Rock2 controls TGFβ signaling and inhibits mesoderm induction in zebrafish embryos

The transforming growth factor β (TGFβ) signal transduction pathway, which involves multiple ligands, including Activin, Nodal, Vg1 and GDFs, is a key player during early development of vertebrate embryos. Among the TGFβ ligands, Nodal proteins are absolutely required for mesendoderm induction and the establishment of left-right asymmetry (Schier, 2003; Shen, 2007; Tian and Meng, 2006). These ligands initiate signal transduction by binding to type I and type II serine/threonine kinase receptors on the plasma membrane, which results in the activation of type I receptors (Shi and Massague, 2003). The activated receptors phosphorylate the downstream effectors Smad2 and/or Smad3, and the phosphorylated Smad2 or Smad3 form complexes with Smad4 and translocate into the nucleus to control the expression of target genes (Jia et al., 2008; Schmierer and Hill, 2007). The multiple steps that are involved in the TGFβ signal transduction pathway allow for precise regulation by many modulators through diverse mechanisms (Chen and Meng, 2004).

Intracellular trafficking of internalized TGFβ receptors has been found to play a role in regulating TGFβ signaling. The membrane-anchored receptors are internalized by clathrin-dependent and clathrin-independent processes (Anders et al., 1997; Di Guglielmo et al., 2003; Lu et al., 2002; Mitchell et al., 2004; Zwaagstra et al., 2001). The internalized receptors are then recycled back to the membrane or degraded via proteasomal or lysosomal pathways (Di Guglielmo et al., 2003; Kavsak et al., 2000; Mitchell et al., 2004). Distinct targeting of the internalized receptors, a process regulated by multiple factors, determines the duration of the intracellular signaling, and thus has profound effects on vertebrate embryonic development (Jullien and Gurdon, 2005; Shim et al., 2006; Zhang et al., 2004).

Targeting and trafficking of intracellular proteins, including internalized receptors, rely on functions of actin-filaments and microtubules (Salinas, 2007; Soldati and Schliwa, 2006). RhoA, a member of the Ras homology family of small GTPases, is a key regulator of the actin cytoskeleton during cell morphogenesis and motility (Etienne-Manneville, 2004; Ridley, 2006). The Rho-associated serine/threonine kinases Rock1 (also termed Rok1 and Rokβ) and Rock2 (also termed Rok2, Rokα and Rho kinase) are immediate downstream targets of RhoA (Amano et al., 1997; Ishizaki et al., 1996; Leung et al., 1995; Matsui et al., 1996); they phosphorylate a variety of substrates, including myosin light chain (Amano et al., 1996; Totsukawa et al., 2000), the myosin-binding subunit (MYPT1) of the myosin-associated form of protein phosphatase 1 (Velasco et al., 2002), LIM kinases (Ohashi et al., 2000; Sumi et al., 2001), adducin (Fukata et al., 1999) and Zipper-interacting protein kinase (Hagerty et al., 2007). Rho kinases may intervene with intracellular transduction and propagation of various extracellular signals by assisting intracellular targeting of the signaling pathway components. Several reports have shown that RhoA and Rho kinases can be activated by TGFβ signaling and are required for mediating the effects of TGFβ signals in epithelial-to-mesenchymal transdifferentiaion, actin filament rearrangement and growth inhibition (Bhowmick et al., 2001; Edlund et al., 2002; Kamaraju and Roberts, 2005). However, it is largely unknown whether RhoA and Rho kinases function to down-regulate TGFβ signal transduction.

Rho kinases have been found to be important regulators of cell contraction, adhesion and migration as well as proliferation and apoptosis (Riento and Ridley, 2003; Shi and Wei, 2007). Many studies using Rho kinase inhibitors have indicated that Rho kinases are involved in various physiological and pathological processes (Shi and Wei, 2007; Wettschureck and Offermanns, 2002). Developmental roles of Rho kinases have been examined in mice by knockout approaches. Rock1 null mice died soon after birth with failure of the closure of the eyelids and the ventral body wall (Shimizu et al., 2005). Most Rock2-deficient mouse embryos died in uterus with placental dysfunction and intrauterine growth retardation; the few survivors were born runts and developed without gross anomalies (Thumkeo et al., 2003). In Xenopus and zebrafish, Rock2 is required for cell convergence and extension movements during gastrulation (Marlow et al., 2002). These reports indicate that Rock1 and Rock2 play important roles in regulating cell movements in vertebrate embryos. However, it remains unclear as to whether they are involved in embryonic patterning during early development.

