Expression of the Arf tumor suppressor gene is controlled by Tgfβ2 during development

It is now well recognized that p19^Arf, encoded by Arf(Cdkn2a - Mouse Genome Informatics) at the mammalian Arf/Ink4a locus (Kamijo et al.,1997), has a range of p53-dependent and -independent effects that contribute to its anti-cancer activity(Sherr, 2006). However, how the expression of p19^Arf is regulated is much less well understood. It was first discovered to be induced in cultured mouse embryo fibroblasts(MEFs) by what has been termed `culture shock'(Kamijo et al., 1997; Sherr and DePinho, 2000) and by the expression of certain oncogenes (de Stanchina et al., 1998; Palmero et al., 1999; Zindy et al., 1998). Implicit in the `oncogene sensor' model for p19^Arf is that Arfexpression is controlled by cell-intrinsic mechanisms. However, this concept has not been rigorously tested. We recently discovered that Arf is also expressed in a subset of perivascular cells enveloping the hyaloid vasculature in the vitreous of the developing eye(Martin et al., 2004; McKeller et al., 2002). The hyaloid vessels are unusual because they abruptly involute in the postnatal period, and Arf is required for this process. It seemed unlikely that cell-intrinsic responses to oncogenic signals would explain this restricted expression pattern.

As we explored candidate factors that might control Arf, we focused on members of the Transforming growth factor β (Tgfβ)family. Tgfβs were initially identified in mammalian fibroblasts transformed with murine or feline sarcoma viruses and in cultured human melanoma cell lines as proteins that alter fibroblast morphology, and promote proliferation and growth in soft agar (De Larco and Todaro, 1978; Marquardt et al., 1983; Marquardt et al., 1984). The three mammalian Tgfβs transduce signals through a heteromeric complex containing type I and II Tgfβ receptor serine/threonine kinases(TβrI and TβrII; Tgfbr1 and Tgfbr2, respectively - Mouse Genome Informatics); binding of Tgfβ to the receptor complex activates a variety of pathways via both Smad-dependent and Smad-independent signals(Bierie and Moses, 2006; Derynck and Zhang, 2003; Massague et al., 2000). Among their effects is the capacity to arrest cell proliferation and block cancer progression, which appears to contrast the original observations. Of the three members in this family, Tgfβ2 was the most interesting because both Arf^-/- (McKeller et al., 2002; Martin et al.,2004) and Tgfb2^-/-(Saika et al., 2001; Sanford et al., 1997) mice display primary vitreous hyperplasia during embryonic development. We investigated the possible relationship between Tgfβ2 and Arfusing complementary cell culture-based models and in vivo models.

Mice in which Arf exon 1β was inactivated(Kamijo et al., 1997) or replaced by a reporter gene encoding green fluorescence protein (Gfp)(Zindy et al., 2003) or encoding β-galactosidase (made in essentially the same way, A.C.M. and S.X.S., see Fig. S5 in the supplementary material) were maintained in a mixed C57BL/6 × 129/Sv genetic background. Tgfb2^+/- mice(Sanford et al., 1997) were purchased from Jackson Laboratories. Primary MEFs from wild-type and Arf^-/- mice were cultivated as previously described(Zindy et al., 1997). Animal studies were approved by the St Jude Children's Research Hospital and the University of Chicago Animal Care and Use Committees.

For in vivo studies, tissue was harvested from euthanized mice and fixed for 4 hours in 1% formaldehyde, 0.2% glutaraldehyde, 0.2% NP-40 and 0.1% SDS;washed in PBS; and stained in 1 mg/ml X-gal, 5 mM K₃Fe(CN)₆, 5 mM K₄Fe(CN)₆, 2 mM MgCl₂, 0.2% NP-40 and 0.1% SDS overnight at 30°C. Tissue was then photographed or processed for either cryostat or microtome sections, used for hematoxylin and eosin (H&E) or immunofluorescence staining for specific proteins.

Gfp, p19^Arf, Pdgfrβ and TβrII were detected by immunostaining mouse tissue as previously described(Silva et al., 2005) using rabbit anti-Gfp (A6455, Molecular Probes); rat anti-p19^Arf(Bertwistle et al., 2004)(provided by C. J. Sherr); goat anti-Pdgfrβ (AF1042, R&D Systems);goat anti-TβrII (SC 33931, Santa Cruz); rabbit anti-phosphoserine 465/467 Smad2 (3849, Chemicon); and species-specific secondary antibodies.

