The anti-proliferative agent hexamethylene bisacetamide (HMBA) belongs to a class of hybrid bipolar compounds developed more than 30 y ago for their ability to induce terminal differentiation of transformed cells. Recently, HMBA has also been shown to trigger HIV transcription from latently infected cells, via a CDK9/HMBA inducible protein-1 dependent process. However, the effect of HMBA on the immune response has not been explored. We observed that pretreatment of human peripheral blood mononuclear cells with HMBA led to a markedly increased production of IL-12 and IFN-γ, but not of TNF-α, IL-6, and IL-8 upon subsequent infection with Burkholderia pseudomallei and Salmonella enterica. HMBA treatment was also associated with better intracellular bacterial control. HMBA significantly improved IL-12p70 production from CD14+ monocytes during infection partly via the induction of type I IFN in these cells, which primed an increased transcription of the p35 subunit of IL-12p70 during infection. HMBA also increased early type I IFN transcription in human monocytic and epithelial cell lines, but this was surprisingly independent of its previously reported effects on positive transcription elongation factor b and HMBA inducible protein-1. Instead, the effect of HMBA was downstream of a calcium influx, and required the pattern recognition receptor and adaptor STING but not cGAS. Our work therefore links the STING-IRF3 axis to enhanced IL-12 production and intracellular bacterial control in primary monocytes. This raises the possibility that HMBA or related small molecules may be explored as therapeutic adjuvants to improve disease outcomes during intracellular bacterial infections.
Burkholderia pseudomallei is a facultative intracellular pathogen and the causative agent of melioidosis, an infectious disease endemic in tropical regions. Treatment of melioidosis is complicated by the intrinsic resistance of the pathogen to many antibiotics. Relapse after the initial infection is common, even after antibiotic therapy (1). Overall mortality ranges from around 15% in Singapore and Australia, to up to 60% in Thailand (2, 3). The closely related yet avirulent B. thailandensis is often used in experimental studies to circumvent logistical difficulties in working with B. pseudomallei, which is designated as a Risk Group 3 pathogen.
Cell-mediated immunity has been shown to be important for the control of melioidosis in animal models (4, 5), and CD4 T cell numbers and IFN-γ production correlate with protection in patients (6–8). We have previously shown that Type 2 diabetic patients, who are at increased risk for developing acute melioidosis and severe disease (9), showed defective IL-12 and IFN-γ production in response to B. pseudomallei infection (10). The production of IL-12 from infected monocytes induced IFN-γ secretion from NK cells, and this in turn activated the microbicidal activity of monocytes to control intracellular bacterial numbers (11). This defective IL-12/IFN-γ axis in diabetic patients was linked to a lower glutathione (GSH) ratio. Although ex vivo addition of free GSH could boost IL-12 production from monocytes upon infection (10), oral supplementation with N-acetylcysteine (a GSH prodrug) in diabetic patients could not reverse the cytokine defect when diabetic peripheral blood mononuclear cells (PBMCs) were infected ex vivo (11). This prompted us to examine whether other molecules have the potential to increase IL-12 production independently of the GSH ratio, during B. pseudomallei infection.
Hexamethylene bisacetamide (HMBA) was discovered in the 1970s as an inducer of terminal differentiation in murine erythroid leukemia cells, and was the lead molecule in a library of polar compounds structurally related to DMSO (12). HMBA is known to cause the growth arrest and differentiation of various transformed cells, a property linked to its bromodomain inhibitory activity (13). After entering phase II trials for acute myelogenous leukemia, the drug was discontinued from further clinical development as the remissions were transient, and the high effective plasma concentrations required led to several side effects in the patients (14). Further development led to the discovery of suberanilohydroxamic acid (SAHA), which was over 100 times more potent than HMBA in inducing terminal differentiation of murine erythroid leukemia cells (15). SAHA was found to possess histone-deacetylase–inhibiting activity, a property not shared by HMBA (16).
Recently, HMBA was found to drive HIV transcription in latently infected cells (17, 18). Increased viral transcription was due to the release of free positive transcription elongation factor B (P-TEFb) from its inactive complex upon HMBA treatment. Free P-TEFb also upregulates expression of the regulatory protein HEXIM-1 by binding to its promoter region (19). Newly synthesized HEXIM-1 protein can negatively regulate P-TEFb activity by directly sequestering the free P-TEFb back into the inactive complex. HEXIM-1 is also capable of inhibiting several other transcription factors including NF-κB, glucocorticoid receptor (GR), and the estrogen receptor α (20–22).
