Scaffold Protein X11α Interacts with Kalirin-7 in Dendrites and Recruits It to Golgi Outposts*

Background: Kalirin-7 is a neuronal Rac1-GEF that regulates spine morphology. X11α is a PDZ scaffolding protein. Results: X11α binds kalirin-7 and modulates its localization to Golgi outposts, GEF activity, and spine morphology. Conclusion: X11α is a novel regulator of small GTPase signaling by recruitment to Golgi outposts. Significance: We identify a novel mechanism for regulation of neuronal GEF localization and function. Pyramidal neurons in the mammalian forebrain receive their synaptic inputs through their dendritic trees, and dendritic spines are the sites of most excitatory synapses. Dendritic spine structure is important for brain development and plasticity. Kalirin-7 is a guanine nucleotide-exchange factor for the small GTPase Rac1 and is a critical regulator of dendritic spine remodeling. The subcellular localization of kalirin-7 is thought to be important for regulating its function in neurons. A yeast two-hybrid screen has identified the adaptor protein X11α as an interacting partner of kalirin-7. Here, we show that kalirin-7 and X11α form a complex in the brain, and this interaction is mediated by the C terminus of kalirin-7. Kalirin-7 and X11α co-localize at excitatory synapses in cultured cortical neurons. Using time-lapse imaging of fluorescence recovery after photobleaching, we show that X11α is present in a mobile fraction of the postsynaptic density. X11α also localizes to Golgi outposts in dendrites, and its overexpression induces the removal of kalirin-7 from spines and accumulation of kalirin-7 in Golgi outposts. In addition, neurons overexpressing X11α displayed thinner spines. These data support a novel mechanism of regulation of kalirin-7 localization and function in dendrites, providing insight into signaling pathways underlying neuronal plasticity. Dissecting the molecular mechanisms of synaptic structural plasticity will improve our understanding of neuropsychiatric and neurodegenerative disorders, as kalirin-7 has been associated with schizophrenia and Alzheimer disease.

The architecture of the dendritic arbor and the distribution, morphology, and strength of synapses on dendrites are crucial determinants of neuronal connectivity within brain circuits. Dendritic spines are small mushroom-shaped protrusions that decorate the dendrite and serve as the sites of most excitatory synapses onto neurons in the mammalian forebrain. Dendritic spine remodeling is an important process in activity-independent and -dependent circuit formation and plasticity during development and adulthood (1). Their dysfunction in disease states may have profound effects on connectivity, as altered dendritic and synaptic morphology has been documented in several neurodevelopmental disorders (2).
Dendritic spine morphogenesis relies on the precise regulation of local actin dynamics (3). Although multiple signaling cascades converge onto the actin cytoskeleton, the Rho-like small GTPase, Rac, has been shown to have profound effects on dendritic spine morphology (4). In cortical neurons, the function of Rac is tightly controlled by the brain-specific guanine nucleotide-exchange factor (GEF) 4 kalirin-7 (5-7). Kalirin-7 has been shown to be a critical coordinator of activity-independent and -dependent changes in dendritic spine morphology in cortical neurons (5,7,8). Functionally, the interaction of kalirin-7 with PDZ domain-containing proteins such as PSD-95 and afadin (also known as AF-6) mediate its localization to the postsynaptic density (9). This localization is thought to be essential for its function, and PDZ proteins may play additional roles in its subcellular localization. However, interactions of kalirin-7 with other PDZ proteins have not been further characterized.
Previously, a yeast two-hybrid screen with the C terminus of kalirin-7 identified a number of interacting PDZ domain-containing proteins, including the X11␣ protein (10). The X11 family of adaptor/scaffold proteins are mammalian homologs of the Caenorhabditis elegans gene LIN-10 (also referred to collectively as mLIN-10) and consist of three members, X11␣, -␤, and -␥, encoded by different genes (ABP1, ABP2, and ABP3) (11,12). X11␣ (also known as Mint1) contains a Munc-18-interacting region, a CASK-interacting region, a phosphotyrosinebinding domain, and two PDZ domains (Fig. 1A). X11␤ (also known as Mint2) has a similar domain structure but lacks the CASK-interacting domain, whereas X11␥ lacks the Munc18and CASK-interacting domains (12). X11␣ and X11␤ have been shown to be expressed in neurons (12,13), exhibiting both presynaptic (14) and postsynaptic localization (13,15,16). Functionally, LIN-10 and its mammalian homologs act as adaptor proteins involved in the trafficking of ion channels and glutamate receptors. In invertebrates, the Lin-10 protein interacts with the motor protein KIF17; this complex mediates the trafficking of NMDA receptors along microtubules (16) and is required for the delivery of AMPA-like receptors to synapses in C. elegans (17). In vertebrates, X11 proteins interact with the small GTPase Rab6 to control the trafficking of amyloid precursor protein (18 -20). In addition, X11 proteins have been shown to directly interact with amyloid precursor protein via their phosphotyrosine-binding domain and are involved in amyloid precursor protein processing and secretion (11,12). Examination of specific X11 family members have revealed that X11␣ promotes the presynaptic localization of N-type calcium channels (21). Recently, X11␤ has been shown to localize to the Golgi apparatus, dendrites, and synapses of hippocampal neurons, where it binds to AMPA receptors in a PDZ-dependent manner and modulates their trafficking (15). However, whether X11␣ also shows a similar localization and can modulate the trafficking of synaptic proteins is unknown.
