Age-Dependent Changes in Glucose Homeostasis in Male Deiodinase Type 2 Knockout Zebrafish

Abstract

Thyroid hormones (THs) are crucial regulators of glucose metabolism and insulin sensitivity. Moreover, inactivating mutations in type 2 deiodinase (DIO2), the major TH-activating enzyme, have been associated with type 2 diabetes mellitus in both humans and mice. We studied the link between Dio2 deficiency and glucose homeostasis in fasted males of two different Dio2 knockout (KO) zebrafish lines. Young adult Dio2KO zebrafish (6 to 9 months) were hyperglycemic. Both insulin and glucagon expression were increased, whereas β and α cell numbers in the main pancreatic islet were similar to those in wild-types. Insulin receptor expression in skeletal muscle was decreased at 6 months, accompanied by a strong downregulation of hexokinase and pyruvate kinase expression. Blood glucose levels in Dio2KO zebrafish, however, normalized around 1 year of age. Older mutants (18 to 24 months) were normoglycemic, and increased insulin and glucagon expression was accompanied by a prominent increase in pancreatic islet size and β and α cell numbers. Older Dio2KO zebrafish also showed strongly decreased expression of glucagon receptors in the gastrointestinal system as well as decreased expression of glucose transporters GLUT2 and GLUT12, glucose-6-phosphatase, and glycogen synthase 2. This study shows that Dio2KO zebrafish suffer from transient hyperglycemia, which is counteracted with increasing age by a prominent hyperplasia of the endocrine pancreas together with decreases in hepatic glucagon sensitivity and intestinal glucose uptake. Further research on the mechanisms allowing compensation in older Dio2KO zebrafish may help to identify new therapeutic targets for (TH deficiency–related) hyperglycemia.

Thyroid hormones (THs) are essential determinants of vertebrate glucose metabolism and insulin sensitivity. In synergism with the pancreatic hormones insulin and glucagon, they contribute to a correct glucose homeostasis (1, 2). Both hyperthyroidism and hypothyroidism are known to induce insulin resistance and poor glycemic control in humans, and there is an increased prevalence of type 2 diabetes mellitus (T2D) in patients with thyroid disorders (3–5). Moreover, an activity-reducing mutation in human deiodinase type 2 (DIO2), the major TH-activating enzyme throughout the body, has been associated with increased fasting glucose and insulin levels, insulin resistance, and increased susceptibility for T2D (6, 7).

Different animal models are being used to unravel the complex link between THs, glucose homeostasis, and T2D. Over the years, studies in rodents showed that THs modulate glucose homeostasis at multiple levels. They are needed for pancreatic β cell development but can impair mature β cell function. Furthermore, they increase glucose uptake in muscle, mainly via glucose transporter (GLUT) 4, and increase hepatic glucose production via gluconeogenesis (8). Moreover, DIO2-deficient mice show signs of T2D: they are insulin resistant, have increased fasting blood glucose levels, and are more prone to diet-induced obesity (9).

More recently, nonmammalian models also made their entry in diabetes research. Indeed, a comparative genomics study in zebrafish, Xenopus, chickens, mice, and humans revealed that the vast majority of carbohydrate/glucose metabolic genes are conserved in all of these species (10). As such, zebrafish quickly became a popular model to study metabolic diseases, including T2D (11–14). However, the link between THs and glucose homeostasis in this model has received limited attention so far. As in rodents, THs are important for zebrafish pancreatic islet development and their maturation during the larval-to-juvenile transition (15, 16). Treatment with T₃ was found to stimulate β cell differentiation and insulin production and to inhibit α cell maturation and glucagon production in both larvae and adults, whereas thyroid gland ablation had the opposite effect (16).

A weakness in many of the studies is the duration of the observation period used. Studying glucose homeostasis at different ages may, however, be important, as a recent report (17) suggested that young adult zebrafish (4 to 7 months) seem to acclimate better to increased glucose exposure than do older fish (>1 year). In this study, we have used two Dio2 knockout (KO) zebrafish lines, showing a pronounced and persistent T₃ deficiency in all tissues tested (18, 19), to study the long-term effect of Dio2 KO/hypothyroidism on glucose homeostasis in depth. Our results confirm a central role for THs in zebrafish glucose homeostasis. Interestingly, older Dio2KO zebrafish gradually mount a compensatory hyperplasia of the endocrine pancreas combined with decreases in glucagon receptors and glucose transporters in liver and gut, allowing a return to normoglycemia after a prolonged period of hyperglycemia at a younger age.

