HIV Protease-Generated Casp8p41, When Bound and Inactivated by Bcl2, Is Degraded by the Proteasome

ABSTRACT HIV protease is known to cause cell death, which is dependent upon cleavage of procaspase 8. HIV protease cleavage of procaspase 8 generates Casp8p41, which directly binds Bak with nanomolar affinity, causing Bak activation and consequent cell death. Casp8p41 can also bind Bcl2 with nanomolar affinity, in which case cell death is averted. Central memory CD4 T cells express high levels of Bcl2, possibly explaining why those cells do not die when they reactivate HIV. Here, we determine that the Casp8p41-Bcl2 complex is polyubiquitinated and degraded by the proteasome. Ixazomib, a proteasome inhibitor in clinical use, blocks this pathway, increasing the abundance of Casp8p41 and causing more cells to die in a Casp8p41-dependent manner. IMPORTANCE The Casp8p41 pathway of cell death is unique to HIV-infected cells yet is blocked by Bcl2. Once bound by Bcl2, Casp8p41 is polyubiquitinated and degraded by the proteasome. Proteasome inhibition blocks degradation of Casp8p41, increasing Casp8p41 levels and causing more HIV-infected cells to die.

Multiple pathways have been described by which HIV-infected cells die as a consequence of HIV infection (reviewed in reference 6). One of these pathways is initiated by the intracellular expression of HIV protease, which, contrary to early reports, is catalytically active within the cytosol (7,8). Expression of HIV protease alone in sufficient amounts is enough to kill some eukaryotic cells, and this phenomenon has been exploited to screen for inhibitors of HIV protease (9). The normal function of HIV protease is to cleave Gag-Pol to allow the initial steps of virus packaging. However, due to its degenerate substrate specificity, HIV protease also cleaves a number of host proteins (10)(11)(12). One host protein cleaved by HIV protease is procaspase 8 (13,14); cells expressing a procaspase 8 mutant that is noncleavable by protease do not die following acute HIV infection in vitro (15). Conversely, certain drug resistance mutations in HIV protease impair its ability to cleave procaspase 8, decreasing Casp8p41 (see below) expression, and result in less CD4 T cell apoptosis than wild-type HIV protease (16).
HIV protease cleaves procaspase 8 between phenylalanines at positions 355 and 356, generating a 41-kDa fragment that we have named Casp8p41. Casp8p41 is seen only in HIV-infected cells (14), and Casp8p41 levels are predictive of future CD4 ϩ T cell losses (16)(17)(18). Because Casp8p41 lacks the catalytic cysteine at position 360 of procaspase 8, it is catalytically inert, yet counterintuitively, it maintains the ability to induce cell death. Once generated, Casp8p41 translocates to the mitochondrion, where it adopts a BH3-like alpha-helical domain that binds to the BH3 groove of Bak, causing Bak activation and pore function that leads to loss of mitochondrial transmembrane potential, release of cytochrome c, and activation of downstream executioners of apoptosis (19)(20)(21). Because Casp8p41 is generated at a step in the HIV life cycle that occurs after integration, it follows that reactivation from latency should also generate Casp8p41. Indeed, our recent work, using cells from ART-suppressed HIV-infected patients, has shown that Casp8p41 is generated following HIV reactivation (22). Moreover, those studies showed that the cells in which latent HIV resides (e.g., central memory CD4 T cells) have an apoptosis-resistant phenotype and elevated expression of Bcl2. This led to the additional observations that Casp8p41 can also bind Bcl2 with nanomolar affinity and that, when Casp8p41 directly binds to Bcl2, the HIV-infected cell is not killed (22). Logically, therefore, inhibiting Casp8p41 binding to Bcl2 (thereby allowing Casp8p41 to bind Bak) enhances the number of HIV-infected cells that die (22,23).
In this study, we tested the hypothesis that, after binding to Bcl2, the Casp8p41-Bcl2 complex is degraded and that blocking the degradation of Casp8p41 leads to increased Casp8p41 levels, greater proportions of HIV-infected cells dying, and a decrease in the number of HIV DNA-containing cells.

