Upregulation of chicken p15^INK4b at senescence and in the developing brain

Replicative senescence of human diploid fibroblasts (HDFs) in tissue culture is primarily determined by the progressive loss of telomeric DNA. However, in other cell types, a senescence-like arrest that precedes telomere attrition can be triggered by the accumulation of the cyclin-dependent kinase inhibitor p16^INK4a (Serrano and Blasco, 2001; Sherr and DePinho, 2000; Wright and Shay, 2000). The induction of p16^INK4a in this setting might relate to the stressful conditions of tissue culture; indeed, other forms of cellular stress, such as the addition of oncogenic Ras, can elicit a senescence-like arrest in relatively young cells irrespective of telomerase expression (Serrano et al., 1997; Wei et al., 1999). In practice, therefore, the lifespan of cells in culture is determined by the relative severities of culture stress and telomere attrition. Spontaneous immortalisation of human cells is extremely rare unless steps are taken to disable the mechanisms underlying senescence. Typical strategies include the use of DNA tumour virus oncoproteins, such as SV40 T-antigen, to inactivate the retinoblastoma (pRb) and p53 tumour suppressors (Shay et al., 1991). In most instances, this simply delays the proliferative arrest engendered by telomere loss and leads to a crisis situation in which chromosome fusion and breakage drives cells into apoptosis (Serrano and Blasco, 2001; Sherr and DePinho, 2000; Wright and Shay, 2000).

These observations with human cells stand in stark contrast to the large body of research on cultured rodent cells, recently enhanced by the availability of mouse embryo fibroblasts (MEFs) from mice with specific genetic defects. Long telomeres and widespread expression of telomerase imply that the limited lifespan of normal MEFs, typically only 10-20 population doublings (PDs), must reflect culture stress (Serrano and Blasco, 2001; Sherr and DePinho, 2000; Wright and Shay, 2000). Immortal MEFs arise at a relatively high frequency, almost always accompanied by some perturbation of the p53 pathway. The pressure to disarm p53 is in part caused by the progressive accumulation of p19^ARF, an upstream regulator of p53 stability (Sherr and DePinho, 2000).

Interestingly, p19^ARF (p14^ARF in humans) and p16^INK4a are encoded by the same genetic locus, officially designated CDKN2A, in which the second exon is translated in different reading frames (Drayton and Peters, 2002; Sherr, 2001). Two transcripts are produced that initiate at separate promoters and encompass different first exons, designated exon 1β (for ARF) and exon 1α (for p16^INK4a). Both products are implicated in tumour suppression and senescence, but their relative importance appears to differ in different cell types or species. For example, whereas p16^INK4a and ARF both accumulate as primary MEFs reach the end of their proliferative lifespan (Kamijo et al., 1999; Palmero et al., 1997; Zindy et al., 1997), the effects appear confined to p16^INK4a in senescent HDFs (Alcorta et al., 1996; Brookes et al., 2002; Hara et al., 1996; Serrano et al., 1997; Wei et al., 2001). Conversely, the spontaneous immortalisation of MEFs is generally accompanied by abrogation of the p53-ARF pathway rather than p16^INK4a, and genetic ablation of either p53 or ARF in transgenic mice generates MEFs that are inherently immortal (Kamijo et al., 1997).

To shed further light on the roles of INK4a and ARF, and resolve some of the species differences, we recently isolated the corresponding locus in chickens (Kim et al., 2003). We identified transcripts encoding INK4b and ARF respectively, but chicken cells lack the capacity to encode INK4a owing to a partial duplication of the region encompassing exon 1β in place of the putative exon 1α. Given the proposed role of p16^INK4a in both spontaneous and oncogene-induced senescence in mammals, it was important to establish whether either or both of the INK4b and ARF products contribute to senescence in chicken cells. Here, we show that the levels of chicken ARF mRNA remain unchanged during prolonged passaging of chicken embryo fibroblasts (CEFs) in culture, although they do respond to some of the oncogenic signals that are known to activate mammalian ARF. By contrast, INK4b RNA and protein accumulate in senescent CEFs and the gene is transcriptionally silenced in two immortalised chicken cell lines. However, knockdown of INK4b or ARF expression with short-hairpin RNA (shRNA) is not sufficient to immortalise CEFs.

