Defining HIV and SIV Reservoirs in Lymphoid Tissues

A primary obstacle to an HIV-1 cure is long-lived viral reservoirs, which must be eliminated or greatly reduced. Cure strategies have largely focused on monitoring changes in T cell reservoirs in peripheral blood (PB), even though the lymphoid tissues (LT) are primary sites for viral persistence. To track and discriminate viral reservoirs within tissue compartments we developed a specific and sensitive next-generation in situ hybridization approach to detect vRNA, including vRNA+ cells and viral particles (“RNAscope”), vDNA+ cells (“DNAscope”) and combined vRNA and vDNA with immunohistochemistry to detect and phenotype active and latently infected cells in the same tissue section. RNAscope is highly sensitive with greater speed of analysis compared to traditional in situ hybridization. The highly sensitive and specific DNAscope detected SIV/HIV vDNA+ cells, including duplexed detection of vDNA and vRNA or immunophenotypic markers in the same section. Analysis of LT samples from macaques prior to and during combination antiretroviral therapy demonstrated that B cell follicles are an important anatomical compartment for both latent and active viral persistence during treatment. These new tools should allow new insights into viral reservoir biology and evaluation of cure strategies.


INTRODUCTION
Due to the accessibility and ease of sampling peripheral blood (PB) in a clinical research setting and the prevailing thought that PB accurately mirrors systemic human immunodeficiency virus (HIV)/simian immunodeficiency virus (SIV) infection and dynamics, most studies of the size, decay kinetics, and features of viral reservoirs in HIV-infected individuals have relied on the longitudinal monitoring of plasma viral loads and infected cell subsets within PB mononuclear cells (PBMCs), including latently infected resting CD4+ T cell subsets [1]. However, since HIV/ SIV infections are primarily diseases of lymphoid tissues (e.g., lymph nodes, spleen, mucosal-associated lymphoid tissue [MALT]) in which the vast majority of HIV/SIV-infected cells and viral repositories reside [2][3][4][5][6][7][8][9][10][11], the assumption that the PB accurately reflects what goes on within these tissues is largely conjecture. We therefore continue to evaluate the lymphatic organ system itself largely by in situ technologies [9] as an essential component of a comprehensive assessment of viral reservoirs and persistence in vivo and to fully recognize the therapeutic potential of HIV-1 curative strategies.
The assessment of HIV/SIV viral reservoir size and phenotype has primarily been performed by approaches that require disruption of the tissue and/or preparation of single cell suspensions so that quantitative measurements can be performed [12,13]. In addition to losing important spatial information, the processing of cells from whole tissues may result in: i) misinterpretation of cell phenotypes (i.e. cell surface marker expression), ii) changes in viral expression patterns, iii) limited recovery of certain tissue resident cells, and iv) loss of cells due to processing induced death. Thus, while these approaches will continue to provide valuable information, we and others [14] strive to develop and use novel in situ technologies to visualize and quantify HIV-1 and SIV infections in anatomically intact native tissue environments to understand the types of cells and anatomic structures in which the virus is produced and how it is stored in follicles and persists in latently or covertly infected cells [3,5,[8][9][10][11][15][16][17][18].
While these classic in situ technologies remain to be valuable in characterizing SIV and HIV-1 infection and persistence in lymphoid tissues (LT), there is ample room for improvement in approaches that are less labor intensive, simpler, and faster than current in situ hybridization (ISH) methods with radiolabeled probes or chromogenic detection; more facile and reproducible than in situ Polymerase chain reaction (PCR) approaches in routinely detecting vDNA+ cells in formalin-fixed paraffin embedded (FFPE) tissues, a prerequisite for detection of latently infected cells in situ [3,14] ; and approaches to simultaneously detect vRNA and vDNA in the same tissue section as a valuable tool to identify covertly infected transcriptionally inactive vDNA+ / vRNAcells in tissues. We show here: 1) that an optimized next-generation ISH platform (termed RNAscope [19,20]) for the rapid detection of vRNA (with results obtained within 1 day) has sufficient sensitivity to reliably detect single virions in B cell follicles (BCF) in FFPE tissue sections, 2) that an approach for the detection of vDNA in situ (referred to as DNAscope) reliably and readily detects vDNA+ cells, and 3) that we have developed an in situ method to simultaneously visualize vRNA and vDNA in the same tissue section and thereby identify transcriptionally latent infections (vDNA+/vRNA-cells) in LTs. These new, highly sensitive in situ hybridization approaches applied to LT samples from macaques prior to and during combination antiretroviral therapy (cART) document the importance of BCFs in active, latent, and persistent infections during treatment. These data underscore the utility of new and sensitive ISH tools that can provide additional insight into viral persistence, reservoir establishment, tissue compartmentalization, and reservoir phenotype in the local in vivo tissue environment.

