CDK8 inhibitors antagonize HIV-1 reactivation and promote provirus latency in T cells

ABSTRACT Latent HIV-1 provirus represents the barrier toward a cure for infection and is dependent upon the host RNA Polymerase (Pol) II machinery for reemergence. Here, we find that inhibitors of the RNA Pol II mediator kinases CDK8/19, Senexin A and BRD6989, inhibit induction of HIV-1 expression in response to latency-reversing agents and T cell signaling agonists. These inhibitors were found to impair recruitment of RNA Pol II to the HIV-1 LTR. Furthermore, HIV-1 expression in response to several latency reversal agents was impaired upon disruption of CDK8 by shRNA or gene knockout. However, the effects of CDK8 depletion did not entirely mimic CDK8/19 kinase inhibition suggesting that the mediator kinases are not functionally redundant. Additionally, treatment of CD4+ peripheral blood mononuclear cells isolated from people living with HIV-1 and who are receiving antiretroviral therapy with Senexin A inhibited induction of viral replication in response to T cell stimulation by PMA and ionomycin. These observations indicate that the mediator kinases, CDK8 and CDK19, play a significant role for regulation of HIV-1 transcription and that small molecule inhibitors of these enzymes may contribute to therapies designed to promote deep latency involving the durable suppression of provirus expression. IMPORTANCE A cure for HIV-1 infection will require novel therapies that can force elimination of cells that contain copies of the virus genome inserted into the cell chromosome, but which is shut off, or silenced. These are known as latently-infected cells, which represent the main reason why current treatment for HIV/AIDS cannot cure the infection because the virus in these cells is unaffected by current drugs. Our results indicate that chemical inhibitors of Cdk8 also inhibit the expression of latent HIV provirus. Cdk8 is an important enzyme that regulates the expression of genes in response to signals to which cells need to respond and which is produced by a gene that is frequently mutated in cancers. Our observations indicate that Cdk8 inhibitors may be employed in novel therapies to prevent expression from latent provirus, which might eventually enable infected individuals to cease treatment with antiretroviral drugs.

T he ability of HIV-1 to persist within infected cells as an integrated provirus, even during suppressive antiretroviral therapy (ART), represents the main barrier to achieve ART-free HIV remission or cure (1,2). Various strategies to inactivate or elimi nate latent viral reservoirs have been proposed, many of which involve approaches to modulate proviral expression (1,3). Like most cellular genes, latent HIV-1 proviruses produce stochastic transcripts through transcriptional noise (4,5), and there is further speculation that such transcripts may contribute to HIV reservoir maintenance in individuals receiving ART (4). Consequently, it has been proposed that therapies capable of durably suppressing basal stochastic expression might allow treatment to be removed without the risk of viral rebound. This rationale is the basis for the HIV remission strategy known as "block and lock, " where intervention would be applied to durably suppress basal and stochastic provirus expression (6)(7)(8). Much focus toward this objective has centered on the viral Tat protein, which binds the nascent TAR RNA and recruits P-TEFb, containing CDK9 and Cyclin T1, which phosphorylates proteins at the core promoter to encourage transcriptional elongation by RNAPII (9). Inhibitors of Tat block the positive feedback effect caused by this factor in cells producing basal provirus transcripts (10,11). Additional potential block and lock targets under current investigation include factors required for transcriptional initiation or elongation, including inhibitors of CDK9 (12).
Expression of chromosomally integrated HIV-1 provirus in T cells is regulated by multiple signaling pathways that are activated in response to stimulation by cytokines, innate immune responses, and engagement of the T cell receptor with antigen-pre senting dendritic cells. Responses to these signals are controlled by sequence-specific transcriptional activator proteins bound to the 5′ LTR enhancer region, including NFκB, NFAT, GABP/Ets, STAT1/3, AP1, TCF-1/LEF, and RBF-2 (TFII-I, USF1/2) (13,14). These factors stimulate transcription from the HIV-1 LTR promoter through interactions that cause the recruitment of general transcription factors (GTFs) for RNA Polymerase (Pol) II, in addition to co-activator complexes, including the mediator (15,16). Several factors bound to the LTR comprise the RBF-2 complex (TFII-I, USF1/2) which promotes transcriptional elongation by recruiting CDK9/P TEFb through interaction with the co-activator TRIM24 (17,18).
The RNA Polymerase II mediator kinase submodule is comprised of CDK8, or its paralog CDK19, cyclin C, and the regulatory factors Med12 and Med13. This submodule transiently associates with the core mediator, is recruited to promoters by transcriptional activators (19), and acts to modulate transcription by phosphorylating sequence-specific activators as well as GTFs (20). Defined GTF substrates for CDK8/19 include the RNA Pol II B220 C-terminal domain and mediator subunits (19,21,22). Phosphorylation of transcriptional activators by CDK8/19 can have positive or negative effects on transcrip tion depending on the functional effect of the modification (19). CDK8-dependent phosphorylation of β-catenin TCF/LEF (23), NFκB p65 (24), and STAT1/3 (25), factors that regulate expression of HIV-1 provirus in response to T cell signaling, enhances transcription. In contrast, the Notch intracellular domain is negatively regulated by CDK8 phosphorylation, which promotes its degradation (26). Consistent with observations that CDK8 phosphorylation has both positive and negative effects on transcription factors, alterations in the CDK8 and cyclin C (CCNC) genes have been implicated as both oncogenes and tumor suppressors. CDK8 is overexpressed in a variety of can cers, including breast and colorectal carcinomas, and malignant melanomas, where expression is associated with tumor progression (23). In contrast, CDK8 overexpression produces tumor suppressive effects of cancers promoted by Notch or EGFR signal ing (27). The significance of alterations in mediator kinase components and cancer progression has spurned the development of specific CDK8/19 inhibitors (28)(29)(30), several of which are in clinical trials for ER-positive breast cancers and acute myeloid leukemia (31). Despite that HIV-1 expression is regulated by at least three factors whose activity is controlled by CDK8, TCF/LEF, NFκB, and STAT1/3, the role of CDK8/19 and the kinase module for regulation of HIV-1 provirus expression and response to T cell signaling has not been characterized.
In this report, we examine the effect of CDK8/19 inhibitors and CDK8 knockouts on expression of HIV-1 provirus. We find that Senexin A and BRD6989, structurally distinct chemical inhibitors specific for CDK8/19 kinase activity, impair reactivation of HIV-1 provirus in cell line models of latency and encourage the establishment of immediate latency in newly infected T cells. Inhibition of CDK8/19 impairs recruitment of RNA Pol II to the LTR promoter. Furthermore, CRISPR-mediated knockout of the CDK8 gene impairs the expression and activation of HIV-1 in response to several latency reversal agents (LRAs). Interestingly, disruption of CDK8 does not entirely mimic the effect of CDK8/19 kinase inhibition, suggesting that CDK8 and CDK19 do not function redundantly for the activation of proviral transcription. Collectively, our observations indicate that CDK8/19 inhibitors, including those currently in clinical trials for the treatment of various cancers, may prove useful for therapies to eliminate latently infected cells by suppressing HIV-1 expression.

