|Home | About | Journals | Submit | Contact Us | Français|
γ-Secretase is a widely expressed multi-subunit enzyme complex which is involved in the pathogenesis of Alzheimer disease and hematopoietic malignancies through its aberrant processing of the amyloid precursor protein (APP) and Notch1, respectively. While γ-secretase has been extensively studied, there is a dearth of information surrounding the activity, composition, and function of γ-secretase expressed in distinct cellular populations. Here we show that endogenous γ-secretase complexes of hematopoietic origin are distinct from epithelial derived γ-secretase complexes. Hematopoietic γ-secretase has reduced activity for APP and Notch1 processing compared to epithelial γ-secretase. Characterization of the active complexes with small molecule affinity probes reveals that hematopoietic γ-secretase has an atypical subunit composition with significantly altered subunit stoichiometry. Furthermore, we demonstrate that these discrete complexes exhibit cell-line specific substrate selectivity suggesting a possible mechanism of substrate regulation. These data underscore the need for studying endogenous γ-secretase to fully understand of the biology of γ-secretase and its complexity as a molecular target for the development of disease therapeutics.
γ-Secretase processes a wide range of substrates including the amyloid precursor protein (APP), Notch, ErbB4, Nectin1α, N-cadherin and other type I transmembrane proteins, reflecting the multitude of physiological functions of this protease (1-2). Cleavage of APP by γ-secretase releases beta amyloid (Aβ) 40 and 42 peptides. Oligomerization of these peptides, specifically Aβ42, is thought to play a causative role in the pathogenesis of Alzheimer's disease (AD) (3-4). Notch1 is another well studied substrate of γ-secretase and it is known to play a pivotal role in lymphocyte development (5). Aberrant Notch1 signaling contributes to T cell acute lymphoblastic leukemia and is associated with a variety of cancers (6-10). γ-Secretase activity is associated with a multi-subunit transmembrane complex comprised of at least four known subunits: presenilin (PS), nicastrin (Nct), anterior pharynx defective (Aph), and presenilin enhancer 2 (Pen2) (11-12). PS is thought to contain the catalytic core of the enzyme as it has previously been shown that active site-directed inhibitors specifically label PS and mutation of either of the two transmembrane aspartates results in loss of γ-secretase activity (13-16). Aph and Nct have been shown to contribute to complex stability and substrate recognition, respectively, while Pen2 is necessary for the final maturation of γ-secretase (17-22). PS exists as two homologs, PS1 and PS2, which have redundant but non-overlapping functions and are found in mutually exclusive complexes (23-24). Additionally, Aph exists as three isoforms which have also been shown to be part of mutually exclusive complexes: Aph1aL, Aph1aS, and Aph1b (25-26). Recent data have suggested that there are six possible γ-secretase complexes consisting of Nct and Pen2 with either PS1 or PS2 and one of the three Aph isoforms (2 PS × 3 Aph) with a 1:1:1:1 stoichiometry (27-28). Recent data have also revealed that TMP21 and CD147 may be putative γ-secretase subunits but their contribution to enzyme activity and their role in the regulation of γ-secretase has not been fully elucidated (29-31).
Previous studies examining the activity and regulation of γ-secretase have utilized the exogenous expression of one or more of the components or substrates of γ-secretase in a limited number of cellular models. However, it is known that exogenous overexpression of PS1 can lead to the total replacement of endogenous PS1 along with the accumulation of full length of PS, which is not observed under endogenous conditions (32). Furthermore, we have previously demonstrated that γ-secretase complexes exist in a dynamic equilibrium and exogenous expression of individual subunits can cause perturbations to this equilibrium resulting in alteration of enzyme activity (33). Therefore, in order to elucidate the physiological function and regulation of γ-secretase, it is critical to characterize endogenous γ-secretase complexes isolated from different cellular sources and tissue systems.
In the present study we have used a small molecule approach to characterize γ-secretase complexes from diverse cellular origins. We reveal that endogenous γ-secretase derived from hematopoietic origins is distinct from classically studied epithelial derived γ-secretase. Hematopoietic derived γ-secretase displays an atypical subunit composition with an altered PS1-NTF:PS1-CTF ratio and a unique activity profile compared to epithelial γ-secretase.
