A total of 181 patients aged 37 to 86 years (median, 56) with endometrial cancer were retrospectively enrolled from the First Affiliated Hospital of Wannan Medical College (Wuhu, China). Inclusion criteria were as: i) Female; pathologically diagnosed as endometrial cancer in the past 3 years; no other malignant tumors nor serious chronic diseases; can be contacted and agree to participate in the project and sign an informed consent form. Exclusion criteria were set as: tissue sample retained in pathology department was too small; tumor cells in the sample was less than 10%; patients lost contact or were unwilling to participate in the research project.

These patients were diagnosed with endometrial cancer from April 2021 to November 2022 and tissues were collected. This was performed between November 2022 and June 2023. The present study was approved by the Ethics Committee of the First Affiliated Hospital of Wannan Medical College (approval no. 2022-110) and each patient provided written informed consent for their clinical information as well as their genomics data (from PCR and NGS) to be reported in the journal. Tumor and matched adjacent non-tumor tissues were collected from all patients and the MMR status was verified using IHC analysis of MSH2, MSH6, PMS2 and MLH1 protein expression levels. All IHC results were tested and reported by pathologists in the pathology department of the hospital. Antibodies including MLH1 (cat. no. ZM-0154, ZSGB-bio), MSH2 (clone FE11, cat. no. ZA-0622, ZSGB-bio, China), MSH6 (clone EP49, cat. no. ZA-0541, ZSGB-bio, China), and PMS2 (clone EP51, cat. no. ZA-0542, ZSGB-bio, China), were stained using Dako's automated staining system (LINK48, Dako, CA, USA) with 1:1,000 dilutions. All staining procedures were performed according to the manufacturer's recommendations and previous study (24).

Surgically specimens were fixed in 10% neutral-buffered formalin for 24–72 h at room temperature and embedded in paraffin. DNA was extracted from 10-µm-thick sections of formalin-fixed paraffin-embedded (FFPE) tumor tissue blocks using the GeneRead DNA FFPE kit (Qiagen, GmbH), according to the manufacturer's instructions. Samples were analyzed using MSI-PCR and NGS. After extraction, DNA quality was evaluated by 1% agarose gel electrophoresis and the concentration of all samples was quantifed using the Qubit dsDNA HS Assay kit (Termo Fisher Scientifc, Waltham, MA, USA) with a Qubit 3.0 Fluorometer.

For each of the five microsatellite loci, namely BAT-25, BAT-26, NR-21, NR-24 and MONO-27, four plasmids were synthesized by Sangon Biotech (China). These included a wild-type fragment and deletions of the wild-type, consisting of 1-, 2- and 3-bp deletions. Plasmids were utilized as spike-in fragments and mixed into the DNA of noncancerous endometrial tissue at a ratio of 1:1.

MSI-PCR analysis was performed using the MSI Analysis System (Promega Corporation) (25) as previously described (26,27). Briefly, the five microsatellite loci, namely BAT-25, BAT-26, NR-21, NR-24 and MONO-27, and two pentanucleotide repeats PENTAC and PENTAD, were amplified in a single multiplex 25 µl PCR reaction. PENTAC and PENTAD were used as reference genes to detect potential contamination. Fluorescently labeled primers used for MSI-PCR analysis are supplied by Sangon Biotech and listed in Table SI. The following thermocycling conditions were used for the PCR: Initial denaturation at 95°C for 11 min and 96°C for 1 min; 10 cycles of 94°C for 30 sec, ramp 68 sec to 58°C, hold for 30 sec, ramp 50 sec to 70°C and hold for 1 min; 20 cycles at 90°C for 30 sec, ramp 60 sec to 58°C, hold for 30 sec, ramp 50 sec to 70°C and hold for 1 min; 60°C for 30 min; hold at 4°C. PCR products were analyzed using a 3500 Genetic Analyzer (Thermo Fisher Scientific, Inc.). GeneMapper 6.0 (Thermo Fisher Scientific, Inc.) was used to determine the size differences between tumor samples and adjacent tissues. A tumor was defined as exhibiting MSI-H if ≥2 markers were unstable, and MSS was defined according to the presence of ≤1 unstable mononucleotide markers in the tumor sample. The term ‘unstable’ was used for markers with a shift of ≥2 bp, or if the shoulder pattern extended the range of the smallest peak by ≥2 bp in the tumor allele.

