β8 Integrin Binds Rho GDP Dissociation Inhibitor-1 and Activates Rac1 to Inhibit Mesangial Cell Myofibroblast Differentiation*

αvβ8 integrin expression is restricted primarily to kidney, brain, and placenta. Targeted αvor β8 deletion is embryonic lethal due to defective placenta and brain angiogenesis, precluding investigation of kidney αvβ8 function. We find that kidney β8 is localized to glomerular mesangial cells, and expression is decreased in mouse models of glomerulosclerosis, suggesting that β8 regulates normal mesangial cell differentiation. To interrogate β8 signaling pathways, yeast two-hybrid and co-precipitation studies demonstrated β8 interaction with Rho guanine nucleotide dissociation inhibitor-1 (GDI). Selective β8 stimulation enhanced β8-GDI interaction as well as Rac1 (but not RhoA) activation and lamellipodia formation. Mesangial cells from itgb8-/- mice backcrossed to a genetic background that permitted survival, or gdi-/- mice, which develop glomerulosclerosis, demonstrated RhoA (but not Rac1) activity and α-smooth muscle actin assembly, which characterizes mesangial cell myofibroblast transformation in renal disease. To determine whether Rac1 directly modulates RhoA-associated myofibroblast differentiation, mesangial cells were transduced with inhibitory Rac peptide fused to human immunodeficiency virus-Tat, resulting in enhanced α-smooth muscle actin organization. We conclude that the β8 cytosolic tail in mesangial cells organizes a signaling complex that culminates in Rac1 activation to mediate wild-type differentiation, whereas decreased β8 activation shifts mesangial cells toward a RhoA-dependent myofibroblast phenotype.

␣v␤8 integrin expression is restricted primarily to kidney, brain, and placenta. Targeted ␣v or ␤8 deletion is embryonic lethal due to defective placenta and brain angiogenesis, precluding investigation of kidney ␣v␤8 function. We find that kidney ␤8 is localized to glomerular mesangial cells, and expression is decreased in mouse models of glomerulosclerosis, suggesting that ␤8 regulates normal mesangial cell differentiation. To interrogate ␤8 signaling pathways, yeast two-hybrid and co-precipitation studies demonstrated ␤8 interaction with Rho guanine nucleotide dissociation inhibitor-1 (GDI). Selective ␤8 stimulation enhanced ␤8-GDI interaction as well as Rac1 (but not RhoA) activation and lamellipodia formation. Mesangial cells from itgb8؊/؊ mice backcrossed to a genetic background that permitted survival, or gdi؊/؊ mice, which develop glomerulosclerosis, demonstrated RhoA (but not Rac1) activity and ␣-smooth muscle actin assembly, which characterizes mesangial cell myofibroblast transformation in renal disease. To determine whether Rac1 directly modulates RhoA-associated myofibroblast differentiation, mesangial cells were transduced with inhibitory Rac peptide fused to human immunodeficiency virus-Tat, resulting in enhanced ␣-smooth muscle actin organization. We conclude that the ␤8 cytosolic tail in mesangial cells organizes a signaling complex that culminates in Rac1 activation to mediate wild-type differentiation, whereas decreased ␤8 activation shifts mesangial cells toward a RhoA-dependent myofibroblast phenotype.
Integrins are a family of ␣ and ␤ heterodimeric extracellular matrix receptors that mediate vital cell functions, including adhesion, migration, proliferation, and survival. Eighteen ␣ and eight ␤ subunits assemble to form 24 heterodimers. The ␤8 subunit is a 769-amino acid polypeptide that partners exclusively with ␣v (1). Although ␤8 expression was initially reported to be restricted to brain, placenta, and kidney (1), subsequent reports have documented its expression in lung and eyelid (2,3). ␤8 gene deletion causes embryonic lethality due to impaired placenta and brain angiogenesis (4), which is similar to the phenotype observed in ␣vϪ/Ϫ mice (5), indicating that ␤8 is the major ␣v partner during development (5,6). Kidney ␤8 localization and function have not been described.
Similar to other cell surface receptors, ligand binding to integrins leads to generation of intracellular signals and cytoskeletal rearrangement, i.e. outside-in signaling. The major site for soluble signaling molecule and cytoskeleton binding is the ␤-subunit cytosolic tail, and the ␣-subunit generally serves a regulatory function (7), although two groups have recently shown that ␣-subunits may also direct signal transduction pathways (8,9). A unique feature of integrins is that independently generated intracellular signals can induce conformational changes to extracellular integrin domains, which enhances ligand affinity and/or extracellular matrix assembly, i.e. inside-out signaling. Neither outside-in nor inside-out signaling pathways have been described for ␣v␤8. The ␤8 cytoplasmic tail is 66 amino acids in length (10,11) and is predicted to contain an ␣-helix that extends from residues 730 -744 (psiphred) but no definable protein interaction domains (SMART, Motifscan) or sequence homology with other integrins (BLAST). ␤-Integrin domain swapping experiments showed that the ␤8 cytosolic tail does not directly affect cell adhesion (10), further suggesting that ␤8 signaling is distinct from other integrins.
