Research Article

Targeted Protein Degradation and Synapse Remodeling by an Inducible Protein Kinase

See allHide authors and affiliations

Science  21 Nov 2003:
Vol. 302, Issue 5649, pp. 1368-1373
DOI: 10.1126/science.1082475

Abstract

Synaptic plasticity involves the reorganization of synapses at the protein and the morphological levels. Here, we report activity-dependent remodeling of synapses by serum-inducible kinase (SNK). SNK was induced in hippocampal neurons by synaptic activity and was targeted to dendritic spines. SNK bound to and phosphorylated spine-associated Rap guanosine triphosphatase activating protein (SPAR), a postsynaptic actin regulatory protein, leading to degradation of SPAR. Induction of SNK in hippocampal neurons eliminated SPAR protein, depleted postsynaptic density–95 and Bassoon clusters, and caused loss of mature dendritic spines. These results implicate SNK as a mediator of activity-dependent change in the molecular composition and morphology of synapses.

Synaptic activity can induce a variety of changes within postsynaptic neurons, ranging from transient posttranslational modifications to altered programs of gene expression. Long-lasting forms of synaptic plasticity require new gene expression and protein synthesis (13). Some activity-inducible genes may mediate the conversion of short-term responses to long-term changes by altering synaptic structure (4, 5).

Numerous activity-inducible genes have been identified (6, 7). Notably, few protein kinases are known to be inducible by synaptic activity at the mRNA level; the best-characterized induction profiles are those of SNK and FGF-inducible kinase (FNK) (8). However, the roles of these polo family kinases in plasticity are unknown.

One reasonable expectation is that synaptic remodeling will involve the dismantling and/or reorganization of key cytoskeletal and scaffolding protein complexes. In the postsynaptic density (PSD) of mammalian excitatory synapses, actin is the major cytoskeletal element, and scaffold proteins of the PSD-95 family are important for assembling glutamate receptors with their signaling-cytoskeletal complexes (9, 10). One PSD-95–interacting partner, SPAR [spine-associated Rap guanosine triphosphatase (GTPase) activating protein (GAP)], is a multidomain postsynaptic protein that controls dendritic spine shape by regulating actin arrangement as well as signaling by the small GTPase Rap (11). Actin dynamics and Rap activity are both regulated by synaptic activity and involved in synaptic plasticity (1216). With its postsynaptic location in the N-methyl-d-aspartate (NMDA) receptor–PSD-95 complex, SPAR is an attractive candidate for mediating activity-dependent remodeling of synapses.

Interaction of SPAR and SNK. We screened for SPAR-binding proteins with the use of individual domains of SPAR (Fig. 1A) as bait in the yeast two-hybrid system (17). SPAR contains two actin regulatory domains, termed Act1 and Act2, a GAP domain specific for Rap, a PDZ domain of unknown function, and a C-terminal region (termed GKBD) that binds specifically to the guanylate kinase domain of PSD-95 (11). When the Act2 domain was used to screen a brain cDNA library, one of the positive clones (clone 19) isolated was SNK, initially identified in fibroblasts as an mRNA transcript induced by mitogenic stimulation (18). Clone 19 encoded roughly the C-terminal half of SNK protein (amino acids 395 to 682; hereon termed SNKc) (Fig. 1A), including a motif characteristic of the polo family of kinases (19, 20). Full-length SNK also bound to Act2 in the yeast two-hybrid assay. Neither full-length SNK nor clone 19 interacted with the GKBD region of SPAR (Fig. 1A).

Fig. 1.

Interaction between SNK and SPAR. (A) Schematic diagram showing domains (with amino acid coordinates) of SPAR and SNK proteins (3′ untranslated region of clone 19 representedby a black line) and table showing domain interactions of SNK, FNK, and SPAR in the yeast two-hybrid system using the constructs indicated. (B) GST precipitation assay. Lysates prepared from COS-7 cells expressing myc-tagged SPAR, SPARΔAct2, or Act2 domain alone were incubated with purified GST or GST-SNKc fusion protein, precipitated with glutathione-sepharose, and immunoblotted with myc antibody. IN, 10% of input lysate. (C) Coimmunoprecipitation of SNK and SPAR. Lysates from COS-7 cells cotransfected with full-length SNK(K108M) and SPAR were immunoprecipitated with nonimmune rabbit IgG or rabbit polyclonal SPAR antibodies and immunoblotted for SPAR and SNK. IN, 10% of input lysate. Molecular weights shown in kD.

