|Home | About | Journals | Submit | Contact Us | Français|
γ-Secretase is an intramembrane protease complex that mediates the Notch signaling pathway and the production of amyloid β-proteins. As such, this enzyme has emerged as an important target for development of novel therapeutics for Alzheimer disease and cancer. Great progress has been made in the identification and characterization of the membrane complex and its biological functions. One major challenge now is to illuminate the structure of this fascinating and important protease at atomic resolution. Here, we review recent progress on biochemical and biophysical probing of the structure of the four-component complex and discuss barriers and potential pathways toward elucidating its detailed structure.
Alzheimer disease (AD) is a highly prevalent neurodegenerative disorder defined by the presence of abundant proteinaceous deposits -- extraneuronal amyloid plaques and intraneuronal neurofibrillary tangles -- in limbic and cortical regions of the human brain (St George-Hyslop, 2000). An amyloid plaque is a fibrous aggregate composed principally of amyloid β-proteins (Aβ) of 39 – 43 amino acids, whereas a neurofibrillary tangle is mainly composed of aggregated, hyper-phosphorylated tau, a cytosolic protein that is normally associated with microtubules. The causes of the common, late-onset form of AD are largely unknown, and the molecular pathway between extracellular Aβ accumulation and the development of the intraneuronal tangles is poorly understood. Certain autosomal dominant genetic alterations can predispose people to the disease, but many individuals without a known genetic predisposition develop AD in the sixties and beyond. The Aβ cascade hypothesis postulates that in both autosomal dominant forms that have accelerated rates of Aβ production and apparently “sporadic” forms that accumulate Aβ later in life, an initial elevation of soluble Aβ monomers and then oligomers and their subsequent aggregation and deposition into amyloid fibrils leads gradually to widespread synaptic, neuronal and glial dysfunction accompanied by declining memory and ultimately a multifaceted, profound dementia (Hardy and Selkoe, 2002; Shankar et al., 2008).
Aβ is a proteolytic product of amyloid β-protein precursor (APP), a large (770 amino acids maximum) type 1 transmembrane glycoprotein widely expressed in mammals (Fig. 1) (Hardy and Selkoe, 2002). The physiological activities of the precursor are partially understood, and APP has been implicated in cell adhesion, cell signaling, protease inhibition and development (Wolfe and Guenette, 2007). The large extracellular domain of APP has an E1 subdomain containing a copper-binding motif (Barnham et al., 2003), an alternatively spliced Kunitz-type protease inhibitor domain (Tanzi et al., 1988), and an E2 subdomain composed of two connected coiled coils (Wang and Ha, 2004). The single transmembrane domain (TMD) of APP is located near the carboxyl end, and the Aβ peptides originate in part from this domain. There is evidence that the APP TMD can dimerize (Beel et al., 2008; Gorman et al., 2008), and mutations of the Glycine residues in the GxxxG motif attenuate the TMD dimerization strength and reduce production of Aβ42 (Munter et al., 2007). APP undergoes ectodomain shedding in which its large, receptor-like N-terminal region is cleaved off, either by α-secretase at position Lys687 (APP770 numbering) to generate an 83-residue membrane-retained stub or by β-secretase at position Asp672 to generate a 99-residue membrane stub. Both C83 and C99 are substrates for a third membrane-associated protease called γ-secretase. The Aβ peptides are produced from C99; the cleavage products from C83 by γ-secretase are N-terminally truncated forms of Aβ referred to as p3. Peptide hydrolysis mediated by γ-secretase is unusual because the cleavage occurs within the hydrophobic lipid bilayer.
γ-Secretase belongs to a diverse family of Intramembrane-Cleaving Proteases (I-CLiPs) (Wolfe and Kopan, 2004). γ-Secretase is now known to be composed of 4 integral membrane proteins: presenilin, nicastrin, Aph-1, and Pen-2 (Edbauer et al., 2003; Kimberly et al., 2003; Takasugi et al., 2003), with presenilin being an aspartyl protease that is the catalytic subunit (De Strooper et al., 1998; Wolfe et al., 1999). The identification of these proteins was a result of convergent efforts along two initially unrelated lines of research (Selkoe and Wolfe, 2007): in the pathobiology, searching for the enigmatic γ-secretase that generates Aβ, and in development biology, searching for components of the Notch signaling pathway. The 4 subunits are now known to be both necessary and sufficient for active γ-secretase. However, there are other factors, for example, TMP21 and CD147, that have been reported to modulate γ-secretase activity (Chen et al., 2006; Zhou et al., 2005). Numerous mutations have been identified in both the proteolytic component, presenilin, and the substrate, APP, that influence the production of Aβ peptides (Goate et al., 1991; Hardy and Selkoe, 2002; Scheuner et al., 1996). The γ-secretase complex also functions in many other cellular processes. Perhaps the most well-studied to date is the Notch signaling pathway, which functions in the development of metazoans by mediating cell fate specification. In this role, γ-secretase releases the Notch intracellular fragment that subsequently enters the nucleus and regulates transcriptional programs (De Strooper et al., 1999; Struhl and Greenwald, 1999).
