Introduction

Lynch syndrome (LS, MIM: 120435) is one of the most prevalent autosomal dominant hereditary cancer susceptibility syndromes in humans and is caused by alterations in one of four mismatch repair (MMR) genes, namely, MLH1, MSH2, MSH6 and PMS2, or a structurally intact MSH2 epigenetic inactivation caused by a large deletion in the non-MMR gene epithelial cellular adhesion molecule (EPCAM)1,2. The MMR system plays a crucial role in maintaining genomic stability and preventing cancer within a biological system by rectifying errors arising during DNA replication and recombination2,3. Inactivation of the second allele typically occurs due to small pathogenic variants and gene loss, resulting in an impairment of the MMR system. DNA mismatch repair deficiency (dMMR) inactivates the ability to repair DNA mismatches, inducing accelerated carcinogenesis4. LS increases the risk of developing several cancers throughout life, including cancers of the colon, rectum, endometrium, stomach, small bowel, biliary tract, pancreas, renal pelvis and/or ureter, bladder, kidney, ovary, brain, or prostate, which are known as LS-associated cancers because of their significantly greater frequency compared to that of the average population5,6,7.

Tumors that occur in LS can develop at any age but often arise in young people, often have a family history, and have the possibility of developing multiple synchronous and metachronous tumors8,9. dMMR in LS often results in high-level microsatellite instability (MSI-H) and the absence of MMR protein(s) in tumor tissue, which provides valuable shortcuts for identifying patients with LS1. The United States Food and Drug Administration (FDA) has now approved the use of immune checkpoint inhibitors for solid tumors exhibiting MSI-H or dMMR, making them a viable treatment option for LS-associated tumors10,11. Next-generation sequencing (NGS) of tumors provides an alternative method for LS screening12,13. Tumor screening and clinical history are both helpful in identifying patients who may have LS, but the identification of heterozygous pathogenic or likely pathogenic (P/LP) variants in one of the MMR genes (or EPCAM) is the definitive diagnosis of LS2. Importantly, with the wide application of NGS technology and multigene testing, the spectrum of LS-associated tumors has also continued to expand, which is beneficial for monitoring and risk management of patients with LS14,15,16. The tumor screening, identification, surveillance and therapy of patients with LS, as well as surgical management of colorectal and gynecological cancers, have been recognized by consensus guidelines17,18,19,20,21,22,23,24.

Although LS quadruples the risk of brain tumors occurring, brain tumors are relatively rare and represent the uncharacterized tumor type in LS, and there is no sufficient consensus on guidelines and management recommendations17,18,19,20,21,22,23,24,25,26,27. Among central nervous system (CNS) tumors, glioma is the most commonly associated with LS, and medulloblastoma and neuroblastoma have also been reported; MSH2 pathogenic variants are more common than other MMR mutations in CNS tumors28. In addition, patients with LS can also develop LS-unrelated cancers, such as breast cancer and sarcomas, because the data are insufficient to demonstrate that the risk of developing these cancers is increased in individuals with LS5,14,29,30,31,32. Age, MSI level, genomic alterations in tumors and other characteristics may differ among LS patients with different tumor types; thus, additional studies are needed to address the features of LS patients with noncanonical tumors14. Although some non-CNS LS-unrelated tumors and CNS LS-related tumors have been reported to develop in a few patients with LS and have been published as case reports and small retrospective cohorts, systematic studies with large sample sizes lack the description of the comprehensive characteristics of the population, especially comparisons with non-CNS LS-related tumors.

Here we study LS from the tumor perspective and investigate the similarities and differences in clinical and genomic features across multiple cancer types in CNS LS-related and non-CNS LS-unrelated tumor groups compared with those in the non-CNS LS-related tumor group, which will improve LS diagnostics, prevention, management and therapy.

Results

Patient enrollment

The clinical characteristics of the enrolled patients with LS

A total of 238 samples from 228 pan-cancer patients with LS were enrolled and divided into 3 groups according to the tumor location and tumor spectrum of the LS: the CNS LS-related tumor group (n = 68), the Non-CNS LS-related tumor group (n = 117), and the Non-CNS LS-unrelated tumor group (n = 53). The clinical characteristics of the three groups are shown in Table 1. There were significant differences in age among the three groups, with the age of the CNS LS-related tumor group being lower than that of the other two groups (p < 0.001). Germline variant gene (p < 0.001) and MSI status (p < 0.001) were also significantly different, but there was no significant difference in gender distribution among the three groups.

