Oncogene-induced senescence

INTRODUCTION

Oncogene induced senescence (OIS) was first observed following expression of an oncogenic form of RAS in normal human fibroblasts.¹ This process resembles replicative senescence, a phenomenon observed in aged cultures of primary human fibroblasts which is generally attributed to telomere attrition and resultant loss of proliferative capacity.² While the in vivo relevance of OIS was at first questioned, its significance was strongly supported by a dependence on tumour suppressor networks that impinge on the p53 and pRb pathways.^1,3-6 More recently, the demonstration that certain oncogenes can elicit a senescence response in vivo provided unambiguous evidence that OIS is an important mode of tumour suppression.^7-9 Moreover, while most mechanistic studies of OIS have been conducted in fibroblasts in vitro, it is important to note that a similar phenomenon can be observed in a range of cell lineages in vivo.^7-11

While its importance is now beyond doubt, there are many key questions that remain unanswered with regard to OIS. How homogeneous is this response in different cellular contexts and which aspects of the phenotype are of fundamental importance? Activated oncogenes such as H-Ras^V12 arrest cell cycle progression beyond the quiescence restriction point in G1 or G2 with no concomitant block in cell growth.¹² A hypertrophic senescent condition ensues characterised by a flat and enlarged morphology, an increase in acidic β-galactosidase activity and dramatic chromatin condensations affecting proliferation promoting genes such as cyclin A.^13,14 Recent studies suggest that a definition of OIS on the basis of phenotypic traits of human primary fibroblasts undergoing Ras-induced senescence may be too restrictive and that there are differences between human and mouse fibroblasts as well as between fibroblasts from different tissues sources, and indeed other cell types.¹⁵ Another key question is whether the process is invariably an aberrant response to mitogenic signalling and if so how important is DNA damage signalling and chromatin remodelling for execution of the phenotyope? And finally, which factors define the relative stability of OIS in different cell backgrounds? Is it the strength of the signal, its duration or the cellular context in which it is expressed that is the key?

To shed light on these questions, we begin with a review of the essential players and pathways in OIS, focusing mainly on evidence gleaned from murine systems where genetic ablation and inhibitor studies have been combined to increase our understanding. The requirement for specific pathways and players will be reviewed particularly in the context of Ras-induced senescence. We will then consider evidence for the Runx gene family that have recently emerged as important mediators of OIS. Finally we will attempt to build these observations into a coherent picture and suggest fruitful ways ahead to understand and exploit this knowledge in the control of cancer cell behaviour.

1. INDUCTION PHASE: GROWTH SIGNALS AND PROLIFERATIVE STRESS

Critical requirement for mitogenic signalling pathways in Ras-induced senescence

Oncogenic Ras is initially mitogenic in primary fibroblasts but with time an antiproliferative response ensues and cells acquire a characteristic senescent morphology.¹ Paradoxically the Raf-MEK-ERK signalling cascade that provides cells with constitutive mitogenic signals also triggers Ras-induced cell cycle arrest through accumulation of p16^Ink4a, p19^Arf, p21^Waf1 and p53.^1,3,16 Further analysis in human fibroblasts revealed that the RAS-MEK-ERK signalling cascade directly activates p38MAPK and that this event is absolutely required for the onset of premature senescence.¹⁷ Perturbation of p38MAPK signalling has also been demonstrated to modulate Ras-induced senescence in mouse tumour models^18,19 and studies in primary murine cultures have revealed an essential role in replicative and stress (culture shock) induced senescence.²⁰ Together these data identify p38MAPK as a critical sensor of cellular stress and a major component of the senescence response.

