The efficacy of DNA MMR enzyme immunohistochemistry as a screening test for hypermutated gliomas

A subset of gliomas has DNA repair defects that lead to hypermutated genomes. While such tumors are resistant to alkylating chemotherapies, they may also express more mutant neoantigens on their cell surfaces, and thus be more responsive to immunotherapies. A fast, inexpensive method of screening for hypermutated gliomas would therefore be of great clinical value. Since immunohistochemistry (IHC) for the DNA mismatch repair (MMR) proteins Msh2, Msh6, Mlh1, and Pms2 is already used to screen for hypermutated colorectal cancers, we sought to determine whether that panel might have similar utility in gliomas. MMR IHC was scored in 100 WHO grade I-IV gliomas with known tumor mutation burdens (TMB), while blinded to TMB data. Eight of 100 cases showed loss of one or more MMR proteins by IHC, and all 8 were hypermutated. Among the remaining 94 gliomas with intact MMR IHC, only one was hypermutated; that tumor had an inactivating mutation in another DNA repair gene, ATM. Overall accuracy, sensitivity, and specificity were 99%, 89%, and 100%, respectively. The strongest correlates with hypermutation were prior TMZ treatment, MGMT promoter methylation, and IDH1 mutation. Among MMR-deficient hypermutated gliomas, 50% contained both MMR-lost and MMR-retained tumor cells. Together, these data suggest that MMR IHC could be a viable front-line screening test for gliomas in which immunotherapy is being considered. They also suggest that not all cells in a hypermutated glioma may actually be MMR-deficient, a finding that might need to be considered when treating such tumors with immunotherapies.


INTRODUCTION
Gliomas are the most common primary brain tumors in adults [19]. Standard of care includes surgical resection followed by adjuvant radiation and temozolomide (TMZ), a DNA alkylating agent [24]. However, tumors nearly always recur and lose sensitivity to adjuvant therapy.
Next generation sequencing (NGS) is the current gold standard for detecting DNA MMR defects and quantifying TMB, but larger panels are costly and have prolonged turnaround time.
Targeted NGS panels, such as Glioseq [18], are less expensive and faster, but usually do not screen for MMR mutations and do not cover enough of the genome to reliably determine TMB.
A standardized quartet of IHC stains (Msh6, Msh2, Mlh1, and Pms2) is used to detect loss of normal MMR gene expression in colorectal cancers [22]. Because most pathology laboratories already have this MMR IHC panel for routine use, we sought to determine its utility as a screening test for hypermutated gliomas.

METHODS
The cohort consisted of 100 World Health Organization (WHO) grade I-IV gliomas from   the Northwestern Nervous System Tumor Bank with known TMB and MMR gene mutations, as   determined by Tempus xT NGS covering approximately 600 genes (Table 1) (Cell Marque MRQ-28 (288M-15), 1:50). Formalin-fixed, paraffin-embedded 4 µm thick tissue sections were baked at 60°C for 30-60 minutes before being deparaffinized and re-hydrated.
Antigen retrieval for Msh6, Msh2, and Pms2 was achieved using a Universal Retrieval (Abcam) buffer in a decloaking chamber reaching 110°C for 5-20 minutes. Antigen retrieval for Mlh1 used a citrate buffer (pH 6) in a decloaking chamber reaching 110°C for 10 minutes. Slides were cooled to room temperature and washed in TBS before neutralizing endogenous peroxidase (Biocare Peroxidase 1). Slides were then treated with a serum-free casein background block (Biocare Background Sniper) before pre-incubation in a 10% goat serum block for 60 minutes.
Primary antibody was added to the slides for overnight incubation at 4°C. After incubation, slides were washed in 3 5-minute washes with TBS-T before incubating in HRP polymer (Biocare MACH 4 Universal HRP Polymer). Reaction products were visualized with DAB (Biocare Betazoid DAB Chromogen Kit). Slides were counterstained with hematoxylin, dehydrated, and mounted with xylene-based mounting media.
Each IHC marker was examined under light microscopy by two independent reviewers (MM and CH) while blinded to NGS data and TMB. Each IHC marker was scored as "retained" or "lost." In cases with lost MMR expression, the pattern (homogeneous versus heterogeneous) was noted. Nonneoplastic cells within each glioma (e.g. lymphocytes, endothelial cells) served as internal positive controls. Interobserver discrepancies were resolved by reviewing equivocal cases together. Regression analyses were performed using http://vassarstats.net/multU.html All rights reserved. No reuse allowed without permission.
author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/19012005 doi: medRxiv preprint