As part of the original characterisation of chicken INK4b and ARF, we showed that HA-epitope-tagged versions of both proteins were able to cause cell-cycle arrest when overexpressed in human cells (Kim et al., 2003). To enable us to perform analogous experiments in chicken cells, we used an amphotropic retrovirus to express the receptor for mouse ecotropic viruses in early-passage CEFs (McConnell et al., 1998). The resultant pool of G418-resistant cells was thus susceptible to subsequent infection with murine leukaemia virus (MuLV)-based recombinant retroviruses carrying suitable drug-resistance markers. Immunoblotting with an antibody against the HA epitope confirmed that the infected cell pools expressed exogenous INK4b and ARF (Fig. 1A). As anticipated, both INK4b and ARF were able to block proliferation of CEFs as judged by 5-bromo-2-deoxyuridine (BrdU) incorporation or by comparing relative cell numbers by staining with Crystal Violet (Fig. 1B). Importantly, introduction of ARF caused a significant upregulation of endogenous p53 and p21^CIP1 (Fig. 1A) (Kim et al., 2003).

We also used these cells to validate polyclonal antisera raised against the predicted C-terminal peptides of chicken ARF and p15^INK4b. Thus, the SK8 antibody detected a band with an apparent molecular weight of 16 kDa that coincided with 2×HA-tagged chicken ARF (Fig. 1C). It also recognised a background band of 14 kDa that was present in CEF cells infected with the empty retrovirus control. Under the conditions used, the SK8 antibody did not detect endogenous ARF in the vector-only controls or in uninfected CEFs. This is reminiscent of the difficulties we and others have encountered in visualising p14^ARF in human fibroblasts; however, in the latter system, the endogenous protein becomes detectable if the cells are transduced with a retrovirus encoding E2F1 (Llanos et al., 2001). In line with these observations, expression of human E2F1 in CEFs resulted in the appearance of a 10-12 kDa protein that was specifically recognised by the SK8 antiserum (Fig. 1D). Note that the endogenous protein is smaller than the 2×HA-tagged version and runs ahead of the 14 kDa background band. However, the mobility of endogenous ARF does not exactly conform to its calculated molecular weight of 7 kDa, perhaps because of its unusual amino acid composition (37% arginine). Had splicing enabled the protein to read into the eponymous alternative reading frame, its predicted molecular weight would have been 24.4 kDa (Kim et al., 2003). Thus, detection of endogenous chicken ARF with the SK8 antibody confirmed that the protein comprises only 60 amino acids encoded exclusively by exon 1β.

The SK14 polyclonal antibody detected a protein of approximately 20 kDa that coincided with 2×HA-tagged chicken INK4b (Fig. 1C) and recognised an endogenous protein of approximately 15 kDa, consistent with the predicted molecular weight of chicken INK4b (Fig. 1D). The protein was detectable in normal CEFs and the levels increased slightly upon expression of E2F1.

Having found that chicken ARF can be upregulated by E2F1, it was of interest to know whether other signalling pathways that impact on mammalian ARF also operate in chicken cells, and whether agents that are known to activate mammalian INK4a have analogous effects on chicken INK4b. Given the technical problems associated with the SK8 antibody, these studies were primarily conducted at the RNA level. CEF cultures expressing the ecotropic receptor were infected with the pBABE retrovirus encoding the G12V variant of human H-Ras, E2F1 or SV40 large T-antigen. After selection in puromycin, the cell pools were analysed by immunoblotting to confirm expression of the exogenous proteins (Fig. 2A). Equivalent amounts of total RNA were analysed by northern blotting using probes derived from the unique 3′-untranslated regions of INK4b and ARF. As shown in Fig. 2B, introduction of either Ras, E2F1 or SV40 large T-antigen led to increased expression of both INK4b and ARF RNA. By contrast, we saw little if any change in INK4b or ARF levels with a retrovirus encoding human Myc (data not shown) or with the avian retroviruses MC29 and MH2, which encode oncogenic derivatives of chicken Myc (Fig. 2C). Based on precedents in mammalian cells, Myc might have been expected to activate ARF but suppress INK4b (Staller et al., 2001; Zindy et al., 1998).

