Identification of a mechanochemical checkpoint and negative feedback loop regulating branching morphogenesis

Abstract

Cleft formation is the initial step in submandibular salivary gland (SMG) branching morphogenesis, and may result from localized actomyosin-mediated cellular contraction. Since ROCK regulates cytoskeletal contraction, we investigated the effects of ROCK inhibition on mouse SMG ex vivo organ cultures. Pharmacological inhibitors of ROCK, isoform-specific ROCK I but not ROCK II siRNAs, as well as inhibitors of myosin II activity stalled clefts at initiation. This finding implies the existence of a mechanochemical checkpoint regulating the transition of initiated clefts into progression-competent clefts. Downstream of the checkpoint, clefts are rendered competent through localized assembly of fibronectin promoted by ROCK I/myosin II. Cleft progression is primarily mediated by ROCK I/myosin II-stimulated cell proliferation with a contribution from cellular contraction. Furthermore, we demonstrate that FN assembly itself promotes epithelial proliferation and cleft progression in a ROCK-dependent manner. ROCK also stimulates a proliferation-independent negative feedback loop to prevent further cleft initiations. These results reveal that cleft initiation and progression are two physically and biochemically distinct processes.

Introduction

Mechanical forces play a pervasive role in regulating morphogenetic changes during embryonic organ development. Cells sense and respond to changes in their microenvironment by contracting and pulling on the extracellular matrix (ECM) and on neighboring cells. Inside the cell, contractile activity is generated by the interaction of activated non-muscle myosin II with actin. This cellular contractility has been implicated in cellular rearrangements and migration (Keller et al., 2003; Kiehart et al., 2000; Xu et al., 2008), proliferation (Huang et al., 1998), and differentiation (McBeath et al., 2004). However, the molecular mechanisms by which cells coordinate forces with cellular responses resulting in morphogenesis remain unclear.

Branching morphogenesis is a developmental mechanism utilized by many organs, including the salivary gland, lung, kidney, and mammary gland, to increase the epithelial surface area for secretion or absorption (Hogan, 1999). The embryonic submandibular salivary gland (SMG) is a classic model for studying tissue morphogenesis in three dimensions ex vivo (Grobstein, 1953). On embryonic day 12 (E12), the first round of branching begins as multiple clefts, or indentations, are formed in the basement membrane covering the surface of the epithelium. These clefts then extend, or progress, towards the interior of the primary bud. These processes are repeated multiple times to generate the complex, arborized adult structure. While many growth regulatory pathways are implicated in the control of branching (Patel et al., 2006; Tucker, 2007), the nature of the physical changes that occur during these morphogenetic events is not understood.

Several hypotheses have been proposed to explain cleft formation. Localized cell proliferation has long been known to play a critical role in branching morphogenesis (Bernfield et al., 1972; Nogawa et al., 1998), although a direct requirement for proliferation in SMG cleft formation was ruled out in studies using proliferation inhibitors since clefts were still observed to initiate under such conditions (Nakanishi et al., 1987). However, these studies failed to distinguish between cleft initiation and progression, and did not investigate whether once initiated, if such clefts continued to progress in the absence of proliferation. Basement membrane dynamics have also been identified as a possible driving force for SMG cleft formation. Classic studies demonstrated that collagen III preferentially accumulates at sites of cleft initiation (Fukuda et al., 1988; Nakanishi et al., 1988), and more recently, the ECM protein fibronectin (FN) has been identified as a putative cleft initiator molecule (Sakai et al., 2003). siRNA knockdown of FN prevents cleft formation in E12 SMGs, while the addition of exogenous FN promotes epithelial cleft formation and leads to decreased E-cadherin-mediated cell-cell adhesions. However, the relationship between FN and proliferation was not examined in this study, nor was an upstream signal identified. Finally, a requirement for the actin cytoskeleton was identified in early studies. SMGs cultured in the presence of cytochalasin B, an inhibitor of actin polymerization, fail to form clefts (Spooner and Wessells, 1972), in support of a hypothesis that actomyosin contractility is required for branching morphogenesis. Mathematical modeling has also implicated contractile events occurring within the mesenchyme as a key contributor to this process, and suggests that the resulting traction forces produced by such mesenchymal contraction may help drive cleft formation by deforming the epithelial surface (Nakanishi et al., 1986; Wan et al., 2008).

Actin rearrangements and cellular contractility can be controlled by Rho-associated coiled-coil containing kinase (ROCK). ROCK activates myosin light chain 2 (MLC2) through phosphorylation of Ser19 (Amano et al., 1996; Totsukawa et al., 2000), which is required for its interaction with actin (Landsverk and Epstein, 2005). Furthermore, signaling through ROCK and the resulting cytoskeletal contraction regulates FN assembly in fibroblasts, primarily through the unfolding of cell-surface bound FN to expose a cryptic self-assembly site (Baneyx et al., 2002; Yoneda et al., 2007; Zhong et al., 1998).

ROCK activity is implicated in the development of other branching organs. In the embryonic lung, inhibition of ROCK reduces branching morphogenesis, and results in the loss of localized differentials in basement membrane thickness and cell proliferation (Moore et al., 2005). A role for FN in this process was not investigated, despite the fact that FN is also required for lung branching morphogenesis (De Langhe et al., 2005; Sakai et al., 2003). Furthermore, a direct link between basement membrane remodeling and localized proliferation was not established. In the developing kidney, ROCK inhibition was reported to promote ureteric bud branching, despite downstream alterations to the actin cytoskeletal structure (Meyer et al., 2006; Michael et al., 2005). Thus, the molecular functions of ROCK in branching morphogenesis have not been clearly defined, and differences between organs are apparent.

In the current study using the developing embryonic mouse submandibular salivary gland as a model branching organ, we report the existence of a mechanochemical checkpoint regulating the transition of clefts to a stabilized state competent to undergo progression. We report that inhibition of ROCK I or non-muscle myosin II activity prevents passage through this checkpoint, and stalls clefts at the initiation stage. In the presence of ROCK/myosin II inhibition, failure of clefts to progress is due to disrupted FN assembly in the basement membrane and a lack of proliferation by basal epithelial cells. Furthermore, we demonstrate that FN assembly itself promotes epithelial proliferation and cleft progression in a ROCK-dependent manner. Interestingly, we also detect an increased number of initiated clefts that fail to progress in the presence of ROCK I/myosin II inhibitors, indicating the existence of a negative feedback loop that temporarily prevents additional cleft initiations locally in a proliferation-independent manner. These studies provide mechanistic evidence that cleft initiation and progression are two distinct events during SMG branching morphogenesis.

Materials and Methods

Submandibular salivary gland ex vivo organ culture and inhibitor assays

Mouse SMGs were dissected from timed-pregnant female mice (strain CD-1, Charles River Laboratories) at either embryonic day 12 (E12, 1 bud) or E13 (4-5 buds), with the day of plug discovery designated as E0, following protocols approved by the University at Albany IACUC committee. SMGs were microdissected from mandible slices and cultured as previously described (Rebustini and Hoffman, 2009; Sakai and Onodera, 2008). Briefly, mandible slices were removed from embryos with a sterile scalpel and, from these, SMGs were microdissected using sterile forceps under a dissecting microscope (SMZ645, Nikon). The SMGs were cultured on 13 mm, 0.1 μm Nuclepore Track-Etch membrane filters (Whatman) floating on 200 μL of 1:1 DMEM/Ham's F12 Medium (F12) lacking phenol red (Invitrogen) in glass-bottom 50 mm microwell dishes (MatTek Corporation). The medium was supplemented with 50 μg/mL transferrin, 150 g/mL L-ascorbic acid, 100 U/mL penicillin, and 100 μg/mL streptomycin, to make complete DMEM/F12 medium.

