Structure–activity relationship of memapsin 2: implications on physiological functions and Alzheimer's disease

Abstract

Memapsin 2 (BACE1, β-secretase), a membrane aspartic protease, functions in the cleavage of the type I transmembrane protein, β-amyloid precursor protein (APP), leading to the production of amyloid β (Aβ) in the brain. Since Aβ is closely associated with the pathogenesis of Alzheimer's disease, understanding the biological function, particularly the catalytic activities of memapsin 2, would assist in a better understanding of the disease and the development of its inhibitors. The transmembrane and cytosolic domains of memapsin 2 function in cellular transport and localization, which are important regulatory mechanisms for its activity. The catalytic ectodomain contains a long substrate cleft that is responsible for substrate recognition, specificity, and peptide bond hydrolysis. The substrate cleft accommodates 11 residues of the substrate in separate binding subsites. Besides APP, a number of membrane proteins have been reported to be substrates of memapsin 2. The elucidation for the specificity of these subsites and the amino acid sequences surrounding the memapsin 2 cleavage site in these proteins has led to the establishment of a predictive model that can quantitatively estimate the efficiency of cleavage for any potential substrates. Such tools may be employed for future studies of memapsin 2 about its biological function. Herein, we review the current knowledge on the structure–function relationship of memapsin 2 and its relationship in the biological function.

Introduction

It has been more than a decade since several laboratories independently reported the identification of a novel aspartic protease [1–5] and evidence showed that it was the long-sought β-secretase known to play a key role in the pathogenesis of Alzheimer's disease (AD). The name of memapsin 2 (rooted in the membrane-anchored aspartic protease of the pepsin family-2) was given by our group [4] and is recommended by the International Commission in Enzyme Nomenclature. A commonly used acronym, BACE1, is also found in the literature. Memapsin 2, although ubiquitously distributed in many tissue types, is present at a relatively high level in the brain [1]. In the past decade, the close involvement of memapsin 2 in AD has stimulated intense research and attained a better understanding of this protease in its structure and function relationships as well as in the development of its inhibitors, which may lead to the treatment of AD. The latter has been extensively reviewed in the literature [6–9]. Here, we discuss the current understanding on the structure and function relationships of memapsin 2.

Involvement of Memapsin 2 in AD

AD is a form of neurodegenerative disease manifested by progressive memory loss and cognitive deficiency, ultimately leading to a near complete loss of mental function. A wealth of evidence suggests that the excessive level of 40/42-residue peptide amyloid β (Aβ) in the brain is intimately related to the development of AD. Aβ is formed from the proteolysis of its precursor protein, β-amyloid precursor protein (APP), by memapsin 2, and another membrane protease γ-secretase, a path of APP catabolism known as the amyloidogenic pathway. Alternatively, APP can also be processed via a non-amyloidogenic pathway by which the sequential cleavage of APP is mediated through α- and γ-secretases, thereby precluding the production of Aβ (Fig. 1). In the non-amyloidogenic pathway, α-secretase is thought to be the collective activities of several membrane-associated metalloproteases such as ADAM10 and ADAM17. In the amyloidogenic pathway, APP is first cleaved by memapsin 2 between position Met596 and Asp597 (APP695 isoform residue numbers), forming a large N-terminal ectodomain known as soluble APPβ (sAPPβ), and a C-terminal 99-residue membrane-associated fragment called C99. The latter is further cleaved by γ-secretase to produce Aβ and an APP intracellular domain (AICD). In addition to the primary cleavage site on APP that leads to the production of Aβ, memapsin 2 has been reported to cleave APP on a second site, a Tyr–Glu bond between residues 10 and 11 of the Aβ sequence [10,11]. These Aβ peptides, Aβ_11–40 and Aβ_11–42, appear to be more fibrillogenic and neurotoxic than the Aβ resulting from primary cleavage in vitro [12] and are reportedly found in the brain of AD patients.

