Microglia across the lifespan: from origin to function in brain development, plasticity and cognition

Abstract

Microglia are the only immune cells that permanently reside in the central nervous system (CNS) alongside neurons and other types of glial cells. The past decade has witnessed a revolution in our understanding of their roles during normal physiological conditions. Cutting‐edge techniques revealed that these resident immune cells are critical for proper brain development, actively maintain health in the mature brain, and rapidly adapt their function to physiological or pathophysiological needs. In this review, we highlight recent studies on microglial origin (from the embryonic yolk sac) and the factors regulating their differentiation and homeostasis upon brain invasion. Elegant experiments tracking microglia in the CNS allowed studies of their unique roles compared with other types of resident macrophages. Here we review the emerging roles of microglia in brain development, plasticity and cognition, and discuss the implications of the depletion or dysfunction of microglia for our understanding of disease pathogenesis. Immune activation, inflammation and various other conditions resulting in undesirable microglial activity at different stages of life could severely impair learning, memory and other essential cognitive functions. The diversity of microglial phenotypes across the lifespan, between compartments of the CNS, and sexes, as well as their crosstalk with the body and external environment, is also emphasised. Understanding what defines particular microglial phenotypes is of major importance for future development of innovative therapies controlling their effector functions, with consequences for cognition across chronic stress, ageing, neuropsychiatric and neurological diseases.

Abbreviations

BBB

blood–brain barrier

BDNF

brain‐derived neurotrophic factor

BM

bone marrow

CNS

central nervous system

E

embryonic day

gw

gestational weeks

IBA1

ionised calcium‐binding adapter molecule 1

IL

interleukin

LTP

long‐term potentiation

SGZ

subgranular zone

SVZ

subventricular zone

TGFβ

transforming growth factor β

TNFα

tumour necrosis factor α

TSPO

translocator protein

YS

yolk sac

Introduction

In the past decade, studies on microglia have expanded from investigating their function as resident macrophages of the brain and mediators of injury, neuroinflammation

and neurodegeneration (reviewed in (Prinz et al. 2011; Cartier et al. 2014; Prinz & Priller, 2014) to understanding their origins and non‐immunological roles in the central nervous system (CNS). Advances in genetic tools that allowed specific fate mapping of microglia (Ginhoux et al. 2010; Goldmann et al. 2013; Parkhurst et al. 2013; Yona et al. 2013; Wieghofer et al. 2015), in vivo imaging techniques, which revealed microglial activities in the brain milieu (Davalos et al. 2005; Nimmerjahn et al. 2005; Wake et al. 2009; Tremblay et al. 2010; Li et al. 2012), and high throughput gene expression analyses that differentiated microglia from their myeloid relatives and other brain cells (Butovsky et al. 2012; Gautier et al. 2012; Chiu et al. 2013; Hickman et al. 2013; Butovsky et al. 2014), have aided the rapid progress towards unravelling the multiple facets of microglia in shaping brain development and maintenance of homeostasis throughout the lifetime of an organism. Recent data have led to the call for a new approach to the classification of microglia in the family of mononuclear phagocytes (Guilliams et al. 2014), as well as a reconsideration of the description of microglial phenotypes based on their macrophage attributes M1 and M2 (Martinez & Gordon, 2014; Heppner et al. 2015), suggesting a major paradigm shift in the field. In this review we focus on recent findings shedding light on the physiological roles of microglia during development, adulthood and ageing, and discuss their impact on our understanding of diseases.

Origin of brain microglia

The colonisation of the CNS by microglial cells is evolutionarily conserved across vertebrate species and takes place even before formation of the neuroectoderm‐derived glial cell types, i.e. astrocytes and oligodendrocytes (Verney et al. 2010; Schlegelmilch et al. 2011; Swinnen et al. 2013). Unlike neuroectodermal cells, microglia are of myeloid origin. Several studies employing genetic fate mapping tools systematically delineated the source of microglia and confirmed that they arise mainly from yolk sac (YS) primitive macrophages (Ginhoux et al. 2010; Hoeffel et al. 2012; Schulz et al. 2012; Kierdorf et al. 2013; Gomez Perdiguero et al. 2015; Hoeffel et al. 2015), reside in the brain throughout life, and maintain their numbers through a process of self‐renewal (Hashimoto et al. 2013).

Normal blood circulation is essential for seeding of the CNS by YS macrophages. De novo formation and remodelling of blood vessels in the mouse embryo occur between embryonic days (E) 8.0 and E10.0 (Walls et al. 2008). Around E7.0, precursors of haematopoietic and endothelial cells expressing vascular endothelial growth factor receptors migrate from the primitive streak to the proximal YS to form the blood islands (Jones, 2011), where the multi‐lineage c‐kit⁺ erythromyeloid YS precursor cells of microglia reside (Kierdorf et al. 2013). Yolk sac precursors mature from A1 (CD45⁺ c‐kit^lo CX₃CR1⁻ F4/80⁻) to A2 (CD45⁺ c‐kit⁻ CX₃CR1⁺ F4/80^hi) amoeboid macrophages in the blood islands and cephalic mesenchyme before taking on a phenotype of mature macrophages in the neuroepithelium at E10.5 (Roumier et al. 2008; Mizutani et al. 2012). Fate mapping of haematopoietic precursors that express the runt‐related transcription factor Runx1 by genetic targeting between E6.5 and E10.5 revealed that YS microglia are specified between E7.0 and E7.5 (Ginhoux et al. 2010). The absence of brain microglial progenitors despite normal YS haematopoiesis in E9.5–10.5 Ncx‐1knockout mouse embryos with defective blood circulation supports the notion that brain recruitment of YS progenitors depends on a functional circulatory system (Ginhoux et al. 2010).

