Diagnosis of Lynch Syndrome and Strategies to Distinguish Lynch-Related Tumors from Sporadic MSI/dMMR Tumors

Simple Summary Microsatellite instability (MSI) is a hallmark of Lynch syndrome (LS)-related tumors but is not specific, as most of MSI/mismatch repair-deficient (dMMR) tumors are sporadic. Therefore, the identification of MSI/dMMR requires additional diagnostic tools to identify LS. In this review, we address the hallmarks of LS and present recent advances in diagnostic and screening strategies to identify LS patients. We also discuss the pitfalls associated with current strategies, which should be taken into account in order to improve the diagnosis of LS. Abstract Microsatellite instability (MSI) is a hallmark of Lynch syndrome (LS)-related tumors but is not specific to it, as approximately 80% of MSI/mismatch repair-deficient (dMMR) tumors are sporadic. Methods leading to the diagnosis of LS have considerably evolved in recent years and so have tumoral tests for LS screening and for the discrimination of LS-related to MSI-sporadic tumors. In this review, we address the hallmarks of LS, including the clinical, histopathological, and molecular features. We present recent advances in diagnostic and screening strategies to identify LS patients. We also discuss the pitfalls associated with the current strategies, which should be taken into account to improve the diagnosis of LS and avoid inappropriate clinical management.


Introduction
Microsatellite instability (MSI) is a hallmark of tumors developed in the context of Lynch syndrome (LS). LS is a hereditary cancer syndrome with an estimated population frequency of 1:300 individuals. It is characterized by an increased risk of colorectal cancer (CRC) and endometrial cancer but, also, of ovarian, hepatobiliary tract, upper urinary tract, small bowel, gastric, pancreatic, brain, and skin cancers, more often occurring at a young age [1][2][3][4].
LS is mainly caused by dominant germline inactivating variations in the DNA mismatch repair (MMR) genes MLH1, MSH2, MSH6, or PMS2. As a consequence, tumors from LS patients display MSI and a loss of expression of MMR proteins. Diagnosis of LS is important, since patients with LS benefit from adapted protocols of clinical management with surveillance and more radical surgery [5][6][7]. Preventive measures include regular colonoscopic and gynecological surveillance of MMR pathogenic variant carriers, with the frequency and age of initiation dependent on the mutated MMR gene. Patients with LS may also benefit from different approaches for adjuvant therapy [8,9]. However, MSI is not restricted to LS. In fact, only 15-20% of MSI/MMR-deficient (dMMR) tumors can Table 1. Amsterdam II criteria [11] and revised Bethesda Guidelines [15].

Amsterdam II Criteria
(1) Three or more relatives with histologically verified LS-associated cancer (colorectal cancer, cancer of the endometrium, small bowel, ureter, or renal pelvis), 1 of which is a first-degree relative of the others.
(2) Cancer involving at least 2 generations. (3) One or more cancer cases diagnosed before the age of 50. (4) Familial adenomatous polyposis should be excluded.

Revised Bethesda Criteria 1
(1) CRC or endometrial cancer diagnosed at age younger than 50.
(2) Presence of synchronous or metachronous CRC or other LS-associated tumors, regardless of age.
(3) CRC with MSI-high pathologic-associated features (Crohn-like lymphocytic reaction, mucinous/signet cell differentiation, or medullary growth pattern) diagnosed in an individual younger than 60. (4) Patient with CRC or LS-associated tumor diagnosed below 50 in at least 1 first-degree relative. (5) Patient with CRC or LS-associated tumor diagnosed at any age in 2 first-degree or second-degree relatives. 1 Only one of these criteria needs to be met. CRC: colorectal cancer, LS: Lynch syndrome, and MSI: microsatellite instability.
It has nevertheless been shown that a non-negligible proportion of patients with cancer linked to LS were not detected with these criteria and that improvement of the screening strategies was needed [16,17]. Thus, current practices tend towards the universal screening of all CRC and endometrial cancers, especially since the result of a tumor analysis can also have an impact on the therapeutic strategy [14,[18][19][20][21][22][23][24]. Universal screening is also under consideration for tumors from other organs [25,26].
The morphological features observed in MSI endometrial carcinomas include the presence of dense peritumoral lymphocytes (PTLs), prominent TILs, and tumor heterogeneity (defined as two morphologically distinct tumor populations), which are features also identified in MSI/dMMR colorectal cancers. The TIL count has a sensitivity of 85% in predicting the MSI status in endometrioid endometrial carcinoma [47,48]. Some other characteristics have been described, such as a tendency to be located at the lower uterine segment, a higher histological grade, myometrial invasion, endometrioid type, isthmic involvement, and dedifferentiation [41,42,[48][49][50], but with conflicting results and the absence of a significant association with the MSI status in some of the studies [40,47,51].
Studies investigating the pathological characteristics specific to LS-associated endometrial cancers are limited by the small size of cohorts and the variability in the definition of the controls groups. Tumors in LS are less-often diagnosed at an early stage (57% stage I) [18]. A trend towards more non-endometrioid tumors (particularly, clear cell carcinomas, rare serous carcinomas, and malignant mixed Mullerian tumors or carcinosarcomas) and less mucinous differentiation was observed in LS patients compared to sporadic tumors [42,52,53]. Mills et al. showed that more than half of the LS-related endometrial tumors (58%) did not have MSI tumor features, i.e., lower uterine segment location, tumor heterogeneity, TILs, and PTLs [19]. However, a recent study showed a significantly higher density of infiltrating immune cell effectors in LS-associated endometrial cancers compared to sporadic MMR-deficient endometrial cancers, with more CD8+, CD45RO+, and PD1+ T-cells at the invasive margin [51]. In conclusion, morphologic features are not sensitive enough to detect LS-related endometrial cancers, and universal screening is recommended [18][19][20]54,55].  [56]. Aysal et al. showed that morphologic criteria such as TILs, PTLs, and dedifferentiated morphology are not sensitive enough to detect MSI/dMMR ovarian cancers, as these features are present in only 14% of the MSI/dMMR ovarian cases [57]. However, dMMR ovarian cancers have been shown to exhibit significantly increased CD3+ and CD8+ TILs and PDL1+ intra-tumoral immune cells [58].

Urothelial Carcinomas
About 3-12% of upper tract urothelial carcinomas (UTUC) exhibit dMMR, which most commonly manifests as a dual loss of MSH2 and MSH6 or isolated MSH6 loss (up to 86%) [25,[65][66][67]. Among tumors with a loss of MMR proteins, LS represents about 20-30% of the cases [65,67,68]. UTUC is the most common urologic malignancy, occurring in 5% of patients with LS [69,70]. Patients with LS are 22 times more likely to develop UTUC than the general population [71] and have a lifetime risk of developing UTUC of up to 15-20%, with the highest risk among patients with MSH2 mutations (16% vs. 3% to 4% for MLH1 or MSH6) [2]. Some morphologic features have been described in an attempt to discriminate dMMR and MMR-proficient (pMMR) UTUC: increased intra-tumoral lymphocytes, lack of nuclear pleomorphic, inverted growth pattern (or endophytic), and the presence of pushing borders. The inverted growth pattern, at least focally, and increased intra-tumoral lymphocytes have been shown to be predictive of MSI in some studies [66,72], but they did not reach statistical significance in others [25]. Otherwise, there is no significant difference in terms of tumor grade, stage [25], or age at onset, which is comparable to the general population (about 65-70 years old) [65][66][67]. Finally, there are too-few studies with a significant number of genetically characterized patients to identify the characteristics specific to LS. Bladder urothelial cancer as an LS-related cancer is a subject of debate, but a study showed statistically significant differences in their frequencies, with more cancers in MSH2 mutation carriers (4.4%) compared to MSH6 (1.7%) or MLH1 (2.2%) mutation carriers [73].

