Astrocyte regulation of critical period plasticity across neural circuits

Abstract

Critical periods are brief windows of heightened neural circuit plasticity that allow circuits to permanently reset their structure and function to facilitate robust organismal behavior. Understanding the cellular and molecular mechanisms that instruct critical period timing is of broad clinical interest, as altered developmental plasticity is linked to multiple neuro-developmental disorders. While intrinsic, neuronal mechanisms shape both neural circuit remodeling and critical period timing, recent data indicate that signaling from astrocytes and surrounding glia can both promote and limit critical period plasticity. In this short review, we discuss recent breakthroughs in our understanding of astrocytes in critical period plasticity and highlight pioneering work in Drosophila.

Introduction

During development, neural circuits assemble from a variety of distinct neuronal types that are functionally connected by synapses. Neural circuit function is regulated by glia, a heterogeneous group of non-neuronal cells that metabolically support neurons, instruct synaptic transmission, and regulate neural circuit remodeling. This remodeling, or plasticity, is enriched during developmental windows termed “critical periods” (CPs) [1,2]. During CPs, environmental experience drives activity-dependent changes in neural circuit architecture to produce long-term effects in circuit function and behavior [1–10]. For this reason, altered CP timing is thought to underlie neuro-development disorders like epilepsy and autism spectrum disorders. Thus, understanding the mechanisms that control CP timing is of clear clinical importance [11,12].

CPs were first described by Hubel and Wiesel following a series of seminal experiments characterizing ocular dominance in cats [13,14]. Over the past 70 years, ocular dominance plasticity has been extensively studied in mammalian models. Using these systems, it was discovered that the maturation of inhibitory circuits is a key driver of CP timing [15,16]. Interestingly, the speed and duration of CPs differs across brain regions [17], leading to the question: what intrinsic or extrinsic factors might influence inhibitory circuit maturation and in turn, CP timing? These key open questions have recently sparked interest in defining CPs in model systems that are more amenable to rapid forward and reverse genetics. One non-mammalian model organism that has proved valuable to the CP field is Drosophila melanogaster [7,9,10,18] (see Figure 1). The shortened lifespan of Drosophila, expansive toolkit for genetic and optogenetic manipulation of small neuronal subsets, and robust behavioral readouts allow for direct correlation between CP plasticity and mature circuit structure/function. Moreover, Drosophila is amenable to rapid genetic screens to identify molecular mechanisms that dictate CP timing. In this review, we will summarize several validated CP models in the fly and discuss vertebrate CPs that inform our understanding of invertebrate circuits. We will also explore how neuron-glia communication influences the execution and expression of distinct CPs.

Figure 1. Critical Periods in Fly Development.

(a) Overview of defined Drosophila critical periods, in adult and larvae. (b) The olfactory critical period begins day 1 after eclosion (adult hatching) and ends by day 4 after eclosion. Intense exposure to various odorants causes activity-dependent changes to olfactory sensory neurons (OSN) including expression of olfactory receptors and a reduction of OSN synapses with projection neurons located in glomeruli of the antennal lobe. Ensheathing glia utilize a Draper-mediated signaling cascade for experience-dependent phagocytosis of OSN synapses during the critical period. (c) Developmental plasticity of the larval visual system reaches its peak during the second instar larval stage, but the timing of this critical period is not well-defined. Activity-dependent remodeling of ventral lateral neuron dendrites occurs following photomanipulation of the Bolwig organ. Increased light exposure causes a decrease of dendrite dynamicity in ventral lateral neurons. Conversely, a decrease in light exposure causes higher amounts of dendrite dynamicity in ventral lateral neurons. Ventral lateral neuron dendrite outgrowth is regulated in part by astrocyte-derived GLaz. Whether this signaling is also present during a critical period is not understood. (d) The motor circuit critical period begins at embryonic stage 17 and ends at 8 hours after larval hatching. Using optogenetics to activate motor neurons during this critical period causes a decrease in dendrite volume and excitatory synapse number. Silencing motor neurons via optogenetics results in an increase to dendrite volume and excitatory synapse number along with a decrease to inhibitory inputs. Astrocytes suppress critical period plasticity by signaling to motor neurons through nlg2 & 4, through GABA turnover (gat), and by modulating the extracellular matrix (chpf).

