Maternal high fiber diet protects offspring on diabetogenic diet in the Nile rat

1 Neuroscience Research Institute, University of California Santa Barbara, Santa Barbara, CA, USA 2 Department of Molecular, Cellular and Developmental Biology, University of California Santa Barbara, Santa Barbara, CA, USA 3 Department of Cell and Regenerative Biology, University of Wisconsin School of Medicine and Public Health, Madison, WI, USA 4 Regenerative Biology Laboratory, Morgridge Institute for Research, Madison, WI, USA


Introduction
Diabetes is one of the fastest growing health challenges facing us today. In the last 20 years, the number of people living with diabetes increased from 151 million to over 463 million [1]; of those, 90-95% have type 2 diabetes. Type 2 diabetes is a complex disease that is poorly understood. It is characterized by insulin resistance exacerbated by beta cell dysfunction. Lifestyle intervention and current medications including insulin therapy can help regulate blood glucose levels but, in the vast majority of patients, these regimens ultimately fail. Consequently, even with treatment, diabetic patients continue to be at risk for numerous debilitating complications and have approximately a two-fold increased mortality compared to people without diabetes [2]. Hence, prevention of diabetes is a top public health priority around the world.
Diabetic risk is strongly associated with both parental history and eating habits. The odds of developing type 2 diabetes is three-to six-fold higher for adults with at least one diabetic parent compared to adults without parental diabetes [3]. However, genome wide association studies repeatedly reveal that genetic factors insufficiently explain the observed familial aggregation [4,5], suggesting that familial risk is modulated by nongenetic factors, which might include learned eating behaviors [6,7] as well as exposure to fetal and neonatal environment [8]. The culpability of diet is well known but not well defined. All three classes of macronutrientscarbohydrate, fat and protein -have been examined quantitatively and qualitatively in the context of type 2 diabetes, but consensus is sorely lacking. In general, newer studies have proposed that a low-carbohydrate diet could be more effective for managing type 2 diabetes compared to a low-fat diet [9,10]. However, fiber is an exception, unlike simple carbohydrate, fiber is a member of the carbohydrate food group that is associated with a reduced risk of developing type 2 diabetes [11].
Due to the unmistakable familial pattern of type 2 diabetes, parental input is likely part of the etiology. It is suspected that mothers play a major role as the only source of nutrition in the fetal and infant stage. The link between poor maternal nutritional status and detrimental metabolic traits has been consistently reported throughout epidemiology studies [12]. Specifically, the higher risk of type 2 diabetes is associated with adults that had a low birth weight and a catch-up growth during childhood [13,14]. This correlational data is the basis of the "Fetal Origin of Adult Disease" hypothesis by Barker in the 1990s, which has since expanded into the "Developmental Origins of Health and Disease" paradigm that focuses on nutritional insults during development on adult life disease susceptibility [15]. Experimental studies have similarly demonstrated adverse long-term effects using a maternal malnourished model, frequently using a low protein (in association with higher fat or carbohydrate) diet during gestation and lactation in sheep, pigs, and rodents [16][17][18][19]. However, research to date have only focused on the impact of maternal diet in increasing disease susceptibility rather than preventing diabetes. Our study explores the consequence of early nutrition using two opposing maternal diets in shifting the susceptibility of type 2 diabetes.
Here we conducted a longitudinal study of 671 Nile rats (Arvicanthis Niloticus) to investigate how maternal diet affects the offspring's diabetic susceptibility. The Nile rat model presents a similar nutritional etiology and disease progression to type 2 diabetes in humans [20][21][22][23]. In its natural habitat, Nile rats consume a low energy, high fiber diet and do not develop dietinduced diabetes [22]. However, in captivity, when they eat a high energy, low fiber diet, they become diabetic, and eventually develop human-like long-term diabetic complications such as diabetic retinopathy [24]. In the lifelong susceptibility of type 2 diabetes, both exposure to maternal nutrition and self-acquired nutrition are acting in concert. Previous experimental studies have interrogated maternal diet as an isolated variable [25,26] but failed to incorporate offspring diet as a covariate. Additionally, we have included sex as a biological variable, acknowledging the presence of sexual dimorphism in diet-induced type 2 diabetes [27,28]. Using a simple Bayesian network, we incorporated three factors -maternal diet, offspring diet, and sex -in our statistical model to examine the probability of diabetes caused by each factor. Our aim is to provide a definitive study of how maternal diet juxtaposed with offspring diet shifts the longterm risk susceptibility and clarify the role of maternal diet in the prevention of type 2 diabetes.

