Wnt Effector TCF4 Is Dispensable for Wnt Signaling in Human Cancer Cells

T-cell factor 4 (TCF4), together with β-catenin coactivator, functions as the major transcriptional mediator of the canonical wingless/integrated (Wnt) signaling pathway in the intestinal epithelium. The pathway activity is essential for both intestinal homeostasis and tumorigenesis. To date, several mouse models and cellular systems have been used to analyze TCF4 function. However, some findings were conflicting, especially those that were related to the defects observed in the mouse gastrointestinal tract after Tcf4 gene deletion, or to a potential tumor suppressive role of the gene in intestinal cancer cells or tumors. Here, we present the results obtained using a newly generated conditional Tcf4 allele that allows inactivation of all potential Tcf4 isoforms in the mouse tissue or small intestinal and colon organoids. We also employed the clustered regularly interspaced short palindromic repeats (CRISPR)/Cas9 system to disrupt the TCF4 gene in human cells. We showed that in adult mice, epithelial expression of Tcf4 is indispensable for cell proliferation and tumor initiation. However, in human cells, the TCF4 role is redundant with the related T-cell factor 1 (TCF1) and lymphoid enhancer-binding factor 1 (LEF1) transcription factors.


Introduction
The wingless/integrated (Wnt) signaling pathway represents one of the fundamental evolutionarily conserved signaling mechanisms controlling cell specification during embryonic development and in adult tissues. Aberrant activation of Wnt signaling causes a number of diseases, including various types of cancer [1]. Overall, there are at least five distinct branches of Wnt signaling. The best studied is the so-called canonical Wnt pathway, with β-catenin as its key effector [2]. Besides the structural function in adherens junctions, β-catenin accumulates in the cytoplasm of the Wnt ligand-stimulated cells, and it mediates Wnt signal transduction in the nucleus. Nuclear β-catenin associates with DNA-binding proteins of the lymphoid enhancer-binding factor/T-cell factor (LEF/TCF) family (further referred to as TCFs). β-Catenin converts TCFs from transcriptional repressors to activators, and TCF/β-catenin complexes upregulate the expression of Wnt target genes such as c-myc, cyclin D1, CD44, axis inhibition protein 2 (Axin2), and Sp5 transcription factor (SP5). In the absence of the Wnt stimulus, cytosolic β-catenin is marked for 12% of microsatellite-stable (MSS) cancers. Moreover, the TCF4 locus was deleted in a subset of the examined cases [16]. These loss-of-function mutations imply that apart from its physiological role in healthy intestines (see further), the TCF4 status is important for the initiation and/or progression of CRC. Additionally, a genome-wide RNA-mediated interference (RNAi) screen identified TCF4 as a transcriptional repressor, decreasing the Wnt pathway output and restricting CRC cell growth [17].
It is presumed that in the mouse intestine, Tcf4 is crucial for embryonic development and adult tissue homeostasis of the small intestinal and colonic epithelia [18]. However, some results of the Tcf4 targeting experiments are contradictory. The Tcf4 whole-body knockout generated by the insertion of the expression cassette producing hygromycin B phosphotransferase immediately upstream of the high mobility group (HMG) DNA binding domain sequence (so-called HMG box; we designated the modified allele Tcf4 Hyg ) was perinatal lethal due to the absence of the proliferative compartments in the small intestine. However, the colon epithelium in these Tcf4-deficient mice stayed seemingly intact [19]. Conditional Tcf4 inactivation performed by Cre-mediated excision of the floxed sequence encoding the HMG box (we named the modified allele Tcf4 floxHMG ) impaired cell proliferation in both the adult small intestine and the colon [20]. In contrast, Tcf4 gene-driven Cre recombinase-mediated deletion of the first Tcf4 exon (modified allele: Tcf4 flox1 ) induced hyperproliferation of progenitor cells in the small intestinal and colonic epithelium at embryonic day (E) 13.5. This resulted in cell exhaustion, accompanied by the disruption of the small intestinal and colonic architecture at E14.5. Moreover, Tcf4 haploinsufficiency promoted the formation of colonic tumors in multiple intestinal neoplasia (Min) mice, which represent the mouse model of intestinal tumorigenesis initiated by Apc loss [21]. In addition, using the same, i.e., Tcf4 flox1 , allele Angus-Hill and colleagues observed that complete knockout of Tcf4 in the adult colon resulted in the formation of aberrant crypt foci (ACF), which are considered to be the earliest neoplastic lesions during CRC initiation [22]. The latter findings indicated-similarly to the situation observed in human CRC-the tumor-suppressive function of Tcf4.
To address these contradictory results, we employed a newly generated mouse strain harboring exon 5 flanked by loxP sites (allele name: Tcf4 flox5 ). The exon is included in all annotated transcripts in the human and mouse Tcf4 gene, and in addition, it is positioned downstream from the most frequently used transcription start sites that initiate mRNAs encoding either a full-length or truncated version of the protein lacking the β-catenin binding domain [18]. As Cre-mediated excision of exon 5 leads to an open reading frameshift, production of the majority of the Tcf4 protein variants is abolished. The Tcf4 flox5 allele was used to inactivate Tcf4 in embryonic and adult mouse intestinal epithelia using several different Cre drivers. Tcf4 deletion was also performed in organoids derived from the small intestine and colon. In addition, we employed the clustered regularly interspaced short palindromic repeats (CRISPR)/Cas9 system to disrupt the TCF4 gene and its closest paralog TCF3 (alias TCF7L1) in human cells, and tested the effect(s) of the disruption on cell viability and Wnt pathway-driven transcription. Our results indicate the importance of mouse Tcf4, mainly in adult intestinal epithelium homeostasis and intestinal tumor initiation. In contrast, in human cells, the TCF4 function is substituted by other LEF/TCF family members. To generate Tcf4 flox5/flox5 mice, the lacZ-neomycin phosphotransferase expression cassette flanked by FTR sites was removed from the genome by mating Tcf4 lacZ/+ with ACTB-FLPe (purchased from the Jackson Laboratory, Bar Harbor, ME, USA) transgenic mice. Mice of the ACTB-FLPe strain express enhanced FLP recombinase (FLPe) from the ubiquitously active regulatory region of the human β-ACTIN gene [23]. Apc flox14/flox14 mice were purchased from the Mouse Repository (National Cancer Institute, Frederick, MD, USA); Villin-Cre and Villin-CreERT2 animals [24] were kindly provided by S. Robine (Institut Curie, Centre de Recherche, Paris, France). Lgr5-EGFP-IRES-CreERT2 [B6.129P2-Lgr5 tm1(cre/ERT2)Cle/J ], Rosa26R-lacZ [B6;129S4-Gt(ROSA)26 Sortm1Sor/J ], and Rosa26R-tdTomato [B6;129S6-Gt(ROSA)26 Sortm14(CAG−tdTomato)Hze/J ] mice were purchased from the Jackson Laboratory. Animals were housed in specific pathogen-free (SPF) conditions, and genotyped according to the provider's or published protocols. For Cre-mediated recombination, mice were administered using an intraoral gavage of tamoxifen (Sigma-Aldrich, St. Louis, MO, USA; stock 10 mg/mL in ethanol). Prior to gavage, the tamoxifen solution was mixed with mineral oil (Sigma-Aldrich). For tumor initiation experiments, mice were administered with a single dose containing 1 mg of tamoxifen; 5 mg of tamoxifen per dose was used in all other experiments.

