HIV-infected macrophages and microglia that survive acute infection become viral reservoirs by a mechanism involving Bim

While HIV kills most of the cells it infects, a small number of infected cells survive and become latent viral reservoirs, posing a significant barrier to HIV eradication. However, the mechanism by which immune cells resist HIV-induced apoptosis is still incompletely understood. Here, we demonstrate that while acute HIV infection of human microglia/macrophages results in massive apoptosis, a small population of HIV-infected cells survive infection, silence viral replication, and can reactivate viral production upon specific treatments. We also found that HIV fusion inhibitors intended for use as antiretroviral therapies extended the survival of HIV-infected macrophages. Analysis of the pro- and anti-apoptotic pathways indicated no significant changes in Bcl-2, Mcl-1, Bak, Bax or caspase activation, suggesting that HIV blocks a very early step of apoptosis. Interestingly, Bim, a highly pro-apoptotic negative regulator of Bcl-2, was upregulated and recruited into the mitochondria in latently HIV-infected macrophages both in vitro and in vivo. Together, these results demonstrate that macrophages/microglia act as HIV reservoirs and utilize a novel mechanism to prevent HIV-induced apoptosis. Furthermore, they also suggest that Bim recruitment to mitochondria could be used as a biomarker of viral reservoirs in vivo.

examined. Several studies have demonstrated the effects of viral proteins on apoptotic protein expression including Bcl-2, Bax, FLICE inhibitory protein (cFLIP) and X-linked inhibitor of apoptosis (XIAP) 17,18 . However, most of these studies have been performed in primary and T-cell lines, while the mechanisms of extended survival and viral transmission by HIV-infected macrophages are still unknown.
Here we demonstrate that HIV-infected human microglia and macrophages function as viral reservoirs. We show that these cells survive HIV infection for extended periods of time by avoiding apoptosis and serving as long lasting viral reservoirs. We also identify that accumulation of Bim in the mitochondria of surviving HIV-infected cells does not result in apoptosis and that Bim in the mitochondria could be used as a biomarker of viral reservoirs in vitro and in vivo.

HIV infects and induces apoptosis in human microglia, but a small population of HIV-infected microglia survives the infection.
To understand the dynamics of apoptosis in HIV-infected microglia, human fetal microglia were isolated, infected with HIV ADA , and stained with DAPI, TUNEL and HIV-p24 to quantify survival of uninfected and HIV-infected cells for up to 120 days. Uninfected cultures of human microglia underwent sustained apoptosis after 21 days in culture (Fig. 1A). HIV infection of microglial cultures resulted in faster initial apoptosis as compared to uninfected cultures up to 21 days (Fig. 1A, *p ≤ 0.007). However, after 21 days post infection, survival of microglial cultures remained stable and higher than uninfected cultures (Fig. 1A, *p ≤ 1.3 × 10 −4 ). To determine whether uninfected or HIV-infected cells were surviving HIV infection, staining for DAPI (to observe the nuclei), phalloidin (to observe the shape of the cells), and HIV-p24 (to detect HIV infection) was performed using staining and subsequent microscopy analysis. The results showed that most surviving microglia in the HIV-infected cultures were HIV-p24 positive (Fig. 1A, 95.27 ± 4.68%) and corresponded to multinucleated cells (82.84 ± 20.09%), indicating HIV infection protects a small population of HIV-infected microglia from apoptosis.

Surviving HIV-infected microglia become latently infected.
To examine the dynamics of HIV replication within the infected microglia, we measured levels of HIV-p24 in culture supernatants by ELISA. HIV replication reached a plateau between 28 to 36 days post infection and then decayed to undetectable levels after 120 days post-infection (Fig. 1B). Although the medium was concentrated ten times using Amicon filters (50 kDa, EMD Millipore, Germany), no secreted HIV-p24 was detected by ELISA at 120 days suggesting that viral replication becomes silent (data not shown).
