and A.D. (green bars) human brain tissues; data points were analyzed by 2-way ANOVA multiple compassion test (*p ≤ 0.0001 compared to G.M. Control). K Quantification of the myelin length in µm in the W.M. and G.M. of control (black bars), HIV-MND (red bars), HAD (blue bars), and A.D. (green bars) human brain tissues; data points were analyzed by 2-way ANOVA multiple compassion test (*p ≤ 0.05 compared to G.M. Control; n = 3)

The analysis of cross-sections obtained from uninfected control individuals indicates that MBP staining provided a myelin sheath thickness of 5.5 ± 1.1 µm (Fig. 2B). In contrast, the analysis of brain tissues obtained from HIV-infected individuals with MND or HAD and stained for MBP shows a compromised myelin sheath thickness of 4.3 ± 0.8 µm and 3.5 ± 1.1 µm, respectively, in the G.M., but no significant changes were observed in the W.M. (Fig. 2C, D, respectively). Analysis of brain tissues obtained from severe cases of A.D. shows compromised myelin thickness in the G.M. compared to uninfected control and HIV brain tissues (Fig. 2J, *p ≤ 0.0001 as compared to G.M. Control, 3 replicates, 4 different tissues). However, there were no differences in W.M. thickness among control uninfected, HIV-infected (MND or HAD), and A.D. tissues (Fig. 2J, W.M.). In conclusion, our MBP staining data indicate that myelin thickness is reduced in the brains obtained from HIV-infected and severe A.D. cases (Fig. 2J).

The analysis of sagittal uninfected brain sections to observe longitudinal axon structures shows a repetitive pattern consistent with the distribution and distance of the myelin sheath separated by the Ranvier’s nodes (Fig. 2A, F, and K, cartoon, staining, and quantification, respectively). There are no changes in the myelin length among the brains of uninfected and HIV individuals with MND or HAD (Fig. 2K, quantification for the W.M. and G.M.). These data indicate that the distances between the Ranvier’s nodes were constant. In contrast, the analysis of brains from A.D. individuals indicated a significant G.M. myelin length compromise suggesting axonal length and conduction issues (Fig. 2K, A.D., *p ≤ 0.05 compared to G.M. Control, 3 replicates, 3 different tissues). Overall, our data indicate that HIV infection compromises myelin structures, especially the thickness of the myelin sheet in the G.M. However, these data are partially inconsistent with the in vivo data, where most of the damage is observed in the W.M.

Myelin damage in the G.M. is associated with the presence of viral reservoirs

A significant observation is that myelin damage was not uniform within the brain, especially in the W.M. Thus, we hypothesize that myelin damage was associated with the presence of long-lasting HIV-infected cells, also called viral reservoirs. We performed tissue staining for HIV-DNA to demonstrate our hypothesis using a Nef DNA probe and staining for nuclei (DAPI) and MBP or PLP (Fig. 3). In control uninfected and A.D. tissues, no staining for HIV Nef DNA was observed (Fig. 3B and N). Also, as described in the previous figures, MBP and PLP thickness in the G.M.(overall quantification), but not in the W.M., and the myelin sheath length were not affected by long-term HIV infection, but in severe A.D. cases, myelin was highly compromised.

Fig. 3.

Myelin structures are compromised in areas with HIV reservoirs. The human prefrontal cortex was obtained from control uninfected, HIV-infected with MND (HIV-MND) and HAD, and Alzheimer’s Disease (A.D.) patients, used as a positive control, and analyzed by immunofluorescence staining for DAPI (blue), HIV DNA for Nef (green), and MBP (red). DAPI staining was reduced to appreciate the HIV DNA signal. A–D correspond to a representative analysis of control uninfected brains. E–H Correspond to representative staining of an HIV-infected individual with MND and positive HIV DNA cells. I–L Correspond to a representative analysis of brains obtained from HIV-infected individuals with HAD. M–P Correspond to a representative example of uninfected Alzheimer’s (A.D.) brains. Circles in each picture represent areas with HV DNA-positive cells. Q Quantification of the total pixels of MBP/PLP per area unit in the absence (HIV(-)DNA area, black bars) and presence of viral reservoirs (HIV( +)DNA area, white bars). The student’s t test analyzed data points with Welch’s Correction for significance (*p ≤ 0.05 compared to G.M. Control; n = 3 different individuals)

