Thyroid hormone-induced expression of Foxl1 in subepithelial fibroblasts correlates with adult stem cell development during Xenopus intestinal remodeling

In the Xenopus laevis intestine during metamorphosis, stem cells appear and generate the adult epithelium analogous to the mammalian one. We have previously shown that connective tissue cells surrounding the epithelium are essential for the stem cell development. To clarify whether such cells correspond to mammalian Foxl1-expressing mesenchymal cells, which have recently been shown to be a critical component of intestinal stem cell niche, we here examined the expression profile of Foxl1 in the X. laevis intestine by using RT-PCR and immunohistochemistry. Foxl1 expression was transiently upregulated only in connective tissue cells during the early period of metamorphic climax and was the highest just beneath the proliferating stem/progenitor cells. In addition, electron microscopic analysis showed that these subepithelial cells are ultrastructurally identified as telocytes like the mammalian Foxl1-expressing cells. Furthermore, we experimentally showed that Foxl1 expression is indirectly upregulated by thyroid hormone (TH) through Shh signaling and that TH organ-autonomously induces the Foxl1-expressing cells concomitantly with appearance of the stem cells in the tadpole intestine in vitro. The present results suggest that intestinal niche cells expressing Foxl1 are evolutionally conserved among terrestrial vertebrates and can be induced by TH/Shh signaling during amphibian metamorphosis for stem cell development.

epithelium, where the stem cells are localized in the trough of intestinal folds 16 similar to the mammalian ones localized in the crypt 17 . Since this intestinal remodeling can be easily induced by thyroid hormone (TH) both in vivo and in vitro 18 , the amphibian metamorphosis provides us an excellent opportunity to study molecular mechanisms underlying development of the stem cells. Previously, by using the X. laevis tadpole intestine, we experimentally demonstrated that connective tissue cells surrounding the epithelium are required for the adult stem cell development in the presence of TH 19 . In addition, our previous in situ hybridization (ISH) analyses of TH response genes indicated that the connective tissue cells close to the adult epithelial primordia highly express several signaling genes such as bone morphogenetic protein (BMP) 4 20 , patched 1 (Ptc1), Gli1 21 and CD44 22 . These results strongly suggest that the connective tissue cells surrounding the adult epithelial primordia play important roles in the stem cell development as a niche component. However, it is not yet known whether these cells express Foxl1 just like the mammalian subepithelial Foxl1-expressing cells, and if so, what signals upregulate Foxl1 expression in the amphibian remodeling intestine. To address this issue, we here examined the expression profile of Foxl1 in the X. laevis intestine during metamorphosis. Our results indicate that Foxl1 is specifically expressed in the connective tissue cells surrounding the adult epithelial primordia and that its expression is indirectly upregulated by TH through sonic hedgehog (Shh) signaling. In addition, the amphibian Foxl1-expressing cells share common molecular and morphological characteristics with the mammalian ones, suggesting their evolutionally conserved roles as the intestinal stem cell niche.

Upregulation of Foxl1 in the intestine during natural and TH-induced metamorphosis.
To study the temporal gene regulation of Foxl1 during intestinal remodeling, we first carried out real-time RT-PCR using total RNA extracted from the X. laevis small intestine at various metamorphic stages (Fig. 1). Since there are two homeologs of Foxl1 gene (Foxl1.L and Foxl1.S) in the allotetraploid genome of X. laevis 23 , we analyzed their expression levels by using the specific primers for each homeolog. The expression of Foxl1.L mRNA was upregulated at stage 60 (early period of metamorphic climax) and reached its peak at stage 62, when the adult epithelial cells are actively proliferating 12 . Thereafter, its expression decreased toward stage 66 (end of metamorphosis) (Fig. 1a). The expression prolife of Foxl1.S mRNA was similar to that of Foxl1.L mRNA (Fig. 1c).
