Formation of a high order complex between Homer and Shank
Overexpression of Shank in neurons causes enlargement of dendritic spines, and coexpression of a long Homer, Homer1b, with Shank synergistically enhances the effect (Sala et al., 2001
). We were intrigued by how Shank causes the enlargement of dendritic spines with a synergistic effect of Homer. Shank is known to homo-multimerize through interactions between its domains (Baron et al., 2006
; Im et al., 2003
; Romorini et al., 2004
), while Homer1b forms a tetramer (Hayashi et al., 2006
). We hypothesized that the tetrameric Homer crosslinks the multimeric Shank, resulting in the formation of a polymerized matrix structure. Such a complex does not have any theoretical limit in size, and reasonably explains the synergistic effect of Homer and Shank on the size of dendritic spines.
To test this hypothesis, we attempted to reconstitute the high-order complex using purified Shank and Homer1b. For stable expression and purification of Shank, we deleted the ankyrin repeats, the SH3 domain, and the proline-rich region predicted to be PEST sequences, while leaving the Homer binding site intact. This deletion construct retains the PDZ domain, the Homer binding site, and the SAM domain, and was named Shank1CΔPEST based on a naturally occurring alternatively spliced form of Shank1 lacking the amino-terminal domains (Lim et al., 1999
The high-order complex formation was assessed by high-speed centrifugation. When tested individually, neither the purified Shank1CΔPEST nor Homer1b formed precipitate upon ultracentrifugation (). In contrast, when they were mixed at a 1:1 molar ratio, approximately 50% of Shank1CΔPEST and Homer1b were precipitated, suggesting the formation of a high molecular weight complex. On dynamic light scattering, a sensitive method for detecting large complex, Homer1b or Shank1CΔPEST alone had a single scattering peak corresponding to the hydrodynamic radii of 10.3 nm and 4.3 nm, respectively (). In contrast, the mixture of Homer1b and Shank1CΔPEST showed two peaks, one at 12.1 nm and the other at 3.6 μm, the upper limit of our instrument (). The latter peak corresponds to the polymerized complex between Homer1b and Shank1CΔPEST. The first peak, slightly larger than Homer1b alone, shows the interaction between Homer1b and Shank1CΔPEST.
Formation of the high-order complex required the specific interaction between Homer1b and Shank. Neither a Homer1b W24A mutant without the EVH1 domain ligand binding activity (Beneken et al., 2000
), nor Shank1CΔPEST P1497L mutant with a mutation at the Homer binding site (Lim et al., 1999
), formed the precipitate (). Shank1CΔPEST without the SAM domain (Shank1DΔPEST) did not form precipitates with Homer1b either, indicating that multimerization of Shank is required for the high-order complex formation. When varying amounts of Homer1b were incubated with a fixed amount of Shank1CΔPEST, the amount of Shank1CΔPEST in the precipitate was maximum when Homer and Shank were at a 1:1 ratio ().
The formation of the high-order matrix structure was further confirmed by negative stain electron microscopy. When observed individually, Homer1b appeared as highly flexible fibers, which likely represent the coiled-coil domain (). Shank1CΔPEST had a globular structure, which likely represents the oligomerized SAM domains and the PDZ domains (). In contrast, the Homer and Shank mixture showed a mesh-like structure in electron microscopic images (). The mesh appears to contain hubs, possibly representing Shank oligomerized via its SAM domain and linked by the filamentous Homer.
Interaction between Homer-Shank high-order complex and other synaptic proteins
The mesh-like structure observed under electron microscopy is consistent with its ability to incorporate other PSD proteins. We therefore wanted to know if the Homer-Shank complex can recruit other PSD proteins. GKAP, another abundant PSD protein was chosen for its binding ability to Shank and PSD-95 (Naisbitt et al., 1999
; Tu et al., 1999
). When purified carboxy-terminal fragment of GKAP (451–666) was incubated with Homer and Shank, the GKAP fragment co-precipitated with the Homer-Shank complex without affecting the amount of Homer or Shank in the precipitates (). The amount of GKAP in the precipitates reached the plateau at the molar ratio of ~1.4 versus Homer/Shank.
