Allelic Imbalance and Chromothripsis Lead to Diversity in Japanese Tumor Genomes With Whole‐Genome Duplication

ABSTRACT

Whole‐genome analyses have revealed that large‐scale structural variations (SVs) such as whole‐genome duplication (WGD) occur early in the development of many cancers. However, the diversity of chromosomal abnormalities within tumors before and after WGD remains poorly understood. Here, we analyzed various types of Japanese tumor genomes via whole‐genome sequencing and examined the diversity of WGD by focusing on large SVs at the chromosomal level. WGD was detected in 52% of cases, while the frequency of chromothripsis (CT) was 20%. Although aneuploidy via deletion of chromosome arms was common in many cancers, in rare ovarian cancers, all chromosomes were near‐haploidy before WGD. Minor allele analysis revealed that many non‐mutated ohnolog genes drifted down chromosome arms after WGD and returned to normal ploidy, but only 17p, including TP53, which is also an ohnolog, underwent loss of heterozygosity due to arm deletion before WGD in most cancers. TP53 mutations were frequently detected in WGD and CT‐positive tumors, and these SVs strongly correlated with homologous recombination deficiency scores. Furthermore, these tumors had many mutations that continued to generate neoantigens and resulted in worse survival outcomes. Diversity analysis of tumors with WGD will provide a new perspective on structural abnormalities in tumor genomes.

Japanese tumor genomes with whole‐genome duplication are diversified by allele imbalance and chromothripsis. These structural abnormalities are linked to a poor prognosis for cancer patients.

Abbreviations

cGAS

cyclic GMP‐AMP synthase

CT

chromothripsis

GEP

gene expression profiling

HRD

homologous recombination deficiency

STING

signaling effector stimulator of interferon genes

SV

structural variation

TMB

tumor mutational burden

TSVB

tumor structural variation burden

WGD

whole‐genome duplication

WGS

whole‐genome sequencing

1. Introduction

Whole‐genome doubling/duplication (WGD) is the duplication of an entire chromosome within a cell and is estimated to occur in approximately 30% of human cancers [1]. WGD facilitates the acquisition of chromosomal alterations [2] in a background permissive for genomic instability, such as in TP53‐deficient cells, which may promote tumorigenesis [3, 4]. WGD is also considered to have had a significant effect on species diversification, leading to increased biological complexity and the origin of evolutionary novelty [5]. Duplicated genes derived from WGD during evolution are called ohnologs and are refractory to copy number variations (CNVs) [6]. Owing to their sensitivity to dosage balance, focal duplication involving ohnologs is not tolerated in vertebrate genomes, leading to pathological behavior in an ohnolog with CNV [7].

WGD in tumors is also known as a chromosomal event that often causes allelic imbalance, such as loss of heterozygosity (LOH) or aneuploidy, resulting in unequal numbers of maternal and paternal copies. Recently, next‐generation sequencing (NGS) analysis revealed that somatic mutations of TP53 and KRAS with allelic imbalance are associated with poor prognosis in myelodysplastic syndromes and several solid tumors, respectively [8, 9, 10, 11]. Niknafs et al. reported that somatic mutations in the LOH region (persistent mutation) can sustain anti‐tumor immune responses [12]. These persistent mutations are expected to be maintained within tumors under the selective pressure of immunotherapy, creating evolutionary bottlenecks that cancer cells cannot overcome. In evaluating the tumor genome, including chromosome‐level structural variation (SV) via NGS, the pathogenicity of a mutation and the allelic status of its gene provide important information for the treatment of patients with cancer.

The development of NGS enabled the detection of copy number alterations in whole genomic regions and also uncovered the presence of unknown SVs such as chromothripsis (CT) [13]. A hallmark of CT is multiple oscillations between two or three copy number states, with a detection algorithm being developed based on this feature, and more than 50% of CTs in the PCAWG cohort were detected in several tumor types [14]. This chromosomal catastrophic event is suspected to be associated with TP53 dysfunction, and two mechanisms of CT occurrence are proposed [15, 16]. However, the relationship of CT to other structural abnormalities associated with TP53 deficiency, such as WGD and allelic imbalance, and its frequency among races remains poorly understood.