Established critical periods in sensory circuits

Olfaction

CPs of plasticity are well-defined in the olfactory circuits of multiple animal models, including mice and Drosophila [3,19,20]. CPs of plasticity encompass changes to innervation patterns, synapse number, and ultimately, neural activity [15]. In addition, activity-dependent regulation of olfactory sensory neuron (OSN) differentiation and specialization during the fly olfactory CP impacts long-term circuit function [10]. Here, we will describe recent findings that inform our current understanding of the olfactory CP in fly.

In Drosophila, OSNs extend axons from peripheral sensory organs (the antenna and maxillary palp) into the brain to innervate projection neuron dendrites within the antennal lobe, akin to the vertebrate olfactory bulb [21]. These synaptic connections form distinct neuropil structures called glomeruli. There are 43 morphologically distinct glomeruli present in Drosophila [21,22]. OSN-glomeruli innervation patterns are largely hard-wired, and complete loss of sensory experience does not impact overall innervation patterns [22]. Interestingly, within a CP directly after eclosion (hatching) of adult flies, changes in the activity of individual OSN populations can drive experience-dependent remodeling of downstream glomeruli [19,20]. Golovin et al. found that exposure to ethyl butyrate during the olfactory CP significantly reduced the volume of VM7 glomeruli in a dose-dependent manner via Or42 OSN synapse elimination [18]. Importantly, a recent report from Leier and Foden et al. identified long-term reductions to presynaptic terminals and a reduction of spontaneous activity in Or42 OSNs following ethyl butyrate exposure specifically within this window, validating this system as a true model of CP plasticity [23].

Synaptic pruning, or the select removal of specific neuronal synapses, is a process that is often glial-mediated [9,24–26]. In Drosophila, neuronal pruning is performed by three distinct, CNS-derived glial types that each possess key astrocytic functions: cortex glia, ensheathing glia, and astrocytes [27]. Drosophila astrocytes, like mammalian astrocytes, are peri-synaptic glia that regulate neural circuit development and neuronal signaling. Astrocytes are already known to clear neuronal debris in Drosophila olfactory circuits outside of the CP, an essential role that is conserved in mammalian astrocytes [25,26,28–30]. Ensheathing glia are neuropil-associated glia that exhibit elevated phagocytic capacity compared to astrocytes, in particular after neuronal injury [25]. Recent works have highlighted a prominent role for ensheathing glia in experience-dependent pruning of Or42 OSN synapses during the olfactory CP [23,31]. Nelson et al. found that ethyl butyrate exposure during the CP drives experience-dependent infiltration of ensheathing glia processes into VM7 glomeruli [31]. This infiltration is accompanied by Draper (the Drosophila ortholog of pruning receptor MEGF10) receptor activation, Basket (JNK) nuclear translocation, which in turn, causes upregulation of Cheerio (FLNA) [31]. Upregulation of Cheerio (FLNA) is necessary to assist in remodeling of the cytoskeleton to drive further glial process extension and targeted synaptic pruning [31]. Thus, glial-mediated pruning is a key driver of neural circuit remodeling during an olfactory CP in fly.

