Hyper-variability in Circulating Insulin Levels and Physiological Outcomes to High Fat Feeding in Male Ins1−/−:Ins2+/− Mice in a Specific Pathogen-free Facility

Insulin is an essential hormone with key roles in energy homeostasis and body composition. Mice and rats, unlike other mammals, have two insulin genes: the rodent-specific Ins1 gene and the ancestral Ins2 gene. The relationships between insulin gene dosage and obesity has previously been explored in male and female Ins2−/− mice with full or reduced Ins1 dosage, as well as in female Ins1−/− mice with full or partial Ins2 dosage. We report herein unexpected hyper-variability in circulating insulin and physiological responses to high fat feeding in male Ins1−/−:Ins2+/− mice. Two large cohorts of Ins1−/−:Ins2+/− mice and their Ins1−/−:Ins2+/+ littermates were fed chow diet or high fat diet (HFD) from weaning and housed in specific pathogen-free (SPF) conditions. Cohort A and cohort B were studied one year apart. Contrary to female mice from the same litters, inactivating one Ins2 allele on the complete Ins1-null background did not cause a consistent reduction of circulating insulin in male mice. In cohort A, HFD-fed males showed an equivalent degree of insulin hypersecretion and weight gain, regardless of Ins2 dosage. In cohort B, Ins1−/−:Ins2+/− males showed decreased insulin levels and body mass, compared to Ins1−/−:Ins2+/+ littermates. While experimental conditions were held consistent between cohorts, we found that HFD-fed Ins1−/−:Ins2+/− mice with lower insulin levels had increased corticosterone. Collectively, these observations highlight the hyper-variability and range of phenotypic characteristics modulated by Ins2 gene dosage, specifically in male mice.


Introduction
Variations in circulating insulin levels have far-reaching metabolic consequences. In addition to the 20 expected fluctuations in insulin secretion that are associated with blood glucose, circulating insulin 21 levels are affected by a number of hormones and circulating factors, including amino acids, fatty acids, 22 estrogen, melatonin, leptin, growth hormone, glucose-dependent insulinotropic polypeptide, and 23 glucagon-like peptide-1 (see [1]). In mice, the mean 5-h fasted insulin levels in non-obese, 12 week-old 24 males can range from 0.5 to 1.2 ng/mL, across four commonly used strains [2]. In humans, fasting 25 insulin levels can range from 0.04 to 3.43 ng/mL in a nondiabetic adult population [3,4], and evidence 26 suggests that less than half of the variance in fasting insulin can be can be accounted for by genetic 27 variability [5,6]. 28 Mice and rats have two non-allelic insulin genes, with a rodent-specific Ins1 gene that likely arose 29 from the transposition of a reverse-transcribed, partially processed mRNA of the ancestral Ins2 [7]. 30 Ins1 and Ins2 genes reside on different chromosomes in mice [8]. While Ins1 lacks one of the two 31 introns found in Ins2, the murine Ins genes share high homology up to 500 base pairs preceding the 32 transcription initiation site [9]. Ins1 and Ins2 have distinct promoter elements, tissue-and temporal-33 specific expression patterns, and imprinting status [10][11][12][13][14]. In addition, differential translation or 34 processing rates of the two murine preproinsulins have been reported [15,16]. It is therefore possible 35 that levels of the fully processed murine insulin 1 and insulin 2 peptides are divergently susceptible to 36 modulation under various conditions, which could underlie the evolutionary retention of both genes 37 [17]. When one insulin gene is inactivated, elevated transcript and protein level of the non-deleted 38 insulin gene can at least partially compensate for the loss [18], although the exact nature of this 39 reciprocal relationship remains understudied. 