Atrophin Proteins Interact with the Fat1 Cadherin and Regulate Migration and Orientation in Vascular Smooth Muscle Cells*

Fat1, an atypical cadherin induced robustly after arterial injury, has significant effects on mammalian vascular smooth muscle cell (VSMC) growth and migration. The related Drosophila protein Fat interacts genetically and physically with Atrophin, a protein essential for development and control of cell polarity. We hypothesized that interactions between Fat1 and mammalian Atrophin (Atr) proteins might contribute to Fat1 effects on VSMCs. Like Fat1, mammalian Atr expression increased after arterial injury and in VSMCs stimulated with growth and chemotactic factors including angiotensin II, basic fibroblast growth factor, and platelet-derived growth factor BB. Two distinct Atr2 transcripts, atr2L and atr2S, were identified by Northern analysis; in VSMCs, atr2S mRNA expression was more responsive to stimuli. By immunocytochemistry, Fat1 and Atrs colocalized at cell-cell junctions, in the perinuclear area, and in the nucleus. In coimmunoprecipitation studies, Fat1 interacted with both Atr1 and Atr2; these interactions required Fat1 amino acids 4300–4400 and an intact Atro-box in the Atrs. Knock-down of Atrs by small interfering RNA did not affect VSMC growth but had complex effects on migration, which was impaired by Atr1 knockdown, enhanced by Atr2L knockdown, and unchanged when both Atr2S and Atr2L were depleted. Enhanced migration caused by Atr2L knockdown required Fat1 expression. Similarly, orientation of cells after monolayer denudation was impaired in cells with Atr1 knockdown but enhanced in cells selectively depleted of Atr2L. Together these findings suggest that Fat1 and Atrs act in concert after vascular injury but show further that distinct Atr isoforms have disparate effects on VSMC directional migration.

Fat1, an atypical cadherin induced robustly after arterial injury, has significant effects on mammalian vascular smooth muscle cell (VSMC) growth and migration. The related Drosophila protein Fat interacts genetically and physically with Atrophin, a protein essential for development and control of cell polarity. We hypothesized that interactions between Fat1 and mammalian Atrophin (Atr) proteins might contribute to Fat1 effects on VSMCs. Like Fat1, mammalian Atr expression increased after arterial injury and in VSMCs stimulated with growth and chemotactic factors including angiotensin II, basic fibroblast growth factor, and platelet-derived growth factor BB. Two distinct Atr2 transcripts, atr2L and atr2S, were identified by Northern analysis; in VSMCs, atr2S mRNA expression was more responsive to stimuli. By immunocytochemistry, Fat1 and Atrs colocalized at cell-cell junctions, in the perinuclear area, and in the nucleus. In coimmunoprecipitation studies, Fat1 interacted with both Atr1 and Atr2; these interactions required Fat1 amino acids 4300 -4400 and an intact Atro-box in the Atrs. Knockdown of Atrs by small interfering RNA did not affect VSMC growth but had complex effects on migration, which was impaired by Atr1 knockdown, enhanced by Atr2L knockdown, and unchanged when both Atr2S and Atr2L were depleted. Enhanced migration caused by Atr2L knockdown required Fat1 expression. Similarly, orientation of cells after monolayer denudation was impaired in cells with Atr1 knockdown but enhanced in cells selectively depleted of Atr2L. Together these findings suggest that Fat1 and Atrs act in concert after vascular injury but show further that distinct Atr isoforms have disparate effects on VSMC directional migration.
Migration and proliferation of VSMCs 2 in the wall of injured blood vessels are critical activities in the pathogenesis of atherosclerosis and related, clinically important vascular diseases.
Arterial injury strongly induces expression of the Fat1 cadherin, which has distinct effects on VSMC migration and proliferation (1). Fat1 and related, atypical cadherins form a subfamily characterized by large extracellular domains containing 34 cadherin motifs, a variable number of EGF repeats, one or two laminin A/G domains, and a single transmembrane domain (2). In vertebrates, the Fat subfamily consists of four members, Fat1, -2, -3, and -4 (or Fat-J), whereas in Drosophila, two members, Fat and Fat-like, have been identified (2,3). Although the Drosophila fat (ft) mutation, which causes enlargement of all larval imaginal discs, including wing, leg, eye-antenna, haltere, and genital imaginal discs, was first described by Mohr 85 years ago (4), understanding of how Fat proteins control developmental processes has only recently started to emerge.
