Scholarly article on topic 'A Dendritic Golgi Satellite between ERGIC and Retromer'

A Dendritic Golgi Satellite between ERGIC and Retromer Academic research paper on "Biological sciences"

CC BY-NC-ND
0
0
Share paper
Academic journal
Cell Reports
OECD Field of science
Keywords
{}

Abstract of research paper on Biological sciences, author of scientific article — Marina Mikhaylova, Sujoy Bera, Oliver Kobler, Renato Frischknecht, Michael R. Kreutz

Summary The local synthesis of transmembrane proteins underlies functional specialization of dendritic microdomains in neuronal plasticity. It is unclear whether these proteins have access to the complete machinery of the secretory pathway following local synthesis. In this study, we describe a probe called pGolt that allows visualization of Golgi-related organelles for live imaging in neurons. We show that pGolt labels a widespread microsecretory Golgi satellite (GS) system that is, in contrast to Golgi outposts, present throughout basal and apical dendrites of all pyramidal neurons. The GS system contains glycosylation machinery and is localized between ERGIC and retromer. Synaptic activity restrains lateral movement of ERGIC, GS, and retromer close to one another, allowing confined processing of secretory cargo. Several synaptic transmembrane proteins pass through and recycle back to the GS system. Thus, the presence of an ER-ERGIC-GS-retromer microsecretory system in all neuronal dendrites enables autonomous local control of transmembrane protein synthesis and processing.

Academic research paper on topic "A Dendritic Golgi Satellite between ERGIC and Retromer"

Cell Reports

Report

A Dendritic Golgi Satellite between ERGIC and Retromer

Graphical Abstract

Authors

Marina Mikhaylova, Sujoy Bera, Oliver Kobler, Renato Frischknecht, Michael R. Kreutz

Correspondence

kreutz@lin-magdeburg.de

In Brief

It is unclear whether post-endoplasmatic reticulum carriers in dendrites require a Golgi-related compartment for glycosylation or whether they bypass the Golgi. Mikhaylova et al. find a widespread ER-ERGIC-Golgi satellite-retromer microsecretory system in all dendrites of pyramidal neurons through which a spectrum of synaptic transmembrane proteins might pass and even recycle.

Highlights

• A dendritic ER-ERGIC-Golgi satellite (GS)-retromer microsecretory system is described

• GS system contains glycosylation machinery and is localized between ERGIC and retromer

• Synaptic transmembrane proteins pass through and recycle back to the GS system

• GS system enables local transmembrane protein processing in all neuronal dendrites

Mikhaylova et al., 2016, Cell Reports 14,189-199 ciossMark January 12, 2016 ©2016 The Authors

http://dx.d0i.0rg/l 0.1016/j.celrep.2015.12.024

CelPress

Cell Reports

Report

A Dendritic Golgi Satellite between ERGIC and Retromer

Marina Mikhaylova,1'4'6 7 Sujoy Bera,17 Oliver Kobler,3 Renato Frischknecht,2 and Michael R. Kreutz15 *

1RG Neuroplasticity

2Department of Neurochemistry/Molecular Biology

Combinatorial Neuroimaging Core Facility (CNI)

Leibniz Institute for Neurobiology, 39118 Magdeburg, Germany

4Emmy-Noether Group "Neuronal Protein Transport''

5Leibniz Guest Group "Dendritic Organelles and Synaptic Function''

University Medical Center Hamburg-Eppendorf, Center for Molecular Neurobiology, ZMNH, 20251 Hamburg, Germany 6Cell Biology, Faculty of Science, Utrecht University, 3584 Utrecht, the Netherlands 7Co-first author

'Correspondence: kreutz@lin-magdeburg.de http://dx.doi.org/10.1016/j.celrep.2015.12.024

This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

SUMMARY

The local synthesis of transmembrane proteins underlies functional specialization of dendritic microdomains in neuronal plasticity. It is unclear whether these proteins have access to the complete machinery of the secretory pathway following local synthesis. In this study, we describe a probe called pGolt that allows visualization of Golgi-related organelles for live imaging in neurons. We show that pGolt labels a widespread microsecretory Golgi satellite (GS) system that is, in contrast to Golgi outposts, present throughout basal and apical dendrites of all pyramidal neurons. The GS system contains glycosylation machinery and is localized between ERGIC and retromer. Synaptic activity restrains lateral movement of ERGIC, GS, and retromer close to one another, allowing confined processing of secretory cargo. Several synaptic transmembrane proteins pass through and recycle back to the GS system. Thus, the presence of an ER-ERGIC-GS-retromer microsecretory system in all neuronal dendrites enables autonomous local control of transmembrane protein synthesis and processing.

INTRODUCTION

In recent years, it has become apparent that a satellite secretory system exists in neuronal processes that allows for local synthesis, processing, and insertion of synaptic transmembrane proteins (Ye et al., 2007; Hanus and Ehlers, 2008; Ramirez and Couve, 2011; Cui-Wang et al., 2012; Hanus and Schuman, 2013). An open question is whether post-endoplasmic reticulum (ER) carriers in dendrites require a Golgi-related compartment for glycosylation or whether they bypass the Golgi and synaptic transmembrane proteins and are inserted without mature glyco-

sylation (Hanus and Schuman, 2013). In a subset of hippocampal pyramidal neurons, neuronal Golgi localizes to the perinuclear region and dendrites as tubulo-vesicular structures called Golgi outposts (GOs; Horton et al., 2005). GOs have been implicated in sorting, trafficking, and posttranslational modification of post-ER carriers (Horton and Ehlers, 2003; Horton et al., 2005; Ye et al., 2007; Jeyifous et al., 2009), as well as acentrosomal microtubule nucleation (Ori-McKenney et al., 2012). However, GOs exist only in the most proximal part of the apical dendrite or at branch points of only one dendrite and are mainly abundant in neuronal development (Horton and Ehlers, 2003; Horton et al., 2005). A recent study suggests that most dendritic post-ER carriers use a satellite secretory system devoid of classical Golgi membranes and instead pass through a direct pathway of ER, ER-Golgi intermediate compartment (ERGIC), and plasma membrane (Hanus et al., 2014). Taking into account the low abundance of GOs in mature dendrites, as well as their presence in only a subset of mature neurons (Horton et al., 2005; Hanus and Ehlers, 2008; Hanus and Schuman, 2013), locally synthesized proteins might not undergo all processing steps of the canonical secretory pathway and could therefore potentially be functionally different from somatically synthesized ones. Thus, faster and spatially restricted delivery may come at the expense of functional maturity and protein stability. Choy et al. (2014) recently presented evidence for a high density of retromer in dendrites and a role of retromer in fast, local delivery of various neurotransmitter receptors from endosomes to adjacent plasma membrane domains. The classical retromer pathway, however, can be seen from the endosome to the frans-Golgi network (TGN), and the question thus arises as to whether TGN-like elements akin to GOs might exist close to retromer and endosomes.

