Arhgap28 Is a RhoGAP that Inactivates RhoA and Downregulates Stress Fibers

The small GTPase RhoA is a major regulator of actin reorganization during the formation of stress fibers; thus identifying molecules that regulate Rho activity is necessary for a complete understanding of the mechanisms that determine cell contractility. Here, we have identified Arhgap28 as a Rho GTPase activating protein (RhoGAP) that switches RhoA to its inactive form. We generated an Arhgap28-LacZ reporter mouse that revealed gene expression in soft tissues at E12.5, pre-bone structures of the limb at E15.5, and prominent expression restricted mostly to ribs and limb long bones at E18.5 days of development. Expression of recombinant Arhgap28-V5 in human osteosarcoma SaOS-2 cells caused a reduction in the basal level of RhoA activation and disruption of actin stress fibers. Extracellular matrix assembly studies using a 3-dimensional cell culture system showed that Arhgap28 was upregulated during Rho-dependent assembly of the ECM. Taken together, these observations led to the hypothesis that an Arhgap28 knockout mouse model would show a connective tissue phenotype, perhaps affecting bone. Arhgap28-null mice were viable and appeared normal, suggesting that there could be compensation from other RhoGAPs. Indeed, we showed that expression of Arhgap6 (a closely related RhoGAP) was upregulated in Arhgap28-null bone tissue. An upregulation in RhoA expression was also detected suggesting that Arhgap28 may be able to additionally regulate Rho signaling at a transcriptional level. Microarray analyses revealed that Col2a1, Col9a1, Matn3, and Comp that encode extracellular matrix proteins were downregulated in Arhgap28-null bone. Although mutations in these genes cause bone dysplasias no bone phenotype was detected in the Arhgap-28 null mice. Together, these data suggest that the regulation of Rho by RhoGAPs, including Arhgap28, during the assembly and development of mechanically strong tissues is complex and may involve multiple RhoGAPs.


Introduction
The actin cytoskeleton is fundamental to a wide range of cellular functions including cellular contractility, stiffness sensing, tissue formation, cell migration and cell polarity but the molecular mechanisms are complex and not fully understood. Members of the family of Rho guanosine triphosphatases (GTPases) are major regulators of the assembly of actin-based stress fibers along with mammalian diaphanous 1 (mDia) and Rho-associated kinase (ROCK) [1][2][3]. ROCK positively drives the assembly of contractile actin stress fibers by directly phosphorylating the myosin light chain (MLC), and also by inactivating MLC phosphatase [2,4]. Dynamic reorganization of the actin cytoskeleton into stress fibers is essential for fibronectin assembly and is regulated by signaling from Rho GTPases [5,6]. Actomyosin contractility is required for the translocation of fibronectin-bound integrins in specialized cell-matrix adhesions along actin stress fibers, a process that is believed to stretch folded fibronectin dimers to facilitate their assembly [7][8][9]. The mechanism for stress fibermediated ECM assembly is mechano-sensitive (via cell-matrix adhesions) and tightly regulated. For examples, disruption to actin polymerization or loss of tension caused the misalignment of collagen fibrils in a newly synthesized tendon ECM [10,11], and targeted ROCK overexpression in the epidermis led to increased collagen deposition and ECM stiffness [12]. While it is well known that Rho GTPases regulate actin stress fiber assembly, how they are regulated during tissue morphogenesis is less well understood.
Although Rho is a GTPase its rate of GTP hydrolysis slow. Efficient hydrolysis of GTP requires a Rho GTPase activating protein (RhoGAP), which accelerates the hydrolytic activity up to 105-fold [13]. There are over 70 genes encoding proteins that contain a RhoGAP domain [14]. This multitude of RhoGAPs is thought to ensure signaling specificity, for example, via tissuespecific expression, specificity for a single GTPase or signaling pathway, or that some RhoGAPs act as scaffold proteins or effectors for crosstalk between Rho GTPases and other signaling pathways (reviewed by [14]) . We show here that Arhgap28 is differentially regulated during mouse embryonic development. The functions of Arhgap28 have not been reported but its differential expression has been listed in a variety of cDNA microarray studies, as summarized in Table S1. Based on amino acid sequence similarities Arhgap28 is closely related to Arhgap6, Arhgap11a, Arhgap11b, Arhgap18, Arh-gap40, DLC1 (Arhgap7), DLC2 (Arhgap27) and DLC3 (Arh-gap38). Some of these RhoGAPs have been shown to regulate actin reorganization. Knockout of DLC1 is embryonic lethal at E10.5 of mouse development and examination of fibroblasts isolated from E9.5 mouse embryos revealed disrupted stress fibers and focal adhesions [15]. Mice lacking functional Arhgap6 protein are phenotypically normal, despite the fact that Arhgap6 is a RhoGAP for RhoA and causes the loss of actin stress fibers in cultured cells [16]. Arhgap18 also has specificity for RhoA and disrupts actin stress fibers, where knockdown of Arhgap18 can enhance stress fiber formation [17]. On the basis of these studies, we hypothesized that Arhgap28 regulates actin stress fiber assembly.

Sequence alignment predicts that Arhgap28 has a RhoGAP function similar to Arhgap6 and Arhgap18
RhoGAP function is mediated via the RhoGAP domain (represented schematically in Figure 1A), which enhances hydrolysis of GTP by the target Rho GTPase. Alignment of the RhoGAP domains of murine Arhgap6, Arhgap18 and Arhgap28 showed high homology and confirms the presence of a putative catalytic arginine residue (R425) in Arhgap28 ( Figure 1B), suggesting that Arhgap28 has RhoGAP function. No other conserved domains were identified.

