RETRACTED: Btbd7 is essential for region-specific epithelial cell dynamics and branching morphogenesis in vivo

Branching morphogenesis is essential for successful development of many epithelial organs, including the lungs, kidneys, pancreas, salivary glands, mammary glands and prostate. This complex, highly dynamic process remodels a simple epithelial bud into a highly arborized branched organ structure that maximizes epithelial surface area for secretion or absorption. Although many studies have highlighted requirements for numerous signaling molecules and transcription factors during branching morphogenesis (Costantini and Kopan, 2010; Harunaga et al., 2011; Hauser and Hoffman, 2015; Hennighausen and Robinson, 2005; Iber and Menshykau, 2013; Kwon and Larsen, 2015; Shih et al., 2013; Varner and Nelson, 2014), it is not fully understood at the cell and tissue level how such diverse regulatory pathways orchestrate the extensive physical remodeling that shapes branched epithelial tissues. Recent advances in microscopy have established that specific dynamic cell behaviors, such as changes in cell motility, cell-cell adhesion and cell-extracellular matrix interactions, are key functional mediators of this tissue reorganization (Daley and Yamada, 2013; Friedl and Gilmour, 2009; Harunaga et al., 2011; Huebner and Ewald, 2014; Kim and Nelson, 2012; Nelson and Larsen, 2015; Varner and Nelson, 2014). However, how such dynamic cell and tissue behavior is guided by localized regulatory molecules at specific sites within a tissue remains largely unknown.

Although the final mature architecture of major branched organs can vary greatly, each begins development as a simple epithelial outgrowth extending into the surrounding mesenchyme (Wang et al., 2017). Cleft formation is a major mode of branching by which an initial bud splits into two or more new buds. Consecutive rounds of cleft formation or budding with proliferative outgrowth of epithelial end buds arborize the tissues to generate the final complex branched architecture. Although the exact mechanisms of cleft formation may vary between different organs and remain unknown, several studies have revealed a role for the coordinated interplay between dynamic tip cells at the periphery of branching end buds and the surrounding microenvironment. A common theme is that the tips of branching end buds must undergo substantial physical deformation. Thus, the epithelium itself must exhibit a high degree of plasticity so that it can be extensively reshaped and remodeled during morphogenesis.

Localized alterations in epithelial cell migration and cell-cell adhesion represent two major mechanisms by which epithelial end buds could both maintain plasticity and contribute actively to the branching process. Live imaging of organ cultures has revealed that individual epithelial cells of major branching organs are transiently motile during early development (Huebner et al., 2016; Riccio et al., 2016; Schnatwinkel and Niswander, 2013). Cells at the peripheral tips of branching end buds are the most dynamic, with faster migration rates than inner bud cells in the developing salivary gland (Hsu et al., 2013; Larsen et al., 2006); in the pancreas, the outer bud cells migrate with more persistence than their inner bud counterparts (Shih et al., 2016).

Localized differences in cell-cell adhesion also exist in many of these organs, and classical studies established that cell-cell adhesions change throughout development (Menko et al., 2002; Nishimura and Takeichi, 2009; Walker et al., 2008). For example, the tip cells of mammary terminal end buds exhibit reduced apico-basal polarity and decreased cell-cell adhesion characterized by immature adherens and tight junctions compared with stabilized ductal structures (Ewald et al., 2012). Recently, E-cadherin expression was shown to be reduced in pancreatic tip cells compared with inner bud regions (Shih et al., 2016). Although generalized signaling from growth factors and integrin-mediated interactions with the basement membrane can modulate tip cell migration and cell-cell adhesion (Hsu et al., 2013; Riccio et al., 2016; Shih et al., 2016), very little is known about locally expressed regulatory molecules within peripheral end-bud cells that could govern these cellular processes.

We previously identified Btbd7 [BTB (POZ) domain-containing 7] as a critical regulator of cleft formation during in vitro branching morphogenesis that functions via the downregulation of E-cadherin and stimulation of epithelial cell motility (Onodera et al., 2010). Btbd7 also promotes an epithelial-to-mesenchymal transition in cultured tumor cells, with an increase in invasive capability (Fan et al., 2014; Luo et al., 2015; Wang et al., 2014; Yang et al., 2016). Nevertheless, whether Btbd7 is a master regulator of branching morphogenesis in vivo for multiple organs remains unknown. Furthermore, although BTB domain-containing proteins can function as transcription factors in the nucleus, a newly appreciated function involves their interaction with E3 ubiquitin ligases to mediate protein degradation of specific targets (Genschik et al., 2013; Metzger et al., 2012). It is not known whether Btbd7 has roles in the ubiquitin-proteasome-mediated protein degradation that affects branching morphogenesis.

