T₃ Induces Both Markers of Maturation and Aging in Pancreatic β-Cells

Thyroid hormones (THs) mediate development and aging processes in different tissues. They are known to be important in postnatal maturation of many tissues, including the central nervous system, gastrointestinal system, and inner ear (1–3). Yet, THs also affect aging: high levels of triiodothyronine (T₃) are associated with a shortened life span, and lowering the level can increase longevity in mammals and humans (4–8). Moreover, caloric restriction, a model known to delay aging across species, decreases T₃ levels. The mechanism behind this association is not completely understood but is thought a result from interaction between TH and metabolism, growth, and inflammation (rev. in 9). Understanding the molecular mechanisms underlying these opposing effects is crucial to predict and manipulate the various effects of this ubiquitous hormone and to effectively apply them translationally. Given the diverse T₃ effects on different tissues and even within the same tissue at different stages of development, a tight regulatory system that determines the specificity of these effects must be present. In addition to deiodinases and membrane transporters, TH receptors (THRs) are key players determining the specificity of T₃ effects, with the main isoforms, THRA1, THRA2, THRB1, and THRB2, differentially expressed across tissues and developmental stages.

The role of TH in the endocrine pancreas is only partially known. TH plays a part in the functional maturation of the pancreatic β-cells. At birth, rodent β-cells lack glucose-responsive insulin secretion and thus are functionally immature. Previously, we have shown that during postnatal development before weaning, the mRNA of β-cell transcription factor Mafa increases in expression in parallel to acquisition of glucose responsiveness (10) and that MAFA drove this maturation. Over this same time span, systemic TH increases and there is a switch in expression of its receptor isoforms THRA and THRB within the islet (11). We further showed that T₃ was a physiological inducer of MAFA (11) and stimulated MAFA expression and functional maturation in immature human islet tissue (12). These findings have been key to improvements in protocols used to differentiate human embryonic stem cells (hESCs) to glucose-responsive β-cells in vitro (13–15). A recent study showing that TH coordinates pancreatic islet maturation during zebrafish development suggest that the role of TH in the functional maturation of β-cells is preserved across species (16). Aging in pancreatic β-cells is not well understood; yet, p16^Ink4a, an alternative splice product of the Cdkn2a gene (17), is both a marker and effector of β-cell senescence (18,19). We recently showed a decline in function as β-cells age that is accompanied by increased expression of p16^Ink4a (20). However, correlations between T₃ and p16^Ink4a or any other aging marker have not been studied in pancreatic β-cells. It could be problematic for β-cell–replacement therapy if T₃ induced senescence in hESC-derived β-cells.

Given the translation potential of using TH during differentiation of human stem cells to β-cells for replacement therapy, our aims were to determine whether THs mediate aging of β-cells and which THR isoform was involved in the processes of maturation using Mafa as the effector and marker of glucose-induced insulin secretion (10,21) and p16^Ink4a as effector and marker of senescence (18). Herein, we show that T₃ induces both maturation and aging markers by increasing Mafa and p16^Ink4a in rodent primary islets, rodent β-cell–derived cell lines, and insulin-positive cells derived from hESCs. We show that this dual activation by T₃ is dependent on different isoforms of the THR with THRA mediating Cdkn2a (p16^Ink4a) gene transcription while THRB induces that of the Mafa gene. Thus, selective THRB agonists may further improve current stem cell differentiation protocols to produce functional β-cells for replacement therapies in diabetes.

Adult Sprague-Dawley rats (some with litters) from Taconic Farms (Germantown, NY) and adult C57Bl/6J mice from The Jackson Laboratory were kept under conventional conditions with access to water and food ad libitum. Pancreas was excised from anesthetized animals for islet isolation (22). Islets were cultured overnight in RPMI 1640 plus 10% FBS, followed by handpicking to ensure purity. For each postnatal day 7 (P7) sample, islets from 10 pups were pooled; each adult sample of either mice or rats was isolated from one animal, with three to six samples per age. Sex of animals was not recorded or taken into account in the analysis of results. The Joslin Institutional Animal Care and Use Committee approved all animal procedures.

