Kinesin dependent, rapid, bi-directional transport of ER sub-compartment in dendrites of hippocampal neurons

Endoplasmic reticulum (ER) is an organelle that is responsible for the storage and release of Ca²⁺ in eukaryotic cells. In the neuron, Ca²⁺ regulation by ER plays various roles in neuronal activities, e.g. modification of synaptic vesicle release (Smith and Cunnane, 1996), or regulation of nerve growth (Takei et al., 1998) in the axon. Ca²⁺ release from ER has also been reported to be involved in the induction of long-term potentiation (LTP) and long-term depression (LTD) (Harvey and Collingridge, 1992; Kohda et al., 1995; Reyes and Stanton, 1996). In particular, inositol 1,4,5-trisphosphate [Ins(1,4,5)P₃]-induced Ca²⁺ release (IICR) from intracellular calcium stores through Ins(1,4,5)P₃ receptors [Ins(1,4,5)P₃Rs] (Berridge, 1993) has been shown to be involved in depotentiation, suppression and input specificity of LTP, and in the induction of LTD in hippocampal neurons (Fujii et al., 2000; Nishiyama et al., 2000). IICR has also been shown to be required for the induction of LTD in the cerebellar Purkinje cells (Inoue et al., 1998).

ER in the neuron is composed of a continuous network of membrane expanding throughout the cell, up into the head of the dendritic spines and down into the axon (Terasaki et al., 1994; Spacek and Harris, 1997; Berridge, 1998; Petersen et al., 2001). In spite of that, it is reported that Ca²⁺ signal in the neuron is spatially restricted (Berridge, 1998; Finch and Augustine, 1998; Takechi et al., 1998; Nakamura et al., 2002), and it is postulated that induction of synaptic plasticity requires this spatially restricted Ca²⁺ release from ER (Miyata et al., 2000; Wang et al., 2000). Localization of ER membrane and targeting of ER resident Ca²⁺-release channels, Ca²⁺ pumps and Ca²⁺ binding proteins to specific regions in the cell should be an important factor in the spatial regulation of Ca²⁺ release. Previous morphological and biochemical studies showing heterogeneous distribution of Ca²⁺ channels on ER and major ER resident proteins in Purkinje cells (Villa et al., 1991; Volpe et al., 1991; Takei et al., 1992) support this idea. However, the distribution of ER membrane in living neurons is little understood.

In addition, ER in the eukaryotic cells is not a static organelle but it is highly dynamic; ER membrane tubules are shown to be highly dynamic in both in neurons and nonneuronal cells (Lee and Chen, 1988; Dailey and Bridgman, 1989; Sanger et al., 1989; Waterman-Storer and Salmon, 1998), and proteins inserted in the ER membrane are also reported to diffuse freely and rapidly on it in COS-7 cells (Nehls et al., 2000). Because neurons are highly polarized cells, transport of newly synthesized proteins and membrane to the distal region should be crucial for the maintenance of ER functions. In neurons, dynamics of ER in the axon has been well studied. Earlier morphological studies utilizing local cold block showed that smooth ER is transported anterogradely by a distinct mechanism from the fast axonal transport system (Tsukita and Ishikawa, 1980; Ellisman and Lindsey, 1983). Dynamics of ER within the growth cone was also examined and suggested to be microtubule dependent (Dailey and Bridgman, 1989). More recently, we demonstrated that a green fluorescent protein (GFP)-tagged transmembrane region of Ins(1,4,5)P₃R had bi-directional, temperature-sensitive and microtubule-dependent movement that was slower than fast axonal transport in the axon of cultured chick dorsal root ganglion neurons (Aihara et al., 2001). However, the dynamics of ER membrane and of ER resident proteins such as Ca²⁺ channels and Ca²⁺ pumps in the neuron of the central nervous system (CNS) remain to be elucidated especially in the dendrite.

In the present study, we examined the distribution and dynamics of ER labeled with fluorescent proteins in the dendrite of cultured hippocampal neurons. We report here that a sub-compartment of ER exists in a large, vesicular structure that showed rapid, bi-directional movements along the dendrite. This vesicular sub-compartment of ER in living neurons shared an important function of the ER, that is, Ca²⁺ release and uptake. Finally, we show that the movements of the vesicular ER sub-compartment were highly dependent on the integrity of kinesin and microtubules.

The construction of GFP-tagged sarcoplasmic/endoplasmic reticulum calcium-ATPase 2a (GFP-SERCA2a) (Aihara et al., 2001), and GFP-tagged Ins(1,4,5)P₃R type1 (GFP-Ins(1,4,5)P₃R1) (Zhang et al., 2003) have been described previously. To construct ER luminal marker proteins, GFP-KDEL and RFP-KDEL, we fused the ER targeting sequence that corresponds to the N-terminal 17 amino acids of mouse calreticulin to the N terminus of EGFP or DsRed1 (Clontech, Palo Alto, USA), and inserted an ER retention signal peptide, KDEL, to its C terminus by a two stage PCR strategy (Terasaki et al., 1996; Roderick et al., 1997). The resultant PCR products were subcloned into pcDNA3.1/Zeo+ (Invitrogen, Tokyo, Japan). To construct a dominant negative kinesin fused with GFP (GFP-DNKHC), the tail region (859 aa - end) of the kinesin heavy chain was amplified by PCR using a cDNA clone coding for the human kinesin heavy chain xKHC (KIAA0531) as a template. KIAA0531 was kindly provided by Kazusa DNA Research Institute (Chiba, Japan). The PCR product was subcloned into the EcoRI/BamHI site of pEGFP-C1 (Clontech). GFP-LC3, a marker for autophagosome (Kabeya et al., 2000), was kindly provided by Drs Mizushima and Ohsumi (National Institute for Basic Biology, Okazaki, Japan). GFP-MST, C-terminal 597 amino acids fused with GFP, was kindly provided by Dr Gelfand (University of Illinois, IL, USA). All the genes were under the control of a CMV promoter.

