| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
J. Biol. Chem., Vol. 278, Issue 38, 36032-36040, September 19, 2003
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
-ALZHEIMER'S PRECURSOR PROTEIN*
From the
Department of Biochemistry, the ¶Department of Neurology, and ||Center for Neurodegenerative Diseases, School of Medicine, Emory University, Atlanta, Georgia 30322-3050 and IGEN International Corporation, Gaithersburg, Maryland 20877
Received for publication, February 14, 2003 , and in revised form, June 25, 2003.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
-APP) to regions of the Golgi/trans-Golgi network, and can co-immunoprecipitate clathrin. We demonstrate that MINTs bind Arfs through a region of the PTB domain and the PDZ2 domain, and Arf-MINT interaction is necessary for the increased cellular levels of -APP produced by MINT overexpression. Knockdown (small interference RNA) experiments implicate -APP as a transmembrane cargo protein that works together with MINTs. We propose that MINTs are a family of Arf-dependent, vesicle-coat proteins that can regulate the traffic of -APP. |
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
There are four characteristics that all Arf-dependent coats share: direct binding to Arf·GTP, recruitment to the Golgi membrane in a brefeldin A-sensitive manner (which further indicates Arf involvement), binding of cargo proteins, and interactions with accessory proteins, e.g. clathrin (4, 7, 12–15). To test whether additional coat proteins exist that act at the Golgi, we focused on the direct binding to Arfs as the common feature of such proteins. We report here the identification of the three MINT family proteins as functional homologs to GGAs and AP complexes as coat proteins that regulate aspects of vesicle traffic.
MINTs are a family of three proteins. MINT1 and MINT2 are expressed only in neuronal tissues (16), whereas MINT3 is ubiquitously expressed (17). All MINTs share a conserved central PTB domain and two C-terminal PDZ domains (16–19). Although MINT1 was originally described and named for its ability to bind Munc18, a neuronal protein acting at the synapse, it also binds a number of other proteins, most notably the -Alzheimer's protein (-APP), neurexins, and a number of transmembrane receptors (16, 18–28). Mammalian MINT1 and its ortholog in Caenorhabditis elegans, LIN-10, are also found in a stable complex with Cask/LIN-2 and Velis/LIN-7 (20, 29, 30). MINT2 lacks the N-terminal Cask binding domain but binds XB51 and NF
B and can also influence -APP processing (23, 26), whereas MINT3 lacks both the Munc18- and the Cask binding domains and has been less well studied. Except for Munc18 and Cask, binding to other proteins occurs through the PTB and/or PDZ domains of MINTs. The best-studied activities of MINT proteins include roles in traffic and/or processing of -APP and the EGF receptor. MINT family proteins coordinately increase cellular -APP levels and half-life and alter its processing (31–33) when overexpressed in mammalian cells, and MINT1/LIN-10 is required for the proper localization of the EGF receptor (LIN-23) to the basolateral surface of the vulva progenitor cell in C. elegans (34, 35). The functional importance of monomeric MINTs, heterodimeric MINT-Munc18, or the heterotrimeric MINT·Cask·Velis complex (and the possibility of interplay between these three states of MINTs) to neuronal or more general eukaryotic cells is not known.