We discovered that HMBA has an unexpected immune adjuvant property, capable of potently increasing IL-12 and IFN-γ production from PBMCs during infection with intracellular bacteria. This effect was independent of the commonly reported molecular mechanisms of HMBA, namely P-TEFb activation and HEXIM-1 protein upregulation. We further elucidate the mechanism underlying this cytokine enhancement in this study.
Materials and Methods
Ethics approval and donor selection
This study received ethics review and approval from the Domain Specific Review Board at the National Healthcare Group, Singapore. Informed written consent was obtained from all participants. Healthy donors aged 18–45 y were recruited for the study, and involved a one-time draw of 40 ml of blood by venipuncture.
Immunoreagents and chemicals
HMBA, SAHA, DMSO, LPS, resiquimod (R848), N-(6-aminohexyl)-5-chloro-1-naphthalenesulfonamide hydrochloride (W-7), BAPTA-AM, nifedipine, DTT, dexamethasone, and Indo-1 AM were purchased from Sigma-Aldrich. Abs for total STAT1, phospho STAT1 (Tyr701), IRF3 (monoclonal, D83B9), phospho-IRF3 (monoclonal, 4947), phospho-p65 (monoclonal, 93H1), and p65 (monoclonal, D14E12) were purchased from Cell Signaling Technologies. The CDK9 inhibitor 4-[(3,5-diamino-[1H]-pyrazol-4-yl)azo]-phenol was purchased from Cayman Chemicals (CAS number 140651-18-9). The following Abs were purchased from the respective sources: blocking Ab for the type I IFN receptor (clone MMHAR-2) and the isotype control from PBL Assay Science; β-actin Ab (monoclonal, A5441) from Sigma-Aldrich; and polyclonal Ab against HEXIM-1 (sc-398479) from Santa Cruz Biotechnology. The MyD88 inhibitory peptide, control peptide, and the cell lines THP-1 dual, THP-1 dual KO-STING, and THP-1 dual KO-cGAS were purchased from InvivoGen.
PBMCs, CD14+ monocytes, or NK cells were isolated as described previously (11), and resuspended in RPMI 1640 + 10% FBS at the typical densities of 2, 0.25, and 0.25 million cells per ml per well respectively. Overnight bacterial cultures were added to cells at the relevant multiplicity of infection (MOI) assuming that 1 OD600 of bacterial culture has ∼1 × 109 CFU per ml. The plates were centrifuged at 250 × g for 5 min and antibiotics added after 1 h of infection. To enumerate intracellular bacteria, at the respective timepoints cells were centrifuged and lysed in 0.1% Triton X-100 in PBS, and serial dilutions plated on Luria-Bertani agar plates. For the two-step infection model, conditioned medium was obtained by filter sterilization of supernatant from B. thailandensis–infected PBMCs, which were pretreated with or without 2.5 mM HMBA for 4 h. The filtrate was diluted 4× in RPMI 1640 + 10% FBS media, and overlaid on to PMA-differentiated THP-1 cells for 16 h. The cells were then washed 1× in PBS, and overlaid with fresh RPMI 1640 + 10% FBS, and infected with stationary phase B. thailandensis at an MOI of 50:1 for 2 or 6 h, as described above. After the indicated time, the monolayers were washed once in 1× PBS, and intracellular bacteria enumerated by plating.
SDS-PAGE and Western blotting
Cells were lysed in RIPA buffer containing 2× protein and phosphatase inhibitors (Pierce). Typically, 0.35 million PMA-differentiated THP-1 cells, or two million PBMCs were lysed in 100 μl of lysis buffer, and then boiled at 95°C for 5 min in the presence of 1× Laemmli sample buffer (Bio-Rad). Lysates were sonicated to facilitate loading, and then resolved on a 10% SDS-PAGE gel before transferring on to a nitrocellulose membrane and immunoblotted. Blots were visualized using ECL Plus Western Blotting Substrate (Thermo Fisher Scientific) on a Bio-Rad ChemiDoc imager, or manually developed using x-ray film.
Assay for calcium flux.
PMA-differentiated THP-1 cells were detached by trypsinization, resuspended in serum-free RPMI 1640, and loaded with 10 μM Indo-1 AM for 30 min at 37°C. The cells were pelleted to remove excess dye, resuspended in complete RPMI 1640, and equilibrated at 37°C for 15 min prior to measurement on a BD LSR Fortessa X-20 flow cytometer. Each sample was first assayed untreated for 60–90 s, stimulated with 2 μM ionomycin or 10 mM HMBA, and then immediately reassayed. The BUV395 filter set (379/28) was used to measure calcium-bound Indo-1, and the BUV496 (515/30) filter set was used to measure free Indo-1. Calcium flux was plotted as the ratio of bound/free Indo-1, and viewed on a time versus ratio plot.