For experiments using SMG epithelial rudiments, dissected SMGs were incubated in 1.6 U/mL dispase (Roche) prepared in HBSS (Invitrogen) for 25 minutes at 37°C. The dispase was neutralized with 5% BSA in DMEM/F12, and the mesenchyme physically separated from the epithelium with fine forceps. SMG epithelial rudiments were cultured in a final concentration of 6 mg/mL Matrigel (BD Biosciences) diluted in DMEM/F12 containing 20 ng/mL EGF and 200 ng/mL FGF7 (R&D Systems). Five intact SMGs or three epithelial rudiments were used for each experimental condition and all experiments were repeated at least three times.

Pharmacological inhibitors (Calbiochem) were resuspended in DMEM/F12 or DMSO vehicle, as recommended by the manufacturer: Y27632 (688000), H-1152 (555550), (-)-blebbistatin (203391), (+)-blebbistatin (negative control for (-)-Blebbistatin, 203392), and hydroxyurea (400046). Inhibitor-containing media was replaced every 24 hours. Brightfield images were acquired on a Nikon Eclipse TS100 microscope equipped with a Canon EOS 450D digital camera at either 4× (Plan 4×/0.10 NA), 10× (10× Ph1 ADL/0.25 NA), or 20× (LWD 20× Ph1 ADL/0.40 NA) magnification.

Morphometric analysis was performed using MetaMorph™ (Version 6.1, MDS Analytical Technologies), as described previously (Larsen et al., 2003). Fold change was calculated by normalizing to the number of buds or clefts present at the initial time point (2 hours) and displayed using Prism (GraphPad, Version 5.01). Thickness measurements were generated from glands treated for 24 hours, fixed, and stained with rhodamine phalloidin (Invitrogen, 1:350). Thickness was calculated from Z-stacks acquired using a confocal microscope (LSM 510 Meta with Argon, HeNe1, and HeNe2 lasers, Carl Zeiss). For SMG diameter measurements, brightfield tiff files were imported into MetaMorph, and the diameter was measured using the calibrated line tool.

siRNA knockdown of ROCK I, ROCK II, and FN

The following pre-validated duplexed siRNAs (Validated Silencer Select) were obtained from Applied Biosystems: ROCK I, s73018 and s73019; ROCK II, s73021 and s73022; FN1, s66182 and s66183; Negative Control siRNA #1; Negative Control siRNA #2; and Cy3-conjugated Negative Control siRNA #1. SMGs were transfected with 100-500 nM siRNA using RNAiFect (Qiagen) according to the manufacturer's protocol, and cultured for 24-48 hours. Each experiment was repeated 2-4 times for each siRNA with 5 SMGs per group.

SCA-9 cell culture and live FN assays

The mouse SMG-derived epithelial cell line, SCA-9, developed from an induced tumor derived from adult mouse submandibular gland, was maintained as previously described (Barka et al., 2005; Barka et al., 1980). SCA-9 cells were cultured overnight to 70% confluence, washed with 1X PBS, and grown for 18 hours in serum-free DMEM containing vehicle control or inhibitors. To detect fibrillar FN, purified human plasma FN (Akiyama, 1999) that was labeled with Alexa⁶⁴⁷ (Larsen et al., 2006) was added to SCA-9 cell cultures at 0.02 μg/μL in serum-free media. Cells were imaged by time-lapse microscopy for 18 hours or immunostained with L8 monoclonal (Chernousov et al., 1987) to label stretched FN and an anti-FN rabbit polyclonal antibody to label total FN. Imunocytochemistry was performed as described previously (Cukierman et al., 2001), and cells were imaged on an inverted fluorescence microscope (Cell Observer Z1, Carl Zeiss) with a CCD camera (Axiocam MRm, Carl Zeiss) at 63× magnification (Plan-Apochromat 63×/1.4 NA DIC, Carl Zeiss). Sodium deoxycholate (DOC) extractions for fibrillar fibronectin were performed as reported previously (McKeown-Longo and Mosher, 1983), except that cells were collected into 2% DOC extraction buffer with a mini protease inhibitor tablet (Roche) and analyzed by Western analysis, as described.

Time- lapse microscopy

Time-lapse microscopy was performed essentially as previously described (Larsen et al., 2006; Larsen et al., 2003), except that a Cell Observer Z1 (Carl Zeiss) fitted with an environmental chamber (XL-S1, Carl Zeiss) and calibrated for 5% CO₂ at the center of the stage was used to capture images. Phase contrast images were captured every 10 minutes for 20 hours using AxioVision software (Release 4.7.1, Carl Zeiss) at 5× (Plan Neo, NA 0.16), 10× (Plan Neo, 0.3 NA) or 20× (Plan Neo, NA 0.5) magnification. Images were automatically registered in AxioVision and then adjusted for brightness and contrast and converted into movies (AVI format/Cinepak codec) in MetaMorph. For cleft measurements, SMGs were imaged for 12 hours and images imported into MetaVue (Version 6.2r6, Molecular Devices). Stacks were adjusted for brightness and contrast, and equally cropped to assemble movies and montages. Morphometric measurements were acquired with the line tool from two clefts each from three SMGs using the top edge of the cropped box as a reference. See Fig S4A for a detailed description of the measurements and calculations of elongation index (EI), contraction index (CI), and progression index (PI).

Whole-mount immunocytochemistry

Immunostaining was performed essentially as described (Larsen et al., 2003) with all antibody incubations performed overnight at 4°C. SMGs were washed 4 × 10 minutes in 1X PBS-T after each antibody incubation step, and mounted on glass coverslips with Secure-seal imaging spacers (Grace Bio-Labs) in 40 uL BioMeda™ Gel-Mount (Electron Microscopy Sciences) containing 1 mg/ml P-phenylenediamine (PPD) antifade reagent. SMGs were imaged on a Zeiss LSM510 confocal microscope at 10× (Fluar 10×/0.50 NA), 20× (Plan Apo/0.75 NA), or 63× (Plan Apo/1.4 NA) magnification.

Antibodies used and their dilutions from stock are as follows: E-cadherin (Clone 36, BD Biosciences, 1:250), ROCK I (Clone EP786Y, Abcam, 1:250), ROCK II (H-85, Santa Cruz Biotechnology, 1:50), pMLC2 (Ser 19) (#3671, Cell Signaling Technology, 1:50), Myosin IIb (#3404, Cell Signaling Technology, 1:100), Na⁺/ K⁺ ATPase (EP1845Y, Epitomics, 1:100), BrdU (Roche, 1:10), and the heparan sulfate proteoglycan, perlecan (clone A7L6, Chemicon, 1:100). Anti-fibronectin rabbit polyclonal antibody R5836 (1:25) and L8 monoclonal antibody (25 μg/mL) were generously provided by Drs. Kenneth Yamada and Michael Chernousov, respectively. SYBR Green I (Invitrogen, 1:100,000), and Alexa Fluor⁵⁴⁶ Phalloidin (Invitrogen, 1:350) were also used. Cyanine dye-conjugated AffiniPure F(ab’)₂ fragments were used as secondary antibodies (Jackson ImmunoResearch Laboratories, 1:50).

