Introduction

Historically, sex differences were largely ignored or were avoided. Investigators would justify the exclusion of females because they were “too complicated,” largely as a function of ovarian cyclicity that compromised efforts to achieve a “stable” hormonal environment. The obvious correlate (and paradox) of this objection is that there is something about the female ovarian cycle that must be relevant as an explanation for the inferred variance in outcomes that would be observed were females included in studies. Indeed, as evidence accumulated for the ubiquity and consequential nature of sex differences, the importance of sex as the source of untapped factors for resilience and susceptibility—across the medical spectrum—became clear and led to the decision by the National Institutes of Health (NIH) in 2015 to require examination of sex in NIH-funded studies. With the proliferation of reported sex differences, several categories of sex differences can be identified: sex differences in physiology can converge in producing the same outcome, the same physiologic processes can diverge and result in different outcomes, reported sex differences may be of unclear physiologic consequence [1], and finally, reported sex differences may be artefactual (e.g., a product of post hoc secondary analysis in the wake of negative findings in the primary analysis) [2]. The real complexity, however, as well as the potential explanatory power emerges when one attempts to map a role of sex onto an outcome, in our case, affective regulation. Sex can directly impact etiopathogenesis affecting physiologically relevant outcomes (e.g., synaptic pruning in autism), impact fundamental brain processes (e.g., arousal, reward) or peripheral physiological processes (e.g., immune or hepatic function) that can influence symptom expression or response to treatment (e.g., pharmacokinetics or dynamics), differentially affect response to the environment (e.g., stress exposure), elicit different responses from the environment (e.g., cultural consequences of sex), create differential exposure to hormones (which can alter cellular function and response at the receptor, signal transduction, transcriptional and translational levels), alter cellular metabolism in a sex chromosome-specific, hormone-independent fashion, and program physiologic function—sensitivity and resilience—as a consequence of sex chromosome dosage and prenatal sex steroid exposure. By any standard, this represents a daunting complexity.

In this article, our task—definition of the potential contribution of sex differences to the understanding of affective regulation—is further complicated by both the volume of reported findings and the considerable gaps in our knowledge of the substrates and processes underlying affective regulation. Quite simply, the combined scope of this topic precludes comprehensive review. Consequently, we will provide a framework for thinking about the contributions of sex to affective illness focused on four questions: How would sex alter brain function; Are there sex differences in the brain; Are there sex differences in the substrates of affective regulation; How might sex contribute to differential capacity for affective regulation?

Overview of observed sex differences in affective disorders

The most consistent and robust sex difference reported in affective disorders is the twofold increased lifetime prevalence of major depression (MDD) in women compared with men [3,4,5,6]. This increased prevalence has been observed in a variety of countries [5]. Similarly, a two- to threefold increased prevalence of dysthymia and threefold increase in seasonal affective disorder [7] as well as increased lifetime risks for other stress-related disorders (e.g., anxiety disorders and posttraumatic stress disorder (PTSD) [8,9,10]) in women have also been noted. Although the prevalence of bipolar disorder is equi-prevalent in men and women [3, 6, 11] (reviewed in ref. [12]), women are more likely to develop rapid cycling [12] and may be more susceptible to antidepressant-induced rapid cycling [13]. Interestingly, depression prevalence rates are not observed to be higher in girls prior to mid-puberty/menarche [14,15,16,17,18,19,20], possibly reflecting ascertainment bias/reporting bias (depressed boys may be more likely to come to the attention of health care providers than depressed girls) or the possibility that prepubertal major depression is premonitory of bipolar illness [21] or that alterations in ovarian hormone levels proximate to menarche combine with earlier developmental risks in girls to increase vulnerability [22]. With some exceptions, the age of onset [4, 5, 23,24,25,26] (but also see [27,28,29,30]), type of symptoms, severity, and likelihood of chronicity and recurrence [4, 5, 26, 27, 31,32,33] (but also see [34,35,36,37,38,39,40]) display few consistent differences between men and women. Clinically, the following are more likely in women: present with anxiety, atypical symptoms, or somatic symptoms [7, 26, 27, 37, 39, 41, 42]; report symptoms, particularly in self-ratings [7, 26, 41]; report antecedent stressful events [43, 44]; and display increased comorbidity of anxiety and eating disorders [30, 45, 46], thyroid disease [47, 48], and migraine headaches [49], as well as lower lifetime prevalence of substance abuse and dependence [27, 30, 50]. Some sex differences in treatment response characteristics have also been observed, with women (compared with men) more frequently reporting a poor response to tricyclics [51,52,53,54] particularly in younger women [53] but a superior response to selective serotonin reuptake inhibitors (SSRIs) or monoamine oxidase (MAO) inhibitors. [55,56,57], with the caveats that reports are not consistent, the sex difference in efficacy may reflect different tolerance of side effects, and several meta-analyses fail to replicate the observed sex difference [58, 59]. Additionally, anecdotal reports suggest that women derive a greater benefit in their antidepressant response to tricyclic antidepressants from triiodothyronine (T3) augmentation [48, 60]; however, recent controlled trials with T3 augmentation of SSRIs have found neither beneficial overall therapeutic effects nor sex differences in response characteristics [61,62,63,64]. The extent to which some differences in response reflect sex-related differences in pharmacokinetics [65,66,67,68,69,70,71] remains to be determined. In general, although differences in pharmacokinetics have been observed as a reflection of differences in absorption (gastric acidity and transit time), volume of distribution, and metabolism (sex differences in CYP 1A2, 2D6, and 3A4) [58, 72], these differences do not materially influence the dosing of or response to antidepressants. In fact, the hypnotic zolpidem is the only psychopharmacologic agent for which different dosing is recommended on the basis of sex [73]. Finally, some women will experience significant mood disorders during periods of reproductive hormone change including during the luteal phase of the menstrual cycle, puerperium, and the perimenopause, which have no analogs in men. The contribution of these sex-specific disorders to the increased prevalence of depression in women is unclear.

The sex differences in both epidemiologic and clinical observations are increasingly complemented by demonstrations of sex differences in a wide range of genetic and neurobiological measures relevant to affective disorders in humans. It is to these “etiopathogenic” sex differences that we now turn in an attempt to answer the following question: in the absence of extensive or actionable sex differences in affective disorder, why would one think that sex differences are critical to our understanding of affective regulation and dysregulation.

How would sex alter brain function (the locus of affective regulation)?

The data required to answer this question fall into three main categories: hormonal effects, which include acute or activational effects and programming or organizational effects; effects of genomic sex (i.e., effects independent of hormones but dependent on the presence of X vs. Y chromosomes); and environmental effects (which include effects on an organism consequent to being one sex or the other, “downstream” effects of peripheral sex differences (e.g., differences in metabolism), and sex differences shaped by interactions with the environment (e.g., stressors)). While these effects are discussed separately below, it is increasingly clear that many sex differences represent the composite or integrated effects of genetics, environmental, and sex steroid exposures (see Fig. 1). (Of note, throughout this paper we use “sex” to refer to biological sex as contrasted with “gender,” which refers to sexual identify or social role and may be associated with specific environmental challenges.)

