Tgfβ Signaling Directly Induces Arf Promoter Remodeling by a Mechanism Involving Smads 2/3 and p38 MAPK*

We have investigated how the Arf gene product, p19Arf, is activated by Tgfβ during mouse embryo development to better understand how this important tumor suppressor is controlled. Taking advantage of new mouse models, we provide genetic evidence that Arf lies downstream of Tgfβ signaling in cells arising from the Wnt1-expressing neural crest and that the anti-proliferative effects of Tgfβ depend on Arf in vivo. Tgfβ1, -2, and -3 (but not BMP-2, another member of the Tgfβ superfamily) induce p19Arf expression in wild type mouse embryo fibroblasts (MEFs), and they enhance Arf promoter activity in ArflacZ/lacZ MEFs. Application of chemical inhibitors of Smad-dependent and -independent pathways show that SB431542, a Tgfβ type I receptor (TβrI) inhibitor, and SB203580, a p38 MAPK inhibitor, impede Tgfβ2 induction of Arf. Genetic studies confirm the findings; transient knockdown of Smad2, Smad3, or p38 MAPK blunt Tgfβ2 effects, as does Cre recombinase treatment of Tgfbr2fl/fl MEFs to delete Tgfβ receptor II. Chromatin immunoprecipitation reveals that Tgfβ rapidly induces Smads 2/3 binding and histone H3 acetylation at genomic DNA proximal to Arf exon 1β. This is followed by increased RNA polymerase II binding and progressively increased Arf primary and mature transcripts from 24 through 72 h, indicating that increased transcription contributes to p19Arf increase. Last, Arf induction by oncogenic Ras depends on p38 MAPK but is independent of TβrI activation of Smad 2. These findings add to our understanding of how developmental and tumorigenic signals control Arf expression in vivo and in cultured MEFs.

Arf is conserved in amniotes as a gene encoding p19 Arf (p14 ARF in humans), a tumor suppressor that exerts its effects by both p53-dependent and -independent mechanisms (1). Early studies, predominantly in cultured cells, showed that Arf is induced to check proliferation in sequentially passed MEFs 3 in vitro (2,3). Moreover, Arf expression is augmented in the presence of certain oncoproteins, like adenovirus E1A, Myc, E2F, Bcr-Abl, and Ras V12 (3)(4)(5)(6)(7). These findings coupled with the initial failure to identify Arf expression in the developing mouse embryo (8) led to the concept that Arf acts as an oncogene sensor that is induced by cell autonomous mechanisms in response to inappropriate or excessive cell proliferation signals (9).
More recent observations point toward Arf regulators extending beyond oncogenic signals. For instance, mouse Arf expression increases with age in a variety of cells that have not suffered overt oncogenic stress (10). Closer evaluation of the developing mouse embryo shows Arf to be robustly expressed in a temporally and spatially restricted pattern in the developing hyaloid vessels and cornea in the eye and also in perivascular cells flanking the intra-embryonic umbilical vessels (11,12). A developmental function of p19 Arf is only clear in the pericytelike cells in the hyaloid vessels of the primary vitreous, where it prevents overgrowth of the pericytes to prompt the developmentally timed regression of the hyaloid vessels (12). Homozygous Arf deletion results in primary vitreous hyperplasia, ocular lens opacification, retinal dysplasia, and blindness by 2 weeks of age (11,12).
The expanded role of p19 Arf in the embryo raises the question of how developmental signals operate and whether they overlap with oncogenic signals controlling Arf. Using a candidate-gene approach, transforming growth factor ␤2 (Tgf␤2), a member of the Tgf␤ cytokine superfamily, was recently found to be critical for Arf expression at several sites in the developing mouse (13). Supporting the importance of this finding, mouse embryos lacking Tgf␤2 have primary vitreous hyperplasia similar to that observed in Arf Ϫ/Ϫ embryos (14,15). Importantly, these observations can be replicated in vitro because exogenous Tgf␤2 enhances Arf expression in cultured MEFs and maintains a proliferation arrest in an Arf-dependent manner (13), thereby providing a model system to further investigate mechanisms.
