Evaluating cell lines as tumour models by comparison of genomic profiles

Abstract

Cancer cell lines are frequently used as in vitro tumour models. Recent molecular profiles of hundreds of cell lines from The Cancer Cell Line Encyclopedia and thousands of tumour samples from the Cancer Genome Atlas now allow a systematic genomic comparison of cell lines and tumours. Here we analyse a panel of 47 ovarian cancer cell lines and identify those that have the highest genetic similarity to ovarian tumours. Our comparison of copy-number changes, mutations and mRNA expression profiles reveals pronounced differences in molecular profiles between commonly used ovarian cancer cell lines and high-grade serous ovarian cancer tumour samples. We identify several rarely used cell lines that more closely resemble cognate tumour profiles than commonly used cell lines, and we propose these lines as the most suitable models of ovarian cancer. Our results indicate that the gap between cell lines and tumours can be bridged by genomically informed choices of cell line models for all tumour types.

Figure 1. Genomic comparison of TCGA HGSOC samples with CCLE ovarian cancer cell lines suggests overall genomic similarity.

(a) CNA profiles (right, chromosomes 1–22) and mutations (left, in eight selected genes) of HGSOC patient samples from TCGA, top and ovarian cancer cell lines from the CCLE, bottom. The samples are sorted according to decreasing fraction of the genome altered in DNA copy number. Somatic mutations in genes known to be commonly altered in one of the four epithelial ovarian cancer subtypes are indicated on the left, with germline mutations included for BRCA1 and BRCA2 in the tumour samples in addition to the somatic mutations. Note the samples with a low degree of CNA coinciding with wild-type TP53 copies near the bottom of each panel (square bracket). (b) The most frequent genomic alterations identified in HGSOC tumour samples and their occurrence in the ovarian cancer cell line panel: CNAs (left) and mutations (right).

Figure 2. Hypermutated cell lines are outliers.

The comparison of mutation frequency (horizontal) and degree of CNA (vertical) for HGSOC tumour samples (blue) and ovarian cancer cell lines (red) reveals a subset of cell lines (dashed ellipse) with a hypermutator genotype (high mutation frequency, few DNA copy-number changes). The hypermutated cell lines (mutation frequency in parentheses) are: IGROV1 (20.7/Mb), OC316 (19.0/Mb), EFO27 (16.1/Mb), OVK18 (14.4/Mb) and TOV21G (13.4/Mb). Cell lines that on the contrary resemble the tumour samples in key characteristics (as shown below in Fig. 3) are also labelled.

Figure 3. Ranking ovarian cancer cell lines by suitability as HGSOC models.

Both average properties (left) and selected genetic events specific to ovarian cancer (right) can be used to distinguish better and poorer models of HGSOC. Average properties include the histological subtype as determined in the original publication (references in Supplementary Data S1), the citation frequency in the literature as estimator of frequency of use in laboratories, the altered fraction of the genome, the number of mutations per million bases and the correlation with the mean CNA profile of HGSOC tumour samples. The selected genetic events include alterations recurrently found either in HGSOC (mutation of TP53, BRCA1 or BRCA2; amplification of C11orf30 (EMSY), CCNE1, KRAS or MYC; mutation or deletion of RB1) or one of the three other major subtypes of ovarian cancer (mutation in PIK3CA, PTEN, KRAS, BRAF, CTNNB1 or ARID1A; mutation or amplification in ERBB2). The colour gradient underlying the cell line names to the left indicates better (green) versus poorer (red) models of HGSOC according to selected characteristics (TP53 status, correlation with mean CNA profile of TCGA samples, low mutation rate and absence of mutations in the seven ‘non-HGSOC’ genes, see Supplementary Data S1). The hypermutated cell lines described in Fig. 2 are located at the bottom of the table. Note that although HGSOC cell lines are probably at the top and unsuitable cell lines are at the bottom of the table (vertical labels), the order does not signify an exact ranking of cell line models.

Figure 4. Expression-based clustering of all 963 CCLE cell lines from diverse tumour types.

The 5,000 most variable genes were used for unsupervised clustering of cell lines by mRNA expression data. Cell lines are colour-coded (vertical bars) according to the reported tissue of origin (a PDF version that can be enlarged at high resolution is in Supplementary Information, Supplementary Fig. S4); horizontal labels at bottom indicate the dominating tissue types within the respective branches of the dendrogram. Most ovarian cancer cell lines (magenta) cluster together, interspersed with endometrial cell lines. However, some ovarian cancer cell lines cluster with other tissue types (*). Top right panels: neighbourhoods (1) of the top cell lines in our analysis, (2) of cell line IGROV1, and (3) of cell line A2780. For the ovarian cancer cell lines in these enlarged areas, the histological subtype as assigned in the original publication is indicated by coloured letters.