Novel aroylated phenylenediamine compounds enhance antimicrobial defense and maintain airway epithelial barrier integrity

Aroylated phenylenediamines (APDs) are novel inducers of innate immunity enhancing cathelicidin gene expression in human bronchial epithelial cell lines. Here we present two newly developed APDs and aimed at defining the response and signaling pathways for these compounds with reference to innate immunity and antimicrobial peptide (AMP) expression. Induction was initially defined with respect to dose and time and compared with the APD Entinostat (MS-275). The induction applies to several innate immunity effectors, indicating that APDs trigger a broad spectrum of antimicrobial responses. The bactericidal effect was shown in an infection model against Pseudomonas aeruginosa by estimating bacteria entering cells. Treatment with a selected APD counteracted Pseudomonas mediated disruption of epithelial integrity. This double action by inducing AMPs and enhancing epithelial integrity for one APD compound is unique and taken as a positive indication for host directed therapy (HDT). The APD effects are mediated through Signal transducer and activator of transcription 3 (STAT3) activation. Utilization of induced innate immunity to fight infections can reduce antibiotic usage, might be effective against multidrug resistant bacteria and is in line with improved stewardship in healthcare.

Enterobacter spp.), where some of these bacteria are opportunistic pathogens 18,19 . Specifically P. aeruginosa and S. aureus are pathogens causing respiratory tract infections that can be life-threatening for immunocompromised patients especially for those suffering from cystic fibrosis 20 . Therefore alternative therapies to treat infections are urgently needed 18,19 . The induction of endogenous AMPs could be an effective way of treating infections because many MDR strains are susceptible to different AMPs. Several different compounds inducing expression of AMPs to boost innate immunity have been shown effective in animal models and clinical trials for treatment of infectious diseases, e.g. pulmonary tuberculosis 21,22 . Vitamin D3 is a direct inducer of the CAMP gene expression, the gene encoding the antimicrobial peptide LL-37 [23][24][25] . Another potent inducer is phenylbutyrate (PBA), a short chain fatty acid derivative and also a histone deacetylase inhibitor (HDACi) 26 . Interestingly, PBA treatment of Shigella-infected rabbits resulted in clearance of Shigella infection and counteracted the suppression of rabbit cathelicidin  in the gut and lung epithelium 27 . However, PBA has a fast turnover and is converted into phenylacetate by β-oxidation 28 , therefore high doses of PBA are needed to induce AMPs expression in vitro and in vivo. Additional potent CAMP gene inducers described recently are Entinostat and derivatives designated aroylated phenylenediamines (APDs) 29,30 . It has been shown that Entinostat stimulates CAMP gene expression via activation of STAT3 and HIF-1α transcription factors in human colonic epithelial cells 29 . Moreover, oral treatment of Shigella-and Vibrio cholera-infected rabbits with Entinostat improved their survival and restored production of the rabbit cathelicidin CAP-18 in gut epithelial surfaces 30,31 . Entinostat is an HDACi undergoing clinical trials as adjunctive cancer therapy 32 . However, Entinostat has a documented cytotoxicity 33,34 .
In this study we tested if new APDs, designated HO53 and HO56 could stimulate innate immunity responses in airway epithelial cells by enhancing the expression of endogenous AMPs and if that response was effective against the respiratory pathogen Pseudomonas aeruginosa PAO1 strain. We used bronchial epithelial cell lines, exhibiting a basal-like character and with the ability to differentiate towards polarized bronchial epithelium during air-liquid interface culture (ALI). In human bronchial epithelial cell lines, the new APDs markedly induced expression of the CAMP gene (encoding cathelicidin pro-LL-37/LL-37) both in monolayer and in ALI. The CAMP gene served as the reference, but also induction of other innate immunity genes involved in the defense against infections was observed. In the infection model with pretreatment of bronchial epithelial cells with the APDs significantly reduced the number of intracellular bacteria without exhibiting direct antibiotic properties. We could also demonstrate that treatment with one APD (HO53) of ALI cells counteracted the disruptive effect of P. aeruginosa conditioned medium by maintaining the epithelial barrier integrity. Utilizing a specific inhibitor, we showed that STAT3 transcription factor was involved in the HO53 mediated CAMP induction. Taken together, the current study might open up possibilities for using APDs as novel innate immunity modulators for host directed therapy (HDT) of infectious diseases.