Non-Colorectal Digestive Cancers
The prevalence of MSI/dMMR gastric cancers is estimated at 9-13% [74,75]. MSI/ dMMR gastric carcinomas tend to occur as poorly differentiated adenocarcinomas in the antrum of elderly patients and to display abundant T-cell infiltration, intestinal histological type, a lack of lymph node metastases, and TNM (Tumor Node Metastasis) stage I-II [75,76]. In LS, the cumulative risk of developing gastric cancer is estimated to be 2.5-6% [2,4]. LS-associated gastric cancers are mostly of the intestinal type, but the diffuse type is also observed [77]. Helicobacter pylori infection has been reported in about 20% of gastric cancers in LS, with no difference from the general population. The clinical factors associated with the risk of gastric cancer in LS patients are the existence of first-degree relative(s) affected by gastric cancer ( or mixed) are not significantly different between microsatellite stable (MSS) and MSI cases. MSI tumors are significantly more likely to show a solid, mucinous, or signet ring pattern, compared to the typical invasion pattern in MSS tumors (defined as invasive glands without any of the previously mentioned features) [79,80]. In LS, the cumulative risk of cholangiocarcinoma ranges from 0.6% to 3.7% [1,2]. There is a lack of data regarding the characteristics associated with LS-associated cholangiocarcinomas.
Small bowel adenocarcinoma is a rare tumor, about 20% being MSI/dMMR. MSI/ dMMR tumors are generally in the duodenum or jejunum, associated with good prognosis, and nonmetastatic [81]. More than 60% of MSI/dMMR small bowel adenocarcinomas may be LS-related according to the family history and MMR protein expression pattern [81]. The lifetime risk for LS patients to develop small bowel cancer is estimated to be 4-12% [2,4,82]. LS-associated tumors are mainly located in the duodenum (43%) and the jejunum (33%), only 7% being located in the ileum [82].
An MMR deficiency is rare in pancreatic cancer, being observed in 1% to 2% of pancreatic ductal adenocarcinomas, with a larger proportion in intraductal papillary mucinous neoplasm (IPMN)-related tumors (about 7%) [83]. MSI/dMMR pancreatic carcinomas show a higher density of CD8+ T-cell lymphocytes at the invasive front compared with MSS/pMMR tumors. The cumulative lifetime risk of pancreatic cancer in LS varies between 0.5% to 3.9%, depending on the mutated gene [2]. LS-associated pancreatic tumors often have a characteristic medullary appearance, with prominent lymphocytic infiltration [84,85]. Pancreatic ductal adenocarcinoma is the predominant tumor type associated with LS [86], even though IPMN have also been reported in LS [87][88][89].

Sebaceous Tumors
Sebaceous tumors, including sebaceoma, sebaceous carcinoma, keratoacanthoma with sebaceous differentiation, and basal cell carcinoma with sebaceous differentiation, are relatively uncommon in the general population [90]. The majority of tumors are solitary and sporadic, with a predilection for the head and neck area of elderly adults [91]. The prevalence of MMR deficiency in sebaceous tumors ranges between 20% and 60%, with a predominance of MSH2 and MSH6 protein loss [92][93][94][95]. Regarding the histopathological characteristics of sebaceous tumors, it has been shown that the number of TILs was increased in dMMR sebaceous neoplasms compared to pMMR sebaceous neoplasms (16.5 vs. 9.7, p = 0.027) but with overlaps between both lesions that prevent using TILs as an indicator of the MMR status [93]. A large subset (39-62%) of sebaceous neoplasias is part of Muir-Torre syndrome (MTS), a phenotypic variant of LS characterized by sebaceous neoplasia and at least one visceral malignancy [96]. Sebaceous neoplasms are observed in 5-9% of patients with LS/MTS [97]. The mean age for the first sebaceous neoplasia is about 55 y.o. in MMR mutation carriers, compared to 67 y.o. in patients without a mutation [92,94,98]. The association with LS/MTS is higher for multiple sebaceous tumors, sebaceous adenomas (vs. sebaceous carcinomas), tumors with a keratoacanthoma-like architecture, and an onset before 60 y.o. Although locations outside the head and neck seem more specific to MTS, many cases are described in the head and neck [98]. Sebaceous tumors precede a visceral malignancy diagnosis in 22-60% of cases [99], supporting that the screening of sebaceous neoplasia provides an opportunity to identify LS patients [90][91][92]94,100,101]. To identify LS, a universal routine screening of sebaceous neoplasms has been proposed, regardless of tumor location, age at diagnosis, or other clinical characteristics, especially since sebaceous tumors are rare in the general population [26].

Brain Tumors
The proportion of MSI/dMMR brain cancers is very low (2% to 3%) [8,102]. Some studies focused on these tumors (related to somatic MMR mutations, to LS, or to CMMRD (Constitutional MisMatch Repair Deficiency) syndrome, which is caused by biallelic germline MMR mutations) [103,104] and showed that they can acquire a secondarily proofreading defect due to a mutation in the POLE (DNA polymerase epsilon, catalytic subunit) gene, Cancers 2021, 13, 467 6 of 33 resulting in an ultra-mutant phenotype (>100 mutations/Mb) and a unique mutational signature ("MMR first/POLE second") [103][104][105]. Additionally, these tumors seem to be more frequently of a particular and very rare histotype characterized by the presence of multinucleated giant cells [106].
Brain tumors are rare in LS, with a cumulative lifetime risk at 70 y.o. estimated to be between 0.5% and 3.7% (the higher risk is for male MSH2 mutation carriers). However, the question of LS and CMMRD often arises for patients with brain tumors, especially because these tumors often occur at an early age. [2,3,107]. No information is available regarding the specific molecular or phenotypic characteristics of LS-(or CMMRD)-related brain tumors. The histological variant of glioblastoma with multinucleated giant cells is also not specific [106,108,109], but given its rarity (approximately 1% of all glioblastomas), there may be a particular interest in proposing a constitutional MMR gene (and POLE) analysis in patients with a giant cell glioblastoma. The molecular hallmark of LS is MSI, which is a consequence of failure of the DNA MMR system that repair errors occurring during DNA replication. Therefore, MSI testing is currently used to evaluate the functionality of the MMR system. Different panels of microsatellite markers have been used to detect MSI in tumors in order to identify LS. However, two panels have been highly recommended and are widely used, i.e., the "Bethesda panel" (also known as the National Cancer Institute (NCI) panel), which combines two mononucleotide repeats and three dinucleotide repeats [110], and the "pentaplex panel", which contains five nearly monomorphic mononucleotide markers [111,112]. Compared to the Bethesda panel, the pentaplex panel has been shown to be more sensitive to detect instability (89-96% vs. 76-84%) [21,113,114] and does not require the testing of matching normal DNA [112,115]. Mononucleotide markers also seem more specific to detect LS-associated cancers (specificity about 99%) [113,116] and are thus considered the current standard because of their higher accuracy [21].
The clinical sensitivity of the MSI test for the identification of patients with proven LS varies greatly depending on the mutated gene, the tumor type, and the threshold (i.e., the number of positive markers) used to define a tumor as positive [110,114].

