A porcine model of osteosarcoma

We previously produced pigs with a latent oncogenic TP53 mutation. Humans with TP53 germline mutations are predisposed to a wide spectrum of early-onset cancers, predominantly breast, brain, adrenal gland cancer, soft tissue sarcomas and osteosarcomas. Loss of p53 function has been observed in >50% of human cancers. Here we demonstrate that porcine mesenchymal stem cells (MSCs) convert to a transformed phenotype after activation of latent oncogenic TP53R167H and KRASG12D, and overexpression of MYC promotes tumorigenesis. The process mimics key molecular aspects of human sarcomagenesis. Transformed porcine MSCs exhibit genomic instability, with complex karyotypes, and develop into sarcomas on transplantation into immune-deficient mice. In pigs, heterozygous knockout of TP53 was sufficient for spontaneous osteosarcoma development in older animals, whereas homozygous TP53 knockout resulted in multiple large osteosarcomas in 7–8-month-old animals. This is the first report that engineered mutation of an endogenous tumour-suppressor gene leads to invasive cancer in pigs. Unlike in Trp53 mutant mice, osteosarcoma developed in the long bones and skull, closely recapitulating the human disease. These animals thus promise a model for juvenile osteosarcoma, a relatively uncommon but devastating disease.


INTRODUCTION
Animal models of human cancers are crucial for the development of urgently needed diagnostic and therapeutic techniques. Genetically modified mice are widely used, but their small size and short lifespan preclude some preclinical studies. It is, for example, difficult to scale down radiological, thermal or surgical treatment of tumours, or perform longitudinal studies of tumour progression and remission, or longer-term response to therapy.
Mouse and human cancer biology also differ. Murine cells are more easily transformed in vitro than human cells, 1 and the set of genetic events required for tumorigenesis is different. 2 Mouse models may therefore not always provide the best representation of human disease.
Pigs are increasingly important in biomedicine and offer valuable complementary resources for cancer research. The number of pigs genetically modified to replicate human diseases has increased dramatically. 3 Valuable models such as cystic fibrosis and diabetes are established. 4,5 Work is also proceeding towards genetically defined porcine cancer models, such as inactivation of BRCA1 for breast cancer, 6 mutation of TP53 7 and, as we have reported, knockout and conditional activation of mutant TP53, 8 latent oncogenic KRAS mutation 9 and truncating mutations of APC to model colorectal cancer. 10 The relevance of pig cancer models depends on how closely they resemble human disease. Porcine cancer biology is a new field and many fundamental questions remain open. Not least, does replication of single or combined human oncogenic mutation(s) in a pig have an equivalent effect on cell transformation and tumorigenesis?
Sarcomas are a group of tumours originating from the mesenchyme. They fall into two categories: those with diseasespecific chromosomal translocations, and those with complex unstable karyotypes. Sarcomas of the second type, for example, fibrosarcoma, liposarcoma, chondrosarcoma and osteosarcoma, often have TP53 and RB1 mutations. 11,12 Consistent with this, transformation of human mesenchymal stem cells (MSCs) requires disruption of the p53 and Rb tumour-suppressor pathways, stabilisation of MYC, activation of oncogenic RAS and telomere maintenance. 13 We report that targeted mutation of endogenous porcine TP53 and KRAS in primary MSCs results in neoplastic transformation and tumorigenesis. Molecular analysis shows the process closely resembles human sarcomagenesis.