Chemical inhibition of CDK8/19 kinase activity suppresses HIV-1 expression
We examined the effect of small molecule inhibitors of CDK8/19 kinase activity on HIV-1 expression by treating the previously characterized JLat10.6 (32) and mHIV-luci ferase Jurkat (33) human CD4 + T cell lines with two structurally unrelated CDK8/19 inhibitors, Senexin A and BRD6989 (Fig. 1A) (34,35). JLat10.6 Jurkat cells possess a full-length HIV-1 provirus with GFP expressed instead of Nef from the 5′ LTR (36), while Jurkat Tat mHIV-luciferase cells express luciferase as a fusion with gag from the 5′ HIV-1 LTR, but no other HIV-1 proteins (Fig. 1B) (18,33). We did not observe an effect on basal GFP expression in JLat10.6 cells upon treatment with these inhibitors, but induction of GFP expression in cells treated with the PKC agonist PMA was inhibited in a concentration-dependent manner ( Fig. 1C and D). A similar effect was seen with mHIV-luciferase cells although with this reporter line we also recorded a significant reduction in basal luciferase expression upon treatment with Senexin A but not with BRD6989 ( Fig. 1E and F). The inhibitory effect of the CDK8/19 inhibitors on HIV-1 expression was not due to toxicity, as concentrations as high as 100 µM had only minor effects on cell viability ( Fig. S1A and B). Additionally, we examined the effect of the CDK8/19 inhibitors on the expression of HIV-1 provirus in the ACH2 CEM CD4 + T cell and U1 monocyte models of latency, where we likewise observe inhibition of basal and PMA-induced expression of viral mRNA ( Fig. 2A and B) and no toxicity at concentrations where inhibition was detected ( Fig. S1C and D). These observations indicate that chemical inhibition of CDK8/19 kinase activity suppresses HIV-1 expression in several cell line models of provirus latency.