The necessary role of γ-secretase in the pathogenesis of AD commends it as a major target for drug development. But its multiple roles in many essential cellular functions complicate its chronic inhibition to treat or prevent AD. It is currently believed that it will be possible to develop γ-secretase inhibitors or modulators that alter Aβ-generating activity while preserving other essential biological functions of the protease (Wolfe, 2008). γ-Secretase is also a target for developing anti-cancer agents (Shih Ie and Wang, 2007). A high-resolution structure would be invaluable in understanding the biology of the γ-secretase complex and in facilitating inhibitor design.
Human presenilin 1 is a 467-residue, ~50 kDa transmembrane protein. Its homologue, presenilin 2, is four amino acids shorter, missing residues 26–29 near the amino terminus. There is a consensus now, after some controversy, that presenilin has a 9-TMD topology, with a cytosolic amino terminus and a lumenal carboxyl terminus (Fig. 2A) (Henricson et al., 2005; Laudon et al., 2005; Spasic et al., 2006). In the active γ-secretase complex, presenilin is cleaved between residues N292 and V293 into an amino terminal fragment (NTF) (~30 kDa, TMD1-6) and a carboxyl terminal fragment (CTF) (~20 kDa, TMD 7-9). This endoproteolysis appears to be an intramolecular autocatalytic event (Brunkan et al., 2005; Wolfe et al., 1999) that is carried out by the same γ-secretase activity that cleaves other substrates. Such a notion is arose from the observation that the presenilin active site mutants D257A or D385A do not undergo endoproteolysis and cannot cleave substrates (Wolfe et al., 1999). The endoproteolysis of presenilin (or the omission of exon 9 as an exceedingly rare familial AD mutation) apparently removes steric hindrance of the large cytoplasmic loop between TMDs 6 and 7, thus enabling presenilin to take on its enzymatically active conformation. The NTF and CTF remain bound together after the cleavage (Thinakaran et al., 1996).
Regarding the intramolecular arrangement of the TMDs in presenilin, it is known that TMD6 and TMD7, each harboring a catalytic aspartic acid at D257 and D385, are clustered at the active site (Sato et al., 2006; Tolia et al., 2006). In addition, TMD1 has been shown by disulfide crosslinking to be near TMD8 (Kornilova et al., 2006). TMD9 also appears to be near the catalytic pore and might be involved in the initial binding of substrates (Tolia et al., 2008). Substituted cysteine accessibility experiments have demonstrated a hydrophilic pore within the membrane at the active site formed by TMD6 and TMD7 (Sato et al., 2006; Tolia et al., 2006).
Only a fraction of cellular presenilin actually forms the γ-secretase complex (Lai et al., 2003). It is postulated that presenilin may have functions outside the γ-secretase complex. For examples, presenilin has been reported to function as a low conductance Ca2+ leak channel (Tu et al., 2006) and may thus be involved in neuronal Ca2+ signaling and homeostasis (Smith et al., 2005). Presenilin also interacts with and, as a consequence, stabilizes β-catenin (Serban et al., 2005; Zhang et al., 1998). Since β-catenin upregulates the expression of several genes associated with cancers in adult (Moon et al., 2002), presenilin mutations might play a role in causing certain cancers.
