Mutational activation of K-Ras is a key genetic event involved in the initiation of pancreatic carcinogenesis. However, K-Ras generally fails to transform precursor lesions into invasive cancers due to activation of powerful fail-safe programmes that counteract transformation and growth. The importance of cellular senescence, a permanent cell growth arrest, is increasingly being recognised as a critical fail-safe programme in pancreatic carcinogenesis. Emerging evidence suggests that oncogene-induced senescence requires transcriptional induction of the CDKN2A gene locus as well as comprehensive chromatin modifications involved in epigenetic silencing of pro-proliferative genes. Moreover, recent work in pancreatic cancer mouse models proposes that inactivation of the CDKN2A tumour suppressor locus is the molecular switch required for senescence evasion and unleashed K-Ras driven malignant transformation in the pancreas.
Statistics from Altmetric.com
Cellular senescence was first reported by Hayflick and Moorhead in1961 as the ultimate and irreversible loss of replicative capacity occurring in primary somatic cell culture.1 While in the beginning regarded as a cell culture artefact by many investigators, evidence has accumulated during the last decade showing that cellular senescence has an important role in vivo during ageing and also in tumorigenesis.2 A major cause of this permanent growth arrest was found in telomeres, which are non-coding nucleoprotein complexes found in the extremes of chromosomes.3 Telomeres associate with specific proteins to form protective caps that prevent chromosome ends from being recognised as double strand breaks by the DNA repair machinery.3 During successive cellular divisions, telomeres in normal human cells shorten progressively and, when telomeres erode down below a threshold length or are missing completely, the cell ceases to divide itself and becomes senescent.3–6 Thus, telomeres function as a ‘molecular clock’ that determines the proliferative capacity and hence control the life span of cells.7 Importantly, senescence can be induced also in the absence of any detectable telomere shortening or dysfunction by numerous conditions such as cellular stress or oncogene activation.8 ,9 Oncogene-induced senescence (OIS) has emerged as a powerful tumour suppressor mechanism protecting cells from unrestrained proliferation imposed by oncogenic signalling.9–11 For instance, it has been shown that normal cells, when forced to express high levels of oncogenic Ras, undergo a permanent and irreversible cell cycle arrest.10 ,11 OIS is frequently found in premalignant lesions but is essentially absent in advanced cancers, suggesting that malignant tumour cells can find ways to bypass or escape senescence.9 ,11–14 Importantly, genetically engineered mice models including those aiming at pancreatic tumorigenesis driven by oncogenic K-Ras have contributed to a better understanding of the molecular principles in OIS and, in addition, have provided strong evidence that evading senescence is a fundamental step in tumour development and progression (figure 1A).15–18 In this review we discuss the current molecular concept of senescence induction, maintenance and evasion in pancreatic tumorigenesis with a particular focus on the CDKN2A tumour suppressor pathway.
Oncogene-induced senescence (OIS)
Serrano et al10 first reported the potential of oncogenic Ras to induce a permanent cell growth arrest in primary human and mouse cells. Along with the irreversible growth arrest, oncogenic Ras subsequently promoted morphological and molecular changes in affected cells that were indistinguishable from those described in the context of replicative senescence.5 ,9–11 In fact, senescent cells display a number of characteristic features which allow their identification in vitro, regardless of the insulting event. For instance, senescent cells usually adopt a large and flattened cell structure, a vacuole-rich cytoplasm and a large nucleus with prominent nucleoli.11 ,19 They also exhibit global changes in chromatin structure (eg, senescence-associated heterochromatin foci (SAHF)) and are characterised by global reprogramming of gene expression.20 ,21 This includes expression and secretion of multiple cytokines and growth factors, thus indicating that senescent cells are by no means metabolically inactive but rather communicate with each other and with cells of the environment as well.22 Unfortunately, senescent tumour cells lack distinctive morphological features in vivo and, despite great efforts by many laboratories, no single molecular marker