Animals
All animal experiments were approved by the University of California Santa Barbara, Institutional Animal Care and Use Committee, and conducted in accord with the NIH Guide for the Care and Use of Laboratory Animals. Our founder Nile rats originated from the Brandeis University colony of the KC Hayes Laboratory. Nile rats in UCSB are housed at 70-78˚F in a conventional facility with individually ventilated cages and are provided autoclaved Sanichips as bedding material.

Experimental Animal Protocol
Nile rat dams are randomly fed a high fiber diet (Lab Diet 5L3M; Newco Speciality, Rancho Cucamonga, CA, USA) or a diabetogenic diet (Formulab Diet 5008; Newco Speciality, Rancho Cucamonga, CA, USA) [22]. Littermates are randomly assigned to either the high fiber or diabetogenic diet, maintained throughout the study. Using sibling-matched diet assignments allow us to redistribute possible genetic bias into each of the high fiber or diabetogenic diet groups. Random blood glucose (RBG) levels are measured every four weeks starting at weaning age (4 weeks old). Diabetic Nile rats are defined as having RBG>100 based on prior studies [29]. To reduce adverse events from diabetic complications, Nile rats with RBG>500 were euthanized. The number of Nile rats grouped by diet scheme and sex is provided in Supplementary Figure S1, and details from each Nile rat are provided in Supplementary Table 1.

Modeling the probability of diabetes
We used a simple Bayesian model framework [30] consisting of three variables (maternal diet, offspring diet, and sex) to model the probability of diabetes from 4 to 40 weeks at 4-weekintervals (Supplementary Figure S2). The Bayesian network parameter learning and probability inference were implemented by R package ("bnlearn") [31].

Estimating human equivalent age of rodents
To provide human relevance, we presented in parallel the human equivalent age of rodents obtained from the Jackson Laboratory (life phases equivalencies between human and mouse) [32].

