Hyper-variability in Circulating Insulin and Physiological Outcomes in Male High Fat-fed Ins1-/-:Ins2+/- Mice in a Conventional Facility

Insulin is an ancient, multi-functional hormone with essential roles in glucose homeostasis and energy storage. Recently, our group has taken advantage of the ability to limit insulin secretion in vivo by reducing insulin gene dosage to demonstrate that insulin hypersecretion is a requirement for diet-induced obesity. Our previous studies employed male Ins1+/−:Ins2−/− mice that exhibit a complete inhibition of diet-induced hyperinsulinemia relative to Ins1+/+:Ins2−/− littermate controls, as well as female Ins1−/−:Ins2+/− mice with transient, partial reduction in circulating insulin relative to Ins1−/−:Ins2+/+ littermates. In the present study, we sought to extend these studies to male Ins1−/−:Ins2+/− mice on the same chow and high fat diets. Surprisingly, while reduced Ins2 gene dosage appeared capable of reducing Ins2 mRNA, insulin protein levels in these mice were not significantly reduced. Moreover, there was a marked hyper-variability in circulating insulin levels within and between two independent cohorts of mice that persisted over at least the first year of life. In Cohort 1, we observed a paradoxical increase in body weight in some high fat-fed male Ins1−/−:Ins2+/− mice relative to Ins1−/−:Ins2+/+ littermate controls. This phenomenon is consistent with the known satiety effects of insulin and our previous observations with Ins2 can be expressed in the brain. Collectively, our data reveal unexpected complexity associated with the Ins2 gene in male mice, and establish the Ins2 gene as a candidate for studying the effects of modifier genes and/or environmental influences on gene-to-phenotype variability. Further studies are required to define the molecular mechanisms of this phenotypic hyper-variability and to define the role of reduced Ins2 gene dosage in the brain.


Introduction
Insulin genes are highly conserved, playing critical roles in glucose homeostasis in all species 26 studied to date [1,2]. Unlike humans, mice have two insulin genes, Ins1 and Ins2 [3]. Most studies 27 have shown that Ins1 is restricted to pancreatic β -cells, where it contributes to approximately 1/3 of the 28 expressed and secreted insulin [3,4]. The peptide product of the Ins1 gene differs from that of the Ins2 29 gene by two amino acids in the β -chain, at the B9 and B29 location, and is missing two amino acids in 30 the connecting C-peptide [5]. Ins1 also lacks an intron present in Ins2 [5]. Ins2 is the ancestral gene, 31 with gene structure, parental imprinting, and a broad tissue distribution similar to human INSULIN [3, 32 4, 6]. Notably, there is evidence that both mouse Ins2 and human INSULIN are expressed at low levels 33 in within sub-populations of cells in the brain [4,7]. The two murine insulin genes are partially 34 redundant and capable of compensating for the loss of one another [8]. However, some studies, such as 35 those comparing the effect of the expression of the Ins1 versus Ins2 in the thymus in the context of type 36 1 diabetes, have shown that the two genes are not entirely redundant [9]. Outside of type 1 diabetes (i.e. 37 in conditions of relative normoglycemia), the effects of changed Ins gene dosage, and ultimately insulin 38 levels, remain to be fully elucidated. 39 Studies of human populations and animal models of obesity have demonstrated that elevated levels 40 of fasting insulin, known as hyperinsulinemia, precede weight gain [10][11][12][13][14][15][16][17][18][19][20][21][22][23]. Moreover, some studies 41 have suggested that humans with class I allele VNTR in the INSULIN gene produce and release more 42 insulin from the pancreatic islets and are also more susceptible to obesity [24][25][26], although this 43 observation remains controversial [27]. On the other hand, studies of invertebrates with reduced insulin 44 or insulin signalling have reported leaner, smaller bodies, along with increased lifespan [28,29]. 45 Similarly, studies in mammalian models, such as the Zucker fatty rats, have shown that treatment with 46 diazoxide, a compound that reduces insulin secretion, results in reduced weight and improved glucose 47 intolerance [30,31]. Treatment of obese patients with diazoxide is also associated with weight loss in some small clinical trials [32,33]. Lustig and his group found similar results using Octreotide, a 49 somatostatin agonist that binds the sst5 somatostatin receptor, found on β-cells, which inhibits insulin 50 release [34][35][36]. Therefore, such observations have raised the question of whether hyperinsulinemia 51 itself is a primary defect in obesity. Recently, our group has extended the observations that a full 52 complement of insulin genes appears to be required for substantial high fat diet-induced obesity in 53 mammals [4,37]. 54 Here, we report on the phenotype of male Ins1 -/-:Ins2 +/mice fed two different diets in a 55 conventional facility. Surprisingly, the effect of Ins2 gene dosage on circulating insulin peptide was 56 highly variable in these mice, displaying strong cohort dependence. This precluded definitive 57 conclusions about the effects of this gene on weight gain, but provide insight into the regulation and 58 effects of insulin production from this locus. 59 60 61

