Pain correlates with germline mutation in schwannomatosis.

Schwannomatosis has been linked to germline mutations in the SMARCB1 and LZTR1 genes, and is frequently associated with pain.In a cohort study, we assessed the mutation status of 37 patients with clinically diagnosed schwannomatosis and compared to clinical data, whole body MRI (WBMRI), visual analog pain scale, and Short Form 36 (SF-36) bodily pain subscale.We identified a germline mutation in LZTR1 in 5 patients (13.5%) and SMARCB1 in 15 patients (40.5%), but found no germline mutation in 17 patients (45.9%). Peripheral schwannomas were detected in 3 LZTR1-mutant (60%) and 10 SMARCB1-mutant subjects (66.7%). Among those with peripheral tumors, the median tumor number was 4 in the LZTR1 group (median total body tumor volume 30 cc) and 10 in the SMARCB1 group (median volume 85cc), (P=.2915 for tumor number and P = .2289 for volume). mutation was associated with an increased prevalence of spinal schwannomas (100% vs 41%, P = .0197). The median pain score was 3.9/10 in the LZTR1 group and 0.5/10 in the SMARCB1 group (P = .0414), and SF-36 pain-associated quality of life was significantly worse in the LZTR1 group (P = .0106). Pain scores correlated with total body tumor volume (rho = 0.32471, P = .0499), but not with number of tumors (rho = 0.23065, P = .1696).We found no significant difference in quantitative tumor burden between mutational groups, but spinal schwannomas were more common in LZTR1-mutant patients. Pain was significantly higher in LZTR1-mutant than in SMARCB1-mutant patients, though spinal tumor location did not significantly correlate with pain. This suggests a possible genetic association with schwannomatosis-associated pain.

Introduction

Schwannomatosis is a hereditary syndrome that predisposes affected individuals to develop nonintradermal schwannomas and, less commonly, meningiomas.[ While there is phenotypic overlap with neurofibromatosis 2 (NF2), early genetic studies of schwannomatosis ruled out mutations in the NF2 locus as a cause.[ To date, 2 genes have been associated with schwannomatosis, SMARCB1 and LZTR1, making schwannomatosis the only form of neurofibromatosis with locus heterogeneity.[SMARCB1 mutations are found more frequently in familial (45%) than sporadic cases (7%) of schwannomatosis,[ while LZTR1 mutations are found at similar frequency between the 2 (21% vs 20.5%).[

Schwannomatosis is notable for several unique characteristics when compared to NF2, including a much lower prevalence of vestibular schwannomas and meningiomas, a relatively higher prevalence of peripheral schwannomas, and a high prevalence of chronic, severe pain. The pain of schwannomatosis has been associated with tumor burden,[ but not location, in previous studies, raising the question of a genetic contribution to an underlying pain syndrome.

The current theory of tumor development among individuals with either SMARCB1- or LZTR1-associated schwannomatosis involves a 3-step, 4-hit mechanism. A germline mutation in either SMARCB1 or LZTR1 is present on 1 allele, followed by a somatic loss of heterozygosity event on chromosome 22q involving both the NF2 gene and either the SMARCB1 or LZTR1 gene, and then a mutation in the NF2 gene on the retained allele.[ This series of mutational events demonstrates not only the importance of NF2 loss in schwannoma development, but also the genetic heterogeneity among tumors even in the same individual. In addition, this model raises the possibility that the mechanism of pain generation may be distinct from that of tumorigenesis, with pain related to the germline SMARCB1 or LZTR1 mutation, while tumor formation is caused by biallelic loss of NF2.

The paucity of genotype–phenotype correlations identified to date for schwannomatosis patients prompted us to investigate with a novel combination of tools. We used a custom, high-efficiency targeted gene capture system for germline sequencing of relevant genes. To that data, we compared key features of schwannomatosis phenotype, including internal tumor number and size using whole body MRI (WBMRI) and regional MRI, as well as validated measures of patient-reported pain and pain-related quality of life (QOL).

Materials and methods

Cohort determination

The patient population for this study was derived from a pool of 282 individuals who participated in an IRB-approved study of WBMRI.[ Analyzed patients had a clinical diagnosis of schwannomatosis as defined by prior publication, with 2 or more nonintradermal schwannomas, at least 1 with histological confirmation, and no evidence of vestibular tumor on high-quality MRI scan, no known constitutional NF2 mutation, and no first-degree relative with NF2; alternatively, patients could meet clinical diagnostic criteria by having 1 pathologically confirmed nonvestibular schwannoma or intracranial meningioma plus a first-degree relative with schwannomatosis.[ Participants with WBMRI, seen at Massachusetts General Hospital with clinically diagnosed schwannomatosis, with germline DNA available, and who had previously consented to germline genetic testing, were included. Among the 37 meeting these criteria, 7 individuals had known germline mutations determined previously by Sanger sequencing. The 30 remaining individuals underwent capture (described below), followed by targeted next generation sequencing (NGS) of the entire gene regions for NF2, SMARCB1, and LZTR1.

