Functional interactions between Wnt signaling and Runx2

Based on implications for complex Wnt pathway responses in differentiating osteoblasts, we first asked whether changes in the activity or expression of the obligate osteoblast transcription factor Runx2 influenced Wnt signaling. By itself, prostaglandin (PG)E₂, which rapidly activates endogenous Runx2, modestly increased basal TCF/LEF-dependent gene expression, but it potently enhanced the stimulatory effect of gene induction by Wnt1 (Fig. 2A​2A),), or by LiCl and WAg (Fig. 2B​2B).). Notably, whereas activation of the TCF/LEF-sensitive promoter was significantly inhibited when endogenous Runx2 was suppressed by transgenic antisense expression (Fig. 2B​2B,, upper panel), this was particularly striking in PGE₂-activated cells (Fig. 2B​2B,, lower panel). Interaction with the Wnt pathway was Runx2 restricted, at least in part, in osteoblasts because TCF/LEF-dependent gene expression was unaltered by C/EBPδ antisense (Fig. 2C​2C,, left panel), and Runx2 antisense had no effect on TCF/LEF-dependent gene expression in fibroblasts (Fig. 2C​2C,, right panel) that express little or no Runx2 (29). Appropriately, Runx2 antisense significantly suppressed gene expression by plasmid SXN1c synthesized to contain two Runx-binding sites upstream of a viral promoter (Fig. 2D​2D),), or by a 1-kb fragment of the TβRI gene promoter that contains four endogenous Runx REs, in control and PGE₂-activated osteoblasts.

Open in a separate window

Figure 2

Interactions between the Wnt pathway and Runx2 in osteoblasts. In A–C, fetal rat osteoblasts were transfected for 24 h to express reporter plasmids TOP-Flash or FOP-Flash at 150 ng/2 cm², or in panels D–F they were transfected with the Runx2-driven promoters SXN1c or TβRI at 75 ng/2 cm². In panels B–D, the cells were cotransfected with complementary amounts of empty vector (0) or antisense expression plasmids specific for Runx2 (αS-Rx2) or C/EBPδ (αS-C/δ) as indicated, or with empty vector (0) or Wnt1 at 75 ng/2 cm² in panel A or the concentrations shown in panel E. The cells were then treated with vehicle (0) or 1 μm PGE₂ alone or in combination with 20 mm LiCl, 10 μm WAg, or both agents (Li+W) as indicated. Rat kidney fibroblasts were also used in panel C as indicated. Reporter activity was measured after 24 h of treatments. Transgenic Wnt1 expression significantly enhanced TOP-Flash, SXN1c, and TβRI promoter activity in panels A and E, and LiCl and/or WAg significantly enhanced SXN1c and TβRI promoter activity in panel F; PGE₂ significantly enhanced TOP-Flash activity in Wnt1-transfected cells in panel A, and in LiCl and/or WAg-treated cells in panel B, designated by asterisk; transgenic αS-Rx2 expression significantly suppressed TOP-Flash activity in Wnt pathway and/or PGE₂-activated osteoblasts in panels B and C, and SXN1c and TβRI promoter activity in panel D, designated by two stacked asterisks (P < 0.05). PGE₂ had no effect on the FOP-Flash reporter in panel A, αS-C/δ had no effect on TOP-Flash activity in osteoblasts, and αS-Rx2 had no effect on TOP-Flash activity in fibroblasts in panel C. In panel G, nuclear and cytoplasmic extracts from cells treated with vehicle (0), 20 mm LiCl, 10 μm WAg, or both agents (Li+W) were polyacrylamide gel fractionated and probed with antibody to Runx2 (at ~60 kDa), TβRI (at ~55 kDa), or cyclophilin B (cycloB; at ~22 kDa), as indicated, and were visualized by Western blot analysis. In panel H, nuclear extracts from vehicle or LiCl plus WAg-activated cells (Li+W) were combined with a ³²P-labeled probe containing a consensus Runx RE (supplemental Table 2) without or with anti-Runx2 antibody (Ab) as indicated, polyacrylamide gel fractioned, and visualized by autoradiography, where the arrowhead indicates the only large complex that migrated separately from free probe (FP). Fibs, Fibroblasts; Obs, osteoblasts.