We have previously demonstrated that the late-endosomal protein Dpr2 binds to and facilitates lysosomal degradation of TGFβ type I receptors, and that zebrafish Dpr2 functions to inhibit mesoderm induction by Nodal signals during early embryogenesis (Zhang et al., 2004). The function of Dpr2 is evolutionarily conserved in vertebrates (Su et al., 2007). In this study, we identified Rock2 as binding partner of Dpr2. We disclose that, like Dpr2, Rock2 attenuates TGFβ signaling by promoting lysosomal degradation of TGFβ type I receptors in mammalian cells. In zebrafish embryos, Rock2 inhibits mesendoderm induction by Nodal signals. Thus, our findings have revealed novel functions of Rock2, in signaling transduction and in embryonic development.

Using a yeast two-hybrid screen to search for Dpr2-interacting factors in zebrafish, we identified Rock2a as a Dpr2-binding partner. We then looked at whether Rock2a also plays a role in regulating TGFβ signal transduction. We first investigated the effects of zebrafish Rock2a on TGFβ signaling using the ARE-luciferase reporter that is driven by Activin responsive elements (Huang et al., 1995). The expression of ARE-luciferase in Hep3B cells was stimulated by the addition of TGFβ1 (Fig. 1A) or by transfection of the constitutively active (ca) form of type I receptors ALK5 (Fig. 1B) or ALK4 (data not shown), whereas the reporter activity was inhibited by co-transfection of zebrafish Rock2a in a dose-dependent manner. The inhibitory effect of zebrafish Rock2a on ARE-luciferase reporter expression was also observed in HepG2 cells (Fig. 1C). Furthermore, overexpression of human ROCK2 also inhibited ARE-luciferase reporter expression (Fig. 1C). The inhibitory effect of human ROCK2 overexpression on ARE-luciferase expression in response to TGFβ1 stimulation was blocked in the presence of the Rock kinase inhibitor Y-27632 in Hep3B cells (data not shown), indicating that the Rock catalytic activity is required. These results imply that Rock2 functions to impede TGFβ signal transduction, and that this function is evolutionarily conserved.

To investigate the effects of endogenous ROCK2 on TGFβ signaling, we made two siRNA constructs, pSilencer-ROCK2-siRNA1 and pSUPER-ROCK2-siRNA2 that were designed to target different regions of human ROCK2 mRNA. Transfection of either of the constructs resulted in a decrease in HA-ROCK2 expression levels (Fig. 1D). Quantitative RT-PCR analysis revealed that transfection in HEK293T cells with these siRNAs caused significant reduction of endogenous ROCK2 mRNA levels but had little effect on ROCK1 mRNA levels (data not shown). These results indicate that both ROCK2-siRNAs were able to effectively and specifically target ROCK2. When endogenous ROCK2 in Hep3B cells was knocked down with pSilencer-ROCK2-siRNA1 or pSuper-ROCK2-siRNA2, TGFβ1 induction of the ARE-luciferase reporter expression was further enhanced (Fig. 1E). Furthermore, the effect of ROCK2-siRNA1 on TGFβ1-stimulated ARE-luciferase reporter expression was blocked by co-expression of the ROCK2 mutant ROCK2 (PM) (Fig. 1F), which contains six mutations within the ROCK2-siRNA1 recognition site and is thus resistant to ROCK2-siRNA1. These results confirm that ROCK2 is able to inhibit TGFβ signaling in human cells.

Dpr2 has been found to inhibit TGFβ signaling by accelerating the degradation of TGFβ type I receptors (Su et al., 2007; Zhang et al., 2004). We speculated that Rock2 might regulate TGFβ signaling through a similar mechanism. To test this, we first examined the physical interaction between Rock2 and TGFβ type I receptors by co-immunoprecipitation. We found that overexpressed fish Rock2a could associate with co-expressed wild-type (ALK5-FLAG) and constitutively active (caALK5-FLAG) forms of the receptor ALK5 in human HEK293T cells (data not shown). Overexpressed HA-ROCK2 bound to a small amount of endogenous ALK5 without the stimulation of exogenous TGFβ1, but their association was intensified dramatically by the addition of TGFβ1 (Fig. 2A), suggesting that Rock2 preferentially associates with the activated receptor.

