Intestinal Epithelial Cell–Derived LKB1 Suppresses Colitogenic Microbiota

Liver kinase B1 (LKB1), also called serine-threonine kinase 11, plays an essential role in controlling the physiological function of the intestinal epithelium (1–4). Highly expressed in the secretory lineage cell types of the intestinal epithelium, LKB1 is required for the differentiation of goblet cells and Paneth cells (5). Previous studies have shown that deletion of LKB1 in intestinal epithelial cells (IECs) increases the number of secretory lineage cells and impairs the endocrine function of the intestinal epithelium, indicating that LKB1 is a central regulator of intestinal physiology (5). However, whether and how LKB1 plays a role in the immune barrier function of the intestinal epithelium remains unknown. The immune barrier function of the intestinal epithelium has received intense attention lately, as the loss of the immune barrier function often leads to expansion of pathogenic bacteria in the commensal community, resulting in susceptibility to diseases such as colitis and tumorigenesis (6–11).

In this study, we examined the impact of intestinal epithelial-specific LKB1 deficiency on the immune barrier function of the gut epithelium. Intestinal epithelial-specific LKB1 deficient mice exhibited an exaggerated response to dextran sodium sulfate (DSS) challenge in a microbiota-dependent manner. Pyrosequencing analysis of the bacterial community in fecal samples revealed a definitive shift in in the microbiota composition in IEC-specific LKB1 knockout (LKB1^ΔIEC) mice compared with wild-type (WT) littermate controls (LKB1^WT). Transfer of the microbiota from the LKB1^ΔIEC mice by fecal transplant or cohousing sensitized WT mice to DSS-induced colitis. Furthermore, the production of IL-18 and antimicrobial peptides (AMPs) was significantly reduced in the LKB1-deficient colon epithelium. Administration of IL-18 restored the expression of AMPs in the LKB1-deficient colon epithelium, reduced the abundance of the outgrown genera, and rescued intestinal epithelial-specific LKB1-deficient mice from increased sensitivity to DSS challenge. Taken together, our study reveals an IL-18–dependent function of LKB1 in suppressing colitogenic microbiota.

Villin-cre (category: 004586) mice and LKB1^fl/fl (category: 014143) mice were purchased from the Jackson Laboratory and backcrossed onto a C57BL/6 background for at least six generations. We crossed Villin-cre mice with LKB1^fl/fl mice to generate Villin-cre/LKB1^fl/fl mice (in which LKB1 was specifically deleted in the intestinal epithelium) and LKB1^fl/fl mice as littermate controls. All mice were bred and maintained in individually ventilated cages under specific pathogen-free conditions in accredited animal facilities. For cohousing experiments, weaning age LKB1^ΔIEC and littermate control pups of the same sex were sorted into the same cage. Sex-matched mice were used between 8 and 10 wk of age. All animals were housed in an American Association for the Accreditation of Laboratory Animal Care International–accredited facility. Experimental colonies tested negative for the presence of Helicobacter and murine norovirus. All experiments were approved by the Institutional Animal Care and Use Committee of Shanghai Jiao Tong University School of Medicine.

Sex-matched LKB1^WT and LKB1^ΔIEC mice at the age of 8–10 wk were fed with 2.5% DSS (MP Biomedical, unless otherwise specified) in normal drinking water for 7 d, followed with normal drinking water for another 2 d to induce acute colitis. Mouse weights were monitored every day during the colitis model. The humane endpoints for DSS study, in accordance with Institutional Animal Care and Use Committee recommendations from Shanghai Jiao Tong University, was the loss of 25–30% of initial body weight. Scoring for stool consistency was performed as previously described (12, 13). In brief, stool scores were determined as follows: 0) well-formed pellets; 1) soft but still formed; 2) semiformed stools that adhered to the anus; 3) liquid stools that adhered to the anus.

