Ret signaling in ureteric bud epithelial cells controls cell movements, cell clustering and bud formation

The receptor tyrosine kinase Ret plays a major role in the formation and branching morphogenesis of the ureteric bud (UB), a tubular epithelial tree that gives rise to the collecting duct system during kidney development (Costantini, 2016; Davis et al., 2014). Ret is expressed broadly in the nephric duct (ND) and in the UB as it emerges from the ND at E10.5 of mouse embryogenesis. Ret signaling in renal development is activated by GDNF, a protein secreted by the metanephric mesenchyme that binds to Ret via the co-receptor GFRα1. As the UB branches, expression of Ret and Gfrα1 is restricted to the tips of the UB branches, while GDNF is expressed by the mesenchyme surrounding each tip. In knockout embryos lacking either the Ret, Gfra1 or Gdnf gene, there is minimal, if any, UB branching, and the mutant mice usually display renal agenesis (Costantini and Kopan, 2010). In humans, mutations in RET have also been associated with renal agenesis or hypoplasia (Davis et al., 2014; Skinner et al., 2008).

Previous lineage-tracing studies showed that the Ret-expressing cells at the UB tips are the major progenitors of the renal collecting ducts; these cells divide rapidly, some of them remaining at the tips during UB growth and branching, while others exit the tips to give rise to the elongating trunks of the collecting ducts (Riccio et al., 2016; Shakya et al., 2005). Those UB cells that leave the tips turn off Ret expression. It remains to be fully elucidated how Ret signaling affects the behaviors of ND cells and UB tip cells in order to promote UB emergence, growth and branching. We have approached this issue by using a variety of methods to generate genetically mosaic kidneys, in which a subset of UB cells lack Ret, and by studying how these cells behave differently from their neighboring wild-type cells (Costantini, 2019). In chimeric mutant embryos containing Ret^+/+ and Ret^−/− cells in the nephric duct, the Ret^+/+ cells preferentially moved to the position of UB outgrowth in the caudal ND, where they emerged to form the tip of the primary ureteric bud, while the Ret^−/− cells were excluded from the tip. This revealed an early role for Ret in promoting ND cell rearrangements that generate the primary UB tip (Chi et al., 2009b; Shakya et al., 2005).

To investigate how Ret affects cell behaviors during UB branching, we later used mosaic analysis with double markers (MADM) (Zong et al., 2005), a method that generates pairs of Ret^+/+ and Ret^−/− sister cells marked by green versus red fluorescent proteins; these rare recombinant cells and their descendants can be tracked in developing kidneys (Riccio et al., 2016). These studies showed that loss of Ret greatly decreased the ability of a cell to remain in the tip compartment. Although in wild-type kidneys every cell is Ret^+/+, we observed that the level of Ret signaling appeared heterogeneous among wild-type tip cells, as several genes whose expression is stimulated by Ret signaling were expressed more strongly in some tip cells than others (Riccio et al., 2016). Therefore, we proposed a model for wild-type kidneys in which: (1) Ret signaling is maintained in a subset of tip cells, but decreased or extinguished in other tip cells; (2) those cells with high Ret signaling undergo rearrangements allowing them to stay at the new tips during UB branching and elongation; and (3) those with diminished Ret signaling fail to undergo these movements and are left behind to form the UB trunks (Riccio et al., 2016).

This model required further testing to better understand how Ret signaling influences cell behaviors. One observation needing additional investigation is illustrated in Fig. 1A. In this MADM clone, which presumably had arisen by recombination in a UB tip cell, although the Ret^−/− cells (red) were all found in the trunk, the Ret^+/+ cells (green) were located in both tip and trunk. As Ret is not expressed in trunk cells, these Ret^+/+ cells in the trunk must have already turned off Ret expression. This raised the question: did these Ret^+/+ cells leave the tip domain because they first turned off Ret expression (or extinguished Ret signaling), or did they turn off Ret expression as a consequence of leaving the tip, perhaps because of exposure to different signals in their new location?

To address this question, and to better understand how Ret influences UB cell behaviors, we generated a series of mouse lines in which Ret can be turned on permanently in individual UB cells (Fig. 1B). We constructed several Rosa26 alleles that, when exposed to Cre recombinase, delete a floxed transcriptional ‘stop’ sequence, causing the cell and its daughters to express Ret together with the nuclear-localized red fluorescent protein H2B-Tomato. As the Rosa26 locus remains active in virtually all cells during embryogenesis (Soriano, 1999), the expression of Ret by the Rosa26 allele will be maintained regardless of the location of a cell or its state of differentiation. One Rosa26 allele expressed a wild-type form of Ret (‘RetWT’); a second allele expressed constitutively active RetPTC2, which lacks the Ret extracellular domain and signals independently of GDNF (Bongarzone et al., 1993). A third Rosa26 allele, expressing only H2B-Tomato, served as a control. We used these mice to ask how a UB cell and its descendants would behave during branching morphogenesis if they were unable to turn off the expression of a Ret gene – either wild-type, GDNF-dependent Ret or GDNF-independent RetPTC2.

