Transcription of the Transforming Growth Factor β Activating Integrin β8 Subunit Is Regulated by SP3, AP-1, and the p38 Pathway*

Integrin αvβ8 is a critical regulator of transforming growth factor β activation in vasculogenesis during development, immune regulation, and endothelial/epithelial-mesenchymal homeostasis. Recent studies have suggested roles for integrin β8 in the pathogenesis of chronic obstructive pulmonary disease, brain arteriovenous malformations, and select cancers (Araya, J., Cambier, S., Markovics, J. A., Wolters, P., Jablons, D., Hill, A., Finkbeiner, W., Jones, K., Broaddus, V. C., Sheppard, D., Barzcak, A., Xiao, Y., Erle, D. J., and Nishimura, S. L. (2007) J. Clin. Invest. 117, 3551–3562; Su, H., Kim, H., Pawlikowska, L., Kitamura, H., Shen, F., Cambier, S., Markovics, J., Lawton, M. T., Sidney, S., Bollen, A. W., Kwok, P. Y., Reichardt, L., Young, W. L., Yang, G. Y., and Nishimura, S. L. (2010) Am. J. Pathol. 176, 1018–1027; Culhane, A. C., and Quackenbush, J. (2009) Cancer Res. 69, 7480–7485; Cambier, S., Mu, D. Z., O'Connell, D., Boylen, K., Travis, W., Liu, W. H., Broaddus, V. C., and Nishimura, S. L. (2000) Cancer Res. 60, 7084–7093). Here we report the first identification and characterization of the promoter for ITGB8. We show that a SP binding site and a cyclic AMP response element (CRE) in the ITGB8 core promoter are required for its expression and that Sp1, Sp3, and several AP-1 transcription factors form a complex that binds to these sites in a p38-dependent manner. Furthermore, we demonstrate the requirement for Sp3, ATF-2, and p38 for the transcription and protein expression of integrin β8. Additionally, reduction of SP3 or inhibition of p38 blocks αvβ8-mediated transforming growth factor β activation. These results place integrin β8 expression and activity under the control of ubiquitous transcription factors in a stress-activated and pro-inflammatory pathway.

In adult tissues, increased expression of ␤8 has been shown in airway fibroblasts in lung tissue from chronic obstructive pulmonary disease patients and its expression is reduced in adenocarcinomas of the lung and in perivascular astrocytes in brain arteriovenous malformations (1,2,4). ITGB8 has been found to be increased in BRCA-1 positive breast cancers and has been identified as part of a six-gene expression signature predicting lung metastasis from breast cancer (3,22).
The mechanistic basis underlying the regulation of ␤8 expression during development and in normal and adult pathologic tissue remains largely unknown, as the genomic regions regulating ITGB8 have not been defined. Interestingly, a genetic variant in the ITGB8 locus has recently been found to correlate with risk of brain arteriovenous malformation development and the at-risk brain arteriovenous malformation variant was associated with reduced ␤8 protein expression (2). Here we describe the first characterization of the promoter and elements that regulate ITGB8 expression and place it under the control of p38, a stress-activated and proinflammatory protein.
5Ј Rapid Amplification of cDNA Ends (RACE)-RNA was extracted from U373 and 293 cell lines using the RNeasy kit (Qiagen, Valencia, CA). 5Ј RACE was performed using RNA from the above cells and human placental RNA from the SMART TM RACE kit (Clontech, Mountain View, CA) according to the manufacturer's protocol. The ITGB8-specific primer used for second-strand synthesis and subsequent re-amplifications of fragments was GSP2j3 (5Ј-tgttgctggcatcccgagcccgagcttcctcccttgcc-3Ј). Amplified fragments were cloned using the TOPO-TA cloning kit (Invitrogen) and sequenced.
Transcription factor binding site mutant reporter constructs were made by site-directed PCR mutagenesis using the high fidelity Phusion DNA polymerase (Finnzymes USA, Woburn, MA) from Ϫ1280/ϩ69 as template. A fragment spanning from the origin to the AflII site was amplified using the primers, 5Ј-ggtaccgagctcttacgcgtg-3Ј and 5Ј-tttgatcttaagGctAaGcttcccagtaaacggaacaaaaagttacag-3Ј, which mutated the upstream putative CRE site to a HindIII site (mutations indicated by uppercase), then cloned into Ϫ1280/ϩ69 to create CRE (Ϫ41) mt. DMRT5mt was created by amplifying a fragment between the AflII and BspEI sites using primers, 5Ј-cgcgggcttaagatcaaaagac-ccactAtGCAttgcaaaagccc-3Ј and 5Ј-tcccggagcgactggccgagatttctcgctg-3Ј, which mutated the putative DMRT5 site to a NsiI site, then cloned into Ϫ1280/ϩ69. E2Fmt was created by a similar amplification of the same region and cloning as for DMRT5mt using AflII and BspEI, but using the primers, 5Ј-gggcgccgcttaagatcaaaagacccactgtaacttgcaaaagc-3Ј and 5Ј-tcccggagcgactggccgagattActAgTtgctcc-3Ј, which mutated the putative E2F site to a SpeI site. NFBmt was created by amplifying a fragment between the BspE1 and BstAP1 sites in Ϫ1280/ ϩ69 using primers 5Ј-gagaaatctcggccagtcgctccggaaacagcccctg-3Ј and 5Ј-ccccgcagctgctgcaGacaaagtggagtcaagggacctcGActgcccccaacgccc-3Ј, which mutated the putative NFB/c-Rel site and created a PstI site upstream for identification purposes by changing a single nucleotide, then cloned into Ϫ1280/ϩ69. ⌬AP4 was created by a single cleavage of the BstAP1 site, Klenow-filled, then re-ligated, resulting in the deletion of "cagc" from the putative AP4 site. The downstream CRE (Ϫ626) mt was created by amplifying a fragment between the two SacII sites in Ϫ1280/ϩ69 using primers, 5Ј-ttaaccgcggATatctcatgcctcaccaatgtcccgcccacgctgct-3Ј and 5Ј-atataccgcgggggttcctgctcagcaggcgcagggcagcctctgtcat-3Ј, which mutated this putative CRE site to an EcoRV site, then cloned into Ϫ1280/ϩ69. The SPmt was created using this same method as for CRE (626) mt, but using an alternative forward primer, 5Ј-atatccgcggtgacttcatgcct-caccaatgtcGAATTcacgctgctccgagctgt-3Ј, which mutated the putative SP site to an EcoRI site. All constructs were confirmed by sequencing.
