Afadin orients cell division to position the tubule lumen in developing renal tubules

The formation and elongation of epithelial tubules is a central feature of many organs. In the kidney, tubules have a single layer of epithelial cells surrounding a central lumen. Within each tubule, the presence of a continuous lumen is essential to its function: even a small discontinuity would block the passage of filtrate, thereby abolishing its excretory function. Yet how tubules develop a single, continuous lumen with high fidelity is not well understood.

Epithelial tubules are one of the most common tissue types in metazoa, and there is remarkable diversity in the structure and mechanisms by which tubules develop (Marciano, 2017; Iruela-Arispe and Beitel, 2013). Despite differences, a key process in tubulogenesis is the development of cell polarity, the segregation of functionally distinct plasma membrane domains. An epithelial cell can polarize along its apical-basal axis, which orients the cell relative to the underlying extracellular matrix. The apical surface faces the luminal space, the lateral surface contacts adjacent cells, and the basal surface contacts the basal lamina. Additionally, the individually polarized cells must coordinate their polarity with surrounding cells along the long axis of the tubule, in a process referred to as planar cell polarity.

In the developing kidney, epithelial tubules are derived from progenitor cells from two embryonic sources: the cap mesenchyme and the ureteric epithelium (Saxen, 1987). The cap mesenchyme gives rise to nephron segments from the glomerulus to the distal tubule, and the ureteric epithelium becomes the collecting duct. Whereas the latter arises as a branch from a pre-existing tubule, the cap mesenchyme must undergo a mesenchymal-to-epithelial transition to form a nascent epithelial tubule with de novo establishment of apical-basal polarity and generation of a lumen. Study of this process serves as an excellent model to elucidate mechanisms of polarization and tubulogenesis.

In previous studies, we have shown that once a lumen is created, elongation initially occurs by simultaneous extension of the lumen and additional de novo lumen generation (Yang et al., 2013). Shortly thereafter, the lumens of developing nephron tubules become continuous, and further elongation of the lumen coincides with elongation of the tubule (Yang et al., 2013).

The cellular mechanisms by which a continuous lumen is generated are poorly understood. In vitro studies have demonstrated that formation of a continuous lumen correlates with the axis of cell division (i.e. the orientation of the mitotic spindle) (Jaffe et al., 2008; Hao et al., 2010; Durgan et al., 2011). During mitosis, epithelia orient their mitotic spindle parallel to the apical surface and basal lamina. Defects in this spindle orientation correlate with the formation of multiple lumens in vitro (Durgan et al., 2011; Hao et al., 2010; Jaffe et al., 2008). How might this occur? Near the end of cytokinesis, the dividing cell remains interconnected by a bridge-like structure called a midbody. Recent studies have shown that intracellular vesicles carrying apical components localize adjacent to this site, and it has been suggested that the vesicles fuse to this site after abscission to lay down new apical membrane (Klinkert et al., 2016; Schluter et al., 2009; Li et al., 2014). Thus, misorientation of cell division could lead to mispositioning of new apical surface.

To date, the role of oriented cell division in de novo lumen formation and continuity during in vivo tubulogenesis has not been investigated. However, it has been examined for renal tubule elongation and maintenance, with varying results depending on the type of tubule and stage analyzed (Fischer et al., 2006; Saburi et al., 2008; Sims-Lucas et al., 2012; Nishio et al., 2010; Karner et al., 2009). Interestingly, a study analyzing the elongation of developing renal collecting ducts has shown that spindle orientation is random during development, and only becomes oriented after the neonatal period (Karner et al., 2009). However, as collecting ducts are derived from the ureteric epithelia, the mechanisms of tubulogenesis in early nephron tubules from the metanephric mesenchyme, which develop lumen formation de novo, remain unknown. Indeed, early-stage nephron tubules have not been examined for oriented cell division, during lumen formation or elongation.

