The overall pattern of the developing kidney is set in large part by the developing ureteric bud/collecting duct system, and dysgenesis of this system accounts for a variety of clinically significant renal diseases. Understanding how the behavior of cells in the developing ureteric bud/collecting duct is controlled is therefore important to understanding the normal and abnormal kidney. Dact proteins have recently been identified as cytoplasmic regulators of intracellular signaling. Dact1 inhibits Wnt signaling, and Dact2 inhibits transforming growth factor (TGF)-β signaling. Here, we report that Dact2 is expressed in developing and adult mouse kidneys, specifically in the ureteric bud/collecting duct epithelium, a structure whose morphogenesis is controlled partially by TGF-β. When small interfering RNA is used to knock down Dact2 expression in collecting duct cells, they show some constitutive phospho-Smad2, undetectable in controls, and elevated phospho-Smad2 in response to TGF-β. They also show defective migration and, in a monolayer wound-healing assay, they fail to assemble a leading edge “cable” of actomyosin and advance instead as a disorganized mass of lamellipodium-bearing cells. This effect is seriously exacerbated by exogenous TGF-β, although control cells tolerate it well. In three-dimensional culture, Dact2 knockdown cells form cysts and branching tubules, but the outlines of the cysts made by knockdown cells are ragged rather than smooth and the branching tubules are decorated with many fine spikes not seen in controls. These data suggest Dact2 plays a role in regulating morphogenesis by renal collecting duct cells, probably by protecting cells from overly strong TGF-β pathway activation.
- Dapper, cell motility
- wound healing
the internal anatomy of the kidney is determined in large part by the development of the ureteric bud, which branches to form a tree-like collecting duct system and also induces the formation of nephrons, thus determining their positions (reviewed in Ref. 58). Many clinically important congenital diseases of the kidney, including supernumerary ureters, multicystic disease, autosomal recessive polycystic disease, and Kallman's syndrome, arise from abnormal development of the collecting ducts (reviewed in Refs. 43 and 70). Understanding the mechanisms of ureteric bud morphogenesis, especially how cell behavior is regulated, is therefore critical to understanding normal renal development. (8, 23, 46, 51).
Typically, the morphogenetic behavior of epithelial cells is controlled by the integrated influences of multiple positive and negative paracrine and autocrine signals. In the case of branching epithelia such as the ureteric bud, FGFs, HGF, and GDNF exert a generally positive, branch-promoting influence, various bone morphogenic proteins (BMPs) can have positive or negative effects, while transforming growth factor (TGF)-β generally inhibits branching but may promote epithelial-mesenchymal transition (reviewed in Ref. 8). Although most effort in the field of branching epithelia has focused on extracellular signals and their receptors, recent work in insects and mammals has highlighted the importance of intracellular modulators of signaling. Sprouty, for example, attenuates FGF signaling and is important in keeping the response of an epithelium to FGFs under control; without Sprouty, the airways of Drosophila melanogaster produce abnormally many sprouts instead of a well-spaced tree (5, 22). Epithelial morphogenesis in the lungs and kidneys of mammals is also abnormal in the absence or superabundance of Spry proteins, which are the mammalian orthologs of Sprouty (7, 33, 64, 65).
Recently, proteins of the Dact family were identified as cytoplasmic attenuators of signaling through Wnt and TGF-β pathways. Dact1 inhibits Wnt signaling by both the canonical and planar cell polarity pathways, apparently by promoting the degradation of Disheveled (74) and thus removing a critical connection in these pathways (Fig. 1). Dact2 modulates the planar cell polarity pathway in zebrafish (67). It does not interact at all with the canonical pathway in mammalian cells, though, and inhibits the planar cell polarity pathway only weakly, and only when expressed at very supraphysiological concentrations (63). In addition to inhibiting planar cell polarity signaling, Dact2 (but not Dact1) inhibits signaling by proteins of the TGF-β family by targeting type I TGF-β receptors for lysosomal degradation (63) so that the receptors cannot go on to activate Smad2/3; this effect seems to be the main function of Dact2 in mammalian cells. No functional data are yet available for Dact3 (15).
