Hepatocyte nuclear factor 1β controls nephron tubular development

Kidney development is a tightly regulated morphogenic process based on crosstalk between the metanephric mesenchyme (MM) and ureteric bud (UB) (Dressler, 2009). Any dysfunction in this process may lead to a variety of congenital abnormalities of the kidney and urinary tract (CAKUT) (Rumballe et al., 2010). Depending on the nature of the affected compartment, abnormalities in the differentiation program can lead to defects ranging from kidney agenesis to urinary tract malformations.

One of the most prevalent genetic defects responsible for this complex pathology is represented by mutations in the gene that encodes hepatocyte nuclear factor 1β (HNF1B), a divergent homeobox transcription factor. HNF1B mutations are detected in ∼20-30% of patients or fetuses that suffer from renal developmental anomalies, predominantly including renal cysts, hypoplasia or single kidneys (Decramer et al., 2007; Heidet et al., 2010; Ulinski et al., 2006). Even though mutations in HNF1B are known to be responsible for maturity-onset diabetes of the young (MODY) type 5 (Horikawa et al., 1997; Lindner et al., 1999), several studies have consolidated the idea that one of first traits of pediatric patients and fetuses carrying HNF1B mutations is represented by the occurrence of renal malformations (Decramer et al., 2007; Heidet et al., 2010; Ulinski et al., 2006).

The study of animal models for Hnf1b deficiency in renal development and maturation has helped us to understand some of the functions of HNF1β (Bohn et al., 2003; Gresh et al., 2004; Hiesberger et al., 2004; Lokmane et al., 2010; Sun et al., 2004; Verdeguer et al., 2010; Wu et al., 2004). The phenotype of mice lacking Hnf1b depends very much on exactly when (developmental timing) and where (which tissue compartment) the gene is inactivated. For instance, it has been shown that germline inactivation of Hnf1b is embryonic lethal (E7.5) (Barbacci et al., 1999; Coffinier et al., 1999a). During kidney development, the lack of this gene in the UB leads to defective UB branching and to the absence of mesenchymal to epithelial transition due to defective expression of key genes including Wnt9b (Lokmane et al., 2010) (our unpublished observations). In parallel, it has been shown that deletion of Hnf1b in already formed, still elongating, tubular segments leads to a severe polycystic kidney phenotype (Gresh et al., 2004; Verdeguer et al., 2010). In this context, it has been demonstrated that HNF1β is a bookmarking factor that is required for postmitotic reprogramming of the expression of a number of genes whose mutation is frequently associated with cystic kidney disease (Verdeguer et al., 2010). In addition, we have shown that the tubular elongation of nephrons during their postnatal maturation is due to the particular orientation of mitotic spindles. This mitotic alignment is lost in the renal cystic epithelium (Fischer et al., 2006).

In recent years a considerable number of studies have clarified the key molecular mechanisms involved in the first steps of nephron morphogenesis (Boyle et al., 2011; Carroll et al., 2005; Cheng et al., 2007; Stark et al., 1994). However, the signaling pathways and the actors that specify the identity and expansion of tubules in nephron precursors are still mostly unknown. Recent studies have highlighted the crucial role of the Notch signaling pathway in tubular segmentation and glomerular formation. The inactivation of Notch2, one of the Notch receptors, in renal mesenchyme is responsible for the absence of both proximal tubular formation and glomerulogenesis. In addition, a hypomorphic allele of one of its ligands, delta-like 1 (Dll1), is responsible for a drastic defect in proximal tubular formation (Cheng et al., 2007).

Nephron tubular compartments are generated by a proliferative expansion of presumptive tubular cells that emerge from S-shaped bodies. At the morphological level, these nephron precursors are composed of three distinct segments: the upper limb, which will give rise to the distal convoluted tubule (DCT), the mid-limb, which will generate Henle's loop (HL) and proximal tubule (PT) segments; and the lower limb, which will form the blood filtration unit of the nephron, the glomerulus. Recently, analysis of the expression pattern of specific developmental genes has shown that the mid-limb can be further subdivided into specific domains (Thiagarajan et al., 2011; Yu et al., 2012). However, the biological significance of this fine-tuned, patterned gene expression, along with the actual developmental functions of these domains, remain poorly explored.

