Introduction

Over the past few decades, worldwide obesity prevalence has increased dramatically and reached epidemic proportions, especially in western countries1,2. Epidemiological studies have shown that high-fat diet (HFD) and obesity might be interlinked with different brain disfunctions including cognitive impairment3,4, autism-related social deficits5,6,7,8, anxiety9,10,11 and depression9,12,13. Moreover, exposure to either HFD or protein malnutrition during the sensitive perinatal period has been associated with major neurodevelopmental deficits, including autism spectrum disorders (ASD)7,14,15 and other psychiatric disorders16,17,18.

In laboratory mice, several studies have examined the effects of perinatal nutrition (i.e. maternal nutrition prior to conception, during gestation and during lactation) on autism-related social deficiencies and their underlying mechanisms. However, these studies produced inconsistent results, depending on the specific diet regimen and behavioral paradigms employed that were incoherent between different labs19,20. Most research was performed on the effects of maternal HFD on offspring social behavior, and the results indicated on impairments in some autism-related social behaviors21,22,23,24,25. For example, when maternal HFD was initiated at early adulthood no changes were seen in social preference21,22, social novelty preference21, or sensorimotor gating of the adult offspring21,22. A stronger manipulation of 8 weeks on HFD prior to mating reduced reciprocal social interaction of male offspring in a neutral cage23,24, and abolished social preference and social novelty preference in the 3-chamber assay24. Substantially fewer studies examined the effects of perinatal protein malnutrition on the behavioral phenotype23,26, mostly reporting increased anxiety-related behaviors20,27,28,29, with minor to no effects on sociality27,28. Specifically, maternal LPD in mice led to reduced social play in juveniles, without impairing adult social interactions in a neutral cage28, or sociability in the 3-chamber assay27. However, to the best of our knowledge, no study has examined the side-by-side effects of maternal HFD and LPD on the comprehensive autism-related phenotype, including territorial aggression, complex behaviors in a group setting, and pheromone-mediated social behaviors of the offspring.

Here, we harnessed our unique behavioral setup, consisting of all classical autism-related behavioral paradigms30,31,32,33, as well as our high-throughput semi-natural behavioral phenotyping system34,35,36. The semi-natural system allows tracking groups of individually tagged mice within large semi-natural enclosures, continuously over several days, and quantify dozens of individual behaviors, pairwise interactions, and dynamic group organization. We employed these standard and custom-designed settings to comprehensively and rigorously scrutinize the behavioral effects of perinatal exposure to either obesity-inducing high-fat diet, or low-protein diet on the behavior of offspring, in a comparable manner (Fig. 1). We hypothesized that both maternal diets would impair social and autism-related behaviors, but to a different extent. Our results indicate that HFD male offspring showed substantial reduction in sociality towards conspecifics and increase in dyadic aggression, as well as profound increase in autism-related behaviors including cognitive rigidity and repetitive behaviors. A marked reduction in sociality was seen in socially-related chemosensory behaviors. LPD male offspring displayed a milder yet significant reduction in sociality and an overall autism-related phenotype. Moreover, HFD offspring, but not LPD offspring, displayed a distinct set of behavioral repertoires in a group under semi-natural conditions, and exhibited increased dominance within the social hierarchy. Finally, the robust effects in olfaction-related social behaviors as well as previous findings on effects of HFD on olfactory processing and social recognition-related brain regions37, prompted us to examine molecular changes in the olfactory bulbs (OB) between HFD and control mice. Transcriptomic analysis identified altered levels of several autism and neuropsychiatry-related genes in the OB of HFD offspring, which might mediate these deficient behavioral phenotypes.

Figure 1
figure 1

Experimental timeline and layout. (A–G) Schematic illustration of the experimental timeline and the behavioral assays.

Full size image

Experimental procedures

Animals and diets

Female C57BL/6JOlaHsd (C57) mice were purchased from Envigo Laboratories (Rehovot, Israel). At the age of weaning (25–30 days old), females were randomly assigned to either control normal diet (ND, 10% of kcal as fat, 20% kcal as protein, n = 11), high-fat diet (HFD, 60% kcal as fat, 20% kcal as protein, n = 10) or low-protein diet (LPD, 10% kcal as fat, 10% kcal as protein, n = 11) groups. Special diets were products D12450B, D12492 and D11080501, respectively (Research Diets, New Brunswick, NJ, USA). Mice were housed in groups of 3–4 littermates per cage and were given ad-libitum access to food and water throughout the experiment. Following 8 weeks of special diet, a male C57 mouse was added to the cage, and the experimental females were monitored for visual signs of pregnancy. Towards their parturition, pregnant females were placed in separate cages, and kept on the same designated diet until the weaning of their offspring. No apparent differences were noticed between the groups in infanticide. Offspring were weaned to standard chow, and behavioral assays were conducted on the offspring between the ages of 9–13 weeks. No differences were found between different litters of the same maternal diet. At 13–14 weeks of age, following behavioral assays, body composition of male and female offspring was assessed using Echo-MRI (Echo Medical Systems, Houston, Texas). Full experimental timeline is presented on Fig. 1.