The Sprouty story highlighted the importance of cytoplasmic attenuators of positive signals to epithelial morphogenesis, so it seems likely that cytoplasmic attenuators of negative signals will also regulate epithelial behavior. We therefore investigated the expression of Dact2 in the developing ureteric bud/collecting duct of the kidney, an epithelium whose branching morphogenesis is known to be inhibited by TGF-β signaling (4, 35, 54). We found Dact2 to be expressed in the collecting duct system throughout its development. Furthermore, we found that RNA interference (RNAi)-mediated knockdown of Dact2 expression in collecting duct cells makes them hypersensitive to TGF-β and alters their behavior so that they migrate more as individuals than as a coherent epithelium, and are consequently deficient in wound healing, and they produce numerous abnormal spikes in three-dimensional tubule culture. These findings suggest that Dact2 functions to control the morphogenetic behavior of ureteric bud cells.
MATERIALS AND METHODS
This work involved no experiments on living animals. Mouse tissues were obtained from embryos of mice maintained in premises, and by staff, licensed for this purpose by the UK Home Office, and the animals were killed by a method specified in Schedule 1 of the Animals (Scientific Procedures) Act 1986.
RNA was isolated using the SV Total RNA Isolation system (Promega, Southampton, UK). Conventional end-point RT-PCR was performed to confirm the expression of Dact2 in mouse embryonic kidney and in inner medullary collecting duct (mIMCD3) cells. The primers were made according to sequence RefSeq BC058740; the forward primer was GACTACGAGCCGCACTGG, and the reverse primer was GCAGGAGGTGGACAGAGAAC. The PCR amplicons were 315 bp in size. PCR was performed in a Techne TC-312 thermal cycler (94°C for 5 min, then cycles of 94°C for 1 min, 55°C for 1 min, 72°C for 1 min, then a final extension of 72°C for 10 min). Real-time PCR was used to establish the expression profiles of Dact2 on mouse kidneys at embryonic day 11.5 (E11.5), E14.5, E15.5, postpartum day 14 (P14), and adult stage. Expression levels were obtained by normalization to a housekeeping gene, Actb. The forward primer for Dact2 was AGCTGGATGTGAGCAGGTCT, and the reverse primer was GCAGGAGGTGGACAGAGAAC; for Actb, the reference sequence was NM_007393, the forward primer was GATCTGGCACCACACCTTCT, and the reverse primer was GTACATGGCTGGGGTGTTG. Three independent experiments were carried out in 96-well plates on a DNA Engine Opticon (95°C for 15 min, then 40 cycles of 94°C for 15 s, 55°C for 30 s, 72°C for 30 s, MJ Research). Each run included at least two no-template controls. Standard curves were generated using serial twofold dilutions of cDNA containing the sequence of interest for every primer pair.
In situ hybridization.
Riboprobes were made by in vitro transcription using digoxygenin (DIG)-labeled dNTPs (Roche), using linearized plasmids as templates. Labeled riboprobes were isolated by precipitation in LiCl/ethanol. Mouse tissues were fixed in ice-cold methanol (10 min) then in 4% formaldehyde (from paraformaldehyde) overnight, then processed using the protocol of Wilkinson (68) with the modification that Protect RNA (Sigma) was used in all solutions after the proteinase K step. After hybridization and washing, tissues were blocked using 10% sheep serum (Sigma) and 2% wt/vol blocking powder (Roche) in Tris-buffered saline (TBS), then incubated in 1/2,000 alkaline phosphatase anti-DIG in blocking solution overnight. Antibody binding was visualized using 9.4 mg/ml BCIP and 18.75 mg/ml nitro blue tetrazolium in developing buffer (0.1 M NaCl, 0.1 M Tris·HCl, pH 9.5, 50 mM MgCl2, 0.1% Tween 20).