RNA Isolation and Quantitative Reverse Transcription-Polymerase Chain Reaction (qRT-PCR)
Total RNA from cells grown in cultures was isolated using TRI Reagent (Sigma-Aldrich) and reverse-transcribed using RevertAid Reverse Transcriptase (Thermo Fisher Scientific) according to the manufacturer's protocol. For sorted cells and organoids, RNeasy Micro Kit (Qiagen, Hilden, Germany) and MAXIMA Reverse Transcriptase (Thermo Fisher Scientific) were used. qRT-PCR was performed in triplicates using the SYBR Green I Master Mix and LightCycler 480 apparatus (Roche, Basel, Switzerland). Primers are listed in Table S1.

Fluorescent Microscopy
Fluorescent staining of small intestinal organoids was performed as follows. Matrigel-embedded organoids were washed with PBS and fixed in 4% (w/v) paraformaldehyde (Electron Microscopy Sciences, Hatfield, PA, USA) in PBS for 30 min at room temperature (RT). After fixation, organoids were washed in PBS and incubated with 0.1% Triton X100 (Sigma-Aldrich) in PBS for 30 min. After additional washing in PBS, organoids were incubated in 5% goat serum (Vector Laboratories, Burlingame, CA, USA) in PBS and washed again. Incubation with primary anti-Tcf4 antibody (#2569, Cell Signalling) was performed overnight at 4 • C. After washing (PBS), the organoids were incubated (2 h at RT) with a goat anti-rabbit immunoglobulin Alexa Fluor 488-conjugated secondary antibody (Thermo Fisher Scientific); cell nuclei were counterstained with diamidino-2-phenylindole (DAPI; 0.1 µg/mL, Sigma-Aldrich) for 10 min at RT. Stained organoids were kept (and photographed) in Scaleview-A2 optical clearing agent (Olympus) at 4 • C for 1 week. Fluorescent pictures were acquired using spinning disk confocal microscope Dragonfly (Andor) and processed using Imaris software (Bitplane, Belfast, UK).

Statistical Analysis
The results of the quantitative reverse transcription polymerase chain reaction (qRT-PCR) and β-Galactosidase (lacZ) staining were evaluated by a Student's t-test.

Analysis of Tcf4 Expression in the Mouse Small Intestine and Colon
To survey Tcf4 expression in the adult small intestine and colon, we used immunohistochemistry (IHC) detection of the Tcf4 protein. Staining of paraffin-embedded sections with a Tcf4-specific antibody revealed an expression pattern that was similar to results published previously [35]. In the small intestine, the Tcf4 positive result was noticed throughout the epithelium, with slightly less pronounced staining in the upper part of the crypts, where rapidly dividing TA cells are localized. The staining pattern was reproduced irrespective of the position along the rostro-caudal axis of the organ. In contrast, in the colon, the strongest Tcf4 nuclear positivity was noted in differentiated cells located on the colon surface ( Figure 1A). To assign Tcf4 expression to individual cell types present in the small intestinal crypts, we isolated epithelial crypt cells from Lgr5-EGFP-IRES-CreERT2 (further referred to as Lgr5-CreERT2) mice producing enhanced green fluorescent protein (EGFP) and CreERT2 fusion proteins in ISCs and secretory cell precursors [6,36]. The green fluorescent signal and anticluster of differentiation 24 (CD24) surface labeling (the labeling marks epithelial cells located at the lower portion of the crypts [37]) was used to discriminate the differentiation status of epithelial cells. Total RNA obtained from the sorted cell populations was employed for qRT-PCR analysis of all TCFs. The expression levels of intestinal alkaline phosphatase (Alpi) and cryptdins were used as markers of differentiated enterocytes or Paneth cells, respectively, and Olfm4 was used as an additional ISC marker. We observed a slight discrepancy in the level of EGFP and Lgr5 expression in Paneth cells and secretory precursors, probably caused by the mosaic production of EGFP from the knock-in allele [38]. Nevertheless, our analysis showed that the highest levels of Tcf4 mRNA were detected in the cells present at the crypt bottom, especially in the ISCs and Paneth cells. We anticipate that the qRT-PCR analysis precluded direct comparison of the mRNA levels among different genes; nevertheless, in agreement with published data, (relatively) high expression of Tcf1 in ISCs was observed ( Figure 1B). We also employed the Tcf4 lacZ/+ reporter strain generated by the knock-in of the lacZ expression cassette into the Tcf4 locus, to follow Tcf4 intestinal expression. As expected, the lacZ-positivity phenocopied Tcf4 immunohistochemical detection. The only noticeable difference was the absence of blue staining in the crypts of the duodenum and jejunum, and in the villus cells of the ileum. Consequently, lacZ production was undetectable in whole-mount specimens of the ileum ( Figure 1C, D). Since the lacZ enzyme is produced from one allele only, we suggest that the discrepancy between the two staining methods was caused by lower gene dosage, resulting in the decreased sensitivity of the lacZ enzyme detection. Interestingly, PAS-mediated visualization of Paneth and Goblet cells, combined with anti-Tcf4 staining, revealed that in the ileal crypts, the highest, i.e., detectable, levels of Tcf4 proteins were produced in Paneth cells ( Figure 1E; figure legend is on the next page).