In this study, we performed whole‐genome sequencing (WGS) focusing on large structural abnormalities involving copy number changes, including WGD, in Japanese tumors. Specifically, chromosome‐level loss and gain (aneuploid arm) and the effect of ohnolog on allelic imbalance were evaluated in WGD‐positive tumors. We also investigated the frequency of CTs in WGD‐positive tumors and the relationship between these SVs and persistent mutations in the homozygous regions. Furthermore, we performed a survival analysis of subgroups classified based on SVs. Elucidating the diversity of tumor genomes before and after WGD will provide new insights into structural aberrations.

2. Materials and Methods

2.1. Clinical Samples

Tumor samples and peripheral blood were subjected to WGS to identify somatic and germline genomic alterations in patients with neoplasms, and tumors and their surrounding samples were used for gene expression profiling (GEP). Samples were obtained from patients who underwent surgery at the Shizuoka Cancer Center Hospital, and samples with tumors > 100 mg were analyzed. A blood sample from the same patient was used as a control for WGS, and tissue surrounding the tumor was used as a non‐tumor control for GEP. WGS and GEP were performed on a NextSeq 6000 System (Illumina, San Diego, CA, USA) and a SurePrint G3 Human Gene Expression 8 × 60 K v2 Microarray (Agilent Technologies, Santa Clara, CA, USA), respectively. Details of the experimental procedures have been previously described [17, 18]. Tumors were diagnosed by pathologists in our hospital and then classified according to the OncoTree code (Table S1).

2.2. WGS

Whole human genome 150‐bp paired‐end sequencing was performed on DNA extracted from tumor and peripheral blood. A library was constructed from 1 μg of DNA using a TruSeq DNA PCR‐Free High Throughput Library Prep Kit (Illumina). Raw data were converted to FASTQ format using Bcl2fastq v2.20 (Illumina), and the sequenced reads were mapped onto a reference human genome (hs37d5) using the DRAGEN Bio‐IT Platform ver. 3.9 (Illumina) and finally converted into BAM files.

2.3. SV Detection

SV was defined as ≥ 50 bp and was detected and categorized into five classes (translocation, insertion, deletion, duplication, and inversion) using the DRAGEN Structural Variant Caller. The number of detected SVs per kb was defined as tumor structural variation burden (TSVB). CNVs were scored using Sequenza var.3.0.0 [19], which calculates the likelihood of copy number status for various combinations of ploidy and tumor content and reports the most likely value. CNV analysis was performed using BAM files of tumor and control samples from the same patient. Following a previous study [20], “kmin” and “gamma” were set to 200 for the sequenza.extract function. Regions not fully covered on the genome were merged when neighboring regions had the same total and minor copy numbers, and their distance was < 1000 bp. The remaining uncovered regions after this procedure were assumed to have a normal diploid status and were assigned a total and minor copy number of 2 and 1, respectively. Cases with major CN ≥ 2 in > 50% of the whole genomic region were defined as WGD positive. Aneuploidy was counted when gain or loss was detected on > 90% of a chromosome arm. The event and region of CT were evaluated using ShatterSeek ver.1.1 [14] based on the results of SV and CNV analysis using the conditions recommended in the tutorial. To confirm this event, including high and low confidences, we subsequently visualized oscillation and CNs across multiple samples and verified them visually. The INV3 and INV5 tags in the VCF file were assigned as the two types of inversions (head‐to‐head and tail‐to‐tail). Finally, a confidence level was determined for each detected CT event based on the criteria described in the manual of this algorithm. The homologous recombination deficiency (HRD) score was calculated as the sum of the following three scores using the DRAGEN Homologous Recombination Deficiency caller: (1) LOH score, (2) telomeric allelic imbalance (TAI) score, and (3) large‐scale state transition (LST) score. The algorithm was designed to indicate HRD positivity if the calculated sum was high (≥ 42) according to previous studies [21]. Acrocentric chromosomes (chr13, 14, 15, 21, and 22) were analyzed for the q‐arm only because several SV analyses cannot be performed. The detailed WGS pipeline for SV detection has been described previously [22].

