Loss of pleasure (Anhedonia): A sex-dependent, cross-species consequence of unpredictable signals to the perinatal brain

Abstract

Reward seeking is an essential component of human behavior. Anhedonia (reduced motivation for and pleasure in reward) is a core feature of several mental illnesses. Early life adversity associates with later impairments in reward behaviors and related psychopathology; however, specific aspects of early adversity that lead to disrupted reward behaviors are unknown. Capitalizing on evidence that unpredictable signals to the developing brain constitute adversity and influence neurodevelopment, we examined whether exposure to perinatal unpredictable maternal signals associated with diminished reward responsivity, and probed causality using experiments in mice. We analyzed a longitudinal human study (N=158 mother/child dyads) and a controlled prospective mouse study (N=111 pups; 17 litters). In humans, exposure to unpredictable signals during late gestation was assessed by maternal mood entropy (a measure of how disordered or random signals are). In mice, unpredictable maternal care behaviors were experimentally induced during the neonatal period, and maternal care entropy was similarly computed. Aspects of anhedonia were determined by adolescent self-report and diminished behavioral responses to palatable food in young adult mice. Across species, unpredictable maternal signals were associated with reduced reward responsivity in males, but not females. Associations persisted after accounting for income and maternal psychopathology in humans, and for cohort and litter effects in mice. These findings indicate a robust, sex-dependent link between early exposure to maternal unpredictability and disrupted reward processing. The convergence of cross-species results indicates that unpredictability of maternal signals plays a role in shaping reward behaviors in males, with implications for the developmental origins of mental illness.

1. Introduction

Motivation for rewards plays an important role in decision-making, and learning across development. Reduced motivation for and pleasure in rewarding experiences (“anhedonia”) is a transdiagnostic risk factor for mental illness including depression, substance use disorders, schizophrenia and posttraumatic stress disorder (Pizzagalli, 2022). Anhedonia can manifest early in development and is linked to psychopathology risk across the lifespan (Irwin et al., 2024; Luby et al., 2004; Pizzagalli, 2022; Nevarez-Brewster et al., 2026). Adverse experiences during early development increase lifetime risk of mood and substance use disorders, which are characterized by altered reward processing (Hanson et al., 2021; Herzberg and Gunnar, 2020; Kasparek et al., 2023; Luby et al., 2017; Novick et al., 2018). Exposure to adversity may exert particularly potent influences during sensitive periods of rapid brain growth, such as during the perinatal period (Glynn and Baram, 2019; Luby et al., 2020; Nelson and Gabard-Durnam, 2020). Therefore, understanding the impact of perinatal exposure to adversity on reward processing is required to allow prevention and intervention strategies for promoting mental health.

Recent research identified exposure to unpredictable early life signals from caregivers and the environment as a form of adversity that is linked to neurodevelopmental outcomes (Davis and Glynn, 2024; Glynn et al., 2018; Glynn and Baram, 2019; Davis et al., 2017; Spadoni et al., 2022; Jirsaraie et al., 2024; Ellis et al., 2022; Kangas et al., 2022; Xu et al., 2023; Ugarte and Hastings, 2024). Mechanistic animal studies have provided evidence that early postnatal unpredictable maternal signals impact the maturation of reward- and stress-related circuitry, leading to altered reward behaviors (Birnie et al., 2022; Birnie and Baram, 2022; Bolton et al., 2018a, Bolton et al., 2018b; Levis et al., 2022; Molet et al., 2016). Consistent with animal findings, human studies show that early life unpredictability is associated with reward-related phenotypes such as anhedonia during adolescence and adulthood (Glynn et al., 2019; Null et al., 2024; Spadoni et al., 2022). In humans, prenatal maternal mood may provide important environmental signals to the growing fetus via a panoply of physiological processes that influence fetal neurodevelopment and postnatal behavior (Glynn et al., 2018; Jirsaraie et al., 2024; Howland et al., 2021). Thus, patterns of maternal signals provided in the late-gestational and early postnatal (perinatal) period may shape long-term reward circuitry in offspring.

Motivation for investigating sex differences in associations between early exposure to parental unpredictability and later anhedonia is based on well-established sex differences in the rates and symptoms of mental illnesses (Hankin, 2009). The male and female brain differ already in the prenatal period, and males and females respond differently to early life adversity as early as the perinatal period (Davis and Pfaff, 2014; Kooiker et al., 2023; Mueller and Bale, 2008). Our recent experimental animal work indicates that early life exposure to the same unpredictable parental signals exerts sexually dimorphic consequences, with anhedonia-like behaviors only in male rodents (Birnie et al., 2023; Bolton et al., 2018; Bolton et al., 2018; Floriou-Servou et al., 2026; Levis et al., 2021, 2022; Molet et al., 2016; Rincón-Cortés and Sullivan, 2016; Taniguchi et al., 2026). Specifically, adult males with unpredictable early life maternal care seek for natural and drug rewards less than controls (Birnie et al., 2023; Floriou-Servou et al., 2026; Levis et al., 2022). In contrast, exposure to unpredictable early life maternal care in females increases the propensity for seeking of both natural and drug rewards (Floriou-Servou et al., 2026; Levis et al., 2021; Taniguchi et al., 2026). In view of this robust experimental animal data, we tested here the hypothesis that early life unpredictability exerts sex-dependent consequences including male-specific anhedonia.