Intestinal Epithelium-Specific Tcf4 Inactivation Had No Effect on the Embryonic Gut
Next, we intercrossed Tcf4 flox5/flox5 mice with Villin-Cre transgenic mice; the latter mice produce constitutively active Cre driven by the murine villin promoter. The transgene is active in epithelial cells of the small intestine and colon [24]. Interestingly, no obvious phenotype was observed in the developing gut. Despite Tcf4 absence, the intervillus regions of the small intestinal epithelium at embryonic day (E) 17.5 contained proliferating cell nuclear antigen (PCNA)-positive cells. (Figure 2A). The observed phenotype was in sharp contrast to previously documented phenotypes of the mouse models of Tcf4 gene inactivation using Tcf4 Hyg/Hyg animals [19], or by conditional deletion of the Tcf4 floxHMG alleles using PGK-Cre transgenic animals (genotype: PGK-Cre Tcf4 floxHMG/floxHMG ) [20]. The obvious explanation for the observed discrepancy would be incomplete inactivation of "our" Tcf4 flox5 allele. It was documented that the Villin-Cre transgene is active in the entire intestinal epithelium at E12.5, i.e., several days prior to transition from pseudostratified to columnar epithelium [24]. Contrary to the original description, we observed groups of Tcf4-positive cells in the colon epithelium of Tcf4 flox5/flox5 Villin-Cre embryos at E17.5, indicating less efficient Cre-mediated DNA recombination ( Figure 2B). Nevertheless, the small intestinal epithelium appeared to be recombined "completely", excluding the possibility that the absence of the small intestinal phenotype was caused by partial recombination. The PGK promoter drives ubiquitous expression of the Cre enzyme [39]. Consequently, the situation in PGK-Cre Tcf4 floxHMG/floxHMG animals mimics Tcf4 whole-body inactivation achieved in Tcf4 Hyg/Hyg mice. We hypothesized that the Tcf4 function in the cells outside of the epithelium might contribute to the small intestinal defect manifested in Tcf4 Hyg/Hyg and PGK-Cre Tcf4 floxHMG/floxHMG animals. Interestingly, we observed Tcf4 nuclear positivity in subepithelial layers of developing colon tissue at E17.5 ( Figure 2B). In the small intestine, the non-epithelial Tcf4 expression was mainly detected in the putative enteric plexus cells (Figure 2A). To analyze Tcf4-deficient mice, we intercrossed Tcf4 lacZ/+ mice; however, no viable offspring of the Tcf4 lacZ/lacZ genotype were obtained. Therefore, we performed time pregnancies followed by an analysis of embryos at different developmental stages. The analysis revealed the absence of proliferating cells in the intervillus regions of the small intestine, starting at E16.5 ( Figure S1). The phenotype-similar to that seen in Tcf4 Hyg/Hyg or PGK-Cre Tcf4 floxHMG/floxHMG mice-indicated the possible contribution of Tcf4-expressing non-epithelial cells to the formation of intestinal tissue. Nevertheless, the experimental proof (e.g., usage of a Cre driver active in intestinal mesenchymal cells) confirming the hypothesis will require additional experiments.
We never detected any hyperproliferation of progenitor cells in the small intestinal and colonic epithelium as described by Angus-Hill and colleagues [22]. It has been suggested that an N-terminal Tcf4 protein fragment that excludes the DNA binding domain (the HMG box) is expressed from the Tcf4 Hyg (and by analogy from Tcf4 floxHMG ) allele and that the protein interferes with TCF/β-catenin-mediated transcription, thus enforcing the phenotype of the Tcf4 knockout mice [40]. Removal of Tcf4 exon 5 results in a frameshift, leading to the production of a (relatively short) polypeptide that is 189 amino acids long. Thus, we tested the possible repressive activity of the polypeptide in a luciferase-based reporter assay. Nevertheless, (over)expression of the N-terminal Tcf4 fragment had no effect on the TCF/β-catenin-dependent transcription ( Figure S2). Recently, Vacik and co-workers [41] discovered that during embryonic development, a shortened version of Tcf4 protein missing the β-catenin binding domain is produced. The variant contains the entire C-terminal portion of the protein, including the HMG box. Consequently, it acts as a dominant negative (dn) Tcf4 isoform blocking Wnt-dependent transcription. Additionally, the production of mRNA encoding the dnTcf4 variant is driven by an intronic promoter located upstream of exon 5 [41]. Although the specific Tcf4 isoform was predominantly detected in the developing nervous system, the analogous TCF4 isoform was identified using "in silico" analysis in various human tissues, including the intestine [18]. We suggest that Cre-mediated recombination of the first exon in Tcf4 flox1/flox1 mice retains (in contrast to other modified Tcf4 alleles) the expression of dnTcf4, generating the imbalance between the full-length and the dnTcf4 protein. What cause the difference between Tcf4-deficient cells and cells retaining expression of the dnTcf4 form? Obviously, dnTcf4 represses the transcription of Wnt-signaling target genes [34,42,43]. Interestingly, using a chromatin immunoprecipitation (ChIP)-sequencing (ChIP-seq) experiment, Schuijers and colleagues showed that in comparison to full-length TCF4, dnTCF4 preferentially targets DNA elements bound by other LEF/TCF family members [43]. Consequently, the expression profile of Tcf4 null cells and cells expressing the dnTcf4 isoform would differ. Nevertheless, why the dnTcf4-mediated repression contributes to hyperproliferation of the embryonic intestinal epithelium remains unclear. We never detected any hyperproliferation of progenitor cells in the small intestinal and colonic epithelium as described by Angus-Hill and colleagues [22]. It has been suggested that an N-terminal Tcf4 protein fragment that excludes the DNA binding domain (the HMG box) is expressed from the Tcf4 Hyg (and by analogy from Tcf4 floxHMG ) allele and that the protein interferes with TCF/β-cateninmediated transcription, thus enforcing the phenotype of the Tcf4 knockout mice [40]. Removal of Tcf4