Mouse endoscopy and murine endoscopic index of colitis severity (MEICS) scoring was performed as previously described (14, 15). In brief, mice were anesthetized by i.p. injection of 0.8% phenobarbital sodium (200 μl in normal saline) before colonoscopy was performed using a high-resolution mouse endoscopic system (Carl Storz, Tuttlingen, Germany). The severity of colitis was scored using MEICS based on the following parameters: translucency of the colon mucosa (0, transparent; 1, moderate; 2, marked; 3, nontransparent); vascular pattern (0, normal; 1, moderate; 2, marked; 3, bleeding); fibrin visible (0, none; 1, little; 2, marked; 3, extreme); granularity of mucosal surface (0, none; 1, moderate; 2, marked; 3, extreme), and stool consistency (0, normal+solid; 1, still shaped; 2, unshaped; 3, spread).

Intestinal permeability in vivo was monitored by FITC-dextran (average molecular mass 4000 kDa; Sigma-Aldrich) gavage experiment under naive and colitis conditions. In brief, mice were fasted overnight before gavaging with FITC-dextran at 60 mg per 100 g body weight. Then 4 h later, blood samples were collected by cardiac puncture, and sera were obtained after centrifuging at 10,000 × g for 10 min. Fluorescence intensity of serum sample was measured (SpectraMax i3; Molecular Devices), and the concentration of FITC-dextran was determined according to the standard curve generated by the serial dilution of FITC-dextran.

Crypts were isolated as previously described (16). In brief, the large intestine was gently pulled out of the abdominal cavity. Mesentery and fatty tissue were removed with forceps. Large intestine without cecum was dissected, opened longitudinally, and chopped into 1 cm pieces. Intestinal pieces were washed thoroughly with ice-cold PBS three times. After thorough washing in PBS, intestinal pieces were further chopped into 5 mm pieces and incubated in 5 mM EDTA/PBS at 4°C on a rocking platform for 30 min. Crypts were released by shaking the tubes for 2 min and were collected by centrifuging at 200 × g for 10 min at 4°C.

Isolation of lamina propria cells were modified from the protocol described previously (17). In brief, large intestines devoid of Peyer’s patches were gently pulled out of the abdominal cavity. Mesentery and fatty tissue were removed with forceps. The intestines were opened longitudinally and chopped into 1 cm pieces. Intestinal pieces were washed thoroughly with room temperature PBS three times. After thorough washing in PBS, intestinal pieces were incubated in 30 mM EDTA/10 mM HEPES/PBS at 37°C on a rocking platform for 10 min. The epithelial cell layer was removed by shaking for 2 min. Remaining intestinal pieces were further chopped in 5 mm pieces and were digested in collagenase VIII (200 U/ml; Sigma-Aldrich) and DNaseI (150 μg/ml; Sigma-Aldrich) in RPMI 1640 complete medium for 50 min. Then cells were resuspended in 4 ml of 40% Percoll (GE Healthcare) and overlaid onto 2.5 ml of 80% Percoll in a 15 ml tube. Cells were collected at the interphase of the Percoll gradient, washed once, and resuspended in PBS containing 1% FBS. Two million isolated lamina propria cells were processed for flow cytometry and two million were processed for inflammatory cytokines analysis by quantitative real-time PCR (Q-PCR).

Whole colons were removed from given mice from an LKB1^WT or LKB1^ΔIEC genotype, rinsed with serum-free DMEM medium, and weighed. The tissue explants were then cultured for 24 h in regular RPMI 1640 medium with 10% FBS, l-glutamine, penicillin, and streptomycin in a routine cell-culture incubator. Cell-free supernatants were collected by centrifuging at 12,000 × g for 10 min at 4°C and stored in aliquots at −20°C.

Isolated lamina propria cells were resuspended in PBS containing 1% FBS and stained with fluorescence-conjugated mAbs to CD4 (clone GK 1.5), CD45 (clone 30-F11), and Ly6G (clone 1A8) and isotype controls, purchased from BD Biosciences. F4/80 (clone Cl: A3-1) was obtained from Serotech. Abs were diluted at 1:100 when used. Stained cells were analyzed on an LSRFortessa X-20 cytometer and data were analyzed with FlowJo software.