We assessed patterns of maternal signals using entropy, a construct characterizing the unpredictability of stochastic processes (Glynn et al., 2018). Entropy-based metrics quantify the degree to which signals are unpredictable (i.e., random), with higher entropy scores reflecting more unpredictable/random sequences of signals or behaviors. Studies of both human and animal models have used entropy to quantify the predictability of maternal signals (Vegetabile et al., 2019)and have demonstrated that unpredictability of maternal signals have important implications for offspring development (e.g., Davis et al., 2017; Davis et al., 2022). Additionally, in humans, intraindividual variability in item scores on mood-related questionnaires, as quantified by mood entropy, reflect the degree of unpredictability in mood symptoms at a given time point (Glynn et al., 2018). Importantly, mood entropy has been linked with offspring risk for psychopathology above and beyond mood level (e.g., Glynn et al., 2018; Howland et al., 2021). Thus, across species, entropy provides a quantitative index of parental unpredictability, allowing one to determine the extent to which patterns of maternal signals has important implications for development.

In the present study, we examined unpredictable human maternal mood during the third trimester of pregnancy and experimentally induced unpredictable mouse maternal behaviors in the first postnatal week. We highlight the fact that it is not possible to precisely align brain developmental ages in humans and mice because, in the two species, different brain ‘modules’ or circuits develop in distinctly different velocities and trajectories (Avishai-Eliner, 2002; Birnie et al., 2020). However, we consider that the developmental stage of the reward system during the first 4-6 postnatal days in mice is similar to that seen during late gestation in the human (Birnie et al., 2020; Cottam et al., 2025). Across both species, we found that early exposure to unpredictable maternal signals during a distinct period of development is associated with anhedonia in males, but not females, suggesting a causal, sex-specific impact of early unpredictability on reward behaviors in males.

2. Methods

2.1. Study overview

We evaluated the relation between exposure to unpredictability of maternal signals and subsequent anhedonia of the offspring in both humans and mice. In humans, exposure to unpredictable maternal signals was assessed by computing entropy of maternal mood during the third trimester of pregnancy (M = 36.4 weeks of gestation, SD = 1.74). Offspring self-reported on anhedonia symptoms during early adolescence (M = 14.09 years, SD = 1.44). In mice, exposure to unpredictable maternal signals was experimentally manipulated during postnatal days 2 to 10. Entropy rates of maternal sensory signals were computed based on observations of maternal behaviors during postnatal days 4 to 6. Anhedonic behavior was evaluated by measuring consumption of limited access, highly palatable food by the offspring on an assessment date between postnatal days 90 to 210.

2.2. Participants

Human Participants. Study participants included mother-child dyads who participated in a longitudinal study from the prenatal period through to adolescence. Mothers were recruited during their first trimester of pregnancy. Recruitment occurred at affiliated clinics from one of two large university medical centers in Southern California. Participant inclusion criteria included singleton pregnancy, maternal age over 18, English-speaking, and non-smoking. Research was approved by the responsible Institutional Review Board, and all mothers provided written informed consent. Additionally, children provided assent during the adolescent visit.

For the current study, 158 mothers and their children (50% female) were included based on participation during adolescence. Logistic regressions showed that participants included for the present study did not significantly differ from participants in the larger longitudinal study based on demographic characteristics (e.g., maternal education), or maternal psychopathology and parenting behaviors during infancy. Half (50%) of mothers and 44% of children identified as White, non-Hispanic, while 33% of mothers and 35% of children identified as Hispanic. The median income-to-needs ratios of families was 4.48. Table S1 contains detailed participant demographics for the present study.

Mouse Subjects. C57BI/6J males and dams were ordered from Jackson Laboratories (strain #000664) and maintained as breeding pairs in-house. Seventeen litters (8 control, 9 experimental group, see below) were progeny of 7 dams (N = 111 total offspring; 31 control males, 30 experimental males, 27 control females, 23 experimental females). Food and water were available ad libitum. All mice were housed in standard conditions at 68-72 °F and 40-60% humidity on a 12-h light–dark cycle (Lights ON: 0630; Lights OFF: 1830) until behavioral testing. In adulthood and at least 2 weeks before behavioral testing, offspring were moved to a separate, behavior-dedicated reverse light cycle room, assigned a rearing-condition blind testing ID, and single housed (Lights ON: 0000; Lights OFF: 1200). Experimental and control mice were tested for reward behaviors within sex and groups were age-matched (within 2 weeks). Testing was thus conducted in cohorts that always included age-matched control and experimental groups. Mice tested were aged 90-210 days. All procedures were approved by University of California Irvine IACUC (AUP 21-128) and were in accordance with the guidelines from the National Institute of Health.

