Synapses between nerve cells in the mammalian brain are not only extremely numerous but also very diverse with respect to their structural and functional characteristics. This heterogeneity arises despite the fact that a set of common basic protein ‘building blocks’ is shared by many synapses. Among these, postsynaptic scaffolding proteins play a key role. They have the ability to assemble into membrane-tethered lattices and to adopt unique conformational states in different postsynaptic microenvironments, which may represent a key prerequisite of synapse heterogeneity. Analyses of such synaptic superstructures, rather than individual proteins and their interactions, are required to develop a mechanistic understanding of postsynaptic differentiation, synapse diversity, and dynamics.
Neuronal synapses are the key relay and modulation sites of signalling in the central nervous system. Consequently their molecular architecture and the properties of synaptic transmission and plasticity continue to be a major focus in modern neuroscience. Chemical synapses are highly specialized trans-cellular structures made up of plasma membrane specializations from two distinct neuronal compartments in apposition. The nerve cell axon typically contributes the presynaptic apparatus of the synapse, which is specialized for the fast release of neurotransmitters, while the plasma membrane segment apposed to transmitter release sites differentiates to form the receptor-rich postsynaptic signal reception apparatus.
A major challenge in deciphering the structural and functional principles of the postsynaptic apparatus is the vast diversity of synapses themselves; they use different transmitter molecules, display different qualitative and quantitative transmission characteristics, and exhibit distinct modes of plasticity. Despite this variability, there are several components and core organizational principles that appear to be sufficiently adaptable to accommodate for this diversity, making them virtually omnipresent synapse constituents.
The postsynaptic membrane is linked to the presynaptic apparatus through trans-synaptic adhesion systems. The corresponding postsynaptic proteins contact their presynaptic binding partners within the synaptic cleft, span the postsynaptic membrane, and present protein–protein interaction domains to the juxta-membrane cytosol in the vicinity of the synaptic contact. Indeed, members of the neuroligin [1,2], SynCAM (synaptic cell adhesion molecule) , NGL (netrin-G ligand) , or LRRTM (leucine-rich repeat transmembrane)  families of adhesion proteins can locally induce postsynaptic differentiation. This property indicates that the positioning of these proteins on the nascent postsynaptic membrane spatially demarcates the site of deployment for the postsynaptic apparatus.
Cytosolic-binding sites of postsynaptic adhesion proteins can recruit proteins with multiple protein–protein interaction domains that are thought to function as scaffolds [6–8], upon which the postsynaptic apparatus assembles. Scaffolding proteins, in turn, interact with an array of proteins that are concentrated at the postsynapse, including transmitter receptors . For example, PSD-95 (postsynaptic density protein of 95 kDa), an abundant scaffolding protein of glutamatergic postsynapses, interacts with NMDA (N-methyl-D-aspartate) receptors  and TARP [transmembrane AMPA (α-amino-3-hydroxy-5-methylisoxazole-4-propionic acid) receptor regulatory protein]-associated AMPA receptors  through its tandem PDZ (PSD-95/discs-large/ZO-1) domains, and the associated protein Homer interacts with metabotropic glutamate receptors . On the other hand, gephyrin, the major scaffolding protein of both GABAergic and glycinergic inhibitory synapses, interacts with and clusters both glycine and GABAA [Type A GABA (γ-aminobutyric acid)] receptors [13–16].
In addition to these protein–protein interactions, a very large number of additional interactions are thought to take place in the postsynaptic compartment, which has led to a schematized view of the postsynapse as an extensive network of protein–protein links. Centred on scaffolding proteins, interactions, often mediated through generic protein–protein interaction modules such as PDZ and SH3 (Src homology 3) domains extend from trans-synaptic adhesion molecules to transmitter receptors and signalling molecules. While these generalized models have helped tremendously to rationalize the high concentration of postsynaptic receptors across from presynaptic terminals, they present us with a paradox that confounds our understanding of postsynaptic differentiation. They cannot explain how synapses, built with the same core scaffolds, can differ so profoundly in their transmission and plasticity properties, or how promiscuous protein–protein interactions lead to the formation of highly specialized postsynaptic structures. How can generic components and an ‘anything-goes’ architecture drive the different programmes of postsynaptic differentiation that lead to the plethora of distinct synapse types of the central nervous system?
A blueprint for postsynaptic scaffolds: assembling a membrane-tethered lattice
A common feature of scaffolding proteins is their ability to homo-oligomerize. This arrangement allows the formation of a multimeric mesh that appears to be important for postsynaptic organization. At inhibitory synapses gephyrin can trimerize at its N-terminus  and dimerize at its C-terminus . Through these homotypic interactions, gephyrin can form coherent lattices of regular hexagonal modules  that are rich in binding sites for inhibitory receptor proteins. The protein collybistin can tether the gephyrin lattice to the postsynaptic membrane by acting as a protein-lipid adaptor that binds to both gephyrin  and phosphoinositides . At excitatory glutamatergic synapses PSD-95 is tethered to the membrane through palmitate moieties at its N-terminus , while its C-terminus, which contains glutamate receptor-binding sites, extends vertically into the postsynaptic density . In high-resolution cryo-EM (electron microscopy) tomographic images of the postsynaptic density, PSD-95 columns appear linked to a horizontal lattice parallel to the plane of the postsynaptic membrane . Indeed, the scaffolding protein Homer forms tetramers that, together with Shank (SH3 and ankyrin domain containing protein) oligomers, form an extended lattice . This lattice is tethered to membrane-bound PSD-95 through the protein GKAP (guanylate kinase-associated protein) that functions as an adaptor for Shank and PSD-95 .
