HNF1B controls proximal-intermediate nephron segment identity in vertebrates by regulating Notch signalling components and Irx1/2

The mammalian kidney is an essential excretory organ that regulates fluid balance, osmolarity and pH. It also performs blood filtration in order to excrete metabolism end products and drugs. These diverse tasks are accomplished by the nephrons, which are the basic filtration units of the kidney. Each adult mouse kidney contains ∼13,000 nephrons (Cullen-McEwen et al., 2003). The nephron is composed of a glomerulus, which filters the blood plasma, followed by the proximal tubule, the Henle's loop, the distal tubule and the connecting segment, which connects with the collecting duct system. Each of these segments performs highly specialised functions for glucose and solute transport, acid/base balance and water reabsorption. Therefore, correct segmentation of the nephron is crucial for kidney function.

Metanephric kidney development begins with the emergence of the ureteric bud (UB), which undergoes branching morphogenesis and generates the entire collecting system and ureter. Signals from the tips of the UB induce a subset of the surrounding mesenchymal cells to undergo an epithelial transition and establish a polarised renal vesicle (RV) (Costantini and Kopan, 2010). The distal part of the RV grows and connects to the adjacent UB epithelium and rapidly evolves to form the comma-shaped and then the S-shaped bodies (CSBs and SSBs). The SSB is organised into a proximal segment, which is further subdivided into two epithelial layers – the parietal (Bowman's capsule) and visceral (podocyte) – and the future proximal tubule, followed by intermediate and distal segments. This structure grows and further differentiates to form a mature nephron. Mature nephrons are observed by embryonic day (E) 16.5 in mice, but mesenchyme aggregation continues until postnatal day (P) 2; thus, several nephrogenesis stages coexist at the same developmental stage (Hartman et al., 2007).

The patterning and subsequent differentiation of nephron segments are still poorly defined processes. Several genes exhibit regionalised expression shortly after epithelialisation of the RV. The distal domain is defined by the restricted expression of many genes, including the transcription factors Lhx1 and Brn1 (also known as Pou3f3) (Nakai et al., 2003), the Notch ligands Dll1 and Jag1, as well as Bmp2 (Georgas et al., 2009; Kobayashi et al., 2005), whereas the proximal domain is characterised by the high expression of Wt1. Consistent with these expression patterns, Wt1 is required for glomerulus podocyte layer specification (Kreidberg et al., 1993), and Lhx1-deficient RVs fail to regionalise along the proximo-distal axis, lack expression of the Lhx1 transcriptional targets Brn1 and Dll1 and do not progress to the CSB stage (Kobayashi et al., 2005). Brn1 is in turn involved in loop of Henle elongation and distal convoluted tubule formation (Nakai et al., 2003). Members of the Iroquois gene family exhibit a more restricted expression than Brn1, being in the intermediate segment of the developing nephron. Notably, studies in Xenopus have shown that Irx1 and Irx3 are required for the formation of intermediate segments of the pronephros (Alarcón et al., 2008; Reggiani et al., 2007).

Several studies have implicated the Notch pathway in podocyte and proximal tubule fate acquisition. Accordingly, disruption of Notch2 or Rbpj leads to abnormal nephrons that lack podocytes and proximal tubules but which apparently retain distal tubule cell fates, whereas ectopic activation of the Notch pathway promotes the formation of proximal tubule cells (Cheng et al., 2007).

The POU homeodomain transcription factor hepatocyte nuclear factor 1β (Hnf1b; also known as vHnf1 or Tcf2) plays a crucial role in the early differentiation of various cell lineages and organs in vertebrates, including visceral endoderm, pancreas, liver and kidney (Barbacci et al., 1999; Coffinier et al., 2002; Haumaitre et al., 2005; Lokmane et al., 2008; Bohn et al., 2003; Gresh et al., 2004; Sun and Hopkins, 2001; Wild et al., 2000). Heterozygous mutations in human HNF1B cause the complex syndrome known as renal cysts and diabetes (RCAD), which is characterised by severe abnormalities of the kidney, genital tract and pancreas, as well as early onset diabetes (Barbacci et al., 2004; Bellanné-Chantelot et al., 2004; Bingham et al., 2001; Haumaitre et al., 2006; Heidet et al., 2010; Lindner et al., 1999; Zaffanello et al., 2008).

During early mouse kidney development, HNF1B has recently been shown to be required for UB branching and the initiation of nephrogenesis (Lokmane et al., 2010). In particular, we have shown that within the UB HNF1B directly controls Wnt9b, the primary signal from the UB to the adjacent metanephric mesenchyme that promotes mesenchymal-to-epithelial transition and initiates nephrogenesis (Carroll et al., 2005). However, Hnf1b is also expressed in RVs and during all nephrogenesis steps (Barbacci et al., 1999; Lokmane et al., 2010), suggesting a later role during this process. Consistent with this, analyses of human foetuses carrying heterozygous mutations in HNF1B showed decreased numbers of nascent nephron structures, while the vast majority of glomeruli were cystic (Haumaitre et al., 2005).

