Sporadic and Lynch syndrome-associated mismatch repair-deficient brain tumors

Mismatch repair-deficient (MMRD) brain tumors are rare among primary brain tumors and can be induced by germline or sporadic mutations. Here, we report 13 MMRD-associated (9 sporadic and 4 Lynch syndrome) primary brain tumors to determine clinicopathological and molecular characteristics and biological behavior. Our 13 MMRD brain tumors included glioblastoma (GBM) IDH-wildtype (n = 9) including 1 gliosarcoma, astrocytoma IDH-mutant WHO grade 4 (n = 2), diffuse midline glioma (DMG) H3 K27M-mutant (n = 1), and pleomorphic xanthoastrocytoma (PXA) (n = 1). Next-generation sequencing using a brain tumor-targeted gene panel, microsatellite instability (MSI) testing, Sanger sequencing for germline MMR gene mutation, immunohistochemistry of MMR proteins, and clinicopathological and survival analysis were performed. There were many accompanying mutations, suggesting a high tumor mutational burden (TMB) in 77%, but TMB was absent in one case of GBM, IDH-wildtype, DMG, and PXA, respectively. MSH2, MLH1, MSH6, and PMS2 mutations were found in 31%, 31%, 31% and 7% of patients, respectively. MSI-high and MSI-low were found in 50% and 8% of these gliomas, respectively and 34% was MSI-stable. All Lynch syndrome-associated GBMs had MSI-high. In addition, 77% (10/13) had histopathologically multinucleated giant cells. The progression-free survival tended to be poorer than the patients with no MMRD gliomas, but the number and follow-up duration of our patients were insufficient to get statistical significance. In the present study, we found that the most common MMRD primary brain tumor was GBM IDH-wildtype. The genetic profile of MMRD GBM was different from that of conventional GBM. MMRD gliomas with TMB and MSI-H may be sensitive to immunotherapy but resistant to temozolomide. Our findings can help develop better treatment options.


INTRODUCTION
The DNA mismatch repair (MMR) system is responsible for the prevention of genomic instability in cells and is controlled by MMR genes. Those are mutL homolog 1 (MLH1), encoded at chromosome 3p21.3, mutS homolog 2 (MSH2) at chromosome 2p22-21, mutS homolog 6 (MSH6) at chromosome 2p16 and postmeiotic segregation increased 2 (PMS2) at chromosome 7p22.2. MMR deficiency (MMRD) can be caused by germline or sporadic mutations or promoter methylation of MMR genes, which is associated with microsatellite instability (MSI) and tumor mutational burden (TMB). Therefore, it contributes to tumorigenesis, poor outcomes, and acquired drug resistance to alkylating agents that mediate the formation of O 6 methylguanine-containing mismatches 1,2 . TMB is considered a potential biomarker for immune checkpoint therapy 3,4 .
Lynch syndrome is an autosomal dominant hereditary cancer syndrome that was originally reported by Warthin in 1913 and is also known as hereditary nonpolyposis colorectal cancer syndrome [5][6][7] . Lynch syndrome and constitutional MMRD syndrome are caused by heterozygous and homozygous germline mutations in one of the MMR genes, respectively 8,9 . Germline mutations in MSH2 (40-50%) and MLH1 (30-37%) are the most frequent, and MSH6 and PMS2 mutations are found in 7-13% and up to 9% of cases, respectively 9,10 . Patients with Lynch syndrome have a lifetime risk of 50-80% for developing colorectal cancer and 40-60% for developing endometrial cancer and less commonly cancers of the upper urinary tract, hepatobiliary tract, small intestine, ovary, and skin [11][12][13][14] . Lynch syndrome also quadruples the risk of brain tumors, predominantly high-grade gliomas (HGGs) [11][12][13] . For an accurate diagnosis of sporadic and hereditary MMRD tumors, immunohistochemistry (IHC) of MMR proteins in the tumors, molecular studies to detect MMR gene mutations or methylation, MSI, and genetic testing of affected family members are required 15,16 .
Lynch syndrome-associated MMRD tumors often exhibit a MSI-H phenotype 17 . However, since MSI-H is also frequently observed in sporadic colorectal cancers, genetic testing for germline MMR genes is essential 10 .
Although MMRD is well recognized in colorectal and endometrial carcinomas, MMRD brain tumors remain poorly understood. Here, we report patients with nine sporadic MMR mutation-and four Lynch syndrome-associated HGGs with MSI-H and high TMB.