2.3. Measures

2.3.1. Quantifying exposure to unpredictability in maternal signals

Human Maternal Mood Measures. To assess maternal mood, mothers completed questionnaires on their depressive symptoms, state anxiety, and perceived stress at the third trimester and three months postpartum and pregnancy-specific anxiety during the third trimester.

Maternal Depressive Symptoms. The Center for Epidemiologic Studies Depression Scale (Santor and Coyne, 1997; CESD) was used to assess maternal depressive symptoms. The CESD consists of 9 items that are scored on a 4-point Likert scale, ranging from 0 (Rarely or None of the Time) to 3 (Most or All of the Time). A total score was computed by summing all items, with a possible range from 0 to 27. Higher scores reflect more severity in depressive symptoms. There was good internal consistency in our sample (α = .88).

Maternal State Anxiety. The state anxiety subscale of the State-Trait Personality Inventory (Spielberger, 1979) was used to assess maternal state anxiety. The state anxiety subscale consists of 10 items that are scored on a 4-point Likert Scale ranging from 1 (Almost Never) to 4 (Almost Always). In our study, this scale exhibited good internal consistency (α = .90).

Perceived Stress Scale. The Perceived Stress Scale (Cohen et al., 1983; PSS) was used to assess how often mothers felt they were able to manage problems and hassles, and how often they felt nervous or stressed. The PSS contains 14 items that are scored on a 5-point Likert Scale from 0 (Never) to 4 (Very Often). The PSS had good internal consistency in our sample (α = .92).

Maternal Pregnancy Specific Anxiety. The Pregnancy Specific Anxiety Scale (Rini et al., 1999; PSA) was used to assess pregnant women's feelings about their health during pregnancy, the health of the baby, and their feelings about labor and delivery. The PSA contains 10 items (e.g., “I am concerned or worried about developing medical problems during my pregnancy”) that were scored on a 4-point Likert scale from 1 (Not at all) to 4 (Very much). There was good internal consistency in our sample (α = .84).

Mood Entropy. Shannon's Entropy was computed to characterize maternal mood unpredictability (Glynn et al., 2018). Shannon's entropy is a measure of the degree of unpredictability or uncertainty of the occurrence of an event. In the context of the current study, Shannon's entropy determined the predictability of a mother's response on each item within the same session on mood-related questionnaires. For example, mothers consistently selecting 2 for each item of the CESD would have a low entropy score as they are responding predictably. On the other hand, mothers selecting more variable responses would have a higher entropy score. We normalized the entropy scores to facilitate comparisons across scales, such that entropy values range from 0 to 100. Prior work has established the validity of mood entropy scores against variability measures derived from repeated sampling of mood using ecological momentary assessment (Glynn et al., 2018; Howland et al., 2021).

Mood Level. To characterize maternal mood level, we created a composite by standardizing total sum scores for each mood-related questionnaire and creating z-scores. We then averaged across all measures. Higher mood level scores reflect more negative mood.

2.3.1.1. Mouse maternal behavior

Experimental Manipulation. In mice, we imposed a resource scarcity paradigm consisting of limited bedding and nesting (LBN) in the cages. This reliably results in dam stress which, in turn, leads to maternal care behaviors that are fragmented and unpredictable (Ivy et al., 2008; Molet et al., 2016; Rice et al., 2008). The limited bedding and nesting (LBN) protocol was set up and carried out as described previously (Ivy et al., 2008; Maras et al., 2014; Molet et al., 2016; Rice et al., 2008); briefly, on postnatal day (P)2, pups and dams were placed in either a “typical” new cage (one full nestlet, 300 mL corncob bedding) or an LBN new cage (½ nestlet, 120 mL corncob bedding, housing a small-gauge mesh platform elevated 1 inch above the floor such that dropping, but not pups, fall through to cage floor). Following a week in these distinct rearing conditions, pups and dams of both groups were returned to a new, “typical” cage on P10. Maternal care was documented by video recordings.

Maternal Care Scoring and Entropy Calculation. Maternal behavior videos during the first hour of the dark phase from days P4-6 were scored with Behavioral Observation Research Interactive Software (Friard and Gamba, 2016). Trained external observers scored the first hour of the dark phase as previous work (e.g., Ivy et al., 2008) demonstrated that during the dark, “active” phase the consistency of behaviors was more substantially altered. As the videos of maternal care necessarily capture the cage environment as well as the dam, scorers could not be blinded to rearing condition, but were blinded to outcome variables. The ethogram included maternal dam behaviors (licking and grooming, nestbuilding, nursing, moving on the nest, and carrying pups), and non-maternal dam behaviors (self-grooming, moving off nest, and eating and drinking) as in (Molet et al., 2016). After scoring, “non-maternal” behaviors were collapsed into a single category. Pup-oriented behaviors (licking and grooming, nestbuilding, nursing, carrying pups, moving on nest) and “non-maternal” behavior were then processed in R 4.4.0 (R Core Team, 2020) to produce a time-invariant first-order Markov Chain transition matrix, and the entropy of the transition matrix for each scored hour of behavior was derived (Molet et al., 2016).