Interaction between the Homer-Shank complex and other PSD proteins
We also examined whether the Homer1b and Shank complex formation can be regulated by mechanisms related to synaptic plasticity. First, we observed the effect of Homer1a, an activity induced form of Homer (). We added increasing amounts of Homer1a to a fixed amount of Homer1b-Shank1CΔPEST mixture. Precipitation was decreased to half when Homer1a was added at a ratio of 1:1, and fully prevented at a 5-fold excess of Homer1a to Homer1b (). This indicates that a competitive interaction between Homer1a and Homer1b on Shank prevents the complex formation.
/calmodulin-dependent protein kinase II (CaMKII) is a key regulator of synaptic plasticity. Three CaMKII phosphorylation sites have been identified in Homer3a (Mizutani et al., 2008
). We examined the effect of CaMKII-mediated phosphorylation on the formation of a Homer-Shank high-order complex. By adding the purified CaMKIIαactivated with Ca2+
and calmodulin to the mixture of Homer and Shank (), the complex formation between Homer3a and Shank1CΔPEST was inhibited, but not the complex formation between Homer1b and Shank1CΔPEST.
The coiled-coil region of Homer forms a hybrid of dimer and tetramer
The filamentous appearance of Homer, which connects hubs of Shank, suggests the importance of the coiled-coil region of Homer for the Homer-Shank network structure formation. To understand the structural basis of the functional importance of the coiled-coil region, we crystallized the carboxy-terminal 65 residues of Homer1b, which corresponds to one-third of the entire coiled-coil region (). The 1.75 Å crystal structure was solved by the multiple-wavelength anomalous dispersion method, using a L308M mutant designed for selenomethionine labeling, and refined to an R/Rfree of 0.216/0.289 (see Table S1 for statistics). Most of the residues of the crystallized fragments were well ordered, and we could assign most of them to the electron density, with the exception of one or two residues from the both termini. The carboxy-terminal 75-residue fragment of Homer3a was also crystallized, and the 2.9 Å structure was solved by the molecular replacement method using the Homer1b structure as a search model, and refined to an R/Rfree of 0.252/0.287. The root mean square deviations of Cαpositions between each strand of the two structures were 0.72–1.09 Å. Due to the high structural similarity between Homer3a and Homer1b, we use Homer1b for analyses and discussions in the remainder of this study.
Crystal structure of the Homer coiled-coil region
The structure showed an elongated rod-like structure 140 Å in length (). It consists of two pairs of parallel left-handed dimeric coiled-coils intercalating with each other at the very carboxy-termini, where it forms an anti-parallel tetrameric coiled-coil. The diameter of the coiled-coil at the dimeric region of each end is 15 Å, and the diameter at the tetrameric region at the center is 25 Å.
The overall length of the coiled-coil region is ~180 residues which can be translated to ~45 nm. Upstream of the coiled-coil region, there is a hinge region of ~60 residues, and the amino-terminal globular EVH1 domain. As a whole, Homer has a dumbbell-like tetrameric structure with a pair of EVH1 domains located at each end of the tetramer, separated by a dimer-tetramer hybrid coiled-coil ().