The cytosolic tail of other ␤-integrin subunits has been shown to regulate the Ras superfamily of small molecular weight GTP-binding proteins (G-proteins). In particular, the Rho subfamily, which is commonly represented by three members, Rho, Rac, and Cdc42, is required for integrin-dependent cytoskeletal assembly (12). RhoA activation establishes stable, integrin-based focal adhesions at the cell periphery and is characterized in vitro by focal adhesion kinase activation. Rac1 mediates several aspects of integrin-dependent cell migration, including formation of membrane ruffles and lamellipodia at the leading edge of migrating cells and focal complexes at more internal sites. Integrin-associated Cdc42 function is required for development of actin-rich filopodia at the leading edge of migrating cells. Because ␤8 mediates nascent adhesion formation in migration (10,13) and the ␤8 cytosolic tail lacks consensus focal adhesion kinase or talin binding sequences (14), we hypothesized that ␤8 preferentially activates Rac or Cdc42 signaling.
G-proteins act as binary switches, which cycle between the GDP-bound inactive state and the GTP-bound active conformation. Rho family G-proteins are regulated by direct binding to GTP exchange factors (GEFs), 4 GTPase-activating proteins (GAPs), and Rho guanine nucleotide dissociation inhibitors (GDIs). Rho family G-proteins are also regulated by subcellular compartmentalization, as activation requires that lipid-modified G-proteins translocate to membrane microdomains for GTP loading and localization to receptors and effectors. It has been suggested that GDI bi-directionally regulates G-protein localization by chaperoning Rho family G-proteins to selective membrane signaling domains, which permits GEF-regulated GDP-GTP exchange as well as by removal of GDP-bound G-proteins from membrane sites and sequestration within the cytosol (15)(16)(17)(18).
We find that under physiologic conditions, kidney ␤8 integrin is localized to glomerular mesangial cells (MCs), and animal models of glomerulosclerosis are associated with decreased MC ␤8 expression. In vitro, MC ␤8 stimulation leads to ␤8-GDI interaction, Rac1 activation, and suppression of RhoA-regulated, pathologic features. The data suggest that the ␤8 cytosolic tail provides specificity to G-protein signaling and regulates MC phenotype by spatially coordinating GDI-bound Rac1 to discrete domains containing Rac-GEFs and effector molecules.
Animal Models-ROP-Os/ϩ mice, which were purchased from The Jackson Laboratory (Bar Harbor, ME), have a radiation-induced inversion mutation on chromosome 8, which results in a 50 -75% decrease in nephron number and oligosyndactyly. ROP-Os/ϩ mice develop proteinuria and renal lesions, which resemble focal and segmental glomerulosclerosis over 3-6 months and ultimately die from renal failure (20). Mice overexpressing a non-infectious HIV transgene lacking gag and pol genes develop glomerular disease, which is indistinguishable from human HIV-associated nephropathy (HIVAN) (21). Characteristic features include the nephrotic syndrome, focal and segmental glomerulosclerosis, microcystic tubular dilatation, and progression to end-stage renal failure. All protocols were approved by the Institutional Animal Care and Use Committees at Case Western Reserve University.
Northern Blot Analysis-Established methods were followed for Northern blotting (13). Total RNA was extracted using the RNeasy kit (Qiagen, Valencia, CA), and 20 g per lane was fractionated on a denaturing 1.0% agarose, 0.67% formaldehyde gel, transferred to nylon membranes, and cross-linked by UV light exposure. To assess ␤8 integrin mRNA levels, full-length human ␤8 cDNA (1) was used as a template to generate probes, which were labeled with [␣-32 P]dCTP to specific activity Ն1.0 ϫ 10 8 cpm/g DNA (RTS Random Prime DNA labeling system, Invitrogen). Hybridization and high stringency washes were conducted according to previously described methods (13). Blots were stripped and re-hybridized with a randomprimed oligonucleotide probe derived from chicken ␤-actin cDNA template (BD Biosciences) as a control for housekeeping gene expression.
Quantitative Reverse Transcription (RT)-PCR-Real-time quantitative RT-PCR analysis was employed to determine mRNA content in mouse kidney using LightCycler and SYBR Green technology (Roche Applied Science). Each analysis included mouse kidney samples of unknown mRNA concentration and, to confirm amplification specificity, random-primed RNA in the absence of RT or RNA template as negative controls. Total RNA (3 g) was first treated with DNase I and then reverse-transcribed using the Thermoscript RT-PCR System (Invitrogen) in a 20-l volume. Two-l cDNA products were PCR-amplified in buffer containing 2 l of LightCycler-Fast-Start DNA Master SYBR Green I mix (Roche Applied Science), 18 l of hybridization buffer, 5 M gene-specific primers, and 3 mM MgCl 2 . ␤8-Specific primers were 5Ј-ATGCACAATA-ATATAGAAAAA-3Ј (nt 475-495) and 5Ј-TCCTTGTAC-CAATGAAACTG-3Ј (nt 1039 -1058). To quantify and validate RNA integrity, real-time PCR for ␤-actin internal standard was also performed using 5Ј-ATCTGGCACCACACCTTCTA-CAATGAGCTGCG-3Ј (nt 333-364) and 5Ј-CGTCATACTC-CTGCTTGCTGATCCACATCTGC-3Ј (nt 1139 -1169) primers (BD Biosciences Clontech, Palo Alto, CA). Thermocycling conditions were 95°C for 10 min (initial denaturation), then 45 cycles at 95°C for 10 s (denaturation), 52°C for 20 s (annealing), and 72°C for 30 s (extension). PCR products from all primer pairs were subjected to melting curve analysis and then analyzed by agarose gel electrophoresis to confirm amplification of a single product of the predicted size. LightCycler software was used to establish amplification cycle thresholds (C T ), which demarcates the cycle number when sample fluorescence is above background and within the initial logarithmic phase. For each set of probes and primer pairs, serial dilutions validated that control and experimental gene amplification efficiencies were similar. Transcript quantification was determined by the comparative C T (⌬⌬C T ) method (23). Data are expressed as relative ␤8 mRNA abundance, which is defined as the ␤8-integrin level normalized to ␤-actin transcript content within the same sample.
