Maternal diet and obesity shape offspring central and peripheral inflammatory outcomes in juvenile non-human primates

Abstract

The obesity epidemic affects 40% of adults in the US, with approximately one-third of pregnant women classified as obese. Previous research suggests that children born to obese mothers are at increased risk for a number of health conditions. The mechanisms behind this increased risk are poorly understood. Increased exposure to in-utero inflammation induced by maternal obesity is proposed as an underlying mechanism for neurodevelopmental alterations in offspring. Utilizing a non-human primate model of maternal obesity, we hypothesized that maternal consumption of an obesogenic diet will predict offspring peripheral (e.g., cytokines and chemokines) and central (microglia number) inflammatory outcomes via the diet’s effects on maternal adiposity and maternal inflammatory state during the third trimester. We used structural equation modeling to simultaneously examine the complex associations among maternal diet, metabolic state, adiposity, inflammation, and offspring central and peripheral inflammation. Four latent variables were created to capture maternal chemokines and pro-inflammatory cytokines, and offspring cytokine and chemokines. Model results showed that offspring microglia counts in the basolateral amygdala were associated with maternal diet (β=−0.622, p<0.01), adiposity (β=0.593, p<0.01), and length of gestation (β=0.164, p<0.05) but not with maternal chemokines (β=0.135, p=0.528) or maternal pro-inflammatory cytokines (β=0.083, p=0.683). Additionally, we found that juvenile offspring peripheral cytokines (β=−0.389, p<0.01) and chemokines (β=−0.298, p<0.05) were associated with a maternal adiposity-induced decrease in maternal circulating chemokines during the third trimester (β=−0.426, p<0.01). In summary, these data suggest that maternal diet and adiposity appear to directly predict offspring amygdala microglial counts while maternal adiposity influences offspring peripheral inflammatory outcomes via maternal inflammatory state.

1. INTRODUCTION

Obesity is a rapidly growing epidemic that currently impacts 40% of adults living in the US. This public health crisis is largely driven by the consumption of a highly palatable and calorically dense diet, termed the Western-style diet (WSD). This diet is composed of higher percentages of saturated fats and simple sugars compared to other diets such as the Mediterranean diet. Chronic consumption of a WSD, and the corresponding obesity, results in public health consequences ranging from increased risk of developing insulin resistance and diabetes to hypertension and cardiometabolic disease (NHLBI Obesity Education Initiative Expert Panel on the Identification, 1998). The impact of this epidemic on future generations is concerning as increasing numbers of women are classified as obese before becoming pregnant. In 2019, 30% of women were classified as obese prior to pregnancy, an 11% increase from 2016 (Driscoll and Gregory, 2020). Many adverse birth outcomes have been associated with maternal obesity ranging from increased risk of preterm birth, preeclampsia, and increased gestational weight gain during pregnancy (Gaillard et al., 2013). A well-established body of research shows maternal consumption of a WSD and an obese metabolic state during pregnancy have long-lasting adverse impacts on offspring health. Preclinical models utilizing non-human primates (NHP) suggest these consequences include fetal exposure to increased circulating inflammatory factors (McCurdy et al., 2009), impaired placental function and inflammation (Frias et al., 2011), increased pancreatic inflammation (Nicol et al., 2013), and even an increase in the number of microglia in the hypothalamus (Grayson et al., 2010).

Maternal consumption of an obesogenic diet in both pre-clinical (DeCapo et al., 2019; Thompson et al., 2018; Thompson et al., 2017) and clinical populations (Krakowiak et al., 2012; Mehta et al., 2014; Torres-Espinola et al., 2015) is associated with behavioral alterations in offspring. Clinical studies have associated maternal WSD overconsumption with increased risk of offspring neuropsychiatric disorders such as Attention-deficit/hyperactivity disorder (Gustafsson et al., 2020) or autism spectrum disorder (ASD) (Andersen et al., 2018). Results from preclinical studies using animal models support the trends observed in clinical studies. Work in both rodents and NHPs suggests maternal consumption of a WSD increases offspring anxiety and depressive-like behaviors later in life (Edlow, 2017; Sullivan et al., 2010; Thompson et al., 2018). A common overlap shared amongst the clinical disorders is disrupted emotional control.

We postulate that these behavioral changes induced by maternal WSD consumption are due to alterations in fetal brain development. Previous work using our model has demonstrated offspring brain development is impacted by maternal diet. These findings include increased microglial staining in the fetal hypothalamus (Grayson et al., 2010) as well as perturbations to the serotonin and dopamine systems throughout the brain (Rivera et al., 2015; Sullivan et al., 2010; Thompson et al., 2017). Here we expanded this hypothesis, by examining microglial alterations in the amygdala. The amygdala is a brain region of interest because of the integral role it plays in emotional control a behavioral domain that has been shown to be impacted by maternal diet in preclinical and epidemiologic studies. Also, its development has been reported to be altered by exposure to maternal obesity and WSD in both rodents (Glendining et al., 2018) and NHP models (Ramirez et al., 2020). The primate amygdala complex has extensive connections with the neocortex as well as the hypothalamus, brain stem, basal forebrain, hippocampus, and striatum (Amaral, 2002). The amygdala complex contains distinct subnuclei that are associated with various facets of emotional control (LeDoux, 2007). In this study, we examined how consumption of a WSD impacts both specific subnuclei as well as the amygdala as a whole.

Microglia are the main immunocompetent cells of the brain. In response to various stimuli, microglia alter their functional state by proliferation, migration to the site of interests, alteration of their morphology, and activation of cell signaling cascades related to inflammation. During development, aside from their main function fighting infection and responding to damage, microglia play a crucial role in guiding proper neurodevelopment by performing functions such as synaptic pruning (Paolicelli et al., 2011), phagocytosis of neural progenitor cells (Cunningham et al., 2013), and refining network connectivity(Eyo and Dailey, 2013). Due to their integral role in development, many hypotheses implicate microglia in neurodevelopmental disorders (Dunn et al., 2019; Kim et al., 2018).