The primary culture of neurons was prepared from hippocampi of 1-day-old ICR mice, as described previously (Higgins and Banker, 1998). In brief, the hippocampi were treated with 0.5% trypsin for 5 minutes at 37°C, then the cells were dissociated in Hanks' balanced salt solution (HBSS) containing 0.05% DNase I (Invitrogen) and 12 mM MgSO₄. The dissociated cells were plated on poly-L-lysine (Nacalai Tesque, Kyoto, Japan) or laminin (Koken, Tokyo, Japan)-coated coverslips at a density of 4.6×10⁴ to1.3×10⁵ cells/cm² and cultured in Neurobasal Medium (Invitrogen) supplemented with 2.5 mM L-glutamine (Nacalai Tesque), 2.5% (v/v) B-27 (Invitrogen), 1.25% (v/v) insulin-transferrin-selenium-A (Invitrogen) and antibiotics (250 units/ml penicillin and 250 μg/ml streptomycin). The cultures were transfected with 10 μg of DNAs, usually on days 4-5 in vitro, by a standard calcium phosphate method (Köhrmann et al., 1999; Kaether et al., 2000). The transfected cells were used for immunohistochemistry or imaging experiments 2-3 days after the transfection, that is, at days 7 or 8 in vitro.

For labeling hippocampal neurons with endosomal and lysosomal markers, cells were incubated with 1 mg/ml Texas Red-dextran or fluorescein-dextran (M_r 3×10³; Molecular Probes, Eugene, USA) overnight (Nakata et al., 1998). To visualize endocytotic vesicles and late endosomes more clearly, other batches of cells were incubated with these tracers for 20-30 minutes, which resulted in no obvious difference from cells treated overnight (data not shown). Mitochondrial tracers, MitoTracker Red CMXRos (Molecular Probes) and MitoTracker Green FM (Molecular Probes) were used at 20 nM and 5 nM for 10 and 30 minutes at 37°C, respectively. An ER tracer, ER-Tracker Blue-White DPX (Molecular Probes) was used at 0.1 μM for 20-30 minutes at 37°C

For immunostaining, cells were sequencially fixed with 4% paraformaldehyde in phosphate-buffered saline (PBS) for 10 minutes then with 100% methanol on ice. After permeabilization with 0.1% Triton X-100 in PBS for 10 minutes and blocking with 5% skim milk in PBS, the cells were incubated with anti-BiP (1:400 dilution; Affinity Bioreagents, Golden, USA), anti-calnexin (1:400 dilution; Stressgen Biotechnologies, Victoria, Canada), anti-calreticulin (1:400 dilution; Affinity Bioreagents), or 2.5 mg/ml of 18A10, a monoclonal antibody against Ins(1,4,5)P₃R1 (Maeda et al., 1988). Alexa488 or Alexa594-conjugated IgGs (Molecular Probes) were used as second antibodies.

Fluorescence images of neurons were taken under a confocal scanning microscope (FV-300, Olympus, Tokyo, Japan) attached to an inverted microscope (IX70, Olympus) with a 60× objective (NA 1.4 or 1.45, Olympus).

All of the images taken by confocal microscopy were processed and presented after digitally smoothing in order to reduce the noise level. The smoothing filter is implemented using 3×3 spatial convolutions, where the value of each pixel in the selection is replaced with the weighted average of its 3×3 neighborhood. Center pixels are four-fold weighted relative to the surrounding pixels.

For the time-lapse imaging experiments, the culture medium was supplemented with 20 mM Hepes (pH 7.3). The cells were visualized under an inverted microscope (IX70, Olympus) and a 60× objective (NA 1.4, Olympus) using standard filter sets and a Hg lamp.

The temperature was maintained at ∼37°C by a heating chamber surrounding the microscope stage. Sequential images were acquired with a cooled CCD camera (ORCA-ER, Hamamatsu Photonics, Hamamatsu, Japan). Images were taken for 58 mseconds or 117 mseconds every 1.5 seconds.

Photobleaching experiments were performed using the same equipment as above with an additional 488 nm Ar laser. Fluorescence of a part of the dendrite was bleached out by continuous laser illumination of 0.88 mW concentrated to the center of the field of view (20 μm in diameter). During the photobleaching, images were taken every 50 mseconds under continuous illumination.

Data were analyzed using TI Workbench, a custom-made software written by TI. Positions of fluorescent vesicles were plotted against time. The velocity of the vesicular movements was obtained by linear fitting of the slope. Only structures moving in more than three image frames were taken into account. The velocities were calculated for each period of consecutive movements.