Reporter Assays-SEAP reporter constructs were transfected into all cell types using the Lipofectamine 2000 reagent (Invitrogen). A Renilla luciferase reporter construct (Promega, Madison, WI) driven by the TK1 (thymidine kinase) promoter was co-transfected at a 1:50 molar ratio to the SEAP constructs to normalize for transfection efficiency. The medium was replaced on the cells 24 h after transfection then at 48 h both reporter assays were performed. SEAP (Phospha-Light detection system, Applied Biosystems, Foster City, CA) and luciferase (Renilla luciferase assay kit, Promega) were detected using the manufacturer's protocols. Raw data were normalized by subtracting the background signal from cells treated with Lipofectamine 2000 alone, then the resulting SEAP signal was divided by the Renilla luciferase signal. Each experiment was normalized to each other by dividing by the activity of pSEAP basic, which does not possess a promoter driving SEAP expression.
Nuclear Extraction and Electromobility Shift Assays (EMSA)-Nuclear extracts were obtained using the NE-PER kit (Pierce) with protease inhibitors (Protease Inhibitor Mixture Set 1, Calbiochem) and phosphatase inhibitors, 1 mM Na 3 VO 4 and 10 mM NaF. Prior to the final step of the kit protocol, the nuclear extracts were subjected to three flash freeze-thaw cycles.
EMSAs were performed using the non-radioactive, Light-Shift Chemiluminescent EMSA kit (Pierce) with slight modifications. 50 g of nuclear extracts were first incubated with competitor probes (100 fmol) and/or antibodies (0.8 -6 g) in binding buffer (32) including 10 g of bovine serum albumin for 45 min at room temperature. Then the wild type, biotinylated probe was added to the reactions (1 fmol) and incubated for an additional 15 min at room temperature. The reactions were subsequently electrophoresed through a 4% polyacrylamide, non-denaturing gel. The rest of the assay was performed according to the manufacturer's protocol (Pierce).
Plasmid and siRNA Transfections and RT-PCR-Control (catalog numberAM4641), SP1 (ID number 143158), and SP3 (ID number 115338) siRNAs were purchased from Ambion, Inc. (Applied Biosystems). The ATF2 (sc-29205) siRNA and control (sc-37007) were purchased from Santa Cruz Biotechnology, Inc. Additional siRNAs against SP3 (J-023096-09), ATF2 (J-009871-06), and Control (D-001810-10) were purchased from Dharmacon, Inc. (Lafayette, CO). These siRNAs were transfected into primary lung fibroblasts using the electroporation technique by Amaxa Nucleofector II (Amaxa Biosystems, Lonza). For transcript analysis, medium was replaced after 24 h and RNA was extracted at 48 h post-transfection. RNA extraction and cDNA synthesis were performed using the RNeasy and Quantitect kits (Qiagen), respectively. For surface protein extraction, medium was replaced after 24 h, then after 72 h the cells were stained for integrin ␤8 and analyzed by flow cytometry. The p38␣DN and pcDNA constructs (gifts from Jiahuai Han, The Scripps Research Institute, La Jolla, CA) were transfected into lung fibroblasts using Amaxa Nucleofector II as described above. RNA was extracted from these cells 24 h post-transfection.
Statistical Analysis-All statistics performed were by comparing two sets of values using the Students' t test in the graphing and statistical analysis software program, Prism TM 4 (GraphPad Software, Inc., La Jolla, CA); * ϭ p Յ 0.05; ** ϭ p Յ 0.01; *** ϭ p Յ 0.001.

Characterization of the Transcriptional Start Site of ITGB8-
To determine the transcriptional start site (TSS) of ITGB8, the 5Ј end of the transcript was mapped by 5Ј RACE (Fig.  1A). These experiments indicate that the four most 5Ј TSS are Ϫ410, Ϫ404, Ϫ236, and Ϫ217, from the published TSS for ITGB8 (located on the plus strand of human chromosome 7 at nucleotide 20,337,250) (38). This result extends the published 5Ј-untranslated region to a total length of 1,115 nucleotides.