Western analysis

Protein assays and Western blot analysis were performed essentially as previously described (Larsen et al., 2003) except that 5-10 μg of protein was loaded per well on a 1 mm-NuPAGE^R 4-12% Bis-Tris Gel, electrophoresed, and transferred to a PVDF membrane using the Invitrogen XCell Surelock™ system (Invitrogen). Chemiluminescent blots were imaged using a BioRad ChemiDOC XRS™ (BioRad) or X-ray film (ECL Hyperfilm) and bands quantified using Quantity One (Version 4.6.1, BioRad) with normalization to GAPDH or MLC2 (for pMLC2). Each experiment was repeated three times, but a single blot and quantification of that blot are shown in each figure.

Antibodies used and their dilutions are as follows: ROCK I (EP786Y, Abcam, 1:500), ROCK II (H-85, Santa Cruz, 1:200), E-cadherin (Clone 36, BD Biosciences, 1:5,000), Vimentin (Clone 13.2, Sigma, 1:10,000), β-Actin (Clone AC-74, Sigma, 1:10,000), Phospho-Myosin Light Chain 2 (Ser19) (Cell Signaling, 1:1000), Myosin Light Chain 2 (Cell Signaling, 1:1,000), anti-FN rabbit polyclonal antibody (Ken Yamada, 1:1,000), and GAPDH (Fitzgerald, 1:10,000).

Proliferation and apoptosis assays

For cell proliferation studies, intact E13 SMGs were cultured for 24 hours in the presence of inhibitor, vehicle control, FN siRNA, or exogenous human plasma FN and pulsed for 2 hours with BrdU using a BrdU Labeling and Detection Kit (Roche), as described previously (Steinberg et al., 2005), with the following modifications. Following the anti-BrdU antibody (1:10) incubation step, SMGs were blocked and immunostained as described above using an antibody recognizing Na⁺/K⁺ ATPase (1:100) to detect primarily the epithelium, anti-FN antibody (1:25) to assess FN knockdown, and SYBR green to counterstain total nuclei. The BrdU/SYBR green ratio was calculated from confocal stacks spanning the thickness of the entire SMG and graphed in Prism.

Apoptosis was compared by TUNEL staining with an In Situ Cell Death Detection Kit (Roche) according to the manufacturer's instructions but with an additional immunostaining step added to detect E-cadherin and SYBR Green to label total nuclei. TUNEL-positive nuclei were quantified in MetaMorph as the ratio of TUNEL/SYBR green from confocal stacks and graphed in Prism.

Statistical analysis

For all experiments, one-way analysis of variance (ANOVA) was performed with a Bonferroni post-test, except when otherwise indicated, using Prism. All graphs show one representative experiment of at least three.

Results

Pharmacological inhibitors of ROCK stall SMG branching morphogenesis at the cleft initiation stage and allow additional cleft initiations

To determine if ROCK-mediated signaling is a critical component of cleft formation during SMG branching morphogenesis, we treated embryonic day 13 (E13) SMGs with increasing concentrations of Y27632 and H-1152, two structurally distinct pharmacological inhibitors that block the catalytic site of both ROCK isoforms, ROCK I and II (Ikenoya et al., 2002; Uehata et al., 1997). Control SMGs cultured for 24 hours underwent several rounds of branching, while SMGs treated with Y27632 or H-1152 exhibited a morphology typified by larger, less-branched buds, relative to vehicle control. Morphometric analysis revealed a dose-dependent decrease in the number of completed epithelial buds (Fig 1A).

Figure 1.

Pharmacological inhibition of ROCK blocks SMG branching morphogenesis in organ culture. (A) Images of live E13 SMGs treated with increasing concentrations of Y27632 and H-1152 for 24 hours exhibit decreased branching relative to vehicle control (see also Movie S1). Quantification of the number of buds after 24 hours revealed a dose-dependent inhibition of branching in the presence of both ROCK inhibitors. (B) E13 SMGs treated for 24 hours with 140 μM Y27632 or 2 μM H-1152 exhibit an increased number of initiated, but not elongated, clefts on the surface of epithelial buds relative to vehicle control (compare white arrows). Quantification revealed a dose-dependent increase in such immature clefts with both Y27632 and H-1152. (C) Branching morphogenesis of E12 SMGs treated with either 140 μM Y27632 or 2 μM H-1152 for 24 hours is also inhibited at the cleft initiation stage. ROCK inhibition results in an increased number of initiated, but not elongated clefts (compare white arrows). (D) E13 SMG epithelial rudiments cultured for 24 hours with 140 μM Y27632 or 2 μM H-1152 also show multiple initiated immature clefts (see also Movie S1). Quantification reveals both a dose-dependent decrease in completed buds and an increase in small initiated clefts. ANOVA *p<0.05, **p<0.01, N=5. Scale bars, 250 μm (A, C, and D) and 50 μm (B and C).

Since the larger-sized buds in ROCK-inhibited SMGs could be explained by decreased clefting, we investigated how ROCK signaling affects SMG cleft formation by time-lapse imaging of intact E13 SMGs (Movie S1). When SMGs were cultured with either ROCK inhibitor, multiple clefts initiated at the surface of the large buds, but did not continue to lengthen, or progress, as they did in vehicle control-treated glands (white arrows, Fig 1B). In fact, morphometric analysis revealed a dose-dependent increase in the number of initiated, but not elongated, clefts in the presence of both ROCK inhibitors (Fig 1B).

The increased number of initiated clefts observed upon ROCK inhibition suggests that ROCK activity is not required for cleft initiation itself. However, because cleft formation begins at E12, and we added the inhibitors on E13, it is possible that clefts had all ready initiated on E12 when ROCK activity was not perturbed. To directly rule out this possibility, we cultured E12 SMGs in the presence of increasing concentrations of both ROCK inhibitors and performed morphometric analysis on the cleft regions. E12 SMGs treated with both ROCK inhibitors likewise showed a significant inhibition of branching morphogenesis (Fig 1C), thus indicating that ROCK signaling is also involved in the initial round of branching. Furthermore, as in E13 SMGs, there was an increased number of small initiated, but not elongated, clefts in the surface of the epithelium in the presence of the ROCK inhibitors (white arrows, Fig 1C). Taken together, these results conclusively demonstrate that ROCK signaling is not required for cleft initiation, since clefts still initiate under conditions of ROCK inhibition at E12, prior to the first round of cleft formation.

In order to ensure that Y27632 and H-1152 were not toxic, we performed wash-out experiments in which both E12 and E13 SMGs showed significant recovery 48 hours after the wash-out (Fig S1A and B). To confirm that the inhibitors did not induce apoptosis, we performed TUNEL assays after 24 hours of culture, and observed no significant differences in the number of apoptotic nuclei for vehicle control-treated and ROCK-inhibited SMGs (Fig S1C). Taken together, these data indicate that ROCK inhibition with two structurally distinct pharmacological inhibitors blocks SMG branching morphogenesis at the cleft initiation stage both during the initial and later stages of branching in a manner that is non-toxic to developing SMGs.