We report here the generation of a Btbd7 knockout mouse model and demonstrate that Btbd7 is required for successful branching morphogenesis of multiple epithelial organs. Our data reveal enhanced Btbd7 protein localization at the peripheral tips of branching end buds, i.e. in the outer bud cells. Mechanistically, Btbd7 locally enhances motility of these outer bud cells and cell-adhesion dynamics, with no apparent effect on the adjacent inner bud cells. Using in vitro MDCK cells to model outer bud cell behavior, we also demonstrate that experimental elevation of Btbd7 levels results in increased ubiquitylation, internalization and degradation of E-cadherin, accompanied by altered cell motility and cell-cell adhesion that mimics outer bud cell dynamics. We conclude that Btbd7 is enriched locally to regulate epithelial cell dynamics and increase tip plasticity during branching morphogenesis.

We first characterized Btbd7 protein localization in a variety of wild-type mammalian branching organs at embryonic day 13.5 (E13.5). In the developing salivary gland, Btbd7 protein was present in both the epithelium and the mesenchyme (Fig. 1A), but at particularly high levels in the outer salivary epithelial cells closest to the peripheral tips of branching end buds, with less localization in the center of buds and in ductal regions (Fig. 1A, inset). This differential localization was not due to uneven antibody penetration, as immunostaining of thin tissue sections also revealed the same protein localization (Fig. S1A). Furthermore, pre-treatment with a blocking peptide inhibited antibody binding in the salivary gland (Fig. S1B). This protein localization is consistent with the proposed role of Btbd7 in branching morphogenesis, as the peripheral tips of end buds are the regions of active branching where clefts initiate and progress.

In other major branching organs, such as the lung (Fig. 1B) and kidney (Fig. 1C), Btbd7 displayed analogous differential localization in the epithelial cells closest to the tips of expanding lung buds and elongating kidney tubules, with less localization proximally in stabilized ductal regions. The exception was the pancreas, where we observed only low-level cytoplasmic localization throughout, with no preferential localization pattern close to the periphery of end buds (Fig. 1D). Together, these data indicate that Btbd7 is highly localized in dynamic tip regions of epithelial buds of lung, kidney and salivary gland, where cells are actively participating in branching morphogenesis.

We next characterized Btbd7 protein localization at progressive stages of salivary gland development. Between stages E11.5 and E14.5, when the gland is branching rapidly, Btbd7 is present in every cell at the peripheral tips of salivary gland end buds. By E15.5 and later, however, fewer peripheral cells retained Btbd7 expression (Fig. S1C). These results demonstrate a relationship between high Btbd7 protein localization in peripheral tip cells during stages of active branching, with decreased expression in such cells at later developmental stages when differentiation occurs, further suggesting an important role for Btbd7 during branching morphogenesis.

Finally, because the protein localization determined here did not coincide with the Btbd7 mRNA localization reported previously by our laboratory (Onodera et al., 2010), we performed single-molecule fluorescence in situ hybridization (smFISH) of salivary gland whole mounts to more quantitatively examine mRNA distribution. Btbd7 mRNA was ubiquitously expressed across the salivary gland epithelium (Fig. S2A-C) with no peripheral enrichment, indicating that additional level(s) of post-transcriptional regulation enrich Btbd7 protein at the peripheral tips of branching end buds.

To test directly the role of Btbd7 in branching morphogenesis in vivo, we generated a Btbd7 knockout (KO) mouse. We deleted exons 3 and 4 of the Btbd7 gene, which include the start codon as well as both BTB domains, by breeding mice expressing a floxed allele of the gene to mice globally expressing Cre recombinase (see supplementary Materials and Methods and Fig. S3). With increasing gestation time, we observed decreased ratios of KO embryos, indicating death of some embryos between E13 and E17 (Fig. S4A); few Btbd7 KO pups were actually born, and such pups died within a few days of birth. At E13 and E17, Btbd7 KO embryos are smaller than their wild-type or heterozygous counterparts (Fig. S4B-D). Despite this size reduction, we detected no significant difference in heart weights between genotypes, indicating organ selectivity in Btbd7 KO effects (Fig. S4E,F).

We have previously reported that siRNA knockdown of Btbd7 in submandibular salivary gland (SMG) organ explants blocks cleft formation and subsequent branching morphogenesis (Onodera et al., 2010). To establish whether Btbd7 is important for SMG branching morphogenesis in vivo, we compared the morphology of wild-type, heterozygous and Btbd7 KO SMGs from E13.5 embryos. When examined 2 h after dissection from embryos, both wild-type and heterozygous glands were multi-lobular (7.4±3.5 buds for wild type and 4.8±1.7 buds for heterozygous), whereas the majority of Btbd7 KO glands showed severely inhibited branching, in nearly all cases consisting of only a single bud (1.1±0.2 buds, Fig. 2A,B). We observed varying degrees of penetrance of this phenotype: e.g. most KO glands displayed one bud, but a few had two to six buds (Fig. S5A). In some cases, Btbd7 heterozygous glands were initially unbranched (Fig. S5B), but all wild-type SMGs were branched.