In the current study, we examined mechanisms of lumen continuity by utilizing a mouse mutant that we had previously identified as having defects in lumen formation. Mice lacking the F-actin-binding protein afadin in developing nephron epithelia have both delayed lumen formation and discontinuous lumens (Yang et al., 2013). Afadin is a known component of adherens junctions and tight junctions in fully epithelialized cells, where it is an adaptor to a major class of adhesion receptors called nectins (Mandai et al., 2013). It also is an effector for the small GTPases Rap1 and Rap2 (Boettner et al., 2000; Linnemann et al., 1999; Monteiro et al., 2013), and has been shown to interact with a number of different proteins (Mandai et al., 2013). In addition to its role in cell junctions, afadin orients the mitotic spindle in Drosophila neuroblasts (Carmena et al., 2011; Wee et al., 2011) and cortical radial glia (Rakotomamonjy et al., 2017), and has been shown to interact with spindle machinery in vitro (Carminati et al., 2016).

To understand how afadin functions to promote a continuous lumen in tubules, we investigated lumen placement in vitro in 3D epithelial cell culture and in vivo in renal tubules. We show that afadin is required for lumen continuity in a 3D in vitro model of renal epithelial lumen formation, demonstrating that it localizes to the cortex overlying mitotic spindles. Lumen discontinuity is dependent on cell division, and afadin orients the mitotic spindle in vitro. In vivo in developing nephron tubules, a composite measurement of the angle of cell division reveals that cell division is not oriented. This seemingly incongruent result becomes understandable after analysis of the mitotic angle relative to the apical-basal axis. Our analysis demonstrates for the first time that cell division is oriented perpendicular to the apical-basal axis, and this specification of oriented cell division is dependent on afadin. These results support a model in which apical-basal spindle orientation controlled by afadin is necessary to ensure the fidelity of a continuous lumen and normal tubulogenesis.

During nephrogenesis, nephron progenitors from the cap mesenchyme condense to form a pretubular aggregate, and then undergo epithelialization to form a spherical structure with a central lumen called the renal vesicle. The renal vesicle then undergoes proliferation and morphogenesis to form an s-shaped tubule aptly called the s-shaped body. We have previously shown that afadin is required for normal lumen formation in developing nephrons (Yang et al., 2013). Mice with conditional deletion of afadin using the Pax3-cre transgenic line display both a delay in the onset of lumen initiation and then develop discontinuous lumens in developing nephrons (Yang et al., 2013). Deletion of the afadin gene (Afdn) specifically within nephron progenitors (Afdn^f/−; Six2-cre) causes similar defects, with mutant mice ultimately developing severe renal dysplasia and dying at 4-6 weeks of age (Yang et al., 2013). Fig. 1 shows that s-shaped bodies from Afdn^f/f; Six2-cre mice had numerous discontinuous lumens compared with control littermates (Fig. 1A-C, Fig. S1). The lumens were demarcated by Par6 (Pard6b) at the apical surface and ZO1 at apical junctions (Fig. 1B). Some of the later-stage, more elongated nephron tubules also had abnormal morphogenesis and more than one lumen transversely, appearing multi-layered (Fig. 1C).

To ensure that the discontinuous lumen phenotype was not secondary to the delay in lumen creation, we also examined mice with conditional deletion of Afdn using Wnt4-cre, which results in Cre-mediated recombination at the pretubular aggregate/renal vesicle stage. In these mutant mice, afadin was still present in renal vesicles and variably decreased in s-shaped bodies (Fig. S2A). These mutant kidneys had lumens of normal appearance in renal vesicles [27 of 27 renal vesicles from two Afdn^f/f; Wnt-cre postnatal day (P) 0 mice], but 30% of s-shaped bodies had discontinuous lumens (9 of 27 s-shaped bodies from two Afdn^f/f; Wnt-cre P0 mice) (Fig. S2B). The reduced penetrance of lumen discontinuities in Afdn^f/f; Wnt-cre mice compared with Afdn^f/f; Six2-cre mice is likely to be due to residual afadin in the former. These results indicate that the delay in lumen initiation and the lumen discontinuity represent two distinct phenotypes.