RNAi knockdown in mIMCD3 cells.
mIMCD3 cells were obtained from ATCC (ATCC line no. CRL-2123). They were used either untransfected or transfected with SureSilencing Dact2 short-hairpin (sh) RNA plasmids (Superarray Bioscience) carrying sequences representing one of two targets on the Dact2 mRNA, TGGATGTGAGCAGGTCTTCTT and AAGGTCAGAAGAAGAGTTAGT, or a control scrambled sequence, GGAATCTCATTCGATGCATAC, that did not match any mouse cDNA (by a BLAST search), under the control of a U1 promoter. Transfection of cells was performed using Metafectene Pro (Biontex) according to the manufacturer's directions, using 2 μg plasmid with 6 μl Metafectene in 100 μl serum-free and antibiotic-free culture medium. Cells were cultured in “G418 culture medium,” which consisted of a 1:1 mix of F-12 Ham's nutrient mixture and Dulbecco's modified Eagle's medium, with 10% fetal bovine serum, 1% penicillin and streptomycin, and 100 μg/ml G418, and individual colonies were picked and grown as separate clones. Untransfected cells were cultured in the same basic medium but without the G418. Cells were maintained at 37°C, 5% CO2; they were passaged when confluent and not used beyond passage 18 (because pilot experiments had indicated that older cultures sometimes grew unusual clones of large, multinucleate cells).
Suspensions of cells were released from their wells by trypsinization, washed, resuspended in medium for counting, centrifuged, and resuspended to 107 cells/ml in SDS sample buffer (62.5 mM Tris·HCl, 2% SDS, 10% glycerol, 50 mM DTT, 0.1% bromophenol blue). Samples were subject to PAGE on a 7.5% acrylamide gel and electroblotted onto nitrocellulose (Amersham). The membrane was blocked for 1 h in blocking buffer (5% dry milk, 0.1% Tween 20 in TBS), washed 3× in TBS, 0.1% Tween 20, and incubated with 1:1,000 anti-phospho-Smad2 (Cell Signaling Technology) in 5% bovine serum albumin, 0.1% Tween 10 in TBS overnight at 4°C. It was then washed again and probed with 1:2,000 horseradish peroxidase (HRP)-anti-rabbit (Cell Signaling Technology) in 10 ml blocking buffer; the part of the membrane with biotin-labeled molecular weight markers was incubated separately in HRP 1:1,000 anti-biotin (Cell Signaling Technology). The blot was visualized with LumiGlo (Cell Signaling Technology) according to the manufacturer's instructions. It was then washed and reprobed with 1:30,000 rabbit anti-GAPDH as a loading control (Sigma), washed, incubated with 1:4,000 HRP anti-rabbit, washed, and visualized with LumiGlo. Band intensity was quantified using Image J.
In vitro wound-healing assay.
For the in vitro wound-healing assay, 1 × 106 cells/well or, for some experiments where greater confluency was desired, 1.5 × 106 cells/well, were washed 3× in PBS and plated in a six-well plate at 37°C. For cells containing Dact2-targeting or control shRNA plasmids, the medium contained G418 as described above. The following day, a straight scratch wound was created by a 10-μl pipette tip held perpendicular to the coverslip, the clearance of cells from the “wound” being checked by phase-contrast microscopy. Cells were then washed in fresh medium (±G418 as appropriate) and were filmed with a Zeiss Axiovert 200 time-lapse microscope using a ×20 objective. Images were taken at the same position in each well at 15-min intervals over 24 h. Stacks of images of the same well were converted into an AVI movie for the supplementary material, the original stacks being used for quantification of wound closure (supplemental material for this article is available online at the journal website). Wound areas were also measured at 2-h intervals by ImageJ's “measure area” feature. Where appropriate (see results), TGF-β was applied at 10 ng/ml, or Rac was inhibited using 75 μM NSC23766 (13, 16). Some cultures were fixed in 4% formaldehyde (made freshly from paraformaldehyde) 2–6 h after wounding, rinsed with PBS, permeabilized with 0.5% Triton X-100 in PBS for 30 min, then stained in 1/100 FITC-phalloidin (Sigma) in PBS for 1 h, and washed in PBS before being mounted in 50% glycerol on PBS and being examined on a Zeiss Axiovert microscope. Image-J 1.37v was used to measure the length of each apparently continuous (at optical resolution) segment of actin cable (see results for an explanation) using the Segmented Line tool, and each measurement was recorded in micrometers.
