Membrane-anchored Neuregulin β1 sheds its ectodomain as soluble factors. Two proteases that belong to a disintegrin and metalloprotease (ADAM) family are known to cleave Neuregulin β1. One is tumor necrosis factor-α converting enzyme (TACE/ADAM17). The other is Meltrin β (ADAM19). Against our expectation that shedding by ADAM proteases occurs at the cell surface, here we found that Meltrin β mediates the ectodomain shedding of Neuregulin β1 in the Golgi apparatus. Meltrin β was localized in and around the Golgi apparatus in developing sensory neurons. Subcellular fractionation revealed that Meltrin β generated soluble Neuregulin β1 in Golgi-enriched fractions while TACE-cleaved Neuregulin β1 was recovered in lighter fractions. To examine whether Meltrin β-mediated ectodomain shedding occurs in the Golgi apparatus in living cells, we took advantage of different diffusion properties of cleavage products from those of membrane-anchored precursor proteins. Fluorescence correlation spectroscopy (FCS) is the most sensitive method to determine milli∼submillisecond diffusion in vivo. Protease-active Meltrin β caused a shift in autocorrelation function in FCS of green fluorescent protein (GFP)-tagged Neuregulin β1 in the Golgi apparatus, suggesting a conversion of Neuregulin β1 molecules from membrane-anchored to soluble forms in that organelle. The Golgi apparatus is a site of processing Neuregulin β1 by Meltrin β.
Various intercellular signaling and adhesion molecules govern cell–cell interactions during the development of multicellular organisms. The actions of these molecules are regulated not only by transcriptional and translational controls but also by post-translational modifications such as phosphorylation and proteolytic processing. The latter process includes ectodomain shedding, which may activate or inactivate the functions of a membrane protein or substantially change its functional properties (Wong et al. 1989; Hooper et al. 1997; Tanaka et al. 1998; Yamazaki et al. 2003). Although ectodomain shedding of numerous membrane proteins and proteases that act as “sheddases” has been reported, little is known about the exact subcellular sites of ectodomain shedding and, to our knowledge, the dynamic process of ectodomain shedding in living cells has never been observed.
Membrane anchored Neuregulin is one of membrane proteins that are subject to the ectodomain shedding. Neuregulin is a multifunctional growth factor that plays essential roles in the development of the heart and nervous system (Meyer & Birchmeier 1995; Falls 2003). The core domain of the protein has an epidermal growth factor (EGF) motif and functions as a ligand of heterodimeric ErbB receptors. Although some of Neuregulin isoforms are soluble proteins, the majority are membrane anchored, including beta-exon-containing type-I Neuregulin-1a (NRG β1). As was reported with other membrane proteins, phorbol 12-myristate 13-acetate (PMA) treatment of cells results in the immediate disappearance of the ectodomain of Neuregulin 1 from the cell surface. The ectodomain shedding of Neuregulin 1 also occurs in intracellular compartments. However, the exact subcellular sites of these proteolytic events have not been determined (Burgess et al. 1995).
Two proteases that belong to a disintegrin and metalloproteases (ADAM) family are known to participate in the ectodomain shedding of NRG β1. One is the tumor necrosis factor-α converting enzyme (TACE/ADAM17), which participates in cleavage of NRG β1 upon PMA treatment of cells. The other is Meltrin β (ADAM19), which exerts the ectodomain shedding of NRG β1 in a manner dependent on its protease activity (Shirakabe et al. 2001; Wakatsuki et al. 2004).