Analysis of the surviving cells by microscopy and FACS indicates that they maintain their macrophage phenotype including iba-1, CD14, CD68, CD11b/c, CD163, and CSF1R expression as well as their phagocytic function, cytokine secretion (TNF-α, IL-1β, and IL-6), and migratory properties in response to CCL2 as described (19)(20)(21)(22) and replication in human microglia exposed to HIV ADA . Viral replication was measured by HIV-p24 ELISA (n = 4 different donors). Lines with circles represent uninfected cultures. Lines with squares represent HIV replication of human microglia. All points are significantly different from control conditions except by times 0 and 120 days. HIV-p24 secretion was undetectable after 120-150 days in culture. (C) Using the same HIV-infected microglia described in (B), with undetectable replication, after 120-150 days post infection, cells were treated with different factors to reactivate replication. Treatment of latently infected primary cultures of microglia results in viral reactivation. In microglia, SAHA, PHA, LPS, methamphetamine (Meth), and the combination of TNF-α plus IFN-γ results in viral reactivation as compared to control conditions (p < 0.05, n = 3). data not shown). In conclusion, all surviving cells correspond to macrophages or microglia, are infected with HIV, and become latently infected.
Latently HIV-infected microglia can be induced to reactivate viral replication. Critical features of viral reservoirs are extended survival, "hiding" the virus by suppressing viral replication and having the capability of reactivating the virus and spreading it to other cells 12,[23][24][25][26][27] . To demonstrate that surviving HIV-infected microglia are latently infected, we induced HIV reactivation after 120-150 days post infection, when no HIV-p24 production is detected, using several factors known to be involved in CNS and peripheral HIV reactivation 12,24 . Next, we evaluated HIV replication by quantifying HIV-p24 secretion into the tissue culture medium for additional 21 days post treatment (Fig. 1C). Surviving latently HIV-infected microglia were treated with SAHA (N-hydroxy-N′phenyl-octanediamide, suberoylanilide hydroxamic acid, a histone deacetylase inhibitor, 1 and 10 ng/ml), PHA (1 µg/ml), DKK1 (10 and 100 ng/ml), IL-1β (10 U/ml), methamphetamine (1 µM, Meth), LPS (1 µg/ml) or TNF-α (10 ng/ml) and/or IFN-γ (1 ng/ml). DKK1, IFN-γ and IL-1β did not reactivate viral replication (data not shown). However, SAHA, PHA, Meth, LPS and the combination of TNF-α and IFN-γ induced significant viral reactivation (Fig. 1C). In conclusion, the small microglial population that survives acute HIV infection and silences HIV replication can be induced to reactivate the virus upon specific treatments. Thus, these surviving cells fit the criteria of HIV reservoirs.
HIV integrates into the host DNA, produces viral mRNA, replicates, and spreads among human macrophages. Due to the limited numbers of microglia isolated from brain tissue, subsequent experiments were performed using human macrophages. Despite the various manuscripts demonstrating active HIV infection and replication in macrophages 13,[28][29][30][31][32][33][34] , it is still considered controversial that macrophages become infected by HIV mainly because macrophages can store cell-free virions within plasma membrane invaginations for at least a month 15,[35][36][37][38] . Thus, to clarify whether macrophages are truly HIV infected, we analyzed HIV integration by multiple techniques, including fluorescent in situ hybridization (FISH) in combination with immunofluorescence, Alu-PCR, mRNA and HIV protein staining, as well as measurements of cell to cell dissemination at different time points (0, 7, 14, 21 and 28 days post-infection).
Using FISH in combination with immunofluorescence, we detected HIV-DNA Nef integration into the host DNA only in HIV infected cultures as early as 24 h post infection and up to 21 days post-infection ( Fig. 2A,  HIV), where viral replication became undetectable (Fig. 2E). No HIV-DNA Nef staining was detected in uninfected macrophages; only Alu repeats, DAPI and actin showed strong signals as expected ( Fig. 2A, Control). These cultures were 99-100% positive for the macrophage marker Iba-1, indicating no T cell contamination (data not shown). Alu-gag PCR confirmed HIV integration into the host DNA after 7 days post infection (Fig. 2B). Furthermore, analysis of viral RNA expression using RNAscope indicated that HIV gag mRNA was produced in macrophages during the entire time course, while no HIV gag mRNA was detected in uninfected cultures ( Fig. 2C) or using a scrambled probe (data not shown). We also analyzed the intracellular expression of HIV protein p24 (HIV-p24), by cell immunostaining and by ELISA of the culture supernatant ( Fig. 2D and E, respectively). HIV infection of macrophages induced the expression and release of HIV-p24 in a time-dependent manner ( Fig. 2D and E, p ≤ 0.0001, n = 3). The increase in HIV-p24 in the culture supernatant from 7 to 14 days post infection confirmed that HIV infection of macrophages was productive. After 14 days HIV replication decreased indicating that some of the HIV-infected macrophages become latently infected (Fig. 2E). Furthermore, HIV was disseminated in a time-dependent manner: the first cycles of replication only infected 8.2 to 32% of all cells (Fig. 2F, 15.69 ± 12.75%) but after 21 days post infection 100% of the cells were infected (Fig. 2F, *p ≤ 0.0001, n = 3) but HIV replication was undetected (Fig. 2E). Together, these data indicate that HIV efficiently integrates into macrophage DNA, produces viral mRNA, and expresses HIV proteins. Furthermore, the increase in HIV-p24 production in the culture supernatant, as well as the spread of infection over 21 days post infection, indicate that macrophages are productively infected upon exposure to HIV. But also our data demonstrate, like microglia, macrophages become latently infected after 21 days post infection despite that 100% of the cells have integrated HIV DNA.

