Clinical and Molecular Features of Pediatric Cancer Patients with Lynch Syndrome

Abstract

Background:

The association of childhood cancer with Lynch syndrome is not established compared with the significant pediatric cancer risk in recessive constitutional mismatch repair deficiency syndrome (CMMRD).

Procedure:

We describe the clinical features, germline analysis, and tumor genomic profiling of patients with Lynch syndrome among patients enrolled in pediatric cancer genomic studies.

Results:

There were 6 of 773 (0.8%) pediatric patients with solid tumors identified with Lynch syndrome - defined as a germline heterozygous pathogenic variant in one of the mismatch repair (MMR) genes (3 with MSH6, 2 with MLH1, and 1 with MSH2). Tumor analysis demonstrated evidence for somatic second hits and/or increased tumor mutation burden in 3 of 4 patients with available tumor with potential implications for therapy and identification of at-risk family members. Only one patient met current guidelines for pediatric cancer genetics evaluation at the time of tumor diagnosis.

Conclusion:

Approximately 1% of children with cancer have Lynch syndrome, which is missed with current referral guidelines, suggesting the importance of adding MMR genes to tumor and hereditary pediatric cancer panels. Tumor analysis may provide the first suggestion of an underlying cancer predisposition syndrome and is useful in distinguishing between Lynch syndrome and CMMRD.

1. Introduction

Lynch syndrome, or hereditary non-polyposis colorectal cancer syndrome, is caused by germline heterozygous pathogenic variants in one of four mismatch repair (MMR) genes (MLH1, MSH2, MSH6, and PMS2) or specific deletions in the EPCAM gene that inactivate MSH2. Lynch syndrome is considered primarily an adult-onset cancer predisposition syndrome, although pediatric cases have been reported^1,2. Given the rarity of childhood-onset tumors, and lack of surveillance recommendations during childhood, testing for Lynch syndrome in unaffected minors is not recommended³.

Following the discovery of Lynch syndrome, a new unique autosomal recessive cancer predisposition syndrome, constitutional mismatch repair deficiency syndrome (CMMRD), was described resulting from inheritance of biallelic germline pathogenic variants in the same MMR genes⁴. While the CMMRD tumor spectrum overlaps with that of Lynch syndrome, tumor onset in early childhood is characteristic of CMMRD. Testing for MMR proteins by immunohistochemistry (IHC) in tumor and normal tissues can be useful in differentiating between the two disorders but is not routinely performed in assessment of pediatric tumors. In Lynch syndrome, a second somatic alteration in the tumor leads to loss of nuclear staining of the affected MMR protein in the tumor tissue but not adjacent normal tissue. In contrast, in CMMRD, MMR protein nuclear expression is absent in both tumor and normal tissue. Tumors from both disorders reveal a high tumor mutation burden (TMB); presence of ultra-hyper mutated TMB (>100 mutation/MB) is a more common feature of CMMRD tumors⁵. Lynch syndrome tumors with somatic POLE variants can mimic CMMRD type tumors with regard to ultra-hyper mutated TMB².

Adoption of large-scale sequencing platforms (genome, exome, and large cancer panels) for both germline and tumor profiling has increased the rate of Lynch syndrome diagnoses in children. However, it remains unclear if Lynch syndrome contributes to the pathogenesis of childhood-onset tumors, or if the increased reporting is a result of broad genomic testing. In a recent study of pan-cancer germline/tumor panel testing in 751 patients with pediatric solid tumors⁶, heterozygous pathogenic/likely pathogenic (P/LP) variants in an MMR gene consistent with Lynch syndrome were seen in seven patients (0.9%), a rate higher than the 0.30% of MMR P/LP variants reported in general population studies ⁷. However, only one of six tumors analyzed in this study demonstrated tumor microsatellite instability (MSI) High⁶. Here we describe the prevalence of Lynch syndrome due to germline heterozygous P/LP variants in MMR genes in pediatric patients with cancer participating in one of two sequential National Institutes of Health Clinical Sequencing Exploratory/Evidence-Generating Research Consortium (CSER) Projects (BASIC3 and KidsCanSeq Trials)^8,9. We expand on previous knowledge of the prevalence of Lynch syndrome by describing the clinical evaluation of each patient following their cancer diagnosis through study sequencing including the results of physical exams by a medical geneticist, comprehensive germline exome sequencing and follow-up targeted sequencing. Additionally, we describe the results of comprehensive tumor profiling and MMR protein status performed through study enrollment in patients with available tumors to evaluate the contribution of defective MMR and Lynch syndrome to the pathogenesis of pediatric tumors.

