INTACT HIV PROVIRUSES PERSIST IN THE BRAIN DESPITE VIRAL SUPPRESSION WITH ART

Abstract

Objective:

Human Immunodeficiency Virus (HIV) persistence in blood and tissue reservoirs including the brain is a major barrier to HIV cure and possible cause of comorbid disease. However, the size and replication competent nature of the CNS reservoir is unclear. Here we used the intact proviral DNA assay (IPDA) to provide the first quantitative assessment of the intact and defective HIV reservoir in the brain of people with HIV (PWH).

Methods:

Total, intact and defective HIV proviruses were measured in autopsy frontal lobe tissue from viremic (n=18) or virologically suppressed (n=12) PWH. Total or intact/defective proviruses were measured by detection of HIV pol or the IPDA, respectively, through use of droplet digital PCR (ddPCR). HIV-seronegative individuals were included as controls (n=6).

Results:

Total HIV DNA was present at similar levels in brain tissues from untreated viremic and antiretroviral (ART)-suppressed individuals (median: 22.3 vs 26.2 HIV pol copies/10⁶ cells), reflecting a stable CNS reservoir of HIV that persists despite therapy. Furthermore, 8/10 viremic and 6/9 virally suppressed PWH also harbored intact proviruses in the CNS (4.63 vs 12.7 intact copies/10⁶ cells). Viral reservoirs in CNS and matched lymphoid tissue were similar in the composition of intact and/or defective proviruses, albeit at lower levels in the brain. Importantly, CNS resident CD68+ myeloid cells in virally suppressed individuals harbored HIV DNA, directly showing the presence of a CNS resident HIV reservoir.

Interpretation:

Our results demonstrate the first evidence for an intact, potentially replication competent HIV reservoir in the CNS of virally suppressed PWH.

INTRODUCTION

The major barrier to Human Immunodeficiency Virus (HIV) cure is the persistence of latent virus in reservoirs, which is unable to be eradicated by antiretroviral therapies (ART). Tissue reservoirs of HIV are difficult to access (both therapeutically and physically) and potentially exhibit different viral profiles and dynamics compared to blood CD4+ T cell reservoirs ¹. Therefore, characterizing HIV persistence in tissue sites including the brain is critical to HIV treatment and cure approaches.

HIV infects the central nervous system (CNS) during acute infection, likely via infiltration of infected monocytes and/or T cells across the blood brain barrier ², which can spread to CNS resident myeloid cells (perivascular macrophages and microglia) ^3–5, astrocytes ^{6, 7} and pericytes ⁸, thus, establishing a viral reservoir. Whilst early infection events are relatively well-defined, the location, frequency and proviral landscape of the CNS reservoir in ART-suppressed people with HIV (PWH) is unclear. This is problematic as HIV persistence in the CNS may act as a source, and site of, ongoing replication. Furthermore, CNS reservoirs of HIV may contribute to the pathogenesis of HIV-associated neurocognitive disorders which affect ~30% of ART-suppressed PWH ^9–11. Previous qPCR-based studies of humans and animal models have detected viral DNA or RNA in the CNS of ART-suppressed individuals/animals ^{1, 12, 13}, supporting a stable viral reservoir in the brain. However, these studies do not define the proportion/magnitude of proviruses that are ‘intact’ and potentially replication competent, which is <2–11.7% of all proviruses in CD4+ T cells (~88.3 – 98% of total proviruses) ^14–16. Cerebrospinal fluid (CSF) escape studies ¹⁷, and in vivo non-human primate (NHP) latency studies ¹⁸ suggest that the brain, particularly resident myeloid cells, may harbor replication competent genomes. However, the size of the intact proviral reservoir in the brain of ART-suppressed PWH is unclear.

Typical qPCR-based assays for HIV DNA cannot distinguish intact from defective proviruses, thus overestimating the size of the intact, replication competent reservoir. The intact proviral DNA assay (IPDA) is a novel multiplex droplet digital PCR (ddPCR) approach to distinguish and separately quantify intact versus defective HIV proviruses. The assay defines proviruses by a lack of fatal defects such as large deletions in the 5’ region of the packaging signal (ψ) of the provirus as well as APOBEC3G-mediated hypermutations or deletions in the Rev response element (RRE) of HIV envelope (env) ¹⁶. The IPDA has been validated as an accurate measure of intact, replication competent HIV genomes in blood ^{16, 19} and lymph node ²⁰. Furthermore, intact proviruses detected by IPDA correlate with total integrated HIV ²¹ and with quantitative viral outgrowth assays (qVOA) ^{16, 19}, supporting the use of this technology as a reliable surrogate measure of intact, potentially replication competent HIV proviruses in tissues where it is difficult to perform qVOA. Here, we utilized the IPDA to quantify intact and defective proviruses in brain tissue from PWH to characterize the proviral landscape in the CNS.