ER luminal Ca²⁺ imaging was performed as described previously (Hirose et al., 1998; Fujiwara et al., 2001), with some modifications. Cells were loaded with 40 μM mag-fura-2 AM (Molecular Probes) or 20 μM fluo-5N AM (Molecular Probes) for 60 minutes in the culture medium at 37°C. The neurons were then washed in culture medium and incubated for a further 30 minutes to allow complete de-esterification of the intracellular AM esters. After being washed in a physiological salt solution (mM): 150 NaCl, 4 KCl, 2 CaCl₂, 1 MgCl₂, 5 Hepes, 5.6 glucose (pH 7.4), the loaded neurons were permeabilized with 60 μM β-escin (Sigma, St. Louis, USA) for 2 minutes in a Ca²⁺-free internal solution (mM): 1 EGTA, 3.3 ATP, 20 Pipes, 20 NaN₃, 112 potassium methanesulfonate (pH 7.0). The cells were subsequently washed with an internal solution containing ∼300 nM Ca²⁺, which was prepared by mixing CaEGTA and EGTA solutions (Hirose et al., 1998).

Ca²⁺ imaging was carried out at room temperature. Neurons were superfused with the internal solution containing ∼300 nM Ca²⁺, and 1 mM MgATP (Sigma) and 20 μM Ins(1,4,5)P₃ (Dojindo Laboratories, Kumamoto, Japan) were added as indicated. Neurons were visualized under an inverted microscope (IX70, Olympus) equipped with a cooled CCD camera (ORCA-ER) and a filter exchanger (OSP-EXA, Olympus). Fluorescence images of neurons were captured every 2 seconds at excitation wavelength of 340 nm and 380 nm (for mag-fura-2) or 480 nm (for fluo-5N), and the ratios (F₃₄₀/F₃₈₀ for mag-fura-2 and F/F₀ for fluo-5N) of the fluorescence intensities were calculated after subtraction of the background fluorescence and correction for bleaching. The time-course of changes in ratio was further filtered to reduce noise using Igor Pro software (Wavemetrics, Lake Oswego, USA).

Stock solutions of nocodazole (10 mg/ml, Sigma), cytochalasin D (10 mg/ml, Sigma) and latrunculin A (1 mg/ml, Molecular Probes) were prepared in dimethyl sulphoxide (DMSO), and stored at -20°C. To confirm that the cytoskeleton was disrupted, fixed cells were stained with anti-tubulin antibody (Lab Vision, Fremont, USA) or rhodamine-phalloidin (Molecular Probes). We found no changes in the microtubules or actin structure from control cells exposed to 0.1% DMSO (data not shown).

The antisense nucleotide treatment was conducted as described previously (Ferreira et al., 1992; Feiguin et al., 1994; Severt et al., 1999; Kaether et al., 2000). Twenty-four hours after transfection of RFP-KDEL, the neurons were incubated with the oligonucleotides at 5 μM. Twelve hours later, oligonucleotides were again added at 2.5 μM. Treated neurons were used for time-lapse microscopy, 24-28 hours after the first administration of oligonucleotides, that is, 48-52 hours after transfection. We used antisense-oligo DNA corresponding to 5′ flanking -11 to +14 of rat conventional kinesin heavy chain (KHC) (5′-GCCGGGTCCGCCATCTTTCTGGCAG-3′) and sense-oligo DNA (5′-CTGCCAGAAAGATGGCGGACCCGGC-3′) as control. This sequence was reported to suppress the KHC activity in rat hippocampal neurons (Ferreira et al., 1992; Feiguin et al., 1994; Severt et al., 1999; Kaether et al., 2000). Although the mouse KHC has a single displacement of nucleotide within this sequence, this oligonucleotide was also effective in mouse hippocampal neurons (see Results).

Western blotting was performed to measure the effect of antisense DNA treatment against KHC according to the standard method, using the monoclonal anti-KHC antibody (H2; Chemicon International Inc., Temecula, USA). Odyssey Infrared Imaging System (LI-COR Inc., Lincoln, USA) was used to quantify the band intensity.

To visualize ER in dendrites, we transfected plasmid DNAs encoding various ER marker proteins to cultured mouse hippocampal neurons. GFP-KDEL and RFP-KDEL are expected to be located within the ER lumen because these proteins contain an ER targeting signal and a retention signal (Terasaki et al., 1996; Roderick et al., 1997). Typical ER membrane proteins, a Ca²⁺ pump sarcoplasmic/endoplasmic reticulum calcium-ATPase 2a (SERCA2a) and a Ca²⁺-release channel Ins(1,4,5)P₃R type1 [Ins(1,4,5)P₃R1], both of which were tagged with GFP [GFP-SERCA2a and GFPIns(1,4,5)P₃R1, respectively], were also used to label ER membrane.

Neurons expressing GFP-KDEL (Fig. 1A), GFP-SERCA2a (Fig. 1B) and GFP-Ins(1,4,5)P₃R1 (Fig. 1C) were observed with a confocal microscope. Fluorescent signal of these GFP-tagged ER markers were found as dense reticular structures and some bright spots (arrowheads) in the shaft of dendrites. Immunohistochemical labeling of endogenous Ins(1,4,5)P₃R1 and an ER resident protein calnexin (Fig. 1D,E, respectively) showed a continuous structure and spots (arrowheads). The expression patterns of these ER markers in living neurons were quite similar to those of endogenous Ins(1,4,5)P₃R1 and calnexin, that is, they were expressed mainly on dense reticular structures and some bright spots. These results indicate that the pattern of distribution of GFP-tagged ER markers reliably reflects those of the endogenous ER resident proteins in neurons. We also confirmed the co-localization of GFP-KDEL, GFP-SERCA2a, GFP-Ins(1,4,5)P₃R1 and calnexin (Fig. S1).

