Campylobacter concisus Impairs Sodium Absorption in Colonic Epithelium via ENaC Dysfunction and Claudin-8 Disruption

The epithelial sodium channel (ENaC) can increase the colonic absorptive capacity for salt and water. Campylobacter concisus is a common pathogenic epsilonproteobacterium, causing enteritis and diarrhea. It can induce barrier dysfunction in the intestine, but its influence on intestinal transport function is still unknown. Therefore, our study aimed to characterize C. concisus effects on ENaC using the HT-29/B6-GR/MR (epithelial cell line HT-29/B6 transfected with glucocorticoid and mineralocorticoid receptors) cell model and mouse colon. In Ussing chambers, C. concisus infection inhibited ENaC-dependent Na+ transport as indicated by a reduction in amiloride-sensitive short circuit current (−55%, n = 15, p < 0.001). This occurred via down-regulation of β- and γ-ENaC mRNA expression and ENaC ubiquitination due to extracellular signal-regulated kinase (ERK)1/2 activation, predicted by Ingenuity Pathway Analysis (IPA). In parallel, C. concisus reduced the expression of the sealing tight junction (TJ) protein claudin-8 and induced claudin-8 redistribution off the TJ domain of the enterocytes, which facilitates the back leakage of Na+ ions into the intestinal lumen. In conclusion, C. concisus caused ENaC dysfunction via interleukin-32-regulated ERK1/2, as well as claudin-8-dependent barrier dysfunction—both of which contribute to Na+ malabsorption and diarrhea.


Introduction
Campylobacter concisus (C. concisus) is a Gram-negative, hydrogen (H 2 )-utilizing microorganism, first identified in periodontal pockets [1]. Extensive colonization by C. concisus and other anaerobic bacteria contributes to inflammation of the oral mucosa [2,3]. A clinical study first detected C. concisus, zoonotic Campylobacter jejuni/Campylobacter coli and other Campylobacter spp. in fecal samples of children To confirm that HT-29/B6-GR/MR cells retained functional viability at this time point, electrogenic chloride (Cl − ) secretion was determined after the addition of prostaglandin E2 (PGE2) and theophylline (both acting via cyclic adenosine monophosphate (cAMP) stimulation), or the cholinergic agonist carbachol (acting via calcium as second messenger). No significant difference in the increase in ISC was observed between controls and C. concisus-infected cell monolayers, either in response to PGE2 and theophylline, or to carbachol treatment ( Figure 2). This confirmed that cells 48 h post-infection were as functionally viable as control cells. Cl − secretion was determined as peak increase in short circuit current (ΔISC in µA/cm 2 ) 2-3 min after addition of Prostaglandin E2 (PGE2) (10 µM, basolateral side) and theophylline (10 mM, apical and basolateral side) to the Ussing chamber. Peak ΔISC was also measured 2-3 min after addition of carbachol (100 µM, basolateral side). C. concisus-infected cell monolayers were compared to untreated and DBA-stimulated controls (n = 4-5 each, ns = not significant). DBA = dexamethasone, butyrate, and aldosterone. To confirm that HT-29/B6-GR/MR cells retained functional viability at this time point, electrogenic chloride (Cl − ) secretion was determined after the addition of prostaglandin E 2 (PGE2) and theophylline (both acting via cyclic adenosine monophosphate (cAMP) stimulation), or the cholinergic agonist carbachol (acting via calcium as second messenger). No significant difference in the increase in I SC was observed between controls and C. concisus-infected cell monolayers, either in response to PGE2 and theophylline, or to carbachol treatment ( Figure 2). This confirmed that cells 48 h post-infection were as functionally viable as control cells. ENaC-dependent Na + absorption prior to infection. An increase in amiloride-sensitive short circuit current (ISC in µA/cm 2 ) was observed after DBA stimulation compared with unstimulated controls and recorded as ΔISC (Figure 1). Forty-eight hours post-infection, a significant reduction in ΔISC was observed in C. concisus-infected cell monolayers, which was similar to that seen with C. jejuni infection ( Figure 1). To confirm that HT-29/B6-GR/MR cells retained functional viability at this time point, electrogenic chloride (Cl − ) secretion was determined after the addition of prostaglandin E2 (PGE2) and theophylline (both acting via cyclic adenosine monophosphate (cAMP) stimulation), or the cholinergic agonist carbachol (acting via calcium as second messenger). No significant difference in the increase in ISC was observed between controls and C. concisus-infected cell monolayers, either in response to PGE2 and theophylline, or to carbachol treatment ( Figure 2). This confirmed that cells 48 h post-infection were as functionally viable as control cells. Cl − secretion was determined as peak increase in short circuit current (ΔISC in µA/cm 2 ) 2-3 min after addition of Prostaglandin E2 (PGE2) (10 µM, basolateral side) and theophylline (10 mM, apical and basolateral side) to the Ussing chamber. Peak ΔISC was also measured 2-3 min after addition of carbachol (100 µM, basolateral side). C. concisus-infected cell monolayers were compared to untreated and DBA-stimulated controls (n = 4-5 each, ns = not significant). DBA = dexamethasone, butyrate, and aldosterone. Cl − secretion was determined as peak increase in short circuit current (∆I SC in µA/cm 2 ) 2-3 min after addition of Prostaglandin E2 (PGE2) (10 µM, basolateral side) and theophylline (10 mM, apical and basolateral side) to the Ussing chamber. Peak ∆I SC was also measured 2-3 min after addition of carbachol (100 µM, basolateral side). C. concisus-infected cell monolayers were compared to untreated and DBA-stimulated controls (n = 4-5 each, ns = not significant). DBA = dexamethasone, butyrate, and aldosterone.
Gene expression analysis of RNA-Seq data revealed that 1667 genes were affected (p < 0.05) 48 h after C. concisus infection. RNA-Seq data are publicly available at Gene Expression Omnibus (GEO) archive under National Centre for Biotechnology Information (NCBI) website with GEO accession ID 141217 [Campylobacter concisus impairs sodium absorption via ENaC dysfunction and claudin-8 disruption. Available online: https://www.ncbi.nlm.nih.gov/gds/?term=GSE141217 (1 January 2020)]. The p-values, adjusted for multiple testing using the Benjamini-Hochberg procedure, revealed that 186 genes were differentially expressed (adjusted p < 0.05)-of which, 66 genes were up-regulated, In addition, we determined the mRNA expression of ENaC subunits (-α, -β, -γ) in unstimulated controls, as well as in DBA-stimulated controls and C. concisus-infected cells. α-ENaC subunit (SCNN1A) mRNA expression was not significantly altered after DBA-stimulation (Supplementary Figure S1A), whereas βand γ-ENaC (SCNN1B, SCNN1G) mRNA expressions were significantly increased with DBA stimulation (Supplementary Figure S1B, 186 genes were differentially expressed (adjusted p < 0.05)-of which, 66 genes were up-regulated, and 120 genes were down-regulated (Supplementary Table S1). Importantly, the mRNA expression of the pro-inflammatory cytokine interleukin-32 (IL-32) was increased in HT-29/B6-GR/MR cell monolayers 48 h after C. concisus infection ( Figure 4). Furthermore, the downstream signaling pathways and upstream regulators modulating ENaC function was evaluated by bioinformatics prediction using Ingenuity Pathway Analysis (IPA) software (Supplementary Table S2 and 120 genes were down-regulated (Supplementary Table S1). Importantly, the mRNA expression of the pro-inflammatory cytokine interleukin-32 (IL-32) was increased in HT-29/B6-GR/MR cell monolayers 48 h after C. concisus infection ( Figure 4). Furthermore, the downstream signaling pathways and upstream regulators modulating ENaC function was evaluated by bioinformatics prediction using Ingenuity Pathway Analysis (IPA) software (Supplementary Table S2). Furthermore, we analyzed the mRNA expression of different absorptive and secretory transporters that influence the Na + absorption and ENaC function after C. concisus infection through RNA-seq data (Supplementary Table S2). It revealed that the mRNA expression of Na + K + ATPase was not down-regulated after C. concisus infection. The mRNA expression of secretory chloride channels Na-K-Cl cotransporter 1 (NKCC1) and calcium-activated chloride channels (CaCC) was not upregulated (as would have been expected for a diarrheal state) after C. concisus infection. Cystic fibrosis transmembrane conductance regulator (CFTR) does not impact the inhibition of ENaC function by C. concisus either, as the mRNA expression of CFTR was not up-regulated but rather down-regulated. The mRNA expression of NHE3 was not changed after C. concisus infection, which might imply the unaltered electroneutral NaCl-absorption during C. concisus infection.
Densitometry analysis revealed that C. concisus and C. jejuni increased ERK phosphorylation after DBA stimulation ( Figure 5B), indicating that C. concisus and C. jejuni induced ERK activation in parallel with ENaC dysfunction in HT-29/B6-GR/MR cells.
To determine whether C. concisus-induced ERK activation caused functional impairment of ENaC, the specific inhibitor U0126 was used to block ERK activation by upstream inhibition of MEK. C. concisus-induced ENaC dysfunction was then tested again during inhibition of ERK activation. Based on measurements of the amiloride-sensitive increase in ISC 48 h post-infection, U0126 significantly decreased the damaging effect of C. concisus infection on ENaC (Figure 6), suggesting that ERK blockade attenuates C. concisus-induced ENaC dysfunction. Furthermore, we analyzed the mRNA expression of different absorptive and secretory transporters that influence the Na + absorption and ENaC function after C. concisus infection through RNA-seq data (Supplementary Table S2). It revealed that the mRNA expression of Na + K + ATPase was not down-regulated after C. concisus infection. The mRNA expression of secretory chloride channels Na-K-Cl cotransporter 1 (NKCC1) and calcium-activated chloride channels (CaCC) was not up-regulated (as would have been expected for a diarrheal state) after C. concisus infection. Cystic fibrosis transmembrane conductance regulator (CFTR) does not impact the inhibition of ENaC function by C. concisus either, as the mRNA expression of CFTR was not up-regulated but rather down-regulated. The mRNA expression of NHE3 was not changed after C. concisus infection, which might imply the unaltered electroneutral NaCl-absorption during C. concisus infection.
Densitometry analysis revealed that C. concisus and C. jejuni increased ERK phosphorylation after DBA stimulation ( Figure 5B), indicating that C. concisus and C. jejuni induced ERK activation in parallel with ENaC dysfunction in HT-29/B6-GR/MR cells.
To determine whether C. concisus-induced ERK activation caused functional impairment of ENaC, the specific inhibitor U0126 was used to block ERK activation by upstream inhibition of MEK.

