Gamma-secretase is a membrane protein complex comprised of presenilin, nicastrin, Aph-1, and Pen-2

Abstract

gamma-Secretase catalyzes the intramembrane proteolysis of Notch, beta-amyloid precursor protein, and other substrates as part of a new signaling paradigm and as a key step in the pathogenesis of Alzheimer's disease. This unusual protease has eluded identification, though evidence suggests that the presenilin heterodimer comprises the catalytic site and that a highly glycosylated form of nicastrin associates with it. The formation of presenilin heterodimers from the holoprotein is tightly gated by unknown limiting cellular factors. Here we show that Aph-1 and Pen-2, two recently identified membrane proteins genetically linked to gamma-secretase, associate directly with presenilin and nicastrin in the active protease complex. Coexpression of all four proteins leads to marked increases in presenilin heterodimers, full glycosylation of nicastrin, and enhanced gamma-secretase activity. These findings suggest that the four membrane proteins comprise the limiting components of gamma-secretase and coassemble to form the active enzyme in mammalian cells.

Fig. 1.

Aph1α2 is associated with the γ-secretase complex. (a) Schematic of the four candidate members of the γ-secretase complex. PS1 contains two intramembrane aspartates (stars) that constitute the probable active site. The membrane orientations of Aph-1 and Pen-2 are not yet verified. (b) Aph1α2-HA was transfected into COS7 cells (Tfxn) or generated by in vitro transcription/translation (IVT/T) in rabbit reticulocyte lysate. The highest band in the COS7 lysate comigrates with the IVT/T product. (c) Affinity precipitation of Aph1α2 by III-31C but not III-112 resin. Start lysate identical to that in b was subjected to batch mode pull-downs with each resin. Bound proteins (ppt) were blotted with R302 to NCT, Ab14 to PS1 NTF, 13A11 to PS1 CTF, and 3F10 to HA of Aph-1α2-HA. (d) One percent digitonin-solubilized membranes were fractionated on an 18–28% glycerol gradient. Fractions were blotted for NCT, PS1, and Aph-1α2 with antibodies as in c. The γ-secretase complex in COS7 reproducibly migrates into denser gradient fractions than do the complexes in HEK293, CHO, or HeLa membranes (see Fig. 4c, and data not shown). Although the reason for this consistent difference is unknown, we note that the γ-secretase complex also migrates differently in CHAPSO than in digitonin detergent gradients. (e) Fractions 2, 6, and 10 from the glycerol gradient were subjected to co-IP with 3F10 antibody (Aph1α2-HA) or R302 (NCT). Beads were exchanged into 0.25% CHAPSO/Hepes, adjusted to contain 0.0125% phosphatidylethanolamine and 0.1% phosphati-dylcholine, and incubated with C100Flag without or with 31C inhibitor. AICD-Flag was detected with M2 Flag antibody and Aβ with 6E10.

Fig. 3.

PS1, NCT, Aph-1α2, and Pen-2 are sufficient to generate more γ-secretase in mammalian cells. (a) CHO cells were transiently transfected with the indicated constructs. Equal amounts of total protein were blotted for NCT (V5), PS1 holoprotein and NTF (M2 Flag), PS1 CTF (13A11; which recognizes both endogenous and exogenous CTF), Aph-1α2 (3F10), and Pen-2 (M2 Flag). Note the increases in mature NCT and PS1 heterodimers when all four proteins are coexpressed (lanes 11 and 12). Exogenous human PS1 CTF migrates slightly higher than endogenous hamster CTF. (b) Stable CHO cell lines expressing various combinations of human PS1, Aph-1α2, and/or Pen-2. Note the nearly equal abundance of immature and mature forms of NCT in the parental PS70 cells (lanes 1 and 2); this obviates the need to overexpress NCT. There is a slight increase in the immature, ≈110-kDa NCT in the A-5 stable line, suggesting that excess Aph-1α2 may stabilize immature NCT. However, only when PS1, Aph-1α2, and Pen-2 are all overexpressed is there a marked increase in mature NCT and PS1 NTF and CTF. PS70, CHO cells stably expressing PS1; A-5, PS70 cells with Aph-1α2-HA; P-2 and P-22, PS70 cells with Flag-Pen2; γ-23 and γ-30, A-5 cells with Flag-Pen2. Antibodies: R302 (NCT), Ab14 (PS1 holoprotein and NTF), 13A11 (PS1 CTF), 3F10 (Aph-1α2), and M2 Flag (Pen-2). (c) Direct assessment of γ-secretase activity in membrane lysates of the indicated stable cell lines. Membranes from equal numbers of cells were solubilized in 1% CHAPSO, and a C100Flag activity assay was performed (20). Samples were probed for the AICD-Flag (Upper) and Aβ (Lower) products, as in Fig. 2b.

Fig. 2.

Pen-2 is associated with the γ-secretase complex. (a) Affinity precipitation of Pen-2. Pen-2 is pulled down by III-31C resin but not the control III-112 resin. (b) M2 Flag resin was used to co-IP Pen-2 (lanes 1–3), which was then incubated as in Fig. 1e. A Flag-tagged PS1 construct was transfected and analyzed in parallel as a positive control (lanes 4–6).

Fig. 4.

NCT, PS 1, Aph1α2, and Pen-2 remain quantitatively associated after a partial purification of γ-secretase. (a) Digitonin lysate from γ-30 cells was run on a blue native-PAGE gel. Identically run strips of membrane were probed with antibodies as in Fig. 3b. A band containing all four proteins (mature complex) runs at ≈250 kDa. We observed lower migrating bands that contained some, but not all, of the proteins; these may represent intermediates in the assembly of γ-secretase. (b) Membranes from the indicated stable cell lines were solubilized in 1% CHAPSO and subjected to co-IPs. Immunoprecipitations were then probed with antibodies to NCT, PS NTF, PS CTF, Aph1α2, and Pen-2 (as in Fig. 3b). Asterisk, cross-reactive band (because of immunoprecipitation antibody light chains). (c) Digitonin-solubilized γ-30 cell membranes were fractionated on an 18–28% glycerol gradient. γ-Secretase-containing fractions were then pooled and injected over a 31C affinity column. The flow through, wash, and Brij-35 eluate fractions were probed for NCT, PS1 NTF and CTF, Aph-1α2, and Pen-2 as in Fig. 3b. (d) The elute fractions from the 31C column were pooled and subjected to co-IP with X81 (to PS1), R302 (to NCT), 3F10 (to Aph-1α2-HA), and M2 Flag (to Flag-Pen-2). The beads were washed and incubated in an in vitro C100 activity assay. Aβ was detected with 6E10. A longer exposure was needed to detect the Aβ generated from the Aph-1α2 co-IP (lanes 7 and 8), presumably because of a less efficient co-IP or steric interference of the 3F10 antibody.