This was a retrospective study conducted at a single referral hospital. The results of universal screening in consecutive patients with CRC who underwent endoscopic and/or surgical resection at Kyoto University Hospital from 2016 until 2018 were analyzed. To evaluate MMR deficiency in an independent cohort of patients with young-onset CRC, we included patients with CRC who were younger than 50 years of age, whose tumor tissues were obtained at Kyoto University Hospital between 2006 and 2015 and for whom formalin-fixed paraffin-embedded (FFPE) samples were available. We defined young-onset CRCs as tumors diagnosed before age 50, in accordance with the Amsterdam criteria II and the revised Bethesda guidelines (3,4). Medical records were reviewed and clinicopathological information and data related to universal screening were collected. Information on genetic testing was also obtained when applicable. The study protocol was approved by the Ethics Committee of Kyoto University Graduate School and Faculty of Medicine (R0821 and R1978).

Immunohistochemical staining was performed using FFPE blocks. We used antibodies against MLH1 (clone; M1, Roche 518-110215; Roche Diagnostics), MSH2 (clone; G219-1129, Roche 518-101947; Roche Diagnostics), MSH6 (clone; EPR3945, Abcam ab92471; Abcam), PMS2 (clone; EPR3947, Roche 518-110857; Roche Diagnostics), and BRAF V600E (clone; VE1, Abcam ab228461; Abcam). Slides were stained using a Ventana BenchMark ULTRA instrument (Ventana Medical Systems) according to the manufacturer's protocol. Total negativity for nuclear expression of MLH1, MSH2, MSH6, and PMS2 was considered as loss. BRAFV600E was evaluated as positive if diffuse and nearly uniform cytoplasmic staining of tumor cells was present with or without membranous accentuation.

As the first step, IHC for MSH6 and PMS2 was performed, and cases with retained expression of both proteins were regarded as proficient MMR (pMMR). Tumors with loss of expression of either protein were designated as dMMR, and immunostaining for MSH2 and/or MLH1 was added. IHC for BRAF V600E was also performed when a tumor showed loss of MLH1 expression. Tumors showing: i) A combined loss of MSH6 and MSH2, ii) loss of MSH6 alone, iii) loss of PMS2 alone, or iv) a combined loss of PMS2 and MLH1 in conjunction with negative BRAF V600E IHC were considered to indicate possible LS. Tumors with a combined loss of PMS2 and MLH1 with positive BRAF V600E IHC were considered to indicate sporadic dMMR. For patients with multiple synchronous CRCs, the most advanced tumor for each individual was designated as the index tumor if IHC results were concordant. Because patients who have both pMMR and dMMR CRCs are less likely to carry LS, a pMMR CRC was assigned as the index tumor. For patients who had metachronous CRCs during the study period, the first resected tumor was chosen as the index tumor.

Patients with tumors showing a possible LS pattern were designated as GTCs. GTCs were informed of their IHC results at a regular follow-up visit, and those who opted to participate were referred to a specialist for hereditary tumor and genetic counseling, and were asked to consider genetic testing for MMR genes and EPCAM. After retrospective MLH1 methylation analysis, GTCs were divided into suspected sporadic cases (sp-GTC), whose tumors were positive for MLH1 methylation, and bona fide GTCs (bf-GTC) for whom MLH1 methylation was negative (eight cases) or undetermined (two cases; insufficient tissue).

Genomic DNA was extracted from FFPE samples of tumors and from matched normal epithelial tissues using GeneRead DNA FFPE (Qiagen GmbH). To examine BRAF V600E mutations, DNA was processed using AmpliSeq for Illumina Library Plus (Illumina Inc.), and analyzed using MiniSeq (Illumina). The sequencing data were analyzed using the Local Run Manager software and VariantStudio (both from Illumina Inc.). To investigate the methylation status of CpG sites in the MLH1 promoter, a pair of bisulfite sequencing primers (BSP) was designed to amplify the genomic sequence including 11 CpG sites (Fig. S1A) using Methyl Primer Express software v1.0 (Thermo Fisher Scientific, Inc.). Genomic DNA was analyzed by targeted bisulfite sequencing as follows: Bisulfite conversion of DNA was performed using EZ DNA Methylation kit (Zymo Research). Then bisulfite-treated DNA was amplified using BSPs and a KAPA HiFi Uracil⁺ kit (Roche Sequencing and Life Science KAPA Biosystems) and was processed using an Ion Plus Fragment Library kit (Thermo Fisher Scientific, Inc.) to prepare libraries for the Ion PGM sequencing system (Thermo Fisher Scientific, Inc.). Ion Torrent sequencing data were analyzed using TABSAT (http://github.com/tadkeys/tabsat) (22). The percentage of methylation in each CpG site was calculated as 100x (reference C reads)/[(C>T converted reads) + (reference C reads)] and the average of 11 CpG sites was determined to represent the methylation level for each sample. We used in vitro methylated and unmethylated control DNA samples to validate the results of this bisulfite sequencing assay and confirmed that methylation levels in the MLH1 promoter region was detected in quantitative manner (Fig. S1B). Because all the normal epithelial tissues and the known nonmethylated colon cancer cells (LoVo) showed methylation levels ≤2% (data not shown), tumor samples showing methylation levels >10% were considered methylated.