Quantification of gene expression via quantitative RT-PCR
After the appropriate treatment or infection procedure, the cells were lysed with Trizol. In experiments where cells were treated with CDK9 inhibitor, BAPTA-AM, W-7, or nifedipine, the control wells were treated with media containing an equivalent amount of solvent (DMSO/dimethylformamide). Transcripts were quantified after RNA extraction and cDNA synthesis using the iQ SYBR Green Supermix (Bio-Rad), on an iQ5 Thermocycler machine (Bio-Rad). Relative mRNA levels of a test gene in an experimental sample were expressed as a fold change in gene expression, by firstly normalizing to the internal reference gene (18 s) and subsequently normalizing to the expression in the untreated sample, using the 2−∆∆Ct method (23).
Plasmid transfection and small interfering RNA–mediated gene knockdown
HEK293T cells were seeded overnight at a density of 75,000 cells per ml per well in complete DMEM overnight, and then transfected with 100 ng plasmid per well in serum-free media for 6 h using jet-Prime Transfection reagent (Polyplus-Transfections). For the glucocorticoid transcriptional response element (GRE) inhibition assay, after 24 h of transfection with the GRE reporter mix and pFLAG-HEXIM1 or pFLAG-CMV2, cells were lysed with passive lysis buffer and the luciferase content assayed using a Dual Luciferase Reporter Assay System (Promega), according to the manufacturer’s instructions. The luciferase signal was normalized to the Renilla signal from the internal control in each well. Reporter activity was expressed as the fold increase in the luciferase signal relative to the Renilla signal, multiplied by 1000, because the internal Renilla signal was several times stronger than the GRE-driven luciferase signal. For measuring NF-κB promoter activity, pNF-κB–SEAP or pSEAP2-control plasmid (Clontech)–transfected cells were treated with HMBA for 5 h, followed by rTNF-α for 12–15 h. The supernatant was briefly centrifuged and the alkaline phosphatase content was assayed using the Phospha-Light kit (Thermo Fisher Scientific), according to the manufacturer’s instructions. For small interfering RNA (siRNA) knockdown, HEK293T cells were seeded at a density of 60,000 cells per well overnight, and transfected with 100 nM siRNA in 0.4 ml of OptiMEM (Thermo Fisher Scientific). After 6 h, 0.6 ml of OptiMEM + 20% FBS was added to each well. After 24 h of transfection, the cells were treated with 0 or 10 mM HMBA. The cells were lysed in Trizol for transcript quantification after 3–4 h of treatment. RIG-1 siRNA were purchased from GE Dharmacon, whereas the MAVS siRNA were designed in-house.
Statistical analysis was carried out using GraphPad Prism software, version 5. Data were reported as mean ± SD. Statistical significance was determined via either a paired or unpaired t test, as mentioned in the relevant figure legends. Differences were considered significant at p < 0.05.
DMSO and HMBA increases IL-12/ IFN-γ production from PBMCs upon infection
While testing compounds for the ability to enhance the IL-12/IFN-γ axis upon infection with B. pseudomallei, we found that pretreatment of PBMCs with DMSO, a solvent for our many test compounds, enhanced cytokine production (Supplemental Fig. 1A, 1B). However, DMSO has been reported to have multiple cellular and pharmacological effects (24) and had to be used at a relatively high concentration (1% v/v solution of DMSO is 141 mM). Because HMBA and SAHA were developed as differentiation-inducing compounds downstream of DMSO, we tested whether they could similarly induce cytokine production. HMBA increased IL-12 and IFN-γ production upon infection in a dose-dependent manner up to 2.5 mM, whereas SAHA strongly suppressed cytokine production (Supplemental Fig. 1C–F). In uninfected, untreated cells, IL-12 levels were undetectable, whereas IFN-γ levels were <10 pg/ml (detection limit of the ELISA is 1 pg/ml). The concentrations tested (millimolars for HMBA and micromolars for SAHA) were based on the effective concentrations reported in the literature.
We next determined the specificity of cytokine induction by HMBA. The most prominent effect was on IL-12 and IFN-γ (Fig. 1A, 1B). IL-1β levels increased by 2.2-fold but the difference was not statistically significant (Fig. 1C). There were no changes in IL-6 or TNF-α and a significant reduction in IL-8 (Fig. 1D–F).
