Wiring and firing neuronal networks: Endocannabinoids take center stage
Summary of recent advances
Endocannabinoids (eCB) function as retrograde messengers at both excitatory and inhibitory synapses, and control various forms of synaptic plasticity in the adult brain. The molecular machinery required for specific eCB functions during synaptic plasticity is well established. However, eCB signaling plays surprisingly fundamental roles in controlling the acquisition of neuronal identity during CNS development. Recent work suggests that selective recruitment of regulatory signaling networks to CB1 cannabinoid receptors dictates neuronal state-change decisions. In addition, the spatial localization and temporal precision of eCB actions emerges as a novel organizer in developing neuronal networks. Current challenges include fitting novel molecular candidates into regulatory eCB signaling pathways, and defining the temporal dynamics of context-dependent signaling mechanisms underpinning particular neuronal specification events.
Synaptic communication in complex neuronal networks relies on the coincident activity of integrated feed-back mechanisms allowing optimal temporal refinement of activity-dependent synaptic connectivity . Generation and maintenance of the structural and functional coherence of neuronal circuits is the basic cellular principle of higher brain functions. Thus, multiple mechanisms have evolved during brain development to allow the temporal refinement of neuronal excitability. Although redundancy at the level of co-existent signaling networks with partially overlapping functions ensures the continuous adaptation of pre- and postsynaptic components between connected neurons, retrograde synaptic signaling emerges as a uniquely powerful means to tune the temporal and spatial efficacy of synaptic information transfer.
During the past decade, several retrograde messengers have been identified (Figure 1) that exhibit robust differences in their speed of affecting synaptic integration, temporal flexibility and efficiency, and spatial precision [2–7]. Endocannabinoid (eCB) signaling represents a key retrograde signaling pathway  for tuning both homosynaptic and heterosynaptic plasticity  in the postnatal brain. The general molecular paradigm is that eCBs are synthesized postsynaptically in an activity-dependent manner and engage presynaptic CB1 cannabinoid receptors (CB1Rs) on both excitatory and inhibitory afferents thus decreasing neurotransmitter release. Four basic forms of eCB-mediated synaptic plasticity have been described (Figure 1) [8–10]: (i) in depolarization induced suppression of inhibition (DSI), depolarization of a postsynaptic neuron stimulates eCB production that activates presynaptic CB1Rs at GABAergic interneuron terminals, leading to a decrease in inhibitory neurotransmission. eCBs produced in the same fashion acting on CB1 receptors on excitatory neurons evoke depolarization induced suppression of excitation (DSE). (ii) In metabotropic suppression of inhibition (MSI), activation of postsynaptic Gq/11-linked receptors (typically M1 or M3 muscarinic acetylcholine receptors or group I metabotropic glutamate receptors [mGluRs]) leads to production of eCBs activating presynaptic CB1Rs thus decreasing inhibitory neurotransmission. As before, when eCBs produced in this fashion activate CB1Rs on excitatory terminals, the process is termed metabotropic suppression of excitation (MSE). (iii) eCB-mediated long-term depression (LTD) is evoked during sustained stimulation of group I mGluRs, as might happen during prolonged low frequency stimulation of excitatory pathways. LTD can affect either the stimulated pathway (homosynaptic LTD) or a neighboring pathway (heterosynaptic LTD), if the terminals of the neighboring pathway express CB1Rs. (iv) In slow-self inhibition (SSI), eCBs are produced following the repetitive depolarization of a neuron and activate CB1Rs on the same neuron, opening inwardly rectifying potassium channels and causing sustained hyperpolarization of the neuron . SSI is remarkable in that in this form of eCB-mediated plasticity, eCBs are both produced and act in the same cell, as opposed to being retrograde messengers in the other forms. It should be noted that since eCBs can inhibit both excitatory and inhibitory neurotransmission, their net effect at the circuit level, also influenced by the coincident presence of other factors, can be either inhibitory or stimulatory.
Molecular determinants of eCB-mediated retrograde signaling at central synapses appear to be developmentally organized such that they can feed-back to control the earliest events of presynaptic neurotransmitter release during the transition from synaptogenesis to synaptic communication in developing neuronal circuits [12,13]. This leads to the question whether molecular underpinnings of eCB signaling loops acting so efficiently in the postnatal brain subserve particular physiological functions during brain development. The answer appears to be yes, but we are far from understanding the molecular logic and temporal dynamics of eCB signaling networks in the embryonic brain, and how their specific neurodevelopmental functions relate to and define their retrograde control of neurotransmitter release at mature synapses. Important open questions include: where and when eCBs are produced in the developing brain; the molecular identity of eCBs and whether they represent ‘active’ signals; whether respective receptors and intracellular signal transduction cascades differ from those in the postnatal brain; how eCB signaling integrates with other regulatory systems; and how the relative power of this newly-emerging signaling entity contributes to the defining of neurodevelopmental processes. In this review, we focus on recent discoveries establishing eCB-driven cellular identification events in the developing cerebrum, and define a unifying concept of how eCB signaling provides positional signals for excitatory and inhibitory afferents along the dendritic tree of cortical neurons, thus shaping the complexity of cortical connectivity.