2. Methods

2.1. Study participants

All participants were recruited from one of two projects that were each approved by the Baylor College of Medicine Institutional Review Board. Informed consent and age-appropriate assent was obtained from parents/legal guardians and patients. Patients that reached the age of majority during the study window were re-consented. The Baylor College of Medicine Advancing Sequencing into Childhood Cancer Care (BASIC3) study completed analysis on 284 patients <18 years of age newly diagnosed with central nervous system (CNS) and other solid tumors from 2012–2016. The Texas KidsCanSeq (KCS) study enrolled patients <18 years of age with diagnoses of CNS and non-CNS solid tumors, lymphomas, and rare histiocytic disorders from 2018–2021 from six sites in Texas. Clinical information was collected at the time of study enrollment by review of the electronic medical record including medical and family history and a family history questionnaire for the BASIC3 trial. Clinical sequencing results were available for 489 KCS patients at the time of this data analysis. In both studies, parents had the option to submit parental samples (blood or saliva) for clinical testing focused exclusively on germline variants identified in the proband. Patients with Lynch syndrome diagnoses, who had not been previously referred to genetics, were subsequently referred for a genetics evaluation including consideration of the potential for a missed CMMRD diagnosis. In addition, when possible, a patient germline sample was sent for comprehensive germline sequencing and copy number testing for the MMR gene of interest to an independent clinical diagnostic laboratory. Analysis included applying the 2022 version of the comprehensive McGill Interactive Pediatric OncoGenetic Guidelines for Identifying Cancer Predisposition Syndromes (MIPOGG) algorithm^10,11 to predict which children with cancer would be recommended to have genetic evaluation. The algorithm was applied using the information (tumor diagnosis, any specialized pathologic staining or molecular diagnostic testing, age and family history) in the medical record available at study entry.

2.2. Germline Molecular analysis

All genomic testing was performed in College of American Pathologists - and Clinical Laboratory Information Act (CLIA) -certified laboratories. All patients in BASIC3 received clinical exome sequencing of blood in the Whole Genome Laboratory at Baylor College of Medicine as previously described⁸. Germline exome testing for KCS patients was performed in the Human Genome Sequencing Center – Clinical Laboratory (HGSC-CL) at Baylor College of Medicine following the BASIC3 protocol including library construction, exome capture by VCRome, version 2.1 supplemented with PKv1 and PKv2 probes for under-covered regions¹² (Roche NimbleGen), and paired-end sequencing on HiSeq 2000/2500 (Illumina Inc).

Texas Children’s Hospital: For KCS patients, DNA from a germline sample (blood or saliva) was also analyzed at Texas Children’s Hospital on the pediatric cancer focused panel, described in detail below. Germline panel reports included germline variants and gene-level copy number alterations in 35 genes associated with cancer predisposition. While MMR genes were analyzed by exome sequencing, they were not included in the focused panel developed in 2018 that was used for analysis of this cohort. These genes were subsequently added and are now included as a part of panel analysis.

2.3. Tumor analysis

In BASIC3, tumor from frozen samples and germline library pairs were sequenced when available on a single lane of a HiSeq 2000/2500, with a mean coverage of 272X and a target base coverage of 20X at 97.3%. Data analysis and somatic variant detection was performed as described previously⁸. In KidsCanSeq, analysis of tumors for patients with high-risk tumors was an integrated cancer genome profile (integrated tumor exome, transcriptome, and copy number profile) performed at the HGSC-CL. KCS patients also received focused targeted panel testing at Texas Children’s Hospital with a custom-designed next-generation sequencing test (Solid Tumor Comprehensive Panel) for mutations and copy number variants in 124 genes using capture hybridization (Roche NimbleGen) and gene fusions in 81 genes using anchored multiplex PCR (ArcherDX)^13,14. DNA libraries were sequenced 2 × 150 bp on a MiSeq (Illumina) for a mean coverage of 345x. Sequence reads were aligned to the GRCh37 (hg19) reference genome (BWA v0.7.12) and variant calling was performed (NextGENe v2.4.1.2 and Platypus v0.8.1 for SNV and short indels; Pindel v0.2.5 and Delly v0.8.1 for longer indels; CNVkit v.0.9.3 for CNV).