HMBA enhances the IL-12/IFN-γ axis during B. thailandensis and Salmonella Typhimurium infection
The adjuvant effect of HMBA was also observed with two other intracellular bacteria: B. thailandensis (a surrogate model for B. pseudomallei) and Salmonella enterica serovar Typhimurium. HMBA pretreatment of PBMCs increased IL-12, IFN-γ, and TNF-α production upon B. thailandensis infection by 7.6-, 5.3-, and 1.6-fold, respectively (Fig. 2A–C). HMBA showed a statistically significant enhancement of IL-12 and IFN-γ production during Salmonella infection for all four donors (Fig. 2D, 2E). TNF-α production during Salmonella infection decreased marginally in the presence of HMBA (Fig. 2F).
HMBA suppresses NF-κB activation
NF-κB is a central regulator of the inflammatory response and is directly responsible for the transcriptional expression of several proinflammatory cytokines including TNF-α and IL-8. To activate NF-κB in HEK293T cells, we used TNF-α as the cell line has negligible or very low expression of TLRs. Treatment of HEK293T cells with HMBA inhibited the activation of the NF-κB responsive promoter element upon addition of recombinant TNF-α (Supplemental Fig. 1G). HMBA treatment did not inhibit the expression of a control SEAP plasmid, indicating that HMBA did not inhibit SEAP reporter activity nonspecifically (Supplemental Fig. 1H). In THP-1 cells, HMBA treatment also reduced the phosphorylation of p65 in response to LPS stimulation (Supplemental Fig. 1I). Therefore, HMBA’s effect on IL-12/IFN-γ production is independent of NF-κB activation. Suppression of NF-κB activation and dampening of the general proinflammatory cytokine production could explain the relative specificity of HMBA for IL-12 and IFN-γ enhancement upon infection (Fig. 1).
HMBA selectively enhances IL-12p35 and IFN-γ transcription in CD14+ monocytes postinfection
Because HMBA enhances the production of both IL-12 and IFN-γ, we examined whether HMBA acts on both monocytes and NK cells, the main producers of IL-12 and IFN-γ respectively. CD14+ monocytes pretreated with IFN-γ and HMBA before infection had a 5-fold increase in IL-12p70 production (Fig. 3A). This was similar to the fold increase in IL-12p70 caused by HMBA pretreatment during B. pseudomallei infection of total PBMCs (Fig. 1A). However, IFN-γ production during B. pseudomallei infection of HMBA-pretreated NK cells was significantly less than that in NK cells not treated with HMBA (Fig. 3B), in contrast to the data in total PBMCs where HMBA increased IFN-γ production during infection (Fig. 1B). This shows that during infection of total PBMCs, HMBA primarily acts on monocytes to enhance IL-12p70 production, which then primes NK cells for increased IFN-γ secretion. The IL-12p70 protein is composed of two subunits, p35 and p40. At 12 h postinfection in CD14+ monocytes, the expression of the IL-12p35 gene was over 20-fold higher in the HMBA-treated cells compared with the untreated cells (Fig. 3C). This was in contrast to the IL-12p40 gene, where the increase in HMBA treated cells was <2 fold compared with the untreated cells (Fig. 3D).
HMBA upregulates IFN-β, IFN-stimulated genes, and IRF expression in CD14+ monocytes
We found that pretreatment of cells with HMBA was required for the enhancement of IL-12 during infection to be observed (data not shown). This raised the possibility that HMBA is inducing a priming factor for IL-12. During TLR-induced IL-12p70 production in dendritic cells, an early autocrine type I IFN loop has been reported to be required for the robust induction of the IL-12p35 gene via the activation of IRFs (25). Indeed, we found that treating CD14+ monocytes with HMBA induced the expression of IFN-β by ∼10–20-fold at 6 h posttreatment (Fig. 4A). IFN-β induction is specific, as transcript levels of TNF-α were constant, whereas IL-8 showed a minor but significant reduction upon HMBA treatment (Fig. 4A). We also observed that the expression of two IFN-stimulated genes (ISGs), MxA and IFIT1, was upregulated by 5–10-fold in the presence of HMBA (Fig. 4B). IRFs are activated by phosphorylation downstream of TLR or IFN stimulation. We therefore looked at the expression of IRF1, 7, and 8 upon HMBA treatment of CD14+ monocytes. IRF7 was significantly and consistently upregulated by ∼2.5-fold. IRF1 was upregulated weakly by ∼1.5-fold, whereas the expression of IRF8 was not affected (Fig. 4C).
Autocrine sensing of HMBA-induced IFN-β in primary monocytes.
