Testing strategies to reduce morbidity and mortality from Lynch syndrome

Abstract

Objective: Lynch syndrome (LS) has an autosomal dominant inheritance pattern and is associated with increased risk for colorectal cancer (CRC) and other cancers. Various strategies are used to identify patients at risk and offer surveillance and preventive programs, the cost effectiveness of which is much dependent on the prevalence of LS in a population. Universal testing (UT) is proposed as an effective measure, targeting all newly diagnosed CRC patients under a certain age.

Materials and methods: LS cases were identified in a cohort of 572 consecutive CRC patients. Immunohistochemistry was performed in 539 cases, using antibodies against mismatch repair proteins MLH1, PMS2, MSH2, and MSH6. Microsatellite instability and gene mutation screening were performed in 57 cases.

Results: In total 11 pathogenic variants were detected, identifying LS in 1.9% of new CRC cases. Comparing the results with current clinical methods, 2 pathogenic variants were found with Amsterdam criteria and 9 when using either Bethesda guidelines or our institution’s prior clinical criteria. Pathogenic variants in MSH6 were the most common in our series. We also found different outcomes using different age cut offs.

Conclusion: Our study demonstrates that UT of tumors before age on onset at 75 years would most likely be cost-efficient and essentially equivalent to applying the Bethesda guidelines or our institution’s prior clinical criteria on all new CRC.

Keywords:

• Lynch syndrome
• MLH1
• MSH2
• MSH6
• PMS2
• immunohistochemistry
• mutation screening
• universal testing

Introduction

The overall purpose with screening for patients with Lynch syndrome (LS), is to reduce morbidity and mortality in families with the disease. LS is an autosomal dominant inherited syndrome, where men and women have a high risk of bowel- and other cancers [1]. The prognosis of all colorectal cancer and gynecological cancers detected under surveillance is extremely good [1,2]. In Sweden today, more than 500 families with LS are known and under surveillance. Families are typically identified because of a family history of, or early onset of, bowel- or gynecological cancer. Members from such families are referred to genetic counseling and genetic testing at a University Hospital in Sweden. Once LS is diagnosed, patients are offered surveillance programs. Gene carriers are recommended regular colonoscopy screening every to every second year. In 1998, a clinical trial was designed to prescribe daily aspirin as primary prevention in gene carriers with LS. The trial took place between 1999 and 2005 and showed that aspirin reduced the risk of colorectal and associated cancer with almost 50% in these patients [3]. Thus, the well-working preventive programs in the high-cancer risk patients with LS confers that it is very important to diagnose all LS patients.

Only a fraction of LS colorectal cancer is identified, and many protocols have been designed to predict LS in patients and healthy individuals [4]. The original Amsterdam criteria as well as the Bethesda guidelines predicting LS based on family history, age of onset and tumor morphology were revised to better predict LS [5,6]. Many studies have focused on different models to preselect colorectal cancer (CRC) patients for mutation testing of the LS genes and all work well in predicting mutation carriers [4]. Most Swedish protocols use family history and age of onset for referral of patients for genetic counseling. Thus, the first selection of patients for mutation screening is made by the referring doctors, often surgeons, and therefore only based on age of onset, or family history of cancer.

Since microsatellite instability (MSI) and typical loss of DNA mismatch repair (MMR) proteins in LS tumors are the known hallmarks of LS, it was early suggested that universal testing of tumors would be better than current diagnostic protocols [7,8]. In 2008 the Evaluation of Genomic applications in Practice and Prevention Working Group (EWG) published a supplementary evidence review on DNA testing strategies aimed at reducing morbidity and mortality from LS [9]. They concluded that difficulties to obtain a good family history at time of surgery for CRC means that both Amsterdam criteria and Bethesda guidelines suffer from severe limitations in finding all LS cases. They suggest removing family history from the procedure to predict for genetic testing for LS and instead use a preliminary test in all individuals under a certain age, newly diagnosed with CRC. The authors emphasized the need of a comprehensive cost-effectiveness analysis to help inform policy makers about which strategy might be the one to favor [9]. It was suggested performing immunohistochemistry (IHC) or MSI-analysis of all new CRC to diagnose all LS families, so called universal testing (UT) [10].