Sequence analysis indicates that no TATA box is present in 1,213 bp of the 5Ј-flanking sequence to the most 5Ј TSS. A CCAAT site is located 28 bp upstream of the most 5Ј TSS. This newly identified TSS includes a close match to the consensus initiator (INR) element YYANWYY, containing the critical adenosine as the ϩ1 nucleotide (39,40). Hereto forward, the newly identified TSS will be designated as ϩ1. ESTs (DA238257.1 and DA994035.1) representing 5Ј oligocapped mRNA sequences from a 5Ј end-enriched cDNA library from brain tissue show transcripts that map ϩ14 and ϩ43 relative to our newly identified TSS, respectively.
Characterization of the ITGB8 Core Promoter-In silico analysis of sequences 5Ј of the TSS for ITGB8 identified two overlapping promoter regions, one from Ϫ779 to Ϫ635 (McPromoter (41)) and one from Ϫ725 to ϩ322 (Dragon Genome Explorer (42)) ( Fig. 1A). Two separate CpG islands, which are commonly found in promoters, also overlap with these predicted promoter regions (Ϫ882 to Ϫ625; Ϫ311 to ϩ1115) (UCSC Genome Browser (43)).
The Core Promoter of the ITGB8 Gene Is from Ϫ493 to ϩ69 bp Relative to the TSS-Promoter reporter assays were performed using deletion constructs containing 5Ј sequences from Ϫ1280 to ϩ69 relative to the ITGB8 TSS ( Fig. 2A). Cell lines derived FIGURE 1. Characterization of the 5 TSS of the ITGB8 gene with surrounding sequence analysis. A, location of the TSS and other sequence features 5Ј of the ITGB8 gene. The published ITGB8 5Ј untranslated region is indicated in capital letters. The start codon is labeled in bold with an arrow above. The 5Ј RACE results from the TSS are labeled in bold and underlined. The most 5Ј of these results is labeled as ϩ1 (above), determining the numbering for the presented sequence. Predicted CpG islands are represented in italics. The region of significant homology across 28 mammalian species is underlined. The predicted, overlapping promoter regions are highlighted in light gray. Putative transcription factor binding sites are boxed and labeled above. B, homology of putative transcription factor binding sites of interest in the sequences 5Ј of the ITGB8 gene across eight placental mammalian species. Consensus binding sites for the transcription factors are indicated above, the E2F consensus includes both strands in the 5Ј-3Ј orientation. Homologous bases are highlighted in gray.
from cell types that normally express integrin ␣v␤8 initially used for promoter assays were U373 (glioblastoma), H1264 (lung carcinoma), and HeLa (cervical carcinoma) (data not shown) (1). All three cell lines demonstrated expression of the longest construct, indicating that the Ϫ1280 to ϩ69 region contains promoter activity ( Fig. 2A). This construct was also expressed in cells that normally do not express integrin ␤8, human microvascular endothelial cells-1, or primary dermal fibroblasts, suggesting that this region does not contain elements for cell type-specific expression of integrin ␤8 (data not shown). In H1264, U373, and HeLa cells, the 5Ј-deletion construct encompassing Ϫ491 to ϩ69 had similar activity as the full-length construct, demonstrating that this region contains the core ITGB8 promoter ( Fig. 2A). In HeLa cells, the construct Ϫ330 to ϩ69 maintained the same low level of activity as the longer 5Ј-deletion constructs but showed no activity in H1264 or U373 cells, indicating that this region of the core promoter has activity in specific cell types ( Fig. 2A).
When the activities of four constructs, Ϫ623/ϩ69, Ϫ491/ ϩ69, Ϫ623/Ϫ336, and Ϫ330/ϩ69, were tested in primary cells that normally express ␣v␤8, HBEC, adult lung fibroblasts, and fetal astrocytes, the core promoter activity of construct, Ϫ491/ ϩ69, was confirmed (Fig. 2C). The smaller 5Ј-deletion construct, Ϫ330/ϩ69, maintained high expression in fetal astrocytes, supporting that this smaller segment of the core promoter is sufficient to drive expression in specific cell types (Fig. 2C).
The CRE and SP Binding Site in the ITGB8 Core Promoter Are Required to Drive Expression-There are four putative transcription factor binding sites with high degrees of interspecies conservation within the ITGB8 core promoter, NFB/c-Rel, AP4, CRE, and SP (Fig.  1B). We made reporter constructs with mutations in each of these sites (Fig. 3A). Mutations in the CRE (Ϫ41) or SP binding site significantly reduced or completely abrogated expression in both cell lines and primary cells (Fig. 3, A and B). Mutation in the NFB/c-Rel site or deletion of the AP4 site, individually or combined, did not alter reporter expression (Fig. 3, A and B, data not shown). We also performed mutational analysis of CRE (Ϫ626) , DMRT5, and E2F putative transcription factor binding sites upstream of the core promoter and were predicted by in silico analysis (Fig. 1). Mutations in the CRE (Ϫ626) , DMRT5, or E2F binding sites did not significantly change reporter expression (Fig. 3). These data indicate that the CRE (Ϫ41) and SP binding site located in the core promoter (Ϫ491/ϩ69) are required for ITGB8 promoter activity.
