Correction: Caldendrin–Jacob: A Protein Liaison That Couples NMDA Receptor Signalling to the Nucleus

Correction for: 
 
Dieterich DC, Karpova A, Mikhaylova M, Zdobnova I, Konig I, et al. (2008) Caldendrin–Jacob: A protein liaison that couples NMDA receptor signalling to the nucleus. PLoS Biol 6(2): e34. doi:10.1371/journal.pbio.0060034 
 
The version of Figure 3 originally published in this article was not the one intended for publication, and therefore does not correspond to the accompanying legend. The correct version of Figure 3 is published below. 
 
 
 
Figure 3 
 
Jacob's Association with Subcellular Structures 
 
 
 
Additionally, the authors would like to clarify the legend that accompanies Figure 3. The original final sentence is ambiguous, and should be replaced by “Jacob is only present in fraction P, which contains RNA Polymerase II and represents the euchromatin.”


Introduction
Ca 2þ signals triggered by NMDA-type glutamate receptors can result in long-lasting changes of synaptic input and dendritic cytoarchitecture in phenomena commonly referred to as neuronal plasticity. On the contrary, NMDA receptors are also important players in neurodegenerative processes. Although both aspects require gene expression, our knowledge is still sparse concerning how these fundamental processes are regulated at the molecular level. The Janus face of neuronal NMDA receptor signalling is probably best reflected by the fact that the influx of Ca 2þ ions is thought to act as one of the major mediators of synapto-nuclear signalling [1,2] and of excitotoxic cell death [3]. Within this scheme, a prevailing idea is the existence of Ca 2þ microdomains coupled to the activation of synaptic and extrasynaptic NMDA receptors, and transducing incoming Ca 2þ events to different downstream pathways [1][2][3][4]. In a series of elegant studies, Hardingham and colleagues [5][6][7] provided evidence that Ca 2þ influx through synaptic NMDA receptors trigger nuclear cAMP response element-binding protein (CREB) phosphorylation via an extracellular signal-regulated kinase (ERK)-dependent pathway, whereas Ca 2þ influx through extrasynaptic NMDA receptors leads via an ERK-

Identification and Cloning of Jacob, a Caldendrin Binding Protein in Brain
Utilizing yeast two-hybrid screening to identify binding partners for Caldendrin, we obtained eight independent clones of a hitherto uncharacterized gene product, which we termed Jacob. The longest cDNA clone encompassed an open reading frame of 1,596 bp encoding a 532-amino acid (aa) protein with a calculated molecular weight of 60 kDa ( Figures  1 and S1A). Several clones obtained from rat brain cDNA libraries reveal alternatively spliced Jacob transcripts ( Figure  S1B), giving rise to multiple Jacob isoforms with apparently different molecular weights.
Analysis of the primary structure of Jacob revealed a putative N-terminal myristoylation site and several potential phosphorylation sites for protein kinase C (PKC), cAMP-/ cGMP-dependent protein kinases, and protein tyrosine kinases (Figures 1 and S1A). In addition, Jacob harbours a well-conserved bipartite nuclear localization signal (NLS). Interestingly, this NLS is part of an incomplete IQ motif (Figures 1 and S1A), a protein-protein interaction region characteristic for CaM binding [20][21].

Jacob, a Synaptic and a Nuclear Protein
In situ hybridization experiments revealed a strikingly restricted localization of Jacob transcripts in the limbic brain and cortical areas (unpublished data), showing extensive overlap with Caldendrin mRNA expression (see

