Loss of CLCa reduces post-natal survival and body weight

We have previously shown that CLCa KO and CLCb KO mice have phenotypes in immune cells and neurons (9, 22). To further characterise physiological functions of CLCs, heterozygous mice were crossed (Clta^+/− × Clta^+/− and Cltb^+/− × Cltb^+/−) and their offspring compared. When weaned (3 wk of age), only 9.7% of the Clta^+/− cross offspring were homozygous for CLCa KO, significantly lower than the expected 25% (P = 0.0185), indicating that loss of CLCa reduces survival at the pre- or neo-natal stage (Fig 1A). In contrast, 23.7% of Cltb^+/− cross offspring were homozygous for CLCb KO, close to the expected 25% (Fig 1A). Thus, loss of CLCa but not loss of CLCb reduces survival. To identify the developmental stage at which survival is affected, CLCa KO genotypes were analysed pre- and post-natally. At the late embryonic (E18.5) and early post-natal stage (PN1), we observed 30.1% and 27.2% homozygous CLCa KO (Clta^−/−), respectively (Fig 1B), indicating that loss of CLCa does not affect pre-natal survival. At post-natal day 3 (PN3), significantly fewer than expected mice were homozygous CLCa KO (<10%), which remained the case at PN7 and at 4-wk old (Fig 1B) and beyond (data not shown), indicating survival rates are unaffected by the loss of CLCa after PN3. Thus, CLCa has a critical role in post-natal mouse development and/or survival behaviour, such as response to critical feeding cues, within the first 3 d of birth.

The body weight of surviving CLCa KO and CLCb KO pups at PN1 was not significantly different from that of their respective WT and heterozygous littermates (Fig 1C). However, at PN7, both CLCa KO and CLCb KO pups had a significantly lower body weight than WT pups, respectively, 21.52% and 19.1% (Fig 1D). CLCa KO adult male and female mice (over 8 wk of age) maintained a significantly lower body weight than WT littermates, with male mice more affected (20.8% reduction in male CLCa KO body weight at 17–24 wk of age compared with a 12.8% weight reduction for female CLCa KO mice, Fig 1E). No significant difference in the body weight of CLCb KO mice compared with WT littermates above 8 wk of age was observed (Fig 1E).

To establish whether the observed mortality and body weight phenotypes were associated with gross morphological changes in the organs of the CLCa KO and CLCb KO mice, tissue slices from mice of each genotype were analysed by hematoxylin and eosin staining. No abnormalities in the structure or morphology of small intestine, skeletal muscle, kidney, and heart tissues from surviving CLCa KO adult mice or CLCb KO adult mice compared with WT adult mice (aged between 5 and 9 mo) were observed (Fig S1).

Loss of CLCa, but not CLCb, reduces fertility in mice

During routine maintenance of CLCa KO and CLCb KO mouse lines, we observed fewer than expected litters when CLCa KO mice were used for breeding. To analyse this further, breeding cages were set up with combinations of WT, heterozygous, and KO animals derived from heterozygous crosses (Clta^+/− × Clta^+/− or Cltb^+/− × Cltb^+/−), and the number of litters born within 54 d recorded (Fig 2A). Whereas WT breeding pairs produced 1−2 litters during this period, CLCa KO × CLCa KO breeding crosses produced no offspring at all, suggesting loss of CLCa affects fertility. In contrast, loss of CLCb did not reduce the number of litters born (Fig 2A). Crosses of heterozygous Clta^+/− mice with CLCa KO animals also generated a significantly reduced number of litters born within 54 d (Fig 2A). However, breeding WT with CLCa KO animals in either male/female combination produced litter numbers similar to WT breedings during the 54 d period (Fig 2A). A separate analysis over a 4 mo (120 d) period of breeding showed that when either CLC KO strain was bred with WT, no significant difference in the number of litters or neonates per litter was observed (Fig 2B and C). Thus, fertility was reduced between animals with no CLCa or between CLCa KO and animals (male or female) heterozygous for CLCa loss. Although this fertility defect was not observed when one mate was WT, these breeding experiments support that loss of CLCa affects fertility.