Islets were cultured for 4 days in RPMI 1640 (11 mmol/L glucose and 10% charcoal-stripped [CS]-FBS) with/without T₃ (150 pmol/L T₃, equivalent to 7.5 pmol/L free T₃ in 10% CS-FBS) (23).

T₃ concentrations and incubation times were optimized to the specific model being used as well as the readout in which we were interested and vary across experimental models. The concentration used is specified in each model.

We followed the Viacyte surface-attached β-cell differentiation protocol (24,25) with the following modifications: 1) Matrigel-coated silicon rubber as the surface on which hESCs (CyT49) were plated, 2) stage 3 lengthened to 7 days, and 3) a 3-day stage 5 with DMEM-B27 media and 100 pmol/L T₃ or 1 μmol/L GC1. At the end of each stage, cells were harvested and RNA was extracted for real-time quantitative PCR (qPCR) analysis (12).

Total RNA was isolated with a PicoRNA extraction kit (Arcturus) or a QIAGEN kit and reverse transcribed (SuperScript reverse transcriptase; Invitrogen). Real-time qPCR used SYBR Green detection and specific primers (Supplementary Table 1). Samples were normalized to a control gene (S25 for rats, S18 for mice, and TATA binding protein [TBP] for hESCs), and the comparative CT (threshold cycle) method was used to calculate gene expression levels.

INS1 cells maintained in RPMI 1640 (11 mmol/L glucose plus 10% FCS, 10 mmol/L HEPES, 2 mmol/L L-glutamine, penicillin/streptomycin, 1 mmol/L sodium pyruvate, and 20 mmol/L β-mercaptoethanol) were switched to RPMI 1640 (1.6 mmol/L glucose plus 10% CS-FBS) 24 h before treatment and then incubated for 14 h with T₃ (150 pmol/L). Murine insulinoma (MIN)6 cells, maintained in DMEM (high glucose) (DMEM-H) supplemented with 15% FBS, were switched to DMEM-H with 15% CS-FBS and T₃ (10 nmol/L for 24 h, a dose proven to be effective in a dose response curve) and then harvested for chromatin immunoprecipitation (ChIP) or RNA.

We previously reported thyroid response element (TRE) motifs in mouse Mafa proximal gene and coding sequences (11). For the Cdkn2a gene, we used the TRANSFAC/MATCH database (BIOBASE) to analyze the 21 kb mouse Cdkn2a region including 1.5 kb 3′ and 5′ to the coding sequence. The primary criterion by which TREs were filtered was the matrix similarity score. Candidates were optimized by considering 1) proximity to a second putative TR- or retinoid X receptor–binding site, 2) position in the genome, 3) sequence homology to rat and human, and 4) presence in similar TREs using AliBaba. Primers were designed for five putative TREs (Supplementary Table 2).

Antibodies listed in Supplementary Table 3 were used with the Imprint chromatin immunoprecipitation kit (CHP1) (Sigma-Aldrich) following the manufacturer’s instructions. DNA from 250,000 cells was used for each condition in four independent experiments. Samples were analyzed by qPCR using specific primers for three putative TREs, S1, S2, and S3 in the Mafa gene as previously reported (11) and five putative TREs, S1, S2, S3, S4, and S5 for the Cdkn2a gene. The same ChIP samples were analyzed for both Cdkn2a and Mafa TREs.

The human THRA (hTHRA) transcript variant 1 was obtained from OriGene, and the insert was amplified with PCR and cloned into the Xhol and PmeI sites of the BirA recognition peptide-TEV cleavage site (Blrp-TEV) vector (kindly provided by Dr. Christopher Glass, University of California, San Diego). T4 DNA ligase (NEB) was used for ligation, and the ligation product was transferred to Escherichia coli DH5α-competent cells (NEB). Recombinant colonies were cultured overnight, and plasmid extraction was done by miniprep. Successful integration and direction were confirmed by sequencing.

