Autism spectrum disorders (ASD) include a range of neurodevelopmental disorders, commonly characterized by repetitive behaviours, as well as impaired social skills, including verbal and nonverbal communication (1). These behavioral symptoms develop in early childhood and persist throughout life. In recent decades, there has been a major surge in ASD incidence globally (2). Although the precise aetiologies of ASD are complex, and remain to be fully understood, recent evidence points to abnormal synaptic development and function, and/or aberrant immune responses, as potential drivers of ASD symptoms (3–6). Notably, microglial cells participate in these physiological processes and have been strongly associated with ASD development (7–10).

Microglia are the main resident immune cells of the central nervous system (CNS), providing the tissue with innate immune sensing, inflammatory effector functions and tissue repair. As such, they are the main producers of proinflammatory mediators in the context of neuroinflammation (11). Although immunomodulatory roles for microglia in neuroinflammatory and neurodegenerative diseases have been widely described, immune modulation is only one of an extensive array of discrete microglial functions. During CNS development, microglia regulate the number and strategic positioning of neurons and shape neuronal connectivity (10, 12). Moreover, they support gliogenesis and myelination (10, 12–15). Given both their immune and developmental functions, it would be attractive to propose that microglial dysfunction could contribute to neurodevelopmental disorders; either by influencing disease development or driving behavioral symptoms. However, the specific roles that microglial cells play in ASD pathophysiology are still controversial. Although several studies show that autistic individuals suffer from ongoing neuroinflammatory processes, characterized by microglial activation in several discrete regions of the brain (16–19), others dispute the significance of this and suggest that microglia may be intrinsically dysfunctional in their resting state, following a prenatal disruption to homeostatic brain development (20, 21). In this review, we explore both well-established and emerging literature and discuss perspectives on the role’s microglia may play in the development of ASD; both in the context of abnormal immune signaling and altered neuronal connectivity. Given the vast array of peripheral factors that can modulate microglial maturation and function, we further discuss how perturbations in these extrinsic signals, particularly the gut microbiota, might promote microglial dysfunction in the context of the neurodevelopmental disorders.

Microglia are a highly specialised population of myeloid cells that inhabit the healthy CNS parenchyma, representing 5–12% of all cells in the CNS (22). Unlike the other cell types that coinhabit the CNS, microglia are not derived from the neuroectodermal germ layer. Rather, microglial ontogeny has been traced to erythromyeloid precursors, which differentiate into microglial progenitors in the yolk sac during embryogenesis (23, 24). At this stage, differentiation is critically controlled by the transcription factors Pu.1 and Irf8 with other transcription factors, such Runx1 and Jun, also providing a supporting role (21, 23, 24). On day 9.5 after conception (E9.5), microglial progenitors leave the yolk sac to seed the developing CNS in one single wave (22–24). Following an initial burst of proliferation and differentiation, mature microglia then colonize the parenchyma where they persist throughout the life of the host ( Figure 1 ). There, within the healthy CNS, microglial numbers are maintained by gradual self-renewal, independently from the recruitment of any other hematopoietic myeloid cells or progenitors (11, 24, 25).

Within the CNS parenchyma, microglia are imprinted by local environmental cues. Microglial differentiation and maintenance are strongly dependent on their expression of colony-stimulating factor 1 receptor (CSF1R), as well as the two main CSF1R ligands, CSF1 and IL-34 ( Figure 1 ). Depleting either of these ligands reduces microglial cell abundance throughout the CNS. Moreover, the CNS of adult CSF1R-deficient mice are virtually devoid of all microglia (26, 27). By driving a microglia-specific gene signature, TGFβ signaling has recently also been shown to be indispensable for microglia maturation. The marker genes induced by TGFβ, which include Tmem119, Sall1, Tgfbr1, and P2ry12, can readily distinguish microglia from bone marrow-derived macrophages (28).

As tissue-resident macrophages, microglial cells are responsible for the continuous immunosurveillance of the CNS. Inflammatory insults induced by invading pathogens or local injuries trigger their production of immune mediators (29–31). These pathways also facilitate increased phagocytosis of cellular debris and/or pathogens (32). Previously, microglia were thought to be inactive during homeostasis and only activated in response to pathological insults. However, in addition to their “canonical” innate immune functions, recent findings suggest that microglia are intimately involved in CNS development through organising neuronal patterning and fine-tuning synaptic connections (13, 22).

