RFX6 regulates human intestinal patterning and function upstream of PDX1 Free

Deciphering the factors that control regional patterning during endoderm organogenesis remains a major area of investigation in developmental biology (Thompson et al., 2018; Spence et al., 2011a). The mechanisms that drive specific cell migration and identity require specialized gene regulatory networks at precise times, which are regulated by a combination of signaling pathways and transcription factors (TFs), to give rise to diverse organs. Several major pathways, such as WNT, BMP and Hedgehog, have been shown to have a crucial role in the patterning and formation of endoderm organs (Zorn and Wells, 2007, 2009; Roberts et al., 1995; Thompson et al., 2018). TFs are often responsible for driving organ-specific gene regulatory networks and their loss can result in organ agenesis or abnormal patterning. Regulatory factor X6 (RFX6) is a winged-helix TF that has been shown to have an essential role in endoderm organogenesis and is required for endocrine cell formation in pancreatic islets and enteroendocrine cell (EEC) differentiation in the gastrointestinal (GI) tract (Smith et al., 2010; Piccand et al., 2014). Early in embryonic development, RFX6 is broadly expressed in the epithelium of the foregut and proximal intestine, and becomes restricted to the endocrine pancreas and the proximal small intestine at later stages of development. Reduction or loss of Rfx6 in mouse intestine results in gut malrotation, duodenal atresia, malabsorption, and significant reduction of the enteroendocrine lineage (Soyer et al., 2010; Gehart et al., 2019). In humans, variants of RFX6 result in Mitchell–Riley syndrome, an autosomal-recessive syndrome characterized by neonatal diabetes, small bowel atresia, and malabsorption (Kambal et al., 2019; Concepcion et al., 2014; Trott et al., 2020). Recent studies using an induced pluripotent stem cell (iPSC) line derived from an individual with a similar variant of RFX6 showed reduced pancreatic differentiation (Trott et al., 2020), consistent with previous functional studies showing that Rfx6 is involved in normal pancreas development in mouse (Piccand et al., 2014). However, the molecular mechanisms by which RFX6 regulates the patterning and function of the regions of the human intestine remain unknown.

Another TF that has a crucial role in the development of the pancreas and proximal small intestine is pancreas/duodenal homeobox 1 (PDX1). It has been shown that RFX6 and PDX1 are co-expressed in the developing duodenum and endocrine cells in pancreas and small intestine (Soyer et al., 2010; Smith et al., 2010; Yang et al., 2017). PDX1 has an essential role in pancreas, stomach and duodenum morphogenesis and persists in the adult tissues; Pdx1-null mouse models have shown a lack of mature pancreatic tissue and malformation of the gastroduodenal junction with significantly reduced endocrine cells in the small intestine (Offield et al., 1996; Boyer et al., 2006; Burlison et al., 2008; Chen et al., 2009; Fujita et al., 2008; Fujitani et al., 2006a; Stoffers et al., 1997). In the intestinal tract, PDX1 is maximally expressed in the anterior duodenal region with decreased expression in the distal small intestine. Additionally, PDX1 has been reported to be necessary for the production of gastric inhibitory peptide (GIP)-secreting cells, although multiple cell lineages are lost in PDX1-null models (Chen et al., 2009; Jepeal et al., 2005). Despite an essential role for PDX1 in regulating patterning and development of the duodenum, the factors that regulate PDX1 expression are not known.

Over the past decade, several human organoid models have been developed that enable novel strategies to model diseases modeling and organogenesis (Kechele and Wells, 2019). iPSC-derived organoids are three-dimensional structures that mimic the development of an organ and offer a unique way to study developmental mechanisms and organ functions in human tissue (Carpenter and Rao, 2015). Organoids also possess the unique advantage of being generated from iPSCs, meaning they exhibit the same variant as the individual from whom they are derived, allowing for the study of individual-specific phenotypes at different stages of development. iPSC-derived intestinal organoids (HIOs) can be differentiated in vitro and engrafted under the mouse kidney capsule where they become vascularized and mature to mimic fetal intestine with crypt-villus architecture and the ability to absorb nutrients (Spence et al., 2011b; Watson et al., 2014). We recently used these iPSC-derived human organoids to identify specific pathologies in individuals with a PDX1 variant (Krishnamurthy et al., 2022).

We have identified a patient with a compound-heterozygous variant of RFX6 that caused neonatal diabetes and duodenal malrotation and atresia, resulting in functional loss of the RFX6 protein and abnormal gut patterning. We generated iPSCs and HIOs from this patient to interrogate further the specific roles of RFX6 during intestinal development. This approach allowed us to compare directly the biopsy samples of this patient with the HIOs to discover pathophysiological effects of this variant. Here, we demonstrate that RFX6 is required to maintain duodenal identity and function, and that its loss results in ileal-like tissue. We describe the mechanism by which RFX6 drives PDX1 expression, which activates the gene regulatory network that establishes and maintains duodenal identity. Finally, we largely rescue the phenotypes caused by the RFX6 variants using both CRISPR allele correction and inducible re-introduction of RFX6 or downstream PDX1. This study represents a novel approach to the use of complex individual pathologies to uncover the mechanisms at play in human development.

Several signaling pathways and TFs converge to pattern the small intestine into duodenum, jejunum and ileum. These segments carry out different functions during nutrient absorption, but, importantly, also employ differential gene regulatory networks to generate their identities (Haber et al., 2017). One key TF is the winged helix factor RFX6. Early in development, RFX6 is expressed widely throughout the endoderm but becomes restricted to proximal small intestine and pancreas after embryonic day (E) 13.5 in mouse (Soyer et al., 2010; Smith et al., 2010). A patient presented to Cincinnati Children's Hospital Medical Center (CCHMC) with monogenic diabetes, duodenal atresia, intestinal malrotation, annular pancreas and polyps in the duodenum that were surgically removed for analysis. Exome sequencing of the patient and their parents revealed that this patient carried a compound-heterozygous variant of RFX6, with a frameshift mutation (p. Arg347Lysfs*18) in the paternal allele and a nonsense mutation (p. Gln875*) in the maternal allele (Fig. 1A). Analysis of resected polyps showed them to be gastric in identity expressing the stomach marker CLDN18 with only small regions expressing the intestinal marker CDH17 (Fig. 1B). This developmental phenotype suggested that early stages of organ development were impacted, especially the patterning of the gut tube. However, limited access to human samples has made it challenging to perform deep mechanistic investigation into the cause of these phenotypes. To perform deep phenotyping and investigate the developmental basis of these congenital malformations, we recruited this patient into our study and acquired a blood sample for generation of iPSCs. Analysis of these lines showed them to be karyotypically normal, pathogen free, genetically unique as measured by short tandem repeat analyses, and able to differentiate into definitive endoderm and form early gut spheroids (Fig. S1).

