Major Carboxyl terminal fragments generated by Gamma-secretase processing of the Alzheimer’s amyloid precursor are 50 and 51 aa in length

Abstract

Objectives

To understand the cleavage of the amyloid β protein (Aβ) precursor by γ-secretase and determine its changes in a representative Familial (FAD) Alzheimer’s disease (AD) mutation.

Methods

Transfected cells expressing wild type and FAD mutant APP were analyzed for changes in the levels of the major secreted Aβ species and of the corresponding intracellular C-terminal APP fragments (APP intracellular domain, AICD) generated by γ-secretase whereas radiosequencing was used to precisely identify the resulting cleavage site(s).

Results

The AICD fragment(s) generated by γ-secretase cleavage co-migrated in gels with a 50 residue synthetic peptide used as control being, therefore, smaller than the expected 59 and 57 residues based on the classical Aβ size ending at positions 40 (Aβ40) and 42 (Aβ42), respectively. In agreement with previous findings, an FAD mutant form of presenilin 1 (PS1-M139V) significantly increased the longer Aβ42 while showing trends towards reducing Aβ40. AICD levels were reduced by the mutation suggesting that γ-secretase activity may be actually impaired by the mutation. Radiosequence analysis in cells expressing wild type PS1 detected γ-secretase cleavage sites at the Aβ peptide bond L⁴⁹-V⁵⁰ to generate a 50 aa AICD fragment (AICD50) and the Aβ peptide bond T⁴⁸-L⁴⁹, generating an AICD of 51 aa (AICD51). No other cleavage sites were reliably detected.

Conclusions

Based on findings that the FAD mutation that increases Aβ42 also reduces AICD, we propose that γ-secretase activity is impaired by FAD mutations and predict that physiological and environmental agents that inhibit γ-secretase will actually induce AD pathogenesis rather that preventing it. Furthermore, we propose that the cleavage site to generate AICD is naturally ragged and occurs predominantly at two sites 48 and 49 aa from the start of the Aβ sequence. Thus, end specific antibodies to these two sites will need to be generated to study the quantitative relationships between these two cleavages in sporadic AD and FAD.

