Intercellular signaling by ectodomain shedding at the synapse

Abstract

Ectodomain shedding (ES) is a posttranslational protein modification process that plays key roles in health and disease. Many neuronal and synaptic membrane proteins are known to undergo ES, but the complexity of functions regulated by the shed peptides is only beginning to be unraveled. Here, we provide an overview of emerging evidence demonstrating that synaptic ES can mediate autocrine and paracrine signaling. We also discuss how advances in large-scale proteomic analyses are leading to the identification of novel synaptic proteins undergoing ES, as well as the targets and functions of their soluble ectodomains. Finally, we also provide an overview of how cerebrospinal fluid analyses of shed proteins could be used as a potential source of new biomarkers for neuropsychiatric disorders.

Synaptic ectodomain shedding is an emerging mechanism in health and disease.

Membrane proteins play essential roles in health and disease; therefore, their function must be tightly regulated. One of the mechanisms regulating membrane protein properties is Ectodomain (see Glossary) shedding (ES), a process in which a protease cleaves the extracellular portion of transmembrane- and glycosylphosphatidylinositol (GPI)-anchored proteins within 10–35 amino acids from the transmembrane region [1]. In some instances, an intramembrane protease can cleave the protein further in a signaling mechanism called regulated intramembrane proteolysis (RIP). RIP generates a second, shorter extracellular fragment and a cytoplasmic one [1] (Figure 1A). The latter can function in intracellular signaling but will not be discussed here.

Figure 1. Synaptic ectodomain shedding.

A. Numerous transmembrane and GPI-anchored proteins undergo proteolytic processing, resulting in the release of a soluble extracellular fragment in a process called ES. B-D. ES at the synapse can be mediated by MMPs (B), ADAM10 (C) and BACE1, ADAM17, MT3-MMP and MT5-MMP (D). Cleaved proteins are indicated next to their main protease; however, each substrate can be cleaved by multiple proteases. Of note, for several proteins it is still unclear whether they are found only in one side of the synapse or both.

An important distinction to consider is the one between shed and secreted proteins. Although both can be present in the extracellular space of a given tissue, shed proteins are originally membrane-attached and fully function as membrane proteins but can be proteolytically cleaved. On the other hand, secreted proteins are soluble non-membrane attached proteins, destined to be released into the extracellular environment by the secretory pathway. Examples of neuronal soluble proteins are hormones (i.e., somatostatin) and growth factors (i.e., brain-derived neurotrophic factor, BDNF). Along these lines, the term “secretome” refers to the totality of proteins found in the extracellular milieu of a given biological sample (including secreted and shed proteins), while “sheddome” refers only to the subset of proteins in the secretome that undergo ES.

ES plays widespread roles in development and disease, including in oncogenesis, metastasis, immunity, cardiovascular and kidney disease, and within the central nervous system (CNS) too. In the CNS, ES is involved in a range of processes, from brain development [2], neurogenesis [3], axonal regulation [4], to formation and maintenance of synapses [5, 6]. While ES was initially seen as serving the termination of membrane-bound protein function, it is becoming increasingly clear that it has much broader functions, including autocrine and paracrine signaling. Furthermore, dysregulation of synaptic ES has been linked to a range of neurological disorders, including Alzheimer’s disease (AD) [7] and prion disease [8]. As ectodomains shed within the brain can be detected in the CSF, proteomic analyses of this fluid show the promise of identifying novel (patho)physiological mechanisms, and biomarkers for brain disorders. Here, we provide an overview of synaptic ES, focusing on intercellular signaling exerted by shed ectodomains. We also summarize evidence on altered synaptic ES in CNS disorders, and the use of CSF sheddome analyses to evaluate synaptic (dys)function.

Synaptic ectodomain shedding is catalyzed by sheddases.

A growing body of evidence indicates that many synaptically-localized membrane proteins undergo ES (Figure 1B–D). This is supported by work focusing on individual proteins, such as neuroligins (NLGN) and neurexins (NRXN) [6, 9–11], but also by global analyses of the neuronal sheddome [2, 12–14]. ES is catalyzed by proteases, also known as “sheddases”, which cleave their targets on the extracellular side, near the membrane. Matrix metalloproteases (MMPs), “A disintegrin and metalloproteases” (ADAMs) and “β-site APP cleaving enzymes” (BACE1 and BACE2) are among the most common sheddases in the CNS and thus involved in the cleavage of most synaptic proteins. These proteases have been reviewed in detail before [1], hence we will only introduce them as they are relevant to synaptic ES. While the above mentioned sheddases are relatively well characterized, it is important to note that for most shed proteins, the responsible sheddases are not yet known, and the substrates for only a few sheddases have been catalogued. All sheddases, and likely the synaptically acting ones, exert their functions in manners dependent on developmental stage and tissue, brain region, cell type, subcellular location and cellular activity [9, 15–17]. Moreover, each protein can be cleaved by several sheddases and each sheddase can cleave several proteins, adding to the complexity of ES regulation. For an overview of mechanisms regulating ES, see Box 1.

Box 1. Mechanisms of shedding susceptibility.

The observation that one sheddase can cleave numerous substrates (Table 1), leads to the question of how ES is fine-tuned to maintain the delicate balance between health and disease. The molecular mechanisms regulating shedding susceptibility remain largely unknown. These mechanisms include phosphorylation, glycosylation, alternative splicing, protein trafficking, and interaction of the sheddase or the sheddable protein with regulatory proteins.

