Hydrogen sulfide blocks HIV rebound by maintaining mitochondrial bioenergetics and redox homeostasis

  1. Virender Kumar Pal
  2. Ragini Agrawal
  3. Srabanti Rakshit
  4. Pooja Shekar
  5. Diwakar Tumkur Narasimha Murthy
  6. Annapurna Vyakarnam
  7. Amit Singh

  • Curated by eLife

    eLife logo

    Evaluation Summary:

    In the present study Pal and colleagues provide evidence that hydrogen sulfide (H2S) inhibits HIV replication and reactivation by a variety of mechanisms including inhibition of NF-kB and and recruitment of the epigenetic silencer, YY1, to the HIV promoter. They further report that H2S helps to maintaining mitochondrial bioenergetics and redox homeostasis and suggest that inclusion of an H2S donor in current ART regimens may help to achieve a functional HIV-1 cure.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 agreed to share their name with the authors.)

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

A fundamental challenge in human immunodeficiency virus (HIV) eradication is to understand how the virus establishes latency, maintains stable cellular reservoirs, and promotes rebound upon interruption of antiretroviral therapy (ART). Here, we discovered an unexpected role of the ubiquitous gasotransmitter hydrogen sulfide (H 2 S) in HIV latency and reactivation. We show that reactivation of HIV is associated with downregulation of the key H 2 S producing enzyme cystathionine-γ-lyase (CTH) and reduction in endogenous H 2 S. Genetic silencing of CTH disrupts redox homeostasis, impairs mitochondrial function, and remodels the transcriptome of latent cells to trigger HIV reactivation. Chemical complementation of CTH activity using a slow-releasing H 2 S donor, GYY4137, suppressed HIV reactivation and diminished virus replication. Mechanistically, GYY4137 blocked HIV reactivation by inducing the Keap1-Nrf2 pathway, inhibiting NF-κB, and recruiting the epigenetic silencer, YY1, to the HIV promoter. In latently infected CD4 + T cells from ART-suppressed human subjects, GYY4137 in combination with ART prevented viral rebound and improved mitochondrial bioenergetics. Moreover, prolonged exposure to GYY4137 exhibited no adverse influence on proviral content or CD4 + T cell subsets, indicating that diminished viral rebound is due to a loss of transcription rather than a selective loss of infected cells. In summary, this work provides mechanistic insight into H 2 S-mediated suppression of viral rebound and suggests exploration of H 2 S donors to maintain HIV in a latent form.

Article activity feed

  1. eLife

    Author Response

    Reviewer #2 (Public Review):

    (1) Much of the cited literature that is used to make the case for their hypothesis is very old and actually refers to active HIV infection and patient studies prior to ART. Also, the literature they cite regarding the role of H2S as an antimicrobial agent seem to be limited to tuberculosis infection.

    We have revised the list of literature and included more relevant references post- ART era. Recently, the antimicrobial role of H2S is comprehensively examined in the context of tuberculosis. Given the close association of TB with HIV, we thought our study is very timely and essential. However, we would like to point out that the references showing the effect of H2S on infection caused by respiratory viruses are included in the manuscript (7-9). Further, recent findings showing the influence of H2S in the context of SARS-CoV2 infection are also included in the revised manuscript

    (2) The choice of the latently infected model cell lines is rather unfortunate. There are much better defined models out there these days than J1.1 or U1 cells, such as the J-LAT cells from the Verdin lab or the various reporter cell lines generated by Levy and co-workers. In particularly, U1 cells should not be considered as latently infected, as the virus has a defect in the Tat/TAR axis and is mostly just transcriptionally attenuated. It is unclear why the authors only use J-LAT cells for one of the last experiments

    As suggested by the reviewer, we have generated new data using J-LAT cells in the revised manuscript. First, we confirmed that PMA-mediated HIV-1 reactivation in J-LAT cells is associated with the down-regulation of cbs, cth, and mpst transcripts (Figure 1-figure supplement 1C-D in the revised manuscript). Additionally, we have performed several other mechanistic experiments in J-LAT cells to validate the data generated in U1 (see below response to # 3).