A few studies have reported a more widespread distribution of a subset Golgi-related proteins in distal dendrites and spines (for early work, see Torre and Steward, 1996; Gardiol et al., 1999; Pierce et al., 2001), raising the possibility that Golgi-derived membranes might be part of a satellite dendritic microsecretory system. However, conflicting results about the localization and abundance of Golgi proteins have been reported, and the use

Figure 1. pGolt Labels a Widespread GS-Related Secretory Microsystem in Dendrites

(A) GFP-Calneuron-2 localizes to somatic Golgi but Is also found In dendritic clusters In hippocampal primary neurons at 9 DIV. Scale bar, 10 mm.

(B) Heterologously expressed GFP-Calneuron-2 and pGolt-mCherry completely co-localize in primary hippocampal neurons (9 DIV). Scale bar, 10 mm. Right panel: high magnification inlet of the box indicated in the left panel. Scale bar, 2 mm.

(C) In contrast to GFP-Calneuron-2, expression of pGolt has no effect on the surface expression of SEP-GluN1 discernible from cells transfected with mCherry alone. Graph represents mean ± SEM. Student's t test, *p < 0.05.

(legend continued on next page)

of molecular markers to identify the Golgi is complicated by Golgi membranes cycling continuously between the ER and the c/'s-Golgi, as well as between the frans-Golgi and the endo-somes or retromer, introducing ambiguity into the interpretation of steady-state labeling in the absence of functional data and dynamic visualization.

Here, using a Golgi tracker called pGolt, we find Golgi-related organelles in all dendrites of pyramidal neurons and close to the ERGIC and retromer. This Golgi satellite (GS) secretory system contains glycosylation machinery but lacks many protein components for sorting and organization of Golgi cisternae. Moreover, we find that a spectrum of synaptic transmembrane proteins might pass and even recycle through these organelles. Collectively, the data suggest GSs might enable local glycosyla-tion of proteins that can then be recruited to membranes in spatially confined dendritic segments.

RESULTS

pGolt Allows for Visualization of GSs in Hippocampal Neurons

pGolt is based on the transmembrane domain (TMD) of the tail-anchored EF-hand Ca2+-sensor protein Calneuron-2 (Mikhay-lova et al., 2006, 2009). Calneurons are targeted efficient to membranes of the TGN after posttranslational insertion into the ER (Mikhaylova et al., 2009; Hradsky et al., 2011). In a direct comparison, Calneuron-2 exhibited superior co-localization with the TGN Golgi marker syntaxin-6 as compared to Calneuron-1 (Mikhaylova et al., 2009; Figures S1A-S1C). We therefore fused the last 28 amino acids of Calneuron-2 containing its TMD to mCherry in cytosolic orientation and inserted this construct into a pmCherry-N1 vector (Figure S1A). To further improve targeting to the Golgi, we added a 12-amino acid-long tandem ER export sequence from the protein Scap (Sun et al., 2007). We then checked Golgi targeting in comparison to established markers in COS7 cells and found that pGolt efficiently localizes to the Golgi apparatus (GA; Figures S1C and S1D).

Given that Calneuron-2 is highly enriched in TGN membranes, we next set out to test whether pGolt might be applicable to visualize Golgi-related organelles in hippocampal primary neurons. We observed that GFP-Calneuron-2 exhibited a punctate accu-

mulation in dendrites of adult neurons and pGolt-mCherry almost perfectly co-localized with these clusters (Figures 1A and 1B). However, in contrast to Calneuron-2, which inhibits phosphatidylinositol 4-kinase IIIp and thus reduces Golgi-to-plasma membrane trafficking (Mikhaylova et al., 2009; Graham and Burd, 2011), heterologous expression of pGolt had no effect discernible from expression of mCherry alone on surface expression of a super ecliptic pHluorin (SEP)-tagged GluN1 subunit of N-methyl-D-aspartate receptors (NMDARs; Figure 1C).

In live-imaging experiments, dendritic pGolt-mCherry fluorescence clusters remained stationary for periods longer than 15 min (Figures 1D and 1E). Larger clusters (about 1-1.5 mm) exhibited confined bidirectional movements (within the area of about 1 mm) when imaged at a high frame rate (Figure 1E). Fewer and smaller directionally moving vesicles were sometimes seen in dendrites (Figure 1E), and fusion or fission of these carriers from the larger pGolt-positive structures could be observed (Movies S1 and S2). Live imaging over 11 hr revealed that pGolt-mCherry-labeled organelles are rarely transported from soma to dendrites even though they can undergo occasionally long-range anterograde and retrograde trafficking (Figure 1D; Movie S1).

GSs Show a Widespread Distribution in Hippocampal Primary Neurons

We next examined the distribution of GSs in dendrites of hippocampal primary neurons. We found that pGolt-positive organelles are present in most dendrites (>90%) in all transfected pyramidal neurons. In contrast to GOs and extended Golgi, GSs are considerably smaller (Figures 1D and 1G-1K). Moreover, they are not preferentially or exclusively present in the primary apical dendrite but also are found in secondary and tertiary dendrites (Figure 1F), and we found that less than 30% of GSs are located at dendritic branch points in mature neurons (Figure 1F). To further analyze the distribution of pGolt puncta in dendrites, we measured the total mean intensity of pGolt in the major and minor dendrites and calculated a polarity index (PI). For uniformly distributed organelles and proteins, the PI = 0, whereas enrichment in major or minor dendrites results in PI < 0 or PI > 0, respectively. Analysis of the PI of pGolt revealed a uniform distribution of GSs with a small tendency for a stronger signal in the

(D) Overview shows a hippocampal primary neuron (17 DIV) transfected with pGolt-mCherry for 24 hr. pGolt-mCherry localizes to the soma and labels small organelles in dendrites. These structures have a size below 1 mm and are found at branch points, along dendrites, and at the base of spines (inset). See also Movie S1.

(E) Kymographs from ROIs show that GSs are relatively immobile. See also Movie S2.

(F) Quantification of dendrites positive for pGolt versus total number of dendrites (far left), localization of pGolt to dendritic branch points (left), and distribution of pGolt-positive organelles in primary, secondary, and tertiary dendrites of adult hippocampal neurons (11-17 DIV; middle) are shown. The PI of MAP2, GM130, and pGolt (11 neurons, 8-10 DIV for each group) shows high preference of GM130 for the major dendrites (right); pGolt distribution is less polarized.

(G) GSs localized to the base of growing dendritic protrusion during dendritogenesis (8 DIV) and might serve as the supply of membrane material (left). The graph shows the normalized intensity of pGolt3-mCherry in dendritic protrusion in relation to the size of the protrusion as measured overtime (right).