•
Gene. The sensitivity of MSI has been shown to be lower in MSH6 mutation carriers (77% with a pentaplex panel) compared to MLH1 and MSH2 mutation carriers (~90%). This is due to fewer unstable markers and a shorter instability length (in bp) observed for MSH6 deficiency [21,114,[117][118][119]. In contrast, no particular difficulties have been described for the three other genes, including PMS2 [120]. • Tumor type. The sensitivity of MSI has been shown to be lower in non-CRC tumors, especially in endometrium, brain, and urothelial tumors, as a consequence of fewer unstable markers and a shorter instability length [25,69,117,118,[121][122][123][124][125]. For example, MSI is particularly difficult to detect in brain tumors, with less than 20% of brain tumors from patients with LS (or CMMRD) exhibiting MSI [69,105,[126][127][128]. These observations have led to the recommendation to analyze normal DNA in parallel for non-CRC tumors [114,122,124,125]. • MSI-High vs. MSI-Low. Two categories of MSI were initially defined, based on the number of positive markers, i.e., MSI-high (MSI-H), corresponding to an instability at two or more out of the five markers tested, and MSI-low (MSI-L), corresponding to an instability at only one out of the five markers tested) [110,114]. MSI-L colorectal tumors were not shown to differ in their clinicopathologic features or in most molecular  [129]. However, the sensitivity for LS detection generally increases when MSI-L is considered to be a positive test result (>90% vs. 67-100%) but with lower specificity (45-85% vs. 61-93%) [12,123,[130][131][132][133]. As MSI testing is a screening test and not a final diagnostic test, some authors suggest to take into account the MSI-L results to perform constitutional genetic testing [131].
Determination of the MSI status using next-generation sequencing (NGS) and the analysis of sequencing reads at (designated) microsatellite regions has been proposed as an alternative to a Bethesda or pentaplex panel analysis (by a PCR and fragment-length analysis). Although the available data show promising results in detecting MSI tumors in CRC [134], there is a lack of data regarding the performances in other tumors of the LS spectrum, and this requires additional studies [102,135,136].

Immunohistochemistry
Immunohistochemistry (IHC) is also currently used in clinical practice for LS screening. Compared to MSI testing, it allows identification of the putative causative gene/protein. The presence of positive nuclear staining in any tumor cell, with appropriate staining of the internal control cells, is interpreted as a normal expression, whereas a loss of expression with positive internal control cells defines a dMMR tumor. Although some authors suggested to screen tumors only with antibodies against MSH6 and PMS2 proteins (two-stain method) to reduce costs [137,138] (with the subsequent staining of the partner if either is absent), it is generally recommended to test the four MMR proteins, since the two-stain immunohistochemical screening may fail to detect mismatch repair deficiency in some LS tumors [115,139]. For example, staining weaker than the control may be incorrectly interpreted as an intact MMR protein expression, or focal/patchy MSH6 can be retained in the absence of MSH2 [139].
The sensitivity of the IHC test for the identification of patients with proved LS has been estimated to be 83-100%, depending on the studies [17,21,140,141]. However, some expression patterns can lead to misinterpretations. A retained protein expression (diffuse strong intact staining) is well-documented in some cases of MLH1 mutation or hypermethylation [142,143] but can be found with all four MMR genes. The most frequent cases correspond to pathogenic missense variants associated with nonfunctional protein but retained antigenicity [34,[144][145][146], but some cases also involve truncating variations. Certain variants are known to be invariably associated with retained expression of the defective protein, which may lead to the misdiagnosis of LS patients [145][146][147][148]. Approximately 6% of MSI cases escape detection by IHC because of a retained MMR protein expression [146,149]. A weak or partial tumor staining compared to the internal control can also be observed in LS-associated tumors and be a cause of misinterpretation. For example, weak or partial MLH1 staining in MLH1 mutation carriers was reported in 34% of the tumors by Mangold et al. [143]. Sarode et al. and Watson et al. showed that 42-50% of the cases with indeterminate expressions were found to have the MMR germline mutation, the most frequent being the MLH1 germline mutation, followed by the MSH6 mutation [34,150,151]. Thus, weak tumor staining can contribute to missed LS cases when using MMR IHC screening alone. Conversely, preoperative chemotherapy or chemoradiation therapy can be associated with a reduced MSH6 protein expression in the absence of a germline or somatic MSH6 mutation in MSS CRC, leading to the overestimation of dMMR cancers and LS cases [152,153].

Comparison of MSI and IHC for Identification of LS
The concordance between MSI testing and MMR protein IHC results vary depending on the panel used for MSI and the cancer type but is generally very high, especially for CRC (between 92 and 99%) [17,147,149,150,[154][155][156]. A notable exception concerns brain tumors, where IHC has demonstrated its superiority compared to MSI in identifying Cancers 2021, 13, 467 8 of 33 dMMR glioblastomas but with global limited performances for both technics in these tumors [69,126].
Neither the MSI test nor MMR protein IHC has a sensitivity of 100% for the identification of LS cases. An optimal sensitivity is obtained by combining the two tests, which may accurately identify close to 100% of LS-related tumors. Therefore, a combination of the two techniques should be recommended when possible to increase the chances of LS detection [16,20,115,149].
Finally, it is important to keep in mind that a negative test does not exclude definitively the possibility of LS and that the results should be interpreted with the clinical data, i.e., age at diagnosis and personal and family history of LS-related cancers.

Focus on Colorectal Carcinogenesis and Adenomas in LS
Whereas sporadic MSI CRC arise through the serrated pathway, LS-related CRC arise through the conventional pathway [33]. Some conflicts exist about the natural history of CRC in LS. Classically, MMR deficiency is thought to be acquired during the progression from early to advanced adenomas and to be involved in the acceleration of adenoma progression, since tumor progression in many LS patients takes less than three years, contrasting with a mean of 15 years in the general population [157,158]. This hypothesis suggests that tumor initiation does not depend on MMR implication and that the second gene inactivation hit occurs after the formation of polyps, caused by other events such as APC mutations [159]. The correlation of MMR deficiency with a bigger size and higher grade of the adenomas, the presence of more extensive villous architecture, and the prevalence of a subset of small, low-grade MMR-proficient adenomas support the concept of the loss of MMR function as a relatively late event in LS-related CRC [157,160,161].
These findings have been challenged by arguments pointing to the role of MMR mutations in tumor initiation, such as the discovery of a loss of expression of MMR proteins in apparently normal crypts in 25-35% of patients with LS-related CRC [162][163][164]. Indeed, MMR-deficient non-neoplastic intestinal crypt foci (MMR-DCF) were reported to detect LS with excellent specificity (≥99%) [163][164][165]. This implies that MMR protein loss might precede adenoma formation. Furthermore, a high rate of MMR deficiency in LSassociated adenomas (around 70-80%) [157,160,166,167], the complete and homogeneous loss of MMR protein expression in a majority of adenomas, and the rarity of adenomas with a focal loss of MMR proteins (3% in Ahadova et al. [160]) suggest that the MMR system is involved in the initial development of the adenoma as an early event. Nevertheless, it is not clear whether MMR-DCF are cancer precursors in LS, even if MSI was detected in 89% of MMR-DCF [162].
A third pathway, with a direct invasive lesion from the dMMR crypt, could explain the high frequency of interval cancers in patients under regular colonoscopic surveillance [31].
Finally, these conflicting observations could be explained by the existence of multiple tumorigenesis pathways in LS, as proposed by Ahadova et al. [160].
To conclude, even though MMR-DCF detection seems to show an excellent specificity because of the relatively low sensitivity (30-40%) [165], additional studies are needed to demonstrate if it could serve as a biomarker for LS. A study has nevertheless already shown that the sensitivity can be increased with the number of non-neoplastic colonic crypts studied (up to 70% for a median number of colonic crypts analyzed per patient of 3250) [164].
In practice, the combination of MSI and IHC testing in colorectal adenomas allows to screen LS patients with a sensitivity around 70-80% and may be particularly useful when LS is suspected and adenomatous polyps are the only tissues available for analysis [157,160,161,166]. Nevertheless, a negative result does not exclude the presence of LS, especially in the case of low-grade dysplasia adenomas.

