Genomic subtypes and cellular phenotypes of high-grade endometrial carcinoma

doi:10.21203/rs.3.rs-4673415/v1

Download PDF

Article

Genomic subtypes and cellular phenotypes of high-grade endometrial carcinoma

https://doi.org/10.21203/rs.3.rs-4673415/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Recent genomic analyses of endometrial carcinoma (EC) have delineated four distinct subtypes: ultra-mutated (POLE) type, microsatellite instability–high (MSI-H) type, copy number–high (CN-H) type, and copy number–low (CN-L) type. Despite these classifications, the relationship between genomic subtype and histological classification of high-grade EC remains obscure. This study is a comprehensive genomic analysis specifically focusing on high-grade EC.

Examination of 81 high-grade EC tumors revealed that serous carcinoma is frequently characterized by TP53 mutations as described previously. However, the correspondence between genomic subtype and histological classification of high-grade EC was weak overall. Gene expression–based clustering showed that clear cell carcinoma, not specifically associated with any particular genomic subtype, had high expression of glandular/lumenal cell marker genes. We also identified a subset of tumors characterized by the expression of genes related to ciliated cells.

Our findings underscore the significance of considering cellular phenotypes that might reflect cell of origin and differentiation status, alongside genomic subtype, to precisely understand individual tumors of various histological types. Importantly, cellular phenotype might be closely associated with the immunological status of cancer cells; tumors with the glandular/lumenal phenotype are thought to present antigens more efficiently. Such stratification might have clinical implications, potentially impacting post-surgical adjuvant treatment or the use of immune therapy in women with high-grade EC. We propose that RNA clustering is a clinically valuable tool for the precise stratification and treatment planning for high-grade EC, emphasizing its potential to enhance personalized medicine approaches for EC.

Health sciences/Oncology/Cancer/Cancer genomics

Health sciences/Oncology/Cancer/Gynaecological cancer/Endometrial cancer

The incidence of endometrial carcinoma (EC) has increased substantially in recent years, even among younger age groups¹. ECs were generally classified into two types, with Type I characterized as estrogen-dependent (mainly grade 1–2 endometrioid carcinomas) and Type II characterized as estrogen-independent (mainly non-endometrioid (serous or clear cell) carcinomas, carcinosarcomas, and grade 3 endometrioid carcinomas). However, extensive genomic analyses in recent years have identified four distinct molecular subtypes based on genomic abnormalities: (1) ultra-mutated type (also referred to as POLE type), characterized by high tumor mutational burden (TMB) of ≥ 100 mutations per megabase pair (Mbp) through POLE mutations; (2) microsatellite instability–high (MSI-H) type, where mutations accumulate due to mismatch repair mechanism impairment (TMB ≥ 10 Mbp); (3) copy number–high (CN-H) type, with TP53 mutations and notable chromosomal copy number (CN) abnormalities; and (4) copy number–low (CN-L) type or no specific molecular profile (NSMP), which excludes the aforementioned categories^2,3.

The Proactive Molecular Risk Classifier for Endometrial Cancer (ProMisE) has been proposed as a more practical molecular classifier for EC^3,4. In ProMisE, MSI-high can be substituted with mismatch repair deficient (MMRd) on the basis of immunohistochemical (IHC) testing, such as MLH1, MSH2, MSH6, and PMS2 (or PMS2 and MSH6 alone⁵). TP53 status can also be evaluated on the basis of IHC testing of p53. The International Federation of Gynecology and Obstetrics (FIGO) staging of endometrial cancer has been updated in 2023 taking into account molecular subtypes.

However, the correlation between molecular subtype and histological classification of EC is weak, even for low-grade EC (grade 1–2 endometrioid carcinoma). The relationship is more complicated for high-grade EC⁴. The precise relationship between each genomic abnormality and its role in carcinogenesis, as well as the manifestation of distinctive tissue morphologies, remains elusive. For instance, clear cell carcinomas encompassed all four molecular subtypes (POLE, 6%; MMRd, 13%; TP53 abnormal, 34%; and NSMP, 47%) in 32 patients⁶. In 92 patients with carcinosarcoma, the prevalence of the four molecular subtypes was reported to be 11% for POLE, 26% for MSI-H, 53% for CN-H, and 10% for CN-L⁷. Although serous carcinomas are characterized by TP53 mutations (90%) (i.e., CN-H)), a small fraction is categorized as NSMP^8,9. These findings suggest that genomic profiling and information about genetic alteration and genomic instability status (such as MSI and chromosomal instability) is not sufficient to connect morphological or histological features with molecular subtype.

Unsupervised clustering based on RNA sequencing in The Cancer Genome Atlas Program (TCGA) dataset identified three robust clusters, but these clusters were composed of various molecular subtypes². In addition, the correlation between immune markers and molecular subtypes has not been fully analyzed. Recently, ARID1A deficiency was reported to be associated with MMRd and abundance of tumor-infiltrating lymphocytes in EC^10,11. The predominant signature of CD8⁺ cells was reported to be associated with favorable prognosis in serous EC, but not with prognosis in either MSI-high or NSMP EC, according to RNA sequencing–based single sample Gene Set Enrichment Analysis (ssGSEA) and Cell-type Identification By Estimating Relative Subsets Of known RNA Transcripts (CIBERSORT)¹². Therefore, the contribution of immune markers to each molecular subtype is yet to be clarified. Recently, immune checkpoint inhibitors (ICIs) plus platinum-based chemotherapy have been approved by U.S. Food and Drug Administration for primary advanced or recurrent endometrial carcinoma, further highlighting the importance of molecular subtypes and the necessity of other molecular markers to predict the sensitivity to ICIs^13,14.