While most CP research on fly olfaction focuses on activity-dependent changes to glomeruli structure, recent work by Jafari et al. demonstrated that early life experience drives long-lasting changes in odorant receptor gene expression in OSNs [10]. There are about 1500 OSNs that mediate olfaction in the fly, and approximately 60 odorant receptor genes within the fly genome [21]. What determines which of the 60 receptors each OSN expresses? Jafari et al. recently tested the relationship between early life stress, odorant receptor gene expression, and OSN differentiation [10]. The authors found that expression of the olfactory receptor Or59b is altered by early life stress during a brief CP after eclosion [10]. Specifically, Jafari et al. found that stress-induced OR expression is tightly regulated by the chromatin remodelers dLsd1 (Lsd1), Su(var)3–9 (Suv39h2), and Kdm4b (Jmjd2), which alter heterochromatin formation to stabilize Or59b expression levels [10]. This highlights that developmental experiences not only establish permanent changes to neural circuitry, but also chromatin structure. As epigenetic signatures can perdure for a lifetime, characterizing the genomic consequences of CP plasticity to neurons and associated glia could greatly impact our understanding of how developmental experience shapes long-term neural circuit function.

Vision

Compared to other sensory systems, our knowledge of visual CPs in invertebrates remains relatively sparse. Here, we will briefly discuss current understandings of experience-dependent plasticity in the Drosophila visual circuit, while also highlighting CPs of visual plasticity in vertebrate systems. These vertebrate visual CPs will undoubtedly inform continued visual CP research utilizing Drosophila. While there are examples of experience-dependent plasticity in Drosophila visual circuits, it is unclear whether this plasticity is restricted to a particular time in development. In Drosophila larvae, the Bolwig organ is a primitive system for light detection reminiscent of a more mature visual system [32]. The Bolwig organ is comprised of 12 different photoreceptor cells which are derived from neuroectodermal cells during embryogenesis [33]. These photoreceptors form connections with ventral lateral neurons, or LNvs [34]. During meta-morphosis, all but 4 founding cells of the Bolwig organ remain, which then contribute to the adult optic neuropiles [35]. During early larval stages, light exposure increases LNv dendrite dynamicity [36]. Interestingly, during late larval life, the LNvs show a distinct form of plasticity in response to light exposure [34]. When the Bolwig organ senses light, the LNvs become active, causing a reduction to the length of dendrites [34]. The timing of this phenomenon coincides with a behavioral shift from light aversion in early stage larvae to light attraction in late stage larvae and adult flies [37]. A recent study demonstrated that this remodeling is facilitated by astrocyte-neuron lipid shuttling through the astrocyte-derived secreted lipocalin Glial Lazarillo (GLaz), which is homologous to human Apolipoprotein D [38]. Whether changes in astrocyte GLaz mediate the transition from activity-dependent dynamicity at early larval stages, to activity-dependent changes in growth at late larval stages, remains to be tested.

While our knowledge of visual CPs in invertebrates is still unclear, the visual system of multiple vertebrate systems laid the foundation for our understanding of CP biology [1,2,8,39–41]. In mammalian models, pioneering research in cats first identified CPs in the visual system [13, 14]. In short, monocular deprivation (visual occlusion to a single eye) during a brief developmental window causes permanent structural and functional changes to visual cortex to favor the open eye (e.g. ocular dominance); this plasticity was suppressed in the mature cat cortex (beyond 3 months of age) [2,13]. Thus, there is a CP for sensory-dependent remodeling of the visual cortex [2,13,39,40]. Ocular dominance plasticity is well-conserved in other mammals, including mice, which exhibit peak plasticity at postnatal day 28 and limited plasticity by 3 months of age [2,15,42]. Excitatory signaling is the predominant form of neuronal communication prior to the onset of the visual CP, and as noted above, maturation of inhibitory circuits is essential to terminate the CP [43,44]. Visual CP onset can be shifted earlier by overexpression of BDNF to cause premature maturation of inhibitory neurons, or via treatment of benzodiazepines to enhance GABAergic signaling [43–48]. Alternatively, visual CP onset can be delayed by sensory deprivation or by genetic deletion of Gad65 to reduce GABA synthesis [16, 43,49–52]. Thus, the balance of excitatory/inhibitory signaling is important for proper CP timing [15,53–55].

