The genomic and epigenomic evolutionary history of papillary renal cell carcinomas

Abstract

Intratumor heterogeneity (ITH) and tumor evolution have been well described for clear cell renal cell carcinomas (ccRCC), but they are less studied for other kidney cancer subtypes. Here we investigate ITH and clonal evolution of papillary renal cell carcinoma (pRCC) and rarer kidney cancer subtypes, integrating whole-genome sequencing and DNA methylation data. In 29 tumors, up to 10 samples from the center to the periphery of each tumor, and metastatic samples in 2 cases, enable phylogenetic analysis of spatial features of clonal expansion, which shows congruent patterns of genomic and epigenomic evolution. In contrast to previous studies of ccRCC, in pRCC, driver gene mutations and most arm-level somatic copy number alterations (SCNAs) are clonal. These findings suggest that a single biopsy would be sufficient to identify the important genetic drivers and that targeting large-scale SCNAs may improve pRCC treatment, which is currently poor. While type 1 pRCC displays near absence of structural variants (SVs), the more aggressive type 2 pRCC and the rarer subtypes have numerous SVs, which should be pursued for prognostic significance.

Fig. 1. Study design and genomic landscape.

a A schematic illustration of the dissection of multiple tumor samples from the center of the tumor towards the tumor’s periphery, plus metastatic samples in the adrenal gland as well as normal samples. For the analysis, the normal sample more distant from the tumor and with absence of tumor nuclei was chosen as reference. b Summary of subjects and samples that underwent different analyses based on DNA availability: whole-genome sequencing (124 samples from 29 subjects), deep targeted sequencing of cancer driver genes (139 samples from 38 subjects), genome-wide methylation (139 samples from 28 subjects) or SNP array profiling (only tumor samples, 101 samples from 38 subjects). c Tumor genomic alterations across histological subtypes. Shown are genome level changes, such as mutational burden, numbers of structural variants (SV) and retrotransposition events (TE), as well as other genomic alterations (denoted by different colors).

Fig. 2. Phylogenetic trees and oval plots for tumors with three or more samples.

Phylogenetic trees: the trees show the evolutionary relationships between subclones (annotated by different colors). Trunk and branch lengths are proportional to the number of substitutions in each clone cluster. Driver SNV and recurrent somatic copy number alterations are annotated on the trees. Tumor regions containing sample-specific subclones are indicated on the tree leaves. Oval plots: In the top rows the ovals are ordered based on the physical sampling of the tumor regions. Ovals are nested if required by the pigeonhole principle. The first row of the plot with nested ovals is linked by lines to the ovals ordered by the phylogenetic analysis, indicating intermixing of subclones spread across 2 or more tumor regions. In the matrix, each main clone (without solid border) and subclone (with solid border) is represented as a color-coded oval. The size of the ovals is proportional to the CCF of the corresponding subclones. Each column represents a sample. Oval plots are separated into three parts: trunk (top, CCF = 1 in all samples), branch (middle present in >1 sample but not with CCF = 1 in all samples), and leaf (bottom, specific to a single sample). GL germline, amp amplification, DLOH hemizygous deletion loss of heterozygosity, HET diploid heterozygous, NLOH copy neutral loss of heterozygosity, HOMD homozygous deletion, ASCNA allele-specific copy number amplification, BCNA balanced copy number amplification.

Fig. 3. Somatic copy number alterations (SCNAs).

a Genome-wide sample cancer cell fraction (CCF) profiles across tumors. Samples are labeled by histology subgroup and cellularity (or called purity). (b) Size of SCNAs identified in each tumor, separated into clonal alterations (purple), which have a shared breakpoint across all samples from a tumor and subclonal alterations (yellow), which have breakpoints that are found only in a subset of samples from a tumor. Where clonal and subclonal SCNAs are both identified in a tumor, they are shown side by side. The numbers of clonal and subclonal SCNAs are included in the parentheses following the tumor ID, such as pRCC1_1472_01(n = 7,0). Vertical lines separate pRCC1 (left), pRCC2 (middle) and rare subtypes (right). In each box-and-whisker plot, the line dividing the box represents the median; the ends of the box are the lower (Q1) and upper (Q3) quartiles; the whiskers are extended to Q1-1.5xIQR and Q3 + 1.5IQR with IQR = Q3–Q1. Each circle represents a data point of SCNA size. (c) SCNAs of tumor pRCC2_1824_13. Top panel: genome-wide SCNAs on ten primary tumors (T01-T10) and three metastatic samples (M01, M02 and M03); T10 has low purity and has no SCNAs. Bottom panels: metastatic sample-specific SCNAs on chromosome 4 for total copy number log-ratio (red line: estimated total copy number log-ratio; green line: median; purple line: diploid state). DLOH: hemizygous deletion loss of heterozygosity; HET: diploid heterozygous; NLOH: copy neutral loss of heterozygosity; ALOH: amplified loss of heterozygosity; ASCNA: allele-specific copy number amplification; BCNA: balanced copy number amplification.

Fig. 4. Structural variants (SV) and retrotransposition events (TE).

a Frequency of SV events and TE insertions for each sample. b–d Circos plots for SV events for three tumors; involved driver genes are noted. e The distribution of mean cancer cell fraction (CCF) of SVs across tumors. Alu elements originally characterized by the action of the Arthrobacter luteus (Alu) restriction endonuclease, ERVK mouse endogenous retrovirus K, L1 Long interspersed element-1.

Fig. 5. Intratumor heterogeneity (ITH) of methylation profiles.

a Scatter plots of pairwise distance between methylation and single nucleotide variant clusters. LOESS (Locally Weighted Scatterplot Smoothing) fitted curve is shown in red line with 95% confident interval in gray shaded area. Spearman’s rank correlation rho is shown on the bottom right. For the two-sided test of rho = 0, the test statistic S is 719790 based on the algorithm AS 89. The exact P value is 8.66 × 10⁻¹⁵. b Methylation ITH on genomic regions for each sample and tumor subtype. c Unsupervised hierarchical clustering of methylation profiles measured by the methylation level as beta value for the top 1% most variable methylation probes. Sample IDs are followed by the purity estimated by SCNAs or SNV VAF in parentheses. The background colors of the sample IDs represent different histological subtypes and tumor or normal tissue samples.