Plasmid Transfections-Plasmids were transformed into competent DH-5␣ bacterial strain according to the manufacturer's protocol (Invitrogen), extracted using a Maxiprep Kit (Qiagen, Valencia, CA), and amplified by culture in Luria-Bertani broth containing appropriate antibiotics. cDNAs were transiently transfected into CHO-B2/v7 or HEK293 cells according to previously described methods (32,33). Briefly, cells were plated in 6-well dishes (0.25 ϫ 10 6 cells/well) and cultured overnight in Dulbecco's modified Eagle's medium/ F-12 plus 10% fetal bovine serum to achieve 80% confluence. The transfection mixture, which contained 2.0 g of plasmid DNA and 6 l of FuGENE 6 transfection reagent (Roche Applied Science) in 100 l serum-free Dulbecco's modified Eagle's medium (Invitrogen), was mixed for 20 min at room temperature and then added to each well with complete medium for 24 h. Cells were evaluated for protein expression by immunoprecipitation (see the methods below) 24 -48 h post-transfection.
Stable ␤8 integrin transfectants were generated from HEK293 cells as previously described (13). Briefly, cells were cultured to 50% confluence in 10-cm dishes, then transfected with 2 g of human ␤8 integrin cDNA (1) subcloned into pcDNA1-Neo vector using FuGENE 6 reagent. Cells were then cultured in complete media containing G418 (Sigma). Individual G418-resistant clones were isolated by limiting dilution, subcultured, and assayed for ␤8 integrin expression by biotin surface labeling and immunoprecipitation according to published methods (13). Stable transfectants with persistent ␤8 integrin overexpression were utilized after three to five passages.

␤8 Regulates G-protein Signaling
Yeast Two-hybrid Screening-Methods have previously been reported (22) and have been modified as described below. The two-hybrid bait construct was generated by PCR amplification of the cDNA encoding the C-terminal 60-amino acid sequence of human integrin ␤8 (1) with 5Ј-CAGGTGGAATTCCAATG-GAATAGT-3Ј (nt 2116 -2139) and 5Ј-AGAATAGGATC-CTCTAGATG-3Ј (vector sequence 3Ј to stop codon) primers. The resulting fragment was inserted into pAS2-1 vector (MATCHMAKER Two-Hybrid System 3; BD Sciences Clontech) in-frame with the GAL4 DNA binding domain to generate pAS␤8. Two-hybrid analysis was performed in the Saccharomyces cerevisiae strain AH109 (MAT␣ trp1-901 leu2-3112 ura3-52 his3-200 gal4⌬ gal80⌬ LYS2::GAL1-HIS3 GAL2-ADE2 met2::GAL7-lacZ). No transcriptional activation by bait fusion protein alone was detected, as measured by expression of three reporters, lacZ, ADE2, and HIS3 (see Table 1). To select protein partners interacting with ␤8 cytosolic tail, AH109 cells were co-transformed with pAS␤8 and Human Kidney Matchmaker cDNA Library (Clontech) in pACT2 vector, which contained the GAL4 transcriptional activation domain fused to inserts. Transformants were initially selected on Leu Ϫ /Trp Ϫ /His Ϫ media in the presence of 5 mM 3-aminotriazole. Selected clones were further tested for growth on Leu Ϫ /Trp Ϫ /Ade Ϫ and Leu Ϫ / Trp Ϫ /His Ϫ /Ade Ϫ media by replica plating. Positive clones were further tested by ␤-galactosidase expression and sequenced. To exclude nonspecific protein interaction with ␤8 fusion protein, control WTIP constructs (22) were inserted into the activation domain plasmid, co-transformed with pAS␤8, and analyzed for reporter expression. Using a similar strategy, GDI specificity was tested for interaction with WTIP and human ␤3-integrin negative controls. The sequence encoding the entire ␤3 cytoplasmic tail was amplified from the IL-2R/␤3 construct using oligonucleotide primers 5Ј-CAGGCTGATAATGATCTGAG-GATGAC-3Ј containing an NcoI site and 5Ј-ATTGGCCTT-GCCGCCCTGCTCATCTG-3Ј containing a BamHI site, cloned into pAS2-1. To ensure specificity and direct bait-prey interaction, individual plasmid prey constructs were co-transformed in AH109 with pAS␤8 to recapitulate results obtained during the library screening process.
Data Presentation-All data are representative of three to four experiments per condition. Graphical results are presented as the mean Ϯ S.E. unless otherwise indicated.