Time-lapse microscopy with a charge coupled device (CCD) camera revealed that some of the spots labeled with the GFP-tagged ER markers traveled rapidly within the dendritic shaft (GFP-KDEL, Fig. 2A; GFP-SERCA2a, Fig. 2B; and GFPIns(1,4,5)P₃R1, Fig. 2C). Representative movements of the spots labeled with GFP-tagged ER markers are presented in Fig. 2D-F. Spots labeled with GFP-KDEL (D), GFP-SERCA2a (E), and GFP-Ins(1,4,5)P₃R (F) moved along the dendrite in both anterograde (to the periphery) and retrograde (to the soma) directions. Some of them changed direction during the movements. Details of the vesicle movements will be described later. Because these spots moved over several micrometers in the dendrites independently of the dense reticular structures, at least a proportion of the bright spots were considered to be not merely patchy structures as a continuation of the reticular ER membrane, but to be independent sub-compartments of ER which exist for a considerable period of time without interacting with reticular ER.

A photobleaching experiment enabled us to visualize the movement of vesicles more clearly and strengthened our idea that the vesicles are separated from the reticular ER. Fluorescence of a portion of a dendrite expressing GFPSERCA2a (Fig. 2G; see also Movie 1) or GFP-Ins(1,4,5)P₃R1 (Fig. 2H; see also Movie 2) was photobleached by continuous illumination with laser light. Almost all the fluorescence of the reticular ER within the illuminated area were photobleached within 20 seconds and never recovered during the illumination. In contrast, vesicles labeled with GFPSERCA2a or GFP-Ins(1,4,5)P₃R1 rapidly moved into the photobleached region from outside of the photobleached area while still under continuous laser illumination, kept moving for several seconds in the same fashion as those in neurons without such illumination (e.g. Fig. 2B,C), and then disappeared. Because the movements of the fluorescent vesicles were distinct from the other GFP signals on the membrane of the bulk ER, which showed no photo-recovery during the recording period, these vesicles are considered to be specialized compartments of ER that are partitioned from the bulk ER.

The size of the moving vesicles labeled with GFP-KDEL proteins, estimated by differential interference contrast (DIC) images, was relatively large, that is, 0.4-0.8 μm in diameter (0.50±0.1 μm, mean ± s.d., n=44).

Previously, we also demonstrated the movements of vesicles labeled with GFP-tagged transmembrane region of Ins(1,4,5)P₃R in the axons of peripheral neurons (Aihara et al., 2001). Because the direction of vesicle movement observed in that study was solely retrograde, we speculated that these vesicles were transported as a part of the endosomal-lysosomal system probably to be degraded in the cell body. To determine whether the moving vesicles observed in the dendrite were also transported as a part of the endosomal-lysosomal system, we labeled endosomes and lysosomes of neurons with Texas Red-dextran. GFP-KDEL (Fig. 3A), GFP-SERCA2a (Fig. 3B) and GFP-Ins(1,4,5)P₃R1 (Fig. 3C) signals did not overlap with those of Texas Red-dextran both in the cell body and dendrites. We also labeled neurons with MitoTracker Red CMXRos, a mitochondrial marker. The ER marker proteins were located close to mitochondria in hippocampal neurons, as was reported in several cell types (Rizzuto et al., 1998) (for a review, see Rizzuto et al., 2000), but they did not overlap each other (Fig. 3D-F). From these results, we concluded that the vesicular structures labeled by the GFP-tagged ER markers were not endosomes, lysosomes, or mitochondria.

Interestingly, the RFP-KDEL protein, red fluorescent protein (DsRed1) fused with ER targeting and retention signals, which was designed to be localized inside the ER lumen, showed a patchy staining pattern (Fig. 4A) when expressed in hippocampal neurons, although it showed reticular pattern when expressed in COS-7 cells (data not shown). Co-transfection of the cells with GFP-KDEL and RFP-KDEL revealed that the RFP-KDEL and GFP-KDEL proteins co-localized on vesicles, but only GFP-KDEL was detected on the reticular ER (Fig. 4B). RFP-KDEL was also found on the vesicles labeled with GFP-SERCA2a and GFP-Ins(1,4,5)P₃R1 but not in the reticular structure (Fig. S2).

RFP-KDEL-positive vesicles showed rapid bi-directional movements along the dendrites (Fig. 4C; see also Movie 2) in the same fashion as the vesicular ER sub-compartment labeled with GFP-tagged ER markers. Vesicles labeled with RFP-KDEL did not disappear and the movement of vesicles did not change even after a brefeldin A treatment (data not shown). Vesicles labeled with RFP-KDEL had a tendency to accumulate at the branch points of dendrites (Fig. 4A, arrowheads), and some of them seemed to form large vacuoles and were less motile than the discrete smaller vesicles. The diameter of the discrete RFPKDEL-labeled vesicles estimated from DIC images ranged from 0.4 to 1.2 μm (0.7±0.2 μm, mean ± s.d., n=75).