MATERIALS AND METHODS Case summary
Among 740 brain tumors from the archives of the Department of Pathology, Seoul National University Hospital, archived from 2018 to 2021 that were subjected to next-generation sequencing (NGS), 13 MMRD brain tumors were found. The tumors included glioblastoma (GBM) IDH-wildtype (n = 9), including one gliosarcoma, astrocytoma IDH-mutant WHO grade 4 (n = 2), diffuse midline glioma (DMG) H3 K27M-mutant (n = 1) and pleomorphic xanthoastrocytoma (PXA) (n = 1). The proportion of MMRD primary brain tumors in our hospital, including cases of MMRD-associated pineal teratocarcinoma (n = 1) and meningioma (n = 1) that were not included in this study, was~2.0%. The age of the 13 patients ranged from 11 to 78 years (median age: 50 years), and the male-to-female ratio was 1.2:1.
Ten patients had recurrent gliomas. As a result of the NGS study of the initial tumors, two of them were found to have developed MMRD after concurrent chemotherapy and radiotherapy (CCRT) (Supplementary  Tables 1 and 4).
Four patients had Lynch syndrome, confirmed by the germline Sanger sequencing, but the concurrent malignancy was found in two patients who had histories of extracrainal cancers. The #4 patient had Lynch syndrome with multiple cancers; he was diagnosed with prostatic adenocarcinoma (Gleason score 8) at the age of 61 years and colonic and jejunal cancers at the age of 62 years. His colonic tumors showed a mucinous subtype in the mid-ascending colon and poorly differentiated adenocarcinoma with signet ring cell features in the proximal ascending colon and the jejunum. These subtypes and locations of intestinal adenocarcinoma are known to be associated with Lynch syndrome 18 . Immunohistochemically, the ascending colonic and jejunal adenocarcinomas showed a loss of MSH2 and MSH6 proteins in the tumor cell nuclei, but the MLH1 and PMS2 proteins were retained. Interestingly, prostatic adenocarcinoma is not an MMRD tumor with retained expression of all four MMR proteins.
The pedigree chart of these four patients with Lynch syndrome suggested an autosomal dominant inheritance of the disease (Fig. 1). All patients underwent craniotomy and tumor resection. Clinical manifestations are summarized in Table 1.

Magnetic resonance imaging (MRI)
A total of 85% (11/13) of the tumors were located in the supratentorial region (temporal, frontal, frontotemporal, parietal, or occipital lobes or thalamus or corpus callosum and cingulate gyrus or basal ganglia) ( Table 1). The remaining two tumors were found in the posterior fossa including the cerebellum. MRI revealed high-and low-signal-intensity masses on T2 and T1 imaging, with rim or heterogeneous enhancement and perilesional edema in the most patients. All MRI findings suggested HGG (Fig. 2).