Distinct functions have been ascribed to Drosophila Fat and Fat-like. The former acts as a suppressor of hyperplastic growth, as mentioned above, and as a mediator of planar cell polarity signals in development (4,5). Fat has been identified in several recent studies as a regulator of the Hippo signaling pathway, which controls organ size during development through effects on both cell proliferation and survival (6 -10). Fat-like, on the other hand, performs a crucial morphogenetic role in the formation of tubular organs such as the trachea, possibly by acting as an epithelial spacer (11). In contrast to the effect of Fat mutations, impairment of Fat-like expression does not affect imaginal disc development or planar cell polarity (11), but whether or not Fat and Fat-like have redundant or overlapping functions in other settings has not been fully explored.
Information about the function of vertebrate Fat proteins is relatively limited. Mice with homozygous inactivation of the fat1 locus die perinatally with loss of the renal glomerular slit junctions, fusion of glomerular epithelial cell processes, and defects in forebrain and eye development; growth perturbations were not detected during embryonic skin development or in neurospheres derived from fat1 Ϫ/Ϫ mice (12). In cultured cells, Fat1 interacts with Ena/VASP proteins, localizes to filopodial tips and lamellipodia at cellular leading edges, organizes cytoskeletal actin, and promotes migration (13,14). We identified increased Fat1 expression in VSMCs responding to arterial injury and found that knockdown of Fat1 in this cell type limited migration but enhanced proliferation (1); taken together with the findings in fat1 Ϫ/Ϫ mice (12), these results point to cell type-or developmental stage-dependent differences in the ability of Fat1 to regulate growth. Additional evidence for Fat1 function as a tumor suppressor includes the recent report of very frequent homozygous deletion or gene silencing of the fat1 locus in oral squamous cell carcinomas (15). As for the other vertebrate Fat proteins, recent reports suggest that Fat2 supports the migration of squamous carcinoma cells (16), whereas Fat4 controls the orientation of cell divisions and tubule elongation during kidney development in the mouse (17).
Genetic analysis in Drosophila demonstrates that Atrophin functions as a transcriptional corepressor during development, with important roles in diverse processes including segmentation and planar polarity (18). A link between Fat cadherins and Atrophin was first identified in a yeast two-hybrid screen in Drosophila using a fragment of the Fat cytoplasmic domain (5). Comparison of ft and atrophin mutants in Drosophila showed similar defects in planar polarity, and double mutants showed strongly enhanced effects on viability, indicating genetic interaction (5). Whether or not fat-like (ftl) and atrophin interact has not been reported.
Whereas the Drosophila genome encodes a single Atrophin, vertebrate genomes harbor two loci that give rise to Atr1 and two forms of Atr2: a long form (Atr2L, also known as RERE) and a short form (Atr2S); Drosophila Atrophin is most like Atr2L in terms of overall structure, suggesting that this long form of the protein may reflect the ancestral gene (19). Interestingly, the Atr isoforms with extended N-terminal domains (Drosophila Atrophin and vertebrate Atr2L) can interact with histone deacetylases (20 -22), indicating at least one mechanism by which these proteins can act as transcriptional repressors. The importance of the N-terminal sequences is supported in vivo by the observation that, although Atr1-null mice are viable and fertile (19), Atr2L mutant mice die around embryonic day 9.5 with defects in heart looping, telencephalon, and somite development, as well as loss of Sonic hedgehog (Shh) and fibroblast growth factor (Fgf) 8 expression from anterior signaling centers (20).
In view of our previous findings that Fat1 regulates VSMC growth and migration, we asked whether mammalian Atrs and their distinct isoforms might contribute to these Fat1-mediated effects. Our studies show that Atrs, like Fat1, are induced after arterial injury, that Atrs and Fat1 interact physically, and that, again like Fat1, they regulate migration and orientation. Interestingly, however, the different Atr isoforms have distinct effects on these cellular activities; the short Atrs, Atr1 and Atr2S, promote migration and orientation, whereas the long Atr isoform, Atr2L, inhibits these activities. Taken together, these results suggest a new framework, potentially extending from the cell surface to the nucleus, for regulation of VSMC chemotaxis in the post-injury setting.

EXPERIMENTAL PROCEDURES
Rat Carotid Artery Balloon Injury-The rat carotid artery balloon injury model was implemented as described previously (1). All of the procedures were approved by and in accordance with guidelines established by the Albert Einstein Institute for Animal Studies.