However, we identified brain areas with compromised myelin structures associated with cell clusters containing viral DNA at the W.M. and G.M. under MND and HAD conditions (Fig. 3E–H, I-L, respectively, see dotted circles). Viral reservoirs corresponded to microglia/macrophages (Iba-1 positive cells) and astrocytes (GFAP positive cells), and the ratio of both infected cell types depends on the ART length. The re-assessment of myelin compromise described in Fig. 1 and 2, based on the presence of HIV-DNA, indicated that 82.02 ± 11.6% of the areas with HIV-DNA were associated with decreased MBP and PLP expression and compromised myelin thickness compared to uninfected brains (Fig. 3, uninfected controls set to 100 ± 26.3 A.U.). We determine that myelin compromise from HIV DNA-positive cells could reach distances up to 250 ± 12.5 µm, suggesting a mechanism of chronic damage amplification under ART conditions. The quantification of the MBP/PLP expression around viral reservoirs (HIV-positive DNA cells and up to 250 ± 12.5 µm around these cells, Fig. 3Q, white bars) indicates that myelin components were compromised more around viral reservoirs than in areas without viral reservoirs (Fig. 3Q, black bars, MND, white bars compared to black bars, *p ≤ 0.05, n = 3 different tissues per patient, n = 16 and 8 from HIV infected individuals). The lack of difference in myelin damage between HIV-MND and HIV-HAD also was not associated with the numbers of HIV-nef DNA positive cells (more abundant in HAD conditions), systemic replication, CD4 counts, age, sex, ART, and years with HIV (see Table 1) suggesting that even in the ART era chronic myelin damage is still significant. In addition, there are no significant differences in MBP/PLP expression in areas without HIV DNA or at distances over 250 µm. In conclusion, myelin compromise is associated with the presence of viral reservoirs, and the degree of damage is maintained in HIV-MND and HAD cases.

Sulfatide is a key dysregulated lipid in HIV-associated myelin damage: a potential biomarker of HIV CNS disease

To examine further the mechanism of myelin compromise induced by HIV-infected cells, we performed an untargeted lipid screening using MALDI-MSI. Our screening indicates that sulfatide, a key lipid family involved in maintaining myelin structure, was compromised in HIV conditions. Sulfatide comprises 4–7% of all myelin lipids, plays a critical role during myelin compaction, and remains essential in the compacted myelin [35]. However, soluble sulfatides due to myelin damage promote inflammation [25], oligodendrocyte survival [36], and prevent axonal outgrowth [37], but the mechanisms associated with the damage in HIV conditions remain unknown. Importantly, sulfatide species have been proposed to have strong interactions with HIV-gp120, altering viral fusion and HIV-associated dementia (see review [38]). Thus, we examined the expression and distribution of sulfatide in uninfected and HIV-infected (high and low systemic replication) brains by MALDI–MSI.

MALDI–MSI enables us to detect and visualize the distribution of multiple sulfatide species comprised of fatty acids of varying carbon length and saturation. Serial sections were used to perform H&E staining and MALDI–MSI. The analysis of different sulfatide species indicates that in control uninfected conditions, sulfatide is concentrated in the W.M. as expected [39] (examples of 18:0, 18:1, 22:0, 24:0, 24:1, and 26:1, as well as their hydroxylated species, are shown, Fig. 4., and Supplemental Fig. 1). Shorter sulfatide species such as 18:0, 18:1, 22:0 and their hydroxylated species, such as 18:0 O.H., 18:1 O.H., and 22:0 O.H., were not altered in HIV (MND or HAD) and A.D. conditions (Fig. 4A–I). However, the analysis of 24 carbon chain sulfatides (24:0, 24:1 and hydroxylated species, 24:0 O.H. and 24:1 O.H.) indicates an upregulation of these sulfatides, known to be the primary sulfatides comprising mature myelin [40], in the W.M. in HIV conditions, MND and HAD. Moreover, other long-chain sulfatides, 26:1 and 26:1 O.H., were not affected by HIV or A.D. status (Fig. 4 and Supplemental Fig. 1). Surprisingly, the increased levels and specificity of the sulfatide species were similar to the changes observed in myelin tracks in brain tissue from severe cases of A.D. (Fig. 4 and Supplemental Fig. 1). Further, our MALDI-MSI data indicates two critical points; first, myelin damage is specific for major myelin sulfatide species. Second, the damage to the myelin sheath is comparable to the extensive damage observed in brains obtained from A.D. individuals. Supporting these points, structural membrane lipids like P.A. 39:3 (m/z 726.5) or P.E. 40:6 (m/z 790.5) were not affected by HIV infection or the cognitive impairment status, MND or HAD (Supplemental Fig. 2, n = 4 different tissues). They were expressed in the W.M. and G.M. and used as markers to delineate the tissue regions. Overall, our data indicate that HIV infection induces the dysregulation of major myelin sulfatide species and could provide early indicators of CNS compromise in the HIV-infected population.

Fig. 4.