ADAM family proteases are involved in the ectodomain shedding of many membrane and extracellular matrix proteins. However, whether individual ADAM proteases have strict substrate specificities or whether some of them share substrates is not clarified yet. Instead, recent studies showed that these proteases are apparently distinct from each other in terms of their mechanisms of regulation of ectodomain shedding. For example, TACE cleaves the ectodomains of various membrane proteins upon PMA treatment of cells (Peschon et al. 1998; Montero et al. 2000; Hinkle et al. 2004; Horiuchi et al. 2005). Kuzbanian/ADAM10 cleaves the ectodomains of various membrane proteins in response to calcium influx (Nagano et al. 2004; Reiss et al. 2005). Significant differences among ADAM proteases have been also reported in terms of the subcellular compartments in which they cleave the ectodomain. The activities of TACE and Kuzbanian are enhanced when the cholesterol-rich microdomain called the membrane raft is disrupted (Kojro et al. 2001; Matthews et al. 2003; Zimina et al. 2005), while Meltrin β is located in the raft fraction together with NRG β1, which is required for ectodomain shedding of the latter (Wakatsuki et al. 2004). Moreover, TACE and Kuzbanian cause disappearance of the ectodomain of CD44 at focal contacts and at the rear ends of some types of moving cells, respectively, suggesting different subcellular sites of ectodomain shedding between these two (Nagano et al. 2004). These results raise a possibility that each ADAM protease meets and cleaves its substrate at distinct subcellular regions among each other. Determination of exact sites of ectodomain shedding by individual proteases is meaningful, assuming that subcellular sites of growth factor generation could regulate intercellular signaling events temporally and spatially.
On the basis of these previous studies, we aimed to identify subcellular sites at which Meltrin β mediates the cleavage of NRG β1. However, determination of sites of ectodomain shedding may not be easy. The major subcellular locations of each ADAM protease are not always in accord with the sites of ectodomain shedding; that is, the disappearance of ectodomain at focal contacts does not necessarily mean that proteolysis occurs at the cell surface at focal contacts. First of all, anterograde and retrograde intracellular traffics of the protease, its substrate precursors and cleavage products should be taken into consideration. Some ADAM proteins, such as TACE and Meltrin α/ADAM12, change their localizations in response to certain stimuli, making the sites of ectodomain shedding difficult to determine (Itai et al. 2001; Matthews et al. 2003). Whether the protease is expressed as an active form at particular intracellular sites should be also examined.
To overcome such difficulties and to determine the exact sites of cleavage mediated by Meltrin β, we set out to determine them in living cells. For this purpose, we applied fluorescence correlation spectroscopy (FCS) to living cells to monitor the ectodomain shedding of green fluorescent protein (GFP)-tagged NRG β1 molecules. FCS is the most sensitive method to determine milli∼submillisecond diffusion in an area as small as subfemtoliter confocal volume in vivo. As a result, against our expectation that shedding by ADAM proteases occurs at the cell surface in general, we could identify the Golgi apparatus as a site where Meltrin β mediates proteolytic processing of Neuregulin. Meltrin β was localized in and around the Golgi apparatus in developing sensory neurons. Subcellular fractionation substantiated the result of FCS. We discuss differential regulation of growth factors by ectodomain shedding both in the Golgi apparatus and in other subcellular regions (such as at the cell surface). Effective application of FCS in the study of ectodomain shedding is also discussed.
Meltrin β is localized in the Golgi apparatus and its immediate vicinity
Meltrin β is expressed abundantly in the peripheral nervous system and in the heart during its development (Kurisaki et al. 1998; Kurohara et al. 2004). To determine the subcellular localization of Meltrin β, neurons prepared from developing dorsal root ganglia of mouse embryos at embryonic day 16.5 were immunostained with antibodies specific for Meltrin β, Golgi 58K or β COP. Both Golgi 58K (forminotransferase cyclodeaminase) and β COP (the beta subunit of a coatomer complex) are Golgi markers. Meltrin β was found in intracellular vesicles (Fig. 1). Some of these vesicles were Golgi 58K- or β COP-positive, whereas others were not but were in the vicinity of Golgi 58K-positive or β COP-positive vesicles. Meltrin β was undetectable at the cell surface and on Transferrin receptor-positive endosomes (data not shown). These results indicate that Meltrin β was localized mainly in the Golgi apparatus.