2.3.2. Assessing anhedonia behaviors

Anhedonia symptoms in humans. To assess anhedonia symptoms, adolescents completed the Snaith-Hamilton Pleasure Scale (Snaith et al., 1995; SHAPS). The SHAPS is a 14-item questionnaire assessing anhedonic symptoms (e.g., “I would enjoy being with my family or close friends”) on a 4-point Likert scale from 0 (Strongly Disagree) to 3 (Strongly Agree). The scale has been validated in adolescent non-clinical populations (Leventhal et al., 2015). There was good internal consistency in our sample (α = .88). Following the original scoring algorithm (Snaith et al., 1995), we recoded each item as dichotomous (Strongly Agree or Agree = 0; Strongly Disagree or Disagree = 1), with scores ≥3 indicating clinically elevated levels. Approximately 28% of our sample had clinically elevated levels of anhedonia.

Palatable food consumption in mice. Palatable food testing was conducted in the home cage in the home room under red light conditions over 4 consecutive days, including 3 days of habituation (Days 1-3) and 1 testing day (Day 4). For the week prior to testing, mouse cages were not changed or otherwise disturbed. The night before habituation began, a pinch of Cocoa Pebbles (POST™) was placed in each cage and left overnight to minimize novelty aversion, and all pebbles were consumed overnight. Cocoa Pebbles were placed in the cage by gently sliding the cage lid back <5 cm and pouring the cereal in the front right corner. Cages were not otherwise moved or disrupted prior to administration of Cocoa Pebbles. To avoid effects of scent, male and female cages were never opened at the same time and also resided on separate racks in the same room.

Each day began between 1300 and 1530 (early active phase). Approximately 1 g (1.00 g ± 0.05 g) of Cocoa Pebbles per mouse was measured out and placed in the home cage as described above. After 1 h, cages were gently removed from the rack, and the majority of remaining Cocoa Pebbles were quickly removed. Following the 1-h test period on day 4, mice were moved to a new cage and the testing cage was retained to allow for accurate retrieval of all remaining Cocoa Pebbles. Consumption is reported as the difference in weight of Cocoa Pebbles before and after day 4 testing. On test day, all mice ate more than 0.15 g but less than 0.85 g of cereal, with a mean across all cohorts of 0.50 g; this suggests the absence of any floor or ceiling effects in consumption. Mean raw consumption varied by cohort, likely a result of humidity and temperature variations, animal age and other uncontrollable events (Molet et al., 2016; Butler-Struben et al., 2022). Mean raw consumption also varied by sex in typically-reared mice, as expected from previous reports (Birnie et al., 2023; Taniguchi et al., 2026). Therefore, to enable cross-cohort comparisons, z-scores of raw palatable food consumption were computed within sex and cohort. The mean and standard deviation were calculated to compute z-scores for each mouse in the respective sex and cohort, and these z-scores were carried forward for analysis.

2.4. Statistical analysis

Identification of covariates (humans). We selected two covariates to include in our analyses: age at adolescent visit and family income-to-needs ratio. We selected these covariates based on conceptual rationale. First, research shows that there are developmental changes in anhedonia symptoms across adolescence (Bennik et al., 2014). Similarly, income-to-needs ratio was included as a covariate given known associations with parental mood, unpredictability, and child anhedonia (e.g., Glynn et al., 2019,Barch et al., 2016).

Associations between prenatal maternal mood entropy and adolescent anhedonia symptoms (humans). We first ran descriptive statistics and bivariate correlations between our variables of interest (Table S2). Our anhedonia outcome was count distributed, and we selected the best-fitting distribution to use in our modeling based on metrics examining predicted zeros and overdispersion. Our results indicated that the number of anhedonia symptoms endorsed was best characterized by a negative binomial distribution (Fig. S1). Thus, generalized linear models with a negative binomial distribution for the anhedonia symptom score were conducted in Mplus 8.6 (Muthén and Muthén, 1998) to examine sex differences in associations between prenatal maternal mood unpredictability and anhedonia symptoms. We also conducted logistic regressions in R 4.6.0. The primary analyses examined prenatal maternal mood entropy, child sex at birth, and their interaction as predictors of anhedonia. Additionally, we conducted sensitivity analyses to test if associations among our main variables of interest changed with the inclusion of (1) prenatal mood level, (2) postnatal mood entropy, and (3) postnatal mood level. Data visualizations were created using ggplot2 in R 4.4.1.

Degree of unpredictability of maternal care signals and analyses of offspring anhedonia-like behaviors (mice). We first used mixed effect models to compare the maternal care entropy levels in the typical and LBN cages, accounting for random intercepts for each dam. We then ran two mixed effect models with z-scored hedonic feeding (highly palatable food consumption) as the outcome. The first model examined experimental condition (typical vs. LBN cages) as the primary predictor. The second model examined maternal care entropy rates as the primary predictor. In both models, sex was included as a moderator. To account for the hierarchical structure of the data, models included random intercepts for dam and litter nested within dam. Data visualizations were created using ggplot2 in R 4.5.2.

