Purpose:

In-depth characterization of recurrent glioblastoma (rGBM) might contribute to a better understanding of the mechanisms behind tumor progression and enable rGBM treatment with targeted drugs.

Experimental Design: In this study, GBM samples were collected at diagnosis and recurrence from adult patients treated with Stupp protocol. Expression of mismatch repair (MMR) proteins was evaluated by IHC, followed by whole exome sequencing (WES) of tumor samples showing loss of MSH6 reactivity. Established genetic, epigenetic, and immunologic markers were assessed by standard methods and correlated with loss of MMR proteins and patient survival.

Results:

Expression of MMR proteins was partially or completely lost in 25.9% rGBM samples. Specifically, 12 samples showed partial or total MSH6 expression reduction. Conversely, 96.4% of GBM samples at diagnosis expressed MMR markers. WES disclosed lack of variants in MMR genes in primary samples, whereas two MSH6-negative rGBM samples shared a c.3438+1G>A* splicing MSH6 variant with a potential loss of function effect. MSH6-negative rGBM specimens had high tumor mutational burden (TMB), but no microsatellite instability. In contrast, GBM samples with partial loss of MMR proteins disclosed low TMB. MMR-deficient rGBM showed significant telomere shortening and MGMT methylation and are characterized by highly heterogeneous MHC class I expression.

Conclusions:

Multilevel profiling of MMR-deficient rGBM uncovered hypermutated genotype uncoupled from enriched expression of immune-related markers. Assessment of MHC class I expression and TMB should be included in protocols aiming to identify rGBM patients potentially eligible for treatment with drugs targeting immune-checkpoint inhibitors.

Translational Relevance

Defects of the mismatch repair (MMR) system are of increasing interest to the medical community, as they could associate with high tumor mutational burden (TMB) and response to immune-checkpoint inhibitors. However, efforts to identify patients who could be good responders to immune-checkpoint inhibitors should probably take into account both state-of-the-art genetic and immune-related markers. In this regard, we show here that relapsed glioblastoma (GBM) samples completely lacking expression of MMR proteins have high TMB but also heterogeneous expression of MHC class I molecules, which is required for effective killing of tumor cells by cytotoxic T lymphocytes. Assessment of both TMB and MHC class I expression should be considered when developing algorithms to predict response to immune checkpoint inhibitors in GBM and other solid tumors.

Glioblastoma (GBM) is the most malignant type of brain cancer, almost invariably relapsing after standard therapy, which includes surgery, radiotherapy, and temozolomide (TMZ)-based chemotherapy. The genetics of GBM at diagnosis has been extensively investigated, and most driver mutations in GBM have been identified (1, 2). Nevertheless, targeted therapeutic approaches for GBM with clear impact on survival are still unavailable (3). Among various reasons, marked intratumoral genetic heterogeneity could play a leading role, as it endorses rapid tumor adaptation to target therapy (2). Accordingly, it has been shown that only few alterations are shared between GBM at diagnosis and at relapse, supporting the hypothesis of therapy-driven selection of tumor cell subpopulations not prevalent in the initial tumor (4–6). For these reasons, GBM genotyping at relapse along with improved characterization of the tumor microenvironment represent fundamental steps which could be useful in the identification of targets not present at tumor onset.

Immunotherapy is considered a promising therapeutic approach for cancer, but therapeutic results are less compelling in GBM than in other solid tumors, and identification of response predictive biomarkers is strongly needed (7). Anecdotal reports suggest hypermutated pediatric GBM as a good responder to immunotherapy with antibodies against immune-checkpoint inhibitors, such as nivolumab (8). Intriguingly, previous studies uncovered a substantial amount of inactivating mutations of mismatch repair (MMR) genes in relapsed GBM, especially involving MSH6. MMR gene alterations are considered to be selected or caused by TMZ and associated with an extremely high mutational burden (4, 9, 10). These mutations were detected by interrogating cancer genomes with Sanger sequencing or next-generation sequencing. Apart from several case reports in childhood brain tumors associated with germline biallelic MMR gene mutations (8, 11), there is only limited knowledge about expression of MMR-encoded proteins in primary versus relapsed adult GBM. Conversely, assessment of MMR defects is routinely performed by IHC staining for the four main MMR-associated proteins (MSH6, MLH1, MSH2, and PMS2) in colorectal cancer, and results of IHC are strongly correlated with microsatellite instability (MSI; ref. 12).

Here, we evaluated expression of MMR proteins in a set of matched (diagnosis/relapse) adult GBM samples and correlated results with in situ immunologic parameters. We further analyzed the mutational status of MMR genes and the possible association with MSI and hypermutated genotypes. Moreover, we measured longitudinal changes involving established genetic and epigenetic markers of GBM and investigated their association with loss of MMR-associated proteins, immunologic features of the tumor microenvironment, and clinical outcome.

Patients and samples

We conducted this monoinstitutional study with permission from the Institutional Review Board (CE clearance n. 20/2017, released March 20, 2017) and in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki. We retrospectively analyzed all patients with GBM treated at the Veneto Institute of Oncology from May 2006 to May 2015. Written consent was obtained from all patients alive at the time of recruitment. Inclusion criteria for patient enrollment were the following: relapsed GBM after surgery and treatment with radiotherapy and TMZ (Stupp protocol), second surgery at relapse, histologically confirmed diagnosis of GBM at first and second surgeries, sufficient tumoral tissue for molecular analysis in both primary and relapsed GBM. We enrolled 57 patients having matched (diagnosis/relapse) GBM samples. At diagnosis, median age was 53.6 years; 21 patients (37%) were female, and 36 (63%) were male. Forty-six (80.7%) patients received a total resection at first surgery and 11 (19.3%) a partial resection based on radiologic assessment by brain MRI performed after surgery. At diagnosis, Eastern Cooperative Oncology Group Performance Status was 0 in 48 patients and 1 in 9 patients. At relapse, 33 patients (57.9%) received a total resection. After second surgery, all of the patients underwent a second-line treatment with chemotherapy: 35 patients (61.4%) received fotemustine, 12 (21%) TMZ dose-dense, 7 (12.2%) lomustine, and 3 (5.4%) bevacizumab.

Immunohistochemistry

This study planned several molecular and in situ analyses which required significant amounts of tissue. To overcome this problem, we performed IHC analysis on tissue microarrays (TMA). All formalin-fixed paraffin-embedded (FFPE) samples were processed using the Galileo CK3500 Arrayer (www.isenet.it), a semiautomatic and computer-assisted TMA platform, as previously described (13). Briefly, three neoplastic areas were selected and marked on hematoxylin and eosin–stained sections of each neoplastic sample. These areas were transferred onto the paraffin-embedded tissue samples, and the coordinates were recorded. Tissue cores (1 mm diameter) were punched out of these areas and neatly embedded in the donor TMA paraffin block. The three areas were selected as far as possible from each other. In all the cases, the three cores corresponded to at least 15% of the entire neoplastic area, and tumor purity was >70% for all samples, as evaluated by two pathologists (Supplementary Table S1).