Inasmuch as Runx2 and TCF/LEF exhibited codependent roles on gene expression in osteoblasts, we then asked whether the short-term Wnt induction protocol that we employed also modulated Runx2. In this regard, transgenic or soluble factor induction of the Wnt pathway for 24 h enhanced gene promoter activity by SXN1c or by the TβRI gene promoter (Fig. 2​2,, E and F). However, it did not increase the total amount of nuclear Runx2 protein by Western blot analysis (upper panel, Fig. 2G​2G),), or by DNA binding (Fig. 2H​2H)) during this interval, even while it increased the amount of the Runx2-sensitive gene product TβRI by approximately 7-fold (middle panel, Fig. 2G​2G),), suggesting instead that it might increase the transcriptional potential of Runx2.

Physical and functional interactions between TCF/LEF and Runx2

To address whether Wnt pathway induction directly enhanced Runx2 activity, we used Runx2 fused to an exogenous GAL4 DNA-binding domain (DBD) (M1-Runx2), which independently assesses its transcriptional activation potential through a reporter driven by GAL4 DNA REs. Treatment with LiCl to induce the canonical Wnt pathway did not significantly increase M1-Runx2 transcriptional activity. By contrast, WAg was highly effective and was not further influenced by LiCl (Fig. 3A​3A,, upper panel). No activity was induced in cells that express vector M1 encoding only GAL4 DBD (Fig. 3A​3A,, lower panel). Accordingly, transgenic expression of Wnt 1, which drives its effects through a Wnt receptor system that favors β-catenin stabilization (30), also did not directly enhance M1-Runx2 activity (Fig. 3B​3B).). Indeed, as shown in earlier and later evidence throughout this study, in the many instances in which transgenic Wnt1 expression and LiCl were compared directly, Wnt1 expression was invariably as effective as LiCl treatment, giving us confidence in this approach.

Open in a separate window

Figure 3

Runx2 activation and functional interaction with TCF-4 in osteoblasts. In panels A–D and F, fetal rat osteoblasts were transfected for 24 h with GAL4 DBD fusion plasmid devoid of Runx2 (M1), or fused to full-length Runx2 (M1-Runx2 or M1Rx2), or fragments of Runx2 encompassing amino-terminal amino acids 1-92 (M1ad); amino acids 93-228 encompassing the runt domain (M1rd); or carboxyl terminal amino acids 229-513 (M1cd) at 50 ng/2 cm², in combination with reporter plasmid driven by five GAL4 RE (5XGAL4) at 150 ng/2 cm²; or in panel E with TOP-Flash reporter at 150 ng/2 cm². In panel B the cells were cotransfected to express complementary amounts of empty vector (0) or Wnt1 at the concentrations indicated. In panel C and the middle and right panels of F the cells were cotransfected to express empty vector (MVN) encoding VP16 gene transactivation domain, or MVN fused to TCF-4 lacking a β-catenin-binding domain (MVN-TCFt) at 50 ng/2 cm². In panels D and E the cells were cotransfected to express empty vector (0) or TCFt lacking the MVN transactivation domain (pcTCFt) at 75 ng/2 cm². The cells were then treated with vehicle (0) or 1 μm PGE₂ alone or in combination with 20 mm LiCl, 10 μm WAg, or both agents (Li+W) as indicated in panels A, B, E, and F. Rat kidney fibroblasts were also used in panel E, or COS-7 monkey kidney fibroblasts were also used in panel F, as indicated. Reporter activity was measured after 24 h of treatment. WAg alone or Li+W significantly enhanced Runx2-dependent gene activation in panel A; PGE₂ significantly enhanced Runx2-dependent gene activation alone in panel B, designated by asterisk, or in combination with Li+W in the left and middle panels of F, designated by two stacked asterisks; coexpression of MVN-TCFt significantly enhanced Runx2-dependent gene expression in panel C, designated by asterisk, and further increased the stimulatory effects of Li+W without and with PGE₂ in the middle and right panels of F, designated by two stacked asterisks; pcTFCt had no effect on Runx2-dependent gene expression in panel D and significantly enhanced TOP-Flash activity in LiCl- or WAg-activated osteoblasts and significantly suppressed TOP-Flash activity in rat fibroblasts in E, designated by two stacked asterisks; in contrast to osteoblasts, MVN-TCFt significantly suppressed Runx2-dependent gene expression in Li+W-treated COS-7 fibroblasts in the right panel of F, designated by two stacked asterisks (P < 0.05). Fibs, Fibroblasts; Obs, osteoblasts.