Snap frozen log phase growth cell culture pellets of suspension adapted HeLa, suspension adapted 293T, Bjab, Jurkat, and HL60 was obtained from the National Cell Culture Center. Mono and polymorphonuclear white blood cells (WBC) were isolated and pooled from 1 liter of porcine whole blood (Pel-Freez) using Polymorphprep (Axis-Shield) as per the manufacturer's instructions. Following isolation, WBCs were washed twice with phosphate buffered saline and frozen at -80°C until membrane preparation.
L685,458 (L458), compound E, GSI34 and compound 5 was prepared and characterized as previously described (16, 34-35). Antibodies against PS1-NTF, recognizing the 1-25 amino acid residues of PS1, were a kind gift of Dr. Min-tain Lai (Merck Research Laboratories). Antibodies against PS2-CTF (PC235) and PS1-CTF (MAB5232 and AB5308) were purchased from Calbiochem and Chemicon, respectively. Antibodies against Nct and NICD1 (SM320) were generated in our lab. Antibody against Pen2 (ab18189) was purchased from Abcam.
Total membrane was isolated from the indicated cell lines as described (33, 36). Briefly, cell pellets were resuspended in Buffer A (50mM MES pH 6.0, 150mM KCl, 5mM CaCl2, 5mM MgCl2, and protease inhibitors) and lysed by passage through a French Press. After removal of nuclear debris by low speed centrifugation, the total membrane fraction was isolated by pelleting at 100,000g for 1h. A portion of total membranes was solubilized with 1% CHAPSO in Buffer A for 1h 4°C and non-solubilized material was removed by ultracentrifugation at 100,000g for 1h. Protein concentration was determined with the DC Protein Assay Kit (BioRad) as per manufacturer's instructions. In vitro γ-secretase activity was measured as described previously using ECL technology (36-37). Total membrane or CHAPSO solubilized membrane was incubated in Buffer B (50mM PIPES pH 7.0, 150mM KCl, 5mM CaCl2, 5mM MgCl2 and protease inhibitors) with 0.25% CHAPSO, biotinylated APP or Notch1 substrate for 2.5h 37°C (APP substrate) or 2h 37°C (Notch1 substrate). The reaction mix was incubated with ruthenylated G2-10 (Aβ40), ruthenylated G2-11 (Aβ42), or SM320 (NICD1) in conjunction with ruthenylated anti-rabbit in Buffer C (1× PBS with 0.5% (v/v) Tween-20) for 2h at 25°C and streptavidin magnetic beads (Dynel) were added. Aβ40, Aβ42 and NICD1 production was measured by ECL on an ECL reader and expressed as Relative Light Units (RLU).
CHAPSO solubilized membranes were captured as described previously (33). The captured complexes were eluted with 2× Lammeli sample buffer and separated by SDS-PAGE. The following antibodies were used for western blotting: PS1-NTF 1:1000, PS1-CTF 1:1000, Nct 1:1000 and PS2-CTF 1:1000. Anti-mouse or anti-rabbit HRP conjugated (Amersham) secondary antibodies were used in conjunction with standard ECL detection methods. PS1-NTF and PS1-CTF detection was done on a single blot without stripping; the blot was first probed with rabbit PS1-NTF, extensively washed with PBS-Tween, and then reprobed with mouse PS1-CTF. In all cases, blots shown are representative of three or more experiments.
Total membrane or CHAPSO solubilized membrane was photolabeled as previously described (14). Briefly, membrane samples were incubated in the presence or absence of 2μM L458 in Buffer B for 0.5h at 37°C. Compound 5 (20 nM) was added for an additional 1h at 37°C and the samples were crosslinked at 350nm on ice for 0.75h. Following crosslinking, the reaction was denatured with RIPA buffer (50mM Tris pH 8.0, 150mM NaCl, 0.1 % (w/v) SDS, 1% (v/v) NP-40, and 0.5% (w/v) deoxycholic acid) for 1h at 25°C. Streptavidin agarose was added and allowed to incubate overnight at 4°C. The labeled complexes were washed with RIPA buffer and eluted with 2× Lammeli sample buffer. Samples were western blotted as described above. In all cases, blots shown are representative of three or more experiments.