Here, we show that X11␣ and kalirin-7 interact via a PDZmediated interaction in cortical neurons. We find that X11␣ localizes to dendrites and spines of cortical neurons, as well as to Golgi outposts, distal specializations of the Golgi complex in dendrites. At synapses, X11␣ co-localizes with several synaptic proteins; furthermore, FRAP and biochemical analysis reveal that X11␣ is weakly associated with the postsynaptic density (PSD) and is a mobile protein in spines. Furthermore, overexpression of X11␣ drives kalirin-7 accumulation in Golgi outposts, attenuates kalirin-7 GEF activity, and reduces dendritic spine size. Taken together, these data provide a novel mechanism of regulation and localization of small GTPase GEFs in dendrites by recruitment to Golgi outposts.
Cell Culture-Dissociated cultures of primary cortical neurons were prepared from E18 Sprague-Dawley rat embryos and cultured as described (23). Neurons were plated onto coverslips or 60-mm dishes, previously coated with poly-D-lysine (0.2 mg/ml, Sigma), in feeding media (Neurobasal ϩ B27 supplement (Invitrogen) ϩ penicillin/streptomycin ϩ 0.5 mM L-glutamine). Neuron cultures were maintained in the presence of 200 M DL-amino-phosphonovalerate acid (Ascent Scientific) beginning on DIV 4 to maintain neuron health for long term culturing and to reduce cell death due to excessive Ca 2ϩ cytotoxicity via overactive NMDA receptors (23). Cortical neurons were transfected at DIV 24 -28 using Lipofectamine 2000 (Invitrogen) following the manufacturer's recommendations. Transfected constructs were allowed to express for 2 days before being fixed. Human embryonic kidney 293 (HEK293) cells were cultured in DMEM with 10% FBS and penicillin/ streptomycin. Cells were plated onto 6-well plates and grown until 40% confluent, when they were transfected using Lipofectamine 2000. Between 2 and 5 g of DNA was used with 6 l of Lipofectamine 2000 per well. Transfections proceeded for 48 h before cell lysates were harvested.
Co-immunoprecipitations-Cells (HEK293) were harvested in RIPA buffer (in mM: 150 NaCl, 10 Tris-HCl, pH 7.2, 5 EDTA, 0.1% SDS, 1% Triton X-100, 1% deoxycholate, plus inhibitors). Precleared lysates were incubated with 3-5 l of antibody for 2-4 h; 60 l of protein-A-Sepharose was added for 1 h at 4°C, after which samples were washed with 1 ml of RIPA buffer. Co-immunoprecipitation from cortical neurons or synaptosomal preparations of rat cortical tissue homogenates (6) was carried out essentially as described above. Following treatments, cells were lysed in RIPA buffer and sonicated, and cell debris was removed by centrifugation. Samples were then incubated with antibodies for 3 h before the addition of 60 l of protein-Sepharose-A, followed by 3-4 washes with 0.5 ml of RIPA buffer. All experiments were performed at least three times from independent cultures.
Rac1 Activation Assays-To examine activation of endogenous Rac1 in neurons, we used the Rac/cdc42 assay kit from Millipore. Neurons were harvested in 0.5 ml of lysis buffer on ice and sonicated. Lysates were cleared by centrifugation at 14,000 ϫ g for 10 min, and supernatants were incubated with 10 l of PAK-1 p21-binding domain resin for 2 h at 4°C; positive and negative controls were incubated with 10 mM EDTA and 0.1 mM GTP␥S and 0.1 mM GDP, respectively, for 15 min at 30°C (controls were incubated with the resin for 45 min). The resin pellet was washed three times in 0.5 ml of lysis buffer, loaded on SDS-PAGE, and analyzed by Western blotting with the anti-Rac1 monoclonal antibody. Quantification of bands was performed by measuring the integrated intensity of each band and normalizing it for protein loading using ImageJ.
Immunocytochemistry-Cells were fixed in either 4% formaldehyde, 4% sucrose in PBS or at 4°C in methanol pre-chilled to Ϫ20°C for 20 min. Fixed neurons were permeabilized and blocked simultaneously in PBS containing 2% normal goat serum and 0.2% Triton X-100 for 1 h at room temperature. Primary antibodies were added in PBS containing 2% normal goat serum overnight at 4°C, followed by three 10-min washes in PBS. Secondary antibodies were incubated for 1 h at room temperature, also in 2% normal goat serum in PBS. Three further washes (15 min each) were performed before coverslips were mounted using ProLong antifade reagent (Invitrogen).