RESULTS
The Casp8p41-Bcl2 complex is polyubiquitinated. Regulated protein degradation is a prerequisite for cellular physiology, and the ubiquitin (Ub) proteasome system represents a major site of protein turnover. Reversible covalent modification of substrates with polyubiquitin chains drives their association with the proteasome and subsequent destruction. It has previously been documented that a complex of other proapoptotic proteins bound to their inhibitors is subject to proteasome-mediated degradation (e.g., proapoptotic NOXA bound by Mcl1 is degraded by the proteasome) (24). To test the idea that higher-molecular-weight forms of Ub-modified Casp8p41 exist and might also be degraded by the proteasome, immunoprecipitations were performed from cells expressing hemagglutinin (HA)-Casp8p41. Western blotting of the immunoprecipitated material using anti-ubiquitin antibody revealed the existence of higher-molecular-weight species within cells expressing HA-Casp8p41 (Fig. 1A). Furthermore, these higher-molecular-weight species were susceptible to the catalytic activities of a subset of deubiquitinating enzymes, specifically, USP2 and YOD1 ( Fig. 1B  and C). Whether ubiquitin-modified Casp8p41 was the result of Bcl2 binding was next addressed through coimmunoprecipitation and Western blotting using anti-HA and anti-ubiquitin antibodies. Immunoprecipitation of Bcl2 revealed association with HA-Casp8p41 (Fig. 1D, left), and ubiquitin-reactive species that were sensitive to DUB activity were detected regardless of whether Bcl2 or HA-Casp8p41 immunoprecipitation was first performed (Fig. 1D, right). Together, these data indicate that Casp8p41 can be polyubiquitinated. Furthermore, the observation that these species are sensitive to the enzymatic activities of YOD1 and USP2 suggests the ubiquitin chains are linked via lysine 48 (25)(26)(27). This suggests a model in which polyubiquitinated Casp8p41 in complex with Bcl2 is targeted to the proteasome for degradation.
Proteasome inhibition increases Casp8p41 levels. The clinical development of proteasome inhibitors (PIs) has altered the course of protein production diseases, such as multiple myeloma, in which plasma cells produce excessive amounts of monoclonal immunoglobulins. Long-term treatment with PIs over months to years causes semiselective death of myeloma cells (28) and improves the survival of affected patients (29). We tested whether the currently approved PIs bortezomib and ixazomib would block the degradation of Casp8p41. Dose-ranging toxicity studies of bortezomib or the active metabolite of ixazomib (MLN2238, here referred to as ixazomib) were performed in primary CD4 T cells. At doses at or below 10 nM and 100 nM, respectively, which are levels that are achievable in the sera of treated patients (30,31), there was minimal nonspecific toxicity with bortezomib and less toxicity with ixazomib, as determined by active caspase 3 staining ( Fig. 2A), consistent with the superior toxicity profile of ixazomib. Consequently, we used bortezomib at 5 and 10 nM and ixazomib at 50 and 100 nM in future experiments. Jurkat CD4 ϩ T cells were transfected with green fluorescent protein (GFP)-tagged Casp8p41, treated with PIs, and analyzed for GFP-Casp8p41 positivity 6 h later; 10 nM bortezomib increased the percentage of cells positive for GFP-Casp8p41 by 2.5-fold relative to control cells (P ϭ 0.009), and 100 nM ixazomib resulted in a 2.4-fold increase (P ϭ 0.045) ( Fig. 2B and C). This effect was confirmed in primary CD4 T cells infected in vitro with HIV IIIb , treated with bortezomib or control, and assessed for intracellular Casp8p41 positivity using a Casp8p41-specific monoclonal antibody (MAb) (Fig. 2D). Consistent with our previous reports (14,17), Casp8p41 is present in HIV-infected T cells and not in uninfected cells. Furthermore, consistent with proteasome inhibitors increasing GFP-Casp8p41 in transfected cells  (Fig. 2D).

Proteasome inhibitors kill HIV-infected CD4 T cells.