Sensitive detection of SIV vRNA+ cells and viral particles by next-generation RNAscope ISH
ISH methods offer the potential to provide key information about the status of both residual latent infection and active viral replication in the setting of cART, including anatomic information not readily obtained by other methods of virological analysis. To better understand, track and discriminate viral reservoirs within tissue compartments, we first compared the sensitivity and specificity of two current ISH approaches (radiolabeled [R-ISH]-the current "gold standard"-and chromogenic [C-ISH]) to a new next-generation technology (RNAscope), evaluating their utility for detecting vRNA+ cells and vRNA in virions in the follicular dendritic cell network (FDCn) in tissues from SIV-infected rhesus macaques (RMs) ( Figure 1A). Productively infected vRNA+ cells typically have a densely staining spherical signal that encompasses the entire cell body, while virions trapped on the FDCn typically have a diffuse lattice-like signal pattern within B cell follicles (BCF) (Figure 1B, C) [5]. Quantitative image analysis of lymph node sections from 6 SIV-infected RMs at 6-weeks post-infection (wpi) demonstrated no significant difference in the detection of productively infected vRNA+ cells in viremic macaques between R-ISH and C-ISH techniques using standard conditions ( Figure 1D). The next-generation RNAscope approach was as sensitive in detecting productively infected vRNA+ cells compared to both R-ISH and C-ISH techniques, with a trend to greater sensitivity that while not statistically significant (P = 0.094) ( Figure 1D) is consistent with identifying vRNA+ cells with low signal as a result of the more aggressive retrieval process and remarkably low background in the RNAscope protocol. Most important, this next-generation ISH technology was significantly more rapid than either the C-ISH or the gold standard R-ISH approaches, with an assay time of only 8 to 9 hours, compared to 3 days for C-ISH and 6 to 21 days for R-ISH (depending on the length of exposure).
The RNAscope ISH approach, like R-ISH and C-ISH, also detected vRNA in virions on the FDCn within the BCF of chronically SIV-infected RMs; however, RNAscope provided superior visual discrimination of putative viral particles compared to other ISH approaches ( Figure 1C). Pretreatment of tissue sections with RNase completely ablated vRNA detection of both productively infected cells and punctate signal from putative virions within BCF (Figure 2A). The relationship between the estimated number of virions within subjacent follicles of chronic SIV+ RMs derived by R-ISH and RNAscope approaches was linear and highly correlated (Supplemental Figure 1), demonstrating RNAscope as a facile and highly sensitive ISH approach for the detection of vRNA within the FDCn in vivo.    Increased sensitivity to detect vRNA by the next-generation RNAscope approach compared to standard in situ hybridization (ISH) approaches. (A) Schematic to demonstrate key differences between standard radiolabeled ISH (R-ISH), chromogenic ISH (C-ISH) and the next-generation RNAscope approaches, with cartoon representation of a productively infected vRNA+ cell on the left and virion trapped on FDCs on the right. Comparison between the gold-standard R-ISH, C-ISH, and RNAscope ISH techniques at low (B) and high magnification (C) demonstrating the exquisite sensitivity with potential for detection of virions in tissues. Scale bars = 200 µm. (D) Quantitative image analysis of subjacent lymph node sections from 6 chronically infected SIV+ RMs demonstrating a trend for increased sensitivity of RNAscope for detecting vRNA+ cells. P values were based on the Wilcoxon-sign rank test corrected for multiple comparisons.
For quantitative validation of the RNAScope method in detecting viral particles, we documented detection of virions in a binding experiment in which SupT1 CCR5-Hi cells were incubated with serial dilutions of concentrated highly purified infectious SIV (sucrose double banded virus), followed by parallel analysis using both qRT-PCR and RNAscope ISH with quantitative image analysis, as well as by comparative analysis with R-ISH and C-ISH. Bound virions were visualized as an intense red signal on SupT1 CCR5-Hi cells with high concentrations of virus after RNAscope ISH, which decreased to a discrete individual punctate signal in proportion to the dilution of SIV added to SupT1 CCR5-Hi cells ( Figure 2B). While the gold standard R-ISH was similar to RNAscope in detecting virus bound to SupT1 CCR5-Hi cells down to a dilution of 10 5 , the RNAscope method provided superior visual discrimination of viral particles bound to these cells, visualized as a single punctate signal compared to the diffuse silver grain signal surrounding cells using the R-ISH method. Both RNAscope and R-ISH were about 3-orders of magnitude more sensitive at detecting viral particles bound to SupT1 CCR5-Hi cells than the C-ISH method. The estimated number of virions bound per cell, as determined by RNAscope analysis, was highly correlated

D.
with vRNA levels determined by qRT-PCR results over several orders of magnitude ( Figure 2C). The robustness and specificity of detecting vRNA+ cells and viral particles by RNAscope ISH was determined by quantifying the relative false detection rate in lymphoid tissues from SIV-negative RMs (Supplemental Table 1; Supplemental Figure 2). We analyzed 274 high power regions of interest (ROIs) consisting of nearly 70 mm 2 of tissue and found only 2 false positive "vRNA+ cells" and 4 false positive "virions. "   Utility of the next-generation RNAscope ISH in a cART SIV NHP model Having validated the sensitivity and specificity of next-generation RNAscope ISH, we applied this approach to tracking the virological response to cART in lymph node samples obtained from SIV+ RMs before and after 26 weeks of cART and compared these results to parallel measurements performed by qRT-PCR. Eight RMs were infected with SIVmac239 and started on cART at 4 wpi; lymph node biopsies were taken prior to cART initiation (4 wpi) and after 26 weeks of cART (30 wpi), at which time 6 of the 8 RMs had plasma viral loads below assay limits (< 30 SIV vRNA copies/ml) (Supplemental Figure 3). The number of productively SIV vRNA+ cells/10 5 cells measured by RNAscope ISH and vRNA copies/10 5 cell equivalents measured by qRT-PCR from the same lymph node biopsy (RNA extracted from bulk tissue) decayed in parallel and were highly correlated ( Figure 3A-C). RNAscope ISH provided additional anatomical information on the location of SIV vRNA+ cells and the proportional change associated with cART ( Figure 3D-F). Prior to cART (4 wpi), on average 69% of the SIV vRNA+ cells were located in the paracortical