Mosaic expression of Rosa26 alleles was activated by crossing with Ret^CreERT2 (Fig. 1B) and injecting the pregnant mothers with tamoxifen (whose dose was adjusted to generate a small percentage of recombinant cells). In an initial analysis, tamoxifen was injected at E10.5, driving recombination in the emerging UB at E11.0-E11.5 (Fig. 1C). At E12.5, kidneys were explanted, cultured for 2 days and then imaged by confocal microscopy (images shown in this article, except where noted, are maximum projections from full z-stacks, allowing nearly every recombinant cell in the kidney, or organoid, to be detected). As expected (Riccio et al., 2016), in kidneys carrying the control allele Rosa26^H2B-Tomato, the recombinant cells were widely dispersed, with 53±5% located in the UB ampullae and 47±5% in the trunks (Fig. 1D,G; ‘ampulla’ refers to a swollen UB tip; during branching, each tip swells to form an ampulla, which is then remodeled to generate two, or sometimes three, new tips; Short et al., 2014; Watanabe and Costantini, 2004). In contrast, in Rosa26^RetWT kidneys, 90±6% of recombinant cells were found in the UB ampullae, and only 10±6% in the trunks (Fig. 1E,G). Thus, forced expression of RetWT in individual cells of the early UB caused most of them to localize to the terminal ampullae and remain there during several cycles of branching.

Ret signaling is activated by GDNF, which is expressed by cap mesenchyme (CM) (Cebrian et al., 2014; Hellmich et al., 1996; Moore et al., 1996; Pichel et al., 1996; Sanchez et al., 1996) and cortical stromal cells (Combes et al., 2019; England et al., 2020; Magella et al., 2018) surrounding the UB ampullae. GDNF can serve as a chemoattractant for other Ret-expressing cell types (Natarajan et al., 2002; Tang et al., 1998; Young et al., 2001), and the localized expression of GDNF could potentially provide spatial information that guides Rosa26^RetWT-expressing cells to the tips. As one test of this hypothesis, we performed similar studies using Rosa26^RetPTC2. RetPTC2 is a fusion protein containing the Ret kinase domain and the R1a subunit of protein kinase A, resulting from a chromosomal translocation in certain papillary thyroid carcinomas; this protein spontaneously dimerizes, causing the kinase domain to signal constitutively (Bongarzone et al., 1993). Although RetPTC2 lacks a transmembrane domain, it localizes to the plasma membrane by interaction with Enigma, and activates most, if not all, of the downstream signaling pathways employed by wild-type Ret (Arighi et al., 1997; Borrello et al., 1996; Durick et al., 1998; Lorenzo et al., 1997; Mercalli et al., 2001). Because of its GDNF independence, we reasoned that the permanent expression of RetPTC2 in individual UB cells might render them insensitive to the varying levels of GDNF in different regions of the kidney; thus, constitutive RetPTC2 signaling might obscure any GDNF-dependent positional information. However, we found that 95±1% of the Rosa26^RetPTC2 recombinant cells became localized to the UB ampullae (Fig. 1F); thus, constitutive RetPTC2 is similar to wild-type Ret in this regard (Fig. 1G).

To examine the sequence of events that caused the recombinant UB cells, which express either Rosa26^RetWT or Rosa26^RetPTC2, to become localized to the UB tips, we induced Ret^CreERT2 with tamoxifen at E9.5. This drives recombination between E10.0 and E10.5 in the ND and emerging UB, where Ret is expressed. The kidneys were explanted at E11.5, cultured and imaged by confocal microscopy two or three times per day for 3-4 days. Initially, most of the RetWT (Fig. 2 and Fig. S1) or RetPTC2 (Fig. 3) recombinant cells were located in the ampullae of the T-stage UBs. Over the next 3-4 days, as these cells proliferated and the UB branched, most of their daughter cells (91±9% for RetWT, 88±29% for RetPTC2) became localized to the new tips that formed by branching (Table S1); the remaining cells failed to reach a tip as the primary ampulla was reshaped into new tips and trunks, and thus became located in a trunk (e.g. Fig. 2).

Early in the cultures, when a UB ampulla branched, the recombinant cells in that ampulla were usually distributed to both of the new tips. Figs 2B,D and 3A (arrows) show several examples in which one or more recombinant cells moved far away from a group of sister cells near one tip and populated the opposite newly forming tip. As the kidneys grew, the recombinant cells proliferated to form large tight clusters; these denser clusters were no longer split when the tip branched again, and instead remained at only one of the daughter tips (e.g. Fig. 1). Although the nephrons were not labeled with a fluorescent protein, they were visible as condensates in bright-field images, and the clusters of Rosa26^RetWT or Rosa26^RetPTC2 recombinant cells were thus seen to eventually localize very close to the UB tip/distal nephron junctions (Fig. 2A-C; Fig. 3A,C; Fig. S1B). Overall, the behaviors of Rosa26^RetWT and Rosa26^RetPTC2 recombinant cells at the UB tips, during early branching, were very similar.