Cell density in the primary vitreous was assessed using ImagePro Plus software as follows: digital photomicrographs of H&E-stained sections were used to select an area of interest (AOI) encompassing the entire posterior chamber (see Fig. S1 in the supplementary material). The number of pixels within this AOI was noted. Next, all cells (excluding erythrocytes) within the AOI were highlighted, and the number of highlighted pixels was calculated. The cellularity was then determined by dividing the cell pixel number by the total pixel number in the AOI, and it was presented as a percentage. Staining for Ki67 (Mki67 - Mouse Genome Informatics) and TUNEL labeling were performed and quantified such that the number of stained cells was normalized to the vitreous area, determined using ImagePro Plus software as previously described(Silva et al., 2005). The average area of the vitreous in midline sections used in the quantitative studies was the same in wild-type and Tgfb2^-/- eyes at this point: 1.48 × 10⁵ versus 1.66 × 10⁵pixels, respectively (P=0.537).

Stained embryos or slides were photographed using an Olympus BX60 microscope equipped with a SPOT RT Slider camera (Diagnostic Instruments) or using a Zeiss 510 NLO multiphoton/confocal laser scanning microscope.

Wild-type and Arf^-/- MEFs (passage 2) were treated in parallel 1 day following plating, with or without Tgfβ2 (0, 1, 5, 10 ng/ml), Tgfβ1 (5 ng/ml), Pdgfb (50 ng/ml) or an equivalent volume of the relevant vehicle for 24, 48 and 72 hours, at which times cells were harvested for protein or RNA extraction, or cell cycle analysis. Cycloheximide (100μM, Sigma) and SB431542 (10 μM, Tocris Cookson) were used in some studies.

Protein extraction and immunoblotting were performed essentially as described (Zindy et al., 1998)using 4-12% gradient gels (XCell II electrophoresis apparatus, Novex, San Diego, CA, USA) and the following primary antibodies: rabbit anti-p19^Arf (ab80, Abcam, UK); anti-phosphoserine 465/467 (CS3101,Cell Signaling Technology); or goat anti-Hsc70 (Hspa8 - Mouse Genome Informatics; K-19, Santa Cruz Biochemicals, CA, USA) to control for protein loading. Arf^-/- 10T1/2 cells, transduced with Gfp- or Arf-encoding retrovirus(Silva et al., 2005), were included as negative and positive controls.

Total RNA extraction and cDNA preparation were accomplished using RNeasy(Qiagen) and Superscript III RT (Invitrogen, MD, USA) according to the manufacturer's recommendations. Quantitative RT-PCR (qRT-PCR) was performed using the TaqMan ABI PRISM 7700 Sequence Detection System (Applied Biosystems,CA, USA) and gene-specific primers for Arf:5′-TGAGGCTAGAGAGGATCTTGAG-3′ (forward);5′-CGTTGCCCATCATCATCAC-3′ (reverse); and 5′-CGTTGCCCATCATCATCAC-3′ (qRT probe) [designed using the Primer Expression program (Applied Biosystems)]. Gapdh primers and probes were purchased from Applied Biosystems. Primers and probes were used at 200 nM and 50 nM, respectively. RT-PCR reactions were shown to amplify single bands by agarose gel electrophoresis. QRT-PCR data are from three replicate experiments.

LacZ expression in MEFs, prepared and cultivated as previously described (Zindy et al.,1997), was assessed using X-gal staining (see above) following fixation in 0.5% glutaraldehyde, or by measuring β-galactosidase activity in cell lysates prepared by sequential freeze/thaw in 0.25 M Tris pH 7.8,using Galacto-Light Plus (PE Biosystems).

For cell cycle analyses, cells were incubated with BrdU, harvested with trypsin/EDTA, fixed, permeabilized and costained using anti-BrdU antibody(Sigma) followed by FITC-coupled anti-mouse antibody (DAKO) and propidium iodide (Sigma; 10 μg/ml) and analyzed using either a FACScan or FACSCalibur(BD Biosciences, San Jose, CA, USA). Relative cell cycle change represents the percent difference in the fraction of Tgfβ2-treated cells in a particular phase as compared with vehicle treated cells.