Microglia enter the brain rudiment via the leptomeninges and lateral ventricles by E9.5 and distribute throughout the cortical wall from both directions at different speeds with varying rates of proliferation and maturation, based on region and developmental stage (Ginhoux et al. 2010; Arnoux et al. 2013; Swinnen et al. 2013). Growing coverage of the CNS by amoeboid macrophages that eventually evolve into ramified microglia is possible due to rapid proliferation of 40–80% of their population between E10.5 and P0 (Kierdorf et al. 2013). Human amoeboid microglia penetrate the developing cerebral cortex at 4.5 gestational weeks (gw) via the pial surface, ventricles and choroid plexus (Monier et al. 2007; Verney et al. 2010). Radial and tangential migration of microglia was observed towards the putative white matter, subplate and cortical plate layers, while pial cells populate the prospective cortical layer I. A second wave of microglial invasion via the vasculature at 12–13 gw is limited to the white matter (Monier et al. 2007; Verney et al. 2010). In the initial two postnatal weeks of mouse development, microglial cell numbers increase, followed by a gradual decline of 50% from week three to six, after which microglial density stabilises (Nikodemova et al. 2015). The increase in microglial apoptosis is concomitant with a decrease in proliferation, contributing to the overall reduction in microglial numbers (Nikodemova et al. 2015).

Despite the exponential increase in studies focused on microglial origin over the past 5 years, fundamental questions remain regarding their cell cycle dynamics. How long do microglia live? Are long‐lived microglial progenitors found within the brain, and do microglia get replaced by bone marrow (BM)‐derived cells? Several reports have excluded the contribution of definitive haematopoiesis to brain microglia by using specific targeting of haematopoietic stem cells generated in the fetal liver or the aorta–gonad–mesonephros in different myeloid‐specific genetic reporter models for Csf‐1r (colony‐stimulating factor‐1 receptor), Flt3 (Fms‐like tyrosine kinase 3), Myb (myeloblastosis), Runx1, and Tie2 (angiopoietin receptor) (Ginhoux et al. 2010; Schulz et al. 2012; Gomez Perdiguero et al. 2015; Hoeffel et al. 2015). While monocytes may be recruited to the neonatal and adult brain where they differentiate into microglia‐like cells, this is more likely to occur under inflammatory conditions (Simard et al. 2006; Monier et al. 2007; Verney et al. 2010; Ginhoux & Prinz, 2015) and can still be phenotypically distinguished from resident microglia cells (O'Koren et al. 2016). These findings were corroborated using a Kit^MercreMer fate mapping strategy, showing that tissue resident macrophages are derived from haematopoietic stem cells, with the exception of microglia, which represent the only resident macrophages entirely derived from YS precursors in mouse (Sheng et al. 2015).

Interestingly, a Hox8b lineage of haematopoietic mononuclear cells exists within the brain of newborn mice and persists in the adult, contributing a significant percentage of the total microglial population in regions where they are found (Chen et al. 2010). A plausible point of entry for these cells is via the brain ventricular choroid plexus (Shechter et al. 2013), since Hox8b expressing cells were observed to be distributed in a gradient from the pial surface and ventricular lining into the parenchyma during the first two postnatal weeks (Chen et al. 2010). Mice lacking Hox8b expression in microglia reportedly display obsessive‐compulsive‐like behaviour of over‐grooming (Chen et al. 2010). This homeostatic contribution of BM‐derived macrophages to the CNS remains an enigma to be clarified, especially since many fate mapping studies have shown that microglia arise from the YS exclusively. In zebrafish a pulse labelling strategy with temporal–spatial resolution challenged this single source view by showing that the adult wave of brain microglial generation does not have a YS origin, but arises from the ventral walls of the dorsal aorta instead (Xu et al. 2015).

Factors required for microglial development and homeostasis

Recent studies have focused not only on determining the origin of microglia and elucidating how these unique cells renew, but also on identifying factors which may affect their maturation, activation, proliferation and apoptosis under normal physiological conditions. Together they expand the possibilities available to control their effector functions in CNS diseases.

Cell signalling pathways required for microglial proliferation and maturation

The early development of YS microglia precursors is dependent on key transcription factors Pu.1, a member of the Ets family (Rosenbauer & Tenen, 2007) and interferon regulatory factor Irf8, which both function as heterodimers in the determination of brain macrophage phenotype (Beers et al. 2006; Minten et al. 2012; Kierdorf et al. 2013). Furthermore matrix metalloproteinases 8 and 9 are upregulated by A2 and embryonic macrophages during proliferation and invasion of the neuroectoderm (Kierdorf et al. 2013). Another transcription factor that is known to regulate the differentiation of myeloid cells is Runx1, which is expressed in a subpopulation of amoeboid microglia restricted to the ventricles during early postnatal forebrain development (Zusso et al. 2012). Runx1 has been described to mediate microglial proliferation. However by P10, the colocalisation of proliferation marker Ki67 and Runx1 was no longer observed, with microglia shifting toward a ramified morphology, suggesting its role as a maturation factor (Zusso et al. 2012). Nevertheless it is still elusive how Runx1 regulates these events in a spatiotemporal dependent manner. The microRNA miR‐124, which binds the mRNA of transcription factor C/EBPalpha and which in turn downregulates Pu.1, was identified to be specific to microglia among myeloid cell populations (Ponomarev et al. 2011). miR‐124 is functionally conserved in zebrafish and mouse in controlling the motility and phagocytic activities of microglia which transition from amoeboid to ramified forms (Ponomarev et al. 2011; Svahn et al. 2015). In the E13.5 mouse spinal cord, purinergic ionotropic receptor P2X7 additionally mediates the proliferation of embryonic microglia and consequently regulates microglial density (Rigato et al. 2012). While many macrophage and microglial‐specific transcription factors have been uncovered, the interactions between factors and signalling cascades have not been elucidated.