Another distinctive feature of mammalian p15^INK4b is its upregulation by TGF-β in epithelial cells (Hannon and Beach, 1994; Reynisdóttir et al., 1995). It was therefore of interest to determine whether the chicken INK4b gene showed similar regulation. Surprisingly, addition of TGF-β to CEFs resulted in a modest reduction in INK4b RNA levels over a 4-hour time period (Fig. 3A), accompanied by clear evidence for activation and nuclear translocation of Smad2/3 (Fig. 3B). As the growth-inhibitory effects of TGF-β are not generally manifest in fibroblasts, we are not aware of previous experiments that have addressed this issue. To be more consistent with published studies, we sought a chicken epithelial cell system in which the TGF-β family has demonstrable effects on proliferation. Primary cultures of chicken ovarian granulosa cells (cGCs) can be generated using an appropriate balance of follicle-stimulating hormone (FSH) and activin A and these cells undergo arrest when exposed to increasing concentrations of TGF-β (Schmierer et al., 2003). As illustrated in Fig. 3C, exposure of cGCs to increasing concentrations of TGF-β also resulted in decreased expression of chicken INK4b RNA. This is the exact opposite of the effects observed in the HaCaT (human keratinocyte) and MvLu (mink lung epithelium) models, where the induction of p15^INK4b is thought to contribute to TGF-β-mediated growth arrest (Hannon and Beach, 1994; Reynisdóttir and Massagué, 1997; Reynisdóttir et al., 1995). However, cell lines that lack p15^INK4b, as a result of homozygous deletion or silencing of the locus, remain sensitive to TGF-β-mediated arrest (Flørenes et al., 1996; Iavarone and Massagué, 1997; Malliri et al., 1996). It will be interesting to explore why TGF-β has such contrasting effects on INK4b expression and the physiological consequences of up- and downregulation in different cell systems.

Taken together, our findings suggest that chicken INK4b has properties that would enable it to compensate for the absence of INK4a in some contexts, but a key question was whether it could do so at senescence. Primary CEFs were passaged continually in standard tissue culture conditions until they underwent replicative senescence, as indicated by failure to reach confluence in 4 weeks (Fig. 4A). In CEFs, this typically occurs after 30-50 population doublings (PDs) and spontaneous escape from senescence is extremely rare. In this respect, CEFs are more akin to HDFs than to MEFs (Kim et al., 2002; Shay et al., 1991; Swanberg and Delany, 2003; Venkatesan and Price, 1998). The senescent cells adopted an enlarged flattened appearance, stained positively for senescence-associated (SA)-β-galactosidase activity (Fig. 4B), and had a very low BrdU labelling index (approximately 1%). These features would be consistent with the M1 senescence phenotype of HDFs (Wei and Sedivy, 1999) but the senescent CEF cultures released substantial numbers of detached cells with a sub-G0 DNA content (30-40%). By contrast, proliferating CEF cultures had very few sub-G0 cells (<1%), a high BrdU labelling index and negligible SA-β-gal staining.

Total RNA was prepared from the CEF cultures at various stages, including the senescent cells (53 PDs), and equivalent amounts were analysed by northern blotting as in Fig. 2. Whereas the 2.0 kb INK4b RNA was expressed at low levels at 27 PDs, it increased dramatically at later PDs to become maximal at PD53, when the cells were considered senescent (Fig. 4C). A similar increase could be documented at the protein level by immunoblotting with the SK14 antibody (Fig. 4D). By contrast, the levels of the 1.6 kb ARF transcript remained moderately high throughout the lifespan of the culture, and declined in the senescent population (Fig. 4C). A similar decline was noted in the GAPDH transcript, used here as a loading control. However, levels of p21^CIP1 RNA, detected using a chicken cDNA probe, and 18S ribosomal RNA (not shown) were essentially constant.

To compare the relative levels of the INK4b and ARF mRNAs, the same samples were hybridised with a genomic DNA fragment that recognises both transcripts by virtue of the conserved sequences at the beginning of the respective second exons. Although this probe will have a slight bias in favour of the 1.6 kb β transcript, it was evident that the ARF RNA was much more abundant than INK4b RNA in early-passage cells, whereas senescent CEFs had almost equivalent levels of the two transcripts (Fig. 4C). Despite the abundance of the ARF transcript, we were unable to detect the ARF protein with the SK8 antibody at any stage of the lifespan. As noted in other studies (Christman et al., 2005), levels of chicken p53, which could in part reflect ARF function, increased transiently as the CEFs approached senescence (Fig. 4D).

To obtain additional evidence for the role of INK4b in senescence, we sought ways of extending the lifespan of CEFs by ablating pRb and p53 or by knocking down INK4b and ARF with shRNA. Infection with a pBABEpuro vector encoding SV40 large T-antigen resulted in a lifespan extension (8-10 PDs) compared with control CEFs infected with the empty vector (Fig. 5A). Analogous experiments in human fibroblasts generally yield a more robust lifespan extension, although the magnitude of the effect can be quite variable (Brookes et al., 2004).