Loss of CLCa results in pyometra and reduction in uterine glands

Consistent with loss of CLCa-affecting reproduction, we observed that 42.9% of CLCa KO female mice developed a swollen abdomen after 4 mo of age due to an enlarged uterus containing cloudy fluid (Fig 3A and B). The uterus is composed of the outer myometrium and the inner endometrium compartments. The endometrium comprises the stroma, glands surrounded by glandular epithelium and the lumen surrounded by lumenal epithelium. H&E staining revealed gross structural abnormalities within the endometrium of CLCa KO mice with enlarged uteri. The cross-sectional width of the endometrium of these mice was narrower than that in WT animals and the endometrial glandular structures were lost (Fig 3C and D). Quantification confirmed that the number of glands in CLCa KO enlarged uteri was significantly lower than the number of glands in unaffected uteri from CLCa KO, CLCb KO or WT mice (Fig 3D).

An enlarged uterus filled with cloudy fluid can be indicative of pyometra, a uterine infection characterised by inflammation and a “pus-filled” uterus (24). To determine whether CLCa KO mice with an enlarged uterus had developed pyometra, the inflammation status of the uteri of these mice was examined and compared with unaffected uteri of all genotypes (WT, CLCb KO, and CLCa KO). When sections of uterus were stained with antibodies against the neutrophil marker Ly6G, an indicator of inflammation (25), Ly6G-positive cells were detected in the uterine epithelial layer of CLCa KO mice with an enlarged uterus but not in unaffected animals of any genotype (Fig 3E and F). This sign of neutrophil infiltration is consistent with pyometra as the cause of the enlarged uterus in CLCa KO mice.

To further understand the role of the CLCs in the uterus, we investigated the expression levels of CLCa and CLCb in the uterus. Relative levels of CLCa versus CLCb were determined for whole tissue lysates from the uterus, spleen, and brain of WT, CLCa KO (without pyometra), and CLCb KO mice by immunoblotting using the antibody CON.1, recognising the 22 amino acid CLC consensus sequence shared between CLCa and CLCb (26). Thus, the protein levels of the different CLCs can be directly compared within a blot. In the uterus of WT mice, CLCa protein levels were marginally higher than CLCb levels, with CLCa comprising 60.5% ± 3.5% of total CLCs, similar to their relative expression in the liver and not as extreme as CLCa dominance in the spleen (9) (Fig 4A and B). More balanced levels of the two CLC isoforms were observed for nCLCa and nCLCb in the brain (Fig 4A) and CLCa versus CLCb in muscle (9). Note that the apparent dominance of nCLCb in brain tissue is due to the presence of cells with non-neuronal CLCa that co-migrates with nCLCb during SDS–PAGE (as detected by CON.1 in the CLCb KO mice). Only a slight apparent increase in CLCb level was observed in the tissues of the CLCa KO mice, and vice versa, as noted in analysis of other tissues from these mice (9) (Fig 4A). Thus, development of pyometra correlates with CLCa loss and inability of residual or slightly elevated CLCb to support the clathrin function needed for the impaired pathway, revealing differences in the functions supported by the two CLC isoforms.

The relative localisations of CLCa and CLCb in the uterus of WT, CLCa KO, and CLCb KO mice were then determined by immunofluorescence. Both CLCa and CLCb were highly enriched in the lumenal and glandular epithelia of the uterus, with both concentrated at the apical pole of the epithelial cells (Fig 4C and D). This is consistent with a role for CLCa at apical domain of epithelia, previously suggested by the specific binding of the epithelial splice variant of myosin VI to CLCa (21) and suggests an additional role for CLCb.