HIV infection of human primary macrophages results in massive apoptosis, but a small population of HIV-infected cells survives the infection. To determine whether HIV infection results in
survival of some HIV-infected macrophages; cells were stained for nuclei (DAPI staining), actin (phalloidin staining), and TUNEL at different time points (0 to 21 days post infection) and we quantified the numbers of cells in 10 different fields using microscopy. Cultures of uninfected macrophages showed low levels of apoptosis up to 21 days in culture in a similar manner compared with primary microglial cultures. HIV infection of macrophage cultures resulted in a significant decrease in the total number of macrophages in the cultures, as compared to uninfected cultures (Fig. 3B, p ≤ 2.14 × 10 −7 , n = 3). A small but stable population of macrophages survived infection up to 21 days (the last point assayed; Fig. 3C). Most of the surviving cells were positive for HIV-p24 and contained HIV integrated DNA as well as showed clear signs of cell to cell fusion (Fig. 3D, HIV). HIV-infected fused macrophages have been described in vitro and in vivo, but not examined in detail, in multiple tissues even during the ART era [39][40][41][42][43] . In conclusion, these data indicate that similar to HIV-infected microglia, a small population of HIV-infected macrophages survive acute infection and become latently infected.
Surviving fused HIV-infected macrophages are generated by the cell to cell fusion. Currently, they are at least four different mechanisms of multinucleation described in different diseases. First, it has been proposed that cell-to-cell fusion is not truly a result of fusion but instead correspond to phagocytosis of damaged cells 40,44 . The second mechanism is due to incomplete mitosis 40,45 ; however, macrophages are terminally differentiated and do not divide. Third, cell to cell fusion by mechanisms involving IL-4, CD44, SIRP-α, macrophages  fusion receptor (MFR) 46,47 and P 2 X 7 (see review by 48 ). Fourth, a mechanism of HIV-induced cell fusion has between proposed involving gp120, gp41, CCR5, and CD4 49 (Fig. 4A).
To test the third mechanism, we used oxidized ATP (oATP, 100 µM) to block purinergic receptors including P 2 X 7 , and probenecid (500 µM) to block pannexin and connexin hemichannels that control the release of ATP into the extracellular medium. We found that neither of these inhibitors altered cell to cell fusion in response to HIV infection (data not shown), suggesting that cell to cell fusion in HIV-infected condition is by an alternative mechanism. To determine whether the presence of multinucleated cells was the result of fusion or phagocytosis, live cell imaging of uninfected and HIV-infected cultures of macrophages was performed. No cell to cell fusion was detected in uninfected cultures (data not shown). However, exposure of macrophage cultures to HIV resulted in the significant cell-to-cell fusion as early as 2-3 days post-infection (Fig. 4B, 3 days post infection and 30 min of live cell imaging, see * and #). In contrast, using live cell imaging, no phagocytosis was detected in uninfected or HIV-infected conditions (data not shown). To further determine whether cell to cell fusion was present in our Top two circles represent neighboring macrophages infected with HIV and containing HIV Env proteins gp120/gp41 (red bulbs/yellow stalks) in proximity with endogenous CD4 and CCR5 (green bulbs, blue wave). Red outlined insert represents moment just before cell-to-cell fusion in which all components necessary are present on the neighboring cell membrane. The dark arrow represents fusion steps that result in cell-to-cell fusion, where the product of fusion is represented just below the arrow. (B) Representative fusion event captured by timelapse imaging. HIV-infected macrophages were maintained in 60 mm culture dishes until cell to cell fusion was evident by light microscopy, then transferred to an incubated microscope with time-lapse capabilities. In this typical case, cell to cell fusion was captured after 3 days post infection, and fusion of two neighboring cells occurred in approximately 30 minutes. The fusion of two cells is denoted by "*" and " # ", and frames presented identify cellular events that are consistent with cell-to-cell fusion. Frame 1 depicts neighboring cells before fusion. Frame 2 illustrates the fusion point between cells (yellow arrow). Note the cell membrane between the two cells is nearly indistinguishable. Frame 3 illustrates the point at which the borders of the two neighboring cells are indistinguishable, with a southeastern invagination the only indicator that the new cell was once part of a pair. Frame 4 depicts the completion of the fusion event. Note that the size of the resultant cell is nearly the additive size of the original cells marked "*" and " # " from frame 1. (C) Scanning electron micrograph of melding cytoplasm between neighboring macrophages in HIV treated cultures.