METHODS

Tissue cohort

Fresh frozen and formalin-fixed paraffin embedded (FFPE) human autopsy brain tissue (frontal lobe deep white matter) and matched lymphoid tissue (spleen, lymph node or gut-associated lymphoid tissue) from PWH or HIV-seronegative individuals were provided by the National NeuroAIDS Tissue Consortium (NNTC, USA; https://nntc.org ²²). All tissues were acquired and processed with ethics approval (RMIT University, Australia; HREC#20843).

Genomic DNA extractions

gDNA/RNA from fresh frozen brain and matched lymphoid tissues (~10 mg pieces) were extracted using the AllPrep DNA/RNA/miRNA universal kit (QIAGEN; Hilden, Germany), where isolated gDNA was used for HIV pol DNA quantification and RNA for T cell receptor (TCR) quantification, or the DNA extraction kit (Agilent; Santa Clara, CA) for the IPDA, as per manufacturer’s instructions. DNA shearing was minimized during DNA extraction for the IPDA (Agilent) by excluding mechanical homogenization (wide-bore pipetting used instead).

ddPCR

ddPCR (QX200 BioRad; Hercules, CA) was used for HIV pol DNA quantification and IPDA analyses, as previously described ^{16, 23}. Positive droplets per sample were defined by examining negative control wells (no template controls and gDNA from HIV-seronegative biological controls) for false-positive HIV droplets above the set threshold to determine a false-positive rate.

HIV pol DNA quantification

HIV pol and Ribonuclease P/MRP Subunit P30 (RPP30) for total input gDNA quantification was quantified, as previously described ²⁴ (Table S1). A median (IQR) of 482,026 (452,543 – 571,480) cells were screened per sample.

Intact Proviral DNA Assay (IPDA)

Intact (Ψ⁺ and env⁺) and defective (Ψ⁺ or env⁺) HIV genomes were measured by the IPDA using high molecular weight gDNA from homogenized fresh frozen brain and matched lymphoid tissues, as previously described ¹⁶. NL4–3 control plasmids were generated by restriction enzyme excision to remove Ψ and/or env regions, where 5’ overhangs were filled-in with DNA Polymerase I, Large (Klenow) Fragment (NEB) and blunt ends ligated with Blunt/TA ligase master mix (NEB). Additionally, a commercially-synthesized gBlock fragment of a hypermutated env sequence was used for env probe validation experiments (Integrated DNA Technologies; Coralville, IA). Intact HIV copies in PWH were corrected for DNA shearing by calculating the fraction of gDNA that has been sheared between the RPP30 amplicons. Prior to droplet formation, high molecular weight gDNA was digested with the restriction enzyme Xhol (NEB), due to the absence of recognition sites within and between genomic regions of interest. HIV and human RPP30 reactions were conducted independently and in parallel, where copies were normalized to total RPP30 detected, as a measure of input cell number. In each ddPCR reaction, a median of 320ng (300 – 350ng) (HIV) or a median of 40ng (37.5 – 43.75ng) (RPP30) of Xhol digested gDNA was combined with ddPCR mastermix. Ten technical replicates were performed for HIV detection and two technical replicates for RPP30 quantification for each patient sample, where a median (IQR) of 549,576 (518,430 – 636,372) cells were assayed in total. For samples that exhibited low levels of env via IPDA, likely due to HIV polymorphisms within the RRE IPDA target region, a secondary env primer/probe set was used as a single-plex reaction ²³. Primers/probes listed in Table S1.

Amplification of HIV V3 env from CD68+ cells isolated by laser capture microdissection

The nuclei of CD68+ myeloid cells (>50 μm from observed vessels) were isolated from FFPE frontal lobe tissue, as previously described ³. >500 cells/sample were isolated per sample and DNA purified (FFPE DNA extraction kit, QIAGEN, Germany) according to the manufacturer’s protocol. HIV V3 env sequences were amplified from DNA prepared from the nuclei of CD68+ cell populations, as previously described ⁶. 5–10 cell equivalents/μL were used in multiples of 8 independent PCR reactions (V3 primers; Table S1). V3 env was purified from PCR positive reactions (<2+ per 8 PCR reactions performed) and Sanger sequenced. CLC Main Workbench v20 was used to align sequences.