Whole-body MRI and clinical characterization

As a component of our previously published study, patients over the age of 18 years with clinically diagnosed schwannomatosis underwent WBMRI.[ Phenotypic data collected at enrollment included neurologic symptoms and evidence of neuraxis tumors identified on previous regional imaging. All included patients consented to blood and tumor genetic testing, and all such data were linked to WBMRI and clinical data but otherwise rendered de-identified.

Tumor volumetrics

Tumor burden was determined as previously described.[ A board-certified radiologist without knowledge of genotype reviewed each scan to identify tumors. Tumors were segmented using computerized volumetry, and whole-body tumor volume was assessed.[

Pain measurement

Pain phenotype was established using self-reported measures of pain intensity and pain-related QOL. Pain intensity was assessed using the visual analog scale (VAS), in which the patient places a vertical mark on a 10 cm horizontal line drawn between “no pain” on the left and “worst pain ever” on the right, to indicate their current level of pain. The distance of the marks from the left side of the scale were measured to the nearest tenth of a centimeter to obtain the final numerical pain score. Quality of life implications of pain were measured with the Short Form 36 (SF-36), version 1, bodily pain subscale. The SF-36 is a short, patient-completed questionnaire that explores various aspects of QOL over the past month, with higher scores reflecting better QOL and lower scores reflecting worse QOL. Norm-based scores were calculated for each patient using the QualityMetric Health Outcomes Scoring Software 2.0,[ which transforms raw scores into a norm-based score with expected mean of 50 and SD of 10 using data from the general US population.

DNA extraction

As a component of our previously published work,[ individuals consented to germline genetic analysis and voluntarily gave blood samples. Immortalized lymphoblast lines were established from peripheral blood samples as described previously.[ Genomic DNA was extracted from cultured lymphoblast cells using a PureGene DNA isolation kit (Gentra Systems, Minneapolis, MN).

cRNA capture/NGS

Custom biotinylated 120-mer coding ribonucleic acid (cRNA) baits were created to capture the entire gene sequences of SMARCB1, LZTR1, and NF2. Using lymphoblast DNA from each patient, sequencing libraries were prepared using the SureSelect XT Target Enrichment System for Illumina Paired-End Multiplexed Sequencing (3 μg), by Agilent Technologies (G7530-90000). Namely, 3 μg of genomic DNA was sheared to approximately 175 bp fragments using the Covaris Focused-ultrasonicator. DNA fragments were end-repaired, adenylated, ligated to adapter oligos, and then amplified with 5 cycles of polymerase chain reaction (PCR) according to the protocol provided by the manufacturer. After quantification, 750 ng of each amplified DNA sample was hybridized overnight with the Capture Library. Following capture cleanup, each genomic deoxyribonucleic acid (gDNA) library was amplified with an additional 16 cycles of PCR, which also tagged each sample with an index-specific barcode. Final products were quantified using the TapeStation 2200 and pooled for rapid mode sequencing on the Illumina HiSeq 2500, which generated a median 4.09 million paired-end reads of 100 bp. Quality checking of raw sequence reads was assessed by FASTQC (version 0.10.1) (Andrew, S 2010, FASTQC, a quality control tool for high-throughput sequence data http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Further, sequence reads were aligned to human reference genome Ensembl GRCh37 (version 71) using bwa-mem (version 0.7.5a-r418) at its default parameter setting[ (arXiv:1303.3997v2 [q-bio.GN]). Next, alignments were coordinate sorted duplicated reads filtered using Picard Tools's SortSam and MarkDuplicates algorithms, respectively (version 1.95; http://broadinstitute.github.io/picard/). Filtered alignments were further insertion–deletion (indel)-realigned and base-recalibrated using the Genome Analysis Toolkit (GATK, version 3.1.1).[ GATK's HaplotypeCaller was used to call raw variants in each sample with the following argument set; –stand_call_conf 30 –stand_emit_conf 10 –minPruning 3. A high-confidence set of variants was selected from intragenic regions along with 30 kb upstream and downstream regions of captured genes with DP > 20 and GQ > 30 filters and they were further annotated by Annovar (version. 02-01-2016).[ Candidate mutations were defined as exonic or splice site calls that were classified as either nonsense or deleterious missense (as predicted by in silico analysis or previous RNA studies), and were confirmed by Sanger sequencing analysis using an ABI Prism 3730 DNA analyzer. We used CNV-seq to call copy-number variations on intragenic regions of 5 target genes from capture-sequencing data [. We compared 30 subjects with schwannomatosis to an equal number of subjects with NF1 used as reference, with options—log2-threshold 0.2—P-value .05—genome human—global-normalization. Duplicated reads were marked by Picard Tool's MarkDuplicates algorithm (version 1.95) (https://broadinstitute.github.io/picard/) and further removed by SAMtools (version 0 1.18).[

