Early life stress perturbs the function of microglia in the developing rodent brain: new insights and future challenges

Abstract

The role of the innate immune system in mediating some of the consequences of childhood abuse and neglect has received increasing attention in recent years. Most of the work to date has focused on the role that neuroinflammation plays in the long-term adult psychiatric and medical complications associated with childhood maltreatment. The effects of stress-induced neuroinflammation on neurodevelopment have received little attention because until recently this issue has not been studied systematically in animal models of early life stress. The primary goal of this review is to explore the hypothesis that elevated corticosterone during the first weeks of life in mice exposed to brief daily separation (BDS), which is a mouse model of early life stress, disrupts microglial function during a critical period of brain development. We propose that perturbations of microglial function lead to abnormal maturation of several neuronal and non-neuronal cellular processes resulting in behavioral abnormalities that emerge during the juvenile period and persist in adulthood. Here, we highlight recent work demonstrating that exposure to BDS alters microglial cell number, morphology, phagocytic activity, and gene expression in the developing hippocampus in a manner that extends into the juvenile period. These changes in microglial function are associated with abnormalities in developmental processes mediated by microglia including synaptogenesis, synaptic pruning, axonal growth, and myelination. We examine the changes in microglial gene expression in the context of previous work and their possible contribution to specific developmental and behavioral abnormalities seen in BDS mice and in other animal models of ELS. The possible utility of these findings for developing novel PET imaging to assess microglial function in individuals exposed to childhood maltreatment is also discussed.

1. INTRODUCTION

In developed countries, childhood maltreatment in the form of abuse, neglect, or erratic parenting is the most common and preventable cause of abnormal brain development (Kaffman and Meaney, 2007; Teicher and Samson, 2016). Roughly 1.5 million children are abused or neglected each year in the United States, an alarming statistic that has been documented for the past 30 years. Childhood maltreatment is a significant risk factor for almost half (!) of all childhood psychiatric disorders (Green et al., 2010) including depression, anxiety, cognitive impairment, and behavioral dysregulation. In many cases, childhood psychopathologies progress to chronic psychiatric and medical conditions that include depression, anxiety, eating disorders, personality disorders, psychosis, substance abuse, diabetes, cardiovascular disease, arthritis, and cancer (Kaffman and Meaney, 2007; Teicher and Samson, 2016). This burden of adult comorbidity is associated with poor parenting skills that can propagate a transgenerational injury compounding the clinical, social and economic burdens associated with childhood maltreatment (Kaffman and Meaney, 2007). A 2009 report from the Institute of Medicine estimated the total annual cost related to childhood maltreatment at $247 billion, equal to the estimated annual cost for all cancers combined (Kaffman, 2009; Kaffman and Meaney, 2007).

Precisely how early life stress (ELS) causes such diverse and disabling clinical outcomes is not well understood in humans. However, the observations that ELS increases peripheral inflammation and that inflammation appears to play a significant role in many of the psychiatric and medical sequelae of ELS has led several groups to propose that elevated levels of pro-inflammatory cytokines mediate many of the long-term consequences of ELS (Beumer et al., 2012; Fagundes et al., 2013; Nusslock and Miller, 2016). While their models provide a novel hypothesis to explain the adult consequences of ELS, they do not explore whether ELS-induced alterations in the innate immune system may cause neurodevelopmental abnormalities that contribute to childhood and adult psychopathology.

The sensitivity of microglia to stress and their role in guiding many developmental processes make them a particularly attractive cellular target to orchestrate neurodevelopmental abnormalities induced by ELS (Delpech et al., 2016). However, the few studies that have examined the effects of ELS on microglial function in the developing brain have focused on changes in microglial cell number and morphology using perfused tissue and did not include developmental abnormalities that might be linked to these microglial changes (Gomez-Gonzalez and Escobar, 2010; Roque et al., 2015; Wu et al., 2001). In this review we focus on recent work demonstrating that brief daily separation (BDS), a mouse model of ELS, perturbs microglial function in the developing hippocampus (Delpech et al., 2016) and discuss the potential implications of this finding for the developmental and long-term behavioral consequences seen in BDS mice and in other models of ELS.

2. HYPOTHESIS: ABNORMAL MICROGLIAL FUNCTION DURING A CRITICAL PERIOD OF DEVELOPMENT IS RESPONSIBLE FOR SOME OF THE DEVELOPMENTAL AND LONG-TERM CONSEQUENCES OF BDS

2.1 BDS perturbs diverse developmental processes in the hippocampus

We have developed a mouse model of ELS using the highly stress-reactive Balb/cByj strain. Some litters are raised undisturbed in the absence of nesting material (control condition). The absence of nesting material is used to accurately assess levels of maternal care and to mimic impoverished conditions, but does not appear to be stressful, or cause developmental or behavioral abnormalities in Balb/cByj (Wei et al., 2010; Wei et al., 2012). The remaining litters are separated from the dams for 15min daily for the first three weeks of life (BDS condition). During the separation period the pups are scattered individually in a new cage devoid of maternal cues to simulate a chaotic child-rearing condition. Several technical details, unique to the BDS procedure (i.e. the absence of nesting material in the home cage, the scattering of the separated pups in a cage devoid of maternal cues, and the stress reactivity of the Balb/cByj strain) are likely responsible for the robust and opposing outcomes seen in BDS mice compared to those previously reported after brief handling in rat pups (Hess et al., 1969; Levine et al., 1967; Meaney et al., 1989; Wei et al., 2010; Wei et al., 2012).