The upregulation of ISGs and IRF7 suggests that IFN-β induced by HMBA in monocytes is being sensed by cells in an autocrine fashion. Further supporting this model was the observation that treatment of monocytes with HMBA led to the distinct phosphorylation of STAT1, particularly at 10 h posttreatment (Fig. 5A). In monocytes from donor 2, we also observed an upregulation of total STAT1 levels. Autocrine type I IFN sensing has been reported to regulate STAT1 expression as well (26). To inhibit the autocrine sensing of type I IFN, a blocking Ab known to inhibit downstream receptor activation of the type I IFN receptor was used. Stimulation was carried out with the TLR agonist combination of R848 + LPS, which stimulates TLR8 and TLR4 respectively. This agonist combination has been shown to induce IL-12 potently in myeloid cells (27). Blocking the type I IFN receptor reduced IL-12 production upon TLR stimulation in primary monocytes by ∼50% (Fig. 5B). Similarly, HMBA enhancement of IL-12 production during B. thailandensis infection was reduced by 40–50% by blocking the receptor (Fig. 5C). Therefore, an increased type I IFN autocrine loop was at least partially responsible for mediating the effect of HMBA on IL-12 production. Although we showed that both type I IFN and ISGs were upregulated upon HMBA treatment of monocytes, it was unclear if HMBA directly upregulated IFN-β and ISG expression or if this was downstream of type I IFN production. Blocking the type I IFN receptor reduced HMBA-induced IFN-β transcription by ∼40% and completely abrogated the upregulation of the ISGs (Fig. 5D). This strongly suggests that HMBA primarily enhances IFN-β transcription. Autocrine sensing of IFN-β leads to the subsequent upregulation of the ISGs.
Type I IFN induction by HMBA is not dependent on HEXIM-1
In PMA-differentiated THP-1 cells, HMBA treatment increased the expression of both IFN-α and IFN-β genes (Fig. 6A). Expression peaked at ∼4 h, although a significant upregulation was still observed at 8 h posttreatment with 10 mM HMBA (Fig. 6A). HMBA treatment is tightly associated with the induction of HEXIM-1, which can subsequently interact with multiple transcription factors to bring about changes in gene expression. An increase in HEXIM-1 transcript and protein levels was also observed upon HMBA treatment of THP-1 cells (Fig. 6B, 6C). The CDK9 activity of free P-TEFb is required for driving the productive elongation of the HEXIM-1 gene by RNA polymerase II downstream of HMBA addition (19). Short-term treatment of THP-1 cells with a CDK9-specific inhibitor prior to HMBA treatment completely abolished the induction of HEXIM-1, at both the transcript and protein levels (Fig. 6B, 6C), indicating that the inhibition of CDK9 is effective. However, no suppression in the induction of IFN-α and IFN-β by HMBA was observed, demonstrating that the HMBA-mediated induction of type I IFN is independent of the early release of P-TEFb triggered by this compound. This also ruled out a role for newly synthesized HEXIM-1 protein in modulating IFN levels.
HMBA treatment can also phosphorylate existing pools of HEXIM-1, and phosphorylated HEXIM-1 disrupts its association with P-TEFb (28). We hypothesized that the phosphorylated HEXIM-1 from the existing cytoplasmic pool, rather than HEXIM-1 synthesized de novo, could be responsible for the type I IFN induction. To deplete HEXIM-1, we made use of the observation that extended CDK9 inhibition depletes HEXIM-1 protein concentration (29). Indeed, treatment of THP-1 cells with CDK9 inhibitor for 20 h led to a significant reduction in the basal levels of total HEXIM-1 protein (Fig. 6D). As short-term inhibition with CDK9 inhibitor reduced HEXIM-1 transcription and translation, the long-term treatment would have depleted the existing phosphorylated pools. However, long-term treatment did not impair the ability of the CDK9 inhibitor-treated cells to upregulate type I IFN upon HMBA addition (Fig. 6E).
Conversely, we overexpressed HEXIM-1 in human HEK293T cells as it had a higher transfection efficiency than THP-1 cells, to examine if this could increase type I IFN induction. HEK293T cells were responsive to HMBA, and drug treatment increased expression of IFN-α and IFN-β with a corresponding increase in HEXIM-1 expression (Supplemental Fig. 2A). Overexpression of HEXIM-1 increased HEXIM-1 transcript levels (Supplemental Fig. 2B) but did not increase type I IFN transcription (Supplemental Fig. 2C). We verified that the overexpressed FLAG-tagged HEXIM 1 was functional by showing that it could suppress GR-driven transcriptional activity, as had been previously reported (21). Addition of the GR agonist, dexamethasone, to the control GRE-reporter + pFLAG-CMV2–transfected cells led to an increase in the relative luciferase signal by around 21-fold, compared with an increase by 6-fold in the same cells not treated with dexamethasone. However, this was not seen in the GRE-reporter + pFLAG-HEXIM1 transfected cells (Supplemental Fig. 2D). Thus, the overexpressed HEXIM-1 is functional, capable of inhibiting the GR response to dexamethasone as well as suppressing basal GR activity. Taken together, these results show that the HEXIM-1 protein is not linked to type I IFN induction by HMBA.