Quantitative Immunofluorescence-Changes in protein clustering and synaptic localization were quantified using immunofluorescence on fixed neurons and visualized with the indicated antibodies, as described previously (23). Images were taken using a Zeiss LSM5 Pascal confocal microscope and a ϫ63 objective (N.A. 1.4) or using a Leica SP5 confocal microscope and a ϫ63 objective (N.A. 1.4). Images of GFP-expressing neurons or endogenous proteins labeled with Alexa 488 were acquired with excitation/detection at 488/505-530 nm, YFP-GalT images with 496/510 -555 excitation/detection, Alexa 568 images with 568/570 -630 excitation/detection, and Alexa 633 with 647/655-750 excitation/detection. Cultures that were directly compared were stained simultaneously and imaged with the same acquisition parameters. For each condition, 9 -15 neurons from three to five separate experiments and 100 m apical dendrite from each neuron were analyzed. Experiments were performed on sister cultures, and analysis was carried out under blind conditions. Maximum projection images were reconstructed using Metamorph. For fluorescence intensity measurements, the background corresponding to areas without cells were subtracted to generate a "backgroundsubtracted" image. Images were then thresholded equally to include clusters with intensity at least 2-fold above the adjacent dendrite. Regions along dendrites were outlined using the "Perimeters" utility, and the linear density (number/100 m dendrite length), area, and total gray value (total immunofluorescence intensity) of each AMPA receptor cluster were measured automatically. To determine the degree of co-localization between two channels, each channel was thresholded to select distinct puncta as described above. A 100-m dendritic region was selected, and puncta counts were made; puncta smaller than 0.08 m 2 were excluded from analysis. Regions representing the measured puncta in one channel were generated using Metamorph and overlaid on the other channel. Puncta were counted as co-localized if the average intensity within the overlaid region exceeded the threshold. Thresholds were set individually for each antibody and held constant across treatment conditions. In the green/purple color scheme, co-localization is indicated in white. Images were taken in the linear range, to allow an accurate representation. Each channel is shown individually in grayscale.
Time-lapse Imaging and FRAP-Neurons on coverslips were transfected with GFP-tagged X11␣ and 2 days later were preincubated in artificial cerebrospinal fluid (in mM: 125 NaCl, 2.5 KCl, 26.2 NaHCO 3 , 1 NaH 2 PO 4 , 11 glucose, 5 HEPES, 2.5 CaCl 2 , and 1.25 MgCl 2 with 200 M DL-amino-phosphonovalerate acid), after which they were transferred to a stage chamber and imaged at 37°C in artificial cerebrospinal fluid. Square regions encompassing dendritic shaft or spines were photobleached using 10 passes (12 s total) of 100% laser power, which was optimized to quench fluorescence in fixed cells. Recovery fluorescence was acquired using 1% laser power, with images taken every 6 s for 144 s. Dendrites were captured through a ϫ63 objective with 2ϫ averaging. Intensity of recovered fluorescence was measured in the spine region or dendritic shaft region. Focal drift was adjusted for by measuring the intensity of nonbleached area on a different dendrite of the same cell. Healthy neurons with overall pyramidal morpholo-gies expressing GFP-tagged protein were identified and imaged. Kymographs were created using the "polyline kymograph" plugin in ImageJ. Curved processes were straightened using the plugin; resultant kymographs show the process along the x axis and time across the y axis. Following imaging sessions, ϫ20 images of the neurons were taken to ascertain lack of photodamage. Any neurons exhibiting signs of distress were omitted from quantification. Between 7 and 9 cells from three independent cultures were analyzed.
Tissue Preparation and Subcellular Fractionation-Rat cortex was dissected from adult female Sprague-Dawley rats (Harlan) and homogenized in 10 volumes of 20 mM Tris-HCl, pH 7.5, buffer containing protease inhibitors, as described previously (6). Subcellular fractionations and synaptosomal preparations of rat cortical tissue homogenates were performed as described previously (6).
Statistics-Bars represent means, and error bars are standard errors. To identify differences among conditional means, statistical analyses (Student's unpaired t test, analysis of variance) were performed in GraphPad Prism 5. Tukey-b post hoc analysis was used for multiple comparisons.

