|Home | About | Journals | Submit | Contact Us | Français|
Neurofibromatosis type 1 (NF1) is a common, autosomal dominant, tumor-predisposition syndrome that arises secondary to mutations in NF1. Glomus tumors are painful benign tumors that originate from the glomus body in the fingers and toes due to biallelic inactivation of NF1. We karyotyped cultures from four previously reported and one new glomus tumor and hybridized tumor (and matching germline) DNA on Illumina HumanOmni1-Quad SNP arrays (~1 × 106 SNPs). Two tumors displayed evidence of copy-neutral loss of heterozygosity of chromosome arm 17q not observed in the germline sample, consistent with a mitotic recombination event. One of these two tumors, NF1-G12, featured extreme polyploidy (near-tetraploidy, near-hexaploidy, or near-septaploidy) across all chromosomes. In the remaining four tumors, there were few cytogenetic abnormalities observed, and copy-number analysis was consistent with diploidy in all chromosomes. This is the first study of glomus tumors cytogenetics, to our knowledge, and the first to report biallelic inactivation of NF1 secondary to mitotic recombination of chromosome arm 17q in multiple NF1-associated glomus tumors. We have observed mitotic recombination in 22% of molecularly-characterized NF1-associated glomus tumors, suggesting that it is a not uncommon mechanism in the reduction to homozygosity of the NF1 germline mutation in these tumors. In tumor NF1-G12, we hypothesize that mitotic recombination also “unmasked” (reduced to homozygosity) a hypomorphic germline allele in a gene on chromosome arm 17q associated with chromosomal instability, resulting in the extreme polyploidy.
Neurofibromatosis type 1 (NF1) is a common, autosomal dominant, tumor-predisposition syndrome that arises secondary to mutations in the tumor suppressor gene NF1 (Wallace et al., 1990). The disorder is associated with multiple types of benign and malignant tumors that arise from biallelic inactivation of NF1, due to loss of heterozygosity or somatic mutation (Brems et al., 2009a). Glomus tumors are painful benign tumors that originate from the glomus body, a thermoregulatory shunt concentrated in the fingers and toes Glomus tumors present with a triad of hypersensitivity to cold, severe paroxysmal pain and focal pain at the fingertip, and are treated by surgical excision. There can be a delay in diagnosis of many years, at a cost of significant pain and morbidity. The key to diagnosis is clinical suspicion, although magnetic resonance imaging may be useful. Glomus tumors occur sporadically, typically in middle-aged females, who almost always have a single tumor (Stewart et al., 2010). We recently reported genetic, functional and clinical evidence that glomus tumors are part of the tumor spectrum of NF1 (Brems et al., 2009b). When associated with NF1, the glomus tumors are often multi-focal and can recur. The fingers are typically affected, although the tumors can occur in toes as well. Although more common in adults, children can also be affected. Glomus tumors in NF1 are more common than previously recognized and may affect up to 5% of the adult NF1 population. We have recommended that adults with NF1 are screened for glomus tumors by asking the simple question, “Do the tips of your fingers ever hurt, especially when cold or bumped?” (Stewart et al., 2010)
To our knowledge, no cytogenetic analysis has been described in glomus tumors, either sporadic or NF1-associated. In this report, we describe the cytogenetics and high-resolution SNP array-based copy-number analysis of five NF1-associated glomus tumor cultures. Clinical details, NF1 mutation data and oligoarray analysis (but not karyotyping) have been previously reported for three of these tumors (NF1-G1, NF1-G3, NF1-G8) (Brems et al., 2009a); clinical data only has been reported on NF1-G12 (“Leu-8” in (Stewart et al., 2010)). We report the NF1 mutation on NF1-G12 and describe the NF1 mutation in a fifth, previously unreported tumor (NF1-G13)
Glomus cell culture from five pathologically-confirmed glomus tumors from five individuals with NF1 was established by treating tumors overnight with collagenase (160 units/mL) and dispase (0.8 units/mL) at 37°C. Glomus cells were grown to confluency in DMEM/F12 + 10% fetal bovine serum + penicillin + streptomycin and harvested. Clinical details, germline and somatic mutations in NF1 for tumors NF1-G1, NF1-G3 and NF1-G8 have been previously reported (Brems et al., 2009b); for these tumors, germline DNA and DNA from at least one passage were available. The clinical but not molecular details of tumor NF1-G12 were previously reported (“Leu-8” in Table 2 and Figure 3D of Stewart et al., 2010). Germline and somatic mutations in NF1 in tumor NF1-G12 were analyzed as previously reported (Messiaen et al., 2000). For tumor NF1-G12, germline DNA as well as DNA from passages 5 and 8 and from the primary tumor itself (before culture) was available. Tumor NF1-G13 was surgically excised from a 40-year-old man with NF1 and an eight-year history of severe, progressive right 5th digit pain. Magnetic resonance imaging of the digit showed a 4 × 12 mm lesion consistent with a glomus tumor in the lateral right 5th digit. Germline NF1 mutation detection on NF1-G13 was performed as previously reported (Messiaen et al., 2000).