Author Summary
Long-lasting changes in communication between nerve cells require the regulation of gene expression. The influx of calcium ions into the cell, particularly through membrane protein called NMDA receptors, plays a crucial role in this process by determining the type of gene expression induced. NMDA receptors can exert multiple and very divergent effects within neuronal cells by impacting opposing phenomena such as synaptic plasticity and neuronal degeneration. We identified a protein termed Jacob that appears to play a pivotal role in such processes by entering the nucleus in response to NMDA receptor activation and controlling gene expression that governs cell survival and the stability of synaptic cell contacts. Removal of Jacob from the nucleus protects neurons from NMDA receptor-induced cell death and increases phosphorylation of the transcription factor CREB, whereas the opposite occurs after targeting Jacob exclusively to the nucleus. The work defines a novel pathway of synapse-tonucleus communication involved in modelling synapto-dendritic input and NMDA receptor-induced cellular degeneration. [19]). In accordance with in situ hybridization data, Jacob immunoreactivity (IR) was found predominantly in cortex and limbic brain structures, including the amygdala, the thalamus, and the hippocampus (Figure 2A). At the cellular level, particularly intense immunostaining was observed in the somatodendritic compartment of pyramidal cells in cortex ( Figure 2B and 2C) and hippocampus, which closely resembles that seen for Caldendrin [16,19]. Moreover, both proteins extensively colocalize in hippocampal primary neurons ( Figure S2).
At the ultrastructural level, Jacob IR was localized to a subset of asymmetric type I synapses on dendrites of cortical neurons ( Figure 2F-2H). Apart from its synaptic localization, intense label was present in patches in dendrites ( Figure 2E). In these patches, Jacob IR was mainly concentrated at the cortical cytoskeleton. In contrast to Caldendrin, intense Jacob immunolabelling was also seen in neuronal nuclei ( Figure  2D). Patchy IR was found both at the nuclear envelope and in the nuclear matrix. Subcellular fractionation experiments confirmed that Jacob is a synaptic and a nuclear protein.
Differential centrifugation of brain protein fractions demonstrated that Jacob IR is associated with particulate Figure 2. Immunolocalization of Jacob in Rat Brain (A) Low-magnification photograph of a rat brain coronal section through the forebrain (corresponding to Figure 31 in the rat stereotaxic atlas by Paxinos and Watson [64]) showing the distribution of Jacob IR. Immunostaining is particularly prominent in the hippocampus (Hip), the cerebral cortex (Cx), amygdala (Amg), and in nuclei of thalamus (Th) and hypothalamus (Hyp).
(B and C) Apart from a fine neuropil staining, soma and apical dendrites (arrows) of pyramidal cells in the layer V of the cerebral cortex are strongly labelled.
(D-H) Electron micrographs of the parietal cortex. (D) At the ultrastructural level, a patch-like distribution of the reaction product (arrowheads) is detectable in nuclei near the nuclear envelope and throughout the karyoplasm of cortical neurons.
(E) Also in dendrites (d) of pyramidal cells, Jacob IR is not evenly distributed. Arrowheads mark patches of the immunoproduct. (G) Two profiles of immunopositive dendrites (d), which receive synaptic inputs (presynapses are marked by asterisks). In the larger dendrite, the reaction product is not concentrated towards the synaptic structures. fractions, including light and heavy membranes, and is prominently present in synaptosomes, synaptic junctional membranes, and the PSD fraction ( Figure 3A). Jacob, like Caldendrin [16], is tightly associated with the PSD, since extensive Triton X-100 extraction resulted in a further relative enrichment of Jacob IR in the detergent-extracted PSD fraction. Interestingly, immunoblots of the crude nuclear fraction demonstrated the presence of prominent Jacob IR bands in the range of 62-70 kDa, whereas the major bands detected in PSD preparations migrated at 72-80 kDa in SDS-PAGE ( Figure 3A and 3B), a difference that most likely reflects posttranslational modification(s). The architecture of the nucleus includes two overlapping nucleic acid-containing structures that are directly associated with the regulation of gene expression: the chromatin and the nuclear matrix. Therefore, we isolated highly pure . Jacob's Association with Subcellular Structures (A) Distribution of Jacob IR in subcellular fractions from rat brain. Panels show two different immunoblots of subcellular fractions (20 lg/lane) obtained by differential centrifugation from brains of adult rats visualized with Jacob (upper panel) and SAP90/PSD-95 (lower panel) antibodies. Hom, homogenate; LM, light membranes; My, myelin fraction; P2, crude membranes; PSD, postsynaptic density fraction; S2, 13.000 3 g supernatant after removal of cell debris and nuclei; SJ, synaptic junctions; Syn, synaptosomes. (B) Western blots showing the extraction of Jacob from a nuclear protein-enriched fraction; P, remaining pellet after centrifugation; S, extractable supernatant. Nuclei were extracted with the agents indicated. Please note the tight association of Jacob isoforms with the nuclear protein fraction. Same amounts of protein were loaded. The size of marker proteins in kDa is indicated at the right margin. (C) Fractionation of chromatin after partial digestion of nuclei by micrococcal nuclease. P: heterogeneously sized DNA bound to the nuclear scaffold; S1: protein fraction containing mononucleosomal-sized DNA fragments; S2: protein fraction containing a nucleosomal ladder of DNA fragments. Jacob is only present in this fraction, which contains RNA Polymerase II and represents the euchromatin. (D) Chromatin immunoprecipitation of Jacob (Jac-IP) (immunoblot in the upper panel) using the P fraction as starting material leads to the coprecipitation of heterogeneously sized DNA (agarose gel in the lower panel) that is not present in IgG control (IgG). (E) N-myristoylation was examined in wild-type Jacob (WT-Jacob-GFP), myristoylation mutant (DMyr-Jacob-GFP), and GFP-expressing HEK-293 (GFP) cells, which were incubated overnight with [ 3 H] myristic acid. Crude detergent extracts were immunoprecipitated with a polyclonal GFP antibody, subjected to SDS-PAGE, transferred to nitrocellulose, and applied to a PhosphoImager system. Left: immunodetection with a monoclonal GFP antibody; right: autoradiograph. Note the incorporation of radiolabelled myristate into immunoprecipitated WT-Jacob-GFP. Sizes of marker proteins in kDa are indicated at the right margin. (F) Subcellular localization of wild-type (Jacob-GFP), myristoylation mutant (DMyr-Jacob-GFP), NLS mutant (DNLS-Jacob-GFP), and myristoylation-NLS double mutant (DNLS/DMyr-Jacob-GFP) Jacob GFP-chimeras expressed in COS-7 cells for 24 h. For counterstaining of nuclei, cells were embedded in propidium iodide-containing mounting media (red). Very little variability was seen in the subcellular localization of these different constructs (DMyr-Jacob-GFP in 98% of all cases nuclear, whereas DNLS-Jacob-GFP and DNLS/DMyr-Jacob-GFP were in 98% of all cases extranuclear). Scale bar indicates 25 mm. doi: 10.1371/journal.pbio.0060034.g003 nuclear matrix, heterochromatin, and euchromatin fractions. Interestingly, after chromatin fractionation, Jacob was found to be exclusively associated with the RNA polymerase IIcontaining euchromatin ( Figure 3C). Moreover, the Jacobcontaining protein complexes immunopurified from euchromatin also contained significant amounts of DNA ( Figure 3D). In addition, the initial purification of the crude nuclear fraction showed an enrichment of the 62-70 kDa Jacob bands in the nuclear matrix (D. C. Dieterich and M. R. Kreutz, unpublished data). This subcellular localization could be confirmed by subsequent extraction of nuclear protein components, including chromatin from COS-7 cells transfected with a green fluorescent (GFP)-Jacob fusion protein (D. C. Dieterich and M. R. Kreutz, unpublished data). These findings suggest that Jacob is highly enriched at active sites of nuclear gene transcription and mRNA processing.

N-myristoylation of Jacob Is a Prerequisite for Its Extranuclear Localization
Jacob harbours an N-terminal myristoylation consensus motif. Transfection of HEK-293 cells cultivated in the presence of 3 H-myristic acid with a wild-type (WT)-Jacob-GFP construct, subsequent immunoprecipitation with a monoclonal anti-GFP antibody and immunoblotting revealed incorporation of radioactivity at a band immunoreactive to both a polyclonal GFP antibody ( Figure 3E) and Jacob antiserum (unpublished data). No incorporation was seen in controls transfected with GFP alone or with a myristoylation mutant, DMyr-Jacob-GFP, in which the crucial glycine at position 2 was mutated to alanine ( Figure 3E). In contrast to the WT-Jacob-GFP construct, transient transfection of COS-7 cells with the DMyr-Jacob-GFP construct led to an exclusive nuclear localization of the mutant protein ( Figure 3F).
Jacob's primary structure exhibits a well-conserved bipartite NLS between aa 250-265. To test for the functionality of this NLS, we generated a deletion mutant (DNLS-Jacob-GFP) lacking the six basic amino acid residues between 247-252. Transfection of this construct in COS-7 cells resulted in an extranuclear localization of the mutant protein ( Figure 3F). Hence, the bipartite NLS seems to be necessary and sufficient for nuclear import of Jacob as the double mutant DNLS/DMyr-Jacob-GFP is extranuclear in transfected COS-7 cells.