3. Results

3.1. Unpredictable maternal signals (entropy) during late pregnancy portend adolescent anhedonia symptoms in a sex-dependent manner in humans

Predictability of maternal signals was assessed during the third trimester by computing mood entropy, which is a validated method for assessing unpredictability of maternal signals (Glynn et al., 2018; Howland et al., 2021), where higher mood entropy values represent more unpredictable mood. We established that maternal mood entropy scores during the third trimester were approximately normally distributed (Fig. 1A).

Fig. 1.

Sex differences in anhedonia symptoms following exposure to unpredictable maternal signals early in life.

1A. Prenatal maternal mood entropy scores during the third trimester were approximately normally distributed, with higher entropy scores reflecting more unpredictable maternal signals.

1B. Exposure to higher levels of unpredictability (high entropy) was associated with anhedonia symptoms for males (b = 0.02, SE = 0.01, p = .005), but not females (b = −0.00, S.E. = 0.01, p = .801) in humans.

1C. Exposure to higher levels of unpredictability increased the odds of having clinical levels of anhedonia for males (OR = 1.11, 95% CI [1.04, 1.20], but not females (OR = 1.00, 95% CI [0.96, 1.03]).

Adolescent offspring of these mothers (mean age = 14.09 years, SD = 1.44) self-reported their anhedonia symptoms. These identified sex-specific consequences of prenatal mood entropy for adolescent anhedonia symptoms (b = −0.02, S.E. = 0.01, p = .040) (Table 1). Specifically, higher prenatal mood entropy associated with more anhedonia symptoms in males (b = 0.02, SE = 0.01, p = .005), but not females (b = −0.00, S.E. = 0.01, p = .801). Fig. 1B depicts the interaction between prenatal mood entropy and adolescent sex on anhedonia symptoms. Notably, the association between prenatal mood entropy and anhedonia persisted after accounting for established aspects of early life stress such as income-to-needs ratio (Table 1). The sex-specific effects of prenatal mood entropy were replicated when we examined risk of clinically significant anhedonic symptoms (b = −0.11, S.E. = 0.04, p = .008; Fig. 1C). Exposure to higher levels of unpredictability was associated with an approximately 11% increase in the odds of clinically significant anhedonia in males (OR = 1.11, 95% CI [1.04, 1.20]), whereas no meaningful association was observed in females (OR = 1.00, 95% CI [0.96, 1.03]).

Table 1.

Results from generalized linear model with a negative binomial distribution for adolescent anhedonia symptoms examining prenatal mood entropy, child sex (0 = male, 1 = female), and their interaction as predictors of adolescent anhedonia.

Model Predictors	Estimate	S.E.	p	95% CI
Intercept	−2.38	1.21	0.048	[-4.74, −0.02]
Income-to-Needs Ratio	−0.18	0.13	0.146	[-0.43, 0.06]
Age at Adolescent Visit	0.15	0.07	0.034	[0.01, 0.29]
Prenatal Mood Entropy	0.02	0.01	0.005	[0.01, 0.04]
Child Sex	0.94	0.65	0.147	[-0.33, 2.21]
Prenatal Mood Entropy × Child Sex	−0.02	0.01	0.040	[-0.05, −0.00]

Sensitivity analyses demonstrated that associations were specific to prenatal maternal entropy (unpredictability) and remained robust after accounting for prenatal maternal mood level (i.e., average levels of anxiety, stress, and depression; see Table 2). Next, sensitivity analyses evaluated whether prenatal mood entropy predicted adolescent anhedonia after covarying for postnatal influences. The association of prenatal mood entropy and adolescent anhedonia remained after including postnatal maternal mood level in the models (Table 2). When postnatal mood entropy was added to the model, the interaction between prenatal mood entropy and child sex attenuated to p = .073, likely reflecting the moderate correlation between prenatal and postnatal mood entropy (r = 0.45, p < .001).

Table 2.

Sensitivity analysis of model testing prenatal mood entropy, child sex, and their interaction as predictors of adolescent anhedonia.