To knockdown chicken INK4b, we made use of the pRetroSuper vector to facilitate long-term expression of shRNAs in CEFs carrying the ecotropic receptor. After empirically testing 19 different shRNA sequences, we found two that produced a significant downregulation of INK4b, as judged by immunoblotting with SK14 (Fig. 5B). It is presently unclear why multiple shRNAs directed against the coding domain of INK4b proved ineffectual. In the event, long-term knockdown of INK4b with either of the shRNAs resulted in modest but consistent extension of lifespan relative to cells infected with the empty vector (Fig. 5C). Interestingly, knockdown of ARF (Fig. 5B) also caused a modest lifespan extension (Fig. 5C), but we did not observe any additive effect when both INK4b and ARF were targeted (data not shown). The effects of ARF knockdown could reflect its role in the p53 pathway as knockdown of p53 with shRNA produced a lifespan extension of similar magnitude (Fig. 5D). However, as we only achieved a partial knockdown of ARF and p53 in these experiments, the interpretation must be treated with caution.

In view of the marked increase in INK4b expression in senescent CEFs, we were curious to know whether the locus was altered in chicken cells that had escaped senescence. There are relatively few established chicken cell lines and we have thus far restricted our analyses to two classic examples: the DF1 line of spontaneously immortalised CEFs (Himly et al., 1998) and the DT40 B-cell lymphoma line (Buerstedde and Takeda, 1991). Northern blotting revealed that the 2.0 kb INK4b RNA was undetectable in the DF1 and DT40 cells, whereas the 1.6 kb ARF transcript was present in both (Fig. 6A).

The specific absence of INK4b RNA without demonstrable deletion or rearrangement of the locus (not shown) suggested that the promoter region, which bears the hallmarks of a CpG island (Kim et al., 2003), might have been silenced by methylation. A simple way of assessing this is to treat the cells with 5′-aza 2′-deoxycytidine (5′-aza C), an inhibitor of global DNA methyltransferase activity, and asking whether removal of DNA methylation will restore expression of the gene. As illustrated in Fig. 6B, treatment of DF1 cells with 5′-aza C reactivated the expression of INK4b RNA but had no discernible effect on ARF levels. Although these assays were not designed to be quantitative, the levels of INK4b in DF1 cells treated with 5′-aza C were similar to those in late-passage CEFs.

There is still considerable uncertainty about the role of the INK4b-ARF-INK4a locus in vivo, and the only settings in which the genes have been clearly implicated are in stem cell self-renewal, particularly in the context of disrupted polycomb gene function (Valk-Lingbeek et al., 2004) and in the senescence-like state associated with premalignant lesions (Braig et al., 2005; Chen et al., 2005; Collado et al., 2005; Michaloglou et al., 2005). As we do not have access to equivalent systems in chickens, we chose to perform a preliminary study of expression in the chicken embryo. We did not detect a signal with the ARF probe, but the INK4b probe revealed a very distinctive expression pattern in the developing hind-brain. INK4b expression was first observed at stage 13 in the roof-plate of rhombomere 1 (r1) near the isthmus (Fig. 7A,B). As development proceeds, the expression spreads towards the posterior (stage 17; Fig. 7C) but not beyond the r1-r2 boundary (stage 20; Fig. 7D). Expression persists at least until stage 24. Because p15^INK4b can cause cell-cycle arrest, we asked whether the cells expressing INK4b were proliferating. None of the roof-plate cells expressing INK4b (stage 24; Fig. 7E) stained with an antibody against phosphorylated histone H3, a marker of mitosis, indicating that the cells are post-mitotic (Fig. 7E-G).

The initial impetus for characterising the chicken INK4b-ARF-INK4a locus was that it might clarify apparent species or cell-type differences in the regulation of INK4a and ARF and their contribution to senescence. In the event, the lack of an INK4a homologue in chickens made direct comparisons impossible but posed a new set of interesting questions. The most compelling was to determine how chicken cells compensate for the absence of p16^INK4a. In HDFs, both spontaneous and oncogene-induced senescence is accompanied by upregulation of INK4a (Alcorta et al., 1996; Hara et al., 1996; Serrano et al., 1997), whereas in primary MEFs, ARF appears to be the major determinant of proliferative lifespan in tissue culture (Kamijo et al., 1997). Here, we suggest that INK4b assumes a central role in chicken cell senescence.