Specific Complexes Form at the CRE and SP Sites in the Core Promoter of ITGB8-EMSAs using a probe corresponding to the sequence containing the CRE, CCAAT, and SP binding sites in the ITGB8 core promoter were performed in nuclear extracts from primary adult lung fibroblasts (Fig. 4A). Similar binding patterns were observed in nuclear extracts from primary HBEC (data not shown). At least six migrating bands were identified with this probe (Fig. 4A, lane 1). We determined that the slowest migrating five bands are specific protein-DNA complexes, because the wild type competitor completely displaced all five bands (Fig. 4A, lane 2, and Fig. 5B, lane 2). A competitor mutated in the CRE and CCAAT binding sites displaced most of the complexes from bands 1, 2, and all from band 5, but did not disrupt complexes in bands 3 and 4 (Fig. 4A, lane 3). In contrast, a competitor with a wild type consensus CRE (32) displaced bands 3 and 4, but not bands 1, 2, and 5 (Fig. 4A, lane  4). Similarly, a competitor mutated in the SP site also displaced bands 3 and 4, but not bands 1, 2, and 5 (Fig. 4A, lane 5). A competitor mutated for all three sites, CRE, CCAAT, and SP did not displace any bands, confirming that complexes in bands 1-5 are specific to these sites (Fig. 4A, lane 6). These results demonstrate that specific complexes form at the CRE site Integrin ␤8 Expression Is Regulated by Sp3, AP-1, and p38 AUGUST 6, 2010 • VOLUME 285 • NUMBER 32 (bands 3 and 4) and SP site (bands 1, 2, and 5) in the ITGB8 core promoter.
Sp1 and Sp3 Bind to the Putative SP Binding Site in the Core Promoter of ITGB8-The most abundant DNA binding complexes in the EMSAs using the CRE/CCAAT/SP probe are bands 1 and 2, which are specific to the SP site. Antibodies against Sp1 and Sp3 were used to supershift these complexes (Fig. 4B). Sp1 is contained in the complex in band 1 as shown by its supershift using antibodies against Sp1 (Fig. 4B, lane 2). Sp3 is contained in complexes in bands 2 and 5 as shown by their supershifts using antibodies against Sp3 (Fig. 4B, lane 3). Three bands, 1, 2, and 5, were supershifted using antibodies against both Sp1 and Sp3 together, confirming that these complexes contain Sp1 and Sp3 (Fig. 4B, lane 4). These data show that Sp1 and Sp3 from nuclear extracts of lung fibroblasts can bind to the SP binding site in the core promoter of ITGB8.
Sp3 Is Required for Basal Expression of ITGB8-Primary human adult lung fibroblasts were treated with siRNAs against SP1, SP3, or a control, and analyzed by quantitative RT-PCR for integrin ␤8 transcript levels. As shown in Fig. 4C, significant reductions of SP1 or SP3 transcripts were obtained by treatment with their respective siRNAs. For cells treated with siRNA against SP1, SP3 transcript levels were significantly increased, whereas treatment with siRNA against SP3 did not alter SP1 levels (Fig. 4C, left and middle panels). Although siRNA against SP1 showed a trend toward reduction in ITGB8 transcript levels, only siRNA against SP3 significantly reduced ITGB8 levels (Fig. 4C, right panel). This result was confirmed using a second siRNA against SP3, which resulted in a 61% reduction in SP3 transcript levels (p ϭ 0.001) and a 31.5% reduction in ITGB8 transcript levels (p Ͻ 0.0001). This reduction in ITGB8 transcript levels by siRNA against SP3 was also observed at the protein level as determined by flow cytometry (Fig.  4D). These data show that Sp3 is required for basal expression of integrin ␤8 in primary adult lung fibroblasts.
Sp3 Is Required for Integrin ␣v␤8mediated TGF-␤ Activation by Lung Fibroblasts-Primary adult lung fibroblasts were transfected with siRNAs against SP3 or control, then subjected to a TGF-␤ activation assay using blocking antibodies against integrin ␤8 (Fig. 4E). As shown in Fig. 4E, TGF-␤ activation was reduced by 59% in the presence of blocking antibodies against integrin ␤8 using control siRNAtransfected cells. Total TGF-␤ activation was significantly reduced by 72% in cells transfected with siRNA against SP3. Addition of integrin ␤8 blocking antibodies had only a negligible effect on SP3 siRNA-transfected cells. These results indicate that Sp3 is required for transcriptional regulation of integrin ␤8 expression, which in turn regulates ␣v␤8-mediated TGF-␤ activation by lung fibroblasts.
