Assembly of the presenilin γ-/ε-secretase complex


  • P. St George-Hyslop,

    1. Tanz Centre for Research in Neurodegenerative Diseases, University of Toronto, Toronto, Ontario, Canada
    2. Cambridge Institute for Medical Research, Addenbrookes Hospital, University of Cambridge, Cambridge, UK
    Search for more papers by this author
  • P. E. Fraser

    1. Tanz Centre for Research in Neurodegenerative Diseases, University of Toronto, Toronto, Ontario, Canada
    2. Department of Medical Biophysics, University of Toronto, Toronto, Ontario, Canada
    Search for more papers by this author

Address correspondence and reprint requests to P. St George-Hyslop, Cambridge Institute for Medical Research, Wellcome Trust/MRC Building, Room 4.36, Addenbrookes Hospital, University of Cambridge, Hills Road, Cambridge CB2 0XY, UK. E-mails:;


J. Neurochem. (2012) 120 (Suppl. 1), 84–88.


The presenilin complex is composed of four core proteins (presenilin 1 or presenilin 2, APH1, nicastrin, and PEN2). Several endogenous proteins have been reported to selectively modulate the function of the presenilin complexes; these include transmembrane trafficking protein, 21-KD (TMP21), CD147 antigen (basigin), the γ-secretase-activating protein (gSAP), and the orphan G-protein-coupled receptor 3. Because the structure and assembly of these complexes underlies their activity, this review will discuss current work on the assembly of the complex and on presenilin-interacting proteins that regulate secretase activity.

Abbreviations used

anterior pharynx defective 1


amyloid precursor protein


C-terminal fragment


γ-secretase-activating protein


γ-secretase modulator




presenilin enhancer 2


presenilin 1


presenilin 2


transmembrane trafficking protein 21-KD

The presenilin complex is composed of four core proteins, presenilin 1 (PS1; Sherrington et al. 1995) or presenilin 2 (PS2; Rogaev et al. 1995), anterior pharynx defective 1 (APH1; Francis et al. 2002), nicastrin (NCT; Yu et al. 2000), and presenilin enhancer 2 (PEN2; Francis et al. 2002) (Fig. 1), whose presence is required for physiological activity. Because APH1 exists as two separate homologues in humans and one undergoes alternative splicing, there are six combinations of PS1 or PS2 presenilin complexes. In addition, one or more regulatory proteins [e.g. transmembrane trafficking protein, 21-KD (TMP21; Chen et al. 2006) and γ-secretase-activating protein (gSAP; He et al. 2010; St George-Hyslop and Schmitt-Ulms 2010)] whose presence is not essential, interact with a subset of complexes. The presenilin complexes mediate the regulated intramembranous proteolysis of several type I transmembrane proteins (Wolfe et al. 1999; Marambaud and Robakis 2005). Recently, both negative stain electron microscopy and electron cryo-microscopy have been used to investigate the structure of presenilin complexes purified from heterologous cells significantly over-expressing multiple components of the complex. These initial studies have generated a variety of proposed presenilin structures differing in shape and volume including: (i) flat heart-shaped structures with a volume of 560 Å × 320 Å × 240 Å (Ogura et al. 2006); (ii) globular 120 Å diameter structures with a large, cylindrical interior channel of approximately 20–40 Å in diameter that is contiguous with 20 Å openings at the top and bottom surfaces (Lazarov et al. 2006); and (iii) 80 Å × 90 Å × 85 Å globular structures with three solvent-accessible but non-perforating central cavities, and with a putative mass of 230 kDa (Osenkowski et al. 2009). The reason for these differences is unclear, but may relate to over-expression of multiple human presenilin complex components in non-human cells that might cause alteration in the stoichiometry and/or post-translational modifications of the proteins. Contamination of even a small percentage of the population of complexes with aberrant complexes will alter the details of the structure produced by the 3-D reconstruction algorithm. Studies of native human presenilin complexes (or complexes produced by expression of a single tagged human component at near physiological levels) in human cells should resolve this question. Nevertheless, while these models are quite divergent, there is consensus on a transmembrane barrel-like structure with at least a partially patent central aqueous cavity.

Figure 1.