Because Casp8p41 is present only in HIV-infected cells (14,17), increasing its expression might selectively cause the death of those HIV-infected cells as opposed to uninfected cells. This hypothesis was tested using 2 separate models of in vitro HIV infection. First, activated primary CD4 T cells were acutely infected with luciferase-expressing HIV and then treated with dimethyl sulfoxide (DMSO) control, bortezomib, or ixazomib. Bortezomib or ixazomib significantly increased killing of HIV-infected T cell cultures over time (P Ͻ 0.001 each) (Fig. 2E) at doses that did not alter the viability of uninfected primary CD4 T cells ( Fig.  2A). Interestingly, the addition of bortezomib and ixazomib to the HIV-Luc-infected cells initially increased luciferase activity (see below for further discussion), but at later time points, the amount of luciferase activity decreased, consistent with HIV-infected cells dying (P ϭ 0.023 and P ϭ 0.008, respectively) (Fig. 2F). In our second model system, Jurkat T cells infected with GFP-HIV were treated with DMSO or ixazomib, and GFP expression was monitored over time (Fig. 2G). GFP-HIV-infected Jurkat cells treated with ixazomib had more cell death (P Ͻ 0.0001) than cells treated with control DMSO (Fig.  2H), but viability on mock-infected (uninfected) cells was not altered by ixazomib treatment (P ϭ 0.72) (Fig. 2H). As not all cells in HIV-infected cultures are directly infected, we measured death specifically in the GFP-positive cells and found that ixazomib increased the number of dead GFP-positive cells compared to DMSO (P Ͻ 0.0001) (Fig. 2I). Conversely, ixazomib also increased cell death in GFP-negative HIVexposed cells (P Ͻ 0.0001) (Fig. 2J). However, we have previously shown that GFP expression is lost during the process of apoptosis, so it is possible that some of the apparently GFP-negative cells in these cultures were indeed infected (22). Thus, in these experiments, PIs caused minimal toxicity in uninfected and unexposed T cells ( Fig. 2A) yet enhanced killing of HIV-infected and, to a lesser extent, HIV-exposed cells.
Proteasome inhibitors activate the HIV LTR predominantly via NF-B. Above, we observed that ixazomib causes a transient increase in HIV Luc (Fig. 2E) and HIV GFP (Fig. 2F) expression in infected cultures. Since bortezomib has been previously shown to activate NF-B (32) and since the HIV long terminal repeat (LTR) is driven by NF-B, we questioned whether ixazomib might directly activate HIV replication through NF-B. To begin, we used J-Lat 10.6 cells, which express GFP upon HIV reactivation, to assess if ixazomib directly impacts HIV replication. Treatment with ixazomib and with bortezomib led to an increase in GFP expression over time (Fig. 3A) that was coupled with HIV p55 Gag protein expression by Western blot analyses (Fig. 3B).
We next evaluated whether these effects on HIV replication were mediated by NF-B transactivation of the HIV LTR. Luciferase reporter constructs in which either the HIV LTR was linked to luciferase (HIV-Luc) or the HIV LTR missing the NF-B-binding sites was linked to luciferase (HIVΔB-Luc) were transfected into Jurkat T cells, and the transfection efficacy was normalized to that of cotransfected Renilla, as previously described (33). Treatment with bortezomib or ixazomib resulted in a dose-dependent increase in luciferase expression (normalized to that of Renilla luciferase), demonstrating a direct effect on HIV LTR-mediated replication in cells containing HIV-Luc (Fig. 3C). Consistent with this effect being mediated predominantly through NF-B transactivation of the HIV LTR, Luc activity was significantly less augmented in the cells containing the HIVΔB-Luc reporter (Fig. 3D). Internally consistent with these findings, the negative repressor of NF-B, IB, was inactivated (through Ser32 phosphorylation [34]), and the p65 subunit of NF-B was activated (through Ser536 phosphorylation) within 6 h of treating J-Lat 10.6 cells with bortezomib or ixazomib (Fig. 3E).
Selective killing of HIV-infected cells by proteasome inhibitors depends upon Casp8p41. Next, we assessed whether the ixazomib-induced death of HIV-infected cells was dependent upon the presence of Casp8p41. HIV-infected J-Lat 10.6 cells or parental Jurkat CD4 T cells were treated with vehicle control, bortezomib, or ixazomib. Bortezomib and ixazomib increased activated caspase 3 in HIV-infected J-Lat 10.6 cells in a dose-and time-dependent manner (P ϭ 0.014 for 5 nM bortezomib; P ϭ 0.004 for 10 nM bortezomib; P Ͻ 0.0001 for 50 nM and 100 nM ixazomib) ( Fig. 4A and B). The same treatments only minimally increased activated caspase 3 expression in parental, uninfected Jurkat cells ( Fig. 2H and 5D). In the J-Lat cells, bortezomib and ixazomib induced intracellular p24 expression (Fig. 4C, top), cleavage of caspase 3 (Fig. 4C, middle), and cleavage of PARP (Fig. 4C, bottom), consistent with these treated cells undergoing apoptosis.