RESULTS
The cohort of 100 gliomas with NGS and TMB data included 70 grade IV GBMs, 13 grade III astrocytomas, 4 grade II astrocytomas, 2 grade III oligodendrogliomas, 7 grade II oligodendrogliomas, and 4 grade I gliomas ( Table 1). Eight (8%) showed loss of at least one MMR protein by IHC (Figure 1, Table 2). All 8 gliomas with MMR loss had TMB >20, and all 8 had mutations and/or copy number losses in corresponding MMR genes ( Table 2). Of the remaining 92 gliomas with intact MMR IHC, only one was hypermutated (TMB=29.5). That glioma did not have mutations in MSH2, MSH6, MLH1, or PMS2, but instead contained an inactivating splice site mutation in ATM (Table 2). Sensitivity, specificity, and overall accuracy of MMR IHC for identifying hypermutated gliomas in this cohort was 88%, 100%, and 99%, respectively.
All 9 hypermutated gliomas had MGMT promoter methylation and were post-TMZ recurrences, and 5/9 (56%) were IDH1 mutant ( Table 2). Correlation and multiple regression analyses confirmed that the variables most strongly associated with hypermutation were prior treatment with TMZ, MGMT methylation, and IDH1 mutation ( Table 3 and Table 4).
In 4 of 8 gliomas with lost MMR expression, the pattern of loss was clearly heterogeneous, as some tumor cells retained all MMR enzymes, while other cells lost expression of one or more MMR enzymes (Figure 2). All rights reserved. No reuse allowed without permission.
author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/19012005 doi: medRxiv preprint

DISCUSSION
Despite their generally aggressive behavior, gliomas tend to have low TMB relative to most other kinds of cancer [2]. However, gliomas that are hypermutated, either at initial presentation or recurrence, may be ideal targets for immunotherapy. Such gliomas usually show increased numbers of infiltrating CD8+ cytotoxic T cells [17,27], which is consistent with the postulate that hypermutated gliomas are more immunogenic.
Hypermutated gliomas have been a subject of intense investigation for some time, though the reported frequencies of hypermutation vary markedly due to differences in cohort selection. In our screening of over 660 untreated sporadic grade II-IV TCGA gliomas in GlioVis were hypermutated, and 7 contained mutations in DNA repair genes [13]. In our own cohort, 9/100 gliomas were hypermutated, all 9 had been previously treated with TMZ, all 9 had MGMT promoter methylation, and 5/9 were IDH1 mutant ( Table 2). Screening for hypermutationassociated MMR defects therefore appears to be of greatest value in recurrent, post-TMZ gliomas, especially MGMT-methylated and/or IDH1 mutant tumors ( Table 3 and Table 4).
Although the Msh2, Msh6, Mlh1, and Pms2 IHC panel is used to screen colorectal cancer, mutations in other DNA repair genes have also been reported in post-TMZ hypermutated gliomas, including MSH4, MSH5, MLH3, PMS1, POLE, and POLD1 [5, 10, 13,26]. In our own cohort, we found a hypermutated glioma with an inactivating mutation in yet All rights reserved. No reuse allowed without permission. author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/19012005 doi: medRxiv preprint another gene associated with DNA repair, ATM ( Interpretation of MMR IHC in gliomas is relatively straightforward, especially since nonneoplastic cells within the glioma serve as a reliable positive control (Figure 1) (Figure 2 and Table 2) underscores the fact that TMB is just a mathematical average of the specimen that was submitted for NGS, and that subclones may exist in "hypermutated" gliomas that are not necessarily hypermutated. Conversely, hypermutated subclones could potentially exist in tumors whose overall TMB has not yet reached the widely accepted cutoff of 20 mutations/Mb, although we did not see this in our own cohort (not shown).
In sum, DNA MMR enzyme IHC can serve as a rapid, low-cost method of screening for hypermutated gliomas. Highest yield for screening includes recurrent gliomas with MGMT promoter methylation and/or IDH1 mutations. While the current panel used for colorectal cancers has very good sensitivity and excellent specificity, adding more DNA repair markers would further enhance its value. Finally, it may be valuable to consider heterogeneity in hypermutated gliomas as a possible predictor of immunotherapy response. All rights reserved. No reuse allowed without permission. author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which was not peer-reviewed) is the TMZ=temozolomide.
All rights reserved. No reuse allowed without permission.
author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/19012005 doi: medRxiv preprint author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/19012005 doi: medRxiv preprint   author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/19012005 doi: medRxiv preprint