40 We have performed a series of investigations to examine how murine Ins1 and Ins2 gene dosage 41 impacts the onset of high fat diet-induced hyperinsulinemia and the development of obesity. Previous 42 work in our laboratory showed that reducing Ins1 gene dosage (on an Ins2-null background) results in 43 continuous suppression of fasting hyperinsulinemia in male mice, thereby preventing diet-induced 44 obesity [14]. Interestingly, circulating insulin levels were not similarly modulated in the female 45 littermates from this study [14], suggesting the possibility of sex-specific differences in the relationship 46 between insulin gene dosage and circulating insulin levels. In the converse genetic manipulation, 47 reduced Ins2 dosage (on an Ins1-null background) led to high-fat fed female Ins1 -/-:Ins2 +/mice having 48 lower insulin secretion than their Ins1 -/-:Ins2 +/+ controls at a young age, which again corresponded with 49 attenuated obesity [19]. The phenotype of the female Ins1 -/-:Ins2 +/mice was highly consistent between 50 the two large cohorts of animals studied under specific pathogen-free (SPF) conditions, and was 51 congruent to preliminary evaluations in a conventional facility. 52 We report herein on circulating insulin levels and the metabolic phenotype of the male Ins1 -/-:Ins2 +/-53 and Ins1 -/-:Ins2 +/+ littermates of the female mice that were the subject of our recent investigation [19]. 54 Contrary to our expectations, inactivating one Ins2 allele did not cause a consistent reduction of 55 circulating insulin in Ins1-null male mice, which precluded us from properly testing the hypothesis that 56 reduced Ins2 dosage and lower insulin levels would lead to protection from obesity in males. 57 Specifically, we report that across cohorts, the effects of high fat feeding on glucose homeostasis, 58 insulin sensitivity, and weight gain in Ins1 -/-:Ins2 +/male mice varied widely. Moreover, circulating 59 insulin levels were hyper-variable across cohorts in Ins1-null male mice, pointing to sex-specific 60 compensation of insulin homeostasis in these animals. Differences in degree of insulin compensation 61 were associated with corticosterone levels, a marker of stress. Together with the accompanying study, 62 we demonstrate that there is phenotypic hyper-variability within two different animal facilities (one 63 conventional and one SPF). Collectively, these reports along with our previous published work [14,19]  alleles were generated previously [20] and were roughly equal parts C57BL/6 and 129 background. 74 Data presented here are from the male littermates of previously described Ins1 -/-:Ins2 +/+ and Ins1 -/-75 :Ins2 +/females [19], and were therefore collected in the same time-frames and conditions. All animals 76 were predominately handled by the same female researcher. The mice tracked across time were in two 77 major cohorts, born a year apart (cohort A, born October 2011 -December 2011, and cohort B, born 78 October 2012 -February 2013; Fig. 1a (Fig. 1a). Diet assignments were distributed within each litter, based on approximately 85 matching starting body weights between diet groups. Mice were housed under SPF conditions at 21ºC, 86 on a 12:12 h light:dark cycle, in the same room for both cohorts. The vast majority of male mice from 87 both cohorts were individually housed, due to fighting between young cage-mates. 88 most analyses, we used two-way analysis of variance (ANOVA) models within each cohort to assess 156 factors of genotype and diet, and a significant interaction led to one-way ANOVAs comparing HFD-157 fed Ins1 -/-:Ins2 +/+ mice, CD-fed Ins1 -/-:Ins2 +/+ mice, HFD-fed Ins1 -/-:Ins2 +/mice, and CD-fed Ins1 -/-158 :Ins2 +/mice, with Bonferroni corrections. Three-way ANOVAs were used for incorporating the 159 additional factor of cohort or parental effect, and 2-tailed independent t-tests were used to assess 160 differences if there were only two groups to be compared. Analysis of covariance was employed to test 161 energy expenditure with covariates of lean and fat mass. In all cases, we used Levene's test to evaluate 162 the assumption of homogeneity of variance, and where the assumption was violated, logarithmic 163 transformations were applied, generally stabilizing data variance. Graphpad Prism 6.0 software was 164 used to generate and assess the linear regressions. 