To determine whether this Btbd7 KO phenotype was due to developmental delay, we tested whether single-bud Btbd7 KO SMGs could undergo branching in ex vivo culture. We found that such Btbd7 KO SMGs failed to branch much further over a 40-hour culture period, whereas their wild-type counterparts increased their bud numbers approximately ninefold (Fig. 2A,B). Interestingly, heterozygous SMGs branched significantly less over 40 h than wild-type glands, indicating a developmental delay (Fig. 2B). Importantly, size-matched wild-type E12.0 SMGs at the single bud stage and the occasional heterozygous SMG that remained a single bud at E13.5 were both able to undergo branching in ex vivo culture, indicating that the KO phenotype at E13.5 is not simply due to a delay in branching, but instead likely because of a lack of intrinsic branching capacity (Fig. S5B-D). Furthermore, although we found relatively low numbers of viable Btbd7 KO embryos at E17.5, the few salivary glands that we could analyze in surviving embryos appeared very small compared with their wild-type and heterozygous counterparts (Fig. S5E). Thus, suppression of the branching phenotype is maintained at later stages, further suggesting that the Btbd7 KO phenotype involves a major loss of branching morphogenesis and not merely a temporary delay in this process. Finally, this major loss of branching was not due to increased apoptosis or decreased cell proliferation, as we observed no apparent differences in numbers of cleaved caspase 3-positive cells at the tips of epithelial end buds (Fig. S5F) or in numbers of proliferating end-bud cells as determined by EdU incorporation (Fig. S5G,H).

We compared effects of Btbd7 KO on the branching phenotype of other major branching organs besides the salivary gland. Strikingly, Btbd7 KO lungs (Fig. 2C,D) and kidneys (Fig. 2E,F) also exhibited a dramatic decrease in branching morphogenesis in vivo at E13.5 (2 h time-point) and after over 40 h in ex vivo culture (18 and 40 h time-points). The phenotype is not as severe as in the salivary gland, but represented a major suppression of branching; E13.5 Btbd7 KO lungs and kidneys branched poorly compared with size-matched wild-type E12.0 organs (Fig. S6A-D). Interestingly, we observed no significant KO effect on the branching phenotype of E13.5 pancreas (Fig. S6E,F). This absence of inhibition is associated with the minimal Btbd7 protein localization at the tips of expanding pancreatic buds (Fig. 1D). To determine the efficiency of Btbd7 gene knockout at the protein level, we immunostained for Btbd7 and found no Btbd7 signal in all branching organs examined from Btbd7 KO embryos (Fig. S7). Taken together, these results indicate that Btbd7 is an essential regulator of branching morphogenesis in many, but not all, branching organs.

Because Btbd7 is enriched in the outer cells at the tips of branching end buds (Fig. 1), we hypothesized that Btbd7 deletion might affect their behavior. We first compared epithelial bud outer cell morphologies. In wild-type glands, each epithelial end bud consists of an outer layer of relatively columnar cells surrounding an inner mass of more rounded cells. With Btbd7 deletion, the outer cells were significantly more columnar and tightly packed compared with wild-type and heterozygous counterparts, which we quantified as axial ratios of cell height to cell width (Fig. 3A,B). Interestingly, very early stage wild-type E11.5 SMGs exhibited outer cell morphologies similar to Btbd7 KO glands at E13.5 (Fig. 3A,B), suggesting that in the absence of Btbd7, E13.5 SMGs are trapped in an earlier, highly compact, strongly epithelial embryonic state that is inhibitory to branching.

We next compared E-cadherin localization in outer versus inner cells of wild type and Btbd7 KO glands. High-resolution confocal z-stacks revealed that the intensity of E-cadherin localization at cell-cell contacts between epithelial cells was significantly lower in the outer bud cells (Movie 1), as quantified by fluorescence intensity line scans across cell-cell junctions (Fig. 3C-E). Strikingly, in Btbd7 KO salivary glands, outer cell E-cadherin localization was substantially enhanced to levels similar to wild-type inner bud cells (Fig. 3C-E, Movie 1). We also observed a similar increase in peripheral end-bud E-cadherin localization in Btbd7 knockout lungs, but not in the pancreas (Fig. S8A,B). These results indicate that the increased Btbd7 localization in branching outer bud cells is associated with decreased E-cadherin protein localization at cell-cell contacts, whereas in the inner bud cells where Btbd7 localization is low, E-cadherin remains largely unaffected.

To determine whether reduced E-cadherin mRNA levels could account for the lower E-cadherin protein localization in the outer bud cells of wild-type glands, we performed small molecule fluorescent in situ hybridization of Cdh1. Similar to E-cadherin protein, we observed fewer Cdh1 mRNA puncta in the outer peripheral cells compared with inner bud cells (Fig. S8C). In contrast, mRNA localization of two other genes, Gapdh and Emg1, showed no difference between inner and outer epithelial cells (Fig. S8D). However, Btbd7 KO glands failed to show increased Cdh1 mRNA in the outer cells (Fig. S8E,F). Thus, even though transcriptional regulation may contribute to decreased E-cadherin protein in the peripheral tips of epithelial end buds, it is Btbd7 independent; consequently, the Btbd7-dependent alteration in E-cadherin is likely post-transcriptional.