Here, we examined the potential function of Hnf1b during nephrogenesis by conditionally inactivating this gene in murine nephron precursors. These studies show that HNF1B is required for the correct patterning of early nephron structures. Major defects are evident at the SSB stage, with altered morphology, dysregulation of nephron markers and an increase in apoptosis. We further show that Hnf1b inactivation leads to strong downregulation of Notch signalling components as well as of the transcription factors Irx1 and Irx2, implying an important function in the differentiation of a proximo-medial SSB subdomain. Parallel studies in the Xenopus embryo show that overexpression of a previously characterised HNF1B dominant-negative construct (Barbacci et al., 2004) results in strong downregulation of proximal and intermediate pronephric segment markers, whereas the distal segments appear to form normally. Thus, the crucial function of HNF1B in nephron patterning appears conserved through vertebrate evolution.

Mice heterozygous for the Hnf1b null allele (Hnf1b^lacZ/+), with the lacZ gene replacing the first exon of Hnf1b (Barbacci et al., 1999), and for the Wnt4^EGFPCre allele (Shan et al., 2010), in which the EGFPCre fusion cDNA and the Neo selection cassette replace a 100 bp region including the translation start site of the Wnt4 gene, were maintained as heterozygotes. The Hnf1b conditional knockout (cKO) allele carrying loxP sites flanking exon 4 was generated with the support of the GIS Maladies Rares and Institut Clinique de la Souris (ICS). Mice homozygous for the Hnf1b cKO allele, designated as Hnf1b^Flox/Flox, are viable and fertile.

Microinjection of Xenopus laevis embryos was performed as described (Umbhauer et al., 2000). nuc-lacZ mRNA was used as a lineage tracer by X-gal staining. The coding sequence of human HNF1B truncated at L329 (L329X) (Barbacci et al., 2004) was subcloned into the pCS2+ vector (Le Bouffant et al., 2012) for capped mRNA synthesis. Inducible Irx1 (Irx1-GR) is from Alarcon et al. (Alarcon et al., 2008).

Whole E15.5 kidneys were fixed in 4% paraformaldehyde (PFA) and processed for OPT as described (Chi et al., 2011) with some modifications. Fixed kidneys in TBST (Tris-buffered saline with 0.01% Triton X-100 and 10% foetal bovine serum) were stained with anti-nephrin antibodies (a kind gift of K. Tryggvason, Karolinska Institute, Stockholm, Sweden) at 4°C for 3 days and then extensively washed in TBST. Secondary antibody was applied at 4°C for 3 days followed by several washes in TBST. Samples were embedded in 1% low-melting-point agarose. Agarose blocks were placed in absolute methanol, cleared with benzyl alcohol/benzyl benzoate solution (1:2) and images captured with a 3001 OPT scanner (Bioptonics, Edinburgh, UK). Three-dimensional movies were prepared and morphometric parameters calculated with Imaris (Bitplan).

Kidneys from embryos/newborns were fixed, embedded in paraffin, sectioned and analysed by immunohistochemistry as described (Lokmane et al., 2008). We used goat anti-HNF4A (Santa Cruz), anti-LTA (Vector Labs), rabbit anti-NKCC2 (provided by M. Knepper, NIH, Bethesda, MD), rabbit anti-PAX2 (Covance), rabbit anti-phosphohistone H3 (Millipore), rabbit anti-WT1 (Santa Cruz), mouse anti-E-cadherin (BD Transduction Laboratories), mouse anti-pan-cytokeratin (Sigma), rabbit anti-laminin (Sigma), mouse anti-ZO1 (Zymed), rabbit anti-JAG1 (Cell Signaling) and rabbit anti-SOX9 (Chemicon). As secondary antibody we used Cy3-conjugated anti-mouse (Jackson ImmunoResearch), FITC-avidin (Vector Labs) for LTA, and for amplification we used biotinylated antibodies before streptavidin-Alexa 488. Immunostaining with anti-HNF1B (Santa Cruz), anti-calbindin D-28K (Chemicon) and anti-E-cadherin, or with anti-PAX2 and anti-β-catenin was performed on 60 μm vibratome sections.

Apotome acquisitions of β-catenin/PAX2 staining were imported into the tomography program IMOD (bio3d.colorado.edu/imod). SSB and collecting duct membranes were outlined in each image and labelled isosurfaces added to the model. Two-dimensional representative images and 3D movies were created using IMOD and Blender (blender3d.fr).

ISH on paraffin sections and terminal deoxynucleotidyltransferase-mediated dUTP-biotin nick end-labelling (TUNEL) were performed as described (Lokmane et al., 2008; Paces-Fessy et al., 2012). Dll1, Lfng, Osr2, Brn1, Irx1, Irx2, Papss2, Tcfap2b and Wfdc2 probes were generated by PCR (GUDMAP database). Whole-mount ISH of Xenopus embryos was performed as described (Le Bouffant et al., 2012).