Nuclear Jacob Accumulation Results in a Rapid Stripping of Synaptic Contacts and a Simplification of Dendritic Processes
To elucidate functional consequences of nuclear versus extranuclear localization of Jacob in terms of structural plasticity, we transfected hippocampal primary neurons with different mutant (DNLS-Jacob-GFP or DMyr-Jacob-GFP) or WT-Jacob constructs. Transfection of these different mutants had drastic effects on cell morphology. WT-Jacob-GFP-transfected neurons as compared to GFP controls exhibited more, but less-complex, dendritic processes ( Figures 4A-4C and S3A). This effect was astonishingly rapid and was observed already after 6-12 h post-transfection. In sharp contrast to the WT-Jacob-GFP overexpression phenotype, DMyr-Jacob-GFP-transfected cells lost most of their dendritic processes within 12 to 24 h ( Figure 4A-4C). In these cells, the construct exclusively accumulated in the nucleus ( Figure 4A). Interestingly, the density of synaptic puncta was already reduced before the retraction of dendrites became visible, observed as early as 6 h following transfection ( Figure 4D and 4E). No effect on synapse number of WT-Jacob-GFP was seen even 24 h after transfection ( Figure 4E). This strongly suggests that the simplification of the postsynaptic receptive units precedes the retraction of the dendrite. The opposite was found after a lentiviral RNA interference (RNAi)-based knockdown of the nuclear Jacob isoforms harbouring exon 6 with the NLS (RNAi-NLS-GFP). Quantitative immunoblot analysis and immunostainings showed that viral infection of cortical primary cultures led to a specific reduction of these nuclear Jacob isoforms ( Figure 4F, 4G, and 4I). Infected cells showed a slightly increased number of synapses and a more complex dendritic cytoarchitecture ( Figure 4H, 4J, and 4K). Similarly, overexpression of DNLS-Jacob-GFP caused an increase in the number of dendrites, but had no effects on the number of synapses (unpublished data), providing further evidence that the reduction of synaptic contacts directly correlates with the presence of Jacob in the nucleus.
Alternative splicing generates splice isoforms like Dex9-Jacob that contain the NLS but lack large parts of the carboxyterminus ( Figure S1B). We generated a myristoylationdeficient construct of this isoform (DMyr-Dex9-Jacob-GFP), which accumulated in the nucleus after transfection of primary neurons, and its overexpression resulted in a comparable reduction of dendritic complexity and synapse number as those seen with DMyr-Jacob-GFP ( Figure 4A-4E). On the other hand, overexpression of a construct lacking the first 235 amino acids (C-term-Jacob-GFP) had no morphological consequences despite the presence of the NLS and its exclusive nuclear localization ( Figures 4E and S3). Taken together, these data provide strong evidence that the reduction of synaptic contacts directly correlates with the presence of Jacob in the nucleus and that the N-terminal half and the NLS are pivotal for Jacob's morphogenetic impact on dendritic architecture.

Caldendrin Binds the Region of Jacob That Harbours the Bipartite NLS in a Ca 2þ -Dependent Manner and the Binding Cannot Be Competed by CaM
Since the NLS is not only essential, but also sufficient to target Jacob fragments to the nucleus, and the N-terminal half of the protein is crucial and sufficient to elicit the strong pleiomorphic negative effects on neurite and synapse number of nuclear Jacob, we next investigated the Caldendrin binding region in Jacob and vice versa, as well as the functional consequences of Caldendrin binding in more detail. Mapping of binding domains within both proteins was performed using deletion constructs for cotransformation in yeast two-hybrid assays. In Caldendrin, the region containing the first and the second, probably cryptic, EF-hand was found to be essential for Jacob binding ( Figure 5A). Strikingly, in Jacob, we could map the Caldendrin binding region to the central a-helical region that harbours the bipartite NLS. Deletion of the first six basic residues of the NLS led to significantly reduced Caldendrin binding ( Figure 5A).
To substantiate the yeast two-hybrid data, we verified the interaction between Jacob and Caldendrin, employing pulldown assays from brain tissue using a glutathione S-transferase (GST)-Caldendrin fusion protein. In this assay, the interaction of Caldendrin and Jacob was found to be Ca 2þ dependent inasmuch as 1 lM free Ca 2þ in the buffer was required to pull down recombinant Jacob ( Figure 5B). Strikingly, CaM did not bind to Jacob at any Ca 2þ concentration tested ( Figure 5C), and CaM did not compete with GST-Caldendrin for binding to Jacob ( Figure 5D). Further evidence for a bona fide interaction of the two proteins was provided by the binding of Jacob to an anti-Caldendrin antibody column and vice versa ( Figure 5E). These findings are consistent with the colocalization of Jacob and Caldendrin in dendrites and dendritic spines in hippocampal primary neurons observed by confocal laser scans ( Figure  S2B). Importantly, the coimmunoprecipitation of Caldendrin from rat brain required the presence of Ca 2þ and was not visible after addition of EGTA to the precipitation buffer ( Figure 5F). Thus, the interaction of Caldendrin and Jacob in vitro and in vivo is strictly Ca 2þ dependent.
Three-dimensional modelling substantiated the idea that structural differences in Jacob binding surfaces of Caldendrin and CaM [18] can explain the above findings. In crystallographic structures of CaM-peptide complexes, a helical peptide binds to one or both hydrophobic pockets formed by pairs of either EF-hands 1 and 2 or EF-hands 3 and 4 of CaM. The residues forming the two hydrophobic binding pockets are identical in CaM and Caldendrin. However, the binding affinities for the various peptides are given by the size and shape of the pocket, which has been classified as open, semi-open, or closed [22]. This size is dynamically regulated by the Ca 2þ binding states of the EF-hands, which are not rigid structural units, but may function as hinges at low Ca 2þ concentration [23]. Indeed, the second EF-hand of Caldendrin is incapable of binding Ca 2þ , which in turn results in different binding dynamics at the first hydrophobic pocket compared to CaM.
Interestingly, the Jacob sequence does not match any pattern for a typical CaM binding site, but has two incomplete IQ motifs (residues 237-247, aisvfRGyaeR, and residues 260-269, IQrnfRkhlr) within a central region (residues 229-272) containing the bipartite NLS (residues 247-266, PSORTII, [24]). Several CaM binding sequences within a-helices have been defined that contain an IQ motif [20][21]25], which was originally suggested to classify Ca 2þ -independent peptide binding according to the referential CaM/IQ complex. Modelling of the Caldendrin interaction site ( Figure 5G) suggested that the phenylalanine at position 241 is absolutely essential for the protein interaction, and we therefore generated a point mutation at this position (substitution of F to E). This F241E mutant construct indeed showed no interaction with Caldendrin in yeast ( Figure 5A), supporting the assumption that the first part of the central a-helix overlapping with the NLS is the Caldendrin binding region. This a-helical region fits neatly into the hydrophobic pocket generated by EF-hands 1 and 2 and resembles binding to the Ca 2þ sensor protein in the manner of an ''open'' hydrophobic pocket as closely related to the binding of CaMKII to CaM, than the ''semi-open'' binding in MyosinI/CaM. Moreover, this largely excludes the second incomplete IQ motif as an active binding site for Caldendrin. Accordingly, mutations of residues 260 and 261, IQ to GG, yielded only minor differences in the binding properties of Jacob to Caldendrin in yeast ( Figure 5A). In summary, it is predicted that Caldendrin will bind to comparable structures in the presence of a large excess of CaM, suggesting that the Caldendrin-Jacob interaction has evolved independently of CaM-signalling pathways. Even more interesting, binding of Caldendrin to Jacob can be predicted to reduce or inhibit the accessibility of the NLS.