When dissected at E11.5, some of the kidneys (8/15 RetWT and 7/19 Ret-PTC2) also contained a few recombinant cells in the ureter, ND or common nephric duct (CND) (Figs 2E and 3D). As the kidneys grew, the Rosa26^RetPTC2 cells in these locations proliferated extensively and formed large dense clusters, which later evaginated to form ectopic buds (Fig. 3D, Table S2), while Rosa26^RetWT cells in these locations divided very slowly and did not cluster or form buds (Fig. 2E). Within the kidney proper, the recombinant cells in UB trunks behaved similarly: those expressing RetWT divided slowly and did not cluster (Fig. S2B), whereas those expressing RetPTC2 proliferated more extensively, generating dense clusters that progressed to form ectopic buds (Fig. S2D, Table S2). The ureter, ND, CND and the UB trunks all differ from the UB tips in that they are distant from the mesenchymal cells that secrete GDNF. Therefore, the inability of Rosa26^RetWT recombinant cells to form clusters and buds in these locations is likely due to a lack of Ret signaling.

To examine cell behaviors during later kidney development, tamoxifen was injected at E11.5, and kidneys were explanted at E12.5 and cultured for 6-7 days. In Rosa26^RetWT kidneys, any cells that had reached a tip during early UB branching (as well as their daughters) remained there, while proliferating, for the first 5-6 days; but eventually, some of these cells began to exit the tips and populate the adjacent trunks (Fig. S3A-C). Similarly, when Rosa26^RetWT kidneys were examined after developing in vivo to P0, many of the branches whose tips were full of recombinant cells also contained recombinant cells in the adjacent trunks (Fig. S4B,C). In contrast, in Rosa26^RetPTC2 kidneys, virtually all recombinant cells in the tips remained there, both throughout the long-term kidney cultures (Fig. S3 D-F) and in the P0 kidneys in vivo (Fig. S4D-F); in the latter, the tips composed of Rosa26^RetPTC2 recombinant cells were abnormally swollen and cystic (Fig. S4C-E). As discussed below, the ability of some cells expressing Rosa26^RetWT to eventually ‘escape’ from the tips, in contrast to those expressing Rosa26^RetPTC2, may be due to the GDNF-dependence of their Ret signaling. The massive swelling of UB tips composed of Rosa26^RetPTC2 recombinant cells is reminiscent of the cystic UB tips previously observed in kidneys in which RetPTC2 was broadly expressed in the UB, under the Hoxb7 promoter (Srinivas et al., 1999); however, in this earlier study, individual cell behaviors were not examined.

To investigate whether forced expression of RetWT or RetPTC2 altered the rate of UB cell division, we generated time-lapse movies of kidney cultures, in which the intervals between successive mitoses could be measured, because immediately after each UB cell mitosis (in the ampullae/tips) the two daughter cells rapidly move apart (Packard et al., 2013). Movie 1 shows an example of a Rosa26^RetWT kidney, Movie 2 a Rosa26^RetPTC2 kidney and Movie 3 a Rosa26^H2B-Tomato control kidney. These movies reveal recombinant cell behaviors similar to those observed in the time series (Figs 2, 3, S1, S2), with most Rosa26^RetWT or Rosa26^RetPTC2 cells populating a UB tip, while showing a high degree of motility during this process. We previously found that cell cycle length was unchanged in Ret^−/− compared with Ret^+/+ UB cells in MADM clones in UB ampullae (Riccio et al., 2016). Similarly, we found here that in the ampullae, the mean (±s.d.) cell cycle lengths for cells forced to express RetWT (15.0±1.7 h) or RetPTC2 (14.6±1.4 h) were indistinguishable from control UB cells (15.5±4.5 h) (Fig. S5). Therefore, neither the deletion of Ret, nor forced Ret expression or Ret signaling, changes the rate of UB cell division in the ampulla.

The tip-seeking behavior of UB cells that were forced to express Ret during kidney development suggested that these cells may have received some type of guidance cues, instructing them to orient their movements towards sites in the UB epithelium where new tips were emerging. One feature that distinguishes the UB tips from the trunks is their proximity to the cap mesenchyme (CM) and stromal cells that surround each tip. It is possible that secreted factors expressed by the CM or stroma, such as GDNF and FGFs, could serve such a role.

To investigate the importance of the mesenchyme for the UB cell behaviors observed in whole-kidney cultures, we next used a mesenchyme-free system, in which isolated UBs are cultured in Matrigel, an artificial extracellular matrix. When certain growth factors, including GDNF and any of several FGFs, are added to these cultures, the UBs can grow and branch extensively (Qiao et al., 1999, 2001). To further characterize these UB organoids, we first examined expression of the endogenous Ret gene, using a Ret^CFP knock-in allele (Gould et al., 2008). UBs isolated from E11.5 Ret^CFP/+ kidneys were cultured in Matrigel containing GDNF and FGF1, and CFP fluorescence was examined periodically (Fig. S6). Initially, CFP was expressed throughout UB epithelium, but as branching progressed, it became restricted to the tips (n=14/14 organoids). Thus, the UB organoids develop Ret-positive tips and Ret-negative trunks, similar to the UBs in intact kidneys, and are therefore a suitable model system to investigate whether the tip-seeking behavior of Ret-expressing cells occurs in the absence of mesenchyme.