Arf^lacZ/lacZ MEF cells (3 ×10⁶/immunoprecipitation) treated with Tgfβ2 (5 ng/ml) or vehicle for 1.5 hours were crosslinked, sonicated and immunoprecipitated with anti-Smad2/3 antibody (sc6033, Santa Cruz) or IgG (AB-108-C, R&D Systems)as a control. Protein A/G sepharose beads (sc2003, Santa Cruz) were used to collect the protein-chromatin complexes. The beads were washed sequentially with low salt, high salt, LiCl and TE buffers (Upstate ChIP Kit, Millipore)and eluted in 0.1 M NaHCO₃, 1% SDS. Crosslinking was reversed by incubation at 67°C overnight and the genomic DNA was extracted with a Qiagen PCR Purification Kit. A total of 3% of the precipitated DNA and 1%input DNA was amplified by PCR using primer sets for Id1 and different regions of Arf (primer sequences are available upon request). The PCR products were resolved on 1.5 % agarose gels and stained with ethidium bromide.

All quantitative studies were accomplished using three or more samples from separate animals, or three or more separate cell culture experiments. Non-quantitative results from histological studies are representative of findings using at least three embryos of the indicated genotypes, obtained from two or more separate experiments. Statistical significance of any quantitative differences was assessed by Student's t-test.

Currently, the only established developmental activity of p19^Arfis to control the accumulation of cells within the vitreous(Silva et al., 2005). If it acted with Tgfβ2 during this process, we reasoned that the developmental defects in the absence of Arf should parallel those in Tgfb2^-/- embryos. Without Arf, increased numbers of vitreous cells are detectable at embryonic day (E) 13.5, approximately 1 day after its expression is evident (Silva et al., 2005). At that point, there is excess proliferation in the vitreous, a defect that is intrinsic to the Arf-expressing cells(Thornton et al., 2007). Notably, Arf loss does not measurably change the small number of apoptotic cells (Silva et al.,2005). Like Arf^-/- mice, Tgfb2^-/- mice displayed obvious primary vitreous hyperplasia at E18.5 and at birth, whereas Tgfb2^+/- and Tgfb2^+/+ eyes were normal(Fig. 1A; see Fig. S2 in the supplementary material). Quantitative analyses of younger embryos revealed increased cell density in the primary vitreous of Tgfb2^-/-versus wild-type embryos as early as E13.5, and this was associated with increased cell proliferation (Fig. 1A,B; see Fig. S3 in the supplementary material). As in Arf^-/- embryos (Silva et al., 2005), the average percentage of apoptotic cells(3.1±3.8% versus 2.5±2.1%, n=∼350 cells counted for each genotype, P=0.81) was similar in the presence or absence of Tgfβ2, respectively. Proliferation continued in the hyperplastic primary vitreous of Arf^-/- and Tgfb2^-/-embryos at E18.5 (Fig. 1C). Lastly, nearly all of the hyperplastic vitreous cells expressed Pdgfrβ in Tgfb2^-/- embryos (Fig. 1D), again mimicking the findings in Arf^-/-embryos (Silva et al., 2005). Hence, loss of Tgfb2 causes hyperplastic expansion of Pdgfrβ-expressing cells from early stages of primary vitreous formation,as in Arf^-/- embryos.

To begin to address whether Tgfb2 could lie `upstream' of Arf in a linear pathway, we explored how the presence or absence of Arf influenced mitogenic or anti-mitogenic effects of Tgfβ in cultured MEFs. Exposure of wild-type MEFs to Tgfβ2 for 24 and 72 hours decreased the fraction of cells in S phase by 15-25%(Fig. 2A, lanes 5 and 7). This was balanced by increases in the G₀/G₁ fraction at both times and in the G₂/M fraction at 24 hours(Fig. 2A, lanes 1, 3 and 9). Tgfβ2-dependent changes in Arf^-/- MEFs paralleled those in wild-type cells at 24 hours (Fig. 2A, compare lanes 1 and 2; 5 and 6; 9 and 10). However, the pronounced block of cells in S phase and the accumulation of cells in G₁ at 72 hours were not maintained without p19^Arf(Fig. 2A, compare lanes 3 and 4; 7 and 8; 11 and 12). The requirement for Arf correlated with p19^Arf induction by Tgfβ2 at 48 and 72 hours(Fig. 2B, lanes 2 and 3 versus lane 1; see Fig. S4A in the supplementary material); p19^Arf was not induced at 24 hours (see Fig. S4B in the supplementary material). The ability of p19^Arf to maintain Tgfβ2-driven growth arrest may depend on p53 (Trp53 - Mouse Genome Informatics), which cooperates with Tgfβ1 to control proliferation in MEFs (Cordenonsi et al., 2003). However, we did not observe Arf-dependent induction of p21^Waf1 (Cdkn1a - Mouse Genome Informatics), a well-described p53 target, or other Cdk inhibitors (see Fig. S4A in the supplementary material). Moreover, p21^-/- mice did not display vitreous hyperplasia (see Fig. S4C in the supplementary material), and it is hence not the main Tgfβ2 target in the eye.