To determine if ROCK signaling is active within the epithelium itself, we cultured E13 epithelial rudiments in the absence of mesenchyme (Morita and Nogawa, 1999). As with the intact SMGs, ROCK inhibition with Y27632 and H-1152 resulted in a decrease in the number of fully-completed buds and an increase in the number of initiated, but not elongated, clefts relative to vehicle control (Fig 1D, data not shown). Time-lapse microscopy further confirmed these observations (Movie S1). These results demonstrate that ROCK signaling within SMG epithelial cells is an important determinant of branching morphogenesis and that inhibition of such signaling blocks cleft progression, but facilitates the initiation of additional clefts at the surface of epithelial buds even in the absence of mesenchyme. These results do not preclude an additional role for ROCK proteins possibly expressed in mesenchymal cells.

ROCK I, expressed in the epithelium of developing SMGs, is the critical ROCK isoform promoting SMG cleft progression

There are two ROCK isoforms, ROCK I (p160ROCK or ROKβ) and ROCK II (ROKα) (Nakagawa et al., 1996), both of which are sensitive to the ROCK inhibitors. To examine cell-type distribution of ROCK I and II in mouse SMGs, we performed Western analysis on cell lysates derived from the epithelium or mesenchyme (Fig 2A). ROCK I, ROCK II, and their downstream effector, MLC2, were expressed approximately two-fold higher in epithelial extracts than in mesenchymal extracts. To confirm the preferential expression of ROCK I and II in the epithelium, we performed immunocytochemistry on intact E13 SMGs (Fig 2B). In agreement with the Western analysis of cell extracts, higher levels of both ROCK isoforms were detected in the epithelium than in the mesenchyme. Interestingly, ROCK I showed locally elevated regions of expression, whereas ROCK II was more diffusely distributed, which is evident at high magnification (white square, Fig 2B).

Figure 2.

ROCK I is expressed in the epithelium of E13 SMGs and regulates cleft progression. (A) Western analysis of epithelial and mesenchymal cell extracts from E13 SMGs reveals that ROCK I, ROCK II, and MLC2 are expressed in both the epithelium and mesenchyme, shown with quantification of relative band intensity normalized to GAPDH. (B) Immunocytochemistry of ROCK I (green) and ROCK II (cyan) demonstrates that both ROCK isoforms are more highly expressed in the epithelium, demarked with E-cadherin (red), than in the mesenchyme. Confocal images were acquired from the center of each SMG. (C) Transfection of E13 SMGs for 48 hours in organ culture with 400 nM ROCK I siRNA, but not 400 nM ROCK II or negative control siRNA, reduces SMG branching with similar morphological perturbations as the pharmacological ROCK inhibitors, High power images show incomplete clefts in ROCK I siRNA-treated SMGs but not control or ROCK II siRNA-treated SMGs (white arrows). Quantification of siRNA-treated SMGs reveals that knockdown of ROCK I, but not ROCK II, blocks cleft progression and promotes additional initiation events. SMGs treated with ROCK I and II siRNAs in combination exhibit similar fold changes in bud and initiated clefts as SMGs treated with ROCK I siRNA alone. (D) Western blot analysis to confirm knockdown of ROCK I and ROCK II with isoform-specific ROCK siRNAs. Quantification was normalized to GAPDH and expressed as a percentage relative to negative control-treated SMGs. U, untreated; R, transfection reagent only; C, 400 nM negative control siRNA. ANOVA *p<0.05, **p<0.01. Scale bars, 50 and 20 μm (B) and 250 and 50 μm (C).

ROCK I and ROCK II share 65% amino acid identity and some functional overlap (Nakagawa et al., 1996). To determine if there is a critical ROCK isoform(s) with a role in SMG branching, we treated developing salivary glands with isoform-specific ROCK siRNAs. Preferential uptake of siRNA in epithelial cells was observed, as previously reported (Sakai et al., 2003) (Fig S2). Knockdown of ROCK I produced SMG morphologies similar to those observed for the ROCK inhibitors Y27632 and H-1152 (Fig 2C, compare to Fig 1A), whereas there was little effect of ROCK II siRNA after 48 hr. Morphometric analysis indicated a significant inhibition of branching with a dose-dependent decrease in completed buds and an increase in initiated, but not elongated, clefts with ROCK I siRNA knockdown but not with ROCK II or control siRNA (white arrows, Fig 2C). Furthermore, SMGs treated with both ROCK I and II siRNAs in combination exhibited a comparable fold change decrease in completed epithelial buds, and a fold change increase in small initiated clefts as was observed for SMGs cultured with ROCK I siRNA alone (Fig 2C). Western blot analysis was performed to assess the efficiency of ROCK I and II knockdown at the protein level (Fig 2D). Both ROCK I and II siRNAs resulted in approximately 50% isoform-specific knockdown when used at a 400 nM concentration with non-targeting negative control siRNAs having a minimal effect. Together, these results indicate that ROCK I is the primary isoform involved in SMG branching morphogenesis and further validates the use of pharmacological inhibitors as a means of investigating the role of ROCK I during SMG branching.

ROCK I and myosin activity regulate cytoskeletal tension around the periphery of epithelial buds

To determine if inhibition of myosin activity would likewise affect cleft formation, we used the pharmacological inhibitor blebbistatin, which prevents myosin function by locking myosin II in its actin-detached state to prevent contractility (Kovacs et al., 2004). Blebbistatin-treated E13 SMGs showed a similar morphology to that of SMGs treated with Y27632, H-1152, or ROCK I siRNA (Fig 3A, compare to Fig 1A and 2C). Similarly, morphometric analysis revealed a dose-dependent decrease in the number of newly formed buds, and a corresponding increase in the number of initiated, but not elongated, clefts (white arrows, Fig 3A), indicating that myosin II, downstream of ROCK I, is a critical regulator of cleft progression.

Figure 3.

Myosin activity is required for SMG cleft progression and promotes cytoskeletal tension around the periphery of epithelial buds. (A) E13 SMGs cultured for 24 hours with 100 μM blebbistatin show an increased number of initiated, but not elongated, clefts (white arrows). Quantification of morphological changes with increasing concentrations of blebbistatin. (B) Western analysis reveals a dose-dependent decrease in the amount of phosphorylated pMLC2 in the presence of increasing concentrations of Y27632, H-1152, and ROCK I siRNA. (C) Confocal images through the center of immunostained SMGs reveals that pMLC2 (green) is present around the periphery of epithelial buds (white arrows), while myosin IIb heavy chain (red) is present throughout the epithelium in the Z dimension, shown as XY view (left) and XZ projection of a stack through the SMG (right). pMLC2 staining is decreased in epithelial rudiments treated with 140 μM Y27632 for 24 hours relative to vehicle control. (D) Confocal images show ROCK I (cyan) is present at the epithelial surface with pMLC2 (green) and actin (red). (E) E13 SMGs cultured for 24 hours with 140 μM Y27632, 2 μM H-1152, or 100 μM blebbistatin exhibit decreased thickness and expanded diameters relative to vehicle control. ANOVA *p<0.05, **p<0.01. Scale bars, 250 μm (A), 20 μm (C), and 10 μm (D).