BDS causes prolonged elevation of corticosterone during the postnatal period (Wei et al., 2010; Wei et al., 2012), a finding that is consistent with reports in maltreated children (Heim et al., 2000). Exposure to BDS simulates other clinical features seen in maltreated children including elevated markers of peripheral inflammation such as C-reactive protein, reduced myelination, and increased anxiety-like behaviors during the juvenile period and adulthood (Delpech et al., 2016; Wei et al., 2010; Wei et al., 2015; Wei et al., 2012). Adult BDS mice show significant deficits in several hippocampal dependent tasks (Wei et al., 2012) consistent with a large body of work showing abnormal hippocampus development and function in both children and adults that have been maltreated in early life (reviewed in Wei et al. 2015).

Using targeted proteomics of synaptosomal proteins and Golgi staining, we recently demonstrated that exposure to BDS impairs multiple neurodevelopmental processes in the hippocampus including synaptic maturation, synaptic pruning, and the expression of proteins involved in axonal growth, mitochondrial biogenesis and myelination (Wei et al., 2015). Abnormal synaptic density has been reported in other animal models of ELS and reduced myelination has been a consistent finding in several animal paradigms of ELS including non-human primates (summarized in Wei et al. 2015 and Wei et al. 2012). These findings support the validity of BDS as a mouse model of ELS, and raise the question of how BDS perturbs such diverse cellular processes in the developing brain.

2.2 Abnormal microglia function may explain how BDS modifies multiple developmental processes in the hippocampus

Several observations lead us to hypothesize that microglial dysfunction plays a central role in the ability of BDS to perturb normal hippocampal development. First, BDS causes sustained elevation in corticosterone in pups (Wei et al., 2010; Wei et al., 2012). Second, microglia express high levels of the glucocorticoid receptor (GR) and increased levels of corticosterone alter microglial cell number, phagocytic activity, and lead to long-term changes in microglial reactivity (Frank et al., 2015; Walker et al., 2013). Third, microglia are instrumental in guiding several neuronal and non-neuronal developmental processes that are abnormal in BDS and in other models of ELS, including: neurogenesis, synaptogenesis, synaptic pruning, axonal growth, astrocyte maturation, myelination and blood-brain barrier integrity (Harry and Kraft, 2012; Schafer and Stevens, 2015). In other words, changes in microglial function may provide a parsimonious model to explain how BDS modifies a broad range of neurodevelopmental processes (Fig 1).

Figure 1.

Working model: ELS induces the release of stress hormones (A) which directly (B) and indirectly (C) perturb microglia function during postnatal development (D). Alteration in microglial function impairs the normal development of several processes that are regulated by microglia leading to diverse childhood psychopathologies (E). Closure of the critical period renders these developmental abnormalities more difficult to reverse in adulthood (F). Changes in gene expression in microglia persist even in the absence of ongoing stress. These long term-changes contribute to some of the delayed effects of ELS and microglial response to environmental challenges in adulthood and further exacerbate the clinical consequences of ELS (G).

Consistent with this hypothesis we found that BDS alters microglial cell number, morphology, phagocytic activity and the expression of genes implicated in synaptic pruning in the developing hippocampus (Delpech et al., 2016). Many questions related to the role of corticosterone in the effects of BDS on microglia and how alterations in microglial function modify neurodevelopment in BDS remain to be answered. Nevertheless, these findings provide new insights into the mechanisms by which BDS alters microglial function and provide an important first step to rigorously test the role that microglia play in modifying brain development in this paradigm.

2.3 Perturbation in microglial function during a critical period of development leads to long-term changes in neuronal circuitry and behavior

Early disturbances in microglial function can produce abnormal brain connectivity and behavior because microglia are essential for successful development and maturation of circuits that regulate complex behaviors during narrow windows of substantial neurodevelopmental plasticity (Harry and Kraft, 2012; Schafer and Stevens, 2015). During these critical or sensitive periods, the brain initially forms an overly interconnected and inefficient grid that is refined in an activity-dependent manner by pruning of redundant and less active synapses, a process that requires normal microglia function (Paolicelli et al., 2011; Schafer and Stevens, 2015; Tremblay et al., 2010). The deposition of extracellular matrix proteoglycans and myelin related factors stabilizes the refined grid and functions, with other changes, to significantly inhibit circuit remodeling at the end of a critical period (Kaffman and Meaney, 2007; Levelt and Hubener, 2012).

A large body of work has shown that the perturbation of microglial function during specific developmental periods causes structural and behavioral changes that persist into adulthood (Delpech et al., 2016; Harry and Kraft, 2012; Schafer and Stevens, 2015; Zhan et al., 2014). For example, elevated estradiol during a brief developmental window stimulates microglia production of prostaglandin-E2 (PG-E2), which promotes increased spine density in the sexually dimorphic nucleus (SDN) of normally developing male rat pups. This sex-specific increase in spine density persists and programs normal male copulatory behavior in adulthood. Injecting female pups with estradiol or PG-E2 during this sensitive period masculinizes the spine density in the SDN and promotes male copulatory behavior in adult females. Administration of the antibiotic minocycline, a nonspecific microglial inhibitor, to the ventricles of female pups, blocks the ability of PG-E2 or estradiol to increase spine density in the SDN and prevents the development of male copulatory behavior (Lenz et al., 2013). During the early pruning period in the mouse visual system, precisely targeted microglial activity is necessary to remove overlapping synapses from the left and right eyes in the monocular regions of the dorsal lateral geniculate nucleus (Schafer et al., 2012). Microglial activity in the visual cortex is essential to the experience-dependent synaptic plasticity seen in response to monocular deprivation during a specific developmental period in the mouse (Sipe et al., 2016). Ineffective pruning during these sensitive periods leaves behind redundant and inefficient circuits that persist in adulthood (Schafer et al., 2012; Sipe et al., 2016).