(H) pGolt expressed specifically under control of asynapsin promoter in neurons at low levels is equally targeted to neuronal somataand dendritic compartments. Lower panel: high magnification of soma and minor dendrites with a line scan along the somatic Golgi complex and dendritic processes that gave rise to the fluorescence intensity profile shown below.

(I-K) Super-resolution live STED microscopy shows neurons transfected with pGolt-SNAP-mCherry (8 and 9 DIV) and labeled by SiR-647. See also Figure S1F. (I and J) Examples of STED images reveal various shapes and sizes of pGolt-positive organelles with a small fraction of vesicles below 250 nm, most organelles spread between 250 and 750 nm, and a fraction (below 10%) of larger structures (N = 12, n = 61). (K) Live-cell STED imaging shows that pGolt-labeled organelles are relatively stationary but constantly changing shape and fuse with smaller vesicles (white arrow). Scale bar, 1 mm. See also Figure S1.

Figure 2. pGolt-Labeled Organelles Are Sensitive to Golgi Disruption with Brefeldin A and Contain Glycosylation Machinery

(A) Time-lapse imaging was done over 11 hr. Red arrows point to stable dendritic GSs visualized with pGolt-mCherry In control neurons, and red arrows with a white fill point to pGolt-mCherry vesicles disappearing after treatment with brefeldin A. Scale bar, 5 mm.

(B) Both pGolt-mCherry and St3gal5-GFP label somatic and dendritic Golgi compartments (blue arrows) and co-localize with GM130 only in the soma and the GOs deriving from the extended Golgi (red arrow).

(C) pGolt-mCherry-labeled compartments co-localize with overexpressed St3gal5 throughout the dendritic arbor. This panel is related to Figure S2A.

(D) pGolt-mCherry-labeled structures in dendrites contain glycosylation machinery. Fixed hippocampal neurons were stained with AAL-Fluo to visualize fuco-sylated proteins.

(E) pGolt-mCherry-labeled compartments co-localize with post-Golgi carrier marker Rab6a-GFP in extended somatic Golgi and in dendrites. A representative image of a distal dendrite is shown in Figure S2G.

(F) Rablb and pGolt-mCherry co-localize in the soma and to some extent in dendrites.

(G) PSA-NCAM1 co-localizes with pGolt-mCherry puncta but not with ERGIC-53-GFP, as evidenced by immunofluorescence staining with a PSA-NCAM1 antibody. Scale bar, 2 mm.

(H) Quantification of co-localization is shown between pGolt-labeled compartments and St3gal5, AAL-Fluo, PSA-NCAM, Rab6a, and Rablb.

(legend continued on next page)

major dendrite, similar to those of the dendritic marker MAP2. In stark contrast, GM130, an established GO marker, showed strong polarization toward the major dendrite (Figure 1F). In younger neurons (8 days in vitro [DIV]), GSs localize more frequently to the base of growing dendritic protrusion and might, like GOs, serve as the supply of membrane material during dendritogenesis (Figure 1G). In live-imaging experiments, we frequently found a decrease of pGolt-mCherry fluorescence at branch points at GSs that is concomitant with an increase of the area of the growing protrusion (Figure 1G). pGolt, expressed under control of a human synapsin promoter, allows for long-term imaging with comparable fluorescence intensity in somatic and extended Golgi and GSs (Figure 1H).

We next performed a more detailed morphological analysis of GSs, employing live-cell stimulated emission depletion (STED) super-resolution microscopy (Figures 11-1K) based on the far red SiR-SNAP substrate (Figure S1E; Lukinavicius et al., 2013). GSs exhibit not a round or ecliptic shape but rather an irregular and frequently extended and hollow structure mainly in the size range of 250-1,000 nm (Figures 1I and 1J; Figure S1E). Smaller (<250 nm) pGolt-positive structures (<5%), which represent most likely mobile vesicles, frequently fuse with larger stationary GSs (Figure 1K). Larger structures (>1,000 nm) were found in less than 10% of total measured organelles (Figure 1J) and most likely represent, to a large extent, pGolt-labeled GOs.

GSs Are Sensitive to Golgi Disruption by Brefeldin A and Are Not Continuous with Other Dendritic Organelles

We next performed a series of experiments to demonstrate that pGolt-positive GSs visualized in dendrites have a Golgi origin. To this end, we applied brefeldin A to hippocampal primary neurons transfected with pGolt-mCherry. We observed that brefeldin A-induced fragmentation and collapse of the Golgi into the ER also resulted in a slow decrease of punctate pGolt-mCherry fluorescence and a disappearance of pGolt-labeled organelles within the 11 hrof imaging (Figure 2A). pGolt-labeled organelles were not co-localizing with markers for lysosomes (LAMP1), early endosomes (EAA1), endosomal sorting complexes required for transport (ESCRT/TSG101), mitochondria (Mito-tracker), and peroxisomes (PEX14; Figures S1F-S1I).

Rab6a exhibits a dynamic localization at the GA and regulates trafficking in both a retrograde direction (from the early endosomes and Golgi to the ER) and an anterograde direction (from the Golgi to the plasma membrane; Liu and Storrie, 2012). We found considerable Rab6a-GFP fluorescence at pGolt organelles (Figures 2E and 2H) and, to a lesser extent, immunofluorescence of endogenous Rab1b (Figures 2F and 2H), a protein that regulates vesicular transport between ER and Golgi.

GSs Contain Glycosylation Machinery

We reasoned that GSs might serve the requirement for local glycosylation in dendrites as part of a larger satellite microsecre-

tory system and therefore asked next whether glycosylation machinery is present. ST3 p-galactoside a-2,3-sialyltransferase 5 fused to GFP (St3gal5-GFP) showed extensive co-localization with GSs in dendrites (94.8% ± 2.7%; Figures 2B, 2C, and 2H; Figure S2A). Moreover, fluorescein-conjugated Aleuria aurantia lectin (AAL-Fluo), which preferentially binds to fucose linked (a-1,6) to N-acetylglucosamine or fucose linked (a-1,3) to N-ace-tyllactosamine-related structures, strongly stained almost all pGolt-positive GSs (94.6% ± 2.6%; Figures 2D and 2H). The TMD of pGolt is based on Calneuron-2, which is of low abundance in hippocampus (Mikhaylova et al., 2006), but we observed co-localization of St3gal5-GFP with those of endogenous Calneuron-1 in dendrites of hippocampal primary neurons (Figure S2B). NCAM1 can undergo polysialylation (PSA) in response to synaptic activity, and staining with a PSA-specific antibody showed that about 38% of pGolt-labeled organelles contain PSA-NCAM1 (Figures 2G and 2H). As expected ERGIC-53 showed no co-localization with PSA-NCAM1 immunofluorescence (Figure 2G). Taken together, these data indicate that pGolt accumulates in GSs containing glycosylation machinery.

