Arf-dependent regulation of Pdgf signaling in perivascular cells in the developing mouse eye

Abstract

We have established that the Arf tumor suppressor gene regulates mural cell biology in the hyaloid vascular system (HVS) of the developing eye. In the absence of Arf, perivascular cells accumulate within the HVS and prevent its involution. We now demonstrate that mural cell accumulation evident at embryonic day (E) 13.5 in Arf^−/− mice was driven by excess proliferation at E12.5, when Arf expression was detectable in vitreous pericyte-like cells. Their expression of Arf overlapped with Pdgf receptor β (Pdgfrβ), which is essential for pericyte accumulation in the mouse. In cultured cells, p19^Arf decreased Pdgfrβ and blocked Pdgf-B-driven proliferation independently of Mdm2 and p53. The presence of a normal Arf allele correlated with decreased Pdgfrβ in the embryonic vitreous. Pdgfrβ was required for vitreous cell accumulation in the absence of Arf. Our findings demonstrate a novel, p53- and Mdm2-independent function for p19^Arf. Instead of solely sensing excessive mitogenic stimuli, developmental cues induce Arf to block Pdgfrβ-dependent signals and prevent the accumulation of perivascular cells selectively in a vascular bed destined to regress.

Introduction

Cellular and molecular interactions between mural cells and underlying endothelial cells are key determinants of vascular integrity and remodeling (reviewed in Hanahan and Folkman, 1996; Yancopoulos et al, 2000). Mural cells are often broadly referred to as either pericytes, defined by their direct apposition to the abluminal side of endothelial cells beneath the endothelial cell basement membrane, or vascular smooth muscle cells (reviewed in Sims, 1986; Beck and D'Amore, 1997; Gerhardt and Betsholtz, 2003). Mural cell accumulation largely depends on endothelial sources of platelet-derived growth factor (Pdgf)-B (Hellstrom et al, 1999; Gerhardt et al, 2003; Lindblom et al, 2003) and the expression of Pdgf receptor β (Pdgfrβ) in mural cells or their progenitors (Benjamin et al, 1999; reviewed in Hoch and Soriano, 2003). The critical importance of Pdgf signaling is evidenced by the fact that mice lacking Pdgf-B or Pdgfrβ have an embryonic lethal phenotype due to microvascular hemorrhage and edema caused by lack of pericytes (Leveen et al, 1994; Soriano, 1994; Lindahl et al, 1997; Hellstrom et al, 2001).

The preponderance of evidence supports the notion that mural cell investment of maturing vessels governs a ‘plasticity window'—in their absence, existing vessels can regress (Benjamin et al, 1998) or remodel by forming angiogenic sprouts (Gee et al, 2003). Mural cells preserve vascular stability by providing mechanical support to underlying endothelial cells (reviewed in Sims, 1986) and by preventing endothelial cell proliferation (Orlidge and D'Amore, 1987; Hellstrom et al, 2001) and migration (Sato and Rifkin, 1989). Surprisingly, pericyte coverage of individual blood vessels varies widely (reviewed in Sims, 1986), perhaps to confer greater or less plasticity. Understanding regulatory mechanisms for mural cell recruitment may reveal how certain vascular beds are stabilized (or destabilized) during development and how this can go awry in disease.

We recently uncovered evidence that a key mammalian tumor suppressor p19^Arf controls perivascular cell accumulation in the postnatal mouse eye (McKeller et al, 2002; Martin et al, 2004). p19^Arf, encoded by the Arf gene resident at the Ink4a/Arf locus in the human and murine genomes, was initially demonstrated to physically interact with and inhibit the function of Mdm2, a negative regulator of the p53 tumor suppressor. The net effect of Arf expression is p53 activation to induce cell proliferation arrest or apoptosis (reviewed in Sherr, 1998). More recent findings indicate that p19^Arf activity does not entirely depend on p53. For example, ectopically expressed p19^Arf arrests proliferation in mouse embryo fibroblasts (MEFs) lacking p53 and Mdm2 (smaller effects are observed in cells with Mdm2) (Weber et al, 2000). The tumor spectrum also differs in p53^−/− (lymphoma predominance) (Donehower et al, 1992; Jacks et al, 1994) and Arf^−/− (sarcoma predominance) (Kamijo et al, 1997, 1999) mice and in mice lacking both Arf and p53 (broader types; multiple tumors) (Weber et al, 2000). p19^Arf can inhibit ribosomal RNA processing (Sugimoto et al, 2003), NF-κB activation (Rocha et al, 2003), and Myc-mediated transactivation (Qi et al, 2004) independently of p53. The relative importance of these activities to the tumor suppression by p19^Arf is not yet known.

Our previous finding that p19^Arf blocks perivascular cell accumulation within the vasa hyaloidea propria (VHP), the vitreous component of the hyaloid vascular system (HVS), was somewhat surprising. Existing paradigms hold that p19^Arf is an oncogene sensor induced by abnormal or excessive mitogenic stimuli (de Stanchina et al, 1998; Zindy et al, 1998, 2003; Palmero et al, 1999). Arf expression is undetectable or very low through most of mouse embryo development (Zindy et al, 1997, 2003), consistent with the idea that active Arf repression prevents untoward p19^Arf-mediated effects during embryogenesis. That Arf-deficient mice have a developmental eye disease evident at birth implies that Arf is not universally repressed in the developing mouse embryo and that its gene product functions beyond tumor suppression.