Maternal diet, adiposity, and peripheral inflammation have all been proposed as potential mechanisms underlying changes in offspring brains and behavior (van der Burg et al., 2016). However, differentiation of the unique programming effects of these early environmental factors on offspring inflammatory outcomes has remained largely unaddressed. While it is believed the main source of inflammatory burden in obese individuals is due to increased production of inflammatory factors due to increased adipose tissue mass, dietary studies indicate specific nutrients such as saturated fatty acids, simple carbohydrates, and polyunsaturated fatty acids are capable of pro-/anti-inflammatory signaling capabilities. In the present study, we aim to differentiate the unique effects of maternal adiposity, inflammation, or diet on offspring outcomes.

In this study, we utilized a NHP model of diet-induced maternal obesity, where animals consumed a diet modeled after the WSD including higher percentages of fat, especially saturated fats, and sugars. We examined animals at 13-months in an effort to increase translatability to clinical populations in humans. This timepoint in NHPs correspond to approximately 3–4 years in humans (Mattison and Vaughan, 2017; Thorpe et al., 2020), which are some of the earliest timepoints neurodevelopmental disorders (e.g. ASD, ADHD) can be diagnosed (Hyman et al., 2020; Wolraich et al., 2011). The goal of this study was to characterize the complex relations among maternal diet, metabolic, and inflammatory states during pregnancy and to examine how these factors impact offspring inflammatory outcomes. We hypothesized that baseline central and peripheral inflammation in offspring is programmed via metabolically induced inflammation in the mother during gestation. This NHP model is uniquely situated to tease out the independent effects of the various maternal factors on offspring inflammatory outcomes. To our knowledge, this is the first NHP study to examine the direct and indirect effects of chronic WSD consumption, maternal metabolic state, and gestational inflammatory environment on offspring central and peripheral inflammation.

2. METHODS

2.1. Animal Group Formation/Demographics

All animal procedures were in accordance with National Institutes of Health guidelines on the ethical use of animals and were approved by the Oregon National Primate Research Center (ONPRC) Institutional Animal Care and Use Committee.

The present study utilized an established preclinical NHP model (Macaca fuscata) of maternal overnutrition to investigate the impact of maternal consumption of a WSD diet on offspring central and peripheral inflammatory outcomes compared to controls. The macaque develops the full spectrum of metabolic disease associated with human consumption of WSD (Havel et al., 2017) and has similar gestational and neurodevelopmental trajectories (Ryan et al., 2019). The high-fat diet utilized in this model is designed to calorically reflect the typical WSD both prior to and during pregnancy, increasing the translatability and relevance to human maternal overnutrition (Havel et al., 2017).

Maternal metabolic variables were measured prior to conception and during the third trimester to determine the role that maternal metabolic state plays in influencing offspring inflammatory outcomes. Given the association between maternal overnutrition and inflammatory burden and the role of maternal inflammatory factors in fetal programming, maternal systemic cytokine levels were also examined during pregnancy.

Adult female Japanese macaques (Macaca fuscata) were housed in indoor/outdoor pens in groups of 4–12 individuals (male/female ratio 1–2/3–10). All animals were given ad libitum access to water and one of two experimental diets: control diet (CTR; 14.7% kcal fat; Monkey Diet no. 5000, Purina Mills) or WSD (36.6% kcal fat; TAD Primate Diet no. 5LOP, Test Diet, Purina Mills). Breakdown of macronutrients in each diet has been previously reported in Thompson et al. 2017 (Thompson et al., 2017). Females received the majority of their caloric intake exclusively from their experimental diets; the diets were supplemented by daily enrichment of fruits or vegetables. Additionally, once a day WSD females were provided a calorically dense treat (35.7% kcal fat, 56.2 % kcal carbohydrate, 8.1% kcal protein).

Females were allowed to breed with intact males. Each measure included is unique to a singleton pregnancy. Adult females consumed their respective diet for at least 1 year prior to pregnancy with any offspring included in this study. Offspring were born naturally and maintained with their mothers until approximately 8 months of age.

2.1.2. Measurement of Maternal Adiposity

Prior to pregnancy, 122 female animals underwent dual-energy X-ray absorptiometry scans to determine body composition. A Hologic QDR Discovery scanner (Hologic, Bedford, MA, USA) in “Adult Whole Body” scan mode and Hologic QDR Software version 12.6.1 was used to calculate percent body fat. The mean ± standard error of pre-pregnancy adiposity for each diet group is as follows: CTR = 19.96 ± 1.15% body fat; WSD = 27.51 ± 1.42% body fat (Table 1).

Table 1.

Animal Numbers for Maternal Procedures

Maternal Measures	N (per diet group)	Average Measures, mean ± SE		Maternal Age (years), mean ± SE
		CTR	WSD	CTR	WSD
Prepregnancy adiposity	122 (CTR = 61,WSD = 61)	19.96% ± 1.15	27.51% ± 1.42	10.20 ± .35	9.01 ± .30
Third-trimester Inflammatory markers	117 (CTR = 52, WSD = 65)	See Table S1		9.94 ± .39	8.97 ± .27

Note: CTR control Diet; WSD Western-style Diet.

2.1.3. Plasma cytokine measure collection

Plasma samples were collected from 117 adult females in the third trimester and 93 juvenile offspring at 13-months (Tables 1 & 2). Prior to pregnancy, plasma supernatant was aliquoted and stored at −80°C until the time of assay. Plasma concentrations of inflammatory marker levels were determined using a monkey 29-plex cytokine panel (ThermoFisher Scientific, Waltham, MA, USA) following the manufacturer’s instructions. Three plates were used to complete this analysis, originating from the same lot (#1833398A). Concentrations of each cytokine were calculated from a standard control curve. Samples were analyzed on a Milliplex Analyzer (EMD Millipore, Billerica, MA, USA) bead sorter with XPonent Software version 3.1 (Luminex, Austin, TX, USA). Data were calculated using Milliplex Analyst software version 5.1 (EMD Millipore). The inter-assay CV and lower limit of quantification (LLOQ) for each assay are listed in Table S1. Growth factors were not considered for this analysis.

Table 2.