2.4. Whole‐Genome Profiles Other Than SVs

Germline and somatic variations (< 50 bp) were identified using the DRAGEN Somatic Variant Caller and Small Variant Caller (Illumina) and were annotated using Ensembl Variant Effect Predictor ver. 104 [23]. The tumor suppressor gene and oncogene catalogs were extracted from the 1988 cancer‐related genes previously reported [17]. All detected somatic variants were considered in determining the tumor mutational burden (TMB), regardless of variant type (SNV or indel). Tumor content was determined using the “estimated tumor content” provided in the CNV analysis using the Sequenza algorithm. The non‐negative matrix factorization method was used to detect mutational signatures in WGS samples. Mutational signature analysis was performed using MutationalPatterns v3.0 [24]. A CNV profile comprising 48 patterns was constructed from copy number status (0, 1, 2, 3–4, 5–8, ≥ 9 ), heterozygosity status (LOH, heterozygous, homozygous deletion), and region length (< 100 kb, 100 kb‐1 Mb, 1–10 Mb, 10–40 Mb, > 40 Mb) according to a previous report [25]. These profiles were decomposed into an optimal combination of SBS and CN signatures in COSMIC (version 3.1), and signatures described as artifacts were excluded from the analysis. The detailed WGS pipeline, including detailed quality control and annotation, has been described previously [22]. Ohnologs were extracted from the registered list in OHNOLOGS (http://ohnologs.curie.fr/) [26].

2.5. Gene Expression Signature Analysis

For RNA analysis, tumor and peritumoral non‐tumor tissues were immediately placed in RNAlater solution (Thermo Fisher Scientific, Waltham, MA, USA). Purified total RNA for GEP was amplified and fluorescently labeled using a One‐Color Low Input Quick Amp Labeling Kit (Agilent Technologies), and the Cy3‐labeled cRNAs were then hybridized onto a SurePrint G3 Human Gene Expression 8 × 60 K v2 Microarray (Agilent Technologies). Signature analysis was performed using the expression ratio of the tumor and the corresponding non‐tumor tissue (T/N) in the same patient [27]. The expression signature was calculated from the average of genes in unique gene sets corresponding to each individual signature (CDK4/6‐RB [28], T cell‐inflamed [29, 30, 31], cGAS‐STING [32, 33], and Immune evasion [34]).

2.6. Statistical Analysis

The gene expression signature and continuous variables in the SVs analysis were compared using Welch's t‐test. Linear correlations between the two datasets were evaluated using Pearson's correlation coefficient. The log‐rank test was performed for overall survival curves. The Benjamini‐Hochberg procedure (q < 0.01) was used to control for the false discovery rate (FDR), and statistical significance was set at a p value < 0.05.

3. Results

3.1. SVs in the Japanese Tumor Genome

From our WGS cohort (n = 1355), we selected 674 cases that were labeled according to OncoTree code after clinician diagnosis, had an estimated tumor content ≥ 0.3, and had gene expression data (tumor and non‐tumor). Cancer types represented by OncoTree codes are listed in Table S1, and the SV profile for each cancer type is shown in Figure 1A. TSVB varied widely among cancer types and even within the same type. This burden positively correlated with TMB in 27% (6/22) of the cancer types excluding others. Insertions were significantly more frequent in LUSC, COAD, and READ. LUAD showed higher frequencies of inversions and translocations, while HCC showed higher frequencies of duplications and deletions. WGD and CT were more frequent in cancer types with high TSVB, with WGD and CT identified in 52% (350/674) and 20% (138/674) of this cohort, respectively. The frequency of WGD was significantly higher in the Japanese cohort than in the White cohort evaluated using the same algorithm [1, 35] except for breast cancer (Table S2). WGD was evaluated based on major copy number (MCN) [1] and was consistent with 98.2% of WGD cases determined using the LOH/ploidy‐based method adopted in a previous study [36] (Figure 1B). The MCN evaluation detected all cases with inconsistent results (12/674) as WGD‐positive. Not surprisingly, HRD scores, which are the sum of three structural abnormalities, were also higher in tumors with high TSVB. Aneuploidy was rarely observed in WGD‐negative tumors, whereas aneuploidy due to chromosome arm loss occurred significantly more often than aneuploidy due to gain in WGD‐positive tumors.

FIGURE 1.

