Somatic mutations in mtDNA have recently been identified in colorectal tumours. Studies of oncocytic tumours have led to hypotheses which propose that defects in oxidative phosphorylation may result in a compensatory increase in mitochondrial replication and/or gene expression. Mutational analysis of mtDNA in thyroid neoplasia, which is characterised by increased numbers of mitochondria and is also one of the most common sites of oncocytic tumours. has been limited to date. Using the recently developed technique of two-dimensional gene scanning, we have successfully examined 21 cases of thyroid tumours, six cases of non-neoplastic thyroid pathology, 30 population controls, nine foetal thyroid tissues and nine foetal tissues of non-thyroid origin, either kidney or liver. We have identified three different somatic mutations (23%) in papillary thyroid carcinomas. In addition, we have found significant differential distributions of mtDNA sequence variants between thyroid carcinomas and controls. Interestingly, these variants appear to be more frequent in the genes which encode complex I of the mitochondrial electron transport chain compared to normal population controls. These findings suggest first, that somatic mtDNA mutations may be involved in thyroid tumorigenesis and second, that the accumulation of certain non-somatic variants may be related to tumour progression in the thyroid.
In addition to the role of mitochondria in energy production and reactive oxygen species production, recent interest has increasingly centred on their role in apoptosis (Green and Reed, 1998). It is the modulation of these processes and their relationship to cell death and survival that has prompted recent investigations into the role of mitochondria in tumorigenesis (Cavalli and Liang, 1998). A small number of studies have demonstrated somatic mutations in colorectal and gastric carcinomas (Alonso et al., 1997; Burgart et al., 1995; Polyak et al., 1998). However, most studies which have focused on breast, thyroid and renal neoplasms have been limited by small sample sizes and insensitive or incomplete screening techniques and have found conflicting results (Bianchi et al., 1995; Ebner et al., 1991; Horton et al., 1996; Johnson et al., 1996; Muller-Hocker et al., 1998; Tallini et al., 1994; Welter et al., 1989).
Oncocytic tumours, because of their large abundance of mitochondria, are of particular interest. Although the thyroid is one of the most common sites of oncocytic tumours, of particular note is that all thyroid carcinomas, not just Hürthle cell tumours, have increased numbers of mitochondria (Stefaneanu and Tasca, 1979). Studies of these tumours as well as mitochondrial diseases have led to several hypotheses including one in which defects in oxidative phosphorylation are sensed by mitochondrial regulatory mechanisms, producing a compensatory response of increased mitochondrial proliferation and/or gene expression (Ebner et al., 1991; Wallace, 1992).
Mutation analyses of mtDNA in thyroid tumours have been limited to date. Prior studies of tumours of the thyroid have been limited in size and scope (Ebner et al., 1991; Johnson et al., 1996; Muller-Hockner et al., 1998; Tallini et al., 1994). The four studies combined have reported on only 19 cases of thyroid tumours; of these 19, seven were found to have mtDNA sequence variation and 12 were not. Given the size of the mitochondrial genome, the majority of these studies (Ebner et al., 1991; Muller-Hocker et al., 1998; Tallini et al., 1994) only targeted a few selected areas of the genome and relied on restriction analysis. Because of practical technical limitations, there has yet to be a study that systematically analyses the coding regions of the mitochondrial genome in sporadic adult human thyroid tumours comparing different pathologies as well as region-matched population controls.
The entire mitochondrial genome is circular and totals 16.6 kb in length. It encodes 13 polypeptide genes, the 12S and 16S RNA genes, and 22 tRNAs which are involved in oxidative phosphorylation (OXPHOS). OXPHOS comprises five multisubunit complexes (I–V) and all are encoded by a combination of nuclear and mtDNA except for complex II, which is only encoded by nuclear genes. Complex I (NADH dehydrogenase) contains 43 subunits, seven of which are encoded by mtDNA genes (ND1, 2, 3, 4, 4L, 5 and 6) (Wallace, 1999). One subunit of Complex III [cytochrome b (CYT B)], three subunits of Complex IV (COI, II and III) and two subunits of Complex V (ATPase 6 and 8) are encoded by mtDNA genes (Wallace, 1999).