Materials and Methods

Zebrafish husbandry

We used wild-type (WT, AB line) and two Dio2KO zebrafish lines. Both mutants completely lack Dio2 activity due to a 9-bp deletion (deletion mutant) or a 4-bp insertion (insertion mutant) in the dio2 gene upstream of the selenocysteine codon, an essential element of the enzyme’s active site (18). Fish were housed in circulating freshwater thermostated aquaria (27.5 ± 0.5°C) under a 14-hour light/10-hour dark photoperiod. They were fed with formulated feed (morning) and Artemia salina larvae (afternoon). Taking into account the well-known sex effect on glucose homeostasis in mammals (20, 21), experiments were conducted exclusively on males. We used young adult (6 months) and old adult (18 months) WT and Dio2KO fish unless specified otherwise. They were first- or second-generation offspring out of siblings. All fish were fasted for 16 hours prior to each experiment or sampling unless specified otherwise, and they were weighed and measured prior to sampling. All animal tests were approved by the Institutional Ethical Committee of KU Leuven (P091/2017) and executed in accordance with the European Council Directive (2010/63/EU).

Blood glucose levels

Blood glucose levels were measured in fasted fish (unless stated otherwise) following euthanasia in 0.1% tricaine. The tail was cut with a scalpel and blood glucose was determined by placing a glucometer test strip (OneTouch Verio; LifeScan Inc., Beerse, Belgium) directly on blood emerging from the docked tail.

RNA isolation and real-time quantitative PCR

The zebrafish pancreas consists of four pancreatic lobes in which one large principal islet and several smaller and diffuse islets are scattered (22). To be sure to include the entire pancreas we chose to dissect the gastrointestinal (GI) system as a whole (i.e., stomach, liver, intestine, and pancreas with part of the surrounding visceral adipose tissue) as has been done before [e.g., (23)]. The GI system and skeletal muscle tissue (tail region) were dissected from euthanized WT and Dio2KO fish (n = 4 to 10 per genotype and age) and snap-frozen on dry ice. Total RNA was isolated using TRIzol reagent (Invitrogen, Merelbeke, Belgium) according to the manufacturer’s guidelines. Subsequent treatment with DNase I (Invitrogen) excluded genomic DNA interference. Two micrograms of total RNA was used for first-strand cDNA synthesis with SuperScript III reverse transcription and oligo(dT)12–18 primers (Invitrogen), following the manufacturer’s guidelines. Target mRNA was quantified using SYBR Green–based real-time quantitative PCR (RT-qPCR) for proinsulin and proglucagon (ins, gcga, and gcgb); two glucose transporters, GLUT2 and GLUT12 (slc2a2 and slc2a12); two enzymes belonging to the gluconeogenesis pathway, phosphoenolpyruvate carboxykinase and glucose-6-phosphatase (pck1 and g6pca.1); two enzymes involved in the glycolysis pathway, hexokinase and pyruvate kinase (hk1 and pkmb); one enzyme essential for hepatic glycogenesis, glycogen synthase 2 (gys2); the predominant TH-inactivating type 3 deiodinase (dio3b); and the TH-responsive Krüppel-like factor 9 (klf9). Probe-based RT-qPCR was carried out for low-abundance insulin and glucagon receptor mRNA (insra, insrb, gcgra, and gcgrb). All primer pairs and probes are listed in an online repository (24). RT-qPCR was performed in a CFX96 Touch™ real-time PCR detection system (Bio-Rad Laboratories, Temse, Belgium) in a total volume of 10 µL, containing 2.5 µL of diluted cDNA (GI system 1:50 for ins, gcga, gcgb, gcgrb, pck1, g6pca.1, and gys2; GI system/muscle 1:500 for slc2a2, slc2a12, gcgra, and klf9; muscle 1:10 for insra, insrb, gcgra, gcgrb, hk1, pkmb, and dio3b; and GI system 1:10 for dio3b). The reaction mixture contained 1× Maxima Probe/ROX qPCR master mix (Thermo Fisher Scientific, Merelbeke, Belgium) and 250 nM (SYBR Green–based)/500 nM (probe based) forward and reverse primers. The thermal cycling profiles for pancreatic genes, glucose transporters, metabolic enzymes, dio3b, and klf9 included 30 seconds of initial degradation at 95°C, followed by 50 cycles of 15 seconds at 95°C and 30 seconds (or 1 minute for transporters) at 60°C (55°C for slc2a2 in muscle). A similar protocol was used for the receptors, with the exception of more cycles (60 cycles) and an annealing step at 62°C for gcgra. Primer pairs were verified for single peak melting curve, and water was used as a negative control. All samples were analyzed in duplicate. Standard curves were prepared from a 1:5 dilution series (for pancreatic genes, glucose transporters, and enzymes of gluconeogenesis) or 1:2 dilution series (glucose receptors and enzymes of glycolysis) of pooled WT and Dio2KO cDNA. Three housekeeping genes were selected for normalization (25): eukaryotic translation elongation factor 1-α1, like 1 (eef1a1l1), ribosomal protein L13a (rpl13a), and eukaryotic translation initiation factor 1B (eif1b).

