TLR4-Dependent Claudin-1 Internalization and Secretagogue-Mediated Chloride Secretion Regulate Irinotecan-Induced Diarrhea

Irinotecan is a commonly prescribed chemotherapeutic agent used to treat a variety of solid tumors (1, 2). Despite long-standing clinical efficacy, irinotecan is associated with a host of debilitating off target toxicities which severely impact its widespread implementation (3). Of these side effects, gut toxicity presents as a major clinical obstacle in oncology practice with limited therapeutic avenues. Characterized by severe diarrhea, rectal bleeding, pain, and infection, irinotecan-induced gut toxicity is considered the most significant dose-limiting side effect of irinotecan, placing a substantial clinical and economic burden on the provision of optimal cancer care (4, 5). The current understanding of the molecular mechanisms that drive gut toxicity has not yet lead to any advances in its treatment and thus, a better understanding of the underlying biology is required. Recent research has outlined emerging evidence implicating intestinal barrier injury and tight junction disruption in the development of gut toxicity (6, 7); however, few studies have investigated their regulation and involvement in diarrhea.

Tight junctions are highly dynamic signaling complexes critical to gastrointestinal homeostasis and the maintenance of barrier function (8). In the setting of chemotherapy-induced gut toxicity, intestinal barrier dysfunction is most commonly associated with augmentation of key tight junction proteins claudin-1, zonular occludens (ZO)-1, and occludin; all of which are integral to maintenance of the tight junction unit (9, 10). A large body of evidence demonstrates disruption of these proteins, and other claudin subtypes, in a number of gastrointestinal pathologies, particularly those characterized by diarrhea (8, 11), with altered barrier function and tight junction integrity seen following a number of chemotherapeutic agents (8). Dual sugar permeability assays are typically used to noninvasively assess barrier function in patients receiving chemotherapy. Elevations in permeability indices have been identified in a number of chemotherapy patient cohorts and have been shown to correlate with the severity of gastrointestinal symptoms (12–16). Despite these findings, the molecular characteristics of tight junctions have been largely ignored clinically, assessed only in in vivo studies. These studies consistently show decreased expression, redistribution, and phosphorylation of claudin-1, occluding, and ZO-1 (7, 17–19); however, few report these changes in combination with robust permeability data.

It is well-documented that changes in tight junction integrity parallel the onset and severity of gastrointestinal symptoms (20). In vivo studies have shown that increased intestinal permeability, caused by irinotecan treatment, permits endotoxin and bacterial translocation, thus increasing the risk of secondary infection (7). In patients with ulcerative colitis, loss of intestinal barrier function and tight junction integrity has been shown to contribute to diarrhea development via passive backflow of anions and water into the lumen (21). Intestinal barrier dysfunction has also been identified as a key risk factor for secondary complications in patients receiving multidrug chemotherapy (22), increasing the risk of bacteremia, mucosal-related invasive fungal disease, typhlitis, and sepsis (23).

Despite strong clinical evidence indicating that barrier dysfunction permits secondary toxicity and infectious consequences, the mechanisms by which barrier dysfunction and tight junction disruption contribute to irinotecan-induced diarrhea remain unclear. We have recently demonstrated increased intestinal permeability to 4-kDa FITC-dextran in BALB/c mice treated with irinotecan, indicating poor tight junction integrity. Importantly, our findings demonstrated that genetic deletion of Toll-like receptor 4 (TLR4; Tlr4^−/−) protected against the development of barrier dysfunction (24) and reduced the duration and severity of diarrhea. We hypothesize that the ability of TLR4 to recruit powerful downstream inflammatory signals, recognized for their ability to disrupt tight junction proteins (25), makes it a potential driver of barrier dysfunction following chemotherapy. Research shows that TLR4, and its ligand lipopolysaccharide (LPS), are both upregulated following irinotecan (26). This parallels other models of barrier dysfunction and supports the idea that TLR4 activation promotes barrier disruption (27, 28).

The current study therefore aims to investigate the morphology and molecular integrity of intestinal tight junctions in a model of irinotecan-induced diarrhea and gauge the importance of TLR4 in tight junction regulation. In addition, this study aims to assess the secretory profile of the intestine using Ussing chambers to determine the impact of altered anion secretion, secondary to barrier disruption, on diarrhea development.