Table 1 Characteristics of the tumors of enrolled patients with Lynch syndrome in the three groups
Full size table

Figure 1A–D shows the gender and age distributions of the whole cohort and the three groups. In the overall population, there were fewer females than males (female: male = 103:135), and overall age showed a normal distribution, with the peak of both sexes appearing at 51–60 years of age. The age distributions of the three groups were different, especially in the CNS LS-related tumor group. The proportion of patients under 20 years old in the CNS LS-related tumor group was significantly greater than that in the other two groups (18/68, 1/117, 0/53), and the peak value appeared in the 31–40 years old range. The peak age of the two Non-CNS groups both appeared at 51-60 years, but the age began to increase significantly from 30 years in the Non-CNS LS-related group, while from 40 years in the Non-CNS LS-unrelated group, which revealed that the Non-CNS LS-unrelated group may have a later age of onset than the Non-CNS LS-related group.

Fig. 1: The population pyramid of patients with Lynch syndrome (LS) and the distribution of different types of cancer.
Fig. 1: The population pyramid of patients with Lynch syndrome (LS) and the distribution of different types of cancer.The alternative text for this image may have been generated using AI.
Full size image

A–D Gender and age distributions of 238 LS samples (A), Central nervous system (CNS) LS-related tumors (B), Non-CNS LS-related tumors (C), and Non-CNS LS-unrelated tumors (D). E Bar graph maps depicting the percentage of patients with LS in 19 different cancer types obtained from The Cancer Genome Atlas (TCGA) dataset versus the Simceredx dataset across 19 cancer types, and the overall difference. The number of patients with lynch syndrome and overall, in the Simceredx cohort and the TCGA cohort are 477/83170 and 88/15812, respectively. Pearson’s chi-squared test was used to evaluate the statistical difference of the incidence of lynch syndrome between the Simceredx cohort and the TCGA cohort (*p  <  0.05; **p  < 0.01; ***p  < 0.001). Significant difference was observed in endometrial cancer (5.68% in Simceredx vs. 1.66% in TCGA, p = 0.0047, χ² = 7.9845), and liver cancer (0.27% vs. 1.33%, p = 0.011, χ² = 6.4781). Source data are provided as a Source Data file.

Incidence of LS in different cancer types

In this study, LS-associated germline P/LP variations were detected in a total of 19 cancer types. The 3 most common cancers detected were all non-CNS LS-related tumors, namely, EC (5.68%), urothelial cancer (3.59%), and CRC (1.96%). The three cancers with the lowest detected proportions were lung cancer (0.18%), gastric cancer (0.15%) and breast cancer (0.13%). The proportion of patients with EC in our cohort was significantly greater than that in the Cancer Genome Atlas (TCGA) cohort (5.68% vs. 1.66%, p = 0.0047), while the proportion of patients with liver cancer was significantly lower than that in the TCGA cohort (0.27% vs. 1.33%, p = 0.011)33. The incidence ratio of each cancer type in the Simceredx cohort and TCGA cohort is shown in Fig. 1E and the Supplementary Data 1.

Germline mutational feature analysis of the MMR gene

The proportion of different MMR genes in different groups

The number of germline P/LP variant genes in the 238 LS samples is shown in Fig. 2A, including MLH1 (n = 52), MSH2 (n = 69), MSH6 (n = 60), and PMS2 (n = 57), which were classified as novel (n = 47, 19.75%) or previously identified (n = 191, 80.25%) mutations according to whether they were reported or included in the database. Novel mutations indicate that these mutations were neither reported in the literature nor included in the database. The proportions of germline P/LP variant genes in different groups are shown in Fig. 2B. The proportions of MSH2 in the CNS LS-related tumor group, MLH1 in the Non-CNS LS-related tumor group and PMS2 in the Non-CNS LS-unrelated tumor group were the highest. There was no significant difference between the four genes in the CNS LS-related tumor group and the Non-CNS LS-related tumor group. The detection ratio of PMS2 in the Non-CNS LS-unrelated tumor group was significantly greater than that in the CNS LS-related tumor group (43.4% vs. 17.65%, p = 0.036) and Non-CNS LS-related tumor group (43.4% vs. 18.8%, p = 0.021). The percentage of MLH1carriers was significantly greater in the Non-CNS LS-related group than in the Non-CNS LS-unrelated tumor group (31.62% vs. 9.43%, p = 0.021). The detection ratio of MSH2 was significantly greater in the CNS LS-related tumor group than in the Non-CNS LS-unrelated tumor group (42.65% vs. 16.98%, p = 0.043). There was no significant difference in the proportion of MSH6 among the three groups.