Importin-a Binding to Jacob's NLS Is Competed by Caldendrin
The transport of proteins from the cytosol through the nuclear pore complex into the nucleus depends on the binding of Importins to a specific NLS within the cargo. Within this scheme, Importin-a functions as an adapter molecule by binding both the NLS-bearing protein and Importin-b. Structural modelling suggests that Caldendrin binding will potentially occupy Jacob's NLS, thereby masking this binding site for interaction partners that are likely involved in Jacob's nuclear localization. We tested this hypothesis first by confirming an interaction of Jacob with Importin-a. Coimmunoprecipitation of Importin-a1 from rat cortex indeed suggests a potential in vivo interaction of both proteins ( Figure 5H). In pull-down experiments, we found specific binding of myc-his-tagged WT Jacob, but not of the DNLS-Jacob mutant to GST-Importin-a1 ( Figure 5I). The binding of GST-Importin-a1 was not affected by the presence or absence of Ca 2þ (unpublished data). We next investigated whether the binding of Importin-a1 can be competed by equimolar amounts of recombinant Caldendrin. Indeed, (C) Quantification of MAP2-immunopositive neurites after overexpression of different Jacob-GFP cDNA constructs and GFP-vector as a control (n ¼ 70 cells in each group). An asterisk (*) indicates p , 0.05; triple asterisks (***) indicates, p , 0.001. (D) Micrographs depicting Bassoon-immunoreactive synaptic puncta on dendrites of hippocampal primary neurons after overexpression of different Jacob cDNA constructs. Scale bar indicates 2 lm. (E) Quantification of synaptic puncta on dendrites of hippocampal primary neurons after overexpression of different Jacob-GFP cDNA constructs. Double asterisks (**) indicate p , 0.01. (F) Immunoblots demonstrating the knockdown of nuclear Jacob isoforms after lentiviral infection of cortical primary neurons. Immunoreactivity is clearly reduced in nuclear-enriched fractions at DIV21 after infection of cultures at DIV0 with an RNAi virus targeted to knockdown specifically the NLSbearing nuclear isoforms of Jacob (RNAi-NLS-GFP). A scrambled version of the targeting sequence (scrRNA-GFP), however, has no effect. (G) Immunoblot quantification of the knockdown of Jacob nuclear isoforms with the Quantity One software from BioRad. Depicted are three independent experiments. Quantitative immunoblot analysis was done for the upper (#1) and lower band (#2) (n  these studies revealed a competition between Caldendrin and Importin-a1 for binding to Jacob in the presence of Ca 2þ ( Figure 5J). Interestingly, no competition was seen in the presence of EGTA, suggesting that elevated Ca 2þ levels are needed for Caldendrin to mask the NLS in Jacob.