has been identified yet that specifically distinguishes senescent cells from other forms of growth arrest. Senescence should therefore always be characterised by a combination of markers including the two OIS sensing tumour suppressor proteins p16INK4a and p19ARF, encoded by the CDKN2A locus, and molecules that indicate accumulation of DNA damage (eg, yH2AX).23 ,24 In addition, senescent cells often exhibit biomarkers that accompany execution of the senescence programme such as senescence-associated β-galactosidase (SABG).9 ,11 Staining for SABG, though not specific for senescence, is the most accepted and widely used marker for permanent growth arrest when assessed at a suboptimal pH of 6.0. Accordingly, the Tuveson laboratory has most recently reported that SABG staining is indeed the most reliable and specific marker to detect senescent cells in precursor lesions of both murine and human pancreata.16 ,25
RAS signals and senescence induction in mice and humans
The RAS family of GTPases (H-RAS, N-RAS, and K-RAS) comprises three highly conserved proteins with roles in numerous vital cellular functions including proliferation, differentiation and senescence.26 Activating Ras mutations, which are among the most common genetic lesions in human tumours, cause constitutive activation of downstream effector cascades, even in the absence of extracellular stimuli.27 ,28 For instance, oncogenic Ras activates MAPK signalling cascades (eg, ERK, MAPK and p38) to drive senescence in various cell systems (figure 1B).26–29 In addition, other downstream effectors (eg, Pi3K-AKT) and transcription pathways (eg, E2F, c-Myc and nuclear factor of activated T cells (NFAT)) can also significantly contribute to the irreversible growth arrest.26 ,30–32 In line with these findings, which were mainly obtained from in vitro studies, transgenic expression of oncogenic H-Ras (HrasG12V) signalling caused an irreversible senescent phenotype in the mammary gland when expressed at high levels.33 Similarly, transgenic expression of N-Ras (NrasG12D) in murine lymphoid tissues renders lymphocytes highly susceptible to chemotherapy-induced senescence, whereas oncogenic activation of K-Ras causes a senescent phenotype in the murine lung tissue.12 ,14 Moreover, pancreas-specific expression of KrasG12V in transgenic mice promotes an initial burst of proliferation and subsequently leads to development of PanIN precursor lesions before cells stop dividing.9 ,16–18 These precursor lesions then exhibit many features of senescence including positive SABG staining and induction of cell cycle inhibitors. Successful progression of PanIN lesions towards frank adenocarcinoma requires evasion from senescence and this can result from additional genetic defects (eg, mutational inactivation of p53), an inflammatory microenvironment or epigenetic silencing of key tumour suppressor genes such as p16INK4a.34–36 An interesting example for overcoming senescence was recently reported by Kennedy et al18 in transgenic mice with simultaneous mutational activation of K-Ras and its effector Pi3K/Akt pathway. Since simultaneous mutational activation of K-Ras and Pi3K/AKT/mTOR signalling are indeed found in a substantial subset of human pancreatic cancers, and particularly in those with poor survival, these findings might be of clinical relevance for patients with pancreatic cancer.18
In contrast to genetically engineered mouse tumour models in which senescence is an accepted tumour suppressive mechanism, the biological relevance of permanent growth arrest in human malignancies is less clear. Accumulating evidence suggests a role for senescence in tumour defence in human skin carcinogenesis, where nevi with mutational activation of oncogenic BRAF or N-Ras show features of cell growth inhibition and stain positive for senescence-associated markers, indicative of successful protection against melanoma development.37 ,38 A new concept with potential implications in human tumorigenesis has been introduced by Zender and colleagues, showing that premalignant senescent hepatocytes can be cleared through a tumour antigen-directed immune response.39 The cell clearing process requires CD4 T cell activity and represents a powerful tumour defence strategy that contributes to blockade of tumour formation.39 Importantly, the authors provided evidence that this tumour defence mechanism also operates in human livers where it causes rapid clearance of senescent cells by the immune system. Whether the ‘senescence surveillance’ mechanism or a similar mechanism is operative in human pancreatic carcinogenesis remains uncertain.
