Paracrine and autocrine R-spondin signalling is essential for the maintenance and differentiation of renal stem cells

During kidney development, WNT/β-catenin signalling has to be tightly controlled to ensure proliferation and differentiation of renal stem cells. Here we show that the two signalling molecules RSPO1 and RSPO3 act in a functionally redundant manner to permit WNT/β-catenin signalling and their genetic deletion leads to a rapid decline of renal progenitors. By contrast, tissue specific deletion in cap mesenchymal cells abolishes mesenchyme to epithelial transition (MET) that is linked to a loss of Bmp7 expression, absence of SMAD1/5 phosphorylation and a concomitant failure to activate Lef1, Fgf8 and Wnt4, thus explaining the observed phenotype on a molecular level. Surprisingly, the full knockout of LGR4/5/6, the cognate receptors of R-spondins, only mildly affects progenitor numbers, but does not interfere with MET. Taken together our data demonstrate key roles for R-spondins in permitting stem cell maintenance and differentiation and reveal Lgr-dependent and independent functions for these ligands during kidney formation.


Introduction
Nephron endowment is a critical factor for renal health and low number of nephrons have been associated with chronic kidney disease and hypertension 1 . In mammals, nephrogenesis is restricted to the developmental period and involves a dedicated nephrogenic niche at the most cortical region of the forming kidney that fuels successive rounds of nephron production 2,3,4 . The nephrogenic niche consists of three independent progenitor populations that will develop into collecting ducts, stroma (interstitial cells) and nephrons. Nephron progenitors form a condensed cap around the tips of the branching ureter. This cap mesenchyme (CM) can be further subdivided into populations that represent cells of progressive differentiation status depending on the position along their migration trajectories around the ureteric tips. Analysis over the last decade have identified at least four different subpopulations: 1) CITED1 + /SIX2 + that are considered as "groundstate" progenitors; 2) CITED1 -/SIX2 + progenitors; 3) primed CITED1 -/SIX2 + /pSMAD + progenitors; 4) WNT4 + /LEF1 + pretubular aggregates (PTA). Once engaged, PTAs undergo a mesenchyme to epithelial transition (MET) to form epithelialized renal vesicles, which will further differentiate via comma-and S-shaped bodies into segmented nephrons.
The transformation of nephron progenitors into epithelialized nephrons is not an entirely cell-intrinsic process, but requires inductive signals from both the ureteric tip and stromal cells 4 . Of particular importance is WNT9b, a molecule released from the branching ureter that induces canonical WNT/b-catenin signalling, stimulates proliferation and thus ensures self-renewal of kidney progenitors. Accordingly, deletion of b-catenin leads to the loss of progenitor cells 5 . However, canonical WNT signalling is also required for MET 6 and transient activation of b-catenin in isolated progenitors induces epithelialisation 7,8,9 . How the balance between progenitor proliferation and differentiation is achieved is not well understood, but experimental evidence suggests that progenitor proliferation requires low levels of b-catenin activity, whereas genes that are involved in MET are activated in the presence of a strong canonical b-catenin signal 10 . In the context of nephrogenesis, a strong b-catenin response appears to rely on the activation of canonical BMP/SMAD pathway, as progenitor cells leaving the niche are positive for pSMAD and deletion of Bmp7 interferes with MET 11 .
WNT/b-catenin signalling is essential for many organ systems and multiple feedback mechanisms have been identified that control signalling strength at almost every level of the signal transduction pathway. WNT receptor availability at the cell membrane is controlled by RNF43 and ZNRF3, two trans-membrane E3 ubiquitin ligases that induce receptor endocytosis and thus negatively regulate WNT signalling. Their action is counteracted by R-SPONDINs (RSPO1-4), a family of secreted molecules that bind to the Gprotein-coupled receptors LGR4/5/6. Binding to LGRs permits R-spondins to interact with RNF43/ZNRF3 and suppress endocytosis of the WNT receptor complex, thus enhancing WNT signalling. 12 In this study, we investigated a potential role of the R-spondin/LGR axis in controlling renal stem/progenitor behaviour in vivo. We show that Rspo1 and 3 are required to maintain the pool of renal progenitors throughout development by supporting their proliferative capacity and preventing their apoptosis. Moreover, strong R-spondin signal is essential to allow nephron progenitors to engage in differentiation and undergo MET. RSPO1/3 achieve these functions by their ability to activate the WNT/b-catenin signalling pathway, a role that is primarily mediated in an LGR-independent manner.