Ras-induced senescence absolutely requires Raf-MEK-ERK activation but recent studies also implicate the phosphatidylinositol-3-kinase (PI(3)K) / protein kinase B (AKT) pathway. Perturbation of PI(3)K signalling through loss of PI(3)K (p85 α or β subunits) or the use of specific PI(3)K inhibitors renders primary murine embryonic fibroblasts (MEFs) highly susceptible to premature senescence, at least in part through upregulation of p27^Kip1 expression.^21,22 Despite their observed sensitivity to suppression of the PI(3)K pathway, murine fibroblasts remain relatively refractory to negative feedback regulation of PI(3)K signalling which was recently identified as a critical component of Ras-induced senescence in human fibroblasts.²³ Specifically, aberrant activation of Ras or Raf induced multiple genes known to affect Ras-GTP levels including Sprouty, DUSP and RasGAPs. ERK signalling persisted albeit at low levels but the PI(3)K pathway was effectively silenced and AKT activation dramatically inhibited.²³ The Pten tumour suppressor protein normally inhibits PI(3)K signalling but is commonly mutated in human tumours carrying activated Raf mutations. The basis of this co-selection may be acquired resistance to termination of the oncogenic signal and hence a bypass of premature senescence.²⁴

Proliferative stress and DNA damage signalling induced by Ras

Ras-induced senescence is preceded by a strong proliferative burst that terminates with the engagement of a DNA damage checkpoint response (DDR). DNA damage is “sensed” by two Ser/Thr protein kinases: ATM (ataxia-telangiectasia, mutated) and ATR (ATM and Rad3-related) that phosphorylate Histone H2AX Ser¹³⁹ molecules flanking the sites of DNA damage, marking them as foci of active DNA repair.^25,26 In addition ATM and ATR phosphorylate transducer checkpoint kinases CHK1 and CHK2 to direct cell cycle arrest at both the G1 and G2/M phases of the cell cycle. The remarkable stability of the senescent cell cycle arrest in human cells has been attributed to the latter block.^25,26 Until recently the role of DNA damage in premature senescence in murine cells remained ambiguous. p19^Arf null primary MEFs are resistant to Ras-induced senescence but retain sensitivity to DNA damaging agents suggesting that DNA damage is not the primary signal.²⁷ However several reports concur that activated Ras induces a mitogenic response that precedes the onset of senescence in rodent fibroblasts in vitro and benign tumours in vivo.^10,28,29 In a mouse model of chemically induced skin carcinogenesis, γ-H2AX positive lesions were detected in the basal and supra-basal layers of early benign skin papillomas that had accumulated Ras mutations and markers of senescence, providing a clear link between Ras-induced mitogenic signalling, DNA damage and the onset of senescence.²⁸ In addition, ablation of CHK2 or Seladin 1, sensors of DNA damage and Ros accumulation, bypasses Ras-induced senescence in primary rodent fibroblasts and facilitates Ras-induced transformation in nude mice.^28,30 From these data it seems likely that an active mitogenic response and DNA damage signal at least contribute to OIS in murine cells and may act as an early anti-cancer barrier in rodents.

2. Execution phase: engagement of the tumour suppressor network

The central role of p53 in Ras-induced senescence in murine primary fibroblasts

The importance of the p53 tumour suppressor pathway is now well established for OIS in murine fibroblasts. Loss of p53 or its upstream regulator p19^Arf permits cellular transformation of H-Ras^V12-transduced primary MEFs,^1,31 while re-establishment of p53 expression in murine tumours induces features of cellular senescence and tumour regression.^32,33 Over the last decade the list of genetic events that enable escape from Ras-induced senescence has grown (Table 1) but all impinge upon regulation of p53 or its downstream signalling effectors.

Table 1. Genetic defects that permit escape from Ras-induced senescence in murine fibroblasts.

Genetic Defect	Senescence pathway perturbed	Expression of Growth Arrest Markers	Tumourigenic		Cyclin Upregulation	Citation
			in vitro	in vivo
p53 -/-	p53 loss	-	+	+		1
p19Arf -/-	p53 stability	-	+	+		31
CHK2 -/-	p53 function		+	+		28
Seladin 1 KD†	p53 function		+	+		30
PML-/-	p53 function	+	+			53
Gadd45α -/-	p53 stability	+	+	+		44
PRAK -/-	p53 function		+	+		19
DMP1 -/-	p53 stability	-		+		35
ΔNP73	p53 function	-	+	+		39
KLF4	p53 transcription	-	+	+	D1	34
Sh††Btg2	pRb function	+	+	+	D1, E	54
Dec1 -/-	pRb function					55
E2FDB	pRb function	+			A, E	66
p107/pRb -/-	pRb function	+	-	-	E1	63
p130/p107/pRb -/-	pRb function	+	+	+		62
hDril-1	pRb function	+	+		E1	73
C/EBPβ -/-	pRb function	+	-	-	A	70
Runx2 -/-	pRb function	+	+	+	E, A2, B1	29