We have previously established that salivary gland epithelial cells are highly motile during early development, with outer bud cells exhibiting the largest migration velocities when compared with inner bud and ductal epithelial cells (Hsu et al., 2013). Thus, we hypothesized that cell migration might be perturbed in Btbd7 KO glands accompanying the alterations in outer cell morphology and E-cadherin localization. We quantified migration by labeling E13.5 glands with the nuclear marker Hoechst and performing two-photon time-lapse imaging with cell tracking of outer versus inner cell nuclei. In the outer bud cells of Btbd7 KO glands, migration was significantly slower, with shortened path lengths compared with their wild-type and heterozygous counterparts (Fig. 4A-C, Fig. S9A, Movie 2). Strikingly, we observed no such differences between wild-type, heterozygous and KO inner bud cells (Fig. 4A,D,E, Fig. S9B, Movie 2), indicating that Btbd7 preferentially regulates outer bud cell migratory behavior.

To confirm the association of outer bud cell dynamics with cleft formation during salivary gland branching morphogenesis, we examined the morphology of cells immediately adjacent to initiating clefts in fixed E13.5 wild-type glands. Glands were immunostained with a collagen IV antibody to mark early clefts, which appeared as indentations in the basement membrane that were accompanied by lateral displacement of cells labeled for E-cadherin and with a nuclear dye – a clear gap between cells at sites of cleft initiation was indicated by a wide space between nuclei and separation of cells (Fig. S9C, white arrow; Movie 3). In contrast, we observed minimal basement membrane indentations into Btbd7-KO epithelial buds that never initiated into true clefts, with E-cadherin remaining strongly localized at basolateral cell-cell contacts without separation of cells and nuclei (Fig. S9D, Movie 3).

We further analyzed these dynamic nuclear separations in real time to evaluate their association with cleft formation. We first used time-lapse imaging, in wild-type E13.5 SMGs labeled with both a nuclear dye and a non-perturbing fluorescently-labeled collagen IV antibody, to confirm that cleft initiation was accompanied by a dynamic separation of adjacent nuclei at sites of collagen IV ingression where the cleft initiates (Fig. S9E, white arrows; Movie 4). In contrast, clefts in KO glands randomly initiated at the basement membrane but failed to progress and then rapidly regressed; nuclear separations appeared minimal (Fig. S9F, white arrows; Movie 5). We next compared these nuclear separation dynamics in E13.5 Btbd7 wild-type and KO SMGs. While nuclear separations between adjacent outer bud cells occurred frequently all around the bud periphery in wild-type glands, such dynamics were greatly decreased in Btbd7 KO salivary glands, with adjacent outer bud nuclei retaining relatively constant distances from one another (Fig. 4F,G). Taken together, these results indicate that Btbd7 is a key regulator of outer, but not inner, bud cell motility during branching morphogenesis. We propose that such Btbd7-mediated dynamic movements of outer bud cells enables the gland epithelium to maintain a high state of plasticity conducive to cleft formation and changes in tissue shape. Loss of Btbd7 results in reduced cell movements and separation, with minimal cleft formation.

To explore the mechanism by which Btbd7 positively regulates branching morphogenesis in vivo, we used an in vitro MDCK cell culture model where Btbd7 could be stably overexpressed using a doxycycline-regulated Tet-off system (see supplementary Materials and Methods) as a model for SMG Btbd7-expressing outer bud cells. To examine Btbd7 effects on MDCK cell-cell adhesion, we cultured GFP and GFP-Btbd7 overexpressing cells in suspension as hanging drops. Control GFP-overexpressing cells formed large aggregates after overnight culture and exhibited E-cadherin at cell-cell boundaries within the clusters. In contrast, Btbd7-overexpressing cells formed only small loosely adhered aggregates with barely detectable E-cadherin at cell-cell junctions (Fig. 5A). This cell-adhesion phenotype was mimicked by treatment with an E-cadherin function-blocking antibody (Fig. 5A), suggesting that the smaller aggregates formed by GFP-Btbd7 cells results from decreased E-cadherin at cell-cell junctions. We quantified these differences and observed significantly more single cells for samples overexpressing Btbd7 or treated with the E-cadherin antibody compared with control GFP cells (Fig. 5B). Thus, like SMG outer bud cells that express higher levels of Btbd7, MDCK cells overexpressing Btbd7 have less E-cadherin at cell-cell junctions and are less adhesive to one another.

Because Btbd7 in outer bud cells regulates cell motility in salivary glands, we performed time-lapse imaging of GFP-Btbd7 MDCK cells cultured in 2D to analyze their dynamics. Cells expressing Btbd7 appeared more spread within epithelial clusters and migrated faster than control GFP cells, and these effects could be mimicked by treatment with an E-cadherin function-blocking antibody (Fig. 5C).

We next analyzed the relationship between neighboring cells within epithelial clusters. In control GFP-expressing cell islands, neighboring cells usually remained attached to one another as they slowly migrated within the cluster, which we quantified as the change in distance between two initially adjacent nuclei during time-lapse imaging (Fig. 5D,E). In contrast to controls, initially adjacent cells in GFP-Btbd7 cell clusters often dynamically pulled apart from one another (Fig. 5D,E). Again, this effect was partially mimicked by treatment with the E-cadherin function-blocking antibody. We quantified the maximum displacement between nuclei for large populations of control and GFP-Btbd7 cells, and found that the maximum displacement distribution for GFP-Btbd7 cells was shifted to larger values when compared with control cells (Fig. 5F).

