Study design and patient cohorts
Cohort 1: Retrospective analysis regarding survival outcomes
We retrospectively analyzed the cases of a separate consecutive cohort of patients with high-grade EC (n = 85, including endometrioid grade 3, serous, and clear cell histology) treated between January 2002 and December 2017 at Nagoya University Hospital (Nagoya, Japan). Formalin-fixed paraffin-embedded (FFPE) tumor tissues were collected for analysis.
Cohort 2: Pros pective analysis regarding the immunophenotype-molecular classification relationship
The cases of 60 patients with EC treated during the period from January 2019 to December 2022 at Nagoya University Hospital were prospectively enrolled. Following a definitive pathological diagnosis, these patients underwent surgical intervention. Ethical approval of this study was granted by the Institutional Review Board of Nagoya University (IRB approval no.: 2018 − 0400). Both fresh frozen and FFPE tumor tissues were collected for analysis.
Diagnosis and treatment
All 145 of the patients underwent a simple or semi-radical hysterectomy with a bilateral salpingo-oophorectomy, pelvic and/or para-aortic lymphadenectomy. The surgical specimens were staged according to the FIGO 2008 staging system. Postoperative treatment followed the Japan Society of Gynecologic Oncology guidelines [15, 16]. Risk-based adjuvant therapy was administered; low-risk patients received no additional treatment, and the intermediate- to high-risk patients received radiotherapy or platinum-based chemotherapy.
Sanger sequencing for POLE exonuclease domain
Hotspot mutations in exons 9, 13, and 14 of the POLE gene were identified using Sanger sequencing. Genomic DNA extraction was conducted in accord with the manufacturers' protocols for both fresh frozen (NucleoSpin® DNA Rapidlyse, Macherey-Nagel, Düren, Germany) and FFPE tissues (QIAamp DNA FFPE Advanced UNG Kit, Qiagen, Hilden, Germany). The samples were enriched using the Blend Taq Plus (Toyobo, Osaka, Japan). The polymerase chain reaction (PCR) amplification and conditions were as described [17]. A portion of the PCR products was run on a 2% agarose gel in 1× TAE buffer to verify the presence of a single band approx. 200–300 base pairs in size. The rest of the PCR products was then purified using the QIAquick Gel Extraction Kit (Qiagen), following the manufacturer's instructions.
The DNA concentrations were measured with a NanoDrop One spectrophotometer (Thermo Fisher Scientific, Waltham, USA). The subsequent DNA sequencing was outsourced to Eurofins Genomics (Tokyo). For the sequence analysis, we used SnapGene Viewer ver. 6.1.2 (GSL Biotech, San Diego, CA) to examine the waveform patterns. Mutation identification was conducted using the Nucleotide BLAST tool (Basic Local Alignment Search Tool) from the U.S. National Center for Biotechnology Information (NCBI). In this study, we defined 'POLE pathogenic mutations' as the nine mutations on exons 9, 13, and 14: P286R (exon9), M295R (exon9), S297F (exon9), V411L (exon13), L424I (exon13), P436R (exon13), M444K (exon13), A456P (exon14), and S459F (exon14) [18].
Immunohistochemistry analysis
Immunohistochemistry (IHC) was conducted on 4-µm sections of FFPE tumor tissues, which included sections from both the central tumor (CT) and the invasive margin (IM). The primary antibodies used for the IHC included anti-human CD8 (clone C8/144b, Dako, Glostrup, Denmark; 1:100), PMS2 (clone A16-4, Biocare Medical, Walnut Creek, CA; 1:100), MSH6 (clone BC/44, Biocare Medical, 1:100), and p53 (clone DO-7, Dako; 1:100). The sections were deparaffinized and rehydrated, subjected to antigen retrieval in 10 mM sodium citrate (pH 6.0) or 1× Immunoactive (pH 9.0, Matsunami, Osaka, Japan) for 20 min at 95°C in a microwave, and treated with 0.3% hydrogen peroxide in methanol for 20 min after being washed with phosphate-buffered saline (PBS). Blocking was performed using the Histofine SAB-PO kit (Nichirei, Tokyo), followed by overnight incubation at 4°C with the diluted primary antibodies. After the primary antibody incubation, sections were washed in PBS, incubated with biotin-labeled secondary antibody, peroxidase-labeled streptavidin, and developed using 3,3'-diaminobenzidine (DAB) substrate-chromogen for specific time durations: 3 min for p53, 5 min for CD8, and 60 min respectively for PMS2 and MSH6. After DAB development, the sections were rinsed, counterstained with hematoxylin, dehydrated, and mounted.
Our application of immunohistochemistry for PMS2 and MSH6 was based on reports suggesting their effectiveness in screening for mismatch repair deficiency (MMRd) [19]. MMRd was identified by the complete absence of nuclear staining for either protein with internal positive controls including unaltered nuclear staining in adjacent normal endometrium, stromal cells, and inflammatory cells. Abnormal p53 staining (p53abn) was characterized as either a strong, diffuse nuclear staining pattern in > 80% of carcinoma cells, or a complete lack of staining ("null pattern"), using adjacent non-tumor cells as an internal control. Wild-type tumor cells exhibited weak and heterogeneous staining patterns [20].