The cellular origin of high-grade ECs, especially in terms of dynamic changes in the endometrium, is also unknown. Endometrial epithelial cells are classified functionally into secretory cells and ciliated cells. They are classified spatially into lumenal and glandular cells. Lumenal cells form the lumen. Glandular cells form glands in the functional and basal layers¹⁵. The endometrium undergoes dynamic changes throughout the menstrual cycle in response to ovarian hormones. Single-cell and spatial transcriptome analyses of endometrial cells highlight the cellular fate of epithelial lineages in endometrial microenvironments^15–19. However, it is unclear whether these epithelial lineages are associated with molecular or histological subtypes of high-grade EC. To identify the connection between genomic abnormalities and histological classification, we conducted a genomic analysis specifically focusing on high-grade EC.

Classification of high-grade endometrial carcinomas based on the results of exome sequencing

We analyzed 81 tumors classically categorized as Type II, encompassing non-endometrioid tumors as well as grade 3 endometrioid carcinomas (EMG3s). Histological type was confirmed by a gynecologic pathologist (H.Y.) according to the 2020 World Health Organization (WHO) classification²⁰ while blinded to genomic analysis. The analyzed tumors included 16 samples of serous carcinoma, 23 samples of clear cell carcinoma, 21 samples of carcinosarcoma, 19 samples of EMG3, and 2 other samples (including 1 small cell carcinoma and 1 unclassifiable poorly differentiated carcinoma) (Table S1). Among clear cell carcinomas, seven tumors were of mixed type containing endometrioid carcinoma components and two tumors were of mixed type containing serous components. Eight MSI-H tumors of grade 1–2 endometrioid carcinoma were also analyzed.

We conducted exome sequencing (tumor-normal paired) to detect genomic mutations. We subsequently performed mutational signature, chromosomal CN, and CN signature analyses. While the TCGA classification uses chromosome CN alterations as an indicator of chromosome instability, given the lower sensitivity for detecting CN alterations in samples with low tumor content and the inherent ambiguity in defining chromosome instability, we opted to delineate genomic subtypes following the procedure outlined in Fig. 1A. Initially, tumors with POLE exonuclease domain mutations (n = 12, 14.8%) were categorized as POLE type. Tumors not classified as POLE type were classified as MSI-H type (n = 14) on the basis of MSI analysis or the MSI score. Tumors not already classified as POLE or MSI-H type that had TP53 mutations or MDM2 amplifications were designated as TP53 type (n = 35). The remaining tumors were classified as having no specific molecular profile (NSMP, n = 20). We also assigned pathological subtypes according to POLE genotype and histological staining results for p53, PMS2, and MSH6 (Fig. 1a). Furthermore, we determined the TCGA subtype according to POLE genotype and CN status; tumors with CN alteration regions with ≥ 1 Gbp were defined as CN-H (Fig. 1a).

Correlation between histological type and genomic abnormalities in high-grade EC

At first, we sought to identify the genomic abnormalities that underlie each histological type of high-grade EC. The MSI-H type comprised endometrioid carcinomas or mixed types containing endometrioid components (including two carcinosarcomas), except for one serous carcinoma (Fig. 1b). POLE-type tumors were predominantly composed of endometrioid carcinomas but also included one serous carcinoma, two clear cell carcinomas, and two carcinosarcomas. Among the 16 serous carcinomas, including 2 mixed-type tumors with clear cell carcinoma components, 13 tumors (81%) carried TP53 mutations (Figs. 1b and 1c). In general, MSI-H or MMRd was associated with endometrioid carcinomas, and TP53 mutations were associated with serous carcinomas. Only a small proportion of clear cell carcinomas without endometrioid components exhibited MSI-H or MMRd (Fig. 1d). While some degree of correlation was observed, no clear relationship between histological type and genomic subtype was evident among high-grade ECs.

Mutational profile of high-grade endometrial carcinomas

The median number of mutations per Mbp was 169 for POLE tumors, 20.7 for MSI-H tumors, 1.27 for TP53 tumors, 1.27 for NSMP tumors, and 18.5 for grade 1–2 MSI-H tumors (Fig. 2a). Mutations in PTEN and ARID1A were frequently observed in POLE and MSI-H tumors (Figs. 2b, 2c). Germline mutations in MMR genes were identified in three MSI-H tumors (one in MLH1 and two in MSH6). Amplifications of CCNE1, ERBB2, ERBB3, FGFR1, and FGFR3 were predominantly identified in TP53 tumors. Although somatic mutations in BRCA1 and BRCA2 were frequently observed (12% and 16%, respectively, Table S2), most of these mutations were found in POLE or MSI-H tumors, suggesting that their contribution to chromosomal instability is likely limited. Germline mutations in BRCA1/2 genes were not identified. In contrast to previous result², KRAS mutations were more common in NSMP tumors (30%) than in the other tumor types (4.9%) (Figs. 2b, 2c).

Analysis of base substitution patterns revealed that the POLE type is characterized by the dominant POLE signature. In POLE tumors, mutations in mismatch repair (MMR) genes were sporadically observed; those tumors have both the POLE signature and the MMRd signature (Figs. 2a, 3a). It is likely that the accumulation of gene mutations due to POLE mutations affects MMR genes, leading to the acquisition of MSI-H characteristics. By contrast, MLH1 methylation was exclusively observed in MSI-H tumors (Fig. 2b). It is likely that POLE tumors become MMRd at relatively later stages of cancer development, with potentially different MMRd status than those observed in MSI-H tumors that occur in the initial phase of cancer development. Of note, two possible POLE tumors, which carried mutations in POLD1 and the POLD1 mutational signature, were by definition categorized as MSI-H tumors (Figs. 2a, 3a). The most predominant mutational signature in TP53 and NSMP tumors was the Age signature, followed by the apolipoprotein B mRNA editing enzyme, catalytic polypeptide (APOBEC) signature (Figs. 2a, 3a).

The median length of CN gain regions was 0.0 Kbp for POLE tumors, 117 Mbp for MSI-H tumors, 2.12 Gbp for TP53 tumors, 844 Mbp for NSMP tumors, and 212 Mbp for grade 1–2 MSI-H tumors (Figs. 2a, 3b, and 3c). The pattern of CN abnormalities²¹ revealed that TP53 tumors had a notable prevalence of the CN signature related to homologous recombination deficiency (HRD), especially CN signature 18 (CN18) (Figs. 2a, 3d). The frequency of CN17, typically associated with HRD tumors in high-grade serous ovarian cancer and triple-negative breast cancer²¹, was not notably high (Fig. 3d). Given reported correlations between CN18 or CN19 and homologous recombination repair defects coupled with two genome doubling events, high-level accumulation of chromosomal abnormalities associated with TP53 mutations in high-grade EC might contribute to the CN18 signature.