The prevalence of LS in a population is of major importance to evaluate the cost-benefit for the test-all strategy, as is the number of first-degree relatives to be tested. Screening will be more cost-effective the more individuals are involved in the programs. The ability to identify all individuals with LS is limited by the expenditure to do so. Total costs for the genetic investigation and the costs for prevention and treatment are equally of importance. The prevalence of LS varies between 2–5% in different regions. The expenses for molecular testing varies depending on the different health care systems ability to adapt to the rapid development of methodologies used, while costs for prevention programs and treatment might differ less [8].

The present study set up to determine the proportion of LS patients in consecutive series of Swedish CRC patients, and to find out how many cases with LS could at best be diagnosed by UT using all new CRC cases, regardless of age, compared with strategies using the following prescreening criteria; 1) Amsterdam criteria [5], 2) Bethesda guidelines [6], or 3) our institution’s prior clinical criteria [11], and 4) using different age limits (40, 50, 60, 70, 80, and 90 years of diagnosis).

Materials and methods

Patients

Patients were recruited within a national study conducted by the Swedish Low-Risk Colorectal Cancer Study Group [12]. Samples were obtained during 2004 to 2009 from 14 different surgical clinics in the middle of Sweden. Informed consent was obtained from each patient, blood was obtained, and DNA was extracted using routine methods. All surgical specimens underwent evaluation directly after operation by a local pathologist and a detailed protocol was later used to describe tumor morphology [13]. For this study 572 available tumor samples from pathology clinics in Stockholm were used. The samples were coded for fulfilling the Amsterdam criteria [5], Bethesda guidelines [6], or our institution’s prior clinical criteria [11].

Immunohistochemistry

Paraffin-embedded blocks containing formalin-fixed adenocarcinoma from either colon or rectum were used for immunohistochemistry (IHC) of MLH1, MSH2, MSH6, and PMS2. Staining was performed according to the manufacturer’s instructions with the Benchmark ULTRA staining module. The following antibodies were used: mouse monoclonal antibody (clone M1, Ventana), mouse monoclonal antibody (clone G219-1129, Ventana), rabbit anti-human monoclonal antibody (clone SP93, Cell Marque) and rabbit monoclonal antibody (clone EPR3947, Ventana). Tissue sections were interpreted by one pathologist and cases with unusual staining patterns were reviewed by two pathologists. The sample was considered MMR-deficient when displaying total or partial nuclear loss of expression in invasive tumor cells and adjacent non-neoplastic tissue with retained expression, serving as an internal positive control for the sample.

MSI testing

Dissection of tumor cells from paraffin-embedded tissue for MSI were performed  as part of the clinical screening (Department of Clinical Genetics, Karolinska University Hospital, Stockholm). MSI testing was performed using a commercial kit (Microsatellite Instability Multiplex System Kit, Promega Corp, Madison, WI) according to the manufacturer’s instructions. Briefly, genomic DNA prepared from tumor tissue was amplified by PCR in a multiplex PCR, amplifying selected microsatellite markers. Microsatellite analysis was performed using an ABI310 or ABI3730 (Applied Biosystems, Foster City, CA) with the GeneScan 3.1 or GeneMapper software (Applied Biosystems). DNA isolated from normal tissue was used as a normal control when available. In total five mononucleotide markers were analyzed. If three or more markers show a pattern consistent with MSI the tumor was considered MSI-H (MSI high) and if none of the markers displayed MSI the tumor was considered MSS (MS stable) all results in between was labeled as MSI-L (MSI low). MSI-H was considered MSI positive and MSS and MSI-L were considered MSI negative. The classification of MSI-H, MSS, and MSI-L were according to the Bethesda/National Institutes of Health (NIH) guidelines [6].

Mutation screening

Mutation screening was mainly performed using Sanger sequencing. In brief, genomic DNA prepared from a blood sample from the patient was amplified using exon specific PCR. All exons in the analyzed gene/-s were included in the analysis (primers and PCR-conditions are available upon request). The PCR products were then used for direct sequencing using the BigDye Terminator v1.1 Sequencing Kit (Applied Biosystems) according to the manufacturer’s recommendation. Sequences were analyzed using an ABI Prism 3730 or ABI3130XL Sequencer (Applied Biosystems). The chromatograms were evaluated using SeqScape v2.5 or v.3 (Applied Biosystems). Sequencing was also performed with massive parallel sequencing (MPS) using the IonTorrent system (Applied Biosystems). A custom-made design of requested genes was made using AmpliSeq designer (Applied Biosystems). Libraries and sequencing was performed according to the manufacturer’s recommendations (Applied Biosystems). Output data was analyzed using the IonReporter software. Reported sequence variants from the MPS were verified using Sanger sequencing. As reference sequences NM_000249 (MLH1), NM_000251 (MSH2), NM_000179 (MSH6), and NM_ 000535 (PMS2) were used.