RESULTS
Kidney ␤8 Integrin Is Localized to MCs-Our original interest in the ␤8 integrin stemmed from investigation of Fas (CD95)directed pathway activation in renal tubular epithelial cells. Using a hybridization array approach, we found that Fas stimulation up-regulated ␤8 expression in cultured proximal tubule epithelial cells (13). To characterize ␤8 expression in vivo, mouse kidney sections were probed for ␤8 mRNA expression by in situ hybridization, since suitable antibodies for immunohistochemical studies were not available. Kidney ␤8 was unexpectedly expressed in a predominant, glomerular mesangial pattern (Fig. 1, A and C ). Mouse kidney glomerular and tubular mRNA expression was also assessed by RT-PCR and confirmed that ␤8 is expressed primarily in glomeruli (Fig. 1D). ␤8 protein content was determined in cultured MC and tubule cell lines by immunoprecipitation of lysates from biotin surface-labeled cells, which revealed robust expression in rat MCs, and to a lesser extent in the HRPT human proximal tubule cell line (Fig.  1E, upper panel), consistent with previously published data (13). HEK293 cells do not express endogenous ␤8 mRNA or protein (Fig. 1E, upper panel), in agreement with previous reports (43). For this reason HEK293 cells were stably transfected with ␤8 cDNA and used in subsequent experiments. In control studies ␤1 protein expression was noted to be similar between the three cell lines (Fig. 1E, lower panel).
Kidney ␤8 Integrin Expression Is Reduced in Mouse Models of Renal Disease-To elucidate ␤8 function in kidney, mouse models of glomerulosclerosis were examined for ␤8 expression. Northern blots from wild-type and ROP-Os/ϩ kidney (20) demonstrated progressively decreased ␤8 expression over time in the ROP-Os/ϩ group ( Fig. 2A). Kidney ␤8 protein levels were similarly decreased in ROP-Os/ϩ compared with wild-type mice (Fig. 2B). In contrast, ␤1 integrin immunoblots revealed no difference between ROP-Os/ϩ and wild-type kidneys (Fig.  2B). To determine whether this observation is generalized to other models, ␤8 mRNA expression was examined by real-time FIGURE 1. ␤8 integrin is expressed in MCs. In situ hybridization with an alkaline phosphatase detection system (stains blue) using ␤8 integrin riboprobes in mouse kidney sections is shown. A, glomeruli with mesangial staining pattern and the absence of tubule cell labeling (ϫ400 magnification). Alkaline phosphatase activity due to riboprobe trapping is noted within the interstitial space and along tubular basement membranes. B, to assess background labeling, control hybridization was conducted with a sense probe, which yielded no alkaline phosphatase activity (ϫ400 magnification). C, cytoplasmic staining of cells with a mesangial distribution (shown by arrows, ϫ1000 magnification) in a single glomerulus. D, mouse kidney glomeruli and tubules were fractionated by microdissection and Percoll gradient centrifugation, respectively. ␤8 and glyceraldehyde-3-phosphate dehydrogenase (GAPDH ) mRNA were amplified by RT-PCR (30 cycles). E, rat mesangial cells in primary culture (RMC ), human renal proximal tubule cells (HRPT), and HEK293 human embryonic kidney tubule cells were surface-biotinylated, and lysates with equal protein content were immunoprecipitated with rabbit ␤8 integrin anti-sera and probed by immunoblot analysis with peroxidase-conjugated streptavidin (upper panel ). Parallel lysates from the same cell lines (20 g of protein per lane were probed for ␤1 integrin expression by immunoblot analysis (lower panel ). PCR in a mouse model of HIVAN (21) as well as in ROP-Os/ϩ mice. These experiments demonstrated decreased ␤8 mRNA content in HIVAN and ROP-Os/ϩ compared with agematched control kidneys (Fig. 2C). We conclude from these studies that glomerular ␤8 expression is suppressed in models of primary (Os) and secondary (HIVAN) forms of glomerular disease.
␤8 Interacts with RhoGDI-1 (GDI)-Upon ligation with extracellular matrix proteins, integrins undergo conformational changes that permit ␤-subunit cytoplasmic tails to associate with intracellular signaling molecules (44). However, unlike other ␤-integrins, the ␤8 tail contains no predicted protein interaction or signaling domains (11). To identify cell signaling pathways regulated by ␤8, we employed a yeast twohybrid screen with the ␤8 cytoplasmic domain as bait ( Table 1). Screening of 2.5 ϫ 10 6 transformants yielded 18 positive clones, 2 of which corresponded to the C terminus of GDI. Directed yeast two-hybrid assays with ␤3 or ␤8 integrin cytoplasmic tail sequences in pAS2-1 and C-terminal GDI sequence in pACT2 confirmed the ␤8-GDI interaction, whereas GDI did not interact with ␤3 ( Table 1). The ␤8-GDI interaction is of particular interest because gdiϪ/Ϫ mice develop a renal phenotype characterized by glomerulosclerosis and proteinuria (27). Yeast two-hybrid screening also identified Band 4.1B as a ␤8 interactor, which confirms recently published data from McCarty et al. (43). Fig. 3A demonstrates that in rat MCs in primary culture, the U373 human astrocytoma cell line, and HEK293 cells stably expressing ␤8, RhoGDI-1 co-precipitates with ␤8, confirming the yeast two-hybrid findings. In co-precipitation experiments using HEK293 cells overexpressing chimeric receptors, with the IL-2 receptor extracellular domain fused to ␤1, ␤3, or ␤8 cytoplasmic tails, GDI interacted exclusively with IL-2R/␤8 (Fig. 3B), suggesting that the integrin-GDI interaction is specific for ␤8. To address whether ligand occupancy drives ␤8-GDI interaction, rat MCs were plated on vitronectin ligand or poly-L-lysine, which permits cell attachment by an integrin-independent mechanism. ␤8-GDI association was assessed by immunoprecipitating GDI and blotting for ␤8. Fig. 3C shows that lysates from vitronectin-stimulated MCs demonstrated robust ␤8-GDI interaction, whereas co-precipitation was not observed in lysates from poly-L-lysine-treated cells, indicating that ligand stimulation recruits GDI to bind the ␤8 cytosolic tail.
