Matriptase processing of APLP1 ectodomain alters its homodimerization

The amyloid beta peptide (Aβ) is derived from the amyloid precursor protein (APP) by secretase processing. APP is also cleaved by numerous other proteases, such as the type II transmembrane serine protease matriptase, with consequences on the production of Aβ. Because the APP homolog protein amyloid-like protein 1 (APLP1) shares similarities with APP, we sought to determine if matriptase also plays a role in its processing. Here, we demonstrate that matriptase directly interacts with APLP1 and that APLP1 is cleaved in cellulo by matriptase in its E1 ectodomains at arginine 124. Replacing Arg124 with Ala abolished APLP1 processing by matriptase. Using a bioluminescence resonance energy transfer (BRET) assay we found that matriptase reduces APLP1 homodimeric interactions. This study identifies matriptase as the first protease cleaving APLP1 in its dimerization domain, potentially altering the multiple functions associated with dimer formation.

www.nature.com/scientificreports www.nature.com/scientificreports/ matriptase was documented to occur predominantly in epithelial cells of different organs, a growing number of studies have reported expression of matriptase in the brain and/or suggest a role for matriptase in the central nervous system (CNS) [33][34][35][36][37] . Indeed, unregulated matriptase activity disrupts neural tube closure in embryonic mice 33 while its expression in neuronal progenitor cells promotes cell migration and neuronal differentiation 34 . Other studies observed an increase of matriptase transcript levels in a mouse model of AD, especially in activated microglia around amyloid plaques 36,37 .
We have recently shown that matriptase mRNA is expressed in different regions of human brain, with an enrichment in neurons and that it is also present at the protein level in differentiated neurons derived from human induced pluripotent stem cells (hiPSCs) 38 . Moreover we showed that matriptase cleaves the three APP isoforms in the E1 domain at residue 102 38 . This cleavage, although distant from the Aβ sequence, alters the production of Aβ peptide in cellular assays. Since the E1 domain is conserved among members of the APP family and is important for dimerization, we investigated the possibility that matriptase cleaves APLP1 and alters the dimerization/heterodimerization process.
In this study, we show that matriptase interacts with and cleaves APLP1 at a specific residue in its E1 ectodomain. Using a BRET-based assay, we show that addition of matriptase to cells expressing APLP1 disrupts the protein's ability to form homodimers. These events may have important consequences on the physiological and pathological functions of APP family members.

Results
Matriptase interacts with APLP1. In order to explore the possibility that matriptase interacts with APLP1, an APLP1 construct tagged at its N-terminus (extracellular/luminal side) with GFP (containing the signal peptide of containing the mannose-6-phosphate receptor signal peptide) was expressed in HEK293T cells (Fig. 1A) together with matriptase. Immunoblot analysis of cell lysates reveals that GFP-APLP1 is detected as a major 120 kDa form (Fig. 1B), whereas matriptase is detected as a doublet at 95 kDa, which reflects the presence of the full-length 855 amino acid protein and a constitutively processed form at glycine 149 [38][39][40] . Furthermore, matriptase co-immunoprecipitates with GFP-APLP1 but does not co-immunoprecipitate with GFP expressed in the lumen of compartments (GFP containing a signal peptide). The specificity of this interaction was further confirmed using VAMP8-GFP, a transmembrane protein tagged at its luminal side with GFP ( Fig. 1). Finally, a specific co-immunoprecipitation between matriptase and APLP1 was also observed using an APLP1 construct tagged at its C-terminus with Flag ( Supplementary Fig. S2). Together, these results indicate that matriptase interacts in cellulo with APLP1.
To determine whether the co-immunoprecipitation observed between APLP1 and matriptase is a consequence of direct interactions, we then performed in vitro GST pull-down experiments (Fig. 2). N-terminal GST chimeric constructs corresponding to the extracellular or intracellular domains of APLP1 ( Fig. 2A) were expressed www.nature.com/scientificreports www.nature.com/scientificreports/ in bacteria. The purified recombinant GST proteins were then incubated with in vitro translated, 35 S-labeled matriptase and GST-bound proteins were detected by autoradiography. A robust signal was obtained when matriptase was incubated with the GST-APLP1 N-term protein while a weak signal was detected with GST alone or with GST-APLP1 C-term (Fig. 2B). Densitometric analysis revealed a statistically significative difference using GST-APLP1 N-term protein when compared to GST alone (Fig. 2C). These results demonstrate that a direct interaction occurs between matriptase and the ectodomain of APLP1 but not with GST-APLP1 C-term.