MIN6 cells cultured in DMEM-H with 15% FBS and 5 μL/L β-mercaptoethanol were transiently transfected via lipofectamine with either empty Blrp-TEV vector (kindly provided along with the BirA-expressing plasmid by Dr. Christopher Glass), Blrp-TEV hTHRB1 (26), or Blrp-TEV hTHRA. For selection of the transfected cells, they were cultured for 8 days in regular MIN6 media supplemented with puromycin. Selected MIN6 cells were then transiently transfected with BirA plasmid and further selected with geneticin. At day 15 after the first transfection, biotinylated THRB1 or biotinylated THRA was precipitated using streptavidin-agarose beads as previously described (26). DNA was used in qPCRs with SYBR chemistry to detect enrichment of different potential TREs sites.

MIN6 cells were transfected with small interfering (si)RNA against mouse Thra, Thrb, or nonspecific (scrambled) siRNA (Thermo Scientific/Dharmacon, Lafayette, CO) using DharmaFECT following the manufacturer’s instructions (100 nmol/L siRNA and 2.5 μL DharmaFECT in 200,000 cells plated in 24-well plates). After 48 h culture in basal maintenance media, the cells were harvested for RNA and qPCR analysis. Knockdown was confirmed by real-time PCR. Results are presented as fold changes with respect to MIN6 cells treated with nonspecific siRNA.

We used activity of acidic β-galactosidase (β-gal) for in vivo evaluation of cellular senescence. We used a fluorescent substrate (ENZ-KIT 130-0010; Enzo Life Sciences) following the manufacturer’s instructions and quantified β-gal⁺ β-cells by FACS sorting for β-cells on the basis of higher endogenous fluorescence (27) and for β-gal activity as previously described (20). Substrate incubation time was optimized to 1 h at 37°C.

Data are shown as mean ± SEM. For statistical analysis, unpaired Student t tests were used to compare two groups and one-way ANOVA followed by Bonferroni post hoc test was used to compare more than two groups. When samples did not have normal distribution as assessed by a normality test, nonparametric statistics were used. A P value <0.05 was considered significant.

Similar to the T₃ induction of Mafa mRNA in rodent islets and cell lines (11), T₃ increased p16^Ink4a mRNA in cultured rat and mouse primary islets and rodent β-cell lines. p16^Ink4a expression increased 40% in mouse islets treated with T₃ compared with untreated islets and had a similar tendency in MIN6 cells (Fig. 1A). In both rat islets and the rat insulinoma cell line INS1, p16^Ink4a mRNA was fourfold increased after T₃ treatment (Fig. 1B). p19^Arf, the alternative splicing product of the Cdkn2a gene, was not increased in either mouse or rat islets after T₃ treatment (Supplementary Fig. 1A). Expression of p21, another important regulator of β-cell senescence, was tested in adult mouse islets treated with T₃, and its levels were unchanged by the hormone (Supplementary Fig. 1B). The expression of additional β-cell aging markers, Bambi and Igf1r, in response to T₃ was also unchanged (Supplementary Fig. 1C and D). The translational aspect of these findings was tested in hESCs differentiated to insulin-positive cells. The addition of 100 pmol/L T₃ during the last step of differentiation (3 days) increased MAFA expression sixfold (Fig. 1C) with respect to stem cells differentiated without T₃, suggesting the acceleration of maturation. Similarly, as in the rodent models, in these human cells T₃ induced highly variable increases of P16^INK4A (average 3.3-fold increase compared with cells without T₃) (Fig. 1D) and there was a direct correlation between the induction of MAFA and P16^INK4A (Fig. 1E) in the same samples. Thus, T₃ activates both β-cell maturation and senescence pathways in all three species.