Nuclear Trafficking of Jacob Is Induced by NMDA Receptor Activation and Requires Importin-a Binding
An elegant recent study established a role of the classical Importin-mediated nuclear import for synapse-to-nucleus communication [26]. In this study, translocation of Importin-a1 and -a2 from distal dendrites to the nucleus was observed requiring NMDA receptor activity. Under resting conditions, however, dendritic Importins are largely immobile. Potential cargos associated with this translocation are at present unknown. Since Jacob is localized both to synapses and the nucleus, and harbours a bipartite NLS, which is bound by Importin-a1 and masked in a Ca 2þ -dependent manner by Caldendrin, we initially tested whether increased NMDA receptor activity will alter the intracellular localization of Jacob. For this purpose, we stimulated hippocampal primary cultures with NMDA for 3 min and quantified for endogenous Jacob the IR fluorescence signal intensity of propidium iodide-counterstained neuronal nuclei. Jacob IR increased significantly in neuronal nuclei within 30 min after NMDA receptor activation, with highest levels after 2 h ( Figure S4A). Nuclear Jacob IR returned to control levels within 4 h ( Figure  S4A). As previously reported [26], Importin-a1 accumulates in the nucleus in a similar time frame. Interestingly, no recruitment of Caldendrin to the nucleus was observed (unpublished data). Bath application of glutamate led to a significantly increased nuclear accumulation of Jacob IR within a comparable time frame to NMDA receptor activation ( Figure S4B). This accumulation could be completely blocked by coincubation of the competitive NMDA receptor antagonist DL-APV (DL-2-amino-5-phosphonopen-tanoic acid), indicating that activation of NMDA receptors is crucial for glutamate-induced recruitment of Jacob to neuronal nuclei ( Figure S4B). To exclude the possibility that stimulation of primary neurons alters the accessibility of the nuclear Jacob antigen to the antibody, we performed quantitative western blot analysis on neuronal nuclei from organotypic hippocampal slice cultures stimulated with the same protocol. These experiments showed a significant increase in intensity of the two major Jacob nuclear isoforms (62 kDa/70 kDa) 2 h after stimulation ( Figure S4C and S4D). Moreover, application of anisomycin after NMDA receptor activation did not affect the increase of nuclear Jacob IR, indicating that a recruitment of already existing extranuclear protein underlies the increased Jacob levels in hippocampal nuclei, but not de novo protein synthesis ( Figure S4C and S4D).
Using quantitative fluorescence time-lapse microscopy of hippocampal primary neurons transfected with WT-Jacob-GFP or the DNLS mutant, we found that the presence of the NLS is essential for the nuclear translocation of Jacob. Glutamate stimulation of WT-Jacob-GFP-transfected cultures kept in the presence of anisomycin resulted in an increase of somatic and nuclear GFP fluorescence with a time course comparable to that of the endogenous protein ( Figure 6A-6D). The nuclear accumulation of Jacob-GFP, however, was not seen in neurons transfected with the DNLS mutant Jacob-GFP construct ( Figure 6E and 6F), suggesting that the presence of the binding site for Importin-a1 is a prerequisite for Jacob's nuclear accumulation. Importantly, concomitant to the nuclear accumulation of WT Jacob, the GFP fluorescence decreased in proximal and distal dendrites ( Figure 6C), an effect that was absent in DNLS-Jacob-GFP-transfected neurons. This indicates that the presence of the NLS and the interaction with Importin-a1 are not only important for the nuclear import, but are also crucial for Jacob's transport from dendrites to the nucleus.
(B) Interaction of Jacob with GST-tagged Caldendrin (Cald-GST) in a pull-down assay at different Ca 2þ concentrations indicated. A Triton X-100extracted P2 fraction was incubated with glutathione sepharose loaded with a GST fusion protein containing the C-terminal half (aa 137-298) of Caldendrin or with GST alone. pull-down, pellet fraction; unbound, supernatant. (C) Interaction of Jacob with GST-tagged Caldendrin in a competitive pull-down assay with recombinant CaM at 100 lM Ca 2þ . A Triton X-100-extracted P2 fraction was incubated with glutathione sepharose loaded with a GST-fusion protein containing the C-terminal half (aa 137-298) of Caldendrin or with GST alone in the presence or absence of 40-lg recombinant CaM. PD, pull-down fraction; UB, unbound material. After washing, bound proteins were detected by western blotting using the antibody anti-JB150. (D) WT-Jacob-GFP does not bind to CaM in a pull-down assay. Jacob binding to CaM or Caldendrin-GST sepharose was tested in the presence of 100 lM Ca 2þ or 2 mM EGTA. GST sepharose was used as the control. WT-Jacob-GFP was detected by western blotting with a GFP antibody. (E) Co-immunoaffinity purification of Jacob and Caldendrin. Left panel: 15 mg of protein from the 100,000 3 g supernatant of a rat brain membrane extract were loaded either to an anti-Jacob affinity column (anti-Jac) or to an anti-Caldendrin affinity column (anti-Cald). Jacob is co-eluted with Caldendrin and vice versa from the corresponding affinity columns. A number sign (#) indicates Jacob-immunoreactive bands and an asterisk (*) Caldendrin-immunoreactive bands. Right panel: no specific binding was seen to normal serum control beads. 1: Deoxycholic acid (DOC) extract from rat brain cortex; 2: nonbinding proteins (flow-through); 3: specifically bound proteins (eluate). (F) Caldendrin coimmunoprecipitates with Jacob from a rat brain extract in a Ca 2þ -dependent manner. Rabbit anti-Jacob bound to protein A sepharose specifically precipitates Jacob independent from the conditions used (5 mM of EGTA or 100 lM Ca 2þ in the precipitation buffer). Rabbit IgG is used as a negative control (upper panel). Caldendrin was only detected in the precipitate under Ca 2þ conditions (lower panel). (G) Molecular modelling of the Caldendrin-Jacob interaction. Caldendrin binds Ca 2þ -dependently to the first incomplete IQ motif, and blocks the NLS. As indicated, Jacob contains two incomplete IQ motifs. The matching residues are marked in bold. The helical structure has been modelled using the coordinates of Myosin I (pdb: 1wdc), showing that Jacob has a similar CaM site distribution as the template structure. In contrast to myosin I/CaM, Caldendrin binds to site 1 in a complex more closely related to the compact CaMKII/CaM complex. This is in agreement with Caldendrin's binding at high Ca 2þ concentration, whereas IQ motif proteins bind independently of Ca 2þ concentrations. Crucial residues for the interaction are indicated.
(H) Immunoprecipitation of Importin-a1 from a soluble rat brain protein fraction with a Protein-A sepharose-coupled Jacob antibody (JB-150). Importin-a1 was only found in the immunoprecipitate (IP) of the Protein-A sepharose coupled Jacob-antibody, whereas it remained in the prewashing supernatant (S1) in the IgG and Protein-A sepharose control. (I) GST-Importin-a1 pull-down of myc/his-tagged WT and DNLS-Jacob extracted from transfected HEK-293 cells. Only WT Jacob, but not DNLS-Jacob, is found in the pull-down, indicating that the presence of the NLS is essential for the Importin-a1/Jacob interaction. PD, pull-down fraction; UB, unbound material. (J) GST-Importin-a1 pull-down of myc/his-tagged WT Jacob in the presence of equimolar amounts of Caldendrin. Pull-down of Jacob is attenuated in the presence of 2 mM Ca 2þ , but not in the presence of 2 mM EGTA. PD, pull-down fraction; UB, unbound material. To learn more about the role of Caldendrin for the extranuclear retention of Jacob, and to understand the apparently contradictory findings (i.e., NMDA receptor activation with subsequent Ca 2þ influx leading to Jacob's nuclear import and concomitantly Caldendrin binding preventing this process at high synapto-dendritic Ca 2þ levels), we analyzed the transport process of Jacob in more detail using confocal laser scan microscopy. A brief depolarization of hippocampal neurons with KCl also induced a translocation of Jacob and Importin-a1 to the nucleus in the presence of anisomycin ( Figure 7A and 7B). However, the nuclear accumulation of both proteins was less pronounced than after glutamate receptor stimulation. Importantly, this effect was completely abolished in the presence of the NMDA receptor antagonist DL-APV ( Figure 7A and 7B), indicating that raising intracellular Ca 2þ levels by other means than Figure 7. Nuclear Immunoreactivity of Jacob and Importin-a1 upon Different Stimulation Protocols Depicted are confocal images obtained from a single nuclear focal plane. Scale bar indicates 10 lm. All experiments were done in the presence of 7.5 lM anisomycin. Nuclear Jacob and Importin IR were quantified using the Image J software. The region of interest was outlined from DAPI stainings, and nuclear Jacob and Importin-a1 IR were measured as mean grey value (in arbitrary units of pixel intensity) from Z-stacks of two to three nuclear planes, and differences between groups were calculated as relative deviations from control. The nuclear membrane was excluded from the analysis. NMDA receptor activation is not sufficient to drive Jacob and Importin-a1 into the nucleus. NMDA receptors are situated both at synaptic and extrasynaptic sites [27,28]. Bath application of NMDA is considered to affect preferentially, but not exclusively, extrasynaptic NMDA receptors [6,7]. To differentiate between these two populations, we indirectly stimulated hippocampal cultures by incubation with the GABA A receptor antagonist bicuculline. The blockade of inhibitory synapses leads to an increased release of glutamate at synaptic sites, and resulted as expected in an increased accumulation of Jacob and Importin-a1 in the nucleus ( Figure 7C and 7D). This effect, however, was much less distinct as compared to the bath application of NMDA. A coincubation with the noncompetitive NMDA receptor antagonist MK-801 attenuated the nuclear accumulation of Jacob and Importin-a1 to levels indistinguishable from control conditions ( Figure 7C and 7D). Since MK-801 is an irreversible open channel blocker, we took advantage of this fact to differentiate between synaptic and extrasynaptic NMDA receptors. After washout of the drug following stimulation of synaptic glutamate receptors, we applied NMDA to the bath solution to exclusively activate extrasynaptic NMDA receptors. Interestingly, this regime induced a marked nuclear translocation of Jacob and Importin-a1 that was more prominent than the accumulation after stimulation of synaptic NMDA receptors ( Figure 7E and 7F). Synaptic NMDA receptors contain predominantly the NR2A subunit, whereas their extrasynaptic counterparts contain mainly the NR2B subunit [29]. To test the hypothesis that the nuclear translocation of Jacob and Importin-a1 requires activation of NR2B-containing NMDA receptors, we repeated the experiments outlined above in the presence of the NR2B-specific antagonist ifenprodil. Intriguingly, in the presence of ifenprodil, the nuclear import of Jacob and Importin-a1 could be completely blocked after bath application of NMDA ( Figure 7G and 7H). These results show that the nuclear import of these two proteins requires signalling via the largely extrasynaptically localized NR2B-containing NMDA receptors.