Most of the RFP-KDEL-labeled vesicles were also labeled with an anti-Ins(1,4,5)P₃R1 antibody, 18A10 (Fig. 4D). As was the case with GFP-tagged ER markers, neither the endosomal-lysosomal marker fluorescein-dextran (Fig. 4E), nor the mitochondrial marker MitoTracker Green FM (Fig. 4F) co-localized with the RFP-KDEL protein. A possibility that the RFP-KDEL-labeled vesicles are derived from autophagosome was ruled out by co-expression with an autophagosome-specific marker GFP-LC3 (microtubule-association protein 1 light chain 3; see also Fig. S3) (Kabeya et al., 2000). We also confirmed that the RFP-KDEL-positive vesicles contained not only Ins(1,4,5)P₃R1 but also other endogenous ER proteins, such as BiP, calreticulin and calnexin (data not shown). These vesicles were also stained by an ER-specific dye, ER-Tracker Blue-White DPX (data not shown). These results strongly indicate that the RFP-KDEL protein was specifically localized in the vesicular sub-compartment of ER. We therefore used the RFP-KDEL protein as a marker for the vesicular sub-compartment of ER in the following experiments, because its preferential expression in the vesicular sub-compartment enabled more accurate analysis of the properties of the vesicles.

As mentioned above, some of the proteins contained in the reticular ER were also expressed in the vesicular sub-compartment of ER. This led us to hypothesize that the vesicular sub-compartment has similar functions to those of the reticular ER. Thus, we tested whether the vesicular sub-compartment of ER had the ability to take up and release Ca²⁺. We used a `real-time luminal Ca²⁺ monitoring technique' (Hirose et al., 1998; Fujiwara et al., 2001). Neurons were loaded with low-affinity fluorescent Ca²⁺ indicators, mag-fura-2 AM (K_d=∼25 μM) or fluo-5N AM (K_d=∼90 μM), and the cell membrane was permeabilized with β-escin to wash out the indicators from the cytosol and to provide direct access for the externally applied drugs to the organelles. Reticular ER takes up Ca²⁺ by activation of a Ca²⁺ pump in the presence of MgATP and releases Ca²⁺ by binding of Ins(1,4,5)P₃ to the Ins(1,4,5)P₃R (Fujiwara et al., 2001).

Fig. 5A,B show the time-course of changes in the luminal Ca²⁺ concentrations monitored with mag-fura-2 and fluo-5N, respectively, in the dendrites of permeabilized neurons. The red traces represent changes in the Ca²⁺ concentration within vesicular sub-compartments of ER labeled with RFP-KDEL, corresponding to the red ellipses (region of interest, ROI) in the images, and traces of other colors represent ROIs in the adjacent areas probably covering the reticular ER. As in the reticular ER, the Ca²⁺ concentration in the RFP-KDEL-positive vesicles increased with 1 mM MgATP and decreased with 20 μM Ins(1,4,5)P₃ in neurons labeled with mag-fura-2 (in 6 vesicles in 4 cells) and with fluo-5N (in 6 vesicles in 4 cells). Here we only counted vesicles that showed several fold larger changes in absolute fluorescent intensities than those in surrounding areas to minimize the crosstalk of fluorescence from reticular ER. Without regard to this point, MgATP-induced increase and Ins(1,4,5)P₃-induced decrease in the Ca²⁺ concentration were observed in more than 80% of RFP-KDEL-positive ER vesicles (60 vesicles in 28 cells loaded with magfura-2, 37 vesicles in 17 cells loaded with fluo-5N). These results indicate that at least a certain proportion of the vesicular sub-compartment of ER serves as Ca²⁺ stores, in addition to the reticular ER. It was not possible to make a clear distinction between the vesicular sub-compartment and reticular ER in permeabilized neurons labeled with other GFP-tagged ER markers (data not shown). For the present, the labeling with RFP-KDEL is the only way to identify the vesicular sub-compartment of ER in permeabilized neurons.

ER in reticular form has been reported to move dynamically in the cytosol (Lee and Chen, 1988; Dailey and Bridgman, 1989; Sanger et al., 1989; Waterman-Storer and Salmon, 1998). The dynamics of reticular ER and the dynamics of Ins(1,4,5)P₃R on the ER membrane have also been observed in the dendritic shafts of hippocampal neurons, and analysis of their movements is currently under way in our laboratory (K. Fukatsu, H.B., S. Zhang, T.I. and K.M., unpublished observations). In this study, we focused on the movements of the vesicular sub-compartment of ER labeled with GFP or RFP-tagged ER marker proteins in the dendrite. As mentioned above, the movements of the vesicular sub-compartment of ER were rapid and bi-directional (Fig. 2, Fig. 4C, and supplemental movies). The bi-directional movements of the vesicles labeled with ER marker proteins were observed in neurons cultured for 4-21 days. Vesicles were never observed inside the spines.

We measured the velocity of moving vesicles labeled with GFP or RFP-tagged ER markers in neurons cultured for 7-8 days (Fig. 6). The average velocity of the vesicle movements was 0.2-0.3 μm/second, as shown in Table 1. The average, range and the distribution pattern of the velocity were similar for the vesicles labeled with the four ER marker proteins. Taken together with the results of the co-localization experiment (Fig. 4B; see also Fig. S2) and Ca²⁺ imaging (Fig. 5), Ins(1,4,5)P₃R1 and SERCA2a are considered to be transported on the same type of vesicles that are labeled with RFP-KDEL and GFP-KDEL.

