Somatic Mitochondrial DNA Mutations in Neurofibromatosis Type 1-Associated Tumors¹

Neurofibromatosis type 1 (NF1) is an autosomal dominantly inherited disorder with an estimated prevalence range from 1:2,190 to 1:7,800 (1). There are no known ethnic groups in which NF1 does not occur or is unusually common. NF1 is caused by mutations in the NF1 tumor suppressor gene, resulting in reduced Ras-GTPase-activating protein activity of neurofibromin (2-5). Clinical features of NF1 include café au lait spots, axillary freckling, iris hamartomas (Lisch nodules), skeletal abnormalities, and learning deficiencies (4). The most prevalent clinical manifestations are the development of benign plexiform and cutaneous neurofibromas. Plexiform neurofibromas can progress at low frequency to highly malignant peripheral nerve sheath tumors. Cutaneous neurofibromas typically start to develop around puberty, and the number of these tumors increases with age. Other tumor types less frequently associated with NF1 include optical gliomas, myeloid leukemias, phaeochromocytomas, and astrocytomas.

Loss of heterozygosity is thought to be the underlying cause for development of benign neurofibromas (6-10). Indeed, loss of heterozygosity at the NF1 locus has been shown in Schwann cells of cutaneous and plexiform neurofibromas and in malignant peripheral nerve sheath tumors (11-14), whereas other studies were unable to confirm loss of heterozygosity for NF1 in astrocytomas and neurofibromas of NF1 patients (15, 16). The variable expressivity, the plethora of mostly benign tumors, and the variability in loss of heterozygosity for NF1 suggest that the threshold for tumorigenesis is perhaps reduced by systemic modulators (17-19).

The diffuse cutaneous neurofibromas typical for NF1 may arise at multiple sites due to the independent occurrence of tumorigenic mutations. Alternatively, an early mutation in a disseminating precursor cell during development may be responsible for the observed phenotype. Here, we use mutational analysis to identify somatic changes in mitochondria in benign neurofibromas and unaffected tissue of NF1 patients. Characterization of somatic mitochondrial DNA (mtDNA) mutations in multiple tumors of the same patient may illuminate the cellular origin of multiple tumors in NF1 from a common precursor.

The role of mitochondria in tumor development has gained much attention with recent reports of somatic mtDNA mutations in brain, ovarian, esophageal, breast, and colorectal human cancers (20-27). Mitochondria contain multiple copies of circular double-stranded DNA molecules that have a high degree of sequence variations among different individuals (28). In addition to energy production by oxidative phosphorylation, mitochondria play a crucial role in programmed cell death (29-32) and generate DNA-damaging reactive oxygen species as side products of normal function. MtDNA is an easy target for oxidative DNA damage due to the close proximity to reactive oxygen species production, the lack of protective histone proteins, and the limited repair capabilities. The accumulation of reactive oxygen species might also contribute to increased nuclear gene mutagenesis (33).

The characteristics of multisystemic manifestation, variable expressivity, and somatic mosaicism of NF1 as well as the reported associations of neurofibromin with highly energy-dependent microtubules (34) and mitochondria (35) may indicate the presence of specific somatic mtDNA mutations in tumors of NF1 patients. We analyzed the mtDNA mutations in neurofibromas and compared mutations between several distinct tumors of the same individual. The presence of somatic mtDNA mutations in nontumor cells and tumor cells was compared.

Somatic mtDNA mutations in neurofibromas were detected by parallel analysis of mtDNA from pairs of matched tumor and normal blood samples by temporal temperature gradient gel electrophoresis (TTGE). Multiple PCR products comprising the entire mitochondrial genome were used. Somatic nucleotide alterations are identified as differences in the banding patterns between the DNA from matched normal and tumor tissues. These analyses were carried out on a total of 19 patients with cutaneous neurofibromas and 18 patients with plexiform neurofibromas. Figure 1 illustrates the results from TTGE and sequencing analyses.

