Endothelin-converting Enzymes Degrade Intracellular β-Amyloid Produced within the Endosomal/Lysosomal Pathway and Autophagosomes*

Background: Endothelin-converting enzymes (ECEs) degrade β-amyloid (Aβ) peptide. Results: ECE inhibition produces, in addition to extracellular Aβ accumulation, intracellular Aβ accumulation within endosomal/lysosomal and autophagic vesicles. Conclusion: An intracellular pool of Aβ is regulated by ECE activity at the sites of production. Significance: ECE dysfunction may cause intraneuronal Aβ accumulation, which is associated with neurotoxicity early in AD progression. Impairments in Aβ removal are increasingly being considered as a possible cause for the abnormal Aβ build-up typical of Alzheimer disease. Of particular interest is a pool of Aβ that accumulates intraneuronally and may contribute to neuronal toxicity. The mechanism for intraneuronal accumulation, however, is not well understood and is commonly attributed to impaired removal of extracellular Aβ by neurons. Based on the intracellular distribution of the well established Aβ degrading enzymes, ECE-1 and ECE-2, we tested whether impairments in their catalytic activity could lead to intracellular Aβ accumulation. Using SH-SY5Y cells overexpressing wild-type amyloid precursor protein and pharmacological inhibition of endogenous ECE activity, we found that ECEs participate in the degradation of at least two distinct pools of Aβ; one destined for secretion and the other being produced and degraded within the endosomal-autophagic-lysosomal pathways. Although ECE-1 regulates both pools of Aβ, ECE-2 regulates mainly the intracellular pool of the peptide. Consistent with this result, ECE-2 was found to co-localize with markers of the endosomal/lysosomal pathway but not with a trans-Golgi network marker. Furthermore, ECE-2 was detected in autophagic vesicles in cells treated with chloroquine. Under these conditions, ECE inhibition produced significantly higher elevations in intracellular Aβ than chloroquine treatment alone. This study highlights the existence of Aβ clearance mechanisms by ECEs at intracellular sites of production. Alterations in ECE activity may be considered as a cause for increased intraneuronal Aβ in Alzheimer disease.

in the CNS, in different soluble states of aggregation or as extracellular insoluble deposits known as amyloid plaques (1). The accumulation is progressive and widely accepted as an important contributor to the neuronal dysfunction and ultimate neuronal loss characteristic of the disease (2). Early in the disease progression, a pool of A␤ accumulates intraneuronally before extracellular amyloid plaque formation (3)(4)(5)(6), and mounting evidence points to intracellular A␤ (iA␤) as a possible cause of the neuronal toxicity typical of AD (7)(8)(9)(10)(11)(12)(13)(14).
As in the case of extracellular A␤ (eA␤), the mechanism of iA␤ accumulation is unknown. Based on the evidence that A␤ is constitutively secreted to the extracellular space following the sequential cleavage of amyloid precursor protein (APP) by ␤-secretase and ␥-secretase, it has been assumed that iA␤ originates from internalization of secreted A␤. However, the ultimate ␥-secretase cleavage for the release of A␤ is largely conducted at the endosome (15), thus requiring intracellular trafficking of A␤ before secretion. At the same time, A␤-degrading enzymes are active in multiple compartments other than the lysosome and the extracellular space (16). If A␤ degradation is an important contributor to the steady-state levels of A␤ before secretion it is then possible that iA␤ could accumulate due to failure in A␤ catabolism.
To date, several A␤ degrading enzymes have been shown to significantly contribute to A␤ homeostasis (17). Numerous studies in vitro and in animal models support the physiological role of insulin-degrading enzyme, neprilysin (NEP), endothelin-converting enzyme-1 (ECE-1) and ECE-2, among others, in A␤ degradation (16). NEP, ECE-1 and ECE-2 are members of the M13 family of metalloproteases, type II membrane-bound zinc metalloproteases sensitive to the inhibitor phosphoramidon (PA).