In both excitatory and inhibitory scaffolds, the postsynaptic lattice structure depends vitally on the homotypic interaction of the corresponding core scaffolding proteins. Mutations or splice variants that affect the homo-oligomerization state of gephyrin disrupt its postsynaptic accumulation . Similarly, Homer mutants or variants that do not adopt a tetrameric stoichiometry are not properly targeted to excitatory synapses. Indeed, perturbing scaffolding protein homo-oligomerization disrupts postsynaptic differentiation altogether such that receptors do not accumulate and postsynaptic spines do not form . Thus the postsynapse is lost if scaffolding proteins do not form an extended lattice, irrespective of whether the interaction sites that connect adhesion, scaffolding and receptor proteins remain intact. These findings show clearly that the protein–protein interaction network described in the prevailing schemata of postsynaptic organization is insufficient to explain postsynaptic differentiation and assembly of the postsynaptic apparatus. Rather, the driving force of postsynaptic assembly appears to be the establishment of a coherent protein superstructure of membrane-tethered multimeric lattices, i.e. a postsynaptic holo-complex is required rather than chains of one-to-one protein–protein interaction links.
Synaptic configurations and the postsynaptic holo-complex
The notion of a postsynaptic holo-complex implies that complete assembly of the constituents into a regular superstructure is a prerequisite for the complex to form at all. Such a precondition could in fact serve to enforce specificity in a seemingly promiscuously assembled protein complex. Assembly, in this case, would not be driven by the isolated affinity profiles of the various interactions, but rather by the achievement of the thermodynamically most favourable states of the complex as a whole.
Two principal characteristics of postsynaptic architecture are its structural density and the multiplicity of possible interactions among its protein constituents. The first is evident in electron micrographs, where the postsynaptic apparatus appears as an electron-dense thickening of the membrane termed the postsynaptic density. In cellular fractionation procedures, postsynaptic density preparations are enriched from other cellular compartments based on their different density and detergent insolubility . The multiplicity of potential interactions is nicely illustrated by data from protein–protein interaction screens, where promiscuity, rather than specificity, appears to be a common theme [8,28]. This is most probably due to the high prevalence of generic protein–protein interaction domains. With this in mind, the assembly of a postsynaptic holo-complex would resemble something of a LEGO® construction where many pieces fit together, thus allowing for many possible configurations. Steric compatibility of the multiple interaction sites within postsynaptic proteins would impose strict criteria for assembly, thereby limiting the configurations that can persist as stable postsynaptic structures. The assembly configuration of postsynaptic proteins must be permissive for the formation of a regular scaffolding lattice as the structure would disassemble otherwise, and associated proteins would interact with the scaffold and among themselves in a manner that is compatible with the steric limitations of the lattice. The postsynaptic membrane presents an additional dominant structural feature that must be accommodated. Importantly, molecules linking the protein components of the postsynaptic apparatus to its lipid component must adopt an orientation within the complex that allows them access to the membrane.
All these constrains of the postsynaptic milieu would significantly limit the potential ways by which proteins of the postsynapse can interact, in contrast with the many possible combinations of interactions seen in vitro with isolated proteins. Additionally, in favour of the overall thermodynamic stability of the complex, proteins may adopt conformations within the holo-complex that would not be favoured in other environments. Increased avidity of adjoined binding sites on homo-oligomeric components and allosteric effects within the holo-complex could change affinities and cover or uncover binding sites. In other words, proteins may adjust their conformational state as the postsynaptic complex ties itself together, adopting synapse-specific configurations that give rise to new structural and functional properties that are not observed in the cytosol, in solution in vitro or in homogeneous crystals.
An example of a postsynaptic protein that appears to adopt different structures within and outside of the postsynapse is gephyrin. This scaffolding protein multimerizes and forms readily distinguishable superstructures. Splice isoforms or mutations affecting oligomerization lead to the formation of gephyrin structures with distinct morphology, ranging from large cytoplasmic globular deposits termed ‘blobs’ to ‘spike-like’ filaments or small granular aggregates [29,30]. Synaptic gephyrin is distinct from these structures and appears as puncta or slender segments lining the postsynaptic membrane . The various cytoplasmic and the distinct postsynaptic superstructures exemplify the versatility of gephyrin as a structural molecule. In addition, alternative splicing of gephyrin affects its binding of glycine receptors [32,33], while signalling events in gephyrin-containing postsynapses have been implicated in the transition from glycine to GABA receptor-containing postsynapses . These findings indicate that gephyrin assembles in various configurations and, indeed, may have more than one synaptic configuration, thus displaying different properties at distinct synapses or distinct stages of synapse maturation.