During embryogenesis, microglia are the first glial cells to populate the developing CNS. In this early neurodevelopmental phase, they control neurogenesis by releasing neurotoxic or neurotrophic factors that orchestrate the survival, differentiation or apoptosis of neuronal progenitors (33–35). The survival-enhancing role of microglia is supported by findings showing that proliferation and survival of these progenitors is higher when they are co-cultured with microglia than when cultured alone (34). On the other hand, microglial respiratory bursts generate superoxide ions, which trigger the apoptosis of Purkinje cells in the postnatal cerebellum (33). Thus, through the selective release of neurotoxic or neurotrophic factors, microglia can shape the neuronal landscape.

In addition to modulating neurogenesis, microglia play important roles in the development and differentiation of neuronal circuits. From an early stage in postnatal neurodevelopment, microglia eliminate redundant neurons that do not establish functional circuits. Moreover, microglia modulate immature neuronal circuits by engulfing and eliminating dendritic spines at the synapse (10). This process, known as synaptic pruning, is critically important for the normal formation of synapses. Its disruption results in several neuronal abnormalities; including impaired functional connectivity, modifications to dopaminergic circuits, and an imbalance of the excitation-to-inhibition ratio in the cortex (9, 10, 36). Importantly, abnormal synaptic pruning in the CNS of the neonate, or even the developing fetus, could be important in the aetiology of ASD, as discussed later.

Finally, there is now cumulating evidence that microglial cells modulate synaptic plasticity, and subsequently, learning and memory (37–39). This is not only important during early developmental stages as depleting microglia from the CNS of adult mice also results in impaired synaptic plasticity and deficits in learning and memory (13, 40–42). Similar phenotypes are observed when microglia are unable to produce brain-derived neurotrophic factor (BDNF), as shown using conditional and inducible BDNF depletion under the CX₃CR1 promotor (42)

Thus, although microglia have several well-defined roles in neuroinflammation, it is becoming increasingly evident that they also shape neuronal survival and connectivity during development, interpret changes in the local milieu and modulate circuit formation accordingly (11, 43).

To date, microglial cell participation in ASD and other neurodevelopmental disorders has been only speculated. While the causes of ASD are incompletely understood, some of the main symptoms, such as impairment in multisensory processing and integration, have been linked to defects in neurogenesis and the strategic positioning of neurons during CNS development, abnormal synaptic pruning and an altered neuronal excitation/inhibition ratio (44). Additionally, systemic and central inflammation may also be intrinsically involved in the pathogenesis of ASD and several other neurological disorders (45, 46). Considering both the physiological roles microglia play in regulating neurogenesis, neuronal migration and synaptic pruning, and their immunomodulatory roles in the CNS, it seems entirely plausible that aberrant microglial function may be a driving force in the pathogenesis of ASD.

Vargas et al. were the first to show an inflammatory phenotype in post-mortem brains from ASD individuals. In this pioneer work, neuropathologic analysis showed increased microglial activation, characterized by elevated expression of MHC class II, throughout the cerebral and cerebellar cortices in individuals with ASD. Moreover, increased expression of pro- and anti-inflammatory factors, such as such as CCL2, IL-6 and TGF-β, were observed in both the brain and cerebrospinal fluid (CSF) (16). Similar studies have reported increased expression of TNF-α, IL-6, IL-8, GM-CSF, and more recently, IL-18 and IL-37, in post-mortem brain tissue and CSF of children with ASD, suggesting a heightened immune response with associated localized brain inflammation (47–49). Consistent with this apparent microglial and astrocyte immune dysregulation, genome-wide analysis of brain tissue from ASD individuals showed enrichment of markers related to activated microglia and expression of genes associated with “immune and inflammatory” gene ontology categories, compared to neurotypical controls (50, 51). These changes in microglial activation markers in ASD brains were also accompanied by changes in microglial morphology, density and spatial localization (18, 52, 53). Not only do microglia have an increased density throughout the cerebral and cerebellar cortices of ASD patients, but they exhibit cell body enlargement, as well as process retraction and thickening. Filopodia also extend from the processes of ASD-associated microglia (18, 54). The putative microglial dysfunction detected in post-mortem samples has now been further confirmed using In Vivo Positron Emission Tomography (PET). In this study, which focused on young adults with ASD, a radiotracer specific for activated microglia and astrocytes was used to show a marked activation of these cells in several discrete regions of the brain (17).

The combination of neuropathological analyses of post-mortem human brain samples and PET scanning of live human ASD patients has provided compelling evidence to suggest that aberrant microglia and astrocyte immune activation is a common hallmark of ASD. However, due to the small number of samples evaluated, variations in genetic backgrounds, lifestyle choices, medication use and socioeconomic status, more studies are required. In the case of post-mortem studies, the cause of death could also impact brain inflammation. Moreover, it remains to be established whether microglial activation is a secondary effect of aberrant brain development or whether microglia play a causative role in the initiation or manifestation of ASD. For that reason, environmental and genetic rodent models are widely employed to explore the range of contributions microglia make to ASD pathogenesis, including their effects on neuronal migration, neurotransmission, brain anatomy and inflammation.

