Comprehensive Scanning of Somatic Mitochondrial DNA Mutations in Breast Cancer¹

Mitochondria are cytoplasmic organelles that generate energy in the form of ATP through OXPHOS (1).³ Deleterious mtDNA mutations have been reported to cause a broad spectrum of neuromuscular diseases (1). Pathogenic mtDNA mutations are usually heteroplasmic, whereas homoplasmic nucleotide substitutions are often benign polymorphisms. mtDNA is highly polymorphic. Although the biochemical consequence of homoplasmic polymorphisms are considered too subtle to cause any effect on OXPHOS, long-term accumulation of the subtle difference in OXPHOS activity may eventually result in oxidative stress (2). Thus, in the late onset of a disease such as cancer, mtDNA polymorphisms can potentially play a role in modifying the risk of developing cancer. mtDNA is subject to post-zygotic somatic mutation. The mutation rate of mtDNA is at least 10 times higher than that of nuclear DNA. Several reasons for the higher mutation rate have been suggested. First, the mitochondrial genome lacks the protective histone proteins. Second, the DNA repair mechanisms in mitochondria are very limited. Third, mitochondrion is the major site of reactive oxygen species production and, thus, are susceptible to oxidative DNA damage, which is believed to be associated with cancer. The important roles of mitochondria in energy metabolism, generation of reactive oxygen species, aging, and initiation of apoptosis suggest that mitochondria may serve as the switching point between cell death and abnormal cell growth, thus contributing to the neoplastic process (3). Somatic mtDNA mutations have been reported in colorectal, bladder, head and neck, lung, and ovarian cancers (2, 4, 5). The majority of these mutations are homoplasmic. The mutation spectrum varies among tumors of different tissues. However, most of the mutations occur in the D loop region, where the origin of replication and promoter are located. Previous studies were carried out by sequencing the mitochondrial genome. Although direct DNA sequencing is regarded as the gold standard of mutation detection, it is time-consuming, expensive, and impractical for routine analysis. Thus, most studies did not cover the entire mitochondrial genome, and the number of tumors analyzed was limited (2, 5). Some studies were restricted to the hypervariable D loop region. In this report, we applied TTGE method developed recently to scan the entire mitochondrial genome for somatic mutations in breast tumor. TTGE has the capability of distinguishing homoplasmic and heteroplasmic mutations (6, 7). It is sensitive to detect a low percentage of heteroplasmic mutations, and the size of the DNA fragment can be as large as 1 kb. The method is simple and fast. It does not involve the preparation of chemical denaturant gradient gel, and it does not require a stretch of guanine-cytosine base pairs (6, 7).

Frozen breast tumors and their matched normal tissues were obtained from histopathology and tissue bank shared resources of the Lombardi Cancer Center at Georgetown University Medical Center according to the Institutional Review Board approved protocol #92–048. DNA was isolated from frozen tissues using proteinase K and phenol-chloroform. Total DNA was quantified using florescent Hoechst dye H33258 with DYNA QUANT 200 according to the manufacturer’s protocol. DNA was diluted to 5 ng/μl to be used in PCR reactions.

Thirty-two pairs of overlapping primers were used to amplify the entire 16.6-kb mitochondrial genome. The DNA fragments vary in size from 306 bp to 805 bp with an average of 594 bp (6, 7). The amplified fragments totaled 18,953 bp, 14.4% more than the mitochondrial genome of 16,569 bp because of the overlapping regions. The position and sequence of the PCR primers, and the conditions of PCR and TTGE have been published recently (6, 7). On TTGE analysis, a single bandshift represents a homoplasmic DNA alteration, and a multiple-banding pattern represents a heteroplasmic mutation. The DNA fragments from normal and tumor tissues of the same patient were analyzed side-by-side. Any DNA fragments showing differences in banding patterns between the normal and tumor samples were sequenced to identify the exact mutations. DNA sequencing was performed by using a dye terminator cycle sequencing kit (Perkin-Elmer) and an ABI 377 (Applied Biosystem) automated sequencer. To identify mutations in low-percentage mutant mtDNA, the DNA bands containing the mutant were excised from the TTGE gel and PCR amplified before sequence analysis. The results of DNA sequence analysis were compared with the published Cambridge sequence (8) using MacVector 7.0 (Oxford Molecular Ltd., Oxford, England) software. Sequence variations found in both tumor and matched normal mtDNA were scored as germ-line variations. Each was then checked against the Mitomap database.⁴ Those not recorded in the database were categorized as novel mtDNA variations, and those that appeared in the database were reported as polymorphisms or mutations. Any DNA sequence differences between tumor and matched normal mtDNA were scored as somatic mtDNA mutations. To study MSI, 11 STRs in D loop, F, L, ND2, COI, ATP 6, CO III, and ND5 regions were analyzed by TTGE and also fully sequenced.

