Long-Term Use of Proton Pump Inhibitors Disrupts Intestinal Tight Junction Barrier and Exaggerates Experimental Colitis

Abstract

Background

Proton pump inhibitors [PPIs] are widely used to treat a number of gastro-oesophageal disorders. PPI-induced elevation in intragastric pH may alter gastrointestinal physiology. The tight junctions [TJs] residing at the apical intercellular contacts act as a paracellular barrier. TJ barrier dysfunction is an important pathogenic factor in inflammatory bowel disease [IBD]. Recent studies suggest that PPIs may promote disease flares in IBD patients. The role of PPIs in intestinal permeability is not clear.

Aim

The aim of the present study was to study the effect of PPIs on the intestinal TJ barrier function.

Methods

Human intestinal epithelial cell culture and organoid models and mouse IBD models of dextran sodium sulphate [DSS] and spontaneous enterocolitis in IL-10^−/− mice were used to study the role of PPIs in intestinal permeability.

Results

PPIs increased TJ barrier permeability via an increase in a principal TJ regulator, myosin light chain kinase [MLCK] activity and expression, in a p38 MAPK-dependent manner. The PPI-induced increase in extracellular pH caused MLCK activation via p38 MAPK. Long-term PPI administration in mice exaggerated the increase in intestinal TJ permeability and disease severity in two independent models of DSS colitis and IL-10^−/− enterocolitis. The TJ barrier disruption by PPIs was prevented in MLCK^−/− mice. Human database studies revealed increased hospitalizations associated with PPI use in IBD patients.

Conclusions

Our results suggest that long-term use of PPIs increases intestinal TJ permeability and exaggerates experimental colitis via an increase in MLCK expression and activity.

1. Introduction

Proton pump inhibitors [PPIs] block the H⁺K⁺ATPase enzyme system in gastric parietal cells. As a treatment for gastro-oesophageal reflux disease [GERD] and associated conditions, PPIs are the second most commonly prescribed drug class in the USA.^1–3 PPI-induced elevation in intragastric pH and subsequent alterations of gastrointestinal [GI] physiology are known to cause undesired effects on the entire GI tract.⁴ PPIs are associated with an alteration in gastric pH and have been shown to cause dysbiosis. Microbiome alterations contribute to the non-gastric effects of PPIs, especially, in the colon. Recent studies have shown that suppression of gastric acid secretion increases the risk of Clostridium difficile infections^5–7 as well as liver disease.⁸ Of all the medications used in suppressing gastric acid secretion, PPIs are more efficient in the rapid healing of gastric ulcers and symptom relief than H2 blockers in the clinical setting. However, there is a possibility of PPIs causing dysbiosis of the gut microbiota due to gastric acid suppression.^9,10

The intrinsic intestinal barrier is composed of single layer of columnar epithelial cells and interepithelial tight junctions [TJs] residing at the apical-most region of the intercellular space. The TJs act as a paracellular barrier and serve as a first line of defence against paracellular permeation of noxious luminal antigens.^11–15− myosin light chain kinase [MLCK]is an important regulator of TJ permeability and does so by inducing MLC phosphorylation and subsequent energy-driven contraction of peri-junctional actin/myosin filaments, leading to physical separation and opening of the TJ barrier [leak pathway], and affecting paracellular flux of both small-sized molecules and macromolecules.^15–17 Loss of the intestinal TJ barrier and increased paracellular permeability has been shown to be an important contributing factor in the pathogenesis of inflammatory bowel disease [IBD], including Crohn’s disease [CD] and ulcerative colitis [UC] and other inflammatory conditions of the gut.^14,15,18,19 PPIs have been shown to alter the microbiome^3,20–22 and may provoke disease flares and increase hospitalization in individuals with established IBD.^5,23,24 Also, early life PPI use appears to be associated with subsequent IBD risk.²³ To date, only few studies have described the association between PPIs and intestinal permeability, despite intestinal permeability being an important factor in the pathophysiology of IBD. PPIs have been shown to increase the intestinal permeability of the upper GI tract in healthy volunteers^25,26 and under stress conditions²⁷, but the mechanisms were unclear.

In healthy humans, diet can modify colonic microbiota composition and its metabolic activity with consequent changes in the bacterial metabolite concentrations, pH and osmolality of the luminal content.²⁸ Alteration in the pH by PPIs has been shown to cause intestinal dysbiosis but how it affects the gut barrier function is still unclear. The extracellular pH and hyperosmorality have been shown to cause alteration in the actin cytoskeleton and affect endocytosis and intracellular trafficking in the enterocytes. Further, the connection between ion transport, intracellular pH and the TJ barrier have been described previously.²⁹ Luminal hyperosmorality and an increase in pH are observed in both physiological^28,30 and pathological conditions in the intestine, notably in IBD.^31,32 However, little is known about the consequences of PPI-induced alteration in the luminal pH on the function of the intestinal TJ barrier. To study the effect of PPI use on colonic pH and the intestinal TJ barrier and understand the underlying mechanisms, we investigated whether PPIs affect colonic pH and colonic TJ permeability. We also used independent mice models of IBD to study the effect of PPIs on the TJ barrier during experimental colitis. Our results show that long-term administration of a PPI disrupts the colonic TJ barrier via activation of MLCK and exaggerates experimental colitis.

2. Materials and Methods

2.1. Cell culture

Caco-2 cells [passage 20] were purchased from the ATCC and maintained at 37°C in Dulbecco’s modified Eagle medium composed of 4.5 mg/mL glucose, 50 U/mL penicillin, 50 U/mL streptomycin, 4 mmol/L glutamine, 25 mmol/L HEPES and 10% fetal bovine serum as previously described.^16,33 Caco-2 cells were used between passages 22 and 30 in this study. The cells were kept at 37°C in a 5% CO₂ environment. The cells were plated on Transwell filters with 0.4-μm pores [Corning] at high density [1 × 10⁵ cells/cm²] and monitored regularly by visualization with an inverted microscope [Eclipse TS100/100-F; Nikon] and by epithelial resistance measurements. To study the effect of PPIs, Caco-2 cells were treated with omeprazole [Millipore Sigma] [50 μM]. Acidified ethanol [0.15 M HCl in 75% ethanol] was used to activate omeprazole. The transepithelial electrical resistance [TEER] of cells was measured by an epithelial voltohmeter [World Precision Instruments]. Monolayers with a TEER of 450–500 Ω/cm² were used for experiments. Similarly, HIEC-6 (human primary intestinal epithelial cell line [CRL-3266-P15]) cells was purchased from ATCC and maintained at 37°C as shown previously.³⁴ The cells were plated on Transwell filters with 0.4-μm pores [Corning] at high-density [1 × 10⁵ cells/cm²] and treated with PPIs.

