Telomere alterations in neurofibromatosis type 1-associated solid tumors

A total of 426 tumor samples from 256 NF1 patients were studied. NF1 status was based on established clinical consensus criteria [27], with various features abstracted from the clinical records. In addition, a set of 99 MPNST samples (18 non-syndrome associated, 2 schwannomatosis-associated, and 79 of unknown NF1 status) from 81 patients were also studied. All histologic sections were reviewed by a board certified neuropathologist (FJR) and classified to the extent feasible using criteria embodied in the 2016 WHO Classification of Tumors of the Central Nervous System [23] and recently proposed criteria for nerve sheath tumors [26, 30]. Tumor samples included whole sections (N = 364) or sections obtained from 3 different TMAs (N = 161) consisting of NF1-associated gliomas and MPNSTs that have been previously characterized [24, 32, 36]. All ethical standards were followed and the study was performed under IRB approval, with consent or waiver of consent approval.

ALT status was performed and interpreted using previously published criteria and was characterized by the presence of distinct large telomeric DNA signals using fluorescence in situ hybridization (FISH) [12, 13, 39] or chromogenic in situ hybridization (CISH) [39] assays. In brief, tissue slides were deparaffinized, hydrated, and steamed for 25 min in citrate buffer (Vector Laboratories). This was followed by dehydration and hybridization with a Cy3-labeled peptide nucleic acid (PNA) probe complementary to the mammalian telomere repeat sequence [(N-terminus to C-terminus) CCCTAACCCTAACCCTAA]. In the FISH approach, an Alexa Fluor-488–labeled PNA probe specific to human centromeric DNA repeats was also included as a control to assess the validity of the hybridization. Following post-hybridization washes, the slides were counterstained with DAPI following post-hybridization washes. The telomere-specific CISH was previously outlined. Briefly, deparaffinized slides were hydrated, steamed for 25 min in citrate buffer, dehydrated, and hybridized with a Cy3-labeled PNA probe (see above). Sections were blocked against endogenous peroxidase activity with Dual Endogenous Enzyme-Blocking Agent (Dako) for 10 min, incubated with a monoclonal anti-Cy3/Cy5 antibody (ab52060, Abcam, 1:2500) for 1 h at room temperature, then incubated with an anti-mouse secondary antibody (Leica Microsystems) for 30 min, and detected with 3,30-diaminobenzidine (Sigma-Aldrich) after 10 min. Finally, sections were counterstained with hematoxylin, rehydrated, and mounted for visual inspection. In the subset of ALT-negative cases with appropriate hybridization signals in the internal non-neoplastic components, telomere lengths were qualitatively scored by direct visual assessment of stained slides comparing telomere signals from tumor cells with telomere signals from benign cells (entrapped non-neoplastic cells) from the same case. In all cases, signals from benign cells were considered 2+ (normal). Telomere signals in tumor cells of different cases ranged from short (1+) to normal (2+) to long (3+).

Immunohistochemical studies that were previously performed as part of the clinical diagnostic workup were reviewed. However, in most cases immunohistochemical studies were systematically performed using antibodies recognizing either ATRX, DAXX, or H3K27me3.

Immunohistochemical studies were systematically performed in most cases using an ATRX antibody (Rabbit polyclonal, 1:200 dilution, catalog# HPA001906 Sigma-Aldrich). Immunostaining was performed on automated instruments (BenchMark, Ventana Medical Systems, Tucson, AZ, USA). The immunohistochemical protocol included deparaffinization, hydration, antigen retrieval, primary antibody incubation, and detection and visualization as per manufacturer’s instructions. Immunohistochemistry for DAXX (Rabbit polyclonal, 1:100 dilution, catalog# HPA008736, Atlas Antibodies) was performed manually. Sections were incubated with primary antibody for 2 h at room temperature followed by secondary antibody (Leica Microsystems) for 30 min and detected with 3,30-diaminobenzidine (Sigma-Aldrich) after 10 min. Immunohistochemistry for H3K27me3 was performed manually in whole tissue sections using a mouse monoclonal antibody (Abcam, Cat# ab6002, Cambridge, MA) at 1:1600 dilution an incubated overnight at 4 °C. The subset of MPNST cases studied in a TMA were tested for H3K27me3 status as part of a previous study [32]. Immunoreactivity preservation in internal non-neoplastic cell components was required for valid interpretation in all cases to evaluate ATRX, DAXX, or H3K27me3 loss. In cases where the loss was clearly limited to a subset of neoplastic cells, with preserved reactivity in non-neoplastic components, the loss was interpreted as “partial”.