Fig. 2: Germline mutational features of mismatch repair (MMR) genes among different cohorts.
Fig. 2: Germline mutational features of mismatch repair (MMR) genes among different cohorts.The alternative text for this image may have been generated using AI.
Full size image

A Bar graph showing the number of germline pathogenic or likely pathogenic (P/LP) variants in MMR genes across 238 samples. B The proportions of different germline P/LP variant genes within the three cohorts. C Box plot showing the age distribution of different MMR carriers across distinct groups. A total of 238 samples, CNS LS tumor (n = 68), Non-CNS LS-related tumor (n = 117), and Non-CNS LS-unrelated tumor (n = 53) were included in this analysis, with each sample representing a single biological replicate. The box plot displays the median (center line), interquartile range (IQR) between 25th and 75th percentiles (bounds of the box), and whiskers extending to 1.5 times the IQR, and the points beyond are outliers, with “minima” and “maxima” as the absolute lowest and highest values. Colors denote the specific germline gene mutation, including MLH1 (orange), MSH2 (blue), MSH6 (green), and PMS2 (yellow). The p-values of comparisons were evaluated by a two-sided Mann–Whitney test, with multiple comparisons adjusted by the false discovery rate (FDR) approach. The asterisks denote significance levels: *p < 0.05, **p < 0.01, and ***p < 0.001. Source data are provided as a Source Data file.

Lollipop plots of four gene variant sites on the peptide sequence in total and in the different groups are shown in Supplementary Fig. 1. MLH1 is dominated by loss-of-function (LOF) mutations, including frameshift, nonsense, and splicing variants. The mutations had no hot spot distribution area and were evenly distributed throughout the gene. The MSH2, MSH6, and PMS2 genes were also similar.

Age distribution of different germline pathogenic variant genes in different groups

Box plots were drawn to explore the correlation between different germline P/LP variant genes and age in different groups (Fig. 2C). The results revealed that the age of patients harboring PMS2 germline P/LP variants in the Non-CNS LS-related tumor group was significantly greater than that of patients harboring MLH1 germline P/LP variants (p = 0.0499), but beyond that, there was no significant difference in age between the 4 P/LP variant genes in the same group. Comparison of the same gene between different groups revealed that there was no significant difference in the age distribution of patients harboring MLH1 and MSH2 germline P/LP variants among the three groups, while the age of patients harboring MSH6 (p = 0.005; p = 0.016) and PMS2 (p = 0.0019; p < 0.001) germline P/LP variants in the CNS LS-related tumor group was significantly lower than that in the other two groups.

In brief, we found that tumor samples in the three groups exhibited heterogeneity in the distribution of germline pathogenic genes (Fig. 2A, B), and there were significant differences in age among the three groups, with the age of the CNS LS-related tumor group being lower than that of the other two groups, especially in patients harboring MSH6 and PMS2 germline P/LP variants (Table 1 and Fig. 2C).

Somatic mutational landscape and feature analysis

The somatic mutation heatmaps of the three groups

To elucidate the genomic characteristics of LS, we conducted further analysis on the landscape of somatic mutations, TMB, and MSI status across the three groups. To explore the differences at the somatic gene mutation level among the three groups, we generated heatmaps of the top 50 genes in the three groups (Fig. 3A–C). The genomic mutation profiles of the three groups were different, with the largest number of mutations in the Non-CNS LS-related tumor group and the smallest number of mutations in the Non-CNS LS-unrelated tumor group. The mutation frequency top 5 genes in the CNS LS-related tumor group were TP53 (AF 91.18%), SETD2 (AF 67.65%), NF1 (AF 63.24%), SMARCA4 (AF 52.94%) and PDGFRA (AF 51.47%); in the Non-CNS LS-related tumor group were MUC16 (AF 75.21%), ACVR2A (AF 64.96%), FAT4 (AF 64.96%), LRP1B (AF 63.25%) and CIC (AF 61.54%); and in the Non-CNS LS-unrelated tumor group were TP53 (AF 52.83%), EGFR (AF 35.85%), MUC16 (AF 33.96%), CDKN2B (AF 20.75%) and LPR1B (AF 20.75%).