pGolt Efficiently Visualizes GSs in Dendrites

We next asked why this structure might have been largely overlooked in previous studies. Dynamics of neuronal Golgi in den-drites and post-Golgi secretory trafficking in living mammalian neurons has mainly been studied with temperature-sensitive vesicular stomatitis virus G protein (VSVG)-GFP-ts045 (Horton and Ehlers, 2003; Horton et al., 2005). This system gives control on the subcellular localization of VSVG-GFP-ts045 via sequential changes in temperature, but the method has significant downsides because it requires steep temperature shifts in culture conditions, which likely interfere with neuronal function. The GA is a highly dynamic organelle with few stable protein components, and we reasoned that VSVG-GFP-ts045 might exhibit a too-high exchange rate of transiting proteins to visualize GSs for longer periods. To test this notion, we performed a VSVG trafficking assay and compared the distribution of pGolt-tar-geted mCherry to the distribution of VSVG-GFP-ts045 directly after removal of the Golgi block by shifting temperature from 20°C to 32°C (Figures 2I and 2J). Both probes are initially in the same Golgi compartment, further supporting the notion that GSs have a Golgi origin. However, pGolt-mCherry fluorescence remained stationary longer than VSVG-GFP-ts045 at later time points (>20 min; Figure 2K), indicating a longer residing time and lower exchange rate at GSs as compared to VSVG-GFP-ts045.

Another probe used for live imaging of GOs is a fragment of mannosidase II (ManII) containing the TMD and an intraluminal stretch of 100 amino acids coupled to GFP (Ori-McKenney et al., 2012). We observed that this probe mainly accumulates in somatic Golgi (Figures S2C and S2D) and requires a significantly

(I) pGolt-mCherry stained GSs completely co-localize with VSVG-GFP-ts045, a widely used marker to visualize secretory trafficking through Golgi-related organelles.

(J) Kymographs depict the motility of VSVG-GFP-ts045 and pGolt-mCherry in dendrites directly after transferring neurons from 20°C to 32°C. Scale bar, 5 mm. (K) Time-lapse imaging reveals that pGolt, in contrast to VSVG-GFP-ts045, is stably residing at GSs. Scale bar, 5 mm. See also Figure S2.

Figure 3. Dendritic GSs Are in Close Spatial and Temporal Relation with ERGIC and Retromer

(A) Confocal image of live neuron (14-16 DIV) shows GS (pGolt-mCherry) localization In relation to ERGIC-53-GFP. ERGIC-53 Is distributed unevenly. Scale bar, 2 mm. Right panel: distance-fluorescence intensity profiles are shown.

(B) Frequency distribution analysis shows distances between localized centers of pGolt-mCherry and ERGIC-53-GFP positive organelles. The 60 pairs were measured from seven cells.

(C) GS is positioned close to retromer (Vps35-GFP). Scale bar, 2 mm. Right panel: distance-fluorescence intensity profiles are shown.

(D) Frequency distribution analysis shows distances between localized centers of pGolt-mCherry and Vps35-GFP positive organelles. The 64 pairs were measured from 12 cells.

(legend continued on next page)

higher level of expression to visualize vesicular structures in dendritic compartments (Figures S2C and S2D). Imaging of moderately expressing ManII-N100-GFP with stronger illumination and longer exposure times allowed for visualization of dendritic GSs (Figure S2D), but fast bleaching of weakly fluorescent dendritic puncta with illumination can easily induce photo-toxicity in living cells. ManII-N100-GFP-labeled dendritic puncta showed a complete overlap with pGolt-mCherry (Figures S2C and S2D), further demonstrating the Golgi origin of pGolt-labeled compartments.

Finally, overexpressed pGolgi, a Golgi probe based on the amino-terminal 81 amino acids of p1,4-galactosyltransferase (Aoki et al., 1992; Yamaguchi and Fukuda, 1995), remained largely in the soma and could not be used in live-imaging experiments to visualize Golgi-related organelles in dendrites unless extremely high levels of overexpression were achieved (Figure S2C). Collectively, the data suggest that pGolt at a moderate expression level labels GSs in dendrites and exhibits a more restricted localization than other published probes. Moreover, pGolt expression levels can be easily titrated for long-term imaging (i.e., >24 hr) by using p-actin or synapsin promoters (Figure 1H; Figures S1G and S2D) that still allow for visualization of GSs.

Immunoreactivity of cis- and trans-Golgi markers including GM130, syntaxin-6, TGN38, TGN46, and Golgin 97 was only occasionally associated with GSs (Figure S2E; data not shown), although a clear co-localization in somatic Golgi was apparent (Figure S2G). The best co-localization in dendrites was found with syntaxin-6 (Figure S2E). In a subset of cells we found in proximal dendrites co-localization with GM130, which most likely reflects labeling of extended Golgi in primary dendrites (Horton et al., 2005; Figure S2E). We confirmed the specificity of GM130 immunofluorescence at somatic and extended Golgi in primary dendrites with short hairpin RNA (shRNA) protein knockdown (Figure S2F). The relatively sparse GM130 antibody staining in dendrites was not affected by the shRNA knockdown (Figure S2F), suggesting cross-reactivity of the GM130 antibody. Taken together, these results show that many Golgi-matrix proteins have a low abundance, if they are even present, in the dendrites of most neurons.

GSs Are Localized between ERGIC and Retromer

We next asked how the localization of GSs relates to those of other dendritic organelles of the secretory pathway. We found that most GSs (>90%) are close to ERGIC-53-GFP (25 percentile, 2.2 mm, and 75 percentile, 6.2 mm, from center to center; Figures 3A and 3B) and the retromer marker Vps35-GFP (25

percentile, 1.9 mm, and 75 percentile, within 5.0 mm, from center to center; Figures 3C and 3D). Distance-fluorescence intensity profiling revealed a closer relationship of GSs to retromer than to ERGIC (Figures 3A-3D). The localization of pGolt did not overlap with both markers (Figures 3A-3D). We aimed to resolve the spatial relationship among ERGIC, GSs, and retromer in dendrites with triple labeling, co-transfected pGolt-mCherry with Vps35-GFP, and then fixed the cells and stained with an ERGIC-53 antibody. We found that pGolt-labeled GSs are almost invariably in the proximity of ERGIC and retromer (Figures 3E and 3G), whereas early endosomes labeled with an EEA1 antibody distribute uniformly over 10 mm in the vicinity of pGolt puncta (Figure S1I). Moreover, GSs were always positioned closer to fluorescence spots of both markers than Vps35-GFP and ERGIC-53 immunofluorescence to each other (Figures 3E-3G).