Absence of Tcf4 Compromised Cell Proliferation of the Adult Small Intestinal and Colon Epithelia
In young (2-week-old) mice, a slight (nevertheless significant) decrease in the number of crypts and villi was detected in the small intestine of Tcf4 flox5/flox5 Villin-Cre mice (compared to animals with intact Tcf4). In contrast to adult tissue, Tcf4 was mainly present in the crypt epithelium, and not in the cells lining the villi. Tcf4-specific staining was seen not only in wild-type (wt) mice, but also in the crypts of 2-week-old Tcf4 flox5/flox5 Villin-Cre animals, indicating incomplete recombination of the floxed sequences. With respect to the crypt number, the small intestinal epithelium recovered in 10and 25-week-old mice (Figures S3 and S4). In addition, in the adult animals, the Tcf4-positive cells were located mainly on the villi, and Tcf4 staining in the crypts was less prominent. Interestingly, streams of cells producing Tcf4 were also visible on the villi of Tcf4 flox5/flox5 Villin-Cre mice, confirming incomplete recombination of the Tcf4 floxed alleles. The latter observation also indicated a strong selection pressure to maintain the Tcf4 expression in epithelial cells.
In the colon of 2-and 10-week-old mice, the Tcf4 absence resulted in disorganized epithelia containing patches of normal tissue that retained strong nuclear Tcf4 staining in the cells covering the colon surface. Similarly to the small intestine, partial recovery of the epithelium was observed in 25-week-old mice. Nevertheless, in contrast to the small intestine, the majority of colon tissue remained without Tcf4 expression ( Figure S5). The reduced dependency of adult colon tissue on Tcf4 expression in adult mice prompted us to test whether Tcf4 is redundant with other TCFs. However, qRT-PCR or IHC analysis did not show increased expression of any additional LEF/TCF family members in the Tcf4 flox5/flox5 Villin-Cre colon (data not shown). Interestingly, no signs of necrotic death of epithelial cells in the colon-reported after deletion of the Tcf4 flox1 allele [22]-were observed. The effect of Tcf4 deletion in the adult intestinal epithelium was also tested in Tcf4 flox5/flox5 Villin-CreERT2 mice. The modified tamoxifen-sensitive ligand binding domain of the estrogen receptor (ERT2) fused to the Cre enzyme allows for excision of the floxed sequences in a timely manner [44]. Experimental animals were sacrificed 1, 4, 7, and 11 days after tamoxifen administration, i.e., Tcf4 gene inactivation. Immunohistochemical staining revealed the absence of Tcf4 protein in the small intestinal and colon samples as early as one day after tamoxifen administration. Tcf4 absence was accompanied by a loss of proliferating cells in the crypts of both tissues. At day 7, several vigorously proliferating hyperplastic crypts expressing Tcf4 were observed in the small intestine. These crypts at day 11 expanded, forming disorganized epithelium with distorted morphology (Figure 3). The colon architecture was-besides the absence of PCNA-positive cells-seemingly not affected. However, fluorescence-activated cell sorting (FACS) analysis showed reduced proportion of cells expressing CD24, indicating that the tissue underwent cellular (and functional) changes ( Figure S6). In the colon, proliferating cells in the crypts re-appeared at day 7 and, simultaneously, nuclear Tcf4-specific staining was observed in some crypt cells. At day 11, Tcf4 production was clearly visible in differentiated cells on the tissue surface ( Figure 3). Tcf4-deficient animals had to be sacrificed prior to day 12 because of malnutrition precluding any analysis at later time points. Nevertheless, it was evident that in contrast to Tcf4 flox5/flox5 Villin-Cre mice, cells in which the Tcf4 gene escaped "acute" inactivation were the source of epithelial recovery in the Tcf4 flox5/flox5 Villin-CreERT2 colon.
Genes 2018, 9, x FOR PEER REVIEW 12 of 22 differentiated cells on the tissue surface ( Figure 3). Tcf4-deficient animals had to be sacrificed prior to day 12 because of malnutrition precluding any analysis at later time points. Nevertheless, it was evident that in contrast to Tcf4 flox5/flox5 Villin-Cre mice, cells in which the Tcf4 gene escaped "acute" inactivation were the source of epithelial recovery in the Tcf4 flox5/flox5 Villin-CreERT2 colon.