Sensor and effector pathways in senescence
The p19ARF-p53 pathway in senescence
p53 is a central mediator of cell growth control.40 It does not merely monitor several cell cycle checkpoints, but is also involved in the execution of cell cycle arrest and fail-safe programmes such as apoptosis and senescence (figure 2A).41 ,42 Signals of mitogenic oncogenes, such as c-myc, K-Ras or NFAT, lead to activation of p53 which, depending on cell type and stimulus, induces either apoptosis or senescence and consequently leads to the elimination of cells with oncogenic activation (S Singh, personal communication).31 ,41–43 p53 is integrated in a complex network of ‘upstream’ sensors and ‘downstream’ effectors.40 ,42 An important sensor of oncogenic signals to p53 is p19ARF, which is encoded in an alternate reading frame (ARF)—in addition to p16INK4a—by the tumour suppressor locus CDKN2A (figure 2A).43 Activation of p19ARF antagonises the effect of the E3 ubiquitin ligase MDM2 which, in the event of reduced p19ARF expression levels, may act upon p53 to initiate proteasomal degradation. Oncogene induction of p19ARF, on the other hand, protects p53 from MDM2-mediated degradation, thereby contributing to stabilisation of the tumour suppressor gene.40 ,43 In addition to p19ARF, p53 can also be activated in response to ROS-dependent DNA damage (figure 1B), thus placing p53 as a central downstream effector pathway in different routes of oncogenic-induced senescence.44 In the cell nucleus, stabilised p53 binds to promoters of more than 300 target genes with implications for cell growth control.45 One such important p53 downstream target is p21 (WAF1/CIP1), a cell cycle inhibitor encoded by the CDKN1A gene.41–43 ,45 p21 binds to and inhibits the activity of cyclin-CDK2 and cyclin-CDK1 complexes, and thus functions as a regulator of cell cycle progression at the G1 phase (figure 2A).46 ,47 Importantly, stabilised p53 has transcription-dependent as well as transcription-independent functions to regulate important processes of tumour suppression. For instance, p53 interacts with numerous transcriptional activators and inhibitors that possess intrinsic histone-modifying activities, as well as with histone deacetylase complexes that modulate specific local chromatin structures and thus intervene in the regulation of senescence.48 ,49 In addition, various p53 regulators with independent functions in senescence execution can influence its activity by means of post-translational modifications. Prominent examples are PML and PTEN tumour suppressors, which stabilise the transcriptional activity of p53 by altering its phosphorylation and acetylation status in response to Ras signalling activation.13 ,32 In particular, p53 acetylation is possibly of immense significance for both the formation and maintenance of a senescent phenotype.50 In agreement with a key role of p53 in senescence and tumour suppression, mutational p53 inactivation is associated with disruption of senescence programmes and accelerated carcinogenesis in many tumour entities and model systems.13 ,32 ,34 ,46 ,50 In the pancreas, p53 inactivation on chromosome 17p has been reported in about 50–75% of carcinomas, and this genetic defect is usually followed by inactivation of the second allele.51 In the murine pancreatic carcinoma model, both genetic loss and the frequently observed stabilising mutation of p53 (R175P) allow K-Ras to bypass senescence and promote progression of precursor lesions to full-blown cancers.17 ,34
The p16INK4a-Rb pathway and formation of SAHF
p16INK4a belongs to the family of INK4 proteins and cell cycle inhibitors.43 It is encoded together with p15INK4b and p19ARF on chromosome 9p21. p16INK4a and p15INK4b exert critical antiproliferative functions through their abilities to block G1 to S phase cell cycle progression (figure 2A). p16INK4a specifically inhibits cyclin-dependent kinase 4 and 6 (CDK4 and CDK6)-mediated phosphorylation of retinoblastoma protein, thus sequestering E2F factors from DNA accessibility (figure 2A). This event is a decisive step in the inhibition of cell cycle progression and also in senescence initiation.41 ,52 It is accompanied by the recruitment of heterochromatin-forming proteins, which assist in the formation of long-range, hyperdense and transcriptionally inactive chromatin structures (figure 2B).20 ,53 These alterations were first described by Scott Lowe and coworkers as dot-shaped DNA sections with strongly condensed and transcriptionally inactive chromatin,20 and are referred to as SAHF (figure 2B). As a result of extensive reorganisational measures, SAHF-senescent cells consistently and irreversibly protect the organism from activation of E2F controlled cell cycle promoting genes.20 Specific histone modifications including trimethylation of histone H3 to lysine-9 (H3K9me) through Suv39H1 methyltransferase are of special significance for the formation of heterochromatin,54 as they allow docking of heterochromatin-stabilising protein HP1. HP1 binding to H3K9me might then serve as an anchor for further complexes and chromatin modifications that ultimately ensure maintenance of SAHF (figure 2B). The significance of Suv39H1 in senescence was confirmed in a Ras-transgenic mouse model with a Suv39−/− background. In contrast to wild-type Suv39H1-expressing mice that respond to K-Ras activation by the initiation of senescence, Suv39H1-deficient mice demonstrate accelerated tumorigenesis with unhampered proliferation and shortened survival.12 Further important elements of transcriptionally inactive SAHF include loss of histone H1 and the recruitment of high-mobility group A proteins which cooperate with p16INK4a in forming SAHF and maintaining senescence, as well as macroH2A, a marker of transcriptionally inactive chromatin sections.55 ,56 As mentioned above, increased expression levels of p16INK4a are frequently found in very early human PanIN lesions in which p16INK4a might significantly contribute to the induction and maintenance of senescence.35 Inactivation of p16INK4a, on the other hand, is required for tumour progression and is found in more than 90% of human pancreatic carcinomas.36 ,57 ,58 The causes of p16INK4a inactivation are manifold, ranging from homozygotic deletions (40%) and loss of the second allele (LOH) (40%) to point mutations and intensified promoter methylation (about 15%).59–61 The significance of p16INK4a as a central tumour suppressor of pancreatic carcinogenesis has also been confirmed in numerous excellent transgenic mouse models in which genetic loss of p16INK4a caused disrupted senescence, as determined by SABG stainings, and a massive acceleration of K-Ras-induced carcinogenesis.35 ,62
Epigenetic mechanisms in OIS
At the transcriptional level, all the aforementioned senescence pathways are tightly controlled and modulated by the chromatin state. This equally applies to mechanisms involved in senescence induction as well as those implicated in overcoming senescence. Epigenetics is defined as the study of inheritable changes in gene expression caused by mechanisms other than changes in the underlying DNA sequence.63 It refers to functionally relevant adaptation of chromatin structures so as to register and signal or perpetuate changes in its activity state. Chromatin is the functional entity of DNA, composed of nucleosomes that consist of 147 base pairs wrapped around histone octamers (figure 3). Each histone octamer consists of two copies each of the core histones H2A, H2B, H3 and H4.64–66 Transcriptional activation or silencing requires a series of highly dynamic changes in the chromatin structure. Histones, which are the foundation of chromatin, are modified by various post-translational modifications to control packaging density as well as the recruitment of further chromatin-modifying enzymes (figure 3).67 For instance, acetylation of specific lysine residues within histone tails is catalysed by enzymes known as histone acetyltransferases (HATs) such as CBP, p300 and PCAF. Removal of acetylation, on the other hand, is mediated by a growing family of histone deacetylases (HDACs). The interplay between HDACs and HATs negotiates the acetylation status of histones and thus establishes a chromatin structure that either permits accessibility of chromatin remodelling factors and opens promoter sites for transcription, or induces a highly condensed and silent chromatin structure.65 Of note, several recruited chromatin remodelling proteins with roles in pancreatic carcinogenesis are targets of genetic alterations, including ARID, mixed-lineage leukaemia 3 (MLL3) and others.66 ,67 Histone deacetylation by HDACs is frequently followed by histone methylation and forms a solid base for highly repressive chromatin structures such as formation of heterochromatin. The functional importance of HATs, HDACs and histone methyltransferases (HMTs) such as Suv39H1 is highlighted by the fundamental regulatory roles they play in developmental processes, and by the fact that their deregulation has been linked to the initiation and progression of cancers.68
DNA methylation in OIS
