NgCAM is present in somatodendritic endosomes
Previously, intracellular organelles containing EGFP–NgCAM were found in dendrites and somata as well as axons at steady state (Burack et al., 2000
). After infection of cells with a defective adenovirus encoding NgCAM (AdNgCAM) for 24 h, we visualized internal and surface pools of NgCAM with different fluorophores using a sandwich protocol (Kamiguchi and Lemmon, 2000
). We similarly find NgCAM in intracellular somatodendritic compartments ( and B3) as well as in intracellular compartments in the axon ( and B1). The surface pool of NgCAM, on the other hand, is highly enriched on the axon (, B2, and B4). Next, we quantified the extent of cell surface polarization of NgCAM using IP software to determine a “polarity index” (PI) similar to that used by others (Cheng et al., 2002
; Sampo et al., 2003
; see Materials and methods for details). Uniform staining gives a PI value of 1, whereas axonal enrichment gives a PI value of >1, and somatodendritic enrichment gives a PI value of <1. For virally expressed NgCAM, the PI is 6.0, indicating sixfold axonal enrichment ().
Figure 1. NgCAM is found in somatodendritic endosomes positive for transferrin. (A) Hippocampal neurons were infected with AdNgCAM for 24 h, fixed, and stained with a sandwich protocol to detect intracellular (red) and cell surface (green) pools of NgCAM. The axon (more ...)
Figure 2. Quantification of axonal localization of NgCAM and NgCAM mutants. NgCAM constructs are depicted as schematics. The 114-aa cytoplasmic tail of NgCAM is numbered from the first residue of the cytoplasmic tail, Lys1. The ankyrin binding domain is indicated (more ...)
We first wanted to test whether some of these somatodendritic compartments containing NgCAM are endosomes. To this end, we again infected cells with AdNgCAM for 24 h. Subsequently, live cells were incubated with FITC–transferrin for 30 min before fixation. Transferrin is a recycling ligand that is taken up into early and recycling endosomes but is largely excluded from late endosomes and lysosomes (Mellman, 1995
). The cell surface and intracellular pools of NgCAM were differentially stained and analyzed by confocal fluorescence microscopy ( C). A subset of the intracellular vesicular profiles of NgCAM present in the somatodendritic domain colocalized with internalized transferrin ( D; yellow indicates overlap), suggesting that some of the somatodendritic compartments containing NgCAM are endosomes.
Next, we asked whether NgCAM reached these somatodendritic endosomes via the cell surface or by direct transport from the TGN. Thus, we assayed NgCAM uptake by incubating cells expressing NgCAM with antibodies to NgCAM's extracellular domain for 30 min before fixation. Cell surface and internalized antibodies were then differentially stained using the sandwich protocol with only secondary antibodies. We found that surface NgCAM was restricted to the axon (blue), while internalized anti-NgCAM antibodies (red) were detected in the somatodendritic domain in ~50% of NgCAM-expressing cells (). Uninfected control cells did not have detectable staining even after 2 h of incubation with anti-NgCAM antibodies, indicating that nonspecific uptake of antibodies was negligible (unpublished data).
Figure 3. NgCAM is internalized into a transferrin-positive somatodendritic compartment. (A) Hippocampal neurons were infected with AdNgCAM for 24 h and incubated live with both FITC–transferrin (green) and anti-NgCAM antibodies for 30 min at 37°C (more ...)
To further characterize the compartments that accumulate internalized NgCAM, we incubated neurons infected with AdNgCAM with both anti-NgCAM antibodies and FITC–transferrin for 30 min before fixation. We find that internalized NgCAM colocalizes to a significant extent (>60% of profiles) with internalized FITC–transferrin ( B; yellow indicates overlap) in the soma. Taken together, these observations suggest that NgCAM on the cell surface can reach early/recycling endosomes in the somatodendritic membrane domain.
Inhibition of endocytosis disrupts axonal localization of NgCAM
Our observations raised the possibility that NgCAM is internalized from the somatodendritic domain. To test whether endocytosis was necessary for axonal accumulation of NgCAM, we down-regulated endocytosis by expressing a dominant-negative dynamin1(K44A) (Seeger and Payne, 1992
; Damke et al., 1995
; Schmid et al., 2000
; Fourgeaud et al., 2003
). The extent of down-regulation of endocytosis by dynamin(K44A) is dependent on its expression levels, so that in cells expressing low amounts of dynamin(K44A), endocytosis will be inhibited only partially, whereas in cells expressing large amounts of dynamin(K44A), endocytosis will be inhibited more completely. If NgCAM were directly inserted into the axonal plasma membrane, it would not a priori depend on endocytosis for axonal localization and would therefore be expected to be unaffected by coexpression of dynamin(K44A). In contrast, if NgCAM had to be selectively removed from the somatodendritic surface for axonal accumulation, high levels of dynamin(K44A) expression would compromise axonal polarization.
