We have demonstrated that the biochemical components of a specific synapse type from a particular neuronal population can be identified using a combination of genetically engineered mice, affinity purification, and MS. Using our approach, we have prepared a fraction enriched in PF/PC PSDs and identified 65 proteins classified in 11 different functional categories. This dataset provides information on signaling pathways specifically tethered to this synapse, as exemplified by our functional analysis of MRCKγ. It also provides information on the variety of proteins that can be part of the code defining the PF/PC synapse.
Approximately 700 different proteins have been identified in PSD preparations from whole brain [16
]. However, it has been estimated that, given the mass of a single PSD, the copy number of scaffold proteins in a PSD, and an average size of 100 kDa for each synaptic protein, only about 100 different proteins can be expected to be found at one particular type of PSD [34
]. The number of proteins we find in our study is consistent with that estimate. Although our analysis may not have revealed all PF/PC postsynaptic proteins, the successful identification of AMPA receptor subunits in our preparations suggests that any proteins not detected in our sample must be present at low stochiometries in the PSD.
Synaptic protein profiling can reveal novel sets of proteins that allow formulation of specific hypotheses regarding synaptic function. For example, we discovered MRCKγ at PF/PC synapses: this kinase is part of a family that has never been described at synapses. This result was striking since MRCK proteins can respond to small GTPases signaling and have been shown to modulate actin cytoskeleton and cell morphology in nonneuronal systems [31
]. These characteristics immediately suggest a role for these kinases in spine morphogenesis, which we have now shown for MRCKγ using transfection of cultured cortical neurons. Taken together, these data also have implications for the study of neurodevelopmental diseases. Deficiencies in spine length and spine morphology in Purkinje cells have been found in models of mental retardation and Angelman syndrome [35
]. Given the link between another small GTPase-dependent kinase, PAK3, and mental retardation [37
], our results suggest that MRCKγ could participate in the signaling pathways involved in mental retardation and autism spectrum disorders.
Another interesting finding of our study is the presence of a high proportion of proteins involved in phospholipid metabolism and signaling at the PF/PC PSD. A major regulator of the physiology of the PF/PC synapse is the metabotropic glutamate receptor 1 (mGluR1) which induces phosphatidylinositol-4,5-P2 (PIP2) hydrolysis through activation of phospholipase C [8
]. Our results show the presence of MRCKγ and Itpr1 in affinity-purified PF/PC PSDs: these proteins can respond to, respectively, DAG and IP3, which are the metabolites of PIP2 hydrolysis. This further supports the importance of mGluR1 signaling at the PF/PC synapse and extends the number of regulatory pathways potentially activated by mGluR1. Also included in the “phospholipid signaling and metabolism” category in our data are synaptojanin-1 and −2, two PIP2-metabolizing enzymes. These enzymes are best known for their regulation of vesicle recycling at synapses, but have also been found by other biochemical studies at PSDs [16
]. Phospholipid metabolism is known to be critical for the function of the presynaptic side of the synapse, especially vesicle recycling [38
]. It also plays a role in defining the boundaries of the apical pole and the localization of tight junctions in epithelial cells [39
]. Our results suggest that phospholipid signaling also participates in regulating the structure and stability of PSDs. Given the fact that lithium is used as a treatment for schizophrenia and bipolar disorders, and that it might act by modulating phospholipids' metabolism [40
], our results may be particularly relevant for studies of a variety of human neurological disorders. Indeed, it has been suggested that synaptojanin-1 is involved in the cognitive defects observed in Down syndrome [41
], and that PIP2 metabolism may be linked to synaptic dysfunctions in Alzheimer disease [42
The results presented here provide clues to the nature of the “synaptic code” and the types of molecules that may be critical in definition of specific synapse types. As expected from previous studies [2
], we find proteins with classical adhesion domains such as Neph1 and the receptor tyrosine phosphatase RPTPmu. SYG1, the Caenorhabditis elegans
homolog of Neph1, has been shown to define synapse location in vivo [6
], and may play a similar role for the PF/PC synapse. Receptor tyrosine phosphatases play important roles in axon guidance, and have also been shown to control synapse formation [43
]. We also find proteins at the PF/PC synapse with as yet unknown functions in synaptogenesis, such as the BAI receptors or GABA-B receptors. In this regard, it is interesting to note that the GABA-B receptor 1 contains a CCP module in its extracellular domain. This module is also found in proteins of the complement cascade, which have recently been shown to be involved in synapse development [44
]. These proteins, and the majority of the remaining proteins identified in this study, are specifically expressed in Purkinje cells within the cerebellum (see Results
). Since cerebellar granule cells also receive excitatory inputs from mossy fibers, we can conclude that, even within the cerebellum, the synaptic codes for specific synapse types must be quite distinct. This supports the results of expression analysis of proteins identified in bulk synapse preparations showing that receptors and other upstream signaling molecules have a highly variable expression pattern in the vertebrate brain [17
]. Taken together, these data indicate that very different sets of molecules must define different excitatory synapse types.
Although our approach employed the expression of a fusion of GluRδ2 with EGFP in a specific cell type, this basic approach can readily be adapted to characterize a wide variety of synapse types, given the wide range of affinity tags that are now available and the hundreds of BAC vectors that can be used to target expression to specific neurons (http://www.gensat.org
). We anticipate that these additional studies of the biochemical diversity of synapses will be critical for understanding the development and function of specific CNS circuits and their dysfunction in disease [45