^†Knock down

^††Short hairpin RNA

Regulation of p53 function and consequences for Ras-induced senescence

In murine fibroblasts Ras-induced senescence absolutely requires activation of the p53 tumour suppressor pathway.^1,31 Oncogenic Ras converts p53 into a senescence inducer and events that reduce the duration or intensity of p53 signalling inevitably compromise p53-mediated growth arrest and facilitate Ras-induced cellular transformation.⁵ These include negative regulation of p53 expression,³⁴ modulation of p53 protein stability ^16,35-37 or perturbation of p53 function through specific protein:protein interactions or post translational modifications.^38,39 Acetylation of p53 on Lys³⁸² is tightly regulated through the formation of trimeric p53-PML-CBP complexes in the promyelocytic leukaemia (PML) nuclear bodies and is considered an essential component of the senescence response to oncogenic Ras.^38,40,41 p53 mutants that compromise Lys³⁸² acetylation or genetic deletion of PML permit escape from Ras-induced senescence and facilitate anchorage independent growth.³⁸ Moreover, protein kinases such as ATM, HIPK2 (Homeodomain-interacting Protein Kinase-2) and p38MAPK that phosphorylate p53 and facilitate acetylation of Lys³⁸² are required for p53-mediated tumour suppression and can cooperate with oncogenic Ras to induce cellular transformation if their activity is impaired.^40,42-47 These data place p53 downstream of Ras signalling networks and identify it as a key regulator of cell fate decisions between proliferation and senescence-like growth arrest. Mouse models of mammary carcinogenesis and human epithelial cancers support these observations, with tumours proliferating in response to physiological levels of oncogenic Ras but displaying an irreversible senescence-like growth arrest following chronic high level activation.^11,48 A similar correlation has been reported in human fibroblasts.⁶ The functional significance of p53 in human fibroblast senescence is somewhat eclipsed by p16^Ink4a which governs the irreversibility of the phenotype but conditional activation of p53 in p53 deficient human tumour cells has been reported to promote irreversible cell cycle arrest with features of senescence.^49,50 Although the tumour lines described did not express exogenous RAS, both harboured activating RAS mutations consistent with a role for p53 in RAS-induced senescence in, at least some human cells.^51,52

Effectors of p53 signalling that contribute to the onset of cellular senescence

Transcriptional targets of p53 required for cellular senescence include, Gadd45α, PML, Btg2, Dec and miR-34s. Gadd45α and PML positively regulate p53 function thereby amplifying the signal ^44,53 whilst Btg2 and Dec1 act downstream of p53 to modulate E2F dependent gene regulation.^54,55 MiR-34s functionally reproduce elements of p53 activity, including cellular senescence, in a context-specific manner.⁵⁶ Other gene targets associated with cellular senescence include Plasminogen activator inhibitor-1 (PAI-1) and p21^Waf1. PAI-1 is critical for replicative senescence in murine and human fibroblasts ⁵⁷ whereas p21^Waf1 mediates a reversible cell cycle arrest in response to DNA damage or reactivation of p53 but is not required for Ras-induced senescence in primary MEFs.⁵⁸ Preferential recruitment of p53 to the promoters of p21^Waf1 and Gadd45α was recently reported during replicative senescence lending weight to the hypothesis that specific transcriptional signatures coordinate the p53 response to multiple different stress stimuli.⁵⁹