NGS was performed on a NextSeq 500 or Novaseq 6000 (Illumina, Inc.) using a custom amplicon-based gene panel that comprised five microsatellite loci included in the Promega Corporation MSI kit. Initially, libraries were generated with the Hieff NGS™ OnePot Pro DNA Library Prep Kit (Shanghai Yeasen Biotechnology Co., Ltd.) according to the manufacturer's protocol. Briefly, 20 ng fragmented genomic DNA was used to amplify the target regions and amplified products were purified (Table SII). Subsequent rounds of PCR were carried out through the addition of sequencing adapters and barcodes to amplicons. Following the purification of the library, quantification of the DNA library was performed using Labchip GX Touch (PerkinElmer). The libraries with 1 pM concentration were then sequenced using the Novaseq 6000 NGS (Illumina, Inc.) platforms and NovaSeq 6000 SP reagent kit (100 cycles; cat. no. 2002746; Illumina Inc.), according to the manufacturer's instructions using 2X150 bp paired-end reads at an average depth of 5,000× for tissue.

A novel algorithm, MSIPeak, was developed in the present study to determine the MSI status of all samples using NGS read-count distribution. MSIPeak program flow was divided into four main steps, as follows (Fig. 1).

Step I: The sequencing data of each tumor tissue and matched adjacent tissues were read in FastQ format files. For each MSI locus, reads coverage information, including reads count, was extracted from FastQ files.

Step II: Minimum-maximum normalization was performed on the reads count of each microsatellite locus. Values were scaled to the range (0,1) for subsequent data processing.

Here, i represents a single microsatellite locus, x_i represents the reads count prior to normalization, x_new represents the reads count value following normalization, and x_min and x_max represent the minimum and maximum values of the reads count for each locus, respectively.

Reads count values were smoothed using the sliding window. Following normalization and smoothing, peak data of the microsatellite loci of tumor and matched adjacent tissues were analyzed. The local maximum values of each repeat were compared with the values of neighboring points.

Step III: For each peak determined in the tumor and adjacent tissues, peak shift size, peak area difference and Shannon coefficient difference were calculated to score the MSI status of each locus (Fig. 2):

Here, i represents a single microsatellite locus, l_x represents the peak value of the microsatellite locus in the tumor sample and l_y represents the peak value of the microsatellite locus in the matched adjacent tissue.

Here, H_diff represents the area difference of each microsatellite locus peak, × and y refer to the vectors of area values of MSI loci in tumor and adjacent tissues, respectively, i represents a single microsatellite locus, and x_i and y_i represent the area values of the ith MSI locus in tumor and adjacent tissues, respectively.

Here, H represents the Shannon coefficient difference between tumor and adjacent tissues. H_x represents the Shannon-Wiener diversity index of the tumor sample, H_y represents the Shannon-Wiener diversity index of the adjacent sample, i represents a single microsatellite locus, p_i represents the relative abundance of the ith microsatellite locus in the tumor sample and p'_i represents the relative abundance of the ith microsatellite locus in the adjacent sample.

Step IV: The final score for each MSI locus was calculated using the following equation:

When the score was ≥1.10, the MSI status of this locus was considered unstable. After the stability of all five markers had been determined, the MSI status of the patient was evaluated. Samples with two or more unstable markers were considered MSI-H and samples with <2 unstable markers were considered MSS.

