Molecular genetic regulation of Slc30a8 /ZnT8 reveals a positive association with glucose tolerance.

Zinc Transporter 8 (ZnT8), encoded by SLC30A8 , is chiefly expressed within pancreatic islet cells where it mediates zinc (Zn 2+ ) uptake into secretory granules. Whilst a common non-synonymous polymorphism (R325W) , which lowers activity, is associated with increased type 2 diabetes (T2D) risk, rare inactivating mutations in SLC30A8 have been reported to protect against T2D. Here, we generate and characterise new mouse models to explore the impact on glucose homeostasis of graded changes in ZnT8 activity in the β cell. Firstly, Slc30a8 was deleted highly selectively in these cells using the novel deleter strain, Ins1Cre. The resultant Ins1CreZnT8KO mice displayed significant (p<0.05) impairments in glucose tolerance at 10 weeks of age versus littermate controls and glucose-induced increases in circulating insulin were inhibited in vivo . Whilst insulin release from Ins1CreZnT8KO islets was normal, Zn 2+ release was severely impaired. Conversely, transgenic ZnT8Tg mice, overexpressing the transporter inducibly in the adult β cell using an insulin promoter-dependent Tet-On system, showed significant (p<0.01) improvements in glucose tolerance compared to control animals. Glucose-induced insulin secretion from ZnT8Tg islets was severely impaired, whereas Zn 2+ release was significantly enhanced. Our findings demonstrate that glucose homeostasis in the mouse improves as β cell ZnT8 activity increases and, remarkably, these changes track Zn 2+ rather than insulin release in vitro . Activation of ZnT8 in β cells might therefore provide the basis of a novel approach to treating type 2 diabetes.


Introduction.
The regulation of insulin secretion by glucose involves the uptake and metabolism of the sugar by pancreatic β cells (1), stimulation of mitochondrial oxidative metabolism (2), Ca 2+ influx (3) and the exocytosis of the hormone from dense core secretory granules (4,5) where it is stored in a near-crystalline form alongside Zn 2+ and Ca 2+ ions (6). Although it is increasingly accepted that impaired insulin secretion underlies the development of type 2 diabetes (T2D) (7), a disease affecting more than 8 % of the adult population worldwide (8), the mechanisms involved remain poorly understood (9). Nonetheless, disease risk is strongly influenced by both genetic (10) and environmental (11)