The present work elucidates one such function. We show that Arf is expressed in a subset of perivascular cells in the eyes of developing mouse embryos. Its restricted expression suggests that the Arf promoter is not only induced by aberrant mitogens but also responds to specific developmental cues. During embryonic development, p19^Arf controls perivascular cell accumulation in the vitreous by blocking proliferation driven by Pdgfrβ in the Arf-expressing cells. Our in vivo and in vitro analyses show that p19^Arf may accomplish this by dampening Pdgfrβ expression, even in cells lacking Mdm2 and p53. Hence, our results extend existing paradigms for p19^Arf biology by showing that Arf is expressed in a strikingly restricted pattern to couple developmental signals to p53-independent effects and block Pdgfrβ-driven mural cell accumulation in a vascular bed destined to regress in the postnatal period.

Results

Increased cell proliferation in the developing vitreous in Arf^−/− mouse embryos

We previously detected Arf mRNA by RT–PCR and in situ hybridization in the postnatal eye (McKeller et al, 2002; Martin et al, 2004), but Arf expression during embryonic eye development has not been reported. In the present analyses, we took advantage of an engineered mouse in which cDNA encoding the green fluorescent protein (Gfp) replaces Arf exon 1β (Zindy et al, 2003). Like the original Arf^−/− mouse (McKeller et al, 2002), Arf^Gfp/Gfp (effectively Arf^−/−) mice have eye disease while Arf^Gfp/+ mice do not (Zindy et al, 2003). Gfp expression accurately reflects Arf mRNA in Arf^Gfp/+ mice and MEFs derived from them (Zindy et al, 2003). To elucidate putative developmental functions for p19^Arf, we used the Arf^Gfp/+ mouse to define the temporal and spatial expression pattern of Arf in the eye and the earliest developmental defects in Arf^−/− embryos.

In the immediate postnatal period, direct green fluorescence was evident in the retrolental mass in unfixed eyes taken from Arf^Gfp/Gfp mice (Zindy et al, 2003), but relatively weak green fluorescence in developing embryos prevented direct visualization of Gfp. In contrast, antibody-based detection readily showed Gfp expression in the vitreous of Arf^Gfp/+ and Arf^Gfp/Gfp but not wild-type mice evaluated from embryonic day (E) 11.5 to E18.5 (Figure 1A; some data not shown). Occasional Gfp-positive cells were observed along the developing optic tract at E11.5 and the number of vitreous cells with strong Gfp expression seemed to increase from E11.5 to E12.5 in heterozygous embryos (Figure 1A). Gfp-expressing cells were scattered throughout the vitreous from E11.5 to E13.5 and appeared to form linear structures by E18.5 (Figure 1A), consistent with the fact that Arf-expressing cells are perivascular in postnatal mice (Martin et al, 2004). That the Gfp reporter accurately reflected protein expression was verified by immunofluorescence staining using a rat monoclonal antibody directed against p19^Arf (Bertwistle et al, 2004), which labeled vitreous cells in wild-type mice as early as E11.5 (Figure 1B).

Figure 1.

Vitreous cells express the Arf promoter from E11.5 of mouse embryo development. (A, C) Representative immunofluorescence photomicrographs of eyes taken from Arf^Gfp/+ (A, C) and Arf^Gfp/Gfp (C) mouse embryos at the indicated stages of embryonic development (A) and E14.5 (C). Sections stained with DAPI and α-Gfp antibody (red (A); green (C)) as indicated. Note Gfp-positive cells along developing optic tract at E11.5 (arrow) and in the vitreous at each stage. (B) Representative immunofluorescence photomicrographs of E11.5 wild-type mouse eye, following staining with α-p19^Arf antibody. Original magnification: × 200 (A) and × 400 (B, C).

In postnatal Arf^−/− eyes, the principal abnormality is excess accumulation of perivascular cells within the VHP in the retrolental vitreous (McKeller et al, 2002). Relatively few Arf-expressing perivascular cells are present in the VHP in wild-type or Arf^Gfp/+ mice in the postnatal period, whereas the bulk of the cells in the retrolental mass in the Arf^Gfp/Gfp mice express the Arf promoter (Zindy et al, 2003; Martin et al, 2004). Similarly, additional Gfp-expressing cells were evident in developing Arf^Gfp/Gfp versus Arf^Gfp/+ embryos as early as E14.5 (Figure 1C). This implied that p19^Arf functioned early during eye development to check vitreous cell accumulation.

We determined whether the earliest evidence of vitreous cell accumulation correlated with Arf promoter activity in the embryo. Between E12.5 and E13.5, fibroblast-like cells accumulated and persisted in the vitreous of Arf^−/− embryos (Figure 2A), resulting in increased cellular density as compared to heterozygous and wild-type mice (Figure 2B; some data not shown). Differences in cellular density in Arf^−/− and Arf^+/− mice were not apparent at E12.5 (Figure 2A and B). Absence of Arf did not appreciably alter the morphology of the fibroblast-like cells nor did it affect the lens and neuroblastic tissue. Increased cellular density was not likely due to decreased vitreous volume: the vitreous area in sections from E13.5 Arf^−/− eyes (0.42±0.11 mm²) was only slightly smaller than in Arf^+/− and wild-type eyes (0.54±0.17 mm²) (P=0.05). Hence, the increased vitreous cell density primarily represented increased cell number, which became apparent in Arf^−/− mice at E13.5.

Figure 2.

