FGF8 induces chemokinesis and regulates condensation of mouse nephron progenitor cells

Mesenchymal condensation is an essential step in kidney development for the early formation of nephrons. This mechanism consists of reciprocal interactive signalling between mesenchymal cells and their surroundings, the epithelial and stromal cells (Das et al., 2013; O'Brien, 2019; Oxburgh, 2018). In addition to reciprocal signalling, intercellular interactions, cellular morphogenesis, i.e. apoptosis or adhesion, and cell migration play an essential role during the establishment of mesenchymal condensation (Ribatti and Santoiemma, 2014; Scarpa and Mayor, 2016; SenGupta et al., 2021). Cell migration can be influenced by chemical, thermal, galvanic, electrical, gravitational or mechanical stimuli, or combinations of these phenomena. A stimulus can cause a tactic response, in which cell movement is directed to the location of the stimulus, or a kinetic response, i.e. random locomotion, in which the magnitude of the response depends on the intensity of the stimulus (Diehn et al., 1977). Particularly in the presence of chemical gradients, cells can show strong chemotactic or chemokinetic responses.

In mice, around embryonic day (E) 11-11.5, mesenchymal condensation in the nephrogenic niche of the developing kidney results in the formation of the cap mesenchyme (CM) (O'Brien, 2019; Davies, 2016). At the same time, nephron progenitor cells (NPCs) from the CM migrate in a stochastic fashion between the top (or tip region) of the epithelial ureteric bud (UB) and the bottom (or trunk region) (Fig. 1) (Combes et al., 2016; Lindström et al., 2018a; Lawlor et al., 2019). NPC fate is niche region specific and requires reciprocal signals between the UB and the surrounding mesenchymal and stromal cells (Fig. 1) (Das et al., 2013; O'Brien, 2019; Lawlor et al., 2019; Carroll et al., 2005; Reginensi et al., 2013; Brown et al., 2013; O'Brien et al., 2018; Li et al., 2021). NPCs that are located in the tip region of the UB maintain their progenitor state and are thus called true nephron progenitor cells (tNPCs) (Fig. 1) (Lindström et al., 2018a,b; Brown et al., 2013). NPCs that migrate downwards to the trunk region are further primed by factors from the UB, becoming committed NPCs (cNPCs) (O'Brien, 2019; Carroll et al., 2005; Brown et al., 2013; O'Brien et al., 2018). A fraction of the cNPCs in the trunk region starts to form a pretubular aggregate (PTA), initiating nephron formation (O'Brien, 2019; Combes et al., 2016; Lawlor et al., 2019; Carroll et al., 2005; Stark et al., 1994). The regulation of NPC fates, migration and priming have been studied intensely (O'Brien, 2019; Oxburgh, 2018; Mari and Winyard, 2015), but the mechanism underlying the condensation of NPCs to the UB is not yet understood. Various signalling factors, receptors and extracellular matrix molecules have been suggested to play a role in NPC condensation (Fig. 1), but its key regulators remain elusive (Combes et al., 2016; Trueb, 2011; Mathew et al., 2012; Kuure and Sariola, 2020).

Fibroblast growth factors (FGFs) are a family of signalling proteins that govern different aspects of kidney development, including UB branching and maintenance of NPCs (Walker et al., 2016). Deletion or mutations in either FGFs or their receptors (FGFRs) can lead to either kidney agenesis or disorders (Walker et al., 2016). FGF8 is expressed in the mesenchyme and is required for both the regulation of downstream genes involved in PTA formation and cell survival (Carroll et al., 2005; Perantoni et al., 2005; Grieshammer et al., 2005; Huh et al., 2020). Deletion of Fgf8 from kidney primordia leads to a lack of mature nephrons, and eventually to lethality within 24 h of birth (Perantoni et al., 2005). The failure of nephron maturation has been attributed to the lack of expression of Wnt4 and Lim1 (Lhx1), both of which are crucial for mesenchymal-epithelial transition (MET) (Perantoni et al., 2005). However, culturing isolated MM cells from kidneys lacking Fgf8 along with a WNT source (embryonic spinal cord) failed to initiate nephrogenesis (Perantoni et al., 2005; Grieshammer et al., 2005), supporting the notion that WNT4 and FGF8 work independently. Furthermore, when ectopic FGF8 was added in combination with a WNT source, MMs lacking Fgf8 expression formed PTAs (Perantoni et al., 2005). These results indicate that FGF8 enhances WNT4 expression in PTAs but also suggest that FGF8 and WNT4 work independently. Because little is known about the specific role of FGF8 during cap mesenchyme formation, we further characterize its function and show that FGF8 signalling co-regulates both NPC migration and mesenchymal condensation.

When the kidney develops from the posterior intermediate mesoderm (Costantini and Kopan, 2010), the brachyury/T gene is required for the formation of the posterior mesoderm and axial development (Clements et al., 1996). Hence, brachyury/T^Cre-mediated deletion leads to the deletion of Fgf8 in both epithelial and mesenchymal compartments of the developing kidney (Perantoni et al., 2005; Costantini and Kopan, 2010; Clements et al., 1996). To examine more closely the involvement of FGF8 in kidney development, we stained mutant kidneys (Fgf8^n/c;T^Cre) with SIX2, a known NPC marker (O'Brien, 2019; Oxburgh, 2018).

