Tumor-independent Detection of Inherited Mismatch Repair Deficiency for the Diagnosis of Lynch Syndrome with High Specificity and Sensitivity

Lynch syndrome (LS) is the most common hereditary cancer syndrome. Early diagnosis improves prognosis and reduces health care costs, through existing cancer surveillance methods. The problem is finding and diagnosing the cancer predisposing genetic condition. The current workup involves a complex array of tests that combines family cancer history and clinical phenotypes with tumor characteristics and sequencing data, followed by a challenging task to interpret the found variant(s). On the basis of the knowledge that an inherited mismatch repair (MMR) deficiency is a hallmark of LS, we have developed and validated a functional MMR test, DiagMMR, that detects inherited MMR deficiency directly from healthy tissue without need of tumor and variant information. The validation included 119 skin biopsies collected from clinically pathogenic MMR variant carriers (MSH2, MSH6) and controls, and was followed by a small clinical pilot study. The repair reaction was performed on proteins extracted from primary fibroblasts and the interpretation was based on the MMR capability of the sample in relation to cutoff, which distinguishes MMR proficient (non-LS) from MMR deficient (LS) function. The results were compared with the reference standard (germline NGS). The test was shown to have exceptional specificity (100%) with high sensitivity (89%) and accuracy (97%). The ability to efficiently distinguish LS carriers from controls was further shown with a high area under the receiving operating characteristic (AUROC) value (0.97). This test offers an excellent tool for detecting inherited MMR deficiency linked to MSH2 or MSH6 and can be used alone or with conventional tests to recognize genetically predisposed individuals. Significance: Clinical validation of DiagMMR shows high accuracy in distinguishing individuals with hereditary MSH2 or MSH6 MMR deficiency (i.e., LS). The method presented overcomes challenges faced by the complexity of current methods and can be used alone or with conventional tests to improve the ability to recognize genetically predisposed individuals.


Introduction
Lynch syndrome (LS, previously hereditary nonpolyposis colorectal cancer; MIM# 120435), is the most common hereditary cancer syndrome in the world, affecting 1 in 100 to 300 individuals (1,2). Characterized by an inherited mismatch repair (MMR) defect, LS predisposes to a significantly increased, large population datasets have not helped the fact that the majority of LS cases still remain undetected (5)(6)(7). The current diagnostic complexity lies within mandatory tumor testing coupled with the need to identify and verify the pathogenicity of a germline sequence variant (8). DiagMMR, the novel test described in the article, detects inherited MMR deficiency from healthy tissue and helps to make a diagnosis with or without tumor and variant information.
Dominantly inherited LS is commonly diagnosed through the occurrence(s) of colorectal and/or endometrial cancer although, the risk for other cancers is also higher than what is seen in the general population (9). Pathogenic variants in the MMR genes, MSH (MIM# 609309), MLH (MIM# 120436), MSH (MIM# 600678), and PMS (MIM# 600259) cause gene-associated cancer risks in their carriers. The average cumulative risk for developing any cancer by age 75 is highest for MSH carriers (81%) and MLH carriers (77%) (4). The MSH-and MLH-associated lifetime risk for developing colorectal cancer, and MSH-and MSH-associated risk for developing endometrial cancer has been estimated to be over 10-fold, compared with the general population (4,6). Of other LS spectrum cancers, bladder, ureter, kidney, and prostate cancers are more commonly associated with MSH defects, while cancers of the stomach, small bowel, bile duct, pancreas, and ovaries are more frequently associated with MSH and MLH defects (4). However, the cancer penetrance driven diagnosis of LS has inevitably contributed to its underdiagnosis. This is emphasized by the estimation that the majority of pathogenic MMR variant carriers have not yet had cancer, or there has not been an indication of LS based on tumor testing (6,10). There is also a notable number of suspected LS families, in which a variant has not been found (6,11).
LS diagnostics utilize protein expression analyses by IHC to indicate MMR deficiency in tumor tissue through the absence of specific MMR protein(s) expression. However, it is important to note that MMR deficiency does not necessarily degrade the protein (12,13). The other tumor phenotype indicating MMR deficiency is microsatellite instability (MSI), which is also increasingly important for treatment decisions. MSI is, however, not characteristic only of LS but of approximately 15%-20% of sporadic colorectal cancers and up to 45% of sporadic endometrial cancers as well (14)(15)(16)(17). MLH promotor hypermethylation, which is typically causing MLH1 expression loss and high MSI (MSI-H) in sporadic tumors, is not commonly observed in LS tumors (10).
Thus, a hypermethylation analysis can be used to distinguish between hereditary and nonhereditary MLH loss in a tumor. Germline sequencing is used to confirm the presence or absence of a hereditary MMR gene variant. However, it does not always provide a definite LS diagnosis because a single sequence variant may not be found or the pathogenicity and clinical relevance of the found variant(s) are uncertain (18,19 (20). So far, 1,714 nontruncating variants have undergone the InSiGHT expert panel review of multiple points of variant effect information, including available clinical and functional data, but the task is challenging and extends to variants uploaded to the U.S. NCBI's ClinVar database (19,21).
Overall, studies have shown that with the current methods, the adherence to LS screening recommendations is limited (5,22). As such, large-scale genetic screenings have not been predicted to be effective (23), while molecular testing of colorectal cancer and endometrial cancer tumor tissue has been shown to be cost-effective, particularly for individuals <70 years old (23,24). Still, LS diagnostics lack a predictive stand-alone method, which could change the course toward proactive recognition and surveillance of LS carriers prior to cancer appearance; a method for accurate LS identification, while also relieving non-LS families and non-carriers in LS families from lifelong clinical follow-up programs.
Here, we introduce a novel LS carrier test called DiagMMR and demonstrate its exceptional high specificity and sensitivity to detect individuals who have an inherited MMR deficiency linked to the MSH2 or MSH6 proteins. The functional MMR defect is detected from skin fibroblasts, enabling its use for preventive and early LS identification. DiagMMR offers a novel single test tool to improve the current LS diagnostic workup, cancer prediction, and prevention strategies.