C.
T cell zone (TCZ) with an average 27% located within BCF and only an average 3% were located within the medullary cords (MC) of the lymph nodes. The differences between the frequencies of vRNA+ cells within each anatomical site were highly statistically significant before cART ( Figure  3D-F). After cART the total number of vRNA+ cells decreased dramatically, with the relative proportion of the remaining vRNA+ cells decreased in the TCZ and the proportion increased in the BCF until there was no significant difference in the proportion of SIV vRNA+ cells within these two areas, with an average 46% of SIV vRNA+ cells located in each distinct anatomical environment ( Figure 3E-F). Although this change in the proportion of SIV vRNA+ cells within the TCZ and BCF after cART was statistically significant; there was no significant change in the proportion of vRNA+ cells within the MC ( Figure 3F). These results are consistent with persistent low-level viral expression during cART as has previously been reported in HIV-infected cART-suppressed patients [15] and highlight the BCF as a particularly important anatomic site for infected cells during treatment. Consistent with the importance of BCF in viral persistence, we also found that virions associated with the FDCn remained detectable after 26 weeks of cART, albeit at very low levels compared to pre-cART time points ( Figure 3G). While the decrease in virions measured after 26 weeks of cART within BCFs was on average 99.5%, nearly every follicle analyzed had infrequent but discernable virions within BCFs by RNAscope ISH ( Figure 3G), which is again consistent with residual and potentially infectious virus in the FDCn pool at this time as has been shown recently for HIV in well-suppressed patients on cART with undetectable virus in PB [21]. Detection of vDNA in situ using next-generation DNAscope ISH The detection of individual virions by the RNAscope ISH approach encouraged us to explore the potential to detect vDNA+ cells in fixed tissues with this technology, a task beyond the capability of current conventional ISH methods. To test this, we used the 3D8 cell line, derived from chronically-SIVmac316 infected CEM×174 cells with reported 1 copy of integrated SIV DNA [22] and the ACH-2 cell line, a latently infected T cell clone initially reported to contain 1 integrated proviral copy of HIV-1 per cell, derived from A3.01 cells that were infected with HIV-LAV and cloned by limiting dilution [23][24][25]. We modified the RNAscope protocol using sense probes to G.
detect SIV and HIV-1 vDNA in these cell lines by "DNAscope" (see Methods) and analyzed serial dilutions of 3D8 and ACH-2 cells in uninfected CEM cells. DNAscope ISH detected vDNA localized within the nuclei of 3D8 cells by both bright field and fluorescence confocal analysis ( Figure  4A-B). The frequency of cells scored as vDNA+ decreased as the proportion of serially diluted 3D8 infected cells present in a constant number of total cells (infected plus uninfected CEM cells) decreased (Supplemental Figure 4). In addition, the number of vDNA+ cells detected by DNAscope ISH was strongly correlated with the number of vDNA copies per cell equivalent detected by parallel qPCR analysis. (Figure 4C, D). Similar data were obtained with ACH-2 cells and HIV-1 DNAscope ISH (Supplemental Figure 5). Interestingly, while we usually found one punctate signal of vDNA within 3D8 cells, ACH-2 cells frequently contained more than one distinct vDNA signal in the nucleus by DNAscope suggesting that ACH-2 cells frequently contain multiple integration sites as previous reported [26] (Supplemental Figure 5). Assuming the average diameter of CEM cells to be 15 μm [27], and DNAscope measurements having been performed on 4 to 6 μm thick sections, we would expect to detect a vDNA signal in 21% to 29% of undiluted 3D8/ACH-2 cells, in agreement with measurements for both 3D8 and ACH-2 cells (Supplemental . We calculated the percent by computing the probability that the nucleus is contained in a cut section conditional on the cell being in the cut section. We calculated the percent by computing the probability that the nucleus is contained in a cut section conditional on the cell being in the cut section (i.e. the ratio of the probability of the 2 events). If we assume a cut cell pellet sample has thickness 2T and the cut sections are 2s thick, then if we assume that the nuclei of cells are uniformly distributed throughout the section we find that the probability that a cell nucleus is in the cut section is s/T. The probability that a cell intersects a section is the probability that a cell is within a cell diameter (denoted d) of the cut section, hence if we assume that cells are uniformly distributed throughout the cut-cell pellet section we find that this probability is (2s+d)/ (2T). Combining these results we get the expression 1/(1+d/(2s). So if the cell radius is 7.5 and the section thickness is 4 to 6 μm we get the percent range given above. A 95% confidence interval for the regression coefficient we obtain from regressing the counted values on the expected values is (0.15, 0.28), which is consistent with our calculations.  Utility of the next-generation DNAscope ISH for the detection of vDNA+ cells in a cART SIV NHP model Having validated the ability of DNAscope to detect vDNA in cell lines, we used the DNAscope ISH approach to detect vDNA in cells in lymph node tissue sections from acutely SIV infected RMs and compared these results to RNAscope detection of vRNA and SIVp17 protein detection by immunohistochemistry (IHC). RNAscope yielded robust detection of vRNA+ cells, virions (data not shown), and IHC, while less sensitive, was capable of detecting abundant SIVp17+ cells in acutely SIV-infected RMs ( Figure 5A-B). Furthermore, DNAscope ISH readily detected vDNA in lymph nodes from acutely SIV-infected RMs ( Figure 5A-B). Most important, while RNase pre-treatment completely ablated vRNA detection by RNAscope, both IHC detection of SIVp17+ cells and DNAscope ISH detection of vDNA+ cells were unaffected by RNase pre-treatment (Figure 5A), demonstrating that DNAscope ISH specifically detects vDNA in cells in tissue sections.

