Metabolic syndrome and type 2 diabetes mellitus (T2DM) have reached epidemic proportions worldwide (Saklayen, 2018). A major risk factor for the development of T2DM is the quantity and quality of intra-uterine nutrient exposure (Gluckman et al., 2008). Studies have shown that the maternal diet can modify the development of regulatory systems in utero and postnatally. Individual studies of high and low protein maternal diets have shown their influence on body weight, metabolic phenotypes and programming of food intake in offspring (Oken and Gillman, 2003; Frias and Grove, 2012). Our previous mouse study on intrauterine growth restriction (IUGR) induced by a low protein maternal diet demonstrated sex-specific phenotypic differences in offspring at 9 months of age (Bhasin et al., 2009). While the male offspring displayed glucose intolerance and increased adiposity, the female offspring were glucose tolerant, without catch-up growth or adiposity. However, the mechanisms underlying these metabolic changes are yet to be elucidated. In addition, previous studies have not systematically compared the effect of low protein/high carbohydrate (LP/HC) versus high protein/low carbohydrate (HP/LC) maternal diets on modulating tissue-specific gene expression profiles in key metabolic tissues and the associated pathological metabolic variations in offspring. As gene expression differences involved in metabolically relevant pathways in the hypothalamus (the control center of energy balance and metabolism) and liver (critical for lipid and glucose homeostasis) may underlie the differing metabolic phenotypes in adulthood, we aim to investigate how early-life nutritional imbalances contribute to metabolic syndrome outcomes in adulthood.

In the current study, we tested the hypothesis that intrauterine HP/LC and LP/HC diets would subsequently affect adult metabolic phenotypes and long-term gene expression levels in key metabolic sites. Using HP/LC or LP/HC dietary interventions during gestation and lactation, we followed male offspring until 10 months of age and assessed body weight, adiposity, insulin production, glucose clearance, and global gene expression changes in the hypothalamus and liver. We discovered that mice expressed differential programming by perinatal exposure to protein imbalances, and these changes persist into adulthood. Importantly, we found that gene expression changes brought about by perinatal nutritional perturbations revealed enrichment of genes involved in diverse pathways including those important for systemic energy homeostasis.

Our study showed that maternal diets with abnormal nutritional compositions such as HP/LC or a LP/HC diet led to developmental differences, which persisted into adulthood affecting glucose metabolism, body weight, adiposity and gene expression levels in the hypothalamus and liver. Male offspring from mothers exposed to LP/HC diet showed a significantly reduced birth weight, which persisted at weaning (day 28) and up to 10 months of age. Male offspring from the HP/LC group trended toward a higher birth weight, and had a significantly higher body weight at weaning and increased adiposity at 5 months. Additionally, both sets of offspring were found to have an impaired ability to clear glucose in adulthood. The IPGTT conducted in HP/LC and LP/HC litters resulted in a significantly higher AUC for both groups compared with control. Interestingly, the LP/HC diet presented offspring with reduced baseline insulin and glucose levels, whereas the HP/LC diet had a reduced insulin response 30 min after glucose exposure. This led us to investigate what genes and pathways might underlie these phenotypic differences in the offspring. We found diet- and/or tissue-specific genes and pathways (e.g., gap junction in hypothalamus and purine metabolism in liver for HP/LC, neuronal pathways for both diets in hypothalamus, hedgehog signaling for LP/HC in both tissues), as well as shared pathways across tissues and diets (e.g., cell cycle and circadian rhythm).