Clinical Criteria
The hallmarks of a hereditary cancer syndrome include early age at diagnosis, multiple affected family members, and an increased risk of cancers associated with the syndrome. Accordingly, the usual approach to diagnose LS patients is to use the patients' personal and family history to guide towards MMR genetic testing.
Several prediction models based on personal and familial clinical data, some including tumor data (MSI and MMR protein IHC), have been developed to evaluate the risk for a given patient to carry a MMR mutation, e.g., MMRpredict [12], PREMM [168], and MMRpro [169]. However, Tresallet et al. evaluated those algorithms in patients with newly diagnosed CRC and showed that the performances of these prediction models were not better than those of the revised Bethesda Guidelines, some LS-patients being missed by all models (elderly patients with no family history) [170]. These results support systematic MSI/dMMR tumoral screening in all CRC.

Molecular Genetic Diagnosis
LS is caused by heterozygous germline inactivating variations in one of the four key DNA MMR genes: MLH1, MSH2, MSH6, and PMS2. In 2015, MLH1, MSH2, MSH6, and PMS2 accounted for 40%, 34%, 18%, and 8%, respectively, of the 3000 unique germline sequence variants of MMR genes deposited in the International Society for Gastrointestinal and Hereditary Tumours (InSiGHT) database [171]. Point mutations are predominant, but large rearrangements can also be the case of LS, representing 10% of the variants in MSH2 and PMS2, 7% in MLH1, and 2% in MSH6 [171]. Variants of unknown significance (VUS) or class 3 variants account for about 20-30% of the MMR variants identified in patients suspected of having LS. For these patients, in vitro MMR functional assays are an important tool to assess the pathogenicity of these variants [172]. Constitutional epimutations of MLH1 and MSH2, characterized by the hypermethylation of the promoter of these genes in normal tissues, are an alternative cause of LS [173][174][175].
The diagnosis of LS is based on the detection of a germline pathogenic mutation or epimutation in a MMR gene. A tumor analysis, i.e., MSI/dMMR status +/− BRAF mutational/MLH1 methylation status, is of great importance to identify the patients who should benefit from MMR germline genetic testing.
Targeted NGS has progressively replaced Sanger sequencing for the molecular diagnosis of LS in most clinical labs, providing multigene panel testing and cost-effectiveness [176]. This strategy can reliably detect nucleotide substitutions and indel within exons and flanking intronic regions and can also detect copy number variations (CNV) using dedicated algorithms. Multiplex ligation-dependent probe amplification (MLPA, MRC-Holland) can be further added to targeted NGS for the detection or the validation of CNV. An analysis of the PMS2 gene is technically challenging due to the presence of highly conserved pseudogenes that complicate the mutation and CNV detection. Strategies to circumvent this issue include long-range PCR followed by nested PCR and Sanger sequencing or by NGS, cDNA sequencing, MLPA adapted to the potential gene conversion between PMS2 and the pseudogene PMS2CL in the 3' exons, and NGS with adapted read alignment and variants calling for exons 12-15 [177][178][179][180][181][182]. Regarding epimutations, their detection requires specifically looking for them with dedicated technics.
Pitfalls: For a substantial proportion of patients with a MSI/dMMR tumors, even those with early onset CRC [183], no germline mutation is detected in the four MMR genes. In these cases, the MSI/dMMR tumor can be the consequence of an undiagnosed hereditary predisposition to cancer or can be a sporadic tumor. Below are the potential origins of MSI/dMMR tumors without germline MMR variations (Figure 1): Pitfalls: For a substantial proportion of patients with a MSI/dMMR tumors, even those with early onset CRC [183], no germline mutation is detected in the four MMR genes. In these cases, the MSI/dMMR tumor can be the consequence of an undiagnosed hereditary predisposition to cancer or can be a sporadic tumor. Below are the potential origins of MSI/dMMR tumors without germline MMR variations (Figure 1): (1) One potential explanation is a germline alteration in the MMR gene that cannot be detected or is difficult to detect using conventional technics. Some of these alterations that require specific methods are well-known, and their detection can be implemented in routine diagnosis. Shortly after the description of germline deletions of the last exon(s) of the EPCAM gene, leading to epigenetic silencing of the downstream gene MSH2 in epithelial cells [184], it was acknowledged that 3′-end EPCAM deletions can be a frequent cause of unexplained MSH2 deficiency [185,186], and EPCAM deletion detection is now included in the routine genetic diagnosis of LS. These deletions are detected by MLPA, as probes at the 3′-end of EPCAM are included in MLH1/MSH2-P003 and MSH6-P072 kits. They can also be detected by applying CNV detection-specific algorithms to NGS data if EPCAM is included in the multigene panel. Germline epimutations of the MLH1 gene (i.e., constitutional hypermethylation of the MLH1 promoter) are occasionally responsible for (1) One potential explanation is a germline alteration in the MMR gene that cannot be detected or is difficult to detect using conventional technics. Some of these alterations that require specific methods are well-known, and their detection can be implemented in routine diagnosis. Shortly after the description of germline deletions of the last exon(s) of the EPCAM gene, leading to epigenetic silencing of the downstream gene MSH2 in epithelial cells [184], it was acknowledged that 3 -end EPCAM deletions can be a frequent cause of unexplained MSH2 deficiency [185,186], and EPCAM deletion detection is now included in the routine genetic diagnosis of LS. These deletions are detected by MLPA, as probes at the 3 -end of EPCAM are included in MLH1/MSH2-P003 and MSH6-P072 kits. They can also be detected by applying CNV detection-specific algorithms to NGS data if EPCAM is included in the multigene panel. Germline epimutations of the MLH1 gene (i.e., constitutional hypermethylation of the MLH1 promoter) are occasionally responsible for the MMR-deficient phenotype in LS patients, and their detection is now part of the genetic diagnosis strategy of LS. It requires dedicated technics, which can be based on bisulfite conversion of the DNA (pyrosequencing or (quantitative) methyl-specific PCR) or on methylation-sensitive restriction enzymes (MS-MLPA) [187]. Additionally, highly sensitive technics are required due to the low level of methylation observed at MLH1 promoter region in peripheral blood mononuclear cell (PBMC) DNA for some patients with a constitutional MLH1 epimutation [175,188] (see, also, Table S4 in [189]). It should be noted that these technics are not used for MSH2 epimutation detection, since the methylation of MSH2 depends on the expression of the EPCAM gene and is consequently mostly detected in epithelial tissues where EPCAM expression is high and not in PBMC.
Other germline alterations require dedicated technics or the extension of existing methods and are not yet part of the routine diagnostic of LS. Some variations may be located within the promoter, regulatory, or deep-intronic regions of the MMR genes, regions that are not commonly analyzed in routine diagnosis procedures. Moreover, even if detected, these variations are still very difficult to interpret, and their clinical significance remains mostly nonconclusive [190]. There are only a few reports of deep-intronic variants with pathogenic significance. A variation located deep within intron 1 of the MSH2 gene (NM_000251.3:c.212-478T>G) has been well-documented in a family with LS and is now considered as pathogenic [191]. The effect on splicing of a variation located within MLH1 intron 15 (NM_000249.4:c.1732-264A>T) has been recently demonstrated [192].
Complex structural rearrangements, especially CNV-neutral ones such as gene inversions and large intronic insertions, are usually not detected with targeted NGS, since their breakpoints are not covered. Paracentric inversions involving MLH1 [190,193] or MSH2 genes [194][195][196] have been described as a disease-causing mechanism in LS. A balanced translocation disrupting MLH1 has been described in a patient from Iceland [197]. The insertion of a 2.2-kb-long retrotransposon in PMS2 intron 7 has also been described [198]. Morak et al. investigated the prevalence in the German population of five founder CNVneutral structural variants (SV) and identified only one patient with an insertion in PMS2 intron 7, concluding that these SV are not frequent [199]. The detection of CNV-neutral SV requires specific methods, such as full-length cDNA or deep-intronic sequencing [190,200], and the optimal strategy is different depending on the type of SV. As the paracentric inversion involving MSH2 exons 1 to 7 can be a frequent cause of unexplained LS in some populations [201], two specific probes have been added to the commercialized MLPA-kit MLH1/MSH2-P003 (MRC Holland) to detect the rearrangement breakpoint reported in intron 7 [194,195]. However only this specific inversion is detected with these probes, and paracentric inversions involving other regions of MSH2 are not [196]. Insertions of Alu elements, such as the ones reported in the coding sequence of MSH2 [202] or MLH1 [188,203], are another type of SV that are not easily detected.