Sulfatide 24:0, 24:1, and their hydroxylated isoforms are upregulated in HIV-MND and HAD, similar to severe A.D. cases. The human frontal cortex of uninfected (Control), HIV-infected with MND (HIV-MND), HIV-infected with HAD (HAD), and A.D. patients were analyzed by MALDI-MSI. A–D Representative MALDI MS images of uninfected control, HIV-MND, HAD and A.D. brains for sulfatide 24:0 (n = 5, three representative brains are shown). E–H Representative MALDI-MS images for sulfatide 24:1 (n = 5, three representative brains are shown). Statistically significant differences were not observed for 18:0, 18:1, 22:0, and 26:1 sulfatide (See supplemental Fig. 1 for hydroxylated sulfatide species). All sulfatide species analyzed were highly concentrated in the W.M. Sulfatides 24:0 and 24:1 are known to be present in adult myelin and are significantly increased in NeuroHIV. Note that sulfatide compromise was similar to severe A.D. cases, suggesting that damage in HIV in some cases can be comparable to severe A.D. cases. The gradient intensity scale expresses relative intensity. In contrast, P.A. and P.E. (Supplemental Fig. 2) were not affected and were used as markers for the W.M. and G.M., respectively. Analysis of the hydroxylated sulfatide species was performed similarly to the sulfatide species described in the current figure (Supplemental Fig. 1). (I) Quantifying the WM ROI lipid signal intensity for sulfatide 18:0, 18:1, 22:0, 24:0, 24:1 and 26:1 in control, HIV-MND, HIV-HAD, and A.D. conditions. Data points were analyzed by 2-way ANOVA Dunnett's multiple comparisons test for significance (*p ≤ 0.05 compared to Control; n = 5)

Sulfatide compromises the generation of calcium in pace markers astrocytes

Our data indicate that HIV infection and compromised cognition in the HIV-infected population are associated with increased sulfatide levels and myelin structural changes. However, whether soluble sulfatide can alter brain function is unknown. We hypothesize that sulfatide could participate in the bystander damage indicated above. Currently, two critical observations support long-range or bystander-associated damage generated by HIV-infected cells: first, the association of myelin damage with HIV-infected cells, and second, the compromise of myelin integrity suggesting a long-range electrical signaling dysfunction. Our laboratory identified gap junction (G.J.) channels and hemichannels (H.C.) as key systems to amplify bystander damage mediated by viral reservoirs by an IP₃ and calcium-mediated mechanisms even in the absence of viral replication [41, 42]. We determined calcium waves to measure cell-to-cell communication and bystander damage—both critical mechanisms related to viral reservoirs associated dysfunction in neighboring uninfected areas.

As we described, human primary astrocytes were plated on Matek plates in uninfected or HIV_ADA-infected conditions (7 days post-infection) [24, 26]. Calcium imaging analysis of the primary human astrocyte cultures identified two astrocyte populations: pacemakers (with intrinsic calcium oscillations) and nonpacemakers (responders to the pacemakers). Analyzing these cell populations enables us to determine the period amplitude, peak numbers, intensity (area under the curve), and frequency (see an example of the signaling in Fig. 5 and Supplemental Fig. 3). Figure 5A represents all the analyses performed in the cultures, including the area under the curve that provides the calcium baseline (Fig. 5B), the number of calibrated peaks per time unit (discounting background intensity, Fig. 5C), the calibrated peak duration during the time analyzed (discounting background intensity, Fig. 5D), calibrated peak frequency (discounting background intensity) (Fig. 5E), and the calibrated amplitude (discounting background intensity, Fig. 5F) (see methods for details).

Fig. 5.

Soluble sulfatide compromises calcium waves in astrocyte pacemaker cells. Human primary astrocytes were infected with HIV_ADA (20–50 ng/ml HIV-p24) for 7 days and treated with 10 μg/ml sulfatide. After infection, cells were treated with 10 μM fluo-4 to perform calcium imaging for 20 min (still pictures collected every 5 s) in three different plate areas. A Graphical presentation of a typical recording to identify nonpacemaker (NPM) and pacemaker cells. Quantification for the NPM and Pacemaker Cells Control Uninfected, Sulfatide Uninfected, HIV-infected, Sulfatide HIV-Infected of B area under the curve; C number of the total peaks; D the period of each calcium waves; E frequency of the calcium waves; and F the amplitude of the calcium waves. Data points were analyzed by 2-way ANOVA multiple compassion test (*p ≤ 0.05 compared to pacemaker cells Control Uninfected; n = 3 different cultures of different individuals)