We generated recombinant clones expressing either Rosa26^RetWT, Rosa26^RetPTC2 or Rosa26^H2B-Tomato, using a Ret^CreERT2 allele induced with tamoxifen at E10.5 (Fig. 4) (similar results were obtained by infection of isolated UBs with adenovirus-Cre, e.g. Fig. 5 and Movies 4-6). With low-dose tamoxifen (1 mg), the UBs at E11.5 typically contained a few sparsely distributed recombinant cells, which then proliferated as the organoids grew and branched (Fig. 4). In the control Rosa26^H2B-Tomato organoids, the recombinant cells continually dispersed as they proliferated, and did not form clusters (n=12/12) (Fig. 4C). In contrast, in all Rosa26^RetWT (n>20) (Fig. 4A) and Rosa26^RetPTC2 UB organoids (n>20) (Fig. 4B), most of the recombinant cells gradually coalesced to form multiple, discrete clusters (Fig. 4A,B, brackets), which increased in size and gradually became denser, as the recombinant cells overgrew the interspersed wild-type cells. This process resembled the formation of recombinant cell clusters in whole-kidney cultures (Figs S2 and S3, Figs S1, S2). As in whole-kidney cultures (Fig. S5), there was no significant difference between the rates of recombinant cell proliferation in control, Rosa26^RetWT and Rosa26^RetPTC2 organoids (Fig. S7). However, the UB organoids differed from the kidneys in that the clusters did not form specifically at tips but, rather, at apparently random locations in the UB epithelium. This suggested that in whole kidneys, the ‘tip-seeking’ behavior exhibited by Ret-expressing cells during UB branching may be guided by signals from the CM and/or stroma.

Each dense cluster of recombinant cells then began to evaginate from the surrounding UB epithelium, giving rise to a new bud (Fig. 4A,B); thus, the overall pattern of UB organoid branching was partially dictated by the (seemingly random) locations where the recombinant cell clusters had initially formed. These clusters varied widely in size, and the smaller clusters tended to generate smaller buds (yellow brackets) than did the larger clusters (orange brackets) (Fig. 4A,B). Once formed, the new buds generated by the recombinant cells continued to expand in size, and a few eventually started to generate folds (e.g. Fig. 4B, 115-159 h), but they did not further extend and branch, while the other UB tips lacking recombinant cells continued to branch.

In culture medium containing GDNF, the behaviors of Rosa26^RetWT and Rosa26^RetPTC2 recombinant cells in UB organoids were indistinguishable (e.g. Fig. 4A,B). But when cultured in medium lacking GDNF, the Rosa26^RetWT recombinant cells (which require GDNF for Ret to signal) became widely dispersed, rather than clustering (Fig. S8A); in contrast, the Rosa26^RetPTC2 recombinant cells behaved similarly with or without GDNF (Fig. S8B), consistent with the constitutive activity of RetPTC2. It was surprising that mouse UB organoids branched extensively with FGF1 but no GDNF in the medium (Fig. S8), as it had been observed that rat UBs do not branch without GDNF (Qiao et al., 1999); either rat and mouse kidneys differ, or some aspect of the culture conditions may have differed. However, the distinct behavior of the Rosa26^RetWT recombinant cells with or without added GDNF confirms the absence of significant levels of GDNF in the Matrigel or basal culture medium.

The observed clustering of UB cells expressing Rosa26^RetWT or Rosa26^RetPTC2 in UB organoids could involve two distinct mechanisms (Fig. 5A): (1) a group of initially dispersed cells could migrate towards each other (‘active convergence’); or (2) the cells could fail to disperse widely as they proliferate (‘lack of dispersal’). As we could not clearly distinguish between these mechanisms by imaging one or two times per day (e.g. Fig. 4), we performed time-lapse confocal imaging of seven additional Rosa26^RetPTC2 UB organoid cultures over 64 h (Movies 4-6). We analyzed 45 recombinant cell clusters that formed in these organoids, measuring the degree to which the RetPTC2-expressing cells converged during the formation of each cluster (Fig. 5). This revealed a range of behaviors, where some of the clusters formed predominantly through convergent cell movements (e.g. Fig. 5B-D), while, for other clusters, the primary mechanism appeared to be a failure of the cells to disperse as they proliferated (although some convergent cell movement also occurred) (e.g. Fig. 5E-G). These results reveal that forced Ret signaling in scattered cells in the UB epithelium can cause them to move towards each other, and also to remain within a more confined region as they proliferate, unlike control recombinant cells that continue to disperse.

To gain further insight into the mechanism of UB cell clustering, we sought to co-express a fluorescent cell surface marker specifically in the recombinant cells expressing Rosa26^RetWT or Rosa26^RetPTC2, to reveal changes in the epithelial cell shapes (as H2B-Tomato labels only the nuclei). However, we were unable to accomplish this co-expression with existing genetic tools. Instead, we examined UB cells in wild-type cultured kidneys, using Ret^CreERT2 to recombine the Rosa26^mTmG reporter allele in rare UB cells, and then performing time-lapse imaging of these kidneys. 3D rendering of the mGFP-expressing cells allowed us to follow their changes in shape over time (Fig. 6 and Movie 7). This showed that the epithelial cells in a wild-type UB ampulla are highly dynamic in shape, extending and retracting long thin processes in different directions and making transient contacts with other, non-adjacent UB cells. Many cells in the wild-type UB ampulla naturally express high levels of Ret. While the role of Ret signaling, if any, in the observed cell shape changes is unknown, this protrusive activity may play a role in the ability of UB epithelial cells to undergo the clustering behaviors observed in our experiments.