Fecal microbial transplantation (FMT) was performed according to previous reports (18, 19). In brief, recipient mice were pretreated with a mixture of neomycin (1 g/l), ampicillin (1 g/l), metronidazole (0.5 g/l), and vancomycin (0.5 g/l) in normal drinking water for 2 wk or untreated. Recipient mice received FMT every other day for a total of three times before DSS-induced acute colitis. Feces were collected from LKB1^WT or LKB1^ΔIEC donor mice, mixed in PBS (50 mg/ml), and each WT mouse received 300 μl by oral gavage.

Tissues or isolated crypts were lysed with radioimmunoprecipitation assay buffer supplemented with protease inhibitor mixture (Roche) on ice for 30 min. Supernatants were collected after centrifuging at 13,000 rpm for 15 min, 4°C. Western blot analysis was performed using the indicated Abs including LKB1 (1:1000), cleaved caspase-3 (1:1000) from Cell Signaling Technology, β-actin (1:2000; Santa Cruz), and HSP90 (1:2000; Proteintech).

Whole colon tissues or cells were preserved and homogenized in TRIzol reagent (Invitrogen). RNA was extracted according to the manufacturer’s instructions and reverse transcribed into cDNA immediately. Q-PCR was performed with SYBR Green Real-time PCR Master Mix (Toyobo) on a ViiA7 Real-Time PCR System (Applied Biosystems). The primer sequences used are shown in the supplemental information.

For immunohistological analysis, tissue collected from LKB1^WT and LKB1^△IEC mice were fixed in 4% paraformaldehyde, processed, paraffin embedded, and sectioned (5 μm). For immunohistochemical staining, sections were de-paraffinized and Ag retrieval was performed. After permeabilization and blocking, the sections were incubated with primary Abs (anti-CD45, 1:100, MCA1388, Bio-Rad; anti-Ki67, 1:1000, 12202S; CST) overnight at 4°C. HRP-conjugated secondary Ab kits (PV-9004, SP-9001; ZSGB-BIO) were used and the HRP enzymes were developed using DAB substrate kit (550880; BD Pharmingen). The sections were counterstained with hematoxylin, then dehydrated in alcohol and xylene. Quantification of CD45⁺ leukocytes was performed by counting CD45⁺ cells in 10 high-powered fields from each mouse.

For H&E staining, ∼3 mm of the small and large intestine were fixed in 4% paraformaldehyde and embedded into paraffin, then sections were stained with H&E for histological analysis. Sections were scored in a blinded manner on a scale from 0 to 4, based on previous reports (20). In brief: 0, no inflammatory cell infiltration or occasional inflammatory cell in the lamina propria with no mucosal damage; 1, mild inflammatory cell in the lamina propria with mild mucosal damage; 2, moderate inflammatory cell in the lamina propria with moderate mucosal damage; 3, marked inflammatory cell in the lamina propria with marked mucosal damage; 4, transmural extension of inflammatory cell in the lamina propria with extensive mucosal damage and extension into deeper structures of the bowel wall.

Metformin (Beyotime Biotechnology) was reconstituted in normal saline. Mice were administered metformin by daily i.p. injection at 250 mg/kg body weight consecutively for 7 d before being subjected to DSS-induced acute colitis.

Fecal samples were harvested from colons of untreated LKB1^WT and LKB1^ΔIEC mice. DNA was extracted with the QIA FastDNA Stool Mini kit (Qiagen). The extracted DNA was sent to the Beijing Genomics Institute (Shenzhen, Guangdong, China) for sequencing of the V3–V4 hypervariable region of the 16S rRNA gene using primers 341 forward (5′-CCTACGGGNGGCWGCAG-3′) and 802 reverse (5′-TACNVGGGTATCTAATCC-3′) incorporating FLX Titanium adapters and sample barcode sequences (21). Paired-end reads of 250 bp each were generated with the Illumina MiSeq platform from high-quality samples. Sequencing data were demultiplexed, trimmed, and quality controlled with default parameters in Mothur (v1.31.2). Overlapping reads were generated with FLASH (v1.2.11) and clustered de novo to operational taxonomic units (OTUs) with USEARCH at 97% sequence similarity. Representative sequences were assigned a taxonomy using the Ribosomal Database Project Naive Bayesian Classifier v2.2. Chimeras were filtered out with UCHIME (v4.2.40) along with unassigned reads. Differential abundance analysis of OTUs between groups was calculated with Metastats and R (v3.1.1) after a Benjamini–Hochberg false discovery rate correction. For β diversity, weighted UniFrac distances were calculated and used to produce a principal coordinate analysis (PCoA) plot in the ade4 package in R, and an Analysis of Similarity (ANOSIM) was conducted with 999 permutations for the statistical comparison of groups. Phylum-level histograms based on relative abundance were drawn in R.