Three-dimensional collagen gel cell culture.
Collagen I matrix was prepared by mixing 20 μl of 10× PBS, 2.3 μl of 1 N NaOH, 77.7 μl of distilled water, and 100 μl of rat tail collagen type I (BD Biosciences) for each well in a 24-well plate. These components were mixed together in a sterilized tube on ice. Cells (3,000, or for some experiments, 5,000, synchronized by FBS starvation for the preceding 24 h) were added to the collagen mixture and seeded in each well. This low cell density was used to avoid cell-cell contact when cells were seeded. Two hundred microliters of the cells/collagen I gel mixture was plated onto a coverslip at the bottom or a 24-well plate and incubated in a humidified incubator at 37°C for 20 min. Five hundred microliters of G418 culture medium containing 25 ng/ml HGF were carefully added into each well. Cells in collagen gel matrices were cultured at 37°C with 5% CO2, the medium being changed every 3 days. After 7 days, cells were fixed in 4% formaldehyde (made freshly by dissolving paraformaldehyde).
Cells were seeded on sterile 13-mm coverslips in 24-well plates at 2.2–2.5 × 105 cells/well. They were cultured overnight, washed twice with PBS, then fixed with −20°C methanol for 10 min. After fixation, the cells were washed three times with PBS, incubated with mouse anti-vimentin (Abcam) or anti-E-cadherin (BD Transduction Laboratories) at 10 μg/ml, anti-total-β-catenin (catalog no. 61054, BD Transduction Laboratories) at 1/100, or anti-active-β-catenin (catalog no. C1202, AG Scientific) at 1/100, at 4°C for 2 h. The cells were then washed three times with PBS and incubated with anti-mouse IgG-FITC-conjugated secondary antibody (1/100, Sigma) for 2 h at 4°C. After three washes with PBS, the coverslips were mounted on slides, inverted, with DAPI mounting solution (Vectashield, Vector Laboratories). Counting of random images was performed using the ImageJ Cell Counter plug-in.
Dact2 mRNA is expressed in the ureteric bud/collecting duct system of developing kidneys and in the collecting duct-derived mIMCD3 cell line.
Conventional RT-PCR was used to determine whether dact2 is expressed in any part of E14.5 embryonic kidneys, an age chosen because it includes components at almost all stages of nephron and stromal development (31). A clear band representing dact2 cDNA was detected (Fig. 2A). Whole mount in situ hybridization showed that dact2 was expressed in the branching ureteric bud/developing collect duct system of embryonic mouse kidneys (Fig. 2, C–E). There was no apparent difference in levels of expression between the ureteric tips and stalk regions, although expression is diminished in the ureter itself. Dact2 was detectable as early as E11.5 (Fig. 2C) and became stronger with developmental time, the pattern being very clear by E14.5 (Fig. 2E). This rise in strength of expression was supported by quantitative RT-PCR for dact2 mRNA at different stages of development, which showed a rise of dact2 mRNA (relative to the housekeeping gene actb) from E11.5 to a peak at E17.5 and then a fall to an adult expression at about half the level seen at E17.5 (Fig. 2H). The ureteric bud-specific pattern of expression seen in kidneys isolated directly from E14.5 embryos could also be observed in kidneys removed from embryos at E11.5 and grown in organ culture for 3 days (Fig. 2G).