Osteosarcoma is a relatively rare solid tumour but the most common primary bone cancer. It predominantly affects young people and is highly malignant, requiring aggressive surgical resection and cytotoxic chemotherapy. Five-year survival for patients with metastatic osteosarcoma is only around 30%. 14 We report that pigs with heterozygous and homozygous inactivation of TP53 consistently develop osteosarcomas, providing a new model of osteosarcoma at human scale to understand and treat this devastating disease.
alterations, including oncogenic KRAS mutations or MYC amplification, may occur later during tumorigenesis. To show that similar molecular events lead to sarcomagenesis in a porcine model, we mutated the endogenous TP53 and KRAS genes in porcine MSCs.
The porcine TP53 R167H mutation is orthologous to human TP53 R175H . In a first round of gene targeting, we introduced a Cre-inducible latent mutant allele, TP53 LSL-R167H , into porcine MSCs and derived one MSC cell clone homozygous for the mutant allele (TP53 LSL-R167H/LSL-R167H ), which in unrecombined form is a TP53 knockout. 8 For brevity, this genotype is referred to as MSC-P. It was subjected to a second round of gene targeting to introduce a G to A substitution into KRAS codon 12 (G12D) and a floxed transcriptional terminator cassette (LSL) into the first intron, generating the Cre-inducible allele KRAS LSL-G12D (Supplementary Figures S1a-d). The resulting genotype (TP53 LSL-R167H/LSL-R167H , KRAS LSL-G12D/+ ) is termed MSC-PK. One MSC-PK cell clone was transduced with Cre and subclones were isolated that had excised LSL cassettes from both mutant TP53 alleles and the mutated KRAS allele (Figure 1b), termed MSC-PKC.
The MYC gene is often amplified in human cancers 15 and has a pivotal role in transformation of human primary cells. 2,16,17 We added a porcine MYC expression vector to one MSC-PKC cell clone and derived a pool (MSC-PKCM) that showed an average 1.7-fold increase in MYC mRNA expression and a similar increase in MYC protein ( Supplementary Figures S2a and b). These genetic modifications are summarised in Figure 1a.
Reverse transcriptase PCR (RT-PCR) and sequence analysis showed that MSC-PKC and MSC-PKCM cells expressed mutant p53-R167H, wild-type KRAS and mutant KRAS-G12D mRNAs and biochemical evidence demonstrated increased levels of activated GTP-bound Ras and accumulation of mutant p53-R167H protein (Figure 1c), similar to their human and murine counterparts. 18,19 The progressive genetic modifications also resulted in phenotypic changes, such as accelerated cellular proliferation (Figure 1d  The nodule consists mainly of matrigel with fat islands, capillary sprouts and isolated cells with slightly irregular nuclei. Scale bar indicates 500 μm. The arrow indicates an area of higher cellularity. Lower: Mesenchymal cells with slightly irregular nuclei embedded in a myxoid stroma are evident at higher magnification. Scale bar indicates 100 μm. (g) H&E-stained sections of MSC-PKCM derived tumour. This tumour is mostly highly cellular, but a matrix-rich area is still present, for example, as indicated by an asterisk. In the cellular areas, large tumour cells with pleomorphic nuclei are evident. A chronic inflammatory reaction is also present. Scale bar indicates 200 μm.
(MSC-PKC and MSC-PKCM) were xenografted into two immunodeficient mice. After 45 days, MSC-PKC cells formed a small nodule in one of the four injection sites with few atypical cells ( Figure 1f). MSC-PKCM cells gave rise to tumours in three of the four injection sites in the other animal, classified as low-grade fibroblastic sarcomas ( Figure 1g). Two such tumours were explanted and cultured to derive a porcine sarcoma cell line (poSARCO).