␤8 Ligation Stimulates Rac1 and Suppresses RhoA Activation-Because GDI and integrins regulate Rho family G-protein signaling, we next tested for ␤8-dependent G-protein activation.
To further address the role of ␤8-regulated G-protein signaling in MCs, poly-L-lysineand vitronectin-stimulated Rac1 and RhoA activities were determined in itgb8ϩ/ϩ and itgb8Ϫ/Ϫ MC lines (Fig. 5A). Fig. 5B demonstrates Rac1 activation by vitronectin in wildtype MCs. In contrast, Rac1 activity was undetectable after plating itgb8Ϫ/Ϫ MCs on poly-L-lysine or vitronectin. RhoA activity was extremely low in wild-type MCs exposed to poly-L-lysine or vitronectin (some activity was observed with longer film exposures) and constitutively activated in itgb8Ϫ/Ϫ MCs. These data are consistent with Fig. 4 results and demonstrate that ␤8 up-regulates Rac1 and suppresses RhoA activation. In data not shown, MC stimulation with serum (10% fetal calf serum, 5 min) to activate Rho and Rac pathways did not affect ␤8 expression, as determined by RT-PCR.
␤8 Activation Is Associated with Rac1 Release from GDI-For Rac1 activation to occur, GDP-bound Rac1 must first be released from GDI, which may be facilitated by a GDF before interaction with a RacGEF for GTP loading. ␤8 ligand binding or clustering regulates Rac1 signaling as well as GDI interaction with the ␤8 cytosolic tail, consistent with mechanisms whereby the ␤8 tail could function as a GDF or GEF. As an initial test to distinguish between these possibilities, CHO-B2/v7 cells expressing chimeric IL-2R/␤8 or IL-2R/␤3 receptors were clustered with Tac antibodies. Integrin cytosolic tail interaction with GDI or Rac1 was then assessed by immunoprecipitation. As seen in Fig. 6, integrin clustering resulted in GDI interaction with ␤8, but not ␤3, in agreement with Fig.  3 data. Moreover, ␤8 did not interact with Rac1, consistent with GDF, rather than RacGEF activity.
␤8 Regulation of G-protein-dependent Cell Morphology-Characteristic in vitro manifestations of Rho family G-protein activation are actin-rich lamellipodia (Rac1), filopodia (Cdc42), and stress fiber formation (RhoA) (12,46). To test for ␤8 regulation of these morphologic features, HEK293 cells expressing IL-2R⌬ or IL-2R/␤8 were stimulated with Tac antibodies and then labeled for F-actin with phalloidin. As seen in Fig. 7A, IL-2R⌬-stimulated cells were small, with little cytoplasm or actin assembly. In contrast, IL-2R/␤8expressing cells developed marked morphologic changes,  Fig. 3 for the indicated times. C, E, and G, as an alternative method of ␤8 activation, HEK293 cells were stably transfected with chimeric receptor constructs composed of IL-2 receptor extracellular domain fused to ␤8 cytoplasmic domain (IL-2R/␤8), negative control IL-2R extracellular domain only (IL-2R⌬), or negative control ␤8 cytosolic domain only (␤8-cd). Transfected cells were maintained in serum-free media for 24 h before incubation with clustering Tac or isotype control antibody-coated beads. F, GTPase activity was determined in ␤8-transfected CHO-B2/v7 cells plated on vitronectin (VN, 10 g/ml, 30 min) or negative control poly-L-lysine (PLL, 10 g/ml, 30 min). G-protein activity was determined by pull-down assays, whereby whole cell lysates were first incubated with GST beads bound to PAK1 binding domain (B-E ) or rhotekin (F and G). GTP-bound G-proteins were detected by immunoblotting for Rac1 (B and C ), Cdc42 (D and E), or RhoA (F and G). Corresponding whole cell lysates were probed for expression of individual G-proteins by immunoblot analysis (B-G, lower panels). FIGURE 5. ␤8 expression down-regulates RhoA activity. MCs from itgb8ϩ/ϩ and itgb8Ϫ/Ϫ mice were harvested and maintained in primary culture. A, to verify itgb8 gene deletion, whole cell lysates were probed for ␤8 integrin protein expression by immunoblot analysis. B, itgb8ϩ/ϩ and itgb8Ϫ/Ϫ MCs were incubated with poly-L-lysine (PLL, 10 g/ml, 30 min) or vitronectin (VN, 10 g/ml, 30 min) and then assayed for Rac1 and RhoA activities, as described in Fig. 4. with cytoplasmic spreading, broad lamellipodia, rare filopodia, and the absence of stress fibers (Fig. 7B). Taken together with data from Figs. 4 and 5, we conclude that the ␤8 cytosolic tail stimulates Rac1 and Cdc42 and suppresses RhoA activation.