IHC stainings were automatically performed using the Bond Polymer Refine Detection Kit (Leica Biosystems) in the BOND-MAX system (Leica Biosystems) on 4 μm-thick FFPE sections with the primary antibodies for MLH1 (clone ES05; Dako), PMS2 (clone EP51; Dako), MSH6 (clone EP49; Dako), MSH2 (clone FE11; Dako), CD8 (clone C8/144B; Dako), MHC class I (clone EMR8-5; Abcam), PD-L1 (clone 22C3; Dako), and PD-L2 (clone 176611; R&D). Stained slides were jointly evaluated by two pathologists.

MMR status was assessed by IHC according a two-tier classification: 0, absence of nuclear immunoreaction; 1, partial absence of nuclear immunoreaction, defined as an heterogeneous pattern of staining (i.e., coexistence of positive and of at least 30% negative tumor cells in the same tumor spot) or reduced immunostaining in tumor cells in comparison with internal normal controls (e.g., endothelial cells); 2, moderate/strong nuclear positivity. The status of deficient MMR (dMMR) was assessed when a case was completely negative for at least one protein. Partial absence of nuclear immunoreaction was also arbitrarily considered within the frame of alterations in the MMR machinery. This category was formulated according to the observation of a significant (>30%) fraction of negative neoplastic cells (mainly for MSH2 and/or MSH6) in relapsed GBMs in comparison with their matched first-surgery specimens. Notably, these negative cells were not clustered, but intermingled with their positive counterparts (Fig. 1).

Figure 1.

Loss of MMR proteins in relapsed GBM. Representative images of GBM samples showing loss of MSH6 or PMS2 expression by IHC (right plots). The primary GBM samples showed a strong diffuse nuclear expression of the MMR proteins (left plots). In the recurrent samples, residual nuclear positivity for the various markers in the relapse samples is accounted for by normal stromal cells. Magnification of ×20 was used.

Figure 1.

Loss of MMR proteins in relapsed GBM. Representative images of GBM samples showing loss of MSH6 or PMS2 expression by IHC (right plots). The primary GBM samples showed a strong diffuse nuclear expression of the MMR proteins (left plots). In the recurrent samples, residual nuclear positivity for the various markers in the relapse samples is accounted for by normal stromal cells. Magnification of ×20 was used.

Close modal

Density of CD8+ tumor infiltrating lymphocytes was evaluated in the three different tumor cores and described as average of positive cells per high-magnification field. Immunostaining for MHC-I and PD-L1 was defined as a percentage of positive tumor cells, whereas PD-L2 expression was evaluated using semiquantitative criteria on the intensity of the staining: absent or sporadic (0), low expression (1+), moderate expression (2+), and strong expression (3+).

Next-generation sequencing studies

Genomic DNA of nine initial and matched recurrent tumors was extracted from the same FFPE block used for IHC studies with QIAamp DNA Mini kit (QIAGEN), quantified by means of Qubit 2.0 Fluorometer (Invitrogen), and subjected to quality control using the Agilent 2100 Bioanalyzer (Agilent Technologies) prior to DNA fragmentation. Samples were sequenced using the SureSelect Human All Exon V6 60 Mb Kit (Agilent) on HiSeq2500 (Illumina) in paired-end mode (2 × 125 cycles). Libraries were processed with Illumina cBot (Illumina) following the manufacturer's recommendations, and the CASAVA 1.8.2 version of the Illumina pipeline was used for raw data analysis. Demultiplexed and trimmed reads were mapped by BWA to the hg19 human genome assembly (14). Variant calling was performed on nine matched primary-recurrent tumors pairs using MuTect2 (GATK version 3.6; refs. 15–17), considering specimens at diagnosis as tumoral sample and relapses as matched normal and vice versa. The detected variants were then annotated using ANNOVAR (15–17). Variants indicated as a confident somatic mutation (PASS) were considered for further investigation (18). In order to account for the discrepancy in mean coverage between samples and avoid potential false positive in mutation calling, results were further filtered by excluding positions with a variant allele frequency (VAF) lower than 10% and a variant sequencing depth lower than 5 reads (19). This preliminary variants list was shortened considering only the genetic alterations in protein-coding sequences or splicing sites and excluding both common SNPs (MAF > 5%) and synonymous single nucleotide variants (SNV). This generated candidate cohort with only the high confidence functional variations, which was used to determine the “raw” tumor mutational burden (TMB). Moreover, to narrow down the analysis, a list of well-established GBM driver genes and MMR genes (indicated in Supplementary Table S2) was used to focus on somatic variants of interest (20).

MSI

MSI testing was performed by fluorescent PCR and capillary electrophoresis using a panel of 5 polyA mononucleotide markers (BAT25, BAT26, NR21, NR24, and NR27). It has been shown that a panel of quasi-monomorphic mononucleotide repeats obviates the need of normal DNA, and it is more specific and sensitive than using dinucleotide repeats, especially for MSH6 deficient tumors (21, 22).

MGMT promoter methylation assay by pyrosequencing

Five hundred ng of DNA was modified by sodium bisulfite, which converts unmethylated cytosine to uracil, following the manufacturer's instructions of the MGMT plus Kit (Diatech Pharmacogenetics). Modified DNA was subjected to PCR using the Rotor-Gene Q, and during the amplification, thymine was incorporated at the uracil position. PCR products were then sequenced, and the extent of methylation at each of the 10 analyzed CpG sites (CpG 74-83; chr 10:131,265,507–131,265,556) was evaluated by pyrosequencing using the PyroMark Q96 ID system (Qiagen). The assay evaluates a cytosine (position chr10: 131,265,528) not followed by a guanine and therefore unmethylated, to test the incomplete conversion by sodium bisulfite for each sample. Methylation extent is obtained calculating the average of methylation of the 10 CpG sites by the PyroMark CpG software (Qiagen).

An analytical cutoff of 5% was considered, as proposed by the MGMT plus Kit, to discriminate methylated from unmethylated samples.

TERT promoter analysis

Genomic DNA amplification for TERT promoter region (260 bp) containing 228C>T and 250C>T mutation sites, as well as the SNP rs2853669 (349T>C) was performed using the following primers: forward 5′-CCCACGTGCGCAGCAGGAC-3′ and reverse 5′-CTCCCAGTGGATTCGCGGGC-3′. PCR was carried out in a 50 μL volume containing 50 ng DNA, 1X PCR buffer II (Applied Biosystems), 1U AmpliTaq Gold (Applied Biosystems), 75 mmol/L MgCl2, 0.8 mmol/L dNTPs, 5% DMSO, and 0.7 mmol/L of each primer. Amplification was carried out for 40 cycles, including 60 seconds at 95°C, 40 seconds at 64°C, and 1 minute and 15 seconds at 72°C. Successful PCR was determined by 1.5% agarose gel electrophoresis. PCR products were purified with Illustra ExoProStar (GE Healthcare) and sequenced on a 3730xl DNA analyzer (Applied Biosystems). All samples were analyzed in forward and reverse directions.