Structural variation (SV) profile in Japanese tumor genomes. (A) SV in each cancer type in 674 samples. From top: Tumor SV burden (TSVB), correlation between TSVB and tumor mutational burden (TMB), SV type, proportion of whole‐genome duplication (WGD), aneuploidy in tumors with or without WGD, proportion of chromothripsis (CT), and median homologous recombination deficiency (HRD) score. Cancer types are ordered according to median TSVB. The number in the parentheses for TSVB represents the number of samples. A high Pearson correlation between TSVB and TMB indicates that TMB increases with the frequency of SVs. Aneuploidy was counted when gain or loss was detected on > 90% of a chromosome arm in WGD‐positive or WGD‐negative tumors. The HRD score was calculated as the sum of large‐scale state transition (LST), telomeric allelic imbalance (TAI), and loss of heterozygosity (LOH). *p < 0.05, **p < 0.01. Tumor abbreviations follow the OncoTree code. The ‘other’ group contains multiple tumor types that have less than five samples. INS, insertion; INV, inversion; DUP, duplication; DEL, deletion; BND, breakend (translocation). (B) A comparison of alternative WGD detection methods with the present method. Cases above the dashed line are classified as WGD based on loss of heterozygosity and ploidy. (C) Relationship between aneuploidy in chromosome arms and karyotype of tumors. In the upper panel, in the absence of WGD, the karyotype is denoted as 2n. If WGD was detected and the average ploidy was five or higher, the karyotype is defined as 6n; otherwise, it was defined as 4n. The middle and lower panels show the distribution of the average ploidy and the number of aneuploidy events categorized as loss or gain. (D) Minor copy number (CN)‐null and aneuploidy with deletion in WGD‐positive tumors. Imbalance events that produced minor CN‐null (minor CN = 0) in > 90% of chromosome arms were counted. WGD‐positive tumors tend to retain minor alleles even when aneuploidy loss occurs.

When aneuploidy events were tabulated according to karyotype (2n, 4n, and 6n), such events due to chromosome arm loss were prominent in tumors identified as tetraploid (Figure 1C). As these losses increased, the average ploidy decreased linearly toward 2. Chromosome arm gain tended to be more frequent at 5p and 8q, and this trend was observed more frequently in WGD‐positive LUAD and in WGD‐negative hepatocellular carcinoma (HCC) (Figure S1). The gain was characteristically observed on chromosomes 7p and 20 in WGD‐positive tumors, 58% of which were LUAD, COAD, and READ. Aneuploidy due to loss of chromosome arms was observed across the autosomes, especially on chromosomes 17p and 19p, in more than 60% of cases. Loss of chromosome arms strongly negatively correlated with gain, suggesting that loss is less likely to occur on chromosome arms in which gain is predominant.

To determine whether biased aneuploidy occurred, chromosome arms with allelic imbalance, especially those with a minor CN‐null (CN = 0), were scored per tumor (Figure 1D). In this cohort, five cases with minor CN‐null in more than 30 of the autosomal arms were identified (Figures 1D and S2). The top three of these were rare non‐epithelial carcinomas of ovarian origin, and a complete split of B allele frequencies was observed in all autosomes.

3.2. Allelic Imbalance in WGD‐Positive Tumors

To determine when aneuploidy occurred, we evaluated copy number imbalance based on minor alleles of chromosome arms (Figure 2A). Minor CN‐null (CN = 0) was detected more frequently in WGD‐positive tumors than in WGD‐negative tumors, with 60% of 17p harboring minor CN‐null. Somatic mutations of TP53 on 17p were also increased by WGD, and 87% (229/262) of WGD‐positive tumors with TP53 mutations have a minor CN of 0 on 17p (Figure 2B). This result indicated the loss of the wild‐type TP53 allele after WGD. TP53 is also known to be a gene classified as an ohnolog that is associated with WGD [35, 37]. We investigated the copy number changes in other genes classified as ohnologs after WGD (Figure 2C). The copy number of ohnolog genes without mutations was significantly lower than that of the original karyotype (2n), even after WGD. In contrast, ohnolog genes with mutations maintained the copy number of the karyotype (≥ 4n) after WGD. Genes not classified as ohnolog and with or without mutations had a predominant copy number ≥ 4 (Figure S3A). In addition, the top 5 oncogenes classified as ohnolog (EGFR, GNAS, CTNND2, PREX2, and KIT, excluding TP53) detected in this cohort mostly maintained CN ≥ 4 after WGD (Figure S3B). The sequential changes in CN via WGD in TP53 and other ohnolog genes based on the above results are presented as a schematic diagram (Figure 2D). Ohnolog genes that maintained function without mutations tended to have four copies. In contrast, ohnolog genes with mutations that may prevent normal function tended to remain in a heterozygous state with four copies.