Type I IFN induction is STING dependent but does not require cGAS
We next investigated the potential for a cytoplasmic pattern recognition receptor (PRR) in driving type I IFN expression. MyD88 inhibition using an inhibitory peptide suppressed IL-6 production by ∼80% upon stimulation with the TLR2 agonist Pam3CSK4, but did not suppress type I IFN induction by HMBA (Supplemental Fig. 3A, 3B). Knockdown of the RNA sensor RIG-I, or the adaptor protein MAVS using siRNA similarly had no effect on type I IFN induction by HMBA (Supplemental Fig. 3C, 3D). However, knocking out STING significantly reduced the ability of HMBA to induce type I IFN in THP-1 cells (Fig. 7A). STING is a PRR that recognizes cyclic di-AMP and cyclic di-GMP from bacteria, and cyclic GAMP from host (30). Recognition of dsDNA by the cytosolic DNA sensor cGAS triggers the production of cyclic GAMP (31). We found that a THP-1 cGAS knockout cell line had no significant difference in type I IFN induction upon HMBA treatment compared with the THP-1 dual wild type cell line (Fig. 7B). Previous studies have reported a link between STING and the IFN response via the unfolded protein response (UPR) induced upon endoplasmic reticulum (ER) stress (32, 33). However, the transcript levels of three UPR-induced genes, XBP-1 (splice variant), CHOP, and BiP, were not significantly enhanced upon HMBA treatment (Supplemental Fig. 4A). Conversely, treatment with two known inducers of ER stress, thapsigargin and DTT, led to a significant increase in the transcript levels of UPR-induced genes but with no significant effect on type I IFN transcription (Supplemental Fig. 4B, 4C). In conclusion, HMBA-mediated type I IFN induction requires the STING receptor, but is independent of cGAS and does not rely on the ER stress response.
Type I IFN induction by HMBA is sensitive to intracellular calcium levels
HMBA has been reported to induce an early calcium influx in cells (34, 35). We found that in PMA-differentiated THP-1 cells, HMBA induced a small but rapid calcium influx as measured via ratiometric flow cytometry using the Indo-1 dye (Supplemental Fig. 4D, 4E). Blocking influx of extracellular calcium ions using either nifedipine (a calcium channel antagonist), or EGTA (a quencher of extracellular calcium ions) did not affect the induction of type I IFN genes by HMBA (Supplemental Fig. 4F). However, a rise in intracellular calcium levels could be via influx from either the extracellular medium, or via the release of calcium from intracellular ER stores, or both. To quench intracellular calcium ion levels irrespective of their source of origin, the cell-permeant calcium quencher BAPTA-AM was used. Pretreatment with BAPTA-AM completely abolished the induction of IFN-α, IFN-β, and HEXIM-1 by HMBA (Fig. 7C). The inhibition of HEXIM-1 expression was used as an indication of the effectiveness of calcium inhibition, as P-TEFb kinase activity has been reported to be calcium sensitive, requiring phosphorylation of the CDK9 T-loop by calcium/calmodulin (CaM) dependent kinase 1D (36). We next investigated if type I IFN induction by HMBA is also dependent on CaM. CaM is an intracellular calcium sensor employed by several effector molecules to modulate their activity in response to changes in intracellular calcium levels (37). Treating THP-1 cells with the CaM inhibitor W-7 led to a significant reduction in the IFN-α and IFN-β transcript levels by 80 and 70%, respectively (Fig. 7D). We then examined the impact of HMBA treatment on the activation of IRF3, because it is a key transcription factor involved in type I IFN production. HMBA treatment increased the phosphorylation of IRF3 at the serine 396 site within 1 h of treatment (Fig. 7E). Treatment with BAPTA-AM prior to HMBA treatment completely abolished phosphorylation (Fig. 7F). For inhibition experiments with BAPTA-AM and W-7, the maximum assay duration was 4 h, during which time we did not observe any cell death as measured by the lactate dehydrogenase release assay.