In Fgf8^n/c;T^Cre mutant kidneys, we found that the Six2⁺ cell population was diminished and less condensed than in the controls (Fig. 2A,B). The reduced cell number has previously been attributed to increased cell death in Fgf8-deficient kidneys (Perantoni et al., 2005). We hypothesized that the absence of FGF8 signalling additionally leads to decreased cell motility and consequent failure of mesenchymal condensation, and eventually to the termination of nephrogenesis. To test this hypothesis, we first investigated whether the failure of Six2⁺ cells to condensate would also occur in an in vitro culture assay. We used the Trowell culture method to culture Fgf8^n/c;T^Cre kidneys and littermate control kidneys. After 3 days of culture, corresponding to E16.5, we found that the Six2⁺ cells in mutant kidneys were significantly less condensed when compared with their littermate controls (Fig. 2C,D).

More specifically, we found that the CM in mutant kidneys was on average twice as thick as in the controls (Fig. 2E). In accordance with this observation, the Six2⁺ cells in mutant kidneys were strongly dispersed within the niche (Fig. 2F). We therefore wondered whether the number of Six2⁺ cells that are attached to the UB was also affected as a result. Attachment to the UB has previously been suggested to affect the differentiation capacity of mesenchymal cells (O'Brien et al., 2018). We classified Six2⁺ cells as being attached to the UB when their centroid position was closer to the UB tip than a threshold value of 17.4 µm, which corresponds to twice the median position of Six2⁺ cells that we measured in the controls (Fig. 2F). Our choice of threshold value agrees well with the upper quartile of distances that were previously ascribed to the attached state (15.2 µm; Combes et al., 2016). We also found that the percentage of attached cells in our controls (92%, Fig. 2G) was in good agreement with what was previously found (85%, compare with Combes et al., 2016). Finally, we found that in our Fgf8-deficient mutant kidneys, the proportion of Six2⁺ cells that was attached to the UB tip was significantly lower than in the controls (Fig. 2G). Thus, as a next step, we wanted to investigate the effect of FGF8 on the Six2⁺ NPC population in more detail.

The Six2⁺ NPCs dominate the CM population around the UB (Brunskill et al., 2014), and when these cells are primed as cNPCs, a subset of these cNPCs forms a PTA (O'Brien, 2019; Oxburgh, 2018). In an exhaustive study of secreted FGF family members that are required for the maintenance of the tNPC state, it has been suggested that FGF8 fails to maintain the true progenitor state of NPCs (Brown et al., 2011). However, this study made use of a 2D monolayer culture, which differs from the 3D in vivo microenvironment of the developing kidney. Ihermann-Hella et al. and Dapkunas et al. have since developed a 3D culture system for NPCs (Ihermann-Hella et al., 2018; Dapkunas et al., 2019). Much as in the 2D cultures, MM cells form aggregates when cultured with FGF2 in the 3D cultures (Ihermann-Hella et al., 2018; Perantoni et al., 1995). To test whether NPCs would also condense in response to FGF8, we cultured the dissociated NPCs from E11.5 kidneys in a 3D matrix along with FGF2 (positive control), FGF8 or anti-FGF8 antibody to block any FGF8 secreted by MM cells. The chosen monoclonal anti-FGF8 antibody binds selectively to FGF8 in in vitro experiments (Fig. S1). From dissected E11.5 kidneys, MM was separated from the UB, dissociated into single cell suspension, seeded in Matrigel and cultured for 24 h.

In response to both FGF2 and FGF8, the NPCs formed condensates and retained Six2 expression (Fig. 3B,C), while both the control and NPCs treated with anti-FGF8 antibody lost Six2 expression (Fig. 3A,D; Movie 1). To confirm that the NPC population that is positive for SIX2 is also expressing another early marker for condensation, we looked at PAX2, which is a marker for progenitors and early nephron precursors (Dapkunas et al., 2019). Anti-PAX2 antibody stained the NPC condensates in the presence of FGF8 ligand but no staining was observed in the condensates without FGF8 ligand or with anti-FGF8 antibody (Fig. 3F-H). As previously reported by Grieshammer et al. (2005), the absence of FGF8 leads to cell death; a similar observation was made in nephrospheres lacking FGF8 when compared with ectopic FGF8 nephrosphere assays, and a similar result is observed for the live cell number (Fig. 3E). This suggests that FGF8 plays an additional role in NPC commitment along with its role in the formation of NPC condensates. To examine the differentiation stage of NPCs between tNPCs and cNPCs, a qPCR analysis of NPCs in the presence or absence of FGF8 was carried out. tNPC markers such as Cited2, Six2 and Eya1 were maintained by ectopic FGF8 (Fig. 3I). To confirm that the results were not influenced by signals from the UB, the same experiment was carried out with fully dissociated MMs lacking UB. Similar results were obtained with retained tNPC markers when treated with ectopic FGF8 (Fig. 3J). It has been reported that crosstalk between WNT and FGF signalling pathways is linked by modulation of phosphorylation of Gsk3b (Katoh and Katoh, 2006), but we did not observe this in our setting (Fig. S2). Finally, in wild-type kidneys that were cultured in the presence of ectopic FGF8 (Trowell culture method), we observed an expansion of the Six2⁺ population (Fig. 3K). In conclusion, our 3D culture matrix experiments suggest that FGF8 is required for both NPC condensation and tNPC maintenance.

