Entinostat up-regulates the CAMP gene encoding LL-37 via activation of STAT3 and HIF-1α transcription factors

Bacterial resistance against classical antibiotics is a growing problem and the development of new antibiotics is limited. Thus, novel alternatives to antibiotics are warranted. Antimicrobial peptides (AMPs) are effector molecules of innate immunity that can be induced by several compounds, including vitamin D and phenyl-butyrate (PBA). Utilizing a luciferase based assay, we recently discovered that the histone deacetylase inhibitor Entinostat is a potent inducer of the CAMP gene encoding the human cathelicidin LL-37. Here we investigate a mechanism for the induction and also find that Entinostat up-regulates human β-defensin 1. Analysis of the CAMP promoter sequence revealed binding sites for the transcription factors STAT3 and HIF-1α. By using short hairpin RNA and selective inhibitors, we found that both transcription factors are involved in Entinostat-induced expression of LL-37. However, only HIF-1α was found to be recruited to the CAMP promoter, suggesting that Entinostat activates STAT3, which promotes transcription of CAMP by increasing the expression of HIF-1α. Finally, we provide in vivo relevance to our findings by showing that Entinostat-elicited LL-37 expression was impaired in macrophages from a patient with a STAT3-mutation. Combined, our findings support a role for STAT3 and HIF-1α in the regulation of LL-37 expression.

Entinostat mediated induction of the CAMP gene is regulated by STAT3. In order to determine which part of the regulatory region of the CAMP gene that might be involved in the Entinostat-mediated induction, in silico analysis of the CAMP promoter was performed. Sequence analysis using the ConREAL web-based algorithm for the identification of conserved transcription factor binding sites revealed a high-score binding site for STAT3 in position -1890. Additionally, Gombart et al. has previously described another STAT3 binding element around position -458 in the CAMP gene promoter 16 , where + 1 ATG is the translation start site (numbers refer to previous mapping of the gene 36 , Fig. 2). To confirm that STAT3 was involved in Entinostat induced CAMP gene transcription, we employed 5,15-diphenyl-21H,23H-porphine (DPP), a selective STAT3 inhibitor. DPP could reduce, but not completely block, the effects of Entinostat, both on the protein level (measured in CampLuc cells, Fig. 3a) and on the mRNA level (in HT-29 cells, Fig. 3b) in a concentration-dependent manner (1 μ M to 50 μ M).
By performing fractionation of HT-29 cell lysates we found that Entinostat induced translocation of STAT3 from the cytoplasm to the nucleus after 3 h stimulation (Fig. 3c).
Next, we examined the effect of the knock-down of STAT3 on the CAMP gene expression. As shown in Fig. 3e, STAT3 silencing by the two shRNA vectors sh3_STAT3 and sh4_STAT3 significantly decreased the Entinostat-elicited CAMP gene induction. The DEFB1 gene expression was down-regulated by sh3_STAT3, but not by sh4_STAT3. In order to study a possible recruitment of STAT3 to the putative binding sites in the promoter region of the CAMP gene, chromatin immunoprecipitation (ChIP) experiments were performed on HT-29 cells stimulated with either vehicle or Entinostat. However, no enriched binding of STAT3 to either binding site in the CAMP gene promoter was detected upon 3 h stimulation with Entinostat (Data not shown).