Recent development of two-dimensional gene scanning (TDGS) to analyse critical areas, i.e. all areas of known pathogenesis, of the mitochondrial genome (van Orsouw et al., 1998) has prompted us to conduct a comprehensive examination for mtDNA mutations and sequence variation in thyroid tumours. Although only displaying 25% of the mitochondrial genome, this method targets multiple areas including all protein encoding genes, as well as rRNAs and 19 of 22 (86%) tRNAs residing in the genome (Table 1). It is also able to detect a heteroplasmy rate as low as 1% (van Orsouw et al., 1998). Using TDGS, we have successfully scanned 21 cases of thyroid tumours, six cases of non-neoplastic thyroid pathology, 30 population controls, nine foetal thyroid tissues and nine foetal tissues of a second type either kidney or liver. We have identified three somatic mutations, all in papillary thyroid carcinomas (PTCs) (19% of all carcinomas; 23% of all PTCs). In addition, we have found 66 sequence variants, representing 39 distinct variations, present in pathological thyroid samples as well as their paired normal thyroid tissues.
Using TDGS, we analysed the mitochondrial genome from 27 paired pathologic and normal thyroid samples, 30 leukocyte samples from controls derived from the normal population, and nine paired foetal thyroid and non-thyroid tissues (Figures 1 and 2). Interestingly, only the PTCs were found to have somatic mtDNA mutations. Among 13 PTCs, there were three somatic homoplasmic mtDNA mutations (23%; Table 2a,c). In one PTC (multicentric with 0.4 cm nodules), a somatic 7521 G>A mutation was found in the gene encoding tRNAAsp. In a second PTC (low-grade follicular variant), a somatic 10398 A>G was found, which resulted in T113A in the ND3 gene. In the third (1.1 cm PTC), a somatic 15179 G>A, predicted to result in V144M, was found in CYT B. The latter had a lymph node with metastatic PTC. The former two demonstrated no evidence of lymph node metastasis or vascular invasion. None exhibited Hürthle cell changes and none of the individuals had a history of familial thyroid disease. In fact, retrospectively, two individuals reported a positive family history of thyroid disease (1 HCA, 1 PTC) and neither had any mitochondrial variants identified.
Among all samples combined, both pathologic thyroid tissues and normal controls, we identified a total of 136 sequence variants, representing 51 distinct mtDNA variations (Table 2). Twenty-one of the 51 distinct variants have not been previously reported, according to the mitochondrial database, MITOMAP (http://www.gen.emory.edu/mitomap.html). Of these 51 different variations, 16 were predicted to alter the amino acid sequence of the encoded proteins; 24 were predicted to be neutral; eight affected tRNA or rRNA genes, and three were in non-coding regions between genes. All were single nucleotide base substitutions.
Of the 39 sequence variants found in the thyroid samples, 24 (60%) were unique to these samples. That is, they were not found in the mtDNA derived from peripheral blood leukocytes from the normal adult population or from the paired foetal tissue controls (Table 2). Ninety-five per cent (38 of 40) of these variants occurred homoplasmically. One variant, 8270 C>G, which was also homoplasmic, was found in normal thyroid but not its corresponding tumour. The remaining variants were either found in both the thyroid samples and normal population and/or foetal tissue controls (15) or the controls alone (12).
Since most variants found in the thyroid samples, excluding the somatic mutations mentioned above, were found in both the tumour and corresponding normal adjacent thyroid tissues, we wanted to confirm that this was not a tissue-specific phenomenon. We therefore compared paired foetal thyroid and a second organ obtained from therapeutic abortions. All mtDNA variants, which were identified in foetal thyroid tissues, were also found in the second paired non-thyroid tissue.
As multiple sequence variants (1–6 per sample) were identified in the pathologic tissues studied as well as normal population and foetal tissue controls, we sought to determine whether the distribution or quantity of variants that were found in the pathologic tissues differed from those in the control populations. The distributions of the number of mtDNA sequence variants were compared using the Kruskal-Wallis test (Siegel and Castellan, 1988) (Table 3). Comparisons between different pathologic sample types and different variant types were made. The Kruskal-Wallis test compares the distribution of the number of variants per sample, and is most powerful when populations have similarly shaped distributions that are shifted with respect to each other. In order to further examine the differences in the distribution of variants among our samples that were identified by the Kruskal-Wallis tests, we compared the mean number of variants per sample (Table 4) between different pathologies. Given the type of data and small sample sizes, no statistical tests were performed when comparing the means.