TH exerts its effects through binding to THRs that belong to the family of nuclear receptors and bind to TREs in target genes. Upon ligand binding, these receptors upregulate or downregulate transcription of their targets. The presence of both isoforms of THR was verified by qPCR in both mice and rat primary islets (Supplementary Fig. 2) and in the species-specific β-cell lines: MIN6 for mice and INS1 for rats. Both cell lines had higher receptor levels than the primary islets of the same species. However, since different housekeeping genes were used for mouse (S18) and rat (S25) tissues, it is difficult to compare the absolute values between species. When the time courses of T₃ induction of Mafa and p16^Ink4a were studied in INS1 cells, Mafa mRNA expression was significantly upregulated at 6, 14, and 24 h after the addition of a low dose of T₃ (0.1 nmol/L) (Fig. 2A), whereas p16^Ink4a mRNA was significantly increased (4.3- to 5.8-fold) only after 24 h of exposure to higher doses of T₃ (1 and 10 nmol/L) (Fig. 2B). The induction of both Mafa and p16^Ink4a suggests they both contain functional TREs. We previously reported Mafa to be a direct target of TH with two TREs (sites 2 and 3) in the mouse Mafa gene (11), but p16^Ink4a having a TRE is novel. Cdkn2a (which encodes the p16^Ink4a gene) had several potential THR-binding sites with a 70–80% stringency (Fig. 3A and Supplementary Table 2). Using an antibody that recognizes both THR isoforms in ChIP assay, one (site 5) of the five potential TREs in Cdkn2a was significantly enriched (10-fold) compared with IgG negative control (Fig. 3B), supporting p16^Ink4a as a direct target of TH.

Activation of Mafa and p16^Ink4a by different TH receptor isoforms may provide the basis for simultaneous activation by T₃ of both maturation and senescence markers. To test the functionality of the identified THR binding to the Mafa and Cdkn2a TREs, we used THR isoform-specific agonists: CO23, a THRA1-specific agonist (28), and GC1, a THRB-specific agonist (29). P7 rat islets, previously shown to express both THRA and THRB (11), were cultured for 4 days in the presence of 150 pmol/L of either agonist and then analyzed for mRNA expression of maturation (Mafa) and senescence (p16^Ink4a) effectors. Mafa expression was significantly upregulated (60%) in the presence of GC1 (Fig. 4A), and p16^Ink4a was significantly upregulated (threefold) by CO23 (Fig. 4B) in the same set of samples. T₃ treatment of adult islets did not cause a further increase of the transcription factor, since physiological functional levels had already been reached (levels of T₃ adult rat islets were 0.92 ± 0.2 those of the control levels); therefore, P7 rat islets were used, since their Mafa levels were still lower than in the adult, and so the effects of T₃ on Mafa transcription could be evaluated. Similar effects of CO23 and GC1 on Mafa transcription were observed in INS1 cells (Supplementary Fig. 3), validating these results in both primary tissue and cell lines.

The two main isoforms of the THR, Thra and Thrb, share the same TRE motif but have different target genes based mainly on differential temporal and spatial expression; however, it is not known whether specific TR isoforms have gene-specific effects on transcription (30). Since the immunoprecipitation antibodies available are not specific to either isoform, we used a transient double transfection system of the biotinylated THRA and THRB1 (26) in MIN6 cells to identify which THR isoform bound to Mafa and Cdkn2a. In this system, biotinylated THRA or THRB1 is expressed after transfection of plasmids carrying 1) BirA construct that expresses the BirA enzyme that is able to specifically biotinylate the Blrp sequence and 2) the Blrp-TEV-THRA/B1 construct expressing THRA or B1 with a Blrp peptide sequence 5′ to the N-terminus, which is recognized and biotinylated by the BirA enzyme; the biotinylated receptor is expressed only when both plasmids are present. The receptor is pulled down with streptavidin-agarose resin, which binds biotin with high affinity, allowing identification of TREs that bind specifically to it in ChIP assay. With use of this system, site 5 of the Cdkn2a gene was significantly enriched with Blrp-TEV-THRA (Fig. 5A) but not with Blrp-TEV-THRB1 (Fig. 5B). Conversely, site 2 of the Mafa gene was significantly enriched (fourfold) with Blrp-TEV-THRB1 but not with Blrp-TEV-THRA (Fig. 5C). Site 3 of the Mafa gene was not enriched with either (Fig. 5D). These results indicate that THRA binds to the Cdkn2a gene, while THRB1 binds to the Mafa gene. The lack of enrichment of site 5 in the Cdkn2a gene by biotinylated THRB1 in the same pull-down samples that showed enrichment of Mafa site 2 demonstrates that the THRA isoform is the potential regulator of p16^Ink4a transcription via site 5. These ChIP studies are consistent with the agonist studies (Fig. 4) indicating that THRB activates Mafa transcription and THRA p16^Ink4a transcription.