The study was approved by the Animal Ethics Committee of the University of Adelaide and complied with National Health and Research Council (Australia) Code of Practice for Animal Care in Research and Training (2014). Mice were group housed in ventilated cages with 3 to 5 animals per cage. They were housed in approved conditions on a 12- hour light:dark cycle. Food and water were provided ad libitum.

All mice were on a BALB/c background. Forty-two female BALB/c-wild-type (WT) and BALB/c-Tlr4^−/−billy mice (n_total = 84) weighing between 18 and 25 g (10–13 weeks) were used. WT BALB/c mice were obtained from the University of Adelaide Laboratory Animal Service (SA, Australia), and BALB/c-Tlr4^−/−billy mice, back-crossed onto BALB/c for more than 10 generations, were kindly provided by Professor Paul Foster from the University of Newcastle (NSW, Australia) and were originally sourced from Osaka, Japan (29). All BALB/c-Tlr4^−/−billy mice were homozygous null mutants and hence expressed no detectable TLR4 mRNA or protein (24). Mice were treated with a single 270 mg/kg intraperitoneal (i.p.) dose of irinotecan hydrochloride (kindly provided by Pharmacia/Pfizer) prepared in a sorbitol/lactic acid buffer (45 mg/mL sorbitol/0.9 mg/mL lactic acid; pH 3.4; Sigma-Aldrich; d-sorbitol #S1876, lactic acid #252476), which was shown in our previous work to cause reproducible diarrhea with no mortality (24). Control mice received the sorbitol/lactic acid buffer only. All mice received 0.03 mg/kg of atropine subcutaneously immediately prior to treatment to reduce the cholinergic response to irinotecan. Mice were randomly assigned to treatment groups and killed at 6, 24, 48, 72, and 96 hours. Mice were anesthetized using 200 mg/kg (intraperitoneal) ilium sodium pentobarbital (60 mg/mL) and blood was collected from the facial vein. They were killed via transcardial perfusion with cold, sterile 1× PBS (pH 7.4) followed by 4% paraformaldehyde (PFA) in 0.1 mol/L PBS (pH 7.4).

Mice were assessed 4 times daily for response to irinotecan treatment as per Wardill and colleagues, (2016). For clinical data, please refer to the study by Wardill and colleagues (24).

The entire gastrointestinal tract from the pyloric sphincter to the rectum was dissected prior to perfusion with 4% PFA and flushed with chilled 1× PBS (pH 7.4) to remove intestinal contents. Both the small and large intestines were weighed immediately after resection. Samples (1 cm in length) of jejunum, ileum, and colon were collected and (i) drop-fixed using 10% neutral-buffered saline for processing and embedding into paraffin wax or (ii) stored in RNAlater (Sigma Aldrich; #R0901) at −20°C for molecular analyses. Mucosal scrapings were also collected from the jejunum, ileum, and colon, snap-frozen, and stored at −80°C.

Tight junction analysis was performed on the jejunum, ileum, and colon taken from WT and BALB/c-Tlr4^−/−billy mice 24 hours after irinotecan treatment as peak diarrhea severity and serum FITC-dextran were seen at this time point. Tight junction analysis was also performed on 6 vehicle control mice from each genotype to ensure TLR4 deletion did not affect tight junction morphology.