One of the major signaling pathways activated by Ret, and required for normal UB branching, is the Erk MAP kinase pathway (Fisher et al., 2001; Ihermann-Hella et al., 2014). In mammary gland epithelial organoids, sparse expression of a constitutively active, mutant form of the Map kinase MEK (MEK1DD) (Srinivasan et al., 2009) caused new buds to be generated by clusters of the mutant cells (Huebner et al., 2016). To investigate whether constitutive MAP kinase signaling was sufficient to drive the clustering and budding of ureteric bud cells, we used the same Cre-dependent Rosa26^MEK1DD allele to induce MAP kinase signaling in rare cells in UB organoids. We found that the recombinant cells, marked by GFP, remained widely dispersed in the UB organoids and failed to generate new buds (Fig. S9) (n=10/10 organoids). Thus, other signaling pathways downstream of Ret, perhaps in combination with MAP kinase signaling, must be required for the cell clustering and budding induced by Ret.

These experiments were designed to investigate the cellular mechanisms by which Ret signaling promotes the branching morphogenesis of the ureteric bud. Based on previous studies of genetically mosaic kidneys in which some cells lacked the Ret gene (reviewed by Costantini, 2019), we proposed a model (Riccio et al., 2016) in which Ret signaling promotes cell movements within the epithelium of the UB ampullae, towards the newly forming tips. Here, we describe further evidence supporting such a model: in contrast to clonal loss of Ret, which increased the probability that cells would leave the tip (Riccio et al., 2016), the forced expression of Ret in individual UB cells caused them to move to the tips and to remain there during many rounds of branching. Control recombinant cells expressing only H2B-Tomato became widely dispersed throughout the tips and trunks, and did not form clusters. Tip-seeking by RetWT- or RetPTC2-expressing cells may require positional signals from the renal mesenchyme, as it occurred in whole kidneys but not in UB epithelial organoids. In addition, our studies in UB organoids revealed a striking, self-organizing behavior in which scattered UB cells with forced Ret signaling attracted each other, forming dense cell clusters within the epithelium, which then evaginated to form new buds. The ability of Ret to promote these events suggests that similar, Ret-dependent cell behaviors are likely to play an important role in normal branching morphogenesis.

When either wild-type Ret, or constitutively active RetPTC2, was permanently turned on in rare cells of the early ureteric bud as it emerged from the nephric duct, the recombinant UB cells displayed a characteristic series of behaviors as the kidney developed. When first examined at E11.5, the cells were typically dispersed in the epithelium of the T-shaped UB. As the UBs further branched, most or all of the recombinant cells were distributed to the new tips that emerged – usually to several tips; sometimes to a single tip. These cells congregated as they proliferated, transitioning from loosely associated clusters to denser clusters. At the loose-cluster stage, individual recombinant cells were often observed to leave a forming cluster and migrate across the branching ampulla to populate the opposite emerging tip. However, once the clusters became denser, cells no longer left to move to a different tip. This suggests that the formation of dense clusters reflects a progressively increased attraction and/or adhesion between these cells. Although cluster formation at UB tips also involved cell proliferation, clustering was not caused simply by an increase in cell proliferation driven by Ret, as the forced expression of RetWT or RetPTC2 did not increase the rate of division of UB tip cells compared with those in control Rosa26^H2B-Tomato kidneys.

The tip-seeking behavior of Rosa26^RetWT cells in whole kidneys, but not in branching UB organoids lacking mesenchyme, suggests a possible role for the mesenchyme in this behavior (although we cannot rule out alternative reasons why cell behaviors in the organoids might differ). GDNF expressed by the nephron progenitor cells of the cap mesenchyme could potentially serve as a chemo-attractive factor that guides the recombinant UB cells; on the other hand, as each branching UB ampulla is broadly surrounded by CM, it is difficult to envision how the distribution of GDNF could simultaneously guide different recombinant cells within the ampulla to move to multiple nascent tips, as was observed. Another argument against this explanation is that expression of Rosa26^RetPTC2 induced a similar tip-seeking behavior; as RetPTC2 lacks an extracellular domain, its signaling is insensitive to changes in GDNF levels. However, all the kidneys in this study retained two wild-type Ret alleles in every cell, and it remains possible that this endogenous Ret allowed even the RetPTC2-expressing cells to respond to GDNF as a spatial cue. It was interesting that both Rosa26^RetWT and Rosa26^RetPTC2 recombinant cells accumulated in a sub-region of the UB tip close to the connection to the distal nephron (the epithelial cells of which no longer express GDNF). Perhaps a combination of GDNF and other signals from the CM and/or nascent nephrons guide the observed movements of Ret-expressing cells. Interestingly, the overexpression of GDNF in its normal expression domain in vivo (Li et al., 2019), the addition of excess GDNF to kidney cultures (Pepicelli et al., 1997; Sainio et al., 1997) or the knockout of Spry1, a negative regulator of Ret signaling (Basson et al., 2006), all caused swelling of the UB tips, which might be explained by the abnormal retention of cells in the tips; although individual cell behaviors were not followed in those earlier studies, the results are consistent with our observations.

