The neuronal pentraxin family consists of two secreted glycoproteins, NPTX1/NP1 and NPTX2/NP2/NARP, and the NPTX receptor (NPTXR). NPTX1 and NPTX2 cluster AMPARs at excitatory synapses onto inhibitory neurons, interacting directly with the N-terminus of AMPARs or indirectly via NPTXR (Table 1). Secreted NPTX1 interacts with NPTXR and GluA4 at shaft synapses resulting in the clustering of NPTX1 and GluA4-containing AMPARs, which are highly expressed on parvalbumin-positive (PV +) interneurons in the forebrain (Sia et al., 2007; Pelkey et al., 2015). The selective effect of NPTXs on these interneurons can be attributed to their almost exclusive interaction with GluA4 subunits (Sia et al., 2007; Pelkey et al., 2015). NPTX2 is an immediate early gene whose expression is correlated with neuronal activity (O’Brien et al., 1999; Xu et al., 2003; Chang et al., 2010) and mediates homeostatic synaptic scaling (Chang et al., 2010). NPTX2 is released from dense-core vesicles (DCVs) near glutamatergic synapses, multimerizes, and becomes trapped in the Extra-Cellular Matrix (ECM)-containing perineuronal nets that surround PV + inhibitory neuron cell bodies (O’Brien et al., 1999; Reti et al., 2008; Chang et al., 2010). Thus, NPTXs act as extracellular scaffolds that interact with and cluster AMPARs and mediate an activity-dependent mechanism to recruit inhibitory neurons into circuits to dampen excitation. Interestingly, NPTX2 can be detected in the circulatory system (Reti et al., 2008), suggesting it may have other functions and act far beyond its sites of release.

Netrin-1 was discovered as a regulator of neurodevelopmental processes including cell migration, axon guidance, and synapse formation in invertebrates and vertebrates (Ishii et al., 1992; Kennedy et al., 1994; Colón-Ramos et al., 2007). More recently, Netrin-1 was shown to promote excitatory synapse formation (Goldman et al., 2013) and activity-dependent insertion of AMPARs into synapses (Glasgow et al., 2018). NMDAR-mediated Ca²⁺ influx stimulates the release of Netrin-1 from hippocampal dendrites, resulting in autocrine activation of the Netrin receptor Deleted in Colorectal Cancer (DCC), activation of Calcium/Calmodulin Kinase II (CaMKII) and increased surface abundance of GluA1-containing AMPARs (Glasgow et al., 2018, 2020). Netrin-1 released from dopaminergic and GABAergic neurons in the ventral tegmental area similarly acts in an autocrine manner to promote AMPAR-mediated currents (Cline et al., 2023). The Netrin co-receptor Down Syndrome Cell Adhesion Molecule (DSCAM) also promotes AMPAR levels and clustering in cultured Aplysia neurons (Li et al., 2009). Additionally, long-term potentiation (LTP) occludes Netrin-1-induced potentiation of AMPAR EPSCs, and vice versa, indicating that activity-dependent secretion of Netrin-1 is a mechanism of LTP (Glasgow et al., 2018). Conditional deletion of Netrin or DCC from excitatory neurons in the hippocampus results in defects in spatial memory in rodents (Horn et al., 2013; Wong et al., 2019), and shRNA-mediated knockdown of Netrin-1 from spinal dorsal horn neurons impairs pain sensation and decreases the membrane fraction of GluA1-containing AMPARs (Cui et al., 2021). Thus, Netrin is secreted in an activity-dependent manner from dendrites of many cell types and promotes synaptic AMPAR retention, providing an example of a secreted factor being used in an autocrine fashion to strengthen glutamatergic signaling in response to postsynaptic activity. It remains to be seen whether the effects of Netrin-1 are restricted to stimulated synapses or if secreted Netrin-1 can diffuse further away to reach neighboring unstimulated synapses.