Rac1 Regulates MC Myofibroblast Phenotype-A cardinal feature of myofibroblast differentiation, which characterizes MC pathology, is ␣-SMA-containing stress fibers. ␣-SMA expression and assembly are regulated by RhoA in mesenchymal cells, including MCs (47,48). Opposing Rac1 and RhoA signaling in response to ␤8 stimulation, therefore, suggests that Rac1 may regulate MC-myofibroblast transformation by inhibiting RhoA (44, 49 -51). The next set of experiments was designed to test whether Rac1 directly modulates RhoA-dependent myofibroblast differentiation.
G-protein functions are dependent upon interaction with effector molecules, such as the serine-threonine kinase PAK and the Wiscott-Aldrich protein (WASP). As a tool to interrogate the effect of Rac1 or Cdc42 upon RhoA-regulated cell phenotype, a peptide corresponding to Rac or Cdc42, which blocks the interaction with PAK, was fused with the HIV Tat protein (52)(53)(54) and incubated with human MCs in primary culture. ␣-SMA assembly was employed as the assay for functional RhoA activation. At incubation times ranging from 8 to 36 h, compared with cells treated with Tat alone, Tat-Rac (17-32)treated cells and to a lesser extent Cdc42 (17-32) cells displayed a more spread morphology and increased ␣-SMA organization (Fig. 8). The data are consistent with Rac1 suppression of RhoAdependent myofibroblast differentiation.
Implications of MC Rho Family GTPase Signaling by GDI-MCs derived from gdiϪ/Ϫ mice demonstrate enhanced RhoA, but not Rac1 activation (not shown), consistent with divergent GDI regulation of Rho family G-proteins (44). To test for regulation of cell morphology by GDI, quiescent mouse gdiϪ/Ϫ and gdiϩ/ϩ MCs were stained for ␣-SMA or F-actin by immunocytochemical methods. As seen in Fig. 9B, serum-starved gdiϪ/Ϫ cells demonstrated robust ␣-SMA assembly into stress fibers, consistent with associated RhoA activation, whereas wild-type cells displayed few ␣-SMA stress fibers (Fig. 9A). Differences in phalloidin staining were also observed between gdiϪ/Ϫ and gdiϩ/ϩ MCs (Figs. 9, C and D), indicating that stress fibers were derived from ␣-SMA and ␤-actin. In addition, lamellipodia were less prominent in gdiϪ/Ϫ compared with gdiϩ/ϩ MC.
Taken together the data support the hypothesis that ␤8 interacts with GDI to promote wild-type MC phenotype via Rac1dependent suppression of RhoA. In MCs with targeted deletion of itgb8 or gdi, ligand-mediated Rac1 activity is lost, which permits RhoA-dependent myofibroblast features to predominate.

DISCUSSION
Cell-matrix interaction and associated signal transduction pathways are regulated by multiple mechanisms, including cell-specific expression of different ␣-␤ integrin heterodimers. Until now, the only well established MC ␤-integrin subunit was ␤1, which partners with ␣1, and to a lesser extent with ␣2, ␣3, and ␣6 (55). The original ␤8 cDNA clon-FIGURE 6. ␤8 activation is associated with Rac1 release from GDI. HEK293 cells expressing IL-2R⌬, IL-2R/␤8, IL-2R/␤3, or empty vector were incubated with clustering Tac or control antibodies. Chimeric integrins were immunoprecipitated (IP) from cell lysates with Tac IgG, resolved by SDS-PAGE, and then probed for GDI or Rac1 by immunoblot analysis (upper two panels). Fractions of cell lysate before immunoprecipitation were immunoblotted with anti-GDI and anti-Rac1 antibodies (lower two panels) to assess expression. ␤8 Regulates G-protein Signaling JULY 14, 2006 • VOLUME 281 • NUMBER 28 ing data demonstrated abundant mRNA expression in kidney (1), but the current report represents the first description of kidney ␤8 localization, which is most prominent in the MCs. We also found that kidney ␤8 mRNA expression was markedly diminished in two different models of progressive glomerular disease, suggesting that ␤8 may specify the normal MC differentiation state and that loss of ␤8 expression may contribute to glomerulosclerosis pathogenesis. Diminished ␤8 expression is unlikely to be due to scarring and loss of MCs, since kidney ␤1 integrin expression persisted in mouse models of glomerulosclerosis.
MC expression of both ␤1 and ␤8 permits switching between integrin pools to achieve specific responses, such as intracellular signaling, cytoskeleton remodeling, and motility, to dynamic extracellular cues (56,57). Unlike ␤1 integrins, which recognize many matrix proteins, ␣v␤8 is less promiscuous, with mammalian ligands including vitronectin, latent TGF␤, and perhaps laminin-1 and type IV collagen (10,58,59). Type IV collagen isoforms, fibronectin and laminins 8 and 9, reside in normal glomerular mesangium (60), whereas vitronectin, laminin-1, and latent TGF␤ do not (61)(62)(63)(64). We, therefore, deduce that under normal circumstances, MC ␣v␤8 may bind type IV collagen in vivo. Only collagen IV isoforms containing ␣1 or ␣2 chains are expressed in mesangial matrix, and it has not yet been established that ␣v␤8 binds to these specific isoforms. However, mesangial matrix components have not been exhaustively identified, so native MC ␣v␤8 ligands may still be unknown. In vivo, MCs are centrally located in glomeruli and surrounded by extracellular (mesangial) matrix. Kikkawa et al. (65) recently demonstrated that MC projections also adhere to glomerular basement membrane through ␣3␤1 integrin binding to laminin ␣5 chains (65). Therefore, MC ␣v␤8 may also be spatially regulated to interact with a cadre of distinct mesangial or glomerular basement membrane matrix proteins.