2.2. Determination of Caco-2 and HIEC-6 paracellular flux

Permeability was determined by measuring the apical-to-basal flux of the paracellular marker inulin [³H, M_r = 5000]. ³H-inulin was added to the apical solution and radioactivity was measured in the basal solution at 30 and 60 min using a scintillation counter, as described previously.³⁵

2.3. PPI administration and induction of colitis in mice

Studies were approved by the Pennsylvania State University, Institutional Animal Care and Use Committee [Protocol no. 0020211546]. C57BL/6 mice of 9 weeks of age were obtained from The Jackson Laboratory. Generation of MLCK^−/− mice was as described previously.³⁶ Omeprazole was dissolved in 40% polyethylene glycol [PEG 400].^37,38 Mice received either vehicle [40% PEG] or omeprazole [20 mg/kg per mouse] daily by subcutaneous injections for 30 ± 2 days. The doses were chosen according to previous studies.^8,25,38–40 For dextran sodium sulphate [DSS] colitis, after 30 days of vehicle or PPI treatment the mice received 3% DSS [molecular mass: 36 000–50 000 Da; MP Biomedical] in autoclaved drinking water for 7 days.^33,41,42 The body weights of mice were monitored daily, and the disease activity index and histological grading of colitis lesions were determined, as described previously.^33,42 For IL-10 knockout enterocolitis studies, IL-10 null [IL-10^−/− background, 129/SvEv] mice were maintained in a specific pathogen area [SPF] until 9 weeks of age and administered omeprazole or vehicle as described previously. After 30 days of PPI administration the mice were moved to the conventional housing area [non-SPF] for induction of enterocolitis. The body weights of mice were monitored daily, and the disease activity index and histological grading of colitis lesions were assessed, as described previously by us and others.⁴³

2.4. Determination of mouse colonic permeability in vivo and measurement of TEER

The colonic permeability in PPI- or vehicle-treated wild type [WT], MLCK^−/− mice, DSS colitis and IL-10^−/− mice was measured using a recycling colonic perfusion method, as described by us previously.^16,35 In brief, mice were anaesthetized using isoflurane and the colon was isolated following mid-abdominal incision. The colon was cannulated at the proximal and distal ends via a rectal opening with a 0.88-mm-diameter plastic tube. An external recirculating pump was used to recirculate the perfusate of Krebs-phosphate saline buffer for a 2-h perfusion period at a constant flow rate [0.75 mL/min]. The body temperature of the mouse was maintained at 37°C with a temperature-controlled warming blanket. Colonic permeability was assessed by measuring the luminal-to-serosal flux rate of a paracellular probe, fluorescein isothiocyanate-labelled dextran [molecular weight, 10 000 g/mol]. Water absorption was determined by using a non-absorbable marker, sodium ferrocyanide, or by measuring the difference between the initial and final volume of the perfusate. For measurement of TEER, colonic tissues were harvested immediately after euthanasia, cut longitudinally and placed on 0.03-cm²-aperture Ussing chambers [Physiologic Instruments]. TEER [Ω cm²] was calculated from the spontaneous potential difference and short-circuit current. The paracellular permeability was assessed by mucosal-to-serosal fluxes of ³H-inulin and [¹⁴C]-urea, as described by us previously.^33,42,44

2.5. Media preparation for pH studies and measurement of mouse colonic pH

Caco-2 cell medium was prepared for pH measurements as described previously.⁴⁵ Briefly, before being applied to cells in the culture, Caco-2 nutrient medium was adjusted to the appropriate pH using either 10 M sodium hydroxide or 1 M hydrochloric acid, measured on a Beckman F32 pH meter calibrated by a double reference method and allowed to equilibrate in a 5% CO₂ atmosphere. Nutrient medium subjected to 37°C and 5% CO₂ conditions showed an average pH change of only 0.02 ± 0.04 units over 24 h in cell-free conditions [data not shown]. The pH change was measured by using an Ohaus micro pH electrode, model STMICRO5. Colonic pH in mice was measured as described^45,46 with slight modifications. After euthanasia, the mice colons were dissected out, luminal contents were removed by gentle flushing, and the colon pieces were suspended in 100 µL of deionized water. Alternatively, colon tissues were also incubated in the deionized water along with phenol red dye. The pH was determined using an Ohaus pH probe, model ST320, 3 in 1 gel pH electrode.

2.6. Gel electrophoresis and Western blotting

The lysates of colonic mucosa and Caco-2 cells were prepared and processed for SDS-PAGE [sodium dodecyl-sulphate polyacrylamide gel electrophoresis] as described previously.^16,35 Equal amounts of protein were loaded in individual wells on the SDS-PAGE gel. After protein transfer to the membrane, the membranes were probed using anti-occludin; claudin-1, -2, -3 and -5 [Invitrogen]; MLCK [M 7905; Sigma]; MLC [3672; Cell Signaling]; phospho-MLC [3671; Cell Signaling]; p38 MAPK [ab170099; Abcam]; phospho p38 MAPK [ab4822; Abcam]; and β-actin [sc-1615; Santa Cruz Biotechnology] antibodies.

2.7. Confocal immunofluorescence

Immunohistochemistry for Caco-2 and HIEC cells on transwell inserts and mouse colonic tissues was performed by standard methods. The Caco-2 cells transwell inserts and colon cryosections were fixed in acetone and permeabilized with 0.1% Triton X-100 in PBS at room temperature for 5 min. The sections were then blocked in normal serum and labelled with primary antibodies in blocking solution overnight at 4°C. After being washed with PBS, the sections were incubated in Alexa Fluor-488 or Cy-3-conjugated secondary antibodies [Invitrogen]. Prolong Gold antifade reagent [Invitrogen] containing DAPI as a nuclear stain was used to mount the sections on glass slides. The slides were examined using a Leica SP8 confocal fluorescence microscope. Images were processed with LAS X software [Leica Microsystems].