Noelin-1/Olfactomedin-1 is a glycoprotein secreted from neurons in the cortex, cerebellum and hippocampus. Noelin-1 and the related Noelin-2 regulate AMPAR trafficking and function in zebrafish and rodents (Sultana et al., 2014; Nakaya et al., 2017). Noelins are conserved in all vertebrates but do not appear to have an obvious homolog in invertebrates (Boudkkazi et al., 2023). Noelin-1 was originally isolated as an AMPAR-interacting protein from rodent synaptosomes (von Engelhardt et al., 2010; Schwenk et al., 2012). Noelin-1 interacts with GluA1 and GluA2 in heterologous cells and, similar to NPTXs, binds directly to the N-terminal domain of GluA2 (Pandya et al., 2018), resulting in the stabilization of synaptic AMPARs (Díaz-Alonso et al., 2017). Interestingly, the effects of Noelin-1 are modified by the presence of the ECM because Noelin-1 reduces AMPAR mobility on young neurons but not on mature neurons which are surrounded by ECM. Noelin-1 may also link AMPARs with Chondroitin sulfate proteoglycans (CSPGs) in the ECM via Nogo, a Noelin-1 receptor (Frischknecht et al., 2009). A recent study shows that Noelins 1-3 act as polyvalent extracellular scaffolds that stabilize AMPARs by anchoring them to a network of secreted and transmembrane proteins (Boudkkazi et al., 2023). Using high-resolution immuno-EM and electrophysiology recordings, neurons lacking Noelins 1-3 were observed to have 40 and 70% reductions in AMPARs at synapses on excitatory and inhibitory neurons, respectively, and defects in hippocampal LTP. This study suggests that secreted factors like Noelins and their extracellular protein networks are major contributors to LTP (Boudkkazi et al., 2023). Interestingly, chemical LTP increases Noelin-1 levels in the synapse suggesting that activity may increase Noelin-1 secretion to stabilize synaptic AMPARs during plasticity (Pandya et al., 2018).

Brain-derived neurotrophic factor, a member of the neurotrophin family, signals through its tyrosine-kinase receptor, tropomyosin-related Kinase B (TrkB) to regulate several processes including neurogenesis, neuronal survival, axon outgrowth, and synapse formation (Messaoudi et al., 1998; Lee and Hempstead, 2018). LTP stimulates BDNF transcription via the transcription factor CREB (Castrén et al., 1993; Dragunow et al., 1993; Tao et al., 1998, 2002; Hong et al., 2008) and promotes BDNF secretion (Balkowiec and Katz, 2000; Aicardi et al., 2004). Conversely, application of mature BDNF induces LTP at hippocampal synapses (Kang and Schuman, 1995a,b). These early studies suggested that BDNF could regulate AMPAR expression or trafficking during LTP. Indeed, BDNF/TrkB signaling increases transcription and protein expression of GluA1, GluA2, and GluA3, and AMPAR colocalization with PSD-95 in dendrites (Jourdi et al., 2003; Caldeira et al., 2007; Nakata and Nakamura, 2007). Of all secreted factors, the intracellular signaling pathways that mediate the effects of BDNF on synaptic AMPAR levels are probably the best understood. Specifically, BDNF/TrkB complexes activate the Ras/ERK pathway that is involved in surface delivery of AMPARs (Li and Keifer, 2008, 2009; Li and Wolf, 2011; Reimers et al., 2014) and phosphorylates S831 on GluA1 subunits, a critical site for CaMKII and protein kinase C (PKC) binding, promoting AMPAR synaptic delivery and retention (Caldeira et al., 2007; Keifer, 2022). Thus, BDNF is a secreted factor whose expression and secretion are regulated by activity that promotes AMPAR levels at synapses via multiple intracellular mechanisms.

The large glycoprotein Thrombospondin-1 is secreted by astrocytes and was originally shown to induce synapse formation during development, although these synapses initially lack AMPARs and are silent (Christopherson et al., 2005). Thrombospondin-1 also functions in the mature nervous system to reduce synaptic transmission: treatment of cultured rat spinal cord neurons with Thrombospondin-1 increases lateral diffusion and endocytosis of AMPARs and increases levels of inhibitory glycine receptors at synapses (Hennekinne et al., 2013). Thus, Thrombospondin-1 is a glia-secreted factor that regulates synapse development and AMPAR trafficking in mature neurons.

