Tumor Predisposing Post-Zygotic Chromosomal Alterations in Bladder Cancer-Insights from Histologically Normal Urothelium

Abstract

Bladder urothelial carcinoma (BLCA) is the 10th most common cancer with a low survival rate and strong male bias. We studied the field cancerization in BLCA using multi-sample- and multi-tissue-per-patient protocol for sensitive detection of autosomal post-zygotic chromosomal alterations and loss of chromosome Y (LOY). We analysed 277 samples of histologically normal urothelium, 145 tumors and 63 blood samples from 52 males and 15 females, using the in-house adapted Mosaic Chromosomal Alterations (MoChA) pipeline. This approach allows identification of the early aberrations in urothelium from BLCA patients. Overall, 45% of patients exhibited at least one alteration in at least one normal urothelium sample. Recurrence analysis resulted in 16 hotspots composed of either gains and copy number neutral loss of heterozygosity (CN-LOH) or deletions and CN-LOH, encompassing well-known and new BLCA cancer driver genes. Conservative assessment of LOY showed 29%, 27% and 18% of LOY-cells in tumors, blood and normal urothelium, respectively. We provide a proof of principle that our approach can characterize the earliest alterations preconditioning normal urothelium to BLCA development. Frequent LOY in blood and urothelium-derived tissues suggest its involvement in BLCA.

Keywords: bladder carcinoma (BLCA); chromosomal copy number alterations; copy neutral loss of heterozygosity (CN-LOH); cystectomy; loss of heterozygosity (LOH); mosaic loss of chromosome Y (LOY); normal urothelium; post-zygotic mutations; transurethral resection of bladder tumor (TURBT).

Figure 1

Experimental and computational workflow in the study. (A) From each patient (n = 67), a minimum of one tumor specimen and several margins at varying distances were studied. Cancer cell content was determined through histopathological examination, resulting in 145 cancerous, 277 non-cancerous- and 63 blood samples. DNA was genotyped using the Infinium Global Screening BeadChip. (B) Outline of the computational pipeline for detection of chromosomal aberrations in above samples. The Mosaic Chromosomal Alteration (MoChA) caller was used to detect loss, gain and CN-LOH. Alteration profiles of blood and non-cancerous samples from each patient were used to classify them into germline and post-zygotic. Alterations classified as post-zygotic in non-cancerous samples were then compared against at least one matched cancerous profile, with shared change indicating plausible autosomal “cancer precursor candidate”. These candidates were evaluated across the entire cohort, revealing 16 distinct post-zygotic alteration hotspots in non-cancerous samples. The loss of chromosome Y (LOY) was assessed across 380 samples within a subset of 52 male donors. Abbreviations: BLCA—bladder cancer; TURBT—transurethral resection of bladder tumor; CN-LOH—copy-neutral loss of heterozygosity; LOY—loss of chromosome Y; CA—chromosomal alteration; PZM—post-zygotic mutation.

Figure 2

Genomic location and recurrence of cancer precursor candidates (CPCs) in samples of histologically normal urothelium. Analysis of gains (blue), losses (red) and CN-LOHs (green) among 480 CPCs in 64 NC samples of 30 patients yielded 16 hotspots of gain&CN-LOH or loss&CN-LOH type. Well-known cancer genes targeted by these hotspots are displayed as gene symbols, followed by the numbers in parentheses counting affected PM/DM samples and donors. For individual samples, CPCs shared by the same subject were counted several times. Table 3 and Table S5 show the hotspot’s GRCh37 coordinates, lengths and other genes previously implicated in bladder cancer. Abbreviations: CN-LOH: copy neutral loss of heterozygosity; hotspot: the shortest overlapping region of highest recurrence among patients and individual samples. Numbers before /-sign and after /-sign indicate numbers of patients and samples, respectively.

Figure 3

Comparison of performance for LOY estimation between MoChA and ddPCR. The external cohort of blood samples (n = 273) was used to compare LOY estimation via MoChA (%LOY on the X axis) and ddPCR (%LOY on the Y axis). When MoChA detects LOY (blue circles), %LOY agrees well with ddPCR values up to values < 65%. The samples where MoChA does not detect any LOY are represented by red color. Five red triangles benefit from ddPCR verification as they have very high %LOY and these samples do have enough heterogeneous probes within PAR1 for MoChA to reliably compute BAF deviation. The grey area represents the standard error of the linear regression model (%LOY ddPCR) = α + β (%LOY MoChA) where α = 0.02 (95% CI [0.013, 0.036]) and β = 0.99 (95% CI [0.93, 1.05], p < 1 × 10^–3 ***). The Pearson’s correlation coefficient for 67 samples where MoChA detects LOY was r = 0.97 (95% CI [0.96, 0.98], p < 1 × 10⁻³ ***, R² = 0.94). This shows that MoChA is a reliable tool for estimating LOY in blood samples, up to values of about 65%.

Figure 4

Mosaic loss of chromosome Y (LOY) in samples from bladder cancer cohort. This figure shows the median log R ratio (LRR) of the non-PAR region of chromosome Y, adjusted by median LRR of autosomes, and the percentage of cells with LOY (%LOY), estimated via MoChA using B-allele frequency (BAF) deviation of PAR1 region of chromosome Y, for 52 male patients of the BLCA cohort. Blood samples (n = 52, green), cancer samples (n = 111, red) and non-cancerous samples from bladder mucosa (n = 205, blue) are shown. Sample KX47Z_UM99 has been reclassified as a tumor sample after histopathological analysis. Five samples (red crosses) with either gain of chromosome Y (GOY) or XXY genotype are upper outliers and all these are tumor samples. The inset summarizes the number and percent of samples at the conservative threshold of LOY > 10% for three types of samples. ddPCR was used to validate %LOY (shown as percentages after sample label) for outliers without enough heterozygous probes, where BAF deviation is not reliably measured. Out of 7 left-bottom outliers, MoChA was able to detect LOY for n = 2 samples (triangles), but underestimated its cellular fraction, LOY < 10% was correctly detected by MoChA for EXDD8_PT1A sample, and LOY > 10% was missed by MoChA for n = 5 (asterisks) samples, which illustrate the need to complement MoChA’s output with other methods. Three samples in the rightmost part of the LOY cloud, deviating from the linear trend, are where MoChA underestimates %LOY and these were also validated with ddPCR. Abbreviations: PAR, pseudoautosomal region of chromosome Y; PAR1, pseudoautosomal region 1.

Figure 5

Conditional probabilities of LOY co-occurrence between cancerous (C), non-cancerous (NC) and whole blood (WB) samples. Probability of detecting LOY > 10% in at least one C or NC sample is significantly higher if WB sample of this donor has LOY > 10%. Points show medians, horizontal bars show 95% HDI from Bayesian regression, controlling for age, age² and smoking history confounders, removing donors where data were missing. p-values are Bonferroni-adjusted. There are 42 donors with non-missing LOY, age and smoking data in both C and WB tissues, and 43 donors with non-missing LOY, age and smoking data in both NC and WB tissues.