ECE-1 and ECE-2, coded by different genes, are characterized by the ability to process big endothelin-1 into the potent * This work was supported by NINDS, National Institutes of Health Grant R01 NS073512 (to E. A. E.). 1 To whom correspondence should be addressed: vasoconstrictor endothelin-1 (18). Similar to NEP, ECE-1 and ECE-2 are expressed in areas relevant to AD (19), and we have previously demonstrated that in ECE-1 and ECE-2 knock-out mice, there is an increase in endogenous levels of A␤ in brain (20). Although NEP is expressed predominantly on the plasma membrane, the ECE family is more broadly distributed. ECE-1 is composed of four isoforms that are located on the plasma membrane as well as in different intracellular compartments, including the secretory pathway, recycling endosomes, and late endosomes (21). For ECE-2, all four isoforms are strictly intracellular, but their distribution has not yet been properly characterized. Based on the diverse intracellular distribution of the ECE family, we investigated how alterations in ECE activity could lead to iA␤ accumulation in SH-SY5Y human neuroblastoma cells overexpressing wild-type APP as an in vitro neuronal model.
Cell Culture-SH-SY5Y cells stably expressing wild-type human APP 695 were maintained in DMEM supplemented with 10% FBS, glutamine, penicillin/streptomycin, and the selective antibiotic geneticin at 400 g/ml (Invitrogen). Cells were routinely passed by trypsinization. The same medium and maintenance protocol was followed for growing CHO cells. For measurement of A␤ by ELISA, medium and cell extracts were obtained from growing cells at 70 -80% confluency in 12-well plates with 500 l of medium. For transient SH-SY5Y transfection, the "nanojuice" reagent (EMD Chemicals) was used according to the manufacturer's instructions; a booster:DNA ratio of 2:1 was added to DMEM with 10% FBS containing no antibiotics, and cells were incubated for 72 h prior to any assay.
Western Blot-Cells were lysed in cold lysis buffer (PBS with 0.1% Triton X-100 and proteinase inhibitors) and protein homogenates were resolved in Novex 10 -20% Tris-tricine gels and transferred to nitrocellulose membranes. For protein detection, membranes were blocked for 1 h with 10% normal goat serum and incubated with the corresponding primary antibody. Bound antibody was detected with ECL reagents (Millipore) following incubation with the appropriate HRPconjugated secondary antibody. Signal was digitally recorded with the ImageQuant LAS 4000 (GE Healthcare).
Immunofluorescence-Cells were fixed with cold methanol at Ϫ20°C for 10 min. As blocking reagent, 5% BSA containing 0.1% Tween 20 was used. Following incubation with primary antibodies, Alexa Fluor 488 and 546-conjugated secondary antibodies were used for visualization under a Zeiss Axio Imager Z1 fluorescent microscope.
ELISA Measurements-Levels of A␤1-40 were measured from conditioned medium and Triton X-100 cell extracts using a highly specific sandwich ELISA. The 33.1.1 antibody against the N-terminal region of the peptide was used for capture and the HRP-conjugated 40-specific antibody, 13.1.1, was used for detection. A␤ ending at position 42 was detected by capturing with anti-A␤42 antibody 2.1.3 and detecting with 4G8.
Data Analysis and Representation-Figures are representative of individual experiments that were repeated at least three times. GraphPad Prism (version 5.01, GraphPad Software, Inc.) was used for all data analysis and graphical representation. For column graphs, column bars represent the mean Ϯ S.E. Student's t test was used for statistical analysis and a significance of p Ͻ 0.05 was set for acceptance. Statistical significance was indicated by *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001; and ****, p Ͻ 0.0001.

RESULTS
PA Treatment Increases both eA␤ and iA␤-To define the contribution of ECE activity to iA␤ degradation, SH-SY5Y-APP cells expressing endogenous levels of ECEs were treated with 100 M PA for 48 h, and levels of A␤40 and A␤42 were measured by ELISA in medium and cell extracts. Similar to previous studies (26,27), we found that PA treatment produced a significant ϳ3-fold elevation in the level of eA␤40, and 1.8-fold elevation in eA␤42 (Fig. 1, A and B). In the same manner, in cell extracts, PA treatment also produced significant elevations in iA␤40 and iA␤42 ( Fig. 1, C and D).