In the present study, we demonstrate that the lack of HNF1β in the MM leads to the formation of aberrant nephrons that are characterized by glomeruli with a dilated Bowman's space and that lack their typical tubular components. Mutant S-shaped bodies were characterized by the absence of a bulge of epithelial cells in the mid-limb. We show that this phenotype is linked to the defective expression of Irx1, Osr2 and Pou3f3 and to defective Notch signaling activation associated with a downregulation of Dll1.

Specific inactivation of Hnf1b in the MM was obtained using a Cre-loxP strategy. Six2-Cre (Kobayashi et al., 2008), Hnf1b^lacZ/+ and Hnf1b^f/f mice were described previously (Coffinier et al., 2002; Coffinier et al., 1999b). Since Six2-Cre; Hnf1b^f/+ or Six2-Cre; Hnf1b^lacZ/+ mice were indistinguishable from wild-type mice, all these animals are indicated in this study as ‘controls’. Inactivation of Notch2 was achieved with a floxed allele Notch2^tm3Grid and/or with a null allele described previously (BayGenomics, San Francisco, CA, USA) (Tan et al., 2009). The inactivation of Pou3f3 was described previously (Nakai et al., 2003). Animals were maintained in two animal facilities, each of which is licensed by the French Ministry of Agriculture (agreement A 75-14-02, dated April 24, 2007). All experiments conformed to the relevant regulatory standards.

Human fetal renal specimens and fetal tissue for DNA isolation were collected after obtaining informed consent from the parents. The pregnancy had been interrupted during the third trimester because of cystic kidneys, following parental request and acceptance by the Prenatal Diagnosis Center of Lyon (France). Genomic DNA was isolated by standard methods. Mutation screening was performed as described (Heidet et al., 2010). Images of histological sections from a MODY5 patient carrying a c.232G>T, p.Glu78X mutation were provided by Professor Laurent Daniel, CHU Timone, Marseille, France.

Mouse kidneys were dissected from embryos at E13.5, E14.5, E16.5 and E18.5 and from newborn pups (P0). Histological and immunohistochemical analyses were performed on 5 μm paraffin sections. After antigen retrieval in boiling citrate buffer (Dako, S2369) for 15 minutes, sections were treated with the MOM Kit (Vector Labs, BMK-2202) according to the manufacturer's instructions and incubated with antibodies (supplementary material Table S1) for 1 hour at room temperature or at 4°C overnight. After several washes in PBS containing 0.1% Tween 20 (PBST), samples were incubated with Alexa Fluor-conjugated secondary antibodies (Invitrogen; 1:500) for 1 hour at room temperature. Sections were analyzed with a Leica microscope equipped with a Yokogawa CSU-X1M1 spinning disk. Images were analyzed with ImageJ (NIH).

cDNA from genes of interest was obtained by reverse transcription followed by PCR amplification using specific primers (supplementary material Table S2). The templates were either subcloned into pGEM7Z(+) Vector (Promega) or used directly as templates with a specific SP6 or T7 tail to generate digoxigenin-labeled RNA probes with the Digoxigenin RNA Labeling Kit (Roche). ISH experiments were carried out as described (D'Angelo et al., 2010).

cDNA was obtained from total RNA from E13.5, E14.5 and P0 kidneys of control and mutant animals using the SuperScript III Reverse Transcriptase Kit (Invitrogen, 18080-051). Gene expression was analyzed by quantitative (q) PCR using the Stratagene Mx3000P Real-time PCR System with SYBR Green-Rox (Roche, 04 913 914 001) in triplicate. Primers (supplementary material Table S3) were designed using Primer3 software (simgene.com/Primer3). Expression levels in mutants are indicated relative to controls. Results were normalized to the Gapdh expression level. n=3 for control and mutant at E13.5 and P0; n=5 for control and mutant at E14.5.

Nuclei were purified from pooled wild-type embryonic kidneys. Chromatin preparation and ChIP were performed as described (Verdeguer et al., 2010). Quantification of precipitated specific DNA fragments was carried out by qRT-PCR in triplicate. Relative fold-enrichment of DNA fragments was calculated using the following formula: ChIP HNF1β/ChIP IgG. Primers (supplementary material Table S4) were designed using Primer3 software.