Alternative splicing and O-glycosylation occurring near the cleavage site are two mechanisms associated with shedding regulation [105, 106]. Both modulate ES of the synaptic cell adhesion molecule SynCAM1 [107, 108]. O-glycosylations have been reported to modulate APP ES, thereby decreasing Aβ production [107]. Particular N-glycosylation patterns regulate intracellular trafficking of sortilin-related receptor (SORLA), affecting its shedding susceptibility [109]. Neuronal activity-dependent SIRPα phosphorylation is necessary for releasing its ectodomain, which leads to synapse maturation [104]. Another mechanism regulating ES, is the expression of sheddase regulatory subunits that form active sheddase complexes. As an example, expression of rhomboid 1 is necessary for ADAM17-dependent ES of ‘multiple epidermal growth factor- like domains protein 10’ (MEGF10) [110]. Interestingly, sheddable membrane proteins can avoid ES by binding to certain proteins, as described for ADAM17-mediated Neogenin ES [111]. Finally, the presence of polysialic groups determines which sheddase can cleave NCAM; while polysialylated NCAM can be cleaved by ADAM10 and ADAM17, absence of this glycan group makes it only cleavable by ADAM17 [24, 49].

Several mechanisms could lead to alterations in ES regulation, potentially contributing to disease pathogenesis. Mutations can increase the affinity of the substrate for one sheddase over the others. For instance, the APP Swedish mutation increases its processing by BACE1, accelerating Aβ deposition. In contrast, the APP Icelandic mutation, also located near the BACE1 cleavage site, protects from developing dementia and presents with decreased amyloidogenic processing of APP [112]. A premature stop codon in the CNTNAP2 gene has been found in several Old Order Amish children suffering from Cortical Dysplasia Focal Epilepsy (CDFE) syndrome. The mutation creates a frameshift, resulting in the loss of the transmembrane and the cytosolic domains of CNTNAP2 [113]. This would result in the translation of a soluble version of CNTNAP2, very similar to the shed ectodomain of CNTNAP2. However, the CDFE mutant ectodomain would be released by the secretory pathway, while the wild-type ectodomain is processed by activity-dependent ES [20]. It seems reasonable to speculate that many other disease-associated mutations of sheddable membrane proteins could be altering these proteins’ regulation by ES, thereby contributing to disease pathogenesis.

MMP substrates include proteins that play important roles in synaptogenesis and synaptic plasticity, such as NLGNs [18], intercellular adhesion molecule 5 (ICAM5) [19], and CNTNAP2 [20] (Figure 1B). ADAM10 plays important roles in synapse formation and maintenance, as well as in axonal development, by cleaving synaptic cell adhesion molecules (CAMs), including cadherins and the neurexin–neuroligin complex (Figure 1C). Other sheddases have also been described to process synaptic proteins (Figure 1D). ADAM17 regulates neuronal development and plasticity by catalyzing the shedding of several proteins, including neuronal pentraxin receptor (NPTXR) [21], p75 neurotrophin receptor (p75NTR) [22], L1 cell adhesion molecule (L1CAM) [23], and neural cell adhesion molecule 1 (NCAM or NCAM1) [24]. BACE1 is abundant at synapses and sheds several important synaptic proteins, including amyloid precursor protein (APP), seizure related 6 homolog (SEZ6), SEZ6L, SEZ6L2, close homolog of L1 (CHL1), L1CAM, and NCAM1 and 2 [12, 15]. BACE1-mediated shedding has been shown to affect synapse formation, function, plasticity, and behavior in mice [25]. Less studied metalloproteases, such as MT3-MMP and MT5-MMP also regulate ES of synaptic proteins, including N-cadherin (CDH2) [26] and Nogo-66 receptor 1 (NGR1) [27] (Figure 1D). γ-secretase is an intramembrane protease that also plays important roles at synapses by cleaving mainly substrates with long extracellular domains, and frequently requires ES by sheddases before γ-secretase cleaves its target protein within the transmembrane domain [28]. A different mode of membrane protein processing is catalyzed by GPI phospholipases or GPI-phosphodiesterase 2, which cleave GPI-anchored proteins releasing their extracellular domains. This mechanism is less discussed in the literature, but is essential for neurogenesis, and modulates aging and cognition [3, 29].

Several neurophysiological processes are known to regulate sheddase activity and, thus, synaptic ES. Neuronal activity, long-term potentiation (LTP), and memory formation during Morris water maze learning increase MMP9 activity [30]. Fear conditioning, a model of aversive associative learning, increases MMP9 and MMP2 activity in rodents. Moreover, seizure induction by kainate [31] and pilocarpine administration [32, 33] increase MMP activity in rodents. Psychostimulants such as methamphetamine [34] or cocaine [35] are also known to increase protease activity. Other inducers of MMPs include chronic stress [36] and ischemia [37]. Interestingly, several proinflammatory factors (LPS, IL-1, TNFα) stimulate MMP9 and MMP2 expression [38], while anti-inflammatory stimuli negatively regulate them [39].

Important synaptic proteins undergo ectodomain shedding.

Several synaptic proteins have been shown to undergo ES. Below, we discuss some of the most prominent examples (Table 1).

Table 1.

Examples of synaptic proteins known to undergo ectodomain shedding, including the cleaving proteases and regulating stimuli.