Histology

After euthanasia in 0.1% tricaine, the trunk area (between gills and anus) was dissected and fixed overnight at 4°C in 4% phosphate-buffered paraformaldehyde. After 24 hours, samples (n = 3 to 8 per genotype) were rinsed with PBS and transferred to a solution of 30% sucrose in PBS at 4°C. Upon saturation, trunks were embedded in Tissue-Tek^® and cryosections (12 µm) were cut and mounted on glass slides for immunohistochemistry (IHC). For IHC the primary antibodies polyclonal guinea pig anti-insulin (26) (1:200) and mouse anti-human glucagon (27) (1:1000) were used to stain the pancreatic β and α cells, respectively. Thawed slides were permeabilized in PBS/0.5% Triton X-100, after which antigen retrieval was carried out in the oven [20 minutes at 95°C in 10 mM sodium citrate, 0.05% Tween 20 (pH 6.0)]. After blocking [2 hours in PBS/1% BSA/1% dimethyl sulfoxide/0.2% Triton X-100 for insulin and 1 hour in donkey serum (1:5 diluted in Tris-NaCl–blocking buffer) for glucagon], sections were incubated with primary antibodies overnight. The next day, they were incubated 2 hours with secondary antisera conjugated with Alexa Fluor 488 [1:200 goat anti–guinea pig (28) for insulin and 1:200 donkey anti-mouse for glucagon (29)]. Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI; 1:1000 in PBS) and sections were mounted with Mowiol and glass coverslips. All steps were conducted in a humidified chamber at room temperature with gentle agitation. Images were taken with an Olympus FV-1000 confocal microscope. For each fish, the total number of pancreatic sections containing the largest, principal islet was determined. Subsequently, three sections evenly distributed across its rostral-caudal axis (one fourth, one half, and three fourths of the total range) were analyzed using the Fiji image processing software ImageJ. Islet surface was used as a proxy for islet size. The average of surface measurement and cell counts of these three sections/fish was used for further statistics.

To estimate hepatic glycogen storage, coronal cryosections of the abdominal region were stained using the periodic acid–Schiff (PAS) stain kit (395B-1KT; Sigma-Aldrich, Overijse, Belgium) (30). Staining was performed according to the manufacturer’s guidelines but with omission of hematoxylin staining to facilitate quantification of the PAS signal. Slides were then dehydrated by passing them through a series of increasing ethanol concentrations (50% to 100%) and 100% xylol and were mounted with DPX. Representative images were taken of two different lobes of the liver per section with a Zeiss Imager Z1 microscope and this was done for two sections per animal. At least six animals per condition were included. A relative value for PAS-positive signal intensity was obtained by converting the original magenta-colored images to a 16-bit image displaying gray values using ImageJ software, after which the mean gray value was determined within a standard 10-inch² box based on the method described by Nguyen et al. (31). The average value was calculated for each animal and used for further statistics.