Cell Culture and Transfection-Primary culture VSMCs were prepared from rat or mouse aortae and maintained as described previously (1). Human aortic VSMCs were obtained from Clonetics and cultured in Dulbecco's modified Eagle's medium supplemented with 20% fetal bovine serum. VSMC phenotype was validated by immunocytochemistry using an antibody specific for ␣ smooth muscle actin (1:400, Clone 1A4; NeoMarkers), and cells were used between four and eight passages from harvest. 293T and Chinese hamster ovary cell lines (American Tissue Type Collection) were cultured in the medium according to the supplier's recommendations. Angiotensin II, dexamethasone, and transforming growth factor-␤ were obtained from Sigma, and basic fibroblast growth factor, platelet-derived growth factor BB, interferon-␥, and interleukin-1␤ were from U.S. Biological. In stimulation experiments, the cells were rendered quiescent by incubation in medium containing 0.4% horse serum for 72 h prior to the addition of fetal bovine serum or other stimulus. Control cultures received an equivalent amount of vehicle. Total cellular protein or total RNA was extracted at designated time points. In transfection experiments, FuGENE 6 (Roche Applied Science) or TransIT-LT1 reagent (Mirus) were used according to the manufacturers' protocols.
Northern Analysis-Total RNA (10 g) was fractionated on 1.2% formaldehyde-agarose gels and transferred to nitrocellulose filters. A rat Atr2 cDNA probe, corresponding to nucleotides 4495-4922 in GenBank TM entry NM_053885, was generated by PCR, random primed in the presence of 32 P-labeled dCTP, and hybridized to the filters in QuikHyb (Stratagene). The filters were washed to high stringency at 55°C in buffer containing 30 mmol/liter sodium chloride, 3 mmol/liter sodium citrate, and 0.1% SDS. Autoradiography was performed with Kodak XAR film at Ϫ80°C for 48 h.
5Ј-Rapid Amplification of cDNA Ends-To define the 5Ј end of the mouse atr2S transcript, we performed 5Ј-rapid amplification of cDNA ends using Ambion mouse embryo rapid amplification of cDNA ends-ready cDNA according to the manufacturer's instructions. The atr2 gene-specific primers were: 5Ј-TGGGAGACGTGCTGCGATTA-3Ј (outer primer) and 5Ј-ATGCAGCCTCTTCCTTCACCTT-3Ј (inner primer). Amplified cDNA products were cloned and sequenced by standard methodology. This sequence will be supplied as supplemental data.
cDNA Constructs-The DelN-Fat1 construct (13) encodes an N-terminal truncated form of human Fat1 (amino acid residues 3987-4590) in which all the cadherin domains, one EGF motif, and the laminin A-G domain have been removed, and part of the extracellular and all of the transmembrane and intracellular domains (IC) are retained. We generated a similar construct by RT-PCR. A 3XFLAG sequence was integrated in frame into the N-terminal of the cDNA, and the resulting product was subcloned into the p3XFLAG-CMV-13 expression vector (Sigma), yielding an N-terminally FLAG-tagged product defined at the C terminus by the native Fat1 stop codon. Derivative C-terminal truncation constructs, designated DelN-Fat1 4500, DelN-Fat1 4400, and DelN-Fat1 4300, were generated by introducing an additional stop codon at the corresponding amino acid residue (residue 4500, 4400, or 4300) by QuikChange site-directed mutagenesis (Stratagene). The full-length cDNAs encoding human Atr1 and Atr2L (long form, amino acids 1-1506) were obtained from Open Biosystems and subcloned into pCMV⅐SPORT6 (Invitrogen) and pcDNA-DEST40 (Invitrogen), respectively. These cDNA inserts were also cloned in frame into the pCS2-6XMyc vector to provide C-terminal Myc epitope tags and into the pEGFP-C2 vector (Clontech) to provide N-terminal EGFP epitope tags. The Atr2S (short form, amino acids 495-1506) cDNA was obtained by RT-PCR and subcloned similarly into the pCS2-6XMyc and pEGFP-C2 vector. All of the constructs were confirmed by sequencing.
Coimmunoprecipitation-Specific proteins were immunoprecipitated by incubating 300 -500 g of precleared whole cell lysates in immunoprecipitation buffer with 2-5 g of the corresponding antibodies or normal IgG control at 4°C for 2 h, followed by incubation with protein G agarose (Invitrogen) at 4°C overnight with gentle agitation. After extensive washing, the immune complexes were recovered by boiling in sample buffer, and the proteins were detected by Western analysis.