To evaluate proliferation and apoptosis, 100 μm vibratome sections were postfixed for 10 minutes in 4% paraformaldehyde at room temperature and then incubated with peanut agglutinin (Vector Labs, RL-1072) and anti-cleaved caspase-3 antibody (R&D Systems, AF835) or anti-phosphorylated histone H3 antibody (Abcam, Ab5819) for 48 hours in PBS containing 0.1% sodium azide and 0.01% Triton X-100.

Kidneys were cut into 1 mm³ slices, fixed for 1 hour in 3% glutaraldehyde and postfixed in 1% osmium oxide. EPON resin-embedded samples were then cut (70 nm ultrathin sections) with a Reichert Ultracut S ultramicrotome and analyzed with a JEOL 1011 electron microscope.

To analyze the role played by HNF1β in nephrogenesis we used a Cre mouse strain that selectively inactivates Hnf1b in nephron progenitor mesenchymal cells. SIX2 is a transcription factor that is specifically expressed in mesenchymal stem cells that give rise to all nephron epithelial cells (Kobayashi et al., 2008). It has been shown that Six2-Cre transgenic animals (Humphreys et al., 2008) target this stem cell compartment without affecting the UB. With this strategy, embryos carrying an Hnf1b^f/f; Six2-Cre germline genotype should maintain the expression of HNF1β in the UB, which in turn should provide the mesenchyme with all the signaling stimuli necessary for the mesenchymal to epithelial transition (MET) at the origin of nephrogenesis.

Our results showed that MM-specific Six2-Cre-driven Hnf1b-deleted embryos (indicated here as ‘mutant’ embryos) did not suffer from any prenatal lethality and were delivered at the expected Mendelian ratio. However, newborns died between the first and second day after birth (Table 1), a lifespan compatible with severe kidney failure. Macroscopic examination showed that kidneys from mutant newborns were small, with a length ∼70% of that of control littermates (Fig. 1A,B). Consistently, the bladder of mutant embryos was small, suggesting that pups were unable to produce normal levels of urine (Fig. 1C,D). In addition, 15% of newborn pups showed hydroureter and hydronephrosis (data not shown).

At birth, the nephrogenic zone, which is localized in the more external part of the cortex, was of similar thickness in mutant and control kidneys. However, the typical convoluted tubular epithelial structures that are normally visible in the cortex were absent in mutant kidneys (Fig. 1E,F).

In order to characterize the identity of the missing tubular segments, we performed immunostaining and in situ hybridization (ISH) on renal histological sections. Our results showed that, at birth, mutant pups lacked most of the tubular nephron segments that are typically produced by the nephrogenic mesenchyme. In particular, mutant embryos showed a very drastically reduced production of PTs, HLs and DCTs. The use of Lotus tetragonolobus lectin (LTL) showed only rare, scattered PTs (Fig. 1G,H). In a similar way, only very rare HLs and DCTs were detected, as shown by the almost complete lack of Slc12a1 (Nkcc2) and parvalbumin (Pvalb) expression (Fig. 1I-L). In addition, qRT-PCR demonstrated that, at P0, mutant pups had a dramatic decrease in the expression of several mRNAs normally expressed in the PT, HL and DCT (Fig. 1M).

In agreement with the absence of Cre expression in the UB, the collecting ducts (CDs) were unaffected, as demonstrated by the presence of the typical medullar tubular rays in mutant newborn pups (Fig. 1E,F). In addition, immunochemistry, immunofluorescence and qRT-PCR experiments showed that all these tubules were positive for typical markers of CDs, such as Dolichos biflorus agglutinin (DBA) staining (supplementary material Fig. S1A,B) and Wnt9b (supplementary material Fig. S1E) and remained positive for the expression of HNF1β (Fig. 2).

Interestingly, the absence of HNF1β did not significantly impair the differentiation of cell types in glomeruli. Consistent with these findings, podocytes expressed normal levels of Wt1 (supplementary material Fig. S1C,D). Nevertheless, these glomeruli presented with an abnormal shape (see below) and several exhibited visible dilation of the Bowman's capsule (Fig. 1E,F, inset).

Taken together, these observations demonstrated that, in the absence of HNF1β, nephron precursors were somehow able to form glomerular structures but were not competent to generate and expand tubules.