The interaction between SNK and SPAR was confirmed with the use of an in vitro precipitation assay in which GST fused to SNKc precipitated full-length SPAR expressed in COS-7 cells but not a SPAR construct with an internal deletion of the Act2 domain (SPARΔAct2) (Fig. 1B). The isolated Act2 domain also bound GST-SNKc with greater efficiency than full-length SPAR. GST alone failed to bind any of these SPAR constructs.

To test for an association between SPAR and SNK in a mammalian cellular environment, we expressed full-length SPAR and SNK in COS-7 cells. Antibodies against SPAR, but not nonimmune immunoglobulin G (IgG), immunoprecipitated SPAR protein and also coimmunoprecipitated SNK (Fig. 1C). For this experiment, it was necessary to use a point mutant of SNK [Lys108→Met108 (K108M)] that renders the kinase domain catalytically inactive, because cotransfection of wild-type SNK caused a dramatic reduction in the amounts of SPAR protein.

Degradation of SPAR by SNK. SPAR transfected into COS cells in the absence of SNK was readily detected as a band of ∼190 kD. However, when cotransfected with increasing amounts of wild-type SNK expression plasmid, SPAR levels decreased in a dose-dependent manner until virtually undetectable (Fig. 2A). Cotransfected green fluorescent protein (GFP) expressed from the same expression vector was unaffected by SNK (Fig. 2A). Moreover, SNK did not alter the protein expression of several other cotransfected complementary DNAs (cDNAs) in COS cells, including Cib (an SNK-interacting protein) (8), PSD-95, and liprin-α1 (Fig. 2B). Thus, the effect of SNK appeared relatively specific for SPAR, although SPAR is unlikely to be the only substrate of SNK in neurons. Cotransfection of the closely related polo kinase FNK, which also binds to SPAR in yeast two-hybrid assays (Fig. 1A), was similarly effective in reducing SPAR levels, but the unrelated kinase calcium/calmodulin-dependent protein kinase II (CaMKII) did not alter SPAR expression (Fig. 2C).

Fig. 2.

Degradation of SPAR by SNK. (A to E) Whole-cell lysates from COS-7 cells transfected with various combinations of plasmids (as indicated) were immunoblotted for the indicated proteins. (A) COS-7 cells were triple-transfected with SPAR (0.6 μg), SNK (0.1, 0.2, or 0.6 μg), and GFP (0.6 μg) expression plasmids, as indicated. Each construct contained an N-terminal myc epitope tag, allowing simultaneous immunoblotting of all exogenous proteins. Endogenous β-tubulin served as internal control for protein loading. (B) SNK (0.1, 0.2, or 0.6 μg) was cotransfected pair-wise with SPAR, Cib, PSD-95, or liprin α1 (0.6 μg). (C) COS cells were cotransfected with SPAR and either FNK or a constitutively active [Thr286→Asp286 (T286D)] CaMKII mutant. (D) SPAR (0.6 μg) was cotransfected with SNK (0.2 or 0.6 μg) or kinase-dead SNK (K108M) (0.6 μg). Immunoblot is shown at higher magnification to highlight the electrophoretic mobility shift of SPAR and SNK proteins that is dependent on SNK kinase activity. (E) (Left) Anti-myc immunoblot of myc-tagged SNK and SNK(K108M) immunoprecipitated from COS cells. (Right) In vitro kinase reactions of immunoprecipitated SNK and SNK(K108M) with hexahistidine-tagged SPAR purified from COS cells. Asterisk indicates SNK autophosphorylation band; arrowhead, SPAR phosphorylation band. (F) Cells were cotransfected with SPAR and either SNK or SNK(K108M), as indicated, and treated with dimethyl sulfoxide (control), leupeptin (100 μg/ml), MG132 (50 μM), or lactacystin (50 μM). Bracket indicates the higher molecular weight smear of SPAR that appears upon proteasomal inhibition. (G) COS cells were cotransfected with SPAR, myc-tagged ubiquitin (myc-Ubq), and either SNK or SNK(K108M), as indicated. Lysates were immunoprecipitated with SPAR antibodies and immunoblotted for myc. IN, 5% of input lysate. All molecular weights shown in kD.