Immunofluorescence (IF) was carried out on 4-μm sections of jejunum, ileum, and colon, cut on a rotary microtome, and mounted onto FLEX IHC microscope slides (Flex Plus Detection System, Dako; #K8020). IF analysis was performed for key tight junction proteins: claudin-1, ZO-1, and occludin. IF was performed using Dako reagents on an automated machine (AutostainerPlus, Dako; #AS480) following standard protocols supplied by the manufacturer. Briefly, sections were deparaffinized in histolene and rehydrated through graded ethanols before undergoing heat-mediated antigen retrieval using an EDTA-NaOH buffer (0.37 g/L EDTA, pH 8.0). Retrieval buffer was preheated to 65°C using Dako PT LINK (pretreatment module; Dako; #PT101). Slides were immersed in the buffer and the temperature raised to 97°C for 20 minutes. After returning to 65°C, slides were placed in the Dako AutostainerPlus and tissue was blocked using 10% normal horse serum (NHS) in 1× PBS. The primary antibodies were applied for 1 hour using 5% NHS as a diluent. A fluorescently labeled secondary antibody [donkey anti-rabbit or mouse IgG (H+L) secondary antibody, AlexaFluor 568 or 488 conjugate, Invitrogen; #A10042] was applied at 0.8 μg/mL for a further 1 hour, using 1× PBS + 1% BSA (Sigma-Aldrich; #A2058) and 2% FBS (Sigma-Aldrich; #F2442) as a diluent. Slides were washed using 1× PBS, counterstained using 1 μg/mL 4′,6-diamidino-2-phenylidole (DAPI; Life Sciences; #D1306), and coverslipped using an aqueous mounting medium (Fluorshield, Sigma Aldrich; #F6182). Negative controls had the primary antibody omitted. Slides were visualized using the SP5 Spectral Scanning Confocal Microscope (Leica). IF was assessed qualitatively for staining intensity and distribution in a blinded fashion.

IF antibody details are as follows: claudin-1 (Abcam ab15098; 2 μg/mL; 1:100; AlexaFluor anti-rabbit 568 nm); ZO-1 (Invitrogen 61-7300; 2.5 μg/mL; 1:100; AlexaFluor anti-rabbit 568 nm); and occludin (Invitrogen 33-1500; 5 μg/mL; 1:100; AlexaFluor anti-mouse 488 nm).

Total protein was isolated from jejunal, ileal, and colonic tissue samples. Tissue samples (30 mg) were immersed in 300 μL of RIPA buffer containing 150 mmol/L sodium chloride, 1.0% IGEPAL CA-630, 0.5% sodium deoxycholate, 0.1% SDS, 0.1% protease inhibitor cocktail, and 50 mmol/L Tris, pH 8.0 (Sigma-Aldrich; #R0278). Samples were homogenized using the QIAGEN Tissue LyserLT (Qiagen) at 50 Hz for 5 minutes. Homogenates were centrifuged at 11,000 × g for 15 minutes at 4°C, and supernatant was collected, aliquoted, and stored at −80°C. Total protein was quantified using the Pierce BCA protein quantification kit (ThermoFisher Scientific,; #23225) relative to a 8-point standard curve (BSA: 25 μg/mL–2 mg/mL).

Total protein lysates were quantified and 30 μg supplemented with 4 μL Bolt reducing agent (ThermoFisher Scientific; #B0008) and 10 μL Bolt lithium dodecyl sulfate (LDS) sample buffer (ThermoFisher Scientific; #B0007). The total volume was then adjusted to 40 μL with milliQ water. Samples were denatured for 10 minutes at 70°C before being loaded into precast Bolt 4% to 12% Bis–Tris Plus, SDS-PAGE Gels (12-well) (ThermoFisher Scientific; #NW04122BOX). Samples were separated using the Bolt mini gel tank (ThermoFisher Scientific; #A25977) at 150 V for 45 minutes. Proteins were transferred to a polyvinylidene difluoride membrane using the iBlot transfer stacks (ThermoFisher Scientific; #IB24002) and compatible iBlot 2 gel transfer device (ThermoFisher Scientific; #IB21001). The membrane was washed with 1× TBS and Tween-20 (TBST) and stained with Ponceau S red staining solution (Sigma-Aldrich; #09276) to confirm equal loading. The membrane was blocked and probed with primary and secondary antibodies using the iBind western device (ThermoFisher Scientific; #SLF1000) as per manufacturer's guidelines. Western blots were assessed using ImageStudio Lite, version 4.0. Signal intensity was determined relative to local background. Data were presented relative to loading control (GAPDH) and vehicle-treated controls.