Isolation of bacterial genomic DNA was as previously described (13). In brief, mice were killed, then the colons were isolated. The luminal feces were collected. Genomic DNA was extracted from the bacterial pellets with the Qiagen Stool Kit according to the manufacturer’s instructions. Analysis of the abundance of specific intestinal bacterial groups was performed with SYBR Green Real-time PCR Master Mix (Toyobo) on a ViiA7 Real-Time PCR System (Applied Biosystems). Signals were normalized to universal bacterial and normalized data were used to calculate relative levels of 16S rDNA gene expression of indicated bacterial groups.

The p values of clinical scores were determined by two-way multiple-range ANOVA for multiple comparisons, which are specified in the figure legends. Values between three or more groups were determined using one-way ANOVA followed by Bonferroni post hoc test. Other p values were determined by nonparametric Mann–Whitney U test. Unless otherwise specified, all results are shown as mean ± SEM. Statistics for survival were analyzed by the Kaplan–Meier method. A p value < 0.05 was considered significant.

LKB1 is expressed in IECs and critically regulates intestinal homeostasis (5). To determine whether IEC-derived LKB1 plays a role in regulating the immune barrier function of the intestinal epithelium, we crossed Villin-cre transgenic mice with LKB1^fl/fl mice to generate IEC-specific LBK1-deficient mice (LKB1^ΔIEC) and littermate controls (LKB1^fl/fl, designated as LKB1^WT). The efficiency of deletion was validated by Western blot analysis of colonocyte lysate (Fig. 1A). We subjected separately housed LKB1^ΔIEC and littermate controls to DSS water and followed their survival and colitis. LKB1^ΔIEC mice exhibited more rapid weight loss and higher mortality (Fig. 1B, 1C) compared with littermate control mice. In addition, LKB1^ΔIEC mice showed more aggravated colitis symptoms including higher stool scores (Fig. 1D) and increased histological damage (Fig. 1E) compared with littermate controls. High-resolution mouse endoscopy revealed severe bleeding and loose stools in the live LKB1^ΔIEC mice, whereas the LKB1^WT controls only exhibited a mildly altered vascular pattern 8 d after DSS treatment (Fig. 1F). Collectively, the data indicate that LKB1 expression in the intestinal epithelium protects mice from DSS-induced colitis.

We compared the inflammatory response of LKB1^ΔIEC mice and that of littermate control mice in the colon after DSS challenge. The expressions of most proinflammatory cytokines and chemokines were comparable between the colons from LKB1^ΔIEC and littermate controls before DSS challenge (Fig. 2A, left panel). However, LKB1^ΔIEC exhibited much higher expression of these inflammatory genes 8 d after DSS treatment (Fig. 2A, right panel). Furthermore, the numbers of Ly6G⁺ neutrophils and F4/80⁺ macrophages were significantly higher in the LKB1^ΔIEC mice compared with littermate controls (Fig. 2B). The enhanced inflammatory response is associated with aggravated damage to integrity of the intestinal epithelium, as shown by the increased influx of FITC-dextran from the gastrointestinal tract into the systemic circulation (Fig. 2C). Consistent with the severe tissue damage, we also found increased activation of caspase-3 (Fig. 2D) in LKB1^ΔIEC mice compared with littermate controls. In addition, time-course analysis of colon sections using anti-Ki67 staining revealed a dramatically reduced number of proliferating cells in the LKB1^ΔIEC mice after DSS challenge (Fig. 2E). Taken together, our data suggest that expression of LKB1 in IECs is essential for controlling intestinal inflammation.

