The study was carried out initially in adolescent female Wistar rats (Harlan, Barcelona, Spain). Animals were allowed to acclimate for at least 4 weeks before the diet assignation. Rats were handled and individually housed at a 12 h light-dark cycle with temperature of 22 ± 1°C. Two weeks before the diet assignation, food intake and body weight gain were monitored weekly. After the acclimatization period, animals were randomly assigned to control (n = 9) or caloric restriction diet (n = 15). At this stage, no statistical significant difference in body weight among groups was found (average weight 240.7 ± 3.4 g).

Control rats were given free access to standard chow (Standard chow SAFE A04, Panlab, Barcelona, Spain). The standard chow provided 16.1% of protein, 60% carbohydrate, 3.1% fat, 4% fiber, 0.0025% sodium and 2.9 kcal/g as energy content. In contrast, calorie-restricted dams were given a daily amount of food corresponding to 80% of the caloric intake of control dams, which was adjusted according to the body weight (20% of caloric restriction). We selected this food restriction schedule as a moderate calorie-restricted diet has been demonstrated to be sufficient to induce long-lasting alterations in offspring without affecting fertility (Terry et al., 2005; Palou et al., 2010). Water was provided ad libitum in both animal groups.

From the beginning of the experiment, food intake and weight were measured daily and cumulative weight gain and caloric intake were calculated weekly. Estrous cycle was evaluated every day to check cycle regularity and duration of every stage. Two weeks after the beginning of the experimental diets, females were allowed to mate with a male of the same strain. Each male rat was mated with females from both groups. The mating phase occurred in the female cage within 24 h from the proestrus. During the mating phase, food restriction was abolished and animals were allowed to eat ad libitum in order to avoid misinterpretation of individual food consumption or interferences in copulation (Terry et al., 2005; Klingerman et al., 2011). Twenty-four hours after proestrus, the males were returned to their cage and restricted dams carried on with their assigned diets. Then, vaginal smear was evaluated to detect the presence of vaginal plug or spermatozoa. The day these signs of successful mating were found, which confirmed pregnancy, was designated as gestational day 0 (GD 0). In the case of no pregnancy signs, the overall procedure was repeated in the following proestrus period.

During the gestational period, food intake and weight were measured daily and female rats were maintained on the calorie-restricted diet until the GD 20, 2 days previous to birth (GD 20), when restricted diet ended (Figure 1). The birth day was defined as postnatal day 0 (PN0). Within 14 h after birth, pups were weighed and sexed. Litter size was adjusted to eight pups, consisting of four males and four females. The remaining pups were quickly sacrificed by decapitation and brains were collected for further endocannabinoid analysis.

Throughout the lactation period, calorie-restricted and control rat dams were provided with standard chow ad libitum. Food intake and dams’/pups’ weight were measured 3 days per week during this period. At PN 22–23, offspring from both types of perinatal diet were weaned onto standard chow diet (n = 15 and n = 23, for control and caloric restriction group, respectively). Rats belonging to the same litter from each perinatal diet group were housed together (2–3 rats/cage). Dams were sacrificed. During the post-weaning period, weight and food intake were measured weekly. Behavioral studies were performed in adolescence and adulthood (Figure 1). At the 5th postnatal month, three quarters of the male offspring were sacrificed.

All experiments in offspring were performed in males. The term “perinatal” was used to refer to pregestational, gestational and lactation periods.

The evaluation of estrous cycle started 2 weeks before mating and it was performed daily until post-mating day. Every morning, in the 2 h following the beginning of the dark phase, each female was moved to the experimental room. A pipette filled with 0.2 ml of clean water was inserted gently in the vagina. Then, vaginal fluid was collected by aspiration and one drop of it was placed on a glass slide and observed under a light microscope. The stages of estrous cycle identified were proestrus, estrus, metestrus and diestrus (Marcondes et al., 2002). Prolonged diestrus was considered when diestrus lasted for three consecutive days or more.

Food intake was determined by subtracting the amount of each food type left in the cage from the total amount of food provided. In weaned offspring, the individual food intake was determined by dividing the total food intake from each cage by the number of pups per cage. Comparisons among groups were carried out by calculating cumulative caloric intake relative to body weight (Kcal/Kg) as well as weight gain in each period of the study. Furthermore, the percentage of increase in body weight and caloric intake in rat dams was calculated at the end of the caloric restriction diet and specifically, on GD 20 and 21 and during the lactation period, from PN 1–22.