Because TH is known to induce precocious intestinal remodeling, if these Foxl1 homeologs are involved in this process, their expression profiles should be reproduced during metamorphosis induced by exogenouslysupplied TH. Thus, we analyzed their expression in the intestine of premetamorphic tadpoles at stage 54 treated  www.nature.com/scientificreports/ with 10 nM T3 for 1 to 5 days. The expression of Foxl1.L mRNA was significantly upregulated after 2 days of T3 treatment and continued to rise even after 5 days of treatment (Fig. 1b). Again, similar results were obtained for Foxl1.S mRNA, except that its expression was significantly upregulated after 3 days of treatment (Fig. 1d). These results suggest that both Foxl1 homeologs may play some roles in the intestinal remodeling and that their expression is upregulated by T3 in a similar manner. In addition, since the upregulation of gene expression of Foxl1 homeologs was relatively later than that of the direct TH response genes such as Shh 21,24 and was not drastic until 3 days of treatment (up to threefold), it seems likely that their expression is indirectly regulated by T3.
TH indirectly induces Foxl1 expression through upregulation of Shh. The regulation of direct TH target genes is mediated by the thyroid hormone receptor (TR) immediately upon T3 addition and is thus independent of new protein synthesis. To determine whether Foxl1 homeologs are direct transcriptional targets of TH or not, protein synthesis inhibitors, cycloheximide and anisomycin (Chx) were added to the rearing water of premetamorphic tadpoles at stage 54 for 1 h before and throughout 50 nM T3 treatment for 6 h to block protein synthesis 25 (Fig. 2a). The mRNA expression of the positive control, Shh, a known TH direct response gene 24 , was significantly upregulated by T3 even in the presence of Chx. In contrast, the expression of neither Foxl1.L nor Foxl1.S mRNA was significantly upregulated by T3 in the presence of Chx, indicating that their expression is not directly regulated by TH.
In an attempt to find out potential transcriptional regulators of Foxl1 genes, we focused on Shh signaling, since the cells expressing Foxl1 often coexpress Gli1 7 , a transcriptional mediator of Shh signaling 26,27 and Shh is the direct target of T3 21,24 . By using the web tool (http://jaspa r.gener eg.net), we found several putative Gli-binding sites in both Foxl1.L and Foxl1.S genes (not shown) in addition to those reported by Madison et al. 28 . Then, we pretreated premetamorphic tadpoles at stage 54, when the expression level of Shh in the intestine is known to be very low 24 , with DMSO for 4 days followed by treatment with 0, 150 or 600 nM Smoothened agonist (SAG) for 1 day to activate Shh signaling 29 independently of T3. The expression of both Foxl1.L and Foxl1.S mRNAs was significantly upregulated by SAG in a dose-dependent manner (Fig. 2b,c). These results were comparable to that for Gli1, a known target of Shh signaling 30 (Fig. 2d). In contrast, the mRNA expression of intestinal fatty acid-binding protein (IFABP), which is specifically expressed in differentiated absorptive epithelial cells independently of Shh signaling 31 , was not affected by SAG treatment (Fig. 2e). In addition, to inhibit Shh signaling, premetamorphic tadpoles at stage 54 were treated with 2.5 mM cyclopamine (CP), an inhibitor of Shh signaling 32 , in the presence of T3 for 3 days. TH-upregulated expression of both Foxl1 mRNAs was significantly suppressed by CP (Fig. 2f,g). Taken together, our results suggest that TH-activated Shh signaling upregulates the expression of Foxl1 genes through the action of Gli1.
Foxl1 expression profile correlates with intestinal stem cell development. We next immunohistochemically examined spatio-temporal correlations between Foxl1 expression and adult epithelial development in the X. laevis small intestine. Before metamorphic climax, the tadpole small intestine mainly consists of the simple columnar larval epithelium, the immature connective tissue, and thin layers of muscles. Every tissue was negative for Foxl1 throughout the larval period until stage 59 (Fig. 3a), in agreement with very weak Foxl1 expression detected by RT-PCR during this period.