Lentivirus Preparation and Transduction-Lentivirus was produced using the Virapower Lentiviral kit (Invitrogen) according to the manufacturer's directions. Briefly, cDNA encoding full-length of human Atr1 was cloned into the pLenti6-V5-DEST vector, and virus was generated in the 293FT viral packaging cell line. Equal titers of test or vector control virus were used in subsequent experiments. Mouse VSMCs were infected with virus-containing supernatant in the presence of polybrene, and stably transduced cells were selected with Blasticidin. A total of 16 clones were isolated, and Atr1 expression levels were detected by Western blot.
Cell Migration Assay-Cell migration was assessed by 1) in vitro scratch wounding of monolayers and 2) with Transwell 24-well cell culture inserts with 8-m pores (Costar), as described previously (1). For the former, cellular progress was photographed and quantitated by planimetry of the denuded area and converted to distance migrated using Image J software. For Transwell assays, quiescent cells were harvested and added (1 ϫ 10 5 /well) to the insert. Culture medium containing 10% fetal bovine serum as chemotactic agent was added to the lower chamber. After 10 h, nonmigrating cells were removed from upper filter surfaces, and the filter was washed, fixed, and stained. We then photographed fifteen randomly selected 100ϫ fields and counted cells that had migrated to the underside of the filter.
Microtubule Organizing Center (MTOC) Orientation Assay-The cells were grown to confluence, and the monolayer was denuded in a linear stripe using a pipette tip. The cells were fixed 16 h later with 4% paraformaldehyde, and the MTOC was localized by immunolabeling using anti-pericentrin antibody (1:1000, Abcam) and a secondary goat anti-rabbit labeled with Alexa 488 dye (Invitrogen). Cell nuclei and cytoskeleton were identified with DAPI and rhodamine-phalloidin stains, respectively (Invitrogen). The signals were visualized by epifluorescent microscopy, and cells in which the MTOC localized within the 120°sector facing a line parallel to the wound margin were scored positive. At least 100 cells were examined for each condition.

Statistical
Analysis-Experiments were repeated a minimum of three times. Comparisons between two groups were analyzed by Student's t test (p Ͻ 0.05), and comparisons between three or more groups were assessed by analysis of variance with a Bonferroni/Dunn post hoc test (p Ͻ 0.05). The data are presented as the means Ϯ S.E.

Atr1 and Atr2 Are Induced after Vascular Injury and by Growth
Factors-Like fat1 (1), both atr1 and atr2 transcripts increased after rat carotid artery balloon injury. For each isoform, mRNA levels peaked 7 days after injury, with levels ϳ5-, 4-, and 1.8-fold over base line for atr1, atr2S, and atr2L, respectively (Fig. 1A). We also found induction of Atr1 protein in response to serum (Fig. 1B), again reminiscent of the pattern we found with Fat1 (1). Although a similar analysis of Atr2 protein was limited by lack of effective antibodies, Northern analysis showed two distinct atr2 transcripts, which we designate atr2L and atr2S, consistent with previous observations (19). Both atr2L and atr2S transcripts increased in response to serum stimulation, albeit with somewhat different kinetics and a greater induction of atr2S (Fig. 1C). Interestingly, relative atr2 isoform levels showed some cell type dependence, because we found a preponderance of the atr2S isoform in VSMCs and of the atr2L isoform in Chinese hamster ovary cells (supplemental Fig. S1). We also tested the response of Atrs to individual factors with known regulatory roles in vascular remodeling. Atr1 protein levels increased ϳ2.5-fold in VSMCs treated with angiotensin II, basic fibroblast growth factor, platelet-derived growth factor BB, and interleukin-1␤ and decreased in cells treated with interferon-␥ (Fig. 1D). These treatments, like serum treatment, induced atr2S more strongly than atr2L. Among these factors, the effect of interferon-␥ was unique, in that it decreased Atr1 expression while preferentially inducing atr2S mRNA (Fig. 1E).
Atr Proteins Colocalize with Fat1 in Different Subcellular Locations-We then determined the subcellular localization of these proteins. Atr1 staining was prominent in the perinuclear area, with less intense signal present at cell leading edges and within the nucleus. Overall, the pattern of Fat1 staining was quite similar, although the Fat1 signal was more prominent at the leading edges ( Fig. 2A). Overlap of these signals was apparent in the perinuclear area and at cell-cell junctions and leading edges ( Fig. 2A, Merge). Atr2 staining was apparent in cell nuclei and diffusely throughout the cytoplasm and showed overlap with Fat1 in these areas (Fig. 2B). To assess the distribution of the 2 Atr2 isoforms, we transfected cells with constructs encoding epitope-tagged Atr2L or Atr2S. The former localized exclusively to the nucleus, where some overlap with Fat1 was evident (Fig. 2C, Merge). The epitope-tagged Atr2S isoform, on the other hand, was present in the cytoplasm and at cell borders and coincided with Fat1 signal within the nucleus and at cellcell borders (Fig. 2D). Although transfection, epitope tagging, and overexpression can all spuriously affect protein localization, our findings with liposome-mediated transfection of these C-terminal tagged constructs were consistent with studies of endogenous protein (Fig. 2B) as well with studies using N-terminal tags (supplemental Fig. S2) and electroporation (not shown).