Cotransfection of SNK also induced a mobility shift in the SPAR protein on SDS–polyacrylamide gel electrophoresis (SDS-PAGE) from a predominantly single band to a smear of higher molecular weight forms (Fig. 2D, lane 2), consistent with phosphorylation of SPAR by SNK. The kinase-dead form of SNK (K108M) had no effect on either steady-state level or electrophoretic mobility of SPAR. Wild-type SNK itself was observed as a doublet, whereas the K108M mutant exhibited only a single band (Fig. 2D), suggesting that SNK undergoes autophosphorylation. Hexahistidine-tagged SPAR protein purified from COS cells was phosphorylated in vitro by SNK but not by SNK(K108M), indicating that SPAR can be a direct substrate of SNK (Fig. 2E).

E6TP1 (E6 target protein 1), the human homolog of rat SPAR, has been characterized as a target for degradation by the human papillomavirus E6 oncoprotein via the ubiquitinproteasome pathway (21). To test whether the proteasome is involved in SNK-mediated reduction of SPAR levels, we treated SPAR/SNK-transfected COS cells with various inhibitors (Fig. 2F). Leupeptin, an inhibitor of lysosomal degradation, did not prevent the elimination of SPAR by wild-type SNK. In contrast, MG132 and lactacystin, two structurally distinct inhibitors of proteasome-mediated degradation, caused the appearance of heterogeneous higher molecular weight SPAR products in SNK-transfected cells, which may reflect accumulation of polyubiquitinated intermediates. To show directly that SPAR can be ubiquitinated and that this process is stimulated by SNK, we coexpressed myc-tagged ubiquitin (22) with SPAR and SNK. Immunoprecipitated SPAR contained myc immunoreactivity, indicating that SPAR was indeed ubiquitinated, and the level of myc-ubiquitinated SPAR increased in the presence of SNK (Fig. 2G), despite the fact that the total amount of SPAR protein was greatly decreased (Fig. 2, A, B, and D).

We also analyzed the effect of SNK on SPAR protein levels and function by immunocytochemistry (Fig. 3). Expression of SPAR in COS cells, as reported previously (11), dramatically reorganizes the actin cytoskeleton into irregular rodlike or star-shaped clusters, in which exogenous SPAR is colocalized (Fig. 3A). When transfected by itself, SNK was diffusely distributed in COS-7 cells (Fig. 3B3) and had little effect on the pattern of F-actin (Fig. 3B1). However, when SNK and SPAR were cotransfected, SPAR immunoreactivity was greatly reduced (Fig. 3D2) as compared to cells transfected with SPAR only (Fig. 3A2) or cotransfected with SPAR and SNK(K108M) (Fig. 3C2). SPAR average immunofluorescence intensity was 98.57 ± 14.96 arbitrary units/μm2 (mean ± SEM) when cotransfected with K108M, compared with 19.04 ± 11.01 when transfected with SNK (P < 0.005, t test; see Fig. 3E for quantitation). In most cells containing wild-type SNK, the level of SPAR was nearly at background levels (23); however, in some cells SPAR was still detectable as puncta that colocalized with F-actin filaments (Fig. 3, D2 and F). Importantly, in cells cotransfected with SNK and SPAR, the actin cytoskeleton appeared normal even when some SPAR puncta remained (Fig. 3, D1 and F).

Fig. 3.

SNK inhibits SPAR-mediated reorganization of actin cytoskeleton. COS cells were transfected with (A) myc-tagged SPAR alone, (B) hemagglutinin (HA)-tagged SNK alone, (C) myc-SPAR and HA-SNK(K108M), and (D) myc-SPAR and HA-SNK. Cells were immunostained with myc and HA antibodies and counterstained with phalloidin-oregon green to visualize F-actin. Individual channels (green, red, and blue as indicated at top) are shown in gray scale; merged image is shown in color in right-most column (colocalization of green, red, and blue appears white). (E) Distribution of fluorescence intensities of SPAR immunoreactivity per cell cotransfected with SNK (blue) or SNK(K108M) (red). (F) Higher magnification view of the boxed region in (D4), showing remaining SPAR puncta associated with F-actin. Scale bar, 10 μm.

Both the reduction in SPAR staining and the prevention of actin rearrangement by SPAR required an active kinase domain in SNK, because the SNK(K108M) mutant had no effect on SPAR clustering or SPAR-mediated reorganization of F-actin (Fig. 3C). However, we noted that SNK(K108M) colocalized extensively with SPAR/F-actin clusters (Fig. 3C4), consistent with a direct interaction between SNK and SPAR and suggesting that SPAR can recruit SNK to subcellular structures.