Only a limited number of patients with somatic mosaicism of a MMR gene have been described so far [204][205][206], and the implementation of NGS has not drastically increased the number of these observations. Anyway, analysis pipelines dedicated to mosaic detection in NGS data may be useful for a few patients with no history of cancer in the ascendants, at least to exclude somatic mosaicism. Alternatively, MMR analysis in the tumor and subsequent screen for the pathogenic variants in nontumoral tissues with highly sensitive technics might be the strategy to exclude low-level mosaicism and mosaicism not detectable in lymphocyte DNA [207].
An unidentified germline MMR gene variant remains highly suspected in families with an aggregation of tumors of the LS spectrum, especially when these tumors harbor the same dMMR phenotype.
(2) Germline pathogenic variants in genes other than the four key MMR genes can mimic LS. The variable phenotype of MUTYH-associated polyposis can overlap the LS phenotype [208], and biallelic MUTYH germline variants impairing the base-excision repair (BER) pathway can sometimes lead to biallelic somatic variants in MMR genes and to MSI/dMMR tumors [209][210][211]. Colorectal tumors from patients with a germline pathogenic POLE variant (i.e., variant leading to loss of the proofreading function of the polymerase) are usually MSS [212], but germline variants in POLE were sometimes reported in patients with early-onset MSI/dMMR CRC and somatic variants in MMR genes [213]. Mutations of these genes must be considered for patients with MSI/dMMR tumors and cancer-affected relatives, especially when diagnosed with cancer at a young age. These genes are now included in most of the multigene panels analyzed for LS diagnosis.
(3) MMR deficiency can also be observed in sporadic tumors, as the consequence of somatic biallelic inactivation of an MMR gene. The best-known cause of somatic inactivation leading to sporadic tumors is the epigenetic silencing of the MLH1 gene, which is responsible for a large proportion of MLH1-deficient tumors, especially in older patients [16,17,[43][44][45][46]220].
The coexistence of a pathogenic variant on each allele of an MMR gene or of one pathogenic variant with the loss of the second allele (i.e., loss of heterozygosity, LOH) can also be the cause of MSI/dMMR. Depending on the studies, the proportion of unexplained MSI/dMMR colorectal and endometrial tumors (i.e., tumors without a pathogenic germline variant or somatic MLH1 promoter methylation), which are caused by biallelic somatic genetic alterations can range from 17% to 95% [205,[220][221][222][223][224][225]. It has been shown that, for 25% of unexplained MSI endometrial cancers, a somatic variant in POLE could be the cause of these somatic MMR gene alterations [226]. Biallelic somatic MMR inactivation has also been described as a frequent cause of MSI/dMMR sebaceous neoplasms [227]. It should be noted that, unlike tumors with MLH1 epigenetic silencing, tumors with somatic MMR genetic alterations are not diagnosed at a significantly higher age than those from LS patients [220,223].
Patients with inflammatory bowel diseases are at increased risk of developing sporadic CRC, with about 15% of these tumors exhibiting MSI [228]. Other conditions, such as Hodgkin's lymphoma treated with radiotherapy or procarbazine-containing chemotherapy, can increase the frequency of MSI CRC with somatic MMR gene alterations [229].
When no germline MMR variant has been detected, somatic MMR testing significantly reduces the number of patients considered to have a LS and to require lifelong Lynch surveillance protocols. Only a limited number of these patients are concerned by an inherited predisposition to cancer and have a dMMR tumor secondary to germline defects in genes other than the MMR genes (see (2)) or to somatic mosaicism in an MMR gene (see (1)). Most of these cases are truly sporadic and, consequently, nonheritable and should therefore be identified to prevent relatives from unnecessary anxiety and colonoscopies and other surveillance procedures. Even if not yet implemented in the recommendations for routine diagnostics, somatic MMR genetic testing should be offered to patients with a MSI/dMMR tumor and negative germline testing. In our routine practice, we propose tumoral MMR genetic testing to these patients when the tumor occurs below 60 y.o. A recent study evaluated that nonheritable causes of MSI/dMMR are eight-fold more common than LS in patients with CRC diagnosed between age 60 and 70 and 20-fold more common in patients between age 65 and 70 [220]. Consequently, above 60 y.o., a sporadic tumor is the most likely explanation when no germline variant has been identified with the current diagnosis procedures, and this does not require further tumor testing.
(4) False-positives or the misinterpretation of MSI/IHC test results can be as high as 19% of the unexplained dMMR tumors in some studies [225]. In most cases, MSI testing and immunochemistry showed discordant results, highlighting the need to perform both the test for LS screening and to repeat the test to confirm the discordant results. Before gene panel testing was the standard for a Lynch diagnosis, misleading IHC results leading to selective analysis of the wrong MMR gene could also be a cause of undetected variations [190]. (5) There may also be some other cellular mechanisms leading to MSI in tumoral cells that could be further investigated in order to explain MSI tumors without defects in MMR genes. For example, it has been shown that cells lacking the H3K36 trimethyltransferase SETD2 (SET domain containing 2) display MSI owing to the loss of an epigenetic histone mark that is essential for the recruitment of the MSH2-MSH6 complex [230]. It has also been shown that the phosphorylation of PCNA (Proliferating Cell Nuclear Antigen) can inhibit MMR function [231].
The term Lynch-like syndrome (LLS) has been proposed for the patients who present with an MSI/dMMR tumor without MLH1 promoter hypermethylation when no pathogenic germline MMR gene variation is found. Buchanan et al. combined the results of three different colorectal cancer cohorts and determined that LLS represents 59% (95% CI: 55-64%) of the patients with a MSI/dMMR CRC tested for germline MMR gene variations (433 patients in total after the exclusion of tumors with MLH1 hypermethylation) [232]. They also combined the results of four endometrial cancer cohorts, where LLS represented 52% (95% CI: 41-62%) of the patients with a dMMR endometrial cancer tested for germline MMR gene variations (101 patients in total after the exclusion of tumors with MLH1 hypermethylation). Comparable values, ranging from 44% to 61%, were reported in other studies, including more recent ones [222,[233][234][235][236].
As detailed above, there are different explanations for the absence of germline MMR gene variant detection, and some patients with LLS have cancers of hereditary origin, whereas others have cancers of sporadic origin. A few studies investigated the risk of colorectal and extracolonic LS-associated cancers in LLS patients and relatives and identified a lower risk compared to LS and a higher risk compared to sporadic cancer or to the general population [4,237,238]. However, in these studies, the proportion of cases explained by biallelic somatic variations in MMR genes was not determined, and LLS patients were considered as a single group, even if most likely heterogeneous. Similarly, variable mean ages at onset, between 48.8 y.o. and 65 y.o., were found for LLS cases, depending on the studies [32,222,[236][237][238], and that might be due to variable proportions of undiagnosed germline mutations in these studies.
As our ability to identify genetic variations of various types increases, the number of patients with unexplained MSI/dMMR tumors, i.e., Lynch-like patients, should decrease. Patients with EPCAM deletions illustrate this, since they were considered as Lynch-like before this mechanism of MSH2 inactivation was discovered. Similarly, patients with MSI/dMMR tumors explained by somatic mechanism should not be considered as Lynchlike patients, and patients with documented biallelic somatic variations should probably be excluded from this group, just as patients with biallelic somatic MLH1 hypermethylation are. Thus, the precise definition of the different groups of patients with MSI/dMMR tumors has major implications on clinical management and genetic counseling.