Another essential pathway that defines microglial cell number involves Csf‐1R (CD115) (Fig. 1). Embryos are depleted of microglia in Csf‐1R knockout mice, which also have impaired brain architecture (Erblich et al. 2011). Neuron‐derived interleukin (IL)‐34, the second ligand for Csf‐1R, is more critical for regulating microglial cell density than Csf‐1 in the adult brain; however, its requirement during perinatal development is still controversial (Greter et al. 2012; Wang et al. 2012) (Fig. 1). Mature microglia similarly require Csf‐1R signalling, as mice treated with Csf‐1R inhibitors experience an almost complete loss of microglia in adulthood (Elmore et al. 2014). Similarly, mice deficient in the Csf‐1R adaptor protein DAP12 (DNAX activation protein of 12 kDa) reportedly have reduced microglial cell numbers in adults (Otero et al. 2009), but no overt impact on their density during development (Kierdorf et al. 2013). While young DAP12‐deficient mice do not show neurological symptoms up to 4 weeks of age, synaptic function and plasticity are impaired, suggesting its importance in maintaining microglial physiology and microglia–neuron interaction (Roumier et al. 2004) as discussed below. Interestingly the decline in microglial numbers after the third postnatal week was not mitigated by upregulating brain Csf‐1 levels (Nikodemova et al. 2015), indicating the existence of alternative microglial responses to unknown developmental signals that drive the apoptosis of a population of microglial cells during CNS maturation.

Figure 1. Signalling pathways regulating microglial proliferation, maturation and homeostasis in healthy and diseased conditions .

The proper development, differentiation, proliferation and function of both microglia and neurons require a continuous and lifelong crosstalk between the two cell types via complementary ligands and receptors and secreted trophic factors (e.g. cytokines). As microglia (blue) and neurons (brown) mature, they downregulate certain signalling pathways (e.g. Cxcl12‐Cxcr4) and upregulate others (e.g. IL‐34‐Csf‐1R). Dysregulation of either ligands or receptors occurs in a number of disease processes (dysfunctional microglia, purple). Maturation of yolk sac derived microglia within the CNS is dependent on proper signalling via purinergic receptors and cell surface protein Csf‐1R. Similarly, mature microglia require functional Csf‐1R signalling for their maintenance. Microglial ‘signature’ factors recently identified include TGFβ (transforming growth factor β, which appears critical for mediating microglial survival and phenotypic differentiation in the healthy mature brain) and Tmem119 (transmembrane protein 119, a cell surface protein of unknown function). The expression of microglial markers (including IBA1, CD45, MHC class II, and others) is often changed during the course of ageing and disease, coincident with changes in microglial function. R, receptors.

Cell signalling pathways required for microglial homeostasis

Mature microglia in the postnatal brain utilise a vast number of surface molecules in order to quickly respond to their extracellular environment, and are capable of responding to cytokines, chemokines, purines, hormones and neurotransmitters, among others. Similar to other tissue resident macrophages, microglia express common markers such as the fractalkine receptor CX₃CR1, Csf‐1R, the integrin CD11b, surface glycoproteins F4/80 and CD68, ionised calcium‐binding adapter molecule 1 (IBA1), and pan‐haematopoietic CD45, but at lower levels than in perivascular macrophages and blood monocytes at steady state (Greter & Merad, 2013) (Fig. 1). High throughput gene expression studies have identified several genes distinguishing microglia from other cell types in the CNS or in the periphery (Serrats et al. 2010; Butovsky et al. 2012, 2014; Gautier et al. 2012; Chiu et al. 2013). These gene expression studies have also highlighted the phenotypic diversity of mature microglia, notably with respect to other tissue macrophages. However, whether these ‘signature’ factors directly contribute to specifying their motility, morphology and functions remains to be demonstrated (Gautier et al. 2012; Chiu et al. 2013; Hickman et al. 2013; Butovsky et al. 2014). A transmembrane protein with unknown function, Tmem119, was recently described to distinguish resident IBA1‐expressing microglia from non‐parenchymal CNS macrophages of the choroid plexus and meninges, as well as CD163⁺ perivascular cells (Bennett et al. 2016) (Fig. 1). Mannose receptor (CD206) expression was also shown to be limited to these non‐parenchymal CNS macrophages populations in normal, inflamed, injured and diseased CNS (Galea et al. 2005; Gomez‐Nicola et al. 2014). These populations of non‐parenchymal CNS myeloid cells have not been thoroughly characterised thus far and were postulated to derive from definitive haematopoiesis in contrast to YS‐derived microglia (Aguzzi et al. 2013). In addition, the simplified use of ‘M1’ classical activation and ‘M2’ alternative activation as descriptors of microglial phenotypes has fallen out of favour, especially from in vivo studies in disease models (Martinez & Gordon, 2014). Instead, much more attention is being paid to characterising the changes in whole genome expression in response to specific environmental challenges, as opposed to dualistic categorisation into ‘pro‐inflammatory’ versus ‘anti‐inflammatory’ phenotypes.