The primary sequence of the amino-terminal portion of the dimeric region (290–312) shows typical heptad repeats with the a and d positions occupied by aliphatic or small polar amino acids (). Most of the residues at the a and d positions form canonical knobs-into-holes interactions. The e and g positions are occupied by acidic or basic residues and are involved in the formation of inter-molecular salt bridges (). The space between the two helices starts to widen around residue 312 towards the carboxy side, thus allowing accommodation of larger amino acids at a and d positions such as Q319 (a position) and F322 (d position) (). Knobs-into-holes packing is rarely observed within this region. Eventually, the distance becomes sufficiently wide for the intercalation of another dimer to form a tail-to-tail tetramer in a left-handed antiparallel configuration via the carboxy-terminal 30 residues (). In the tetrameric region (326–354), the e positions, in addition to the a and d positions, are occupied by hydrophobic residues, typically leucine and isoleucine (). Unlike the dimeric region, the knobs-into-holes interactions are observed between the residues at the d (A–D and B–C) and e (A–C and B–D) positions (). The residues at the a positions, all occupied by leucines, do not form knobs-into-holes interactions but fill the cavity at the center of the four α-helices, thereby forming a hydrophobic core (). No fixed water molecules were observed in this cavity.
A dimeric mutant of Homer1b does not form a high-order complex with Shank
To explore the functional significance of the unusual antiparallel tetrameric structure, we attempted to disrupt this structure and dimerize by introducing point mutations based on the crystal structure. In the dimeric coiled-coil region, positively or negatively charged residues at the e and the g positions are common and they form either inter- or intra-chain salt bridges to stabilize the dimer, while these positions are occupied by aliphatic residues in the tetrameric region (). I337 (e position) of Homer1 forms a knobs-into-holes interaction with the adjacent chain (). We expected that changing I337 and the corresponding residue at the g position, I332, to a pair of positively and negatively charged residues would destabilize the hydrophobic interaction at the core of the tetramer and result in the formation of a stable dimer through electrostatic interactions between the parallel chains. As expected, the I332R/I337E double mutant had a Stokes radius of 6.8 nm, a significant reduction compared to wild type Homer1b and comparable to that of a deletion mutant of the tetrameric region, Homer1bΔ329 (6.5 nm, ). To determine the oligomerization status unambiguously, we measured the molecular weight by sedimentation equilibrium experiments (). While the wild type had a molecular weight of 179 kDa, corresponding to 4.2mer, the I332R/I337E mutant was 85 kDa, corresponding to 2.0mer, and the Δ329 deletion mutant was 84 kDa, corresponding to 2.1mer.
We tested if wild type Homer1b and Homer1b I332R/I337E interact with each other. We tagged them with HA and myc epitopes, respectively, and coexpressed in HEK293T cells. The proteins were separated on an analytical gel filtration column, and their elution profiles were monitored by Western blotting. Individually expressed HA-Homer1b and myc-Homer1b I332R/I337E were eluted as a single peak at the expected position of tetramer and dimer, respectively. When these two proteins are coexpressed, their elution profiles showed two peaks, both at the tetramer and the dimer molecular weights (). This shows that the dimeric mutant interacts with wild type Homer to make heteromeric dimers, although some of the heteromers still form tetramers.
Next, we studied if the tetramerization of Homer is important for Shank cross-linking activity. In both a high-speed centrifuge assay and a dynamic light scattering assay, the dimeric mutant Homer1b I332R/I337E did not show the formation of the high-order complex (). A small shift in the hydrodynamic radius by the addition of Shank1CΔPEST (from 6.8 nm to 7.5 nm) shows the interaction between the two proteins. This indicates that the specific spatial arrangement of the four EVH1 domains conferred by the coiled-coil domain is important for the Homer-Shank network formation. In contrast, this mutant does not change its interactions with syntaxin 13, which is known to interact with the coiled-coil region of Homer () (Minakami et al., 2000
Tetramerization of Homer1b is required for spine localization of Homer, Shank and PSD-95
Using the Homer1b I332R/I337E mutant, we studied the role of tetramerization of Homer in neurons. We first compared the synaptic localization between the wild type Homer1b and the dimeric mutant. We coexpressed the mGFP-tagged Homer or its mutant with cytosolic red fluorescent protein (DsRed2) to normalize volume, in CA1 pyramidal neurons of organotypically cultured hippocampal slices. Homer1b accumulated in spines with an average spine/dendrite ratio of 3.09 ± 0.04 (mean ± SEM) (). In marked contrast, the localization of the dimeric mutant Homer1b I332R/I332E was significantly reduced (1.49 ± 0.02), comparable to monomeric Homer1a (1.50 ± 0.02). These results show that the tetramer formation is critical for synaptic localization of Homer1b.