Telomere length measurement

Telomere length was determined by multiplex PCR assay as previously described (23). Relative telomere length values were calculated as telomere/single-copy gene (TS) ratio, as previously described (23, 24).

Statistical analysis

The Kaplan–Meier method was used for survival analysis. Survival curves were compared using the log-rank (Mantel–Cox) test. The categorical variables were compared using the χ2 test or Fisher test, and the continuous variables were compared using the Wilcoxon test. The Mann–Whitney U test was used to assess differences in basal biomarker levels between groups of patients grouped according to MMR status and telomeres length in paired GBM samples. Results were classified as statistically significant if their P values were <0.05 All statistical analyses were performed using SPSS version 15.0 statistical software (SPSS Inc.).

Further experimental details are available under Supplementary Methods. An overview of the molecular assays performed with each sample is shown in Supplementary Table S1.

Assessment of MMR status and correlation with hypermutated genome

We initially investigated expression of MLH1, PMS2, MSH2, and MSH6 by IHC in 57 pairs of GBM samples, including tumor samples obtained at diagnosis and relapse. Three samples were not evaluable for technical reasons. Among the n = 54 samples of relapsed GBM analyzed, loss of IHC reactivity for one or more of these markers was found in 14 samples (25.9%); representative examples of IHC patterns are shown in Fig. 1. In 12 of 14 samples, MSH6 reactivity was either completely (n = 4) or partially (n = 8) lost (Table 1; Supplementary Table S3). Concomitant loss of MSH2 expression, which forms heterodimers with MSH6, was found in 3 of these 12 samples lacking MSH6 expression. In the remaining two cases, loss of PMS2 or MLH1 expression was observed. In contrast, GBM samples at diagnosis (n = 56) stained generally positive for these four markers, except for n = 2 samples (3.6%) which had partial loss of MSH2/MSH6 or MLH1/PMS2 expression, in line with the pattern of expression of the same markers in the corresponding relapsed tumor sample (Table 1; Supplementary Table S3).

Table 1.

Evaluation of MMR protein expression (IHC) in paired MMR-deficient GBM samples

Primary GBMRecurrent GBM
MMR statusMLH1PMS2MSH2MSH6MLH1PMS2MSH2MSH6
Positive 12 12 12 12 13 12 11 
Partially negative 
Completely negative 
Not evaluable 
Total 14 14 14 14 14 14 14 14 
Primary GBMRecurrent GBM
MMR statusMLH1PMS2MSH2MSH6MLH1PMS2MSH2MSH6
Positive 12 12 12 12 13 12 11 
Partially negative 
Completely negative 
Not evaluable 
Total 14 14 14 14 14 14 14 14 

We next investigated whether loss of MMR proteins associated with MSI in n = 40 samples by using 5 polyA mononucleotide markers (BAT25, BAT26, NR21, NR24, and NR27). These samples included 12 of 14 samples lacking MMR protein expression at relapse. In no case was MSI detected (Supplementary Table S3), whereas MSI was not analyzed in the remaining samples due to lack of gDNA.

Previous studies in GBM reported that mutations in MSH6 are associated with hypermutated genomes (4, 6). We therefore investigated whether recurrent tumor samples completely lacking MSH6 expression had higher TMB compared with the matched native tumors.

Whole-exome sequencing (WES) was performed on nine matched primary-recurrent GBM: three samples lacking MSH6 expression at relapse, three tumor pairs expressing MMR proteins, and three matched tumors displaying a partial loss of MSH6 protein expression in the relapsed specimen. All samples were successfully analyzed, with 94.1% of aligned bases passing the quality assessment (base call quality of Q20 or higher) and a mean coverage of 110.4X among sequenced samples. On average, 89.4% of coding bases on the targeted exonic regions were covered by at least 30 high-quality reads (see Supplementary Table S4 for more details). In terms of mutational burden, MSH6-negative tumors at relapse (#17/II, #18/II, and #53/II) displayed a “raw” mutational load of about 135-fold higher compared with primary tumor specimens (Supplementary Table S5). This increase concerns both all functional alterations, on average 49 variants at diagnosis compared with 6593 variants at relapse (Fig. 2A), and splicing variants, detected only in recurrent specimens #18/II and #53/II (166 and 614 SNVs, respectively). Interestingly, the MSH6-negative tumor pairs with the lowest TMB (#17/I and II) did not bear any functional mutations in MMR genes, suggesting other mechanisms underlying loss of MSH6 expression, as shown in other studies (25, 26).

Figure 2.

Mutational burden in primary and recurrent GBM tumor samples according to the MMR status. A, Number of functional somatic variants identified in nine primary and recurrent GBM tumor pairs with different MSH6 expression levels (negative #17, #18, and #53; positive #27, #38, and #54; and intermediate #3, #20, and #41). B, Predicted functional readout of the c.3438+1G>A splice-donor variant found in #18/II and #53/II samples on MSH6 protein structure. The splice-donor variant is predicted to cause skipping of the MSH6-exon5 (first plot). The MSH6-exon5 encodes for the C-terminal lever (1,025–1,102) and the nucleotide binding region of the ABC-ATPase domain (1,134–1,141; second plot), which could be affected by the variant (in red on MSH6 structure, 2o8b.pdb).

Figure 2.

Mutational burden in primary and recurrent GBM tumor samples according to the MMR status. A, Number of functional somatic variants identified in nine primary and recurrent GBM tumor pairs with different MSH6 expression levels (negative #17, #18, and #53; positive #27, #38, and #54; and intermediate #3, #20, and #41). B, Predicted functional readout of the c.3438+1G>A splice-donor variant found in #18/II and #53/II samples on MSH6 protein structure. The splice-donor variant is predicted to cause skipping of the MSH6-exon5 (first plot). The MSH6-exon5 encodes for the C-terminal lever (1,025–1,102) and the nucleotide binding region of the ABC-ATPase domain (1,134–1,141; second plot), which could be affected by the variant (in red on MSH6 structure, 2o8b.pdb).