In case no potential pathogenic sequence variant was detected genomic DNA was also analyzed for the possibility of larger deletions/duplications in analyzed genes. This analysis was performed using MLPA [14]. MLPA (assays P003, P008 and P072) was performed according to the manufacturer’s recommendations (MRC-Holland b.v., Netherlands). Data was analyzed using GeneMarker software (SoftGenetics LCC, State College, PA, USA).

Generally, all samples tested before year 2015 was analyzed using Sanger sequencing and MLPA and all samples analyzed 2015 and onwards was tested using MPS and MLPA.

ACMG guidelines [15] were used to classify detected variants, if the variant wasn’t already classified by the InSiGHT variant interpretation committee [16].

Results

Altogether 572 CRC cases were used for IHC or MSI. In total 513 samples were tested using IHC only, 46 samples were tested with MSI only and 13 samples were tested with both MSI and IHC. For MSI-positive samples IHC was also performed to decide what gene to sequence. Samples with aberrant IHC were subject to sequencing and MLPA of selected genes; samples with loss of MLH1 and PMS2 were screened for MLH1, samples with loss of MSH2 and MSH6 were screened for MSH2, samples with loss of MSH6 only were screened MSH6 and samples with loss of PMS2 only were screened for variants in PMS2. If unclear IHC, one or more MMR genes were screened. 27 cases were screened using Sanger sequencing and MLPA and 49 cases were analyzed with MPS and MLPA. In total 13% (77/572) samples showed aberrant IHC and were subject to mutation screening of one to four of the MMR genes. The remaining 495 tumors (MSI-neg or normal IHC) were not used for further testing. In total 11 pathogenic variants were detected and this was considered the best result to be obtained from UT using IHC to select candidates for sequencing within these 572 cases (Table 1). In three cases, variants of unknown clinical significance was detected (data not shown). We compared the results to what would have been the outcome if Amsterdam criteria, Bethesda guidelines or our institution’s prior clinical criteria had been used instead. Amsterdam criteria were fulfilled in five cases, and two samples were sequenced, and both had pathogenic variants (Table 1). Bethesda revised guidelines were fulfilled in 182 cases, and 32 were sequenced. Among these 32 samples nine pathogenic variants were found (Table 1). Our institution’s prior clinical criteria were fulfilled for 131 of the tumors, and IHC suggested Lynch syndrome in 20 of those, which all were sequenced. In total 9 pathogenic variants were found which gives the same detection rate as the Bethesda guidelines. Thus, UT in this series identified Lynch syndrome in 1.9% (11/572) of a consecutive series of new Swedish CRC cases.

Table 1. Summary of findings in this study.

						Selection criteria
Patient		FH^a	IH^b	Pathogenic variant	Age of onset	AC^c	BG^d	PCC^e
1		0	PMS2-	PMS2; c.2113G > A (p.Glu705Lys)	35	0	1	1
2		0	PMS2-	PMS2; del exons 2-9	46	0	1	1
3		0	MSH2-MSH6-	MSH2; c.1906G > C (p.Ala636Pro)	31	0	1	1
4		1	MLH1-PMS2-	MLH1; del exons 10-11	64	2	2	2
5		1	MSH6-	MSH6 c.3013C > T (p.Arg1005*)	90	0	0	0
6		2	MLH1-PMS2-	MLH1 c.546-2A > G	54	0	4	2
7		2	MSH2-MSH6-	MSH6; c.3195_3199del (p.Asn1065Lysfs*5)	57	0	2	3
8		1	MSH2-MSH6-	MSH2; del exons 1-8	61	1	2	2
9		0	MSH6-PMS2-	PMS2; c.2113G > A (p.Glu705Lys)	71	0	0	0
10		2	MSH6-?	MSH6; c.3647-2A > C	71	0	4	2
11		0	MSH6-	MSH6; c.2779dup (p.Ila927Asnfs*8)	64	0	2	3

^a FH (Family History), 0; no family history, 1; one close relative with CRC, 2; at least two more relatives with CRC.