In the brain, the fractalkine receptor CX3CR1 is exclusively expressed on microglia allowing these cells to detect the fractalkine ligand secreted by forebrain neurons. Deletion of CX3CR1 causes abnormal synaptic pruning in the developing hippocampus and prevents synaptic multiplicity in the Schaffer collaterals, which is a form of synaptic strengthening that takes place during the second week of life in normally developing mice. The result is impaired connectivity within the hippocampus and impaired hippocampal dependent memory (Rogers et al., 2011; Zhan et al., 2014). The reduced connectivity in CX3CR1 knockout mice is not restricted to the hippocampus and is also seen between the prefrontal cortex and several other brain regions. Interestingly, reduced connectivity between the ventral hippocampus and the prefrontal cortex is highly correlated with social deficits seen in CX3CR1 knockout mice (Zhan et al., 2014).

Transient depletion of microglia during the first week of life blunts stress reactivity and reduces anxiety and depression-like behaviors in adult rats. These behavioral changes occur despite complete restoration of microglial number by the third week of life (Nelson and Lenz, 2017). Abnormal microglial function early in development also leads to long-term abnormalities in myelination, blood brain barrier integrity, and angiogenesis (da Fonseca et al., 2014; Harry and Kraft, 2012; Pont-Lezica et al., 2011; van Tilborg et al., 2016) demonstrating that disruption of microglial function impacts non-neuronal as well as neuronal cells.

These diverse findings demonstrate that microglial activity during distinct developmental windows is necessary to guide normal development of multiple cell types and disruption of microglial function during these sensitive periods leads to persistent changes in psychiatrically relevant behaviors such as anxiety, depression, and social affiliation. Therefore, the ability of BDS to perturb microglial function during a critical period of neurodevelopment may explain the increased anxiety and the hippocampal-dependent memory deficits seen in adult BDS mice (Wei et al., 2012).

2.4 Long-term changes in microglial activity may cause vulnerability to”secondary hits” that further contribute to functional deterioration in adultanimals exposed to ELS

Several studies have shown that exposure to ELS can alter microglial function in a manner that extends into the juvenile period and adulthood even in the absence of ongoing stress (Delpech et al., 2016; Diz-Chaves et al., 2013; Diz-Chaves et al., 2012; Schwarz et al., 2011; Takatsuru et al., 2015). Such long-term changes in microglial function may be responsible for the delayed cognitive and structural abnormalities reported in the hippocampus with some forms of ELS (Wei et al., 2015). This proposition is supported by data showing that microglia regulate the rates of spine formation and elimination in the motor cortex, and are also necessary to support normal hippocampal dependent function in juvenile and adult mice (Parkhurst et al., 2013). Long-term programming may also alter microglial response to later environmental challenges such as stress, high-fat diet, and infection creating a vulnerability to “secondary hits” that further contribute to deterioration in function in adulthood, see Nusslock and Miller (2016) for an excellent review of this topic.

2.5 The proposed model and its predictions

In figure 1 we propose a working model to explain how ELS modifies microglial function and how these alterations lead to abnormal circuit development that persists into adulthood. We want to emphasize that this model does not assume that microglia are the only cell type affected by ELS, nor does it assume that all developmental and behavioral consequences of ELS are mediated by changes in microglia function in early life. Additionally, ELS is not a homogenous concept and therefore different paradigms may cause different changes in microglial function (see sections 3.3 and 4.1 below). Finally, we recognize that some aspects of this model are speculative and require additional work to verify. The schematics in figure 1 should be considered a working model to direct future work to clarify important elements of the hypothesis.

In the model, increased levels of stress hormones like corticosterone, catecholamines and corticotropin-releasing hormone during early development is the initiating event (Fig 1A). These mediators affect the development of several cell types that closely and reciprocally interact with each other (Schafer and Stevens, 2015). Microglial function is consequently modified in two ways: first, by direct actions of these stress hormones (Fig 1B) and second, by signals that originate from other cell types (Fig 1C). Early disruption of microglial function leads to the abnormal maturation of several neuronal and non-neuronal cellular processes (Fig 1D) in circuits that regulate psychiatrically relevant behaviors during the juvenile period (Fig 1E). During the juvenile period, the deposition of extracellular proteoglycans and other molecules that restrict plasticity (Kaffman and Meaney, 2007; Levelt and Hubener, 2012) renders such developmental abnormalities more difficult to reverse in adulthood (Fig 1F). ELS also alters the capacity of microglia to modify their immediate vicinity and respond to “secondary hits” like exposure to additional stress, infection, or high-fat diet during the juvenile period and adulthood (Fig 1G). These long-term effects further exacerbate the developmental abnormalities and the clinical consequences of ELS.

