Solute Carrier Transporters as Potential Targets for the Treatment of Metabolic Disease

a. Citrate in energy metabolism

NaCT, encoded by the SLC13A5 gene, is the sodium-coupled citrate transporter. At physiological pH, NaCT is selective for citrate over the other TCA cycle intermediates succinate, malate, and fumarate, which are transported with lower affinities (Inoue et al., 2002a,b, 2004). NaCT provides TCA cycle intermediates for the liver and brain, metabolically active tissue, where NaCT is predominantly expressed (Figs. 1 and 2). In the liver, NaCT affects mitochondrial energy production, fatty acid synthesis, cholesterol synthesis, cytosolic glycolysis, and gluconeogenesis (Inoue et al., 2004). In the brain, NaCT is primarily expressed in neural plasma membranes and enables, in concert with NaDC3, an anaplerotic supply of TCA cycle intermediates (Wada et al., 2006; Yodoya et al., 2006). NaCT (also INDY, acronym for “I’m Not Dead Yet”) was first identified in Drosophila melanogaster, where mutations in the mammalian NaCT homolog Indy lead to life span extension (Rogina et al., 2000). Moreover, Indy mRNA was shown to be downregulated in diet-restricted healthy flies, and Indy-mutated long-lived flies share several phenotypes with long-lived caloric-restricted flies. Therefore, NaCT has been proposed to regulate the aging processes at least in part by activating an AMP-activated protein kinase/peroxisome proliferator-activated receptor γ coactivator 1α network (Rogina et al., 2000; Neretti et al., 2009; Wang et al., 2009; Rogers and Rogina, 2014; Schwarz et al., 2015).

b. Solute carrier 13A5 genetic variants

SLC13A5 mutations have been described recently to cause early-onset epileptic encephalopathy (Thevenon et al., 2014; Hardies et al., 2015; Klotz et al., 2016; Schossig et al., 2017; Weeke et al., 2017). In contrast, NaCT has not been directly linked to metabolic disease in humans, and genome-wide association is lacking. However, in liver samples of obese, insulin-resistant patients with NAFLD, SLC13A5 mRNA expression was significantly increased and associated with hepatic steatosis (von Loeffelholz et al., 2017b). Importantly, the study is the first hint of a relationship between NaCT and metabolic disease in humans.

c. Metabolic phenotype of animal models

The association between NaCT and metabolic disease is well characterized in animal models (Table 1). Slc13a5 deletion in mice resulted in a metabolic phenotype resembling caloric restriction (Birkenfeld et al., 2011). Remarkably, Slc13a5 knockout mice were protected from high-fat diet (HFD)–induced and aging-induced obesity, hepatic steatosis, and insulin resistance. The phenotype was linked to an increase in energy expenditure, hepatic mitochondrial biogenesis, and fatty acid oxidation, and a reduction in hepatic lipogenesis (Birkenfeld et al., 2011). Consistent with these mouse data, the knockdown of human SLC13A5 resulted in lower lipid levels in a hepatocarcinoma cell line (Li et al., 2015). Moreover, data obtained from rat primary hepatocytes show an increase in NaCT-mediated citrate transport after glucagon treatment and additionally detect cAMP-responsive element-binding protein-binding sites within the Slc13a5 promotor, suggesting that Slc13a5 is a cAMP-responsive element-binding protein–dependent glucagon target that is induced during periods of fasting and in T2D (Neuschäfer-Rube et al., 2014).

d. Pharmacological inhibition of sodium-coupled citrate transporter

Slc13a5 deletion and pharmacological inhibition ameliorate obesity, insulin resistance, and NAFLD in various model organisms (Birkenfeld et al., 2011; Shulman and Helfand, 2011; Willmes and Birkenfeld, 2013; Neuschäfer-Rube et al., 2014; Huard et al., 2015; Pesta et al., 2015; Brachs et al., 2016; Willmes et al., 2016, 2018; Rogina, 2017; von Loeffelholz et al., 2017a). NaCT has therefore been proposed as a promising target for the prevention and treatment of metabolic diseases. In rats, the inducible hepatic Slc13a5 knockdown at onset of the HFD feeding using 2′-O-methoxyethyl chimeric antisense oligonucleotides improved hepatic glucose production and insulin sensitivity and protected from NAFLD (Pesta et al., 2015), thus reproducing the Slc13a5 knockout phenotype. Moreover, hepatic Slc13a5 knockdown via small interfering RNA improved insulin sensitivity and hepatic triglyceride accumulation in mice (Brachs et al., 2016). These data led to the development of specific small-molecule NaCT inhibitors (Pajor and Randolph, 2007; Sun et al., 2010).

A first proof-of-concept study applied compound 2 (PF-06649298), a small dicarboxylate molecule that was discovered using a substrate-based design strategy and was shown to selectively bind to the NaCT transporter in a competitive and stereosensitive manner. Impressively, compound 2 led to the potent inhibition of NaCT-mediated citrate transport in vitro and in vivo (Huard et al., 2015). In compound 2–treated mice, glucose intolerance on HFD was completely reversed and hepatic di- and triglyceride content tended to be reduced (Huard et al., 2015). This finding suggests an increase in β-oxidation, supported by a tendency of higher β-hydroxybutyrate level. To determine the residues that are involved in binding of compound 2, methods of molecular modeling and site-directed mutagenesis, transport characterization, and cell surface biotinylation of the human NaCT transporter were used (Pajor et al., 2016). Apparently, residues adjacent to the putative citrate-binding site (G228, V231, V232, and G409) affect citrate transport and its inhibition, but also residues located outside the putative citrate-binding site (Q77 and T86) might be relevant (Pajor et al., 2016). Another small-molecule NaCT inhibitor, compound 4a, is less selective, but also reduced citrate uptake and moderately ameliorated glucose metabolism in rodents (Huard et al., 2016). Finally, these data reveal new mechanistic insights and could guide SLC13 inhibitor design.

To sum up, cell-based studies and mouse and human data indicate the potential of targeting the citrate transporter NaCT to ameliorate metabolic disease. Notably, small-molecule NaCT inhibitors are already developed and validated in mouse models of diet-induced obesity.

a. Monocarboxylates in energy metabolism

SLC16A1 is ubiquitously expressed and encodes for MCT1, the first and best-characterized member of the SLC16 family, which transports L-lactate, pyruvate, and ketone bodies across the plasma membrane (Halestrap and Price, 1999) (Figs. 1 and 2). Glycolytic cells produce lactic acid when oxygen is limited. Lactic acid is also a substrate for lipogenesis and gluconeogenesis in the liver, kidney, and adipose tissue, as well as a respiratory substrate for tissues such as the heart, skeletal muscle, and brain. Cells and tissues exhibiting these anabolic and catabolic pathways express high levels of MCT1, MCT2, and MCT4, catalyzing the uptake or efflux of lactic acid (Halestrap and Meredith, 2004; Halestrap, 2012, 2013; Halestrap and Wilson, 2012). This lactate shuttle between different cell types in the same tissue is well described, for example, in the muscle, where glycolytic white muscle fibers secrete lactate via MCT4 that is taken up by oxidative red muscle fibers via MCT1, and in the brain, where glycolytic astrocytes secrete lactate via MCT1 and MCT4 and oxidative neurons take up lactate via MCT2. Strikingly, lactate is the primary carbon source in the TCA cycle and therefore of energy production, highlighting the importance of monocarboxylates in energy homeostasis (Hui et al., 2017). Similar to lactic acid, pyruvate is produced during glycolysis, fueling gluconeogenesis or oxidation in a tissue-specific fashion. The ketone bodies β-hydroxybutyrate and acetoacetate are produced during fasting or ketogenic diets and provide energy for the brain, heart, and skeletal muscle (Halestrap and Meredith, 2004; Halestrap, 2012, 2013; Halestrap and Wilson, 2012).