Normal population samples vs foetal tissues
No significant difference was identified in the distribution of the number of variants between the 30 normal blood samples and the nine foetal thyroid tissue samples. They were, therefore, considered as a single group of ‘normal controls’ (hereafter collectively referred to as ‘controls’) comprising a total of 39 samples.
Pathologic thyroid samples vs controls
When the sequence variants in the 27 cases with thyroid pathology were compared to those of the 39 controls, a significant difference was observed in the distribution of variants (P=0.05; Table 3). Specifically, cases with thyroid pathology appeared to have more variants than controls. There also appeared to be nearly two times as many variants per sample on average among the aggregate abnormal samples compared to the normal samples (2.556 versus 1.385; Table 4).
This difference between the thyroid samples and controls was most likely due to the significant difference between carcinomas and controls (P=0.002; Table 3). The distributions of all variant types between carcinomas and controls differed significantly. There was no difference in distributions between either benign or non-neoplastic samples compared to controls.
Carcinoma cases vs adenoma cases
Interestingly, carcinomas (n=16) and adenomas (n=5) differed significantly only in their distributions of neutral coding variants (P=0.05; Table 3). There was a fourfold difference in the mean number of neutral coding variants that were identified in the adenomas (0.200) versus carcinomas (0.875; Table 4). Only one of five (20%) adenoma samples had a neutral variant in a coding region, compared to 10 of 16 (63%) in the carcinoma samples.
Neoplastic vs non-neoplastic thyroid pathology
Although there was limited power to detect such differences with the small sample sizes available, we found that there were no significant differences between the distribution of total variants in the neoplastic (carcinomas plus adenomas) (n=21) and non-neoplastic (n=6) thyroid samples. However, there was some evidence that these sample groups may differ in the distribution of the number of variants in the genes encoding tRNAs and tRNAs (P=0.05; Table 3). Specifically, none of the non-neoplastic samples had any variants in the RNA encoding genes, while 11 of 21 (52%) neoplastic samples had at least one variant with an average of 0.857 (Table 4) variants per neoplastic sample.
Mean number of variants in electron transport chain
In order to examine whether the mtDNA sequence variants which were identified were possibly occurring at preferential ‘hotspots’ in the protein encoding genes of the electron transport chain, we also computed the mean number of variants per sample by complex (Table 5). Complex I (NADH dehydrogenase) had the greatest difference in the mean number of coding variants between normal (0.513) and pathologic samples (1.074). Although there was also a twofold difference in the mean number of variants per sample in complex 4, the total number of variants was considerably lower (0.179 for normal samples versus 0.370 for abnormal samples).
Systematic examination of the mitochondrial genome has led us to uncover three somatic mutations in PTCs and a significant differential distribution of mtDNA variants in cases with malignant thyroid pathology versus normal controls. Although confounded by a small sample size in benign and non-neoplastic thyroid pathology, the finding of significant differential neutral coding and RNA variant distributions between benign and malignant and between non-neoplastic and neoplastic thyroid cases, respectively, is also interesting.
While finding somatic mtDNA mutations in PTCs is novel, these mutations occur only in a minority of tumours, albeit a significant minority. However, whether the 23% mutation frequency among PTCs is an underestimate is not known. Although TDGS targets all known pathogenic areas of the mitochondrial genome, it only examines segments of specific genes (Table 1). Therefore, variants or mutations lying outside these segments would be missed. In addition, because the tissues were dissected based on their gross pathologies and because TDGS is exceedingly sensitive in detecting mismatch, it is possible that a small amount of tumour contamination of normal thyroid tissue could result in a false positive. Nonetheless, despite this caveat, the finding of somatic mtDNA mutations in a subset of sporadic PTCs is significant and is consistent with the hypothesis that these represent later events, events related to tumour progression and not necessarily initiation.
