Perinatal western‐type diet and associated gestational weight gain alter postpartum maternal mood

Abstract Introduction The role of perinatal diet in postpartum maternal mood disorders, including depression and anxiety, remains unclear. We investigated whether perinatal consumption of a Western‐type diet (high in fat and branched‐chain amino acids [BCAA]) and associated gestational weight gain (GWG) cause serotonin dysregulation in the central nervous system (CNS), resulting in postpartum depression and anxiety (PPD/A). Methods Mouse dams were fed one of four diets (high‐fat/high BCAA, low‐fat/high BCAA, high‐fat, and low‐fat) prior to mating and throughout gestation and lactation. Postpartum behavioral assessments were conducted, and plasma and brain tissues assayed. To evaluate potential clinical utility, we conducted preliminary human studies using data from an extant sample of 17 primiparous women with high GWG, comparing across self‐reported postpartum mood symptoms using the Edinburgh Postnatal Depression Scale (EPDS) for percent GWG and plasma amino acid levels. Results Mouse dams fed the high‐fat/high BCAA diet gained more weight per kcal consumed, and BCAA‐supplemented dams lost weight more slowly postpartum. Dams on BCAA‐supplemented diets exhibited increased PPD/A‐like behavior, decreased dopaminergic function, and decreased plasma tyrosine and histidine levels when assessed on postnatal day (P)8. Preliminary human data showed that GWG accounted for 29% of the variance in EPDS scores. Histidine was also lower in women with higher EPDS scores. Conclusions These findings highlight the role of perinatal diet and excess GWG in the development of postpartum mood disorders.

branched-chain amino acids (BCAA), found mostly in red meats but also in pork, poultry, eggs, and fish. BCAA include leucine (Leu), isoleucine (Ile), and valine (Val), and are three of the nine essential amino acids that must be obtained exogenously via the food supply (Harper, Miller, & Block, 1984). The BCAA are also important during pregnancy and postpartum recovery. These states induce significant metabolic changes, including protein accretion, that are designed to synthesize new maternal and fetal tissues, ensure the fetus receives a continuous supply of nutrients despite intermittent maternal food intake, and support lactation after birth (Kalhan, 2000;Kalhan & Parimi, 2006;Kalhan, Rossi, Gruca, Super, & Savin, 1998). During healthy pregnancy there is a downregulation of the rate of BCAA transamination, especially for leucine, resulting in higher plasma and extracellular fluid levels of BCAA as compared to the nonpregnant state (Jolly et al., 2004;Kalhan and Parimi). When there is excess BCAA consumption as is typical in the Western-type diet, this increases the BCAA load on an already downregulated system, resulting in lower rates of protein turnover and higher whole-body levels of BCAA (Jolly et al., 2004;Kalhan & Parimi, 2006).
We hypothesize that perinatal consumption of a Western diet increases risk for PPD/A by increasing concentrations of plasma BCAA, causing competition for transport of all large neutral amino acids across the blood-brain barrier into the central nervous system (CNS) via the Large neutral Amino acid Transporter 1 (LAT-1) (Fernstrom, 2005). This includes transport of BCAA as well as the amino acids histidine (HIS), tryptophan (TRP), and tyrosine (TYR), which are precursors to key neurotransmitters implicated in PPD/A (Nutt, 2001;Senkowski, Linden, Zubrägel, Bär, & Gallinat, 2003;Werner & Coveñas, 2010) (i.e., histamine, serotonin, dopamine, and norepinephrine). An increase in one of the large neutral amino acids induces decreased uptake of the others, and vice versa (Fernstrom, 2005). Furthermore, the rate of production of these neurotransmitters depends on the availability of their amino acid precursors (Wurtman & Fernstrom, 1975;Young & Gauthier, 1981).
The primary aim of this study was to use a mouse model to investigate the hypothesis that perinatal consumption of a diet high in fat and BCAA, as in a Western-type diet, and associated gestational weight gain (GWG) cause serotonin dysregulation in the CNS resulting in postpartum depressive-like and anxiety-like behavior. The secondary aim was to determine whether we could establish preliminary clinical evidence in a small cohort of mothers for the relationships among diet, GWG, BCAA concentrations, and postpartum mood identified in the animal studies. This work is clinically relevant because diet is a modifiable behavior that is safe to manipulate during pregnancy when based on principles of sound nutrition.