We investigated the mechanistic aspects of Tgfβ2-driven p19^Arf induction. We first explored whether p19^Arfinduction correlated with increased Arf mRNA and promoter activity. The latter was addressed using a new reporter mouse, in which Arfexon 1β was replaced by lacZ cDNA encoding theβ-galactosidase reporter; X-gal staining in the mouse recapitulated the previously described Arf expression in the eye and the testis(Fig. 2C; see Fig. S5 and Fig. S6A,B in the supplementary material). Like Gfp expression in the Arf^Gfp/+ mouse (Martin et al., 2004; Zindy et al.,2003), β-galactosidase activity increased as Arf^lacZ/lacZ MEFs were cultivated using a 3T9 protocol(see Fig. S6C in the supplementary material), and as they grew more confluent(Fig. 2D, lanes 2, 5, 8),paralleling previous findings with native p19^Arf(Kamijo et al., 1997; Sharpless et al., 2004). We observed that Tgfβ2 increased Arf mRNA in wild-type MEFs and Arf promoter transcription in Arf^lacZ/lacZ MEFs at 72 but not 24 hours, correlating with changes at the protein level(Fig. 2D, lanes 8 and 10 versus 2 and 4; see Fig. S7A in the supplementary material), although the reporter detected Arf induction earlier than the qRT-PCR assay. Lastly,p19^Arf expression fell following cycloheximide exposure in Tgfβ2-treated MEFs at least as rapidly as in the control (see Fig. S7B in the supplementary material). Hence, Tgfβ2 controls p19^Arfexpression by inducing its promoter without measurably changing the stability of the protein.

Next we addressed whether Tgfβ1 or platelet derived growth factor B(Pdgfb) shared the capacity to induce Arf in MEFs. The former was tested because it can reverse primary vitreous hyperplasia in Tgfb2^-/- mice (Zhao and Overbeek, 2001), whereas the latter acts as a potent mitogen in MEFs (Silva et al., 2005). All of the effects of Tgfβ1 on Arf expression paralleled those of Tgfβ2 (Fig. 2D, lanes 3, 6, 9 versus 4, 8, 10; see Fig. S7A in the supplementary material). By contrast, Pdgfb did not induce the reporter at 48 or 72 hours(Fig. 2D, lanes 12 and 13;negative data not shown). Therefore, Arf induction in this cell culture model does not simply represent a response to supraphysiological mitogens.

We also tested if the observed effects depended on Smad2 because this signaling protein was phosphorylated in response to Tgfβ2 in MEFs(Fig. 2E). To test this, we exposed the cells to SB431542, an inhibitor of TβrI(Seay et al., 2005). This chemical inhibited Smad2 phosphorylation at serine 465/467 and blocked both p19^Arf and β-galactosidase induction in wild-type and Arf^lacZ/lacZ MEFs, respectively(Fig. 2F,G). We also used a genetic approach to exclude the possibility that this might represent an off-target effect of the chemical. The ectopic expression of the inhibitory Smad6 (Derynck and Zhang,2003) impeded Arf promoter activation by Tgfβ2(Fig. 2H), proving that Smad-dependent signals stemming from the Tgfβ receptor drive p19^Arf expression.