Figure 1.

Processing of APP by α-, β-, γ-secretasesMembrane anchored APP is processed by two pathways. In the amyloidogenic pathway, a soluble APP fragment (sAPPβ) is secreted upon cleavage by β-secretase (memapsin 2). The resulting C99 fragment is cleaved by γ-secretase to produce Aβ. The γ-secretase cut releases the AICD, which may be involved in nuclear signaling. In the non-amyloidogenic pathway, α-secretase cleavage occurs within the Aβ domain and prevents amyloidogenesis. However, a small peptide (P3) is generated by the subsequent cleavage of the C83 fragment by γ-secretase. In addition to P3, the large ectodomain (sAPPα) is secreted.

The regulation of APP catabolism between amyloidogenic and non-amyloidogenic pathways is an important factor that controls the level of Aβ in the brain. The fact that memapsin 2 is the protease involved in the initiation of Aβ production in vivo provides strong evidence to link this protease to AD. Additional evidence supporting the role of memapsin 2 in initiating Aβ generation has been well established in the literature. For example, increased memapsin 2 levels have been found in AD brains [13–17]. Furthermore, APP mutations that increase memapsin 2 hydrolytic rate produced an early onset form of AD [18–21], while another APP mutation resulting in a reduced memapsin 2 hydrolytic rate protected people from AD [22]. Lastly, deletion of memapsin 2 gene in mice abolished Aβ production and rescued transgenic mice from AD-like syndromes [23,24].

Physiological Functions of Memapsin 2 Structural Domains

Memapsin 2 is initially synthesized as an inactive precursor protein in the endoplasmic reticulum. Structurally, the inactive precursor contains an N-terminal signal peptide (residues 1–21) followed by a pro-peptide (residues 22–45), a protease domain (residue 46–460), a transmembrane domain (residues 461–477) and a small cytosolic domain (residues 478–501) (Fig. 2). The N-terminal signal sequence directs transportation of the 65 kDa pro-memapsin from endoplasmic reticulum (ER) to Golgi [25]. Within the Golgi, the signal peptide is removed by furin and the protein undergoes N-glycosylation at four asparagine sites, thereby rendering a mature 70 kDa memapsin 2 [25–28]. Matured memapsin 2 consists of three distinct structural domains. The N-terminal catalytic domain is a typical aspartic protease of the pepsin family containing two canonical active-site DTGS/T motifs and a strong sequence homology with pepsin-like aspartic proteases. The function of the catalytic domain is to cleave protein substrates such as APP and others to be described below. The transmembrane and cytosolic domains of memapsin 2 function in anchoring the protease to the membrane and regulating its cellular trafficking, respectively. Memapsin 2 and its closest homolog memapsin 1 are unique among aspartic proteases of the pepsin family in having the transmembrane and cytosolic domains.

Figure 2.

Domain structure of memapsin 2Memapsin 2 is synthesized as a precursor with an N-terminal signal peptide (residues 1–21) followed by a propeptide (residues 22–45), a protease domain (residues 46–460), a transmembrane domain (residues 461–477), and a small cytosolic domain (residues 478–501). It undergoes glycosylation for asparagines. The DTGS and DSGT motifs contain the catalytic aspartyl residues for protease activity. The dileucine motif (DISLL) is from amino acids 496 to 500.

Transmembrane and Cytosolic Domains Regulate Memapsin 2 Activities through Intracellular Trafficking

The transmembrane domain of memapsin 2 has common characteristics with the structures of other type I transmembrane proteins. Evidence suggests that the transmembrane domain of memapsin 2 may also play an important role in controlling its activity. First, membrane attachment appears to be important for the efficiency of its activity as indicated by a reduction in APP cleavage in cells expressing soluble form of memapsin 2 [29,30]. Since all the known substrates of memapsin 2 are membrane proteins, the efficiency of hydrolysis may be facilitated by the close proximity of the protease and substrates tethered to the membrane. Second, the transmembrane domain contains three half cysteines, which are linked to palmitic acid. Although contradictory reports exist in the literature, such post-translational modifications may facilitate the partition of memapsin 2 into the lipid raft domain of the membrane and facilitate its trafficking to endosomes, an event intimately related to the degradation of substrates like APP [31–33].