Expression of Arhgap28-V5 inhibits RhoA activation and actin stress fiber assembly
To examine if Arhgap28 can regulate RhoA signaling and actin stress fiber formation, we created a V5-tagged Arhgap28 expression clone (Arhgap28-V5; see Figure S1 for details) and transfected SaOS-2 cells. SaOS-2 cells were selected because they form prominent stress fibers when cultured on plastic or glass. Expression of Arhgap28-V5 was confirmed in transiently transfected cells by western blotting using an anti-V5 antibody ( Figure 2A). Active RhoA was examined by a Rhotekin-GST pull-down assay to precipitate GTP-bound Rho. Expression of Arhgap28-V5 caused a reduction in active RhoA compared to cells treated with the transfection reagent only or when transfected with the empty vector (data not shown). In further experiments we performed quantitative ELISA assays for active RhoA, Rac1 and Cdc42 in the presence and absence of expressed Arhgap28-V5 and a mutant Arhgap28 R425A -V5 (R425A-V5) with control samples transfected with empty vector only. Note that the R425A-V5 construct lacks the putative catalytic arginine residue in the GAP domain (see Figure 1). We consistently observed reduction in the amounts of active RhoA (but not Rac1 and Cdc42) in the presence of Arhgap28-V5. The amounts of active RhoA remaining ranged from 62-78% of basal levels of active RhoA observed in controls. The results of the experiment that showed reduction to 78% is shown in Figure 2B; the expression of Arhgap28-V5 but not the R425A-V5 resulted in a significant reduction in basal levels of active RhoA with no significant effect on basal Rac1 and Cdc42 activity. The variability in the results between experiments was most probably because of differences in transfection efficiency between experiments. In separate experiments we prepared stably transfected SaOS-2 cells, expressing Arhgap28-V5. However, for reasons that were unclear to us, the expression levels were very low. Overall, the experiments showed that Arhgap28-V5 was effective at reducing the levels of RhoA.
In further experiments, actin stress fibers were examined by staining with phalloidin and immunofluorescence using the anti-V5 antibody; cells without staining for V5 exhibited prominent stress fibers ( Figure 2C). In all cells expressing Arhgap28-V5, as identified by staining with V5 antibodies, disrupted actin stress fibers were observed. Multiple actin microspikes and membrane ruffles were also observed on the edge of these cells ( Figure 2C, arrows). To test if these morphological changes were attributed to the GTP hydrolysis activity mediated by Arhgap28, cells were transfected with R425A-V5. All cells that stained positively for V5 contained prominent stress fibers ( Figure 2D). These expression studies indicated that Arhgap28 is a RhoGAP for RhoA and its expression negatively regulates stress fibers.
Arhgap28 has a restricted expression pattern through embryonic development We studied the spatial and temporal activation of Arhgap28 during embryonic development using an Arhgap28 gene trap (Arhgap28 gt ) mouse. The gene trap cassette does not disrupt the expression of the wild type Arhgap28 transcript and so the animals develop normally (see Figure S2 for details). In this mouse, the endogenous promoter of Arhgap28 drives the expression of bgalactosidase, which can be localized by X-gal staining. We examined the expression of b-galactosidase in E11.5 to E18.5 embryos. At all time points examined, b-galactosidase activity had a restricted spatial pattern ( Figure 3A). At E11.5 and E12.5, staining was localized to the dorsal region in what appears to contain somatic cells ( Figure 3A arrows in lower panels). At E13.5 to E15.5 the staining in the body of the embryos spread to the limbs and regions where the ribs are formed. At E18.5, staining could be clearly seen in limb bones and in the dorsal portions of the ribs. E18.5 embryos were sectioned and counter stained with Alizarin red. The results showed that cells staining positively for bgalactosidase activity were localized to the calcified portions of long bones and ribs ( Figure 3B). These data suggested to us that Arhgap28 has a role in regulating Rho in the initial stages (from E12.5) of ECM assembly throughout the embryo and becomes restricted to boney tissues at late stages of embryonic development, in the mouse.
Arhgap28 and related RhoGAPs are upregulated during Rho-mediated tissue stiffening in vitro To further explore the possibility that Arhgap28 is involved in the assembly of ECM, we examined the expression of Arhgap28 in a 3D cell culture model of tissue assembly in which embryonic fibroblasts deposit and tension a collagen fibril-rich ECM [11]. In this system, fibroblasts are moved from conventional 2D culture to medium containing fibrinogen and thrombin. The formation of a fibrin gel occurs within 5 minutes to produce a loose gel in which the fibroblasts find themselves suspended. During the next ,10 days in culture, the fibroblasts replace the fibrin with a collagen/ fibronectin-rich matrix [18] and subsequently tension this matrix using non-muscle myosin II-derived forces [19]. Quantitative PCR showed that the expression of Arhgap28 was low in cells cultured on plastic ( Figure 4A). The expression increased when the cells were in fibrin gels but was notably upregulated (15-fold) once the tissue constructs had tensioned (day 13; p,0.05). The expression levels continued to increase and were 35-fold higher (compared to 2D culture) after 14 days of 3D culture under tension (25 days in total; p,0.001; Figure 4A). Arhgap6 and Arhgap18 were also upregulated during tissue construct formation but not to the same extent as Arhgap28 ( Figure 4A). To test if Rho signaling is involved in 3D tissue construct contraction, fully formed tissue constructs (10 days of culture) were treated with lysophosphatidic acid (LPA) to activate Rho and the formation of stress fibers [20,21] or the ROCK inhibitor, Y27632. As shown in Figure 4B, actin polymerization was dramatically affected by these treatments; LPA induced thicker, more prominent actin stress fibers whereas Y27632 resulted in stress fiber shortening and disassembly.
The 3D tissue constructs recoil when unpinned, which is a process driven by non-muscle myosin II-dependent actomyosin contraction [22]. We made use of this feature to develop an assay of Rho-myosin II-dependent contraction of the tissue constructs. Thus, constructs were incubated with either LPA or Y27632 and the length of the unpinned construct was measured during 30 minutes. Tissue constructs treated with LPA significantly contracted to 53%60.02 of the original length in 30 minutes,  Figure 4C). Y27632 treatment significantly inhibited the contraction of the construct, which contracted to only 71.2%60.6 of the original length after 30 minutes (p,0.001; Figure 4C). The T 1/2 for recoil was: 444 s616 for DMSO-treated control tissue constructs, 323 s67 for LPA treated constructs, and 670 s63 for Y27632 treated constructs ( Figure 4D). In addition, expression of Arhgap28 and Arhgap6 was dramatically downregulated at 6 and 24 hours after tissue constructs were unpinned ( Figure 4E). Together, these data show that Arhgap28 is upregulated and Rho signaling is active during contraction of newly synthesized ECM, and the expression of Arhgap28 is inversely related to the stiffness of the mechanical environment.