Close modal

Conversely, an identical MSH6 mutation c.3438+1G>A* affecting a donor splicing site and known to cause aberrant splicing was found in the recurrent tumors of the other two pairs (#18/II and #53/II) with a VAF of 30% and 78%, respectively. This exon-skipping variant confirmed by Sanger sequencing in both aforementioned samples (Supplementary Fig. S1) has a potential loss of function role, because it could affect either mRNA stability or the ABC-ATPase domain, which is required for MSH6 MMR activity (Fig. 2B; ref. 27). Further sample-specific MSH6 nonsynonymous variants falling right before or within functional domains of the MSH6 protein were detected in the same samples along with single or multiple mutations in additional MMR genes (i.e., MSH2, MSH4, MSH5, PMS1, MLH1, and MLH3) but with unknown or uncertain significance (Supplementary Table S6). In line with literature data, comparison of initial tumors with their recurrent counterpart confirmed absence of MSH6 as well as MMR gene mutations at diagnosis (6). Noteworthy, the pair with the highest TMB (#53/I–II) harbored a pathogenic variant in the TP53 gene (p.A276D; VAF = 39%), different from the one found in the initial tumor (p.P151S; VAF = 9.5%) and likely contributing to the augmented mutation frequency in this tumor sample. WES of three additional tumor pairs expressing MMR proteins revealed minor variations of the mutational status between primary (#27/I, #38/I, and #54/I) and relapsed (#27/II, #38/II, and #54/II) specimens, with only 3-fold more somatic functional variants (range, 50–116 vs. 9–45) in recurrent tumors versus the matched native ones (Fig. 2A; Supplementary Table S5). Moreover, both the initial tumors and their relapsed counterpart showed absence of variants in MMR genes. Finally, when the mutational load at first surgery was evaluated, no significant difference was detected between MMR-deficient and MMR-proficient samples (range, 14–115 vs. 9–45). Notably, analysis of three tumor samples (#3, #20, and #41) showing partial loss of MSH6 expression by IHC at relapse did not disclose hypermutated genotype (Fig. 2; Supplementary Table S5). In conclusion, we found that complete lack of immunoreactivity for MSH6 in GBM samples is associated with a hypermutated genotype, and it is not necessarily due to MSH6 mutations, underscoring the possibility that IHC might identify a broader fraction of tumor samples with hypermutated genotype, compared with targeted sequencing of MMR genes.

Analysis of MGMT methylation

MGMT methylation was evaluated for all 57 patients studied at both first and second surgeries. At diagnosis, MGMT promoter was found methylated in 32 (56.1%) and unmethylated in 25 (43.9%) samples. At recurrence, MGMT promoter was evaluated in 56 samples and was found methylated in 26 (46.4%) and unmethylated in 30 (53.6%) patients. Methylation extent ranged from 8% to 63% at diagnosis and from 8% to 55% at recurrence. In this longitudinal analysis, MGMT methylation status changed in 9 patients (15.8%), with seven of nine samples becoming unmethylated and two out of nine samples acquiring MGMT methylation at recurrence. Among samples lacking MMR protein expression which could be evaluated for MGMT status at relapse (n = 14), 4 (28.6%) had unmethylated MGMT promoter and 10 (71.4%) had methylated MGMT promoter (Table 2). Moreover, the majority of cases lacking MMR protein expression at relapse (78.5%) had methylated MGMT promoter at diagnosis (Table 2). Statistical analysis discloses a significant association between MGMT methylation both at diagnosis and relapse and MMR status (Table 2). In conclusion, these results indicate a moderate decrease in MGMT methylation at recurrence, which might underscore—along with the much higher fraction of samples lacking MMR protein expression, which is needed to respond to DNA-damaging agents (28)—reduced sensitivity to TMZ or other alkylating drugs.

Table 2.

Association between MMR status and other biomarkers

MMR status at diagnosisMMR status at relapse
MMR deficientMMR proficientTotalOR (95% CI)PMMR deficientMMR proficientTotalOR (95% CI)P
MGMT           
 M 10 18 28 4.48 0.03 10 14 24 4.46 0.02 
 UM 22 25 (1.08–18.52)  25 29 (1.18–16.95)  
 Total 13 40 53   14 39 53   
EGFRvIII           
 − 12 28 40 2.36 0.47 13 27 40 5.29 0.14 
 + 11 13 (0.45–12.29)  11 12 (0.62–45.53)  
 Total 14 39 53   14 38 52   
EGFR FISH           
 NAMP 10 21 31 4.05 0.08 19 27 1.89 0.35 
 AMP 17 19 (0.78–21.02)  18 22 (0.48–7.40)  
 Total 12 38 50   12 37 49   
MMR status at diagnosisMMR status at relapse
MMR deficientMMR proficientTotalOR (95% CI)PMMR deficientMMR proficientTotalOR (95% CI)P
MGMT           
 M 10 18 28 4.48 0.03 10 14 24 4.46 0.02 
 UM 22 25 (1.08–18.52)  25 29 (1.18–16.95)  
 Total 13 40 53   14 39 53   
EGFRvIII           
 − 12 28 40 2.36 0.47 13 27 40 5.29 0.14 
 + 11 13 (0.45–12.29)  11 12 (0.62–45.53)  
 Total 14 39 53   14 38 52   
EGFR FISH           
 NAMP 10 21 31 4.05 0.08 19 27 1.89 0.35 
 AMP 17 19 (0.78–21.02)  18 22 (0.48–7.40)  
 Total 12 38 50   12 37 49   

NOTE: Statistically significant P values (P < 0.05) are reported in italics.

Abbreviations: AMP, EGFR amplified by FISH; M, methylated; NAMP, EGFR not amplified by FISH; UM, unmethylated; +, EGFRvIII-positive sample by IHC; –, EGFRvIII-negative sample by IHC.

IDH1/2 mutation

IDH status was investigated in all GBM samples by Sanger sequencing. Three IDH1-mutated samples (#26, #47, and #54) were found out of 57 samples analyzed (5.2%). No IDH2 mutations were found. IDH1-mutated samples invariably conserved expression of MMR proteins. IDH status did not change between matched first and second surgery samples.

EGFRvIII mutation and EGFR amplification

EGFRvIII mutation was investigated in 56 and 54 GBM samples at diagnosis and recurrence, respectively. As shown in Table 2, we found by IHC 14 (25%) EGFRvIII+ samples at diagnosis and 12 (22%) at relapse. Overall, variations in EGFR status were detected in n = 15 samples, including 8 samples which lost EGFRvIII reactivity at recurrence and 7 tumor samples which acquired EGFRvIII reactivity at recurrence. Among the 14 recurrence samples lacking expression of MMR proteins, only 1 sample (7.1%) was EGFRvIII+. In contrast, among 38 recurrence samples with MMR protein expression, 11 samples expressed EGFRvIII (28.9%). Moreover, the two GBM samples showing partial loss of MMR proteins at diagnosis (#6 and #46) also lacked EGFRvIII expression. However, statistical analysis did not disclose association between MMR status and EGFRvIII (Table 2).

As EGFR mutations have been reported to coexist with gene amplification in GBM, we studied EGFR amplification by FISH. Overall, we detected EGFR amplification in 20 of 52 (38.5%) and 24 of 52 (46.1%) samples analyzed at diagnosis and relapse, respectively (Table 2). In all cases analyzed and in line with published data (29), we observed a high grade of gene amplification displayed as double minutes, and in all cases but one, the pattern of amplification was homogenous in the tumor population (Supplementary Fig. S2). Six cases of 50 (12%) showed a difference in term of EGFR copy number between first and second surgeries. In particular, two cases were EGFR amplified at first surgery but not at recurrence, whereas the opposite trend was observed in four additional cases. Notably, given the known intratumoral heterogeneity of EGFR in GBM (29), it cannot be ruled out that variations in EGFR status between first and second surgeries may be due to undersampling. Considering the overall concordance between EGFRvIII expression and EGFR gene copy number at second surgery, 10 of 11 (90.9%) EGFRvIII+ samples showed EGFR amplification and 26 of 38 (68.4%) EGFRvIII tumors were nonamplified, with kappa coefficient of 0.183, underscoring poor correlation. Finally, statistical analysis disclosed no association between MMR status and EGFR amplification (Table 2).