Demonstrating iA␤ accumulation is particularly challenging due to cross-reactivity of specific A␤ antibodies with APP or cleavage products of APP that contain a portion of, or the full A␤ sequence. To define the specificity of the iA␤ measurements by our ELISA system, cells were treated with the ␥-secretase inhibitor DAPT to stop A␤ production and increase intracellular accumulation of C-terminal fragments (CTFS) (28). As seen in Fig. 1, A and B, levels of eA␤40 and eA␤42 after a 48-h treatment with 100 M DAPT or PA ϩ DAPT were nearly undetectable. For iA␤ analysis in the absence of PA, A␤40 and A␤42 measurements were similar with and without DAPT treatment ( Fig. 1, C and D), indicating a significant background signal in the cell extracts that is not due to A␤ or CTFs. Importantly, the significant increases in iA␤40 and iA␤42 measured in cells treated with PA were completely blocked by treatment with DAPT, confirming that PA produces iA␤40 and iA␤42 accumulation.
As additional confirmation of PA-mediated iA␤ accumulation, we repeated the experiments with the CNS-derived H4 cell line transfected with wild-type APP (29). Similar to the results obtained with SH-SY5Y-APP cells, treatment of H4-APP cells with 100 M PA for 48 h produced significant increases in both eA␤ and iA␤ levels ( Fig. 1, E and F). Interestingly, H4-APP cells secrete substantially less A␤ than SH-SY5Y-APP cells and only 4-fold higher than wild-type SH-SY5Y cells (data not shown), and yet H4 cells treated with PA accumulated intracellular A␤ at levels greater than SH-SY5Y-APP cells.
To verify that the A␤ accumulation was due to impaired degradation rather than increased production, we analyzed by Western blot the levels of full-length APP and CTFs, as a measure of secretase activity. Western blot results using a C-terminal APP antibody showed that APP was preferentially processed by ␣-secretase, as seen by higher levels of CTF-␣ than CTF-␤ and that PA did not alter APP processing (Fig. 2B). As expected, in cells treated with DAPT, we detected a significant increase in the levels of CTFs. This increase was not affected by PA treatment. Also, by immunocytochemistry, we observed that PA treatment did not alter APP distribution in the cell ( Fig.  2A), a potential cause for altered A␤ levels (15).
Using N-terminal A␤ antibody 33.1.1 (the capture antibody for our ELISA system), we verified iA␤ accumulation by immunocytochemistry. In PA-treated cells, immunoreactivity with a punctuated pattern was detected in the cell body and surrounding the nuclei ( Fig. 2A), whereas in control cells, only faint intracellular staining was observed.
Finally, the accumulation of A␤ in cell extracts, detected by sandwich ELISA (Fig. 1), and immunocytochemistry ( Fig. 2A) was further confirmed by Western blot using either A␤1-16 antibody 33.1.1 or A␤17-24 antibody 4G8. A 4-kDa immunoreactive band was detected in extracts from SH-SY5Y-APP cells treated with PA (Fig. 2C). This band was completely absent in extracts of cells co-treated with PA and DAPT.
NEP Does Not Contribute to A␤ Accumulation in SH-SY5Y Cells-PA inhibits the activity of members of the NEP and ECE families and, less potently, angiotensin-converting enzyme (30). Thus, to differentiate the contribution of the NEP and ECE family members to iA␤ removal, we tested the effect of a more selective ECE inhibitor, CGS 35066 (31). As shown in Fig. 3, a 24-h treatment with CGS 35066 increased the levels of both iA␤ and eA␤ in a dose-dependent manner, with maximal effects similar to that of PA. In contrast, with the potent NEP inhibitor thiorphan, we found no significant increase in either iA␤ or eA␤ at doses up to 10 M after 48 h. These results indicate that NEP and other thiorphan/PA-sensitive proteases do not have a major role in A␤ degradation in our cellular model, consistent with previous reports showing that NEP is poorly expressed in SH-SY5Y cells (32). Increased levels of A␤ by PA are likely due . eA␤40 and eA␤42 production were blocked by treatment with the ␥-secretase inhibitor DAPT. PA also produced a significant accumulation of iA␤40 (C) and iA␤42 (D) in these cells (p ϭ 0.0004 and p ϭ 0.0225, n ϭ 3). In cell extracts, ELISA measurements remained unchanged upon treatment with DAPT or DAPTϩPA, indicating that signal measured in non-treated groups is attributable to background and that the increase in iA␤ signal with PA treatment is a true increase in A␤ peptide levels. Similar to results obtained with SH-SY5Y-APP cells, treatment of H4-APP cells with PA produced significant accumulations of eA␤40 (E) (p ϭ 0.0028, n ϭ 3) and iA␤40 (F) (p ϭ 0.0058, n ϭ 3).
to specific inhibition of ECE family members (33), although we cannot completely rule out the presence of an unidentified ECE-like enzyme.
