Background
Obesity has been characterized as a state of chronic inflammation, displaying increased macrophage and adipocyte release of interleukins and inflammatory cytokines as well as macrophage accumulation in adipose tissue [
1,
2]. The presence of this low-grade systemic inflammation contributes to the pathogenesis of metabolic diseases (i.e. insulin resistance, diabetes, cardiovascular disease) [
3,
4]. While weight loss, through calorie reduction, can reduce obesity-associated inflammation [
5], there is increasing interest in identifying foods and micronutrients that can dampen the inflammatory profile independent of weight. For instance, there is strong epidemiological evidence to suggest a beneficial effect of dairy consumption on adiposity and inflammation [
6,
7], but randomized trials in humans are inconclusive [
8‐
10]. Differences in specific dietary components or dairy foods in these studies may be responsible for the disparity of results. Furthermore, it can be difficult to dissociate inflammatory outcomes from changes in adiposity, since the latter directly correlates with white adipose tissue (WAT) macrophage infiltration and inflammatory markers over a wide range, at least in mice [
11]. Therefore, it is important to identify which components of dairy and by what mechanism dairy foods exert anti-inflammatory and anti-obesity effects.
One proposed mechanism by which dairy can improve inflammatory status is through dairy-derived calcium’s ability to suppress circulating levels of 1,25-dihydroxyviatmin D
3 (calcitriol).
In vivo evidence has shown increased dietary calcium, both with and without dairy, diminished inflammation and oxidative stress in genetically-modified obese animals, with the interpretive caveat that adiposity was also reduced under these conditions [
12]. Increased dietary calcium dampens circulating calcitriol, and calcitriol has been shown to activate pro-inflammatory pathways by inducing reactive oxygen species generation and cytokine release in adipocytes and macrophages [
13,
14]. However, anti-inflammatory actions of calcitriol have also been reported in humans [
15], and serum calcitriol has been reported to be significantly lower in obese verses non-obese human subjects [
16]. In diet-induced obese (DIO) mice fed high calcium or high-calcium plus dairy, circulating calcitriol did not correlate with alterations in systemic or WAT inflammation [
17,
18]. Thus, changes in calcitriol due to dietary calcium cannot fully explain the apparent anti-inflammatory effects of dairy.
Non-calcium components in dairy foods could also play a role in obesity-associated inflammation and regulation of metabolism. Dairy bioactives such as lactalbumin, angiotensin-converting enzyme inhibitor and branched-chain amino acids may modify metabolic and other physiological functions [
19‐
23]. Additionally, evidence indicates that dairy proteins may have beneficial effects on appetite in humans [
24], glucose metabolism and muscle protein synthesis [
25], as well as improve fasting lipid and insulin levels [
26,
27]. In animal models of obesity, inclusion of dairy or dairy-based proteins in the diet reduces weight gain and adiposity, even in the absence of lower energy intake. In some studies in DIO mice fed a high fat diet containing a dairy matrix (e.g., a calcium rich, non-fat dry milk (NFDM) protein and carbohydrate-based diet), energy intake was modestly increased, resulting in lower feed efficiency [
17,
18]. The etiology of these effects on food intake and energy balance remains to be ascertained.
Systemic energy balance is delicately regulated by multiple interactive systems, including a suite of endocannabinoids (eCBs) and oxylipins, low abundance signaling lipids present in the open circulation and tissues. To date, no attention has been given to the impact of a dairy-based diet on these lipid mediators, despite their roles in inflammation, food intake behaviors, thermogenesis, and a variety of other processes [
28,
29]. Oxylipins, oxygenated products of fatty acid metabolism that include the arachidonic acid-derived eicosanoids, have numerous bioactive properties including both pro- and anti-inflammatory effects [
30], constrictive and dilatory vascular effects [
31], and effects on insulin signaling [
29,
32]. The eCBs are lipid mediators that elicit effects on energy homeostasis and immune function through interactions with the cannabinoid type 1 and type 2 (CB1 and CB2) receptors and transient receptor potential cation channel 1 (TRPV1). eCBs may also bind and activate peroxisome proliferator-activated receptor γ (PPARγ), a transcription factor that regulates genes to promote lipid storage (and immunomodulation) [
33]. The archetypical eCBs are the arachidonate-derived 2-archidonoylglycerol (2-AG) and N-arachidonoyl-ethanolamine (AEA). In addition to neuromodulatory effects on appetite, eCB signaling has been implicated in the pathogenesis of obesity by actions that promote lipogenesis in adipose and liver, and may also regulate insulin sensitivity [
34‐
36].