| Animals
All procedures were approved by the Duke University Institutional Animal Care and Use Committee. Juvenile (P28) female C57BL/6 mice were obtained from Charles River Laboratories (Raleigh, NC) in three separate cohorts (n = 189) and housed in individually ventilated cages with ad libitum access to food (PicoLab Mouse Diet 5058, Lab-Diet, Philadelphia, Pennsylvania) and filtered water. Adult breeder males were placed with 2-3 females each for breeding, for a maximum of 3 weeks. The colony was maintained at 22°C on a 12:12-h light-dark cycle (lights on at 7 AM). Sentinel animals were housed in the colony room and screened periodically for the presence of common rodent diseases; all screens were negative.
The HFD and HFD/BCAA diets provided 4.7 kcal/g; the LFD provided 3.84 kcal/g; and the LFD/BCAA provided 3.85 kcal/g. The complete nutritional profiles of each diet are available online (http:// www.researchdiets.com). Females were weighed prior to diet assignment (baseline weights did not differ significantly by diet group), and every 3 days thereafter until parturition (the final peripartum weight was thus collected between E16 and E18, see Figure 2d). Food intake was measured by weighing the food every 3 days and dividing the change in food weight by the number of females in the cage. After parturition, to minimize disruption when dams were alone with their litters, maternal weights and food consumption were recorded on P10, P14, P21, and P28.
Females not pregnant after 3 weeks with males were tested after 13 weeks on diet (to mirror the time pregnant females remained on diet prior to sacrifice) in EZM, OFT, and FST, and sacrificed to collect blood and brain samples for later analysis (n = 6-15/group; pregnancy rates varied nonsignificantly by group).

| Maternal care
Maternal care was assessed on days P2-P8 for 3 hr each day, twice during the light phase (0900-1000, 1500-1600 hr) and once during the dark phase (2100-2200 hr). Raters (n = 5) trained by an experienced (Löwe, Kroenke, Herzog, & Gräfe, 2004) coder to >95% reliability and blinded to experimental condition live scored the instantaneous occurrence (frequency) of the following behaviors every 5 min for the duration of the hour: being on the nest, arch-backed nursing, blanket nursing (lying down with little to no arch in the back to nurse pups), F I G U R E 1 Experimental timeline for animal studies. P35-P40 female mice were placed on one of four diets (LFD/Control, HFD/Control, LFD/BCAA, or HFD/BCAA) and fed ad libitum for 6 weeks before the introduction of adult breeder males. Dams (n = 16-20/diet group) were singly housed once visibly pregnant, were allowed to give birth naturally, and remained on the diets throughout pregnancy and lactation. All litters were observed and maternal behavior scored three times daily from P2 to P8. Eight days after parturition, half of the dams (n = 8-11/ diet group) underwent behavioral assessments in elevated zero maze (EZM), open field test (OFT), and Porsolt forced-swim test (FST), blood draws, and sacrifice to obtain brain tissue for analysis. For the remaining dams (n = 8-9/diet group), pups were weaned on P28, and immediately afterward dams underwent behavioral assessments, blood draws, and sacrifice for brain tissue analysis as at P8

| Elevated zero maze
Elevated zero maze is a widely used test for anxiety-like behavior in rodents (Bradburn, Frankel, Baker, & Pergamit, 1991). The maze has an elevated (49.5 cm high) circular lane (4.5 cm wide) divided into four quadrants. Two opposite quadrants are enclosed by walls

| Open field test
Open field test was used to assess activity levels and anxiety-like behavior (Schauder, Zavelberg, Langer, & Herbertz, 1987). Animals were placed in a square enclosure 40 cm by 45.1 cm with walls 34.9 cm high and allowed to explore freely for 10 min. Total distance traveled, mean speed, and time mobile were used as metrics of overall activity, while time in the surround (the area closest to the walls of the maze, ~50% of total area) and percent of distance traveled in the surround were used to quantify anxiety-like behavior. All behaviors were recorded using ANY-maze video-tracking software (Stoelting Co., Wood Dale, IL, USA).