    (3) It is further unclear why the authors perform most of the experiments using U1 cells, which are considered promonocytic, but in the end seek to demonstrate the influence of H2S on latent HIV-1 infection in CD4 T cells. Performing all experiments in J1.1 or better J-LAT cells would have seemed more intuitive.

    The choice of U1 was based on our earlier studies showing that U1 cells uniformly recapitulate the association of redox-based mechanisms and mitochondrial bioenergetics with HIV-latency and reactivation (10-12). We have validated key findings of U1 cells in J1.1 and J-Lat cell lines. We genetically and chemically silenced the expression of CTH in J-Lat cells and examined the effect on HIV-1 reactivation. Consistent with U1 and J1.1, genetic silencing of CTH using CTH-specific shRNA (shCTH) reactivated HIV-1 in J-Lat (Figure 2-figure supplement 1F-G in the revised manuscript). Supporting this, pre-treatment of J-Lat with non-toxic concentrations of a well-established CTH inhibitor, propargylglycine (PAG) further stimulated PMA-induced HIV-1 reactivation (Figure 2-figure supplement 1H-I in the revised manuscript). Altogether, using various cell line models of HIV-1 latency, we confirmed that endogenous H2S biogenesis counteracts HIV-1 reactivation.

    (4) The authors suggest that H2S production would control latent HIV-1 infection and reactivation. Regarding the idea that CBS, CTH or possibly MPST would control latent infection as a function of their ability to produce H2S from different sources, there are several questions. First, if H2S is the primary factor, why would the presence of e.g. MPST not compensate for the reduction of CTH? Second, why would J1.1 and U1 cells both host latent HIV-1 infection events, however, their CBS/CTH/MPST composition is completely different? Third, natural variations in CTH expression caused by culture over time are larger than variations caused by PMA activation.

    These questions are important and complex. CBS, CTH, and MPST produce H2S in the sulfur network. CBS and CTH reside in the cytoplasm, whereas MPST is mainly involved in cysteine catabolism and is mitochondrial localized. The lack of compensation of CTH by MPST could be due to the compartmentalization of their activities. Furthermore, CTH and CBS activities are regulated by diverse metabolites, including heme, S-adenosyl methionine (SAM), and nitric oxide/carbon monoxide (NO/CO). In contrast, MPST activity responds to cysteine availability. How substrates/cofactors availability and enzyme choices are regulated in the cellular milieu of J1.1 and U1 is an interesting question for future experimentation.

    Moreover, the tissue-specific expression/activity of CBS and CTH dictates their relative contributions in H2S biogenesis and cellular physiology (13). Some of these factors are likely responsible for differential expression of CBS, CTH, and MPST in J1.1 and U1 cells. Regardless of these concerns, viral reactivation uniformly reduces the expression of CTH in U1, J1.1, and J-Lat. While we cannot completely rule out natural variations in CTH expression over prolonged culturing, in our experimental setup CTH remained stably expressed and consistently showed down-regulation upon PMA treatment as compared to untreated conditions.

    (5) Also, the statement that H2S production as exerted per loss of CTH would control reactivation is not supported by the kinetic data. In latently HIV-1 infected T cell lines or monocytic cell lines, PMA-mediated HIV-1 reactivation at the protein level is usually almost complete after 24 hours, but at this time point the difference between e.g. CTH levels only begins to appear in U1 cells. The data for J1.1. are even less convincing.

    We have performed the kinetics of p24 production and CTH in U1 cells. We showed that the levels of p24 gradually increased from 6 h and kept on increasing till the last time point, i.e., 36 h post-PMA-treatment (Fig. 2D in the revised manuscript). The p24 ELISA detected a similar kinetics of p24 increase in the cell supernatant (Fig. 2E in the revised manuscript). The CTH levels show reduction at 24 h and 36 h. Based on these data, we report that HIV-1 reactivation is associated with diminished biogenesis of endogenous H2S. We have not made any claims that depletion of CTH precedes HIV reactivation. However, our CTH knockdown data clearly showed that diminished expression of CTH reactivates HIV-1 in the absence of PMA, which is consistent with our hypothesis that H2S production is likely to be a critical host component for maintaining viral latency.