Total RNA from embryonic kidneys, reverse transcription and qRT-PCR were performed as described (Paces-Fessy et al., 2012). ChIP was performed on E14.5 kidneys (Heliot and Cereghini, 2012). Primers are listed in supplementary material Table S1.

To investigate the specific function of Hnf1b during nephrogenesis we used Wnt4(EGFPCre) mice, which are reported to mediate recombination of the Rosa26 locus in pretubular aggregates, before the RV stage (Shan et al., 2010), when HNF1B is first detected. Males compound heterozygous for Wnt4^+/EGFPCre and Hnf1b^lacZ/+ were crossed with Hnf1b^Flox/Flox females carrying loxP sites flanking Hnf1b exon 4, thus obtaining offspring of genotype Wnt4^+/EGFPCre; Hnf1b^lacZ/Flox (referred to as mutants), Wnt4^+/+; Hnf1b^lacZ/Flox and Wnt4^+/+; Hnf1b^+/Flox (referred to as control) and Wnt4^+/EGFPCre; Hnf1b^+/Flox (referred to as heterozygotes).

To assess the efficiency of Hnf1b inactivation using the Wnt4(EGFPCre) line, we examined whether HNF1B protein was still present in early mutant nephron structures. HNF1B is present in UB branches and collecting system nuclei. During nephrogenesis, it is first expressed at low levels in the distal part of RVs (Fig. 1A), then at high levels in distal CSBs (Fig. 1B) and in a proximo-distal gradient at the SSB stage (Fig. 1C,D), but is absent from the most proximal region that forms the glomerulus. Yet, HNF1B is expressed in the future Bowman's capsule of the SSB and then in all mature nephron segments (Fig. 1C,D; data not shown). In mutants, HNF1B was detectable only in UB and collecting duct nuclei, but not in the RVs nor in their derivatives (Fig. 1E-G; data not shown). Analysis of whole mutant RVs and CSBs showed sporadic unique HNF1B⁺ cells, but neither patches of HNF1B⁺ cells nor mixed positive/negative cells, thus suggesting efficient Hnf1b inactivation. qRT-PCR showed a reduction of wild-type Hnf1b transcript levels of 53% and 70%, respectively, at E14.5 and E16.5, compared with controls (Fig. 1H), further suggesting loss of Hnf1b function in nephron precursors, but not in collecting ducts.

Thus, Wnt4(EGFPCre) activity leads to efficient Hnf1b inactivation from the RV stage, when HNF1B is first expressed, allowing us to assess its specific function during nephrogenesis.

We obtained mutant newborns that were indistinguishable from normal littermates at the expected Mendelian ratio (23%, n=40) (Table 1). However, mutants died within the first 2 days after birth. Gross histological analysis showed a reduction of tubular structures in mutant kidneys (Fig. 2A-D). Notably, the normally easily visible proximal convoluted tubule clusters (Fig. 2E,F) and the medullar elongated Henle's loop tubules (Fig. 2G,H) were not detected in P0 mutants. They also showed a decrease in their medial area, although kidney weights were similar to controls (Table 1). The total number of mature glomeruli was evaluated by OPT using the mature glomerulus marker nephrin. We observed a reduction of 43.5% in mutant kidneys. OPT analysis also uncovered a higher density of mature glomeruli within the mutant medulla region, whereas in controls mature glomeruli are rather homogenously distributed at the periphery and medulla region (Table 1; supplementary material Fig. S1, Movies 1, 2). Moreover, in mutants, 12-16% of glomeruli became cystic with variable enlargements of the Bowman's space and they occasionally exhibited hydronephrosis (supplementary material Fig. S2).

A fraction of compound heterozygous Wnt4^+/EGFPCre; Hnf1b^+/Flox kidneys exhibited, in addition to cystic glomeruli, enlarged tubules, a phenotype that was stronger at E15.5 than at P0 (supplementary material Fig. S2). This phenotype is not the consequence of a genetic interaction between Wnt4 and Hnf1b, as it was not observed in Wnt4^+/EGFPCre; Hnf1b^lacZ/+ heterozygotes, but appears to result from a potential dominant-negative effect of the Hnf1^Δe4 allele (supplementary material Fig. S2; data not shown). Since Wnt4^+/EGFPCre; Hnf1b^+/Flox heterozygotes did not present any of the nephrogenesis defects observed in our mutants, their phenotype will be reported separately.

We first examined the expression of specific nephron segment markers. In both control and mutant kidneys, WT1 was expressed in the most proximal part of SSBs and the podocyte layer (Fig. 3A,B). At E15.5, the layer of podocytes of developing glomeruli appeared thicker and relatively disorganised compared with that of controls (Fig. 3C,D; supplementary material Fig. S3). These defects were not observed when glomeruli became mature (Fig. 3E′,F′; data not shown). Two markers of podocyte maturation, nephrin and ZO1 (TJP1 – Mouse Genome Informatics), were detected normally in mutant E15.5 glomeruli. We also observed similar nephrin/WT1 ratios in controls and mutants, suggesting that the glomerulus capillary network and their maturation were unaffected (Fig. 3A,B,E-F′).