All three somatic mutations found in the PTCs are missense mutations. As such, it is difficult to assign functionality to missense mutations, and this is particularly so for mtDNA alterations. In general somatic missense mutations noted in the tumour and not the corresponding germline/transmitted DNA from the same individual usually suggests pathogenecity. Further, the majority of heritable mtDNA mutations found in neurodegenerative diseases are missense (reviewed by Simon and Johns, 1999). Two of the three somatic mutations identified in PTCs (10398 A>G in ND3 and 7521 G>A in tRNAAsp) were also found in the heritable (not somatic) mitochondrial genome of other individuals with non-PTC tumours as well as normal controls. This suggests that when these mutations are somatic, they are specific to PTCs and that the 10398A>G and 7521G>A variants might not be totally innocuous despite being found in the transmitted mitochondrial genome of other individuals. One might speculate that these somatic mtDNA mutations are low penetrance modifiers of tumour risk which, along with other mtDNA alterations may augment this risk. Alternatively, it is also possible that these represent mtDNA damage from oxidative stress. In other words, cause and effect cannot be separated at present.
A recent comprehensive study demonstrating somatic mitochondrial mutations in colorectal tumours found that, of 88 novel sequence variants, only 12 were somatic (Polyak et al., 1998). This, together with prior studies which demonstrated the presence of mtDNA variants in both tumour and adjacent tumour tissue (Bianchi et al., 1995; Johnson et al., 1996) and our own findings, prompted us to hypothesise that non-somatic variants may also play a role in tumorigenesis. Although additionally supported by our findings that there were significant differences in the distribution of the number of variants between, for example, carcinomas and controls, we were, nonetheless, concerned that these non-somatic variants represented either tissue specific variants or an admix of tumour and normal cells. We demonstrated that this was unlikely since: (1) there were no differences in the distribution of variants among nine distinct sets of foetal thyroid tissue paired with another non-thyroid tissue (kidney or liver) from the same; and (2) the distribution in the foetal tissues was similar to that from peripheral blood leukocyte mtDNA from the adult normal control population. Hence, our observations are likely specific for cases with thyroid pathology.
Because thyroid tissues do have high energy requirements (Wallace, 1999), should the accumulation of certain non-somatic variants affect the efficiency of OXPHOS, individuals harbouring those variants may be predisposed, in some way, to develop thyroid neoplasia. For example, if increasing numbers of mtDNA variants, particularly seen in the carcinoma cases in our study, cause alterations in OXPHOS, then our data would suggest that these specific variants might be involved in or associated with tumour progression. This would be in agreement with the recently proposed model in colorectal tumours where somatic mitochondrial mutations, together with other polymorphic variations in mtDNA, would result in subtle changes that would generate slightly higher levels of reactive oxygen species, thereby producing an environment conducive to cell proliferation (Polyak et al., 1998). This, in addition to the high mutation rate of mtDNA (10 times that of nuclear DNA) (Wallace, 1994) and findings that MnSOD levels decrease in tumour cells (Oberley and Oberley, 1988), suggests that tumour cells and specifically mtDNA in these cells are especially susceptible to free radical damage. Therefore, whether the accumulation of mtDNA variants overall is correlated with thyroid tumorigenesis and/or progression, or is the result of the neoplastic process remains to be investigated. Study of these variants and their possible functional impact on OXPHOS may help us better understand the role of mitochondria in tumorigenesis.
Materials and methods
Twenty-one tumours (one Hürthle cell adenoma, four follicular adenomas, 13 PTCs, one follicular carcinoma, one insular carcinoma, one medullary carcinoma) and six noncarcinomatous pathologic samples (three cases of multinodular hyperplasia and three cases of Hashimoto's thyroiditis) and their corresponding normal adjacent tissue were obtained at the time of surgery, carefully dissected by expert pathologists, and snap frozen.
Peripheral blood leukocytes obtained from 30 random normal adult volunteers were used for a control population. These specimens were stripped of all identifiers for complete anonymization. In addition, nine anonymized paired foetal thyroid tissues and kidney or liver tissues were used as controls for analysing the tissue specificity and frequency of variants found. All foetal tissues were obtained from therapeutic, and therefore, most likely normal, first trimester abortions. All tissues were obtained according to Human Research Committee protocols (Brigham and Women's Hospital, Boston, MA, USA).
