TGS-inducing shRNAs protect Jurkat E6-1 cells from HIV-1SF162 replication
To confirm the shRNAs silencing activity we performed a time-course of infection on Jurkat E6 cells (Fig. 1a, top right) and measured the levels of HIV-1 replication via RT-assays, qRT-PCR of HIV-1 gag-mRNA and qPCR of Alu HIV-1 integrated DNA. Suppression of viral replication by 100% matched shRNAs in this setting, is expected to translate as resistance to reactivation during latency. Previous experiments in our laboratory have demonstrated that stable expression of PromA in MOLT-4 cells provides prolonged and potent protection for over 1 year [21]. Therefore, to further challenge and possibly disrupt RNA-induced epigenetic silencing in cells expressing TGS-inducing shRNAs, we used an extremely high amount of HIV-1SF162 virus, 1125 mU/uL per 3 × 105 cells. This also increased the likelihood of having all the cells infected with at least one provirus. Despite the high virus inoculum, on Day 10 (D10) there was at least 10-fold difference in the levels of RT activity between controls and protected cell lines (Fig. 1b, upper left panel), confirming that PromA, 143 and PromA/143 combined expressing cells were able to specifically repress HIV-1 transcription (Fig. 1b, upper right panel). A/143 showed some variation at D10, but the levels remained within the same range (> 10 μU/mL) as those of PromA and 143. Similarly, HIV-1 gag-mRNA expression was > 1000 times less in PromA, ~ 80 times less in 143, and > 100 times less in PromA/143, compared to the infected parental cell line (Fig. 1b, lower left panel). Also, the levels of integrated HIV-1 DNA were approximately the same in all the cell lines (Fig. 1b, lower right panel). This was not unexpected because the experimental conditions were directed to infect all the cells with at least one viral particle and the shRNAs do not prevent integration of proviral DNA. These data confirmed previous observations that these constructs are able to suppress HIV-1 replication during a robust infection challenge, while 2–3 mismatches in the target shRNA sequence disabled this protective effect.
Protective shRNAs provide resistance to HIV-1 reactivation during LRA treatments
We previously showed that stable expression of shPromA, sh143 or shPromA/143 provided protection from HIV-1 reactivation when challenged with TNF, SAHA or a combination thereof [9]. This, in addition to the strong silencing observed during the time course infection, prompted us to challenge the transduced and Parental J-Lat 9.2 cell lines with a panel of potential LRAs (Fig. 2) to assess if combined expression of the two protective shRNAs resulted in stronger or broader protection to a wider range of stimuli.
We first quantitated the levels of shRNA expression in all JLat 9.2 cell lines to determine whether each shRNA construct was expressed to the same degree. We observed similar shRNA expression across all J-Lat 9.2 cell lines, with no significant differences observed (Fig. 2a). The loop sequence was not present in the scrambled Control construct or the untransduced cell line and was therefore not detected in these cell lines.
After treatment with TNF the protected cells lines showed the lowest levels of GFP+ cells (Fig. 2b) and the lowest levels of GFP expression (MFI) (Fig. 3a), both compared to the parental cell line and specificity controls. This indicates decreased proviral reactivation and decreased transcriptional activity from reactivated proviruses. The highest concentration of TNF in viremic HIV-1 infected patients is 100 pg/mL(0.1 ng/mL), (27), while 5 ng/mL is the highest reported during fatal acute sepsis (37, 62). At 0.1 ng/mL the proportion of GFP+ cells in PromA and PromA/143 cells was significantly less compared to Parental (Fig. 4a, left panel), though, the overall extent of reactivation was extremely low across all cell lines and therefore not significant (Fig. 4a, lower). At 5 ng/mL of TNF, both the percentage of GFP+ cells and levels of GFP expression were significantly lower in PromA, 143 and PromA/143 (p = 0.0005 and p = 0.02, p = 0.003 and p = 0.0005, and p < 0.0001 both, respectively) (Fig. 4b). In addition, cell viability was slightly more affected in PromA and M2, than any of the other cell line (Fig. 5a).