Clinical Presentation
Several clinical features are indicative of LS-related MSI/dMMR cancers: (1) First, personal and family histories of LS-related cancers are strong indicators of LS [20,27,59,233,239,243,244]. Synchronous LS-related tumors are also indicative of LS. The identification of a few additional colorectal polyps at the time of CRC diagnosis is suggestive of LS [239]. Multiple sebaceous neoplasias are a strong indicator of LS/Muir-Torre syndrome [92,94,100,244]. On the other hand, synchronous primary carcinomas of the ovary and endometrium are unlikely to be part of LS [57,245,246]. Certain types of MSI/dMMR cancers are more predictive of LS, e.g., small-bowel adenocarcinoma [233]. More than 60% of MSI/dMMR small-bowel adenocarcinomas may be LS-related [81].
(2) Second, the age of cancer onset, which is much lower in patients with LS. Overall, the mean age of diagnosis is more than 10 years earlier than that of sporadic MSI/dMMR cases (48 y.o. vs. 60.5 y.o.) [239]. For example, ovarian cancers occur at a mean age of 45 y.o., i.e., about 20 years earlier than the sporadic cancers, with more than 60% (62-85%) of the tumors diagnosed before 50 y.o. compared to 13% in sporadic cases and about 90% diagnosed before 60 y.o. [59,60,63]. Therefore, an onset before the age of 50-60 y.o. for a LS-associated type of tumor is a strong indicator of LS. However, about 50% of CRC are diagnosed at an age older than 50 y.o. and 17% after 60 y.o. [16,17], and LS women frequently present with endometrial cancer at an age older than 50 y.o. (50-90%, depending on the studies) or even 60 y.o. (15-45%) [20,[43][44][45]). Similarly, 25% of patients with proven LS/Muir-Torre syndrome present their first sebaceous tumor after 60 y.o. [92]. Moreover, the mean age at cancer diagnosis is much higher in MSH6 mutation carriers (about 56 y.o. for CRC and 55 y.o. for endometrial cancer) compared to MLH1, MSH2, and PMS2 mutation carriers [119,120,247]. Therefore, an early onset is a strong indicator of LS, but LS cannot be ruled out by an older age at diagnosis.
(3) Third, the gender and tumor location for CRC, sporadic MSI CRC being significantly more frequent in elderly women and located in the proximal colon, while LS-related CRC are observed in both genders and occur at equivalent frequencies in the proximal colon and in the distal colon and rectum [27]. Consequently, patients with left-sided or rectal tumor locations are more likely to have LS [32].
(4) The body mass index (BMI) is also an indicator for patients with endometrial cancer. Given that obesity is a known risk factor of sporadic endometrial cancer, a BMI within the normal range may support the hypothesis of LS [45,239].
(5) Finally, several histopathological features are commonly observed in MSI/dMMR tumors, which include a poor differentiation (for CRC) or over-representation of some histotypes (for CRC, endometrium, and ovary) and prominent TILS and PTLs [33,42,47,50,53,56,57,248,249], but these do not discriminate LS-related and sporadic tumors. Some features may be more specific to LS, such as the presence of conventional adenomas in CRC and of MMR-deficient colonic crypts in peritumoral mucosa [162][163][164] and an intra-tumoral immunological pattern with much higher CD8+ immune cells in endometrial carcinomas [51], which may be helpful for the identification of LS.

Molecular Presentation
LS-related and sporadic MSI/dMMR tumors develop through different mechanisms of MMR gene inactivation. As a consequence, sporadic MSI/dMMR tumors display unique molecular features, which are routinely used to distinguish them from LS-related tumors.
Most of the sporadic MSI/dMMR tumors are due to inactivation of the MLH1 gene by methylation of its promoter. An immunohistochemistry analysis of the MMR proteins is thus the first step in discriminating LS and sporadic MSI/dMMR tumors, since the combined loss of MSH2 and MSH6 or the isolated loss of MSH6 or PMS2 will argue for a LSrelated tumor. In contrast, the detection of MLH1 protein loss is not specific and requires a complementary analysis to distinguish a LS-related tumor caused by a MLH1 constitutional alteration from a sporadic tumor caused by acquired somatic hypermethylation of the MLH1 promoter. Of note, zonal heterogeneity of the staining with areas of the tumor showing an abrupt loss of MLH1 and PMS2 protein expression may be an indicator of acquired inactivating alterations of MLH1 (primarily, MLH1 promoter hypermethylation), arguing for sporadic cancer rather than LS-related cancer [250,251]. The isolated loss of PMS2 can also be due to MLH1 promoter hypermethylation [240].
While this reasoning is applicable to the majority of tumors, a notable exception is sebaceous neoplasia. Indeed, most of sebaceous MSI/dMMR tumors display a loss of MSH2 and MSH6 proteins, which is widely considered as highly predictive of LS, but a MSH2 germline mutation is detected in only one-third of cases [92,94,244], suggesting that a significant proportion of MSI/dMMR sebaceous tumors are sporadic. Accordingly, biallelic somatic MMR inactivation has been described as a frequent cause of MSI/dMMR in sebaceous neoplasms [227].