C.
DNAscope can also be combined with immunofluorescence detection for phenotypic analysis of the infected cell type. In acutely SIV-infected RMs the vast majority of vDNA+ cells in lymph nodes were T cells ( Figure 5C), and many infected T cells at the time of peak infection contained 2 or more distinct punctate vDNA signals within their nuclei. This finding is consistent with similar results in ACH-2 cells and with several reports of superinfection of cells through virological synapses mediated by cell-to-cell transmission [28,29] (Figure 5C-D). There were vDNA+ macrophages that appeared to be authentically infected ( Figure 5D), but the vDNA detected in other macrophages was associated with ingested infected T cells (Supplemental Figure 6), as has been previously described [30]. We quantified vDNA+ cells before and after cART in 8 RMs, as described in Figure 4, and compared these results to qPCR results from the same tissues. There was a strong correlation between the number of vDNA+ cells quantified by qPCR and DNAscope ( Figure 5E), with a decrease in the total number of vDNA+ cells after 26 weeks of cART. In addition, DNAscope revealed a proportional enrichment of the remaining vDNA+ cells within the BCF relative to other lymph node anatomic sites after suppressive cART, similar to the vRNA+ cell results, once again highlighting the BCF as an important site of viral persistence during cART ( Figure 5F). Consistent with vDNA+ cells being more stable during long-term cART than vRNA+ cells, we see an average of approximately 5-fold reduction in vDNA+ cells compared to about a 24-fold reduction in vRNA+ cells after 26 weeks of cART, but a strong association still exists between the number of vRNA+ and vDNA+ cells before and during cART (Supplemental Figure  7). We have generated, validated, and applied both the RNAscope and DNAscope approaches for the detection of HIV-1 vRNA and vDNA in clinically relevant patient samples infected with a multitude of HIV-1 clades (data not shown), and these HIV-1 lineage-specific reagents are now available to the scientific community.

E. F.
Duplexed RNAscope / DNAscope ISH for the combined detection of vRNA and vDNA Finally, we reasoned that the ability to unambiguously identify and quantify latently HIV/SIV infected cells that are transcriptionally silent in their native in vivo environment could provide needed insight into reservoir persistence and maintenance; thus, we developed an approach that combined RNAscope with DNAscope. This combination approach used sense probes (vDNA) targeting the 5ʹ gag-pol portion of the SIV/HIV genome and anti-sense probes (vRNA) targeting genes in the 3ʹ half of the genome (vif, vpx, vpr, tat, env and nef), as well as the TAR element in the 5ʹ genome (Supplemental Table 3). This approach was capable of distinguishing transcriptionally active cells (vRNA+, vDNA+) from transcriptionally inactive, potentially latent, infected cells or cells harboring transcriptionally incompetent proviruses (vRNA-, vDNA+) in the same tissue section (Figure 6). .