DNA was extracted from frozen pathologic and normal thyroid tissue pairs, and foetal tissues, using the QIAmp Tissue Kit (Qiagen Inc., Valencia, CA, USA). DNA was extracted from peripheral leukocytes using standard protocols (Ausubel, 1994–1998). These protocols result in the extraction of both genomic and mtDNA.
All primers used were obtained from Gibco–BRL (Rockville, MD, USA) and were identical to those previously described (van Orsouw et al., 1998). A PTC-200 thermal cycler (MJ Research, Watertown, MA, USA) was used. Long-distance PCR products of 16.4 kb in length were obtained using the LA PCR Kit (Takara, Japan). Multiplex short PCRs of two multiplex groups (I and II) and singleplex short PCRs were carried out as previously described (van Orsouw et al., 1998).
Because of the potential for the presence of homoplasmy, both groups of multiplex PCR fragments were also subjected to a heteroduplexing reaction with a positive ‘normal’ control, i.e. pooled human DNA (Promega, Madison, WI, USA), prior to electrophoresis. The heteroduplexing reaction was carried out as previously described (van Orsouw et al., 1998).
TDGS and DGGE
For TDGS, a semi-automated Two Dimensional Gene Scanning System (TDGS-4002, C.B.S. Scientific Co., Solana Beach, CA, USA) was used (Dhanda et al., 1998). This TDGS was similar to the DGGE-4000 apparatus previously described (van Orsouw et al., 1998); it possessed the additional ability to run the first (horizontal) and second (vertical) dimensions sequentially without gel manipulation. This obviated both the needs to excise bands from the first dimension and for a second gel for the second dimension.
Similar to van Orsouw et al. (1998), a 1 mm thick 10% polyacrylamide (acrylamide : bisacrylamide=37.5 : 1) gel with a urea formamide (UF) gradient between 25 and 55% (100% UF is 7.0 M urea and 40% formamide) was used. The 55% UF solution also contained 10% glycerol. Gradient gels were created by mixing the two solutions (25% UF and 55% UF) in a linear gradient former (Gibco–BRL, Rockville, MD, USA). After polymerization of the gradient gel, a nondenaturing stacking gel comprised of 10% polyacrylamide was poured on top of the gradient gel. A TDGS comb was then inserted.
After heteroduplexing, 5 μl each of multiplex groups I and II from a single sample were combined and mixed with loading buffer (0.25% xylene cyanol, 0.25% bromophenol blue, 15% Ficoll, and 100 mM Na2 EDTA) as previously described (van Orsouw et al., 1998). Electrophoresis for the first (horizontal) dimension was conducted at 50°C for 3.5 h at 250 V in 0.5×TAE (50×TAE=2 M Tris-Acetate and 50 mM EDTA, Gibco–BRL, Rockville, MD, USA). For the second (vertical dimension), the temperature was increased to 60°C and then electrophoresis was carried out for 13.5 h at 100 V. Gel-based two-dimensional ‘fingerprints’ were visualised after staining with Ethidium Bromide (1 μg/ml), and recorded using a gel documentation system (Alpha Innotech Corp., San Leandro, CA, USA).
For one-dimensional (standard) DGGE, 0.75 mm thick 10% polyacrylamide with identical denaturing gradient gels and a 10% nondenaturing polyacrylamide stacker, were poured. Combs used in DGGE contained multiple wells, allowing single fragment analysis. Conditions were identical to the second dimension mentioned above.
Fragments of interest were amplified using non-GC-clamp primers. Products were gel and column purified using the Wizard PCR-Prep kit (Promega, Madison, WI, USA). Sequence analysis was then carried out using the Applied Biosystems Prism Dye Terminator Cycle Sequencing Ready Reaction Kit (Perkin-Elmer Corp., Norwalk, CT, USA). The products of cycle sequencing were electrophoresed on a 6% Long Ranger gel (FMC Bioproducts, Rockville, MD, USA) and analysed on an Applied Biosystems Model 373A automated DNA sequencer (Perkin-Elmer Corp.).
Sequences were compared against human mitochondrial DNA sequence, GENBANK accession #J01415 as well as a comprehensive mitochondrial databank, MITOMAP (1999).