Cellular density and cell proliferation increase in the vitreous of Arf^−/− mice between E12.5 and E13.5. (A) Representative photomicrographs of midline, hematoxylin and eosin (H&E)-stained section of eyes taken from E12.5 and E13.5 Arf^+/− or Arf^−/− mouse embryos. (B) Quantitative analysis of average cellular density (and standard deviation) in the vitreous at E12.5 and E13.5 in Arf^−/− versus Arf^+/− mice. Differences at E13.5 (^*) were statistically significant (P=0.0002) whereas those at E12.5 were not (P=0.55). (C) Representative photomicrographs of vitreous cells from E13.5 Arf^−/− (left) and Arf^+/− (middle) and E12.5 Arf^+/− (right) eyes following staining for Ki67 and TUNEL (arrow), as indicated. (D) Quantitative analysis shows average percentage (and standard deviation) of Ki67-positive cells within vitreous of wild-type, Arf^+/−, or Arf^−/− eyes at E12.5 and E13.5. Differences between Arf^−/− and Arf^+/− or Arf^+/+ were statistically significant at each time point (P<0.002), while differences between Arf^+/− and Arf^+/+ at E13.5 were not (P=0.27). (E) Representative immunofluorescence photomicrograph of vitreous cells from E13.5 Arf^Gfp/+ mouse depicts Gfp (a), DAPI (b), Ki67 (c), and dual Gfp and Ki67 (d). Arrows show Gfp-positive cells, negative for Ki67; arrowheads show Gfp-negative cells, positive for Ki67. Lens (asterisk) provides internal negative control for Gfp and Ki67 stains. Original magnification: × 400 (A, C, E).

Because p19^Arf regulates cell proliferation by both p53-dependent and -independent mechanisms (Kamijo et al, 1997; Weber et al, 2000), we determined whether it arrested vitreous cell proliferation between E12.5 and E13.5. Immunohistochemical staining for the cell proliferation marker Ki67 labeled vitreous cells in wild-type, Arf^+/−, and Arf^−/− eyes at E12.5 and E13.5 (Figure 2C; some data not shown), but the relative number of Ki67-positive cells was significantly higher in Arf^−/− eyes (Figure 2D). With rare exception (see below), the Arf-expressing cells in Arf^Gfp/+ heterozygous embryos were negative for Ki67 (Figure 2E). Few apoptotic cells were identified in the vitreous at E12.5 (three of 120 nuclei counted were TUNEL positive in Arf^+/− versus 0 of 162 in Arf^−/− embryos) (Figure 2C and data not shown). We conclude that p19^Arf controlled vitreous cell accumulation primarily by arresting proliferation from E12.5 although less pronounced proapoptotic effects might also have contributed.

Quantitative analysis at E13.5 suggested a trend toward increased Ki67-positive cells in Arf^+/− versus wild-type mice (Figure 2D). Although no pathology was evident before E13.5 or in postnatal eyes (McKeller et al, 2002), inspection of the Arf^+/− vitreous at E14.5 demonstrated clusters of fibroblast-like cells, present in all heterozygous eyes (n=6) (Figure 3B, arrowhead) but not in wild-type eyes (n=8) (Figure 3A) examined in a blinded fashion. Cells often clustered near linear structures (Figure 3B, arrow) thought to represent developing hyaloid vessels. Dual immunofluorescence labeling for Gfp and CD31, an endothelial cell marker, in Arf^Gfp/+ eyes at E14.5 and E15.5 confirmed that the cellular clusters were composed of Arf-expressing cells (Figure 3C and D, arrowhead) adjacent to CD31-positive vessels (Figure 3C, arrow). Occasional Gfp-expressing cells in the clusters expressed Ki67 (Figure 3D). Hence, Arf haploinsufficiency provided a sufficient genetic insult to allow localized accumulation of Arf-expressing cells adjacent to some developing hyaloid vessels, which may provide a mitogenic signal (see below). But loss of one Arf allele was not sufficient to allow the unchecked vitreous cell proliferation and obvious eye pathology that occurs uniformly in Arf^−/− eyes.

Figure 3.

Arf-expressing cells cluster near developing hyaloid vessels in Arf^Gfp/+ embryos. Representative photomicrographs of H&E-stained paraffin sections (A, B) and immunofluorescence-stained cryostat sections (C, D) through the midline sections of eyes of wild-type (A) and heterozygous E14.5 (B, C) and E15.5 (D) mouse embryos. Note vessel-like structure (arrow (B, C)) near the cluster of vitreous cells (arrowhead (B–D)) in Arf^Gfp/+ eyes. Vessel-like structure expresses CD31 (arrow (C)). Clustered cells express Gfp (green (C); red (D)). Several Gfp-expressing cells in the cluster are Ki67 positive (green (D)). Sections in panels B–D are from separate Arf^Gfp/+ embryos. Original magnification: × 100 (A, B) and × 400 (C, D).

Arf is expressed in mural cells within the developing hyaloid vessels

The timing and localization of Arf-expressing cells in the vitreous and their clustering near developing endothelial tubes in Arf^Gfp/+ eyes suggested that they may represent progenitors of the hyaloid vasculature. We used immunofluorescence staining to determine whether the Gfp-positive cells in the embryonic vitreous were of endothelial or mural cell lineage.

In the postnatal period, cells expressing the Arf promoter in the vitreous of heterozygous or null mice do not express CD31 (Martin et al, 2004). Similarly, expression of Gfp and CD31 did not overlap in the eyes of E15.5 embryos (Figure 4Ab) nor was there coexpression of Gfp and the Vegf receptor Flk-1 in E13.5 Arf^Gfp/Gfp embryos (AC Martin and SX Skapek, unpublished). Thus, we found no evidence that Arf-expressing cells in the vitreous were of endothelial cell lineage.