DNAscope detection of HIV vDNA+CD68+ myeloid cells

HIV DNA+ myeloid cells in FFPE fixed human frontal cortex tissue sections (10 μm) were identified using DNAscope (V-HIV1-CladeB-sense) and CD68 antibody (KP-1, DAKO, Glostrup, Denmark; 1:400, as previously described ²⁵, with the following modifications. Nuclei was labelled with Hoechst (Invitrogen, Waltham, MA; 1:750 in TBS; 15 minutes), and autofluorescence quenched with TrueBlack Lipofuscin Quencher (Biotium, Fremont, CA; 1:50 in 70% EtOH; 30 seconds). Slides were cover slipped and imaged using the PALM microbeam (Zeiss, Oberkochen, Germany).

Quantification of CD3+ T cells in brain tissue

CD3+ T cells were identified in FFPE frontal cortex tissue section (6 μm) by chromogenic immunohistochemistry analysis using anti-CD3 (rabbit polyclonal, 1:100; DAKO), as previously described ^{3, 6}. Whole tissue samples were imaged (VS120 slide scanner Olympus, Shinjuku, Japan) and CD3+ T cells (relative to total nuclei) were quantified using HALO imaging software (Indica Labs v3.3).

Quantification of TCR in brain tissue

Relative expression of TCRbC mRNA in brain tissue was quantified by qPCR. cDNA was synthesized (SuperScript® III First-Strand Synthesis System; Thermo Fisher) and quantified by SYBR-based qPCR (Quantinova, QIAGEN) using a QuantStudio 5 real-time PCR machine (Thermo Fisher). Cycle settings: 95°C for 2 mins, 95°C for 5 sec, 60°C for 18 sec. Relative gene expression was standardized using the ΔΔCt method to 60S ribosomal protein L13 (RPL13). Primers listed in Table S1.

Statistics

Comparisons made using non-parametric Mann-Whitney U (unpaired) and Wilcoxon tests (paired). Pearson analyses used for correlative analyses. All analysis was performed using GraphPad Prism (version 9.2.0, GraphPad Software, La Jolla California USA).

RESULTS

HIV DNA persists in the CNS of PWH despite viral suppression with ART

To first quantify and characterize the presence of HIV in the CNS, ddPCR for HIV pol and the RPP30 gene to determine total cell input, was performed using DNA extracted from frontal lobe brain tissue. Fresh frozen brain tissue was obtained from HIV-seronegative individuals (n=6), viremic PWH (median plasma viral load: 17,500 HIV RNA copies/mL; median CSF viral load: 408 HIV RNA copies/mL; n=18) and virally suppressed PWH (VS PWH; all undetectable plasma and CSF HIV viral load; CD4 T cell count [median]: 355 cells/mm³; n=12) provided by the NNTC (Table 1). Viral suppression was defined as a minimum 2-year period of ART-controlled HIV viremia with undetectable plasma and CSF viral loads (a single blip of 250 HIV RNA copies/mL >6 months from death was allowed). The median time of viral suppression of VS PWH was 6.2 years (IQR: 2.57 – 7.04).

Table 1.

Clinical cohort characteristics

		HIV+
	HIV−	Viremic	Virally suppressed
N	6	18	12
Age	55 (39 – 74)	44 (35 – 62)	59 (44 – 67)
Male sex (%)	2 (33.3)	11 (61.1)	12 (100)
HIV parameters
Viral load (plasma)a	-	56,129 (15,090 – 207,526)	UD
Viral load (CSF)a	-	594.5 (340.3 – 5973)	UDb
CD4 T cells (mm3)		72.0 (5.50 – 185)c	355 (239 – 785)d
Nadir CD4 T cells (mm3)		8.00 (2.00 – 72.0)c	78.0 (54.0 – 161)d
Current ART (%)		0 (0%)	12 (100%)
Viral suppression (yrs)e		N/A	6.2 (2.6 – 9.1)

ART: antiretroviral therapy; CSF: cerebrospinal fluid; UD: undetectable

^aHIV RNA copies/mL

^bData missing for 4 donors

^cData missing for 13 donors

^dData missing for 1 donor

^eViral suppression defined by undetectable plasma viral load for > 2 years (1 blip <250 HIV RNA copies/mL > 6 months from death permitted)

HIV DNA was detected in brain tissue from all PWH, as measured by ddPCR for HIV pol (Fig 1A). Importantly, the levels of HIV DNA in brain tissue from VS PWH were similar to those from viremic PWH (median: 26.2 vs 22.3 HIV pol DNA copies/10⁶ cells, P>0.05), indicating that viral suppression of HIV in the blood does not influence HIV DNA levels in the CNS. HIV DNA levels in the CNS of viremic PWH correlated with CSF viral load (ρ=0.655, P=0.03; Fig 1B), but not with plasma HIV RNA levels (ρ=0.143, P=0.58; Fig 1C), suggesting that the CNS is a unique tissue compartment of HIV. Due to the definition of ‘viral suppression’ as an undetectable viral load (i.e. below the level of detection of clinical assays), no associations between HIV viremia in plasma and HIV DNA in brain tissue of VS PWH were made. Together, these findings predict the presence of a HIV reservoir in the CNS which is not associated with plasma HIV viral load levels or substantially influenced by long-term viral suppression in the blood by ART.