Fig. 3: The landscape plot and tumor mutation burden (TMB) analysis of patients with LS.
Fig. 3: The landscape plot and tumor mutation burden (TMB) analysis of patients with LS.The alternative text for this image may have been generated using AI.
Full size image

A–C Somatic mutation heatmaps of CNS LS-related tumors (A), non-CNS LS-related tumors (B), and non-CNS LS-unrelated tumors (C). Distinct colors correspond to specific mutational functions and clinical information, as indicated. Each column represents an individual patient, with the top 50 genes and mutation frequencies listed on the y-axis. D, E Box plot showing the comparison of TMB across the three groups (D) and the TMB of different germline mutant genes in different groups (E). A total of 191 patients have detected TMB values, with n = 22 in CNS LS tumor, n = 117 in Non-CNS LS-related tumor, and n = 52 in Non-CNS LS-unrelated tumor. The box plot displays the median (center line), interquartile range (IQR) between 25th and 75th percentiles (bounds of the box), and whiskers extending to 1.5 times the IQR, and the points beyond are outliers, with “minima” and “maxima” as the absolute lowest and highest values. Colors denote the specific germline gene mutation, including MLH1 (orange), MSH2 (blue), MSH6 (green), and PMS2 (yellow). The p-values of comparisons were evaluated by a two-sided Mann–Whitney test, with multiple comparisons adjusted by the FDR approach. (ns, p > 0.05; *p  <  0.05; **p  <  0.01; ***p  <  0.001). Source data are provided as a Source Data file.

TMB in different groups and in different germline variant genes

The box plots show the differences in TMB among the three groups (Fig. 3D). The median TMB was 42.27 (0–639.01), 51.47 (0.71–412.06) and 3.55 (0–90.07) in the CNS LS-related tumor group, Non-CNS LS-related tumor group and Non-CNS LS-unrelated tumor group, respectively. The TMB in the Non-CNS LS-unrelated tumor group was significantly lower than that in both the Non-CNS LS-related tumor group (p < 0.001) and the CNS LS-related tumor group (p < 0.001).

To better explore the differences in TMB between the germline P/LP variant genes in different groups, we constructed box plots of the TMB of different germline variant genes in different groups (Fig. 3E). Intragroup comparison revealed that the TMB of patients harboring the other three genes was greater than that of patients harboring PMS2 in the total population and in the non-CNS LS-related tumor group (p < 0.05). In the Non-CNS LS-unrelated tumor group, the TMB of patients harboring the MLH1 gene was greater than that of patients harboring other genes (p < 0.05). The results showed that the TMB of samples with MSH2 and MSH6 germline variants in the CNS LS-related tumor and Non-CNS LS-related tumor groups was greater than that in the Non-CNS LS-unrelated tumor group (p < 0.001), and the TMB of samples with PMS2 germline variants in the Non-CNS LS-related tumor group was greater than that in the Non-CNS LS-unrelated tumor group (p < 0.005).

These results suggest that the TMB features in the CNS LS-related tumor group are more similar to those in the Non-CNS LS-related tumor group.

MSI status in different groups and in different germline variant genes

The corresponding MSI status of samples with different germline pathogenic variant genes in different groups is shown in Fig. 4A, B. Compared with the other two groups, the Non-CNS LS-related tumor group had a greater proportion of MSI-H tumors, especially in the samples with MLH1 and MSH2 germline P/LP variants. The proportion of MSI-H tumors in samples with MLH1 variants was the highest, followed by that in samples with MSH2 variants. In the Non-CNS LS-unrelated tumor group, MSH-H only existed in some samples with MLH1 variants and in 1 patient with a PMS2 variant. In the CNS LS-related tumor group, no samples with MSI-H status were found among the samples harboring MSH6 and PMS2.

Fig. 4: Distribution patterns of microsatellite instability (MSI) status of different genes in different groups and characteristics of patients with LS harboring germline co-occurring susceptibility mutations.
Fig. 4: Distribution patterns of microsatellite instability (MSI) status of different genes in different groups and characteristics of patients with LS harboring germline co-occurring susceptibility mutations.The alternative text for this image may have been generated using AI.
Full size image

A The numbers of different MSI status in different germline genes among the three groups are depicted, with distinct colors representing various MMR genes. The sample number is indicated on the y-axis. B Pie chart indicating the distribution of MSS tumors (blue), MSI-H tumors (dark orange) and MSI-L tumors (yellow) with respect to different germline genes among the three cohorts. C The TMB, MSI and germline MMR gene signatures of different cancer types for each sample were analyzed. The tumor types are indicated on the x-axis. TMB is indicated on the y-axis. The shape indicates the MSI status. The colors indicate the germline genes. D The characteristics of patients with LS harboring germline co-occurring susceptibility mutations. The different colors correspond to different characteristics, as indicated. Each column represents an individual patient. Source data are provided as a Source Data file.

