Cleavage of APP by the α-secretases, ADAM10 and ADAM17/TACE, is of physiological importance because it precludes the formation of the toxic Aβ peptide. Moreover, the secreted APP ectodomain fragment, sAPPα, promotes neural plasticity and neurogenesis, and exhibits neuroprotective properties [6
]. G-protein coupled receptors, and compounds such as phorbol esters that directly activate PKC, increase cleavage of sAPPα by α-secretases, but the mechanism is not fully understood [6
]. In the present study, we examined the effect of inhibiting dynamin-dependent endocytosis on the activity of ADAM10 toward its substrate APP.
Although other ADAMs, notably ADAM17, are capable of cleaving APP, the expression of APP in adult mouse brain overlaps to a much greater extent with ADAM10 than with ADAM17 [37
]. Moreover, APP shedding is nearly abolished in primary neurons from mice null for ADAM10 [38
], and in cell lines following knockdown of ADAM10 [39
], indicating that ADAM10 is the physiologically relevant α-secretase in brain. Co-expression of ADAM10 with a dominant-negative dynamin mutant, dyn I K44A, increased surface expression of both mature and immature forms, indicating that endocytosis of ADAM10, like that of APP [16
] is dynamin-dependent. An earlier report showed that endocytosis of membrane-type matrix metalloprotease 1 (MT1-MMP) is inhibited by a dynamin K44A mutant, resulting in increased cleavage of its substrate pro-MMP2 [40
]. Thus, modulation of internalization could be a general mechanism regulating the activity of transmembrane metalloproteases toward their substrates.
Several laboratories have examined the effects of kinase activation on ADAM trafficking and protease activity. PMA caused down-regulation of ADAM17/TACE surface expression in Jurkat and THP-1 cells [41
], and a transient increase followed by a decline in HeLa cells. The latter response was blocked by an inhibitor of mitogen-activated protein kinase kinase, and by mutation of threonine 735 to alanine of ADAM17/TACE [42
]. PMA also increased degradation of the mature processed form of ADAM17/TACE, but not ADAM10, in HEK293 cells [43
], and promoted the translocation of ADAM10 to the cell membrane in glioblastoma cells [44
]. Taken together, the evidence indicates that PKC-regulated activation of ADAM10, or ADAM17/TACE, may be associated with alterations in trafficking in some cell types; however, neither PMA nor M3 receptor activation appear to regulate cell-surface expression of ADAM10 in HEK cells (Figure and data not shown). Thus, the ability of these activators to promote APP ectodomain shedding in our model is not mediated by inhibition of internalization of either ADAM10, or its substrate APP [17
Overexpression of ADAM10 increased constitutive, but not carbachol-evoked release of APP695
from HEK-M3 cells (Figure ). Even in the absence of carbachol, surface APP695
was greatly reduced in ADAM10/APP695
co-transfectants (Figure ), suggesting that under these conditions the response to carbachol is limited by the availability of APP695
. This interpretation is supported by an earlier report, which showed that ADAM10 overexpression caused a much greater fold-increase in basal than in PMA-stimulated APP shedding from HEK cells, whereas a dominant negative ADAM10 mutant inhibited PMA-stimulated APP shedding by approximately 75% [45
]. Interestingly, surface levels of immature ADAM10, but not mature ADAM10, were increased in cells co-transfected with APP695
, relative to levels in cells expressing ADAM10 and empty vector (Figure ). This suggests that the association of APP with ADAM10 inhibits the processing of ADAM10 by proprotein convertases, but how or where in the cell this occurs is unknown.
In addition to increasing surface expression of ADAM10, the dynamin mutant caused a marked elevation in cellular levels of an ADAM10 CTF, the abundance of which was further increased by the γ-secretase inhibitor L-685,458 (Figure ). This is consistent with earlier reports that ADAM10 is subject to proteolysis by ADAM9 or 15, generating a CTF that is a substrate for γ -secretase [19
], and suggests that the cleavage of ADAM10 occurs at the cell surface. As with surface expression of ADAM10, generation of the ADAM10 CTF was not affected by carbachol (Figure ), in contrast with an earlier report that PMA increases ADAM10 shedding in SH-SY5Y neuroblastoma cells [20
]. The reasons for this discrepancy are unclear, but could be related to cell-specific differences, which also appear to underlie the varying effects of PKC activators on ADAM10 trafficking. Because of the potential signaling role of the ADAM10 CTF [19
] it would be of interest to identify factors that regulate generation of this fragment. One possible mechanism could involve alterations in ADAM10 endocytosis resulting from direct interactions with other cell surface proteins. An example of such a mechanism is the modulation of APP internalization and processing by members of the low-density lipoprotein receptor-related protein (LRP) family [46
]. Notably, binding to LRP1B retains APP at the cell surface, increasing its cleavage by α-secretase, and reducing Aβ generation [46
]. Similarly, F-spondin, a secreted signaling molecule implicated in neuronal development and repair [48
], binds to both the APP ectodomain and the apolipoprotein receptor ApoE2, promoting sAPPα release, and inhibiting cleavage by β-secretase [49
]; both these effects are consistent with reduced internalization. We predict that binding partners of ADAM10 could affect its processing in a similar way.
Evidence from multiple laboratories indicates that amyloidogenic APP processing occurs in lipid rafts, whereas processing by ADAMs is confined to non-raft regions [31
]. Disruption of raft domains by depleting membrane cholesterol shifts APP processing in favor of non-amyloidogenic processing. This could reflect the displacement of APP from raft domains, and away from β- and γ-secretases [31
], but might also be due in part to inhibition of endocytosis [52
]. Although dynamin I K44A expression increased surface levels and ectodomain cleavage of mature ADAM10, there was no accompanying change in the partitioning of either full-length ADAM10 or its CTF in raft and non-raft domains (Figure ). Thus, the effects of the dynamin mutant on ADAM proteolysis are likely simply to reflect the accumulation of ADAM10, and possibly of the ADAM that targets it, at the cell surface. To our knowledge, the effect of dynamin inhibition on the distribution of APP in membrane subdomains has not been addressed. However, like ADAM10 itself, the ADAM10 substrate N-cadherin underwent increased proteolysis in the presence of dyn I K44A, without altering its distribution between Triton-soluble and insoluble membrane domains (Figure ).
ADAM10 has numerous substrates [53
], and alterations in its trafficking to or from the cell surface could affect its activity towards any or all of them, thereby regulating their biological activity. In the case of APP and N-cadherin, this could inhibit their ability to promote cell-cell adhesion, neuronal development and synaptogenesis [54
], but at the same time would increase the signaling functions of their liberated C-terminal fragments [35
]. Increased generation of the ADAM10 CTF, also a consequence of decreased internalization, might exert biological effects of its own [19
], but these remain to be elucidated.