Remarkably, at P0, neither Ncc (Slc12a3 – Mouse Genome Informatics) transcripts, which encode a Na⁺-Cl^– co-transporter of the distal tubule, nor NKCC2 (SLC12A1 – Mouse Genome Informatics), a Na⁺-K⁺-2Cl^– co-transporter specific to the thick ascending limb of Henle's loops, was detected in mutant kidneys (Fig. 3G-J). Moreover, none of the early proximal tubule markers HNF4A, cubulin (Cubn – Mouse Genome Informatics) and megalin (LRP2 – Mouse Genome Informatics), nor LTA, a marker of mature proximal tubules, was detected (Fig. 3K-N; data not shown). These markers were also absent at E16.5, indicating that these defects were not due to a de-differentiation process (data not shown).

Although glomeruli appeared to have formed normally in mutants, cells forming the neck segment that connects with the proximal tubule exhibited a more columnar morphology with larger basal nuclei than controls (Fig. 3O,P). Moreover, this tubule appeared laterally inserted into the Bowman's capsule (Fig. 3P,T; data not shown).

Immunodetection of pan-cytokeratin and E-cadherin on adjacent sections showed that nephron tubules, which are E-cadherin⁺/pan-cytokeratin^–, were not detected in mutant medullar regions (Fig. 3Q-R′), suggesting an absence of Henle's loops. Interestingly, E-cadherin/calbindin co-staining of E16.5 vibratome sections showed that mutant nephrons appeared to comprise only a glomerulus linked to the collecting duct by a short tubule (Fig. 3S,S′), whereas control kidneys showed immature nephrons with distal tubules, primitive Henle's loops and proximal tubules (Fig. 3T,T′).

These data together strongly suggest that neither primitive nor mature nephron segments are formed in mutants.

We next examined whether RVs were correctly epithelialised and polarised by analysing the expression of several key regulatory molecules and epithelial markers. Both control and mutant kidneys exhibited a well-formed extracellular matrix and tight junctions, as stained by laminin and ZO1, respectively, suggesting the correct epithelialisation of mutant RVs (Fig. 4A-D). Fgf8, as well as Lhx1, Brn1 and Bmp2, were also expressed normally in a polarised manner within the mutant RVs (Fig. 4E-L), as well as in the distal part of CSBs (Fig. 4E-L; data not shown).

Together, these results indicate that Hnf1b-deficient RVs correctly initiate the segmentation process but fail to establish nephron segment fates at later stages.

At the SSB stage, developing nephrons present a convoluted morphology with two slits that generate two large curves and the glomerular crevice. This is associated with the regionalised expression of several key markers, prefiguring the future nephron segments.

At this stage, PAX2 exhibits segmented expression and is strongly reduced in the future proximal region, whereas in mutants PAX2 remained highly expressed in this region (Fig. 5A). qRT-PCR revealed an increase in Pax2 transcripts in mutant kidneys both at E14.5 and E16.5 (Fig. 5Y).

We subsequently performed high-resolution 3D reconstructions using β-catenin/PAX2 immunostaining on E15.5 vibratome kidney sections. Mutant SSBs displayed a clearly less compacted structure with expanded slits, and substantially underdeveloped regions that normally contribute to the proximal and intermediate segments (Fig. 5A; supplementary material Movies 3, 4). Remarkably, a similar persistence of PAX2 expression has been reported in Notch2 mutants (Cheng et al., 2007), raising the possibility that our mutants exhibit a similar proximal fate acquisition defect.

We therefore examined the expression of components of the Notch pathway. The Notch ligands Jag1 and Dll1 are both expressed in the distal part of RVs and then, at the SSB stage, are restricted to the precursors of the proximal tubule (Fig. 5B,D). In mutants, expression of Dll1 and Jag1 was induced normally in distal RVs. However, the burst of expression of these ligands at the CSB stage (Fig. 5B,D) appeared severely affected (Fig. 5C,E). Interestingly, at this stage we also observed an even stronger downregulation of Lfng, a modulator of the Notch pathway (Fig. 5F,G). As mutant CSBs progressed into SSBs, the expression of these genes was further decreased and the prospective proximal tubule territory appeared strongly reduced (Fig. 5C,E,G). Downregulation of Dll1, Jag1 and Lfng transcript levels was further confirmed by qRT-PCR, by 53%, 19% and 55%, respectively (Fig. 5Y).

By contrast, the expression of Notch1 and Notch2 receptors was unaffected in mutant kidneys. The Notch effector Hes5, which is expressed specifically in the future proximal territory (Piscione et al., 2004), was strongly downregulated (to 32% of controls), whereas the expression of Hes1 and Hey1 remained unaffected (Fig. 5Y; data not shown). These data together suggest that HNF1B might be required initially at the CSB stage for modulating the expression of Notch ligands and subsequently for maintaining their expression in the SSB prospective proximal segment.

