The data presented in this study indicate that the perinuclear aggregates formed in cells expressing EB1-ΔN2-GFP are aggresomes, although the underlying reason for their formation by this fusion protein remains unclear. Two mechanisms seem possible. First, the initial aggregation occurs as a direct consequence of the interaction between EB1-ΔN2-GFP and p150glued, perhaps coupled to aberrant retrograde transport of this complex, and aggresome formation occurs subsequent to this. A second possibility is that EB1-ΔN2-GFP misfolds, becomes refractive to proteolytic degradation and thereby triggers the formation of an aggresome. Neither EB1-GFP nor any of the other GFP-tagged EB1 deletion mutants generated in our laboratory, including those that do not interact with microtubules but retain an ability to bind APC and p150glued, form aggresomes in transfected cells [3
]. This suggests that the first possibility is less likely. Furthermore, our Western blotting analyses indicated that cell extracts from cultures transfected with the EB1-ΔN2-GFP construct contained significantly more fusion protein than those from cultures transfected with the EB1-GFP construct (Fig. ). Since the transfection efficiency for the EB1-GFP and EB1-ΔN2-GFP constructs was similar, equal amounts of transfected cell extracts were analysed and the base expression vector used for both constructs was identical, the best explanation for the higher levels of EB1-ΔN2-GFP relative to EB1-GFP in our experiments is that EB1-GFP was turned over normally within the cell whereas EB1-ΔN2-GFP was not. This data therefore supports the second mechanism suggested above. However, if aggregation is a response to EB1-ΔN2-GFP misfolding, then the presence of both p150glued and APC in the aggregates suggests that this misfolding is partial since the binding sites for both of these proteins appears to be functional (Figs. and ; ref [3
]). We therefore propose that EB1-ΔN2-GFP in COS-7 cells adopts a conformation where APC and p150glued binding are maintained but the fusion protein is recognised by the cellular stress response pathway for misfolded proteins, perhaps as a result of a disordered N
-terminal region arising from the loss of the CH domain from the EB1 N
-terminus. As this partially misfolded protein is resistant to degradation it accumulates within the cytoplasm and an aggresome is formed.
Other investigators have previously described the formation of aggresomes by a cytosolic GFP fusion protein (GFP-250) in living cells [14
]. EB1-ΔN2-GFP aggresome formation appeared to correlate well with that observed for GFP-250, particularly during aggresome assembly following nocodazole treatment and wash out. However, previous ultrastructural studies have suggested that aggresomes represent an "aggregate of aggregates" formed by an accumulation of individual particles without further coalescence into a cohesive structure [12
]. Our time-lapse studies suggest that this is not the case with EB1-ΔN2-GFP. In our system aggresomes exhibited dynamic microtubule-dependent linear extensions. These typically resembled beads on a string with one end remaining attached to the main structure of the aggresome. Detachment and anterograde movement of individual particles from aggresomes was never seen. These observations indicate that the brighter structures in EB1-ΔN2-GFP aggresomes must be linked in some way and, since extensions were never seen to fully detach from aggresomes, this linkage is strong enough to resist the forces generated by microtubule motors. Together with our data showing that microtubules are not necessary to maintain the pericentrosomal location of pre-existing EB1-ΔN2-GFP aggresomes, this suggests that these structures possess an inherent cohesiveness and are not simply a collection of individual elements maintained in close proximity to the centrosome by dynein motor activity. At present it remains unclear whether this dynamic behaviour is a universal feature of aggresomes or is restricted to those formed by EB1-ΔN2-GFP. However, we note that associations with kinesin II have recently been described for both p150glued [23
] and APC [24
]. Since both proteins are clearly present in EB1-ΔN2-GFP aggregates these interactions could potentially contribute to the dynamic behaviour of these structures.
Recent studies of aggresome formation by mutant CFTR and SOD proteins in cells treated with proteosome inhibitors indicated that the dynein/dynactin components CDIC, p50dynamitin and p150glued were all recruited to aggresomes [15
]. This raises the possibility that the p150glued observed in EB1-ΔN2-GFP-induced aggresomes was present as a normal cofactor for aggresome formation rather than as a specific EB1-ΔN2-GFP ligand. Furthermore, an increase in the amounts of p150glued and p50dynamitin in the detergent insoluble fraction of cells containing mutant SOD-induced aggresomes was reported [25
]. In our experiments no increase in the amount of insoluble full-length forms of these proteins was observed. However, in contrast to previous studies we found lower molecular weight p150glued species in the detergent-insoluble fraction of cells expressing EB1-ΔN2-GFP. The p150glued antibody used in this study recognises an epitope at the N
-terminus of the protein [26
], the same region that mediates the EB1-p150glued interaction [3
] but distinct from the regions responsible for the interactions with other dynein/dynactin subunits [8
]. It seems possible that the bands detected in our immunoblots represent N
-terminal p150glued fragments tightly bound to insoluble EB1-ΔN2-GFP and inaccessible to the proteosome. An analogous mechanism has been proposed to explain the limited proteolytic processing of NF-κB precursors, where it is suggested that a close association with a partner molecule inhibits the processive degradation of the NF-κB p50 domain [27
]. To explain the discrepancies between our data and that presented in the study of mutant SOD-induced aggresomes we therefore propose that in the latter case the dynactin complex is recruited to but not trapped within aggresomes as part of a normal cellular response, whereas in our system the aggresome-associated p150glued is tightly bound to EB1-ΔN2-GFP. Subsequent degradation of exposed regions of the molecule destroys the binding sites for p150glued-associated dynein/dynactin subunits, precluding their stable co-incorporation into the aggresome. However, binding to proteolytically-resistant EB1-ΔN2-GFP molecules preserves the p150glued fragments detected by the monoclonal antibody used in our study.