We then examined the effect of SAHA, a pan-histone deacetylase inhibitor (HDACi) [22, 23] which has been extensively studied in vivo as an HIV-1 LRA [3, 4]. A concentration of ~ 0.335 μM has been previously used as an in vitro equivalent of 400 mg, the protein-unbound pharmacological concentration after single dose of Vorinostat(ZOLINZA®) (63–64), also known as SAHA. At ~ 0.56 μM SAHA, the nearest to 0.335 μM tested, we found no significant differences in the % GFP+ cells nor in the levels of GFP expression (Figs. 2c, 3b, 4c). At higher concentrations the protected cell lines showed the lowest proportion of reactivated cells, and these cells had reduced GFP expression, with PromA and A/143 showing ~ 3-fold less compared to Parental (Figs. 2c, 3b). Interestingly, while 143 showed the least proportion of reactivated cells, these cells were expressing GFP at levels that paralleled those of the M2 control cell line, indicating that the sh143 induced-TGS is more susceptible to SAHA reactivation (Figs. 2c, 3b). Overall, SAHA reactivated a greater proportion of cells from the unprotected cell lines compared to TNF and the induced expression was much lower (Figs. 2c, 3b). Additionally, viability was substantially affected at higher concentrations of SAHA (Fig. 5b).
We next combined TNF with SAHA aiming to disrupt silencing via two different reactivation pathways. We decreased the dose-range of SAHA to evaluate the pharmacologically relevant concentration of ~ 0.335 μM in combination with a concentration of TNF that falls within the high range observed in the plasma of HIV-1+ patients (1.5 ng/mL) [24]. Increasing concentrations of the combined treatment SAHA/TNF induced the highest percentage of GFP+ cells in all the cell lines with A/143 showing the lowest (Figs. 2d, 3c, 4d), and were increasingly toxic (Fig. 5c). Although, not significant, PromA, 143 and PromA/143 showed ~ 2 fold lower GFP expression and thus less proviral reactivation compared to Parental, up to supra physiologic concentrations of 3.13 μM for SAHA and 12.5 ng/mL for TNF (Fig. 3c). After this point, GFP expression declined in all the cell lines.
Activation of PKC pathways in conjunction with NF-kB can induce potent reactivation of latent HIV-1. Therefore, we sought to induce activation of different PKC isoforms along with the NF-κB pathway by treating the cells with increasing concentrations of Bryostatin [25] in combination with a fixed concentration of TNF (5 ng/mL). The combined treatment reactivated comparable percentages of GFP+ cells in all the cell lines (Fig. 2e). However, the protected cell lines showed the lowest GFP expression levels revealing impairment of proviral transcription, while the controls revealed increased GFP expression at all concentrations tested (Figs. 2e, 3d). Cell line 143 appeared less susceptible to disruption by the combined treatment than PromA. In addition, cell viability was comparable across all cell lines (Fig. 5d).
In order to obtain insight into the epigenetic profile induced by the protective shRNAs, we investigated the effect of inhibiting the histone lysine methyltransferases (HKMTs), SUV39H1 and EZH2, using Chaetocin [26] and DZNep [27], respectively. These have been previously used to indirectly assess the epigenetic profile of latent HIV-1 [26, 27]. SUV39H1 mediates the trimethylation of lysine 9 in histone 3 (H3K9me3,) whereas EZH2 mediates the trimethylation of lysine 27 in histone 3 (H3K27me3) [28]. Both, Chaetocin and DZNep were not efficient reactivation treatments under the conditions tested. Chaetocin at its highest concentration reactivated a maximum of ~ 40% in the Parental cell line, 143 and PromA, while A/143 showed the least percentage of reactivated cells (Fig. 2f). Expression was only detected in the specificity controls at 25 nM, while the protected cell lines did not show expression at any concentration (Fig. 3e). Chaetocin was highly toxic at its highest concentrations (Fig. 5e). DZNep was the least efficient treatment for reactivation of latent HIV-1, reactivating a maximum of ~ 1% GFP+ cells in PromA and M2 at 100 μM (Fig. 2g). GFP expression was low, but only detected for the specificity controls and the Parental cell lines with peak effect at 25 nM, while undetected in PromA, 143 and PromA/143 across the concentrations tested (Fig. 3f). Viability was mostly affected in PromA (Fig. 5f).
Altogether, the shRNAs showed a differential ability to protect the cells from HIV-1 reactivation depending on the stimuli. The data indicated that protective shRNAs inhibit HIV-1 reactivation at concentrations of LRAs considerably beyond those likely to be relevant in vivo, and that dual expression of shPromA and sh143 generally provided broader protection across a variety of stimuli.