63
The Ins1 -/and Ins2 -/mice were previously generated by Jacques Jami (INSERM) and are described 64 elsewhere. [8] A neo cassette was used to disrupt the Ins1 gene and replace most of its sequence. A β -65 geo (Neo/LacZ) cassette was used to disrupt most of the Ins2 gene sequence. [

93
We placed 8-week old mice (N = 3-5) from each group in PhenoMaster indirect calorimetry cages 94 (TSE Systems Inc., Chesterfield MO) for three complete days. The cages also measured food, drink 95 and body weight as well as activity using infrared beam grid in the x, y and z axes. All cages were 96 contained in an environmental chamber to ensure constant temperature (21°C). The room's light cycles 97 were from 7 am -7 pm. Data collected from the first 4 hours were not included in the study. The 98 average of data collected from each of the 3 days were presented as a prototypical day for each 99 genotype, as in our previous publications [4,37]. 100 101

102
At one year of age, mice were euthanized for the purpose of tissue collection. The following tissues 103 were collected: pancreas, epididymal fat pads, soleus muscle, liver, brain, kidney, spleen, heart, thymus 104 and tibia. Some samples were snap-frozen in liquid nitrogen and stored in -80°C freezer. The rest of the 105 samples were fixed in 4% paraformaldehyde (PFA) for tissue sectioning. For removal of non-bone 106 tissue the tibias were incubated in 2% KOH for for physical measurements. Camera (Roper Scientific). Image analysis was done using the Slidebook software (Intelligent Imaging 118 Innovations) as previously described [38]. 119 120

121
Most results are expressed as means ± SEM. The area under the curve (AUC) was used to measure 122 statistical significance in different groups for the glucose tolerance and insulin secretion tests and is 123 described elsewhere [4]. The area over the curve (AOC) was used for insulin tolerance studies and are 124 also described in detail elsewhere [4]. SPSS 15.0 software or Prism 5 (Graphpad) software was used to 125 perform the statistical analyses. Two-way ANOVAs were used to compare factors of genotype and 126 diet, or alternatively in the case of significant interactions, one-way ANOVAs with Bonferroni 127 corrections were used. We used Levene's test to validate homogeneity of variance. In all statistical 128 analyses, if p < 0.05 the differences were considered to be significant. was not statistically different between genotypes, but showed a tendency to be decreased in Ins1 -/-138 :Ins2 +/mice compared to littermate control Ins1 -/-:Ins2 +/+ mice (p = 0.095; Fig. 1C). 139 Immunofluorescent staining of islets from year-old mice (Fig. 1D) mirrored the measurements of 140 insulin protein in isolated islets at 8 weeks (Fig. 1C). We did not detect significant differences in 141 pancreatic β -cell area between groups (Fig. 1E), in contrast to Ins1 +/-:Ins2 -/male mice in our previous 142 report [4]. year-old Ins1 -/-:Ins2 +/mice compared to Ins1 -/-:Ins2 +/+ mice (Fig. 1F). Fasting insulin was also 160 significantly different between cohorts at 8 and 52 weeks of age (Fig. 1F). Collectively, these data  (Fig. 1F), there were significant differences between cohorts for glucose-stimulated insulin 170 secretion values at 7 and 52 weeks of age. However, in spite of differences in value magnitude, there 171 tended to be similar patterns of the effects of diet and genotype across both cohorts for stimulated 172 insulin secretion and blood glucose response to intraperitoneal glucose or insulin, and so the cohorts 1 0 were pooled for these data. Glucose-stimulated insulin secretion was higher in high fat-fed mice 174 compared to chow-fed mice at 7 weeks of age, but there were no statistically significant differences in 175 stimulated insulin secretion detected between Ins1 -/-:Ins2 +/mice and their Ins1 -/-:Ins2 +/+ littermate 176 controls at any time point (Fig. 2B). This indicates that a single allele of the Ins2 gene is sufficient to 177 generate enough insulin to mount an appropriate response to glucose in these male mice, unlike the 178 previously described male Ins1 +/-:Ins2 -/mice [4] or female Ins1 -/-:Ins2 +/mice [37]. Effects of genotype 179 on glucose tolerance were very modest, with Ins1 -/-:Ins2 +/mice exhibiting slightly worsened glucose 180 tolerance than their Ins1 -/-:Ins2 +/+ littermates only at 25 weeks (Fig. 2C). In addition, a minimal degree 181 of HFD-induced glucose intolerance was observed by 52 weeks of age (Fig. 2C). Interestingly, high 182 fat-fed Ins1 -/mice (regardless of Ins2 gene dose) were paradoxically insulin hypersensitive at all of the 183 time points from 12 weeks onward (Fig. 2D). On chow diet, 6 week-old mice with reduced Ins2 gene 184 dosage tended to be hypersensitive to exogenous insulin, but this was not observed in older mice (Fig.  185   2D).