Synaptic protein	Protease	Stimuli	References
APP	ADAM10 BACE1 γ-secretase*	Neuronal activity Nicotinic receptor activation NMDAR activation	[41,42]
CD44	MMP9	5-HT7R activation	[56]
CDH2	MT5-MMP ADAM10	Bicuculline-induced neuronal activity NMDA-induced neuronal activity	[26, 50]
CHL1	BACE1 γ-secretase*	N.D.	[12, 57]
CNTNAP2	MMP9	cLTP	[20]
EFNA2	ADAM10	EFNA2/EPHA3 binding	[4]
ICAM5	MMP9	NMDAR and AMPAR activation	[19, 51]
L1CAM	MMPs ADAM10 BACE1	PMA-induced PKC signaling NMDA incubation	[23, 53, 55]
NCAM	ADAM10 ADAM17 BACE1	EFNA5/EPHA3 binding	[15, 24, 49]
Nectin1	ADAM10	Constitutive NMDAR activation	[52]
NGL3	MMPs γ-secretase*	LTD induced by NMDA Low frequency stimulation (LTD)	[58]
NGR1	MT3-MMP	N.D.	[27, 103]
NLGN1	MMP9 ADAM10	NMDA+glutamate Interaction with sNRXN Pilocarpine-induced epileptic seizures	[5, 6, 18]
NLGN2	MMP9	NMDA+glutamate	[18]
NLGN3	ADAM10 MMP9	NMDA+glutamate	[18, 48]
NOTCH	ADAM10	Binding to Delta ligands	[66]
NPTXR	ADAM17/TACE	mGluR1/5-dependent LTD	[21]
NRG1	BACE1 ADAM10 ADAM17	NMDA Kainate	[62, 63]
NRXNs	MMP γ-secretase	KCl-induced neuronal activity	[45]
NRXN3(3	ADAM10 ADAM17/TACE α-secretase γ-secretase*	N.D.	[46, 47]
SEZ6 SEZ6L	BACE1	N.D.	[59]
SIRPα	MMPs	KCl Activity-induced tyrosine phosphorylation	[70, 104]

N.D. not determined.

^*Not sheddase, intramembrane cleavage.

APP is a single-pass transmembrane protein abundant at synapses. Its improper proteolytic cleavage is an emblematic example of ES contributing to a brain disorder. The amyloidogenic pathway is mediated by the activity of BACE1 followed by γ-secretase, which generates the neurotoxic amyloid-β (Aβ). Processing of APP by anti-amyloidogenic α-secretases (e.g., ADAM10) followed by γ-secretase, results in the production of the neuroprotective soluble APPα (sAPPα) [40], limiting the production of toxic Aβ. APP proteolytic processing has been described to be activity- and nicotinic receptor-dependent [41, 42]. Interestingly, sublethal NMDA receptor activation in cultured cortical neurons causes a shift from α-secretase to β-secretase APP processing, promoting Aβ production [17]. Proteolytic cleavage of APP has been reviewed extensively, hence we will not discuss it in detail [43, 44].

Neurexins (NRXN1-3) are a family of transmembrane proteins involved in the development and maturation of synapses. In the CNS they are located mainly pre-synaptically. NRXNs undergo KCl-induced ES regulated by MMPs and then they are further cleaved by γ-secretase [45]. NRXN3β undergoes shedding mediated by ADAM10/17 [46] and α- and γ-secretases [47]. CNTNAP2 is a CAM located at synapses and is a prominent neurodevelopmental disorder risk factor. Proteomic profiling of the neuronal sheddome in vitro and human CSF together with super resolution imaging techniques revealed that CNTNAP2 undergoes activity-dependent shedding mediated by MMP9 [20].

Neuroligins (NLGN1-4) are a family of postsynaptic proteins that interact with presynaptic neurexins through their extracellular domains. Neuroligins promote the maturation of presynaptic terminals and regulate synaptic plasticity. Activity-dependent shedding of NLGN1, 2 and 3 has been reported so far [18]. NLGN1 is cleaved by MMP9 and ADAM10 upon activation of NMDA-R in vitro and by pilocarpine-induced seizures in vivo [5, 6, 18]. NLGN2 undergoes activity-dependent shedding mediated by MMP9 [18]. NLGN3 is cleaved by ADAM10 and MMP9 also in an activity-dependent manner [18, 48].

Classical synaptic CAMs also undergo ES. NCAM is an adhesion molecule that controls axon and dendrite development, as well as synaptic plasticity. NCAM ES is catalyzed by ADAM10 via ephrinA5 (EFNA5)/EphA3 interaction [49]. ADAM17-mediated ES of NCAM regulates neurite outgrowth in vitro [24]. Interestingly, NCAM1 and NCAM2 are differentially processed by BACE1 in vivo, depending on the brain region and developmental stage. A study in mice examined the cleavage of NCAM1 and NCAM2 by BACE1 at different developmental timepoints, including postnatal day 10, and at 4 and 12 months of age. NCAM1 and NCAM2 ES was observed at all ages in the olfactory bulb. In the hippocampus, however, NCAM1 ES was observed only at postnatal day 10, and not during adulthood, while NCAM2 was not processed by ES in this brain region at any age [15]. At synapses, CDH2 is cleaved by MT5-MMP and ADAM10 upon increased neuronal activity induced by bicuculline or NMDA treatment [26, 50]. The neuron-specific ICAM5 is present in filopodia and immature dendritic spines and has an important role in higher-order cognitive functions. ICAM5 shedding is regulated by NMDA-R and AMPA-R in an MMP-dependent manner [19, 51]. Nectin1 is a CAM that localizes pre- and post-synaptically and plays important roles in synapse maturation. It undergoes ADAM10-dependent ES, in the pre- and post-synaptic compartments, both constitutively and upon NMDA-R activation [52]. L1CAM plays significant roles in synaptic function and neurodevelopment, including processes such as neurite outgrowth and myelination. L1CAM ES is regulated by MMPs, ADAM10, ADAM17 and BACE1, and it is stimulated by PKC activation and NMDA incubation. ES of L1CAM is involved in cell adhesion, migration, and neurite outgrowth [12, 53, 54]. KO of l1cam in zebrafish was shown to cause axonal growth abnormalities and hydrocephalus; however, L1cam ES plays a role only in hydrocephalus and not in axonal outgrowth, as shown in rescue experiments with soluble and uncleavable forms of L1cam in zebrafish [55]. CD44 is cleaved by MMP9 upon serotonin receptor 5-HT7R activation inducing synapse remodeling [56]. CHL1 is a cell adhesion molecule that undergoes BACE1-dependent ES, and posterior γ-secretase cleavage. This modulates semaphoring-3A-mediated growth cone collapse. CHL1 ES by BACE1, induces growth cone collapse which is stopped when the membrane-bound CHL1 fragment is further cleaved by γ-secretase [57]. NGL3 (netrin-G ligand-3), a postsynaptic CAM that trans-synaptically interacts with the leukocyte antigen-related (LAR) family of receptor tyrosine phosphatases, undergoes proteolytic cleavage upon NMDA-induced long-term depression (LTD) and low-frequency stimulation, catalyzed by MMPs and γ-secretase/presenilin [58]. Experiments in mice indicated that the cleavage of SEZ6 and SEZ6L, which plays roles in synaptic connectivity and motor coordination, is catalyzed by BACE1 in the brain [59].