Atr Proteins Coimmunoprecipitate with Fat1-To evaluate possible physical association of Fat1 and Atrs, we immunoprecipitated endogenous Atr1 from VSMC lysates and probed for Fat1, which was readily detectable in these immunoprecipitates, but not in control samples (Fig. 3A). This finding using whole cell lysates is consistent with a direct Atr-Fat1 interaction or indirect association based on coexistence in larger multi-protein complexes. We then mapped the domains required for this association. Progressive deletions starting from the Fat1 C terminus (Fig. 3B) were tested by cotransfection with Myc epitope-tagged Atr1 or Atr2L and FLAG-tagged Fat1 derivatives. As shown in Fig.  3C, Fat1 constructs with full-length and slightly truncated cytoplasmic domains retained the ability to coimmunoprecipitate with Atrs (lanes 4 and 5), whereas those with deletions between amino acids 4500 and 4400 lost much of this activity; further deletions to amino acid 4300 eliminated coimmunoprecipitation. We obtained similar results in studies with the interleukin 2R-mouse Fat1 IC domain fusion protein series of internal deletion constructs (1); these experiments implicated both the FC1 (amino acids 4297-4395) and FC2 (amino acids 4395-4497) domains (supplemental Fig. S3). These results point to Fat1 sequences overlapping domains FC1 and FC2, which we previously also identified as interacting domains for ␤-catenin (1). Residues critical for some interactions with Atrs have been localized to the Atr-box (Fig.  3D) in the C-terminal domain of Atr (21). We found that these residues were also important for the interaction of Atr1 and Atr2 with Fat1, because mutation of two conserved leucines within the Atr-box caused a substantial reduction in the binding (Fig. 3E, lane 3

versus lane 4 and lane 5 versus lane 6).
Selective Inhibition of Atr Expression by siRNA-We then developed effective siRNA reagents to allow direct testing of the effects of loss of Atr function in VSMCs. Atr1-selective siRNAs to three distinct targets were transfected into cells at high efficiency and evaluated by Western analysis and qRT-PCR. In each individual case, Atr1 expression was substantially reduced, but a pool of the siRNAs had no additional effect (Fig.  4A). Knockdown mediated by siRNA 1297 was highly efficient at 2 and 3 days after transfection, with some loss of effect by 6 days (Fig. 4B). For siRNAs targeting atr2, we detected knockdown by qRT-PCR caused by a lack of an effective antibody for Atr2 Western analysis. siRNAs targeting the 5Ј region of atr2 (amino acids 689 and 1116) selectively decreased atr2L transcripts, whereas 3Ј region (amino acid 4622) or pooled siRNAs knocked down both atr2L and atr2S transcripts (Fig. 4C). Again, knockdown of ϳ80% by a single siRNA, 4622, persisted through 2 and 3 days after transfection, with some loss of effectiveness by 6 days (Fig. 4D). These siRNAs did not show unintended targeting of other atr isoforms (supplemental Fig. S4).
Atr Knockdown Does Not Affect VSMC Growth-Fat1 knockdown allows robust increases in VSMC proliferation (1). In view of the Fat1 and Atr interactions characterized in Figs. 2 and 3, we speculated that the loss of Atr might also affect VSMC growth. We used siRNA-mediated knockdown of Atr1 and Atr2, separately and together, and evaluated proliferation by

. Colocalization of Atrs and Fat1 in VSMCs.
Shown is immunofluorescence analysis of Fat1 and endogenous rat Atr1 (A), endogenous human Atr2 (B), transfected, epitope-tagged rat Atr2L (C), and transfected, epitope-tagged rat Atr2S (D). In the merged color images, Fat1 staining is red, Atr staining is green, and areas of colocalization (arrows and arrowhead) are yellow. Nuclei stained with DAPI are also shown. Fat1, Atr1, and Atr2 were detected with specific antibodies against the respective proteins; epitope-tagged Atr2 isoforms were detected with anti-c-Myc antibody. Scale bars, 10 m.
bromodeoxyuridine incorporation. These studies provided, however, no evidence that changes in Atr1 or Atr2 expression affected proliferation (supplemental Fig. S5).