HO53 and HO56 induce CAMP gene expression in bronchial epithelial cell lines (BCi and VA10).
Entinostat has been confirmed as a potent inducer of AMPs, with effects against bacterial infections in animal models 30,31 , but is known to possess cytotoxic properties 33,34 and has limited solubility in aqueous solutions. Based on the structure activity relations found in the first studies on APDs 30 , we started to optimize the AMP-inducing aroylated phenylenediamines (APDs) by designing and synthesizing new alternative compounds. The criterion was to reduce toxicity, while retaining efficient induction of AMPs and the design was based on making more hydrophilic APDs. Using the previously described luciferase reporter HT29 colonic cell line for expression-analysis of the antimicrobial peptide LL-37 35 , we identified, among the novel APDs, HO53 and HO56 ( Fig. 1a; Supplementary Figs S1 and S2; Supplementary Methods) as interesting AMP-inducers with high activity but reduced toxicity.
The present study is focused on the effect of the two new compounds HO53 and HO56 (Fig. 1a) and with Entinostat as comparison. HO53 and HO56 enhanced CAMP gene expression in BCi-NS1.1 cells (BCi) at 24 h post treatment in a dose dependent way ( Fig. 1b; Supplementary Fig. S3b and S3c). Entinostat also induced CAMP gene expression in BCi cells at 24 h but in contrast to HO53 and HO56, the induction was not dose dependent ( Fig. 1b; Supplementary Fig. S3a). A similar pattern of induction with HO53 and HO56 was observed for the bronchial epithelial cell line VA10 (Supplementary Fig. S4). A broad range of HO53 and HO56 concentrations (2.5-250 µM) (Supplementary Fig. S3b and S3c) was tested and for further experiments we selected 75 µM for both compounds. These concentrations represented a low dose that significantly induced CAMP gene expression and had low effect on cytotoxicity and proliferation of BCi cells as compared to Entinostat, but comparable to PBA (Supplementary Fig. S5a and S5b). We used lower concentrations of Entinostat (2.5-50 µM) to keep the DMSO (solvent) concentration in the cell culture medium lower than 1% (v/v). However, there was no difference in the CAMP gene induction between the various concentrations of Entinostat in BCi cells and therefore we decided to use 10 µM in following experiments because of low cytotoxicity ( Supplementary Fig. S5).
Next, we monitored CAMP gene expression over time (0-72 h) with the selected concentration of Entinostat (10 µM), HO53 and HO56 (both 75 µM) (Fig. 1c-e). Notably, the CAMP gene expression was increased significantly after 12 h and 24 h of treatment with HO56 and HO53, respectively and reached maximum expression at 24 h that was maintained up to 48 h post-treatment, but declined at 72 h. A similar effect was observed for Entinostat, where the maximal fold induction was after 24 h and lasted until 48 h, but declined at 72 h. Furthermore, we tested cooperation of HO53 and HO56 with the known CAMP gene inducers PBA (2 mM) and vitamin D3 (1α,25-dihydroxyvitamin D3; 100 nM) (Fig. 1f). We observed a synergistic effect between vitamin D3 and the two compounds at 24 h. However, no cooperation was observed upon treatment with PBA and HO53 or HO56 (Fig. 1f). Consistent with these findings, the synergistic effect after co-treatment with vitamin D3 was also reflected at the protein level by Western blot analyzes, where higher amount of pro-LL-37 was secreted to the cell culture medium (Fig. 1g). HO53 and HO56 separately and in combination with PBA did not affect the pro-LL-37 experiments ± SEM; statistical significance was calculated in comparison to HO53 (red) and HO56 (black) using two-way ANOVA with Dunnett's multiple comparisons test. (g) Concentrated culture medium after 24 h stimulation was used for