2.4. Tumor mutation burden (TMB) and mutation signature analysis

TMB was calculated from the tumor exome (33.6 Mb of coding region) using the total number of non-synonymous somatic variants detected in the coding region adjusted for the size. TMB is reported following classifications based on the rates described by Goodman et al and Grobner et al: low (1–5 mutations/mb), intermediate (6–19 mutations/mb), high (≥ 20 mutations/mb)¹⁵ and ultra-hyper mutated (>100 mutations/mb)¹⁶.

Mutation signature analysis and visualization was performed as previously described¹⁷. Briefly, the 96-substitution classification signature¹⁸ with all substitutions converted to the pyrimidine of the Watson-Crick base pair and representing the immediate trinucleotide context (5’-x-3’) of the mutated base signatures were generated using the Mutational Patterns Bioconductor package in R¹⁹. The optimal non-negative linear combination of COSMIC signatures was then fitted to the substitution classification profile, followed by calculation of the cosine similarity between the tumor mutation signature in patients and COSMIC signatures.

2.5. Mismatch repair IHC

Given the MMR mutation information, tumor specimens were sent to a reference Pathology laboratory for MMR IHC. FFPE blocks with adequate representation of tumor was sectioned at 3 μm thickness and stained with antibodies to MLH1 (Ventana, M1 mouse monoclonal primary antibody, 1 microgram/ml, Roche, Germany), MSH2 (Ventana anti-MSH2 (G19-1129) mouse monoclonal primary antibody, 20 microgram/ml, Roche, Germany), MSH6 (Ventana anti MSH6 (SP93) rabbit monoclonal primary antibody, ~ 1microgram/ml, Roche, Germany) and PMS2 (Ventana anti PMS2 (A16-4) mouse monoclonal primary antibody, ~1 microgram/ml, Roche, Germany). Stromal cells and background lymphocytes served as internal positive control.

3. Results

Of 773 patients enrolled to date, six (0.8%) were identified with germline heterozygous pathogenic variants in MMR genes: MSH6 (n=3), MLH1 (n=2), MSH2 (n=1). The age of diagnosis ranged from 1 to 17 years, and tumor pathology included a variety of CNS and non-CNS tumor types: neuroblastoma in a 1yr old (MSH6, c.1438dupG), Hodgkin lymphoma in a 15yr old (MSH6, c.3439-2A>G), pilocytic astrocytoma with anaplastic transformation in a 17yr old (MSH6, c.2150_2153del), glioblastomas in a 9yr old (MLH1, c.116+1G>A) and 13yr old (MSH2, c.1697delA), and rectal adenocarcinoma in 15yr old (MLH1, c.307-2A>G) (Table 1). The MIPOGG algorithm was applied to all 6 patients (see Methods). The patient with early onset rectal carcinoma (case 6) met MIPOGG criteria for genetic evaluation based on a classic Lynch syndrome tumor. Of the remaining 5 patients, based on available information at study entry, they did not meet criteria. Of note, 3 had gliomas, 2 of which were subsequently documented (see below) to have abnormal IHC for mismatch repair genes and would meet MIPOGG criteria for a genetics referral at that point^10,11.

TABLE 1.