TERT studies

TERT promoter sequencing analysis was performed in 56 pairs of primary tumor and recurrences. At primary surgery, we identified 41 mutations (73.2%) in the TERT promoter region: 31 (75.6%) 228C>T and 10 (24.4%) 250C>T; these two mutations were mutually exclusive. When comparing primary tumor versus recurrence, 49 cases showed concordance regarding the TERT promoter mutational status, in 4 cases the initial 228C>T TERT promoter mutation was not found at recurrence, and in 3 cases 228C>T TERT promoter mutation was detected at recurrence, but not in the primary tumor (Supplementary Table S7). The lack of detection of TERT promoter mutation at primary tumor, compared with matched recurrence, may depend on the detection limit of Sanger sequencing method when the tumor cells bearing TERT promoter mutation were present at “sub-clonal” level despite the specimen contains more than 70% of tumor tissue. Of interest are the cases of lack of detection of TERT promoter at relapse compared with matched primary tumors; this observation is in agreement with others (30) and likely indicates that the relapse originated from expansion of TERT-unmutated cells (31). Regarding frequencies of rs2853669 SNP, 28 patients (50%) had TT genotype, 3 (5.4%) CC genotype, and 25 (44.6%) were heterozygous. As expected, primary tumor and matched recurrence showed the same rs2853669 genotype.

Measurement of relative telomere length (TS) was obtained from 52 primary tumors and matched recurrences. TS values in primary tumors and recurrences were stratified according to TERT promoter mutational status and SNP genotype and reported in Table 3. Of note, TS significantly shortened in tumors carrying 228C>T mutation [median interquartile range (IQR), 2.01 (1.50–2.93) in primary tumors vs. 1.69 (1.20–2.25) in recurrences, P = 0.016], regardless of the time lapse between first and second surgeries (Table 3).

Table 3.

Relative telomere length in paired GBM samples

PatientsTS primary tumorTS recurrenceP
NMedian (IQR)Median (IQR)Time between diagnosis and relapse (months) Median (IQR)
All 2.07 (1.53–2.75) 1.79 (1.35–2.31) 0.129 15.8 (9.8–26.1) 
n = 52     
TERTp WT 2.30 (1.67–3.24) 2.12 (1.77–3.86) 0.918 19.5 (15.3–30.2) 
n = 13     
TERTp MUT (228+250) 1.95 (1.4–2.77) 1.67 (1.26–2.24) 0.052 15.1 (9.4–26.0) 
n = 39     
TERTp228C>T 2.01 (1.50–2.93) 1.69 (1.20–2.25) 0.016 14.2 (9.2–26.1) 
n = 30     
TERTp250C>T 1.67 (1.23–2.59) 1.56 (1.42–2.16) 0.820 15.3 (10.2–23.5) 
n = 9     
rs2853669 TT 1.85 (1.46–2.86) 1.74 (1.18–2.47) 0.196 17.9 (11.1–28.4) 
n = 25     
rs2853669 TC+CC 2.21 (1.53–2.75) 1.97 (1.49–2.41) 0.380 14.0 (9.4–26.2) 
n = 27     
MMR proficient 1.84 (1.47–2.77) 1.99 (1.35–2.43) 0.729 15.2 (9.7–24.6) 
n = 37     
MMR deficient 2.07 (1.61–2.32) 1.67 (1.07–1.82) 0.048 17.1 (9.0–26.1) 
n = 13     
PatientsTS primary tumorTS recurrenceP
NMedian (IQR)Median (IQR)Time between diagnosis and relapse (months) Median (IQR)
All 2.07 (1.53–2.75) 1.79 (1.35–2.31) 0.129 15.8 (9.8–26.1) 
n = 52     
TERTp WT 2.30 (1.67–3.24) 2.12 (1.77–3.86) 0.918 19.5 (15.3–30.2) 
n = 13     
TERTp MUT (228+250) 1.95 (1.4–2.77) 1.67 (1.26–2.24) 0.052 15.1 (9.4–26.0) 
n = 39     
TERTp228C>T 2.01 (1.50–2.93) 1.69 (1.20–2.25) 0.016 14.2 (9.2–26.1) 
n = 30     
TERTp250C>T 1.67 (1.23–2.59) 1.56 (1.42–2.16) 0.820 15.3 (10.2–23.5) 
n = 9     
rs2853669 TT 1.85 (1.46–2.86) 1.74 (1.18–2.47) 0.196 17.9 (11.1–28.4) 
n = 25     
rs2853669 TC+CC 2.21 (1.53–2.75) 1.97 (1.49–2.41) 0.380 14.0 (9.4–26.2) 
n = 27     
MMR proficient 1.84 (1.47–2.77) 1.99 (1.35–2.43) 0.729 15.2 (9.7–24.6) 
n = 37     
MMR deficient 2.07 (1.61–2.32) 1.67 (1.07–1.82) 0.048 17.1 (9.0–26.1) 
n = 13     

NOTE: P values <0.05 are indicated in bold.

Abbreviations: IQR, interquartile range; MUT (228+250), mutated at 228C>T or 250C>T; n, number; P, P value evaluated by Wilcoxon Signed Rank Test; TERTp, TERT promoter; TS, relative telomere length; WT, wild-type.

A hallmark of genomic instability is the presence of extremely short telomeres (30). Given the interest in genetic changes associated to the MMR defects, we also stratified telomere lengths according to the IHC evaluation of MMR expression (Table 3). Although tumors with conserved MMR protein expression showed no significant difference in TS values between primary tumor and matched recurrence [median (IQR), 1.84 (1.47–2.77) vs. 1.99 (1.35–2.43), respectively, P = 0.729], telomeres shortened significantly between primary tumor and recurrence in MMR-deficient tumors [median (IQR), 2.07 (1.61–2.32) vs. 1.67 (1.07–1.82), respectively, P = 0.048; Table 3]. Notably, this difference did not depend on time lapse within primary tumors and recurrences because this interval was not significantly different between patients stratified according to MMR protein expression [median (IQR), 15.7 (9.7–24.6) months in MMR-proficient vs. 17.1 (9.0–26.1) months in MMR-deficient, P = 0.707; Table 3]. Moreover, tumor purity was similar in the two groups compared (Supplementary Table S1), making it unlikely that differences at this level might affect telomere length evaluation. In conclusion, these results highlight the possibility that telomere shortening could be a feature of the genome instability secondary to MMR defects.