b. Monocarboxylates and cancer

To date, MCT1 has primarily been linked to the pathogenesis of cancer, whereas little is known about its potential as target for other indications (Jones and Morris, 2016; Fisel et al., 2018). Because lactate metabolism is critical for tumor growth and MCT1 is an important monocarboxylate transporter in various cancers, this transporter could have relevance in anticancer therapy. The role of monocarboxylate transporters in tumor tissue has been well described: cancer cells rely on anaerobic glycolysis to supply energy (Warburg effect) and produce lactate, which is transported out of the cell via monocarboxylate transporters (Jones and Morris, 2016). As MCT1, MCT2, and MCT4 act in a proton-coupled mechanism, these carriers transport protons out of the cell, thereby attenuating intracellular acidosis. Secreted lactate is transported by SLC16 proteins into neighboring cells that rely on oxidative phosphorylation. Lactate shuttling in tumor tissue between cells with different energy needs is mediated by MCT1, MCT2, and MCT4. Therefore, these transporters may serve as tumor markers and drug targets. Of note, the MCT1 inhibitor AZD3965 is in phase I clinical testing (www.clinicaltrials.gov) as an anticancer drug. Multiple studies validated the efficacy of the therapeutic principle in cancer cell lines and an in vivo model of xenograft tumors, confirming the inhibition of bidirectional monocarboxylate transport by AZD3965 and illustrating reduced glycolytic activity and consequently reactivation of oxidative phosphorylation to maintain cell survival (Bola et al., 2014; Polański et al., 2014; Beloueche-Babari et al., 2017).

c. Solute carrier 16A1 genetic variants

Using AZD3965 in treating metabolic disease is not an obvious choice because genetic proof of an association between SLC16A1 gene locus and diet-induced obesity and T2D in human subjects is lacking. Nevertheless, MCT1 is linked to metabolic traits in humans and in mice (Table 1). In two independent studies, homozygous and heterozygous SLC16A1 loss-of-function gene mutations were linked to recurrent, severe ketoacidosis in humans (van Hasselt et al., 2014; Balasubramaniam et al., 2016). Ketoacidosis is caused by excess circulating ketone bodies in the face of insulinopenia and is generally a severe complication of diabetes (Nuttall, 1965; Winegrad and Clements, 1971). Targeted exome sequencing in an index patient and her family members revealed nine rare SLC16A1 variants (van Hasselt et al., 2014). Subsequent analysis of more patients with unexplained ketoacidosis identified seven additional variants, including six truncating mutations likely resulting in loss of protein function. Mutations were correlated with disease severity and MCT1 protein expression and activity. Therefore, MCT1-mediated transport of ketone bodies into extrahepatic tissues may be essential in preventing acidosis during periods of high ketone body turnover (van Hasselt et al., 2014). Although heterozygous SLC16A1 mutations can result in clinical symptoms, heterozygous relatives of patients did not develop ketoacidosis. Additional environmental and/or genetic factors may exacerbate symptoms. The hypothesis is supported by the symptom-free intervals seen in the patients (van Hasselt et al., 2014).

Promotor-activating SLC16A1 mutations have been identified in patients with exercise-induced hyperinsulinemia, a condition characterized by periods of severe hypoglycemia after physical exercise and linked to syncope (Meissner et al., 2001; Otonkoski et al., 2003, 2007). Excess SLC16A1 expression in pancreatic β-cells may increase circulating insulin via a mechanism linking increased pyruvate uptake and catabolism to higher ATP levels and insulin secretion (Otonkoski et al., 2007). MCT1-related pathophysiology in humans deserves further evaluation because MCT1 inhibition may be beneficial in treating exercise-induced hyperinsulinemia, but could also promote ketoacidosis.

d. Metabolic phenotype of animal models

Exercise-induced hyperinsulinemia and hypoglycemia were reproduced in transgenic mice with β-cell–specific Slc16a1 overexpression (Pullen et al., 2012). Genetic Slc16a1 deletion further suggested that the transporter affects whole-body energy homeostasis (Lengacher et al., 2013). Slc16a1 null mutation in mice is embryonically lethal, whereas heterozygous animals develop normally. Interestingly, heterozygous Slc16a1 knockout mice are resistant to diet-induced obesity and show less insulin resistance and hepatic steatosis compared with littermate controls. Reduced weight gain was associated with decreased hepatic and white adipose fat accumulation, potentially mediated by AMP-activated protein kinase secondary to reduced substrate availability (Carneiro et al., 2017). Metabolic changes in MCT-deleted mice support the concept that monocarboxylate transport is important for metabolic control. A putative role of MCT1 in metabolic disease is also supported by increased brain expression in diet-induced obese and genetically obese ob/ob and diabetic db/db mice (Pierre et al., 2007). Chronic hyperglycemia elicits a similar response in rat brains (Canis et al., 2009), which governs energy homeostasis. Slc16a1 was also upregulated in mouse liver, kidney, and small intestine by the peroxisome proliferator-activated receptor α (PPARα), an important mediator of lipid catabolism that is linked to obesity (König et al., 2008).

In summary, multiple publications implicate monocarboxylate transporters in body weight regulation (Carneiro and Pellerin, 2015). Despite the metabolic phenotype of Slc16a1 haploinsufficient mice and altered Slc16a1 expression in obesity, MCT1 was never discussed as a metabolic treatment target, and no genetic evidence has been described in human GWAS. However, clinical testing of the MCT1 inhibitor AZD3965 holds promise.

a. Discovery of a novel type 2 diabetes risk gene

A hitherto unnoticed and uncharacterized member of the SLC16 family has recently attracted interest in metabolic research: MCT11, the transporter encoded by the SLC16A11 gene. SLC16A11 sequence variants predicted T2D risk in a GWAS. The SIGMA type 2 diabetes consortium analyzed SNPs in Mexicans and other Latin Americans and identified a haplotype associated with T2D (Williams et al., 2014). The risk haplotype includes one silent and four missense SNPs in the SLC16A11 gene. Three of four amino acid–substituting SNPs are predicted to be tolerated (rs117767867, rs75418188, rs75493593), whereas one SNP (rs13342692) is predicted to be damaging. The risk to develop T2D increases by approximately 20% per haplotype copy. Subjects carrying the risk haplotype develop T2D on average 2.1 years earlier and at 0.9 kg/m² lower body mass index (BMI) than noncarriers. The stronger association in younger and leaner individuals may reflect a role of MCT11 in T2D pathogenesis independently of adiposity. Remarkably, the risk haplotype is present in 48% of Native American and in 12% of Asian populations, but rare or absent in European or African samples. Gene expression analysis in human tissues illustrated high SLC16A11 expression in the liver, salivary glands, and thyroid (Fig. 2). Moreover, human MCT11 protein localizes to the endoplasmic reticulum (Fig. 1), but only marginally to the plasma membrane (Williams et al., 2014; Rusu et al., 2017). Metabolite profiling of SLC16A11-expressing HeLa cells showed a substantial increase in triacylglycerol (TAG) and a modest increase in diacylglycerol (DAG) concentrations, whereas lysophosphatidylcholine, cholesterol ester, and sphingomyelin lipid concentrations were decreased. Perturbed lipid metabolism, especially TAG and DAG accumulation, promotes insulin resistance and T2D. Possibly, MCT11 modulates T2D risk in part through actions on hepatic lipid metabolism.

b. Solute carrier 16A11 genetic variants

The findings regarding the SLC16A11 risk haplotype in Mexicans (Williams et al., 2014) inspired subsequent investigations. The SLC16A11 polymorphism rs13342692, predicted to be the crucial SNP in the risk haplotype, was not significantly associated with T2D in Mayas (Lara-Riegos et al., 2015). However, the study presented significant associations between rs13342692 and plasma insulin levels and homeostasis model assessment for insulin resistance in nonobese nonaffected risk allele carriers, implicating a higher susceptibility to metabolic disorders for risk allele carriers. Additionally, SLC16A11 risk haplotype carriers exhibit decreased insulin action, and transaminase levels are increased in carriers with diabetes (Almeda-Valdes et al., 2019). The diabetes risk haplotype was further confirmed in a Hispanic/Latino cohort in the United States with an association to diabetes in the Mexican-American, but not in non-Mexican Hispanic/Latino groups (Hidalgo et al., 2019).

