In vitro evidence for senescent multinucleated melanocytes as a source for tumor-initiating cells

Oncogenic signaling in melanocytes results in oncogene-induced senescence (OIS), a stable cell-cycle arrest frequently characterized by a bi- or multinuclear phenotype that is considered as a barrier to cancer progression. However, the long-sustained conviction that senescence is a truly irreversible process has recently been challenged. Still, it is not known whether cells driven into OIS can progress to cancer and thereby pose a potential threat. Here, we show that prolonged expression of the melanoma oncogene N-RAS61K in pigment cells overcomes OIS by triggering the emergence of tumor-initiating mononucleated stem-like cells from senescent cells. This progeny is dedifferentiated, highly proliferative, anoikis-resistant and induces fast growing, metastatic tumors. Our data describe that differentiated cells, which are driven into senescence by an oncogene, use this senescence state as trigger for tumor transformation, giving rise to highly aggressive tumor-initiating cells. These observations provide the first experimental in vitro evidence for the evasion of OIS on the cellular level and ensuing transformation.

up to 30% of human cancers. 7,8 This indicates that senescence bypass is a key feature of cancer development. The fact that many premalignant tissues with tumorigenic potential display features of senescence has led to the concept that OIS precedes transformation, and tumors arise from senescent tissue. 1,5,6,9 However, as OIS was long considered to be irreversible, it was not clear how this transformation process can take place. Recently, there has been accumulating evidence that OIS can be reversed under certain circumstances on the cellular level. Specifically, H-RAS 12V induces senescence in fibroblasts, which is caused by ribonucleotide reductase (RRM2) suppression and accompanying dNTP reduction. 10 Interestingly, the forced re-expression of RRM2 is able to overcome senescence. Still, is not known whether a senescent primary cell might give rise to cancer. While the later steps of tumor progression are fairly well understood, early events in tumorigenesis such as the transition of a benign senescent lesion to a tumor are still enigmatic.
Here, we reveal that long-term NRAS 61K activation in melanocytes triggers a strong senescent phenotype characterized by multinucleation, which then is followed by the post-senescence generation of tumor-initiating cells with stem cell-like properties. The results demonstrate that senescence in melanocytes can be overcome on the cellular level and can also be a source for malignant cancer cells.

Results
Multinucleated cells are frequently found in human nevi. When we analyzed randomly chosen sections of paraffin-embedded acquired melanocytic nevi, we found multinucleated nevus cells in 13/28 samples (46%). In most cases, samples contained few multinucleated cells, amounting to less than 1% of the nevus cell population, although some nevi showed a higher abundance of these abnormal cells, reaching up to 23% (Figure 1a). In vitro, such bi-or multinucleation of non-transformed cells occurs under conditions of senescence. [11][12][13][14] The strong nuclear disturbance induces an efficient cell-cycle block in non-transformed cells. The causes leading to multinucleation can be manifold and encompass UV damage, elevated levels of reactive oxygen species (ROS) and oncogenic stress, for example, by the melanoma oncogene N-RAS 61K that efficiently induces OIS in melanocytes. 11,12,14,15 To investigate the dual role of oncogenic RAS as senescence inducer and tumor driver, we overexpressed N-RAS 61K in a vector with simultaneous GFP coexpression in normal human epidermal melanocytes (NHEM). As expected, we observed an N-RAS 61K -induced stop in cell-cycle proliferation (Supplementary Figure S1) as well as the induction of biand multinucleation by N-RAS 61K (Figures 1b and c). Although OIS is widely considered to be a tumor-suppressive process, we found that cultivation of oncogene-expressing primary human melanocytes for more than 3 weeks led to the generation of proliferating cell clones (Figures 1d and e).
The cells overgrew the non-transfected cells, and clusters of viable cells were found in the supernatant, indicating anoikis resistance (Figures 1f and g). These cells lacked melanocyte differentiation markers such as DCT and TYRP1 (Figure 1h).