1
Ins1 -/-:Ins2 +/+ mice did not gain weight on a high fat diet (Fig. 3A), leading us to speculate that the 198 hypersecretion of the Ins1 gene product could have been required for the proper storage of lipids in 199 adipose tissue. Remarkably, however, the majority of Ins1 -/-:Ins2 +/mice in the first cohort exhibited 200 striking weight gain on the high fat diet (Fig. 3A). Surprisingly, these somewhat paradoxical 201 differences were not observed in a second cohort (Fig. 3B). Together, these data illustrate a hyper-202 variability of the physiological response to reduced Ins2 gene dosage in these male mice. Ins1 -/mice (Fig. 4A), reminiscent of mice lacking adipocyte insulin receptors [39,40]. Circulating 214 leptin tended to be proportional to epididymal fat pad weight and fat-to-lean ratio (Figs. 4 B,C,F). We 215 did not detect differences in the circulating free fatty acids between groups in this cohort (Fig. 4D). 216 Together, these data suggest that, in some currently un-defined conditions, reducing Ins2 expression 217 can be associated with facilitating greater weight gain, via a mechanism that is active specifically in the 218 context of a high fat diet. 219 220 Given the statistical similarity between genotypes for the average circulating insulin levels prior to 228 one year of age, these observations hint at the possibility of altered local effects of Ins2 in the brain. It 229 is well established that insulin can act in the brain as a satiety factor [41,42]. We have confirmed that 230 Ins2 is expressed in multiple regions of the brain that can potentially control and project to feeding, 231 reward and memory centers, raising the possibility that central Ins2 gene expression may regulate food 232 intake [4]. In this first cohort, where reduced Ins2 gene dosage was associated with increased weight 233 gain on high fat diet, there was a tendency for Ins1 -/-:Ins2 +/mice to show elevated food intake on HFD 234 versus CD, whereas lean Ins1 -/-:Ins2 +/+ littermates tended to to reduce caloric intake in response to high 235 fat feeding (Fig. 5A). Other parameters such as activity and energy expenditure were not statistically 236 different between any of the groups (Figs. 5B-E). Taken together, these experiments indicate that, 237 within this cohort, reduced Ins2 gene dosage led greater weight gain on high fat diet, which may have 238 been associated with increased relative food intake. To our surprise, reduced Ins2 gene dosage did not translate into consistent differences in circulating 249 insulin. We observed a large degree of variability within and between two cohorts of mice. These 250 observations are in contrast to our experience with female mice of the same genotypes on the same 251 diets, and in contrast with our experience reducing Ins1 gene dosage in male mice [4,37]. 252 Insulin is the most studied hormone in biology, yet the present study provides new insight into the 253 range of physiological functions that can be modulated by insulin. Specifically, our data point to 254 somewhat unique roles for, and post-transcriptional regulation of, the Ins2 gene, relative to the Ins1 255 gene. Elegant studies in flies and worms have demonstrated that deleting specific insulin genes 256 increases lifespan and prevents diseases associated with adiposity, and clearly indicate that individual 257 insulin-like peptide genes have distinct physiological functions despite signalling through a single 258 receptor [28,29]. It has been previously shown that the mouse Ins1 and Ins2 genes have opposing 259 effects on type 1 diabetes incidence in the NOD mouse due to the induction of thymic tolerance by 260 Ins2, [43,44]. Specifically, Ins2 expression was associated with reduced incidences of type 1 diabetes 261 and the reverse was true for the expression of the Ins1 gene [43,44]. Our previous observation that the 262 pancreatic-specific Ins1 is dose-dependently required for diet-induced obesity in male mice [4], defines 263 the first specific role for Ins1 outside the context of type 1 diabetes. Similarly, a follow-up study 264 revealed that a modest and transient reduction in circulating insulin in female Ins1 -/-:Ins2 +/mice 265 relative to Ins1 -/-:Ins2 +/+ littermate controls was sufficient to provide long-term protection from diet-266 induced obesity [37]. In the present study, the high fat diet was unable to consistently increase fasting 267 insulin or beta-cell mass in a statistically significant manner. This meant that we were unable to 1 4 formally test the hypothesis that reducing hyperinsulinemia by reducing Ins2 gene dosage might protect 269 these mice from high-fat diet-induced obesity, as we could in our previously published studies [4,37]. 