Immunologic correlates

Finally, we sought to investigate whether lack of MMR proteins associated with other immune-related markers in tumors. To this aim, we evaluated by IHC the CD8+ T-cell infiltrate in these tumors. Due to the limited amount of tissue available, this analysis could be performed in only 51 of 57 samples, including all 14 samples with reduced expression of MMR proteins. Results show that there was a trend—albeit not significant—toward increased numbers of CD8+ cells in tumors lacking partially or completely expression of MMR proteins (Supplementary Fig. S3; Table 4). We also observed that PD-L1 and PD-L2 were marginally expressed in GBM, and their expression was independent from MMR status (Table 4; Supplementary Table S8). Finally, staining with an anti-MHC class I antibody disclosed marked heterogeneity in the percentage of MHC class I–positive cells in the samples (range, 0%–95%; Supplementary Table S8) without correlation with MMR status (Table 4).

Table 4.

Immunologic markers expression in paired GBM samples

MarkerMMR at relapseNMeanSDMedianP valuea
CD8 (HPF) Deficient 14 10.07 7.79 7.5 0.09 
 Proficient 38 6.45 5.94  
 Total 52 7.42 6.61  
HLA I (%) Deficient 14 52.86 35.12 65 0.26 
 Proficient 37 36.75 36.61 30  
 Total 51 41.18 36.59 40  
PD-L1 (%) Deficient 14 1.5 4.11 0.71 
 Proficient 37 2.05 7.11  
 Total 51 1.9 6.39  
MarkerMMR at relapseNMeanSDMedianP valuea
CD8 (HPF) Deficient 14 10.07 7.79 7.5 0.09 
 Proficient 38 6.45 5.94  
 Total 52 7.42 6.61  
HLA I (%) Deficient 14 52.86 35.12 65 0.26 
 Proficient 37 36.75 36.61 30  
 Total 51 41.18 36.59 40  
PD-L1 (%) Deficient 14 1.5 4.11 0.71 
 Proficient 37 2.05 7.11  
 Total 51 1.9 6.39  

Abbreviation: HPF, high power field.

aMann–Whitney U.

Clinical correlates

Median progression-free survival (time from surgery to progressive disease assessed by investigator according to Macdonald or RANO criteria) analyzed by Kaplan–Meier curves was 12.8 and 6.1 months from first and second surgeries, respectively. Median overall survival (OS, time from surgery to death or last follow-up) was 32.1 and 11.5 months from first and second surgeries, respectively (Supplementary Fig. S4). At the time of analyses, 12 patients were still alive. OS from relapse was similar in MMR-proficient and MMR-deficient subsets (Supplementary Fig. S5). Among other markers analyzed, only MGMT methylation assessed at diagnosis correlated with improved survival, as expected (Supplementary Fig. S6).

Main conclusions

A summary diagram illustrating the main findings in MMR-deficient GBM samples is shown in Supplementary Fig. S7.

In this study, loss of expression of MSH6 or other MMR proteins was found in about 25% relapsed GBM but only rarely at diagnosis, in line with results of previous genetic studies. These IHC findings are of increasing interest to the medical community, as they could be associated with high TMB, a genetic feature of the tumor which in several tumor types correlates with response to immune-checkpoint inhibitors. In all three MMR−/− tumor samples analyzed, WES disclosed increased TMB, compared with matched samples at diagnosis, or with MMR+/+ samples. Moreover, two of three samples (#18 and #53) shared a known MSH6 mutation which was validated by Sanger sequencing and was predicted to cause loss of function of the protein. In the third sample analyzed (#17), which did not bear MSH6 mutations, loss of MSH6 expression is putatively accounted for by epigenetic mechanisms, as suggested by other studies (25, 26). Notably, sample #17 had a relatively lower TMB at relapse compared with the two samples harboring MSH6 mutations, although clearly higher than TMB of MSH6-proficient samples. Moreover, three samples (#3, #20, and #41) with partial loss of MSH6 expression had TMB similar to MMR-proficient samples. We conclude that complete loss of MMR protein expression correlates with increased TMB in the GBM samples analyzed, in line with recent findings by other studies (31, 32). Due to the relatively low number of samples analyzed by WES (9 of 54 samples), further studies are required, on the one hand, to determine the TMB of samples showing partial loss of MMR proteins and, on the other hand, to conclude whether assessment of MMR protein expression by IHC is an adequate surrogate assay for TMB for patient stratification purposes.

In contrast to the clear-cut increase in mutations detected by WES, we could not demonstrate MSI in any of the MMR-deficient GBM samples analyzed, thus confirming previous reports indicating that MSI is often not present in extraintestinal tumors associated with Lynch syndrome (33). However, MMR-deficient tumors had increased telomere shortening compared with MMR-proficient tumors, which might underscore increased genomic mutations of this tumor subset. This result is in line with previous observations demonstrating that deficiency in MMR is associated with instability in telomeric DNA and increases the rate of telomere shortening (34, 35).

From the clinical viewpoint, although we selected a subgroup of patients having a good prognosis, long OS due to the second surgery, and good performance status at diagnosis, survival data indicated that MMR status (assessed at relapse) was not a prognostic factor in this cohort. Among other markers analyzed, only MGMT promoter methylation was associated with improved OS both at diagnosis and at relapse. MGMT methylation was also associated with MMR status, as MMR-deficient samples were predominantly characterized by methylated MGMT promoter (Table 2). Due to the known requirement of the MMR system for the cytotoxic activity of TMZ (36), it can be hypothesized that MMR inactivation is a second possible target along the same pathway and contributes to make these tumors resistant to TMZ treatment. The heterogeneous treatments received by the patients after relapse along with the relatively small number of MMR-deficient samples did not enable us to further investigate this hypothesis here.

With regard to immunotherapy, many clinical trials are currently evaluating the effects of immune-checkpoint–inhibitory antibodies in GBM, but only a small subset of patients appear to experience a response (8); indeed, in the Checkmate 143 trial, a randomized phase III study evaluating the efficacy of nivolumab, an immune-checkpoint inhibitor, versus bevacizumab, an antiangiogenic drug, in patients with recurrent GBM, there was no statistically significant difference in OS between the two arms: 9.8 months versus 10.0 months of OS in nivolumab and bevacizumab treatment, respectively; noteworthy, 8% of patients who received nivolumab had a response to treatment, and the median duration of response was 11.1 months (37). This may be related to the relatively low mutational rate of gliomas, the general lack of a considerable amount of intratumoral lymphoid cells, and the presumably poor degree of penetration of therapeutic antibodies into the brain (7, 38). Accordingly, we found marginal numbers of CD8+ T cells in GBM samples, and average numbers were comparable in MMR-deficient and MMR-proficient samples (Table 4), thus confirming results by Hodges and colleagues (32). Efforts to identify patients who could be good responders to immune-checkpoint inhibitors should probably take into account both state-of-the-art genetic and immune-related markers. In this regard, our findings indicate that in addition to expression of MMR proteins and assessment of TMB, it should also be considered MHC class I expression, which is required for effective killing of tumor cells by cytotoxic T lymphocytes. We found high heterogeneity in MHC class I expression both within MMR-deficient and MMR-proficient tumors, with values of positivity ranging from 0% to 95%. We speculate that MHC class I–negative tumors could be resistant to immunotherapy, despite bearing alterations in the MMR system. In our study, MHC class I was marginally (0%–30%) expressed by 5 of 14 MMR-deficient GBM samples analyzed. Intriguingly, mutations of the gene encoding β2-microglobulin, a protein required for MHC class I expression and antigen presentation, have been reported associated with acquired resistance to pembrolizumab in patients affected by solid tumors (39, 40). Assessment of MHC class I expression should be considered when developing algorithm to predict response to immune-checkpoint inhibitors in GBM and other solid tumors.