    (6) Figure 2F. PMA is known to induce an oxidative stress response, however, in the experiments the data suggest that PMA results in a downregulated oxidative stress response. Maybe the authors could explain this discrepancy with the literature. In fact, both shRNA transductions, scr and CTH-specific seem to result in a lower PMA response.

    In our experiment, PMA treatment for 24 h results in down-regulation of oxidative stress genes. However, the effect of PMA on the oxidative stress responsive genes is time-dependent. In our earlier publication, we showed that 12 h PMA treatment induces oxidative stress responsive genes in U1 cells (12), whereas at 24 h, the expression of genes is down-regulated (10). Genetic silencing of CTH resulted in elevated mitochondrial ROS and GSH imbalance, which is in line with a further decrease in the expression of oxidative stress responsive genes as compared to PMA alone. As a consequence, PMA-treatment of U1-shCTH induced HIV-1 reactivation, which supersedes that stimulated by PMA or shCTH alone.

    (7) Given that the others in subsequent experiments use GYY4137, which is supposed to mimic the increased release of H2S, the authors should have definitely included experiments in which they would overexpress CTH, e.g. by retroviral transduction. Specifically in U1 cells, which seemingly do not express CBS, overexpression of CBS should also result in a suppressed phenotype

    We have explored the role of elevated H2S levels using GY44137. Treatment with GYY4137 suppressed HIV reactivation in multiple cell lines and primary CD4+ T cells. As suggested by the reviewer, overexpression of CTH could be another strategy to validate these findings. However, since the transsulfuration pathway and active methyl cycle are interconnected and share metabolic intermediates (e.g., homocysteine), overexpression of CTH could disturb this balance and may lead to metabolic paralysis. Owing to these potential limitations, we used a slow releasing H2S donor (GYY4137) to chemically complement CTH deficiency during HIV reactivation. We thank the reviewer for this comment.

    (8) Figure 4F: The authors need to explain how they can measure a 4-fold gag RNA expression change in untreated cells. Also, according to Figure 4A, 300 µM GYY produces much less H2S than 5mM, yet the suppressive effect of 300 µM GYY is much higher?

    The four-fold-expression in untreated cells is likely due to leaky control of viral transcription in J1.1 cells (14-16). However, to avoid confusion, we have replotted the results by normalizing the data generated upon PMA mediated HIV reactivation with the PMA untreated cells in the revised manuscript (Figure 4F in the revised manuscript). The suppressive effect of GYY4137 at the lower concentration is intriguing but consistent with the findings that high and low concentrations of H2S have profound and distinct effects on cellular physiology (3,17). One possibility is that the high concentration of H2S induces mitochondrial sulfide oxidation pathway to avert toxicity. This might modulate mitochondrial activity and ROS, resulting in the suppression of GYY4137 effect. Consistent with this, higher concentrations of H2S have been shown to cause pro-oxidant effects, DNA damage and genotoxicity (3,18). We have discussed these possibilities in the revised manuscript

    (9) Initially, the authors argue "that the depletion of CTH could contribute to redox imbalance and mitochondrial dysfunction to promote HIV-1 reactivation"(p. 9). Less CTH would suggest less produced H2S. However, later on in the manuscript they demonstrate that addition of a H2S source (GYY4137) results in the suppression of HIV-1 replication and supposedly HIV-1 reactivation. This is somewhat confusing.

    We show that depletion of endogenous H2S by diminished expression of CTH (U1-shCTH) resulted in higher mitochondrial ROS and GSH/GSSG imbalance. Both of these alterations are known to reactivate HIV-1 and promote replication (10,11,19). The addition of GYY4137 chemically compensated for the diminished expression of CTH, and prevented HIV-1 reactivation in U1-shCTH. These events are expected to suppress HIV-1 replication and reactivation. We have made this distinction clear in the revised manuscript.

    (10) CTH, or for that matter CBS or MPST do not only produce H2S, however, they also are part of other metabolic pathways. It would have been interesting and important to study how these metabolic pathways were affected by the genetic manipulations and also how the increased presence of H2S (GYY4137) would affect the metabolic activity of these enzymes or their expression.