Molecular analysis of transformed MSCs and sarcomas
Microarray analysis was carried out to identify transcriptional changes associated with the stages of porcine MSC transformation. Figure 2a and Supplementary Table S1 show expression profiles for six groups of genes associated with cellular transformation: cellular transformation, telomere maintenance, cell cycle control, p53 target genes, apoptosis, and chromosomal instability (CIN). Cellular transformation. TP53 was not expressed in uninduced MSC-PK cells but upregulated after Cre activation. MYC expression increased in MSC-PKCM owing to exogenous expression. KRAS expression increased at each stage of cellular transformation and was the highest in poSARCO cells, likely owing to spontaneous gene amplification, as in many tumours. 20 Quantitative PCR (Q-PCR) analysis revealed two KRAS copies in wild-type MSCs, MSC-PKC and MSC-PKCM, while poSARCO had six. Sequence analysis of KRAS mRNAs (exon 1 to exon 4) detected higher expression of KRAS-G12D than wild-type KRAS in poSARCO (Figure 2b).
PoSARCO cells also exhibited increased levels of activated GTPbound Ras protein ( Figure 2c).
Primary human cells can escape cellular senescence through simultaneous inactivation of p53 and Rb. 21 The CDKN2A gene products p16INK4α and p14ARF (p19ARF in mice) regulate the Rb and p53 pathways, [22][23][24] but their roles differ between humans and mice 25 and little is known about the pig. As shown in Figure 2a, RB1 expression remained low in the MSC derivatives, but p14ARF (ARF) expression varied with the stages of cellular transformation. The microarrays used did not include p16INK4α, so p16INK4α and p14ARF expression was determined by Q-RT-PCR. Figure 3a shows that p16INK4α expression was diminished in MSC-PK cells, almost absent in MSC-PKC cells and similarly low in MSC-PKCM cells. Investigation of 21 CpG dinucleotides in the 5′ region of p16INK4A revealed that none were methylated in wild-type MSCs and progressively greater methylation in uninduced MSC-PK, activated MSC-PKC cells and the MSC-PKCM pool (Figure 3b).
Telomere maintenance. No reactivated TERT expression was observed (Figures 2a and 3c) in transformed MSC derivatives, but there was evidence for activation of the ALT mechanism, as frequently observed in human MSC-derived sarcomas. 26 BLM, FEN1, FANCD2 and, to a lesser extent, FANCA were upregulated. Genes involved in telomere elongation and capping, RAD51, RAD54B and BRCA2, were also upregulated, while shelterins TERF1 and TERF2 and POT1 remained low, possibly indicating telomere deprotection (Figure 2a, Supplementary Table S1).  Table S1).
p53 target genes and genes associated with apoptosis. Reduced expression of the key feedback regulator MDM2 indicates that mutant p53-R167H has impaired ability to transactivate its targets. This extends to genes important for cell cycle arrest, such as CDKN1A, CDKN1B and GADD45A, expression of which was lower than in wild-type MSCs and unaffected by induction of p53-R167H expression. Impaired capacity for apoptosis is indicated by lower expression of components of the death receptor pathway, for example, FAS, TNFSF10 (TRAIL) and TNFRSF10B (KILLER/DR5); and of the mitochondrial pathway, for example, BBC3 (PUMA), BAX and CASP6 (Figure 2a, Supplementary Table S1).