The association between TP53 mutations and CN abnormalities is not always apparent. The classification of NSMP or CN-L tumors is inherently challenging. However, some differences were observed, such as a higher frequency of KRAS mutations and fewer CN amplified genes in NSMP tumors compared to TP53 tumors (Figs. 2b, 2c, 3c). The scarcity of mutations in CTNNB1, which were reported to be common in NSMP and CN-L tumors²², might be associated with the high-grade nature of tumors in this study.

In summary, the overall mutation profile aligned well with previous studies^2,9,22,23. However, differences in mutation frequencies compared to the TCGA dataset, which might reflect differences in pathological status and racial background, as well as the predominance of copy number signatures associated with HRD status in the TP53 tumors, highlight the need for further investigation.

Clustering of tumors based on the global gene expression profile

To understand the complicated relationships between molecular and histopathological types in high-grade EC, we conducted an analysis of gene expression profiles. We chose 5,000 genes with notably higher expression levels and significant variability among samples for clustering. In the principal component analysis, EMG3s, carcinosarcomas, and clear cell carcinomas had lax clusters, suggesting a potential tendency to form clusters, while serous carcinomas had greater dispersion (Fig. 4a). To detect more distinct clusters, dimensionality reduction was performed using t-distributed stochastic neighbor embedding (t-SNE), resulting in one cluster of normal tissues and three clusters of tumor tissues (Fig. 4b). Clusters 1, 2, and 3 were enriched for clear cell carcinoma, EMG3, and carcinosarcoma, respectively. Serous carcinoma was distributed across all clusters (Fig. 4c).

To understand the biological implication of the clusters, we examined gene expression patterns in detail. Cluster 1 had high expression of genes that are markers for glandular/lumenal cells of the endometrium, such as WNT7A, HNF1B, and FOSL2 (Figs. 4d, S1) ^18,19. By contrast, Cluster 2 had elevated expression of marker genes for ciliated cells, including RFX3 and TPPP3, along with ESR1 and PGR (Figs. 4d, S1). These results suggest the potential for classifying or stratifying ECs based on the expression of genes involved in endometrial differentiation in addition to estrogen responsiveness. In Cluster 3, high expression of Hedgehog signaling genes GLI1 and GLI2, as well as PRL, was observed. Elevated expression of mesenchymal-associated genes like TWIST1, ZEB1, and CDH2, consistent with the mesenchymal characteristics of carcinosarcoma, was also noted (Figure S1).

In line with the above observations, Cluster 3 possessed basal-like or stem-like cell characteristics according to gene set enrichment analysis (GSEA) (Fig. 4e). Conversely, Cluster 1 was characterized by high activity in the TNF or NF-κB pathways, suggesting that Cluster 1 tumors have a more active tumor immune response than Cluster 2 or 3 tumors, predicting a higher efficacy of immune checkpoint inhibitor (ICI) therapy. Of note, MSI-H type is distributed across Clusters 1, 2, and 3. While response to ICIs can be predicted on the basis of differences in MSI status, tumor characteristics reflected in this clustering might also serve as additional predictor of responsiveness to ICIs. Differences in the expression of marker genes were observed for each histological subtype, with clear cell carcinoma and serous carcinoma showing a glandular/lumenal phenotype and EMG3 exhibiting a tendency toward a ciliated phenotype or estrogen responsiveness (Figs. 4f, S2). High expression of genes associated with the glandular/lumenal phenotype was more pronounced in clear cell carcinoma than in serous carcinoma.

Taken together, these results suggest the potential usefulness of gene expression levels associated with the differentiation state or phenotype of endometrial cells for stratifying and classifying high-grade EC. This led to the speculation that there might be differences in the cell of origin or differentiation states among EMG3, clear cell carcinoma, and serous carcinoma.

Validation of clustering with selected genes and the TCGA dataset

We conducted another round of clustering to classify high-grade EC using genes associated with endometrial cell differentiation. According to reported single-cell RNA sequencing analysis^18,19 and initial clustering results, we selected 96 relevant genes (Table S3). We identified one normal tissue cluster and three tumor clusters, with some discordance from the initial clustering of tumors based on 5,000 genes (Figure S3). This consistency supports using differentiation-related genes for high-grade EC stratification.

To further validate our analysis, we analyzed TCGA data obtained from University of California, Santa Cruz (UCSC) Xena²⁴. Clustering via t-SNE revealed three tumor clusters and one normal cluster (Figure S4a). Similar to the clustering of high-grade EC in this study, Cluster T1 was characterized by high expression of genes associated with glandular/lumenal cells. Cluster T2 was characterized by high expression of genes associated with ciliated cells (Figures S4, S5). Cluster T3 was mainly composed of carcinosarcomas. Cluster T1 was mainly composed of serous carcinomas, possibly due to the lack of clear cell carcinomas in the TCGA dataset. Despite this difference, which is discussed in detail in the supplementary materials, the results indicate that marker gene–based classification is also feasible within the TCGA dataset.

Differentiation signature of endometrial carcinomas

To generalize the clustering procedure described above, we selected 8 genes considered to represent the glandular/lumenal phenotype, 11 genes considered to represent the ciliated phenotype, and 7 genes considered to represent the carcinosarcoma phenotype. We performed ssGSEA to calculate the score of each signature (Table S4). Signature scores consistent with the clustering results were obtained (Fig. 5a). Endometrioid tumors (EMG3s) exhibited a strong ciliated phenotype, clear cell carcinomas exhibited a strong glandular/lumenal phenotype, and serous tumors exhibited a moderate but apparent glandular/lumenal phenotype (Fig. 5b). Similar results were obtained for the TCGA dataset (Figs. 5c, 5d). These findings demonstrated that cell phenotype signatures determined on the basis of gene expression profiles are strongly associated with histological classification.