Study Design and Sample Collection
The purpose of the study was to clinically validate the DiagMMR test and determine its specificity and sensitivity for MSH and MSH caused LS detection. Study included people who had been confidently identified as LS cases. Thereafter, we demonstrated the utility of DiagMMR in a small clinical pilot study including individuals whose LS diagnosis had remained inconclusive. Because of the lack of other diagnostic functional methods, we used sequencing as a reference test for the method validation. While sequencing does not reveal function of proteins, it is considered the gold standard test for LS diagnosis. DiagMMR test results were interpreted either as normal (proficient) MMR or reduced (deficient) MMR based on the MMR capability of the sample in relation to cutoff, which distinguishes MMR proficient from MMR-deficient function.
Sample collection was organized at the Helsinki University Central Hospital (HUCH) and Jyväskylä Central Hospital, Finland. For the clinical validation, skin samples were collected from LS pathogenic variant carriers during their colonoscopy control visits, from controls, and from LS family members who had been called for genetic counseling as risk members but did not know their own carrier status yet. From them, the skin samples were collected after genetic counseling visit and tested in parallel to blood sample collection and sequencing (performed at HUCH). A separate postvalidation pilot cohort included seven LS suspected individuals based on tumor studies but whose sequencing result was either inconsistent or the variant was not found.

Clinical Validation Samples
Altogether, 119 skin samples were collected for the validation study. This cohort included samples from carriers (n = 28) whose variant in MSH or MSH was confirmed by sequencing (LS carrier), and controls not suspected to carry a germline pathogenic variant (n = 65) or in whom a pathogenic variant  (Fig. 1). The InSiGHT LOVD database was further used (see Discussion) for determining the pathogenicity of variants included in the study (20).

DiagMMR
All samples were routinely treated according to the DiagMMR protocol outlined in Fig. 2. The DiagMMR method has been previously optimized with 259 skin samples for the detection of MMR deficiency caused by an inherited pathogenic variant in the MSH or MSH gene. The method includes the entire process from sampling to fibroblast culture, functional MMR efficiency testing, and results interpretation.

Sampling and Fibroblast Culture
All dermal biopsies were collected from inner forearm using a ≥2 mm biopsy at 2-3 mm depth. Samples kept at room temperature were dissected within 4 days for primary cell culture. To promote tissue adherence and cell growth, tissue dissections were dried to a cell culture flask in humid conditions. After 14 days of initial cell growth in standard fibroblast cell culture conditions, cells were trypsinized off the vessel and expanded in subsequent culture to approximately 1 × 10 8 cells.

Protein Extraction
Nuclear proteins were extracted as described previously (25). In brief, cells were collected and treated with cold isotonic and hypotonic buffers prior to the disruption of cell membranes. Nuclei were then centrifuged prior to the desalting of the protein sample extracted. Nuclear proteins were quantified using Qubit 3.0 fluorometer (Thermo Fisher Scientific, RRID:SCR_020311).

MMR Efficiency
To distinguish between MMR proficiency and deficiency, the functional MMR efficiency was assessed as described previously (25,26). Briefly, 80 μg of nuclear proteins were incubated at 37°C with 50 ng of circular GT heteroduplex substrate for 1 hour. The reaction was terminated with proteinase K and clarified of proteins with phenol:chloroform:isoamyl alcohol prior to substrate DNA precipitation and subsequent restriction analysis with Eco31I and BglII. The BglII enzyme is specific for repaired substrate only and allows the quantitation of MMR repair in each reaction. Repair efficiency was visualized with gel electrophoresis using a 1% agarose gel ran at 100 V for 1 hour and quantified by GeneTools (SynGene, RRID:SCR_005663). Repair efficiency is measured as a percentage of repaired DNA of total amount of DNA in each reaction.