X11␣ and Kalirin-7 Interact and Co-localize in Neurons-
In a previous study, the C terminus of kalirin-7 was shown to interact with a fragment containing the two PDZ domains of X11␣ in a yeast two-hybrid screen ( Fig. 1A) (10); however, this interaction has not been shown in neurons. To validate this interaction, we performed co-immunoprecipitation experiments from rat brain and transfected heterologous cells. We first used an antibody recognizing the spectrin region of kalirin (kal-spect) to immunoprecipitate kalirin proteins from rat cerebral cortex synaptosomal preparations (Fig. 1B), and we probed them using an antibody that specifically detected X11␣ (Fig.  1C). In these experiments, X11␣ co-immunoprecipitated with kalirin-7, indicating that these two proteins form a protein complex. Notably, X11␣ and kalirin-7 did not co-precipitate in the presence of kalirin-7 preimmune serum or when the kalspec antibody was preadsorbed to its antigen (GST-spectrin domains) (Fig. 1B). To determine whether this interaction was dependent on the C terminus of kalirin-7, we performed coimmunoprecipitation experiments from HEK293 cells transfected with Myc-tagged kalirin-7 or its C-terminally truncated mutant that lacks the C-terminal PDZ-binding domain (kal7-⌬CT). Consistent with our data from rat cerebral cortex, X11␣ co-immunoprecipitated with kalirin-7 (Fig. 1D). However, X11␣ did not co-immunoprecipitate with myc-kalirin-7-⌬CT, indicating that the C-terminal portion of kalirin-7 was critical for its interaction with X11␣ (Fig. 1D). These data show that kalirin-7 interacts with X11␣ in the cerebral cortex, and this is likely mediated by the interaction of the kalirin-7 C-terminal tail with the PDZ domains of X11␣.
As kalirin-7 and X11␣ co-immunoprecipitated together, we sought to determine whether these proteins co-localized along dendrites and dendritic spines of cortical neurons. Endogenous kalirin-7 and X11␣ were detected by immunofluorescence using specific antibodies. As described previously, kalirin-7 was present along dendrites and at synapses at 24 days in vitro (DIV) ( Fig. 1E) (7). Consistent with previous suggestions (11,12), X11␣ was also present along dendrites and at a subset of spinelike structures (Fig. 1E). Moreover, X11␣ and kalirin-7 were found to co-localize along the dendrite and in a subset of spinelike structures along dendrites of cortical pyramidal neurons (Fig. 1E). Quantification of co-localization revealed that ϳ70% of X11␣ was positive for kalirin-7, whereas only 50% of kalirin-7 puncta was positive for X11␣ (Fig. 1E). Taken together, these data indicate that kalirin-7 interacts with X11␣ in the cerebral cortex, via its C-terminal tail. Furthermore, X11␣ and kalirin-7 co-localize in a subset of punctate structures along the dendrites of cultured cortical pyramidal neurons.
X11␣ Is Present in Excitatory Synapses and Spines-Although X11␣ has been suggested to be present along dendrites and at the PSD (11,12,14), a detailed analysis of X11␣'s subcellular distribution has not been performed. Interestingly, a previous study using a pan-X11 antibody, which detects both ␣ and ␤ isoforms, observed X11/mLIN-10 within synaptosomes prepared from cultured hippocampal neurons (15). This suggests that X11␣ is enriched in synapses, consistent with our observations that X11␣ co-localizes with kalirin-7. Thus, we investigated the localization of X11␣ in dendrites of mature cultured cortical pyramidal neurons (DIV 24) using immunofluorescence microscopy and an X11␣-specific antibody (Fig. 1C). X11␣ localized to punctate structures along dendrites and in dendritic shafts where it partially co-localized with the postsynaptic protein PSD-95, a marker for excitatory synapses ( Fig.  2A). Moreover, X11␣ also partially co-localized with the AMPA receptor subunit GluA1 (Fig. 2B). Quantification of co-localization revealed that approximately 55% of X11␣ puncta was positive for PSD and 50% of PSD-95 puncta was positive for X11␣. Conversely, ϳ60% of X11␣ puncta was positive for GluA1, and ϳ65% of GluA1 puncta was positive for X11␣ (Fig. 2C). These data indicate that a subpopulation of X11␣ is enriched at the postsynaptic density of excitatory synapses.
To further characterize the localization of X11␣ at synapses, we immunostained GFP-expressing neurons for the endogenous protein (Fig. 2D). Consistent with its co-localization with PSD-95 and GluA1, X11␣ was observed in a subset of dendritic spines (Fig. 2, D and E, yellow arrows). In addition, X11␣ could also be seen in large clusters along the dendrites (Fig. 2D, open arrowheads); X11␣ clusters were especially prominent at the dendrite branch points and were observed at the base of spines (Fig. 2E). To analyze the distribution of X11␣ further, we exam-

X11-␣ Modulates Kalirin-7 Activity
ined X11␣ clusters within spines or along the dendritic shaft. X11␣ clusters along the dendritic shaft exhibited increased fluorescence intensity compared with clusters within dendritic spines (Fig. 2F). Collectively, these data demonstrate that X11␣ is present at a subset of excitatory synapses, within dendritic spines and along the dendritic shaft of cortical neurons. Yellow arrows indicate sites of co-localization, and yellow arrowheads indicate X11␣ puncta that do not localize with PSD-95 (overlay of green and magenta ϭ white). B, X11␣ partially co-localizes with AMPA receptor subunit GluA1 in cultured rat cortical pyramidal neurons. Yellow arrows indicate sites of co-localization, and yellow arrowheads indicate X11␣ puncta that do not localize with GluA1. C, bar graphs are quantitative measures of respective co-localization of immunofluorescent puncta (PSD-95 and X11␣; GluA1 and X11␣). D and E, representative confocal images of cortical neurons (DIV 24) expressing GFP and pRK5-X11␣. Overexpressed pRK5-X11␣ localizes to dendritic spine heads (yellow arrows) and dendritic shaft (yellow arrowheads). Red arrows indicate dendritic spines containing little or no X11␣ signal (D). X11␣ accumulated at the base of spines and at dendritic branching points (yellow arrowheads) (E). F, mean fluorescence intensity of X11␣ signal in dendritic spines is increased compared with X11␣ signal in dendritic shaft (*, p Ͻ 0.05, Student's unpaired t test). Scale bars, 5 m (A, B, and D) and 1 m (E).