Among the factors identified so far, TGFβ (transforming growth factor β) (Fig. 1) could be critical for mediating microglial survival and phenotypic differentiation, considering that microglial density is drastically reduced in the CNS of TGFβ receptor‐deficient mice (Butovsky et al. 2014). TGFβ signalling induces microglia to adopt a ramified morphology concomitant with reduced levels of CD86, MHC class II and CD11c, and upregulation of CX₃CR1 and IBA1 in vitro (Abutbul et al. 2012). Through its interaction with microglial CX₃CR1 receptor, neuron‐derived fractalkine (CX₃CL1) is considered an OFF signal keeping microglia in a non‐activated state (Biber et al. 2007) (Fig. 1). It is necessary for the timely recruitment of microglial cells into the early postnatal hippocampus and cerebral cortex (Hoshiko et al. 2012; Ueno et al. 2013; Paolicelli et al. 2014). The loss of CX₃CR1 leads to a transient reduction of microglial numbers during the early postnatal period, and concomitant deficit in microglia‐mediated synaptic pruning, weakened synaptic transmission, and decreased functional brain connectivity (Paolicelli et al. 2011; Zhan et al. 2014), with consequences for cognitive behaviours discussed below.

Modulation of microglial process remodelling by purinergic signalling

Microglia are the most dynamic cells of the mature CNS, constantly remodelling their processes (Davalos et al. 2005; Nimmerjahn et al. 2005; Wake et al. 2009; Tremblay et al. 2010; Li et al. 2012), even following death of the organism (Dibaj et al. 2010). Several studies have focused on elucidating the mechanisms regulating microglial cell migration and process motility this past decade (Tremblay et al. 2011), revealing that purinergic signalling through microglial P2RY12 (Fig. 1) drives microglial response to laser injury in vivo (Davalos et al. 2005; Haynes et al. 2006), process remodelling in retinal explants (Fontainhas et al. 2011), as well as filopodial extension in mouse models of status epilepticus (Eyo et al. 2014) and neuropathic pain (Gu et al. 2016). In addition, dendritic neuronal NMDA receptor activation was shown to trigger release of ATP inducing the outgrowth of microglial processes in acute mouse hippocampal slices, suggesting a novel form of neuron–microglia communication that is mediated by purines (Dissing‐Olesen et al. 2014) (Fig. 1). More recently, it was found that extracellular calcium reduction induced microglial processes to converge at distinct sites, targeting neuronal dendrites independently from neuronal action potential firing in mouse brain slices and in vivo. This process is mediated by purinergic signalling through microglial P2RY12 receptors, suggesting that microglial interactions with neurons are guided by cerebral calcium reduction in the healthy brain (Eyo et al. 2015). The source of purinergic signalling, whether astrocytic or neuronal, however, remains unclear due to conflicting results between studies, which could arise from differences in the environmental contexts under investigation.

Heterogeneity of microglia in CNS milieus

Gene expression analyses have identified subtle differences in microglial expression profiles between brain regions under normal physiological conditions (de Haas et al. 2008; Doorn et al. 2015). The attraction of microglia into the ventricular/subventricular zone (SVZ) of the developing cortex is influenced by CXCL12 secreted from basal progenitors lining the ventricular zone via microglial CXCR4 (Arno et al. 2014) (Fig. 1). Genetic ablation of basal progenitors reduces microglial recruitment, while reduction of microglial numbers affects the number of neuronal progenitors in the ventricular and SVZ (Arno et al. 2014). In the white matter, it was unveiled that microglia specifically upregulate the ubiquitin‐specific protease 18, which under healthy conditions mitigates tissue destruction due to the presence of tonic interferon signal in this milieu (Goldmann et al. 2015). Microglia in the adult SVZ and rostral migratory stream also comprise a morphologically and antigenically distinct phenotype, distinguishable by lower expression of purinergic receptor P2RY12, and its lack of ATP‐driven chemotaxis (Ribeiro Xavier et al. 2015). Taken together, specific dynamic interactions between developing neural progenitor cells, neurons and microglia in each microenvironment may be required to determine the proper recruitment of microglia, as well as the maturation and maintenance of local neuronal circuits.

Differences between sexes

In addition to regional heterogeneity, sex‐related differences in microglial cells were lately uncovered. Microglial density was shown to differ between males and females across stages of the lifespan and brain regions that include the preoptic area, the hippocampus, parietal cortex and amygdala, under steady state conditions (Mouton et al. 2002; Schwarz et al. 2012; Lenz et al. 2013), and in response to chronic stress (Bollinger et al. 2015). These effects could be modulated by sex hormones, considering that isolated microglial cells express both oestrogen and progesterone receptors (Sierra et al. 2008). Microglia act in concert with the nervous and endocrine systems during sex determination in development. Inhibition of microglial activation with minocycline (which also affects astrocytes and other inflammatory mediators in a non‐specific manner) during the critical stage for sexual differentiation prevented the shift towards masculinity, defined by an increase in dendritic spine related proteins in the preoptic area (Lenz et al. 2013). Microglial involvement with pain processing, through P2X4R‐induced release of BDNF (brain‐derived neurotrophic factor) (Fig. 1), was further observed in male but not female mice, indicating a mechanistic difference between sexes (Sorge et al. 2015). This effect was mediated by testosterone, as treatment of female animals with testosterone resulted in microglia‐facilitated neuropathic pain instead of T cells‐driven processing in females (Sorge et al. 2015). Overall, the genetic or environmental factors that modulate the sex‐linked determination of microglial morphology, density and function in their respective domains are largely unclear. However, recognition of the differential regulation of microglial ontogeny prompts a critical assessment of existing data based on a particular sex.