At stage 60, when adult stem cells become detectable as small islets stained red with pyronin (primordia of the adult epithelium) between the larval epithelium and the developing connective tissue in the entire small intestine ( Fig. 3b) 12 , cells positive for Foxl1 became detectable only in the connective tissue underneath the epithelium ( Fig. 3c). At stage 61 when both the adult epithelial primordia and the connective tissue rapidly increase in cell number (Fig. 3d), the connective tissue cells positive for Foxl1 increased in number close to the adult epithelial primordia (Fig. 3e) and then, at stage 62 when the adult epithelium mostly replaces the larval one ( Fig. 3f), most of the connective tissue cells surrounding the adult epithelium were positive for Foxl1 (Fig. 3g). Thereafter, as multiple intestinal folds form towards stage 66, the connective tissue cells positive for Foxl1 rapidly decreased in number and became mostly localized in the trough of intestinal folds (Fig. 3h), where the stem cells are localized 13 . In agreement with the immunohistochemical analysis described above, ISH analysis showed that Foxl1 mRNA was specifically expressed in the connective tissue cells (Fig. 3i) underneath the adult epithelial primordia expressing c-Myc, one of the intestinal stem cell markers 15 , at stage 62 (Fig. 3j).
To investigate the spatio-temporal correlation between the cells positive for Foxl1 and the adult epithelial primordia more directly, double-immunofluorescence labeling was performed in the X. laevis intestine at stage 61. Cells positive for Foxl1 were localized in the connective tissue close to the adult epithelial primordia positive for Msi1 (Fig. 4a), one of the stem cell markers common to the mammalian and amphibian intestines 14,33 . In addition, most of the cells positive for Foxl1 were also positive for Gli1 (Fig. 4b), whose expression agrees well with its mRNA expression shown by our previous ISH study 21 .

Ultrastructure of subepithelial fibroblasts surrounding adult epithelial primordia.
To obtain more information about the morphological characteristics of the Foxl1-expressing cells, we observed by electron microscopy the connective tissue of the X. laevis intestine during stages 61-62, when almost all of the connective tissue cells just beneath the adult epithelial primordia express Foxl1. During this period, the subepithelial connective tissue cells were mainly irregular-shaped fibroblasts. Characteristically, they possessed many long and thin processes whose diameter was mostly less than 200 nm (Fig. 5a) In addition, some of the processes had a moniliform aspect with dilations where the cell organelles often existed (Fig. 5b). Through these long processes, the subepithelial cells often made contact with the adult epithelial primordia (Fig. 5a,b). These ultrastructural Scientific Reports | (2020) 10:20715 | https://doi.org/10.1038/s41598-020-77817-1 www.nature.com/scientificreports/ characteristics coincide well with those of telocytes possessing long processes termed "telopodes" 34 , as well as the Foxl1-expressing cells in the mammalian adult intestine 35 .
TH organ-autonomously upregulates Foxl1 expression in vitro. Finally, to clarify whether the Foxl1-expressing cells are organ-autonomously induced by TH in vitro, we cultured the small intestine isolated from X. laevis stage 57-tadpoles in the presence or absence of T3. In its absence, the larval epithelium remained simple columnar, and every tissue was negative for Foxl1 throughout the cultivation (Fig. 6a). By contrast, in the presence of T3, although the intensity of Foxl1 immunoreactivity was lower than in vivo, cells positive for Foxl1 became detectable around culture day 5 (Fig. 6b) when adult epithelial primordia appear in vitro (Fig. 6c) 14 . They were localized in the connective tissue close to the adult epithelial primordia expressing Msi1 (Fig. 6c) just like  www.nature.com/scientificreports/ those during natural metamorphosis. These results indicate that the Foxl1-expressing cells originate from the tadpole intestine itself before metamorphic climax.

Discussion
In the present study, we have shown by using RT-PCR that the expression of Foxl1 gene homeologs, both Foxl1.L and Foxl1.S, temporally correlates well with stem cell development in the X. laevis intestine during metamorphosis. In addition, although it is difficult to distinguish the cells expressing each homeolog of Foxl1 gene by immunohistochemistry or ISH at present, we found that Foxl1 is specifically expressed in the connective tissue cells, which often coexpress Gli1 and are morphologically identified as telocytes just like the mammalian Foxl1-expressing cells essential for maintaining the epithelial cell-renewal 8 . These findings strongly suggest that Foxl1-expressing cells common to the mammalian ones appear in the amphibian intestine when the stem cells develop during metamorphosis. Therefore, it should be an interesting future topic to clarify the role of Foxl1 in stem cell development of the metamorphosing intestine.