DISCUSSION
The persistence of HIV-1 reservoirs despite apparently effective cART remains a formidable obstacle to achieving sustained virologic remission in HIV-infected individuals in the absence of continued cART [1,[31][32][33]. While the resting CD4+ T-cell reservoir in PB has been the most intensely studied and characterized HIV-1 reservoir to date, it has long been known [34] that the LT are the principal site [2,3,5,[9][10][11] where virus is produced, that the virus is stored in the FDCn in quantities estimated to be 40-to 50-fold larger than in PB [5], and that it is harbored as transcriptionally inactive proviruses in CD4+ T cells with population sizes of the order of 10 10 cells in an untreated 70-kg individual [3,9]. It is thus not surprising given the large number of cells with potentially reactivatable proviruses, continuing replication in LT under conditions of relatively low levels of ARVs [15,35] and infectious virions in the FDCn pool, that recrudescent infection on treatment interruption is rapid and multifocal in the LT reservoir [16].
It also follows that experimental approaches to tracking this LT reservoir are essential in assessing strategies to achieve sustained remissions when cART is stopped. To that end, we describe here fast and facile in situ technologies to track productively infected cells, the FDCn pool and cells that harbor transcriptionally inactive proviruses before and during cART, and the changes effected by a reactivation strategy, e.g., increases in the proportion of vDNA+ and vRNA+ cells. While in situ PCR was developed to detect a single copy of lentiviral DNA in a cell [36] and has been used with ISH to reveal the large population of covertly infected vDNA+ vRNA-CD4+ T cells in LT in HIV-1 infected individuals [4] , the DNAscope and combined DNAscope/RNAscope methods with the ability to detect vDNA and vRNA in a single tissue section are significant advances in assessing the effectiveness of reactivation strategies in vivo. The remarkable specificity of this approach with little-to-no background "noise" is likely due to the requirement of a pair of target double Z probes (ZZ), each possessing a different type of tail sequence, being directly juxtaposed to each other forming a contiguous target region for subsequent hybridization. This specificity coupled with the exceptional sensitivity of this approach achieved by utilizing a combination of ZZ probes in a cocktail probe mixture (40-85 ZZ pairs with approximately 2 to 4.2 kilobase (kb) of genome coverage) and an amplification cascade via sequential hybridization similar to the branched DNA (bDNA) allows for the visualization and discrimination of individual virions or vDNA within the nuclei of infected cells.
There is a growing appreciation of the importance of the FDCn pool as a reservoir from which infection can be reignited on discontinuation of cART. Earlier studies identified the FDCs as cells that HIV [3] and SIV [37] do not infect, but instead binds these viruses via complement receptors on their prolific processes in a state where infectivity is maintained [21,38]. There is thus a large reservoir [5,39] of virus well positioned to maintain and reignite infection. Moreover, in the chronic stage of infection, the BCF is a preferential site for continued infection and immunologically privileged in the sense that generally virus-specific CD8+ T cells are not effectively recruited into BCFs [40][41][42][43].
For all these reasons, the rapid, sensitive, and quantitative assessment of virus in the BCFs on the FDCn by RNAscope represents a significant methodological advancement to evaluate immunologic therapies and otherwise target this reservoir. When applied in longitudinal analyses in SIV-infected macaques before and during cART, these novel next-generation ISH approaches documented that: 1) the massive viral particle repository trapped on the FDCn within the BCF is significantly diminished with cART, as previously shown in humans [5, 6, 10] but persists at low yet demonstrable levels in most follicles; 2) cART dramatically reduces the number of vRNA+ cells within the TCZ and BCFs, but persistently low levels of vRNA+ cells remain detectable in all compartments within secondary lymphoid tissues; and 3) within the decreased total viral nucleic acid positive cells remaining in the setting of cART, the proportion of vRNA+ and vDNA+ cells decreased in the TCZ, but increased in the BCFs. Collectively, these findings underscore the importance of the BCF as a site of persistent infection and reservoir within the LT compartment, one that must be targeted effectively by immune clearance and other strategies to eradicate infection following transmission or for a functional cure.
Combining RNAscope and DNAscope with immunofluorescent protein detection represents a new enabling technology to characterize the cellular sources of virus production and cellular sanctuaries of persistent infection. We confirm the predominance of infection in T cells during acute SIV infection, and expect that by using antibodies to phenotype the activation state we will find in the future the predominance of infection in ostensibly resting cells in transmission and acute systemic stages, as well as after cART [44,45]. While we documented visually ingested infected T cells by macrophages, these combined technologies not only distinguished phagocytosis from authentic infection but also enabled assessing potential macrophage reservoirs [46] within LT and other sanctuaries, such as the CNS. Collectively the in situ approaches we describe should provide anatomic and cellular information to complement qPCR/qRT-PCR approaches to thereby better understand the establishment and maintenance of viral reservoirs and to evaluate the in vivo impact of HIV-1 cure strategies.

Animals and tissue collection
All rhesus macaques (Macaca mulatta) were housed and cared for in accordance with the Association for the Assessment and Accreditation of Laboratory Animal Care (AAALAC) standards in AAALAC-accredited facilities. Tissues were utilized from 16 rhesus macaques of Indian origin treated according to protocols approved by the Institutional Animal Care and Use Committee (IACUC) of the National Cancer Institute. Our cohort consisted of the following animals: SIV-negative (n = 2), acute SIV+ (2 wpi; n = 4), chronic SIV+ (8-14 wpi; n = 2), and a longitudinal cohort with samples collected before cART (4 wpi) and after 26 weeks of cART begun at 4 wpi (30 wpi; n = 8). Prior to study inclusion, all animals were free of cercopithicine herpesvirus 1, SIV, simian type-D retrovirus, and simian T-lymphotropic virus type 1. Animals were either intravenously or intrarectally inoculated with SIVmac239 (1 ng p27; 4/6/07 stock; provided by Ronald Desrosiers) or SIVmac239X [47,48]. As described previously, the cART regimen employed included a daily subcutaneous injection of the co-formulated reverse transcriptase inhibitors (R)-9-(2-phosphonylmethoxyypropyl) adenine (PMPA; tenofovir; 20 mg/kg; Gilead Biosciences) and beta-2′, 3′ dideoxy-3′-thia-5-fluorocytidine (FTC; emtricitabine; 40 mg/kg; Gilead Biosciences), and the integrase inhibitors L-870812 ('812; 20 mg/kg; Merck) and L-900564 ('564; 10 mg/ kg; Merck), each given via the oral route twice daily, and the protease inhibitors darunavir (DRV; 600 mg; Janssen Therapeutics, purchased from the National Institutes of Health [NIH] pharmacy) and ritonavir (RTV; 100 mg; AbbVie, purchased from the NIH pharmacy), each also given via the oral route twice daily as previously described [48]. Lymph nodes were collected at necropsy and/ or surgical extraction before (4 wpi) and 26 weeks after cART (30 wpi) and immediately fixed in freshly prepared neutral buffered 4% paraformaldehyde (PFA) for 24 hours at room temperature. After fixation for 24 hours, fixative was replaced with 80% ethanol and tissues were paraffin embedded as previously described [48,49].
Plasma viral load, cell-associated viral RNA, and cell-associated viral DNA determination Plasma viral loads, cell-associated viral RNA, and cell-associated viral DNA determination were performed as previously described [47]. SIV DNA and RNA values were normalized to cell equivalents (CE), determined by co-amplification of a target sequence in the rhesus (or human) CCR5 gene.