CDK8/19 antagonizes HIV-1 reactivation by latency reversing agents
We examined the effect of CDK8/19 inhibitors on stimulation of HIV-1 expression by additional signaling agonists. Treatment with a combination of the phorbol ester PMA and ionomycin mimics T cell activation by stimulation of the MAPK, PKC, and calcineurinsignaling pathways downstream of the T cell receptor (37,38). PMA stimulates PKC, activating the MAPK and IKK-IκB-NFκB pathways, while ionomycin causes the release of intracellular calcium resulting in activation of NFAT through calcineurin-mediated dephosphorylation (14). We found that HIV-1 expression in both the JLat10.6 and mHIVluciferase cell lines treated with PMA, alone or in combination with ionomycin, is inhibited by Senexin A and BRD6989 ( Fig. 3A and B). Similarly, the CDK8/19 inhibitors also significantly impair HIV-1 activation in response to Ingenol 3-angelate (PEP005) (Fig. 3A and B), a latency reversal agent which activates NFκB (39,40).
Latent HIV-1 provirus can also be reactivated by treatments which affect chromatin modifications or association of epigenetic modifiers with the LTR in the absence of signaling agonists (41). The histone deacetylase inhibitor (HDACi), suberanilohydroxamic acid (SAHA), and the BRD4 inhibitor JQ1 are well-characterized latency reversing agents (LRAs) (40,42). SAHA induces expression of HIV-1 to a comparable level as PEP005 in the JLat10.6 cell line, and we found that this effect was also inhibited by Senexin A and BRD6989 (Fig. 3A). The effect of SAHA on HIV-1 expression in the mHIV-luciferase line was not as pronounced, and here, we observed only a minor but significant effect of the CDK8 inhibitors (Fig. 3B). JQ1 caused significantly lower levels of reactivation in both reporter cell lines, which was only significantly affected by the CDK8/19 inhibitors in the mHIV-luciferase cell line ( Fig. 3A and B). These results demonstrate that CDK8/19 kinase activity is required for the most robust reactivation of HIV-1 expression in response to a variety of LRAs.

Effect of CDK8/19 inhibition on cell growth
Above, we investigated the effect of Senexin A and BRD6989 on HIV-1 expression by employing temporally acute treatments. To further examine the cellular impact of CDK8/19 inhibition, we assessed the growth of Jurkat T cells exposed to a range of drug concentrations for 4 days (Fig. 4). On day 3, 10 µM of Senexin A was found to slightly impair cell expansion while 50 µM of the compound elicited cell-cycle arrest (Fig. 4A). In contrast, cell-cycle progression was less affected by BRD6989 with concentrations as high as 50 µM having minor observable effects on cell division (Fig. 4B). Importantly, no effect on cellular viability throughout the 4 days examined was observed for any of the employed Senexin A or BRD6989 concentrations ( Fig. 4C and D).

Inhibition of CDK8/19 encourages establishment of HIV-1 latency
Having determined concentrations of Senexin A and BRD6989 that have no observed toxicity and minimal impact on cell cycle progression, we next examined the role of CDK8/19 for the establishment of immediate latent infection in Jurkat T cells using the replication incompetent Red-Green-HIV-1 (RGH) virus. RGH is a dual reporter virus where GFP is expressed from the 5′ LTR and mCherry from an internal constitutive PGK pro moter, thus allowing latently infected cells (mCherry+/GFP−) to be discriminated from productively infected cells (mCherry+/GFP+) as early as 24 h post infection by flow cytometry (Fig. 5A and B) (43,44). Consistent with previous observations (44,45), the proportion of productively infected cells peaks at 4 days post infection and then declines as the provirus succumbs to silencing (Fig. 5C, Ve) (45). In contrast, infection of cells treated with Senexin A or BRD6989 over the course of the analysis produced a significantly lower proportion of productively infected cells, with the largest discrepancy occurring at 7 days post infection ( Fig. 5B and C). Similar results were observed with infection of the SupT1 human T cell line ( Fig. S2A and S2B) and with an additional dual HIV-1 reporter virus, HIV GKO ( Fig. S2C and S2D) (46). These results indicate that inhibition of CDK8/19 kinase activity in newly infected cells encourages the establishment of immediate latent provirus.
In addition, we examined whether treatment of cells with the CDK8/19 inhibitors post infection affected the proportion of latent or productively infected cells. For this, we treated cells with the CDK8/19 inhibitors 4 days post infection and measured the proportion of latently and productively infected cells 24 h later (Fig. 5D). In this experi ment, we found that Senexin A caused an ~20% reduction in productive infections, while BRD6989 caused a more modest ~10% decrease as compared to cells treated with a vehicle control (Fig. 5E). This result indicates that effects of the CDK8/19 inhibitors are not temporally restricted to immediate events in HIV-1 infection, including entry and formation of integrated provirus, but rather must inhibit expression of the virus post integration, which is consistent with the effect of these compounds on reactivation of provirus reporter by latency-reversing agents. the CDK8/19 inhibitors or a vehicle control prior to infection with the replication incompetent RGH dual-reporter virus. Proviral activity was assessed by flow cytometry 3 days post infection (Fig. 6A). Consistent with the results described above with immortal ized T cell lines, we found that Senexin A and BRD6989 cause a significant decrease in the proportion of productively infected cells, represented by the expression of both mCherry and GFP as compared to mCherry alone (Fig. 6B), where Senexin A inhibited to a greater effect than BRD6989 ( Fig. 6C and D). Furthermore, we found that proviruses that managed to establish productive infections in the presence of Senexin A or BRD6989 displayed lower levels of transcriptional activity than their untreated counterparts (Fig.  6D, Productive Inf.). Notably, the CDK8/19 inhibitors did not affect the viability (Fig. 6E) and only had a slight impact on division (Fig. 6F) of the primary CD4 + T cells over the course of the treatment similar to the results obtained from Jurkat T cells (Fig. 4). Finally, primary CD4 + T cells treated with either CDK8/19 inhibitor showed no difference in Full-Length Text susceptibility to HIV-1 infection (Fig. 6G). Collectively, these results indicate that Senexin A and BRD6989 inhibit HIV-1 expression in primary T cells, an effect which encourages the formation of latency.