 The presenilin complex is composed of four core components [presenilin 1/presenilin 2 (blue); APH1 (yellow), PEN2 (red) and nicastrin (green)], and is the protease which performs the final cleavage of a variety of type I transmembrane proteins including APP (shown multi-coloured) and Notch (not shown). A number of potentially regulatory proteins associate with the presenilin complex including TMP21 (orange) and gSAP (purple) either directly via their TM domains (TMP21) or by a bipartite interaction with both the presenilin complex and with the juxtamembrane domain of APP (gSAP). It is unclear whether they are all interacting with the same regulatory domains, and whether this domain is also the site for interaction of the non-steroidal, anti-inflammatory gamma-secretase modulator compounds. However, they selectively modulate the presenilin-dependent cleavages at residues 49/50 (ε-site), 45/46/47 (ζ-site); 42/43 (minor but pathological γ-site); 39/40 (major γ-site), and 37/38 (minor γ-site).

Because the structure and assembly of these complexes underlies their activity, this review will discuss current work on the assembly of the complex and on presenilin-interacting proteins that regulate secretase activity.

APH1 and nicastrin – an initial scaffolding complex

Work by several groups using two-dimensional Blue Native gel electrophoresis or chromatographic fractionation methods have clearly established the existence of a stable, low abundance sub-complex containing APH1 and immature glycosylated forms of nicastrin (Arawaka et al. 2002; Yang et al. 2002; Gu et al. 2003). The formation of this intermediate APH1-NCT scaffold is regulated by the ER retention factor, retrieval to ER 1 protein, which blocks APH1 interaction with NCT during the early stages of presenilin complex assembly (Kaether et al. 2007; Spasic et al. 2007). The sequence of events leading to the formation of a mature presenilin complex may involve a direct binding of the APH1-NCT scaffold to PS1 (or PS2) followed by the incorporation of PEN2. Alternatively, the APH1-NCT pre-complex combination may bind directly to a comparable PS1-PEN2 structure to generate the mature, active presenilin complex (Shirotani et al. 2004; Capell et al. 2005).

The existence of alternative splice forms and homologous forms of APH1 (two in humans, three in mice) have led to the suggestion that different APH1 isoforms might have slightly different catalytic activities. This hypothesis is supported by the fact that knockout of APH1a results in lethality that is not seen following ablation of APH1b (Ma et al. 2005; Serneels et al. 2005). Furthermore, some, but not all studies, suggest that some APH1 isoforms (e.g. APH1b) might be particularly important in the production of Aβ (Serneels et al. 2005). If correct, targeting these APH1b complexes might be therapeutically attractive. However, the precise function of APH1 within the presenilin complex remains elusive. Several polar residues within APH1 transmembrane domains have been shown to contribute to the assembly and activity of the mature presenilin complexes (Pardossi-Piquard et al. 2009b), and it has been speculated that these residues are involved in substrate presentation with the secretase (Chen et al. 2010).

PEN2 addition activates presenilin endoproteolysis

Studies using PEN2 knockdown have shown that PEN2 has important functions in stabilizing the complex and permitting endoproteolytic cleavage of PS1 into the PS1 N-terminal fragment and PS1-C-terminal fragments (PS1-CTF) [some studies have suggested that PEN2 may be necessary for the PS endoproteolytic cleavage to occur (Steiner et al. 2002; Prokop et al. 2004, 2005)]. In the absence of PEN2, the complex is degraded by the proteasome. The PS-subunit stabilizing function of PEN2 depends on: (i) the length and overall sequence of its C-terminal domain (Steiner et al. 2002; Prokop et al. 2004, 2005); and (ii) on the proximal two-thirds of the PEN2 transmembrane domain 1 (Kim and Sisodia 2005b). The ‘NF’ sequence within TM4 or PS1 is the minimal PS1 motif that is required for binding with PEN2 (Kim and Sisodia 2005a).

Current data suggest that following the binding of pen-2, the TM6-TM7 loop domain of PS1/PS2, which is inserted into the transmembrane channel, is cleaved by endoproteolysis using the same ragged cleavages as observed with γ- and ε-cleavages (Fukumori et al. 2010). This cleavage of the TM6-TM7 loop is the final event which liberates the lumen and activates the complex for cleavage of other substrates.

Modifiers of presenilin complex activity

Several endogenous proteins have been reported to selectively modulate Aβ production, but spare Notch cleavage. These include TMP21 (Chen et al. 2006), CD147 antigen (basigin) (Zhou et al. 2005), the gSAP (He et al. 2010), and the orphan G-protein-coupled receptor 3 (Thathiah et al. 2009) (although the mechanism of the latter is by altering the subcellular localization of PS1 complexes).