Astrocytes and microglia secrete the protein SPARC which antagonizes synapse formation induced by hevin, another astrocyte-secreted factor (Jones et al., 2011; Kucukdereli et al., 2011). Consistent with a role for SPARC in negatively regulating excitatory transmission, hippocampal neurons from SPARC knock-out mice exhibit increased mESPC amplitude and frequency. Moreover, when cultured with SPARC knock-out astrocytes, hippocampal neurons exhibit increased surface levels of GluA1 and GluA2. Intriguingly, secretion of SPARC from astrocytes is dependent on synaptic activity, suggesting a model whereby increased synaptic activity leads to SPARC release and downregulation of AMPAR surface levels and signaling (Jones et al., 2011). These studies indicate that SPARC is an activity-regulated secreted factor that inhibits glutamatergic signaling by reducing surface levels of AMPARs at synapses, although how SPARC controls intracellular signaling pathways in neurons to impact AMPAR trafficking remains to be determined.

Astrocytes contribute to the ECM surrounding neurons by secreting proteoglycans such as CSPG. Frischknecht et al. (2009) showed that digestion of the ECM backbone with hyaluronidase results in increased AMPAR mobility in the membrane and increased short-term plasticity (STP), suggesting that the ECM restricts exchange of desensitized synaptic AMPARs with naïve extrasynaptic AMPARs during STP (Frischknecht et al., 2009). Since CSPG levels increase during postnatal development, they may contribute to synapse maturation by limiting plasticity (Pizzorusso et al., 2002). Consistent with this, digestion of ECM increased synapse number in a hippocampal/astrocyte co-culture system and electrophysiological recordings indicated that the synapses were immature (Pyka et al., 2011). Thus, components of the ECM like CSPG and its interacting proteins are required to stabilize synaptic AMPARs and promote synapse maturation. It will be interesting to determine how CSPG expression or secretion from astrocytes is regulated to control synaptic AMPARs and maturation.

Glypicans are conserved heparan sulfate proteoglycans (HSPGs) comprising 6 family members in mammals that are anchored to the plasma membrane via glycosylphosphatidylinositol links. Allen et al. (2012) biochemically isolated Gpc4 and Gpc6 from astrocyte-conditioned media as factors that induce functional retinal ganglion cell (RGC) synapses. Both Gpc4 and Gpc6 promote cell surface clustering of GluA1-containing AMPARs and increased excitatory transmission (Allen et al., 2012). Knock-out of Gpc4 in mice decreases GluA1 and reduces excitatory transmission in the developing hippocampus. The authors proposed that glypicans induce immature synapses containing GluA1 which then subsequently mature by incorporating GluA2/3 and GluA4 subunit-containing AMPARs. Although glypicans can be proteolytically cleaved and released from the plasma membrane, it is not clear how astrocytes regulate the release of glypicans.

Other studies have revealed the mechanism by which Gpc4 regulates AMPARs. Mammalian glypicans and the Drosophila glypican homolog Dally like interacts with Leukocyte common antigen-related (LAR) receptor tyrosine phosphatases (RPTPs) family members (Johnson et al., 2006; Takahashi and Craig, 2013; Ko et al., 2015) and postsynaptic Leucine-Rich Repeat TM4 (LRRTM4) (de Wit et al., 2013; Siddiqui et al., 2013). These interactors promote excitatory synapse formation and function in mice (Dunah et al., 2005; Linhoff et al., 2009; Siddiqui et al., 2013; Soler-Llavina et al., 2013) by increasing surface GluA1-containing AMPARs (Schwenk et al., 2012; Shanks et al., 2012; Siddiqui et al., 2013). Additionally, astrocyte-secreted Gpc4 stimulates release of NPTX1 to promote AMPAR clustering at synapses (see above). Disruption of the NPTX1-AMPAR interaction with antibodies blocks Gpc4-mediated synapse formation, and loss of Gpc4 or LAR family member RPTPδ results in accumulation of NPTX1 in presynaptic terminals, reduced GluA1 clusters and decreased synapse number (Farhy-Tselnicker et al., 2017). These data suggest a model whereby astrocyte-released Gpc4 interacts with presynaptic RPTPδ resulting in secretion of NPTX1 from neurons and postsynaptic AMPAR clustering.