Based on this model, early interventions that preserve normal microglial function could reduce some of the developmental and behavioral abnormalities seen in individuals exposed to ELS. Neuroimaging techniques to monitor microglial activity in early development and later stages could provide important information underlying functional outcomes and help identify vulnerabilities to psychopathologies throughout the lifespan (see section 4).

3. GENOMIC STUDIES PROVIDE IMPORTANT INSIGHTS INTO HOW BDS ALTERS MICROGLIAL FUNCTION IN THE DEVELOPING HIPPOCAMPUS

3.1 BDS causes microgliosis in the hippocampus of 14-day old pups

We first examined the effects of BDS on cell number, density, and morphology of microglia in the hippocampus of 14 and 28-day old mice. We focused the analysis on male mice because the behavioral effects of BDS are more pronounced in male mice (Wei et al., 2010). Postnatal day 14 (P14) was chosen because the number and density of microglia in the hippocampus peak at P14 (Paolicelli et al., 2011), suggesting an important role for microglia at this age. Also, the separation procedure causes prolonged elevation of corticosterone in P14 BDS pups (Wei et al., 2010; Wei et al., 2012) allowing us to examine microglial morphology and function when the pups are actively stressed. P28 was selected because BDS mice are no longer exposed to separation stress at this age (the last day of separation is P21) and synaptogenesis, synaptic pruning, axonal growth and myelination are abnormal at this age (Wei et al., 2015). P28 also provides an important age to assess the developmental trajectory of BDS because synaptogenesis, axonal growth, and neurogenesis reach adult-like levels at this age in the normal hippocampus (Wei et al., 2015).

We found that BDS increased the number, density, and surface area of microglia in the hippocampus of 14-day old pups (Delpech et al., 2016). Similar changes in microglial morphology have been reported in the hippocampus of adult rodents exposed to stress (Beumer et al., 2012) and in 15-day old rat pups exposed to maternal separation (Roque et al., 2015), indicating that microglia are sensitive to stress across multiple rodent species and ages. The effects of BDS on microglial number and morphology were no longer present in the hippocampus of 28-day old mice suggesting that these changes require exposure to ongoing stress. This interpretation is consistent with work showing that the administration of the glucocorticoid receptor antagonist RU486 blocked microglia proliferation in adult stressed mice (Nair and Bonneau, 2006).

3.2 Exposure to BDS alters the expression of genes implicated in the proliferation, migration, and phagocytic activity of microglia

To investigate how BDS modifies the number and function of microglia we characterized the expression of 540 immune-related genes in microglia isolated from the developing hippocampus. A highly homogenous population of microglia was purified from the developing hippocampus of a single pup using a Percoll step-gradient followed by flow cytometry. This protocol was used to harvest microglia from mice injected with lipopolysaccharide (LPS) endotoxin or saline, and identified over 150 gene transcripts, including many known targets of LPS such as IL-1 , IL-6, and Tnf-α, that were significantly elevated in microglia isolated from LPS-injected mice. These findings demonstrated the method’s ability to distinguish “activated” and “non-activated” microglia, and confirmed its utility in characterizing BDS effects on microglia function in the developing hippocampus (Delpech et al., 2016).

Of the 286 genes transcripts detected in P14 microglia, 58 were significantly altered by BDS (See table 1 for a summary of some of the genes and their cellular function). Place Table 1 around here. BDS modified the expression of several genes known to regulate the proliferation and survival of microglia including Csf1, Tgfbr2, and Caspase 2, pointing to possible mechanisms for the increase in microglia seen at this age in BDS mice (Delpech et al., 2016).

Table 1.

Summary of some of the genes that are modified by BDS in microglia harvested from the developing hippocampus and their cellular function.

Cell function	Examples of genes
Microglia proliferation and survival	Csf1, Csf3r, Casp2, casp8, Tgfbr2, CD81
Cell migration and process extension	Ccl4, CX3Cr1, Ccr6, Itga5, CD29/Itgb1
Phagocytic activity	C1qa, C1qb, CD81, Itgam, Ncf4, TLR9
Immune function, cytokine production	Tgfbr2, IL-6, IL-1a, Fyn, Jak3
Antigen presentation	H2-Ob, H2-Eaps
Glucocorticoid resistance	FKBP5

BDS reduced the expression of CX3CR1 by 50% (p= 0.07) and increased C1qa levels by 50% at P14 (Delpech et al., 2016). These genes play an important role in normal synaptic pruning (Paolicelli et al., 2011; Schafer et al., 2012) and their dysregulation in postnatal microglia may explain the spine abnormalities seen in BDS mice (Wei et al., 2015). Recent work has shown that elevated C1qa secreted by microglia can induce a form of reactive astrocytosis that is toxic to neurons (Liddelow et al., 2017) raising the possibility that such reactive astrocytosis may be present in the developing hippocampus of BDS mice.

Increased microglial expression of integrin 1 (Itgb1/CD29) seen in BDS mice has also been reported in adult rats exposed to chronic restraint stress (Hinwood et al., 2013). Itgb1/CD29 increases in response to activation of P2Y12R on microglia, an increase that is necessary to direct microglial process extension toward an ATP gradient (Ohsawa et al., 2010). P2Y12 receptor activation is also necessary for the removal of less-active synapses in response to monocular deprivation (Sipe et al., 2016). It will be interesting to test whether blocking P2Y12 receptor activation prevents the increase in microglial Itgb1/CD29 levels and corrects some of the synaptic and behavioral abnormalities seen in BDS mice.