Histopathology and immunohistochemistry (IHC) of MMRD brain tumors
Neutral formalin-fixed paraffin-embedded (FFPE) tissues were cut into slices of 3 μm thickness for H&E staining and IHC. Tissue sections were stained with anti-IDH1 R132H (H09) monoclonal antibody (1:100 dilution, Dianova, Hamburg, Germany), anti-ATRX polyclonal antibody HPA001906 IHC staining was carried out using a standard avidin-biotin-peroxidase method with a BenchMark ULTRA system (Roche Diagnostics). The primary antibodies used in this study are listed in Table 2.
We used a proper positive control. Most cases had internal positive controls on the slides (Fig. 3), and for the negative control, we omitted the primary antibodies. The Ki67 labeling index was calculated on virtual Leica Biosystems slides (Aperio ScanScope system) using the SpectrumPlus Nuclear Algorithm n9 image analyzer. The positive controls for PD1 and PD-L1 were a known PD1/PD-L1-positive tumor and positive lymphocytes.
Complete loss of expression of MLH1, PMS2, MSH2, or MSH6 in tumor cell nuclei on IHC indicated loss of the respective protein, and heterogeneous loss of expression was defined as a mixture of areas with loss of expression and retained expression. According to Graham et al.'s paper, heterogeneous MSH6 loss is uncommon but exists and is usually  caused by MSI and instability of the MSH6 exon 5 polycytosine tract but is not associated with a germline MSH6 mutation 19 .
In the in vivo state, MLH1/PMS2 and MSH2/MSH6 form two functional pairs. When either MLH1 or MSH2 is lost, the partner protein is destabilized and degraded, resulting in the loss of the partner MMR protein. However, the opposite is not true; the absence of PMS2 or MSH6 does not affect stability because MLH1 and MSH2 can bind to and stabilize other molecules 20 . Therefore, we carefully examined the protein expression of these pairs. Histopathology was reviewed by two pathologists (HK and SHP) according to the histopathological criteria defined by the 2021 WHO classification 21 and cIMPACT-NOW updates 22 .
DNA and RNA extraction for next-generation sequencing (NGS), MGMT promoter (MGMTp) methylation studies, and microsatellite instability studies Representative areas of the tumor from FFPE tissue on H&E-stained sections with at least 90% tumor cell content were outlined for macrodissection. DNA/RNA extraction was performed from these FFPE tissues using the Maxwell ® RSC DNA/RNA FFPE Kit (Promega, USA) according to the manufacturer's instructions. For MSI polymerase chain reaction (PCR) studies using Bethesda's five-marker panel and for the germline MMR mutation study, paired tumor and normal tissue samples were used for genomic DNA extraction. Definitively normal tissue adjacent to the brain tumor was used as the normal counterpart. If it was difficult to find normal tissue on the H&E-stained slide, we verified that it was normal tissue with Ki-67, EGFR, or TP53 immunostaining. If no normal tissue was present in the brain tumor biopsy sample, biopsied extracranial normal tissue or blood was used as the normal counterpart. Genomic DNA was subjected to PCR with fluorescently labeled oligonucleotide primers for five microsatellite loci (BAT25, BAT26, D2S123, D5S346, and D17S250), followed by capillary electrophoresis on an ABI 3100 genetic analyzer (Applied Biosystems, Foster, CA, US). The instability of the investigated loci was defined as a change in the length of the PCR product in a tumor sample compared to the length of the PCR product in the paired normal sample. MSI status was classified as MSI-H if the sample showed instability at two or more microsatellite loci, MSI-low (MSI-L) if the sample showed instability at one locus, and microsatellite stable (MSS) if there was no instability.

NGS and pipelines of analysis of the somatic mutations
NGS studies were performed with tumor DNA extracted from FFPE tumor tissue and NEXTSeq Dx505 using a customized brain tumor gene panel (The FIRST brain tumor panel established by the Department of Pathology, SNUH, and approved by the Korea Food and Drug Administration), which assesses 207 brain tumor-associated genes and 54 fusion genes, including 4 MMR genes (Supplementary Table 2). Fusion genes were sequenced using RNA. Somatic mutations were detected using the Genome Analysis Toolkit (GATK) Mutect2 v4.1.4.1. with default parameters 23 . To avoid germline variant contamination, we used the gnomad.hg19.vcf Genome Aggregation Database (gnomAD) 24 and 1000 g_pon.hg19.vcf files, which include a normal panel for 1000 genomes. The files were provided by the GATK resource bundle. After calling somatic mutations, all variants were annotated by ANNOVAR (https://doc-openbio.readthedocs.io/projects/ annovar/en/latest/) 25 .
We extracted recent 20 cases of IDH-mutant and 60 cases of IDHwildtype grade 4 gliomas from our hospital NGS data and we compare the number of mutations between MMRD gliomas and non-MMRD gliomas.

R programming
Clinical information, mutations, and copy number variations were summarized with Oncoprint data, which were generated using the R package ComplexHeatmap (version 2.7.