MLH1 Promoter Methylation
As mentioned above, sporadic MSI/dMMR tumors are generally associated with hypermethylation of the MLH1 promoter, which coincides with the CpG island methylator phenotype (CIMP) [252,253]. CpG island methylation is a physiological process related to aging, which concerns a number of specific genes in cancer, among which is MLH1, leading to their inactivation [253]. In contrast, methylation of the MLH1 promoter is rare in LS. A MLH1 promoter analysis in tumors displaying loss of the MLH1 protein is thus very helpful in the distinction of LS-related and sporadic tumors. Importantly, only the methylation status of the proximal region of the promoter is of interest. Indeed, only hypermethylation in the proximal region (usually referred to as regions C and D) correlates with an absence of MLH1 protein expression, region C being the most specific [254,255].
About 75% of MSI and MLH1-deficient CRC unselected for age at diagnosis or a family history of cancer display MLH1 promoter hypermethylation, the proportion of tumors with MLH1 promoter hypermethylation increasing with age [16,17,220]. This proportion ranged from 0% in MSI/dMMR CRC below 40 y.o. to 84% in MSI/dMMR CRC between 65 and 69 y.o. in the large prospective cohort reported by Vos et al. [220]. Conversely, hypermethylation of the MLH1 promoter is uncommon in LS, occurring in less than 6% of tumors from MLH1 mutation carriers (for review, see [242]). Unmethylated MLH1 has a sensitivity of 94% and a specificity of 88% for the identification of MLH1 mutation carriers in patients with a MSI/dMMR CRC with a loss of MLH1 [241].

BRAF Mutations
Eight-ten percent of all CRC display a somatic mutation in the BRAF (MIM*164757) gene. Most of BRAF mutations correspond to the NM_004333.4:c.1799T>A, p.Val600Glu (V600E) (78%), which activates the MAPK pathway [258]. This mutation is tightly associated with MSI and MLH1 promoter hypermethylation in CRC [252]. The BRAF V600E mutation may have a role in the initiation and promotion of colorectal tumorigenesis through the serrated neoplasia pathway [259][260][261]. About 60% (40-73%) of sporadic MSI/dMMR CRC exhibit the BRAF V600E mutation. In contrast, the BRAF V600E mutation is almost never detected in LS-related CRC (1.4%) [241,[262][263][264][265][266][267]. Consequently, for tumors with a loss of the MLH1 protein, the detection of a BRAF V600E mutation is a strong negative predictor of LS. The absence of a BRAF V600E mutation has a sensitivity of 98.6% but a specificity of 66% for the identification of MLH1 mutation carriers in patients with a MSI/dMMR CRC with a loss of MLH1 [241]. Of note, no information is available about non-V600E mutations, which account for about 20% of all BRAF mutations identified in CRC. However, these mutations are uncommon in MSI CRC (6%) and seem to represent a distinct (biologically and clinically) category of tumor [258,268].
BRAF V600E IHC has been proposed as an alternative to the BRAF mutation molecular analysis, with very variable results. Several studies suggested a very high sensitivity and high specificity of V600E-specific IHC for the detection of BRAF-mutated CRC [276][277][278] (sensitivity, 100% and specificity, 98.8-100%), while others pointed out insufficient sensitivity, specificity, and robustness for the use in clinical practice (sensitivity, 51-85% and specificity 68% [279] and sensitivity, 35-71% and specificity, 74% [280]). Anyway, BRAF IHC does not seem robust enough to be currently used as an alternative to the BRAF molecular analysis.

BRAF Mutation Analysis vs. MLH1 Methylation Analysis
BRAF mutation testing is more common than methylation testing in clinical laboratories, since it is commonly used for therapeutic purposes. Moreover, it is technically easier than a methylation analysis. However, the clinical utility of BRAF mutation detection for LS screening is limited, as it is restricted to CRC. Moreover, BRAF mutations concern only about 60% of sporadic MSI/dMMR CRC, which implies that about 40% will necessitate a subsequent MLH1 methylation analysis.
A methylation analysis is technically more challenging than a BRAF mutation analysis (and attention must be paid to the region analyzed). However, the clinical value of the MLH1 promoter methylation analysis is much higher for the identification of MLH1 mutation carriers among patients with CRC [241,242] and is also applicable to endometrial cancers and probably to other tumor types.
Several cost-effective strategies have been proposed to optimize the identification of LS among MSI/dMMR CRC with a loss of MLH1: (1) a BRAF mutation analysis alone, (2) a MLH1 methylation analysis alone, and (3) a BRAF mutation analysis followed by a MLH1 methylation analysis in cases with no BRAF mutation. A MLH1 hypermethylation analysis does not only outperform a BRAF mutation analysis but is also more cost-effective. Using a BRAF mutation analysis as a sole test would increase the referral rates for genetic testing by two-to-three-fold compared with a methylation analysis of MLH1 alone. However, the hybrid approach may facilitate the wider implementation of LS screening without significantly increasing the cost [281][282][283].
Pitfalls: Although MLH1 promoter hypermethylation and BRAF V600E mutation are strong negative predictors for LS, the data must be interpreted with caution.
Regarding MLH1 promoter hypermethylation, although its negative prediction value for LS is high, the detection of MLH1 hypermethylation does not exclude a diagnosis of LS. Indeed, MLH1 promoter hypermethylation may occasionally be the cause of inactivation of the second allele at the somatic level (second hit) in patients with a germline MLH1 mutation [263,284,285]. A quantitative MLH1 methylation analysis has been proposed as a mean to distinguish biallelic and monoallelic methylation [286], but this implies an accurate evaluation of the percentage of tumor cells in the sample, which is challenging in a number of cases. Moreover, MLH1 constitutional epimutation can be the cause of LS [173,175,187]. Similarly, BRAF V600E mutations have been occasionally detected in CRC from proven MMR mutation carriers [187,241,262]. Of note, BRAF mutations may be more common in CRC from patients with MLH1 constitutional epimutation (3/30 cases identified with an MLH1 epimutation [187]) compared to CRC from LS patients with a classical MMR gene mutation. Although rare, one should be aware that the use of BRAF mutation and MLH1 methylation testing to rule out the diagnosis of LS will ignore a few individuals with LS. Moreover, patients with primary MLH1 epimutation generally do not have a family history of cancer. Therefore, germline MMR mutation testing should still be considered in patients with clinical features highly suggestive of LS, and MLH1 epimutation testing should be performed in those with a cancer onset below the age of 60 y.o. and a MLH1-methylated tumor, irrespective of the BRAF mutation status and of the family history.
Conversely, the absence of MLH1 promoter hypermethylation does not always mean that the patient is affected by LS, since MSI/dMMR tumors may alternatively be the consequence of two somatic mutations or of mutations in another gene than the MMR ones (see Section 3). Such pitfalls underline the importance of interpreting molecular data cautiously.