When the analyzed mtDNA sequences from blood were compared with those of the published sequence (36), 71 germ line sequence variations were revealed (Table 1). These do not represent all the variations because only the mtDNA PCR products of regions that showed somatic mutations in the paired tumor sample were sequenced. Nine of the germ line variations detected are novel. Many of the remaining reported polymorphisms occurred in multiple individuals. Among them, A73G and T16519C are common polymorphisms in various ethnic groups. The apparent high frequencies of A263G and 303-309insC are because these are polymorphisms in the Cambridge reference sequence (37, 38). Although germ line variations are generally considered silent, missense mutations such as the novel A265V (C5263T) alteration in the mitochondrial electron transport chain complex NADH dehydrogenase subunit 2 may have a functional effect.

The complete results of our analysis of NF1-associated cutaneous and plexiform neurofibromas are listed in Table 2. The overall percentage of neurofibromas with somatic mtDNA mutations was similar to that found for glioblastoma and medulloblastoma but lower than those in lung, breast, and oral cancers (20-27).

Somatic mtDNA mutations were detected in 9 of 18 plexiform neurofibromas (Table 2). Four of the plexiform tumors with mtDNA mutations harbored a single alteration; five cases had >1 mutation. A total of 27 different somatic mtDNA mutations were identified. All mutations were found in the displacement loop (D-loop) region. Insertions or deletions in the nucleotide positions (np) 303 to 309 poly(C) region were detected in multiple tumors. This region has been reported to be the somatically unstable mutation hotspot of breast cancer (25, 39). All nucleotide substitutions were transitions between T/C and A/G, consistent with oxidative DNA damage. Fourteen of 27 of the somatic mtDNA mutations in the plexiform tumors were alterations from homoplasmic state in blood to homoplasmic state in tumor. A shift from homoplasmy in blood to heteroplasmy in the corresponding tumor was found in six cases, and a shift from heteroplasmy in blood to homoplasmy in tumor was detected in four cases. Three pairs of matched samples had heteroplasmic nucleotide substitutions in both blood and tumor, but there were detectable differences in the degree of heteroplasmy as assessed by quantitative comparison of the nucleotide peak amplitudes in the corresponding sequence profiles (Table 2).

One patient with plexiform tumor, T173, harbored nine somatic mtDNA mutations. To rule out an error in DNA sampling, identity analysis on tumor T173 and corresponding blood sample B174 samples was done. Identical alleles were detected at five polymorphic sites: the short tandem repeat in intron 3 of the PAH gene (chromosome 12), the CTG repeats of the myotonin protein kinase gene (disease gene for myotonic dystrophy, chromosome 19), the CAG repeats of the androgen receptor gene (X chromosome), and the SCA1 (chromosome 6) and SCA3 (chromosome 4) genes (data not shown). These results support that tumor T173 indeed had nine somatic mtDNA mutations, four of which are novel. The cause for this unusually large number of tumor-specific somatic mtDNA mutations is not clear. Apparently, the point mutations in the origin of H-strand replication and the termination-associated sequence regulate the mtDNA synthesis and transcription in the tumor. A large number (>6) of somatic mtDNA mutations also occurred in ∼5% to 10% of medulloblastomas and breast and lung cancers (25, 40).

Cutaneous neurofibromas were analyzed from a total of 19 patients. Seven of these tumors had somatic mtDNA mutations (Table 2), and all of them occurred in the hypervariable D-loop region. NF1 patients usually have multiple cutaneous neurofibromas ranging from <10 to >1000. Surprisingly, multiple cutaneous neurofibromas resected from distinct anatomic sites on an affected individual always shared identical somatic mtDNA mutations (Tables 2 and 3): Analysis of tumor samples T104 and T105 (Fig. 2C) revealed a change in a short poly(C) sequence at np 303 to 309 in the conserved sequence block (CSB), which shifted from C8/C9 heteroplasmy in the blood sample (B106) of the same patient to near homoplasmy of C8 in both tumor samples (Table 3). Analysis of mtDNA samples from tumors T107 and T108 showed a shift from np 303 to 309 C7 homoplasmy in the matched blood sample B109 to C7/C8 heteroplasmy in each of the tumors. Tumors T119 and T120 both harbor the same homoplasmic T16304C mutation when compared with the corresponding blood DNA B121, which is homoplasmic for the wild-type T16304. Comparable changes were detected in tumors and blood of patients 4, 5, and 6 (Table 3). In patient 7, we detected a high proportion of np 303 to 309 C8 in blood and in tumor samples, whereas no D-loop mutations were found in patient 8. Tumor tissues from different parts of the same tumor also showed the identical mtDNA mutations with similar degrees of plasmy (Table 3, patients 4 and 5).