Using a two-hybrid strategy, we identified MINTs as binding partners of activated Arf3. Additional data confirmed that MINTs bind directly and preferentially to activated Arfs and that this interaction occurs through their PTB and PDZ2 domains. Treatment with brefeldin A, a specific inhibitor of Arf exchange factors, rapidly reversed MINT localization to Golgi membranes. Knockdown in the expression of either MINT3 or -APP results in commensurate changes in the distribution of the other protein, extending the functional linkage of Arf to MINT to -APP. Thus, these data reveal that MINTs share all the characteristics of Arf-dependent coat proteins and -APP is implicated as a transmembrane cargo for MINT vesicles. The implications of novel cellular roles for MINTs and -APP in vesicle traffic to pathological conditions leading to Alzheimer's disease are discussed.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
-APP antibodies used in this study were mouse monoclonals 26D6 (Sibia) and 22C11 (Chemicon), raised against sequences in the A and N-terminal portions, respectively. Mouse monoclonal antibodies raised against MINT1 (#M75920, BD Transduction Laboratories), MINT2 (#M76120, BD Transduction), MINT3 (#M93620, BD Transduction), and HA epitope (12CA5, BAbCO), or rabbit polyclonal antisera directed against giantin (BAbCO), 
-adaptin (M-300, Santa Cruz Biotechnologies), clathrin heavy chain (H-300, Santa Cruz Biotechnologies), or rabbit IgG (1–5006, Sigma Chemical Co.) were each obtained from commercial sources. The -COP rabbit polyclonal antibody was raised against "EAGE peptide" as previously described (36). Co-immunoprecipitation—COS-7 cells were transiently transfected using FuGENE 6 (Roche Applied Science), according to the manufacturer's specifications. Twenty-four hours after transfection, cells were collected and lysed with lysis buffer (20 mM Tris, pH 7.4, 100 mM NaCl, 1 mM MgCl2, 1 mM CaCl2, 10% glycerol, 1% Nonidet P-40) that also contained a protease inhibitor mixture (Sigma). Nuclei and insoluble material were removed from the cell lysate by centrifugation for 5 min at 14,000 x g. The supernatants were collected and incubated in fresh tubes with antibody for 4 h at 4 °C, before the addition of protein-G-Sepharose beads and a further 1-h incubation. The beads were then collected by centrifugation in a microcentrifuge and washed three times with phosphate-buffered saline (PBS) containing 1% Nonidet P-40. SDS sample buffer (37) was added to the beads, and the samples were boiled for 5 min before loading onto polyacrylamide gels.
Indirect Immunofluorescent Cell Staining—Cells were grown on coverslips and then fixed with 3.7% formaldehyde for 20 min at room temp before being permeabilized for 10 min with 0.2% saponin in PBS containing 10% goat serum. After permeabilization, the cells were incubated with PBS containing 0.2% saponin, 10% goat serum, and primary antibodies for 1 h at room temperature. The coverslips were washed three times with PBS containing 10% goat serum before a second incubation with Alexa 488 anti-rabbit IgG (Molecular Probes) and Alexa 594 anti-mouse IgG (Molecular Probes) in PBS/0.2% saponin/10% goat serum for 1 h. The coverslips were washed three times with PBS/10% goat serum, washed twice with PBS, and mounted in PBS/0.1 M N-propyl gallate/50% glycerol.
Arf-effector Binding Assay—Direct binding of an effector to Arf often results in a change in the steady-state binding of GTP to Arf (38), and therefore radioligand binding can be used as an assay for effector binding (39, 40). Binding of [-35S]GTPS to Arf was determined as previously described (38, 41, 42). Arf3 (1 µM) was incubated with GTPS (10 µM,[-35S]GTPS, 2500 cpm/pmol) and GST-MINT2 or POR1 (5 µM) at 30 °C in binding buffer (20 mM HEPES, pH 7.4, 1 mM EDTA, 1 mM dithiothreitol, 100 mM NaCl, 0.5 mM MgCl2, 3 mM L-
-dimyristoyl phosphatidylcholine, 0.1% sodium cholate, and 100 µg/ml bovine serum albumin). Aliquots (10 µl) were diluted into 2 ml of ice-cold buffer (25 mM Tris-Cl, pH 7.4, 100 mM NaCl, 10 mM MgCl2, and 1 mM dithiothreitol). Samples were rapidly filtered onto 25-mm BA85 nitrocellulose filters, and retained radionucleotide was determined by liquid scintillation counting.
Small Interference RNA—Knockdown in the expression of either human -APP or MINT3 was achieved by transfection of cells with pSUPER-based plasmids containing inserts that generate snap-back RNAs 19 nucleotides long with a loop of 9 nucleotides, as described in Brummelkamp et al. (43). Sequences targeted were in the middle of the open reading frames in -APP (5'-GAAGGCAGTTATCCAGCAT-3') or MINT3 (5'-GATGCTCTGCCACGTATTC-3'). Each sequence is 100% identical to regions of the open reading frame of the targeted messages, and BLAST searches did not find those sequences in any other message.
HeLa cells were transfected using LipofectAMINE reagents (Invitrogen), according to the manufacturer's instructions. After a series of time course experiments, 4 days was chosen as the optimal time for knockdown in protein expressions. Cells were then trypsinized, washed in PBS, and lysed in buffer containing 25 mM HEPES, pH 7.4, 100 mM NaCl, and 0.5% Triton X-100. Protein concentrations of whole cell lysates were determined using the Bio-Rad protein assay. Parallel coverslips were fixed and prepared for immunofluorescence studies, as described above.