We searched for molecular mechanisms of Btbd7 regulation of E-cadherin and epithelial cell motility dynamics in MDCK cells. We detected no difference in E-cadherin mRNA levels between GFP and GFP-Btbd7-expressing MDCK cells (data not shown), suggesting that Btbd7 downregulates E-cadherin at the protein level in this model. To test this hypothesis, we evaluated E-cadherin turnover in control and Btbd7-overexpressing cells in the presence of cycloheximide to block new protein translation. E-cadherin turnover was more rapid in Btbd7-overexpressing cells, with loss of most E-cadherin at 4 h compared with 8 h in GFP control cells (Fig. 6A). We confirmed this result by immunostaining with an E-cadherin antibody specific to the extracellular region of E-cadherin. Only small remnant puncta of E-cadherin remained on GFP-Btbd7-expressing cells (Fig. S10A). If these cells were detergent-permeabilized prior to immunostaining, E-cadherin showed diffuse cytoplasmic localization, whereas control GFP cells had strong localization at cell-cell contacts (Fig. S10B). We conclude that elevated levels of Btbd7 enhance the internalization and subsequent turnover of the cell-cell adhesion molecule E-cadherin.

Transmembrane protein internalization and turnover can be regulated by the ubiquitin-proteasome pathway, where a series of ubiquitin ligases function together to ubiquitylate the protein substrate, resulting in its endocytosis and targeting to the proteasome (Fig. 6B). As high Btbd7 reduces E-cadherin at the cell surface, we examined the possible role of the ubiquitin-proteasome pathway. We first tested whether Btbd7 enhances E-cadherin ubiquitylation. Immunoprecipitation of E-cadherin from GFP versus GFP-Btbd7-expressing cells revealed elevated ubiquitylation of E-cadherin in GFP-Btbd7 lysates (Fig. 6C, Fig. S10C). Furthermore, such ubiquitylated E-cadherin was greatly decreased when cells were treated with PYR-41 (Yang et al., 2007), a pharmacological inhibitor of ubiquitin activation by E1 ubiquitin ligases (Fig. 6B,C, Fig. S10D). These results indicate that E-cadherin is more highly ubiquitylated when Btbd7 expression is elevated.

To dissect the relationship between Btbd7, ubiquitin-proteasome activity and loss of E-cadherin, we treated GFP-Btbd7 cells with inhibitors at other specific points in the pathway (Fig. 6B). Interestingly, E-cadherin localization at cell-cell contacts was partially restored by treatment with PYR-41 and dynasore (Macia et al., 2006), which inhibit ubiquitylation and dynamin-mediated endocytosis, respectively (Fig. 6B,D, white arrowheads). Although treatment with the general proteasome inhibitor lactacystin increased total E-cadherin levels in GFP-Btbd7 cells, its localization remained primarily cytoplasmic rather than at cell-cell contacts (Fig. 6D, yellow asterisk).

Finally, we tested whether inhibition of ubiquitin-proteasome activity in GFP-Btbd7 expressing cells could partially rescue cell-cell adhesion and decrease cell motility. Treatment of GFP-Btbd7 cells with PYR-41 and dynasore, but not lactacystin, resulted in increased formation of cell aggregates in hanging-drop cultures (Fig. 6E). Furthermore, inhibition of ubiquitylation or endocytosis, but not proteosomal degradation, also reduced the velocity of GFP-Btbd7 cells, as well as the separation of initially adjacent cells (Fig. 6F,G, Fig. S10E). Taken together, these in vitro studies indicate that Btbd7 enhances the ubiquitylation, endocytosis and degradation of E-cadherin in MDCK cells, which results in decreased cell-cell adhesion and increased dynamic cell motility.

We next tested whether our in vitro findings could explain outer versus inner cell behavior during branching morphogenesis. When wild-type SMGs were treated with PYR-41 or dynasore at E12.5 and E13.5, both inhibitors strongly reduced branching morphogenesis, with large distended buds and significantly fewer clefts (Fig. 7A-C).

We next characterized outer bud cell morphology in wild-type E13.5 SMGs treated with either PYR-41 or dynasore. In both cases, outer bud cells appeared more tightly packed and columnar, with higher axial ratios and similar appearance to Btbd7 KO SMGs (Fig. 7D,E). Furthermore, E-cadherin was more strongly localized to cell junctions in the outer bud cells, as indicated by fluorescence intensity line scanning (Fig. 7D,F), with no change in its localization in the inner bud cells. Overall, the inhibitor-treated wild-type SMGs phenocopied Btbd7 KO SMGs at E13.5.