We surveyed the expression of a panel of inflammatory cytokines in the colons of LKB1^ΔIEC and littermate control mice in untreated and DSS-challenged colons (Figs. 2A, 3A). Interestingly, whereas most of the surveyed transcripts were elevated in the colons of LKB1^ΔIEC mice (Fig. 2A), IL-18 expression was significantly reduced in the colons of both untreated and DSS-challenged LKB1^ΔIEC mice compared with those of LKB1^WT controls (Fig. 3A). Colonic explants from untreated LKB1^ΔIEC mice also secreted less IL-18 protein after overnight culture (Fig. 3B). Because IECs are a major source of IL-18 in the colon, we examined expression of IL-18 in purified IECs from LKB1^ΔIEC and littermate LKB1^WT controls using Western blot. LKB1-deficient IECs showed diminished pro–IL-18 and mature IL-18 (Fig. 3C) compared with the IECs from littermate controls. Further intensity quantification showed the ratio of the pro- to cleaved forms of IL-18 are similar between LKB1^WT and LKB1^ΔIEC mice, suggesting that there is a defect primarily in the upregulation of pro–IL-18 (Fig. 3C). Taken together, the data indicate that LKB1 is required for pro–IL-18 expression in the IECs.

Recent studies have suggested that impaired IL-18 reduces the expression of AMPs such as Retnlß, Ang4, and Itln1 in the colon (22, 23). Indeed, we also found a significant decrease in the expression of Retnlß, Ang4, and Itln1 in IECs from untreated LKB1^ΔIEC mice (Fig. 3D). Reduced AMPs in the intestine are known to alter the configuration of gut microbiota (10, 24–26). Defective IL-18 production from IECs has been associated with dysbiosis, which confers colitis susceptibility. Real-time PCR analysis of 16S rDNA from fecal samples of untreated LKB1^ΔIEC and littermate control mice showed significant changes in the several known phyla present in the mouse intestine, with a marked increase of Proteobacteria, Actinobacteria, and Prevotella, and a noticeable decrease of Bacteroides, Firmicutes, and Lactobacillus in LKB1^ΔIEC mice (Fig. 3E). Collectively, the data suggest that LKB1 deficiency in IECs leads to an altered abundance of specific bacteria in the commensal community.

To comprehensively assess the impact of IEC-specific LKB1 deficiency on the gut microbiota composition, we performed pyrosequencing on fecal samples from separately housed LKB1^ΔIEC and littermate control mice. PCoA of the pyrosequencing data indicated that the commensal community in LKB1^ΔIEC mice was distinct from that of littermate controls (Fig. 4A, Supplemental Table I), both before and after DSS challenge. Interestingly, DSS-induced colitis promoted the outgrowth of specific bacterial taxa in LKB1^ΔIEC mice but not littermate controls (Fig. 4B). To exclude any possible impact of cage effects on the different microbiota composition between LKB1^ΔIEC and LKB1^WT mice, we performed pyrosequencing on fecal samples from four additional cages of untreated LKB1^ΔIEC and LKB1^ΔIEC mice. Consistently, PCoA analysis showed that the microbiota of the fecal samples clustered primarily in accordance with the mice genotypes (Fig. 4C, Supplemental Table I) but not cage allocation. The data suggest that LKB1 expression in IECs exerts a restraining force on the microbiota, preventing the colonization of a specific subgroup of bacteria. Outgrowth of specific pathogenic bacteria has been previously reported in several immune compromised mouse strains, including the IL-18–deficient mice. These bacteria can be transferred between hosts by cohousing as a result of coprophagia. Analysis of fecal samples from cohoused LKB1^ΔIECand LKB1^WT littermate control mice showed a converging trend in the composition of their microbiota (Fig. 4D, 4E, Supplemental Table I), suggesting that microbiota in the LKB1^ΔIEC mice is transmissible.