Among previously published NGS-based MSI studies, MSIsensor (21) and MANTIS (22) were widely used analytical methods (2,5,28,29). The calculation principles of these two algorithms are markedly different from MSIPeak (Table I). Therefore, MSIPeak was compared with the MSIsensor and MANTIS algorithms. MSISensor2 (30), an upgraded version of MSIsensor, and Mantis were run according to their manuscript, for the analysis of in-house whole-exome sequencing (WES) data from 25 endometrial cancer samples. The WES library was constructed using the commercial Hi-Exon 35 Panel and supporting library construction kit (cat. no. P10016-96, Shanghai HeYin Biotechnology Co., LTd.). A total of 50 ng fragmented genomic DNA was used for a capture-based library (Table SIII) according to the manufacturer's protocol of the library construction kit. After the quantification of the DNA library by the Labchip GX Touch (PerkinElmer), WES library with 1 pM concentration was performed on the same Novaseq 6000 NGS (Illumina, Inc.) platforms and NovaSeq 6000 SP reagent kit (100 cycles; cat. no. 2002746; Illumina Inc.), according to the manufacturer's instructions using 2×150 bp paired-end reads at an average depth of 150×. To obtain clean reads, FASTQ files from tumor tissue and white blood samples were done by fastp (https://github.com/OpenGene/fastp, version 0.19.3). Clean reads were mapped to the reference genome (hg38/GRCh38) by Burrows-Wheeler aligner (BWA, https://github.com/lh3/bwa, version 0.7.12-r1039) and perform alignment processing by SAMtools (https://github.com/samtools/samtools, version 0.1.19–96b5f2294a). The quality score was recalibrated using GATK (https://github.com/broadinstitute/gatk, version 4.1.0.0) to generate the final binary SAM (BAM) files used for subsequent analyses. Lastly MSI status was detected using the MANTIS (version v1.0.5) (22) and MSISensor2 (Version 0.1) (30).

The chi-square test was used to compare the frequencies of MSI-H and MSS tumors identified through PCR and NGS, with the dMMR and pMMR status determined by IHC. For chi-square test analysis, P<0.001 was considered to indicate a statistically significant difference in PCR and NGS in ability to detect MSI-H. Cohen's κ was calculated to evaluate the level of agreement between IHC-based and molecular-based methods, PCR and NGS. A Cohen's κ of P<0.001 was considered to indicate a statistically significant difference between methods. All data presented in figures and tables are reported as percentages for categorical comparisons. P<0.05 was considered to indicate a statistically significant difference. The statistical analyses were performed using R software (version 4.3.2; RStudio).

The present study demonstrated that spike-in DNA sample profiles with 3-bp deletions displayed notable left-shifts, which could be distinguished from those of the wild-type fragments (Fig. 3A). Spike-in DNA sample profiles with 2-bp deletions exhibited ambiguous extensions on the left shoulder of BAT25 and NR24, compared with the wild-type fragments. However, there were no notable shifts in BAT26, NR21 and MONO27 with 2-bp deletions or in the five markers with 1-bp deletions (Fig. 3A and B).

Results obtained using MSI-NGS are presented as peaks, which were comparable with those obtained using MSI-PCR (Fig. 3C). Peaks of the spike-in DNA samples with 1-bp deletions exhibited subtle shifts compared with those of the wild-type fragments. However, the score was more than the threshold of 1.10 (Fig. 3D). Spike-in DNA samples with 1- or 2-bp deletions exhibited shifts and scores that were indicative of MSI (Fig. 3C and D).

A total of 39 endometrial cancer samples were identified as dMMR and the remaining 142 samples were identified as pMMR (Table II). Within the 39 dMMR samples, 16 were classified as MSI-H using PCR testing and 36 were identified as MSI-H using NGS (Table II). The concordance between IHC and NGS was significantly higher compared with that between IHC and PCR (Cohen's κ=0.492 vs. 0.872; P<0.001; Table II). All 16 dMMR/MSI-H samples confirmed using IHC and PCR were also defined as MSI-H using NGS (Fig. 4; Tables SIV and SV). In addition, a further 20 MSI-H samples were identified using NGS alone, with the profiles exhibiting minor shifts that did not meet the criteria for MSI-H based on PCR analysis (data not shown).

Among the 25 samples with available WES data, 8 dMMR and 17 pMMR cases were identified using IHC analysis. MSIPeak, MSISensor2 and MANTIS consistently classified the 17 pMMR samples as MSS (Fig. 5A). However, MSIPeak was the only algorithm to identify all 8 dMMR samples as MSI-H, while MSISensor2 and MANTIS classified 2 and 3 dMMR samples as MSS, respectively (Fig. 5A and B; Table SVI).