In cellulo cleavage of APLP1 by matriptase. When APLP1 was co-transfected with matriptase, a GFP-APLP1 fragment migrating at 33 kDa was detected in the cell lysate (Fig. 1B). This 33-KDa fragment would correspond to the GFP tag (25 kDa) and a portion of the N-terminal domain of APLP1 (8 kDa). To further characterize cleavage of APLP1 by matriptase, HEK293T cells were transfected with GFP-APLP1 in combination with matriptase or with a catalytically inactive form of matriptase (S805A). Physiologically, substrate cleavage by matriptase occurs at the plasma membrane or in the extracellular medium 32,41 . Therefore, cell lysates and culture medium were harvested to detect APLP1 cleavage products. The GFP-tagged APLP1 fragment of 33 kDa was detected in both cell lysates conditioned medium of cells transfected with wild-type matriptase but not in cells transfected with the inactive matriptase S805A (Fig. 3A). This result demonstrates that APLP1 is processed by catalytically active matriptase into soluble forms found in the media of transfected cells. This cleavage by catalytically active matriptase was also validated by co-expressing HAI-1, the physiological inhibitor of matriptase. Indeed, no cleavage products are detected when GFP-APLP1 and matriptase are co-transfected with HAI-1 ( Supplementary  Fig. S5). Finally, when matriptase-2, a protease closely related to matriptase, is co-expressed with GFP-APLP1, no cleavage products are observed ( Supplementary Fig. S6) supporting the idea of a specific cleavage by matriptase.
Because matriptase is shed in the extracellular space as a soluble form 32,41 , we investigated whether exogenous addition of an enzymatically active soluble form of matriptase cleaves GFP-APLP1 expressed in HEK293T cells. A GFP-tagged APLP1 fragment of 33 kDa was detected in the conditioned culture medium of GFP-APLP1 expressing cells incubated with recombinant soluble active matriptase but not with soluble inactive matriptase S805A (Fig. 3B). Altogether, these results suggest that active matriptase can cleave GFP-APLP1 at the cell surface to generate soluble APLP1 forms. www.nature.com/scientificreports www.nature.com/scientificreports/

Identifying matriptase cleavage sites on APLP1.
To identify the precise matriptase cleavage sites within APLP1 extracellular domain, mass spectrometry (MS) analysis was performed on APLP1 fragments generated following in vitro incubation of purified GST-APLP1 ectodomains with or without recombinant soluble matriptase. A major cleavage site was found at arginine 124 of APLP1 ( Fig. 4A and Supplementary Fig. S7). This cleavage site is predicted to yield N-terminal fragments with molecular weights of 9.6 kDa. Taking into account the molecular weight of the N-terminal GFP tag (25 kDa), this is consistent with the molecular weight of the fragment observed in Fig. 3 (33 kDa). Interestingly, similar to APP 38 , the matriptase cleavage site was found in the E1 domain of APLP1. In order to confirm this cleavage site, Arg-124 was mutated to Ala (R124A) in GFP-tagged APLP1. HEK293T cells expressing either wild-type GFP-APLP1 or GFP-APLP1-R124A were then incubated with (or without) soluble matriptase and analyzed for the presence of APLP cleavage products in the conditioned medium (Fig. 4B). Expression of the R124A mutant completely abolished the formation of the GFP-APLP1 cleaved fragment indicating that this mutant is resistant to cleavage by matriptase. These results confirmed Arg-124 as the primary matriptase cleavage site in APLP1. Taken together, these results indicate that, similarly to APP, matriptase cleaves in the E1 domain of APLP1.