All experimental procedures were approved by and conducted in strict compliance with the Institutional Animal Care and Use Committee of the Weizmann Institute of Science, and in accordance with ARRIVE guidelines.

Behavioral assays

Behavioral assays were conducted continuously on all animals, according to the order presented in Fig. 1, and as detailed below. The resident-intruder assay was prioritized to be conducted first, so that it would not be impacted by prior testing, as it measures both aggression and sociality in a vast array of behavioral parameters.

Resident-intruder assay

The assay was conducted in the mice's home cage as previously described36,38,39,40. Briefly, the resident males were introduced with a sexually naïve 5 week-old intruder males swabbed with 140 µl urine collected from sexually mature and experienced male mice. Behavior was recorded and later analyzed using the Observer software (Noldus) by a trained observer blind to the experimental conditions. Behavioral parameters scored included olfactory investigation, allogrooming, aggression and sexual behaviors, to examine ASD-related reduced sociability41.

Olfactory preference assay

The assay was conducted in a three-chamber apparatus as previously described38,39. Briefly, following a habituation day where the mice were placed in the three-chamber apparatus for 15 min, the mice were tested in the apparatus for their preference of female or male-soiled bedding vs. clean bedding during 2 consecutive days (the order of the sex of the bedding stimuli was preference test was conducted as previously describedounter-balanced between the mice). Following 10 min habituation, 30 ml of soiled/clean bedding were placed in polycarbonate cups and stationed in the corners of the two side chambers. Interaction time of the mice’s center mass with each cup was analyzed using the Ethovision software (Noldus). The preference index was measured as the time spent sniffing the stimulus bedding divided by the total time spent sniffing either bedding, and used to examine ASD-related reduced sociability41.

Three-chamber social assay

The three-chamber social preference test was conducted as previously described31,32. Briefly, the test consisted of habituation, followed by 2 consecutive phases: (i) Social approach, with a wire-cage containing an unfamiliar mouse (stranger mouse) situated in one side-chamber and an empty cage (object) on the opposite side, and; (ii) Preference for social novelty, when an additional unfamiliar mouse was placed in the wire cage that had been empty during the previous session. Time spent in each chamber was scored using the Ethovision software (Noldus) and used to examine ASD-related reduced sociability41.

Locations of different stimuli were counterbalanced between animals. Stimulus mice were 5 weeks old CD1 mice, placed in the wire cage 5 min prior to the experiment.

Wet T-maze assay

The wet T-maze was conducted as previously detailed31,33 using a T-shaped Plexiglas chamber filled with 15 cm of clear water with a transparent escape platform submerged 0.5 cm below water level. Animals had five trials during each of the four experimental days, placed each trial in the starting arm facing the wall, and were allowed to swim until locating the hidden platform, or until 90 s have passed. Inter-trial interval was > 5 min. On the first and second days, the platform was located in one arm, while on the third and fourth days it was located in the opposite arm. Latency to climb on the platform and the number of correct turns were measured manually, to examine ASD-related cognitive rigidity41. Mice that failed to learn the location of the platform (i.e. did not show at least 60% of correct turns during the 2nd day) were excluded from the analysis.

Open field test

The assay was conducted as previously described31. Briefly, mice were placed in a 30 × 30 × 25 cm cage for 10 min. Time spent in the center and number of visits to the center were quantified using the Ethovision software (Noldus). Behaviors of self-grooming, climbing, and exploring the arena were quantified using the Observer software (Noldus) by a trained observer blind to the experimental conditions, and used to examine ASD-related repetitive behavior41.

Computation of autism composite score

Based on the previously established scoring method for autism-related severity in mouse models31,32,42,43, the scores of each mouse in nine behavioral parameters, representing the core symptoms of autism44, were z-standardized according to the control ND group, such that higher values represent more severe autism-related behaviors. Relevant measures included: (A) Cognitive rigidity: no. of correct turns on day 3 of the water T-maze. (B) Stereotypic/repetitive behavior: self-grooming duration in the open-field assay. (C) Social interactions deficiency: social preference index in the 3-chamber test (calculated as (time with unfamiliar mouse)/(time with unfamiliar mouse + time with object)), social novelty index (calculated as (time with unfamiliar mouse)/(time with familiar mouse + time with unfamiliar mouse)), duration of aggressive behavior and duration of non-social behavior during the resident-intruder assay. (D) Pheromone-mediated social deficiency: duration of olfactory investigation in the resident intruder assay, and preference index for male/female soiled-bedding. The average Z score of all nine parameters was designated as the autism composite score.

