Where canonical protein degradation pathways are concerned, synaptic protein degradation via the ubiquitin-proteasome system (UPS) has undoubtedly received the most attention. Indeed, a significant number of synaptic proteins have been shown to undergo ubiquitination and/or degradation in a ubiquitination-dependent manner (e.g. ^{31, 37, 49, 57– 67} ). Moreover, pharmacological suppression of proteasomal activity has been shown to affect global and/or synaptic levels of synaptic proteins ^{30, 49, 50, 57, 68, 69} (reviewed in ^{2, 5, 70– 72} ). Yet the interpretation of such findings is not always straightforward. First, it often remains unknown at what point during a synaptic protein’s life cycle ubiquitination-dependent degradation occurs, a matter that cannot be ignored given the central role of UPS-mediated degradation in quality-control processes (see below). Second, ubiquitination does not necessarily imply degradation, as ubiquitination can act as a signal for other downstream events or as an effector of protein function without affecting synaptic protein levels ^{73– 75} (see also ⁷⁶ ). Third, under physiological conditions, ubiquitination is continuously countered by deubiquitinating enzymes; thus, ubiquitination cannot be taken as unequivocal evidence for imminent proteasomal-based degradation ^{77, 78} . Fourth, ubiquitination can also mark proteins for lysosomal degradation or autophagy (see below). Finally, many of these findings are based, at least in part, on pharmacological agents that inhibit proteasomal activity; proteasome inhibition, however, can lead to unexpected results, including suppressed synthesis of most proteins and enhanced synthesis of others (e.g. ^{1, 33, 79– 85} ). Unfortunately, methods typically used for studying the effects of proteasomal inhibition (e.g. Western blots and quantitative immunohistochemistry) are incapable of separating effects on protein degradation from effects on protein synthesis, as they report total protein quantities. In part, this problem can be bypassed by pharmacologically inhibiting protein synthesis; this manipulation, however, severely limits experiment duration and has been shown to affect protein degradation pathways (e.g. ⁸⁶ ). Pulse-chase experiments based on radioactive amino acids are better in this respect but, as mentioned above, are not well suited for studying the degradation of long-lived proteins because of the conflicting requirements for long labeling periods and low concentrations of unlabeled amino acids, and the compromised availability of essential amino acids this entails.

In a recent study ³³ , metabolic labeling with stable isotopes was used to specifically measure the degree to which proteasomal inhibition slowed the degradation of hundreds of proteins expressed by cultured rat cortical neurons using an approach that overcame most of the aforementioned limitations. Somewhat surprisingly, highly effective proteasomal inhibitors did not slow the degradation of most identified proteins (1,530, including 176 synaptic proteins in total). In other words, apart from relatively small sets of neuronal and synaptic proteins, degradation of most identified proteins continued at normal rates, even when proteasome-mediated degradation was effectively inhibited. This analysis, however, also illustrated the inherent limitations of probing synaptic protein degradation through catabolic inhibition ( Figure 1). Given the slow degradation rates of many synaptic proteins, short periods of catabolic inhibition might barely affect protein abundance. As an example, consider a protein with a half-life of 7 days (such as the postsynaptic density [PSD] protein ProSAP1/Shank2 ³² ). Here, inhibiting its catabolism for 4, 10, or 24 hours would be expected to increase its abundance by merely ~1.6%, 4%, and 9%, respectively. Thus, without prior knowledge of the half-lives of the proteins of interest, as well as a good appreciation of measurement sensitivity, results of such experiments remain inconclusive. In the aforementioned study ³³ , known half-lives for identified proteins were used to calculate the expected impact of catabolic inhibition, and these values were compared with experimental measurements and constrained by calibrations of measurement sensitivity. This analysis allowed the authors to conclude that proteasomal inhibitors did not slow the degradation of many, if not most synaptic proteins, although findings remained inconclusive for some synaptic proteins whose particularly long half-lives (>5 days) precluded sufficiently accurate measurements.

Although these findings might seem surprising at first, they are in good agreement with prior large-scale studies carried out in human colon cancer ⁷⁷ , HeLa ⁸⁷ , human Jurkat ⁷⁸ , and osteosarcoma cells ¹ . In all of these studies, practically no changes in the abundance of thousands of proteins were observed following 5–8 hours of proteasomal inhibition, in spite of dramatic changes in protein ubiquitination ^{77, 87} .

A possible complication in the interpretation of these studies relates to the consistent observation that prolonged proteasomal inhibition induces proteotoxic stress, driven by the buildup of unfolded/misfolded proteins ^{88, 89} . This stress, in turn, triggers cellular responses, such as the unfolded protein response (UPR), which leads to a generalized shutdown of protein synthesis (reviewed in ⁴⁷ ). Interestingly, the synthesis of synaptic proteins appears to be particularly sensitive to this response ³³ (see also ⁹⁰ for a recent review on relationships between the UPR and neurodegeneration), further confounding interpretations based on total protein abundance measurement in proteasomal inhibition experiments.