Fig. 1
figure 1

Schematic depiction of the multiple levels at which sex influences brain function. Sex is a ubiquitous, context-creating modulator of brain and behavior, accomplished through both organizational effects that program subsequent brain sensitivities and development, and activational effects that acutely impact neural function. Sex influences the internal environment in which brain function occurs (e.g., differential exposure to stress or immune soluble molecules) as well as modulating the impact of the external environment (e.g., diet or stressors, particularly in the prenatal environment, or even social responses from others based on sex). Sex chromosomes impact brain development directly, may impact physiology through differences in exposure to gene products (e.g., sex-linked genes or differences in gene dosage), and alter brain function developmentally and activationally through sex-determined gonadal function and differential exposure to sex hormones. Sex differences in peripheral organs (e.g., adipose, liver) lead to differential exposure of the brain to hormones as well as medications (through effects on metabolism). The sexome refers to the cumulative array of sex-related modulatory effects on intracellular molecular interactions. Sex differences appear at all levels of neural organization, from cell to circuit. Finally, reported sex differences in meta-cognitions may influence perception and processing of environmental stimuli, thus influencing affective generation and regulation (references appear in the text)

Hormonal effects

Programming/organizational effects

Reproductive hormones quite literally shape brain architecture as well as subsequent sensitivities. The classical studies of Phoenix, Gorski, and Arnold [74,75,76] established that exposure of the brain to reproductive steroids during critical periods of development influenced the development of behavioral capacities (e.g., aggression, sex behaviors) and further “programmed” the brain so as to elicit a different behavioral response in adulthood upon re-exposure to reproductive steroids. Exposure of the perinatal rodent to testosterone increases local exposure to estradiol (E2) (an aromatized metabolite), which is responsible for masculinizing the brain and permitting subsequent “male” behaviors. These programmed capacities were complemented by observations of sex differences in brain morphology and synaptic organization [77, 78].

During critical periods of brain development (i.e., in utero and, in humans, possibly puberty) sex steroids have the capacity to regulate many if not all of the processes (and signaling molecules) involved in the regulation of functional brain development, including neuroplasticity and epigenesis, as well as immune factors (e.g., microglia) relevant to sex-specific brain development. For example, in the Syrian hamster, sex steroids secreted during puberty regulate levels of spinophilin, synaptophysin, glyceraldehyde 3-phosphate dehydrogenase (GAPDH), and synaptic pruning of the medial amygdala and thereby influence social and mating behaviors [79]. In rodents, during embryonic and early postnatal development, estradiol and testosterone permanently masculinize several behaviorally relevant brain structures (e.g., bed nucleus of the stria terminalis (BNST) and medial preoptic area (POA)) through modulation of the enzymes regulating epigenesis (e.g., DNA methyltransferase) [80, 81]. Indeed, masculinization of both the POA and BNST (and its behavioral consequences) can be prevented by administering histone deacetylase (HDAC) inhibitors [81,82,83]. Finally, in neonatal rats E2 regulates microglial number and morphology/activation in the preoptic area, which in turn regulate the process of masculinization and subsequent related behaviors [84]. These programmed differences in brain structure and function in animals potentially manifest at multiple levels of physiologic function from genetic to cellular to circuit level and have provided a major source of inference about the role of gonadal steroids in human brain function and behavior.

Acute/activational effects

There is virtually no element of neural function that is not regulated by reproductive hormones. This is unsurprising for several reasons: (1) reproductive steroid receptors are among the oldest signaling molecules (e.g., the E2 receptor existed for millions of years prior to the appearance of E2 in the steroid metabolic cascade) [85], a fact that helps to explain why, as intracytoplasmic molecules, reproductive steroid receptors act as a point of convergence of multiple intracellular signaling pathways; and (2) without the ability to coordinate reproductive motivated behavior and biology, species do not last long. Teleology aside, from cell to circuit, neural activity is modulated by reproductive steroids.

Neural structure

In addition to programmed modification of neural structure, reproductive steroids can acutely modify neural wiring/connectivity [86, 87]. For example, E2 acts through membrane-initiated signaling involving metabotropic glutamate receptor 1 (mGLUR1) to phosphorylate and deactivate cofilin, an actin severing protein. The resulting change in the cellular cytoskeleton produces acute (within 30 min) appearance of new synaptic spines, which will become permanent if followed by activity-dependent depolarization [86, 88, 89]. As such, E2 can influence both acute and sustained synaptic wiring.

Neural excitability

For at least 30 years, it has been known that reproductive steroids and their metabolites (e.g., the progesterone metabolite allopregnanolone) are able to acutely—within minutes—and more chronically modulate neural excitability [90,91,92]. The mechanisms involved in the acute, membrane-initiated signaling include direct binding of ion channels and of ligand gated ion channels, activation of G-protein-coupled receptors (leading to G-protein regulation of ion channels), and second messenger-mediated modulation of membrane conductance [88, 93, 94]. E2 directly binds and potentiates L-type voltage gated calcium channels [95], acts through estrogen receptor-beta (ER-β) to increase calcium-activated potassium currents (through BK (Big Potassium) potassium channels) to rescue neuronal excitability after O2/glucose deprivation [96], and directly binds and activates Slack potassium channels [97]. In addition to binding ion channels and receptors to alter conductance, E2 can alter genes that determine membrane properties underlying intrinsic excitability. With long-term E2 deprivation, CA1 neurons show decreased intrinsic excitability, less efficient generation of stimulated action potentials and long-term potentiation (LTP), and loss of sensitivity to the acute immediate/early regulatory effects of E2 [98].

Regulation of neural cell function

Reproductive hormones regulate virtually all elements of neuronal (and glial) function, including intracellular signaling, transcription, epigenetic modification of transcription, and (protein) translation. Hormonal regulation of the membrane concentrations of canonical neurotransmitters influences the extent and balance/nature of cell signaling. In turn, through both genomic and nongenomic effects, reproductive steroids like E2 regulate signal transduction through direct effects on Ca2+/calmodulin-dependent protein kinase II (CaMKII; calcium), protein kinase A (PKA; cyclic adenosine monophosphate (cAMP)), extracellular signal-regulated kinase (ERK; mitogens, growth factors), phosphatidylinositol-3-kinase (PI3K)/Akt (growth factors, insulin), and G proteins (cAMP, calcium, PI3, etc) [93, 99,100,101,102]. Not surprisingly, the effect seen depends on the cellular context (e.g., mitogen-activated protein kinase (MAPK) is increased by E2 in neurons and decreased in glia) [103, 104]. E2, through its receptor, regulates gene expression through DNA estrogen response elements (ERE) as well as by tethering to activator protein-1 (AP1), specificity protein-1 (SP1), and nuclear factor (NF)-κB, permitting the regulation of genes without EREs [105]. At the level of transcription, E2 regulates all three RNA polymerases, thus influencing general transcription as well as the production of micro-RNAs, ribosomal RNA, and even transfer RNAs [106]. Indeed, up to one-third of the 500 or so transfer RNAs are robustly and rapidly upregulated by short-term exposure to E2 [106]. Via these means, E2 regulates not only short-term transcription but also potently upregulates the entire protein biosynthetic apparatus. The epigenetic machinery by which environmental events alter transcription is influenced at multiple levels by E2; i.e., regulation of histone acetylation (through effects on both histone acetyltransferase and histone deacetylase) and methylation (through DNMT3b [107, 108]). These effects are at least in part mediated through rapid signaling effects of E2 (e.g., ERK, PI3K), leading, for example, to increased histone acetyltransferase activity and acetylation of H3 (not H4) and target genes, decreased HDAC2 (but not HDAC3) levels, and increased DNMT3b (but not DNMT1) [109]. These effects in hippocampus mediate long-term memory formation.