Animal Numbers for Juvenile Procedures

Juvenile Measures	N (per diet group)	N (per sex)
Immunohistochemistry	24 (CTR/CTR = 8, WSD/WSD = 8, WSD/CTR = 8)	F = 12, M = 12
13-month Inflammatory markers	93 (CTR/CTR = 28, WSD/CTR = 36, CTR/WSD = 11, WSD/WSD = 18)	F = 42, M = 51

Note: CTR control Diet; WSD Western-style Diet.

2.2. Immunohistochemistry Methodology

A fluorescent immunohistochemistry (IHC) experiment was performed to quantify microglial cell density in the right amygdala of 13-month juvenile animals. Tissue was obtained from the Obese Resource tissue bank. At 13.25 ± 0.71 months juveniles were euthanized and brain tissue was collected. Euthanasia was performed by ONPRC Necropsy staff and adhered to American Veterinary Medical Association Guidelines on Euthanasia in Animals and ONPRC standard operating procedures and guidelines. Briefly, animals were sedated with ketamine HCl (15–25 mg/kg i.m.) transported to the necropsy suite and deeply anesthetized with a surgical dose of sodium pentobarbital (25–35 mg/kg i.v.). After sufficient anesthetic depth was reached, animals were exsanguinated from the aorta while cerebral perfusion was performed via the carotid artery. Perfusion consisted of an initial flush of 0.5–1 liter of 0.9% heparinized saline followed by 1–2 liters of 4% paraformaldehyde buffered with sodium phosphate (NaPO4, pH 7.4) to fix tissue. After perfusion, the brain was removed, sectioned into regional blocks, and placed in 4% paraformaldehyde for 24 hours at 4°C to complete fixation. Brain tissue blocks were then transferred to 10% glycerol buffered with NaPO4 for 24 hours and finally transferred to 20% glycerol solution for 72 hours. Tissue blocks were frozen in −50°C 2-methylbutane and then stored at −80°C until sectioning. Coronal sections (35 μm) of the temporal lobe were collected in 1:24 series using a freezing microtome and stored in cryoprotectant at −20°C until immunohistochemistry was performed.

Twenty-four animals balanced for maternal measures such as age and offspring sex (12 females) were included in this study. Diet manipulation involved long-term maternal consumption of a control or WSD which was maintained for offspring post-weaning until tissue collection (WSD/WSD and CTR/CTR). One group of offspring received a diet intervention at weaning when they were transitioned from WSD to the control diet (WSD/CTR). Standard IHC methodology was used. Briefly, 35 μm thick tissue sections containing the amygdala were bath washed in a KPBS solution. Sections were then blocked for 1 hour in 2% Normal Donkey Serum (NDS; Jackson ImmunoResearch Cat #017-000-121), 0.4% Triton-X100 (Fischer Bioreagents Cat# BP151-500) KPBS solution. Following blocking, sections were transferred to a primary antibody solution (2% NDS in KPBS) containing 1:1500 rabbit anti-IBA-1 (Wako FujiFilm Cat# 019-19471 Lot# PTE0555) and 1:100 mouse anti-Acetylcholinesterase (Abcam Cat# ab2803 Lot# GR3242309-4). Sections were incubated at room temperature for 2 hours before being transferred to 4°C where they remained for 22 hours.

After primary incubation, sections were removed from 4°C and washed in KPBS. Tissue was then transferred to secondary antibody solution (2% NDS in KPBS) containing 1:1000 donkey anti-rabbit Alexa 488 (ThermoFisher Cat# A21206) and 1:1000 donkey anti-mouse Alexa 555 (ThermoFisher Cat# A31570). Sections were incubated for 2 hours at room temperature. After secondary incubation, sections were again washed in KPBS. Next, tissue sections were mounted onto gelatin-subbed slides, coverslipped using Prolong Gold Antifade (ThermoFisher Cat# P36930) and stored at 4°C until imaging.

2.3. Automated Cell Counting

2.3.1. Acquisition

Z-stacks of the right amygdala were acquired using a Leica SPE point scanning confocal with an HC Fluotar L 25x magnification 0.95 NA W VISIR objective. Each stack was 293.91 μm by 293.91 μm, with 45 frames separated by 0.57 μm to acquire information through the entire volume of each region of interest. Seven sections, representing the whole rostral to caudal aspect of the right amygdala, from each animal were analyzed. To examine specific subregions of the amygdala, boundaries of each region were determined using an AChE stain with a 5x objective (Figure 1), referencing an adolescent non-human primate atlas (Paxinos et al., 2009). To have a representative count for each subregion, three images were acquired and averaged for each subregion per section. Once images were collected, 20 Z-stacks were selected at random and assessed for the presence of signal. As frames 3 through 43 contained signal for all Z-stacks this range was selected for analysis to ensure all images represented the same volume of tissue. A custom FIJI macro was recorded and used to generate max projections.

Figure 1.

Schematic of experimental paradigm

2.3.2. Analysis

A selection of representative images, balanced for diet, sex, and subregion, were used to make a microglia identification guide. In brief, microglia were defined by an observable circular cell body, brighter on the boundaries and dimmer in the center. To be counted as a microglia, a cell needed to have at least a single process present with positive staining for IBA-1. Cell bodies were also roughly similar in size (around 40 μm²), so any potential microglia that appeared noticeably larger or smaller than other cells in the image were rejected. Masses of processes without an associated cell body or otherwise-identifiable microglia touching the edge of the image were rejected. Following training, two experimenters counted microglia in 35 images selected at random to use as a “gold standard” to compare to automated counts.

A FIJI macro was written to allow fast identical application of the methodology to every image (Schneider et al., 2012). In brief, to reduce background noise a Gaussian blur (sigma=60) was applied to the image, and an Otsu local threshold (radius=15) was used to separate any IBA1+ signal from the background. The FIJI function “Analyze Particles” was used to count microglia cell bodies. Cell bodies were distinguished from processes or small patches of autofluorescence using a size filter of 40 μm². The circularity filter was set to 0.05–1.00; this range excluded individual thin processes without excluding microglia of a diversity of morphologies. Fiji then generated an output image, which contains a table of object counts (cells and any false positives that passed the filters) as well as an image showing outlines of the objects counted. This automated method was accepted when it captured at least 85% of the manually counted cells per image in the 35-image test set. Each outlined object was then assessed by the experimenter for false positives, and any images that failed the automated method due to poor staining were recorded and excluded from analysis.