2.8. Colonic epithelial cells and RT-PCR for MLCK mRNA expression

The Caco-2 cells and the mouse colonic tissues were collected in Trizol. Total RNA was isolated and RNA concentration was determined by absorbance at 260/280 nm and an equal amount of total RNA from each sample was reverse transcribed into cDNA using the QuantiTect Reverse Transcription kit [Qiagen]. The reverse transcription [RT] reactions were performed in a thermocycler [Pikoreal96; Thermo Scientific] using TaqMan Gene Expression Assay for MLCK [id no. Mm00653039_m1; Applied Biosystems].

2.9. Human patient data

Data were obtained from TriNetX, a web-based tool for population cohort research that provides access to electronic medical records including visit information, demographics, diagnoses, procedures and laboratory test values. We used the TriNetX Research Network, which encompasses electronic medical records from approximately 64 million patients from 48 healthcare organizations across the USA. The database is de-identified and users do not have access to personal or protected health information. Matching of study cohorts was based on demographic and common [ICD codes K51 and K50] comorbid conditions that have been reported as IBD with the use of PPIs or nor during the years 2002–2021. We assessed the balance of covariates through standardized mean difference, with a preset absolute threshold of >0.1 indicating imbalance. Cohorts were compared on the outcomes of hospitalization, using Kaplan–Meier analysis with a log-rank test and a Cox proportional hazard model, both of which are built in to TriNetX. In addition, we used the built-in scaled Schoenfeld residual to test for proportionality. The patients’ statistics were matched accordingly. Study outcomes were the risk for hospitalization after IBD diagnosis with the use of a PPI or not.

2.10. Establishment of human intestinal organoid and paracellular permeability

Fresh human intestinal organoids were maintained in a 37°C 5% CO₂ atmosphere with media change every 3 days and passaged as described previously.⁴⁷ The organoids were treated with PPI [omeprazole, 50 µM or veh] for 24 h and paracellular permeability was determined by co-incubating organoids with 10% [v/v] FITC-D4 [1 mg/mL, Sigma] at 37°C. Permeation of the marker FITC-D4 from the basal to luminal side of the organoids was assessed using confocal microscopy [Leica SP8; Leica Microsystems] and ImageJ software as shown previously⁴⁸

2.11 Statistical analysis

Data are reported as mean ± standard error of the mean [SE]. Whenever needed, data were analysed by using an analysis of variance [ANOVA] [SigmaStat; Systat Software]. A Tukey’s test was used for post hoc analysis between treatments following ANOVA [p < 0.05].

3. Results

3.1. PPI reduced intestinal TJ barrier function via an increase in MLCK activity and expression

To study the effect of a PPI on intestinal epithelial TJ barrier function, filter-grown Caco-2 monolayers were treated with various concentrations of omeprazole, and the TJ barrier function was assessed by measuring TEER and apical-to-basolateral flux of paracellular markers. Differentiated Caco-2 cells displaying a TEER of around 400 Ω.cm² were used in all experiments. The PPI [omeprazole, 50 μM, a dose based on our dose curve studies; Supplementary Figure S1] caused a significant decrease in Caco-2 TEER compared to control vehicle-treated cells [Figure 1A]. The PPI also caused a significant increase in mucosal-to-serosal flux of radiolabelled inulin [³H, mol. radius = 15 Å] [Figure 1B] and 10-kD dextran [mol. radius, 23 Å, not shown], indicating that the PPI increased intestinal TJ paracellular permeability. A similar reduction in TEER and increase in inulin flux was observed in Caco-2 cells treated with other PPIs, lansoprazole [40 μM] and pantoprazole [20 μM] [Supplementary Figure S2], indicating that the omeprazole-induced increase in TJ permeability extends to other PPIs. There was no evidence of a cytotoxic effect of the PPI at the concentration of 50 µM, with up to 72 h of exposure, in Lactate dehydrogenase (LDH) release assays [Supplementary Figure S3]. Collectively, based on these findings, the 50 µM concentration of PPI [omeprazole] was used in further studies. The possible effect of this PPI on the expression of TJ proteins, including occludin, claudin-1, claudin-2 and zonula occludin-1, revealed that it did not affect expression of the TJ proteins except for a moderate increase in claudin-2 [Supplementary Figure S4]. To rationalize the PPI-induced increase in large molecule paracellular flux, the effect of a PPI on Caco-2 MLCK expression and activity was examined. MLCK is an important mediator of TJ barrier function and increases TJ permeability via cytoskeletal reorganization. Previous studies from our group have shown that MLCK is an important regular of the intestinal TJ barrier and can increase intestinal TJ permeability under various pathological stimuli.^35,49 The Caco-2 cells treated with a PPI showed a significant increase in MLCK protein expression [Figure 1C] and MLCK activity as assessed by phospho-MLC [pMLC] expression [Figure 1D]. Moreover, pharmacological inhibition of MLCK with ML-7 prevented the PPI-induced increase in pMLC [Figure 1D]. Using confocal microscopy, increased pMLC staining was observed at the cell membranes in PPI-treated Caco-2 cells [Figure 1E]. In light of the PPI-induced increase in MLCK protein, we examined if the PPI affects MLCK gene expression. In RT-PCR studies, it caused a significant increase in MLCK mRNA expression [Figure 1F]. These findings indicated that the PPI causes activation of the MLCK gene, consistent with the increase in MLCK expression and MLCK activity. Also, PPI treatment caused a significant increase in Caco-2 media pH [Figure 1H]. These findings suggest that the PPI caused an increase in intestinal TJ permeability, associated with MLCK activation and an increase in extracellular media pH.

Figure 1.

A PPI induced an increase in intestinal tight junction permeability. [A] PPI [omeprazole, 50 µM] caused a decrease in transepithelial electric resistance [TEER] in Caco-2 cells when compared to control- or vehicle-treated cells. [B] The PPI also caused an increase in ³H inulin flux compared to vehicle-treated cells. [C] PPI treatment caused an increase in MLCK and phosphor-MLC [pMLC] protein expression; the MLCK inhibitor ML-7 prevented PPI-induced increase in pMLC [C and D]. β-Actin is shown as a loading control. The densitometry for MLCK and pMLC expression is included in C and D, respectively. [E] PPI-treated cells showed increased pMLC staining under confocal immunofluorescence microscopy. Green: actin, red: pMLC, blue: nuclei. Scale bar = 10 μm. Quantification of pMLC fluorescence intensity is shown in the bar graph. PPI treatment increased the MLCK mRNA expression [F] in Caco-2 cells. PPI treatment changed Caco-2 media pH [G]. n > 5 per group. Representation of more than three blots in C and D. *p < 0.05 vs Veh, **p < 0.01 vs 72 h of PPI. Unpaired t-test.^.