Brefeldin A Treatment—Cells were treated with 10 µM brefeldin A for 0, 2.5, 5, and 10 min before fixing with 3.7% formaldehyde. The cells were stained as indicated.
Yeast Two-hybrid Assay—A human, fetal brain cDNA library was screened with an activating mutant of human Arf3 (Arf3-Q71L, I74S) as bait using a two-hybrid screening protocol as previously described (4). -Galactosidase and histidine auxotrophy assays were performed.
Reverse two-hybrid screening was performed to identify point mutations in MINT2 that result in the loss of binding to Arf3-Q71L, I74S. Random mutations were generated in the MINT2-(500–749) insert by polymerase chain reactions performed under conditions of reduced stringency and gap repair of yeast vectors with PCR products. Yeast two-hybrid assays and selection of colonies made deficient in -galactosidase activity were performed as previously described (39, 44). Mutated inserts were sequenced and moved into mammalian expression vectors for further analysis. Sixteen mutants were identified that expressed the MINT2 fragment to comparable levels as starting material. Three different mutations in PDZ2 were identified in this way, and those mutations were then moved into the full-length MINT2 open reading frame in the pcDNA3.1 mammalian expression vector. These point mutant full-length MINT2 inserts were also sub-cloned into pACT2, to confirm their loss of binding to activated Arfs.
Membrane/Vesicle Preparations—Clathrin-coated vesicles (CCVs) were prepared from fresh rat liver as described previously (45, 46). Briefly, a microsome fraction was prepared from fresh rat liver by differential centrifugation. Crude CCVs were isolated from this fraction by centrifugation in 12.5% Ficoll 400/12.5% sucrose and further purified using discontinuous sucrose gradients. TGN membranes were also prepared from fresh rat liver, using a protocol modified from Seaman et al. (47). Briefly, a post-nuclear supernatant (PNS) was purified from rat liver homogenate by centrifugation (1500 x g for 10 min) and loaded onto a step gradient consisting of 15 ml of 6% Ficoll layered over 10 ml of 18% Ficoll layered over 4 ml of 45% Nycodenz, all dissolved in 0.25 M sucrose, 10 mM HEPES, pH 7.4, 1 mM MgCl2. The gradient was spun at 100,000 x g in a Beckman SW28 rotor for 2 h at 4 °C. TGN membrane fractions were collected at the 6–18% interface.
MINT2 Truncations—Primers were designed to amplify a series of MINT2 truncations using the polymerase chain reaction. After sequencing, these truncated MINT2 mutants were sub-cloned into pACT2 for use in yeast two-hybrid assays.
Image Acquisition—Images were acquired using a Zeiss LSM 510 Axiovert 100M confocal scanning laser fluorescence microscope with x63 optics. The graphics in the figures were assembled using Adobe Photoshop.
Replication of Experiments—Every experiment described herein was repeated at least twice with essentially the same results. Most were repeated more than twice.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
-galactosidase assays but each only in the presence of the Arf3-Q71L, I74S, and not with an unrelated bait. The inserts from the three positives were sequenced and found to encode different lengths of a single open reading frame, encoding human MINT2 (residues 252–749, 475–749, and 500–749). There are three human MINT proteins (MINT1–3, also known as X11, -, and - or X11, X11-like, and X11-like2; accession numbers Q02410
[GenBank]
, Q99767
[GenBank]
, and O96018
[GenBank]
, respectively) that all share a conserved central PTB domain and two C-terminal PDZ domains (16, 17, 19). The shortest insert identified from the library screen (MINT2-(500–749)) encoded both PDZ domains plus an additional 30 residues at the C terminus of the (
200 residues) PTB domain. The domain organization of the MINT proteins and some of the N-terminal truncation mutants are depicted in Table I. Although MINT1 and -2 are the most highly related family members and each is expressed only in neurons, MINT3 is ubiquitously expressed and shorter (575 residues), lacking both the Munc18 and Cask binding domains from the N terminus (17). To test whether MINT1 and MINT3 also bind to activated Arfs, constructs homologous to the shortest MINT2 insert were made and, like MINT2, bound activated Arf3 and Arf4 but not the unactivated Arfs, activated Arl1 (see Table I), or unrelated bait proteins (data not shown). Activated Arfs also bound full-length MINT2 (see Fig. 2) and binding does not require the second mutation, Arf3-I74S. Thus, the interactions between Arfs and MINTs are promoted by GTP binding and are different from those previously described in that mutation of isoleucine 74 does not interfere with binding. Representatives of both class I (Arf1–3) and class II (Arf4 and -5) Arfs bind MINTs, suggesting that it is likely a feature of all the soluble Arf (Arf1–5).