In American Indians, carriers of the SLC16A11 rs75493593 risk allele with a BMI < 35 kg/m² showed an increased risk for T2D and a decreased T2D risk with higher BMI (Traurig et al., 2016). Similarly, a longitudinal follow-up on the BMI and diabetes status of individuals yielded a higher hazard ratio in the leanest BMI and vice versa. The longitudinal data set also raised the question of whether rs75493593 allele carriers with lower BMI are prone to T2D, or whether they lose weight as a consequence of T2D onset. In one study cohort, individuals carrying the risk allele lost more weight after T2D onset, highlighting the possibility that additional factors influence the association between BMI and the risk allele in T2D. However, the finding was not reproduced in another study cohort (Traurig et al., 2016). Individuals carrying the risk allele displayed increased carbohydrate and reduced lipid oxidation rates (Traurig et al., 2016). Differential levels of lipid metabolites have also been observed in SLC16A11-expressing HeLa cells (Williams et al., 2014). Reduced mitochondrial β-oxidation due to reduced mitochondrial density with subsequent DAG accumulation predisposes development of insulin resistance and T2D (Samuel et al., 2010; Birkenfeld and Shulman, 2014; Goedeke et al., 2019). Therefore, lower lipid oxidation rates in a SLC16A11 risk allele carrier could conceivably confer increased T2D risk. The SLC16A11 risk allele also affected expression of the adjacent RNASEK gene, encoding for the ribonuclease K. The interaction decreases RNASEK expression in the adipose tissue, muscle, and whole blood of risk allele carriers. Possibly, both genes contribute to the phenotypic associations observed in this study. Because altered hepatic and adipose RNASEK expression were not detected in another study of risk haplotype carriers (Rusu et al., 2017), the significance of this observation is unclear.

c. Potential causal role of monocarboxylate transporter 11

Recent data showed that the SLC16A11 diabetes risk haplotype affects MCT11 by reducing its mRNA expression, but also by inhibiting transporter translocation to the plasma membrane (Rusu et al., 2017). Specifically, SLC16A11 gene expression was reduced in liver samples taken from carriers of the risk haplotype in a gene dose-dependent manner. Additionally, the authors show that the four coding variants associated with T2D risk reduce cell-membrane expression of the transporter. Diminished interaction with the chaperone protein basigin may have contributed to reduced transporter activity. Finally, to analyze the consequences of reduced MCT11 function, the authors knocked down SLC16A11 in primary human hepatocytes and observed that intracellular acylcarnitines, DAGs and TAGs, were significantly increased. DAGs and ceramides mediate insulin resistance via activation of novel protein kinases and subsequent serine phosphorylation of the insulin receptor and/or the insulin receptor substrate 1 (Birkenfeld and Shulman, 2014; Petersen et al., 2016). Thus, MCT11 may affect cellular fatty acid and lipid metabolism, thereby promoting insulin resistance and T2D. However, the exact molecular pathway remains unknown. Reduced cell surface localization of T2D risk variants may be critical in T2D pathogenesis. Despite this, only a very small portion of the wild-type protein localizes to the plasma membrane, and it is unclear whether MCT11 serves as an intracellular and/or plasma membrane transporter. MCT11 may translocate to the plasma membrane in physiological conditions requiring its transport activity. Furthermore, other SLC16 family members also transport pyruvate, and additional substrates may be transported by MCT11. Perhaps some substrates are specifically transported by MCT11 under particular physiological or metabolic conditions. If so, compensatory substrate transport by other monocarboxylate transporter may not suffice, when MCT11 function is reduced through genetic risk variants. Clearly, cellular localization and the regulation of MCT11 in health and disease require further study.

d. Metabolic phenotype of animal models

The first data on a MCT11-deficient mouse model were published in 2019 (Zhao et al., 2019b). Whereas constitutive Slc16a11 knockout did not result in severe metabolic consequences, even in animals fed HFD, reconstitution of knockout mice with mutant Slc16a11 in the liver using a liver-specific adeno-associated virus resulted in a metabolic phenotype. The phenotype comprised increased blood and liver triglycerides and reduced glucose tolerance and insulin resistance. Hepatic reconstitution of mutant Slc16a11, but not of the wild-type Slc16a11, results in Lipin-1 upregulation. Therefore, the observed increase in liver triglycerides may have been mediated through Lipin-1, an enzyme involved in triglyceride synthesis. In contrast to previous studies indicating that reduced MCT11 function promotes T2D risk (Rusu et al., 2017), the authors proposed that gain-of-function mutations in Slc16a11 foster T2D. However, functional characterization of mutant Slc16a11 was not shown, and hepatic reconstitution with wild-type Slc16a11 also resulted in lipid accumulation. Although these findings suggest that MCT11 contributes to hepatic lipid metabolism, they do not prove that gain of function explains the observed detrimental metabolic phenotype in their MCT11 mouse model.

In conclusion, although the detailed mechanism of MCT11 action in the development of T2D is still not clear, the causal role of MCT11 in T2D is strongly supported by genetic association combined with in vitro genetic and functional data predicting the loss of protein function to be disease promoting. Thus, enhancing MCT11 function may hold “the promise as a potential therapy for T2D” (Stadler and Farooqi, 2017). Therapeutic approaches could increase gene expression or protein level, restore MCT11 protein interaction with basigin for correct plasma membrane localization, or increase transport activity of the protein.

a. Mitochondrial uncoupling and thermogenesis

UCP1, also known as thermogenin, is encoded by the SLC25A7 gene and is expressed in brown adipose tissue (Fig. 2). In contrast to white adipose tissue, which is the primary site of energy storage, brown adipose tissue contains large amounts of mitochondria. Through uncoupled mitochondrial respiration, brown adipose tissue regulates thermogenesis. UCP1 is responsible for the mitochondrial proton leak (Nicholls et al., 1978) and is the UCP, which mediates energy dissipation as heat (Golozoubova et al., 2001; Nedergaard et al., 2001). The term “mitochondrial proton leak” defines a non-ATP–consuming process that takes place when inner mitochondrial membrane proton conductance is increased and the generated proton gradient is insufficient such that oxygen consumption is uncoupled from ATP production. Because mitochondria of brown adipose tissue are particularly permeable to protons, proton leakage in this tissue directly dissipates energy as heat, explaining the thermogenic activity of brown adipose tissue (Nicholls and Locke, 1984). Brown adipose tissue thermogenesis plays a key role in the development of obesity in animal models (Rothwell and Stock, 1979; Lowell et al., 1993; Feldmann et al., 2009; Kajimura and Saito, 2014) and in humans (Nedergaard et al., 2007; Cypess et al., 2009; Saito et al., 2009; van Marken Lichtenbelt et al., 2009; Kiefer, 2017; Leitner et al., 2017).

b. Solute carrier 25A7 genetic variants

Numerous human studies have linked SLC25A7 polymorphisms to obesity and T2D (Table 1). Most association studies focused on the rs1800592 polymorphism in the SLC25A7 promotor region that has been associated with reduced SLC25A7 mRNA expression in obese subjects (Esterbauer et al., 1998). The datasuggest that reduced protein function increases susceptibility to metabolic disease. However, findings across more than 30 human genetic studies, including subjects from different populations worldwide, varied. Some studies showed associations between one or more SLC25A7 polymorphisms with obesity, diabetes, or other metabolic syndrome characteristics. Others did not show associations between SLC25A7 polymorphisms and metabolic traits (Brondani et al., 2012). Data remain conflicting regarding the effect of the allelic variants. For example, whereas the G-allele of the rs1800592 polymorphism (-3826A>G) was significantly associated with increased BMI (Chathoth et al., 2018), another study reported that G-allele carriers had lower body weight and fat mass and homozygous G-allele carriers had lower frequency of T2D compared with patients carrying the A-allele (Nicoletti et al., 2016). Other studies did not detect any association between the SLC25A7 polymorphism and BMI, obesity, or T2D risk (de Souza et al., 2013; Brondani et al., 2014a,b).