Factors associated with infiltration of CD8 + T cells

Next, we explored the immunological implications of the findings described above. In comparison to microsatellite stable (MSS) tumors, there was more CD8 + T cell infiltration in MSI-H tumors, aligning with previous reports on MSI-H colorectal cancer or EC^25–27 (Fig. 6a). However, there were no apparent differences in CD8 + T cell infiltration among clusters or histological types, which is also the case with the TCGA dataset (Figs. 6b, 6c). The expression profile of programmed cell death ligand 1 (PD-L1) as evaluated with IHC analysis was similar to that of CD8, which was caused by the expected positive correlation between PD-L1 and CD8 (Fig. 6d). This is partly because of the higher CD8 + T cell infiltration observed in a subset of Cluster 1 tumors with a glandular/lumenal phenotype. While this finding appears contradictory to the abundance of EMG3 in Cluster 2 (tumors with a ciliated phenotype), which is enriched for MSI-H and POLE tumors, the results were consistent with the analysis of the TCGA dataset. There was higher expression of HLA-A and HLA-B, which are essential for antigen presentation in Cluster T1 (Fig. 6e). Consistent with the observation in the TCGA dataset, we observed enhanced expression of HLA-A and HLA-B in Cluster 1 even when MSI tumors, which have abundant lymphoid cells, were excluded (Fig. 6f). Increased antigen presentation capacity in Cluster 1 tumors with a glandular/lumenal phenotype might have resulted in CD8 + T cell infiltration levels comparable to those in Cluster 2, which include MSI-H tumors that are known to carry many tumor-specific antigens. Interestingly, expression of HLA-A and HLA-B was positively correlated with a glandular/lumenal score calculated above (Figs. 6g, 6h), suggesting the interplay among cell phenotype, histological classification, and immunological status. This positive association was more pronounced when analysis was restricted to MSS tumors (Fig. 6i).

Consistent with prior studies¹⁰, ARID1A mutations in MSI-H tumors were associated with higher CD8 + T cell infiltration (Fig. 6j). Although ARID1A impairment has been reported to induce mismatch repair deficiency¹¹, in this cohort, no alterations in MSI score, tumor mutational burden, or the number of indels were observed due to the loss of ARID1A function. Expression of HLA-A and HLA-B and glandular/lumenal score were higher in ARID1A-defective MSI-H tumors than in ARID1A-intact MSI-H tumors, although the difference was not statistically significant (Fig. 6j). These findings suggest the presence of factors influencing immune response in endometrial carcinomas other than MMRd; ARID1A and cellular phenotype might play a role in modulating CD8 + T cell infiltration.

In this study, we conducted genome and histological analysis to gain insights into the molecular basis of high-grade EC. Several observations deepened our understanding. First, a strong association between TP53 type and serous carcinoma was found, consistent with previous studies^2,8. We noticed that TP53 tumors in this cohort of high-grade EC had distinct chromosomal copy number abnormalities characterized by multiple rounds of genome doubling events. Notably, the frequency of CN17, characteristic of BRCA1/2 loss of function, was relatively low. The mutational signature SBS3, characteristic of BRCA1/2 loss of function, was not detected. This aligns with prior research indicating that SBS3 is observed only in a minority of ECs^28,29. These findings suggest a molecular basis for chromosomal instability distinct from HRD derived from BRCA1/2 loss of function, as seen in high-grade serous ovarian cancer and triple-negative breast cancer^30,31. The high frequency of CCNE1 amplification and low frequency of BRCA1/2 mutations also support the abundance of homologous recombination proficiency in high-grade ECs.

A relatively high proportion of our cohort had POLE mutations. This finding aligns with previous observations of POLE mutation enrichment in high-grade EC^22,32,33. While the possibility of a higher frequency of POLE mutations in the Japanese population cannot be excluded²³, the current tumor cohort is not suitable for formal comparisons because the samples were selected to encompass a sufficient number of distinct, representative high-grade histological types. On the other hand, the high proportions of TP53 or CN-H tumors, along with low proportions of NSMP or CN-L tumors, are considered reflective of the high-grade nature of this cohort. Consistent with prior reports, we reaffirmed the lack of a clear relationship between genomic subtype and histological classification.

In this study, we identified molecular bases for histological classifications through gene expression analysis. Consequently, it became evident that gene expression states related to the differentiation of endometrial cells are strongly correlated with histological type. Thus, for a detailed understanding of high-grade EC, it is crucial to consider the cell of origin or differentiation state, as reflected in gene expression profiles. Specifically, serous carcinoma and clear cell carcinoma exhibit strong glandular/lumenal traits while EGM3 shows prominent ciliated traits.

Furthermore, we inferred immunological status by evaluating the expression of genes associated with immunoreactivity, including HLA genes. In serous carcinoma or Cluster 1 tumors characterized by glandular/lumenal features, we observed enhanced activity of immune response–related signaling pathways, such as the TNF and NFKB pathways, along with increased expression of antigen-presenting HLA-A and HLA-B. Higher expression of HLA-A and HLA-B is expected to result in effective antigen presentation and anti-tumor immune response³⁴. Therefore, in EC, an abundance of tumor-specific antigens does not necessarily correlate with the strength of immune responses or antigen presentation. In accordance with previous reports^27,35,36, it seems plausible that there could be tumors with many tumor-specific antigens but low antigen presentation function, or tumors with relatively few tumor-specific antigens but high antigen presentation function. Importantly, our data suggested that HLA-A and HLA-B expression might be closely related to the glandular/lumenal characteristics of tumors. The immune responsiveness of tumors with ARID1A mutations has attracted attention in recent years^10,11,37. ARID1A mutations in epithelial cells were reported to be involved in the upregulation of pro-angiogenic and pro-lymphangiogenic factors and enrichment of lymphatic endothelial cells in endometriosis¹⁷. In this study, CD8 + T cell infiltration was higher in ARID1A-lost tumors compared to ARID1A-retained tumors in both MSS and MSI tumors, which is contradictory to a previous report³⁷. This discrepancy may be due to differences in the composition of histological types or the methods for accessing CD8 + T cell infiltration. These findings underscore the need for a more detailed understanding of the relationship between immunological status, genomic status, and cellular phenotypes of ECs for optimal use of immunotherapies such as ICIs.