Campylobacter concisus Decreases Transepithelial Electrical Resistance and Promotes Changes in Tight Junction Protein Expression
To investigate the barrier function of infected cell monolayers 48 h post-infection, transepithelial electrical resistance (TER) was measured 20 min after adding amiloride, when ENaC was completely blocked. Under these conditions, TER reflected paracellular sealing by TJ proteins. In C. concisus-infected cell monolayers, TER 48 h post-infection was decreased compared with DBA-stimulated controls ( Figure 7A). This is direct evidence that C. concisus impaired paracellular barrier function. We also examined changes in TJ integrity at the molecular level, and using RT-qPCR, found claudin-8 (CLDN8) mRNA expression to be decreased ( Figure 7B).

Campylobacter concisus Decreases Transepithelial Electrical Resistance and Promotes Changes in Tight Junction Protein Expression
To investigate the barrier function of infected cell monolayers 48 h post-infection, transepithelial electrical resistance (TER) was measured 20 min after adding amiloride, when ENaC was completely blocked. Under these conditions, TER reflected paracellular sealing by TJ proteins. In C. concisusinfected cell monolayers, TER 48 h post-infection was decreased compared with DBA-stimulated controls ( Figure 7A). This is direct evidence that C. concisus impaired paracellular barrier function. We also examined changes in TJ integrity at the molecular level, and using RT-qPCR, found claudin-8 (CLDN8) mRNA expression to be decreased ( Figure 7B). We also analyzed the expression of different TJ proteins by Western blotting and densitometric analysis, which indicated that C. concisus decreased claudin-8 expression and increased occludin expression when compared with DBA-stimulated controls ( Figure 7C). The expression of the other TJ proteins was not affected by C. concisus infection.
In order to further study the functional importance of this change in claudin-8, parallel experiments were done to determine the effect of C. concisus infection on the subcellular distribution of claudin-8 using confocal laser-scanning microscopy (CLSM). We observed subcellular redistribution of claudin-8 protein signals away from the TJ. Z-stack analysis of CLSM images (C) Western blots and the corresponding densitometric analysis were performed to detect changes in tight junction protein expression after C. concisus infection compared with controls after DBA stimulation (n = 3-5, * p < 0.05, *** p < 0.001, ns = not significant). DBA = dexamethasone, butyrate, and aldosterone.
We also analyzed the expression of different TJ proteins by Western blotting and densitometric analysis, which indicated that C. concisus decreased claudin-8 expression and increased occludin expression when compared with DBA-stimulated controls ( Figure 7C). The expression of the other TJ proteins was not affected by C. concisus infection.
In order to further study the functional importance of this change in claudin-8, parallel experiments were done to determine the effect of C. concisus infection on the subcellular distribution of claudin-8 using confocal laser-scanning microscopy (CLSM). We observed subcellular redistribution of claudin-8 protein signals away from the TJ. Z-stack analysis of CLSM images revealed that claudin-8 was delocalized from TJs and accumulated as intracellular aggregates in C. concisus-infected cell monolayers, whereas clear co-localization of zonula occludens protein-1 (ZO-1) and claudin-8 was observed in TJs in control cell monolayers ( Figure 8). revealed that claudin-8 was delocalized from TJs and accumulated as intracellular aggregates in C. concisus-infected cell monolayers, whereas clear co-localization of zonula occludens protein-1 (ZO-1) and claudin-8 was observed in TJs in control cell monolayers ( Figure 8). To confirm cell viability 48 h after C. concisus infection, cell proliferation rate and cytotoxicity of the cells were tested using the CCK-8 (Cell Counting Kit-8) assay. This revealed no significant differences in the cell viability after C. concisus infection when compared with controls ( Figure 9), indicating that C. concisus-induced paracellular barrier defects were independent of cytotoxicity.  To confirm cell viability 48 h after C. concisus infection, cell proliferation rate and cytotoxicity of the cells were tested using the CCK-8 (Cell Counting Kit-8) assay. This revealed no significant differences in the cell viability after C. concisus infection when compared with controls ( Figure 9), indicating that C. concisus-induced paracellular barrier defects were independent of cytotoxicity. revealed that claudin-8 was delocalized from TJs and accumulated as intracellular aggregates in C. concisus-infected cell monolayers, whereas clear co-localization of zonula occludens protein-1 (ZO-1) and claudin-8 was observed in TJs in control cell monolayers ( Figure 8). To confirm cell viability 48 h after C. concisus infection, cell proliferation rate and cytotoxicity of the cells were tested using the CCK-8 (Cell Counting Kit-8) assay. This revealed no significant differences in the cell viability after C. concisus infection when compared with controls ( Figure 9), indicating that C. concisus-induced paracellular barrier defects were independent of cytotoxicity.

Campylobacter concisus Impairs Sodium Absorption via ENaC in the Colon of IL-10 −/− Mouse
The abiotic IL-10 −/− mouse is an ideal model to study the functionality of inflamed intestine following experimental Campylobacter jejuni infection [27,28]. This mouse model was used to determine the transport effects of C. concisus in vivo, particularly ENaC-mediated Na + absorption in infected distal colon. Similar to our in vitro cell monolayer model, changes in Isc across tissues obtained from infected IL-10 −/− mice were measured in Ussing chambers. Six days post-infection, C. concisus infection in IL-10 −/− mouse colon caused a decrease in the amiloride-sensitive I SC when compared with control mice, indicating marked ENaC dysfunction ( Figure 10).

Campylobacter concisus Impairs Sodium Absorption via ENaC in the Colon of IL-10 −/− Mouse
The abiotic IL-10 −/− mouse is an ideal model to study the functionality of inflamed intestine following experimental Campylobacter jejuni infection [27,28]. This mouse model was used to determine the transport effects of C. concisus in vivo, particularly ENaC-mediated Na + absorption in infected distal colon. Similar to our in vitro cell monolayer model, changes in Isc across tissues obtained from infected IL-10 −/− mice were measured in Ussing chambers. Six days post-infection, C. concisus infection in IL-10 −/− mouse colon caused a decrease in the amiloride-sensitive ISC when compared with control mice, indicating marked ENaC dysfunction ( Figure 10). In order to test the viability of colonic tissue in Ussing chambers 6 h after aldosterone stimulation, we determined changes in ISC (∆ISC, µA/cm 2 ) after stimulation of electrogenic Cl − secretion by prostaglandin E2 (PGE2) and theophylline, and its subsequent inhibition by bumetanide. As shown in Table 1, in all mucosae Cl − secretion was stimulated by PGE2 and theophylline and inhibited by bumetanide. No significant differences in ∆ISC were observed between controls and C. concisus-infected IL-10 −/− mice, indicating that mucosal viability was maintained (Table 1). Furthermore, impedance measurements indicated that there were no differences in epithelial resistance (R epi ) and subepithelial resistance (R sub ) between the two groups (Table 1), which implies that colonic ENaC dysfunction induced by C. concisus in IL-10 −/− mice was independent of epithelial barrier changes (e.g., leaks or tissue destruction).  In order to test the viability of colonic tissue in Ussing chambers 6 h after aldosterone stimulation, we determined changes in I SC (∆I SC , µA/cm 2 ) after stimulation of electrogenic Cl − secretion by prostaglandin E 2 (PGE2) and theophylline, and its subsequent inhibition by bumetanide. As shown in Table 1, in all mucosae Cl − secretion was stimulated by PGE2 and theophylline and inhibited by bumetanide. No significant differences in ∆I SC were observed between controls and C. concisus-infected IL-10 −/− mice, indicating that mucosal viability was maintained (Table 1). Furthermore, impedance measurements indicated that there were no differences in epithelial resistance (R epi ) and subepithelial resistance (R sub ) between the two groups (Table 1), which implies that colonic ENaC dysfunction induced by C. concisus in IL-10 −/− mice was independent of epithelial barrier changes (e.g., leaks or tissue destruction).