c. Metabolic phenotype of animal models

The idea that UCP1 dysfunction promotes obesity has been extensively investigated in mouse models. As early as 1978, defects in thermogenesis were demonstrated in genetically obese ob/ob mice that were cold-sensitive (Himms-Hagen and Desautels, 1978). Transgenic mice with the primary deficiency of brown adipose tissue lacking UCP1 became obese at a young age in the absence of hyperphagia. This finding indicates that brown adipose tissue–deficient mice have increased metabolic efficiency (Lowell et al., 1993). However, Slc25a7 knockout mice were cold-sensitive, indicating defective thermoregulation, but the animals did not become obese on either a standard diet or HFD. Loss of UCP1 could therefore be compensated in part by UCP2 that is induced in brown fat of UCP1-deficient mice. Other signals that are involved in the regulation of body weight could also contribute to the finding (Enerbäck et al., 1997).

Transgenic mice with Slc25a7 expression in white adipose tissue exhibit markedly reduced HFD-induced obesity in association with a comparable total oxidative capacity in white and brown adipose tissue (Kopecky et al., 1995; Kopecký et al., 1996a,b; Baumruk et al., 1999). Ectopic Slc25a7 expression in epididymal fat through adenovirus vector injection improves glucose tolerance and decreases food intake in both diet-induced and genetically obese mouse models (Yamada et al., 2006). Transgenic Slc25a7 expression in mouse skeletal muscle increased muscle oxygen consumption dramatically and protected from diet-induced obesity, improved insulin sensitivity, and increased metabolic rate at rest and with exercise (Li et al., 2000). In a genetic obese mouse model, tetracycline-inducible Slc25a7 expression in skeletal muscle lowered body weight, increased oxygen consumption, improved glucose tolerance, and unexpectedly lowered blood pressure (Gates et al., 2007).

d. Mitochondrial uncoupling to treat metabolic disease

The so-called browning of white adipose tissue has received much attention in recent years. Today, it is known that UCP1 can be induced in white fat cells by cold as well as pharmacological, nutritional, and endogenous stimuli (Ost et al., 2017). Although the presence of brown fat depots in adult humans was questioned, SLC25A7-expressing adipocytes in adults were shown in multiple studies and also associated with lower mass and/or activity in obese and older subjects (Cypess et al., 2009; Saito et al., 2009; van Marken Lichtenbelt et al., 2009; Virtanen et al., 2009; Leitner et al., 2017).

Obesity treatment aims to reduce energy intake and/or increase energy expenditure. Therapeutic intervention reduces energy intake either by affecting satiety centers in the brain such as sibutramine (Bray et al., 1999; James et al., 2000; Ryan, 2000) and glucagon-like peptide 1 receptor agonists (O’Neil et al., 2018), or by decreasing the efficiency of intestinal absorption such as orlistat (Sjöström et al., 1998). Chemical agents such as 2,4-dinitrophenol or carbonyl cyanide p-(trifluoromethoxy) phenylhydrazone increase energy expenditure by uncoupling mitochondrial oxidative respiration. However, life-threatening side effects limit the use of such compounds (Parascandola, 1974). More recently, liver-specific mitochondrial protonophores, called controlled-release mitochondrial protonophores, produce mild hepatic mitochondrial uncoupling, ameliorated hepatic steatosis, and insulin resistance in a HFD-fed rat model and steatohepatitis in a rat model of nonalcoholic steatohepatitis (Perry et al., 2015). The treatment also prevented hepatic steatosis, insulin resistance, and T2D in lipodystrophic mice (Abulizi et al., 2017). Given this proof of concept, mitochondrial uncoupling targeting UCP1 received renewed interest to modulate whole-body energy homeostasis and to treat obesity and related metabolic diseases (Ost et al., 2017).

Taken together, mouse and human studies indicate an essential role of UCP1 in body weight regulation, T2D, and arterial hypertension. Therefore, modulation of brown adipose tissue thermogenesis to increase energy expenditure could be an effective strategy against obesity. Reduced SLC25A7 gene expression in obese subjects indicates reduced transporter activity as causative for the related disease phenotypes, and mouse models clearly support the idea of protection from diet-induced obesity due to ectopic Slc25a7 expression. Most importantly, a very recent publication identified the chemical compound linifanib as a potent inducer of UCP1 expression in primary inguinal adipocytes in vitro and in vivo (Zhao et al., 2019a). Of note, linifanib (ABT-869), an orally bioavailable, small-molecule receptor tyrosine kinase inhibitor, has been reported to exhibit potent antiangiogenic and antitumor effects in preclinical studies and is currently undergoing phase III trials (McKeegan et al., 2015; Horinouchi, 2016). Thus, additional research on the effect of a specific UCP1 activator in the context of obesity seems to be enormously promising, and existing linifanib preclinical studies could facilitate its application in obese subjects. In contrast to existing anti-obesity drugs, an UCP1 activator might have additional metabolic benefits regarding diabetes and hypertension. However, the safety profile will need to be carefully studied due to the possibility of hyperthermia, and the parallel effect of receptor tyrosine kinase inhibition in the patients must be critically investigated.

a. Physiological functions irrespective of mitochondrial uncoupling

UCP2, encoded by SLC25A8, was identified in 1997 as novel mitochondrial uncoupling protein that, unlike UCP1, is widely expressed in human tissues (Fleury et al., 1997; Gimeno et al., 1997) (Fig. 2). UCP3, product of the SLC25A9 gene predominantly expressed in skeletal muscle, was identified shortly thereafter (Boss et al., 1997; Lowell, 1999) (Fig. 2). Because the two genes are localized adjacent on chromosome 11, they may result from an ancestral gene duplication. In contrast to UCP1, the physiological functions of UCP2 and UCP3 are less well defined and may include the export of fatty acids or fatty derivates from the mitochondrial matrix (Moore et al., 2001; Goglia and Skulachev, 2003; Schrauwen et al., 2003; Schrauwen and Hesselink, 2004; Bouillaud et al., 2016). Other physiological functions involve antioxidant mechanisms, which may maintain the balance between reactive oxygen species production and clearance. A possible side effect of proton leaks is a higher flow rate through the electron–transport chain, and therefore a decrease in the half-life of radical intermediates, which can donate its electron to oxygen and create reactive oxygen species (Dalgaard and Pedersen, 2001). UCP2 and UCP3 might dissipate the proton gradient to prevent reactive oxygen species generation, thereby regulating oxidative stress and cell signaling (Echtay et al., 2002; Mailloux and Harper, 2011; Bouillaud et al., 2016).

