Reviewing Editor: Luca Persani

Conflict of interest declaration: None declared

Correspondence to:
Stefano Mariotti, MD, Professor of Endocrinology, Dipartimento di Scienze Mediche “M.
Aresu” SS 554 I-09042 Monserrato CA, Italy e-mail: mariotti@medicina.unica.it

ABSTRACT
It has been estimated that 40-60% of the variation of serum TSH levels is under genetic control. In keeping with this notion, polymorphisms of several genes potentially involved in the control of thyroid function have been linked to serum TSH concentrations. Genome-wide association scan (GWAS) is a powerful tool to simplify genetic analysis of complex traits and diseases. By genotyping >360,000 single nucleotide polymorphisms (SNP) in a large cohort of 4,300 Sardinian subjects, we recently identified a strong association (p = 1.3 x 10^-11) between alleles of the SNP rs4704397 and serum TSH. This association was confirmed in two genetically unrelated cohorts from Tuscany and the Old Order Amish and contributed to the 2.3% of the total variation of circulating TSH concentration. The rs4704397 SNP is located in intron 1 of the phosphodiesterase 8B (PDE8B) gene, encoding a highaffinity cAMP-specific phosphodiesterase abundantly expressed in thyroid tissue. This suggests that different PDE8B variants may modulate c-AMP-dependent thyroid hormone secretion and affect by feed-back pituitary TSH production. So far at least one independent study confirmed that the minor A allele of the rs4704397 SNP is associated with higher serum TSH in a cohort of pregnant women. In conclusion PDE8B is an important gene involved in controlling serum TSH concentration in normal individuals. Further studies are needed to ascertain whether and to what extent PDE8B may also represent a candidate gene for thyroid dysfunction and/or response to treatment.

Introduction
TSH secreted by anterior pituitary is the key regulator of thyroid function and its secretion is in turn strictly controlled through negative feed-back by circulating thyroid hormone concentrations. In the absence of hypothalamus-pituitary failure, serum TSH is therefore a sensitive indicator of thyroid function, with high and low levels indicating hypo- and hyperfunction of the thyroid gland. “Normal” reference range of serum TSH concentrations as derived from population studies carried out in euthyroid subjects without clinical, ultrasound or serological evidence of subtle underlying thyroid disease is generally referred between 0.4 and 4.0 mU/ml. However, recently both lower (2.5 mU/ml) (1) and higher (6-7 mU/ml) serum TSH concentrations (2) have been proposed as upper normal range values for adult young-to-middle aged and elderly (> 80 years) subjects, respectively.
Independently from the criteria employed to calculate TSH normal reference levels, it is clear that, together with serum thyroid hormone, in a given population intra-individual circulating TSH variability is comprised in a narrower range when compared to the general inter-individual variability (3). The precise mechanisms underlying the remarkable inter-individual variation of serum TSH have not been fully elucidated. It is believed that very mild thyroid dysfunction, together with other environmental factors such as diet, smoking and medication concur with the genetic background to determine the individual “set-point” of the hypothalamus-pituitary thyroid axis (4). As detailed in the next paragraph, evidence has been provided that polymorphisms of several key genes involved in thyroid function control may account for differences in serum TSH concentration. In this brief review, attention will be focused on the phosphodiesterase 8B (PDE8B) gene, recently identified as important controller of serum TSH concentration by genome-wide association scan (GWAS) on different unrelated populations in the context of the Progenia study coordinated by Silvia Naitza, Manuela Uda and Antonio Cao in Cagliari (Italy) and by David Schlessinger in Baltimore (U.S.A.).

Genetic control of serum TSH concentration: studies of candidate gene known to affect thyroid hormone secretion activity and metabolism
On the basis of studies carried out on twins, the heritability of serum TSH levels has been estimated between 40-65% (4). To identify the genes involved in the control of serum TSH several studies have been initially focused on candidate genes involved in different known thyroid hormone pattern.



As summarized in Table 1, the results obtained provided evidence that polymorphisms of TSH receptor, deiodinases, thyroid hormone transporters and thyroid hormone receptor do account for significant serum TSH variability.