To identify possible somatic genetic alterations associated with ALT, we performed next generation sequencing on the 9 ALT-positive tumors with available tissue (6 gliomas and 3 MPNST). Adjacent formalin-fixed, paraffin-embedded sections cut at 5-μm thick were scrapped into 2 ml tubes for DNA extraction, focusing on areas enriched for neoplastic cells. An automated Siemens Tissue Preparation System (Siemens Healthcare Diagnostics, Inc., Tarrytown, NY) was used for DNA extraction and purification. Quantification of Genomic DNA was performed using the Qubit 2.0 Fluorometer (Life Technologies, Carlsbad, CA). DNA libraries were prepared using Agilent SureSelect-XT reagents (Agilent Technologies, Inc., Santa Clara, CA). Genomic regions of interest were captured using an Agilent custom-designed bait set covering the full coding regions of 644 cancer associated genes, although exon 1 is poorly covered for some genes. Genes of interest included NF1, SUZ12, EED, ATRX, DAXX, and BRAF. Sequencing of libraries was achieved to an average unique read depth of greater than 500X using Sequencing by Synthesis (SBS) 2 × 100 base pairs (bp) paired-end cluster generation on the Illumina HiSeq 2500 plataform (Illumina, Inc., San Diego, CA). FASTQ files were generated from Binary Cluster Files (.bcl) using the Illumina bcl2fastq v1.8.4 software with parameters set as per vendor’s specifications. FASTQ files were aligned to the human genome reference hg19 (GRCh37) using the Burrows-Wheeler Aligner v0.7.10 algorithm with default settings. BAM files (.bam) were generated using Picard Tools v1.119 and variant calling was performed using in-house variant caller algorithm (MDLVC v5.0) cross referenced with HaplotypeCaller (Genome Analysis Tool Kit 3.3) under discovery mode in the coding regions of target genes. All variant calls were inspected using Integrated Genomics Viewer v2.3.4 (IGV; Broad Institute, MIT Harvard, Cambridge, MA) and annotated with dbSNP v150 and COSMIC v82 databases. Common single nucleotide polymorphisms (SNPs; population allele frequency > 50%) were excluded from analysis. Matched normal (tissue or peripheral blood) was not available for comparison. Mean target coverage of the full coding regions (including exon 1) of ATRX, DAXX, and NF1 was 688 reads (330 to 930 average base read depth), 811 reads (406 to 1129 average base read depth), and 823 reads (423 to 1089 average base read depth), respectively, across all cases.

All characteristics were described using proportions, ranges, means, medians, and standard deviations as appropriate. Proportions were compared using Chi-Square or Fisher’s exact tests as appropriate. Survival rates were visualized using Kaplan–Meier curves and separations of curves were assessed using the log-rank test. Overall survival was calculated from the time of first pathologic diagnosis to death. Recurrence-free survival was calculated from the time of first pathologic diagnosis to the first evidence of tumor growth per clinical record, or time of death. In further analyses in the glioma and MPNST cohorts, Cox-proportional hazard models were used to assess the independent effects of either telomere length or ALT on overall survival while adjusting for age and grade. Statistical analyses were performed using GraphPad Prism version 8.0 (San Diego, CA), SAS version 9.4 (Cary, NC), and R version 3.3.2.

A total of 78 CNS tumors from 70 patients with NF1 were analyzed, including 70 primary biopsies or resections and 8 surgeries/recurrences. Clinicopathologic and molecular features are outlined for all CNS tumors (Additional file 1: Table S1) and summarized for the gliomas in Table 1. In the glioma cohort, ALT was present in 23 (of 70; 32.9%) tumors. When comparing groups, ALT was present in the majority of high-grade gliomas 14 (of 23; 60%) compared to 9 (of 47; 19%) low-grade gliomas, a difference that was statistically significant (p = 0.0009, χ² test). When focusing on different pathologic subtypes, ALT was present in only 2 (of 26, 8%) NF1-associated PA compared to 21 (of 44, 47%) other NF1-CNS (non-PA) tumors (p = 0.001, Fisher Exact Test). When assessing additional tumor samples/recurrences, there was a perfect concordance among multiple resections from the same patient, with ALT present in 3 and absent in 5 pathology samples from the same patient. Among 23 cases with ALT, ATRX loss was present in 17 (74%) (14 complete, 3 partial) and preserved in 6 (26%). In contrast, DAXX was preserved in 11 (of 11) cases. Finally, among the ALT-negative cases, 3 (of 42; 7%) cases tested demonstrated partial ATRX loss, while DAXX was preserved in 14 (of 14) cases.