Metabolic assay

The metabolic assay based on energy expenditure relied on the conformational change of the water-soluble sodium salt resazurin into the fluorescent resorufin via metabolic reduction by NADH₂ (32). Fish (n = 5 to 7 per genotype) were placed individually in square glass containers in 70 mL of resazurin solution (0.002 mg/mL tank water). Containers without fish served as blanks. During the test, fish remained in the zebrafish facility (normal temperature and circadian cycle) and could move freely but they were not fed. Water samples were taken at 48 hours and stored at −20°C until fluorescence measurement. Analysis was performed as described previously (32) and fluorescence intensity was normalized to the weight of the individual fish (expressed per gram).

Statistical analysis

RT-qPCR data were analyzed using qbase+ software (Biogazelle, Gent, Belgium). Statistical analysis on all results was performed using GraphPad Prism 8 software (GraphPad Software, San Diego, CA). Relative values from RT-qPCR and metabolism measurements were ln transformed before analysis. Data were analyzed by two-way ANOVA for effects of genotype, age, and genotype–age interaction. A Sidak multiple comparisons post hoc test was used to indicate differences between WT and Dio2KO animals at a given age. A P value <0.05 was considered statistically significant for all analyses.

Results

A summary of the two-way ANOVAs for all measurements is shown in Table 1.

Effects on body weight and blood glucose levels

Previous studies showed that after an initial growth delay, Dio2KO animals catch up and by 4 months of age male Dio2KO fish have a similar body weight as WT fish (18). This was confirmed in the current study where all fish were weighed and measured prior to sampling, and no significant differences in body weight or in body mass index were found between mutants and corresponding WT animals across all experiments and at any age (data not shown).

Fish typically have lower blood glucose levels than do other vertebrates (33). Average levels in our fasted WT zebrafish were ∼50 mg/dL, which is within the range reported in several other zebrafish studies [e.g., (11, 34)]. Measurements in the insertion mutant showed significantly increased blood glucose levels at 6 and 9 months, which had returned to WT levels at 18 and 24 months (Fig. 1A). Analysis of the deletion mutant allowed to confirm a similar phenotype and to show that blood glucose levels gradually declined and eventually normalized between 11 and 15 months of age (Fig. 1B).

We additionally measured blood glucose levels in fed fish of 9 months of age. One hour after feeding, Dio2KO fish (insertion mutant) showed significantly increased blood glucose levels (124 ± 11 mg/dL, n = 12) compared with corresponding WT fish (69 ± 6 mg/dL, n = 10) (unpaired Student t test, P = 0.0004), confirming their hyperglycemia also in a fed status.

Effects on (pro)insulin and (pro)glucagon expression

To determine the potential causes for the transient hyperglycemia, we measured changes in expression profiles of proinsulin and proglucagon, precursors of the two major hormones controlling the glucose balance in circulation. Expression was measured in the entire GI system to include all proinsulin- and proglucagon-producing cells from the diffuse pancreatic islets as well as enteroendocrine cells. In contrast to proinsulin, proglucagon can give rise to multiple functional peptides. Zebrafish gcga encodes glucagon, glucagon-like peptide (GLP)-1, and GLP-2 whereas gcgb encodes glucagon and GLP-1 (35). Importantly and different from the situation in mammals, GLP-1 in zebrafish (and teleosts in general) is not produced in the intestine but in the pancreatic islet cells and has an activity similar to glucagon in the regulation of blood glucose levels (36).

(Pro)insulin (ins) mRNA levels were highly upregulated in Dio2KO insertion mutants at 6, 9, 18, and 24 months (Fig. 2A). Overall (pro)glucagon a (gcga) and (pro)glucagon b (gcgb) expression was also upregulated, although the differences between Dio2KO and WT animals in the Sidak post hoc test following two-way ANOVA were often not significant due to relatively large SEMs (Fig. 2B and 2C). Analysis of the deletion mutant at 6 and 18 months showed a similar profile with in this case a strong and significant upregulation of both (pro)insulin and (pro)glucagon expression (Fig. 2D–2F), again confirming the identical phenotype.

Owing to the limited number of fish available as a result of poor reproduction (19) and based on the fact that both mutants clearly showed the same phenotype, we further focused on one young and one older stage, 6 months and 18 months, respectively, and performed experiments on either the insertion line (RT-qPCR analyses) or the deletion line (IHC and metabolic assays).