In rodents, genetic manipulation of microglia can profoundly alter CNS function, culminating in behavioral abnormalities resembling those found in ASD. For example, mice lacking the gene encoding CX₃CR1 exhibit ASD-like behaviours, including social deficits (9, 10). Also known as the fractalkine receptor, CX₃CR1 is a chemokine receptor that facilitates direct contact between microglial cells and CX₃CL1 (fractalkine)-expressing neurons; an interaction known to suppress microglial cell activation and IL-1β production following peripheral immune stimulation (55). Signaling through the fractalkine/CX₃CR1 axis is required for the optimal recruitment of microglia to specific CNS locations during embryogenesis (56). As such, Cx3cr1-deficient mice have fewer microglia present in the CNS during early postnatal development, resulting in altered synaptic pruning and subsequent deficits in neuronal connectivity throughout life (9, 10). Similar to CX₃CR1, microglial expression of immunoglobulin superfamily-member, triggering receptor expressed on myeloid cells 2 (TREM2), is fundamental for synaptic pruning during prenatal neurodevelopment (57). TREM2 signalling transduction has a central role in promoting microglial activation (11) and variants in TREM2 have been linked to different types of neurological diseases, including multiple sclerosis, Parkinson’s and Alzheimer’s diseases (58–62). Recent studies in mice show that the absence of this receptor results in defective remodelling of neuronal synapses, dysregulated excitatory/inhibitory neurotransmission, impaired neuronal connectivity and behavioral defects reminiscent of ASD (57). The expression of TREM2 was also significantly reduced in post-mortem brain tissue from individuals with ASD compared to neurotypical controls. This ASD-associated reduction in TREM2 expression was most prominent in samples collected from patients with severe symptoms, showing a negative correlation between TREM2 levels and ASD severity (57).

Recent studies have also shown that elevating protein synthesis, induced exclusively in microglia via overexpression of the translation initiation factor eIF4E, is sufficient to impair synaptic formation and drive the manifestation of ASD-like behaviors in young mice (63). Indeed, mutations that inactivate negative regulators of translation, such as in PTEN (phosphatase and tensin homolog), TSC1/2 (tuberous sclerosis complex 1/2), and FMR1 (fragile X mental retardation protein), are thought to cause ASD in a proportion of patients (64–68). Xu and collaborators suggested that defects in these ubiquitously expressed genes can alter microglial cell function sufficiently to drive ASD. In addition to an increased phagocytic potential, these microglia exhibit reduced mobility and impaired synaptic pruning, culminating in higher synapse density and higher excitatory neurotransmission compared to wild type mice, ultimately driving the development of ASD-like behaviors (63). Similar to the phenotype observed in mice that overexpressed eIF4E, the frequency, phenotype and function of microglia in the prefrontal cortex, hippocampus and striatum of Pten-deficient mice was substantially altered when compared with their wildtype littermates (63, 69, 70). Collectively, these studies add further weight to the hypothesis that aberrant microglial cell functions may help to drive the pathophysiology and behavioural symptoms associated with ASD.

Other models of autism in which risk genes are depleted in rodents to model symptomatic ASD variants, such as Rett syndrome and fragile X syndrome, are also related to microglial-dependent synaptic modulation. In Fmr1-deficient mice, a model of fragile X syndrome, microglia are increased in terms of size and abundance compared to those from wild-type littermates, and these physiological changes were associated with reduced microglial-mediated synaptic pruning (63, 71). As it is caused by loss-of-function mutations in the gene encoding methyl-CpG binding protein 2 (MECP2), Rett syndrome is modeled using variations of Mecp2-deficient mice. The specific deletion of Mecp2 in murine microglial cells triggers an overproduction of glutamate, altering neuronal morphology and impeding the formation of synapses (72). Abnormal microglia–synapse interactions, and increased expression of inflammatory genes in macrophages and microglia, were also observed in mice lacking Mecp2 (13, 73). These studies imply a pathological role for microglial cell dysfunction in ASD, apparently without the context of neuroinflammation. However, since both resting and activated microglia are able to secrete cytokines, neurotoxic and neurotrophic factors, as well as other soluble factors that have been implicated in ASD, it is possible that microglia use these mediators to influence a diverse range of neuronal functions and sculpt synaptic connections (19, 20, 74).