Regulation of neural transmission

Evidence from animal studies and some human studies document the widespread actions of sex steroids (and their metabolites) on neurotransmission, including effects on the glutamate, GABA, serotonin, dopamine, cholinergic and noradrenergic systems, as well as on the function of several important neuropeptides (e.g., brain-derived neurotrophic factor (BDNF), oxytocin) (see refs. [86, 92, 110,111,112,113,114,115]). The actions of sex steroids impact multiple aspects of neurotransmitter physiology. For example, E2 (and progesterone) regulate multiple aspects of dopaminergic function, affecting the synthesis, release, and metabolism of dopamine [116,117,118] and modulating dopamine receptor expression and function [92, 110, 119, 120]. However, in keeping with the complexities and diversity of dopamine’s neuroregulatory profile, the effects of E2 on dopaminergic function vary in a brain region-specific manner. In rodents, E2 increases stimulated dopamine release in the dorsolateral striatum but inhibits dopaminergic activity in the nucleus accumbens, as well as alters dopamine receptor density and striatal dopamine uptake sites, whereas ovariectomy reduces striatal dopamine receptor density, extracellular dopamine levels, and behaviors mediated by the striatal dopaminergic system [110, 115, 121]. Thus, even within the reward network, E2 can differentially alter the balance between dopaminergic tracts (i.e., striatum vs. accumbens) [122] in female mice. In contrast, it is well established that E2 exerts anti-dopaminergic effects on the anterior pituitary and hypothalamus, where it inhibits dopamine synthesis and prolactin cell responsiveness to dopamine [123]. By modulating dopamine function in the striatum, nucleus accumbens, and prefrontal cortex (PFC), sex steroids may impact both reward and working memory network function. Similar data support the potentially important roles of sex steroids in central glutaminergic and serotonergic system functions, both of which are relevant for affective disorders in women [124,125,126]. Further, the neurosteroid metabolites of progesterone (allopregnanolone) and testosterone (androsterone) are high-affinity, allosteric modulators of the GABA receptor complex at physiologic concentrations [127]. Most important, the widespread nature of sex steroid regulation of central neurotransmission suggests the capacity for these steroids to modulate the cross-talk among different receptor systems in the complex regulation of behavior.

In women, positron-emission tomography (PET) imaging has been employed to evaluate the effects of ovarian steroids on serotonergic and dopaminergic neurotransmitter system function. Many of these findings parallel observations in animal studies and demonstrate in humans the potential for ovarian steroids to regulate these neurotransmitter systems. Studies have employed a range of PET ligands (MAO-A binding (11C harmine), serotonin transporter binding (11C DASB, 11C MADAM), serotonin 2A receptor binding (18F deuteroaltanserin), dopamine receptor binding (11C raclopride)) across several different hormonal conditions in healthy women (i.e., postpartum, across the menstrual cycle, pre- and post-hormonal therapy, or in the context of gonadotropin-releasing hormone (GnRH) agonist-induced ovarian suppression). Overall, findings show an effect of hormonal state on most of these measures (although no menstrual cycle-related differences were observed in 11C raclopride binding [128]). With the exception of two studies using different serotonin transporter ligands, both showing that estrogen decreased binding [129, 130], and one showing an increase in serotonin 2A receptor binding after estradiol therapy [131], the direction of the effects related to either high or low levels of ovarian steroids are inconsistent across studies and could reflect differences in study conditions and the age of the women studied [132,133,134].

Circuit regulation

Reproductive hormone-dependent changes in cell function lead to both structural and functional changes in neural circuitry. Srivastava et al. [99] observed that E2 may rapidly alter neuronal “wiring” by increasing synaptic spines, which become permanent if followed by activity-dependent depolarization [89]. As mentioned above, rapid E2-stimulated membrane-initiated signaling increases immediate/early gene transcription (in a methylation-dependent and, hence, epigenetic fashion), acts through mGLUR1 to induce dendritic remodeling, and alters AMPA receptor trafficking and LTP, thus modifying both neuronal morphology/structure and connectivity [86].

In humans, neuroimaging studies have employed a range of experimental conditions (e.g., across the menstrual cycle, during ovarian steroid hormone manipulation protocols, and before and after ovarian steroid replacement in menopausal women) to examine neural targets of ovarian steroids in women. Both PET and functional magnetic resonance imaging (fMRI) measures have been employed with a variety of paradigms to examine specific neural systems. Neuroregulatory effects of both E2 and progesterone have been documented in working memory (dorsolateral prefrontal cortex (DLPFC), hippocampus) [135,136,137], reward (orbitofrontal cortex (OFC), amygdala, striatum) [138,139,140,141], default mode (medial prefrontal cortex (mPFC), rostral anterior cingulate) [142,143,144] (although also see refs. [145,146,147]), emotional processing (amygdala, OFC, anterior cingulate cortex (ACC)) [148,149,150,151,152,153], and hubs within the salience network (insula) [149, 154, 155]. Thus, ovarian steroids have the potential to modulate many of the functional brain networks underlying alterations in affective state [156,157,158,159,160,161,162].

Sex chromosome effects

In an elegant experiment in mice, De Vries et al. [163] (also see ref. [164]) created a four-core genotype in which Sry—the gene that codes for the factor that results in the development of testes—was transferred to an autosome, thereby permitting assessment of the effects of sex chromosomes independent of gonadal/hormonal effects. They clearly demonstrated the complexity of sex chromosome and sex steroid influences on brain development and that several sex differences in physiology and behavior were determined in a sex chromosome/hormone-independent fashion. For example, in addition to directing the development of the testes and testosterone exposure, the presence of the Sry gene can influence neuronal response to stress well before there is any evidence of testes development or the secretion of testosterone [165]. Indeed, both X and Y chromosomes contain multiple genes relevant to brain development [166,167,168,169]. With the exception of genes shared by X and Y chromosomes, males contain genes only from one X chromosome (from the mother), whereas females have one maternal copy and one paternal copy of each X-related gene, either one of which could convey disease risk or resilience. To maintain homeostasis in overall X-gene expression, the dosage of X-linked genes is balanced by the epigenetic silencing of one copy of the female X chromosome (i.e., X inactivation of parent-of-origin genes—either the maternal or paternal). It is believed that this process results in an equal selection for maternal vs. paternal X silencing, although the origin of the silenced X may vary in a tissue-specific manner and can change with aging [170,171,172,173,174]. Further complexity has been identified in women, in whom a relatively large number of X-linked genes (up to 15%) escape silencing (including those X–Y paired genes thought to share similar functional properties), and consequently gene dosage effects can arise in women whereby a larger proportion of X-related genes are expressed [175,176,177]. The potential relevance of genes escaping X inactivation to disease processes in women has recently been defined in T and B lymphocytes from women who show bi-allelic gene expression (i.e., expression of genes from both X chromosomes) in several immune-related genes that could contribute to the increased prevalence of autoimmune diseases (i.e., Lupus) in women [178]. Thus, the expression of X-chromosome genes in women can be influenced by individual differences in the parent-of-origin genes being expressed (i.e., imprinting), X inactivation (and the escape of X inactivation—including skewing of the proportions of paternal or maternal alleles present within cells [179]), and, possibly, the impact of interactions between the inactivated X chromosome and other autosomes [180,181,182].