b. Solute carrier 25A8 and solute carrier 25A9 genetic variants

The first SLC25A8 and SLC25A9 gene polymorphisms were identified in Danish persons with obesity (Urhammer et al., 1997, 1998). In the last two decades, allelic variants of the genes have been studied in different populations (Table 1). Dalgaard and Pedersen (2001) and Dalgaard (2011) reviewed physiologic al functions of both transporters and known polymorphisms that are associated with obesity and T2D. Three common SLC25A8 polymorphisms are well studied in humans: the promoter variant rs659366 (-866G>A), the missense variant rs660339 (c.164C>T; p.A55V), and a 45-bp insertion–deletion polymorphism in the 3′ untranslated region. The rs659366 promoter minor A-allele most likely directs higher rates of transcription from the SLC25A8 promoter compared with the G-allele (Dalgaard, 2011). Conflicting studies show the polymorphism to be associated with a reduced risk of obesity (Esterbauer et al., 2001; Jun et al., 2009), but also with increased obesity prevalence (Dhamrait et al., 2004; Ochoa et al., 2007; Kring et al., 2008). Several studies revealed no associations, suggesting that UCP2 has only a modest effect, if any, on the development of obesity (Dalgaard, 2011). Similarly, human data on β-cell function and T2D are partially conflicting. Heterozygous SLC25A8 mutations were reported in patients suffering from congenital hyperinsulinemia (Gonzalez-Barroso et al., 2008). Because SLC25A8 null mutations are related to hyperinsulinemia, the A-allele of the rs659366 variant, resulting in increased gene expression, is expected to show association with decreased β-cell function and ultimately with T2D (Dalgaard, 2011). Indeed, decreased glucose-stimulated insulin secretion was observed in two studies (Sesti et al., 2003; Sasahara et al., 2004), and decreased basal insulin secretion was reported among A-allele carriers (Esterbauer et al., 2001). Others did not reproduce the associations (Dalgaard et al., 2003; Sesti et al., 2003; D’Adamo et al., 2004; Ji et al., 2004; Bulotta et al., 2005). Moreover, both alleles of the rs659366 variant were associated with T2D (Sasahara et al., 2004; Lyssenko et al., 2005; Gable et al., 2006). Both increased risk of hypertension (Ji et al., 2004) as well as decreased risk of coronary artery disease in subjects with T2D (Cheurfa et al., 2008; Palmer et al., 2009) have been reported to be associated with the A-allele of the rs659366 polymorphism. Given the conflicting data, larger samples with well-characterized participants are required to detect the true effect of an identified variant (Dalgaard, 2011). A more recent meta-analysis of 42 studies evaluated associations of the three common SLC25A8 gene polymorphisms with overweight and obesity: the T-allele of rs660339 polymorphism was associated with an increased risk of obesity, the A-allele of the rs659366 promotor polymorphism protected from obesity, and the 45-bp insertion/deletion polymorphism was not associated with obesity susceptibility (Zhang et al., 2014).

The promoter variant rs1800849 (-55C>T) of the SLC25A9 gene is common and well-studied (Cassell et al., 2000; Dalgaard et al., 2001a,b; Berentzen et al., 2005). The T-allele results in increased SLC25A9 expression in skeletal muscle and may raise resting energy expenditure (Schrauwen et al., 1999c; Kimm et al., 2002). Low resting energy expenditure promotes weight gain. Indeed, one study revealed a negative correlation between SLC25A9 expression and BMI (Schrauwen et al., 1999b). Furthermore, the T-allele was associated with reduced BMI, increased high-density lipoprotein–cholesterol levels (Schrauwen et al., 1999c; Liu et al., 2005; Hamada et al., 2008), reduced T2D risk, and higher plasma total cholesterol and low-density lipoprotein cholesterol concentrations (Meirhaeghe et al., 2000). However, multiple studies did not observe associations between the rs1800849 polymorphism and metabolic rate, obesity, BMI, insulin secretion, or T2D (Walder et al., 1998; Dalgaard et al., 2003; Berentzen et al., 2005; Ochoa et al., 2007; Hsu et al., 2008). Some showed a reverse association of the T-allele with higher BMI and waist circumference (Otabe et al., 2000; Herrmann et al., 2003; Lindholm et al., 2004). Jia et al. (2009) provided a comprehensive overview of SLC25A9 polymorphisms and noticed the discrepancies between studies that might be explained partially by influences of adjacent loci in the same haplotype, as well as the exclusion of key lifestyle factors in the statistical analyses.

The potential role of both genes in metabolic disease is further supported by studies indicating that expression of SLC25A8 and SLC25A9 predicts diabetes onset in humans. Variation in the SLC25A8-SLC25A9 gene cluster is associated with an increased T2D risk (Gable et al., 2006). Moreover, reduced UCP3 protein levels were reported in individuals with prediabetes and T2D (Schrauwen et al., 2006). Recent meta-analyses detected significant associations between the SLC25A8 and SLC25A9 polymorphisms and BMI, obesity, and T2D (de Souza et al., 2013; Brondani et al., 2014a,b). However, associations markedly differed between different populations; a recent cross-sectional study of European adolescents failed to identify any association between SLC25A8 or SLC25A9 variants and adiposity, whereas the C-allele of the SLC25A7 rs6536991 intron variant was associated with a lower risk of overweight (Pascual-Gamarra et al., 2019). Mechanisms and phenotypes are not clear to date, and, as for every GWAS, interindividual differences, including ethnic, lifestyle, environmental and genomic factors, and nutritional characteristics, result in different effects of described SLC25A8 and SLC25A9 polymorphisms on obesity and related phenotypes. These factors could interfere with UCP2 and UCP3 physiology and pathophysiology. Regarding limited consistency between studies, further human studies and functional characterization of both transporters could help to solve the unanswered questions.

c. Metabolic phenotype of animal models

Mouse studies support links between UCP2 and UCP3 and obesity and related disorders, but also present varying phenotypes. Increased Slc25a8 mRNA levels have been observed in obesity-resistant A/J mice compared with obesity-prone B6 mice. Moreover, HFD induced Slc25a8 in white adipose tissue. The findings suggest that SLC25A8 might play a role in the development of hyperinsulinemia and obesity (Fleury et al., 1997). UCP2 does not seem to regulate body weight or cold-induced thermogenesis in mice (Arsenijevic et al., 2000; Zhang et al., 2001). UCP2-deficient mice exhibited increases in insulin and lower blood glucose levels. Furthermore, pancreatic islets lacking UCP2 feature increased ATP levels and increased insulin secretion. The finding indicates that UCP2 negatively regulates glucose-stimulated insulin secretion possibly linking obesity and β-cell dysfunction. Indeed, UCP2 is upregulated in obese ob/ob mice, and UCP2 deficiency markedly improved their diabetic phenotype (Zhang et al., 2001). However, contrary to results in UCP2-deficient mice of mixed 129/B6 strain background (Zhang et al., 2001), glucose-stimulated insulin secretion in UCP2-deficient islets of congenic mouse strains was significantly decreased, and these mice also did not become diabetic (Pi et al., 2009). Increased oxidative stress in UCP2-deficient mice may contribute to declining β-cell function, but overall, the pathophysiological role of UCP2 is unclear.

UCP3 deficiency in mice does not affect body weight regulation, exercise tolerance, fatty acid oxidation, or cold-induced thermogenesis. Yet, reactive oxygen species are more abundant in skeletal muscle, indicating that UCP3 protects from oxidative stress (Gong et al., 2000; Vidal-Puig et al., 2000). Transgenic mice overexpressing human SLC25A9 in skeletal muscle are hyperphagic. Nevertheless, body weight and adipose tissue mass are reduced, together with lower fasting plasma glucose and insulin levels and an increased glucose clearance rate (Clapham et al., 2000; Horvath et al., 2003). Strikingly, UCP2 and UCP3 deficiency in mice is associated with increased production of reactive oxygen species that is known to be involved in insulin resistance (Houstis et al., 2006). Therefore, modulation of UCP2 and UCP3 activity may protect from insulin resistance in addition to obesity.