Table 1 lists the DNA samples from cultured glomus tumor cells, uncultured (frozen) tumor and germline (peripheral white blood cells) hybridized to HumanOmni1-Quad SNP arrays (~1 × 106 SNPs; Illumina, San Diego, CA) according to the manufacturer’s instructions. Allele-specific copy-number changes, ploidy and loss of heterozygosity (LOH) were determined on DNA from primary tumor, matching cell cultures and peripheral white blood cells using ASCAT (version 2.0) (Van Loo et al., 2010). Borders of permissible tumor ploidy were set between 1.6n and 6n to allow for the extreme polyploidy of the NF1-G12 tumor, except for cases NF1-G12_P5 (ploidy range set from 3n to 6n, for consistency with other NF1-G12 passages) and NF1-G13_tumor (ploidy range set from 1.6n to 3n, for consistency with the cultured NF1-G13 tumor). Copy-number analysis was performed using GenomeStudio v.20011.1 (Illumina Inc., San Diego, CA), Partek Genomic Suite 6.5 (Partek Inc., St. Louis, MO) and Nexus 5.0 (BioDiscovery Inc., El Segundo, CA) software, all per the manufacturer’s instructions. The Partek analysis included selection of “Copy number analysis,” followed by “Detecting amplifications and deletions,” “Finding regions in multiple samples,” and “Finding overlapping genes.” “Genomic segmentation” was chosen as the partitioning algorithm (default parameters). The analysis in Nexus was performed using the default settings as well. For this analysis, we analyzed six tumor-germline pairs. Four pairs (NF1-G1, NF1-G3, NF1-G8 and NF1-G13) were hybridized to HumanOmni1-Quad arrays, as noted above. Lower resolution oligonucleotide array-CGH has been reported previously on NF1-G1, NF1-G3 and NF1-G8; no copy-number changes were observed (Brems et al., 2009b). The remaining two glomus tumor samples (NF1-G10_GT1 and NF1-G10_GT3) arose in two different fingers from the same individual with NF1 and were hybridized to Illumina HumanHap550v.3 SNP-arrays (~5 × 105 SNPs), as previously reported (Brems et al., 2009b). Note that for NF1-G1, NF1-G3 and NF1-G8 tumors DNA was isolated from early passage cell cultures, whereas for NF1-G13 and NF1-G10_GT1 and NF1-G10_GT3 tumors DNA was isolated from primary tumor tissues. For the NF1-G13 tumor, DNA was available from the primary tumor and matching low-passage cell culture. The raw SNP genotype data from the HumanOmni and HumanHap550 chips were analyzed separately in GenomeStudio and then combined in either Partek or Nexus applications. Manifest files (.bpm) for both types of chips were based on the NCBI36/hg18 human genome build. We searched for copy-number changes in at least three of the six tumors using either Partek or Nexus software, and then sought significant differences identified by both methods. We excluded the NF1-G12 tumor-germline pair from copy-number analysis, given the extreme polyploidy of the tumor genome.