Ephrins and Eph receptors are another class of synaptic proteins that undergo shedding. EphA and -B receptors are the largest subfamily of receptor tyrosine kinases and their ligands, ephrins, are transmembrane (ephrin-B, EFNB) or GPI-anchored (ephrin-A, EFNA) proteins localized on opposite cells. Signaling by Eph-ephrin interactions regulate diverse processes including repulsive and attractive axon guidance, dendritic spine remodeling, and synaptic plasticity [4, 60]. Eph-induced axonal retraction requires proteolytic cleavage of EFNA2 by ADAM10 following binding to its receptor EphA3 in vitro [4].

Several other molecules that do not belong to the families discussed above, also undergo ES. NPTXRs are localized to excitatory synapses, where they bind to AMPA-R and regulate AMPA-R dependent plasticity. ADAM17-mediated NPTXR shedding is induced by mGluR1/5 dependent LTD-like stimuli [21]. NGR1 mediates axonal growth inhibition, regulates axonal regeneration, and negatively regulates plasticity. NGR1 undergoes ES catalyzed by MT3-MMP, promoting excitatory synapse formation [27]. Moreover, NGR1 ectodomain administered in vivo promoted the erasure of fear memories in mice [61]. Finally, neuregulin1 (NRG1) is a neurotrophic protein that binds to ERBB3 and ERBB4, and plays important roles in neurodevelopment. NRG1 is shed by BACE1, ADAM10 and ADAM17, processes which in zebrafish have also been shown to regulate myelination [62]. In rats, NRG1 ES is activated by kainate-induced seizures [63]. Interestingly, zebrafish expressing a NRG1 probe fused with mCherry and GFP to the extracellular and intracellular domains, respectively, showed that its ES occurs preferentially at axonal locations despite being expressed also in somas [16].

Physiological functions of synaptic ectodomain shedding.

Many of the physiological functions of ES are only beginning to be uncovered, and the function of most shed ectodomains is currently unknown. However, several biological roles for synaptic ES can already be distinguished.

Termination of activity

For synaptic CAMs ES often causes loss of cell-cell contacts and synapse weakening, as removal of trans-synaptic adhesive protein-protein interactions reduces bidirectional synaptic signaling necessary for synaptic assembly/maturation [5, 6, 64]. NLGN1 ES inhibits synaptic maturation and decreases neurotransmitter release. It also attenuates excitatory postsynaptic potential (EPSC) frequency and amplitude, by terminating the interaction between NLGN1 and presynaptic NRXN [5, 6]. Proteolytic cleavage of NLGN3 also reduces synaptic strength: re-expression of NLGN3 in NLGN1-3 tripleknockdown organotypic hippocampal slices enhanced inhibitory and excitatory synaptic transmission. This effect was abrogated by PMA-induced NLGN3 ES [48]. In cultured cortical neurons, activity-dependent proteolytic cleavage of NLGN1 decreased vesicle release probability and mEPSC frequency by destabilizing its interaction with NRXN1 [5]. ADAM10-dependent CDH2 cleavage negatively regulates dendritic spine morphology in vitro [65].

ES of synaptic receptors may lead to the termination of receptor effects and may even give rise to decoy receptors. For example, NOTCH ES creates an extracellular inert product that is trans-endocytosed and degraded [66]. Eph/ephrin-mediated cell repulsion require termination of the high-affinity Eph-ephrin interaction by either endocytosis or ectodomain cleavage [4].

Consistent with synapse weakening effects, ES is sometimes associated with LTD. For example, NPTXR cleavage is induced by LTD-like stimuli and promotes AMPA-R internalization [21], while the shed ectodomain of NLGN1 decreases synaptic activity by activating presynaptic mGluR2 [67].

Interestingly, in some instances, ectodomain cleavage is a mechanism to remove an inhibitory signal. NGR1 ES by MT3-MMP removes a break on synaptogenesis, as either NGR1 ES or application of sNGR1 promote excitatory synapse formation [27].

Intercellular signaling by synaptic ectodomain shedding.