De novo protein synthesis is required for full inducing effect by Entinostat. We previously reported that the mechanism of PBA-induced CAMP gene involves de novo protein synthesis 14 . To investigate whether this was also the case for Entinostat, we employed the protein synthesis inhibitor cycloheximide (CHX) in HT-29 cells. At 1 μ g/ml of CHX, the Entinostat induced expression of the CAMP gene was significantly attenuated (by 50-60%), suggesting a secondary effect of this compound on the transcription of the CAMP gene (Fig. 4). , and valproic acid (VPA, 0.5-2 mM). Vitamin D 3 (Vit D, 100 nM) was also tested, alone or in combination with PBA (2 mM) or Entinostat (2.5 μ M, black bar). The luciferase activity was assayed, and the results are expressed as luminescence relative to vehicle (ctrl). Graph is representative of at least 3 experiments (Compiled with data from Nylen et al. 19 ). (b) Stimulation of HT-29 cells by Entinostat (250 nM-1 mM), PBA (2 mM) and PBA/Entinostat in combination with Vitamin D 3 (Vit D, 100 nM) at different concentrations was determined for induction of the CAMP gene expression by qRT-PCR. Significantly altered expression is indicated vs ctrl: *p < 0.05; induction vs PBA: # p < 0.05; induction vs Entinostat (Entino): ¤ p < 0.05. (c) Treatment of HT-29 cells with Entinostat (2.5 μ M) at different time-points (6-48 h) was analysed for the induction of CAMP gene expression by qRT-PCR. Significant altered expression vs ctrl is indicated: *p < 0.05. (d) The expression, of the genes DEFB1 and DEFB4 encoding HBD1 and HBD2, was assessed in HT-29 parental cells by qRT-PCR after stimulation with Entinostat (2.5 μ M) or PBA (2 mM). Significantly altered expression vs ctrl is indicated: *p < 0.05. Figure 2. Analysis of transcription factor binding sites on the CAMP gene promoter. The CAMP gene promoter contains putative binding sites for the transcription factors STAT3 and HIF-1α as underlined in bold. Sequence analysis included 3000 bp proximal to the CAMP gene translational start site (+ 1) using the CONserved Regulatory Elements anchored Alignment (CONREAL) algorithm (http://conreal.niob.knaw.nl). Transcriptional start site is indicated as 5´UTR. Additional binding sites that are underlined are adopted from Gombart et al. 16 . As a control experiment, CHX in combination with Vit D was analysed, and the induction of CYP24, a known Vit D responsive gene, which is independent of de novo protein synthesis, was assessed by qRT-PCR ( Supplementary  Fig. S3). A potent increase of CYP24-gene expression was observed, despite treatment with 1 μ g/ml of CHX, ruling out the chance of a non-specific down-regulation of gene expression.

HIF-1α expression and activation is required for CAMP gene induction by Entinostat.
Since the expression of CRAMP, the mouse cathelicidin, depends on HIF-1α 31 , we postulated that Entinostat could regulate the human CAMP gene via this transcription factor. As shown in Fig. 5a, stimulation with Entinostat caused an up-regulation of HIF1A expression on the mRNA level.
Next, we addressed whether increased HIF1A mRNA correlated to enhanced downstream activity. A reporter vector named pTRAF 37 , which allows the imaging of HIF-1α activation via the yellow fluorescent protein YPet in HEK293 cells, was utilized. We found that both PBA (2 mM) and Entinostat (1-25 μ M) caused a robust activation of HIF-1α (Fig. 5b). In Fig. 5c, one representative field of cells stimulated with Entinostat compared to cells incubated with vehicle alone is shown.
In silico analysis of the CAMP promoter (see Fig. 2) revealed one high-score binding site for HIF-1α in position -568 (where + 1ATG is the translation start site). Therefore, ChIP analysis was performed on HT-29 cells to assess the recruitment of HIF-1α to the CAMP gene promoter. As shown in Fig. 5d, the proximal region of the CAMP promoter showed a significant enrichment of bound HIF-1α . As a proof of concept, HIF1A gene expression was down regulated using shRNA, which prevented Entinostat-induced CAMP gene expression (by ~80-90%) (Fig. 5e). Interestingly, STAT3 knock-down with shRNA constructs resulted in a down-regulation of HIF-1α and completely prevented Entinostat-induced increase of HIF1A mRNA (Fig. 5f). STAT3 mRNA levels were not significantly affected by HIF-1α silencing, whereas the two vectors used (sh2_HIF-1α and sh4_HIF-1α ) had opposing effects on DEFB1 gene expression (Fig. 5e). These results implicate that regulation of the CAMP gene in human macrophages, at least in part, involves STAT3 that in turn can regulate HIF1α (Fig. 5g).