Finally, we investigated the molecular mechanisms underlying the movement of the vesicular sub-compartment of ER. To determine whether the cytoskeleton, including microtubules and actin filaments, is involved in the vesicle movements, we tested the effects of drugs that disrupt the cytoskeleton on the movement of the vesicular sub-compartment of ER labeled by RFP-KDEL. Nocodazole (10 μg/ml) was used for microtubule disruption, and latrunculin A (1 μg/ml) or cytochalasin D (10 μg/ml) was used for actin disruption, and we confirmed that these treatments were effective in cultured mouse hippocamoal neurons by immunocytochemical staining with anti-tubulin antibody (for microtubules) and rhodamine-phalloidin (for actin filaments) (data not shown). Treatment of hippocampal neurons with nocodazole did not change the overall position of the ER, although fragmentation of ER in the reticular structures in small dendritic branches was sometimes observed; latrunculin A and cytochalasin D had no apparent disruptive effect on the structure of reticular ER. The distribution pattern and number of the vesicles labeled with RFP-KDEL were not affected by any of these drugs. In contrast to the minor effects of the drugs on the reticular ER, the movements of the vesicular sub-compartment of ER were severely affected by nocodazole treatment. The population of moving vesicles in the dendrites was significantly decreased by nocodazole treatment (P<0.02, Mann-Whitney U-test), but not by latrunculin A treatment (Fig. 7). Nocodazole treatment also decreased the velocity of vesicle movements by ∼40% in both anterograde and retrograde directions (P<0.005, Mann-Whitney U-test), while latrunculin A did not have any significant effect on the velocity of vesicles (P>0.2, Mann-Whitney U-test) (Fig. 8 and Table 2). We also used cytochalasin D to disrupt the actin cytoskeleton, and the result was much the same as that obtained with the latrunculin A treatment. Nocodazole treatment combined with latrunculin A did not result in complete inhibition of the movement as well as treatment with nocodazole alone (Fig. 7), although the velocity of the vesicle was more severely affected by the combination of the drugs than by nocodazole alone (Fig. 8 and Table 2). These results suggest that the movements of the vesicular sub-compartment of ER are highly dependent on the integrity of the microtubules, and the contribution of the actin cytoskeleton to the movement is minor, if any.

Next, we investigated a candidate motor protein that may be responsible for vesicle movements. An ER membrane protein, kinectin, was shown to be associated with the microtubule motor protein kinesin (Toyoshima et al., 1992) and to be responsible for the membrane movements in vitro (Kumar et al., 1995), suggesting that kinesin could be involved in the movements of the vesicular sub-compartment of ER. To test this possibility, we examined the effects of overexpression of a dominant-negative kinesin, that is, the tail of the kinesin heavy chain (KHC) fused with GFP (GFP-DNKHC) and treatment with an antisense oligonucleotide against KHC on the movements of the RFP-KDEL-labeled vesicles in the dendrites. Western blot analysis confirmed the decrease in KHC protein levels in antisense-treated cells: the amount of KHC protein in the antisense-treated cells was 40-50% of untreated cells and 50-70% of sense-treated cells (data not shown). Both expression of dominant negative KHC and antisense treatment against KHC did not affect the overall distribution of ER, and the labeling patterns of mitochondrial markers and endosomal/lysosomal markers (data not shown). This result is inconsistent with a previous report (Feiguin et al., 1994) showing that KHC antisense treatment induced loss of ER from the neurites, which may be explained by the difference in the level of reduction of kinesin (>80% of untreated cells in the previous study) and the age of the neurons (DIV1 in the previous study).

In neurons expressing GFP-DNKHC, the number of motile vesicles (Fig. 7) and the velocity of the vesicular movement (Fig. 8 and Table 2) were significantly reduced as compared with those in control neurons expressing GFP instead of GFPDNKHC (by ∼73% for the number of motile vesicles and ∼35% for the velocity; P<0.005, Mann-Whitney U-test). The antisense oligonucleotide treatment also significantly decreased the number of moving vesicles by ∼88% (P<0.001, Mann-Whitney U-test) (Fig. 7), and the velocity of vesicles by ∼50% (P<0.005, Mann-Whitney U-test) compared with those in the sense-treated cells (Fig. 8 and Table 2). These data strongly suggest that the movement of vesicular sub-compartment of ER is dependent on the microtubule motor protein, kinesin.

We have examined the distribution and dynamics of ER in dendrites of living hippocampal neuron using fluorescent proteins that are localized on the ER membrane (GFPSERCA2a and GFP-Ins(1,4,5)P₃R1) or inside the ER lumen (GFP-KDEL and RFP-KDEL).

First, we showed that a part of the ER exists as vesicular structures that moved dynamically within the dendrites (Figs 1, 2 and 4), in addition to the well-known dense and reticular structure. Because the signal of the ER marker GFP-KDEL, which is expected to diffuse throughout the continuous ER lumen, was found in the vesicular as well as the reticular forms, and because vesicles labeled by GFP-tagged ER marker proteins moved over ∼10 μm for at least 20 seconds while being independent of the reticular ER, we consider that at least some proportion of the vesicles are spatially separate from the reticular ER for a considerable period of time. The photobleaching experiment showing that the movements of fluorescent vesicles were quite different from the other GFP signals on the membrane of the bulk ER also supports this idea (Fig. 2G,H; see also Movies 1 and 2). Because the fast moving vesicles labeled with ER proteins ran in the photobleached area in dendrites and kept running in it for at least 15 seconds, the lifetime of the vesicular sub-compartment of ER, which is separate from bulk ER, may be at least ∼15 seconds.