AP-1 Complexes Bind to the Putative CRE Binding Site in the Core Promoter of ITGB8-EMSA competition experiments using the CRE/CCAAT mutant competitor with the CRE/ CCAAT/SP probe revealed two faint bands that co-migrate with bands 1 and 2, suggesting that these bands represent complexes that are also specific to CRE or CCAAT, labeled 1Ј and 2Ј (Fig. 5A, lane 2). Thus, supershift analysis for CRE binding and AP-1 transcription factors was performed in the presence of the CRE/CCAAT mutant competitor for better visualization of bands 1Ј and 2Ј (Fig. 5A, lanes 2-6). Antibodies against either ATF-2 or c-Jun displaced band 1Ј and partially displaced bands 3 and 4 (Fig. 5A, lanes 4 and 6). We hypothesize that the antibodies are binding to their protein targets in such a way that it causes their dissociation from the DNA, although it is also possible that these complexes are being supershifted, but are not detected in this non-radioactive EMSA. Longer exposures show a very faint, supershifted band in the lane with the extract con- taining c-Jun antibodies (data not shown). These data suggest that an AP-1 complex containing ATF-2 and c-Jun in nuclear extracts of lung fibroblasts bind to the CRE site in the core promoter of ITGB8.
Other AP-1 transcription factors were analyzed for their ability to bind to the CRE site by EMSA supershift analysis. Antibodies against JunD and JunB showed a significant reduction in the intensity of DNA-binding complexes in bands 3 and 4 (Fig. 5A, lanes 10 and 11). However, antibodies against c-Fos, Fra-1, and Fra-2 did not cause differences in the binding pattern (Fig.  5A, lanes 12-14). A slower migrating complex could be seen at longer exposures in some supershift experiments using antibodies against JunD or JunB (data not shown). In summary, these data suggest that AP-1 complexes containing JunD and JunB from nuclear extracts of lung fibroblasts bind to the CRE site in the core promoter of ITGB8.
ITGB8 Expression Is Regulated by ATF-2 and p38-ATF2 siRNA significantly reduced ATF2 levels (Fig.  5B), which led to the reduction of ITGB8 transcript levels by ϳ45% (Fig. 5B). A second siRNA against ATF2 confirmed this result, with a 45% reduction in ATF2 transcript levels (p ϭ 0.0048) and a 36% reduction in ITGB8 transcript levels (p Ͻ 0.0001). The reduction in ITGB8 transcript levels was functionally significant because a 40% reduction in integrin ␤8 protein levels was also observed in cells treated with siRNAs against ATF2 (Fig. 5C). These results suggest a role for an AP-1 complex containing ATF-2 in regulating integrin ␤8 expression.
AP-1 is primarily regulated by the MAPK signaling pathways (46,47). The MAPK pathways are classically divided into the ERK, JNK, and p38 pathways (48). To determine which MAPK signaling pathway may be involved in the regulation of integrin ␤8 expression through AP-1, primary adult lung fibroblasts were treated with chemical inhibitors to each of these pathways then analyzed for surface expression of integrin ␤8 using flow cytometry (Fig. 6A). Only the p38 inhibitor, SB202190, significantly reduced integrin ␤8 surface expression (Fig. 6A). A similar reduction by this inhibitor was also seen at the transcript level (Fig.  6B). SB202190 inhibited phosphorylation of a canonical target of p38, HSP 27, and also inhibited the dual phosphorylation of ATF-2, another known target (49) (Fig. 6C). These data suggest a role for p38 in regulating ITGB8 expression through activation of ATF-2 in lung fibroblasts.
Integrin ␣v␤8-mediated TGF-␤ Activation by Lung Fibroblasts Is Regulated by the p38 Pathway-Primary adult lung fibroblasts were treated with SB202190 and subjected to a TGF-␤ activation assay (Fig. 6D). As shown in Fig. 6D, treat-FIGURE 4. Sp3 is required for ITGB8 expression and ␣v␤8-mediated TGF-␤ activation. A, EMSA in adult lung fibroblasts using a biotinylated probe (sequence indicated above) covering the CRE, CCAAT, and SP transcription factor binding sites in the core promoter with or without the indicated unlabeled competitors. Wt is the unmutated competitor. CRE/CCAAT mt, SP, or triple mt indicates mutations in each or all of the indicated sites. CRE wt is an unrelated CRE consensus site competitor (32). The figure is a composite from a single representative gel. B, EMSA supershift analysis using polyclonal antibodies against Sp1, Sp3, or Sp1 and Sp3 combined. Supershifted complexes are indicated by arrows. The figure is a composite from a single representative gel. C, quantitative RT-PCR results for SP1, SP3, and the ITGB8 in adult lung fibroblasts transfected with siRNA against SP1, SP3, or control (Ϯ S.E.). The measured transcript is labeled above each respective graph. D, flow cytometry for surface expression of the integrin ␤8 subunit on adult lung fibroblasts transfected with siRNA against SP3 or control (Ϯ S.E.). MFI, mean fluorescence intensity. E, TGF-␤ activation assays of adult lung fibroblasts treated with siRNA against SP3 or control using control or anti-␤8 blocking antibodies (Ϯ S.E.). * ϭ p Յ 0.05; ** ϭ p Յ 0.01; *** ϭ p Յ 0.001. ment with either the blocking antibody against ␣v␤8 or SB202190 significantly reduced TGF-␤ activation by 74 and 95%, respectively. Although the p38 inhibitor essentially abolished all of the TGF-␤ activation by these cells, this effect was not significantly different from the effect of integrin ␤8 blocking antibodies. To determine the specificity of the inhibitor, a plasmid expressing a dominant-negative isoform of p38␣ (p38␣DN) was transfected into lung fibroblasts (50). Overexpression of p38␣DN (Fig. 6E, left panel) caused a significant decrease in ITGB8 levels by ϳ35% (Fig. 6E, right panel). Furthermore, overexpression of p38␣DN also caused a 65% decrease in ␣v␤8-mediated TGF-␤ activation (Fig. 6F). These data suggest that p38␣, through the regulation of integrin ␤8 expression, regulates ␣v␤8-mediated TGF-␤ activation by lung fibroblasts.