CLCa loss alters endometrial epithelial cell identity

The advent of pyometra suggests a defect in epithelial integrity, leading to inflammation. To further assess the consequences of CLCa loss on endometrial epithelia, we analysed the levels of FOXA2 expression in the uterus of WT, CLCa, and CLCb KO mice. FOXA2 is a transcription factor that regulates epithelial differentiation and development, and its loss affects gland development, uterine function, and fertility (27, 28, 29). In the mature endometrium, expression of FOXA2 is confined to glandular epithelium and absent from epithelia bordering the uterine lumen, as seen the uterine tissue from WT mice (Fig 5A and B). Ectopic expression of FOXA2 was observed in epithelial cells at the uterine lumen of CLCa KO animals with pyometra and also, to a lesser extent, in the CLCa KO mice that did not suffer from pyometra (Fig 5A and B). In contrast, loss of CLCb did not change FOXA2 expression in the uterus, with expression seen only in the glandular epithelium (Fig 5A and B). FOXA2 has previously been shown to regulate the proliferation of endometrial epithelia (30, 31). To assess whether ectopic expression of FOXA2 in the lumenal epithelia of the uterus of CLCa KO mice altered cell proliferation, we analysed the expression of the postulated endometrial stem cell marker SRY-box transcription factor 9 (SOX9) (32) and the proliferation marker Ki67. However, the expression of both SOX9 and Ki67 in endometrial epithelial cells was unaffected by the loss of CLCa or CLCb compared with WT animals (Fig S2A–C), indicating that the ectopic expression of FOXA2 in the endometrial epithelia does not affect proliferation of these cells. The endometrial epithelium normally consists of simple columnar epithelial cells; however, overexpression of FOXA2 in the endometrium has previously been shown to induce epithelial stratification (33). We therefore analysed the expression of keratin 14 (K14), a marker of stratified squamous epithelial cells (34). As expected, the endometrial epithelia of WT, CLCb KO, and CLCa KO mice without pyometra did not express K14 (Fig 6A and B). In contrast, K14-expressing cells were found in endometrial epithelia in uteri from 40% (two of five analysed) CLCa KO mice with pyometra (Fig 6A and B). Consistently, the cells expressing K14 had a more squamous shape (Fig 6A), indicating a change in epithelial subtype in the uterus of some of the mice with pyometra.

FOXA2 overexpression in mouse uterine tissue has previously been shown to alter expression of FOXA2 target genes (33). To test whether FOXA2 transcriptional activity was elevated in CLCa KO mice, we analysed the expression of FOXA2 target genes including Ltf and Muc1 in the uterus of WT, CLCa KO, and CLCb KO mice and CLCa KO mice with pyometra. Only inconsistent expression of these target genes was observed (Fig S3A), but these were whole uterine tissue samples, and not specifically epithelia. To follow-up, we then transfected a construct encoding Flag–FOXA2 into Ishikawa tissue culture cells of endometrial epithelial origin (which do not express FOXA2) to see if transcription of previously identified FOXA2 target genes (33, 35), and keratin 14, was induced. Whereas the expression of several genes (LIF, LTF, MUC1) remained unchanged after ectopic FOXA2 expression, expression of several other genes (AREG, C3, IHH, and KRT14) was variable. Notably, transcription of the pancreatic trypsin inhibitor SPINK1, which is not endogenously expressed in Ishikawa cells, was significantly increased with Flag–FOXA2 expression (Fig S3B–E). Thus, although inconsistent, ectopic FOXA2 expression did alter transcription of at least a subset of target genes in uterine epithelia, as reported elsewhere (33, 35).

The structure and integrity of the endometrial epithelia at the lumen in the CLC KO mice was further assessed by immunofluorescent staining of epithelial adherens junction component E-cadherin and the apical tight junction protein ZO-1. WT, CLCa KO with and without pyometra, and CLCb KO mice all displayed normal distribution of both E-cadherin and ZO-1, with ZO-1 distributed at the apical surface of the endometrial epithelium and E-cadherin distributed at the basolateral membrane (Fig S4A and B). These findings suggest that epithelial junctions are formed normally in all CLC KO mice. Thus, at the resolution detectable by confocal microscopy, loss of CLCa or CLCb does not alter the organisation or polarity of the endometrial epithelia. The expression and distribution of mucin 1 (MUC1) was also examined because MUC1 is found at the apical surface of uterine epithelia and acts as a barrier against microbial invasion (36). Loss of this barrier could therefore lead to infection and pyometra. However, a layer of MUC1 was observed at the apical surface of endometrial epithelia of the uterine lumen from WT, CLCa KO, and CLCb KO animals (Fig S4C), suggesting that a loss of apical MUC1 is not the cause of pyometra in CLCa KO animals. Thus the pyometra phenotype may be linked to regional disruption of the epithelium by aberrant conversion to stratified epithelial cells, this differentiation defect being the only defect detected in our analysis.