FGFs have been shown to act as chemoattractants that trigger a chemotactic response (Makarenkova et al., 2009; Bae et al., 2012), i.e. the migration of a cell along the concentration gradient of the chemoattractant. To determine how FGF8 affects NPC motility, we placed a FGF8-soaked bead in a 3D Matrigel matrix containing MM cells from E11.5 kidneys and tracked cells (Movie 2). Cell tracking revealed that MM cells migrated significantly faster compared with the control bead experiments, but cellular motion was stochastic and lacked directionality (Fig. 4; Movie 2). This shows that FGF8 has mainly a chemokinetic effect, i.e. an impact on the speed of movement, rather than a chemotactic effect.

Here, we note that we did not measure the FGF8 concentration profile and thus cannot exclude the possibility that meaningful gradients did not emerge due to rapid dispersion of FGF8 in Matrigel. Previous measurements in zebrafish showed that the diffusion coefficient of FGF8 is high in aqueous environments (Yu et al., 2009). However, a slow-moving fraction of FGF8 with a reduced diffusion coefficient has been shown to persist due to interactions with heparan sulphate proteoglycans (HSPGs), which are also present in Matrigel and have been shown to play a crucial role in the extracellular distribution of growth factors, modulating morphogen signalling and transport (Makarenkova et al., 2009; Yan and Lin, 2009; Matsuo and Kimura-Yoshida, 2014; Stapornwongkul and Vincent, 2021; Krishnan Harish et al., 2022 preprint).

Ultimately, we find that the measured cell speeds in our FGF8 bead assays (4.6±0.5 µm/h; Fig. 4N) are comparable with those measured by others in the UB tip region (Combes et al., 2016; Tikka et al., 2022). Interestingly, in the presence of FGF8-soaked beads, most cells formed large aggregates, resulting in a collective movement resembling swarm behaviour, which is reflected in the bifurcation of the track straightness measurements of FGF8 versus control (Fig. 4E,F; Movie 2). In the control bead experiments, only a few small aggregates were formed, presumably owing to low levels of Fgf8 expression by some MM cells.

In summary, we find that FGF8 triggers a chemokinetic response of MM cells from E11.5 kidneys. This observation of undirected cellular motility agrees with previous quantifications where MM cells were found to move semi-stochastically and in a swarming-like manner (Combes et al., 2016).

We next decided to test the impact of FGF8-induced chemokinesis on NPC condensations at the UB using computational modelling. UB outgrowth starts at around E10.5 when the metanephric mesenchyme is in a diffuse thickened state, inducing the patterning of the MM (Davies, 2016; Carroll et al., 2005; Perantoni et al., 1995; Mugford et al., 2008; Xu et al., 2014; Munro et al., 2017). At around E11, diffuse weak expression of Fgf8 by MM cells coincides with the emergence of a well-defined cap mesenchyme (Davies, 2016; Carroll et al., 2005; Munro et al., 2017). Close to the UB, MM cells are rather immotile (Combes et al., 2016). Interestingly, sonic hedgehog (SHH), a repressor of Fgf8 expression, is secreted from UB cells during early nephrogenesis (Yu et al., 2009; Lin et al., 2001). The SHH gradient emanating from the UB likely results in the lower Fgf8 expression that is observed closer to the UB (Chen et al., 2016). A gradient of autocrine FGF8 signalling and thus chemokinesis would be in agreement with the previous observation that cell speeds increase with distance from the UB (Combes et al., 2016). In the same study, it was found that MM cells experience a subtle attraction towards the UB, indicating the presence of a UB-secreted chemotactic factor. WNT11 represents a likely candidate, as it is secreted already around E10.5-E11 by UB tip cells and is required for stable NPC-UB attachment (Carroll et al., 2005; O'Brien et al., 2018; Kispert et al., 1996; Uchiyama et al., 2010).

To analyse the interplay of chemical signalling and cell motility during mesenchymal condensation, we asked whether a model consisting of (1) FGF8-induced mesenchymal cell motility, (2) WNT11based chemoattraction and (3) SHH-induced repression of FGF8 close to the UB can explain mesenchymal condensation around the UB (Fig. 5A). In the computational model, NPCs are initially randomly positioned in a niche bordering a flat patch of the ureteric epithelium (Fig. 5E). The NPCs are assigned velocities that depend on the local FGF8 concentration, while the direction of movement is chosen randomly (Fig. 5A,B). Epithelial cells are secreting weak concentrations of both WNT11 and SHH, while scenarios of different levels of Fgf8 expression by MM cells are explored (Fig. 5F,G; Fig. S3). Mechanically, all cells can adhere to each other when being closer than a threshold distance and detach when moving apart. Simulating this model results in an effective motility gradient with a trap-like region close to the epithelium (Fig. 5C,D,F; Fig. S3). At weak concentrations of FGF8, WNT11 and SHH, only the MM cells closest to the ureteric epithelium show strong displacement towards the ureteric epithelium, but most cells that are further away remain within a few cell diameters of their initial positions (Fig. 5C,F; Movie 3). At intermediate concentrations of FGF8 (twofold increase), most cells aggregate close to the basal surface of the ureteric epithelium (Fig. 5C,D,G). At higher FGF8 concentrations (threefold increase), cells aggregate everywhere in the niche, which is in line with our bead experiments (Fig. 5C; Movie 2, Movie 3). Last, we also observe FGF8 concentration peaks at locations where cells aggregate, resulting in swarm-like motility (Movie 3).