Senexin A and BRD6989 do not inhibit HIV-1 Tat and block HIV reactivation in a cell-specific manner
Originally isolated from the murine sponge Corticium simplex (47), the steroidal alkaloid cortistatin A (CA) was found to inhibit CDK8/19 kinase activity (48,49). Because of its limited availability from nature, analogs of CA were developed, including didehydro-Cortistatin A (dCA) (50,51). Subsequently, dCA was found to suppress HIV-1 expression by a mechanism proposed to involve inhibition of Tat, the HIV-1 transactivator of transcription, independently from effects involving inhibition of CDK8/19 activity (11,12,52,53). Given these reported effects of dCA on HIV-1 Tat, we examined whether the CDK8/19 inhibitors Senexin A and BRD6989 had a similar effect on Tat function. To this end, we transfected HEK293T cells with an HIV-1 LTR reporter plasmid where GFP is expressed from an IRES, alone or in combination with viral Tat protein (Fig. 7A). As expected, transfection of HEK293T cells with the reporter co-expressing Tat resulted in substantially elevated GFP expression compared to the reporter lacking Tat (Fig. 7B). Intriguingly, neither of the CDK8/19 kinase inhibitors had an effect on the expression of

Full-Length Text
Journal of Virology GFP from either of these reporters (Fig. 7B), indicating that these compounds likely do not directly affect Tat function, unlike the reported mechanism of dCA (12,53). Consis tent with this observation, we note that HIV-1 provirus in the ACH2 cell line possesses a point mutation in TAR rendering it defective to Tat transactivation (54), while the provirus in U1 cells expresses defective Tat protein (55). As shown above (Fig. 2), Senexin A and BRD6989 both inhibit reactivation of HIV-1 expression in these lines. Furthermore, Senexin A and BRD6989 are not structurally related to dCA (34,35), and these results indicate they must have partially distinct mechanism(s) of action. Considering that the CDK8/19 inhibitors antagonize HIV-1 expression in Jurkat and primary T cells but did not affect basal or Tat-activated expression of the LTR-IRES-GFP reporter in HEK293T cells (Fig. 7B), we wondered whether the effect of these compounds on HIV-1 expression may be cell type-specific. To further examine this, we used the HeLaderived TZM-bl cell line which bears an integrated HIV-1 LTR-β-Gal-luciferase reporter (56). We observed induction of luciferase expression in this line in response to PMA, or Full-Length Text the LRAs PEP005, SAHA, and JQ1. However, treatment with Senexin A or BRD6989 had no effect on this response (Fig. 7C). These observations indicate that effects of CDK8/19 kinase activity on HIV-1 expression may be specific to cells of leukocyte lineage. Relating to this, we note that it was previously reported that knockdown of CDK8 and/or CDK19 does not affect HIV-1 infection in HeLa cells, where dCA has an inhibitory effect (12). This supports a view that dCA must have additional inhibitory target(s) apart from the CDK8/19 kinase, such as Tat, and that the effect of mediator kinase for HIV-1 expression may be unique to T cells.