Several lines of evidence thus suggested that transcription factors in the TCF/LEF and Runx2 families functionally interact to favor each of their activities. With regard to Runx2 transcriptional activity, however, this appears not to involve either rapid stabilization of Runx2 or changes in β-catenin levels that occur in response to LiCl. Consequently, we next asked whether this might involve its physical interaction with TCF/LEF proteins (31), which, in this case, is unrelated to β-catenin. To do so, we used a β-catenin-independent derivative of the TCF/LEF gene family member TCF-4 (25) fused to the transactivation domain of herpes simplex virus protein 16 (MVN-TCFt), and measured its ability to drive two-hybrid-dependent gene expression in combination with M1-Runx2 and the GAL4-sensitive reporter in osteoblasts. By this analysis, M1-Runx2 activity was significantly increased by coexpression of MVN-TCFt, consistent with the ability of transgenically expressed Runx2 and TCF-4 to coimmunoprecipitate (31). Amino-terminal (amino acids 1–92; M1ad) and carboxyl-terminal (amino acids 229–513; M1cd) fragments of Runx2 fused to the GAL4 DBD had little or no endogenous gene activation potential, whereas the Runt domain containing fragment of Runx2 (amino acids 93–228; M1rd) was 3- to 4-fold greater in activity compared with M1 alone. Small increases in activity occurred between the various Runx2 domain fragments and MVN-TCFt, but none were as potent as full-length M1-Runx2 (Fig. 3C​3C).). Transgenic expression of TCFt fused to the Herpes virus protein 16 (VP16) gene transactivation fusion domain did not increase gene expression in combination with vector M1 encoding only GAL4 DBD (Fig. 3C​3C),), and expression of TCFt without the VP16 gene transactivation fusion domain did not augment basal M1-Runx2 activity (pcTCFt; Fig. 3D​3D),), revealing that transgenic TCFt expression itself did not increase basal GAL4 reporter gene expression or M1-Runx2 transcriptional potential (Fig. 3D​3D).). Nonetheless, transgenic expression of TCFt enhanced TCF/LEF-dependent gene promoter activity through TOP-Flash in both control, LiCl-, and Wag-induced osteoblasts (left panel, Fig. 3E​3E).). Inasmuch as TCFt lacks β-catenin binding potential, these findings established that β-catenin was not an obligate component for TCF/LEF activity in osteoblasts. In notable contrast to osteoblasts, TCFt was not stimulatory in fibroblasts. Rather, lack of the β-catenin binding by TCFt produced a dominant-negative effect in fibroblasts where it virtually eliminated the potent stimulatory effect of LiCl (right panel, Fig. 3E​3E).

We then asked whether soluble factor activation of either factor enhanced the transcriptional activity of M1-Runx2 or the functional two-hybrid complex formed by M1-Runx2 and MVN-TCFt. Osteoblasts treated with LiCl and WAg to induce the endogenous Wnt pathway, and with PGE₂ to activate M1-Runx2, induced greater M1-Runx2 activity than treatment with any agent individually or any two agents together (left panel, Fig. 3F​3F).). Moreover, induction of both pathways even further enhanced two-hybrid-dependent gene expression by M1-Runx2 and MVN-TCFt (center panel, Fig. 3F​3F).). Again by contrast to osteoblasts, expression of MVN-TCFt suppressed basal and Wnt pathway-induced M1-Runx2 transcriptional activity in COS-7 cells (right panel, Fig. 3F​3F),), confirming disparate functional effects that can occur in osteoblasts and fibroblastic cells.

Wnt modulates TGF-β activity

Given the stimulatory effect of Wnt pathway induction on Runx2 and the supportive role of Runx2 on the expression of the primary signal transducing TGF-β receptor, TβRI, in osteoblasts, we then asked whether it also controlled the responsiveness of these cells to TGF-β treatment. Transgenic Wnt1 enhanced both activator protein 1 (AP-1)-dependent gene expression through 3TPLux, and Smad-dependent gene expression through SBE4 in TGF-β-treated osteoblasts (Fig. 4A​4A,, upper and lower left panels). These effects could be separated in part because the stimulatory effect of TGF-β on AP-1 activity was relatively more sensitive to WAg (Fig. 4A​4A,, upper right panel) whereas TGF-β-induced Smad activity was relatively more sensitive to LiCl (Fig. 4A​4A,, lower right panel). Still, both transcriptional effects by TGF-β were more fully stimulated with both inducers. Wnt pathway induction also increased nuclear protein binding to AP-1 and Smad REs by EMSA in the absence of TGF-β treatment but did not further enhance the potent effect of TGF-β on nuclear protein binding (Fig. 4B​4B).