Since phosphorylation of MLC2 on Ser19 is required for its activation, and this site is a target of ROCK (Amano et al., 1996; Totsukawa et al., 2000), we investigated whether the treatment of SMGs with Y27632, H-1152, and ROCK I siRNA caused a decrease in phosphorylated MLC2 (pMLC2). Both ROCK inhibitors as well as ROCK I siRNA demonstrated a dose-dependent decrease in the amount of pMLC2 (Ser19) relative to MLC2 (Fig 3B). Immunocytochemistry revealed that immunostaining for pMLC2 decreased when SMG epithelial rudiments were treated with Y27632 or H-1152 (Fig 3C and data not shown), while myosin heavy chain IIb did not. Significantly, confocal stacks of SMGs in the Z dimension revealed that pMLC2 was primarily localized to the periphery of epithelial buds, while myosin IIb was present throughout. This peripheral localization was substantially decreased when SMGs were cultured with Y27632 or H-1152 (Fig 3C and data not shown). Furthermore, subpopulations of pMLC2 localized to both regions of the cytosol and to patches along the plasma membrane, in a pattern similar to that of both ROCK I and actin (Fig 3D). Together, these data provide molecular evidence that ROCK I functions upstream of myosin during SMG branching morphogenesis to regulate actomyosin-based contractility at the periphery of epithelial buds. Taken together with the finding that the number of cleft initiation events at the epithelial surface increases when contractility is inhibited (Fig 1B, 2C, and 3A), these data indicate that ROCK I may promote a steady state surface tension that prevents multiple initiations from occurring within a local area of the epithelial surface during normal development.

Given that ROCK I affects MLC2 activation, we expected to detect morphological changes on the tissue level. We found that SMGs treated for 24 hours with Y27632, H-1152, or blebbistatin showed a significantly reduced thickness in the Z dimension relative to their vehicle control-treated counterparts (Fig 3E). Furthermore, morphometric analysis revealed a corresponding increase in the diameter of SMGs treated with the ROCK and myosin inhibitors relative to vehicle controls (Fig 3E). These changes in tissue morphology cannot be attributed to changes in apoptosis (Fig S1C, and data not shown). Together, these measurements are indicative of a loss of cytoskeletal contractility at the tissue level when ROCK I and myosin II are inhibited.

Inhibition of ROCK and myosin activity decreases SMG cell proliferation

We next questioned what mechanism might regulate the transition of an initiated cleft to one competent for progression, as well as the molecular and cellular basis for cleft progression itself. Because recent studies suggest that enhanced cytoskeletal contractility can promote cell proliferation (Chen, 2008; Wozniak and Chen, 2009), we hypothesized that cell proliferation might be involved in cleft progression despite previous reports that deny a role for proliferation in SMG cleft initiation (Nakanishi et al., 1987). To this end, we measured SMG cell proliferation at 24 hrs following a BrdU pulse in the presence of vehicle control, Y27632, H-1152, or blebbistatin. ROCK and myosin-inhibited SMGs exhibited significantly decreased cell proliferation, as compared to their vehicle control-treated counterparts (Fig 4A and 4B, Fig S3A), and with a magnitude similar to that of hydroxyurea (HU), which blocks the cell cycle in S phase by inhibiting ribonucleotide reductase (Karon and Benedict, 1972) (Fig S3A and S3C). Of note, these findings demonstrate that the observed inhibitor-induced increase in SMG diameter (Fig 3E) is not due to increased proliferation.Further, these results indicate that ROCK I and myosin II-mediated stimulation of proliferation may be involved in cleft formation.

Figure 4.

Inhibition of ROCK and myosin activity decreases SMG proliferation. (A) Proliferation was assessed in E13 SMGs cultured for 24 hours in the presence of 140 μM Y27632, 2 μM H-1152, 100 μM blebbistatin, or 0.5 mM HU with a BrdU assay: BrdU (proliferating nuclei, green), Na⁺/K⁺ ATPase (preferential for epithelial membranes, red). Proliferating nuclei are localized at the distal ends of progressing clefts (white arrows) in vehicle control SMGs, but are absent from the proximal cleft base (arrow heads), apparent in confocal images. Such BrdU-positive nuclei are greatly reduced with all inhibitor treatments. (B) Quantification of proliferating nuclei expressed as BrdU-positive nuclei to total nuclei (SYBR green-positive). (C) SMGs cultured for 24 hours with 0.5 mM HU exhibit initiated but not progressing clefts, apparent at higher magnification (white arrows). Quantification of the number of completed buds and initiated clefts is shown for SMGs treated for 24 hours with increasing concentrations of HU, in comparison with Y27632, H-1152, and blebbistatin. ANOVA **p< 0.01, N=5. Scale bars, 10 μm (A) and 250 μm (C).

Interestingly, we observed a large proportion of BrdU-positive nuclei around the periphery of epithelial buds, in the outer basal layer of epithelial cells in control SMGs. Because ROCK and myosin inhibition resulted in a global loss of BrdU-positive nuclei, there was a consequent decrease in the number of proliferating cells in this outer basal layer (Fig S3B, compare white arrows). Since the periphery of the bud is where clefts initiate, we hypothesized that this localized decrease in epithelial cell proliferation might contribute to the failure in cleft progression that occurs when ROCK signaling is perturbed. Examination of progressing clefts in vehicle control SMGs revealed that BrdU-positive nuclei are present at the distal-most regions (Fig 4A, arrows), but not in the proximal regions closer to the cleft base (Fig 4A, arrow heads). In contrast, these BrdU-positive nuclei were absent in the presence of both ROCK and proliferation inhibitors.

To further explore the relationship between ROCK-mediated effects on proliferation and cleft formation, we performed morphometric analysis on HU-treated SMGs and observed a dose-dependent decrease in the number of completed epithelial buds when compared with vehicle control-treated SMGs (Fig 4C and Fig S3D). Furthermore, while clefts initiated normally in the presence of HU, they failed to progress, similar to clefts in the presence of the ROCK and myosin inhibitors (Fig 4C, compare to Fig 1B, Fig 2C, and Fig 3A). However, unlike ROCK and myosin inhibitor treatments, which increase the number of initiated clefts, HU-treated SMGs showed no significant difference in the number of initiated clefts relative to vehicle control (Fig 4C). These results suggest that ROCK- and myosin-mediated effects on the number of initiated clefts are independent of cell proliferation. However, since clefts in the HU-treated SMGs fail to progress, the subsequent transition to progression-competent clefts appears to require cell proliferation.

SMG cleft progression requires both a contractive and proliferative contribution

Since ROCK and myosin inhibitors both decrease cell proliferation and contractility, we investigated the relative contribution of inward-driven contractile forces versus outward-driven bud expansion forces to cleft progression. We performed time-lapse imaging of SMGs grown in the presence or absence of ROCK, myosin, or proliferation inhibitors (Movie S2) and morphometric analysis on the cleft regions. During cleft progression, the distance that the base of the cleft moved inward was designated as the contractive index (CI). The expansion index (EI) was established as a metric of the opposing end bud outgrowth, and the progression index (PI) as a metric of cleft length (see Fig 5A for graphical depictions of the indices and Fig S4A and B for definition of the measurements and additional time-points, respectively). Consistent with our previous findings, the PI was significantly decreased in all inhibitor-treated SMGs (Fig 5C). In vehicle control-treated SMGs, bud outgrowth progressed significantly away from the starting point, defined as a positive EI, and the base of the cleft progressed slightly inwards toward the interior of the bud, defined as a negative CI (Fig 5B and 5C). Treatment of SMGs with HU resulted in SMGs with significantly reduced proliferation and a coordinately reduced level of EI, but with a CI indistinguishable from control (Fig 5B and 5C). In contrast, when SMGs were treated with ROCK or myosin inhibitors, we observed a significantly reduced EI, but also a reversal of the CI into a net positive contribution, consistent with a loss of cytoskeletal contractility at the surface of the bud. Together these findings lead us to conclude that cleft progression is due primarily to a ROCK/myosin II-induced outgrowth of epithelial cells in regions adjacent to the progressing cleft and to a lesser extent from an inward-directed contractile force.