Figure 4C shows the TMB, MSI and germline variant genes in different cancer types for each sample with TMB information. In the non-CNS LS-unrelated tumor group, the TMB was generally low. In the CNS LS-related tumor group, 6 glioma samples (PLS041, PLS046, PLS124, PLS143, PLS185, PLS228) had a TMB greater than 300 but were MSS, and we found each patient harboring mutations in the DNA polymerase genes, POLE or POLD1, which are associated with a hypermutated phenotype. The count and statistical results of the MSI status corresponding to different genes in each group are shown in Supplementary Data 2.

The exceptional cases in LS

Co-occurring germline susceptibility mutations in LS

We identified 10 patients with LS harboring germline co-occurring susceptibility mutations in genes other than MMR genes. These co-occurring mutated genes were primarily involved in the DNA damage repair pathway, including ATM, ATR, AXIN2, ERCC2, FANCA, FANCD2, MUTYH, NF1 (Fig. 4D). Notably, seven cases occurred in the non-CNS LS-related tumor group, including 6 patients with CRC exhibiting characteristics of MSI-H and TMB-H. In contrast, the remaining three patients in the CNS LS-related tumor group and the non-CNS LS-unrelated tumor group had MSS. Intriguingly, two patients in the CNS LS-related tumor group were over 50 years old, suggesting that co-occurring autosomal dominant tumor-related genes, such as NF1 variants, may drive tumorigenesis in these patients (Fig. 4D).

Our analysis of tumor germline susceptibility genes provides valuable insights into the underlying drivers of disease development in patients with LS, aiding in personalized management and avoiding diagnostic and treatment biases.

Multiple primary cancers in LS

Among the patients in our cohort, 9 (3.9%) were diagnosed with multiple primary tumors, with 19 samples. Table 2 shows the characteristics of LS patients with multiple primary cancers. Two-thirds (66.7%) of these patients had tumors in two different organs, while one-third (33.3%) had tumors in the same organ (e.g., the intestine or lung). Notably, 44.4% (4/9) of the patients with multiple primary tumors had different MSI status at different lesion sites, and all of these patients had tumors in two different organs. Patients with multiple primary tumors were older (average age: 55.2 years) than those with a single primary tumor. Additionally, females (n = 6) may have a greater probability of developing multiple primary tumors than males (n = 3). Understanding the characteristics of multiple primary tumors associated with LS is critical for accurate diagnosis, treatment planning, and management of the disease.

Table 2 Characteristics of LS patients with multiple primary cancers
Full size table

Discussion

Our study represents a large-scale study of LS patients in China, including a total of 229 patients with LS in 19 cancer types identified from 44,040 pan-cancer samples. In this study, 238 final enrolled samples from 228 patients were divided into 3 groups according to tumor location and tumor spectrum of LS: the CNS LS-related tumor group (n = 68), the Non-CNS LS-related tumor group (n = 117), and the Non-CNS LS-unrelated tumor group (n = 53). We studied LS from the tumor perspective and analyzed clinical and genomic characteristics and the similarities and differences among the 3 groups. The exploration of age, incidence, germline alterations, and somatic mutation characteristics in different groups is conducive to improving LS identification, management and treatment.

The characteristics of the non-CNS LS-related tumor group in our study were similar to those reported in previous studies, with incidence rates of EC (5.68%) and CRC (1.96%)34,35,36, a mean age of 52.8 years9, a proportion of MLH1 and MSH2 genes with germline variants being the highest5,34, a median TMB being high (51.47, from 0.71 to 412.06), and a ratio of MSI-H being 70.1%8. The median ages of the carriers of the MLH1, MSH2, MSH6 and PMS2 genes were 49.5, 51, 52, and 56 years, respectively, and the patients harboring PMS2 germline P/LP variants were significantly older than the patients harboring MLH1 germline P/LP variants (Fig. 2C), indicating that the trend of the age distribution in the Non-CNS LS-related tumor group was consistent with previous reports5. The TMB of patients with variants in the other three MMR genes was greater than that of patients with PMS2 variants in the total population and in the non-CNS LS-related tumor group.

