Renal functional units known as nephrons undergo patterning events during development that create a segmental array of cellular compartments with discrete physiological identities. Here, from a forward genetic screen using zebrafish, we report the discovery that transcription factor AP-2 alpha (tfap2a) coordinates a gene regulatory network that activates the terminal differentiation program of distal segments in the pronephros. We found that tfap2a acts downstream of Iroquois homeobox 3b (irx3b), a distal lineage transcription factor, to operate a circuit consisting of tfap2b, irx1a and genes encoding solute transporters that dictate the specialized metabolic functions of distal nephron segments. Interestingly, this regulatory node is distinct from other checkpoints of differentiation, such as polarity establishment and ciliogenesis. Thus, our studies reveal insights into the genetic control of differentiation, where tfap2a is essential for regulating a suite of segment transporter traits at the final tier of zebrafish pronephros ontogeny. These findings have relevance for understanding renal birth defects, as well as efforts to recapitulate nephrogenesis in vivo to facilitate drug discovery and regenerative therapies.
Vertebrate kidney ontogeny involves the reiterative formation and degradation of up to three structures from the intermediate mesoderm: the pronephros, mesonephros and metanephros (Saxen, 1987). In amniotes, the metanephros serves as the final kidney, whereas lower vertebrates such as fish and frogs use the mesonephros. Each kidney version contains functional units called nephrons (Dressler, 2006). Nephrons comprise a blood filter, tubule and collecting duct, which modify the filtrate in a stepwise fashion to perform excretion and osmoregulation. Occurring in ∼1 in 500 births, congenital anomalies of the kidney and urinary tract (CAKUT) are a common birth defect that causes pediatric end-stage renal disease (ESRD) (Airik and Kispert, 2007; Song and Yosypiv, 2011). One etiology across these diverse conditions is aberrant nephrogenesis stemming from genetic dysregulation (Schedl, 2007). To this end, it is imperative to understand the signals that coordinate nephron formation.
The zebrafish pronephros has emerged as a genetically tractable vertebrate for studying the mechanisms of nephron development (Naylor et al., 2017). The embryonic zebrafish is transparent in nature, and the pronephros is structurally simple, consisting of two bilateral nephrons with conserved proximal and distal segments (Wingert et al., 2007; Wingert and Davidson, 2008). Furthermore, zebrafish nephrons undergo fundamental processes such as the mesenchymal-to-epithelial transition (MET) of renal progenitors, establishment of apical-basal polarity, lumen formation and ciliogenesis (Gerlach and Wingert, 2013). Although there has been recent progress in understanding nephron segmentation (Desgrange and Cereghini, 2015; Lindström et al., 2015; Chung et al., 2017), the pathways that dictate segmental terminal differentiation are far from understood.
Transcription factors play a central role in operating the genetic networks that orchestrate nephron formation (Desgrange and Cereghini, 2015). Advances in single cell RNA sequencing and gene expression analysis in embryonic murine and human kidneys have identified genes that mark distinct regions of developing nephrons (Lindström et al., 2018a,b). One such gene is transcription factor AP-2 alpha (TFAP2A), which clusters with developing medial/distal nephron tubule signatures (Lindström et al., 2018b). TFAP2A is a member of the AP-2 transcription factor family (consisting of AP-2α, AP-2β, AP-2γ, AP-2δ and AP-2ε), which share conserved dimerization and DNA-binding motifs (Fig. S1). AP-2 factors bind to GC-rich promoter sequences, and can homodimerize and heterodimerize with one another (Eckert et al., 2005). During ontogeny, these factors exercise redundant and unique functions depending on the tissue context (Eckert et al., 2005).
Tfap2a and family member Tfap2b have overlapping expression patterns in neural crest derivatives, surface ectoderm and the kidney (Moser et al., 1997a; Knight et al., 2003, 2005). During mammalian nephrogenesis, Tfap2a is broadly expressed in the developing tubule, where Tfap2b is more restricted to distal regions of the renal vesicle, comma and S-shaped body (McMahon et al., 2008; Harding et al., 2011). Surprisingly, the elimination of Tfap2a and Tfap2b in mice results in very different phenotypic outcomes. Tfap2a knockout mice die perinatally with severe pleiotropic features that include craniofacial alterations, incomplete neural tube closure, and hypoplastic hearts and kidneys (Schorle et al., 1996; Zhang et al., 1996; Brewer and Williams, 2004; Brewer et al., 2004). In contrast, Tfap2b null mice exhibit patent ductus arteriosus and die shortly after birth with acute renal failure involving elevated apoptosis (Moser et al., 1997b; Hilger-Eversheim et al., 2000; Wang et al., 2018). Because Tfap2b knockout causes less severe phenotypes, the gene has been proposed to share redundant functions with Tfap2a (Eckert et al., 2005; Kerber et al., 2001). An example in support of this relationship is that Tfap2a plays a more dominant role than Tfap2b during branchial arch development in mice (Van Otterloo et al., 2018).
Genetic defects in the AP-2 factors are associated with several human diseases. Autosomal dominant TFAP2A mutations cause branchio-oculo-facial syndrome (BOFS), which primarily affects craniofacial tissue (Milunsky et al., 2008). Additionally, human TFAP2A lesions are associated with multicystic dysplastic kidney, but the mechanisms are unexplored. Dominant-negative mutations in human TFAP2B cause Char syndrome, which affects heart, face and limb development (Satoda et al., 2000). Despite the renal phenotypes associated with Tfap2a and Tfap2b deficiency in rodents, these genes have not been examined further in the context of kidney development. Nevertheless, Tfap2a/tfap2a has been extensively studied in the vertebrate neural crest, where it facilitates specification and differentiation through a complex genetic regulatory network (Knight et al., 2003, 2005; Holzschuh et al., 2003; Barrallo-Gimeno et al., 2003; O'Brien et al., 2004; Li and Cornell, 2007; Hoffman et al., 2007; Kwon et al., 2010; Van Otterloo et al., 2010; de Crozé et al., 2011; Wang et al., 2011; Bhat et al., 2012; Green et al., 2014; Kantarci et al., 2015; Seberg et al., 2017). These studies provide a valuable framework with which to consider the roles of Tfap2a/tfap2a in other tissues.