Next, we tested whether overexpression of Rock2 promoted degradation of ALK4 or ALK5. When fish Rock2a was co-overexpressed, protein levels of caALK4 or caALK5 were markedly reduced in a dose-dependent fashion (Fig. 2B). Even in the presence of cycloheximide (CHX), a protein synthesis inhibitor, transfection of fish Rock2a resulted in apparent reduction of the caALK5 level (Fig. 2C) and led to a shortened half-life for caALK5 (Fig. 2C'). This implied that the Rock2a-induced caALK5 reduction was not due to impaired transcription or translation. Similarly, transfection of human ROCK2 caused reduction of the caALK5 level in the presence of CHX (Fig. 2D), suggesting a conserved function of Rock2 from fish to human. Consistent with these observations, knockdown of ROCK2 in human HEK293T cells by ROCK-siRNAs gave rise to an increase in the amount of overexpressed caALK5 (Fig. 2E). Thus, Rock2 controls TGFβ signaling by accelerating the degradation of TGFβ type I receptors.

We then made efforts to identify the pathway(s) through which Rock2 promotes degradation of ALK4 and ALK5. We found that the presence of the lysosomal inhibitors NH₄Cl, bafilomycin A1 or chloroquine, but not of the proteasome inhibitor MG132, prevented Rock2a-induced caALK5 degradation (Fig. 3A). We obtained similar results for caALK4 (data not shown). Thus, Rock2 promotes the receptor degradation through the lysosomal pathway but not the proteasomal pathway. We then adopted a biotinylation method (Lu et al., 2002) to investigate the effect of Rock2 on the degradation of TGFβ receptors anchored on the cell surface. In the precipitates pulled-down with streptavidin 3 hours after activation of internalization, the amount of endogenous cell surface ALK5, detected with anti-ALK5 antibody, was markedly reduced by overexpression of HA-ROCK2 (Fig. 3B). Taking these results together, we hypothesize that Rock2 binds to and promotes the degradation of endogenous ALK5 internalized from the cell surface.

As a serine/threonine kinase, Rock2 exerts its biological effects by phosphorylating substrates (Riento and Ridley, 2003). We then looked at whether the kinase activity of Rock2 was required for its ability to inhibit TGFβ signaling. We first used the ROCK kinase inhibitors Y27632 (Ishizaki et al., 2000) and HA1077 (Davies et al., 2000) to address this issue. Treatment with either of the inhibitors in mammalian Hep3B cells resulted in elevated levels of TGFβ1-induced ARE-luciferase reporter expression (Fig. 3C). The presence of Y27632 or HA1077 also caused an increase in caALK5-HA levels and attenuated the HA-ROCK2-induced caALK5-HA degradation (Fig. 3D; data not shown for HA1077 treatment). These results suggest that the kinase activity of ROCKs is required for the inhibitory effect on TGFβ signal transduction. We further tested the role of Rock2 kinase activity in the inhibition of TGFβ signaling by using several mutant forms of human ROCK2 (Amano et al., 1997; Amano et al., 1999; Tan et al., 2002), including the constitutively active mutant ROCK2-CAT, which retains the N-terminal region (6-553 aa) only, ROCK2-CAT KD (identical to ROCK2-CAT except for loss of kinase activity) and the dominant negative mutant ROCK2-RB/PH(TT), which carries the C-terminal region (1125-1388 aa) but lacks RhoA-binding activity. Like the full-length ROCK2, ROCK2-CAT overexpression inhibited the caALK5-stimulated ARE-luciferase reporter expression in Hep3B cells (Fig. 3E). By contrast, overexpression of ROCK2-CAT KD had no effect, but ROCK2-RB/PH(TT) overexpression enhanced the reporter expression (Fig. 3E). Furthermore, overexpression of dnRock2a, a dominant negative mutant of zebrafish Rock2a that is similar to ROCK2-RB/PH(TT) (Marlow et al., 2002), elevated caALK5-HA protein levels in HEK293T cells (Fig. 3F). Taken together, these data support the idea that the kinase activity of Rock2 is required for regulating TGFβ type I receptor degradation.