Western blot analyses using monoclonal antibody against LL-37. Positive control included 2 ng of synthetic LL-37 peptide. GAPDH analyzed in cell lysates was used as a loading control. The representative Western blot is one of 3 independent experiments. The full-length blots are presented in Supplementary Figure S10a. For all qRT-PCR experiments the CAMP gene expression was normalized to TUBB (tubulin-β) reference gene and presented as fold change of the expression compared to control cells (Ctrl) set as 1. Symbols for statistical analysis indicate p-values *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. www.nature.com/scientificreports www.nature.com/scientificreports/ with Entinostat, HO53 and HO56, the expression of LCN2 at mRNA level was significantly increased (approximately 20-30 times) (Fig. 2a). The analyses at protein level revealed low induction of lipocalin 2 (also called Neutrophil gelatinase-associated lipocalin, NGAL) by Entinostat at 24 and 48 h and the representative relative NGAL/GAPDH ratio normalized to control was 1.85 and 1.45, respectively ( Fig. 2b and Supplementary Table S2). In contrast to Entinostat, the induction of lipocalin 2 with HO53 and HO56 was prominent at 24 h (the representative relative NGAL/GAPDH ratio normalized to control was 6.56 and 6.98 respectively) and decreased at 48 h (2.24 and 2.23 for HO53 and HO56, respectively) ( Fig. 2b and Supplementary Table S2). HBD1 encoding the human β-defensin-1 antimicrobial peptide 37 was significantly induced by Entinostat, HO53 and HO56 at mRNA level (Fig. 2c). Significant induction of human β-defensin-1 antimicrobial peptide at the protein level analyzed by ELISA was observed at 48 h (Fig. 2d). In addition, the S100A8 gene encoding a unit of calprotectin antimicrobial protein 38 was significantly induced by Entinostat (approximately 10 times) after 24 h stimulation, whereas induction with HO53 and HO56 was not significant (Fig. 2e). In contrast, the expression of lysozyme (LYZ), lactoferrin (LTF) and β-defensin 2 (HBD2) at mRNA levels was not detected in monolayer BCi cells. In summary the APD compounds affected multiple innate antimicrobial effectors in bronchial undifferentiated epithelial cells. (d) secretion of hBD-1 peptide measured by ELISA in cell culture supernatants. (e) S100A8 expression was analyzed by qRT-PCR. TUBB (tubulin-β) was the reference gene in qRT-PCR. Each bar represents mean value of 3 independent experiments ± SEM; statistical significance was calculated in comparison to the control group using two-way ANOVA with Dunnett's multiple comparisons test; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Significant changes are highlighted.
www.nature.com/scientificreports www.nature.com/scientificreports/ HO53 and HO56 affect cytokine profiles in bronchial epithelial cell line (BCi). We also evaluated the effect of Entinostat, HO53 and HO56 on the induction of the pro-inflammatory mediators IL-8 (CXCL8) and TNFα. Upon induction with Entinostat and HO56 a significant induction of mRNA for IL-8 was observed (Fig. 3a). However, at the protein level significantly increased secretion of IL-8 in cell supernatants was observed for the three inducing compounds (Fig. 3b). Induction of TNFα mRNA expression for three inducing compounds was observed only at 24 h ( Fig. 3c) but at the protein level significant increased secretion was detected only in cell supernatants from Entinostat and HO53 stimulated culture at 48 h (Fig. 3d). In summary the APD compounds Entinostat, HO53 and HO56 affected the expression and release of specific cytokines and chemokines, indicating potent enhancement of innate immunity defenses in bronchial epithelial cells.

HO53 and HO56 enhance antibacterial activity of the human bronchial epithelial cells (BCi).