TSH receptor gene (TSHR) polymorphisms
A common TSHR polymorphism (Asp727Glu) has been firstly associated to lower serum TSH levels in a small cohort of 156 healthy Dutch Caucasian blood donors (5), and this finding was confirmed in a subsequent study on 756 Dutch twin pairs (6). In the latter study, however, evidence was provided that the contribution of TSHR Asp727Glu to the genetic variability was very small. Quite recently, two single nucleotide polymorphisms (SNPs) in the promoter/enhancer region of the TSHR gene (rs10149689 G and rs12050077 A) were found associated with increased serum TSH (7). Interestingly, the allelic frequency of these 2 SNPs as well as that of the GA haplotype was significantly increased in Ashkenazi Jewish centenarians and in their offspring compared to controls (7), suggesting that in humans a heritable phenotype characterized by higher serum TSH is associated with longevity. Loss-of-function mutations in the TSHR gene are well known cause of Increased serum TSH and different degrees of hypothyroidism in single patients and in families (8), but recently two inactivating TSHR gene mutations were found to be responsible of the high frequency of hyperthyrotropinemia in an Israeli Arab-Muslim consanguineous community (9). It is therefore conceivable that unrecognized minor loss-of-function mutations may also contribute to the general variation of serum TSH levels.

Deiodinases genes (DIO1, DIO2, DIO3) polymorphisms
Several polymorphisms of the deiodinase genes have been found associated with different levels/ratios of circulating thyroid hormones, and to a lesser degree, TSH (10). Both type 1 (D1) and type 2 (D2) deiodinase activate the prohormone T4 by conversion to T3 through 5’-deiodination, while type 3 (D3) deiodinase acts as inactivating enzyme through 5-deiodination of T4 to inactive rT3. The gene most consistently associated to serum thyroid hormone concentration is DIO1, encoding for D1 protein. Two polymorphisms of the DIO1 gene (D1a-C/T and D1b-A/G) have been extensively studied (11-13): carriers of the T allele of D1a-C/T had higher serum rT3, lower T3 and lower T3/rT3 ratio (indicative of lower 5‘-deiodinating activity), while the D1b-G allele was associated with higher serum T3 and T3/rT3 ratio (indicative of higher 5‘-deiodinating activity) (13). A similar circulating thyroid hormone pattern suggestive of increased 5’-deiodinating activity has been found associated with Callele carries of a further DIO1 gene SNP (rs2235544 C/A (14). Increased serum T4 levels were found to be associated to a DIO2 gene polymorphism of (D2-ORFa-Asp3) in young but not in elderly (11) subjects, possibly as a consequence of the age-dependent decrease in the D2/D1 activity ratio in the 5’-deiodination process of T4 (12). None of the above polymorphisms were found to be associated with significant changes in serum TSH concentration. On the other hand, and more relevant to the purpose of the present review, a common DIO2 gene variant (Thr92Ala) has been found involved in determining the dose of L-T4 needed to normalize serum TSH in thyroidectomized patients, since Ala/Ala homozygous patients needed higher L-T4 dose as compared with patients carrying the Thr92 variant (15). Moreover, another common DIO2 SNP (rs225014), although not associated with circulating thyroid hormone level, appears to influence the clinical effectiveness of L-T4 therapy (16). In particular, the carriers of the rarer genotype CC showed impaired psychological well-being on L-T4 and enhanced response to combined L-T3/L-T4 therapy, which was not observed in patients carrying the more common TT genotype.
Thyroid hormone transporters (MCT8, MTC10, OATP1B1) gene polymorphisms
Polymorphisms of thyroid hormone transporters genes MCT8, MTC10, OATP1B1 have been studied in rather small populations (10). No effect on serum TSH and inconstant effects on circulating T3, T4, rT3 have been reported so far (17).
Thyroid hormone receptor gene polymorphisms
The consequences of polymorphisms in the thyroid hormone receptor (THR) alpha (THRA) and beta (THRB) genes on serum thyroid hormone and TSH levels have been thoroughly investigated (18). No association was found between SNPs and serum thyroid hormone (T4, T3 And rT3) concentrations, while only the intronic SNP THRB-in9-G/A was inconsistently associated to serum TSH levels, with higher concentrations found in carriers of the A/A genotype.
Taken together the above studies provide evidence that subtle differences in most of the key genes involved in thyroid hormone secretion, metabolism and action on peripheral tissues may be relevant in determining the individual “set-point” of the hypothalamus-pituitary-thyroid axis. Further insight on the mechanisms involved in determining circulating TSH level may be derived from genome-wide scan studies which may lead to the identification of genes unknown or not previously related to the control of thyroid function. The recent identification of PDE8B as a significant determinant of serum TSH levels is an example of the importance of this approach and shall be discussed in detail in the next paragraph.