Growing evidence shows that soluble ectodomains function as intercellular signals to activate downstream pathways in synapse development, plasticity, and brain circuit function. The complexity of effects exerted by shed ectodomains involves autocrine and paracrine signaling, causing diverse effects on various cell types and during different developmental stages (Table 2, Key Table). As such, the same ectodomain may have different, sometimes even opposite roles in different situations, or the shed ectodomain may have opposite roles to the full-length protein. Of note, the technical approach most often used to study the effects of shed ectodomains has been the exogenous application of the peptides. Thus, in some instances, it cannot be specified whether the effects described for these ectodomains are due to paracrine or autocrine signaling. Moreover, paracrine effects have been most frequently studied, probably due to technical reasons. Below we discuss the most prominent examples of autocrine/paracrine effects exerted by shed ectodomains of synaptic proteins.

Table 2.

Autocrine/paracrine effects of shed synaptic ectodomains.

Synaptic protein	Type of effect	Cell type/system	Developmental stage	Effect	Reference
APP (Aβ)	Autocrine and paracrine	Cultured hippocampal slices (rat)	Postnatal day 6–7	↓Spine density and plasticity	[41]
CNTNAP 2	Paracrine	Cultured cortical neurons (mouse)	Mature neurons	↓Neuronal synchrony	[20]
ICAM5	Bath application. Autocrine?	Cultured hippocampal neurons (mouse and rat)	Mature neurons	Filopodia elongation, fexcitatory transmission	[51, 69]
L1CAM	Autocrine/paracrine	CHO cells	N.A.	Cell migration	[53]
NCAM	Paracrine	Neocortex and hippocampus (mouse)	Adult	↓Synapses in amygdala	[68]
NRG1	Paracrine	Swann cells (zebrafish)	Larva	Peripheral nervous system myelination	[62]
NRXN1β	Acute bath application. Paracrine?	Cultured hippocampal neurons (rat)	Immature neurons	Neurite outgrowth	[9]
	Acute bath incubation. Paracrine?	Autaptic hippocampal neurons (rat)	Mature neurons	↑Synaptic transmission
	Prolonged bath incubation. Paracrine?	Autaptic hippocampal neurons (rat)	Mature neurons	↓Synaptic transmission
NRXN3β	Paracrine	Granule cell to newborn neurons (mouse)	Adult	Spine maturation	[46]
	Autocrine	Newborn neurons (mouse)	Adult	Axonal development
SIRPα	Paracrine	Cultured hippocampal neurons (mouse)	Immature neurons	Presynaptic maturation	[70]
αNRXN	Paracrine	Neuromuscular junction (C. eleaans)	Adult	↓ACh release	[10]

Several autocrine/paracrine roles have been reported for neurexins and neuroligins. In C. elegans, ES of postsynaptic neurexin inhibits acetylcholine release by binding to and inhibiting presynaptic α2δ subunits associated with Ca_V2 channels at the neuromuscular junction, thereby regulating synaptic transmission [10] (Figure 2A). This interaction is evolutionarily conserved despite different synaptic orientations of neurexin in C. elegans and mouse. Soluble NRXN3β (sNRXN3β) secreted by mature granule cells in the adult mouse hippocampus, increased spine density on hippocampal newborn neurons in a paracrine manner. In contrast, sNRXN3β secretion by newborn neurons had no effect on their own spine density but induced axonal maturation in an autocrine manner [46]. In cultured hippocampal neurons before synaptogenesis, acute bath application of sNRXN1β increased Ca²⁺ influx via N-type Ca²⁺ channels, and it increased neuritogenesis; in mature neurons, the Ca²⁺ influx elicited by sNRXN1β was NMDAR dependent instead [9]. These effects were probably paracrine, because NRXNs are mostly presynaptic, and the effects were NLGN dependent, which is mostly postsynaptic. On the other hand, while an acute exposure to sNRXN1β increased glutamatergic synaptic transmission in mature neurons, a prolonged incubation inhibited it [9].

Figure 2. Examples of paracrine signaling by soluble ectodomains shed from synaptic proteins.

Proteins of origin, sheddases, regulation of shedding, soluble ectodomains, protein targets and biological actions are indicated for the ES of NRXN1 [10] (A), CNTNAP2 [20] (B), NLGN1 [67] (C) and SIRPα [70] (D).

Another member of the neurexin superfamily, CNTNAP2, also undergoes ES [20]. CNTNAP2 ectodomain decreased neuronal network activity and enhanced Ca²⁺ extrusion in hippocampal slices in a PMCA2-dependent manner. Interestingly, colocalization of shed CNTNAP2 with the presynaptic marker VGLUT1 increased upon chemical LTP (cLTP). We hypothesize that upon activity, shed CNTNAP2 binds to and activates presynaptic PMCA2. This increases presynaptic Ca²⁺ extrusion, which in turn acts as a feedback inhibitor of neuronal activity [20] (Figure 2B). sNLGN1 activates presynaptic metabotropic glutamate receptor 2 (mGluR2) in a paracrine manner, decreasing glutamate release from mossy fibers, and synaptic activity [67] (Figure 2C).

The soluble ectodomains of several CAMs also have autocrine/paracrine functions. A transgenic mouse expressing the NCAM ectodomain in the neocortex and the hippocampus exhibited a paracrine reduction of synapses in the amygdala and the cingulate and frontal association cortices, but not in the hippocampus, which demonstrates a lack of autocrine effect [68]. Early studies showed that L1CAM ES promotes CHO cell migration by autocrine/paracrine stimulation via αvβ5 integrin receptors [53]. ES of ICAM5 is an interesting example of bidirectional regulation of synaptic development; the full-length protein inhibits dendritic spine maturation, while the shed ectodomain promotes it [51]. The effect of sICAM5 is lost in ICAM5 deficient hippocampal cultures, which could be pointing to a dominant negative effect of sICAM5. This observation also indicates that the effect of sICAM5 might be autocrine, as ICAM5 is mostly expressed postsynaptically. Moreover, bath application of sICAM5 increased excitatory transmission, and insertion of GluA1 in the neuronal surface in cultured hippocampal neurons [69].