We observed that HO53 and HO56 treatment enhanced production of AMPs and other innate immunity factors in human bronchial epithelial cells. Therefore, we tested if HO53 and HO56 treatment was effective in inducing a functional antimicrobial response in BCi cells against the respiratory pathogen Pseudomonas aeruginosa strain PAO1 (Fig. 4). BCi cells were treated for 24 h with HO53 and HO56, then infected for 1 h with multiplicity of infection (MOI) ~40, remaining extracellular bacteria were eliminated by treatment with gentamicin and the number of intracellular bacteria were enumerated as colony forming units (CFU). The number of intracellular PAO1 was significantly lower after HO53 (~75%) and HO56 (~60%) treatment in comparison to the PAO1 number in untreated control cells (100%; equal to ~1.0 × 10 4 intracellular PAO1) (Fig. 4a). The treatment with low doses of gentamicin (0.5 µg/ml), a cell impermeable antibiotic, served as a positive control and reduced PAO1 entry to the cells about 50%. We performed an analogous experiment to the infection assay but in the absence of the BCi cells (Fig. 4b). After 1 h direct exposure of PAO1 to HO53 and HO56 in the cell culture medium, there was no significant differences in CFU counts (Fig. 4b). Furthermore, HO53 and HO56 did not inhibit PAO1 growth in Luria Bertani (LB) medium over time and did not kill bacteria after 2 h of direct exposure (Fig. 4c,d).
To verify if the reduction of the intracellular PAO1 number was caused by induced antimicrobial polypeptides, we excluded additional antimicrobial effector systems that are involved in the defense of the epithelial surfaces, by analyzing the production of reactive oxygen species (ROS) (Supplementary Fig. S6a) and nitric oxide (NO) (Supplementary Fig. S6b, S6c) 39,40 . These findings indicated that the reduction of intracellular bacterial number www.nature.com/scientificreports www.nature.com/scientificreports/ was due to induced antimicrobial responses in human bronchial epithelial cells, but not any direct activity of the two APD compounds.

HO53 and HO56 induce several innate immunity genes in polarized bronchial epithelial cells.
In order to transfer our findings into a more clinically relevant context, we tested if HO53 and HO56 treatment was effective in stimulating antimicrobial responses in polarized mature epithelium, mimicking the human bronchi 41 , we analyzed expression of AMPs in differentiated BCi cells. The mRNA levels of LCN2 and S100A8 were significantly higher compared to untreated cells, especially after 24 h post-treatment with HO56 (~4 and ~40 fold, respectively) ( Fig. 5a,b). We also observed increased protein levels of lipocalin 2 (NGAL) and S100A8 after 24 and 48 h of treatment with HO53 and HO56. However, induction of lipocalin 2 was more pronounced after 48 h (Fig. 5c). Further, we analyzed the effects of HO53 and HO56 treatment on the expression of the genes CAMP, HBD1 and LYZ in differentiated BCi cells (ALI). The CAMP gene expression was on a similar level as the cells in monolayer (upregulated about 15-times) but that effect was not observed after 48 h of treatment with HO53 and HO56 (Fig. 5d). The enhanced expression of HBD1 (Fig. 5e) and LYZ (Fig. 5f) upon treatment with HO53 and HO56 was significantly higher (~10 times and 3 times, respectively). Furthermore, the expression of pro-inflammatory cytokines/chemokines was not significantly upregulated, except the expression of IL1B after www.nature.com/scientificreports www.nature.com/scientificreports/ 24 h of treatment with HO53 ( Supplementary Fig. S7d). In contrast to monolayer cells, the expression of TNF/ TNFα and CXCL8/IL-8 was not significant (Supplementary Fig. S7). In summary, we observed a different induction profile of antimicrobial effectors and cytokines/chemokines for differentiated cells in comparison to monolayer cells.

HO53 counteracts disruptive effect of P. aeruginosa conditioned medium on airway epithelial integrity.