To determine whether afadin played a similar role in the ureteric epithelial tubules, we conditionally deleted Afdn using the HoxB7-cre line. As described earlier, these tubules arise by branching/budding from a pre-existing tubule, and their lumens form by extension of a pre-existing lumen, not de novo. We examined 22 3D reconstructions of ureteric tubules from three Afdn^f/f; HoxB7-cre neonatal mice, and all had fully continuous lumens (Fig. S3A). However, residual afadin was observed by immunostaining in nearly half of all ureteric tubules (48%, 13 of 27 from two Afdn^f/f; HoxB7-cre mutants) (Fig. S3B). Because half of the ureteric tubules did not appear to have afadin, it is reasonable to conclude that afadin does not have a major role in lumen continuity of ureteric epithelia.

To elucidate the mechanism by which afadin regulates lumen formation, we utilized a well-established 3D model of renal epithelial lumen formation employing Madin–Darby canine kidney (MDCK) cells (Marciano et al., 2011; O'Brien et al., 2001). When individual MDCK cells are cultured in a 3D gel of extracellular matrix, they proliferate and form fluid-filled spheres called cysts that are lined by a monolayer of polarized cells. Lumen formation can be studied at discrete stages that are marked, in part, by the localization of podocalyxin (Podxl1; Podxl) (Fig. 2A). Podxl1 is initially peripheral (Fig. 2Aa), and then is internalized to form intracellular vesicles (Fig. 2Ab, arrows). These vesicles fuse at the cell-cell junction, forming the apical membrane initiation site (AMIS) (Fig. 2Ac), which expands to become what is referred to as a pre-apical patch (PAP) (Fig. 2Ad), and ultimately forms an open lumen (not shown). This lumen expands as the size of the cysts increases (Fig. 2Ae). Using these stages, we found that afadin underwent a dynamic relocalization during lumen formation. Afadin was first localized to the cell-cell interface and was excluded from the AMIS (Fig. 2Ac). It remained excluded from the PAP and became restricted to apical cell junctions (Fig. 2Ad).

To determine the role of afadin in lumen formation, we generated MDCK cells stably depleted of afadin using shRNA. Afadin knockdown (Af KD) cysts were devoid of afadin by immunostaining (Fig. 2B, compare arrows in control and Af KD) and western blot (Fig. 2C). To control for off-target effects, we generated afadin knockout (Af KO) cell lines using gene editing (Fig. 2B,C). Afadin KD and KO showed similar phenotypes, as described below.

We first wanted to assess the role of afadin in the timing of lumen initiation using the MDCK cyst assay. Although we had previously shown that mice with conditional deletion of Afdn using Pax3-cre have a significant delay in lumen initiation (Yang et al., 2013), this phenotype could have arisen from several factors, including a delay in the mesenchymal-to-epithelial transition, differentiation, or lumen initiation. To study the process in vitro, we quantified the early stages (20 h) of lumen formation in two-cell cysts with and without afadin (Fig. 2D). This analysis revealed that among the early-stage afadin KO cysts there were fewer cysts with a single lumen at 20 h. The afadin KO cyst group had more cysts at the AMIS stage, suggesting that afadin is required for timely initiation of the PAP stage. The results demonstrate that afadin is required for timely lumen initiation in vitro.

In addition to a delay in lumen initiation, the most pronounced phenotype we observed was a multi-lumen phenotype at later time points (Fig. 2E,F), similar to the murine in vivo phenotype (Fig. 1). Indeed, it is this phenotype that will be the focus of the rest of the paper. Many of the 5-day cysts lacking afadin contained several lumens, as illustrated in Fig. 2E. We did note, however, that there was no obvious loss in polarity per se. The basolateral marker cadherin 6 and apical marker podocalyxin seemed to have normal localization compared with controls in 5-day cysts, and in two-cell cysts that had already formed a lumen (Fig. 2E). Similarly, ZO1, a marker of tight junctions, appeared to be correctly localized.