Alonso A, Martin P, Albarran C, Aquilera B, Garcia O, Guzman A, Oliva H and Sancho M . 1997 Electrophoresis 18: 682–685
Ausubel F, Brent R, Kingston RE, Moore DD, Seidman JG, Smith JA and Struhl K ed . 1994–1998 Current Protocols in Molecular Biology John Wiley and Sons, Inc, USA
Bianchi MS, Bianchi NO and Bailliet G . 1995 Cytogenet Cell Genet 71: 99–103
Burgart LJ, Zheng J, Shu Q, Strickler JG and Shibata D . 1995 Am J Pathol 147: 1105–1111
Cavalli LR and Liang BC . 1998 Mutat Res 398: 19–26
Dhanda RK, van Orsouw NJ, Sigalas I, Eng C and Vijg J . 1998 Biotechniques 25: 664–668 670 672–675
Ebner D, Rodel G, Pavenstaedt I and Haferkamp O . 1991 Virchows Arch B Cell Pathol Incl Mol Pathol 60: 139–144
Green DR and Reed JC . 1998 Science 281: 1309–1312
Horton TM, Petros JA, Heddi A, Shoffner J, Kaufman AE, Graham Jr SD, Gramlich T and Wallace DC . 1996 Genes Chrom Cancer 15: 95–101
Johnson D, Tallini G, Silna D, Liang M and Wong L . 1996 Am J Pathol 149: 1786
Muller-Hocker J, Jacob U and Seibel P . 1998 Ultrastruct Pathol 22: 91–100
Oberley LW and Oberley TD . 1998 Mol Cell Biochem 84: 147–153
Polyak K, Li Y, Zhu H, Lengauer C, Willson JK, Markowitz SD, Trush MA, Kinzler KW and Vogelstein B . 1998 Nat Genet 20: 291–293
Sigel S and Castellan N . 1998 Nonparametric Statistics for the Behavioural Sciences, Second edn McGraw-Hill New York
Simon DK and Johns DR . 1999 Annu Rev Med 50: 111–127
Stefaneanu L and Tasca C . 1979 Endocrinologie 17: 233–239
Tallini G, Ladanyi M, Rosai J and Jhanwar SC . 1994 Cytogenet Cell Genet 66: 253–259
van Orsouw NJ, Zhang X, Wei JY, Johns DR and Vijg J . 1998 Genomics 52: 27–36
Wallace DC . 1992 Science 256: 628–632
Wallace DC . 1994 Proc Natl Acad Sci USA 91: 8739–8746
Wallace DC . 1999 Science 283: 1482–1488
Welter C, Kovacs G, Seitz G and Blin N . 1989 Genes Chromosomes Cancer 1: 79–82
JJ Yeh thanks James Becker for his continued support. The authors wish to acknowledge Jennifer B Kum, Sean McGrath and Wendy M Smith for technical advice. Partially funded by P30CA16058 from the National Cancer Institute (Ohio State University Comprehensive Cancer Center), a Boston University School of Medicine Surgical Research Fellowship (to JJ Yeh), and generous gifts from the Brown and Abrams families (to C Eng). PLM Dahia is a recipient of a postdoctoral fellowship from the Susan G Komen Breast Cancer Research Foundation (to C Eng).
About this article
Cite this article
Yeh, J., Lunetta, K., van Orsouw, N. et al. Somatic mitochondrial DNA (mtDNA) mutations in papillary thyroid carcinomas and differential mtDNA sequence variants in cases with thyroid tumours. Oncogene 19, 2060–2066 (2000) doi:10.1038/sj.onc.1203537
- thyroid tumours
- sequence variants
Detection of novel mitochondrial mutations in cytochrome C oxidase subunit 1 (COX1) in patients with familial adenomatous polyposis (FAP)
Clinical and Translational Oncology (2019)
Mitochondrial 10398A>G NADH-Dehydrogenase Subunit 3 of Complex I Is Frequently Altered in Intra-Axial Brain Tumors in Malaysia
Brain Tumor Research and Treatment (2018)
Journal of Clinical Investigation (2018)
Defining the momiome: Promiscuous information transfer by mobile mitochondria and the mitochondrial genome
Seminars in Cancer Biology (2017)