The mIMCD3 cell line is derived from mouse inner medullary collecting duct (53) and has been much used to study the cell biology of ureteric bud morphogenesis (24, 26, 49, 56, 57). As might be expected from their provenance, mIMCD3 cells also express dact2 mRNA (Fig. 2B).
Knockdown of dact2 mRNA causes enhanced sensitivity to TGF-β and a migration defect in mIMCD3 collecting duct-derived cells.
Given the pattern of Dact2 expression in the branching epithelia of the kidney, and the potential for Dact2 to act as a modulator of morphogenetic signaling (Fig. 1), we sought to test the effect of Dact2 on epithelial cell motility and morphogenesis in a variety of assay systems. To do this, we compared the behavior of cells with normal amounts of dact2 mRNA with that of cells in which dact2 mRNA expression had been reduced by RNAi. Our attempts to achieve deep (>80%) and reliable knockdown of dact2 mRNA in intact kidney rudiments, using techniques for renal RNAi that we have published elsewhere (9), failed to produce adequate knockdown. The use of fluorescent test siRNAs suggested that this failure was because too little siRNA was taken up by the ureteric buds; we have already remarked on the relatively poor uptake of siRNAs into this tissue (28). For this reason, we used the mIMCD3 collecting duct-derived cell line as an alternative means of studying the effects of dact2 knockdown on collecting duct cell behavior and morphogenesis.
Plasmids encoding an antibiotic resistance gene and either a control shRNA, or an shRNA targeting dact2 mRNA, were introduced into mIMCD3 cells, and clones of the transfected cells were grown under G418 selection pressure. Their levels of dact2 expression were compared using quantitative RT-PCR: several clones were obtained in which dact2 expression was knocked down by varying amounts, down to 10% of normal in the most knocked down clones. Control cells showed no detectable Smad2 phosphorylation in the absence of added TGF-β (Fig. 3, A and B), but those in which dact2 had been knocked down showed a modest signal even under these conditions, suggesting a weak constitutive activation of the TGF-β pathway, possibly from “noise” in the receptor system that is normally attenuated by the activity of Dact2 protein. In the presence of TGF-β, dact2 knockdown cells exhibited a significantly enhanced level of Smad2 phosphorylation, and this continued to rise at concentrations where the control cells' activity was flattening toward a plateau. There was, on the other hand, no evidence for enhanced canonical Wnt signaling. Control and knockdown cells both showed mainly membrane and cytoplasmic total β-catenin (Fig. 3, c and d) with no evidence for more nuclear β-catenin in knockdowns. In both control and knockdown cells, staining for active β-catenin was weak and confined only to some cells; neither the intensity nor the frequency of these cells appeared to be higher in knockdowns (Fig. 3, e and f), suggesting that the activity of the canonical Wnt pathway, affected by Dact1 (74), was unaffected by loss of Dact2 in these cells.
One of the simplest assays of the morphogenetic behavior of epithelial cells is the healing of a linear scrape wound in a monolayer (47). Untreated cells, or cells carrying the control plasmid, showed typical epithelial behavior: they advanced steadily across the wound from both sides, moving as an orderly sheet of cells with a smooth leading edge (Fig. 4A). Their advance halved the width of the open wound by 3 h and filled it completely by 6 h (Fig. 4, A and b). Cells expressing only 10% of normal dact2 mRNA also migrated into the wound, but they did so much more slowly; these wounds were still about two-thirds open by 6 h, by which time control cells had sealed their wound, and took 18 h to close completely (Fig. 4, A and B). Furthermore, the leading edge of the moving epithelium looked much less smooth and coordinated than in controls. These effects, on the speed of wound healing and cellular morphology, were observed in experiments with an independent clone of cells, carrying a different dact2 targeting sequence that also expressed dact2 at ∼10% of normal, suggesting that the effect on cell behavior was due to loss of Dact2 and not to an off-target effect, to some effect of plasmid integration site or to a random mutation acquired during cloning.