Results
In a laboratory environment, Nile rats readily develop diabetes when fed a standard rodent chow [22], thereby referred to as a diabetogenic diet (specifically in the context of the Nile rat). Conversely, they do not develop diabetes when fed a high fiber diet. Both diets used are commercial diets based on their robustness to induce and prevent diet-induced diabetes in the Nile rat. In this study, longitudinal RBG measurements were used to access the development and progression of diabetes. Methods of monitoring blood glucose involving fasting were avoided because of the potential confounding effect of regularly fasting our animals in a long-term study. More importantly, RBG has been verified to be a reliable early biomarker and preferred clinical parameter for predicting both incidence of diabetes and its severity in the Nile rat [29].
Offspring on diabetogenic diet were protected against diabetes when mothers consumed a high fiber diet We observed a large difference between the diabetogenic diet offspring from mothers that consumed a high fiber diet compared to a diabetogenic diet. As shown in Figure 1, in the group where offspring and mothers share the same diabetogenic diets, the average RBG depicts an upward trend from 4 weeks old and clearly reflects a diabetic value from 12 weeks old. The upward trend continues until around 36 weeks old, when the average RBG plateaus at around 300mg/dL. In contrast, in the group with mothers on a high fiber diet, the average RBG wavered around 100mg/dL and did not reach a diabetic value until around 28 weeks, suggesting a delay in onset of diabetes of about 16 weeks. Furthermore, with each subsequent reading at a later timepoint, the difference between the diabetogenic maternal diet group and the high fiber maternal diet group gets larger, suggesting that in addition to the delayed onset of hyperglycemia, the progression of diabetes is additionally slowed down.
Offspring on a high fiber diet did not develop diabetes even when mothers consumed a diabetogenic diet Interestingly, a high fiber diet conferred long-term protection of offspring against diabetes regardless of maternal diet. In offspring that were fed a high fiber diet after weaning, there was no discernable difference in the average blood glucose trajectory between those that were from mothers with a high fiber diet or those that were from mothers with a diabetogenic diet. The average RBG in both groups never reached the diabetic threshold of RBG>100 mg/dL, from 4 to 40 weeks old (Figure 2). This pattern was observed in both sexes (Supplementary Figure  S3).
Protective effect of the high fiber maternal diet was more pronounced in males The blood glucose lowering effect of the maternal high fiber diet on diabetogenic diet offspring was observed in both sexes. Males benefitted more because they had a more severe hyperglycemic profile compared to the females when on a diabetogenic diet. However, the maternal high fiber diet protection lowered the mean RBG trajectory to a similar normal level in both male and female offspring (Figure 3).
High fiber maternal diet reduced the long-term risk susceptibility of diet-induced diabetes In our experimental design, we have three factors -maternal diet, offspring diet, and sex -that affect the risk susceptibility of diet-induced diabetes. To account for all three factors, we built a statistical model using a simple Bayesian network from 671 Nile rats with RBG values ranging from 4 to 40 weeks old. From our model, we calculated the probability of diabetes in diabetogenic diet offspring within the high fiber maternal diet group compared to the diabetogenic maternal diet group. As shown in Figure 4, the probability of diabetes in the diabetogenic maternal diet group increased from 28.74% at 12 weeks old (equivalent to 20 years old in a human) to 81.77% at 36 weeks old (equivalent to 38 years old in a human). In comparison, the high fiber maternal diet group increased from 13.21% at 12 weeks old to 32.87% at 36 weeks old. This data is replotted according to sex in Figure 5.