| Forced-swim test
Forced-swim test is a widely used test for depressive-like behaviors in rodents (Castagné et al., 2011) that was selected to preclude the changes to dietary intake inherent with other tests, such as sucrose consumption, and the dangers associated with the tail suspension task in accommodating high body weights due to diet and pregnancy. The apparatus is a Plexiglas cylinder (20.3 cm diameter × 49.5 cm high) filled two third full with 22°C water. Each mouse was lowered into the center of the water-filled cylinder, and then scored for the duration of immobility for 6 min. Immobility was defined as the absence of all swimming, except for the subtle motions required to keep the head above water. In our tests, immobility time was not observed to be dependent on the body weight of the animal. This test is stressful for the animals; therefore, it was conducted last on the day of sacrifice to avoid unwanted behavioral effects in other tests. Animals were sacrificed within 3-5 min of the end of the forced-swim test to minimize the effects of stress on brain neurochemistry.

| Blood and brain collection
Blood (100-150 μl) was drawn via the facial maxillary gland using a 5-mm lancet and collected into microcentrifuge tubes containing EDTA. Whole-blood glucose was assessed using a glucometer (TRUE2go; CVS Pharmacy, Inc., Woonsocket, RI, USA). Mice were then anesthetized via ketamine/xylazine (430 mg/kg ketamine; 65 mg/kg xylazine, i.p.), and transcardially perfused with ice-cold saline for 2 min. Whole brains were rapidly extracted and dissected into hippocampus (HIPP; cut into two halves) and prefrontal cortex (PFC), and snap frozen in liquid nitrogen. Blood was centrifuged at 1,200 g at 4°C for 10 minutes, and plasma was removed and stored along with brain tissue at −80°C until assayed. Briefly, 1 μl of cDNA was added to 12 μl of master mix containing specific primers for genes of interest. We designed primer pairs for all genes (Gapdh, Htr1a, Htr2c, Slc6a4, Drd1, and Drd2; see Table 1) as previously described (Tai et al., 2010). Designed primers were obtained from Integrated DNA Technologies, Inc. (Coralville, IA, USA). Optimal annealing temperatures for each primer pair were determined by running a temperature gradient, and specificity was verified by melt-curve analysis. For analysis, we determined the threshold cycle (CT) for each reaction and calculated relative gene expression using the 2 −ΔΔCt method (Livak & Schmittgen, 2001;Rao, Huang, Zhou, & Lin, 2013;Tai et al., 2010), using Gapdh as a reference gene.

| High-performance liquid chromatography with electrochemical detection (HPLC-EC)
Two hundred and fifty microliters of ice-cold standard buffer

| Human subjects
The Duke University Institutional Review Board approved all procedures. Data were analyzed from 17 women who participated in a pilot randomized trial examining health coaching for women who exceeded Institute of Medicine (IOM) recommendations for GWG, based on prepregnancy body mass index (BMI). Eligible participants included primiparous women between 18 and 45 years of age with full-term, uncomplicated, singleton pregnancies who were recruited within 24 hr of delivery from the birthing centers at Duke Regional Hospital and the Duke University Medical Center. Exclusion criteria included current or recent (within the last 6 months) smoker, inability to read and understand English, and complications of delivery that would increase risk of blood draw (e.g., dehydration, anemia). Data included for the present analysis were obtained at baseline via the electronic medical record or at 4-6 weeks postpartum via standardized self-report instruments using a computer-assisted self-interview (CASI) prior to participation in any intervention procedures.

| Blood samples
A fasting (≥8 hr) 60-ml blood sample was obtained by antecubetal venipuncture within 48 hr of delivery and prior to hospital discharge. Samples were drawn into chilled EDTA-treated vacutainers, immediately transferred on ice to the laboratory, and spun for 10 min at 1200 g using a balanced, refrigerated (4 o C) centrifuge.
The plasma, sera, and buffy coats were then transferred to 1.7ml polypropylene microcentrifuge tubes and stored at −70°C until assayed.