Survival analysis
Kaplan-Meier survival analysis was performed, and IDH-wildtype GBMs and IDH-mutant WHO grade 4 astrocytoma with intact MMR were compared. Fig. 2 The brain MRI images of case 4 with Lynch syndrome. A sagittal T1-weighted (postcontrast), B axial T2-weighted, and C T2 FLAIR MRI results, showing an~6 cm-long diameter enhancing mass with perilesional edema in the right occipital lobe. Case 2 (glioblastoma IDHwildtype) D sagittal T1-weighted (postcontrast), E axial T2-weighted, and F T2 FLAIR MRI results, revealing an~5.7 cm heterogeneous mass in the right parietal lobe and midline shift.
The control cases were previously established cohort in Park et al.'s paper 28 . PFS was defined as the time from first surgery for the brain tumor to disease progression, while OS was defined as the time from the first surgery for the brain tumor to death.
Among the four MLH1-mutant tumors, complete loss of both MLH1 and PMS2 IHC in the tumor cells was present in three cases (Patient #1, 9, and 10), but one case showed complete loss of MLH1 and heterogeneous loss of the partner protein PMS2 (Patient #11). Four MSH2-mutant tumors showed complete loss of MSH2 IHC but a heterogeneous loss of the partner protein MSH6 (Patient #2, 4-6) (Fig. 3F, G) (Table 3). Three (Cases #3, 12, and 13) of four MSH6-mutant tumors had MSH6 loss only (Fig. 3C, G), but both MSH6 and PMS2 losses were also found in one MSH6-mutant tumor (Patient #8). These results were expected because it is already known that MSH2 loss can result in a heterogeneous loss of the partner protein MSH6 as a result of MSI and instability of the MSH6 polycytosine tract, but MSH6-mutant tumors are known to have no partner protein loss 19 .
PDL-1 was weakly positive in 1% of tumor cells in three patients (cases #6, #8, and #9), but it was not expressed in the other cases. PD-1 was positive in a few immune cells in case #6 and #8 (positive in up to 4 cells/HPF) and was not expressed in the other tumors ( Table 3). The Ki-67 labeling index was relatively high (Table 1).

Molecular analysis and next-generation sequencing (NGS) study
MSI-H was found in 50% (6/12) of patients ( Supplementary Fig. 1) including all four Lynch syndrome-associated cases, MSI-L was found in one patient (8%), and five patients exhibited MSS. The MSI study could not be performed in the remaining one case because there was no normal tissue (Table 3). MGMTp methylation was found in 53.8% (7/13) of tumors, and among GBMs and gliosarcomas, 55.5% (5/9) of them had MGMTp methylation. The NGS studies found MMR gene mutations as well as multiple pathogenic mutations and variants of uncertain significance (Fig. 4) The MMR gene mutations were verified by IHC (Fig. 3). Notably, the MSH2 p.Tyr405* mutation found in patients with Lynch syndrome is a known germline variant but has never been reported as a somatic mutation in the OnkoKB and Cosmic databases 29 .   Table 1). The number of nonsense mutation was higher in MMRD-HGG (average 23.0-23.6) and than HGGs without MMRD (average 4.7-5.8) (Fig. 5). Of the 3 MMRD gliomas with low mutation numbers (5 SNPs), GBM IDH-wt (Case #7) had many copy number aberrations, but the remaining DMG (Case #7) and PXA (Case #13) did not have many copy number aberrations, eventhough they had MMRD (Supplementary Table 3).
Germline studies of MMR genes by Sanger sequencing revealed germline mutations in three cases (Cases #3, #8, and #13, Supplementary Fig. 2). In one remaining patient, the germline study could not be performed because there was no normal tissue   or blood. This patient had a solitary brain tumor only, and additional clinicopathological reviews suggested sporadic MMRD brain tumors.

