Construction of GFP Fusion Proteins
The aim of our study was to analyze the trafficking of APP in living neurons and compare it with the trafficking of another axonal membrane protein, the synaptic vesicle protein p38. Therefore, we tagged the human APP695 at the C terminus with a spectral variant of GFP, YFP5 (Pepperkok et al., 1999
). Two other fusion proteins, consisting of p38 and EGFP (p38-EGFP) and p38 and another spectral GFP variant, ECFP (p38-ECFP), were constructed. The YFP/CFP mutants were chosen to allow simultaneous visualization of two different proteins in a single cell (Ellenberg et al., 1999
). The fusion proteins are schematized in Figure .
Figure 1 Schematic diagram of GFP fusion proteins.YFP was fused to the C terminus of human APP695 (left). EGFP and ECFP were fused to the C terminus of rat p38 (right). In both constructs the FP is in the cytoplasm. The double line represents the plasma membrane. (more ...)
FP-tagged Fusion Proteins Are Expressed as Full-Length Proteins and Sorted to Axons in Hippocampal Neurons
Before analyzing the transport of the fusion proteins by video microscopy, we ensured that the proteins were expressed as full-length proteins and that they were correctly localized. This was especially important in the case of APP-YFP, because the APP is subject to a variety of potential proteolytic cleavages. For example, extensive intracellular α- or β-secretase cleavage would generate an FP-tagged cytoplasmic tail and transmembrane domain of APP indistinguishable from intact protein by fluorescent microscopy. To check for full-length expression, hippocampal neurons were transfected with APP-YFP cDNA and analyzed by Western blot. Using a GFP-specific antibody, a prominent band with the size of the fusion protein (130 kDa) was detected in cell lysates of transfected cells (Figure A, left). As expected for a transmembrane protein, no signal was detected in the medium. Only a minor degradation product is visible in the cell lysate of transfected neurons (Figure A, arrow). To substantiate this finding, the blot was stripped and reprobed with an APP-specific antibody (Figure A, right). The antibody detected full-length APP-YFP as well as endogenous rat APP. If a significant amount of APP-YFP would be cleaved by α-secretase, one would expect more of the soluble APP fragment in the medium of transfected cells compared with control cells. This is not the case (Figure A, right). Similarly, cell lysates of neurons transfected with p38-EGFP cDNA show only the full-length fusion protein when probed with an antibody against GFP (Figure B). The minor lower band reflects unspecific cross-reaction of the antibody (Figure B, asterisk).
Figure 2 Transfected GFP fusion proteins are expressed as full-length proteins. (A) Cell lysates (C) and media (M) of 8-d-old APP-YFP–transfected (+) and mock-transfected (−) neurons were analyzed by SDS-PAGE and immunoblotting with GFP (more ...)
Because we wanted to study axonal transport, we next analyzed whether the fusion proteins would be sorted to axons after transient transfection. Therefore, neurons cultured for 6–9 d were transfected with APP-YFP and p38-EGFP cDNA, respectively, and subjected to immunofluorescence using antibodies against GFP and MAP2, a marker for dendrites. Microscopic analysis of both constructs showed that they were sorted to axons as expected (Figure ). In addition, a somatodendritic staining was observed for both constructs. This has been found for a variety of exogenously expressed axonal and apical proteins and could be due to saturation of the axonal sorting machinery (reviewed by Winckler and Mellman, 1999
). In the case of APP-YFP a somatodendritic localization of at least some of the protein could be due to transcytosis from the axon (Simons et al., 1995
; Yamazaki et al., 1995
Figure 3 GFP-labeled fusion proteins are sorted to axons. Seven- to 10-d-old neurons were transfected with APP-YFP (A and B) or p38-EGFP (C and D) cDNA, fixed after 1 d, and processed for immunofluorescence using anti-GFP antiserum (αGFP) and anti-MAP2 (more ...)
Taken together, these data suggest that APP-YFP and p38-EGFP are useful markers to study axonal transport in living hippocampal neurons.