X11-␣ Modulates Kalirin-7 Activity
X11␣ Is a Mobile Protein in Dendrites and at Synapses-Based on our localization data, X11␣ localizes to spines and dendritic shaft. To examine the possibility that X11␣ is a mobile protein, we examined the mobility and persistence of X11␣ at synapses and within the dendritic shaft in live cells. We monitored GFP-tagged X11␣ in live mature cortical neurons (DIV 24) also expressing mCherry, allowing visualization of neuronal morphology. Similar to the endogenous protein, GFP-X11␣ localized to both dendrites and dendritic spines (Fig. 3A). To monitor the behavior of GFP-X11␣ puncta, we used kymograph analysis to plot to the fluorescence intensity of X11␣ puncta in the dendrite. Within this region, motile and stationary X11␣ puncta could be visualized (Fig. 3, B and C); stationary puncta were seen as vertical bands, and mobile puncta are indicated by diagonal bands within the kymograph. The percentage of X11␣ puncta that were moving, assessed over a period of 3 min, was ϳ45% (data not shown), indicating that a subpopulation of X11␣ is a mobile protein within the shaft (Fig. 3C). Interestingly, kymograph analysis of X11␣ puncta within dendritic spines also demonstrated that GFP-X11␣ was being trafficked in spines (Fig. 3D). Occasionally, GFP-X11␣ puncta could be observed moving to or from the dendrite into a spine (Fig. 3D), strongly suggesting that X11␣ can traffic to and from synapses.
To quantitatively examine the mobility of X11␣ puncta in the dendritic shaft and within spines, we performed FRAP experiments. A lack of fluorescence recovery in FRAP experiments indicated an immobile protein, although a recovery of fluores-cence indicated the exchange of the bleached molecule in the spine or dendrite with unbleached molecules from other pools of protein. Cultured cortical neurons (DIV 24) were transfected with GFP-tagged X11␣ and imaged in artificial cerebrospinal fluid, and regions encompassing individual spines or dendritic shaft were photobleached (Fig. 4, A and B). Intensity of recovered fluorescence was measured in the spine region or dendritic shaft region. We found that GFP-X11␣ fluorescence signal was effectively depleted by bleaching and partially recovered after photobleaching in both dendritic shaft and spine, as would be expected for a mobile protein (Fig. 4, A-C). Indeed, over 60% of GFP-X11␣ recovered within 150 s, indicating that a large portion of X11␣ is mobile. Interestingly, the absolute mobile fraction was similar in the spine and dendritic shaft (% mobile fraction: spine, 62.7 Ϯ 8.2; shaft, 60.4 Ϯ 7.0; p ϭ 0.84; Fig. 4D). However, the fluorescence recovery pattern was not the same; the recovery of GFP-X11␣ fluorescence within the dendritic shaft occurred more rapidly within the first 30 s as compared with the recovery within spines (Fig. 4C). Consistent with this, the time at which 50% of equilibrium fluorescence level is recovered (1 ⁄ 2 ) from GFP-X11␣ in the dendritic shaft was significantly lower than in the dendritic spines (1 ⁄ 2 (seconds): spine, 67.3 Ϯ 8.2; shaft, 29.4 Ϯ 5.2; p Ͻ 0.001; Fig. 4E). This difference in mobility dynamics for X11␣ in each compartment suggests that X11␣ turnover in spines is slower than in the dendritic shaft, potentially due to increased stability of X11␣ within spines. Overall, these data demonstrate that X11␣ is a mobile protein, with a high degree of mobility within both spines and shaft, but it is partly stabilized in spines as compared with the shaft.
As our FRAP data demonstrate that X11␣ is a mobile protein at synapses, we examined the strength of association of X11␣ with the PSD using subcellular fractionation and detergent solubilization experiments. We first prepared synaptosomes using a HEPES/sucrose gradient and PSD fractions by solubilization of synaptosomes using Triton X-100, SDS, or deoxycholate from rat cortical tissue. X11␣ was present in P1, S2, and P2 homogenates indicating that this protein is abundant in synaptosomes (Fig. 4F), similar to other proteins, including kalirin-7, localized to spines (6,24). Interestingly, X11␣ could be extracted by all detergents tested, even weaker detergents like Triton X-100 or deoxycholate (Fig. 4F). In comparison, spi-nophilin, another PDZ domain-containing protein that is tightly associated with the PSD, was not readily extracted by Triton X-100 (Fig. 4F), consistent with previous reports (25). Thus, based on our biochemical data, X11␣ appears to associate weakly with the PSD, which would be consistent with a mobile protein capable of trafficking to and from synapses.