Other Potential Markers
Other markers have been evaluated for their ability to identify LS, some of which showed a potential interest in the distinction of LS and sporadic MSI/dMMR cancers.
Sporadic CRC with MSI and the hypermethylation of MLH1 belong to the group of CIMP tumors [253]. CDKN2A (p16) is among the most frequently methylated loci [253,287,288]. The loss of p16 expression has been shown to be present in 30% of tumors with MLH1 promoter hypermethylation but in none of the tumors from patients with LS, suggesting that p16 immunochemistry may be used as a surrogate marker for MLH1 hypermethylation [289,290].
Certain proteins, such as ANXA10 (Annexin A10), are highly expressed in colorectal tumors, arising through the serrated neoplasia pathway [291]. The expression of the ANXA10 protein is observed in about 45% (42-49%) of sporadic CRC, irrespective of the presence of a BRAF mutation, but in a low proportion of LS-related CRC (5-12%), suggesting that ANXA10 immunohistochemistry may be used as a supportive marker in association with BRAF to distinguish LS and sporadic MSI/dMMR CRC [291,292].
PIK3CA mutations may have be interesting for identifying tumors with two somatic MMR gene mutations, excluding a diagnosis of LS. Indeed, PIK3CA mutations have been shown to be more frequent in CRC with somatic mutations (67% (14/21)) compared to LS-associated CRC (22% (4/18)) (and CRC with MLH1 hypermethylation, 20% (2/10)) [293]. Mutations in PIK3CA were also present in all endometrial cancers with somatic mutations in the MMR genes tested (n = 13) [293]. However, additional studies are needed to evaluate the potential utility of this test in discriminating sporadic and LS-related tumors.
Recently, tumor mutational signatures that provide insights in the etiology of the tumorigenesis processes have been identified [294]. Signature 6 is strongly associated with MSI/MMR deficiency in CRC, but several other signatures have been associated with MMR deficiency in a variety of cancers [294,295]. Signatures 6 and 15 have been shown to be more prevalent in MMR-deficient sebaceous tumors from patients with LS compared with MMR-proficient tumors [296]. However, whether these tumor mutational signatures are able to distinguish LS-related from sporadic MSI/dMMR tumors is currently unknown.

MMR Gene Analysis in Tumors
The risk of cancer in families without a germline mutation has been shown to be lower than the one in families with proven LS but higher than the risk of cancer in families with sporadic cancer and in the general population [4,237,238], justifying surveillance procedures. However, it is now clear that a majority of cases are due to acquired biallelic somatic mutations. Thus, identifying the cause of MSI/dMMR is crucial to guide the clinical management of patients and their families. Further analysis of the four key MMR genes in the tumor of MSI/dMMR mutation-negative individuals and identification of two inactivating mutations in the tumor allow to exclude the patient and their relatives from heavy LS-specific surveillance protocols.
Moreover, in a recent study, the tumor sequencing of MMR genes among other genes, in addition to assessing the BRAF and MSI status in a series of patients with CRC, showed better sensitivity than MSI or IHC, followed by a BRAF mutation analysis (100% vs. 89.7%), with equal specificity (95.3% vs. 94.6%) for the identification of LS [297]. This result needs to be confirmed, as tumor sequencing is technically challenging (especially for CNV detection), and the cost-effectiveness has to be determined. However, this raises the question of introducing tumor sequencing first, with or without an MLH1 methylation analysis, as a replacement for the current LS screening tests. The identification of one or more MMR mutations would lead to genetics counseling and further germline MMR analysis.

Optimal Strategy for Discrimination of LS-Related and Sporadic MSI/dMMR Tumors
A strategy for distinguishing LS and sporadic tumors is proposed in Figure 2. However, it is important to keep in mind that there is no marker allowing the perfect discrimination of LS-related and sporadic cancers in the context of screening. As indicated above, even the detection of MLH1 promoter hypermethylation or of a BRAF mutation (in CRC) does not exclude definitively the possibility of LS, and it remains important to consider the results in light of the clinical data, i.e., the patient's personal and family history of cancers. Moreover, although these markers are used to detect LS, they should not be considered as diagnostic markers of LS. Only the identification of a germline pathogenic variant (or epimutation) in one of the four MMR genes or EPCAM will confirm the diagnosis of LS and lead to appropriate management of the patient and his relatives. even the detection of MLH1 promoter hypermethylation or of a BRAF mutation (in CRC) does not exclude definitively the possibility of LS, and it remains important to consider the results in light of the clinical data, i.e., the patient's personal and family history of cancers. Moreover, although these markers are used to detect LS, they should not be considered as diagnostic markers of LS. Only the identification of a germline pathogenic variant (or epimutation) in one of the four MMR genes or EPCAM will confirm the diagnosis of LS and lead to appropriate management of the patient and his relatives. Figure 2. Molecular strategy to distinguish LS-related from sporadic MSI/dMMR tumors. a Should be interpreted with the patient's clinical data and family history of cancer; thus, a strong suspicion of LS may lead to a further germline MMR analysis. b Although rare, the hypotheses of germline MLH1 epimutation and of methylation as the 2nd hit should be considered and may lead to further germline MLH1 analysis. c Although rare, the hypothesis of a mosaic MMR mutation should be considered. Abbreviation: LS, Lynch syndrome, CRC: Colorectal Cancer The absence of germline mutation (or epimutation) detected in MMR genes (and in MUTYH or POLE) should lead to a further tumor analysis to look for biallelic somatic MMR gene mutations. This is of importance for appropriate genetic counseling and clinical management. Somatic MMR mutations are likely sporadic events. Therefore, the identification of two MMR mutations exclusively in the tumor (a mosaicism should have been excluded) will allow the reassurance of patients and their relatives and their exclusion from LS-specific surveillance programs. This led us to propose a tumoral MMR gene analysis in patients who developed their MSI/dMMR tumor before the age of 60 when no Figure 2. Molecular strategy to distinguish LS-related from sporadic MSI/dMMR tumors. a Should be interpreted with the patient's clinical data and family history of cancer; thus, a strong suspicion of LS may lead to a further germline MMR analysis. b Although rare, the hypotheses of germline MLH1 epimutation and of methylation as the 2nd hit should be considered and may lead to further germline MLH1 analysis. c Although rare, the hypothesis of a mosaic MMR mutation should be considered. Abbreviation: LS, Lynch syndrome, CRC: Colorectal Cancer The absence of germline mutation (or epimutation) detected in MMR genes (and in MUTYH or POLE) should lead to a further tumor analysis to look for biallelic somatic MMR gene mutations. This is of importance for appropriate genetic counseling and clinical management. Somatic MMR mutations are likely sporadic events. Therefore, the identification of two MMR mutations exclusively in the tumor (a mosaicism should have been excluded) will allow the reassurance of patients and their relatives and their exclusion from LS-specific surveillance programs. This led us to propose a tumoral MMR gene analysis in patients who developed their MSI/dMMR tumor before the age of 60 when no germline variant has been identified; as for patients above 60 y.o., a sporadic tumor is the most likely explanation and does not require further tumor testing [220].

Conclusions
The methods for LS diagnosis have considerably evolved in recent years and so have the tumoral tests for LS screening and for the discrimination of LS-related and sporadic MSI/dMMR tumors, leading to the improved identification of LS patients. In this review, we provided an update on the available clinical, histopathological, and molecular criteria that are evocative of LS. We also focused on the different mechanisms and pitfalls that can lead to potential misinterpretations and to inappropriate clinical management. A strategy taking into account these issues was proposed. Even if there are still a few cases with unknown etiology of the MSI/dMMR, the proposed screening strategy makes it