Sex chromosome effects in non-brain tissues can also potentially contribute to altered brain development and sex-specific disease risk. For example, in gonadectomized mice in the four-core genotype experimental manipulation, XX mice—regardless of the type of gonad present—had greater food intake (and adiposity) during daylight hours and exhibited a greater risk for a metabolic syndrome-like phenotype [183,184,185]. These effects represented abnormal X gene dosage due to the escape of several X genes from inactivation and could impact both circulating sex steroid levels (due to the presence of steroid synthetic enzymes in adipose tissue) and inflammatory processes, either of which influence brain function. The potential widespread physiologic impact of different sex chromosomes and gene dosage is underscored by observations that genes function in networks, and alterations in the expression of even a single gene may alter the function of that network. It is not surprising, therefore, that sex differences in transcriptional profiles vary by tissue and are as high as 72% in liver, 68% in adipose tissue, 55% in muscle, and 14% in brain [186]. Finally, mitochondrial genes are only transmitted from the mother, and, therefore, natural selection can act only through women to optimize mitochondrial gene function [187]. The corollary of this process is that the function of mitochondrial genes may not be optimized in men and, therefore, potentially conveys disease risk in men. Moreover, studies suggest that some mitochondrial DNA contains functional EREs, and since ER-β can be imported into mitochondria, it is possible that sex steroids can also regulate mitochondrial gene function [188, 189].

Environmental effects

It is axiomatic that the environment shapes the brain, and there are a multitude of ways in which sex interacts with the environment in this process. First, women are subject to very different environmental responses than men, with many of the resultant effects consequent for brain development. Sex-related traumatic sequelae, for example, are enduring and profound, altering (at the least) cognitive, behavioral, and physiological response to subsequent stressors. Second, environmental stimuli (e.g., stressors) may be processed in a sex-dependent fashion (see below). Third, sex-related differences in peripheral organs (e.g., hepatic function) can expose the brain to a different “environment” [190]. Fourth, the maternal–placental–fetal unit contributes to fetal development in several ways, including populating serotonin neuronal concentrations in the mouse and human fetal brain [191]. Finally, sex may select a different environment that affects subsequent risk, as seen by Markle et al. [192] in the study of the role of the microbiome in risk for diabetes. These investigators were able to eliminate the overwhelming sex difference in the risk of diabetes in a strain of mice (non-obese diabetic (NOD))—four times higher in females—by transplantation of gut contents from an adult male to the young female. This remarkable effect was dependent on a sustained increase in testosterone induced in the females by the transplanted microbiota.

Are there sex differences in the brain?

Structure

Despite the presence of several well-documented sex differences in brain morphology, innervation, and regional composition in animals [193], many of which may not be observed within the same order (e.g., rats vs. mice [194]), only a few robust sex differences in humans have been demonstrated consistently. First, women with Turner syndrome (with only a single X chromosome) have smaller MRI-measured volumes of several brain regions, including the hippocampus, caudate, and parieto-occipital cortex, suggesting the role of X-chromosome imprinting or X-chromosome dosage on gray matter morphology [195]. Sex differences in gray matter (GM) volume have also been observed in normal men and women, with women observed to have a higher percentage of GM volume (relative to white matter) and cerebrospinal fluid, despite controlling for total intracranial volume [196, 197], and greater volumes of the orbital frontal cortices, but not hippocampus, amygdala, or DLPFC [198] (although see ref. [199] in which increased hippocampal volume is observed in girls compared with boys after puberty). In contrast, a recent meta-analysis reported that men had higher gray matter densities than women in several brain regions, including amygdala, hippocampus, insular cortex, and putamen [200]. Indeed, a postmortem study found that men have higher cortical synaptic density in all layers of the temporal neocortex [201]. Sex differences in the effects of aging on cortical volumes [202] and inter- and intra-hemispheric connectivity (as described below) have also been reported [203, 204]. Many of these sex differences emerge during adolescence, and some studies suggest a critical regulatory role of androgens and the androgen receptor in the developmental trajectories of these differences [205,206,207,208,209,210]. Thus, both the X-chromosome and sex steroid exposures across the lifespan potentially can alter structural volumes in men and women. A more recent analysis of 1400 adults in which male and female patterns of GM, white matter, and connectivity were defined a priori on the basis of a separate cohort of scans demonstrated a substantial overlap in sex-specific patterns in all brain regions examined. Although the analytic methods employed are controversial and the cohort was assembled from several studies employing differing methodologies (see refs. [211,212,213,214,215,216]), these data suggest that in humans individual differences contribute more variance to brain structure and connectivity than does sex [217]. Several caveats deserve mention. It is likely that the impact of structural sex differences on affective regulation is brain region specific, and, therefore, overall gray and white matter volumes might be less relevant. Moreover, the methods employed to measure regional volumes vary considerably across studies (see above). Thus, in contrast to the animal literature, inconsistencies in structural brain assessments at the human level probably reflect a range of methodologic complexities and signal the need for additional research to resolve these issues and identify meaningful and generalizable findings.

Cell activity/signaling differences

The functional end of sex effects in the brain is the cell (neuron, glia). Multiple sex differences in cellular signaling have been observed and attributed to both sex-related programming and acute hormonal effects. For example, Huang and Woolley [218] showed that although E2 acutely potentiated excitatory postsynaptic potentials equally in males and females, E2-suppressed, rapid, perisomatic synaptic inhibitory transmission occurred only in females. This E2 suppression of inhibition (in oophorectomized females) was mediated through ER-α, mGluR1, and endocannabinoid signaling, and Tabatadze et al. [219] observed sex differences in both E2-dependent and E2-independent regulation of the hippocampal endocannabinoid system. They demonstrated that ER-α/mGluR1/IP3R signaling is regulated by E2 only in females—despite comparable levels of ER-α, mGluR1, IP3R—and E2-independent hippocampal tonic endocannabinoid release exists in female (but not male) rats. These findings of presumed programmed sensitivity parallel those of Boulware et al. [220] and Meitzen et al. [221] showing that rapid, E2-stimulated CREB phosphorylation, which occurs only in females, can be eliminated by postnatal exposure to E2 or testosterone (and hence, again, is an organized sensitivity). Although the mechanisms underlying these post-receptor differences in ER-α signaling are unknown, Oberlander and Woolley have identified a striking sex difference in the role of ERs in E2 regulation of glutamate signaling in the hippocampus: glutamatergic presynaptic signaling is mediated by ER-α in males and ER-β in females, while postsynaptic signaling is mediated by ER-β in males and GPR30 in females [222]. Another organized sex difference (i.e., unaffected by varying hormonal levels) in signaling is that of corticotropin-releasing factor (CRF)-mediated signaling in the locus coeruleus [223]. There, the differential coupling of CRF receptor 1 to β-arrestin-2 (more in males) favors Gs-mediated pathways (PKA, ERK) in females vs. Gs-independent pathways (Rho, Src, ERK) in males [224]. This has striking implications for the processing of stress-related signals as described below. In addition to organized sex-dependent sensitivities, many sex differences in neural signaling reflect differential exposure to acute, regulatory effects of reproductive steroids. These include E2-stimulated BDNF transcription (with corresponding increased expression in females compared with males during the high E2 phase of the estrus cycle), phosphorylated (activated) axonal Trk-B (the receptor for BDNF), and associated synaptic plasticity [225]. Particularly noteworthy is the demonstration that many observed sex differences in gene expression (and that link to sex-typical behaviors) depend on adult exposure to testosterone in males but are independent of adult ovarian hormones in females [193].