Matriptase processing of APLP1 alters its homodimerization. APP family members can form homotypic and heterotypic cis and trans dimers mainly through their N-terminal E1 domains 17 . Given that we defined the matriptase cleavage site in the E1 domain of APLP1, we next examined whether this cleavage impairs www.nature.com/scientificreports www.nature.com/scientificreports/ its dimerization. We set up a novel Bioluminescence Resonance Energy Transfer (BRET2) assay to quantify APLP1 cis-homodimerization 42 . In this assay, when GFP10 (acceptor molecule) is in close vicinity to Renilla luciferase II (RlucII, donor molecule) in the presence of its substrate coelenterazine, a BRET2 signal is generated and can be measured (Fig. 5A). To detect APLP homodimerization, GFP10 and RLucII moieties were fused to the C terminus of APLP1 (Fig. 5B). The BRET2 pair, APLP1-GFP10/-RLucII exhibits a saturation curve for a fixed concentration of donor (RLucII) and an increasing concentration of the acceptor, GFP10 (Fig. 5C). This result reflects a typical saturation profile consistent with a specific interaction between APLP1 monomers. On the other hand, this saturation plateau is not reached using an increasing concentration of the plasma membrane expressed angiotensin II type 1 receptor, (AT1R-GFP10) that does not interact with APLP1. These results confirm the homodimeric interaction for APLP1 proteins and validate our BRET2 approach to monitor dimerization.
We then assessed the impact of matriptase cleavage on the homodimeric interaction of APLP1 by expressing fixed concentrations of the APLP1 BRET constructs with increasing concentration of transfected active (WT) or inactive (S805A) matriptase. BRET2 signal reduction is expected if dimerization is impaired. Indeed, a BRET2 signal decrease was detected with increasing concentrations of active (WT) or inactive (S805A) matriptase for APLP1 homodimers (Fig. 5D). This suggests that interaction between matriptase and APLP1 affects the ability of APLP1 to remain as dimers. The effect on dimerization seen with active and inactive matriptase is in accordance with co-immunoprecipitation assays demonstrating an interaction between APLP1 and both matriptase forms. Furthermore, analyzing the dose dependent BRET2 response curves, we observed a leftward shift of the curves generated with active matriptase compared to the inactive protease. These results suggest that the catalytically active matriptase was more potent at disrupting APLP1 homodimers than inactive matriptase, probably due to its ability to cleave the E1 domain of APLP1. Statistical analysis of the data set of each curve show a statistical difference between the two curves.

Discussion
For several years, research efforts have focused on the general understanding of the pathophysiology of AD and on APP processing for the development of therapeutic approaches. In this context, several novel proteases have been recently identified to cleave APP and to alter Aβ production 43 . However, much less is known about the processing of APP homologs APLPs and whether their proteolytic fragments have biological functions in the CNS or in peripheral tissues. Here, we identify matriptase as a novel protease that cleaves APLP1 at a specific residue in its E1 ectodomain, which impacts its ability to form homodimers, a key molecular event regulating its functions. www.nature.com/scientificreports www.nature.com/scientificreports/ Matriptase is one of the most studied TTSPs and its expression was considered limited to epithelial cells. However, several recent studies have shown it is expressed in human brain regions 38 , neuronal progenitor cells 34 , microglia, activated astrocytes around amyloid plaque and neuronal IPSC cells [36][37][38]44 . We and others previously showed that matriptase cleaves APP 38,45 , a novel substrate for matriptase. Here, we find that matriptase interacts with and cleaves APLP1. Physiologically, this interaction and cleavage could occur in the brain where expression of matriptase is low but expression of APLP1 is high as well as in peripheral tissues, such as pancreas, where both proteins are expressed 6 .
We observed that matriptase interacts with and cleaves APLP1 in cellulo. Interaction with matriptase has been confirmed in vitro for APLP1's ectodomain. Accordingly, mass spectrometry analysis and mutagenesis studies have identified Arg124 within the sequence GSRR 124 ↓SGSC in the E1 domain of the extracellular region of APLP1 as the major matriptase cleavage site. To our knowledge, this specific arginine has never been identified as a processing site in APLP1 by other proteases. Interestingly, we previously reported that matriptase cleaves at Arg102 of APP's E1 domain 38 . Sequence alignment of matriptase cleavage sites of APP and APLP1 highlighted their location at the same position in the GFLD region of their E1 domains ( Supplementary Fig. S8), indicating that this well-exposed region is favorable for matriptase processing. In summary, we now report that both APP and APLP1 are cleaved by matriptase in a conserved region of their E1 domains.