To investigate the dysfunction of mitotic checkpoints suggested by deregulated expression of cell cycle control genes, cells were treated with nocadazole to disrupt the mitotic spindle and induce mitotic arrest. The number of cells in G 0 /G 1 , S and G 2 /M phase determined by flow cytometry is shown in Figure 4a. Figure 4b shows the ratio between nocodazole treated and untreated cells in G 1 phase. Human colon cancer cell lines known to have CIN and mitotic checkpoint deficiency (SW480, CaCo2) and chromosomally stable human cell lines of non-tumour origin (HEK293) or from colon cancer (HCT116, microsatellite unstable) provided comparisons. G 1 -peak ratios ranged from 0.05 in HEK293, which show the strongest blocking response to nocodazole, to 0.2 for CaCo2 cancer cells, which showed a weak response. Transformed porcine MSCs showed elevated ratios of cells in G 1 phase after nocodazole treatment, similar to the CIN-positive cell lines SW480 and CaCo2, and poSARCO cells displayed an even higher ratio ( Figure 4b).
In summary, these data suggest that porcine MSCs resemble human MSCs in requiring perturbation of the p53, Rb, KRAS and MYC signalling pathways combined with a spontaneous telomerase-independent immortalisation step to convert to a fully transformed phenotype.
TP53 inactivation in pigs is sufficient for development of osteosarcomas We determined whether pigs carrying the latent TP53 LSL-R167H mutation 8 in heterozygous or homozygous form would spontaneously develop tumours. Animals were regularly observed for alterations in health and wellbeing, and individuals were slaughtered periodically for necropsy examination.
To date, nine heterozygous knockout pigs aged up to 32 months have been examined by necropsy, as summarised in Table 1. There was no evidence of tumours or other abnormalities in animals o 16 months. Of the five older animals, four had tumours (ID: 34,36,47,60), and one animal (ID: 49) had disseminated calcifying, ossifying lesions but no identifiable primary tumour. Figure 5a shows gross morphology and histology of tumours from two heterozygous and one homozygous animal.
Lesions from three pigs (animal ID: 34,36,47) were determined by histology as osteoblastic osteosarcomas. Two animals (animal ID: 34, 36) had osteosarcomas within long bones (metaphyseal region of the tibia and tuber olecrani) (Figure 5a, top). The other animal (animal ID: 47) had a tumour in the nasal conchae, classified as an osteoblastic and chondroblastic osteosarcoma (Figure 5a, bottom). Animal 60 showed a calcified fibrous tumour of the mandible.
We examined two F2 generation homozygous TP53 LSL-R167H/LSL-R167H knockout pigs (animal ID: 242, 336). These both grew more slowly than wild-type and heterozygous siblings, weighing 85-90 kg at 8 months while normal weight for this age and breed is 120-130 kg. Other than reduced size, neither showed ill effects or signs of distress in early life, but at 8 months pig 242 quickly lost condition and exhibited laboured breathing, and pig 336 developed paralysis of the hind legs. Both were slaughtered. Necropsy of pig 242 revealed  Figure 5b shows gross morphology and histology of the skull and left femur. Necropsy of pig 336 revealed multifocal osteoblastic osteosarcoma in the vertebral column, right femur and right ulna. The eighth thoracic vertebra was fractured, probably causing the limb paralysis.
Over the past 6 years, similar necropsies have been performed on 60 wild-type pigs aged between 12 and 48 months at the Bavarian Animal Health Service, none of which displayed any tumours. Moreover, there have been very few reports of spontaneous osteosarcomas in wild-type pigs. 31,32 Radiation resistance Several mutant p53 isoforms confer resistance to radiation. 33 p53deficient MSCs or MSCs expressing mutant p53 show increased resistance to 137 Cs irradiation (Figure 6a). Similarly, cells derived from an osteosarcoma on the left tibia of homozygous TP53 knockout pig 242 showed increased radioresistance compared with control cells derived from a tumour-free bone of the same animal ( Figure 6b). Tumour-derived and normal cell isolates were characterised as positive for osteocalcin, osteonectin, type 1 collagen, alkaline phosphatase and vimentin and negative for EpCam.