Open in a separate window

Figure 4

Wnt pathway activation selectively regulates aspects of TGF-β and BMP-4 activity in osteoblasts. In panel A, fetal rat osteoblasts were transfected for 24 h with reporter plasmid driven by three consensus AP-1 REs (3TPLux) at 150 ng/2 cm², or four consensus Smad REs (SBE4) at 75 ng/2 cm² alone or in combination with empty vector (vec) or expression plasmid encoding Wnt1 at 75 ng/2 cm² as indicated. Reporter activity was measured after 24 h of treatment. TGF-β alone, and WAg alone, designated by asterisk, or Wnt 1 expression, WAg, or Li+W significantly enhanced the effect of TGF-β on 3TPLux activity, designated by two stacked asterisks; Wnt 1 expression, TGF-β, or Li+W significantly enhanced SBE4, designated by asterisk, and LiCl or Li+W significantly enhanced the effect of TGF-β on SBE4 activity, designated by two stacked asterisks (P < 0.05). In panel B, nuclear extracts from cells treated for 24 h with vehicle (0) or LiCl plus WAg (Li+W), alone or in combination with 120 pm TGF-β1, were combined with ³²P-labeled probes containing consensus AP-1 or Smad RE (supplemental Table 2) as indicated, polyacrylamide gel fractioned, and visualized by autoradiography, where large complexes migrated separately from free probe (FP). In panel C, osteoblasts were treated for 24 h with vehicle (0), TGF-β at 12 pm for DNA synthesis, or 120 pm for collagen synthesis, or BMP-4 at 1 nm, without or with Li+W as indicated, and pulse labeled with [³H]thymidine to measure new DNA synthesis or [³H]proline to measure new collagen synthesis during the last 2 h of treatment. TGF-β and BMP-4 each significantly enhanced DNA synthesis, and their effects were significantly suppressed by Li+W. TGF-β significantly enhanced collagen synthesis, designated by asterisk, whereas Li+W significantly suppressed collagen synthesis in TGF-β- or BMP-4-treated cells, designated by two stacked asterisks (P < 0.05). In D, osteoblasts were treated for the times indicated with vehicle (0), 120 pm TGF-β, or 1 nm BMP-4, without or with Li+W as indicated, and cytoplasmic extracts were used to measure alkaline phosphatase activity in vitro. TGF-β significantly suppressed, BMP-4 significantly enhanced, designated by asterisks, and Li+W significantly suppressed the stimulatory effect of BMP-4 on alkaline phosphatase activity, designated by two stacked asterisks, after 48 or 72 h of treatment (P < 0.05). In F and G, osteoblasts were treated for 24 h with vehicle (0), 120 pm TGF-β, or 1 nm BMP-4, without or with Li+W as indicated, and mRNA levels were measured by RT-PCR. Single asterisks designate significant differences from untreated cells, and two stacked asterisks indicate significant inhibitory effects by Li+W in combination with TGF-β or BMP-4 (P < 0.05). alk phos, Alkaline phosphatase; T, TGF-β; B, BMP-4.

In contrast to these transcriptional effects, Wnt pathway induction suppressed the stimulatory effect of TGF-β on DNA and collagen synthesis (Fig. 4C​4C)) without reversing its inhibitory effect on alkaline phosphatase activity (Fig. 4D​4D).). Although other studies suggest that Wnt might induce some BMP-dependent responses in osteoblasts through interactions among Wnt receptors, kinase cascades, and downstream transcriptional components (32), Wnt pathway induction also limited DNA and collagen synthesis and potently suppressed alkaline phosphatase activity in response to BMP-4 (Fig. 4​4,, C and D).