Tcf4-Deficient Intestinal Stem Cells s Do Not Contribute to Intestinal Homeostasis
Subsequently, we employed Rosa26R-lacZ reporter mice to trace the fate of ISCs upon Tcf4 ablation. Rosa26R-lacZ animals harbor the lacZ gene integrated downstream of the Rosa26 promoter. However, although the Rosa26 locus is ubiquitously active, lacZ mRNA is produced only after Cremediated removal of the floxed transcriptional blocker placed upstream of the lacZ gene [45]. Recombination of the floxed sequences was induced in adult Tcf4 flox5/flox5 Rosa26R-lacZ Lgr5-CreERT2

Tcf4-Deficient Intestinal Stem Cells s Do Not Contribute to Intestinal Homeostasis
Subsequently, we employed Rosa26R-lacZ reporter mice to trace the fate of ISCs upon Tcf4 ablation. Rosa26R-lacZ animals harbor the lacZ gene integrated downstream of the Rosa26 promoter. However, although the Rosa26 locus is ubiquitously active, lacZ mRNA is produced only after Cre-mediated removal of the floxed transcriptional blocker placed upstream of the lacZ gene [45]. Recombination of the floxed sequences was induced in adult Tcf4 flox5/flox5 Rosa26R-lacZ Lgr5-CreERT2 and control Tcf4 +/+ Rosa26R-lacZ Lgr5-CreERT2 animals by a single dose of tamoxifen. LacZ staining was followed at several points after tamoxifen administration. At day 1, lacZ-positive cells were observed at the bottom of the small intestinal and colonic crypts in both mouse strains. At day 5, streams of blue cells expanded from the crypts in animals of both genotypes. However, in Tcf4-deificient epithelium, the streams frequently separated from the crypt bottoms. Twelve days after recombination, continuous "ribbons" of labelled cells emanating from the crypts and reaching the top of the villi or colonic surface were observed in control mice. However, in Tcf4-deficient animals, the majority of blue cells disappeared from the intestine ( Figure 4A). Subsequent quantification confirmed substantially reduced numbers of persistently labelled crypts in Tcf4-deficient animals when compared to mice with the wt Tcf4 alleles ( Figure 4B). and control Tcf4 +/+ Rosa26R-lacZ Lgr5-CreERT2 animals by a single dose of tamoxifen. LacZ staining was followed at several points after tamoxifen administration. At day 1, lacZ-positive cells were observed at the bottom of the small intestinal and colonic crypts in both mouse strains. At day 5, streams of blue cells expanded from the crypts in animals of both genotypes. However, in Tcf4deificient epithelium, the streams frequently separated from the crypt bottoms. Twelve days after recombination, continuous "ribbons" of labelled cells emanating from the crypts and reaching the top of the villi or colonic surface were observed in control mice. However, in Tcf4-deficient animals, the majority of blue cells disappeared from the intestine ( Figure 4A). Subsequent quantification confirmed substantially reduced numbers of persistently labelled crypts in Tcf4-deficient animals when compared to mice with the wt Tcf4 alleles ( Figure 4B).

Tcf4 Loss Affects the Size of Apc-Deficient Small Intestinal Tumors
To elucidate the role of Tcf4 in intestinal tumorigenesis, we used mice with the conditional Apc allele containing exon 14 flanked by loxP sites (Apc flox14/flox14 ). The Cre-mediated excision of the floxed exon changes the reading frame downstream of the deletion and leads to the production of a truncated, non-functional Apc polypeptide [46]. Animals harboring the floxed or wt Tcf4 alleles, i.e., Apc flox14/flox14 Tcf4 flox5/flox5 Lgr5-CreERT2 or Apc flox14/flox14 Tcf4 +/+ Lgr5-CreERT2 mice, respectively, were treated with a reduced dose of tamoxifen (1 mg per animal) to increase their survival. In the small intestine, concomitant inactivation of Tcf4 and Apc resulted in a significant decrease of the size of

Tcf4 Loss Affects the Size of Apc-Deficient Small Intestinal Tumors
To elucidate the role of Tcf4 in intestinal tumorigenesis, we used mice with the conditional Apc allele containing exon 14 flanked by loxP sites (Apc flox14/flox14 ). The Cre-mediated excision of the floxed exon changes the reading frame downstream of the deletion and leads to the production of a truncated, non-functional Apc polypeptide [46]. Animals harboring the floxed or wt Tcf4 alleles, i.e., Apc flox14/flox14 Tcf4 flox5/flox5 Lgr5-CreERT2 or Apc flox14/flox14 Tcf4 +/+ Lgr5-CreERT2 mice, respectively, were treated with a reduced dose of tamoxifen (1 mg per animal) to increase their survival. In the small intestine, concomitant inactivation of Tcf4 and Apc resulted in a significant decrease of the size of neoplastic lesions. Interestingly, IHC analysis showed that in some PCNA-positive, i.e., proliferating tumor parts, Tcf4-specific staining was absent, indicating that the proliferation of transformed cells is Tcf4-independent ( Figure 5A). Due to the less efficient Lgr5-CreERT2-mediated removal of the floxed sequences [7,47], less abundant neoplastic lesions were formed in the colon. Nevertheless, Tcf4 inactivation had no effect on the size and amounts of colonic (micro)adenomas ( Figure 5B). Moreover, in contrast to the small intestine, all colonic lesions retained Tcf4-specific residual staining, suggesting incomplete Tcf4 inactivation. Moreover, we detected increased expression of Tcf3 in neoplastic colon tissue, suggesting that the decreased dosage of Tcf4 was compensated for by increased Tcf3 production ( Figure 5C). Since Tcf3 expression was never linked to (hyper)active Wnt signaling, a mechanistic explanation for its elevated production in Apc-deficient neoplastic lesions is unclear. Contrary to our results, Angus-Hill and co-workers found that increased expression of additional LEF/TCF family members Lef1 and Tcf1 in colon tumors of Min mice harboring only one intact Tcf4 allele (in comparison to tumors with both alleles functional). In the same study, no significant change in Tcf3 expression was observed [22]. We suggest that in larger, more progressed tumors, the reduced dosage of Tcf4 activity might be compensated for by Tcf1 and/or Lef1, whereas in early lesions, the Tcf4 role is substituted by Tcf3. neoplastic lesions. Interestingly, IHC analysis showed that in some PCNA-positive, i.e., proliferating tumor parts, Tcf4-specific staining was absent, indicating that the proliferation of transformed cells is Tcf4-independent ( Figure 5A). Due to the less efficient Lgr5-CreERT2-mediated removal of the floxed sequences [7,47], less abundant neoplastic lesions were formed in the colon. Nevertheless, Tcf4 inactivation had no effect on the size and amounts of colonic (micro)adenomas ( Figure 5B). Moreover, in contrast to the small intestine, all colonic lesions retained Tcf4-specific residual staining, suggesting incomplete Tcf4 inactivation. Moreover, we detected increased expression of Tcf3 in neoplastic colon tissue, suggesting that the decreased dosage of Tcf4 was compensated for by increased Tcf3 production ( Figure 5C). Since Tcf3 expression was never linked to (hyper)active Wnt signaling, a mechanistic explanation for its elevated production in Apc-deficient neoplastic lesions is unclear. Contrary to our results, Angus-Hill and co-workers found that increased expression of additional LEF/TCF family members Lef1 and Tcf1 in colon tumors of Min mice harboring only one intact Tcf4 allele (in comparison to tumors with both alleles functional). In the same study, no significant change in Tcf3 expression was observed [22]. We suggest that in larger, more progressed tumors, the reduced dosage of Tcf4 activity might be compensated for by Tcf1 and/or Lef1, whereas in early lesions, the Tcf4 role is substituted by Tcf3.