X11␣ Localizes to Golgi Outposts in Dendrites-It is now generally accepted that many neurons contain both somatic and discrete discontinuous Golgi complexes (also known as Golgi outposts) located in dendrites. These satellite Golgi complexes are thought to play an important role in controlling the synthesis, sorting, processing, and trafficking of synaptic proteins (26,27). In neuronal cells, both X11␣ and -␤ have been suggested to associate with the Golgi complex (12, 15), although whether Square regions encompassing spines or dendritic shaft were photobleached (white dotted square) using 10 passes (12 s total) of 100% laser power, which was optimized to quench fluorescence in fixed cells. Recovery fluorescence was acquired using 1% laser power, with images taken every 6 s for 144 s. Intensity of recovered fluorescence was measured in spine region or dendritic shaft region (yellow dashed ovals). Focal drift was adjusted for by measuring intensity of nonbleached area on a different dendrite of the same cell. C, normalized intensity of GFP-X11␣ signal during fluorescence recovery after photobleaching. D, percentage of GFP-X11␣ that is mobile in spines and dendritic shaft, as measured by percentage of fluorescence that recovers after photobleaching. E, 1 ⁄2 (time at which 50% of equilibrium fluorescence level is recovered) for regions of interest in spines and dendritic shafts. 1 ⁄2 of GFP-X11␣ in dendritic shaft is significantly shorter than dendritic spines. F, relative abundance of X11␣ and spinophilin in rat cortex PSD fractions. Synaptosomes were prepared from rat cortical tissue using a HEPES/sucrose gradient, separated by SDS-PAGE, and analyzed by Western blotting using X11␣ (upper) and spinophilin (lower) antibodies. Fractions indicated are as follows: P1, nuclear pellet; S2, supernatant from crude synaptosomal fraction; P2, crude synaptosomal pellet. The P2 fraction was further solubilized in buffers with detergents of different strengths, and soluble (S) or insoluble and particulate (P) fractions were separated by centrifugation. Detergents used were as follows: TX, Triton X-100; SDS, sodium dodecyl sulfate; Doc, deoxycholate. X11␣ was present in P1, S2, and P2 fractions and was extracted from synaptosome fractions by all detergents tested, suggesting a weaker association with the PSD. **, p Ͻ 0.001. Scale bar, 1 m. X11␣ is present in Golgi outposts along dendrites is not known. We thus examined in detail the localization of X11␣ to Golgi complexes and related structures. In agreement with previous studies (15), X11␣ localized in the vicinity of the Golgi complex (Fig. 5A), as indicated by proximity to TGN38, a marker of trans-Golgi complexes. To further investigate the relationship between X11␣ and the Golgi network throughout the neuron, we examined the co-localization of X11␣ with giantin, a marker for cis-Golgi and medial Golgi complexes located in the soma and in dendrites (28). Consistent with co-localization with TGN38, X11␣ immunoreactive puncta partially overlapped with giantin in the soma (Fig. 5, B and C, panel i). In the soma and proximal regions of the primary dendrite, giantin was enriched in tubulovesicular structures in the soma and proximal regions of the primary dendrite (Fig. 5C, arrows), and X11␣ was more abundant in vesicular structures (Fig. 5C, panel i,  open arrowheads). Consistent with previous studies (28), giantin was present in vesicular structures in distal portions of the dendrite, consistent with the presence of Golgi outposts (Fig. 5, B and C, panel ii) (26,29). Remarkably, a subset of X11␣ puncta co-localized with giantin in these structures along dendrites (Fig. 5, B and C, panel ii, yellow arrows). Recently, it has been suggested that the medial, cis-, and trans-compartments of Golgi complexes are disconnected the dendrites of Drosophila neurons (30). Therefore, to assess whether X11␣ preferentially localized to different Golgi compartments, we next examined the localization of X11␣ with the trans-Golgi compartment. Thus, we labeled Golgi outposts with fusion proteins consisting of the N-terminal 81 amino acids of the human ␤-1,4-galactosyltransferase and YFP (YFP-GalT) (31). Once more, we found that X11␣ and YFP-GalT co-localized in punctate structures in distal dendrites (Fig. 5, D and E, arrows). Interestingly, only 27.9 Ϯ 8.74% of X11␣ co-localizes with giantin and 43.8 Ϯ 6.11% with YFP-GalT (Fig. 5F), suggesting that is predominantly located in trans-Golgi compartments. Interestingly, X11␣ was found to co-localize with both giantin and YFP-GalT at dendritic branch points (Fig. 5G), a region where Golgi complexes consisting of medial, cis-, and trans-Golgi compart-

X11-␣ Modulates Kalirin-7 Activity
ments (30). Taken together, these data indicate that X11␣ is preferentially associated with a vesicle pool in proximity of the Golgi apparatus, as well as with dendritic Golgi outposts.