Prenatal Models	Estimate	S.E.	p	95% CI
With Concurrent Mood Level
Intercept	−2.07	1.16	0.073	[-4.34, 0.19]
Income-to-Needs Ratio	−0.16	0.13	0.241	[-0.42, 0.11]
Age at Adolescent Visit	0.14	0.07	0.038	[0.01, 0.28]
Prenatal Mood Entropy	0.02	0.01	0.054	[0.00, 0.04]
Prenatal Mood Level	0.11	0.14	0.422	[-0.16, 0.38]
Child Sex	0.92	0.66	0.162	[-0.37, 2.21]
Prenatal Mood Entropy × Sex	−0.02	0.01	0.038	[-0.05, −0.00]
With Postnatal Mood Entropy
Intercept	−1.85	1.16	0.109	[-4.11, 0.42]
Income-to-Needs Ratio	−0.18	0.13	0.171	[-0.44, 0.08]
Age at Adolescent Visit	0.15	0.07	0.023	[0.02, 0.29]
Prenatal Mood Entropy	0.03	0.01	0.000	[0.01, 0.05]
Postnatal Mood Entropy	−0.02	0.01	0.006	[-0.03, −0.01]
Child Sex	0.55	0.62	0.376	[-0.66, 1.75]
Prenatal Mood Entropy × Sex	−0.02	0.01	.073a	[-0.04, 0.00]
With Postnatal Mood Level
Intercept	−2.18	1.19	0.066	[-4.52, 0.15]
Income-to-Needs Ratio	−0.16	0.13	0.229	[-0.41, 0.10]
Age at Adolescent Visit	0.14	0.07	0.046	[0.00, 0.28]
Prenatal Mood Entropy	0.02	0.01	0.013	[0.00, 0.04]
Postnatal Mood Level	0.06	0.09	0.506	[-0.12, 0.25]
Child Sex	0.94	0.66	0.153	[-0.35, 2.24]
Prenatal Mood Entropy × Sex	−0.02	0.01	0.041	[-0.05, −0.00]

^a The Entropy × Sex Interaction term does not reach statistical significance at the p < .050 level, though the effect size is similar.

3.2. Unpredictable maternal care (entropy) precedes sex-dependent, anhedonia-like behaviors in mice

Mouse pups were exposed to fragmented and unpredictable maternal care during the early postnatal period, an age in which the maturity of the reward circuitry of the mouse resembles that of late gestation in humans (Cottam et al., 2025). We quantified unpredictable sequences of maternal care using entropy rates and found that in LBN cages, entropy was significantly higher than in control cages (b = 0.32, 95% CI [0.05, 0.58], p = .022; Fig. 2A) after accounting for variances attributed to dam.

Fig. 2.

Sex differences in anhedonia-like reward behaviors in mice exposed to unpredictable maternal signals early in life.

2A. Raising the litters in cages with limited bedding and nesting (LBN) materials induced unpredictable sequences of maternal care behaviors (high entropy) during the first week of life (comparable to late gestation and early postnatal life in the human)

2B. Rearing in the LBN cages is associated with anhedonia-like behaviors in male but not female mice. Male C57Bl/6J mice exposed to a the LBN environment consumed significantly less palatable food compared with those raised in typical cage environment; In contrast, no effect of experimental condition is apparent in female mice.

2C. Highly unpredictable sequences of maternal behaviors (high entropy) was associated with anhedonia-like behaviors in male mice, suggesting that entropy is the key feature of the LBN rearing that drives later life reward behaviors. Dimensionally, in male C57Bl/6J mice, the degree of maternal care unpredictability (higher entropy) was significantly associated with less palatable food consumption. This effect was not present in females.

Young adult mice with a history of early life exposure to high entropy maternal care and age-matched controls were tested for consumption of palatable food. A linear mixed-effects model (Table 3) showed a significant experimental condition × sex interaction (b = −0.99, 95% CI [−1.66, −0.32], p = .004) on palatable food consumption, such that this consumption differed by experimental group in males but not females. Among males, palatable food consumption was significantly lower in the mice reared in LBN environments compared to those reared in the typical environment (b = −0.84, SE = 0.26, p = .006; Fig. 2B). There were no significant differences among females (b = −0.16, SE = 0.28, p = .578). Furthermore, sex moderated associations between maternal care entropy and palatable food consumption (Table 4, b = −1.44, 95% CI [−2.71, −0.17], p = .027) such that higher maternal care entropy associated with lower palatable food consumption for males (b = −1.53, SE = 0.48, 95% CI [−2.49, −0.56]) but not females (b = −0.09, SE = 0.50, 95% CI [−1.10, 0.92]). We included in these models random intercepts for dam, and litter nested within dam. Variances attributable to dam and litter were negligible (τ² ≈ 0).

Table 3.

Mixed-effect model showing that the effects of Experimental Condition (Exp Condition) on palatable food consumption varies in mice.

Palatable Food Consumption (z-scored)
Predictor	Estimate	95% CI	p	df
(Intercept)	−0.11	−0.46 – 0.25	0.558	104.00
Sex (1=Male, 0=Female)	0.52	0.06–0.98	0.026	104.00
Exp Condition [ELA]	0.16	−0.36 – 0.68	0.554	104.00
Sex × Exp Condition [ELA]	−0.99	−1.66 – -0.32	0.004	104.00
Random Effects
σ2	0.77
τ00DamID:Litter.ID	0.00
τ00DamID	0.01
N DamID	7
N Litter.ID	17
Observations	111
Marginal R2/Conditional R2	0.115/NA

Table 4.

Mixed-effect model showing that the effects of maternal care entropy on palatable food consumption varies by sex in mice.