We next analyzed the dynamic spatio-temporal association of all three organelles and found that the residing time when they are close to one another is expectedly low, with a docking time below 1 min in more than 40% of all events recorded (Figures 3H and 3I). The residing time close to retromer and GSs exhibited a bimodal distribution in which more than 50% of interactions lasted less than 2 min and almost 40% of all docking events were longer than 4 min (Figures 3H and 3I). This finding might explain the tighter association of retromer and GSs compared to ERGIC after fixation of cells.

Like other dendritic organelles, pGolt-labeled GSs exhibited confined bidirectional movements when imaged at a high frame rate, and these oscillations were decreased after removing inhibitory tone with the GABAA-receptor antagonist bicuculline (Figures 3J and 3K). After stimulation with brain-derived neurotrophic factor (BDNF; BDNF signaling is a well-known activator of protein translation), the rapid confinement of motility was even more pronounced (Figures 3J and 3K). Immediately after application of BDNF, ERGIC-53-GFP and pGolt-mCherry mobility was simultaneously halted in register to each other (Figure 3L; Movie S3), further indicating a close spatial relationship between ERGIC and GS.

GSs Are Part of a Satellite Dendritic Microsecretory System

NMDARs are diverted from the somatic ER into a specialized ER subcompartment that bypasses somatic Golgi, merging instead with dendritic GOs (Jeyifous et al., 2009). We found that the GFP-tagged GluN1 subunit of NMDARs prominently associated with pGolt-positive GSs (Figures 4A and 4D; Movie S4). A recent

(E) Confocal image of a hippocampal neuron transfected with pGolt-mCherry and Vps35-GFP and stained for endogenous ERGIC-53. Right panel: high-magnification ROIs.

(F) Fluorescence intensity traces reveal the ROIs shown in (E), measured across indicated white lines.

(G) Distribution profile shows relative distances between centers of ERGIC-, pGolt-mCherry-, and Vps35-GFP-labeled compartments.

(H and I) Kymograph analysis shows the spatio-temporal interaction of ERGIC with GS (H) and GS with retromer (I). Examples of short-term (left) and long-term (right) association between organelles are depicted. Frequency distribution of residing time is plotted below.

(J) Kymograph analysis shows activity-dependent dynamics of pGolt-labeled GSs. Increased synaptic activity (bicuculline, 50 mM) reduced the bidirectional mobility of pGolt organelles (left panel), and this effect is even stronger after stimulation with BDNF (100 ng/ml; right panel). (K) Displacement analysis shows dendritic GSs (pGolt-mCherry) upon different stimulation conditions.

(L) Kymograph of ERGIC-53-GFP and pGolt-mCherry mobility demonstrates a close spatial relation between ERGIC and GS. Treatment with 100 ng/ml BDNF decreases the mobility of both markers in a synchronized manner. See also Movie S3. See also Figure S3.

report indicates that the synaptic cell adhesion protein Neuroli-gin-1 (NLG1) directly associates with GluN1 via a cis-interaction and might be an auxiliary subunit of GluN1 that associates with the receptor shortly after biogenesis (Budreck et al., 2013). In line with this hypothesis, we found that NLG1-GFP is frequently present in GSs, indicating an association with GluN1 before transport to the plasma membrane (Figures 4B and 4D). GluN1 diverts the somatic Golgi upon association with scaffolding molecules like SAP97 or CASK (Jeyifous et al., 2009), and the synaptic scaffolding protein Homer-1 can be found at pGolt-positive GSs (Figures 4C and 4D).

Surprisingly, we found that after transfection, all GFP- or myc-tagged synaptic transmembrane proteins tested, including Syn-decan-2, APP, Neuroplastin55, and GluA1, are present in GSs (FiguresS3Aand S3B). However, BDNF- orNPY-GFPexocytotic secretory vesicles did not (NPY, unless highly expressed) or only to a low extent (BDNF) pass through GSs (Figure S3C). Thus, processing of several different synaptic transmembrane proteins via this unconventional regulated secretion route is conceivable, and it is unlikely that only some proteins will be able to access this route.

A fluorescence recovery after photobleaching (FRAP) analysis of pGolt-EGFP fluorescence revealed a high recovery rate, as expected for a Golgi-related structure with approximately 50% recovery after 5 min (Figure S4A). Enhancing synaptic activity only modestly increased the exchange rate (Figure S4B). To directly assess whether this reflects increased demand for replenishment of synaptic membrane proteins, we also determined FRAP for GluN1- and NLG1-GFP. The FRAP curves for both proteins show slower recovery than pGolt-GFP (Figure S4A), and enhancing activity does not result in higher exchange rates at least for 5 min (Figures S4C and S4D).

Proteins Are Transported Out of GSs to the Plasma Membrane and Recycled Back

Rab11 has been localized to both Golgi and recycling endo-somes, and Rab11-positive recycling endosomes are one of the main routes for activity-dependent insertion of a-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors (Correia et al., 2008). Rab11 is known to regulate recycling of endocy-tosed proteins and post-Golgi transport (Welz et al., 2014). We found a dynamic association of Rab11a-GFP and GSs in live-imaging experiments with frequent docking of Rab11a-GFP positive vesicles (Figures S3D and S3E). In the final set of experiments, we therefore wanted to determine whether proteins are transported out of GSs to the nearby plasma membrane and might recycle back from there. To address these questions, we first checked whether NLG1-photo-activatable GFP (paGFP) after photo-activation accumulates at the membrane. We used pGolt-positive regions in dendrites as regions of interest (ROIs) and photo-converted paGFP in these ROIs. Time-lapse live imaging revealed that NLG1-paGFP accumulated at the membrane (Figures 4E and 4F). Within 30 min following photo-conversion, NLG1-paGFP fluorescence prominently accumulated at the dendritic plasma membrane close to pGolt structures (Figures 4E and 4F). Taken together, the data show that proteins exit from GSs in distal dendrites for plasma membrane trafficking.

In a recent study, indirect evidence was found that NMDARs might be one of the cargos delivered to the dendritic membrane via retromer (Choy et al., 2014). We therefore tested whether retromer may be involved in the recycling of NMDARs via GSs and performed live cell staining of myc-tagged GluN2B to label surface-exposed NMDARs with a myc antibody and thus follow up on the fate of internalized receptors. We observed an increased association of surface-labeled GluN2B with retromer over a period of 90 min following antibody application (Figures 4G and 4H). In addition, we observed localization of surface-labeled GluN2B in 38.8% of GSs after 30 min and 44.35% after 60 min (Figures 4I and 4J), which suggests that internalized GluN2B-containing NMDARs can pass through a GS-related organelle and that they might be recycled to the plasma membrane. Upon enhanced synaptic activity induced by bicuculline, we found a modest but statistically significant decrease of internalized NMDARs in GSs 60 min after stimulation (35.2%), suggesting that synaptic activity might modulate recycling of NMDARs via this satellite secretory system (Figures 4I and 4J). Thus, GSs might receive both anterograde traffic (i.e., from the ER) and retrograde traffic (i.e., from the plasma membrane, endosomes, and retromer).