Figure 5.

Cleft progression requires both a proliferative and contractile contribution. (A) Schematic diagram of the expansion index (EI, cyan dotted line), representing epithelial outgrowth, and the contractive index (CI, red dotted line), representing cleft base ingression, relative to a fixed starting point (base of initiated cleft, black dotted line). The progression index (PI) is a metric of cleft length. Measurements and calculations are detailed in Fig S7. (B) Montage of selected images from widefield time-lapse imaging of cleft regions (see Movie S2) with the base of clefts (time 0, black dotted line), extent of bud outgrowth (final time point, cyan line), and base of clefts (final time point, red line) indicated for control, 140 μM Y27632, 100 μM blebbistatin, and 0.5 mM HU-treated E13 SMG. (C) Quantification of EI, CI, and PI (in AU). ANOVA *p<0.05, **p<0.01, N=6. Scale bar, 50 μm.

ROCK-mediated actomyosin contractility is required for the assembly of fibronectin (FN) into the basement membranes of branching SMGs

Thus far, we have demonstrated that cleft progression requires both a contractive and proliferative contribution. We next sought to determine the molecular mediators that integrate both of these cellular processes. Previous studies have also demonstrated that the ECM protein fibronectin (FN) is required for cleft formation during SMG branching morphogenesis (Sakai et al., 2003). Based on these studies and others implicating ROCK-mediated actomyosin contractility in FN matrix assembly in fibroblasts (Yoneda et al., 2007; Zhong et al., 1998), we questioned whether the morphological perturbations observed upon ROCK and myosin inhibition result from altered FN dynamics during SMG development. Western analysis of E13 SMG cell lysates treated for 24 hours with vehicle control or ROCK and myosin inhibitors showed no change in the amount of total endogenous FN (Fig 6A). We used immunocytochemistry to determine if there was a change in the localization of FN within the basement membranes of SMGs treated with ROCK or myosin inhibitors and ROCK I siRNA. As shown in Fig 6B, FN is present continuously throughout the entire basement membrane of vehicle control-treated E13 SMGs, co-localizing with the basement membrane proteins laminin-111 and the heparan sulfate proteoglycan, perlecan (data not shown). In contrast, SMGs cultured in the presence of Y27632, H-1152, or blebbistatin exhibit disorganized FN within their basement membranes, with areas lacking FN and other areas having punctate stitches of FN (Fig 6B, compare white arrows). Similarly, treatment of SMGs with ROCK I siRNA also resulted in severe FN disorganization within their basement membranes, with minimal effects associated with negative control or ROCK II siRNAs (Fig 6B). Significantly, HU treatment did not result in disorganized FN in the basement membrane, as did the ROCK and myosin inhibitors (Fig S5). These results suggest that ROCK I-stimulated myosin activity is required for localized FN assembly into the basement membrane of developing SMGs, and that the FN disorganization observed in the basement membranes of ROCK- and myosin-inhibited SMGs is not due to a decrease in ROCK-stimulated cell proliferation.

Figure 6.

ROCK signaling is required for FN assembly in the basement membranes of branching SMGs. (A) Western analysis for FN was performed on extracts of E13 SMGs treated for 24 hours with 140 μM Y27632, 2 μM H-1152, or 100 μM blebbistatin. Quantification of total endogenous FN, normalized to GAPDH and expressed relative to vehicle control, demonstrates no significant difference between treatments. (B) SMGs treated with Y27632, H-1152, blebbistatin, or ROCK I siRNA exhibit punctate, disorganized FN (green) in the basement membrane surrounding epithelial buds (white arrows). E-cadherin (red) marks epithelial buds in confocal images acquired at the center of SMGs. (C) SMGs and SCA-9 cells treated with Y27632, H-1152, or blebbistatin exhibit decreased DOC-insoluble FN relative to vehicle control. Quantification of SMG and SCA-9 cell DOC extracts was normalized to vimentin and GAPDH, respectively, and expressed relative to vehicle control. Vimentin (DOC insoluble) and actin (DOC soluble) confirmed separation of the two fractions. (D) Immunostaining of total exogenous FN matrices (green) by the L8 antibody (red) is decreased in SCA-9 cells treated with 50 μM Y27632 (white arrows), as observed with widefield images. (E) Confocal images acquired from the center of E13 SMG epithelial rudiments treated with 140 μM Y27632 show a reduction in L8 staining (red) of the basement membrane around the periphery of epithelial buds relative to vehicle control-treated SMGs. Scale bars, 5 μm (B), 20 μm (D), and 20 and 10 μm (E).

Since cell surface-associated, unassembled FN is soluble in deoxycholate (DOC), whereas fully extended, fibrillar FN is not (Sechler et al., 1996), we performed DOC extractions on ROCK- and myosin-inhibited intact SMGs to detect levels of fibrillar FN. Despite the fact that large amounts of FN are produced by mesenchymal cells, which may mask changes in FN assembly by the epithelium, decreased levels of assembled, fibrillar FN were detected in the DOC insoluble fraction in the presence of the inhibitors relative to vehicle control-treated samples (Fig 6C).

To assess the effect of ROCK inhibition on FN assembly specifically in salivary epithelial cells, we used the SCA-9 cell line (Barka et al., 2005; Barka et al., 1980). We determined that SCA-9 cells do assemble a fibrillar FN matrix, and that assembly of this matrix is disrupted when either ROCK or myosin is inhibited (Fig S6A), with FN primarily localized within short stitches and punctate dots along the cell periphery, reminiscent of FN localization within the basement membranes of ROCK- and myosin-inhibited SMGs (Fig 6B). To determine if these punctate regions represent unassembled FN bound to the cell surface, we performed time-lapse imaging of SCA-9 cells cultured with fluorescently labeled FN (Alexa⁶⁴⁷-FN) in the presence or absence of ROCK and myosin inhibitors. We first confirmed that Alexa⁶⁴⁷-FN was efficiently incorporated into pre-existing FN fibrils by comparing immunostaining with endogenous FN (Fig S6B). We then examined FN matrix assembly in these cells using time-lapse microscopy. We observed that Alexa⁶⁴⁷-FN was continuously incorporated into a fibrillar matrix over 18 hours in vehicle control-treated cells, but was evident only as cell surface-associated puncta in ROCK- and myosin-inhibited cells (Fig S6C).