These results suggest that the tumor-specific mtDNA mutations may have already been present in nontumor cells and accumulated in the neurofibromas. To examine this, we analyzed mtDNA isolated from unaffected skin and matched cutaneous neurofibromas from three NF1 patients. Mutation analysis of these skin and tumor samples was focused on the D-loop and its surrounding region, because the data obtained from 37 neurofibromas revealed that all somatic mtDNA mutations occurred in the D-loop region. In one patient (patient 8), no somatic mtDNA D-loop mutations were found in the tumor and skin samples (Table 3). The other two sets of tumors displayed somatic mtDNA mutations (patients 4 and 5). A progressive increase in mutant mtDNA content was showed among blood, skin, and the neurofibromas in both of these patients (Table 3). In addition, the identical mutation was present in the skin, in different parts of the tumor, and in separate tumors from different locations of the same individual.

Neurofibromas are mixed-cell tumors composed predominantly of Schwann cells and fibroblasts, with a minority of perineureal cells, neurons, mast cells, and endothelial cells. To determine which cell type in an apparently homoplasmic cutaneous neurofibroma contains the mutated mitochondria, mutational analysis was done on laser capture microdissected cells. S100-positive Schwann cells, S100-negative cells, and endothelial cells were dissected from neurofibromas 583 and 584 (patient 4) and 592 and 593 (patient 5; Table 3). Fibroblasts and epithelial cells were also dissected from unaffected skin of the same patients. MtDNA from the separate cell types was analyzed for the presence of the previously identified D-loop mutation 303-309insC C7/C8.

In tumors 583 and 584, S100-positive Schwann cells contained almost exclusively the C8 mtDNA mutation (Fig. 2A-C). S100-negative tumor-derived cells, presumably fibroblasts, were also highly enriched for the C8 mtDNA mutation (data not shown). Although carefully selected, we cannot completely exclude the possibility that the S100 cell fraction contains neoplastic Schwann cells, as it is notoriously difficult to distinguish between the different cell types in neurofibromas. Tumor-derived nonneoplastic endothelial cells contained only C7 mtDNA (Fig. 2D-F), as did epithelial cells obtained from the skin sample 586 of the same patient (Fig. 2G). Dermal fibroblasts from skin sample 586 contained C7 and C8 mtDNA at a ratio of ∼30:70 (Fig. 2H). Similar results were obtained for the different cell types of tumor samples 592 and 593 and matched skin sample 588. The results of the cell type–specific mtDNA analysis are summarized in Table 4.

In the analysis of NF1-associated tumors, one single somatic mtDNA mutation was detected in each cutaneous neurofibroma, whereas an average of three mutations was detected per plexiform neurofibroma. This compares with an average of one to three somatic mtDNA mutations for other tumors (20, 22, 25, 27). It is noteworthy that 22 of 27 mutations in plexiform neurofibromas are nucleotide substitutions compared with only 2 of 7 mutations in cutaneous neurofibromas. One interesting observation is that mutations at np 204 and 207 occurred three times in three unrelated patients. This result suggests that either the np 204 and 207 are mutation hotspots or the mutant mitochondria have selective growth advantage.

All of the mtDNA somatic mutations identified in our study occurred in the hypervariable D-loop region of the mitochondrial genome. This is unique because numerous studies on lung, breast, ovarian, bladder, head and neck, glioblastoma, and oral cancers showed that 20% to 70% of somatic mtDNA mutations were found in coding regions (20-32, 40). The pathologic significance of mutations in noncoding regions of the mitochondrial genome is currently unknown. Apparently, mutations in the D-loop influence the origin of replication and promoter region and may affect the mitochondrial biogenesis, transcription, and protein expression (41, 42).