Next, we considered whether Smads directly influence the Arfpromoter because multiple potential Smad binding elements (SBEs) flank Arf exon 1β (Fig. 3C). Chromatin immunoprecipitation (ChIP) using an antibody recognizing Smad2 and Smad3 is described to show Smad binding at promoters like that driving mouse Id1 expression(Smith et al., 2009). In our experiments, Smad2/3 binding was evident at baseline levels and it increased 1.5 hours following exposure of Arf^lacZ/lacZ MEFs to Tgfβ1 (Fig. 3A). To interrogate the Arf gene, we designed 12 PCR primer sets spanning the-2.2 kb to +1.1 kb region flanking exon 1β(Fig. 3B,C; primer sequences are available upon request). ChIP assay shows that Smad2/3 binding was observed at amplicons 1 and 3 proximal to the first exon in Arf^lacZ/lacZ MEFs 1.5 hours following the addition of Tgfβ1 (Fig. 3C,D). Binding at amplicons 2 and 4 was also observed, but the signal was weaker. Selective Smad2/3 binding to the chromatin was not convincing at other regions,including those with putative SBEs. Interestingly, the relatively small amount of binding observed at amplicon 2 suggests that Smad2/3 might bind to two distinct regions in amplicons 1 and 3. Thus, although increased Arfexpression was not readily detected until 48 hours following the addition of either Tgfβ1 or 2, increased Smad2/3 binding is present in the Arf promoter shortly after Tgfβ exposure.

We felt it was important to verify that our findings in MEFs were relevant to Arf regulation in the mouse eye. Immunofluorescence staining detects p19^Arf in subnuclear foci in the mouse testis(Bertwistle et al., 2004). p19^Arf was similarly detected in vitreous cells in Tgfb2^+/+ and Tgfb2^+/- embryos at E13.5, but not in the Tgfb2^-/- littermates stained in parallel (Fig. 4A). To determine whether decreased p19^Arf represented decreased transcription in vivo, we employed Arf^Gfp/+ and Arf^lacZ/+ reporter mice. As expected, vitreous hyperplasia developed at E15.5 in Arf^Gfp/Gfp embryos that were heterozygous for Tgfb2, and nearly all of the vitreous cells expressed the Arf promoter (Fig. 4B, panels a and b), consistent with our observations in Arf^lacZ/lacZ embryos(Fig. 2C). Hyperplasia was also evident in Arf^Gfp/Gfp Tgfb2^-/- embryos(Fig. 4B, panel c) (formally establishing that Tgfβ2 does not provide essential mitogenic signals to drive vitreous hyperplasia), but Gfp staining showed Arf promoter activity to be markedly decreased (Fig. 4B, panel d). Similar observations using Tgfb2^-/-Arf^lacZ/+ embryos at E13.5(Fig. 4C) proved that the effects of Tgfβ2 on Arf expression were not specific to a single reporter and did not depend on an Arf-deficient state. Consistent with potentially direct signaling, immunofluorescence staining revealed that Gfp and TβrII expression overlapped in some of the vitreous cells of Arf^Gfp/+ embryos (Fig. 4D). Furthermore, phosphoSmad2 was present in nuclei in the primary vitreous in E13.5 wild-type embryos and throughout the retrolental mass in newborn Arf^-/- mice(Fig. 4E; additional data not shown). Because essentially all of the cells in the retrolental mass expressed the Arf gene in the embryo (Fig. 4B) and in the newborn period(Martin et al., 2004), we conclude that Tgfβ2 directly impacts cells expressing Arf.

Finally, a survey of the Arf^lacZ/+ embryos revealed two new sites where Arf is expressed during development: within the stroma of the developing cornea at E12.5 and E13.5(Fig. 5Aa,b) and around the umbilical arteries within the abdominal/pelvic cavities between E13.5 through to the early postnatal period (Fig. 5C; see Fig. S8A in the supplementary material). Expression in the cornea stroma at this stage is interesting because Tgfβ2 induces extracellular matrix proteins such as collagen I and lumican, enhancing cornea thickness (Saika et al.,2001). However, there is no obvious thinning in the Arf^-/- cornea (see Fig. S9 in the supplementary material). Arf expression in the umbilical arteries, which are essential only during embryo development (like the hyaloid vasculature), parallels the pattern in the vitreous in that the cells are perivascular in location and some coexpress Pdgfrβ (Fig. 5Cc; see Fig. S8B in the supplementary material). However, aspects of umbilical artery biology appeared unaffected by Arf loss. For example, the atresia that typically develops in the left umbilical artery between E13.5 and E14.5 (Warot et al.,1997) still occurs without p19^Arf(Fig. 5Cb). In both the cornea and the umbilical arteries, absence of Tgfβ2 diminished Arfexpression (Fig. 5Ac,d; Fig. 5B,D). Although the functional relevance of this pathway in umbilical artery and cornea development is not yet clear, we can confidently conclude that Tgfβ2-mediated control of Arf promoter activity extends beyond the vitreous of the eye.