Normal embryonic development of Arhgap28-null mice
To test further the possible role of Arhgap28 in tissue formation, we generated a functional Arhgap28-null mouse by crossing Arhgap28 gt mice with Cre transgenic mice. LoxP sites flank exons 7, 8 and 9 of the Arhgap28 gt gene. Therefore, Cre recombinase would be predicted to produce mice harboring an Arhgap28 del7-9 (Arhgap28 del ) allele ( Figure S3A). Using PCR readouts with genomic DNA we were able to identify wild type, heterozygous and null mice ( Figure S3B). The absence of exons 7-9 causes a frame shift and a smaller transcript that can be detected by RT-PCR using primers that span exons 6 to 11 ( Figure S3C). Thus, transcripts from the defective allele would produce a truncated protein lacking the RhoGAP domain. The results showed that heterozygous Arhgap28 del mice were viable and fertile. Litters produced by cross breeding heterozygotes had a close to normal Mendelian distribution of genotypes: 30%69 wild type; 49%65 heterozygotes; and 22%68 homozygotes (6SEM; from a total of 37 pups, in 8 litters). Arhgap28 del embryos (E15.5) were examined histologically and no apparent differences were observed compared to wild type mice ( Figure S3D). To confirm the loss of Arhgap28 in bone tissues, RNA was isolated from the tibia and fibula of P0 wild type and Arhgap28 del mice, and overlapping PCRs were performed ( Figure 5A). Fulllength Arhgap28 was detected in wild type bone tissues and overlapping PCRs confirmed that exons 7-9 were absent in the Arhgap28 transcript of Arhgap28 del bone tissues ( Figure 5B). The products of these reactions were sequenced to confirm that the transcript expressed in Arhgap28 del bone encoded a 367 amino acid-long protein lacking a functional RhoGAP domain ( Figure  S4).
Due to the lack of an obvious phenotype in heterozygous or null mice, we investigated the possibility that other RhoGAPs (e.g. Arhgap6 and Arhgap18) were compensating for the absence of Arhgap28. Quantitative PCR showed that the truncated Arh-gap28 del transcript was detectable at low levels in tissues from Arhgap28 del mice, which suggests that this truncated transcript (lacking the sequences encoding the RhoGAP domain) is at least partially stable. However, it was clear that the mutant transcript was at significantly lower levels compared to the expression of the Arhgap28 transcript in wild type mice (p,0.001; Figure 6A) presumably due to nonsense mediated mRNA decay. We also detected a significant 2-fold upregulation of Arhgap6 expression in Arhgap28 del tissues (p,0.001). Noteworthy, no significant differences were found in the expression of Arhgap18 in Arhgap28 del tissues ( Figure 6A). The data show that in tissues devoid of functional Arhgap28, there is compensatory upregulation of Arhgap6. We have previously shown that the two RhoGAPs were upregulated during ECM assembly and tensioning and were downregulated in response to lack of tension, which indicate that the Arhgap6-Arhgap28 pair might co-regulate the same Rho signaling pathway for actin reorganization. We also examined the expression of Rho GTPases. A small but significant upregulation in RhoA expression, but not in expression of Rac1, Cdc42 or RhoQ, was found in Arhgap28 del bone tissues compared to wild type (p,0.05; Figure 6B). Again, these data were indicative of a role of Arhgap28 in RhoA regulation.
Microarray comparison of gene expression between bone tissues of wild type and Arhgap28 del mice While Arhgap6 expression was upregulated in Arhgap28 del bone tissues there was also a significant increase in RhoA gene expression, which suggests that Rho signaling might still be altered in Arhgap28 del tissues despite Arhgap6 compensation. To determine if global changes in gene expression occurred, microarray analyses were performed in which gene expression from bone tissues (tibia and fibula) of wild type and Arhgap28 del mice were compared. The integrity of the RNA samples and the microarray readouts were analyzed (see Figure S5 for details). A total of 45037 probe sets were detected of which 363 of these showed significant differential expression of $62-fold difference (q,0.05). A heat map generated by hierarchical clustering of all the probe sets, based on similarities in the expression level and the expression profile is shown in Figure S5D.
As a cautionary note, three probe sets detected Arhgap28 del expression in bone from knockout animals (see Table 1). This was not unexpected because a 39 microarray was used and the mutant Arhgap28 del transcript contains the endogenous 39-end of the Arhgap28 transcript (see Figure 5). In contrast to the qPCR analyses performed previously, readouts for Arhgap6 did not show a fold change of $62 in Arhgap28 del bone ( Table 1). Expression of Arhgap18 and Rho GTPases (RhoA, Rac1, Cdc42 and RhoQ) also showed no differential expression in Arhgap28 del bone. The discrepancy between the microarray and the qPCR analyses could be because qPCR is a more quantitative and sensitive method of Arhgap28 Regulates RhoA and Stress Fiber Assembly PLOS ONE | www.plosone.org detecting gene expression. Interestingly, the actin genes detected by this microarray study, Acta1 and Actc1, were both significantly downregulated (,3-fold) in the bone tissue of Arhgap28 del mice (q,0.05; Table 1). The collagen genes expressed in bone tissues, Col1a1, Col1a2 and Col10a1, were not significantly different in Arhgap28-null bones.