Shed ectodomains of several other types of proteins also function as paracrine signals. For example, activity-dependent ES of signal regulatory protein-α (SIRPα) is a trans-synaptic mechanism that promotes presynaptic maturation by the interaction of sSIRPα and the presynaptic CD47 in a paracrine manner [70] (Figure 2D). The soluble EGF-like domain of NRG1 is released by BACE1 and ADAM17 cleavage, and rescues myelination abnormalities in BACE1 KO zebrafish in a paracrine manner [62]. The effects of the proteolytic processing of APP to generate Aβ have been described to be autocrine and paracrine. Aβ production was sensitive to neuronal activity and nicotinic receptor blockade, and reduced the number and plasticity of dendric spines in an autocrine and paracrine manner [41].

Collectively, these findings illustrate the richness of effects exerted by shed ectodomains depending on the developmental stage, cell type, directionality of effect (autocrine/paracrine), or exposure time. The mechanisms differentiating between autocrine vs paracrine signaling remain to be elucidated, although different affinities for presynaptic or postsynaptic binding partners, or competition with the full-length protein probably play a role. In some instances, both the shed ectodomain and the full-length protein are known to bind the same target. Two known examples are the sCNTNAP2-PMCA2 [20] and sAPP-integrin [71] interactions. However, it is still unknown whether both protein species bind their targets with the same affinity. Moreover, soluble ectodomains with the same binding profiles as the full-length protein, may play dominant negative effects. This may occur when binding of the full-length protein to its target induces an intracellular cell signaling cascade, which would be impeded by the soluble version due to the lack of the transmembrane and the cytosolic domains.

Targets of shed synaptic ectodomains

To date, few studies have identified binding partners for synaptically shed ectodomains or examined the functional consequences of this interaction. In one of the studies addressing these issues, the authors used affinity pull-down assays of recombinant sAPP interacting with synaptosomal proteins, coupled with mass spectrometry, and showed that sAPP binds to GABA_B1A receptors, which modulated synaptic transmission and plasticity [72]. Earlier studies have shown that sNRXN1 binds Ca_V2 channels reducing acetylcholine release at the C. elegans and mammalian neuromuscular junction [10]. Lastly, in a recent study from our group, sCNTNAP2, which interacts with and activates the extrusion pump PMCA2, was shown to decrease cytosolic Ca²⁺ [20]. These emerging reports underscore the potential of identifying novel targets of shed ectodomains in revealing processes controlled by autocrine/paracrine signals.

Global analyses of neuronal ectodomain shedding and the CSF

While most of the reports address ES of individual proteins, a growing number of studies have examined the sheddome globally, facilitated by recent advances in proteomics. These studies typically involve the analysis of conditioned media from cultured neurons, sometimes followed by validation in mice. They led to the identification of novel substrates for known sheddases such as BACE1 or ADAM10 [2, 12], but also new biological functions of ES [20]. Several methods have been developed to analyze neuronal sheddomes, such as “secretome protein enrichment with click sugars” (SPECS), or its recent version “high-performance SPECS” (hiSPECS) [13]. They have been used in conjunction with sheddase inhibition or knockouts to identify substrates. These studies allowed the generation of comprehensive lists and databases of shed proteins. One such searchable database, SheddomeDB, compiled experimentally validated shed membrane proteins [73] (available at http://120.126.86.226:8080/).

In most cases, these investigations were not specifically designed to address the synaptic sheddome. To gain insight into global patterns of synaptic ES, a recent study from our group analyzed datasets from SPECS and liquid chromatography tandem mass spectrometry (LC-MS/MS) analyses of media from cultured neurons [20]. Based on a bioinformatic analysis, the most enriched biological process was found to be ‘positive regulation of synapse assembly’. It was also found that the neuronal sheddome is enriched in risk factors for neurodevelopmental disorders. Interestingly, this unbiased analysis underscored the role of ES in synapse regulation and synaptopathologies, which is supported by previous information on individual protein’s ES.

The CSF is the most readily accessible source of proteins derived from the brain in living humans, hence it has the potential to offer a readout of brain function. Several proteomic analyses have been performed to detect brain-enriched proteins in the CSF, including in human CSF extracted by lumbar puncture. Studies in mice show that the genetic or pharmacological manipulation of sheddases such as BACE1 [74] and ADAM10 [75] modulates the levels of ectodomains in the CSF. There is a notable overlap between the in vitro neuronal and CSF sheddome datasets, suggesting that the CSF sheddome mirrors the neuronal sheddome [20]. Moreover, autism spectrum disorder (ASD) and schizophrenia (SZ) risk gene-encoded proteins were enriched in the cleaved, but not in the secreted CSF fraction. While requiring further investigation, this supports the idea that proteomic analysis of the CSF might be useful to study synapse pathophysiology. Indeed, the human CSF (hCSF) has been used to understand synapse pathology, as discussed later.

Implications of synaptic ectodomain shedding for neuropsychiatric disorders

Altered ES, including of synaptic proteins, has been associated with CNS diseases. This suggests broader roles of synaptic ectodomains in pathogenic processes or as biomarkers, which remain to be explored. The most studied example of deleterious ES is the amyloidogenic APP cleavage in AD. However, ES has also been implicated in other pathologies including glioma proliferation [11] or microglial activation [76]. As discussed next, both global proteomic ES analyses and investigations of specific synaptic ectodomains are helping advance current understanding of how synaptic ES contributes to neuropsychiatric disorders.

Individual synaptic ectodomain alterations in specific CNS diseases.