TGS-inducing shRNAs recruit AGO1 and HDAC1, and maintain the epigenetic repressive mark H3K27me3 at the HIV-1 promoter during TNF-induced reactivation
To investigate the epigenetic changes occurring at the HIV-1 promoter during the LRA challenges, we treated all J-Lat 9.2 cell lines for 48 h with 5 ng/mL of TNF; equivalent to the highest pathological TNF concentration reported in human serum during sepsis [29]. We used this higher concentration to induce levels of GFP expression detectable via ChIP assays enabling the comparison of epigenetic marks between reactivation and latency. We performed ChIP assays on sorted live-GFP+ or GFP− cells (Fig. 1a, lower right), and analysed changes in the expression of several markers of heterochromatin associated with the 5′LTR using an ordinary two-way ANOVA to compare the normalized % Input to determine whether the shRNAs modified epigenetic profiles. We use the term “latent” when referring to the sorted GFP− population in which the provirus had not been reactivated, and “reactivated” when referring to the GFP+ population in which the provirus reactivated from latency following TNF treatment.
Significant interactions between the effects of the cell lines (shRNA-transduced or Parental) and the condition (latent or reactivated) were identified in the epigenetic profile of the HIV-1 LTR, for the relative presence of AGO1, HDAC1, H3K27me3, H3K9me2 and H3K9me3, but not H3K9Ac. Simple main effects analyses indicated significant differences within each of the factors, “in between” cell line and condition, for all these epigenetic related marks and proteins (Additional file 1: Table A1). These data indicate that some shRNAs modify the epigenetic profile of HIV-1 during latency, during TNF reactivation or during both. The specific interactions between cell lines and conditions were further identified (See Additional file 1: Table A2 for the Summary of P values) and are explained below.
We first evaluated the changes in the levels of repressive and activating epigenetic marks. “In between” comparisons did not find significant differences in levels of H3K27me3 between any of the transduced cell lines and Parental, during latency (Fig. 6a, left). In contrast, PromA/143 demonstrated significantly higher levels of H3K27me3 during TNF driven HIV-1 reactivation when compared to Parental (Fig. 6a, middle) (p < 0.0001) indicative of an overdrive mechanism maintaining closed chromatin despite the reactivation stimulus. This provides an explanation of how reactivation is limited in these circumstances. In addition, H3K27me3 levels were considerably higher in PromA/143 during reactivation compared to the levels in latency (p < 0.0001) (Fig. 6a, right). In contrast 143 did not show any difference in H3K27me3 levels during reactivation by TNF compared to the parental cell line (Fig. 6a, left and middle), but did show a significant increase of this mark when comparing between latency and TNF reactivation (p = 0.004)(Fig. 6a, right). The levels of H3K27me3 in PromA were not different to Parental during both conditions (Fig. 6a, left and middle), nor post activation compared to latency (Fig. 6a, right).
Only 143_3M had significant lower levels of H3K9me2 in comparison to Parental during HIV-1 latency (p = 0.04) (Fig. 6b, left) and showed a significant decrease during HIV-1 reactivation (p = 0.01) (Fig. 6b, middle). Intriguingly, H3K9me2 was not affected by TNF treatment in any of the other cell lines, except in the 143, in which it showed an increase compared to Parental (p = 0.002) (Fig. 6b, right).
We found no differences in the levels of H3K9me3, across cell lines during HIV-1 latency when compared to Parental (Fig. 6c, left). However, upon TNF stimulation M2, PromA/143 and Control cell lines completely lost this epigenetic mark from the HIV-1 promoter (Fig. 6c, middle). In fact, the levels of this mark were so low for these cell lines that the calculated relative % Input fell below the background and hence the negative values indicate a profound depletion of the epigenetic mark. When comparing the levels of H3K9me3 between HIV-1 latency and HIV-1 reactivation in each cell line, all the transduced cell lines appeared to show a decrease upon treatment with TNF, though this decrease was only significant in M2, PromA/143 and Control (Fig. 6c, right).
Consistent with the more limited interaction of H3K9me3 between the two factors, cell lines and transcriptional conditions, there were no significant differences in the levels of H3K9Ac in any of the conditions, or between latency to reactivation (Fig. 6d, Additional file 1: Table A2). This indicates that acetylation of this residue is not affected by the addition of the shRNAs during latency and that TNF treatment has the same effect on all the cell lines tested.