Members of the Tgf␤ superfamily frequently modulate the transcription of key target genes through Smad proteins, which directly transduce Tgf␤ receptor activation to the nucleus. In addition, Smad-independent signaling through p38 MAPK, ERK, PI3K/Akt, and JNK provide alternative mechanisms of gene activation (16). In this manuscript we demonstrate that Tgf␤ and p19 Arf act on cells of the same lineage; that Arf is required for the anti-mitogenic effects of Tgf␤ in vivo and that both Smad-and p38 MAPK-dependent mechanisms underlie Arf induction by Tgf␤.
Histology Studies-Female mice pregnant with embryonic day (E) 13.5 litters received BrdU (10 mg/g in PBS) by intraperitoneal injection 4 and 2 h before euthanasia using CO 2 . Whole embryos were fixed in 4% paraformaldehyde in PBS for 4 h at 4°C and then equilibrated in 20% sucrose overnight at 4°C. Fixed embryo heads were embedded in TBS Tissue Freezing Media (Fisher) before cryostat sectioning. Hematoxylin-and eosin-staining was performed using 5-m sections as previously described (2,3). For BrdU staining, 5-m sections were blocked in 10% donkey serum, 0.1% Triton X-100, PBS at room temperature and then stained using sheep ␣-BrdU polyclonal antibody (1:100, Fitzgerald Industries International; Action, MA) at room temperature for 90 min. The primary antibody was detected with a Dylight 488-conjugated donkey anti-sheep secondary antibody (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA). Sections were mounted in VectaShield mounting media with DAPI (Vector Laboratories, Inc; Burlingame, CA) and visualized using a Leica DM IRB fluorescent microscope at 400ϫ magnification. The fraction of DAPI-positive cells in the vitreous that were BrdU-positive was determined using at least three embryos from two or more different litters. Quantification was verified by two individuals who were blinded to the genotypes. Photomicrographs were obtained using an Optronics camera and MagnaFire 2.1C imaging software (Optronics, Goleta, CA).
Cell Culture, Western Blot Analysis, and ␤-Galactosidase Assay-Early passage wild type and Arf lacZ/lacZ MEFs were treated with either Tgf␤1 (5 ng/ml), Tgf␤2 (5 ng/ml), Tgf␤3 (10 ng/ml), or BMP2 (150 ng/ml) or an equivalent volume of vehicle (4 mM HCl) for 1.5-72 h. Arf-null 10T1/2 cells were transduced with Gfp-or Arf-expressing retrovirus as negative and positive controls for Western blotting. In some experiments SB203580 (20 M), SB431542 (10 M), SP600125 (10 M), U0126 (10 M), and LY 294002 (5 M) were applied to MEFs for 20 min before Tgf␤2 or vehicle was added to the culture media. In some studies, wild type MEFs or Arf lacZ/lacZ MEFs were seeded at 50% confluence in 6-or 12-well plates 24 h before transfection using of siRNA targeting specific genes (or scrambled siRNA as a control) (0.5 M) using DharmaFECT 2 (Dharmacon) according to the manufacturer's instructions. In some studies, Tgfbr2 fl/fl MEFs were infected with adenovirus encoding Cre recombinase or red fluorescent protein as a control for 48 h before exposure to vehicle or Tgf␤2 for Western blotting for p19 Arf . In some experiments, early passage wild type MEFs were infected with retrovirus encoding H-RAS V12 and treated with SB203580 (20 M) and SB431542 (10 M) concurrently. For all studies Western blotting and ␤-galactosidase assays were performed in wild type and Arf lacZ/lacZ MEFs, respectively, as previously described (13). Experimental findings were confirmed in at least two independent experiments, with quantitative data from ␤-galactosidase assays pooled from all representative experiments.
As a control for SB203580 activity, the p38 MAPK assay (Cell Signaling Technology) was used according to the manufacturer's instructions. Briefly, phosphorylated p38 MAPK was immunoprecipitated from the corresponding cell lysate, and its activity was determined by in vitro kinase assay measuring phosphorylation of its substrate ATF-2.
Statistical Analysis-Quantitative data are presented as the mean Ϯ S.D. from three or more representative experiments. Statistical significance (p value Ͻ0.05) was calculated using Student's t test.