TMP21 binds to presenilin complexes and modulates amyloid precursor protein (APP) cleavage, but not Notch cleavage (Chen et al. 2006). Knockdown of TMP21 increases γ-secretase-mediated Aβ production while having no effect on ε-secretase-mediate Notch cleavage and amyloid intracellular domain production. Examination of amyloid intracellular domain-regulated elements such as p53 and neprilysin following TMP21 suppression indicates no significant effect on their expression but is accompanied by an elevation of Aβ processing (Pardossi-Piquard et al. 2005; Alves Da Costa et al. 2006; Dolcini et al. 2008). TMP21 is expressed in neurons that co-localize with presenilin complex components, and its expression is decreased in Alzheimer disease cases as compared to unaffected controls (Vetrivel et al. 2008).

Using a domain swapping approach, the TM domain of TMP21 was found to be critical for the effect of TMP21 on presenilin catalytic function (Pardossi-Piquard et al. 2009a). This was confirmed though investigation of a synthetic peptide corresponding to the TMP21 transmembrane helix. The isolated TMP21 TM peptide, but not the comparable p24a TM peptide, was able to inhibit Aβ production in a dose-dependent fashion. These observations suggest that the transmembrane domain of TMP21 is likely to be involved (directly or indirectly) in the interaction with the presenilin complex.

Other studies have revealed that TMP21 does not appear to be associated with presenilin complexes that bind to γ-secretase inhibitor affinity-columns (Vetrivel et al. 2008). This has been interpreted to mean that TMP21 is associated with inactive complexes, which is consistent with the inhibitory activity of TMP21 in presenilin complexes. However, in agreement with the concept that TMP21 has a regulatory role, only a small proportion of the total cellular PS1 complexes contain TMP21. It also remains to be determined which components of presenilin complexes TMP21 interacts with, as well as the mechanism by which TMP21 modulates the function of the complex (i.e. direct binding to catalytic domain, modulation of structure and/or substrate presentation).

More recently, the gSAP was found to also be a modulator of γ-site cleavage while not affecting ε-site cleavage. gSAP is a ∼16 kDa protein derived by the proteolytic cleavage of a larger, uncharacterized protein, previously termed pigeon homologue protein (PION) (He et al. 2010). gSAP binds both to imatinib (a compound that selectively decreases the production of Aβ while sparing the cleavage of Notch) and to CTF of both PS1 and APP. gSAP differentially regulates the γ- and ε-site cleavages of APP-CTF, and is also highly substrate-specific, having no effect on Notch cleavage.

Although it seems from the present data that gSAP and TMP21 can bind to presenilin complexes, it is unclear how they inversely modulate the γ- and ε-site cleavages of APP-CTF. Additional studies are also needed to document the stoichiometry of the various cleavage products generated by complexes that are modulated by TMP21 and gSAP. Specifically, because TMP21 and gSAP either do not affect or even activate ε-cleavage of APP, the fate of the N-terminal fragment which is not cleaved by γ-site cleavage is unclear. Is this large N-terminal fragment simply left in the membrane as the equivalent of Aβ-49? Is it degraded by some other mechanism that does not generate traditional Aβ species ending at residues 40 and 42?

It seems likely that TMP21 and gSAP act upon the different allosteric sites. Current evidence suggests that TMP21 acts by its TM domain (Pardossi-Piquard et al. 2009a), whereas gSAP’s effect is mediated by gSAP’s binding to the juxtamembrane cytoplasmic domain of the substrate. Similar questions have been asked about a class of small molecules related to non-steroidal anti-inflammatory drugs (γ-secretase modulators, GSMs). GSMs inhibit γ-secretase activity while sparing ε-secretase activity and Notch processing. Conflicting evidence suggests that these compounds might bind both the APP substrate and the presenilin complex (Kukar et al. 2008). If so, it is conceivable that GSMs and gSAP might affect the same components of the presenilin complexes and their substrates, while TMP21 affects a different component of the presenilin complexes.

In conclusion, the presenilin complex represents a fascinating biological machine that is assembled from at least four core proteins (PS1 or PS2, APH1, PEN2 and nicastrin). These four core proteins are sufficient for cleavage of multiple different, non-homologous type 1 transmembrane proteins, with the cleavage occurring through the substrates’ TM domains. The presenilins are the catalytic sites for the enzyme. The role of the other proteins is unclear, possibly subserving roles in substrate-binding (nicastrin) or in the assembly and stabilization of the complex (aph1, pen2). In addition, several other proteins (e.g. TMP21, gSAP) may have a modulatory role on a subset of the enzymatic functions of the presenilin complexes, and they do so in a way that is of interest to drug design because they imply that the γ-site and ε-site cleavages may be separately regulated. It remains to be determined, however, whether these proteins act via similar molecular mechanisms to γ-secretase modulator compounds,

Conflict of interest

The authors declare no conflicts of interest.