FIGURE 2.

Allelic bias with LOH in WGD‐positive tumors. (A) Frequency of minor allele CN in each chromosomal arm. The p arms of acrocentric chromosomes (chromosomes 13, 14, 15, 21, and 22) were excluded from copy number analysis. (B) Minor allele CN and variant allele frequency (VAF) of TP53 mutations in tumors with or without WGD (left panel), and Minor allele CN in TP53‐mutated tumors (right panel). In tumors with TP53 mutations, chromosome 17p maintained TP53‐mutated alleles through allelic imbalance without minor alleles. **p < 0.01; n.s., not significant. (C) Difference in the number of tetraploid‐maintaining ohnolog genes with and without somatic mutations in WGD‐positive tumors. (D) Schematic representation of the predominant changes in the chromosomal arms encoding TP53 and ohnolog genes before and after WGD. Although TP53 is an ohnolog, because TP53 mutations and deletions occurred prior to WGD, WGD produced homozygous TP53 mutations while maintaining the diploid karyotype (2n). Consequently, the homozygous state with mutations in both alleles remained even after WGD. Chromosomal arms containing non‐mutated ohnologs tended to drift down and return to their original karyotype (2n) after WGD, whereas those containing ohnologs with mutations tended to maintain the karyotype obtained at WGD.

3.3. CT in WGD‐Positive Tumors

Cancer type with higher TSVB showed a higher proportion of WGD and CT detection (Figure 1). We counted CT events on each chromosome in WGD‐positive and ‐negative tumors (Figure 3A) and observed the highest frequency in tumors with WGD, although more than half of the CT detected in GEJ and HCC was not accompanied by WGD. The occurrence of these events strongly correlated with chromosome length but was higher for chromosome 16, a relatively small chromosome, with 55% of these events occurring in STAD, HCC, and IDC (Figure 3B).

FIGURE 3.

Association of WGD with other structural abnormalities. (A) Number of CT events detected in WGD‐positive tumors for each cancer type. (B) Correlation of CT with chromosome length. (C) HRD scores in WGD and CT‐positive tumors. *p < 0.05, **p < 0.01. Significant differences (p < 0.05) were observed in all four event combinations involving wild‐type TP53. In TP53‐mutated tumors, all combinations were significantly different (p < 0.05) except those that were not significant (n.s.). The HRD score is defined as the sum of the loss of heterozygosity (LOH) score, telomere allelic imbalance (TAI) score, and large‐scale state transition (LST) scores calculated by WGS.

3.4. Effect of WGD and CT on HRD Detection

Cancer types with higher TSVB also had higher HRD scores calculated from WGS (Figure 1). We evaluated how the presence or absence of WGD and CT events affected HRD‐related indicators. HRD scores were highest in WGD‐ and CT‐positive tumors and lowest in tumors in which both were undetectable; TP53 mutations significantly increased scores in this classification (Figure 3C). Compared with WGD‐positive tumors, WGD‐negative tumors tended to have a lower proportion of TAI, one of the components of the HRD score (Figure S4A). The frequency of TP53 mutation was also higher in WGD‐ and CT‐positive tumors, and a strong positive correlation was observed between the incidence of TP53 and HRD score by cancer type (Figure S4B,C). The frequency of BRCA1/2 somatic and germline mutations ranged from 4% to 14% (Figure S4D). In addition, we performed an analysis using the CN signature, which is another measure that is associated with HRD. The contribution of CN2 (tetraploidy) and CN17 (HRD) was higher in CT‐positive tumors, and by cancer type, CN2 was implicated in COAD and READ and CN17 in HNSC and AMPCA (Figure S4E). These results suggested that WGD‐ and CT‐positive tumors tended to have high HRD scores, which are likely elevated by TP53 mutations.