Inhibition of CDK8/19 inhibits recruitment of RNAPII to the LTR
Regulation of transcription from the HIV-1 LTR, like most cellular genes, involves the recruitment of RNA Pol II and histone-modifying complexes to the promoter as well as regulation of transcriptional elongation and mRNA processing (1,57). To examine which of these steps in HIV-1 expression may be affected by CDK8/19 kinase inhibition, we assessed recruitment of RNA Pol II to the LTR using ChIP-qPCR. We found that treatment of unstimulated cells with Senexin A caused a significant decrease in association of RNA Pol II at the LTR RBE3 and RBE1 cis-elements that are located upstream of the enhancer region (−130) and near the transcriptional start site (−22), respectively (Fig. 8A, compare Ve and SenA). Treatment with PEP005, a PKC agonist which imparts activation of NFκB, resulted in a significant increase in LTR-associated RNAPII (Fig. 8A, compare Ve and PEP005), but this effect is blocked by the CDK8/19 inhibitor Senexin A (Fig. 8A, compare PEP005 and PEP005/SenA). These results indicate that CDK8/19 kinase activity is necessary for the efficient recruitment of RNA Pol II to the LTR.
Next, we examined the effect of CDK8/19 kinase inhibition on the localization of CDK8 and CDK19 to the HIV-1 LTR. Treatment of cells expressing Flag-tagged CDK8 or CDK19 with Senexin A does not appear to affect abundance of these proteins (Fig. 8B), but interestingly, we observe increased association of both kinases with the LTR in cells treated with this inhibitor (Fig. 8C and D). It is possible that direct interaction of the CDK8/19 kinase module negatively regulates HIV expression, which is consistent with a previous report indicating that CDK8 is evicted from the HIV-1 LTR upon T cell stimulation and induction of virus expression (58). These observations would support a view that interaction of mediator kinase module with the promoter has a negative regulatory effect on transcription, a function that was also indicated from experiments with yeast (59) and transcription reactions in vitro (60). However, decreasing promoter transcrip tional activity through TFIIH inhibition was found to enrich core promoter-associated mediator in yeast (59). It is possible that suppression of LTR transcription in response to CDK8/19 inhibition causes lagged kinetics that allows our ChIP experiment to capture Flag-tagged CDK8 or CDK19.

CDK8 and CDK19 are activators of HIV-1 expression
The results above demonstrate that chemical inhibition of the CDK8/19 kinases inhibits HIV-1 basal and activated expression in Jurkat T cells and normal CD4 + lymphocytes. All current small molecule inhibitors of the mediator kinases do not differentiate between effects on CDK8 and CDK19 (29,31). Although these kinases are generally considered to have overlapping or redundant function for the mediator kinase module (31,61), the role of CDK19 is less well understood and some evidence suggests specific functions for CDK19 (62). To assess the specific role of CDK8 in regulating HIV-1 expression, we used shRNA-mediated knockdown of CDK8 in both the mHIV-Luciferase (Fig. S3A) and JLat10.6 ( Fig. S3B) cell lines. We found that knockdown of CDK8 by shRNA in the mHIV-luciferase cell rendered the integrated provirus significantly impaired for PMA-induced luciferase reporter expression (Fig. S3C). Curiously, we found that depletion of CDK8 expression by shRNA in the JLat10.6 cell line did not inhibit induction of GFP upon PMA treatment (Fig.  S3D). To examine the effect of CDK8 on HIV-1 transcription in more detail, we performed CRISPR-Cas9 gene editing to generate mHIV-luciferase (Fig. 9A, lanes 2-7) and JLat10.6 ( Fig. 9A, lanes 9-11) CDK8 knockout (CDK8 KO) cell lines. Consistent with results from shRNA knockdown, we found that all of the CDK8 KO mHIV-luciferase cell lines displayed a significant defect for reactivation of expression in response to PMA (Fig. 9B). However, similar to results with shRNA knockdown, we did not observe an effect of the CDK8 KO in JLat10.6 cells on proviral expression in response to PMA (Fig. 9C). These observations indicate that CDK8 function is required for reporter gene induction in the mHIVluciferase but not in the JLat10.6 line. Considering that the chemical inhibitors of CDK8/19 inhibit PMA-induced HIV-1 expression in both of these lines (Fig. 1), the differential effect of CDK8 disruption in the mHIV-luciferase and JLat10.6 lines may indicate functional differences between CDK8 and CDK19, a possibility we discuss further below. We do note that JLat10.6 cells typically express slightly higher levels of CDK8 protein than mHIV-luciferase cells as determined by immunoblotting ( Fig. 9A and S4).
Previously, we observed differential expression of HIV-1 in response to PMA between the mHIV-luciferase and JLat10.6 CDK8 KO Jurkat cell lines (Fig. 9). These are two distinct clonal lines, and as such, we sought to examine the effect of CDK8 abrogation on the expression of a heterogenous population of proviruses. For this analysis, we used the replication incompetent Red-Blue-HIV-1 (RBH) dual HIV-1 reporter which expresses all viral gene products apart from Nef and Env, and where BFP is expressed from the 5′ LTR and mCherry from an internal CMV promoter (Fig. 10A) (43). Wildtype (WT) or CDK8 KO Jurkat T cells were infected with RBH, and the populations were subsequently treated with a vehicle control (Ve) or an LRA 4 days post infection; flow cytometry was performed 1 day later to examine BFP and mCherry expression ( Fig. 10B; Fig. S5). Importantly, we found that the proportion of productively infected cells in the otherwise untreated population was significantly lower in CDK8 KO Jurkat cells relative to WT (Fig. 10C, Ve). The proportion of productively infected cells treated with the PKC agonists PMA and PEP005 was significantly increased, but interestingly, this proportion was nearly identical in the CDK8 KO line (Fig. 10C, PMA, PEP005). In contrast, we found that treatment of RBHinfected CDK8 KO cells with the histone deacetylase inhibitor SAHA, or the bromodomain inhibitors JQ1 and IACS-957 (18) failed to recapitulate the level of productive infections observed for wild-type cells (Fig. 10C, SAHA, JQ1, IACS-9571). Luciferase expression was measured following 4 h of treatment and is displayed as relative light units (n = 3, mean ± SD).