Pathogens exhibit different strategies to evade the host epithelial surfaces. P. aeruginosa can disintegrate the junctions between epithelial cells and enter tissue through the paracellular space 42 . Another strategy is the production of virulence factors that can be injected into the host cells or secreted by bacteria to the environment, e.g. rhamnolipids 43 . We have shown that Pseudomonas aeruginosa PAO1 conditioned medium can disrupt epithelial barrier integrity 44 . Here we investigated if the treatment with HO53 can counteract the disruptive effect of PAO1 conditioned medium (Fig. 6). BCi cells differentiated in ALI culture were pretreated with HO53 for three consecutive days and challenged with PAO1 conditioned medium applied on the apical side of ALI culture for 24 h. Azithromycin was used as a positive control (Supplementary Fig. S8). After 3 h exposure of control BCi cells to PAO1 conditioned medium, trans-epithelial electrical resistance (TEER) decreased from ~1200 Ω × cm 2 to ~200 Ω × cm 2 and the disruptive effect was also observed after 6 h of exposure (Fig. 6a). The recovery of the tight junctions (TJs) integrity was noted after 24 h post challenge. In contrast to the control cells, the pretreatment of the ALI culture with HO53 did not lead to a pronounced TEER drop (~1000 Ω × cm 2 to ~700 Ω × cm 2 ) and the epithelial barrier integrity was restored after 5 h. To illustrate counteraction of TJ disruption, we performed confocal microscopy after 6 h of PAO1 conditioned medium challenge. Consistent with the TEER results, the disruption of zonula occludens (ZO-1) and occludin was observed in the control group, while HO53 treated group maintained TJs integrity (Fig. 6b). We further analyzed levels of the tight junction proteins occludin and claudin-1 at selected time points (0, 3, 6, 24 h) by Western blot analysis (Fig. 6c). The band corresponding to the main occludin isoform was faintly detected after 3 and 6 h of PAO1 conditioned medium challenge in the control group (the western blot analyses shows the occludin degradation/processing product, Supplementary Fig. S9) and did not reach the same expression level after 24 h. However, in the cells pretreated with HO53 occludin was not degraded. After 3 h of challenge the occludin level was reduced, while after 6 h the expression was restored. Notably, induction of occludin expression was detected upon HO53 treatment, whereas there was no change in claudin-1 level (Fig. 6c).
In conclusion, HO53 counteracted disruptive effect of PAO1 conditioned medium on respiratory epithelium via a novel mechanism affecting tight junctions.

STAT3 activation is required for HO53 mediated CAMP gene induction. It has been shown that
Entinostat induces CAMP gene expression in gut epithelial cells via activation of STAT3 and HIF-1α transcription factors 29 . HO53 is structurally related to Entinostat and here we investigated if the molecular pathways www.nature.com/scientificreports www.nature.com/scientificreports/ involved in the HO53 induced CAMP gene expression in bronchial epithelial cells follows similar pathways as described for the colon epithelial cell line HT-29 29 . For that, we examined the role of STAT3 by using the STAT3 inhibitor Stattic (Fig. 7a,b). BCi cells were pretreated with increasing doses of the Stattic inhibitor (5 µM, 10 µM, and 20 µM) for 30 min followed by stimulation with HO53 for 24 h. We observed significant dose-dependent Stattic mediated decrease in CAMP gene expression (Fig. 7a). Transcription factor HIF-1α expression was also decreased in a dose-dependent manner as for the CAMP gene (Fig. 7b). Furthermore, we investigated if HO53 treatment resulted in enhanced STAT3 expression or post-translational modifications (Fig. 7c). The expression level of STAT3 and phosphorylated-STAT3 was increased in a time dependent manner after treatment with HO53. The level of the phosphorylated STAT3 increased after 4 h and reached the maximum after 6 h of treatment with HO53. We did not observe any difference in the acetylation at Lys685 of STAT3 after treatment with HO53. However, modifications of other lysine residues of STAT3 cannot be excluded. HIF-1α expression level was gradually elevated at 4, 6 and 8 h and at 24 h there was a prominent enhanced level of HIF-1α. In summary, STAT3 is most likely a central regulator of the CAMP gene induction by HO53 although other STAT3 modifications need to be evaluated to explain the detailed mechanism of the inducing effect of HO53.

Discussion
An increasing number of infections caused by antibiotic resistant pathogens contribute to higher morbidity/mortality rate and generate high costs for the health care system 45 . Therefore, development of alternative strategies to conventional antibiotic therapy is urgently needed. Host directed therapy (HDT) based on inducing innate immunity by enhancing expression of endogenous antimicrobial components or counteracting pathogen-mediated suppression of first line defenses could be an alternative 27 . HDT could limit the selection of antibiotic resistant strains and might be used against multidrug-resistant (MDR) bacteria.