165 166 168 We have recently studied the effects of reducing Ins2 gene dosage in Ins1-null female mice, and 169 found consistent outcomes for insulin secretion and body composition across cohorts [19]. While 170 characterizing male siblings from the same cohorts, we noticed that a number of measured parameters 171 showed dramatic inconsistencies between cohorts A and B, precluding us from pooling the data from 172 these two cohorts. For instance, we observed that in cohort A males, fasting insulin was significantly 173 higher in HFD-fed mice than CD-fed mice at all measured time points across a year (Fig. 1b), and 174 glucose-stimulated insulin secretion was higher at 8, 15, and 52 weeks of age (Fig. 1c). Unexpectedly, 175 there were no significant differences in circulating insulin levels (either fasting or glucose-stimulated) 176

Variability in Insulin Secretion Between Cohorts of Male Mice
between Ins1 -/-:Ins2 +/+ and Ins1 -/-:Ins2 +/mice, at any point up to one year in cohort A (Figs. 1b,c). In 177 contrast, cohort B Ins1 -/-:Ins2 +/mice tended to have lower fasting insulin levels than their Ins1 -/-:Ins2 +/+ 178 littermates (Fig. 1b), and at 27 weeks of age they had significantly reduced glucose-stimulated insulin 179 secretion (Fig. 1d). In addition, Ins1 -/-:Ins2 +/+ mice were the only cohort B males showing more 180 glucose-stimulated insulin secretion on HFD than CD at the early 8-week time point (Fig. 1d). 181 However, the HFD-induced elevation of basal and glucose-stimulated insulin secretion was, in general, 182 quite modest for most males in cohort B (Figs. 1b,d). Notably, by 52 weeks all groups of cohort B male 183 mice clearly had lower average insulin levels than cohort A males (Fig. 1b). 184 To evaluate potential mechanisms underlying cross-cohort changes in circulating insulin levels in 185 male mice, we assessed insulin mRNA, protein, and islet function in two distinct groups. Due to the 186 longitudinal nature of the experiments, we could not use islets from cohort A or B mice. However, a 187 separate group of 25 week-old male mice tended to show a genotype effect for in vivo fasting insulin (p 188 = 0.056), as well as similar raw insulin values to those of cohort B mice (Figs. 1b, 2a). Islets from these 189 25 week-old Ins1 -/-:Ins2 +/mice had an expected reduction in Ins2 mRNA compared to Ins1 -/-:Ins2 +/+ 190 islets (Figs. 2b). Interestingly, the significant reduction in Ins2 mRNA did not correspond to significant 191 genotype differences in islet insulin protein content (Fig. 2c), suggesting the involvement of post-192 transcriptional compensation. Consistent with the lack of a difference between genotypes for islet 193 insulin content, dynamic secretion by islets from 25 week-old HFD-fed Ins1 -/-:Ins2 +/mice was not 194 reduced compared to Ins1 -/-:Ins2 +/+ islets, and in fact 25 week-old Ins1 -/-:Ins2 +/islets appeared to have 195 the capacity for a marginally increased 2 nd phase response to KCl stimulation (Fig. 2d). We also evaluated 70 week-old islets from a group of HFD-fed mice that had shown no obvious 213 genotype differences in fasting insulin levels at 52 weeks of age (Fig. 2e). This allowed us to confirm 214 that the genetic manipulation also led to reduced Ins2 mRNA in older HFD-fed Ins1 -/-:Ins2 +/male mice 215 (Fig. 2f), with a similar capacity for compensation at the level of islet insulin content (Fig. 2g) as was 216 evident at 25 weeks (Fig. 2c). The only detected differences in dynamic in vitro islet secretion were 217 minimal, showing that at 70 weeks, Ins1 -/-:Ins2 +/islets did not secrete quite as much insulin in the 2 nd 218 phase of glucose stimulation as Ins1 -/-:Ins2 +/+ islets (Fig. 2h). Collectively, these data show that 219 although the experimental genetic manipulation did successfully reduce Ins2 expression in the male 220 mice, there was evidence of compensation with respect to insulin protein content and islet secretory 221 capacity. The capability for insulin production and/or secretory compensation may have accounted for 222 the lack of consistent differences in circulating insulin levels between Ins1 -/-:Ins2 +/+ and Ins1 -/-:Ins2 +/-223 male mice. 224 225

226