In the X. laevis intestine, we have previously shown that TH directly upregulates the expression of Shh only in the epithelium and that Shh expression reaches its peak in the adult epithelial primordia that actively proliferate 36,37 . By contrast, a Shh receptor, Ptc1, and its downstream transcription factors Glis are specifically expressed in the connective tissue cells 21 . In the present study, we demonstrated that TH indirectly upregulates the expression of Foxl1, both Foxl1.L and Foxl1.S homeologs, by activating Shh signaling and that the expression www.nature.com/scientificreports/ of Foxl1 at least partly overlaps with that of Gli1, whose expression is transiently up-regulated in the connective tissue during metamorphic climax and reaches its maximum level at stage 62 21 similar to the Foxl1 expression. Taken together with the fact that Foxl1 is a target of Gli proteins, we here propose that, in the amphibian intestine, TH induces the adult epithelial primordia to express Shh, which in turn acts on the nearby connective tissue cells to express Foxl1 via Gli1 in a paracrine manner. In the developing mammalian intestine, the expression of Foxl1 has been also shown to be up-regulated by Shh through Gli binding sites 28 , but it still remains unknown whether TH regulates the Foxl1 expression. Although the origin of the Foxl1-expressing cells remains unclear, our culture study indicated that they are derived from the tadpole intestine itself before metamorphic climax.
Since the Shh receptor Ptc1 is expressed only in the connective tissue cells 21 , it can be concluded that the Foxl1expressing cells originate from some cells residing in the connective tissue of the tadpole intestine. Thus, our next step should be directed to clarify what type of larval connective tissue cells can be induced by Shh signaling to differentiate into telocytes that express Foxl1 by making use of transgenic frogs in which the Foxl1-expressing cells are genetically labeled. Our previous studies have elucidated several signals important for the stem cell development in the X. laevis intestine, such as Shh, BMP4, and Wnts 16 . It should be noted that the expression profile of Foxl1 shown by the present study spatio-temporally coincides well with that of BMP4, which is highly expressed in almost all of the connective tissue cells surrounding the adult epithelial primordia and is also upregulated by TH/Shh signaling 38 . Thus, it seems highly likely that Shh-upregulated expression of BMP4 is mediated by Foxl1, as reported in the mammalian intestine where BMP4 is a downstream target of Foxl1 28,39 . Furthermore, in the mammalian intestine, the Foxl1-expressing cells have recently been shown to be a source of Wnts essential for maintaining the epithelial cell-renewal 5,6 . So far, in the X. laevis intestine, Wnt2b and Wnt5a have been identified as TH response genes 40 . In addition, we have previously shown that both canonical and non-canonical Wnt signaling pathways are activated in the stem cells during development 22,41,42 . Thus, to clarify which Wnt genes are the targets of Foxl1 and contribute to the formation of the stem cell niche, the expression and functional analyses of Wnt genes should be further performed in the X. laevis intestine.
In conclusion, we have shown here that subepithelial Foxl1-expressing cells, which shares common molecular and morphological characteristics with mammalian ones, appear by the inductive action of TH/Shh signaling in the amphibian intestine during metamorphosis. To understand the evolutionally conserved molecular mechanisms underlying the stem cell niche formation, it is a challenging future issue to clarify how and what kinds of signals are exchanged between the Foxl1-expressing cells and the nearby stem cells during postembryonic development by making use of gene knockout technologies in the present animal model.