Development of novel stable compounds for the induction of innate immunity would be beneficial. To approach this aim, a reporter cell line with the CAMP gene fused to the luciferase gene was established for screening induction of the CAMP gene expression 35 . Screening of compound libraries and selected HDAC inhibitors resulted in the identification of Entinostat, working against Shigella and Vibrio infections in rabbits 30,31 . Recently, www.nature.com/scientificreports www.nature.com/scientificreports/ a similar strategy for identification of other innate immunity inducers was described based on the induction of the defensins genes HBD2 and HBD3 46,47 . Another interesting approach was through the pattern recognition receptor NOD2 (Nucleotide Binding Oligomerization Domain Containing 2), when activation of this receptor by N-phosphonacetyl-L-aspartate (PALA), was identified as a potent inducer of innate immunity. Expression of HBD2 and CAMP were induced by PALA and antimicrobial activity was demonstrated in skin explants against bacteria including methicillin-resistant Staphylococcus aureus (MRSA) 48 .
In the current study we describe that the novel APD compounds HO53 and HO56 induced a broad spectrum of antimicrobial effectors in bronchial epithelium. Both compounds are developed as AMP inducers with increased water solubility, lower cytotoxicity and reduced effects on cell proliferation as compared to Entinostat. We mainly used the BCi-NS1.1 (BCi) cell line 41 but the induction of the CAMP gene expression was also confirmed in the VA10 cell line 49 . The gene induction profiles of HO53, HO56 and Entinostat were similar with reference to time but not with reference to the concentration. The dose studies utilizing BCi cells revealed a concentration dependent response for HO53 and HO56, but not for Entinostat ( Fig. 1b; Supplementary Fig. S3) most likely due to higher cytotoxicity of the latter. Interestingly, when HO53 and HO56 were tested with the known innate immunity inducers vitamin D3 and PBA, a synergistic induction of the CAMP gene expression was noted with vitamin D3 but not with PBA. This suggests that APDs and PBA possibly act through the same signaling pathways. PBA and Entinostat are known HDAC inhibitors 28,32 , however today this activity is referred to as lysine deacetylases inhibition (KDACi) because apart from histones also cytoplasmic signaling pathway proteins are affected 50 . Indeed, the suggested mechanism of Entinostat activity included acetylation of the cytoplasmic www.nature.com/scientificreports www.nature.com/scientificreports/ signaling protein STAT3 29 . Therefore, possible KDAC inhibition by the novel APD compounds must be experimentally confirmed.
For both novel APDs we found induced antimicrobial activity in BCi cells against Pseudomonas aeruginosa (PAO1) and the disruptive effect of PAO1 on epithelial integrity was counteracted by HO53. This double effect of HO53 on the epithelial barrier might be potentially beneficial in vivo for fending off bacterial intruders and blocking bacterial translocation. Thus, HO53 would be an interesting candidate to treat and/or prevent infections. Interestingly, the antimicrobial activity was assessed against Pseudomonas aeruginosa that is prevalent in immune compromised lungs, including cystic fibrosis patients. The characteristics of the cellular response to treatment with HO53 suggest an effective HDT compound that might be beneficial for cystic fibrosis patients and used against MDR strains.
Induction of the CAMP gene expression served as a reference for the selection of HO53 and HO56 compounds. We selected additional genes encoding peptides/proteins with known defense functions at epithelial surfaces such as lipocalin 2, HBD1, S100A8, lysozyme and lactoferrin to verify whether the induced expression would mimic in vivo defenses. Significant induction of lipocalin 2, HBD1 and S100A8 expression was observed at mRNA and protein levels, indicating the APD compounds as powerful activators of a broad spectrum defense. Kinetics of mRNA and protein induction were not strictly correlated, where the most pronounced deviation was for HBD1 expression with a fold change of mRNA from 50 to 100, while at protein level only approximately 2 times changed (Fig. 2c,d). Similar differences were also observed for lipocalin 2, indicating a regulation at translational level in the cells. However, the current overall expression profile shows a broad induction of multiple polypeptides. In addition, we estimated the influence of APDs on reactive oxygen species (ROS) and nitric oxide (NO) production. ROS was not affected but the expression of NOS2 (encoding inducible nitric oxide synthase iNOS) was enhanced. However, this enhancement was not confirmed on effector level by the Griess method, measuring nitrate in solution probably due to low concentrations. The observed antimicrobial effect in our system is unlikely due to enhanced NO production. However, low induction of the iNOS protein may contribute to the antimicrobial effect in vivo. Thus, the main antibacterial activity in our model is most likely due to increased expression of antimicrobial proteins/peptides and could even be enhanced in the in vivo situation with complete processing of active components.