We first determined whether we could phenocopy the pathologies of this patient with intestinal (duodenal) organoids generated from their iPSC lines. We first generated HIOs in vitro, at which stage they represent first trimester human duodenum (Spence et al., 2011b). To promote further development and maturation, HIOs were transplanted into immunocompromised mice and allowed to develop for an additional 12 weeks, at which time they are similar to early third trimester human duodenum (Fig. S2A) (Singh et al., 2023). Analysis of HIOs generated from a control (wild-type, WT) iPSC line confirmed them to be duodenal in nature, as we previously reported (Spence et al., 2011b), expressing PDX1 and CDX2 with little-to-no expression of gastric markers (Fig. S2B-E). We found that RFX6 protein was expressed in duodenal HIOs in vitro and those matured in vivo (Fig. 1C, Fig. S3A-D). In contrast, HIOs made from this patient's iPSCs with RFX6 mutations (RFX6 Mut) had low RFX6 expression (Fig. 1D). RFX6 Mut HIOs had large patches of gastric tissue expressing CLDN18, reminiscent of the gastric polyps found in the patient. Further phenotyping of RFX6 Mut HIOs showed a wide range of abnormalities, including increased goblet cells and a decrease in EECs with loss of many subtypes (Fig. 1B,E, Fig. S4A). Interestingly, the HIOs exhibited a loss of PDX1 expression, which is a key driver of the duodenal gene regulatory network (Boyer et al., 2006; Chen et al., 2009). Furthermore, duodenal HIOs had inappropriate expression of the distal intestinal marker SATB2, a TF that has been recently shown to be a driver of ileal-colonic fate (Fig. 1E,F) (Gu et al., 2022). As a control for this regional phenotype, we generated distal HIOs that are ileal in nature by addition of FGF and CHIR99021 which further pattern the developing hindgut (Tsai et al., 2017; Dessimoz et al., 2006). RFX6 is not normally expressed in the distal small intestine, and, as predicted, distal small intestinal organoids were unaffected by the RFX6 variant (Fig. 1E,F).

Having identified multiple previously unappreciated pathologies in RFX6 Mut HIOs, we wanted to investigate whether these were similarly found in tissue from the patient. Given the pathologic nature of the duodenal polyps, we chose to histologically compare HIOs to a biopsy from a visually normal region of the duodenum from this patient. In all cases, the HIO phenotypes were also observed in the biopsy sample, including reduced PDX1, increased SATB2, increased goblet cells, and a reduction in most EEC populations (Fig. 1E,F, Fig. S4B-D). As expected from our immunofluorescence analysis, expression of markers notably reduced in RFX6 Mut duodenal HIOs included several proximally enriched EEC-peptides, such as GIP, ghrelin and motilin (Fig. S4C), suggesting RFX6 has an important role in the differentiation of intestinal EECs in humans (Piccand et al., 2019). This confirmed previous murine studies that had reported loss of EECs in Rfx6-null mice (Gehart et al., 2019; Piccand et al., 2019). These data suggest that RFX6 variants cause mispatterning of the developing GI tract, with loss of duodenal identity and inappropriate expression of distal markers.

To investigate more comprehensively the impact of the RFX6 variants on intestinal development, we isolated tissue from transplanted HIOs for RNA sequencing (RNA-seq). These transplanted organoids often display some degree of heterogeneity because they are each grown in a different animal with immunocompromised health; additionally, the different background in the parental line from a healthy to a mutant iPSC line can add some variability. Thus, we mainly focused on genes that were differentially expressed after pooling the replicates together. As a baseline, we determined that in vitro Mut HIOs (35D, i.e. 35 days of culture of the HIO, which is the endpoint in vitro and the time at which they are transplanted into mouse) consistently display intestinal markers (e.g. CDX2), but also exhibit phenotypes seen in biopsies, such as loss of PDX1 and increased expression of SATB2 compared with WT HIOs (Fig. S5A). To establish further the variance pre-transplant, we generated RNA-seq data from in vitro HIOs; principal component analysis (PCA) of transcriptomes from WT and RFX6 Mut duodenal 35D HIOs showed 85% variance between WT and Mut whereas between replicates there was only 6% variance. Additionally, the heatmap showed over 2000 differentially expressed genes (Fig. S5B). In addition, we found altered expression of HOX genes in our in vivo HIOs that are known to regulate downstream targets involved in anterior-posterior (AP) patterning in developing embryos of species ranging from flies to humans (Fig. S5C) (Hajirnis and Mishra, 2021). In the GI tract, HOX genes are expressed in the mesenchyme, the epithelium or both, and follow a proximal-to-distal gradient going from the duodenum to the colon. In RFX6 Mut duodenal HIOs, distal HOX genes, such as HOXC10, HOXA10 and HOXA11, were increased in expression compared with WT duodenal HIOs, suggesting that the duodenum of the patient with the RFX6 variant exhibits more ileal identity. However, known proximal HOX genes did not decrease in RFX6 Mut duodenal HIOs, suggesting that there is a mix of duodenal and ileal-like gene expression patterns in RFX6 Mut HIOs. We analyzed additional mesenchymal markers in our in vivo HIOs to determine whether RFX6 variants impact mesenchyme and showed significantly decreased expression of HAND1 and FOXF1, which have been shown to expressed in the duodenal mesenchyme at these stages of gut tube development in mice (Fig. S5D) (Han et al., 2020).

Gene ontology (GO) term analysis of all the genes downregulated in RFX6 Mut duodenal HIOs revealed loss of biological processes associated with duodenal-specific functions (Fig. 2B). For example, the duodenum is a key segment of the small intestine for lipid uptake and metabolism (DiPatrizio and Piomelli, 2015). RFX6 Mut duodenal HIOs showed decreased expression of MTTP, APOC3 and RBP2, which are involved in lipid metabolism processes, compared with WT duodenal HIOs (Fig. 2A) (Iqbal and Hussain, 2009; Hussain, 2014). We observed a trend toward increased expression of genes involved in ileal-specific functions, such as bile acid absorption and vitamin B12 processing, although they were not significant (Fig. 2A) (Ticho et al., 2019; Thompson and Wrathell, 1977). Given that RFX6 variants impacted the expression of lipid-handling genes, we investigated whether RFX6 Mut HIOs had impaired lipid transport. We exposed the lumen of transplanted organoids to the fluorescent lipid BODIPY and quantified lipid uptake (Carten et al, 2011). We first compared BODIPY uptake in WT duodenal HIOs compared with WT ileal HIOs and, as expected, observed robust uptake of BODIPY into the epithelium of duodenal, but not WT ileal, HIOs. In contrast, RFX6 Mut duodenal HIOs did not efficiently absorb lipids (Fig. 2C). Quantification of lipid uptake showed a two- to threefold reduction in RFX6 Mut duodenal HIOs compared with control HIOs. Taken together, these data suggest that the RFX6 variant affects AP patterning during development of the duodenum, resulting in reduction of proximal small intestine functions, such as lipid absorption, and an enrichment of an ileal signature.