Network differences

The existence of sex- and hormone-related effects on brain region structural connectivity and activation provide a basis for inferring differential network processing between sexes. Multiple sex differences in the rat basolateral amygdala were recently described, including increased neuronal firing rates, more dendritic spines, and greater sensitivity/responsivity to glutamate (via iontophoresis) in females [226]. Further, both organized and activated sex differences in dendritic structure in the locus coeruleus and hippocampus provide examples of “wiring” differences and synaptic remodeling that are associated with differential circuit function [112, 224]. Suggestions of sex-related differential cortical dynamics in humans are primarily derived from three sources: neurocognitive studies showing better female performance on memory and social cognition tasks vs. better male performance on spatial and motor speed tasks (but see refs. [227,228,229]; electroencephalography (EEG) studies demonstrating greater laterality in male brains [230]; and diffusion MRI tractography studies [204] (but also see ref. [231]). MRI studies with graph theoretical analysis of anatomical connectivity suggest that, despite overlap, women show significantly higher overall connectivity as well as increased local and global efficiency (after controlling for brain size) [232]. Ingalhalikar et al. [204] demonstrated prominent sex differences in cortical connectivity patterns, whereby men showed greater within-hemispheric connectivity (favoring coordination of perception and action), whereas female brains have greater between-hemispheric connectivity, facilitating “communication between analytical and intuitive processing modes.” Sex differences in connectivity patterns (both functional and structural) were also observed by Tunc et al. [233] and Satterthwaite et al. [234] (but see ref. [235]). Nonetheless, in a large MRI study, Joel et al. [217] concluded that although sex differences in brain and behavior are observed, the overlap in all brain regions between men and women is extensive, and the internal consistency within even a single brain is far less common than variability; i.e., individuals are mosaics, with volume or connectivity of brain regions (and behaviors) spanning the male/female continuum in a region (and behavior)-specific fashion.

Are there sex differences in the substrates of affect regulation?

Many physiologic systems and processes have been implicated as contributing to the etiopathogenesis of depression. These include neurotransmission, neuroplasticity, stress axis, immune function, and neural and genetic network regulation. As described below, sex-related differences as well as reproductive steroid-related modulation have been observed in each of these processes. (The existence of reproductive steroid regulatory effects provides a basis for inferring sex-dependent differential function of these systems, but, as noted above, differences at the molecular level may converge to result in similar physiologic endpoints.) Several examples follow, with greater attention paid to dysregulation of immune function, stress reactivity, and neural network function.

Neurotransmission

As described above, the effects of reproductive steroids on neurotransmission are profound, affecting all levels of function including neurotransmitter synthesis and metabolism, receptor synthesis and trafficking, and signal transduction. Not surprisingly, therefore, there are many reported sex differences, many of which are brain region specific, in the concentrations, receptors, and metabolites of classical neurotransmitters as well as in their elicited intracellular signals and cellular actions. These are reviewed elsewhere (see refs. [86, 92, 110, 111, 114, 115]).

Neuroplasticity

Alterations in neuroplasticity, including neurogenesis, cell death, and synaptic remodeling, are some of the fundamental processes that underlie the development of sex differences in the brain [194]. Sex steroids differentially regulate many of the molecules involved in neuroplasticity (e.g., neurotrophin secretion), as well as several of the physiologic systems (e.g., glutamatergic and GABA-ergic) [112, 236] regulating the opening and closing of critical developmental windows [237]. Thus, exposures to sex steroids could regulate sex-specific “opportunities” for the effects of physiologic events (e.g., puberty) or adversity to differentially impact developmental (re)programming and the instantiation of behavioral risk over the life cycle. Sex differences in the mechanisms of neuroplasticity have been reported, particularly in the hippocampus, a brain region implicated in affective dysregulation [238]. In knock-out mice, calcium/calmodulin kinase kinases appear to play a more significant role in hippocampal neuroplasticity in males compared to females [239], whereas females but not males show synapse induction within the hippocampus by E2 (although testosterone will induce synapse formation in males (in vivo) [240]). Similarly, the pattern of hippocampal synaptic remodeling of CA3 dendrites after chronic stress exposure is observed in males but not females [241] (reviewed in ref. [113]). The relevance of these differences in hippocampal neuroplasticity to sex differences in human brain function (or disease) remains to be clarified. Additionally, several sex differences in neuroplasticity after administration of ketamine (an N-methyl-d-aspartate (NMDA) receptor antagonist used for treating depression [242]) have been reported in socially isolated rodents [243]. For example, ketamine reversed the decreased spine density in the medial prefrontal cortex in males but not females [244]. These differences also likely reflect differences in the impact of social isolation on male and female mice [245, 246]. Recently, female mice were observed to have threefold increased levels of hydroxynorketamine, the ketamine metabolite with rapid antidepressant-like actions, in the absence of NMDA receptor inhibition [247]. Thus, although a recent report from human studies did not identify sex differences in the antidepressant effects of ketamine in treatment-resistant depression [248], it is possible that sex differences in the sensitivity to ketamine could emerge in larger samples.

There is considerable evidence that reproductive steroids modulate neuroplastic processes implicated in depression and/or the antidepressant response. E2, for example, does the following: acts like antidepressants (and opposite to stress) in stimulating BDNF [249]; increases activity of the transcription factor cAMP response element-binding protein (CREB) [250] and trkA (neurotrophic tyrosine kinase receptor type 1) [251, 252]; and decreases glycogen synthase kinase-3β in rat brain [253], the same direction of effects as seen with mood stabilizers. Interestingly, sex differences are observed in the facilitative effects of E2 on neuroplasticity. Although the effect on synaptic activity (i.e., potentiation of glutamatergic synapses in CA1 hippocampal slices) is similar, the mechanism involves ER-β (and possibly GPER1) in females vs. ER-α in males [222].

Neuroprotective effects of E2 and progesterone (or its neurosteroid metabolites) have also been described in neurons grown in serum-free media or those exposed to glutamate, amyloid-β, hydrogen peroxide, ischemia, or glucose deprivation [254,255,256,257,258] (see also ref. [259]). Some of these effects appear to lack stereospecificity (i.e., are not classical steroid-receptor mediated) and may be attributable to the antioxidant properties of E2 or the GABAA-modulating effects of allopregnanolone, although more recent data support steroid receptor-mediated mechanisms of action. Gonadal steroids may also modulate cell survival through effects on cell survival proteins (e.g., Bcl-2, BAX), signaling pathways (e.g., MAPK, Akt), intracellular calcium regulation, metabolism of amyloid precursor protein and Aβ, or through enhancing mitochondrial respiratory efficiency [259,260,261,262]. Through increased oxidative capacity and efficiency of neuronal mitochondria, E2 promotes neuronal bioenergetics and protects neurons against multiple toxins, including free radicals, excitotoxins, Aβ, and ischemia [260]. Damage from oxidative stress to mitochondria promotes apoptosis and cell death, and both estrogen receptor-dependent and -independent neuroprotection at the level of the mitochondria have been described. For example, E2 bound to ER-β can be transported into the mitochondria, where it binds an ERE in mitochondrial DNA [188, 189, 263] and produces a range of anti-apoptotic proteins that maintain the integrity of the mitochondrial membranes. Thus, in addition to its other neurotrophic actions [263], E2 could improve mitochondrial respiratory efficiency by directly inducing transcriptional activity in mitochondrial DNA and prevent the oxygen free radicals that are believed to adversely affect mitochondrial energetics in depression.

Neural and gene network function

Gene networks

A study of gene co-expression networks showed marked differences between patients with MDD and controls, but strikingly there was little overlap in the altered transcriptional network connectivity patterns in males and females with depression or in mice subjected to chronic variable stress [264]. Of note in this study, a hub gene—Dusp6—was downregulated in a sex-specific fashion in both depressed women and stressed mice, regulated cell signaling and ventromedial PFC pyramidal cell excitation (only in females), and, when downregulated, increased behavioral sensitivity to stressors; i.e., identical manipulations of the same gene led to major differences in both the physiological and behavioral effects in males and females. The sex-specific transcriptional signatures in depression were recently replicated by Seney et al. [265], who observed that of the 700 to almost 900 genes differentially expressed in cortico-limbic regions in men and women with MDD compared with controls, only 73 genes were differentially expressed in both men and women, and 52 of these changed in the opposite direction. Additionally, postmortem studies of men and women with MDD have reported a sex difference in the expression within the dorsolateral prefrontal cortex of multiple glutamatergic genes, with increased expression in women [266]. Finally, several clock genes have been reported to differ in a sex-specific manner, possibly in keeping with reports that women have an overall phase advance in several measures of circadian rhythmicity compared with men [267,268,269].