Arf Is Required for Anti-proliferative Effects of Tgf␤ in Vivo-
Arf expression is blunted in the developing eyes of Tgf␤2 Ϫ/Ϫ embryos (13), and both Arf Ϫ/Ϫ and Tgf␤2 Ϫ/Ϫ embryos have primary vitreous hyperplasia evident at embryonic day (E) 13.5 (13,15,21). These observations imply that Tgf␤2 lies "upstream" of Arf. Furthermore, they are consistent with the idea that p19 Arf is required for the anti-proliferative effects of Tgf␤2 in the mouse eye. To formally test this hypothesis, we took advantage of a transgenic mouse line in which Tgf␤1 is expressed from the ␣A-crystallin promoter (here called AC-Tgf␤1 mice) (20). Indeed, previous studies demonstrated that transgenic expression of Tgf␤1 can rescue the Tgf␤2-dependent primary vitreous hyperplasia in E18.5 embryos, although the mechanism for the rescue has not been elucidated (22). This finding coupled with our prior observation that Tgf␤1 can activate the Arf promoter in cultured MEFs in a manner comparable with Tgf␤2 (13) allowed us to explore the epistatic relationship between Tgf␤2 and Arf in vivo.
Arf-expressing Cells in the Eye Originate from Wnt1-expressing Neural Crest-The fact that some Arf-expressing cells coexpress T␤rII in the eye (13) indicates that Tgf␤2 might signal directly to cells expressing p19 Arf . To further address this, we took advantage of the fact that members of the Sommer laboratory (23) previously showed that blocking Tgf␤ signals in cells derived from a Wnt1-expressing lineage leads to primary vitreous hyperplasia. This was accomplished by breeding Wnt1-Cre mice into the Tgf␤r2 fl/fl mouse strain in which exon 4 Tgfbr2 can be conditionally deleted by Cre-mediated recombination (19). We investigated whether primary vitreous hyperplasia would similarly occur in animals in which Arf is inactivated in the same lineage.
We accomplished this by breeding Wnt1-Cre mice with animals in which with exon 1␤ of Arf is flanked by LoxP sites (17). These Arf fl/fl animals were previously used with Arf Cre/Cre and Pdgfr␤ fl/fl animals by Gromley et al. (17) to formally show that conditional deletion of Pdgfr␤ anatomically and functionally rescues the postnatal Arf Ϫ/Ϫ eye phenotype. As a control, we confirmed that Cre promoted recombination in cells populating the primary vitreous by analyzing Wnt1-Cre, Rosa26 lacZ/ϩ mouse embryos. 4 Histological examination of eyes from postnatal day (P) 15 Wnt1-Cre, Arf fl/fl mice revealed a hyperplastic retrolental mass that was not observed in Wnt1-Cre, Arf fl/ϩ or Arf fl/fl littermates ( Fig. 2A; additional data not shown). Vitreous hyperplasia was also observed in Wnt1-Cre, Arf fl/fl embryos at E13.5 (Fig. 2B), and this correlated with increased BrdU incorporation in these cells (Fig. 2, C and D); both of these findings reflect the phenotype of Arf Ϫ/Ϫ mice (11,21). This finding demonstrates that the cells that are controlled by p19 Arf arise in the Wnt1 expressing neural crest and further supports a model in which Tgf␤ directly influences the cells expressing Arf.
Tgf␤1, -2, and -3 Induce p19 Arf Expression in Cultured MEFs-Having confirmed the functional relationship between these two proteins, we sought to better define the mechanisms by which Tgf␤ controls Arf. We took advantage of a cell culture system using early passage MEFs from wild type E14.5 embryos. To understand the kinetics of p19 Arf induction, we treated the MEFs with Tgf␤2 for 1.5, 24, and 48 h. In this context, p19 Arf protein was minimally increased at 24 h and was significantly higher at 48 h (Fig. 3A), suggesting that p19 Arf induction was not an immediate Tgf␤2 response.