The total number of lumens within a cyst reflects a balance between the creation of new lumens and fusion of existing lumens. We found that at an early time point (2 days), nearly 50% of control cysts had developed multiple lumens (Fig. 3A). This percentage decreased over time (Fig. 3A), suggesting that lumen fusion predominates over the generation of new lumens in normal cysts, consistent with prior reports (Cerruti et al., 2013). To examine this process in cysts lacking afadin, we quantified the number of lumens in control and Af KO cysts at 2 days and 4 days (Fig. 3B). Consistent with our time course data, at 4 days, control cysts had an increase in single-lumen cysts (Fig. 3B) and a decrease in the number of lumens per cyst (Fig. 3C) compared with those at 2 days. However, in Af KO cysts there was both a small increase in the single lumen cysts and an increase in the number of cysts with >4 lumens (Fig. 3B, compare Af KO at day 2 with Af KO at day 4), such that the average total number of lumens per cyst was unchanged (Fig. 3C). Because there was an increase in the percentage of cysts with >4 lumens, this suggested that the multi-lumen phenotype may be, at least in part, due to the creation of new lumens.

We next sought to determine the cellular mechanisms that contribute to creation of excessive new lumens in Af KO cysts. Recent data has shown that cell division is coupled with lumen formation and lumen position (Bañón-Rodríguez et al., 2014; Jaffe et al., 2008). We examined whether afadin controls lumen number via a role in cell division. We blocked/inhibited cell division with thymidine treatment, which has previously been shown to slow cell division (Banon-Rodriguez et al., 2014), in control and Af KO cysts at 2 days. We added thymidine to 2-day cysts rather than 1-day cysts because 1-day cysts had few cells and generally zero or one lumen. The addition of thymidine caused a 15% decrease in the total number of lumens per cyst in Af KO cysts (P≤0.01) (Fig. 3C). Although this was a consistent but modest reduction, it suggested that the excessive number of lumens formed in the absence of afadin is linked to cell division, and led us to the next set of experiments.

Misorientation of the mitotic spindle has been associated with a multi-lumen phenotype (Durgan et al., 2011; Hao et al., 2010; Jaffe et al., 2008), which is thought to be due to mispositioning of the newly generated apical surface that occurs after cell division. To determine whether afadin contributes to the orientation of the mitotic spindle in renal epithelia, we first examined afadin localization during mitosis in early MDCK cysts. Afadin localized to the lateral cell cortex adjacent to the spindle poles, marked by α-tubulin, during metaphase and anaphase (Fig. 4Aa-Ac, compare with 4Ad-Af). Afadin also localized to the area surrounding the midbody during cytokinesis (Fig. S4).

We examined the orientation of cell division relative to the apical-basal axis in control and Af KO cysts at 5 days in both metaphase and anaphase. Anaphase was examined to limit errors because the spindle rotates in prometaphase (Reinsch and Karsenti, 1994). The spindle angle was measured relative to a radial vector (defined as the apical-basal axis) from the center of the cysts. The Af KO cysts had misorientation of mitotic spindles in both metaphase and anaphase (Fig. 4B,C). Because many of the Af KO cysts at 5 days had multiple, misplaced lumens, we also considered that the lumen itself could potentially act as a cue for spindle orientation, with misplaced lumens redirecting spindle orientation. Thus, we quantified the orientation of cell division in early-stage cysts (20 h) that contained zero or one lumen (Fig. 4D). These results showed that Af KO cysts had loss of spindle orientation very early in cyst formation. This suggested that the defect in orientation of cell division is primary, occurring prior to the formation of multiple lumens. Thus, we concluded that afadin regulates the apical-basal orientation of cell division in vitro in renal epithelial cells.

Because the absence of afadin caused a spindle orientation defect, its role in midbody placement, which forms during late cytokinesis, could not be independently assessed in 3D models (or in vivo). It is worth noting that the absence of afadin did not lead to multi-nucleated cells (not shown), demonstrating that it does not have an essential role in cytokinetic abscission.