Mature memapsin 2 is transported to the cell surface via the constitutive secretory pathway. On the cell surface, both memapsin 2 and APP are located in ‘lipid rafts’, distinct membrane domains characterized by high concentrations of cholesterol glycosphingolipids [34]. While on the cell surface, APP cleavage by memapsin 2 is minimal because at pH 7, memapsin 2 has only residual activity. Both memapsin 2 and APP are endocytosed into the endosomal system (interior pH about 5), where the acidic interior can effectively accommodate APP cleavage by memapsin 2 [28,35]. The observation that a truncated recombinant memapsin 2, lacking the transmembrane or cytosolic domain (M2_ED), can be co-endocytosed with APP from the cell surface to endosomes suggests that the endocytosis of memapsin 2 is at least partially dependent on the interaction of its ectodomain with APP [36]. The co-endocytosis of APP and memapsin 2 is also facilitated by the binding of ApoE to its receptor apolipoprotein E receptor 2 (ApoER2) through the formation of an APP/X11/ApoER2 complex [37]. From the endosomal system, memapsin 2 is recycled back to the Golgi and eventually to the cell surface through a phosphorylation (S498)-dependent interaction of the acid cluster dileucine motif (ACDL) on memapsin 2 and GGA1 or retromer family of proteins [38–42]. Another GGA family member, GGA3 can also bind with phosphorylated memapsin 2 and direct it from ER to the lysosome for degradation [43–46] (Fig. 3).

Figure 3.

Cellular transport of memapsin 2Memapsin 2 is a type I transmembrane protein that is synthesized in the ER, post-translationally modified in the Golgi and secreted to the cell surface where it undergoes coupled endocytosis with APP, thereby resulting in the presence of both proteins in the endosomal compartment ApoE-containing lipoprotein particles can trigger the endocytosis of complexes formed by ApoER2, APP, and memapsin 2. This event is mediated by adaptor protein X11. Memapsin 2 cleaves APP in the acidic endosomal environment that constitutes the first cleavage event in the amyloidogenic pathway leading to Aβ production. From the endosomes, memapsin 2 is recycled back to the cell surface through the secretory pathway. Phosphorylation of the dileucine motif on memapsin 2 results in an interaction with GGA or the retromer complex, which allows for the transport of memapsin 2 from the endosomal system to the Golgi. Once in the Golgi, memapsin 2 is transported back to the cell surface and remains unphosphorylated.

Catalytic Domain, Active Site and Enzyme Specificity

Like other aspartic proteases, the crystal structure of memapsin 2 ectodomain has two nearly independently folded similar halves, each contributing structurally to an elongated active site cleft located between the lobes [47,48]. Although the folding of the memapsin 2 peptide chain is nearly the same as in other aspartic proteases of the pepsin family, there are some distinct structural features. The most unique among these are the four insertion loops on the molecular surface in an area adjacent to the N-terminal of the substrate cleft, forming a larger molecular boundary and contributing to a more open and accessible active site cleft. These four loops are not directly involved in substrate recognition, thus, their function may involve yet undiscovered molecular contacts that are important for their biological function.