factors.
A non-synonymous variant in the SLC30A8 gene associated with elevated T2D risk was identified by genome-wide association studies (GWAS) in 2007 (12). Expressed almost exclusively in pancreatic β and α cells (13)(14)(15), SLC30A8 encodes a secretory granule-resident zinc transporter, ZnT8, implicated in the accumulation of zinc within these organelles and thus in insulin storage (16). Given these likely roles, SLC30A8/ZnT8 has been mooted as a potentially tractable new target for personalised disease therapy.
Subsequent functional studies on the expressed ZnT8 protein (14,17) demonstrated that the risk (R325) variant is a less active zinc transporter than the protective (W) form.
Consequently, possession of risk alleles seems likely to impair insulin crystallisation and storage. Supporting this view, mice inactivated globally (14,18) or selectively in the β cell (15,19) for Slc30a8 revealed striking abnormalities in the formation of dense cores within insulin granules. Surprisingly, however, measurements of insulin release from isolated islets from Slc30a8 null mice revealed either no change (18) or improved (14,19) glucosestimulated insulin secretion from isolated islets or the perfused pancreas, and unchanged insulin content. Despite this, glucose homeostasis and circulating insulin levels were both lowered in ZnT8 null animals. Providing a possible explanation for this conundrum, Tamaki and colleagues (19) demonstrated that the enhanced release of Zn 2+ ions alongside insulin in W-variant carriers suppresses insulin clearance (and presumably non-productive insulin signalling) by the liver, favouring insulin action on this, as well as other tissues (notably adipocytes and skeletal muscle). An observed increase in C-peptide:insulin ratio in human Rcarriers supported this model, since the mature hormone, but not proinsulin, is expected to be cleared by the liver. Moreover, Slc30a8 elimination from the mouse has no effect on insulin processing (14,18), arguing against a β cell-autonomous action of the variant on the release of mature versus partially processed forms. Together, the above findings have stimulated the search for activators of the transporter which, by favouring Zn 2+ accumulation by β cell secretory granules, may eventually prove useful in the clinic.
However, and challenging the above view, a recent study based largely on Swedish, Finnish and other Northern European populations, but also including individuals from elsewhere, identified rare (< 0.1% of the population) nonsense (truncating) or mis-sense mutations in the SLC30A8 gene. Unexpectedly, the carrier population showed a ~3-fold enrichment for healthy individuals versus those with T2D, implying a protective role for the mutant transporter.
Although only a small number of carriers was involved (345 in total of ~150,000 subjects sequenced) a range of structurally distinct variants was found in cohorts with differing ancestry, providing evidence that the SLC30A8 mutations, rather than other polymorphisms in the same linkage disequilibrium (LD) block, were likely to explain the changes in disease risk.
The above findings are nonetheless difficult to reconcile with the observed increase in T2D risk in carriers of the common risk alleles. Although an activating effect of the identified mutants on the remaining allele cannot be excluded absolutely, an alternative explanation (20) is that a complex interplay between insulin storage and Zn 2+ release by β cells, and downstream effects on target tissues including the liver, results in a bimodal (bell-shaped) dependence of T2D risk on ZnT8 activity. Thus, modest decreases in β cell ZnT8 activity, as observed in carriers of the common risk (R) variants, may act chiefly by lowering β cell Zn 2+ secretion, thus enhancing insulin clearance by the liver. On the other hand, a more substantial lowering of ZnT8 activity, engendered by rare loss-of-function alleles, may lead to a more dramatic increase in insulin release from the pancreas, an effect outweighing impaired Zn 2+ release and altered insulin clearance.
The impact of deleting ZnT8 from the β cell in mice has also been the subject of some debate.
Thus, one recent study (21) reported that global knockout on a pure C57BL6 background exerted no effects on glucose tolerance, in contrast to findings on more mixed backgrounds (14,18). Moreover, several previously-reported β cell-selective deletion models are complicated by deletion in other tissues, including the brain, when Cre deleter strains (notably RIP2Cre and Pdx1), with activity in these tissues (22), are used. Correspondingly, RIP2Cre:ZnT8 mice gain more weight versus controls on a high fat diet than observed with globally deleted animals (23). The latter findings argue for a role for ZnT8 in a small number of neuronal cells in which the Pdx1 or Ins2 promoter may be at least transiently active during development or at later stages. On the other hand, the mouse insulin promoter 1 Cre (MIP Cre) used in (19) may also be affected by the co-expression of growth hormone encoded by the cDNA included in this transgene (24).
Our first goal here was therefore to explore the impact of deleting ZnT8 more specifically in the β cell, and on a pure C57BL6 background, using a new deleter strain in which the Ins1 promoter, which is inactive in brain and other tissues (22) drives expression of Cre after introduction into the endogenous locus ("knock-in") (25,26). Importantly, Ins1Cre mice do not express the growth hormone (GH) minigene, unlike both RIP2Cre (24) and MIPCre mice (27), and mice bearing the transgene alone display no abnormalities in glucose tolerance (22) (G.R. unpublished results).
Up to now, there have been no attempts to examine the effect of over-expressing ZnT8 selectively in the β cell, thus mimicking one of the likely actions of agents capable of stimulating the activity of the transporter. Our second goal here was therefore to generate a series of transgenic mouse lines in which ZnT8 expression is under the control of rat insulin promoter Tet-On system (28).
We demonstrate that highly selective deletion of ZnT8 in the β cell leads to dense core granule mis-formation and glucose intolerance. By contrast, overexpression of the transporter in the β cell in adults leads to improved glucose tolerance but reduced insulin secretion, whilst Zn 2+ release is markedly enhanced. A positive relationship thus pertains between β cell ZnT8 expression (and Zn 2+ secretion), and glucose tolerance. If reflective of human physiology, these results lend weight to the view that ZnT8 activation might prove beneficial in the context of T2D.