To begin to characterize endogenous γ-secretase complexes, we determined γ-secretase activity levels in a panel of human cell lines derived from epithelial (HeLa and HEK293T) or hematopoietic (Bjab, Jurkat and HL60) origins. In vitro γ-secretase activity was determined using total membrane or CHAPSO solubilized membrane fractions and recombinant APP or Notch1 substrates. The epithelial cell lines, HeLa and 293T, exhibited the highest specific γ-secretase activity for processing of APP. The rate of Aβ40 and Aβ42 production for HeLa and HEK293 cell membrane was 2-5 fold higher than the hematopoietic cell lines, Bjab, Jurkat, and HL60 (Figure 1a. right and middle panels, p value < 0.01). Since the Aβ40 and Aβ42 products were assayed using different antibodies, we compared their relative rates. The relative rate for Aβ40 and Aβ42 production was very similar for both total membrane and solubilized membrane γ-secretase activity, indicating that all cell lines exhibited a similar preference for the Aβ40 over the Aβ42 cleavage site of APP (Figure 1a, right panel). Similar to APP γ-secretase activity, HeLa and 293T had the highest levels Notch1 γ-secretase activity than Bjab, Jurkat, and HL60 (Figure 1b). The hematopoietic cell lines had between 5 and 50 fold less Notch1 γ-secretase activity (Figure 1b, p<0.01). To compare relative rate, we calculated the relative Aβ40:NICD1 rate ratio from total membrane for the various cell lines. HeLa, 293T and Bjab all had a similar Aβ40:NICD1 ratio of approximately 30 indicating that their γ-secretase complexes have similar substrate selectivity for APP compared to Notch1. Jurkat and HL60 had a Aβ40:NICD1 ratio of 81 and 184, respectively, indicating that γ-secretase complexes in these cell lines have altered substrate selectivity and strongly favor APP as a substrate compared to HeLa, 293T, and Bjab (Figure 1c).
The combined in vitro γ-secretase activity data suggest that epithelial and hematopoietic cell lines possess distinct complexes which exhibit differing levels of activity and substrate selectivity. We examined the protein levels of Nct, PS1-NTF/CTF, and PS2-CTF by western blotting to determine if a reduction in the core catalytic and substrate binding subunits could account for the altered levels of relative activity. An equal amount of either total membrane or solubilized membrane was separated by SDS-PAGE and western blotted for the indicated proteins. HeLa and 293T cell lines had slightly more Nct, PS1-NTF, and Pen2 compared to the hematopoietic cell lines Bjab, Jurkat, and HL60 for both total membrane and solubilized membrane fractions (Figure 2a, upper two panels and bottom panel, respectively). Strikingly, Bjab, Jurkat, and HL60 had significantly less PS1-CTF compared to HeLa and 293T for both total membrane and solubilized membrane fractions (Figure 2a, the third panel). We confirmed this phenomenon using two other PS1-CTF antibodies with different epitopes (data not shown). To ensure that the differences in PS1-NTF/CTF level between the cell lines was not do to an artifact created during membrane preparation or solubilization, we measured the expression of PS1-NTF and PS1-CTF in whole cell lysates of the five cell line (Figure 2b). Like the western blot data from membrane and solubilized membrane, the hematopoietic cell lines had reduced levels of PS1-CTF compared to the epithelial cell lines but roughly equal amounts of PS1-NTF. Therefore, the hematopoietic cell lines displayed a distinct PS1-NTF:CTF ratio compared to the epithelial cell lines. PS2-CTF expression did not correlate with cell line lineage as Nct and PS1-NTF/CTF expression did. The 293T cells had the highest expression of PS2-CTF, HeLa and Bjab had slightly less PS2-CTF than 293T and Jurkat and HL60 had undetectable levels of PS2-CTF (Figure 2a, the fourth panel
In order to ensure that the observed increase in the relative PS1-NTF:CTF ratio of the hematopoietic cell lines was truly reflective of a cell linage specific characteristic, we quantitated the amount of PS1-NTF and PS1-CTF in HeLa and Bjab CHAPSO solubilized membrane. Increasing amounts of protein (1.25 to 5 μg) of HeLa (Figure 2c, left panel) or Bjab (Figure 2c, right panel) were resolved by SDS-PAGE and western blotted for PS1-CTF and PS1-NTF to determine the dynamic range of detection. The relative band intensity was determined by measuring the area under the curve (AUC) using ImageJ quantification software and the linear range of detection for both PS1-NTF and PS1-CTF for HeLa (left panel) and Bjab (right panel) was determined to be between 2.5μg and 5μg of solubilized membrane (Figure 2c). We therefore resolved 3μg of solubilized membrane in quadruplicate by SDS-PAGE and determined the relative AUC for PS1-NTF and PS1-CTF for HeLa and Bjab (Figure 2d). Solubilized membrane from HeLa had approximately 4 fold more PS1-NTF than from Bjab but approximately 15 fold more PS1-CTF. These data confirmed that Bjab exhibited a significantly different PS1-NTF:CTF ratio compared to HeLa (p value <0.001).