Semi-natural enclosures

Experiments in the semi-natural enclosures were conducted as previously detailed34,35,36. Briefly, two weeks before the experiment mice were anesthetized by a solution of Ketamine (100 mg/kg)/Xylazine (23 mg/kg) and implanted subcutaneously with two RFID microchips. Each arena was occupied by 6 unfamiliar adult male offspring, two individuals of each maternal diet group (i.e. ND, HFD, LPD), for 6 days. On the 7th day, two unfamiliar adult female mice were added to each arena. Total recording time for each experimental day was 4 h in accumulation, during the dark period (i.e., 1/3 of the total dark phase of 12 h were recorded). The semi-natural enclosure contained a large central exploratory arena (L × W × H, 119.2 × 119.2 × 80 cm) surrounded by 8 external standard mouse cages that were connected with short Perspex tubes to the central arena. The central arena was videotaped from the ceiling by infrared-sensitive cameras, and the floor contained RFID antennas connected to a central computer. The mice were allowed to freely roam the arena and the external cages throughout the experiment. The arena floor was covered with bedding and equipped with transparent hiding shelter boxes, bridges, a free-access feeder (containing standard chow) and a water container. The mice were tracked using time-synchronized video and RFID data sets, and their full trajectories were detected automatically and analyzed as previously described34,36. Multiple individual, dyadic, and group-level behaviors were extracted automatically as previously described34,36 based on the calculated distance between the mice, their location within the arena, and their calculated velocities. Then, all behavioral parameters for each mouse in the data set were averaged across each experimental stage (i.e. 1, 2 & 3) and normalized to calculate the linear discriminant analysis (LDA), a supervised method for dimensionality reduction, as previously described34,45. Formation of hierarchy for each group of mice was calculated using the Glicko rating system based on pairwise chasing/being-chased interactions, continuously updating the ratings following each event, as previously detailed34,36.

ELISA

At the end of the behavioral assays mice were sacrificed by administration of overdosed phenobarbital and blood was collected into heparin coated tubes. Samples were centrifuged at 1000 g for 10 min and the plasma (upper layer) was stored at − 80 °C. Oxytocin (K048-H, Arbor Assays, Inc., USA, kit sensitivity 17.0 pg/ml), testosterone (K032-H, Arbor Assays, Inc., USA, kit sensitivity 9.92 pg/ml) and corticosterone (ADI-901-097, Enzo Life Sciences, Inc., USA, kit sensitivity 27.0 pg/ml) were measured by ELISA kits according to the manufacturers' instructions. Plasma samples for testosterone quantification were pre-extracted according to the manufacturer’s instructions. All samples were measured in duplicates.

Microarray & real-time PCR

Upon sacrifice of the mice, olfactory-bulbs (OB) were removed manually. Total RNA extraction (PerfectPure RNA Tissue Kit, 5 Prime), reverse transcription into cDNA (High Capacity cDNA Reverse Transcription Kit, Applied Biosystems) and real-time PCR (Power SYBR Green PCR Master Mix, Applied Biosystems) were conducted as previously described31,39. OB RNA samples from 6 HFD and 6 control ND mice were pooled in pairs and analyzed on a GeneChip® Mouse Gene 2.0 ST Array (Affymetrix, UK). Microarray results were analyzed using Metascape46 and Enrichr47 to identify clusters of gene ontology, and complemental searches were performed in the SFARI gene database48,49, MGI database50, and in the literature. The top eight genes with differential expression between the groups that produced usable primers were verified by quantitative real-time PCR (qPCR). The primers for actin-beta, tyrosine hydroxylase (TH), dopamine receptor D2, oxytocin, and oxytocin receptor were previously detailed31. Additional primers used were: Erdr1, F: GACGGACTCCACAAGGTGC, R: CATTTCTGTACGCAGTCAGGC; Tceal6, F: TTCCTTGAGCTGTCCTGGTTAC, R: TCCAGCTTGCCTTCGTTTTC; IL11R1a, F: TACAGAGCATCTTGCGTCCTG, R: GTCGGTATTGCAACCGGAAC; TrpC5, F: TGAAGGCCCGACATGAGTTC, R: CAGCATGATCGGCAATGAGC; Chrdl1, F: CCGAGTATGCAGAGGGGATG, R: GGGAAGACCTCCAGCTTGTC; Dhrs4, F: AGCTGATAACCACGGCTCTG, R: AAACCTTGTCCCACACCTCC; Tceal9, F: CTCCCTTTAGCCTTGCAGAC, R: GGCAGGGTTTCATCTTGTG; H2afv, F: TTATCAAGGCCACCATAGCC, R: ATGTGGCCTTTGTCTCGTC; AVP, F: CTACGCTCTCCGCTTGTTTC, R: GGGCAGTTCTGGAAGTAGCA; AVPR1a, F: GGGATACCAATTTCGTTTGG, R: AAGCCAGTAACGCCGTGAT.

Statistical analysis

All statistical analyses were performed using STATISTICA software (Tibco, Palo Alto, CA, USA). Standard behavioral assays and gene expression levels were analyzed by the Mann–Whitney test with FDR correction for multiple comparisons. Social preference and social novelty were analyzed using the Wilcoxon signed-rank test. Correlations between weight, gene expression and behavioral parameters were calculated using Spearman's correlation.

Parameters in the semi-natural arena and ELISA were analyzed using repeated-measures ANOVA, followed by post- hoc Tukey’s or Fisher tests. All results are displayed as means ± SEM. #p ≤ 0.1, *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001.