We conclude from these simulation results that the chemokinetic effect of FGF8 enables the niche-wide distribution of NPCs. This allows them to reach the vicinity of the UB and also to enter the sphere of influence of epithelial factors that support the immobilization of NPCs. The corresponding motility gradient that appeared in the simulations (Fig. S3) is in agreement with experimental observations (Combes et al., 2016). The simulations also show that excess FGF8 can override the guidance of epithelial signalling and prevent mesenchymal condensation.

Deletion of Fgf8 before gastrulation is lethal as cells lose the ability to migrate away from the primitive streak (Perantoni et al., 2005; Sun et al., 1999). Deletion of Fgf8 in the MM using Pax3^Cre mice leads to the same phenotype, and new-born mice die shortly after birth (Grieshammer et al., 2005). We wondered whether FGF8 also plays a role in later stages of kidney development. In situ hybridization of Fgf8 revealed its expression in upper cells forming PTAs and this expression was still maintained in the top cells of renal vesicles that form the future comma and S-shape bodies in wild-type kidneys (Fig. 6A-E). To investigate its effect during early nephron formation at PTA stage, we used later expression tissue-specific Cre mouse lines (see Materials and Methods). Cre was under the promoter of Wnt4 or Pax8 genes. By using this strategy, we generated mutant Fgf8^n/c progeny with 50% frequency, while Fgf8^n/+ mice were used as littermate controls. First, to be sure that Fgf8 did not have any function in the UB, we first deleted Fgf8 from the UB using HoxB7^Cre mice; as expected, the mutant had no phenotype and the kidneys were similar to those of the wild-type mice (Figs S4C and S5A-D).

Second, we deleted Fgf8 from MM cells by employing two tissue-specific Cre recombinase mice (under the promoter of Pax8 and Wnt4 genes). Pax8^Cre is expressed in both the MM and the UB tip (GUDMAP; Fig. S4A) (Bouchard et al., 2002; Harding et al., 2011), and Wnt4^Cre is expressed only in the MM (Fig. S4B; Shan et al., 2010). We found that the Fgf8 deletion from the MM using either Pax8^Cre or Wnt4^Cre led to smaller kidneys when compared with littermate controls (Fig. 7A-C). On closer inspection, we found that in both cases kidneys had fewer mature nephrons when compared with littermate controls. Kidneys of Fgf8; (Pax8^Cre) revealed an arrest in S-shaped-body structures whereas the Fgf8; (Wnt4^Cre) showed comma-shaped body structures (Fig. 7D-I, higher magnification). Because the complete loss of FGF8 function during kidney development results in a failure of nephron formation around the S-shaped body stage (Perantoni et al., 2005; Grieshammer et al., 2005), these results raised the issue of whether there is still some Fgf8 expressed in the Fgf8^n/c;Pax8^Cre and Fgf8^n/c;Wnt4^Cre kidneys. Therefore, we analysed the Fgf8 expression in Fgf8^n/c;Pax8^Cre kidneys. Functional Fgf8 RNA is expressed at a lower level than in the control litter analysed by qPCR, suggesting that the remaining Fgf8 expression causes a hypomorph kidney phenotype. We observed levels of Fgf8 expression that were around 40% of controls in E12.5 kidneys lacking Fgf8 (Fig. 7J). These data demonstrate that FGF8 is required within the developing kidney to support the further development of the nephrons, and the reduced Fgf8 expression during nephrogenesis induces hypomorph phenotypes.

tNPCs in the tip region of the UB are marked by Six2 expression along with Cited1, Cited2 and Eya1 (Fig. 1) (O'Brien, 2019; Mugford et al., 2009; Brown et al., 2015; Tanigawa et al., 2016). Wnt9b that is expressed and secreted by the UB regulates the transition from tNPCs to cNPCs (Fig. 1) (Carroll et al., 2005). This transition is marked by a decrease of Cited1 and Cited2 expression (O'Brien, 2019; Brown et al., 2013; Mugford et al., 2009). At the same time, the expression of Wnt4 in an aggregated subset of cNPCs that are located at the tip-trunk interface of the UB indicates the onset of nephron formation (Fig. 1) (Lawlor et al., 2019; Stark et al., 1994). When we compared the expression of Six2 in wild-type kidneys with tissue-specific deletions of Fgf8 in Fgf8^n/c;Pax8^Cre (MM and UB tip) or Fgf8^n/c;Wnt4^Cre (MM only) kidneys, we found an untypical Six2⁺ expression patterns, indicating disorganized NPCs (Fig. 8A-C). Deletion of Fgf8 in Fgf8^n/c;Pax8^Cre kidneys did not seem to alter the expression of Wnt9b when compared with the littermate controls (Fig. 8D-F) and, as expected, Wnt9b was not affected in Fgf8^n/c;Wnt4^Cre kidneys. On the other hand, the expression of Wnt4 was decreased in both Fgf8^n/c;Pax8^Cre and Fgf8^n/c;Wnt4^Cre kidneys (Fig. 8H,I) as compared to the littermate controls (Fig. 8G). Further, in both Fgf8^n/c;Pax8^Cre and Fgf8^n/c;Wnt4^Cre kidneys, condensation of NPCs expressing Eya1 and Six2 around the UB was perturbed, while the expression of Cited1 was still maintained in the Fgf8^n/c;Wnt4^Cre kidneys but not in the Fgf8^n/c; Pax8^Cre kidneys (Fig. 8K,L,N,O). Together, these data show that these anchor genes for the NPC compartment have an incorrect pattern where Fgf8 is deleted via either Pax8^Cre or Wnt4^Cre mouse lines.