Genetic control of serum TSH concentration: genome-wide scan studies and the identification of PDE8E as determinant of serum TSH levels.
The SardiNIA team of the Istituto di Neurogenetica e Neurofarmacologia, Consiglio Nazionale delle Ricerche of Cagliari, Italy in collaboration with Laboratory of Genetics, National Institute on Aging of Baltimore, USA has recently carried out several extensive genome-wide association scan (GWAS) studies in a large Sardinian cohort (Progenia study) in whom the founder-population structure can simplify genetic analyses of complex traits and diseases (19). This approach lead to identification of several genes associated with susceptibility to asthma, obesity-related traits, uric acid and lipid levels, height and severity of β-thalassemia (see (20) for detailed references). To evaluate the genes involved in serum TSH control, a total of 4,300 subjects selected from a sample of 6,148 individuals to represent the largest available families, were genotyped with the 500K Affymetrix mapping Array Set or the 10K Mapping Array System. A total of >360,000 SNPs were tested for association with serum TSH levels with an additive model (20). The results obtained, recently published in the American Journal of Human Genetics (20), revealed 3 SNPs (rs4704397, rs6885099 and rs2046045) with genome-wide significance (p< 10^-10 association) at a single chromosome 5 locus (Fig. 1). These three SNPs were all in strong linkage disequilibrium and lied in intron 1 or upstream the phosfodiesterase 8B (PDE8B) gene: among them the SNP showing the strongest association (rs4704397) explained 2.3% of the variance of TSH levels.
To confirm this association, three additional genetically unrelated cohorts were analyzed, one of 1,858 individuals enrolled in the Progenia study but unrelated to the first group, the second (1,164 subjects) form Tuscany (Italy) enrolled in the InCHIANTI study (21) and the third of 1,136 individuals from the founder population of the Old Order Amish (22).


Fig. 1. Results of Genome-wide Association Scan for TSH levels. For each marker the p value of the association is plotted and the position of PDE8B is annotated (From (20), with Publisher’s permission).

The results obtained confirmed the SNP association found in the original sample. In particular, in all populations studied, the A/A genotype of rs4704397 SNP was associated with higher median serum TSH when compared to A/G and G/G genotypes. Of particular interest the finding that this association was maintained either excluding or including subjects with evidence of associated thyroid disorders, representing a substantial proportion (1093/8479 = 12.8%) of the entire population studied. Moreover, although the difference did not reach the level of statistical significance, there was a trend toward a stronger effect of the A/A allele on serum TSH of subjects with underlying thyroid disorders. Although very preliminary, this finding suggest that PDEB8 gene could be included between the potential candidate markers for development and/or progression of thyroid dysfunctions (possibly through different expression of PDEB8 activity in the thyroid tissue.
Although the association of PDEB8 gene and TSH levels was very strong, this explained, as previously stated, only 2.3% of the general TSH variation, a small part of the total genetic component of this trait (~50%), which may be also explained by other genes, as briefly discussed in previous paragraphs. We therefore selected 24 candidate genes potentially involved in TSH secretion, activity and regulation of thyroid function and tested their association with TSH levels by GWAS analysis. Evidence of association was obtained for some genes such as THRB and TSHR, whose correlation with serum TSH levels has been shown in other studies (see previous paragraphs) and other candidate genes such as TG (encoding thyroglobulin, involved in thyroid hormone synthesis) and PDE4D and PDE7B (encoding for other phosphodiesterases involved in the inactivation of cAMP).

Hypothetic mechanisms linking PDE8B to serum TSH levels.
The PDE8B gene encodes a high-affinity phosphodiesterase catalyzing the hydrolysis and inactivation of cAMP (Fig.2) (23).