Fig 1. HIV DNA persists in the CNS of PWH despite long term suppression with ART.

(A) HIV pol DNA copy number in human frontal lobe brain tissue from viremic (n=18) or virally suppressed (n=12) people with HIV (PWH) or HIV-uninfected individuals (n=6) as measured by droplet digital PCR. (B) Correlative analysis of HIV pol DNA copy number in human brain tissue relative to HIV RNA copy number in cerebrospinal fluid or (C) plasma in untreated viremic PWH only (n=18). Median and interquartile ranges shown; comparisons made using Mann-Whitney U tests. Correlative analyses assessed by non-parametric Pearson’s correlations with Pearson’s rho and p values shown. P<0.05 considered statistically significant. Note: CSF viral load data unavailable for n=4.

CNS tissues from VS PWH harbor intact HIV proviral genomes

To date, identifying the presence of intact proviruses in the CNS of large cohorts of VS PWH has been difficult due to (i) the requirement of a large amount of high quality, high molecular weight DNA for advanced sequencing procedures, including full-length individual proviral sequencing (FLIPS) and/or (ii) the inability to perform qVOA analyses using autopsy material. To overcome these obstacles we optimized the IPDA to quantify both intact and defective HIV genomes in the CNS via multiplex ddPCR. The IPDA is a duplex ddPCR assay which targets HIV proviral DNA that is integrated in genomic DNA of host cells during viral infection. HIV integration is an essential step for HIV replication. Using two primer probe sets that target either the 5’ or 3’ end of the integrated HIV provirus, the simultaneous detection of both probe sets (separated by 7564bp base pairs) in a single lipid droplet demonstrates the presence of an ‘intact’ near full length HIV provirus integrated into host DNA which may be capable of viral replication. This assay differs from other single or dual target ddPCR assays that can only detect and quantify short regions of integrated HIV DNA (i.e. HIV gag, pol etc) as it simultaneously detects two regions of HIV that are commonly deleted or hypermutated in defective proviruses ^{12, 26}. The IPDA was validated using HIV NL4–3 plasmid controls that mimic intact or defective proviruses; generated via restriction enzyme digestion, as previously described (Fig 2A ¹⁶). As expected, only intact (ψ⁺env⁺), 3’ defective (ψ⁺ only) or 5’ defective proviruses (env⁺) were present in respective controls, validating primer specificity in our hands (representative 2D amplitude ddPCR plots shown; Fig 2B). To remove the possibility of hypermutated sequences in HIV env being amplified and mistakenly quantified by the IPDA, amplification of env involved a VIC-labelled probe that bound to intact/non-hypermutated sequences, and an unlabeled competitor probe that favorably binds to hypermutated sequences in the RRE, as previously described (¹⁶; Fig 2C). ddPCR was performed using varying mixed ratios of both full-length HIV NL4–3 and a commercially-synthesized gBlock fragment of a hypermutated env sequence. HIV env copies correlated with NL4–3 input, but not levels of hypermutated gBlock fragment DNA (Fig 2C), confirming the specificity and sensitivity of these primers to effectively discriminate and quantify non-hypermutated HIV env. To further confirm the sensitivity of the assay, the IPDA was performed on gDNA from peripheral blood mononuclear cells of uninfected donors (n=3) spiked with a series of ACH-2 cell equivalents, which contains a single provirus per cell (Fig 2D). As expected, the HIV DNA copy number of HIV ψ (ρ=0.98, P=0.006) or env (ρ=0.99, P=0.004) measured by the IPDA strongly correlated with input HIV (based on ACH-2 cell input; Fig 2D), directly validating the sensitivity and specificity of the IPDA in our hands.

Fig 2. Validation of IPDA specificity.