CNS tumors, as an extracolonic manifestation of LS, appear to be relatively infrequent, with the occurrence risk quadrupling, and the cumulative lifetime risk of brain tumors is significantly greater in individuals with MSH2 mutations than in those with variants in other MMR genes28,37,38. In the CNS LS-related tumor group in our study, the MSH2 gene (42.65%) accounted for the greatest proportion of the four MMR germline P/LP genes, which is consistent with previous reports and highlights the importance of screening for the MSH2 gene in CNS tumors25,28. Both the mean age (38 years) and the age of patients harboring MSH6 and PMS2 germline P/LP variants in the CNS LS-related tumor group were significantly lower than those in the other two groups, while there was no significant difference in the age distribution among the 4 MMR genes in the CNS LS-related tumor group. The TP53 gene was the most frequently somatically mutated gene in the CNS LS-related tumor group, with an AF of 91.18%. There was no significant difference in the TMB distribution between the CNS LS-related tumor group and the Non-CNS LS-related tumor group, indicating that the TMB features were more similar between the two groups. However, in the CNS LS-related tumor group, MSI-H tumors were found only in samples with MLH1 (6/10) and MSH2 (11/29) germline P/LP variants. The CNS LS-related tumor group had a lower proportion of MSI-H tumors than the Non-CNS LS-related tumor group. This finding is consistent with previous reports indicating that MMRDness scores vary across different tissues, with brain tissue typically exhibiting lower scores compared to blood and gastrointestinal tissues39. Compared to non-CNS LS-related tumors, CNS-LS-related tumors are relatively rare and may manifest at an earlier age but have similar characteristics of MMR deficiency and have some slight genetic heterogeneity. Our research serves as a fundamental reference for the diagnosis and treatment application of CNS-LS-related tumors.

In the Non-CNS LS-unrelated tumor group in our study, the proportion of patients with PMS2 (43.4%) was the highest, and this ratio was significantly greater than that in the other two groups; however, the proportion of patients with MLH1 was significantly lower in the Non-CNS LS-unrelated group than in the Non-CNS LS-related group. It is widely believed that individuals carrying the P/LP PMS2 variant have a much lower risk of developing cancer than other MMR carriers, although individual cancer risk may be influenced by various factors40,41. In the previous study42, PMS2 gene had the highest carrier rate among the four MMR genes in colorectal cancer cases. In our study, due to differences in the classification of cancer types and potential variations in ethnicity, PMS2 gene did not have the highest proportion in the non-CNS LS-related tumor group, which includes colorectal cancer. LS patients may develop sporadic tumors not associated with MMR deficiency. These tumors may represent background cancer risk unrelated to LS. Our tumor-based analysis describes features of tumors arising in LS carriers but does not imply that every tumor is causally related to LS. PMS2 variants have lower penetrance and higher prevalence in the general population. In the LS-unrelated group, the prevalence probably more resembles the prevalence of pathogenic MMR mutations as seen in the normal population, in which PMS2 variants are most prevalent. The predominance of PMS2 in the LS-unrelated group likely reflects background distribution, and most of these tumors might indeed be “LS-unrelated” in biology. This fact does not exclude that tumors in the LS-unrelated group are caused by MMR deficiency, but it is not unlikely that in most cases the tumors ack a second hit and are MMR-proficient. Future studies are needed to clarify which tumors in Lynch syndrome carriers are causally linked to mismatch-repair deficiency and which are not.

The mean age of the Non-CNS LS-unrelated tumor group was 55.5 years, and the proportion of patients with LS among all cancer types in this group was low, with a maximum of 0.6%. This group had the smallest number of mutations, and the TMB in this group was significantly lower than that in the other two groups. In addition, the TMB of samples with MSH2 and MSH6 germline variants in this group was lower than that in the other two groups, and the TMB of PMS2 germline variant samples in this group was lower than that in the non-CNS LS-related tumor group. The TMB of patients harboring the MLH1 gene was greater than that of patients harboring other genes in this group. MSSs accounted for 83% of this group, and MSH-H existed mostly in samples with MLH1 variants. Non-CNS LS-unrelated tumors exhibit molecular features (late onset, low TMB/MSS) consistent with sporadic cancers not driven by MMR deficiency, which indicates that these tumors may be truly unrelated to LS.

People with LS should receive lifelong cancer screening starting in adulthood, and should be counseled in the proper way, irrespective of a cancer diagnosis or diagnosed tumor type. Surveillance of non-CNS LS-related tumors, such as CRC, EC, and urothelial cancer, has been recommended for individuals with LS at an early age16. However, there is not enough evidence to support increased screening above average risk in Non-CNS LS-unrelated tumor screening recommendations. Our findings suggest that patients with non-CNS LS-unrelated tumors should not be grouped together with patients with non-CNS LS-related tumors when considering genetic counseling or clinical management, and the later onset of non-CNS LS-unrelated tumors may also justify specific guidelines for surveillance that are tailored to this tumor type. Revised international clinical guidelines for MMR carriers for different tumor groups should be based on a comprehensive assessment of their associated cancer risks.