Introduction

APP is a type I integral membrane protein with a short half-life due to its cleavage in the juxtamembrane region by an enzyme named α-secretase ^1–3. This cleavage yields a secreted derivative, sAPPα (612–687 aa), which includes the first 16 aa of the Aβ sequence and a membrane-bound fragment, CTFα (83 aa), which is further cleaved by in the intramembrane protease, γ-secretase, to yield a 3 kDa fragment (P3 or Aα) starting at Aβ position 17 ⁴. In contrast, BACE1 cleaves APP on the N-terminal side of Aβ (β site) to generate the secreted derivative, sAPPβ (596–671 aa), and a membrane-bound fragment, CTFβ (99 aa), which is cleaved to Aβ of 4 kDa by γ-secretase ⁴. While most (>90%) secreted Aβ is 40 aa long, the accumulated deposit in the AD brain is mostly 42/43 aa long and these longer forms selectively increase in FAD mutations on APP or presenilins 1 (PS1) or 2 (PS2). A large body of literature has demonstrated that PS1 and PS2 are active subunits of γ-secretase, which is a complex of four integral membrane proteins: PS1/2, anterior pharynx 1 (Aph1), presenilin enhance 2 (Pen2) and Nicastrin (NCT) that can be coexpressed in yeast to generate the active enzyme ^{5, 6}. Aβ42 tends to form multimeric complexes more readily than Aβ40 and some of the oligomeric intermediates are neurotoxic in culture, supporting a hypothesis that a neurodegenerative cascade triggered by Aβ aggregates (amyloid hypothesis) causes AD ^{7, 8}. The longer Aβ42 forms increases in FAD and provides the strongest support for the amyloid hypothesis. Approximately eight years after the discovery of Aβ, one group isolated the APP intracellular fragment, AICD, from brain ⁹. At the same time, we developed an assay to show that γ-secretase cleavage yields substantial quantities of a cytoplasmic APP-CTF and named it CTFγ (a.k.a. AICD, Cγ, CTFε, AID) that is lost in cells due to rapid degradation ^{10, 11}. Our study triggered several follow up studies that identified the AICD cleavage site as the junction of Aβ49–50 and the fragment size as 50 aa and called it the ε site ^12–14. One group studied the membrane Aβ fragment more extensively and demonstrate the presence of substantial quantities of Aβ46 and called it the ζ site ¹⁵. Other studies described a novel serial cleavage pathway where the initial γ-secretase cleavage at Aβ49 generates additional cleavages in 3 aa intervals to generate Aβ of 38–43 aa ^{15, 16}. Detection of an alternative minor cleavage site at Aβ48 led to the hypothesis that Aβ48 generates Aβ45 and Aβ42 while Aβ49 is processed to Aβ46, Aβ43 and Aβ40, suggesting that the misprocessing occurs early in the sequential γ-secretase cleavage ^{17, 18}. However, these studies failed to detect any Aβ49 and suggest that Aβ49 is rapidly converted to Aβ46 but Aβ48 is more stable and give rise to lower levels of Aβ45 and Aβ42 ^17–19. Moreover, it was proposed that Aβ46 couldn’t give rise to Aβ42, based on inhibitor and kinetic analyses ¹⁷. However, these studies are contradicted by site-directed mutagenesis studies providing strong evidence that Aβ46 can generate Aβ40, Aβ42 and Aβ38 ^{16, 20}. As discussed earlier, mass spectroscopy studies showed that ε cleavage by γ-secretases generates mostly AICD50 and small quantities of AICD51 ²¹. Since matrix effects and differential ability to quantitatively detect fragments are known to affect the interpretation of mass spectroscopy data, we chose to detect the exact cleavage site of AICD by radiosequencing ^{22, 23}. Our data confirm that primary cleavage of APP by γ-secretase to generate AICD is predominantly at residues Aβ48 and Aβ49 providing independent evidence that there is no hint of cleavage at Aβ40 or Aβ42. We also describe a second cleavage that generates AICD51 and provide evidence that these are the only two major cleavage sites for AICD production. Further, we previously showed that inhibition of γ-secretase by an FAD mutation increases Aβ42, and provides an alternate model for the increase in Aβ42 generation based on the reduced rate of turnover of its immediate precursors using antisense PS1 RNA ²⁴. These studies support a hypothesis that inhibition of γ secretase may be a cause of AD, a problem that needs to be evaluated in typical late onset AD as independently proposed by Jie Shen and us previously ^{25, 26}.

Materials and Methods

Cells and reagents

Chinese hamster ovary (CHO) transfected with the 695 aa neuronal variant of wild type APP (2B7) has been described previously ²⁷. Transient transfections of wild type and mutant PS1 was carried out using HEK293H (Invitrogen (Carlsbad, CA), which is very efficiently transfected by lipofectamine 2000 (Invitrogen, Carlsbad, CA) as described previously (29). Radioactive ³⁵S methionine was purchased from Perkin Elmer-New England Nuclear (Massachusetts, USA). Antibodies O443, BAN50, BCO5 and BA27 against the C-terminal 20 aa of APP, 1–16 aa of Aβ, Aβ42 end and Aβ40 end were described previously. Peptide standards were from rPeptide (Augusta, GA) for Aβ and Calbiochem (EMD, San Diego, CA) for AICD fragments of 50 and 57 aa. Mutant APP and presenilin (PS)-1 constructs were described previously ^{24, 28}. Precast Bis-Tris gels and related reagents were obtained from Invitrogen (Carlsbad, CA). Other reagents and cell culture media and assay kits were obtained from either VWR or Fisher Scientific as described previously.