DISCUSSION
Based on our data, we propose the following model of X11␣/ kalirin-7 interaction (Fig. 8). Kalirin-7 is targeted to dendritic spines through its interactions with several PDZ domain-containing proteins, including PSD-95 and afadin. In dendritic spines, kalirin-7 enhances Rac1 activation, leading to actin cytoskeletal rearrangements, spine growth, and recruitment of AMPA receptors. In addition to PSD-95, kalirin-7 also interacts with X11␣ in spines. This interaction with X11␣ may inhibit kalirin-7's Rac1-GEF activity, similarly to what has been observed with PSD-95 (10). At the same time, X11␣ promotes the removal of kalirin-7 from spines and the sequestration to Golgi outposts present in the dendrite. This leads to a reduction of kalirin-7 activity in spines and its enrichment in Golgi outposts, where it could potentially participate in compartmentspecific signaling pathways.
We have previously shown that X11␣ interacts with the C terminus of kalirin-7 in a yeast two-hybrid assay, but whether this occurred in physiological preparations was not clear. Here, we show that X11␣ and kalirin-7 interact in native brain tissue and in cortical neurons and,, moreover, that this interaction occurs in a PDZ-dependent manner. Kalirin-7 interacts with a number of distinct classes of PDZ domain-containing proteins, and many such interactions appear to fulfill different functions and mediate the association of kalirin-7 with different protein complexes (32). For example, PSD-95 mediates the association of kalirin-7 with glutamate receptors, ErbB4, and potentially serotonin receptors (7,8,10,33). Afadin mediates the associa-tion of kalirin-7 with N-cadherin complexes and EphB receptors (5,34). Interestingly, several of the proteins found to interact with the C terminus of kalirin-7 in a yeast two-hybrid screen, including PSD-95 and afadin, are involved in membrane trafficking, suggesting roles for kalirin signaling in trafficking (10).
X11␣ and its homolog X11␤ are mLin-10 family members encoded by different genes, both present in neurons. Although these proteins play important roles in axons and the presynaptic terminal (35,36), several laboratories have investigated their dendritic roles. Stricker and Huganir (15) have examined the functions of X11/mLin-10 in neuronal dendrites and synapses, with special emphasis on X11␤. They have shown that X11␤ is highly enriched in the trans-Golgi complex of neurons and is present in punctate structures in dendrites. However, X11␤ was only occasionally present in spines and synapses. Our data indicate that X11␣ is enriched in dendrites and spines, as well as Golgi outposts and the Golgi complex. Interestingly, Stricker and Huganir (15) show that X11␤ associates with GluA1 and GluA2 subunits of the AMPA receptors, and it affects GluA1 surface expression. In addition, Setou et al. (16) have shown that mLin-10 interacts with KIF17 to mediate the trafficking of the NR2B subunit of NMDA receptors in dendrites. Here, we show that in addition to structures in dendrites, X11␣ is present in synapses, and it co-localizes with the postsynaptic and excitatory synapse proteins, PSD-95, GluA1, and kalirin-7.
Our live-cell imaging data also demonstrate that X11␣ is a moderately mobile protein. Comparisons of the half-recovery time (1 ⁄ 2 ) for GFP-X11␣ with that previously published for SAP102 and PSD-95 (37) reveal that X11␣ is less mobile than these synaptic proteins in spines: 1 ⁄ 2/2 for both SAP102 and PSD-95 was around 45 s (37), although our data indicate that X11␣ has a 1 ⁄ 2 of ϳ67 s. Importantly, both SAP102 and PSD-95 are considered to be highly mobile proteins at synapses (37); therefore, our data suggest that X11␣ is a moderately mobile protein in the spine. Moreover, the increased stability of X11␣ FIGURE 8. Model of X11␣/kalirin-7 interaction. Kalirin-7 enhances Rac1 activation in dendritic spines, leading to actin cytoskeletal rearrangements, spine growth, and recruitment of AMPA receptors. X11␣ binding to kalirin-7 inhibits its GEF activity, recruits it to Golgi outposts in the dendrite, and induces spine shrinkage. DECEMBER 19, 2014 • VOLUME 289 • NUMBER 51 in comparison with SAP102 or PSD-95 further supports the suggestion that X11␣ weakly or transiently interacts with other proteins in the PSD, such as kalirin-7. Interestingly, GFP-X11␣ displayed differential mobility dynamics in dendritic spines versus the dendritic shaft. This difference in mobility dynamics is likely due to an increased residency of X11␣ within spines. However, differences in mobility may also be due to the mechanism(s) that regulate X11␣ trafficking within these two cellular compartments. Moreover, the diffusion coefficient of GFP-X11␣ in both compartments was not significantly different (diffusion coefficient (m 2 /s)): spine, 0.0025 Ϯ 0.00057, and shaft, 0.0024 Ϯ 0.00066; p ϭ 0.96, graph not shown), suggesting that the observed difference in mobility was not due to a differential ability of GFP-X11␣ to diffuse into an area of similar size in either structure. This further supports the idea that the difference in mobility dynamics of this protein in each compartment is due to either an increase in stability of X11␣ in spines or differential trafficking mechanisms. However, we need to note that the spine neck may also play a part in increasing the way that X11␣ is trafficked, as well as its residency within the spine. Indeed, to assess how membrane topology influences the mobility and diffusion of X11␣, it would be necessary to take into account the morphology of the subcellular compartment being investigated (38). Nevertheless, these data suggest that X11␣ may fulfill multiple functions in these locations, likely modulating the localization and function of dendritic and synaptic proteins. We also report that X11␣ is present at synapses where it weakly associates with the PSD. Furthermore, the weak presence of X11␣ in detergent fractions is consistent with the subcellular localization of kalirin-7 (6, 10), strengthening the suggestion that these proteins are capable of interacting in vivo.