Previously, we showed that co-expression of Meltrin β with NRG β1 markedly enhanced the ectodomain shedding of NRG β1 (Shirakabe et al. 2001). In addition, the Neuregulin1 gene and the Meltrin β gene are actively transcribed and translated in developing neurons at similar stages (Meyer et al. 1997; Shirakabe et al. 2001). We thus examined whether the Golgi apparatus is the site of ectodomain shedding of NRG β1 by Meltrin β. If Meltrin β cleaves NRG β1 in the Golgi apparatus after translation in the endoplasmic reticulum (ER), released NRG β1 should be found in the former organelle. On the basis of this idea, NRG β1 with a hemagglutinin peptide-tag at its N-terminus (HA-NRG β1) was produced in the cells and the intracellular distribution of membrane-bound and soluble HA-NRG β1 was determined.
Post-nuclear supernatants of N1E115 neuroblasts and COS7 cells, which expressed HA-NRG β1 together with wild-type Meltrin β (WT), empty vector or Meltrin β (EQ), a mutant of Meltrin β that has a single amino acid substitution in the active site of the protease domain, were separated by Nycodenz (Histodenz) gradient centrifugation. Each fraction was subjected to immunoblotting using antibodies against HA, ERp72 (as an ER marker) or Golgi 58K. ERp72 and Golgi 58K were mainly contained in fractions 6–10 and in fractions 1–5, respectively (Figs. 2 and 3). Golgi 58K(+)ERp72(–) and Golgi 58K(–)ERp72(+) fractions were assessed as Golgi- and ER-enriched fractions, respectively. Fraction 6, which often showed Golgi 58K(+)ERp72(+), was considered as a transient or intermediate fraction between the ER and the Golgi apparatus. The bottom Golgi 58K(+)ERp72(+) fraction (fraction 10) was assumed to contain cell debris. The top Golgi 58K(+)ERp72(+) fractions (fractions 1 and 2) might include lighter membrane fractions such as endosomes, secretory vesicles and particulated plasma membrane, as well as cytosolic components. In NIE115 neuroblasts, high levels of endogenous Meltrin β were produced. Although endogenous or exogenously introduced Meltrin β was found not only in the Golgi fractions but also in ER fractions, prodomain-excised, active forms of Meltrin β were found in the Golgi 58K(+) (fractions 1–5) and intermediate (fraction 6) fractions. The majority of NRG β1 molecules in these fractions were soluble forms not only in the exogenous Meltrin β-treated cells but also in empty vector-treated cells (Fig. 2A,C). In contrast, expression of Meltrin β (EQ) resulted in the suppression of ectodomain shedding in Golgi-enriched and intermediate (fractions 3–6) fractions (Fig. 2B), while fractions 1 and 2 contained soluble NRG β1. To exclude the possibility that ectodomain shedding with Meltrin β was mediated by TACE, the post-nuclear supernatant of TACE protease-deficient (TACE−/–) cells transfected with HA-NRG β1 together with Meltrin β (WT) or with Meltrin β (EQ) was analysed similarly. Ectodomain shedding of HA-NRG β1 was enhanced by Meltrin β (WT) in the Golgi apparatus in a TACE-independent manner (Fig. 2D). Similar results were obtained in COS7 cells. Golgi 58K(+) (fractions 1–5) fractions of the cells transfected with Meltrin β (WT) exclusively contained shed, soluble NRG β1, whereas ER-enriched fractions (fractions 7–9) of these cells contained only uncleaved molecules. In contrast, a substantial proportion of NRG β1 remained uncleaved in Golgi 58K(+) fractions of the cells transfected with empty vector or Meltrin β (EQ) vector (Fig. 3). Taken together, these results suggested that Meltrin β generated soluble NRG β1 in the Golgi apparatus, which was delivered to the lighter membrane fractions.