We also observed heterogeneity between cohorts in a longitudinal analysis of glucose homeostasis, 227 as might be expected from cross-cohort variability in insulin levels. In conjunction with the sustained 228 HFD-induced elevation of circulating insulin in cohort A mice, from 14 weeks onwards HFD-fed males 229 from cohort A showed significant whole-body insulin resistance compared to their CD-fed littermates 230 (Fig. 3a). HFD-fed mice in cohort A also had a modest degree of glucose intolerance compared to mice 231 on CD, at 8 and 50 weeks (Fig. 3b). In contrast, there were no statistically significant differences in 232 insulin sensitivity or glucose tolerance observed between CD-and HFD-fed groups of cohort B mice, 233 consistent with a limited response to HFD-feeding (Figs. 3c,d). Ins1 -/-:Ins2 +/+ and Ins1 -/-:Ins2 +/cohort B mice did not appear to cause robust negative repercussions for 246 glucose homeostasis (Figs. 3c,d). However, in cohort A, Ins1 -/-:Ins2 +/mice were slightly but 247 significantly more glucose intolerant than their Ins1 -/-:Ins2 +/+ littermate controls at each measured time 248 point across a year. Closer examination of the responses to intraperitoneal glucose stimulation shows a 249 trend for a delayed or sustained peak in blood glucose in cohort A Ins1 -/-:Ins2 +/male mice (Fig. 3b). Ins2 allele did not notably affect either circulating insulin levels (Figs. 1b,c) or HFD-induced growth 281 ( Fig. 4c) in cohort A males, since all HFD-fed mice showed equivalent weight gain compared to 282 CD-fed littermates, particularly from 20 weeks onward (Fig. 4c). Although young Ins1 -/-:Ins2 +/mice 283 were smaller than their Ins1 -/-:Ins2 +/+ littermates for a limited period in cohort A, this did not persist 284 (Fig. 4c). HFD-fed mice were heavier than CD-fed mice for a similar duration in cohort B as in cohort 285 A, but the significantly smaller mass of cohort B Ins1 -/-:Ins2 +/mice compared to their Ins1 -/-:Ins2 +/+ 286 littermates continued throughout most of the year (Fig. 4d). Notably, a difference in body mass 287 between Ins1 -/-:Ins2 +/+ and Ins1 -/-:Ins2 +/mice of cohort B was detectable in male pups as young as 2 288 days of age (Fig. 4e), and it is possible that there could have also been similar size differences in 289 neonatal pups of cohort A (not measured) that might have contributed to the genotype effect on body 290 mass observed in young cohort A mice (Fig. 4c). 291 We attempted to evaluate different factors that could have contributed to phenotypic differences 292 between cohorts A and B. We did not detect obvious means by which parental imprinting of the Ins2 293 allele might have accounted for the observed variability, as both breeding pair options (i.e. either the 294 dam or the sire donating the disrupted Ins2 allele) contributed to cohort A and B experimental animals. 295 There were no statistically significant differences between offspring of dams versus sires with the 296 disrupted Ins2 allele when the factor of "parental effect" was incorporated into analyses of body mass 297 or fasting insulin levels (with both cohorts pooled); any patterns suggestive of a possible parental effect 298 in one cohort were either not present or showed opposite trends in the other cohort (data not shown). 299 There was also no apparent distinction in the mean or distribution of litter sizes between cohorts A and 300 B (Fig. 4f). 301 As a limited indicator of environmental stressors, we also measured 4 h-fasted corticosterone levels 302 at 27 weeks. Interestingly, there tended to be a greater number of mice with elevated corticosterone 303 levels in cohort B, and modest trends suggested that there were higher average corticosterone levels in 304 HFD-fed Ins1 -/-:Ins2 +/males from cohort B, compared to cohort A (Fig. 4g). Closer examination of 305 HFD-fed animals showed that across both cohorts, fasting insulin levels in HFD-fed Ins1 -/-:Ins2 +/male 306 mice were inversely correlated to corticosterone levels (r 2 = 0.38, p < 0.01), and this relationship was 307 not evident in HFD-fed Ins1 -/-:Ins2 +/+ mice (r 2 = 0.00, p = 0.95; Fig. 4h). Interpretation of these data is 308 limited, as it is based on using a single measurement of corticosterone for each individual animal as a 309 marker of 'stress' during an