Caldendrin Binding Targets Jacob Outside the Nucleus Only after Synaptic NMDA Receptor Stimulation
To follow up this hypothesis in more detail, we transfected a GFP-Caldendrin construct into hippocampal primary neurons. Expectedly, overexpression of Caldendrin blocked the increase of endogenous nuclear Jacob IR after synaptic stimulation at day in vitro (DIV) 16, indicating that the interaction with Caldendrin masks the bipartite NLS of Jacob ( Figure 8A and 8B). However, after stimulation of extrasynaptic NMDA receptors, overexpression of Caldendrin attenuated Jacob's nuclear import much less efficiently ( Figure 8A and 8B). We therefore checked whether RNAi knockdown of Caldendrin ( Figure S5) affects the nuclear trafficking of Jacob differentially after synaptic and extrasynaptic NMDA receptor stimulation. We found that the nuclear immunofluorescence for Jacob was significantly increased in cells with reduced Caldendrin levels after enhancing synaptic activity with bicuculline at DIV 16 ( Figure  8C-8E), whereas the Caldendrin knockdown had no effect on Jacob's nuclear import after activation of extrasynaptic NMDA receptors. This points to a regulatory function of this protein-protein interaction in nuclear trafficking of Jacob after enhanced synaptic activation that is related to the competitive accessibility of Jacob's NLS for either Caldendrin or Importin-a binding.

Jacob Is Part of the CREB Shut-Off Pathway
The predominant Ca 2þ -and NMDA receptor-activated signalling pathways to the nucleus in neurons funnel through the activation of the transcription factor CREB [1][2]. Previous work has shown that extrasynaptic NMDA receptor activation results in a dephosphorylation of CREB at Ser133 (pCREB) that renders it transcriptionally inactive and, therefore, constitutes a CREB shut-off signal [7,30]. Because Jacob was most efficiently targeted to neuronal nuclei after extrasynaptic NMDA receptor activation, we next addressed the question of whether the presence or absence of Jacob in the nucleus affects the phosphorylation of CREB at this crucial serine residue. As a first proof of principle, we explored whether nuclear overexpression of the DMyr-Jacob-GFP construct significantly reduced the levels of pCREB in hippocampal primary neurons as compared to untransfected or GFP-transfected controls under resting conditions ( Figure  9A and 9B). Indeed, infection of cortical primary cultures with a Semliki Forest virus-expressing DMyr-Jacob-GFP led to drastically reduced pCREB levels as evidenced by quantitative immunoblotting, whereas total CREB levels were not affected ( Figure 9C and 9D). To more rigorously test the hypothesis that Jacob is part of the CREB shut-off signalling pathway, we induced a knockdown of nuclear Jacob using plasmid-based RNAi constructs targeting exon 6-containing isoforms of the protein and subsequently stimulated extrasynaptic NMDA receptors with the protocol outlined above. We found that nuclear knockdown of Jacob completely abolished the reduction of pCREB observed after stimulation of extrasynaptic NMDA receptors ( Figure 9E and 9F). These data point to a critical role of Jacob for survival of hippocampal primary neurons after triggering the CREB shut-off pathway. We therefore decided to assess next whether the absence of Jacob in the nucleus enhances neuronal survival after triggering CREB shut-off with the stimulation of extrasynaptic NMDA receptors. To this end, we chose in situ TdT-39end labelling to visualize DNA fragmentation in hippocampal primary neurons as a measure of apoptotic cell death. Using this assay, we found that the number of neurons showing fragmented DNA after sustained extrasynaptic NMDA receptor activation was clearly reduced under conditions of nuclear knockdown of Jacob as compared to untransfected cells from the same plate or independent GFP-transfected controls from other plates ( Figure 10A and 10B). Accordingly, the number of condensed propidium iodide-positive nuclei after nuclear knockdown of Jacob was reduced in the same manner as compared to controls ( Figure 10C and 10D).

Knockdown of Nuclear Jacob Prevents Stripping of Synaptic Contacts Induced by Sustained NMDA Receptor Activation
A prominent consequence of bath application of NMDA to primary cultures is the reduction of synaptic contacts within a few hours [34]. Importantly, we found that this reduction requires gene transcription. Coincubation of NMDA with actinomycin-D, an inhibitor of RNA Polymerase II, com-  pletely blocked the loss of synaptic contact sites in treated cultures 4 h after stimulation ( Figure 11A and 11C). Thus, in line with previous work, loss of synapses appears to be an early event of structural breakdown in cultures treated with bath application of NMDA and requires gene transcription. To further strengthen the point that Jacob is upstream of a transcription-dependent cell death pathway following excessive extrasynaptic NMDA receptor activation, we performed a plasmid-based RNAi knockdown of nuclear Jacob isoforms. Using this approach, we found that the knockdown of Jacob in the nucleus not only prevented CREB shut-off, but also preserved the structural integrity of transfected neurons. As evidenced by immunostainings with the presynaptic marker bassoon quantified 4 h after stimulation, the number of synapses in cultures transfected with RNAi targeting of nuclear Jacob isoforms was essentially the same compared to cultures transfected with a control RNAi vector (scrRNA-GFP). Most importantly, however, after bath application of NMDA, the number of synapses dropped only in cultures transfected with the control vector but not in Jacob RNAitransfected cultures ( Figure 11B and 11D). Thus, the absence of nuclear Jacob prevents not only CREB shut-off and subsequent neuronal degeneration, but also early events of structural disintegration related to loss of synaptic input.