Figure 4.

Coexpression of the Arf promoter and mural cell proteins in the developing vitreous. (A, B) Representative immunofluorescence photomicrographs of E15.5 (A) and E13.5 (B) Arf^Gfp/+ eyes stained for Gfp (green (A and B, c)), CD31 (red (A, a)), NG2 (red (B, a)), and DAPI (blue (A and B, d)). Panel b shows dual red/green fluorescence. Note lack of overlapping CD31/Gfp expression (A, b) and cell membrane staining for NG2 in some Gfp-expressing cells (arrows (B)). Original magnification: × 400. (C) Representative immunofluorescence (a–c, e–g) and phase-contrast (d) photomicrographs from cryostat sections of an E13.5 Arf^Gfp/Gfp mouse. Gfp expression (green (a, e)) is confined to vitreous cells where it overlaps with Pdgfrβ (red (b, f); dual red/green (g)). Note that other Pdgfrβ-expressing cells adjacent to pigment epithelial cells (arrow (b)) do not express Gfp (arrow (a)). DAPI stain (c). Original magnification: × 100 (a–d) and × 400 (e–g). (D) Representative confocal photomicrographs of E11.5 Arf^Gfp/+ mouse embryo following dual immunofluorescence labeling for Pdgfrβ (red (a)) and Gfp (green (a)) and DAPI (b). Relatively few vitreous cells behind the lens (L) express Gfp (green, arrow) compared to abundant Pdgfrβ-expressing cells (red, arrowhead). Original magnification: × 400.

Mural cells lack true lineage-defining markers, but their anatomic localization and the expression of either NG2 chondroitin sulfate proteoglycan (Ozerdem et al, 2001) or Pdgfrβ (Benjamin et al, 1999; reviewed in Hoch and Soriano, 2003) can place them in this lineage. As in the postnatal mouse eye (Martin et al, 2004), we observed that some Gfp-positive cells in both Arf^Gfp/+ and Arf^Gfp/Gfp embryos coexpressed NG2 (Figure 4B; some data not shown). Gfp expression also overlapped with Pdgfrβ in vitreous cells from E11.5 through at least postnatal day 1 (Figure 4C and D; additional data not shown). Of note, not all of the Pdgfrβ-positive cells in the eye expressed the Arf promoter (Figure 4C, arrows). Moreover, at E11.5, when relatively few Gfp-positive cells were evident, Pdgfrβ was readily detectable (Figure 4D). Although the degree of overlap varied somewhat in Arf^Gfp/+ and Arf^Gfp/Gfp embryos at E13.5 (see below), the findings indicated that the Arf-expressing cells in the embryonic vitreous likely represented mural cells or their progenitors.

p19^Arf can regulate Pdgfrβ and block Pdgf-B-driven cell proliferation in pericyte-like cells and MEFs independent of Mdm2 and p53

Pdgf-B/Pdgfrβ signaling is essential for pericyte and VSMC recruitment and proliferation during mouse development (reviewed in Hoch and Soriano, 2003). Because of overlapping Arf and Pdgfrβ expression, we considered whether p19^Arf may influence Pdgf-B signaling to control mural cell accumulation. We first addressed this in vitro using mouse 10T1/2 cells. These are embryonic fibroblast-like cells that express Pdgfrβ, Ang-1, and Vegf (SX Skapek, unpublished data), suggesting that they have pericyte-like properties. Moreover, they migrate toward cocultured endothelial cells growing as a monolayer (Hirschi et al, 1998) or in tube-like structures on Matrigel (JD Thornton and SX Skapek, unpublished data) and they localize to blood vessel walls when injected into mice (Hirschi et al, 1998). Southern blotting showed they had biallelic deletion of the Arf gene (SX Skapek, unpublished data), but they were efficiently transduced with retrovirus to express p19^Arf (Figure 5A).

Figure 5.

Arf expression blocks Pdgf-B-driven cell proliferation in 10T1/2 pericyte-like cells. (A) Immunofluorescence photomicrographs of 10T1/2 cells transduced with Arf-expressing (Arf) or control (Gfp) retrovirus and stained using anti-p19^Arf antibody or DAPI as indicated. Note focal subnuclear localization of ectopic p19^Arf (arrowhead). (B, C) Quantitative analysis of 10T1/2 cell number following transduction with Gfp (G)- or Arf (A)-encoding virus, with or without exposure to Pdgf-B, as indicated. Cells were harvested on day 1 (24 h) or day 2 (48 h) as indicated. Lanes 1 and 2, and 3–6 (B) and lanes 1–6 and 7–18 (C) represent separate experiments. Cell numbers are presented as average percentage (and standard deviation) of transduced cells not treated with Pdgf-B from triplicate samples. (D) Representative histograms (top) and cumulative data (bottom) from FACS analysis of PI-stained cells following 24 h exposure of cells to 50 ng/ml Pdgf-B. Data are presented as average percent S-phase fraction (and standard deviation) using duplicate samples from two or three representative experiments.