Treatment, follow-up of patients, and survival analysis (PFS and OS)
After the surgery, nine patients with GBM and DMG were treated with CCRT with temozolomide (TMZ), and one patient with GBM (Case #6) was treated with postoperative radiotherapy (PO-RT) only. However, the remaining three patients did not receive adjuvant therapy (Table 1). The intestinal carcinoma of a Lynch syndrome patient (Case 4) who had been treated with postoperative adjuvant 5-fluorouracil, leucovorin, and oxaliplatin did not recur for 8 years. Instead, this patient's MMR-intact prostatic adenocarcinoma metastasized to multiple bones, including the rib, thoracic spine, sacrum, and pelvic bones, during the last 7 years, despite radiation therapy, chemotherapy (docetaxel and abiraterone/prednisone), and androgen deprivation therapy. Four patients (31%) died from diseases, 10 patients (77%) had recurrences of tumors and the remaining 3 patients did not have enough follow-up period (~2 months in all 3 patients). Patient #1 recurred after 1 year of treatment, despite gross total resection (GTR) of the tumor plus CCRT and gamma knife stereotactic radiosurgery, and died in 57 months after the initial surgery. Patient #2 died at 12 months after GTR with no adjuvant therapy, and patient #10 died at 22 after GTR and CCRT with recurrences. Case #11 with astrocytoma IDH-mutant recurred in 39 months and died in 59 months after GTR. One patient (Case #12) with sporadic astrocytoma IDH-mutant WHO grade 4 did not recur. PFS of the patients with recurrent tumors was 1 month to 42 months.
Kaplan-Meier survival analysis showed a trend for lower PFS in patients with MMRD-HGG than those in patients without MMRD-HGG (p = 0.69, p = 0.64) (Fig. 6A, C). However, all the results did not have a statistical significance due to the small number of cases, short follow-up duration, and also better patient care. (P>0.05) (Fig. 6B, D).
DISCUSSION MMRD brain tumors are very rare, accounting for~2% of primary brain tumors, and are also histopathologically diverse, including GBM, astrocytoma, oligodendroglioma, gliosarcoma, anaplastic PXA, medulloblastoma, and neuroblastoma 30 . Among them, GBM and high-grade astrocytoma are the most common sporadic or 21% and 12% of cases, respectively. There was one case each with frameshift and insertion mutations.
Inactivating mutations of TP53 and chromosomal instability following the loss of MMR function are common genetic abnormalities 37 . However, because colorectal carcinomas usually have diploid or near-diploid DNA content with a few chromosomal aberrations, colorectal carcinomas with MMRD usually do not have inactivating mutations of TP53 and chromosomal instability 37  amplification or mutation of PDGFRA. In addition, copy number aberrations of various genes and many other VUS, suggesting a high TMB. Because of the presence of mutations in the TP53, NF1, and ATM genes, concomitant or underlying Li-Fraumeni, neurofibromatosis type 1, or ataxia-telangiectasia cancer syndrome needed to be ruled out. However, these cancer-predisposing syndromes require germline mutations for diagnosis, and the associated tumor types are different from those seen in our Lynch syndrome cases; Li-Fraumeni syndrome-associated cancers are usually sarcomas, breast cancers, brain tumors, and leukemias 42 . Brain tumors of Li-Fraumeni syndrome can appear as low-grade gliomas or HGGs 43 . Patients with neurofibromatosis type 1 typically have neurofibroma, optic nerve glioma, or malignant peripheral nerve sheath tumor, and patients with ataxia-telangiectasia syndrome usually have non-Hodgkin lymphoma and leukemia 42 .
Hypermutation can occur in recurrent tumors after TMZ treatment via MMRD and MSI-H-related mechanisms, whereas these alterations are extremely rare in primary brain tumors 44 . MMRD is often associated with MSI, and it is one of the mechanisms behind acquired resistance to the alkylating chemotherapeutic agent TMZ in gliomas 44 .
MMRD is also related to TMB and neoantigen loads, therefore, can be a target of immunotherapy. Generally, MMRD tumors have MSI-H 17 . MSI-H is generally uncommon in sporadic brain tumors, but if it is present, it may represent MMR gene germline mutation carriers 45 . Although one study reported that MSI is rare in Lynch syndrome-associated brain tumors 46 , it can occur as a result of the loss of MMR function 17 . Our 4 Lynch associated HGG had MSI-H.