Loss of Atr1 Impairs, and Loss of Atr2L Enhances, VSMC Migration-Decreased Fat1 expression has also been linked to impaired cell migration (1, 13, 14, 24). We used siRNA knock-down to test how Atr expression affected the ability of VSMC to migrate in response to denudation injury of a confluent cellular monolayer. Knockdown of Atr1 (atr1 siRNA 1297) decreased the distance that VSMC migrated after injury; an additive effect was seen when expression of both Fat1 and Atr1 was inhibited (Fig. 5A). We also examined the physical distribution of Atr1 and Fat1 within VSMCs responding to monolayer injury. The large arrow in Fig. 5B, drawn orthogonally to the line of denudation, represents the presumed direction of migration. Atr1 and Fat1 were both apparent at filopodial and lamellipodial tips extending toward the denuded area, and Atr1 colocalized with F-actin-rich structures (arrowheads).
To extend this analysis, we tested Atr2 effects on migration. Interestingly, atr2 siRNA 4622, which targets a site common to both short and long atr2 isoforms, had no effect on migration, but atr2 siRNAs 689 and 1116, which target sites present only in atr2L transcripts, each increased the distance that cells migrated (Fig. 5C, left panel). Further evaluation showed that fat1-specific siRNA cotransfected with atr2 siRNA 4622 inhibited migration and prevented the increased migration seen with the atr2 siRNA 689 (Fig. 5C, center  panel). Knockdown targeting all Atrs yielded no net change in migration (Fig. 5C, right panel).
Studies using Transwell assays of cell migration yielded results (Fig.  5D) that were qualitatively similar to those obtained with monolayer denudation. In addition, we studied the effect of combined knockdown of Atr1 and Fat1, which caused a modestly more significant decrease in cell migration compared with individual knockdown (Fig. 5D,  upper left panel). Together, these findings suggest that Atr1, like Fat1, promotes VSMC migration, whereas Atr2L inhibits it, with the latter effect requiring Fat1 expression. Because knockdown of both Atr2 isoforms did not change migration, we suggest that Atr2S, like Atr1, promotes migration, but this effect is masked by concurrent loss of Atr2L with siRNAs targeting sites shared by both isoforms. To . The cell lysates were tested to confirm efficient expression of all transfected constructs (input, bottom panels). D, schematic showing highly conserved Atr-box sequence in Atr1 and Atr2, and location of mutated residues. E, coimmunoprecipitation and Western analysis of transfected FLAG epitope-tagged DelN-hFat1 and Mycepitope tagged Atr proteins, with wild type (wt) or mutated Atr-box sequences. The cell lysates were tested to confirm efficient expression of all transfected constructs (input, bottom panels). The data are representative of three independent experiments. our knowledge, Atr proteins have been not been previously linked to regulation of migration.
Loss of Atr1 Impairs, and Loss of Atr2L Enhances, VSMC Orientation-Work in Drosophila supports a role for Atr in the establishment of planar cell polarity (5). VSMCs in culture do not normally show planar polarization but do respond to injury of the cellular monolayer with directed migration into the denuded area (25,26). We hypothesized that perturbation of VSMC orientation toward the denuded area could contribute to the changes in net migration that we found with knockdown of Atr1 or Atr2L (Fig. 5). To test this idea, we performed scratch wound injury of confluent mouse VSMC monolayers treated with atr siRNAs and assessed cellular orientation by staining the MTOC with a pericentrin-specific antibody (supplemental Fig. S6). Knockdown of Atr1 or Fat1 decreased the percentage of cells along the margin that oriented toward the wounded area; combined Atr1-Fat1 knockdown yielded additional misorientation (Fig. 6A); Targeting of atr2S and atr2L with siRNA 4622 had no net effect on orientation, but knockdown with atr2 siRNA 689, which selectively targets atr2L, caused a significant increase in the percentage of VSMCs oriented toward the wound area; this increase was eliminated by concurrent knockdown of Fat1 (Fig. 6B). Knockdown of both Atr1 and Atr2 had no net effect on orientation (Fig. 6C). Overall, these findings echo the effects of atr1-and atr2-selective siRNAs on migration and support the idea that control of cellular orienta-tion toward the denuded area contributes to the migration effects of Atr knockdown measured in our scratch wound assays.