Cell lines and SIV binding experiment
To calibrate our validation studies for detection of vDNA by DNAScope we used well-characterized cell lines containing a reported single integrated provirus, 3D8 T cells (for SIV) and ACH-2 cells (for HIV-1) [22][23][24][25]. In proportional reconstitution experiments, 3D8 or ACH-2 cells were mixed with uninfected CEM cells at different ratios (1:0, 4:1, 1:1, 1:4, 0:1) to obtain serial dilutions of cells containing viral genomes in a cell pellet of 10-20 x 10 6 total cells. Cells were washed in PBS and aliquoted for qPCR or, after fixation in 4% paraformaldehyde (PFA) at room temperature for 24 hours with mild aggregation on a slow speed rocker, for DNAScope analysis. Fixed cells were washed to remove PFA, which was replaced with 80% ethanol followed by gentle mixing. Cells were then washed to remove ethanol and suspended in 4 drops of liquefied Histogel biopsy gel (Fisher Scientific Catalog No.: NC9150318) pre-warmed to 50 o C. After mixing in the Histogel, cells were centrifuged at 2,000 x g for 5 minutes (using a Fisher Scientific accuSpin™ Micro 17 microcentrifuge) and placed at 4 o C until the Histogel solidified, at which point the gel was covered by ethanol 80% and the cell pellet was paraffin embedded. SupT1 CCR5 HI cell binding studies were performed by incubating 10-fold serial dilutions of well-characterized SIV CP-mac virus (stock = 1.12 x 10 -5 µg SIV capsid) with SupT1 CCR5 HI cells (5-10 x 10 6 ) for 2 hours on ice to allow virion binding but not internalization, followed by washing 3 times with cold PBS. Cells at each virus dilution were then aliquoted for qRT-PCR or fixation in 4% PFA at room temperature for 24 hours with mild aggregation on a slow speed rocker and processed for Histogel and paraffin embedding for RNAScope analysis, as described above.