Quantification of the calcium baseline or area under the curve indicates neither HIV infection nor sulfatide treatment changed the calcium baseline in nonpacemaker (NPM) and pacemaker astrocytes (20 min post sulfatide treatment or after 7 days postinfection, Fig. 5B). However, the calcium peaks in uninfected pacemaker cells were reduced by sulfatide treatment (Fig. 5C, *p ≤ 0.05, n = 3 independent experiments), but not in NPM cells or the pacemaker cells in HIV-infected conditions (Fig. 5C). Periods were reduced by HIV infection and treatment with sulfatide recovered the number of calcium wave periods to control levels (Fig. 5D, *p ≤ 0.05, n = 3). No changes in the periods were found in NPM cells (Fig. 5D). Analysis of the calibrated frequency of the calcium signals indicates no changes in uninfected or HIV-infected NPM cells and uninfected pacemaker cells (Fig. 5E). However, when sulfatide was added to pacemaker cells in HIV-infected cultures the calibrated frequency was reduced (Fig. 5E, *p ≤ 0.0001, n = 3, see representative videos in Supplemental Fig. 3). The analysis of the amplitude of the calcium signals in uninfected and HIV-infected cultures in the presence and absence of sulfatide indicated no changes (Fig. 5F).

Overall, our data indicate that sulfatide added to uninfected and HIV-infected astrocyte cultures did not affect calcium signaling in NPM cells. However, sulfatide could compromise pacemaker cells in uninfected and HIV-infected conditions. These data indicate that HIV infection, combined with sulfatide, alters the calcium signaling between the pacemaker and nonpacemaker cells (responders), compromising the coordination between both cell types.

Soluble sulfatide compromises gap junctional communication

Our laboratory and others identified that G.J. channels are essential to coordinate calcium waves [41, 42]. Also, HIV infection of astrocytes upregulates Cx43 expression and GJ communication [43, 44], and blocking GJ communication prevents bystander damage induced by the virus [41, 42]. However, the participation of host factors such as sulfatide in regulating Cx43 channels is unknown.

We evaluated whether treating primary human astrocyte cultures with soluble sulfatide (10 µg/ml) altered Cx43 and ZO-1 expression using qRT-PCR and Western Blot analysis (Fig. 6). Sulfatide treatment of uninfected astrocyte cultures increased Cx43 mRNA expression in all the cultures analyzed (from different individuals) at different time points (6, 12 or 24 h) (Fig. 6A, 6.16 ± 3.39 folds at the peak of Cx43 mRNA, *p ≤ 0.05 at peak upregulation analyzed, n = 5).

Fig. 6.

Soluble sulfatide upregulates Cx43 and ZO-1 expression in human primary astrocytes. Human primary astrocytes were treated with 10 μg/ml of sulfatide for 6, 12, and 24 h; Cx43 mRNA and protein expression were analyzed. In all experiments, sulfatide treatment increased the expression of Cx43 and ZO-1 mRNA and protein at different time points, probably due to the interdonor variability of human primary astrocytes. Thus, the maximal peak was considered to calculate the data. A qRT-PCR for Cx43 and GAPDH mRNA using untreated (control, white bar) and sulfatide-treated (gray bar) astrocyte cultures. Sulfatide treatment increased Cx43 mRNA expression; data points for Cx43 were analyzed by student’s t test with Welch’s Correction for significance (*p ≤ 0.05, n = 5). B Representative immunoblotting for Cx43 (phosphorylated and unphosphorylated isoforms, P3, P2, P1, and N.P.) and GAPDH in control and sulfatide-treated human astrocytes for 6, 12, 24 and 48 h. The lysates were divided into two equal portions; one was treated with bovine intestinal alkaline phosphatase (A.P.) to assess the Cx43 total amount versus phosphorylation. C Western blot quantification for total Cx43 using the protein lysates of untreated (control, white bar) and sulfatide-treated (gray bar) astrocyte cultures. Sulfatide treatment increased total Cx43 protein expression; data points for Cx43 normalized by GAPDH were analyzed by student’s t test with Welch’s Correction for significance (*p ≤ 0.05, n = 5). D Western blot quantification for Cx43 using the protein lysates of untreated (control, white bar) and sulfatide-treated (gray bar) astrocyte cultures digested with bovine intestinal alkaline phosphatase (A.P.). Sulfatide treatment increased total Cx43 protein expression (phosphorylated and unphosphorylated); data points for Cx43 normalized by GAPDH were analyzed by student’s t test with Welch’s Correction for significance (*p ≤ 0.05, n = 5). (E) qRT-PCR for ZO-1 (a Cx43 partner at the plasma membrane) and GAPDH mRNA using untreated (control, white bar) and sulfatide-treated (gray bar) astrocyte cultures. Sulfatide treatment increased ZO-1 mRNA expression; data points for ZO-1 distributed normally were analyzed by student’s t test with Welch’s Correction for significance (*p ≤ 0.05, n = 5). F Immunoblotting and Western blot densitometry band quantification for ZO-1 in control and sulfatide-treated human astrocytes for 6, 12, and 24 h. As a loading control, GAPDH was used. Sulfatide treatment increased ZO-1 protein expression. Data points for ZO-1 normalized by GAPDH were analyzed by student’s t test with Welch’s Correction for significance (*p ≤ 0.05, n = 5)