Excitatory synapse maturation is marked by a switch from calcium-permeable GluA1-containing to calcium-impermeable GluA2-containing AMPARs. Blanco-Suarez et al. (2018) identified Chordin-like 1 (Chrdl1) as an astrocyte-secreted factor that promotes synapse maturation by increasing surface levels of GluA2-containing AMPARs at synapses on rat RGCs, thus inducing the switch from calcium-permeable to calcium-impermeable AMPARs (Blanco-Suarez et al., 2018). Interestingly, Chrdl1 knock-out mice exhibit a shift toward immature synapses which is associated with increased plasticity in the visual cortex. The authors propose a model where Gpcs act first to promote clustering of GluA1-containing AMPARs at immature synapses followed by Chrdl-1 which promotes the switch to GluA2-containing AMPARs typically associated with mature synapses. Consistent with this model, astrocyte expression of Gpc4 and Gpc6 is highest in the first two postnatal weeks when immature synapses are developing while Chrdl1 peaks in the cortex later at the time of synapse maturation in mice (Cahoy et al., 2008; Allen et al., 2012). It remains to be determined how Chrdl1 secretion is regulated and how Chrdl1 regulates intracellular trafficking to promote GluA2-containing AMPARs. Given that CSPGs also promote synapse maturation, it will be informative to investigate how Chrdl-1 and CSPG interact to regulate maturation.

The proinflammatory cytokine TNFα is secreted from astrocytes or microglia and promotes surface levels of AMPARs in hippocampal neurons in vitro and in vivo (Beattie et al., 2002). TNFα regulation of AMPARs is not involved in hippocampal LTP or LTD, but is required for homeostatic synaptic scaling (Stellwagen and Malenka, 2006) in which chronic changes in synaptic activity result in a global adjustment of synaptic strength by altering AMPAR levels (O’Brien et al., 1998; Turrigiano et al., 1998). Synaptic scaling up is blocked if TNFα is scavenged by soluble TNFR1 or if neurons are co-cultured with TNFα knock out glia (Stellwagen and Malenka, 2006). Mechanistically, TNFα increases the exocytosis of GluA2-lacking AMPARs in hippocampal and cortical neurons (Ogoshi et al., 2005; Stellwagen and Malenka, 2006; He et al., 2012) via PI3 kinase (Stellwagen and Malenka, 2006) while simultaneously promoting the internalization of GABA receptors (Stellwagen et al., 2005). The effects of TNFα are brain region-specific. TNFα has the opposite effect on AMPARs in the mouse dorsolateral striatum where it promotes internalization of GluA1-containing AMPARs from medium spiny neurons and reduces corticostriatal synaptic strength (Lewitus et al., 2014). In the striatum, TNFα treatment activates protein phosphatase 1 (PP1) via DARPP-32 which promotes AMPAR endocytosis by increasing the dephosphorylation of GluA1 on S845 and S831.

Wingless (Wg)/Wnt is secreted from neurons and glia and regulates pre- and postsynaptic development via the receptor Frizzled at the glutamatergic Drosophila neuromuscular junction (NMJ) (Packard et al., 2002; Mathew et al., 2005; Miech et al., 2008). Loss of Wg results in decreased synapse numbers and abnormal glutamate receptor (GluR) distribution (Packard et al., 2002). Interestingly, Wg is secreted from glia in addition to motor neurons at the Drosophila NMJ. Loss of Wg in neurons or subperineurial glia results in increased GluR cluster size and intensity, however, these synapses are correlated with reduced quantal content. Thus, Wg secreted from both glia and neurons is required for the normal distribution and function of GluRs. Interestingly, the source of Wg matters because Wg secreted from motor neurons, but not glia, regulates NMJ size, suggesting that neuron-derived Wg regulates NMJ structure. Although it is not clear how the Wg receptor Frizzled signals to regulate GluR trafficking, vertebrate Wnt7a promotes synapse formation in hippocampal neurons via postsynaptic CaMKII (Ciani et al., 2011) which is known to promote synaptic AMPAR levels during plasticity (Hayashi et al., 2000).