Here, we report the novel zebrafish nephron segment mutant, terminus (trm), which was isolated in a forward haploid genetic screen. By employing whole-genome sequencing, we identified a mutation that blocks normal splicing of tfap2a. Although tfap2a deficient nephrons have normal distal segment pattern formation, epithelial polarity and ciliogenesis, they experience a block in other aspects of terminal differentiation, resulting in the loss of solute transporter expression within the distal early (DE) and distal late (DL) segments, which are analogous to the mammalian thick ascending limb (TAL) and distal convoluted tubule (DCT) (Wingert et al., 2007). We found that tfap2b functions redundantly and downstream of tfap2a to turn on the distal nephron solute transporter program. Furthermore, tfap2a articulates with Iroquois homeobox transcription factors irx1a and irx3b, which regulate distal nephron identity. Our study reveals that tfap2a controls a gene regulatory network to serve as a gatekeeper of terminal differentiation during distal nephron segment development, and thus establishes a new paradigm for nephrogenesis mechanisms during kidney formation.
A forward genetic screen identifies tfap2a as a novel regulator of nephron development
There remain many gaps in our knowledge of the genetic blueprint used to orchestrate nephron formation. The embryonic zebrafish pronephros is a tractable genetic model for nephron segmentation research (Gerlach and Wingert, 2013). At 24 h post-fertilization (hpf), the pronephros is fully formed and exhibits a very simple organization consisting of two parallel nephrons (Fig. 1A), making cellular changes easy to detect (Poureetezadi and Wingert, 2016). Each nephron has a series of proximal and distal segments that reabsorb and secrete molecules, and a collecting duct to transport waste (Fig. 1A) (Wingert et al., 2007). To identify renal regulators, we performed a forward genetic screen. We obtained maternal gametes from the F1 generation, applied ultraviolet light-inactivated sperm to generate F2 haploid embryos and assayed for nephron segment defects by whole-mount in situ hybridization (Kroeger et al., 2014). We applied a mixture of riboprobes to identify alternating compartments of the pronephros: podocytes (P) (wt1b), proximal convoluted tubule (PCT) (slc20a1a) and the DE (slc12a1) (Fig. 1B). Through this multiplex assay, we isolated the nephron mutant terminus (trm), in which DE segment development was affected based on abrogated slc12a1 expression within the pronephros (Fig. 1B).
We used whole-genome sequencing to clone the trm mutation (Leshchiner et al., 2012; Ryan et al., 2013). Using SNPtrack software, the lesion was mapped to chromosome 24, where tfap2a was identified as a candidate due to a G>A substitution that was predicted to disrupt splicing at the splice donor of exon 1c (Fig. 1C). PCR sequencing of trm mutants and wild-type siblings confirmed this mutation (Fig. 1C). Transcript analysis revealed four aberrant tfap2a spliceoforms in trm compared with wild-type (Fig. 1D) embryos. One aberrant transcript encoded an in-frame addition of 38 amino acids (Fig. 1D), which may possess native function or have dysfunctions associated with protein folding or stability. The other three transcripts encoded premature stop codons (Fig. 1D). These aberrant trm transcripts are predicted to truncate the essential transcriptional activation and DNA-binding domains in Tfap2a (Fig. 1D).
Next, we explored whether the loss of tfap2a function in trm mutants was the sole origin of their renal phenotype. We performed a complementation test using tfap2am819, which encodes a nonsense allele, followed by phenotype assessment with whole-mount in situ hybridization and genotyping. Compound trm+/−;tfap2am819+/− heterozygote embryos displayed a failure to complement due to the loss of kcnj1a.1 expression in the DE segment; furthermore, they exhibited reduced dlx2 expression in the pharyngeal arches, consistent with neural crest alterations, but retained kcnj1a.1 ionocyte expression (Fig. 1E). As tfap2a is essential for neural crest and epidermis differentiation, we assessed whether trm evinced defects in these tissues. Live morphology analysis at 4 days post fertilization (dpf) revealed that trm formed abnormal craniofacial cartilage and pericardial edema similar to tfap2a morpholino (MO) knockdown (Fig. 1F). RT-PCR analysis confirmed that this tfap2a MO effectively disrupts splicing (Fig. S2). Alcian Blue staining revealed that trm possessed defects in Meckel's cartilage and pharyngeal arches (Fig. 1G), consistent with the tfap2a mutant alleles lockjaw and mont blanc (Knight et al., 2003; Barrallo-Gimeno et al., 2003). trm also displayed altered craniofacial vasculature (Fig. S3). Interestingly, whole-mount immunofluorescence showed that Tfap2a had a nuclear expression pattern in the DE, where Slc12a1 localized to the apical margin of DE tubule cells within wild-type animals at 24 hpf (Fig. 1H). There was little to no pronephric expression of Tfap2a or Slc12a1 in trm (Fig. 1H), consistent with the conclusion that trm encodes a loss-of-function allele. In sum, these results show that trm mutants exhibit many features of tfap2a deficiency, and reveal for the first time that tfap2a is needed for nephrogenesis – specifically for proper emergence of the DE segment.
tfap2a and tfap2b are co-expressed dynamically during pronephros development and function redundantly to induce regimens of distal segment solute transporter genes
Of the AP-2 family members, only tfap2a and tfap2b are expressed in the pronephros (Knight et al., 2005; Sugano et al., 2017). Zebrafish tfap2a and tfap2b genes are closely related, and share overall 65% amino acid sequence identity with similar DNA-binding and transactivation domains (Knight et al., 2005). Zebrafish tfap2a and tfap2b are also conserved with other vertebrates (Fig. S1) (Knight et al., 2005). We performed whole-mount in situ hybridization over the time span of nephrogenesis to further investigate the spatiotemporal expression of tfap2a and tfap2b. tfap2a and tfap2b were expressed broadly in renal progenitors, which are mesenchymal in nature, at the 10-somite stage (ss) (Fig. 2A). tfap2a expression was restricted to the distal pronephros by the 28 ss, when the maturing precursors complete MET, tubulogenesis and segmentation (Fig. 2A). Through fluorescent in situ hybridization, we confirmed that tfap2a and tfap2b were co-expressed in the renal progenitor domains at the 10 ss (Fig. 2B). Fluorescent in situ hybridization at the 20 ss and 28 ss revealed that tfap2a had a slightly broader expression pattern across the distal pronephros compared with tfap2b (Fig. 2B), aligning with the DE and DL segment domains. Differential tfap2a/b expression was noted rostrally and in the posterior pronephric duct region at the 20 ss and 28 ss, where only tfap2a transcripts were detected (Fig. 2B). Using fluorescent in situ hybridization, we confirmed that tfap2a and tfap2b are expressed within the pax2a+ renal progenitor stripes at the 10 ss (Fig. 2C). Furthermore, we found that tfap2a and slc12a1 transcripts colocalized in the DE at the 28 ss (Fig. 2D). These results indicate that tfap2a and tfap2b expression in renal progenitors, and subsequently in differentiating distal nephron segments, occurs during a time window that positions them as possible participants in nephrogenesis.