Finally, we investigated whether expression of jejunal-specific genes were altered in RFX6 Mut organoids. The jejunum-specific transporters SLC2A2, SLC2A5 and SLC34A2 (Anderle et al., 2005), which participate in glucose, fructose and phosphate uptake, were not expressed in WT or Mut HIOs. This could be because HIOs are still fetal even after maturation in vivo, and many transporters are not expressed until exposed to nutrients, which happens postnatally and leads to the functional differences between the segments of the small intestine (Schubert, 2011; Lema et al., 2020).

The patient in this study carries compound-heterozygous variants in both parental alleles, with the paternal mutation (p. Arg347Lysfs*18) introducing a premature stop codon that deletes 563 amino acids from RFX6. The maternal allele is a nonsense mutation that deletes 52 amino acids. Because the patient's parents did not present with clinical manifestation of RFX6 deficiency, we reasoned that conversion of one mutant allele to a WT allele would be sufficient to restore a more normal duodenal phenotype. Previous reports of RFX6 variants have shown that the larger the deletion, the more severe the symptoms (Kambal et al., 2019). We therefore used CRISPR/Cas9 to convert the paternal mutation into a WT RFX6 allele (Fig. 2D) (Haeussler et al., 2016; Ran et al., 2013; Liang et al., 2017). RNA-seq analysis of the transplanted HIOs revealed that the corrected duodenal HIOs were much more similar to the WT duodenal HIOs than to RFX6 Mut duodenal HIOs (Fig. S5E,F). PCA showed that the primary difference caused by RFX6 variants (PC1;76%) had been completely reversed by CRISPR correction of the paternal allele. However, PC2 showed that there were still differences between control lines and CRISPR-corrected lines. For example, several genes and biological processes specific to the proximal small intestine showed different levels of expression between the WT and corrected HIOs. This could be due to the different genetic backgrounds of the WT control line, or due to reduced RFX6 activity as a result of the maternal variant.

A closer analysis of specific factors that regulate duodenal patterning showed that expression levels of PDX1, GATA4 and several regionally expressed HOX genes was restored in CRISPR-corrected HIOs (Fig. S5G). We also saw restored expression of genes involved in duodenal function, including lipid metabolism (Fig. S5G). CRISPR-corrected duodenal HIOs did not exhibit evidence of gastric polyps, had significantly decreased expression of the ileal-colonic marker SATB2, and goblet cell numbers more comparable to WT duodenal organoids (Fig. 2E). Together, these data show that the rescue of the paternal allele significantly restored a duodenal phenotype to RFX6 compound-heterozygous HIOs, suggesting that RFX6 has a crucial role in establishing duodenal identity and function. However, the expression of some genes was not restored to the levels of control HIOs generated from iPSC lines of different genetic background. This could be due to haploinsufficiency or gene expression variability between organoids generated from diverse genetic backgrounds.

The AP regional identity of the developing gut tube is established at early stages of development. To investigate the mechanisms leading to the abnormal patterning in RFX6 Mut intestine, we performed RNA-seq analysis during the gut tube patterning stage of differentiation (D7) (Fig. 3A) (Spence et al., 2011a,b). This stage resembles ∼E8.5-9.5 in mice, at which point Rfx6 is already expressed (Soyer et al., 2010; Smith et al., 2010). We first confirmed that RFX6 expression in human cultures initiates in this window of time and found that expression starts at D5 and increases through D7 of differentiation (Fig. S3C); PCA analysis of RNA-seq data displayed a 94% variance between WT and RFX6 Mut and over 4000 differentially expressed genes (Fig. S6A). Most of the differentially expressed genes were downregulated in the RFX6 Mut line, suggesting a failure to activate an RFX6-mediated transcriptional program (Fig. S6A). Lineage-tracing experiments in mice have shown that Rfx6 expression is restricted to the epithelium at this stage of development (Smith et al., 2010). Consistent with this, we observed that RNA counts of crucial endodermal patterning regulators, such as CDX2, GATA4, GATA6, HNF4A, were significantly reduced in the D7 gut tube endoderm cultures from RFX6 Mut iPSC lines (Fig. 3B). Consistent with our observation that the patient duodenum had areas of mixed organ identity, we used a list of organ markers from a published atlas and found that RFX6 Mut cultures lost small intestinal organ markers (Fig. 3C) (Yu et al., 2021).

In addition to epithelial patterning defects, we observed a surprising number of expression differences in key signaling factors that regulate endodermal patterning through a series of epithelial–mesenchymal interactions (Zorn and Wells, 2007; Walton and Gumucio, 2021; McCarthy et al., 2020) (Fig. 3D). In particular, WNT3, WNT5B, BMP4, IHH and many downstream targets of these pathways were reduced in RFX6 Mut cultures. Given that WNT signaling is involved in the expression of CDX2, the master regulator of intestinal identity (Sherwood et al., 2011; Chawengsaksophak et al., 1997), it is likely that reduced WNT signaling in D7 cultures causes a delay in CDX2 expression. Known CDX2 targets that are involved in early intestinal development (Kumar et al., 2019) were also reduced in the RFX6 Mut line, suggesting that the abnormal patterning is due to low CDX2 expression (Fig. S6B).

Although RFX6 is expressed in the epithelium, RFX6 variants cause patterning changes in both the epithelium and mesenchyme of D7 cultures. To begin to separate the direct effects of the RFX6 variants in the epithelium and indirect effects on the mesenchyme, we looked at published dataset from Han et al. which mapped endoderm versus mesoderm specific markers in early gut development (Han et al., 2020). We identified endoderm- and mesoderm-specific markers and performed GO analysis. Interestingly, the biological processes that were impacted in our RFX6 Mut cultures included intestinal patterning, epithelium development, tube development and morphogenesis (Fig. S6C,D). Together, these data suggest that there is an initial delay of intestinal patterning as observed by low CDX2 expression in RFX6 Mut cultures. However, at later stages of differentiation, CDX2 is expressed in both control and RFX6 Mut HIOs and at those stages duodenal markers are lost and more distal markers are expressed.

Finally, we investigated whether the CRISPR correction of the paternal allele rescued the patterning of the HIOs at D7 using PCA of RNA-seq data. CRISPR correction restored the PC1 component (76% variance) to nearly that of WT, suggesting that correction of one allele allows transcription of many downstream targets (Fig. 3E). There was, however, an 18% variance in PC2 from the WT and the corrected group, due to their background difference. Transcripts that were restored in the CRISPR-corrected cultures included components of the WNT and BMP pathways known to regulate gut tube patterning (Fig. 3F). When comparing the CRISPR-corrected line with both the Mut and WT there were also differentially expressed genes (Fig. 3E, Fig. S6E,F). Lastly, we found that correction of one RFX6 allele was sufficient to rescue the expression of several key TFs involved in hindgut development and patterning, such as CDX2 and HNF4A (Fig. 3G). Taken together, these data suggest that, although RFX6 expression is specific to the epithelium, it functions to regulate patterning and signaling in the epithelium. Additionally, CRISPR correction of the paternal allele is sufficient to restore duodenal patterning in organoids; however, there remain transcriptional differences in organoids that retain the maternal variant allele. This suggests that the maternal allele may not be fully functional.

