Stratification of clear cell renal cell carcinoma (ccRCC) genomes by gene-directed copy number alteration (CNA) analysis

Abstract

Tumorigenic processes are understood to be driven by epi-/genetic and genomic alterations from single point mutations to chromosomal alterations such as insertions and deletions of nucleotides up to gains and losses of large chromosomal fragments including products of chromosomal rearrangements e.g. fusion genes and proteins. Overall comparisons of copy number alterations (CNAs) presented in 48 clear cell renal cell carcinoma (ccRCC) genomes resulted in ratios of gene losses versus gene gains between 26 ccRCC Fuhrman malignancy grades G1 (ratio 1.25) and 20 G3 (ratio 0.58). Gene losses and gains of 15762 CNA genes were mapped to 795 chromosomal cytoband loci including 280 KEGG pathways. CNAs were classified according to their contribution to Fuhrman tumour gradings G1 and G3. Gene gains and losses turned out to be highly structured processes in ccRCC genomes enabling the subclassification and stratification of ccRCC tumours in a genome-wide manner. CNAs of ccRCC seem to start with common tumour related gene losses flanked by CNAs specifying Fuhrman grade G1 losses and CNA gains favouring grade G3 tumours. The appearance of recurrent CNA signatures implies the presence of causal mechanisms most likely implicated in the pathogenesis and disease-outcome of ccRCC tumours distinguishing lower from higher malignant tumours. The diagnostic quality of initial 201 genes (108 genes supporting G1 and 93 genes G3 phenotypes) has been successfully validated on published Swiss data (GSE19949) leading to a restricted CNA gene set of 171 CNA genes of which 85 genes favour Fuhrman grade G1 and 86 genes Fuhrman grade G3. Regarding these gene sets overall survival decreased with the number of G3 related gene losses plus G3 related gene gains. CNA gene sets presented define an entry to a gene-directed and pathway-related functional understanding of ongoing copy number alterations within and between individual ccRCC tumours leading to CNA genes of prognostic and predictive value.

Fig 1. Workflow “analysis of copy number alterations (CNAs)”.

Three different workflow branches (WB I–III) conducting CNA analysis have been applied leading to CNA gene sets that were validated by published and external data.

Fig 2. Gains and losses of CNA genes ordered at 4p16.3.

Cytoband-plots of 4p16.3 show CNAs of selected HRO tumours. Corresponding data are available at Table F in S8 File. The height of bars documents number of tumours that share CNAs. Assignments of colourings are dark = G3 nominator genes, medium gray = G1 nominator genes, light gray = genes without any preference for Fuhrman malignancy grades.

Fig 3. Gains and losses of CNA genes ordered at 8p23.1.

Cytoband-plots of 8p23.1 show CNAs of selected HRO tumours, corresponding data at Table G in S8 File. Assignments of colourings are dark = G3 nominator genes, medium gray = G1 nominator genes, light gray = genes without any preference for Fuhrman malignancy grades.

Fig 4. CNA genes in HRO tumour G3_541 at cytobands 3p14.3, 6q21 and 7q22.1.

Cytoband-plots of HRO tumour G3_541 depict three cytobands affecting almost all genes (bars are present), data at Tables H-J in S8 File. Assignments of colourings are dark = G3 nominator genes, medium gray = G1 nominator genes, light gray = genes without any preference for Fuhrman malignancy grades.

Fig 5. Correlation of cytobands based on gene losses.

Unsupervised hierarchical clustering by average linkage and Euclidian distance, representing a correlation-matrix displaying Pearson correlation of gene losses between cytobands specified by p-values below 10⁻¹⁴ (Table C in S5 File).

Fig 6. Correlation of ccRCC tumours based on gene gains in cytobands.

Unsupervised hierarchical clustering by average linkage and Euclidian distance, representing a correlation-matrix displaying Pearson correlation between HRO tumours based on CNA gains per cytoband specified by p-values below 10⁻¹⁴ (Table C in S5 File).

Fig 7. CNAs of all HRO tumours assigned to PI3K pathway.

Decomposition of the KEGG Pathway PI3K shows a colour coded map of genes. The shapes are:

1. Red = Loss

2. Blue = Gain

3. Yellow = Both states (Gain/Loss)

4. Rectangle = Uncoloured, Gene is not affected

5. Rectangle = Grade (1/3) equally affected, colour shows aberration

6. Circle = Grade 3, colour shows aberration

7. Diamond = Grade 1, colour shows aberration

Corresponding data sets are enlisted at Table G in S2 File.

Fig 8. Heatmap of CNA gene sets derived from 4 HRO G3 tumours.

Unsupervised hierarchical clustering by average linkage and Euclidian distance of all HRO tumours based on a list of CNA genes with a p-value below 0.1 were shared by 4 G3-patients which are similar to most of the G1-patients (determined by unsupervised hierarchical clustering of the corresponding correlation-matrix, Table A in S13 File). Data are presented at Tables F and G in S13 File.

Fig 9. Patient group assignments of HRO and Swiss data sets.

CNA based patient group assignments include Fuhrman grades and survival outcome. Classifications are highlighted (extended data under S16 File).

Fig 10. CNA-heatmap of HRO and Swiss tumours (gene set HRO201).

A: Hierarchical clustering of CNA genes (gene set HRO201) regarding HRO is shown. In total, 8 gene subgroups are obtained by unsupervised hierarchical clustering by average linkage and Euclidian distance, data at Table A and B in S14 File. Corresponding patient groups as well as Fuhrman gradings are shown, see Fig 9 as well. B: Swiss data set (GSE19949) was analysed as in Fig 10A based on Gene set HRO201. Data at Table A and B in S16 File. Concerning HRO and Swiss tumours, corresponding patient groups as well as Fuhrman gradings are shown, see Fig 9 as well.

Fig 11. Kaplan-Meier survival analysis of HRO cohort patients.

A: visualizes survival data of HRO patients stratified by three CNA patient groups based on gene set HRO201 leading to classifications as assessed in Table A in S16 File. The hazard ratio (HR) of the HRO cohort concerning both two groups (A, Z vs. B-F) is 21.76 (95% confidence interval is [2.58, 183.2]). The logrank test between both two groups yields a p-value of 5.83·10⁻⁵. B: is based on survival data categorized by Fuhrman grades G1, G2 and G3, number of patients in parenthesis (data: Tabla A in S16 File). Comparing Fuhrman grades G1 and G3 using Cox proportional hazards regression model leads to a hazard ratio close to infinity since all G1 and G2 patients are survivors.