Compared to APP, much less is known about the proteolytic cleavage of APLPs and the functional consequence of its processing. APLPs, like APP, undergo shedding of their ectodomain by αand β-secretases, followed by γ-secretase-mediated intramembrane proteolysis [46][47][48] . Although the physiological role of the shed forms of APP is starting to emerge, the function of shed APLPs is less well defined. However, different forms of soluble APLPs have been detected in human cerebrospinal fluid [49][50][51] and it has been shown that a soluble N-terminal fragment of the closely related member APLP2 binds Death receptor 6 (DR6) 52 . In addition to secretases, another protease, rhomboid protease RHBDL4, cleaves multiple sites within APLP ectodomains resulting in several N-and www.nature.com/scientificreports www.nature.com/scientificreports/ C-terminal fragments 53 . These cleavages occur in the lumen of the endoplasmic reticulum and regulate cell surface APLP levels 53 . Matriptase is thus one of the few proteases reported to cleave APLP1 ectodomains, potentially generating fragments that can have biological activities or altering processing by other proteases.
We report here that inactive matriptase can alter APLP1 dimerization and furthermore, that the active form is even more potent at disrupting this dimer. This suggests that modulation of APLP1 dimerization involves two distinct mechanism i.e. through direct protein-protein interaction and proteolytic cleavage in the E1 domain. The E1 domain of APP family members has been previously reported to mediate homo-and heterophilic interaction of APP/APLPs 17,23 . Deleting the GFLDs of APP and APLP2 inhibited cis homo-and hetero-dimerization of APP and APLP2 but only marginally affected APLP1 dimerization 23 , suggesting that the E2 domain is crucial for initiating APLP1 interaction 21 . The BRET homodimerization assay indicated that cleavage by matriptase in the GFLD region reduces the capacity of APLP1 to form cis homodimers. Although mass spectrometry analysis indicates that the main cleavage by matriptase occurs in the GFLD domain, we cannot exclude cleavage at other sites or that indirect effects of matriptase activity are also involved in this inhibition of APLPs cis interaction. Moreover, given that the E1 domain is also a key interface for APP/APLPs trans dimerization 17 , we expect that matriptase cleavage would affect trans interaction of APP family members. Cis and trans homo-and hetero-dimerization of APP family members has been involved in multiple functions including cell signaling 54 , neurite outgrowth 55 , neuronal cell adhesion 21 , trans-cellular adhesion 17 , synapse formation and function 15,56,57 , as well as Aβ peptides production 22,23,25 . Thus, membrane-bound or soluble matriptase cleavage of APP family members could regulate their functions by altering their capacity to dimerize in the brain or in peripheral tissues.
In conclusion, this study identifies matriptase as a novel APLP1 cleaving protease altering its ability to form dimers, an important feature that is known to affect its functions. Importantly, evaluating the impact of those events in a physiological context would provide novel insights into how APLP1 function is regulated in vivo and to which extent this regulation could influence pathological states.

Methods
Antibodies and reagents. Anti-GFP rabbit polyclonal antibodies (pAbs) and anti-human matriptase pAbs were purchased from Clontech Molecular Probes (Eugene, OR, USA), and Bethyl Laboratories (Montgomery, TX) respectively and were used as described previously 38 .