The common 4977-bp deletion was analyzed by PCR. The mtF8295 forward primer and mtR13738 reverse primer were used for the detection of deleted mtDNA, which yielded a 466-bp fragment. The mtF8259 and mtR8600 were used for the detection of wild-type mtDNA, which produced a DNA fragment of 306 bp. Both tumor and adjacent normal tissues were analyzed.

mtDNA from 19 pairs of tumor and matched normal breast tissue was analyzed by TTGE followed by sequencing of the DNA fragments showing different banding patterns, either the homoplasmic single bandshift or the heteroplasmic multiple bands on TTGE gel. Fig. 1 is an example of such analysis. The multiple banding pattern shown on TTGE analysis (Fig. 1,A) suggests the presence of a heteroplasmic mutation. This was followed by direct sequencing, which identified the heteroplasmic C16147T mutations (Fig. 1 B). Because the DNA was not derived from a microdissected tumor, it is possible that the apparent heteroplasmic mutation is attributable to the contamination of the surrounding non-neoplasitic cells.

Fourteen of 19 (74%) tumors had somatic mtDNA mutations (Table 1), and 12 of them had mutations in the hypervariable D loop region. Six tumors had 1 mutation, and each of the remaining 8 tumors had multiple somatic mutations with a total of 27 mutations (Table 1). Every somatic mutation occurred only once except for the insertion or deletion in the poly C region of nucleotide position 303–309. Among the 27 somatic mutations, 1 was in rRNA (3.7%), 4 in mRNA (14.8%), and 22 in the hypervariable D loop region (81.5%). Seventeen DNA alterations were at the homoplasmic state in tumor tissue, and 5 of them were heteroplasmic in the surrounding normal tissue. Again, the heteroplasmy could be attributable to the contamination with the surrounding tumor tissue, or the morphologically normal tissue may have already undergone molecular changes. The remaining 10 mutations were heteroplasmic in tumor tissue. Eight of them were changing from the homoplasmic state in the surrounding normal tissue, and 2 of them were also heteroplasmic in the surrounding normal tissue but with a quantitatively different proportion of the mutant mtDNA. This phenomenon is consistent with the progressive feature of tumorigenesis.

DNA fragments showing different banding patterns between normal and tumor tissue on TTGE analysis were sequenced to identify the presumed somatic mutations (Table 1). Meanwhile, when the sequence of normal tissue was compared with that of the published Cambridge sequence (8), numerous germ-line sequence variations were found (Table 2). A total of 102 distinct germ-line variations have been identified from the sequenced fragments. These do not represent all of the sequence variations in the specimens analyzed, because only the DNA regions that show somatic mutations by TTGE are sequenced. Twenty-one of these variations are novel, and 81 of them have been recorded in the Mitomap database (3). Nineteen of the 21 (90.5%) novel variations occurred only once, and each of the remaining 2 variations occurred twice (Table 2), whereas 34 of the 79 reported polymorphisms occurred multiple times. Among them, A73G, A263G, 303–309 ins C, 311–315 ins C, C7028T, A8860G, and T16519C, which occurred in >50% of the cases, represented polymorphisms in Cambridge sequence that had been revised and updated recently (8).

The common 5-kb deletion (np 8469 to np 13447) in tumor tissue was analyzed by PCR method using forward primer mtF8295 and reverse primer mtR13738. This method is sensitive enough to detect 0.01% of deleted mtDNA. Deletion was not detected in any of the breast tumors. To study MSI, 9 mononucleotide tract, a dinucleotide, and a trinucleotide repeat at np 303, 311, 514, 3566, 6692, 9478, 12385, 12418, 12981, 13231, and 16184, were analyzed by both TTGE and sequencing. Homoplasmic or heteroplasmic insertions or deletions were found in np 303–309 poly C tract only. They were not found in any other microsatellite regions. The insertion or deletion in np 303–309 is probably attributable to the hypervariability of the D loop region rather than a reflection of MSI (see “Discussion,” below). This differs from the high frequency of MSI observed in colon and gastric cancers (9, 10, 11).

In this report we present a comprehensive study of somatic mtDNA mutations in human breast cancer. Among the 22 distinct somatic mutations, 10 have been reported in other types of cancers (4, 5, 12, 13, 14, 15), and 13 were observed for the first time. The pathogenic role of each mutation in tumorigenesis is currently unknown, although some of these mutations are located in structurally/functionally important regions. For example, a nonconserved missense mutation, like the replacement of Leu with Pro in ATP synthase subunit 6, may be of significance, but more extensive biochemical and molecular studies will be necessary to determine the effect of this mutation on energy metabolism in tumor cells. A survey of the recent reports reveals that most of the studies on somatic mtDNA mutations in cancers focused on D loop and regions containing microsatellites (Table 3). Between 7 and 80% of the mitochondrial genome was studied with various methods including single-strand conformation polymorphism, two-dimensional gene scanning, and manual sequencing (2, 4, 5, 9, 11, 12, 13, 16, 17). The most comparable study was the investigation of 10 ovarian carcinomas by whole mitochondrial genome sequencing (4). Somatic mtDNA mutations were found in 60% (6 of 10) of ovarian carcinomas, and 33% of the mutations were in the D loop region (4). In this study, 74% (14 of 19) of breast tumors had mtDNA mutations, and 81% of the mutations were in the D loop region. Study of a large number of specimens in a wide diversity of human neoplasms will be necessary to determine the mtDNA mutation spectrum in various cancers. The heteroplasmic patterns observed (Table 1) may represent the true phenomenon rather than contamination, to suggest that DNA alteration has taken place at the molecular level before the gross morphological change and that tumorigenesis is a progressive process.