In order to characterize the mechanisms underlying the phenotypes observed we aimed to identify the first detectable abnormalities that led to these drastic developmental defects. In agreement with previous studies (Harding et al., 2011; Lokmane et al., 2010; McMahon et al., 2008), immunofluorescence experiments showed that HNF1β was not expressed in cap mesenchyme (Fig. 2A, arrow). HNF1β was first detected in renal vesicles, with a clear polarized distal pattern in a limited number of cells adjacent to the UB, in the prospective tubular compartment (Fig. 2A). The first steps of MET were preserved in kidney from embryos carrying a Six2-Cre-driven Hnf1b deletion (Fig. 2B). In addition, the characteristic mutually exclusive distal/proximal expression pattern of specific genes (Georgas et al., 2008) was preserved. In both control and mutant embryos, vesicles were characterized by polarized expression of Jag1 in the distal portion of the vesicle, whereas Wt1 expression was restricted to the proximal compartment (Fig. 2C,D) (Georgas et al., 2008).

In mutant embryos, comma-shaped bodies did not display any detectable morphological abnormalities, whereas the structure of S-shaped bodies was significantly altered. In control embryos, HNF1β was expressed at high levels in the prospective tubular segments of comma-and S-shaped bodies (Fig. 2E,G). Of note, HNF1β was expressed in the prominent and protruding bulge of epithelial cells that forms the first bend between the mid and the lower limb, clearly visible in sagittal sections. In the absence of HNF1β, this typical structure did not form (Fig. 2G,H,K-N). In coronal sections of S-shaped bodies of control embryonic kidneys, we observed the mid-limb cut in a perpendicular orientation, almost circular in shape. This structure is typically positioned over a concentric, concave layer of columnar epithelial cells that are strongly positive for Wt1 (which gives rise to the typical podocyte precursor ‘smiling’ crown) (Fig. 2I). In mutant embryos, this stereotypical ‘smiling’ structure was replaced by a ‘mustached’ configuration (Fig. 2J). Mutant embryonic ‘S-shaped’ structures were characterized by the presence of a mid-limb that was directly connected to the prospective urinary space in the lower limb compartment. In spite of this drastic defect, the most proximal part of the S-shaped body (lower limb) developed normally, giving rise to precursors of podocytes and Bowman's capsule epithelial cells, as illustrated by the expression of Wt1 (Fig. 2L). The unaltered structure of the lower limb in mutant embryos correlated with the complete absence of HNF1β expression in podocytes and their precursors (Fig. 2G).

In summary, these results showed that the absence of HNF1β in nephron precursors prevented the formation of the prominent and protruding bulge of epithelial cells at the bend between the mid and lower limbs.

It is known that the emergence, expansion and first differentiation of PTs and HLs begin in the more advanced and mature S-shaped bodies already at E14.5. In qRT-PCR assays on cDNA prepared from total kidneys of control embryos we detected the expression of typical tubular markers (Hnf1a, Hnf4a, Vil1 and Slc12a1) already at this stage. Their expression further increased just before birth (Fig. 3A, Fig. 2). In agreement with the defective tubule formation and expansion visible in newborns, the expression of all these markers was drastically decreased in mutant embryos already at E14.5 (Fig. 3B).

Most of these genes are known to contain several HNF1 binding sites in their transcriptional control regions (Tronche et al., 1997) (M.P. and S.G., unpublished). The defective expression of these genes might have been ascribed to a potential direct transcriptional activation effect of HNF1β. The concomitant loss of expression of all these genes could have contributed to the aberrant formation and expansion of nephron tubular components. However, in agreement with previous observations (Harding et al., 2011; McMahon et al., 2008), our inspection clearly revealed that S-shaped bodies in wild-type embryos were still devoid of significant expression (protein or mRNA) of these genes. Hnf1a, Hnf4a and Vil1 were expressed only at later developmental steps of differentiation and specifically when PTs have already emerged from S-shaped bodies (Fig. 3C; data not shown). Since mutant embryo ‘S-shaped’ bodies never expanded these tubular structures, we concluded that the defective expression of these genes was simply a consequence, and not the origin, of the developmental defect.

Altogether, these experiments indicated that the absence of tubular components in mutant newborn animals is not due to a progressive loss of normally differentiated tubules. Our results indicated that, in the absence of Hnf1b, the process of tubular expansion and differentiation was completely impaired.