During pancreas development, the PDX1 gene is bound by RFX6 protein, suggesting that this gene may be directly regulated by RFX6. In the duodenum, CRISPR correction of one allele in the RFX6 Mut line restored expression of many TF genes, in particular PDX1, which is essential for duodenal patterning (Fig. 4A). We further developed a doxycycline (DOX)-inducible PSC line in which we can temporally induce RFX6 expression in the mutant background. Induction of RFX6 in foregut endoderm was sufficient to induce endogenous PDX1 mRNA and protein expression (Fig. 4A, Fig. S7A-C). Although these findings showed that RFX6 is both necessary and sufficient to regulate PDX1 expression in cultured human foregut endoderm, we wanted to perform similar loss- and gain-of-function experiments in the context of an embryo. To do this, we used the amphibian Xenopus, a vertebrate model in which endoderm development and GI tract organogenesis is highly conserved (Zorn and Wells, 2009; Rankin et al., 2021). Xenopus rfx6 shares a similar expression pattern as observed in mammals (Pearl et al, 2011; Smith et al., 2010), which we verified by in situ hybridization in foregut progenitors at Nieuwkoop and Faber stage (NF) 20 [∼1 day post-fertilization (d.p.f.)] and in pancreas, stomach and duodenum progenitors at NF33 (∼2.5 d.p.f.) (Fig. S8A). Knockdown of rfx6 in Xenopus using a previously validated translation-blocking morpholino (MO; Pearl et al., 2011). We tested the efficiency of the MO knockdown by titrating the MO for its ability to block a tagged Rfx6 protein (FLAG tagged) and showed immunostaining of the FLAG-tagged protein at NF20 (Fig. S8B,C). This MO revealed a loss of pdx1 at NF33 and microdissection of the gut tube at NF43 (∼4 d.p.f.) revealed a hypoplastic, malrotated GI tract with strong reduction of the pdx1 stomach/duodenum domain, a phenotype similar to that of the mouse Rfx6^−/− knockout (Fig. 4B, Fig. S8D,E) (Smith et al., 2010). Interestingly, we observed an expansion of the distal hindgut satb2⁺ domain into the midgut and foregut territories at NF33, suggesting an evolutionarily conserved function for rfx6 in regulating vertebrate GI tract patterning (Fig. 4B). Injection of mRNA encoding human RFX6 was able to rescue these molecular changes at NF33 and NF43 (Fig. 4B).

We next utilized Xenopus to interrogate further the functional impact of the two observed variants in human RFX6. We constructed dexamethasone (DEX)-inducible versions of human WT RFX6 and the two variants, in which the ligand-binding domain of the human glucocorticoid receptor (GR) is fused to the N terminus of RFX6; this GR-Dex induction system has been widely used for decades in Xenopus to control TF activity at a temporal level (Sive and Bradley, 1996; Zorn et al., 1999; Rankin et al., 2021). mRNA encoding the GR-TF fusion is synthesized in vitro and injected into embryos; the mRNA is translated in vivo by the embryo and the GR domain functionally sequesters the TF in the cytoplasm until addition of DEX to the culture buffer, at which point the TF can translocate to the nucleus and regulate target gene expression (Fig. 4C).

We compared the ability of human WT RFX6 and the two variants to induce pdx1 ectopically and suppress satb2 by targeted injections into the Xenopus hindgut endoderm (Fig. 4D), utilizing a hindgut explant dissection assay that also allowed for assessment of direct induction by pre-treatment of explants with cycloheximide (CHX) before temporal DEX TF induction (Fig. 4C,D). WT human RFX6 and the p.Gln875*stop variant were both able to robustly induce pdx1 and suppress satb2 in Xenopus hindgut endoderm, whereas the p.Arg347Lys.fs18*stop variant had no activity (Fig. 4D), suggesting this variant creates a null, non-functional allele. Similarly, in the CHX direct target assay, both the WT and p.Gln875*stop variant were able to induce pdx1 in the presence of CHX, suggesting that pdx1 is a direct target of RFX6 in vivo, whereas again the p.Arg347Lys.fs18*stop variant had no activity (Fig. 4E). As a positive control for CHX effectiveness, we assayed the expression of the pancreas TF gene ptf1a, induction of which by RFX6 was largely blocked by CHX treatment (Fig. S8F). Together, these results suggest that the p.Gln875*stop variant, resulting in a loss of the last 52 amino acids of the C terminus of RFX6, does not affect RFX6 transcriptional activity, whereas the more severe variant, resulting in p.Arg347Lys.fs18*stop, creates a null, non-functional allele in vivo.

During pancreas development, RFX6 has been shown to bind to DNA at sites near to the PDX1 gene; however, no functional analysis of RFX6-binding sites in PDX1 regulatory domains has been performed in any context. To investigate further the transcriptional regulation of pdx1 by RFX6, we bioinformatically examined a known, evolutionarily conserved enhancer upstream of human PDX1, of which the orthologous mouse region is sufficient to drive reporter expression in pancreas, stomach and duodenum (Gannon et al., 2001; Fujitani et al., 2006a), indicating functional enhancer activity. Evolutionary comparison of this human PDX1 upstream enhancer region (hg19_dna range=chr13:28491064-28492299) revealed significant conservation amongst several genomes; using the TF motif searching tool CisBP (Weirauch et al., 2014), six potential RFX-binding motifs were predicted (Table S1). Analysis of published ChIP-seq studies in human PSCs during directed endoderm differentiation for the key endoderm identity TFs FOXA2 and GATA4 (Vinckier et al., 2020), as well as for the midgut/hindgut TF CDX2 (Kumar et al., 2019), revealed binding of these important TFs to this conserved human PDX1 upstream enhancer (Fig. 4F), suggesting it also regulates developmental expression of PDX1 in human endoderm. We thus utilized the Xenopus model to interrogate potential RFX6-dependent regulation of this predicted human PDX1 enhancer. Luciferase reporters were constructed containing the WT human PDX1 upstream enhancer region or a version in which all six CisBP-predicted RFX motifs were mutated (Fig. 4G). Injection of the WT Human PDX1:luc reporter into foregut endoderm (targeting the progenitor domain of the pancreas, stomach, duodenum) revealed robust reporter activity, whereas injection of the enhancer:luc reporter into the satb2⁺ hindgut domain resulted in negligible activity (Fig. 4H,I) at both NF20 and NF33. Endogenous Xenopus Rfx6 was necessary for reporter activation in the foregut, as Rfx6 knockdown by MO prevented foregut-targeted PDX1 enhancer reporter activity. In support of this, ectopic injection of WT human RFX6 mRNA into the satb2⁺ hindgut domain resulted in robust PDX1 enhancer activity (Fig. 4I), and mutation of the six CisBP-predicted RFX motifs in the human PDX1 enhancer abolished reporter activity in the foregut domain as well as negated the ability of WT human RFX6 to activate the reporter in the hindgut (Fig. 4I). We further verified that the human p.Gln875*stop RFX6 variant retained ability to activate the human PDX1 enhancer, whereas the p.Arg347Lys.fs18*stop variant had no ability to activate the enhancer (Fig. S8G). Together, these results suggest that RFX6 controls vertebrate GI tract patterning in part by direct activation of PDX1 via a conserved upstream enhancer.