Immunophenotyping
The assessment of TILs in this study followed the guidelines established by the International Immuno-Oncology Biomarker Working Group [21]. We did not differentiate between intra-tumoral TILs and stromal TILs during this evaluation. After capturing stained slide images with a VS120-S5 (Evident, Tokyo, Japan), CD8+ TILs were quantified automatically in both the CT and IM with the use of QuPath ver. 0.3.0 [22], as illustrated in Fig. 3A,B. This quantification was performed over five distinct areas, each measuring 0.25 mm² (0.0625 mm² per square). The average number of CD8+ TILs per square millimeter was calculated for these regions.
Tumors with a CD8+ TIL density ≥ 1,000 cells/mm² in both the CT and IM regions were classified as 'inflamed' phenotype. The tumors with a CD8+ TIL density < 1,000 cells/mm² in the CT but > 1,000 cells/mm² at the IM were designated as the 'excluded' phenotype. Conversely, tumors were classified as the 'desert' phenotype when the density of CD8+ TILs was < 1,000 cells/mm² in both the CT and IM areas.
The ProMisE molecular classification
We conducted the molecular classification of tumors using an adapted approach from the ProMisE methodology, aligned with the steps outlined in the World Health Organization (WHO) classification [8]. This approach involved a sequential assessment of specific molecular markers. Initially, all tumor samples underwent Sanger sequencing to identify mutations in the POLE exonuclease domain, specifically targeting exons 9, 13, and 14 as noted above. Tumors harboring pathogenic mutations in these regions were classified as 'POLEmut.' Next, the tumors exhibiting a complete absence of nuclear staining for PMS2 or MSH6 protein by IHC were categorized as 'MMRd.' Then, the tumors demonstrating abnormal p53 expression patterns by IHC were classified as 'p53abn.' Finally, tumors that did not exhibit any of the aforementioned molecular characteristics were classified as 'NSMP,' indicating a no specific molecular profile.
Survival analysis
We defined progression-free survival (PFS) as the duration from the initiation of a patient's treatment to the point of observed disease progression. Overall survival (OS) was determined as the time span from the commencement of treatment to either the death of a patient due to any reason or the patient's last confirmed survival status, with data collected up until March 2023. We used the Kaplan–Meier method to estimate the 145 patients' PFS and OS rates. This approach allowed us to generate survival curves, providing a visual and statistical representation of the survival probabilities over time for the patients in Cohort 1. The survival analysis was conducted exclusively for cohort 1 due to limitations in Cohort 2, i.e., the short follow-up period and the low number of recorded events.
RNA-sequencing analysis
The RNA-sequencing analysis was carried out on samples from the 40 of the 60 patients in Cohort 2 from whom RNA samples were available. The extraction of RNA from fresh frozen tumor tissue was done using the NucleoSpin RNA Plus kit (Macherey-Nagel), following the manufacturer's guidelines. We measured the total RNA concentration with the NanoDrop One spectrophotometer. RNA sequencing was then performed by Novogene Japan (Tokyo). The obtained raw FASTQ data were uploaded to Galaxy, an open-source web-based platform tailored for data-intensive biomedical research. Quality control of the data was executed using FastQC and Trimmomatic. The clean, paired-end data were then processed for gene expression quantification using the Kallisto quant tool, referencing the GENCODE GRC38.p13 transcript (genecode. v41.transcript).
Post-processing, the data were aggregated using the tximport package (ver. 1.18.0) in R software (ver. 4.0.3) and RStudio. For the subsequent analyses, scaled transcripts per million (TPM) counts were used. The TPM counts were processed with the use of the web portal for integrated differential expression and pathway analysis (iDEP) (iDEP 2.01; http://bioinformatics.sdstate.edu/idep, accessed April 30, 2024). We also used iDEP 2.01 for a principal component analysis (PCA). A gene set enrichment analysis (GSEA) was conducted employing the GSEA software (ver. 4.3.2), allowing for the identification of significantly altered pathways and gene sets in the dataset.
Statistical analyses
We used GraphPad Prism software, ver. 9.2.0 (GraphPad Software, San Diego, CA) for the statistical analyses. To compare the relationships between different groups, we applied two distinct statistical tests depending on the data structure and distribution: the Wilcoxon matched-pairs signed rank test was used for paired data comparisons, and the nonparametric Mann-Whitney U-test was applied for unpaired data sets. To compare distributions between the observed and expected data, we used the χ2-test (parts-of-whole section).
The Kaplan-Meier method was used for the survival analyses, i.e., the PFS and OS rates. This allowed us to plot survival curves and estimate survival probabilities over time. The differences in survival rates between groups were evaluated by the log-rank test. Throughout the analyses, a p-value threshold < 0.05 was set for determining statistical significance. This threshold helped ensure that the observed differences or associations were not due to chance, thereby bolstering the reliability of our findings.