Our analysis focused on subregions that had representation from 90% of animals and section. These regions included the basolateral dorsal (BLD), basolateral intermediate (BLI), basolateral ventromedial (BLVM), basomedial magnocellular (BMMC), lateral (LA), and paralaminar (PAL) amygdala nucleus. Two animals needed to be removed from analysis due to poor staining as a result of imperfect fixing of the brain tissue at necropsy.

2.4. Data analysis/Statistics

Data were analyzed using a structural equation modeling (SEM) framework (Schumacker and Lomax, 2004) using Mplus 7.4 (LK and BO, 1998–2012). The robust maximum likelihood estimator was used for models considering continuous outcome variables and the weighted least squares means estimator was used for models that considered categorical outcome variables. These estimators accommodate non-normally distributed data by adjusting standard errors using the Huber-White sandwich estimator.

Model fit was assessed using standard fit indices, the Comparative Fit Index (CFI), the Tucker-Lewis Index (TLI), and the root mean square error of the approximation (RMSEA). CFI and TLI values above .90 and RMSEA values below .08 indicate adequate model fit (Bentler and Bonnett, 1980; Browne and Cudeck, 1992). Non-independence of observations (i.e., the nesting of offspring with the same mothers) was handled using the Mplus Cluster command. Missing data was handled using full information maximum likelihood (Enders, 2001).

2.4.1. Data Cleaning Process

Inflammatory marker data were processed prior to analysis. This method was applied to more accurately capture variability in inflammatory marker concentrations that may be lost when large numbers of concentration values fall below the lower limit of quantification (LLOQ). Briefly, inflammatory marker concentrations below the LLOQ were replaced with values equal to LLOQ/2 (Beal, 2001). The number of subjects that had values below the LLOQ for each inflammatory marker were noted. Next, a log2 transformation was performed on the inflammatory marker values. Base 2 transformations are more physiologically relevant and usually provide more normal distributions. Outliers 5 standard deviations above the mean were removed. In both maternal and juvenile inflammatory marker measures, no values were removed as outliers. After outliers were screened, mean and SD were recalculated. Next, values 3 SD above the mean were Winsorized (Blaine, 2018). Inflammatory markers that produced more than 55% of the values below the LLOQ were discarded and not considered in analysis. The inflammatory markers that were removed at this step were FGF-Basic, GMCSF, IL-4, IL-5, IL-10, and MIG. Inflammatory markers where 11.5%–55% of concentration values were below the LLOQ were transformed into categorical variables with bins of equal amounts, where the first category is comprised of the values below the LLOQ and the remaining values are placed into equal sized bins (e.g., for juvenile IL-17, 0 = all values below the LLOQ (20), 1 = 19 values, 2 = 18 values, 3 = 18 values, 4 = 18 values). Summary of the inflammatory markers that were excluded or transformed into categorical variables are summarized in Table S1.

2.4.2. Data Reduction

Confirmatory factor analysis (CFA) was used to reduce the high dimensionality of the cytokine data. Based on widely accepted categorizations of inflammatory signaling molecules, we created four latent variables that captured unique aspects of inflammatory signaling: maternal pro-inflammatory cytokines, maternal chemokines, offspring cytokines, and offspring chemokines. Maternal and offspring chemokine latent variables were indicated by MCP-1, Eotaxin, RANTES, ITAC, MDC, and IP10. The maternal pro-inflammatory cytokine variable was indicated by IL-12, MIF, TNFα, IFNγ, and IL-1b. The offspring cytokine variable was indicated by MIF, TNFα, IFNγ, IL-1b, IL-2, IL-6, IL-15, IL-17, and IL-1RA (Table S2).

2.4.3. Hypothesis Testing

After determining which latent variables to include in analysis we used SEM to investigate the relative effects of maternal diet, maternal metabolic state, and maternal inflammation on offspring central and peripheral inflammatory outcomes. This statistical modeling technique has the advantage of being able to simultaneously estimate complex relationships between maternal metabolic and inflammatory factors and offspring outcome measures, including both direct (e.g., WSD→offspring inflammation) and indirect (e.g., WSD→adiposity→inflammation) effects. Model estimation proceeded as follows. First, we examined the influence of WSD and adiposity on offspring central and peripheral inflammation.

Specifically, the offspring inflammation measure was regressed on maternal adiposity and maternal WSD. Additionally, maternal adiposity was regressed on WSD. Offspring central and peripheral inflammatory variables were considered in separate models. Second, maternal inflammation was added to this model as a mediator of the effect of WSD and adiposity on offspring inflammation. Specifically, maternal inflammation was regressed on WSD and maternal adiposity. Additionally, offspring inflammation was regressed on maternal inflammation. Maternal cytokines and chemokines were assessed in separate models. The statistical significance of indirect effects were tested using the model indirect command.

3. RESULTS

Our group has previously characterized the metabolic changes experienced by dams consuming a WSD compared to a control diet in our NHP model (Elsakr et al., 2021; Thompson et al., 2017). Briefly, pregnant dams consuming a WSD have increased body fat percentage, and impaired insulin sensitivity and glucose metabolism. However, in our NHP model we do not observe a significant difference in gestational weight gain between diet groups during pregnancy (Thompson et al., 2017).

3.1. Confirmatory Factor Analysis Model Fit

Prior to hypothesis testing, four confirmatory factor analyses were used to create latent variables (described above) to model various inflammatory profiles. All latent variables, including maternal chemokines [χ² (4) = 4.088, p = 0.394, CFI = 0.999, TLI = 0.997, RMSEA = 0.014], maternal pro-inflammatory cytokines, [χ² (3) = 4.073, p = 0.254, CFI = 0.989, TLI = 0.965, RMSEA = 0.055] offspring chemokines [χ² (9) = 7.69, p = 0.57, CFI = 1.000, TLI = 1.033, RMSEA = 0.000], and offspring cytokines [χ² (22) = 26.622, p = 0.226, CFI = 0.972, TLI = 0.954, RMSEA = 0.048], fit the data adequately. All factor loadings were statistically significant (p<.05), and the variances of the latent variables were significant (Table S2). Together, these results suggest that these latent variables represent a statistically sound and appropriate way to consider these cytokine data.