We also measured the ability of each of the closely related Tgf␤1, -2, and -3 proteins to enhance Arf expression by using MEFs derived from an Arf lacZ/lacZ reporter mouse in which the first exon of Arf is replaced by lacZ cDNA (13). Arf lacZ/lacZ MEFs derived are functionally Arf Ϫ/Ϫ , and ␤-galactosidase expression can be used as a surrogate for Arf promoter activity. Our previous work showed that the time course for ␤-galactosidase induction by Tgf␤2 parallels that of p19 Arf protein ( Fig.  3A) (13). All three of the related Tgf␤ proteins induced ␤-galactosidase expression in the MEFs (Fig. 3B, lanes 2, 4, and 6). The relatively small increases in the Arf promoter in Arf lacZ/lacZ MEFs as compared with p19 Arf induction (compare Fig. 3, A, lanes 5 and 6 to B, lanes 3 and 4) indicated that increased Arf transcription might be complemented by additional, post-transcription effects such as transcript stabilization or increased translation. In contrast, BMP-2, another Tgf␤ superfamily member (16,24), failed to induce ␤-galactosidase (Fig. 3C). These findings suggest that Tgf␤1, -2, or -3 act through the classical pathway involving ALK-5 and T␤rII rather than BMP type I receptors (ALK-2, ALK-3, and ALK-6) (16,24).
We previously showed Arf and T␤rII to be co-expressed in some pericyte-like cells in the primary vitreous in the mouse (13). To directly confirm the importance of this receptor in Arf gene activation, we used MEFs derived from the above-mentioned Tgfbr2 fl/fl mouse. When the MEFs were infected with adenovirus encoding Cre recombinase, T␤rII expression fell (Fig. 3D, lanes 3 and 4 versus 1 and 2). Tgf␤2 augmented p19 Arf in Tgfbr2 fl/fl MEFs when they were infected with control adenovirus encoding red fluorescent protein (Fig. 3D, lane 2 versus lane 1) but not in MEFs in which Cre recombinase had diminished T␤rII expression (Fig. 3D, lane 4 versus lane 3).

Smads 2 and 3 Cooperatively Induce p19 Arf in Response to
Tgf␤2-Given the role that Smads 2 and 3 play in response to Tgf␤2, we sought to define the relative importance of these two proteins in Arf regulation. We used gene-specific siRNA to knock down their expression singly and in combination in wild type and Arf lacZ/lacZ MEFs (Fig. 4, A and B, top panel; the supplemental figure). Despite the very low Smad3 expression as compared with Smad 2, knockdown of either one impaired Tgf␤2 induction of the native Arf locus and ␤-galactosidase in wild type and Arf lacZ/lacZ MEFs, respectively; interestingly, targeting both proteins further limited Arf induction (Fig. 4, A and  B, bottom panel). These findings indicate that Smad2 and -3 cooperatively control Arf induction by Tgf␤2.
The p38 MAPK-dependent Pathway Also Mediates Tgf␤2 Effects on p19 Arf -To investigate whether Smad-independent signals also influence Tgf␤2 induction of p19 Arf , we used a panel of chemical inhibitors targeting a variety of pathways recognized to be influenced by Tgf␤ (Fig. 5A). As a control, we included SB431542, an inhibitor of T␤rI, which blocks Smad2 phosphorylation and Tgf␤2-driven induction of p19 Arf and lacZ expression in wild type and Arf lacZ/lacZ MEFs, respectively (13) (Fig. 5, B, lanes 5 and 6, C, lanes 11 and 12, and D, lane 7).
Application of SB203580, a chemical inhibitor of p38 MAPK, blunted Arf expression in both models (Fig. 5, B, lanes 3 and 4,  and C, lanes 3 and 4) but did not interfere with Smad 2 phosphorylation (Fig. 5D, lane 3). Consistent with a previous publication, SB203580 was not able to block p38 MAPK phosphorylation (25,26), but it did block the phosphorylation of its downstream target ATF-2 (Fig. 5E). Of note, blocking T␤rI activity with SB431542 also dampened p38 MAPK activation, consistent with existence of parallel pathways downstream from this receptor (Fig. 5D, lane 7). Although the absence of a measurable effect on Smad2 phosphorylation suggests that SB203580 may act independently of Smads, we cannot formally exclude the possibility that p38 MAPK inhibition does influence Smad-dependent activity.
