Molecular genetics of syndromic and non‐syndromic forms of parathyroid carcinoma

Abstract Parathyroid carcinoma (PC) may occur as part of a complex hereditary syndrome or an isolated (i.e., non‐syndromic) non‐hereditary (i.e., sporadic) endocrinopathy. Studies of hereditary and syndromic forms of PC, which include the hyperparathyroidism‐jaw tumor syndrome (HPT‐JT), multiple endocrine neoplasia types 1 and 2 (MEN1 and MEN2), and familial isolated primary hyperparathyroidism (FIHP), have revealed some genetic mechanisms underlying PC. Thus, cell division cycle 73 (CDC73) germline mutations cause HPT‐JT, and CDC73 mutations occur in 70% of sporadic PC, but in only ∼2% of parathyroid adenomas. Moreover, CDC73 germline mutations occur in 20%–40% of patients with sporadic PC and may reveal unrecognized HPT‐JT. This indicates that CDC73 mutations are major driver mutations in the etiology of PCs. However, there is no genotype–phenotype correlation and some CDC73 mutations (e.g., c.679_680insAG) have been reported in patients with sporadic PC, HPT‐JT, or FIHP. Other genes involved in sporadic PC include germline MEN1 and rearranged during transfection (RET) mutations and somatic alterations of the retinoblastoma 1 (RB1) and tumor protein P53 (TP53) genes, as well as epigenetic modifications including DNA methylation and histone modifications, and microRNA misregulation. This review summarizes the genetics and epigenetics of the familial syndromic and non‐syndromic (sporadic) forms of PC.


INTRODUCTION
Parathyroid carcinoma (PC) is a rare endocrine malignancy accounting for 0.005% of all cancers and <1% of primary hyperparathyroidism (pHPT) (Hundahl, Fleming, Fremgen, & Menck, 1999;Ruda, Hollenbeak, & Stack, 2005). Data from the Surveillance, Epidemiology, and End Results cancer registry showed a 60% increase in PC incidence from 1988 to 2003, which, in part, may be due to increased screening of serum calcium and an increased number of patients undergoing surgery for asymptomatic pHPT (Lee, Jarosek, Virnig, Evasovich, presence of macronuclei in many tumor cells, qualifies for a diagnosis of PC, whereas the presence of only one to three of these features, qualifies for a diagnosis of APA, which is considered to have features of carcinomas that lack unequivocal evidence for invasive growth (Bondeson, et al., 2004;Chan, 2013;DeLellis, 2011;Kumari, Chaudhary, Pradhan, Agarwal, & Krishnani, 2016). Indeed, using such clinicopathological criteria only ∼15%-35% of the prospectively diagnosed PC cases will continue to behave in a malignant manner, whereas ≤50% of PCs will have an initial diagnosis of benign disease (Gill, 2014;Marsh, Hahn, Howell, & Gill, 2007). Prospective diagnosis of PC is important, as cure can only be achieved following complete surgical resection.
Therefore, a broader understanding of the molecular and hereditary basis of PC would provide insight to improve pre-and post-surgical diagnosis and staging.
The aim of this review is to summarize the current knowledge of the molecular and hereditary basis of PC. A PubMed and EMBASE literature search was undertaken on July 1, 2016 using the search term: "parathyroid carcinoma", its free and controlled vocabulary EMTREE and MeSH synonyms, cross-referenced with "genetics," "epigenetics," "mutations," "CDC73," "hyperparathyroidism-jaw tumor," "familial isolated hyperparathyroidism," "MEN1," "MEN2," and its free and controlled vocabulary EMTREE and MeSH synonyms. There were no restrictions on language, publication type, or date. Additionally, reference lists from all major reviews were examined for citations that did not appear in the PubMed or EMBASE search.

CLINICAL FEATURES OF PARATHYROID CARCINOMA
PC may occur as part of a complex syndrome and hereditary disorder, or as a non-hereditary (i.e., sporadic) and isolated (i.e., non-syndromic) endocrinopathy (Table 1 and Figure 1). PC most commonly occurs as a sporadic non-syndromic disorder. The hereditary syndromes asso-ciated with PC include the hyperparathyroidism-jaw tumor (HPT-JT) syndrome, the multiple endocrine neoplasia (MEN) type 1 (MEN1) and type 2 (MEN2) syndromes, and potentially the non-syndromic familial isolated primary hyperparathyroidism (FIHP), which may be clinically difficult to distinguish from the MEN1 and HPT-JT syndromes ( Figure 1).
A palpable tumor is found in ∼50% of PC patients, whereas it is rarely identifiable in patients with benign pHPT (Holmes et al., 1969;Wynne et al., 1992). More than 90% of PC cases involve functioning tumors with plasma PTH concentrations 3-10 times higher than