BRET-based biosensor assays. HEK293T  Co-immunoprecipitation and immunoblotting. HEK293T cells were plated in 60-mm culture dishes and transfected with indicated constructs and co-immunoprecipitation assays were performed as previously described 38 . Briefly, after 48 h, cells were washed twice with PBS and lysed in 50 mM Tris buffer (pH 7.4) containing 150 mM NaCl, 1% triton x-100, and protease inhibitors for 1 h at 4 °C and were then centrifuged 20 min at 15,000 g. Supernatants were incubated with GFP-TrapA (Chromotek, Germany) overnight at 4 °C, washed with 50 mM Tris buffer (pH 7.4) containing 150 mM NaCl, 1% triton x-100, and protease inhibitors three times. Bound immune complexes were boiled in Laemmli sample buffer and proteins separated by SDS-PAGE. Immunoblotting was performed as described previously 38 . DnA constructs. Mammalian expression vectors APLP1 were kindly provided by Dr. Gerhard Multhaup (McGill University, Montreal). APLP1 was subcloned in pCMV5 downstream of the signal peptide (SP) of mannose-6-phosphate receptor and GFP (as previously described 38,58 ) to produce an APLP1 tagged at its N-terminus (luminal/extracellular side) with GFP. Using this plasmid construction as a backbone, a SP-GFP control (GFP expressed in luminal compartment) was produced by introduction of a stop codon between GFP and APLP1 and the mutant GFP-APLP1-R124A was produced by substitution of an Arg to Ala at position 124 in GFP-APLP1 sequence. Both constructs were obtained using the QuikChange Lightning site-directed mutagenesis kit (Agilent Technologies). APLP1 containing the N-terminus (residues 39-581) and C-terminus (residues 597-650) coding sequences were subcloned in pET41a (Novagen). For BRET assays, APLP1 coding sequence was inserted upstream of GFP10 and RLucII in a pIRES-Hygromycin backbone using the NEBuilder HiFi DNA Assembly cloning kit as recommend by the manufacturer's instructions (New England Biolabs, MA, USA). Recombinant matriptase (residues 596-855) construction used for bacterial expression (pQE30 vector, Qiagen, Mississauga, ON, Canada) and S805A-matriptase-pcDNA3.1 have been described previously 39,59 . GFP-VAMP8 cDNA was kindly provided by Dr. Steve Jean (Université de Sherbrooke) and has been previously described 60 . All constructs were submitted to nucleotide sequencing before being used.
Glutathione S-transferase pull-down assays. GST pull down assays were performed as described previously 38 . GST fusion proteins expressed in Escherichia coli BL21 were purified on glutathione-Sepharose 4B beads (Pharmacia, Piscataway, NJ, USA) according to the manufacturer's instructions. Using the TNT T7 rabbit reticulocyte Quick Coupled Transcription/Translation system (Promega, San Luis Obispo, CA, USA), the www.nature.com/scientificreports www.nature.com/scientificreports/ 35 S-labelled in vitro translation products of pcDNA3.1-human APLP1 and matriptase were prepared in the presence of [ 35 S] EasyTag EXPRESS labelling mix (73% Met/22% Cys; 41,000 Ci/mmol, Perkin Elmer). The in vitro translated products were incubated with 10 μg of purified GST or GST-fusion protein in 20 mM Tris-HCl buffer (pH 7.4) containing 150 mM NaCl, 1% triton x-100, protease inhibitors for 2 h at 4 °C. Beads were washed four times with the same buffer. Bound proteins were eluted with Laemmli buffer, resolved by SDS-PAGE and visualized by autoradiography.
Mass spectrometry analysis. GST fusion proteins were expressed in Escherichia coli BL21 and purified on glutathione-Sepharose 4B beads (Pharmacia, Piscataway, NJ, USA) according to the manufacturer's instructions. For 2 hours at 37 °C, bound proteins were incubated in 100 μL of 100 mM Tris-HCl (pH 8.5) containing or not 100 nM of recombinant soluble WT matriptase. The supernatant was then collected, lyophilised and suspended in 25 μL of 10 mM HEPES/KOH (pH7.5). As previously described 38 , mass spectrometry analysis was performed using an Orbitrap QExactive mass spectrometer (Thermo Fisher Scientific) and data were processed, searched, analyzed, and quantified using the MaxQuant software package version 1.5.2.8 employing the Human Uniprot database (16/07/2013, 88,354 entries).

Treatment of cells with Recombinant Matriptase.
Cleavage assays in cellulo using purified recombinant matriptase were performed as described previously 38 . Briefly, culture medium of HEK293T was removed and replaced with 2 mL of serum-free HCELL-100 medium (Wisent, St-Bruno, QC, Canada) containing 5 nM of recombinant soluble human WT matriptase or mutant S805A. The conditioned medium was collected after 36 h incubation and concentrated with Amicon Ultra Centrifugal filters 3,000 NMWL (Merck Millipore Ltd., T., C., Co Cork IRL). Both cell lysates and concentrated conditioned medium were then analyzed by SDS-PAGE and immunoblotting.

Statistical analysis.
Experiments were performed at least in triplicate and a one sample t-test with a hypothetical value set to 1 for GST alone was used for GST-Pull Down quantification analysis. A two-way ANOVA with a Sidak's multiple comparisons test was used in BRET assay.