In vivo relevance of the STAT3-mediated pathway in the regulation of the CAMP gene. Next, the in vivo relevance of STAT3 in the regulation of the CAMP gene was investigated in immune cells from one patient with hyper-IgE syndrome (HIES). HIES patients bear dominant-negative STAT3 mutations leading to reduced STAT3 signalling, and they suffer from recurrent staphylococcal and candida infections, pneumonia and eczema 38 . We previously found that HIES patients have an impaired release of AMPs in nasal fluid in response to pathogens 39 . Therefore we analysed the response to Entinostat, in terms of CAMP gene induction, in HIES macrophages compared to macrophages from healthy controls. Interestingly, Entinostat induced the CAMP gene in healthy control macrophages, but hardly at all in HIES macrophages (Fig. 6), suggesting a crucial role of STAT3 in the regulation of the CAMP gene in human macrophages.

Discussion
The HDAC inhibitor Entinostat is a potent inducer of the human antimicrobial peptide LL-37 expression 20 , as shown in several cell types, and it works in synergy with the active form of vitamin D. Here, we dissected the molecular basis of this induction and unravelled a complex interplay between the transcription factors STAT3 and HIF-1α in the regulation of the CAMP-gene encoding LL-37. Our data suggest a model where Entinostat activates STAT3, which subsequently leads to activation of HIF-1α and downstream transcription of the CAMP gene.
HDAC inhibitors have been reported to induce AMP-expression. For example, the non-selective HDAC inhibitors trichostatin A and sodium butyrate up-regulate the expression of LL-37 in human airway epithelial-40 , in gastric-, hepatocellular-41 and colon epithelial-cells 13 . The HDAC inhibitor phenylbutyrate (PBA) has previously been shown to up-regulate LL-37 expression in lung epithelial cells and in monocytes 14 . In our luciferase  ChIP assay for HIF-1α recruitment to the CAMP gene promoter was performed in HT-29 cells stimulated for 3 h with or without Entinostat. Significance vs ctrl: *p < 0.05. (e) HIF-1α knockdown was performed in HEK293 cells by transfecting with either a control vector (shSCR) or with shRNA specifically targeting HIF-1α (sh3_ HIF-1α and sh4_ HIF-1α ) transcript. Cells were then stimulated with or without 2.5 μ M Entinostat for 24 h, and analysed for the expression of the genes CAMP, HIF1A, STAT3, and DEFB1 (HBD1) expression by qRT-PCR. Significance vs each respective ctrl: *p < 0.05; significance vs untransfected (untrans) cells stimulated with Entino: § p < 0.05. (f) HEK293 cells were transfected with either a control vector (shSCR) or with shRNA specifically targeting STAT3 (sh3_STAT3 and sh4_STAT3) transcript. Cells were then stimulated with or without 2.5 μ M Entinostat for 24 h, and then analysed for HIF-1α expression by qRT-PCR. Significance vs untransfected (untrans) control (ctrl) cells: *p < 0.05. (g) Representation of a proposed mechanism where Entinostat activates STAT3, possibly by inhibition of deacetylation of the transcription factor, which in turn increase expression of the HIF1A gene and stabilizes the protein product HIF-1α . Induced CAMP gene expression is activated by direct binding of HIF-1α .
reporter assay Entinostat dramatically induced the expression of LL-37 20 . This induction was substantially higher than with other more potent HDAC inhibitors and at a much lower concentration than the positive control PBA 20 . Histone acetylation plays a critical role in the regulation of gene transcription by causing chromatin remodelling and allowing the binding of transcription factors to regulatory DNA-elements 42 . The molecular basis of CAMP induction by PBA is, however, more complex and partly mediated via the vitamin D receptor VDR 43 , but may also depend on de novo protein synthesis 14 . Entinostat -like PBA -is known to be an HDAC inhibitor and thus regulates the expression of many genes. However, we have not observed any clear correlation between documented potency of HDAC inhibition and induction of LL-37 19,20 . Given that Entinostat and several other related compounds was superior to all other HDAC-inhibitors analysed 19 , we hypothesized that additional mechanisms, apart from HDAC inhibition, are involved in Entinostat-mediated regulation of the CAMP gene.