Cells were infected with a recombinant adenovirus encoding dynamin(K44A) for several hours before either transfection with a plasmid encoding NgCAM or infection with AdNgCAM. Cells were fixed 20 h later and stained against surface NgCAM and internal dynamin(K44A). We observed a significant change in the distribution of NgCAM in cells expressing dynamin(K44A). A representative experiment is shown in A. The percentage of cells that expressed NgCAM preferentially on the axonal surface decreased dramatically, while the proportion of uniformly expressing cells increased. This observation suggests that endocytosis is a necessary step for axonal accumulation of NgCAM. Surprisingly, we also observed a novel population of cells (~20%) expressing NgCAM primarily on the somatodendritic domain, i.e., showing reversed polarity ('). Furthermore, this population was only observed in cells expressing high levels of dynamin(K44A), but never in control cells. As is the case with dominant-negative approaches in general, the extent of inhibition is dependent on the expression level of the dominant-negative protein. In accordance with this, we observe no effects of low levels of dynamin(K44A) on axonal localization of NgCAM ('). At high levels of dynamin(K44A) expression where endocytosis was most likely effectively blocked, NgCAM accumulated on the somatodendritic domain (').
Figure 4. Inhibition of endocytosis leads to loss of axonal polarization of NgCAM. (A) Neuronal cultures expressing NgCAM in the presence (gray bars; DNdynamin) or absence (black bars; control) of dominant-negative dynamin (DN-dynamin[K44A]) were (more ...)
Following the transport kinetics of NgCAM
Our observation of somatodendritically enriched NgCAM raised the possibility that NgCAM is first targeted to the somatodendritic surface, reinternalized, and then transcytosed to the axon. We therefore decided to study the kinetics of NgCAM transport from the TGN to the plasma membrane. To this end, we applied a pulse-chase approach making use of the secretion inhibitor brefeldin A (BFA) (Chardin and McCormick, 1999
). BFA is a highly specific blocker of Arf1-dependent membrane transport in both the Golgi and endosomes, and its effects are readily reversible in all cell types studied, including cultured hippocampal neurons (Craig and Banker, 1994
; Cid-Arregui et al., 1995
; Jareb and Banker, 1997
). As retrograde Golgi transport is still active in the presence of BFA, Golgi membranes and associated proteins flow back into the ER, forming a hybrid ER/Golgi compartment (Klausner et al., 1992
). Membrane proteins, including newly synthesized NgCAM, accumulate in the ER/Golgi compartment during BFA treatment and fail to reach the plasma membrane (unpublished data). As the BFA block is reversible, a wave of transport of the accumulated membrane proteins can be observed after removal of BFA (i.e., “pulse”). The progress of this pulse can be observed by fixing the cells after intervals following BFA washout (i.e., “chase”).
We infected neuronal cultures with AdNgCAM for 4 h, treated overnight with BFA to stop Golgi trafficking, washed the BFA away, and fixed the cells at various intervals thereafter. Internal and surface pools of NgCAM were differentially stained as before. Strikingly, surface expression of NgCAM was first detectable on the somatodendritic domain, i.e., with reversed polarity ( A). Cells with somatodendritic NgCAM were also observed when the primary antibody was added to live cells (unpublished data) and then fixed, rather than after fixation without permeabilization. Indeed, 4 h after washing out BFA, 30% of infected cells expressed NgCAM with reversed polarity ( B, circles), while most of the rest (67%) showed only intracellular staining ( B, diamonds, stippled line). Of the cells with detectable surface expression at 4 h of chase, 92% show somatodendritic restriction of NgCAM. This somatodendritic pool could then be quantitatively chased to the axonal plasma membrane over time ( B, squares). These observations suggest that during biosynthetic transport from the Golgi, NgCAM is first inserted into the somatodendritic plasma membrane.
Figure 5. Kinetic analysis of NgCAM transport reveals a somatodendritic transport intermediate. (A) After infection with AdNgCAM, NgCAM was accumulated in the ER/Golgi compartment by an overnight treatment with BFA. Transport of the accumulated membrane proteins (more ...)
To corroborate this unexpected finding, an alternative pulse-chase assay was used that avoided the use of BFA. After infection with AdNgCAM, membrane traffic was arrested by incubating the cells at 19°C for 12 h (Scales et al., 1997
). Subsequently, the cells were warmed to 37°C to allow membrane traffic to proceed (“pulse”), and then the cells were fixed at several time intervals after warming (“chase”). We obtained qualitatively similar results with this second approach: after arrest of membrane traffic and subsequent release, NgCAM was first detected on the cell surface in the somatodendritic domain ( C). Again, this population could be chased to the axon over time. The kinetics of axonal appearance are somewhat faster after cold block release than after BFA block release. This is likely an indication that the Golgi block is more readily reversed after cold treatment than after BFA treatment, resulting in a tighter chase of NgCAM to the surface.
We also wanted to see if we could detect somatodendritic NgCAM without using any TGN transport block. Therefore, we fixed and stained AdNgCAM-infected cells at early time points after infection (21 h), when NgCAM first appeared at the cell surface. Surface expression was low at this point, and only ~25% of infected neurons had any detectable surface expression at all. We could detect a small percentage of cells with somatodendritic enrichment of NgCAM (unpublished data). We therefore consider it unlikely that the initial somatodendritic surface appearance is an artifact caused by the use of BFA. Rather, our kinetic data suggest that NgCAM first appears at the somatodendritic surface and not at the axonal membrane domain.