Adiposity was estimated by calculating the percentage of abdominal fat weight over total body weight at the time animals were sacrificed. Rats were weighed immediately before death and sacrificed by decapitation after administration of Equitesin^® (3 mg/kg). Then, perirenal and perigonadal fat deposits were dissected and weighed. The sum of both types of fat was used to determine the percentage of abdominal fat.

At PN0, male offspring chosen to be sacrificed from the two experimental groups were decapitated during the 2nd/3rd h of the dark phase and brains were quickly removed and frozen at −80°C until brain region isolation. To avoid the possibility of variable outcomes among litters, brains from at least three litters per group were used to carry out endocannabinoid measurements (control pups n = 7–5–3 and pups from restricted dams n = 16–11–9, for hypothalamus, hippocampus and olfactory bulb, respectively). For the isolation of the brain regions selected, brains were thawed in cold Tris-HCl buffer (50 mM, pH = 7.40) and the entire hypothalamus, right hippocampus and right olfactory bulb were quickly isolated and immediately frozen at −80°C until lipid extraction. The overall isolation procedure was carried out in less than 7 min for all animals to avoid ex vivo production/degradation of endocannabinoids.

For lipid extraction, pre-cooled steel balls of 5 mm were added to pre-cooled tubes containing the tissue. A solution of deuterated endocannabinoids (AEA-d4, 2-AG-d5, AA-d8, MAEA, oleoylethanolamide (OEA)-d2, palmitoylethanolamide (PEA)-d4 and 1-AG-d5, Cayman Chemicals, Ann Arbor, MI, USA) in acetonitrile was added to the tissue along with 300 μL of ice-cold 0.1 M formic acid and 300 μl of ethylacetate/hexane (9:1, v/v). Then, the samples were homogenized with a TissueLyser II (Qiagen, Hilden, Germany) for 60 s at 30 Hz. Subsequently, the samples were centrifuged for 10 min at 5000 g and 4°C. The organic phase was removed and evaporated under a gentle stream of nitrogen at 37°C. The aqueous phase was further used for protein content determination. The lipid extract was re-dissolved in 50 μL acetonitrile/water (1:1, v/v) and quantitative analysis of the endocannabinoid levels was carried out by liquid chromatography-multiple reaction monitoring (LC-MRM). The concentrations of internal standards, as well as the calibration curves, were set and tailored using test hypothalamic hippocampal and olfactory bulb tissues. LC/MRM conditions for quantitative analysis of the endocannabinoids were set as endocannabinoid levels were normalized to the corresponding protein content of the tissues as previously reported (Wenzel et al., 2013; Bindila and Lutz, 2016).

For protein quantification, the BCA method (bicinchoninic acid assay) was used and the measurements performed by using a FLUOstar Galaxy (BMG Lab technologies).

Anxiety-related behaviors were evaluated in handled animals with the elevated plus maze, at the 7–8th PN weeks (adolescence period). The elevated plus-maze (Panlab, Barcelona, Spain) consisted of a cross-shaped platform made of black and gray plastic. The platform was elevated 65 cm from the floor and had two opposing open arms (50 cm × 10 cm) and two closed arms of the same size. A central area of 10 cm connected all arms. The closed arms were fenced by 50-cm high opaque walls. The light intensity was adjusted at 150 lux in the open arms and 80 lux in the closed arms. The test was performed after 5 h from the beginning of the dark phase. When the test started, each rat was placed in the central area facing one of the open arms and opposite to the experimenter position. Then, the animal was allowed to explore the maze freely for 5 min. After this time, the rat was moved back to the home cage and the maze was cleaned. The number of open-arm and closed-arm entries and the percentage of time spent on open and closed arms were determined by a computer-controlled system recording the interruptions of infrared photo beams located along each arm. Data were analyzed by using the MAZE soft software (Panlab, Barcelona, Spain). Animals that fell off the maze during the test were excluded from the analysis.

Locomotor activity and anxiety-related behavior were evaluated with the open field test in the 2 days following the elevated plus maze test. The open field consisted of a square arena (80 cm × 80 cm and 40 cm high) virtually divided into a peripheral zone and a central zone (40 cm × 40 cm). It was made of plywood and was located in an experimental room illuminated with low light intensity (30 lux). The test was performed 5 h after the beginning of the dark phase. Each rat was positioned in the center of the open field and was allowed to explore freely for 5 min. After this time, the rat was moved back to the home cage and the arena was cleaned. A video camera installed above the arena was connected to a monitor and a video tracking motion analysis system (Smart, Panlab, Harvard Apparatus, Spain), which measured the total distance traveled (cm) and mean speed (cm/s). The program calculated the percentage of time spent in the central area as well as the number of entries to the center zone, as an index of anxiety-like behavior.