Overexpression of Atr1 Enhances VSMC Migration and Orientation-To complement the loss-of-function studies described in Figs. 5 and 6, we tested the effect of increased Atr1 expression. For this analysis, we transduced mouse VSMCs with a lentivirus encoding full-length Atr1, selected individual colonies, and identified several clonal lines stably expressing different levels of Atr1 (Fig. 7A). We used line Atr1-2, with a modest level of overexpression, and line Atr1-6, with strong overexpression, in Transwell and orientation assays. Strong Atr1 overexpression yielded a significant increase in migration (Fig. 7B) and orientation (Fig. 7C) compared with VSMCs transduced with vector alone, whereas modest overexpression showed a nonsignificant trend in the same direction.

DISCUSSION
Atrs were first identified because of their pathogenetic role in dentatorubral-pallidoluysian atrophy, a clinical syndrome in which polyglutamine expansions of Atr1 lead to progressive neurodegenerative disease (27)(28)(29). Recent studies show that Atrs have essential functions in metazoan development. Drosophila embryos lacking maternal Atrophin display complex developmental phenotypes, with failures of segmentation, dorsoventral patterning, and neuro- genesis, and mosaic adult flies with clonal lineages bearing atrophin mutations show pleiotropic effects, including accumulation of ectopic wing vein material, notal clefts, nonautonomous planar polarization defects in the eye (18), and position-dependent patterning defects in the leg, in which ventral Atrophin-deficient clones exhibit lateral-distal phenotypes (18,30). Similarly, mice lacking Atr2L show defective patterning in multiple aspects of early development, with specific failure in ventralization of the anterior neural plate, loss of heart looping, and irregular partitioning of somites (20). On the other hand, Atr1, despite its association with clinical neurodegenerative disease, is not required for developmental patterning, perhaps because of functional redundancy with the other vertebrate short form of Atr, Atr2S (19).
Although Atrs have frequently been described as transcriptional corepressors, the molecular basis for Atr function is not entirely clear. A model to explain bidirectional and isoform-dependent Atr transcriptional function has recently been proposed, in which interaction with histone deacetylases confers transcriptional repressive activity to the isoforms that bear extended N-terminal domains (Drosophila Atrophin and vertebrate Atr2L), whereas C-terminal association with the histone acetyltransferase p300 (all Atr isoforms) supports positive transcriptional activity, as seen with Atr1 and Atr2S (19). Nevertheless, a number of examples that do not fit this relatively straightforward model have been reported: 1) Atr1 (a short form) tethered to DNA recruits repressive activity in cultured cells (18,31,32); 2) Atr1 and Drosophila Atrophin (a long form) each repress transcription when tethered to DNA in vivo (18); and 3) Drosophila Atrophin has been characterized as positive regulator of gene expression in the proximoventral leg (30). Several possible explanations for these discrepancies can be suggested. Atrs may associate with additional unknown tran- In the merged color images, Atr1 staining is green, Fat1 or F-actin staining is red, and areas of colocalization (arrowheads) are yellow. The arrow points orthogonally to the wound edge. C, scratch wounding migration assay of mouse VSMCs transfected with CTL or specific atr2 siRNA along or in combination with fat1 or atr1 siRNAs, as indicated. The method and analysis are as described above, except that migration proceeded for 30 h for the right panel. D, Transwell migration assay of mouse VSMCs transfected with control or specific atr or fat1 siRNA along or in combination, as indicated. Fifteen photomicrographic fields were counted per condition, and the values were averaged for each filter. The data show the means Ϯ S.E. *, p Ͻ 0.05; **, p Ͻ 0.01, versus control. The data are representative of three independent experiments.
scriptional regulators, interactions may be controlled by posttranslational mechanisms not yet identified, or some of the reported effects may be indirect. In addition, short Atr isoforms are found outside the nucleus (Fig. 2), but whether they have specific functions in the cytoplasm is unknown. The link between Atrophin-mediated gene regulation and the cellular effects that underlie critical Atrophin roles in development and neurodegenerative disease are also not fully understood. Lost expression of a diffusible signal may underlie planar polarity defects seen in the Drosophila eye with loss of Atrophin function (5). In Atr2L-deficient mice, misregulation of Shh and Fgf8 expression during development may undermine proper formation and function of critical anterior midline, anterior neural ridge, and apical ectodermal ridge signaling centers (20). Increased Atr levels appear to be particularly toxic in neural tissues, because increased apoptosis was found in neuroblastoma cells with increased Atr2L expression (33) and in Drosophila neurons with increased Atrophin expression that results from the loss of negative regulation by the micro RNA mIR-8 (34).