3.2. Offspring microglial counts in specific amygdala subregions were associated with maternal diet and adiposity but not circulating inflammatory factors.

Results from the model used to test the influence of maternal diet, adiposity, and gestation length during pregnancy on offspring microglial counts in the lateral amygdala are presented in Figure 3. Previous work has suggested that length of gestation can have profound and lasting effects on neurodevelopment (Davis et al., 2011; Espel et al., 2014), so we included this measure in our models for this study. This model fit the data well (CFI = 0.981, TLI = 0.885, RMSEA = 0.063). Maternal WSD was significantly associated with decreased offspring microglial counts (β_{Diet → offspring microglia counts} = −0.622, SE = .195, p<0.01), when accounting for adiposity and gestational length. We also found evidence that WSD indirectly affected microglial counts via WSD’s effects on pre-pregnant adiposity (β_{indirect WSD → pre-pregnant adiposity → offspring microglia counts} = 0.207, SE = .093, p<0.05). Specifically, maternal WSD was associated with increased adiposity prior to pregnancy (β = 0.349, SE = .087, p<0.05) compared to CTR animals, which in turn was associated with greater microglial cell counts (β = 0.593, SE = .180, p<0.01). Finally, we found that length of gestation significantly influenced microglial counts when accounting for maternal diet and adiposity (β_{gestation length → offspring microglia counts} = 0.164, SE = .214, p<0.05) where increases in gestation time were associated with increased microglia number. In sensitivity analyses aimed at determining if this effect was specific to a particular subregion of the amygdala, this model was re-run 6 times, one model for each of the 6 subregions described in the “automated cell counting analysis” methods above. In these models, the total microglial count in each of the subregions considered in place of our primary analysis variable. In these supplemental analyses, maternal variables were only associated with the BLD and BLI subregions. These subregions followed a similar pattern and directionality of association as observed between maternal variables and the lateral amygdala (Table 2). Previous studies indicate that microglia are highly responsive to local cues in the parenchyma (De Biase et al., 2017) and perform crosstalk events with the periphery (Liu et al., 2020). To account for the possibility that offspring circulating inflammatory factors were driving the changes we see in microglia counts in the amygdala we tested if offspring cytokines or chemokines were associated with microglia counts in the amygdala and found that there was no association (β =0.360, SE = .324, p=0.266, and β =0.243, SE = .266, p=0.362 respectively). Additionally, to account for offspring diet consumed post-weaning, we ran sensitivity analyses including offspring diet as an indicator regressed onto microglial cell counts and found no significant effects in any subregion (data not shown).

Figure 3.

Path analysis model including maternal metabolic state measures and 13-month offspring Lateral amygdala microglial cell counts. Indirect mediated pathways through pre-pregnant adiposity significantly predict juvenile offspring microglial cell counts in the Lateral subregion of the right amygdala. Black lines indicate significant direct effects (* = p<0.05, ** = p<0.01) and are labeled with a beta value.

To test if maternal circulating inflammatory factors had a direct effect on the number of microglia in the amygdala of offspring, SEM was used to examine the association between maternal chemokines and the number of microglia counted in the lateral amygdala of offspring (Figure 4A). This model fit the data fit adequately (χ² (9) = 9.906, p=0.358, CFI = 0.994, TLI = 0.985, RMSEA = 0.029). Maternal chemokines did not significantly predict microglial counts in offspring lateral amygdala (β = 0.135, SE = .214, p=0.528). Maternal pro-inflammatory cytokines followed a similar trend of results as maternal chemokines (Figure 4B). Additionally, all other amygdala subregions did not show significant association with maternal circulating inflammatory factors (data not shown).

Figure 4.

Maternal third trimester inflammatory markers do not significantly predict offspring microglial cell counts. Models of maternal chemokine and proinflammatory cytokine latent variables are presented with constituent cytokine protein markers. Corresponding standardized factor loadings are listed in Table S1. Both maternal A) Chemokines and B) pro-inflammatory cytokines do not significantly predict microglia number in the lateral amygdala at 13 months. Grey dashed lines indicate measurements that were estimated but were non-significant (p<0.05) and are labeled with a beta value and specific p-value.

3.3. Offspring peripheral inflammatory factors were associated with maternal adipose-derived decrease in circulating chemokines.

Previous work by our group has suggested that maternal WSD influences circulating maternal chemokine levels via its effects on maternal adiposity (Thompson et al., 2018). Here, we confirm our previous findings and take a further step by suggesting that maternal adiposity-derived third trimester chemokines are associated with offspring cytokines and chemokines at 13-months. Figure 5A presents the results from the SEM used to examine the association between the maternal diet, adiposity and chemokines and offspring cytokines latent variables. This model fit the data adequately (χ² (107) = 117.63, p = 0.227, CFI = 0.961, TLI = 0.950, RMSEA = 0.026). We found higher levels of maternal chemokines were associated with lower levels of offspring cytokines (β = −0.389, SE = .117, p<0.01). We also found a significant indirect effect of maternal WSD on offspring cytokine levels via maternal adiposity-induced increases in maternal chemokine levels (β_{indirect WSD → pre-pregnancy adiposity → maternal chemokines → offspring cytokines} = 0.058, SE = .238, p<0.05). Specifically, maternal WSD was associated with increased adiposity (β = 0.353, SE = .088, p<0.01) compared to controls, which in turn was associated with decreased maternal chemokines (β = −0.426, SE = .133, p<0.01), which resulted in decreased offspring circulating cytokines at 13-months (β = −0.389, SE = .117, p<0.01). Figure 5B presents the results from SEM used to examine the association between the maternal chemokines and offspring chemokines latent variables. The data fit this model adequately (χ² (65) = 79.27, p = 0.110, CFI = 0.942, TLI = 0.919, RMSEA = 0.039). We found maternal chemokine levels were positively associated with offspring chemokine levels (β = 0.298, SE = .149, p<0.05). Similar to Figure 5A, we found maternal WSD was associated with increased maternal adiposity (β = 0.373, SE = .084, p<0.01), which in turn was associated with decreased maternal chemokines (β = −0.413, SE = .112, p<0.01), which were then associated with a decrease in offspring chemokines (β = 0.298, SE = .149, p<0.05). We also ran models examining the influence of a maternal proinflammatory cytokine latent variable on offspring peripheral measures, however no significant effects were found (data not shown). Figure 6A presents results from the model used to test the impact of maternal diet and adiposity on offspring circulating cytokines. This model fit the data adequately (χ² (38) = 46.08, p = 0.173, CFI = 0.952, TLI = 0.931, RMSEA = 0.040). Maternal diet was not significantly associated with offspring peripheral circulating inflammatory factors (β_{WSD → offspring cytokines} = −0.099, p=0.488). Similarly, pre-pregnant adiposity (β_{pre-pregnancy adiposity → offspring cytokines} = 0.114, p = 0.425) was not significantly associated with offspring cytokines. The same trend of results held true for the model that included offspring chemokines in place of offspring cytokines (Figure 6B). Additionally, we performed regression analyses examining the relationships between maternal obesity, maternal and offspring diet and their effects on individual cytokine levels (Table S3).