Since HNF1B is also expressed in the parietal part of SSBs, we analysed Osr2 expression. As expected, its expression was strongly reduced but not absent in the mutant SSB future proximal segment. Interestingly, however, Osr2 expression was maintained in the mutant SSB parietal layer, suggesting that Bowman's capsule precursors were correctly specified (Fig. 5H,I,Y). Note also that HNF4A, which is initially expressed in the prospective proximal region of late SSBs, was completely lost in mutants, both at the protein and transcript levels (Fig. 5J,K,Y), thus suggesting that HNF1B, as reported in other tissues (Lokmane et al., 2008), is required for Hnf4a induction.

In our mutants, the SSB intermediate region also appeared affected and Henle's loops were not formed. Brn1-deficient mice present defective maturation of Henle's loops and distal tubules, but primitive loops of Henle tubules are initially formed (Nakai et al., 2003). In our mutants, Brn1 was induced normally and then restricted to the prospective distal segment (Fig. 4; data not shown).

To examine whether the prospective intermediate domain was correctly specified, we analysed several regulators identified as exhibiting polarised SSB expression patterns. The transcription factor Sox9 and the bifunctional enzyme Papss2 display a dynamic expression pattern during nephrogenesis, both being initially restricted to the distal RV (data not shown). As nephron development proceeds, Sox9 exhibits a gradient of expression from the intermediate to the distal domains of the SSB, whereas Paspss2 is initially restricted to the intermediate segment and then to the proximal region (Fig. 5L,N) (Georgas et al., 2008; Reginensi et al., 2011). In mutant SSBs, we observed a decrease of Sox9 and Papss2 expression in the intermediate region (Fig. 5M,O). qRT-PCR in E14.5 mutant kidneys showed a 20% reduction in Sox9 and Papss2 transcripts (Fig. 5Y).

Recently, Iroquois transcription factors, particularly Irx1 and Irx3, have been identified in Xenopus as required for pronephros intermediate tubule morphogenesis. Irx1, Irx2 and Irx3 are also expressed in a highly restricted manner in the intermediate segment of the mammalian SSB, suggesting a similar function (Alarcón et al., 2008; Reggiani et al., 2007). Remarkably, the restricted expression of Irx1 and Irx2 in the intermediate part of the SSB was completely lost in mutants (Fig. 5P-S). qRT-PCR confirmed the reduction of Irx1 and Irx2 transcripts, by 98% and 96%, respectively (Fig. 5Y). Irx3 ISH of control and mutant kidneys gave only very weak signals. We found, however, that the transcript levels of Irx3 in mutant kidneys were decreased by only 21% when compared with controls (Fig. 5Y). A similar decrease in Papps2 and Sox9 RNA levels was observed in mutant kidneys (Fig. 5Y).

Regarding prospective distal segment acquisition, the SSB distal expression of several regulatory molecules (Bmp2, Fgf8, Brn1) was unaffected (Fig. 4; data not shown). Tcfap2b and its target Wfdc2 have recently been reported to be expressed in a distal-medial gradient in the SSB (Yu et al., 2012). ISH detected both in control and mutant distal SSBs, although compared with controls Tcfap2b expression was globally weaker. Wfdc2 qRT-PCR revealed a moderate reduction, which might be related to the intermediate territory reduction (Fig. 5T-W,Y).

Taken together, the mutant SSB morphology along with the abnormal expression of several key molecules strongly suggest that Hnf1b deficiency results in the absence of a subdomain encompassing part of the prospective proximal tubule and part of the adjacent intermediate region, which normally expresses Irx1/2. To confirm this, we analysed kidney sections co-stained for JAG1 and SOX9. These markers normally overlap in a medial SSB subdomain and are likely to establish a boundary between these two regions. In mutant SSBs, no cells co-expressing JAG1 and SOX9 were observed (Fig. 5X). Interestingly, a few distal markers (such as Wfdc2) were maintained in the oldest mutant nephron tubules (data not shown), suggesting that these tubules might derive essentially from the distal SSB region. Thus, initial distal fate seems to be acquired correctly, but further differentiation into mature distal tubule is severely compromised. Concerning the small JAG1/Dll1-positive territory, we have been unable to detect any proximal marker, including the more proximal markers Spp2 and Fbp1 (data not shown).

Altogether, these analyses indicate that HNF1B is a key determinant of the acquisition of a novel proximo-medial subdomain of the early SSB, very likely through the modulation of Notch signalling components and the induction of Iroquois genes.

We then took advantage of HNF1B ChIP sequencing data recently generated for E14.5 kidneys (C.H., O. Bogdanovic, A.D., S. LeGras, I. Davidson, J. L. Gomez-Skarmeta and S.C., unpublished) to identify potential HNF1B targets expressed in the different SSB territories. We first examined genes known and/or proposed to be involved in proximal tubule differentiation and that were dysregulated in our mutants. We found that HNF1B is bound to the Lfng, Dll1 and Hnf4a promoter regions, a result further validated by three independent ChIP assays (Fig. 6A). In addition, we found HNF1B fixation peaks in the promoter regions of several other genes also expressed in the future proximal/intermediate tubule in SSBs (Georgas et al., 2009; Yu et al., 2012), including Cdh6, Pcsk9 and Tcfap2b, which were all downregulated in our mutants (Fig. 6B). The restricted expression patterns of these genes in distinct SSB subcompartments suggest an additional important function of HNF1B in the regulation of a whole set of genes expressed in developing nephron tubules (Fig. 6C).