HMBA treatment enhances intracellular bacterial control
We had previously linked the IL-12 and IFN-γ response to improved intracellular bacterial load in a PBMC model of infection (10). Therefore, we explored the effect of HMBA treatment on bacterial control. HMBA pretreatment improves B. pseudomallei intracellular bacterial control from 67- to 224-fold reductions without HMBA treatment to 143- and 452-fold reductions in bacterial numbers in two donor PBMCs at 6 h postinfection (Fig. 8A). B. pseudomallei infection of PBMCs is acute and can only be examined at short timepoints due to host cell lysis. We also examined infection with another intracellular bacteria S. enterica Typhimurium over a longer period of infection. Pretreatment of PBMCs with HMBA reduced intracellular bacterial counts at 24 h postinfection by ∼50%, as compared with untreated, infected cells (Fig. 8B). This effect was not due to increased lysis of infected cells in the presence of HMBA, as the number of viable cells at 24 h postinfection was not significantly different between the two treatment groups (data not shown).
During an in vivo infection, the cytokines produced upon bacterial encounter by the early responding cells serve to prime the second wave of infiltrating immune cells. To model a possible role for HMBA in bacterial control by the infiltrating immune cells, we established a two-step infection model with B. thailandensis. PBMCs were pretreated with or without HMBA, and then infected with B. thailandensis (first infection). The spent supernatant (containing cytokines) was then filter-sterilized to remove all bacteria and used to prime PMA-differentiated THP-1 cells. These primed cells were then infected with B. thailandensis for 6 h (second infection). Exposure of THP-1 cells to HMBA-treated, infected PBMC supernatants improved intracellular bacterial control by 75–90%, compared with priming with infected PBMC supernatants without HMBA treatment (Fig. 8C).
The cytokines IL-12 and IFN-γ play a central role in the development and co-ordination of a robust immune response to intracellular bacteria. IL-12 is mainly produced from infected myeloid cells, and can enhance the production of IFN-γ from NK and T cells, and the differentiation of CD4 T cells into Th1 cells (38). IFN-γ acts on phagocytes to enhance microbicidal activity and Ag presentation, and is a priming signal for IL-12 production (39). In this study, we identify a novel role for HMBA in improving the IL-12/IFN-γ axis in response to intracellular bacterial infections. This enhancement was most prominent during infection with B. pseudomallei and B. thailandensis, but was also observed during infection with S. Typhimurium. HMBA acts on monocytes to increase the transcription of the p35 subunit of IL-12p70. which is often rate limiting for the production of bioactive IL-12 (40). This was in part due to its previously unreported ability to upregulate type I IFN production in a variety of different human cell types. Surprisingly, HMBA is unable to induce type I IFN in both mouse primary cells and cell lines that we tested, although these cells responded to HMBA by upregulating HEXIM-1 (data not shown). Correspondingly, we could not detect increased IL-12 production in HMBA-treated, mouse bone marrow–derived macrophages after bacterial infection. This shows that the drug has different effects in mice versus humans, and suggests that the immune modulatory role of HMBA we have uncovered is distinct from its anticancer property, which acts on both mice and humans. We also found that the mechanism behind the induction of type I IFN by HMBA is distinct from that driving HIV transcription in latently infected cells (28).
Instead, we identified that the induction of type I IFN by HMBA requires calcium signaling and the ER-resident STING adaptor. Early studies reported that HMBA triggers a calcium influx in murine erythroleukemia cells that drives the differentiation of these cells (34, 35). Because BAPTA-AM treatment completely inhibited type I IFN induction by HMBA, but inhibiting extracellular calcium alone (by nifedipine or EGTA) did not, our results suggest that the Ca2+ influx is originating, at least in part, from the ER stores. This is in accordance with a study that reported HMBA triggers a release of calcium, in part from the inositol 1,4,5-trisphosphate–sensitive intracellular Ca2+ stores (41).
Our findings join a series of recent studies reporting that type I IFNs can be induced in a STING-dependent manner in the absence of canonical PRR ligands. Nonpathogen-derived stimuli such as thapsigargin (an inhibitor of the ER Ca2+ ATPase) and alcohol (in hepatocytes of alcoholic liver disease patients) can induce ER stress, followed by downstream activation of the STING-TBK1-IRF3 axis (32, 33). However, we did not find evidence of any ER stress-induced HMBA. It has also been reported that unscheduled membrane fusion, for example between the viral envelope and the host cell membrane, can trigger the production of type I IFN and ISGs through STING (42). It is not clear how membrane fusion translates to STING activation, but interestingly, this activation was found to be cGAS independent (42, 43). Similarly, it is presently unclear how HMBA triggers STING. HMBA bears little resemblance to the cyclic nucleotides that act as the physiological ligands, and thus activation is likely via an indirect process.