Cytogenetic analysis and chromosome counts were performed with DAPI banding and spectral karyotyping (SKY, Applied Spectral Imaging Inc, Carlsbad, CA) on metaphase spreads from all five glomus tumors cell cultures. Higher passage cell culture was used in the karyotyping experiments: NF1-G1 (passage 11), NF1-G3 (passage 13), NF1-G8 (passage 14), NF1-G12 (passage 9) and NF1-G13 (passage 3). Metaphase slide preparations were made after mitotic arrest with colcemid (0.015 μg/mL, 2–4 hours) (GIBCO, Gaithersburg, MD), hypotonic treatment (0.075 mol/L KCl, 20 minutes, 37°C), and fixation with methanol–acetic acid (3:1). Mosaic counts were performed whenever possible. Due to the high level of heterogeneity of karyotypes of NF1-G12, we performed a clustering analysis of individual karyotypes in order to identify a predominant class. Clustering was performed in BRB-Array Tools microarray analysis software, v3.8.0 (Biometric Research Branch, Division of Cancer Treatment and Diagnosis, NCI/NIH, Bethesda, MD). Counts for each individual chromosome in 13 metaphases were used as an input. Clustering was performed using centered correlation and average linkage per the developer’s instructions.
In general, NF1-associated glomus tumors share many histologic similarities to sporadic ones: namely the cuboidal cells surrounding blood vessels, smooth muscle actin (SMA) positivity, and benign appearing rounded nuclei with moderate to abundant cytoplasm. Both tumors NF1-G12 and NF1-G13 were large and arose secondary to mitotic recombination of chromosome arm 17q; in addition, tumor NF1-G12 was also visible as a mass at the base of the nail bed. Tumor NF1-G12 was a fibrous fragment consisting of monotonous population of small pericyte-like cells with regular, round nuclei and no atypical nuclei; the features were characteristic of a glomus tumor. Immunohistochemistry showed diffuse expression of alpha-SMA in the pericyte cells, consistent with glomus tumor; the endothelial cells of the blood vessels expressed CD31. Tumor NF1-G13 was readily recognizable as a glomus tumor but was composed of mostly larger epithelioid cells with abundant pale eosinophilic cytoplasm consistent with epithelioid glomus tumors, or oncocytic changes in a glomus tumor (Slater et al., 1987; Shin et al., 1990) (Supplementary Figure 1).
Table 1 summarizes the germline and somatic mutations in NF1 for the five samples. In tumors NF1-G12 and NF1-G13, there is evidence of copy-neutral loss of heterozygosity of chromosome arm 17q not observed in the germline sample (Figure 1), consistent with a mitotic recombination event. This was likely an early or initiating event since, unlike other copy-number abnormalities, there is no diminution of the abnormality with increased passage number. Since chromosomal band 17q11.2 harbors NF1, mitotic recombination is the likely explanation of biallelic inactivation of NF1 in these tumors. No mutation in NF1 was identified in tumor DNA from NF1-G12, other than the known germline mutation.
There were few cytogenetic abnormalities observed in NF1-G1, NF1-G3, NF1-G8 or NF1-G13 (Table 1); ASCAT analysis was consistent with disomy for all chromosomes (Table 2). Glomus tumor NF1-G12 featured extensive polysomy across all chromosomes (Table 2, Supplementary Table 1 and Figure 2), but no re-arrangements or abnormalities in morphology, as determined by SKY (Supplementary Figure 2). We observed intra-tumoral heterogeneity in chromosome copy number in tumor NF1-G12. First, no two identical karyotypes were observed (Supplementary Table 1). Second, there was little similarity across clones in chromosome count, and the highest pair-wise correlation was ~0.5 (Figure 3). Third, in NF1-G12, we observed three different groups of chromosomes: near-tetraploidy (chromosomes 1, 7, 10, 11, 14, 15, 16, 22, X), near-hexaploidy (chromosomes 2, 3, 4, 5, 6, 8, 9, 21) and near-septaploidy (chromosomes 12, 13, 17, 18, 19, 20) (Table 2). The variety of abnormalities was most prominent in the primary tumor and decreased in passages 5 and 8 (Table 2 and Supplementary Figure 3 D, E, F). The decrease in heterogeneity of aberrations correlated with increased passage number and increased normal cell admixture (as determined by ASCAT), likely reflecting the growth of normal cells in culture (Supplementary Figure 3). The karyotyping was performed on cells from passage 9 (9.4% aneuploid), and thus under-estimates the true percentage of abnormal chromosomes in the original tumor, as estimated by ASCAT from frozen tumor DNA (Table 2).