Altered levels of individual soluble ectodomains of synaptic proteins have been detected in the CSF of patients with numerous neurological and psychiatric disorders (Table 3). There could be several reasons for such alterations in shed ectodomains in the CSF, including altered expression levels or transport of the cleavable proteins, genetic variants that affect the cleavage, aggregation of pathogenic cleaved ectodomains, altered expression levels or targeting of sheddases, altered neuronal activity levels which modulate shedding, altered number of synapses, or other yet unknown mechanisms.

Table 3.

Synaptic ectodomains detected in hCSF in CNS diseases.

Protein	Disorder	Levels	Reference
sCNTNAP2	ASD	Decreased	[20]
sICAM5	Acute encephalitis	Increased	[87]
	MS	Decreased	[89]
	Temporal lobe epilepsy	Increased	[88]
sLICAM	SZ	Decreased	[77]
	Dementia, AD	Increased	[78]
	Glioblastoma	Increased	[90]
sNCAM	SZ	Increased	[77]
	MS	Decreased	[86]
NRXN2and 3	Preclinical AD	Decreased	[96]
	Prodromal AD	Increased
	AD	Increased
sp75NTR	AD	Decreased	[79]

Increased levels of sNCAM have been detected in the CSF of patients with SZ [77], and a transgenic mouse overexpressing sNCAM displayed SZ-related abnormal behaviors including deficits in sensory gating and emotional memory, as well as decreased dendritic spine density [68]. Decreased sL1CAM has also been found in SZ CSF [77], while elevated sL1CAM has been found in the CSF of dementia patients, including AD [78]. Decreased levels of p75NTR ectodomain have also been detected in the CSF of AD patients [79].

As for autism, proteins encoding ASD susceptibility genes were found to be enriched within the sheddome, but not in the secreted components of the secretome, both in neurons in vitro and in the human CSF [20]. Moreover, decreased levels of sCNTNAP2 in the CSF were observed in a group of individuals with ASD [20]. This points to a potential role of ES in ASD, a hypothesis that needs further validation.

While a role for ES in epilepsy has not yet been extensively investigated, a number of studies suggest it. Several sheddases including MMP9 and ADAM10 are altered in epilepsy patients [80–82], and their manipulations modify seizure phenotypes in mice [83–85].

Altered levels of soluble ectodomains have also been reported in several other neurological disorders. In multiple sclerosis (MS) patients, sNCAM levels were lower when compared to healthy controls, while treatment with disease-modifying therapies increased them [86]. Whether this reversal in the levels of sNCAM is a reflection of neuroplasticity/neurorepair mechanisms remains to be elucidated. sICAM5 levels in the CSF also change in pathophysiological conditions: they are increased in acute encephalitis [87] and temporal-lobe epilepsy [88], and decreased in MS [89]. On the other hand, treatment of a mouse model of MS with sICAM5, ameliorated the disease’s symptoms [89]. Finally, patients with glioblastoma and brain metastasis present increased levels of sL1CAM in the CSF [90].

Global analyses of CSF in CNS diseases

Global proteomic analyses of CSF have been performed in patients with several neurodegenerative disorders, including AD [91, 92], amyotrophic lateral sclerosis (ALS) [93], and sporadic Creutzfeldt-Jakob disease (sCJD) [94]. In many of the related studies, synaptic proteins have been of particular interest, given the broadly accepted role synapses play in pathogenesis. In ALS, WD repeat-containing protein 63, amyloid-like protein 1, SPARC-like protein 1, and cell adhesion molecule 3 have been proposed as candidate biomarkers [93]. Synaptic proteins like proSAAS, apolipoprotein J, neurosecretory protein VGF, phospholemman, and chromogranin A were proposed as early biomarkers in AD [95]. CSF analyses in different stages of AD demonstrated that certain synaptic proteins are decreased in preclinical patients before CSF markers of neurodegeneration or symptoms are observed [96]. In contrast, in later stages of neurodegenerative disorders including ALS, Parkinson’s and AD, elevated levels of synaptic proteins in the CSF were observed, and these changes are thought to reflect synapse loss [79, 93, 96–98]. Finally, CSF proteome analyses of major depressive disorder patients also showed a decrease in synaptic proteins, including NRXN3 and NPTXR [99], and bipolar disorder 1 patients showed reduced levels of several brain-expressed proteins including testican-1, TNFRSF21, CADM3, and ADAM22 in the CSF [100].

Overall, these studies indicate that synaptic changes might be detected in the CSF and that events occurring at different stages of the pathophysiological cascade could be measured with a relatively low-invasive technique such as a lumbar puncture. Indeed, the CSF has been proposed as a source for biomarkers of synapse function or pathology in the brain [91, 93, 95]. However, it must be considered that changes in the synaptic CSF proteome could be reflective of synapse loss, altered ES, or disease-associated cell death. Of note, there have been some efforts to define the synaptic sheddome by bioinformatically filtering synaptically expressed proteins from the CSF sheddome, however none of the published data addresses the synaptically-originated sheddome. Thus, proteomic studies are needed to shed light on how the synaptic sheddome is altered in CNS disorders.

Soluble ectodomains as potential therapeutics in CNS diseases

In preclinical studies, soluble synaptic ectodomains have also been described to have therapeutic utility. For example, in an AD mouse model, the p75NTR ectodomain was shown to be neuroprotective against amyloid-β toxicity, as it improved memory performance, reduced Aβ levels in the brain, and attenuated inflammation and dendritic spine loss [101]. In hippocampal slices, sAPPα, but not sAPPβ generated by the amyloidogenic pathway, protected against Aβ-associated dendritic spine loss and increased tau phosphorylation [40]. sNRX1β also showed neuroprotective properties in cultured hippocampal neurons exposed to H₂O₂ and potassium deprivation [9]. An NCAM mimetic peptide, has shown neuroprotective properties in cultured cerebellar and dopaminergic neurons [102]. In addition, treatment of a MS mouse model with sICAM5, ameliorated the disease symptoms [89].