Antibody details are as follows: claudin-1 (Invitrogen 51-9000; 5 μg/mL; 1:50; IRDye anti-rabbit 800 CW); ZO-1 (Invitrogen 61-7300; 6.25 μg/mL; 1:40; IRDye anti-rabbit 800 CW); occludin (Invitrogen 71-1500; 1.25 μg/mL; 1:400; IRDye anti-rabbit 800 CW); and GAPDH (Abcam ab15822; 0.5 μg/mL; 1:2000; IRDye anti-chicken 680 CW).

Total RNA was isolated from jejunal, ileal, and colonic mucosal scrapings samples using the Macherey-Nagel NuceloSpin RNA/Protein Purification Kit as per manufacturer's instructions (Macherey-Nagel; #740933.250). Once eluted, RNA was stored at −80°C. Total RNA yield and purity were assessed using the BioTek Synergy Mx Microplate Reader (BioTek), TAKE3 plate, and Gen5 (version 2.00.18) software.

Five hundred nanograms of total RNA was reverse transcribed using the iScript cDNA Synthesis Kit (BioRad; #1708890) as per manufacturer's instructions. RT-PCR was performed using the Rotor-Gene 3000 (Corbett Research). Amplification mixes contained 1 to 2 μL of cDNA sample (100 ng/μL), 5 μL of SYBR green fluorescence dye, 2 to 3 μL of RNase-free water, and 0.5 μl of each forward and reverse primers, prediluted to 50 pmol/μl, to make a total volume of 10 μL.

Primer details are as follows: mouse claudin-1 (NM_016674.4) QIAGEN Mm_Cldn1_1_SG QuantiTect Primer Assay (QT00159278); T_m 55°C; mouse ZO-1 (NM_001163574.1) F: 5′-ACTCCCACTTCCCCAAAAAC-3′; R: 5′-CCACAGCTGAAGGACTCACA-3′; 166 bp; T_m 52/54°C; (30); mouse occludin (NM_008756.2) F: 5′-ACTGGGTCAGGGAATATCCA-3′; R: 5′-TCAGCAGCAGCCATGTACTC-3′; 192 bp; T_m 52/54°C; (30); mouse GAPDH (NM_008084.2) F: 5′-CCTCGTCCCGTAGACAAAATG-3′; R: 5′-TCTCCACTTTGCCACTGCAA-3′; T_m 52/54°C; (31); mouse 18S rRNA (NR_003278.3) F: 5′-TCGGAACTGAGGCCATGATT-3′; R: 5′-TTTCGCTCTGGTCCGTCTTG-3′; 100 bp; T_m 52/54°C; (31).

Thermal cycling conditions for ZO-1 and occludin included a denaturing step at 95°C for 15 minutes, followed by 45 cycles of denaturation at 95°C for 10 seconds, annealing at 52°C for 15 seconds, and extension at 72°C for 20 seconds. Claudin-1 cycling conditions were as follows: denaturing step at 95°C for 15 minutes, followed by 45 cycles of denaturation at 94°C for 15 seconds, annealing at 55°C for 30 seconds, and extension at 72°C for 30 seconds. All samples were run in triplicate with the inclusion of blank, negative controls (no cDNA template). Primer efficiency was evaluated using standard curves, and experimental threshold (C_T) values were calculated by the Rotor Gene 6 programme. C_T values were used to quantify relative mRNA expression of each tight junction protein using the ΔC_T method, where relative expression = |{2^{ - [ {{C_{\rm{T}}}( {{\rm{target}}} ) - {C_{\rm{T}}}( {{\rm{housekeeper}}} )} ]}}$|⁠. 18S and GAPDH were assessed for their suitability as housekeeping/reference genes (32).