An intriguing possibility is that ␤8 could regulate TGF␤ signaling, which has been implicated in myofibroblast differentiation (66). Recent reports have demonstrated that the latencyassociated peptide portion of the latent TGF␤ complex may be a ␤8 ligand (59). Latent TGF␤ is not abundantly expressed in normal glomeruli, but it is induced in animal models of glomerular disease and secreted by MCs (63,64,67). Unlike latencyassociated peptide ligation to ␣v␤6, which induces a conformational change in the integrin that leads to direct TGF␤1 activation (68), ␣v␤8-regulated TGF␤1 activity requires concomitant latency-associated peptide cleavage by the membrane-associated metalloproteinase, MT1-MMP (59). MT1- . Tat-peptide transduction efficiency range was 75-100%. Cells were fixed in paraformaldehyde, mounted with 4Ј,6-diamidino-2-phenylindole-containing media to label nuclei (blue), then incubated with Texas Red-streptavidin and counterstained with anti-␣-SMA antibodies followed by fluorescein isothiocyanate-conjugated secondary antibody. C, the percentage of MCs demonstrating ␣-SMA assembly was calculated as described under "Materials and Methods" from 20 random fields per coverslip by an observer blinded to experimental conditions. Comparisons were then made between a population of MCs that took up the Tat-Rac (17-32), Tat-Cdc42 (17-32), or HIV-Tat peptides. Data represent the mean Ϯ S.E. from four separate experiments. FIGURE 9. Implications of MC Rho family GTPase signaling by GDI. MCs harvested by microdissection from wild-type (A and C) and gdiϪ/Ϫ (B and D) mice were maintained in primary culture. Cells were changed to media containing 0% serum for 24 h before fixation in paraformaldehyde. Cells were labeled with anti-␣-SMA antibodies followed by fluorescein isothiocyanateconjugated secondary antibody (A and B) or Alexa 568-conjugated phalloidin (C and D). Representative ϫ400 fluorescence images are shown.
MMP is also inducible in MCs but generally only in pathologic conditions (69). We, therefore, speculate that in disease states, latent TGF␤ may displace the natural extracellular matrix ligand for ␤8 to initiate glomerular injury, which includes down-regulation of ␤8 expression and bioactive TGF␤ release within metalloproteinase-rich microenvironments.
Our studies represent the first characterization of a ␤8-regulated signal transduction pathway, namely activation of the small molecular weight G-protein, Rac1, and to a lesser degree, Cdc42. The canonical integrin signaling pathway is initiated by focal adhesion kinase followed by downstream activation of Rho family members, Rho, Rac, and Cdc42. However, focal adhesion kinase is not predicted to bind ␤8, and it was not identified as a ␤8 interactor by yeast two-hybrid assays. Furthermore, although other integrins have been shown to activate Rac1, including IL-2 receptor ␤1 and ␤3 integrin chimeras (39), Rac1 activation in association with integrin-GDI binding is unique to ␤8. That ␤8 stimulation activated Rac1, but not RhoA, is consistent with antagonism between Rac and Rho pathways in other systems, including ␤1 integrin-mediated adhesion to fibronectin (44). It was originally demonstrated that Rac1 can activate RhoA in fibroblasts, although weakly and with delayed kinetics (12,46). Several subsequent reports in epithelial and mesenchymal cells have shown that Rac1 downregulates RhoA activity (70 -73), which was corroborated by our data. RhoA activation induces stress fiber and focal adhesion formation (44), whereas activated Rac1 prevents these processes (44,49) through PAK-dependent inhibition of myosin heavy chain phosphorylation (50) and myosin light chain kinase activity (51). Stress fibers are predominantly composed of ␤-actin. However, MCs also express ␣-SMA, which assembles into stress fibers only during myofibroblast differentiation in pathologic states.
Taken together, the in vivo and in vitro studies support a model (Fig. 10) of constitutive MC ␣v␤8 activation by mesangial matrix or glomerular basement membrane ligands which stimulates Rac1-bound GDI shuttling to ␤8-containing microdomains to facilitate GDI release of Rac1 for membrane insertion in proximity to appropriate GEFs and other effectors. This scheme is consistent with recent reports describing spatial regulation of G-proteins to sites of activation by GDI in mammalian (16) as well as plant cells (74). Downstream, Rac1-directed signals maintain the normal differentiation state by suppression of RhoA-regulated ␣-SMA stress fiber formation. In the context of glomerular disease, MC ␤8 expression is diminished, which leads to altered Rac1 targeting, decreased Rac1 activity, and stimulation of RhoA-dependent ␣-SMA assembly. The model, therefore, predicts that the shift from Rac1 to RhoA signaling drives MC pathophysiology. The unique finding that GDI interacts with the ␤8 cytosolic tail in the context of G-protein activation is not easily rationalized with data demonstrating that G-protein-GDI interaction is unnecessary for Rac1 and Cdc42 activation (75,76) or that the sole function of GDI is to modulate G-protein activity by cytosolic sequestration. Based upon a recent report by Moissoglu et al. (77) demonstrating that GDI inhibits Rac1 membrane targeting, one unifying explanation is that ␤8 may enhance Rac1 membrane association and activation by sequestering free GDI, i.e. GDI not complexed with Rac1 (78,79). Alternatively, GDI down-regulation of Rac1 and Cdc42 may be peculiar to systems employing overexpression of constitutively active G-proteins and, therefore, may not reflect endogenous G-protein function (18,80).