Promoter analysis revealed that 76% of the transcriptional changes seen in microglia from the hippocampus of P14 BDS mice could be attributed to alterations in the activity of Creb1, Sp1, and RelA. Although these transcription factors play a critical role in many immune functions including cell cycle progression, cytokine production, cell migration, phagocytosis, and cellular differentiation, their role in programming microglial function in the developing brain has not been studied (Delpech et al., 2016). This analysis allows us to move beyond the effect of BDS on specific transcripts and focus on the role that master regulators such as RelA or SP1 play in modifying microglial function and hippocampal development in BDS mice (see also section 3.5 on the role that PU.1 may play in modifying neurodevelopment in BDS mice).

Given that BDS altered expression of several genes implicated in phagocytosis, (see table 1) we tested the ability of microglia acutely isolated from the hippocampus of P14 control and BDS pups to phagocytose opsonized E. coli particles conjugated with pH-sensitive fluorescent dye (i.e. pHrodo-E. coli). Using this ex-vivo assay, we found no differences in phagocytic activity between BDS and control P14 microglia (Delpech et al., 2016). This finding should be interpreted with caution given the artificial nature of these particles. Additional work is needed to test the effect of BDS on phagocytic activity of more relevant substrates such as synaptosomes, neural stem cells, myelin, or axonal segments.

3.3 Different forms of ELS cause different neuroinflammatory responses in the developing hippocampus

Microglia isolated from the hippocampus of P14 BDS mice showed a complex immune response that included both pro and anti-inflammatory changes. BDS increased the expression of some pro-inflammatory genes (e.g. Fyn, Irak4) while reducing the expression of other pro-inflammatory genes (e.g. IL-1 , RelA). BDS did not increase levels of classic pro-inflammatory cytokines such as IL-1 , TNF- or IL-6 in microglia isolated from the hippocampus at P14 (Delpech et al., 2016). In contrast, a recent study found a 10–15 fold increase in the levels of IL-1β mRNA in the hippocampus of 15-day old rat pups exposed to 3hrs of maternal separation (Roque et al., 2015). Such a robust induction of IL-1β resembles levels seen after LPS administration (Delpech et al., 2016) and is significant because exogenous administration of IL-1β postnatally impairs myelin formation and causes long-term cognitive deficits (Favrais et al., 2011). The different pro-inflammatory responses between the two paradigms might be due to the length of the maternal separation (15min vs. 180min) or genetic differences between Balb/cByj mice and Sprague Dawley rat pups. We do not believe that these differences are due to tissue processing (isolating microglia vs. hippocampal dissection) because we did not find an increase in IL-1β mRNA or protein when the hippocampus was rapidly dissected at P14 in BDS pups. These findings underscore that ELS is not a homogenous entity and that different models are likely to cause different immune responses in the developing brain. The key unanswered question is what mechanisms are responsible for these differences (see also sections 3.7 and 4.2).

3.4 BDS perturbs the maturation of microglia in the developing hippocampus

Despite the fact that the number and morphology of microglia were similar in the hippocampus of 28-day old BDS and control mice, microglia isolated from the hippocampus of 28-day old BDS mice showed an increase in ex-vivo phagocytic activity and altered expression of many genes (Delpech et al., 2016). These findings highlight the limitations of using cell counting and morphological analysis of fixed tissue to characterize microglia function.

The overall transcriptional response seen in microglia isolated from the hippocampus of P28 BDS mice was anti-inflammatory. However, there were several examples of pro-inflammatory genes that were elevated in BDS mice including a significant increase in IL-6 as well as reduced expression of anti-inflammatory genes such as Tgfbr2. This complexity may be due to an underlying heterogeneity in the microenvironment that surrounds individual microglia cells in the hippocampus including differences between sub-regions of the hippocampus (i.e. CA1 vs. dentate gyrus) or proximity to blood vessels and needs to be investigated further (see section 4.2).

There was relatively little overlap in the lists of immune genes affected by BDS at P14 and P28. We have replicated these findings in two independent cohorts of BDS mice indicating that these age specific effects are robust and reproducible. These genomic results suggest that BDS affects microglia differently at these two ages and are consistent with other age-specific changes in microglial function such as microgliosis at P14 and increased phagocytic activity at P28 (Delpech et al., 2016).

Several observations demonstrate that microglia in the hippocampus undergo important changes between P14 and P28 in normally developing mice. First, the density of microglia decreases three fold from day 14 to 28. Second, microglia isolated from the hippocampus at day 28 show increased expression of many inflammatory genes and reduced phagocytic activity. Promoter analysis indicates that alterations in the transcriptional activity of PU.1, RelA, Creb1, and Sp1 account for these normal developmental changes. These four transcription factors account for 86% of the genes differentially regulated in microglia harvested from the hippocampus of 28-day old control and BDS mice. These findings suggest that BDS perturbs the normal maturation of microglia in the developing hippocampus. This interpretation is supported by the observation that the phagocytic activity of microglia isolated from P28 BDS mice resembles the phagocytic activity seen in P14 control and P14 BDS microglia, and was significantly higher than phagocytic activity in microglia from P28 control mice (Delpech et al., 2016).