Dact2 is believed to function mainly by attenuating TGF-β signaling, and we have shown that knockdown of dact2 in our cells results in elevated Smad2 phosphorylation (Fig. 3, A and B). If Dact2 is normally needed to buffer cells against excessive activation of the Smad2 pathway, it would be expected that the behavior of cells expressing too little Dact2 would show too little buffering and be abnormally sensitive to excess TGF-β. We tested this idea by applying TGF-β to the wound-healing system. Control cells showed considerable robustness against interference by 10 μM exogenous TGF-β, and the time course of wound closure was only slightly affected (Fig. 5A). Cells in which Dact2 was knocked down, on the other hand, had lost this robustness, and the same concentration of TGF-β inhibited their already-delayed healing so much that closure was not achieved even by 22 h (Fig. 5A). The elevated sensitivity of dact2 knockdown renal cells to TGF-β, in terms of both pSmad activation (Fig. 3, A and B) and inhibited wound healing (Fig. 5A), strongly suggests that Dact2 is important to normal function of these cells with respect to their response to TGF-β signaling. In contrast, pharmacological inhibition of the Rac GTPase, which is a target of Dact2 in lower vertebrates but apparently not in mammals (Fig. 1), neither rescues nor worsens the effect of dact2 knockdown (Fig. 5B). This further supports the idea that, in mammals, Dact2 is important mainly or exclusively in TGF-β signaling.
Knockdown of dact2 mRNA alters actin organization in migrating mIMCD3 cells.
Examination of control and knockdown mIMCD3 epithelial monolayers moving into a scrape wound showed that cells at the leading edges had different morphologies. Control cells showed no obvious lamellipodia or filopodia and remained in very close association with each other (Fig. 6a). Many knockdown cells at the leading edge showed an almost fibroblastic morphology, with a clear lamellipodium and occasional short filopodia and with much weaker associations with neighboring cells, to the extent that some cells seemed to be migrating independently, completely unattached to the monolayer (Fig. 6b).
Extensive work on wound healing, both in simple culture models like the one presented here and in embryonic epidermis in vivo, has highlighted the importance of actin microfilaments to normal hole closure. Cells at the wound edge organize thick bundles of actin microfilaments that run parallel to the wound edge, in a process organized by the small GTPase Rho, its effector ROCK, and ROCK's effector, myosin light chain kinase (1, 34, 36). Myosin-mediated contraction of these filament bundles closes the hole by purse-string contraction. If production of the filament bundles is inhibited, for example, by blocking Rho activation by genetic techniques, or by depolymerizing actin using cytochalasins, closure of the wound takes place abnormally slowly (1) and eventually occurs by lamellipodial crawling (69). The effect of dact2 knockdown on wound-healing behavior in our cells is similar to the effects of inhibiting actin filament formation in these published reports; we therefore examined actin organization in our control and knockdown cultures.
In control cultures, cells abutting the wound showed prominent actin microfilaments arranged parallel to the wound edge and appearing to form, at the resolution of light microscopy, a continuous cable (Fig. 6c). In dact2 knockdown cultures, most leading edge cells failed to make this cable, although a few did apparently produce a faint one (Fig. 6d). It is notable that there was a clear negative correlation between cells having a motile morphology (lamellipodium, etc.) and having this actin cable. To quantify the difference in actin cable formation, we measured the lengths of apparently continuous actin cables at the wound edges of control cells, knockdown cells, and cells never transfected with even control plasmids. As well as there being significantly fewer cables in knockdown cells, they were only about half as long (Fig. 6e). The cells with control plasmid had cables of the same length as completely untransfected cells, as expected. Cell proliferation, another mechanism for closing a gap, was unaffected as measured by 5-bromo-2-deoxyuridine incorporation following wounding; controls showed 23% (σ = 4%) of cells incorporating 5-bromo-2-deoxyuridine, knockdowns 25% (σ = 8%; insignificant at P = 0.15 by ANOVA).
Knockdown of dact2 mRNA alters surface morphology of cysts and tubules made by mIMCD3 cells in three-dimensional culture.