As shown above, fish Rock2a is able to downregulate TGFβ signaling by facilitating the degradation of ALK4 and ALK5 receptors in mammalian cells. We wondered whether Rock2a could also regulate TGFβ signaling in zebrafish embryos. Because Nodal signaling is essential for mesoderm induction (Tian and Meng, 2006), we looked at whether rock2a had an effect on mesoderm induction in zebrafish embryos. A previous study by Marlow et al. demonstrated in zebrafish that rock2a plays a role in cell convergence and extension movements during gastrulation by mediating non-canonical Wnt11 signaling (Marlow et al., 2002). To avoid interpretation ambiguity due to abnormal convergence and extension movements, we examined the expression of the Nodal targets goosecoid (gsc), chordin (chd) and no tail (ntl) at the sphere stage, well before the onset of gastrulation, following injection of zebrafish rock2a mRNA into one-cell embryos. As shown in Fig. 4A,A', overexpression of rock2a led to reduction or elimination of gsc, chd and ntl expression in the dorsal margin where dorsal mesoderm precursors are located, whereas injection of lacZ mRNA had little effect on gsc and ntl expression. Unexpectedly, injection of lacZ mRNA also caused reduction of chd expression, which might have occurred through unknown mechanisms. To confirm the in situ hybridization results, we performed quantitative RT-PCR analysis. As shown in Fig. 4B, rock2a overexpression caused significant reduction in gsc, ntl and chd transcript levels, which excludes the possibility that the change of the maker expression patterns is owing to abnormal convergent extension movements. By contrast, neither quantitative RT-PCR nor whole-mount in situ hybridization detected any decrease following rock2a overexpression, even if no increase, in the expression of bozozok (boz) (Fig. 4B,C), which is a direct target of canonical Wnt/β-catenin signaling (Leung et al., 2003; Ryu et al., 2001). This implies that rock2a does not downregulate canonical Wnt/β-catenin signaling in vivo. Thus, rock2a overexpression negatively regulates the expression of the Nodal targets during mesoderm induction, irrespective of the canonical Wnt signaling.

The next issue to investigate was whether rock2a inhibited mesoderm induction by antagonizing Nodal signaling in zebrafish embryos. As reported before (Rebagliati et al., 1998), overexpression of squint (sqt), a zebrafish Nodal gene, caused dramatic expansion of gsc, chd and ntl expression domains at the sphere stage (Fig. 5A). The inductive effect of sqt was obviously compromised by co-injection of rock2a mRNA. Zebrafish taram-a encodes a TGFβ-related type I receptor required for Nodal signaling, and overexpression of its constitutively active form tar^* induces mesendodermal marker expression (Aoki et al., 2002; Renucci et al., 1996). We found that induction of gsc, chd and ntl expression by injection of tar^* mRNA was also inhibited by co-injection of rock2a mRNA (Fig. 5B). Lefty1 (also known as Antivin) is a potent antagonist of Nodal ligands, and its overexpression in zebrafish embryos eliminates the formation of most mesendodermal tissues (Cheng et al., 2000; Thisse and Thisse, 1999). When 10 pg of lefty1 mRNA was injected, the expression levels of gsc, chd and ntl were decreased at the sphere stage (Fig. 5C). When the same amount of lefty1 mRNA was co-injected with 200 pg of rock2a mRNA, most of the injected embryos showed further decrease in the expression levels of these markers. Taken together, these data indicate that rock2a overexpression is able to block mesoderm induction by ectopic Nodal signaling and to enhance the inhibitory effect of lefty1 on mesoderm induction in vivo.

To obtain direct evidence of the effects of rock2a on Nodal signaling in fish embryos, we examined phosphorylated Smad2 levels after overexpression of sqt and rock2a. As shown in Fig. 5D, injection of sqt mRNA led to an increase in the phospho-Smad2 levels, but co-injection with rock2a mRNA compromised the stimulatory effect of sqt, while the total amount of Smad2 remained comparable. These results demonstrate that rock2a downregulates the Nodal signaling upstream of Smad2 phosphorylation in fish embryos, which is consistent with the fact that Rock2 accelerates the turnover of TGFβ receptors in mammalian cells.