Critical role of pRb-regulated pathways in murine senescence

The retinoblastoma (Rb) tumour suppressor gene product plays a key role in premature senescence acting as a crucial gatekeeper of cell cycle progression.⁶⁰ In human cells activation of pRb correlates with the implementation of an irreversible cell cycle arrest that cannot be overcome by subsequent inactivation of pRb whereas in murine cells inactivation of pRb, like p53, results in reversal of the senescent phenotype, suggesting that pRb is required not only for the initiation of the program but also for maintenance of the senescent state.⁶¹ pRb and the closely related proteins p107 and p130 are required for oncogenic transformation by Ras but deletion of pRb and p107 is sufficient for escape from Ras-induced senescence.^62,63 These results demonstrate that in the absence of both pRb and p107 murine fibroblasts are resistant to the antiproliferative effects of the p19^Arf/p53 tumour suppressor network and suggest alternative pathways by which pRb family proteins complement p53 to implement Ras-induced senescence. Possibilities include regulation of E2F:pRb repressor complexes ^64,65 and/or the formation of distinct chromatin structures on E2F responsive promoters to regulate the permanence of the senescent state.^66,67

E2F mediated transcriptional repression is required for Ras-induced senescence

E2F dependent gene transactivation is required for mitogen-induced cell cycle re-entry but during normal cell cycle progression or under conditions of proliferative arrest, E2F associates with hypophosphorylated pRb family proteins to form transcriptional repressor complexes that target multiple cell cycle genes including the G1/S cyclins.^64,65 Disabling this response renders MEFs resistant to senescence induced by p19^Arf, p53 or Ras^V12.⁶⁶ Partial impairment through loss of p16^Ink4a or p21^Waf1 and therefore incomplete dephosphorylation of pRb family members is not sufficient illustrating the extremely tight regulation imposed on this pathway.^58,68,69 A significant proportion of genetic events that enable escape from Ras-induced senescence do so against high level expression of growth inhibitory factors along with deregulated cyclin gene expression, suggesting a checkpoint defect downstream of the Ras-ERK-p53 signalling cascade (Table 1). Closer examination of these cooperating events reveals a strong preference for mutations that antagonise E2F/pRb mediated gene repression. These include physical disruption of the complex,^66,70 genetic loss or hyperphosphorylation of pRb family members,^{54,55,62,63,71} or ectopic expression of E2F responsive genes such as cyclin E1 and B-Myb.^72,73 Collectively these data provide definitive evidence of the critical role played by E2F-mediated transcriptional repression during premature senescence.

SWI/SNF chromatin remodelling factors and induction of premature senescence

Molecular features that define E2F:pRb mediated gene repression in murine senescence are poorly understood. During normal cell cycle progression E2F:pRb recruits SWI/SNF chromatin remodelling factors to CDE/CHR (cell cycle dependent element / cell cycle gene homology region) repressor elements in the promoter regions of E2F target genes such as Cyclin A and Cyclin E.⁷⁴ CDE/CHR elements are also responsive to p53 following DNA damage and SWI/SNF complexes have been linked to the formation of γH2AX DNA damage foci following ionising irradiation of NIH3T3 fibroblasts.^64,74-76 Whilst no obligatory relationship appears to exists between SWI/SNF and the onset of murine senescence, two mutually exclusive ATPase components of SWI/SNF have been linked to cell cycle control and tumour suppression in a number of different systems.⁷⁷ Brahma related gene 1 (Brg1) induces senescence-like growth arrest and increased resistance to nuclease digestion in Rat mesenchymal stem cells,⁷⁸ while Brahma (Brm) accumulates in growth-arrested murine fibroblasts and is downregulated during Ras-induced transformation of NIH3T3 cells.⁷⁹ It may also be significant that senescent murine cells express cytologically detectable foci of condensed nuclei and chromatin that are indistinguishable from those associated with SWI/SNF-dependent repression of E2F responsive genes in proliferating cells.^65,74,80

