Direct observation of microtubule and cell junction associated GFP-APC in living cells
To date, a junctional association for mammalian APC has only been observed using immunofluorescence microscopy. Immunostaining studies of APC distribution are open to criticisms concerning antibody specificity and fixation artefacts and are unable to yield data about the dynamic behaviour of the protein. Previous attempts to examine this localisation using GFP fusion proteins were hampered by the toxic side effects associated with the expression of full-length APC in mammalian cells [5]. To address this issue and to further define the subcellular distribution of APC we attempted to examine the distribution of a GFP-APC fusion protein in a variety of mammalian cell lines. In most of these transfection efficiencies were low and GFP-APC expression was poorly tolerated. However, our data indicated that COS-7 cells represented a system in which good transfection efficiencies could be achieved and expression levels suitable for imaging were tolerated without obvious short-term effects upon cell morphology or viability. However, since COS-7 cells are not epithelial in origin, we first confirmed that they expressed APC (Figure 2, panel A) and that the endogenous APC in confluent COS-7 cells was localised to the cell cortex along with the known junctional proteins cadherin (Figure 3, panels A-C), β-catenin (Figure 3, panels D-F), and actin (Figure 3, panels G-I). COS-7 cells were found to possess full-length APC protein and this endogenous APC was found to be closely associated with junctional proteins, confirming that these cells do indeed form junctions when confluent. We were therefore confident that we could use this cell line to study dynamic microtubule-associated, as well as junction-associated localisations of our GFP-APC constructs. The integrity of GFP-APC constructs used for localisation studies in COS-7 were confirmed by immunoprecipitation, SDS-PAGE and Western blotting of transfected cell extracts (Figure 2, panel B+C). The largest of these fusion proteins (Figure 2, panel B, lanes 1–3), consisting of full-length GFP-APC and minimal deletion mutants, are indistinguishable by band locations on Western blots with the gel and blotting systems used in this study because of their large size.
A variety of GFP-APC distributions could be observed in COS-7 cells using time-lapse fluorescence microscopy, all consistent with either previous immunostaining studies or known interactions of mammalian APC, suggesting that our GFP-APC imaging was a good reflection of endogenous APC behaviour. The first was an association with microtubule tips at specific regions near the cell periphery (Figure 4, panel A (arrowheads); additional file 1). This consisted of labelling of microtubule distal segments and the presence of discrete GFP-APC puncta or clusters on or just behind the microtubule tip. Cells expressing GFP-APC were co-stained with anti-α-tubulin, confirming that the GFP-APC clusters were indeed microtubule associated (Figure 4, panels B-D). GFP-APC puncta on the end of these microtubules were observed to move at speeds consistent with microtubule growth (18 ± 8.4 μm/min; n = 28 puncta from 3 cells where movement could be continuously observed for at least 12s). As previously shown by workers using Xenopus GFP-APC, GFP-APC puncta located behind the microtubule tip underwent anterograde movement towards the tip [6]. Peripheral deposition of GFP-APC puncta by shrinking microtubules could also be observed (Figure 4, panels E-H; additional file 2). These observations closely resembled both previous reports of APC distribution based upon immunostaining (e.g. [3–5] and the APC-GFP distribution and behaviours described in Xenopus cells [6]).
However, we noted in COS-7 cells that peripheral GFP-APC puncta on microtubule ends frequently underwent retrograde linear movements with average velocities of 21.6 ± 2.4 μm/min (Figure 4, panels I-L (arrowhead); additional file 3). Typically this movement occurred over distances of less than 10 μM. In some cases portions of an original puncta were deposited within the cytoplasm while the remainder continued retrograde movement on the microtubule end (Figure 4, panels M-P arrowheads; additional file 4, examples also apparent in additional file 3). We also noted that the depolymerising microtubules tipped by GFP-APC puncta could undergo pause and then re-growth with the puncta of GFP-APC still attached (Figure 4, panels Q-T arrowhead; additional file 5). Interestingly, close examination of sequences such as additional file 5 indicates that separate GFP-APC puncta decorating the distal segment of a microtubule are swept up into a single tip-associated structure as the microtubule shrinks but re-separate into a string of beads-like distribution during microtubule regrowth, before finally beginning to re-coalesce into a single structure at the tip of paused microtubules near the cell membrane. The behaviour of GFP-APC at microtubule distal tips is clearly therefore very complex. Retrograde trafficking of GFP-APC was not documented in previous studies [6, 11]. Furthermore, these movements seem likely to represent puncta associated with shrinking microtubule tips and not with growing tips looping back from the cell edge or retrograde transport of GFP-APC puncta along microtubules since the microtubule distal segment could clearly be identified by GFP-APC labelling in many cases. We therefore conclude that mammalian APC can remain associated with shrinking microtubule tips at the cell periphery, a novel observation for this protein.