A previous study carried out by Marlow and coworkers (Marlow et al., 2002) indicated that knockdown of rock2a with antisense morpholino oligonucleotides in zebrafish embryos fails to cause any detectable phenotypes, probably due to the presence of the abundant maternal deposit of Rock2 proteins (Marlow et al., 2002). Because the dominant negative mutant dnrock2a can attenuate the degradation of TGFβ receptor by interfering with endogenous ROCK2 in mammalian cells (Fig. 3F), we evaluated the biological function of endogenous Rock2 proteins by injecting dnrock2a mRNA into one-cell embryos and examining gsc, chd and ntl expression using whole-mount in situ hybridization and quantitative RT-PCR analysis at the sphere stage. In situ patterns of these markers were not altered obviously by dnrock2a overexpression (data not shown). However, quantitative RT-PCR analysis revealed that overexpression of dnrock2a resulted in significant increases in ntl and chd transcript levels, although the gsc transcript level was only marginally increased (Fig. 6A). These results suggest that endogenous Rock2 is required for mesoderm induction during the early development of zebrafish embryos.

We also took another approach, dnrock2a and lefty1 co-injection, to investigate the requirement of endogenous Rock2 for mesendoderm induction. As shown in Fig. 6B,B', lefty1-induced reduction of gsc, ntl and chd expression at the sphere stage was compromised by co-injection of dnrock2a mRNA. Overexpression of lefty1 led to reduction of the expression of the endodermal marker sox32 (also a Nodal target) at the 75% epiboly stage and of ntl expression in the axial mesoderm at the bud stage, but caused expansion of the expression domains of the ventral ectodermal marker gata2 at the 80% epiboly stage (Fig. 6C,C'). These effects of lefty1 overexpression were also alleviated by co-expression of dnrock2a. Our observations support the idea that endogenous Rock2 proteins play an inhibitory role in mesendoderm induction during zebrafish embryogenesis.

To confirm the functional relationship between Rock2 and Dpr2, we first tested their physical interaction by co-immunoprecipitation assay. As shown in Fig. 7A, overexpressed HA-Rock2a in HEK293T cells was co-immunoprecipitated with Myc-Dpr2, and vice versa. Moreover, endogenous human ROCK2 was found to associate with overexpressed Myc-tagged mouse Dpr2 in HEK293T cells (Fig. 7B). We next investigated whether the functions of Dpr2 and Rock2a in regulating TGFβ signaling were interdependent. Like the full-length Rock2a, transfection of HA-Rock2aΔDpr2, a Rock2a mutant that was depleted of 1068-1110 aa and failed to interact with Dpr2 (data not shown), attenuated the caALK5-stimulated expression of the ARE-luciferase reporter in Hep3B cells (Fig. 7C), suggesting that the effect of Rock2a is independent of its Dpr2-binding ability. When Dpr2 in HEK293T cells was knocked down by Dpr2-siRNA1 (Su et al., 2007), co-expression of HA-Rock2a was still able to effectively accelerate caALK5-HA degradation (Fig. 7D). Thus, it appears that Rock2a inhibits TGFβ signaling in a Dpr2-independent manner. By contrast, transfection of Dpr2 had little effect on the TGFβ1-stimulated ARE-luciferase reporter expression in Hep3B cells upon knockdown of endogenous ROCK2 by pSilencer-ROCK2-siRNA1 (Fig. 7E). Consistent with this observation, overexpression of Dpr2 failed to promote caALK5-HA degradation when ROCK2 was knocked down in HEK293T cells (Fig. 7F). These data imply that the inhibitory effect of Dpr2 on TGFβ signaling relies on ROCK2 function.

We then investigated whether ROCK kinase activity is required for binding to and facilitating the function of Dpr2. We generated a ROCK2 mutant, ROCK2K121G, by mutating K121 to G121 at the ATP-binding site. Like wild-type ROCK2, ROCK2K121G was able to bind to mouse Dpr2 (mDpr2) (Fig. 7G), suggesting that the kinase activity of ROCK2 is not required for associating with Dpr2. However, the presence of the ROCK kinase inhibitor Y-27632 reduced the inhibitory effect of fish Dpr2 overexpression on the TGFβ1-stimulated ARE-luciferase expression (Fig. 7H). Thus, the kinase activity of ROCK2 is dispensable for binding to Dpr2 but is necessary for assisting the function of Dpr2 in the control of TGFβ signaling.