Contradictory observations of MSI in breast cancers have been reported. Anbazhagan et al. (18) evaluated >10,000 PCR of STRs in noncoding regions and found that nuclear MSI was uncommon in human breast cancer, whereas Richard et al. reported recently MSI in nuclear and mtDNA in breast cancer (17). The methods used to detect the MSI were either fragment size analysis or RFLP analysis, which may not be the most accurate means because of the interference of shadow bands and the possibility of incomplete digestion, respectively. We investigated the mtDNA MSI by both TTGE and sequencing methods, and found that insertion and deletion were detected only in the D loop region of np 303–309 where a stretch of 7 Cs was located. MSI was not detected in any of the other 10 STR regions. Three specimen pairs had T16189C substitution that resulted in a stretch of 10–13 Cs in the region. None of them showed obvious insertion or deletion of Cs. These observations implied that the np 303–309 is probably a mutation hot spot rather than the reflection of a true MSI. Study of nuclear MSI in these specimens is currently underway. The sequence between np 299 and np 315 is a conserved sequence block of mtDNA. Length variations in this region may play an important role in regulating mtDNA replication. The role of somatic mtDNA mutations in tumor progression has not been investigated. It is possible that mutations in the conserved regions, origins of replication, promoters of transcription, or transcription factor binding sites, may affect the total amount of mitochondrial transcripts and mature proteins. Ultimately, the overall OXPHOS activity of the mitochondria may be affected. Four mutations were found in the coding region. Three were in the NADH dehydrogenase subunit 2, which did not result in amino acid change. One mutation, T9131C in the ATP synthase subunit 6, resulted in the substitution of hydrophobic Leu residue with the secondary amino acid Pro at amino acid position 122. This T9131C variation has been observed and was reported as a polymorphism. Whether this mutation plays a role in tumorigenesis requires additional investigation. Certain polymorphisms may also play important roles in modifying cancer risk or the process of tumorigenesis for the same reasons. Worth mentioning is the T3398C mutation. This mutation, a Met to Thr at the amino acid position 31 in ND1, has been observed in a patient with progressive external ophthamoplegia and cardiomyopathy. There are two novel germ-line sequence variations, T14110C in ND5 changes a Phe to Leu at amino acid 592 toward the COOH termius, and G14207A in ND6 changes a Thr to Ile at amino acid 137. It should be noted that transcription of mitochondrial genome produced two polycistronic primary transcripts that are processed by endonuclease to yield the mature rRNA, tRNA, and mRNA molecules. Thus, mutations anywhere in the genome affecting the folding and secondary structure of the RNA precursors are potentially detrimental to RNA processing. Additional biochemical and molecular studies of RNA processing and protein expression shall be investigated in cell cultures to elucidate the biological effect of these sequence variations before any pathogenic significance can be assigned. In addition, many of the somatic mtDNA mutations in tumor may represent passenger mutations that do not play any primary role in tumorigenesis.

We recently reported a comprehensive screening of mtDNA mutations in patients suspected of mitochondrial disease. It demonstrated that the novel germ-line mutations span the entire mitochondrial genome. When the number of mutations occurring in each region was normalized to the size of the region, the tRNA and noncoding regions were 3 and 27 times, respectively, more susceptible to mutations than the mRNA region. This is consistent with the data in the Mitomap database⁴ in that most of the disease-causing mutations are in tRNA genes. In this study, somatic mutations in tRNA were not found. When the number of somatic mutations was normalized to the size of the region, it was found that the D loop region was ∼60 times more susceptible to mutation than the coding region (22 mutations in 1,122 bp of D loop region versus 5 mutations in 15,358 bp of coding regions). These results suggest that the mechanism of germ-line mutation and somatic mutation is different, although tissue specificity should also be considered. The displacement three stranded region is more susceptible to mutations but more so in cancer cells, perhaps because of the increase in cell growth. Additional investigation of the biochemical consequences of mtDNA mutations in disease and various types of tumors will provide insight regarding the roles of mitochondria in the pathogenesis of neuromuscular diseases, tumorigenesis, apoptosis, and aging.

We thank John Park for technical assistance. The tumor tissue specimens were provided by the Histopathology and Tissue Bank Shared Resources at Georgetown University Medical Center.