To explore potential genetic relationships between tfap2a and tfap2b, we performed loss-of-function experiments. trm exhibited significantly reduced tfap2b expression in the distal pronephros and the hindbrain region (Fig. 2E,F). Analysis of tfap2b function was performed with MO and mutant analysis. tfap2bsa10090−/− encodes a G>A substitution, which disrupts an essential splice site between exons 3 and 4 (Fig. S3). A tfap2b MO was designed and verified to interrupt splicing, where it caused inclusion of intronic sequence that encoded a premature stop codon and truncated peptide (Fig. S4). In both tfap2bsa10090−/− mutant embryos (Fig. 2G,H) and tfap2b MO knockdowns (Fig. S4), tfap2a expression was unaffected throughout the embryo, including the pronephros. In combination, these genetic studies suggested that tfap2a may act upstream of tfap2b during nephrogenesis.
Because tfap2a and tfap2b function redundantly in the development of other tissues, we next determined whether these two factors acted similarly in the pronephros (Knight et al., 2005; Van Otterloo et al., 2018; Seberg et al., 2017; Bassett et al., 2012; Jin et al., 2015). To interrogate this, we performed combination knockdown and mutant studies, and assayed expression of the solute transporters that characterize the distal nephron segments at 24 hpf. tfap2b deficiency alone had no detectable effects on distal solute transporter expression (Fig. 3). trm and tfap2a morphants exhibited significant reductions in slc12a1, slc12a3 and clcnk expression compared with wild-type embryos (Fig. 3). At 4 dpf, trm mutants still failed to express the DE solute transporter transcripts slc12a1 and kcnj1a.1 (Fig. S3). Additionally, trm mutant tubules demonstrated abrogated apical expression of Slc12a1 protein at 4 dpf (Fig. S3). However, the pronephros was functional between 2 and 3 dpf based on active renal clearance, which normally initiates during this time, thereby ruling out developmental delay (Fig. S3). Knockdown of tfap2b in trm mutants and knockdown of tfap2a in tfap2bsa10090−/− mutants caused a more severe slc12a3 reduction than tfap2a deficiency alone (Fig. 3). Notably, 5/9 tfap2bsa10090−/− mutants injected with tfap2a MO exhibited collecting duct cysts and/or failure of the cloaca to fuse (Fig. 3A, Fig. S4). These results suggest that the tfap2bsa10090 allele induces additional distinct phenotypic outcomes than the tfap2b MO. Interestingly, these phenotypes emulate the distal tubule/collecting duct cysts in Tfap2b knockout mice, although, in the zebrafish pronephros, cysts were observed only in the absence of both Tfap2a and Tfap2b (Moser et al., 1997b; Hilger-Eversheim et al., 2000; Wang et al., 2018). There was not a statistically significant reduction in the slc12a1 or clcnk domain length when tfap2b knockdown was performed in trm mutants versus tfap2a deficiency alone (Fig. 3). By comparison, tfap2bsa10090−/− mutants injected with tfap2a MO and tfap2a/2b morphants had statistically significant reductions of the slc12a1, slc12a3 and clcnk pronephros expression domains versus tfap2a deficiency alone. Although we have not observed appreciable Tfap2a protein by immunofluorescence in trm, this approach is not quantitative and it is reasonable to expect that a low level of Tfap2a could be produced owing to the existence of the aforementioned isoform. Thus, we hypothesize that lower Tfap2a expression is the reason why double tfap2a/tfap2b morphants have a slightly stronger phenotype compared with knockdown of tfap2b in trm. In light of this phenotypic spectrum, we concluded that the development of the distal nephron program is sensitive to the dose of functional tfap2 alleles. Taken together, these genetic studies reveal that the concerted activity of tfap2a and tfap2b is necessary to fully turn on distal solute transporter programs, where tfap2a plays a more prominent role in this process upstream of tfap2b.
tfap2a is necessary and sufficient for DE differentiated cell expression signature
Next, we wanted to determine whether provision of wild-type tfap2a transcripts could rescue DE development in trm. Activation of a heat shock-inducible tfap2a transgene at the 8 ss restored kcnj1a.1 expression in trm mutants comparable to wild-type levels based on absolute length measurements (Fig. 4A,B). This result further underscores the conclusion that tfap2a deficiency is the single, specific cause of the trm phenotype. We then employed tfap2a gain-of-function studies using an inducible hs:tfap2a transgenic line and microinjection of tfap2a mRNA in wild-type embryos. tfap2a overexpression caused a significant expansion of the kcnj1a.1 pronephros expression domain, which normally marks the DE (Fig. 4A,B). Over 70% of heat shock-treated hs:tfap2a embryos exhibited an expanded kcnj1a.1 domain, whereas non-heat-shocked embryos had normal domains (Fig. S5). In heat-shocked hs:tfap2a, the increased kcnj1a.1 expression localized to cdh17+ kidney cells at 24 hpf (Fig. S5). About 9% (12/132) of tfap2a cRNA-injected embryos had an increased kcnj1a.1 pronephros domain; however, 64% (85/132) were dysmorphic. Compared with transgenic overexpression, this lower penetrance is likely due to effects of tfap2a that alter gastrulation (Li and Cornell, 2007). Interestingly, ectopic kcnj1a.1+ cells appeared to form in both proximal and distal segment regions in these gain-of-function experiments (Fig. 4, Fig. S5).