DISCUSSION

In this study, we provide evidence for a Golgi-related satellite microcompartment in dendrites that is more widespread than previously described GOs (Horton et al., 2005). The results argue for the possibility that ion channels and transmembrane proteins produced in more distally dendritic branches can pass through this probably simplified Golgi microcompartment, and the data suggest that locally synthesized transmembrane proteins may use a direct ER-ERGIC-Golgi plasma membrane pathway. In line, pGolt-labeled structures are positive for ST3 p-galactoside a-2,3-sialyltransferase 5, ManII, and VSVG-GFP-ts045. A subset of pGolt-labeled organelles shows co-localization with Rab1b and Rab6a—small guanosine triphosphatases associated with the Golgi membranes and ER-to-Golgi and Golgi-to-plasma membrane carriers, respectively. In addition, we could show that proteins passing through this structure are inserted into the nearby plasma membrane and that a spectrum of synaptic transmembrane proteins co-localize with pGolt-labeled GSs.

GSs are found at the interface of ERGIC and retromer, a complex of proteins that is crucial in recycling transmembrane receptors from endosomes to the TGN. It was recently shown that p-adrenergic receptors and potentially ionotropic glutamate receptors are sorted for local endosomal membrane insertion via retromer (Choy et al., 2014). Thus, the close spatial relationship between retromer and GS suggest that this Golgi-related organelle might receive retrograde traffic of synaptic receptors. In accord with this notion, we found recycling of GluN2B-con-taining NMDARs through retromer and GSs, possibly after removal of sugar residues, and it is tempting to speculate that local re-glycosylation of the receptor might take place in this dendritic microsecretory system in confined dendritic segments. Finally, synaptic plasticity depends on sugar modification of synaptic membrane proteins like PSA of NCAM (Rutishauser, 2008; Senkov et al., 2012). Along these lines, it is plausible that this

Figure 4. GS System Is Involved in Delivery of Synaptic Proteins to the Plasma Membrane and Might Participate in Recycling of NMDARs

(A) The GFP-tagged GluNI subunit of the NMDAR associates with pGolt-mCherry-labeled GSs. Scale bar, 5 mm. See also Movie S4.

(B) NLG1-GFP is present in pGolt-labeled GSs. Scale bar, 5 mm.

(C) A fraction of the synaptic scaffolding protein Homer-1-GFP co-localizes with pGolt-mCherry. Scale bar, 5 mm.

(D) Quantification of the percentage of co-localization for (A)-(C).

(E and F) Live imaging shows hippocampal neurons transfected with pGolt-mCherry and NLG1-paGFP. Steals of single-plane images show individual frames for photo-converted NLG1-paGFP. Fluorescence intensity profiles along the perpendicular line across the dendrite near the photo-conversion point demonstrate increase of NLG1-paGFP fluorescence at the membrane over time. Scale bar, 5 mm.

(G) Internalized myc-NR2B is found at Vps35-GFP positive retromer-associated organelles. Live staining with anti-myc was followed by fixation at different time points (30, 60, and 90 min) and labeling with secondary antibody conjugated with Alexa 568. Scale bar, 2 mm.

(H) Quantification of percentages shows Vps35-GFP positive organelles overlapping with surface-labeled myc-GluN2B at different time points.

(I) Internalized myc-GluN2B is found at pGolt-mCherry positive Golgi-related organelles. Live staining with anti-myc was followed by fixation at different time points (30 and 60 min) and labeling with secondary antibody conjugated with Alexa 488. Synaptic activity enhances the internalization rate of GluN2B. Scale bar, 2 mm.

(J) Quantification of myc-GluN2B intensity shows live-stained puncta overlapping with pGolt3-mCherry (30 and 60 min). Graph represents mean ± SEM. Student's t test, *p < 0.05. See also Figure S4.

widespread GS system could be important for establishment of clustered plasticity, a form of synaptic plasticity restricted in a functionally segregated dendritic segment (Govindarajan et al., 2011). ERGIC post-ER carriers are mobile under resting conditions, and synaptic activity restricts this mobility in a CamK- and KIF-17-dependent manner (Hanus et al., 2014). It will be interesting to see how GS relates to ERGIC and retromer in the context of synaptic plasticity.

A question that arises in light of the widespread distribution of GSs is whether they differ from GOs only in size or also in function. pGolt will allow study of whether GSs are assembled locally or derive from somatic Golgi like GOs (Quassollo et al., 2015) and whether neuronal activity contributes to their formation (Thayer et al., 2013). At present, it is unclear whether both organelles can substitute for each other. GSs are devoid of staining for many Golgi-membrane markers, indicating a low abundance of these proteins, but pGolt labels both GOs and GSs. It appears unlikely that a GS system is an integral part of modifying, sorting, and packaging macromolecules for cell secretion like classical GA. We reason that the widespread distribution of GSs makes a local insertion of transmembrane proteins passing through this structure via retromer likely and that tightly regulated sorting and packaging processes might be dispensable. In Drosophila neurons, cis-, medial-, and trans-compartments of Golgi are often disconnected in dendrites in vivo (Zhou et al., 2014). Thus, the failure to detect Golgi cisternae in dendrites might be related to simplification of the Golgi structure and probably to no necessity existing for controlled local exit and sorting of post-Golgi cargo.

EXPERIMENTAL PROCEDURES

Cell Culture, Transfection, and Immunocytochemistry

Rat hippocampal primary cultures were prepared as described previously (Dieterich et al., 2008). For live-cell-imaging experiments, primary neurons were plated on 18 mm glass coverslips or 35 mm m-Dishes (Ibidi). Neurons were transfected with the cDNA plasmids using Lipofectamine 2000 (Invitro-gen) according to a protocol described by Kapitein et al. (2010). For experiments in young neurons, cells were transfected between 7 and 8 DIV. In the 12-48 hr following transfection, cells were either fixed for immunostaining or used for live cell imaging. For experiments in older neurons, neurons were transfected 13-19 DIV, and fixed and live-cell-imaging experiments were performed between 14 and 20 DIV.

Transfection of COS7 cells and immunocytochemical staining was done as described previously (Hradsky et al., 2011; Karpova et al., 2013). More details are included in Supplemental Experimental Procedures.