Figure 5.

Maternal adiposity-induced chemokines influence offspring peripheral inflammatory outcomes. Models of maternal chemokine and offspring chemokine and cytokine latent variables are presented with constituent cytokine protein markers. Corresponding standardized factor loadings are listed in Table S1. A) Indicates the path analysis model including maternal diet, adiposity, and chemokine latent variables influence on offspring cytokine latent variable. Model fit statistics suggest adequate model fit with RMSEA = 0.026, CFI= 0.961, and TLI = 0.95. B) Indicates the path analysis model including maternal diet, adiposity, and chemokine latent variables influence on offspring chemokine latent variable. Model fit statistics suggest adequate model fit with RMSEA = 0.039, CFI= 0.942, and TLI = 0.919. Black lines indicate significant direct effects (* = p<0.05, ** = p<0.01) and are labeled with a beta value. Grey dashed lines indicate measurements that were estimated but were non-significant.

Figure 6.

Maternal metabolic state does not directly influence offspring peripheral inflammatory markers. Models of offspring chemokine and cytokine latent variables are presented with constituent cytokine protein markers. Corresponding standardized factor loadings are listed in Table S1. A) Path analysis model includes maternal diet and adiposity measures and 13-month offspring cytokine latent variable. Model fit statistics suggest adequate model fit with RMSEA = 0.040, CFI= 0.952, and TLI = 0.931. B) Path analysis model includes maternal metabolic state measures and 13-month offspring chemokine latent variable. Model fit statistics suggest adequate model fit with RMSEA = 0.015, CFI= 0.993, and TLI = 0.990. Black arrows indicate significant direct effects and are labeled with β value. Dashed grey arrows indicate non-significant estimations. WSD, Western-style diet. * = p<0.05.

4. DISCUSSION

We hypothesized that maternal consumption of an obesogenic WSD during pregnancy would impact juvenile offspring inflammatory outcomes via obesity-induced maternal inflammation. The current study provides a novel examination of the unique effects of several maternal factors on the developing offspring through the use of structural equation modeling. For the first time, we found that the number of microglia in the basolateral amygdala of juvenile NHP offspring were associated with maternal diet and adiposity, but not maternal circulating inflammatory factors. We also found that juvenile NHP offspring peripheral inflammatory outcomes in the form of basal circulating cytokine and chemokines levels were associated with maternal obesity-induced circulating chemokine levels during the third trimester of gestation. These results suggest that maternal diet, adiposity, and inflammation differentially influence various offspring inflammatory outcomes.

The precise mechanisms by which maternal WSD and adiposity during gestation influence microglia number or circulating inflammatory markers in offspring are not fully elucidated. The findings of this research informs several possible mechanisms that warrant exploration in future work. First, we originally hypothesized that increases in maternal adiposity influences the developing offspring through an inflammatory mechanism. Increases in adiposity are associated with alterations in circulating inflammatory markers, such as increases in proinflammatory cytokines and decreases in chemokines (Aye et al., 2014; Thompson et al., 2018). This obesity derived inflammatory state is then able to either directly or indirectly impact the developing fetus (Figure 7). This hypothesis appears to be true when considering offspring peripheral circulating inflammatory marker outcomes as we saw that decreases in maternal circulating chemokines influenced offspring chemokine and cytokine levels.

Figure 7.

Conceptual figure describing the results observed in this study. Consuming a WSD leads to increased adiposity, this increased adiposity can affect central and peripheral inflammatory outcomes in juvenile offspring. Peripherally, adiposity tissue leads to decreased maternal circulating chemokines which in turn increases cytokines and decreases chemokines in circulation of juvenile offspring. Centrally, increased maternal adiposity leads to increased microglial counts in juvenile offspring. Created with BioRender.com.

4.1. Mechanisms by which Maternal Diet and Obesity-Induced Inflammation Impacts Fetal Development

Adipose tissue is hormonally active and maintains dynamic communication with the immune system through cytokine release and macrophage recruitment (Kershaw and Flier, 2004). Obesogenic diets alone can stimulate cytokine release through adipocyte Toll-like receptor 4 (TLR-4) activation (Rogero and Calder, 2018). The increased adipose tissue mass associated with obesity exacerbates inflammation and contributes to constitutive cytokine expression (Kern et al., 2001; Sindhu et al., 2015). Adipose tissue hypertrophy is associated with large, fragile adipocytes (Monteiro et al., 2006) and increased tissue hypoxia and reactive oxygen species generation (Engin, 2017). This contributes to non-apoptotic adipocyte death, a pro-inflammatory process that leads to macrophage recruitment, proliferation, and cytokine release from adipose tissue (Cinti et al., 2005). The overall effect of obesity and WSD consumption is termed “meta-inflammation,” a chronic condition of low-level elevated inflammation and immune activation (Li et al., 2018). These changes in maternal immune and inflammatory profile during pregnancy may impact fetal development to produce the alterations in basal levels of inflammatory markers we observed in juvenile offspring.