3.2. PPI-induced MLCK up-regulation is MAPK-dependent

In the following studies, the signaling pathways that mediate PPI-induced increases in MLCK gene activity and Caco-2 TJ permeability were examined. Since MAP kinases are known to play an important role in MLCK gene activity,¹⁶ the role of MAP kinase pathways in PPI modulation of MLCK gene activity was examined. We found that omeprazole causes an increase in phosphorylation of p38 MAPK [Figure 2A] and not JNK and ERK MAPK [data not shown]. Moreover, pretreatment with the p38 kinase inhibitor SB-203580 prevented the PPI-induced increase in MLCK promoter activity as well as protein expression [Figure 2B and C]. Overall, these data suggest that the PPI induced an increase in MLCK gene activity and protein expression in a p38 MAPK-dependent manner.

Figure 2.

PPI-induced increase in MLCK protein expression is associated with p38 MAPK activation. [A] The PPI caused an increase in phosphorylation of p38 MAPK. Total p38 MAPK [T-p38] is shown as a loading control. [B] The p38 MAPK inhibitor SB-203580 [10 µM] prevented the PPI-induced increase in MLCK promoter activity in a luciferase assay. [C] Inhibition of p38 MAPK with SB-203580 prevented the PPI-induced increase in MLCK protein expression [72 h]. n = 5 and representation of more than three blots. *p < 0.05 vs Vehicle, **p < 0.01 vs PPI. One-way ANOVA.

3.3. PPI increases mouse colonic TJ permeability by activating MLCK

Intestinal epithelial TJ permeability is known to be regulated by MLCK via MLC phosphorylation.^16,42,50,51 The above in vitro studies showed that the PPI-induced increase in intestinal epithelial TJ permeability is MLCK-dependent. Next, we investigated colonic paracellular permeability in PPI-treated mice [30 days, 20 mg/kg], using an in vivo method of whole length colonic recycling perfusion, as shown previously.^16,35 The baseline colonic 10-kD dextran flux showed a small but significant increase in mice treated with the PPI when compared to control, vehicle-treated mice [Figure 3A]. As an alternative method to assess colonic permeability, the colonic tissues from PPI-treated and control, vehicle-treated mice were mounted on the Ussing chambers to study colonic TJ permeability. The average colonic TEER in PPI-treated mice was significantly lower than that in control, vehicle-treated mice [40 and 52 Ω.cm² colonic TEER in PPI- and vehicle-treated mice, respectively] [Figure 3B]. Similarly, the ³H inulin permeability was significantly increased in the colon of PPI-treated mice compared to vehicle-treated mice [Figure 3C].

Figure 3.

PPI caused an increase in intestinal tight junction permeability in mice. PPI [omeprazole, [20 mg/kg body weight] caused an increase in 10-kDa dextran flux [A], decrease in TEER [B] and increase in ³H inulin flux serosal-to-mucosal flux when compared to vehicle-treated mice colon. PPI administration to the mice caused an increase in MLCK mRNA [D] and MLCK protein expression [E] in the colon tissue compared to vehicle-treated mice. The relative densitometry for MLCK protein expression is included in panel E. [F] The PPI caused an increase in pMLC staining in mice colon. Green: actin, red: pMLC, blue: nuclei. Scale bar = 25 μm. The relative fluorescence intensity for pMLC is shown in the bar graph. [G] The PPI changed the colonic pH in the mice, compared to vehicle-treated mice colon. n > 5 per group; unpaired t-test, *p < 0.05 vs vehicle.

In the following series of studies, the role of MLCK in the PPI-induced increase in mouse intestinal TJ permeability was examined. PPI treatment caused a marked increase in MLCK mRNA level in mouse colonocytes, as determined by real-time PCR [Figure 3D]. In control, vehicle-treated mice there was only minimal protein expression of MLCK in the colonic tissue. However, PPI administration caused an increase in colonic tissue MLCK protein expression [Figure 3E]. The PPI also caused an increase in pMLC protein expression in mouse colonocytes, as shown in Figure 3E. Similarly, under confocal immunofluorescence microscopy, PPI-treated mice colon showed a marked increase in pMLC staining on the colonic apical membrane [Figure 3F]. Together, these data suggested that the PPI increases mouse colonic TJ permeability by activating MLCK. We further examined the colonic luminal pH of the mice treated with the PPI and vehicle and our data showed that omeprazole caused a significant increase in colonic luminal pH [Figure 3G].

3.4. Severity of DSS colitis is aggravated in PPI-treated mice

Clinical as well as experimental colitis is marked by increases in intestinal paracellular permeability and epithelial TJ barrier disruption.^13,17,33,41 Thus, we assessed the effect of a PPI on the colonic permeability in DSS colitis by an in vivo method of whole-length colonic recycling perfusion with Texas red-labelled dextran [10 kDa] as a macromolecular marker. We found that the severity of experimental colitis after 7 days of oral administration of 3% DSS was significantly higher in the mice treated with the PPI [30 days of omeprazole + 7 days of DSS] when compared with the DSS only. The loss of body weight and the disease activity index [DAI]^52,53 were significantly increased in PPI- and DSS-treated mice when compared with DSS only [Figure 4A and B]. On gross examination, DSS plus PPI colons were shortened and dilated with bloody contents [Figure 4C]. The ratio of colon length to body weight was significantly lower in the DSS plus PPI mice compared with DSS-only mice [p < 0.001] [Figure 4D]. Furthermore, the PPI plus DSS mice had much more severe histopathological changes compared to DSS mice in terms of colonic mucosal erosion, neutrophilic and mononuclear infiltration in the lamina propria, loss of crypts, and oedema in the muscularis layer of the colon [histological score: 3.95 ± 0.15 and 3.13 ± 0.17 for PPI plus DSS and DSS alone mice, respectively, on a scale of 0–4; p < 0.001] [Figure 4E]. Colonic permeability was assessed after PPI treatment [30 days, 20 mg/kg] followed by 7 days of DSS treatment by in vivo recycling perfusion of 10-kDa dextran flux as described by us previously,^16,33,41,42 and was found to be comparatively increased in PPI plus DSS-treated mice compared to DSS-only mice [Figure 4F]. As an alternative approach for measuring intestinal permeability, the colonic tissues from PPI plus DSS and DSS-only mice were mounted on Ussing chambers as described previously. The ³H inulin flux in the PPI plus DSS group was significantly higher when compared to DSS-only mice colon [Figure 4G]. Interestingly, our data revealed that the colonic luminal pH in DSS mice was significant increased compared to WT control mice [Figure 4H]. There was no significant difference in the colonic pH of the PPI plus DSS and DSS-only group [data not shown]. Overall, the DSS-induced increase in clinical severity and inflammation was found to be aggravated by PPI administration.