In addition to presenilin, the other three components in γ-secretase complex are Aph-1, nicastrin, and Pen-2 (Fig. 2A). Aph-1 has a topology of seven TMD with a cytosolic carboxyl terminus (Fortna et al., 2004). Pen-2 is a 2-TMD hairpin-like protein with both ends in the lumen (Crystal et al., 2003). Nicastrin is a type-1 membrane glycoprotein with a large lumenal domain. The ectodomain of Nicastrin shares a common fold with the peptidases (Fagan et al., 2001). The ectodomain is involved in the assembly and maturation of the complex (Chavez-Gutierrez et al., 2008) as well as its conformational change upon activation of the γ-secretase complex (Shirotani et al., 2003). There are up to 16 potential N-glycosylation sites, although glycosylation doesn’t appear to be necessary for the function of γ-secretase (Herreman et al., 2003).
γ-Secretase appears to follow a defined and stepwise assembly pathway. Initially, nicastrin and Aph-1 form a binary subcomplex (Shirotani et al., 2004), and this half-complex is stable in the detergent n-dodecyl-β-D-maltopyranoside, which dissolves the 4-member complex (Fraering et al., 2004a). Subsequently, presenilin binds to the nicastrin-Aph-1 subcomplex to form a hetero-trimeric subcomplex (Takasugi et al., 2003). Finally, Pen-2 joins the trimeric subcomplex, resulting in the formation and activation of the mature γ-secretase complex (LaVoie et al., 2003; Niimura et al., 2005). The addition of Pen-2 may allow the endoproteolysis of presenilin to occur, thereby activating the protease. The γ-secretase complex might have already fully assembled in ER, but there is some evidence that its full activation may occur in a slightly acidic export compartment downstream of the ER (Kim et al., 2007).
There has been controversy concerning the stoichiometry of the γ-secretase complex, both in terms of the overall oligomerization state and the copy number of individual subunits, particularly presenilin. For example, the dimerization of presenilin at the catalytic core has been suggested (Hebert et al., 2003; Schroeter et al., 2003). The calculated molecular weight based on protein sequence, assuming one copy each of the four subunits, is about 200 kDa. Purified γ-secretase runs on some Blue Native gels at around 500 kDa, implying that γ-secretase might be a dimeric complex. However, molecular sizing based on electrophoresis can be unreliable, particularly in the case of membrane protein complexes, due to different migration characteristics of these hydrophobic proteins with respect to water-soluble molecular weight markers (Osenkowski et al., 2009). Glycosylation of nicastrin adds another layer of uncertainty, because the strongly charged sialic acid might bind disproportionally to the dye molecule, increasing the apparent weight of the complex in gels.
Recent work with purified and functionally active complexes revealed a 1:1:1:1 stoichiometry of the 4 γ-components (Sato et al., 2007). Consistent with this stoichiometry, the absolute mass of the purified γ-secretase as measured by scanning transmission electron microscopy (STEM) is 230 kDa (Osenkowski et al., 2009). The extra ~ 30 kDa above the theoretical mass may be accounted for by the glycosylation state of nicastrin. Therefore, it is likely based on the latest evidence that γ-secretase is a monomeric complex with one copy each of its four components, although the data cannot role out the possibility of the presence of an extra Pen-2 subunit. We also note that this stoichiometry is established for the detergent purified γ-secretase complex from the over-expressing system, in which every γ-component has one or more exogenous tags. Therefore, it remains to be seen if this conclusion can be applied to the endogenous complexes in vivo. However, oligomerization, even if can occur under some conditions, is unlikely to play a significant role in γ-secretase activity, because purified monomeric γ-secretase complex is functional (Cacquevel et al., 2008; Fraering et al., 2004b).
Fig. 2B summarizes the overall architecture of γ-secretase complex, based on pair-wise interactions, partial complexes in different types and concentrations of detergents (Fraering et al., 2004a) and chemical crosslinking (Sato et al., 2006; Schroeter et al., 2003; Steiner et al., 2008; Thinakaran et al., 1998). First, there is a tight interaction between nicastrin and Aph-1. The NTF and the CTF of presenilin further interact with Pen-2 and nicastrin, respectively. There are two mammalian isoforms of presenilin (PS1 and 2) and three of Aph-1 (Aph1α-short, 1α-long and Aph-1β), resulting in six possible γ-secretase complexes (Shirotani et al., 2007; Wakabayashi and De Strooper, 2008). Despite their potential differences in specific activities (Serneels et al., 2005), the architecture of these γ-secretases appears to be essentially the same (Steiner et al., 2008).