The recombinant cells expressing Rosa26^RetWT or Rosa26^RetPTC2 remained at the tips during many cycles of branching over 4-5 days of culture. To answer the question posed in the Introduction, this suggests that when wild-type tip cells leave the tips and enter the adjacent trunks during this period of extensive branching, as many of them normally do, they must have previously turned off (or reduced the level of) Ret expression or Ret signaling. While the expression of several Ret-regulated genes (and hence the level of Ret signaling) appears to vary widely among wild-type UB tip cells (Riccio et al., 2016), it is not known to what extent the Ret expression level varies among tip cells; perhaps a single cell RNA-seq analysis of UB cells will address this question.

After 5-6 days of kidney culture, or after development to birth in vivo, the restriction of Rosa26^RetWT cells to the tips was apparently relaxed, as some of the cells left the tips to populate the adjacent trunks. However, cells expressing constitutive Rosa26^RetPTC2 always remained restricted to the tips, showing that constitutive Ret signaling prevents exit from the tip. Thus, the Rosa26^RetWT cells that leave the tips apparently have reduced Ret signaling activity, despite continued Ret expression. One possible explanation of the difference is that, after the UB ampullae fill with Rosa26^RetWT recombinant cells, those cells at the base of the ampulla, which are more distant from the cap mesenchyme, are exposed to lower levels of GDNF, reducing their Ret signaling level, whereas the GDNF-independent RetPTC2 remains fully active in UB cells regardless of their location, blocking their exit from the tip.

The behaviors of Rosa26^RetWT and Rosa26^RetPTC2 recombinant cells in mesenchyme-free UB organoids, which grow and branch in an environment where the added growth factors are uniformly distributed in the surrounding extracellular matrix, showed interesting similarities, as well as differences, with their behaviors in intact kidneys. As in the intact kidneys, the recombinant cells in UB organoids first formed loosely associated clusters, which then became denser as the interspersed wild-type cells were displaced. Unlike intact kidneys, however, the recombinant cells in UB organoids did not move to the tips of the organoid (indeed, they often formed before the organoids had developed distinct tips), but, instead, they generated clusters at apparently random locations in the epithelium. Time-lapse imaging of Rosa26^RetPTC2 organoids revealed that these clusters formed either by convergent cell movements, or by a failure of the cells to disperse as they proliferated or a combination of these mechanisms. Like wild-type UB tip cells (Packard et al., 2013), the individual Rosa26^RetWT and Rosa26^RetPTC2 recombinant cells underwent mitosis-associated cell dispersal (MACD) immediately after they divided (Fig. S5, Movies 1, 2, 4-6); MACD causes the two daughter cells to initially occupy non-adjacent positions in the epithelium. However, further clonal dispersion can involve other mechanisms (e.g. divergent cell movements or the proliferation of the intervening cells), and, in this case, clonal dispersion was inhibited despite normal MACD. The clustering of Ret-expressing cells could be explained either by cell-cell attraction at a distance or by increased adhesion among the cells. The mechanism by which Ret-expressing cells adhere to and/or attract each other is likely to involve cell adhesion molecules or secreted proteins whose expression is increased by Ret/GDNF signaling.

Because only the recombinant cell nuclei were differentially labeled (with H2B-Tomato) in these experiments, we could not discern the cell outlines to determine when the recombinant cells made contacts with each other. However, in time-lapse movies of wild-type kidneys, in which a few random UB tip cells were labeled by membrane-associated GFP, we observed that the labeled epithelial cells formed long transient membranous protrusions that sometimes contacted non-adjacent UB cells (Fig. 6 and Movie 7). Furthermore, we previously reported that, in 2D cultures of dissociated wild-type fetal kidney cells (Unbekandt and Davies, 2010), the UB cells employed long membranous protrusions to contact each other and to generate UB cell clusters (Leclerc and Costantini, 2016). In a similar 2D culture of dissociated kidney cells, it was observed that wild-type UB cells formed clusters via random motion coupled with differential adhesion (Lefevre et al., 2017); in contrast, our time-lapse movies revealed that, in the epithelium of UB organoids, a group of initially distant Rosa26^RetPTC2 recombinant cells often converged in an apparently non-random manner (Fig. 5B-D). If Ret signaling increased this protrusive activity, or increased cell-cell adhesion, it might promote the observed clustering. Interestingly, we previously found that GDNF upregulated the expression of several potential cell-adhesion molecules in cultured wild-type ureteric buds (Lu et al., 2009), although the roles of these specific molecules remain to be investigated.