The reduced differences in the microbiota of cohoused LKB1^ΔIEC and littermate controls prompted us to test whether cohousing can change the colitis phenotype in LKB1^ΔIEC mice. We subjected cohoused LKB1^ΔIEC and littermate controls and separately housed LKB1^ΔIEC and littermate controls to a DSS-induced colitis model. Cohousing conferred the littermate control mice increased susceptibility to DSS challenge (Fig. 5A, 5B). The data led us to hypothesize that the microbiota in LKB1^ΔIEC was colitogenic. To test this hypothesis, we first performed FMT to transfer the bacterial community from the LKB1^ΔIEC mice to the LKB1^WT littermate controls. FMT led to a marked increase of Proteobacteria, Actinobacteria, and Prevotella, and a noticeable decrease of Bacteroides, Firmicutes, and Lactobacillus in LKB1^WT mice receiving LKB1^ΔIEC fecal homogenate (Supplemental Fig. 1A). The FMT with LKB1^ΔIEC microbiota conferred the LKB1^WT mice with increased sensitivity to DSS challenge (Fig. 5C, 5D), suggesting that the colitis sensitivity in LKB1^ΔIEC mice is imparted by the altered commensal community. To further test this hypothesis, we transferred the microbiota from littermates of separately housed LKB1^ΔIEC and LKB1^WT to unrelated WT recipients and tested their response to DSS challenge. FMT was able to recapitulate the differential abundance of specific taxa from donor mice in unrelated WT recipients (Supplemental Fig. 1B). Consistent with our hypothesis, fecal transplant with feces homogenate from separately housed LKB1^ΔIEC mice, but not LKB1^WT littermate controls, sensitized the unrelated WT recipient mice to DSS challenge (Fig. 5E–G). Finally, we tested whether antibiotic treatment can reduce the severity of colitis in LKB1^ΔIEC mice in response to DSS challenge. We pretreated the LKB1^ΔIEC mice with a mixture of antibiotics for 2 wk to deplete most of the bacteria. Pretreatment with an antibiotic mixture was sufficient to ameliorate the colitis phenotype in separately housed LKB1^ΔIEC mice (Fig. 5H, 5I). Taken together, our data indicate that LKB1 deficiency in the IECs nurtures a transmissible microbiota that confers increased sensitivity to DSS-induced colitis.

The association of a reduced expression of IL-18 and AMP levels in LKB1^ΔIEC mice suggests that IEC-derived LKB1 might be required for IL-18–mediated induction of AMPs. In support of this, administration of recombinant IL-18 into the LKB1^ΔIEC mice increased the expression of AMPs after 3 d of consecutive injection (Fig. 6A). We then performed a serial injection of recombinant IL-18 into LKB1^ΔIEC mice and examined the composition of the microbiota by 16S rDNA analysis. IL-18 administration reduced the abundance of bacterial strains outgrown in LKB1^ΔIEC mice (Fig. 6B) and partially desensitized the mice from DSS challenge (Fig. 6C–F, Supplemental Fig. 2). Taken together, the results indicate that IEC-derived LKB1 regulates the microbiota composition of the intestinal epithelium by promoting the expression of IL-18 and AMPs.

Lack of LKB1 signaling appears to be associated with decreased IL-18 expression, and it remains to be determined how LKB1 may regulate IL-18 production. LKB1 is an upstream kinase that phosphorylates and activates AMPK in response to low energy stress in many tissue types (27–29). Activation of AMPK leads to inhibition of mTOR complex 1, reducing energy consumption process including cell growth, and protein synthesis. We asked whether IEC-derived LKB1 protects mice from DSS-induced colitis through the LKB1-AMPK pathway. Western blot analysis of colonocyte lysates from untreated LKB1^ΔIEC and littermate controls did not show notable differences in AMPK phosphorylation, suggesting that the LKB1-AMPK axis may not be operative in colonocytes at the steady state (Fig. 7A). Nevertheless, we still attempted to pharmacologically activate AMPK in LKB1^ΔIEC mice to see if it can rescue the colitis phenotype with metformin, a Food and Drug Administration–approved AMPK activator. Pretreatment with metformin for 5 d induced AMPK phosphorylation in LKB1-deficient IECs (Fig. 7A). However, metformin treatment failed to attenuate the severity of colitis in LKB1^ΔIEC mice (Fig. 7B–E) or restore IL-18 expression (Fig. 7F). The data suggest that reactivating AMPK is not sufficient to restore the protective impact of LKB1 in IECs.