In order to determine if the increased apoptosis in HIV-infected cells induced by ixazomib was dependent on Casp8p41 expression, we utilized JB-6 cells, which are procaspase 8 deficient and therefore resistant to HIV-induced apoptosis (15). Our strategy exploited our previous observation that procaspase 8-deficient JB-6 cells reconstituted with HIV protease-resistant procaspase 8 (Casp8RN) do not die following HIV infection, whereas JB-6 cells reconstituted with HIV protease-cleavable wild-type procaspase 8 (Casp8) do die from HIV infection (15). JB6 cells were infected with vesicular stomatitis virus (VSV)-G-pseudotyped HIV and 2 days later transfected with enhanced green fluorescent protein (EGFP)-Casp8, EGFP-Casp8RN, or EGFP empty -ector control ( Fig. 4D and E). The cells were then treated with ixazomib or DMSO, and apoptosis (active caspase 3) in infected (p24 ϩ ) cells expressing GFP (a marker of procaspase 8) was measured. Ixazomib increased active caspase 3 expression in HIVinfected (p24 ϩ , -bright, and -dim) cells expressing wild-type procaspase 8 (P ϭ 0.005) (Fig. 4F), but not in HIV-infected cells expressing the mutant, noncleavable procaspase 8 or in cells expressing GFP alone. The same effect was noted when the analysis was restricted to p24-bright cells only (Fig. 4G,   [WT]-Casp8-positive cells, respectively, as opposed to active caspase 3 staining [MFI of 285 in both ixazomib-and control-treated Casp8RN-expressing cells]) in order to decrease the possibility that the p24-positive cells were false positive. These data indicate that the increase in apoptosis induced by ixazomib is dependent upon expression of procaspase 8 and the ability of procaspase 8 to be cleaved by HIV protease.
Proteasome inhibitor treatment kills infected cells more than uninfected cells in coculture. We next evaluated the effects of bortezomib and ixazomib on HIVinfected J-Lat 10.6 cells compared to the parental non-HIV-infected Jurkat cells using a clonotypic assay in which we evaluated cell survival within mixed populations.
First, we confirmed that HIV-infected J-Lat 10.6 cells and uninfected Jurkat T cells have similar propensities to undergo apoptosis in response to non-HIV apoptotic stimuli (CH-11, a Fas agonist, and camptothecin, a topoisomerase inhibitor) (Fig. 5A (Fig. 5D and E).
We have previously shown in primary CD4 ϩ T cells from ART-suppressed patients that reactivation from latency induces Casp8p41 expression (15). Therefore, we postulated that proteasome inhibitor treatment of cells from ART-suppressed HIV-infected donors might (i) reactivate HIV, (ii) stabilize Casp8p41, and (iii) result in the death of more HIV-infected cells than uninfected cells. Since clinical experience indicates that ixazomib does not cause significant leukopenia (35,36) whereas bortezomib is associated with significant gastrointestinal and hematological toxicity (37), we used ixazomib for these studies. To first assess the clinical relevance of our findings, we infected activated primary CD4 T cells with a clinical dual-tropic HIV-1 isolate (HIV-1 RF ), treated with ixazomib or control, and assessed for caspase 3 activation over time (Fig.  5F). Ixazomib increased active caspase 3 expression in HIV-1 RF -infected and -exposed cells (P Ͻ 0.0001), but not in mock-infected cells. This is consistent with our findings in laboratory-adapted HIV-1 strains (compare Fig. 2E and H) and suggests that ixazomib's effect may be clinically relevant.