Interaction of an AP-1 Complex Containing ATF-2, c-Jun, and Sp3 with the ITGB8 Promoter Requires p38 Signaling-Chromatin immunoprecipitation assays using antibodies specific for the phosphorylated form of ATF-2 (pATF-2) or all forms of ATF-2, c-Jun, and Sp3 were performed on adult lung fibroblasts treated with the p38 inhibitor, SB202190, to assess their association with the ITGB8 promoter (Fig. 7A). Chromatin immunoprecipitates (pATF-2, ATF-2, c-Jun, or Sp3) from untreated cells increased PCR amplicons from the core promoter regions, P1 and P2, but not from P3, which is outside of the ITGB8 core promoter (Fig. 7A). Chromatin immunoprecipitates (pATF-2, ATF-2, or c-Jun) of fibroblasts treated with SB202190 had a reduced PCR signal, relative to untreated fibroblasts, at these ITGB8 promoter sites (Fig. 7A). These results demonstrate that ATF-2, c-Jun, and Sp3 are associated with the ITGB8 core promoter and form a higher order complex that is dependent on p38 signaling.

DISCUSSION
Integrin ␤8 Expression and ␣v␤8-mediated TGF-␤ Activation Are Regulated by Sp3, AP-1, and the p38 Pathway-Here we report the first identification and characterization of the promoter for human ITGB8, a gene that is crucial for the regulation of TGF-␤ function in vivo. We identify and characterize two essential cis-acting promoter elements, one that binds Sp1 and Sp3 and one that binds an AP-1 transcriptional complex containing ATF-2. Both Sp3 and ATF-2 are required for efficient ITGB8 transcription and protein expression. We show that the regulation of ITGB8 expression is dependent on p38, a MAPK upstream of ATF-2. Finally, we demonstrate that Sp3 or p38, through regulation of integrin ␤8 expression, regulates ␣v␤8-mediated TGF-␤ activation. Altogether, these results suggest a mechanistic basis for the regulation of ␣v␤8-mediated TGF-␤ activation by stress-mediated and pro-inflammatory signaling pathways.
The Identification of the ITGB8 Core Promoter-The ITGB8 core promoter contains, in close proximity, a CCAAT, CRE, and SP binding site, and an INR consensus. These elements are found commonly in eukaryotic core promoters usually residing within 100 bp of the TSS (51,52). Indeed, the ITGB8 core promoter (Ϫ491 to ϩ69 relative to the TSS), which contains the minimal sequence required to drive gene expression in all cell lines in this study, contains all of these elements. However, other yet unidentified regulatory elements reside in the sequence between Ϫ330 and Ϫ491 because a smaller core promoter construct (Ϫ330 to ϩ69), which contains the CCAAT, CRE, SP1, and INR, drives efficient expression only in specific cell types. Thus, cell type-specific regulatory elements are required for efficient core promoter activity in the region between Ϫ330 and Ϫ491. We tested putative NFB or AP4 binding sites, which are found in the Ϫ330 to Ϫ491 sequence, singularly or combined, and found neither was required for core promoter activity. However, we cannot exclude the possibility that either site, or both, in combination with another currently unidentified site within the Ϫ330 to Ϫ491 region, is required. Transcription factor matrices do not reveal additional highly conserved consensus transcription factor binding sites in the Ϫ330 to Ϫ491 region and thus, the exact functional sequence(s) and cognate transcription factor(s) remain to be elucidated.
The Sp3 Transcription Factor Regulates ITGB8 Expression-The SP consensus sequence located at Ϫ19 in the core promoter is required for activity, because a deletion mutant at this site completely abolishes the activity of the promoter construct and knock-down of SP1 and SP3 reduce expression of ␤8 by 29 and 62%, respectively. Although Sp1 and Sp3 are thought to be functionally redundant, SP3 knock-down is more effective in reducing ␤8 expression than SP1, possibly due to a compensatory increase in SP3 expression with SP1 knock-down, which likely masks any effect on ␤8 expression.
The SP family of transcription factors is highly conserved, consists of Sp1-9, and binds with high selectivity to the SP consensus binding sequence, the GC box (53). Sp1-4 form a subgroup of the SP family that contain homologous activation domains and C-terminal zinc finger regions, which mediate DNA binding specificity (54). We limited our investigation to Sp1 and Sp3 because they are expressed in tissues where ␤8 is known to be expressed and have been shown to regulate expression of a wide array of integrin genes (55)(56)(57)(58)(59)(60)(61)(62)(63)(64)(65). Their functions partially overlap as demonstrated by their knock-out phenotypes in mice (66 -68). Sp1 and Sp3 are required for survival and are generally involved in cellular differentiation, proliferation, and organogenesis (69).