|
|
MINTs Bind Arfs Directly—Preference for binding activated Arfs over wild type proteins in two-hybrid assays is a strong indication that the binding is GTP-dependent (4, 39, 49, 50). To determine whether MINT2 binds Arf directly, we first expressed the shortest MINT2 insert pulled from the library screen (MINT2-(500–749)) in bacteria as an N-terminal fusion protein with glutathione S-transferase (GST-MINT2-(500–749)). Many Arf effectors have been shown to increase the binding of GTPS to Arf, probably by increasing the affinity for activating nucleotides (38). Addition of the purified protein (5 µM) to the GTPS binding assay resulted in 100% increase in steady-state level of bound GTPS to Arf3 (1 µM, Fig. 1). This is comparable to the effects of any of the GGAs on Arf in this assay (38). Recombinant POR1 was used as a positive control in this study, because it has previously been shown to have this activity (38). These results confirm that the binding of MINT2 to Arf is direct and that neither the Q71L nor the I74S mutations are needed to bind Arf.
|
MINTs Bind Arfs through PDZ2 and PTB Domains to Alter the Processing of -APP—To better define the one or more regions of MINTs that bind activated Arfs, a series of truncation mutants were generated and assayed in yeast two-hybrid assays. All three MINT2 inserts cloned from the cDNA library produced very similar activities in assays for -galactosidase activity (Fig. 2) and histidine auxotrophy (data not shown). Truncation of the shortest clone, MINT2-(500–749), to delete the entire PTB domain, MINT2-(521–749), resulted in the loss of -galactosidase activity and a weakened histidine auxotrophy. Further truncation of the linker region between the end of the PTB domain and the start of PDZ1 caused the complete loss of two-hybrid activities (see Fig. 2). PDZ2 was also found to be essential for interaction with activated Arf as truncation of only PDZ2 from the C terminus, to yield MINT2-(500–644), was sufficient to completely abrogate binding. These data implicate the C-terminal portion of the PTB domain and PDZ2 in the binding of MINTs to activated Arfs.
To determine residues and domains in MINT2 that are critical to the binding of Arf, we performed reverse two-hybrid assays and screened for the loss of binding to Arf3-Q71L, as described previously (39, 50). Random mutations were introduced into the open reading frame of MINT2-(500–749) and inserted into the appropriate two-hybrid vector by PCRs under conditions of reduced stringency followed by gap repair of plasmids in yeast. Sixteen MINT2 mutants exhibited diminished interaction with Arf3-Q71L yet were expressed in yeast to levels comparable to parental control. These mutants were sequenced to identify single mutations responsible for loss of binding to Arf. Three such mutations in the PDZ2 domain (MINT2-E701K, MINT2-N721D, and MINT2-I715T) were selected for further testing. Mammalian expression and yeast two-hybrid vectors were constructed that drove expression of full-length human MINT2 carrying each of these point mutations. Although full-length MINT2 yields lower -galactosidase activity than MINT2-(500–749) in two-hybrid assays, in each case the single mutation in context of the full-length MINT2 caused a further loss in binding to Arf3-Q71L in two-hybrid assays (see Fig. 2). This observation supports the earlier conclusion that PDZ2 is required for binding activated Arf.