The differential regulation of the THR isoforms was confirmed using a siRNA-based knockdown system in MIN6 cells. Both Thra (Fig. 6A) and Thrb (Fig. 6B) expression were reduced by 40–50%. Under basal conditions, p16^Ink4a mRNA was decreased 10% with siThra (Fig. 6C), but Mafa mRNA levels were unchanged with either siThra or siThrb (Fig. 6D). Igf1r mRNA, a β-cell aging marker that we have previously shown to be downregulated after p16^Ink4a knockdown (20), was significantly decreased with both siThra and siThrb (Fig. 6E).

Since T₃ increased expression of the senescence effector p16^Ink4a in islets, we evaluated whether T₃ upregulation of p16^ink4a expression induced the cellular senescence phenotype using senescence-associated acidic β-gal activity to identify and quantify senescent β-cells. After islets from 5-month-old mice were treated with T₃ for 72 h, the percentage of β-gal⁺ β-cells significantly increased (Fig. 7A). This increase in senescent β-cells underscores the physiological importance of T₃-induced p16^Ink4a expression and is in line with our previous observation of 4–7% β-gal⁺ Ins⁺ cells in islets obtained from 3 to 6 months of age (20).

To evaluate the translational value of the differential regulation by THR in human β-cells, we tested the effects of GC1 compared with T₃ in the final stages of differentiation of hESC. We previously published (12) that this differentiation protocol resulted in 17% of C-peptide–positive and 9% glucagon-positive cells under control conditions. Moreover, at the end of stage 4, 70% of cells expressed PDX1 protein and 24% expressed NKX6.1. By qPCR at this stage, the cells expressed insulin, PDX1, NKX6.1, and NEUROD1 mRNA but lacked MAFA expression. They also expressed THRA and THRB. At the end of stage 5 under control conditions, 16% expressed C-peptide, whereas 9% expressed glucagon. At the end of stage 5, T₃ treatment resulted in a 20-fold increase in Insulin mRNA and a fivefold increase in transcription factors MAFA and NKX6.1, twofold increase in CPE, and a tendency for increase in PCSK1 and PCSK2 mRNA. Scattered double hormone-positive cells were observed rarely. HESC cells treated with GC1 had significantly decreased P16^INK4A mRNA levels (Fig. 7B) with no consistent increase in MAFA. However, four of five samples had induction of MAFA or MAFB after treatment with GC1 (Supplementary Fig. 4). From a functional point of view, GC1 treatment of differentiated hESCs increased insulin secretion at 16.8 mmol/L glucose (Fig. 7C) and significantly increased insulin content (Fig. 7D) compared with those treated with T₃, which we believe resulted from an increase in the amount of insulin per cell as the β-cell matured, as we have previously shown (11,12). These findings underline the translational potential of our findings.

In summary, our results indicate that in pancreatic β-cells, T₃ upregulates the transcription of Mafa and p16^Ink4a genes through its interaction with different THR isoforms that bind directly to TREs. Through the two THR isoforms, TH simultaneously regulates both maturation and cellular senescence markers/effectors in β-cells, but a selective effect can be achieved using specific concentrations and timing or receptor isoform agonists.