An AP1 Complex Is Required for ITGB8 Core Promoter Activity-The CRE site located at Ϫ40 is also required for ITGB8 core promoter function. Many different transcription factors and complexes are capable of interacting with CRE, such as CRE-binding protein (CREB), CREBbinding partner (CBP), and a multitude of AP-1 factors (70). AP-1 is a transcription factor complex typically composed of a homo-or heterodimer of a member of the Jun and Fos or ATF transcription factor families including c-Jun, JunB, JunD, c-Fos, FosB, Fra-1, Fra-2, ATF-1-4, and JDP (Jun dimerization partner) (46). In fact, several different AP-1 factors such as ATF-2, c-Jun, JunD, and JunB bound to the CRE site in the ITGB8 promoter.
ATF-2 bound to the consensus CRE site in the core promoter of ITGB8 in a complex with c-Jun, probably in a heterodimer, because antibodies against ATF-2 and c-Jun supershifted the same complex in EMSA and both ATF-2 and c-Jun bound to the core promoter region in chromatin immunoprecipitations. Additionally, some, but not all, of the components of one of the ATF-2 complexes could be supershifted with antibodies to JunB and JunD, suggesting that a multimeric AP-1 complex can form around the CRE on the ITGB8 core promoter. However, it is likely that the functional complex at the CRE is one that contains ATF-2 and c-Jun, because Jun-Jun dimers have a lower affinity for CRE than ATFs/CREBs, whereas c-Jun is a potent transcriptional activator (47).
We confirmed that ATF-2 regulates integrin ␤8 expression because knock-down of ATF2 caused a significant reduction in integrin ␤8 levels. Both transcript and surface levels were reduced by ϳ40 -45%, which could either be due to residual ATF-2 protein levels after ATF2 knock-down, incomplete knock-down of ATF2, and/or compensation by other AP-1 factors in the absence of ATF-2.
ATF-2 is part of a subfamily of AP-1 transcription factors, the ATF/CREB family, which contains a homologous DNA binding domain that mediates interaction with consensus CRE or 12-O-tetradecanoylphorbol-13-acetate-response elements (TRE) and are regulated by cAMP-dependent and stress-activated kinase pathways (70,71). ATF-2 is the most extensively characterized and is ubiquitously expressed (71). ATF2 Ϫ/Ϫ mice die either perinatally due to severe respiratory distress due to deficient pulmonary differentiation, or succumb later to defective immune responses to microbial infection (72,73). These data . MFI, mean fluorescence intensity. B, quantitative RT-PCR results for ITGB8 expression in adult lung fibroblasts treated with SB202190, normalized to GAPDH and ␤-actin and relative to control (Ϯ S.E.). C, immunoblot for phosphorylated HSP 27 and dual-phosphorylated ATF-2 from nuclear extracts from adult lung fibroblasts treated Ϯ SB202190. Immunoblot for the nuclear localized proteins, lamins A and C, was used as a loading control. D, TGF-␤ activation assays of adult lung fibroblasts treated with anti-␤8 blocking antibodies or SB202190 (Ϯ S.E.). E, quantitative RT-PCR results for MAPK14 (p38␣) and ITGB8 in adult lung fibroblasts transfected with plasmids expressing a p38␣ dominant-negative isoform (p38␣DN) or the empty vector control, pcDNA (Ϯ S.E.). The measured transcript is labeled above each respective graph. F, TGF-␤ activation assays of adult lung fibroblasts transfected with plasmids expressing a p38␣ dominant-negative isoform (p38␣DN) or the empty vector control, pcDNA. Percentage (%) of ␣v␤8-mediated TGF-␤ activation shown (Ϯ S.E.). * ϭ p Յ 0.05; ** ϭ p Յ 0.01; *** ϭ p Յ 0.001. and the discovery that ATF-2 regulates ITGB8 expression also support a role for integrin ␤8 in the lung and the immune response. ATF2 is mapped to a region on chromosome 2q that is subject to loss of heterozygosity in multiple cancer types and specific alleles of ATF2 are linked to a subset of lung cancers suggesting that ATF-2 may operate as a tumor suppressor in these cancers (74). We have previously shown that integrin ␤8 behaves as a tumor suppressor in lung cancer cells in vitro and in vivo (4). Therefore, ATF-2 may act as a tumor suppressor in lung cancers at least in part via regulation of integrin ␤8 expression.
Sp1/Sp3 and AP-1 are often found as components in large multimeric transcription factor complexes that can interact with each other (75,76). The close proximity of the SP and CRE consensus sequences in the ITGB8 promoter suggest these complexes may interact to regulate ␤8 expression. Although we did not identify any complexes that contained both Sp1 or Sp3 and AP-1 factors by EMSA analysis, it remains plausible that these adjacent complexes interact to drive expression of integrin ␤8, but remain either transient or with insufficient affinity to be detected by EMSA. In fact, the results from the chromatin immunoprecipitation experiments suggest that these transcription factors do interact in a higher order chromatin com-plex at the ITGB8 core promoter because they associate with the same regions of DNA.