3.5 Reduced levels of the master regulator PU.1 in postnatal microglia may orchestrate many of the microglial, neurodevelopmental and behavioral consequences of BDS

PU.1 is a master regulator whose expression is both necessary and sufficient to induce macrophage-like cellular phenotype (Kierdorf and Prinz, 2013; Smith et al., 2013). The brain of a PU.1 knockout mouse is devoid of microglia, a deficit that is thought to be responsible for the reduced neural stem cell proliferation and astrogenesis seen in these mice during embryonic development (Antony et al., 2011). Ectopic expression of PU.1 is sufficient to induce macrophage-like cellular phenotype in fibroblasts and mast cells and PU.1 knockdown reduces the survival of adult microglia and reduces their phagocytic activity towards amyloid-beta1–42 peptide (Kierdorf and Prinz, 2013; Smith et al., 2013). PU.1 knockout mice die at around birth due to infection (Kierdorf and Prinz, 2013) and therefore little is known about the role that PU.1 plays in postnatal microglia. Our work shows that PU.1 levels increase in hippocampal microglia during the second and third weeks of life, a process that appears to coordinate the normal maturation of these cells (see section 3.4 above). PU.1 is highly expressed in microglia and its expression is down regulated in microglia from BDS mice at both P14 and P28. Reduced PU.1 levels account for 74% of all the genes whose expression was modified by BDS at P28 consistent with the notion that BDS perturbs the maturation of microglia by reducing expression of PU.1 (Delpech et al., 2016).

It is not clear whether reduced PU.1 in postnatal microglia serves an adaptive function that compensates for abnormal development caused by BDS and/or whether it contributes to the developmental and behavioral abnormalities seen in BDS mice. It is important to consider that reduced microglial PU.1 at P14 may adaptively reduce microglial cell number while simultaneously compromising synaptic pruning. We are now testing whether microglial PU.1 reduction in P14 control pups is sufficient to reproduce some of the developmental and behavioral abnormalities seen in juvenile BDS mice and whether PU.1 reductions make pups more sensitive or more resilient to the developmental and behavioral consequences of BDS.

3.6 A possible role for corticosterone in altering microglial function

The mechanisms by which BDS alters microglial function in the developing hippocampus are yet to be identified. One likely candidate is corticosterone (Fig 1A), which is elevated in BDS pups (Wei et al., 2010; Wei et al., 2012). Glucocorticoids have an anti-inflammatory role followed by a delayed pro-inflammatory “priming effect” (Bellavance and Rivest, 2014; Frank et al., 2015). This complex immune modulation is consistent with the mixed pro and anti-inflammatory pattern seen in microglia harvested from the developing hippocampus of BDS mice. Elevated corticosterone increases microglia proliferation and is associated with increased phagocytic activity (Bellavance and Rivest, 2014; Nair and Bonneau, 2006), findings also seen in BDS microglia. Microglia isolated from the hippocampus of 14-day old BDS pups show increased expression of several known transcriptional targets of GR including FKBP5, Fyn, Jak3, and C1qa (Klengel et al., 2013; Rao et al., 2011; Walker, 1998), suggesting a direct effect of corticosterone on postnatal microglia (Fig 1B). The ability of GR to antagonize RelA activity (Rao et al., 2011; Ratman et al., 2013) may explain why altered RelA activity accounts for more than half of the transcriptional changes seen in microglia harvested from the hippocampus of 14-day old BDS pups (Delpech et al., 2016). To address these issues we are now testing whether a microglia-specific deletion of the GR can restore normal microglial function, synaptic development, and anxiety-like behavior in BDS mice.

3.7 Mechanisms of long-term change in microglial function induced by ELS

The altered gene expression and increased phagocytic activity seen in P28 BDS mice is consistent with previous work showing that exposure to ELS leads to long-term changes in microglial function that persist in adulthood. For example, in vivo two photon microscopy has shown that exposure to maternal separation during the first two weeks of life enhanced the surveillance rate of microglia present in the adult sensory cortex (Takatsuru et al., 2015). Tactile stimulation in adult control mice led to a transient increase in microglial ramification (i.e. increased number of processes present in an individual microglia cell) that was restored to baseline within 20 minutes. In contrast, the same tactile stimulation increased microglial ramification over a three-hour period in adult mice exposed to early maternal separation. These alterations in microglial surveillance rate may contribute to an enhanced sensitivity to pain reported in mice exposed to maternal separation (Takatsuru et al., 2015). This important preclinical finding that may explain the increase in somatization and pain-related disorders seen in individuals exposed to childhood maltreatment (Pieritz et al., 2015).

Work by Diz-Chaves et al. (2013) provides another example of long-term alterations in microglial function induced by ELS. Elevated IL-1β and TNF- mRNA were found in the hippocampus of adult male mice that had been exposed to prenatal stress. Prenatally stressed adult mice also showed increased microglia in the dentate gyrus after LPS injection, an increase not seen in control mice (Diz-Chaves et al., 2013). These findings are intriguing given the ability of corticosterone to induce long-term changes in microglial response to LPS (reviewed in Frank et al., 2015) and support the notion that ELS leads to long-term changes in microglial response to “secondary hits” in adulthood (Fig 1G).

