PDGF-induced ODG progresses more rapidly in p27 deficient mice
p27 acts as a tumor suppressor in several mouse models of cancer (Blain et al., 2003
; Cipriano et al., 2001
; Di Cristofano et al., 2001
; Fero et al., 1998
; Martins and Berns, 2002
; Park et al., 1999
; Shaffer et al., 2005
). Loss of p27 expression correlates with decreased survival in human ODG (Cavalla et al., 1999
; Kamiya and Nakazato, 2002
; Korshunov et al., 2002
), suggesting that p27 might play a contributing role in ODG. To determine if loss of p27 had an effect on PDGF-driven gliomagenesis, we crossed nestin-tva (Ntv-a) mice onto a p27 mutant (p27D51/Δ51
) background. The protein produced by this allele lacks the amino terminal 51 amino acids and does not bind cyclin-cdk complexes. Tumor formation was induced with the RCAS/tv-a system. HA tagged PDGF was sub-cloned into the RCAS vector (RCAS-PDGF-HA) (Shih et al., 2004
) to promote constitutive expression of PDGF. Chicken DF1 cells were transfected with RCAS-PDGF-HA, and virus-producing cells were directly injected into the frontal cortex of newborn pups to allow for viral infection of proliferating cells in the subventricular zone. We used transgenic mice that express the tv-a receptor under the nestin promoter, which restricts infection to neural/glial progenitors, glial restricted progenitors, OPC and astrocyte precursors.
, and p27D51/D51
mice were injected with 104
DF1 cells that produced the RCAS-PDGF-HA virus and evaluated for 12 weeks. Mice were sacrificed earlier if they became morbid. Virtually all tv-a negative mice, regardless of their p27 genotype, were still alive at the end of 12 weeks, and as expected, none had any evidence of tumor when sacrificed. Thus the following descriptions focus solely on those mice that express tv-a. The onset of morbidity was equivalent in the p27D51/D51
animals, with 50% having to be sacrificed by the fourth week compared to nine weeks for p27+/+
mice (). H&E staining demonstrated that >95% of the mice sacrificed had a tumor. Tumors were diffuse and consisted of small cells; there was evidence of perivascular and perineuronal satellitosis, white matter tracking, and subpial infiltration (Liu et al., 2007
). At the end of 12 weeks, only 6% of p27D51/D51
mice and 20% of p27+/D51
mice remained alive compared to 38% of the p27+/+
mice. Virtually all the mice that survived to 12 weeks had tumors. The differences in survival were statistically significant by log-rank analysis. Thus, genetic deficiency of p27, either at one or both alleles, was sufficient to increase morbidity and mortality associated with enforced expression of PDGF.
p27 deficiency causes a decrease in survival in PDGF-induced oligodendrogliomas
To determine if p27 deficiency affected the nature of the disease, three tumors from p27+/+
animals were randomly selected for further analysis by immunohistochemistry. Both the p27+/+
tumors were positive for the oligodendrocyte marker olig2 (Liu et al., 2007
) and the stem cell marker sox2 (Avilion et al., 2003
; Ferri et al., 2004
) (), and negative for the neuronal marker NeuN (Jin et al., 2003
). With the exception of trapped astrocytes, the tumors were also negative for GFAP. Thus, the presence of two p27 alleles contributed to better prognosis and longer survival when ODG was driven by expression of the growth factor PDGF.
p27 is a haploinsufficient tumor suppressor that is not dose-dependent in PDGF-induced ODG
In most other tumor models, haploinsufficiency at the p27 locus manifests as dosage-dependence—heterozygous mice succumb to disease at a rate intermediate to that observed in wild type or nullizygous mice (Fero et al., 1998
; Park et al., 1999
). However, in PDGF-induced ODG, symptom free survival was equivalent for p27+/D51
mice and p27D51/D51
animals (P<0.5, ), suggesting that in this tumor type p27 might not be dose-dependent. We determined whether p27 protein was still produced by the wild-type allele. Nuclear expression of p27 was observed in both p27+/+
tumors, but not in the p27D51/D51
tumors (). Thus the tumor suppressive effect of p27 is not dosage dependent in this tumor type.