Multinucleation and CIN
TP53 mutations are associated with CIN and human osteosarcomas have genome-wide DNA instability. 34 Osteosarcoma cells isolated from TP53 heterozygous pig 47 and TP53 homozygous knockout pig 242 were checked for mitotic abnormalities, as were the p53-deficient MSC clones (MSC-P, MSC-PK), cell clones with activated TP53-R167H mutation (MSC-PKC, MSC-PKCM) and tumour-derived poSARCO cells. Multinucleated cells with fragmented nuclei were common in all samples ( Supplementary  Figures S3a and b). Many osteosarcoma-derived cells showed nuclear abnormalities, such as giant nuclei, multinucleated cells, micronuclei, lagging chromosomes (anaphase bridges) and abnormal spindle apparatus ( Supplementary Figures S3b and c).

DISCUSSION
In vitro studies of sarcomagenesis with human and mouse MSCs revealed that murine MSCs can be transformed by abrogation of p53 function alone. 35 Human MSCs are more refractory, requiring disruption of both the Rb and p53 pathways to bypass senescence in combination with ectopic expression of TERT, oncogenic HRAS V12 and SV40 small T antigen. 13 We report that porcine MSCs resemble human MSCs, requiring perturbation of p53, KRAS and MYC signalling pathways with spontaneous Rb pathway inactivation and telomerase-independent immortalisation steps to convert to a fully transformed phenotype. Immortalisation via the ALT mechanism rather than telomerase reactivation accords with findings from human mesenchymal malignancies. 26 There have been two other reports of neoplastic transformation of porcine primary cells in vitro, based on overexpression of randomly integrated oncogenic transgenes, 36,37 but these are generally expressed at non-physiological levels. Our findings are based on targeted mutation of endogenous genes, more closely mimicking natural lesions that underlie cancer.
We demonstrate that null mutation of TP53 results in spontaneous osteosarcomas in both heterozygous (420 months) and homozygous (7-8 months) form. Rapid disease onset is an important practical advantage for future use of this model and compares well with mice, where homozygous Trp53 knockout animals develop neoplasms around 6 months of age. 38 The natural lifespan of a pig is approximately 10-15 years.
The effects of osteosarcoma and the necessary surgery can be devastating. There is an urgent need for animal models to Patients with Li-Fraumeni syndrome and hereditary retinoblastoma are predisposed to develop osteosarcomas. 41,42 Disruption of p53 often leads to genomic instability, defective nocodazoleinduced mitotic spindle checkpoints and resistance to radiation, all of which we observed. We showed that cells derived from porcine osteosarcomas are resistant to radiation, consistent with radioresistance of human osteosarcomas. Similar to human osteosarcoma cell lines and tumour samples from patients, porcine osteosarcoma cells displayed nuclear abnormalities and atypical mitotic figures (anaphase bridges, abnormal spindle apparatus). Moreover, metaphase spreads of porcine osteosarcoma-derived cells displayed highly variable chromosome numbers, again similar to human data. 43 As up to 50% of human osteosarcomas have alterations in the TP53 gene 44 and various cytogenetic abnormalities, 45 we consider that our porcine osteosarcoma model provides a valuable platform for studying such genetic changes. Genetically modified mice have so far been the natural focus for modelling osteosarcoma. Germline Trp53 inactivation in mice results in diverse cancers, with~25% osteosarcomas in heterozygotes and~4% osteosarcomas in homozygotes, which mainly develop lymphomas. 46 The high incidence of other tumours is a problem that has motivated development of improved mouse osteosarcoma models with Cre-mediated conditional deletion of Trp53 in the osteogenic lineage, sometimes in combination with Rb1. These models show highly penetrant osteosarcoma formation but have been criticised because murine primary tumours predominantly affect the axial skeleton, while human osteosarcomas are most common in the long bones of the limbs. 47 Work in rats is less advanced than in mice, but a similar mixed tumour spectrum has been reported, with approximately half of heterozygous Trp53 knockout rats developing osteosarcomas, while most homozygotes develop haemangiosarcoma. 48 Yucatan pigs have been described that carry the TP53 R167H mutation. 7 Pigs heterozygous for this mutant allele were reported to be free of tumours up to 30 months, while homozygous pigs mainly developed lymphomas and some osteogenic tumours. These findings differ from ours, perhaps owing to the different breed, the p53 mutation or both.
Many different p53 mutations have been analysed in humans and mice. Human TP53 R175H (orthologous to porcine TP53 R167H ) is thought to impart gain-of-function properties with dominantnegative effect. Mutant p53-R175H inhibits wild-type p53 interaction with promoter elements, 49 advances angiogenesis, 50 and promotes epithelial-mesenchymal transition. 51 Genotype-phenotype analysis of Li-Fraumeni families revealed that patients carrying such a mutation in the core p53 DNA-binding domain show more highly penetrant cancer phenotypes with higher incidence and earlier onset than those with TP53 inactivation mutations. 52,53 These observations are also reflected in mouse studies. Mice heterozygous for Trp53 R172H (orthologous to human TP53 R175H and porcine TP53 R167H ) developed more tumours and a different tumour spectrum than heterozygous knockout mice. 54 We previously showed that p53-deficient porcine cells develop a transformed phenotype with more rapid cell doubling, growth in semi-solid medium and resistance to the chemotherapeutic drug doxorubicin and that these characteristics were more pronounced in porcine cells that express mutant p53-R167H, 8 suggesting that porcine p53-R167H is a gain-of-function mutation, as per murine and human data. An important advantage of our model is that the TP53 knockout allele can be activated to express mutant p53-R167H, so it will, in future, be possible to generate animals that express p53-R167H for direct comparison with p53 inactivation in the same breed.
To the best of our knowledge, ours is the first report of TP53 knockout pigs. The anatomical locations of osteosarcomas observed are mainly the long bones, skull and mandible. Analysis of homozygous p53 deficiency is so far restricted to two animals but indicates accelerated tumour development and, unlike rodents, no change in tumour spectrum relative to heterozygous knockout.
Our priority is now to further investigate how accurately the model represents the human disease. The initiation and progression of human osteosarcoma are not well understood and we hope the porcine model will help elucidate the molecular pathways and driver mutations involved. Imaging studies will enable non-invasive longitudinal investigation of individual animals to identify early-stage tumours, analyse disease progression and investigate metastasis. Several practical advantages of the model are already apparent, not least its simplicity, with no need for tissue-specific gene inactivation or multiple engineered mutations. It is also helpful that osteosarcomas form relatively late in heterozygotes, allowing normal breeding and expansion of the herd, and with short latency in homozygotes that can be produced as required. Nine TP53 LSL-R167H/+ and two TP5 LSL-R167H/LSL-R167H Landrace pigs aged 7.5-32 months, male and female, were generated and raised in our own facilities with food and water provided ad libitum.