Our observations strongly suggest that these vesicles are specialized sub-compartments of ER that are separated from the bulk ER. However, it is still unclear whether the vesicular sub-compartment is completely separate from the reticular ER during its movement. Previous studies using electron microscopy clearly showed the existence of vesicles carrying endogenous ER proteins such as calsequestrin, SERCA and Ins(1,4,5)P₃R, which are entirely separate from the ER cisternae in the neuronal dendrites (Villa et al., 1991; Volpe et al., 1991), and the existence of tubulovesicular smooth ER in the axons (Tsukita and Ishikawa, 1980; Metuzals et al., 1997; Tabb et al., 1998). These observations suggest that there are vesicles carrying ER-specific proteins that are completely separated from the rest of the ER. Morphological studies may be required to clarify whether the vesicular sub-compartment of ER, revealed in this study, corresponds to those previously observed structures or not. Whether the vesicular ER is finally absorbed by the bulk ER by fusion or whether it is totally separate from the bulk ER also remains to be elucidated. We also showed that this vesicular sub-compartment of ER conveys some of the endogenous ER proteins (Fig. 4), but were distinct from other membranous organelle such as endosomes, lysosomes, mitochondria and autophagosomes (Figs 3, 4; see also Fig. S3). An important finding demonstrated in this study is that the vesicular sub-compartment of ER had the ability to function as a Ca²⁺ store, just like the reticular ER (Fig. 5). Although it is a widely accepted idea that the Ca²⁺ storage site in the neuron is a continuous structure (Berridge, 1998; Petersen et al., 2001), the finding that the vesicular sub-compartment of ER could serve as a Ca²⁺ storage site suggests the existence of another category of Ca²⁺ store in the neuron.

The uneven distribution of ER markers, that is, RFP-KDEL resided only in the vesicular structure while other GFP-tagged ER markers resided in both the vesicular and reticular structures, may reflect the fact that the ER-resident proteins often distribute unevenly in the ER structures. One of the well known example is the uneven distribution of calsequestrin in ER: calsequestrin is condensed into the vacuolar structure because of its tendency to form oligomers (Wuytack et al., 1987; Villa et al., 1991; Volpe et al., 1991; Villa et al., 1993; Gatti et al., 2001). RFP (DsRed1) also has a strong tendency to form oligomers (Baird et al., 2000). The expression pattern of DsRed2-KDEL (Zhang et al., 2003) that is expected to form less aggregates than RFP (DsRed1)-KDEL was intermediate to those of GFP-KDEL and RFP-KDEL: DsRed2-KDEL labeled both reticular ER and the vesicular sub-compartment, and the latter was labeled much more intensely by DsRed2-KDEL than by GFP-KDEL (unpublished observation). These results suggest that the vesicular structure that RFP-KDEL highlights is not a totally different structure from ER but overlaps with the well-known vesicular structure stained by several ER markers (Villa et al., 1991; Volpe et al., 1991). RFP-KDEL might be sorted exclusively to the vesicular ER fraction by similar mechanisms to those for calsequestrin.

In the present study, we demonstrated for the first time that the vesicular sub-compartment of ER was rapidly transported within dendrites. The movement of the vesicles was bidirectional, with the direction of movement altering occasionally, at an average velocity of 0.2-0.3 μm/second (Fig. 6 and Table 1). The movements were quite similar to those of the `large phase-dense organelles (>∼0.4 μm)' that were previously observed by video-enhanced phase contrast microscopy (Overly et al., 1996).

The movements of reticular ER have been hypothesized to depend on three mechanisms, namely, (1) sliding movement of ER along microtubules, (2) movement of microtubules to which ER is attached, and (3) movement dependent on the polymerization of the microtubules to which ER is attached (Waterman-Storer and Salmon, 1998). In contrast, the movements of vesicular sub-compartment of ER seemed to depend mainly on a single mechanism, that is, a sliding movement along the microtubules induced by, at least, a plus-end directed motor protein, kinesin (Figs 7, 8 and Table 2). Although we cannot exclude the possibility that other motor proteins such as dynein may also be involved in the movement, the bi-directional movement of the vesicles by kinesin could be explained by the mixed polarity of the microtubules in the dendrites of hippocampal neurons (Baas et al., 1988). It is interesting that the movement of mRNA in the dendrite of neurons, which has been suggested to be kinesin-dependent (Severt et al., 1999), is similar to the movements of the ER sub-compartment in several respects; both movements are bi-directional and the velocity of movement of mRNA granules [∼0.1 μm/second on average (Knowles et al., 1996; Rook et al., 2000)] is similar to that of the ER sub-compartment (0.2-0.3 μm/second). In the dendrite, AMPA-type glutamate receptors are also transported by kinesin (Kim and Lisman, 2001; Setou et al., 2002). How can various cargoes such as mRNA, AMPA receptors and ER sub-compartment be targeted to their precise positions in the dendrite by a single class of motor protein, kinesin? The molecular mechanisms underlying the kinesin-dependent transport processes in the dendrite need further studies, such as has been well elucidated for the axonal transport processes (for a review, see Almenar-Queralt and Goldstein, 2001).