RFX6 and PDX1 have overlapping expression patterns in the nascent duodenal endoderm and both genes are required for normal duodenal development. Our data show that PDX1 expression is lost in RFX6 variants and suggest that PDX1 is a direct RFX6 target. This suggests that some or all of the phenotypes associated with loss of RFX6 could be due to loss of PDX1 expression. To identify PDX1-dependent versus -independent roles of RFX6, we used RFX6 Mut iPSCs to generate tetracycline-inducible RFX6 or PDX1 constructs whereby we could induce expression of either gene at specific time points (Fig. S7A,B) and identify downstream transcriptional responses. We validated that a 48 h DOX induction of RFX6 was sufficient for PDX1 expression (Fig. S7C). We differentiated RFX6 Mut iPSCs and induced expression of either RFX6 or PDX1 separately after D4 of differentiation and maintained induced expression until harvest at D7, the gut tube stage (Fig. 5A). PCA of gut spheroids with induced RFX6 and PDX1 revealed that gut tube endoderm in which RFX6 had been induced was more similar to that of the WT than to that of cultures in which PDX1 had been induced (Fig. 5B). A comparison of differentially expressed genes in response to RFX6 or PDX1 expression showed that RFX6 expression rescued 58% of the transcripts that were downregulated in the RFX6 Mut cultures. In contrast, tetracycline-induced PDX1 expression rescued only 25% of the transcripts that were lost in the RFX6 Mut cultures (Fig. 5C, Fig. S7D,E). From these comparisons, we identified which differentially expressed genes in the RFX6 Mut gut tube cultures were PDX1 dependent (rescued by PDX1 expression) compared with those that were not rescued by PDX1 expression. GO analysis of genes that appeared to be in the RFX6-PDX1 regulatory axis highlighted duodenal development (Fig. 5D). Expression of either RFX6 or PDX1 in the RFX6 Mut line was able to rescue several key gut-patterning markers, such as CDX2, SATB2 and GATA4, although PDX1 induction was not able to rescue RFX6 expression, supporting the suggestion that PDX1 acts downstream of RFX6 (Fig. 5E). Whereas the majority of PDX1-regulated genes were also regulated by RFX6, 67% of the 2070 differentially expressed genes in the RFX6 Mut samples were not regulated by PDX1, suggesting that these are regulated by RFX6 independently of PDX1 (Fig. 5D, Fig. S7F). In particular, RFX6 appeared to target expression of numerous genes involved in signal transduction pathways that play essential roles in gut tube patterning and development, such as the WNT, BMP and Hedgehog pathways. RFX6 significantly upregulated transcription of multiple ligands, antagonists and receptors, suggesting that RFX6 impacts intestinal development through regulation of these important signaling pathways (Fig. 5F). Lastly, we observed that RFX6 induction was also able to upregulate other members of the RFX family, mainly RFX4-7, whereas PDX1 failed to induce most RFX genes (Fig. 5G). Taken together, these data show that RFX6 is responsible for the correct patterning of the small intestine by activation of PDX1 in the gut tube epithelium as well as by regulating signaling pathway components that are known to be involved in gut tube patterning.

We identified that PDX1 expression in the RFX6 Mut gut tube endoderm was sufficient to rescue early expression of duodenal patterning markers. However, it was not clear whether early rescue of duodenal patterning was stable at later stages of development. To test this, we used the PDX1 DOX-inducible line to express PDX1 during the early stages of duodenal development that occur in vitro and then we extended development for an additional 10 weeks by transplanting HIOs into immunocompromised mice (Fig. 6A). We found that expression of PDX1 during HIO growth in vitro was not sufficient to maintain duodenal identity during the added 10 weeks of development in vivo (Fig. 6B,C). Moreover, we observed restoration of transcripts involved in proximal small intestine-specific functions, such as lipid metabolism, transport and absorption (Fig. 6D), including genes such as APOB, APOA4 and APOC3 being upregulated in RFX6 Mut HIOs that had DOX-induced PDX1 expression (Fig. 6E), suggesting that this phenotype associated with loss of function of RFX6 could be restored by PDX1. However, if we maintained PDX1 expression in vivo by feeding transplanted mice with DOX-containing chow, we observed restoration of markers of normal duodenal patterning, mainly the lack of SATB2 expression and maintenance of PDX1 expression in the epithelium, and lower numbers of goblet cells (Fig. 6F,G). Although we also saw an increase in the number of EECs in PDX1-expressing HIOs, they were still lower in number compared with controls (Fig. 6F,G). One possible reason is that EECs have higher levels of PDX1 than other epithelial cell types and DOX-containing chow might not induce high enough levels of PDX1 expression to promote robust EEC differentiation. We therefore augmented DOX levels by injecting transplanted mice with an additional 10 μg/kg dose of DOX 5 days before harvesting the HIOs. This resulted in a rescue of the number of EECs in RFX6 Mut HIOs to nearly WT levels (Fig. 6H). Our data suggest that PDX1 not only initiates the patterning of the duodenum and midgut, but is required to be maintained to keep duodenal function and identity (Fig. 6I). Altogether, these data show that expression of PDX1 can partially rescue the phenotypes seen in RFX6 Mut duodenal HIOs and that PDX1 must be actively maintained to preserve duodenal identity.

In this study, we generated patient-derived intestinal organoids to model Mitchell–Riley Syndrome, which is caused by a compound-heterozygous variant of RFX6. Mitchell–Riley syndrome is a recessive syndrome characterized by neonatal diabetes, pancreatic hypoplasia, intestinal atresia, chronic diarrhea and malrotation (Concepcion et al., 2014). We uncovered previously unappreciated roles of RFX6 in early patterning of the small intestine and maintenance of duodenal identity, plus we were able to characterize possible downstream targets. We demonstrated that HIOs are an ideal model in which to mimic development and to manipulate genetic rescue experiments to characterize the mechanism of a disease (Spence et al., 2011b; Krishnamurthy et al., 2022). Genetic correction by CRISPR/Cas9 was used to show that one allele of RFX6 was sufficient to pattern the small intestine correctly and maintain duodenal identity. A tetracycline-inducible model also demonstrated that PDX1 can rescue the phenotypes of the RFX6 variant as a downstream target. This study provides additional evidence for the use of iPSC-derived organoids as an alternative to correct variants and generate healthy tissue for future transplantation. Here, we describe the role of RFX6 in intestinal development, patterning and maintenance. This RFX6 variant misregulates several signaling pathways in both the epithelium and mesenchyme, which leads to a delay in CDX2 expression, potentially due to the deranged crosstalk between mesoderm and endoderm. We showed that RFX6 is required to promote a duodenal identity and without it we saw conversion to a more distal pattern as marked by SATB2 and HOX genes. We also showed that the establishment of duodenal identity by RFX6 involves the direct regulation of PDX1 and that PDX1 expression was sufficient to rescue some, but not all, of the RFX6 transcriptional program. At later stages of HIO development, loss of RFX6 led to loss of duodenal function as measured by lipid transporter activity. Lastly, we validated that RFX6 is required for enteroendocrine differentiation as shown in mouse and human (Gehart et al., 2019; Piccand et al., 2019).