Palatable Food Consumption (z-scored)
Predictor	Estimate	SE	95% CI	p	df
(Intercept)	0.04	0.43	−0.82 – 0.89	0.931	99.00
Sex (1=Male, 0=Female)	1.43	0.60	0.24–2.63	0.019	99.00
Maternal Care Entropy	−0.09	0.47	−1.02 – 0.84	0.852	99.00
Sex × Entropy	−1.44	0.64	−2.71 – -0.17	0.027	99.00
Random Effects
σ2	0.78
τ00DamID:Litter.ID	0.00
τ00DamID	0.00
N DamID	7
N Litter.ID	16
Observations	106
Marginal R2/Conditional R2	0.107/NA

We also tested if these associations were specific to palatable food consumption, or if there are similar effects on weight gain during the first ten days of life. A linear mixed-effects model showed that there were no associations of experimental condition, sex, or their interaction with postnatal day 10 weight (Table S3). The mixed-effects model included random intercepts for cohort, dam, and litter nested within dam, with most of the random variance attributable to differences between dams (τ² = 0.30) and litters (τ² = 0.09), while variance attributable to cohort was negligible (τ² ≈ 0).

4. Discussion

Patterns of signals during sensitive periods sculpt the maturation of the mammalian brain (Birnie and Baram, 2022; Hubel and Wiesel, 1962). The current cross-species research provides new evidence that exposure to unpredictable patterns of signals during the perinatal period has sex-specific consequences for reward behaviors. This conclusion is supported by the parallel results in humans and in controlled, prospective studies in experimental mice.

Our work addresses several knowledge gaps in the growing literature demonstrating that patterns of maternal signals during early development play an important role in the maturation of brain circuits (Birnie and Baram, 2022; Davis et al., 2017; Davis and Glynn, 2024; Glynn et al., 2018; Glynn and Baram, 2019; Howland et al., 2021). First, our data support that the perinatal period constitutes a sensitive window during which patterns of maternal signals matter. Other sensitive periods (e.g., adolescence) or signals exist and influence the developing brain; however, in both mice and humans we demonstrate here the sensitivity of perinatal brain reward systems to patterns of external stimuli. Second, this work is the first evaluation of the implications of prenatal maternal mood entropy on the reward system in human adolescents. Prior work has shown associations of early exposure to unpredictability and adult anhedonia (e.g., Glynn et al., 2019). Here we show that the association is evident already during adolescence and persists after adjusting for levels (e.g., severity) of maternal mental health symptoms as well as other factors such socioeconomic status. Third, we identified sex-specific consequences on reward behaviors of early life exposure to maternal unpredictability. Although most research has focused on the increase in sex-differences around puberty (Rodríguez-Montes et al., 2023), sex differences can emerge as early as the fetal period, and males and females respond differently to early environmental signals during the perinatal period (Davis and Pfaff, 2014; Mueller and Bale, 2008; Wheelock et al., 2019). Here we show a clear sex-difference in the consequences of perinatal exposures on adolescent phenotypes. Finally, our preclinical studies identified sex-specific effects of early exposure to unpredictability on anhedonia-like behaviors, which were manifest in males only (Bolton et al., 2018; Levis et al., 2022; Molet et al., 2016). The concordance between sex-specific effects of unpredictability on anhedonia across species supports the robustness of our observations. Notably, the cross-species approach suggests that the timing or type of adversity at specific periods of development may explain established sex differences in mental health problems following early life adversity.

There are several strengths of the current study. Notably, we employed a prospective and longitudinal design in both humans and mice. We used entropy as an operational and validated measure of unpredictability in both species. We enhanced observational human studies in which shared genetic and environmental contributions to associations between mothers and children are difficult to disentangle with controlled experimental exposure to unpredictability in mice, lending credence to the notion of a direct link between unpredictable maternal signal patterns and subsequent sex-specific deficits in reward behaviors.

While a key strength of the current paper was the cross-species examination of a shared association, this alignment was imperfect. To match developmental periods for experienced adversity, we assessed gestational maternal signals in humans but postnatal maternal signals in mice. An additional consideration is that prenatal and postnatal maternal mood entropy were moderately correlated, suggesting continuity in the predictability of maternal signals across the perinatal period. We conducted sensitivity analyses to examine whether associations with prenatal mood entropy were independent of postnatal mood entropy. The attenuation of the prenatal association after adjustment for postnatal mood entropy likely reflects shared variance between prenatal and postnatal maternal mood entropy, making their independent contributions difficult to disentangle statistically. Importantly, mood entropy associations remain after covarying for mood level, indicating an independent contribution of unpredictability above and beyond effects of level. Prenatal and postnatal unpredictability share variance and may reflect a broader pattern of maternal signal unpredictability spanning the perinatal period. Accordingly, the independent contributions of prenatal versus postnatal exposures could not be fully disentangled in the current study, and future work with larger samples and repeated assessments across development will be important for identifying potential timing-specific effects.