    We fully agree with the reviewer. In fact, our NanoString data show that upon CTH knockdown (U1-shCTH), MPST levels were down-regulated and CBS remained undetectable (Fig. 2F in the revised manuscript). Additionally, GYY4137 treatment induced the expression of CTH but not MPST upon PMA addition (Fig. 5A in the revised manuscript). We have incorporated these findings in the revised manuscript. Given that CBS and CTH catalyzed at least eight H2S generating steps and two cysteine-producing reactions, the modulation of CTH by HIV is likely to have a widespread influence on transsulfuration pathway and active methyl cycle intermediates. Our future strategies are to generate a comprehensive understanding of sulfur metabolism underlying HIV latency and reactivation. These experiments require multiple biochemical and genetic technologies with appropriate controls. We hope that the reviewer would agree with our views that these experiments should be a part of future investigation. We thank the reviewer for this comment.

    (11) H2S has been reported to cause NFkB inhibition by sulfhydration of p65; as such, the findings here are not particularly novel or surprising. Also, H2S induced sulfhydration is rather not targeted to a specific protein, let alone a HIV protein, making this approach a very unlikely alternative to current ART forms.

    We believe that NF-kB inhibition is not the only mechanism by which H2S exerts its influence on HIV latency. Recent studies point towards the importance of the Nrf2-Keap1 axis in sustaining HIV-latency (20). Our data suggest an important role for Nrf2-Keap1 signaling in mediating the influence of H2S on HIV latency. Additionally, recruitment of an epigenetic silencer YY1 is also affected by H2S. Interestingly, YY1 activity is modulated by redox signaling (21), suggesting H2S could be an important regulator of YY1 activity in HIV-infected cells. We have so far, no evidence for viral proteins targeted by H2S. However, experiments to examine global S-persulfidation of host and HIV protein are ongoing in the laboratory to fill this knowledge gap. Lastly, our findings raise the possibility of exploring H2S donors with the current ART (not as an alternate to ART) for reducing virus reactivation. We have tone down the clinical relevance of our findings.

    (12) The description of the primary T cell model used to generate the data in Figure 6 is slightly misleading. Also, the idea of this model was originally to demonstrate that "block and lock" by didehydro-cortistatin is possible. In this application, the authors did not investigate whether GYY4137 would actually induce a HIV "block and lock" over an extended period of time.

    As suggested by the reviewer, we have cited the didehydro-cortistatin studies as the basis of our strategy. Our idea was to adapt the primary T cell model to begin understanding the role of H2S in blocking HIV rebound. Our results indicate the future possibility of investigating GYY4137 to lock HIV in deep latency for an extended period of time. However, comprehensive investigation would require long-term experiments and samples from multiple HIV subjects. In the current pandemic times with overburdened Indian clinical settings, we cannot plan these experiments. However, we hope our data form a solid foundation for HIV researchers to perform extended “block and lock” studies using H2S donors.

    (13) However, the authors never provide evidence that endogenous H2S is altered in latently HIV-1 infected cells (which may actually be an impossible task). By the end of the manuscript, the authors have not provided clear evidence that the effects of e.g. CTH deletion would be mediated by the production of H2S, and not by another function of the enzyme. Similarly, the inability of stimuli to trigger efficient HIV-1 reactivation following the provision of unnaturally high levels of H2S is not surprising given reports on the effect of GYY4137 as anti-inflammatory agent and suppressor NF-kB activation. Unless the authors were to demonstrate a true "block and lock" effect by GYY4137 the data will likely have limited impact on the HIV cure field.

    It's difficult to measure H2S levels in the latently infected primary cells due to the assay's sensitivity and the insufficient number of cells latently infected with HIV-1. However, in the revised manuscript we have clearly shown that cysteine levels are not affected by CTH depletion and cysteine deprivation does not reactivate HIV-1. These results indicate that the effects of CTH depletion are likely mediated by H2S. This is consistent with our data showing that GYY4137 specifically complement CTH deficiency and blocks HIV-1 reactivation in U1-shCTH. Further, we carried in-depth investigation to show that the effect of GYY4137 is not due to impaired activation of CD4+ T cells.