Multinucleated melanocytes give rise to proliferationcompetent progeny. To follow the fate of multinucleated cells after longterm N-RAS 61K stimulation, we used murine instead of human melanocytes, as replicative exhaustion can (h) RT-PCR analysis (28 cycles) of TYRP1 and DCT expression in NHEM control cells and two N-RAS 61K -transgenic NHEM cell clones, which grew out after several weeks of cultivation and developed anoikis resistance. The RT-PCR is representative of three independent biological replicates. Ribosomal S14 was used as a reference be prevented in these cells under well-established in vitro culture conditions by the addition of tetradecanoyl-12,13phorbolacetate (TPA). 16 We have shown previously that N-RAS 61K expression in melan-a murine melanocytes similar to NHEM cells leads to a multinucleated phenotype. This is caused by N-RAS induction of ROS and is accompanied by p53 signaling and senescence-associated β-galactosidase (SA-β-Gal) expression. 11 We used a doxycycline-inducible N-RAS 61K vector to drive long-term expression in melanocytes for extended time periods. After 14 days of doxycycline treatment, a large fraction of cells had become senescent, as demonstrated by their multinucleated phenotype and SA-β-Gal staining, which mostly correlated (Figure 2a, middle panel). In addition, proliferation was impeded, comparable to control cells which are not expected to proliferate in the absence of TPA (see Figure 5e: 'ctrl' and 'N-RAS 61K '). Importantly, the overall levels of oncogenic N-RAS were only slightly enhanced (Supplementary Figure S2A), thus showing that this phenotype is visible under conditions which mimic the in vivo situation. N-RAS 61K expression went along with activation of the MAPK and PI3K pathways, as seen by enhanced P-ERK1/2 and P-AKT levels (Supplementary Figure S2B). The N-RAS 61K mediated senescence is characterized by activation of the p53 pathway, as indicated by p19-ARF induction as well as enhanced DNA damage signaling, which was visible as enhanced γ-H2AX and P-p53 levels. p21 levels were barely altered and p16 levels were not visible (Supplementary Figure S2B). After 2-3 weeks, we observed that multinucleated cells were sometimes surrounded by small mononucleated cells (Supplementary Figure S2C). Furthermore, after 3-4 weeks of N-RAS 61K expression, proliferating cells appeared that overgrew the cell culture and formed three-dimensional cellular aggregates typical for transformed cells in vitro (Figure 2a, lower panel). Concurrently, we noted the appearance of viable, detached cells in the culture supernatant. Replating of such floating cells was followed by reattachment before they again gave rise to detached cells. We termed these cells 'N-RAS 61K -AR' (for 'anoikis resistant').
To trace the origin of the proliferating cells, we conducted time lapse movies of multinucleated N-RAS 61K cells after longterm oncogene stimulation. To resolve single cells, we transfected the N-RAS 61K cells with a plasmid, which allows to separately visualize membranes (red-purple) and nucleus (green). We recorded several highly multinucleated cells performing asynchronous cytokinesis, as for example shown in Figure 2b To compare the features of naïve N-RAS 61K cells, N-RAS 61K -AR cells that arise after senescence induction, and melan-a control cells (transfected with pTRE2hyg control vector, but simply termed 'melan-a' hereafter) we monitored proliferation under different cell culture conditions. N-RAS 61K -AR cells but not melan-a or N-RAS 61K cells grew efficiently in the absence of TPA ( Figure 2e) and under growth factordeprived conditions (Supplementary Figures S3A and B).
The unexpected proliferation behavior of N-RAS 61K -AR cells derived from senescent cell cultures was further confirmed by their capacity to efficiently form soft agar colonies ( Figure 2f). Furthermore, when we analyzed four different N-RAS 61K -AR cell clones, we noted several chromosomal aberrations, indicating that the cells had undergone a phase of genetic instability (Supplementary  Table S1).

N-RAS 61K -AR cells are tumorigenic and dedifferentiated.