Chromatin remodelling and maintenance of the senescent phenotype

Maintenance of the senescent phenotype requires stable repression of E2F target genes through pRb recruitment of chromatin remodelling factors.⁶⁷ In human cells these assemble as specialised domains of facultative heterochromatin called senescence-associated heterochromatic foci (SAHFs). Formation of SAHFs depends on intact pRb and p53 tumour suppressor pathways and the presence of PML nuclear bodies to assemble protein complexes prior to their translocation to the chromatin.⁸¹ The stability of the arrest in these cells relates to the intrinsic levels of p16^INK4A and the intensity of SAHF formation. IMR90 human fibroblasts, which express high levels of p16^INK4A, exhibit pronounced SAHFs and an irreversible senescent phenotype whereas BJ human fibroblasts senesce with poorly pronounced SAHFs and low level expression of p16^INK4A, and can be stimulated to divide after disruption of the p53 pathway.^67,82 By comparison, murine fibroblasts express poorly defined SAHFs and inactivation of either the pRb or p53 tumour suppressor pathways results in reversal of the senescent phenotype.^61,67,83 Indeed a mouse mammary tumour model of Ras-induced senescence describes dramatic chromatin condensations that lack recognisable molecular features of SAHFs such as hypoacetylated histones, methylated histones H3 on Lys 9 (H3K9me) or heterochromatin protein-1(HP-1).^11,13 Moreover murine fibroblasts deleted for SIRT-1, an NAD-dependent histone deactylase, retain sensitivity to Ras-induced senescence whereas loss of a generic chromatin remodelling agent, ING1, leads to a partially defective senescence-like response to oncogenic Ras that correlates with impaired heterochromatin formation.^84,85 In addition, Eμ-N-ras transgenic mice heterozygous for a mutation in a histone H3 methyltransferase, SUV39h1, succumb more rapidly to T-cell lymphoma than wild-type controls,⁷ but MEFs lacking Suv39h1/h2 remain susceptible to H-Ras^V12-induced senescence.⁶⁷ These data suggest that alternative modes of chromatin silencing are operative in murine senescence where relative stability is governed by the chromatin remodelling factors engaged.

3. The role of Runx2 in Ras-induced senescence

We recently reported that, in common with activated Ras, all three Runx proteins induce a senescence-like growth arrest in primary MEFs that depends on an intact p19^Arf/p53 tumour suppressor pathway.^29,86 This process provides a barrier to the expression of the oncogenic potential of the Runx gene family that seems to depend on a strong survival advantage of Runx-expressing cells.^86,87 The resemblance to Ras-induced senescence is far from complete, however, as primary cells expressing ectopic Runx undergo growth arrest without a preliminary phase of abortive replication.²⁹ These results suggest that ectopic Runx may be activating the downstream execution phase of OIS, bypassing the requirement for replicative stress signals. Consistent with this interpretation, we have shown that primary MEFs lacking Runx2 are resistant to Ras-induced senescence.²⁹ As we have suggested, the essential role of Runx2 in this context may be due to its relatively abundant expression in primary fibroblasts compared to other family members rather than a unique Runx2-specific function.

Our previous explorations of Runx2^-/- MEFs showed that these cells are refractory to H-Ras^V12-induced premature senescence despite hyper-activation of p38MAPK and the induction of a cascade of growth inhibitors and senescence markers, including p19^Arf, p16^Ink4a, p53 and p21^Waf1. A rationale for their continued growth was provided by the observation that Runx2^-/- cells also display elevated expression of S/G2/M cyclins and their associated cyclin dependent kinase (CDK) activities. This deregulation was evident both in the presence and absence of Ras^V12 and hence appeared to be intrinsic to the Runx2 null phenotype and presumably sufficient to override the effects of growth inhibitory signals.²⁹ The basis of this cyclin gene deregulation remains to be established, as does the possibility of lesions elsewhere in the Ras-induced failsafe process. Recent studies to investigate this phenomenon further are now described.