The most common somatic mtDNA mutations identified in our study are insertions or deletions at np 303 to 309. Mutations at np 303 to 309 were not observed in mtDNA from 40 muscle tissues of individuals ranging in age from 0 to 65 years.⁵

MtDNA mutations may occur randomly, and multiple mutations may or may not occur simultaneously. To reach a homoplasmic state may require some mechanism for selection. Homoplasmy for mtDNA mutations in tumors can also be caused by random segregation after a sufficient number of cell divisions (43). Thus, mutations in the origin of replication (D-loop region) may provide a replicative advantage of these mutant mtDNAs, or homoplasmy for the D-loop mutation in tumors may be the result of random segregation. The last mechanism is unlikely to explain the observed distribution of mtDNA mutations in individual neurofibromas. The independent appearance of the same mtDNA mutation in distinct tumors of the same patient in six analyzed individuals argues against a random process. This is supported by the observation that multiple mutations in tumors of the same patient were almost always found in a homoplasmic state. In addition to random segregation and replicative advantage, mitochondrial segregation dynamics and maintenance may also be influenced by the genetic background of cells (44). Because neurofibromin is closely associated with microtubules and mitochondria (34, 35), haploinsufficiency of NF1 may perhaps affect the stability and promote segregation of specifically mutated mitochondria in neurofibromas.

The present data indicate the presence of somatic mtDNA mutations in a heteroplasmic state in cells of nontumor tissues of NF1 patients. The mutated mitochondria, particularly the np 303 to 309 C8 genotype, accumulate and ultimately dominate in the multiple cutaneous neurofibromas of a patient. This specific mitochondrial distribution pattern, in association with NF1 haploinsufficiency, may be indicative of a lowered tumorigenic threshold in NF1. Furthermore, the mitochondrial signatures of plexiform and cutaneous neurofibromas of NF1 patients vary, underlying their separate developmental origins. Although the present findings cannot exclude a mechanism of frequent mutational events in the NF1 gene, leading to multiple cutaneous neurofibromas, they support the notion that multiple cutaneous neurofibromas in NF1 patients may arise from a widely distributed cell carrying an early somatic mitochondrial mutation.

Patients with NF1 were recruited through the Departments of Neurosurgery and Neurogenetics, Massachusetts General Hospital, Harvard University and the Department of Neurology, Klinikum Nord Ochsenzoll (Hamburg, Germany) according to institutional review board–approved protocols. Patients were phenotypically characterized for features of NF1, including number, location, and size of cutaneous neurofibromas. Only patients with a clear diagnosis of NF1 according to NIH criteria (45) were included in this study. Cutaneous and plexiform neurofibromas were removed during routine surgery, dissected into multiple aliquots, and frozen immediately. Two or more cutaneous neurofibromas resected from different anatomic sites on each individual were obtained from 19 patients. For three of these patients, skin samples were biopsied from an area overlaying the resected cutaneous neurofibroma and from an area distal to the tumor (Table 2, patients 4, 5, and 6). A single plexiform neurofibroma sample was obtained from each of 18 patients.

The age of patients with cutaneous tumors ranges from 16 to 50 years with a mean age of 37.6 years. The age of patients with plexiform tumors ranges from 8 to 73 years with a mean age of 28.8 years.

DNA was isolated from frozen tissues using proteinase K and phenol/chloroform extraction method. DNA was extracted from peripheral blood lymphocytes using a modified nonenzymatic method (46). Total DNA was quantified using fluorescent Hoechst dye H33258 (Sigma, St. Louis, MO) with DyNA Quant 200 (Amersham Biosciences, Uppsala, Sweden)according to the manufacturer's protocol and diluted to 5 ng/μL to be used in PCR reactions (47).

DNA isolated from 19 pairs of matched blood and cutaneous neurofibroma samples and 18 pairs of matched blood and plexiform neurofibroma samples was used for mutational analysis of the mitochondrial genome by TTGE.