To further test roles of ubiquitylation and endocytosis in E-cadherin localization, we treated E13.5 wild-type SMGs with EGTA to disrupt E-cadherin-mediated cell-cell adhesion, internalization, degradation (Troyanovsky et al., 2006) and dissociation of epithelial bud cells. Cell surface-exposed E-cadherin staining was lost within 3 h of EGTA treatment. After membrane permeabilization by detergent, immunostaining confirmed that E-cadherin was lost from cell-cell junctions and internalized to a diffuse cytoplasmic localization in SMG epithelial cells (Fig. S11A,B). In contrast, when SMGs were pre-treated with PYR-41 or dynasore to inhibit ubiquitylation or endocytosis, respectively, E-cadherin remained at cell-cell junctions, and buds appeared relatively intact (Fig. S11C). Inhibiting proteasomal degradation with lactacystin did not prevent EGTA-induced loss of E-cadherin from cell-cell contacts and epithelial bud dissociation, but elevated levels of internalized cytoplasmic E-cadherin (Fig. S11C). These findings mimic our findings concerning Btbd7-induced loss of E-cadherin. Taken together, these results extend to intact living glands our in vitro findings that perturbation of the ubiquitin-proteasome pathway can regulate E-cadherin retention at epithelial cell junctions.

Branching morphogenesis establishes the basic architecture of numerous organs, yet the specific mechanisms by which it reshapes mammalian tissues are still incompletely understood. Here, we report that Btbd7 is required for the in vivo branching morphogenesis of salivary gland, lung and kidney. We find that Btbd7 specifically functions at the tips of epithelial end buds, where it enhances cell motility and cell-cell adhesion dynamics without appreciably affecting these processes in inner bud cells (Fig. 7G). Furthermore, mechanistic analyses using an in vitro MDCK cell model of outer bud cell behavior unexpectedly revealed that Btbd7 promotes the ubiquitylation, internalization and degradation of E-cadherin with major effects on epithelial cell motility and cell-cell adhesion. These novel findings establish that Btbd7 can locally govern epithelial cell dynamics specifically within the outer tip cells during branching morphogenesis.

Collective cell migration has been implicated as a cellular mechanism that drives branching morphogenesis (Daley and Yamada, 2013; Ewald et al., 2008; Friedl and Gilmour, 2009). Some branching systems, including Drosophila trachea and angiogenic sprouting, initiate new branches through a form of invasive collective cell migration in which protrusive tip cells extend filopodia into the surrounding ECM (Ochoa-Espinosa and Affolter, 2012). In contrast, mammalian epithelial branching organs undergo collective migration in the absence of these invasive leading-edge membrane protrusions (Ewald et al., 2008; Schnatwinkel and Niswander, 2013). Live imaging of the salivary gland (Hsu et al., 2013; Larsen et al., 2006), lung (Schnatwinkel and Niswander, 2013), kidney (Riccio et al., 2016), mammary gland (Huebner et al., 2016) and pancreas (Shih et al., 2016) have all revealed that the peripheral bud tip cells in these organs are more dynamic compared with inner bud and ductal cells.

In some developing organs, mechanisms contributing to this differential cell migration have been identified. In salivary gland and pancreas, integrin-mediated signaling from the basement membrane appears to maintain enhanced velocity or directionality of outer bud cell migration (Hsu et al., 2013; Shih et al., 2016). Inhibition of integrin signaling switches the migratory behavior of outer cells to resemble the inner cells. Associated with the integrin requirement for elevated outer bud cell velocity in salivary glands, previous work suggested that Btbd7 is induced by the extracellular matrix protein fibronectin (Onodera et al., 2010), a matrix protein that is required for ex vivo branching of salivary gland, lung and kidney (Sakai et al., 2003).

In the developing kidney and mammary gland, growth factor-mediated signaling may play an analogous role. Increased Ret signaling at the tips of kidney tubules promotes cell migration (Riccio et al., 2016), and FGF receptor activation at the tips of mammary terminal end buds signals to stimulate tip cell migration (Huebner et al., 2016). Based on the results presented here, we now propose that Btbd7 is a crucial regulator of differential cell migration in the outer bud cells of developing salivary glands. Because genetic ablation of Btbd7 results in major disruptions of lung and kidney branching, we suggest that Btbd7 may play a similar role in the previously described differential cell migration of peripheral cells in these organs.

Our laboratory previously reported enrichment of Btbd7 mRNA in the clefts of branching salivary glands and lungs (Onodera et al., 2010), which differs from the antibody staining pattern reported here. We therefore performed single molecule FISH for Btbd7 mRNA, which established quantitatively that Btbd7 is ubiquitously expressed throughout the salivary gland epithelium. Differences between the in situ hybridization methods likely account for the different findings: the original study used several rounds of tyramide amplification to detect Btbd7 mRNA, which we believe resulted in nonspecific signals. smFISH is a highly specific technique used successfully previously to quantify mRNAs in a variety of developmental systems.