Supplementary Tables 2, 3 and 4 summarize the copy-number analyses of the six glomus tumor-germline pairs in Partek and Nexus, respectively. Genes (or loci) with significant differences (nominal P value < 0.05, with no multiple testing correction) in copy-number in at least 3 out of 6 tumors are listed in the tables. Specifically, two genes (WDR17 and C16orf11) and one microRNA (MIR1267) were deleted in at least three of the six tumors studied using two different methods (Partek and Nexus). The Partek analysis identified 24 such genes, containing only deletions, while Nexus identified ten genes with copy-number differences: three with amplifications and seven with deletions. The median sizes of aberrant regions were similar in the two analyses: 1.9 and 1.5 kilobases in Partek and Nexus, respectively. There were two common genes in the two lists (WDR17 and C16orf11) and one microRNA (MIR1267). All three genes carried deletions in tumor tissue but were normal in germline samples.
This is the first study, to our knowledge, of the cytogenetics of glomus tumors, and the first to report biallelic inactivation of NF1 secondary to mitotic recombination of chromosome arm 17q in multiple NF1-associated glomus tumors. Three of the five studied tumors harbored discrete pathogenic germline and somatic mutations in NF1. Two of the five tumors (NF1-G12 and NF1-G13) had evidence of mitotic recombination of chromosome arm 17q; since the gene NF1 resides in chromosome band 17q11.2, mitotic recombination causes loss of heterozygosity and is thus the initiating event in the tumorigenesis of the glomus tumor. Four of the five glomus tumors studied were diploid with little evidence of aneuploidy or structural rearrangement by mosaic-count cytogenetic and SNP copy-number analysis. In only one of the four diploid tumors (NF1-G13) could DNA from the original frozen tumor be directly studied (the other three were cultured), and thus we may have underestimated the frequency of polyploidy. One tumor (NF1-G12) was polyploid across all chromosomes; this polyploidy was not a tissue-culture artifact and was also observed by analysis of allele-specific copy-number of DNA from the primary tumor. We identified two genes (WDR17 and C16orf11) and one microRNA (MIR1267) with copy-number changes (all deletions) in at least three of the six tumors studied using two different methods (Partek and Nexus) (Supplementary Tables 2, 3 and 4). We are hesitant to draw many conclusions from such a small sample set. MicroRNAs are short non-coding RNAs involved in post-transcriptional regulation of gene expression; MIR1267, located in the same locus but on the opposite strand of WDR17, has unknown function. Similarly, C16orf11 encodes a poorly characterized hypothetical protein. In a large sequencing study of medulloblastoma, a single missense mutation in C16orf11 was identified in one sample (Parsons et al., 2008). Lastly, WDR17 is expressed in retina and testis and is a candidate gene for autosomal recessive retinitis pigmentosa (Stohr et al., 2002; Roni et al., 2007; Geisert et al., 2009). In the COSMIC database, missense and synonymous mutations in WDR17 have been occasionally reported in medulloblastoma, ovary carcinoma and malignant melanoma (Wellcome Trust Sanger Institute COSMIC database, accessed September 15, 2011).
Mitotic recombination of 17q is a common, even predominant, mechanism for LOH of NF1 in NF1-associated tumors. Mitotic recombination has been reported in dermal neurofibromas (62%) (Garcia-Linares et al., 2011), spinal neurofibromas (75%) (Upadhyaya et al., 2009), plexiform neurofibromas (46%) (Steinmann et al., 2009), NF1-associated juvenile myelomonocytic leukemia (47%) (Steinemann et al., 2010) and gastrointestinal stromal tumors (Stewart et al., 2007). No evidence of correlation between the presence of mitotic recombination of chromosome arm 17q in NF1-associated tumors and tumor phenotype (e.g., size, growth rate or histology) has been reported. We note that the two glomus tumors with mitotic recombination were both unusually large and NF1-G13 harbored some unusual histologic features.