Results.
Impaired glucose tolerance and insulin secretion in Ins1Cre:ZnT8 fl/fl mice. β cell-selective deletion of ZnT8 with a variety of Cre deleter strains (eg. RIP2 (15) and MIP (19)) display varying degrees of recombination at extrapancreatic sites, due to ectopic expression of Cre. By contrast, Ins1Cre knockin mice display no detectable expression of the recombinase in the brain, only very minor recombination in other islet cells (<3% of α cells in utero), but > 94% recombination in β-cells (25,29). We therefore used this model to inactivate ZnT8 selectively in β cells (Fig. 1A). Confirming efficient deletion of the endogenous ZnT8 alleles in the β-cell with Ins1Cre, islets from Ins1Cre +/-:ZnT8 fl/fl (Cre + ) mice showed >80 % reduction in ZnT8 mRNA levels (****p<0.001; two way ANOVA; Crevs. measures two-way ANOVA; n=14 Cre -& 13 Cre + ) in response to a 3g/Kg body weight glucose injection, consistent with impaired insulin secretion or enhanced clearance of the hormone. Insulin sensitivity measured using an insulin tolerance test was unchanged in both male ( Fig 2F) and female (Supp. Fig. 2) Ins1Cre +/-:ZnT8 fl/fl mice, as assessed at 10 or 8 weeks, respectively. These findings are in-line with those in global (14) or mouse insulin 1promoter deleted animals (19).  Fig 3A) and, similarly, glucose-stimulated insulin release was not different between null and wild-type islets assayed during perifusion at 16.7 mmol/L glucose (Supp. islets respectively). Finally, β cell-β cell connectivity (32), known to contribute to the regulation of insulin release from intact islets, was unaltered in ZnT8 null islets (Fig. 3D,E).
The above studies thus demonstrate that deleting ZnT8 selectively in the β cell leads to normal insulin, but abnormal Zn 2+ secretion, in vitro, but markedly lower post-stimulation insulin levels and glucose tolerance in vivo.

Improved glucose tolerance in ZnT8 transgenic mice.
To explore the impact of increasing ZnT8 activity in β cells we next generated transgenic mice in which the expression of the transporter was under the control of a bi-directional tetracycline-inducible promoter (38). β cell-selective induction was then achieved by activation of a Tet-On transactivator expressed selectively in β cells under the control of the rat insulin 2 promoter (RIP7-rtTA) (Fig. 6A). Whilst 7 founders were produced, we selected two (#31, #23; copy numbers 5 and 13, respectively) for further analysis (Supp. Fig 4). ZnT8 Tg+; p<0.0001; Student's t-test; n=3 each genotype). The expression of human ZnT8 protein was also apparent by Western blotting using an antibody specific for the human protein ( Fig. 6E).
Females from founder #31 displayed significant improvements in glucose tolerance at both 10- ( ZnT8 Tg+ islets secrete less insulin but more Zn 2+ in response to glucose. Assayed in vitro, insulin secretion from isolated islets derived from 10-14 week old female mice in response to high glucose (16.7 mM) was significantly reduced ( Fig. 8A; 0.94 ± 0.21 ng/mL vs. 0.40 ± 0.05 ng/mL, ZnT8 Tg-vs ZnT8 Tg+ respectively; p<0.05; two-way ANOVA, n=10-13 replicates). No differences were apparent in the response of transgenic islets to stimulation with incretin or depolarisation with KCl (Fig. 8A Staining pancreatic slices for insulin and glucagon (Fig 8D) revealed no changes in β or α cell mass nor the ratio of β to α cells (Fig 8E-G).