HIV-infected cultures, scanning electron microscopy of macrophages in the process of fusion was performed (Fig. 4C, inset). Our data indicate a clear continuous membrane between the two cells in the process of fusion (Fig. 4C, arrow). Together these data suggest that the generation of multinucleated cells in response to HIV infection was a product of cell to cell fusion and not an artifact of phagocytosis, ATP-mediated cell-to-cell fusion or incomplete mitosis.

HIV fusion inhibitors decreased HIV replication but increased the survival of HIV-infected macrophages. To examine whether interaction between HIV-infected cells and host receptors (CD4 and CCR5)
involved in cell to cell fusion participates in the survival of HIV-infected macrophages, cultures were infected with HIV for 24-48 hours to enable successful HIV integration and replication and then treated with HIV fusion blockers T20 (1 µg/mL) or TAK779 (1 µg/mL). While cell to cell fusion was not altered at any time point in the HIV-infected macrophage cultures treated with TAK779 and T20, we did observe a significant (∼99%) decrease in HIV replication (Fig. 5A). Furthermore, TAK779 and T20 treatment of HIV-infected cultures resulted in a higher numbers of HIV-infected surviving macrophages after 21 days post infection, suggesting that these fusion inhibitors increased the survival of the infected cells (Fig. 5B). TAK779 and T20 treatment also protected non-multinucleated cells present in the HIV-infected cultures from apoptosis (Fig. 5C). Microscopy analysis of the surviving cells in the HIV-infected cultures treated with TAK779 at 21 days indicates that fusion was still present, and most cells had undetectable to minimal HIV-p24 staining (Fig. 5D, HIV + TAK-21 days, compare to Fig. 3D). These results indicate that after initial infection and replication, antiretrovirals such as TAK779 and/ or T20 extended the survival of latently infected cells. In conclusion, multinucleation was independent of viral production after initial infection, and HIV fusion inhibitors had an unexpected protective effect on the survival of infected macrophages.

HIV-infected macrophages are protected from HIV-associated apoptosis.
A key event in apoptosis is the formation of the transition pore in the mitochondrial membrane and the release of mitochondrial factors, such as cytochrome C, into the cytoplasm 50,51 . We have reported that HIV infection of human astrocytes results in bystander apoptosis of uninfected neighboring cells by a mechanism that involves the release of cytochrome C (CytC) into the cytoplasm 52,53 . CytC does not induce apoptosis in the infected astrocytes, but the spread of IP 3 and calcium signals through gap junction channels causing apoptosis of uninfected neighbor cells 52,53 . It is well established that the equilibrium between expression, protein-protein interaction, and formation of the transition pore by pro-apoptotic and anti-apoptotic proteins determines the apoptotic fate of the cell [54][55][56] . For instance, anti-apoptotic Bcl2 protein promotes maintaining mitochondrial outer membrane pore integrity by sequestering pro-apoptotic proteins that participate in pore opening 57 . T cell lines engineered to overexpress Bcl2 have been used as a model to study latency and reactivation 58 . However, primary cells that survive infection with clinically relevant strains of HIV have never been found to contain elevated levels of Bcl2 59-61 and if these mechanisms operate in surviving HIV-infected macrophages is unknown.