Loss of CLCa or CLCb disrupts epithelial cyst polarity in vitro

Analysis of uterine tissue suggested that CLCs play a role in epithelial cell differentiation. However, despite the high level of expression of CLCa and CLCb in these cells, loss of either CLCa or CLCb did not result in observable organisational defects in the epithelial layer of the uterine lumen of CLCa KO or CLCb KO mice. The mice used in this study are a constitutive knock-out model, and it is possible that independent compensatory pathways or some substitute functions shared by CLCa and CLCb may have been activated and are sufficient to obscure individual functions of each CLC isoform in these cells. To address this possibility, we assessed the ability of cultured epithelial cell lines to form 3-dimensional cysts in vitro upon acute knockdown of either CLC or clathrin heavy chain (CHC17). Ishikawa cells (human endometrial adenocarcinoma cells) and Caco-2 cells (human colorectal adenocarcinoma cells) were treated with control siRNA or siRNA targeting CLCa, CLCb, or CHC17 for 72 h. Treated cells were then trypsinised and seeded as single cells embedded within the extracellular matrix substrate Matrigel and grown for 4–6 d to allow cysts to form before analysis by immunofluorescence. Clathrin component depletion was confirmed by immunoblotting lysate from Ishikawa and Caco-2 cells treated with siRNA for 72 h (Fig S5B and E), and immunoblotting showed sustained depletion of clathrin components during 2D culture of Ishikawa cells for another 6 d (the period of cyst differentiation) (Fig S5B). Persistence of clathrin component depletion for both Ishikawa cells and Caco-2 cells was confirmed by immunofluorescence in 6-d cysts, with some variability in Caco-2 cysts (Fig S5A, C, and D). Ishikawa and Caco-2 cells treated with control siRNA formed symmetrical, spherical cysts with a single lumen surrounded by an F-actin-rich apical membrane (Fig 7A and B). In both Ishikawa and Caco-2 cysts, CLCa and CLCb displayed distinct distribution patterns relative to each other. In the cysts, CLCa had a more ubiquitous distribution whereas CLCb localisation was enriched towards the apical surface of the cells (Fig 7A and B). The 3D-cysts formed by Ishikawa and Caco-2 cells treated with control siRNA displayed the expected basolateral distribution of E-cadherin and ZO-1 localised tightly around the actin-rich lumen (Fig 7C). A complete loss of this apico-basal polarity was seen in Ishikawa or Caco-2 cysts depleted of CLCa or CHC17 and also for Caco-2 cysts depleted of CLCb. In all of these cases, neither E-cadherin nor ZO-1 were polarised (Fig 7C–E), and there was no symmetrical distribution of cells around a lumen (Fig 7C–E). Analysis of E-cadherin, ZO-1, and actin distribution in Ishikawa cysts depleted of CLCb showed that some apico-basal polarity was attained in these cysts, with E-cadherin showing a degree of basolateral distribution, and greater symmetry observed across the cyst than for CLCa-depleted cysts (Fig 7C). When cyst size was quantified, Caco-2 cysts were found to be reduced in size when depleted for CLCa, CLCb, or CHC17, whereas this was not observed consistently for Ishikawa cysts treated in the same way (Fig 7F and G). However, DAPI staining, which typically reveals a uniform ring of cells in control-treated cysts, showed aberrant positioning of cells in the CLC-depleted Ishikawa cysts, and we frequently observed an abnormal number of lumens within the cysts (Fig 7C). During in vitro cyst growth, lumen formation is directly linked with apico-basal polarity development and organisation, as a single lumen develops next to the forming apical membrane initiation site and subsequent pre-apical patch (37, 38). Thus, loss of apico-basal polarity leads to abnormal lumen development. Therefore, to quantify the disruption of apico-basal polarity organisation in cysts depleted of CLCa, CLCb, or CHC17 observed in Fig 7C–E, a luminogenesis assay was performed. Cells were treated with siRNAs before growing cysts for 6 d. For the final 24 h, cysts were treated with cholera toxin to expand the lumen, then fixed, and the number of lumens were scored according to the phenotypes shown for Caco-2 cysts in Fig 7H. Phenotypes were detected using phalloidin staining of apical actin rings. The apico-basal polarity of these cysts was assessed by ZO-1 and E-cadherin staining, which confirmed that loss of a single lumen detected by phalloidin coincided with disruption of these polarity markers (Fig S6). ∼80% of control-treated cysts contained the expected single lumen for both Caco-2 cysts and Ishikawa cysts, which was significantly reduced in all cases by treatment with siRNA targeting CLCa, CLCb, or CHC17 (Fig 7I and J). Thus, although constitutive loss of either CLCa or CLCb did not appear to disturb the apico-basal polarity of the endometrial epithelia in the knock-out mice, acute knockdown of either CLCa or CLCb was sufficient to disturb the apico-basal polarity of 3D cysts from endometrial and colonic epithelial cells in vitro, suggesting the CLCs play an important role in epithelial cell differentiation. The more severe phenotypes seen for both cysts upon CLCa siRNA treatment compared with CLCb siRNA treatment is consistent with a dominant role for CLCa in epithelial development.