Because disruption of spindle angle (as a result of loss of afadin) correlated with multiple lumens in vitro, and absence of afadin caused multiple, discontinuous lumens in vivo, we decided to examine orientation of cell division in developing nephron tubules. We first imaged transverse sections in 2D from developing control nephrons. We examined mitotic cells in anaphase in a late renal vesicle (Fig. 5A,A′) and developing tubule (Fig. 5B,B′) and observed that the orientation of division appeared to be parallel to the luminal/apical surface, i.e. perpendicular to the apical-basal axis of the epithelium. This is consistent with our observations in MDCK cysts (see Fig. 4B). We also observed that the midbody, the transient structure that is formed near completion of cytokinesis, is placed apically in developing nephrons (Fig. 5C,C′). This also is consistent with in vitro data that have demonstrated that cytokinesis occurs asymmetrically in epithelia (Reinsch and Karsenti, 1994; Morais-de-Sá and Sunkel, 2013), and suggests that cell division occurs perpendicular to the apical-basal axis.

Within tubules, oriented cell division occurs when the mitotic spindle (and nuclear chromatid) vector is parallel to the longitudinal axis and perpendicular to the apical-basal axis of the tubule (Fig. 6A, a). Defects in either longitudinal (Fig. 6A, b) and/or apical-basal orientation (Fig. 6A, c) lead to misoriented cell division. Mitotic spindle orientation in dividing renal epithelial cells can be determined by measuring the angle (θ) between the vector comprising the long axis of the tubule (the reference vector) and the vector of the mitotic spindle/chromatids, shown schematically in Fig. 6B. Additionally, to assess its apical-basal orientation specifically, one can also measure a projected angle relative to the apical-basal axis of the epithelium (φ), as shown in Fig. 6C (Tang et al., 2011). To measure these angles, we generated 3D reconstructions of z-stacks from immunostained developing nephrons. We analyzed NCAM-positive tubules, which included s-shaped bodies and slightly more mature tubules. Measurements of earlier stages were not possible owing to difficulties in determining the reference vector.

A representative z-stack of a control tubule (Fig. 6D) shows that although chromatid/spindle vectors do not appear to be oriented parallel to the long axis of the tubule (Fig. 6Db,Dc), the φ angles are oriented at almost 90° relative to the apical-basal axis (Fig. 6Db1,Dc1). It also illustrates that dividing cells ‘round up’ and move near the center of the tubule (Fig. 6Db1,Dc1), as has been demonstrated in epithelial cultures (Reinsch and Karsenti, 1994) and in elegant imaging studies of developing ureteric epithelium (Packard et al., 2013).

Using our 3D reconstructions, we measured the orientation of cell division using the angle θ and the apical-basal angle φ (Fig. 6E,F). In control nephron tubules, θ appeared to be uniformly distributed (P=0.34, D'Agostino and Pearson normality test). This result in developing nephrons is similar to previously published results for developing collecting ducts, which also do not have oriented cell division (Karner et al., 2009). Our measurements of φ added additional information, showing that the orientation of cell division in the apical-basal axis was not uniformly distributed (P<0.001, D'Agostino and Pearson normality test), but skewed toward higher angles (skewness= −1.6). Thus, the apical-basal component of cell division is oriented.

We next examined the role of afadin in oriented cell division in vivo. In Afdn^f/f; Six2-cre tubules, analysis of θ showed that spindle orientation distribution across the entire angular range was not different from control tubules. However, there was a difference in φ between controls and mutant tubules (mean±s.d.: 77±8 and 57±23, P<0.001) (Fig. 6E), with mutants having a more uniform distribution of φ. Whereas control tubules had no mitoses with φ<50°, 35% of mutant tubules had φ<50° (Fig. 6F). Together, our data demonstrate in vivo the importance of afadin in regulating apical-basal orientation of cell division and continuous lumen formation.