Remarkably, we also found that HNF1B was recruited to several sequences highly conserved within a 700 kb region between the mouse Irx1 and Irx2 genes (at –430,229 bp and –418,342 bp with respect to the Irx2 transcription start site). This intergenic region contains several enhancers and exhibits an evolutionarily conserved architecture that brings the promoter of the two genes together in the same chromatin landscape, which is essential for their co-regulation (Tena et al., 2011). Consistent with this, we found a small HNF1B fixation peak in the Irx2 promoter (between –587 bp and +50 bp), without an identifiable HNF1B binding motif, suggesting chromatin loop formation (Fig. 6A). Further studies are required to confirm the importance of the far upstream sequences in the regulation of the Iroquois gene cluster during nephrogenesis, but these results taken together lead us to suggest that an HNF1B→IRX1/2 cascade might contribute to the acquisition of intermediate segment identity (Fig. 6C).

The less convoluted morphology of the mutant SSB suggested either a decrease in proliferating cells or an increase in apoptosis. We quantified mitotic cells using an anti-phosphohistone H3 antibody. At E16.5, global proliferation appeared unaffected. We observed, however, a moderate decrease of proliferating cells in mutant CSBs (by ∼14%) and a stronger decrease in the future proximal tubule of SSBs (∼42%) (Fig. 7A-C), whereas proliferation in other kidney structures was unaffected. Interestingly, Notch2 mutants exhibit a similar decrease in the proliferation of proximal region precursors (Cheng et al., 2007). Thus, Hnf1b activity might be required to maintain normal proliferation levels of proximal tubule precursors, probably, at least in part, via the Notch pathway.

To define whether apoptosis was perturbed in mutant kidney, we preformed TUNEL assay at E16.5. We found an increase in apoptosis in epithelia of mutant kidneys, but not in the stroma. By quantifying the localisation of TUNEL-positive cells, we observed a strong increase in apoptosis in late mutant ‘nephron tubules’ (7.6-fold). In addition, we observed a more moderate increase of apoptotic cells in late SSBs (∼3-fold) (Fig. 7D-F). Fgf8 has been shown to be required for nephron tubule survival (Grieshammer et al., 2005). Yet, we observed that it was correctly induced and maintained in mutant SSBs, suggesting an FGF8-independent apoptosis pathway (Fig. 4C; data not shown).

We propose that this increased apoptosis is a consequence of defective territory specification at the SSB stage, without excluding a direct function of HNF1B in cell survival.

Recent molecular analyses have revealed an extensive conservation of nephron segmentation between the Xenopus pronephros and the mammalian metanephros (Raciti et al., 2008). Hnf1b is expressed in the Xenopus pronephric field of the neurula and its expression is maintained throughout the entire pronephros, with the highest levels at tailbud stages in the proximal part of the pronephros (M.U., unpublished). Given these observations and the anatomical simplicity of the Xenopus pronephros, we investigated the effect of expressing a dominant-negative human HNF1B construct (HNF1B-DN) on pronephros segmentation. This construct encodes a truncated protein (L329X) that contains the N-terminal dimerisation domain and the POU homeodomain of HNF1B, binds DNA and heterodimerises efficiently with the wild-type protein (Barbacci et al., 2004). Selected amounts of HNF1B-DN mRNA were injected into Xenopus embryos (supplementary material Fig. S4) and only those with normal morphology were considered.

We first validated the dominant-negative effect of HNF1B-DN overexpression by confirming the brain defects described in zebrafish vhnf1 (hnf1ba) mutants (Lecaudey et al., 2004; Sun and Hopkins, 2001; Wiellette and Sive, 2003) (supplementary material Fig. S4). We then investigated the effect of HNF1B-DN ectopic expression on pronephros development by ISH analysis of pronephric segment markers at the tadpole stage (stages 33-35) (Fig. 8A-N). Expression of slc4a4 (NBC1) in the proximal part of the pronephric tubule was strongly inhibited on the injected side (by 90%, n=32). Notably, the most proximal expression of pax2 was also strongly affected (83%, n=24) and the three nephrostomes were absent (83%, n=24). By contrast, pax2 expression in the distal pronephric tubule was similar to that on the control side. The expression of slc12a1 (NKCC2), which is normally detected in the intermediate tubule and first distal segment, was strongly inhibited, especially in the intermediate segment (86%, n=58). Consistent with this, irx1 and irx2, the expression of which is confined to the first segment of the intermediate tubule, were strongly downregulated [89% (n=28) and 69% (n=29), respectively]. Moreover, irx3, which is expressed in the intermediate tubule and the last proximal segment, was also diminished (78%, n=28). The most anterior part of evi1 expression that corresponds to the intermediate segment was often decreased (59%, n=34), whereas the distal expression remained normal (100%, n=34). Interestingly, pronephros defects caused by HNF1B-DN were partially rescued by an IRX1 construct (Irx1-GR) induced at the early neurula stage, further suggesting that HNF1B is upstream of Irx1 in a complex regulatory circuit controlling nephron patterning (supplementary material Fig. S5).