MATERIALS AND METHODS Reagents
Two 7-and 10-week-old female NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice were purchased from Jackson Lab (Bar Harbor, ME, USA) and maintained in specific pathogen-free facilities with food and water provided ad libitum.

Necropsy examination and tumour analysis
Pigs were humanely killed and complete necropsy examination was carried out at the Tiergesundheitsdienst Bayern (Bavarian Animal Health Service). Specimens were fixed, embedded, sectioned and stained by standard methods. Bone specimens were first decalcified in Ossa Fixona (Waldeck GmbH, Münster, Germany).

Porcine primary cells and cell lines
Porcine primary cells and cell lines were derived in house and regularly screened for mycoplasma contamination. Cell lines of human origin (Hek293, HCT1116, SW480,CaCo2) were obtained from ATCC or DSMZ (Braunschweig, Germany), cultured in Dulbecco's Modified Eagle's Medium (DMEM) with foetal calf serum (7%), 1% Pen/Strep and 1% glutamine and tested every 2 months for mycoplasma infection. To avoid contamination and phenotype changes, cells were kept as frozen stocks and cultured for 4 weeks maximum.
Generation of TP53 R167H , KRAS G12D double gene targeted porcine MSC clones Derivation and culture of porcine TP53 gene-targeted MSC clones has been described previously. 8 The vector KRAS-BSR (Supplementary Figure S1a) is the same as vector KRAS-NEO, which has been described previously, 9 but contains the blasticidin resistance gene (bsr) instead of neo.
KRAS-BSR targeted cell clones were identified by: 5′ and 3′ junction PCR, RT-PCR, and restriction fragment-length polymorphism to confirm the G to A mutation, as described elsewhere. 9 Cre-mediated induction of TP53 R167H and KRAS G12D alleles was confirmed by PCR and RT-PCR analysis as described elsewhere. 8,9 All PCR primers used and diagnostic fragments amplified are shown in Supplementary Table S2 Cre recombinase protein transduction Cre protein was produced in vitro with the vector pTriEx-HTNC (Addgene plasmid 13763, Cambridge, MA, USA) as described 55,56 and transduced as described previously. 8 After 96 h, cells were plated by limiting dilution into 96-well plates to derive single-cell clones.

Bisulphite sequencing
In all, 500 ng genomic DNA samples were converted using the EpiTect Fast Bisulphite Conversion Kit (Qiagen, Germantown, MD, USA) by standard methods. A 266-bp P16INK4A fragment (NCBI accession number AJ316064.1: 299…565) was amplified using primers BS-p16F and BS-p16R with GoTaq Polymerase (Promega, Madison, WI, USA). Thermal cycling conditions were: 30 min, 95°C; then 35 cycles of: 30 s, 95°C; 30 s, 54°C; 30 s, 72°C; followed by 5 min, 72°C. Amplified PCR products were processed as described previously. 57 Copy number variation analysis Q-PCR was carried out using Fast SybrGreen PCR MasterMix and run on 7500 Fast Real-Time PCR System (both from Applied Biosystems, Foster City, CA, USA). Relative quantification of KRAS copy number was performed with primers CNV-KRASF and CNV-KRASR. GAPDH (glyceraldehyde 3phosphate dehydrogenase) was amplified using primers CNV-GAPDHF and CNV-GAPDHR. Relative quantification of copy number alterations was performed in 10 μl with 10 ng genomic DNA. Thermal cycling conditions were 10 min, 95°C; then 40 cycles of: 15 s, 95°C; 1 min, 60°C. Samples were run in triplicate, data were normalised to GAPDH and copy numbers were calculated by the ΔΔC T method.
Quantitative real-time RT-PCR Relative quantification of gene expression was carried out by two-step Q-RT-PCR using Fast SybrGreen PCR MasterMix. Primers used were: p16INK4α RT-p16F and RT-p16R; p14ARF RT-p14F and RT-p14R; MYC RT-MycF and RT-MycR; and GAPDH RT-GAPDHF and RT-GAPDHR. Reactions were performed in 10 μl. Thermal cycling parameters were as above. Samples were assayed in triplicate, relative expression was normalised to GAPDH expression and fold-differences were calculated by the ΔΔC T method.