Here we analyzed the expression of proteins involved in mitochondrial outer membrane (MOM) pore integrity and Cytochrome C-mediated apoptosis, mitochondrial fusion, and apoptosome formation including Bcl-2, Bak, Bax, Bim, mfn-1, XIAP, apaf-1, mcl-1, hsp70, mfn2, hsp27, AIF, and caspase -3 and -9. No changes in protein levels or protein cleavage of these protein was detected (data not shown and Fig. 6A). The only protein affected was Bim. We found that only Bim expression was upregulated in HIV-infected macrophage cultures after 21 days post-infection, or the time when only a few latently HIV-infected macrophages remain in culture ( Fig. 6A and B). In contrast, Bim expression was stable in uninfected cultures (Fig. 6B). Also, no changes in the three Bim isoforms (EL, L, and S) were detected in our cultured macrophages (data not shown) aside of the overall increased expression. These multiple Bim isoforms are generated by alternative splicing and have been proposed to have different functions 62 . Thus, trafficking of the protein into the mitochondria or protein-protein interaction with other apoptotic proteins may be affected. Because Bim is a highly pro-apoptotic protein 63 , the observation that Bim expression was increased in the few surviving HIV-infected macrophages (Fig. 6B) suggests that Bim may have an alternative function in these cells or HIV blocks its apoptotic function.

HIV prevents initiation of apoptosis and formation of the apoptosome.
To examine the mechanisms by which HIV prevents apoptosis in a small population of HIV-infected macrophages, we examined protein expression levels of Apaf-1, CytC, and caspase-3. We detected no changes in expression or activation of caspase-3 in response to HIV infection and no changes in the molecular weights of apoptotic proteins that would result from caspase-mediated cleavage or complex formation with other proteins (Fig. 6C). Also, to identify whether apoptosis-inducing factor (AIF) or cytochrome C are released from the mitochondria into the cytoplasm, cell fractionation was performed in uninfected, and HIV-infected macrophages after 21 days post infection. Our results indicate that AIF and CytC are retained inside of the mitochondria (Fig. 6D). Thus, the apoptotic process is blocked. Finally, to determine whether the mitochondria of the surviving HIV-infected macrophages are functional, we used mitotracker, a membrane potential-sensitive dye. No changes in mitochondrial membrane potential were detected in the surviving HIV-infected macrophages as compared to uninfected cells, indicating that mitochondria are still functional (data not shown). Together, these results indicate that apoptosis is not initiated in HIV-infected macrophages that survive acute HIV infection.

Increased Bim mitochondrial association in HIV-infected macrophages and microglia in vitro and in vivo.
Bim is a highly pro-apoptotic protein in most cell types 63 that is nevertheless highly upregulated in surviving HIV-infected macrophages where apoptosis is blocked (Fig. 6B). Normally Bim associates with the cytoskeleton, but upon apoptosis activation, Bim is recruited into the mitochondrial membrane to trigger apoptosis 64 . Using cell fractionation experiments, we determined that bim was also recruited or sequestered into Scientific RepoRts | 7: 12866 | DOI:10.1038/s41598-017-12758-w the mitochondria specifically in surviving latently HIV-infected macrophages (Fig. 6D, mito fraction, HIV+). To examine whether Bim also associated with mitochondria in vivo, we used human tissues (lymph nodes and brains) obtained from uninfected and HIV-infected individuals without detectable viral replication and performed immunostaining for Bim, VDAC (a mitochondrial protein), HIV-p24, and DAPI.
As expected, analysis of uninfected lymph nodes and brains showed no staining for HIV-p24 protein (Fig. 7, control-LN and control brain, A and C). Furthermore, in uninfected tissues diffuse Bim and VDAC staining was observed in lymph nodes (Fig. 7A) and brains (Fig. 7C). In contrast, in HIV-infected tissues obtained from individuals on effective ART at the time of death (6-24 years of AR treatment), a few low expressing HIV-p24 infected cells were detected (Fig. 7, HIV-LN and HIV-brain, B and C). These infected cells showed increased expression of Bim in perfect colocalization with VDAC, a mitochondrial protein (Fig. 7B and D), consistent with our in vitro Fusion inhibitors TAK and T20 reduced HIV-p24 production collected from the supernatant ( # p ≤ 0.0148 as compared to HIV alone, n = 3). Control cultures did not produce an ELISA signal above background (UI). Supernatant from HIV alone cultures contained 196.5 ± 76 pg/mL HIV-p24, a significant increase from control cultures (*p = 0.0112, n = 3). Supernatant collected from HIV infected cultures treated with TAK or T20 also contained a significant amount of HIV-p24 compared with control cultures, 16 ± 8 pg/mL, and 5 ± 2 pg/mL respectively (*p ≤ 0.0123, n = 3). The amount of HIV-p24 in the supernatant of infected cultures treated with TAK or T20 did not significantly differ from each other. The amount of HIV-p24 in supernatant from HIV alone cultures was significantly higher than both TAK779 and T20 treated cultures ( # p ≤ 0.015, n Negative controls using IgGs and control sera did not show any staining (data not shown). Together, these data indicate that Bim expression is increased and it is recruited to the mitochondria in surviving HIV-infected macrophages/microglia, suggesting that Bim may be used as a biomarker to identify HIV reservoirs in vivo.