Because SB203580 might have off-target effects, we used a genetic approach to confirm the importance of p38 MAPK in Tgf␤-mediated Arf regulation. Consistent with our results from chemical inhibition, even partial knockdown of p38 MAPK blocked the effects of Tgf␤2 on Arf expression (Fig. 6, A and B).
Tgf␤ Fosters Chromatin Remodeling of the Arf Gene-We previously demonstrated Smad 2/3 had the capacity to bind to two regions ϳ2.2 and ϳ1.5 kb upstream of the Arf translation initiation codon in Arf lacZ/lacZ MEFs (13); however, we did not understand how Tgf␤2 influenced the binding, the kinetics of the effect, and any other coincident changes to the Arf gene. To begin to address these issues, we first examined the binding of Smad2/3 to these regions in vehicle-and cytokine-treated MEFs. We used wild type MEFs to preclude adverse positional effects from insertion of the lacZ-Neo cassette into the Arf locus in Arf lacZ/lacZ MEFs; however, more limited analysis of these MEFs showed essentially the same results (data not shown).
We observed that Smad2/3 binding at both the distal and proximal Tgf␤2 binding elements (formerly P3 and P7, respectively, in Freeman-Anderson et al. (13)) increased within 1.5 h after the addition of Tgf␤2 (Fig. 7Ba); the effects were more pronounced at the promoter proximal element. In contrast, Tgf␤2 did not significantly enhance Smad2/3 binding over base line at the other, previously identified distal site (formerly called P1 in Freeman-Anderson et al. (13)) (data not shown) or at putative Tgf␤2 binding elements in the first intron (formerly P11 in Freeman-Anderson et al. (13)) (Fig. 7, A and Ba, lanes [3][4][5][6]. Smad binding can foster histone modification and the recruitment of the RNA Polymerase II (RNA Pol II) complex at Tgf␤-responsive genes (27,28). In our experiments, histone H3 acetylation paralleled changes in Smad2/3 binding at the distal and proximal elements 1.5 h after Tgf␤2 treatment; histone acetylation was not observed at the Necdin promoter, a locus that remains silent in MEFs (Fig. 7Bb, lanes 1-4 versus 5 and 6). In contrast to early Smad2/3 binding and histone acetylation, RNA Pol II binding at the proximal site was absent at 1.5 h but became detectable by 24 h after Tgf␤ (Fig. 7Bc).
Tgf␤ Treatment Increases Arf Transcription-We used quantitative, real-time RT-PCR to explore how the events at the Arf promoter correlated with increased transcription. The mature Arf mRNA transcript increased in wild type MEFs from nearly base line at 24 h through 72 h after Tgf␤2 addition to the culture media (Fig. 8, top panel) in a time course paralleling p19 Arf induction (Fig. 3A). Similarly, quantita-tive, real-time RT-PCR of the primary, unprocessed Arf transcript increased (Fig. 8, bottom panel). The parallels between RNA Pol II localization and Arf primary transcript induction indicate that Tgf␤2 enhances Arf transcription between 24 and 72 h even though Smad2/3 binding and histone modification near exon 1␤ are enhanced as early as 1.5 h after exposure to this cytokine.