What is now clear beyond doubt is the crucial role the UPS plays in rapidly degrading newly synthesized, misfolded ⁴⁷ , excess, or superstoichiometric proteins ³⁹ . It is thus possible that UPS-mediated degradation of some synaptic proteins occurs primarily during the earliest phases of their life cycle, namely immediately after synthesis and initial processing (see above). In fact, for both γ-Aminobutyric acid (GABA) _A ⁴⁹ and GABA _B receptors ⁵⁰ , this was proposed to serve as an important regulatory step. Conversely, UPS-mediated degradation might play a secondary role once the proteins are localized within synapses, at least under baseline conditions. In agreement with this possibility, it was suggested that some recently synthesized, particularly large, proteins are degraded rapidly after their synthesis in a UPS-dependent manner and at rates unexpected from their known lifetimes ³³ . Along these lines, a recent electron cryotomography study suggests that in the absence of proteotoxic stress, only 20% of 26S proteasomes in cultured hippocampal neurons are actively engaged in substrate processing, which might be taken as evidence for their importance in handling excess or misfolded proteins ⁹¹ .

This proposition does not negate the importance of locally acting, activity-regulated, UPS-mediated synaptic protein degradation ^{2– 8} . Thus, for example, a series of studies has demonstrated and characterized the redistribution of proteasomes to dendritic spines following synaptic stimulation or activation of postsynaptic N-methyl-D-aspartate (NMDA)-type glutamate receptors ^{92– 96} . Along these lines, NMDA receptor activation was shown to drive PSD remodeling through UPS-mediated degradation of PSD-95 (a major postsynaptic scaffolding protein ⁶⁸ ). Similarly, an activity-inducible kinase (serum-inducible kinase) was shown to mark UPS-mediated degradation of another PSD protein (spine-associated Rap GTPase-activating protein), leading to subsequent synapse loss ⁹⁷ . Interestingly, the degradation of this kinase was also found to be UPS mediated. Changes in activity levels were also shown to drive UPS-mediated degradation of the postsynaptic scaffolding protein GKAP/SAPAP ⁶⁶ , although, in this case, the actual degradation of synaptic GKAP seems to occur only after transporting it away from synapses to centralized locations. Interestingly, local proteasomal activity also seems to play important roles at very early stages of synapse formation, namely during the extension of nascent dendritic spine precursors ⁹⁸ . On the presynaptic side, UPS-mediated degradation was implicated in activity-dependent changes in global and synaptic levels of Rim and Munc-13 ⁹⁹ and in global levels of Bassoon and liprin-α ⁶⁹ . Intriguingly, the large presynaptic active zone proteins Bassoon and Piccolo have been shown to locally suppress the ubiquitination and degradation of multiple presynaptic proteins, possibly by directly regulating the activity of the E3 ligase Siah1 ¹⁰⁰ (see also ⁷⁵ ).

While local, activity-regulated UPS-mediated degradation of certain synaptic proteins seems to be functionally important, this form of degradation does not generalize to many other synaptic proteins ^{2, 5, 30, 66, 69, 99} . Moreover, the degree to which such local processes affect synaptic protein contents under basal conditions appears to be quite modest. For example, in spontaneously active primary cultures of mouse and rat cortical neurons, pharmacologically suppressing proteasomal function for 10–24 hour did not increase synaptic levels of PSD-95 and PSD-93 (24 hours ⁶⁶ ), Munc13-1 (10 hours ¹⁰¹ ), PSD-95, Shank3/ProSAP2, SV2A, Synapsin I, or Bassoon (10–24 hours ³³ ) and only slightly increased synaptic Rim levels ³³ (incidentally, the half-life of Rim1 was estimated here to be ~3 days, which is more compatible with the spatiotemporal constraints discussed above as compared to a prior estimate of 0.7 hours ³¹ ). Finally, as mentioned above, suppression of proteasomal activity did not significantly slow constitutive degradation rates of most synaptic proteins ³³ . It is thus possible that local, activity-regulated UPS-mediated degradation mainly serves to shape synaptic properties in spatially and temporally constrained manners ²⁵ , with the constitutive degradation of most synaptically residing proteins relegated to other pathways.

An alternative catabolic route, which is receiving increasing attention in the context of synaptic protein degradation, is autophagy. A major form of autophagy is macroautophagy, a primary mechanism used by cells to degrade cytosolic complexes and membranous organelles such as mitochondria. Here, double-membraned structures engulf portions of cytoplasm to form autophagosomes, which ultimately fuse with lysosomes, where their contents are degraded. Mutant mice deficient for autophagy-specific genes in the nervous system (e.g. ^{102– 105} ) suffer from progressive neurological deficits, accumulation of abnormal cytoplasmic inclusions in neurons, and ultimately neuron loss and premature death. Along these lines, macroautophagy has been suggested to play important roles in removing cytosolic components and membranous organelles from remote axonal regions and targeting them for somatic degradation ^{106– 108} . Similarly, roles for autophagy in presynaptic proteostasis, axonal membrane homeostasis (e.g. ^{104, 109– 113} ), and even degradation of entire synaptic vesicles (“vesiculophagy”) have been suggested ¹¹⁴ . Finally, the active zone protein Bassoon, previously shown to locally suppress UPS-mediated synaptic protein degradation ¹⁰⁰ (see above), has recently been shown to also control presynaptic autophagy through interactions with Atg5, an E3-like ligase essential for autophagy ¹¹⁵ .