In this study, we have demonstrated that Rock2, which was identified as a binding partner of Dpr2, acts to inhibit TGFβ signal transduction in mammalian cells. Mechanistically, Rock2 binds to and promotes the degradation of endocytosed TGFβ type I receptors via the lysosomal pathway. In zebrafish embryos, rock2a overexpression inhibits mesoderm induction by Nodal signals. Interference with endogenous Rock2 proteins by injecting a dominant negative mutant (dnrock2a) of rock2a compromises the inhibitory effect of lefty1 on mesendoderm marker expression, suggesting an involvement of endogenous Rock2 proteins in mesendoderm induction during early embryogenesis. Thus, our study discovered a novel role of Rock2 in modulating TGFβ/Nodal signaling and in controlling mesendoderm induction.

Rock1 and Rock2 have been found to play pivotal roles in cell motility. They regulate actin-cytoskeleton assembly and cell contractility by phosphorylating specific substrates, which ultimately exert their effects on the cytoskeleton (Riento and Ridley, 2003). We found that Rock2 could promote lysosomal degradation of endocytosed TGFβ type I receptors. Hypothetically, this effect of Rock2 may be associated with cytoskeleton-dependent intracellular trafficking of the endosomes containing TGFβ receptor towards lysosomes. Several studies have shown that RhoA and Rho kinases regulate endocytosis of transferrin and epidermal growth factor receptors (Lamaze et al., 1996), low-density lipoprotein receptors (Hrboticky et al., 2002), and equine herpesvirus 1 (Frampton et al., 2007). Kaneko et al. demonstrated that Rho kinases might inhibit endocytosis of epidermal growth factor receptors by phosphorylating endophilin A1, a key regulator of clathrin-mediated endocytosis (Kaneko et al., 2005). Accelerated degradation of the endocytosed TGFβ receptors by Rock2 might be ascribed to its ability to phosphorylate substrates that facilitate transportation of the TGFβ-receptor-bound endosomes. Further studies are needed to test this hypothesis.

On the one hand, we found that Rock2 was physically associated with Dpr2 and TGFβ type I receptors; on the other hand, Dpr2 is primarily located in the late endosomes, and it also binds to endocytosed TGFβ type I receptors (Zhang et al., 2004). What then is the relationship between Rock2 and Dpr2 in controlling TGFβ signaling? We note that Dpr2-induced degradation of TGFβ receptors is largely Rock2-dependent in mammalian cells, whereas the effects of Rock2 on TGFβ-receptor degradation are independent of Dpr2 function. We hypothesize that Dpr2 binds to TGFβ receptor(s) located on endosomes and presents such receptors to Rock2, which promotes movement of such `labeled' endosomes towards lysosomes along the cytoskeleton. Most probably, other unidentified adaptor proteins may have a function similar to Dpr2, which recognizes and presents TGFβ receptors to Rock2.

In zebrafish, rock2a is maternally expressed and its transcripts are distributed in all blastomeres, implying that it may be involved in specification of cell fate during early embryogenesis. A previous study found that knockdown of rock2a with antisense morpholinos failed to cause any detectable defects in early zebrafish embryos, even in convergence and extension movements during gastrulation (Marlow et al., 2002). One reason for this could be the presence of the abundant maternally derived Rock2a protein that is not targetable by morpholinos. Another reason could be the presence of other rock genes in the zebrafish genome, e.g. rock2b, rock2bl and probably multiple rock1 genes, which play redundant functions, in addition to rock2a. Further genetic studies are needed to disclose the requirement of endogenous Rho kinases in embryonic patterning.