Hyperactivation of p38MAPK signalling is associated with elevated levels of Gadd45α

Hyperactivation of p38MAPK and escape from Ras-induced senescence in Runx2^-/- cells places Runx2 downstream of Ras/p38MAPK/p53 in the execution phase of the pathway. Gadd45α is a transcriptional target of p53 that amplifies the Ras/p38MAPK signal through a positive feedback loop and interacts with proteins such as CDC2, PCNA and p21^Waf1 to negatively regulate cell cycle progression at G1 and S/G2/M.⁴⁴ Murine fibroblasts lacking Gadd45α fail to undergo Ras-induced senescence and are permissive for cellular transformation despite induction of a full complement of growth arrest effectors.⁴⁴ Given the accelerated transit of Runx2^-/-/Ras^V12 MEFs through S/G2/M, their biochemical resemblance to Gadd45α null MEFs and the conserved Runx and p53 binding sites in intron 3 of the Gadd45α gene promoter,^29,44 we were interested to examine whether Runx2 was required for Ras^V12-induction of Gadd45α expression in this cell background. Surprisingly, quantitative reverse transcriptase PCR (qtRT-PCR) showed a robust activation of Gadd45α mRNA transcription in response to activated Ras that was greater in the absence of Runx2 (Figure 1). These results indicate that induction of the positive signalling loop between p38MAPK, p53 and Gadd45α ⁴⁴ is retained in Runx2 null MEFs but that execution of Gadd45α-mediated growth arrest is in some way disabled, leading to a checkpoint defect that is clearly revealed in the presence of H-Ras^V12.

Figure 1.

Loss of Runx2 promotes Gadd54α transcription in response to activated Ras. cDNA was prepared from littermate matched Runx2^-/- and WT cultures transduced with H-Ras^V12 or the PURO control vector and plated for 3 to 13 days in culture. Gadd45α expression was assayed at each timepoint by relative quantification to the day 3 WT sample and normalized to endogenous control hprt. The data is displayed as raw RQ values and is representative of two independent experiments on two littermate matched Runx2^-/- and WT lines.

PI(3)K signalling persists in H-Ras^V12- transduced primary MEFs and is unaffected by Runx2 disruption

The PI(3)K/AKT signalling cascade is an essential component of the early mitogenic response to RAS and is negatively regulated in human fibroblasts as they undergo premature senescence.²³ Since Runx-induced senescence was not accompanied by cellular proliferation and loss of Runx2 facilitates Ras-induced transformation, we were interested to determine the effects of Runx2 on PI(3)K/AKT signalling. To this end we compared the levels of phosphorylated AKT1 in wildtype (WT) and Runx2 null MEFs following retroviral transduction with activated Ras. As shown in Figure 2A the levels of phosphorylated AKT1 remained high and comparable over an 11 day growth curve in both genetic backgrounds despite robust activation of senescence-like growth arrest in WT cells over this time period.²⁹ These results suggest that PI(3)K signalling is preserved in the absence of Runx2 and that in contrast to observations in human fibroblasts, negative feedback regulation of this pathway is not required for the onset of senescence in murine primary fibroblasts.

Figure 2.

PI(3)K signaling is refractory to loss of Runx2 but uncoupled from p27^Kip1 regulation. (A) Total protein was extracted from early passage (p4) littermate matched Runx2 null and WT lines following retroviral transduction with H-Ras^V12 (Ras) and plating in culture for up to 11 days. The blot was probed against antibodies to phospho-AKT (Cell Signalling #9271) or AKT1/2 (sc-8312) as a loading control and is representative of two independent experiments. (B) Total protein was extracted from early passage (p4) littermate matched Runx2 null and WT lines directly after retroviral transduction with H-Ras^V12 (R) or the PURO (P) vector control. The blot probed against an antibody to p27^Kip1 (BD transduction labs. 610242). α Actin (sc-1616) was used as a loading control. (C) Western blot analysis as in (A) probed against antibodies to p27^Kip1 and α Actin. (D) Immunoprecipitation (IP) of lysates extracted from H-Ras^V12 transduced Runx2 null and WT lines with an antibody to Cyclin D1 (Biosource DCS11). The blots were subsequently immunoblotted (IB) against antibodies to p27^Kip1 and α cyclin D1. (E) Cell cultures as described in (D) were plated on poly-L-lysine coated chamber slides and labeled with an antibody to p27^Kip1 and a second antibody coupled to FITC. DAPI containing mountant (Vector Labs) was used to visualize the nuclei.