Western blot analysis of astrocyte cultures indicates that sulfatide increased the Cx43 protein levels at all the time points analyzed, 6, 12, and 24 h (Fig. 6B, Cx43 and GAPDH as a loading control). However, there are no changes in the phosphorylation ratio of the protein, including nonphosphorylated, phosphorylated isoform 1 (P1), phosphorylated isoform 2 (P2), and phosphorylated isoform 3 (P3), and hyperphosphorylated (H.P., not shown). To demonstrate that the total Cx43 was increased over the phosphorylation, the lysates were incubated with alkaline phosphatase (A.P.) to eliminate any protein phosphorylation (Fig. 6B, D). The quantification of total unphosphorylated Cx43 protein confirmed our previous data that sulfatide increased Cx43. The quantification of the Cx43 protein upregulation indicates an increase of the total amount, Cx43 2.10 ± 0.62, and digested Cx43, 1.30 ± 0.28 (Fig. 6C and D, respectively. *p ≤ 0.05 at peak upregulation analyzed, n = 5 different astrocyte cultures from different individuals). A critical partner of Cx43 at the plasma membrane is the adaptor protein, ZO-1 [45]. Notably, ZO-1 interaction regulates G.J.’s dynamic turnover maintaining G.J. channels in the plasma membrane in a functional state. When ZO-1 separates from Cxs, it induces G.J.s closure and transitioning for endocytosis [45, 46]. qRT-PCR analysis for ZO-1 and GAPDH mRNA demonstrated that sulfatide treatment also upregulates ZO-1 mRNA as compared to untreated astrocytes (Fig. 6E, *p ≤ 0.05 timepoints at peak expression analyzed, n = 4). Furthermore, Western blot analysis indicated that sulfatide treatment increased ZO-1 protein (Fig. 6F, *p ≤ 0.05 at peaks expression analyzed, n = 3). These results suggest that extracellular or soluble sulfatide upregulates Cx43 and ZO-1 expression, both critical components of calcium wave coordination and cell-to-cell communication.

Soluble Sulfatide induces the maintenance of Cx43 at the plasma membrane

Untreated control and sulfatide-treated cultures were analyzed by confocal and image analysis for Cx43, ZO-1, actin (phalloidin), and nuclei (DAPI). 3D reconstructions indicate that Cx43 and ZO-1 are localized at the plasma membrane and internal vesicular stores in control conditions (Fig. 7A). Sulfatide treatment for 6, 12, or 24 h increased Cx43 and ZO-1 staining determined by the numbers of positive pixels (Fig. 7B, quantification in 7C, *p ≤ 0.05 at peak expression analyzed, n = 3), in agreement with our mRNA and protein data (Fig. 6A–F). Analysis of specific ROI, plasma membrane, G.J. plaques, and intracellular vesicles indicates an upregulation in each compartment. Colocalization analysis between Cx43 and ZO-1 was performed using Pearson’s colocalization [47]. In control, untreated conditions, most Cx43 colocalized with ZO-1, but sulfatide treatment increased the colocalization of both proteins (Fig. 7D, *p ≤ 0.05 at peak expression analyzed, n = 3). Overall, soluble sulfatide increased the expression and maintained the channels and the interacting proteins on the membrane, suggesting that releasing sulfatide from myelin tracks can profoundly affect gap junctional communication. Our data on neuroHIV indicates that G.J. are essential for spreading toxicity and apoptosis even in the ART era; thus, the release of sulfatide can further increase the damage radius.

Fig. 7.

Soluble sulfatide increased Cx43-containing G.J. at the plasma membrane in human primary astrocytes. Human primary astrocytes were treated with 10 μg/ml of sulfatide for 6, 12, and 24 h; expression and distribution of Cx43-containing channels were analyzed by confocal microscopy. In all experiments, sulfatide treatment increased the expression and distribution of Cx43 but at different time points, probably due to the interdonor variability of human primary astrocytes. Thus, the maximal peak was considered to calculate the data. Astrocytes were stained for Cx43 (Alexa 488, green staining), ZO-1 (Alexa 647, white staining), Phalloidin (Alexa Texas red, red staining), and the nucleus (DAPI, blue staining) in untreated (control) and sulfatide-treated conditions. A Correspond to control astrocytes cultures. The inset indicates the plasma membrane colocalization between Cx43 and ZO-1. B Corresponds to Sulfatide treated astrocytes. The inset indicates the increased expression and localization of Cx43 and ZO-1 at the plasma membrane. The last panel represents the merge of all colors. Arrows indicate the area in the inset. Bar: 110 μm. C Quantification of the total pixel number for Cx43 at the maximal peak of expression induced by Sulfatide (gray bar) compared to untreated conditions (control, white bar) and sulfatide-treated conditions. Data points for Cx43 were analyzed by student’s t test with Welch’s Correction for significance (*p ≤ 0.05, n = 3). D Quantification of the colocalization of Cx43 and ZO-1 in untreated (control, white bar) and sulfatide-treated (gray bar) conditions. For significance, the student’s t test analyzed data points for Cx43 and ZO-1 colocalization with the Holm − Sidak method (alpha = 0.05)