^b IH, immunohistochemistry loss of protein.

^c AC, revised Amsterdam criteria, 0, the criteria are not met, 1, at least three close relatives with CRC, one of whom is younger than 50 years of age, 2, at least three close relatives with CRC or HNPCC-associated cancer, one of whom is younger than 50 years of age [5,10].

^d BG, revised Bethesda guidelines, 0, the criteria are not met, 1, patient with CRC younger than 50 years of age, 2, presence of synchronous, metachronous CRC or other HNPCC-associated tumor, 4, CRC or HNPCC-associated tumor in at least one close relative younger than 50 years of age [6].

^ePCC, our institution’s prior clinical criteria, the clinical criteria used in previous study, 1, CRC in a patient younger than 50 years of age, 2, CRC in at least one close relative younger than 60 years of age, 3, CRC in at least two close relatives regardless of age [11].

Loss of MLH1 and PMS2 in 58 cases had the lowest detection rate of mutations among samples screened for mutations, only two patients with pathogenic variants in MLH1 were found, 54 and 64 years old. IHC loss of MSH2 and MSH6, MSH6 alone and PMS2 alone all were better predicting a pathogenic variant (Table 2).

Table 2. Results from immunohistochemistry and mutation screening.

Protein	MLH1-/PMS2-	MSH2-/MSH6-	PMS2-	MSH6-
IH-loss	58	3	6	7
Mutations	2	3	3	3
IH-loss, age <70	25	3	4	2
Mutations	2	3	2	1

An upper age limit for UT of tumors has been suggested. To find out whether UT using various age-limits would be feasible we compared the outcome applying different age cut offs (Table 3). It is quite clear that ages of onset show a wide range evenly spread from before 40 up to 75. The benefit using an age limit would obviously mean a lower number to screen with IHC, but at the same time many mutations will be lost. Screening all diagnosed before 75 years at age of onset to find 10 of the 11 pathogenic variants, would have needed IHC to be done on 377 tumors (Table 3). If Bethesda guidelines or our institution’s prior clinical criteria had been applied on all and 9 pathogenic variants found, 144 and 130 tumors respectively would have to be used for IHC. Both Bethesda and our institution’s prior clinical criteria would have missed the PMS2 pathogenic variant in a 71 years old patient with no family history. The case with CRC at 90 years of age with no known family history of disease could only be detected if all tumors were screened using IHC.

Table 3. Distribution of included number of patients and the number of detected pathogenic variants of totally 11 in each group respectively.

Group of patients^a	Patients	Patogenic variants
< 40 years	8	2
< 50 years	30	3
< 60 years	115	5
< 70 years	280	8
< 80 years	450	10
< 90 years	563	0
< 100 years	572	11
Amsterdam criteria	5	2
Bethesda guidelines	14	9
Prior clinical criteria	130	9

^aResults grouped according to age and clinical criteria.

Discussion

The prevalence of LS in a population is of importance for the cost-efficiency of the strategy of universal testing, since the more gene carriers could be offered preventive measures the higher cost-efficiency. In our previous study aiming to define the proportion of LS in a consecutive series of 3214 newly diagnosed CRC cases in Sweden, less strict criteria of family history and age of onset were used to suggest tumors for IHC/MSI. This approach resulted in a frequency of Lynch syndrome of 2% in our population [12]. The present study, used 572 cases from the local hospitals in Stockholm, Sweden and applied universal testing with IHC or MSI on all those cases and with the same result. However, both studies show an underestimation of the proportion of LS among newly diagnosed colorectal cancer in Sweden. The first one [12] missed some of the old patients and cases without family history, and the present study would have missed LS patients with MSI-negative and normal IHC in their tumors. These are most likely patients with MSH6 mutations, but also MLH1 mutations which are not rare in our population [11,17]. Only sequencing of the MMR genes in all new cases could give us a more better proportion of mutations, but not even this approach would find all mutations since rare EPCAM mutations or even mutations in the other MMR genes could be missed by a sequencing approach. A previous study using prescreening of tumors before sequencing in a population in Ohio, USA, found LS in 2% [18]. A larger study with a similar design to study two populations (Ohio and Finland) found 3% with LS [19]. Thus, it is likely that Sweden has a LS frequency among new CRC cases of 2–3%.