Rho kinases act on many different substrates, implying that they may regulate multiple signaling pathways. Although Rock2 proved to be a negative regulator of TGFβ signaling in this study, several other studies have found that RhoA and Rho kinases positively regulate convergence and extension movements by mediating non-canonical Wnt signaling in zebrafish and Xenopus gastrulas (Jopling and den Hertog, 2005; Kim and Han, 2005; Marlow et al., 2002; Zhu et al., 2006). Our unpublished data indicate that Rock2a may also be a positive regulator of the canonical Wnt/β-catenin signaling pathway. Since these signaling pathways are all involved in embryonic patterning, weak phenotypes caused by genetic manipulations of Rock activity in zebrafish, as shown in this study, may be expected. It will be interesting to investigate how RhoA and Rocks differentially regulate different signaling pathways.

pCS2-Rock2a and pCS2+MT-dnRock2a were generous gifts from Lilianna Solnica-Krezel (Vanderbilt University Medical Center, Nashville, TN); ROCK2 (human cDNA clone KIAA0619) was a kind gift from KAZUSA DNA Research Institute (Chiba, Japan). Zebrafish Rock2 and human ROCK2 were subcloned into pCMV5 with the addition of an HA tag at the N-terminus. N- or C-terminal deletion mutants of Rock2 were generated according to standard molecular techniques. For human ROCK2 mutants, CAT encoding the catalytic domain of ROCK2 (6-553 aa) was inserted into pCMV5 with an HA tag, and RB/PH (TT) (941-1388 aa, with the point mutations N1036T and K1037T) inserted into pCMV5 with a FLAG tag (Amano et al., 1997; Amano et al., 1999; Tan et al., 2002). ROCK2 (PM), a human ROCK2 mutant, was made by mutating the ROCK2-siRNA1-recognizing sequence to GTAAGTCGGTCAACACGCG (substituted bases are underlined). Deletion and point mutations were generated by a PCR-based QuickChange Site-Directed-Mutagenesis (Stratagene, La Jolla, CA) method with Pfu polymerase. All of the inserts were verified by sequencing. The details of clone strategy and primer sequences are available upon request.

Two siRNAs targeting human ROCK2 were designed and constructed by vector-based siRNA technology. Their targeting sequences are 5′-GCAAATCTGTTAATACTCG-3′ (for ROCK2-siRNA1) and 5′-ATCAGACAGCATCCTTTCT-3′ (for ROCK2-siRNA2). Control (ctr) siRNA was designed to target the sequence 5′-AGCGGACTAAGTCCATTGC-3′, and was used as a negative control. Oligonucleotides of ROCK2-siRNA1 and ctr siRNA were inserted into ApaI/EcoRI sites of the vector pSilencer 1.0-U6 (Ambion, Austin, TX), and those of ROCK2-siRNA2, ROCK2 as well as ctr siRNA were cloned into BglII/HindIII sites of the vector pSUPER.retro (OligoEngine, Seattle, WA). Dpr2-siRNA1 was reported previously by Su et al. (Su et al., 2007).

Total RNAs were extracted from zebrafish embryos at the one-cell, 40% epiboly, 5-somite and 24-hour stages using Trizol reagent. The cDNA library was constructed using the BD Matchmaker Yeast Two-Hybrid System (Clontech, Palo Alto, CA) according to the manufacturer's instructions.

To screen Drp2-interacting proteins, the Drp2 N-terminal region of 479 amino acids was cloned into the bait vector pGBKT7. Yeast strain AH109 transformed with the bait was mated with Y187 cells that were pre-transformed with the zebrafish cDNA library, followed by screening with the BD Matchmaker Two-Hybrid System (Clontech).

HEK293T and HeLa cells were maintained in DMEM with 10% fetal bovine serum (FBS, Hyclone, South Logan, UT) in a water-saturated atmosphere of 5% CO₂. Hep3B and HepG2 cells were cultured in the same conditions, except that DMEM was replaced with MEM. Transfection for HEK293T and HepG2 cells was performed using the calcium-phosphate precipitation method, and transfection for Hep3B and HeLa cells was carried out using lipofectamine (Gibco BRL, Gaithersburg, MD). To keep a consistent total amount of transfected plasmids for each sample in one experiment, corresponding empty vector was added to the amount required. Reagents for treating cells included TGFβ1 (PeproTech Asia, Rehovot, Israel); cycloheximide, NH₄Cl, chloroquine and Y27632 (Sigma); and bafilomycin A1, MG-132 and HA1077 (Calbiochem).