Oncogenic RAS Activates Arf Independently of Tgf␤ Signaling to Smad2/3-Arf was initially described as an oncogene sensor that checks inappropriate or excessive cell proliferation signals. For example, ectopic expression of oncogenic H-RAS V12 activates Arf via the Raf-ERK-Dmp1 pathway (29). We investigated whether the developmental signaling pathway outlined above was relevant to Arf induction by oncogenic RAS expression in MEFs.   NOVEMBER 12, 2010 • VOLUME 285 • NUMBER 46

JOURNAL OF BIOLOGICAL CHEMISTRY 35659
Ectopic expression of human H-RAS V12 induced p19 Arf 48 h later in early passage, wild type MEFs (Fig. 9, lanes 4 versus 1). Inhibition of T␤rI phosphorylation of Smad2/3 by SB431542 had no obvious effect Arf induction (Fig. 9, lane 5 versus 2), but it did slightly increase base-line p19 Arf (Fig. 9, lane 2 versus 1). In contrast, inhibition of p38 MAPK using SB203580 significantly blocked p19 Arf induction by H-RAS V12 (Fig. 9, lane 6  versus 3). Of note, p19 Arf induction by ectopically expressed FIGURE 5. Smad-independent pathways influence Tgf␤ induction of Arf. A, shown is a schematic diagram showing Smad-dependent and Smad-independent and specific inhibitors for each target: SB43, SB431542; SB20, SB203580; SP60, SP600125; U01, U0126; LY29, LY294002. B and C, shown is a representative Western blot for the indicated proteins from wild type MEFs (B) and ␤-galactosidase activity in Arf lacZ/lacZ MEFs (C) exposed to Tgf␤2 (T2) or vehicle (V) for 48 h and either DMSO or the indicated chemical inhibitors. Quantitative analysis (C) shows that Tgf␤2 significantly changed ␤-galactosidase activity when compared with relevant vehicle (p Ͻ 0.013); however, the induction of ␤-galactosidase was significantly blunted by SB203580 and SB431542 when compared with DMSO (p ϭ 0.0015 (#) and Ͻ0.001 (*), respectively), and it was significantly enhanced by U0126 when compared with DMSO (p ϭ 0.047 (#)). D, a representative Western blot shows the corresponding targets for individual inhibitors in Arf lacZ/lacZ MEFs exposed to Tgf␤2, and the indicated chemical inhibitor confirms that LY294002 blocks AKT phosphorylation (lane 4), U0126 blocks p42/44 phosphorylation (lane 6), and SB431542 blocks Smad2 and p38 MAPK phosphorylation (lane 7). E, representative Western blot using Arf lacZ/lacZ MEFs exposed to the indicated chemical inhibitors or DMSO for 20 min shows that SB20 blunts anisomycin-stimulated, p38 MAPK-dependent phosphorylation ATF-2. F, a representative Western blot using Arf lacZ/lacZ MEFs exposed SP600125 for 20 min blunts UV-stimulated phosphorylation of JNK. cyclin D1 was not measurably affected by either chemical inhibitor. 5 Thus, p38 MAPK is needed for full induction of Arf by both Tgf␤2 and oncogenic RAS in MEFs; in contrast, Smad2/3 phosphorylation is only required for Arf induction by Tgf␤2.

DISCUSSION
The central role that p19 Arf plays as a tumor suppressor in incipient cancer cells is well established. In contrast, relatively little is known about Arf regulation in physiological contexts. We recently demonstrated that the Tgf␤2 gene is essential for Arf induction at several sites in the developing mouse embryo (13). Here we further explore the functional relationships between Tgf␤ and p19 Arf in vivo, and we uncover molecular mechanisms underlying Arf induction by this signaling protein.
First, we demonstrate that the capacity for Tgf␤ to arrest cell proliferation in vivo in the developing eye strictly depends on its ability to induce p19 Arf . Hence, understanding of how it induces Arf becomes central to knowing how it acts in this developmental capacity. Second, we use a genetic approach to show that T␤rII and p19 Arf both act in cells of the same, neural crest-derived lineage to prevent primary vitreous hyperplasia; this provides in vivo evidence for cell-intrinsic signaling from T␤rII to Arf. Third, the complementary use of chemical inhibitors and genetic manipulations define the signaling pathways extending from Tgf␤1, -2, or -3 binding to the T␤rI/II complex to Arf gene activation. Our findings indicate that Smads 2 and 3 and p38 MAPK are necessary for full p19 Arf induction. We have characterized the kinetics of Smad2/3 binding in DNA elements 5Ј to Arf and subsequent changes in the chromatin architecture of the locus. Interestingly, although these events begin within a short 1.5-h timeframe after Tgf␤2 treatment, Arf transcription, measured by RNA Pol II binding and primary Arf transcript increase, are not detected until 24 -48 h later. Last, we show that Arf induction by oncogenic RAS shares the p38 MAPK arm of this developmental pathway, whereas activation of Smad2/3 is dispensable.