Our study has several limitations. First, we did not address the efficacy of ICIs. Second, the analyzed cohort was biased toward advancing pathophysiological understanding of various histological classifications, limiting the assessment of clinical implications. To translate the results of this study into clinical practice, it is necessary to proceed with research in more clinically oriented settings focused on each disease entity. Finally, cellular origin cannot be clarified on the basis of cell marker expression alone. Methylome analysis with both tumor and normal epithelial cells of various types in the endometrium might be helpful.

In conclusion, we found that RNA clustering is a valuable tool for the precise stratification and understating of the histology-related biology of high-grade EC, suggesting its potential to enhance personalized medicine in EC, along with molecular or genomic subtype.

Patients and pathological review

This study was conducted in accordance with the Declaration of Helsinki and received approval from the ethics committee of the Chiba Cancer Center (approval number: M04-001). Patients with high-grade EC who underwent surgical resection at Juntendo University from 2006 to 2022 (14 patients, approval numbers: M09-0551, M20-0007), The University of Tokyo Hospital from 2010 to 2022 (51 patients, including 8 patients with grade 1–2 endometrioid carcinomas, approval numbers: G0683, 2022083Ge), or National Cancer Center Hospital from 2012 to 2015 (24 patients, approval number: 2022-003) were enrolled in this study. Written informed consent for research use of the samples and clinical information was obtained from all participants. Tissue samples from the patients at the National Cancer Center Hospital were provided by the National Cancer Center Biobank, Japan.

Pathological diagnoses were initially made according to the 2020 WHO classification²⁰ (https://tumourclassification.iarc.who.int/welcome/) by at least two pathologists at each institution. Subsequently, a gynecologic pathologist (H.Y.) conducted central review to confirm histological type. Only tumors diagnosed as high-grade EC such as serous carcinoma, clear cell carcinoma, EMG3, undifferentiated/dedifferentiated carcinoma, carcinosarcoma, or small cell carcinoma were included in the study. Representative histological images are shown in Supplementary Figure S6. All pathologists were blinded to genomic analysis results and prognostic information. All specimens resected during surgery were appropriately fixed in 10% neutral buffered formalin for 12–48 hours immediately after the collection of frozen tissue samples. Frozen tumor samples were used for genomic analysis. Formalin-fixed paraffin-embedded sections, including representative tumor areas, were provided for IHC analysis. Clinicopathological data for each patient were retrospectively obtained from medical records.

Whole-exome sequencing

Tumor tissues were collected immediately after surgery, cut into small pieces, and stored at − 80°C until use. Genomic DNA was extracted from surgically resected specimens using the QIAamp Fast DNA Tissue Kit (Qiagen, Venlo, Limburg, Netherlands). Peripheral blood samples were stored at − 30°C. Genomic DNA was extracted from peripheral blood samples using the QIAamp DNA Blood Midi Kit (Qiagen). Whole-exome sequencing (WES) libraries were generated using the Twist Library Preparation EF system (Twist Bioscience, South San Francisco, CA) with enzymatic fragmentation, the Twist Universal Adaptor system, and the Twist Fast Hybridization Target Enrichment system. Briefly, 50 ng of genomic DNA was enzymatically fragmented to 200–300 bp, followed by end-repair, A-tailing, and pair-end index adaptor ligation. Pre-capture amplification was conducted with KAPA HiFi DNA polymerase (KAPA Biosystems, Roche Diagnostics, Indianapolis, IN, USA). Exonic fragments from approximately 750 ng of amplified products were enriched using the Twist Comprehensive Exome Panel as a probe. Massively parallel, paired-end sequencing of sample libraries was performed with a NovaSeq6000 sequencer (Illumina, San Diego, CA, USA).

Analysis of sequence data

Reads from paired-end WES were aligned to the human reference genome (hg38) using the Burrows-Wheeler Aligner (http://bio-bwa.sourceforge.net/) and Bowtie 2 (http://bowtie-bio.sourceforge.net/bowtie2/index.shtml). Somatic (synonymous and non-synonymous) mutations were called using an in-house caller and publicly available mutation callers: Genome Analysis Toolkit (https://gatk.broadinstitute.org/hc/en-us), MuTect2 (https://gatk.broadinstitute.org/hc/en-us/articles/360037593851-Mutect2), and VarScan 2 (http://varscan.sourceforge.net/). Mutations were discarded if any of the following criteria were met: total read number < 20, variant allele frequency (VAF) in tumor samples < 0.05, mutant read number in germline control samples > 2, mutation in only one genome strand, or variant present in normal human genomes of the 1,000 Genomes Project dataset (https://www.internationalgenome.org/) or the in-house database. Gene mutations were annotated using SnpEff (http://snpeff.sourceforge.net). CN was analyzed using an in-house pipeline that determined the log R ratio (LRR) as follows: (I) homozygous (VAF ≤ 0.05 or ≥ 0.95) or heterozygous (VAF 0.4–0.6) single-nucleotide polymorphisms (SNPs) were selected from the genomes of the related normal samples in the 1,000 Genomes Project database; (II) normal and tumor read depths for the selected SNPs were adjusted according to the G+C percentage of a 100-bp

Analysis of mutational signatures

Mutations with VAF ≥ 0.1 that passed the quality filter were subjected to mutational signature analysis with SigProfilerExtractor³⁹ (https://github.com/AlexandrovLab/SigProfilerExtractor). To summarize results, we aggregated single base substitution (SBS) categories into broader categories. The age category was derived by summing the values from SBS1 (Age) and SBS5 (Age). The APOBEC category combined values from SBS2 (APOBEC) and SBS13 (APOBEC). Similarly, the POLE category was created by summing values from SBS10a (POLE), SBS10b (POLE), SBS14 (POLE), and SBS20 (POLD1). The MMRd category was based on values in SBS15 (MMRd). SBS87 was used as is. According to the description provided on the Catalogue Of Somatic Mutations In Cancer (COSMIC) website (https://cancer.sanger.ac.uk/signatures/), we also aggregated CN signature categories into broader categories. The HRD-TandemDup category was derived by summing the values from CN17 (HRD), CN18 (unknown), CN19 (unknown), and CN20 (unknown).