DAVID online tool was used to perform gene ontology analyses on probe sets showing a differential expression $62-fold. The analysis showed that 57 probe sets detected genes that were downregulated and 306 probe sets detected genes that were upregulated, in the mutant samples. DAVID identified 10 functional annotation clusters in the genes that were downregulated. The top 3 annotation clusters and the genes that are overrepresented in these clusters are listed in Table 2. All three clusters contained genes indicative of a cartilage ECM, for example, Col2a1, Col9a1, Hapln1 (hyaluronan and proteoglycan link protein 1), Matn1 (matrillin 1) and Matn3, and some genes that are linked to negatively regulating bone mass, Agtr2 (angiotensin II receptor, type 2) and Dlk1, which encodes a transmembrane Arhgap28 is upregulated during ECM assembly and downregulated during Rho-dependent tissue recoil. Primary embryonic chick fibroblasts were seeded into a fibrin gel for the formation of 3D cell cultures containing aligned ECM. A. The expression of Arhgap6, Arhgap18, Arhgap28, Col1a1, and Fn were quantified by qPCR (n = 3). B. 3D tissue constructs were treated with 5 mM LPA or 10 mM Y27632 for 2 hours in serumfree conditions and changes to the actin cytoskeleton were observed by phalloidin staining. Bars = 50 mm. Arrows indicate the alignment of the tissue. C. The recoil baseline, which is the calculated length of the construct at the end of the exponential fit (see Materials and Methods for details). D. The half-time, which is the time required to achieve half of the recoil baseline. E. The expression of Arhgap6, Arhgap18 and Arhgap28 were quantified by qPCR during tissue recoil (n = 3). Fold changes in gene expression were normalized to Gapdh (2 2DDCt values). Bars show SEM. *** and * indicate significant differences found, p,0.001 and p,0.05, respectively, one way ANOVA. doi:10.1371/journal.pone.0107036.g004 Arhgap28 Regulates RhoA and Stress Fiber Assembly PLOS ONE | www.plosone.org protein called delta-like 1 homolog. These data suggest that a bone phenotype could still be expected upon closer analysis.
For genes that were upregulated in Arhgap28-null bone tissue, 65 annotation clusters were identified and the top 3 clusters are listed in Table 3. The genes that were over-represented in the first cluster were genes involved in targeting proteins for ubiquitination. For example, this cluster contained 13 E3 ubiquitin protein ligases, including March3 (membrane-associated ring finger (C3HC4) 3), March5, Mybp2 (MYC-binding protein 2); 3 peptidases; and a gene, Psmd14, which encodes a regulatory subunit of 26S proteasome (see Table 3 for full list of genes). The second cluster contained genes that promote actin polymerization, including Pafah1b1 (platelet-activating factor acetyl-hydrolase 1b), Tmod1 (tropomodulin 1) and Diap3 (also known as mDia2); actin nucleation (Spire1); and genes involved in linking the actin cytoskeleton to the plasma membrane including Utrn (utrophin) and Spna1 (spectrin). The third cluster of over-represented genes that were upregulated in Arhgap28 del bone tissue contained zinc finger proteins (Rbm5, Neil3 and Fus), which bind DNA or RNA. To summarize, the GO analyses showed that loss of Arhgap28mediated RhoA signaling causes: (i) down-regulation of cartilage ECM genes; (ii) up-regulation in genes involved in targeting proteins for degradation by ubiquitination; and (iii) up-regulation of genes involved in anchorage of the actin cytoskeleton to the plasma membrane. Whether or not Arhgap28 regulation of Rho signaling is involved functionally in cartilage homeostasis, protein degradation of anchorage of actin to the plasma membrane will require further investigation. However, the results presented in Figure 2 are indicative of a role for Arhgap28 in actin stress fiber polymerization at the plasma membrane. Figure 5. Comparison of the full-length transcripts of wild type and Arhgap28 del alleles. A. Overlapping products of PCR reactions 1, 4, 6-7, 9-14 by specific primers (numbers in gray) are sequenced. ATG and TAG indicates the start and stop codons, respectively. Diagram not drawn to scale. B. RNA was isolated from the tibia and fibular of P0 mice, and the 10 overlapping RT-PCRs were performed and visualized by gel electrophoresis. Products were subsequently purified and confirmed by DNA sequencing. Reactions in the green boxes will not produce a product with cDNA from Arhgap28 del (see Figure S4). The results of the GO analyses of Arhgap28-null bone tissues showing down-regulation of genes encoding cartilage ECM molecules and the fact that Arhgap28 expression was localized to pre-bone structures in embryos, prompted us to explore the possibility of a bone phenotype in the knockout mice, especially because mutations in Col2a1, Col9a1, Matn1 and Matn3 are associated with bone dysplasia (for reviews see [23][24][25]). Thus, we examined the skeletons of mice by X-ray at 10 weeks of age, when the skeletal growth has reached maturity. Skeletal analyses revealed no significant differences in the lengths of femur and tibia between wild type, heterozygotes and Arhgap28 del mice (Figure 7). As a measure of intramembranous ossification, we measured the intercanthal distance and found no significant differences between wild type, heterozygous and Arhgap28-null animals ( Figure 7). Together, these data suggest that although genes associated with bone dysplasia were downregulated in the Arhgap28 del mouse there is no bone length phenotype, presumably because of compensation from Arhgap6.