We analyzed the induction of different innate immunity effectors expression both in monolayer BCi cells and air liquid interphase (ALI), thereby comparing basal like cells with differentiated epithelial cells 41 . Several differences were observed: (1) the mRNA expression for lysozyme (LYZ) was not detected in monolayer but was inducible in differentiated cells (Fig. 5f), (2) induction of S100A8 expression was more pronounced in ALI culture (S100A8 could not be detected in undifferentiated cells). (3) In contrast, the HBD1 expression was more pronounced in undifferentiated cells. (4) The expression of pro-inflammatory effectors like, TNF/TNF-α and CXCL8/IL-8 was significantly enhanced in monolayer cells (Fig. 3) but not in ALI cultured cells, except IL1B expression ( Supplementary Fig. S7). These differences indicate that the undifferentiated cells are more sensitive to external stimuli and trigger NF-kB regulated responses, which might be linked to their basal functions. The different responses can also reflect different transcription factors setup and chromatin accessibility in the basal cells versus polarized cells.
In the experiments for approaching molecular mechanism and epithelial integrity we selected HO53 because of lower cytotoxic effects on BCi cells, higher yields from the synthesis and better solubility than HO56. One suggested mechanism for induction of innate immunity by Entinostat included two steps: first, activation of STAT3 by acetylation and second, subsequent increase of HIF-1α 29 . Notably, the transcription factor HIF-1α has been confirmed important for transcription of innate immunity genes 51 . Because the new APDs are structurally related to Entinostat, we tested the effect of HO53 in relation to the STAT3 transcription factor using a specific inhibitor -Stattic. By blocking STAT3 we observed significant reduction of CAMP and HIF1A expression upon HO53 treatment. Furthermore, upon HO53 treatment there was a time dependent increase of the STAT3 protein and in particular a pronounced effect on the phosphorylation of STAT3, whereas the acetylation status at Lys685 was unchanged. Gradually increased expression of HIF-1α was detected with time after stimulation of the cells by HO53. In conclusion, STAT3 seems to be an important mediator of the APD response but more detailed studies on STAT3 modifications are needed that might define the cellular target of HO53.
In gut epithelia STAT3 mediated expression of occludin was shown to enhance tight junction function and prevent bacterial translocation 52 . HO53 treatment led to increased occludin expression (Fig. 6) that could be an explanation for limited disruption of tight junctions (TJs) caused by Pseudomonas. This type of counteraction was initially identified on polarized lung epithelium for the antibiotic azithromycin (AZM) by an unknown mechanism 44 . Here, we confirmed the effect by AZM (Supplementary Fig. S8), but interestingly the HO53 effect seemed to be more potent in rescuing the barrier integrity than AZM. Together all these effects of HO53 do not only underline the double action of the molecule but also highlights the importance of STAT3 in epithelial immunity 53,54 .
Our results warrant continuation in animal infection models and motivate pharmacodynamics and pharmacokinetic studies. Usage of the APDs could re-establish the niche for the natural microbiota and avoid selection of resistant bacteria in line with improvement of stewardship in healthcare.