Animals

CLCa or CLCb KO mice (both C57BL/6 background) were established as previously described (9, 22). All mice used in this study (including WT, heterozygotes and KOs used for breeding pairs in fertility studies) were produced by breeding Clta^−/+ or Cltb^−/+ heterozygotes. For fertility studies, individual breeding pairs of the indicated genotype were housed continuously in a breeding cage, and the litter size and frequency were monitored for 54-d or 4 mo as indicated. All animal procedures and breeding were conducted according to the Animals Scientific Procedures Act UK (1986) and in accordance with the ethical standards at University College London.

Immunoblotting

Mouse tissue was harvested and snap-frozen in liquid nitrogen and stored at −80°C until further use. Tissue was homogenized in lysis buffer (50 mM Tris–HCl, pH 7.4, 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS, 1 mM EDTA) with complete EDTA-free proteinase inhibitor cocktail (Roche) and then incubated on ice for 1 h. After centrifugation at 10,000g, 4°C, 10 min, the supernatant was taken and the protein concentration was determined by BCA assay (Thermo Fisher Scientific). Proteins were separated by SDS–PAGE gel and transferred to Protran nitrocellulose membrane (0.2 μm; GE Healthcare). Immunoblotting was performed by incubating with primary antibodies diluted in TBS (20 mM Tris, pH 7.6, 150 mM NaCl) containing 2% BSA overnight at 4°C. Primary antibodies used were anti-CLC (CON.1, made in house) (26) and anti-tubulin (1:10,000; Abcam). Blots were incubated with IRDye800/700-conjugated secondary antibodies (1:5,000; LI-COR Biosciences), 1 h, room temperature. Proteins were detected with LI-COR Odyssey Imager (LI-COR Biosciences) and analysed using Image Studio Lite 5.2 software (LI-COR Biosciences). Immunoblotting of lysate from cultured cells was performed using the same protocol.

Histology and immunohistochemistry

Mouse uteri were removed and fixed in 4% PFA (pH 7.4; Sigma-Aldrich) at 4°C overnight. Specimens were dehydrated through graded ethanol solutions, paraffin-embedded, and cut into 5 μm sections. Sections were then deparaffinized and hydrated through graded ethanol solutions. For antigen retrieval, sections were incubated in sodium citrate buffer (10 mM sodium citrate, 0.05% Tween 20, pH 6.0), EDTA buffer (1 mM EDTA, 0.05% Tween 20, pH 8), or Tris-EDTA buffer (10 mM Tris base, 1 mM EDTA, 0.05% Tween 20, pH 9.0) for 20 min at 95°C. Slides were incubated with blocking buffer 1 (2% BSA, PBS) for 15 min and blocking buffer 2 (1% BSA, 0.3% Triton, PBS) for 15 min before incubation with primary antibodies diluted in blocking buffer 2 at 4°C overnight. Primary antibodies and antigen retrieval method used were anti-CLCa (1:200, citrate buffer; Sigma-Aldrich), anti-CLCb (1:200, citrate buffer; Sigma-Aldrich), anti-FOXA2 (1:100, EDTA buffer; Cell Signaling), anti-ZO-1 (1:100, citrate buffer; Thermo Fisher Scientific), E-cadherin (1:100, citrate buffer; BD), Ki-67 (1:200, EDTA buffer; Cell Signaling), anti-SOX9 (1:200, citrate buffer; Cell Signaling), anti-K14 (1:100, citrate buffer; Abcam), anti-mucin 1 (1:200, Tris–EDTA buffer; Abcam) anti-Ly-6G (1A8)-Alexa 647 (1:50, citrate buffer; BioLegend), β-catenin Alexa 488 (1:50, EDTA; BD). Slides were washed with PBST (0.1% Tween 20; PBS) and incubated with Alexa Fluor–conjugated secondary antibody (1:500; Thermo Fisher Scientific) and DAPI (1 μg/ml; Thermo Fisher Scientific) for 1 h. After PBST washes, slides were mounted with ProLong Diamond antifade media (Thermo Fisher Scientific). Images were obtained using a Leica TCS SP8 inverted laser–scanning confocal microscope with a PLAN APO 40X 1.3NA oil immersion objective, using four laser lines. Dyes were sequentially excited at 405 nm (DAPI), 488 nm (Alexa Fluor 488), 543 nm (Alexa Fluor 555), and 633 nm (Alexa Fluor 647). 1,024 × 1,024 pixel images where acquired using a photomultiplier detector, with the following emission wavelengths: DAPI: 410–470 nm; Alexa Fluor 488: 500–545 nm; Alexa Fluor 555: 560–610 nm; Alexa Fluor 647: 645–695 nm. Images were captured using Leica LAS X software and subsequently analysed using Image J.