The Fisher's exact test or the chi-squared test was used to analyze categorical data. All P-values were two-sided and P<0.05 was considered to indicate a statistically significant difference. All analyses were carried out using JMP 15.2 (SAS Institute, Inc.)

The results of the universal screening are summarized in Fig. 1. A total of 540 CRCs from 503 patients were treated by either surgical or endoscopic resection. Three patients who had either known familial adenomatous polyposis, Crohn's disease, or ulcerative colitis were excluded from the LS screening. An additional 37 CRCs were excluded because IHC was not performed for undetermined reasons. For the 32 patients who possessed multiple CRCs, only the index tumors were included for further analyses.

Thus, the universal screening procedure was performed in 463 patients with CRC. The patients' characteristics are shown in Table I. Representative images of IHC are shown in Fig. S2A-E. There were 427 pMMR and 34 dMMR tumors; the expression status of PMS2 and/or MSH6 was undetermined in two cases. No tumors showed a combined loss of PMS2 and MSH6. After additional IHC for MLH1, MSH2, and BRAF V600E, 15 dMMR tumors showed the sporadic pattern. Another tumor was considered to be sporadic despite showing combined MLH1 and PMS2 loss with negative BRAF V600E IHC, because it arose within a sessile serrated lesion, which is known to be a major precursor of sporadic dMMR CRC (23). The remaining 18 patients had dMMR tumors with a possible LS pattern and were designated as GTCs. After they were informed of their IHC results at a regular follow-up visit, 11 GTCs opted to consult hereditary tumor specialists and/or the clinical genetics unit, and eight of them subsequently underwent genetic testing. Germline analyses revealed two pathogenic mutations in MMR genes, NM_00251.2(MSH2):c.84_85(p.Lys29X) and NM_00249.3(MLH1):c.1692delC:(p.Ile565PhefsX26), leading to identification of two LS probands (0.4%) (Fig. 2A).

Of the 461 index tumors for which MMR status was successfully determined by IHC, 427 were pMMR and 34 were dMMR (Table II). We then evaluated the proportion of dMMR tumors according to various clinicopathological findings. The dMMR CRCs displayed a bimodal distribution with age with peaks in younger (<50 years old) and older (≥60 years old) patients. In addition, dMMR tumors were significantly enriched within right-sided tumors, and high-grade and mucinous adenocarcinoma, and other histology compared with low-grade adenocarcinoma. Detailed information about the dMMR CRCs is presented in Table SI.

Although we did not incorporate assays for BRAF V600E mutations and MLH1 methylation assays into the universal screening process, we retrospectively analyzed them to evaluate the validity of our procedure. Our data showed that BRAF V600E IHC was able to determine BRAF V600E mutational status with a sensitivity and a specificity of 86.7 and 76.5%, respectively (Table SII). MLH1 promoter methylation analysis showed that, among the 32 dMMR tumors tested, all 16 non-GTCs with the sporadic IHC pattern were positive for MLH1 methylation. Eight of 10 dMMR tumors in the GTC group that showed MLH1 protein loss and negative or undetermined BRAF V600E IHC were also positive for MLH1 methylation, indicating that they were more likely sporadic than LS (Table SIII).

After the retrospective MLH1 methylation analysis demonstrated positive MLH1 methylation in the majority of dMMR tumors, we divided the GTCs into two groups: sp-GTC, whose tumors showed MLH1 methylation indicative of a sporadic pattern and bf-GTC without positive-MLH1 methylation. As shown in Table III, universal screening identified dMMR tumors, bf-GTCs, and LS probands in 7.3, 2.2, and 0.4%, respectively, of patients with CRC. Given the low rate of detection of bf-GTC and LS, we then tested a range of age-restricted strategies. Screening of patients with CRC who were <50 years old reduced the required number of tests to 33 (7.1% of all CRC), and enabled a remarkable concentration of bf-GTCs and LS probands. However, seven of 10 bf-GTCs remained untested. Screening of patients with CRC who were <70 years old required 232 tests (50.1% of all CRC), and enabled to detect eight of 10 bf-GTCs and both LS probands identified by universal strategy. In contrast, screening of 231 patients with CRC who were ≥70 years old led to detection of only two bf-GTCs and no confirmed LS cases. (Table III).

Next, we investigated the frequencies of those who underwent genetic testing in GTCs. Included in the 19 patients with dMMR CRC who were ≥70 years of age were nine GTCs and two bf-GTCs. Of these, three GTCs (3/9, 33.3%) underwent genetic testing, all of whom were negative. In contrast, all three patients <50 years of age who had dMMR CRC were bf-GTCs. Two of these underwent genetic testing, leading to the identification of two LS probands (Fig. 2B). Another woman <50 years old had both colon and endometrial cancers with loss of MSH2/MSH6 expression, and the multiple occurrences of CRCs in her family fulfilled the Amsterdam criteria II. Despite the high level of suspicion of LS, she has not undergone genetic testing.

To test the enrichment of dMMR tumors and GTCs in an independent cohort of patients with young-onset CRCs, we utilized historical cases from before the start of universal screening in our hospital. As shown in Table IV, evaluation of MMR status in a total of 111 patients with young-onset CRCs (<50 years old) revealed 10 (9.0%) dMMR tumors including nine GTCs and one non-GTC. There was greater enrichment of dMMR tumors among patients <40 years of age. Detailed information about the dMMR CRCs is presented in Table SIV.