Discussion
Presently, effective prevention measures are needed to curb the steep rising rate of type 2 diabetes. The susceptibility of type 2 diabetes is subjected to cumulative events throughout life, possibly starting from conception, described as a life course perspective. Early life has been identified as a sensitive period of phenotypic plasticity largely modulated by maternal nutrition. Thus, a better understanding of how maternal diet modulates disease susceptibility would inform us on effective ways to prevent type 2 diabetes.
Here, we use two opposing diets that either induce or prevent diabetes in the Nile rat to understand how maternal nutrition shifts the disease risk of type 2 diabetes. Our experimental design is different in a few ways. First, we used a lifelong maternal diet in the dams. In other experimental studies, the maternal diet is typically restricted to pregnancy and lactation.
However, the influence of maternal diet could include the period of oogenesis that occurs preconception. We also considered the possibility that abruptly switching diets just prior to conception may trigger altered behavioral changes due to interrupted eating habits not pertinent to the maternal nutrition. Second, maternal diet and offspring diet were treated as two independent variables, providing us an opportunity to observe the interplay between early life nutrition and later life nutrition. This added layer of information cannot be obtained from studies using maternal diet as the sole experimental variable. Additionally, we considered sex as a biological variable, versus the majority of diabetic studies in animal models where only males are represented. Third, the risk susceptibility of diet-induced diabetes was analyzed in a longitudinal manner. Diet-induced diabetes has a complicated disease profile where onset and severity are highly variable, both in the human population and the Nile rat model. Thus, we generated a detailed longitudinal dataset to capture dynamic changes over time. Also, our dataset for this study is unusually large, consisting of RBG values from 671 Nile rats from 4 to 40 weeks old. This allows us to generate a robust Bayesian model with three nodes: maternal diet, offspring diet, and sex.
We found that a maternal high fiber diet reduces the probability of diabetes in offspring that consumed a diabetogenic diet after weaning (Figure 4). The ratio for the percentage of calories provided by carbohydrate, fat, and protein were 67:10:23 for the high fiber diet and 56:17:27 for the diabetogenic diet. Although the ratios of macronutrients were not dramatically different, the percentage of crude fiber was 23% for the high fiber diet and only 4% for the diabetogenic diet. The beneficial effect of fiber on diet-induced diabetes in the Nile rat had previously been demonstrated by Bolsinger [29]. A recent epidemiological meta-analysis reports a 10-20% reduction in type 2 diabetes incidence associated with higher dietary fiber using data from 48,468 people in 17 studies [33]. The observed protective effect could be due to the presence of dietary fiber that promotes an anti-diabetic microbiome by increasing the amount of short-chain fatty acid [34]. Another possibility is that fiber reduces the intake calories, and the high fiber diet is essentially a low-calorie diet. Future studies using purified diets can be used to assess the specific effects from dietary fiber.
The focus of maternal diet (marker of early life nutrition) stems from observational studies on adults that lived through famines where early life undernutrition was consistently associated with detrimental metabolic health consequences in adult life [35]. To explain these findings, the predictive adaptive response hypothesis was proposed, where a mismatch between early life nutrition and later life nutrition resulted in reduced fitness compared to similarly matched nutrition throughout life [36]. However, a recent human study investigating different combinations of early and later life environmental conditions revealed that adverse early life conditions are detrimental to later health across all environments, not only in mismatched situations [37]. Our two-by-two factorial design provided two groups with mismatched diets and two groups with matched diets. Nile rats with the high fiber maternal diet and diabetogenic offspring diet had a diabetes profile that was better than Nile rats with matched diabetogenic diets but worse than Nile rats with matched high fiber diet. Furthermore, Nile rats with the diabetogenic maternal diet and high fiber offspring diet did not get diabetes, similar to Nile rats with matched high fiber diets. Therefore, our study does not support the predictive adaptive response hypothesis. Rather, a high fiber diet is beneficial to prevent diabetes whether consumed in early life through the maternal feeding or self-acquired later in life.
In conclusion, we showed compelling evidence that improving maternal nutrition, such as adding high fiber, can be very effective to prevent diet-induced diabetes. The progression of diabetes in the Nile rat offspring (both sexes) fed a diabetogenic diet is depicted as a series of random blood glucose levels from 4 to 40 weeks old. The offspring of mothers with a high fiber diet, on average, experienced a delayed onset of diabetes and progressed more gradually to higher levels of RBG compared to offspring of mothers with a diabetogenic diet. The error bars reflect a 95% confidence interval of the data. The progression of diabetes in the Nile rat offspring (both sexes) fed a high fiber diet is depicted as a series of random blood glucose levels from 4 to 40 weeks old. When the offspring were fed a high fiber diet, they do not develop diabetes neither from mothers with a high fiber diet nor from mothers with a diabetogenic diet. The error bars reflect a 95% confidence interval of the data. A. In male offspring on diabetogenic diet, the protective effect from a maternal high fiber diet was evident from 12 to 40 weeks old. B. In female offspring on diabetogenic diet, the protective effect from a maternal high fiber diet was evident from 24 to 40 weeks old. The error bars reflect a 95% confidence interval of the data. We used a simple Bayesian network that incorporated maternal diet, offspring diet and sex to calculate the probability of diabetes due to maternal diet differences. When the offspring is on diabetogenic diet, the probability of diabetes is much reduced in the group with a maternal high fiber diet compared to a maternal diabetogenic diet.  Table S1: Detailed information of Nile rats used in this study including sex, maternal diet, offspring diet and random blood glucose (RBG) levels from 4-40 weeks old. All animals with RBG>500 mg/dL were euthanized for humane reasons. We then transformed all RBG values over 500 mg/dL to 500 mg/dL.   Age (weeks) 4