Western blot analysis for AICD

CHO-2B7 cells were cultured in 100 mm dishes in Opti-MEM (Invitrogen, Carlsbad, CA) containing 10% fetal bovine serum to ~ 70% saturation. The cells from two plates were scraped with a rubber policeman and resuspended in ice cold 50 mM HEPES, 0.15M NaCl, pH 7.0 containing the Roche protease inhibitor cocktail (HBS) and differentially centrifuged at 2500 rpm (600 × g) for 10 min and the supernatant (S1) was centrifuged at 10,000 rpm (12,000 × g) for 15 min to collect the pellet (P2). The pellet was resuspended in ice cold HBS, washed and resuspended in the same buffer. The washed P2 was incubated at 37°C for 2 h and centrifuged once again at 14,000 rpm (20,000 × g) for 15 min to remove the membrane and recover the supernatant containing AICD as described previously ¹¹. Supernatants were separated on precast 10% Bis-Tris gels and electrophoresed using the MES buffer system to ensure separation of low molecular weight fragments in the 4–6 kDa range. The gels were transferred onto nitrocellulose membranes (VWR) in a transfer buffer consisting of 25 mM Tris, 191 mM glycine and 20% methanol using a Bio-Rad wet transfer apparatus for 90 min at 90 V. The blots were dried overnight, washed extensively in distilled water for 15 min, soaked in boiling phosphate buffered saline (Fisher Scientific) for 10 min, rinsed in distilled water and blocked for 1–2 h in 10% newborn bovine serum (NBS) in 50 mM Tris pH 8.0, 0.15 M Nacl (TBS) containing 0.05% Tween-20 (TBST). We find that this step improves signal of nearly all Western blots and therefore extends beyond antigen retrieval. The blocked membranes were washed extensively washed with TBST (6 × 10 min) before incubation with O443 (10,000 fold dilution) in TBS containing 2% NBS for 1–2 h. Following extensive wash in TBST as described earlier, the blots were probed with horseradish peroxidase-labeled goat anti-rabbit antibody (Jackson Immunoresearch, 1:7500) for 1 h and washed extensively again in TBST. The blots were then incubated in super signal west pico chemiluminescent substrate (Fisher Scientific) and exposed to x-ray film to capture the signal. The blots were scanned using an Umax scanner with transparency attachment and quantified by the Molecular Dynamics Storm software-image quant. The procedure detects as little as 0.1 fmole of synthetic AICD-C50 and AICD-C57 peptides.

Radiolabeling and enzyme assay

Ten dishes of 100mm were seeded with CHO-2B7 cultures and grown to 70% confluence. Cells were washed once with Hank’s balanced salt solution (Gibco), equilibrated in methionine-deficient medium (Gibco) for 30 min and subjected to metabolic labeling for 16 h in methionine-deficient medium containing 1 mCi of ³⁵S methionine (ICN), as previously reported ²². AICDs were generated as described above and immunoprecipitated with the O443 antibody against the 20 C-terminal residues of APP, separated on 10% Bis-Tris gels with MES buffer and transferred to Immobilon P (Millipore) using a Bio-Rad (Hercules, CA) Western blotting apparatus, as described by the manufacturer. The membrane was exposed to X-ray film overnight and the radioactive band was identified, marked and excised along with surrounding regions for Edman degradation using a protein sequencer (Applied Biosystems; Foster City, CA). The released amino acid from each sequencing cycle was recovered and ³⁵S Methionine radioactivity assessed in a Beckman LS-250 liquid scintillation counter.

Statistical Analysis

Aβ and AICD from wild type and mutant PS1-expressing cells were compared by paired two tailed Student’s t test using GraphPad Prism version 6.

Results

Most AICD released comigrates with C50 rather than C57

CHO-2B7 cells membranes were fractionated to obtain the P2 fragment and incubated in vitro for γ-secretase to release the cytoplasmic C-terminal domain, AICD or CTFγ. Western blot analysis of the supernatants from the in vitro assay was carried out as described in Materials and Methods. It shows that most of the C-terminal fragment obtained in the in vitro γ-secretase assay comigrates with C50 suggesting that it is the major peptide generated by γ-secretase cleavage (Fig 1). A high level of MG132 (50 and 200 nM), a proteasome inhibitor that also inhibits γ-secretase ²⁹, reduces production of the AICD fragment (Fig 1). These studies indicate that little, if any AICD of 57 or 59 residues that complements Aβ40 or Aβ42 is generated by γ-secretase cleavage.