Collecting duct-derived cell lines such, as Madin-Darby canine kidney and mIMCD3 cells, make cysts when cultured in three-dimensional collagen gels (27, 66). When appropriate ramogenic growth factors are added to the culture medium, these cysts go on to form extended, branching tubules (41, 42). These have become a standard system for the study of the mechanisms of ureteric bud branching (56, 57).
When we set up control mIMCD3 cells in 3-dimensional culture, they formed rounded cysts with smooth edges (Fig. 7a). Knockdown cells were still able to form cysts, but these had much rougher, less-defined edges (Fig. 7b). Both control and dact2 knockdown cells were able to form branching tubules (Fig. 7, c and d). Those made by the knockdown cells were again less smooth edged, having significantly more spikes protruding from them (Fig. 7, d and e).
Dact2 knockdown cells show evidence of epithelial-mesenchymal transition.
The type of motility shown by dact2 knockdown cells in the wound-healing and the three-dimensional tubulogenesis assays suggests behavior more typical of mesenchymal cells than of the epithelial starting population. We therefore compared the expression, in control and dact2 knockdown cells, of E-cadherin, a marker characteristic of epithelial cells, and vimentin, a marker characteristic of mesenchymal cells. Cells harboring control plasmid expressed E-cadherin (Fig. 8a) but fewer than 1% expressed vimentin (Fig. 8, c and e). Most cells harboring the Dact2 knockdown plasmid lost their expression of E-cadherin (Fig. 8b) and gained expression of vimentin (Fig. 8, d and e). They therefore demonstrated the classic signs of epithelial-mesenchymal transition.
In all of these assays, then, wound healing, cyst formation, and tubulogenesis, mIMCD3 cells in which dact2 had been knocked down showed a phenotype with much more invasive, motile character than did controls, and also expressed markers characteristic of mesenchyme. This suggests that dact2 normally plays an important role in restraining mesenchymal character and in keeping cells epithelial so that they move as a group rather than by cells' individual locomotive efforts.
In results, we reported that dact2 is expressed in the ureteric bud/collecting duct system of developing kidneys and that, when its expression is knocked down in a collecting duct cell line, the cells become more sensitive to TGF-β, both biochemically and behaviorally. Furthermore, the morphogenetic behavior of those cells changes: rather than moving as an organized group of cells and producing smooth-edged branching tubules, they move more individually with spiky leading edges, and, when arranged as tubules, they make numerous, fine spikes.
The main function of Dact2 in intracellular signaling is to promote degradation of Alk5 (TGFβRI); in mammals, it has no detectable effect on canonical Wnt signaling and affects noncanonical Wnt signaling only when massively overexpressed (63). Dact2's normal function in the ureteric bud may therefore be reasonably assumed to be to limit the activation of Smad2/3, and thence Smad4, by members of the TGF-β family (see Fig. 1). The developing kidney expresses a number of TGF-β molecules, including activin, several BMPs, and TGF-β1 and -2 (Table 1). When applied to cultured kidney rudiments, most of them are strong inhibitors of ureteric branching (4, 20, 35, 54). The partial exception is BMP7, which inhibits at high doses but may have a stimulatory influence at subnanomolar doses (49). The molecules also inhibit branching in three-dimensional mIMCD3 cultures (21, 49). Dact2's main function in the ureteric bud may therefore be to attenuate the branch-inhibiting signals of the TGF-β family so that branching is still possible but its direction and probability can perhaps still be steered by gradients in TGF-β, as seems to happen in the mammary gland (29, 45).