Molecular logic of endocannabinoid signaling sculpted by developmental principles
2-Arachidonoylglycerol (2-AG) and anandamide (AEA), members of the eCB family of neuroactive lipids, are primarily synthesized by sn-1-diacylglycerol lipase α/β (DAGLα/β)  and α/β-hydrolase 4/glycerophosphodiesterase 1 (ABDH4/GDE-1)  and bind to cannabinoid receptors in the brain (Figure 1) and at the periphery. 2-AG and AEA promiscuously activate CB1, CB2 (CB1/2R), and other cannabinoid receptors including GPR55 . However, the identity of eCBs and bioactive lipids stimulating GPR55 remains ambiguous: AEA and lysophosphatidylinositol, but not 2-AG, palmitoylethanolamine or virodhamine were shown to activate GPR55 [17–20]. CB1R and GPR55 are expressed on neurons, whereas CB2Rs are primarily found on microglia in the adult [16,17]. CB1R is one of the most abundantly expressed G protein-coupled receptors (GPCRs) by neurons and is selectively targeted to axons and synapses. In contrast, GPR55 expression appears significantly regionalized  with the identity and subcellular distribution of this receptor being unknown. Enzymatic inactivation of 2-AG and AEA primarily involves monoglyceride lipase  and fatty-acid amide hydrolase , respectively. eCB signaling in the cerebrum is principally restricted to excitatory synapses between pyramidal cells and inhibitory synapses of cholecystokinin (CCK)-containing GABAergic interneurons [3,23]. The spatial confinement of eCB signaling to particular synapse populations supports the concept that neurons may simultaneously express molecular determinants of multiple retrograde signaling systems and possess the capacity for domain-specific recruitment of particular signaling machineries to subsynaptic microterritories along their dendrites. Accordingly, several spatially-segregated retrograde feed-back mechanisms have recently been proposed (Figure 1): dendritic release of Wnt family ligands regulates neurotransmitter release at cerebellar mossy fiber-granule cell synapses ; quantal glutamate release suppresses inhibitory inputs originating from parvalbumin-containing GABAergic basket cells on cortical pyramidal cells ; while brain-derived neurotrophic factor (BDNF) released from secretory granules upon afferent stimulation enhances synaptic plasticity of both excitatory and inhibitory cortical synapses [5,6,24]. A striking similarity amongst the Wnt, glutamate and BDNF/TrkB signaling systems is that they play crucial roles during brain development through coordinated control of cellular positioning, identification, and presynaptic and postsynaptic differentiation [4,25–29]. Thus, we argue that it is not a mere coincidence that three signaling systems critical for axon remodeling and presynaptic development during early organization of neuronal networks also play key roles in regulating synaptic activity at mature synapses. This suggests that the eCB system, like the other signaling cassettes, might have a dual function and play unexpectedly fundamental roles in both the wiring and firing of the nervous system.
Expression dictates function: context-dependent signaling at CB1 cannabinoid receptors
The existence of eCB ligands and CB1Rs in the developing rodent and human brain triggered an initial wave of interest when CB1Rs were unequivocally identified as the targets of Δ9-tetrahydrocannabinol (THC) from cannabis . It took almost another decade for the cellular specificity, functions, and interacting partners of eCB signaling networks affecting CNS patterning to emerge. In fact, both 2-AG and AEA are present in the developing CNS from very early stages of differentiation, though 2-AG levels are characteristically a magnitude higher. Although absolute extracellular eCB levels may not reflect true signaling competence, we postulate that 2-AG is the prime eCB in the developing brain, nonetheless, AEA might play a role in regulating 2-AG bioavailability by controlling its levels and physiological efficacy .