Figure 1.

AICD derived from CHO-2B7 cells migrate with the APP-C50 synthetic peptide.

CHO-2B7 cells were cultured, membranes isolated for an in vitro γ-secretase assay, separated on Tris-Tricine 10–20% gels (Life Technologies) along with synthetic APP-C50 and APP-C57 standards and analyzed by Western blotting with the O443 antibody as described in Materials and Methods and reported previously ¹¹. MG132 was used to inhibit γ-secretase cleavage as described previously. 15 and 10 kDa markers are indicated and AICD migrates at ~ 7 kDa.

Radiosequencing identifies two cleavage sites for AICD generation

Although the Western blot data in Fig 1 are definitive in not showing larger AICD fragment, a problem with this approach is that the method has a limited resolution in its ability to identify the exact cleavage sites and sizing on gel can be affected by conformation and post-translational modification. Therefore, we used Edman degradation to radio-sequence the AICD fragment generated from ³⁵S-methionine-labeled 2B7 cells. The products of Edman degradation were analyzed by scintillation counting; the number of cycles necessary to detect the radioactive methionine indicates the length of the peptide that extends towards the N-terminus of the radioactive residue. Fig 2 and Table 1 show the distribution of the radioactive amino acid peaks obtained in the first 12 cycles of Edman degradation analysis in a representative experiment. Two radioactive peaks were detected at cycles 2 and 3 whereas the rest of the cycles consistently rendered background levels of radioactivity, indicating the presence of two fragments –one containing methionine at position 2 and the other at position 3–identifying the cleavage sites at Aβ peptide bonds L⁴⁹-V⁵⁰ and T⁴⁸-L⁴⁹ and the generation of AICD50 and AICD51 (Fig 3). No fragments complementing Aβ40 (AICD59) or Aβ42 (AICD57) were detected.

Figure 2.

Distribution of ³⁵S-methionine by cycles of Edman degradation.

CHO-2B7 cells were cultured and radiolabeled and AICD generated from the membranes were immunoprecipitated with O443 as described in Materials and Methods. The quantity of ³⁵S-methionine released in each cycle after Edman degradation was determined by scintillation counting and the percentage of counts per minute versus total were plotted as described in Materials and Methods.

Table 1.

Yield of ³⁵S-methionine from AICD radiosequencing indicates that the cleavage occurs after Aβ48 and Aβ49

Cycle	% yield	Predicted AICD	Predicted Aβ
1	3.4	C49	AB50
2	55.0	C50	AB49
3	37.8	C51	AB48
4	0.0	C52	AB47
5	3.4	C53	AB46
6	0.0	C54	AB45
7	−0.7	C55	AB44
8	−0.7	C56	AB43
9	1.7	C57	AB42
10	0.0	C58	AB41
11	0.0	C59	AB40
12	0.0	C60	AB39

Figure 3.

Target sequence for γ-secretase cleavage showing the major cleavage sites that generate AICD.

The sequence of the C-terminal APP 101 residues starting 2 aa before Aβ to Aβ99 is shown. The known cleavage sites are indicated by “|” and major cleavage sites identified by Edman degradation analysis are shown below the target sequence and labeled AICD50 and AICD51. The reference radioactive methionine detected by radiosequencing is indicated in bold by an asterisk M*.

These studies confirm previous mass spectroscopy findings regarding γ-secretase cleavage generating AICD fragments of 50 and 51 aa ¹⁸. The study detects no AICD fragment other than the 50 and 51 aa forms suggesting that we need to study the biological effects of these two forms instead of using AICD57 and AICD59 predicted from the Aβ40 and Aβ42 cleavage sites.