Caveats in studying the synaptic sheddome

There are several important caveats of studies addressing the synaptic sheddome. In experimental studies, the endogenous targets of ectodomains may also interact with the full-length, membrane-attached, non-cleaved proteins. Competition of exogenous ectodomains with endogenous binding of its target or its source protein may be a confound.

There are also caveats inherent to mass spectrometry analyses. Mass spectrometry could miss some proteins due to low peptide counts and results could be influenced by biochemical sample preparation. Proteomic analyses of CSF carry caveats as well, in particular in human subjects. When analyzing human CSF in brain disorders, incorrect diagnosis or comorbidity with other diseases could be an important confound. Other parameters that could influence human CSF studies are medication, age differences, alterations in blood-brain barrier permeability, or contamination with blood during CSF extraction. Replications, independent validation, and consistent methodology are thus essential to ensure reproducibility. Validation of results in larger independent cohorts of patients and longitudinal analyses of the CSF proteome during the different phases of the diseases will be also crucial to determine future reliable biomarkers. Another potential confound could be extracellular vesicles present in the CSF, however, their levels are typically low compared to shed proteins and can be eliminated by ultracentrifugation. Proteins could also be present in the CSF due to cell death or damage. They could also originate from other tissues apart from the brain, as the CSF is derived from the blood plasma.

Concluding remarks and future perspectives

Initially considered as a mechanism that terminates membrane protein function, ES is now considered as a key posttranslational mechanism regulating cell biology, which is supported by its role in the pathogenesis of disorders such as cancer and AD. Shed synaptic ectodomains are emerging as neuronal modulators that function in conjunction with classic neurotransmitters and neuromodulators. They have the potential of exerting autocrine and/or paracrine effects, as they can diffuse away from the cell of origin. Moreover, the effects of shed ectodomains are complex, as they depend on the developmental stage, cell/tissue of origin and the target cell. Despite significant progress over the last years, there are still many questions that need to be addressed, particularly in the context of synaptic ectodomain shedding. These include the comprehensive identification of shed synaptic proteins, their binding partners, their modulatory effects, and the molecular mechanisms regulating their ES (see Outstanding Questions). Moreover, exogenous ectodomains or smaller fragments derived from them could be further explored as therapeutic approaches for brain disorders.

Outstanding Questions Box.

Which synaptic proteins undergo ES? What is the full extent of the synaptic sheddome, and which soluble ectodomains have roles in paracrine and autocrine signaling? How do soluble ectodomains affect the function of individual neurons and neuronal circuits? How do they affect behavior?

How do physiological conditions, such as neuronal activity, neuromodulators, immune signals, alter synaptic ES globally? What individual proteins and pathways are preferentially affected?

What are the targets of shed synaptic ectodomains? How do ectodomains affect the function of these targets and off-target cells?

How does synaptic ectodomain shedding participate in pathogenic processes in the CNS? Can faulty ES give rise to brain disorders or contribute to disease progression? Do mutations in sheddase substrates alter their ES in a manner that could contribute to disease?

How is the synaptic sheddome reflected in the CSF? Can CSF shed ectodomains originating from central synapses be used to monitor biochemical processes in the brain? Can the analysis of the sheddome reflected in the CFS be exploited for biomarkers?

Can exogenous ectodomains or bioactive peptides derived from them be employed therapeutically in CNS disorders?

The evidence for altered synaptic ES in neuropsychiatric disorders is growing. This is perhaps not surprising given the central roles synapses play in brain pathogenesis, and the fact that many membrane proteins undergo ES. However, better understanding of the global function and regulation of synaptic ES in the context of CNS (patho)physiology is needed. In this regard, proteomic analyses of the synaptic sheddome in various conditions could reveal novel pathogenic pathways, biomarkers, and drug targets. As shed synaptic proteins can be detected in the CSF, this fluid has the potential of offering information on synapse (patho)physiology. Yet, the relationship of the synaptic sheddome and the CSF proteome remains to be systematically investigated.

Highlights.

• Numerous synaptic proteins undergo ectodomain shedding, often regulated by neuronal activity, and many more await discovery.

• Previously thought as a modality for protein inactivation, emerging evidence indicates that ectodomain shedding generates soluble ectodomains which act as paracrine signals to modulate neuronal development, function, plasticity, and circuit properties.

• Faulty neuronal ectodomain shedding is associated with a growing number of neurological disorders.

• Shed ectodomains originating from central synapses are detectable in the cerebrospinal fluid (CSF), and their monitoring could provide insights into brain processes and potentially serve as CSF biomarkers for brain disorders.

• Large-scale proteomic analyses are poised to identify novel synaptic proteins undergoing ectodomain shedding, as well as new functions and targets for soluble ectodomains, raising the possibility that soluble ectodomains could be developed into novel therapeutics.

Glossary

Ectodomain

the extracellular portion of a transmembrane domain or a GPI anchored protein. It could be part of the full-length protein or soluble, upon shedding

Sheddase

protease that catalyzes the cleavage, and the shedding, of membrane protein ectodomains

Secretome

all proteins found in the extracellular milieu of a given biological sample (including secreted and shed proteins)

Sheddome

subset of proteins in the secretome that undergo ES and thus originally possess at least one transmembrane domain or a GPI anchor