Discussion
The first main finding was that C. concisus infection impaired ENaC activity in colonic epithelial cells, which was reflected by a decrease in amiloride-sensitive I SC and the transcriptional down-regulation of βand γ-ENaC subunits in our HT-29/B6-GR/MR (HT-29/B6 colonic epithelial cells stably transfected with glucocorticoid receptors (GR) and mineralocorticoid receptors (MR)) cell model in vitro. The HT-29/B6-GR/MR epithelial cell model is the only steroid hormone-sensitive intestinal cell model available. Basic glucocorticoids levels like 50 nM dexamethasone in the present study are necessary for a localization of the de novo expressed ENaC subunits in the apical enterocyte cell membrane. Butyrate inhibits histone deacetylation and thereby intensify βand γ-ENaC subunit expression via increased binding of the transcription factor SP3 and histone acetylation [29]. Put together, the HT-29/B6-GR/MR epithelial cell model also allowed us to investigate the intracellular cell signaling pathways that regulate or impair ENaC function.
During diarrheal states, ENaC-mediated electrogenic sodium absorption is activated in the distal colon as a reserve absorption system to minimize the loss of Na + . C. concisus impaired ENaC-mediated Na + absorption in this cell model. ENaC dysfunction has been identified as a pathomechanism that reduces the overall transport capacity for Na + and directly contributes to Na + malabsorption and watery diarrhea, a predominant intestinal symptom in C. concisus infection [8]. Furthermore, C. concisus was frequently detected in fecal samples of diarrheal patients [4,5]. Interestingly, C. concisus is the main non-zoonotic Campylobacter species identified so far in human specimens and a source of infection is yet to be identified in the environment or animals. Indeed, for many years it was unclear whether colonization of C. concisus in the human intestinal mucosa is cause or consequence of intestinal inflammation. Nielsen and co-workers demonstrated that C. concisus induced barrier dysfunction by epithelial apoptosis and moderate TJ changes in HT-29/B6 cells [9]. The study also supported the pathogenetic principle of a paracellular leak-flux mechanism exhibited by C. concisus to induce diarrhea. However, a clinical epidemiological observation found that C. concisus-infected patients present prolonged watery diarrhea with the milder intestinal inflammatory outcome and less fever compared to C. jejuni-infected patients [8]. The symptom of watery diarrhea correlates with our experimental finding that C. concisus impairs ENaC-dependent sodium absorption in the distal colon leading to watery rather than bloody diarrhea which is frequently induced by other cytotoxic enteropathogens. This feature of C. concisus infection was also reflected by our experimental findings of unchanged cell viability with retention of active Cl − secretion, defined TJ changes with claudin-8 dysregulation, and no induction of lesions or cytotoxic destruction of the tissue after C. concisus infection.
As the second main result, we showed that C. concisus induced ERK activation in HT-29/B6-GR/MR cell monolayers. In the corresponding blockade experiment, ERK inhibition with U0126 ameliorated ENaC dysfunction after C. concisus infection, which is direct evidence that C. concisus infection impaired ENaC function via ERK activation. This means the bacteria not only caused general cell damage, but they also decreased ENaC function via ERK activation. Moreover, a reduction in the mRNA expression of regulatory ENaC subunits (-β and -γ) 48 h after C. concisus infection (Figure 3) indicated that C. concisus dysregulates ENaC function. Similar mechanisms of functional ENaC dysregulation were previously reported in Crohn's disease and ulcerative colitis as well as for lymphocytic colitis [18][19][20].
Tumor necrosis factor (TNF)-α was identified as an important pro-inflammatory cytokine that could down-regulate colonic ENaC expression in different studies [18,30]. C. concisus also induced the release of pro-inflammatory cytokines such as interleukin-8 and TNF-α from intestinal epithelial cells, macrophages and/or THP1 immune cells [31]. Hence, we presumed that TNFα-mediated ERK activation might contribute to ENaC dysfunction in C. concisus infection, as previously demonstrated for Crohn's disease [19] and lymphocytic colitis [20]. However, our RNA-Seq analysis indicated that IL-32 is the cytokine with the highest mRNA expression change rather than TNF-α in C. concisus infection. The bioinformatics prediction through Ingenuity Pathway Analysis (IPA) from our RNA-Seq data indicated that C. concisus could promote ERK activation via IL-32, which might lead to ENaC dysfunction in TNF-α-independent pathway (scheme, Figure 11). IL-32 has been reported to induce activation of ERK in fibroblast-like synoviocytes in rheumatoid arthritis [32] and human calcified aortic valves [33].
Interestingly, IL-32 can also be activated by interferon-γ (IFN-γ) and interleukin-1β (IL-1β) or through bacterial lipopolysaccharides (LPS) according to our Ingenuity Pathway Analysis (IPA) analysis (Supplementary Table S3). Thus, it may be reasonable to conclude that in the presence of the sub-epithelial immune compartment, cytokines like IFN-γ and IL-1β released in response to C. concisus infection could intensify the ENaC dysfunction observed in the HT-29/B6-GR/MR cell model in our present study. activation of ERK in fibroblast-like synoviocytes in rheumatoid arthritis [32] and human calcified aortic valves [33]. Figure 11. Scheme of Campylobacter concisus-induced impairment of ENaC-dependent Na + absorption in colonic epithelial cells. In diarrheal state, glucocorticoid and mineralocorticoid receptors (GR and MR) are activated by glucocorticoids (GC) like dexamethasone and mineralocorticoids (MC) like aldosterone respectively. GR and MR forms a heterodimer, binds to steroid-responsive element (SRE) and activates the expression of β/γ ENaC genes with the subsequent protein synthesis to enhance the electrogenic Na + absorption. Butyrate inhibits histone deacetylation to intensify the expression of β/γ ENaC genes. GC (dexamethasone), MC (aldosterone) and butyrate were the activators of ENaC used in our experimental setup [17]. From the findings of our study, the red arrows in the figure represent C. concisus-induced transcriptional down-regulation or dysregulation of protein synthesis. The green arrows in the figure represent the transcriptional up-regulation and activation of signaling pathways (ERK pathway) by C. concisus. C. concisus promoted the transcriptional up-regulation of interleukin-32 (IL-32), which might increase its protein expression (indicated by upward green arrow) and lead to activation of neural precursor cell expressed developmentally down-regulated protein (NEED4-2)dependent ubiquitination of ENaC (via ERK) and impair ENaC-dependent Na + absorption. The dotted green lines represent the predicted activation of IL-32 by upstream regulators IFN-γ, IL-1β and bacterial LPS (lipopolysaccharides), which could contribute to ERK activation leading to ENaC dysfunction in DBA-stimulated controls 48 h after C. concisus infection (bioinformatics prediction from RNA-Seq data by Ingenuity Pathway Analysis (IPA) software). The mRNA expression of NEED4-2 in the HT-29/B6-GR/MR cell model is confirmed through RNA-Seq data. However, the regulation of NEDD4-2 via ERK1/2 and ubiquitination of ENaC which might lead to disassembly of ENaC subunits from epithelium is also a prediction through IPA. The reduction in claudin-8 protein expression by C. concisus is indicated by red color. The reduction in claudin-8 expression and protein redistribution perturbs the ionic paracellular barrier and leads to back leakage of Na + into the apical side, contributing to the net loss of Na + .
