Single parathyroid adenomas, which cause approximately 85% of the cases of primary hyperparathyroidism 1 are well-differentiated, benign, clonal tumors, which produce hypercalcemia through excessive secretion of parathyroid hormone (PTH). Activation of the cyclin D1 oncogene and inactivation of the MEN1 tumor suppressor gene are established pathogenetic contributors, but understanding of molecular pathogenesis in this disease remains incomplete. The mitochondrial genome has been identified as a possible target for somatic mutations that may promote tumorigenesis 2345678910. Mitochondrial abnormalities, including changes in structure, number, and respiratory enzyme components and transport systems, have been observed in many cancers. Homoplasmic somatic mutations of the mitochondrial genome have also been identified in many tumor types 2345678910. However, the pathogenetic significance of the mutations detected in those tissues is unclear.

The human parathyroid is an intrinsically low-turnover tissue 11, as such, benign tumors arising from this tissue have relatively low proliferative rates. The mitochondrial genome is an especially attractive potential target for mutations that could drive tumorigenesis in such a tissue. Mitochondrial DNA replicates frequently and independently of the nuclear genome 12, and over many years the potential for enhanced DNA damage from locally generated reactive oxygen species (ROS) may further enhance the accumulation of somatic mutations in the mitochondrial genome.

In many endocrine organs, especially the parathyroid glands, with age, increasing numbers of oxyphil cells appear 1314. The oxyphil cells of normal parathyroid glands have a characteristic eosinophilic granular cytoplasm (Figure 1) which on electron microscopic examination is filled with abnormally shaped mitochondria 1516. In contrast to the chief cells, the oxyphil cells in normal glands lack the organelles associated with protein synthesis and secretion, and lack immunohistochemically identifiable chromogranin and PTH 16. The majority of parathyroid adenomas are composed predominately of chief cells (Figure 1), however there is a subset of hyperfunctioning parathyroid adenomas that contain greater than 90% oxyphil cells and are classified as oxyphil adenomas These tumors, in addition to having a cytoplasm filled with abnormal mitochondria, contain the organelles of protein synthesis and secretion and contain histochemically identifiable PTH. While the molecular basis for this phenotypic change is unknown, it has been hypothesized that mitochondria may accumulate in oxyphil cells due to abnormalities of the mitochondrial genome17, supported by the finding of increased numbers of oxyphil cells in normal glands with increasing age, analogous to increased mtDNA damage that occurs with aging 16.

We therefore sought acquired somatic mutations of the mitochondrial genome in both typical chief cell parathyroid adenomas and oxyphil cell parathyroid adenomas using an especially rigorous approach, and compared these findings with the mitochondrial genomes of normal and hyperplastic parathyroid glands.

A total of 393 homoplasmic sequence variants were identified (8–45 variants per sample) as compared to the standard Anderson mtDNA sequence 19. Of the 393 variants, 375 were single base pair substitutions, 13 were single base pair insertions or deletions (4 deletions and 9 insertions), 2 were 2 base pair deletions, 2 were 2 base pair insertions and 1 was an 8 base pair deletion. Evaluation utilizing the mitochondrial database, Mitomap 18, indicated that of the 393 variants, 117 were in non-coding regions, 244 were in protein coding genes and 32 were predicted to affect rRNA or tRNA. Novel sequence variants are shown in Table 1.

Eleven somatic mutations were identified in 6 of 18 chief cell adenomas (Table 2 and Figure 2). Two tumors contained mutations that were predicted to result in early stop codons in either the ND4 or ND6 genes, both subunits of mitochondrial Complex I. Three tumors contained mutations that were predicted to result in non-synonymous changes in subunits of either Complex I or Complex III. Two tumors contained mutations only in the highly polymorphic, but non-coding D-loop region.

Sixteen somatic mutations were identified in 9 of 12 oxyphil adenomas. Four tumors contained mutations that were predicted to result in early stop codons in either the ND1, ND4 or ND5 genes, all subunits of mitochondrial Complex I. Two tumors contained non-synonymous changes in Complex I subunits, one tumor contained a mutation in CO1 not predicted to change the amino acid sequence and one tumor contained a mutation affecting a tRNA gene. One tumor contained a D-loop mutation only. All mutations appeared to be homoplasmic, with the exception of one, which appeared to be heteroplasmic, a G to A transition at position 559, within the D-loop region. Homoplasmy refers to the presence of a particular sequence variant or mutation in virtually every copy of the mitochondrial genome within a single tumor cell sample (and within the limits of sensitivity of the sequencing technology), while heteroplasmy refers to the presence of more than one such detectable allele 20. Heteroplasmy within a tumor would be indistinguishable from homoplasmy in a sub-population of the tumor's cells, since each tumor's cells are being analyzed collectively rather than on an individual cell-by-cell basis.

Mutations affecting Complex I genes, which represent 38.2% of the mitochondrial genome, accounted for 55.6% of all mutations and 22.2% of mutations occurred in the highly variable D-loop region.

To determine if the observed trend towards increased numbers of likely or potentially functionally significant mtDNA mutations in oxyphil adenomas as compared to chief cell adenomas was statistically significant, Fisher's Exact test was performed yielding a significant p-value of 0.02.