Currently little is known about the mechanisms regulating the long-term changes in microglial gene expression and function caused by ELS. The generation of stable chromatin states that yield long-term changes in the expression of microglial genes is one possible mechanism. The longevity of brain microglia and their role in mediating some of the behavioral consequences of stress in later life (Beumer et al., 2012; Harry and Kraft, 2012; Iwata et al., 2016) make this is an attractive possibility. Such stable chromatin changes have been reported in neuronal populations regulating stress reactivity in adult animals exposed to ELS (Kaffman and Meaney, 2007). Similar mechanisms may be at work in microglia. Indeed, Schwarz et al. (2011) have found that rats that were handled during the first three weeks of life showed reduced drug-seeking behavior when injected with morphine in adulthood. This protective effect was associated with reduced DNA methylation in a regulatory element of the IL-10 gene in microglia in the nucleus accumbens (NAc) of handled rats. The reduced microglial DNA methylation was associated with elevated IL-10 expression that was detected in handled rats as early as P10 and continued to be elevated in the adult NAc. The evidence that elevated IL-10 in the NAc might attenuate drug seeking behavior comes from data showing that while acute injection of morphine induces immune activation in the NAc, administration of anti-inflammatory agents like minocycline or manipulations that increase IL-10 levels in the NAc block drug-seeking behavior (Schwarz et al., 2011). These findings demonstrate that experiences early in life can cause stable epigenetic alterations resulting in long-term changes in microglial gene expression in circuits that regulate drug-seeking behavior in adulthood.

We identified a few genes whose expression is modified by BDS in microglia at P14 and P28, and may be regulated by changes in DNA methylation (Delpech et al., 2016). The most interesting is the Tgf- receptor 2 (Tgfbr2). Preliminary data show that Tgfbr2 levels are also reduced in microglia isolated from the hippocampus of adult BDS mice indicating long-term transcriptional regulation of this gene. Tgfbr2 is one of the most highly expressed genes in postnatal microglia (Delpech et al., 2016), and Tgf-signaling is essential for maintaining the molecular signature of microglia as well as cell proliferation and immune quiescence (Butovsky et al., 2014). A microglia-specific deletion of Tgfbr2 causes altered morphology and increased microglia number (Buttgereit et al., 2016) in a manner that resembles the microgliosis seen in 14-day old BDS pups. Deletion of Tgfbr2 also increases microglial expression of pro-inflammatory cytokines such as IL-1β and Tnf- and other surface molecules such as CD45 consistent with the idea that Tgfbr2 suppresses immune activation in microglia (Buttgereit et al., 2016). We did find an increase in CD45 but not IL-1 or Tnf- in microglia isolated from P14 BDS mice. The absence of elevated levels of IL-1 and Tnf-might be due to the relatively mild reduction in Tgfbr2 in BDS (about 40% reduction in BDS compared to 90% reduction in the Tgfbr2 knockout). A finding showing that 3 hours of daily maternal separation causes more severe reduction in Tgfbr2 levels in microglia may explain why prolonged maternal separation but not BDS increases IL-1β levels in the developing hippocampus (see section 3.3 above). Deleting Tgfbr2 in fibroblasts causes up regulation of the de-novo DNA methylase DNMT1 protein and increased DNA methylation of many genes (Yamashita et al., 2008). If similar findings are confirmed in microglia, reduced Tgfbr2 may also provide a mechanism by which BDS causes long-term changes in DNA methylation, gene expression, and function in adult microglia.

3.8 BDS increases peripheral immune responses during the postnatal period

Peripheral inflammation is tightly linked to microglial function in some paradigms of adult stress (Hodes et al., 2014; Iwata et al., 2016; Nusslock and Miller, 2016; Wohleb et al., 2014). This prompted us to investigate the effects of BDS on several peripheral markers of inflammation. We found that BDS increased levels of total CD11b−positive cells and CD11b/SSC^hi/Ly6C^int (granulocytes), but not CD11b/SSC^low/Ly6C^hi (monocytes) in the blood of P14 pups. These changes were no longer present at P28, suggesting that they require ongoing stress (Delpech et al., 2016). Recruitment of CD11b/CD45^high monocytes from the periphery into the brain has been shown to play a role in modifying microglial function and increasing anxiety-like behavior in adult male mice exposed to repeated social defeat (Wohleb et al., 2014). However, we found no change in the number of CD11b/CD45^high monocytes isolated from the hippocampus at either P14 or P28 (Delpech et al., 2016).

One of the most consistent peripheral immune markers associated with childhood maltreatment is elevation of the acute phase reactant, C-reactive protein (Baumeister et al., 2016; Coelho et al., 2014). We found that BDS caused a significant increase in levels of C-reactive protein (CRP) in the plasma of both P14 and P28 mice indicating a robust activation of this peripheral marker even in the absence of ongoing stress (Delpech et al., 2016). Elevated levels of CRP have been shown to increase blood brain barrier permeability (Closhen et al., 2010) allowing blood products to access the brain’s parenchyma and modify microglial function (Fig 1C). Perivascular microglia play a role in maintaining the integrity of the blood brain barrier and disturbing their function may further compromise its integrity (da Fonseca et al., 2014). We have not yet tested whether BDS increases blood brain barrier permeability but Gomez-Gonzales and Escobar (2009) have shown increased blood brain barrier permeability during the postnatal period in rat pups exposed to either prenatal or postnatal stress (Gomez-Gonzalez and Escobar, 2009).