Despite being found bound to either Munc18 or in the heterotrimeric MINT·Cask·Velas complex, there is only one published biological effect of any MINT; overexpression of MINT1 or MINT2 increased the half-life and cellular levels of -APP and decreased its processing into the neurotoxic A peptides (31, 33, 51). To begin to assess the biological importance of the Arf-MINT interaction, we asked if there existed a correlation between the expression of mutant MINT proteins that do not bind Arfs and cellular -APP levels. As previously shown (32), overexpression of MINT2 in HEK293 cells increased cellular levels of -APP (see Fig. 3, compare lanes 1 and 2). In contrast, three of three point mutations in PDZ2 of MINT2 that exhibit reduced binding to Arf3-Q71L also produced clearly blunted effects on cellular -APP levels (Fig. 3, compare lane 2 to lanes 3–5), despite the fact that similar levels of MINT2 and mutants were expressed (Fig. 3, lower panel). These data are consistent with MINT binding to Arfs playing a role in the regulation of -APP level, half-life, and/or processing. The mutant MINT2 proteins were able to increase cellular -APP levels to above controls. At this point, we cannot determine whether this is the result of the partial retention of Arf binding (that cannot be quantified in yeast two-hybrid assays) or whether there also exists an Arf-independent effect of MINT2 on -APP.
|
MINTs Localize to the Golgi in a Brefeldin A-sensitive Manner—We used indirect immunofluorescence to determine the locations of endogenous MINT2 in primary cultures of rat cortical neurons and of MINT3 in a number of mammalian cells. Staining of MINT2 and MINT3 was predominantly perinuclear and highly overlapping with that of -COP (Fig. 4, A and C), mannosidase II (Fig. 4, E and G), Arf, and TGN38 (data not shown), indicating a Golgi location. Several Arf effectors, particularly coat proteins, are recruited to membranes as a result of their direct binding to Arfs. We used brefeldin A, a specific inhibitor of Arf activating proteins, to determine if this was the case for membrane-associated MINTs. Within 2 min of exposure to brefeldin A, both Arfs (data not shown) and MINTs (Fig. 4, B and F) were released from membranes and became more diffuse. At this early time point there is no discernible change in Golgi morphology, as visualized with the marker of the Golgi lumen, mannosidase II (Fig. 4H). These results show that the localization of MINT2 and MINT3 to the Golgi in live cells is a result of their direct binding to activated Arf. Direct binding to Arf, Golgi localization, and brefeldin A sensitivity are all features shared by the family of Arf-dependent coat proteins, termed GGAs, as well as adaptins (e.g. AP-1), and COPI (4, 14, 15, 32).
|
MINTs Can Co-immunoprecipitate Clathrin—Because some Arf-dependent adaptor/coat proteins, including GGAs and AP-1 (but not COPI), have been shown to co-immunoprecipitate clathrin (53–55) we tested whether or not MINTs share this property. Existing MINT antibodies did not work well in immunoprecipitation trials, so MINT2 or MINT3 were expressed as N-terminal HA fusion proteins in COS-7 cells. Indirect immunofluorescence revealed that these fusion proteins localized to the Golgi, comparably to the endogenous proteins, at early time points (data not shown). AP-1 and COPI were used as positive and negative controls for clathrin binding, respectively, as each is an Arf-dependent adaptor/coat that localize to Golgi membranes. As expected, antibodies to -adaptin, a subunit of AP-1, co-immunoprecipitated clathrin (Fig. 5, lanes 2 and 6), whereas those to -COP (Fig. 5, lanes 3 and 7), a subunit of COPI, or an IgG control (Fig. 5, lanes 4 and 8) did not. The larger amount of clathrin brought down with -adaptin antibodies likely reflects the greater abundance of AP-1 in cells. Clathrin was found to specifically co-immunoprecipitate with MINT2 or MINT3 (Fig. 5, lanes 1 and 5, respectively), indicating that MINTs and clathrin interact in cell lysates. Some of the controls are shown below the clathrin blots and demonstrate that the antibodies used to IP the coat proteins are effective and highly specific. Thus, MINTs share all the previously described biochemical features of AP-1 and GGAs as Arf-dependent coat proteins.
|
To further examine the association between MINTs and clathrin, we asked if MINT3 is enriched in clathrin-coated vesicle or TGN preparations. Fresh rat livers were used to prepare TGN and CCV, which were immunoblotted for clathrin, AP-1, COPI, MINT3, and APP (see Fig. 5B). As expected, both clathrin and AP-1 were enriched in TGN membranes over the post nuclear supernatant (PNS) and very highly enriched in CCVs. Although the presence of MINT3 in the TGN preparation was somewhat variable we consistently observed the clear enrichment of MINT3 in CCVs. -APP was enriched both in TGN and CCV preparations, over the PNS. The presence of -COP in our TGN preparation may indicate contamination of the TGN with other Golgi membranes, but it is clear that -COP was not enriched in CCVs. Neither the co-immunoprecipitation results nor the cell fractionation data indicate a direct interaction between MINTs and clathrin, and the absence of an identifiable clathrin binding motif in MINTs is a weak argument against such a direct link. However, the data in Fig. 5 suggest that MINTs function in cells, at least part of the time, on membranes or vesicles containing clathrin.