The chocolate preference test was performed at adolescence (8–9th PN weeks) and adulthood (12–13th PN weeks). At the beginning of the test, animals were single-housed in new cages provided with both types of food (standard chow and the mixture of chocolates) and water ad libitum. Food intake for both types of food and animal weight was determined 24 h after the beginning of the test. Chocolate preference was calculated as the percentage of chocolate eaten over total food provided, a measure that was independent of the body weight of the animal and that only reflects food preference.

All data are expressed as mean ± SEM. Statistical analysis of results was performed by using the GraphPad Prism version 5.0 program (GraphPad Software Inc., San Diego, CA, USA) and SPSS 15.0 for windows (SPSS Inc., Chicago, IL, USA). Weight gain over the time and caloric intake were analyzed by one-way repeated measures analysis of variance (ANOVA). Multiple comparisons were assessed by Bonferroni post hoc test. To assess the differences among groups on fertility-related parameters the chi-squared test was adopted. Results from the chocolate preference test were analyzed by two-way ANOVA with group (control vs. free-choice animals) and age period used as variables. Further analyses were performed by using the Student t-test, when data passed normality requirements (D’Agostino-Pearson test) or U Mann Whitney test. A p-value below 0.05 was considered statistically significant.

Control pups were born between GD 21 or 22, whereas all pups from calorie-restricted dams were born at GD 22 (data not shown). All pups were born during the light phase. At birth, Student’s t-test analysis revealed that offspring from preconceptional/pregnancy calorie-restricted dams did not differ in weight from control pups, either when analysis was done in both sexes together or in male and female separately (data not shown). However, litter size tended to be lower in the diet restriction group (11.67 ± 0.81 vs. 8.73 ± 1.01; Student’s t-test, t = 2.005, p = 0.05). Indeed, a significant decreased little size was found in female (6.111 ± 0.67 vs. 4.067 ± 0.49, Student’s t-test, t = 2.484, p < 0.05) but not in male pups.

The endocannabinoid system is a relevant homeostatic network of signals controlling energy expenditure and metabolism (Matias and Di Marzo, 2007; Bermudez-Silva et al., 2010). However, its role in developmental programming has not been studied in depth. To our knowledge, this is the first study investigating the levels of endocannabinoids and NAEs at birth after maternal exposure to restricted feeding during the preconceptional/pregnancy period. Interestingly, we found that offspring from calorie-restricted mothers presented in general reduced levels of endocannabinoids and their precursor, AA, in hypothalamus and hippocampus and a similar tendency in the olfactory bulb, as compared to controls. Reduced AA levels may be explained in part because AA (20:4n-6) derives from linoleic acid (18:2n-6), which is an essential fatty acid obtained from dietetic sources. Therefore, decreased availability of nutrients after a prolonged calorie-restricted maternal diet might have affected the levels of the precursor (AA) of the two predominant endocannabinoids (AEA and 2-AG). Conversely, reduced eCB levels, which are AA precursors as well, may also have determined in part the reduced levels of AA.

In the hypothalamus, we detected a strong reduction in the level of the main endocannabinoids and AA. These results are in agreement with Matias et al. (2003), showing decreased levels of AEA in weaning offspring from dams previously exposed to a maternal caloric restriction during pregnancy and/or lactation. However, in our experiments we detected more robust alterations in endocannabinoid and NAE levels likely because preconceptional diet restriction may induce more exacerbating effects in the offspring, as reported in both cohort studies (Roseboom et al., 2006) and animal models of undernutrition (Edwards and McMillen, 2002; McMillen et al., 2004; Zhang et al., 2013), despite the lack of change in weight at birth. Consistently, we found reduced endocannabinoid levels in normoweight pups. However, the lack of association between endocannabinoid levels and pup weight is in disagreement with Matias et al. (2003), showing decreased endocannabinoid levels in underweight pups. However, differently from Matias et al. (2003), we measured the endocannabinoids at birth. Calorie-restricted rat dams tended to weigh less than controls at birth and their endocannabinoid levels might not have been restored to normal. Thus, the endocannabinoid levels in new born rats might reflect the maternal status, due to the fact that long fatty acids can be transferred through the placenta and the fetus are not fully able to modify fatty acid structures (Keimpema et al., 2013). Future research will address these possibilities.