To analyze tumorigenicity in vivo, melan-a, N-RAS 61K and N-RAS 61K -AR cells were injected subcutaneously into nude mice. Melan-a and N-RAS 61K cells gave rise to subcutaneous, pigmented, nevus-like structures (Figures 3a and b  To gain insight into the process leading to the generation of the highly aggressive N-RAS 61K -AR cells from a previously senescent cell culture, expression profiling of N-RAS 61K cells at different times of doxycycline stimulation and of N-RAS 61K -AR cells was performed. We found the expression of melanocyte differentiation markers to be highly regulated in response to N-RAS 61K expression (Figures 4a and b). Concurrent with senescence progression, differentiation markers, such as Tyrp1, Mlana, Dct, Sox10 and Mitf, were strongly induced after 14 days. However, expression of most differentiation markers decreased after 28 days of N-RAS 61K induction and in N-RAS 61K -AR cells. In contrast, Mitf levels were still higher in N-RAS 61K -AR when compared with the 6day sample. It is known that a certain level of MITF is similarly maintained in human melanoma, consistent with its role for melanocyte and melanoma proliferation and survival. 17 Concomitant with the decrease of differentiation genes, numerous genes, which are typical for neuronal tissues or the neural crest, were induced (Figures 4c and d). Many of these genes were still expressed in N-RAS 61K -AR cells. During embryonic development, melanocytes arise from the neural crest lineage. Re-expression of embryonic or neuronal genes is associated with a stem cell-like phenotype and is a hallmark of aggressive human melanomas. 18,19 To test whether the N-RAS 61K -AR cells display features of stem-like cells, we looked at the expression of embryonal and stem cell markers. Interestingly, Pdpn and Nanog were among the genes whose expression was markedly increased in N-RAS 61K -AR cells (Figure 4e). In particular, NANOG is considered as a master transcription factor for the maintenance of the undifferentiated state and  cellular self-renewal. Using a NANOG-driven GFP reporter, 20 we found a large fraction of N-RAS 61K -AR cells with activated Nanog promoter activity, which was not observed in N-RAS 61K cells ( Figure 4f). N-RAS 61K -AR cells also displayed a reduction of cytoplasmic volume and consequently a higher nuclear/cytoplasmic ratio compared with N-RAS 61K cells ( Figure 4g, Supplementary Movie S3), which is a morphological feature of stem-like and dedifferentiated cells.
Further analysis of the genes distinguishing N-RAS 61K -AR cells from all other samples led to two prominent groups: (1) positive regulation of proliferation and (2) meiosis genes (Supplementary Figure S5). The group 'positive regulation of proliferation' included numerous genes encoding growth factors (e.g., Kitl, Vegfa, Hbegf, Btc and Areg) and growth factor receptors (Pdgfra, Flt1) as well as enzymes involved in generation and transmission of growth-promoting signaling molecules (Ptgs2, Fabp4) (Supplementary Figure S5A and B). These data suggest that the AR cells are able to provide a number of autocrine signals, which sustain cell growth even under growth-restricting conditions such as starvation or withdrawal of TPA (as reported above). Indeed, ERK1/2 activity, which efficiently transmits growth-promoting signals in melanocytes, was increased in N-RAS 61K -AR cells and reached higher levels than in doxycycline-stimulated N-RAS 61K cells (Supplementary Figure S5C).
Among the genes included in the meiosis signature (Supplementary Figures S5D and E), Spo11 is also known as cancer/testis gene 21 and it displays copy number gains in different cancer types such as colorectal and gastric cancer (www.oncomine.org). Cyp26b1 is involved in retinoic acid metabolism and has a role in neural crest-derived tissue development. 22 Meiosis-related genes are often upregulated in melanoma cells that are particularly prone to so-called 'meiomitosis', describing the partial expression of meiosis machinery in mitotic cells. 23 Taken together, the expression profile provides evidence for melanocyte dedifferentiation, which is a typical feature of invasive melanoma cells, [24][25][26] upon long-term N-RAS 61K activation and reveals the expression of stem-like and meiosis-associated genes in N-RAS 61K -AR cells.

Senescent melanocytes give rise to tumor initiating cells C Leikam et al
The development of anoikis resistance is dependent on prior senescence. To verify that the cell population that gives rise to anoikis-resistant cells is truly senescent, we made use of the fact that p53 is strongly induced during senescence in N-RAS 61K -expressing multinucleated melanocytes. 11 N-RAS 61K cells were transiently transfected with the p53-specific GFP reporter plasmid. The pGreenFire-p53 plasmid is a lentiviral reporter vector containing a p53 responsive element, thus directly displaying p53 activity. A strong GFP signal was visible in senescent multinucleated N-RAS 61K cells, but not in control melan-a cells (Figure 5a). To obtain a 100% senescent melanocyte population, N-RAS 61K cells were strictly sorted for high GFP intensity and large cell size by flow cytometry (Figure 5b) and were then seeded at low density. These isolated multinucleated cells were termed N-RAS 61K -s (s for 'sorted'). They were able to repopulate the culture dish with small cells after 3 weeks ( Figure 5c). Furthermore, anoikis-resistant cells appeared after the cells became confluent (Figure 5d), confirming that they originate from the senescent cells. The anoikis-resistant cells were harvested from the supernatant ('N-RAS 61K -s-AR'), and their growth was monitored in comparison with N-RAS 61K -and N-RAS 61K -s cells. As expected, N-RAS 61K -s and N-RAS 61K -s-AR, but not the parental N-RAS 61K cells displayed a strong proliferation potential both in the absence of TPA and in growth factor-reduced serum (Figure 5e). In addition, they were able to efficiently grow in soft agar (Figure 5f).