Lumen continuity is essential to renal tubule function, but our knowledge of the mechanisms that promote its formation is limited. In addition, the role of oriented cell division in generating a continuous lumen in vivo is unknown, although it has been studied in other contexts (Fischer et al., 2006; Saburi et al., 2008; Sims-Lucas et al., 2012; Nishio et al., 2010; Karner et al., 2009). In this study, we sought to identify cellular mechanisms that regulate the generation of a continuous lumen in vivo in renal tubules. We demonstrate that afadin determines lumen continuity and directs spindle orientation, which we find to be highly oriented perpendicular to the apical-basal axis.

We initially utilized a 3D cyst model in which the absence of afadin in vitro recapitulates the phenotype of multiple discontinuous lumens previously described in developing renal tubules. Specifically, there is a delay in lumen formation, and then multiple, discontinuous lumens are generated when afadin is deleted from MDCK cell cysts. We found that once the initial lumen is generated, the cysts have a clear demarcation of apical and basal plasma membranes, despite their multiple lumens, suggesting that afadin is required to position the apical surface.

The 3D in vitro studies of lumen generation suggest that both luminal fusion and creation of new lumens contribute to final lumen number. During early cystogenesis, nearly half of all cysts have more than one lumen, generally two, but this proportion dramatically decreases over time. This suggests that fusion to a single lumen predominates over the generation of new lumens in control cysts.

In contrast, cysts lacking afadin have a high number of lumens during early cystogenesis, which do not decrease over time, at least initially (from day 2 to day 4). Although the percentage of afadin-depleted cysts with a single lumen increases over time, this is offset by increased lumen number in the remaining cysts. Together, these data suggest that luminal fusion is not blocked in the absence of afadin, and that additional new lumens are generated.

There is growing evidence in the literature that cell division plays an important role in the creation of new apical/luminal surfaces. Indeed, our in vitro experiments show that inhibition/slowing of cell division decreased the number of lumens in cysts lacking afadin. Studies of the first mitotic event in MDCK cysts have demonstrated that during cytokinesis, vesicles carrying apical membrane components localize in close proximity to the cytokinetic bridge microtubules and then fuse to the apical membrane initiation site at the cell-cell interface between the newly divided cells (Klinkert et al., 2016; Schluter et al., 2009; Li et al., 2014). After the initial lumen is formed, in subsequent mitotic events, the axis of cell division (i.e. the orientation of the mitotic spindle) is oriented parallel to the basal lamina and apical surface in vitro. Both polarity proteins and the spindle machinery direct spindle orientation (Schober et al., 1999; Wodarz et al., 1999; Hao et al., 2010), and loss of one of these proteins leads to both misoriented spindles and multiple lumens. This suggests that the axis of cell division is important for maintaining the position of the apical surface, thus allowing for further extension of a single lumen after cell division.

It is important to note that although cell division is generally oriented in MDCK cysts, as mentioned previously, many cysts initially develop more than one, generally two, lumens. However, corrective mechanisms, e.g. luminal fusion, exist to return the cyst to a single lumen state. It is likely that the defects in the afadin-depleted cysts overwhelm the capacity of these corrective mechanisms.

In contrast to in vitro studies, examination of oriented cell division in renal tubules has led to differing results (Fischer et al., 2006; Saburi et al., 2008; Sims-Lucas et al., 2012; Nishio et al., 2010; Karner et al., 2009). These discrepancies might be due to differences in tubule stage (early and late developmental stages, and mature tubules), location (cortical versus medullary), and type of tubule. Our own studies of developing nephron tubules found that overall cell division is not oriented, as measured by the angle (θ) between the long axis of the tubule and the spindle vector. Similar results have been obtained in developing ureteric epithelial tubules (Karner et al., 2009), suggesting that random orientation of θ during embryogenesis might be a general feature of renal epithelia. Our result also initially suggested correct spindle orientation is not required for lumen continuity (or tubule morphogenesis), which implied that other factors, such as cellular movements, might be responsible for generating tubules with a continuous central lumen. However, spindles can be oriented by planar polarity and apical-basal polarity, and measurement of θ encompasses both. Because only defects in apical-basal polarity would be expected to cause lumen discontinuity, we performed additional analysis to measure the apical-basal projection angle (φ). Measurement of φ showed that cell division is consistently oriented nearly perpendicular to the apical-basal axis, indicating that spindles are indeed oriented by apical-basal polarity in embryonic tubules. Consistent with this, we found that midbody placement in these tubules is apical, as has been reported for in vitro epithelia (Morais-de-Sá and Sunkel, 2013). Thus, in nephron tubules, although overall cell division is not oriented, the component corresponding to apical-basal polarity is oriented.