Although the endocannabinoid system has received little attention in the context of nutritional programming, proteins interacting with endocannabinoids, such as leptin and brain-derived neurotrophic factor (BDNF) have been shown to be altered after maternal exposure to undernutrition (Di Marzo et al., 2001; Keimpema et al., 2014). For instance, after maternal undernutrition, alterations in the surge of leptin and BDNF levels during specific developmental stages have been associated to disruption in the development of the hypothalamic circuit and metabolic disorders later in life (Bouret et al., 2004; Yura et al., 2005; Coupé et al., 2009). These data lead to hypothesize that the endocannabinoid system might be involved in the regulation of these processes as well. Indeed, during the prenatal and postnatal periods endocannabinoids levels reach a peak in concentration, which correlates with neurodevelopment processes occurring in these specific age stages (Berrendero et al., 1999). Moreover, alterations of endocannabinoid signaling by prenatal administration either of cannabinoid receptor agonists, such as THC (Δ⁹-Tetrahydrocanabinol), or antagonists, is associated to impairment in neuronal activity, cortical connections and emotional behavior (Rodríguez de Fonseca et al., 1991; Antonelli et al., 2005; Bernard et al., 2005; Moreno et al., 2005; de Salas-Quiroga et al., 2015). Importantly, our data suggest that prolonged prepregnancy-gestational undernutrition could have altered the normal fluctuations of endocannabinoid levels and thus, the establishment of functional circuitries involved in metabolism, ultimately leading to metabolic abnormalities later in life, as proposed by Keimpema et al. (2013). Supporting this hypothesis, we have recently analyzed the expression of cannabinoid CB1 and CB2 receptors mRNA in the hypothalamus of adult offspring born from mothers exposed to pre- and gestational moderate caloric restriction and we have found a clear upregulation in the expression of both receptors, indicating long-term alterations in the offspring as result of maternal undernutrition (Ramírez-López submitted).

Interestingly, although the levels of non-cannabinoid NAEs seem to be less influenced by diet (Hansen and Diep, 2009), we have found decreased levels of PEA in hypothalamus. PEA has been shown to exert anti-inflammatory (Hoareau and Roche, 2010) and antiobesity effects (Mattace Raso et al., 2014). Taken into account this evidence, PEA level deregulation in early life may predispose to diseases associated to inflammation, such as obesity and metabolic syndrome (Rana et al., 2007). Further research should be pursued to corroborate these possibilities.

We also found reductions in AEA levels and a strong tendency to decreased AA levels in hippocampus. It has been shown that circuits connecting hippocampus and hypothalamus are involved in the control of the hedonic aspects of eating (Petrovich, 2013) and that hippocampal endocannabinoid levels can be altered according to different types of diet (Rivera et al., 2012). For instance, decreased levels of hippocampal endocannabinioids have been reported after exposing young animals to prolonged starvation (Hanus et al., 2003). Moreover, we have recently reported decreased levels of hippocampal endocannabinoids after maternal exposure to a high-fat and low protein-containing diet (Ramírez-López et al., 2016). Taking into account that the endocannabinoids play an important role in the regulation of emotional processes and memory formation in this specific brain region (Lutz et al., 2015), these findings raise the question of whether decreased hippocampal endocannabinoid levels in critical windows of development may lead to altered emotional responses later in life (Ramírez-López et al., 2016).

Finally, in the olfactory bulb, another brain structure involved in the regulation of metabolism and feeding behavior, we only found a tendency towards decreased AEA levels, similarly to that found in the hippocampus and hypothalamus. Although these modifications were subtle, they still could have led to alterations in metabolism and feeding behavior later in life, considering the role of endocannabinoid signaling in odor perception and food intake (Soria-Gómez et al., 2014). Future studies are necessary to confirm these speculations.

Although offspring were weaned on standard chow diet, male offspring form calorie-restricted dams started to gain more weight as compared to controls at the beginning of adulthood and exhibited overweight starting from the 15th PN week. Increased weight was accompanied by augmented abdominal adiposity, although caloric intake was not changed in these animals. These long-lasting effects suggest that altered endocannabinoid levels, and presumably signaling, might be associated with inadequate wiring, which could directly trigger a pathological state. This deregulation may also induce subtle alterations in neuronal connectivity of brain structures involved in metabolism, which would eventually increase the vulnerability to diseases later on in life, as proposed by the direct/double hit hypothesis (Keimpema et al., 2013).