3.5. Persistent SNVs (pSNVs) in WGD‐Positive Tumors

Allelic imbalance, including minor CN‐null, was detected more frequently in WGD‐positive tumors than in WGD‐negative tumors (Figure 2A). Nonsynonymous substitutions in minor CN‐null alleles continuously produce proteins with abnormal amino acid sequences because normal alleles are not present. We counted persistent SNVs that continue to generate neoantigenic source mutations through genes with minor CN‐null. SNVs with non‐synonymous substitutions in protein‐coding genes with a minor CN of 0 were defined as pSNV. The pSNVs were highest in WGD‐ and CT‐positive tumors and lowest in tumors in which both were undetectable (Figure 4A). The distribution of pSNVs varied among cancer types and exhibited a trend of accumulation similar to that of TSVBs (Figure S5A). To evaluate the potential of pSNVs to generate immunogenic neoantigens, we compared the proportion of predicted binding complexes with IC₅₀ values below 100 nM between pSNVs and normal SNVs in MHC class I and class II alleles. The pSNVs exhibited similar or higher frequencies of predicted neoantigens compared to normal SNVs (without minor CN‐null), particularly in MHC class I (p < 0.05), while no significant difference was observed in MHC class II (Figure S5B). In addition, gene expression‐based signature analysis showed that the CDK4/6‐RB signature involved in the cell cycle was significantly higher in WGD‐ and CT‐positive tumors, while no significant differences were observed between the subgroups for the T‐cell‐inflammatory and cGAS‐STING signatures involved in cancer immunity. Additionally, gene expression signatures associated with immune evasion, which decrease in tumors with a high neoantigen load, were reduced in WGD‐positive and CT‐positive tumors (Figure S6).

FIGURE 4.

Persistent SNV (pSNV) and gene expression signatures in WGD and CT‐positive tumors. (A) Distribution of pSNVs. SNV, single nucleotide variation; CN, copy number. **p < 0.01. (B) Gene expression signatures of CDK4/6‐RB, T cell‐inflamed, and cGCAS‐STING.

3.6. Overall Survival in WGD‐Positive Tumors

Finally, we investigated the overall survival of cases whose survival could be tracked in this cohort (Figure 5A). Overall survival was worse for cases with WGD‐ and CT‐positive tumors, while a better prognosis was observed for CT‐positive tumors. No bias was found in the cancer types in the WGD‐ and CT‐positive tumors (Figure 5B). In multivariate Cox regression analysis, the group with both WGD and CT (WGD+/CT+) showed a significantly increased risk of death compared to the WGD‐/CT‐ group (HR = 2.033, p = 0.018, Table S3). These results suggest that WGD and CT can be used as prognostic markers across cancer types.

FIGURE 5.

Survival outcomes in WGD and CT‐positive tumors. (A) Kaplan–Meier plot of overall survival. (B) Distribution of tumor types used for survival outcomes.

4. Discussion

International whole‐genome analysis has challenged the detection of WGD and CT in pan‐cancer, revealing the frequency of these structural abnormalities [14, 38]. Whole‐genome analysis of tumors has been performed worldwide. We also performed WES and WGS of Japanese tumor genomes and found that the detection of somatic mutation and CNV was dependent on tumor content and racial differences [20, 39]. In particular, WGS analysis revealed that the detection accuracy of CNV is lower than that of SNV and SV in several algorithms when the estimated tumor content is less than 0.3. This fact raises concerns regarding the effect of low tumor content on WGD, wherein total copy number is a criterion for assessment, and CT detection, wherein copy number oscillation is a criterion. Therefore, we excluded samples with low tumor content (< 0.3) from the analysis to accurately evaluate WGD and CT.

Previous reports have shown that allelic imbalance, including LOH, is a common event in WGD‐positive tumors [40, 41]. We integrated aneuploidy and minor copy number variation into the WGD analysis and found that WGD‐positive tumors exhibit diversity in aneuploidy and in allelic imbalance. Consistent with the findings of previous reports, aneuploidy with loss of chromosome arms was significantly more common in WGD‐positive tumors than in WGD‐negative tumors. An average of 9.2 homozygous chromosome arms (minor CN‐null) was detected in WGD‐positive tumors, implying that these chromosome arms had a single‐allele loss prior to WGD. WGD acts as a buffer against harmful allele loss by increasing the number of CN states that chromosomes can adopt [41, 42]. Consequently, WGD is frequently observed after allele loss. Furthermore, in WGD‐positive non‐epithelial ovarian cancers (ovarian yolk sac tumor, immature ovarian teratoma, and ovarian malignant germ cell tumor), allelic imbalance was observed on all autosomes. The cellular origin of these tumors is known to involve germ cells [43, 44], and our results strongly suggested, for the first time using WGS, that these tumor genomes may have undergone haploidization. In other WGD‐positive tumors with allelic imbalance in ≥ 30 chromosome arms, heterozygosity is maintained in several chromosome arms, suggesting that diploidy occurs after acquiring near‐haploidy in normal or precancerous cells (not germ cells). In WGD‐positive tumors where excessive allelic imbalance is detected by WGS, the tumor type and ploidy state must be carefully analyzed.