The aberrant conformation of mutant S-shaped structures might have been ascribed to a defective proliferation/apoptotic balance in the precursors of S-shaped bodies (vesicle or comma-shaped bodies). To evaluate this possibility we monitored and compared the occurrence of proliferative and apoptotic events by counting the number of phosphohistone H3 PS10-positive and cleaved caspase-3-positive cells, respectively. This scoring was performed in 100 μm vibratome sections in order to evaluate the occurrence of proliferative and apoptotic events in structures throughout their whole extent. These experiments showed that mutant S-shaped bodies exhibited significantly decreased proliferation specifically in the upper and mid-limbs. These segments were also characterized by a significantly increased number of apoptotic events (Fig. 3D,E).

In the lower limb, we did not see any difference in the average number of proliferative or apoptotic events. This observation paralleled the fact that, in the lower limb, Hnf1b is not expressed in podocyte precursors and is expressed only at low levels in the prospective capsular cells.

It is worth noting that early nephron precursors (vesicles and comma-shaped bodies) were unaffected by any defective proliferation or excessive apoptosis. Aberrations in these processes only occurred in already deformed S-shaped segments (Fig. 3D,E). This indicated that the defective proliferation and excessive apoptosis can very likely be ascribed to late, secondary effects, possibly linked to the deformation or to the lack of differentiation induced by the absence of HNF1β.

Recent studies have clearly indicated that the gene expression pattern in nephron precursors is more complex than previously thought. In particular, large-scale analysis of the expression of transcriptional regulatory factors has revealed novel molecular subdomains in developing renal structures (Thiagarajan et al., 2011; Yu et al., 2012). We took advantage of these findings in order to identify the specific nature of the deformation of S-shaped structures in mutant embryos. As a first approach, based on qRT-PCR analysis, we monitored genes that have been shown to be expressed in subdomains of the mid-limb, including Osr2, Pou3f3 and Irx1 (Yu et al., 2012). Our results showed significant downregulation in their expression already at E13.5 (Fig. 4A). In control embryos at this stage, Cdh6 is predominantly expressed in renal vesicles. With the progressive appearance of comma- and S-shaped bodies, Cdh6 also becomes specifically expressed in the first prospective PT precursor cells (Cho et al., 1998; Mah et al., 2000). In mutant embryos, Cdh6 expression was normal at E13.5 but drastically decreased, together with Osr2, Pou3f3 and Irx1, at E14.5 (Fig. 4B).

The mid-limb of S-shaped bodies is characterized by the expression of Pou3f3 in a relatively large segment, whereas the expression patterns of Osr2 and Irx1 tend to be more restricted to the most proximal part of the mid-limb. ISH experiments revealed that the expression of these two genes was significantly reduced in mutant S-shaped bodies (Fig. 4C-F). In a similar way, using an immunohistochemical approach, we showed that the expression of Jag1 was also decreased in the mid-limb of the S-shaped body (data not shown). In parallel, we also monitored the levels of expression of Lhx1 and Pax2, genes that are known to be controlled by HNF1β in other kidney compartments (Lokmane et al., 2010) and that have been shown to play a role in kidney development (Dressler et al., 1993; Kobayashi et al., 2005). Our qRT-PCR analysis did not show any decreased expression of Lhx1 or Pax2 at E14.5 (Fig. 4K). At later steps, at E19.5, these two genes showed aberrant, persistently high expression in the abnormal S-shaped bodies of the mutant and in the rare and primitive tubular segments that were formed (Fig. 4G-J). In addition, qRT-PCR performed on newborn mutant pups showed clear upregulation of Lhx1 and Pax2 expression (Fig. 4K). The persistent expression of these genes suggests that, even if Hnf1b mutant embryos developed some short and primitive tubular components, they did not fully differentiate and still expressed markers of earlier steps of their developmental program.

It has been shown that Notch signaling plays a crucial role in the specification and differentiation of the most proximal structures of nephrons during development. Interestingly, mice carrying a hypomorphic allele of Dll1, one of the ligands of the Notch receptors, exhibit a phenotype reminiscent of that of our Hnf1b mutant embryos. In fact, the defective expression of Dll1 leads to drastically reduced formation of PTs (Cheng et al., 2007). Interestingly, Dll1, Notch intracellular domain (NICD) and Hes5 are detected in the epithelial cells that form the bulge at the bend between the mid and lower limb of the S-shaped bodies (Cheng et al., 2007; Piscione et al., 2004).