Besides the effects of dietary long chain PUFAs on circulating oxylipins and eCBs concentrations [
37,
38], the impact of other nutrients on these systems is not known. Given the putative role of dairy and calcium in adiposity and inflammation, we hypothesized that these nutrients would impact signaling lipid profiles, and we tested this under conditions of both developing and pre-existing obesity. Using targeted quantitative metabolic profiling we performed a secondary analysis of archived plasma from a previous study investigating the effects of dairy and calcium on inflammatory phenotypes in a DIO mouse model [
17,
18]. As previously reported, animals fed a high fat diet rich in calcium in a NFDM-based matrix showed decreased adiposity, markedly reduced steatosis, and lower adipose inflammation in comparison to a high-calcium, soy protein-based high fat diet [
17,
18]. In the current study, signaling lipid profiling, gene expression, and gross morphometric parameters were used to describe phenotypes in high fat fed mice that best discriminated NFDM protein and carbohydrate-fed mice from animals fed soy protein/sucrose carbohydrate-based diets. We contend that the identified shifts in oxylipin and endocannabinoid tone may be associated with biochemical and physiological processes that drive the markedly different energy balance and inflammatory phenotype observed in with NFDM feeding in mice.
Discussion
A diet rich in dairy has been associated with improvements in body composition, metabolic disease risk markers and inflammatory status; however, the mechanisms by which dairy exerts its effects are still being elucidated. In the current study, we identified unique profiles of oxylipins and endocannabinoids in the plasma of DIO mice fed a diet rich in NFDM. As previously reported, these NFDM fed animals displayed decreased adiposity, markedly reduced steatosis, and lower adipose inflammation in comparison to control/high-calcium, soy protein based high fat diets [
17,
18]. The results described herein highlight that these phenotypic differences may be associated with alterations in circulating bioactive lipids.
Biosynthesis of oxylipins is catalyzed by cyclo-oxygenase (COX), lipoxygenase (LOX), and cytochrome p450 (CYP) enzymes [
29]. The bioactive products of these pathways have broad cellular effects on vascular [
47], hepatic [
48], and adipose physiology [
32], and, depending on the specific metabolites involved, can exert both pro-inflammatory and anti-inflammatory effects. In the current investigation, dietary treatment resulted in distinct changes to patterns of plasma lipids emanating from the CYP epoxygenase pathways, but not LOX and COX pathways. Specifically, mice fed a NFDM based diet displayed decreases in several species of diol fatty acids, products of soluble epoxide hydrolase (sEH) enzyme activity, compared to controls or mice fed high calcium alone. These effects were observed in two separate cohorts of mice, each with disparate weight outcomes in response to NFDM and HighCa, suggesting that changes in diol biochemistry were unique to the NFDM diet and weight/adiposity-independent.
The major CYP-derived products include epoxide fatty acids (i.e. arachidonate-derived epoxyeicosatrienoic acids (EETs)). Recent evidence suggests EETS have beneficial effects on adipose and liver insulin sensitivity through the reduction of ER stress [
49,
50], and that DHETs, products of EET hydrolysis by sHE, exert their own opposing activity (DHETs increase ER stress and attenuate insulin sensitivity). Our findings are consistent with the role for epoxide/sEH/diol activity in mediating inflammation and insulin signaling. While epoxides were similar across dietary treatment groups, decreased plasma diols observed in the plasma of NFDM fed mice were associated with this group’s improved glucose tolerance, dampened inflammation, and reduced liver steatosis, both in developing obesity (Cohort 1) [
17] and pre-existing obesity (Cohort 2) [
18].
The mechanism by which NFDM decreases plasma diols is unclear, but could result from direct dietary effects on relevant biochemical pathways and substrates flowing through the CYP epoxygenase/sEH pathways, or proceed secondary to metabolic and inflammatory phenotypes associated with these diets. While fatty acid composition was the same for all experimental diets, other components of dairy, such as dairy protein, may alter appetite-regulating hormones [
24], thereby decreasing total food intake and overall total fatty acid substrates. However, cumulative energy intake was not a discriminating variable in the PLS-DA model, did not cluster with diol fatty acids, and was not reduced in NFDM-fed mice [
17,
18]. Alternatively, diet-associated signals or the attenuated obesity phenotype in NFDM mice could have reduced tissue sEH expression and/or activity. It has been previously reported that sEH protein is increased in adipose and liver of DIO mice and obese humans [
50], presumably decreasing the beneficial activity of epoxides. In the current investigation,
Epxh2 (sEH) mRNA expression was elevated in the liver and reduced in the adipose of NFDM fed animals during developing obesity, but these changes were not apparent in the presence of pre-existing obesity. However, changes in sEH activity have been reported despite no changes in mRNA expression [
51]. Furthermore, it is unclear how plasma pools of oxylipins reflect tissue specific enzyme activity, and whether paracrine activity is more important to the phenotype observed. Future studies will be needed to explore if NFDM alters tissue-specific diol production and sEH activity or protein abundance. Nevertheless, we have demonstrated for the first time that circulating oxylipins are altered in DIO mice fed a diet rich in NFDM, a diet that improves inflammatory status and glucose homeostasis.