Results

Maternal HFD induces reduction in offspring number and long-term female-biased impairments in offspring physiology

Female mice were fed with either HFD, ND (control), or LPD diet from their weaning onwards, and co-housed with a male at adulthood to allow breeding. At weaning age, the offspring of all diet groups were transferred to standard rodent chow until the end of the experiment. We found that HFD females gained significantly more weight compared to the ND females (supplementary Fig. 1, F = 13.3, p < 0.001), and HFD caused a significant reduction in the number of offspring per litter (ND = 7.7 ± 0.3, HFD = 4 ± 0.6, LPD = 3.3 ± 0.3, Z = 2.13, p < 0.05). However, unexpectedly, this reduction was only due to reduced number of female offspring (Z = 1.97, p < 0.05) and not in the number of male offspring (Z = 0.62, p > 0.5, Fig. 2A). HFD offspring of both sexes weighed more than ND offspring upon weaning (males, Z = 3.5, p < 0.001; females, 3.17, p < 0.01; Fig. 2B) and maintained this overweight as adults (males, Z = 2.73, p < 0.01; females, 2.69, p < 0.01; Fig. 2C). However, only female HFD offspring displayed higher percentage of body fat compared to controls, while male offspring did not differ in their body fat between the groups (males, Z = 0.96, p > 0.3; females, Z = 2.69, p < 0.01; Fig. 2D). These findings suggest that female offspring are more susceptible to the long-term physiological effects of HFD at the perinatal period.

Figure 2
figure 2

Maternal high-fat diet reduces sociality and induces an autism-related phenotype in male offspring. (A) Number of male and female offspring per litter in the ND and HFD groups. (B and C) Weights of male and female offspring measured at weaning (B) and adulthood (C). (D) Percentage of fat mass out of total mass measured at adulthood. (E) Schematic illustration of the resident-intruder assay. (F) Percentage of aggressive animals during the assay. (G–M) Quantification of aggressive (G–I) and social (J–M) behaviors during the 15 min assay. (N) Preference of a stranger mouse over an object in the 3-chamber assay (i.e. social approach), in ND and HFD male offspring. (O) Preference of a novel stranger mouse over the familiar mouse in the 3-chamber assay (i.e. social novelty), in ND and HFD male offspring. (P) Preference of a conspecific soiled-bedding over clean bedding in the olfactory preference assay. (Q) Schematic illustration of the wet T-maze assay and percentage of correct turns during days 1–4 of the assay (O). (R–T) Open-field assay, quantification of self-grooming (R), distance traveled (S) and number of visits to the center (T). nF1 ND = 7–10, nF1 HFD = 8–13. Data is presented as mean ± SEM. *p ≤ 0.05, **p ≤ 0.01.

Full size image

Maternal HFD impairs social behavior and increases aggression and autism-related phenotypes of male offspring

Upon adulthood, offspring male mice were subjected to a series of standard behavioral assays to examine their social behavior as well as non-social autism-related phenotypes. Behavioral analysis showed a consistent reduction in sociality, alongside increased aggression, in HFD male offspring (Fig. 2E–P). In the resident-intruder assay, rates of aggressive animals were significantly higher in the HFD group compared to the control ND (Fig. 2F p < 0.05). Also, duration of aggressive behavior (Z = 2.03, p < 0.05) and number of aggressive events (Z = 2.1, p < 0.05) were higher in the HFD group, and the latency to present aggression was significantly shorter compared to the ND group (Z = 2.33, p < 0.05, Fig. 2G–I). In addition, HFD males exhibited lower levels of olfactory investigation (Z = 2.59, p < 0.01, Fig. 2J–K) and allogrooming of the intruder (Z = 2.38, p < 0.05, Fig. 2L–M) compared to the controls, while there was a marginally significant increase in their duration of non-social behaviors (ND = 380.4 ± 27.9, HFD = 525.5 ± 41.9, Z = 1.81, p = 0.07). In the 3-chamber social preference assay, HFD males failed to show the typical preference for a social stimulus seen in the ND males (i.e. social approach, ND, Z = 2.24, p < 0.05; HFD, Z = 1.49, p > 0.1; Fig. 2N), as well as the typical preference for social novelty (i.e. social memory, ND, Z = 2.38, p < 0.05; HFD, Z = 1.33, p > 0.1; Fig. 2O). In the olfactory preference assay38,39, HFD males showed reduced preference for investigating female-soiled bedding, compared to ND males (Z = 2.58, p < 0.01, Fig. 2P). Finally, HFD male offspring displayed autism-related behavioral deficits of cognitive rigidity and repetitive behaviors. This was manifested in reduction of correct turns during the 3rd day of the T maze assay (Z = 2, p < 0.05, Fig. 2Q) and a marginally significant increase in repetitive self-grooming during the open field assay (Z = 2.14, p = 0.055, Fig. 2R). Both ND and HFD group displayed significant difference between day 3 and day 4 of the T maze assay in their latencies to mount the platform (NDday3 = 12.42 ± 1.24, day 4 = 6.04 ± 0.69, p < 0.05; HFDday3 = 14.91 ± 2.5, day 4 = 8.87 ± 2.57, p < 0.05). In addition, no differences were measured between the groups in the primary learning during days 1–2 of the T-maze assay (p > 0.5), or in the distance traveled or the number of center visits during the open field assay (p > 0.1, Fig. 2Q–T).