The clear role that Arf plays in mouse development implies that its control must extend beyond that provided by the deregulated signals accompanying the activation of certain oncogenes. Whereas this regulatory paradigm focuses on cell-intrinsic signaling, our findings provide the first evidence that Arf expression can be controlled by specific cell-extrinsic signals from Tgfβ2 during development and in MEFs. We can now begin to integrate our findings with emerging knowledge regarding the regulation of p19^Arf (Gil and Peters, 2006; Kim and Sharpless, 2006).

First, although there is some evidence that its expression is controlled by post-transcriptional mechanisms (Colombo et al., 2005), our findings using two different reporter systems indicate that the Tgfβ-dependent induction depends largely, and perhaps exclusively, on transcriptional activation. Unfortunately, regulatory mechanisms guiding transcription of Arf and the flanking Ink4a (Cdkn2a - Mouse Genome Informatics) and Ink4b(Cdkn2b - Mouse Genome Informatics) genes are complex and incompletely understood. The activity of certain Polycomb group proteins(notably Bmi1) leads to the methylation and silencing of Arf, Ink4aand to a lesser extent Ink4b in MEFs(Jacobs et al., 1999). In the developing mouse, the relative importance of Bmi1-mediated repression of specific genes at this locus is cell type-dependent(Bruggeman et al., 2005; Molofsky et al., 2005). The entire Ink4b/Arf/Ink4a locus can also be silenced by binding of the DNA replication protein Cdc6, leading to the recruitment of histone deacetylases to a regulatory domain in the Ink4b promoter(Gonzalez et al., 2006). Of note, neither p16^Ink4a nor p15^Ink4b were induced with p19^Arf by Tgfβ2 in our model (see Fig. S4A in the supplementary material), which speaks against Tgfβ2-dependent modification of the entire gene locus. Certain other transcription factors such as Tbx2 (Jacobs et al.,2000), E2f1 and E2f3 (Aslanian et al., 2004; DeGregori et al., 1997), Dmp1 (Inoue et al., 1999) and pokemon (Zbtb7a - Mouse Genome Informatics)(Maeda et al., 2005) act on sites found close to exon 1β. Others act more remotely; for example, FoxO family members bind a regulatory element approximately 8 kb 3′ of exon 1β (Bouchard et al.,2007). Mechanisms by which other transcription factors like p53(Kamijo et al., 1998), Myc(Zindy et al., 1998) and Twist(Maestro et al., 1999)positively or negatively regulate mouse Arf are even less clear. Despite the fact that our ChIP studies show Smad2/3 binding to sites flanking Arf exon 1β, neither Tgfβ1 nor Tgfβ2 induced a plasmid reporter containing ∼2.5 kb 5′ and ∼1.3 kb 3′ to exon 1β transiently transfected into mouse 10T1/2 fibroblasts [which lack Arf (Thornton et al.,2007)], and into WT MEFs (negative data not shown). These negative data might imply that Arf induction is mediated by mechanisms that are not reflected in a transiently expressed reporter. They could also indicate that Smad-dependent changes near the promoter must cooperate with other cis regulatory elements that are further removed from the promoter.

Another important mechanistic issue relates to exactly how Tgfβ2 signals get to the Arf promoter. We have some insights, which are as follows: first, our detection of both TβrII and phosphoSmad2 in Arf-expressing cells in vivo indicates that Tgfβ2 signaling can directly impact the cells that express the Arf promoter; second,because Arf induction correlates with Smad2 phosphorylation and can be blocked by chemical inhibition of TβrI and the ectopic expression of Smad6, it seems safe to conclude that a Smad-dependent process is required in MEFs; lastly, our ChIP assay using a Smad2/3-specific antibody showed that Smad proteins directly bind to the Arf gene. That Smad2 phosphorylation is detected in Arf-expressing cells in the embryo suggests but does not prove it to be the crucial Smad in vivo. Its candidacy is strengthened by the fact that Smad3^-/- mice, which are viable, appear to have normal eyes (Banh et al., 2006; Zhu et al.,1998). Formally evaluating the role of Smad2 may require its knockout in cells destined to express the Arf promoter because Smad2^-/- mice suffer early embryonic lethality(Nomura and Li, 1998; Weinstein et al., 1998). The genetically engineered mice needed to accomplish this experiment are not currently available.