Discussion
The mechanical stiffness of musculoskeletal tissues is directly related to the organization of collagen fibrils in the ECM. For examples, the strongest tensile tissues such as tendon and ligament have collagen fibrils arranged in parallel register, which presumably is the best organization to resist uniaxial force; in bone the collagen fibrils provide a template for mineralization. Furthermore, the collagen fibrils are pre-stressed by cells to ensure that tissues can respond directly to applied forces. It is poorly understood how the tissue-specific arrangement and pre-stressing of collagen fibrils is achieved. Tissue stiffening requires cellular contraction via actin stress fibers, which are regulated by Rho GTPases. How these Rho GTPases are regulated during tissue development to ensure that the ECM is at optimal stiffness is unknown.
Here, we show that Arhgap28 is activated in bone tissues before birth and during the assembly of a stiff ECM. Experiments expressing Arhgap28-V5 suggest that Arhgap28 is a negative regulator of RhoA and actin stress fiber formation. Arhgap28deficiency does not appear to affect bone development, which is most likely due to functional redundancy between Arhgap28 and a closely related RhoGAP, Arhgap6. It will be important in future work to investigate the mechanisms of how Arhgap28-regulated actin contractility determines stiffening of the ECM and to understand how RhoGAPs crosstalk regulates Rho and actin remodeling within developing musculoskeletal tissues. Cellular tension is generated by actin stress fibers [26]. In this context, the closely related Arhgap6 and Arhgap18 regulate the formation of actin stress fibers via RhoA [16,17] and in this study, expression of Arhgap28-V5 caused similar effects. Sustained RhoA activity has inhibitory effects on Rac1-and Cdc42-activated lamellipodia and filopodia formation [27,28], which helps explain the appearance of actin microspikes and membrane ruffles in Arhgap28-overexpressing cells although there was no detectable activity of Arhgap28 against Rac1 and CDC42 in the assays used here.
Cells respond to stiff extracellular matrices via Rho-activated actin stress fibers [29][30][31] and overactive RhoA signaling is linked to cancer [see 32 for review]. Here, loss of RhoGAPs, such as DLC1, is associated with cancer [33], which suggest that RhoA signaling affects cell fate. Overexpression of RhoA and Rho GTPases are linked to cancer and the stability of RhoA transcripts in cancer cells has been shown to be a result of altered polyadenylation signals [34], further suggesting why it would be biologically important to have more than one RhoGAP regulating Rho GTPases.
There are a few studies that show that the differentiation of mesenchymal stem cells into an osteogenic lineage can be influenced by Rho/ROCK signaling, for examples see [35][36][37]. Surprisingly, mice expressing a dominant-negative RhoA have a bone sclerotic phenotype, which suggests that lack of RhoA signaling enhances mineralization [38]. These disagreements linking Rho/ROCK and bone development could be due to the need for a balanced signal and crosstalk between Rho/ROCK and growth factors rather an all-or-nothing response. RhoA expression was also significantly upregulated in bone tissues of Arhgap28-null mice. RhoA expression is activated by transcription factor Myc [39], but Myc was not differentially regulated in the microarray comparison between wild type and Arhgap28 del bone tissues. Due to the large family of mammalian RhoGAPs, it was predicted that Arhgap28-deficiency might be compensated for by functional redundancy between Arhgap28 and other RhoGAPs. Indeed, absence of Arhgap6 in mice does not cause an overt phenotype, presumably because of the observed compensatory mechanisms [16]. It was surprising in our studies that Arhgap18 was not also upregulated because both Arhgap6 and Arhgap18 have been shown to negatively regulate RhoA and actin stress fibers. These results reveal the potential of a novel co-regulatory mechanism for RhoA signaling and actin stress fibers by Arhgap6 and Arhgap28. How these RhoGAPs are activated during the patterning of ECM is unknown. Matrix assembly and detection of the ECM occurs at cell-matrix adhesion sites, therefore it is hypothesized that Arhgap28 is activated by signals downstream of cell-matrix adhesions in similar mechanisms described for other RhoGAPs. For examples, upon activation of integrin b1, p190-RhoGAP is activated via tyrosine phosphorylation at the Nterminus by Src [40], and more recently, the identification of protein-protein interaction domains within DLC1 suggest direction interaction between DLC1 and cell-matrix adhesion proteins tensin, talin and focal adhesion kinase [41,42].
Importantly, although there was no overt phenotype in the Arhgap28 del mice (perhaps due to compensation by Arhgap6) the microarray study revealed downregulation in Acta1 and Actc1 genes, which encode isoforms of actin found most abundantly in muscle. Downregulation in both these genes has also been observed in a microarray study of chondrocyte differentiation in vitro [43] and downregulation of Acta1 was observed in bone in response to mechanical loading [44], which suggest of change in the cells of Arhgap28 del bone tissues. Gene ontology analyses also revealed that genes that encode ECM molecules associated with a cartilage tissue (including Col2a1, Col9a1, Matn3 and Comp) were downregulated in Arhgap28-null bone tissues. Mutations in these genes are associated with bone dysplasia [for reviews, see 23,24,25]. Bone tissues from Arhgap28 del mice also showed an upregulation of genes involved in ubiquitination and actin  45]. The target for many of these E3 ligases is unknown. The genes that stood out include Hace1, which catalyzes the ubiquitination of Rac1 [46], Mycbp2, which can bind Myc [47] but whether or not it is inhibitory or enhancing for RhoA transcription is unknown. The differential regulation of these genes is indicative of a response caused by Arhgap28-deficiency, which might be independent of RhoA signaling if Arhgap6 compensates for Arhgap28.
In conclusion, we describe experiments that point to Arhgap28 being a functional negative regulator of RhoA during the formation of actin stress fibers in cells of mesenchymal origin. However, knockout experiments in mice indicated that Arhgap28 is not autonomous but is functionally redundant in the presence of other RhoGAPs including Arhgap6 and possibly Arhgap18. Such redundancy presumably underpins the cell's ability to generate actin stress fibers during tissue formation.