Soluble sulfatide enhanced gap junctional communication

We performed scrape loading-dye transfer assay experiments to determine whether the changes in calcium, Cx43/ZO-1 expression and localization resulted in functional changes in gap junctional communication (Fig. 8A, cartoon). Briefly, a monolayer of astrocytes was scratched in the presence of two dyes, Lucifer yellow (L.Y.), which crosses G.J., and Dextran, which does not cross G.J. due to its size. L.Y., but not dextran, is taken up by the damaged cells and diffused via gap junctions into the intact neighboring cells [48].

Fig. 8.

Soluble sulfatide increased gap junctional communication in human primary astrocytes. Human primary astrocytes were treated with 10 μg/ml of sulfatide for 6 and 12 h, and then the scrape loading assay was performed to evaluate gap junctional communication. A Graphical presentation of the scrape loading assay. B, E Representative fluorescent image for Lucifer Yellow (L.Y.) in untreated (control) and sulfatide-treated conditions. The white line represents the scratch site. Bar: 250 μm. C–G Representative fluorescent images for L.Y. in control and sulfatide-treated conditions in the presence of two well know gap junction blockers, 18α‐glycyrrhetinic acid (AGA) or Octanol (OCT). The white line represents the scratch site. (H) Quantification of the diffused distance of L.Y. from the injury site in μm. Normalization was performed using dextran-Alexa Fluor 594 (dye not permeable to G.J.). Treatment of human primary astrocytes with sulfatide (gray wave) resulted in increased L.Y. intensity compared to untreated (control, beige wave) conditions. Multiple t tests analyzed L.Y. intensity vs. diffusion data points with the Holm − Sidak method (alpha = 0.05) for statistical significance (n = 7). L.Y. diffusion was G.J. dependent as preincubation of untreated (control, beige curve) astrocytes with AGA (green curve) or OCT (dark green curve) prevented the diffusion of L.Y., and the staining was captured primarily by the cells in the scrape area. Preincubation of Sulfatide (grey curve) astrocytes with AGA (pink curve) or OCT (orange curve) prevented the diffusion of L.Y., and the staining was retained primarily by the cells in the scrape area. Multiple t tests analyzed data points for L.Y. diffusion vs. intensity with the Holm − Sidak method (alpha = 0.05) for Control + AGA* p ≤ 0.05 and Sulfatide + AGA# p ≤ 0.05, n = 7 independent cultures from different individuals

In control untreated conditions, L.Y., but not dextran, diffused up to 1400 µm, indicating a strong gap junctional communication (Fig. 8B, H correspond to the quantification of L.Y. intensity). The preincubation of the cultures with 18α‐glycyrrhetinic acid (AGA, 50 µM) or Octanol (OCT, 500 µM), two well-known G.J. blockers, prevented the diffusion of L.Y. (Fig. 8C, D, *p ≤ 0.05, n = 3). Sulfatide treatment increased the L.Y. diffusion distance but with a higher L.Y. intensity than control conditions (Fig. 8E, G, quantification of L.Y. intensity, *p ≤ 0.05, n = 3). Pre-application of AGA or OCT in sulfatide-treated cultures prevented the L.Y. diffusion and increased L.Y. intensity in response to the treatment reducing them to 500 and 600 µm, respectively (Fig. 8F, G, H for the quantification, #p ≤ 0.05, n = 3). Overall, our data indicate that Sulfatide increased gap junctional communication in astrocytes, even in the absence of HIV components, supporting the hypothesis that sulfatide release in HIV-infected tissues due to the presence of viral reservoirs not only corresponds to myelin compromise but also has a profound increase in cell to cell communication that can further contribute to CNS dysfunction.

Discussion

Our data demonstrated that viral reservoirs are associated with chronic HIV-mediated CNS damage even in the current ART era, especially in myelin-rich areas compromising its structure and integrity. Viral reservoirs increase sulfatide levels that compromise gap junctional communication and associated calcium signaling, both essential components of proper brain function. Thus, we propose that increased sulfatide levels are a key mediator of CNS damage in the HIV-infected population and could be a potential biomarker of early or chronic stages of HIV CNS dysfunction.