Our findings are consistent with other studies showing similar alterations in offspring exposed to a maternal caloric restriction and weaned on a normocaloric diet (Desai et al., 2005; Breton et al., 2009; Suzuki et al., 2010; García et al., 2011; Lukaszewski et al., 2011). Moreover, these effects have been shown either in offspring that were underweight (Desai et al., 2005) or normoweight at birth (Palou et al., 2010). Interestingly, adult obesity in previously undernourished fetus has been associated to early catch-up growth (Eriksson et al., 1999; Desai et al., 2005). Although we do not show direct evidence of early catch-up growth in calorie-restricted offspring, the fact that they were born normoweight and displayed increased weight gain during the first days of life suggests that these animals may be underweight during the fetal period but may have recovered at the end of the nutritional insult (when dams were switched to normal diet). Therefore, normal weight at birth may result from an accelerated catch-up growth, as similarly reported in animal models of maternal caloric restriction (Desai et al., 2005; Yura et al., 2005; Suzuki et al., 2010).

On the other hand, obesity in offspring exposed to maternal undernutrition has been also associated to hyperphagia (Vickers et al., 2000; Desai et al., 2005; Breton et al., 2009; Palou et al., 2010). However, our animals did not display modifications in the caloric intake relative to body weight consistent with the evidence that undernutrition in early life does not necessarily alter food intake (Yura et al., 2005; Sebert et al., 2009). Indeed, it has been reported that once offspring from animal models of maternal caloric restriction reach the same body weight as controls, their hyperphagic behavior ends (Lukaszewski et al., 2011). Under this framework, the increased body weight could be also explained by a reduction in energy expenditure linked to deregulation in endocannabinoid levels during early life, in accordance to the role of these bioactive lipids in food intake, energy balance and energy expenditure (Matias and Di Marzo, 2007; Bermudez-Silva et al., 2010).

To explain the increased adiposity found in the adult offspring from calorie-restricted dams, additional mechanisms have been proposed. For instance, inadequate nutritional conditions in early life induce upregulation of adipogenic signaling pathways, increased expression of genes involved in adipocyte differentiation (Guan et al., 2005), upregulation of lipogenic transcription factors (Desai et al., 2008) and alterations in sympathetic innervations of adipose tissue (García et al., 2011). Interestingly, the endocannabinoid system has been involved in adipogenesis and fat accumulation (Matias and Di Marzo, 2007). However, the role of the endocannabinoid system in perinatal programming has not been elucidated yet and, in the current study, we have not addressed the potential impact of maternal diet restriction on endocannabinoid signaling alterations in the periphery, such as liver, muscle or adipose tissue. However, recent preliminary data obtained using this model indicates changes in the expression of the enzymes controlling production and degradation of endocannabinoids in these peripheral tissues in adult offspring, suggesting that the observed alteration at birth might extend to the adulthood (Ramírez-López submitted). Although we did not monitor the evolution of these changes, one may speculate that endocannabinoid level alterations at birth and/or alterations in peripheral tissues involved in energy expenditure could have contributed to the increased adiposity found in adulthood. Further investigation are required to address these possibilities.

Because of the importance of the long-term consequences of maternal undernutrition, we also explored the behavioral phenotype of adult offspring born from calorie-restricted dams. We have observed enhanced anxiety-related responses in adolescence, as measured by the elevated plus maze test, as previously described in humans (Nomura et al., 2007) and in animal models of maternal caloric restriction in early pregnancy (Erhard et al., 2004; Levay et al., 2008). Although various mechanisms have been proposed regarding the altered emotional responses following prenatal undernutrition (i.e., dysfunctions in HPA and sympathetic adrenal (Levay et al., 2010), our observation of alterations in the endocannabinoid system suggests its involvement given the predominant role played by this system in emotional control a prominent role in emotional control (Lutz, 2009; Lutz et al., 2015). In the present study, we found reduced endocannabinoid levels specifically in hypothalamus and hippocampus. Decreased levels of 2-AG in hippocampus have been associated to anxiety-related responses (Jenniches et al., 2016; Ramírez-López et al., 2016) and altered endocannabinoid signaling after cannabinoid agonist administration in early development leads to decreased emotional reactivity (Antonelli et al., 2005). Therefore, these data suggest that impairment in endocannabinoid signaling in the developing brain could lead to neurobehavioral alterations in the offspring. Nevertheless, further investigations must be pursued to establish whether decreased endocannabinoid levels at birth are correlated to the development of anxiety later in life.