Crosstalk with the periphery

Even though microglia reside in the apparently immune‐privileged CNS, an increasing number of studies point to an inter‐relationship between these cells and the periphery in the absence of pathology. Microglia are constantly modulated by blood to brain cytokine diffusion and transport during infection, immune‐related molecules secreted from endothelial cells forming the blood–brain barrier (BBB), and peripheral immune signals from the autonomic nervous system as reviewed in (Dilger & Johnson, 2008). In particular, the gut microbiome, which tightly regulates mood and cognition, was implicated in the normal development and maintenance of microglial cell homeostasis. Rats exposed to helminths prior to birth have different microglial tiling and reduced microglial response to early‐life immune challenges (Williamson et al. 2015). Mice bred in germ‐free conditions have increased microglial density and immaturity, with the effects regulated by short‐chain fatty acids derived from bacterial fermentation by‐products of microbiota (Erny et al. 2015). Rodents raised without gut microbiota also display increased BBB permeability and decreased expression of tight junction proteins on the endothelial cells forming their BBB (Braniste et al. 2014). It is even possible that the introduction of short‐chain fatty acids into otherwise germ‐free mice could allow signalling to peripheral splenic macrophages, which then traffic into the CNS, promoting the maturation of microglia (Mosher & Wyss‐Coray, 2015). These data demonstrate that the steady state of the CNS immune system does not only depend on the responses of microglia to their local environment, but also on a complex system of interactions between the peripheral and central nervous systems.

Physiological functions of microglia in brain development and beyond

Microglia are crucial regulators of brain development and homeostasis via neuronal–microglial interactions, synaptic modelling, scavenging of cellular debris and secretion of trophic factors. How does a single cell type multitask? Adult microglia appear uniform morphologically during steady state conditions, but they are functionally heterogeneous in their physiological responses, which may be attributed to their local environment and dependent on the ongoing neuronal activity (Li et al. 2012; Schafer et al. 2012; Arnoux et al. 2013; Clark et al. 2015).

Microglial regulation of neuronal death and survival

As phagocytes, activated microglia proliferate and accumulate in areas presenting high densities of apoptotic neurons to facilitate neuronal turnover during developmental cell death (Marin‐Teva et al. 2004; Peri & Nusslein‐Volhard, 2008; Swinnen et al. 2013). These events are very well coordinated. For instance, increased rates of cell death in neurogenic regions such as the SVZ along the lateral ventricles lead to the release of macrophage migration inhibitory factor (MIF) triggering microglial proliferation (Arno et al. 2014). When massive neuronal death is induced by neonatal alcohol exposure, microglia in the P7 cortex respond with an acute increase in TNFα (tumour necrosis factor α), IL1β, and CD68 levels and decrease of P2RY12, consistent with a pro‐inflammatory state (Ahlers et al. 2015) (Fig. 1). This probably encourages the clearance of dead neuronal cells, and subsequent deactivation occurs 48 h post ethanol injection – just after the peak of phagocytosis (Ahlers et al. 2015).

Apart from the clearance of dead cells, microglial phagocytosis controls the number of neural precursors in proliferative regions of the developing telencephalon in perinatal rats and macaques (Cunningham et al. 2013). Pharmacological interventions or maternal immune activation disrupting microglial cell numbers inversely change the pool size of neural progenitors and affect proper CNS development (Cunningham et al. 2013). However, further investigations are still required to understand the actual impact of modulating the number of neural precursor cells during early development on adult behaviour. Besides being scavengers, microglia directly support neuronal survival, neurogenesis and oligodendrogenesis in vitro (Butovsky et al. 2006; Walton et al. 2006) and in vivo during pre‐ and postnatal development (Ueno et al. 2013; Arno et al. 2014; Shigemoto‐Mogami et al. 2014). Microglia lining the subcortical white matter tracts in the early postnatal cerebral cortex support the survival of layer V neurons through secretion of the trophic factor insulin‐like growth factor 1 (Ueno et al. 2013) (Fig. 1). Activated microglia in the early postnatal SVZ support neurogenesis and oligodendrogenesis via the release of pro‐inflammatory cytokines such as IL‐1β, IL‐6, TNFα, and IFNγ (interferon γ) (Shigemoto‐Mogami et al. 2014) (Fig. 1). They are also necessary for maintaining homeostasis of the pool of basal progenitors in the forebrain throughout neurogenesis from E14 to E17 (Arno et al. 2014).

In the healthy mature brain, neurogenesis continues throughout adulthood in well‐defined areas where the existence of neural stem cells persists. The two consensus areas in rodents and human are the SVZ and hippocampal subgranular zone (SGZ) (Sierra et al. 2014). In the SGZ, microglial phagocytosis was shown to actively eliminate the excess newborn neurons that die by apoptosis (Sierra et al. 2010; reviewed in Sierra et al. 2014). In the SVZ, depletion of the microglial phenotype expressing low levels of P2RY12 hampered the survival and migration of newly generated neuroblasts to the olfactory bulb (Ribeiro Xavier et al. 2015), suggesting the importance of this unique microglial subpopulation in maintaining the adult olfactory circuitry.

Roles of microglia in neuronal wiring

Once neuronal circuits are established, microglia mainly contribute to the refinement of synaptic connections in the healthy brain (Wake et al. 2009; Tremblay et al. 2010; Bialas & Stevens, 2013). Microglial involvement with the formation of axonal tracts or engulfment of dendritic spines and/or axon terminal fragments during CNS development is conserved from Drosophila (Watts et al. 2004) to rodents (Paolicelli et al. 2011; Schafer et al. 2012; Squarzoni et al. 2014), and presumably humans, based on their topological relationship with axonal tracts described postmortem (Cho et al. 2013).