To examine more closely whether tfap2a overexpression induced other nephron segments to exhibit features of a DE program, we performed double fluorescent in situ hybridization on hs:tfap2a embryos heat-shocked at the 10 ss. We detected slc12a1+ cells within the slc12a3+ DL that co-expressed both markers (Fig. 4C,C′). Generation of a fluorescent intensity plot of slc12a1 signal in a heat-shocked hs:tfap2a animal revealed ectopic slc12a1 peaks above baseline levels in the DL, which were absent in the wild-type sample (Fig. 4D). Heat-shocked hs:tfap2a animals also had slc12a1+ cells in the proximal domain that co-expressed slc9a3 (Fig. 4E,E′). Similarly, the slc12a1 fluorescent plot had multiple slc12a1 signal spikes above background levels in the proximal nephron domain, which were absent in the wild-type control (Fig. 4F). These phenotypes induced by tfap2a overexpression greatly contrast the wild-type situation, where there are sharp, clear boundaries between neighboring segment domains (Fig. 4C,E). Additionally, tfap2a overexpression at the 8 ss caused slc12a3+ cells to invade the slc9a3+ domain (Fig. S5). In summary, these results indicate that tfap2a overexpression is sufficient to sway the differentiation profile of proximal and distal late cell types by triggering the misexpression of DE and DL solute transporter genes.
tfap2a drives DE terminal differentiation program
Previous studies have demonstrated tfap2a can regulate terminal differentiation of various cell types, including neural crest, melanocytes, statoacoustic ganglion neurons, noradrenergic neurons and the trophoblast lineage (Barrallo-Gimeno et al., 2003; Kantarci et al., 2015; Seberg et al., 2017; Greco et al., 1995; Kim et al., 2001; Pfisterer et al., 2001; Handwerger, 2009). This literature, in light of our loss- and gain-of-function results, led us to hypothesize that tfap2a controls the terminal differentiation of distal nephron cells. To explore this notion, we first wanted to determine whether the DE and other nephron segment domains were patterned correctly in trm mutants.
To assess nephron segment pattern formation in trm mutants, we performed double whole-mount in situ hybridization to assess the segment domains located adjacent to the DE: in this case the pan-proximal (slc9a3+) and DL (slc12a3+). trm mutants exhibited a domain of slc9a3 expression comparable with wild-type embryos (Fig. 5A). In both wild-type and trm embryos, this pan-proximal region was followed by a gap situated at the position normally occupied by the DE and the DL, which is smaller in trm mutants (Fig. 5A). The intact sequence of the pan-proximal, gap/placeholder and then the DL segment suggested that the DE segment ‘footprint’ was present in trm mutants, and thereby consistent with normal pattern formation (Fig. 5A). Additionally, trm mutants exhibited normal slc20a1a and trpm7 expression domains, which mark the PCT and PST segments, respectively (Fig. S6). tfap2a morphants also had normal proximal segments, as well as the DE footprint (data not shown). These results indicate that tfap2a-deficient embryos undergo normal segmental patterning of the nephron tubule.
We then examined development of the corpuscle of Stannius (CS) in trm, which is an endocrine gland situated between the DE and DL (Cheng and Wingert, 2015). We used whole-mount in situ hybridization to assess expression of stanniocalcin 1 (stc1) transcripts, a specific CS marker. Compared with wild-type embryos, trm mutants exhibited severely reduced stc1 expression (Fig. S6). Recent studies have suggested that the CS transdifferentiates from distal tubule cells (Naylor et al., 2018), which is consistent with our finding that trm mutants exhibit perturbations in both distal tubule and CS development. In summary, these data rule out the occurrence of possible fate switches with adjacent nephron cell types as the reason for abrogated DE marker expression.
Next, we determined whether trm mutant cells occupying the DE region were specified as kidney. To do this, we individually assayed for genes that are expressed robustly throughout the entire nephron tubule. Identical to wild-type embryos, trm mutants showed no gaps in expression of cdh17 and hnf1ba, illustrating that mutant DE cells were fated to a kidney lineage identity (Fig. 5B). Furthermore, we assessed whether alterations in cell proliferation or cell death occurred from loss of tfap2a. There was no perceivable difference in cell death based on Acridine Orange staining compared with wild-type embryos at 24 hpf (Fig. 5D). pH3+ cell numbers were also similar in the DE of trm mutants and wild-type embryos at 24 hpf, suggesting a normal proliferation rate (Fig. 5D,E). Finally, we determined whether undifferentiated mutant DE cells undergo apoptosis later, similar to undifferentiated tfap2a-deficient neural crest (Barrallo-Gimeno et al., 2003). There were no visible caspase 3+ cells in the trm pronephros at 4 dpf, but elevated cell death was observed in the notochord compared with wild-type embryos (Fig. 5F).
Renal progenitor differentiation in the zebrafish pronephros entails an MET of these mesenchymal cells, along with establishment of polarity, lumen formation and the formation of cilia. Therefore, we sought to determine whether trm mutant DE cells exhibited any differentiated features of the nephron tubular epithelium. Differentiated pronephros cells exhibit apical-basal polarity and form either a single cilium or multiple cilia by 24 hpf (Gerlach and Wingert, 2013, 2014; McKee et al., 2014; Marra and Wingert, 2016; Marra et al., 2016, 2019a). Using immmunofluorescence to analyze the expression of basolateral marker Na,K-ATPase and the apical adaptor complex aPKC in the DE at 4 dpf, we found both proteins were properly localized in trm when compared with wild-type embryos, indicating that epithelial polarity was correctly established within the nephrons (Fig. 5G,G′). Furthermore, trm mutants had a clearly discernible nephron lumen, suggesting normal tubulogenesis (Fig. 5G,G′). To scrutinize potential alterations in DE cell morphology, we measured nuclear area (labeled by DAPI) and cell surface area (cell membranes marked by Na,K-ATPase) in 4 dpf animals. We found that trm−/− had smaller nuclei and overall cell size than wild-type controls; these mutant cell attributes potentially indicate an altered terminal differentiation state (Fig. 5H,I).
Next, we combined fluorescent in situ hybridization of slc12a1 with whole-mount immunofluorescence of acetylated α-tubulin to determine whether cilia formation occurs within the mutant DE segment region (Marra et al., 2017). At 24 hpf, cilia arrangement and morphology in trm was comparable with wild-type embryos (Fig. 5J). This indicates that cilia assembly occurred normally in the mutant DE cells, which were visualized based on their nearly abrogated slc12a1 signal (Fig. 5J). We confirmed slc12a1 mRNA signal was significantly decreased in trm−/− mutants by comparing the fluorescent intensity values in the DE segment with wild-type controls (Fig. 5K). In corroboration with these results, comparison of fluorescent intensity plot profiles from trm−/− and wild-type embryos revealed that no individual mutant slc12a1 signal extended above the mean wild-type threshold value (Fig. 5L). Taken together, these data indicate that mutant DE cells exhibit mature epithelial qualities; however, they do not express specific solute transporters, which are indicators of terminal differentiation and ultimately dictate segment-specific physiological functions. From this, we conclude that trm mutants exhibit a unique block in the terminal differentiation of distal nephron cells, which involves the acquisition of segment-specific solute transporter proteins; this is a process distinct from MET, polarity establishment, tubulogenesis or ciliogenesis programs.