Effects on insulin- and glucagon-producing cell numbers

To find out whether the increased ins and gcg expression was accompanied by an increase in the number of insulin/glucagon producing cells, we performed IHC staining on cryosections of 6-month and 18-month WT and Dio2KO fish of the deletion mutant line (Fig. 3A–3H). For each animal, we measured the size of the principal islet (using glucagon-positive cells to determine islet borders), total cell number and density (based on DAPI staining), and number of insulin- and glucagon-positive cells. None of these parameters differed significantly between WT and Dio2KO fish at 6 months (Fig. 3I–3L). However, at 18 months both islet size (Fig. 3I) and total cell number were strongly increased in Dio2KO fish compared with WT fish (3.5-fold and 3.2-fold, respectively). Cell density in the islet was unaffected, pointing to cell hyperplasia rather than hypertrophy (Fig. 3J). Both insulin-positive and glucagon-positive cell numbers were highly increased in mutant compared with WT fish (Fig. 3K and 3L). The relative increase in Dio2KO mutants at 18 months compared with 6 months was however higher for β cells (×1.95) than for α cells (×1.46).

Effects on insulin/glucagon receptor expression

We measured mRNA levels of insulin receptors (insra, insrb) in skeletal muscle and glucagon receptors (gcgra, gcgrb) in the GI system and skeletal muscle, using the insertion mutant line. The insra expression was significantly downregulated in muscle of Dio2KO compared with WT fish at 6 months. At 18 months both insra and insrb expression in muscle tended to be lower in Dio2KO fish (Fig. 4A and 4B). Expression of glucagon receptors in muscle and the GI system was not different between genotypes at 6 months, but at 18 months gcgrb in muscle and both gcgra and gcgrb in the GI system were clearly downregulated in Dio2KO fish (Fig. 4C–4F).

Effects on glucose transporter expression

Other crucial players in the regulation of glucose homeostasis are the glucose transporters. Based on literature data we chose to investigate GLUT2 (slc2a2) and GLUT12 (slc2a12). In humans, increased hepatic glucose efflux and reduced muscle glucose influx, by GLUT2 and GLUT4, respectively, play an important role in the loss of glycemic control in T2D and insulin resistance (37). Zebrafish slc2a2 is structurally and functionally very well conserved and is highly expressed in intestine, liver, and muscle (38, 39). GLUT4, the main insulin-regulated glucose transporter in humans, is not present in zebrafish but its function is taken over by GLUT12, which is present in almost all tissues (39, 40). We found no significant changes in slc2a2 or slc2a12 levels in skeletal muscle of insertion mutants compared with WT fish at either age (Fig. 5A and 5B). In contrast, expression of both transporters was significantly decreased in the GI system of mutants at 18 months (Fig. 5C and 5D).

Effects on glycolysis, gluconeogenesis, and glycogenesis

Glycolysis and gluconeogenesis are two important biochemical processes in the control of glucose homeostasis by respectively catalyzing the breakdown and generation of glucose. The process of glycolysis mainly occurs in muscle tissue, where hexokinase (hk1) and pyruvate kinase (pkmb) catalyze the first and last step, respectively. Gluconeogenesis is largely accomplished in the liver, where phosphoenolpyruvate carboxykinase (pck1) catalyzes the first (rate-limiting) step and glucose-6-phosphatase (g6pca.1) the final one. Hepatic glycogenesis, catalyzed by glycogen synthase 2 (gys2), also contributes to the regulation of circulating glucose by providing temporary carbohydrate storage.

In skeletal muscle, we found a strong downregulation of hk1 and pkmb in Dio2KO fish of the insertion line at 6 months whereas expression levels were similar to WT fish at 18 months (Fig. 6A and 6B). No differences were observed for pck1 at either age in GI system samples (including the liver), whereas g6pca.1 was significantly downregulated in mutants at 18 months only (Fig. 6C and 6D). Hepatic gys2 expression was higher in Dio2KO fish at 6 months but lower at 18 months. This latter decrease was accompanied by a clear decrease in PAS staining intensity, indicating decreased glycogen content (Fig. 6E and 6F).