Figure 4.

PPI administration exacerbates DSS-induced colitis in mice. PPI administration increased DSS-induced loss of body weight [A] and increased the disease activity index [B] compared to the DSS-only group. [C] Gross images of colons in DSS colitis revealed shortening of colon length in PPI-treated DSS mice compared to DSS mice. The colon length/body weight ratio was significantly lower in the PPI-treated DSS group compared with the DSS-only group [D]. *p < 0.05 vs WT DSS. [E] Histological examination showed loss of colonic surface and crypt epithelium, haemorrhages, and diffuse inflammatory cell infiltration in DSS colon. PPI administration severely obliterated colonic mucosa during DSS colitis. H&E stain, black scale bar = 50 μm. Representation of several microscopic areas, from three or more samples in each group. The histological score of DSS colitis were significantly increased in PPI-treated DSS mice compared to DSS-only mice. *p < 0.01 vs Control [vehicle], # vs DSS vehicle. In Ussing chamber studies following DSS colitis, colonic 10-kDa [F] and inulin flux [G] was higher in PPI-treated DSS mice when compared to DSS-only mice. *p < 0.05 vs Veh, ****p < 0.0001 vs Vehicle DSS. [H] DSS colitis altered the colonic pH [H]; n = 5 per group. ****p < 0.0001. ANOVA, Brown–Forsythe test and Bartlett’s test.

3.5. MLCK and pMLC analysis in PPI plus DSS colitis

In investigation of MLCK and pMLC analysis in DSS colitis following PPI administration, we found that PPI administration further aggravated the DSS-induced increase in colonocyte MLCK mRNA [Figure 5A]. We then examined MLCK and pMLC protein levels in colonic tissue from PPI plus DSS and DSS mice. PPI administration with DSS colitis caused a significant increase in the MLCK and pMLC protein level in the colon when compared to DSS-only mice [Figure 5B]. Under confocal immunofluorescence microscopy, PPI plus DSS mice colon showed a marked increase in pMLC staining on the colonic apical membrane when compared to DSS-only mice [Figure 5C].

Figure 5.

PPI administration exacerbates DSS-induced increase in MLCK activity and expression. The PPI with DSS caused a further increase in MLCK mRNA expression compared to only DSS treatment in mice colon [A]. [B] PPI and DSS-treated mice colon showed an increase in MLCK and pMLC protein expression compared to only DSS-treated colon tissue. The densitometry for pMLC expression is shown in the bar graph. [C] PPI and DSS-treated mice colon showed increased staining for pMLC when compared to only DSS mice colon. The relative fluorescence intensity for pMLC is shown in the bar graph. Green: actin, red: pMLC, blue: nuclei. Scale bar = 25 μm. ***p < 0.01 vs Veh DSS, *p < 0.001 vs Veh DSS. Unpaired t-test.

3.6. PPI promotes MLCK expression, TJ barrier disruption and colonic inflammation in IL-10 knockout mice

As an alternative colitis model, we examined the inflammation-promoting effects of a PPI in an IL-10 knockout [IL-10^−/−] spontaneous enterocolitis model.⁴³ In this model, the IL-10^−/− mice do not show any symptoms as long as they are maintained in specific pathogen-free [SPF] housing but develop colitis in several weeks after transfer to conventional housing. The IL-10^−/− mice were treated with vehicle or PPI [omeprazole] for 30 days in SPF housing and then were maintained in SPF housing or moved to conventional housing area. Body weights and DAI were monitored as indicators of colitis induction. As shown in Figure 6A, after moving to a conventional area, PPI-treated IL-10^−/− mice showed a significant drop in body weights at 6 weeks compared to IL-10^−/− mice without PPI. Similarly, the DAI score was higher in PPI-treated IL-10^−/− mice compared to IL-10^−/− mice without PPI [Figure 6B]. Colonic permeability was assessed by in vivo recycling perfusion of 10-kDa dextran flux at 10 weeks, and was markedly increased in vehicle-treated IL-10^−/− mice which were moved to conventional housing compared to age-matched vehicle-treated mice maintained in SPF housing [Figure 6C]. Moreover, PPI treatment further increased the 10-kDa dextran flux compared to vehicle-treated IL-10 knockout mice in a conventional area, suggesting exaggeration of intestinal barrier dysfunction caused by the PPI [Figure 6C]. Further studies showed that colonic epithelial MLCK protein expression was significantly higher in IL-10^−/− mice in the conventional area that were treated with a PPI compared to vehicle control group treatment [Figure 6D]. The IL-10^−/− mice treated with a PPI exhibited more inflammatory cell infiltrations in the colonic mucosa and much higher mean histological scores compared to IL-10^−/− mice without PPI [Figure 6E]. Interestingly, the IL-10^−/− mice with enterocolitis showed a significant increase in luminal pH when compared to control IL-10^−/− mice [Figure 6F]. Overall, we found that in the IL-10^−/− enterocolitis model, PPI pretreatment increased MLCK expression and activity, exacerbated intestinal TJ permeability, and worsened the inflammation.

Figure 6.

PPI administration exacerbates enterocolitis in IL-10 knockout mice. The IL-10^–/– mice maintained in a specific pathogen-free [SPF] area [S] and a cohort of mice were moved to a conventional area [C] to induce enterocolitis. [A] The percentage reduction in body weight in IL-10^−/− mice in the conventional area was further aggravated by PPI treatment, compared to the group without PPI. * p < 0.05 vs IL-10 + PPI [S], **p < 0.001 vs IL-10 [C]. The disease activity index was increased in IL-10^−/− mice in the conventional area treated with PPI when compared to untreated IL-10^−/− mice [B]. *p < 0.05 vs IL-10 + PPI [S], **p < 0.001 vs IL-10 [C]. In Ussing chamber studies following enterocolitis in IL-10^−/− mice, colonic 10-kDa flux was higher in IL-10^−/− mice in the conventional area than untreated IL-10^−/− mice [C]. *p < 0.001 vs IL-10 [S], *p < 0.0166 vs IL-10 [C]. PPI-treated IL-10^−/− mice in the conventional area showed an increase in MLCK and pMLC protein expression compared to non-PPI IL-10^−/− mice [D]. p < 0.001 vs IL-10 [S]. Histological examination showed increased loss of colonic surface epithelium and diffuse inflammatory cell infiltration in IL-10^−/− mice in the conventional area treated with a PPI when compared to untreated IL-10^−/− mice [E]. The histological score of IL-10^−/− mice treated with the PPI was significantly increased when compared with untreated IL-10^−/− mice. H&E stain, black scale bar = 25 μm. Representation of several microscopic areas, from three or more samples in each group [F]. *p < 0.001 vs IL-10 [S], **p < 0.001 vs IL-10 [C] The colonic pH in IL-10^−/− mice in the conventional area was increased compared to IL-10^−/− mice in the SPF area [G]. ****p < 0.0001 vs IL-10 [S] [S ANOVA, Brown–Forsythe test and Bartlett’s test].