Ethics
The care and use of all mice in this study was carried out in accordance with UK Home Office regulations, UK Animals (Scientific Procedures) Act of 1986 under the UK Home Office licence (PPL 40/3485). No experimental procedures were performed on live animals. All animals were sacrificed by Schedule 1 cervical dislocation by trained personnel, and all efforts were made to minimize suffering.

V5-His-tagged murine Arhgap28 and Arhgap28 R425A expression clones
For the generation of a C-terminus V5-His-fusion Arhgap28 clone, the sequence encoding V5-His (GKPIPNPLLGLDST-(His)6) was introduced immediately 59 of the stop codon of the endogenous, larger Arhgap28 isoform cDNA. The cDNA clone of the short Mus musculus Arhgap28 isoform (BC066788.1; Source BioScience Geneservice) was digested with EcoRI and BsmBI (New England Biolabs, Hertfordshire, UK) following the manufacturer's instructions to produce a 1123 bp fragment of the 59 portion of the Arhgap28 clone. For the preparation of the 39 portion of endogenous Arhgap28 transcript, a 1209 bp fragment was amplified from mouse fibroblast cDNA (made with random hexamers) using a forward primer 59 of a unique endogenous BsmBI restriction recognition site 39-AAG ATT TGG GTT GAC CGA GAC G-59 and a reverse primer that created a new BstBI site 39-AAT TCG AAG GGC TTG ATG ACC C-59 immediately before the stop codon. iProof High-Fidelity DNA Polymerase (Bio Rad Laboratories) was used following the manufacturer's protocol. The PCR fragment was sequenced and cut with BsmBI and BstBI  Arhgap28 Regulates RhoA and Stress Fiber Assembly PLOS ONE | www.plosone.org restriction enzymes (New England Biolabs). pcDNA6/V5-His C vector was prepared by digestion with EcoRI and BstBI restriction enzymes (New England Biolabs). The three cut fragments were purified using QIAquick Gel Extraction kit (Qiagen) and ligated using T4 DNA Ligase (New England Biolabs). Arhgap28 R425A clone containing a codon change from CGA to GCA (Genscript) was subcloned into pcDNA6/V5-His C between EcoRI and BstBI sites. DNA was purified from each culture using QIAprep Spin Miniprep kit (Qiagen) following the manufacturer's protocol.

3D tissue constructs and recoil assays
Primary embryonic chick fibroblasts were isolated from E14 chick metatarsal tendons. For the formation of 3D tissue constructs in vitro, cells were seeded into fibrin gels as described previously [11]. To modulate Rho signaling in vitro tissue constructs were washed with phosphate buffered saline (PBS) three times and then equilibrated for 30 minutes with serum-free medium and then treated for 30 minutes with final concentration of 5 mM lysophosphatidic acid (LPA; Sigma) in 0.1% (v/v) DMSO or 10 mM Y27632 (ROCK inhibitor; Sigma) in 0.1% (v/v) DMSO, diluted in serum-free medium, all at 37uC, 5% CO 2 in a humidified environment. For the recoil assay, the tissue constructs were treated for 2 hours and was then cut from one suture to allow contraction to occur. The constructs were imaged using a digital single lens reflex camera at a fixed focal point for 30 minutes at 10second intervals. The length of the constructs in each image was measured using ImageJ software. The length was converted to a percentage of the original and the means for each experimental group were calculated. The mean values were then fitted to a 3parameter exponential decay function using SigmaPlot (Systat Software Inc.). One-way ANOVA and a Dunnett's test were used to determine significance differences between the derived T 1/2 and recoil baseline values compared to control constructs incubated in DMSO.

RNA and PCRs
RNA was isolated with TRIzol reagent (Invitrogen) and treated with DNase (Promega) following manufacturers' protocols. cDNA was synthesized using TaqMan Reverse Transcriptase reagents (Applied Biosystems) and used for analyses by PCR and qPCR. A list of primers used can be found in Table S2. All PCR products produced from primers were validated and confirmed by DNA sequencing using the BLAST software against the NCBI nucleotide database and aligned with the expected sequence. For qPCR analyses, the 2 -DDCt method [48] was used to analyze relative fold changes in gene expression compared to the experimental control group. A two-sample t-test or one-way ANOVA followed by a Dunnett's test were performed to determine significance differences compared to the control sample.