Effects on metabolic rate

Insulin resistance and T2D belong to the group of metabolic diseases, and THs are important regulators of metabolic rate. Therefore, we finally estimated metabolic rate in fish from the deletion mutant line using a noninvasive test measuring energy expenditure in freely swimming animals based on reduction of resazurin by NADH₂ (32). Fish were fasted for the duration of the experiment. Energy expenditure was significantly lower in mutants at 6 months but had returned to WT levels at 18 months (Fig. 7). This normalization occurred despite the persistent T₃ deficiency as indicated by a pronounced decrease in expression of the TH-inactivating enzyme dio3b and the TH-responsive gene klf9 in GI system and skeletal muscle in mutants at 18 months (data not shown).

Discussion

THs are known to modulate glucose homeostasis at multiple levels, and their actions are influenced by factors such as developmental stage and feeding status (8). Both hyperthyroidism and hypothyroidism have been associated with complications in insulin signaling, glucose intolerance, and metabolic diseases such as T2D, but the interactions between THs, hyperglycemia, and insulin resistance remain elusive (4, 5, 41). Moreover, the consequences of hyperthyroidism have been described in more detail than those of hypothyroidism where observations are often quite divergent (3, 42). We used Dio2KO zebrafish to investigate the link between THs and glucose homeostasis in more detail in male 16-hour-fasted animals. Earlier studies have shown that these mutants experience T₃ deficiency in all tested tissues, including the GI system and skeletal muscle [(18, 19) and data not shown]. Our results confirm several of the findings obtained in hypothyroid/DIO2-deficient mammals but at the same time reveal some interesting differences showing compensatory mechanisms allowing older Dio2KO zebrafish to counteract dysglycemia.

Effects of Dio2 deficiency on glucose homeostasis in young adults

THs have a role in morphological and functional maturation of the endocrine pancreas as shown in both rodents and zebrafish (15, 16). Nevertheless, by the time our T₃-deficient Dio2KO mutants reached adulthood (shown at 6 months of age), both pancreatic islet size and cell density were normal and the number of β and α cells were similar in Dio2KO and WT fish. However, expression of both ins and gcg (gcga and gcgb) were increased in Dio2KO fish and they were clearly hyperglycemic both in a fasted (shown at 6 and 9 months) and a fed state (shown at 9 months). These findings are partially in line with those of another zebrafish study (16) where researchers found that ablation of the thyroid gland in larvae at 20 days postfertilization did not change pancreatic islet morphology by 30 days postfertilization but increased gcg expression and resulted in hyperglycemia. Ablation of the thyroid gland in adult zebrafish (age not specified) yielded similar results when tested after 3 days (16). However, in contrast to our findings, their treatment decreased ins expression at both stages (16). It is difficult to define exactly why the effect on ins expression differs between both studies, but one possible factor is the duration of T₃ deficiency. In the early stages of diabetes in mammals, β cells show a compensatory response to hyperglycemia by increasing insulin production (43) as found in our mutants, which have been T₃ deficient for months. The short period of lack of TH production (3 to 10 days) and resulting hyperglycemia in the other study (16) may not have been sufficient to show this type of compensatory response. The increase in ins expression in our Dio2KO zebrafish is also in line with a number of studies in humans with the activity-reducing Thr92Ala polymorphism in DIO2 showing fasting hyperinsulinemia (6, 7, 44), as well as with observations in DIO2KO mice (9).

At 6 months of age the higher insulin production in mutants was apparently insufficient to restore normoglycemia. Although increased glucagon production may be at least partially responsible, several other factors can lead to hyperglycemia despite the presence of high insulin levels. It has been suggested in humans that hyperthyroidism leads to insulin resistance predominantly at the level of gluconeogenesis in liver whereas hypothyroidism leads to insulin resistance at the level of glucose utilization in peripheral tissues such as skeletal muscle (45). Unless accompanied by a concomitant decrease in hepatic gluconeogenesis (46), this can lead to hyperglycemia. We therefore measured expression of insulin receptors in skeletal muscle and of glucagon receptors and the insulin-regulated glucose transporters GLUT2 (slc2a2) and GLUT12 (slc2a12) in skeletal muscle and the GI system (including liver, intestine, pancreas, and part of the visceral adipose tissue). The only significant difference at 6 months was a decrease in insra expression in skeletal muscle. However, this decrease was accompanied by a strong reduction in muscle hexokinase (hk1) and pyruvate kinase (pkmb) expression, pointing toward a clear disruption in glycolysis and hence muscle glucose utilization. The changes at the level of skeletal muscle agree with the fact that hypothyroidism in rats is associated with muscle insulin resistance (47) and also fits with the observation that 6-month-old Dio2KO zebrafish show a decreased motility (data not shown) and freely moving animals have a lower energy expenditure than do WT animals (Fig. 7).