The spectrum of pathogenic variants was different from our previous study of clinical based LS screening [11,17]. In our previous study clinical relevant variants were mostly found in MLH1 and MSH2, reflecting that those two genes were the first known to cause LS. Also, these patients were selected for referral based on their higher penetrance, resulting in a stronger family history. In this study the results are more likely to represent the true proportion of mutations causing LS. In fact, PMS2 mutations could be the most common MMR gene mutations in our population, considering the low penetrance associated with these mutations [1]. The finding of many PMS2 mutations in early onset CRC conflicts with the idea of low penetrance, however, those early onset cases often lack a family history of cancer even if there are many gene carriers among relatives. The fact that many PMS2 mutations are found in early onset cases with a minimum of family history suggests that the risk for CRC contributed by the PMS2 mutation alone could be modest, and only when modified by other risk factors to result in early onset CRC.

Universal testing has been using different age limits to optimize cost-efficiency by prescreening, those under 50, 60 or 70 for MSI. Typically, more LS patients were found by increasing ages [20]. In our study, it was clear that age of onset up to at least 75 years was common. It was evident that predictive prescreening using Amsterdam criteria was not useful, while both Bethesda guidelines and our institution’s prior clinical criteria were almost as good as universal testing. Most important for interpretation of the results are the conclusions from other studies that, although the general population is willing to participate in genetic screening for CRC [21], incomplete clinical follow-up, reducing referrals and genetic testing, will limit the cost-efficiency from universal testing or any other prescreening method used [22].

The cost-efficiency of universal testing of all new CRC still seems unclear. Universal testing was considered important to implement by one study [20], estimated to be less clear by another [23] and not cost efficient by others [24]. One study compared nine diagnostic strategies and found all predictive model strategies cost-effective compared to no screening [25]. Even when age limits were raised from 50 to 60 and 70 years – predictive models were still cost-effective [25]. The cost for the mutation detection is of importance for the cost efficiency, why we did not use the best pre-selection method, which would have been to use both MSI and IHC on all tumors, followed by methylation test and BRAF testing if IHC showed loss of MLH1 protein, followed by mutation testing. Not performing MSI on all cases is not a severe limitation since MSI and IHC is concordant to a high degree [26]. However, a small proportion of LS tumors, in particular those with at MSH6 or PMS2 mutation, can be false negative regarding both MSI testing and IHC [27]. Thus, there is a possibility that we slightly underestimated the number pf pathogenic variants in this cohort, since we did not sequence them all. We sequenced all 59 samples with IHC loss of MLH1-PMS2 for mutations in MLH1, instead of doing methylation test and BRAF testing first as recommended to avoid unnecessary mutation testing. Still, the expenses for combined methylation test and BRAF are similar to the expenses for sequencing in our laboratory, which explains why this approach was reasonable in this study. Other factors of even more importance for cost-efficiency is how the clinical routines regarding diagnostics and surveillance of LS patients and families are designed.

We know today that surveillance reduces mortality from cancer in LS families [1,2], and that the surveillance programs are well understood and tolerated among the population [28]. Thus, it is important to diagnose all families with LS as a mean of cancer prevention. Our study demonstrates that universal testing of tumors before age of onset at 75 years or applying the Bethesda guidelines or our institution’s prior clinical criteria on all new CRC would most likely be cost-efficient. In any case, the outcome would be highly dependent on to what degree the test was applied, and the patient referred for counseling.

Today, massive parallel sequencing has made it possible to sequence all new tumors for somatic mutations, which can be used in clinical treatment decisions as well as to find germline mutations. When this strategy is implemented, germline mutations possibly associated with LS could be identified at the time of treatment. Counseling will then be of importance to determine if the patient has LS in order to offer cancer preventive programs to family members.

Acknowledgements

We are grateful to the patients for their contribution and Berith Wejderot and Josefin Persson for technical assistance and Patrick Joost as well as Denis Nastic for support.

No potential conflict of interest was reported by the authors.

Disclosure statement

Additional information

Funding

The study was supported by grants from the Swedish research council [521-2012-1897], the Swedish Cancer Society [16-0458] and grants provided by the Stockholm County Council (ALF project).