In the present study, we found that Atrs are induced in injured arteries and in cultured VSMCs exposed to growth factors. In view of previous similar findings with the atypical cadherin Fat1 (1) and the reported interaction of Drosophila Atrophin and Fat (5), we sought to test the role of Atrs in the prominent VSMC activities provoked by these in vivo and in vitro manipulations. Although Fat1 inhibits growth and enhances migration of VSMCs (1), our present studies demonstrate that loss of Atr expression affects VSMC migration and orientation, but not growth; thus Atrs may participate in part, but not all, of the Fat1-mediated effects on VSMC activities. Interestingly, these results are consistent with descriptions of  . Effect of Atr1 overexpression on VSMC migration and orientation in scratch-wounded monolayers. A, Western analyses of Atr1 overexpression in mouse VSMC stable transfectants. After Blasticidin selection, cell lysates from a vector-transfected clone (pLenti6), Atr1 mass cultures (Atr1mass) as well as several Atr1 clonal cell lines (Atr1-1 to Atr1-9, respectively) were immunoblotted with antibody against Atr1. The blot was also probed for Actin as loading reference. B, effect of Atr1 overexpression on VSMC migration in scratch-wounded monolayers. Atr1-2, Atr1-6, and pLenti6 vector alone mouse VSMC were allowed to migrate 28 h after wounding of monolayer, and the migration distance was evaluated as described above. The data show the means Ϯ S.E. *, p Ͻ 0.05. C, effect of Atr1 overexpression on VSMC orientation in scratch-wounded monolayers. Atr1-2, Atr1-6, and pLenti6 vector alone mouse VSMC were fixed 16 h after wounding and stained to show MTOCs. The data show the means Ϯ S.E. *, p Ͻ 0.05. The data are representative of three independent experiments.
Atr mutant phenotypes in Drosophila, which show defects in cell fate and patterning similar to Fat mutants but lack the tumor suppressor phenotype found in the latter (5). Although Atrophin regulates planar cell polarity in Drosophila, and Atr2L is required for proper formation of critical signaling centers in early vertebrate development, a specific effect of Atrs on cellular migration and/or orientation has not been reported previously.
Re-expression of fetal developmental genes in the setting of vascular injury is well known (35). Distinct regulation of Atr isoforms after injury suggests, however, a level of control of Atr activity that has evolved since the divergence of metazoan lineages leading to insects and vertebrates. One possible explanation for the distinct effects of different Atr isoforms on migration and orientation is that although Atr2L acts to limit migration and orientation, this effect is opposed by the short isoforms, Atr1 and Atr2S. The short isoforms lack the N-terminal domains that interact with histone deacetylases (21,22,32) and conceivably could interfere with gene repression mediated by long forms of Atr (19,36). Such a model is compelling in part given the preferential induction of the short isoforms by promigratory growth factors and by arterial injury (Fig. 1); in these settings, inhibition of Atr2L-mediated gene repression by induction of the short isoforms might allow transcriptional reprogramming of VSMCs to support a more migratory phenotype. In addition, we surmise that the (nonessential) short isoforms may have evolved to add a higher layer of regulation to the (essential) evolutionarily conserved long form of Atr. This also raises the question of whether analogous but necessarily distinct mechanisms of Atrophin regulation exist in Drosophila, which have only a single long Atrophin isoform.
We developed evidence along several lines that support the validity and significance of the Atr-Fat1 interaction; both proteins 1) show a similar pattern of regulation after vascular injury, which suggests a shared biological function, 2) colocalize within cells, 3) coimmunoprecipitate at endogenous levels of expression, and 4) depend on discrete sequences for their physical interaction, which argues against nonspecificity of the interaction. In addition, we found that enhanced VSMC migration mediated by Atr2L knockdown requires Fat1 expression. The mechanism underlying the opposing effects of Fat1 and Atr2L on migration is not yet clear. As discussed above, it may be that Atr2L, acting as a transcriptional repressor, controls expression of gene products directly involved in cellular migration as promoted by Fat1; by antagonizing Atr2L-mediated gene repression, the short Atr isoforms also enhance migration. Because Fat1 and Atrs interact at multiple cellular locations, alternative mechanisms must also be considered. For example, it is possible that Fat1-Atr1 or -Atr2S interactions at or near the cell surface could modify interactions of the Fat1 IC domain with Ena/VASP proteins (13,14), thereby promoting migration via effects on actin cytoskeletal remodeling; alternatively, Fat1 IC domain in the cell nucleus (1,24,37) interacting with Atrs could affect Atr-mediated transcriptional activities at migration-related target genes. Future studies will address the precise mechanistic link between Fat1-Atr interactions and effects on VSMC activities and determine the significance of these interactions for VSMC migration and vascular disease.