Both crystal structure and biochemical data indicated that the active-site cleft of memapsin 2 accommodates 11 substrate residues (subsites S7 to S4′), instead of 8 substrate residues (S4 to S4′), which is common for most aspartic proteases [49–52]. The residue preferences of the memapsin 2 subsites are of great interest for understanding the physiological function of this protease and for assisting in the design of inhibitors. The complete subsite specificity of memapsin 2 has been determined [52–54] using quantitative kinetic parameters, relative k_cat/K_m, which revealed a generally broad specificity for this protease. This indicates that each subsite can accommodate a number of amino acids. These studies contributed to design synthetic substrates, potent inhibitors, and assess physiological protein substrates (see below). However, the contribution of an individual subsite to the overall specificity of memapsin 2 varies a great deal. In general, the subsites near the scissile bond, e.g. S1 and S1′, are more important in determining the specificity than the outside subsites. Crystal structures of memapsin 2 bound to transition-state analogues have revealed the locations of these subsites on the protease tertiary structure. Since the substrate of memapsin 2 is in an extended structure [48], the subsites are alternately located on opposite sides of the active site cleft. As will be discussed below, a predictive model for memapsin 2 specificity has been established that the subsites P4, P3, P2, P1, P1′, and P2′ are more important than the other sites. Figure 4 shows the binding of the most favorable sidechains in these six subsites. It can be seen that in each case, these subsites are independent enclave areas that create strong interactions between the side chains and the protease residues.

Figure 4.

Stick model illustrating the interactions of substrate sidechains with six subsite residues of memapsin 2The six substrate side chains (Glu, Leu, Asp, Leu, Ala, and Val shown in green) are bound in subsites P4, P3, P2, P1, P1′, and P2′, respectively. In each subsite, the close interaction between the memapsin 2 residues (in the center) and the substrate sidechains are shown. Images are taken from a crystal structure of memapsin 2 complexed with inhibitor OM00-3 (PDB 1M4H [87]).

Protein Substrates of Memapsin 2

The identification of the physiological substrates for memapsin 2 is important in the understanding of its biological function. Such information may assist in assessing potential side effects of memapsin inhibition in the clinical treatment of AD. In addition to APP, a number of protein substrates for memapsin 2 have been reported including neuregulins 1, 3 (NRG1, 3) [55–57], APLP1, 2 [58–62], voltage-gated sodium channels β1–4 (VGSC β1–4) [63–66], beta-galactoside alpha 2,6-sialyltransferase (ST6Gal1) [67–69], the cell adhesion protein P-selectin glycoprotein ligand 1 (PSGL-1) [70], lipoprotein receptor-related protein (LRP) [62,71], and interleukin-1 receptor type II (IL-IR2) [72]. Most of these proposed substrates are type I transmembrane proteins, with the exception of NRG1, 3 (multiple transmembrane proteins), and ST6GAL (type II transmembrane protein) (Fig. 5).

Figure 5.

Substrate sequences recognized by memapsin 2Memapsin 2 substrates are all transmembrane proteins including APP, NRG1, NRG3, APLP1, APLP 2, VGSC β1–4, beta-galactoside alpha 2,6-sialyltransferase (ST6Gal1), the cell adhesion protein PSGL-1, LRP, and IL-IR2. Neuregulin 1 is unique as multiple-membrane protein. ST6Gal 1 is a Golgi-resident protein, unique as a type II transmembrane protein, whereas others are most type I transmembrane proteins that, like APP, are likely all cleaved by memapsin 2 in the endosomes. The arrow indicates the cleavage site in each substrate. For APP, the cleavage between M and D is the primary cleavage site and the other is the alternative cleavage site.

Neuregulins (NRG1 and NRG3)

Besides APP, neuregulins (epidermal growth factor-like factors, NRGs) are the best established protein substrates for memapsin 2 [55–57,73,74]. Processing of neuregulin by memapsin 2 produces an NRG1 N-terminal domain that binds to ErbB receptor and regulates myelination of axons in the nervous system. Mice with memapsin 2 gene deletion produce hypomyelination of axons in both the central and peripheral nervous systems during prenatal development [55,56]. Moreover, the abrogation of memapsin 2 cleavage on NRG1 also impairs the re-myelination process of injured sciatic nerves in adults [57].