Moderate CB1R levels in neural progenitors persist throughout pre- and postnatal life. Likely eCB actions on neural stem and progenitor cells are underscored by ample DAGL expression in neurogenic telencephalic niches [32,33], CB1R/CB2R and metabolic enzyme expression in neurospheres, and by the sensitivity of neural stem cells to pharmacological and genetic disruption of eCB signaling [32–35] (Figure 2). A robust increase in CB1R expression in immature post-mitotic neurons becomes evident as soon as neuronal lineage-commitment occurs [36–38]. The presence of an eCB-rich transient territory marked by DAGL-expressing cells at the outermost border of the subventricular zone (SVZ) facing the first layer of CB1R-positive post-mitotic neurons (corresponding to cells entering the intermediate zone in the cerebrum)  suggests that a propulsive eCB tone exists in the developing neocortex that facilitates radial migration of immature pyramidal cells and GABAergic interneurons  from the SVZ and deep migratory stream, respectively (Figure 2). Neurons undergoing axonal polarization and morphogenesis selectively target CB1Rs to their elongating axon . CB1R expression levels peak as synaptic connectivity is established by cortical pyramidal cells (embryonic day 14–16 in mouse ) and GABAergic interneurons (> embryonic day 17 ). Once synaptogenesis terminates, CB1R expression adjusts such that ideal tuning of synaptic efficacy can occur. eCB-driven neuronal specification however relies on the temporal control of eCB synthesis and inactivation such that eCB actions are defined by (i) the identity of endogenous ligands momentarily available to initiate signaling events; (ii) the chronospecificity of available receptors underpinning specific neuronal identification events; (iii) selective recruitment of signal transduction machineries to activated receptors such that complex cellular regulatory networks required for cellular state-change decisions are activated ; (iv) the coherence of signaling events at the genomic level such that temporally compartmentalized transcriptional regulation of neuronal differentiation is precisely executed .
From birth to functional integration in complex circuitries, neuronal development is regulated by a continuum of cellular state-specific differentiation factors, whose concerted, ordered (inter-)actions ultimately define neuronal fate. Multimodal interplay amongst co-existing signaling systems will define the relative power of eCB actions whilst shaping the central nervous system. An increasingly accepted rule is that receptor homo- or heterodimers form functional units of GPCR signaling. The ability of opioid , A2A adenosine , and D2 dopamine receptors  to heteromerize with CB1Rs provides a mechanism to generate differential G-protein coupling at the CB1R, thus directly modulating physiological output. Alternatively, eCB signaling at CB1Rs can act as either upstream regulator (transactivation, BDNF/TrkB ) or downstream signal effector (N-cadherin/FGF-2 at FGF receptor [38,44]) thereby providing differential control of axonal growth and guidance. The physiological significance of multimer receptor mosaics as signaling ‘hubs’ is being recognized as a primary level of differential regulation of neuronal function. However, added levels of signaling complexity exists with regards to eCBs: (i) the 2-AG precursor diacylglycerol (DAG) is abundantly generated upon activation of, among others, Gαi and Gαq-coupled GPCRs by phospholipase C cleavage of polyphosphoinositide. Thus, coincident signaling through several GPCRs may affect 2-AG production by regulating DAG precursor bioavailability. (ii) AEA can modulate the activity of various ion channels  thus affecting neuronal excitability. Promiscuous AEA actions in the developing CNS may amplify primary eCB actions on CB1Rs. (iii) Agonist stimulation of both CB1R and GPR55 activates distinct signaling pathways. When both receptors are present, the prevailing cellular response has been proposed to depend on the activation state of integrins. When CB1R is linked to unclustered integrins (β1), AEA stimulation induces Gi/o signaling and downstream activation of spleen tyrosine kinase (Syk)-mediated activation of Erk1/2, nuclear translocation of NF-κB and concomitant inhibition of phosphoinositide 3-kinase (PI3K) . Since GPR55 activation mobilizes Ca2+release from intracellular store through PI3K , inhibition of this enzyme has the potential to uncouple some aspects of GPR55 signaling. Once heteromultimeric integrins become clustered by e.g. activation of Mn2+ or diminished extracellular [Ca2+], CB1R is released from the β1 integrin subunit and Syk-mediated PI3K inactivation ceases. Consequently, GPR55 signaling is freed from CB1R-induced repression resulting in intracellular Ca2+ mobilization and differential transcription factor activation. Overall, these data promote the concept that the cellular signaling context drives differential engagement of cannabinoid receptors and selective recruitment of signal transduction cascades upon agonist stimulation. The cumulative impact of cannabinoid receptor activation thus clearly exceeds that of each individual receptor type.