Neural networks

Disturbances of amygdalar activity or connectivity have been implicated in affective disorders [270], and sex differences have been described in amygdala activation patterns [271] and connectivity [272]. Of note, the amygdala is differentially activated in men and women as a function of the valence or nature of the affective stimulus, positive or sexual in men and negative in women [273].

Sex differences have been observed in resting state functional connectivity, with several studies finding increased connectivity within statistically defined modules or sensorimotor resting networks (i.e., increased intra-network/module connectivity) in women and increased inter-network or cross-module connectivity in men [234, 274], with the caveat that studies also have reported increased intra- and cross-modular connectivity in girls compared to boys (Reding et al., unpublished observations) as well as other sex-specific patterns of connectivity in boys and girls [275,276,277]. As emphasized by Mak et al. [277], the divergence in some of these functional connectivity findings reflect differing analytic methods (e.g., independent component analyses vs. seed-based vs. data-driven analyses), potential presence of negative affective symptoms on the day of scanning, and differing ages of participants (e.g., prepuberty vs. adults). Nonetheless, abnormalities in resting state functional connectivity within the default mode network (DMN) have been observed in depressed men and women [157, 158, 278]. However, few if any studies report sex differences in DMN functional connectivity in depression. Recent studies do suggest an association between depressive rumination and functional connectivity between the DMN and the subgenual prefrontal cortex (sgPFC) [157]. Since studies also suggest that women experience a greater amount of rumination during depression than men [279], one could imagine that a sex difference specifically in DMN functional connectivity with the sgPFC may be detected. Alternatively, if confirmed, the presence of sex differences in DMN network activity in asymptomatic men and women, and the absence of comparable differences in depression, could inform our understanding of depressive illness.

A sex difference has been reported in an emotional/arousal network, with men showing greater activation (fMRI blood-oxygen-level dependent (BOLD)) of the anterior cingulate gyrus, OFC, hippocampus, and mPFC when watching negatively valenced/high arousal pictures compared with women during mid-cycle of the menstrual cycle [271]. Using a similar paradigm, these investigators previously also demonstrated that activation patterns in many of these brain regions in women decreased during mid-cycle compared with the early follicular phase, suggesting that higher estradiol levels at mid-cycle attenuated the activity of this network [148]. Thus, sex differences in network level function may also reflect the regulatory effects of sex steroids. This observation is lent further support by the recent discovery of an estradiol-regulated reward circuit in mice [280], again suggesting the contribution of activational effects of sex steroids to differential network function in depression.

Immune function

Multiple lines of evidence suggest that immune dysfunction contributes to the risk for depression. Raison and Miller [281] have proposed the pathogen:host defense model, which suggests that immune activation and stress perception have “co-evolved” to generate sickness behavior (protective in intent) in response to environmental threats and challenges. Additionally, multiple studies identify elevated immune activation markers and cytokine levels (e.g., C-reactive protein, interleukin (IL)-6, IL-1B, and tumor necrosis factor-α (TNF-α)) in patients with major depression compared with controls [282, 283]. Multiple sex differences have been described for immune function, stress response (see below), and the interaction of stress and immune function in depression.

Sex differences in immune function or effects include the following. (1) In response to immune activation, females experience more adverse behavioral effects, including increased immobility on the forced swim test and decreased sucrose preference in rodents [284,285,286] and increased depression and social disconnection in women [287, 288]. (2) In the immune response itself, in humans some immune cells increased equally in males and females after stress, some to a greater extent in males, some more in females regardless of menstrual cycle phase, and some only in females on oral contraceptives, findings again consistent with both programming and activational sources of sex differences [289]. Further, although lipopolysaccharide-stimulated microglial IL-1B expression in vitro is increased in neonatal males compared with females, it is suppressed by E2 in males and increased by E2 in females, suggesting that not only is the sex difference hormonally responsive, but the effect of the same hormone is opposite in males and females [290]. (3) Sex differences exist in the susceptibility to neuroimmune-related disorders. Autoimmune disorders (including multiple sclerosis (MS), lupus, rheumatoid arthritis, and thyroid disease) show a dramatically increased prevalence (2–9-fold) in women [291], consistent with the increased susceptibility to experimental autoimmune encephalitis (EAE), an animal model of MS, in female rodents [292]. Notably, in the EAE model, the same trigger (myelin basic protein) leads to decreased lymph node immune cells, decreased reactive cells, and decreased cytokines, as well as increased spleen-derived “anti-inflammatory” cytokines in males [292, 293]. The same stimulus, then, produces both different immune responses and different delivery of immune signals to the brain. Despite the female predominance of EAE, E2 improves EAE severity in both males and females [294]. Indeed, while multiple studies support the immunomodulatory effect of E2, suggesting its suppression of the microglial “inflammatory” state and dose-dependent suppression of the synthesis of a range of cytokines (TNF-α, IL-1B, MCP2) [295], the role of E2 in the regulation of immune response is complex and likely both cell- and context-dependent [294]. Sex steroids are key regulators of immune cell phenotype and function, with demonstrated roles in the regulation of both adaptive and innate immunity. Androgens and estrogens have been shown to regulate immune cell proliferation, differentiation, and apoptosis, as well as cytokine and immunoglobulin production [296,297,298]. Of note, sex steroids are synthesized de novo in the brain (neurons, glia), and hence their immunomodulatory effects can occur locally, representing paracrine and autocrine rather than classically endocrine effects (reviewed in ref. [299].

Stress axis

The importance of antecedent stress, stress sensitivity, and stress axis dysregulation in affective illness is, at this point, axiomatic. Stress in relation to affective disorders can be viewed from three distinct perspectives—stress as an environmental stimulus, the stress axis as an outcome measure, and the stress axis as a mediator of change in physiology (e.g., neural network or immune function) or behavior. Sex differences in stress are reviewed elsewhere [300,301,302], but several deserve mention in relation to affective disorder. First, as noted above, sex may elicit different responses from the environment (i.e., women are subject to different, potentially stressful interactions because of their sex). Second, women are twice as likely to experience stress-related disorders (MDD, anxiety, PTSD, obesity (also eating disorders and most phobias)) and more likely to develop them after trauma, physical abuse, or maternal distress during infancy [303,304,305,306,307,308]. Prenatal stress results in earlier behavioral problems in boys, with girls showing stronger effects (and increased amygdalar volumes) later [309, 310]. Third, the hypothalamic–pituitary–adrenal (HPA) axis may respond differently as a function of sex and is regulated by reproductive steroids. Sex differences in measures of HPA axis activity take the form of differences in magnitude, effects in one sex but not the other, and opposite effects. For example, men have greater stimulation of adrenocorticotropic hormone (ACTH) and cortisol compared with women after the Trier Social Stress Test (TSST) [311], and both CRH- and exercise-stimulated ACTH and cortisol are greater in men than women even in the absence of differences in reproductive steroids (i.e., under GnRH-induced hypogonadism [312], although also see ref. [302]). Early trauma is positively associated with basal ACTH in women and negatively associated in men, while severe trauma is associated positively with ACTH response to CRH in men, but not women [313]. Of note, this literature is filled with ostensible inconsistencies, no doubt a function of the nature, duration, and timing (developmentally) of the stressors as well as the measures obtained.