p27^Kip1 accumulates but is mislocalised in response to H-Ras^V12 in Runx2 null primary fibroblasts

AKT negatively regulates multiple growth inhibitory substrates including p27^Kip1. Inhibition of PI(3)K/AKT signalling and upregulation of p27^Kip1 expression has been previously reported to induce premature senescence in primary MEFs.²¹ To determine whether this signalling network was uncoupled in the absence of Runx2 we examined the levels of p27^Kip1 protein in WT and Runx2 null fibroblasts in response to activated Ras. To our surprise p27^Kip1 expression levels were elevated in the absence of Runx2 relative to control fibroblasts. This was apparent immediately after retroviral transduction and puromycin selection (Figure 2B), when activated Ras is reported to reduce p27^Kip1 transcription and protein stability,^88,89 and was sustained for a further 11 days in culture with no accompanying positive staining for senescence associated β-galactosidase, (Figure 2C).²⁹ p27^Kip1 transcripts remained responsive to Ras and equivalent to WT controls over the time course indicating that transcriptional regulation was intact in the Runx2 null genetic background (data not shown). Post translational regulation of protein stability represents an alternative but significant mode of p27^Kip1 control.⁹⁰ Cyclin E/CDK2 phosphorylates and targets p27^Kip1 for ubiquitin-mediated proteolysis and events that interfere with this process, such as sequestration by cyclin D kinase complexes or cytoplasmic mislocalisation, effectively increase p27^Kip1 stability but render it functionally inactive as an inhibitor of cell cycle progression.^90,91 Examination of the subcellular distribution of p27^Kip1 in Runx2-null cells transduced with H-Ras^V12 revealed a higher proportion sequestered in cyclin D1 complexes or localised to the cytoplasm relative to WT controls (Figure 2D & E). These data suggest that signalling pathways downstream of PI(3)K are perturbed in the absence of Runx2 to favour p27^Kip1 stability and functional inactivation but argue against a role for p27^Kip1 in Ras-induced premature senescence in wild type MEFs.

Why is S/G2/M cyclin gene expression deregulated in Runx2 null fibroblasts?

We previously demonstrated deregulated expression of multiple cyclin genes (A2, B1 and E1) and their associated cyclin dependent kinase activities in Runx2 null MEFs, which remained elevated in the presence of H-Ras^V12.²⁹ Interestingly, acute loss of pRb in primary MEFs dramatically upregulates cell cycle genes including cyclins A2, E1, B1 and B2,⁹² suggesting that functional loss of this key regulator might be the source of the Runx2 null defect. Examination of pRb family protein expression and phosphorylation status revealed this not to be the case, however (data not shown), indicating that the answer to this question must lie elsewhere.

The SWI/SNF complex is an essential component of E2F:pRb mediated cyclin gene repression during normal cell cycle progression and quiescence.^64,65 In addition, SWI/SNF nucleosome remodelling and Runx2 recruitment is required for effective expression of the osteocalcin gene ⁹³ and defective subnuclear targeting of Runx2 has been reported to bypass senescence and promote immortalisation of primary osteoblasts.⁹⁴ If Runx2 forms an integral component of the SWI/SNF complex then it is conceivable that loss of Runx2 directly contributes to deregulated cyclin gene expression and consequent escape from Ras-induced senescence in murine fibroblasts. Regulation of cell cycle entry and transit at the level of composition of the SWI/SNF chromatin remodelling complex is not unprecedented. In addition to the core BRG1 or BRM ATPases that are associated with proliferation and differentiation respectively,⁹⁵ SWI/SNF comprises seven more non-catalytic subunits including an ARID (AT Rich Interacting Domain) containing DNA binding protein. Recruitment of particular ARID proteins was recently reported to have specific and opposing effects on c-myc regulation and cell cycle control.⁹⁶

To determine whether Brg1 or Brm expression was perturbed by loss of Runx2 we performed qtRT-PCR on WT and Runx2 null MEFs in the presence and absence of activated Ras. As shown in Figure 3 Brm and Brg1 transcripts were upregulated in the absence of Runx2. The trend persisted in the presence of activated Ras despite consistent repression of Brm levels as previously reported.⁷⁹ These observations are once again unexpected and not consistent with a single loss of expression model with regard to Brm and Brg1. However the functional status of these proteins and the components of the SWI/SNF complexes remain to be explored.