In this study, we describe a novel function of LKB1 in suppressing colitogenic microbiota by modulating IL-18 expression. We found that LKB1 deficiency in IECs was essential in suppressing intestinal inflammation in a microbiota-dependent manner. LKB1 deficiency in the intestinal epithelium nurtured colitogenic microbiota, which conferred increased susceptibility to colitis in WT recipients. We found that LKB1 expression in IECs was required for the production of IL-18 and AMPs. Administration of exogenous IL-18 restored the expression of AMPs in the LKB1-deficient colon epithelium, reduced the growth of the outgrown strains, and rescued the intestinal epithelial-specific LKB1-deficient mice from increased sensitivity to DSS challenge. Taken together, our study reveals a role for LKB1 in regulating the immune barrier function of the intestinal epithelium.

Ablation of LKB1 in epithelial cells significantly reduced the production of IL-18 and AMPs from IECs, potentially creating a permissive environment for the growth of the colitogenic bacteria. Recent studies reported that NLRP6 deficiency promotes the expansion of the colitogenic Prevotella as a result of defective IL-18 production from IECs (22). Interestingly, we also observed an increased abundance of Prevotella in LKB1^ΔIEC mice, further supporting the idea that LKB1 deficiency in IECs disabled the immune barrier function required for containing the pathogenic bacterial species. The primary source of IL-18 in the colon is the IECs (30). IL-18 has been shown to exert a protective role during the early phase of mucosal immune responses. The production of IL-18 is tightly regulated at LKB1-deficient IECs, suggesting that LKB1 may regulate IL-18 expression at the transcriptional level. LKB1 is also widely known as a tumor suppressor (31–34). LKB1 phosphorylates and activates AMPK when the adenosine 5'-monophosphate/ATP ratio is high. The phosphorylation of AMPK is crucial for its activation, which subsequently phosphorylates both tuberous sclerosis 2 and regulatory-associated protein of mTOR to inhibit mTOR complex 1 activity (29, 35–37). Even though our results suggested that activating AMPK fails to rescue the colitis phenotype in LKB1^ΔIEC mice, LKB1 might impact on other substrates thus modulating IL-18 production in the colon epithelium. Moreover, it should be mentioned that metformin is not a specific activator of AMPK. Metformin may induce multiple other signaling pathways as well (which may interfere with them), and its use may not have been ideal to try to answer our hypothesis. Future studies are required to determine the impact of LKB1 deficiency on the activation of downstream signaling pathways in the IECs to understand the molecular mechanism of LKB1-dependent IL-18 expression.

The bacterial community in the gut establishes a symbiotic relationship with the host, supplying critical metabolites and modulating the host immune system. Disruption of the symbiosis between the host and the microbiota, termed dysbiosis, is associated with the pathogenesis of a variety of diseases, including inflammatory bowel diseases (6, 7, 38, 39) and colon cancer (8, 40–42). Importantly, dysbiosis is not equivalent to the outgrowth or anguish of a single strain of bacteria; it represents an alteration of the whole bacterial community. In our study, we noted that the microbiota in cohoused LKB1^ΔIEC and littermate control mice showed a trend toward convergence but maintained a unique microbial signature, suggesting a possibility that the transmission of colitis susceptibility from LKB1^ΔIEC mice into cohoused LKB1^WT mice might be triggered by colitogenic microbiota from cohoused LKB1^ΔIEC mice. Two causal mechanisms are possible given our results. First, colitogenic bacteria from the LKB1^ΔIEC mice transfer to LKB1^WT mice and facilitate susceptibility to colitis directly, or bacteria from the LKB1^ΔIEC mice engage in microbe-microbe interactions with bacteria from the LKB1^WT mice to facilitate susceptibility.

Loss-of-function germline mutations of LKB1 are associated with Peutz–Jeghers syndrome (PJS), an autosomal dominant disorder. PJS patients develop characteristic hamartomatous polyposis and have a higher risk of developing colon carcinoma during their lifetime (43–46). Given the discoveries of our study, considerable effort should be dedicated to determining whether dysbiosis contributes to the increased cancer risk in PJS patients.

The authors have no financial conflicts of interest.