Localization of MINT3 and -APP Are Mutually Interdependent—With the evidence that MINTs share features of coat proteins with GGAs and AP-1, we next sought to identify transmembrane proteins that serve as cargo to aid in the Arf-dependent recruitment of MINTs to Golgi membranes. -APP is a ubiquitous transmembrane protein that traffics the secretory and endocytic pathways and contains an NPXY sorting motif that has previously been shown to bind the PTB domain of MINT1 (19, 56–59). Traffic and proteolytic processing of -APP are highly regulated, because they result in the production of both the neurotoxic A peptides that accumulate in Alzheimer's disease and a C-terminal fragment, recently shown to enter the nucleus and alter gene transcription (60). To assess the role of -APP in the binding of MINTs to Golgi membranes, we used siRNA techniques to deplete cells of -APP or MINT3 and monitored the impact on the immunocytochemical distribution of the proteins.
HeLa cells express only a single MINT, MINT3, whose distribution (Fig. 6D) overlaps extensively with that of -APP (Fig. 6A), markers of the Golgi compartment (e.g. giantin, see Fig. 6G), and Arfs (data not shown). Transient transfection of cells with plasmids directing expression of a small, snap-back RNA resulted in knockdown in the levels of expression of -APP (Fig. 6, B, E, and H) or MINT3 (Fig. 6, C, F, and I), as visualized by cell staining and confirmed by immunoblotting (Fig. 6J). -APP in HeLa cells is seen in puncta throughout the cytosol and at high density in the perinuclear region. The most evident consequence of siRNA of -APP was the loss of the concentrated, perinuclear -APP staining (Fig. 6B) seen in control cells (Fig. 6A). Accompanying the loss of -APP at the Golgi, staining of MINT3 was altered to a less compacted but still perinuclear location in transfected cells (Fig. 6E). In contrast, giantin staining was not effected like MINT3 by the loss of -APP, indicating preservation of the Golgi and a specific effect on MINT3 staining. As expected, knockdowns in other, unrelated proteins (including Arl2 or Arf6) had no effect on the staining patterns of either MINT3 or -APP (data not shown). Because the sequences used to knockdown expression of MINT3 and -APP were unique to those messages and completely unrelated to one another, we conclude that the mutually interdependent changes observed between the two proteins result from specific changes in the ability of MINT3 or -APP to properly localize in the absence of the other. Note that in the presence of lowered concentrations of -APP the MINT3 still binds to Golgi membranes, consistent with the earlier conclusion that binding to Arf may be sufficient for localization to Golgi membranes. The failure of MINTs to concentrate normally in -APP-depleted cells could reflect either an arrest at an Arf-bound state or a re-localization as a result of binding to different transmembrane cargo receptors. These data are consistent with the hypothesis that MINTs get recruited to membranes as a result of their binding to Arfs and then get further concentrated at exit sites as a consequence of binding to -APP and other cargo.
|
We hypothesize that MINTs act in concert with Arfs to recruit -APP into budding vesicles at the Golgi. A prediction from this model is that, in the absence of MINT, the -APP should be less able to concentrate at the Golgi. To test it, we knocked down expression of MINT3 and looked for effects on -APP staining. When MINT3 expression was decreased by siRNA (Fig. 6, F and J), -APP distribution was more diffuse and appeared less able to concentrate at the Golgi (Fig. 6C). Again, these mutual effects of MINT and -APP are specific, because depletion of either protein by siRNA does not alter the Golgi compartment as a whole (Fig. 6, G–I). Effects of knockdowns of -APP or MINT3 on staining patterns were quantified, as shown in Fig. 7, and are consistent with the fact that we typically achieve 60–80% transfection efficiency under the conditions used in these experiments. In addition, we consistently found that the cellular levels of -APP were reduced by knockdown in MINT3 levels (Fig. 6J). This is consistent with and extends previous data demonstrating that increased expression of MINTs resulted in an increase in the cellular levels of -APP (see Fig. 3) (31, 33, 56, 57).