Beyond synaptic pruning, glial cells are essential for shaping neuronal functions like synaptic development, maturation, and signaling. Indeed, astrocytes regulate the development of both excitatory and inhibitory synapses through various secreted and membrane bound cues [56,57]. For example, astrocyte secretion of Hevin regulates synapse formation and is necessary for ocular dominance plasticity [58]. For this reason, astrocytes are well-poised to regulate CP timing [7,40,59]. Indeed, in a classic experiment by Müller and Best (1989), the authors demonstrated that transplantation of immature astrocytes into the adult cat visual cortex is sufficient to reopen CP plasticity [60]. These experiments were recently recapitulated in mouse. Ribot et al. further showed that astrocytes increase expression of Connexin 30 during the CP, which downregulates RhoA-MMP9 signaling to allow for maturation of chondroitin sulfate proteoglycans into perineuronal nets [40]. Perineuronal net formation is required to drive maturation of inhibitory neurons and in turn, CP closure [40]. These findings were among the first to identify astrocyte-derived signaling pathways that regulate CP timing. Additionally, Lee et al. found that synapse remodeling following monocular deprivation is reliant on astrocyte MEGF10 (Draper) [61]. These findings further support the roles of astrocytes in visual CP regulation and identify conserved roles for astrocyte Draper/MEGF10-mediated synaptic pruning in the execution of CP plasticity. Additional research is needed to identify if this conserved astrocyte function is also present in the Drosophila visual circuit.

Recent work in zebrafish has also expanded our understanding of CP plasticity in non-mammalian systems, which will undoubtedly contribute to our understanding of the evolutionary significance of CPs. The zebrafish visual circuit displays several windows of CP plasticity that affect the function of neurons at distinct levels of the circuit. In the optic tectum, Avitan et al. tracked changes in the organization of spontaneously active tectal neurons during zebrafish development [8]. Here, they discovered a critical window from 5 to 6 days post fertilization (dpf) where experience shapes activity of neurons residing in the neuropil and paraventricular layers of the optic tectum to refine prey-capture behavior [8]. During a similar developmental window (5 to 7 dpf), Xie et al. found that visually-experienced zebrafish larvae had higher optomotor responses to stimuli compared to naïve animals [62]. More recently, Hageter, Starkey, and Horstick described an additional CP in the visual thalamus that spanned from 2 to 4 dpf [63]. This CP shapes the development of an evolutionarily conserved escape response: turn direction preference following loss of visual stimulation (e.g. darkness) [63]. Thus, CP plasticity shapes several visually-evoked behaviors in zebrafish. While it is not yet known whether astrocytes might regulate CP plasticity in zebrafish, the timing of astrocyte morphological maturation supports a potential role for zebrafish astrocytes in CP closure. Zebrafish astrocytes are born by 2 dpf and mature both morphologically and functionally by 6 dpf [64]. In the mouse visual circuit [40] and Drosophila motor circuit [7], astrocyte morphological maturation was a major predictor of CP timing, and astrocyte-synapse contact was essential for CP closure [7,40]. Given that these three CPs of visual plasticity in zebrafish terminate between 4 and 7 dpf, when astrocytes first make synaptic contacts [64], it is possible that astrocytes similarly regulate CP closure in the zebrafish visual circuit.

Established critical periods in motor circuits

In both the olfactory and visual systems, external stimuli like odorant or light exposure are used to modulate neural circuit activity. These stimuli represent naturally occurring phenomenon used to acclimate the neural circuit to the surrounding environment and can be easily manipulated in a laboratory environment. There are unique challenges to studying motor circuit CPs. Given that the motor circuit CP overlaps with embryonic development (at least in Drosophila), motor circuit plasticity is dependent on intrinsic changes to patterns of motor neuron (MN) activity [7,9,24,54]. These challenges may be the reason that motor circuit CPs are less defined in vertebrate models [65,66]. Suppressing MN activity in developing zebrafish enhances motor dendrite dynamicity, though it is unclear whether stimulus-dependent changes are restricted to a particular time of development [65]. Activity-dependent changes to MN dendrites have also been noted in mammalian systems, though these changes have not been studied in the context of a developmentally-confined CP [66]. Here, we will focus on the well-defined Drosophila motor CP.