extended time period. However, it appears that those HFD-fed 310 Ins1 -/-:Ins2 +/male mice that reached the highest fasting insulin levels at 27 weeks (predominately 311 individuals from cohort A) tended to have lower corticosterone levels, whereas increased stress may 312 have dampened the capacity of HFD-fed Ins1 -/-:Ins2 +/male mice to compensate for reduced Ins2 313 dosage at the level of basal insulin secretion. If mice in cohort B did experience more stressful 314 conditions than cohort A animals (Fig. 4g), this could be one of potentially many contributing factors 315 underlying phenotypic differences in circulating insulin levels between cohorts A and B. 316 Heterogeneity in our data precluded pooling cohorts A and B to generate averaged results. However, 317 although experimental genetic and dietary manipulations did not have consistent effects in both 318 cohorts, we did observe a comparable range of fasting insulin values and body masses for both cohorts 319 (Figs. 1b, 4c,d). To indirectly evaluate whether differences in fasting insulin might be underlying body 320 weight alterations in this model, we examined the relationship between these two variables across all 321 year-old mice. There was a positive correlation between body mass and fasting insulin levels at this age 322 (r 2 =0.55, p<0.0001; Fig. 4i). Therefore, while we did not observe consistent effects of reducing Ins2 323 gene dosage on circulating insulin and obesity in male mice, in general, these data support the concept 324 that reduced insulin levels are associated with attenuated body weight and obesity. 325 326

327
We further characterized cohort B mice, as they showed a sustained divergence in body mass 328 between Ins1 -/-:Ins2 +/and Ins1 -/-:Ins2 +/+ groups (Fig. 4d). First, we used metabolic cages to examine 329 the in vivo energy balance of HFD-fed mice at 17 weeks, an age when both cohorts showed similar 330 trends for body masses. Although HFD-fed Ins1 -/-:Ins2 +/mice exhibited a slight elevation in activity 331 levels during the early hours of the dark period (Fig. 5a), it did not appear that there were genotype 332 differences in whole-body energy expenditure (Fig. 5b), respiratory exchange ratio (Fig. 5c), or food 333 intake (Fig. 5d) to account for the disparities in weight gain between HFD-fed Ins1 -/-:Ins2 +/and Ins1 -/-334 :Ins2 +/+ males. 335 We assessed body composition longitudinally with DEXA. Consistent with the evidence suggesting 346 that lower Ins2 dosage led to generally reduced growth in male neonatal pups (Fig. 4e), reductions in 347 both adiposity and fat-free mass contributed to the smaller size of cohort B Ins1 -/-:Ins2 +/male mice. 348 Ins1 -/-:Ins2 +/+ animals had significantly higher fat mass than Ins1 -/-:Ins2 +/males on both diets, at all 349 measured time points (Fig. 6a). A similar pattern was also evident for fat-free masses, particularly at 350 the older ages (Fig. 6b). However, since reductions in fat mass were proportionally greater than 351 reductions in fat-free mass for Ins1 -/-:Ins2 +/versus Ins1 -/-:Ins2 +/+ males up to 60 weeks of age (Figs.  352   6a,b), it is clear that an attenuation in adiposity contributed to the reduced body mass of cohort B 353 Ins1 -/-:Ins2 +/male mice, compared to their Ins1 -/-:Ins2 +/+ controls. In the group of 25 week-old males 354 with similar insulin levels as cohort B mice (Figs. 1b, 2a), both subcutaneous and visceral white 355 adipose tissue depots were smaller in Ins1 -/-:Ins2 +/versus Ins1 -/-:Ins2 +/+ mice, in addition to the reduced 356 sizes of these depots in CD-mice compared to HFD-fed mice (Fig. 6c). Furthermore, the Ins1 -/-:Ins2 +/-357 mice had smaller interscapular brown adipose tissue depots than their Ins1 -/-:Ins2 +/+ littermates, and a 358 slight reduction in the size of a mixed triceps surae muscle group (Fig. 6c). Ins1 -/-:Ins2 +/+ littermates, we did not observe notable genotype differences in the fasting levels of 374 circulating lipids and metabolic factors at 40 weeks of age. All HFD-fed mice had higher cholesterol 375 and non-esterified fatty acids than CD-fed animals, without significant differences in triglycerides 376 levels ( Fig. 6d-f). In spite of differences in adipose tissue mass, leptin was similarly elevated in all 377 cohort B HFD-fed mice compared to CD-fed mice (Fig. 6g), as was resistin (Fig. 6h). Based on levels 378 of interleukin 6, there did not appear to be significant differences in inflammatory state between groups 379 (Fig. 6i). Glucose-dependent insulinotropic polypeptide levels were similarly elevated in all HFD-fed 380 mice, compared to CD-fed animals (Fig. 6j), but no significant differences were detected in 381 concentrations of peptide YY (Fig. 6k). These data suggest that while a combination of attenuated 382 adiposity and reduced fat-free mass contributed to the smaller body weights of cohort B Ins1 -/-:Ins2 +/-383 males compared to their Ins1 -/-:Ins2 +/+ littermate controls, cohort B HFD-fed Ins1 -/-:Ins2 +/males 384 nonetheless displayed many of the expected characteristics of high fat feeding. 385

388
The initial aim of our work was to test the hypothesis that reducing Ins2 gene dosage on an Ins1-null 389 background would prevent HFD-induced hyperinsulinemia, and thereby protect against obesity in male 390 mice. Contrary to our expectations, inactivating one Ins2 allele did not cause a consistent reduction of 391 circulating insulin in Ins1-null male mice -not even the transient suppression of insulin hypersecretion 392 that was consistently evident in their female Ins1 -/-:Ins2 +/littermates at a young age [19]. We report 393 that under some conditions, Ins1-null males with reduced Ins2 mRNA were capable of producing 394 nearly equivalent circulating insulin levels as Ins1 -/-:Ins2 +/+ males, albeit possibly with subtle 395 differences in secretory patterns that could have contributed to modest glucose intolerance. This clearly 396 distinguishes these Ins1 -/-:Ins2 +/males from the Ins2-null male mice with reduced dosage of the Ins1 397 gene, as Ins1 +/-:Ins2 -/male mice experienced a sustained suppression of hyperinsulinemia [14]. Our 398 findings show that Ins1 -/-:Ins2 +/male mice exhibit phenotypic hyper-variability across cohorts with 399 respect to insulin levels, glucose homeostasis, and weight gain with chronic high fat feeding. 400 In the current study, all HFD-fed mice in the first experimental cohort, cohort A, showed notable 401 insulin hypersecretion and weight gain, without significant effects of reduced Ins2 dosage. In contrast, 402 cohort B tended towards a less pronounced degree of HFD-induced insulin hypersecretion and 403 peripheral insulin resistance. In addition, in cohort B there seemed to be a sustained reduction in insulin 404 levels and body mass in Ins1 -/-:Ins2 +/mice compared to their Ins1 -/-:Ins2 +/+ littermate controls, without 405 detected changes in food intake or energy expenditure. These two cohorts from the same colony were 406 studied approximately one year apart, under consistent experimental conditions in a controlled SPF 407 facility. Despite this, by one year of age the average differences in fasting insulin levels between the 408 two cohorts were considerably more pronounced than the difference between having one or two 409 functional Ins2 alleles (in either cohort). 410 It is important to note that pronounced phenotypic variability between cohorts of Ins1 -/-:Ins2 +/male 411 mice, particularly with respect to body mass, was also evident within another animal facility. We 412 cannot explain the widely diverse phenotypic responses to reduced Ins2 gene dosage in male mice. 413 However, it is clear that cross-cohort phenotypic variability in Ins1 -/-:Ins2 +/males has been observed in 414 two distinct facilities to a degree that was not observed in their female littermates [19], nor in Ins2-null 415 male or female mice with full or partial Ins1 expression [14], despite the fact that these similar mouse 416 models were studied by our group in the same time frames and under similar conditions [14,19]. 417 Therefore, in Ins1-null male mice, the phenotypic outcomes of Ins2 gene modulation appear to be 418 susceptible to a wide range in variability. 