There is a well-cataloged history of studies exploring the effects of maternal diet alterations on diseases in adult offspring, specifically within the context of metabolic perturbations (Langley-Evans et al., 1994; Rees et al., 2000; Daenzer et al., 2002; Zambrano et al., 2006). Previous reports have associated both maternal high- and low-protein diets with increased blood pressure, while maternal LP diet results in reduction of body weight in offspring (Jahan-Mihan et al., 2015). Our results indicate that the inadequacy of the low protein maternal diet induces low birth weight and lower body weights throughout life but no difference in adiposity. Additionally, consistent with our previous findings, LP/HC offspring had diminished glucose clearance rates (Bhasin et al., 2009). The HP/LC maternal diet outcomes indicated that too much stimulation of the nutrient-sensing pathways in utero, while not resulting in higher birth weights, did result in higher body weight at weaning, and higher adiposity and glucose intolerance in adulthood. There are conflicting results regarding the physiological impact of increased maternal protein intake on offspring in previous studies. One study exploring high protein maternal diet on offspring reported no change in body weight, degree of adiposity, or glucose intolerance (Thone-Reineke et al., 2006), while another reported a reduction in body weight with an increase in adiposity (Daenzer et al., 2002). Differences in results may be due to variations in maternal dietary composition, specifically the ratio of protein to complementary nutrients and the total energetic intake; one diet had a protein to carbohydrate ratio of 1.12 while the other was 0.91. Our study utilized a protein-carbohydrate ratio of 0.35, which is significantly lower than the other two studies and may contribute to the variation in results recorded by different groups.

In our study, we essentially found a U-shaped curve in terms of the relationship between maternal protein intake and glucose tolerance in the offspring’s adult life. It is predicted that optimal levels for many essential nutrients follow a U-shaped curve: levels that are too low as well as too high result in detrimental health outcomes. Sufficient vitamin A, for instance, is essential in pregnancy to prevent diaphragmatic hernia in the fetus (Clugston et al., 2010), but excess vitamin A results in cleft palate (Ackermans et al., 2011). High protein diets have been advocated as a way to achieve weight loss or prevent weight gain in humans, and our findings suggest that this type of diet, while effective in preventing weight gain, would not be beneficial to the developing offspring. Furthermore, the HP/LC diet resulted in high baseline glucose levels compared with controls at 10 months of age. Therefore, the current trend of providing high protein to low birth weight infants in Neonatal Intensive Care Units (Hay and Thureen, 2010) to combat postnatal growth restriction, may have far reaching implications to these infants that may prove detrimental toward subsequently developing glucose intolerance, and needs close monitoring upon follow-up.

Our hypothalamic transcriptome and pathway analyses in the offspring indicate alterations in tissue gene programming, likely through the developmental consequences resulting from embryonic nutritional imbalances, that connects maternal diet with long-term physiological outcomes in offspring. Key hypothalamic pathways such as synaptic transmission, neural restructuring and neural development were particularly altered by the HP/LC maternal diet suggesting that these pathways may lead to abnormal neural circuitry in the hypothalamus, which plays a key role in energy balance, ultimately contributing to dysregulation of metabolic homeostasis including increased fat storage and glucose intolerance in adulthood. These findings corroborate a recent study where a high fat maternal diet led to significant alterations in hypothalamic gene expression, resulting in a disruption of regular hypothalamic circuitry through ECM and dopamine availability pathway perturbations (Barrand et al., 2017). While the pathways are not the same between maternal diets, they all reflect changes in neuronal signaling due to maternal dietary alterations. In addition, mitochondrial dysfunction, specifically with regard to oxidative phosphorylation within the hypothalamus has been reported to be associated with neurodegeneration and anorexia in anx/anx mice (Lindfors et al., 2011). This seems to link with our data showing enrichment of oxidative phosphorylation pathways within hypothalamic tissue for both diets and the development of feeding aberrations observed in a previous study (Lindfors et al., 2011). Finally, our finding that Ebf1 expression is significantly altered by HP/LC maternal diet is suggestive of a metabolic disorder and dysregulation of neural development, as previous studies have linked this gene with childhood obesity and neuronal differentiation (Moreno and Bronner-Fraser, 2005; Comuzzie et al., 2012).