Discussion
In this report, we describe new mouse models for ZnT8 which provide insights into the pathogenic mechanisms likely to be involved in the actions of human alleles associated with increased T2D risk.
Ins1CreZnT8 KO mice showed dramatic changes in secretory granule morphology and plasma insulin level under glucose stimulation, similar to findings previously reported in global ZnT8 KO mice (14), in mice with conditional ZnT8 alleles deleted with the more promiscuous RIP2 Cre (15), or with MIPCre which also expresses GH (24). The present results thus confirm that such morphological changes are likely to be a β cell-autonomous event, and to reflect impaired Zn 2+ uptake into dense core granules in the absence of ZnT8.
Despite the exaggerated glucose excursions and smaller plasma insulin increases observed in response to intraperitoneal injection of the sugar in these animals ( Fig. 2B & E, respectively), islets derived from Ins1CreZnT8KO mice displayed unaltered glucose-stimulated insulin secretion in vitro. By contrast, glucose-induced Zn 2+ release from these islets was reduced by > 80%, in line with earlier results with global ZnT8 KO mice (36), presumably reflecting impaired Zn 2+ accumulation by secretory granules. These findings reinforce the recent proposal (19,35) that impaired β-cell Zn 2+ secretion and de-inhibition of insulin receptor endocytosis leads to exaggerated clearance of mature insulin by the liver.
We extend support for the above view by showing that glucose tolerance is improved in a new model in which ZnT8 is selectively over-expressed in the β cell. Remarkably, insulin secretion from islets isolated from these mice was barely stimulated by glucose, whereas Zn 2+ release was increased by >50%. Nonetheless, fasting insulin levels tended to be increased in ZnT8Tg animals (Fig. 7D) and these levels were further strongly increased by intraperitoneal glucose injection (Fig. 7D). Thus, in ZnT8Tg animals, elevated Zn 2+ secretion may act both to impair insulin clearance through the internalisation of insulin receptors (19), and possibly also to enhance insulin signalling. At the molecular level, possible actions of the released Zn 2+ included inhibition of insulin receptor dephosphorylation by protein tyrosine phosphatase B1 (PTPB1) (39), or of phosphatidylinositol (3,4,5) phosphate (PIP 3 ) degradation by phosphatase and tensin homologue on chromosome ten (PTEN) (40).
These studies show that, by manipulating Slc30a8 expression selectively in the mouse β cell using molecular genetics, a near-linear relationship exists in this species between ZnT8 levels and glucose tolerance (Fig. 9A). Note that the study of mice deleted for just one conditional Slc30a8 allele was not feasible with the breeding strategy used here though, in earlier studies with global Slc30a8 null mice (14), we noted that heterozygous (ZnT8 +/-) mice displayed intermediate glucose tolerance between wild-type and homozygous null animals, consistent with the current findings. Importantly, the changes in peak glucose observed in the present study were best correlated to Zn 2+ release from the islet (Fig. 9B): this was essentially eliminated by Slc308 deletion (Fig. 4D) but enhanced when the transporter was overexpressed (Fig. 8B). By contrast, insulin release in vitro was inversely correlated with ZnT8 expression (Fig. 8C).
Interestingly, we observed differences between the impact of Slc308 deletion on male and female mice, with only males showing defective glucose tolerance over the age range examined. In contrast, both male and female mice deleted globally for the transporter (31) displayed glucose intolerance at six weeks of age, whereas only males were intolerant at 12 weeks. The reasons for the differences between the impact of Slc308 deletion between sexes is presently unknown, but may in part reflect the intrinsically greater inter-measurement variability in females resulting from the reproductive cycle, and/or the lower insulin sensitivity of male animals which imposes a greater metabolic stress on the β-cell.
Surprisingly, this position was reversed in transgenic animals with the greater penetrance of ZnT8 over-expression observed in females. In this case, the underlying mechanisms are less clear but might reflect sex-specific differences in the handling of enhanced Zn 2+ loads by the liver or other target tissues (41).
Whether control of hepatic insulin clearance and/or action via Zn 2+ assumes the same importance in man, where the much larger diameter and volume of the portal vein may mean greater dilution of Zn 2+ after release from the β cell -and hence lowered action on the liveris unclear. Nonetheless, and arguing for this possibility, carriers of risk (R) SLC30A8 alleles show lowered C-peptide:insulin levels, consistent with the more efficient uptake of the latter by hepatocytes when Zn 2+ levels are lowered (42).
The present approach further demonstrates the feasibility of using mouse genetics to explore the mechanisms through which T2D risk genes, identified in GWAS studies (12), act.
Intriguingly, we provide additional evidence that the actions of SLC30A8 involve interactions between multiple tissues (β cells and liver), despite the tight restriction of the expression of this gene to the endocrine pancreas (13). Whether SLC30A8 variants also influence the release of glucagon may require further investigation: global inactivation of the gene exerted little effect on glucagon release from islets, though detailed in vivo analysis involving hypoglycemic clamps were not reported in these studies (15). Fadista et al (43) recently reported a strong positive correlation between glucagon and SLC30A8 expression in human islets, consistent with a role for SLC30A8 variants in controlling glucagon production.
In the light of the present results, the possibility that other GWAS genes expressed in multiple tissues, e.g. TCF7L2 (44,45), might act via extrapancreatic sites to regulate insulin secretion, would seem worthy of careful investigation. Of note, TCF7L2 is an upstream regulator of the mouse Slc30a8 (46) and human SLC30A (47) genes, and as a "master" regulator of T2D susceptibility (47), might act in part via ZnT8 to modify β cell Zn 2+ release and insulin clearance.