In the present study, actin filaments were shown to have little contribution to the transport of vesicular sub-compartment of ER in the dendritic shaft (Figs 7, 8, and Table 2). However, myosin V, one of the actin motors, has been reported to be involved in the movement of vesicular ER in the squid giant axon (Tabb et al., 1998). This discrepancy may have arisen not only from the difference of the cell types but also from the difference of the cellular compartments, i.e., axon or dendrite. The axon and the dendrite have diverse properties. Although the GFP-tagged ER markers labeled vesicular structures both in the axon (Aihara et al., 2001) and the dendrite in this study, the vesicles labeled with GFP-tagged ER markers moved only retrogradely in the axon (Aihara et al., 2001), while the vesicles moved in both directions in the dendrite in this study. Furthermore, the movement of the vesicles was not sensitive to nocodazole in the axon (Aihara et al., 2001) while it was sensitive in the dendrite. Thus, the mechanisms underlying the movement and other properties of the vesicles may be different in the axon and the dendrite. Although the previous work (Aihara et al., 2001) did not show involvement of actin filaments in vesicle movement, it is tempting to speculate that the movement of the vesicular sub-compartment of the ER in axons might be largely dependent on actin filaments while those in dendrites are not.

The role of actin filaments and myosin V in the transport of the vesicular sub-compartment of ER in dendrites should be further examined, because actin-dependent movements of ER vesicles could be crucial in actin-rich domains, such as around the subcortical regions or the dendritic spines (Dekker-Ohno et al., 1996). For the present, our attempt to evaluate the contribution of myosin V to the transport of the vesicular sub-compartment of ER using GFP-tagged dominant negative myosin V (GFP-MST) has not been successful, because mouse hippocampal neurons overexpressing GFP-MST were not viable (unpublished observation), although overexpression experiments of dominant negative myosin V have been reported in many cell lines (Bahadoran et al., 2003; Gross et al., 2002; Hales et al., 2002; Rodriguez and Cheney, 2002; Rudolf et al., 2003). Myosin V might play more crucial roles in cell survival in cultured neurons than in conventional cell lines. Furthermore, experiments using myosin Va knockout mice, where myosin Vb and Vc might compenstate for the absence of myosin Va, could be useful to address this issue.

Finally, we would like to speculate on the physiological roles of the moving ER vesicles. One possibility is that the vesicular sub-compartment of ER just transports newly synthesized ER proteins from the cell body to distal parts of the dendrites. Vesicles labeled with ER marker proteins traveled at an average velocity of 0.2-0.3 μm/second (Table 1). The speed of the reticular ER movement in newt lung cells was reported to be 0.07 μm/second (Waterman-Storer and Salmon, 1998), and the velocity of the ER membrane proteins [SERCA2a and the transmembrane region of Ins(1,4,5)P₃R1] examined in the axons of chick dorsal root ganglion neurons was also 0.07 μm/second (Aihara et al., 2001), implying that ER proteins on the vesicular sub-compartment can be transported 3-4 times faster than those on the reticular ER. Because neurons are highly polarized cells, transport of the vesicular sub-compartment of ER that contains ER resident proteins could greatly contribute to the urgent supply of ER proteins to the distal parts of the dendrites.

Another possibility is that the vesicular sub-compartments are targeted to special regions of the dendrites to function as active Ca²⁺ stores. The finding that the vesicular sub-compartment of ER takes up and releases Ca²⁺ supports this idea (Fig. 5). Interestingly, Ins(1,4,5)P₃R-mediated Ca²⁺ waves begin predominantly at branch points on the main apical shaft of the dendrites of rat hippocampal CA1 pyramidal cells (Nakamura et al., 2002) and the ER sub-compartment had a tendency to accumulate at branch points of dendrites in this study, even though the vesicles existed not only at branch points of main shafts but also at those of minor branches (Fig. 4A). Thus, it is tempting to speculate that the transport of the vesicular sub-compartment of ER is involved in the spatial regulation of Ca²⁺ signaling within the neuron, which is postulated to be important for synaptic plasticity (Finch and Augustine, 1998; Miyata et al., 2000; Wang et al., 2000).

In this study, we examined, for the first time, the dynamics of the ER sub-compartment in the dendrites of mouse hippocampal neurons. We found that the sub-compartment of ER is transported rapidly and bi-directionally in a microtubuleand kinesin-dependent manner. We also demonstrated that this ER sub-compartment could constitute a Ca²⁺-release pool. The functional dynamics of the vesicles is of interest, as it may be a key to understanding the spatial regulation of neuronal Ca²⁺ signaling and various other neuronal activities, such as synaptic plasticity.

We are grateful to the Kazusa DNA Research Institute (Chiba, Japan) for providing the human cDNA clone KIAA0531, N. Mizushima and Y. Ohsumi, National Institute for Basic Biology (Okazaki, Japan) for providing GFP-LC3, Dr V. I. Gelfand for providing GFP-MST clone (University of Illinois, IL, USA), Y. Inoue (University of Tokyo, Tokyo, Japan) for giving us advices on analysis of vesicle movement, and Olympus Corporation (Tokyo, Japan) for the assistance with the laser system for the photobleaching experiment. We also thank T. Nakamura (Calcium Oscillation Project, Tokyo, Japan) for valuable discussions, M. Iwai and Y. Tateishi (University of Tokyo, Tokyo, Japan) for assistance with plasmid construction, and N. Ogawa (RIKEN, Saitama, Japan) and K. Fukatsu (University of Tokyo, Tokyo, Japan) for technical assistance. This study was supported by a grant from the Ministry of Education and Science of Japan (to T.I., M.H. and K.M.).