Entinostat also enhanced the transcription of the DEFB1 gene encoding human β -defensin 1 (HBD1), an important peptide of innate defences at epithelial surfaces 44 . However, the DEFB4 gene encoding β -defensin 2 (HBD2) was not affected by Entinostat stimulation. It is plausible that the expression of additional AMPs can be modulated by Entinostat. This possibility is quite appealing, as the induction of an array of AMPs is favourable for the host during infection. Interestingly, AMPs are often co-regulated 5,45 and it is also known that LL-37 is downregulated by several pathogenic bacteria 18,46 . We therefore consider LL-37 as a marker for a healthy epithelial barrier and as a representative for innate effectors. The combination of Entinostat with vitamin D results in a synergistic up-regulation of LL-37. The molecular mechanism behind this synergy has yet to be elucidated. For Entinostat alone, however, Shen et al. 23 showed increased acetylation of STAT3 upon Entinostat challenge. STAT3 in turn is linked to host defence and inflammation, and in silico analyses indicate the presence of STAT3 responsive elements in the promoter of the CAMP gene. Therefore we hypothesized that STAT3 could mediate Entinostat-elicited induction of LL-37 expression in our model. Indeed, blocking STAT3 signalling pathways with either a pharmacological inhibitor or with RNA-silencing strategies reduced LL-37 induction by Entinostat, providing evidence for a critical role of this transcription factor in the induced expression of the CAMP gene. Entinostat activated STAT3 signalling in HT-29 cells, as shown by the enhanced transcription of BCL2, a known STAT3 downstream target gene 47 . We also observed a translocation of STAT3 from the cytoplasm to the nucleus, suggesting that upon Entinostat stimulation, STAT3 is accumulated in the nuclear compartment, where it can bind and activate target genes. In a murine model of infection with Citrobacter rodentium, a pathogen that mimics Escherichia coli infection in humans, the expression of AMPs, such as RegIIIγ and Pla2g2a is dependent on STAT3 activation in the intestine. Furthermore, STAT3 deletion causes increased susceptibility to Citrobacter rodentium infection, with higher bacterial load, severe gut inflammation and dissemination of bacteria to distant organs 48 . Interestingly, we could not detect recruitment of STAT3 to the CAMP promoter, suggesting an indirect role for this transcription factor in the regulation of the CAMP gene. Further, we observed an induction of the gene HIF1A (encoding HIF-1α ) after treatment with Entinostat. HIF-1α is a master regulator of the homeostatic response to hypoxia and activates the transcription of many target genes 33 . HIF-1α activation classically occurs via hypoxia-induced stabilization of the HIF-1α subunit. However, oxygen-independent induction of HIF-1α expression has been documented. Lipopolysaccharide causes HIF-1α accumulation in macrophages through transcriptional and translational activation, in an hypoxia-independent fashion 49 . NFκ B, a key regulator of the immune response to infections, was found to mediate bacteria-elicited increase of HIF-1α mRNA in macrophages 50 . Tumour necrosis factor-α (TNF-α ), another crucial inflammatory mediator, induced HIF-1α expression in macrophages under normoxia 51 , providing another link between inflammation and HIF-1α stabilization in immune cells. T-cell receptor ligation enhanced HIF-1α expression, especially in the pro-inflammatory Th17 cells, via a STAT3 dependent mechanism 52 . STAT3 also mediates IL-6 and TGF-β elicited induction of HIF-1α mRNA 53 . Here we did not observe an involvement of NFκ B in the inducing effect of Entinostat on the CAMP gene expression (Supplementary Fig. S2). On the other hand, STAT3 silencing completely abrogated the up-regulation of HIF-1α mRNA caused by Entinostat, providing evidence for a critical role of STAT3 in the transcriptional regulation of the HIF1A gene. Next, we asked whether increased expression of HIF-1α was reflected by a functional activation of this transcription factor. The results obtained with the pTRAF reporter system clearly demonstrated an activation of HIF-1α by Entinostat, which was detected as increased production of the corresponding fluorescent protein downstream of a HIF-1α binding element in the promoter region of the reporter plasmid. Interestingly, ChIP analyses showed a recruitment of HIF-1α to the CAMP promoter in the proximity of a HIF-1α binding site upon Entinostat stimulation. Thus, the next question to address was whether HIF-1α was needed for LL-37 induction by Entinostat. By using HIF-1α -targeting shRNA we demonstrated that the knock down of this transcription factor significantly reduced the inducing effect of Entinostat on LL-37 expression. We did not observe a clear effect of HIF-1α silencing on DEFB1 gene expression, suggesting a different mechanism for the induction of this gene by Entinostat. Notably, HIF-1α knock down did not affect STAT3 mRNA levels, indicating that in the interplay between these two transcription factors, STAT3 is not transcriptionally regulated by HIF-1α .