γ-Secretase is the only intramembrane protease to date that functions as a multi-subunit protein complex; other I-CLiPs are single-protein enzymes. The obvious question is: what are the roles for each component and why are there so many components in γ-secretase? Whereas presenilin has been shown to function as the catalytic subunit, the exact roles of the other three subunits are not clearly defined, beyond that they are necessary for γ-secretase activity. One explanation for the requirement of multiple components might be that γ-secretase acquired over evolution the ability to cleave peptide bond rather non-specifically, so that there is a need to effectively sequester the active site using other protein partners, in order to prevent unintended proteolysis. More subunits might also afford fine-tuning the activity of γ-secretase in response to its environmental cues.
In addition to APP and Notch, γ-secretase processes more than 80 other transmembrane protein substrates (Beel and Sanders, 2008; Hemming et al., 2008; Wakabayashi and De Strooper, 2008). No consensus sequence motif can be recognized among these substrates. So are there general characteristics that make a protein a γ-secretase substrate? The answer seems to be yes. The general requirements are: (1) a type I transmembrane helix; (2) a small ectodomain, usually resulting from prior shedding by an α-secretase-like protease; (3) permissive determinants in the juxtamembrane and cytoplasmic domains (Hemming et al., 2008); and perhaps (4) residues that can destabilize the helical configuration of the TMD around the cleavage site (Beel and Sanders, 2008). Since both γ-secretase and its substrates reside in the confined two-dimensional space of the lipid bilayer, it was previously hypothesized that γ-secretase might have little or no sequence specificity and cleave virtually any hydrophobic α-helical domain. In this sense, γ-secretase might be considered as a membrane equivalent of the cytoplasmic proteasome (Kopan and Ilagan, 2004; Small, 2002). However, recent studies suggest that there are specific determinants that make a polypeptide a substrate of γ-secretase and other presenilin-like intramembrane proteases (Hemming et al., 2008; Martin et al., 2008; Ren et al., 2007). Nevertheless, there are certain parallels between γ-secretase and the proteasome: (1) both undergo a defined assembly process; (2) both require a maturation step before becoming enzymatically active: in the proteasome, the propeptide is auto-cleaved as the final maturation step, and in γ-secretase, the large cytoplasmic loop between TMD6 and TMD7 in presenilin is also auto-cleaved to activate the complex. Proteasomal substrates are tagged with a ubiquitin chain and translocated in an ATP-dependant manner through an entrance for proteolysis. In analogy to the proteasome, one might ask then, if there is a defined gating mechanism, a subunit that acts in substrate recognition and/or translocation, or a tagging system that marks substrates for γ-secretase cleavage? It will be interesting to see whether certain domains of the 4-member complex or one of the postulated the γ-secretase modulatory proteins play such roles. Again, elucidating the structure of γ-secretase should help us understand some of these questions.
The structure of recombinant human γ-secretase complex expressed in CHO cells was initially studied by negative stain EM and 3D image reconstruction (Lazarov et al., 2006). γ-Secretase was found to have a spherical structure about 8–10 nm in diameter with a low-density interior (~2 nm) that was suggested to be a water accessible proteolytic chamber. However, another study using a γ-secretase preparation from Sf9 cells produced a much larger, Y-shaped structure at 30–50 nm in size with an unusually large central pore (> 10 nm) (Ogura et al., 2006). While the exact reason for this discrepancy is unclear, we suspect that differences in the γ-secretase constructs might be responsible. In our γ-30 CHO cell line, there was no His-tag on any of the four components (Lazarov et al., 2006). In the Sf9 expression system, 3 out of the 4 protein components, i.e. presenilin-1, nicastrin and Aph-1, contained a His-tag (Ogura et al., 2006). The 6XHis-tag is positively charged and is capable of altering the oligomeric state of membrane proteins (Amor-Mahjoub et al., 2006; Li et al., 2004; Wu and Filutowicz, 1999). Thus, it is possible that the three His-tags in each γ-secretase might have induced artificial oligomerization of the Sf9-purified γ-secretase complex, potentially explaining the unusually large observed size (Ogura et al., 2006).