The delay between Smad binding to the Arf gene and subsequent increases in Arf promoter and transcript levels was unexpected. Early Smad-dependent effects of Tgf␤ on gene transcription are often evident within several hours of receptor activation (e.g. Gomis et al. 30)). Indeed, in cultured MEFs, early cell proliferation arrest by Tgf␤2 proceeds independently of Arf (13) and is likely mediated by earlier events like repression of Cdk4 (31) or induction of Cdk inhibitors like p21 Cip1 or p15 Ink4b (32,33). We can speculate that the delayed Arf induction (although still initiated as an immediate Smad-dependent response) could have evolved to allow p19 Arf to primarily contribute to the maintenance of a Tgf␤-driven proliferation arrest.
At a molecular level, the delay between Smad binding and transcriptional activation suggests the need for the recruitment of other transcription factors to Arf regulatory sequences before or coincident with RNA Pol II binding. Candidate transcription factors already implicated as direct Arf regulators (by virtue of binding to the Arf promoter) include E2f 1 and 3 (34), Dmp1 (35), Pokemon (36), and FoxO3a (37), which positively or negatively regulate mouse Arf expression. Of these, FoxO3a (which binds to intronic DNA ϳ8 kb 3Ј of exon 1␤) (37) is particularly interesting because FoxO proteins directly interact with Smads 2/3 and 4 to induce p21 Cip1 (38). Functional cooperation between FoxO and Smad proteins is also found in a cluster of genes that seem to control stress and adaptive cell signaling responses at least in cultured HaCaT cells (30). However, this work focused on immediate responses within 3 h, a time point at which Arf transcription is not observed. In our preliminary studies, Tgf␤2 did not alter the levels of FoxO3a in MEFs, nor did inhibition of Akt (a negative regulator of FoxO (39)). Additional work is required to confirm or dispel the importance FoxO proteins in Arf regulation in the eye.
That a p38-dependent signaling pathway also contributes to the regulation of Arf by Tgf␤ is consistent with the growing understanding of cross-talk between Smad-dependent and -independent effectors of Tgf␤ (16). p38 MAPK has previously been implicated in Ink4/Arf regulation; for example, decreased expression of Wip1 phosphatase in Ppm1d Ϫ/Ϫ MEFs increases both p16 Ink4a and p19 Arf in a p38 MAPK-dependent manner (40). It also contributes to other Tgf␤ effects like the arrested DNA synthesis in primary mouse vascular smooth muscle cells at 48 h (26), induction of type I collagen in a cultured retinal pigment epithelial cell line at 24 h (41), and increased ␣-smooth muscle actin expression at 48 -72 h in cultured primary human fibroblasts (42). We previously showed that some of the Arfexpressing perivascular cells also express ␣-smooth muscle actin in the newborn mouse eye and that ectopic p19 Arf expression in 10T1/2 fibroblasts (in which Arf is deleted) can promote 5 Y. Zheng and S. X. Skapek, negative data not shown. ␣-smooth muscle actin and smooth muscle myosin expression 48 -96 h later (43). Conceivably, the induction of Arf with smooth muscle proteins may represent part of a Tgf␤-and p38 MAPK-dependent transcriptional routine leading to the maturation and cell cycle arrest of a subtype of vascular smooth muscle cells surrounding the hyaloid vessels.
How Smad and p38 MAPK signaling cooperates to induce Arf is not clear. Direct cooperation between the two is possible, and our preliminary studies indicate that SB203580 slightly decreases Smad2/3 occupancy at the proximal element, but the decrease is not statistically significant (negative data not shown). Instead, the positive interactions we found may lie in the ability of p38 MAPK to activate potential transcriptional co-factors, such as p300, C/EBP␤, or ATF2 (16,44,45). Of these, p300 is known to cooperate with Smads to modulate histone acetylation (46). Putative binding sites for C/EBP␤ and ATF-2 are present in genomic DNA flanking the Arf first exon. C/EBP␤ was shown to be required for Ras V12 -mediated senescence in MEFs but it was not needed for p19 Arf induction by this oncogene (47); instead, p19 Arf negatively regulates C/EBP␤ (48). ATF2 has a variety of activities that could play a role in cancer biology, but there is no clear connection to Arf (49).