Discussion
The data presented here demonstrate that a small population of human microglia and macrophages survive acute HIV infection, that these surviving cells are HIV-infected but protected from apoptosis, and that HIV replication is silenced but can be reactivated in these cells. This indicates that these cells act as viral reservoirs. We found that formation of these reservoirs is associated with macrophage fusion and that cell fusion inhibitors extend the survival of HIV-infected cells. Furthermore, we found that while apoptosis is blocked at an early step, the proapoptotic protein Bim is upregulated and recruited into the mitochondria in latently infected macrophages both in vitro and in vivo. Thus, we propose that Bim association with the mitochondria is a potential biomarker of latently HIV-infected macrophages/microglia in vivo.
The presence of fused macrophages, also known as multinucleated giant cells (MNGC), has been reported extensively in the context of AIDS 65 . Furthermore, several studies of non-AIDS conditions indicate the presence of fused cells in lymphoid organs [66][67][68][69][70][71] , HIV-associated lymphoepithelial cysts of the parotid gland 71 , and colonic mucosa 72 . Thus, these fused cells are present in vivo, even during effective ART, and are found in virally advantageous regions that are part of or are in close contact with lymphoid areas that can support viral reservoirs and reactivation.
Also, we found that HIV fusion inhibitors T20 and TAK779 unexpectedly resulted in extended survival of a subpopulation of uninfected and HIV-infected macrophages, probably due to the reduction in HIV replication (Fig. 2E). These results indicate that some kinds of ART, i.e. those that prevent cell fusion, may extend the survival of HIV reservoirs by avoiding the toxic effects of high replication. Our hypothesis involves two different mechanisms of cell protection induced by TAK779 and T20. First, a protective effect of these antiviral drugs in latently HIV-infected cells mediated by prevention of cell death of HIV-infected cells as described in the results, but there may be a second mechanism of cell protection of uninfected cells, by preventing binding of the virus to CD4 and CCR5. A similar mechanism of cell death protection mediated by CCR5 has been described in several cell types [73][74][75][76] . This point requires further exploration because this indicates that some ART may have negative effects on the survival of HIV reservoirs. Although the HIV genome does not encode any apoptosis inhibitor proteins, several HIV proteins such as Tat can upregulate host anti-apoptotic proteins including Bcl-2, FLIP, XIAP and C-IAP2 [77][78][79][80] . Also, HIV-Nef induces phosphorylation and inactivation of Bad, suggesting that alterations in the apoptotic process contribute to the survival of HIV-infected T cells 81 . However, we did not detect changes in expression of these apoptotic proteins in our latently HIV-infected macrophages. However, we cannot discard changes in protein-protein interactions, conformational changes, post-transcriptional modifications, or activity of the transition pore on the mitochondria altered by HIV infection. These studies will be the focus of future research. However, several laboratories have identified that the formation of transition pore is affected in other cell types, because overexpression or antagonist of Bcl-2 can control HIV replication, reactivation, susceptibility to apoptosis, but the mechanism is unclear [82][83][84][85][86] .
Also, our previous studies of human astrocytes indicate that in infected astrocytes the transition pore in the mitochondrial membrane is formed, and CytC is secreted from the mitochondria into the cytoplasm. HIV, however, blocks the subsequent formation of the apoptosome [87][88][89][90] . In contrast, we did not detect any evidence for the formation of the transition pore or secretion of mitochondrial factors into the cytoplasm in latently infected macrophages, indicating that HIV blocks formation or the function of the transition pore early during the apoptotic process. Thus, the mechanism of how HIV promotes the survival of HIV-infected microglia and macrophages is different from those operating in T cells and astrocytes.