After the Rosa26^RetWT or Rosa26^RetPTC2 recombinant cells in UB organoids formed dense clusters, these clusters evaginated from the epithelium, giving rise to new buds composed entirely of the recombinant cells; this revealed a remarkable self-organizing ability induced by Ret signaling. It is not clear what triggers the transition from cell clustering to budding, but it appears to be independent of cluster size, as smaller clusters typically made smaller buds, rather than waiting to bud until they had grown to a consistent size (Fig. 4). Within intact kidneys, a similar budding was often observed after Rosa26^RetPTC2 recombinant cells formed clusters in the ureter or UB trunks (Fig. 3C, Fig. S2D). In contrast, the dense clusters of Rosa26^RetPTC2 at the UB tips within intact kidneys did not appear to generate extra buds, suggesting that the behaviors of these cells is modulated by interactions of the UB with the CM and/or nephrons.

In conclusion, these studies further support a model (Riccio et al., 2016) in which the decision of individual UB cells to remain in the progenitor population at the tips, or to exit the tips and give rise to the elongating trunks, is regulated by Ret signaling. Our work also reveals three types of UB cell behaviors promoted by Ret signaling: tip seeking, cluster formation and epithelial budding. Although these processes undoubtedly contribute to the important role of Ret signaling in normal kidney development, their specific functions in UB branching morphogenesis, and how they are coordinated, remain to be fully elucidated. The ability to turn on Ret in individual, labeled cells using Rosa26^RetWT and Rosa26^RetPTC2 mice, together with other tissue-specific Cre lines, should also be useful for studies of the role of the Ret gene in other tissues during normal development and disease processes.

Human RetPTC2 cDNA was provided by Dr Marco Pierotti (Istituto Nazionale Tumori, Milan, Italy) and mouse wild-type Ret9 cDNA by Dr Vassilis Pachnis (National Institute for Medical Research, London, UK). H2B:tdTomato was assembled by PCR cloning H2B together with tdTomato using the following primers for stitching: pENTR4_H2B_F, 5′-TCAGTCGACTGGATCCCGTATAGCGGCCGCATATTACCATGCCAGAGCCAGCG-3′; H2B_R, 5′-TCGCCCTTGCTCACGCGCTGGTGTACTTGGTGATGGCCTTAGTACC-3′; tdTomato_F, 5′-AAGTACACCAGCGCGTGAGCAAGGGCGAGGAGGTCATCAAAGAGTTCATGC-3′; and tdTomato_R, 5′-GTCTAGATATCTCGAGGCCATCGATGTGCGGCCGCTTTACTTGTACAGCTCGTCC-3′. To achieve co-expression of RET variants with a fluorescent reporter, H2B:tdTomato was cloned in frame to the 3′ cDNA sequences of RetWT and RET PTC2, connected by the viral T2A sequence, using the primers: pENTR4_PTC2_F, 5′-TCAGTCGACTGGATCCCGATAGCGGCCGCATAACCATGGAGTCTGGCAGTACC-3′; Ret_T2A_R, 5′-GTCACCGCATGTTAGCAAGCTTCCTCTGCCCTCTCCACTGCCGAATCTAGTAAATGCATGG-3′; T2A_H2B_F, 5′-GAAGCTTGCTAACATGCGGTGACGTCGAGGAGAATCCTGGCCCAATGCCAGAGCCAG-3′; tdTomato_R 5′-GTCTAGATATCTCGAGGCCATCGATGTGCGGCCGCTTTACTTGTACAGCTCGTCC-3′; and pENTR4_RetWT, 5′-TCAGTCGACTGGATCCCGATAGCGGCCGCATTAACCATGGCGAAGGC-3′.

All constructs were BamHI/NotI cloned into the pENTR4 entry plasmid (Invitrogen, 11818-010) and subsequently recombined into pRosa26-DEST vector (Hohenstein et al., 2008), using Gateway cloning methods (Invitrogen). Sequence-confirmed pRosa26-DEST targeting vector plasmids were linearized and electroporated into KV1 ES cells. Primers to screen for homologous recombinant clones were: G1 for the external 5′ arm primer 5′-TAGGTAGGGGATCGGGACTCT-3′ and the internal G2 reverse primer 5′-GCGAAGAGTTTGTCCTCAACC-3′, which generate a 1.3 kb PCR 5′ arm fragment (Nyabi et al., 2009). Targeted ES clones were injected into C57BL/6 blastocysts to generate chimeric male mice, which were bred to C57BL/6 females to transmit the alleles.

All animal use was approved by the Institutional Animal Care and Use Committee. Other mutant mouse strains employed were: Hoxb7/myrVenus [Tg(Hoxb7-Venus*)17Cos/J] (Chi et al., 2009a), Ret^CFP (Ret^tm2.1Heno) (Gould et al., 2008), Ret^CreERT2 [Ret^{tm2(cre/ERT2)Ddg}] (Luo et al., 2009), Rosa26^Mek1dd [Gt(ROSA)26Sor^{tm8(Map2k1*,EGFP)Rsky}] (Srinivasan et al., 2009) and Rosa26^mTmG [Gt(ROSA)26Sor^{tm4(ACTB-tdTomato,-EGFP)Luo}] (Muzumdar et al., 2007).