tfap2a functions downstream of irx3b and upstream of irx1a in the distal pronephros
We next wanted to understand the genetic relationship of tfap2a with known segment patterning factors. Previous studies have shown that Irx3/irx3b are required for the development of slc12a1+ distal tubule cells in Xenopus and zebrafish, respectively (Reggiani et al., 2007; Wingert and Davidson, 2011; Marra and Wingert, 2014). Because of this requirement, we selected irx3b as a candidate for investigation. To determine whether tfap2a and irx3b are co-expressed during pronephros development, we performed fluorescent in situ hybridization. tfap2a and irx3b transcripts colocalized in developing distal nephron cells at the 20 ss (Fig. 6A). Because of the tfap2a/irx3b overlapping expression patterns, we rationalized that these factors could be interacting in the same developmental pathway and next performed knockdown experiments to explore this. Upon tfap2a knockdown, the irx3b expression domain was unchanged (Fig. 6B). However, in irx3b crispants and morphants, the tfap2a expression domain was significantly truncated (Fig. 6B,C). The regional loss of tfap2a transcripts in irx3b-deficient animals equates to the DE segment address. We also observed disrupted tfap2a expression in migrating neural crest streams in irx3b knockdown animals (Fig. 6B). To further validate whether tfap2a acts downstream of irx3b, we performed rescue experiments in irx3b morphants. Overexpression of the hs:tfap2a transgene was unable to rescue kcnj1a.1 expression in irx3b knockdowns (data not shown). We postulate that this is because irx3b deficiency causes loss of hnf1ba expression within the DE progenitor compartment, and that the cells are not competent to respond to Tfap2a (Naylor et al., 2013). These results suggest that tfap2a likely activates the DE program downstream of irx3b.
Previous studies have shown that Irx1 and Irx3 are dually required for Xenopus pronephros development (Reggiani et al., 2007; Alarcón et al., 2008). Importantly, loss of Irx3 in Xenopus results in abrogation of Irx1 expression (Reggiani et al., 2007). Furthermore, irx1a is expressed in the putative DE (Cheng et al., 2001). Therefore, we chose Iroquois homeobox family member, irx1a, as the next gene for analysis. In wild-type embryos, irx1a transcripts were colocalized with the DE marker slc12a1 (Fig. 6D,D′). In trm mutants and tfap2a morphants, irx1a expression was nearly abrogated, with only a few remaining nephron cells expressing transcripts (Fig. 6E,G). When we induced overexpression of tfap2a at the 8 ss, the irx1a expression domain length was significantly expanded, indicating that tfap2a functions to activate irx1a expression directly or indirectly (Fig. 6F,G). To determine whether irx1a is sufficient to initiate DE solute transporter genes, we performed overexpression experiments by injecting irx1a cRNA into wild-type embryos. irx1a overexpression caused a drastic expansion of both the kcnj1a.1 and slc12a1 expression domains (Fig. 6H-J) (Fig. S7). We also discovered that irx1a overexpression caused formation of ectopic eyes in a subset of animals (Fig. S8). This finding is consistent with previous reports on the requirement of irx1a during zebrafish eye development (Cheng et al., 2006; Choy et al., 2010). Interestingly, upon irx1a cRNA injection into trm−/−, we were able to partially rescue kcnj1a.1 expression length and area (Fig. 6H-J). These results indicate that irx1a functions downstream of tfap2a, where it is sufficient to promote expression of kcnj1a.1. Given this partial rescue, however, other genes may operate under the control of tfap2a to fully turn on the DE terminal differentiation program. Taken together, our genetic analyses suggest a model in which tfap2a coordinates a genetic regulatory network, likely through direct and indirect interactions, that controls the terminal differentiation of distal nephron segments (Fig. 7).
Here, we have shown that tfap2a is required for distal nephron segment differentiation. We propose a model in which trm mutant cells progress through normal nephron developmental checkpoints until the final stage of differentiation (Fig. 7A). Our data support the conclusion that tfap2a deficiency does not affect derivation of renal progenitors from intermediate mesoderm, as hnf1ba and cdh17 expression was unaffected. Furthermore, trm mutants undergo nephron specification and epithelialization, with normal segmental patterning, proper localization of polarity proteins and ciliogenesis. However, mutant cells appear poised in a specified state, as they fail to turn on a suite of distal solute transporters. This novel discovery disentangles the control of the solute transporter transcriptome from other differentiation processes, such as MET and polarity establishment in renal progenitors. Thus, we propose that a Tfap2 genetic circuit controls differentiation of distal nephron epithelium within the zebrafish pronephros (Fig. 7B). In this genetic network, we propose that tfap2a functions upstream of tfap2b. However, our data lead us to conclude that tfap2a and tfap2b function synergistically in developing nephrons to turn on distal solute transporter gene targets, a level of redundancy that likely serves to amplify and reinforce this specific differentiation signal. This model is supported by the findings that elimination of tfap2a led to defects in solute transporter expression, where elimination of tfap2b alone had no consequence. Our study does not reconcile whether tfap2a and tfap2b interact directly, or whether the suite of targets are direct. For example, it is possible tfap2a binds to the tfap2b promoter region functioning as a transcriptional activator, or that Tfap2a heterodimerizes with Tfap2b to affect transcription of targets. These potential biochemical mechanisms are crucial areas for future investigation. The observation of collecting duct cysts after knockdown of tfap2a in tfap2b genetic mutants suggests, however, that these factors also have roles in tubulogenesis or maintenance that warrant further investigation. This finding is also intriguing, as it suggests there may be conserved roles of tfap2b because Tfap2b mouse knockouts display kidney cysts (Moser et al., 1997b). Furthermore, our work thus far supports the conclusion that tfap2a acts in the same pathway as Iroquois homeobox genes irx3b and irx1a. Our genetic experiments indicate that irx3b promotes tfap2a expression, and tfap2a functions upstream of irx1a (Fig. 7B). Several Irx genes have been previously implicated as necessary for nephron segmentation in zebrafish and frogs (Reggiani et al., 2007; Wingert and Davidson, 2011; Marra et al., 2019b). Importantly, Iroquois factors likely play conserved roles in the mammalian nephron, as Irx3 and Irx1 define intermediate segment territories in developing S-shaped bodies and have been linked with intermediate tubule defects (Reggiani et al., 2007; Heliot et al., 2013; Massa et al., 2013). Defining whether Irx3b directly regulates tfap2a, and whether Tfap2a directly regulates irx1a, will also be important to discern in future studies.