Immediately following dissection, samples of distal colon were mounted into Ussing chambers (Physiologic Instruments; #EM-CSYS-8) for electrophysiological analyses. Briefly, the colon was cut longitudinally along the mesenteric attachment and placed onto a 0.1-cm² aperture slider (Physiologic Instruments; P2303A). The tissue was mounted and continually bathed in an oxygenated, glucose-fortified Ringer solution at 37°C with the following composition (composition in mmol/L/L: NaCl 115.4; KCl 5; MgCl₂ 1.2; NaH₂PO₄ 0.6; NaHCO₃ 25; CaCl₂ 1.2; and glucose 10). Tissues were voltage-clamped to zero potential difference by the application of short-circuit current (I_sc) and baseline was established. Tissues were allowed to equilibrate for 20 minutes and baseline I_sc (μA/cm²) and conductance (S/cm²) were recorded. Tissues were pretreated with amiloride (20 μmol/L) to inhibit the apical epithelial sodium channel before being treated with forskolin (adenosine 3′5′-cyclic monophosphate agonist; 10 μmol/L) and carbachol (Ca²⁺ agonist; 100 μmol/L), applied to the apical chamber. The I_sc response was then measured and determined as the change in I_sc following agonist administration (ΔμA/cm²), representing stimulated chloride secretion.

Data were compared using Prism version 7.0 (GraphPad Software). A D'Agostino–Pearson omnibus test was used to assess normality. When normality was confirmed, 2-way ANOVA with appropriate post-hoc testing was performed to identify statistical significance between groups. In other cases, a Kruskal–Wallis test with Dunn multiple comparisons test and Bonferroni correction were performed. P < 0.05 was considered significant.

Qualitative analysis of IF for tight junction proteins showed membranous staining for all tight junction proteins in vehicle-treated WT and BALB/c-Tlr4^−/−billy mice. No apparent differences were noted in any tight junction protein between genotypes receiving vehicle (Figs. 1–3; vehicle control).

Claudin-1 staining presented with sharp apical intensities (Fig. 1B and H; arrowheads; red staining) and membranous staining down the apicolateral border of the enterocyte. Marked claudin-1 internalization was evident at 24 hours after irinotecan treatment in WT mice (Fig. 1C and D, red staining; arrows). This was particularly evident in the ileum (Fig. 1C and D), with complete loss of membranous staining specificity in some areas. Claudin-1 internalization was seen as uniform cytoplasmic staining. Evidence of claudin-1 internalization was also seen in the colon of WT animals following irinotecan treatment, characterized by cytoplasmic staining and loss of membrane specificity (Fig. 1K and L). The degree of claudin-1 redistribution was comparatively less in BALB/c-Tlr4^−/−billy mice treated with irinotecan in both gut regions. Membranous staining remained intact, although loss of apical staining intensity was evident (Fig. 1G, H, O, and P).

IF staining for ZO-1 (Fig. 2; red staining) showed focal areas of proteolysis in the ileum (Fig. 2D; arrowheads) and colon (Fig. 2L; arrowheads) of WT mice 24 hours after irinotecan. These focal areas of protein disruption were particularly evident in areas of epithelial injury (identified in our previous study; ref. 24), often occurring alongside phenotypically normal tight junction staining (Fig. 2D and L; arrows). Similar changes in occludin expression were also seen in the ileum and colon of WT mice, with focal areas of proteolysis corresponding with frank epithelial damage (Fig. 3D and L, arrowhead; green staining). Staining appeared uniform for BALB/c-Tlr4^−/−billy mice treated with irinotecan (Fig. 2G, H, O and P; Fig. 3G and H, O/P).

Despite significant redistribution of claudin-1 and focal areas of ZO-1/occludin disruption following irinotecan treatment, Western blot analysis revealed no quantitative changes in tight junction protein expression in any mouse group (Fig. 4).

Average RNA yield using the Macherey-Nagel NuceloSpin Protein/RNA Isolation Kit was 464.44 ± 276.47 ng/μL with an average A^260/280 of 2.07 ± 0.06. cDNA conversion resulted in an average yield of 2,077.69 ± 347.44 ng/μL with an average A^260/280 of 1.76 ± 0.11. Linear regression slope analysis revealed suitable primer efficiencies for ZO-1 (relative to 18S: 0.083; relative to GAPDH: −0.080) and occludin (relative to 18S: −0.084; relative to GAPDH: −0.080). Linear regression slope analysis for claudin-1 showed disparate primer efficiencies relative to each housekeeper. Housekeeping stability was analyzed using the ΔC_T method, where relative expression = |{2^{ - [ {{C_{\rm{T}}}( {{\rm{housekeeper\ in\ controls}}} )\ - \ {C_{\rm{T}}}( {{\rm{housekeeper\ in\ treated}}} )} ]}}$|⁠. 18S was eliminated as a suitable housekeeper on the basis of differential expression in vehicle- and irinotecan-treated mice (data not shown).