In addition to genetic factors, environmental factors also modify microglia function, affecting brain development, synaptic connectivity and CNS immune responses (75). Indeed, the behavioural abnormalities that are observed in mouse models in which environmental risk factors are the driving forces behind ASD development, such as the maternal immune activation (MIA) model, are similar to those induced by genetic modification (75, 76). The MIA model, in which pregnant mice are challenged with polyinosinic-polycytidylic acid [Poly(I:C)] or lipopolysaccharide (LPS) during embryonic development (E9–12), was developed based on numerous epidemiological studies that have linked prenatal infection in humans to the development of several neurological disorders, including ASD, in the offspring (13, 77).

MIA remotely triggers the induction of multiple cytokines in the fetal brain of rodents, subsequently leading to abnormal neurodevelopment. Due to their rapid ability to respond to inflammatory signals and their role in modulating neuronal function and connectivity, microglial cells have been implicated in driving this disorder (78). However, contradicting results cloud the ability to draw clear conclusions on how and when MIA shapes microglial functions in the developing offspring. While some studies showed increased expression of microglial activation markers in adult offspring exposed to MIA in utero, others did not uncover any postnatal differences in microglial phenotype as a consequence of MIA. The most consistent microglial changes were found during pre- or perinatal developmental stages, suggesting that transient perturbations in microglial function might have life-long effects on neuronal patterning, functional connectivity and behaviour (79–82). Supporting this hypothesis, studies using genome-wide chromatin accessibility assays revealed a series of temporally distinct developmental stages, both pre- and perinatal, during which the susceptibility of microglia to immune mediators and other environmental cues was increased (21, 24). Microglia from newborns exposed to maternal immune activation showed an untimely downregulation of genes that are typically expressed during this early stage of development, such as Spi1 (the gene encoding Pu.1) and Irf8, and instead exhibited a transcriptional phenotype more akin to that of adult microglia (21, 24). Although these alterations were transient, the authors suggested that accelerated microglial maturation could have sufficient detrimental consequences in the developing brain to induce and maintain neurological disorders that continue long after the microglia phenotype is restored (21).

Together, these data strengthen the notion that microglia can play a fundamental role in driving neurodevelopmental disorders, including ASD, via their effects on neuroimmune pathways, synaptic remodelling, neuronal survival and connectivity. Understanding the main factors that induce microglial dysfunction and identifying developmental timepoints when the CNS is most susceptible to the impacts of microglial dysregulation would help to identify novel therapeutic targets and prophylactic strategies to better treat or prevent ASD.

Although genetic factors can majorly influence the risk of ASD development, epidemiological and preclinical studies estimate that 50% of ASD pathophysiology is driven by non-heritable factors, suggesting that environmental factors may play an equally prominent role (83, 84). However, while progress has been made towards gaining an understanding of the genetic components that drive ASD, environmental risk factors are less understood. Several recent studies suggest that prenatal, perinatal and postnatal factors act synergistically to induce the development of ASD (85). Maternal diet and lifestyle, as well as exposure to infection, environmental chemicals and drugs during critical periods of CNS development, can induce various congenital malformations, culminating in a latent and long-term impact on brain function, and enhancing the risk of ASD development in the offspring (86–90). Many environmental components that are vertically transmitted via mother-to-child interactions can influence brain development during peri- and postnatal periods, whilst horizontally transferred external factors, i.e. those that are not dependent on the maternal interface, have the capacity to interfere with the maintenance and progression of ASD symptoms (91, 92).

We are still unravelling the complex tri-directional relationships linking the microbiota with microglial function and ASD development. Here we describe recent evidence implicating microglia in ASD development, and discuss how environmental risk factors, particularly gut dysbiosis, could compromise the immunological and neurological functions of microglia to drive permanent changes in the brain. We have also highlighted how perturbations in the gut microbiota during prenatal and neonatal periods, induced by antibiotics, dietary changes or infections, could compromise microglial function, thus altering brain function and increasing the risk of ASD. Recent studies have begun to clarify the significant influence the gut microbiota has on microglial phenotype during steady-state, and in numerous models of neurological disorders. However, more research is required to identify the precise mechanisms by which microglia and the gut microbiota collude to drive neurodevelopmental disorders, particularly in humans. We hope that future studies, using metabolomics assays and advanced next-generation sequencing platforms, will reveal specific microbial communities or molecules associated with ASD pathogenesis or alleviating symptoms, as well as the precise molecular mechanisms involved. This could pave the way for the identification of novel treatment targets and/or the rational design of probiotics to treat or prevent ASD.

All authors listed have made a substantial, direct, and intellectual contribution to the work, and approved it for publication.

MD-F is supported by a Beverley Phillips Postdoctoral Fellowship and a Cumming School of Medicine Postdoctoral Fellowship.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.