Animals and treatment. Tadpoles of the South African clawed frogs (Xenopus laevis) at Nieuwkoop and
Faber (NF) stages 11 from 54 (premetamorphosis) to 66 (end of metamorphosis) were reared in the laboratory and used throughout the experiments. Some tadpoles at stage 54 were treated with 10 nM 3, 3′, 5-triiodo-Lthyronine (T3) (Sigma-Aldrich, St. Louis¸ MO, USA), whose stock solution was prepared at 50 μM in 10 mM www.nature.com/scientificreports/ NaOH, for 1 to 5 days. To inhibit protein synthesis, cycloheximide and anisomycin (Chx) dissolved in dimethyl sulfoxide (DMSO) were added to the rearing water of premetamorphic stage 54-tadpoles at 20 and 25 mg/l, respectively (final concentration of DMSO was 0.1%). It has been reported that Chx treatment at this concentration for 1 h inhibits more than 99% of protein synthesis in the stage-54 tadpoles 25 . 0.1% DMSO was used for the control. After 1 h of Chx treatment, T3 was added at 50 nM, and the tadpoles were additionally treated for 6 h. To activate Shh signaling, Smoothened agonist (SAG, Enzo Life Sciences, Farmingdale, NY, USA) dissolved in DMSO at 3 or 0.75 mM was used 29 . Premetamorphic tadpoles at stage 54, when the expression level of Shh is very low in the intestine 24,36 , were pretreated with 0.02% DMSO for 4 days. Then, the tadpoles were treated with 0, 150 or 600 nM SAG for 1 day (final concentration of DMSO was 0.02%). To inhibit Shh signaling, cyclopamine (CP, Fujifilm Wako pure chemical, Osaka, Japan) dissolved in 100% ethanol at 5 mM was used 32 . Premetamorphic tadpoles at stage 54 were treated with 10 nM T3 in the presence of 2.5 mM CP or 0.05% ethanol vehicle for 3 days. At least three tadpoles were analyzed for each stage or day of T3, SAG or CP treatment. All experiments using the X. laevis animals were approved by the Animal Use and Care Committee of Nippon Medical School and were carried out in accordance with the relevant guidelines and regulations set by this committee. In addition, to distinguish conventionally adult stem/progenitor cells from the remaining larval cells during stages 60-62 when the larval-to-adult epithelial replacement occurs, other sections were stained with methyl green-pyronin (Muto, Tokyo, Japan) for 5 min 12 . The adult stem/progenitor cells were stained intensely red with pyronin because of their RNA-rich cytoplasm 45 , whereas the larval epithelial cells undergoing apoptosis were stained much weaker during this period. Furthermore, some other sections were double-immunostained at room temperature for 1 h with mixtures of the rabbit anti-Foxl1 (1:100) and the mouse anti-Musashi-1 (Msi1) (1:10; Abnova, Taipei, Taiwan) or anti-Gli1 antibodies (1:20; NovusBio, Centennial, CO, USA). They were then incubated with a mixture of Alexa Fluor 568-conjugated anti-rabbit IgG (1:500; Molecular Probes, Eugene, OR, USA) and Alexa Fluor 488-conjugated anti-mouse IgG antibodies (1:500; Molecular Probes) and analyzed by fluorescence microscopy. All the control sections showed only background levels of signals, if any (data not shown). According to the manufacturer's information, the human immunogen sequences used to raise the antibodies above are nearly identical to the X. laevis counterparts; anti-Foxl1 (96% to Foxl1.L, 100% to Foxl1.S), anti-Msi1 (92% to both Msi1.L and S), and anti-Gli1 antibodies (78% to both Gli1.L and S). At least three intestinal fragments were examined for each developmental stage, and six for each experimental time point after T3 treatment.