Clinical and tumor characteristics of pediatric cases of Lynch syndrome

Case	Study	Age at DX (yrs)	Tumor Diagnosis	Gene Variant	Tumor 2nd hit	TMB summary	IHC tumor/normal
1	K	1	Neuroblastoma	MSH6 NM_000179.2 c.1438dupG	No	Low (0.24 mut/MB)	Retained/Retained
2	K	15	Hodgkin lymphoma, nodular sclerosis	MSH6 NM_000179.2 c.3439-2A>G	NA	NA	Unknown
3	K	17	Pilocytic astrocytoma with anaplastic transformation	MSH6 NM_000179.2 c.2150_2153del	SNV c.3646+2T>G	Ultra-hyper (>200 mut/MB)	MSH6 loss/Retained
4	K	9	Glioblastoma	MLH1 NM_000249.3 c.116+1G>A	Deletion (LOH)	Intermediate (7.26 mut/MB)	MLH1&PMS2 loss/Retained
5	B	13	Glioblastoma	MSH2 NM_000251 c.1697delA	NA	NA	Unknown
6	B	15	Rectal adenocarcinoma	MLH1 NM_000249 c.307-2A>G	NA	NA	MLH1&PMS2 loss/Retained

B=BASIC3, K=KidsCanSeq; DX=Diagnosis; yrs=years; NA=Not available; LOH=Loss of heterozygosity; SNV=single nucleotide variant; TMB=Tumor mutation burden; IHC=Immunohistochemistry

Subsequent medical genetic evaluation occurred in 5 cases: physical examination revealed no skin findings or other features typical of CMMRD including café au lait macules, axillary freckling, and neurofibromas²⁰. Additional independent germline testing at an external laboratory for mutations and copy number variants in each case confirmed the MMR variants to be heterozygous without any second variant or copy number alteration detected as would be expected in CMMRD. Patient 5 died of glioblastoma before genetics evaluation, however, there was no history of skin findings for the patient and evaluation of two siblings revealed no skin findings. Comprehensive gene analysis revealed the known variant in both siblings but did not detect a second MMR variant. This was the only case in which genetic testing of minor siblings was performed given the inability to determine whether the proband had a missed CMMRD diagnosis. In all 6 cases, the variant was inherited from a parent (one paternal and five maternal). In two patients (case 2 and 5), additional family history suggestive of Lynch syndrome was reported by the family following results disclosure.

Tumor analysis (genomic profiling and/or immunohistochemistry) was completed in 4 of 6 patients with distinct tumors (neuroblastoma, astrocytoma, glioblastoma, and rectal adenocarcinoma) with available samples as detailed in Table 1. MMR IHC was performed in the tumor from 4 patients demonstrating loss of staining in the affected MMR protein(s) in 3 tumors (MSH6 in the astrocytoma (case 3), MLH1 and PMS2 in both the glioblastoma (case 4) and rectal adenocarcinoma (case 6)) with retention in tumor-adjacent normal tissue. Although the tumor from the rectal adenocarcinoma was not analyzed further, a second hit in the form of a somatic loss of function variant or deletion of the wildtype MMR gene allele was detected in tumor from the astrocytoma and glioblastoma (cases 3 and 4). This is consistent with the heterozygous germline variant playing a role in cancer development in these two tumors. Both of these tumors had increased TMB with an intermediate (7.26 mut/MB) TMB in case 4, and an ultra-high TMB (>200 mut/MB) in case 3. Interestingly, the astrocytoma also harbored a somatic POLE exonuclease domain mutation (p.Pro436Ser) consistent with somatic POLE/POLD1 variants, previously reported in MMR-deficient tumors of adult Lynch syndrome patients associated with ultra-high increase in TMB. Mutation signature analysis revealed high correlation with signature 6 pattern (cosine values 0.90 and 0.86) in both tumors 3 and 4, like that reported in other MMR-deficient tumors¹⁸.

In only one of four patients (case 1), with tumor available for analysis, there was no evidence for a second hit in the MMR gene of interest, no protein loss by IHC in tumor and normal samples, and low TMB. Thus, this is the only patient with tumor available where the germline MMR variant did not appear to play a role in tumor (neuroblastoma) development and mutation burden.

4. Discussion

Our finding of Lynch syndrome in 0.8% of patients in the BASIC3 and KCS cohorts analyzed via germline exome and tumor analysis is similar to the 0.9% rate recently reported by Fiala et al from analysis of pediatric patients analyzed on the MSK-IMPACT^™ panel⁶. Although a small number of patients, the tumor analysis of our patients suggests that Lynch syndrome is playing an important role in tumorigenesis in 3 of 4 tumors analyzed. Whether causative of the proband’s current cancer or incidental, a diagnosis of Lynch syndrome has implications for surveillance for second malignancies in the patient and may impact cancer treatment decisions including eligibility for enrollment in clinical trials and use of targeted therapies such as immune checkpoint inhibitors^21,22. Given that not all pediatric patients currently undergo tumor genomic sequencing or IHC, a germline finding may be the first clue which subsequently leads to tumor analysis with the potential to identify treatment options. In addition, even when incidental, it may lead to the identification of other at-risk family members leading to preventative screening measures in adults as was demonstrated by all six variants in this cohort having been inherited from a parent.