Because only a small percentage of expressed PS is engaged in active γ-secretase complexes, western blotting alone does not distinguish between active and inactive γ-secretase complexes (23, 33, 38). Therefore, it is possible that the increased relative PS1-NTF:CTF ratio observed for the hematopoietic cell lines does not reflect the actual ratio in the active complex. We recently developed a second generation active site affinity probe, compound 5 (Figure 3a), which has been used to examine mouse brain endogenous active γ-secretase complexes (35). We measured the ability of compound 5 to inhibit in vitro γ-secretase activity in solubilized membrane fractions using a recombinant APP substrate; compound 5 had equal potency in all five cell lines. Compound 5 was incubated with solubilized membrane fraction in the presence or absence of L458 and the bound complexes were isolated with strepavidin agarose under native conditions, resolved by SDS-PAGE, and western blotted for the indicated proteins. In all cases, capture by compound 5 was blocked by inclusion of excess L458 indicating that the capture was specific (Figure 3b). Similar amounts of both Nct and PS1-NTF were captured with compound 5 from all five cell lines (Figure 3b, top two panels). On the contrary, significantly less PS1-CTF was captured by compound 5 from Bjab, Jurkat, and HL60 in comparison to HeLa and 293T (Figure 3b). Overall this caused a 2-20 fold difference in the relative PS1-NTF:CTF ratio for the hematopoietic cell lines compared to the epithelial cell lines. These data parallel the western blotting data and confirm that active hematopoietic γ-secretase complexes contain a distinctive PS1-NTF:PS1-CTF ratio compared with the epithelial protease.
In order to confirm that this difference was truly refect the ratio of PS1-NTF and PS1-CTF within the active γ-secretase complex in cell membranes, we analyzed photolabeled species using compound 5. Compound 5 also contains a photoreactive benzophenone group which can be used to covalently label the active site of γ-secretase (Figure 3a, grey highlight), which has the same core structure of L852,505 (14). Our recent studies indicated that L852,505 also label PS1-NTF when the labeled proteins were detected with anti-PS1 antibody against the 1-25 amino acid residues of PS1 (39), which was not recognizable using an antibody against the 1-14 amino acid residues of PS1 (14). Compound 5 was incubated with either solubilized membrane or total membrane in the presence or absence of excess L458. After irradiation, photolabeled proteins were solubilized, captured, eluted and analyzed by western analysis. Compound 5 was able to label roughly equal amounts of PS1-NTF from both solubilized and total membrane fractions of all five cell lines (Figure 3c). Labeling was completely blocked by inclusion of excess non-biotinylated L458, indicating that the labeling was specific (Figure 3c). Much less PS1-CTF was photolabeled in Bjab, Jurkat, and HL60 compared to HeLa and 293T (Figure 3c, bottom panel) for both solubilized membrane and, even more strikingly, total membrane. In order to reduce the variation between different Western blot analyses, we determined the relative ratio PS1-NTF:PS1-CTF in each cell line and then normalized them against HeLa membrane (i.e. the normalized ratio for HeLa membranes equals 1.) This conversion allows direct comparison of blots from separate studies. Consequently, we averaged normalized ratios from Fig. 3b and 3c and found the ratio in Bjab, Jurkat, and HL60 is significantly higher than in HeLa and 293T (Figure 3d). These analyses rule out that the differences between these cells resulted from variations of CHAPSO solubilization and support our finding that hematopoietic cell lines contain an atypical γ-secretase complex which has an altered relative PS1-NTF:CTF ratio. Moreover, we found that a positive correlation (R = 0.97) exists between normalized NTF:CTF ratio and Aβ40:NICD1 ratio (Figure 3e).