To further confirm the results obtained from our in vitro experiments (cultured Fgf8^n/c;T^Cre kidneys and the effect of FGF8 on MM cells in the bead culture experiment), we stained NPCs and UB cells using fluorescent markers on Fgf8^n/c;Pax8^Cre and Fgf8^n/c;Wnt4^Cre mice that delete Fgf8 later in nephrogenesis (after E11.5). To analyse the complete PTA formation, we have selected the E16.5 stage where most of the PTAs are already formed. Staining of Six2⁺ NPCs in E16.5 kidneys obtained from crossing Fgf8^n/c and Pax8^Cre revealed a thicker cap mesenchyme at the tips of UBs, suggesting differences in niche composition when compared with littermate controls (Fig. 9A,B). A thicker layer of Six2⁺ cells in the tip region of the UB indicates that NPCs either failed to fully condense around the UB or that they were not primed as cNPCs (Fig. 9C). Similar results were obtained when Fgf8 was deleted using Wnt4^Cre (Fig. 9D,E). With Wnt4^Cre-mediated deletion of Fgf8, the Six2⁺ population failed to condense around the UB tip, as seen after Pax8^Cre-mediated deletion of Fgf8 (Fig. 9F). LIM-class homeodomain transcription factor 1 (LHX1) is a known crucial marker for nephron patterning and maturation (Kobayashi et al., 2005; Liu et al., 2018). In E16.5 wild-type and Fgf8^n/c;Wnt4^Cre kidneys, LHX1 stained RV, CB and SB structures, whereas in Fgf8^n/c;Pax8^Cre kidneys the expression was only found in RV and CB, demonstrating that nephron maturation is delayed (Fig. S6A-D). These results confirm the observations from the in situ hybridization experiments (Fig. 8), suggesting that the Six2⁺ NPCs accumulate around the UB tip and that NPC induction is impeded, inducing a delay of nephron development.

Intercellular signalling between NPCs and UB cells is key for mammalian kidney development, and it is known to be tightly controlled. Failure of the expression of key genes such as Six2, Fgf8, Wnt4, Wnt9b and others leads to developmental defects or even to embryonic lethality (O'Brien, 2019; Oxburgh, 2018; Stark et al., 1994; Perantoni et al., 2005; Grieshammer et al., 2005). FGF8 is also known to be involved in the activation of the kidney-specific genes Wnt4 and Lim1 (Lhx1) (Perantoni et al., 2005; Huh et al., 2020).

In this work, we have used several cre-based mouse lines to establish that Fgf8 expression is located in the MM and imparts its function on the nephron progenitor cells. Deletion of Fgf8 from the MM resulted in embryonic kidneys that lacked mature nephrons, which led to smaller hypomorphic kidneys and postnatal death. As Fgf8-deficient kidneys lack Wnt4 expression, Fgf8 is required for Wnt4 expression (Perantoni et al., 2005; Grieshammer et al., 2005). Although it is known that WNT4 initiates MET (Stark et al., 1994), in accordance with previously published results, we also observed that, independently of WNT4, FGF8 is required for the condensation of NPCs (Perantoni et al., 2005). We found that in kidneys where FGF8 signalling was blocked using an anti-FGF8 antibody, NPCs did not condense to form pretubular aggregates. But upon removal of the antagonizing agent, PTA formation was recovered. This indicates that, even though FGF8 is upstream of WNT4, it regulates NPC condensation, which is itself required for PTA formation before WNT4-induced MET of PTAs.

FGFs are also involved in the maintenance of NPCs during early nephrogenesis (Brown et al., 2011). Although previously only a modest effect of FGF8 on cap marker transcription was observed in 2D cultures (Brown et al., 2011), our results with 3D cultures show that without FGF8, Six2 expression is lost, but when ectopic FGF8 is added and in the presence of the UB, the expression of Six2 is maintained when compared with the vehicular control. Similarly, the expression of other tNPC markers, such as Cited1, Cited2 and Eya1 is also maintained in the presence of FGF8. We explain the discrepancy with the previous observations through improved culture conditions, as it has been shown that culturing NPCs in a 3D micro-environment leads to an improvement in nephrogenic potential (Li et al., 2016). In conclusion, FGF8 seems to be required for the expression of tNPC markers and thus tNPC maintenance, although it might not determine NPC fate, as cNPCs that form PTAs also express FGF8.