To mimic what occurs in the vitreous, we evaluated whether Pdgf-B promoted 10T1/2 cell accumulation in vitro and whether Arf expression could block it. Ectopically expressed Arf was functional in these cells, as it decreased their accumulation in response to 10% FBS by 20–40% over a 24–48 h period (Figure 5B, compare lanes 2, 4, and 6 with 1, 3, and 5). The addition of 5 or 50 ng/ml of Pdgf-B augmented 10T1/2 cell accumulation and this effect was blunted in cells expressing p19^Arf (Figure 5C, lanes 1–3, 7–9, and 13–15 versus 4–6, 10–12, and 16–18). We performed quantitative analyses to determine whether the effects of Pdgf-B and p19^Arf were primarily due to effects on the cell cycle or cell survival. Exposure to Pdgf-B for 24 h increased the fraction of 10T1/2 cells in S phase (Figure 5D, lanes 2 and 6 versus 1 and 5). Arf transduction prevented cell cycle phase changes (Figure 5D, lanes 4 and 8 versus 3 and 7). Interestingly, exposure of Arf-expressing cells to 50 ng/ml Pdgf-B further limited 10T1/2 cell accumulation as compared to Arf-transduced cells cultured without or with only 5 ng/ml Pdgf-B (Figure 5C, compare lanes 6, 12 and 18 with lanes 4, 5, 10, 11, 16, and 17). This may have resulted from slightly enhanced apoptosis. The sub-G₁ fractions in repeated experiments were as follows: 1.74 and 0.71% in untreated Gfp-transduced cells; 0.43 and 0.83% in Gfp-transduced cells treated with 50 ng/ml Pdgf-B; 1.54 and 1.73% in untreated Arf-transduced cells; and 3.43 and 3.78% in Arf-transduced cells treated with 50 ng/ml Pdgf-B. Hence, in vitro studies of 10T1/2 cells recapitulated in vivo findings in that Arf expression limited Pdgf-B-driven pericyte-like cell accumulation by arresting cell proliferation while smaller effects on apoptosis also occurred.

Because Arf promoter activity only partially overlapped with Pdgfrβ expression in the developing vitreous and Pdgfrβ may decrease with mural cell maturation (Benjamin et al, 1998; Hellstrom et al, 1999), we tested whether ectopic p19^Arf modulated Pdgfrβ expression in cultured 10T1/2 cells. Pdgfrβ expression decreased after transduction with Arf-encoding retrovirus (Figure 6A, lanes 4–6 versus 1–3). Decreased Pdgfrβ protein on day 3 in Arf- versus Gfp-transduced cells correlated with decreased mRNA (Figure 6A, lane 6 versus 3 and Figure 6B, lane 4 versus 3) while Pdgfrβ protein and mRNA were similar in Arf- versus Gfp-transduced cells on day 1 (Figure 6A, lane 4 versus 1 and Figure 6B, lane 1 versus 2). Dampening of Pdgfrβ expression was not simply secondary to the fact that Arf-transduced cells were growth arrested because Pdgfrβ expression was similar in proliferating 10T1/2 cells versus those arrested by confluence (Figure 6C).

Figure 6.

Downregulation of Pdgfrβ mRNA and protein by p19^Arf. (A) Representative Western blots for indicated proteins in 10T1/2 cells transduced with Gfp- or Arf-expressing retrovirus. (B) Representative Northern blot (top) and phosphoimage densitometry (bottom) for Pdgfrβ and β-Actin in 10T1/2 cells transduced with retrovirus and harvested as indicated. (C) Representative Western blots for indicated proteins and cell cycle analysis of 10T1/2 cells cultivated for 3 days as a subconfluent (SC) or confluent (C) monolayer.

We used genetically engineered MEFs to determine whether Arf-dependent arrest of Pdgf-B-driven cell proliferation and downregulation of Pdgfrβ depended on p53, which is intact in 10T1/2 cells (Syljuasen et al, 2001). First, we verified that the effects of ectopic Arf expression on Pdgfrβ were similar in Arf^−/− MEFs and 10T1/2 cells (Figure 7A, lanes 1 and 3 versus 2 and 4). The apparent difference in the kinetics of Pdgfrβ downregulation in 10T1/2 cells versus Arf^−/− MEFs likely reflects the fact that additional factors can control its expression in cultured cells (see, for example, Vaziri and Faller, 1995). In contrast to these cells, ectopic p19^Arf did not downregulate Pdgfrβ in Arf^−/−, p53^−/− (DKO) MEFs, nor did it block Pdgf-B-mediated cell proliferation on day 3, despite high transduction efficiency and p19^Arf expression (Figure 7B, lanes 1 and 3 versus 2 and 4; some data not shown). Interestingly, concurrent loss of Mdm2 restored the capacity of ectopically expressed Arf to dampen Pdgfrβ in Arf^−/−, Mdm2^−/−, p53^−/− (TKO) MEFs (Figure 7C, lanes 1, 3, and 5 versus 2, 4, and 6) and to block cell proliferation stimulated by Pdgf-B (Figure 7D, lanes 1–3 versus 4–6). Our findings indicated that Arf-dependent reduction in Pdgfrβ could be achieved independent of p53 in TKO MEFs and suggested that the presence of endogenous Mdm2 might interfere with p19^Arf when expressed in DKO MEFs.

Figure 7.