Loss of RFX6 caused several phenotypes in the intestinal organoids; however, we wanted to determine whether these could be corrected. CRISPR correction of one RFX6 allele rescued the phenotypes at the hindgut development stage and at the more mature intestine stage, which showed that one WT allele is sufficient to allow RFX6 function in intestinal development and duodenal identity maintenance. We also attempted to rescue the RFX6 variant phenotypes with PDX1. It has been shown that PDX1 is co-expressed in endocrine cells with RFX6, and PDX1 null models have been shown to have intestinal atresia (Chen et al., 2009; Boyer et al., 2006; Fujita et al., 2008; Fujitani et al., 2006a; Yang et al., 2017). A DOX-inducible construct allowed us to introduce PDX1 at different time points, which showed that PDX1 induction can correct the intestinal phenotypes caused by the RFX6 variant as early as the hindgut development stages. Additionally, we showed a significant overlap in differentially expressed genes when inducing RFX6 and PDX1 expression, respectively, which is evidence for redundancy in the roles of these TFs in intestinal maintenance. Finally, we showed that PDX1 not only has to be expressed during the gut developing stage, but is required to be continually expressed to maintain duodenal identity. We were able to identify a plethora of phenotypes caused directly and indirectly by loss of RFX6, which suggests that it has a role in regulating the crosstalk between mesenchyme and epithelium during gut development. We were able to identify the epithelial-specific mechanism of duodenum maintenance via PDX1 and identify candidates that could be involved in the intermediate steps in the regulation of signaling pathways between mesenchyme and epithelium. Additionally, along with evidence from both parents with normal RFX6 function and evidence from our CRISPR correction and Xenopus data, we were able to determine that the maternal allele retains transcriptional activity whereas the paternal allele is equivalent to complete loss of function. This provides an insight into RFX6 function relating to each domain of the protein. Further studies are needed to analyze the structural-functional correlation analysis in more detail; however, our data suggest that the paternal allele, having the DNA-binding domain intact, still loses nearly all RFX6 transcriptional activity.

In summary, we developed an HIO model from a patient with compound-heterozygous variants in RFX6 that led us to identify new pathologies of the duodenum of individuals with such variants. We demonstrated that the paternal allele was null and that CRISPR correction was sufficient to restore duodenal identity, in part through regulation of PDX1, consistent with studies in the developing pancreas in which PDX1 was a target (Trott et al., 2020; Cheng et al., 2019; Piccand et al., 2014). Further to this point, we identified a conserved enhancer in PDX1 that is regulated by RFX6 during gut development in embryos. Consistent with our findings, while this manuscript was in revision, a different paper was published that supports our data of RFX6 having a crucial role in gut patterning (Nakamura et al., 2024); however, their main focus was early endoderm development and did not focus on divergence of downstream transcriptional networks. Some limitations of our study were that there is variability between organoids and specially between parental cell lines; additionally, the differentiation protocol requiring kidney capsule transplantation is technically challenging and lengthy. However, this study helps uncover the spatiotemporal role of RFX6 in the development of the intestine and the maintenance of intestinal function with the characterization of a downstream target in PDX1. We were able to identify a mechanism by which the duodenum is initially patterned and demonstrate how its maintenance is also orchestrated by RFX6, either via PDX1 or independently. This study uses a combination of in vitro, in vivo and ex vivo models including an embryonic model (Xenopus) to validate transcriptional mechanisms and abnormal patterning caused by a variant of RFX6. Finally, our work provides evidence of the translational potential of PSC-derived organoids and genetic technologies that are broadly translatable to other organs, which opens the path for engineered tissue to be an alternative for transplantation after genetic correction.

The iPSC lines were generated from the patient as previously described (Zhu et al., 2016). Informed consent was obtained from the patient for the collection of whole blood. Peripheral blood mononuclear cells were isolated and used to generate an iPSC line at the Pluripotent Stem Cell Facility at CCHMC. iPSCs were maintained in feeder-free culture. Cells were plated on hESC-qualified Matrigel (BD Biosciences) and maintained at 37°C with 5% CO₂ with daily removal of differentiated cells and replacement of mTeSR1 media (STEMCELL Technologies). Cells were passaged routinely every 4 days using Dispase (STEMCELL Technologies). HIOs were generated according to protocols established in our lab (Spence et al., 2011b; Kechele and Wells, 2019; Tsai et al., 2017) and experiments with human iPSCs were approved by the CCHMC ESCRO committee (Protocol #EIPDB2713).

HIOs were removed from Matrigel 28-35 days after spheroid generation, and were transplanted under the kidney capsule of immune-deficient NOD.Cg-Prkdc^scidIl2rg^tm1Wjl/Szj (NSG) mice (Watson et al., 2014). NSG mice were maintained on Bactrim chow for a minimum of 2 weeks prior to transplantation and thereafter for the duration of the experiment (8-14 weeks) except for DOX-chow during transplant for specific experiments. Mice received additional intraperitoneal DOX injection (1 ml of 0.5 µg/ml solution per kg body weight) 5 days before harvest.

CRISPR/Cas9 was used to correct the mutation in the paternal allele in the RFX6 patient iPSC line. The guide RNA used was ACATAAAAATTGGGAACAGT. A phosphorothioated single-stranded DNA donor oligo (T*A*A*TGCATTTTTTAACAATGAAGCATTTAACACATAGCCTTCTTTGTAGCTTATTAGCAGACATAAGAAATTTTGCTAAgAAcTGGGAgCAaTGGGTTGTTTCATCCTTGGAAAACTTGCCAG*A*A*G; substitutions are underlined and asterisks indicate another nucleotide) was designed to include the nucleotide insertion according to published methods (Liang et al., 2017). The iPSCs were reverse transfected with plasmid and donor oligos using TranIT-LT1 (Mirus). After green fluorescence-activated cell sorting, iPSCs were re-plated in Matrigel. Single clones were manually excised for genotyping, expansion and cryopreservation. Correctly targeted clones were identified by PCR, enzyme digestion, and Sanger sequencing.