Accumulated iA␤ Does Not Have an Extracellular Origin-There is no clear understanding of the origin and mechanism of iA␤ accumulation in neurons. One possibility is an overloading of the mechanism of eA␤ removal via endocytosis. Another possibility is that A␤ accumulates due to dysfunction in an undefined step prior to secretion. To study whether PA was inhibiting the degradation of endocytosed A␤, we treated empty vector-transfected SH-SY5Y cells with A␤ at levels similar to those secreted by SH-SY5Y-APP cells treated with PA. If iA␤ originates from internalization of the extracellular pool, incubation with this amount of exogenous A␤ should produce increases in iA␤ similar to those observed in SH-SY5Y-APP cells treated with PA. Fig. 4 shows the results of 48 h of treatment with 5 nM A␤ and 100 M PA. Levels of eA␤ and iA␤ were measured soon after A␤ addition and again after 48 h. A␤ measurement after the 48-h incubation period showed a dramatic decrease in the level of exogenous A␤ in the presence or absence of PA, indicating internalization or nearly complete degradation by proteases insensitive to PA. Levels of iA␤ were measured to determine whether the peptide had been internalized and accumulated within the cell. There was no increase in iA␤ accumulation following treatment with 5 nM A␤, suggesting that if SH-SY5Y cells endocytose A␤ (34,35), any internalized peptide is degraded by non-PA sensitive proteases.
A␤ Accumulates within A␤-producing Compartments-The intracellular location of members of the ECE family coincides with compartments where CTF-␤ is cleaved by ␥-secretase to release A␤ (see "Discussion"). Thus, we tested whether iA␤ accumulated in A␤-producing compartments. Immunostaining with an antibody against the C terminus of APP revealed a punctuated staining pattern that was not seen with an N-terminal APP antibody (not shown) and colocalized with iA␤ (Fig. 5). Also, the punctuated iA␤ staining localized with rab7-positive compartments (marker for late  endosome), suggesting that iA␤ is degraded at the site of production within the late endosome.
ECE-2 Resides in the Endosomal/Lysosomal Pathway-ECE-2 has not been as thoroughly investigated as ECE-1. Based on its low pH optimum (5-5.5) and substrate affinities similar to those of ECE-1, it has been postulated that ECE-2 resides in the TGN (36). However, this has not been demonstrated convincingly. We confirmed by RT-PCR that SH-SY5Y cells and H4 cells express ECE-2 (data not shown), but endogenous levels of the protein were only weakly detected with commercially available antibodies. Therefore, we evaluated the cellular localization of the enzyme in transfected cells by immunocytochemistry. We observed that in ECE-2 overexpressing SH-SY5Y cells, ECE-2 partially co-localized with rab5 (early endosomal marker), rab7, and LAMP-1 (lysosomal marker) but not with TGN38 (TGN marker) (Fig. 6). Results suggest that ECE-2 . Levels of eA␤ in A␤-treated cells were decreased after 48 h in the presence or absence of PA, suggesting that eA␤ removal is not dependent on ECE activity. The decrease in eA␤, however, did not reflect an increase in iA␤ indicating that ECEs do not degrade internalized eA␤ or that all eA␤ was degraded in the extracellular space by proteases not sensitive to PA.