These observations represented the major GFP-APC distribution in cells imaged in subconfluent cultures although other minor localisations were also seen. GFP-APC was occasionally observed at structures resembling the centrosome (Figure 5, panel A, arrowhead). This localisation was confirmed in fixed COS-7 cells expressing GFP-APC co-stained with γ-tubulin (Figure 5, panels B-D). A centrosomal localisation for APC was recently examined in detail by other investigators [12]. In less than 5% of cells a motile comet-like distribution within the cell interior was seen (Figure 5, panel E; additional file 6). These comets moved with an average velocity of 20.4 ± 6 μm/min (n = 29 comets from 3 cells where movement could be tracked for at least 12s). This was essentially indistinguishable from the distribution and behaviour observed for the microtubule tip-tracking APC ligands EB1 and EB3 when expressed as GFP fusion proteins in COS-7 cells (JMA, unpublished data). GFP-APC expressing COS-7 cells fixed and co-stained for EB1 showed that the GFP-APC comets did indeed co-localise with EB1 staining at the end of microtubules (Figure 5, panels F-H). Finally, GFP-APC puncta were occasionally observed in the cell interior (Figure 5, panel I-L; additional file 7). Over the time course of imaging these were either immobile or underwent rapid translocations at peak speeds consistent with microtubule motor-mediated transport (in the region of 2–3 μm/s). Previous reports have described an interaction between APC and kinesin-associated transport complexes [13] that could explain this behaviour.
In cultures that were seeded and transfected to be largely confluent during imaging a further GFP-APC distribution was seen. This consisted of a discontinuous array of punctate structures associated with peripheral membranes contacting other cells (Figure 6, panel A; additional file 8). This junctional localisation was confirmed by co-immunostaining for β-catenin (Figure 6, panels B-D) and actin (Figure 6, panels E-G), both of which confirmed that GFP-APC was closely associated (but not co-localised) with junctional proteins at the cortex. In comparison with the microtubule-associated GFP-APC distributions these cortical structures were far less dynamic. Tracking analysis indicated that they moved with an average velocity of 0.3 ± 0.36 μm/min (n = 70 puncta from 4 cells continuously observed for at least 10 min). This movement appeared to directly correlate with remodelling of the cell periphery during long-term changes in cell shape, indicating that these structures were essentially immobile relative to the cell cortex.
The junctional GFP-APC localisation was dominant in cells having extensive contacts with neighbouring cells, particularly those in densely confluent regions of an imaging dish. However, in regions of lower cell density we observed that both the microtubule and cell junction-associated GFP-APC distributions could be found in different regions of the same cell (Figure 6, panel H; additional file 9). The junctional population seen in these cells was unlikely to represent an overexpression artefact since it was present at low GFP-APC fluorescence intensities and no phenomena indicative of APC overexpression were seen (for example, GFP-APC decorated microtubule bundles). Observation of cells with both junctional- and microtubule-associated GFP-APC confirmed the very different dynamic behaviours of these protein populations. The microtubule-associated GFP-APC localisation was restricted to free cell edges and cell vertices and was highly motile whereas the junctional localisation was only present at sites of cell-cell contact and was essentially immobile. Dynamic GFP-APC microtubule-associated clusters were not seen in the vicinity of cell junctions, although when present the EB1-like localisation to growing microtubule distal tips was. We therefore suggest that the two GFP-APC populations might reflect the localised regulation of APC interactions in specific cellular regions.