To test the effect of an independent OIS inducer, we used the melanoma-associated receptor tyrosine kinase Xmrk, an EGFR family protein that strongly drives MAPK and PI3K pathways and induces a senescence phenotype with accompanying proliferation inhibition similar to N-RAS 61K . 11,27 Again, senescence of oncogene-expressing melan-a cells was followed by the generation of three-dimensionally growing cells and anoikis resistance (Figure 6a). Similar to the observation for N-RAS 61K -AR cells, Nanog promoter activity was also induced by this oncogene (Figure 6b).
Senescence induced by oncogenes is not immediately executed after oncogene activation, but only comes into action after a previous proliferation boost caused by strong, growthpromoting signals. 28 To determine whether a proliferation stimulus per se leads to the observed dedifferentiation and anoikis resistance, we cultivated the parental cell line melan-a for long time periods in the presence of the growth stimulus TPA, an efficient activator of PKC and the MAPK pathway. Apart from the addition of TPA, melan-a cells underwent the same treatment as N-RAS 61K cells. After TPA-treated melan-a cells had reached confluency, pigmentation increased ( Figure 6c). Upon 28 days of treatment, we could neither observe viable, detached melan-a cells in the supernatant nor the three-dimensional cell colonies we observed in N-RAS 61K cells (Figures 6c and d). Furthermore, these cells showed the induction neither of the pluripotency markers Pdpn and Nanog nor of the male meiosis marker Spo11, whereas the melanocyte markers Mlana, Tyrp1 and Dct were generally upregulated (though the data only reached significance in case of Mlana) (Figure 6e). Thus, the mere induction of proliferation is not sufficient to allow the generation of AR-cells.
Anoikis resistance does not depend on oncogene expression levels. To investigate whether a change in N-RAS 61K expression could affect senescence phenotype or anoikis resistance, we generated an independent N-RAS 61K transgenic cell line, which allows very fine-tuned expression induction accompanied by inducible GFP fluorescence. We applied 10, 25, 100 and 1000 ng/ml doxycycline to these cells and followed their behavior over time. From 25 ng/ml on, expression was visible on GFP level (Supplementary Figure S6A) as well as NRAS RNA level (Supplementary Figure S6B). At 0 and 10 ng/ml, no GFP or RNA signals were detected. Interestingly, senescence was only visible in those cells, which were treated with doxycycline concentrations starting from 25 ng/ml (Supplementary Figure S7A). In accordance, all these cells developed anoikis resistance (Supplementary Figure S7B) and dedifferentiation as well as upregulation of the AR marker genes Pdpn and Nefl ( Supplementary Figures S7C and E). Although a slight downregulation of differentiation genes was even detected after treatment with 10 ng/ml doxycycline, the stemness gene Pdpn and the neuronal gene Nefl were barely changed under these conditions. These data show that the degree of N-RAS 61K expression is not important for senescence and anoikis resistance once a critical, but small threshold level is reached.
Signaling pathways involved in senescence-dependent anoikis resistance. Next, we analyzed the pathways necessary for mediating the generation of N-RAS 61K -AR cells by using inhibitors, which were added to the cells at the start of the experiment. We first blocked the MAPK and PI3K pathways, which are both direct effectors downstream of N-RAS 61K , using the small molecule inhibitors PD184352 or LY294002, respectively. MEK inhibitor PD184352, but not PI3K inhibitor LY294002 prevented the  Figure S8A). In both cases, the cells were incapable of producing anoikis-resistant progeny (Figure 7a). Similar results were obtained with a different PI3K inhibitor (GDC-0941, Supplementary Figure S8C).