No potential conflicts of interest were disclosed.

Conception and design: S. Indraccolo, G. Lombardi, A. Della Puppa, M.P. Gardiman, V. Zagonel

Development of methodology: A. Pambuku, M.P. Gardiman, V. Zagonel

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): G. Lombardi, M. Fassan, S. Giunco, G.N. Fanelli, A. Pambuku, A. Della Puppa, D. D'Avella, M.P. Gardiman, V. Zagonel

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Indraccolo, M. Fassan, L. Pasqualini, A. Gasparini, S. Nalio, P. Fiduccia, L. Bonaldi, R. Bertorelle, V. Zagonel

Writing, review, and/or revision of the manuscript: S. Indraccolo, G. Lombardi, M. Fassan, L. Pasqualini, S. Giunco, A. Gasparini, A. Pambuku, M.P. Gardiman, A. De Rossi, V. Zagonel

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): G. Lombardi, L. Pasqualini, A. Pambuku

Study supervision: S. Indraccolo, A. Della Puppa, A. De Rossi, V. Zagonel

Other (analysis of MGMT and IDH1/2): R. Marcato

Other (analysis of MSI): C. Candiotto

Other (FISH analysis and interpretation of results): L. Bonaldi

The authors are grateful to Christina Drace for the English revision of the article and to Vincenza Guzzardo, Martina Verza, Valeria Sacchetto, and Eleonora Bergo for technical assistance.