Using human brain tissues, we identified that specific sulfatides (24:0, 24:1 and their hydroxylated species) were upregulated in HIV-infected individuals with MND, HAD, and Alzheimer’s cases. Long-chain sulfatides 24:0 and 24:1 are the major species present in mature myelin. Shorter chain sulfatides with 16 to 22-carbon nonhydroxylated and hydroxylated fatty acids play a crucial role in oligodendrocytes maturation; thus, they modulate their expression in the premature myelin stages [40]. Our findings showed significant changes for the long-chain 24:0 and 24:1 sulfatides (hydroxy fatty acid and nonhydroxy fatty acid species) in W.M. of the HIV-infected population with mild and severe cognitive impairment. These sulfatide species are important for maintaining and functioning myelin sheath structures [40]; therefore, their altered metabolism correlates with a direct compromise of mature myelin structure observed for the HIV-MND and HAD conditions analyzed and for A.D. cases. Interestingly, sulfatide dysregulation observed within the W.M. of HIV-MND and HAD-affected individuals did not correlate with PA/PE expression and distribution, indicating a lack of uniform cell damage/destruction within these tissues. Myelin structural protein compromise observed in the immediate vicinity of HIV reservoirs indicates a specific mechanism of dysregulation induced by latently infected cells. Thus, we propose that damage could be spread from viral reservoirs and compromise myelin integrity even without systemic viral replication. The concentric mechanisms of damage associated with viral reservoirs indicate that tissue reservoirs can generate chronic damage in the absence of viral replication, and novel mechanisms of amplification of damage need to be considered to reduce the devastating consequences of NeuroHIV.

Further, using mild/severe A.D. cases as a positive control of brain damage provided several important points in the analysis of brains obtained from HIV-infected individuals. First, some of the damage can reach myelin compromise severity observed in A.D. cases; Second, HIV damage can be expanded from viral reservoirs into neighboring uninfected areas of the brain, suggesting a unique amplification mechanism independent of systemic viral replication. Third, the HIV-mediated damage and sulfatide dysregulation can provide a unique mechanism of myelin compromise that may contribute to the motor abnormalities observed in the HIV-infected population.

Our data demonstrate that viral reservoirs are associated with significant bystander damage to large brain areas. As described by several groups, smaller cortical volumes, thinner cortical thickness, and poorer cognitive function are associated with the appearance of W.M. hyperdensities [10, 49, 50]. However, the nature of these hyperdensities is unknown. Neurobiological changes have been reported to occur soon after initial HIV infection and worsen in the absence of ART, resulting in early signs of cognitive compromise [51–53]. Later, during chronic infection with ART, the total numbers of individuals with the cognitive disorder remain equal to the early infection events. However, the cases of HAD have decreased in severity into more mild forms of cognitive impairment, suggesting that damage, even in the absence of active viral replication, remains a major public health issue. In these cases, cognitive impairment is independent of systemic viral replication and immune reconstitution [54]. Our data also indicate that myelin compromise was not associated with age, gender, systemic HIV replication (plasma and CSF), CD4 counts, ART regimen, and reported years with HIV, suggesting that ongoing brain damage in the HIV-infected population is dependent on viral reservoirs and unknown mechanisms of toxicity.

A potential mechanism of toxicity in the HIV-infected population due to the upregulation of sulfatide and mild myelin compromise could be the immune detection of these self-antigens. Research in neurodegenerative disorders, including HIV and M.S., supports this idea, indicating that the immune system can recognize myelin components as foreign antigens and establish a strong immune response. We demonstrated that low levels of soluble MBP induce microglia/macrophage activation resulting in cytokine and chemokine release [55]. These data indicate that mild myelin damage also contributes to damage amplification by immune activation. In addition, sulfatide in immune cells helps lymphocyte homing and increases monocyte differentiation and adhesion [56–58]. Also, natural killer cells, innate-like T cells, could recognize lipid antigens, including sulfatide, by a CD1-dependent mechanism to elicit a strong immune response [59–61]. CD1 sulfatide presentation to type II natural killer cells and T cells regulates the Th1, Th2, and Th17 immune responses, and they could participate in autoimmune diseases [62], indicating a potential autoimmune role for the free sulfatide increase we observed to be induced by HIV. Long-chain sulfatides are more efficient in natural killer cell activation, and the generation of soluble sulfatide has been described in autoimmune hepatitis and experimental autoimmune encephalomyelitis involving an L-selectin-dependent mechanism [63, 64]. Soluble sulfatide also has coagulation properties by binding fibrinogen, laminins, thrombospondin, and selectin in platelets [65–67]. Also, sulfatide has been proposed to be a biomarker of carcinoma [38, 68], viral replication [69–71], and tuberculosis [72], suggesting that sulfatide plays a key role in several pathogens’ and genetic conditions. These conditions have been described in HIV-infected individuals and probably contribute to viral reservoir survival and chronic inflammation.