(A) Primer binding map for HIV Ψ (blue), HIV env (green) and RPP30 (blue and green arrows; same spacing as the Ψ and env amplicons (7564 bp) to correct for DNA shearing in a separate multiplex ddPCR). XhoI (red arrow) binding site for gDNA digestion for the IPDA is shown. Maps of NL4–3 plasmid controls including restriction enzyme cut sites (black arrow): (i) full-length NL4–3, (ii) 5’ deletion, (iii) 3’ deletion or (iv) 5’ + 3’ deletion. (B) Representative IPDA 2D-amplitude plots using 1000 DNA copies of NL4–3 plasmid controls shown in A assayed on a background of PBMC gDNA from HIV seronegative donor. Intact (HIV Ψ + and env+; orange), 5’ deleted (HIV env+ only; green), 3’ deleted (HIV Ψ+ only, blue) and 5’+3’ deleted NL4–3 plasmid controls (HIV ψ− and env−; grey). Bottom left quadrants also represents droplets with no provirus. (C) Map of env probes specific for the RRE region consisting of two adjacent G->A APOBEC3G sites (dashed boxes). An unlabelled hypermutated competitor probe preferentially binds to hypermutated proviruses, preventing the binding of the VIC-labelled intact probe, which only hybridizes to intact proviruses to release fluorescent signal away from 3’ minor-groove binder quencher (Q). Specificity of the env probes was validated using varying copies of HIV NL4–3 spiked with a synthetic gBlock fragment of a hypermutated env sequence (average env copies from n=3 experiments shown). (D) Correlative analysis of expected and IPDA-measured frequencies of ψ and env copies of HIV. gDNA from PBMCs of HIV seronegative donors was spiked with cell equivalent concentrations of ACH-2 cells and subjected to the IPDA. Non-parametric Pearson’s rho and significant p-values (P<0.05) shown. N=3 experiments performed.

The IPDA was performed on brain tissue from PWH where adequate CNS tissue (>4.5 × 10⁵ cells) was available (n=10 VS PWH; n=11 viremic PWH) and 6 HIV-seronegative individuals as biological controls. IPDA results were corrected for DNA shearing based on the frequency of RPP30 double positive droplets using a separate multiplex ddPCR for two regions of the RPP30 gene with similar genomic spacing to the Ψ and env amplicons (~7.5kb; Fig 2 ¹⁶). Minimal gDNA shearing was observed in all samples and within expected levels ¹⁶ (median droplet shearing index [IQR]: brain gDNA 0.17 [0.13 – 0.23] and lymphoid tissue gDNA: 0.24 [0.16 – 0.38]; Fig 3A). Intact proviral genomes were present in brain tissue from the majority of viremic PWH (8/10 individuals, median [IQR] copies/10⁶ cells: 4.63 [1.93 – 350.7]; Fig 3B and C), accounting for ~11.1% of all proviruses (Fig 3D). The major proportion of proviral DNA in brain tissue from viremic PWH was either 3’ defective (45.0 [21.2 – 448] copies/10⁶ cells, ~43.1% of total proviruses) or 5’ defective (36.9 [12.3 – 807] copies/10⁶ cells, ~45.8% of proviruses; Fig 3C and D). Importantly, intact proviral genomes were also present in CNS tissue from the majority of VS PWH (6/9 individuals, 12.7 [0 – 67.8] copies/10⁶ cells; ~15.5% of total proviruses; Fig 3C). Similar to findings in viremic PWH, the majority of proviruses detected in VS PWH were either 3’ (24.7 [7.74 – 107] copies/10⁶ cells; Fig 3C and D; ~37.2% of proviruses) or 5’ defective (59.4 [11.2 – 162] copies/10⁶ cells; Fig 3C and D; ~47.1% of proviruses).

Fig 3: Intact proviral genomes persist in the CNS despite long term viral suppression as measured by the intact proviral DNA assay (IPDA).

(A) The droplet shearing index (DSI) was determined in brain (n=27) or lymphoid (n=7) tissue by duplex ddPCR for two regions of the RPP30 gene and represents the fraction of gDNA that has been sheared between the amplicons. Median and interquartile range shown. (B) Representative 2D-amplitude ddPCR plot of proviruses in brain tissue from a virally suppressed PWH. (C) Copy number (standardized to 10⁶ cells) or (D) frequency of proviral genomes defined as intact (HIV ψ ⁺ and env⁺; orange), 3’ defective (ψ ⁺; blue) or 5’ defective (env⁺, green) from CNS frontal lobe tissue from viremic (n=10) or virally suppressed (VS; n=9) PWH or HIV-uninfected individuals (n=6) as measured by the IPDA. Median and interquartile ranges shown. (E) Correlative analysis of HIV pol DNA and intact, (F) 3’ defective or (G) 5’ defective proviruses in frontal lobe brain tissue from PWH. Non-parametric Pearson’s rho and p values shown. P values <0.05 considered statistically significant.