The finding that expression of phosphoenolpyruvate carboxykinase (pck1) and glucose-6-phosphatase (g6pca.1), two key enzymes for hepatic gluconeogenesis, was not decreased in the GI system of hypothyroid Dio2KO fish is somewhat surprising because in rodents expression of Pck1, the rate-limiting enzyme in gluconeogenesis, is stimulated by THs via TH receptor β and is upregulated in liver of thyrotoxic rats (48, 49). Our findings are, however, in line with the above-mentioned suggestion that hypothyroidism in humans induces insulin resistance mainly in peripheral tissues such as skeletal muscle (45).

One additional intriguing observation is the absence of changes in glucose transporter expression in young adult Dio2KO compared with WT zebrafish. Early studies in rats showed that T₃ increases expression of Slc2a2 (GLUT2) and Slc2a4 (GLUT4), the main transporters involved in mammalian hepatic glucose influx/efflux and skeletal muscle glucose influx, respectively (50, 51). Furthermore, a TH receptor binding site has been identified in the rat GLUT4 gene promotor (52). GLUT2 is structurally and functionally well conserved between zebrafish and mammals, whereas GLUT4 is absent in zebrafish and functionally replaced by GLUT12 (38, 40). Our present results do not point to a direct effect of T₃ on the expression of either zebrafish glucose transporter but a detailed study of the genes’ promotor region should provide additional information in this regard.

Effects of Dio2 deficiency on glucose homeostasis in older adults

Unless adequately treated, T2D in humans progressively gets worse. After an initial phase of compensation for insulin resistance, prolonged hyperglycemia gradually impairs β cell function and reduces β cell mass (53). Moreover, the proliferative and regenerative capacities of mammalian β cells severely decline with age (54, 55). In contrast to mammals, zebrafish pancreatic islets have a high regenerative capacity and zebrafish can regenerate β cells throughout their entire life either by neogenesis or by transdifferentiation of α cells into β cells (56, 57). Nevertheless, some reports indicated that also in zebrafish the ability to react to insulin resistance/glucose exposure diminishes with age (17, 23). Therefore, we also included older animals in our study and surprisingly found that circulating glucose levels in mutants normalized around 1 year of age. At 15, 18, and 24 months, Dio2KO zebrafish were normoglycemic.

More detailed analysis of 18-month-old fish showed that ins as well as gcga/b expression was still increased compared with WT fish. However, in contrast to what we found at 6 months, this was accompanied by a more than threefold increase in surface of sections through the principal pancreatic islet as well as in islet cell number in Dio2KO compared with WT fish. This points to a compensatory response that at that stage was more efficient in compensating for the hyperglycemic effects of T₃ deficiency. The increased insulin production (number of β cells almost doubled compared with 6-month-old fish) may have been enough to overcome muscle insulin resistance as suggested by the normalization of glycolysis. At the same time, the increased glucagon production by the higher number of α cells was counteracted by other mechanisms, as discussed below.