Although astrocytes are considered nonexcitable cells, their activation results from variations in cytosolic calcium and voltage-gated channels; normally, these calcium signals in astrocytes respond to the synaptic activity and regulate the release and uptake of gliotransmitters such as ATP and TNF-α, or D-serine, which further modulate neuronal synapses and BBB function by a lipid dependent mechanism [73]. The close association between calcium signaling in astrocytes is essential for pre-and postsynapse function by modulating neuronal synchronization and firing patterns [73]. Our findings sustain that sulfatide compromised the triggering of calcium waves and could have significant consequences on neuronal activity, BBB integrity and blood flow control. All these mechanisms require further examination, especially for myelin components, due to their importance in HIV and traumatic brain injury (TBI) and M.S.

Astrocytic calcium signals relay in several key components to generate, amplify and coordinate long-range signaling, including the pacemaker cells, IP₃R activation, and diffusion of calcium and IP₃ via G.J. IP₃R and G.J. Any alterations in these mechanisms have been associated with disease [74, 75]. Our data in HIV-infected astrocytes indicate that HIV infection and latency decreased IP₃ degradation in infected cells to enable diffusion by G.J. into surrounding uninfected cells. In HIV-infected astrocytes, the high levels of IP₃ do not result in calcium overload because nef binds to the IP₃R1, preventing its proper activation and subsequent apoptosis. However, in surrounding uninfected cells communicated by G.J., nef is not present, and diffused IP₃ can trigger a calcium response. [26, 41, 42]. Cx43 stability and G.J. maintenance in HIV-infected astrocytes depended on residual HIV-tat protein expression in glial reservoirs [41, 44, 76]. Our data in vivo and in vitro indicate that latently infected astrocytes use G.J. and IP₃ to amplify apoptosis and toxicity into neighboring uninfected cells (endothelial, neurons, and astrocytes) [26, 42, 44, 77]. HIV-latently infected cells were resistant to apoptosis and toxicity by a mechanism that prevents the proper IP₃ and IP₃R signaling on the endoplasmic reticulum membrane and loss of interactions with the mitochondria [42]. Normally, all these components work together to signal and prevent toxicity. However, how important could astrocyte infection and dysregulated signaling be in the HIV-infected population? A single astrocyte can communicate or enwrap hundreds of synapses [78–80], but also astrocytes/astrocyte endfeet and their calcium waves are essential to coordinate blood flow, BBB permeability, and the gliovascular unit to provide and exclude material to neurons/glia cells [81–84]. Thus, the potential implications of HIV, sulfatide, and calcium wave dysregulation are significant. HIV-infected cells could amplify toxicity to neighboring uninfected cells, as observed in our data in myelinic areas. Here, we propose that local myelin compromise further contributes to local inflammation and neuronal dysfunction associated with the presence and altered signaling provided by viral reservoirs.

In addition, our finding that free sulfatide levels are increased within the W.M. of neuro-HIV-affected individuals and that free sulfatide can promote GJ-controlled cell-to-cell communication identifies a potential host-directed therapeutic target to control HIV-induced bystander damage. Indeed, therapeutic modulation of specific lipid-related pathways has successfully prevented or reduced inflammatory diseases. Now, we are proposing a similar approach to chronic HIV. Our data that CNS damage is associated with viral reservoirs provides a unique mechanism of bystander damage associated with sulfatide and gap junctional communication. Currently, the best-described viral reservoirs are resting CD4⁺ T lymphocytes, but in the brain, the main reservoirs are microglia/macrophages and a small astrocyte population [85, 86]. However, the brain is “insulated” due to the presence of the BBB, and it has been proposed that the virus sequences in the brain had different evolution than other tissues [87–89]. In agreement, it has been shown that rebounding HIV is genetically different from circulating viral reservoirs suggesting that viruses come from other tissues and cell types, including the brain [88, 90–92]. Thus, examining the different kinds of viral reservoirs and the mechanisms of bystander damage is essential to understanding and treating the cognitive decline observed in at least 50% of the HIV-infected individuals under ART.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Study concepts and design were undertaken by BP, and EAE Data acquisition and analysis were undertaken for DD BP and EAE Manuscript suggestions and writing was provided for all authors.

Funding

This work was funded by The National Institute of Mental Health grant, MH128082, the National Institute of Neurological Disorders and Stroke, NS105584, from the UTMB Sealy Institute for Vaccine Sciences and the UTMB Institute for Human Infection &Immunity (to EAE).

Data availability

The data sets generated are all available upon reasonable request.

Declarations

Conflict of interest

The authors declare no conflicts of interest.

Ethical approval

This study was performed according to the principles of the Declaration of Helsinki, and all studies were reviewed and approved by the host institutions.