Statistics

Summary statistics were compared between groups using a Wilcoxon test for 2-group comparisons and a Kruskal–Wallis test for 3-group comparisons of tumor number, total body tumor volume, pain scores, SF-36 bodily pain subscores, and other continuous variables. Chi-square statistics were used to compare binomial variables between groups, including presence of intracranial schwannomas, café-au-lait macules, and other clinical features. Spearman coefficients were used for comparisons of pain with either tumor number or total body tumor volume. Missing data were omitted from analyses due to a sample number too small for imputation. SAS version 9.4 was used for all statistics.

Ethical approval

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. The studies were reviewed and approved by the Partners Healthcare Institutional Review Board.

Results

From our original 51 patient cohort of schwannomatosis patients who underwent WBMRI, 14 were excluded from this study due to having no lymphoblast DNA available for sequencing. Another 7 patients previously underwent Sanger sequencing analysis and exhibited a germline SMARCB1 mutation; these were included here in the correlation analysis but not the targeted sequencing.[ For the remaining 30 subjects, we used a custom biotinylated cRNA library to capture the entire gene regions of SMARCB1, LZTR1, and NF2, which were then analyzed by targeted NGS. The median library size was 4 million paired-end reads with 7.7 million uniquely mapping individual reads. Median coverage was 1060× with 72% efficiency of capture within the target regions. The results of genetic analysis are shown in Table 1.

Table 1

Mutational analysis of schwannomatosis patients.

No patients had deleterious NF2 mutations or copy number variants on chromosome 22q in the germline. Of those sequenced by capture, 5/30 (16.7%) had a germline LZTR1 variant, 8/30 (26.7%) had a germline SMARCB1 variant, and 17/30 (56.6%) had no identified mutation. Mutations called by capture sequencing were verified by Sanger sequencing. Sanger confirmations of novel mutations are shown in Figure 1.

Figure 1

Sanger sequencing confirming novel mutational calls in LZTR1.

To the 30 capture-sequenced samples, we added the 7 samples for which mutations in SMARCB1 were already known by prior Sanger sequencing. Then, the 37 patients were divided into 3 groups based on their germline mutation: LZTR1 (n = 5), SMARCB1 (n = 15), or no identified alteration (n = 17). Patient characteristics for each group are shown in Table 2. There was no significant difference between age at diagnosis, median tumor number, or median total body tumor volume between the 3 groups. No intracranial meningiomas were identified in any patient in this cohort. Patients with LZTR1 mutations had a significantly higher prevalence of spinal schwannomas than patients with SMARCB1 mutations (100% vs 40%, P = .0195). Further, patients with LZTR1 germline mutations reported significantly higher pain intensity than SMARCB1 patients (3.9 vs 0.5, respectively; P = .0414), and they reported significantly worse QOL, as assessed by SF-36 pain burden scores, than the SMARCB1 group (P = .0106). Notably, inclusive of all mutation groups, neither pain intensity nor pain-related QOL was significantly different among patients with spinal schwannomas compared to patients without spinal schwannomas (P = .5071 and P = .3393, respectively). Additional phenotypic characteristics were compared between groups and were not significant (Table 2).

Table 2

Phenotype correlations by germline mutation.

Pain scores were correlated with tumor number and total body tumor volume within groups. For all patients, VAS pain intensity scores correlated with total body tumor volume (rho = 0.32471, P = .0499), though not with number of tumors (rho = 0.23065, P = .1696) (Fig. 2). Within each gene grouping, no such correlations were found (data not shown).

Figure 2

Correlations between pain and tumor burden. A significant correlation was found between patient-reported pain score and total body tumor volume, but not with tumor number.