Endocannabinoid define synapse positioning
Establishment of long-range excitatory axons by pyramidal cells precedes the neurochemical specification and synapse patterning of GABAergic interneurons during corticogenesis. Prominent DAGLα/β localization to pyramidal cell axons is required for axonal elongation through cell-autonomous signaling [33,37] (Figure 3). De novo synthesized 2-AG can exert either autocrine regulation via CB1Rs distributed along the longitudinal axis of the growing axon, or paracrine signaling amongst neighboring axons, thus maintaining the integrity of an eCB-rich axonal trajectory. The lack of CB1Rs in glutamatergic growth cones suggests that directional steering decisions are independent of CB1Rs. Fasciculating excitatory axons define an ‘eCB protomap’ by demarcating eCB-rich hot-spots and eCB-sparse corridors (Figure 3A). This eCB-based spatial matrix provides permissive and repulsive niches for GABAergic axons navigating to their postsynaptic targets. Local GABAergic axons lack 2-AG synthesis capacity and use target-derived 2-AG as navigational cues. Consequently, when an inhibitory afferent approaches an eCB-rich brain microdomain of excitatory axons, agonist stimulation of CB1Rs will activate RhoA GTPases, initiating a repulsive response by collapsing the side of the growth cone facing the 2-AG gradient (Figure 3B). Thus, spatial navigation of GABAergic growth cones will be restricted to eCB-sparse microenvironments. Through 2-AG release, pre-existing excitatory axons will thus determine the spatial distribution of inhibitory afferents along the dendritic arbors of pyramidal neurons. This novel mechanism of eCB-driven synapse segregation is further supported by the fact that CB1R and DAGL levels peak in cortical neurons throughout the period of axonal elongation with progressively reduced CB1R expression and increased somatodendritic DAGL targeting as synaptogenesis terminates [12,14,33,36,37]. Another striking feature of our proposed model of neuronal network formation is that the ratio of CB1R and DAGL levels between glutamatergic and GABAergic neurons remains steady throughout neuronal specification and postnatal function. Lastly, the selective recruitment of DAGLα to dendritic spines of adult neurons  reflect postsynaptic arrangements (see also Figure 1A) corresponding to eCB-rich hot-spots during corticogenesis.
The neurodevelopmental impact of THC exposure during pregnancy has recently gained considerable attention given the long-lasting effects of prenatal cannabis use on emotional control, social behaviors, and cognition in affected offspring . THC may be deleterious acting as an agonist or antagonist. Since THC is an intermediate potency, low efficacy CB1R agonist, THC can potentially antagonize the actions of the more efficacious 2-AG in vivo. Thus, in additional to activating CB1Rs, THC effects may be antagonistic when efficacious eCBs are present [10,47]. Physiological situations favoring antagonistic THC actions include low receptor density (e.g., pyramidal cells) or limiting downstream signaling molecules (e.g., during early specification events). Intriguingly, high-affinity antagonists at CB1Rs including, among other compounds, AM251, rimonabant, and taranabant, appear to exert even more powerful effects on developing glutamatergic neurons [33,37] by inhibiting axon fasciculation and postsynaptic target selection. A cell-autonomous eCB tone provided by DAGL activity in developing axons may define the cellular basis for undesired effects of CB1R antagonists, and account for the cortical delamination and perturbed long-range axon patterning induced by AM251 and rimonabant [33,37]. Alternatively, deleterious drug effects may be due to inverse agonism at CB1Rs. We conclude that eCBs can reach very high local concentrations but are secreted in a spatially and temporally precisely coordinated fashion. However, cannabinoid receptor antagonists engage their cognate receptors indiscriminately thus impacting the chronodynamics of eCB signaling in the developing brain.
The eCB system is emerging as a key regulatory signaling network fundamental to the wiring of the brain during development with an array of functions ranging from lineage segregation of stem cells to refinement of synaptic functions in complex neuronal networks. We have recently experienced a quantum leap in understanding the molecular hierarchy and signaling principles selectively governed by eCB signals in the embryonic brain. Nevertheless, the key unresolved question remains to define the exact importance of eCB signaling and its interactions with other developmentally-regulated signaling entities. The recent identification of robust developmental phenotypes in mice with cell-type specific conditional deletion of CB1Rs [12,33,37] challenged our prevailing concept on the limited significance of eCB signaling in the developing brain. Instead, these findings argue that promiscuity at the level of ligands, receptors, and signal transduction pathways provides an essential redundancy allowing exceptional cellular adaptation of eCB signaling: losing a single component will bear modest effects on neurodevelopment. However, losing more than one may be catastrophic.
This work was supported by the Swedish Medical Research Council (T.H.), Scottish Universities Life Science Alliance (T.H.), Alzheimer’s Association (K.M., T.H.), European Molecular Biology Organization Young Investigator Programme (T.H.), European Commission (HEALTH-F2-2007-201159; T.H.), National Institutes of Health grants DA023214 (T.H.), DA11322 (K.M.), DA15916 (K.M.), DA21969 (K.M.), and BBSRC (P.D.).
Conflict of interest statement
The authors have no conflict of interest with any aspect of data or concepts presented in this review.