The current results are in line with previous reports, suggesting that HIF-1α plays a key role in host innate immunity. HIF-1α deficient macrophages displayed reduced migration and bacterial clearance 29 , and reduced bacterial phagocytosis 30 . Mice lacking myeloid HIF-1α showed higher susceptibility to infection with Streptococcus pyogenes and decreased neutrophil production of granule proteases and the murine cathelicidin peptide CRAMP 31 . Conversely, activation of HIF-1α supported myeloid cell production of host defence factors and improved bactericidal activity 31 . In addition, pharmacological stabilization of HIF-1α enhanced phagocyte-mediated clearance of methicillin-resistant Staphylococcus aureus 54 . Moreover, our results indicate that STAT3 is involved in the regulation of HIF-1α , since silencing of STAT3 by shRNA abrogates Entinostat-mediated induction of HIF-1α mRNA.
To further corroborate an in vivo role for STAT3 in AMP-expression, we obtained cells from a patient with hyper-IgE syndrome (HIES). Interestingly, Entinostat gave little induction of LL-37 in macrophages obtained from this HIES patient, with an impaired STAT3 signalling, compared to control macrophages from two healthy controls. Furthermore, we have previously demonstrated that HIES patients have an impaired release of AMPs in nasal fluid in response to pathogenic bacteria 39 . Therefore our results implicate that STAT3 is involved in the regulation of the CAMP gene in human macrophages.

Materials and Methods
Ethical statement. Written informed consent was obtained from the HIES patient and healthy controls, who were recruited from the Immunodeficiency Unit, Karolinska University Hospital, Stockholm, Sweden. The study was approved by the regional ethical committee in Stockholm (dnr 2011/116-31/4) and all experiments were performed in accordance with the approved guidelines at Karolinska Institutet, Stockholm, Sweden.

Nomenclature of the transcription factor HIF-1. The HIF-1 transcription factor consists of an
α -and a β -subunit. Herein we use primers, shRNA and antibodies raised against the inducible HIF-1α subunit. Throughout the text we therefore use HIF-1α for describing both the dimer and the α -subunit. HIF1A is the name of the gene encoding HIF-1α .
Cell culture. The human colonic epithelial cell line HT-29 was obtained from the American Type Culture Collection: HTB-38 (Rockville, Md., USA) and was cultured in RPMI 1640 (Gibco, Carlsbad, CA, USA), supplemented with 10% Fetal Calf Serum (FCS; Gibco), 100 μ g/ml streptomycin and 100 U/ml penicillin (Invitrogen, Carlsbad, CA, USA), in a 5% carbon dioxide atmosphere at 37 °C. HT-29 cells stably transfected with a plasmid containing the CAMP gene ( henceforth termed MN8CampLuc cells), including the upstream promoter region of approximately 3000 base-pairs (bp) fused with the firefly luciferase gene 19 , were grown under the same conditions as the parental HT-29 cells 19 . The human embryonic kidney 293 cell line (HEK293) was cultured in Dulbecco's modified Eagle's medium (DMEM; Gibco) and Eagle's Minimum Essential Medium (MEM) (ATCC, Manassas, Virginia, USA) supplemented with 10% FBS (GE Healthcare) as well as 100 U/ml penicillin and 100 μ g/ml streptomycin (Biochrom, Berlin, Germany) in a 5% carbon dioxide atmosphere at 37 °C.