The immune response during a healthy pregnancy is a complicated, less understood process that requires a delicate balance between pro and anti-inflammatory responses to provide protection from external pathogens while simultaneously preventing the rejection of the developing fetus by the mother’s body. Longitudinal studies in humans suggest that there is a reduction in pro-inflammatory cytokine production coinciding with a progressively increasing expression of anti-inflammatory biomarkers (Graham et al., 2017). However, there are conflicting results across both human and animal models that are likely due to variations in study design, such as species, specific biomarkers examined, and the possible lack of inclusion of underrepresented populations (Ander et al., 2019; Gillespie et al., 2016). Despite conflicting results in what is considered “typical” progression of the inflammatory response during pregnancy, disruptions to this finely tuned process are proposed to impact fetal brain development. This has been consistently shown in epidemiological studies of maternal immune activation, where children born to mothers who experienced an infection display increased risk for behavioral outcomes associated with schizophrenia and autism spectrum disorder (Brown and Meyer, 2018). Interestingly, these same associated risks for neuropsychiatric disorders overlap strongly with outcomes of animal models where offspring are exposed to maternal consumption of a WSD (Davis and Mire, 2021).

In addition to obesity-induced inflammatory changes, specific nutrients that are consumed in excess when eating a WSD (saturated fat, simple carbohydrates) have been shown to have inflammatory signaling capabilities, but the precise mechanisms are not fully elucidated. The route by which saturated fatty acids (SFAs) enact inflammatory effects is heavily debated. Some evidence suggests SFAs such as palmitate can stimulate TLR4 receptors (Rocha et al., 2016). However, other studies claim that SFAs do not directly stimulate TLR4 receptors, but instead enact their effects through obesity-induced increases in circulating LPS which reprograms macrophage metabolism to be more sensitive to subsequent stimulation by SFAs and induces further inflammatory signaling (Erridge and Samani, 2009; GI et al., 2018). Similarly, the inflammatory actions of simple carbohydrates may be indirect. It is postulated that chronic consumption of high amounts of sugar promotes lipogenesis in the liver eventually leading to lipotoxicity (Softic et al., 2016). This lipotoxicity is then suggested to be involved in inflammatory processes such as reactive oxygen species formation (Alkhouri et al., 2009).

4.2. Offspring Microglia Influenced by Maternal Diet and Obesity but Not Maternal Circulating Inflammation

We found that microglia counts were specifically impacted by maternal diet, adipose tissue mass, and length of gestation, and not by maternal or offspring circulating inflammatory molecules. This finding contradicts the proposed mechanism of maternal obesity-derived inflammation directly programming microglia. We see in Figure 3 that maternal diet and adiposity have unique but opposing effects on microglial cells counts in the offspring. This finding may suggest a compensatory response by offspring to maternal WSD exposure on microglial number in the brain. Additionally, the unique effects of maternal diet and adiposity may result in alterations in other aspects of microglia such as morphology, which has not been captured in this current study. One possible mechanism by which increased adiposity may exert its effects on the number of microglia in the brain of offspring is through adipose tissue-derived hormones, such as leptin, which has been suggested to influence fetal brain development (Figure 8) (Valleau and Sullivan, 2014). The roles of other adipokines such as, adipokine and neuregulin, have not been extensively studied in offspring brain development. Additionally, it has been shown that obesity during pregnancy affects sex steroid concentrations depending on fetal sex (Maliqueo et al., 2017). This is important because microglia are known to be highly responsive to modulation by sex hormones during gestation (VanRyzin et al., 2019). This mechanism is currently understudied but could provide much needed context for maternal adiposity’s ability to influence offspring microglial number.

Figure 8.

Conceptual figure illustrating potential mechanisms underlying adiposity’s ability to increase microglia counts in the amygdala of juvenile offspring. The first mechanism shown here includes increased adipose mass leading to increased circulating fatty acids that can cross both the placental and blood-brain barriers to impact microglial membrane composition as well as function. Second, increased adipose tissue could release abnormal levels of adipose-derived hormones such as adiponectin, leptin, neuregulin, as well as impact steroid hormones such as estrogen and testosterone. These altered levels of hormones could impact microglial state and possibly lead to increased microglia counts in the amygdala. Created with BioRender.com.

Excess fatty acids crossing the placenta into fetal circulation is a potential mechanism shared by maternal diet and adiposity for influencing offspring neural outcomes. Maternal obesity is known to reduce the efficiency of adipose tissue fatty acid turnover which results in increases in circulating fatty acids as well as ectopic fatty acid accumulation (Spalding et al., 2017). Additionally, maternal diet, outside the context of obesity, can modulate the levels of lipid content in fetal brains, as well as microglia membranes specifically (Bowen and Clandinin, 2005; Destaillats et al., 2010; Rey et al., 2018).

This is important given the role of membrane lipids in cellular signaling. Finally, under homeostatic conditions microglia express several lipid-sensitive receptors such as triggering receptor expressed on myeloid cells 2 (TREM2), cluster of differentiation 36 (CD36), and Toll-like receptors (TLRs) (Mauerer et al., 2009). These lipid-sensitive receptors also perform traditional inflammatory signaling. This overlap may provide a key connection between the increased microglial number in the amygdala and maternal overconsumption of fats in an obesogenic diet.