FIGURE 5. A␤ accumulates in cellular compartments rich in APP and/or
CTFs. In PA-treated SH-SY5Y-APP cells, iA␤ immunoreactivity co-localized with APP/CTFs detected with a C-terminal APP antibody and rab7, indicating that iA␤ accumulation occurs at a site of production within the late endosomal/lysosomal pathway. FIGURE 6. Intracellular distribution of ECE-2. SH-SY5Y-APP cells transiently transfected with ECE-2 (variant 2) were co-stained with antibodies against ECE-2 and markers of various intracellular compartments. Co-localization of ECE-2 with rab5, rab7, and lamp-1, markers of early endosome, late endosome, and lysosome, respectively, demonstrated that ECE-2 is mostly located in the endosomal/lysosomal pathway. In contrast, ECE-2 did not co-localize with TGN38, a marker of the trans-Golgi network, suggesting that ECE-2 does not conduct its activity within the secretory pathway.
A␤-degrading activity appears restricted to the endosomal/lysosomal pathway and not the secretory pathway in neuronal cells. (21), we next studied whether ECE-2 and ECE-1b directly participate in iA␤ degradation. SH-SY5Y-APP cells were transiently transfected with either enzyme prior to incubation with PA. The relative inefficiency of transfection in this cell line permits the comparison of A␤ accumulation within ECE-overexpressing cells and neighboring non-transfected cells. First, we observed that both ECE-1b and ECE-2 partially co-localized with APP and/or APP CTFs (Fig.  7). In addition, in SH-SY5Y-APP cells treated with PA, overexpression of either ECE-2 or ECE-1b dramatically lowered accumulation of iA␤ (expressed as the average number of A␤-positive granules per cell) as compared with surrounding non-ECE transfected cells (Fig. 8, A and B). Even though these cells were treated with PA, we suspect that intracellular ECE activity was not completely inhibited in the overexpressing cells. This hypothesis is supported by the observation that reductions in extracellular A␤ concentration in ECE-1-transfected CHO cells are not reversed 100% by PA treatment at the concentration used in this experiment (see Fig. 10).

ECE Inhibition Produces Greater iA␤ Accumulation under Conditions of Impaired Autophagic/Endocytic Flux-Macroau-
tophagy is a distinct lysosomal degradative pathway specialized in the removal of long-lived proteins, organelles, and other macromolecules by the formation of double membrane vesicles known as autophagosomes (37). Autophagosomes ultimately fuse with lysosomes, resulting in degradation of their contents. Some late endosomes may fuse with autophagosomes upstream of the lysosome, and these and nascent autophagosomes can both be referred to as autophagic vesicles (AVs). There is evidence that A␤ is targeted to degradation by macroautophagy (38), whereas entrapment of membrane domains rich in ␥-secretase and CTF-␤ within autophagosomes provides them with the capacity to produce A␤ (39,40). Under normal conditions, autophagic-lysosomal fusion is very efficient in neurons, and autophagic vesicles are transient and difficult to detect. In AD, however, dysfunction in this pathway results in accumulation of autophagosomes in the brain (41).
After localizing ECE-1b and ECE-2 to late endosomal compartments, we next tested whether these enzymes also contribute to A␤ degradation within AVs. SH-SY5Y-APP cells were therefore treated with chloroquine (CQ), a lysosomotropic agent that produces AV accumulation by inhibiting AV/lysosomal fusion. Treatment with CQ for 48 h produced a dose-dependent decrease in eA␤ levels (Fig. 9A). Although CQ decreased A␤ secretion, CQ at the highest dose (10 M) produced a significant increase in iA␤. Furthermore, combined PAϩCQ treatment produced a significant dose-dependent increase in iA␤, as compared with the treatment with PA or CQ alone.
Immunocytochemical analysis of the autophagosome marker LC3B showed that treatment with CQ (5 M) produced a redistribution from a cytosolic to a vesicular pattern, consistent with AV accumulation (Fig. 9B). Co-immunostaining with A␤ antibody 33.1.1 demonstrated that iA␤ was accumulating mainly in LC3B positive vesicles in CQϩPA treated cells. Treatment with PA alone did not alter LC3B distribution (data not FIGURE 7. Co-localization studies of ECE-2 and ECE-1b with APP. SH-SY5Y-APP cells transiently transfected with either ECE-2 or ECE-1b were stained with the corresponding ECE antibody along with a C-terminal APP antibody. Both ECE-2 and ECE-1b showed partial co-localization with compartments rich in APP and/or CTFs.  FEBRUARY 22, 2013 • VOLUME 288 • NUMBER 8 shown). By Western blot, we observed a similar accumulation of CTF-␣, CTF-␤, and CTF-␥ in CQ and CQϩPA treated groups, confirming the processing of APP C-terminal fragments by ␥-secretase within AVs (Fig. 9C).