A unique feature of surviving latently infected macrophages is the significant upregulation and recruitment of proapoptotic protein Bim into the mitochondria (Fig. 6B). Bim downregulation in cancer cells is related to extended survival, metastasis and improved response to cytotoxic agents (reviewed in 91,92 ). We propose that upregulation and sequestration of Bim into the mitochondria in surviving infected macrophages reflects an early step in the apoptotic pathway, which is subsequently blocked by HIV because Bim recruitment is not associated with apoptosome formation, secretion of CytC or AIF into the cytoplasm, or activation of caspases. The upregulation of Bim is currently used as a promising cancer therapeutic because upregulation of this protein can result in apoptosis 63 . However, in latently HIV infected cells in vitro and in vivo, Bim did not lead to apoptosis despite its recruitment into the mitochondria. Also, as indicated in the result section, there are no changes in expression or ratios the three different isoforms of Bim. Thus, the mechanism by which Bim is blocking apoptosis of latently HIV-infected macrophages could be related to changes in host protein-protein interaction, the opening of the transition pore, or direct binding of HIV proteins to the pore. Clearly, HIV is altering mitochondrial function/signaling/metabolism 93,94 , but the interplay between metabolism, survival and viral silencing/reactivation is totally unknown and currently is under active investigation in several laboratories.
An additional explanation of the role of Bim in the surviving HIV infected macrophages is its participation in the metabolism of viral reservoirs. Survival of HIV infected macrophages results in mitochondrial fusion and changes in cell metabolism to promote survival of these cells including the use of alternative sources of fuel such as amino acids. Thus, we propose that bim plays a role in metabolic regulation of the survival of these infected cells. Interesting, Bim during the formation of the transition pore also can interact indirectly with enzymes involved in mitochondrial metabolism such as creatine kinase, and hexokinase 95 . Furthermore, both enzymes participate in HIV replication and viral reservoir formation 93,96 , but whether both set of proteins could regulate HIV reservoir metabolism is still under active investigation.
In summary, we have obtained compelling evidence that macrophages/microglia function as HIV reservoirs, and that these viral reservoirs are formed by a novel mechanism involving Bim upregulation and recruitment to mitochondria, which may be useful as a biomarker of viral reservoirs in vivo. These insights, together with the observation that fusion inhibitors increase the size of the latently HIV-infected macrophage pools, need to be incorporated into the current HIV reservoir paradigms and considered during the ongoing efforts to achieve HIV eradication.

Materials and Methods
Reagents. All reagents were purchased from Sigma (St. Louis, MO) except in the places that are indicated otherwise. HIV ADA , TAK779, T20, soluble CD4 (sCD4), were from the NIH AIDS Research and Reference Reagent Program (Germantown, MD). Primers and probes were obtained from Biosynthesis (Lewisville, TX) and PNAbio (Newbury Park, CA). Medium, penicillin/streptomycin (P/S), dyes and secondary antibodies were obtained from Thermo-Fisher (Waltham, MA). Human AB serum and FBS were from Lonza (Walkersville, MD). HEPES was from USB (Cleveland, OH). HIV-p24 ELISA was obtained from Perkin-Elmer (Waltham, MA). Antibodies to HIV-p24 were obtained from Genetex (Irvine, CA). All other antibodies were purchased from Sigma, Santa Cruz (Santa Cruz, CA) or Abcam (Cambridge, MA). Purified mouse IgG 2B and IgG 1 myeloma protein were from Cappel Pharmaceuticals, Inc. TUNEL was obtained from Roche Ltd (Germany). All experiments were performed under the regulations of Rutgers University and the NIH.
Microglia isolation. Human fetal CNS tissue was used as part of an ongoing research protocol approved by Rutgers University (IRB protocols Pro2012001303 and Pro20140000794). Microglia was established as previously described 97 . Briefly, the tissue was minced and shaken. The slurry was passed through a 250 µm nylon mesh filter followed by a 150 µm filter, washed once with HBSS, and then with complete DMEM (DMEM plus 25 mM HEPES, 10% FCS, 1% penicillin-streptomycin, 1% non-essential amino acids). Cells were resuspended and seeded at 9 × 10 7 per 150 cm 2 flask for 12 days. The medium, containing microglia, was then removed and centrifuged for 5 minutes at 220 × g. The microglia was resuspended in complete DMEM and seeded according to the experiment required. Microglia were infected with HIV and medium was collected every 3 days to measure HIV replication. All experiments performed in microglia were determining HIV-replication as determined by ELISA and apoptosis or survival as determined by microscopy. The main limitation of these cultures is the numbers of microglia.
Monocyte isolation and macrophage culture. Human monocytes were isolated from leukopaks obtained from the New York Blood Center. Peripheral blood mononuclear cells (PBMCs) were isolated by differential centrifugation using a Ficoll gradient (GE Healthcare, Piscataway, NJ). Adherent cells were cultured for 7 days in the presence of 10 ng/ml macrophage colony stimulating factor (Miltenyl Biotec, San Diego, CA) in RPMI 1640 with 10% FBS, 5% human AB serum, 1% P/S, and ten mM HEPES to differentiate the cells into macrophages.