While we had previously shown that TH was a physiological driver of β-cell maturation via its induction of the β-cell transcription factor Mafa, the current study suggests that TH is directly involved in β-cell aging via the induction of the senescence β-cell marker p16^Ink4a. Further, we have shown here that these dichotomous effects can occur simultaneously and that these opposing effects on β-cell physiology are due to T₃ binding to different thyroid receptor isoforms that selectively induce Mafa and p16^Ink4a gene expression. THRB1 binds to Mafa at site 2 TRE, while THRA binds to site 5 TRE of the Cdkn2a gene. These findings have increased importance, since the recent protocols for hESC-derived insulin-positive cells (13–15) have included TH to induce functional maturation. The simultaneous induction of the aging marker p16^Ink4a and potential acceleration of cellular senescence could be detrimental for the therapeutic use of hESC-derived β-cells. It would be ideal to induce the positive effects of MAFA upon β-cell function without inducing accelerated aging that might hamper the overall function or proliferative capacity of β-cells. Our studies provide a molecular basis to circumvent the increased p16^Ink4a expression by using specific THRB agonists that are able to selectively induce Mafa expression without inducing p16^Ink4a. Even though further studies are needed to fully translate these findings into hESC differentiation protocols, our data suggest that the use of specific THR agonists could optimize the final functional maturation step without compromising the ability of these cells to respond to growth factors. Additionally, different incubation times and concentrations might also favor one transcript over the other.

In vivo the differential effects of the THR isoforms are enhanced by the temporal differences in their expression, since a direct comparison of both isoforms from a single species has failed to demonstrate differences in T₃ binding properties (31). A switch of THR isoforms occurs in postnatal rat β-cell development (11), with predominance of THRA mRNA from P2 to P9, then similar levels of THRA and THRB until P15, and then a 2:1 dominance of THRB expression through adulthood. At the protein level, as reflected by immunostaining, β-cell nuclei were predominantly stained for THRA through P15, after which point nuclear THRB protein dominates, with lower levels of THRA. Although this isoform switch might initially seem contradictory to the proposed effects of each isoform, it is consistent with the physiological changes that accompany β-cell maturation. THRA, the THR isoform usually expressed in neonatal tissues, is expressed in β-cells from neonatal rodents. However, circulating T₃ levels are very low and do not induce p16^Ink4a at this early age. As circulating and local T₃ levels begin to increase as induced by changes in dediodinases expression (increase in Dio2 and a decrease in Dio3 expression), expression of THRB is induced (11) leading to increased β-cell Mafa and acquisition of glucose-responsive insulin secretion. In adult mature β-cells THRB expression dominates and maintains the high levels of Mafa, while low levels of THRA allow the senescence effector and mediator p16^Ink4a expression to be minimal. However, increasing T₃ further in adult β-cells does not further increase Mafa expression but can induce p16^Ink4a leading to senescence.

The protein biotinylation for ChIP was particularly useful for testing the isoform binding to the different TREs, since the commercial THR antibodies for immunoprecipitation are limited in their specificity for THR isoforms. This system has the advantage of providing an antibody-independent means to detect direct interaction between the receptors and TREs. Moreover, the high affinity of biotin-avidin interaction provided greater enrichment of target DNA, improved signal-to-noise ratio, and a more robust ChIP assay (32). With this system, we were able to identify specific THRB1 binding to site 2 TRE of Mafa and THRA to site 5 of Cdkn2a. The specificity of effect on these genes by receptor isoform was consistent with our findings of p16^Ink4a induction by a THRA functional agonist, CO23; decreased p16^Ink4a with Thra siRNA; and induction of Mafa mRNA both by THRB1 binding to its site 2 TRE and the functional THRB agonist GC1. A second TRE (site 3) within the Mafa gene that was immunoprecipitated with antibodies against both isoforms of THR (11) was not precipitated using the biotinylated system.