Downregulation of Pdgfrβ by p19^Arf can occur in the absence of Mdm2 and p53 in MEFs. (A–C) Representative Western blots for indicated proteins in Arf^−/− (A), Arf^−/−, p53^−/− (B), and Arf^−/−, Mdm2^−/−, p53^−/− (C) MEFs transduced with Gfp- or Arf-expressing retrovirus. (D) Quantitative analysis of the relative fraction of cells that are BrdU positive in Arf^−/−, Mdm2^−/−, p53^−/− MEFs transduced with the indicated retrovirus and harvested on day 2 following 1.5 h exposure to 5 or 50 ng/ml Pdgf-B. Results are presented as percent (and standard deviation) of triplicate samples from a representative experiment.

p19^Arf blocks Pdgfrβ signaling to limit vitreous cell accumulation

We considered whether endogenous p19^Arf was associated with decreased expression of Pdgfrβ. A strict correlation between the presence of p19^Arf and the level of Pdgfrβ was not always observed in serial-passed wild-type and Arf^−/− MEFs (AC Martin and SX Skapek, unpublished). To better assess this in vivo, we determined whether the presence of one functional Arf allele correlated with loss of Pdgfrβ in the vitreous of E13.5 Arf^Gfp/+ embryos. At this stage, nearly all Gfp-positive vitreous cells also expressed Pdgfrβ in Arf^Gfp/Gfp mice while Arf^Gfp/+ mice had significantly fewer double-positive cells (P=0.001) (Figure 8A and B). This decrease was offset both by more Pdgfrβ-positive cells lacking detectable Arf expression and by more Arf-expressing cells with no detectable Pdgfrβ (schematically depicted in Figure 8C). Increased Pdgfrβ single-positive cells in heterozygous animals may be due to dampening of the Arf promoter in cells with functional p19^Arf or it may simply reflect lower Gfp protein in mice heterozygous for the reporter gene. Increased Gfp-positive (Arf-expressing), Pdgfrβ-negative cells in Arf^Gfp/+ mice supports the concept that endogenous p19^Arf dampens Pdgfrβ in vivo. The relatively small number of Gfp-positive, Pdgfrβ-negative cells may be due to suboptimal repression of the receptor as Arf heterozygous mice have subtle defects in mural cell accumulation (Figure 3). Alternatively, it could reflect ‘disappearance' of Arf-expressing cells if the Arf promoter decreases with further mural cell maturation.

Figure 8.

Pdgfrβ expression correlates with Arf status in the developing mouse vitreous. (A) Representative confocal photomicrographs of cryostat sections from E13.5 Arf^Gfp/+ (a) and Arf^Gfp/Gfp (b) mice following dual immunofluorescence staining for Gfp (green) and Pdgfrβ (red). Note some cells express only Gfp (arrowhead) or only Pdgfrβ (arrow) in the heterozygous eye. (B) Quantitative analysis of Gfp- and Pdgfrβ-expressing cells stained as in (A), expressed as average percent (and standard deviation) of single- or double-positive cells. (C) Schematic diagram depicting potential model in which p19^Arf may both (1) check the expression of its own promoter to block the accumulation of Arf-, Pdgfrβ-double-positive cells and (2) promote the subsequent downregulation of Pdgfrβ expression with mural cell maturation. (D) (a) Documentation of Arf and Pdgfrβ genotype in embryos and (b) photomicrographs of H&E-stained sections of E14.5 and E15.5 mouse embryo eyes of the indicated genotypes. Data are representative of 10 eyes of each genotype examined from three separate litters. (E) Schematic diagram showing that p19^Arf blocks Pdgfrβ-driven vitreous cell accumulation in a manner that is not strictly Mdm2 or p53 dependent. Original magnification: × 400 (A) and × 100 (D).

To establish a causal relationship between Pdgfrβ and vitreous cell accumulation, we tested whether loss of Pdgfrβ lessened the phenotype in Arf^−/− embryos. We analyzed Arf^−/− embryos with heterozygous or homozygous Pdgfrβ deficiency at E14.5 and E15.5 because Pdgfrβ^−/− embryos manifest developmental defects after E16 (Soriano, 1994). The average number of vitreous cells in Arf^−/−, Pdgfrβ^−/− embryos (1.64±1.11 cells/2500 μm²) (n=10) was significantly decreased as compared to Arf^−/−, Pdgfrβ^+/− embryos (7.46±1.40 cells/2500 μm²) (n=8) (P<0.001) (Figure 8D). These findings indicated that Pdgfrβ was required to drive vitreous cell accumulation in the absence of Arf.

Discussion

Regulation of the Pdgf-B/Pdgfrβ axis is particularly important in the vitreous because excess cells accumulate in mice expressing a mutated form of Pdgf-B that lacks the retention motif to limit its mitogenic effects (Lindblom et al, 2003). Our experiments have addressed whether the Arf tumor suppressor gene product regulates this signaling pathway. We have established that Arf was expressed in and blocked the proliferation of mural cells within the vitreous early during its development. That the phenotype in Arf^−/− embryos was corrected in those also lacking Pdgfrβ demonstrated a requirement for p19^Arf to control Pdgfrβ-dependent effects in the eye. Without p19^Arf protein, nearly all of the vitreous cells expressing the Arf promoter also expressed Pdgfrβ while the number of Arf-expressing cells with detectable Pdgfrβ decreased in the presence of one normal Arf allele. Arf expression in cultured pericyte-like cells and MEFs suppressed Pdgfrβ and blocked Pdgf-B-driven proliferation in a manner separable from Mdm2 and p53. Taken together, our findings show that Arf plays an essential role to check Pdgfrβ signals driving vitreous accumulation (Figure 8E).