Additionally, we discovered an intriguing phenotype when we globally overexpressed tfap2a, where slc9a3+ proximal tubule and slc12a3+ distal tubule cells ectopically co-express slc12a1, a marker of the DE (Fig. 5). This indicates that renal progenitors are competent to respond to Tfap2a, which is sufficient to activate the DE differentiation circuit. These mixed segment identities induced by tfap2a gain of function parallel a phenotype recently described as lineage infidelity, which was observed in differentiating nephrons of Hox9/Hox10/Hox11-knockout mice, where individual cells dually express markers of more than one nephron segment (Magella et al., 2018; Drake et al., 2018).
With the advent of next-generation sequencing technologies, recent studies have identified new targets within the Tfap2 genetic regulatory network (Seberg et al., 2017; Van Otterloo et al., 2018). Similar analyses of trm kidneys and comparison with the mammalian kidney Tfap2a CHIP-seq data set (Pihlajamaa et al., 2014) will help to identify putative targets governing terminal differentiation of distal nephron cells. Interestingly, the majority of differentially expressed genes in Tfap2a/2b murine branchial arches were associated with regions corresponding to poised histone marks (Van Otterloo et al., 2018). Thus, Tfap2a may regulate transcription during nephrogenesis via an intermediate factor or chromatin modifier as part of the gene regulatory network. In support of indirect regulation, Tfap2a acts as a tissue-specific pioneer factor in the epididymis to modify chromatin structure and activate androgen receptor signaling (Pihlajamaa et al., 2014). We speculate that Tfap2a and Tfap2b operate both directly and indirectly to regulate expression of GRN components responsible for terminal differentiation of distal nephron tubules. To explore this prospect, ATAC-seq and ChIP-seq could be used to determine chromatin accessibility in the tfap2a-deficient pronephros.
Tfap2a has roles in cell specification, patterning, survival and differentiation, depending on the tissue type (Eckert et al., 2005). Interestingly, in the developing mouse kidney, Tfap2b is required for the maintenance and survival of renal epithelium (Moser et al., 1997b). In Tfap2b-null mice, distal tubules and collecting duct cells undergo a massive wave of apoptosis. Histological sectioning of mutant kidneys revealed numerous cysts in the distal tubules and collecting ducts. Ultrastructural analysis showed epithelial cells lining the cysts had a flat dedifferentiated phenotype (Moser et al., 2003). Furthermore, Tfap2b-deficient mice manifest defective ion homeostasis and inability to concentrate urine, suggesting defective distal solute transporters. Tfap2 factors may exhibit conserved functions in activating solute transporter programs during the differentiation process of distal tubules in both zebrafish pronephric development and mammalian metanephric development. The cysts and elevated apoptosis present in distal tubules and collecting ducts of Tfap2b−/− mice may arise as a secondary consequence of undifferentiated nonfunctional epithelium. In corroboration with these murine studies, we found that tfap2a knockdown in tfap2b−/− mutants initiated cyst formation in the distal-most region of the pronephric duct. Interestingly, Tfap2a loss in mice has more catastrophic consequences than Tfap2b loss, as the animals die perinatally and exhibit severely hypoplastic kidneys. Similar observations hold true in the zebrafish, as we found that tfap2a mutants exhibit more severe phenotypes than tfap2b mutants, suggesting that Tfap2a is the more dominant factor. However, it remains a possibility that the genetic requirements differ for zebrafish pronephric kidney development and the mammalian metanephros, as the latter involves creation of convoluted nephrons within an intricate tissue architecture resulting in the final kidney form.
Based on recent studies in the zebrafish model, upstream candidates for regulating tfap2a/2b may occupy the prostaglandin signaling pathway, which controls the balance of DE and DL territories during zebrafish nephrogenesis (Poureetezadi et al., 2016), or be transcription factors such as mecom, ppargc1a, tbx2a/2b or emx1, which regulate DL development (Li et al., 2014; Drummond et al., 2016; Chambers et al., 2018; Morales et al., 2018). Additional network candidates that may crosstalk with tfap2a/2b include sall1 and sox11. In mammals, Sall1 has been found to be a crucial factor in TAL segment development, which is analogous to the zebrafish pronephros DE (Basta et al., 2017). In the murine kidney, Sox11 is also necessary for loop of Henle ontogeny, where Sox11-deficient kidneys have significantly reduced expression of Slc12a1, Irx1 and Irx2 (Neirijnck et al., 2018).
Knowledge about the terminal differentiation programs of each nephron segment has central importance for understanding kidney disease and to advance regenerative medicine. Human BOFS is associated with the occurrence of dysplastic kidneys, but the underlying mechanisms are not known. Our zebrafish trm mutant provides an opportunity to model aspects of BOFS at the molecular level of the nephron. With regard to kidney engineering, current groups face major challenges in generating differentiated nephron structures in kidney organoid cultures (Hariharan et al., 2015; Chambers et al., 2016; Oxburgh et al., 2017; Takasato and Little, 2017). However, growing mouse and, in particular, human kidney organoids is an immensely promising technology for studying kidney development, modeling renal disease and performing nephrotoxicity assays (Morizane and Bonventre, 2017). Reconstructing the mammalian nephron requires understanding of the correct signals needed to guide stem cells down the appropriate differentiation paths to generate highly specialized compartments of cells. Although fully differentiated nephrons have yet to be achieved in organoid cultures, the discovery of terminal differentiation factors, such as tfap2a and tfap2b, can herald progress in this crucial aspect of the kidney organoid field. In summary, our work indicates that further elucidation of the Tfap2a/TFAP2A gene regulatory network in zebrafish, murine and human nephron progenitors can shed valuable insights into nephron differentiation and congenital renal disease.
MATERIALS AND METHODS
Ethics statement and zebrafish husbandry
Adult zebrafish were maintained in the Center for Zebrafish Research at the University of Notre Dame Freimann Life Science Center. All studies were performed and supervised with by the University of Notre Dame Institutional Animal Care and Use Committee (IACUC), under protocol numbers 13-021 and 16-025. Tübingen strain animals were used for wild-type experiments. Zebrafish embryos were raised in E3 embryo media, staged and fixed as described previously (Westerfield, 1993; Kimmel et al., 1995).
Whole-mount and fluorescent in situ hybridization
Whole-mount and fluorescent in situ hybridization were performed as described previously (Marra et al., 2017; Brend and Holley, 2009; Lengerke et al., 2011; Cheng et al., 2014) with antisense RNA probes. Probes were synthesized using IMAGE clone template plasmids for in vitro transcription (Wingert et al., 2007; Wingert and Davidson, 2011). Digoxigenin-labeled probes were used to detect wt1b, slc20a1a, slc12a1, dlx2a, kcnj1a.1, tfap2a, tfap2b, clcnk, slc12a3, slc9a3, cdh17, hnf1ba, irx3b, irx1a, trpm7 and stc1. Fluorescein-labeled probes were used to detect tfap2b, slc12a1, slc12a3, tfap2a and pax2a. For all gene expression studies, each analysis was performed in triplicate with sample size of at least n=20 for each replicate. Representative animals were imaged and absolute length measurements were collected.