Notably, the IPDA was also performed for an additional viremic PWH and an additional VS PWH. As low HIV env signal was detected for these individuals, a phenomenon known to occur in a small number of cases due to polymorphisms within the env primer/probe region ²³, an additional round of ddPCR was performed using previously published alternate env primers (data not shown). This confirmed failure of env amplification and these samples were removed from analysis. HIV-seronegative controls (n=6) showed no indicators of intact, 3’ or 5’ defective proviruses, ruling out that levels detected were due to experimental contamination.

The levels of intact, 3’ and 5’ defective proviruses in the CNS all positively correlated with levels of HIV pol DNA (P<0.05 for all, Fig 3E–G), indicating that the level of intact and defective proviruses in the CNS is proportionally associated with reservoir size. Together, these findings provide the first evidence for an intact proviral HIV reservoir in the CNS of a subset of VS PWH.

CNS and lymphoid tissue reservoirs exhibit a similar frequency of intact and defective proviruses

To compare the relative size and nature of the intact and defective HIV reservoir in the CNS versus lymphoid tissue reservoirs, the IPDA was performed on lymphoid tissues of a subset of individuals where lymphoid tissue was available (n=7; 2 individuals were excluded from analysis due to inefficient env amplification as noted above). Intact proviral genomes were detected in lymphoid tissues of all individuals tested (5/5), with a trend to higher intact HIV copy number present in lymphoid tissue vs CNS tissue (median HIV intact copy number CNS vs. lymphoid tissue: 3.78 vs 110 copies/10⁶ cells, Fig 4). Similar to findings in the CNS (Fig 3C), the majority of proviruses in lymphoid tissues were either 3’ (233 [120 – 6757] copies/10⁶ cells) or 5’ defective (302 [154 – 1483] copies/10⁶ cells; Fig 4A). A tread to higher absolute levels of all forms of HIV proviruses in lymphoid tissues relative to matched CNS tissues was observed, however, this did not reach statistical significance. No difference was observed in the frequency of intact proviruses between the CNS and matched lymphoid tissues (Fig 4B), indicating that the composition of the proviral landscape in the CNS reflects that in peripheral lymphoid tissues.

Fig 4. CNS reservoir size relative to lymphoid reservoirs in people with HIV.

(A) Copy number (standardized to 10⁶ cells) or (B) frequency of proviral genomes measured by the IPDA from CNS frontal lobe tissue and matched lymphoid tissue (either spleen, lymph node or gut) from viremic (n=3, closed symbols) or virally-suppressed PWH (n=2, open symbols), where proviruses are defined as intact (ψ⁺ and env⁺; orange), 3’ defective (ψ ⁺; blue) or 5’ defective (env⁺, green). Median and interquartile ranges shown; comparisons made using paired Wilcoxon tests, P<0.05 statistically significant.

CNS resident myeloid cells harbour HIV DNA in VS PWH

To confirm cellular reservoirs of HIV in the brain, CNS cells containing HIV viral DNA (vDNA) were assessed in FFPE tissue from viremic and VS PWH by in situ hybridization DNAscope. HIV vDNA+ myeloid cells (CD68+) were present in frontal lobe tissue from both viremic and VS PWH (representative images shown in Fig 5). HIV vDNA+ CD68+ cells were >50 μm from blood vessels indicating infection of parenchymal cells, and not perivascular macrophages, directly demonstrating the presence of HIV-infected resident myeloid cells in the brain. No HIV DNA signal was detected in HIV-seronegative biological controls. As HIV-infected T cells passaging into the brain may also contribute, in part, to CNS reservoir measures by qPCR/ddPCR ²⁷, CD3+ T cells were quantified in matched FFPE brain tissues by immunohistochemistry. CD3+ T cells were observed in CNS tissue from both viremic and VS PWH at similar levels, contributing to ~0.5% of total cells in frontal lobe (median: 6,251 CD3+ T cells/10⁶ CNS cells in VS PWH). Importantly, the frequency and number of T cells in the CNS (Fig 6A; n=20), or gene expression levels of TCR RNA levels in matched frozen frontal lobe tissue of PWH (n=11) did not correlate with HIV pol measures in the CNS (Fig 6B; P>0.05 for all), suggesting that the level of T cells in the brain did not predict HIV DNA levels in the brain.

Fig 5. CNS resident myeloid cells in virally suppressed PWH harbour HIV DNA.