We investigated the possible involvement of defective Notch signaling activation in the onset of the phenotypes observed in Hnf1b mutant embryonic kidneys. We first assessed the expression of Notch2 and Jag1, key factors that have been shown to be controlled by HNF1α and HNF1β in the intestine (D'Angelo et al., 2010). qRT-PCR experiments showed that at E13.5 the expression of these two genes was similar in mutant and control embryonic kidneys (Fig. 5A). This level of expression was maintained at E14.5, with no differences between genotypes (Fig. 5B). By contrast, we saw significant downregulation of Dll1 accompanied by significant downregulation of Hes5, one of the effector target genes of Notch signaling (Fig. 5A,B).

To assess the functional implication of Notch signaling pathway activation in the Hnf1b mutant phenotype we examined the possible genetic interaction between Hnf1b and Notch2 in the specification and expansion of PTs. Our results showed that double heterozygous mice (Hnf1b^+/–; Notch2^+/–) at P0 had a modest but consistent reduction (∼20%) of PTs compared with control animals (either wild-type or single-heterozygous mice; supplementary material Fig. S2). In a similar way, we also tested the possible genetic interaction between Hnf1b and Pou3f3. In this case, however, PT extents were similar in double-heterozygous and control newborns (data not shown).

In agreement with previous studies, our ISH experiments showed that Dll1 is particularly strongly expressed in the epithelial bulge of control S-shaped bodies. We showed that this expression was drastically decreased in Hnf1b mutants (Fig. 5C,D). The defective expression of Dll1 in mutants could have been ascribed to the lack of formation of this particular segment. Interestingly, Dll1 is known to be expressed in earlier precursors, including vesicles and comma-shaped bodies, structures that apparently formed normally in our mutant embryos. When we assessed the expression of Dll1 in these earlier structures, we could clearly see that Dll1 was significantly downregulated in vesicles and comma-shaped bodies in mutant embryos (Fig. 5E-H). These results indicated that HNF1β is probably involved in the transcriptional activation of Dll1 in nephron precursors.

To investigate whether Dll1 could be directly controlled by HNF1β we assessed its direct binding in vivo by chromatin immunoprecipitation (ChIP) experiments on embryonic kidney chromatin. As a first step, we identified the position of the best candidate HNF1 binding sites using a previously described in silico approach (Verdeguer et al., 2010). This led to the identification of a particularly well conserved site 2 kb upstream of the transcription start site of the Dll1 gene together with a second site much farther away (302 kb downstream). Our results showed that both sites are bound in vivo (supplementary material Fig. S3). The combination of the ISH and ChIP results suggests that the absence of bulge formation in Hnf1b-deficient kidneys might be a direct consequence of the defective expression of Dll1 as a direct target of HNF1β already in early nephron precursors (vesicles and comma-shaped bodies).

Our in silico approach also led us to discover evolutionarily conserved HNF1 binding sites around the Osr2 and Irx1 genes. Our ChIP results showed that one site, detected close to Osr2 (24.1 kb upstream of the putative transcription start site) and three sites around Irx1 (14 kb, 65 kb and 66 kb upstream of the putative transcription start site), were significantly bound in vivo (supplementary material Fig. S3). Since expression of Osr2 and Irx1 is not detected in the deformed S-shaped bodies of Hnf1b mutants, one possibility is that their defective expression could simply be due to the absence of formation of their territory. However, we cannot rule out the possibility that the expansion of this crucial territory might also be controlled by HNF1β via the transcriptional activation of these two genes.

After the formation of S-shaped structures, the tubular components of the nephron normally proliferate and form their typical long tubules, including the PT, HL and DCT. In this respect, nephron spatial organization is characterized by the fact that the final portion of the HL ends with the macula densa, a tubular structure that is intimately connected with the vascular pole of the same glomerulus. The spatial organization of Hnf1b-deficient nephrons was characterized by a very short primitive tubule emerging from the glomerulus in continuation with the presumptive macula densa. Normally, the urinary pole of a glomerulus is located directly opposite (at 180 degrees) the vascular pole. In our mutant embryos, the absence of tubular expansion had probably led to a deformation of the glomeruli, such that the two poles were forced close to each other (Fig. 6A-C). In addition, the vascular tuft, instead of showing its typical constriction at its base (by the formation of the vascular pole), was characterized by an atypical distended conformation. In spite of this alteration, at the ultrastructural level the glomerular structures were globally normal, with podocytes and their foot processes surrounding capillaries with fenestrated endothelial cells (supplementary material Fig. S4). As mentioned above, mutant glomeruli had dilations of the Bowman's (urinary) space, forming glomerular cysts.