In the prenatal mouse brain, microglia regulate the wiring of forebrain circuits, controlling the outgrowth of dopaminergic axons into the forebrain and the laminar positioning of neocortical interneurons subsets (Squarzoni et al. 2014). In the postnatal brain, microglia‐mediated synaptic pruning is similarly required for the activity‐dependent wiring of neural circuits (Paolicelli et al. 2011; Schafer et al. 2012). Microglial phagocytosis of synaptic elements is documented during postnatal development, adolescence, adulthood and normal ageing, within the thalamus, cerebral cortex and hippocampus (Tremblay et al. 2010, 2012; Paolicelli et al. 2011; Schafer et al. 2012; Milior et al. 2016), suggesting a crucial role in synaptic plasticity and behavioural adaptation to the environment (Tremblay et al. 2010), notably in response to chronic stress (Milior et al. 2016). In the developing retino‐geniculate system, microglial sculpting of neural circuits was demonstrated to involve TGFβ, which modulates the expression of complement protein C1q, which in turn mediates the downstream complement protein C3 tagging of synapses (Stevens et al. 2007) for elimination via CR3/C3 signalling (Schafer et al. 2012; Bialas & Stevens, 2013) (Fig. 1). Mice lacking TGFβ receptor II in retinal neurons had reduced synaptic localization of complement proteins, as well as impaired microglial engulfment of retinal ganglion cell inputs (Bialas & Stevens, 2013).

During adolescence and adulthood, depleting microglia using transgenic expression of the diphtheria toxin (Parkhurst et al. 2013) or treatment with a Csf‐1R inhibitor (Elmore et al. 2014) also increased synaptic density in the hippocampus (Rice et al. 2015). It was also shown using a transgenic strategy (CX₃CR1^CreER crossed with BDNF^flox) that preventing microglial release of BDNF decreases the formation of cortical dendritic spines in vivo during motor learning (Parkhurst et al. 2013) (Fig. 1). Similar findings of a decreased spinogenesis, paralleling the impairment of synaptic elimination, were obtained upon microglial depletion achieved by crossing the CX₃CR1^CreER model with diphtheria toxin receptor (DTR)‐inducible R26^iDTR mice (Parkhurst et al. 2013). These observations support the notion that microglia are important effectors of structural plasticity in the healthy mature brain, mediating both the formation and elimination of synaptic elements.

Microglial regulation of neuronal activity and synaptic plasticity

Healthy microglia additionally produce a broad spectrum of signalling molecules, from cytokines to neurotransmitters and extracellular matrix proteins which are capable of regulating neuronal activity, as well as synaptic activity and functional plasticity (Bessis et al. 2007; Bechade et al. 2013; Ji et al. 2013). In the larval zebrafish, dynamic interactions between microglial processes and neuronal cells bodies regulate neuronal activity, reducing both spontaneous and visually evoked activities in vivo (Li et al. 2012). In rodents, several ex vivo studies in hippocampal slices revealed various alterations of synaptic maturation, activity and functional plasticity using knockout mice where microglia‐specific signalling pathways are compromised (Wu et al. 2015). For instance, fractalkine signalling deficiency in CX₃CR1 knockout mice impaired the functional maturation of thalamo‐cortical synapses during the first postnatal week (Hoshiko et al. 2012). Loss‐of‐function mutation or deletion of microglial DAP12, CD200R or CX₃CR1 similarly resulted in reduced or enhanced hippocampal long‐term potentiation (LTP), the most commonly studied paradigm of synaptic activity and plasticity, during adulthood (Roumier et al. 2004, 2008; Maggi et al. 2009; Costello et al. 2011; Rogers et al. 2011) (Fig. 1). CD200R interacts with the glycoprotein CD200 expressed by neurons, oligodendrocytes and astrocytes (Costello et al. 2011). Several studies also revealed modulatory effects of fractalkine on the electrophysiological properties of excitatory synapses, involving NMDA and AMPA receptors, as well as d‐serine, in acute hippocampal slices from juvenile and adult mice (reviewed in Paolicelli et al. 2014). Nevertheless microglial involvement and the downstream pathways related to these effects on neurotransmission remain to be investigated.

Consequences of microglial dysfunction on cognition

Given the plurality of microglial functions, conditions resulting in defective microglia during brain development can lead to impaired clearance of cellular debris and compromised neural connectivity. Dysfunctional or perturbed microglial homeostasis could have direct consequences on the onset of severe neurodegenerative or neuropsychiatric disorders at an early age or later in adulthood. For example, young boys deficient for the ALD protein encoded by ABCD1 gene suffer from a debilitating brain demyelinating disease known as X‐linked adrenoleukodystrophy, which could be halted by partial reconstitution of myeloid and lymphoid populations via gene therapy (Cartier et al. 2009) Previously, glial cells in MECP2 mutants lacking the methyl‐CpG‐binding protein were directly implicated in Rett syndrome, an X‐linked autism spectrum disorder, due to neuronal dysfunction (Maezawa & Jin, 2010; Lioy et al. 2011; Derecki et al. 2012). Derecki et al. (2012) proposed that BM transplantation of wild‐type ‘microglia’ could ameliorate the disease symptoms. However, a collaborative investigation using several mouse models of Rett could not replicate the amelioration of motor deficits and breathing abnormalities, as well as extended lifespan, leading to the termination of clinical trials for BM transplants in children with Rett syndrome (Wang et al. 2015).