Mononuclear cells were isolated from human peripheral blood of healthy volunteers or a hyper-IgE (HIES) patient by density gradient centrifugation with Ficoll-Paque Plus (GE Healthcare, Little Chalfont, UK) and then seeded into 6-well plates in serum free RPMI. After 2 hours, medium was replaced with fresh RPMI containing 10% FCS supplemented with 50 ng/ml macrophage colony-stimulating factor (M-CSF) for differentiation to the M2 macrophage subset. After 7 days, cells were stimulated for 24 h with Entinostat for qRT-PCR experiments.
Luciferase activity assay. MN8CampLuc cells were seeded into 96-well plates in duplicates at a density of 6 × 10 4 cells/well and cultured for 48 h in RPMI growth medium. Cells were then exposed to either vehicle (control) or test compounds for 24 h. Isovaleric acid, Isobutyric acid and valproic acid were obtained from Sigma-Aldrich. Entinostat and Vorinostat were purchased from LC laboratories (Woburn, MA, USA). Following treatment, cells were harvested and lysates were assayed for luciferase activity using the Luciferase assay kit (Promega, Madison, WI, USA) in accordance to manufacturer's instructions. Luminescence was recorded using an Infinite M200 microplate reader (TECAN, Infinite, Männedorf, Switzerland).
Cell lysate fractionation. HT-29 cells were seeded in 10 cm dishes and treated with or without 2.

Chromatin immunoprecipitation (ChIP).
HT-29 cells were grown to 80% confluency in 10 cm culture dishes, treated with 2.5 μ M Entinostat or control and harvested 3 h post treatment. Cells were then cross-linked utilizing 1% formaldehyde for 10 min at room temperature with gentle agitation. Cross-linking reaction was stopped by the addition of glycine to a final concentration of 0.125 M for 5 min at RT. Cells were washed with cold PBS, harvested, and resuspended in 400 μ l lysis buffer (50 mM HEPES pH 8.0, 1 mM EDTA, 0.5 mM EGTA, 140 mM NaCl, 10% glycerol, 0.5% NP-40, 0.25% Triton X-100, and 1 mM PMSF, supplemented with Complete Protease Inhibitor Cocktail tablets) and sonicated by the Bioruptor Plus (Model UCD-300, Diagenode, Liège, Belgium) for 5 cycles of 30 s with a 30 s rest period between pulses. The chromatin-containing supernatants obtained after centrifugation were incubated with 1 μ g of antibodies specific for STAT3 (Santa Cruz Biotechnology) or HIF-1α (BD Biosciences, Franklin Lakes, NJ, USA) overnight at 4 °C. The following day the supernatants were incubated with 50 μ l protein A/G Sepharose (50% slurry; Santa Cruz Biotechnology) under gentle agitation for 2 h at 4 °C. The pellets were then washed twice with 1 ml washing buffer (10 mM Tris-HCl pH 8.0, 1 mM EDTA, 0.5 mM EGTA, 200 mM NaCl, and 1 mM PMSF), three times with washing buffer containing an increased salt concentration (500 mM NaCl), and once again with ordinary washing buffer. After washing, the pellet was resuspended in 110 μ l Tris-EDTA (TE) buffer (10 mM Tris-HCl pH 8.0 and 1 mM EDTA) with 1% SDS and the cross-links were reversed by overnight incubation at 66 °C. Using a PCR purification kit (Promega), DNA was isolated and eluted in 50 μ l elution buffer. Immunoprecipitated samples and total input were analysed by PCR using two sets of primers: ChIP1 primers, designed to amplify a 324 bp region of the CAMP gene promoter, were used to assess recruitment of HIF-1α and STAT3 transcription factors proximal to the CAMP gene. ChIP1 primers were; 5′ -GCCACCGTGCCCTGCCTCATTCATCAATTC-3′ (fw), −440 from + 1ATG; 5′ -GGGTGTGGGCTGGGGTTTGCTTTA-3′ (rev), −116 from + 1 ATG). ChIP2 primers, designed to amplify a 207 bp region of the CAMP promoter, were used to study the distal STAT3 binding site; 5′ -AGCTAGAGCACCAAACAGGG-3′ (fw), -1939 from + 1 ATG ; 5′ -CACGTATGCCCCCATCACAT-3′ (rev), −1732 from + 1 ATG) (see Fig. 2).