ROS have an important role in melanoma development and pro-tumorigenic signaling. [29][30][31] RAS activation leads to the induction of NADPH oxidase isoforms (Figure 7b, upper image and Weyemi et al. 32 ). This is accompanied by the generation of ROS. 32,33 Accordingly, activation of oncogenic N-RAS in melan-a cells causes DNA damage and p53 activation, as described above. To elucidate the role of DNA damage and senescence induction in N-RAS 61K -AR development, we applied the antioxidant glutathione ethyl ester or inhibited NADPH oxidase (diphenyle iodonium, DPI), ATM (caffeine) or p53 (pifithrine). DPI, caffeine and pifithrine, but not the glutathione component, were able to prevent the generation of multinucleated (Supplementary Figures S8A and B) and anoikis-resistant cells (Figures 7a and b, lower image), consistent with the involvement of DNA damage and senescence induction in this process. To test whether the inhibitors that prevent anoikis resistance are only involved in Figure 5 Only senescent NRAS 61K cells give rise to anoikis-resistant progeny. (a) Melan-a control cells (upper images) and N-RAS 61K cells (lower images) were cultivated in the presence of doxycycline (Dox) before they were transiently transfected with the pGreenFire-p53 reporter plasmid. Note the strong GFP staining for the multinucleated N-RAS 61K melanocytes. Scale bars, 75 μm. (b) After 1 week of doxycycline stimulation, p53-GFP-transfected N-RAS 61K cells were sorted for large cell size (as characteristic for the senescent multinucleated cells) and strong GFP positivity. (c) Appearance of cells 2 weeks after sorting (left) and 1 week later (right). Scale bars, 35 μm. (d) After 6 weeks, supernatant from sorted cells was transferred to a new dish and was allowed to attach overnight, before a crystal violet staining was performed. Please note that due to the low initial cell density in the dishes, the amount of cells in the supernatant is much lower than in previous figures, thus giving only rise to weak crystal violet staining. Arrows indicate the presence of some exemplary cell colonies (left). A magnification shows the accumulation of many cells to one colony, suggesting that the cells that gave rise to the colony were present as spheroid in the cell culture supernatant (right). Scale bar, 50 μm.  Figure S8D). Again, anoikis resistance was prevented, suggesting that the escape from senescence is also be affected by MAPK-PI3Kand DNA damage signaling.
In their tissue of origin, melanocytes encounter hypoxic conditions. We therefore performed a comparative analysis of N-RAS 61K -induced anoikis resistance and gene regulation under normoxic (20% O 2 ) and hypoxic (1% O 2 ) conditions. Importantly, N-RAS 61K was also able to trigger OIS in hypoxia (Supplementary Figure S9). Similarly, differentiation genes were downregulated (Figure 7d) and the stemness marker Pdpn as well as the neuronal marker Nefl were strongly upregulated under hypoxic conditions. The cells were also able to undergo anoikis resistance, albeit at lower efficiency compared with N-RAS 61K cells kept under normoxic conditions ( Figure 7c). To check whether continuous N-RAS 61K signaling is only required for the entry into senescence or whether it also affects senescence escape, we additionally investigated the effect of doxycycline withdrawal (thereby switching off N-RAS 61K ). Under normoxic and hypoxic conditions, continuous N-RAS 61K signaling was required even after senescence entry to allow all the features of anoikis resistance (Figures 7c-f).

Discussion
Despite the fact that premalignant tissue often displays features of senescence, senescence is generally considered as a tumor-suppressive mechanism, which either stalls the tumorigenesis of premalignant cells (in case of OIS) or blocks the progression of established tumors (in case of therapyinduced senescence (TIS) of tumor cells). It was recently shown that TIS is induced by a combination of IFN-γ and TNF, which leads to impressive control of tumor growth. 34 However, the idea that senescence exclusively operates as tumor-suppressive mechanism was previously challenged by the observation that, in their niche, senescent cells secrete factors that provide autocrine inhibitory signals, but also stimulate neighboring premalignant cells or cause low levels of tumor-promoting inflammation. 35,36 Furthermore, senescent cells harbor epigenetic markers, which are also found in cancer. 37 Here we provide experimental evidence that OIS can be overcome on the cellular level and describe the generation of highly tumorigenic melanoma cells, characterized by dedifferentiation and stem-like features, from senescent, multinuclear pigment cells.