Tcf4-Deficient Organoids Displayed Impaired Growth
To address whether Tcf4 absence affects the growth of intestinal epithelial cells in vitro, organoid cultures were established from the crypts explanted from the small intestine and colon of Tcf4 flox5/flox5 Villin-CreERT2 mice; control organoids were derived from Tcf4 flox5/flox5 animals. Recombination of the floxed sequences was induced by 4-OHT, and the organoid growth and morphology was monitored. In accordance with data published previously [20], Tcf4-deficient organoids grew comparably to Tcf4 flox5/flox5 organoids until day 4 after 4-OHT addition to the culture medium. Since day 4, Tcf4-deficient organoids showed intermitted budding, and started to release dead cells from the lumen. Additionally, the organoids were unable to restart their growth after a single passage ( Figure 6A and Figure S7A). qRT-PCR analysis of total RNA isolated from organoids three days after 4-OHT treatment showed a clear decrease in the level of Tcf4 mRNA in CreERT2-expressing organoids. This was accompanied by robust downregulation of proliferating cell markers (Ki67, cyclin D1) and Wnt signaling responsive genes (Axin2, Lef1, Tcf1) [48][49][50]. The expression levels of Axin2 paralog Axin1, i.e., the gene that is not regulated by the Wnt pathway, did not differ between Tcf4-deficient and control samples. Interestingly, Tcf3 expression was two-fold upregulated in colon organoids after Tcf4 inactivation. Moreover, in agreement with the FACS analysis, Tcf4 ablation decreased production of CD24 in colon organoids ( Figure 6B and Figure S7B).

Tcf4-Deficient Organoids Displayed Impaired Growth
To address whether Tcf4 absence affects the growth of intestinal epithelial cells in vitro, organoid cultures were established from the crypts explanted from the small intestine and colon of Tcf4 flox5/flox5 Villin-CreERT2 mice; control organoids were derived from Tcf4 flox5/flox5 animals. Recombination of the floxed sequences was induced by 4-OHT, and the organoid growth and morphology was monitored. In accordance with data published previously [20], Tcf4-deficient organoids grew comparably to Tcf4 flox5/flox5 organoids until day 4 after 4-OHT addition to the culture medium. Since day 4, Tcf4deficient organoids showed intermitted budding, and started to release dead cells from the lumen. Additionally, the organoids were unable to restart their growth after a single passage (Figures 6A  and S7A). qRT-PCR analysis of total RNA isolated from organoids three days after 4-OHT treatment showed a clear decrease in the level of Tcf4 mRNA in CreERT2-expressing organoids. This was accompanied by robust downregulation of proliferating cell markers (Ki67, cyclin D1) and Wnt signaling responsive genes (Axin2, Lef1, Tcf1) [48][49][50]. The expression levels of Axin2 paralog Axin1, i.e., the gene that is not regulated by the Wnt pathway, did not differ between Tcf4-deficient and control samples. Interestingly, Tcf3 expression was two-fold upregulated in colon organoids after Tcf4 inactivation. Moreover, in agreement with the FACS analysis, Tcf4 ablation decreased production of CD24 in colon organoids ( Figures 6B and S7B).  Next, we derived small intestinal or colon organoids from Apc flox14/flox14 Tcf4 flox5/flox5 Lgr5-CreERT2, Apc flox14/flox14 Tcf4 +/+ Lgr5-CreERT2, and Apc +/+ Tcf4 +/+ Lgr5-CreERT2 mice, and induced single Apc or double Apc/Tcf4 inactivation by 4-OHT. Apc inactivation in organoids harboring the unmodified Tcf4 gene led to the formation of fast growing "spheroids" displaying a cyst-like morphology lacking the budding crypt domains (the morphological change was mainly evident after organoid passage). However, simultaneous deletion of Apc and Tcf4 caused organoid demise after organoid culture splitting ( Figure 7A and data not shown). Subsequent qRT-PCR analysis revealed that the Apc inactivation caused robust upregulation of Lef1 mRNA and a moderate increase in the expression levels of other Wnt signaling target genes Axin2, Tcf1, and cyclin D1 in both colonic and small intestinal organoids. Additionally, Apc-deficient small intestinal organoids displayed increased levels of Tcf3 mRNA. However, concomitant deletion of Apc and Tcf4 led to a substantially diminished expression of all tested Wnt target genes, including Axin2, Lef1, Lgr5, and, Tcf1, and crypt base cells marker CD24 ( Figure 7B and Figure S8). Finally, we derived organoids from the hyperplastic small intestine of Apc flox14/flox14 Tcf4 flox5/flox5 Lgr5-CreERT2 mice treated with tamoxifen. As the neoplastic tissue contained Tcf4-negative tumor cells ( Figure 5A), we hypothesized that in growing tumors, some transformed ISCs might lose their dependency on Tcf4. Control organoids were established from tamoxifen-treated Apc flox14/flox14 Tcf4 +/+ Lgr5-CreERT2 mice, and the Tcf4 protein was visualized in both types of organoids (that were growing as typical tumor spheroids) using fluorescent microscopy. Nevertheless, tumor organoids derived from both mouse strains produced similar levels of Tcf4 (Figure 8). This showed that Tcf4 is indispensable for organoid establishment or growth, even in the absence of Apc. Next, we derived small intestinal or colon organoids from Apc flox14/flox14 Tcf4 flox5/flox5 Lgr5-CreERT2, Apc flox14/flox14 Tcf4 +/+ Lgr5-CreERT2, and Apc +/+ Tcf4 +/+ Lgr5-CreERT2 mice, and induced single Apc or double Apc/Tcf4 inactivation by 4-OHT. Apc inactivation in organoids harboring the unmodified Tcf4 gene led to the formation of fast growing "spheroids" displaying a cyst-like morphology lacking the budding crypt domains (the morphological change was mainly evident after organoid passage). However, simultaneous deletion of Apc and Tcf4 caused organoid demise after organoid culture splitting ( Figure 7A and data not shown). Subsequent qRT-PCR analysis revealed that the Apc inactivation caused robust upregulation of Lef1 mRNA and a moderate increase in the expression levels of other Wnt signaling target genes Axin2, Tcf1, and cyclin D1 in both colonic and small intestinal organoids. Additionally, Apc-deficient small intestinal organoids displayed increased levels of Tcf3 mRNA. However, concomitant deletion of Apc and Tcf4 led to a substantially diminished expression of all tested Wnt target genes, including Axin2, Lef1, Lgr5, and, Tcf1, and crypt base cells marker CD24 (Figures 7B and S8). Finally, we derived organoids from the hyperplastic small intestine of Apc flox14/flox14 Tcf4 flox5/flox5 Lgr5-CreERT2 mice treated with tamoxifen. As the neoplastic tissue contained Tcf4-negative tumor cells ( Figure 5A), we hypothesized that in growing tumors, some transformed ISCs might lose their dependency on Tcf4. Control organoids were established from tamoxifen-treated Apc flox14/flox14 Tcf4 +/+ Lgr5-CreERT2 mice, and the Tcf4 protein was visualized in both types of organoids (that were growing as typical tumor spheroids) using fluorescent microscopy. Nevertheless, tumor organoids derived from both mouse strains produced similar levels of Tcf4 (Figure 8). This showed that Tcf4 is indispensable for organoid establishment or growth, even in the absence of Apc.  bottom show organoids one day after splitting. Scale bar: 200 μm; (B) qRT-PCR analysis of organoids derived from the indicated mouse strains. Total RNA was isolated from four (parallel) organoid cultures 48 hr after 4-OHT or ethanol addition to culture media. Diagrams show the relative expression levels of the indicated genes in ethanol-treated organoids when compared to 4-OHTtreated organoids of the corresponding genotype. RNA levels were normalized to the Actb mRNA levels; Ubb represents an additional housekeeping gene. Error bars indicate SDs.