LS is an autosomal dominant genetic disorder linked to a high risk of cancer and appears to be common in multiple ethnicities1. We compared the occurrence ratio of different cancer types in the Simceredx cohort to that in the TCGA cohort. The proportion of patients with EC in our cohort was significantly greater than that in the TCGA cohort, while the proportion of patients with liver cancer was significantly lower than that in the TCGA cohort, which may be related to ethnicity33. Significant differences in genomic mutation profiles, mutation types and genes with high frequencies of mutations among the three groups were also observed. This observation implies that LS exhibits significant heterogeneity, indicating that its diverse pathogenesis is driven by distinct genetic alterations. The observed differences in age at onset, MSI, and TMB can serve as pilot data for future risk-prediction models and guideline revisions. This study offers valuable reference points for future mechanistic investigations and clinical decision-making, providing a robust hypothesis-generating foundation for both basic research and prospective trials. To improve the management of CNS LS-related tumors and non-CNS LS-unrelated tumors, further comprehensive multiomics analyses are needed.

In the present study, we found that ∼4.4% (10/229) of LS patients had co-occurring pathogenic variants in non-MMR hereditary cancer genes. These co-occurring mutations were predominantly associated with the DNA damage repair pathway. The identification of a second pathogenic mutation in the cancer susceptibility gene is an unexpected finding. Most hereditary cancer syndromes are Mendelian disorders characterized by complex phenotypes due to the influence of additional independently inherited genetic variations and/or environmental factors43. An increasing number of clinical phenotypic overlaps among distinct inherited cancer syndromes have been documented29,44. Among the patients included in our cohort, 9 were diagnosed with multiple primary tumors. Notably, a significant proportion of these patients (2/3) exhibited tumors in distinct organs. Individuals with LS exhibit an elevated lifetime risk of developing various malignancies and multiple primary cancers compared to the general population, and the risk of multiple secondary tumors in patients with LS is of concern9,45. For patients diagnosed with LS, multigene panel testing of multiple primary tumors is recommended to ascertain the underlying etiology of the neoplasms.

There are several limitations to our study. First, our study used DNA-based NGS tests, which have some limitations in the detection of some types of variation, such as large genome rearrangements (LGRs) and epigenetic changes in MMR genes, although these variants are rare but often associated with the etiology of LS46,47. Second, constitutional mismatch repair deficiency (CMMRD) has been identified in individuals with biallelic pathogenic germline mutations in MMR genes48,49. In our study, among the 20 patients under the age of 25, two patients had two germline variants in the same MMR gene (PLS081 PMS2 P and LP; PLS227 MLH1 LP and VUS, respectively), which strongly suggests a diagnosis of CMMRD. It has been reported that primary mismatch repair-deficient gliomas in children, adolescents, and young adults represent a distinct patient group with similar characteristics that can benefit from immune checkpoint inhibitor therapy50. Given the small number of only two cases with an extremely low population incidence of CMMRD, we have not excluded them at this stage. In the future, the newly published method39,51 can be considered to collect more confirmed cases of CNS CMMRD and LS according to the current differential diagnosis recommendations of CMMRD52, and explore and distinguish their characteristics, which have important clinical value. However, our study was retrospective and did not include family verification, so the diagnosis of CMMRD of the two patients could not be conclusively confirmed.

In conclusion, our study investigates the clinical and genomic characteristics and the similarities and differences among CNS LS-related tumors, non-CNS LS-related tumors, and non-CNS LS-unrelated tumors, which hold paramount importance for patient management and precision-based therapeutic interventions.

Methods

Patient enrollment

Our study was approved by the Ethical Committee of Xijing Hospital of Digestive Diseases, Fourth Military Medical University (KY20252370-F-1) and the Institutional Review Board of Nanjing Simcere Medical Laboratory Science (NSML-IRB-202406-MS03). All adult participants or guardians provided written informed consent. Neither sex nor gender was considered in the study design, since the primary focus of this study was unrelated to sex or gender. Participants received no compensation.

A total of 229 patients diagnosed with LS harboring heterozygous germline P/LP variants in at least one of the MMR genes (MLH1, MSH2, MSH6, and PMS2) were enrolled in this study. The 239 tumor samples from these 229 LS patients included 19 samples from 9 patients with multiple primary cancers. One sample from a patient with cancer of unknown primary origin was excluded. Considering that samples from different sites in patients with multiple primaries might be assigned to different groups, 238 samples from 228 LS patients were ultimately enrolled and used for all subsequent analyses. Baseline characteristics information, including age, gender and cancer types of all patients, was collected. Among the 228 enrolled patients, 132 were male and 96 were female, with a median age of 52.5 years (range: 0–79).