“In-between” multiple comparisons revealed significantly lower levels of AGO1 at the HIV-1 LTR for the specificity controls M2 (p = 0.01) and 143_3M (p = 0.005) cell lines during HIV-1 latency, when compared to Parental (Fig. 7a, left panel). These differences were not observed for PromA, 143 or PromA/143 (Fig. 7a, left panel). In contrast during reactivation, PromA, 143 and PromA/143 showed significant higher levels of AGO1 at the HIV-1 promoter (all p < 0.0001), when compared to Parental (Fig. 7a, middle). “In-within” multiple comparisons determined that this increase in AGO1 during reactivation was highly significant (p < 0.0001) compared to levels of AGO1 during latency (Fig. 7a, right).
For HDAC1, M2, 143_3M and Control, all showed significant lower levels during latency (p < 0.0001, p = 0.001 and p = 0.0006, respectively), conversely, PromA, 143 and PromA/143, did not (Fig. 7b, left). HDAC1 levels of Control and M2 showed a significant increase from latency to reactivation (Fig. 7b, right) but these levels were no different to those of Parental during TNF reactivation (Fig. 7b, middle); whereas 143 and PromA/143 showed a significant increase during reactivation (both, p < 0.0001) (Fig. 7b, middle), and the magnitude of this increase was significant when compared to the levels during HIV-1 latency (Fig. 7b, right).
Together these data support specific recruitment of AGO1 and HDAC1, in addition to H3K27me3, to the HIV-1 LTR by shPromA, sh143 and shPromA/143 during TNF reactivation conditions and is consistent with these constructs acting specifically at the HIV-1 LTR, to maintain epigenetic repression. The protective constructs appear to induce the relative maintenance of certain repressive epigenetic marks despite the presence of drivers of reactivation (Fig. 7c). These results explain how proviral transcription was impaired in the small percentage of protected cells in which the provirus reactivated.
Maintenance of H3K27me3 is induced by protective shRNAs during reactivation with TNF
The coexistence of H3K4me3 and H3K27me3 in promoter regions is associated with poised or inducible genes (Reviewed in (30)). Their coexistence or bivalency in the HIV-1 promoter may be characteristic of inducible latent proviruses. The ordinary two-way ANOVA did not identify a significant interaction between the cell lines (shRNAs) and these epigenetic marks during latency (Additional file 1: Table A3.). Further, there were no differences in the absolute % Input of H3K4me3 between the transduced cell lines and Parental (Fig. 8a, left and middle). Only 143_3M cell line showed slightly less H3K27me3 compared to Parental cell line (Fig. 8a, middle) (p = 0.02).
However, post hoc Holm Šídák multiple comparisons (Additional file 1: Table A4) found significant higher levels of H3K27me3 compared to H3K4me3 within all the cell lines, during HIV-1 latency (PromA, M2, 143 and A/143: p > 0.0001; Control p = 0.009) (Fig. 8a, right). These data are consistent with the latent state of the provirus in J-Lat 9.2 cells, suggesting H3K27me3 is imposing a strong repressive signal while H3K4me3 allows for transcriptional activation upon stimulation.
Phosphorylation of RNA Pol II at Serine 2 of indicates the presence of an elongating polymerase a the Latent HIV-1 promoter
RNA Pol II can be processive or non-processive depending on the phosphorylation state of, Serine 5 (pSer5) and 2 (pSer2) within the Carboxyl-terminal domain (CTD). The phosphorylation status of these residues is differentially associated with initiation and productive elongation of transcription [30, 31]. Statistical analyses identified an interaction between the cell lines and the phosphorylated species of RNA Pol II (p = 0.02) (Additional file 1: Table A3). The “in-within” cell line multiple comparisons determined a significant higher % Input of RNA Pol II pSer2, compared to that of pSer5, during HIV-1 latency in PromA (p = 0.009), 143 (p < 0.0001), PromA/143 (p < 0.0001) and M2 (p = 0.05) (Fig. 8b, right and Additional file 1: Table A4 top). M2 showed the lowest % Input for both marks (Fig. 8b, left and middle). Post-Hoc Holm Šídák “in-between” cell line comparisons (Additional file 1: Table A4, bottom) determined the levels of RNA Pol II pSer2 were significantly higher in 143 (p = 0.03) and in PromA/143 (p = 0.009), compared to Parental (Fig. 8b, left). No significant differences were identified for RNA Pol II pSer5 (Fig. 8b, middle). These data suggest the TGS mechanism induced by shRNAs PromA, 143 and A/143 possibly involves stalling of the elongating pSer2 RNA Pol II early after transcription initiation, interfering with efficient transcription elongation during reactivation stimuli.