Finally, although SLC25A8 and SLC25A9 polymorphism may be associated with obesity, diabetes, or related phenotypes, their impact on these phenotypes seems to be modest with large heterogeneity between studies (Dalgaard and Pedersen, 2001). To date, the effects of UCP2 and UCP3 modulation remain elusive because human and mouse data did not clearly demonstrate whether the transporters should be activated or inhibited for therapeutic intervention. The functional characterization based on human association studies and on mouse models is very complex and needs further evaluation. Nevertheless, the link between UCP function and dysfunction and metabolic disease is well established and supports the idea of targeting UCPs.

a. Mitochondrial aspartate–glutamate transport

The aspartate–glutamate carrier 2 (AGC2), also known as citrin, adult-onset citrullinemia type II (CTLN2), or aralar2, is encoded by the SLC25A13 gene and facilitates the calcium-dependent exchange of cytoplasmic glutamate with mitochondrial aspartate across the inner mitochondrial membrane (Fig. 1). AGC2 is a key component of the malate–aspartate shuttle and essential to export aspartate from the mitochondrial matrix to the cytosol. Aspartate is required for urea synthesis from ammonia and alanine, protein synthesis, gluconeogenesis from lactate, as well as the oxidation of cytosolic NADH (Palmieri, 2004). The transporter is the only aspartate–glutamate carrier isoform expressed in the liver (Fig. 2). Hence, in the liver, mutations in the SLC25A13 gene cannot be compensated by another transporter system and can cause two age-dependent autosomal–recessive diseases: neonatal intrahepatic cholestasis by citrin deficiency (NICCD) and CTLN2, an accumulation of ammonia and other toxic substances in the blood.

b. Solute carrier 25A13 genetic variants

Mutated SLC25A13 has been identified in patients suffering from type II citrullinemia and proposed to affect urea cycle function (Kobayashi et al., 1999). In 2001, two reports linked SLC25A13 mutations to a neonatal liver disease (Ohura et al., 2001; Tazawa et al., 2001). Multiple case reports with over 50 different SLC25A13 gene mutations further underpin AGC2 dysfunction in human metabolic disease (Table 1). Hepatocytes have limited capacity to take up extracellular aspartate. Therefore, both AGC2-related disease phenotypes are mainly caused by reduced cytosolic aspartate and hepatic NADH reoxidation (Palmieri and Monne, 2016). AGC2 deficiency decreases hepatic argininosuccinate synthetase activity, causing hyperammonemia and citrullinemia. Additionally, reduced malate–aspartate shuttle activity increases the cytosolic NADH/NAD⁺ ratio, resulting in the inhibition of glycolysis and alcohol metabolism. The mechanisms explain carbohydrate and alcohol intolerance in affected patients, which exacerbates symptoms, and preference for high-protein diets containing aspartate. High-carbohydrate diets increase hepatic triglycerides in some patients, likely due to compensatory malate–citrate shuttling transferring reducing equivalents to the mitochondria, thereby producing acetyl-CoA in the cytosol and stimulating fatty acid synthesis (Palmieri, 2004).

Neonatal NICCD is characterized by transient intrahepatic cholestasis, fatty liver, hepatomegaly, ketotic hypoglycemia, hypoproteinemia, and aminoacidemia. Due to liver dysfunction, some patients experience hepatitis, jaundice, decreased coagulation factors, hemolytic anemia, and bleeding diathesis. NICCD is usually benign, and symptoms disappear after a few years. In adulthood, some patients develop CTLN2, which is associated with hypoproteinemia and hyperammonemia, leading to a neurological manifestation, including encephalopathy and neuropsychiatric symptoms (Palmieri, 2004, 2013; Palmieri and Monne, 2016). Some adult patients also manifest fatty liver and hyperlipidemia unrelated to obesity or ethanol consumption, pointing to the essential function of the aspartate–glutamate transporter in hepatic metabolism. Finally, SLC25A13 mutations have been linked to steatosis, nonalcoholic steatohepatitis, and hepatocellular carcinoma (Takagi et al., 2006; Tsai et al., 2006; Fukumoto et al., 2008; Komatsu et al., 2008; Chang et al., 2011). For patients with CTLN2, liver transplantation is one of the most promising prospects. However, dietary supplementation can ameliorate symptoms; patients with NICCD usually respond to medium-chain triglyceride supplements and lactose-restricted formulas, whereas medium-chain triglyceride supplements with low-carbohydrate formulas are recommended for CTLN2 (Hayasaka et al., 2012, 2014, 2018; Hayasaka and Numakura, 2018).

c. Metabolic phenotype of animal models

Beneficial responses to dietary interventions have also been shown in Slc25a13 knockout mice, indicated through increased food intake and maintained body weight following supplementation with protein or medium-chain triglycerides (Saheki et al., 2012). Initially, the Slc25a13 knockout mouse model failed to mimic CTLN2 and did not show changes in glucose, amino acid, or ammonia metabolism. At least in mice, AGC2 deficiency may not be sufficient to elicit liver disease (Sinasac et al., 2004). Because enhanced glycerol phosphate shuttle activity may be compensating for the loss of AGC2 function, Slc25a13 Gpd2 (glycerol 3-phosphate dehydrogenase 2) double-knockout mice were generated. The model developed citrullinemia, hyperammonemia, hypoglycemia, and a fatty liver, displaying features of human AGC2 deficiency (Saheki et al., 2007).

Taken together, the phenotypes of AGC2 deficiency have been extensively investigated in human beings and in mice (Saheki and Song, 2005; Hayasaka and Numakura, 2018). However, AGC2 was never suggested as target to treat fatty liver, possibly because dietary interventions and liver transplantation are well established treatments for NICCD and CTLN2. Moreover, pharmacological activation of the transporter seems to be challenging. AGC2 deficiency associated with fatty liver is only one symptomatic feature in humans and has never been directly related to obesity or metabolic disease. Given the complex physiological function of the hepatic aspartate–glutamate carrier, additional research on the role of AGC2 in obesity-related metabolic disease may be worthwhile and could generate novel therapeutic concepts beyond the two well-known human phenotypes.

a. Mitochondrial carnitine–acylcarnitine transport

The broadly expressed mitochondrial carnitine–acylcarnitine carrier (CAC), product of the SLC25A20 gene, catalyzes the exchange between cytosolic acylcarnitine that enters the mitochondria and intramitochondrial-free carnitine that is transferred into the cytosol (Figs. 1 and 2). Thereby, CAC is a key component of the carnitine shuttle system. The system transfers acyl groups from acyl-CoA to carnitine by the outer mitochondrial membrane carnitine palmitoyltransferase 1, translocates acylcarnitines across the inner mitochondrial membrane by CAC, and finally transfers the acyl groups from acylcarnitines to CoA inside mitochondria by carnitine palmitoyltransferase 2. The mechanism imports acyl groups of cytosolic long-chain fatty acids into mitochondria. In mitochondria, acyl groups are oxidized by the β-oxidation pathway, the major energy source for heart and skeletal muscle during periods of fasting and physical exercise (Indiveri et al., 2011; Palmieri, 2013). As this pathway is fundamental in cellular energy metabolism, CAC transporter dysfunction leads to a severe disease known as CAC deficiency (Table 1).

b. Solute carrier 25A20 genetic variants

CAC deficiency was described in 1992 as an autosomal recessive disorder (Stanley et al., 1992). The first pathogenic SLC25A20 gene mutation was reported in 1997 (Huizing et al., 1997). Over 40 additional mutations have since been described as resulting in the complex disease pattern that clinically manifests with episodes of coma upon fasting, cardiomyopathy, hypotonia, muscle weakness, respiratory distress, seizures, and hepatic dysfunction (Indiveri et al., 2011; Palmieri and Monne, 2016). Multiple circulating metabolites are abnormal in this condition. Key findings include hypoglycemia secondary to oxidation of glucose instead of ketone bodies in the muscle during fasting, as well as hypoketosis, hyperammonemia, dicarboxylic aciduria, and increased levels of transaminases and long-chain acylcarnitines. The clinical symptoms are well characterized and based on the energy deficit in tissues that rely on fatty acid oxidation. Moreover, elevated acylcarnitine levels may damage the heart and skeletal muscle, whereas hypoketosis, hyperammonemia, and hypoglycemia result in the neurological phenotype (Indiveri et al., 2011; Palmieri, 2013). The disease can occur in a mild or severe form, with remaining protein activity of about 5% or below 1%, respectively. CAC pathophysiology, common clinical features, and interventions of the disease have been reviewed recently (Indiveri et al., 2011; Yan et al., 2017). The severe phenotype is usually present in the neonatal period with a high mortality rate within the first year of life due to cardiac arrest and respiratory failure (Yan et al., 2017). Therefore, early appropriate diagnosis and treatment are necessary to manage the highly lethal disorder and to ameliorate disease progression. Therapeutic approaches include intravenous glucose injection to suppress fatty acid mobilization and oxidation, as well as ammonia detoxification through arginine administration. Long-term management requires strict diet with frequent meals to avoid fasting periods, high-carbohydrate intake, medium-chain triglyceride supplementation, and limitation of long-chain fatty acids (Indiveri et al., 2011; Yan et al., 2017).