    Lastly, since CTH catalyzed multiple reactions during H2S production, we cannot rule out the effect of other metabolites in this process. However, we think that this is outside the scope of the present study. Our study focuses on understanding of how H2S modulates redox, mitochondrial bioenergetics, and gene expression in the context of HIV latency. These understandings are likely to positively impact future studies exploring the role of H2S on HIV cure.

  2. eLife

    Evaluation Summary:

    In the present study Pal and colleagues provide evidence that hydrogen sulfide (H2S) inhibits HIV replication and reactivation by a variety of mechanisms including inhibition of NF-kB and and recruitment of the epigenetic silencer, YY1, to the HIV promoter. They further report that H2S helps to maintaining mitochondrial bioenergetics and redox homeostasis and suggest that inclusion of an H2S donor in current ART regimens may help to achieve a functional HIV-1 cure.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 agreed to share their name with the authors.)

  3. eLife

    Reviewer #1 (Public Review):

    Hydrogen sulfide (H2S) is best known for it smell of rotten eggs and its high toxicity due to interference with oxygen transport. Recent studies suggest, however, that H2S is involved in various physiological and pathological processes and might exert beneficial therapeutic effects at very low doses. Thus, H2S-releasing compounds are considered for the treatment of e.g. cardiovascular diseases. In the present study, Pal et al. report that H2S also plays an important role in reactivation of HIV from latency. The first show that HIV reactivation reduces the levels of endogenous H2S in cell lines. They further show that suppression of the H2S producing enzyme CTH enhances HIV reactivation and modulates the expression of cellular genes associated with apoptosis and mitochondrial function. The authors also provide evidence that endogenous production of H2S or release from small molecule H2S donors allows U1 and Jurkat cells to maintain normal redox homeostasis and mitochondrial bioenergetics and suggest that this promotes HIV-1 latency. Finally, they provide evidence that the H2S donor GYY4137 modulates Nrf2, NF-kB, and YY1 pathways and suppresses HIV reactivation in latently infected T cells from HIV-infected individuals. The study addresses an important topic from a different angle. It is comprehensive and provides insights into the mechanisms underlying the effect of H2S on HIV latency. In addition, the impact of H2S donors on reactivation of HIV is of significant interest. Limitations of the study are that potential risks and problems associated with H2S as therapeutic agent are not addressed. H2S is highly toxic; thus, dosage will be a major challenge. Treatment would not only affect cells harboring latent HIV but all cells and the H2S-mediated mechanisms proposed to suppress HIV reactivation, such as inhibition of NF-kB and altered metabolism, would be expected to cause side-effects. Finally, the authors propose to combine H2S donors with ART to suppress HIV reactivation. However, effective ART usually prevents viral replication with high efficiency. Thus, rebound after treatment interruption (and not under ART) is the problem. In the end, a functional cure only makes sense if it doesn't need to be combined with ART and does not require daily treatment.

  4. eLife

    Reviewer #2 (Public Review):

    While the idea of H2S acting as a regulator of HIV-1 latency is interesting, there is a long list of inconsistencies that hamper the enthusiasm for this manuscript.

    Much of the cited literature that is used to make the case for their hypothesis is very old and actually refers to active HIV infection and patient studies prior to ART. Also, the literature they cite regarding the role of H2S as an antimicrobial agent seem to be limited to tuberculosis infection.

    The choice of the latently infected model cell lines is rather unfortunate. There are much better defined models out there these days than J1.1 or U1 cells, such as the J-LAT cells from the Verdin lab or the various reporter cell lines generated by Levy and co-workers. In particularly, U1 cells should not be considered as latently infected, as the virus has a defect in the Tat/TAR axis and is mostly just transcriptionally attenuated. It is unclear why the authors only use J-LAT cells for one of the last experiments.

    It is further unclear why the authors perform most of the experiments using U1 cells, which are considered promonocytic, but in the end seek to demonstrate the influence of H2S on latent HIV-1 infection in CD4 T cells. Performing all experiments in J1.1 or better J-LAT cells would have seemed more intuitive.