Wide-Field and Confocal Laser Scanning Microscopy

Live-imaging experiments were performed at 37°C and 5% CO2 in Tyroid's solution containing 10 mM HEPES pH 7.9, 145 mM NaCl, 5 mM KCl, 2 mM MgCl2, 2 mM CaCl2, and 10 mM glucose or conditioned neurobasal medium. Wide-field fluorescent image acquisition was performed using an inverted microscope (Axio Observer D1; Zeiss) equipped with an electron microscopy charge-coupled device camera (Evolve 512; Photometrics) controlled by MetaMorph (Molecular Devices) and VisiView (Visitron Systems) software, using a Zeiss 63x oil objective, Plan Apochromat, numerical aperture 1.4. The filters used are the EGFP ET filter set (exciter 470/40, emitter 525/50, dichroic 495 LP) or DAPI/FITS/Texas Red filter set (exciter 407/494/576, emitter 457/530/628, dichroic 436/514/604) from Chroma Technology. Confocal time lapses and fixed cell images were taken with 63x oil objective as single plane or z stacks (300 or 500 nm z step), depending on the sample, using a Leica SP5 microscope equipped with a krypton-argon-ion laser (488/

568/647 nm) and an acousto-optic-tunable filter for selection and Intensity adaptation of laser lines. Maximum intensity projections were calculated from each fluorescence channel of the image stack and analyzed with ImageJ (http://rsb.info.nih.gov/ij). For analysis of docking time among GS, ERGIC, ret-romer,and recycling endosomes, two-color confocal lapses of512 x 256 pixel were recorded with 63x oil objective and 4x confocal zoom as single planes, with a 5 s interval between individual frames. More details on FRAP, photo-activation, and STED imaging are included in Supplemental Experimental Procedures.

Data Analysis

The distribution of pGolt-positive compartments along the dendrites was quantified using ImageJ for image processing, and the number of puncta was counted manually. For measuring the distances between centers of organelles labeled by pGolt-mCherry, ERGIC-53-GFP, and Vps35-GFP, pGolt-mCherry, or EEA1, one of the markers was selected as center and a 10 mm lane was drawn in both proximal and distal directions along the dendritic fragment containing puncta of GFP, mCherry, or Alexa 647. Intensity profiles were plotted using ImageJ, and the distances between fluorescence intensity peaks in overlaid profiles were calculated. The proximity of localization was measure in GraphPad Prism using distribution frequency analysis. The docking time between organelles was analyzed by tracking individual vesicles and by kymograph analysis using ImageJ. For measuring the fluorescence intensity of NLG1-paGFP, a line perpendicularly crossing the dendrite outside of the photo-activation area was selected. The intensity profile, where the x axis represents distance along the line and the y axis is the pixel intensity, was plotted over different time points using ImageJ. Using kymograph analysis (ImageJ), motility of pGolt, VSVG, ERGIC-53, and other vesicular markers was analyzed. For determination of velocities and calculation of displacement, particles were tracked using the built-in "track object'' application in MetaMorph.

SUPPLEMENTAL INFORMATION

Supplemental Information includes Supplemental Experimental Procedures, four figures, and four movies and can be found with this article online at http://dx.doi.org/10.1016/j.celrep.2015.12.024.

AUTHOR CONTRIBUTIONS

M.M., S.B., O.K., and R.F. performed the experiments and analyzed the data. M.M., S.B., R.F., and M.R.K. designed the experiments. M.M. and M.R.K. wrote the manuscript.

ACKNOWLEDGMENTS

The authors gratefully acknowledge the professional technical assistance of C. Borutzki, S. Hochmuth, M. Marunde, J. Bar, and S. Wehrmann. We thank Dr. Kai Johnsson for providing the SiR-SNAP substrate. This work was supported by grants from the Bundesministerium für Forschung und Technologie (BMBF; Energi), Deutsche Forschungsgemeinschaft (DFG; Kr1879 3-1, 5-1, 6-1 SFB 779 TPB8), and Leibniz Foundation (Pakt für Forschung); Grant EU-FP7 MC-ITN NPlast to M.R.K.; a Schram grant to R.F.; and a European Molecular Biology Organization (EMBO) Long-Term Fellowship co-funded by MarieCurie Actions (EMBO ALTF 884-2011) and DFG Emmy-Noether Programm (Ml 1923/1-1) to M.M. O.K. is supported by the DFG grant SCHE 132/18-1.

Received: May 22, 2015 Revised: November 8, 2015 Accepted: November 25, 2015 Published: December 31, 2015

REFERENCES

Aoki, D., Lee, N., Yamaguchi, N., Dubois, C., and Fukuda, M.N. (1992). Golgi retention of a frans-Golgi membrane protein, galactosyltransferase, requires cysteine and histidine residues within the membrane-anchoring domain. Proc. Natl. Acad. Sci. USA 89, 4319-4323.

Budreck, E.C., Kwon, O.B., Jung, J.H., Baudouin, S., Thommen, A., Kim, H.S., Fukazawa, Y., Harada, H., Tabuchi, K., Shigemoto, R., et al. (2013). Neuroligin-1 controls synaptic abundance of NMDA-type glutamate receptors through extracellular coupling. Proc. Natl. Acad. Sci. USA 110, 725-730. Choy, R.W., Park, M., Temkin, P., Herring, B.E., Marley, A., Nicoll, R.A., and von Zastrow, M. (2014). Retromer mediates a discrete route of local membrane delivery to dendrites. Neuron 82, 55-62.

Correia, S.S., Bassani, S., Brown, T.C., Lise, M.F., Backos, D.S., El-Husseini, A., Passafaro, M., and Esteban, J.A. (2008). Motor protein-dependent transport of AMPA receptors into spines during long-term potentiation. Nat. Neuro-sci. 11, 457-466.

Cui-Wang, T., Hanus, C., Cui, T., Helton, T., Bourne, J., Watson, D., Harris, K.M., and Ehlers, M.D. (2012). Local zones of endoplasmic reticulum complexity confine cargo in neuronal dendrites. Cell 148, 309-321. Dieterich, D.C., Karpova, A., Mikhaylova, M., Zdobnova, I., Konig, I., Landwehr, M., Kreutz, M., Smalla, K.H., Richter, K., Landgraf, P., et al. (2008). Caldendrin-Jacob: a protein liaison that couples NMDA receptor signalling to the nucleus. PLoS Biol. 6, e34.

Gardiol, A., Racca, C., and Triller, A. (1999). Dendritic and postsynaptic protein synthetic machinery. J. Neurosci. 19, 168-179.

Govindarajan, A., Israely, I., Huang, S.Y., and Tonegawa, S. (2011). The dendritic branch is the preferred integrative unit for protein synthesis-dependent LTP. Neuron 69, 132-146.