APP-YFP Is Transported along Axons in Elongated, Fast-moving Tubules
To visualize the transport of APP-YFP in the axons of living hippocampal neurons, we transfected cells with APP-YFP cDNA and analyzed the transport at 37°C by video microscopy. Typically, 24 h after transfection, expression of exogenous proteins could be observed in 1–20% of neurons. Transfected cells that appeared healthy by phase-contrast and GFP fluorescence were imaged by video microscopy. Figure (video) shows six consecutive frames (0.9 s apart) of a time-lapse sequence collected from axons of transfected neurons. Long tubular structures were moving along the axons, mainly anterogradely (Figure , arrows) but occasionally retrogradely (Figure , arrowheads). In addition to the moving tubules, stationary or slow moving vesicles were present (Figure , asterisk). The tubular appearance of the transport carriers was not due to an artifact created by the movement of a structure during exposure time (see MATERIALS AND METHODS) or due to overexpression, because tubules were observed also in neurons expressing very low amounts of APP-YFP. The tubules were of variable length and could be as long as 10 μm. The length of a tubule was not correlated to its speed. Occasionally a tubule could be seen to stop, without losing the tubular appearance, then eventually shrinking to a vesicle before elongating again and resuming movement (our unpublished results). The tubules are not mitochondria, which also show a tubular appearance in axons (Morris and Hollenbeck, 1995
), as shown by staining with rhodamine 123 (our unpublished results). Only in very few cases did the APP-YFP tubules change direction; in many cases they transversed the field of view without stopping.
Figure 4 APP-YFP is transported in long tubules along axons (video). Nine-day-old neurons were transfected with APP-YFP cDNA and analyzed 24 h later by video microscopy. Frames were taken every 0.9 s; 6 subsequent frames of 50 frames (total length, 45 s) are shown. (more ...)
To quantitate these movements, axonal transport in a number of transfected neurons was recorded. Evaluation of 8 cells from three different experiments showed that of 156 clearly discernible structures 32% (50 particles, all vesicular) did not move or only moved over a short distance (<3 μm). Forty-eight percent (75) moved anterogradely over long distances, 97.3% (73) of which were long tubules and 2.7% (2) of which were vesicular. The remaining 20% (31) were moving retrogradely, 71% (22) of which were tubular, the others vesicular.
We next tracked all structures moving >3 μm and calculated their velocity. APP-YFP–containing transport carriers were moving anterogradely at an average velocity of 4.5 ± 1.5 μm/s with maximal velocities of up to 9 μm/s. Retrogradely APP-YFP–containing structures moved slower, on average 2.9 ± 1 μm/s. Transport carriers could be followed on average for 53 μm before leaving the field of view or plane of focus or, rarely observed, stopping.
p38-EGFP Is Transported in Different Carriers
Are other axonal membrane proteins also transported in elongated tubules? To test this, neurons were transfected with p38-EGFP cDNA and analyzed by video microscopy (Figure , video). Fluorescent vesicles could be observed along the axons of transfected cells. They showed two types of movement: bright fluorescent vesicles without movement or moving over short distances (Figure , arrowheads) and tubulovesicular structures, often less bright, moving either anterogradely or retrogradely (Figure , arrows). The overall appearance, however, was strikingly different compared with the movement of APP-YFP carriers. No tubules or only short ones could be observed, and carriers rarely moved over long distances (Figure ). The dynamics of p38-EGFP–labeled organelles are reminiscent of those of recycling endosomes recently described by Prekeris et al., (1999)
. We therefore tested to what extent the p38-EGFP–containing organelles could be labeled with an endosomal marker. To this end, we labeled p38-EGFP–transfected neurons with Rh-dextran and analyzed them by two-color video microscopy (Figure ). Moving p38-EGFP–labeled tubulovesicular structures were not costained with the endosomal marker, suggesting that they are transport vesicles rather than endosomes (Figure , arrowheads). Moving Rh-dextran–labeled organelles were often devoid of p38-EGFP staining (Figure , arrows). However, we sometimes observed moving structures containing both p38-EGFP and Rh-dextran (our unpublished results). In addition to the moving organelles, many immobile short tubulovesicular organelles containing both markers were observed (Figure , asterisks). Taken together, these data suggest that the organelles labeled by p38-EGFP are a mixture of endocytic and exocytic vesicles.
Figure 5 p38-EGFP is transported in tubulovesicular structures (video). Seven- to 8-d-old neurons were transfected with p38-EGFP cDNA and analyzed 24 h later by video microscopy. Frames were taken every 3 s; two subsequent frames of 70 (total length, 210 s) are (more ...)
Figure 6 Moving p38-EGFP–containing vesicles do not contain an endosomal marker. Eight-day-old neurons were transfected with p38-EGFP cDNA, labeled 22 h later for 1–2 h with 1 mg/ml Rh-dextran, and analyzed by two-color video microscopy as described (more ...)
