Prostate cancer (PC) remains one of the major causes of death in men today. With a better understanding of the disease pathogenicity at the molecular level, especially coupled with the advancement made in personalised medicine regimes, the overall survival and quality of patients' lives post-treatment has largely benefited. Here, using whole exome sequencing, we show that the differential tumorigenic potential of the widely used murine prostate cancer cell lines (TRAMP-C1, 2 and 3) is likely reflected by its Trp53 mutational status. Furthermore, we also unravel that the tumorigenic TRAMP-C1 and C2 cells harbours the same stop-gain variant of Trp53 compared with the non-tumorigenic TRAMP-C3 cells, which harbour a wild-type Trp53. This observation opens up the possibility to use these cell lines to model PC in a defined manner.

Prostate cancer (PC) remains one of the main prevailing causes of mortality in men globally, with an estimated detection of 220,800 new cases and at least 12% or above fatality rate^[1] being reported every year. A number of factors have been linked to the disease pathogenesis^[2]. The role of TP53 and typically its mutational status have historically been debated in PC^[3][4]. However, recently p53 levels have been shown to dictate the molecular classifications of PC^[5].

Several mouse models have been developed to study PC^[6], including the so called TRAMP mouse, which develop prostate malignancies owing to the over-expression of prostate-specific SV40 large T antigen transgene^[7]. Besides being useful for studying PC pathogenesis, a tumour from a TRAMP mouse has also been the source of the now widely used mouse prostate cancer (TRAMP-C) cell lines, which are transplantable in syngeneic C57BL/6 mice and hence can be used in drug efficacy studies^[8]. It was shown that, out of the three cell lines (TRAMP-C1, C2 and C3) developed from three different regions of the same tumour, TRAMP-C3 does not confer tumorigenic potential in vivo, a finding that remains unexplained to date^[8]. The only analysis of potential genetic causes of the tumorigenic potential of the cell lines that was performed was to analyse expression of large T antigen and p53 protein by qRT-PCR and immunohistochemistry, respectively. A high p53 expression would mean it is mutant since it would not be able to trigger its own degradation via Mdm2^[9]. Their analyses showed that neither was expressed; therefore the authors claimed that the tumour cells had silenced the large T antigen but remained wild type for p53 (hence low expression).

Intra-tumour heterogeneity plays an important role across malignancies and is most likely the cause of variations seen within a given tumour^[10]. We hypothesised that the underlying genomic variations may be the probable cause of differential tumorigenic potential seen in these cell lines. To validate this, we subjected the three TRAMP-C cell lines (TRAMP-C1, C2 and C3) to whole-exome sequencing.

To define the genetic differences between the widely used TRAMP-C1, -C2 and -C3 cell lines.

To be able to couple genetic changes to drug responses^[11], we set out to investigate the mutation spectrum of the widely used murine prostate cancer cell lines (TRAMP-C1, 2 and 3) by exome sequencing. We found ~5000 single nucleotide variation (SNV) and Insertion-deletion (InDel) mutations per cell line, around 10% of which were in regions that would alter the amino acid sequences (so-called non-synonymous or frameshift mutations; Fig. 1A and Suppl. Table 1). Comparison of the genes mutated in the different cell lines showed that 17 genes were mutated only in TRAMP-C1 and C2 cells (Fig. 1B). To investigate if any of these genes are mutated in cancer, we mined The Cancer Genome Atlas (TCGA;cbioportal.org). Not surprisingly, the most frequently mutated, gene was TP53 (Trp53 in mouse). Of the other genes, none were recurrently mutated but LRRC15 is part of the 3q26-29 region of the genome that is frequently amplified in many cancers^[12]. However, this gene is not recurrently mutated arguing against the importance of the heterozygous mutation we found here in TRAMP-C1 and C2 cells.

We find that both TRAMP-C1 and C2 cells have the same frameshift deletion in Trp53 (V271fs), whereas TRAMP-C3 cells harbour a wild type Trp53 (Fig. 1B, C and Suppl. Table 1). We therefore speculate that this mutation governs the enhanced tumorigenicity of the TRAMP-C1/C2 cell lines.

A likely underlying cause of in vivo tumorigenic potential across the TRAMP-C cell lines is Trp53 mutation status. Hence, the absence of TP53 mutations claimed in these cell lines, and possibly in human PC, may be not due to a lack of mutations but because of the fact that immunohistochemistry has been the golden standard to assess mutational status. This method calls stop-gain or deletion mutations as wild-type because of a lack of expression of ‘mutant’ protein.

To formally prove that the inactivating mutation of Trp53 is tumorigenic, additional experiments are needed, for example, Trp53 gene deletion/correction in the different cell lines in a reciprocal manner. An additional limitation is that we cannot rule out other mechanisms of tumorigenesis such as epigenetic changes, among others.

Exome Sequencing

The cell lines were purchased from ATCC (Manassas, Virginia, USA) and cultured as per their recommendation. Genomic DNA was extracted using NucleoSpin® Tissue kit (MACHEREY-NAGEL GmbH & Co. KG, Germany) as per the manufacturer’s protocols. Exome sequencing was outsourced to BGI, China. Briefly, exome capture was performed using SureSelect Target Enrichment System Capture Kit (Agilent Technologies, USA) and sequenced on an Illumina HiSeq2000 (Illumina Inc., USA) for 100PE reads.

Bioinformatics analysis

Quality assessment of the sequence reads was performed by generating QC statistics with FastQC. Low-quality reads were filtered with PRINSEQ^[13] (minimum fragment length and minimum quality mean were set to 30 and 20, respectively). Read alignment to the Mus musculus MM10 genome was done using BWA^[14] with default parameters. After removal of PCR duplicates (Picard tools) and file conversion (samtools^[15]) quality score recalibration, Indel realignment and variant calling were performed with the GATK package^[16]. Variants were annotated with ANNOVAR^[17] using SNP and gene information from the Mouse Genome Project at the Sanger Institute and the UCSC^[18].

This work was supported by grants from the Swedish Cancer Society, the Swedish Research Council, the Region Västra Götaland (Sahlgrenska University Hospital, Gothenburg), the Knut and Alice Wallenberg Foundation, the Sahlgrenska Academy and BioCARE - a National Strategic Cancer Research Program at University of Gothenburg (to JAN), and from the Adlerbertska Forskningsstiftelsen, Assar Gabrielsson Foundation, the W&M Lundgren Foundation and Sahlgrenska Universitetssjukhusets stiftelser (Sahlgrenska University Hospital, Gothenburg) (to JB).

We would like to thank Dr. Erik Larsson Lekholm for graciously providing server space.

The authors declare no conflicts of interest.

Not applicable.

No fraudulence is committed in performing these experiments or during processing of the data. We understand that in the case of fraudulence, the study can be retracted by ScienceMatters.