In the liver, circadian rhythm and metabolic pathways such as phosphate metabolism and metabolic regulation were prominent, indicating that these pathways may be major effectors of metabolic syndrome outcomes due to altered maternal diets. Our findings that a number of circadian rhythm related pathways (Table 1), including rhythmic processes and transcriptional regulation via Per2, are altered within hepatic tissue under a LP/HC maternal diet is an interesting outcome of the current study and is supported by previous work done in intrauterine growth restricted male rat models (Freije et al., 2015). Additionally, physiological evidence directly supports the role of circadian rhythm dysregulation in increasing the risk for obesity, type 2 diabetes, and other metabolic syndrome traits (Blattler et al., 2012), thus supporting our physiological findings. Notably, around 10% of transcripts oscillate diurnally within the liver (Panda et al., 2002), many of which are involved in energy homeostasis and metabolism, further supporting the potential pathophysiological repercussions of hepatic circadian dysregulation. Furthermore, the expression of Slc5a6, a biotin transporter, is dependent on cell autonomous circadian clock genes Clock and Bmal1 (He et al., 2016). Notably, biotinylation is an essential posttranslational modification of a number of mammalian carboxylases, including acetyl-CoA carboxylase, which has been linked to non-alcoholic fatty liver disease (Tessari et al., 2009).

We also report the alteration of MAPK (for HP/LC) and hedgehog (for LP/HC) signaling pathways, both of which have been previously linked with the facilitation of insulin signaling. Disruption of insulin production or transduction of its signal is intimately associated with generation of metabolic syndrome (Guo, 2013). More specifically, our data suggests that the HP/LC maternal diet disrupts the MAPK activity within the hypothalamus. MAPK facilitates hypothalamic insulin signaling involved in determining the organism’s energy and nutrient homeostasis (Mayer and Belsham, 2009; Schultze et al., 2012). The hedgehog-signaling pathway was predominantly affected by the LP/HC condition for both the liver and hypothalamus. The activation of this pathway is highly correlated with the severity of liver damage in patients with non-alcoholic fatty liver disease and is closely linked with hepatic repair mechanisms. Additionally, hedgehog perturbation is involved in both congenital and adult onset metabolic syndrome (King et al., 2008; Choi et al., 2011; Guy et al., 2012).

We acknowledge that this is a discovery study and the findings provide the framework for future avenues of investigation. In particular, pathohistological and structural analysis of both the hypothalamus and liver tissues are necessary to show association between the highlighted pathways/genes and tissue pathology. Additionally, protein level assays and perturbation experiments are necessary to validate the functional and physiological outcomes of the potential causal genes highlighted by our current study. Moreover, it would be interesting to explore further the classification of adiposity and the relevant abundance of white versus brown adipose tissue. This would help characterize more precisely the effects of nutritional perturbation on adipogenesis. A potential limitation of this study is a lack of further validation of our Affymetrix data via RT-PCR, although the validity of microarrays has been highly supported in the literature (Morey et al., 2006; Wang et al., 2006). Additionally, our study mainly focused on identifying the pathways perturbed by diets instead of individual DEGs. The reliability for the significant pathways is high because the random chance to have multiple significant DEGs from the same pathway is extremely low.

Overall, the prevalence of metabolic disorders such as T2DM and metabolic syndrome in modern societies is growing and the etiology is clearly complex, involving diet, physical activity, and possibly endocrine disrupting chemicals in the environment (De Coster and van Larebeke, 2012). Here, we demonstrate that either a deficiency or an excess of protein in early life can result in adverse effects on metabolic outcomes in adulthood, especially with respect to glucose intolerance and T2DM-like phenotypes. Importantly, these early life perturbations altered gene pathways in adulthood, which may underlie the observed metabolic dysregulation, although the cause and effect relationship remains to be further explored. Our unique study design directly comparing the effects between low and high protein maternal intake on both metabolic phenotypes of the offspring and multi-tissue transcriptomes offers a comprehensive understanding of the role of maternal protein imbalance in metabolic health. Based on our findings, a balanced approach to maternal and postnatal nutrition toward achieving optimal outcomes is necessary for a normal physiological phenotype in the offspring.

LJM designed and conducted the experiments, researched the data, analyzed all the data, and drafted the manuscript. QM, SL, and CP analyzed the gene expression data. MB analyzed and summarized all the data, and wrote the manuscript. SX and WT conducted the experiments. JW researched and analyzed the data. SD developed experimental design and assisted in phenotypic experiments and interpretation. AJL developed the experimental design. XY supervised the data analysis and manuscript writing. All authors reviewed and edited the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.