In vivo and in vitro analyses indicated that one mechanism by which p19^Arf might accomplish this is by downregulating Pdgfrβ expression. It must be emphasized, though, that p19^Arf can control proliferation by multiple, p53-dependent and -independent mechanisms operating with differing kinetics (Weber et al, 2000). Indeed, we observed this in 10T1/2 cells in which Arf-mediated cell cycle arrest preceded Pdgfrβ downregulation (compare Figures 5D and 6A). The early arrest was likely secondary to p53-dependent induction of the cyclin-dependent kinase inhibitor p21^Waf1 evident on day 1 (Supplementary Figure 1). How Arf-dependent downregulation of Pdgfrβ relates to other recognized p53-independent functions of p19^Arf is not yet known. Moreover, whether downregulation of Pdgfrβ is the sole mechanism by which p19^Arf functions in the eye must be critically evaluated, for example, by testing if transgenic expression of Pdgfrβ under the control of the Arf promoter recapitulates the Arf^−/− eye phenotype.

Certain mechanistic aspects of Arf-dependent dampening of Pdgfrβ signaling merit emphasis. First, a number of observations allowed us to conclude that decreased Pdgfrβ was not merely secondary to p19^Arf-dependent cell cycle arrest or decreased confluence as follows: (1) Pdgfrβ did not measurably change in proliferating cells versus those arrested by confluence (Figure 6C) nor did it change in Gfp-expressing cells over the course of an experiment (Figure 6A, lanes 1–3), even though the S-phase fraction decreased due to changes in confluence (Figure 5D, compare lanes 1 and 5). (2) Pdgfrβ expression was dampened by p19^Arf on day 3 (Figure 6A, lane 6 versus 3) at a point when the S-phase fraction was similar in Gfp- and Arf-transduced cells (Figure 5D, compare lanes 7 and 5). (3) As mentioned above, 10T1/2 cells transduced with Arf are arrested and less confluent than Gfp-transduced cells on day 1, but little difference in Pdgfrβ is evident (Figure 6A, lane 4 versus 1). Second, the parallel decrease in Pdgfrβ mRNA and protein (Figure 6A and B) suggested that p19^Arf influenced its transcription or message stability. The former possibility is intriguing because p73, a p53-related protein, represses the Pdgfrβ promoter when expressed in cultured NIH 3T3 cells but p53 cannot (Hackzell et al, 2002). A functional relationship between p19^Arf and p73, though, has not been established. Finally, depression of Pdgfrβ in Arf-expressing MEFs could occur in the absence of p53 as long as the MEFs also lacked Mdm2. That endogenous Mdm2 can hinder p19^Arf in p53^−/− cultured MEFs was previously proposed (Weber et al, 2000), but its significance for eye development is not clear.

We previously established that Arf is required for VHP regression in the mouse and that its failed involution causes secondary lens and retina defects that mimic the human eye disease, persistent hyperplastic primary vitreous (PHPV) (McKeller et al, 2002). The disease has apparently complete penetrance in Arf^−/− mice and complete absence in Arf^+/− mice despite the subtle embryonic phenotype reported here. Whether human ARF defects contribute to PHPV pathogenesis is not yet proven. Consistent with our mouse model, eye abnormalities were not reported in two small kindreds with ARF exon 1β haploinsufficiency (Randerson-Moor et al, 2001; Hewitt et al, 2002). Abnormalities in p16^Ink4a, also encoded at the Arf/Ink4a locus, do not result in PHPV in Ink4a^−/− mice (Martin et al, 2004) or in humans with heterozygous or homozygous INK4A mutations (Yakobson et al, 2000; Della et al, 2001; Pavel et al, 2003).

The role of p53 in VHP involution has been harder to understand. Certain inbred p53^−/− mice develop a PHPV-like process with variable severity and penetrance (Reichel et al, 1998; Ikeda et al, 1999). We and others reported that p53^−/− mice in a mixed C57BL/6 × 129/Sv genetic background usually have normal eyes (Ikeda et al, 1999; McKeller et al, 2002) and we proposed that p19^Arf might regulate parallel, p53-dependent and -independent pathways to control vitreous cell accumulation (Martin et al, 2004). Dampening of Pdgfrβ signaling may represent one such p53-independent pathway.

That we did not observe Arf promoter activity widely in mural cells enveloping other vessels implies that Arf-dependent modulation of mural cell biology might somehow contribute to the ‘transient' nature of the VHP. The simplest mechanism by which p19^Arf might destabilize this vasculature is by limiting mural cell coverage. Integrity of the underlying vessels would then depend on other sources of angiogenic factors like Vegf derived from the retina (Martin et al, 2004) or lens (Mitchell et al, 1998). Dampening of Pdgfrβ signals by p19^Arf might also influence other aspects of mural cell biology like their expression of angiogenic factors to more directly destabilize the vessels. Either concept has potential implications for tumor suppression by p19^Arf because there is some evidence that pericytes influence tumor vascular biology (Benjamin et al, 1999; Bergers et al, 2003; Gee et al, 2003; Jain and Booth, 2003). Conceivably, Arf-dependent regulation of perivascular cells—to prevent their accumulation or alter their expression of angiogenic factors—might destabilize developing tumor vasculature and prevent the ‘angiogenic switch' required for tumorigenesis (Hanahan and Folkman, 1996).

Perhaps our most surprising finding was the localized expression of Arf in the developing eye. Induction by excessive mitogenic or oncogenic signals does not fully explain its selective expression in a subset of cells in the relatively homogenous embryonic vitreous. The temporal and spatial correlation of Pdgfrβ and Arf expression suggest that Pdgf-dependent signals may induce the Arf promoter. But these are likely not sufficient because Pdgf-B is generally required for mural cell accumulation in developing blood vessels (reviewed in Hoch and Soriano, 2003) whereas Arf is not widely expressed in the embryo (AC Martin and SX Skapek, unpublished data). Elucidating the specific developmental cues driving Arf expression may help predict settings where it has tumor suppressor effects: if Arf is only induced by certain mitogens or in particular cells or cellular environments, its tumor suppressor properties may be similarly restricted. Understanding how it is controlled during eye development may ultimately lead to ways to re-activate its expression in that subset of human cancers where the gene is intact but its expression is repressed.