Not surprisingly, sex differences have been reported in many components of the stress response, including differences in elicited changes in brain structure, non-HPA physiologic response (e.g., immune response noted above), and behavior. Chronic stress causes atrophy of dendrites in medial prefrontal cortex and hippocampus (CA3 pyramidal cells) only in male rodents [314, 315], whereas projections from medial prefrontal cortex to amygdala show (estrogen-dependent) increased spine density and dendritic expansion after chronic stress only in females [316]. These data are complemented by behavioral studies demonstrating that stressors like tail pinch increase associative learning (e.g., classically conditioned eyeblink response) in male rodents and produce learning deficits in females [317, 318]. Further, stress effects on learning are mediated by different brain regions and circuits in males and females (i.e., the mPFC, particularly in its connection to the amygdala, is critical to disrupted learning in females but not to enhanced learning in males, which involves activation of the BNST (not seen in females)) [319]. It should be noted that the sex differences in learning after acute stress are reversed for chronic stress, in which memory deficits are observed in males, but not females [236, 320]. As described above, the same repeated stress paradigm producing similar deficits in genetically identical male and female mice yields amazingly distinct, sex-dependent profiles of differentially expressed genes (about 20% overlap) compared with unstressed control animals [264]. Further, subchronic variable stress induces multiple depression-associated behavioral deficits (e.g., decreased sucrose preference, latency to eat in a novel environment) in female but not male mice, with, again, sex-dependent difference in the transcriptional response to stress in the nucleus accumbens [321]. A variety of findings converge in suggesting that women differentially process stressful stimuli: negative, arousing stimuli evoke faster and greater EEG responses in women [322], and emotion-evoking tasks produce greater activation of the locus coeruleus in women than men [323]. In parallel, the dendritic structure and post CRF receptor signaling in the locus coeruleus in rodents favor an exaggerated response to stress in females [324].

Effects of gonadal steroids

Many of the myriad sex differences in stress processing are attributable to differential exposure to reproductive steroids, which, with their metabolites, play a major role in modulation of the stress response. The CRF receptor has an ERE [325], and many components of the HPA axis vary across the estrus or menstrual cycle. Basal and stimulated ACTH and cortisol levels are higher in female rodents and during proestrus (when E2 and progesterone levels are high) [326,327,328]. These sex differences are eliminated by ovariectomy, as are many of the sex differences in brain region-specific structural and functional effects of stress noted above (e.g., the expansion of dendrites and spines in the mPFC–amygdala projections in females [316] or the sex-divergent effects of acute stress on classical eyeblink conditioning [319]). Of note, metabolites of reproductive steroids also regulate the response to stress. Thus, metabolites of progesterone (allopregnanolone) and testosterone (dihydrotestosterone) both dampen the response to stress in rodents, the former through acute effects on the GABA receptor and the latter through ER-β [329, 330].

How might sex contribute to differential capacity for affective regulation?

Affect regulation is dynamic. The ability to change and regulate affective state (whether externally or internally facilitated) appears at an early age, and while this capability may be “programmed” by genomic or early environmental events (contributing to “disposition”), it is also modified as a function of experience and environment throughout life. This view is consistent with the following: (1) the known “cumulative” effects of stressors (leading to different genomic expression patterns in response to subsequent stress [331] and greater susceptibility to affective disturbance); (2) other environmental (e.g., hormonal or experiences with mastery and control) and genomic contributions to susceptibility and resilience; and (3) the plasticity of affective regulation that allows for both increased susceptibility to affective disturbance and development of capacities/strategies for decreasing affective dysregulation.

Sex differences in affective disorder—particularly the presence of sex-specific, reproductive-related mood disorders in women (postpartum depression (PPD), premenstrual dysphoric disorder (PMDD), and perimenopausal depression)—provide unparalleled insights into this dynamic process of affective regulation. Employing hormone manipulation paradigms involving blinded reproductive steroid administration, often in the context of GnRH-induced ovarian suppression, we have been able to demonstrate a clear role for ovarian steroids in both PMDD and PPD [332, 333]. Thus, despite normal reproductive endocrine function, the depressed state in these disorders is triggered by levels or changes in ovarian steroids that are without impact on affect in women lacking a history of PMDD or PPD. These affective disorders, then, represent the combination of a regulatory role for ovarian steroids on affective state with a susceptibility that permits normal reproductive endocrine signals to precipitate a dysphoric affective state. Put differently, women with these affective disorders are differentially susceptible to reproductive steroids such that a normal signal produces an abnormal behavioral state. How might this process—as well as its absence in men—be understood?

Cortical/genetic networks in males and females function differently or are differentially responsive to stimuli like stress that load on development of affective dysregulation

Differential processing of affective-relevant stimuli

Consistent with the animal studies noted above, the same stimulus may be processed differently in men and women. Supporting examples (also described above) include the following: (1) Sex differences have been described in cortical networks relevant for affect; and (2) stress activates the brain differently in males and females. Stress leads to activation changes in the PFC (increased) and in the OFC (decreased) of males, whereas females show activation of limbic structures [334]. Further, different neural strategies are employed to cognitively control emotions (e.g., cognitive control decreases amygdalar activity in males but increases activity in the ventral striatum, ACC, and frontal regions in females) [335, 336]. Thus, networks may be differentially elicited or differentially deployed in response to affective stimuli.

Women process affective-related/social stimuli differently

Women experience emotional stimuli as more arousing than men and experience increased free recall, particularly for personal life events as well as increased memory for emotional events [337]. Affective valence category-specific sex differences in emotional appraisal have also been described [338]. Women have greater sensitivity to others, increased self-awareness, and increased capacity to manage new situations [339]. Women ruminate more than men [279], which may load on susceptibility to depression. As with the “orchid/dandelion dichotomy” [340, 341], whereby the environmentally sensitive child will do worse under environmentally stressful and better under nurturing conditions than the environmentally insensitive child, so the “increased sensitivity” of women to socio-affective stimuli may allow for greater emotional flexibility (under good conditions) but greater stress-related adverse consequences under unfavorable conditions. Sex differences in social and other risk factors for experiencing depression have been described by Kendler and Gardner [342]: failures in interpersonal relationships (marital dissatisfaction, interpersonal loss, loss of social support) play a stronger etiologic role in depression for women, while stressful life events in the past year, history of childhood sexual abuse, failure to achieve expected goals, and lowered self-worth are more prevalent in male depression.

How might these different stressors acquire differing impact on susceptibility to affective dysregulation in men and women? Independent of cultural expectations that impact the emotional salience of events, reproductive steroids influence network development and regulation such that environmental events are differentially processed. This is true at an organizational level as well, where E2 or testosterone will shape both circuit development AND subsequent sensitivity to (hormonal) stimuli [74,75,76]. Differential processing can be seen at the level of both neural and gene networks. As described above, Ingalhalikar et al. [204] demonstrated that the connectivity patterns in the brains of women may result in different processing of information, favoring better integration of between-hemisphere, analytical, and intuitive modes. Further (also described above), several studies demonstrated markedly different transcription profiles in men and women in association with major depression [264, 265]. These observations are consistent with those of Gaiteri et al. [343] that E2 modulates the synchrony of the gene interaction networks that are most disturbed in depression, suggesting a means by which pathological neural activity can be transformed (through E2 regulation of transcription and synaptic formation) into enduring cellular changes over time and across brain regions. So, the same genes are differentially organized and expressed within different cells and brain regions, with the differential expression under stressful circumstances potentially compromising cross-regional coordinated activity and adaptive behavioral strategies.