Fig. 2. Intracellular signalling pathways following activation of membrane-bound and soluble cyclases. Adenylate and guanylate cyclases generate cyclic adenonosine monophasphate (cAMP) form ATP and cyclic guanosine monophophate (cGMP) form GTP which acts as second messengers inducing cellular responses. Phosphodiesterases (PDE) in turn inactivate cAPM and cGMP in inactive AMP and GMP, respectively. To date, eleven different PDE families including several isoforms and splice variants are known. Some PDE families are specific for cAMP (PDE4, PDE7 and PDE8, in red), other are specific for cGPM (PDE5, PDE5 and PDE9, in blue) and several (PDE1, PDE2, PDE3, PDE10 and PDE11, in violet) show dual specificity.

Because PDE8B is undetectable in the pituitary (24), we infer that it could act primarily in the thyroid by inactivating cAMP produced after TSH stimulation. Indeed, of the 5 major isoforms of PDE8B, the major isoform PDE8B1 and minor isoforms PDE8B2 and PDE8B3 are abundantly expressed in the thyroid (23,25,26). PDE8B could therefore influence serum TSH levels through its effect on TSHdependent thyroid hormone synthesis and secretion (Fig. 3). Interestingly, in our study some evidence of association with serum TSH levels was provided for other genes such as PDE4D, PDE7B (both specific for cAMP (23)) and PDE10A (a phosphodiesterase with dual specificity for cyclic nucleotides and stimulated by cAMP (23)). Further studies, however, involving differential transfection of thyroid cells with different phopshodiesterase variants, are needed to confirm their actual contribution to intracellular cAMP levels and to thyroid functional activity.


Fig. 3. Hypothetical link between PDEE8 and TSH levels. PDE8B activity may affect intrathyroidal cAMP concentrations and subseqent thyroid hormone secretion which in turn, via hypothalamic-pituitary negative feedback, modulate TSH production

If the above hypothesis will provide to be correct, it should also be considered that mutations of PDE8B gene could be responsible of some cases of inappropriately high serum TSH in the absence of thyroid autoimmunity or loss-of-function mutations in THRB or TSHR genes.
Several studies already support the biologic relevance of PDE8B activity in human diseases. A PDE8B mutation leading to increased intracellular cAMP has been described pituitary adenomas (24) and in a case of Cushing’s syndrome with bilateral hyperplasia (27). Increased cAMP-degradating PDE8B activity has also been found in autonomous hyperfunctioning thyroid adenomas and interpreted as a compensatory mechanism to the constitutively activated cAMT pathway typical of these tumours (28). PDE8B upregulation has also been reported in Alzheimer’s brain (29) and PDE8B activity has been linked to insulin secretion (30).
Finally, our finding that other cAMP-specific phosphodiesterase genes (PDE4D and PDE7B) are also significantly linked to serum TSH levels strongly supports the concept that intrathyroidal cAMP concentration represent a key point in the regulation of hypothalamic-pituitary-thyroid axis.

Other studies linking PDE8B to serum TSH
To our knowledge, only one aditional study has been published so far on the relationship between PDE8B gene and serum TSH levels (31). In this paper, the PDE8B SNP rs4704397 (showing the strongest association in our study) was evaluated in a large cohort of 877 pregnant women at 28 week of gestation. The results obtained confirmed the involvement of this SNP in the control of serum TSH concentration; in particular, the AA genotype was associated with the higher median serum TSH concentration (2.16) while lower levels (1.84 and 1.73) were found in the AG and GG genotypes, respectively (p=0.004). No association was found between rs4704397 SNP and FT4, FT3, or anti-thyroid peroxidise (anti-TPO) autoantibody, as well as with offspring birth weight or gestational age at the delivery. This study has important practical implications since the current guidelines (32) for treating subclinical hypothyroidism in pregnant women are still based on a fixed cut-off since specific serum TSH reference ranges taking into account the main genetic contributions to individual variation which is always narrower than that of the general population.

Conclusions
A recent GWAS provided compelling evidence for a role of PDE8B gene (encoding a phopshodiesterase specifically catalyzing hydrolysis of cAMP) in the control of circulating TSH concentration and this finding has been reproduced in a cohort of pregnant women. This phenomenon is probably mediated through different degradation activities of intrathyroidal cAMP, leading to different thyroid hormone secretion rates. In keeping with this concept, other cAMP-specific phopshodiesterase genes(PDE4D and PDE7B) have also been found by GWAS associated to serum TSH levels. Further studies are needed to ascertain whether and to what extent PDE8B may also represent a candidate gene for thyroid dysfunction and/or response to treatment.