Voltage-gated sodium channels subunits β1–4

Voltage-gated sodium channels play an essential role in the initiation and propagation of action potentials in neurons. These channels are large membrane protein complexes consisting of one α subunit and at least one β subunit as their auxiliary unit [75]. To date, four β subunits (VGSC β1–4) have been identified. Similar to APP processing, β2 and β4 subunits of VGSC are subjected to sequential cleavage by memapsin 2 and γ-secretase, resulting in the generation of β-intracellular domains (β-ICDs) [63–65]. Cleavage of VGSC β1 or β3 by memapsin 2 is less efficient than VGSC β2 and β4. Recent studies revealed that cleavage of sodium channel protein beta subunits by BACE1 is linked to epileptic seizures, which contributes to the epileptic behaviors observed in BACE1-null mice [76,77].

APLP1 and APLP2

APLP1 and APLP2 are APP homologues and have been proposed to be substrates of memapsin 2 [59,60]. However, the sequences of APLPs at the proposed memapsin 2 cleavage site are very different from that in APP. Although both APLP1 and APLP2 are cleaved by α- and γ-secretases, only APLP2 is processed by memapsin 2 in cultured cells [59] and this is supported by the absence of APLP2 processing (but not APLP1) in memapsin 2^−/− mice [61].

Beta-galactoside alpha 2,6-sialyltransferase

ST6Gal1 is a Golgi-resident sialyltransferase secreted out of the cell after proteolytic cleavage, which can be enhanced by memapsin 2 overexpression [67,69]. Enriched in liver, secretion of ST6Gal1 from the liver into the plasma is known to be up-regulated during the acute-phase response and lead to increased sialylation of soluble glycoprotein [68]. In a hepatic pathological rat model, simultaneous increases in memapsin 2 mRNA and ST6GalI protein level were detected. Consistently, memapsin 2^−/− mice exhibit decreased level of plasma ST6Gal1, supporting the notion that ST6Gal1 is a physiological substrate for memapsin 2 [78].

P-selectin glycoprotein ligand 1

PSGL-1 modulates leukocyte adhesion to endothelial cells during the inflammatory response in brain and peripheral tissue. In HEK-293 cells, PSGL-1 undergoes proteolytic processing by α-secretase, memapsin 2, and γ-secretase, thereby liberating cleavage products. In primary cells derived from memapsin 2-deficient mice, no cleavage fragment of PSGL-1 was detected, suggesting an in vivo role of memapsin 2 in PSGL-1 proteolysis [70].

Interleukin-1 receptor II

IL-1R2 is a decoy receptor of cytokine IL-1. It can bind IL-1 but is incapable of initiating downstream signals; thus, it is a regulator for IL-1 functions. IL-1R2, a type I transmembrane protein, undergoes shedding of its ectodomain. Increased proteolytic processing and secretion of IL-1R2 has been linked to AD pathogenesis, and a recent report indicates that all three secretases (α-, β-, γ-secretase) appear to cleave IL-1R2 [72].

Lipoprotein receptor-related protein

LRP, a type I integral membrane protein, mediates the endocytosis of an array of proteins including APP and apolipoprotein E that are known to be linked to the development of AD [79]. LRP has been shown to associate with memapsin 2 in cultured cells resulting in the processing of LRP light chain [71].

The above listed membrane proteins have been shown to be cleaved by memapsin 2 in cultured cells. However, many of these findings have not been substantiated in animal experiments. Thus, whether they all are memapsin 2 substrates in vivo should be taken with caution. In cellular experiments both the protease and the potential substrates are overexpressed, thus, creating an enhanced opportunity for cleavage. For example, although memapsin 1 (BACE2) cleaves peptide bonds between Phe19 and Phe20, and Phe20 and Ala21 in the Aβ domain under physiological conditions, overexpressed memapsin 1 can also cleave APP at the β-secretase site in cultured cells [80–82]. However, mice with memapsin 2 gene deletion, but memapsin 1 expression, produce virtually no Aβ [23,83,84], suggesting that memapsin 1 does not play a significant role for APP processing in vivo.