Importantly, the combination of uniform distribution of Btbd7 mRNA expression in the epithelium versus Btbd7 protein enrichment at the peripheral tips of branching end buds implies post-transcriptional regulation of Btbd7 for regional protein localization. We speculate that because the pattern of microperforations in the basement membrane at the tips of end buds in most major branching organs (Harunaga et al., 2014) resembles the Btbd7 protein localization in this study, epithelial protrusions known to extend through the microscopic holes may contact Btbd7-inducing factors such as fibronectin (Larsen et al., 2006; Sakai et al., 2003). Besides such fibronectin/integrin signaling, growth factors might also regulate Btbd7 protein levels by preferential communication to peripheral cells through these basement membrane microperforations at end-bud tips.

Our results also indicate that Btbd7 is a key downregulator of E-cadherin localization at cell-cell junctions in the dynamic outer bud cells of the developing salivary gland. In wild-type E13.5 glands, we observed reduced E-cadherin localization at cell-cell contacts between outer bud cells compared with inner cells, consistent with less mature cadherin-based adhesions in these highly migratory outer cells. Interestingly, Cdh1 mRNA also shows reduced expression in these cells, indicating some form of transcriptional downregulation that helps suppress E-cadherin levels in outer cells. In contrast, with deletion of Btbd7, E-cadherin protein was enhanced at cell-cell junctions of the outer bud cells, while remaining unchanged in inner bud cells. The outer bud cells also became more columnar and tightly packed, with minimal cleft formation. Because we did not observe a difference in Cdh1 mRNA by smFISH in the outer cells of Btbd7 KO versus wild-type salivary glands, Btbd7-mediated regulation of E-cadherin is likely post-transcriptional and post-translational, as suggested by our in vitro experiments in MDCK cells showing that Btbd7 promotes E-cadherin ubiquitylation and degradation.

Interestingly, previous studies of cell-cell adhesion molecule expression and localization across different stages of salivary gland development reported that cell-cell adhesions are primarily mature when the epithelium initially invaginates, but then undergo downregulation as the epithelium initiates rapid branching. Afterwards, during differentiation, cell-cell adhesions are re-established (Baker, 2010; Menko et al., 2002; Walker et al., 2008). In fact, a direct comparison of early E11.5 wild-type glands prior to initiation of branching to Btbd7 KO E13.5 glands revealed remarkably similar patterns of outer cell morphologies and E-cadherin localization.

Consequently, we propose a model in which Btbd7 acts on outer tip cells to transition them from a highly epithelial, strongly adhesive state to a more dynamic, less adhesive, migratory state conducive for initial cleft formation and branching (Fig. 7G). Indeed, inhibiting E-cadherin function increased MDCK cell migration rates and dynamics. Furthermore, a previous study in salivary gland organ cultures demonstrated that inhibition of E-cadherin had minimal effects on outer cell migration but significantly increased migration rates of the inner cells (Hsu et al., 2013). These results are clarified by our model in which the outer cells have reduced functional E-cadherin at cell-cell contacts, whereas the inner cells retain cell surface E-cadherin, which restrains their motility.

Surprisingly, we found using an MDCK cell model in culture that Btbd7 promotes the ubiquitylation, endocytosis and degradation of E-cadherin, and that this mechanism increases the migration dynamics of epithelial cells. Other BTB domain-containing proteins have been shown to interact with the ubiquitin-proteasome pathway by serving as adaptor molecules linking E3 cullin ubiquitin ligases to substrate proteins (Bennett et al., 2010; Chen et al., 2009; Genschik et al., 2013; Metzger et al., 2012; Perez-Torrado et al., 2006). An attractive hypothesis could propose Btbd7 as a direct link between an E3 ligase and a substrate molecule such as E-cadherin. However, we were unable to detect a direct interaction between GFP-Btbd7 and E-cadherin in MDCK cells (data not shown). Thus, Btbd7-mediated effects on E-cadherin ubiquitylation are likely to be indirect, possibly in conjunction with an E3 ligase such as Hakai that can directly target E-cadherin (Aparicio et al., 2012; Fujita et al., 2002). As Btbd7 protein is present in the nuclei of peripheral bud cells, future studies should explore its potential role in a broader nuclear program promoting epithelial cell dynamics. In terms of the in vivo relevance of the MDCK model studies, experimental inhibition of ubiquitylation and endocytosis in wild-type salivary glands blocked branching morphogenesis and resulted in similarly increased E-cadherin localization at cell-cell contacts of outer cells. We propose that ubiquitylation, endocytosis and proteolysis are the mechanisms by which Btbd7 downregulates E-cadherin to enhance the epithelial cell dynamics at the periphery of branching salivary glands associated with cleft formation.

More generally, we find that Btbd7 is essential for morphogenesis of two other organs with seemingly divergent modes of branching and eventual functions. For example, lung branching is highly stereotyped with three branching modes termed domain branching and planar or orthogonal bifurcation, generating an organ for gas exchange (Metzger et al., 2008). Kidney development melds two distinct embryonic tissues – the ureteric bud and metanephric mesenchyme – to form nephrons and connecting ducts for excretion of urine (Costantini and Kopan, 2010). Salivary gland branching involves a relatively stochastic process of progressive bud formation with final function involving water, protein and mucin secretion. Although these three different organs have divergent mechanisms of branching and ultimate functions, we found that all of them were highly dependent on Btbd7. In striking contrast, pancreas morphogenesis remains unaffected by this genetic ablation of Btbd7. Interesting future investigations of the commonalities versus differences in mechanisms underlying these patterns of dependence on Btbd7 should provide further insight into morphogenetic mechanisms.