R-spondins are dynamically expressed during kidney development
To understand the role of R-spondins during kidney development in mice, we first mapped the expression of the four members of this gene family using qPCR and in situ hybridisation analysis. Whereas Rspo2 and Rspo4 were undetectable in developing kidneys (Suppl. Fig.1A), Rspo1 and Rspo3 could be found as early as E10.5 within SIX2 + renal progenitors (Suppl. Fig. 1B and 13 ). Interestingly, Rspo3 marked only a proportion of SIX2 positive cells, suggesting this population to be heterogeneous already at this early age  Fig. 1C). Interestingly, by E18.5 Rspo3 was virtually absent from nephron progenitors, but strongly expressed in the cortical stromal compartment (Fig. 1B ii), indicating a shift of expression towards the stroma. Strong Rspo3 signal was also detected in stromal cells lining ducts of the renal papilla (Fig. 1B iii).

RSPO1 and RSPO3 are essential to maintain the pool of kidney progenitors
Lack of Rspo1 is compatible with life 14 and kidneys isolated from knockout mice showed no discernible abnormalities (data not shown). Mice carrying a constitutive deletion of Rspo3 die at E9.5 due to placental defects 15,16 . To overcome this early lethality, we employed a conditional allele for Rspo3 (Rspo3 fl/fl ) 17 in combination with a range of Creexpressing lines (Suppl. Fig. 2). Tamoxifen (Tam) induction at E11.5 in presence of a CAGG:CreER TM driver 18 efficiently abolished Rspo3 expression three days after induction (E14.5) (Suppl. Fig. 1A) and resulted in a mild reduction of progenitor cells (Suppl. Fig. 3B).
To test whether Rspo1 and Rspo3 may act in a functionally redundant manner, we induced deletion at E11.5 and analyzed the renal phenotype at E14.5 and E18.5 (CAGG:CreER TM ; Rspo1 -/-, Rspo3 fl/fl -from now on called DM). Depletion was efficient, and no compensatory up-regulation of Rspo2 or Rspo4 was detected in double mutant kidneys (Suppl. Fig. 1A).
Macroscopic observation at E18.5 showed severe renal hypoplasia in R-spondin DM, when compared to control littermates (Fig. 1C, D). Hematoxylin and Eosin (H&E) staining of kidneys at E14.5 indicated a reduced nephrogenic zone and a near-complete absence of nephrons (Fig. 1E). Molecular analysis confirmed a dramatic loss of SIX2 + progenitors and an absence of forming nephrons ( Fig. 2A). The remaining SIX2 positive cells appeared scattered, when compared to the condensed mesenchyme seen in controls. To test whether the reduction of progenitors was caused by defects in proliferation or apoptosis, we next quantified BrdU incorporation and TUNEL staining ( Fig. 2B-C; Suppl. Fig. 4). In order to measure early events of R-spondin deletion within progenitors, we allowed the first wave of nephrons to form, TAM-induced CRE activity at E12.5 and analysed samples 2 days thereafter. Quantification of BrdU labelled SIX2 + progenitors indicated a significant 40% reduction of proliferation in this compartment (P=0.0002) (Fig. 2B). In addition, a significant increase of apoptosis was observed (P=0.0319 for SIX2 + cells) (Fig. 2C). Stromal Rspo3 maintains nephron progenitors during late stages of kidney development Rspo3 is produced by both, stromal and nephron progenitors, but at later stages expression shifts to a predominantly stromal expression (compare Fig. 1Bi with 1Bii). To evaluate the contribution of stromally derived RSPO3, on kidney development, we specifically depleted its expression in this compartment using the constitutively active