Data showed no change in the relative mRNA expression of occludin or claudin-1 in WT or BALB/c-Tlr4^−/−billy mice following irinotecan treatment (Fig. 5). ΔC_T analysis revealed a significant decrease in the mRNA expression of ZO-1 (relative to GAPDH) in the colon 24 hours after irinotecan treatment in WT mice (Fig. 5H).

Ussing chamber studies indicated changes in baseline short-circuit current (I_sc) and conductance in irinotecan-treated WT mice (Fig. 6A–D). There was a significant difference in baseline I_sc between WT and BALB/c-Tlr4^−/-billy mice 24 hours after irinotecan treatment (Fig. 6A; WT: 103.65 ± 36.32 μA/cm²; BALB/c-Tlr4^−/−billy: 28.6727 ± 10.03; *, P = 0.0168). Increased baseline conductance, a measure of paracellular permeability, was seen in WT mice at 24 hours after irinotecan treatment (Fig. 6B and D; 53.19 ± 6.46 S/cm², +105.62% relative to WT vehicle controls; #, P = 0.0008). There was no change seen in BALB/c-Tlr4^−/−billy mice at any time after irinotecan compared with vehicle controls.

WT and BALB/c-Tlr4^−/−billy mice exhibited increased ΔI_sc after carbachol administration 24 hours (Fig. 7A; WT 100.35 ± 18.37 μA/cm²; #, P = 0.022; BALB/c-Tlr4^−/−billy: 102.72 ± 18.80 μA/cm²; ⁁, P = 0.023) and WT at 48 hours (99.75 ± 25.22 μA/cm²; #, P = 0.0244) after irinotecan. Irinotecan-treated WT mice also showed increased ΔI_sc in response to forskolin at 72 hours (82.18 ± 16.54 μA/cm²; #, P = 0.025). No change was seen in irinotecan-treated BALB/c-Tlr4^−/−billy mice in response to forskolin compared with vehicle controls (65.60 ± 18.28 μA/cm²; P = 0.999). Unstimulated (baseline) current and secretory responses to both carbachol/forskolin did not correlate with diarrhea severity (Fig. 7C and D; baseline: r² = 0.08, carbachol: r² = 0.04, forskolin: r² = 0.04). # denotes a change relative to untreated vehicle controls in WT mice; # denotes a change relative to untreated vehicle controls in BALB/c-Tlr4^−/−billy mice; * denotes a significant difference between WT and BALB/c-Tlr4^−/−billy mice, where P < 0.05.

Irinotecan is a commonly prescribed chemotherapeutic agent; however, its therapeutic efficacy is often limited by its severe gastrointestinal side effects, notably late-onset diarrhea (4). Diarrhea significantly impacts on patients' clinical outcomes (5); however, the underlying mechanisms remain unclear and therapeutic interventions limited (33). It is becoming increasingly clear that irinotecan causes significant intestinal barrier disruption, characterized by increased paracellular permeability and tight junction breakdown (7, 24). Our previous research has shown that BALB/c-Tlr4^−/−billy mice are protected from developing irinotecan-induced intestinal barrier disruption, supporting the idea that TLR4-dependent mechanisms are critical in tight junction disruption (24). Results from the current study support this hypothesis, showing improved tight junction integrity in irinotecan-treated BALB/c-Tlr4^−/−billy mice compared with their WT counterparts. This is the first study to compare functional in vivo permeability, ex vivo electrophysiological measures of barrier function, and morphological assessment of tight junction proteins in the setting of irinotecan-induced gut toxicity.