Interestingly, IL-32 can also be activated by interferon-γ (IFN-γ) and interleukin-1β (IL-1β) or through bacterial lipopolysaccharides (LPS) according to our Ingenuity Pathway Analysis (IPA) analysis (Supplementary Table S3). Thus, it may be reasonable to conclude that in the presence of the sub-epithelial immune compartment, cytokines like IFN-γ and IL-1β released in response to C. concisus infection could intensify the ENaC dysfunction observed in the HT-29/B6-GR/MR cell model in our present study. Figure 11. Scheme of Campylobacter concisus-induced impairment of ENaC-dependent Na + absorption in colonic epithelial cells. In diarrheal state, glucocorticoid and mineralocorticoid receptors (GR and MR) are activated by glucocorticoids (GC) like dexamethasone and mineralocorticoids (MC) like aldosterone respectively. GR and MR forms a heterodimer, binds to steroid-responsive element (SRE) and activates the expression of β/γ ENaC genes with the subsequent protein synthesis to enhance the electrogenic Na + absorption. Butyrate inhibits histone deacetylation to intensify the expression of β/γ ENaC genes. GC (dexamethasone), MC (aldosterone) and butyrate were the activators of ENaC used in our experimental setup [17]. From the findings of our study, the red arrows in the figure represent C. concisus-induced transcriptional down-regulation or dysregulation of protein synthesis. The green arrows in the figure represent the transcriptional up-regulation and activation of signaling pathways (ERK pathway) by C. concisus. C. concisus promoted the transcriptional up-regulation of interleukin-32 (IL-32), which might increase its protein expression (indicated by upward green arrow) and lead to activation of neural precursor cell expressed developmentally down-regulated protein (NEED4-2)-dependent ubiquitination of ENaC (via ERK) and impair ENaC-dependent Na + absorption. The dotted green lines represent the predicted activation of IL-32 by upstream regulators IFN-γ, IL-1β and bacterial LPS (lipopolysaccharides), which could contribute to ERK activation leading to ENaC dysfunction in DBA-stimulated controls 48 h after C. concisus infection (bioinformatics prediction from RNA-Seq data by Ingenuity Pathway Analysis (IPA) software). The mRNA expression of NEED4-2 in the HT-29/B6-GR/MR cell model is confirmed through RNA-Seq data. However, the regulation of NEDD4-2 via ERK1/2 and ubiquitination of ENaC which might lead to disassembly of ENaC subunits from epithelium is also a prediction through IPA. The reduction in claudin-8 protein expression by C. concisus is indicated by red color. The reduction in claudin-8 expression and protein redistribution perturbs the ionic paracellular barrier and leads to back leakage of Na + into the apical side, contributing to the net loss of Na + . We used amiloride, which selectively inhibited apical ENaC-mediated Na + entry. In parallel with the initial aldosterone-dependent and ENaC-mediated increase and the subsequent amiloridedependent inhibition of Na + absorption, an initial decrease and subsequent amiloride-induced increase in TER were observed. These changes in TER simply reflect the opening and closure of the ENaC and do not give any information on the effect of C. concisus infection on the paracellular barrier function. However, the overall TER after amiloride directly reflects the integrity of the paracellular barrier in HT-29/B6-GR/MR cell monolayer. Thus, a direct comparison becomes possible between infected and control monolayers. In this context, TJ protein claudin-8 plays a crucial role as it seals the paracellular space, in order to prevent the back leakage of Na + into the apical compartment [25].
The third important finding of our study shows that C. concisus-induced Na + malabsorption is accompanied by a decrease in TER, mediated by a reduction in the mRNA and protein expression of claudin-8. This might point to a paracellular barrier dysfunction promoted by C. concisus either as a parallel or as a subsequent effect to ENaC dysfunction. Moreover, a redistribution of claudin-8 from the TJ domain of the cells to intracellular compartments was observed after C. concisus infection. Hence, we could confirm that C. concisus impairs Na + absorption in colonocytes not only by ENaC dysfunction but also by claudin-8 disruption, leading to a loss of Na + via the paracellular pathway (scheme, Figure 11). A possible explanation for the proposed paracellular barrier dysfunction promoted by C. concisus, comes from one of our previous studies on claudin-8 regulation in response to ENaC activity [25], which revealed that ENaC stimulation also induces claudin-8 expression. Hence, we could ascertain that functional ENaC impairment by C. concisus could contribute to claudin-8 expression changes and might contribute to paracellular barrier dysfunction.
Previously, claudin-8 down-regulation was determined in C. jejuni infection in the human colon mucosa. However, this had not been linked with Na + malabsorption and was rather discussed in the context of a general pro-inflammatory barrier dysfunction [26]. From the findings of our present study on C. concisus, a specific contribution of claudin-8 to the loss of Na + seems much more likely. Very few claudins exist with paracellular channel function, like claudin-2 and -15 which are predominantly expressed in proximal intestinal segments like the jejunum. Other claudins like claudin-8 have rather sealing functions. Together with the expression of distinct transport proteins in the plasma membrane of the enterocytes, TJ proteins define the properties of a specific intestinal segment to be leaky or tight. Thus, a co-regulation of specific claudins and the corresponding channels is predictable but not yet shown. The signaling connection between MAPK and claudin-8 was shown in Yersinia enterocolitica infection and JNK [34] and colorectal cancer and ERK [35]. From this, we can hypothesize that the ENaC and claudin-8 co-regulation could happen via ERK/MAPK in Campylobacter infection. However, this should be confirmed only after a detailed investigation.
In our present study, a similar pathomechanism was seen after C. concisus infection as previously described in lymphocytic colitis, in which claudin-8 disruption and ENaC dysfunction synergistically promote watery diarrhea [19,36]. Interestingly, a clinical study found that 12% of C. concisus-infected patients with prolonged diarrhea developed microscopic colitis (lymphocytic colitis is a subtype of microscopic colitis) in a six-month follow-up period [37]. We also observed a significant increase in the expression of the TJ protein occludin in HT-29/B6-GR/MR cell monolayers after infection with C. concisus. This could be a result of host cell autophagy modulation required for intracellular survival of C. concisus [38,39]. Besides, occludin was demonstrated not to affect the ionic barrier properties of intestinal epithelia in an occludin knockout mouse model, since the transepithelial electrical resistance did not differ between occludin-deficient mice and their wild-type littermates [40,41]. Hence, the increase in occludin expression has to be interpreted as an independent phenomenon regardless of the paracellular barrier change induced by C. concisus.
In order to confirm the effects of C. concisus on ENaC function in vivo, we employed the abiotic IL-10 −/− mouse model. Mice display a strong physiological colonization resistance of the intestine due to the mouse-specific gut microbiota composition and are therefore protected from infection with enteropathogens including C. jejuni [42,43]. Furthermore, mice are per se approximately 10,000-fold more resistant to LOS and lipopolysaccharide (LPS), the major cell wall constituents of C. jejuni or other Gram-negative bacteria [44,45] as compared to humans [46]. It was recently shown that the abiotic IL-10 −/− mice (gut microbiota depleted by broad-spectrum antibiotic treatment) are effectively colonized by C. jejuni upon peroral infection and develop key features of acute human Campylobacteriosis [27]. The main reasons for these severe C. jejuni-induced immunopathological responses in the acute stages of enterocolitis in mice are (i) the absence of colonization resistance following microbiota depletion and (ii) the lack of IL-10 enhancing susceptibility of mice to C. jejuni LOS [27,47]. In consequence, abiotic IL-10 −/− mice infected with C. jejuni display a pronounced LOS-induced and Toll-like receptor (TLR)-4-dependent innate and adaptive immune response in the intestine [27]. Since then, abiotic IL-10 −/− mouse model has been successfully employed in many studies [48][49][50][51]. In recent studies, the abiotic IL-10 −/− mouse model was also used to determine the barrier protective and anti-inflammatory effects on C. jejuni infection using curcumin or vitamin D [52,53].
The first experimental infection of mice by C. concisus to study inflammatory effects on the intestine was carried out in Bagg Albino/c (BALB/c) mice [54]. However, this study reported poor colonization of C. concisus in the intestine of BALB/c mice without any substantial inflammation and proposed an improved mouse model from a different mouse strain for future investigations. In the current study, we achieved successful bacterial colonization of the colon in C. concisus-infected abiotic IL-10 −/− mice (C57BL/6 strain). Further, in concordance to our in vitro model, a decrease in ENaC-dependent Na + transport could be measured in the colon of C. concisus-infected IL-10 −/− mice 6 days post-infection. This gives us a solid piece of evidence for C. concisus-induced ENaC dysfunction in vivo. However, we have not observed any differences in R epi and R sub values between the colon mucosa of controls and C. concisus-infected mice. This indicates that C. concisus neither induces massive inflammation nor impairs epithelial barrier function in the colon but can still cause ENaC dysfunction.
Taken together, C. concisus impairs ENaC-dependent Na + absorption via down-regulation of βand γ-ENaC mRNA expression and ERK activation. The mRNA expression of pro-inflammatory cytokine IL-32 is up-regulated after C. concisus infection, which might contribute to ERK activation, in turn leading to ENaC dysfunction. In parallel, C. concisus disrupts claudin-8 and facilitates back leakage of Na + ions. Thus, C. concisus induces ENaC dysfunction via ERK activation and claudin-8-dependent barrier dysfunction-both of which contribute to Na + malabsorption and diarrhea.