Yes Yes No
No Patient with proven (or suspected) PC, including APA, PA or pHPT <35 years, recurrent pHPT, or multiglandular PA (Table 7) Obtain detailed family history to ascertain if first-degree relatives are affected and, if appropriate, screen asymptomatic relatives for hypercalacemia and/or parathyroid nodules. CDC73 germline mutation identified F I G U R E 1 A genetic testing approach to patients with parathyroid carcinoma. PC, parathyroid carcinoma; APA, atypical parathyroid adenoma; PA, parathyroid adenoma; pHPT, primary hyperparathyroidism; HPT-JT, hyperparathyroidism-jaw tumor; FIHP, familial isolated primary hyperparathyroidism; MEN1, multiple endocrine neoplasia type 1; MEN2, multiple endocrine neoplasia type 2; FHH, familial hypocalciuric hypercalcemia normal upper limit, whereas plasma PTH concentrations 2-3 times higher are typically found in benign pHPT (Holmes et al., 1969;Wynne et al., 1992). Recently, a population-based study reported a positive predictive value of >80% for PTH levels ≥10 times higher than the upper normal limit (Schaapveld et al., 2011). Most PC patients have severe hypercalcemia at presentation (calcium >14 mg/dl, i.e., >3.50 mmol/l), whereas in benign pHPT calcium levels are generally 1-2 mg/dl (i.e., 0.25-0.50 mmol/l) above normal (Wang & Gaz, 1985;Wynne et al., 1992;Chen et al., 2003). Plasma alkaline phosphatase activity is more commonly elevated in patients with PC than benign pHPT as a result of bone involvement (Silverberg et al., 1990;Chen et al., 2003). However, there is considerable overlap of these elevations of plasma calcium and PTH concentrations and alkaline phosphatase activity in patients with PC and PA, thereby making it difficult to rely upon them for establishing an unequivocal diagnosis of PC. However, PC patients have been reported to have elevated levels of urinary human chorionic gonadotropin subunits, particularly the hyperglycosylated isoforms, which are associated with an increased risk of hip fracture and death, and this difference from patients with benign pHPT requires further study (Rubin, Bilezikian, Birken, & Silverberg, 2008).

SYNDROMIC AND HEREDITARY FORMS OF PARATHYROID CARCINOMA
The syndromic and hereditary forms of PC are associated with germline mutations of the cell division cycle 73 (CDC73) gene, also referred to as the hyperparathyroidism type 2 (HRPT2) gene, MEN type 1 (MEN1), and rearranged during transfection (RET) genes (Table 1). The RET mutations, which are activating, are dominant at the cellular level, and only one copy of the mutated gene is required for tumor development. However, for the MEN1 and CDC73 mutations, which are inactivating and recessive at the cellular level, two mutations are required for a tumor to develop: for the hereditary tumors, these two recessive mutations comprise one germline and one somatic mutation that may involve a chromosomal loss and be detected as loss of heterozygosity (LOH) in the tumor. Such tumors may also occur sporadically, that is, without a family history and without inheritance of the germline mutation, and in these patients, both the recessive mutations will have likely occurred as somatic mutations in the tumor. This genetic model of neoplasia involving two recessive mutations in the development of tumors is known as Knudson's two-hit hypothesis. The genetic mechanisms involved in the etiology of the MEN1 and HPT-JT syndromes due to MEN1 and CDC73 mutations are consistent with Knudson's two-hit hypothesis (Knudson, 1971;Thakker, 1993).

Hyperparathyroidism-jaw tumor (HPT-JT)
HPT-JT (MIM# 145001) is a rare syndrome characterized by pHPT, fibro-osseous lesions (ossifying fibroma) of the mandible and maxilla, and tumors of the kidney and uterus (Jackson, 1958;Bradley et al., 2005b). Parathyroid tumors, of which 15% are carcinomas, are generally the first manifestation, and occur in >90% of HPT-JT TA B L E 2 Spectrum of diseases associated with CDC73 mutations

Hyperparathyroidism-jaw tumor
Diagnosis may be established in individuals with: • pHPT and ossifying fibroma(s) of the maxilla and/or mandible, or • pHPT and a direct relative with HPT-JT, or • Ossifying fibroma(s) of the maxilla and/or mandible and a direct relative with HPT-JT

CDC73-related familial isolated primary hyperparathyroidism
Diagnosis may be established in individuals with: • pHPT and a CDC73 germline pathogenic variant, and • At least one relative with pHPT, and • Absence of ossifying fibromas and exclusion of other causes of familial hyperparathyroidism

CDC73-related parathyroid carcinoma
Diagnosis may be established in individuals with: • PC and a CDC73 germline pathogenic variant Germline mutations often occur simultaneously with somatic mutations. PC is the most severe form of CDC73 associated diseases and may occur isolated or in the context of HPT-JT or FIHP pHPT, primary hyperparathyroidism; HPT-JT, hyperparathyroidism-jaw tumor; PC, parathyroid carcinoma; FIHP, familial isolated primary hyperparathyroidism.