Changes in reproductive hormones may regulate affective state change

Several lines of evidence support this idea. (1) Teleologically, the reproductive hormonal regulation of sexual receptivity/motivation facilitates reproductive efficiency (sex during periods of optimal fertility). As such, behavioral sensitivity to hormone levels or changes is “hard-wired.” (2) Changes in hormones have clear regulatory function. For example, the rate of change of blood cortisol concentration exerts a rapid (time-delayed) (5–30 min) feedback action on pituitary ACTH release [344], and fluctuating levels of the neurosteroid allopregnanolone induce altered function of GABA receptors through subunit-dependent conformational changes in the receptor [345]. (3) Changes in reproductive steroids modify the excitatory/inhibitory (E:I) balance that regulates behavioral state transitions and those between neural networks [346]. For example, postpartum E2 withdrawal impairs GABA-ergic inhibition and long-term depression in the basolateral amygdala (BLA) via downregulation of the GPR30 estrogen receptor (also called GPER) [347], and GPR30 activation decreases stress-induced anxiety by maintaining E:I balance in the BLA [348]. Similarly, allopregnanolone increases GABA-ergic interneuronal activity, and low progesterone concentrations are associated with increased amygdala activity in women [349]. Deficient inhibitory interneuronal function leads to a “tuning deficit” with decreased inhibitory filtering, increased “noise,” disturbed behavioral state transitions, and both persistence of and failure to suppress the hyper-excitable DMN when the cognitive control network is activated [346]. (4) Changes in reproductive hormones in PMDD precipitate affective state change. Blind administration of either E2 or progesterone to women in whom PMDD-related affective disturbance had been eliminated by ovarian suppression resulted in precipitation of depression [332]. As it was unclear whether the precipitating event was a change (increase) in the hormone or the exceeding of a threshold level of hormone, a study was designed to administer reproductive steroids continuously for 3 months in association with ovarian suppression. Following the initial precipitation of depression in concert with hormone exposure, the affective state resolved and failed to reappear during the remaining 3 months of the stabilized hormone administration, thus demonstrating that the change in hormone (rather than the level achieved) was the inciting stimulus [350]. Similarly, prevention of the luteal phase-related increase in the progesterone metabolite allopregnanolone by the administration of dutasteride, a blocker of allopregnanolone synthesis, prevented the switch into the PMDD-related dysphoric state [351], thus mirroring the stress-like behavioral state observed following changes in allopregnanolone in rodents [345].

Substrates of differential sensitivity

The literature is replete with models (kindling, pharmacologic sensitization, time-dependent sensitization, learned helplessness) demonstrating that behavior and its underlying physiological substrates are highly context dependent. Genes and environment interact continuously, dynamically, and in a way that can enduringly change subsequent response to the same stimulus [352, 353]. Sex, both endocrine and genetic, creates a context that shapes the nervous system and helps program subsequent responses (despite the considerable overlap between sexes that exists for most brain structures and functions). This capacity to create differential sensitivities is modeled in the interaction of genes and environment that results in susceptibility or resilience to stressful stimuli. Brain region-specific manipulation of the transcription of genes implicated in affective disorder and displaying sex-dependent regulation by stress (e.g., DNMT3a, Dusp-6) alters the sensitivity to stress and susceptibility to stress-induced, depression-like behaviors (increased with over-expression of the former and downregulation of the latter) [264, 321]. Even within members of a genetically similar inbred strain of mice, individual differences in stress sensitivity can be identified and, more remarkably, successfully conveyed to other mice with bone marrow transplantation [354]. These observations suggest that differences in behavioral sensitivity can occur through transcriptional responses to the environment that are enduring. Certainly, as noted above, some sex differences in sensitivity appear early and are sex chromosome (rather than sex hormone) dependent. In vitro studies of mouse embryos prior to sex differentiation (and hence sex hormone exposure) show that chromosomal sex drives different cellular responses to stressors and different transcriptional responses to exogenously administered sex hormones [355]. It is intuitively compelling, however, that differences in affective sensitivity reflect an ongoing and sculpting dialog between environment and genome, even if evidence of differential sensitivity exists at a cellular level. In support of this hypothesis is the observation by Dubey et al. [356] that lymphoblastoid cell lines from women with a reproductive endocrine-related mood disorder, PMDD, show, compared with controls, increased expression of a family of epigenetic modifying enzymes (ESC/E(Z)) as well as differential response to exogenous application of E2 and progesterone.

Conclusions and future directions

We still know regrettably little about how affect is regulated. That said, for heuristic purposes, affective regulation can be parsed into the following: neural basis for affective experience (circuit formation/activation and synchronized firing), for changing circuit function (switching between states; cortical oscillation/dynamics), for dysregulation of circuit function (dysfunctional states and disturbed kinetics of state changes), and for susceptibility to sustained dysregulation of circuit function (genomic/transcriptional capacities for modulating interactions with the environment). On one hand, sex differences can be observed in these processes at almost any level of investigation, thus underscoring two critical points: (1) the failure to study both sexes will give us a false sense of understanding and deprive us of physiological insights; and (2) all physiology is context dependent, and sex, like developmental stage, age, past history, and genetic background, is a context (and a particularly powerful one as we have described). On the other hand, we simply do not understand the meaning and relevance of many of the sex differences that we have detailed above or that have been described elsewhere. Despite the multitude of studies documenting sex differences in the brains of lower animals, with the exception of the hypothalamus (in which cyclic pulsatile gonadotropin secretion is present in women but not men) most sex differences in the human brain have relatively modest effect sizes, with considerable overlap between sexes.

Studies of sex differences in human affective regulation are largely characterized by their inconsistencies, which might lead one to conclude that these differences are neither meaningful nor actionable. It is important, however, to bear in mind that human studies, particularly those of the brain, entail multiple methodologic complexities. (1) It is immensely more difficult to control for potential confounds—particularly in studies of brain and behavior—in humans than animals. (2) The expense of human studies often precludes the recruiting of sample sizes sufficiently powered to detect interaction effects between sex and the specific outcome measures. (3) Intra-sex variability often exceeds between-sex variability (which does not mean that relevant sex differences are absent or inconsequential). (4) The methodologic factors responsible for many of the current inconsistencies are legion and include diagnostic methods, sample age, statistical methods, reproductive state (prepubertal, premenstrual, postmenopausal), and menstrual cycle phase. This is particularly true when examining imaging studies, in which the following variables are seen: brain region (region of interest), volumetric measurement strategy, imaging method (PET vs. structural MRI vs. functional MRI), nature of imaging analysis (e.g., activation vs. connectivity), and nature of stimulus (e.g., cognitive vs. affective), to name just a few. Systematic efforts to employ common, validated methods and rigorous study designs in adequately powered studies will greatly improve our ability to detect and interpret meaningful sex differences in affective regulation.

Although a picture is emerging in which affective disorders may represent a “convergent” process (i.e., different physiologic routes to the same behavioral phenotype), we nonetheless must avoid the temptation to prematurely ascribe etiopathogenic meaning to cross-sectional snapshots of observed differences. Rather, we should attempt to develop common methodologies for examining the substrates of affective regulation, leverage big data, and thereby enable sex differences to help illuminate the neural antecedents of behavior, detect novel sources of variance, and pressure test our assumptions about pathophysiology.