Increased peripheral immune activation has been reported in other paradigms of ELS (Bronson and Bale, 2014; Hennessy et al., 2009; Roque et al., 2015). For example, prenatal stress increased levels of IL-6 in the placenta and caused hyperactivity in the dark-light test in adult male, but not female, offspring. Administering the anti-inflammatory drug, aspirin, during prenatal stress normalized IL-6 in the E12.5 placenta and restored normal locomotor activity in adult male offspring (Bronson and Bale, 2014). These findings are intriguing because maternal immune activation with poly I:C at E12.5 also increased IL-6 in the placenta, an increased that was necessary to induce a transient upregulation of IL-6 transcription in the cerebellum (Wu et al., 2017). It is currently unclear whether microglia are responsible for this transient increase in IL-6 levels in the cerebellum of E12.5 pups, but this increase in IL-6 seems essential for inducing some of the long term changes associated with maternal immune activation including a reduction in the number of calbindin-positive Purkinje cells in the cerebellum and social deficits (Wu et al., 2017).

4 CHALLENGES AND FUTURE DIRECTIONS

4.1 Additional characterization of microglial gene expression

Although our work shows that BDS alters microglial function in the developing hippocampus, it is not clear that transcriptional changes in microglia drive any of the developmental and behavioral abnormalities seen in BDS mice. Future work should clarify whether dysregulation of specific genes in microglia such as PU.1, GR, Tgfbr2, Cq1a, and Itgb1/CD29 affect the development of neuronal and non-neuronal cells in BDS mice (see Fig 1D). Since BDS and many other rodent models of ELS cause more pronounced behavioral changes in male mice (Loi et al., 2015; Wei et al., 2010), it will be important to test whether similar transcriptional changes also occur in postnatal microglia from female mice.

Unbiased whole genome methods, like mRNA-seq or microarray, can shed light on the effects of BDS on microglial functions not represented by the 540 gene immune panel used by our lab. Important missing information includes expression of neurotrophic factors, genes involved in cytoskeleton rearrangement, chromatin remodeling factors, ion channels, and micro-RNAs. Proteomic approaches are also needed to assess protein levels and post-translational modifications. Such work could identify novel targets for positron emission tomography (PET) imaging to characterize microglial function in live animals and humans (see also section 4.2 below).

More work is also needed to confirm the genomic and proteomic BDS findings using in-situ methods in intact tissue. This approach could help determine if Tgfbr2 mRNA in microglia is uniformly reduced across the hippocampus or localized to specific microenvironments. If Tgfbr2 was found to be reduced in microglia close to blood vessels, this might indicate that peripheral signals from the circulation are responsible for this change and help explain the complex immune pattern found at P14 and P28. Such an approach could also determine the pattern of microglial Tgfbr2 expression in other developing brain regions like the amygdala and prefrontal cortex.

4.2 Generalizability and clinical implications

We are now testing whether other models of ELS such as low bedding (Rice et al., 2008) and unpredictable 1-hr maternal separation during the postnatal period affect microglial cell number, phagocytic activity, and gene expression in the developing hippocampus of Balb/cByj mice. The results of these studies are preliminary but they suggest that different paradigms cause very different alterations in microglial function, highlighting the need to clarify the mechanisms by which different models of ELS alter postnatal microglia function. We are not aware of any studies that have examined the effects of ELS on microglia function in the developing brain of non-human primates. Such studies would be an important bridge to related studies in children. Our hope is that the findings summarized here and advances in imaging will inspire translational work in non-human primates and humans.

A relevant clinical breakthrough has been the development of PET ligands to examine microglial activation in humans. This method has recently been used to demonstrate microglial activation in individuals at ultra-high risk for psychosis or diagnosed with schizophrenia (Bloomfield et al., 2016), depression (Setiawan et al., 2015), or chronic fatigue syndrome (Nakatomi et al., 2014). So far it has not been used to assess microglial activation in young adults exposed to ELS. Given that ELS is associated with a broad range of psychopathologies including psychosis, depression, and chronic fatigue syndrome (Heim et al., 2006; Teicher and Samson, 2016) these recent PET findings might reflect underlying exposure to ELS. Significantly, the ligand target in the studies mentioned was originally identified using microglia activated by LPS, and may not be appropriate to evaluate microglial changes stimulated by stress. It is likely that more subtle and diverse microglial phenotypes will be characterized over time (Buttgereit et al., 2016). In fact, our findings indicate that BDS causes very different microglial activation compared to LPS and our preliminary data suggest that BDS does not increase expression of TSPO, which is the receptor used to detect microglia activation in current PET imaging. Animal models can help guide the development of novel PET ligands with greater specificity and sensitivity for detecting microglial responses to stress in humans.

5 CONCLUDING REMARKS

Recent work has shown that BDS perturbs microglial function in the developing hippocampus while impairing several developmental processes that are mediated by microglia such as synaptogenesis, synaptic pruning, axonal growth and myelination. The identification of specific genes whose expression is modified by BDS in postnatal microglia provides an important step towards clarifying the role that microglia dysfunction plays in causing some of the developmental and behavioral abnormalities seen in BDS mice. Demonstrating that microglia mediate some of the developmental abnormalities seen in BDS mice will provide a novel mechanism to explain how stress early in life affects such diverse developmental processes in ways that persists into adulthood and will hopefully inspire translational work to examine this issue in humans.

Highlights.

• ELS perturbs microglial function during a critical period of brain development

• ELS impairs multiple developmental processes that are regulated by microglia

• ELS modifies the activity of key transcription factors in postnatal microglia

• ELS causes long-term changes in microglia function