To address whether the Wnt pathway regulates the steady state levels of osteoblast-related mRNAs in TGF-β- or BMP-4 treated cells, relative transcript abundance was determined by RT-PCR. Transcripts encoding the α1-chain of type I collagen were by far the most abundant, by 500-fold relative to alkaline phosphatase, 200- to 1000-fold relative to the various TβRs, and 200,000-fold relative to osteocalcin (Fig. 4E​4E).). Of note, the level of β-actin mRNA was not useful as an invariant control because activation of the Wnt or TGF-β pathways highly increased its expression without inducing TβRII or TβRIII (top panel, Fig. 4F​4F).). Still, earlier studies showed that TβRIII mRNA expression varies significantly with differentiation, growth factor, or hormone treatments (33,34,35,36). Thus, possible variations in mRNA were determined relative to TβRII. Wnt pathway induction enhanced mRNA encoding the Runx2-sensitive gene TβRI, and this was increased further by TGF-β (top left panel, Fig. 4G​4G).). Analogous relative changes occurred on mRNA encoding the α1-chain of type I collagen (bottom left panel, Fig. 4G​4G),), also thought to be Runx2 sensitive. Unlike its combinatorial effect with TGF-β, Wnt failed to augment BMP-4 activity on TβRI and α1 collagen I mRNA (left panels, Fig. 4G​4G).). By contrast, and consistent with their individual or combined effects on enzyme activity, Wnt and TGF-β pathway induction suppressed alkaline phosphatase mRNA, and Wnt pathway induction suppressed the potent stimulatory effect of BMP-4 (top right panel, Fig. 4G​4G).). Costimulatory effects between the Wnt and TGF-β pathways were not evident on osteocalcin mRNA, another Runx2-sensitive gene, whereas Wnt pathway induction was suppressive in BMP-4-activated cells (bottom right panel, Fig. 4G​4G),), perhaps reflecting the low basal levels of osteocalcin mRNA in osteoblasts (Fig. 4E​4E)) and the need for a potent gene coinducer like dihydroxyvitamin D₃ (37,38). Therefore, effects by Wnt pathway induction on mRNA expression were related to changes in protein expression or activity, or to the presence of Runx2 promoter elements in some genes but not in others, revealing multiple regulatory events in osteoblasts that require still further study.

Plasmids

Local Wnt expression was induced with expression plasmid encoding prototypical Wnt1. Wnt-dependent gene expression was assessed with luciferase reporter plasmid TOP-Flash containing four TCF/LEF response elements (RE), FOP-Flash in which the TCF/LEF-binding elements in TOP-Flash are mutated, or a 0.94-kb fragment of the rat oxytocin gene promoter (NCBI reference sequence NC_005102.2) that contains multiple TCF/LEF-binding elements, as revealed by MatInspector (http://www.genomatix.de) analysis. Runx-dependent gene expression was assessed with luciferase reporter plasmid SXN1c containing two copies of a Runx RE from the rat TβRI, or a 1.0-kb fragment of the native rat TβRI gene promoter. TGF-β-dependent gene expression was assessed with luciferase reporter plasmid 3TPLux containing three AP-1 REs, or SBE4 containing four Smad REs (29,46,54,55). Endogenous transcription factor levels were reduced with transgenic expression plasmids containing restriction fragments of mouse Runx2 or rat CCAAT enhancer binding protein (C/EBP)δ cloned in reversed orientation within vector pSV7d (46,55,56,57). Runx2 transcriptional activation was assessed with expression plasmids encoding full-length or various fragments of Runx2 fused to a GAL4 promoter DBD within vector M1 (M1-Runx2), and luciferase reporter plasmid containing five GAL4 REs (5XGAL4) (55,56,57,58). Physical and functional interactions between TCF-4 and Runx2 were assessed by cotransfection with expression plasmid encoding a β-catenin-independent derivative of TCF-4 (25) fused to the transactivation domain of Herpes virus protein 16 (VP16) within vector MVN (MVN-TFCt), M1-Runx2, and luciferase reporter plasmid 5XGAL4.

Transfections

Promoter-reporter, expression, antisense, and empty vector plasmids were pretitrated for optimal expression efficiency and transfected with reagent TransIT LT1 (Mirus Corp., Madison, WI). Cell cultures at least 70% confluent (~35,000 cells per cm²) were exposed to 25–75 ng per cm² of expression plasmid DNA and/or 37.5–75 ng per cm² of reporter plasmid DNA for 16 h in medium containing 4% fetal bovine serum. Cells were then treated as indicated in each figure in serum-free medium and lysed, and extracts were analyzed for reporter gene activity by comparison with positive and negative control plasmids, and corrected for differences in protein content (46,55,56,57).