In addition to the well described role of GDI as a negative regulator of G-protein activation, GDI interaction with Rac1 and Cdc42 has been associated with G-protein activation by mechanisms such as GDI-regulated inhibition of GTPase activity (15,(81)(82)(83), shielding G-proteins from protease cleavage (84) or trafficking G-proteins to appropriate membrane domains (85). To accommodate dual roles for GDI in G-protein activation and inactivation, it has been postulated that GDI dynamically regulates G-protein signaling by chaperoning G-proteins from cytosol to membrane activation domains and by removal of G-proteins from membrane sites and retention as an inactive cytosolic complex (18). Because ␤8 ligation was associated with GDI binding and Rac1 activation, we hypothesized that the integrin could acts as a GEF or GDF (86 -90). Because Rac1 was not detected in the complex with activated ␤8, our data are more consistent with ␤8 functioning as a GDF rather than a GEF.
␤8 stimulation by multiple strategies resulted in Rac1 but not RhoA activation. Because GDI binds all three classes of small molecular weight G-proteins (Rac1, RhoA, Cdc42), we speculate that if ␤8 has GDF activity, it may discriminately regulate G-protein pathways. Microinjection of fibroblasts with radixin, the first described GDF, resulted in activation of RhoA, but not Rac1 (87), thereby establishing a precedent for selective G-protein regulation by GDFs. Specificity of GDI release may also be regulated by post-translational modification and conformational changes of GDI. For example, PAK phosphorylation of Ser-101 and Ser-174 GDI residues caused release and activation of Rac1 but not RhoA (91). By analogy, ␤8-dependent PAK FIGURE 10. Schematic model of ␤8-integrin regulation of Rho family G-proteins in MC differentiation. The figure on the left represents the normal state, wherein MC differentiation is regulated by ␣v␤8-dependent docking of RhoGDI-1 with the ␤8 cytoplasmic tail, which stimulates release of Rac1 from GDI for Rac-GEF interaction. The figure on the right represents the pathologic state, modeled by either decreased ␤8 or GDI expression, resulting in up-regulated RhoA-dependent ␣-SMA assembly, which defines pathologic, myofibroblast differentiation. activation could represent an additional mechanism for preferential Rac1 activation. Finally, a recent model proposes that integrins are juxtaposed with lipid rafts containing specific G-proteins and effectors, and signal amplification is achieved by integrin prevention of lipid raft microdomain internalization (85). This study, therefore, suggests that the ␤8-GDI complex could be targeted to lipid domains, which are enriched for Rac1 rather than RhoA effectors (87,91).
Data from gdiϪ/Ϫ mice (27) support biologic relevance of the ␤8-GDI interaction in MCs. Despite ubiquitous GDI expression in normal mice, a limited number of gdiϪ/Ϫ phenotypes was observed. The most profound was renal dysfunction, which included massive proteinuria and premature death due to renal failure. Histologic examination revealed glomerulosclerosis in a mesangial distribution. MC morphology was not addressed in detail, although gdi gene deletion in other mesenchymal cells enhanced stress fiber formation (92). MC-to-myofibroblast transition, characterized by ␣-SMA stress fiber formation, is a recognized glomerular disease feature. Evidence that ␤8 and GDI may be partners within a complex that regulates MC differentiation include the following. (a) ␤8 and GDI co-precipitate, (b) kidney expression of ␤8 is most prominent in glomerular mesangium, and gdiϪ/Ϫ mice have a mesangial phenotype, and (c) itgb8Ϫ/Ϫ and gdiϪ/Ϫ MCs exhibited pathologic, myofibroblast features including enhanced RhoA activity and RhoA-dependent ␣-SMA stress fiber organization.
In conclusion, the ␤8 integrin is localized to kidney glomerular MCs in vivo and in vitro, and animal models of glomerulosclerosis are associated with decreased MC ␤8 expression. In vitro, ␤8 stimulation leads to ␤8-GDI interaction, Rac1 and Cdc42 (but not RhoA) activation, and suppression of pathologic MC features. Taken together the data suggest that under basal, physiologic conditions, the ␤8 cytosolic tail provides specificity to G-protein signaling and regulates MC phenotype, perhaps by docking and sequestering GDI, to permit Rac1 targeting to discrete signaling domains. A corollary effect of ␤8dependent Rac1 activation is concomitant RhoA suppression. Diminished MC ␤8 expression may impair Rac1 targeting and activation, thereby permitting MCs to develop a RhoA-regulated myofibroblast phenotype.