The frequency of 17p minor CN‐nulls in Japanese WGD‐positive tumors exceeded half the total and was even higher in tumors with TP53 mutations. Recent studies have shown that loss of TP53 function promotes WGD [45, 46]. Our results support this, indicating that TP53 mutations and LOH should be considered as antecedents of WGD. We also found that ohnolog genes without mutations have a predominant copy number of 2 after WGD. Ohnologs, which are important dosage‐sensitive elements of the genome, have been implicated in some of the deleterious phenotypes observed in pathogenic CNVs [7, 47], and our results suggest that ohnolog duplications are also deleterious in cancer cells. In contrast, mutated ohnologs had a predominant copy number of 4 after WGD. This trend was particularly observed for major oncogenes such as EGFR or GNAS, suggesting that mutated ohnologs do not behave as duplicated genes in cancer cells because the mutated genes have acquired an oncogenic function. In other words, even if these genes increase to four copies after WGD, two of the genes will remain non‐mutated with normal function, while the other two will become oncogenic (Figure 2D). Ohnologs that are classified as oncogenes function as genes with different roles and are likely to avoid the effects of dosage balance, making negative selection less likely after WGD. The incorporation of the concept of ohnologs, a key concept of biological evolution, may provide novel insights into cancer development.

WGD and CT strongly correlated with HRD score, with TP53 mutations being involved in these correlations. Loss of normal TP53 function causes chromosomal instability, leading to structural abnormalities [48, 49]. Meanwhile, the typical indicators used to discriminate HRD are scores of structural alterations (LOH, telomeric allelic imbalance, and LST), which are only surrogate HRD markers and consider structural alterations that occur when normal homologous recombination repair is not completed. HRD scores increase in tumors with TP53 mutations in breast, lung, and gastric cancers [50, 51, 52]. Similar to our study, these reports suggest that chromosomal instability mediated by TP53 mutations may influence HRD scores, although evidence of a causal relationship between HRD and TP53 mutations is lacking. It is reasonable to assume that WGD and CT‐positive tumors have high HRD scores and that chromosomal instability due to TP53 mutations induces WGD and CT, rather than dysfunctional repair causing WGD and CT. However, BRCAness, which has a DNA repair dysfunction phenotype without BRCA1/2 abnormalities, is also known to be associated with TP53 mutations [53, 54]. Our results also showed that even in tumors without TP53 mutations, WGD‐ and CT‐positive tumors had higher HRD scores than negative cases. Therefore, the possibility that HRD causes CT and WGD cannot be excluded. Future in vitro experiments considering TP53 mutations are needed to evaluate the relationship of HRD with WGD and CT.

Somatic mutations in homozygous chromosomal regions are considered to increase the response rate to immune checkpoint inhibitors by generating persistent neoantigens, and WGD promotes the production of these persistent mutations [12]. In our cohort, persistent mutations generated by allelic imbalance were significantly higher in WGD‐ and CT‐positive tumors, and the immune evasion scores were also lower. Furthermore, pSNVs retained a comparable or even greater capacity to present high‐affinity neoantigens via MHC class I and II, similar to heterozygous SNVs. This suggests that, despite being expressed solely from the mutant allele, pSNVs may continue to generate immunogenic neoantigens through mutant‐only transcription and translation. However, no significant differences in the two signatures associated with immune checkpoint inhibitor response rates were found. Like Jekyll and Hyde, WGD‐positive tumors also result in aneuploidy, which lowers the response to immunotherapy [55]. This raises the possibility that WGD and CT‐positive tumors contain a population that accumulates persistent somatic mutations but are less responsive to immune checkpoint inhibitors. In contrast, WGD and CT‐positive tumors exhibit high CDK4/6 activity based on the CDK4/6 RB expression signature. For such tumors, new immunotherapies that aim to enhance tumor antigen presentation by targeting the CDK4/6‐RB pathway have been proposed [56]. The use of CDK4/6 inhibitors against WGD and CT‐positive tumors may enhance tumor antigen presentation and consequently promote anti‐tumor immunity against potentially aberrant protein‐prone tumors. Cell experiments will be necessary in the future to evaluate this possibility.