During adulthood, impaired microglial remodelling of neuronal circuits could severely impair learning and memory functions. Fractalkine signalling deficiency in CX₃CR1 knockout mice, microglial‐BDNF deletion and microglial depletion similarly resulted in cognitive impairment, when assessing motor behaviour, and fear conditioning where the amygdala plays a critical role (Rogers et al. 2011; Parkhurst et al. 2013) (Fig. 1). The CX₃CR1 knockout mice also showed altered performance in the Morris water maze testing hippocampus‐dependent learning and memory functions (Maggi et al. 2011; Rogers et al. 2011). Nevertheless, no significant deficits of novel object recognition, a paradigm which depends on the prefrontal cortex, were detected in these models, suggesting the recruitment of different mechanisms (Parkhurst et al. 2013; Zhan et al. 2014). In addition, 21 days of microglial depletion with the Csf‐1R inhibitor did not induce cognitive deficit under various behavioural schemes (Elmore et al. 2014), indicating that longer treatment periods might be required to modify the behaviour. Indeed, microglial depletion using clodronate or oral administration of a Csf‐1R inhibitor only transiently impaired spatial memory without modifying sociability (Torres et al. 2016).

Dysregulation by the environment

Exposure to environmental risk factors (such as air pollution, omega‐3 deficiency, chronic psychological stress, infection, etc.) has profound effects on microglia and the inflammatory milieu in the brain, as well as cognition (Bilbo, 2013; Castanon et al. 2015). These effects are particularly pronounced when exposure occurs during critical windows of development (Knuesel et al. 2014). The underlying mechanisms need to be further investigated, especially in the context of neuropsychiatric disorders, but fractalkine signalling is emerging as an important mediator of microglial response to the environment (Maggi et al. 2011; Milior et al. 2016). In particular, the CX₃CR1 knockout mice display social interaction deficits associated with autism spectrum disorders in humans, both early in life and adulthood (Zhan et al. 2014) (Fig. 1). CX₃CR1 knockout mice exposed to an enriched environment fail to enhance hippocampal LTP and to improve learning performance in the Morris water maze (Maggi et al. 2011). CX₃CR1 knockout mice are also resistant to anxiety‐like behaviour (Wohleb et al. 2014) and microglial alterations (Hellwig et al. 2015; Milior et al. 2016) induced by chronic stress. In particular, microglial phagocytosis of synaptic elements was implicated in the adaptation of the brain and behaviour to chronic stress (Milior et al. 2016). In addition, blockade of stress‐induced microglial activation by minocycline or IL‐1 receptor antagonist overexpression prevents microglial apoptosis and decline upon chronic stress and the emergence of cognitive impairment (Hinwood et al. 2012; Kreisel et al. 2014).

Microglial activation, senescence and priming in normal ageing

Two compelling studies based on heterochronic parabionts (i.e. young and aged mice with linked circulatory systems) claimed that young blood, or its derived factor, was able to reverse cognitive impairment, improve synaptic plasticity, enhance brain vasculature, and increase neurogenesis in aged mouse (Katsimpardi et al. 2014; Villeda et al. 2014). Is ageing one of the greatest risk factors for neurodegenerative diseases? What is the physiological relevance of microglia in essential coping mechanisms during late adulthood and normal ageing? We explore these questions by examining the various microglial phenotypes observed in the aged CNS and their relevance to healthy ageing.

Microglial activation

In the absence of pathology, aged microglia have higher expression of pro‐inflammatory genes and antigen presenting markers such as complement components, TLR (Toll‐like receptor) signalling, inflammasomes, scavenger and MHC class II antigens; increased production of pro‐inflammatory cytokines and reactive oxygen species (ROS); while anti‐inflammatory cytokines including IL‐10 and TGFβ1, and microglial activation inhibitory factors such as CD200 and fractalkine receptors are down‐regulated (Mosher & Wyss‐Coray, 2014) (Fig. 1). These parameters indicate that as humans and rodents age microglia take on a phenotype consistent with a pro‐inflammatory response, even in the absence of overt pathological stimuli (Luo et al. 2010; Choi & Won, 2011; Wong, 2013; Sierra et al. 2014; Ojo et al. 2015; Barrientos et al. 2015 a). These findings were confirmed with positron emission tomography in aged individuals (>50 years old) showing increased detection of the radiotracer (R)‐[11C]PK11195, a marker of translocator protein (TSPO) mainly expressed on the outer mitochondrial membrane of activated microglia (Schuitemaker et al. 2012; Wehrspaun et al. 2015). Microarray gene expression analyses also revealed that surface receptors mediating neuronal–microglial crosstalk and anti‐inflammatory genes are reduced in aged individuals (Schuitemaker et al. 2012; Wehrspaun et al. 2015). Although another group showed that the microglial activation might be related to TSPO polymorphisms rather than ageing (Suridjan et al. 2014), TSPO is still considered an important regulator of microglial activation and phagocytosis through macroglia–microglia interaction (Karlstetter et al. 2014; Wang et al. 2014).

Microglial senescence

Despite their increase in pro‐inflammatory signalling, aged microglia are often regarded as ‘dystrophic’ or ‘senescent’ as they lose the ability to respond adequately to external stimuli and injury (Streit, 2006; Wasserman et al. 2008). Dystrophic microglia were previously identified in aged human brain in situ (Streit et al. 2004, 2009) and in vitro (Caldeira et al. 2014). One study proposed that following stress‐induced depressive‐like condition, dystrophic microglia (including apoptosis, reduction in their numbers within the hippocampus and reduced expression of activation markers) might contribute to reduced hippocampal neurogenesis and increased depressive‐like behaviour in rodents (Kreisel et al. 2014). Compared to 3‐month old mice, aged microglia in 2‐year old mice show different levels of morphological changes, including ∼25% increase in density, ∼50% enlargement in area of soma, shorter and thicker processes with ∼20% reduction in process motility, ∼60% increase in soma motility, and accumulation of cellular debris (Tremblay et al. 2012; Hefendehl et al. 2014). Microglial activation, process motility, and migratory velocity were also diminished in aged retina and cerebral cortex in response to injury (Damani et al. 2011; Hefendehl et al. 2014).