Cell lines

Ishikawa cells (ECACC) were grown in minimum essential medium (Sigma-Aldrich) supplemented with 5% FBS (Thermo Fisher Scientific), 1% non-essential amino acids (Sigma-Aldrich), 2 mM L-glutamine (Thermo Fisher Scientific), and 100 U/ml penicillin/100 μg/ml streptomycin (P/S) (Thermo Fisher Scientific). Caco-2 cells were grown in DMEM (Thermo Fisher Scientific) supplemented with 10% FBS, 2 mM L-glutamine and P/S.

Cyst formation and siRNA treatment

For cyst formation, Ishikawa and Caco-2 cell lines were transfected with 20 nM siRNA (QIAGEN) using jetPRIME (Ishikawa) or INTERFERin (Caco-2) (both Polyplus Transfection) transfection regent following manufacturer’s instructions. Target sequences for siRNA used were CLTA: 5′-AA AGA CAG TTA TGC AGC TAT T-3′, CLTB: 5′-AAG GAA CCA GCG CCA GAG TGA-3′, CLTC: 5′-AAG CAA TGA GCT GTT TGA AGA. 72 h after transfection, cells were trypsinised into single cells. 6,000 cells were mixed with 30% Matrigel (Corning) in PBS and plated in a well of an eight-well EZ slide (Millipore) which had been pre-coated with 5 μl Matrigel. Cells were incubated in a 37°C, 5% CO₂ incubator for 30 min to allow Matrigel to set before then being supplemented with 400 μl culture medium. Culture medium was changed every 2 d. Cholera toxin (0.1 μg/ml; Sigma-Aldrich) was added 24 h prior fixation where indicated.

Immunofluorescence staining

Cysts were washed twice with PBS and fixed in 4% PFA at room temperature for 1 h before unreacted PFA was quenched with glycine. After PBS washes, cysts were permeabilized with 0.1% Triton X-100 in PBS for 30 min, followed by blocking with blocking buffer 1 for 1 h. Cysts were then incubated with primary antibodies diluted in blocking buffer overnight at 4°C. Primary antibodies used were anti-CLCa (1:250; Sigma-Aldrich), anti-CLCb (1:250, LCB.1, made in house (23)), ZO-1 (1:100; Thermo Fisher Scientific), E-cadherin (1:100; BD). After PBS washes, cysts were incubated with Alexa Fluor 488 or 647 conjugated–secondary antibody (1:500; Thermo Fisher Scientific) and Alexa 546–conjugated phalloidin (1:50; Thermo Fisher Scientific) diluted in blocking buffer for 1.5 h at room temperature. Cells were washed with PBS and stained using 1 μg/ml DAPI in PBS (Thermo Fisher Scientific) before slides were mounted with ProLong Diamond antifade media (Thermo Fisher Scientific). Images were obtained using a Leica TCS SP8 inverted laser-scanning confocal microscope with a HC PLAN APO 63X 1.40 NA CS2 oil immersion objective, using four laser lines. Dyes were sequentially excited at 405 nm (DAPI), 488 nm (Alexa Fluor 488), 543 nm (Alexa Fluor 546), and 633 nm (Alexa Fluor 647). 1,024 × 1,024 pixel images where acquired using a photomultiplier detector, with the following emission wavelengths: DAPI: 410–470 nm; Alexa Fluor 488: 500–545 nm; Alexa Fluor 546: 565–595 nm; Alexa Fluor 647: 645–695 nm. Images were captured using Leica LAS X software and subsequently analysed using Image J.