    The authors suggest that H2S production would control latent HIV-1 infection and reactivation. Regarding the idea that CBS, CTH or possibly MPST would control latent infection as a function of their ability to produce H2S from different sources, there are several questions. First, if H2S is the primary factor, why would the presence of e.g. MPST no compensate for the reduction of CTH? Second, why would J1.1 and U1 cells both host latent HIV-1 infection events, however, their CDB/CTH/MPST composition is completely different? Third, natural variations in CTH expression caused by culture over time are larger than variations caused by PMA activation.

    Also, the statement that H2S production as exerted per loss of CTH would control reactivation is not supported by the kinetic data. In latently HIV-1 infected T cell lines or monocytic cell lines, PMA-mediated HIV-1 reactivation at the protein level is usually almost complete after 24 hours, but at this time point the difference between e.g. CTH levels only begins to appear in U1 cells.
    The data for J1.1. are even less convincing.

    Figure 2F. PMA is known to induce an oxidative stress response, however, in the experiments the data suggest that PMA results in a downregulated oxidative stress response. Maybe the authors could explain this discrepancy with the literature. In fact, both shRNA transductions, scr and CTH-specific seem to result in a lower PMA response.

    Given that the others in subsequent experiments use GYY4137, which is supposed to mimic the increased release of H2S, the authors should have definitely included experiments in which they would overexpress CTH, e.g. by retroviral transduction. Specifically in U1 cells, which seemingly do not express CBS, overexpression of CBS should also result in a suppressed phenotype.

    Figure 4F: The authors need to explain how they can measure a 4-fold gag RNA expression change in untreated cells. Also, according to Figure 4A, 300 µM GYY produces much less H2S than 5mM, yet the suppressive effect of 300 µM GYY is much higher?

    Initially, the authors argue "that the depletion of CTH could contribute to redox imbalance and mitochondrial dysfunction to promote HIV-1 reactivation"(p. 9). Less CTH would suggest less produced H2S. However, later on in the manuscript they demonstrate that addition of a H2S source (GYY4137) results in the suppression of HIV-1 replication and supposedly HIV-1 reactivation. This is somewhat confusing.

    CTH, or for that matter CBS or MPST do not only produce H2S, however, they also are part of other metabolic pathways. It would have been interesting and important to study how these metabolic pathways were affected by the genetic manipulations and also how the increased presence of H2S (GYY4137) would affect the metabolic activity of these enzymes or their expression.

    H2S has been reported to cause NFkB inhibition by sulfhydration of p65; as such, the findings here are not particularly novel or surprising. Also, H2S induced sulfhydration is rather not targeted to a specific protein, let alone a HIV protein, making this approach a very unlikely alternative to current ART forms.

    The description of the primary T cell model used to generate the data in Figure 6 is slightly misleading. Also, the idea of this model was originally to demonstrate that "block and lock" by didehydro-cortistatin is possible. In this application, the authors did not investigate whether GYY4137 would actually induce a HIV "block and lock" over an extended period of time.

    The authors claim that the major conclusion of their study is that HIV-1 reactivation is coupled to depletion of endogenous H2S, which is associated with dysfunctional mitochondrial bioenergetics, in particular suppressed OXPHOS, GSH/GSSG imbalance, and elevated mitoROS. However, the authors never provide evidence that endogenous H2S is altered in latently HIV-1 infected cells (which may actually be an impossible task). By the end of the manuscript, the authors have not provided clear evidence that the effects of e.g. CTH deletion would be mediated by the production of H2S, and not by another function of the enzyme. Similarly, the inability of stimuli to trigger efficient HIV-1 reactivation following the provision of unnaturally high levels of H2S is not surprising given reports on the effect of GYY4137 as anti-inflammatory agent and suppressor NF-kB activation. Unless the authors were to demonstrate a true "block and lock" effect by GYY4137 the data will likely have limited impact on the HIV cure field.

  5. Version published to 10.7554/elife.68487 on eLife
  6. Version published to 10.1101/2021.04.21.440760v1 on bioRxiv