Indeed, the huge increase in pancreatic islet size at 18 months was accompanied by some other changes compared with the observations at 6 months. Relating to hormone sensitivity, we found that glucagon receptor expression was strongly decreased in the GI system of Dio2KO zebrafish and to a lesser extent also in skeletal muscle. Adult zebrafish liver has much higher expression levels of glucagon receptors than does intestine (58), so the prominent decrease in the GI system can be attributed mainly to the liver. Hepatic glucagon resistance fits with the observation that gluconeogenesis in liver seems to be reduced (decrease in g6pca.1) and could be an additional factor contributing to normalization of blood glucose levels. A clear difference was also found in the expression of the glucose transporters slc2a2 and slc2a12, which were now both decreased in the GI system of mutants. Both transporters are highly expressed in adult zebrafish intestine and liver (38, 39). In rodents, GLUT2 is known to be important for glucose absorption at the level of the intestine (59), and glucose absorption is decreased in hypothyroid rats (60). It is also the major transporter for hepatic glucose exchange (61). GLUT12 is known to contribute to intestinal glucose absorption in mouse intestine and human colon cells, and its function is affected by reduced insulin sensitivity (62). Therefore, decreased expression of glucose transporters in liver/intestine of older Dio2KO zebrafish should also help to lower blood glucose levels by decreasing intestinal glucose uptake and hepatic glucose efflux. By decreasing intestinal and hepatic uptake, GLUT2 may also help to explain the observed decrease in hepatic glycogenesis. Alternatively, normal expression of slc2a12 at the level of skeletal muscle should allow sufficient GLUT12-mediated glucose influx. Together with the normalized expression of the key enzymes hk1 and pkmb, this suggests that glycolysis in muscle is no longer reduced at 18 months. Although there is not necessarily a causal relationship, this would provide the necessary energy for the normalized energy expenditure in freely moving Dio2KO 18-month-old zebrafish. The exact reasons for this normalization in spite of the persistent T₃ deficiency remain to be identified.

Conclusions

Figure 8 summarizes the main findings and hypotheses derived from this study. We found that Dio2KO zebrafish show an impaired glucose metabolism at young adult age, with a downregulation of insulin sensitivity and the glycolytic enzymes hexokinase and pyruvate kinase in muscle tissue. This results in hyperglycemia, despite upregulation of insulin expression in the GI system (representing their islet as well as gut-associated endocrine cells) with, however, also glucagon upregulation. As the animals age, a restoration of glycemic control is observed, with pronounced islet hyperplasia and increased β and α cell number. The increased insulin production may allow muscle insulin resistance to (at least partially) be overcome. At the same time, two additional compensatory changes occur in the GI system. Glucagon receptors are downregulated together with glucose-6-phosphatase, indicating hepatic glucagon resistance and decreased gluconeogenesis. GLUT2 and GLUT12 are also downregulated, which points to decreased glucose uptake at the level of the intestine. The compensation observed in Dio2KO zebrafish after an initial period of hyperglycemia clearly differs from the situation in mammals where long-term hyperglycemia typically results in decompensation and loss of pancreatic β cells. As such, adult Dio2KO zebrafish do not faithfully reproduce the changes occurring in (TH deficiency–induced) T2D. In contrast, the multiple compensatory mechanisms leading to normoglycemia in older Dio2KO zebrafish, despite persistent T₃ deficiency, provide interesting starting points for further investigation. This may lead to the identification of new therapeutic targets for (TH deficiency–induced) hyperglycemia or dysglycemia in general.

Acknowledgments

The authors thank Lut Noterdaeme, Lieve Geenen, Arnold Van Den Eynde, and Véronique Brouwers for technical assistance, Lien Andries for help with collecting the metabolic assay samples and analysis of the RT-qPCR data, Prof. Peggy Biga (University of Alabama at Birmingham, Birmingham, AL) for advice on the metabolic assays, and Prof. Tom Wenseleers (KU Leuven) for advice on the statistical analyses.

Financial Support: A.M.H. was supported by a fellowship from the IBSA Foundation for Scientific Research and subsequently by a postdoctoral fellowship from the Research Council of KU Leuven.

Additional Information

Disclosure Summary: The authors have nothing to disclose.

Data Availability: The datasets generated during and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.

Abbreviations:

Abbreviations:
 
• DAPI

  4′,6-diamidino-2-phenylindole

 
• DIO2

  type 2 deiodinase

 
• GI

  gastrointestinal

 
• GLP

  glucagon-like peptide

 
• GLUT

  glucose transporter

 
• IHC

  immunohistochemistry

 
• KO

  knockout

 
• PAS

  periodic acid–Schiff

 
• RT-qPCR

  real-time quantitative PCR

 
• T2D

  type 2 diabetes mellitus

 
• TH

  thyroid hormone

 
• WT

  wild-type

References and Notes