Cell Culture and Campylobacter Infection
HT-29/B6-GR/MR cell monolayers (epithelial cell line HT-29/B6 transfected with glucocorticoid and mineralocorticoid receptors; [15]) were used to determine functional ENaC activity in vitro. Three days after DBA stimulation, cell monolayers were washed and incubated with heat-inactivated 10% h-f FCS without any antibiotic supplements for 24 h. Four days post-DBA stimulation, TER values were recorded with chop-stick electrodes and cell monolayers were infected with C. concisus (C. concisus AAuH 37 UC oral [55]) or C. jejuni (C. jejuni wild-type strain 81-176) on both apical and basolateral sides of the cell monolayers at a multiplicity of infection (MOI) of 400 ( Figure 12). After infection, cell monolayers were incubated in a H 2 -containing atmospheric condition [microaerophilic/CO 2 -enriched gas pack (BD GasPak EZ CampyPak container system sachets, BD Biosciences, San Jose, CA, USA) and 10% hydrogen gas 0.082 g of sodium borohydride (NaBH 4 ) in 10 mL of distilled water in 2.5 L airtight plastic jar] at 37 • C for approximately 30 h. Cell monolayers were then placed in a humidified atmosphere at 37 • C. Forty-eight hours post-infection, cell monolayers were used for amiloride-sensitive short circuit current (I SC ) measurements in Ussing chambers, total RNA isolation, Western blot analysis, confocal laser-scanning microscopy (CLSM, Zeiss LSM 780, Jena, Germany) and CCK-8 assay.
Biosciences, San Jose, CA, USA) and 10% hydrogen gas 0.082 g of sodium borohydride (NaBH4) in 10 mL of distilled water in 2.5 L airtight plastic jar] at 37 °C for approximately 30 h. Cell monolayers were then placed in a humidified atmosphere at 37 °C. Forty-eight hours post-infection, cell monolayers were used for amiloride-sensitive short circuit current (ISC) measurements in Ussing chambers, total RNA isolation, Western blot analysis, confocal laser-scanning microscopy (CLSM, Zeiss LSM 780, Jena, Germany) and CCK-8 assay.