Redundancy of LEF/TCF Family Members in Human APC-Deficient Cells
Active Wnt signaling represents a hallmark of the majority of CRC [16]. In addition, initial studies indicated that TCF4 is the major mediator of aberrant Wnt signaling in CRC cells [11,14]. Our recent analysis of various gene expression databases indeed confirmed that TCF4 displays-among the LEF/TCF family members-the highest expression in the human intestine and colon. Nevertheless, the same analysis also showed that besides TCF4, all other TCFs are expressed in healthy or tumor intestinal tissue [18]. The latter finding was verified experimentally by immunoblotting, which showed that LEF/TCF family members are produced in APC-deficient SW480 CRC cells and M1 cells (Figures 9A and S9A). M1 cells were generated from parental HEK293 by transcription activator-like effector nucleases (TALEN)-mediated targeting of the APC locus. Consequently, the cells produce a truncated form of APC, leading the constitutive Wnt pathway activity, thus mimicking the situation in the majority of CRC. To test the TCF4 contribution to TCF/βcatenin-dependent transcription, we employed the CRISPR/Cas9 system to disrupt TCF4 and/or its closest paralog TCF3. We generated TCF3/TCF4 single-or double-deficient SW480 and M1 single cell clones. The clones were viable and did not change their proliferation rate when compared to the parental cells ( Figure S10). Total RNA isolated from several cell clones obtained from each gene targeting experiment was used to analyze the expression levels of Wnt-responsive genes AXIN2 [48, Figure 8. Small intestinal organoids derived from the transformed epithelium produce Tcf4. Fluorescent microscopy images of tumor organoids (spheroids) derived from the small intestine of Apc flox14/flox14 Tcf4 +/+ Lgr5-CreERT2 (Apc KO/KO Tcf4 +/+ ) and Apc flox14/flox14 Tcf4 flox5/flox5 Lgr5-CreERT2 (Apc KO/KO Tcf4 KO/KO ) mice 28 days after tamoxifen administration. Samples were stained using an anti-TCF4 antibody (green fluorescence) and they were counterstained using diamidino-2-phenylindole (DAPI) stain (blue nuclear fluorescence). Scale bar: 100 µm.