X11-␣ Modulates Kalirin-7 Activity
An interesting observation in this study was that X11␣ associates with Golgi outposts in dendrites and modulates the localization of kalirin-7 to these structures. Although the Golgi apparatus localizes in a perinuclear region of the neuronal soma, cellular compartments called "Golgi outposts" have also been identified in dendrites (28,31). Golgi outposts are often localized near synapses or at dendritic branch points (29). Recently, it has been shown that not all Golgi outposts contain the biochemcially distinct medial, cis-, and trans-Golgi compartments (30). Indeed, although we observed co-localization of X11␣ with giantin, a resident protein found in the medial and cis-Golgi compartment protein, a larger portion of the protein was found to co-localize with the trans-Golgi compartment marker, YFP-GalT. Importantly, it should be noted that it is not possible to determine whether X11␣'s co-localization with giantin or YFP-GlaT indicates that it present in only single compartment Golgi or whether a subset are multicompartment Golgi. Moreover, the implications of X11␣'s apparent predominance in trans-Golgi compartment is currently unknown, as the functional significance of single or multicompartment Golgi in dendrites is currently unknown (30). However, X11␣ is present at branch points where it co-localizes with both giantin or YFP-GalT, indicating that X11␣ is present in Golgi complexes consisting of medial, cis-, and trans-Golgi compartments. Golgi outposts play central roles in local post-endoplasmic reticulum trafficking of proteins used in dendrite growth and synaptic plasticity (27). These structures may function as a passive site for docking inactivated kalirin-7, to be released when needed. However, kalirin-7 may fulfill specific functions in Golgi outposts, possibly associated with protein trafficking, as kalirin has been linked to the trafficking of transmembrane proteins, such as peptidylglycine ␣-amidating monooxygenase (39). Future studies will need to explore the functional impact of the targeting of kalirin-7 and X11␣ to Golgi outposts. One potential role may be in actin remodeling in Golgi outposts. Actin remodeling by small GTPases plays an important role in the morphology and function of the Golgi apparatus (27), and kalirin-7 may play such a regulatory role in Golgi outposts.
We have previously reported that disrupting the interaction of PSD-95 with kalirin-7 removes it from synapses and inhibits its GEF activity (7,10). Our data suggest hat X11␣ is capable of modulating kalirin-7's presence at synapses by recruiting it to Golgi outposts, and overexpression of X11␣ results in a reduction in dendritic spine size, a consequence of inhibiting kalirin's GEF activity and removing it from synapses (7,10). Therefore, it possible that by recruiting kalirin-7 to Golgi outposts, one outcome would be to reduce its GEF activity, thus resulting in a reduction in spine size. Such inhibitory interactions might protect spines from overactivation of Rac1, which has detrimental effects. Inhibition of kalirin-7 GEF activity by another protein, DISC1, has been shown to protect spines from the detrimental effects of Rac1 overactivation that might lead to disease (40). Interestingly, X11␣ modulation of kalirin-7 localization opposes that of PSD-95; PSD-95 recruits kalirin-7 to spines, whereas X11␣ seems to remove kalirin-7 from spines. However, both proteins may reduce kalirin-7's GEF activity (10). Thus, X11␣-mediated regulation of kalirin-7 localization thus might be a more general mechanism of repressing kalirin-7 GEF activity in spines.
Proper regulation of dendritic spine morphology is a critical component of healthy brain function. Conversely, disruptions in the mechanisms and proteins that control spine morphology have consistently been associated with neurodevelopmental, psychiatric, and neurodegenerative diseases (2). Abnormal kalirin-7 signaling has also been linked with several disorders, including schizophrenia, Alzheimer disease, and Alzheimer disease with psychosis (32). X11/mLIN-10 proteins have been implicated in amyloid precursor protein processing (41) and apoE receptor trafficking (42), and their deletion decreases amyloid production in Alzheimer mouse models (41). Future investigations of kalirin-7/X11␣ interaction might provide insight into the pathogenesis of neurological disorders.