Altogether, these results showed that podocyte differentiation was unaltered but that the global architecture of the capillary tuft was distorted leading to a highly deformed global structure of the glomerulus.

Fetuses carrying heterozygous mutations of HNF1B present with a set of complex and extremely variable renal abnormalities ranging from kidney agenesis to hypodysplasia and/or cysts (Decramer et al., 2007; Heidet et al., 2010). These abnormalities can be linked to an impaired differentiation process in MM-derived structures due to the defective expression of HNF1β and/or its target genes.

In an attempt to compare the pathological features observed in MODY5 patients with those in our mouse model, we analyzed paraffin sections from a MODY5 fetus (Patient 1) carrying a complete genomic deletion affecting all exons of HNF1B. Histological analysis of kidney sections revealed the presence of specific focal areas characterized by clusters of glomeruli, without their typical surrounding set of cortical convoluted tubules (Fig. 6D). Immunofluorescence experiments confirmed that there were only very few LTL-positive tubular sections (Fig. 6F). In addition, some of these glomerular structures were dilated and their capillary tuft was spread, similar to what was observed in the mutant mouse embryos (Fig. 6E,G). In addition, the analysis of paraffin sections from a MODY5 child (Patient 2) carrying a c.232G>T, p.Glu78X mutation showed the presence of glomerular cysts arranged next to one another and that lacked the typical surrounding tubular structures (Fig. 6H). In addition, we again observed the spreading of glomerular tufts that was typical of mutant mouse embryos (Fig. 6I).

Patients who carry HNF1B mutations (MODY5 patients) suffer from complex developmental dysfunction of kidney morphogenesis, including kidney agenesis, dysplasia and cyst formation (Decramer et al., 2007; Heidet et al., 2010; Ulinski et al., 2006). Previous studies indicated that the expression of Hnf1b in the UB of mouse embryonic kidneys is absolutely required for the expression of several key genes, including Wnt9b, which encodes an essential secreted factor for the condensation of the MM (Lokmane et al., 2010) (our unpublished observations). The absence of HNF1β in the UB leads to a complete block of nephrogenesis before the typical MET.

Very little is known about the role of HNF1β during the first steps of nephrogenesis, once the gene is programmed to be turned on in the mesenchyme-derived epithelial structures. In the present study, using a Six2-Cre inactivation strategy, we showed that the expression of HNF1β is not required for the first steps of nephron formation in mouse. The absence of this transcription factor does not perturb the typical formation and segmentation of renal vesicles and comma-shaped bodies. The first visible consequence of HNF1β deficiency becomes evident only in S-shaped bodies. In these structures, the most proximal portion of the mid-limb is deformed and the bulge of epithelial cells that probably contains the precursors of the PT and HL is missing (Fig. 7). This defect could be linked to the decreased expression of Dll1, as a hypomorphic Dll1 allele can lead to defective expansion of renal tubules. We showed that Dll1 expression is decreased in earlier nephron precursors (vesicles and comma-shaped bodies) before the appearance of any deformation (visible only in S-shaped bodies). In addition, our genetic interaction studies showed a modest but consistent reduction of proximal tubular expansion in Notch2; Hnf1b double-heterozygous newborns. Finally, HNF1β protein binds to the transcriptional control regions of the Dll1 gene in vivo. All these considerations suggest that HNF1β might indeed play an important role in the expression of Dll1. It is worth noting that the phenotype of our Hnf1b mutant overlaps only partially with that induced by the deletion of Notch2 (Cheng et al., 2007). In fact, Notch2 mutant embryos completely lack the most proximal part of the nephron including the PT and glomeruli. In addition, HNF1β is required for the differentiation and expansion of DCTs, a process that does not require Notch2 (Cheng et al., 2007). Altogether, these considerations indicate that the transcription factor HNF1β is involved in the formation of all nephron tubular components.