Unfolding of dimeric FN molecules bound to cell surface integrin receptors unmasks a cryptic self-association site enabling the self-association of FN molecules to form a fibrillar matrix (Zhang et al., 1997; Zhong et al., 1998). The L8 monoclonal antibody binds to a site within the I₉ and III₁ modules in FN that is only exposed when FN is integrin-bound and stretched by cytoskeletal-mediated tension (Chernousov et al., 1987; Zhong et al., 1998). We investigated whether a reduction in actomyosin contractility was directly responsible for the reduced FN matrix assembly observed in SCA-9 cells and salivary gland cultures. Inhibitor-treated SCA-9 cells were incubated for 18 hours with Alexa⁶⁴⁷-FN followed by immunostaining with the L8 monoclonal antibody to label FN under tension and a total anti-FN antibody to detect the entire FN matrix. For SCA-9 cells treated with vehicle control alone, L8 recognized a FN matrix that was indistinguishable from that detected with both the polyclonal antibody and Alexa⁶⁴⁷- FN (Fig S7A). In contrast, L8 staining was markedly decreased in cells cultured with Y27632, H-1152, or blebbistatin (Fig 6D and Fig S7B, compare white arrows). Similar results were observed in E13 SMG epithelial rudiments, with L8 monoclonal antibody reactivity detected primarily within the basement membranes of vehicle control-treated rudiments, as determined by co-localization with Alexa⁶⁴⁷-FN, which was previously shown to incorporate into the epithelial basement membrane of E12 SMG epithelial rudiments (Larsen et al., 2006) (Fig 6E). In contrast, such immunoreactivity is significantly decreased in E13 rudiments cultured in the presence of Y27632 (Fig. 6E), H-1152, or blebbistatin (data not shown), instead exhibiting only remnant puncta of L8 staining. We conclude that ROCK I-mediated actomyosin contractility is required for the stretching of cell surface-bound FN into an assembly-competent state in salivary gland epithelial cells, and that when either ROCK I or myosin activity are inhibited, FN fibril assembly is compromised.

Fibronectin promotes SMG epithelial cell proliferation and cleft progression in a ROCK-dependent manner

We next questioned whether the defect in FN assembly into the basement membranes of ROCK- and myosin-inhibited SMGs was directly linked to the reduction in proliferation observed upon ROCK and myosin inhibition, and hence the lack of cleft progression. To this end, we cultured E13 SMGs for 24 hours in the presence of increasing concentrations of FN siRNA, as previously reported for E12 SMGs (Sakai et al., 2003). FN siRNA, but not a non-targeting negative control siRNA, resulted in a dose-dependent inhibition of branching due to a decrease in the number of cleft formation events, which could be rescued by the addition of exogenous FN to the culture medium (Fig 7A and Fig S8A). Interestingly, such rescue was dependent upon ROCK activity, since the addition of Y27632 at the same time as exogenous FN to FN siRNA-treated SMGs completely abrogated this effect (Fig 7A). We performed Western analysis on SMG cell lysates treated with increasing concentrations of FN siRNA to confirm the efficiency of FN knockdown, and observed approximately 50% knockdown when siRNA was used at a 500 nM concentration (Fig S8B). Together, these results confirm that FN promotes cleft progression not only in E12 SMGs, as was previously reported (Sakai et al., 2003), but also in E13 SMGs. Furthermore, since the FN siRNA rescue requires ROCK activity, these results suggest that it is FN assembly in the basement membrane, and not just FN itself, that is specifically required for cleft progression.

Figure 7.

FN promotes epithelial cell proliferation in a ROCK-dependent manner. (A) Treatment of E13 SMGs for 24 hours with 500 nM FN siRNA, but not negative control siRNA, inhibits branching morphogenesis. Such inhibition can be rescued by the addition of 0.5 mg/mL exogenous plasma FN to the culture media, but not in the presence of 140 μM Y27632. Quantification of the number of buds after 24 hours in culture under the indicated conditions. (B) E13 SMGs were cultured for 24 hours under the indicated culture conditions and pulsed for 2 hours with BrdU. BrdU nuclei (cyan) are greatly reduced in the presence of 500 nM FN siRNA, which can be rescued by the addition of 0.5 mg/mL exogenous FN to the culture media, but not in the presence of 140 μM Y27632. E13 SMGs cultured for 24 hours in the presence of increasing concentrations of exogenous FN exhibit increased incorporation of BrdU label, but not in the presence of 140 μM Y27632. FN knockdown and incorporation were confirmed by staining for FN (green). Confocal images were acquired at the center of each SMG. Quantification of BrdU-positive nuclei, expressed as a ratio to the total number of SYBR green-positive nuclei. ANOVA *p<0.05, **p<0.01, N=5. Scale bars, 250 μm (A), 100 μm (B).

We have demonstrated that cleft progression requires cell proliferation. To assess the effect of FN knockdown on SMG cell proliferation, E13 SMGs cultured for 24 hours in the presence of FN siRNA were subjected to a BrdU pulse. We observed a dose-dependent decrease in the ratio of BrdU-positive to total nuclei with increasing concentrations of FN siRNA, as confirmed by co-staining for FN (Fig 7B and Fig S8C). This defect in proliferation could be partially rescued by the addition of exogenous FN to FN-siRNA treated SMGs, but also required ROCK activity, since FN siRNA-treated SMGs cultured with both exogenous FN and Y27632 at the same time exhibited a level of proliferation comparable to that of glands cultured with FN siRNA alone (Fig 7B). To determine whether FN alone was sufficient to promote SMG epithelial cell proliferation, we cultured intact E13 SMGs for 24 hours with increasing concentrations of exogenous FN added to the culture media. We observed a dose-dependent increase in the ratio of BrdU-positive nuclei (Fig 7B), as well as an increase in the number of buds (Fig S9A and B), both of which were also dependent upon ROCK activity. SMGs treated with the same concentrations of BSA did not show a change in BrdU-positive nuclei or in the number of buds, thereby indicating that these changes are specifically caused by FN. We conclude that ROCK I-mediated actomyosin contractility is required for the assembly of FN into the basement membrane of branching SMGs, which in turn stabilizes an initiated cleft to facilitate its progression by enhancing localized epithelial cell proliferation. The point at which a threshold of contractility is reached within an initiated cleft to facilitate the transition to a progressing cleft, we refer to as a mechanochemical checkpoint.

Discussion

The initial step of salivary gland branching morphogenesis is the formation of clefts in the basement membrane surrounding a primary epithelial bud. Following initiation, clefts progress, or extend, to a certain point before terminating, and the process begins anew. Prior to this study, it was unclear if cleft progression was a continuation of cleft initiation or a separate event. Here, we describe a crucial role for ROCK I-mediated signaling through its downstream effector MLC2 in the differential regulation of both cleft initiation and progression during SMG branching morphogenesis. Our data support a role for ROCK I/myosin II-mediated contractility in regulating a checkpoint facilitating the transition of initiated clefts to a progression-competent state. Once a cleft passes through this checkpoint, the major contributor to cleft progression is cell proliferation with a contractile component. Our data also supports a role for ROCK I/myosin II in a negative feedback loop that maintains the basement membrane in a contracted state that is not immediately competent to initiate additional clefts (Fig 8).

Figure 8.

Proposed model for ROCK I-mediated actomyosin contractility during SMG cleft formation. ROCK I signaling regulates a checkpoint by which an initiated cleft undergoes a transition to a progression-competent, stabilized cleft, requiring localized contractility-mediated FN assembly. FN promotes localized epithelial cell proliferation, which generates an outward-directed expansion force that overcomes surface tension resulting in cleft progression, ROCK I-mediated actomyosin contractility also stimulates a negative feedback loop, temporarily preventing initiation of additional clefts.

We propose a model for cleft progression in which a localized increase in ROCK I-mediated actomyosin contractility reaches a certain threshold that triggers the assembly of FN fibrils in the cleft region, which in turn promotes epithelial cell proliferation in the regions adjacent to the stabilized cleft. The outward directed force generated by such proliferative outgrowth would overcome the steady state surface tension at the periphery of epithelial buds, thus resulting in cleft progression (Fig 8). Our results indicate that proliferation is not required for cleft initiation, which is in agreement with a previous report (Nakanishi et al., 1987), but demonstrate that progression of initiated clefts does depend upon cell proliferation.