Even though changes induced by the siRNA knockdown of Thra and Thrb were minimal, we should point out that that only 40–50% of receptor knockdown was achieved in spite of our use of maximal doses of both DharmaFECT and siRNA. Difficulty with transfection is an inherent characteristic of these cells. Furthermore, it is difficult to directly translate these results from a cell line into a physiological setting, since immortalized cells may have changed the regulation of cell cycle inhibitors such as p16. Interestingly, Igf1r is likely a downstream target of p16, since knocking down p16 with siRNA in Min6 cells significantly reduced Igf1r levels but Igf1r knockdown did not influence p16 mRNA (20). Thus, a longer period of treatment with T₃ may be needed to see effects on these other aging markers.

There may be further complexity of the effects of T₃. T₃ was shown to mediate senescence and DNA damage in cultured mouse embryonic fibroblasts and in liver and kidney, but not heart, of hyperthyroid mice (33). In contrast to our finding that THRA directly induces p16^Ink4a mRNA, these authors focused on the role of THRB inducing the Cdkn2a pathway in the tissues with strong THRB expression. It is important to remember that the Cdkn2a encodes two different senescence effectors p16^Ink4a and p19^Arf; however, we saw no increased expression of p19^Arf in adult rodent islets.

The induction of p16^Ink4a for 10 days in β-cells of 3- to 4-week-old transgenic mice has been reported to result in increased function of β-cells (34), leading to the conclusion of a novel functional benefit of senescent β-cells. However, the β-cells of 5- to 6-week-old mice may not yet all be fully mature; islets from 3-week-old rats are glucose responsive but without the robustness of a 2- to 3-month-old rat (35). Additionally, in contrast to this recent report, we showed that β-cells that had high levels of p16^Ink4a expression had decreased glucose-induced insulin response (20). By understanding the roles of THRA and THRB in maturation and senescence of β-cells, we may now be able to discriminate between processes that are regulated during maturation and those of senescence.

The dichotomous role of T₃ upon maturation and aging is not exclusive to β-cells; it has been observed in other tissues and in different organisms (9). From an evolutionary point of view, it has been proposed that during early life, TH mediates growth and maturation, optimizing the health and reproductive chances of different organisms (9), partly by decreasing proliferation. However, as the organism ages, T₃ accelerates this process, impairing repair mechanisms and limiting longevity. Caloric restriction, which prolongs life span, has been shown to decrease circulating T₃ levels, providing additional evidence of the important role of T₃ on aging. The molecular mechanisms behind the aging effects of T₃ are thought to involve the IGF-1/insulin pathway, since low T₃ levels, such as those induced by caloric restriction, decrease circulating growth hormone levels with resultant decreased IGF-1 and insulin levels (9). In rats, the effects of T₃ on growth hormone are direct, since a TRE has been identified in the promoter region of growth hormone (36).

In conclusion, we have shown that T₃ induces both maturation and aging effectors through direct binding of different isoforms of THR to the key regulators of these processes, Mafa and p16^Ink4a. Since the effects are conserved across species and experimental models including humans, our results provide a molecular basis for development of specific highly targeted THRB-based strategies to promote β-cell maturity without accelerating aging.

Funding. This study was supported by grants from the National Institutes of Health (DK-093909 and DK-110390 [to S.B.-W.] and DK-036836 [Joslin Diabetes Research Center]) and by JDRF grant 1-2011-591 (to S.B.-W.). This work was also supported by the Diabetes Research and Wellness Foundation and an important group of private donors.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. C.A.-M. and S.B.-W. conceived the project and wrote the manuscript. C.A.-M., T.B.L., M.M., A.D., H.R., and P.R. researched data. T.S., A.M.Z., and P.R.L. provided THR agonists. A.H., C.C., and A.S. provided critical discussions and edited the manuscript. All authors reviewed the manuscript. S.B.-W. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the 74th Scientific Sessions of the American Diabetes Association, San Francisco, CA, 13–17 June 2014, and at the Keystone Frontiers in Islet Biology and Diabetes Symposia, Keystone, CO, 4–8 February 2018.