Ethical approval
All animal procedures were approved by the home office according to the Animals (Scientific Procedures) Act 1986 of the United Kingdom (PPL 70/7349).

Generation of β-cell selective knockout mice by Ins1Cre-driven recombination
ZnT8 floxed mice (ZnT8 fl/fl ) were generated by GenOway, Lyon, France (15). This involved the insertion of a LoxP site together with a flippase recognition target flanked neomycin selection cassette within intron 1 and a single distal LoxP site within the upstream exon 1 containing the translational start codon. ZnT8 fl/fl animals were then bred with the Ins1Cre deleter strain, to produce 50% β-cell specific knockout animals (Ins1Cre +/-ZnT8 fl/fl ) and 50% littermate controls (Ins1Cre -/-ZnT8 fl/fl ). Note that, in contrast to RIP2Cre (24) and MIP2Cre (ADA), Ins1Cre mice do not express a GH cassette and the transgene alone does not affect glucose tolerance (25,29). Animals were maintained in a pathogen-free facility under a 12 h light/12 h dark cycle with free access to water and food.

Generation of β-cell specific transgenic mice
Plasmid pCDNA3, containing the human ZnT8 (W325 form) coding sequence with the addition of a single COOH-terminal c-Myc epitope tag (14), was digested with XhoI, blunt-end filled and further digested with NotI. The digested hZnT8-Myc DNA fragment was gel purified and cloned into plasmid pBI-L Tet (Clontech) between NotI and PvuII sites. This generated a plasmid with a bidirectional tetracycline-regulated promoter driving expression of both hZnT8-Myc and firefly luciferase. The positive clone was further confirmed by DNA sequencing using a pBI-L internal primer GAAAGAACAATCAAGGGTCC and a hZnT8 primer ACACTAGCACGCCAGTCACC.
The expression cassette was excised from the plasmid backbone by AatII and hZnT8 -Luc-/RIP7-rtTA + (ZnT8 Tg-). All offspring were genotyped for both the hZnT8 and RIP7-rtTA genes (49). Mice were treated with 0.5 g/L doxycycline from 5 weeks of age.