Representative images of HIV DNA+ CD68+ myeloid cells in frontal brain tissue from virally suppressed (VS HIV+; n=2) or viremic PWH (n=2) as determined by in situ hybridization for HIV DNA. HIV DNA (red), CD68 (green) or nuclei (blue) shown. Images acquired on a PALM Robo 4.2 at x 63 magnification. HIV DNA myeloid cells identified by white boxes. HIV DNA⁺DAPI⁺ and HIV DNA⁺CD68⁺DAPI⁺ insets of colocalization shown. Scale bars: 20 μm (x63 image) and 5 μm (x3.5 digital zoom - insets).

Fig 6. T cells infiltrating the CNS are not predictive of HIV DNA levels in the brain.

(A) Correlative analysis of frequency of infiltrating CD3+ T cells (as measured by immunohistochemistry) and HIV pol DNA levels (as determined by droplet digital PCR) in frontal brain tissue from virally suppressed (n=10) or viremic PWH (n=10; groups combined shown; n=20). (B) Correlative analysis of relative expression of T cell receptor (TCR) mRNA and HIV pol DNA levels from fresh frozen human brain of PWH (n=11). Parameters log transformed and spearman rho and P value shown. P<0.05 considered statistically significant.

To further demonstrate the presence of CD68+vDNA+ cells in the brain, LCM of the nuclei of parenchymal myeloid cells from viremic (n=2) and VS PWH (n=5) was performed. FFPE tissue was stained by immunohistochemistry for CD68, and the nuclei of >500 CD68+ cells that were >50 μm from blood vessels were collected. Nuclei from cells were isolated, DNA purified and triple nested PCR analysis was performed using primers specific for HIV V3 prior to sequence analysis. HIV DNA was detected in the nuclei of CD68+ cells isolated from both viremic (2/2 individuals) and VS PWH (4/5 individuals; sequences shown; Fig 7). HIV V3 sequences showed diversity between isolates, each of which were distinct from the laboratory strain NL4.3, indicating that sequences were true clinical isolates and not laboratory introduced contamination. Collectively, these findings directly demonstrate the presence of a resident CNS reservoir in the brain of VS PWH.

Figure 7.

HIV V3 protein sequences derived from CD68+ myeloid cells isolated from brain tissue of virally suppressed (VS) or viremic PWH. The nuclei of CD68+ myeloid cells (labelled by immunohistochemistry) were isolated from FFPE brain tissue from viremic (n=2) or VS PWH (vs; n=5) by laser capture microdissection prior to lysis, PCR and sequence analysis. Two sequences per patient shown. Changes in amino acids relative to HIV laboratory strain NL4.3 shown.

DISCUSSION

The size and proviral landscape of the HIV reservoir in the CNS is presently unclear. Here we utilized state-of-the-art molecular approaches coupled with a well-characterized cross-sectional tissue bank to provide the most comprehensive evidence for an intact proviral HIV reservoir in the CNS of PWH. HIV DNA was present at similar levels in brain tissue from both viremic and VS PWH, reflecting findings in NHP studies that the HIV reservoir in the brain is not dramatically affected by long-term viral suppression with ART ^{1, 28}. Importantly, intact proviral genomes were detected in brain tissue from both viremic and VS PWH, supporting the presence of an intact HIV reservoir in the brain. Finally, HIV DNA was identified in CD68+ myeloid cells in brain tissue, directly demonstrating the presence of a CNS resident reservoir in the CNS of VS PWH.

Our observation of intact proviral genomes in brain tissue, identified by the IPDA, strongly suggests the presence of a replication competent reservoir in the CNS. These data support the results of previous studies that have shown HIV RNA ^{29, 30} and anti-HIV antibodies in the CSF despite up to 10 years of suppressive ART ³¹, which has previously been the most convincing evidence for ongoing HIV replication in the CNS. Furthermore, cell-associated HIV DNA and/or RNA in the CSF has been associated with poorer neurocognitive outcomes, implying a role for ongoing viral replication/persistence in the CNS with the development of neurocognitive disorders ³². However, analysis of CSF has raised questions as to whether CSF viremia is directly related to ongoing CNS replication or possible infiltration from the periphery due to incomplete viral suppression due to drug resistance ^{33, 34}. Therefore, our findings within brain tissue, and more specifically, myeloid cells provide further evidence for a localized CNS reservoir in the brain. Intact proviral genomes were detected in the CNS of 6/9 VS PWH, whilst intact genomes were detected in peripheral lymphoid tissues of all individuals. One explanation is that individuals without detectable intact proviruses in the CNS may have harbored a smaller reservoir prior to ART initiation, which is supported by our findings that the number of intact proviruses directly correlated with HIV pol DNA levels (Fig 3). Alternatively, as others have shown the preferential decay of cells containing intact proviruses with time ^{35, 36}, it is possible that these individuals harbored a more effective immune response within the CNS. Together, our findings support the presence of an intact, potentially replication competent HIV reservoir in the CNS of a proportion of VS PWH.