Electrophysiological Determination of ENaC Function In Vitro
Forty-eight hours post-infection with Campylobacter spp., HT-29/B6-GR/MR cell monolayers grown on filters were mounted in Ussing chambers (epithelial surface area of 0.6 cm 2 ; Institute of Clinical Physiology, Charité, Berlin). The composition of the bathing solution in the Ussing chambers was as follows: Na + 140.0 mM; Cl − 123.8 mM; K + 5.4 mM; Ca 2+ 1.2 mM; Mg 2+ 1.2 mM; HPO4 2− 2.4 mM; H2PO4 − 0.6 mM and HCO 3− 21.0 mM. The solution was gassed with carbogen gas (95% O2 and 5% CO2) by bubble lift. Temperature was maintained at 37 °C, pH 7.4. TER (Ω·cm 2 ) and short circuit current (ISC; µA/cm 2 ) were recorded using voltage clamp devices (CVC6, Fiebig Hard & Software, Berlin, Germany). Cell monolayers were allowed to stabilize and ENaC-dependent Na + transport recorded as a decrease in ISC (∆ISC; µA/cm 2 ) 20 min after the apical addition of the ENaC blocker amiloride (100 µM; Sigma-Aldrich, St. Louis, MO, USA). For the complete inhibition of ENaC in the colonic epithelium and mucous-producing HT-29/B6-GR/MR cells, which is covered by a mucous layer, an amiloride concentration of 100 µM was employed. This concentration is ten-fold higher than concentrations usually used to completely block the ENaC in kidney cell models (10 µM), but still specific for Na + transport via the ENaC, as NHE3, the other transport system for Na + in the apical cell membrane of colonocytes, is only affected by amiloride concentrations at 1 mM [17]. To ensure that epithelial cells were functionally viable during ∆ISC measurements, electrogenic chloride Cl − secretion by the cells was determined by measuring the increase in ISC at end of each experiment in response to the addition of theophylline (10 mM) and PGE2 (10 µM), or carbachol (100 µM).