| Postpartum mood
Approximately 1  Specifically, the EPDS includes measurement of symptoms of both depression and anxiety, and total scores have been validated to identify risk for either disorder (Brouwers et al., 2001). In this study, a cut-off score of 10 was used for the EPDS, which has reported sensitivities and specificities of 70-90% in multiple studies, is between the high (e.g., 12, 13) and low (e.g., 6, 7, 8) cut-off scores used previously (Cox et al., 1987;Eberhard-Gran et al., 2001), and is a recommended cut-off score for psychiatric referral (Eberhard-

| Gestational weight gain (GWG)
GWG was calculated as the percent change in BMI (%ΔBMI) from prepregnancy to last recorded pregnant weight (occurring no more than 1 week prior to delivery). Data were abstracted from the medical record to obtain BMI, calculated using the standard formula (weight

| Data analysis and statistics
All data were collected without knowledge of treatment group and analyzed using SPSS statistical software (IBM, Armonk, NY, USA).
All animal measures were analyzed using two-way (HFD X BCAA) ANOVAs. If heterogeneous variance was noted between groups, outliers were first identified and removed via Tukey's IQR method.
Following significant F scores for interactions, post hoc comparisons (Tukey's HSD) were performed to further distinguish among groups, and all differences were considered statistically significant if p < .05.
Descriptive statistics were calculated for human subjects' age, EPDS scores, and percent change in BMI from prepregnancy to delivery. A simple correlation and coefficient of determination were calculated relating BMI change to EPDS score. Plasma amino acids in human samples were analyzed via nonparametric Wilcoxon rank-sum tests (also known as Mann-Whitney U tests), due to the disparity in sample size between the groups.

| Diets high in fat and BCAA generate an obese phenotype in female mice
Both HFD and BCAA animals gained significantly more weight than con- palatability and/or energy density of these diets (Coppola et al., 2013).
However, HFD/BCAA animals gained significantly more weight per

| BCAA-supplemented diets induce a PPD-like phenotype, and diets high in fat induce anxietylike behavior
At 8  which may also indicate elevated anxiety (Detke, Rickels, & Lucki, 1995;Heisler et al., 1998). Together, these data suggest that HFDinduced anxiety is not restricted to the early postpartum period, as was the case for the BCAA effect.

| HFD and BCAA dams exhibit altered maternal care behaviors
During the critical first postnatal week, LFD/BCAA dams had lower nest scores relative to control dams [based on a nest quality score of 0 to 3 performed by observers blind to experimental group; significant HFD X BCAA interaction, F(1,62) = 7.05, p < .05; post hoc, p < .05; In contrast, none of these differences was found in the brains of P28 dams (data not shown), consistent with the behavioral effects.
HFD dams exhibited changes in the opposite direction, that is, a   Figure 5f]. None of these changes was evident in P28 dams (data not shown), by which time the PPD/A-like phenotype had disappeared.
We also examined the prefrontal cortex, which receives substantial dopaminergic innervation from striatal regions. We found a simi- dams (data not shown). There were no significant differences in dopamine turnover rates, due to a similar trend for a decrease in DA itself, although this did not reach significance. There were also no significant differences in serotonergic neurochemistry in this region (Figure 5g).

| Western-type diets and PPD/A are associated with altered metabolomic profiles in both mice and humans
One long-term goal of this body of research is to identify a circulating biomarker that could aid diagnosis of PPD/A, as well as elucidate its Mean EPDS scores were 6.76 (SD = 4.69; range 0-18), and using a cut-off score of 10, 23.5% of the sample (N = 4) had significant risk for PPD/A.
As it is difficult to control or manipulate diet in humans, we used excess GWG as a proxy for excess caloric intake (Stuebe, Oken, & Gillman, 2009), which we hypothesized to be a risk factor for the development of PPD/A. Accordingly, we analyzed the association between %ΔBMI during pregnancy in our cohort and EPDS scores, F I G U R E 5 Diets high in fat and BCAA alter postpartum monoamine neurochemistry. (a) Expression of Htr1a, the gene that encodes the 5-HT1a serotonin receptor, did not change due to diet in the hippocampus of P8 dams. (b) Expression of Slc6a4, which encodes the 5-HTT serotonin transporter, was significantly decreased in the hippocampus of P8 BCAA dams, but increased by HFD. (c) Expression of Drd2, the gene that encodes the D2 dopamine receptor, was significantly decreased in the hippocampus of BCAA dams, but increased by HFD at P8. (d) HPLC-EC analysis of monoamine neurotransmitter levels in the hippocampus of P8 dams revealed no significant differences due to diet in serotonin turnover rate. (e-f) However, BCAA supplementation induced marked decreases in dopamine turnover rates in the hippocampus at P8, as represented by the DOPAC/DA ratio (e) and HVA/DA ratio (f). (g-i) HPLC-EC analysis of monoamine neurotransmitter levels in the prefrontal cortex of P8 dams revealed no significant differences due to diet in levels of the serotonin metabolite 5-HIAA (g), but marked differences in levels of the dopamine metabolites DOPAC (h) and HVA (i), just as in the hippocampus of P8 dams. Data are mean ± SEM, n = 7-11/group. and discovered a significant positive correlation [r s (19) = 0.54, p < .05; Figure 7a]. This correlation has an R 2 of 0.29, meaning that the %ΔBMI accounts for 29% of the variance in EPDS scores.
Analysis of plasma amino acids showed that of the 15 amino acids measured, changes were observed only in histidine, where mean levels were significantly lower (Wilcoxon p = .045) in women with PPD/A risk versus without PPD/A risk (Figure 7b), which is concordant with the findings in mice, as described above.