The p38 Pathway Regulates ␤8 Expression-We demonstrate that ␤8 expression is regulated by p38 and that ␣v␤8-mediated TGF-␤ activation is p38-dependent, linking, for the first time, the p38 pathway to the regulation of TGF-␤ activation through integrin ␤8 expression. In ␤8 expressing cells, ATF-2, a known target of the p38 MAPK pathway, is phosphorylated, and therefore activated, in a p38dependent manner. Phosphorylated ATF-2 and c-Jun both bind to the ITGB8 core promoter in a p38-dependent manner. In contrast, Sp3 binds to its cognate site in the ITGB8 core promoter in a p38independent manner.
Taken together, we hypothesize a model for the regulation of integrin ␤8 expression through p38, phospho-ATF-2 and Sp3 (Fig. 7B). By EMSA, we showed that Sp3, ATF-2, and c-Jun directly bind to their cognate sites in the core promoter and that these sites are required for promoter expression. We showed that both the "P1" region, which contains these sites, and an adjacent region, "P2," of the ITGB8 promoter are required for full promoter expression as determined by reporter assays. However, we were unable to determine which putative transcription factor binding sites within the P2 region were required for promoter expression. As expected, chromatin immunoprecipitation of Sp3, ATF-2, or c-Jun resulted in their enrichment at the P1 and P2 regions. This enrichment at the P2 region was unexpected because we were unable to show direct binding of Sp3, ATF-2, or c-Jun to sequences in the P2 region by EMSA. This result could either be due to larger DNA fragments that span the directly adjacent "P1 and P2" regions present in the nuclear sonicates or to a direct or indirect association of Sp3, ATF-2, or c-Jun with the P2 region. Evidence in support of the later is that Sp3 dissociates with P2 but maintains its association with the P1 region in cells treated with the p38 inhibitor. . ATF-2, c-Jun, and Sp3 association with the ITGB8 core promoter requires p38 signaling. A, chromatin immunoprecipitation of ATF-2, c-Jun, or Sp3 in adult lung fibroblasts treated with the p38 inhibitor, SB202190, with PCR amplification of regions from the ITGB8 promoter (Ϯ S.E.). Genomic locations of the amplicated regions are indicated in the schematic below the graphs. Essentially identical results were obtained using antibodies against phospho-ATF-2 and total ATF-2 and thus, results were pooled. B, hypothetical model for regulation of the ITGB8 promoter by p38, ATF-2, c-Jun, and Sp3. p38 phosphorylates ATF-2, which heterodimerizes with c-Jun to bind to the CRE site in the ITGB8 core promoter. Along with Sp3, which is already bound to its cognate SP site adjacent to the CRE site in the ITGB8 core promoter, these transcription factors form a higher-order chromatin complex that interacts either directly with the P2 region or indirectly through yet unidentified transcription factors to activate transcription of ITGB8. * ϭ p Յ 0.05.
Therefore we hypothesize a model where p38 phosphorylates ATF-2, which binds to c-Jun, and facilitates interaction with the CRE site in the ITGB8 P1 region of the core promoter, which is in close proximity to Sp3 constitutively bound to its SP site. A higher-order chromatin complex forms containing Sp3 and AP-1 that interacts with an adjacent (P2) region in the core promoter either directly or through as of yet unknown factors, which activates the basal transcriptional machinery and transcription of the ITGB8 gene.
The p38 pathway is activated by a wide range of cellular stresses and inflammatory cytokines and also regulates the expression of many pro-inflammatory cytokines (49). Hence, p38 is critical for normal immune and inflammatory responses. Many of the studies defining p38 function rely on relatively specific p38 pyridinylimidazole inhibitors such as SB202190 (80). There are four mammalian isoforms of p38, and these compounds selectively inhibit both p38␣ and p38␤ isoforms, but not p38␥ or p38␦ (81). Knock-out studies indicate a dominant role for p38␣ in vivo (82,83). Thus, p38␣ is the most likely isoform to be involved in regulation of ␤8 expression, which we confirmed using a dominant-negative isoform of p38␣ and causing a reduction in ITGB8 expression by ϳ35%. This modest reduction in ITGB8 levels is similar to the reduction caused by knock-down of ATF-2, suggesting that p38 has its effect primarily through its activation of ATF-2. Although we cannot exclude the involvement of the other p38 isoforms in regulating ITGB8 expression, p38␤ is the only other isoform that is reportedly inhibited by SB202190, suggesting that it might also contribute to the regulation of ITGB8 (81).
TGF-␤ is a crucial mediator of immune and epithelial-mesenchymal homeostasis through diverse effects on cellular differentiation, proliferation, and apoptosis (23). Previous studies have demonstrated important roles for AP-1, p38, and TGF-␤ in normal epithelial-mesenchymal homeostasis and wound healing of the skin and lung (23,47,84,85). We have also shown that integrin ␤8 regulates epithelial-mesenchymal and endothelial-mesenchymal homeostasis in the airway and brain, respectively (1,2,15,17). Here we show a direct link between p38, AP-1, and TGF-␤ in these processes via regulation of integrin ␤8 expression.