CDC73
HPT-JT is an autosomal dominant disease due to germline mutations of the CDC73 gene (Tables 1-3). CDC73, which is comprised of 17 exons ( Figure 2A) and is located on chromosome 1q31.2, encodes the protein, parafibromin, which is associated in the polymerase associated factor  Yart et al., 2005). Functions attributed to parafibromin include the downregulation of cyclin D1 expression and direct interaction with -catenin resulting in the activation of transcription of target genes ( Figure 2C) (Woodard et al., 2005;Mosimann, Hausmann, & Basler, 2006;Zhang et al., 2006;Bradley et al., 2007). Parafibromin also has a role in embryonic development regulating genes involved in cell growth and survival ( Figure 2C) (Wang et al., 2008).
About 75% of HPT-JT patients will have germline CDC73 mutations within the coding region (Table 3 and Figure 2A Bradley et al. (2005b). h Reported as HPT-JT, but occurrence of jaw tumors, which may not always occur in HPT-JT patients, was not detected in any family members. i Reported in other publication as a possible FIHP case, but the frequent recurrence, presence of APA and renal and uterine tumors favors the diagnosis of HPT-JT . j Initially reported as FIHP by Masi et al. (2008). k Initially reported as FIHP by Howell et al. (2003). l It is possible this is a case of HPT-JT associated with PC since: the patient was diagnosed with three renal cysts, while "a maternal cousin had jaw pain and presumably bone destruction of the jaw, termed a 'hole in the jaw' ." Furthermore, histological description of the proband's parathyroid gland was consistent with an APA ("…vascular and capsular invasion, but no definitive features of PC were identified") and disease recurrence on the contra-lateral side (again with diagnosis of APA) suggests a more malignant behavior. m Reported as a germline mutation in a later publication, but inconsistency between patients' gender and age are observed . n Unclear if this kindred was included in the previous study of Howell et al. (2003). o For detailed information of the effect of CDC73 mutation on splicing please consult Hahn, McDonnell, and Marsh (2009 whole exon or gene deletions that are not readily detected by polymerase chain reaction (PCR) and sequencing, mutations in unidentified genes, or epigenetic modifications (Carpten et al., 2002;Cetani et al., 2004;Bradley et al., 2005b;. Approximately 55% of reported germline CDC73 mutations are associated with HPT-JT, and these comprise: 60% frameshift, 26% nonsense, and 3% loss of initiator methionine mutations that are predicted to result in parafibromin truncation or loss of protein transcription; 5% missense; and 5% splice site mutations (Newey, Bowl, Cranston, & Thakker, 2010 are missense mutations; 6% are reported from patients with sporadic adenomas, and of these 50% are missense and 50% are nonsense mutations; and 3% are reported from patients with sporadic ossifying fibromas of the jaw and all of these are frameshift insertion or deletions (Newey et al., 2010). Moreover, the same germline CDC73 mutations may be associated with HPT-JT, FIHP, and sporadic PC in different patients; for example, the c.679_680insAG, p.Arg227LysfsX31 mutation has been reported to occur in patients with HPT-JT (Figure 2 and   Table 3), FIHP ( Figure 2 and Table 4) and sporadic PC ( Figure 2 and Table 5), and the c131+1G > A mutation has been reported to occur in patients with HPT-JT and FIHP (Tables 3 and 4) (Carpten et al., 2002;Shattuck et al., 2003b;Cetani et al., 2004;Simonds et al., 2004;Bradley et al., 2005a;Newey et al., 2010). Thus, there is a lack of genotypephenotype correlation and the underlying mechanisms for this variability remain to be elucidated (Howell et al., 2003;Shattuck et al., 2003b;Thakker, 2016).