For inducible overexpression, we obtained human RFX6 cDNA in pENTR221 (from ORF Clone Collection ID IOH27199), which was then cloned into lentiviral destination vector pInducer20 (gift from Stephen Elledge, Addgene #44012) using Gateway cloning methods with LR Clonase II (Invitrogen). Successful cloning was confirmed using Sanger sequencing. Lentiviral particles were generated by CCHMC Viral Core. iPSCs were transduced by addition of lentiviral supernatant to the culture medium immediately after passaging, which was then replaced with fresh medium after 24 h of exposure. The cells were selected with G418 (100 μg/ml) for 4 days, beginning on the following day. At this point, the colonies had normal morphology and normal characteristics, whereas mock-transduced cells were dead. The cell line was the maintained under standard culture conditions, with intermittent exposure to G418 to maintain resistant cells. Inducible RFX6 expression was validated by qPCR 24 h after exposure to DOX (0.5 μg/ml) at the pluripotent cell stage. In the same way, a PDX1-inducible line was generated as previously described (Krishnamurthy et al., 2022). They were maintained in mTeSR with intermittent G418 selection.

The initial amplification step for all samples was carried out using the Ovation RNA-Seq System v2 kit (Tecan Genomics). The assay was used to amplify RNA samples to create double-stranded cDNA. The concentrations were measured using the Qubit dsDNA BR assay. The cDNA size of each sample was determined using an Agilent HS DNA chip. Libraries were then created for all samples. Specifically, the Illumina protocol, the Nextera XT DNA Sample Preparation Kit, was used to create DNA library templates from the double-stranded cDNA. The concentrations were measured using the Qubit dsDNA HS assay. The size of the libraries for each sample was measured using the Agilent HS DNA chip. The samples were placed in a pool. The concentration of the pool was optimized to acquire at least 40 million reads per sample. The investigator requested the use of Paired-End 100 bp Reads, so for sequencing a NovaSeq S1 (200 cycles) v1.5 flow cell was used. Libraries were sequenced on the Illumina NovaSeq 6000 system following the manufacturer's protocol.

RNA-seq data were analyzed using Computational Suite for Bioinformaticians and Biologists (CSBB, v3.0.0) using the ‘Process-RNASeq_PairedEnd’ and ‘Generate-TPM_Counts_Matrix’ modules (https://github.com/praneet1988/Computational-Suite-For-Bioinformaticians-and-Biologists). FASTQs had adapters trimmed using a tool called bbduk, and quality was checked using fastqc. Bowtie2+RSEM was used to align and count transcripts, using reference genome (Genome Reference Consortium Human Build 38). From the RSEM results, count/TPM matrices were generated for the genes/isoforms. Further analysis and data visualization was performed using RStudio v4.2.1 with the following package: DESeq2, ggplot2, plyr, ggrepel, EnhancedVolcano, pheatmap (Anders and Huber, 2010; Wickham, 2011).

Tissue was fixed in 4% paraformaldehyde, cryopreserved in 30% sucrose, embedded in OCT, and frozen at −80°C before cryosectioning. Cryosections (8 μm thick) were mounted on Superfrost Plus slides and permeabilized, blocked and stained according to standard protocols (as described by Spence et al., 2011b). Primary antibodies used are listed in Table S2 and all secondary antibodies were conjugated to Alexa Fluor 488, 546/555/568 or 647 (Invitrogen, A-21206, A-21202, A-11055) and used at 1:500. Images were acquired using a Nikon A1 GaAsP LUNV inverted confocal microscope and NIS Elements software (Nikon). Quantification and analysis were carried out using Nikon software (version 4.11.0).

RNA was extracted using the Nucleospin RNA extraction kit (Macharey-Nagel) and reverse transcribed into cDNA using Superscript VILO (Invitrogen) according to the manufacturer's instruction. qPCR primers were designed using NCBI PrimerBlast. Primer sequences are listed in Table S3. qPCR was performed using Quantitect SYBR Green PCR kit (QIAGEN) and a QuantStudio 3 Flex RT PCR System (Applied Biosystems). Relative expression was determined using the Delta Cycle method and normalizing to PPIA (cyclophilin A). Samples from at least three independent passages were used for quantification.

WT adult X. laevis frogs were purchased from the National Xenopus Resource Center (NXR, Woods Hole, MA, USA; RRID:SCR_013731). Ovulation, in vitro fertilization, embryo de-jellying, and microinjection were performed as described (Sive and Bradley, 1996). Human RFX6 cDNA clone was purchased from Horizon Discovery (OHS6084-202637733) and Gateway sub-cloned from its entry vector pENTR223 into the expression vector pCSf107mT-Gateway-3′myc (Addgene #67617) using clonase (Thermo Fisher Scientific, 11791020) according to the manufacturer's instructions. SacII-linearized plasmid template was used to make mRNA for injection using the Ambion mMessage mMachine SP6 RNA Synthesis Kit (Thermo Fisher Scientific, AM1340) according to the manufacturer's instructions.

The WT human RFX6 coding region (NCBI Reference Sequence: NM_173560.4) and mutated versions (a 1 bp mutation nucleotide C2623T or a 13 bp deletion nucleotides 1040-1052 of the coding sequence) were also commercially synthesized with 5′EcoRV and 3′SpeI ends and cloned into the pCDNA3.1 vector by GenScript. To construct the inducible GR-RFX6 WT or mutants used in CHX direct target assays, the respective RFX6 WT or mutant versions were released from pCDNA3.1 via EcoRV/SpeI digests, gel purified, and cloned 5′EcoRV 3′SpeI into the pT7TS-GR-HA vector. XbaI-linearized plasmid templates were used to make mRNA for injection using the Ambion mMessage mMachine T7 mRNA Synthesis Kit (Thermo Fisher Scientific, AM1344) according to the manufacturer's instructions. One-hundred picograms total mRNA for WT or mutant GR-RFX6 RNAs were injected at the 8-cell stage into ventral-vegetal blastomeres to target the hindgut (50 pg per cell). At stage 20, hindgut explants (containing both endoderm and mesoderm) were microdissected in 1× Modified Barth's Saline (MBS) + 50 μg/ml gentamycin sulfate (gent; MP Biochemicals, 1676045) and cultured in 0.5× MBS+0.2% fatty acid-free bovine serum albumin (Fisher Scientific, BP9704)+50 μg/ml gent with the following concentrations of factors: 1 μM DEX (Sigma-Aldrich D4902); 1 μM CHX (Sigma-Aldrich, C4859). In CHX experiments, explants were treated for 2 h with CHX prior to DEX+CHX treatment for 6 h.