In summary, we report a novel mechanism in which a single regulator, Btbd7, locally promotes the rapid turnover of a cell-cell adhesion molecule, E-cadherin, to increase outer cell plasticity in branching organs. Its function is crucial for salivary gland, lung and kidney branching morphogenesis.

The Btbd7 targeting vector was constructed by recombineering (Cotta-de-Almeida et al., 2003; Liu et al., 2003; Warming et al., 2005) and transfected into W4-129S6 mouse ES cells (Taconic) as described in the supplementary Materials and Methods. Primers used for the generation of the targeting construct and for confirmation of successful recombination in ES cells are presented in Tables S1 and S2, respectively. To generate chimeric mice, targeted ES cells were injected into 72 h blastocysts from C57BL/6 mice and implanted into pseudopregnant females. The offspring were mated to obtain KO mice as described in the supplementary Materials and Methods and shown in Fig. S3B. All procedures performed were approved by the NIDCR Animal Care and Use Committee.

DNA from tail biopsies was obtained using the DNeasy Blood and Tissue kit (Qiagen). Genotyping of animals was performed by PCR amplification with the primers listed in Table S3 using MiFi polymerase (BioLine) or Takara ExTaq polymerase (Takara/Clontech).

All mice used in this study were bred and euthanized according to NIDCR-approved animal study protocols. Submandibular salivary glands (SMGs), lungs, kidneys and pancreas were dissected from mouse embryos between embryonic day 11.5 and 17.5 (E0 was defined as the day of plug discovery). All dissected organs were cultured and bud numbers quantified as described in the supplementary Materials and Methods.

Immunostaining was performed essentially as previously described (Daley et al., 2009; Hsu et al., 2012; Larsen et al., 2003), with antibodies used listed in the supplementary Materials and Methods. smFISH for Btbd7, Cdh1, Gapdh or Emg1 was performed on E13.5 SMGs using Stellaris FISH probes (LGC Biosearch Technologies). mRNA puncta were quantified from confocal images using ImageJ and expressed as puncta per 100 µm². For details, see the supplementary Materials and Methods.

EdU incorporation assays were performed using a Click-iT EdU kit (Invitrogen) as described previously (Harunaga et al., 2014). For further details, see the supplementary Materials and Methods.

E13.5 SMGs were cultured overnight on 0.2 µm pore filters prior to live imaging. Glands were cultured in media containing a 1:100 dilution of OxyFluor (Oxyrase), 0.5 µg/ml Hoechst 33342 (Sigma) and 5 µg/ml labeled collagen IV antibody (Millipore) for 3 h, and live imaging was performed on an inverted Nikon A1R MP confocal system (Nikon). For cell migration measurements, outer and inner cell nuclei were manually tracked using the ImageJ MTrackJ plug-in. For details, see the supplementary Materials and Methods.

GFP or GFP-Btbd7 Tet-off MDCK II cells (Clontech) were cultured as previously described (Onodera et al., 2010). Doxycycline was removed from the media to induce GFP-Btbd7 expression 4 days before analysis. Detailed information on hanging-drop assays, immunostaining, immunoprecipitation, cell migration, quantification of cell migration and inhibitor studies is provided in the supplementary Materials and Methods.

MDCK II Tet-Off tTA control cells or MDCK cells expressing GFP-Btbd7 were cultured as previously described (Onodera et al., 2010) except without antibiotic and stained with 1 µg/ml Hoescht (Invitrogen). Cells were imaged for 6 h using an Axiovert 135TV microscope (Carl Zeiss) and cell migration quantified using either the ImageJ plugin mTrackJ or an automated MATLAB point tracking algorithm. For details, see supplementary Materials and Methods.

For all experiments, one-way or two-way analysis of variance (ANOVA) was performed with a Bonferroni multiple comparison post-test, as described in the figure legends, using GraphPad Prism5. All organ culture experiments were repeated three to five times with embryos from at least three pregnant mothers. All in vitro cell culture experiments involved at least three biological replicates with the number of technical replicates given in figure legends. Sample sizes of sufficient power were chosen based on similar published research. Although experiments were not randomized, all experiments with mouse embryos were blinded because investigators did not know the embryo genotype until after the experiment.

We thank Susan Yamada for technical advice and Jeff Hsu for technical assistance with ES cell colony picking. We also thank Glenn Longnecker, Ashok Kulkarni and the NIDCR Gene Targeting Core for ES cell injections to generate founder mice and the NIDCR Veterinary Resources Core for help with animal caretaking. Special thanks to Jennifer Symonds for the kind donation of tissue sections, and to Yoshihiko Yamada, Jennifer Symonds and Joshua Collins for thoughtful suggestions on the manuscript.