Our results also demonstrate that afadin mutants have disrupted apical-basal orientation of cell division and discontinuous lumens in developing nephron epithelial tubules and in in vitro cysts. Thus, defects in the apical-basal orientation of cell division could contribute to the luminal defects in afadin mutants. Because spindle angles are generally measured relative to the basal, not apical, surface of a tubule or cyst, we considered that perhaps the multiple, ectopic apical surfaces/lumens in afadin mutants were acting as ‘orienting cues’, which would make spindles appear misoriented with respect to the basal surface. However, examination of early 3D cysts showed that those lacking afadin have spindle misorientation even prior to the multiple lumen phenotype.

Although this study was conducted to examine regulation of lumen continuity and placement, it is difficult to fully separate lumen continuity from tubule elongation/morphogenesis, as these processes are likely to be interrelated. Some late-stage nephron tubules lacking afadin have morphogenetic defects, indicating that lumen placement correlates with morphogenesis. However, in afadin mutant mice, defects in lumen continuity precede morphogenetic defects. Additionally, afadin KO cysts have abnormal lumen placement with normal spherical structure, indicating that a morphogenetic defect does not cause the defect in lumen placement and spindle orientation in vitro. Thus, abnormal lumen placement in the absence of afadin is likely to be the primary defect. Our results are most consistent with a model in which absence of afadin results in loss of apical-basal spindle orientation, which leads to multiple, misplaced lumens, contributing to morphogenetic defects, as shown schematically in Fig. 6G.

Together, our results support a model in which afadin positions the placement of newly generated apical/luminal plasma membrane at least in part through its role in spindle orientation, leading to lumen continuity. This novel finding for the role of afadin requires further mechanistic study and is the focus of our ongoing work. Additionally, although our analysis has focused on the role of spindle orientation and apical-basal positioning of the lumen, afadin has a well-known role in adherens junction formation and has also been implicated in directional cell migration (Mandai et al., 2013; Sawyer et al., 2011). Both of these processes could also affect lumen continuity and/or tubule morphogenesis in vivo. Indeed, cell movements have been shown to resolve lumen discontinuities in in vitro epithelial tubes (Kim et al., 2015). Future studies, particularly those using live imaging, will be needed to elucidate the roles of these various processes in tubulogenesis.

We crossed conditional Afdn (Afdn^f/f) females to Afdn^f/+; Six2-cre^Tg (Kobayashi et al., 2008) males to obtain Afdn^f/f; Six2-cre mutant mouse embryos. Mice with conditional deletion of afadin with Wnt4-cre (Mugford et al., 2009) and HoxB7-cre (Yu et al., 2002) were also generated. Mice were maintained on mixed genetic backgrounds and genotyped by standard PCR. Embryonic day (E) 17.5 kidneys of both sexes were harvested and fixed for 2 h in 4% paraformaldehyde in PBS. Procedures were performed according to UTSW-IACUC-approved guidelines.

MDCK type II cells were grown and imaged as described previously (Marciano et al., 2011; Mostov and Deitcher, 1986) and tested for mycoplasma contamination. Afadin shRNA was based on published shRNA from Y. Takai (Sato et al., 2006) and is shown in Table S1. For the control, we utilized a scramble shRNA, as we have done previously (Marciano et al., 2011). shRNAs were cloned into pLKO.1 puromycin. Viral transduction of MDCK cells was performed as previously described (Bryant et al., 2010).

For CRISPR knockout lines, sgRNA oligonucleotides were designed using the CRISPR Genome Engineering Resources (http://crispr.mit.edu/). Knockout (KO) clones were generated and selected according to a published protocol (Cong and Zhang, 2015). Clones were verified for loss of afadin using western blot. The sgRNAs used to generate clones for this study are included in Table S1.