3.7. Alkaline pH causes rearrangement of the actin cytoskeleton via activation of p38 MAPK and MLCK

Homeostasis of intracellular and extracellular pH is important for numerous cellular processes including proper protein folding, enzyme activities and protein synthesis.^54–56 Chronic administration of PPIs has been shown to suppress gastric acid secretion, alter GI luminal pH, and induce intestinal bacterial overgrowth and dysbiosis.^{8,20,21,27,57} Interestingly, our studies revealed that long-term administration of a PPI in WT mice altered colonic pH [Figure 3G]. To further examine if the PPI-induced intestinal TJ barrier disruption is due to alteration in the extracellular pH, we examined the effect of alkaline media [extracellular alkalosis] on the TJ barrier in Caco-2 cells. The normal media pH for the Caco-2 cell line was in the range 7.3–7.6. Caco-2 cells were exposed to alkaline media [pH 8.5] and TEER and inulin flux were measured to assess the TJ barrier function, as shown previously. Media with pH 8.5 caused a decrease in Caco-2 TEER compared to control Caco-2 cells incubated in pH 7.5 media [Figure 7A]. Elevated pH in the media also caused a significant increase in trans-epithelial flux of inulin [Figure 7B], indicating that pH itself can cause the alteration in intestinal TJ barrier. Since alkaline pH and not acidic pH has been shown to cause bleb formation and activation of MLCK,⁵⁸ we further examined the possibility that the alteration in pH affects MLCK expression, and studied the effect of media pH on MLCK protein expression. Similar to our findings with the PPI, we found that pH 8.5 media caused an increase in Caco-2 MLCK protein expression [Figure 7C]. As MLCK activity has been shown to be regulated by p38 MAPK⁵⁹ we examined the effect of altered media pH on activation of MAPK. Our studies showed that pH 8.5 media increased phosphorylation of p38 MAPK when compared to normal pH media in Caco-2 cells [Figure 7D]. These data suggested that changes in the extracellular pH activates p38 MAPK signalling and thus may cause actin rearrangement and TJ barrier disruption via activation of MLCK, resulting in increased intestinal TJ permeability. We also noted that replacement of pH 8.5 media with normal media [pH 7.5] allowed recovery of the TEER and reduction in inulin flux to near control levels [Figure 7A and B], further supporting the impact of alkaline extracellular pH on TJ barrier function.

Figure 7.

Alkaline pH causes activation of MLCK and disrupts the TJ barrier. Exposure of Caco-2 cells to media of pH 8.5 caused a decrease in TEER [A] and increase in ³H inulin flux [B] in Caco-2 cells when compared to control media of pH 7.5. The changes in TEER and inulin flux caused by pH 8.5 media were reversible, n = 6 [R]. *p < 0.0001 vs Media pH 7.5. Media of pH 8.5 caused an increase in MLCK protein expression [C] and p38 MAPK activation [D]. The densitometry for MLCK protein expression and phospho-p38 MAPK expression is shown by bar graphs in the respective panels. Representation of more than three blots. *p < 0.05 vs control [pH 7.5].

3.8. PPI-induced increase in intestinal TJ permeability and exaggeration of DSS colitis is prevented in MLCK^−/− mice

Given the role of the PPI in activation of MLCK and increase in colonic TJ permeability, we examined the effect of both the PPI and DSS in MLCK^−/− mice. The MLCK^−/− mice were administered with omeprazole [30 days, 20 mg/kg] and then also further exposed to 7 days of DSS colitis as described earlier. The PPI-induced increase in intestinal permeability as well as the severity of chronic DSS colitis was attenuated in MLCK^−/− mice compared to WT mice. The MLCK^−/− mice in the PPI plus DSS group showed no difference in body weights or DAI when compared to the DSS-only group [Figure 8A and B]. Compared to WT mice treated with the PPI and DSS, the colonic 10-kDa dextran flux was also lower in MLCK^−/− mice, indicating preservation of TJ barrier function during PPI and DSS colitis [Figure 8C].The histological changes of complete colonic mucosal obliteration, loss of crypts, and mucosal and submucosal mononuclear infiltration were not commonly seen in MLCK^−/− mice treated with the PPI and DSS when compared to PPI and DSS-treated WT mice and the PPI did not change the DSS histology score in MLCK^−/− mice [Figure 8D]. Overall, these findings underline the role of MLCK in PPI-induced disruption of the intestinal TJ barrier.

Figure 8.

Severity of PPI-induced DSS colitis and epithelial permeability is attenuated in MLCK^−/− mice. Percentage reduction in the body weights [A] and disease activity index [B] was comparable in PPI + DSS colitis and DSS-only group in MLCK^−/− mice. [C] DSS-induced increase in 10-kDa dextran flux and PPI-mediated exaggeration of 10-kDa dextran flux in WT mice was prevented in MLCK^−/− mice. [D] Histological examination of the colon did not reveal significant changes in DSS or DSS + PPI colon in MLCK^−/− mice. n = 5, H&E stain, black scale bar = 25 μm. **p < 0.01, as indicated. ANOVA, Brown–Forsythe test and Bartlett’s test.