Transient transfection of Flag–FOXA2

For transient DNA transfection, Ishikawa cells were seeded at 70,000 cells/well of a six well dish. The following day, cells were transfected with 0.75 μg Flag–FOXA2 (number 153110; Addgene) using jetPRIME transfection reagent (Polyplus Transfection), following manufacturer’s instructions. Control samples were mock transfected using transfection reagent without DNA. 24 h post transfection, cells were collected for immunoblotting or mRNA extraction.

mRNA extraction, cDNA synthesis, and qPCR

mRNA was collected from cells in six well dishes using TRIzol reagent (Invitrogen) following manufacturer’s instructions. Briefly, cells were lysed and homogenised using 1 ml TRIzol reagent per well. Lysates were transferred to a microfuge tube and incubated for 5 min before adding 0.2 ml chloroform and mixing. After incubating for 3 min, samples were centrifuged at 12,000g for 15 min at 4°C. The aqueous phase was subsequently collected and RNA precipitated by incubation with 0.5 ml isopropanol, 10 min, 4°C. Samples were centrifuged at 12,000g for 10 min at 4°C, and then pellets washed in 1 ml 70% ethanol before a final centrifugation at 7,500g, 5 min, 4°C. The supernatant was discarded and pellet air dried for 5 min before resuspending in 30 μl RNase-free water. cDNA was synthesised from the isolated mRNA using the LunaScript RT SuperMix Kit (New England Biolabs) following manufacturer’s instructions. qPCR was performed using a Chromo4 PTC-200 Real-Time PCR Detector system (Bio-Rad) with SYBR-Green JumpStart Taq ready mix (Merck). PCR conditions were 94°C for 2 min, followed by 40 three-step cycles of 94°C for 15 s, 60°C for 30 s, and 72°C for 30 s. GAPDH was used as the endogenous gene controls. Melt curves were used to determine that a single PCR product was produced by each primer pair. Each qPCR was performed in triplicate with the following primers: AREG (forward: GTGGTGCTGTCGCTCTTGATA; reverse: CCCCAGAAAATGGTTCACGCT), CTNNB1 (forward: CACAAGCAGAGTGCTGAAGGTG, reverse: GATTCCTGAGAGTCCAAAGACAG), C3 (forward: GTGGAAATCCGAGCCGTTCTCT, reverse: GATGGTTACGGTCTGCTGGTGA), FOXA2 (forward: CCCCACAAAATGGACCTCAAG, reverse: GAGTACACCCCCTGGTAGTAG), GAPDH (), IHH (forward: AACTCGCTGGCTATCTCGGT, reverse: GCCCTCATAATGCAGGGACT), KRT14 (forward: TGAGCCGCATTCTGAACGAG, reverse: GATGACTGCGATCCAGAGGA), LIF (forward: CTGTTGGTTCTGCACTGGAA, reverse: CCCCTGGGCTGTGTAATAGA), LTF (forward: ATGCTGGAGATGTGGCTT, reverse: CCTTTCGGCTTTATTTGGT), MUC1 (forward: TGCCGCCGAAAGAACTACG, reverse: TGGGGTACTCGCTCATAGGAT), RNF43 (forward: AAAATCCAGCCTCTCTGCCC, reverse: ACAACCACACTGGCTGTGAA), SPINK1 (forward: TGACTCCCTGGGAAGAGAGG, reverse: AGTCCCACAGACAGGGTCAT), Ltf (forward: ATCTCTGTGCCCTGT, reverse), Muc1 (forward: TTCCAACCCAGGACACCTAC, reverse: ATTACCTGCCGAAACCTCCT).

Statistical analysis

Statistical analysis was performed using Prism 7 (GraphPad). Methods of comparison are stated in figure legends. Mean ± SE is shown.