ECE-2 Is Present in Autophagic
Vesicles-Next, we tested whether ECE-1b and/or ECE-2 were present in AVs. As seen in Fig. 9D, upon CQ treatment, ECE-2 co-localized within LC3B positive vesicles. In contrast, we observed little co-localization of ECE-1b in these compartments. Collectively, our data supports that AVs are a source of iA␤ production, but levels are tightly regulated by degradation. In this cellular model, ECE-2, but not ECE-1b, appears to be a major contributor to A␤ degradation within AVs. Our data also suggest that under certain pathological conditions, espe-cially where autophagic/endocytic/lysosomal flux is impaired, iA␤ accumulation may occur with no change, or even a decrease, in eA␤.
ECE-2 Does Not Regulate eA␤-To gain further knowledge of how iA␤ degradation by ECE-1 and ECE-2 impacts secreted A␤, we worked with CHO cells, which do not endogenously express members of the ECE family (42). Transfection of CHO cells with all the different isoforms of ECE-1 confirmed previous findings about its role in modulating eA␤ (27). Overexpression of ECE-1 isoforms present on the plasma membrane (a and c) as well as intracellular isoforms (b and d) (21) significantly decreased the levels of eA␤40 and eA␤42 (Fig. 10). These effects were reversed by PA treatment, indicating that they were dependent on the activity of the enzymes. FIGURE 9. ECEs participate in clearance of iA␤ within autophagosomes. Treatment with CQ for 48 h produced a dose-dependent decrease in eA␤40 levels, even in the presence of PA (A). For iA␤, CQ at the highest dose (10 M) produced a significant increase in iA␤ (p ϭ 0.0075, n ϭ 3), and combined PAϩCQ treatment produced a significant dose-dependent increase in iA␤, as compared with the treatment with PA alone (p Ͻ 0.0001, n ϭ 3 for 10 M CQ). By immunocytochemistry (B), we observed a redistribution of the autophagosome marker LC3B (in red) from a cytosolic to a vesicular pattern in CQ-treated cells, consistent with AV accumulation. Co-immunostaining with A␤ antibody 33.1.1 (in green) demonstrated that iA␤ was accumulating mainly in LC3B-positive vesicles in CQϩPA-treated cells (B). By Western blot using a C-terminal APP antibody, we observed an accumulation of CTF-␣, CTF-␤, and CTF-␥ in CQ-and CQϩPA-treated groups, confirming the processing of APP C-terminal fragments by ␥-secretase within AVs (C). In CQ-treated cells, ECE-2 (in green) co-localized within LC3B-positive vesicles (in red) in ECE-2-overexpressing cells. In contrast, we observed little co-localization of ECE-1b with LC3B in CQ-treated cells (D).
In contrast to the effect of intracellular ECE-1 isoforms on A␤ secretion, overexpression of ECE-2 in CHO cells resulted in little or no effect on eA␤ levels (Fig 10). Similar results were obtained with ECE-2 variants 2 (data not shown) and 5, both of which are expressed in brain. To confirm that transfected ECE-2 cells were expressing an active form of the enzyme, ECE activity in membrane fractions was quantified using a big endothelin-1 conversion assay (36) (data not shown). These results indicate that the trafficking and subcellular localization of intracellular ECE-1 and ECE-2 differ in ways impacting their interaction with distinct pools of A␤. ECE-2 appears not to traffic through the secretory pathway but is active predominantly in late endosomes and AVs where it degrades an intracellular pool of A␤ not destined for secretion.