Recent studies have demonstrated that the production of type I IFN, which was traditionally associated with viral infections, is a commonly elicited pathway during intracellular bacterial infections as well. Although several TLRs are capable of inducing type I IFN via the TRIF or MyD88 pathway (TLR3, 4, 7, and 9), the majority of type I IFN induced during these intracellular bacterial infections is reported to originate upon cytosolic PRR engagement (44). Type I IFN production can be deleterious to the host in the context of bacterial infections, as it can be associated with a suppression of the innate immune response (45). For example, type I IFN production during Listeria monocytogenes and Mycobacterium tuberculosis infection in mice was linked to a suppression of IL-12 and TNF-α production (46, 47). In contrast, we observe that the production and autocrine sensing of type I IFN plays an important role in robust IL-12 production in monocytes, in agreement with what has been observed in dendritic cells (25). A possible explanation for this discrepancy could lie in the level of type I IFN induced. In our study, HMBA induced type I IFN production in CD14+ monocytes by only 10–20-fold (Fig. 4A). Exposure to a subthreshold level of type I IFN has been reported to prime cells and improve their response to stimulation with other cytokines, particularly IFN-γ by favoring STAT1 activation (26, 48). On the other hand, high levels of type I IFN can dampen myeloid cell responsiveness to IFN-γ by downregulating the type II IFN receptor expression (44).
Thus, we present a model where HMBA drives the phosphorylation of IRF3 in a calcium/CaM-sensitive and STING-dependent manner. Initial IRF3 activation leads to the production of IFN-β in primary monocytes, which is then sensed in an autocrine fashion via the type I IFN receptor. This autocrine sensing primes the cells for improved IL-12p70 production via the upregulation of IRFs, which can then be activated during infection and drive further IL-12p35 transcription (49). The IL-12p35 gene promoter contains an IRF-binding element (50, 51). In particular, this site is critical for IRF-1–mediated IL-12p35 transcription (50, 52). An upregulation in IRF1 transcription was indeed observed within primary monocytes upon HMBA treatment in our work. In contrast, the ability of HMBA to directly phosphorylate IRF3 is one explanation of why blocking the type I IFN receptor only inhibited IL-12 production by ∼40–50%. The activated IRF3 can directly bind to, and enhance transcription of, the IL-12p35 gene during infection (53).
The suppression of NF-κB activation by HMBA could explain the striking specificity for IL-12/IFN-γ enhancement observed in our study. This suppression has been attributed to the P-TEFb–dependent upregulation of HEXIM1 (19, 22). IFN-γ stimulation of macrophages is well known to enhance TLR-mediated and TNF-α–mediated NF-κB activation (54, 55). Activation of NF-κB can in turn induce the transcription of various proinflammatory cytokines and chemokines, including TNF-α, IL-1, IL-6, and IL-8 (56). In the absence of a suppressive effect of HMBA on NF-κB activation, enhanced IFN-γ production in our PBMC model of infection would have resulted in a pan-proinflammatory cytokine enhancement.
The use of adjunctive therapies during infection has attracted increasing attention in recent years, in part triggered by the global epidemic of antibiotic resistance, and in part driven by a paucity of novel antimicrobial agents in the clinical trial pipeline (57). As opposed to traditional antimicrobial therapies that target the pathogen, enhancing the host immune response is less likely to lead to the development of antibiotic resistance. Such adjunctive therapies are also more likely to be applicable to a broader range of pathogens, as the host immune system often employs a conserved set of responses against distinct classes of pathogens. HMBA is not an attractive adjuvant for further development because of the relatively high concentrations needed, poor pharmacokinetics, and reported side effects. However, screening of existing libraries of polar hybrid compounds, as well as libraries of HMBA derivatives (58, 59) may yield hits with superior immune adjuvant activity and better pharmacokinetic profiles. Therapeutic compounds that have been reported to enhance type I IFN production during viral infection can also be re-examined for the ability to favorably modify the immune response via enhanced IL-12 production during intracellular bacterial infections (60, 61).
We thank Deepa Rajagopalan and Sudhakar Jha for designing siRNA for knockdown of MAVS and for helpful discussion. We are grateful to Fatimah Bte Mustafa from the Department of Microbiology for phlebotomy assistance, and Paul Hutchinson from the Flow Cytometry Laboratory, Centre for Life-Sciences, for assistance with flow cytometry. We also acknowledge support by the Yong Loo Lin School of Medicine BSL-3 Core Facility, National University of Singapore, and National University Health System.
This work was supported by Singapore National Medical Research Council NMRC/CBRG12nov035 and NMRC/CG/013/2013. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
The online version of this article contains supplemental material.
Abbreviations used in this article:
glucocorticoid transcriptional response element
multiplicity of infection
peripheral blood mononuclear cell
pattern recognition receptor
positive transcription elongation factor B
small interfering RNA
unfolded protein response
The authors have no financial conflicts of interest.