Recently, an improved structure of γ-secretase was determined by cryo-EM at a nominal resolution of 12 Å (Osenkowski et al., 2009). Cryo-EM reveals a protein structure itself, rather than a surface envelope of the protein structure as contoured by heavy metal stain, which is the case in the negative stain method. Therefore, the cryo-EM method provides a potentially more reliable structure by circumventing several shortcomings in negative stain EM, such as uneven staining, stain-induced flattening, and possible stain penetration to the interior of a protein structure. The cryo-EM structure was determined from a new human γ-secretase preparation (purified from our S20 CHO cell line) that contained a 6XHis-tag at the carboxyl terminus of nicastrin (Cacquevel et al., 2008). It was verified that this single His-tag did not cause oligomerization of the γ-secretase complex. The overall size and shape of the S20 γ-secretase was essentially the same as the γ30 cell-derived γ-secretase that had no 6XHis-tag (Lazarov et al., 2006). The new cryo-EM structure reveals three smaller low-density interior regions; these regions do not coalesce to form a single chamber as observed in the negative stain structure (Lazarov et al., 2006). The cryo-EM structure has better definition than the previous stained structure in that the ~4 nm thick transmembrane region and four extracellular density domains are resolved (Osenkowski et al., 2009) (Fig. 3).
γ-Secretase is one of the four families of intramembrane proteases. Since the crystal structures of two intramembrane proteases, the serine protease Rhomboid and the metalloprotease S2P, were solved recently (Fig. 3) (Ben-Shem et al., 2007; Feng et al., 2007; Lemieux et al., 2007; Wang et al., 2006; Wu et al., 2006), we asked whether these structures resemble the γ-secretase. Unfortunately, these six-TMD single-polypeptide proteases share no sequence or structural similarity among themselves or with γ-secretase (Urban and Shi, 2008). However, all reside in similar hydrophobic bilayer environments and perform virtually the same task of hydrolyzing peptide bonds; thus, the intramembrane proteases all face the same requirements of admitting the transmembrane helical substrates as well as water molecules into their respective catalytic sites located within the bilayer. Therefore, a side-by-side comparison of the structures of these proteases might still be meaningful (Fig. 3). Indeed, there appears to be a vertically oriented groove on the surface in the transmembrane portion of each of these structures (Fig. 3A). These grooves likely represent the initial substrate docking sites (Kornilova et al., 2005). Binding of a substrate to the docking site on an intramembrane protease should induce a lateral gate opening in the protease (Feng et al., 2007) and a helix-to-coil conformational switch at the cleavage site in the substrate structure (Beel and Sanders, 2008), such that the substrate could enter into the catalytic site and be ready for processing. The vertical sections of these structures show sizable cavities that open to the cytoplasm and/or extracellular regions, indicating that water has access to the interior of the enzymes where the catalytic sites reside, even in the absence of their respective substrates (Fig. 3B). Thus, the intramembrane proteases appear to have solved the conundrum of hydrolyzing peptide bonds in a hydrophobic environment by simply creating a hydrophilic, water-accessible cavity in the greasy membrane (Urban and Shi, 2008). In all three cases, the water pathway appears to reach as far as the catalytic site – but does not cross the membrane. This might be important to preserve membrane potentials (Ha, 2007; Urban and Shi, 2008).
The low-resolution cryo-EM map of γ-secretase has whetted our appetite for a higher resolution structure. However, a crystal structure of the full γ-complex may not be on the horizon just yet. The complex maturation pathway (Spasic and Annaert, 2008), the general requirement of certain lipids for activity (Osenkowski et al., 2008), the multiple glycosylations of nicastrin, the tendency to break into subcomplexes in detergents that are more appropriate for structural studies (Fraering et al., 2004a), the existence of multiple isoforms of Aph-1 and presenilin resulting in at least 6 different γ-complexes (Wakabayashi and De Strooper, 2008), and the relatively small quantity of the γ-complex that can be produced for crystallization trials (Cacquevel et al., 2008) are all formidable barriers to achieving the crystal structure of this 19-TMD complex. However, much progress has been made in recent years in over-expressing mammalian membrane proteins (Tate, 2001), in robot-based nano-crystallography that significantly reduces the required amount of highly purified specimen (Bodenstaff et al., 2002), and in the availability of synchrotron-based x-ray diffraction equipment enabling the use of ever-smaller crystals (Jiang and Sweet, 2004). Considering how seemingly impossible an atomic structure of an ion channel or a transporter was twenty years ago and how many such crystal structures are available now (Gouaux and Mackinnon, 2005), it might be reasonable to expect that the atomic structure of γ-secretase will be achieved in the not-so-distant future.