Methylation of CpG dinucleotides is the most important modification of DNA and, hence, plays a crucial role in epigenetic expression control. DNA methylation preferentially occurs in CpG islands and is mediated by specialised enzymes known as DNA methyltransferases (DNMTs). CpG islands are 500–2000 base pairs long cis-regulatory sequences with a GC content of about 60%. DNA methylation plays an important role in promoter inactivation, chromatin condensation and genomic imprinting.69 ,70 Promoter methylation usually causes inactivation of genes either by means of direct arrest of interaction with activators of expression or indirectly through extensive modifications of chromatin structure. At least three DNMTs play important roles in this process: DNMT1, DNMT3A and DNMT3B.70 DNMT1 is the most common form in somatic cells and appears to be mainly responsible for maintaining the state of methylation during cell replication and, therefore, for passing on the methylation profile to the daughter cell. Interestingly, DNMT1 is strongly overexpressed in pancreatic carcinoma where it is associated with a poor prognosis.71 Consistently, aberrant DNA methylation has been observed in both early- and late-stage human pancreatic tumours.57 Aberrant DNA methylation includes hypermethylation of specific DNA regions and also a global reduction in the quantity of methylated cytosines (hypomethylation).57 ,70 ,72 ,73 In fact, tumours are frequently marked by global hypomethylation74 ,75—a scenario that would explain epigenetic activation of numerous oncogenes in pancreatic carcinoma.76 In contrast, de novo methylation of CpGs in promoter regions of genes that are involved in the regulation of cell growth and senescence frequently lead to chromatin condensation and gene silencing.77 In addition, methylated DNA sequences can serve as identification sites for subsequent recruitment of histone modifiers and chromatin remodelling proteins, resulting in stable transcriptional silencing.78 This ‘self-reinforcing’ mechanism requires activation of DNMT1, which does not merely maintain DNA methylation but also creates a bridge to important histone modifiers including HDACs and HMTs (eg, Suv39H1). The degree of DNA packaging is therefore defined by DNA methylation and post-translational histone modifications. In pancreatic cancer, apart from PTEN and p27Kip1, the p73 tumour suppressor—and especially the CDKN2A locus which codes for the key senescence effector genes p16INK4a and p19ARF—are affected by self-reinforcing processes, as discussed in detail below.
The CDKN2A locus: a bona fide target for epigenetic silencing in cancer
Clearly, the CDKN2A locus is a key regulator of OIS in many organs including the pancreas.36 ,79 The locus is tightly controlled by a complex network of transcription regulators, histone modifying enzymes and chromatin reading proteins. During senescence, repressive molecules such as Polycomb group (PcG) members (see below) are released from the CDKN2A locus so that the chromatin structure opens and transcription of p16INK4a and p19ARF can be initiated.80 ,81 This change in the chromatin structure is strongly associated with demethylation of histone H3at lysine 27 (H3-K27me3) and may result from inactivation of the Ezh2 methyltransferase or, as shown most recently, from induction of jumonji domain-containing 3 (JMJD3), a demethylase that targets lysine residues in histone tails (figure 4).82 The locus is then prone to transcriptional activation by signalling pathways or cellular stress. Key players are Trithorax (TrxG) proteins that activate the CDKN2A locus in various cell lines and under senescence-inducing conditions.83 Mechanistically, the SWI/SNF complex of chromatin remodelling TrxG proteins displaces PcG proteins, DNMTs and HDACs from the CDKN2A locus and promotes acetylation of histone H4 at lysine 16 (H4-K16ac), a characteristic histone mark that labels for transcriptional gene activation. Like SWI/SNF, the HMT MLL1 belongs to the TrxG family and contributes to transcriptional activation of the CDKN2A locus through its ability to control histone modifications.83 ,84 Thus, release of repressive elements from the CDKN2A gene locus is critical for successful transcriptional induction of p16INK4a and p19ARF tumour suppressor pathways during senescence. Most importantly, silencing of the CDKN2A tumour suppressor locus is the molecular switch required for senescence evasion and unleashed K-Ras driven malignancy in cancer. Silencing of CDKN2A can be initiated by action of the PcG proteins. PcG are master repressor complexes that antagonise TrxG functions and stably silence transcription through chromatin modifications.80 ,81 There are two core PcG complexes, each of which is made up of multiple subunits. The mammalian PRC2 complex contains enhancer of zeste homologue 2 (Ezh2), a HMT that trimethylates histone H3 at lysine 27 (H3-K27me3), which is characteristic of inactive chromatin (figure 4).80 ,84 ,85 Ezh2 is transiently upregulated during pancreatic acinar cell regeneration, where it controls proliferation of pancreatic progenitor cells and contributes to silencing of the CDKN2A locus through increased H3-K27 trimethylation.84 ,85 Established repressive histone marks will then allow recruitment of PRC1 to form a stable inactive chromatin state so that transcriptional memory is maintained. Recent evidence suggests