HIF-1α activation with the pTRAF reporter vector. Activation of HIF-1α was studied using the pTRAF (plasmid for transcription factor reporter activation based on fluorescence) reporter plasmid according to Johansson et al. 37 . Activation of HIF-1α promotes the expression of YPet by binding to a suitable response element, guiding the expression of this fluorescent protein. The fluorescence can be quantified and is directly correlated to the degree of HIF-1α activation.
Briefly, HEK293 cells were seeded at a density of 18000 cells/well in 96-well plates (Biocoat Collagen I plate) approximately 24 h before transfection with 0.05 μ g DNA, 5 μ l OptiMEM and 0.1 μ l TurboFect (Invitrogen) diluted in 55 μ l medium per well. The DNA mixture (DNA, OptiMEM and TurboFect) was first incubated for 20 min, before the complete medium was added. After approximately 20 h of transfection cells were exposed to Entinostat at different concentrations for 24 h. To prepare the samples for analysis, cells were treated with 40 ng/ml Hoechst for 30 min to stain the nuclei and subsequently fixed in 2% ice-cold paraformaldehyde for 10 min in RT. The fixed cells were covered with PBS and fluorescence was measured using the Operetta ® High Content Imaging System (PerkinElmer). For single cell quantification, seven fields for each cell culture well, covering edges and centre, were recorded in two channels for fluorescence detection of HIF-1α (Ypet) (excitation: 490-510 nm; emission: 520-560 nm) and Hoechst (excitation: 360-400 nm; emission: 410-480 nm). The exposure times were fixed for each channel, with all samples analysed in the same settings. Determinations of fluorescence signals were subsequently performed using the Columbus (PerkinElmer Waltham, MA, USA,) and Excel (Microsoft Redmond, WA, USA) computer programs. Briefly, individual cells were identified based upon the Hoechst staining and the corresponding cytosols were defined using the fluorescent signal that accorded to HIF-1α expression. The intensities of these signals were quantified as total signals within the defined cell area. The resulting single-cell results were subsequently exported from the Columbus software and further analysed using the Excel and GraphPad Prism (GraphPad Software, San Diego, CA, USA) computer programs. To determine accumulated responses on cell culture-level, all single-cell signals within an experiment were summarized and corrected for total cell numbers. For more detailed information, see Johansson et al. 37 .
STAT3 and HIF-1α knock-down experiments. HEK293 cells were seeded in 24-well plates and grown to 80% confluency. Then the cells were transiently transfected with 1 μ g of either a scramble vector (shSCR) or with shRNA vectors specifically targeting STAT3 transcript (sh3_STAT3, TRCN0000329886 and sh4_STAT3, TRCN0000020843) or HIF-1α transcript (sh2_ HIF-1α , TRCN0000003809 and sh4_HIF-1α , TRCN0000003811) (all plasmids were obtained from Sigma-Aldrich). The vectors and 2 μ l of Turbofect were diluted in OptiMEM to a final volume of 100 μ l. Twenty-four hours post-transfection, the medium was replaced and cells were treated with either vehicle or Entinostat for an additional 24 h. The following day, cells were harvested and assayed by qRT-PCR as described above.