Atypically, the polyploid tumor (NF1-G12) was readily apparent on visual inspection at the base of the nail. Mitotic recombination is rare, although it has been reported as an important mechanism in other genetic disorders including ichthyosis with confetti (Choate et al., 2010). In previous work (Brems et al., 2009b), we described biallelic inactivation in seven glomus tumors, all secondary to discrete mutations or LOH in NF1, and none secondary to mitotic recombination. Glomus tumors are rare and thus caution needs to be exercised in drawing conclusions from small sample sets. However, this report increases the number of molecularly characterized glomus tumors (NF1-GT12 and NF1-GT13) to nine, of which two (22%) from two separate populations (North America and Belgium) arose secondary to mitotic recombination, suggesting that it may be a not-infrequent mechanism in the reduction to homozygosity of the NF1 germline mutation in these tumors.
There are numerous reported examples of NF1-associated tumors, typically malignant peripheral nerve sheath tumors, neurofibromas and leukemias, with high ploidy, although most have very complex karyotypes (Mertens et al., 2000; Wallace et al., 2000; Kobayashi et al., 2006; Liu et al., 2010; Mitelman et al., 2011). In NF1-G12 we observed high ploidy in multiple karyotypes in late-passage glomus cells but no other apparent chromosomal rearrangement other than the mitotic recombination of chromosome arm 17q (Supplementary Figure 2). This is consistent with allele-specific copy-number determination by ASCAT on DNA from the frozen (and therefore never cultured) tumor that showed near tetra-, hexa- and septaploidy in the majority of chromosomes. The diversity of observed ploidy in NF1-G12 suggests intra-tumoral heterogeneity and extensive sub-clonal structure and was not observed in the other glomus tumors.
Cell culture of tumors can distort the proportions of normal and non-normal cells; paradoxically the most abnormal, genomically deranged cells in vivo often do not grow well in vitro and can be overgrown by stromal cells (Buehring and Williams 1976; Gazdar et al., 2010). This can be seen with the increasing passage number of tumor NF1-G12; the magnitude of copy-number abnormalities decreases from the non-cultured primary tumor with passage number (Table 2 and Supplementary Figure 3 D, E, F). This is likely secondary to the growth of normal cells in culture. Thus it is possible that the cultures for NF1-G1, NF1-G3, NF1-G8 and NF1-G12 do not reflect the copy-number abnormalities present in the original tumor. Since the karyotyping was performed on higher passage tumor cell culture, the number of cytogenetic abnormalities present in the original tumor may be underestimated.
Maintenance of chromosomal stability is important in cellular homeostasis, and numerous chromosomal instability (CIN) genes have been identified (Chandhok and Pellman, 2009), including multiple candidates on chromosome arm 17q (e.g., MYC, ERBB2, NME1, MPO and MECOM). High ploidy is not uncommon in hematologic malignancies (Pan et al., 2009), but is less common in solid tumors (Chow and Poon, 2010) and may be either a cause or a consequence of tumorigenesis (Holland and Cleveland, 2009). More broadly, it remains controversial whether aneuploidy in general is a cause or a consequence of tumorigenesis (Chandhok and Pellman, 2009). The mitotic recombination of chromosome arm 17q was likely the initiating event in the tumor since it resulted in reduction to homozygosity of the mutated germline NF1 allele. Abnormalities of chromosome 17 have been implicated in other high-ploidy solid tumors (Olaharski et al., 2006; Bettio et al., 2008). Further, isochromosome 17q is one of the most frequent nonrandom genomic alterations in neuroectodermal tumor, medulloblastoma, chronic myeloid leukemia, acute myeloid leukemia and myelodysplastic syndrome (Carvalho and Lupski, 2008). Although mitotic recombination of chromosome arm 17q is common in NF1-associated tumors, high ploidy in the absence of other cytogenetic rearrangements is not, especially in benign tumors. We observed mitotic recombination in chromosome arm 17q in our tumor NF1-G13, but no polyploidy. Thus, in NF1-G12, we hypothesize that mitotic recombination also “unmasked” (reduced to homozygosity) a hypomorphic germline allele in a gene on chromosome arm 17q associated with chromosomal instability.