| DISCUSSION
This study represents the first reported evidence that perinatal consumption of a Western-type diet high in fat and BCAA, and associated excess GWG, may increase the risk for PPD/A, findings that warrant future investigation to validate the relationship among perinatal diet, PPD/A, and associated mechanisms. In mice, consuming a diet high in BCAA induced a depressive-like phenotype that was not observed in nonpregnant females, suggesting PPD risk increases as a function of the interaction between pregnancy and components of the Western diet. This phenomenon is further compounded by the addition of high fat, which we and others have shown induces anxiety-like behavior (Bilbo & Tsang, 2010;Bolton & Bilbo, 2014). Moreover, we report for the first time that this evidence of a relationship between GWG and PPD/A risk identified in mice may extend to humans, although future studies with larger samples sizes should validate these findings.
In our study, women who gained more gestational weight as measured by percent change in BMI reported more symptoms of PPD/A. This study is also the first to identify a potential amino acid biomarker for PPD/A risk. Plasma metabolomic profiling showed that levels of histidine, an α-amino acid, were significantly lower both in the animals fed BCAA-supplemented diets and in the women with PPD/A risk. Histidine is a precursor to histamine, which in the central nervous system modulates food intake as well as regulates other neurotransmitter systems associated with depression, including serotonin, dopamine, and norepinephrine (Hough, 1999;Kano et al., 2004;Nutt, 2008). Tyrosine levels in the animal plasma were also significantly de- Contrary to the prediction from our model in Figure 8, we found slightly decreased, or unchanged, levels of circulating BCAA in BCAAsupplemented mouse dams. However, the regulation of circulating BCAA levels is complex, and plasma levels of BCAA are unlikely to be reliable markers of BCAA status (Tom & Nair, 2006). A decrease in plasma BCAA could indicate increased oxidation, decreased protein breakdown, or increased protein synthesis in the muscle, but this could also indicate that BCAA are being transported across the blood-brain barrier at a higher rate, thus inhibiting the transport of other large neutral amino acids into the brain. We were also surprised to find no clear changes in serotonin levels or metabolites in animals with PPD/A-like behavior. Rather, there were significant changes in dopamine metabolites in the brain, and circulating tyrosine and histidine. However, the etiology of depression is complex, and in recent years, other studies have found roles for dopamine and histamine in depression and anxiety (Raber, 2005;Tye et al., 2013;Zweifel et al., 2011). Research on the maternal brain in humans has also revealed that new mothers exhibit greater activation of the dopaminergic system of the caudate nucleus in response to the body odor of newborns than nulliparous women (Lundström et al., 2013).  (Goodman, 2004;Oʼhara & Swain, 1996). P28 in mice is after the pups will naturally begin to wean themselves and switch to solid food (~P25), so it is at the end of the lactation period (Curley et al., 2009;Hall & Williams, 1983). It is thus possible that the behavioral and neurochemical changes we observed could be linked to lactation, and specifically, that we identified a susceptibility in the BCAA group that is unmasked by lactation. Future studies should explore this possibility.
In our animal model, the observed changes in maternal mood were also associated with alterations in maternal care, which may have serious consequences for offspring health and development. Indeed, F I G U R E 7 Human studies reveal that gestational weight gain is positively associated with PPD risk, and plasma histidine should be investigated further as a prospective biomarker for PPD. (a) In our cohort of women, % ΔBMI during pregnancy is positively correlated with self-reported depression (EPDS scores), such that % ΔBMI accounts for 29% of the variance in depression scores. (b) Metabolomics analysis of postpartum human plasma identified histidine as the only amino acid that was significantly changed in women with PPD risk, which is in agreement with our