Workflow diagram giving overview of warm autopsy.

The four steps of the study are shown on the left. The right shows in detail the organization of the warm autopsy committee, as well as the operating table layout in implementation.

Clinical course and serial procurement of the patient.

(A) The clinical course of disease progression and treatment in this index case. The bottom figure shows the pathology of biopsies (200x magnification). (B) The pathology of metastatic sites procured by warm autopsy (200x magnification). All the hematoxylin and eosin (H&E) stains were performed using standard techniques. (C) Tissue and blood samples taken at different times. Two lymph node metastases samples taken on day 210 were not sequenced. R, right; L, left; LN, lymph node. (D) The type of sequencing performed. The abbreviations of samples are indicated. FF, fresh frozen; FFPE, formalin-fixed paraffin-embedding.

Intratumoral genetic heterogeneity of prostate primary tumors.

(A) The regional distribution of nonsynonymous mutations in primary tumors. The heat map indicates the presence of a mutation (purple) or its absence (white) in the individual tumor. Right showed the gene names of driver mutations. The TP53 mutation was not detected in the PB2 sample using WGS due to insufficient total depth, and the number in square indicate the alt_reads/total_reads/allele_fraction. (B) The figure shows the allele fraction distributions plotted by mutation number (left vertical axis) and density (right vertical axis). Tumor purity provided by FACETS in three samples is also indicated. (C) Copy number profile of chr17 and the LOH of TP53 and CDK12. Shown from upper to lower are the total copy number log-ratio (the log ratio of total read depth in the tumor versus that in the normal), allele-specific log-odds-ratio (the log odds ratio of the variant allele count in the tumor versus in the normal), and corresponding integer (total, minor) copy number calls provided by FACETS. (D) The figure demonstrates how somatic mutations accumulate in a CN-LOH (TP53) and Loss (CDK12) chromosome. (E) Cancer cell fractions and clusters of mutations inferred by PyClone for pairs of samples. Blue density areas reveal the mutation clusters present at clonal or subclonal levels, and the manually colored circles provide the localization of mutation clusters in different samples. Driver mutation genes present in the cluster are marked in red. (F) The clonal evolution tree of the primary tumor. The length of branches connecting clones is proportional to the number of mutations contained, and the driver events identified are marked on the tree.

Intratumoral genetic heterogeneity and clonal evolution of prostate metastatic tumors.

(A) An overview of somatic alterations detected in 11 tumors. Each panel displays the number of mutations in coding region, nonsynonymous mutations, the number of segments for copy number alterations, and the tumor purity, respectively. (B) Overview of the analyzed driver genomic alterations in the primary tumor and metastases. (C) The clonal evolution tree of the primary tumor and metastases inferred by ClonEvol. Except for cluster 12 private to LV2M, which is manually added, all the CCF clusters were calculated by PyClone. The branch length is scaled by the log2 ratio of the number of mutations in the individual clone. The potential driver events are highlighted. (D) The emergence and movement of clones in the spread of metastasis. The color-coded arrows depict the seeding events and the acquisition of mutations, and the sequence of events is ordered according to the clonal evolution relationship. Plus (+), the acquisition of subclone.

CDKN1B alterations in prostate cancer metastasis.

(A) The position distribution of CDKN1B somatic mutations in this patient and in the deduplicated samples from cbioprotal. The circles are colored with respect to the different mutation types, and the size represents the number of patients with the mutation. (B) Scratch assay of 22RV1 cells transfected with shRNA (n = 3), which corresponds to Supplementary Fig 7D. (C, D) Images of migrating and invading cells tested by using Transwell assays for 22RV1 cells transfected with shRNA (n = 3). P values were determined by two-tailed Student’s t test.

Intratumor heterogeneity of DNA methylation and epigenomic evolution in prostate metastatic tumors.

(A) The figure shows the intratumoral heterogeneity of methylation patterns depicted by unsupervised hierarchical clustering using the top 1% of CpG sites (n = 150,000) with the greatest difference. Normal samples were excluded for the hierarchical clustering but were input into row clustering. (B) DNA methylation Intratumor heterogeneity on genomic regions. The median methylation variability on the right of the figure was calculated by the range of CpG sites (maximum level - minimal level) between tumors. (C) Epigenomic clonal evolution tree inferred from DNA methylation distance matrices. Lengths of trunks and branches were inferred using the top 1% of CpG sites (same as Fig 5A, see Supplemental Methods). Color coding is the same as in Fig 4F. (D) Genomic clonal evolution tree inferred from CCF (cancer cell fraction, left) and VAF (variant allele fraction, right) distance matrices. (E) The correlation between epigenomic distance matrices and genomic distance matrices (CCF, left; VAF, right). LOESS fitted curve and 95% confident interval are presented. Rho coefficient (R) and P value (P) are assessed by Spearman’s rank correlation. (F) The difference in methylation levels of CpG island in promoter region of known prostate cancer driver genes (www.genome.jp/pathway/ko05215) between each tumor and three normal prostate samples. Hyper-methylation and hypo-methylation were defined as difference of more than 20%. White cells in the heatmap represent differences below 20%.