The gephyrin-associated molecule collybistin is another example of a protein that appears to acquire distinct properties in its postsynaptic configuration. Collybistin acts as a protein-lipid adaptor by binding to gephyrin  and, via a PH (pleckstrin homology) domain, to membrane lipids . In pyramidal neurons, collybistin is critical for the association of the gephyrin scaffold with the postsynaptic membrane . In heterologous cells that lack synapses, collybistin remains cytoplasmically aggregated with gephyrin. However, N-terminal truncation of collybistin induces aberrant tethering of gephyrin to the plasma membrane, even in the absence of other synaptic components. These observations indicate that the collybistin N-terminus is a negative regulator of the adaptor properties of collybistin. Thus the N-terminal domain prevents collybistin from expressing its synaptic properties out of the synaptic context, whereas within the postsynaptic complex, collybistin adopts a different configuration in which its adaptor properties are no longer inhibited. Indeed, the postsynaptic cell adhesion protein neuroligin 2 activates collybistin at nascent synapses by binding to its inhibitory N-terminus and thereby triggers the formation of a tripartite postsynaptic complex consisting of neuroligin 2, collybistin and gephyrin, which initiates the assembly of the postsynaptic receptor and signalling machinery .
Postsynaptic protein function revisited
The emergence of distinct protein properties specifically within the postsynaptic apparatus has far reaching implications for our understanding of synapse assembly and function. Indeed, the fact that the postsynaptic environment can change the properties of a given protein profoundly extends the versatility of the postsynapse. This way, synapses can display distinct properties that are the product of distinct configurations of postsynaptic proteins, despite being composed of the same core components.
The recruitment of different receptor types is central to postsynaptic diversity. Processes of receptor dynamics and diversity, such as the developmental transition from NR2A to NR2B subunit-containing NMDA receptors in the hippocampus , the activity dependent recruitment of AMPA receptors , the medley of synapse-specific GABA receptor subunit combinations , the existence of mixed GABAergic and glycinergic synapses [39,40] and transitions from GABAergic to glycinergic transmission , all occur with apparently little variation of the core components of the postsynaptic scaffolding. PSD-95 and several related PDZ domain proteins contribute to this type of receptor dynamics at glutamatergic synapses , but whether a differential range of postsynapses with these scaffolding proteins plays a role in synapse diversity remains to be explored.
The solution to this apparent paradox may lie in the configurations of postsynaptic scaffolding proteins and potential transitions between these configurations. A change in one component of the postsynapse, such as a phosphorylation event arising from activity-dependent signalling, may have drastic consequences for the properties of the entire postsynaptic holo-complex as it rearranges itself to accommodate the modification. A striking example of an activity-dependent change that would have drastic consequences in postsynaptic structure is the recruitment of CaMKII (Ca2+/calmodulin-dependent kinase II). CaMKII forms a homo-dodecamer, which, once recruited, becomes the single most abundant component of excitatory postsynapses . The sheer magnitude of this recruitment event has led to the proposal that, apart from its role as a kinase, CaMKII may additionally have a distinct structural role at the excitatory postsynapse .
The intimate interplay between the components of the postsynaptic apparatus requires us to re-examine the properties of known components in the context of a postsynaptic holo-complex. The growing list of postsynaptic proteins arising from systematic proteomic approaches, together with the emerging description of the postsynaptic interactome, allow the study of postsynaptic proteins in systems that reconstitute aspects of the postsynaptic holo-complex. In vitro analyses of protein complexes, reconstitution of postsynaptic elements in heterologous cells, and structural analysis of co-crystallized proteins can uncover potential new states of known proteins and help to decipher the properties of their synaptic configurations. The structure–function relationships uncovered through these analyses will further allow us to genetically address the role of such processes in synaptic transmission and plasticity through the use of knock-in technology to target distinct functionalities of a protein. Though a daunting task, a composite approach to the analysis of postsynaptic proteins may allow us to develop a mechanistic understanding of postsynaptic differentiation, synapse diversity and dynamics.
Synaptopathies: Dysfunction of Synaptic Function: A Biochemical Society Focused Meeting held at The Hotel Victoria, Newquay, U.K., 2–4 September 2009. Organized and Edited by Nils Brose (Max Planck Institute for Experimental Medicine, Göttingen, Germany), Vincent O'Connor (Southampton, U.K.) and Paul Skehel (Centre For Integrative Physiology, Edinburgh, U.K.)
I thank Nils Brose (Max Planck Institute of Experimental Medicine, Göttingen, Germany) for critical comments on this manuscript, and Frédérique Varoqueaux (Max Planck Institute of Experimental Medicine, Göttingen, Germany) for continued support and advice.
My own work in the laboratories of Frédérique Varoqueaux (F.V.) and Nils Brose (N.B.) was supported by the Max Planck Society, the German Research Foundation [grant number GRK 521 (to F.V.)], the Deutsche Forschungsgemeinschaft Research Center for Molecular Physiology of the Brain [grant number FZT-103 (to F.V. and N.B.)], the European Commission (EUSynapse) [grant number LSHM-CT-2005-019055 (to N.B.)] and the Cure Autism Now Foundation (to F.V.).