RSPO1 and RSPO3 are required for mesenchyme-to-epithelial transition
To test the role of progenitor-released R-spondins, we next generated Rspo1 -/-; Six2:Cre; Rspo3 fl/fl mice (Six2:Cre-DM). In this strain, the only remaining R-spondin is released from the stromal compartment (Fig. 4A). At E18.5 Six2:Cre-DM embryos possessed hypodysplastic kidneys that histologically displayed defects in nephrogenesis and a complete absence of glomeruli ( Fig 4B). Histological and immunofluorescent analysis of The above phenotype is reminiscent of defects seen in mutants for Wnt9b, a signalling molecule that is released from the branching ureter. WNT9b activity has been shown to activate two classes of b-catenin dependent genes: Class II genes that require lower levels of b-catenin signalling and are found predominantly in nephron progenitors; Class I genes such as Wnt4 that depend on strong b-catenin activation and are highly expressed in pretubular aggregates 5,10 . In situ hybridisation (Ish) analysis revealed that loss of R-spondins did not impact Wnt9b expression (Fig. 4Di&ii), but dramatically reduced the expression of Class II (Bmp7, Tfrc, Crym1, Uncx4, Etv5 and Slc12a) target genes within progenitors ( Fig. 4D and Suppl. Fig. 5C). Interestingly, Tfrc was absent from mesenchymal cells, but maintained to be expressed in the ureteric epithelium suggesting that progenitorspecific R-spondin expression is dispensable for its activation in the collecting duct system.
The class I target Wnt4, a key regulator of MET 19 , was also undetectable within the nephrogenic zone, but remained expressed in medullary stromal cells (Fig. 4D ix&x).
Six2:Cre-DM kidneys showed a small reduction of nephron progenitors (Fig. 4C). To exclude the possibility that the lower number of progenitors interfered with the MET process, we employed the Wt1:CreER T2 strain (Fig. 5A), which in response to Tamoxifen activation, induced deletion primarily within nephron progenitors (Suppl. Fig. 2

and 6).
Wt1:CreER T2 -DM embryos kidneys contained large numbers of SIX2 + progenitors (Fig. 5B) that failed to differentiate. Indeed, expression of LEF1, a direct target and interaction partner of b-catenin that is also a marker of committed pretubular aggregates, was virtually absent from the nephrogenic region of Rspo1/3 double mutant mice ( Fig 5B). Six2:Cre and Wt1:CreER T2 driven deletion thus showed a very similar phenotype. The higher number of progenitors in WT1:CreER T2 induced mutants is likely due to less efficient deletion of Rspo3 and, as a consequence, persistence of slightly higher levels of R-spondin activity.
Commitment of renal progenitors to nephron differentiation involves canonical BMP7 signalling, which allows cells to fully respond to the WNT/b-catenin pathway and undergo MET 11 . In situ hybridization indicated a dramatic downregulation of Bmp7 in Six2:Cre-DM kidneys (Fig. 4D). Canonical BMP signalling induces phosphorylation of SMAD1/5 and we therefore evaluated the phosphorylation status of this signal transducer.
Whereas progenitors in control kidneys that engaged in MET were positive for pSMAD1/5 ( Fig. 5C), the nephrogenic zone in Six2:Cre-DM kidneys was completely devoid of staining. pSMAD staining in medullary stroma was not affected by this deletion. Taken together these data demonstrate that R-spondins are required for Bmp7 activation, which in turn permits priming of renal progenitors for MET.
LGRs are dispensable for MET R-spondins act by associating and internalising the ubiquitin E3 ligases ZNRF3/RNF43. Expression analysis using RNA-Scope demonstrated RNF43 to be specifically expressed within the branching ureter, whereas ZNRF3 showed a more widespread expression pattern, covering also the nephrogenic niche (Suppl. Fig.7). Rspondin interaction with ZNRF3/RNF43 is believed to be primarily mediated via binding to their cognate receptors LGR4-6 and mutations in Lgr4 or Lgr4/5 have been shown to have mild defects in kidney development 20,21,22 . To determine to what extent RSPO1/3 activity in kidney development depends on the LGR receptor family we mapped their expression using RNA-Scope analysis. Lgr4 showed the broadest expression pattern with mRNA detected in virtually all cell types of the developing kidney (Fig. 6A). Particularly strong staining was found in the distal part of comma shaped bodies that extended into the intermediate segment at the S-shaped body stage. In addition, we detected strong staining in medullary stromal cells that line the developing ureter. Lgr5 mRNA was virtually absent from the CM, but could be found within ureteric tips, a proportion of medullary stroma cells and the distal segment of S-shaped bodies, as previously reported using a lacZ reporter strain 23 . Lgr6 mRNA showed the most restricted expression pattern and could only be detected in PTAs of forming nephrons (Fig 6A).
Based on these findings we hypothesized that LGR4 was likely to be the main receptor mediating RSPO function in renal progenitors. To test this hypothesis, we took advantage of a mouse strain carrying an Lgr4 null allele that was previously generated by our laboratory 24 . Quantification of progenitor cells at E14.5 indicated a reduction of 37% (p=0.0006) in Lgr4 -/mutants compared to control kidneys (Fig. 6B, C), a cell loss that was considerably inferior to that observed in R-spondin mutants. Moreover, MET appeared to be unaffected in LGR4 mutants as exemplified by the activation of LEF1 in PTAs and the presence of WT1 positive cells in proximal proportion of forming nephrons (Fig. 6D). To test whether LGR5 and/or LGR6 may compensate for the lack of LGR4, we next analyzed wholebody Lgr4/5/6 triple knockout kidneys at E14.5, the latest developmental stage these embryos can be collected. The absence of LGR4/5/6 had little effect on early renal development and progenitors were able to undergo MET and form WT1+ glomeruli ( Fig   6E). We conclude that while LGR4 appears to mediate some R-spondin activity within the nephrogenic niche, the majority of signalling occurs in an LGR-independent manner.