MEN1
MEN1 is an autosomal dominant disease, due to germline mutations of the MEN1 gene located on chromosome 11q13. Parafibromin also regulates cell growth, via cyclin D1 and Wnt signaling, and embryonic development via genes involved in cell growth and survival. H19, H19 fetal liver mRNA; IGF1 and IGF2, insulin-like growth factor 1 and 2; IGFBP4, insulin-like growth factor binding protein 4; HMGA1 and HMGA2, high mobility AT-hook 1 and 2; HMGCS2, 3-hydroxy-3-methylglutaryl-Coenzyme A synthase 2 and mothers against decapentaplegic homolog (mad, Drosophila) (SMAD) family members to inhibit transforming growth factor-b (TGF-b) and bone morphogenetic protein-2 (BMP-2) signaling, and forkhead transcription factor checkpoint suppressor 1 (CHES1) in an S-phase checkpoint pathway response to DNA damage; genome stability entails interactions with subunit of replication protein (RPA2) and Fanconi anemia complementation group D2 protein (FANCD2) that is involved in DNA repair; cell division includes interactions with nonmuscle myosin II-A heavy chain (NMHC II-A), glial fibrillary acidic protein (GFAP) and vimentin; and in proliferation, the reported interactions are with non-metastatic cells 1 protein (NME1) and activator of S-phase kinase (ASK) (Thakker, 2014 Bassett et al., 1998). To date >1,800 MEN1 mutations have been reported and ∼40% of these mutations are frameshift, followed by ∼25% nonsense, ∼20% missense mutations, and ∼10% splice site mutations (Lemos & Thakker, 2008; The Universal Mutation Database, 2017). Therefore, >70% of mutations are predicted to lead to truncated, and thus inactivated, forms of menin, with the majority of missense mutations resulting in the mutant menin being targeted to the proteasome, thereby reducing its ability to act as a tumor suppressor (Lemos & Thakker, 2008;Lemos et al., 2009). However, 5%-10% of MEN1 cases do not harbor mutations in the MEN1 gene (Bassett et al., 1998;Lemos & Thakker, 2008 e It is possible this is a case of HPT-JT associated with PC since: the patient was diagnosed with three renal cysts, while "a maternal cousin had jaw pain and presumably bone destruction of the jaw, termed a 'hole in the jaw' ." Furthermore, histological description of the proband's parathyroid gland was consistent with an APA ("…vascular and capsular invasion, but no definitive features of PC were identified") and disease recurrence on the contra-lateral side (again with diagnosis of APA) suggests a more malignant behavior. f All mutation carriers (n = 3) of this kindred developed PC. g Kindred originally reported by Williamson et al., and classified as HPT-JT by Carpten et al. and FIHP by Bradley et al., and associated with PC by Carpten et al., Bradley et al., and Yu et al. (Williamson et al., 1999;Carpten et al., 2002;Bradley et al., 2005a;Yu et al., 2015). h Reported as a FIHP family, but no information was provided on the pHPT status of the mutation carriers. i Additional clinical details about these kindreds are provided by Corbetta et al. (2010) and Vaira et al. (2012). j Studies reported by the same group, therefore it is not possible to exclude that equal mutations described in different publications are from the same proband/kindred. k Mutation was incorrectly reported in the original publication and was posteriorly updated by Warner et al. (2004) and Cardinal et al. (2005). l The authors collected 165 MEN1 mutations in patients with MEN1, but seven probands/kindreds exhibited FIHP phenotype (i.e., only pHPT) and were included here. m In a posterior publication, this mutation was identified by the same group in a kindred with MEN1 syndrome, and it is unclear if there were two different kindreds with the same mutation or if it was an update of the previous kindred (Cardinal et al., 2005). n Presence in the probands/kindreds of: rt renal cysts/lesions, and/or ut uterine tumors (if the presence of renal cysts or uterine tumors was unknown, one "?" was added next to the previous superscripts; hjt? was added if the absence of jaw tumors was unknown), and/or lip lipoma, and/or tn thyroid nodules, and/or hca hypercalcemia, and/or hcu hypercalciuria, and/or uccr urine calcium/creatinine clearance ratio < and ∼30% of patients, respectively (Howe, Norton, & Wells, 1993).
The RET gene encodes a receptor tyrosine-protein kinase involved in cell proliferation, neuronal navigation, cell migration, and cell differ-  (Table 6). All patients were men and all had PC metastasis at diagnosis. RET mutations were identified in two of these patients, and these comprised a c.1852T > C, p.Cys618Arg mutation, whose germline or somatic origin was not defined, and a germline c.1901G > A, p.Cys634Tyr mutation. The metastatic PC from the patient with the RET Cys634Tyr mutation had additional somatic genetic abnormalities involving LOH at loci from chromosomes 1, 2, 3p, 13q, and 16p (Jenkins et al., 1997).

CDC73
CDC73 mutations occur in 8% of FIHP patients (Pontikides et al., 2014). The majority of CDC73 germline mutations associated with FIHP are frameshift or nonsense, predicting premature truncation of parafibromin, thereby supporting a tumor suppressor function (Table 4 and Figure 2A and B).

CASR
The calcium sensing receptor (CASR) gene, locate on chromosome 3q13.33, encodes a G-protein coupled receptor that is predominantly expressed in the parathyroids and kidneys, where it respectively regulates PTH secretion and renal tubular calcium reabsorption appropriate to the prevailing calcium concentration (Thakker, 2004). The CaSR is also expressed in other tissues where its function remains to be elucidated (Thakker, 2004). CASR mutations occur in 2% of FIHP patients (Pontikides et al., 2014). To date 10 kindreds with CASR mutations associated with FIHP have been reported, and all of them had heterozygous CASR mutations that were predicted to be inactivating (Table 4). However, no PC case has been reported in any individual from these FIHP kindreds. FIHP patients with MEN1 and CASR mutations are generally younger and have multiglandular disease, whereas patients with CDC73 mutations have a disproportionally high prevalence of PC (Warner et al., 2004;Iacobone et al., 2007).