In addition to possible effects of oxylipin changes in supporting NFDM-associated phenotypes, the endocannabinoid system and structurally-related analogs were considered due to their deep involvement in the control of food intake and energy homeostasis [
52]. In this study, several species of MAG, including 2-AG, increased in the plasma of DIO mice fed NFDM compared to the Control or HighCa diets, under the conditions of both developing obesity and pre-existing obesity. 2-AG has been shown to be a more efficacious cannabinoid receptor agonist than AEA [
53], while other 2-MAGS have been shown to potentiate the activity of eCBs (i.e. 2-LG) [
54] or have GPR119 agonist activity (i.e. 2-OG) [
55]. Others have reported that 2-AG is increased in peripheral tissues of animals fed a high-fat diet (HFD) [
56,
57], and circulating 2-AG levels positively correlate with BMI and intra-abdominal adiposity in humans [
58,
59], although this observation has not been universal [
40]. An association between eCB tone and obesity has been hypothesized to exacerbate obesity-related dysmetabolic conditions by increasing food intake and liver fatty acid synthesis through cannabinoid type-1 (CB1) receptor signaling [
36,
52,
60]. However, in the current study, increases in circulating 2-AG levels in NFDM fed animals were concurrent with markedly reduced liver steatosis, even in Cohort 2 where body weight differences from Control animals were minimal [
17,
18]. Similar findings were reported in a mouse model where tissue levels of 2-AG were artificially elevated by utilizing a monoglyceride lipase (
Mgl) global knockout, the primary degrading enzyme for 2-AG [
61]. These animals did not demonstrate the expected hyperphagia and increased weight gain when fed a high fat diet and exhibited improved insulin sensitivity and considerably less liver steatosis, similar to NFDM-fed mice seen herein. The authors also reported desensitization in liver CB1 receptor signaling in
Mgl knockouts. Although liver CB1 and CB2 mRNA were not detected in the present study, we did observe a significant decrease in CB1 and CB2 receptor mRNA in the RP WAT of NFDM fed animals, at least in Cohort 1 where increases in 2-AG were much more robust. Decreased adipose CB1 receptor gene expression was also associated with increased systemic 2-AG concentrations in viscerally-obese subjects [
58]. While speculative, this raises the possibility that, in NFDM-fed mice, negative feedback regulation of cannabinoid receptor signaling by elevated 2-AG contributed to the lean phenotype.
In addition to 2-AG, several other species of MAG were elevated in the plasma of NFDM fed mice, prompting us to quantify the mRNA abundances of enzymes involved in synthesis (fatty acid liberation via lipolysis) and degradation of these metabolites in RP-WAT, liver, and nodose ganglia (containing vagal afferent neurons), sites that represent tissues responsive to endocannabinoid actions. In our multivariate model, mRNA abundances for these targets weakly contributed to the separation of treatment groups. However, it is acknowledged that mRNA expression may not reflect protein levels or enzyme activity. Alternatively, elevations of plasma 2-AG observed here may be a regulatory response to anti-inflammatory signals stemming from the NFDM diet. Endocannabinoids, including 2-AG, exert immunomodulatory effects across many different cell types and tissues [
62], and it has been reported that plasma 2-AG is decreased under acute inflammatory exposure [
63].
Furthermore, enzyme activity and eCB signaling in tissue sites not sampled herein may contribute to the plasma pool of metabolites or drive the observed phenotype. For instance, intestinal eCB signaling influences food intake via vagal afferents [
64‐
66] and models of intestinal inflammation increase eCB tone [
67,
68]. Also, cross-talk between the eCB system and the gut microbiome can alter adipose physiology [
69] and recent studies demonstrated that administration of
Akkermansia muciniphila, a bacterium associated with an improved metabolic profile, increased intestinal levels of the monoglycerols 2-AG, 2-OG and 2-PG [
70]. Intriguingly, we have detected significant differences in the microbiome composition of the mice fed NFDM compared to the other dietary groups [
71], raising the possibility that intestinal alterations are responsible for elevations in systemic 2-AG and possibly other metabolites.
It is important to address whether the distinct pattern of lipid metabolites seen here is due to a direct dietary effect or a secondary consequence of the phenotype. Studies addressing dietary influences on oxylipins and eCBs primarily focus on amount and composition of dietary fat [
38,
72,
73]. In the current study, the source and macronutrient content of fat were identical across treatment groups and was presumably not a driver in the treatment associated phenotypes. Furthermore, high-calcium alone is not sufficient to exert the effects seen here, since HighCa and NFDM fed animals, whom both received 1.5% calcium in their diets, discriminated independently from one another in our multivariate models. In future studies it will be interesting to determine if NFDM effects on lipid and enzyme phenotypes, reported herein, are modulated by altering dietary Ca intake. Regardless, there is evidence that the lipidomic patterns seen here are influenced by diet and are not simply a consequence of a lean or obese phenotype. First, lipid metabolites that best discriminate treatment differences in our multivariate model cluster independently from body weight, adiposity, adipose inflammatory markers, and food intake. Second, the pattern of discriminating factors observed in our model was observed in both Cohort 1 (developing obesity) and Cohort 2 (pre-existing obesity) studies despite markedly different feeding paradigms and body weight outcomes.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
TND, AHK, APT, SHA conducted animal model research and/or sample analyses; TND, AHK, JWN conducted statistical analysis; TND, AHK, JWN, SHA interpreted results; and TND drafted the manuscript with edits from all authors. All authors read and approved the final manuscript.