Immunofluorescence
Cells fixed with freshly prepared 4% (w/v) paraformaldehyde in PBS for 20 minutes at RT, then washed with PBS containing 0.1% (v/v) Tween 20 three times and blocked with 1% (w/v) bovine serum albumin, 0.1% (v/v) Triton X-100 in PBS containing 0.1% Tween 20 for 1 hour. Cells were incubated with primary antibodies, mouse monoclonal anti-V5 epitope or the appropriate control IgGs, diluted in the blocking buffer overnight at 4uC. Cells were washed with PBS containing 0.1% (v/v) Tween 20 for 10 minutes three times and incubated with Alexa Fluor 594conjugated antibodies (Invitrogen) and Atto 488-conjugated phalloidin (Sigma) diluted with the blocking buffer for 1 hour at room temperature, protected from light. Stained cells were washed with PBS containing 0.1% (v/v) Tween 20 for 10 minutes three times and mounted using Vector Shield containing DAPI (Vector Laboratories). Fluorescent images were taken using a digital camera attached to an Olympus BX51 and captured using MetaVue imaging software (Molecular Devices).

Reporter and knockout mice
Using homologous recombination in agouti C57BL/6N embryonic stem (ES) cell, the L1L2_Bact_P targeting gene trap cassette (including the genes encoding b-galactosidase and neomycin) was introduced into intron 6 of the Arhgap28 gene by the NIH Knockout Mouse Project (KOMP; CA, USA). In addition, exons 7-9 are flanked by loxP sites. ES cells containing the Arhgap28 gene trap allele (Arhgap28 gt ) were injected into C57BL/6J blastocysts to produce germ line-transmitting chimeras (prepared at the University of Oulu, Finland). Chimeric males were assessed on coat color and mated with wild type C57BL/6J females to produce heterozygous progeny, which were then mated to obtain Arhgap28 gt mice. For the generation of Arhgap28 deleted exons 7-9 (Arhgap28 del ) mutants, male Arhgap28 gt mice were mated with females from a deleter Cre transgenic mice (a gift from M. Briggs, University of Newcastle, UK) to ablate the loxP-flanked exons 7-9 in all tissues. F1 offspring heterozygous for the knockout allele were then mated to generate knockout (Arhgap28 del ) mice.

Genotyping
Mice were genotyped using genomic DNA extracted from ear punches of adult mice or amniotic sacs of embryos using 200 mg/ ml Proteinase K (Invitrogen) in buffer containing 17.6 mM Nlauroyl sarcosine, 100 mM NaCl and 5% (w/v) Chelex 100 resin (Bio Rad Laboratories). Genotypes were determined using specific primer pairs (see Table S2) using BioMix Red PCR reagents (Bioline). The annealing temperature for the wild type/mutant allele PCR was 53uC. The annealing temperature for the Arhgap28 del allele was touch-down from 70-60uC for 10 cycles followed by 20 cycles at 60uC. The annealing temperature for the Cre transgene was 50uC.

Whole mount X-gal stain
For detection of beta-galactosidase expression, E10.5-E15.5 embryos were fixed for 1 hour in 3.7% (w/v) PFA in PBS pH 8 at room temperature. For older embryos the skin was removed before being fixed for 1 hour in 3.7% (w/v) PFA in PBS pH 7.4 at room temperature. Embryos were then washed with PBS containing 0.1% (v/v) Triton X-100 for 15 minutes twice. Each embryos was then stained with 20 ml of freshly prepared X-gal staining solution (1 mM X-gal (Qiagen), 5 mM potassium ferricyanide (K 3 Fe(CN) 6 ), 5 mM potassium ferrocyanide (K 4 Fe(CN) 6 ), 1 mM MgCl 2 in PBS containing 0.1% (v/v) Triton X-100) for 24 hours at 37uC. After staining, the embryos were rinsed with PBS and post-fixed in 4% (w/v) PFA in PBS overnight at 4uC. After post-fix, the embryos were either processed for paraffin embedding or dehydrated in 70% (v/v) ethanol for 6 hours and cleared with glycerol (10 ml of each 30, 50 and 80% (v/v) glycerol in 1% (w/v) potassium hydroxide and then 100% glycerol), incubating at 37uC for 2-3 days each.

Histology
Sagittal sections (6 mm thick) were cut from wax embedded embryos and mounted for haematoxylin and eosin staining. Slides were stained using an automated stainer and cleared into Histo-Clear (Thermo Fisher Scientific). For Alizarin red staining, sections were then washed with distilled water and stained with 2% (w/v) Alizarin red pH 4.2 for 10 minutes, washed with distilled water and dehydrated. Images were captured using a Carl Zeiss Axiocam Colour CCD camera with associated AxioVision software.
Arhgap28 Regulates RhoA and Stress Fiber Assembly

Comparative gene expression microarrays
For the comparison of bone tissues between wild type and Arhgap28 del mice tibia and fibula were dissected from three P0 neonatal litters each from defined breedings. Bones were incubated in 1000 U/ml bacterial collagenase type 4 (Worthington Biochemical Corporation) in 0.25% (w/v) trypsin (Invitrogen) for 25 minutes at 37uC with agitation every 10 minutes. The bones were then removed of any excess muscle and cartilage tissues and washed in PBS. RNA was isolated using a dismembrator as described previously [49]. Integrity and measurement of total RNA was performed using Agilent 2100 Bioanalyzer (Agilent Technologies). RNA was amplified by two-cycle cDNA synthesis, and then labelled cRNA was synthesized and hybridized to Mouse Genome 430 2.0 GeneChip arrays (Affymetrix). Microarray data are available in the ArrayExpress database (www.ebi.ac.uk/ arrayexpress) under accession number E-MTAB-2296. Microarray data sets were analyzed by dChip (DNA-Chip) Analyzer to normalize the array readouts [50]. Normalized readouts were analyzed using the Robust Multichip Average method as described by [51]. Principal component analysis (PCA) was employed to confirm that different variables were present as a quality control for the arrays. p values for each probe set were generated by Limma t-test and q values were subsequently generated by applying false discovery rate correction. Gene ontology analysis was performed on probe sets that have detected fold changes greater than 2 using Database for Annotation, Visualization and Integrated Discovery (DAVID) online tool [52].