On the postsynaptic side, macroautophagy has been implicated in (activity-induced) degradation of a number of synaptic proteins (reviewed in ¹¹⁶ ) such as α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)-receptor subunits ¹¹⁷ , whereas genetic suppression of macroautophagy was shown to lead to an excess of excitatory synapses ¹¹⁸ . Interestingly, macroautophagy has been studied mainly in the context of nutrient starvation and other forms of stress, but neither starvation nor pharmacological inhibition of mammalian target of rapamycin seem to induce overt macroautophagy in neurons ¹⁰⁷ (but see ¹¹⁷ ). This would be congruent with a constitutive role for macroautophagy in synaptic proteostasis ¹⁰⁷ (see also ¹¹⁹ ; reviewed in ¹⁵ ), although the flux and identities of synaptic proteins or protein complexes degraded in this fashion remain largely unknown.

A related degradation pathway is endosomal microautophagy, a process through which late endosomes directly engulf cytoplasmic material. A recent study suggests that this process may play a key role in the degradation of numerous synaptic proteins, at least in Drosophila ¹²⁰ . In this remarkable study, the chaperone Hsc70-4, initially identified by its ability to deform membranes, was shown to play a crucial role in microautophagy and degradation of Comt/NSF, Unc-13, and EndophilinA and the formation of multivesicular bodies (MVBs) containing multiple synaptic vesicles. Furthermore, a bioinformatic survey of about 170 rat synaptic proteins suggested that 53% of these contain at least one microautophagy motif. The involvement of Hsc70-4 in this process is interesting not only because of its function as a synaptic chaperone but also because it is a central player in a third type of autophagy known as chaperone-mediated autophagy (CMA), in which complexes of Hsc70 and target proteins are selectively translocated across the lysosomal membrane and subsequently degraded. Although CMA was deemed unlikely in Drosophila ¹²⁰ , the possibility that CMA as well as microautophagy play important roles in mammalian synaptic protein degradation remains intriguing ¹⁶ .

Another interesting finding in the aforementioned study concerns the fate of synaptic transmembrane proteins. No evidence was found for microautophagy-mediated degradation of three such proteins (Synaptotagmin1, Syntaxin1A, and the vesicular glutamate transporter Vglut). This, and the observation that synaptic transmembrane proteins harbor fewer microautophagy motifs, led the authors to suggest that synaptic transmembrane protein degradation is mediated by “classical” lysosomal pathways, in agreement with prior studies regarding synaptic vesicle-associated protein degradation in flies ^{121, 122} . Here, evidence was provided that endosomes served as sorting stations for synaptic vesicle proteins endocytosed as part of the synaptic vesicle cycle, sending dysfunctional proteins to lysosomes for degradation. Furthermore, the removal of such presumably dysfunctional proteins was associated with improved presynaptic functionality, a conclusion later supported by a study on the consequences of impaired retrograde transport of presynaptic endosomal cargos ¹²³ . Targeting to lysosomes seems to involve ubiquitination and the formation of MVBs whose formation depends on the endosomal sorting complex required for transport machinery. The universality of this presynaptic protein sorting and degradation process is supported by a recent study carried out in rat hippocampal neurons ⁵¹ . Intriguingly, this study suggests that different synaptic vesicle proteins are degraded at different rates (SV2, Synaptotagmin1 » VAMP2 » Synaptophysin, Vglut1) and in manners differentially affected by manipulations of network activity levels, arguing against the turnover of synaptic vesicles as discrete units (as discussed above) and further highlighting the importance of sorting processes in the specific regulation of particular synaptic protein degradation.

On the postsynaptic side, considerable evidence suggests that AMPA receptors (transmembrane synaptic proteins) are continuously cycled between the plasma membrane and endosomal compartments; interestingly, all four AMPA-receptor subunit types (GluA1–4) can undergo activity-dependent ubiquitination, which is thought to determine whether internalized AMPA receptors are targeted for lysosomal degradation or recycled back to the membrane (reviewed in ¹²⁴ ; see also ¹²⁵ ). Multiple studies have demonstrated a crucial role for the E3 ligase Nedd4 in this form of regulatory ubiquitination ^{63, 126– 129} . Proteasomal inhibitors were shown to strongly slow Nedd4 degradation rates ³³ , perhaps explaining the prominent effects of proteasomal inhibitors on AMPA receptor surface expression ¹³⁰ .

In common with AMPA receptors, ubiquitination of GABA _A receptors (more specifically, their γ subunits) seems to determine whether internalized GABA _A receptors are targeted for lysosomal degradation or recycled back to the membrane ¹³¹ . GABA _B receptor degradation seems to follow a similar route ¹³² . Collectively, these and additional findings point to the importance of endosomal sorting and lysosomal degradation in the catabolism of synaptic transmembrane proteins, in particular synaptic vesicle proteins, postsynaptic receptors, and possibly others (e.g. ^{133, 134} ; reviewed in ⁷⁰ ).