Real
In situ hybridization (ISH). X. laevis genomic region containing Foxl1.L gene that consists of a single exon common to both Foxl1.L and Foxl1.S homeologs was cloned from the genomic DNA extracted from the intestine at stage 62 by PCR with primers (5′-GGG CAG ATA TTA CTG ACT TGA TAG TAG GAG C-3′ and 5′-CCT GAG ATA ATA GGA GTG TAT GGG AAC AGG AC-3′). The amplified DNA fragment was used as a template for the nested PCR performed with primers (5′-AAT ATT GAA TTC CTT CCA CTA CAT CCT CAC ATT GCC CAT GTA TGTG-3′ and 5′-AAT ATT CCC GGG TTA AAC AGC ATT TGG AGT GTG ACA GGT TTT CCAC-3′). The PCR product (ca. 5 kb) was double-digested with EcoRI and XmaI and inserted into pBSII-KS+ vector predigested with the same enzymes. Then, the DNA sequence was checked (pBSII-KS+_Foxl1.L(full)). X. laevis c-Myc.S (stem cell-associated transcription factor) 15 cDNA was isolated by PCR with primers (5′-GAG ACC AAT CGC ATG GCA GG-3′ and 5′-GGA AGA GAC AAC CTT GAT ATAG-3′) using a cDNA prepared from the stage-62 intestine as a template followed by the nested PCR with primers (5′-AAT ATT ACC GGT GCC GCC ACC ATG GCA GGA AAG ATG CCT CTT AATG-3′ and 5′-TAT TAC TAG TCT ACT TGT CGT CGT CAT CCT Scientific Reports | (2020) 10:20715 | https://doi.org/10.1038/s41598-020-77817-1 www.nature.com/scientificreports/ TGT AGT CGA CAA AGT TCC TCA GCT GTT GGA G-3′). The amplified DNA was double-digested with AgeI and SpeI and inserted into pT7Ts vector 46 predigested with the same enzymes and sequenced (pT7Ts_c-Myc-FLAG). Nearly full-length of c-Myc.S coding region (ca. 1.2 kb) was digested out from this plasmid with BbvCI (followed by blunt-ending) and AgeI, and inserted into pBSII-KS+ vector predigested with HincII and XmaI (pBSII-KS+_c-Myc probe). These plasmids were linearized to synthesize sense and antisense probes with T7 and T3 RNA polymerase, respectively, by using digoxigenin (DIG) RNA Labeling Mix (Roche Applied Science, Indianapolis, IN, USA). The probes were digested by alkaline treatment (40 mM NaHCO 3 , 60 mM Na 2 CO 3 ) into ca. 200 base-long. Intestinal fragments were isolated from the anterior part of the small intestine just after the bile duct junction in tadpoles at stage 62 and fixed in MEMFA followed by cryosectioning. Tissue sections were prepared at 7 μm and subjected to ISH. ISH was performed by using the sense or antisense probe as previously described 47 . Sense probes did not show any specific signal (not shown). Photographs were taken by using a digital CCD color camera (WRA YCA M-ALASKA, Wraymer, Osaka, Japan) attached to an optical microscope (BX51, Olympus, Tokyo Japan).
Transmission electron microscopy (TEM). Small fragments isolated from the anterior part of the small intestine were fixed in 2.5% glutaraldehyde in 0.1 M cacodylate buffer (pH7.4) at 4 °C for 2 h and postfixed in 1% osmium tetroxide in the same buffer at 4 °C for 2 h. They were then stained en bloc with saturated uranyl acetate and embedded in epoxy resin. Transverse ultrathin sections were stained with 4% lead citrate for 5 min and were examined with a JOEL 200CX electron microscope.
Organ culture. Tissue fragments were isolated from the anterior part of the small intestine just behind the bile duct junction in X. laevis tadpoles at stage 57 (before metamorphic climax) and were cultured for 5 days at 26 °C as we previously described 14,48 . Briefly, they were slit open lengthwise and placed on membrane filters (type HAWP; Millipore, Bedford, MA, USA) put on steel grids in culture dishes. The medium contains 60% Leibovitz's L-15 medium (Gibco-BRL, Grand Island, NY, USA) supplemented with 100 IU/ml of penicillin, 100 μg/ml of streptomycin, and 10% charcoal-treated fetal bovine serum (Gibco-BRL). To induce metamorphic changes, 20 nM T3, 5.0 μg/ml insulin, and 0.5 μg/ml hydrocortisone (Sigma-Aldrich, Louis, MO, USA) were added to the medium. The culture medium was changed every other day.
Western blot analysis. Using pBSII-KS+_Foxl1.L(full) as the template, Foxl1.L coding region fused with FLAG tag was amplified with primers (5′-AAT ATT ACC GGT CCG CCA CCA TGG ACT ACA AGG ATG ACG ACG