To understand how FGF8 might contribute to the formation of the cap mesenchyme, we used a computational model combining FGF8-induced chemokinesis with weak repressive and attractive chemical signals released from the epithelium. Our simulations showed that this results in graded motility perpendicular to the UB surface and thus a trap-like region near the UB that immobilizes passing NPCs. In our model, the specific balance between FGF8-induced NPC motility and epithelial signalling is evident. How this balance is achieved, whether through the precise control of morphogen gradients or based on heterotypic interactions between NPCs and UB cells, remains to be elucidated.

Our results suggest that FGF8 is an autocrine chemokinetic factor expressed in the metanephric mesenchyme and is required for the condensation of NPCs while being involved in the maintenance of tNPCs. It is known that the FGF8 ligand interacts with several FGF receptors and that this interaction can also be modulated by heparan sulphate proteoglycans, which consequently affect the spatio-temporal gradient of FGF8 concentration (Krishnan Harish et al., 2022 preprint). Yet self-organized motility, such as we observed in the NPCs with FGF8, may provide adaptability to changes in the microenvironment. To fully understand how FGF8 mediates its function, a detailed elucidation of the ligand-receptor interaction in vivo is required.

In this work, the mouse experiments were conducted in accordance with the Finnish and EU legislation. The Finnish National Animal Experiment Board approved all animal experiments, and experiments were conducted under internal licences issued by the Laboratory Animal Centre of the University of Oulu, Finland. To delete Fgf8 from mouse kidneys, we crossed Fgf8^Δ2,3/+ males with Pax8^Cre/+ female. The progeny were genotyped and females with a Pax8^Cre/+;Fgf8^n/+ genotype were crossed with Fgf8^{Floxed/Floxed} males. The progeny were genotyped and embryos with the genotype Fgf8^n/c;Pax8^Cre/+ were selected for the experiments. A similar strategy was used for Wnt4^eGFPCre/+ and HoxB7^Cre/+. To delete Fgf8 using T^Cre, a similar strategy was employed to that of Perantoni et al. (2005). Briefly, females with genotype Fgf8^{Floxed/Floxed} were crossed with males with genotype T^Cre/Cre;Fgf8^Δ2,3/+ and progeny with a genotype of T^Cre/+;Fgf8^n/c were used for this study. Expression of Cre and deletion of Fgf8 were assessed by genomic PCR, as described by Perantoni et al. (2005). Timed matings were checked at noon for the presence of a vaginal plug, which was considered to be E0.5 when identified. To obtain unbiased kidney samples, male and female embryos were collected from pregnant females that were euthanized with CO₂ followed by cervical dislocation, as per the institutional guidelines. Embryos were collected in sterile PBS and kidneys were dissected in 1×PBS with calcium and magnesium. Dissected kidneys were further treated depending on the experiment. Details of the mouse lines used in this work can be found in Table S1.

For light microscopy, paraffin wax-embedded kidney sections were stained with Hematoxylin and Eosin, following standard procedures according to Veikkolainen et al. (2012).

Embryonic kidneys were collected and dissected at E11.5, E12.5, E16.5 and P0 from genotyped mutant and littermate controls. Samples were fixed in 4% paraformaldehyde (PFA) overnight at 4°C and dehydrated in serial dilutions with 25% ethanol→50% ethanol→75% ethanol→100% ethanol in water with an incubating time in each step of 45 min or until the sample sank to the bottom of the test tube. At this point, the samples were transferred to a clearing solution (xylene), twice, for 60 min each. Samples were moved to melted paraffin wax, thrice, for 60 min each, and then these samples were carefully embedded in paraffin wax blocks and stored at 4°C. Embedded samples were sectioned (14 µm) with Lecia microtome, and slides were prepared with up to four samples on each slide. To perform staining, sections were selected and slides were prepared by heating at 55°C to melt the paraffin wax. To deparaffinize, slides were incubated in xylene solution, twice, for 5 min each. To remove xylene, slides were incubated in the 100% ethanol, twice, for 5 min each, and then hydrated in 95% ethanol→75% ethanol→50% ethanol in water for 5 min each. Antigen retrieval was performed in 1 mM EDTA-NaOH solution (pH 8.0) or 10 mM sodium citrate-citric acid solution (pH 6.0) in pressure cooker for 10 min (Pileri et al., 1997). After antigen retrieval, samples were incubated, for 60 min, in blocking solution [1×PBS+0.01% Triton-X100+serum (serum was selected based on antibody reactivity)]. Incubation duration and temperature were tested for each primary antibody and used accordingly. Images were acquired on an Olympus BX51WI upright microscope with Hamamatsu ORCA-ER digital camera.