Maternal LPD induces an autism-related phenotype without overt physiological impairments in male offspring

Similarly to the HFD group, LPD also caused a significant reduction in the number of female offspring per litter (Z = 2.19, p < 0.05, Fig. 3A). However, LPD offspring of both sexes did not differ from control ND offspring in their weight or fat mass at any of their measurements (p > 0.5, Fig. 3B–D). In the resident-intruder assay, there was a marginally significant increase in the number of aggressive animals (p = 0.09) and reduction in the aggressive latency (Z = 1.83, p = 0.07) in the LPD group compared to controls (Fig. 3E–F). Notably, LPD male offspring displayed reduced allogrooming behavior (Z = 3.34, p < 0.001) and increased non-social behavior (Z = 3.72, p < 0.001, Fig. 3G–I). However, there were no significant differences in their overall levels of aggression or olfactory investigation towards the intruder (p > 0.2, supplementary Fig. 2A). LPD male offspring displayed intact social preference in the 3-chamber social approach assay (Z = 2.49, p < 0.05), but failed to display the typical preference of social novelty (Z = 1.6, p > 0.1, supplementary Fig. 2B). Notably, LPD male offspring displayed a marked decrease in percentage of correct turns during day 3 of the T maze assay (Z = 2.9, p < 0.01, Fig. 3J), although no differences were found between the groups during days 1–2 of the assay (p > 0.5). In addition, LPD offspring failed to show improvement in the latency to mount the platform during day 4 of the assay (p > 0.2), unlike ND offspring (p < 0.05, Fig. 3K), and displayed a marginally significant increased repetitive self-grooming behavior (Z = 1.92, p = 0.055, Fig. 3L). No differences between the groups were measured in the remaining parameters measured (supplementary Fig. 2C, D).

Figure 3
figure 3

Maternal low-protein diet impairs physiology and sociality of offspring. (A) Number of male and female offspring per litter in the ND and LPD groups. (B and C) Weights of male and female offspring measured at weaning (B) and adulthood (C). (D) Percentage of fat mass out of total mass measured at adulthood. (E–I) Quantification of aggressive and social behaviors during the 15 min assay. (J–K) Wet T maze assay. Quantification of correct turns during the 4-day assay (J) and comparison of the latency to mount the platform on days 3–4 of the assay (K). (L) Quantification of self-grooming behavior during the open-field assay. nF1 ND = 7–10, nF1 LPD = 7–12. Data is presented as mean ± SEM. #p < 0.1, *p ≤ 0.05, **p ≤ 0.01, ***P ≤ 0.001.

Full size image

Maternal malnutrition induces an autism-related phenotype in male offspring

To evaluate the overall autism-related severity phenotype of the offspring mice, nine comprehensive behavioral parameters were Z-standardized and averaged to calculate the autism severity composite score (see methods, Fig. 4A, B), as previously detailed31,32. Both HFD and LPD male offspring displayed a marked and significant increase in their autism severity score compared to the control ND offspring (ZHFD-ND = 3.18, p < 0.01; ZLPD-ND = 2.81, p < 0.01), indicating that prolonged perinatal exposure to either diet induces a broad autism-related behavioral phenotype (Fig. 4B). Moreover, autism severity was significantly correlated with the offspring adulthood weight (R = 0.6, p < 0.01, Fig. 4C), demonstrating the massive long-term impact of perinatal diet and metabolism on adult autism-related deficiencies in male mice.

Figure 4
figure 4

Maternal malnutrition induces an autism-related phenotype. (A and B) Autism severity calculation. Nine behavioral parameters (see methods) were standardized and averaged to four deficiency scores of cognitive rigidity, repetitive behavior, social interactions deficiency and pheromone-mediated social deficiency scores (A), and to total autism composite score (B). (C) Spearman's correlation of autism severity score and adult weight of male offspring exposed to perinatal ND, HFD or LPD. nF1 ND = 7, nF1 HFD = 8, nF1 LPD = 7. Data is presented as mean ± SEM. #p < 0.1, *p ≤ 0.05, **p ≤ 0.01.