HIV infection and replication. After seven days in culture to enable differentiation, macrophages were inoculated with 20-50 ng/ml HIV ADA for 24 hours, and then apoptosis, fusion, and expression of apoptotic proteins was examined. Supernatants were collected, and the medium was changed every 24 hours until 7, 14, 21 and 28 days post-inoculation. Viral replication was analyzed by HIV p24 ELISA according to the manufacturer's instructions.
Live cell imaging. To assess fusion, macrophages were imaged using a Zeiss AxioObserver Z1 with an LD Plan-Neofluar 5X/0.4 10x air objective lens and a Zeiss AxioCam MRm camera using Axiovision software. Stage and objectives were housed within an incubation chamber maintained at 37 °C and 5% CO 2 .

HIV integration by fluorescent in situ hybridization (FISH). Uninfected and HIV-infected mac-
rophages were placed in an ultra-clean glass and incubated in probes directed to HIV-Nef DNA, and Alu repeats as well as immunofluorescence for Iba-1, a macrophage marker, and actin as well as DAPI to label all the nuclei. The resolution of our equipment corresponds to 20 nm per pixel. Thus, colocalization of DAPI, HIV-Nef DNA, and DNA-Alu correspond to integrated HIV DNA. This staining method enables us to detect few copies (even a single copy of HIV integrated DNA) of integrated DNA in the host DNA, however, in this manuscript we did not quantify copy numbers, we only quantified negative versus positive cells. Thus, we expect perfect colocalization between DAPI, Alu repeats, and the HIV integrated DNA is HIV DNA is inserted into the host DNA.
Alu-gag PCR. Integration of HIV into the host genome was detected by Alu-gag PCR as described previously with minor variations 98 . The system was calibrated using OM-10 cells and diluted OM-10 cells into millions of uninfected Hela cells to quantify the lowest numbers of copies possible in 10 8 to 10 12 cells.
In situ Gag RNA analysis. To examine the formation and localization of gag mRNA, RNAscope was used (Newark, CA). However, we analyzed the data using fluorescence because localization of the RNA is more informative than by colorimetric analysis. We used the same protocol described by the provider in combination with immunofluorescence to detect DAPI (to label nuclei) and Iba-1 (a macrophage marker) in the same cell.
Immunofluorescence. Human macrophages, HIV-infected and uninfected, were grown on glass coverslips, fixed and permeabilized in 70% ethanol for 20 min at −20 °C or fixed in 4% paraformaldehyde and permeabilized with 0.01% Triton-X for 2 minutes. Cells were incubated in TUNEL reaction mixture (Roche, Germany) at 37 °C for 1 h, washed three times with PBS and incubated in blocking solution for 30 min at room temperature. Cells were incubated in blocking solution for 30 min at room temperature and then in primary antibody (anti-HIV-p24 or isotype controls: both 1:50) overnight at 4 °C. Cells were washed several times with PBS at room temperature and incubated with phalloidin conjugated to Alexa Fluor 488 (Thermo-Fisher, Carlsbad, CA) to identify actin filaments and/or the appropriate secondary antibody conjugated to FITC (Sigma, St. Louis, MO) for 1 h at room temperature, followed by another wash in PBS for 1 h. Then, cells were mounted using anti-fade reagent with DAPI. Cells were examined by confocal microscopy using an A 1 Nikon (Tokio, Japan) to quantify the total numbers of cells as well as TUNEL positive cells.
the mitochondrial antibodies described and developed with HRP (see original blots in Supplemental Fig. 1). Densitometric analysis was performed using NIH ImageJ software. The original gels are presented in supplemental figure 1.
Human tissue sections. Human tissues were collected as part of the IRB-approved for Rutgers University and the Manhattan HIV Brain Bank and National NeuroAIDS Tissue Consortium (NNTC). Sections of 15 to 25 µm thickness were processed for immunofluorescence and confocal microscopy as described above (n = 11, four uninfected and five HIV-infected with no viral replication detected for 6-24 years, see Table 1 for details).
Statistical analysis. Statistical analyses were used to determine the significance of data from all experiments. Significance was assessed by determining the validity of the null hypothesis that states that all treated groups were the same as their respective controls or HIV infection alone. Origin 8 software was used to test the null hypothesis by comparing the relative value to a theoretical mean of 1 using a two-tailed, two sample t-test with a 95% confidence intervale.