Whole-mount immunofluorescence studies were completed as described previously (McCampbell et al., 2015). To assess Tfap2a protein expression, anti-Tfap2a (1:50) (LifeSpan Biosciences, LS-C87212-100) was used. To examine Slc12a1/2 protein expression, anti-T4 supernatant (1:500) (Developmental Studies Hybridoma Bank, Na-K-Cl cotransporter, AB_528406) was employed. To analyze proliferation, anti-phospho-Histone H3 (Ser10) antibody (1:200) (Millipore, 06-570) was used (Kroeger et al., 2017). As an indicator of cell death, anti-caspase 3 (1:100) (BD Biosciences, 559565) was used. For cilia studies, anti-acetylated α-tubulin (1:400) (Sigma-Aldrich, T6793) was used (Marra et al., 2017, 2019a,b). Monoclonal anti-NaKATPase supernatant (1:35) (Developmental Studies Hybridoma Bank, α6F) and anti-aPKC (1:250) (Santa-Cruz, SC-216) were applied to embryos incubated in 0.003% PTU to prevent pigmentation and fixed in Dent's solution (80% methanol, 20% DMSO) overnight at 4°C (Gerlach and Wingert, 2014). For immunofluorescence, the following fluorescently conjugated secondary antibodies were used to detect primary antibodies listed previously at a 1:500 dilution: anti-mouse, anti-rabbit and anti-goat (Invitrogen, A11031, A11061 and A11015). 4,6-diamidino-2-phenylindole dihydrochloride (DAPI) (Invitrogen, D1306) was used to stain nuclei.
Image acquisition and statistical analysis
A Nikon Eclipse Ni with DS-Fi2 camera was used to image whole-mount in situ hybridization samples. A Nikon C2 confocal microscope was used to image whole-mount and fluorescent in situ hybridization, and immunofluorescence samples. The polyline tool in Nikon Elements imaging software was used to measure gene expression domains. Fiji (Image J) software was used to process images and collect data for fluorescent intensity plots. All graphs were generated using GraphPad Prism software. A minimum of three representative samples for each control and experimental group were imaged and measured. Averages and standard errors were calculated. Unpaired t-tests or one-way ANOVA tests were used for statistical analyses.
Mutagenesis, whole-genome sequencing and genotyping
Wild-type zebrafish were exposed to ethylnitrosurea and haploids generated as described (Kroeger et al., 2014). Whole-genome sequencing was performed as described (Leshchiner et al., 2012). Pools of 20 trm mutants and 20 wild-type siblings were identified by whole-mount in situ hybridization analysis for slc12a1 (DE) expression. DNA isolation was conducted using the DNAeasy blood and tissue kit (Qiagen). Whole-genome sequencing results were interpreted using SNPtrack software (Leshchiner et al., 2012; Ryan et al., 2013). Isolation of genomic DNA from individual trm animals was performed and PCR amplification of the tfap2a locus was completed using the following primers: forward 5′-TTTGAACGCTGGCCACCGCCACCTCGCCCTACAATTATTGTTGGCTTGATTTAATTTGCACGTTCGTTTTTGATTTGTCCTTCTGAATTTCACGTCTTTT-3′; reverse 5′-AAATGTTTGGTTTTCGTTTACCAGTTAAAATCCTACCGAAAGGCAAAGGAAATTAACAATTAACCACAGCTCACATGAAGAAAATCTTTGTAATAGCCTT-3′. For all studies, trm mutants were confirmed by genotyping and/or abrogated dlx2 expression. The tfap2bsa10090 mutant strain (ZIRC) employed for genetic studies was genotyped using the following primers: forward 5′-GTTTGGATGTATGGGCAGCTCTGGATTGCTTTCAATTGTTC-3′; reverse 5′-TCCCCAACAGTCACTTTATATTTGGATGTGGAGCTAAGAAG-3′.
The QIAquick PCR Purification Kit was used to purify PCR product and sequenced with the forward primer by the University of Notre Dame Genomics Core Facility. Genotyping of hs:tfap2a transgenic [Tg(hsp70:tfap2a)×24], which was a generous gift from Bruce Riley (Texas A&M University, USA), was conducted by performing PCR amplification (34 cycles, 60°C annealing) of the transgene and running product on a 1% agarose gel. The following primers were used: forward 5′-CTCCTCTCAATGACAGCTG-3′; reverse 5′-ATGGCGGTTGGAAGTCTGAA-3′.
To activate the heat-shock inducible tfap2a transgene, heterozygous transgenic embryos were incubated at 38°C for 30 min as described previously (Bhat et al., 2012; Kantarci et al., 2015). For rescue and gain-of-function studies, transgenic embryos were heat shocked at the 8 ss. For fluorescent in situ hybridization gain-of-function studies, transgenic embryos were heat shocked at the 10 ss. For cRNA synthesis, the open reading frame of tfap2a was subcloned into the pCS2 vector. The primers used for subcloning were: forward 5′-GATCATCGATGCCGCCACCATGTTAGTGCACAGTTTTTCCGCGATGGATC-3′; reverse 5′-GATCTCTAGATCACTTTCTGTGCTTCTCATCTTTGTCACC-3′. For in vitro transcription, the tfap2a template was linearized using the Not1 restriction enzyme. tfap2a cRNA (50 pg) was injected into one-cell stage embryos. For irx1a overexpression and rescue experiments, a pCS2 construct was created to allow for in vitro synthesis of full-length irx1a cRNA with EcoR1 and Xho1 sites flanking the open reading frame and an SP6 site for generating sense RNA. irx1a cRNA (80 pg or 70 pg) was injected into one-cell stage embryos for overexpression and rescue experiments, respectively. RNA runoff reactions to produce irx1a and tfap2a cRNA were performed using the mMessage Machine Sp6 kit (Ambion).