Transcriptome sequencing

Ribosomal RNA depletion with NEBNext® rRNA Depletion Kit v2 (New England Biolabs, Ipswich, MA, USA) was performed on 500 ng of RNA extracted from clinical samples. Sequencing libraries were prepared using the NEBNext® Ultra™ II RNA Library Prep Kit for Illumina® (New England Biolabs) and sequenced over 150 bp from both ends using the NovaSeq6000 sequencer (Illumina). The expression level of each gene was calculated using DESeq2 (https://bioconductor.org/packages/release/bioc/html/DESeq2.html). Gene set enrichment analysis and single sample GSEA were performed with the GSEAPY package, version 1.0.6 ⁴⁰ (https://gseapy.readthedocs.io/en/latest/gseapy_example.html#Prerank-example).

Microsatellite instability testing and detection of MLH1 promoter methylation

Using labeled primers, microsatellite instability was analyzed with polymerase chain reaction (PCR) at five microsatellite loci: BAT25, BAT26, NR-21, NR-24, and MONO-27. MLH1 promoter methylation was assessed with PCR after digestion of genomic DNA with a methylation-sensitive restriction enzyme as previously described⁴¹, with slight modifications. Genomic DNA (200 ng) was digested in a volume of 20 µL reaction buffer by HinP1I (Nippon Genetics, Tokyo, Japan), followed by heat inactivation of restriction enzymes. Digested DNA (20 ng) was subjected to 25 cycles of multiplex PCR in a total volume of 10 µL reaction buffer using the Type-it Microsatellite PCR Kit (Qiagen). PCR products were electrophoresed on a 3730xl DNA Analyzer (Applied Biosystems, Thermo Fisher Scientific, Waltham, MA, USA) and analyzed using Peak Scanner 2 software (Applied Biosystems).

TCGA data analysis

We obtained batch effect normalized RNA-seq data for uterine corpus endometrial carcinoma pan-cancer and uterine carcinosarcoma pan-cancer from UCSC Xena²⁴. Metadata, such as histological type and subtype information, were acquired from both UCSC Xena and cBioPortal (https://www.cbioportal.org/). The analysis was conducted using 595 samples (including adjacent normal tissues) and 17,507 genes (with no missing values) for which gene expression data and metadata were available.

Immunohistochemical analysis and interpretation

All IHC tests, including those for p53 (DO7, prediluted, Dako, Glostrup, Denmark), ARID1A (HPA005456, rabbit polyclonal, 1:2000, Sigma-Aldrich, St. Louis, MO, USA), PMS2 (EP51, prediluted, Dako/Agilent), MSH6 (EP49, prediluted, Dako/Agilent), estrogen receptor (ER) (SP1, prediluted, Ventana Medical Systems, Tucson, AZ, USA), progesterone receptor (PR) (1E2, prediluted, Ventana Medical Systems), CD8 (4B11, 1:200, Dako), and PD-L1 (E1L3N, 1:400, Cell Signaling Technology, Danvers, MA, USA) were performed as described in our previous study^23,42. IHC testing for p53, ARID1A, PMS2, MSH6, CD8, and PD-L1 was performed with the Link 48 autostainer (Dako/Agilent). IHC testing for ER and PR was performed using the BenchMark ULTRA autostainer (Ventana/Roche Diagnostics). After deparaffinization, tissue sections were stained using the antibodies mentioned above and then counterstained with hematoxylin.

Pathologists evaluated all IHC slides according to the following criteria. For p53 staining, tumors with either a strong and diffuse nuclear staining pattern (> 80% of carcinoma cells) or a staining pattern entirely negative of carcinoma cells (null pattern) were considered to have aberrant p53 staining pattern indicating TP53 alterations. Staining of the adjacent nontumor cells served as an internal positive control. Tumors with weak or heterogeneous staining patterns were considered to have wild-type TP53. In addition, subclonal mutant p53 immunostaining was defined as a combination of wild-type patterns and one or more mutant patterns, with each present in at least 5% of tumor cells⁴³. The loss of ARID1A nuclear expression in tumor cells was classified as either homogeneous (negative in almost all tumor cells) or focal (regionally negative)⁴². Because IHC for PMS2 and MSH6 can reportedly be used instead of the four-antibody panel (MLH1, MSH2, MSH6, and PMS2) for MMRd screening⁵, MMRd status was defined as the complete loss of nuclear staining for PMS2 or MSH6 proteins in this study. Internal positive controls with intact nuclear staining included the adjacent normal mucosa, stromal cells, and inflammatory cells. The proportion of positive tumor cells with nuclear staining, regardless of intensity, was measured for ER and PR expression. CD8 + T cells were analyzed using HALO software (Indica Lab, Corrales, NM, USA). Specifically, the tumor area was manually annotated by a pathologist (S.S.), the number of CD8 + T cells and their area within that region were measured, and the number of CD8 + T cells per unit area was calculated. PD-L1 expression was assessed by calculating the combined positive score (CPS)⁴⁴. CPS was calculated by dividing the number of PD-L1–positive cells (including viable tumor cells, lymphocytes, and macrophages) by the total number of viable tumor cells and multiplying by 100. Representative IHC staining results are shown in Supplementary Figure S7.

Statistical analyses

Python (version 3.11.4) and SciPy (version 1.11.1) were used for statistical analyses. Comparisons of quantitative variables between groups were conducted using the nonparametric Wilcoxon signed-rank test or analysis of variance (ANOVA) followed by Tukey’s post hoc test. Correlations were assessed with Pearson’s correlation coefficients. P values < 0.05 were considered statistically significant.