The previously validated translation-blocking MO against Xenopus Rfx6 (Pearl et al., 2011) was purchased from GeneTools and injected at the 8-cell stage (1.5 ng per cell) into each dorsal vegetal blastomere to target the foregut. MO sequences were as follows: Rfx6-MO1: 5′-AATTGGCATTTCACCGGGTTCAGGC-3′; a negative control Rfx6 mismatch (MM MO) with five bases altered: 5′ -AATaGGgATTTgACCcGGTTCAcGC-3′ (mismatch bases indicated in lowercase, bold, underlined).

Xenopus explants were dissected from embryos of two or three separate fertilization/injection experiments, homogenized and frozen on dry ice in 350 μl of Buffer RA1 containing BME (Macherey-Nagel Nucleospin RNA isolation kit, 740955), and stored at −80°C. RNA was extracted according to the manufacturer's instructions and 500 ng RNA was used in cDNA synthesis reactions using Superscript Vilo Mastermix (Thermo Fisher Scientific, 11755050), also according to the manufacturer's instructions. qPCR reactions were carried out using PowerUp Mastermix (Thermo Fisher Scientific, A25742) on an ABI QuantStudio3 machine. Xenopus RT-qPCR primer sequences were as follows: pdx1.L forward: 5′-GTTCCCTCAGCTGCTTATCG; pdx1.L reverse: 5′-TACCAAGGGGTTGCTGTAGG; ptf1a.L forward: 5'-ATGGAAACGGTCCTGGAGCA; ptf1a.L reverse: 5'-GAGGATGAGAAGGAGAAGTTG. Relative expression, normalized to the ubiquitously expressed odc1, and then set relative to uninjected, untreated hindgut explants, was determined using the 2^−ΔΔCt method. Graphs display the average 2^−ΔΔCt value±s.d. Statistical significance (P<0.05) was determined using parametric, two-tailed, paired t-tests, relative to uninjected, untreated hindgut explants. Each black dot in the RT-qPCR graphs represents an independent biological replicate containing four explants.

In situ hybridization of Xenopus embryos was performed as described (Sive and Bradley, 1996) with minor modifications. Briefly, embryos were fixed overnight at 4°C in MEMFA (0.1 M MOPS, 2 mM EGTA, 1 mM MgSO4, 3.7% formaldehyde), washed 2×10 min in MEMFA without formaldehyde, dehydrated directly into 100% ethanol, washed four to six times in 100% ethanol, and stored at −20°C in 100% ethanol for at least 24 h. Proteinase K (Thermo Fisher Scientific, AM2548) on day 1 was used at 2 µg/ml for 10 min on stage NF20 embryos and 5 µg/ml for 10 min on NF33 embryos; hybridization buffer included 0.1% SDS; the RNase step was omitted; and anti-digoxygenin (DIG)-alkaline phosphatase antibody (Sigma-Aldrich, 11093274910) used at 1:5000 in MAB buffer (100 mM maleic acid, 150 mM NaCl, pH 7.5)+20% heat-inactivated lamb serum (Gibco, 16070096)+2% blocking reagent (Sigma-Aldrich, 11096176001). Full-length rfx6.L and pdx1.L anti-sense DIG-labeled in situ probes were generated using linearized plasmid cDNA templates with 10× DIG RNA labeling mix (Sigma-Aldrich, 11277073910) according to the manufacturer's instructions. For immunostaining of the NF20 foregut region (Fig. S6), embryos were fixed for 1 h in MEMFA, dehydrated directly into Dent's post-fixative (80% methanol/20% DMSO), and stored at −20°C. Embryos were serially rehydrated through a methanol series (75%, 40%, 25%), bisected, and blocked for 2 h in PBS+10% normal goat serum (Jackson ImmunoResearch, 017-000-121)+1% DMSO+0.5% Triton X-100. Primary antibody incubation in the same blocking solution was carried out overnight (12-16 h) at 4°C using mouse anti-FLAG M2 (Sigma-Aldrich, F3165, 1:1000) and goat anti-FoxA2 (Santa Cruz Biotechnology, sc-6554, 1:500). After five 1 h washes in PBS+0.1% Triton X-100, embryo halves were incubated in secondary antibodies (donkey anti-mouse Cy2, Jackson ImmunoResearch, 715-225-150, 1:1000; donkey anti-goat Cy5, Jackson ImmunoResearch, 705-175-147, 1:1000) diluted in PBS+0.5% Triton X-100 overnight at 4°C. Embryo halves were then washed five times, 30 min per wash, in PBS+0.1% Triton X-100, dehydrated in 100% methanol, cleared in Murray's clear (80% benzyl benzoate, 20% benzyl alcohol), and imaged with a Nikon A1R confocal microscope.

A 1236 bp sequence of the human PDX1 upstream ‘area I II III’ enhancer (>hg19_dna range=chr13:28491064-28492299) was screened for RFX6 responsiveness because the orthologous mouse Pdx1 enhancer region has previously been shown to drive reporter expression in foregut and duodenum (Gannon et al., 2001; Fujitani et al., 2006b). WT or a mutant version of the human PDX1 enhancer with six CisBP-predicted RFX-binding motifs mutated (Weirauch et al., 2014) were commercially synthesized by GenScript and cloned into the pGL4.23 firefly luc2/miniP vector (Promega, E8411). Embryos (16- to 32-cell stage) were co-injected with 5 pg of pRL-TK:Renilla luciferase plasmid (Promega, E2241)+80 pg of the pGL4.23 luc2/miniP Human PDX1 enhancer:firefly luciferase plasmid (WT or 6RFX motif mutant) into either a D1 blastomere to target foregut or a D4 blastomere to target hindgut. The following amounts of MOs or mRNAs were also injected in enhancer:luciferase assays: 3 ng total of Rfx6-MO1 or 5 bp mismatch-MO; 100 pg WT or mutant human RFX6 mRNAs.

Each biological replicate contained a pool of five embryos, obtained from two or three separate fertilization/injection experiments, which were frozen on dry ice in a minimal volume of 0.1× MBS and stored at −80°C. To assay luciferase activity, samples were lysed in 100 µl of 100 mM TRIS-Cl pH 7.5, centrifuged for 10 min at ∼13,000 g and then 50 µl of the clear supernatant lysate was used separately in firefly (Biotium, 30085-1) and Renilla (Biotium 300821) luciferase assays according to the manufacturer's instructions. Relative luciferase activity was determined by normalizing firefly to Renilla levels for each sample. Graphs show the average relative luciferase activity ±s.d. with black dots showing values of biological replicates. Statistical significance was determined by parametric, two-tailed, paired t-test (*P<0.05).

Data are presented as the mean±s.e.m. Significance was determined using appropriate tests in GraphPad Prism, with P>0.05 being considered not significant.

We are thankful for the support provided by the confocal imaging core, gene expression core, pathology core and pluripotent stem cell facility at CCHMC. CCHMC is supported by a Digestive Diseases Research grant (P30 DK078392). We thank all the members of the Wells and Zorn labs for reagents and feedback. All figures were created using Adobe Illustrator, BioRender.com and GraphPad Prism.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.202529.reviewer-comments.pdf