Discussion and conclusions

Whereas 83% of surveyed schwannomatosis patients report chronic pain (≥3 months duration), and 75% take chronic pain medications (Plotkin, unpublished results), understanding the determinants of schwannomatosis-related pain is critical in order to create accurate laboratory models that can be generated and studied. To that end, we used a combination of WBMRI data, a novel high-throughput method for genotyping schwannomatosis germline samples, and patient reported pain levels and pain-related QOL to investigate genotype–phenotype correlations in 37 individuals with schwannomatosis. While mutation groups had no significant difference in tumor burden, there was nonetheless a significant difference in pain intensity between groups, with self-reported pain being significantly greater in patients with a germline LZTR1 mutation than those with a germline SMARCB1 mutation. Similarly, pain-related QOL was significantly worse (nearly one standard deviation) among the former group. Although the incidence of spinal schwannomas was much greater in patients with LZTR1 mutation than SMARCB1 mutation, pain and pain-related QOL were not significantly different between patients with and without spinal schwannomas without regard to mutational status, indicating that tumor location is not the primary driver of pain. We also showed here that patient-reported pain scores correlate with total body tumor volume among patients with schwannomatosis, but not with number of tumors. This stands in agreement with our previous work.[ Overall, our data indicate that schwannomatosis-associated pain may not only stem from total body tumor volume, but may also be related to the germline mutation that predisposes to schwannomatosis.

The reason for differences in pain between mutational groups is unknown, but may be due to function of SMARCB1 and LZTR1 genes. SMARCB1 is known to be involved in regulating expression of genes throughout the genome, due to its association with the SWI/SNF human chromatin remodeling complex. The function of LZTR1 is currently less well established, but has been identified as a likely tumor suppressor gene and may have a functional link with SMARCB1 through nuclear receptor corepressor (N-CoR) interactions.[ Individuals with no identified mutation were not included in the primary analysis due to the presumed genetic heterogeneity of this population, a limitation of any genotype-phenotype study in this disease.

Patients with an LZTR1 mutation were found to have spinal schwannomas significantly more frequently than the other schwannomatosis patients in this study. This may represent a mutation-specific tumor location predisposition, similar to sporadic olfactory groove meningiomas with SMO mutations.[ Our understanding of this genotype–phenotype relationship will improve as more patients with schwannomatosis are characterized.

Our custom cRNA capture library is an efficient sequencing method for high-throughput genotyping, and provides high-quality genetic information for identification of single nucleotide variants or indels. Additionally, when pooled with individuals with non-22q disease, this method provides sufficiently high quality data for copy number variant analysis. While the present study only included analysis of exons, splice sites, and 3′ untranslated regions (3′ UTR), introns were also captured and sequenced and will be analyzed in future work for deleterious variants within known effector sequences.

Whole body MRI provides high-quality phenotypic information regarding tumor burden in schwannomatosis, and has revealed large differences in overall tumor burden between individuals within the same mutation groups. The WBMRI in our study included only large field-of-view coronal STIR images, which are not reliably able to identify intracranial or intraspinal tumors. Regional MRI data were only available where it had been clinically indicated. Also of note, WBMRI slice thickness was 5 mm in this study, thus small (yet potentially painful) tumors may have been present but beyond the resolution of WBMRI. These limitations will be addressed in future studies by the addition of dedicated brain and spine imaging and using smaller slice thickness.

Together, capture sequencing and WBMRI provide high quality data for identifying genotype–phenotype correlations in schwannomatosis. Our finding of increased pain among patients with LZTR1 mutations is both important and timely, given that severe, chronic pain is common among patients with schwannomatosis, and that there are large-scale international efforts to create schwannomatosis-associated pain models and to find therapies for this unique population. While the small sample in our analysis is reflective of the rarity of this disease, further study of the identified genotype-phenotype correlation is warranted.

Author contributions

JTT: Study concept and design, acquisition of data, analysis and interpretation, critical revision of manuscript.

MJS: Analysis and interpretation, critical revision of manuscript

JAW: Acquisition of data, analysis and interpretation, critical revision of manuscript

SE: Acquisition of data, analysis and interpretation, critical revision of manuscript

MET: Study concept and design, acquisition of data

VM: Study concept and design, acquisition of data, analysis and interpretation, critical revision of manuscript

VR: Study concept and design, acquisition of data

WC: Study concept and design, acquisition of data, analysis and interpretation

GJH: Study concept and design, acquisition of data, analysis and interpretation

MAB: Study concept and design, acquisition of data, analysis and interpretation, critical revision of manuscript

MS: Analysis and interpretation, critical revision of manuscript

AS: Acquisition of data, analysis and interpretation

JFG: Study concept and design, analysis and interpretation, critical revision of manuscript, study supervision

SRP: Study concept and design, acquisition of data, analysis and interpretation, critical revision of manuscript, study supervision