Full size image

Maternal high-fat diet impairs social behaviors within a group in male offspring

In order to determine the effects of maternal diet on naturalistic social behaviors in a group setting, male offspring were placed in semi-natural arenas for 6 days. On the 7th day, two unfamiliar female mice were also introduced into the semi-natural setup to test for the effect on group dynamics and to measure female-directed behaviors (Fig. 5A). Behavioral tracking divided the experiment into three separate stages, consisting of day 1 (i.e. stage 1, ‘exploration’), days 2–6 (i.e. stage 2, ‘stability’), and day 7 (i.e. stage 3, 'social stimulus') (Fig. 5B-D) as previously described34,35. First, during stage 1 and stage 2 (Fig. 5E), we noticed that HFD offspring displayed significantly higher levels of social interactions (comprised of total events of chasing, being chased and approaching, p < 0.05, Fig. 5F). To further evaluate the behavioral repertoires within the group, multiple behavioral parameters for each stage were analyzed using linear discriminant analysis (LDA)45,51. LDA showed a clear separation between the HFD group and the other two groups during all 3 experimental stages (Fig. 5G–H, supplementary Fig. 3A; supplementary Table 1). Specifically, during stage 1 of the experiment LDA showed the highest separation between HFD and control ND offspring on the LD1 axis, with LPD male offspring unseparated from the controls, while during stage 2 LPD offspring were placed between ND and HFD on the LD1 axis. The main parameters separating HFD from the other offspring on LD1 were types of social interactions, namely approaching, chasing and being chased, as well as hiding and center time alone (Fig. 5G-H). LDA for stage 3 was employed focusing on sex-specific social interactions (e.g. male-male chasing, male–female chasing, etc.) and showed a similar separation of HFD offspring from ND controls, but also separation of LPD offspring (supplementary Fig. 3A, supplementary Table 1). Despite the marked differences in behavioral repertoires, and increased aggression seen in the resident intruder assay, HFD did not differ from the other groups in their dominance Glicko scores, and LPD did not differ from control ND offspring (FMaternal diet = 1.1, p > 0.3, Fig. 5I, supplementary Fig. 3B). Yet, when we analyzed the magnitude of dominance between the most dominant (i.e. alpha and beta individuals) and submissive (i.e. gamma, delta, epsilon, and zeta) mice, we found significant differences only within the HFD offspring (FRank = 22.6, p < 0.01; pND > 0.1, pHFD < 0.05, pLPD > 0.4; Fig. 5J–L), suggesting that maternal HFD increases behavioral territoriality and therefore dominance relations within a group setting. Finally, we analyzed the amounts of sex-specific social behaviors within the group during stage 3 of the experiment, after introduction of two unfamiliar females (Fig. 5M). Surprisingly, we noticed that HFD male offspring chased these females significantly less (Z = 2.05, p < 0.05), while no differences were seen in their male-male chasing during stage 3 (Z = 0.037, p > 0.7, Fig. 5N), suggesting impaired sexual motivation, as also observed in the olfactory preference assay (Fig. 2P).

Figure 5
figure 5

Maternal high-fat diet alters behavioral repertoire of male offspring within a group. (A) Schematic illustration of the semi-natural experiment. (B–D) Selected behavioral parameters during the 3 stages of experiment. (E) Representative ethograms (background in dark blue) of selected individual (up) and social (down) behaviors in a group of 6 mice, during the first 6 days of experiment. (F) Quantification of total social interaction events (i.e. chasing, being-chased and approaching) during the first (day 1) and second (days 2–6) stages of the semi-natural experiment, in HFD, LPD and ND male offspring. (G–H) LDA of multiple behavioral parameters, segregating HFD from LPD and ND offspring during stage 1 (G) and stage 2 (H) of the experiment. For legibility of the entire plot, only the relevant parameters are labeled with arrows. A full list of parameters with their loading coefficient for LD1 and LD2 is presented in supplementary table 1. (I) A representative plot of dynamic hierarchical Glicko scores in a group of 6 mice during the first 6 experimental days. (J–L) Comparison of daily hierarchical Glicko scores between the most dominant (i.e. alpha & beta) and submissive (gamma, delta, epsilon, zeta) mice in ND (J), HFD (K) and LPD (L) perinatal diet groups. (M–O) Quantification of male-to-male and male-to-female social behaviors in HFD, LPD and ND offspring during the 3rd stage of the experiment. N = 8 per group. Data is presented as mean ± SEM. #p < 0.1, *p ≤ 0.05, **p ≤ 0.01.

Full size image

On the other hand, LPD male offspring showed a specific increase in male–male together events (Z = 3.11, p < 0.01), with no change in male–female together events (Z = 0.05, p > 0.9), compared to control ND offspring (Fig. 5O).

Maternal HFD increases OT plasma levels and alters gene expression in the brain of male offspring