The decrease in proliferation that we observed upon treatment of SMGs with FN siRNA and blebbistatin suggests that the proliferative effects of ROCK I signaling are primarily mediated through FN. Indeed, it has recently been shown that FN enhances proliferation in a mammary epithelial 3D cell culture model (Williams et al., 2008) and that cardiac myocyte proliferation is regulated in part by FN binding to β1 integrins (Ieda et al., 2009). However, we cannot rule out the possibility that to some degree ROCK I can promote SMG cell proliferation independently of FN and independent of myosin. Interestingly, a recent study demonstrated that signaling through ROCK modulates the G1/S transition by many mechanisms in mouse NIH 3T3 fibroblasts, some requiring ROCK-mediated effects on the actin cytoskeleton and others not (Croft and Olson, 2006). Since growth factor-dependent pathways signal through ERK1/2 to promote proliferation in branching SMGs (Kashimata et al., 2000; Steinberg et al., 2005), and a contracted actin cytoskeleton is implicated in the maintenance of robust MAPK signaling (Croft and Olson, 2006; Welsh et al., 2001), cross-talk may exist between growth factor and ROCK I-stimulated proliferation in branching SMGs.

We identified FN assembly as a crucial downstream effect of ROCK I/myosin II signaling in cleft progression. Previous studies demonstrated a requirement for FN and its receptor integrin α5β1 in the process of SMG cleft formation (Sakai et al., 2003) and inward-directed FN assembly in progressing clefts (Larsen et al., 2006), yet signals regulating FN assembly were not identified. Here, we demonstrate that signaling through ROCK I and myosin II drives organized FN matrix assembly in branching SMGs. These results, in combination with the previously demonstrated involvement of FN in cleft formation (Larsen et al., 2006; Sakai et al., 2003), are consistent with a requirement for FN assembly in the basement membrane to stabilize clefts, rendering them competent for progression. We propose that such stabilization is required for passage through a mechanochemical checkpoint that senses the contractile state of the basement membrane at the initiated cleft to facilitate the transition of initiated clefts into a progression-competent state. Although the biochemical composition of this checkpoint is not known, it likely requires inside-out integrin signaling.

Our results are consistent with other studies showing that ROCK-mediated actomyosin contractility promotes the unfolding of cell surface-bound FN in 2D cell cultures to expose a cryptic self-association site that enhances its assembly into a fibrillar matrix (Baneyx et al., 2002; Yoneda et al., 2007; Zhong et al., 1998). We detect a decreased amount of FN in the DOC insoluble fraction of ROCK- and myosin-inhibited SMGs, although this decrease is much more obvious in obvious in our 2D cell culture experiments with SCA-9 cells. This is not altogether surprising, given that DOC insoluble FN has recently been reported to be difficult to detect in in vivo Xenopus extracts (Dzamba et al., 2009). Taken together with our results with the L8 monoclonal antibody, however, we conclude that ROCK I-mediated actomyosin contractility is an important determinant of FN assembly in branching SMGs, primarily due to its ability to promote the extension of FN into an assembly-competent state. Interestingly, in embryonic lung, the blockade of ROCK signaling with Y27632 also inhibits branching. Such effects were suggested to be a result of defects in the actomyosin-dependent contraction of the basement membrane, and a consequent loss of localized differentials in basement membrane remodeling and epithelial cell proliferation (Moore et al., 2005). While FN is also required for branching in the lung (De Langhe et al., 2005; Sakai et al., 2003), a role for ROCK in this mechanism was not investigated. Furthermore, no direct links were established between basement membrane remodeling and localized cell proliferation. This is the first report to demonstrate that ROCK/myosin II signaling regulates FN assembly within the basement membrane in cleft formation during branching morphogenesis, and that FN itself specifically induces epithelial cell proliferation to stimulate cleft progression.

Interestingly, we identified unexpected roles for ROCK I/myosin II-mediated signaling in cleft initiation. The observed decrease in proliferation of the subset of proliferating cells localized to distal cleft regions when ROCK I or myosin activity were inhibited correlated with incomplete clefts failing to progress. However, this decreased proliferation did not account for the increased numbers of these small initiated clefts in the surface of the epithelium when ROCK I or myosin was inhibited, since treatment with HU did not significantly increase the number of such clefts. This indicates that cleft initiation events relate to ROCK I/myosin II-mediated changes in cellular contractility. Our results are compatible with a model in which decreased contraction in the presence of ROCK I or myosin inhibitors destabilizes the basement membrane, allowing additional cleft initiations in a proliferation-independent manner during normal development. We propose that ROCK I-mediated actomyosin contractility at the bud periphery constitutes a negative feedback loop that locally maintains the basement membrane in a contracted state until completion of the currently progressing cleft, thus preventing further cleft initiations. Indeed, such an occurrence would not be without precedence, since a recent study by Fischer et al. in a 3D model of endothelial cell branching morphogenesis identified ROCK-mediated myosin II activation and the resulting cortical tension as a negative regulator of pseudopodial branch initiation (Fischer et al., 2009).

A general assumption regarding regulation of branching morphogenesis has been that the primary point of regulation is cleft initiation itself, with a possible initiating signal of mesenchymal origin. In fact, the mechanism of cleft initiation remains unknown although putative initiator proteins have been identified, one of which is a class 3 semaphorin (Chung et al., 2007). However, our results presented here suggest that instead of a pre-determined cleft initiation site regulated by an initiator signal, the entire epithelial surface may be competent to undergo cleft initiation events, and that the primary point of regulation may be the checkpoint prior to cleft progression. This model may explain why epithelial tissue can undergo branching in the absence of mesenchyme.

The effect of ROCK signaling in other branching organs is is reported to differ from what we report here. In the lung, ROCK signaling positively regulates epithelial branching (Moore et al., 2005), while in the embryonic kidney, ROCK was reported to be a negative regulator of this process (Korostylev et al., 2008; Meyer et al., 2006). However, upon closer examination, some similarities between the proposed mechanisms of ROCK signaling in these different systems are apparent. We here identified a role for ROCK in FN assembly in the basement membrane during cleft progression, whereas in lung, a ROCK-dependent change in basement membrane thickness was reported (Moore et al., 2005). When isolated ureteric buds are treated with ROCK inhibitors, they exhibit very small buds that do not elongate ductal structures (Meyer et al., 2006), suggesting a role for ROCK signaling in an analogous cleft progression checkpoint during kidney morphogenesis. Further studies will be required to understand specific similarities and differences in ROCK-stimulated signaling during branching morphogenesis.

In conclusion, we report evidence for a ROCK I/myosin II-controlled checkpoint regulating the transition of clefts from initiation to progression. Lack of cleft progression is associated with decreased cytoskeletal contractility, decreased FN assembly, and subsequent decreased proliferation. Additionally, in the presence of ROCK I/myosin II inhibition, additional cleft initiation events occur in a proliferation-independent manner due to a loss in contractility in the basement membrane at the periphery of epithelial buds. Our data indicate that cleft initiation and progression are, in fact, two distinct morphological events during SMG branching and that a checkpoint regulates the transition from one to the other. To the best of our knowledge, this is the first published report describing a checkpoint regulating the process of branching morphogenesis.