Full-Length Text
Beyond assessing the ratio of productive versus latent infections, we took a closer examination of the effect of CDK8 KO on LTR-driven BFP expression. This analysis was performed by independently evaluating the BFP mean fluorescence intensity (MFI) of all infected cells (mCherry+/BFP− and mCherry+/BFP+, All Inf.) and the productive proviral infections (mCherry+/BFP+, Productive). Comparison of BFP expression of all infections showed that CDK8 KO cells were associated with decreased LTR activity upon treatment with a vehicle control (Ve), SAHA, JQ1, or IACS-9571, but no difference was apparent upon stimulation with PMA or PEP005 (Fig. 10D, All Inf.). This is expected as we are examining BFP expression for the whole of integrated proviruses and the results mirror the effect of CDK8 KO on the productive infection ratio for the indicated treatment (Fig.  10C). Next, we examined LTR-driven BFP expression of the proviruses that established productive infections (mCherry+/BFP+ population). Here, we observed a slight but nonsignificant decrease in viral transcription of CDK8 KO cells following treatment with PMA, PEP005, or SAHA (Fig. 10D). However, the productive proviral infections of CDK8 KO cells following incubation with a vehicle control, JQ1, or IACS-9571 displayed greatly reduced LTR expression as compared to wildtype (Fig. 10D, Productive). Collectively, these results demonstrate that CDK8 regulates basal HIV-1 transcription as well as activation in response SAHA, JQ1, and IACS-9571. However, CDK8 is not required for viral induction in response to PMA or PEP005. As the CDK8/19 dual inhibitors reduce HIV-1 expression in response to these agonists ( Fig. 1 and 3), CDK19 must be sufficient to mediate LTR induction in response to PKC activation as mediated by PMA or PEP005.

Full-Length Text
The RBH reporter virus allowed us to assess a heterogenous population of proviruses and identify latency reversal pathways that are strongly or weakly dependent on CDK8. For instance, CDK8 was not required for viral activation as induced by the PKC agonists PMA or PEP005 but was necessary for LTR responsiveness to the histone deacetylase inhibitor SAHA, the BRD4 bromodomain inhibitor JQ1, and the TRIM24 bromodomain inhibitor IACS-9571 ( Fig. 10C and D). To characterize this further, we examined the reactivation of wildtype and CDK8 KO JLat10.6 cells in response to the HDACi, SAHA. Although JLat10.6 cells did not require CDK8 for PMA-mediated reactivation (Fig. 9C), we found that CDK8 null cells displayed substantially dampened responsiveness to SAHA (Fig. 10E and F). Altogether, these results indicate that CDK8 and CDK19 may have overlapping functions for viral response to PMA or PEP005 but divergent function for reactivation by the HDAC inhibitor SAHA.

CDK8/19 inhibitors suppress HIV-1 reactivation in CD4 + T cells
Because inhibition of CDK8/19 impairs HIV-1 expression in multiple models of latency as well as normal CD4 + T cells, we wondered if these inhibitors could also block reactivation of virus in primary CD4 + T cells from individuals with HIV-1 receiving suppressive ART. Treatment of CD4 + T cells from five participants with a combination PMA and ionomycin stimulated HIV-1 mRNA expression in all cases (Fig. 11A, PMA/Ion). However, Senexin A inhibition of CDK8/19 kinase activity robustly limited induction of HIV-1 viral expression by PMA and ionomycin in CD4 + T cells from all ART-treated participants (Fig. 11A, Full-Length Text PMA/Ion/SenA). We observe a similar effect on the expression of IL2 in these samples (Fig. 11B), which is consistent with IL2 and HIV-1 both being regulated by T cell signaling Full-Length Text (13). Senexin A caused minor alterations, but more divergent effects on the expression of CD69, a marker of T cell activation in these treatments (Fig. 11C). Importantly, Senexin A treatment had no discernable effect on cell viability at concentrations where we observe inhibition of HIV-1 provirus reactivation (Fig. 11D). Collectively, these observations indicate that chemical inhibitors of CDK8/19, including Senexin A, may represent potential therapies to suppress HIV-1 provirus for elimination of latently infected cells in individuals on ART.