Given the broad role of Arf in both tumor suppression and eye development, our findings may help us understand certain human diseases. For instance, so far the molecular pathogenesis of persistent hyperplastic primary vitreous, which the Arf Ϫ/Ϫ model mimics (11,12), is unclear. Occasional familial cases of this disease suggest an underlying genetic basis (50 -52). Elucidating the complete series of components necessary for Tgf␤mediated Arf transcription will allow us to potentially interrogate the genomic DNA extracted from either involved tissue samples or from blood of diseased patients in a more focused way. Because a persistent hyperplastic primary vitreouslike disease can also develop in somatic mosaic mice in which Arf is missing in only a subset of cells in the mouse (43), such an analysis should be accomplished in a way that can also detect mosaicism of the key gene(s).
We hope that our findings may also provide some insight into tumor suppression by Arf in incipient, oncogene-stressed cancer cells. This may involve both how Arf is induced and how Tgf␤ acts as a tumor suppressor. Arf induction by oncogenic RAS, E1A, or Myc in cultured cells occurs over ϳ48 h (4, 5), a time course that is similar to induction by Tgf␤ (Fig. 3A). It is interesting that Tgf␤s were initially described in oncogene-transformed fibroblasts (53,54). Furthermore, using cultured mouse keratinocytes, v-Ras Ha and Tgf␤1 cooperatively induce p19 Arf and p16 Ink4a , which is also encoded at the Arf/Ink4a locus, and base-line p19 Arf expression is decreased in Smad3 Ϫ/Ϫ keratinocytes transduced by v-Ras Ha -transduced cells (55). These findings imply that Arf induction by certain oncogenes might be driven by an autocrine or paracrine loop. In our analysis of cultured MEFs, RAS V12 does not require T␤rI/Smad signaling, although full Arf induction by ectopic RAS depends on p38 MAPK. Because tissue-specific factors control Arf induction by oncogenic Ras (56), a more robust investigation into the relationship between Arf regulation by Tgf␤ and by different oncogenes will require analysis of a variety of cell types and contexts. A, a schematic diagram shows Distal, Proximal. and Intron regions of Arf locus used in ChIP assays. B, shown is quantitative analysis of representative ChIP assays of using wild type MEFs exposed to vehicle (V) or Tgf␤2 (T2) for 1.5 h (a-c) or 24 h (c) after serum deprivation. A ChIP assay was carried out using antibodies specific to Smad2/3 (a), histone H3 acetylated at lysines 9 and 14 (H3-Ac) (b), and RNA polymerase II (RNA Pol II) (c). Immunoprecipitated DNA and input DNA were amplified with primers for genomic regions indicated in A or for the Tgf␤2 non-responsive gene Necdin. p values are as follows: a, 0.081 (*), 0.001 (#), 0.55 (@); b, 0.045 (*), 0.183 (#), 0.738 (@); (c) 0.059 (*) for Tgf␤2 versus corresponding vehicle.
The anti-tumor effects of Tgf␤ have been well described (57,58), and disruption of components of the pathway is particularly common in certain cancers, like those involving the cervix, gastrointestinal epithelium, and liver (59 -61). In other contexts, Tgf␤ appears to promote tumorigenesis (58,62). One might reconcile these apparent discrepancies if the presence or absence of Arf, which is required for anti-proliferative effects of Tgf␤2 in the eye, also determines the effectiveness of Tgf␤mediated tumor suppression. Shown is quantitative analysis of real time, RT-PCR using total RNA isolated from wild type MEFs exposed to vehicle (Veh) or Tgf␤2 for the indicated times. Differences in transcript level between Tgf␤2-and vehicle-treated MEFs are significant for the mature transcript at 48 and 72 h (p Ͻ 0.04 (*) and Ͻ 0.005 (#)) and for the primary transcript at 72 h (p ϭ 0.039 (@)). FIGURE 9. p38 MAPK signaling is needed for Arf induction by ectopically expressed RAS, but inhibition of T␤rI does not. Representative Western blot for the indicated proteins using lysates from wild type MEFs, exposed to the indicated chemical inhibitors at the time of transduction using Gfp-or H-RAS V12 -expressing retrovirus. Lysates were collected 48 h after transduction. Fluorescence microscopy detection of Gfp from the murine stem cell virus-based vectors showed no differences in transduction efficiency across the samples.