animal model PPD/A in humans have been linked to adverse child outcomes, including later cognitive deficits and emotional disturbances (Feldman et al., 2009;Murray & Cooper, 1996;O'CONNOR, Heron, & Glover, 2002), and PPD specifically has been associated with offspring growth retardation and risk for being underweight (Rahman, Iqbal, Bunn, Lovel, & Harrington, 2004). Similarly, we observed that BCAA mothers had smaller pups at weaning. HFD/BCAA dams also had fewer pups survive to weaning, which may be a consequence of poor maternal care and neglect, or may be related to poor health of the pups due to maternal diet, but independent of maternal behavior. Further research is needed to distinguish between these possibilities.  (Detke et al., 1995;Heisler et al., 1998). It is also possible that HFD may be protective against depressive-like behavior, although future studies are needed to confirm this hypothesis.
Excessive self-grooming, observed in HFD dams, is one of the hallmark symptoms in an OCD mouse model, which is also associated with increased anxiety-like behaviors (Shmelkov et al., 2010;Welch et al., 2007). Notably, studies in humans have found that up to one third of women report obsessive-compulsive behaviors in the early postpartum period, often in association with negative maternal mood (Miller, Hoxha, Wisner, & Gossett, 2015).
While the current studies suggest that perinatal dietary intake, and specifically consumption of a Western diet and associated GWG, increases PPD/A risk, additional studies are needed to validate this relationship and the proposed mechanism of action in humans.
Controlled feeding trials in pregnant women allow for the most dietary regulation and may provide the data necessary to validate findings from the current animal models. Additionally, behavioral interventions designed to change dietary components during pregnancy (e.g., reduce proportions of unhealthy fats and BCAA/red meats consumed) should be conducted to determine whether altering these components lowers incidence of PPD/A for women with and without other biobehavioral risk factors for PPD/A. Finally, studies that consider the role of genetic risk for mood disorders and obesity in the development of PPD/A should include metabolomic profiling to determine whether associated metabolite patterns may be used as a biomarker for PPD/A risk.
Findings from this study must be considered in light of several limitations. First, the sample size for the human cohort was small, and findings may not be generalizable to all women with PPD/A risk.
Second, animal models, although extremely valuable for translational research in many instances, cannot always translate directly to clinical populations (Seok et al., 2013;Takao & Miyakawa, 2015). The model of a Western diet that we utilized enabled us to study the components of high fat and high BCAA separately and in interaction with each other to address our mechanistic hypothesis, but at the cost of including a high carbohydrate or high sugar component, which has also been implicated in a Western-type diet. Third, we did not have data on prenatal dietary intake or mood among human participants. However, previous studies have shown that excess GWG is associated with total energy consumed prenatally, consumption of fried foods, which are high in both saturated fats and BCAA, and consumption of a Western diet (Stuebe et al., 2009). Moreover, excess GWG was associated with higher EPDS scores, even without consideration of dietary patterns.
F I G U R E 8 Proposed mechanism for how perinatal consumption of a Western diet may result in increased postpartum depression risk. BCAA levels increase as a function of normal pregnancy and excess dietary consumption, and compete with the amino acid precursors (e.g., histidine, tyrosine, and tryptophan) to mood-altering neurotransmitters, for transport across the blood-brain barrier via their common transporter, LAT-1, thereby limiting their uptake. Because synthesis of these neurotransmitters, including histamine, dopamine, and serotonin, is dependent on the availability of their precursor amino acids, increased concentrations of BCAA decrease their production and thus, increase PPD risk Thus, the relationship of GWG to PPD/A should be investigated fur-