Redundancy of LEF/TCF Family Members in Human APC-Deficient Cells
Active Wnt signaling represents a hallmark of the majority of CRC [16]. In addition, initial studies indicated that TCF4 is the major mediator of aberrant Wnt signaling in CRC cells [11,14]. Our recent analysis of various gene expression databases indeed confirmed that TCF4 displays-among the LEF/TCF family members-the highest expression in the human intestine and colon. Nevertheless, the same analysis also showed that besides TCF4, all other TCFs are expressed in healthy or tumor intestinal tissue [18]. The latter finding was verified experimentally by immunoblotting, which showed that LEF/TCF family members are produced in APC-deficient SW480 CRC cells and M1 cells ( Figure 9A and Figure S9A). M1 cells were generated from parental HEK293 by transcription activator-like effector nucleases (TALEN)-mediated targeting of the APC locus. Consequently, the cells produce a truncated form of APC, leading the constitutive Wnt pathway activity, thus mimicking the situation in the majority of CRC. To test the TCF4 contribution to TCF/β-catenin-dependent transcription, we employed the CRISPR/Cas9 system to disrupt TCF4 and/or its closest paralog TCF3. We generated TCF3/TCF4 single-or double-deficient SW480 and M1 single cell clones. The clones were viable and did not change their proliferation rate when compared to the parental cells ( Figure S10). Total RNA isolated from several cell clones obtained from each gene targeting experiment was used to analyze the expression levels of Wnt-responsive genes AXIN2 [48,51] and SP5 [52]. Representative results are summarized in Figure 9B and Figure S9B. In SW480 cells, the TCF3 absence had no effect on the levels of AXIN2 and SP5 mRNA. Moreover, in single TCF4 knockout cells, we observed either a negligible effect or a slight reduction in the expression levels of both tested genes. Additional disruption of TCF3 in TCF4-deficient cells either had no additive effect, or it further potentiated TCF4 loss ( Figure 9B).
In M1 cells, single or double disruption of TCF3 and/or TCF4 had virtually no effect on AXIN2 and SP5 expression ( Figure S9B). Since the obtained results implied redundancy of TCF3/4 proteins with LEF1 and/or TCF1, we used RNAi to downregulate the production of the latter factors. As shown in Figure 9C and Figure S9C, the most robust decrease in AXIN2 and SP5 expression levels was observed in single TCF4-deficient cells treated simultaneously with siRNA against LEF1-and TCF1-specific siRNAs. TCF3/TCF4 double deficiency had-depending on the individual clone analyzed-either no additional effect, or it further decreased the mRNA levels of the analyzed Wnt target genes. In summary, the results showed mutual interchangeability among TCFs in human cells. Moreover, although we cannot exclude the possibility that different LEF/TCF family members regulate different sets of target genes, we did not observe any transcriptional repression activity of the TCF4 (or TCF3) protein.  51] and SP5 [52]. Representative results are summarized in Figures 9B and S9B. In SW480 cells, the TCF3 absence had no effect on the levels of AXIN2 and SP5 mRNA. Moreover, in single TCF4 knockout cells, we observed either a negligible effect or a slight reduction in the expression levels of both tested genes. Additional disruption of TCF3 in TCF4-deficient cells either had no additive effect, or it further potentiated TCF4 loss ( Figure 9B). In M1 cells, single or double disruption of TCF3 and/or TCF4 had virtually no effect on AXIN2 and SP5 expression ( Figure S9B). Since the obtained results implied redundancy of TCF3/4 proteins with LEF1 and/or TCF1, we used RNAi to downregulate the production of the latter factors. As shown in Figures 9C and S9C, the most robust decrease in AXIN2 and SP5 expression levels was observed in single TCF4-deficient cells treated simultaneously with siRNA against LEF1-and TCF1-specific siRNAs. TCF3/TCF4 double deficiency had-depending on the individual clone analyzed-either no additional effect, or it further decreased the mRNA levels of the analyzed Wnt target genes. In summary, the results showed mutual interchangeability among TCFs in human cells. Moreover, although we cannot exclude the possibility that different LEF/TCF family members regulate different sets of target genes, we did not observe any transcriptional repression activity of the TCF4 (or TCF3) protein.

Conclusions
TCF4 is the major nuclear mediator of canonical Wnt signaling in the mouse intestine and human colorectal cancer cells. However, several groups have reported discrepant results related to the TCF4 function in the embryonic or adult mouse gut tissue. Moreover, recent genetic analysis of human tumor specimens indicated a possible tumor suppressive role of TCF4 in a significant fraction of colorectal carcinomas. We employed a newly generated floxed Tcf4 allele that allows inactivation of all potential Tcf4 isoforms produced in the mouse tissues. The allele was combined with several intestinal-specific Cre drivers to perform continuous or timed Tcf4 gene ablation in intestinal stem cells or throughout the epithelium of the small intestine and colon. Additionally, we utilized the CRISPR/Cas9 system to disrupt the TCF4 gene and its closes homolog TCF3 in two human cell lines and quantified the impact of the disruption on expression of Wnt signaling target genes AXIN2 and SP5.
Targeted deletion of Tcf4 in the adult gut was accompanied by a loss of proliferating cells in both the small intestine and colon. Moreover, lineage tracing experiments showed that adult Tcf4-deficient small intestinal and colon stem cells do not contribute to epithelial self-renewal. During embryogenesis, epithelial expression of the Tcf4 gene was seemingly less essential. The absence of any (strong) phenotype in the Tcf4-deficient developing gut might be caused by incomplete recombination of the floxed allele and/or by direct involvement of non-epithelial Tcf4-expressing cells in the intestinal epithelium formation.
We did not observe a tumor suppressive effect of Tcf4. In fact, concomitant deletion of Tcf4 and Apc resulted in a significant decrease in the size of small intestinal tumors. Moreover, all colonic lesions retained residual Tcf4 expression. Additionally, Tcf4 appeared essential for the growth of small intestinal or colon organoids irrespective of the Apc status. The Tcf4 necessity was mainly manifested during expansion of the organoid cultures. Contrary to the results obtained in the mouse, TCF4 (and TCF3) knockout in APC-deficient human cells had no remarkable effect on cell growth or transcription of the Wnt signaling target genes. Subsequent siRNA experiments confirmed the redundancy of TCF4 with LEF1 and TCF1.
In summary, our results showed the importance of the Tcf4 Wnt effector, mainly in the mouse model of adult intestinal epithelium homeostasis and tumor initiation. In human cells, other TCF/LEF family members substitute for the TCF4 role, probably due to different "wiring" of the intracellular signaling mechanism.