Discussion
The Caldendrin-Jacob Interaction Constitutes a New

Mechanism of NMDA Receptor-To-Nucleus Communication
In this study, we identified a novel neuronal protein pathway that is well suited to couple NMDA receptor signalling to the cell nucleus and to trigger long-lasting changes in the cytoarchitecture of dendrites and the number of spine synapses. This novel pathway particularly couples activation of NR2B-containing NMDA receptors to morphogenetic signalling via the nuclear trafficking of Jacob. At resting conditions, Jacob is attached to extranuclear compartments in an either Importin-a bound or unbound state (see also Figure 12). Ca 2þ influx through synaptic and extrasynaptic NMDA receptors is followed by a translocation of Importin-a from synapses and dendrites to the nucleus, and we propose that Importin-a-bound Jacob will be concomitantly recruited to the nucleus. Moreover, the presence of the NLS is essential for Jacob's translocation, indicating that trafficking from dendrites to the nucleus and not only nuclear import already requires the classical Importin pathway. This is reminiscent of previous data showing NMDA receptor-dependent Importin trafficking from dendrites to the nucleus [26], and establishes Jacob as the first identified cargo of this trafficking event. Accordingly, we always found a tight correlation between Jacob's and Importin-a1 nuclear translocation. Caldendrin binding can mask the bipartite NLS of Jacob in competition with Importin-a and thereby prevent its nuclear trafficking ( Figure 12). However, in contrast to Importin-a binding, this requires high Ca 2þ levels and not only NMDA receptor activation ( Figure 12). We propose that Caldendrin will target Jacob to spine synapses after enhanced synaptic activation ( Figure 12). In support of this hypothesis, we could provide evidence that activation of NR2B-containing NMDA receptors, which are mainly located at extrasynaptic sites, is crucial for the nuclear import of Jacob and Importin-a1. Interestingly, we found that blocking this receptor subtype did attenuate the nuclear accumulation of both proteins after stimulation of synaptic NMDA receptors that contain predominantly, but not exclusively, the NR2A subunit [29][30][31][35][36][37][38]. This suggests the intriguing possibility that the nuclear Jacob-Importin pathway is physically coupled to NR2B-containing NMDA receptors and that the presence or absence of Ca 2þ -bound Caldendrin in the respective synapto-dendritic compartment will decide whether local Jacob shuttles to the nucleus or not.

Caldendrin and Jacob: A Protein Liaison in Control of the CREB Shut-Off Pathway
On the basis of the characteristics and consequences of its nuclear import, we found conclusive evidence that Jacob is part of the CREB shut-off pathway. The most prominent nuclear target of neuronal NMDA receptor signalling is the transcription factor CREB [1][2][8][9]. Subsequent to its phosphorylation at serine 133, pCREB triggers gene expression crucially involved in processes of synaptic plasticity and neuronal survival [8][9]. Analysis of this pathway has demonstrated that synaptic NMDA receptors strongly activate CREB-dependent gene expression, whereas extrasynaptic NMDA receptors trigger a CREB shut-off [7]. A most intriguing finding in recent years has been that the antagonistic signalling of extrasynaptic versus synaptic NMDA receptors resembles their opposing actions on the activation of ERK kinase [6,[39][40][41]. Activation of synaptic NMDA receptors is coupled to the Ras-ERK pathway and subsequent CREB phosphorylation, whereas extrasynaptic NR2B-containing receptors promote dephosphorylation and inactivation of the Ras-ERK-pathway [6,[39][40][41]. One caveat of this scenario, however, is that shutting down Ras-ERK alone cannot explain the shut-off of CREB since other mechanisms, and here prominently nuclear CaMK-IV, should be in principal sufficient to phosphorylate CREB in the absence of ERK activity [8][9]. Thus, the opposing influence of both types of NMDA receptors after bath application of NMDA requires another mechanism that will actively trigger CREB shut-off. Our data suggest that the same conditions that trigger shut-off of CREB and the Ras-ERK pathway drive Jacob into the nucleus. Overexpression of Jacob in the nucleus-without activating these pathways-is sufficient to attenuate CREB phosphorylation, and a nuclear knockdown of Jacob prevents CREB shut-off as well as neuronal cell death after triggering the pathway. Finally, the rapid loss of synaptic contacts, one of the hallmarks of bath application of NMDA in hippocampal primary cultures, was prevented by reducing the amount of nuclear Jacob. Noteworthy in this regard is the observation that CREB shut-off cannot be induced in young cultures (,DIV 7) [30,42], a developmental stage at which Jacob protein levels are very low (unpublished data). We therefore propose that nuclear Jacob is an essential component of CREB shut-off that might be actively involved in rendering CREB in a dephosphorylated state.
What is Jacob's physiological role in the nucleus? In initial experiments, we could not establish a direct binding of Jacob to CREB although both proteins are found in the overlapping fractions after gel filtration of nuclear protein complexes (unpublished data). Therefore, it is conceivable that Jacob is indirectly coupled via CREB-binding proteins to the CREB signalosome. To further support a role in gene expression, we provided substantial evidence that Jacob is highly enriched in two nuclear compartments associated with gene transcription and pre-mRNA processing. Jacob is abundant in euchromatin fractions and therefore present at active sites of gene transcription. The protein harbours long stretches of basic amino acid residues, which are well suited for DNA binding, although no known DNA binding motif was identified in its primary structure. Particularly with regard to the phenotype of its nuclear overexpression that involves a rapid destabilization of synaptic contacts and a retraction of dendrites, and which cannot be explained entirely by CREB shut-off, it is reasonable to assume that Jacob will be part of additional nuclear signalling events.