Materials and methods

Mice and cell lines

Mice in which Arf exon 1β was inactivated (Kamijo et al, 1997) or replaced by a reporter gene encoding Gfp (Zindy et al, 1998) were bred to maintain a mixed C57BL/6 × 129/Sv genetic background. Pdgfrβ-deficient mice (Soriano, 1994) were provided by P Soriano (Fred Hutchinson Cancer Research Center). Genotyping methods have been described (Soriano, 1994; Kamijo et al, 1997; Zindy et al, 2003). Animal studies were approved by the St Jude Children's Research Hospital Animal Care and Use Committee.

Mouse 10T1/2 pericyte-like cells (CCL-226) (ATCC) were maintained in DMEM with 10% FBS supplemented with L-glutamine and penicillin/streptomycin (pen/strep). Primary MEFs from wild type; Arf^−/−; Arf^−/−, p53^−/−; and Arf^−/−, Mdm2^−/−, p53^−/− mice (provided by Charles J Sherr and Martine F Roussel, St Jude Children's Research Hospital) were cultivated as previously described (Zindy et al, 1997).

Histology studies

Embryos removed from euthanized, pregnant mice were fixed in 4% paraformaldehyde (PFA) in PBS and processed for paraffin or cryostat sections as described previously (Martin et al, 2004). Immunofluorescence staining for Gfp in Arf^+/Gfp and Arf^Gfp/Gfp mouse embryo eyes was performed as previously described (Martin et al, 2004). Dual immunofluorescence staining was carried out using rabbit α-Gfp and either rat α-CD31 or goat α-Pdgfrβ. Staining for p19^Arf in wild-type embryos was accomplished with a rat monoclonal antibody directed against mouse p19^Arf (Bertwistle et al, 2004). Primary antibodies were detected using species-specific secondary antibodies (Jackson Immunoresearch Laboratories). Cell proliferation was assessed by staining for Ki67 proliferating cell nuclear antigen. Apoptosis was assessed by terminal deoxynucleotidyl transferase (TdT)-mediated dUTP nick end labeling (TUNEL). Digital photomicrographs were obtained with an Olympus BX60 microscope equipped with a SPOT RT Slider camera (Diagnostic Instruments) or with a Zeiss 510 NLO multiphoton/confocal laser-scanning microscope. (Additional details for histology studies are provided in Supplementary data.)

Quantitative analysis of vitreous cell density in embryos of different genotypes was performed by counting the total number of cells (excluding nucleated red blood cells) within the vitreous in comparable midline eye sections of stained slides. Cell number was presented relative to the area of the vitreous. TUNEL-, Ki67-, and Pdgfrβ-positive cells were presented as a percent of the total number of vitreous cells. Pdgfrβ expression data were determined using confocal microscopy, whereas others were assessed using light microscopy ( × 400 magnification in all cases). Summary data (Figures 2B, D, and 8B) are from analyses of 6–17 embryos from six timed pregnancies, presented as mean value and standard deviation. Statistical significance was assessed by t-test.

Retroviral production

Retrovirus was produced as described previously (Zindy et al, 1998), using the control pMSCV-IRES-Gfp plasmid or one containing mouse Arf cDNA 5′ to the IRES (provided by Martine F Roussel, SJCRH).

Cell proliferation analysis

Subconfluent 10T1/2 cells or MEFs were infected with Arf-encoding or control retrovirus (Zindy et al, 1998). At 16 h after infection, the culture media were replaced. After 20 h, 5 or 50 ng/ml of Pdgf-B (R&D Systems) was added to the media; this point represents day 0 for experiments using transduced cells. Pdgf-B was replenished each day. Cells from triplicate wells were removed and counted using a hemocytometer on days 1, 2, and 3. Cell cycle phase was assessed in Arf- or control-transduced cells by flow cytometry (FACScan, Becton Dickinson) of Gfp-sorted cells that were stained with propidium iodide (PI). For cell cycle analyses, Pdgf-B was added 1.5 or 24 h prior to analysis. S phase was also monitored by cultivating cells in 10 μM BrdU followed by staining with the APC-BrdU Flow kit (BD Biosciences Pharmingen).

Gene expression studies in cultured cells

Western analyses were conducted using subconfluent cells, transduced with Arf-encoding or control retrovirus and harvested on days 1, 2, and 3 for protein extraction as described (Zindy et al, 1998). Proteins were detected by chemiluminescence using α-Pdgfrβ, α-p19^Arf, or α-Hsc70 primary with species-specific secondary antibodies. The Northern-Max kit (Ambion) was used to perform Northern blotting of total RNA isolated from 10T1/2 cells harvested at days 1 and 3. The blot was probed with ³²P-labeled Pdgfrβ and β-actin probes and scanned using Storm Phosphoimager (Amersham Pharmacia Biotech). Pdgfrβ mRNA abundance was normalized using β-actin transcript. mRNA extracted from cultured 10T1/2 cells was used for RT–PCR. (See Supplementary data for details of Western blotting, PCR primer design, and cDNA probe synthesis.)

Supplementary Material

Supplementary Information