Notch signalling has also been shown to play, as in mice, a predominant role in patterning the Xenopus pronephros anlagen by regulating proximal fate (McLaughlin et al., 2000; Taelman et al., 2006). Consistent with our findings in mice, the expression of dll1, which is normally localised to the dorsoanterior portion of the pronephric anlagen, was either absent or strongly reduced (by 90%, n=49) at the early tailbud stages (stages 24-25) (Fig. 8O,P). At this stage, evi1 and pax8 expression was either similar to that on the control side [77% (n=45) and 74% (n=35), respectively] or slightly reduced (Fig. 8Q,R; data not shown).

In contrast to these observations, the overexpression of two human HNF1B mutants in Xenopus embryos has recently been shown to result in modest effects on pronephros development, although proximal tubules were the most affected (Sauert et al., 2012). Since the mutants employed in that study either retained transcriptional activity (P328L329del) or did not bind DNA (A263insGG), it is likely that they were less efficient than our construct in decreasing Hnf1b activity.

Together, these results indicate that in Xenopus Hnf1b is required for proximal and intermediate tubule fate and further suggest that, as in mice, it may act on these patterning events through the Notch pathway and the Iroquois genes within a relatively complex regulatory circuit

We report here that conditional Hnf1b inactivation in nephron progenitors leads to a glomerulus that is connected to the collecting system by a short tubule that has acquired initial distal fates. RV epithelisation and polarisation are unaffected, but the SSBs exhibit abnormal regionalisation and morphology. Molecular analyses reveal an unsuspected role of Hnf1b in the maintenance of high expression levels of Notch signalling components and the induction of transcription factors that are expressed in restricted regions of nascent nephrons. Together, these data uncover a novel and crucial role of HNF1B in the acquisition of a proximal-intermediate nephron segment fate, probably through the direct control of several key regulators.

Interestingly, the earliest difference observed between control and mutant at the CSB stage is a dysregulation of key components of Notch signalling, including the modulator Lfng and the ligands Dll1 and Jag1. The Notch pathway has been shown to be required for proximal tubule and podocyte fate acquisition (Bonegio et al., 2011; Cheng et al., 2007; Wang et al., 2003). Hnf1b inactivation, however, does not phenocopy the previously described Notch mutants and leads to relatively normal glomerulus development. It remains possible that the temporal requirement of Notch signalling for glomerulus formation takes place earlier, at the RV or early CSB stages. In contrast to Dll1, Jag1 is expressed in pretubular aggregates, and at the SSB stage has a more extended expression than Dll1 into the most proximal region that will form the glomerulus (Fig. 5). Thus, the Jag1-induced Notch pathway could be specifically involved at earlier stages in podocyte fate acquisition. Alternatively, the residual expression of Notch ligands in mutant SSBs could be sufficient to induce glomerulogenesis but not to acquire a proximal tubule fate. Both possibilities are supported by the phenotype reported for hypomorphic Dll1 mutant mice, which present defective proximal tubule formation but normal glomerulogenesis (Cheng et al., 2007).

Additional proteins modulate the activity of the Notch pathway and we show here that Lfng, which exhibits a very similar expression pattern to Dll1, is strongly downregulated in mutant CSBs. We also show that HNF1B is recruited in vivo to Lfng promoter sequences, as it is to the Dll1 promoter, suggesting direct transcriptional regulation. Although the function of Lfng in nephron patterning is unknown, we favour the hypothesis that HNF1B integrates proximal tubule fate acquisition by the Notch pathway primarily through the regulation of Lfng, in addition to Dll1. Further studies are needed to better understand the precise role of particular Notch ligands in SSB patterning.

The phenotype observed in our mutants is not restricted to the control of components of the Notch signalling pathway. We also noted that Hnf4a, a known HNF1B target expressed in the prospective proximal region of the late SSB, is not induced. More importantly, Hnf1b inactivation also leads to a severe intermediate tubule fate defect, and Irx1 and Irx2, which are normally highly restricted to this SSB territory, are virtually undetectable. The regulatory network involved in the specification of this particular region is unknown, as are the consequences of Irx1/2 loss of function in mice. We cannot exclude the possibilities that lack of a proximo-intermediate segment in our mutants is either due to HNF1B acting primarily through the Iroquois genes that secondarily affect more proximal fate specification or, alternatively, to an earlier defective proximal specification, via the observed deregulation of the Notch pathway, that subsequently would perturb intermediate segment fate acquisition. Precise analyses of Dll1 and Lfng, as well as of Irx1/2 conditional mutants, are required to assess these possibilities. However, detailed expression analyses in early mutant SSBs suggest that, at this stage, there is a reduction of a proximo-medial territory, although not an absence (Fig. 5). In addition, if the complete absence of HNF4A and Irx1/2 reflects the lack of the whole territory that normally expresses these genes, one would have expected the absence of a larger segment encompassing the entire prospective proximal and intermediate segments, and this does not seem to be the case.