DISCUSSION
Expression of chromosomally integrated HIV-1 provirus in T cells is dependent upon cellular factors that bind the LTR enhancer region and are regulated by signaling pathways stimulated by T cell receptor engagement (14,63). Of central importance to this response is NFκB, which is activated by phosphorylation and inhibition of its cytoplasmic inhibitor I-κB, allowing translocation of NFκB p65 RelA to the nucleus where it binds two sites within the LTR enhancer region and activates transcription (64,65). Previous work showed that CDK8/19 is co-recruited with NFκB to promoters and that CDK8/19 inhibitors impair transcriptional activation by NFκB (24). Additional transcrip tional activators regulated by CDK8 also bind the 5′ HIV-1 LTR, including β-catenin TCF/LEF (23) and STAT1/3 (25), which is consistent with our findings that CDK8/19 inhibitors, CDK8 knockdowns, and CDK8 gene knockouts limit reactivation of HIV-1. We find that CDK8/19 inhibitors impair induction of HIV-1 provirus in cell line models and newly infected CD4 + T cells, and in response to a variety of latency-reversing agents, including PEP005, SAHA, and JQ1. Importantly, we find that the CDK8/19 inhibitor Senexin A prevents reactivation of latent virus expression in CD4 + T cells from individuals with HIV receiving suppressive antiretroviral therapy. These results indicate that chemical inhibitors of CDK8/19, some of which are presently in clinical trials for various cancers (66), could be useful to provide longstanding and durable suppression of the latent HIV-1 reservoir.
We find that Senexin A impairs recruitment of RNA Polymerase II to the LTR promoter, which suggests that it causes a defect in transcriptional activation by LTR-bound transcription factors. Treatment of T cells with the PKC agonist PEP005 causes significantly enhanced association of RNA Pol II with the LTR, which likely represents the effect of LTR-associated NFκB p65 and recruitment of general transcription factor complexes, including mediator and P-TEFb (67,68). Stimulation of NFκB p65 by TNF-α was shown to cause recruitment of TFIIH/ CDK7, which was identified as a rate-limiting event for reactivation of provirus from latency (58). We find that Senexin A inhibits association of RNA Pol II to the LTR in untreated cells and those stimulated with the PKC agonist PEP005 (Fig. 8A). Latent HIV-1 provirus is associated with RNA Pol II that is paused immediately downstream of the transcriptional start site, following synthesis of the nascent TAR RNA (69), and activation of HIV-1 involves stimulation of elongation from the viral promoter through recruitment of P-TEFb containing CDK9 and cyclin T1 (70). However, the factors and mechanisms involved in establishing paused RNA polymerase II and associated GTFs at the latent HIV-1 core promoter have not been determined. Our results indicate that CDK8 activity might be required for the establishment of paused RNA Pol II at the promoter. Consistent with this possibility, the mediator kinase module, including CDK8, was found associated with the latent virus promoter (58).
The CDK8/19 inhibitors Senexin A and BRD6989 have similar effects on CDK8 and CDK19 in vitro and presumably do not discriminate for inhibition of these paralogs in vivo (35,71). Consequently, it is interesting that these compounds cause only partial inhibi tion of HIV-1 reactivation in cell line models, but that knockout of the CDK8 gene in the mHIV-luciferase line nearly completely inhibits reactivation in most knockout clones (Fig.  9B). This observation may indicate that CDK8 has a structural role for regulation of HIV-1 transcription in addition to catalytic function. Furthermore, although depletion of CDK8 by shRNA-mediated knockdown or genetic ablation caused significant inhibition of HIV-1 expression in the mHIV-luciferase line, they have no effect on PMA or PEP005-stimulated GFP reporter expression in the JLat10.6 line. Potential mechanistic differences between these kinases might depend upon specific properties of the JLat10.6 and mHIV-luciferase cell lines, for example, differences in the site of chromosomal integration. Also, we note that luciferase is expressed in the mHIV-luciferase as a fusion with Gag from an unspliced transcript, whereas expression of GFP in JLat10.6 is dependent upon production of a spliced sub-genomic RNA. Spliced HIV-1 transcripts, including that encoding Nef, which is replaced by GFP in the JLat10.6 line (36), are expressed early in infection or reactivation from provirus latency. Splicing of sub-genomic transcripts is suppressed later in infection by the viral Rev protein, which enables synthesis of full-length genomic RNAs (72). In contrast, expression of luciferase as a fusion with Gag in the mHIV-luciferase cell line does not require mRNA splicing (Fig. 1B) (73). Alternatively, the differential effect of CDK8 and CDK19 between reactivation of HIV-1 in these lines may be dependent upon the expression of additional HIV-1 gene products, as the JLat10.6 line is capable of express ing all viral proteins, apart from Nef (1), whereas coding sequences for all viral accessory factors is deleted in the mHIV-Luciferase line (73).
Additionally, there was an apparent overlap in phenotype between CDK8 KO cells and CDK8/CDK19-inhibited cells when we examined the heterogenous population Full-Length Text of proviruses that was generated by RBH infection. In response to SAHA, JQ1, and IACS-9571, CDK8 ablation mimicked the effect of CDK8/19 inhibition, causing suppressed LTR transcriptional activity. However, unlike CDK8/19 inhibition, loss of CDK8 had no effect on proviral induction in response to the PKC agonists PMA or PEP005. Furthermore, activation of HIV-1 expression in CDK8 null JLat10.6 cell lines shared a similar pattern to RBH proviruses. Given these observations, it is likely that HIV-1 reactivation in response to SAHA, JQ1, and IACS-9571 is largely dependent upon CDK8, while CDK19 is generally sufficient for PKC-mediated transcriptional induction. The mechanism(s) contributing to these differences are not understood but likely represent distinct functions for CDK8 and CDK19 as Senexin A and BRD6989 inhibit HIV-1 expression in response to all of these stimuli.
The mediator kinases were previously shown to have distinct functions, where it was shown that CDK19 but not CDK8 has a significant role for regulation of cell cycle progression of mouse hematopoietic stem cells (62). We also observed that Senexin A and BRD6989 did not have an effect on HIV-1 expression in HEK293T or HeLa cells, indicating possible cell-type-specific functions for the mediator kinase. This latter result could account for a previous report indicating that the mediator kinases do not affect HIV-1 expression (74). Overall, a more detailed understanding of the mechanistic role of CDK8 and CDK19 for regulation of transcription will be required to resolve these discrepancies.
Latent HIV-1 proviruses are known to produce sporadic transcripts through transcrip tional noise (75,76), such that even the most strongly repressed provirus will occasion ally produce viral transcripts (4,45). Considering these observations, it is possible that stochastic expression of latent provirus may contribute to maintenance of the latent population in tissue compartments where ART may not be capable of preventing local spread of virus (4). Consequently, one proposed therapeutic strategy, designated block and lock, would involve intervention to suppress stochastic expression of provirus in patients on ART to promote deep latency, which may enable the elimination of the latently infected population by cell lifespan decay (6,8). The results presented here indicate that inhibitors of the CDK8/19 mediator kinases may be an important contribu tion toward this strategy.
Peripheral blood mononuclear cells (PBMCs) from participants with HIV-1 on ART were isolated from whole blood by density gradient centrifugation using Lymphoprep and SepMate tubes (StemCell Technologies) and cryopreserved. Upon thawing, PBMCs were cultured in RPMI supplemented with 10% FBS, penicillin(100 units/mL), streptomy cin (100 g/mL), and L-glutamine (2 mM). All primary cells were incubated in a humidified 37°C and 5% CO 2 atmosphere. Samples from participants were collected with written informed consent under a protocol jointly approved by the research ethics boards at Providence Health Care/UBC and Simon Fraser University (certificate H16-02474).