For each transfection experiment, a plasmid DNA mixture consisting of 0.5 μg ARE-luciferase, 0.15 μg FAST2, 25 ng Renilla, 0.15 μg caALK5 or caALK4 (optional), and other plasmids as indicated in the figures was equally divided and used to transfect cells in six-well plates. Unless otherwise stated, cells were harvested with passive lysis buffer (Promega, Madison, WI) at 30 hours post-transfection. As for TGFβ stimulation, cells were washed with serum-free medium and incubated for an additional 12-16 hours with the addition of 50 or 100 pM TGFβ1 in MEM plus 0.2% FBS and then analyzed. Luciferase activity was measured by a luminometer (Berthold Technologies, Oak Ridge, TN), and normalized for transfection efficiency by the internal control Renilla and for correction with empty vector controls. Experiments were repeated in triplicate, and the data are represented as the mean ± s.d. of three independent experiments.

For the receptor expression level assay, HEK293T cells were transfected with 0.3 μg caALK5-HA or caALK4-HA, 0.1 μg GFP or GST and other plasmids as indicated per well in a six-well plate. Unless otherwise stated, the cells were incubated for 30 hours before analysis. As for co-immunoprecipitation, transiently transfected HEK293T cells in 100-mm dishes were harvested at 40 hours post-transfection; if required, cells were starved for 6 hours and then stimulated with 100 pM TGFβ1 for 2 hours before harvest. Cell lysate preparation, immunoprecipitation and western blotting were preformed as described previously (Zhang et al., 2004).

At 36 hours post-transfection, control cells or Rock2-transfected HeLa cells were labeled with 0.5 mg/ml NHS-SS-biotin (Pierce, Rockford, IL) for 40 min at 4°C, and then were incubated at 37°C in the presence of 100 pM TGFβ1 for 0 or 3 hours before harvested. The biotinylated cell surface proteins were precipitated with streptavidin beads (Pierce) and then eluted. ALK5 in the eluted proteins was detected by immunoblotting using anti-ALK5 antibody (Santa Cruz Biotechnology, Santa Cruz, CA).

Wild-type Tuebingen strain was used. Embryos were incubated in Holtfreter's solution at 28.5°C and staged according to Kimmel and coworkers (Kimmel et al., 1995). RNA synthesis, microinjection and whole-mount in situ hybridization were performed essentially as described previously (Zhao et al., 2003).

Total RNA was extracted from a pool of 120 injected embryos at the sphere stage by RNeasy Mini Kit (Qiagen, Valencia, CA). According to the manufacturer's instructions (Promega), cDNAs were synthesized from 2 μg RNA using oligo-dT primer in a volume of 20 μl. The volume of each cDNA sample was brought up to 100 μl in dH₂O and then quantitative real-time PCR was carried out with SYBR green I (OPE, Shanghai, China) in the M×3000P real-time PCR system (Stratagene). Expression levels of β-actin were analyzed in parallel as an internal control and each sample was run in triplicate. Primers for boz were 5′-TACTGCTGCTGCGTTCCA-3′ (forward) and 5′-TGCTCCGTCTGGTTGTCG-3′ (reverse); primers for other genes were as described by others (Liu et al., 2005; Xiong et al., 2006). For testing ROCK2-siRNAs specificity in HEK293T cells, primers for human ROCK1 were 5′-GACCTGTAACCCAAGGAGATG-3′ (forward) and 5′-CAGGAAAGTGGTAGAGTGTAGGA-3′ (reverse), those for human ROCK2 were 5′-CAACTGTGAGGCTTGTATGAAG-3′ (forward) and 5′-TGCAAGGTGCTATAATCTCCTC-3′ (reverse), and those for the control gene GAPDH were 5′-ATCACTGCCACCCAGAAGAC-3′ (forward) and 5′-CAGTGAGCTTCCCGTTCAG-3′ (reverse).

Uninjected or injected embryos at the sphere stage were dechorionized, deyolked and resuspended in the TNE solution (4 μl per embryo), which contained 100 mM NaF, 1 mM orthovanadate (NaVO4), and a protease inhibitor cocktail (Roche Diagnostics GmbH, Mannheim, Germany). Total proteins were separated by 12% SDS-PAGE gels, immunoblotted with anti-phospho-Smad2 and anti-Smad2/3 antibodies (Cell Signaling Technology, Danvers, MA).