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
-APP. We then extended the functional significance of the -APP-MINT interaction to live cells by demonstrating that reducing the expression of either protein resulted in alteration in the ability of the other to properly localize at Golgi membranes. Given the facts that: 1) -APP is a transmembrane protein with a sorting motif in its cytoplasmic tail that binds directly to MINTs and traffics the endomembrane system, 2) Arfs are soluble GTPases whose principal function is the recruitment of coat proteins to (predominantly Golgi) membranes, and 3) MINTs are recruited to membranes by Arfs where they bind -APP, can co-immunoprecipitate with clathrin, and are important to the Golgi localization of -APP, we propose that MINTs function in cells as a new family of Arf-dependent coat proteins whose cargo includes -APP. The number of coat proteins or complexes that confer specificity to vesicles involved in membrane traffic has more than doubled in recent years with the identification of AP-3 (10, 61), AP-4 (3, 9, 62), and the three GGA proteins (4–6). With a larger number of distinct coated vesicles emanating from or targeted to a common donor compartment, there is increasing likelihood of differences in destinations and rates of flux through the donor compartment. The Golgi/TGN is central to vesicle traffic and the most likely compartment to require multiple different types of vesicles to enter and leave. We now can identify nine different types of coated vesicles at the Golgi/TGN (AP-1, COPI, AP-4, GGA1–3, and MINT1–3), all of which are Arf-dependent.
The demonstration that MINTs interact with Arfs in cells offers insights into the actions of both proteins. The PTB and PDZ2 domains are both required for the proper localization of MINT3 at Golgi membranes (58). This observation was interpreted as evidence for two different receptors on Golgi membranes that coordinate the recruitment of MINT3. Our observations that the same two domains are required for binding solely to Arfs and that activated Arfs are required for binding to Golgi membranes in cells suggests that Arf binding may be sufficient to localize MINT3 on Golgi membranes. However, the facts that MINTs also bind directly to -APP and that the staining profile of MINT3 in cells is altered by loss of -APP expression are also consistent with its playing a role in MINT binding to Golgi. Thus, we conclude that Arf and -APP each play roles in the binding of MINTs to Golgi membranes. Similarly, the reciprocal loss of Golgi staining with knockdowns of MINT3 and -APP is evidence of their interdependence, consistent with their proposed functions in membrane traffic.
Neither PTB nor PDZ domains have previously been shown to bind Arf. Only the C-terminal portion of the PTB domain is needed to facilitate Arf binding (see Table I). This is an atypical interaction between a protein and a PTB domain. In contrast, the crystal structure of the PTB domain of MINT1 bound to a -APP-derived peptide reveals details of how most proteins bind PTB domains (59). Interestingly, the portion of the PTB domain implicated in binding Arf is almost all in one long -helix and makes few contacts with the -APP peptide. This may allow Arf and -APP to bind simultaneously. Alternatively, it also places the two protein ligands in proximity that may facilitate or require sequential binding of Arf and -APP.
PDZ domains are 100-amino acid protein-interaction domains that are rich in hydrophobic residues and often function in the assembly and localization of protein complexes at membrane surfaces (63–65). PDZ domains bind 4–5 residue motifs that are followed by the C terminus or a stable -turn. The C termini of Arfs are highly charged and not likely candidate PDZ binding sites. Examination of Arf structures revealed one candidate internal PDZ binding domain. The sequence 54ETVTY58 is (i) conserved in all mammalian Arfs, (ii) very close to the consensus class I PDZ binding motif described in Harris et al. (64, 65), (iii) followed by a -turn, and (iv) in the interswitch region (66–69). Binding of GTP to Arfs results in changes in the switch I and switch II regions, homologous to those in Ras proteins and a 2-residue shift in the interswitch region that makes it more solvent-accessible (66, 69). The speculated binding of this nucleotide-sensitive region to PDZ2 would explain the GTP-dependent binding of Arfs to MINTs and would entail a novel type of interaction between a regulatory GTPase and an effector.