Data Availability
Raw data for this study were generated at the University of Helsinki (Helsinki, Finland) and LS CancerDiag Ltd (functional data) and Helsinki University Hospital (clinical data) and are not publicly available due to patient privacy requirements. Derived data (Tables 1 and 2) supporting the findings of this study are available within the article.

Clinical Validation and Diagnostic Accuracy of DiagMMR
In total, 119 dermal biopsies were tested for DiagMMR clinical validation. Of the 119 samples, 93 demonstrated inherited functional MMR proficiency (no-LS) and 24 MMR deficiency (LS) by DiagMMR, while two samples were repeatedly within 5% of the cutoff, inside the "gray area" (Fig. 3A) The ROC curve for the 117 samples (excluding two gray area samples) confirms the previously defined cut-off level to be optimal (Fig. 3B). The test performance data indicate that increasing sensitivity from current level would rapidly increase FP fraction and therefore the selected cut-off level can be considered optimal. The AUC value reaches an excellent level of 0.97 indicating high overall accuracy.

Postvalidation Clinical Pilot
Immediately after clinical validation, seven individuals whose LS status had remained inconclusive, were studied with DiagMMR (Helsinki University Central Hospital). After testing the blinded samples, the DiagMMR test results were compared with the available clinical data, which revealed that in cases where tumor tissue had shown MSH2 and/or MSH6 protein expression loss with IHC, an MMR gene variant was found in only one case, and here it is pathogenicity could not be confirmed ( Table 2).
The DiagMMR results of the clinical pilot samples are illustrated in Fig. 4 (44,45). IHC inability to assess the functionality of the expressed protein contributes to its false-negative rate (nonfunctional protein expressed), while changes restricted to the tumor tissue reflect the FP rate. This is particularly relevant for pathogenic missense variants as well as for double somatic MMR gene defects which are particularly common in endometrial cancer and are at least part reason for the challenges in detecting LS in patients with tumor IHC and MSI findings (12,46). MSI analyses have proven extremely informative for tumor treatment decisions but are unable to distinguish LS tumors from sporadic colorectal cancers with MSI due to MLH promotor hypermethylation (47). In all cases though, current LS diagnostics rely heavily on tumor-derived information and sequencing, with deficiencies in the detection of large deletions and insertions in the coding region and base errors occurring in deep intronic sequences and noncoding regulatory regions (48). Thus, the need for functional assays in translating sequence alterations into clinically actionable diagnoses persist (49)(50)(51). In LS, the mechanism causing a high risk for cancer is the dominantly inherited MMR deficiency, no matter where in the genome the pathogenic variant is, what type it is, and whether the protein is degraded or not in a tumor. To our knowledge, DiagMMR is currently the only validated and CE marked functional test for LS (IVDD 98/79/EC). The method is based on our previous work with the functional in vitro MMR assay (26) used for the functional assessment of recombinant MMR proteins. The in vitro MMR assay was further developed taking advantage of our research results showing that the functional assessment of MMR gene variants and nuclear proteins of cells with reduced MMR gene mRNA expression is effective in recognizing reduced MMR function (25,31,(52)(53). The novelty and inventiveness of the DiagMMR test lies in the fact that it can measure the DNA mismatch repair capability directly from the proteins extracted from the individual's primary fibroblast cells. The test specifically detects the weakening of MMR function in normal human cells. However, the test does not show that this inherited impairment of MMR efficiency in noncancerous cells, although the susceptibility is known to exist, leads to malignancy or other biological consequences without requiring other events in the cell. Although the use of blood cells is generally considered convenient in diagnostics, their suitability for DiagMMR testing has not been demonstrated.
Using clinical diagnoses confirmed by sequencing as a reference standard for identifying MSH and MSH variant carriers, we show that DiagMMR has exceptionally high specificity (100%, no FP), sensitivity (89%), and accuracy (97%). The high AUC value of 0.97 demonstrates the methods ability to detect LS. The lack of FP results is expected as the assay cutoff was based on the lowest repairs measured in a set of control samples to aim for high specificity. Indeed, all controls were interpreted as MMR proficient (normal), except for one which could not be classified as MMR proficient due to its proximity to the cutoff. This and an MSH carrier sample also positioning within 5% of the cutoff were interpreted as samples neither MMR proficient nor deficient. Overall, individuals included in the clinical validation carried 20 different MSH/MSH variants, which all were interpreted as pathogenic in the hospital. Twelve of those were classified in the InSiGHT LOVD database as pathogenic [10 pathogenic (class 5) and 2 likely pathogenic (class 4)], two were listed as pathogenic but not confirmed, and six clinically categorized pathogenic variants were not listed in the database (20  Resources. M. Nyström: Conceptualization, resources, supervision, funding acquisition, writing-review and editing.