DISCUSSION
The biogenesis of A␤ is a complex multi-step process involving the traffic of APP through the TGN to the cell surface and clathrin-mediated internalization into endosomes, where the sequential ␤and ␥-cleavage of APP takes place. A␤ is then secreted via endosomes or the TGN. Alternatively, CTF-␤ fragment may traffic unprocessed from the endosome to TGN (43) or to the lysosomal pathway (44) where A␤ production may still occur. As evidence suggests, APP sorting and secretase activity are critically important factors modulating A␤ production. The steady-state levels of A␤, however, also depend on its degradation rate. Because A␤ is believed to be continuously secreted, A␤ catabolism has primarily been considered an extracellular and lysosomal process (after internalization of extracellular A␤). Although this may hold true for certain A␤ degrading enzymes such as NEP or cathepsins, catabolism within other cellular compartments by enzymes such as the ECEs is less well characterized.
In addition to its strong expression in endothelial cells, ECE-1 is expressed in neurons in different areas of the CNS, including areas relevant to AD, and contributes to the metabolism of different neuropeptides as well as receptor resensitization within the endosomal system (45,46). The intracellular ECE-1d and ECE-1b isoforms are localized in recycling endosomes with ECE-1b present also in late endosomes. Both intracellular ECE-1 isoforms, however, have also been shown to be transiently expressed at the plasma membrane and secretory pathway (21).
We have previously established that ECE-1a and ECE-1b regulate extracellular levels of A␤ by degrading the peptide primarily before secretion and not in the extracellular space (27). Here, we show that all 4 ECE-1 isoforms are capable of regulating eA␤ accumulation. These findings, combined with the cellular distribution of the enzymes, demonstrate that A␤ is actively degraded within the vesicles where it is produced. In addition, the ϳ2-3-fold elevation in secreted A␤ upon ECE inhibition highlights ECE activity as a key factor in determining levels of secreted A␤, perhaps as influential as secretase activity.
In this report, we also demonstrate that ECE dysfunction produces, in addition to increased eA␤, an accumulation of iA␤ that can be detected by ELISA, Western blot, and immunocytochemistry. The majority of the accumulated iA␤, however, seems independent from eA␤ and is not destined for secretion, judged by the ability of ECE-2 overexpression to prevent its build-up without influencing extracellular levels of the peptide. This result suggests that A␤ accumulating intracellularly is not being produced within early/recycling endosomes and instead originates from an APP pool sorted to other compartments outside of the secretory system. Co-localization of iA␤ with rab7 shows the late endosome as a site of iA␤ accumulation, consistent with findings in AD brains (14).
Although part of the A␤ produced within recycling endosomes could be routed through the late endosomal/lysosomal pathway, the specific accumulation of CTFs in the late endosome and co-localization with iA␤ demonstrates that iA␤ is produced within this pathway. Therefore, parallel to ECE-1 cleavage of eA␤ prior to secretion, ECE-1b (although we do not exclude other isoforms) and ECE-2 appear to degrade A␤ at the late endosomal site of production. Furthermore, under conditions of impaired lysosomal flux, ECE activity helps to prevent excessive A␤ accumulation within AVs. Co-localization of ECE-2 with LC3B suggests that ECE-2 is at least partially responsible for A␤ clearance within the autophagic pathway. Our results demonstrate that efficient degradation of iA␤ occurs upstream of the lysosome.
In conclusion, we demonstrate that a large amount of A␤ is degraded presecretion, challenging the concept that all A␤ is constitutively secreted after production. We also show that inhibition of ECE activity reveals an elusive pool of iA␤ that is produced within the endosomal/lysosomal pathway but effi- FIGURE 10. Overexpression of ECE-2 does not influence eA␤ levels. A, transfection of CHO cells, which do not express endogenous ECEs, with ECE-1 isoforms a, b, c, or d, resulted in significant decreases in eA␤40 and eA␤42, which were mostly reversed by inhibition of the overexpressed enzyme with PA. In contrast, ECE-2 overexpression had little to no effect on extracellular accumulation of A␤. FEBRUARY 22, 2013 • VOLUME 288 • NUMBER 8 ciently degraded by ECEs under normal conditions. The possible physiological relevance of this iA␤ pool remains to be established. Considering the results, we propose that intraneuronal A␤ in AD does not originate from internalization of secreted peptide but from impairments in ECE activity. As alterations in the endosomal/lysosomal and autophagic pathways are observed early in the AD course and coincide with iA␤ accumulation (6,14,47), they could represent an underlying cause of ECE dysfunction and increase in iA␤.