HIV-1 and SIV RNA in situ hybridization
Radiolabeled and chromogenic HIV-1 and SIV in situ hybridization (R-ISH and C-ISH, respectively) were performed as previously described with some modifications [15,17]. In brief for R-ISH, after acetylation with acetic anhydride and dehydration, slides were hybridized at 45°C overnight with a 35 S-labeled riboprobe containing 0.5 mM aurintricarboxylic acid in the hybridization mix. After extensive washes and ribonuclease treatment, tissue sections were dehydrated, coated in Ilford K5 or Kodak NTB emulsion diluted with glycerol and ammonium acetate, exposed at 4°C for 11 days, and developed and fixed per manufacturer's instructions. Slides were counterstained with hematoxylin, dehydrated, and mounted with Permount.
Chromogenic HIV-1 and SIV in situ hybridization (C-ISH) was performed as previously described [17]. In addition, we utilized RNAscope for SIV and HIV-1 detection [50]. Each series of specific SIV and HIV-1 target probes was designed to hybridize to viral RNA in gag, pol, vif, vpr, tat, rev, env, nef, and vpx genes (for SIV) (Supplemental Table 2). In addition to probes for SIV and HIV-1 clade B described in this report, we have generated and validated RNAscope probes for HIV-1 clades A, CRF_AE, C, and D with similar results to SIV (Supplemental Table 2; data not shown). The target probe design strategy was described previously [19]. Briefly, each target probe contained an approximately 25-base region complementary to the corresponding SIV (or HIV-1) plus-RNA strand (transcribed transcripts or whole transcribed genome) of each gene, a spacer sequence, and a 14-base tail sequence (conceptualized as a Z). A pair of target double Z probes (ZZ), each possessing a different type of tail sequence, hybridized contiguously to a target region (approximately 50 bases). The two-tail sequences together formed a 28-base hybridization site that binds to a signal preamplifier, which initiates a signal amplification cascade via sequential hybridization, similar to the branched DNA (bDNA) method described previously [20], followed by chromogenic enzymatic detection (horseradish peroxidase using 3, 3-diaminobenzidine [DAB] or tyramide-cyanine 3.5 [PerkinElmer] or alkaline phosphatase using Fast Red substrate). This approach targeted about 4.5 kb of the viral genome. The double-Z probe design strategy ensures superior background control because it is highly unlikely that a nonspecific hybridization event will juxtapose a pair of target probes along an off-target molecule to form the 28-base hybridization site required for binding of the preamplifier and also because a single 14-base tail sequence will not bind the preamplifier with sufficient strength to result in successful signal amplification. Sections of tissues or cell pellets (4-6 μm) were mounted on Superfrost Plus microscope slides (Fisher Scientific), heated at 60°C for 1 hour, dewaxed in xylenes for 10 minutes, and then placed in ethanol 100% for 5 minutes before air drying. RNAscope was performed as previously described [19]. First, slides were incubated with RNAscope Pretreat 1 reagent (endogenous peroxidase block; ACD) for 10 minutes at room temperature. Heat-induced epitope retrieval was performed by boiling sections in RNAscope Pretreat 2 buffer (a citrate buffer [10 nmol/L, pH 6]; ACD) for 30 minutes, immediately washed in double distilled water, and then dehydrated in 100% ethanol for 5 minutes before air drying. Hydrophobic barrier pen was applied to encircle the section, then the slides were incubated with diluted (1:5) RNAscope pretreat 3 reagent (protease digestion solution; 2.5 ug/mL) for 20 to 25 minutes at 40°C using a HybEZ hybridization oven (ACD). Sections were rinsed 3 times in double distilled water and then incubated with pre-warmed target probes (20 nmol/L of each oligo probe) in hybridization buffer A (6X SSC [1XSSC is 0.15 mol/L NaCl, 0.015 mol/L Na-citrate], 25% formamide, 0.2% lithium dodecyl sulfate, blocking reagents) and incubated for 2 hours at 40°C. Slides were washed in wash buffer (0.1X or 0.05X SSC, 0.03% lithium dodecyl sulfate) and incubated with amplification reagents as described in the RNAscope 2.0 HD detection protocol. Amplifier 1 (2 nmol/L) in hybridization buffer B (20% formamide, 5X SSC, 0.3% lithiumdodecyl sulfate, 10% dextran sulfate, blocking reagents) at 40°C for 30 minutes; Amplifier 2 (a proprietary enhancer to boost detection efficiency) at 40°C for 15 minutes; Amplifier 3 (2 nmol/L) in hybridization buffer B at 40°C for 30 minutes; Amplifier 4 (2 nmol/L) in hybridization buffer C (2X SSC, blocking reagents) at 40°C for 15 minutes; Amplifier 5 (a proprietary signal amplifier) at room temperature for 30 minutes; Amplifier 6 (a proprietary secondary signal amplifier) at room temperature for 15 minutes. After each hybridization step, slides were washed with wash buffer three times at room temperature. Before detection, the slides were rinsed one time in 1X TBS Tween-20 (0.05% v/v). Amplification 6 contained alkaline phosphatase (or horseradish peroxidase) labels, and chromogenic detection was performed using FastRed as substrate to generate red signal, DAB to generate a brown signal, or tyramide-cyanine 3.5 (PerkinElmer) for fluorescence detection. Red chromogen development was performed following the RNAscope 2.0 HD detection protocol and reagents, brown chromogen development using ImmPACT™ DAB (Vector Laboratories), and fluorescent detection using tyramide-cyanine 3.5 Plus (PerkinElmer) (chromogen incubation time varied between 2 and 8 minutes). Slides were counterstained with haematoxylin or DAPI (4ʹ,6-diamidino-2-phenylindole) and mounted in Permount (Fisher Scientific) or Prolong ® Gold (ThermoFisher Scientific). Slides mounted in Permount were scanned at high magnification (×400) using the ScanScope AT2 System (Aperio Technologies), yielding high-resolution data from the entire tissue section. Fluorescent slides mounted with Prolong ® Gold (Invitrogen) were imaged on an Olympus FV10i confocal microscope using a 60x phase contrast oil-immersion objective (NA 1.35) and applying a sequential mode to separately capture the fluorescence from the different fluorochromes at an image resolution of 1024 x 1024 pixels.

HIV-1 and SIV DNA in situ hybridization
Viral DNA detection (DNAscope) was performed by utilizing the sense probe targeting the reverse strand for each viral lineage of interest or cellular DNA utilizing a sense probe targeting the complementary CCR5 strand (Supplemental Table 2; data not shown). We modified the RNAscope procedure by adding an RNase tissue pretreatment step (with ribonucleases A [25 μg/ml; Fisher Scientific] and T 1 [25 units/ml; Roche Diagnostics] in 1× Tris-buffered saline [TBS; Boston BioProducts] containing 0.05% Tween-20 [TBS-Tw] for 30 minutes at 37°C) following the RNAscope pretreat three step, which was followed by a short denaturation step in which we incubated the slides at 60°C with warmed (60°C) sense probes for 10 to 15 minutes, and then immediately transferred the hybridization chamber to an oven set at 40°C and hybridized overnight (between 18 to 21 hours). Amplification and detection were performed as described for RNAscope previously, using 0.5X wash buffer for all washing steps. DNAscope validation was performed on 3D8 and ACH-2 cells diluted in uninfected CEM cells (generated as described previously) to make cell preparations at different concentrations ranging from 0% 3D8 or ACH2 cells to 100% 3D8 or ACH2 cells. Cell pellets were embedded in paraffin blocks, after which 4 to 6 µm sections were cut, and SIV DNA in situ hybridization was performed using DNAscope.