DISCUSSION
Controlled proliferation and differentiation of renal progenitors is essential for the establishment of sufficient numbers of nephrons that ensure healthy kidney function throughout life. Our analysis has established R-spondins as novel key regulators that ensure the maintenance of a healthy pool of nephron progenitors throughout kidney development and permit MET during nephrogenesis (Fig. 7).
Consistent with their partially overlapping expression pattern, R-spondins appear to act in a functionally redundant manner in kidney development and single gene deletion of Rspo1 or Rspo3 had only mild effects on progenitor numbers (Suppl. Fig. 3). However, whereas Rspo1 is present in all SIX2 + cells, Rspo3 expression is restricted to a subpopulation that -based on their location at the outermost cortex -appear to represent 'ground-state' progenitors. Interestingly, by E18.5 Rspo3 expression is lost from the nephron progenitor compartment, which is likely to reflect the progressive aging/differentiation of these cells.
Indeed, age-dependent changes of nephron progenitors that affect their proliferation capacity and lifespan have been described, previously 25 . At late stages of embryogenesis, Rspo3 expression becomes almost exclusively restricted to stromal progenitors, which is consistent with recent single sequencing data at E18.5 that highlighted Rspo3 as a marker for the stromal compartment 26 . Stromal expression appears to be important for nephron progenitor maintenance, as Rspo3 deletion in this compartment resulted in progenitor loss at later stages of development. Renal progenitor cells are thus sandwiched between stromal cells releasing RSPO3 and the WNT9b producing ureteric bud that ensures their survival at later stages of development.
In previous studies, R-spondins, and in particular Rspo3, have been shown to enhance both canonical and non-canonical WNT signalling 27,28,29 . Our analysis demonstrates that upon R-spondin deletion direct downstream targets of b-catenin, such as Tafa5 5 and Axin2 30 are reduced in nephron progenitors indicating that in kidney development R-spondins activate canonical WNT signalling. Indeed, the observed phenotypes largely mimic those seen upon loss of b-catenin activity ( 5, 6, 9 and this study).
These findings are consistent with recent in vitro data that described Rspo1 to enhance canonical signalling upon treatment of a mesonephric cell line (M15) with WNT9b (Ref. 31 ).
The considerably different phenotypes observed in CAGG-Cre (or Foxd1-Cre) driven deletion of R-spondins, which impacted progenitor survival, and Wt1-Cre (or Six2-Cre) driven excision that interfered with MET, may at a first glance seem surprising. However, the ubiquitously expressed CAGG-CreER TM strain completely abolished R-spondin expression in the kidney and thus effectively blocked b-catenin signalling within progenitors. Since bcatenin signalling is essential for proliferation and survival, progenitor cell numbers rapidly declined in these mutants. By contrast, Six2:Cre and Wt1:CreER T2 induced deletions did not interfere with stromal Rspo3 expression, which bestowed sufficient b-catenin signalling in progenitors to permit survival. This hypothesis is compatible with a model in which low and high levels of b-catenin signalling regulates Class II and Class I target genes, respectively 10 . However, the concept of lower b-catenin signalling in uncommitted progenitors seems contradictory considering that they are exposed to higher levels of Rspondins (RSPO1+RSPO3), when compared to cells that engage in MET (only RSPO1). This dilemma can be resolved when introducing an additional "switch" that permits the activation of class I targets (e.g. Wnt4/Lef1) once progenitors leave the cortical niche.