To further confirm whether an atypical γ-secretase exists in hematopoietic cells, we carried out the same study using white blood cells (WBC) isolated from whole porcine blood. Mononuclear and polymorphonuclear leucocytes were separated from anticoagulated porcine blood, pooled, and the total membrane fraction was isolated. Porcine WBC γ-secretase exhibited a rate 3,674 Aβ40 units/min/μg and 1,225 NICD1 units/min/μg total membrane. Porcine WBC Aβ40 activity could be inhibited by L458 in a dose dependant manner (Figure 4a). Western blotting revealed that WBC had less Nct and PS1-NTF when compared to HeLa total membrane, consistent with the lower in vitro activity levels (Figure 4b, top two panels). Interestingly, porcine WBC PS1-NTF was slightly larger than HeLa PS1-NTF, likely reflecting species specific processing of PS. Like the other hematopoietic cell lines examined, WBC had very low expression of PS1-CTF; this was confirmed with two different antibodies (Figure 5b, panels 3 and 4). Therefore, we conclude that porcine WBC, like the human hematopoietic cell lines, has much less PS1-CTF relative to PS1-NTF.
To confirm that the western blotting data reflects the composition of active γ-secretase complexes, we used compound 5 to covalently photolabel PS1-NTF and PS1-CTF from total membrane. While compound 5 was clearly able to photolabel WBC PS1-NTF, labeling of WBC PS1-CTF was undetectable whereas equal amounts of HeLa PS1-NTF and PS1-CTF were labeled (Figure 4c). These results are identical to the compound 5 photolabeling of total membrane from Bjab, Jurkat, and HL60 (Figure 4, right panel) and further confirm that cells of hematopoietic origin have an atypical active γ-secretase complex which exhibits a greatly increased relative PS1-NTF:CTF ratio.
We have clearly established that cells of hematopoietic origin contain distinct γ-secretase complexes compared to the classically studied epithelial derived γ-secretase; hematopoietic γ-secretase has lower overall activity, differing relative substrate selectivity, and an atypical subunit composition. We wanted to determine if hematopoietic γ-secretase had a unique pharmacological profile which could be used to distinguish it from epithelial γ-secretase. In vitro APP γ-secretase activity in solubilized membrane was measured in the presence of increasing concentrations of three structurally diverse small molecule γ-secretase inhibitors: a peptidiomimetic (L458), a sulfonamide (GSI34), and a benzodiazepine (Compound E). There was no difference between the cell lines in the potency of the small molecule inhibitors screened indicating that these small molecules cannot distinguish between the two classes of γ-secretase (Figure 5).
The existence of discrete γ-secretase complexes has long been inferred from biochemical and expression data (25, 40-42). However, the findings described here are the first which demonstrate the existence of unique complexes using purely endogenous sources. We have clearly shown that hematopoietic γ-secretase is distinct from the classically studied epithelial γ-secretase in both activity and subunit composition; hematopoietic γ-secretase has lower in vitro activity, displays unique substrate selectivity, and has an atypical PS1-NTF:CTF stoichiometry incorporated into active complexes compared to epithelial γ-secretase.