Next, telomere lengths were evaluated in the ALT-negative tumors, which was possible in 30 cases. Representative images of telomere lengths in NF1-associated gliomas are illustrated (Fig. 1a-c). Telomere lengths were normal in 17 (57%) cases and abnormally long in 12 (40%). Thus, ALT was a feature of predominantly high-grade or non-PA NF1-associated gliomas, while abnormally long telomeres (in the absence of ALT) are present in a subset of low-grade astrocytomas.

Next, we studied 197 samples obtained from 174 MPNSTs from 152 patients (93 NF1-associated, 16 non-syndrome-associated, 1 Schwannomatosis-associated, and 42 with unknown syndrome status given incomplete clinical history). Clinicopathologic and molecular features are outlined for all nerve sheath tumors (Additional file 1: Table S1) and summarized for the MPNST and neurofibroma cohorts in Table 1. ALT was present in 25 (of 148; 17%) MPNSTs from unique patients. In contrast, ALT was present in only 2 (of 193) neurofibromas obtained from 77 NF1 patients, a statistically significant difference (p < 0.0001, Fisher Exact Test). Notably, the 2 ALT-positive neurofibromas were obtained from a single patient (case 78) who underwent malignant transformation into an ALT-positive MPNST (Additional file 1: Figure S1). Tumors from this patient demonstrated partial ATRX and DAXX protein loss but lacked definite ATRX and DAXX mutations (Additional file 1: Table S1 and Table S2).

We assessed telomere lengths in a subset of ALT-negative MPNST cases (n = 30), identifying 14 (47%) cases with short, 15 (50%) cases with normal, and 1 (3%) case with long telomere lengths. Representative images of telomere lengths in MPNST are illustrated in Fig. 1d-f. When limiting to only the NF1-associated group, ALT was present in 9 (12%), while telomere lengths were abnormally short in 9 (of 24 ALT-negative cases; 38%), normal in 14 (58%), and long in a single case (4%). When limiting to only the sporadic group, ALT was present in 2 (of 14; 14%) MPNSTs, and telomere lengths were all short in the 5 evaluable ALT-negative cases. A total of 183 (of 191) ALT-negative neurofibromas were evaluable for telomere lengths and all cases displayed normal telomere lengths.

Within the ALT-positive MPNSTs, 6 (of 10) from distinct patients had ATRX protein loss, which was complete in 2 and partial in 4. Partial DAXX loss was present in 1 (of 10) cases. In this group, all 10 cases showed retained H3K27 trimethylation. In addition, one MPNST case displayed ATRX protein loss and was ALT-positive in the primary, and also displayed ATRX protein loss and was ALT-positive in 3 additional distant metastases, suggesting that ATRX loss and ALT are maintained during tumor progression. When assessing the ALT-negative cases, 11 (of 53; 21%) had partial ATRX loss and 3 (of 32; 9%) had partial DAXX loss. These findings collectively suggest that abnormal telomere lengths are frequent in MPNST, in contrast to their benign neurofibroma counterparts, and are characterized by the presence of ALT or abnormally shortened telomeres.

Next, we evaluated other rare NF1-associated solid tumors (n = 46). A single malignant phyllodes tumor of the breast demonstrated ALT; however, retained ATRX protein expression. In contrast, none of the other rare NF1-associated tumors demonstrated ALT (Additional file 1: Table S3), including gastrointestinal stromal tumors (GIST), pheochromocytomas, glomus tumors, juvenile xanthogranulomas, and duodenal neuroendocrine tumors.