Evidence from knockout studies suggests this switch to involve Bmp7-dependent activation of pSMAD signalling at the boundary between the niche and committed cells 11 . Since progenitors in Six2:Cre-DM/Wt1:CreER T2 -DM mice have significantly decreased levels of Bmp7 (a class II target of b-catenin 9 ), progenitor cells fail to activate pSMAD signalling, lack expression of class I targets and as a consequence do not epithelialize.
We have also addressed the potential role of LGR4/5/6, the cognate receptors of Rspondins, in conveying their action. Lgr4 single mutants have been described previously to show dilated collecting ducts and reduced kidney size 20 that could be traced back to a reduction in renal progenitors 21 . Our own analysis with an independent Lgr4 knockout allele confirms this phenotype, although duct dilation was only seen at late stages of kidney development ( 24 and data not shown). In R-spondin mutants we did not observe dilated collecting ducts, which might be explained by persistent expression of Rspo3 in medullary stroma using the tissue specific deletion in stromal or nephron progenitors. Lgr5 was found to be virtually absent from uninduced progenitors, but was strongly expressed within the ureteric tip and the distal part of the developing nephron, two sites of strong b-catenin activity. To our knowledge, no renal abnormality have been associated with Lgr5 mutations and the previously published Lgr4/Lgr5 double mutants show a kidney phenotype comparable to the Lgr4 single mutant 22 . Our RNA-Scope analysis revealed Lgr6 expression to be highly restricted to the pretubular aggregates and renal vesicle. Surprisingly, Lgr4/5/6 triple knockouts are associated with only a mild loss of renal progenitors that are capable to undergo MET, a phenotype that does not phenocopy the dramatic loss of renal progenitors and their inability to undergo MET correlated with the absence of Rspo1 and Rspo3. Due to the embryonic death around E14.5, later stages could not be analysed. Taken together these data suggest that Lgr4, but not Lgr5/6, contributes to the transduction of R-spondin signalling. More importantly, the mild phenotype observed in triple mice when compared to Rspo1/3 mutants, demonstrates that R-spondins can act in an LGRindependent manner during kidney development. A similar LGR-independent action has recently been described for Rspo2 in limb and lung development 32 . Interestingly an alternative mode of action for RSPO2 and RSPO3 has recently emerged that potentiates the WNT/b-catenin pathway through an interaction with heparan sulfate proteoglycans (HSPG) 33 . In Xenopus, RSPO3 has been shown to interact with Syndecan4 to induce noncanonical (PCP) signalling 29 and Syndecan1 has been suggested to present WNT and Rspondins in multiple myeloma 34 . Further analysis will be needed to identify the proteins responsible for mediating R-spondin action in the kidney.
The number of nephrons varies dramatically in the human population with a count of 250000 and 2.5 million glomeruli per kidney being considered as normal. However, lower nephron numbers have been directly linked with hypertension and a higher risk of developing renal diseases 1 . Our study has identified R-spondins as regulators that maintain the nephrogenic niche during development and we can speculate that changes in expression levels or protein function in these genes may be associated with renal disorders.
Interestingly, the RSPO3 locus has been identified in two studies to be linked with renal diseases including abnormal blood urea nitrogen (BUN), a hallmark for glomerular filtration dysfunction 35,36 . Given these studies and our findings mice, it will be of interest to further investigate a potential role of R-spondin variants in the predisposition to renal disease.