A recent study in C. elegans showed that the VEGF homolog PVF-1 is secreted from muscle and regulates surface levels of the calcium-permeable AMPAR GLR-1 in upstream neurons (Luth et al., 2021). Interestingly, PVF-1/VEGF is expressed and released from muscle to mediate GLR-1 surface levels in pre-motor interneurons that reside two synaptic layers upstream of the neuromuscular junction. Loss of the VEGF Receptor (VEGFR) homologs VER-1 and VER-4, or PVF-1 reduces GLR-1 cell surface levels, resulting in impaired glutamate-dependent locomotor behaviors (Luth et al., 2021). Local recycling of GLR-1 to the plasma membrane is notably unchanged in VER mutants, consistent with the idea that VEGFR signaling in C. elegans promotes long-range AMPA receptor trafficking, perhaps via retromer (Luth et al., 2021).

In vertebrates, VEGF is expressed in vascular endothelial cells throughout the body including in the brain and spinal cord (Licht et al., 2011), but also in neurons (Ogunshola et al., 2002; Schiera et al., 2007), microglia (Bartholdi et al., 1997) and astrocytes (Ijichi et al., 1995). VEGF is also expressed in multiple muscle cell-types such as cardiac myocytes (Levy et al., 1995), vascular smooth muscle (Ishida et al., 2001) and skeletal muscle (Germani et al., 2003). VEGF expression can be induced by ischemia and, in the case of skeletal muscle, is known to both promote angiogenesis and act in an autocrine manner to regulate myoblast function to promote new muscle growth (Germani et al., 2003). Interestingly, VEGFR family members are expressed in neurons and glia and can be induced in astrocytes and microglia following injury (Ogunshola et al., 2002; Choi et al., 2007a,b).

Several studies indicate that VEGF regulates glutamatergic transmission in vertebrates (Cammalleri et al., 2011; Licht et al., 2011; De Rossi et al., 2016). Inhibiting or stimulating VEGF signaling in adult mice has respective effects on LTP in the hippocampus and glutamate-dependent fear conditioning (Licht et al., 2011). VEGFRs localize to the postsynaptic density of mouse hippocampal neurons, and although VEGF alone failed to increase AMPA receptor trafficking or LTP (Cammalleri et al., 2011; De Rossi et al., 2016), co-administration of VEGF and NMDA activated PKC and CaMKII, increased the delivery of calcium-permeable GluA1-containing AMPARs to synapses, and increased excitatory synapse number. Conversely, conditional knockout of VEGFR2 in neurons impaired LTP and fear-related memory consolidation, without altering basal transmission (De Rossi et al., 2016). Collectively, these findings suggest that in the hippocampus, VEGFR2 signaling may promote activity-dependent strengthening of glutamatergic transmission and resultant behavior by promoting the synaptic incorporation of AMPARs. Together these studies indicate that VEGF regulates AMPAR levels and trafficking in vertebrates and invertebrates, however, the source of VEGF in vertebrates is not clear. Given that PVF-1/VEGF is secreted from muscle to regulate AMPARs in C. elegans, it is tempting to speculate that muscle-secreted VEGF may, under some circumstances, also act at a distance to regulate AMPARs in the brain in vertebrates.

Muscle activity during exercise has beneficial effects on energy metabolism and brain function (Cotman et al., 2007; Mattson, 2012). Interestingly, previous studies identified irisin as a muscle-released factor that signals to the brain. FNDC5 is a transmembrane protein expressed in the brain and skeletal muscle that is cleaved and released as a peptide called irisin during exercise. Irisin enters the circulation and directs fat metabolism and thermogenesis and also mediates many of the beneficial effects of exercise on cognition (Boström et al., 2012). Peripheral injection of FNDC5 or irisin increases irisin protein levels in the hippocampus without altering irisin mRNA levels in the brain (Wrann et al., 2013; Lourenco et al., 2019; Islam et al., 2021), suggesting that irisin can cross the blood-brain barrier. Once in the brain, irisin induces the expression of BDNF in the hippocampus (Wrann et al., 2013), which as discussed above, can regulate AMPAR expression and trafficking. Interestingly, BDNF expression is increased after exercise in the hippocampus (Vaynman et al., 2004; El Hayek et al., 2019; Lourenco et al., 2019) and also in skeletal muscle (Cuppini et al., 2007; Szuhany et al., 2015) and can be released into the circulatory system (Pan et al., 1998; Matthews et al., 2009).