Previous evidence of direct binding of MINTs to other proteins (22, 23, 27, 28) suggests that -APP is unlikely to be the only cargo involved in MINT-dependent traffic and that perhaps they serve a more general role in traffic to the plasma membrane. This is consistent with genetic and cell biological evidence from C. elegans, in which the MINT1 ortholog, LIN-10, is required for proper localization of the EGF receptor to the basolateral surface of vulva precursor cells (34, 35). The observations that LIN-2/Cask, LIN-7/Velis, and LIN-10/Mint1 are all required for LET-23/EGF receptor localization and that the three gene products (LIN-2/LIN-7/LIN-10) are all found in a cytosolic, trimeric complex indicates that the three function together to regulate the traffic of LET-23. It is possible that the trimeric complex is also recruited to membranes by Arfs, because the Cask binding domain is located near the N terminus of MINT1, thus leaving the Arf binding domain free. However, Cask and Velis are clearly not required for the actions of all MINTs in membrane traffic, because MINT2 and MINT3, foci of our study, lack the Cask binding domain and should not be able to assemble into such a trimeric complex. LIN-10 is also required for the basolateral localization of the GLR-1 glutamate receptors in C. elegans, but in this case LIN-2 and LIN-7 are not involved (70). Whether MINT1 is active in the transport of -APP in nerve cells, either alone of when bound to Munc18 or Cask/Velis, remains to be shown. MINT1 and MINT2 can also form a stable heterodimer with Munc18, are proposed to have additional role(s) in neurons at synapses, and are likely to be distinct from the proposed role in -APP traffic (21, 51).
The binding of both -APP and Arfs to the PTB and PDZ domains found in all MINTs suggests that all three MINTs are capable of regulating vesicle coating at the Golgi. MINTs and -APP have previously been shown to co-localize at Golgi membranes and vesicles (17, 57, 58). In addition, MINT1 and a kinesin (Kif17) were found on vesicles carrying the N-methyl-D-aspartate receptor along neuronal dendrites (25). -APP was recently shown to be subject to proteolytic processing and present, along with kinesin and kinesin-linked cargos, on vesicles moving down axons in the sciatic nerve (71). It will be important to learn if one or more MINTs are found on those same vesicles. The co-localization of MINT3 and -APP in HeLa (Fig. 6) cells and the reciprocity in their altering localization with each other when expression was compromised by siRNA are evidence of a close functional and physical relationship between a MINT and -APP that, we speculate, exists in all cells.
The ubiquity of Arfs, -APP, and MINTs in mammalian cells is consistent with highly conserved roles in membrane traffic. The presence of two different sorting motifs in the C-terminal, cytoplasmic tail of -APP may suggest the need to interact with at least two different coated vesicles during the lifetime of the protein (52, 72); e.g. one may be required to direct traffic of -APP to the cell surface and the second to an endocytic compartment. The observations, that either increases (31, 51, 56) or decreases (this report) in the levels of MINTs in cells result in parallel changes in -APP levels, implicate MINT-dependent traffic as a potential target for novel drugs aimed at altering the progression of Alzheimer's disease. This highlights another important consequence of the findings that cells possess such diversity in Arf-dependent coats, the possibility of designing inhibitors of membrane traffic with greater specificity than are currently available.
|
FOOTNOTES |
|---|
** To whom correspondence should be addressed: Center for Neurodegenerative Diseases, Emory University School of Medicine, 1510 Clifton Rd., Atlanta, GA 30322-3050. Tel.: 404-727-3561; Fax: 404-727-3746; E-mail: rkahn{at}emory.edu.
1 The abbreviations used are: Arf, ADP-ribosylation factor; -APP, -amyloid precursor protein; PBS, phosphate-buffered saline; siRNA, short-interfering RNA; TGN, trans-Golgi network; EGF, epidermal growth factor; HA, hemagglutinin; GTPS, guanosine 5'-3-O-(thio)triphosphate; CCV, clathrin-coated vesicles; PNS, post-nuclear supernatant; GST, glutathione S-transferase; X-gal, 5-bromo-4-chloro-3-indolyl--D-galactopyranoside.
|
ACKNOWLEDGMENTS |
|---|
-APP (26D6)) and Kelley Moremen and Marilyn Farquhar (mannosidase II) for the generous gifts of antibodies. We also thank Reuven Agami (Netherlands Cancer Institute) for the gift of the pSUPER vector. Members of the Kahn laboratory, Howard Rees, and the Center for Neurodegenerative Diseases provided invaluable help and discussions that contributed to this work. |
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