419 Phenotypic variability, in general, is poorly understood, but likely affects many long-term animal 420 studies. There is evidence that the in utero and neonatal environments (e.g. [24][25][26][27]), gut microbiome 421 composition (e.g. [28][29][30]), and exposure to different stressors, including temperature (e.g. [31][32][33]), 422 noise (e.g. [34,35]), social hierarchy (e.g. [36,37]), and even the sex of the researchers working with 423 animal subjects [38], can have far-reaching effects on many physiological parameters. Additional 424 considerations include animal background strain or sub-strain, and genetic drift within a colony (e.g. 425 [2,[39][40][41][42]). These variables can confound experimental results through such means as altering the 426 endocrine milieu, or changing gene expression levels, directly or via the epigenome. However, it 427 should be noted that we attempted to control for many of these potentially confounding variables at a 428 reasonable level in our investigation. 429 There are numerous other factors that may have played a part in the observed phenotypic 430 heterogeneities in our investigation. For instance, the murine Ins2 gene has been shown to be subject to 431 developmental stage-dependent and tissue-specific genomic imprinting [11,43,44]. We considered the 432 possibility that the disrupted Ins2 allele may have had variable effects depending on whether it was 433 inherited from the maternal or the paternal side, particularly as our mice lacked the potential for 434 compensatory Ins1 expression. Although we did not observe obvious, consistent parental effects on the 435 experimental animals, genomic imprinting is a complex system that has not yet been fully elucidated 436 (see [44][45][46]), and a potential role cannot be fully ruled out. 437 As the experimental cohorts were separated temporally, another potential explanation for cross-438 cohort variability is subtle environmental changes across the years (although the cohorts were roughly 439 matched for seasons, since they were approximately one year apart). There were no differences 440 between cohorts with respect to the average number of siblings sharing their in utero and neonatal 441 environments, nor in numerous controlled parameters. However, in long-term experiments it is not 442 always possible to avoid such environmental perturbations as minor earthquakes, construction periods 443 around a facility, and so forth. At 27 weeks, an age when it was becoming clear that the two cohorts 444 were diverging in their patterns of insulin secretion and weight gain, a single blood sample per mouse 445 provided limited indication that levels of the stress hormone corticosterone might have been slightly 446 higher in cohort B males, compared to mice from cohort A. Effects may vary depending on duration or 447 type of stressor, but there is evidence that chronic stress [47,48] or glucocorticoid exposure itself [49-448 52] can lead to reduced insulin secretion in rodents. Interestingly, the glucocorticoid receptor has been 449 shown to bind to a negative regulatory element of the human INS gene [53]. In our results, there was a 450 negative correlation between basal insulin levels and corticosterone across both cohorts in HFD-fed 451 Ins1 -/-:Ins2 +/males. Interestingly, an under-powered, rudimentary assessment of plasma samples 452 available from prior experiments showed that 27 week-old HFD-fed Ins1 -/-:Ins2 +/+ and Ins1 -/-:Ins2 +/-453 males in a conventional animal facility (see accompanying article) tended to have even higher 454 corticosterone levels and lower fasting insulin levels (data not shown) than in the SPF facility. We 455 suggest that in the current study, reduced exposure to stress, signified by decreased plasma 456 corticosterone, may have partially accounted for some HFD-fed Ins1 -/-:Ins2 +/male mice having the 457 ability to produce nearly equivalent amounts of fasting insulin as their Ins1 -/-:Ins2 +/+ littermates. 458 However, multiple other contributing factors likely influenced these outcomes. 459 The hypothetically environmental-dependent or stress-dependent ability to compensate for reduced 460 Ins2 dosage at the level of insulin translation and/or secretion was only observed in male Ins1 -/-:Ins2 +/-461