Other Genes
MEN1, CDC73, CASR, and GCM2 mutations may not be found in over 60% of FIHP patients (Pontikides et al., 2014). Interestingly, one study has reported a 1.7 Mb interval of significant genetic linkage for FIHP on chromosome 2p13.3-14, although conservative mutations involving the protein phosphatase 3 regulatory subunit B alpha (PPP3R1) and prokineticin receptor 1 (PROKR1) genes, which are in this interval, were not identified (Warner et al., 2006).
These will be reviewed.

CDC73
Approximately 40% of CDC73 mutations identified in patients with sporadic PC are germline mutations (Table 5 and Figures 2 and 3), and of these CDC73 mutations ∼65% occur in exons 1, 2, and 7, and the majority are frameshift or nonsense mutations, resulting in premature protein truncation and loss of protein function (Table 5) (Marsh et al., 2007;Newey et al., 2010). Moreover, a non-random gain of mutated CDC73 alleles has been reported in PC, and this suggests that aberrant CDC73 expression may also be important in the pathogenesis of PC (Yu et al., 2015). For example, a recent study reported that 4 of 22 (∼20%) PCs had a 3-5 copy number gain of mutant alleles, with three of these four PCs also having loss of the wild-type CDC73 allele through focal deletion or loss of the whole chromosome arm (Yu et al., 2015).
The identification of germline CDC73 mutations in patients with apparently sporadic PC is important, as it indicates that the patient and relatives are at risk of developing HPT-JT-associated tumors. Such germline CDC73 mutations are reported to occur in 20%-40% of patients with apparently sporadic PC, and somatic CDC73 mutations have been reported to occur in ∼40%-100% of apparently sporadic PCs (Table 5 and Figure 2) (Howell et al., 2003;Shattuck et al., 2003b;Cetani et al., 2004;Guarnieri et al., 2012). Moreover, LOH involving the CDC73 locus, on chromosome 1q31.2, is reported to occur in 50%-55% of sporadic PCs, and loss or reduced nuclear expression of parafibromin, detected by immunohistochemical (IHC) analysis, has been reported in >70% of PCs (Haven, van Puijenbroek . In contrast, germline CDC73 mutations were not found in patients with sporadic PAs, or in patients with hyperplastic parathyroids, and somatic CDC73 mutations and LOH of chromosome 1q have been reported to occur in <5% and <5%-10% of sporadic PAs, respectively (Carpten et al., 2002;Howell et al., 2003;Cetani et al., 2004;Krebs, Shattuck, & Arnold, 2005;Yip et al., 2008).
The absence of parafibromin nuclear staining, detected by IHC analysis, has been reported to occur in 15% of APAs, and <5%-20% of PAs, and it seems that the ability to distinguish between PC, APA, and PA using parafibromin IHC appears to be lower than CDC73 mutational analysis (Tan et al., 2004;Gill et al., 2006;Juhlin, et al., 2006;Cetani et al., 2007;Guarnieri et al., 2012;Cetani et al., 2013;Hu, Liao, Cao, Gao, & Zhao, 2016). However, one study has reported that the absence of parafibromin nuclear staining, detected by IHC analysis has a sensitivity of ∼70% and specificity of 95% for diagnosis of PCs (Hu et al., 2016). These findings indicate that CDC73 mutations are major driver mutations in the etiology of PCs.

MEN1
About 40%-50% of PCs have LOH of chromosome 11q, which is the location of the MEN1 gene, and >35% of PCs have combined LOH of 11q and 1q, which is the location of CDC73 (Figure 3). Combined LOH of 11q and 1q is rarely observed in PAs, and these findings suggest that MEN1 may be involved in PC pathogenesis (Dwight et al., 2000;Haven et al., 2004). In addition, somatic MEN1 mutations have been reported to occur in <15% PCs, in contrast to the higher frequencies of 35% and >45% of somatic MEN1 mutations and LOH involving chromosome 11 in sporadic PAs, respectively (Haven et al., 2007;Newey et al., 2012). Thus, the involvement of the MEN1 gene is likely to be a rare occurrence in PCs.