We focused on SVs at the chromosomal level and determined the frequencies of WGD and CT in Japanese tumor genomes. However, although our NGS pipeline for the detection of SVs, including WGD and CT, uses Illumina chemistry, it is not completely identical to that of previous WGS cohorts. This difference may influence SV detection. Although we evaluated the CN loss (CN = 0) of ~40 samples for several highly homogeneous genomic regions using PCR, we did not verify this for large genomic regions or with a statistically valid number of samples. In addition, our pan‐cancer analysis was performed on surgical cases at our hospital; thus, there is a bias toward cancer types because it does not affect treatment and diagnosis. This bias may influence the frequency of CT and WGD. This cohort should be expanded in future research to perform a more detailed analysis according to tumor type.

In conclusion, the present study focusing on chromosomal alterations highlights the hidden diversity of WGD. The TP53 mutations and ohnolog genes affected chromosomal gain and loss before and after WGD, causing allelic imbalance. Furthermore, WGD‐ and CT‐positive tumors accumulated persistent mutations that can keep generating potential neoantigens and result in worse survival outcomes. Characterization of WGD‐positive tumors based on SVs may provide new perspectives for the development of new biomarkers and immunotherapeutic strategies.

Author Contributions

Keiichi Hatakeyama: conceptualization, formal analysis, funding acquisition, investigation, writing – original draft. Takeshi Nagashima: data curation, investigation, methodology, writing – review and editing. Sumiko Ohnami: investigation. Shumpei Ohnami: investigation. Koji Maruyama: investigation. Keiichi Ohshima: data curation, formal analysis. Yuji Shimoda: investigation, methodology. Akane Naruoka: investigation. Hirotsugu Kenmotsu: investigation. Kenichi Urakami: investigation. Yasuto Akiyama: investigation. Ken Yamaguchi: supervision, writing – review and editing.

Ethics Statement

All aspects of this study were approved by the Institutional Review Board of Shizuoka Cancer Center (approval number 25–33). Frozen and blood specimens were collected and used to predict pathogenic germline alterations. Appropriate informed consent, including the possibility of secondary findings such as those identified in blood‐based constitutional analysis, was obtained from the sample donor with the approval of the Ethics Review Board so as not to disadvantage them. Human studies followed the ethical guidelines for clinical application in accordance with the Declaration of Helsinki.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: List of OncoTree codes used in this study.

Table S2: cas70159‐sup‐0001‐DataS1.pdf. Comparison of WGD frequency with cohorts using the same algorithm.

Table S3: cas70159‐sup‐0001‐DataS1.pdf. Univariate and multivariate analysis of mortality.

Figure S1: cas70159‐sup‐0001‐DataS1.pdf. Circos plot and B allele frequency (BAF) of a representative whole‐genome duplication (WGD) positive case with a high frequency of minor copy number (CN)‐null and aneuploidy with deletion.

Figure S2: cas70159‐sup‐0001‐DataS1.pdf. Distribution of aneuploidy in each chromosome arm.

Figure S3: cas70159‐sup‐0001‐DataS1.pdf. Alteration in chromosomal arms containing genes other than ohnologs.

Figure S4:. Influence of WGD and chromothripsis (CT) on homologous recombination deficiency (HRD).

Figure S5: cas70159‐sup‐0001‐DataS1.pdf. Distribution of persistent SNVs (pSNVs) and frequency of neoantigens predicted from pSNV sequence.

Figure S6: cas70159‐sup‐0001‐DataS1.pdf. Immune evasion (IEV) score in WGD and CT‐positive tumors.

Data Availability Statement

The dataset presented in the current study has been submitted to the NBDC Human Database (https://humandbs.dbcls.jp).