Figure 3.

Loss of Runx2 confers increased expression of SWI/SNF ATPases but does not prevent Ras-dependent repression of Brm. cDNA was prepared as for Figure 1.

PROSPECTS

Where do the RUNX transcription factors belong in the overall scheme of premature senescence? Alternative possibilities are indicated by numbers 1-6 in Figure 4. Upstream of p53, we have evidence that Rac1 activity is elevated in p53 null primary MEFs in response to ectopic Runx1expression (unpublished observation AK). Rac1 was previously identified as a critical mediator of ROS production in response to activated Ras and has been associated with both cellular senescence and DNA repair.^60,97 In addition, Runx1-dependent expression of p19^Arf correlates with a senescent phenotype in primary MEFs,⁹⁸ and Runx1 has been identified as a key regulatory subunit of HIPK2 that facilitates PML-dependent phosphorylation of p53 and cell cycle arrest.^45,99,100 While these are plausible mechanisms to link Ras signalling to Runx, our data in the Runx2 null genetic background strongly argues for a defect downstream of p53, since escape from Ras-induced senescence occurred against high background levels of all the major senescence markers including p53.²⁹ Similarly, downstream effectors of p53 appeared to be intact in the absence of Runx2, despite evidence that at least some of these, notably p21^Waf1 and Gadd45α may function as Runx targets. Indeed, Gadd45α appeared to be hyperactivated, consistent with our previous observations of p38MAPK signalling.²⁹

Figure 4.

Possible perturbation of the cellular senescence machinery by Runx transcription factors. 1. Runx1 elevates Rac1 activity (unpublished observations-A.K.). 2. Runx1 transcriptionally activates p19^Arf expression. 3. Runx1 is cofactor for HIPK2 which facilitates acetylation of p53 on Lys³⁸². 4. Runx and p53 binding sites feature in multiple p53 regulated genes. 5. Loss of Runx2 is associated with deregulated cyclin gene expression normally repressed by E2F:pRb:SWI/SNF complexes. 6. Runx proteins interact with chromatin remodeling factors recruited during senescence to mediate stable gene repression.

The most compelling explanation for the failure of Runx2 null cells to undergo H-Ras^V12-induced senescence remains the elevated expression of the S/G2/M cyclin genes, implying a defect at the level of E2F:pRb:SWI/SNF dependent gene repression. The Runx proteins interact with a wide range of co-repressors including histone deacetylases and histone methyltransferases previously identified in the SWI/SNF chromatin remodelling complex and in SAHFs.^93,101,102 It is tempting to speculate that Runx2 comprises an integral component of the SWI/SNF complex, repressing cyclin gene expression at multiple points in the cell cycle to effect an irreversible senescence-like growth arrest. Moreover, this model may be extended to a general role for the Runx family. In this regard it is interesting that Runx1 is required for N-Ras-dependent growth arrest in haematopoietic progenitor cells,¹⁰³ while RUNX1 loss of function mutations have been associated with activated Ras signalling in AML and myelodysplastic syndrome/AML.^103-105 However, it should also be noted that direct genetic changes to Runx genes may not be necessary as cytoplasmic mislocalisation of RUNX proteins, notably RUNX3, has been reported in early breast and gastric cancers.^106-108 An interesting further prospect is that loss of RUNX expression or function may serve as an alternative means by which cancer cells acquire the ability to overcome failsafe blocks to cell cycle progression. A corollary of this hypothesis is that pathways and effectors of RUNX tumour suppressor inactivation may be amenable targets for therapeutic intervention in cancers of multiple lineages.