Supplementary Figure 1. Photomicrographs of tumor NF1-G13 showing histologic features consistent with epithelioid glomus tumor, or oncocytic changes in a glomus tumor. The lesional cells are characterized by small rounded nuclei with inconspicuous nucleoli and contain abundant pale eosinophilic cytoplasm. A. H&E stain, original magnification 400X. B. H&E stain, original magnification 600X.
Supplementary Figure 2. Spectral karyotyping (SKY) analysis of a metaphase from cell 612-07 from glomus tumor NF1-G12 tumor showing a 184, XXXX karyotype and no evidence of gross rearrangements.
Supplementary Figure 3. ASCAT profiles of glomus tumor culture and tumor DNA from this study. X-axis = chromosome number; Y-axis = copy-number. Purple = total copy-number; Blue = copy number of the minor allele. A: NF1-G1, early passage; B: NF1-G3, early passage; C: NF1-G8, early passage; D: NF1-G12, passage 5; E: NF1-G12, passage 8; F: NF1-G12 tumor (also depicted in Figure 1B); G: NF1-G13, passage 2; H: NF1-G13 tumor. Note mixture of near tetra-, hexa- and septaploidy in frozen tumor tissue NF1-G12 (panel F), which decreases with increased passage number and decreased aberrant cell fraction (panels D and E). See Table 2 for count of ploidy in all chromosomes.
Supplementary Table 1. Summary of ploidy from 13 (9.4%) aneuploid metaphases from 139 karyotyped metaphase spreads from NF1-G12; balance of 126 (90.6%) were diploid. No other chromosomal abnormalities were observed. Karyotyping was performed on culture passage 9 and likely underestimates the percentage of aneuploid cells in the original tumor.
Supplementary Table 2. Genes/loci with significant differences (nominal P value < 0.05) in copy-number in at least 3 out of 6 tumor-germline pairs, as identified by Partek analysis. Genes/loci on this list that were also identified by Nexus analysis are highlighted in boldface. Genomic coordinates are based on the NCBI36/hg18 human genome build.
Supplementary Table 3. Genes/loci with significant differences (nominal P value < 0.05) in copy-number in at least 3 out of 6 tumor-germline pairs, as identified by Nexus analysis. Genes/loci on this list that were also identified by Partek analysis are highlighted in boldface. Genomic coordinates are based on the NCBI36/hg18 human genome build.
Supplementary Table 4. Samples that harbor deletions in C16orf11, WDR17 and MIR1267, as determined by both Nexus and Partek analyses.
Supported by: The work was supported in part by the Division of Intramural Research of the National Human Genome Research Institute (NHGRI), the Division of Cancer Epidemiology and Genetics (DCEG) of the National Cancer Institute’s Intramural Research Program and in part with federal funds from the National Cancer Institute, National Institutes of Health, under Contract No. HHSN261200800001E. PVL and HB are postdoctoral researchers of the Research Foundation – Flanders (FWO) at KULeuven. This work is supported by the Fonds voor Wetenschappelijk Onderzoek (FWO)-Vlaanderen (G.0578.06) (EL), the cancer foundation ‘Stichting tegen Kanker’ (C.0011-204-208) (EL), and a Concerted Action Grant (GOA/11/010) from the KULeuven. EB is supported by a Research Fellowship from the “Vlaamse Liga tegen Kanker”.
We thank Settara Chandrasekharappa, Ph.D, Ursula Harper and Marypat Jones (all NHGRI) and Salma Chowdhury, Kedest Teshome and Aurélie Vogt of the NCI DCEG Core Genotyping Facility for assistance with SNP genotyping. We thank Stephen Wincovitch (NHGRI) for assistance with microscopy and Ludwine Messiaen, Ph.D (University of Alabama, Birmingham) for assistance with NF1 genotyping.
The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products or organizations imply endorsement by the U.S. Government.