Previous studies have suggested that the core γ-secretase subunits are arrayed with a 1:1:1:1:1 stoichiometry (27-28, 43). Our lab has previously shown that the complex stoichiometry is dynamic and can be altered by exogenous expression of γ-secretase subunits or PS FAD mutants (33). Our work here further supports this dynamic nature of the γ-secretase complex stoichimoetric arrangement and, importantly, does so using purely endogenous cell models. While we can not conclude the precise amount each subunit present, it is clear that the relative ratio of PS1-NTF to PS1-CTF is clearly different between hematopoietic and epithelial γ-secretase. The finding that the relative PS1-NTF:CTF ratio is greatly altered in hematopoietic γ-secretase is surprising as γ-secretase is a diaspartyl protease with both catalytic aspartates required for activity (13). The γ-secretase complex is formed in a tightly regulated stepwise fashion and subunit stability is dependent on complex formation: an initial sub-complex is formed between Nct and Aph, followed by the incorporation of full length holo-PS, and finally incorporation of Pen2 (11, 17, 44-46). The incorporation of Pen2 promotes the endoproteolysis of holo-PS into its stably associated PS-NTF and PS-CTF fragments; this endoproteolysis converts an inactive complex into a catalytically competent enzyme, partitions the catalytic aspartates into separate fragments, and is required for γ-secretase activity (12, 47-50). It is likely that catalytic activity of any γ-secretase complex is driven by a 1:1 array of PS1-NTF:PS1-CTF which maintains the catalytic diaspartyl stoichiometry. Therefore, the reduced catalytic activity seen in hematopoietic γ-secretase compared to epithelial γ-secretase can be accounted for by a reduction in the amount of PS1-CTF which would decrease the number of 1:1 NTF:CTF catalytic units in the active complexes. The unpaired hematopoietic PS1-NTF units may contribute the altered substrate selectivity seen for Jurkat and HL60 γ-secretase as these cell lines exhibited the most robust increase in relative PS1-NTF:CTF ratio and altered substrate selectivity in comparison to the other cell lines examined; indeed our data demonstrated that the relative ratio of NTF:CTF does in fact correlates with the relative rate ratios of γ-seccretase for Aβ40 and NICD1 production (see Figure 3e). However, the precise mechanism of the NTF:CTF ratio and its correlation with substrate specificity as well as the exact physiological function of the catalytically unpaired PS1-NTF in hematopoietic γ-secretase requires further investigation.
In addition, these data also indicate that the levels of PS1-NTF and CTF in hematopoietic γ-secretase appear to be independently regulated. There have been reports which have shown that it is in fact possible for PS1-CTF expression levels to vary independently of PS1-NTF levels. During zebrafish embryogenesis, PS1-CTF levels were shown to increase relative to holo-PS levels, suggesting selective regulation of PS1-CTF (51). Selective phosphorylation of PS1-CTF has also been shown to play a role in regulation of PS1-CTF levels. Kirschenbaum et al. previously demonstrated that PS1-CTF levels could be reduced independently of PS1-NTF levels through phosphorylation by glycogen synthase kinase-3β (52). Likewise, overexpression of cyclin-dependent kinase-5/p35 has been shown to lead to the selective stabilization of phosphorylated PS1-CTF in 293T cells (53). Further investigation is required to clarify how such an atypical γ-secretase complex is assembled and/or regulated in cells of hematopoietic origin.
γ-Secretase is known to process a wide range of type I transmembrane proteins with limited primary sequence homology (1, 54). The exact mechanism by which γ-secretase achieves substrate regulation is unknown. It is intriguing to note that the cell line specific complexes identified here exhibit unique patterns of relative substrate selectivity for the processing of APP and Notch1. These data, therefore, suggest that distinct γ-secretase complexes with preferential substrate processing abilities may be one potential mechanism of substrate regulation. The physiological substrates and function of these discrete γ-secretase complexes remains to be investigated.
In sum, this work further supports the existence of multiple and distinct endogenous γ-secretase complexes and underscores the complexity of γ-secretase as a molecular target for the development of therapeutics for the treatment of Alzheimer disease and other disease states.
We thank Dr. Min-tain Lai for PS1 NTF antibody.
This work is supported by NIH grant R01-AG026660 (YML) and Alzheimer Association (YML). LP and JWC were supported by Institutional Training Grant T32 GM073546.
Completing Interests Statement: The authors declare that have no competing financial interests.