DNA extraction, library preparation and genomic sequencing

Frozen fresh tumor tissues or formalin-fixed paraffin-embedded tumor slides of 238 tumor samples and paired blood samples were collected and sequenced at Nanjing Simcere Diagnostics Laboratory (Simceredx, Nanjing, China), the College of American Pathologists (CAP, Certificate ID: 1856000) and Clinical Laboratory Improvement Amendments (CLIA, Certificate ID: 99D2289380) co-accredited laboratories, using hybridization capture-based targeted NGS of the 551-gene panel (n = 181), 131-gene panel (n = 46), or the 353-gene panel (Neuro-Onco360) (n = 11). The gene lists for each panel are provided in Supplementary Data 3.

A Concert® gDNA Tissue Extraction Kit was used to extract genomic DNA (gDNA) from tumor tissues, and gDNA from paired leukocytes was extracted using a magnetic gDNA kit (TIANGEN®). Then, the gDNA was quantified through a Qubit dsDNA HS Assay Kit with a Thermo Fisher Scientific Qubit Fluorometer, and its quality was evaluated using an Agilent 4200 TapeStation (Agilent). Two hundred nanograms of the above-qualified gDNA were enzymatically snipped into fragments of 200–300 bp. The ends of the gDNA fragments were then repaired using the KAPA Hyper DNA Library Preparation Kit (Roche Diagnostics). The VAHTSTM Universal DNA Library Prep Kit for Illumina® (Vazyme) was used to conduct A-tailing and to produce indexed paired-end adaptors for the SimcereDx Illumina platform. After eliminating unwanted contaminants using Agencourt AMPure XP beads (Beckman Coulter), a pre-library was generated by polymerase chain reaction (PCR).

The 150 bp paired-end sequencing of the final qualified DNA library was performed using the Illumina NovaSeq 6000 platform according to the manufacturer’s recommendations. The fastp software (V.2.20.0) can obtain various quality control indicators for each sample, trim the adapter, and eliminate low-quality bases53. The following step was the alignment of the paired-end read sequences with the UCSC hg19/GRCh37 reference genome using the Burrows–Wheeler Aligner (BWA-MEM v.0.7.17) algorithm.

Bioinformatics analysis of sequencing data

Subsequently, we called and annotated the single-nucleotide variation (SNV) and insertion/deletion (Indel) mutations using VarDict (v.1.5.7) and InterVar tools, respectively54,55. We then screened public databases, such as variants in the 1000 Genomes Project (August 2015) and the Exome Aggregation Consortium Browser28, for common single-nucleotide polymorphisms. The fusion genes were accessed through Factera (v1.4.4)56, while the copy number variations (CNVs) were obtained using CNVkit (dx1.1)57.

The cutoff set for the minimal somatic variant allele frequency was 2%. The tumor mutation burden (TMB) was defined as the sum of nonsynonymous somatic mutations in the protein-coding region per megabase (muts/Mb). The MSI was determined based on the homopolymer repeat loci on the detected panel with sufficient coverage, and it was calculated as the percentage of unstable loci58. According to the thresholds, the MSI values were then transformed into MSI-high (MSI-H), MSI-low (MSI-L), and MS-stable (MSS) values.

Germline variants sequenced by paired blood samples were screened by VarDict (v.1.5.7) and then annotated to public databases, including gnomad (v3.1.2), CLINVAR (202308), dbNSFP (v42a), COSMIC (v98), and the Simceredx database. After filtration, the pathogenicity assessment for each mutation was performed according to the integrated outcomes of InterVar and CLINVAR databases. According to the American College of Medical Genetics and Genomics (ACMG) guidelines, germline mutations were stratified into five categories: P, LP, benign, likely benign, and variants of unknown significance (VUS), in which P and LP variants are typically selected for further analysis59.

Lynch syndrome analysis in The Cancer Genome Atlas (TCGA) cohort

We obtained data from the publication by Huang et al.33, including information on cancer types, germline genes with P/LP variants, specific mutations and variant classifications and so on. From this dataset, we identified samples with germline variants in MMR genes, which are associated with LS. We then calculated the incidence of LS across different cancer types and further compared these incidences with those in our own cohort.

Statistical analysis

R version 4.3.1 and Excel were used for all analyses. The top mutated somatic genes and germline variant genes were analyzed and visualized utilizing the R package maftools and ComplexHeatmap60,61. Other figures were generated using the R packages ggplot2 and Excel. The differences between the continuous variables were analyzed using the Wilcoxon test, while the chi-square test was chosen for the large sample analysis. The differences between different groups of the matrix pie charts were generated through Fisher’s exact test. A p-value < 0.05 was considered to indicate statistical significance.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.