A sphere-forming assay was performed as described previously with minor modifications (Ihermann-Hella et al., 2018). MMs from E11.5 kidneys from CD-1 mice were dissociated with TripLE Express (Gibco) or with a mix of collagenase type 3 (Worthington Biochemical, LS004180) and DNase I (New England Biolabs, M0303S) for 5 min in 37°C in Hepes buffer (pH 7.35) at 37°C for 15 min. To obtain a single-cell suspension, after stopping the activity of the enzymes with complete media, the cell suspension was strained through 0.45 µm cell strainer. The total obtained cell suspension was divided into four equal parts, from which each part was mixed with two parts of Matrigel (Corning, 354277) and allowed to attach for 10 min at 37°C, 5% CO₂ and 95% humidity. Media containing a Src-kinase inhibitor (10 µM PP2; Tocris Bioscience, 1407) and/or FGF8b (100 ng/ml; R&D Systems, 423-F8) 24 h was added to the polymerized mix cells and Matrigel.

To remove nephrospheres from the embedded Matrigel, plates were chilled on ice for 1 h on a shaking platform. Liquefied Matrigel solution was collected and centrifuged at 10,621 g at 4°C for 10 min. The pellet was washed twice with DEPC-treated 1×PBS at 2655 g at 4°C for 5 min and flash-frozen in liquid nitrogen until required. RNA extraction was performed using RNeasy mini (Qiagen, 74104) and cDNA synthesis was performed using the First Strand cDNA synthesis kit (Thermo Fisher Scientific, K1612), where 100 ng of RNA was used as a template. cDNA was diluted 1:1 with PCR grade water; for the qPCR reaction, 2 µl cDNA was mixed with 1.2 µl each of forward and reverse primer (Table S3) along with 0.6 µl of PCR water and 5 µl Brilliant Sybr Green III qPCR master mix (Agilent Technologies, 600882). qPCR was carried out at CFX96 Touch System (Bio-Rad) with a program of 95°C for 10 min, 40 cycles of 95°C for 20 s, 60°C for 20 s and 72°C for 20 s, followed by a melt curve. qPCR was performed with three biological replicates in three technical replicates.

For the validation of anti-FGF8b antibody, a Flag-tagged FGF8b clone was obtained from Genscript CloneID: OMu22892D (NCBI Nucleotide: NM 001166361.1) and was overexpressed in CHO-K1 cells (ATCC, CCL-61). The media were collected from FBS-starved CHO-K1 cells expressing FGF8b and precipitated using the TCA method as described previously (Fic et al., 2010). Western blotting was carried out with controls (commercially bought rmFGF8b and cell lysate of validated protein-expressing Flag tag and CHO-K1 cell lysate).

For GSK3β quantification, a nephrosphere culture was established (see ‘Nephrosphere’ section). The cells were extracted as described in the section ‘Real time qPCR’. Cells were lysed using 1×RIPA cell lysis solution (Cell Signaling, 9806) supplemented with cOmplete, Mini, EDTA-free Protease Inhibitor Cocktail (Roche, 04693159001). Protein quantity estimation was performed using the a BCA Protein Assay Kit (Pierce, 23225). 50 mg of total protein was loaded onto in-house prepared 12.5% SDS-PAGE separating gel and 6% stacking gel, and separation was performed for 90 min at 110 V at room temperature. Transfer was carried out onto NCP Porablot Membrane (Macherey-Nagel, 12807411) for 90 min at 90 V at 4°C. The membrane was blocked with 5% BSA solution and all the primary antibodies were incubated overnight at 4°C, while secondary antibodies were incubated for 60 min at room temperature (Table S2). The detection was performed using LumiGLO Reagent (Cell Signaling, 7003S) on a Fujifilm LAS-3000 Imager. For sequential protein detection, the antibodies were stripped away, by incubating the membranes with 0.2 M NaOH solution for 15 min at room temperature and re-blocking with BSA. The protein quantitation was performed using an ImageQuant TL8.1 (Cytiva LifeSciences).

The non-radioactive section in situ hybridization technique was performed as described previously (Junttila et al., 2015). The used cDNA probes Wnt4, Wnt9b, Six2, Eya1 and Cited1 were gifts from Prof. Thomas Carroll (University of Texas Southwestern Medical Center, Dallas, TX, USA).

Dissected E11.5 kidneys were cultured in a Trowell culture system. After 3 days of culture with or without rmFGF8b, kidneys were dissociated into a single-cell suspension. Cells were fixed with 4% PFA and permeabilized with cytofix/cytoperm (BD Biosciences, 554714) as per the manufacturer's instructions. Samples were stained using primary anti-SIX2 antibody (Table S2) for 30 min on ice, washed and then stained using secondary goat anti-rabbit AlexaFluor 488 (Table S2) for 30 min on ice and washed thoroughly. Samples were scored on a FACSCalibur (BD Biosciences). Three biological repeats were carried out for each condition while maintaining the protocol and template for sample acquisition.

Sections of E16.5 Pax8^Cre;Fgf8^n/c and E11.5 T^Cre;Fgf8^n/c kidneys were cultured for 3 days in Trowell culture, fixed and stained using anti-SIX2 (Table S2) and anti-TROMA (Table S2) antibodies and counterstained with Hoechst 33342 (Thermo Fisher Scientific, H3570). Imaging was carried out with Zeiss LSM 780 confocal microscope and samples were analysed on a Zen Blue (2012 edition, Zeiss). The distance between the pair of Six2⁺ cells closest and most distant to the UB was measured repeatedly along the UB at intervals of one cell length to determine the thickness of the cap mesenchyme. Additionally, the dispersion of NPCs was quantified as the Euclidean distance between NPCs and UB surfaces. Using Bitplane Imaris 9.6.0 and the Imaris modules Measurement Pro and Vantage, NPC positions were quantified using fluorescence intensity-based spot detection; UB surfaces were segmented using fluorescence intensity-based 3D segmentation. To determine the proportions of attached and free or unattached Six2⁺ cells in both controls and mutant kidneys, cells were classified as attached when their centroid positions were closer to the UB than twice the median distance (17.4 µm) of all centroid positions of the Six2⁺ cells in the control kidneys.