Analysis of the phenotype of mesenchyme-specific Hnf1b-deficient kidneys showed that PTs and HL formation seems to derive from a specific subdomain of S-shaped bodies. This process, which is promoted via the activation of Notch signaling, leads to the specification of a set of cells, probably already in the comma-shaped body. Expansion in the number of these cells gives rise to a particular territory in the more proximal part of the mid-segment of the S-shaped body. This particular subdomain of the S-shaped body, which is organized as a protruding bulge of epithelial cells, is probably at the origin of the future expansion of the whole tubular component of the nephron between the glomerulus and the macula densa. At the molecular level, specific markers are expressed in this particular domain of the S-shaped body, such as Osr2 and Irx1. Interestingly, the downregulation of Irx1 in Xenopus leads to abnormal morphogenesis of the intermediate pronephric tubules (Alarcón et al., 2008). In parallel, Osr2 has been shown to be involved in proximal segment differentiation in Xenopus and zebrafish (Tena et al., 2007), but its role in mammalian kidney morphogenesis is probably redundant as its inactivation in mouse does not lead to any renal abnormality (Lan et al., 2004). Interestingly, Osr2 been shown to act as a repressor that negatively regulates the expression of Pax2 in zebrafish and Xenopus (Tena et al., 2007). This property might explain the persistently high expression of Pax2 observed in the primitive tubules of mutant embryos. Even if the defective expression of Irx1 and Osr2 does not play any direct causal role in our Hnf1b mutant mouse phenotype, the presence of evolutionarily conserved HNF1 binding sites that are bound in vivo suggests that they could constitute a direct transcriptional cascade that could participate in the differentiation and expansion of tubules in combination with the role played by HNF1β at later steps of differentiation (Fig. 7).

Hnf1b-deficient S-shaped structures can generate a glomerulus with podocytes that can reach maturity with the development of normal foot processes. This observation is somewhat compatible with the fact that HNF1β is not expressed in podocytes and their precursors. The only proximal nephron structure that expresses HNF1β is the Bowman's capsule. In this context, we observed some dilation of the Bowman's (urinary) space and the formation of glomerular cysts. This glomerulocystic phenotype is observed with variable penetrance in MODY5 patients who carry mutations in HNF1B in the heterozygous state (Bingham et al., 2001). Intriguingly, in mouse, the heterozygous inactivation of Hnf1b does not cause any significant phenotype unless mice carry compound heterozygosity for both Hnf1b and other transcription factors (Paces-Fessy et al., 2012). However, the complete deletion of Hnf1b in either the UB or in the mesenchyme or in already formed tubules gives rise to the same traits that are observed in MODY5 patients with variable penetrance. This indicates that, somehow, the haploinsufficiency in MODY5 patients might reduce the protein levels below a critical threshold. This could lead to the appearance of a phenotype in the UB leading to defective branching and MET (hypodysplasia). In addition, haploinsufficiency in already formed nephron precursors could lead to defective tubular expansion accompanied by glomerular cysts. These phenotypes are recapitulated in mouse embryo models in which it has been demonstrated that HNF1β controls the expression of crucial effectors of kidney morphogenesis. Interestingly, the inactivation of Hnf1b in nephron progenitors using a Wnt4-Cre line and another Hnf1b floxed allele gave rise to the same phenotype. This confirms the crucial role of HNF1β in nephron tubular specification (Heliot et al., 2013).

In conclusion, our studies have demonstrated that HNF1β plays an essential role in the generation of renal tubules. In particular, HNF1β is required for the formation of a specific mid-limb subcompartment (epithelial bulge) of S-shaped bodies, which is probably at the origin of the expansion of PTs and HLs via the activation of a set of crucial kidney development genes (Fig. 7).

We thank Moshe Yaniv for helpful discussions and critical reading of the manuscript; Professor Laurent Daniel (CHU Timone, Marseille, France) for providing images of histological sections of a kidney from a MODY5 patient; and Raphael Kopan for providing Notch2 mouse models. We thank Silvia Cereghini for sharing unpublished data. We are very grateful to the platform of Imagerie Cellulaire, Morphologie et Histologie and the animal house facilities of the Cochin Institute.