Besides attempting crystallization of the entire γ-complex, another route for the X-ray crystallography approach would be to solve the structures of individual components or domains of complex. In this regard, signal peptide peptidase (SPP), a homolog of presenilin (Narayanan et al., 2007), the nicastrin ectodomain (Shah et al., 2005), Aph1, and PEN-2 might each be a good target for crystallography or NMR structure determination. The presenilin holoprotein has been efficiently expressed alone in Sf9 insect cells and purified from microsomes (Tu et al., 2006). This provides a unique opportunity for solving the structure of the catalytic subunit of γ-secretase. When the high-resolution structures of these γ-secretase components become available, interpretation of the lower resolution EM structure of the intact γ-complex will be more meaningful.
In recent years, the methodology of single particle cryo-EM has advanced dramatically, to a point that near-atomic resolution structures for a number of very large and highly symmetrical biological assemblies such as the icosahedral or helical virus have been achieved (Jiang et al., 2008; Yu et al., 2008; Zhang et al., 2008). The attainable resolution for several very large protein complexes without any symmetry has also been significantly advanced to well within the nanometer resolution range (Agirrezabala et al., 2008; Chandramouli et al., 2008; Matadeen et al., 1999). It might thus be possible for the atomic resolution cryo-EM structure of a very large asymmetrical complex to be achieved. However, large and tightly packed protein complexes with masses well over 1 MDa are a minority in biology. The vast majority of the known protein complexes are in the size range of 100 – 400 kDa, a range to which γ-secretase belongs. From a structural biologist’s point of view, it is a difficult middle ground, since this range is somewhat too big for x-ray crystallography and too small for single particle cryo-EM.
Cryo-EM likely will continue to play an important role in our understanding of the structure and function of the γ-secretase complex. In the absence of atomic structure of the γ-secretase by crystallography, more work by cryo-EM might further improve the resolution of the existing structure of the γ-secretase complex; it might be possible to determine the EM structure at better than 10 Å resolution. A structure at this level of resolution is far more desirable or informative than the 12 Å structure we have obtained, because many of the predicted transmembrane helices would be resolved, so that homology-based modeling or docking of the component atomic structures, once available, into the EM structure could be carried out. Furthermore, the potential substrate docking site and the catalytic site could also be visualized by determining the cryo-EM structures of a catalytically inactive (aspartate-mutant) γ-secretase in complex with substrates or helical peptide inhibitors (Bihel et al., 2004). When an atomic structure is finally achieved by x-ray crystallography, the EM structure may serve to cross-validate the physiological relevance of crystal structures that could be influenced by the crystalline packing. An interesting case in point is the role cryo-EM played in supporting the protein-lipid interface movement model of the voltage sensor paddle as proposed from the crystal structures of potassium channels (Jiang et al., 2004).
Within the cryo-EM single particle approach, despite its theoretical potential for determining relatively small protein structures down to 40 kDa (Henderson, 1995), several practical issues, including beam-induced specimen charging and specimen movement, have so far hindered its realization. Another problem in single particle cryo-EM is the reliance on phase contrast generated by deeply under-focusing the objective lens. A large defocus value not only scrambles the higher frequency information in the micrographs that makes accurate information recovery difficult but also significantly cuts off the very low frequency information that is essential for image alignment. The emerging technique of phase plate enables in-focus electron imaging while still providing high image contrast (Cambie et al., 2007; Danev and Nagayama, 2001). This technique will likely make a significant difference in cryo-EM structural determination of relatively small protein complexes such as γ-secretase.
The cryo-EM method also encompasses another approach that is called electron crystallography (Fujiyoshi and Unwin, 2008). With this approach, detergent-purified membrane proteins are reconstituted back into a lipid bilayer, forming two-dimensional crystalline sheets. The structures then are determined by extracting the amplitudes from electron diffraction patterns and the phases from high-resolution electron images of these crystalline sheets (Glaeser et al., 2007). Membrane protein structures solved by this method are invaluable in evaluating the x-ray crystal structures that are necessarily solved in detergents (Fujiyoshi and Unwin, 2008). For γ-secretase, the lipid environment appears to be particularly important, since its proteolytic function occurs within the lipid bilayer, and this function is regulated by the lipid composition (Osenkowski et al., 2008). Many 2D crystals of membrane proteins reported so far do not diffract to a high resolution; nevertheless, there are several examples where atomic resolution structures have been solved by electron crystallography (Fujiyoshi and Unwin, 2008; Gonen et al., 2005). Therefore, we hope that electron crystallography will also contribute to the structural assault on the γ-secretase complex.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.