As demonstrated here, tumor type or age of onset cannot differentiate between Lynch syndrome and CMMRD. Tumor and germline analysis both play a role in the evaluation of Lynch syndrome and CMMRD and they may occur in different sequence. Tumor analysis can be useful in two distinct clinical contexts. First, MMR IHC can effectively distinguish between these two syndromes in patients with germline variants, particularly given difficulty in identifying all PMS2 alleles. Second, in children who do not meet current recommendations for germline genetic evaluation at the time of diagnosis, tumor-only genomic profiling may provide the first clue to a potential germline condition in a child and reveal the MMR gene of interest. For example, cases 3 and 4 would have met MIPOGG criteria for genetic evaluation based on MMR IHC and/or the tumor genomic profile information.

Lynch and CMMRD have significantly different surveillance recommendations as multimodal screening for CMMRD patients in early childhood has been shown to reduce morbidity²³ whereas no screening is advised for Lynch syndrome in childhood. The correct diagnosis also allows accurate discussion of reproductive risk and cancer risk, particularly for siblings as well as other family members given the dominant versus recessive modes of inheritance.

Concordant with data from previous research demonstrating that 34% of variants considered to have moderate- to high-penetrance were unexpected in a pediatric oncology population⁶, our results indicate that Lynch syndrome may be among the diagnoses missed in patients with pediatric cancer. For this reason, including MMR genes on all tumor pediatric cancer panels is important particularly as MMR IHC is not routinely done on all pediatric cancer types. This data on unexpected findings also supports consideration of offering germline genetic testing that includes MMR genes to all newly diagnosed patients with pediatric cancer. Of note, the MMR genes are included in the American College of Medical Genetics (ACMG) secondary findings list²⁴ and thus may be evaluated in any pediatric patient who is undergoing exome or genome analysis. It is possible that there could be other genetic modifiers that result in an early cancer onset for some patients with Lynch syndrome. Future research into genetic modifiers in larger patient cohorts will be an important area of study. The diagnosis of Lynch syndrome in pediatric patients raises the subsequent question of genetic testing in healthy minor siblings and screening recommendations in this population. In our cohort, genetic testing of minors was only completed in the case of a deceased sibling where CCMRD could not be excluded in the proband. The impact of such a shift on testing practices and service delivery models, as well as that of a Lynch syndrome diagnosis on psychosocial status and patient-oriented outcomes in children, are critical topics for further study and clinical guidelines.

Abbreviations

BASIC3

Baylor College of Medicine Advancing Sequencing into Childhood Cancer Care

CLIA

Clinical Laboratory Information Act

CNS

Central nervous system

CMMRD

Constitutional mismatch repair deficiency

CSER

Clinical Sequencing Exploratory/Evidence Generating Research

FFPE

Formalin-Fixed Paraffin-Embedded

HGSC-CL

Human Genome Sequencing Center-Clinical Laboratory

IHC

immunohistochemistry

KCS

KidsCanSeq

MMR

Mismatch repair

MSI

Microsatellite instability

P/LP

Pathogenic/likely pathogenic

TMB

Tumor mutation burden

Data Availability

All participants were enrolled in one of two sequential National Human Genome Research Institute and National Cancer Institute Clinical Sequencing Exploratory/Evidence-Generating Research Consortium (CSER) Projects (BASIC3 and KidsCanSeq Trials). As a part of these protocols participants consent for data sharing. These cases are being submitted to the National Library of Medicine ClinVar database²⁵ at https://www.ncbi.nlm.nih.gov/clinvar/ and dbGaP Genotypes and Phenotypes²⁶ https://www.ncbi.nlm.nih.gov/gap/ under projects number phs001683.v2.p.1and phs002378.v1.p1.