the existence of multiple compositions of mammalian PRC1 complexes containing chromobox homologue proteins (eg, CBX2, CBX4 and CBX8), ring finger proteins 1 or 2 (RING1 and RING2) and the stem cell marker Bmi1. Bmi1 was the first PcG protein to be implicated in cancer.86 It was initially identified as an oncogene that cooperates with myc in lymphomagenesis,86 and subsequent studies reported a robust expression of Bmi1 in various human cancers including pancreatic cancer.87 ,88 Bmi1 promotes regeneration of pancreatic acinar cells and triggers escape from cellular senescence through transcriptional silencing of the CDKN2A locus.88 ,89 Mechanistically, Bmi1 represses the gene locus by mediating the loading of CDC6 onto a cis-acting replication origin several thousand nucleotides upstream of the transcriptional start site,90 causing replication-coupled transcriptional repression of the CDKN2A locus. An additional and fascinating new layer in epigenetic control of the CDKN2A locus is defined by non-coding RNAs (ncRNAs) (figure 4). Although ncRNA transcripts are not translated into proteins, they are by no means junk RNAs as previously suspected.91 Rather, ncRNAs can recruit histone modifiers and therefore can actively participate in transcription regulation. One such important ncRNA is ANRIL, a long non-coding RNA transcript that is antisense to, and overlaps with, the CDKN2A locus.92 ,93 ANRIL is directly involved in epigenetic silencing of the CDKN2A locus through recruitment of PcG proteins (PRC2) and other key modifiers for gene silencing.93 ,94 However, it is not clear at this point whether ANRIL is overexpressed in pancreatic cancer and, if so, whether this contributes to long-term silencing of the CDKN2A locus as one could imagine.
Finally, our group has most recently unravelled a novel and alternative mechanism in p15INK4b gene silencing that might have crucial implications in overcoming senescence during pancreatic carcinogenesis.95 We found that the calcium/calcineurin responsive transcription factor NFATc2 targets the p15INK4b promoter for inducible and sequential heterochromatin formation and gene silencing. NFAT binding to its cognate promoter site induces stepwise recruitment of the HMT Suv39H1, causes local H3-K9 trimethylation and allows docking of the heterochromatin protein HP1γ to stabilise the repressor complex (figure 5).95 Conversely, inactivation of NFAT disrupts this repressor complex assembly and local heterochromatin formation, resulting in restoration of p15INK4b expression and induction of cell cycle arrest in pancreatic cancer cells.95
Taken together, these recent findings demonstrate that the CDKN2A locus is under the tight control of complex signalling cascades and epigenetic mechanisms. This is true for gene activation during OIS as well as silencing of the tumour suppressor locus when senescence is bypassed during the development of cancer.
Conclusions and future perspective
In the pancreas, activation of the K-Ras oncogene leads to the formation of premalignant PanIN lesions whose progression to invasive carcinoma is prevented by the initiation of senescence. For only a few years have we known that senescence is by no means an in vitro phenomenon but is an integral part of a cell defence programme that protects cells from transformation and progression into invasive tumour cells. The central aspects of OIS comprise activation of the CDKN2A gene locus which encodes for the master tumour suppressor pathways p19ARF and p16INK4a as well as the subsequent formation of transcriptionally inactive SAHF. Genetic or epigenetic inactivation of the CDKN2A gene locus or disruption of the key effector pathways may disturb the balance between oncogenic stimulation and cellular defence and thus favour tumour progression towards invasive carcinomas. In particular, epigenetic silencing mechanisms offer novel therapeutic strategies in cancer treatment. For instance, inhibitors of DNA methylation and histone modifications (eg, HDAC inhibitors) have shown partly convincing effects in many in vitro models and mouse cancer models, including those for pancreatic carcinoma. Their clinical applicability and benefits are currently being investigated in several preclinical and clinical trials.
We apologise to the many investigators for the inability to contain all references on this topic in this review. We thank Dr Sandra Baumgart and Dr Albrecht Neesse for their critical reading of the manuscript and helpful advice.
Contributors Both authors wrote and designed this review article.
Funding This work was generously supported by the Deutsche Forschungsgemeinschaft to VE (KFO210, SFB-TR17) and the German Cancer Research Foundation (Stiftungs-Professur).
Competing interests None.
Provenance and peer review Commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.