SNK promotes synapse loss. To determine the effect of SNK on endogenous SPAR in neurons, we infected cultured hippocampal neurons with Sindbis virus expressing SNK or SNK(K108M) and analyzed the expression pattern of SPAR by immunocytochemistry. The exogenous SNK protein, immunostained by its N-terminal FLAG epitope tag, appeared punctate and enriched in dendritic spines (Fig. 4A1, inset), although at high expression levels it filled the dendritic shafts as well. In SNK-infected neurons, SPAR immunoreactivity was virtually undetectable, whereas neighboring uninfected spiny neurons exhibited the typical punctate SPAR staining along dendrites (Fig. 4A2) (11). The SNK(K108M) mutant also concentrated in spines but had little effect on SPAR immunoreactivity, which remained brightly punctate and indeed colocalized with K108M in dendritic spines (Fig. 4B) (SPAR immunofluorescence intensity with K108M was 1.07 ± 0.07 arbitrary units/μm2, mean ± SEM, compared to 0.39 ± 0.03 with wild-type SNK, P < 0.001, t test, normalized to nearby uninfected control neurons).

Fig. 4.

SNK induces loss of SPAR, PSD-95, and Bassoon in hippocampal neurons. Hippocampal neuron cultures were infected at DIV18 to DIV21 with Sindbis virus expressing SNK [(A), (C), (E), and (F)] or SNK(K108M) [(B), (D), and (G)]. Cells were coimmunostained for exogenous SNK and endogenous SPAR, PSD-95, or Bassoon, as indicated, with merged image shown in color (right column). For (A), (B), and (E), a higher magnification view is also shown of the dendritic segment enclosed in white box. For (C), (D), (F), and (G), only the high magnification views of dendrites are shown. Asterisk in (A) indicates an untransfected neuron. Scale bars, 10 μm.

To address whether SNK has more general effects on postsynaptic structure, we analyzed the distribution of PSD-95, a major scaffold protein of the PSD. In virtually all SNK-infected neurons, PSD-95 immunostaining was much reduced compared with uninfected cells (Fig. 4C) or cells infected with SNK(K108M) (Fig. 4D). The effect of SNK on PSD-95 levels is probably indirect, because SNK did not cause PSD-95 degradation in heterologous cells (Fig. 2B), in either the presence or the absence of SPAR (23). At the time point examined (18 to 24 hours postinfection), we also observed in the majority of cells that the loss of punctate PSD-95 immunoreactivity was most pronounced in proximal dendrites but relatively spared in distal dendrites (Fig. 4E). We speculate that the elimination of PSD clusters as a function of distance from the soma results from the earlier “capture” of newly induced SNK by proximal spines and synapses. Immunostaining intensity for Bassoon (a presynaptic marker protein) also decreased in proximal dendrites of SNK-infected neurons as compared with SNK(K108M)-infected or uninfected neurons (Fig. 4, F and G), consistent with an overall loss of synapses.

Because SPAR promotes spine growth and elaboration (11), we reasoned that SNK-induced loss of SPAR should cause morphological changes in spines. SNK-infected neurons showed a decrease in the density of mature dendritic spines and a corresponding increase in immature filopodia (Fig. 5, B and E). Filopodia comprised 1.5 ± 0.3% of total protrusions in control GFP-transfected neurons, compared to 32.9 ± 2.9% for SNK-transfected and 3.1 ± 0.7% for K108M-transfected neurons (P < 0.001 SNK versus GFP). Accordingly, SNK-infected neurons showed a striking increase in length of dendritic protrusions (compare Fig. 5, A and B; quantified in Fig. 5D). The average dendritic protrusion length in SNK-infected cells increased to 1.92 ± 0.09 μm from 1.09 ± 0.03 μm (GFP control, P < 0.001). In contrast, the kinase-dead SNK(K108M) had little effect on spine length (Fig. 5D) but increased the density of dendritic spines to about double the control values (Fig. 5, C and E). The mechanism for this surprising spine-promoting effect is unclear but may involve either an independent function of SNK unmasked by the inactivation of the kinase domain or a dominant-interfering effect of the K108M mutant on endogenous, ongoing SNK activity that promotes spine loss. The effect of SNK(K108M) on spine numbers is unlikely to be solely a result of an elevated level of SPAR, because overexpression of SPAR does not increase spine density (11). The loss of mature spines accompanied by formation of filopodia seen with wild-type SNK is similar to the effects of dominantnegative SPAR overexpression in neurons (11), suggesting that the morphological effects induced by SNK may be mediated in large part by elimination of SPAR.