Duplex vRNA and vDNA in situ hybridization
Simultaneous visualization of both vRNA and vDNA was performed by combining two custom probe sets-one sense (targeting the vDNA coding strand) and one anti-sense (targeting vRNA transcripts)-covering different regions of the viral genome (Supplemental Table 3) using the DNAscope protocol (without RNase treatment). We utilized a customized RNAscope 2.0 HD Multiplex detection protocol (ACD) for detecting vRNA (Fast Red) followed by detection of vDNA (DAB-Brown) allowing for the visualization of both vRNA and vDNA in the same tissue section.

Quantitative image analysis
To quantify vRNA copies from R-ISH stained tissues, photographic images using epifluorescence were taken with a digital camera, and the TIFF images were used to analyze the area occupied by silver grains using Photoshop (Adobe Systems, San Jose, California) with Fovea Pro 4 (Reindeer Graphics, Asheville, North Carolina), corrected by the background silver grain density of the slide. Section weights were estimated from their volume (5-μm thickness × their area). The number of copies of vRNA was calculated as follows: silver grains observed/cell × 2 disintegrations/ grain × 1 Ci ÷ 2.2x10 12 disintegrations/min ÷ exposure time (min) ÷ specific activity of probe (Ci/ mmole probe RNA) × 6.02 × 10 20 copies/mmole = vRNA copies/cell) [5,6,10]. For virions associated with the follicular dendritic cell network (FDCn), silver grains were enumerated over follicles, excluding grains over vRNA+ cells, and the number of vRNA copies in the follicle estimated as just described, dividing by 2 based on 2 vRNA copies/virion to estimate the number of virions.
To quantify RNAscope stained tissues, whole slides were scanned at 40× magnification with an AT2 slide scanner (Aperio Technologies, Vista, California), and regions of interest (ROIs), including B cell follicles (BCF) were saved as TIFF images for analysis. ROIs were analyzed with Photoshop CS6 (Adobe Systems, San Jose, California) using Noiseware 5 (Imagenomic, Vienna, Virginia) for noise reduction and Fovea Pro 4 (Reindeer Graphics, Asheville, North Carolina) for the analysis. To estimate the number of virions in BCF, ROIs were segmented by reducing noise with Noiseware, duplicating the ROI image, converting the duplicate from RGB to CMYK color mode, thresholding the magenta channel, and then copying and pasting the revised image as a new layer on the original image; the threshold value was determined empirically for each set of slides. To refine the segmentation, Fovea's fill holes command was used to consolidated positive areas, and then the reject features command was applied to remove any objects < 2 pixels or touching an edge of the image. Stained areas of each ROI were recorded with Fovea's area fraction command. Image quantitation data were recorded in text files, and analyzed images were saved in Photoshop format. The fraction of BCF ROIs that stained for vRNA by RNAscope was divided by the average virion size calculated from discretely resolved virions from both SupT1 CCR5+ cells and LTs stained with RNAscope and Fast Red (average virion size = 0.9695 μm 2 ).
To quantify the number of productively infected vRNA+ cells from RNAscope stained tissues, Fast Red staining in an ROI was segmented by reducing noise with Noiseware, duplicating the ROI image, converting the duplicate from RGB to CMYK color mode, thresholding the magenta channel, and then copying and pasting the revised image as a new layer on the original image. The segmentation was refined by morphological opening and closing using Fovea's Euclidean distance map-based morphology operations; when counting positive cells, a morphological opening removed small objects, and then a closing consolidated positive areas within cells. Watershed segmentation was also performed to segment individual stained objects and the Fast Red-positive objects were counted with Fovea. The data were recorded in a text file, and the analyzed image was saved in Photoshop format. The number of vRNA+ cells was calculated based on area (mm 2 ) or per 10 5 cells. Nuclei in an ROI were segmented by thresholding the red channel with the Pun algorithm in Fovea, morphological opening, watershed segmentation, and then counting the segmented objects. The number of vRNA+ cells was divided by the estimated number of cells determined by the nuclei segmentation in that image and multiplied by 10 5 .
The number of vDNA+ cells was determined by either manually counting by at least two individuals who were blinded to the proportion of 3D8 or ACH2 cells on each slide or quantified using a similar approach to virion quantification with noise reduction and thresholding the CMYK magenta channel as above, and rejecting objects < 9 pixels or touching an edge of the image.

Statistical analyses
Statistical analyses were performed using Prism (v6.0; GraphPad Software, La Jolla, California). The Mann-Whitney U test was used for comparisons between groups; the Wilcoxon matchedpairs Signed-Rank test was used for comparisons within groups, and the Pearson's correlation coefficient (Pearson's r) was used to measure of the linear correlation between two variables after log10 transformation of the data. Averaged data were presented as the arithmetic mean ± standard error of the mean (SEM). P values less than 0.05 were considered significant.