Previous reports have shown that microtubules can be resistant to Nocodazole treatment in cells overexpressing microtubule-associated GFP-APC [14]. In order to test the effect of Nocodazole on transfected COS-7 cells possessing cortical and microtubule-associated GFP-APC we first incubated cells with 5 μg/ml Nocodazole for 1 hour (Figure 7, panels A-C). We noted that even after a one-hour incubation cells containing a small number of stable microtubules could be found (Figure 7, panel B, arrows). In transfected cells these remnant microtubules were decorated with GFP-APC (Figure 7, panels A-C). However, cortical APC could be clearly distinguished at the cell cortex in the absence of microtubules and appeared unaffected by Nocodazole treatment (Figure 7, panels A-C, arrowheads). We next examined the effects of microtubule depolymerisation on GFP-APC-cortical localisations in living cells. Figure 7, panels D-G and additional file 10 shows a cell in which both junctional (Figure 7, panels D-G arrowheads, also see additional file 10) and microtubule (Figure 7, panels D-G arrow, also see additional file 10) GFP-APC populations are present. At the beginning of the recording Nocodazole was added to a final concentration of 5 μg/ml. As previously reported by other investigators [6], the GFP-APC localization to microtubule distal tips was rapidly lost and the cortical clusters at microtubule ends gradually dispersed. However, the behaviour and intensity of junctional GFP-APC was unaffected by Nocodazole, even in subsequent experiments where the drug concentration was raised to 20 μg/ml for 30 min (data not shown). We conclude that the junctional association of GFP-APC is independent of microtubules. We next wanted to study the response of junctional GFP-APC to treatment with the actin depolymerising drug Cytochalasin D. Preliminary experiments with this drug at standard experimental concentrations of 10 μg/ml led to rapid cell rounding (data not shown). This made it impossible to study the effects of Cytochalasin D on cortical GFP-APC dynamics in these cells. We therefore decided to use low concentrations (1 μg/ml) of drug in order to study the effect of gradual, but not total, loss of cortical actin in COS-7 cells. Treatment of COS-7 cells expressing a GFP-actin construct [15] confirmed that at low concentrations Cytochalasin D perturbed actin integrity at the cell cortex in a stereotypical way. Figure 8, panels A-D, and additional file 11 show a GFP-actin expressing COS-7 cell treated with 1 μg/ml Cytochalasin D. At the beginning of the experiment the cell possesses a cortical actin belt around the cell periphery. Over the time course of the experiment this cortical belt can be seen to weaken and break at points within the cell (Figure 8, panels A-D arrowheads and additional file 11), These breakages lead to the contraction of the cortical actin belt along cell edges and the accumulation of actin at cell vertices. We next examined the results of low-level Cytochalasin D treatment in GFP-APC expressing cells. The results of a typical experiment are shown in Figure 8, panels E-H, and additional file 12. The cell of interest in this sequence has both cortical- and microtubule-associated GFP-APC. The earliest response, beginning around 10 min after drug addition, was a slow contraction of the whole cell; this was followed by breakage of the cell-cell contacts (Figure 8, panels E-H black arrow, and additional file 12) and the contraction of the GFP-APC puncta along the cell edge (Figure 8, panels E-H white arrow indicates direction of contraction, and additional file 12). This data closely resembled the effects of Cytochalasin D on the junction-associated cortical actin ribbon in COS-7 cells expressing GFP-actin (Figure 8, panels A-D, and additional file 11). When confluent GFP-APC expressing COS-7 cells were treated with 1 μg/ml Cytochalasin D for 30 min before fixation and staining with phalloidin an association of GFP-APC with both cytoplasmic actin aggregates (Figure 8, panels I-K, arrows) and cortical actin (arrowheads) was seen, confirming that the distribution of GFP-APC was closely linked to that of actin in these cells. These drug experiments lead us to conclude that in transfected COS-7 cells two spatially separate populations of GFP-APC exist at the cell periphery and can co-exist in the same cell. The first is highly dynamic and associated with microtubules near free edges and cell vertices, while the second is comparatively static and is closely associated with actin at sites of cell-cell contact.