RB1
The retinoblastoma 1 (RB1) tumor suppressor gene, located on chromosome 13q14.2 encodes a protein (RB1) that is a negative regulator of the cell cycle. The active hypophosphorylated form of RB1 binds to the transcription factor E2 promoter binding factor 1 (E2F1) and leads to cell cycle arrest, whereas the phosphorylated form of RB1 allows dissociation from E2F1 and leads to transcription of E2F1 target genes that are involved in cell progression through G1 phase of the cell cycle (Asghar, Witkiewicz, Turner, & Knudsen, 2015). RB1 also maintains chromatin structure by stabilizing constitutive heterochromatin through stabilization of histone methylation (Gonzalo et al., 2005;Dyson, 2016). The RB1 gene has been implicated in the pathogenesis of PC, as allelic loss of RB1 has been observed in ∼30%-100% of PCs and decreased RB1 expression has been reported in >85% of PCs ( Figure 3) (Cryns et al., 1994b;Dotzenrath et al., 1996;Szijan et al., 2000). This contrasts with the low rate (i.e., <5%) of RB1 allelic loss in PAs, and no loss of RB1 expression (Cryns et al., 1994b). However, no RB1 somatic mutations have been identified in PCs, although RB1 allelic loss has been reported to be associated with PC recurrence and aggressive PA (Pearce et al., 1996;Shattuck et al., 2003a).

TP53
Tumor protein P53 (TP53), is a tumor suppressor gene, which is located on chromosome 17p13.1 and encodes a protein (p53) that is a transcription factor whose level and post-translational modification state are altered in response to cellular stress to induce growth arrest or apoptosis. Activated p53 suppresses cellular transformation by inducing growth arrest, apoptosis, DNA repair, and differentiation in damaged cells (Brosh & Rotter, 2009). TP53 allelic loss has been reported in 1 of 3 PCs studied (Figure 3), whereas TP53 overexpression has been observed in ∼10% of PAs (Cryns, Rubio, Thor, Louis, & Arnold, 1994a;Kishikawa et al., 1999). However, a somatic TP53 missense mutation (c.743G > A, p.Arg248Gln) has been reported in anaplastic PC cells, but not in differentiated PC cells, suggesting an association between this TP53 mutation and anaplastic transformation (Hakim & Levine, 1994;Tamura et al., 2009). The TP53 Arg248 residue is part of DNA binding domain (DBD) that interacts directly with the minor groove of DNA, and the p.Arg248Gln mutation is reported to result in the loss of DNA binding via the DBD (Ng et al., 2015). Interestingly, such TP53 mutations affecting Arg248 are reported to be present in ∼4% of all cancers (Petitjean et al., 2007).

CCND1
Cyclin D1 (CCND1), also known as parathyroid adenoma 1 (PRAD1), is an oncogene located on chromosome 11q13.3, that encodes cyclin D1, a 295-amino acid protein that is a component of the cyclin D1-cyclin-dependent kinase 4 (CDK4) complex that phosphorylates RB1 and thus inhibits the actions of RB1 in regulating G1/S transition (Arnold et al., 1992). Overexpression of cyclin D1 occurs in ∼65%-90% of PCs, but in <40% of PAs and ∼60% of parathyroid hyperplasia (Hsi, Zukerberg, Yang, & Arnold, 1996;Vasef, Brynes, Sturm, Bromley, & Robinson, 1999;Haven et al., 2004). The overexpression of cyclin D1 is associated with PC cell proliferation and a Ki-67 index of ≥5% . Overexpression of CCND1 gene may be associated with a 2-3 copy number gain of CCND1, which has been found to occur in five out of seven (∼70%) PCs (Figure 3), in contrast to the reported copy number gain of CCND1 in only three out of 14 (∼20%) PAs (Zhao et al., 2014). The increased CCND1 copy number in the PCs was associated with higher CCND1 mRNA levels and protein expression. However, the mechanisms linking CCND1 and PC tumorigenesis remain unknown.
One hypothesis is that the potent inhibition of CCND1 expression by CDC73 may be lost after "two hits" on the CDC73 gene, which may then trigger CCDN1 disinhibition and tumorigenesis (Woodard et al., 2005).
EZH2 represses, through histone modification H3K27me2/3, the tumor suppressor gene hypermethylated in cancer 1 (HIC1), which is involved in controlling growth of parathyroid cells and is reported to be decreased in PCs and PAs (Svedlund et al., 2012).

APC
Adenomatous polyposis coli (APC) is a tumor suppressor gene located on chromosome 5q22.2, that encodes a 2,843-amino acid protein, which inhibits canonical Wnt signaling by controlling -catenin ubiquitination and proteolysis. Loss of APC expression has been reported in PCs, although APC mutations and copy number changes have not been observed, thereby suggesting that APC may be involved in epigenetic mechanisms (Figure 3) (Juhlin et al., 2010;Svedlund et al., 2010;Andreasson et al., 2012;Newey et al., 2012;Yu et al., 2015).
Thus, APC expression is reported to be lost in 75% of PCs (Figure 3), but maintained in 100% of PAs (Juhlin et al., 2009). Quantitative real time PCR (qRT-PCR) and Western-blot analysis has also revealed that APC mRNA is either undetectable or very low, and that APC protein expression is undetectable in PCs (Svedlund et al., 2010). These alterations in APC expression in PCs may involve hypermethylation of the APC promoter 1A, and indeed methylation levels of APC promoter 1A CpGs were found to be significantly higher in PCs (>85%) than normal parathyroids (>15%); this was associated with decreased APC expression and accumulation of active nonphosphorylated -catenin (Svedlund et al., 2010). Moreover, treatment of PC cultured cells with the DNA methylation inhibitor 5-aza-2 ′ -deoxycytidine (decitabine) resulted in re-expression of APC mRNA, APC protein, and reduced cell viability, thereby suggesting that decitabine could be an additional option in the treatment of patients with recurrent or metastatic PC (Svedlund et al., 2010).