In addition to its proven ability to cause cell-cycle arrest (Fig. 1), its activation by oncogenic Ras (Fig. 2) and its dramatic accumulation in senescent CEF cultures (Fig. 4), the most persuasive evidence for the importance of INK4b is its transcriptional silencing in immortal chicken cell lines (Fig. 6). This echoes the inactivation of p16^INK4a that is observed in approximately 75% of established human cell lines (Ruas and Peters, 1998) and the loss of expression of INK4a that can be experimentally observed during the emergence of immortal clones from primary cultures of human cells (Noble et al., 1996). In many cases, this occurs through de novo methylation of the INK4a promoter (Foster et al., 1998; Huschtscha et al., 1998), and we suggest that a similar mechanism applies to the silencing of INK4b in DF1 cells (Fig. 6B). Importantly, selection for the loss of human INK4a persists in cells that have been transduced with telomerase or have alternative telomere maintenance mechanisms (Noble et al., 2004; Taylor et al., 2004; Tsutsui et al., 2002).

There is also a strong selection against p53 function during the immortalisation of human cells, and it has been reported that all established chicken cells have inactivated p53 (Kim et al., 2001; Ulrich et al., 1992). Despite such strong credentials for a role in senescence, experimental knockdown of INK4a or inactivation of p53 in human fibroblasts provides a relatively modest lifespan extension, particularly when evaluated in cell pools (Beausejour et al., 2003; Bond et al., 2004; Brookes et al., 2004; Itahana et al., 2003; Wei et al., 2003; R.J. and G.P., unpublished observations). This is presumably because senescence reflects the integration of several different signals, as well as potential heterogeneity in the response of individual cells. The relatively modest lifespan extension achieved by shRNA-mediated knockdown of chicken INK4b, ARF or p53 (Fig. 5) is therefore in line with the findings with human fibroblasts.

The most likely explanation for the limited lifespan extension would be the continued erosion of telomeres. CEFs are thought to resemble HDFs in terms of resistance to immortalisation and lack of telomerase activity, but the situation is complicated by the fact that chicken cells contain both macro- and micro-chromosomes with different size classes of telomeric repeats, some of which are interstitial (Delany et al., 2000; Kim et al., 2002; Lejnine et al., 1995; Swanberg and Delany, 2003; Venkatesan and Price, 1998). At this juncture, the dynamics of telomere erosion and repair remain unclear, as does the sensitivity of CEFs to different forms of culture stress. Thus, INK4b, ARF and p53 might contribute to the effectiveness of the response, but are presumably not essential for the implementation of the telomere-based arrest.

A role for p15^INK4b in senescence is not without precedent. For example, p15^INK4b levels have been shown to increase with population doublings in human T-lymphocytes and in mammary epithelial cells in which p16^INK4a has been epigenetically silenced (Erickson et al., 1998; Sandhu et al., 2000). Moreover, there is consistent evidence that p15^INK4b is specifically methylated or deleted in various leukaemias and lymphomas (Batova et al., 1997; Herman et al., 1997; Herman et al., 1996; Malumbres et al., 1997; Uchida et al., 1997). These findings imply that, in some settings, p15^INK4b can serve as a bona fide tumour suppressor, as either an adjunct or an alternative to the proliferative restraints imposed by p16^INK4a. It is tempting to speculate that, in chicken cells, p15^INK4b has re-assumed the role it fulfilled prior to the duplication of the INK4a and INK4b genes.

Against this background, the physiological role of p15^INK4b has been enigmatic. Despite widespread expression in cultured cells, there have been no indications of a role in embryonic development and/or differentiation that would tally with regulation by members of the TGF-β family. Our preliminary survey of chicken INK4b expression in early stages of embryogenesis provides the first evidence that INK4b might have a specific role in development, although further genetic experiments would be required to confirm this proposition. Expression was observed in the hind-brain, in roof-plate cells in the dorsal midline of rhombomere 1. This corresponds to a group of cells that will eventually be lost when the two halves of the cerebellum meet and fuse. Although this event happens later in development than the embryos at stage 13 to stage 24 that we have examined thus far, activation of p15^INK4b could prevent the roof-plate cells in this region from proliferating, in order to keep the two halves of the cerebellum close together as they grow and develop. As well as consolidating these ideas, it will be interesting to determine whether the expression pattern seen in the chicken embryo is replicated in the mouse, and whether closer inspection of the relevant sections of the mouse hind-brain reveals evidence of p16^INK4a expression. Mice that are nullizygous for either INK4b or INK4a are not reported to have neurological problems, but the phenotypic manifestations could be quite subtle and the two genes could have overlapping or compensatory functions.