No somatic mutations occurred in any of the hyperplastic parathyroid glands. The one normal parathyroid for which paired thyroid tissue was available also contained no somatic mutations. No clustering of germline sequence variants occurred in any of the samples studied.

Mitochondrial abnormalities have been observed in many cancers and previous studies have demonstrated altered energy metabolism in cancer cells 2345678910. Homoplasmic somatic mutations of the mitochondrial genome have been identified in various tumor types. The mitochondrial genome is an especially attractive potential target for mutations that could drive tumorigenesis in a low turnover tissue such as the human parathyroid, given the frequent replication of mitochondrial DNA, independent of the nuclear genome and the enhanced potential for mitochondrial DNA damage by ROS. One might expect an increased potential for accumulation of somatic mutations in the mitochondrial genome in such low replicative tissues as the parathyroid over many years. Our data provide initial support for these hypotheses, and suggest that other slow-growing tumors arising from low-replication tissues should be similarly analyzed.

Mitochondrial DNA mutations in parathyroid adenomas tend to affect Complex I genes more often than other tumor types and more often then expected by chance alone. Additionally, D-loop mutations appear to be underrepresented in parathyroid adenomas compared to other tumor types. In contrast to many other studies, this study rigorously examines the entire mitochondrial genome in matched tumor/normal pairs from the same individual while many analogous studies of other types of tumors focused primarily on coding regions and/or the D-loop region or used population rather than control DNA from the same patient. Due to the wide range of mutation frequencies in the literature and discrepancies in study design, statistical analyses between our data and those from other related studies cannot generally be performed. The importance of observed differences between parathyroid adenomas and other tumor types remains unclear.

Our results raise the possibility that NADH dehydrogenase subunit genes might play a key role in tumorigenesis, perhaps especially in low-replication tissues. Few studies have directly examined this issue. The contribution of specific mtDNA mutations, affecting the ATP synthase subunit 6 2122, to tumorigenesis has been examined, however little is known about the potential pathogenetic significance of the majority of mtDNA mutations identified to date. The contribution of specific germline sequence variants as predisposition alleles to specific tumor types has been suggested, however the majority of germline mtDNA sequence variants appear unlikely to directly contribute to tumorigenesis because of the frequency of such changes in individuals without tumors. Muller-Hocker, et al. examined respiratory chain function in normal parathyroid glands and hyperplastic parathyroids in uremic patients23, finding defects in Complex III and IV in both normal and diseased glands. They also reported that chief cells in 20% of uremia-associated (secondary) hyperplastic parathyroid glands had defects of the respiratory chain, and that heart and muscle cells had more respiratory chain defects than parathyroid oxyphil cells. Together with reports of respiratory chain defects in normal, adenomatous, and hyperplastic parathyroid glands, the preferential clustering within NADH dehydrogenase subunit genes of the somatic mitochondrial mutations in our series of adenomas reinforces the intriguing suspicion that these mutations may be playing an integral role in parathyroid tumorigenesis. Given that mitochondria have recently been implicated in the regulation of intracellular calcium homeostasis 2425 and apoptosis 26, and that calcium sensing and signaling are crucial to parathyroid function and regulated PTH secretion, the possible relationship between calcium homeostasis and mitochondrial complex I enzyme function deserves further scrutiny.

Our data also provide additional support for the hypothesis that mitochondria accumulate in oxyphil cells due to abnormalities of the mitochondrial genome. Previous studies of Hurthle cell 27 and other oncocytic tumors 28 have demonstrated an association between a large deletion of mtDNA (Δ4977) and oncocytic/oxyphilic tumors in comparison to non-oncocytic tumors of the same tissue type. However, given its presence in both normal and tumor cells of various tissues, the potential pathogenetic significance of such a finding remains unclear. As such, the presence or absence of this deletion was not addressed by our experimental approach and was considered beyond the scope of this manuscript. Functioning oxyphil cell parathyroid adenomas are more likely to contain somatic mtDNA mutations than chief cell adenomas and the somatic mutations in oxyphil adenomas are more likely to be functionally significant. These differences between oxyphil and chief cell adenomas are statistically significant with a p-value of 0.02 (Fisher's Exact Test). Our results show marked similarity to the recent findings in thyroid tumors by Gasparre et al that disruptive mutations (nonsense and frameshift mutations) in Complex I genes are associated with an oncocytic phenotype 29. It is therefore quite plausible that these molecular genetic lesions may make key contributions to determining the oxyphilic phenotype, as mitochondria may increase their replication rate as a compensatory mechanism for altered energy metabolism due to mtDNA mutations. Further studies of normal oxyphil cells that could shed additional light on this hypothesis should be encouraged.

The author(s) declare that they have no competing interests.

JC-G participated in the design of the study, carried out molecular genetic studies and sequence analyses, and drafted the manuscript. TT participated in the design of the study and carried out molecular genetic studies and sequence analyses. SIR performed histological analyses for all tumor samples. AA participated in the design and coordination of the study, data analysis and interpretation. All authors participated in revising drafts of the manuscript, and all authors read and approved the final manuscript.