GSK3B
Glycogen synthase kinase 3 beta (GSK3B) protein expression has been reported to be lost in <35% of PCs and ∼5% of PAs (Juhlin et al., 2009). The GSK3B gene, located on chromosome 3q13.33, encodes a 420-amino acid enzyme regulating glycogen synthesis, Wnt, and PI3-kinase/AKT signaling pathways. However, loss of GSK3B expression was not associated with any increase of -catenin or cyclin D1 expression, thereby suggesting that GSK3B may act through a pathway different to the classical Wnt/ -catenin pathway in the etiology of PC (Juhlin et al., 2009). This would be consistent with results from several studies that have reported that abnormal nuclear expression ofcatenin is not a characteristic of PC (Semba, Kusumi, Moriya, & Sasano, 2000;Juhlin et al., 2009;Cetani et al., 2010).

PRUNE2
Prune homolog 2 (PRUNE2) germline and somatic mutations, comprising three missense mutations (one germline mutation in a PC without CDC73 mutations; two somatic mutations in two PCs without
Moreover, a somatic MLL2/KMT2D missense mutation (c.2522G > T, p.Cys841Phe) was reported in a PC, although in silico analysis has predicted that this is a likely tolerated/benign variant (Kasaian et al., 2013). Interestingly, menin and parafibromin, which are found to be mutated in such PCs, also interact with the histone methyltransferase SUV39H1 and function as transcription repressors by inducing H3K9 methylation (Rea et al., 2000;Yang et al., 2013). In addition, PCs

Role of microRNAs in parathyroid carcinoma
MicroRNAs (miRNAs) are small, 19-25 nucleotides long, non-coding RNAs, that function as negative regulators of gene expression by decreasing translation or increasing degradation of the target mRNA.