For cysts treated with thymidine, we first performed a dosage curve to establish the concentration of thymidine that would inhibit MDCK cell proliferation but not result in cell death by 48 h (data not shown). We added this dose (2 µM thymidine) to cysts at 48 h, and analyzed the effect on lumen number 48 h later.

Immunostaining was performed as previously reported (Yang et al., 2013). Briefly, kidney sections were permeabilized with 0.3% Triton X-100/PBS (PBST) and blocked with 10% donkey sera/PBST. For cells, samples were permeabilized with 0.3% PBST and blocked with 0.7% fish skin gelatin in TBS with 0.025% saponin (TFS). Samples were incubated with primary antibodies overnight (4°C), then with fluorophore-conjugated secondary antibodies. Kidney sections were mounted with Prolong Gold (Invitrogen). Confocal imaging was performed on a Zeiss LSM510 META laser scanning confocal microscope. Images were analyzed using ImageJ software. Images were minimally processed and re-sampled to 300 dpi using Adobe Photoshop.

MDCK cysts were lysed in ice-cold buffer containing 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, 20 mM sodium phosphate (pH 7.2), 150 mM NaCl, 50 mM Tris-HCl, 1 mM NaF, 1 mM Na₃VO₄, 1 mM PMSF and a protease inhibitor cocktail (Roche). Lysates were centrifuged at 17,000 g for 15 min at 4°C to remove insoluble aggregates, and SDS-PAGE and western blotting were performed.

For kidney immunostaining, primary antibodies were used at 1:100 unless stated otherwise: afadin (Sigma, A0224), NCAM [Developmental Studies Hybridoma Bank (DSHB), 5B8; GeneTex, H28-123], α-tubulin (AbD Serotec, MCA77G), Par6b (SCBT, sc-67392). For cell culture immunostaining, primary antibodies were used at 1:200 unless stated otherwise: afadin (Sigma, A0224), α-tubulin (AbD Serotec, MCA77G), Podocalyxin (DSHB, 3F2/D8), ZO1 (Millipore, MABT11), cadherin 6 (a gift of G. Dressler, University of Michigan, MI, USA), and Phalloidin 647 (Invitrogen, 42008A). For cell western blots, we used afadin (1:1000, Sigma, A0224) and β-actin (1:5000, Abcam, ab8227) antibodies with HRP-conjugated secondary antibodies (Jackson Immunoresearch) at 1:5000-1:10,000.

In MDCK cells, immunofluorescence was performed with antibodies to α-tubulin and podocalyxin to visualize the mitotic spindle and apical surface of cysts. Cysts were imaged using a Zeiss LSM 510 META confocal laser-scanning microscope at the largest cross-sectional diameter (equatorial plane). Spindle coordinates (x₁,y₁) and (x₂,y₂) of a mitotic cell were used to generate vector V_s. In 5-day cysts, the midpoint (x_m,y_m) of V_s was calculated and the center (x_c,y_c) of each cyst was determined using an ImageJ macro. A radial vector V_c was generated from (x_c,x_c) and (x_m,x_m). The projected angle φ between V_s and V_c was defined as the apical basal spindle angle, and was calculated as φ=acos((V_c · V_s)/(||V_c|| ||V_s||)). In early cysts (two cells), the reference vector was obtained from the cell-cell interface, and calculations were performed similarly.

In vitro experiments were performed in triplicate and are representative of at least two similar experiments. All data shown are mean±s.d. Statistical significance was performed using an unpaired, two-tailed Student's t-test unless stated otherwise.

We thank G. Dressler (U. Michigan) for the cadherin 6 antibody, N. Tang (NIBS, Beijing) for advice on spindle measurements, the UTSW Live Cell Imaging Center, and T. Carroll and M. Shiloh (UTSW) for reviewing the manuscript. We thank A. McMahon (USC) for Wnt4-cre mice.