Western Blot Assessment of Tight Junction Protein Expression
Forty-eight hours post-infection, TERs of cell monolayers were recorded 20 min after measuring the amiloride-induced changes in ISC. Cell monolayers of HT-29/B6-GR/MR were then prepared prior to evaluating changes in TJ protein expression. Control and infected cells were scraped carefully from the cell monolayers and subjected to total cell lysis using a lysis buffer (150 mM NaCl, 10 mM Tris  For the complete inhibition of ENaC in the colonic epithelium and mucous-producing HT-29/B6-GR/MR cells, which is covered by a mucous layer, an amiloride concentration of 100 µM was employed. This concentration is ten-fold higher than concentrations usually used to completely block the ENaC in kidney cell models (10 µM), but still specific for Na + transport via the ENaC, as NHE3, the other transport system for Na + in the apical cell membrane of colonocytes, is only affected by amiloride concentrations at 1 mM [17]. To ensure that epithelial cells were functionally viable during ∆I SC measurements, electrogenic chloride Cl − secretion by the cells was determined by measuring the increase in I SC at end of each experiment in response to the addition of theophylline (10 mM) and PGE2 (10 µM), or carbachol (100 µM).

Western Blot Assessment of Tight Junction Protein Expression
Forty-eight hours post-infection, TERs of cell monolayers were recorded 20 min after measuring the amiloride-induced changes in I SC . Cell monolayers of HT-29/B6-GR/MR were then prepared prior to evaluating changes in TJ protein expression. Control and infected cells were scraped carefully from the cell monolayers and subjected to total cell lysis using a lysis buffer (150 mM NaCl, 10 mM Tris buffer pH of 7.5, 0.5% Triton X-100, and 1% SDS). The concentration of the proteins isolated was estimated by the Pierce bicinchoninic acid (BCA) assay (Thermo Scientific, Waltham, MA, USA) according to manufacturer's instruction. Proteins were resolved using 12.5% SDS-PAGE gel, and 15 µg of proteins were used from each sample. The resolved proteins were electro-transferred to PVDF nitrocellulose membranes (Thermo Scientific, Waltham, MA, USA) using the Trans-Blot system (Bio-Rad Laboratories, Inc., Hercules, CA, USA) at 25 V for 15-17 min.

Western Blot Assessment of ERK Phosphorylation
Forty-eight hours post-infection, HT-29/B6-GR/MR cell monolayers were washed and incubated with RPMI media lacking h-f FCS and supplemented with 2 mg/mL gentamycin (Gibco, Carlsbad, CA, USA) for 3 h to remove and kill all residual bacteria on the apical and basal sides of the cell monolayers. Cell monolayers were then washed with heat-inactivated 10% h-f FCS to completely remove gentamycin, after which cells were DBA stimulated and re-infected with C. concisus or C. jejuni for 5, 15, 30, 60 and 120 min before detecting protein phosphorylation by Western blotting. Cells were scraped carefully from control and the infected monolayers, and removed using complete cell lysis buffer (pH 7.5) supplemented with phosphatase inhibitors (20 mM Tris, 150 mM NaCl, 1 mM Triton X-100, 1 mM EDTA, 1 mM PMSF, 2.5 mM Na + pyrophosphate, 1 mM β-glycerophosphate, 1 mM Na + orthovanadate, 1 mM EGTA, 1µg/mL leupeptin, complete protease inhibitor cocktail (Roche, Mannheim, Germany)).
Proteins were isolated from lysed cells and their concentrations determined. Proteins were then resolved and electro-transferred to PVDF nitrocellulose membranes (Thermo Scientific, Waltham, MA, USA). PVDF membranes were blocked, incubated with primary and secondary antibodies and evaluated for specific proteins using FUSION FX7, as described in the Section 4.3. Primary antibodies used to determine ERK phosphorylation were Rb anti-p-ERK1/2, Rb anti-ERK1/2 (Cell Signaling Technology Europe B.V., Frankfurt am Main, Germany) and M anti-β-actin (Sigma Aldrich, St. Louis, MO, USA). Densitometric analysis of Western blots was performed by normalizing band intensities of p-ERK1/2 and ERK1/2 (total ERK1/2) to their respective β-actin intensities.