GENETIC TESTING IN CLINICAL PRACTICE FOR PATIENTS WITH SUSPECTED PARATHYROID CARCINOMA
Distinguishing between PC (malignant disease) and PA (benign disease) on the basis of clinical and histological features is difficult and frequently not possible. This is because there is considerable overlap in the clinical features, including elevations of plasma calcium and PTH concentrations, and alkaline phosphatase activity, between patients with PC and PA (Wang & Gaz, 1985;Silverberg et al., 1990;Wynne et al., 1992;Chen et al., 2003). Moreover, histological examination may also not be able to reliably distinguish between PC, APA (an intermediate category between PC and PA), and PA (Bondeson, et al., 2004;DeLellis, 2011;Chan, 2013;Kumari et al., 2016). For example, it has been reported that use of pathological criteria has been associated with ≥50% of PCs, which actually behaved in a malignant manner being initially considered to be benign, and <15% of PCs were successfully diagnosed prospectively (Gill, 2014). However, CDC73 mutational analysis and parafibromin immunostaining has been reported to be more reliable, with: CDC73 mutations being identified in >75% of PCs but in <1% of PAs (Shattuck et al., 2003b;Krebs et al., 2005;Gill, 2014); and loss of nuclear parafibromin immunostaining occurring in >95% of PCs, but in <1% of PAs (Tan et al., 2004;Gill et al., 2006;Meyer-Rochow et al., 2007). Moreover, even in the absence of a family history of PC or PAs, >30% of patients with PCs have a CDC73 germline mutation, thereby indicating they had an unrecognized HPT-JT syndrome or FIHP (Shattuck et al., 2003b;Cetani et al., 2004;Gill, 2014).
These observations indicate that parafibromin immunostaining is useful for diagnosis of PC, and that genetic testing for germline  (Heinrich, Blanke, Druker, & Corless, 2002;Willmore-Payne, Holden, Tripp, & Layfield, 2005); or B-Raf proto-oncogene, serine/threonine kinase (BRAF) for melanoma and papillary thyroid cancer (Davies et al., 2002;Kimura et al., 2003). However, targeted therapies are not available for PC and at present genetic testing for somatic CDC73 mutations using parathyroid tumor DNA may not be clinically useful for establishing the diagnosis or staging, especially as such tumors may contain multiple mutations. For example, whole exome sequence analysis of PCs and PAs have reported that the number of somatic mutations in these tumors vary between 3-176 and 2-110, respectively, and that <50% of these tumors may have MEN1 mutations (Cromer et al., 2012;Newey et al., 2012;Yu et al., 2015). Moreover, our analysis, that has compared the frequency of somatic CDC73 mutations in the catalogue of somatic mutations in cancer (COSMIC) database with the frequency of germline CDC73 mutations in the exome aggregation consortium (ExAC) database, has revealed that there are ∼65-fold more somatic non-synonymous mutations than germline CDC73 mutations. This increased frequency of somatic CDC73 mutations is similar to that occurring for the neurofibromin 1 (NF1) gene, another tumor suppressor, in which somatic nonsynonymous mutations were ∼80-fold more frequent than germline mutations. Recent studies have raised doubts about the pathogenicity of such mutations (Check Hayden, 2016;Lek, et al., 2016;Minikel, et al., 2016;Walsh et al., 2017), and while the clinical significance of such somatic mutations remains unknown, it would seem prudent at present, to reserve their investigation for research purposes only.
Genetic testing for germline CDC73 mutations may be helpful in clinical practice in several ways including: (1) confirmation of the high risk for developing PC and associated syndromic and hereditary forms of PC, so that appropriate screening for associated tumors (e.g., PC, uterine tumors, renal tumors) can be undertaken; (2) implementation of appropriate treatment (e.g., early parathyroidectomy for patients with HPT-JT because of increased risk of PC); (3) identification of family members who may be asymptomatic but harbor the mutation and therefore require screening for tumor detection and early treatment; and (4) identification of the 50% of family members who do not harbor the familial germline mutation and can therefore be relieved of the anxiety burden of developing tumors, while reducing the cost to the individuals and their children, and also to the health services in not having to undertake unnecessary biochemical and radiological investigations (Newey & Thakker, 2011;Thakker et al., 2012;Eastell et al., 2014;Thakker, 2016).
A genetic testing approach in a patient with a proven or suspected PC (Figure 1) could be as follows. The indications for undertaking such genetic testing in a patient are occurrence of: a proven or suspected sporadic PC or APA; a PA in association with an ossifying fibroma, early-onset uterine tumor, renal cysts or tumor, or endocrine tumor (e.g., pancreatic neuroendocrine or pituitary tumor); PA or pHPT occurring <35 years of age; recurrent pHPT, multiglandular parathyroid disease or hyperplasia or FIHP (Figure 1 and these disorders in a relative will help to guide decisions for appropriate germline mutational analysis of the MEN1, RET, CDC73, or CASR genes, with the results of the tests further guiding clinical management and treatments (Figure 1). In the absence of a family history, the diagnosis of sporadic PC should be considered, and germline mutational analysis of CDC73 undertaken as >30% of such patients will have a germline CDC73 mutation, and will therefore be at high risk of developing HPT-JT-associated tumors (Figure 1) Eastell et al., 2014;Wells et al., 2015;Thakker, 2016). The first degree relatives of these patients, even if asymptomatic, should also be offered tests for germline CDC73 mutations as these will help to identify if they have inherited the CDC73 mutation and are therefore at high risk of developing HPT-JT-associated tumors (Figure 1), or not inherited the CDC73 mutation in which case they can be reassured and have the burden of anxiety of developing PC and HPT-JT-associated tumors removed. Patients with sporadic PC who do not have CDC73 mutations, are likely to have another etiology for their disease and should be offered the opportunity of participating in research studies to elucidate the genetic abnormalities causing this rare disorder Eastell et al., 2014;Wells et al., 2015;Thakker, 2016).
Thus, identification of germline mutations would be helpful in the clinical management of PC patients and their families.

CONCLUSIONS
PC is a rare endocrine cancer, presenting typically with symptoms of hypercalcemia and predisposition to recurrence and metastasis. A definitive diagnosis of PC, which is usually based on histological analysis, is often only made retrospectively. Improvements in predicting the predisposition to PC and in diagnosis of PC are required, to facilitate improvements in patient care. To this end, molecular genetic studies have helped in identifying the underlying causes of PC and a genetic approach (Figure 1) can be helpful for the management of patients.
Molecular genetic studies have revealed CDC73 mutations to be major driver mutations in the etiology of PCs and defining and implementing clinical indications (Table 7) for CDC73 mutation analysis will aid in future management and counseling of patients at risk from PC and PCassociated syndromes such as HPT-JT. The genetic etiology causing PC involves other genes, which include MEN1, RET, and PRUNE2, as well as epigenetic mechanisms, alterations in miRNA expression and potentially as yet unidentified genes. PC is a rare neoplasm, and it is therefore essential that collaborative efforts that pool scarce tumor material and increase sample size are pursued to facilitate the identification of the genetic etiology of PC by next generation sequencing methodologies. These approaches are likely to yield important insights into the causative mechanisms for PC and to improved methods at detecting and diagnosing PCs, whose translation into the clinic are likely to lead to improved treatments and outcomes for patients.

DISCLOSURE STATEMENT
The authors declare no conflict of interest.