Fig. 5.

SNK induces loss of dendritic spines and increased filopodia. Hippocampal neurons were infected with Sindbis virus expressing GFP (A) or wild-type SNK (B) or SNK(108M) (C), as indicated. Neurons were stained with antibodies against GFP [(A)] or SNK [(B) and (C)]. Representative dendritic segments (boxed) are shown below at higher magnification. (D) Cumulative frequency distribution of dendritic protrusion lengths on neurons infected with GFP, wild-type SNK, or SNK(K108M), as indicated. (E) Density of filopodia and spines. Asterisk, P < 0.05 compared to total SNK protrusions; double asterisks, P < 0.001 compared to GFP control. Scale bars, 10 μm.

Induction of endogenous SNK. The above experiments mimicked SNK induction by viral transfection of the cDNA. What is the effect of induction of endogenous SNK in neurons? Basal levels of SNK are barely detectable by western blot in unstimulated hippocampal cultures (Fig. 6L). Previous studies showed that SNK mRNA expression is induced in brain by electrical stimulation or drug-induced seizures (8). We found that picrotoxin (100 μM, 18 to 24 hours), an inhibitor of γ-aminobutyric acid type A receptors, potently induced SNK protein levels in dissociated hippocampal culture (Fig. 6L). Bath application of glutamate (20 μM) and depolarization by KCl (55 mM) also increased SNK levels, but to a lesser degree (23). Induced SNK was observed as a doublet at ∼80 kD, comigrating with recombinant SNK expressed in COS cells (Fig. 6L). The degree of induction of SNK by picrotoxin increased from day in vitro (DIV) 7 to DIV24, correlating with synapse formation and maturation over this period (23).

Fig. 6.

Synaptic activity induces endogenous SNK and leads to loss of SPAR and PSD-95. (A to H) Hippocampal cultures (DIV24) stimulated with picrotoxin (100 μM, 18 hours) [(C), (D), (G), and (H)], or unstimulated[(A), (B), (E), and (F)] were double-stained for SNK (red) and either SPAR or PSD-95 (green), as indicated, with merge in color. [(B), (D), (F), and (H)] Higher-magnification views of dendrite segments from (A), (C), (E), and (G) (boxed regions). (G1, inset) Higher-magnification view of boxed region in red showing spine enrichment of endogenous SNK (arrowhead). Scale bars, 10 μm. (I) Integrated intensity of PSD-95 puncta as a function of distance from the cell body in uninfected neurons or neurons infected with Sindbis-SNK or SNK(K108M). Asterisk, P < 0.001 for SNK versus K108M. (J) Integrated intensity of Bassoon puncta as a function of distance from the cell body in uninfected neurons or neurons infected with Sindbis-SNK or SNK(K108M). Asterisk, P <0.05 for SNK versus K108M. (K) Integrated intensity of SNK immunoreactivity and SPAR and PSD-95 puncta as a function of distance from the cell body in unstimulated and picrotoxin-stimulated neurons, normalized to the highest value for a given antibody. Asterisk, P < 0.001 for stimulated versus unstimulated. (L) Hippocampal cultures (DIV24) were treated with picrotoxin (100 μM) for 24 hours in combination with APV (50 μM), CNQX (40 μM), nimodipine (nimo, 5 μM), FK506 (2 μM), or ascomycin (asco, 2 μM) as indicated, or with MG132 (50 μM) or lactacystin (10 μM) in the absence of picrotoxin. Total lysates were immunoblotted for SNK and CaMKII. Histograms show quantitation of SNK levels normalized to unstimulated level (mean ± SEM of three experiments).

Because picrotoxin indirectly increases excitatory synaptic transmission, we tested whether glutamate receptors were required for the induction of SNK. Indeed, a combination of APV (50 μM) and CNQX (40 μM), antagonists of NMDA and AMPA receptors, respectively, blocked SNK induction (Fig. 6L). Inhibition of L-type voltage gated calcium channels (VGCCs) by nimodipine (5 μM) also prevented SNK induction, as did FK506 (2 μM) and ascomycin (2 μM), two independent inhibitors of the calciumregulated protein phosphatase PP2B (also called calcineurin) (Fig. 6L). These results suggest that picrotoxin-mediated SNK induction is triggered by calcium entry through VGCCs, presumably secondary to depolarization via activation of ionotropic glutamate receptors. Inhibition of proteasomes by either MG132 (50 μM, 24 hours) or lactacystin (10 μM, 24 hours) strongly elevated SNK levels, even in unstimulated cultures (Fig. 6L).