Knockdown and RT-PCR experiments
All morpholino oligonucleotides were synthesized by Gene ToolsC. tfap2a MO-splice (tfap2a MO4) targets the exon 2-intron 2 splice site: 5′-AGCTTTTCTTCTTACCTGAACATCT-3′. tfap2b MO-splice (tfap2b MO1) targets the exon 4-intron 4 splice site: 5′-GCCATTTTTCGACTTCGCTCTGATC-3′ (Knight et al., 2005). irx3b MO-ATG (irx3b MO2) targets the start site: 5′-ATAGCCTAGCTGCGGGAGAGACATG-3′. Morpholinos were solubilized in DNase/RNase-free water to create 4 mM stock solutions and stored at −20°C. The stocks were diluted as follows for microinjection: tfap2a-MO, 1:12; tfap2b-MO, 1:10; irx3b, MO, 1:10. One-cell stage embryos were injected with ∼3 nl of morpholino. All splice-blocking MOs were verified by RT-PCR. irx3b crispants were generated using methods previously described (Naylor et. al., 2018). gRNAs were synthesized targeting the first exon (5′-GGCGCGGAGATCTCGGTCAC-3′) and second exon (5′-GGATGCAGAAAAACGAGATG-3′) using the T7 Megascript kit (Ambion). gRNAs (400 ng/μl) and 0.8 μM of Cas9 protein were injected into one-cell stage embryos (5 nl droplet size). Crispants were then verified by a T7 endonuclease assay, where products were separated on a 1.5% agarose gel. PCR primers for the T7 endonuclease assay were: forward primer, 5′-TCCCGCAGCTAGGCTATAAGTA-3′; reverse primer, 5′-ACGGGATCAAATCTGAGCTATT-3′. Transcript analysis of tfap2a and tfap2b splicing in wild-type embryos, wild-type siblings, trm mutants, tfap2a morphants and tfap2b morphants was performed using RT-PCR (Galloway et al., 2008). In brief, RNA was isolated from pools of about 20 embryos, cDNA was synthesized using random hexamers (Superscript IV, Invitrogen) and PCR was performed with the following primers: trm mutant transcript analysis forward, 5′-GCATTGCATCTAA-AGGGCAGACGAA-3′ and reverse, 5′-TAAGGGTCCTGAGACTGCGGATAGA-3′; tfap2a MO-splice transcript analysis forward, 5′-CCCTATCCATGGAATACCTCACTC-3′ and reverse, 5′-GATTACA-GTTTGGTCTGGGATGTGA-3′; tfap2b MO-splice transcript analysis forward, 5′-AGTGC-CTGAACGCGTCTCTGCTTGGT-3′ and reverse, 5′-TGACATTCGCTGCCTTGCGTCTCC-3′; tfap2bsa10090 mutant transcript analysis forward, 5′-GCGATGAATATTCTGGATCAGTCCGTC-3′ and reverse 5′-CTCCGATTAGTCCGTCTTTGTTCATC-3′. For tfap2a MO and tfap2b MO transcript analysis, bands were gel extracted, purified and sequenced. For trm and tfap2bsa10090 mutant transcript analysis, bands were gel extracted, purified and cloned into the pGemTEasy vector (Promega), and minipreps were sequenced.
Alcian Blue staining and o-dianisidine staining
Alcian Blue cartilage staining was performed as previously described (Neuhauss et al., 1996). In brief, larvae were fixed at 4 dpf for 2 h at room temperature in 4% PFA. Larvae were bleached for 1 h in 10% KOH, 30% H2O2, 20% Tween diluted in distilled water. Samples were digested with proteinase K (10 mg/ml) diluted to 1× for 20 min. Samples were stained in 0.1% Alcian Blue (Sigma) dissolved in 70% ethanol/5% concentrated HCl overnight, shaking at room temperature in glass vials. Larvae were destained using acidic ethanol for 4 h, dehydrated by an ethanol series and stored in glycerol. O-Dianisidine staining was performed as described previously on 4 dpf larvae to visualize blood and vasculature (Wingert et al., 2004).
Acridine Orange assay
Acridine Orange (AO; Sigma A6014; 100×) staining was performed on wild-type embryos and trm mutants to analyze cell death (Westerfield, 1993). In brief, a 50×AO stock solution (250 µg/ml) was made. At 24 hpf, embryos were incubated in 1:50 AO solution (made from 50×stock) diluted in 0.003% PTU/E3 media protected from light for 1 h. Embryos were then washed three times with 0.003% PTU/E3, and then imaged with a dissecting microscope under the GFP filter in 2% methylcellulose/0.02% tricaine.
Dextran clearance assay
To assess kidney function in wild-type embryos and trm mutants, clearance assays using fluorescent 40 kDa dextran-fluorescein (FITC) (Invitrogen) were completed. Embryos were treated with 0.003% PTU at 24 hpf. At 2 dpf, embryos were anesthetized with 0.02% tricaine and dextran-FITC was injected into circulation. Live fluorescent imaging was performed 1 h after injection and 24 h after injection. Embryos were live-imaged with a dissecting microscope under the GFP filter in methylcellulose/0.02% tricaine.
We thank Bruce Riley for sharing the inducible hs:tfap2a zebrafish transgenic. We thank the staff of the Department of Biological Sciences for support and the Center for Zebrafish Research at the University of Notre Dame for their tremendous dedication and care of our zebrafish aquarium. We thank Elvin E. Morales and Hannah M. Wesselman for help with protocol establishment for crispant generation. Finally, we thank all the members of our lab for their support, discussions and insights about this work.
Conceptualization: B.E.C., G.F.G., R.A.W.; Methodology: B.E.C., G.F.G., W.G., R.A.W.; Validation: B.E.C., G.F.G., R.A.W.; Formal analysis: B.E.C., G.F.G., E.G.C., K.H.C., A.E.L., R.A.W.; Investigation: B.E.C., G.F.G., E.G.C., K.H.C., A.E.L., I.L., R.A.W.; Resources: R.A.W.; Data curation: B.E.C., G.F.G., E.G.C., K.H.C., A.E.L., I.L., W.G., R.A.W.; Writing - original draft: B.E.C., R.A.W.; Writing - review & editing: B.E.C., G.F.G., E.G.C., K.H.C., W.G., R.A.W.; Visualization: B.E.C., R.A.W.; Supervision: B.E.C., R.A.W.; Project administration: R.A.W.; Funding acquisition: R.A.W.
This work was supported by the National Institutes of Health (R01DK100237 to R.A.W.). We are grateful to Elizabeth and Michael Gallagher for a generous gift to the University of Notre Dame on behalf of their family for the support of stem cell research. The funders had no role in the study design, data collection and analysis, decision to publish, or manuscript preparation. Deposited in PMC for release after 12 months.
The authors declare no competing or financial interests.