Tight junctions are critical in maintaining gastrointestinal health and homeostasis. Despite this, they are highly plastic structures vulnerable to posttranscriptional and -translational modification by a variety of pathological cues (34, 35). Tight junction disruption has been identified following treatment with a number of chemotherapeutic agents, both preclinically (6, 7) and clinically (12, 15, 36); however, the mechanisms that underpin their breakdown remain unclear. To date, many studies have shown architectural abnormalities (15), functional alterations (12, 37), and downregulation of key tight junction proteins such as claudin-1, ZO-1, and occludin (7, 18). The present study did not identify any alterations in the protein expression of these tight junction units in any region of the gut. Instead, significant derangement of these proteins was identified, characterized by severe cytoplasmic redistribution and disassembly of the tight junction unit. Internalization of tight junction proteins is well-recognized to contribute to poor barrier function and loss of tight junction apposition (35). In the current study, cytoplasmic redistribution of claudin-1 was seen at 24 hours; where peak barrier dysfunction and diarrhea were detected.

In addition to demonstrating tight junction disruption in response to irinotecan, this study has also provided evidence implicating TLR4-dependent mechanisms in tight junction regulation. This is a growing area of research, with the interaction between TLR4 and its ligand, LPS, as well as its powerful downstream effects on inflammation providing a strong rationale for its involvement. Evidence suggests that LPS/TLR4-dependent tight junction disruption occurs via direct epithelial processes (27). This has been shown using in vitro models of LPS-induced barrier dysfunction, with LPS administration resulting in TLR4-dependent activation of focal adhesion kinase (FAK) and tight junction disruption in Caco-2 monolayers. Importantly, siRNA silencing of TLR4 prevented LPS-induced disruption. This was also confirmed with in vivo, intestinal epithelial-specific knockdown of TLR4.

Despite a growing body of evidence showing direct TLR4-mediated regulation of tight junctions, it is likely that this mechanism occurs in concert with cytokine-mediated tight junction disruption. Evidence exists supporting a role for proinflammatory cytokine-mediated tight junction disruption (35, 38) explaining the observation of intestinal barrier disruption in diseases characterized by inflammation (21, 39). Conversely, evidence also suggests that loss of anti-inflammatory cytokines can be detrimental in regulation of barrier function (40, 41). This molecular crosstalk between inflammatory cytokines and tight junctions is compelling with regards to irinotecan-induced gut toxicity, as peak IL1β, IFNγ, and TNF levels coincide with intestinal barrier dysfunction and cytoplasmic redistribution of claudin-1 in WT mice treated with irinotecan (24). Although associative links have been identified between a proinflammatory state and tight junction disruption, much of the research has been conducted in in vitro models as in vivo models can be complicated by cytokine-dependent immune cell recruitment and activation within the mucosa (11).

A particularly strong case for IL1β-mediated tight junction disruption exists given TLR4 acts as its main upstream regulator (42). Addition of IL1β to growth media has been shown to directly increase epithelial permeability and decreased expression of key tight junction proteins in Caco-2 cells (representative of the small intestinal epithelium; ref. 43). Similarly, TNF and IFNγ have been shown to synergistically disrupt tight junctions through protein rearrangement and subsequent phosphorylation of myosin II regulatory light chain (MLC; refs. 44, 45). Given that BALB/c-Tlr4^−/−billy mice showed no significant change in IL1β, TNF, and IFNγ (24), it is possible that proinflammatory cytokine production drives barrier dysfunction in a TLR4-dependent manner. Furthermore, the barrier protective effects of anti-inflammatory cytokines, such as IL10, also warrant further investigation as research suggests these cytokines are downregulated following chemotherapy (46).

The clinical consequences of intestinal barrier dysfunction are becoming increasingly recognized in the setting of chemotherapy-induced gut toxicity. We have shown significant barrier dysfunction at 24 hours which precedes serum detection of LPS (24), supporting the study by Blijlevens and colleagues that describes that loss of barrier function is critical in preventing secondary, systemic toxicity. LPS translocation has also been suggested to underpin cancer-related sickness behavior (47) and secondary toxicities such as chemotherapy-induced pain (48). In addition to permitting LPS translocation, loss of barrier integrity has also been shown to result in bacterial translocation and colonization in mesenteric lymph nodes and the spleen following treatment with irinotecan, thus increasing the risk for infection, graft-versus-host disease (49), and sepsis (22). It is also likely that loss of paracellular integrity allows greater exposure, and subsequent activation, of the innate mucosal immune system, driving a heightened inflammatory response following cytotoxic insult. Barrier dysfunction is therefore likely to exacerbate direct cytotoxic injury in the gut, thus worsening clinical outcomes for patients. By understanding the mechanisms that lead to barrier dysfunction, therapeutic interventions may be targeted to prevent local toxicity transitioning to systemic toxicity, reducing the associated risk such as infection, sepsis, and pain.