The TMB could not be verified in our cases due to the limitations of the targeted gene panel (207 brain tumor-targeted genes and 54 fusion genes), but most of our cases possibly had TMB because of many pathogenic and likely pathogenic mutations and VUS. However, three cases (1 GBM IDHwt, 1 PXA, 1 DMG) had less than five SNPs with variable copy number aberrations (Fig. 5, Supplementary Table 1).
Among MMRD-associated tumors, the number of methylated genes is known to be the lowest in brain tumors and the highest in colorectal cancers 47 . The methylation of MGMTp is known to occur in approximately half of MMR-intact gliomas; therefore, the incidence of MGMTp methylation in MMRD gliomas could be similar to that in MMR-intact gliomas 48 . Methylation of MGMTp was observed in about half (53.8%) of our cases.
According to a recent study, replication repair-deficient (RRD) HGGs have a global methylation pattern distinct from that of replication repair-intact HGGs 49 . This methylation pattern varies according to key driver mutations; for example, the IDH1 R132H and H3F3A K27M mutations, and the location of the tumor. Even in the same RRD HGG case, the methylation patterns of the initial tumor and recurrent tumor can be different. In addition to the methylation pattern, hypomethylation in specific gene pathways associated with critical cellular functions occurs in RRD HGGs, which can be used as a target for treatment. Therefore, methylation patterns should be studied to help classify and treat MMRD gliomas.
Loss of nuclear MMR protein was 100% correlated with MMR gene mutation in our cases, suggesting that IHC is sufficient to identify MMRD in brain tumors. MLH1/PMS2 and MSH2/MSH6 form two functional pairs in vivo. Loss of MLH1 or MSH2 destabilizes and degrades the partner protein, so MLH1/PMS2 and MSH2/MSH6 pair losses are common 50 . However, the opposite is not true; the absence of PMS2 or MSH6 does not affect the stability of the partner protein, because MLH1 and MSH2 can bind to and stabilize other molecules 20 . Therefore, the expression of these protein pairs must be carefully investigated. Most of our cases showed these patterns.
MMRD is often associated with MSI-H and TMB, which can contribute to poor outcomes 1,2 . Unexpectedly, PFS and OS were not significantly different between glioma patients with and without MMRD (P>0.05) (Fig. 6), possibly due to the limited number of MMRD cases and the short follow-up duration. To determine the biological behavior of MMRD brain tumors, more large-scale, well-designed prospective studies are required. Randomized controlled trials are necessary to improve the treatment of MMRD patients. PD1/PDL1 IHC staining was almost negative in our cases. Identifying patients with MMRD brain tumors is important for appropriate treatment strategies for patients with sporadic MMRD tumors and family members at risk for Lynch syndrome or CMMRD. Since the PD1/PDL-1 IHC staining was negative in most of our MMRD brain tumors, other options, such as direct identification of MMRD via NGS or IHC studies in primary brain tumors, may be needed to determine the indications for immunotherapy. These MMRD gliomas could be sensitive to immunotherapy but resistant to TMZ 51 .
In conclusion, we analyzed nine sporadic MMRD and four Lynch syndrome-associated gliomas in this study, representing a rare event, about 2.0% of the primary brain tumor cases in our hospital. Most (92%) cases were grade 4 except for one PXA, which was WHO grade 2. MMRD developed after CCRT in 2 cases, which were both IDH-mutant and IDH-wildtype gliomas. These MMRD gliomas contained many pathogenic and benign or likely benign mutations and VUS, suggesting high TMB, but one GBM, DMG, and PXA did not have TMB despite the MMRD. Half of our MMRDgliomas and all Lynch syndrome-associated GBMs had MSI-H. Genetic profile of MMRD-associated GBMs was different from that of conventional GBMs. The MMRD GBM did not have EGFR amplification, PTEN homozygous deletion, or concurrent 7p gain and 10q loss. TERTp mutation was found in only 33% (3/9) of IDHwildtype MMRD GBMs. MGMTp methylation was found in 54% of our MMRD cases. The PFS of our MMRD patients had a tendency of early recurrence. More studies are needed in the form of clinical trials of immunotherapy for MMRD brain tumors.

DATA AVAILABILITY
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.