On a more speculative level, our findings encourage the view that the primordial duplication of the INK4b gene, to create INK4a, could have enabled a degree of functional specialisation but that vestiges of the original roles of INK4b might have been maintained. Thus, although p16^INK4a became the predominant player in senescence, these functions were originally within the remit of p15^INK4b and may still be in certain lineages. In the absence of p16^INK4a, chicken cells presumably rely more heavily on p15^INK4b-mediated defences. The information and tools developed in this work should assist in unravelling the situation and facilitate attempts to bypass senescence in CEFs.

Primary chicken embryo fibroblasts (CEFs) were grown at 39°C in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% chicken serum (GIBCO-BRL). The DF1 line of immortalised chicken fibroblasts (Himly et al., 1998) was purchased from the ATCC (CRL-12203) and grown in DMEM with 10% FCS at 39°C. The DT40 bursal lymphoma line was obtained from J.-M. Buerstedde (Institute for Molecular Radiology, Munich, Germany) and grown in RPMI-1640 medium with 10% FCS and 1% chicken serum at 41°C. Cell proliferation assays using Crystal Violet staining were performed as described (McConnell et al., 1998). BrdU incorporation was assessed using the Boehringer Mannheim BrdU labelling and detection kit II. The cells were grown in 48-well culture dishes, labelled with 10 mM BrdU for either 1 hour or 16 hours, and BrdU incorporation was measured by counting the proportion of positively stained cells.

CEF and DF1 cells were infected with an amphotropic retrovirus encoding the receptor for mouse ecotropic retroviruses and selected in G418 (400 μg/ml) as previously described (McConnell et al., 1998). After selection, the cell pools were infected with recombinant ecotropic retroviruses encoding human H-Ras, SV40 large T-antigen, human E2F-1, or human Myc, all based on the pBABEpuro vector. Alternatively, the cells were infected with pRetroSuperPuro containing shRNA sequences against chicken INK4b, ARF and p53 (see below). Cells were infected in the presence of polybrene (4 μg/ml) for 2 hours at 37°C, the medium was then replaced, and the following day the cells were placed in selective medium containing puromycin (1.25 μg/ml). CEF cultures were also infected with the MC29 and MH2 avian retroviruses using supernatants obtained from producer lines provided by P. Vogt (The Scripps Research Institute, La Jolla, CA).

INKb #2: 5′-TATATGTGCTTGTCGGTTA-3′; corresponding to nucleotides 1787-1805 in the INK4b sequence (accession number AY138247). INK4b #4: 5′-TGAATTCGGTTCTTACGTG-3′; corresponding to nucleotides 1697-1715 in the INK4b sequence (accession number AY138247). ARF: 5′-CGCTGCTCCGGCGCATCTT-3′; corresponding to nucleotides 211-229 in the ARF sequence (accession number AY138245). P53: 5′-AATCGGTCACCTGCACTTACT-3′; corresponding to nucleotides 377-397 in the p53 cDNA sequence (accession number X13057).

Total and polyadenylated RNA were prepared from cultured cells using the RNeasy kit (Qiagen) and the Poly(A) Pure™ kit (Ambion), respectively. Samples of total (30 μg) or polyadenylated RNA (3 μg) were fractionated by electrophoresis in agarose-formaldehyde gels and transferred to nylon membranes following standard protocols. The blots were hybridised at 42°C in ULTRAhyb™ buffer (Ambion) and washed according to the manufacturer's protocol. The PCR primer sequences used to generate the specific probes were presented elsewhere (Kim et al., 2003).