Our previous research shows that intestinal permeability best reflects the clinical progression of diarrhea following irinotecan treatment (24). This parallels clinical findings in patients with inflammatory bowel disease, with barrier function and tight junction integrity correlating with the onset, severity, and duration of diarrhea (20). A wealth of studies show associative links between barrier dysfunction and diarrhea; however, the mechanisms by which barrier dysfunction contributes to diarrhea remain unclear. It has been suggested that passive “leak flux” mechanisms may be involved (21), where loss of barrier integrity allows passive leakage of solute and water into the lumen of the gut. The current study did not assess passive ion and/or water movement following irinotecan treatment; however, electrophysiological analyses of the distal colon revealed hyperresponsiveness following treatment with chloride secretagogues in both animal strains following irinotecan. This suggests an active secretory component, independent of TLR4, may be involved in the development of irinotecan-induced gut toxicity. However, dissecting its role in diarrhea development is difficult as neither secretagogue-induced chloride secretion nor unstimulated baseline current correlated with diarrhea. In addition, conclusions cannot be drawn regarding paracellular sodium movement, regulated by the pore-forming claudin-2 protein (50). Future studies should therefore focus on the synergism between barrier dysfunction, active chloride secretion, and paracellular sodium/water fluxes in chemotherapy-induced diarrhea.

Tight junction disruption is a hallmark trait of many pathological states. A wealth of research now implicates poor tight junction integrity following treatment with various chemotherapeutic agents. We have shown that irinotecan treatment causes tight junction disruption, characterized by claudin-1 internalization and barrier dysfunction via TLR4-dependent mechanisms. We have outlined mechanisms by which TLR4 may regulate tight junction disruption (i) through direct epithelial events mediated through luminal LPS and (ii) through its downstream effects on inflammation. In reality, it is likely that these mechanisms occur in concert with one another given the highly multifactorial biology of irinotecan-induced gut toxicity. Nonetheless, TLR4 appears to be critical in the development of barrier dysfunction, with overarching effects on gut toxicity. In addition, this study shows active chloride secretion following irinotecan treatment, irrespective of TLR4, and provides a novel avenue for the treatment of irinotecan-induced diarrhea.

R.J. Gibson reports receiving a commercial research grant from Onyx Pharmaceuticals and has received speakers' bureau honoraria from and is a consultant/advisory board member for Mundipharma. No potential conflicts of interest were disclosed by the other authors.

Conception and design: H.R. Wardill, J.M. Bowen, J.K. Coller, R.J. Gibson

Development of methodology: H.R. Wardill, J.M. Bowen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H.R. Wardill, Y.Z.A. Van Sebille, K.R. Secombe, I.A. Ball

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H.R. Wardill, J.M. Bowen, J.K. Coller, R.J. Gibson

Writing, review, and/or revision of the manuscript: H.R. Wardill, J.M. Bowen, Y.Z.A. Van Sebille, J.K. Coller, R.M. Logan, R.J. Gibson

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y.Z.A. Van Sebille, I.A. Ball

Study supervision: J.M. Bowen, R.M. Logan, R.J. Gibson

We would like to thank Mr. Anthony Wignall for his help conducting the animal study, as well as Professor Paul Foster from the University of Newcastle for supplying the TLR4-null mice.

The study was supported by Florey Medical Research Foundation Doctor Chun Chung Wong, Madam So Sau Lam Memorial Postgraduate Cancer Research Top Up Scholarship 2015/16, and Australian Postgraduate Award (to H.R. Wardill and Y.Z.A Van Sebille), and Ray and Shirl Norman Cancer Research Trust Project Grant (to R.J. Gibson, J.K. Coller and J.M. Bowen).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.