3.9. PPI increased intestinal TJ permeability and pMLC in primary human epithelial intestinal cells and intestinal organoids

To support our results from Caco-2 cells and mice studies, we further studied the effect of the PPI on intestinal TJ permeability in filter-grown HIEC monolayers and in 3D human colonoids. The HIEC cells and 3D colonoids were treated with PPI [omeprazole, 50 μM] as described in the Methods. In HIEC cells the TJ barrier function was assessed by measuring apical-to-basolateral flux of radiollabeled inulin [³H, mol. radius = 15 Å]. The PPI [omeprazole, 50 μM] caused a significant increase in apical-to-basolateral flux of radiolabelled inulin [Figure 9A], indicating that it increased intestinal TJ paracellular permeability. Similarly, to rationalize the PPI-induced increase in large molecule paracellular flux, the effect of the PPI on HIEC pMLC expression was examined. Under confocal microscopy, increased pMLC staining was observed at the cell membranes in PPI-treated HEIC cells when compared to vehicle-treated HIEC cells [Figure 9B]. Furthermore, the intestinal permeability was measured in colonoids treated with omeprazole by adding 4-kDa FITC dextran to the media. PPI treatment caused a significant increase in the presence of 4-kDa FITC dextran in the colonoid lumen compared to untreated colonoids, as shown in Figure 9C. Similarly, confocal immunofluorescence showed that PPI-treated colonoids had a significant increase in pMLC expression when compared with untreated colonoids [Figure 9D]. Overall, these results supported our Caco-2 and mice studies showing a PPI-induced, MLCK-dependent increase in intestinal TJ barrier permeability.

Figure 9.

The PPI increased intestinal TJ permeability and phosphor-MLC in HIEC cells and intestinal organoids. [A] The PPI [50 µM] caused a significant increase in ³H inulin flux compared to [vehicle] untreated cells. *p < 0.002 vs Veh. [B] PPI-treated HIEC cells showed increased pMLC staining under confocal immunofluorescence microscopy. Green: pMLC, red: actin, blue: nuclei. Scale bar = 10 μm. Quantification of pMLC fluorescence intensity is shown in the bar graph. *p < 0.05 vs Veh. PPI treatment significantly increased 4-kDa FITC dextran increase in intestinal colonoids when compared to untreated intestinal colonoids. Quantification of the FITC dextran is shown in the bar graph. ***p < 0.001 vs Veh. [D] PPI treatment in intestinal colonoids caused a significant increase in pMLC staining by confocal immunofluorescence microscopy. Green: pMLC, red: actin, blue: nuclei. Scale bar = 10 μm *p < 0.001 vs Veh. n > 5 per group for A and C and representation of >15 cells or colonoids in B and D. Student’s t-test.

3.10. Long-term use of PPIs increases hospitalization in IBD patients

PPIs have been commonly prescribed to IBD patients, including UC and CD, to alleviate gastric acid-related symptoms.^5,24 PPIs have also been associated with an altered microbiome and intestinal dysbiosis. Microbiome alteration or dysbiosis is also a key factor in IBD.^20,22 Patients with IBD are often prescribed a PPI but the impact of PPIs on IBD patients is not clear.^5,24 Recently, an association of PPI use with flares in IBD has been shown and the regular use of PPIs was associated with an increased risk of IBD and its subtypes.⁶⁰ To further extend the previous observations, we assessed if PPI administration in IBD patients is associated with any adverse effects and conducted a study comparing the effect on a PPI on IBD patients using TriNetX. We used a modified algorithm recommended by TriNetX to identify adult patients [aged ≥18 years] with a diagnosis of IBD [UC and CD]. The IBD patients were divided into two study cohorts based on treatment with and without a PPI within the last 10 years. The total number of matched patients was 45 151 and the association with hospitalizations was studied. Our data revealed that PPI administration in IBD patients is associated with a significant increase in hospitalization [Figure 10]. Although further studies are needed to establish the impact of PPIs on IBD patients, our findings clearly indicate that the recent rise in the use of gastric acid-suppressing [PPI] medications might contribute to the increased incidence of complications associated with IBD.

Figure 10.

PPI use in IBD patients is associated with adverse outcomes. The use of a PPI in patients with IBD over a 5-year period was associated with an increased risk of hospitalization. n = 42 454. ***p < .0001. Student’s t-test.

4. Discussion

A defective intestinal epithelial TJ barrier is an important pathogenic factor that contributes to the development of intestinal inflammation and systemic inflammatory responses by allowing increased intestinal permeation and systemic circulation of gut-derived bacterial antigens.^16,33,50,61 Loss of intestinal TJ barrier function forms a basis for human intestinal disorders and numerous experimental models of colitis.^16,35,41,62 Since the emergence of PPIs in the 1970s, they have been widely used for the treatment of a variety of gastric acid-related diseases. However, emerging evidence shows an increasing association of the use of PPIs with Clostridium difficile-associated diarrhoea, intestinal bacterial overgrowth and dysbiosis.^8,20,21,57 Recent studies have also shown that PPIs have the potential to alter the gut microbiome through pH-dependent and -independent mechanisms.^57,63 It is also evident that PPIs, via suppression and alteration of the gastric pH, affect the pH of the entire digestive tract. A strong association between the use of PPIs and IBD⁶⁰ has been demonstrated. PPIs have also been shown to provoke disease flares and increase hospitalization in individuals with established IBD.^23,24 The role and status of colonic pH in IBD is not clear and only a few such studies have reported a wide range of colonic pH in IBD.⁶⁴ Moreover, the impact of PPI use on tight junction barrier in IBD is yet not clear. Our studies show that the PPI not only altered colonic pH but also caused an increase in colonic TJ permeability.

The focus of the present study was on examining the effect of a PPI on colonic TJ barrier and colonic homeostasis. Our in vitro studies show that omeprazole treatment causes activation of p38 MAPK and MLCK activation in Caco-2 cells. We were also able to show increased TJ permeability and MLCK activation in HIEC cells as well as human colonoids. MLCK is a central regulator of TJ permeability and increases TJ permeability via actin cytoskeletal rearrangement. Our previous studies have shown that lipopolysaccharide [LPS] induces an increase in intestinal TJ permeability via activation of MLCK.^16,35,49 Inflammatory mediators such as TNF-α and IL-1β have been shown to increase intestinal TJ permeability by increasing MLCK expression and kinase activity.^65–67 In our studies, high pH itself activated p38 MAPK in Caco-2 cells, which is consistent with similar findings in other experimental models including the amphibian heart.^54,68–70 In our experiments, the casual association between PPI-induced alkaline pH and MLCK activation was further supported by normalization of p38 MAPK and MLCK activity after correction of the pH of the media as well as a lack of PPI-induced increase in colonic TJ permeability in MLCK knockout mice. In an another study linking pH and MLCK, Luqmani et al.⁵⁸ showed that alkaline, rather than acidic, pH induces dramatic cellular morphological changes, through activation of MLCK and re-organization of cortical actin. Consistently, multiple studies have proposed the use of PPIs for altering the acidic tumour environment and promoting tumour cell death.^39,71,72 The role of pH in actin cytoskeleton rearrangement has previously been demonstrated,⁵⁸ but TJ barrier disruption by pH has not been well studied. Alternatively, it has been shown that an increase in extracellular pH promotes intracellular calcium concentration and mitochondrial reactive oxygen species generation.⁷³ An increase in intracellular calcium plays an important role in activation of MLCK^74,75 and PPI-induced alteration of the intracellular calcium concentration needs further evaluation.