4.3. Offspring peripheral inflammatory state impacted by maternal obesity-induced circulating inflammation

We found that offspring peripheral inflammatory outcomes were indirectly programmed by maternal adiposity’s impact on circulating maternal chemokines during the third trimester. Previous work examining maternal circulating inflammatory factors have focused on single factors (e.g., IL-6, CRP, TNFα), but here, through the use of confirmatory factor analysis and the creation of latent variables, we are able to take a novel approach to examine how many related factors (i.e., chemokines and cytokines) exert their effects in concert as a measure of the overall inflammatory burden variable rather than individually. With that said, even though certain maternal inflammatory factors may load well onto a latent variable (indicating that they are capturing a shared underlying construct), it is possible that individual markers may exert specific effects on the developing offspring. It is important to explore this in future research as these specific effects could be targeted for therapeutic intervention in the future. The offspring immune system can be programmed by several different maternal factors such as psychosocial stress and nutrition (Palmer, 2011), as well as maternal immune activation (Mandal et al., 2013). Two major maternal-fetal immune transfer mechanism are the placenta prenatally and breast milk postnatally. Our group has previously reported that breast milk from WSD dams contained lower levels of eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA) as well as lower levels of total protein than CTR dams (Grant et al., 2011). It is plausible that the reduction in circulating chemokines of obese individuals in our study is perceived by the developing offspring’s immune system at the placental interface during gestation or through breast milk postnatally.

4.4. Limitations

The current study contained a few limitations. In addition to alterations in number, microglia change other aspects of their phenotype to respond to various cues. This current study is limited by only measuring counts, but additional measures that should be considered in future work are morphology (ramified vs ameboid), gene expression, and the proteins downstream for various inflammatory signaling cascades. Collecting these additional measures provides a more comprehensive picture of the overall state of offspring microglia. Furthermore, in this study we examined microglia, which are only one of the cell types in the brain. This decision was guided by previous work performed by our group which suggested that microglia were impacted by maternal WSD in the hypothalamus at a fetal 3^rd trimester timepoint (Grayson et al., 2010). However, there are other important glial cells, such as astrocytes, that may be impacted by maternal diet and adiposity during development that have the capacity to influence behavior outcomes. Additionally, the small sample size of offspring in the IHC experiment may have limited our ability to detect only large effects where we may not have been powered to detect small to intermediate effects sizes. Our models, like most statistical tests, capture linear associations among our study variables. We did not find evidence of non-linear associations among our focal variables; however future research should further explore this possibility. Finally, another limitation of this study is based on the translatability of the specific diet used here. The WSD utilized in this study was created to reflect the average American diet and may not reflect the actual dietary practice of any one individual.

4.5. Conclusion and Future Directions

While we did not examine the effects of maternal diet, adiposity and inflammatory state on offspring behavioral outcomes in this study, previous work by our group and others shows strong evidence that consuming a WSD and the resulting metabolic and inflammatory outcomes influence offspring behavior, such as increased anxiety and depressive behaviors (Ramirez et al., 2020; Sullivan et al., 2010; Thompson et al., 2018). Thus, our findings may have implications for understanding offspring risk for mental health disorder. The precise mechanisms by which maternal factors influence offspring behavior remain to be fully elucidated but studies suggest that modulation of maternal and offspring inflammatory systems may play a role (Gustafsson et al., 2020; Thompson et al., 2018). Additionally, findings from this study that implicate alterations in microglia in the juvenile amygdala indicate a possible mechanism for the behavioral alterations observed previously in the offspring as the amygdala is known to be involved with emotional regulation. While the amygdala is a key brain region involved in behavior, future studies should examine other brain regions involved in modulating behavior such as the pre-frontal cortex.

Interventions that target and modify the maternal immune system may be useful for ameliorating the programming effects on offspring circulating inflammatory factors as well as behavior. Currently, one of the more promising intervention routes in the context of maternal nutrition, is omega-3 (n-3) polyunsaturated fatty acid (PUFA) supplementation. N-3 PUFA consumption is generally reduced when individuals are consuming a WSD compared to other diets such as a Mediterranean diet. Preclinical studies that examine n-3 PUFAs consumption during pregnancy suggest n-3 supplementation may protect against offspring behavioral risk associated with increased maternal adiposity (Gustafsson et al., 2019) as well as modify microglial phagocytic capacity (Madore et al., 2020; Madore et al., 2014). Further work to understand strategies that may be able to relieve adverse health outcomes in offspring born to obese mother is an important field of research.

Figure 2.

Representative image of AChE stain in the right amygdala at 4x magnification. Dashed lines indicate approximate locations of subregions based on reference of the non-human primate atlas (Paxinos et al., 2009). The inlet provides a representative image of IBA-1 positive cells at 25x magnification that were used in analysis of microglia counts. AA=anterior, BLD=basolateral dorsal, BLI=basolateral intermediate, BLVM=basolateral ventromedial, BMMC=basomedial magnocellular, Ce=central, LA=lateral, Me=medial, PAL=paralaminar.

Table 3.

Results of Models Relating Maternal Metabolic State and Offspring Microglial Counts in Amygdala Subregions

	Total			Lateral			Basolateral-dorsal			Basolateral-intermediate
Direct Effects	β(SE)	P	95 % CI of β	β(SE)	P	95 % CI of β	β(SE)	P	95 % CI of β	β(SE)	P	95 % CI of β
Maternal WSD → Amygdala region	−.443(.292)	.129	−.925, .038	−.554(.207)	.007	−.336, .196	−.483(.226)	.033	−.855, −.111	−.368(.252)	.144	−.782, .046
Pre-Pregnancy Adiposity → Amygdala region	.433(.269)	.100	.000, .886	.628(.166)	.001	.354, .901	.471(.215)	.029	.117, .824	.521(.206)	.011	.182, .860
Gestation Length → Amygdala region	.114(.092)	.215	−.037, .264	.164(.077)	.034	.036, .291	.226(.080)	.005	.095, .358	.138(.096)	.153	−.021, .297
Indirect Effect Paths
Maternal WSD → Pre-Pregnancy Adiposity → Amygdala region	.155(.107)	.145	−.020, .331	.219(.094)	.016	.069, .369	.166(.091)	.068	.016, .315	.183(.094)	.050	.029, .336

Note: WSD = Western Style Diet; Bold numbers indicate significance at p<0.05.

Highlights:

• Offspring amygdala microglia count were associated with maternal diet and adiposity

• Microglia number in the offspring were not associated with maternal inflammation.

• Offspring cytokines were positively associated with maternal chemokines.

• Offspring chemokines were negatively associated with maternal chemokines.

• Maternal diet and adiposity exert unique effect on offspring inflammatory outcomes.