The accumulation of bi-and multinuclear cells is a common feature of senescent cells in vitro, and DNA damage, caused by oncogenic or replicative stress, often mediates senescence. 38 We observed that the development of N-RAS 61K -AR cells was prevented by inhibiting senescence mediators such as NADPH oxidases and p53. Importantly, cells that were specifically isolated according to their high p53 expression gave rise to N-RAS 61K -AR cells, thus indicating that senescence induction is a pre-requisite for the observed anoikis resistance.
The PI3K and MAPK pathways are both required for the development of anoikis-resistance. MEK inhibition of N-RAS 61K cells prevented the development of multinucleated cells and consequently AR development. By contrast, PI3K inhibition allowed multinucleation, but prevented anoikis resistance, implicating a role for this pathway at later stages of AR development and highlighting a co-operativity between the MAPK and PI3K pathways in generating tumor-initiating cells from senescent cells. In line with these observations, it was recently shown that high PI3K activity, enabled by PTEN depletion, helps to overcome BRAF V600E -induced senescence. 39 Although we have observed this phenomenon in our experimental setting in vitro and clinical confirmation is still missing, the relevance of our findings for animal models and human melanoma is supported by several observations from other groups. First, mice with lung-specific expression of oncogenic RAS first develop senescent pre-malignant lung adenomas, which only later develop into adenocarcinomas. 6 As all affected lung cells express oncogenic RAS, and this induces senescence, the tumorigenic process is likely initiated after the induction of senescence. Second, staining of the senescence mediator and DNA damage marker p53 is retained in human melanoma with a history of high cumulative sun exposure, and is even associated with higher Breslow thickness, 40 suggesting residual pro-senescent signaling in melanoma and its compatibility with tumor maintenance. Finally, polyploid cells, which we identified as a source for the malignant offspring, are not only found in melanoma, but also in melanocytic nevi, as shown by us and others, 41,42 and a significant proportion of melanomas arise from pre-existing nevi. 43,44 In lentigo maligna (LM), also called melanoma in situ, multinucleated 'star burst cells' are present in up to 85% of patient samples. 15 LM mainly occurs in chronically sundamaged skin and is associated with UV light exposure as well as p53 positivity, 40,45 suggesting a causal relationship between UV-induced DNA damage, which is a strong senescence trigger, and the appearance of starburst cells.
In general, multinucleated or polyploid cells are at increased risk of becoming cancerous through a combination of genomic instability that facilitates the appearance of further mutations 46 and aberrant gene expression owing to the presence of multiple copies of many genes. Interestingly, irradiated cells of different origin displayed a behavior comparable to our observations, which was termed 'neosis'. 47 Here, the neotic cells also adapted a transformed phenotype. The 'budding' or 'neotic' behavior strongly resembles the process of 'depolyploidisation' by asymmetric division, which was recently described for tetra-or polyploid tumor cells. Such cancer cells can undergo a ploidy cycle where aneuploid cells give rise to para-diploid cells, thereby re-aligning to normal cell-cycle regulation and reducing the risk of lethal accumulation of DNA damage. 48,49 A shared phenomenon is the upregulation of meiotic genes such as Spo11, which is associated with depolyploidisation. This upregulation might reveal a functional parallel between asexual ploidy reduction, for example, occurring in cancer, and sexual ploidy reduction, occurring during meiosis. 50 Importantly, depolyploidization was also described to go along with increased levels of OCT4, NANOG and SOX2. 47 Stem cell-like features have also been noted in the progeny of polyploid giant cancer cells, 51 suggesting that multinucleated senescent cells and polyploid cancer cells share crucial pro-tumorigenic features.
Our data illustrate a mode of overcoming senescence that takes place in response to naturally occurring oncogenes such as the melanoma oncogene N-RAS 61K . This process is accompanied by the loss of melanocyte markers and by the increase of neuronal-and neural crest-like expression patterns. Importantly, the cells display features of highly dedifferentiated cells, such as a high nuclear-cytoplasmic ratio and increased expression of the transcription factor NANOG, which might explain their aggressive growth and early metastasis in vivo. Similar to our observations in melanocytes, transformed pigmented melanoma cells also show the tendency to express enhanced levels of stem cell markers or markers of melanoma initiating cells when they are forced into dedifferentiation, 25,52 thus suggesting a common epigenetic and pro-tumorigenic program that can not only be triggered in transformed melanoma cells, but also in senescent melanocytes.