Effect of the dimeric mutant form of Homer on the localization of synaptic proteins
Given the importance of tetramerization of Homer1b for the formation of the high-order complex with Shank in vitro, we studied if Homer tetramerization is required for spine localization of Shank. We transfected dissociated cultures of hippocampal neurons with myc-Homer1b or its dimeric mutant. Transfected neurons were identified by anti-myc antibody staining, and the distribution of endogenous Shank was detected by an anti-Shank antibody. The expression of Homer1b I332R/I337E clearly reduced the number and the intensity of Shank clusters ().
To study if the expression of the dimeric mutant also affected the spine localization of other PSD proteins, we immunostained PSD-95, another major PSD protein which binds to receptors. The expression of Homer1b I332R/I337E significantly reduced the number of PSD-95 clusters (). These results suggest that the expression of the dimeric mutant of Homer interferes with normal synaptic localization of Shank by preventing the formation of a high-order complex, and also affects the localization of other synaptic proteins, such as PSD-95.
Tetramerization of Homer1b is necessary for maintenance of dendritic spine structure and synaptic function
Next, we investigated whether the tetramerization of Homer1b is required for the integrity of dendritic spine structure. We introduced myc-Homer1b or myc-Homer1b I332R/I337E, along with GFP as a volume-filler, into neurons in hippocampal dissociated culture and measured the morphology and density of spines. Compared with neurons expressing Homer1b, those expressing the dimeric Homer1b I332R/I337E had significantly reduced dendritic spine density (). The remaining spines were significantly longer and slightly thinner, typical of immature spines, though the difference in spine width did not reach statistical significance (). The same experiments using hippocampal organotypic slice culture showed a similar effect of myc-Homer1b I332R/I337E (data not shown). We also tested the effect of siRNA against Homer1 on the dendritic spine structure (). The siRNA reduced the number of mature spines, which is consistent with the phenotype of Homer1b I332R/I337E overexpression. There were phenotypic differences of the immature spines, that neurons expressing Homer1b I332R/I337E had fewer, longer spines, while those with Homer1 siRNA had increased number of shorter spines. This could be due to the difference in the mechanisms of suppression between these two methods. Importantly, the effect of siRNA was reversed by a rescue construct of Homer1bR
with silent mutations at the siRNA target sequence, but not with a similar rescue construct with dimeric mutation (Homer1bR
I332R/I337E). In addition, while the coexpression of Homer1b with Shank significantly increased the size of Shank clusters compared with Shank alone as previously reported (Sala et al., 2001
), the effect was not observed in the Homer1b I332R/I337E mutant ().
Effect of the dimeric mutant form of Homer on dendritic spine structure and synaptic transmission
Finally, to assess the functional significance of Homer tetramerization on synaptic transmission, we analyzed the AMPA-R and NMDA-R mediated excitatory postsynaptic current (EPSC) of neurons transfected with Homer1b or Homer1b I332R/I337E (). The expression of Homer1b caused small and insignificant increases of both AMPA-R and NMDA-R-EPSC compared with untransfected neurons (). On the other hand, expression of Homer1b I332R/I337E caused significant decreases in the both currents. The effect is likely due to the combined effect of reduced size and number of dendritic spines and reduced amount of synaptic protein at the synapse. The AMPA/NMDA ratio did not change with the expression of the wild type or the mutant Homers, suggesting that the effect of the dimerization equally affected the two receptor populations (). These results indicate that tetramerization of Homer1b is important for structure and function of dendritic spines and synapse, through its ability to form a high-order complex with Shank.