p27 deficiency does not affect PDGF expression from RCAS vectors, alter PDGF signaling pathways, or increase proliferation in incipient tumor cells
It was formally possible that p27 loss exerted its effects in this system by altering PDGF signaling. Unlike many solid tumors, ODG are diffuse and contain highly motile cells. The masses are heterogeneous and contain both oligodendrocyte tumor cells and uninfected oligodendrocytes, making direct measurement of proliferation or apoptotic indices on tumor cells problematic. We therefore employed an approach that allows us to overcome these limitations. We previously demonstrated that cell lines can be developed that recapitulate the molecular and cellular features of incipient tumor cells (Dai et al., 2001
; Liu et al., 2007
; Soos et al., 1996
). Post-natal day 1 Ntv-a positive p27+/+
brains were cultured for 4 days and infected with an RCAS virus that expresses PDGF-HA under control of the viral LTR and GFP under the control of the SV40 promoter (Becher et al., 2008
). Following infection and passaging for 6 weeks, we generated wild type (WT1) and p27 deficient (KO1) cultures in which greater than 99% of the cells expressed GFP (, upper panels). RCAS integration was confirmed by fluorescence in situ hybridization (, lower panels).
Wild type and p27 deficient glial cells lines were generated from primary brain cultures by infection with the RCAS/PDGF/GFP virus
Western blotting revealed that both WT1 and KO1 PDGF-infected glial cell lines evaluated at early passage (≤p20) expressed equal amounts of HA-tagged PDGF and activated phospho-PDGFRβ (, compare lanes 1 and 3). Levels of phospho-Erk1/2 (pERK1/2), phospho-p70S6kinase (pp70S6K) and cyclin D1 were similar between the genotypes, but there was a modest increase in the expression of p21 and cyclin A in the KO1 cells (, compare lanes 1 and 3). When cells were treated with the receptor tyrosine kinase inhibitor PTK787, levels of pERK1/2, pp70S6K, and cyclin D1 were decreased to similar extents in both WT1 and KO1 cells. Levels of Rad51, a protein involved in DNA repair, were equivalent. Although the amount of PDGFR phosphorylation, cyclin A and p21 also decreased in the KO1 cells relative to controls, the decreases were not as strong as those observed in the WT1 cells (, compare lanes 2 and 4). Nevertheless, treatment reduced DNA replication as measured by BrdU incorporation in both WT1 and KO1 cells (). There was a trend towards PTK787 resistance in the p27 deficient culture, although this was not statistically significant, and apoptosis was equivalent in WT1 and KO1 cells in the presence or absence of PTK787 (data not shown). Thus, p27 deficiency did not appear to grossly alter PDGF signaling, or the requirement of PDGF for proliferation.
Wild type and p27 deficient PDGF/GFP-infected glial cell lines have similar growth and proliferation rates at early passage and are sensitive to PDGFR kinase inhibition
In early passage cell populations, the growth and proliferation rates of the p27+/+ and p27D51/D51 PDGF- infected glial cultures were equivalent (). As passage number increased, the p27 deficient cells exhibited an accelerated growth rate () and increased proliferation, based on BrdU incorporation (P<0.02) () compared to the wild type cells. Apoptotic indices, measured either by FACS analysis of Annexin V staining or immunoblot analysis of cleaved caspase 3, were unaffected (data not shown). These data suggest that loss of p27 did not affect the early growth response to the oncogenic signal, but that over time p27 deficiency was associated with the evolution of a more proliferative cell type.
Karyotypic abnormalities are more frequent in p27 deficient PDGF-infected glial cell lines
One non-cell cycle related mechanism by which p27 deficiency might contribute to tumor development is to increase genomic instability (Chibazakura et al., 2004
; Payne et al., 2008
; Shaffer et al., 2005
; Spruck et al., 1999
; Strohmaier et al., 2001
)g. We therefore addressed whether there was an increase in genomic instability during the early neoplastic changes occurring in PDGF-expressing glial cells. An additional set of independently derived wild type (WT2) and p27 deficient (KO2) cultures was established. More p27 deficient cells had an abnormal karyotype in early passage cultures (KO1 and KO2), compared to early passage wild type cultures (WT1 and WT2) (). The few abnormal karyotypes observed in the WT1 and WT2 cultures consisted of aneuploid and tetraploid cells. By contrast, many of the abnormalities observed in p27 deficient cultures also included translocations, segmental duplications and inversions (data not shown and ). A few representative karyotypes of normal cells and abnormal cells are shown in Supplementary Figure 1
. We attempted to culture primary brain tumors from tumor bearing mice for karyotype analysis. Only 1 of 5 primary tumors from p27+/+
mice proliferated in tissue culture whereas 2/2 p27D51/D51
tumors proliferated well in culture. Thus, we were only able to analyze one set of age matched p27+/+
tumor cultures. The p27D51/D51
cells had an increased number of abnormal cells compared to p27+/+
cells that was statistically significant (). The increased number of abnormalities suggests that the degree of chromosomal instability is elevated in p27 deficient cells, or that the mechanisms that eliminate such cells from the culture are incapacitated in p27 deficient cells.