Microglial priming

Microglial priming is widely observed in aged brains from humans, non‐human primates and rodents (Streit & Sparks, 1997; Sheffield & Berman, 1998; Godbout et al. 2005; Frank et al. 2006; VanGuilder et al. 2011). Microglial priming with ageing may additionally result from the accumulation of misfolded and aggregated proteins, such as amyloid‐β and mutant huntingtin, in Alzheimer's and Huntington's diseases, respectively (reviewed in Cartier et al. 2014). This age‐related priming induces an exaggerated inflammatory response to a secondary stimulus (Cunningham, 2013; Perry & Holmes, 2014; Matt & Johnson, 2015). According to early human studies, elderly individuals were more susceptible to peripheral infections, and these infections increased mortality rates for individuals over 65 years of age (Pinner et al. 1996). Extensive studies performed on rodents have produced data that support the hypothesis of higher susceptibility to neurodegeneration due to hyperactivation of microglia during pathological ageing. Peripherally delivered lipopolysaccharide (LPS) stimulation exaggerated gene and protein expressions of pro‐inflammatory cytokines, triggered oxidative stress, and further reduced social behaviour and locomotor activity in aged rodents (Godbout et al. 2005; Sierra et al. 2007; Matt & Johnson, 2015; Barrientos et al. 2015 b). More importantly, prolonged activation can make microglia resistant to regulation, impairing their responses to IL‐10, TGFβ1 and IL‐4, and become irresponsive to neurons signalling through CX₃CL1‐CX₃CR1 and CD200‐CD200R. These changes could disrupt the fine coordination between the immune and nervous systems (Barrientos et al. 2010, 2015 a; Cunningham, 2013; Norden & Godbout, 2013; Wong, 2013; Biber et al. 2014; Perry & Holmes, 2014; Ojo et al. 2015).

Dark microglia and consequences on cognition

In addition, we recently described a new ‘dark’ microglial phenotype in mouse hippocampus, cerebral cortex, amygdala and hypothalamus that is prevalent during chronic stress, fractalkine signalling deficiency (CX₃CR1 knockout model), normal ageing, and Alzheimer's disease pathology (Bisht et al. 2016). Dark microglia exhibit several signs of oxidative stress, including a condensed, electron‐dense cytoplasm and nucleoplasm (giving them a ‘dark’ appearance similar to mitochondria), mitochondrial disruption, and endoplasmic reticulum dilatation (i.e. the best characterised sign of oxidative stress at the ultrastructural level). These cells appear to be extremely active, frequently reaching for synaptic clefts, while extensively encircling axon terminals, dendritic spines and entire synapses with their highly ramified and thin processes. They also express CD11b (composing CR3 involved in synaptic pruning) in their processes encircling synaptic elements (Bisht et al. 2016). Our observations suggest that dark microglia could be implicated in the loss of synapses, which is the best pathological correlate of cognitive decline across ageing and several diseases, including the highly prevalent Alzheimer's disease. In particular, dark microglia could represent a subset of hyperactive microglia that become stressed as a result of their hyperactivity under adaptive pressure, leading to dysregulated interactions with synapses. Considering that chronic stress accelerates cellular ageing, potentiates oxidative stress and neuroinflammation, and predisposes to various diseases across the lifespan, therapeutic interventions aimed at promoting microglial resilience to stress could be crucial for promoting healthy ageing.

Overall, these findings indicate that normal ageing can lead to various microglial abnormalities ultimately affecting their physiological roles in brain plasticity and cognition. While it is still unclear whether microglia are the primary players in neurodegenerative diseases, changes in aged microglia should be considered to provide insights into the development of diagnosis and therapeutic approaches for such diseases (Siskova & Tremblay, 2013).

Summary and outlook

Over the past decade microglia have emerged as important contributors to normal brain physiology. As a consequence, the field recently experienced an explosive growth. Within this context, our review aims at providing a critical update on their unique origin, the factors regulating their maturation and homeostasis, their physiological roles in brain development, function and plasticity, and the implications of their loss or dysfunction to cognitive decline during normal ageing and in diseases. Since microglia actively maintain health during normal physiological conditions, it is of prime importance to elucidate the cellular, molecular and epigenetic mechanisms that determine their phenotype and underlie their motility, phagocytic behaviour, synaptic interactions and release of various mediators modulating cognition. The development of effective treatments requires better understanding of those mechanisms governing microglial effector functions across the lifespan.

Additional information

Competing interests

The authors declare no competing financial interests.

Biography

Marie‐Ève Tremblay has been an assistant professor at Université Laval since 2013. Her research focuses on elucidating the roles of microglia in the loss of synapses which best correlate with the impairment of learning and memory across chronic stress, ageing and various diseases. Her long‐term goal is to help develop new therapies using myeloid cells as vectors for effecting targeted changes in neuronal circuits, to promote stress resilience and healthy ageing, and to improve the cognitive functions in individuals suffering from post‐traumatic stress disorders, and various neuropsychiatric and neurodegenerative diseases.