An FAD mutant PS1 reduces AICD levels

PS1 M139V is an FAD mutation that increases the levels of Aβ42 and represents an example of the major hypothesis in the AD field that toxic amyloid oligomers are responsible for the characteristic neurodegeneration and dementia ³⁰. To determine whether AICD is affected by this FAD mutation, HEK293H cells were transiently transfected with either APPwt + PAG3 vector or APPwt and PS1-M139V in PAG3. Both these transiently transfected cells showed detectable Aβ40 and Aβ42 levels by a standard ELISA assays (Fig 4). The data (Fig 4, Left panel) show that the relative Aβ42 levels increase in cells transfected with the FAD mutant PS1 as reported by several groups previously ³¹. In addition, the studies show a significant drop in the levels of AICD levels in the cells expressing mutant PS1 (Fig 4, Right panel). However, there was no significant change in Aβ40 levels. These results suggest that FAD mutations may lead to partial inhibition of γ-secretase rather than modify its specificity and thereby act as dominant-negative mutations.

Figure 4.

FAD mutant PS1 M139V increases Aβ42 and reduces AICD levels.

APP and either wild type or mutant PS1 were transiently transfected into HEK293H cells or CHO cells and the secreted Aβ40 (pM) and Aβ42 (pM) detected by a standard sandwich ELISA assay with means and standard deviations plotted (Left panel). Membranes from the cells were isolated and assayed for changes in AICD levels by Western blotting quantified by densitometry showing means and standard deviations. Unpaired two-tailed Student’s t test for Aβ42 (t=3.496, df=6, P=0.0129*) and AICD (t=2.789, df=4, P=0.0494*) showed significant differences between wild type and mutant cells, but Aβ40 (t=1.748, df=6, P=0.131) did not show statistically significant difference.

Discussion

The finding that FAD mutations subtly alter APP processing to generate increased levels of Aβ42 has been a major foundation of the amyloid hypothesis and has been a widely repeated finding ⁷. However, only a few studies have examined the effects of PS1 mutations on AICD with varying results ^{21, 32}. In this study, we show that an FAD mutant form of PS1 increases Aβ42 and reduces AICD levels. The drop is significant (Fig 4) and suggests that FAD mutations can reduce γ-secretase activity, at least with respect to AICD generation. We have previously shown that antisense RNA against PS1 increases Aβ42 like some inhibitors of γ-secretase, which also leads to the same conclusion ²⁴. Furthermore, we have previously shown that overexpression of all γ-secretase subunits together in HEK293H cells increases Aβ and AICD, but actually reduces the Aβ42/Aβ40 ratio conversely showing that increasing γ-secretase reduces processing to the longer Aβ42 form and therefore reinforces the conclusion that the FAD-associated increase in Aβ42 may be due to reduction of γ-secretase ³³. However, it is important to note that AICD production may be independent of Aβ given that γ-secretase has two activities, the first named ε-cleavage cuts CTFβ to Aβ45–49 and the second γ-cleavage to generate secreted Aβ of 39-43 residues (Fig 5) ^{19, 21}. In addition to Aβ48/49, one study identified a critical intermediate, Aβ46, in Aβ40 and Aβ42 generation and named this as ζ cleavage without attempts to detect the corresponding AICD ³⁴. The current study suggests that there is no AICD corresponding in size to forms other than Aβ48/49, supporting the view that they are intermediates derived from further processing of these longer Aβ forms^{16, 19, 20, 34–36}. Given a sequential cleavage pattern from Aβ48/49 to Aβ45/46 to Aβ42/43 to Aβ39/40, a slow or impaired γ-secretase will likely release the longer Aβ42/43 forms into the media rather than convert it to the shorter 39/40 forms. We have also proposed that accumulation of the immediate substrate, CTFβ, may lead to increased release of Aβ in spurts, which may facilitate aggregation ³. However, if the proximal substrate for secreted Aβ42 is Aβ45–49, the phenomenon may fostered by accumulating these intermediates rather than CTFβ.. Indeed, we find that partial γ-secretase inhibition can yield very high amounts of Aβ40 and Aβ42 presumably by accumulation CTFβ and longer Aβ intermediates (Barnwell et al, manuscript in preparation).