Functional Blockade of ERK to Determine the Changes in ENaC Function In Vitro
Upstream MEK inhibitor U0126 (Biogems International, Inc. Westlake Village, CA, USA), which functionally blocks ERK activity, was used to determine the changes in ENaC-dependent Na + transport (∆I SC ) after C. concisus infection in HT-29/B6-GR/MR cells. For this purpose, we used the in vitro infection model as described in Figure 11 and Section 4.1 with few modifications. Four days post-DBA stimulation, the apical and basolateral compartments of the cell monolayers were treated with the functional MEK inhibitor U0126 at a concentration of 10 µM supplemented along with heat-inactivated 10% h-f FCS without any antibiotic supplements. Then, the cell monolayers were incubated at 37 • C in humidified atmosphere (95% air/5% CO 2 ) for 2 h. Then, the cell monolayers were infected with C. concisus on both apical and basolateral compartment of the cell monolayers at MOI of 400. Following infection, the cell monolayers were incubated in a special microaerobic atmospheric condition (as described in Section 4.1) but only for approximately 4 h. Then, the cell monolayers were replaced at 37 • C in humidified atmosphere. Forty-eight hours post-infection, ENaC-dependent Na + transport (∆I SC ) was determined as described in Section 4. To confirm viability of the colonic epithelium after 6 h of aldosterone exposure, the Cl − secretory response of the epithelium was assessed by measuring increases in I SC at end of each experiment after the addition of theophylline (10 mM) and prostaglandin E 2 (10 µM). Subsequently, inhibition of the stimulated Cl − secretion was assessed by measuring decreases in I SC after the addition of bumetanide (100 µM) to the basolateral compartment.
In both control and inflamed colon, I SC measurements could be influenced to different degrees by the thickness of the subepithelial tissue layers. Thus, in addition to I SC measurements, the total transepithelial resistance (R total ) and subepithelial resistance (R sub ) of colonic samples were recorded via one-path impedance spectroscopy, as described previously [57]. Epithelial resistance (R epi ) was determined by subtracting R sub from R total . To ensure that changes in Isc accurately reflected changes in active transport, the contributions from subepithelial tissue were taken into account by calculating the ratio R total /R epi , as described previously [20,26].

Ethics Statement
Animal experiments were carried out in the animal facility at the Forschungseinrichtung für Experimentelle Medizin (Charité-Universitätsmedizin Berlin) according to the German animal protection law (approval number G0172/16 (13 October 2016), LaGeSo Berlin).

RNA-Seq Expression Analysis
Total RNA was obtained from HT-29/B6-GR/MR cells using the mirVana TM miRNA Isolation Kit (Ambion, Life Technologies, Carlsbad, CA, USA). RNA sequencing was performed using the TrueSeq Stranded Total RNA method on a NovaSeq TM 6000 Sequencing System (https://www.illumina.com/) with quality scores of ≥80%.
The reads from RNA-Seq were mapped against the human genome GRCh38 release 97 and sorted using the STAR aligner version 2.7.1a in a two-pass mode [58]. First-pass read mapping utilized coordinates from Ensembl annotation release 97 as a framework. Second-pass mapping added splice sites that were found in the first run. Count tables containing gene-read coverages were obtained using the feature Counts function of the Bioconductor package Rsubread [59], with coordinates from aforementioned Ensembl annotation and default parameters.
The Bioconductor package DESeq2 [60] was used to quantify the differential expression of genes between two conditions in form of log2-fold changes with their corresponding p-values. p-values were adjusted for multiple testing using the Benjamini-Hochberg procedure. Pathway analysis was performed with Ingenuity Pathway Analysis software (IPA, Qiagen Silicon Valley, Redwood, CA, USA) to evaluate the C. concisus-dependent changes in the expression of different genes that regulate ENaC function. Fastq files containing the unprocessed raw reads from sequencing and a raw counts matrix table are publicly available at Gene Expression Omnibus (GEO) archive under National Centre for Biotechnology Information (NCBI) website with GEO accession ID 141217 [Campylobacter concisus impairs sodium absorption via ENaC dysfunction and claudin-8 disruption. Available online: https://www.ncbi.nlm.nih.gov/gds/?term=GSE141217 (1 January 2020)].
In addition, counts per million and log-transformed counts per million (CPM) normalization was performed using CPM function of the Bioconductor package edgeR [61], and the gene expression of IL-32 was determined using CPM.

Statistical Analysis
All data are expressed as the mean value ± standard error of the mean (SEM). Statistical analyses were performed with GraphPad Prism (GraphPad Software version 5.0, Inc., San Diego, CA, USA). For data in Figures 1, 3-7 and 9, the unpaired t-test with Welch's correction for unequal variances was applied. For data that were not normally distributed ( Figure 10 and Table 1), the Mann-Whitney U-Test was used. To compare data sets from three different samples (data of Figure 2), two-way ANOVA with Bonferroni-Holm adjustment was used. p < 0.05 was considered statistically significant.

Conclusions
Campylobacter concisus impairs ENaC-dependent Na + absorption via down-regulation of βand γ-ENaC mRNA expression and ERK activation. The up-regulated mRNA expression of pro-inflammatory cytokine IL-32 after C. concisus infection might contribute to ERK activation, in turn leading to ENaC dysfunction. Besides, C. concisus disrupts claudin-8 and facilitates back leakage of Na + ions. Hence, C. concisus induces ENaC dysfunction via ERK activation and claudin-8-dependent barrier dysfunction-both of which contribute to Na + malabsorption and diarrhea.

Conflicts of Interest:
The authors declare no conflict of interest.