Our studies clearly indicated that the PPI causes an alteration in colonic pH in vitro and in vivo. Colonic pH has a number of important physiological effects, including bile acid solubility and the availability of cations such as Ca²⁺.⁷⁶ The large intestine contains most of the human GI microbiome in part because the colonic pH of 5.5–7.0 is permissive for the growth of many microbial species.^57,77 It has been known for some time that the growth of pathogenic Escherichia coli under simulated gut conditions is inhibited by reduced pH.^78,79 Thus, dysbiosis is an obvious effect of altered GI luminal pH. Interestingly, recent studies have shown that PPIs are associated with increased intestinal permeability via dysbiosis of gut microbiota under stressed conditions²⁷ and with the use of non-steroidal anti-inflammatory drugs.⁹ Long-term use of PPIs has been shown to be associated with changes in the microbiota and biliary hyperplasia in a rat model.^8,40 Moreover, the role of gastric acid suppression by PPIs in exaggerating steatohepatitis was directly linked to overgrowth of enterococci, a pathobiont, indicating an important impact of PPI-induced changes in GI pH on GI tissue and the luminal microbiome.⁸ Interestingly, intestinal epithelial MLCK-activated brush border fanning by IFN-γ has been shown to promote adherence and internalization of normally non-invasive enteric bacteria.⁸⁰ The present study emphasizes an important role of MLCK in shaping the relationship between intestinal epithelial lining and the luminal microbiome. Whether MLCK plays a direct role in gut dysbiosis is not yet clear, but an MLCK-induced increase in intestinal permeability was shown to activate intestinal immune responses, supporting a role of MLCK in crosstalk between the luminal microbiome and the intestinal mucosa.⁸¹ We acknowledge limitations of this study in terms of identifying the mechanism or direct target of the PPI in colonic epithelial cells. In this regard, PPIs may target lysosomal V-ATPase, affecting the TJ barrier via altered intracellular trafficking and degradation of TJ proteins or F-actin rearrangement. It has been shown previously that the H⁺-V-ATPase subunit C binds to the actin cytoskeleton linking V-ATPase and the actin-based cytoskeleton⁸² and that inhibition of V-ATPase results in F-actin rearrangement.⁸³ Also, ion channels involved in pH_i regulation such as NHE3 are known to regulate the TJ barrier via actin cytoskeletal rearrangement.²⁹ Although we did not find any effect of PPIs on the NHE3, PPIs may target other colonic ion channels such as ENac, NHE-1, NKCC-1-2, non-gastric H⁺K⁺ATPase, Na⁺K⁺ATPase, DRA or CFTR either directly or indirectly by alterations in pH_i. Thus, whether chronic PPI administration disrupts the TJ barrier by affecting any pH-sensitive colonic channels directly or via alteration in pH_i remains to be elucidated. Moreover, although our in vitro experiments indicated a direct effect of the PPI on epithelial TJs, dysbiosis associated with PPI-induced change in luminal pH can contribute to the PPI-induced epithelial TJ dysfunction under in vivo conditions.

Our studies also demonstrated that the PPI-induced increase in intestinal TJ permeability exaggerates experimental colitis. In two independent DSS and IL-10^−/− spontaneous enterocolitis models, PPI treatment aggravated TJ barrier disruption, disease activity and intestinal inflammation. In our previous studies, we have shown that genetic and pharmacological manipulations that compromise intestinal TJ barrier integrity increase susceptibility to experimental colitis.^33,41,42 In line with those observations, our present study clearly indicates that PPI administration increases susceptibility to colitis. A recent study has linked the use of PPIs to the risk of IBD [hazard ratio 1.43; 95% confidence interval 1.23–1.66 after adjusting for sociodemographic characteristics, lifestyle factors and PPI clinical indications].⁶⁰ These observations clearly align with the PPI-induced increase in intestinal permeability in our mice models of colitis. Furthermore, we showed that IBD patients who have used a PPI for more than 3 years have an increase in hospitalization events when compared to IBD patients without PPI use [45 151 matched patients, odds ratio of 2.408, p < 0.001], suggesting that PPI use in IBD patients might worsen the disease and increase hospitalizations. Since IBD is associated with an increase in gut permeability, it will be important to investigate if PPI use aggravates gut permeability in IBD patients and results in IBD flares and hospitalizations.

In conclusion, we showed that long-term use of PPIs leads to an MLCK-dependent increase in mouse colonic permeability and exaggerates experimental colitis. The potential casual role of a defective TJ barrier in IBD and the need for an intact TJ barrier for remission in IBD warrants further investigations to understand the mechanisms of the PPI-associated increase in intestinal permeability.

Funding

This research work was supported in part by Crohn’s & Colitis Foundation Award 694583 [MN], Department of Medicine Innovation Award from the Penn State College of Medicine [MN], National Institute of Diabetes and Digestive and Kidney Diseases grant DK-106072 [TM] and DK114024 [PN]. The content is solely the responsibility of the authors and does not necessarily represent the official views of the funding agencies. The authors also acknowledge support from the Peter and Marshia Carlino Fund for IBD Research.

Conflict of Interest

No conflicts of interest, financial or otherwise, are declared by the author[s].

Author Contributions

M.N., D.M. and T.M. conceived the research; M.N. designed the research; M.N., P.L., N.M., V.D., I.K. A.G., K.S., A.W., W.D. and P.N. performed experiments and analysed data; G.Y., W.D. and W.K. provided resources; M.N., T.M., D.M. and P.N. interpreted the results; S.D. collected and analysed data; M.N., N.M., S.D. and P.N. prepared figures and drafted the manuscript; M.N., D.M., P.N. and T.M. edited and revised the manuscript. All authors approved the final version of the manuscript.

Data Availability Statement

The authors confirm that the data supporting the findings of this study are available within the article and its supplementary materials.