A sphere-forming assay was modified to induce cell motility in response to FGF8. Metanephric mesenchyme cells were dissociated into a single-cell suspension and an equal amount of cells were divided and mixed with Matrigel. BSA-soaked or FGF8 (100 µg/µl) -soaked agarose blue beads (Affi-Gel Blue Gel, Bio-Rad, 1537302) were carefully placed in individual wells of a four-chamber 35 mm glass-bottom dish (Cellvis, D35C4-20-0-N). The Matrigel cell mixture was carefully applied to the surface of the beads and placed in a pre-equilibrated microscopic chamber maintained at 37°C and 5% CO₂. Time-lapse was performed with Leica SP8 falcon 20× water immersion lens for 24 h. Cells were tracked using the Imaris (v9.6; BitPlane) cell tracking functionality. Cell tracks from different samples (n=5 each) were pooled and analysed using the R package CelltrackR (Wortel et al., 2021).

Parameter values for the simulation of mesenchymal condensation in the nephrogenic niche. One length unit of the simulation box corresponds to half a micron, i.e. and 2 · 10⁵ iterations correspond to 12 h of developmental time, i.e. ⁠. The diffusion coefficient D of the morphogens is related to the LB relaxation time τ via (Tanaka et al., 2015). The gradient length of an exponential gradient c(x)=c₀e^−x/λ is the distance at which the concentration has decreased to c(λ)=c₀/e. is the median intracellular concentration. Mass source strength as well as production and degradation rates refer to the PhysicalNodes, which represent the discrete grid on which the fluid and the morphogens live (Tanaka et al., 2015). The area of an NPC or an epithelial cell comprises a few hundred PhysicalNodes; the simulation box comprises exactly 300 · 300 PhysicalNodes. Morphogen dynamics are described in the section ‘Setup’. Our model uses a cell-based Lattice-Boltzmann Immersed-Boundary simulation framework for morphogenetic problems (Tanaka et al., 2015). Model parameter values are shown in Table S4.

We simulated mesenchymal condensation within a 150 µm² section of the nephrogenic niche containing randomly dispersed NPCs and a flat patch of ureteric epithelium for an interval of 12 h, corresponding to embryonic days E10.5 to E11 (Fig. 5D, Movie 3).

At the end of each simulation, the distance of all NPC centroids to the centroid of the ureteric epithelium was measured. All data were pooled and visualized as a boxplot (Fig. 5C) using R (R Core Team., 2020). The significance level was determined based on a Wilcoxon signed-rank test and added to the boxplot.

Cells are represented by highly resolved 2D polygonal geometries with a cortical tension established by elastic forces between GeometryNodes (Tanaka et al., 2015). Similarly, cell adhesion is realized by elastic forces between neighbouring cells. GeometryNodes are added or removed when the distance between GeometryNodes, i.e. cell size, changes. Similarly, adhesions are created or removed based on a distance threshold between intercellular GeometryNodes. The cells are immersed in a Newtonian fluid and no-flux boundary conditions are imposed on the domain boundaries.

Morphogens are produced and degraded within cells, can freely diffuse within the entire domain and through cell boundaries, and are advected by motile cells. Parameter values are shown in Table S4.

Random motility is established by applying 2D velocities to each GeometryNode, where velocities are picked from a normal distribution whose standard deviation is proportional to the median local FGF8 concentration (Fig. 5B). Similarly, a weak attractive force is established by picking positive 1D velocities (directed towards the UB) from a uniform distribution where the upper bound is proportional to the median local WNT11 concentration.

We thank Ms Paula Haipus, Ms Johanna Kekolahti-Liias and Ms Hannele Härkman for excellent technical assistance, Prof. Maxime Bouchard for providing Pax8^Cre, Fgf8^{Floxed/Floxed} and Fgf8^Δ2,3/+, and Prof. Andy McMahon for Hoxb7^Cre mouse line. We also thank Tiina Jokela for collecting the mouse samples and establishing the protocol for genotyping. We thank Dr Alan O. Perantoni and Dr Mark B. Lewandoski for providing us with T^Cre;Fgf8^n/c mouse embryonic kidneys. Dr Veli-Pekka Ronkainen is acknowledged for helping us set up the time-lapse imaging and the Vainio lab for helpful discussions. M.M. thanks Harold Gomez for discussions on image analysis and Lisa Conrad for comments on the manuscript. Confocal imaging was conducted at the Light Microscopy Unit of Biocenter Oulu, University of Oulu. Imaging was performed at the Biocenter Oulu Light Microscopy Core Facility, University of Oulu, Finland, supported by Biocenter Finland.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.201012.reviewer-comments.pdf