Given the very low frequency of HIV-infected cells in blood circulation (38), it is not possible to monitor the death of latently infected cells compared to uninfected cells; we therefore analyzed the HIV DNA levels before and after treatment in patient samples. CD4 T cells from 11 HIV-positive ART-suppressed patients were treated with ixazomib or control, and the HIV DNA content was measured by assessing total cell-associated HIV DNA content with digital-droplet PCR (ddPCR) (39). Ixazomib treatment reduced total cell-associated HIV DNA by a median of 35% (interquartile range [IQR], 7% to 42%; P ϭ 0.007) (Fig. 5G, top). To independently verify this result, we treated cells from an additional 8 patients with ixazomib, and an independent laboratory, blinded to sample identification, analyzed the integrated HIV DNA content using a previously validated assay (40,41). Ixazomib treatment also reduced integrated HIV DNA in the second group (median reduction in integrated HIV DNA, 69% [IQR, 58 to 89%]; P Ͻ 0.001) (Fig.  5G, bottom).

DISCUSSION
There is increasing appreciation that the biological processes that favor survival of cancer cells also operate in cells containing the transcriptionally silent reservoir for HIV; examples are epigenetic modifications of DNA and enhanced expression of checkpoint inhibitors (42,43). Following the initial description of the HIV protease Casp8p41 pathway of HIV-infected-cell killing (13), the simultaneous advances in understanding of the control of proapoptotic Bak (44) allowed mechanistic insights into how the catalytically inert Casp8p41 protein might bind and activate Bak to initiate cell death. Indeed, we now know that Casp8p41 adopts an alpha-helical conformation that functions as a BH3-like domain that binds with nanomolar affinity to the BH3 groove in Bak (21), causing the death of HIV-infected cells. As other Bcl2 family members contain BH3 grooves and therefore can bind proteins containing BH3 domains (45), our finding that Casp8p41 binds Bcl2 and Bak with similar affinities (22) provides a mechanistic basis for how cells that generate Casp8p41 might escape the prodeath effects of Casp8p41. Indeed, selectively occupying the BH3 groove in Bcl2 with venetoclax allows Casp8p41 to interact with Bak, and infected-cell death is favored (23). In the current report, we extend this understanding to show that the Casp8p41-Bcl2 complex is polyubiquitinated and targeted to the proteasome for degradation. Blocking this pathway with ixazomib or bortezomib allows Casp8p41 to accumulate, thereby revers-ing the Bcl2 restriction of HIV-infected-T cell death. Also of additional benefit is the previously unreported finding that ixazomib independently drives HIV reactivation, which is, of course, a necessary step in generating Casp8p41 production. At present, it is unknown if Casp8p41 is expressed in HIV-infected macrophages or if proteasome inhibitor treatment of infected macrophages would have the same effect as in T cells. This will be important to study, as macrophages are naturally resistant to HIV-induced cell death and contribute to the HIV-1 reservoir (46).
Our data demonstrate that an oral proteasome inhibitor that is clinically dosed once weekly functions dually to reactivate HIV and to increase Casp8p41 expression, thereby inducing apoptotic death of T cells in which HIV replicates (Fig. 6). Killing is preferential for HIV-infected cells because cell death is dependent in part upon the presence of functional Casp8p41 (Fig. 4F), which is produced only by HIV protease (13,14). It should be noted that cell death is also increased to some degree in ixazomib-treated, GFPnegative Jurkat T cells exposed to HIV-GFP (Fig. 2J), as well as ixazomib-treated, GFP-negative J-Lat cells (Fig. 4A). However, we have previously shown that GFP expression alone does not differentiate infected versus uninfected dying cells, as GFP expression is lost during apoptosis (22). Also, these GFP-negative cells were exposed to HIV; therefore, individual cells, even if uninfected, could be exposed to other deathinducing stimuli, such as FasL, HIV gp120, or HIV Vpr. Therefore, the important experimental controls used were uninfected, unexposed cells.
This preferential, proapoptotic effect of ixazomib causes two surrogate measures of HIV reservoir size (total cell-associated and integrated HIV DNA levels) to decrease following single-dose treatment ex vivo. Acknowledging that the full clinical significance of ex vivo HIV DNA content reductions is unknown, and given that a significant fraction of HIV DNA is replication or otherwise defective (47), additional studies investigating the impact of multidose ixazomib on HIV dynamics using large-volume leukophoresis samples are ongoing via a pilot clinical trial of ixazomib in ARTsuppressed HIV-positive patients (ClinicalTrials.gov identifier NCT02946047).