Further, although maternal entropy was used for both species, the type of maternal signal necessarily differed between humans and mice. Similarly, the means of measuring anhedonia was necessarily different across species; rodent anhedonia assays require behavioral testing. Rodent literature indicates that consumption of palatable food under limited, intermittent access (as done here) is “non-homeostatic” or “hedonic feeding” (Buesing et al., 2025). We thereby used consumption of highly palatable food during a limited period to probe pleasurable experience in rodents. In the current work, a single measure of anhedonia was obtained in each species-the SHAPS in humans and palatable food consumption during a limited period in mice. Whereas the SHAPS is thought to measure primarily motivational aspects of anhedonia, the mouse task, reduced palatable food consumption under low effort conditions, might be more indicative of consummatory anhedonia. The present findings should be considered together with our prior work demonstrating that adult male mice and rats exposed to early life unpredictability display seem to have both motivational and consummatory anhedonia-like behaviors (Floriou-Servou et al., 2026; Taniguchi et al., 2026). Lastly, while most prior work in rodents found, in addition to decreased reward motivation in males, an augmented motivation for natural and drug rewards in females exposed to early life unpredictability, we did not see that typical female phenotype here (Birnie et al., 2022; Floriou-Servou et al., 2026; Kooiker et al., 2024; Levis et al., 2021; Molet et al., 2016; Taniguchi et al., 2026). It should be noted that female mice exposed to unpredictability are consistently more motivated than controls in high-effort tasks, whereas there is more variability in low-effort tasks such as the one used in the present study. Another potential explanation for the absence of a female effect here is the use of highly-inbred C57Bl/6J mouse strain (Newmyer et al., 2019). Alternatively, there may be associations between unpredictability and other reward-related constructs in females that we did not assess in the current study. An additional limitation of our study is the limited number of dams (n = 7) suggesting that random effects estimates should be interpreted cautiously. Analyses of pup weight gain suggested the models are able to detect dam-level clustering, and litter-level sensitivity analyses reported in the supplementary materials further support the robustness of our main findings.

It could be argued that in both humans and mice, other factors rather than maternal unpredictability were the key determinant of reward-related behavioral outcomes. In the LBN cages, scarcity of resources might be a plausible cause for later life anhedonia-like behavior. However, the strong correspondence of the degree of unpredictability (higher entropy) to poorer reward behaviors in male mice (Fig. 2C) supports a key role for maternal behavior patterns as a driver of the deficits in reward behaviors later in life. Similarly, in humans, maternal mood entropy may be associated with other factors that have been implicated in anhedonia and psychopathology risk more broadly (e.g., prenatal mood symptoms and socioeconomic status; Hernandez et al., 2024; Davis and Sandman, 2012; Glynn et al., 2018). The persistent relation of entropy and anhedonia when other early-life experience factors, such as income-to-needs ratios, are included in the model (Table 1) supports a role for mood entropy itself influencing adolescent outcomes.

In summary, the present findings contribute to our understanding of the early etiology of anhedonia and reward-related behaviors. We demonstrate that exposure to unpredictable maternal signals during the perinatal period has lasting, sex-specific consequences for the development of reward-related behavior. These insights offer a foundation for future prevention and intervention strategies that target exposure to patterns of maternal care to reduce risk for anhedonia and related mental health outcomes.

Ethics statement

Human subjects research was approved by the responsible Institutional Review Board at the University of California Irvine (Protocols, 2002-2441 and 2013-9622). All mothers provided written informed consent. Additionally, children provided assent during the adolescent visit. All study procedures with mice were approved by University of California Irvine IACUC (AUP 21-128) and were in accordance with the guidelines from the National Institute of Health.

Funding information

Supported by NIH grants P50 MH096889; R01NS041298, R01HD028413, R01HD051852 HD040967 and the Donald Bren Foundation. A.M. Zhou was supported by a NIH postdoctoral training grant (T32MH015442). M.R. Tetzlaff was supported by the UCI MSTP training grant (T32GM008620).

CRediT authorship contribution statement

Anna M. Zhou: Conceptualization, Data curation, Formal analysis, Methodology, Validation, Visualization, Writing – original draft, Writing – review & editing. Madison R. Tetzlaff: Data curation, Formal analysis, Investigation, Methodology, Project administration, Writing – original draft, Writing – review & editing. Laura M. Glynn: Funding acquisition, Investigation, Methodology, Project administration, Supervision, Writing – review & editing. Curt A. Sandman: Funding acquisition, Investigation, Methodology, Project administration, Writing – review & editing. Hal S. Stern: Formal analysis, Methodology, Supervision, Writing – review & editing. Tallie Z. Baram: Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Writing – original draft, Writing – review & editing. Elysia Poggi Davis: Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Writing – original draft, Writing – review & editing.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Handling Editor: Prof R Lawrence Reagan

Appendix A. Supplementary data

The following is/are the supplementary data to this article:

Data availability

The data that support the findings of this study will be openly available at https://digitalcommons.du.edu/.