Recombination of Rosa26 alleles was induced by crossing females homozygous for the Rosa26 allele with males heterozygous for Ret^CreERT2 mice and homozygous for Hoxb7/myrVenus; pregnant females were injected i.p. at the stages indicated in the text, with a single dose of tamoxifen empirically adjusted (0.25-1.0 mg) to yield a low percentage of recombinant UB cells. Kidneys were explanted at E11.5 or E12.5 and cultured on Transwell clear filters (Costar 3450) as described previously (Costantini et al., 2011). In some cases, instead of injecting the mother with tamoxifen, we cultured the explanted kidneys in medium containing 50-100 nM 4-OH tamoxifen for 1 h. Kidney cultures were periodically imaged on a Zeiss LSM710 or LSM800 confocal inverted microscope with 10× objective, then returned to the incubator. In some cases, a corresponding bright-field image was collected with a T-PMT transmitted light detector. After culture, some kidneys were fixed in paraformaldehyde, cleared with FocusClear (Cedarlane Labs) and re-imaged. Maximum projections were generated from the z-stacks using Zen2 Blue Lite software (Zeiss), and recombinant cells in the UB ampullae versus trunks were counted manually. To examine kidneys that had developed in vivo, tamoxifen was injected at E12.5, kidneys were dissected from P0 pups, fixed in 4% paraformaldehyde, embedded in low-melting temperature agarose and sectioned at 60 µM on a Leica vibratome.

E11.5 kidneys were treated with collagenase and the UBs isolated with tungsten needles, as described previously (Costantini et al., 2011). In some cases, the mother was first injected with tamoxifen at E10.5 to activate recombination by Ret^CreERT2; in other cases, the isolated UBs were incubated for 20 min at 37°C with adenovirus-Cre (a gift from Dr Stephen Goff, Columbia University, New York, NY, USA) diluted in DMEM/F12 medium (see below) to ∼10⁶ pfu/ml, then washed four times in DMEM/F12. The UBs were cultured in a 40 μl drop of Matrigel (Corning 356237) in a Nunc four-well dish (Thermo Scientific 144444) or in a cylindrical well (0.5 inch diameter, ePlastics.com) glued with Aqueon silicone sealant (Central Aquatics) to the glass bottom of a MatTek P35G-0-20-C culture dish. The Matrigel was submerged in DMEM/F12 (Gibco 10565-018) containing 10% fetal bovine serum, 5-10% BSN cell-conditioned medium (prepared as described by Zhang et al., 2012), 1× antibiotic-antimycotic (Gibco 15240-096), 125 ng/ml recombinant hGDNF (R&D Systems) and 500 ng/ml FGF1 (Peprotech) (Qiao et al., 2001). BSN cells were a gift from Dr Sanjay Nigam (University of California at San Diego, USA) and were not re-authenticated or tested for contamination.

For time-lapse imaging of cultured kidneys (Fig. S5 and Movies 1-3), kidneys were grown on Transwell clear filters (Costar) with 100 nM 4-OH tamoxifen for 24 h, then transferred to normal medium and cultured in an environmental chamber on the stage of a Zeiss epifluorescence microscope with 10× objective, using Zen Blue software (Zeiss). UB organoids for time-lapse imaging (Movies 4-6) were cultured as described above, in glass-bottom dishes in an environmental chamber (Costantini et al., 2011) on the stage of a Zeiss LSM710 confocal microscope; a z-stack of images was collected with a 10× objective every hour using Zen Black software (Zeiss). Time-lapse imaging of Rosa26^mTmG kidneys (Fig. 6, Movie 7) was performed as previously described (Packard et al., 2013).

To estimate the degree to which Rosa26^RetPTC2 recombinant cells underwent convergent movements during cluster formation in UB organoids, we used Zen software (Carl Zeiss) to measure the x, y and z coordinates of the Tomato+ nucleus of each cell contributing to a cluster, at two times: first when the nuclei were maximally dispersed (e.g. 40 h in Fig. 5B) and again after the cells had formed a dense cluster (e.g. 49 h in Fig. 5B). The degree of dispersion at each of the two time points was calculated as follows: the coordinates of the central position in the cluster (x_avg, y_avg and z_avg) were calculated as the average of x, y and z values for all the nuclei in the cluster; the distance (d) from each nucleus (i) to the central position was calculated as d=SQRT[(x_i – x_avg)²+(y_i – y_avg)²+(z_i − z_avg)]; the d values for the nuclei in each cluster were averaged as a measure of dispersion, d_avg. Then a ‘convergence index’ (c) was calculated as c=d_avg-initial÷d_avg-final.

We thank Zaiqi Wu for technical assistance; Chyuan-Sheng Lin of the Genetically Modified Mouse Models Shared Resource, Columbia University Irving Medical Center, for generating chimeric mice; Theresa Swayne of the Confocal and Specialized Microscopy Shared Resource (CUIMC) for help with microscopy and image processing; and Hideki Enomoto, Andrew Ewald, Stephen Goff, Sanjay Jain, Vassilis Pachnis and Marco Pierotti for reagents and mouse strains.