Certain mechanistic facts must still be elucidated. Like in other scenarios, we suspect that Smads cooperate with other transcription factors to enhance Arf transcription. The region bound by Smad2/3 in our experiments contains many putative transcription factor binding sites,including potential sites for C/EBPβ (also known as Cebpβ), a known Smad-interacting protein (Ross and Hill,2008) (TFSEARCH result; http://www.cbrc.jp/htbin/nph-tfsearch). However, its role in Arf regulation is not evident from the literature. The rapid localization of Smad2/3 to the promoter, and the delayed increase in measurable Arf expression, challenges us to more broadly consider the role Smads might play in this process. For example, Smad binding might initiate a cascade of events that make the chromatin more accessible to other transcriptional activators, even those like FoxO that act at a distance. As such, gaining insight into the Tgfβ-dependent changes in DNA methylation and histone status is likely to be informative. Lastly, although Smad2/3 binding to the promoter is detectable, given the fact that Smad6 preferentially inhibits BMP signaling(Goto et al., 2007; Kirkbride et al., 2008), it is possible that BMPs; Smads 1, 5 and 8; and TβrIII might also contribute to Arf regulation.

Our findings shed light on the molecular basis for the crucial developmental process guiding the maturation of the primary vitreous into the avascular and largely acellular secondary vitreous, a developmental process that is essential for normal vision, and this might be relevant to human eye disease. We previously established that without Arf, vision is severely compromised because the vitreous hyperplasia obscures the optic lens and destroys the retina (Martin et al.,2004). This ocular defect resembles severe PHPV, a disease long known to be caused by failed maturation of the primary vitreous and persistence of the hyaloid vessels(Goldberg, 1997; Haddad et al., 1978). Tgfβ2 resides upstream of Arf in this developmental program. Interestingly, this observation is also consistent with a recent finding that selective inactivation of TβrII using Cre recombinase expressed from a neural crest-specific promoter appears to mimic the Arf^-/-eye phenotype (Ittner et al.,2005). Based on these mouse models, we suggest that a search for the molecular basis for the human disease should include studies of human ARF, TGFB2 and the yet-to-be-defined components of the signaling pathway.

Tgfβ2 loss leads to a complex developmental phenotype that includes multiple craniofacial, cardiac, pulmonary, urogenital and skeletal defects (in addition to the ocular abnormalities mentioned here)(Sanford et al., 1997). Molecular mechanisms underlying these other defects are still elusive;certainly, the absence of these other defects in Arf^-/-mice indicates that, unlike the primary vitreous hyperplasia, they do not arise merely due to failed induction of Arf. This also implies that,although Tgfβ2 is required for Arf expression at certain sites,it is not sufficient at others. We are taking advantage of our complementary cell culture-based model and the Arf^lacZ/+ reporter mouse to identify putative cooperating signaling proteins. More refined molecular or physiological studies will be required to determine the functional consequences of loss of Arf in the cornea and umbilical artery where the Tgfβ2/p19^Arf pathway is intact but anatomic abnormalities are not apparent in the absence of Arf.

Beyond development and eye disease, our finding of a linear pathway between Tgfβ and p19^Arf has potential implications for the tumor suppressor actions of these proteins. It is interesting that both Tgfβ1 and p19^Arf block early papilloma formation triggered by DMBA/TPA in the mouse (Cui et al., 1996; Kelly-Spratt et al., 2004);however, Tgfβ1 does not hinder progression of the papilloma to an invasive cancer (Cui et al.,1996). Conceptually, anti-cancer effects of Tgfβ in this model could be a result of p19^Arf induction, and Arfdeletion might coincide with loss of the tumor suppressive effects of Tgfβ. Dynamic regulation of p19^Arf at different phases of cancer development, like the sequential induction of Arf during Myc-driven lymphomagenesis, has been attributed to the accumulation of additional oncogenic events (Bertwistle and Sherr, 2007). Our observations support an alternative hypothesis that extracellular signals from Tgfβ, derived from either tumor cells or stroma elements, might contribute to Arf regulation at different stages of tumor development and progression.