This work was supported by grants from Fondazione Giovanni Celeghin (to V. Zagonel) and IOV 5 × 1000 Intramural Research Grant–Genomica dei Tumori (to G. Lombardi).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Diamandis
P
,
Aldape
KD
. 
Insights from molecular profiling of adult glioma
.
J Clin Oncol
2017
;
35
:
2386
93
.
2.
Qazi
MA
,
Vora
P
,
Venugopal
C
,
Sidhu
SS
,
Moffat
J
,
Swanton
C
, et al
Intratumoral heterogeneity: pathways to treatment resistance and relapse in human glioblastoma
.
Ann Oncol
2017
;
28
:
1448
56
.
3.
Touat
M
,
Idbaih
A
,
Sanson
M
,
Ligon
KL
. 
Glioblastoma targeted therapy: updated approaches from recent biological insights
.
Ann Oncol
2017
;
28
:
1457
72
.
4.
Hunter
C
,
Smith
R
,
Cahill
DP
,
Stephens
P
,
Stevens
C
,
Teague
J
, et al
A hypermutation phenotype and somatic MSH6 mutations in recurrent human malignant gliomas after alkylator chemotherapy
.
Cancer Res
2006
;
66
:
3987
91
.
5.
Kim
J
,
Lee
IH
,
Cho
HJ
,
Park
CK
,
Jung
YS
,
Kim
Y
, et al
Spatiotemporal evolution of the primary glioblastoma genome
.
Cancer Cell
2015
;
28
:
318
28
.
6.
Wang
J
,
Cazzato
E
,
Ladewig
E
,
Frattini
V
,
Rosenbloom
DI
,
Zairis
S
, et al
Clonal evolution of glioblastoma under therapy
.
Nat Genet
2016
;
48
:
768
76
.
7.
Sampson
JH
,
Maus
MV
,
June
CH
. 
Immunotherapy for brain tumors
.
J Clin Oncol
2017
;
35
:
2450
6
.
8.
Bouffet
E
,
Larouche
V
,
Campbell
BB
,
Merico
D
,
de Borja
R
,
Aronson
M
, et al
Immune checkpoint inhibition for hypermutant glioblastoma multiforme resulting from germline biallelic mismatch repair deficiency
.
J Clin Oncol
2016
;
34
:
2206
11
.
9.
Cahill
DP
,
Levine
KK
,
Betensky
RA
,
Codd
PJ
,
Romany
CA
,
Reavie
LB
, et al
Loss of the mismatch repair protein MSH6 in human glioblastomas is associated with tumor progression during temozolomide treatment
.
Clin Cancer Res
2007
;
13
:
2038
45
.
10.
Yip
S
,
Miao
J
,
Cahill
DP
,
Iafrate
AJ
,
Aldape
K
,
Nutt
CL
, et al
MSH6 mutations arise in glioblastomas during temozolomide therapy and mediate temozolomide resistance
.
Clin Cancer Res
2009
;
15
:
4622
9
.
11.
Johannesma
PC
,
van der Klift
HM
,
van Grieken
NC
,
Troost
D
,
Te Riele
H
,
Jacobs
MA
, et al
Childhood brain tumours due to germline bi-allelic mismatch repair gene mutations
.
Clin Genet
2011
;
80
:
243
55
.
12.
Remo
A
,
Fassan
M
,
Lanza
G
. 
Immunohistochemical evaluation of mismatch repair proteins in colorectal carcinoma: the AIFEG/GIPAD proposal
.
Pathologica
2016
;
108
:
104
9
.
13.
Saraggi
D
,
Galuppini
F
,
Remo
A
,
Urso
EDL
,
Bacchin
D
,
Salmaso
R
, et al
PD-L1 overexpression in ampulla of Vater carcinoma and its pre-invasive lesions
.
Histopathology
2017
;
71
:
470
4
.
14.
Li
H
,
Durbin
R
. 
Fast and accurate short read alignment with Burrows-Wheeler transform
.
Bioinformatics
2009
;
25
:
1754
60
.
15.
Fuentes Fajardo
KV
,
Adams
D
,
Mason
CE
,
Sincan
M
,
Tifft
C
,
Toro
C
, et al
Detecting false-positive signals in exome sequencing
.
Hum Mutat
2012
;
33
:
609
13
.
16.
McKenna
A
,
Hanna
M
,
Banks
E
,
Sivachenko
A
,
Cibulskis
K
,
Kernytsky
A
, et al
The Genome Analysis Toolkit: a MapReduce framework for analyzing next-generation DNA sequencing data
.
Genome Res
2010
;
20
:
1297
303
.
17.
Wang
K
,
Li
M
,
Hakonarson
H
. 
ANNOVAR: functional annotation of genetic variants from high-throughput sequencing data
.
Nucleic Acids Res
2010
;
38
:
e164
.
18.
Cibulskis
K
,
Lawrence
MS
,
Carter
SL
,
Sivachenko
A
,
Jaffe
D
,
Sougnez
C
, et al
Sensitive detection of somatic point mutations in impure and heterogeneous cancer samples
.
Nat Biotechnol
2013
;
31
:
213
9
.
19.
Kjallquist
U
,
Erlandsson
R
,
Tobin
NP
,
Alkodsi
A
,
Ullah
I
,
Stalhammar
G
, et al
Exome sequencing of primary breast cancers with paired metastatic lesions reveals metastasis-enriched mutations in the A-kinase anchoring protein family (AKAPs)
.
BMC Cancer
2018
;
18
:
174
.
20.
Frattini
V
,
Trifonov
V
,
Chan
JM
,
Castano
A
,
Lia
M
,
Abate
F
, et al
The integrated landscape of driver genomic alterations in glioblastoma
.
Nat Genet
2013
;
45
:
1141
9
.
21.
Goel
A
,
Nagasaka
T
,
Hamelin
R
,
Boland
CR
. 
An optimized pentaplex PCR for detecting DNA mismatch repair-deficient colorectal cancers
.
PLoS One
2010
;
5
:
e9393
.
22.
You
JF
,
Buhard
O
,
Ligtenberg
MJ
,
Kets
CM
,
Niessen
RC
,
Hofstra
RM
, et al
Tumours with loss of MSH6 expression are MSI-H when screened with a pentaplex of five mononucleotide repeats
.
Br J Cancer
2010
;
103
:
1840
5
.
23.
Gianesin
K
,
Noguera-Julian
A
,
Zanchetta
M
,
Del Bianco
P
,
Petrara
MR
,
Freguja
R
, et al
Premature aging and immune senescence in HIV-infected children
.
AIDS
2016
;
30
:
1363
73
.
24.
Rampazzo
E
,
Bertorelle
R
,
Serra
L
,
Terrin
L
,
Candiotto
C
,
Pucciarelli
S
, et al
Relationship between telomere shortening, genetic instability, and site of tumour origin in colorectal cancers
.
Br J Cancer
2010
;
102
:
1300
5
.
25.
Edelbrock
MA
,
Kaliyaperumal
S
,
Williams
KJ
. 
Structural, molecular and cellular functions of MSH2 and MSH6 during DNA mismatch repair, damage signaling and other noncanonical activities
.
Mutat Res
2013
;
743–744
:
53
66
.
26.
Gazzoli
I
,
Kolodner
RD
. 
Regulation of the human MSH6 gene by the Sp1 transcription factor and alteration of promoter activity and expression by polymorphisms
.
Mol Cell Biol
2003
;
23
:
7992
8007
.
27.
Warren
JJ
,
Pohlhaus
TJ
,
Changela
A
,
Iyer
RR
,
Modrich
PL
,
Beese
LS
. 
Structure of the human MutSalpha DNA lesion recognition complex
.
Mol Cell
2007
;
26
:
579
92
.
28.
Lynch
HT
,
Snyder
CL
,
Shaw
TG
,
Heinen
CD
,
Hitchins
MP
. 
Milestones of Lynch syndrome: 1895–2015
.
Nat Rev Cancer
2015
;
15
:
181
94
.
29.
Eskilsson
E
,
Rosland
GV
,
Solecki
G
,
Wang
Q
,
Harter
PN
,
Graziani
G
, et al
EGFR heterogeneity and implications for therapeutic intervention in glioblastoma
.
Neuro Oncol
2018
;
20
:
743
52
.
30.
Heidenreich
B
,
Rachakonda
PS
,
Hosen
I
,
Volz
F
,
Hemminki
K
,
Weyerbrock
A
, et al
TERT promoter mutations and telomere length in adult malignant gliomas and recurrences
.
Oncotarget
2015
;
6
:
10617
33
.
31.
Johnson
BE
,
Mazor
T
,
Hong
C
,
Barnes
M
,
Aihara
K
,
McLean
CY
, et al
Mutational analysis reveals the origin and therapy-driven evolution of recurrent glioma
.
Science
2014
;
343
:
189
93
.
32.
Hodges
TR
,
Ott
M
,
Xiu
J
,
Gatalica
Z
,
Swensen
J
,
Zhou
S
, et al
Mutational burden, immune checkpoint expression, and mismatch repair in glioma: implications for immune checkpoint immunotherapy
.
Neuro Oncol
2017
;
19
:
1047
57
.
33.
Gylling
AH
,
Nieminen
TT
,
Abdel-Rahman
WM
,
Nuorva
K
,
Juhola
M
,
Joensuu
EI
, et al
Differential cancer predisposition in Lynch syndrome: insights from molecular analysis of brain and urinary tract tumors
.
Carcinogenesis
2008
;
29
:
1351
9
.
34.
Mendez-Bermudez
A
,
Royle
NJ
. 
Deficiency in DNA mismatch repair increases the rate of telomere shortening in normal human cells
.
Hum Mutat
2011
;
32
:
939
46
.
35.
Pickett
HA
,
Baird
DM
,
Hoff-Olsen
P
,
Meling
GI
,
Rognum
TO
,
Shaw
J
, et al
Telomere instability detected in sporadic colon cancers, some showing mutations in a mismatch repair gene
.
Oncogene
2004
;
23
:
3434
43
.
36.
Thomas
A
,
Tanaka
M
,
Trepel
J
,
Reinhold
WC
,
Rajapakse
VN
,
Pommier
Y
. 
Temozolomide in the era of precision medicine
.
Cancer Res
2017
;
77
:
823
6
.
37.
Reardon
DA
,
Omuro
A
,
Brandes
AA
,
Rieger
J
,
Wick
A
,
Sepulveda
J
, et al
Randomized phase 3 study evaluating the efficacy and safety of nivolumab vs bevacizumab in patients with recurrent glioblastoma: checkmate 143
.
Neuro-Oncology
2017
;
19
(
suppl_3
):
iii21
.
38.
Preusser
M
,
Lim
M
,
Hafler
DA
,
Reardon
DA
,
Sampson
JH
. 
Prospects of immune checkpoint modulators in the treatment of glioblastoma
.
Nat Rev Neurol
2015
;
11
:
504
14
.
39.
Le
DT
,
Durham
JN
,
Smith
KN
,
Wang
H
,
Bartlett
BR
,
Aulakh
LK
, et al
Mismatch repair deficiency predicts response of solid tumors to PD-1 blockade
.
Science
2017
;
357
:
409
13
.
40.
Zaretsky
JM
,
Garcia-Diaz
A
,
Shin
DS
,
Escuin-Ordinas
H
,
Hugo
W
,
Hu-Lieskovan
S
, et al
Mutations associated with acquired resistance to PD-1 blockade in melanoma
.
N Engl J Med
2016
;
375
:
819
29
.