Data availability

Raw sequencing data will be deposited in the Japanese Genotype-phenotype Archive hosted by the DNA Data Bank of Japan⁴⁵ under the accession number JGAS00000000NNN (In the process of submission, NBDC number: J-DS000878).

Acknowledgements

We would like to thank Saori Sugaya, Reina Takeyama, and Noriko Sakurai for technical support. We would like to thank Zenis (https://www.zenis.co.jp/index.html) for English language editing. National Cancer Center Biobank is supported by the National Cancer Center Research and Development Fund, Japan. This study was supported by Grants-in-Aid for Scientific Research [21H02772 (M. Kawazu), 22K09566 (A.T.), 21K15072 (J.L.), and 24K02584 (K.O.)], grants from the Japan Agency for Medical Research and Development (AMED) [JP21cm0106502 (M. Kawazu), JP22ck0106723 (M. Kawazu), JP22ck0106724 (M. Kawazu), JP23ama221528 (M. Kawazu), JP22zf0127009 (H.M.), JP22wm0325014 (A.T.), and JP23wm0325057 (A. T.)], a Chiba Prefecture Research Grant (M. Kawazu), a Seiichi Imai Memorial Foundation Research Grant (A.T.), a grant from the Taiju Life Social Welfare Foundation (J.L.), a grant from the Takeda Science Foundation (M. Kawazu), a grant from the Princess Takamatsu Cancer Research Fund (M. Kawazu and S.I.), a grant from the Uehara Memorial Foundation (M. Kawazu), a grant from the NOVARTIS Foundation (Japan) for the Promotion of Science (M. Kawazu and S.I.), a grant from the Kobayashi Foundation for Cancer Research (M. Kawazu), a grant from the Kashiwado Memorial Foundation (M. Kawazu), a grant from the Kondou Kinen Medical Foundation (S.I.), a grant from the Toyoaki Scholarship Foundation (S.I.), and a grant from the Hirose Foundation (S.I.).

Conflict of interests statement

M. Kawazu received research honoraria from Pacific Biosciences unrelated to this study. J.L. received a research grant from the NVIDIA Corporation unrelated to this study. H.M. serves as a board member of CureGene, and received research grants from Asahi Kasei, Gilead/Kite, Konica-Minolta, and Ambry unrelated to this study. K.O. received research honoraria from Konica Minolta, Co., Ltd., and AstraZeneca K.K., and lecture fees from Konica Minolta, Co. Ltd, Chugai Pharmaceutical Co., Ltd., and AstraZeneca K.K. unrelated to this study. All other authors declare that they have no competing financial interests.