Following behavioral testing, male offspring mice were sacrificed and their blood and brains were further analyzed to explore for potential underlying mechanisms. Since marked behavioral alterations were seen mostly in the HFD group, we continued our molecular mechanistic analyses in HFD and ND offspring. First, we measured the levels of circulating oxytocin, testosterone and corticosterone in the blood of HFD and control ND mice, and found a significant increase of oxytocin in the HFD group (F = 4.56, p < 0.05, Fig. 6A). Testosterone and corticosterone levels did not differ significantly between the offspring groups (FTestosterone = 0.4, p > 0.6; FCorticosterone = 2.27, p > 0.1; Fig. 6B, C). As HFD males displayed reduced olfactory investigation behaviors in both resident intruder and olfactory preference assays, we dissected their olfactory bulbs and performed microarray gene profiling. Microarray results produced 107 genes that had > 1.5-fold significant difference in expression levels between ND and HFD (p < 0.05), which were further analyzed for their gene ontology and associated literature (see methods, Fig. 6D, E, Supplementary table 2). Notably, almost 30% of the genes were associated with autism spectrum disorders (ASD) or with other neuropsychiatric disorders, and 5 genes were associated with aggressive behavior, suggesting potential underlying mechanisms for the ASD-related phenotype seen in HFD offspring. In addition, ~ 20% of the genes were associated with endocrinology and hormone signaling, and 32 genes were associated with metabolism and energy homeostasis, indicating the devastating long-term effects of maternal HFD on the physiology of the offspring. Finally, we found 31 genes that were non-coding, suggesting a strong effect on transcription regulation (Supplementary table 2). We further verified the expression levels of the 8 genes with the highest difference that produced usable primers, using real-time quantitative PCR (qPCR), and were able to validate the upregulation of 3 of them in the OB of HFD male offspring compared to controls (ZErdr1 = 2.16, p < 0.05; ZChrdl1 = 1.84, p = 0.065; ZTrpC5 = 2.32, p < 0.02; Fig. 6F). In addition, qPCR showed significant elevation in dopamine receptor type 2 (D2) levels in OB of HFD males compared to control NDs (Z = 2.48, p < 0.05), without changes in the levels of tyrosine hydroxylase (TH), oxytocin receptor or vasopressin receptor type B1 (Fig. 6F). Moreover, we found significant correlations between expression levels of these genes in the OB and behavioral parameters related to ASD. Specifically, expression levels of Chrdl1 and Trpc5 were positively correlated with the autism severity score, while levels of Erdr1, Trpc5 and D2 were negatively correlated with pheromone-mediated female preference. In addition, Trpc5 expression levels were correlated with adulthood weight of the offspring (Supplementary Fig. 4).

Figure 6
figure 6

Maternal HFD alters molecular markers in male offspring. (A–C) Quantification of circulating hormone levels in HFD and ND male offspring, Oxytocin (A), Testosterone (B) and corticosterone (C). (D–E) microarray of olfactory bulb (OB) samples taken from ND and HFD male offspring. (E) Heatmap of gene expression levels in the microarray and association of each gene to specific behavioral/molecular functions related to ASD, social behavior and communication. (F) Relative mRNA expression levels of selected genes in the olfactory bulbs of ND and HFD male offspring. nF1 ND = 6–8, nF1 HFD = 6–9. Data is presented as mean ± SEM. #p < 0.1., *p ≤ 0.05.

Full size image

Discussion

In the past few decades, the world has been witnessing an emergence of an obesity pandemic, especially in western countries52, associated with growing availability of food and increasing rates of fat consumption53. Several lines of evidence point to a transgenerational effect of HFD consumption and obesity, as offspring exposed to maternal HFD show multiple behavioral and neuronal deficits21,24,29,54,55,56,57,58. The obesity pandemic has been co-occurring with a worldwide gradual increase in rates of ASD59,60, while accumulating epidemiological data has been suggesting a link between the two6,61,62. Alongside with obesity and fats consumption, increasing evidence also suggest neurodevelopmental deficiencies and risk of ASD following malnutrition and insufficient protein consumption during critical developmental stages63,64,65.

In our study, female mice kept on either HFD or LPD delivered significantly fewer female offspring than control ND females, and female offspring also suffered from increased percentage of fat mass. This effect is in line with previous results22, where maternal HFD produced sex-specific effects on offspring physiology37,66, and increased the proportion of male-biased litters under conditions of environmental stress67.

Adult male HFD offspring displayed autism-related phenotype, including increased aggression and reduced sociality, consistent with previous studies22,24,55,68,69,70,71. Impaired sociality was especially noticed in reduced social olfactory investigation towards an unfamiliar intruder and conspecific female bedding. HFD male offspring displayed additional behavioral deficits related to autism, namely cognitive rigidity and repetitive behaviors. Importantly, the autism severity composite score was significantly higher in HFD male offspring compared to control ND offspring. Notably, we have previously shown that individual exposure to HFD worsens autism severity score in the ASD-related BTBR mouse model, and that severity score was positively correlated with weight31, emphasizing the massive impact of HFD on ASD-related phenotype.