Western blots

Equal amounts of protein were fractionated on 12.5% sodium dodecyl sulfate-polyacrylamide gel and electroblotted onto polyvinylidene fluoride membranes (NEN Life Science Products, Boston, MA) along with prestained molecular weight markers. Blots were blocked in 5% fat-free powdered milk, probed with a 1:1000 or 1:3000 dilution of primary antibody [Santa Cruz Biotechnology, Inc. (Santa Cruz, CA) or BD Transduction Laboratories (Lexington, KY)]. Reactive bands were visualized with a 1:2,500 dilution of species specific anti-IgG linked to horseradish peroxidase (The Jackson Laboratory, Bar Harbor, ME) and chemiluminescence (Western Lightning, PerkinElmer Life Sciences, Wellesley, MA). Inasmuch as various proteins differ in relative abundance, exposure times or image intensities were varied to produce analogous stain to background ratios and achieve more accurate determinations of relative differences by densitometry (26).

mRNA analysis

Total RNA was extracted with acid guanidine monothiocyanate, precipitated with isopropyl alcohol, dissolved in sterile water (33), and purified further using the RNeasy Plus Mini Kit (QIAGEN, Chatsworth, CA). RNA was reverse transcribed with the iScript cDNA Synthesis Kit, and products were amplified and quantified using specific primers (eurofins mwg/operon; Supplemental Table 2) and the iQ SYBR Green Supermix in a CFX96 real-time PCR detection system (Bio-Rad Laboratories, Inc., Hercules, CA).

EMSA

Double-strand oligonucleotide DNA probes containing specific nuclear factor REs (Supplemental Table 3) were labeled with [³²P]dCTP and Klenow fragment of Escherichia coli DNA polymerase I and gel purified. Extracted nuclear protein (3 μg) from control or induced cells was supplemented with ³²P-labeled probe, and protein-bound DNA complexes were resolved on a 5% nondenaturing polyacrylamide gel and examined by autoradiography (46,55,56,57).

DNA synthesis

Cells were incubated with 5 μCi/ml [methyl-³H]thymidine (80 Ci/mmol) for the last 2 h of culture. Cells were lysed in 0.1 m sodium dodecyl sulfate and 0.1 m sodium hydroxide, collecting the precipitate formed with 10% trichloroacetic acid, and scintillation counting (33,51).

Collagen synthesis

Cells were incubated with 5 μCi/ml [³H-2,3]proline (2.5 Ci/mmol) for the last 2 h of culture. Cell layers were lysed by freeze thawing and extracted with 0.5% Triton X-100. Samples were precipitated with cold 10% trichloroacetic acid, and acid-insoluble material was collected by centrifugation. Precipitates were extracted with acetone, dried, and rehydrated. [³H-2,3]proline incorporated into collagenase-digestible protein (collagen) was determined with bacterial collagenase free of nonspecific protease activity (33,51).

Alkaline phosphatase activity

Cell layers were lysed by freeze thawing and extracted with 0.5% Triton X-100. Enzyme activity was assessed in cell extracts by hydrolysis of p-nitrophenol phosphate, measured at 410 nm after 30 min of incubation at 37 C. Data are expressed as picomoles of p-nitrophenol released per minute per μg protein (33,51).

Statistical analysis

Differences were assessed by one-way ANOVA with Tukey post hoc analysis using SigmaStat software (Jandel Scientific, San Rafael, CA) from nine or more replicate samples and two or more independent cell preparations. Western blots were from at least two studies and mRNA levels were from three or more studies. A significant difference was assumed by a P value below 0.05.

We thank Drs. John Wysolmerski (Yale University, New Haven, CT), Joan Massague (Sloan Kettering Memorial Institute, New York, NY), Yoshiaki Ito (National University of Singapore), Peter A. Rotwein (Oregon Health Sciences University, Portland, OR), Bert Vogelstein (Johns Hopkins University, Baltimore, MD), Ivan J. Sadowski (University of British Columbia, Vancouver, British Columbia, Canada), and Peter G. Schultz and Jun Liu (Scripps Research Institute, La Jolla, CA) for some of the vectors, reporter, or expression plasmids; and to Drs. Scott A. Heibert (Vanderbilt-Ingram Cancer Center, Nashville, TN), Joseph A. Madri, and Dianne Duffey (Yale University), for some of the antibodies used in these studies. We also thank Dr. Madri (Yale University) for his helpful comments and advice during manuscript preparation.