Crosbie, E. J. et al. Endometrial cancer. Lancet 399, 1412–1428 (2022).
Levine, D. A. Integrated genomic characterization of endometrial carcinoma. Nature 497, 67–73 (2013).
Talhouk, A. et al. A clinically applicable molecular-based classification for endometrial cancers. Br J Cancer 113, 299–310 (2015).
Kommoss, S. et al. Final validation of the ProMisE molecular classifier for endometrial carcinoma in a large population-based case series. Ann Oncol 29, 1180–1188 (2018).
Hall, G. et al. Immunohistochemistry for PMS2 and MSH6 alone can replace a four antibody panel for mismatch repair deficiency screening in colorectal adenocarcinoma. Pathology 42, 409–413 (2010).
DeLair, D. F. et al. The genetic landscape of endometrial clear cell carcinomas. J Pathol 243, 230–241 (2017).
Gotoh, O. et al. Clinically relevant molecular subtypes and genomic alteration-independent differentiation in gynecologic carcinosarcoma. Nat Commun 10, 4965 (2019).
Bogani, G. et al. Uterine serous carcinoma. Gynecol Oncol 162, 226–234 (2021).
Momeni-Boroujeni, A. et al. Genomic landscape of endometrial carcinomas of no specific molecular profile. Modern Pathology 35, 1269–1278 (2022).
Sun, D. et al. ARID1A deficiency associated with MMR deficiency and a high abundance of tumor-infiltrating lymphocytes predicts a good prognosis of endometrial carcinoma. Transl Oncol 33, 101685 (2023).
Shen, J. et al. ARID1A deficiency promotes mutability and potentiates therapeutic antitumor immunity unleashed by immune checkpoint blockade. Nat Med 24, 556–562 (2018).
Dessources, K. et al. Impact of immune infiltration signatures on prognosis in endometrial carcinoma is dependent on the underlying molecular subtype. Gynecol Oncol 171, 15–22 (2023).
Eskander, R. N. et al. Pembrolizumab plus Chemotherapy in Advanced Endometrial Cancer. N Engl J Med 388, 2159–2170 (2023).
Mirza, M. R. et al. Dostarlimab for Primary Advanced or Recurrent Endometrial Cancer. N Engl J Med 388, 2145–2158 (2023).
Tsolova, A. O., Jain, V. & Critchley, H. O. D. Mapping the non-pregnant uterus cell-by-cell. Dev Cell 57, 421–423 (2022).
Tan, Y. et al. Single-cell analysis of endometriosis reveals a coordinated transcriptional programme driving immunotolerance and angiogenesis across eutopic and ectopic tissues. Nat Cell Biol 24, 1306–1318 (2022).
Fonseca, M. A. S. et al. Single-cell transcriptomic analysis of endometriosis. Nat Genet 55, 255–267 (2023).
Garcia-Alonso, L. et al. Mapping the temporal and spatial dynamics of the human endometrium in vivo and in vitro. Nat Genet 53, 1698–1711 (2021).
Wang, W. et al. Single-cell transcriptomic atlas of the human endometrium during the menstrual cycle. Nat Med 26, 1644–1653 (2020).
WHO Classification of Tumours Editorial Board. Female Genital Tumours WHO Classification of Tumours, 5th Edition, Volume 4. (2020).
Steele, C. D. et al. Signatures of copy number alterations in human cancer. Nature 606, 984–991 (2022).
Stelloo, E. et al. Improved Risk Assessment by Integrating Molecular and Clinicopathological Factors in Early-stage Endometrial Cancer—Combined Analysis of the PORTEC Cohorts. Clinical Cancer Research 22, 4215–4224 (2016).
Kobayashi Kato, M. et al. Clinical impact of genetic alterations of CTNNB1 in patients with grade 3 endometrial endometrioid carcinoma. Cancer Sci 113, 1712–1721 (2022).
Goldman, M. J. et al. Visualizing and interpreting cancer genomics data via the Xena platform. Nat Biotechnol 38, 675–678 (2020).
Kawazu, M. et al. HLA Class I Analysis Provides Insight Into the Genetic and Epigenetic Background of Immune Evasion in Colorectal Cancer With High Microsatellite Instability. Gastroenterology 162, 799–812 (2022).
Howitt, B. E. et al. Association of Polymerase e–Mutated and Microsatellite-Instable Endometrial Cancers With Neoantigen Load, Number of Tumor-Infiltrating Lymphocytes, and Expression of PD-1 and PD-L1. JAMA Oncol 1, 1319 (2015).
Talhouk, A. et al. Molecular Subtype Not Immune Response Drives Outcomes in Endometrial Carcinoma. Clinical Cancer Research 25, 2537–2548 (2019).
de Jonge, M. M. et al. Frequent Homologous Recombination Deficiency in High-grade Endometrial Carcinomas. Clinical Cancer Research 25, 1087–1097 (2019).
Ashley, C. W. et al. Analysis of mutational signatures in primary and metastatic endometrial cancer reveals distinct patterns of DNA repair defects and shifts during tumor progression. Gynecol Oncol 152, 11–19 (2019).
Cancer Genome Atlas Network. Comprehensive molecular portraits of human breast tumours. Nature 490, 61–70 (2012).
Cancer Genome Atlas Research Network. Integrated genomic analyses of ovarian carcinoma. Nature 474, 609–15 (2011).
Church, D. N. et al. Prognostic Significance of POLE Proofreading Mutations in Endometrial Cancer. JNCI: Journal of the National Cancer Institute 107, (2015).
McConechy, M. K. et al. Endometrial Carcinomas with POLE Exonuclease Domain Mutations Have a Favorable Prognosis. Clinical Cancer Research 22, 2865–2873 (2016).
Kawase, K. et al. High Expression of MHC Class I Overcomes Cancer Immunotherapy Resistance Due to IFNγ Signaling Pathway Defects. Cancer Immunol Res 11, 895–908 (2023).
Dou, Y. et al. Proteogenomic Characterization of Endometrial Carcinoma. Cell 180, 729-748.e26 (2020).
Palomero, J. et al. Biomarkers of tumor-reactive CD4 ⁺ and CD8 ⁺ TILs associate with improved prognosis in endometrial cancer. J Immunother Cancer 10, e005443 (2022).
Yamashita, H. et al. Evaluation of ARID1A as a Potential Biomarker for Predicting Response to Immune Checkpoint Inhibitors in Patients with Endometrial Cancer. Cancers (Basel) 16, 1999 (2024).
Shen, R. & Seshan, V. E. FACETS: allele-specific copy number and clonal heterogeneity analysis tool for high-throughput DNA sequencing. Nucleic Acids Res 44, e131–e131 (2016).
Islam, S. M. A. et al. Uncovering novel mutational signatures by de novo extraction with SigProfilerExtractor. Cell Genomics 2, 100179 (2022).
Fang, Z., Liu, X. & Peltz, G. GSEApy: a comprehensive package for performing gene set enrichment analysis in Python. Bioinformatics 39, (2023).
Sato, K. et al. Fusion Kinases Identified by Genomic Analyses of Sporadic Microsatellite Instability–High Colorectal Cancers. Clinical Cancer Research 25, 378–389 (2019).
Kato, M. K. et al. Loss of ARID1A Expression as a Favorable Prognostic Factor in Early-Stage Grade 3 Endometrioid Endometrial Carcinoma Patients. Pathology and Oncology Research 27, (2021).
Singh, N. et al. p53 immunohistochemistry is an accurate surrogate for TP53 mutational analysis in endometrial carcinoma biopsies. J Pathol 250, 336–345 (2020).
Chiba, Y. et al. Trop-2 expression and the tumor immune microenvironment in cervical cancer. Gynecol Oncol 187, 51–57 (2024).
Kodama, Y. et al. The DDBJ Japanese Genotype-phenotype Archive for genetic and phenotypic human data. Nucleic Acids Res 43, D18–D22 (2015).

Yes there is potential Competing Interest. M.K. received research honoraria from Pacific Biosciences unrelated to this study. J.L. received a research grant from the NVIDIA Corporation unrelated to this study. H.M. serves as a board member of CureGene, and received research grants from Asahi Kasei, Gilead/Kite, Konica-Minolta, and Ambry unrelated to this study. All other authors declare that they have no competing financial interests.

Kawazusuptablechecked240624.docx
Supplementary dataset 1
Kawazusupchecked240624.docx
Supplementary materials

Download PDF

Version 1

posted

You are reading this latest preprint version

Genomic subtypes and cellular phenotypes of high-grade endometrial carcinoma

Status:

Version 1

Abstract

Figures

Introduction

Results

Classification of high-grade endometrial carcinomas based on the results of exome sequencing

Correlation between histological type and genomic abnormalities in high-grade EC

Mutational profile of high-grade endometrial carcinomas

Clustering of tumors based on the global gene expression profile

Validation of clustering with selected genes and the TCGA dataset

Differentiation signature of endometrial carcinomas

Factors associated with infiltration of CD8 + T cells

Discussion

Methods

Patients and pathological review

Whole-exome sequencing

Analysis of mutational signatures

Transcriptome sequencing

TCGA data analysis

Immunohistochemical analysis and interpretation

Statistical analyses

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1