The consequences of expressing an HNF1B dominant-negative construct for pronephros development in Xenopus further substantiate the results of Hnf1b inactivation in mouse nephron progenitors. We observed strong downregulation of both early proximal and intermediate tubule markers, whereas distal markers were not, or only slightly, affected. These studies demonstrate that Hnf1b is required for proximal and intermediate tubule fate in Xenopus and further suggest that Hnf1b might function, as in mice, through the Notch signalling pathway via Dll1 and the Iroquois genes in these patterning events. However, further studies are required, both in Xenopus and mouse, to uncover the precise regulatory circuit that links these genes during nephron segmentation.

Detailed transcriptional profiling of the developing mouse kidney highlighted a strong statistical association between the expression of HNF1B in the developing proximal tubules and the presence of well-conserved HNF1 binding sites in the promoters of many genes that are highly expressed in this nephron structure (Brunskill et al., 2008). More than 40% of these genes were found bound by HNF1B in the ChIP-Seq data. Moreover, HNF1B ChIP-Seq on E14 kidney, which is at the precise stage when nascent nephron structures predominate, indicates that HNF1B is recruited to the regulatory sequences of several key genes induced either in restricted prospective or intermediate segment SSB domains, further substantiating the crucial role of HNF1B in SSB patterning. Based on these observations and our findings in Hnf1b mutants, it is tempting to speculate that HNF1B function is required for the transcriptional induction of Irx1/2 and Hnf4a taking place only at SSB stages: cells lacking expression of these key regulators would subsequently lose their identity, probably being extruded or removed by apoptosis.

In our mutants we observe a moderate decrease in proliferation in CSBs, followed by a strong increase in apoptosis in late SSBs. Interestingly, a similar decrease in proliferation has been described in Notch2 mutants (Cheng et al., 2007). However, and in contrast to our mutants, no significantly enhanced apoptosis was observed in Notch2 mutants. Analysis of Fgf8 hypomorphic mutants has indicated a key role of FGF8 in SSB tubule progenitor survival (Grieshammer et al., 2005). Intriguingly, the kidney phenotype of Fgf8 hypomorphic mutants is very similar to that of our Hnf1b mutants. However, the expression of neither Fgf8 nor the receptors Fgfr1 and Fgfr2 is affected in our mutants (data not shown). We have also been unable to identify potential survival factors that might be regulated by HNF1B. We conclude that increased apoptosis is probably a secondary consequence of the abnormal specification of prospective nephron segments.

Surprisingly, glomerulogenesis appears weakly affected, with a slight increase in WT1 staining at the early renal corpuscle stage, when the podocyte layer begins to form a ‘cup’ around the capillary bundle. Subsequently, at P0, we observe apparently normal glomeruli and urine in mutant bladders. However, some glomeruli become cystic and it is interesting to note that, in humans, mutations in HNF1B are frequently associated with glomerulocystic disease (Heidet et al., 2010; Zaffanello et al., 2008). Previous analysis of the cystic dysplastic kidneys of two human foetuses carrying heterozygous mutations in HNF1B (Haumaitre et al., 2006) have also shown a global decrease of structures, labelled by either LTA, NKCC2 or UMOD, suggesting that decreased levels of HNF1B are associated with defective or delayed nephrogenesis. Thus, our data represent not only a significant advance in comprehension of the regulatory networks controlling the early steps of nephron segmentation but might also provide deeper insights into the complex RCAD disease associated with heterozygous mutations in HNF1B. Remarkably, conditional inactivation of Hnf1b in nephron progenitors using the Six2-cre line and another Hnf1b floxed allele resulted in the same phenotype described here, further confirming the crucial role of HNF1B in nephron patterning (Massa et al., 2013).

In conclusion, our results show that HNF1B is an essential transcriptional regulator that, in addition to its known roles in UB branching and induction of nephrogenesis, is required for normal SSB patterning and subsequent morphogenesis of all nephron segments. It appears to function both non-cell-autonomously via the modulation of various components of the Notch signalling pathway and cell-autonomously through the direct induction of key regulators in specific SSB subdomains, potentially participating in proximal and Henle's loop segment fate acquisition. Our study also uncovers a previously unappreciated function of a proximal-medial subcompartment in global nephron patterning: SSBs with defective fate acquisition of this domain fail to differentiate all nephron segments despite the initial acquisition of distal fate.

The cover image was taken by A. Desgrange. We thank P. Mailly for help in 3D reconstructions; J. L. Skarmeta for discussions and reagents; M. Pontoglio and E. Fisher for sharing unpublished data; R. Schwartzmann for confocal imaging; and the electron microscopy service of IFR83-Pierre Marie Curie University.