Furthermore, quantitative analysis of multiple cells at short times after BFA release with IP software showed that in cells with “somatodendritic/reversed polarity” ( B, circles), >80% of the fluorescence signal is on the somatodendritic domain. The initial somatodendritic delivery therefore appears to be the predominant pathway followed by the majority of NgCAM molecules leaving the TGN and not a minor salvage pathway traveled by only a small percentage of missorted molecules. Based on these findings, we propose the following model for NgCAM delivery to the axon: (1) NgCAM travels from the TGN primarily to the somatodendritic plasma membrane, (2) the protein is internalized, and (3) NgCAM traffics from somatodendritic endosomes to the axonal plasma membrane by transcytosis.
Direct and indirect axonal pathways exist
We next sought to determine the regions of NgCAM that contain the trafficking signals necessary for transcytosis. Numerous studies have demonstrated that the cytoplasmic tail of NgCAM contains signals important for trafficking and localization of the molecule (Kamiguchi and Lemmon, 1998
; Kamiguchi et al., 1998
). Thus we generated a truncated NgCAM construct, NgCAM(CT3), lacking all but three amino acids of its cytoplasmic tail (). Neuronal cultures were simultaneously infected with AdNgCAM (CT3) and Adp75NGFR. P75NGFR is present on both axons and dendrites and serves to visualize both axonal and somatodendritic domains ( A, inset). We found that NgCAM(CT3) is axonally enriched at steady state (; A), indicating that axonal targeting information is encoded in NgCAM's ectodomain (Sampo et al., 2003
Figure 6. The cytoplasmic tail of NgCAM is required for transcytotic delivery to the axon. (A) Hippocampal neurons 8–10 DIV were infected with recombinant adenoviruses expressing NgCAM(CT3) for 30 h, fixed, and stained with mouse monoclonal antibodies against (more ...)
To analyze whether NgCAM(CT3) still reaches the axonal membrane via transcytosis, we studied the transport kinetics of NgCAM(CT3) using the BFA pulse-chase assay. In contrast to full-length NgCAM, NgCAM(CT3) was never detected on the somatodendritic surface (circles) after BFA washout ( B). We first detected surface expression after 2 h of chase. Even at these earliest times of detectable surface expression, NgCAM(CT3) was already axonally enriched on the plasma membrane ( B, squares). Therefore, NgCAM(CT3) appears to be directly delivered from the TGN to the axonal plasma membrane. Therefore, both direct and transcytotic axonal pathways exist.
The initial somatodendritic delivery of NgCAM requires the Y33RSL motif
Next, we asked which cytoplasmic tail sequences of NgCAM are required for the initial somatodendritic insertion. Previously, the Y33
RSL motif has been demonstrated to control endocytosis of NgCAM through its interaction with the clathrin adaptor complex AP-2 (Kamiguchi et al., 1998
). Interestingly, tyrosine-based endocytosis signals are sometimes colinear with tyrosine-based basolateral and/or somatodendritic sorting signals (Matter et al., 1992
To test the idea that the Y33RSL motif might be needed for initial somatodendritic targeting before endocytosis, we generated a recombinant construct, NgCAM(Y33A), in which the Y33RSL motif was disrupted by mutating tyrosine33 to alanine (). Again, neuronal cultures were simultaneously infected with Adp75NGFR to visualize all processes ( B). We found that NgCAM(Y33A) was well polarized to the axon at steady state (; ). To determine how NgCAM(Y33A) arrived at the axon, we used the BFA pulse-chase assay. Similarly to NgCAM(CT3), NgCAM(Y33A) was enriched on the axonal plasma membrane ( C, squares) even at the earliest times of detectable surface expression. No somatodendritic population was ever observed, suggesting transport via a direct pathway.
Figure 7. The YRSL motif is required for transcytosis. Neurons (8–10 DIV) were coinfected with AdNgCAM(Y33A) (A) and defective adenoviruses encoding the uniformly expressed p75/NGFR (B) for 30 h. NgCAM(Y33A) is axonally enriched (arrows mark axon, arrowheads (more ...)
Transport along a direct pathway is expected to be independent of endocytosis. We, therefore, tested whether down-regulation of endocytosis would affect the axonal localization of NgCAM(Y33A) by coexpressing dynamin (K44A) and NgCAM(Y33A). We found that axonal localization of NgCAM(Y33A) was unaffected by coexpression of dynamin(K44A) ( D), and therefore appeared independent of endocytosis, consistent with direct routing to the axon. Therefore, these data suggest that a mutation in the Y33RSL motif prevents the initial somatodendritic delivery of NgCAM and changes the axonal transport pathway from transcytotic to direct. Therefore, the Y33RSL motif might serve as a somatodendritic targeting determinant facilitating the sorting of NgCAM at the TGN.