Materials and Methods
Cell culture. N-RAS 61K and Hm hi -cells as well as melan-a cells were presviously described and cultivated as reported earlier. 11,53 In addition to these described pTRE2-hyg-N-RAS 61K cells, which were derived from one cell clone, another inducible N-RAS 61K transgenic cell line was generated. We used the vector pSB-ET-iE (M. Gessler, Dept. of Developmental Biochemistry, University of Wurzburg), which allows integration of N-RAS 61K by sleeping beauty-mediated transposition. Here, the responsive T6 promoter drives expression of N-RAS 61K and EGFP, the latter being separated from the NRAS gene by an IRES site. After transposition, these cells were selected with 1 μg/ml puromycin for 2 weeks, and the oligoclonal cell population was used for further experiments. N-RAS 61K was cloned into the vector using NheI/AfeI restriction enzyme sites. For long-term treatment, cells were cultured in starving medium (Dulbecco's Modified Eagle's Medium with 10% dialyzed, fetal calf serum (Gibco/Invitrogen, Karlsruhe, Germany) for 3 days before the assay. Induction of the oncogene was generally done using 1 μg/ml doxycycline for the indicated timespans, if not indicated otherwise. In general, medium containing doxycycline was changed 2-3 times per week.
The vectors pBabe-puro-H2-eGFP and pBabe-MN [EF1a-red membrane and green nucleus]-2APuro were delivered into the cells by retroviral infection according to standard protocols. pBabe-puro-H2-eGFP contains a fusion protein of histone 2B and enhanced GFP, thus leading to green fluorescence of the nucleus. pBabe-MN [EF1a-red membrane and green nucleus]-2APuro vector expresses a fusion product of the membrane targeting sequence MGCIKSKRKDNLNDDE with mCherry, followed by a T2A site and a fusion protein of histone 2B with eGFP.
RNA isolation and real-time PCR. RNA isolation was performed with TrIR solution (ABGene, Hamburg, Germany) from at least two independent biological Senescent melanocytes give rise to tumor initiating cells C Leikam et al replicates. Whole RNA was reversely transcribed using the RevertAidTM First Strand cDNA Synthesis Kit (Fermentas, Leon-Rot, Germany). Fluorescence-based quantitative real-time PCR was performed using the iCycler (Bio-Rad, Munich, Germany), and for each gene, three independent real-time PCRs were performed with cDNA from each biological replicate. The sequence of oligonucleotides used can be found in the Supplementary information (Supplementary Table 3). Gene expression was normalized to hypoxanthine-guanine phosphoribosyltransferase. Relative expression levels were calculated applying REST software.
SA-β-Gal staining. Cells were washed with phosphate-buffered saline (PBS; pH 7.2) and fixed with 3.7% formaldehyde in PBS for 5 min at room temperature. After washing, they were stained in the dark for 12 h at 37°C (using 1 mg/ml X-Gal, 40 mM citric acid/sodium phosphate buffer (pH 6.0), 5 mM potassium ferricyanide, 5 mM potassium ferrocyanide, 150 mM NaCl and 2 mM MgCl 2 ). Cells were then washed with PBS and were examined by light microscopy.
Soft agar growth. One milliliter of 1.2% agar was mixed with 1 ml 2 × DMEM supplemented with 10% FCS and plated onto 6-well plates. Upon polymerization, the solid agar was overlain with an equal amount of soft agar mix (0.6% agar) containing 4 × 10 4 cells per well. The polymerized soft agar was overlain with 100 μl of D10 medium every second day.
In vivo growth. Nude mice (NMRI, Harlan strain) were subcutaneously injected with 2.5 × 10 6 melan-a, N-RAS 61R or N-RAS 61R -AR cells per flank. Where indicated, 2 mg/ml doxycycline (Sigma-Aldrich) dissolved in 10% saccharose solution was supplied ad libitum by administration to the drinking water. After 4 weeks, N-RAS 61R -AR mice were killed due to considerable tumor load. In all other mice, growth of injected cells was monitored for 10 weeks after injection before killing the mice. All experiments were in accordance with the institution's guidelines.