Thirty-two pairs of overlapping primers were used to amplify the entire mitochondrial genome by PCR (47). The PCR-amplified DNA fragments vary from 306 to 805 bp long. The positions and sequence of the PCR primers and the PCR and TTGE conditions were as described recently (47). Briefly, the DNA template, after the initial denaturation at 94°C for 5 minutes, was amplified with 35 cycles of 94°C for 45 seconds, 55°C for 45 seconds, and 72°C for 45 seconds and completed by 4 minutes of extension at 72°C. The PCR products were denatured at 95°C for 30 seconds and slowly cooled to 45°C for 45 minutes at a rate of 1.1°C/min. The reannealed homoduplexes and heteroduplexes were maintained at 4°C until TTGE analysis was done on a D-Code apparatus (Bio-Rad Laboratories, Hercules, CA). Five microliters of denatured and reannealed PCR products were loaded onto a polyacrylamide gel (acrylamide/bis 37.5:1) prepared in 1.25× Tris-acetate-EDTA buffer containing 6 mol/L urea. Electrophoresis was carried out at 145 V for 4 to 5 hours at a constant 1°C/h to 2°C/h temperature increment (47). The temperature range was determined by computer simulation from the melting curve of the analyzed DNA fragment (MacMelt software, Bio-Rad Laboratories). The gels were stained with 2 mg/L ethidium bromide for 5 minutes and imaged with a digital CCD gel documentation system (high-performance ultraviolet transilluminator, Ultraviolet Products, Upland, CA).

On TTGE analysis, a single-band shift represents a homoplasmic DNA alteration, and a multiple-band pattern represents a heteroplasmic mutation (48). Any DNA fragments showing different banding patterns between matched blood and tumor sample pairs were sequenced to identify the exact mutations.

To rule out the possibility that some mutations may not be detectable by TTGE, we randomly chose six samples that did not show TTGE-positive banding patterns and sequenced 10 coding regions containing stretches of six to eight homopolynucleotides. No mutations were found.

Confirmation of the nucleotide alteration was carried out by direct DNA sequencing of the purified PCR product using the original PCR primers and a BigDye terminator cycle sequencing kit (Perkin-Elmer, Foster City, CA) and analyzed on ABI 377 (Applied Biosystems, Foster City, CA) automated sequencer. All mutations identified were confirmed by repeat PCR and sequencing. The results of DNA sequence analysis were compared with the published Cambridge sequence (37) using MacVector 7.0 (Oxford Molecular Ltd., Oxford, United Kingdom) software. Sequence variations found in both tumor and blood mtDNA were scored as germ line variations. Each was then checked against the MITOMAP database (36). Variations not recorded in the database were categorized as novel mtDNA variations, and those appearing in the database were reported as polymorphisms. Any mtDNA sequence differences found between a tumor sample and its corresponding blood sample were scored as somatic mtDNA mutations specific to the tumor.

Microdissection was done using a PixCell II (Arcturus, Mountain View, CA) laser capture microscope according to the manufacturer's instructions. Frozen 10 μm sections were prepared on uncoated slides from cutaneous neurofibromas and unaffected skin tissues from the same patients (Table 2, patients 4 and 5). The sections were stained in H&E immediately before microdissection. Adjacent control slides were stained with antibody to S100 protein to confirm histologic and morphologic differentiation of Schwann cells within the neurofibromas. S100-negative cells, S100-positive Schwann cells, and endothelial cells were dissected from neurofibromas. Epithelial skin cells and dermal fibroblasts were dissected from skin samples. About 200 to 300 S100-positive Schwann cells, S100 negative cells, and morphologically distinguishable endothelial cells from neurofibromas, epidermal epithelial cells, and dermal fibroblast from normal skin were dissected for DNA extraction (PicoPure extraction kit, Arcturus). The D-loop region of the mtDNA was amplified by PCR and directly sequenced.

We thank Dr. Mia MacCollin for valuable discussions and continuing support of this project.