Cell cultures. The human bronchial epithelial cell line BCi-NS1.1 (BCi) immortalized with retrovirus expressing human telomerase (hTERT) was from dr Matthew S. Walters, Weill Cornell Medical College, New York NY, USA 41 . An E6/E7 viral oncogene immortalized human bronchial epithelial cell line VA10 has been described previously 49 . Both cell lines were cultured in Bronchial/Tracheal Epithelial cell growth medium (BEGM) (Cell Applications, 511A-500) supplemented with retinoic acid (Cell Applications, 511-RA) and Penicillin-Streptomycin ((20 U/ml, 20 µg/ml, respectively) (Life Technologies, 15140122)) at 37 °C and 5% CO 2 . The ALI (air-liquid interface) culture of BCi cells were maintained as described previously 41 . Cells in monolayer were treated with HO53 and HO56 by direct addition to the culture medium. Differentiated BCi cells were used for experiments when TEER value of ≥1000 Ω × cm 2 was reached and then HO53 and HO56 were added to the lower chamber for indicated period of time. Azithromycin was used as a control.
Bacterial culture. The overnight culture of Pseudomonas aeruginosa PAO1 strain was diluted in LB medium to OD 590 = 0.05 and cultured at 37 °C with 180 rpm shaking until the bacterial subculture reached the mid-log phase. (2019) 9:7114 | https://doi.org/10.1038/s41598-019-43350-z www.nature.com/scientificreports www.nature.com/scientificreports/ Ros detection. The H 2 O 2 level in the cell culture medium after 24 h of treatment with HO53 and HO56 (both at 75 µM) was measured using ROS-Glo TM H 2 O 2 assay (Promega, G8820) according to the manufacture's protocol.
Agarose gel electrophoresis. Semi-quantitative PCR products were separated on 1.5% agarose gel containing ethidium bromide (0.5 µg/ml). Electrophoresis was run at 80 V for 30 min using 1x TAE (Tris-acetate-EDTA) buffer. β-tubulin was used as a control for the quantification of the band intensity using ImageJ software.

Preparation of PAO1 conditioned medium and challenge of differentiated BCi cells. Wild-type
(WT) P. aeruginosa strain PAO1 was used to prepare bacterial conditioned medium. Shortly, the bacteria were cultured in Dulbecco's Modified Eagle Medium F-12 Nutrient Mixture (DMEM/F12) + 2% UltroserG (UG) at 30 °C and shaking was at 180 rpm for 5 days. Bacterial culture supernatants were collected, vortexed thoroughly, centrifuged, and filtered through 0.22 mm pore-size filter (GE Healthcare and Life Science, Whatman, Germany). BCi cells were cultured at the ALI for 3 weeks with medium changed every 2-3 days to get differentiated cells, followed by placing HO53 (75 µM diluted in DMEM/F12 + 2% UG) in the basal chamber of the transwell insert for 3 consecutive days, while Azithromycin was used as positive control as described previously 44 . Next, differentiated cells were challenged with PAO1 conditioned medium and TEER was measured after every hour. Samples for Western blot and confocal microscopy analyzes were collected at selected time points.
Immunofluorescence staining and confocal microscopy. BCi cells growing on ALI filters were fixed using chilled methanol at 4 °C overnight followed by chilled acetone. Briefly, staining was done as follows: filters were hydrated with IF buffer (PBS + 0.3% Triton X-100), blocked with 10% FBS, washed and incubated with a primary antibody overnight at 4 °C. The following primary antibodies were used: rabbit anti-occludin and mouse anti-ZO-1. Next day the filter was washed and incubated with a secondary antibody for 2 h. For immunofluorescence staining, isotype-specific Alexa Fluor secondary antibody conjugates were used and DAPI was used to stain nuclei. The filter was then washed with IF buffer and finally rinsed with water. Cell culture transwell filters were mounted in Fluoromount ™ Aqueous Mounting Medium (F4680-Sigma) and coverslips were placed over the filters. Images were captured using Olympus fluoview Fv1200 confocal microscope at 30x magnification. Z-scans were performed by taking series of images at the same location with fixed focal intervals. statistical analysis. Results are presented as mean ± standard error of mean (SEM) from at least three independent experiments, otherwise it is indicated in the figure legends. One-or two-way ANOVA with post-hoc Dunnett's or Sidak's multiple comparisons test were used to determine significance of the data. p values are included in the figure legends. The statistical analysis were performed with GraphPad Prism 6 software (Graph Pad, USA). The Western blots are representative of at least two independent experiments.

Data Availability
The data generated during and/or analysed during the current study are available from the corresponding author on reasonable request.