Memapsin 2 Cleavage Predictive Model

From the kinetic data on the complete memapsin 2 subsite specificity [53,54], an empirical algorithm has been developed [54]. In this predictive model, the likelihood of memapsin 2 cleavage at any position of amino acid sequence was calculated from the identities of the residues in six subsites (P4, P3, P2, P1, P1′, and P2′) and expressed as the preference constants, a term which correlates to the kinetic constant k_cat/K_m. As a test, the comparison of actual and predicted hydrolytic efficiency of 12 sequences derived from protein substrates was found to have excellent agreement with a correlation coefficient of 0.97 (Fig. 6). The good correlation appears to support a basic assumption of the model namely the binding of each subsite is largely independent from each other.

Figure 6.

Comparison between the memapsin 2 hydrolytic efficiency and the hydrolytic efficiency calculated using the predictive model The calculated relative k_cat/K_m of different substrates is plotted to the relative k_cat/K_m of different substrates determined by experiments. The correlation coefficient for the predicted data to experimental data is 0.97.

This predictive model may be useful in several areas. As the list of memapsin 2 protein substrates grows longer, it will be of interest to determine the locations of the cleavage sites in new protein substrates. It is also scientifically interesting to learn the efficiency of these cuts in order to assess their physiological consequences. Such data are usually generated from protein chemistry and kinetic studies that are tedious to obtain. Thus, a quantitative predictive model to identify and assess the potential memapsin 2 cleavage sites would aid the initial evaluation and facilitate further experiments. For example, proposed memapsin 2 cleavage sites on substrates APLP1 were extremely inefficient [54], it becomes questionable if the proposed processing is indeed physiologically significant. In addition, this is the first predictive model for an aspartic protease involved in the substrate recognition of many subsites. Similar predictive models may also be established for other proteases with complicated subsite specificities.

Proteolytic Specificities of Memapsin 1 and Memapsin 2

Memapsin 1 (BACE2) is the closest homolog of memapsin 2 and both proteases are type I membrane anchored proteins with highly homologous crystal structures of their catalytic domains. The subsite specificities of these two proteases are very similar [53,85] and memapsin 1 hydrolyzed efficiently some of the same peptide bonds, including the β-secretase cleavage site of APP [80–82]. Although these two proteases are present in many of the same tissues, their physiological activities appear to be distinct. For example, memapsin 1 does not have a significant physiological function in APP process and memapsin 2 does not cleave memapsin 1 substrate Tmem27 in vivo [86]. The mechanism to achieve this distinct activity may lie in the regulation of their expression and subcellular transport, thus this is an area of research that may be of great interest in the future.

Perspectives

Memapsin 2 was identified over a decade ago as the long sought β-secretase that executes the first cleavage on APP leading to the production of Aβ. Because of the importance of this protease in the pathogenesis of AD, memapsin 2 has been intensely studied. As reviewed above, we now understand, in some depth, the structure–function relationships of this protease to its catalytic functions including its subsite specificity. Such knowledge, including the solving of its crystal structure and developing kinetics properties, has contributed to the development of memapsin 2 inhibitors, which remains as an area of high scientific and clinical interests. The progress in this field has led to several clinical trials for memapsin 2 inhibitors. It is hopeful that such efforts will ultimately lead to establishing effective drugs for AD treatment. The structural information of memapsin 2, however, has not been well interpreted in the cellular regulation and trafficking of this protease. Since these areas are intimately related to the normal physiological functions of memapsin 2 and the pathogenesis of AD, future research to unveil such information would be of great scientific and medical interest.

Funding

This work was supported by the grants from the National Institute of Health (AG18933 to J.T.), the Natural National Science Foundation of China (81130042 and 31171323 to L.C.), and the University Innovation Team Support Plan of Liaoning Province (LT2011011 to L.C.).