270 Clearly, additional studies with greater statistical power and more diet groups would be required to 271 formally rule in or out a role for hyperinsulinemia stemming from the Ins2 gene in diet-induced 272 obesity. 273 One interesting observation from the present study was that male Ins1 +/-:Ins2 -/from the first cohort 274 were heavier than their Ins1 +/-:Ins2 -/littermates fed the same high-fat diet. Considerable mouse-to-275 mouse heterogeneity was observed in these measurements, but they were relatively consistent within 276 each mouse over time. Taken at face value, these observations suggest the possibility that Ins2 277 expression in the brain made have played a role in this weight gain, because only the Ins2 gene is 278 robustly expressed in the brain [4], and because we only detected modest and late-onset differences 279 between genotypes in circulating insulin levels. The central nervous system is known to play important 280 roles in peripheral energy homeostasis and body weight regulation [45][46][47]. For example, it has been 281 reported that insulin receptor knockout in the brain leads to increased high fat food intake and obesity 282 [48], suggesting the possible presence of a local signalling network. The presence of small amounts of 283 insulin protein and mRNA in the mammalian brain has long been reported [49](reviewed in [4]). The 284 production of Ins2 mRNA and protein in specific brain regions was confirmed by our group using Ins1 -285 /and Ins2 -/as negative controls, and Ins2 βGal/+ mice as positive staining controls [4]. 286 Since insulin has been proposed to be a satiety factor [50], downregulation of insulin expression 287 and/or action may be expected to increase food intake. Consistent with this notion, Ins1 -/-:Ins2 +/mice 288 in our first cohort tended to show increased high fat food intake and were obese when compared to the 289 high fat-fed Ins1 -/-:Ins2 +/+ littermate controls. Although our experimental treatment was expected to 290 reduce Ins2 gene dosage in multiple tissues, including the pancreas, thymus and brain, several lines of 291 evidence could potentially hint that a partial reduction of brain Ins2 was associated with the diet-292 dependent hyperphagia and obesity, at least in some specific conditions. Other investigators have 293 shown that Ins2 knockout in the thymus had no effect on body weight [51], further suggesting that the 294 brain could have play a role in this phenomenon. Our work should open new avenues for investigating 295 the biology of insulin in neurons and their connections. However, it is imperative that these data are 296 interpreted with caution, as the second cohort of Ins1 -/-:Ins2 +/mice did not gain additional weight with 297 high fat feeding. A complete understanding of the role of Ins2 in the central nervous system will 298 require conditional Ins2 knockout mice. 299 The phenotypes we observed in the present study were hyper-variable, sex-specific, and diet-300 dependent with respect to reduction of the Ins2 gene. The animal facility where this work was done no 301 longer exists, so it is impossible to formally repeat these studies in the same environment. Similar 302 studies were also conducted in a modern specific-pathogen free facility, and cohort-dependent 303 variability is described in a companion paper. Remarkably, in this distinct animal facility we observed 304 opposite outcomes for the effect of Ins2 gene dosage on weight gain than those which were shown in 305 the current experiments. Collectively, the experience of our laboratory is that modulation of the Ins2 306 gene, in the absence of Ins1, leads to striking variability in body weight. The source of this variability 307 will be investigated by our collaborators specializing in epigenetics and translational control.   Week 5 Week 11 Week 25 Week 50 Week 6 Week 12 Week 26 Week 51 Week 13 Week 27 Week 52