p27 deficiency does not diminish the G2/M checkpoint invoked by the DNA damage response
Increased genomic instability in tumor cells might occur as a consequence of defects in DNA repair or checkpoint fidelity. To evaluate these possibilities, we irradiated early passage cells to induce DNA double strand breaks and evaluated three proteins involved the DNA damage response: Atm, Chk2, and γH2AX. Thirty minutes after γ-irradiation, levels of phospho-ATM, the activated form of ATM kinase, increased in both WT1 and KO1 cells () and returned to near basal levels by three hours. Chk2 phosphorylation was maximally increased by 30 minutes following irradiation, and increased levels were maintained over the duration of the experiment. γH2AX, a marker of DNA double strand breaks (Foster and Downs, 2005
), also increased in both wild type and p27 deficient cells by 30 minutes following IR. We did not observe a significant difference in γH2AX accumulation between wild type and p27 deficient cells before or after irradiation ( and data not shown). Similar results were observed in WT2 and KO2 cells (data not shown). Together, these results suggest that activation of the DNA damage response, at least in response to ionizing radiation, was largely equivalent in the wild type and p27 deficient cells.
p27 deficient cells activate the DNA damage response and maintain the G2/M checkpoint
We next looked downstream to G2/M checkpoint activation in response to γ-irradiation (). Phospho-histone H3 is a convenient flow cytometric marker for mitotic cells. Cells with unrepaired DNA double-strand breaks arrest prior to mitosis and are negative for phospho-histone H3. When WT1 cells were exposed to either 1 or 2 Gy, the number of mitotic cells decreased approximately 47% or 70%, respectively, compared to unirradiated controls. In KO1 cells, the number of mitotic cells decreased 66% and 88%. Pretreatment with caffeine, an inhibitor of ATM kinase activity (Sarkaria et al., 1999
), abrogated G2/M checkpoint activation in both WT1 and KO1 cells (). Similar results were obtained with WT2 and KO2 cells (data not shown). Thus p27 deficient cells appear capable of pausing at the G2/M checkpoint to repair DNA damage. To determine if G2/M checkpoint activation was intact in vivo
, we irradiated tumor bearing mice (10 Gy), and performed phospho-histone H3 immunohistochemistry three hours after irradiation. We also analyzed tumors from unirradiated control mice. The percent of phospho-histone H3 positive cells was similar between p27+/+
unirradiated tumors (). No phospho-histone H3 positive cells were identified in either p27+/+
tumors three hours after irradiation. We also looked at the induction of apoptosis. Similar amounts of cleaved caspase 3 and TUNEL staining were seen at 3 and 12 hours, respectively, in wild type and mutant animals (data not shown). These results confirm that G2/M checkpoint activation was intact in vivo
p27 deficiency affects the persistence of Rad51 foci formed and phosphorylation of the C-terminus of BRCA2 in response to DNA damage
DNA double strand breaks are repaired by either Rad51 mediated homologous recombination or non-homologous end joining (NHEJ). p27 deficient cells display reciprocal translocations, implying that NHEJ is intact (Weinstock et al., 2006
). Since homologous recombination has been proposed to be the main DNA repair pathway utilized by cancer cells (Powell and Bindra, 2009
) and is favored in cells in G2, we focused our attention on homologous recombination.
We irradiated early passage wild type and p27 deficient cells and scored Rad51 foci over the next 24 hours. Formation of Rad51 nuclear foci is a quantitative measure of DNA repair by homologous recombination (Haaf et al., 1995
; Raderschall et al., 1999
). Since normal S-phase cells may contain up to five such foci, we required that a cell have greater than five foci to score positive. The percentage of wild type and p27 deficient cells containing Rad51 nuclear foci was similar one hour after irradition (), but there was a significant difference at three hours. Fewer KO1 cells (34%) formed Rad51 foci compared to WT1 cells, (62%) (*P<0.001) (). Similar results were obtained in WT2 and KO2 cells (data not shown). By 24 hours, WT1 cells had basal levels of Rad51 foci positive cells. However, more KO1 cells contained Rad51 foci, 11% compared to 3% of WT1 cells (**P<0.05) (), suggesting that p27 deficient cells were unable to repair double strand breaks as efficiently as wild type cells.
p27 deficient cells are defective in Rad51 nuclear foci formation and contain elevated GST-TR2 phosphorylation activity in response to γ-irradiation
The measurement of Rad51 foci positive cells could reflect differences in Rad51 protein levels or in the distribution of cells in various phases of the cell cycle. However, Rad51 protein levels and propidium iodide stained flow cytometric profiles of WT1 and KO1 cells were comparable at all time points (). Levels of Mre11 and Ku70, proteins involved in sensing DNA damage and NHEJ, respectively, were equivalent before and after irradiation in both wild type and p27 deficient cells (). These results indicate that the accumulation or maintenance of Rad51 nuclear foci in response to irradiation was perturbed in p27 deficient cells, suggesting that p27 contributes to homologous recombination-dependent DNA repair mechanisms.