Mice
The experiments described in this paper were carried out in compliance with the French and international animal welfare laws, guidelines and policies and were approved by the local ethics committee (PEA N°: NCE-2014-207 and PEA N°: 2018060516474844 (V2)).

Embryo collection
Embryos were collected from time-mated females considering the date when the vaginal plug was observed as embryonic day 0.5 (E0.5). Embryos from E11.5 to E18.5 were fixed with 4% paraformaldehyde (PFA) in phosphate buffer saline (PBS) overnight at 4°C or at room temperature, if used for RNAScope analysis. The day after, embryos were washed in PBS, dehydrated through an Automatic tissue processor (Leica TP1020) and embedded in paraffin.

In situ hybridization
Tissues were fixed overnight in 4% paraformaldehyde, progressively dehydrated and embedded in paraffin. 7 µm thick sections were cut then rehydrated and hybridization was  Table 1 and incubated overnight at 4°C. The following day, after 3 washes in PBS, secondary antibodies were diluted 1/500 in PBS and applied on sections for 1H at room temperature. After 3 washes in PBS, sections were mounted in a 50% glycerol medium.

Immunofluorescence and histological analysis
For immunofluorescence experiments, tissues were fixed overnight in 4% paraformaldehyde, progressively dehydrated and embedded in paraffin. 5 µm thick sections were rehydrated, boiled in a pressure cooker for 2 min with Antigen Unmasking Solution (Vector laboratories) and blocked in PBS solution containing 10% normal donkey serum, 0.1% tween and 3% BSA. All antibodies were applied overnight at 4 °C at the concentrations listed in Table 1. Secondary antibodies were diluted 1:500 and applied at room temperature for 1 h. For histological analysis 5 µm thick sections were stained with haematoxylin and eosin according to standard procedures.

Quantitative RT-PCR analysis
RNA was extracted from embryonic samples using RNeasy Mini-or Micro-kit (Quiagen), following the manufacturer's instructions. Reverse transcription was performed using M-MLV reverse transcriptase (Invitrogen) in combination with Random Hexamers (Invitrogen). The cDNA synthesised was used as a template for qPCR reaction performed using the Light Cycler® SYBR Green I Master Kit (Roche). Expression levels were normalized for Sdha. Primers are described in Table 2.

Quantification of proliferating or apoptotic renal progenitor and stromal cells
A combination of anti-BrdU and anti-SIX2 antibodies were used to detect proliferating renal progenitors. Apoptotic cells were labeled with TUNEL kit (Roche) and renal progenitor and stromal cells identified using anti-SIX2. Results were obtained after scoring the percentage of BrdU or TUNEL positive nuclei/total SIX2+ or MEIS+ cells in 10 sections collected throughout the entire embryonic kidney.

Quantification of renal progenitors
SIX2 + progenitors were counted in three consecutives sagittal sections of kidney samples through the renal pelvis and the average number was reported on the graph.

Statistical analyses
Data are shown as mean±s.e.m. Analyses were performed according to the two-tailed unpaired Student's t-test, *p < 0.05 **p < 0.01, ***p < 0.001, ****p<0.0001. The letter "n" refers to the number of individual samples. Multiple comparisons tests were performed according to one-way ANOVA.       Immunofluorescence (IF) staining with anti-SIX2 antibodies was achieved on same sections to identify renal progenitors (red). ISH and IF signals were overlaid (green and red respectively).
Expression patterns within the nephrogenic niche are schematized to the right of the panel.
Note the persistence of Rspo3 expression in stromal cells upon progenitor specific deletion