We found little staining of SNK in unstimulated neurons by immunocytochemistry (Fig. 6, A1 and E1). After picrotoxin stimulation, endogenous SNK immunoreactivity increased specifically in the somatodendritic compartment, with a preferential concentration in more proximal dendrites at the time point examined (18 to 24 hours after stimulation) (Fig. 6, C1 and G1). The native induced SNK was also enriched in dendritic spines (Fig. 6G1, inset), consistent with it being targeted to synapses by an interaction with postsynaptic proteins such as SPAR.

The same picrotoxin-stimulated neurons were immunostained for endogenous SPAR and PSD-95. In unstimulated cells, SPAR puncta (Fig. 6, A2 and B2) and PSD-95 puncta (Fig. 6, E2 and F2) were evenly distributed along dendrites. Picrotoxin-treated cells showing high levels of SNK in proximal dendrites exhibited a loss of both SPAR (Fig. 6, C2 and D2) and PSD-95 (Fig. 6, G2 and H2) in the same regions, leading to a complementary distribution of SNK and SPAR/PSD-95 in activated neurons. Loss of SPAR and PSD-95 was most marked in the proximal dendrites, where SNK immunoreactivity was highest (Fig. 6K). The loss of SPAR was more severe than for PSD-95 (Fig. 6K). The similarity between this effect and that of specifically overexpressing recombinant SNK (quantified in Fig. 6, I and J) suggests that SNK induction may be a major mechanism underlying the elimination of SPAR and PSD-95 after picrotoxin stimulation.

Ubiquitination and proteasome-mediated proteolysis play important roles in synapse development and plasticity (24, 25). However, little is known about the specific mechanisms and targets of ubiquitination in synapses and how they are regulated by activity. Our study reveals a mechanism for structural plasticity of synapses in which an activity-inducible kinase (SNK) is targeted to spines and executes the degradation of a key PSD protein (SPAR) involved in actin remodeling, Rap signaling, and spine morphogenesis (11).

An attractive possible mechanism for the enrichment of SNK in spines is through direct binding to SPAR. We propose that the physical interaction between SPAR and the kinase selects SPAR (and possibly other nearby postsynaptic proteins) for phosphorylation by SNK and that this phosphorylation leads to specific degradation of SPAR by the ubiquitin-proteasome pathway. Although the related polo-like kinase has been shown to phosphorylate and activate anaphase-promoting complex, a multisubunit ubiquitin ligase (E3) (26), as well as the proteasome itself (27), we found no stimulation of general proteasome activity by overexpression of SNK in hippocampal cultures (fig. S1). Interestingly, SNK itself appears to be a substrate for proteasomal degradation; this instability as a protein allows for finer temporal control of SNK action.

As a result of elimination of SPAR (and perhaps other unidentified synaptic substrates of SNK), PSDs and synapses are lost, leading to depletion of mature dendritic spines and formation of filopodia (generally considered the exploratory precursors of spines). Elimination of SPAR by SNK is associated with the loss of the PSD marker PSD-95. Because SPAR is a large multimodular scaffold protein of the PSD with the ability to regulate the actin cytoskeleton and Rap, it is quite plausible that SPAR degradation would lead indirectly to dismantling of the PSD and depletion of PSD-95. Reduction in PSD-95 could contribute to loss of mature spines on the basis of the observation that PSD-95 overexpression increases spine size and number (28). Whether the morphological loss of spines is a result of elimination of PSD-95 and associated postsynaptic proteins, or vice versa, is uncertain. Nevertheless, because SPAR is known to be important for spine morphogenesis (11), controlling the levels of postsynaptic SPAR represents a powerful way to control postsynaptic structure.

What might be the physiological role of SNK induction and subsequent synapse remodeling? Given that SNK is induced by activity at the mRNA level in the cell body (i.e., not in a local synapse-specific manner), the relatively long time frame of SNK induction (many hours), and its negative effect on synapses and spines, we propose that SNK may be involved in the homeostatic (“synaptic scaling”) mechanisms that dampen synaptic function after elevated activity (29). In such a negative feedback model, SNK acts globally to set the overall level of activity in a neuron, maintaining stability in the face of local synaptic modifications that result from correlation-based plasticity mechanisms (30).

Supporting Online Material

www.sciencemag.org/cgi/content/full/1082475/DC1

Materials and Methods

Fig. S1

References and Notes

View Abstract

Navigate This Article