In summary, given the severity of CAC deficiency and the limited success of existing therapies, research on new therapeutic approaches is needed. Moreover, whether direct CAC modulation could ameliorate other disorders, including obesity-related metabolic disease, deserves to be studied. However, more common genetic polymorphisms may not result in a dramatic loss of protein activity and have not been associated with other disease phenotypes. Slc25a20 knockout mice are not available to test its obesity-related phenotype under HFD feeding. Keeping in mind the crucial function of the transporter in energy metabolism and the severe metabolic defects in CAC deficiency, research in this direction should be considered.

a. Mitochondrial transport of adenine nucleotides

The mitochondrial ATP-Mg/phosphate carrier protein (APC)1, encoded by SLC25A24, is broadly expressed and facilitates exchange of adenine nucleotides, including ATP-Mg, ATP, ADP, and AMP, and phosphate between mitochondrial matrix and cytosol (Figs. 1 and 2). The transporter regulates adenine nucleotide concentrations in the mitochondrial matrix, thereby affecting mitochondrial adenine-nucleotide–dependent enzymes, which regulate gluconeogenesis from lactate, urea synthesis, mitochondrial DNA replication, transcription, and protein synthesis, among others (Palmieri and Monne, 2016).

b. Solute carrier 25A24 genetic variants

SLC25A24 was proposed as a novel susceptibility gene for low fat mass in humans (Urano et al., 2015) (Table 1). Using SNP arrays, three SNPs in the SLC25A24 gene were associated with body fat percentage and BMI (rs491785, rs519129, and rs547364) in Japanese postmenopausal women (Urano et al., 2015). Subjects harboring the major alleles had lower body fat percentage and BMI and the rs491785 major allele associated with lower SLC25A24 mRNA expression in human preadipocytes. No additional gene variants linked to disease phenotypes in other populations and/or larger study cohorts are known.

c. Metabolic phenotype of animal models

Slc25a24 knockout mice are resistant to HFD-induced obesity with significantly reduced white adipose tissue and liver weights (Urano et al., 2015). The phenotype was linked to reduced adipocyte differentiation marker expression and reduced hepatic triglycerides in Slc25a24 knockout mice (Urano et al., 2015).

In conclusion, APC1 may be linked to obesity in humans and mice. Further studies are needed to verify these data and clarify APC1-regulated mechanisms and (patho)physiology in more detail.

a. Zinc and insulin biosynthesis

The most prominent member of the SLC30 family in terms of metabolic disease is ZnT8, product of the SLC30A8 gene. The transporter is mainly expressed in the membrane of insulin secretory granules in pancreatic β-cells (Figs. 1 and 2) and provides zinc for insulin synthesis (Chimienti et al., 2004). Zinc concentrations are particularly high in β-cells, especially in insulin secretory granules, where it forms crystalloid-like hexamers with insulin. The process maintains insulin maturation, storage, and stability (Emdin et al., 1980; Hutton, 1989; Foster et al., 1993; Dodson and Steiner, 1998). The critical role of zinc in insulin biosynthesis was first proclaimed when Scott and Fisher (1938) observed concomitant reductions in insulin and zinc in the pancreas, but unaltered zinc concentrations in the livers of patients with diabetes. Serum zinc concentrations are also significantly reduced in patients with diabetes (Garg et al., 1994). The complex relationship between zinc and both type 1 and type 2 diabetes has further been established (Chausmer, 1998; Maret, 2017; Fukunaka and Fujitani, 2018). Zinc deficiency in patients with diabetes led to the idea of zinc supplementation to promote β-cell function. The intervention showed beneficial metabolic effects in humans and in T2D animal models (Taylor, 2005; Jayawardena et al., 2012).

b. Solute carrier 30A8 genetic variants

Whether zinc deficiency in patients with diabetes is the cause or consequence of the disease was not clear for a long time. However, a potentially causal relationship was discovered with the identification of a SLC30A8 sequence variant that increased T2D susceptibility in various populations worldwide (Table 1). Furthermore, the transporter was identified as a major autoantigen in human type 1 diabetes (Wenzlau et al., 2007). The link to T2D arose with the identification of the SLC30A8 SNP rs13266634 as risk loci for T2D in three independent GWAS (Saxena et al., 2007; Scott et al., 2007; Sladek et al., 2007). The first study analyzed polymorphisms in French T2D patients and identified four novel loci, including SLC30A8. Two additional studies in Finnish and Swedish subjects described the same SLC30A8 sequence variant (Saxena et al., 2007; Scott et al., 2007). The rs13266634 polymorphism leads to a single base change from C to T with consequent amino acid change from arginine (R) to tryptophan (W) at position 325 (c.973C>T; p.R325W). Increased T2D risk was associated with the major C-allele and accordingly the R325 protein variant. The polymorphism is common in individuals of European and East Asian descents and associated with a 14% increase in diabetes incidence per risk allele (Cauchi et al., 2010). The first functional studies demonstrated that the R325 risk variant is less active than the low-risk W325 variant (Nicolson et al., 2009; Kim et al., 2011) and provided evidence that ZnT8 activation could minimize diabetes risk. In fact, SLC30A8 overexpression in insulinoma cells enhanced insulin secretion during hyperglycemia. Thus, ZnT8 may promote zinc storage in insulin granules and thereby the insulin secretory pathway (Chimienti et al., 2006; Sladek et al., 2007). The ability to transport and store zinc rather than the extracellular zinc status is important for zinc accumulation in β-cells. Therefore, SLC30A8 overexpression could counteract zinc depletion in diabetes, which likely participates in β-cell mass loss (Chimienti et al., 2006). According to the effect of SLC30A8 overexpression on insulin secretion, subjects homozygous for the C-risk allele exhibit a decrease in first-phase insulin release after intravenous glucose administration (Boesgaard et al., 2008).

Influences of endogenous plasma zinc and dietary zinc supplementation on insulin secretion or T2D risk have been related to SLC30A8 sequence variants. Glucose-stimulated insulin secretion in healthy nondiabetic Amish individuals with and without the rs13266634 polymorphism was tested before and after zinc supplementation. Contrary to the previous hypothesis, individuals with the lower-risk CT/TT genotypes showed attenuated insulin responses to glucose stimulation compared with homozygous C-risk allele carriers without zinc supplementation (Maruthur et al., 2015). However, after 14 days of zinc supplementation, the increase in insulin response to glucose was not significantly changed in the homozygous risk allele carriers compared with baseline and was increased further in the nonrisk CT/TT-allele carriers (Maruthur et al., 2015). A prospective study of a Swedish cohort presents zinc supplementation as lowering the risk of T2D, but does not find significant modification of the zinc–T2D associations by the SLC30A8 rs13266634 polymorphism (Drake et al., 2017). However, T2D risk was lowest among nonrisk allele carriers on zinc supplements. A Chinese population presented the inverse association between plasma zinc levels and diabetes. The association was more pronounced among individuals carrying the TT genotype of SLC30A8 rs13266634 polymorphism than those carrying CT or CC genotypes (Shan et al., 2014). The attenuated inverse association of zinc level and T2D risk in the subjects carrying the C-risk allele supports the data sets on zinc supplementation (Maruthur et al., 2015; Drake et al., 2017). Apparently, beneficial zinc effects on metabolism are attenuated in risk allele carriers.