In a variety of assay systems, TGF-β causes epithelial cells to express a more mesenchyme-like phenotype. Examples of cells that apparently undergo a complete epithelial-mesenchymal transformation in subconfluent two-dimensional cell culture include NP-1 renal tubular cells, mammary epithelial cells, hepatocytes, thyroid epithelial cells, and pulmonary alveolar cells (17, 19, 25, 48, 62, 72). The appearance of the TGF-β-treated cells in these systems is very similar to those of our dact2 knockdown cells at the wound margin, except perhaps that TGF-β tends to produce strong stress fibers that were not an obvious feature of our system. In particular, in mammary cells it has recently been reported that TGF-β causes cells to migrate as individuals rather than en masse (17). As might be expected from the in vitro experiments mentioned above, TGF-β appears to be an important natural driver of epithelial-mesenchymal transitions in normal development of several organs including the heart and the palate (44, 50). It also drives epithelial-mesenchymal transitions in response to injury, for example, in the lens of the eye and in the kidney (55, 73). The loss of E-cadherin and gain of vimentin that we see as a response to Dact2 knockdown in our cells suggests that TGF-β signaling, when freed from the restraint normally provided by Dact2, is powerful enough to induce epithelial-mesenchymal transition in our cells, too, at least in these simple culture systems. In whole animals, dact2 knockout has only a modest phenotype, connected with enhanced TGF-β-driven keratinocyte motility and thence improved wound healing in the context of full-thickness adult skin (37), which heals by scarring, a different situation from the scar-free, single layer of epithelium in our culture model, probably because of redundancy between Dact2 and other modulators of TGF-β signaling such as FKBP12, Smad7, Smurf1, and STRAP. Culture systems are often more sensitive to perturbation, and therefore more powerful as an assay for function, than is an intact animal; in the urinary collecting duct itself, for example, HGF (71) and neurturin (10) are revealed as positive regulators of branching although their roles are masked by redundancies in an intact knockout embryo (60, 61).
The presence of “motility” structures (lamellipodium, etc.) on mesenchymal cells may create the impression that these are better suited to movement than epithelial cells, and therefore make the observation that TGF-βs inhibit ureteric bud branching seem paradoxical. Such an impression would be false, however, for the ability of epithelial cells to move collectively as a coordinated sheet or tube is superior to that of isolated cells that result from a mesenchyme-to-epithelium transition. This is illustrated both in our wound-healing assay and those of other researchers (1). The presence of actin cables, organized by Rho, ROCK and myosin light chain kinase, is critical to coordinated closure of wounds in many systems (1, 34, 36), although it can be less important than Rac-driven crawling in some cells lines such as Madin-Darby canine kidney cells (14). The actin cables are clearly disrupted or missing in our Dact2 knockdown cells. We have shown elsewhere that organized actin cables are a feature of ureteric bud branching and that interference with their production, using cytochalasins or drugs that inhibit the action of ROCK or of myosin light chain kinase, inhibits branching (38).
The discovery of a growing number in intracellular molecules that limit pathway activation, such as Dact2, Sprouty, SPRED, etc., has emphasized that understanding negative regulation is essential to understanding the coupling between diffusible ligand and cellular response (3, 11). Negative signaling has been proposed to be important in limiting both the amplitude and duration of pathway activation (11). Here, we observed that loss of the negative TGF-β regulator Dact2 resulted not only in a change to a more mesenchymal phenotype but also much less stability of behavior in the face of excess TGF-β (control cells were little affected by it in a wound-healing assay, whereas knockdown cells were highly susceptible). This suggests that one role of intracellular pathway inhibitors such as Dact2 may be to buffer cells against fluctuating levels of extracellular ligand and to keep their behavior stable across a range of ligand concentrations. Such a function may be important in development, for example, in gradients, and in adult life in maintaining tissue homeostasis in the face of environmental noise. Expression of different internal inhibitors, or different concentrations of them, may also be one mechanism by which different cell types show different responses to the same extracellular ligand.
This work was funded by EU Framework 6 (Kidstem) and the NC3Rs.
No conflicts of interest, financial or otherwise, are declared by the authors.
We thank Trudi Gillespie and Paul Perry for help with microscopy and Mathieu Unbekandt and Veronika Ganeva for useful discussions.
- Copyright © 2010 the American Physiological Society