Deletion analysis of the APC-cell junction association
To further define the domain in APC responsible for mediating its junctional localisation we made a number of GFP-APC deletion constructs for analysis in COS-7 cells. It seemed possible that the junctional localisation of APC cells could be mediated by an interaction with the junctional protein Discs Large (DLG), a known ligand of APC [16]. This interaction requires the extreme C-terminus of the APC molecule. We therefore made a construct, pEGFP-APCΔC (Figure 1, panel B), which directed the expression of full-length GFP-APC lacking its final 54aa. APC also dimerises using heptad repeats at the N-terminus of the protein and deletion of the first 58 amino acids of APC prevents this [17, 18]. Since any construct made by us might potentially be able to localize to junctions by interacting with endogenous APC, we therefore made a second construct, pEGFP-APCΔNΔC (Figure 1, panel C). This directed the expression of GFP-APC lacking both its final 54aa and first 207aa, which would both abolish DLG binding and prevent dimerisation. Expression of these constructs in transfected cells was confirmed by western blotting (Figure 2, panel B). In transfected COS-7 cells examined by time-lapse fluorescence microscopy the distribution of both of these fusion proteins closely resembled that of full-length GFP-APC, with the full range of both microtubule associated (not shown) and junctional APC localisations observed (Figure 9, panels A and B, arrowheads). This indicated that DLG binding was not required to recruit APC to the cortex in COS-7 cells and suggested that APC dimerisation was also unnecessary. A third construct, pEGFP-APC-C1 (Figure 1, panel D), drives the expression of a protein consisting of the last 170 amino acids of APC fused N-terminally to EGFP and would be predicted to interact with DLG. It showed a diffuse cytoplasmic localization within transfected cells (Figure 9, panel C), suggesting that a DLG interaction alone is insufficient for APC targeting to junctions in COS-7 cells. The final construct we examined, pEGFP-APC-N (Figure 1, panel E), directed the expression of a GFP-APC molecule truncated just after its armadillo repeat motifs. This construct showed three distinct localizations within the cell. The first was to the centrosome (Figure 9, panel D, arrow), confirmed by co-immunostaining of transfected cells for the centrosomal marker γ-tubulin (Figure 9, panels E-G). This observation confirms previous studies indicating a centrosomal localization for truncated APC proteins [19]. The second localization was to small, sometimes motile puncta within the cytoplasm (Figure 9, panel D; additional file 13). As noted above, the armadillo repeat region of APC has been shown to mediate an association with kinesin [13]. This association may potentially mediate this localization, but confirming this hypothesis was beyond the scope of the present study. The third localization was to sites of cell-cell contact in a distribution that resembled that previously seen with the full-length GFP-APC protein (Figure 9, panel D, arrowheads; additional file 13), confirmed by co-immunostaining for β-catenin (Figure 9, panels H-J. These data strongly suggest that the APC localization to cell junctions is mediated by a domain lying within the first 746aa of the protein. Unlike our longer GFP-APC fusion proteins, GFP-APC-N did not localise in dynamic clusters or decorate microtubule distal tips in specific regions of the cell periphery, nor did it localise to growing microtubule ends in the cell interior, consistent with a role for the APC microtubule and EB protein binding domains in mediating these localisations.
Our data indicated that C-terminally truncated GFP-APC molecules retained an ability to associate with sites of cell-cell contact in transfected cells. The pEGFP-APC-N (Figure 1, panel E) construct used in this study is similar to some of the more extremely truncated APC proteins expressed in colorectal cancer cells. Our observations might therefore predict that mutant APC should also localise to sites of cell-cell adhesion in cancer cells, as long as these cells were capable of assembling functional intercellular junctions. We therefore looked for colon cancer cell lines containing similar truncations to our pGFP-APC-N construct. The cell line COLO320 expresses an APC protein truncated at amino acid 811. Unfortunately however, this cell line displays very poor cell-cell adhesion so it was not possible to look at an APC junctional localisation in this cell type. We therefore examined the distribution of endogenous APC in the human colorectal tumour cell lines Caco-2 and SW480. Caco-2 cells possess one mutant copy of APC and express a protein truncated at amino acid 1367. These cells form adhesive junctions and although the truncated APC protein is longer than our minimal construct the question of whether C-terminally truncated APC proteins could localise to the cortex could still be addressed. The polyclonal M-APC antibody and the monoclonal antibody ALI12-28 would both be predicted to recognise the truncated APC protein. Unlike a previous report [5], we found that truncated APC localized to sites of cell-cell adhesion in Caco-2 cells following immunostaining with the polyclonal M-APC antibody (Figure 10, panels A-C) or the monoclonal ALI-12-28 antibody (Figure 10, panels D-F), but not with monoclonal antibodies specific for the APC C-terminus (Figure 10, panels G-I). Cells were also co-stained for actin to confirm the presence of adhesive junctions in the confluent cultures (Figure 10, panels J-L). The colon cancer cell line SW480, containing an APC protein truncated at amino acid 1338, produced similar results to those seen with Caco-2 cells (data not shown).