The efficiency of homologous recombination can be modulated by cdk activity (Aylon et al., 2004
; Esashi et al., 2005
; Huertas et al., 2008
). Cdk-mediated phosphorylation of Ser3291 within the C-terminus of BRCA2 blocks BRCA2/Rad51 interactions and prevents Rad51-mediated DNA repair (Esashi et al., 2005
). Since loss of p27 may lead to enhanced cdk activity, we monitored cdk activity in p27 deficient cells using TR2, a C-terminal BRCA2 peptide containing Ser3291, as a substrate. WT1 and KO1 cells were irradiated (5 Gy) and lysed at 0, 1, 2, 3, or 4 hours post-irradiation. Lysates were incubated with GST or GST-TR2 bound to Sepharose beads and labeled with [γ-32
P]ATP (). Phosphorylation of TR2 decreased by 50% one hour after irradiation in WT1 cells (black bars) and increased back to basal levels by 3 hours (). By contrast, phosphorylation of TR2 remained unchanged in KO1 cells (open bars) at all time points, suggesting that p27 deficient cells contained elevated GST-TR2 phosphorylation activity after exposure to γ-irradiation.
To determine if cdk activity was altered in p27 deficient cells, kinase assays were performed using Histone H1 as a substrate. Cyclin B1-associated kinase activity was reduced after irradiation in WT1 cells (Supplementary Figure 2A
, black bars). However, there was a more modest decrease in KO1 cells (Supplementary Figure 2A
, open bars), suggesting that cyclin B1-associated kinase activity could be partially mediating the BRCA2-TR2 phosphorylation activity in p27 deficient cells. Cdk2 kinase activity decreased in WT1 cells (Supplementary Figure 2B
, black bars) after irradiation with an even longer response observed in KO1 cells (Supplementary Figure 2B
, open bars). These results demonstrate that p27 deficient cells are responsive to DNA damage inducing checkpoints and suggest that elevated kinase activity associated with p27 deficiency following ionizing radiation may perturb the function of BRCA2.
The frequency of chromatid breaks is increased in p27 deficient cells
Cells that fail to repair DNA double-strand breaks during the G2/M checkpoint either die, or “adapt” to the checkpoint by progressing into mitosis with chromatid breaks. To determine if p27 deficiency increased the number of chromatid breaks in mitotic cells, wild type and p27 deficient cells were irradiated with 1 Gy, and allowed to recover for 1.5 hours prior to the addition of colcemid. No chromosome breaks/gaps or chromatid breaks were observed in unirradiated WT1, WT2, KO1, or KO2 cells (data not shown). We therefore focused our analysis on the presence of chromatid breaks, an indicator of repair defects in G2 when homologous recombination occurs. There were two-fold more mitotic cells with greater than three chromatid breaks in KO1 cultures compared to WT1 cultures (P=0.02) (), and seven-fold more mitotic cells with greater than three breaks in KO2 cells compared to WT2 cells (P<0.01) (data not shown) There was a higher incidence of Rad51 foci in p27 deficient cells at 24 HR post-irradiation (). To determine if this correlated with an increase in unrepaired breaks, we irradiated WT1 and KO1 cells with 5 Gy and analyzed 100 metaphases 24 HR after irradiation. There was an increase in the number of mitotic cells containing unrepaired breaks, 12% in KO1 versus 4% in WT1 (P=0.03) (). When metaphase spreads were prepared from unirradiated primary tumor cultures, we observed more mitotic cells with unrepaired breaks in the p27D51/D51 primary tumor culture compared to the age matched p27+/+ primary tumor culture (P<0.02) (). These findings suggest that p27 deficient cells were less proficient at repairing DNA double strand breaks in G2 and eventually progressed to mitosis with broken chromatids. Thus, loss of p27 diminishes the cell's ability to repair DNA double strand breaks, leading to chromosomal instability, a hallmark of cancers with poor prognosis.
p27 deficient cells contain increased chromatid breaks in response to γ-irradiation
Mitotic chromatid breaks in primary tumors