Graham, T.R., and Burd, C.G. (2011). Coordination of Golgi functions by phos-

phatidylinositol 4-kinases. Trends Cell Biol. 21, 113-121.

Hanus, C., and Ehlers, M.D. (2008). Secretory outpostsforthe local processing

of membrane cargo in neuronal dendrites. Traffic 9, 1437-1445.

Hanus, C., and Schuman, E.M. (2013). Proteostasis in complex dendrites. Nat.

Rev. Neurosci. 14, 638-648.

Hanus, C., Kochen, L., Tom Dieck, S., Racine, V., Sibarita, J.B., Schuman, E.M., and Ehlers, M.D. (2014). Synaptic control of secretory trafficking in dendrites. Cell Rep. 7, 1771-1778.

Horton, A.C., and Ehlers, M.D. (2003). Dual modes of endoplasmic reticulum-to-Golgi transport in dendrites revealed by live-cell imaging. J. Neurosci. 23, 6188-6199.

Horton, A.C., Racz, B., Monson, E.E., Lin, A.L., Weinberg, R.J., and Ehlers, M.D. (2005). Polarized secretory trafficking directs cargo for asymmetric dendrite growth and morphogenesis. Neuron 48, 757-771. Hradsky, J., Raghuram, V., Reddy, P.P., Navarro, G., Hupe, M., Casado, V., McCormick, P.J., Sharma, Y., Kreutz, M.R., and Mikhaylova, M. (2011). Post-translational membrane insertion of tail-anchored transmembrane EF-hand Ca2+ sensor calneurons requires the TRC40/Asna1 protein chaperone. J. Biol. Chem. 286, 36762-36776.

Jeyifous, O., Waites, C.L., Specht, C.G., Fujisawa, S., Schubert, M., Lin, E.I., Marshall, J., Aoki, C., de Silva, T., Montgomery, J.M., et al. (2009). SAP97 and CASK mediate sorting of NMDA receptors through a previously unknown secretory pathway. Nat. Neurosci. 12, 1011-1019.

Kapitein, L.C., Schlager, M.A., Kuijpers, M., Wulf, P.S., van Spronsen, M., MacKintosh, F.C., and Hoogenraad, C.C. (2010). Mixed microtubules steer dynein-driven cargo transport into dendrites. Curr. Biol. 20, 290-299. Karpova, A., Mikhaylova, M., Bera, S., Bar, J., Reddy, P.P., Behnisch, T., Ran-kovic, V., Spilker, C., Bethge, P., Sahin, J., et al. (2013). Encoding and trans-

ducing the synaptic or extrasynaptic origin of NMDA receptor signals to the nucleus. Cell 152, 1119-1133.

Liu, S., and Storrie, B. (2012). Are Rab proteins the link between Golgi organization and membrane trafficking? Cell. Mol. Life Sci. 69, 4093-4106. Lukinavicius, G., Umezawa, K., Olivier, N., Honigmann, A., Yang, G., Plass, T., Mueller, V., Reymond, L., Correa, I.R., Jr., Luo, Z.G., et al. (2013). A near-infrared fluorophore for live-cell super-resolution microscopy of cellular proteins. Nat. Chem. 5, 132-139.

Mikhaylova, M., Sharma, Y., Reissner, C., Nagel, F., Aravind, P., Rajini, B., Smalla, K.H., Gundelfinger, E.D., and Kreutz, M.R. (2006). Neuronal Ca2+ signaling via caldendrin and calneurons. Biochim. Biophys. Acta 1763, 1229-1237.

Mikhaylova, M., Reddy, P.P., Munsch, T., Landgraf, P., Suman, S.K., Smalla, K.H., Gundelfinger, E.D., Sharma, Y., and Kreutz, M.R. (2009). Calneurons provide a calcium threshold for trans-Golgi network to plasma membrane trafficking. Proc. Natl. Acad. Sci. USA 106, 9093-9098. Ori-McKenney, K.M., Jan, L.Y., and Jan, Y.N. (2012). Golgi outposts shape dendrite morphology by functioning as sites of acentrosomal microtubule nucleation in neurons. Neuron 76, 921-930.

Pierce, J.P., Mayer, T., and McCarthy, J.B. (2001). Evidence for a satellite secretory pathway in neuronal dendritic spines. Curr. Biol. 11, 351-355. Quassollo, G., Wojnacki, J., Salas, D.A., Gastaldi, L., Marzolo, M.P., Conde, C., Bisbal, M., Couve, A., and Caceres, A. (2015). A RhoA signaling pathway regulates dendritic Golgi outpost formation. Curr. Biol. 25, 971-982. Ramirez, O.A., and Couve, A. (2011). The endoplasmic reticulum and protein trafficking in dendrites and axons. Trends Cell Biol. 21, 219-227. Rutishauser, U. (2008). Polysialic acid in the plasticity of the developing and adult vertebrate nervous system. Nat. Rev. Neurosci. 9, 26-35. Senkov, O., Tikhobrazova, O., and Dityatev, A. (2012). PSA-NCAM: synaptic functions mediated by its interactions with proteoglycans and glutamate receptors. Int. J. Biochem. Cell Biol. 44, 591-595.

Sun, L.P., Seemann, J., Goldstein, J.L., and Brown, M.S. (2007). Sterol-regu-lated transport of SREBPs from endoplasmic reticulum to Golgi: Insig renders sorting signal in Scap inaccessible to COPII proteins. Proc. Natl. Acad. Sci. USA 104, 6519-6526.

Thayer, D.A., Jan, Y.N., and Jan, L.Y. (2013). Increased neuronal activity fragments the Golgi complex. Proc. Natl. Acad. Sci. USA 110, 1482-1487. Torre, E.R., and Steward, O. (1996). Protein synthesis within dendrites: glyco-sylation of newly synthesized proteins in dendrites of hippocampal neurons in culture. J. Neurosci. 16, 5967-5978.

Welz, T., Wellbourne-Wood, J., and Kerkhoff, E. (2014). Orchestration of cell surface proteins by Rab11. Trends Cell Biol. 24, 407-415. Yamaguchi, N., and Fukuda, M.N. (1995). Golgi retention mechanism of beta-1,4-galactosyltransferase. Membrane-spanning domain-dependent homodi-merization and association with alpha- and beta-tubulins. J. Biol. Chem. 270, 12170-12176.

Ye, B., Zhang, Y., Song, W., Younger, S.H., Jan, L.Y., and Jan, Y.N. (2007). Growing dendrites and axons differ in their reliance on the secretory pathway. Cell 130, 717-729.

Zhou, W., Chang, J., Wang, X., Savelieff, M.G., Zhao, Y., Ke, S., and Ye, B. (2014). GM130 is required for compartmental organization of dendritic Golgi outposts. Curr. Biol. 24, 1227-1233.