Figure 5.

Model showing APP processing by γ-secretase at the γ- and ε- sites

APP is cleaved to CTFα and CTFβ, which are cleaved by γ-secretase at the ε site to release AICD of the indicated sizes with relative levels. The most common cleavage site confirmed in this study based on AICD length (Table 1) should give rise to Aβ49 but this fragment has not been reported, whereas considerable amounts of Aβ48 are efficiently detected ¹⁹. This study detects substantial amounts of AICD of 51 aa consistent with the reports showing Aβ48. The final step consists of γ-site cleavages, which occur at residues 42 and 43. Although Aβ43 is readily detected in lysates, unlike Aβ42, it is not detected in media ¹⁹. In contrast a second γ-site cleavage at residues 38–40 release secreted Aβ primarily of 40 residues. However, it is possible that most of these γ-site cleavages are independent of each other even though they may be dependent on the ε-site cleavage.

The model that inhibition of γ-secretase to explain the generation of differential amounts of Aβ40 and Aβ42 by simple enzyme kinetics rather than by invoking a change in enzyme specificity is more parsimonious and provides a simple kinetic mechanism to explain the three observed phenomena: 1. Multiple FAD mutations spanning the entire sequence of PS1 increase Aβ42/Aβ40 ³¹; 2. Treatment with antisense RNA against PS1 or γ-secretase inhibitors increases Aβ42/Aβ40 ratios ²⁴; and 3. Overexpression of all γ-secretase subunits reduces Aβ42/Aβ40 ratios ²⁸.

Another implication of our model is that γ-secretase inhibition may lead to increases in secreted Aβ levels simply by accumulating substrate (CTFβ and longer Aβ forms) and releasing them in spurts. This type of accumulation-release mechanism should result in a high local Aβ concentration and facilitate Aβ aggregation to its toxic oligomers ³⁷. Thus, even if Aβ oligomers are the only cause of AD-related neurodegeneration, one of the major triggering events may be the inhibition of γ-secretase. Since γ-secretase is an essential enzyme in the whole organism complete inhibition may lead to rapid death. Even complete inhibition limited to some cell populations and may lead to accumulation of high levels of membrane-bound fragments: CTFα and CTFβ, which may trigger inflammatory pathways to clear the dysfunctional cells as recently noted in the γ-secretase deficiency associated with acne inversa ^{25, 38}. Although partial γ-secretase inhibition may be less damaging, it can clearly affect critical processes such as endothelial tight junction barrier by facilitating vascular endothelial growth factor ^{39, 40}. Thus, partial γ-secretase inhibition may increase Aβ and be damaging by multiple mechanisms that include impairment of signaling by other regulatory substrates, inflammation in response to accumulation of misfolded proteins in the membrane, and impaired membrane function due to congestion in the protein trafficking pathways caused by accumulated peptides ^{26, 41}. Aβ is certainly one of the major accumulated peptides and could be the major toxin in initiating the disease process. However, Aβ40 and Aβ42 represent metabolites that are generated by normal processing of APP that is highly conserved in higher vertebrates. Under normal conditions, it is well tolerated and does not accumulate. The finding that FAD mutations inhibit γ-secretase but increase Aβ42 suggests that partial inhibition of γ-secretase can be the cause of AD pathogenesis ^{3, 26}. Prevention of AD must therefore involve preservation of γ-secretase rather than its inhibition.

Conclusion

Due to extensive efforts spanning behavior, imaging, neuropathology, genetics and toxicology, Aβ accumulation is recognized as an important trigger in dementia of the Alzheimer type. This study presents a slightly different twist to the hypothesis by presenting evidence that FAD mutations may actually impair γ-secretase, the final step in Aβ biogenesis, suggesting that even if Aβ toxicity is the trigger in AD, γ-secretase inhibition may be an incorrect approach for its treatment. In addition, the study presents details on intracellular γ-secretase cleavage products of APP and identifies fragments whose function and toxicity need to be evaluated with respect to their role in neurodegeneration and dementia.