In contrast to the HFD, adult LPD male offspring did not display any overt physiological effects. However, they did show social deficits, alongside increased cognitive rigidity and increased repetitive behavior resembling to those observed in the HFD group, yet slightly milder. This was also manifested in their higher autism severity scores compared to ND controls. Conversely, a study in African striped mice showed that exposure to low protein diet induced an increase in aggression and reduced pro-social behavior in a social competition assay, but only if diet exposure occurred post-weaning, and not at the perinatal period72. In laboratory mice, male offspring exposed to perinatal low protein diet showed impaired social memory in their home-cage73, and reduced juvenile social play28. However, low-protein male offspring mice kept in post-weaning isolation showed no social deficit and even displayed reduced repetitive behavior27. Notably, in a group settings LPD male offspring displayed increased pro-social affiliative behavior specifically towards other males, under conditions of sexual competition (i.e. following introduction of intruder females). This behavior resonates to our previous findings in similar semi-natural conditions, where vomeronasal-deficient mice displayed a mixture of pro-social with aggressive behaviors, reducing the degree of sexual dimorphism between males and females34. Thus, it is possible that LPD perinatal diet reduces male-typical characteristics of male offspring behavior, shifting the social behavioral repertoire of male offspring closer to that of females.

Under semi-natural conditions in a group setting, HFD male offspring displayed a distinct set of behavioral repertoires throughout the 3 experimental stages, compared to the ND and LPD groups, with the main behavioral feature differentiating the HFD offspring being increased aggressive chasing, and increased total social interactions which include aggressive chasing and being-chased. The aggressive phenotype was also manifested in increased magnitude of social dominance relations between dominant and submissive individuals seen in HFD offspring. These findings extend the previous knowledge on aggressive behavior and ASD-related reduced sociality of HFD offspring24,68,74,75 to the settings of group organization and social hierarchy. That is to say, we demonstrate for the first time the core ASD-related symptom of impaired sociality in HFD offspring in a naturalistic group settings, as we have previously shown in the classic ASD-related BTBR model35. Moreover, upon introduction of female intruders to the semi-natural enclosures, HFD male offspring displayed reduced typical chasing towards the females compared to ND and LPD male offspring. Considering that maternal HFD also impaired typical olfactory investigation of female-soiled bedding in the male offspring, it is suggested that maternal HFD induces deficits in appetitive pheromone-mediated male sexual behavior76,77. Indeed, several studies have demonstrated impairments in the reproductive system of offspring males exposed to perinatal HFD78, including hypogonadism79, as well as lower levels of luteinizing hormone80,81, follicle-stimulating hormone82, kisspeptin and GnRH79. Perinatal HFD also induced early sexual puberty83, impaired spermatogenesis84 and reduced consummatory sexual behaviors in adult male rats82. In our study design, maternal HFD did not alter baseline circulating testosterone or corticosterone levels, similarly to previous research57. However, maternal HFD did increase circulating oxytocin level, in line with previous studies showing increased expression of oxytocin receptor85 and modulation of specific oxytocinergic pathways86 in rodents exposed to perinatal HFD. Although multiple studies have linked increased oxytocin levels with increased sociality, in contrast to our findings, it has also been shown that oxytocin rather enhances the salience of social cues and can promote aggressive behaviors36,87,88, including in semi-natural conditions89. To further explore for possible neural mechanisms underlying these effects of perinatal HFD on offspring neurodevelopment90,91,92, and inspired by the robust reductions in social olfactory behaviors, we analyzed the whole transcriptome of olfactory bulbs taken from HFD and control ND male offspring. Transcriptome analysis and further real-time PCR validations produced several genes that were differentially expressed in the HFD offspring, including Erdr1, Chrdl1, Trpc5 and dopamine receptor D2. Furthermore, expression levels of Chrdl1 and Trpc5 were correlated with autism severity, while levels of Erdr1, Trpc5 and D2 were correlated with pheromone-mediated social deficiency. Notably, Erdr1 has been previously shown to be upregulated in cortices of P30 mice offspring exposed to maternal HFD21, and dysregulated in mouse models of autism spectrum disorders93,94. In humans, the TRPC5 gene was found to have missense variants in autistic patients95,96. Trpc5 belongs to a family of Trpc genes that play a critical role in olfactory processing of rodents97. Indeed, several studies in mice have shown that high-fat diet (HFD) induces sensory processing deficits98, especially olfactory deficits99,100,101, while previous studies have shown olfactory sensory impairments in autistic patients102,103, and in autism-related mice models104,105,106. In addition, the increased expression of D2 in the OB of HFD offspring suggests a possible involvement of dopamine signaling in mediating the social behavioral effects of maternal HFD, as we have previously shown with individual exposure to HFD in a mouse model of autism31.

Taken together, our research reveals novel behavioral phenotypes of male offspring exposed to prolonged perinatal malnutrition, extending previous results of heightened aggression in HFD offspring to a setting of social conflict within a group. Our findings suggest impaired appetitive sexual behavior under both neutral as well as competitive naturalistic conditions. Autism-related social effects were seen both in overweight HFD offspring, as well as in LPD offspring which did not present overt effects on weight. Nevertheless, the effects of perinatal diet were more robust in the HFD offspring, consistent with epidemiological findings62. Moreover, we show a heavy and broad effect of perinatal malnutrition on ASD-related phenotype, displaying all core symptoms of ASD in HFD offspring and robustly increasing autism severity in both HFD and LPD offspring. Finally, we provide putative molecular mechanisms that link perinatal HFD with ASD-related phenotype specifically in sensory processing pathways, and can pave the way to future therapeutic potentials.