Nuclear Jacob Induces Pleiomorphic Negative Effects on Synapto-Dendritic Cytoarchitecture
The nature of such signalling events will be obviously related to the circumstances of Jacob's nuclear trafficking. CREB shut-off has been largely assigned so far to pathophysiological insults, including spill-over of glutamate after excessive stimulation or reversal of glutamate transporters in the context of epileptic seizures or brain ischemia [3]. This view, however, probably has to be extended because in recent years, a number of observations raise the possibility that the activation of extrasynaptic NR2B-containing NMDA receptors can occur in a physiological context. It was shown that in several brain regions, sustained synaptic activation causes spillover of synaptically released glutamate to nonsynaptic sites [43][44][45][46][47][48][49]. In addition, sustained synaptic activation favours nonsynaptic release of glutamate from astrocytes [50][51][52], and it has been suggested that this glia-neuron transmission via extrasynaptic NMDA receptors has profound effects on non-Hebbian types of neuronal plasticity [53]. Moreover it was also claimed that activation of extrasynaptic NMDA receptors might directly induce heterosynaptic long-term depression at certain synapses in close proximity [54]. The evolving concept behind these studies is the idea of homeostatic scaling of synaptic input. Homeostatic plasticity refers to a process by which principal neurons in particular constantly adjust the integration of synaptic input to optimize the contribution of a single synapse with reference to its location in the dendrite and the synchronized activity in a given neuronal network [55,56]. A major aspect of homeostatic plasticity is the fact that uncontrolled potentiation of synapses will induce a ceiling effect characterized by epileptic activity and a decoupling of a given neuron from the dynamics of presynaptic input. Homeostatic plasticity reflects the necessity to either remove certain synapses that contribute less efficiently to the optimal activity within a neuronal network or to reduce the level of potentiation of synapses in this network. Jacob's nuclear accumulation and its rapid morphogenetic effects are in favour for a role in the regulation of plasticity-related gene expression related to homeostatic synaptic plasticity. Interestingly, this role includes a stripping of synaptic contacts that precedes the simplification and regression of dendritic processes. It is therefore conceivable that the loss of synapses is the initial trigger for the retraction of dendritic arbors. Moreover, this process is surprisingly rapid, indicating that synapses are actively destabilized. This in turn suggests that Jacob either blocks an essential nuclear signalling event required to prevent the removal of synaptic input or regulates the expression of genes that will actively destabilize synapses. It is likely that the CREB shut-off pathway will be part of this mechanism, but it is unclear whether it is sufficient to trigger solely the course of events following Jacob's nuclear import.
Caldendrin and Jacob Provide a Novel Mechanism of Neuronal Ca 2þ Signalling A further intriguing aspect of this study is that it provides the first demonstration that an EF-hand CaM-like Ca 2þ sensor protein regulates the nuclear localization of a protein by competitive binding to its NLS in a Ca 2þ -dependent manner. The significance of this novel mechanism of neuronal Ca 2þ signalling is further underscored by the fact that binding of Caldendrin is specific in that its ancestor and closest relative in brain, CaM, did not bind to Jacob at any Ca 2þ concentration tested. This is of importance since CaM levels are probably more than a magnitude higher in neurons than those of Caldendrin [18], and Ca 2þ binding affinities are comparable between both proteins [57]. Computer modelling based on templates from crystallized structures shows that the outer surface of solvent-exposed amino acids, particularly EF-hand 2, which seems to be crucial for binding to Jacob, and another recently identified binding partner light chain 3 (LC3) [58] are covered by residues that clearly differ between CaM and Caldendrin [18]. Accordingly, LC3, a component of the microtubular cytoskeleton, apparently does not bind to CaM [58]. The principal specificity of Caldendrin protein interactions is further supported by the observation that very few mutations occurred in this region during vertebrate development and that none of these mutations affected the solventexposed amino acids of EF-hand 2 [18]. Thus, the singularity of the Caldendrin surface is intrinsic and independent from insertions or deletions, and we therefore suggest that this is probably due to adaptations of its surface to a specific localization and function in neurons of higher vertebrates.
How could this singularity with respect to other Ca 2þ binding proteins relate to Caldendrin's neuronal function? In contrast to the interaction with Jacob, Caldendrin binding to most of its interaction partners is Ca 2þ independent, as already described above for the LC3 interaction [57]. For instance, Ca 2þ -, CaM-, and ATP-independent interaction of the C-terminal half of Caldendrin/CaBP1 was demonstrated for the inositol trisphosphate receptor ((InsP 3 R) [59][60]. The functional consequence of Caldendrin binding is a reduction of InsP 3 -triggered intracellular Ca 2þ release [59,60]. At the synapse, a Ca 2þ -independent binding was reported for L-type voltage-dependent Ca V 1.2 Ca 2þ channels [61,62]. This interaction will probably lead to increased Ca 2þ currents following synaptic activation and thereby indirectly via increased synaptic activity could promote Caldendrin's and possibly Jacob's synaptic localization. Low synaptic activity and, hence, low synapto-dendritic Ca 2þ levels will instead favour Caldendrin's binding to the InsP 3 R. It is therefore conceivable that Caldendrin can thereby directly lower Ca 2þ levels in dendritic microdomains, and in consequence, negatively regulate its own association with Jacob. Therefore, a switch of binding partners could directly relate to Caldendrin's role in regulating Jacob's nuclear transition. Along these lines, it can be predicted that keeping the delicate balance between Jacob's nuclear and extranuclear localization via Caldendrin binding will provide a powerful regulatory mechanism in the transformation of dendritic Ca 2þ signals into morphogenetic signals for the dendritic cytoarchitecture of principal neurons under pathophysiological and probably also under physiological conditions. cortical neurons at DIV0. After 3 wk, cells were either fixed with PFA for immunostaining or harvested and prepared for SDS-PAGE.
For the preparation of Semliki Forest particles and infection of primary cortical neurons, pSFV-Helper2, pSFV-DMyr-Jacob-EGFP, or pSFV-EGFP after in vitro transcription were cotransfected into CHO-K1 cells with Lipofectamine2000 (Invitrogen) according to the supplier's manual. After 24 and 48 h, the culture medium containing the budded particles was harvested. Viral particles were concentrated by ultracentrifugation through 10% sucrose, the pellet was resolved in Tris-buffered solution overnight at 4 8C. Aliquots of the particles were stored at À80 8C after shock freezing. For infection of primary cortical neurons, the particles were activated by chymotrypsin and further diluted with OptiMEM. High-density cortical cultures were infected at DIV16 and harvested 24 h later. Neurons were homogenized in 20 mM Tris buffer containing protease and phosphatase inhibitors, and solubilised in SDS buffer. The protein concentration was determined by amido black test, and equal amounts were loaded for SDS-PAGE.
Statistical analysis. Statistical analysis was performed with ANOVA and subsequent Bonferroni's Multiple Comparison test. Data are presented as mean 6 standard error of the mean (SEM). A level of p , 0.05 was considered statistically significant.
Computer modelling. Details about the computer models can be found under Supplementary Materials and Methods in Text S1.     Text S1. Supplementary Results and Materials and Methods Found at doi:10.1371/journal.pbio.0060034.sd001 (73 KB DOC).

Accession Number
The Protein Data Bank (http://www.pdb.org/pdb/home/home.do) accession number for the structural model discussed in this paper is 1wdc.