Western blotting analysis
Protein lysates were prepared as described previously. 8 Total protein (40 μg) was loaded in each lane and separated by 15% sodium dodecyl sulphate-polyacrylamide gel electrophoresis. Blotting and detection of p53 and GAPDH was as described previously. 58 Porcine MYC was detected with anti-c-MYC (N-262, Santa Cruz Biotechnology, Heidelberg, Germany) diluted 1:250.
10 μg Raf-1 RBD agarose for 3 h at 4°C with gentle agitation. Western blotting and detection were carried out according to the manufacturer's instructions.

Growth characteristics of porcine cells
Cycling and non-cycling cells in the G 0 /G 1 peak were differentiated by 2-h pulse-labelling with 5-ethynyl-2'-deoxyuridine and then staining with the Click-iT EdU Flow Cytometry Assay Kit (ThermoFisher). For the focus formation assay, 1 × 10 5 cells were plated onto 10-cm cell culture dishes and cultured for 4 weeks, and the medium was changed every 2-3 days. Anchorage-independent growth assay of primary and transformed porcine cells was carried out as described. 8 Tumour growth and isolation of porcine sarcoma-derived tumour cells Groups of 1 × 10 7 cells suspended in chilled DMEM were mixed 1:1 with high concentration matrigel basement membrane matrix (BD Biosciences, Bedford, MA, USA) and implanted subcutaneously into NSG mice. Mice were humanely killed, tumours excised, fixed, sectioned and H&E (haemotoxylin and eosin) stained by standard methods. Cells were derived from a minced piece of tumour digested in DMEM supplemented with 200 U/ml collagenase type IV (Worthington, Lakewood, NJ, USA) at 37°C for 24 h, then centrifuged at 100 r.c.f. for 5 min. Cells were resuspended in standard porcine MSC medium, 8 cultured at 37°C, 5% CO 2 and passaged at regular intervals.

Establishment of cell cultures from porcine osteosarcoma
Cell cultures were established from osteosarcomas as described above for tumour xenografts, cultured and passaged by standard methods.
Gene expression profiling and data analysis cDNA was prepared from total RNA by standard methods. Gene expression experiments were conducted on custom Sus Scrofa 12-plex microarrays containing 142 073 probes targeting 17 261 genes. Probe selection and array design is described elsewhere. 59 In all, 200 ng double-stranded cDNA was Cy3-labelled using the 'One colour DNA Labelling Kit' (Roche NimbleGen, Madison, WI, USA). A minimum of three technical replicates of each sample were loaded, and raw data intensities were collected using a NimbleGen MS200 scanner. Background corrected, quantile normalised and RMA (Robust Multi-Array Analysis) normalised probe intensities were generated using the NimbleScan v2.6 software (Roche NimbleGen). Average signal intensities were calculated for each probe set and log2 transformed. Data were analysed using the Partek Genomics Suite software, v6.6 (Partek, St Louis, MO, USA). Significant differentially expressed genes were identified for a false-discovery rate P-value of 0.05 and twofold change.

Cell cycle analysis
Mitotic arrest was induced with 100 ng/ml nocadazole for 20 h. Cells were harvested, washed, resuspended in ice-cold Ca 2+ /Mg 2+ -free phosphatebuffered saline with 1 mg/ml glucose, fixed in 70% ice-cold ethanol overnight at 4°C and incubated 1 h at room temperature in the dark with 50 μg/ml propidium iodide and 0.4 mg/ml RNaseA (Qiagen). Cells were washed, resuspended in phosphate-buffered saline with 0,5% (v/v) bovine serum albumin and 0.01% (w/v) NaN 3 and strained before fluorescenceactivated cell sorting analysis. Cell cycle profiles were measured using a FACSCalibur device (BD Biosciences, San Jose, CA, USA) with the CellQuest Pro software (BD Biosciences). Doublets, dead cells and debris were excluded, and single-parameter DNA histograms were analysed with the FlowJo software (Tree Star, Ashland, OR, USA) to identify G 1 and G 2 /M peaks and the S-phase populations from univariate distribution curves.
Cell irradiation clonogenic assay Aliquots of 5 × 10 5 cells were irradiated with 2, 6 and 10 Gy from a 137 Cs source (GE Healthcare Buchler, Braunschweig, Germany), seeded in six-well plates and cultured under standard conditions overnight. A total of 1 × 10 2 cells were seeded on 10-cm dishes and cultivated under standard conditions for 14 days. Plates were fixed and stained with 0.1% crystal violet by standard methods, and colonies were counted by two observers.