Although the first functional studies of the SLC30A8 risk allele discovered in 2007 predicted reduced transporter activity, SLC30A8 sequence variants lowering diabetes risk were described in 2014 (Flannick et al., 2014). Twelve novel SLC30A8 sequence variants, resulting in a truncated protein and loss of transporter activity, predicted reduced T2D risk. This publication was groundbreaking because previous functional studies of ZnT8 suggested that reduced zinc transport increases T2D risk, whereas SLC30A8 loss-of-function mutations provide strong evidence that haploinsufficiency protects against T2D, and ZnT8 inhibition may be suitable to treat T2D. However, the novel identified variants are rare.

c. Metabolic phenotype of animal models

The potential causal role of ZnT8 in T2D has been extensively investigated in mouse models (Table 1). HFD-fed Slc30a8 knockout mice are glucose intolerant and diabetic with less responsive β-cells (Lemaire et al., 2009). Similarly, chow-fed Slc30a8 knockout mice showed reduced glucose-stimulated insulin secretion and lower zinc content of secretory granules, but an overall mild metabolic phenotype (Nicolson et al., 2009; Pound et al., 2009). Because ZnT8 is not only expressed in β-cells, but also in pancreatic α-cells (Nicolson et al., 2009), the role of ZnT8 in the two different islet cell populations was further addressed using a β-cell–specific Slc30a8 knockout mouse. HFD-fed whole-body Slc30a8 knockout mice became remarkably obese, hyperglycemic, hyperinsulinemic, insulin resistant, and glucose intolerant, whereas mice with specific Slc30a8 deletion in β-cells showed an attenuated metabolic phenotype with a similar body weight to control mice (Hardy et al., 2012). These data suggest that ZnT8 contributes to the risk of developing T2D through β-cell– and non-β-cell–specific effects (Hardy et al., 2012). Indeed, the metabolic function of ZnT8 and interacting environmental factors, genetic background, gender, and age ultimately determine the variable metabolic phenotype of whole-body Slc30a8 knockout (Lemaire et al., 2009; Nicolson et al., 2009; Pound et al., 2009, 2012) and β-cell–specific Slc30a8 knockout models (Wijesekara et al., 2010; Hardy et al., 2012; Tamaki et al., 2013; Mitchell et al., 2016). However, most of the studies display abnormalities in insulin crystallization and storage, impaired glucose tolerance, and lower circulating insulin levels. Thus, reduced ZnT8 activity alters glucose homeostasis, thereby increasing T2D risk (da Silva Xavier et al., 2013). Vice versa, Slc30a8 transgenic mice overexpressing the transporter in β-cells show markedly improved glucose tolerance (Mitchell et al., 2016).

Unexpectedly, some studies showed unaltered or improved insulin release after glucose stimulation of isolated Slc30a8 knockout β-cells (Lemaire et al., 2009; Nicolson et al., 2009; Tamaki et al., 2013; Mitchell et al., 2016). The contradictory observation of lower plasma insulin levels in vivo, but improved insulin secretion in vitro, could be explained by a functional link between Slc30a8 deletion in β-cells and hepatic insulin clearance (Tamaki et al., 2013). Zinc secreted along with insulin from pancreatic secretory granules could conceivably affect neighboring endocrine cells and the liver. By inhibiting clathrin-dependent insulin endocytosis, zinc suppresses hepatic insulin clearance. Insulin uptake in hepatocytes increases as a result of the lower pancreatic zinc release in Slc30a8 knockout mice. Indeed, the study also showed low circulating insulin levels in the face of insulin hypersecretion from pancreatic β-cells. Because ZnT8 affects hepatic insulin clearance, genetic dysregulation of this could contribute to T2D.

HFD-fed transgenic mice overexpressing the human SLC30A8 high-risk R325 variant exhibit increased pancreatic zinc and proinsulin levels and decreased insulin and glucose tolerance, confirming the detrimental effect of the human risk allele (Li et al., 2017a). Conversely, overexpression of the low-risk W325 variant decreases pancreatic zinc and proinsulin levels together with increases in insulin and glucose tolerance. The beneficial metabolic response to SLC30A8 overexpression strongly suggests that ZnT8 activation could ameliorate metabolic disease. This idea is further supported by the fact that pancreatic ZnT8 protein levels are downregulated in diabetic disease models such as db/db and Akita mice in the early stages of T2D (Tamaki et al., 2009).

In contrast to previously described mouse phenotypes and gene expression data stressing potential benefits of ZnT8 activation, a recent study of knockin mice carrying a human SLC30A8 loss-of-function variant yielded unexpected findings. The model featured increased glucose-induced insulin secretion, likely reflecting increased capacity of β-cells to secrete insulin. The study suggested for the first time that ZnT8 inhibition could elicit beneficial metabolic responses (Kleiner et al., 2018), supporting the finding of SLC30A8 loss-of-function variants that reduced T2D risk (Flannick et al., 2014). Although loss of function seems to be beneficial in this mouse line, the variability of mouse phenotypes makes it difficult to judge how ZnT8 must be modulated to achieve metabolic protection.

d. Pharmacological activation or inhibition of zinc transporter 8

The diabetic phenotypes of the knockout mice and human data on SLC30A8 genetics and zinc supplementation support the idea that ZnT8 augmentation could improve T2D. However, discrepant data concerning the metabolic effect of the risk allele raise the question of whether the SLC30A8 risk variant causes an increase or a decrease in zinc transport activity. Accordingly, whether ZnT8 activation or inhibition should be pursued as T2D treatment remains unclear. Whereas in mice loss of ZnT8 function results in impaired glucose tolerance, human data are contradictory, with studies describing high-risk and low-risk variants with loss of function. The SLC30A8 R325 risk variant displayed lower zinc transport (Nicolson et al., 2009), but was proposed to be hyperactive in another study (Merriman et al., 2016) that is supported by the finding that the diabetes risk genotype CC at rs13266634 is associated with higher islet zinc concentrations (Wong et al., 2017), along with the identification of rare T2D-risk–reducing SLC30A8 loss-of-function mutations (Flannick et al., 2014). The following concept may explain the paradoxical finding that high-risk and low-risk variants both present lower protein activity (Rutter and Chimienti, 2015): assuming the activity of the risk variant is 30% lower compared with the nonrisk variant, zinc transport into granules would be reduced by 15% and 30% in heterozygous and homozygous C-allele carriers. In contrast, heterozygous loss-of-function variants result in 50% lower activity. Consequently, the rare loss-of-function mutations result in a lower transport activity compared with the rs13266634 risk variant. The risk variant R325 results in moderate reduction of ZnT8 activity, lowering β-cell zinc secretion and enhancing hepatic insulin clearance. In contrast, more severe reduction in ZnT8 function increases insulin secretion from β-cells, compensating the impaired zinc release and insulin clearance. This model argues for a complex dose–response relationship between ZnT8 activity and diabetes risk (Rutter and Chimienti, 2015). Species-specific differences may explain the increased risk of hyperglycemia in Slc30a8 knockout mice with the simpler model of progressively impaired insulin action due to increased hepatic clearance. Moreover, the differential role of ZnT8 in insulin secretion and glucose homeostasis could be explained in part by an age-dependent effect (Rutter and Chimienti, 2015). This idea originated from the observation of impaired glucose tolerance and abnormal insulin granules in younger (6–8 weeks old), but not in older Slc30a8 knockout mice (Nicolson et al., 2009). Furthermore, pancreatic islet cell survival in the face of hypoxia was enhanced in older (>12 weeks) Slc30a8 knockout mice, but not younger animals (Gerber et al., 2014). However, whether findings in rodents can be translated to human disease remains unclear, and particularly the interaction with age, gender, drugs, or physical activity may influence ZnT8 function.

Taken together, both rodent and human data highlight the significance of ZnT8 function in glucose homeostasis and the pathological consequences of ZnT8 malfunction. However, drug interventions targeting ZnT8 are likely complex given the proposed bell-shaped model of the of ZnT8 function in the development of T2D. Clearly, effects of both activating and inhibiting drugs will need to be investigated in suitable models prior to clinical trials in human beings (Rutter and Chimienti, 2015).