Islet Isolation
Mice were euthanised by cervical dislocation and pancreatic islets isolated by collagenase digestion as preciously described (50). Given the sex-and age-dependent differences between mouse lines, islets used for ex vivo analysis were obtained from mice of the appropriate sex and, importantly, at an age where an in vivo phenotype was apparent i.e. 10 week old male Ins1Cre +/-::ZnT8 fl/fl mice and 10 -14 week old female Rip7rTta +/-::ZnT8Tg +/mice.

Quantitative real-time PCR
Total islet RNA was extracted using Trizol reagent (Invitrogen, Paisley). After reverse transcription, relative expression was assessed using SYBR Green (Invitrogen, Paisley).
Primers were designed using PerlPrimer and gene expression was normalised to β-actin (Actb). dilution, secondary-Alexa Fluor 568 1:500) and sealed using Vector Shield Antifade Hard Set reagent (Vector Laboratories). β cell mass was determined as described (51). Data capture was performed using a Zeiss AxioObserver and a 40x/0.75NA objective. β/α cell mass was calculated using the threshold plugin for ImageJ (NIH), as previously detailed (52).

Intraperitoneal Glucose and Insulin Tolerance Tests
Glucose (1 g /Kg body weight) was injected into the abdomen of mice that had been fasted overnight. Blood glucose measurements were taken at 0, 15, 30, 60, 90 and 120 min. using an automatic glucometer (Accucheck). Insulin tolerance tests are performed as per glucose tolerance test but animals were fasted for 5 h prior to 0.75 U insulin/Kg bodyweight insulin injection.

Plasma Insulin Measurements
Mice fasted overnight were injected with glucose (3 g glucose/Kg body weight) and blood from the tail vein was collected into heparin coated tubes (Sarstedt, Beaumont Leys, UK) at 0, 15 and 30 min. Plasma was separated by centrifugation at 2000g for 10 min, 5 μL of blood plasma was used to measure insulin levels using an ultrasensitive mouse insulin ELISA kit (CrystalChem, IL, USA).  (32).

ZIMIR imaging
ZIMIR imaging was performed as previously described (36). Briefly, isolated islets were incubated (37°C, 95% O2/5% CO2) in ZIMIR (1 μM) for 30 min and imaged in bicarbonate buffer solution supplemented with 1 μM EDTA to improve the signal-to-noise ratio. ZIMIR was excited at 491 nm and emitted signals captured at 525 nm. After acquisition, islets were divided into sub-regions before extraction of intensity over time to allow analysis of amplitude and area under the curve (AUC) of glucose-stimulated ZIMIR responses.

Cytosolic free Zn 2+ measurements
Zn 2+ measurements were acquired as previously described (32). Briefly, islets were dispersed onto coverslips before infection with adenovirus containing the FRET-based Zn 2+ sensor eCALWY4. Steady-state fluorescence intensity ratio citrine/cerulean (R) was measured, then maximum and minimum ratios were determined to calculate free Zn 2+ concentration using the following formula: [Zn 2+ ]=Kd(Rmax−R)/(R−Rmin). The maximum ratio (Rmax) was obtained upon intracellular zinc chelation with 50 μM TPEN and the minimum ratio (Rmin) was obtain upon Zn 2+ saturation with 100μM ZnCl2 in the presence of the Zn 2+ ionophore, pyrithione (5μM).

Transmission Electron Microscopy
Isolated islets were fixed in Vincenzo's fixative (2 % PFA, 2.5 % glutaraldehyde, 3 mM CaCl 2 , 0.1 M sodium cacodylate buffer (pH 7.4)) for 20 min at 37°C initially followed by a further 2 h at room temperature and finally overnight at 4°C. Electron microscopy was performed as previously described (53).

Statistical analysis.
Values represented are the mean ± SEM. Statistical significance was assessed using either Student's t-test or the Mann-Whitney U test depending on data distributions. Two-way ANOVA (with Bonferroni or Sidak multiple comparison test) was used to examine the effect of multiple variables. Statistical analyses were performed using Graph Pad Prism 6.0, ImageJ and IgorPro.
Values represent mean ± SEM.