To quantify the movement of p38-EGFP fluorescent transport carriers, time-lapse microscopy of transfected neurons was analyzed as for APP-YFP. From eight cells from three independent experiments, 165 clearly discernable fluorescent carriers were tracked. Fifty-six percent (92) of p38-EGFP–containing carriers were not moving or showed brownian motion. These most likely corresponded to organelles of endocytic origin. Forty-four percent (73) showed movement over >3 μm, without directional bias. The average velocity of anterogradely moving carriers was 0.9 ± 0.9 μm/s, more than four times slower than APP-YFP–containing carriers. Retrogradely moving carriers were slightly faster, on average 1.2 ± 1.2 μm/s. p38-EGFP moving structures could be tracked on average only 6.6 μm, eight times shorter than APP-YFP–containing tubules. A comparable p38-GFP has been shown to move in a similar manner in dorsal root ganglion neurons (Nakata et al., 1998
). Because all p38-EGFP–labeled organelles, endosomes and transport vesicles, move with dynamics strikingly different from those labeled with APP-YFP, these data suggest that the two proteins are transported in different carriers. A velocity profile of APP-YFP– and p38-EGFP–containing carriers is shown in Figure , and the findings are summarized in Table .
Figure 7 Velocity profile of axonal transport of APP-YFP in comparison with p38-EGFP. Neurons were transfected with APP-YFP and p38-EGFP cDNA, respectively, and analyzed by video microscopy. Moving structures were tracked, and the velocity was calculated as described (more ...)
Characteristics of FP-tagged APP and p38
In Doubly Transfected Neurons, APP-YFP and p38-ECFP Are Transported in Different Carriers
APP-YFP and p38-EGFP transport carriers move in a strikingly different manner along axons, suggesting that they are sorted to different transport carriers. To directly analyze the sorting and differential transport of the two proteins in single axons, we cotransfected neurons with both APP-YFP and p38-ECFP cDNAs. Most transfected cells expressed both exogenous proteins, however, often in differing amounts. Neurons expressing approximately similar amounts of both proteins were chosen and analyzed by two-color video microscopy (Figure ). When frames of either the YFP or CFP channel were displayed separately, they looked identical to the single transfected neurons: APP-YFP moving in long tubules (Figure A, left, video), p38-ECFP moving slower and in vesicular structures (Figure A, middle, video). Display of the merged frames shows the different movement of the two proteins along the same axon (Figure A, right, video). The long tubules transporting APP-YFP did not contain p38-ECFP (Figure B, arrows). Conversely, moving p38-ECFP–containing structures did not appear to have APP-YFP (Figure B, triangle). Occasionally, slow or nonmoving structures containing both proteins were seen, which are possibly endosomes or multivesicular bodies (Figure B, arrowheads), because slow-moving, nontubular APP-YFP carriers can be labeled with Rh-transferrin (our unpublished results).
Figure 8 Two-color video microscopy of doubly transfected neurons shows little colocalization of APP-YFP and p38-ECFP (video). Seven- to 8-d-old neurons were doubly transfected with APP-YFP and p38-ECFP cDNA, and two-color video microscopy was performed as described (more ...)
Taken together these data suggest that FP-tagged APP and p38 are sorted to and transported in carriers that differ in their morphology, their velocity, and their displacement. This implies that these carriers have a different molecular composition and are transported by the action of different motor proteins.
APP-YFP Transport Is Kinesin Dependent
Treatment of hippocampal neurons with antisense oligonucleotides against KHC and subsequent immunofluorescence suggested a role for kinesin in the fast axonal transport of APP (Ferreira et al., 1993
; Yamazaki et al., 1995
). To test the effect of kinesin on the transport of the APP-YFP tubules described here, we treated transfected neurons with KHC antisense or sense oligonucleotides. The transport of APP-YFP was then analyzed by video microscopy (Figure , video). In transfected neurons treated with sense oligonucleotides, the previously observed pattern of movement was observed: tubules moving over long distances and almost never changing their direction during recording of 30–70 frames (Figure , left, the track display of one structure is shown). In contrast, APP-YFP tubules showed a different type of movement when neurons were treated with antisense oligonucleotides: The shape of the tubules was unchanged, but their velocity was reduced. In addition, many of the observed moving structures changed their direction of movement several times (exemplified in Figure , right, where the tubule changed direction four times). Quantitation of 28 cells from four experiments showed that the velocity of APP-YFP tubules in antisense-treated neurons was on average reduced to 60% of the control velocity, and that 19 of 66 (29%) of the observed tubules changed their direction at least once, compared with 3% in control neurons. Interestingly, both anterograde and retrograde transport were affected, similar to observations in squid axons (Brady et al., 1990
). These data demonstrate that the transport of APP-YFP tubules is microtubule and kinesin dependent.
Figure 9 KHC antisense treatment interferes with axonal transport of APP-YFP (video). Seven- to 8-d-old neurons were transfected with APP-YFP and incubated overnight with sense (s) and antisense (as) oligonucleotides corresponding to KHC. Video microscopy was (more ...)