In Drosophila, the motor circuit matures during late embryogenesis, and altered neuronal activity during this window impacts larval locomotor behavior [7,9,24, 54]. During late embryonic development, neurons within the ventral nerve cord (akin to the vertebrate spinal cord) transition from random firing to waves of patterned spontaneous network activity (PaSNA), which coincides with the development of rhythmic muscle contractions required for larval hatching [55]. The development of PaSNA is dependent on mechanosensory feedback, and inhibition of mechanosensitive chordotonal neurons suppresses PaSNA in embryogenesis and causes long-term deficits in larval foraging behavior [55]. Additionally, ectopic activation of MNs during this window of development enhances susceptibility to electroshock-induced seizures in late larval life [9, 24, 54]. Thus, changes in neural activity during late embryogenesis impacts both homeostatic and stress-induced motor behaviors in mature Drosophila larvae.

While it was clear that late embryogenesis represented an essential window for activity-dependent motor circuit development, the trajectory of the motor CP within the CNS remained unknown until recently. Leveraging optogenetics and high resolution imaging, Ackerman et al. demonstrated that late embryogenesis is a developmental window of heightened activity-dependent remodeling of motor circuit structure, and that this structural plasticity rapidly declines during early larval life [7]. This finding was congruent with a previously defined CP of motor plasticity at the Drosophila neuromuscular junction [67]. After the CP was closed, neuronal activity no longer had an impact on motor circuit structure.

Having characterized the trajectory of the fly motor CP (peak at embryonic stage 17, closed by 8 hours after larval hatching), Ackerman et al. next sought to determine the cellular and molecular mechanisms that drive CP closure. Drosophila astrocytes are peri-synaptic glia that, like mammalian astrocytes, regulate neuronal function and longevity. The authors found that astrocytes invade the motor neuropil just prior to CP closure, and that genetic ablation of astrocytes was sufficient to extend CP plasticity. Through genetic screening, the authors identified several MN-astrocyte signaling pathways, including Neurexin-Neuroligin signaling, as necessary and sufficient for CP closure [7]. Finally, the authors demonstrated that transient extension of the CP caused permanent deficits in motor behavior [7]. These findings parallel the discoveries by Ribot et al. in the mouse visual circuit [16]. Indeed, both studies identified astrocyte extracellular matrix cues and regulation of excitation/inhibition as key mediators of CP closure [40]. In the future, it will be interesting to test whether these astrocyte-derived signaling pathways function in parallel, or in a single pathway, to stabilize motor circuit function and function.

Concluding thoughts

Recent studies identified essential roles of astrocytes in CP closure in two different model organisms (mouse and Drosophila) and in multiple different neural circuit CPs (visual, motor, and olfactory) [7,40]. Within the critical period, astrocytes used conserved pathways to remodel circuits (MEGF10/Draper) in an activity-dependent manner [23,31,61]. Furthermore, astrocytes used core mechanisms (e.g. extracellular matrix) to shut down CP plasticity [7,40]. There is much left to understand regarding the roles of astrocytes in CP timing and CP plasticity more broadly. For example, in the classic Müller and Best study, it was shown that immature astrocytes can reopen CP plasticity in the visual circuit [60]. Does this result reflect an increased ability of immature astrocytes to prune neural circuits, or might immature astrocytes express specific cues that stimulate neural plasticity? Furthermore, recent data demonstrate that other glial cell types (e.g. oligodendrocytes, oligodendrocyte precursor cells, and microglia) also modulate CP plasticity [68]. How might glia communicate with one another to sculpt plasticity across the brain? By leveraging our understanding of fundamental glial functions in development, we anticipate that astrocytes, and glia more broadly, will become the keys to unlock CP plasticity.

Data availability

No data was used for the research described in the article.