To study genetic alterations associated with the development of ALT, we performed NGS in 4 MPNSTs (3 ALT-positive and 1 ALT-negative) and 6 ALT-positive gliomas, 4 of which were also analyzed as part of a prior study [31]. Sequencing results of ALT-positive cases are provided by case in Additional file 1: Table S2. In the ALT-positive glioma group, all 6 had pathogenic NF1 mutations, with a high VAF suggestive of LOH (n = 4) or a second mutation (n = 1). Three (of 6) had ATRX mutations, 2 had homozygous CDKN2A/B deletions, but none had a DAXX mutation.

All 3 ALT-positive MPNST had pathogenic NF1 mutations and lacked EED, SUZ12, or DAXX mutations. Case 110 had an ATRX p.E511Kfs*3 mutation with ATRX protein loss (Fig. 2). Case 78 had an ATRX variant (p.Q929E) with a 67% VAF and partial ATRX protein loss; whereas case 53 lacked an ATRX mutation, but contained two RECQL4 variants (Additional file 1: Table S2). RECQL4 is a DNA helicase and alterations in genes associated with DNA repair could lead to ALT, in the absence of the better known alterations in the chromatin remodelers ATRX and DAXX. RECQL4 variants were also present in two ALT-positive, ATRX intact NF1-gliomas. The ALT-negative MPNST with sequencing data displayed NF1 and SUZ12 mutations. Next, we analyzed publicly available gene sequencing data from cBioPortal [6, 11] and identified missense DAXX mutations in 2 (of 15) MPNSTs, both NF1-associated. These tumors lacked PRC2 (EED, SUZ12) mutations, and no mutations were present in RBL2 or SP100, two altered telomere maintenance genes found to be enriched in ALT-positive leiomyosarcomas [7]. Collectively, these findings suggest that in a subset of NF1-associated tumors, ALT may be mediated by genes other than ATRX and DAXX.

In the NF1-glioma cohort, median overall survival for patients with an ALT-positive tumor was only 18 months and significantly differed compared to patients with either long (69 months) or normal telomeres (not reached) (p < 0.0001, log rank test). When restricting the analysis to cases with evaluable telomere lengths, there was also a trend for intermediate survival for patients with long telomeres and better survival for patients with normal telomeres, although the differences in survival appeared to be driven mostly by the ALT-positive group (p = 0.0021, log rank test) (Fig. 3a, b). When modeling in a Cox univariate model, we observed consistent patterns for altered telomeres lengths (Long: HR = 3.99, 95% CI = 0.70–22.77; ALT HR = 13.05, 95% CI = 2.35–72.43) when compared to normal telomere lengths (Additional file 1: Table S4).

Since age and grade are some of the most important prognostic factors for central nervous system tumors, and ALT was associated with adults and higher grade tumors in our prior study [39], we performed multivariate analyses adjusting for these two factors. While there was not a statistically significant difference by telomere length group (Additional file 1: Table S4), ALT status remained significantly associated with a worse outcome in the model (HR = 3.62, 95% CI 1.17–11.17; p = 0.03) (Table 2).

In the NF1-MPNST cohort, overall survival was decreased for patients with ALT-positive tumors, intermediate for those with normal telomeres, and increased for those with short telomeres (p = 0.0052). When evaluating the entire MPNST cohort (including sporadic tumors and tumors with unknown NF1-status), median survival for patients was 28 months for ALT-positive tumors, 31 months for ALT-negative tumors, 32 months for normal telomeres, and not reached for short telomeres (Additional file 1: Figure S2a,b). These survival differences were similar in the NF1-MPNST group (p = 0.0016) (Fig. 3c,d).

In contrast to the glioma cohort, there were no statistical significant differences in overall survival between ALT-positive and ALT-negative NF1-MPNSTs (p = 0.20). Similarly, using a Cox univariate model, the altered telomere length groups did not differ (Short HR = 0.13, 95% CI = 0.02–1.12; ALT HR = 1.93, 95% CI = 0.68–5.49) when compared to the normal telomere group. However, in a multivariate model with age and grade, the short telomere group displayed a significantly better overall survival difference (HR = 0.03, 95% CI = 0.01–0.29; p = 0.003) compared to the normal telomere group (Table 3). When considering only ALT status, there were no significant differences with ALT-positivity in the NF1-MPNST cohort (HR = 1.21, 95% CI = 0.41–3.59; p = 0.70) (Table 3) or in the entire MPNST group (including sporadic and unknown status) (Additional file 1: Table S5 and Table S6).