Similar to studies in peripheral sites such as blood cells and lymph node ^{14, 16, 19, 20}, the majority of proviruses identified in the CNS were either 5’ or 3’ defective (~47 vs 37% of total proviruses, respectively) which may have important implications on CNS health. Whilst defective HIV proviruses are incapable of generating replication competent virions, some can still generate viral proteins including Nef and Tat ^{37, 38} which are neurotoxic and well known to induce CNS cellular activation, neuronal damage and are associated with adverse cognitive outcomes ^39–42. Many of these proteins have been identified in the CNS of virally suppressed human and NHPs ⁴³, supporting a role for defective proviruses in contributing to the generation of neurotoxic proteins. Additionally, studies in blood have shown that cells containing defective proviruses may be more prone to immune targeting by cytotoxic T cells ^{38, 44}, suggesting that similar immune activation phenomena may occur in the CNS. In our study, no inverse association between 5’ or 3’ defective proviruses and years of viral suppression was observed (data not shown), albeit limited by small patient numbers, suggesting that the presence of defective proviruses may not necessarily result in increased immune clearance of infected cells in the brain. Therefore, the persistence of high levels of defective HIV proviruses in the CNS, capable of generating neurotoxic proteins despite viral suppression, may contribute to ongoing immune activation in the CNS.

Our findings of HIV vDNA+ cells in the CNS by both in situ hybridization methods and PCR of HIV in isolated myeloid cells provides direct evidence for a CNS HIV reservoir in VS PWH. These findings confirm and extend previous studies by us and others who have identified HIV-infected myeloid cells ^{4, 5}, astrocytes ^{6, 7} and pericytes ⁸ in the CNS of both viremic and VS PWH. Of note, in vivo NHP studies have shown that HIV-infected T cells can also passage into the CNS which may also contribute to HIV DNA levels in the brain, particularly in the meninges during acute, untreated infection ⁴⁵. Whilst T cells undoubtably traffic into the brain and may, in part, contribute to HIV viral persistence in the CNS, we did not find any association between T cell infiltration and HIV pol DNA or IPDA measures. Therefore, our findings directly demonstrate the presence of a CNS resident HIV reservoir, and whilst infected T cells may contribute to HIV DNA levels in the CNS, T cells are not the sole reservoir in frontal white matter of VS PWH.

Levels of HIV pol DNA, intact or defective proviruses did not correlate with HIV plasma load in this study, supporting the CNS as a unique tissue compartment. It is noteworthy that plasma viral loads were determined by clinically available assays prior to death, so no single copy HIV RNA analyses were performed on either plasma or CSF. This may be of interest in future studies, as HIV RNA levels below the traditional limit of detection of clinical assays (i.e. 50 HIV RNA copies/mL plasma) could potentially contribute to reservoir size in blood in some cases ⁴⁶.

This study has some limitations which need to be considered. Whilst the IPDA is a validated surrogate measure of intact HIV proviruses and correlates well with qVOA outgrowth assays ^{16, 19}, we acknowledge that not all proviruses deemed ‘intact’ in this study are likely to be replication competent, as the IPDA primers do not assess the full genome. Future studies with techniques such as FLIPs may offer additional insight of the intact nature of the CNS reservoir. However, due to the low throughput nature of FLIPS, our use of the IPDA in this study allowed for the assessment of the near-intact reservoir in a large number patients which would not have been feasible by other approaches. As this study utilized an autopsy tissue bank, it is possible that DNA fragmentation may have occurred post-mortem leading to a higher proportion of proviruses deemed 5’ or 3’ defective than present in vivo. We acknowledge that it is possible that a gap in treatment may exist between last viral load test and death for some individuals. While treatment history was not provided at time of death, we found no association between the time between last viral load test (performed whilst subjects were alive) and autopsy and HIV pol DNA levels in the brain (P=0.203, ρ=−0.418). This study characterised the HIV reservoir in the frontal cortex as cognitive impairment in PWH is associated with this site. Future studies assessing the size and composition of HIV reservoirs in different regions within the CNS would be of interest.

In summary, we utilized a well-characterized cross-sectional bank of brain tissue to provide the first evidence for an intact proviral reservoir in the brains of PWH which importantly, appears to persist despite long-term viral suppression. These findings highlight the importance of the CNS as a known and substantial reservoir of HIV that must be considered in both HIV treatment and cure strategies.

Supplementary Material

Table S1

Table S1. PCR primers and probes