Luciferase reporter assays
For TZM-bl cells, 2 × 10 4 cells were plated with 100 µL DMEM per well in 96-well plates. Following 24 h, cells were incubated with the indicated concentration of drug for 4 h and luciferase expression was measured. For Jurkat mHIV-luciferase cells, 1 × 10 5 luciferaseexpressing cells were plated with 100 µL media in 96-well plates. Luciferase activity was measured after the indicated time of treatment. Measurements were performed using the Superlight luciferase reporter Gene Assay Kit (BioAssay Systems) as per the manufacturer's instructions and 96-well plates were read in a VictorTM X3 Multilabel Plate Reader.

Flow cytometry
Cells were treated as indicated in the figure legends. For flow cytometric analysis, human T cell lineages were suspended in PBS, while HEK293T cells were suspended in PBS containing 10% trypsin-EDTA to prevent aggregation. A BD Biosciences LSRII-561 system was used for flow cytometry with threshold forward scatter and side scatter parameters being set so that a homogenous population of live cells was counted (Fig. S6). FlowJo software (TreeStar) was used to analyze data and determine the indicated MFI.

Statistics and reproducibility
All replicates are independent biological replicates and are presented as mean values with ± standard deviation shown by error bars. The number of times that an experi ment was performed is indicated in the figure legends. P-values were determined by performing unpaired samples t-test with the use of GraphPad Prism 9.0.0. Statistical significance is indicated at *P < 0.05, **P < 0.01, or ***P < 0.001, with n.s. denoting non-significant P ≥ 0.05.

ACKNOWLEDGMENTS
We thank Andy Johnson and Justin Wong of the UBC Flow Cytometry Facility for performing FACS analysis as well as for assistance with flow cytometry. We thank the laboratory staff at the BC Centre for Excellence in HIV/AIDS for processing PBMCs from study participants.