Bone length measurements
X-rays of mice were produced using a Flaxitron x-ray specimen radiograph system (Flaxitron Bioptics) and x-ray film (GE Healthcare). Bone measurements were taken from scanned radiographic images using ImageJ software. One-way ANOVA followed by a Dunnett's test were used to determine significance differences compared to wild type. Figure S1 Cloning of a V5-tagged Arhgap28 overexpression construct. A. Difference in C-terminal ends of the two major Arhgap28 isoforms from amino acid residue 679. PCR strategy to determine the sequence of endogenous Arhgap28 transcript expressed in primary mouse fibroblasts. Overlapping products of PCR reactions 1 to 12 by specific primers (numbers in grey) were sequenced. Reactions in the boxes are unique to either the larger or the smaller isoform. ATG and TAG indicates the start and stop codons, respectively. Diagram not drawn to scale. B. RT-PCR products with RNA was isolated from primary mouse fibroblasts. A cDNA clone of the smaller Arhgap28 isoform was used as control template. C. Arhgap28 was cloned into a pcDNA6 vector where a V5-His6 tag was introduced into the Arhgap28 sequence immediately 59 of the stop codon. D. RNA was isolated from NIH3T3 fibroblasts transiently transfected with Lipofectamine only (L), empty vector (E) or Arhgap28-V5 clones (#6 or #8). RT-PCRs were performed to detect the expression of housekeeping gene, GAPDH (221 bp) or Arhgap28-V5 using specific primers (329 bp). No RT controls confirmed the absence of plasmid DNA contamination. E. Protein was isolated for western blotting to detect V5-tagged protein expression. A positive signal was detected at ,90 kDa. The predicted molecular weight of Arhgap28-V5 is 85 kDa. Control lysates were included -Talin-V5 (60 kDa), mock transfection (2) and myristoylated-FAK-V5 (+; 175 kDa). (TIFF) Figure S2 Arhgap28 gt mice are normal and express Arhgap28 due to unsuccessful gene trapping. A. Schematic showing the genotyping strategy for identifying the presence of the gene trap cassette targeted to the Arhgap28 gene. B. DNA was isolated from wild type, Arhgap28 +/gt and Arhgap28 gt/gt mutant neonatal tail tendons cells to confirm genotypes. DNA from wild type (+/+) animals will only produce a 493 bp product whereas DNA from homozygous (gt/gt) animals will only produce a 354 bp product and DNA from heterozygous (+/gt) animals will produce both bands. C. RNA was also isolated and RT-PCR was performed to detect the expression of Arhgap28 transcript spanning from exon 6 to 11, Col1a1 and Gapdh. D. Sagittal sections of wild type (+/+), heterozygous (+/gt) and homozygous Arhgap28 gt mutant (gt/gt) embryos at gestation day E15.5 stained with H&E and close-up of the limbs. (TIFF) Figure S3 Arhgap28 del mice express an Arhgap28 transcript lacking exons 7 to 9. A. Schematic showing the genotyping strategy for identifying the presence of the gene trap cassette targeted to the Arhgap28 gene and for the detection of Arhgap28 del7-9 KO allele. B. Representative gel image of genotyping PCR products. In the first genotyping PCR which distinguishes between wild type (493 bp) or mutant Arhgap28 allele (either the Arhgap28 gt or Arhgap28 del allele; 354 bp). The second genotyping PCR tests for the presence of the Arhgap28 del allele, the product of the mutant Arhgap28 allele after Cre recombinase-mediated DNA excision (400 bp). The third genotyping PCR tests for the presence of the deleter Cre transgene (350 bp). C. RNA was isolated from wild type, Arhgap28 +/del and Arhgap28 del/del pups and RT-PCR was used to detect expression of wild type Arhgap28 (634 bp) and Arhgap28 del (338 bp) transcripts spanning from exons 6 to 11. RT-PCRs for Col1a1 and Gapdh was used as loading controls. D. Sagittal sections of wild type and homozygous Arhgap28 del embryos at gestation day E15.5 stained with H&E. (TIFF) Figure S4 Translated sequences of Arhgap28 transcripts. Amino acid sequences transcripts of Arhgap28 from wild type, predicted Arhgap28 del , actual Arhgap28 del bone tissue. Sequence in blue is a result of remainder sequence from Cremediated recombination. Sequence in green is the RhoGAP domain. (TIFF) Figure S5 Quality control of microarray comparing wild type and Arhgap28 del bone tissues. A. Agarose gel showing total RNA isolated from triplicate samples of P0 tibia and fibula tissues from wild type and Arhgap28 del/del mice. B. PCA mapping of variability between the array samples. C. dChip analysis of microarray data sets. (1) Triplicate samples of RNA from tibia and fibula of P0 Arhgap28 +/+ (wt) and Arhgap28 del/del (del) mice indicated by the number at the end of each array name.