For PCR-based analyses, cDNA was generated using 2 μg of total RNA, oligo dT primer and MuLV reverse transcriptase for 1 hour, in a total volume of 22 μl, as recommended by the supplier (Invitrogen). A sample (1 μl) of cDNA was then used as template for PCR amplification with primers based on the 3′-untranslated regions of chicken INK4b and ARF, with GAPDH as a control: INK4b-F: 5′-CACCCCTCCCCTGATTCATTG-3′; INK4b-R: 5′-CGGACACACAGAAGCGTTCG-3′; ARF-F: 5′-CGCTTTGCTCGCTGCCCC-3′; ARF-R: 5′-TTGTTGAATCGGATGCCTC-3′; GAPDH-F: 5′-CCACAACACGGTTGCTGTATCCAA-3′; GAPDH-R 5′-TGCAGGTGCTGAGTATGTTGTGGA-3′. Amplification was carried out for 33 cycles of 94°C for 1 minute, 60°C (53°C for ARF and GAPDH) for 1 minute and 72°C for 2 minutes, followed by a final incubation at 72°C for 10 minutes. Products were analysed by electrophoresis in a 1.2% agarose gel and visualised by staining with ethidium bromide.

For immunoblotting, cells were harvested in lysis buffer containing 50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 2 mM EDTA, 1% Triton X-100, 1% SDS and protease inhibitors, and the protein concentration was determined using the BCA assay (Pierce). Samples (50 μg) were then boiled with 0.2 volumes of 5× Laemmli buffer, fractionated by SDS-PAGE in 12% gels, transferred to Immobilon-P (Millipore) and processed as previously described (Stott et al., 1998). Proteins were detected using horseradish peroxidase (HRP)-conjugated anti-mouse, anti-rabbit (Amersham) or anti-goat (Santa Cruz) antibodies and ECL (Amersham). For immunoprecipitation, lysates were prepared in NP40 lysis buffer (Stott et al., 1998). Approximately 1 mg of protein was immunoprecipitated overnight at 4°C with 5 μl of rabbit polyclonal antiserum pre-mixed with 20 μl of protein A beads (Pierce). The precipitated proteins were fractionated by SDS-PAGE in 12% gels and processed as above.

Rabbit polyclonal antisera against chicken ARF (SK8) and p15^INK4b (SK14) are described in the text. Chicken p53 was detected using the HP64 mouse monoclonal antibody (supplied by T. Soussi, Institut Curie, EA3493, 75970 Paris, France). Chicken p21^CIP1 was visualised using a polyclonal antibody provided by A. MacLaren and D. Gillespie (Beatson Institute for Cancer Research, Glasgow, UK). A monoclonal antibody against SV40 large T-Antigen was provided by the CRUK Monoclonal Antibody Production Service. The pan-Ras antibody (#OP41) and β-actin (#CP01) antibodies were purchased from Oncogene Science, and the monoclonal antibodies against human E2F1 (KH95) and the HA epitope (F-7) were from Santa Cruz (#sc-251 and #sc-7392, respectively). Smad2/3 were detected using a monoclonal antibody from Transduction Laboratories as described (Pierreux et al., 2000).

Chicken ovarian granulosa cells (cGCs) were isolated and dispersed from granulosa sheets by collagenase digestion. Dispersed cells were plated in DMEM containing 5% FCS, 25 ng/ml activin A and 50 ng/ml follicle-stimulating hormone (FSH) at approximately 50% confluence. After 24 hours, the cells were split 1:1.5, to maintain 50% confluence, into DMEM plus 5% FCS, without additional growth factors. Cells were incubated overnight and then treated with varying concentrations (2-25 ng/ml) of TGF-β. RNA was prepared after 4 hours and samples (20 μg) of total RNA were analysed by northern blotting as described.

Chicken embryo whole-mount and section in situ hybridisation for INK4b was performed as described (Henrique et al., 1997) using a probe derived from the 3′-untranslated region of INK4b (Kim et al., 2003). Immunohistochemistry was performed using standard procedures. Anti-phospho-histone H3 (Upstate Biotechnology) immunostaining was performed following in situ hybridisation for INK4b and nuclei detected with TO-PRO-3 iodide (Molecular Probes) in SlowFade Light Antifade kit (Molecular Probes).

We are grateful to J.-M. Buerstedde for DT40 cells, to P. Vogt and H. Filoteo for MC29 and MH2 viruses, to T. Soussi for providing an antibody that detects chicken p53, and to A. Maclaren and D. Gillespie for an antibody against chicken p21^CIP1. Thanks are also due to M. Mitchell for assistance with bioinformatics, to I. Mason for interpretation of the in situ hybridisation data, to F. Nicolás for help with the Smad2/3 immunofluorescence, and to C. Hill for comments on the manuscript. K.K. and B.S. were supported by a grant from the Austrian Science Foundation, project SFB-604.