The post-nuclear supernatant of COS7 cells transfected with vectors for TACE and HA-NRG β1 was also separated by Histodenz gradient centrifugation, and each fraction was subjected to immunoblotting using the antibodies mentioned above. Most of the HA-NRG β1 molecules were uncleaved in the Golgi-enriched and intermediate fractions (fractions 3–6), whereas a substantial proportion of truncated forms were found in the top fractions (fractions 1 and 2). Thus, TACE might cleave HA-NRG β1 in subcellular regions other than the ER or the Golgi apparatus (Fig. 3D).
Design of GFP-NRG β1 to detect ectodomain shedding in living cells
Next, we tried monitoring the ectodomain shedding of NRG β1 by Meltrin β in living cells. To study this, we took advantage of the extremely high time resolution of FCS. FCS provides information on how long the fluorescent molecules are retained in the subfemtoliter confocal volume. As shown in Fig. 4A, we constructed NRG β1 fused to GFP at the N-terminus (GFP-NRG β1). If newly synthesized NRG β1 underwent proteolytic processing in the Golgi apparatus after translation, GFP-containing molecules in the Golgi apparatus would be free from the lipid bilayer and smaller, as a whole, than those in the ER, where GFP-NRG β1 would be intact. Because membrane proteins are known to show at least one order slower diffusional motion than soluble proteins, shedding of NRG β1 should result in a clear difference in autocorrelation function G(τ) (Fig. 4B). The shed NRG β1 confined in the ER–Golgi lumen should show a 39%−60% reduction in diffusion from the theoretical diffusion constant estimated from the Einstein–Stokes equation (Olveczky & Verkman 1998). However, the effects of confinement on diffusional motion should be negligible compared with those of restriction by membrane-anchoring.
We first determined whether nascent GFP-NRG β1 could be trapped early in the secretory pathway and reach the cell surface properly in living cells. GFP-NRG β1 was expressed in mouse fibroblast L929 cells for 12 h at 37 °C and was observed by confocal microscopy. GFP-NRG β1 was found to be localized at the cell surface, in the Golgi apparatus and in the ER. When the incubation temperature was subsequently shifted to 20 °C for 1 h, GFP-NRG β1 accumulated mainly in the Golgi apparatus. When the incubation temperature of the cells was then shifted back to 37 °C for 30 min, GFP-NRG β1 appeared at the cell surface again. Thus, newly synthesized GFP-NRG β1 was transported properly (Fig. 5A–C). The ectodomain shedding of GFP-NRG β1 mediated by Meltrin β generated N- and C-terminal fragments of the expected sizes (data not shown).
FCS measurements reveal that Meltrin β mediates ectodomain shedding of GFP-NRG β1 in the Golgi apparatus in living cells
GFP-NRG β1 vector was introduced together with Meltrin β (WT) vector into COS7 cells using the glass microbead method and incubated for 2.5 h. A significant difference (P < 0.01) was observed between the FCS in the ER and that in the Golgi apparatus in the same cells (Fig. 6, steady state). A higher proportion of rapid diffusional motion was detected in the Golgi apparatus than in the ER, suggesting that membrane-anchored GFP-NRG β1 molecules were cleaved and showed faster diffusional motion in the Golgi apparatus than in the ER (Fig. S1, steady state, WT; a clear difference in autocorrelation function G(τ) between ER and Golgi). In contrast, co-transfection of empty vector or Meltrin β (EQ)vector did not produce such differential composition of diffusion or FCS autocorrelation curves for the two organelles (Fig. S1, steady state, EQ and vector).
To check the reliability of these FCS measurements, we compared the autocorrelation functions and the best-fit two-component simulation profiles (Fig. S2). The good fit between the experimental and theoretical values indicates that the autocorrelation functions of these data were reliable, assuming that there were two main mobile components (rapidly diffusing and slowly diffusing). According to previous observations on the diffusion of soluble and membrane-anchored proteins (Saito et al. 2003), the diffusion constants of membrane proteins are generally one order lower than the soluble forms. In this study, because most of FCS data were best-fitted with the two components model in which the fast component was ∼300 µs and the slow was ∼12 000 µs, we assumed that there are two populations whose diffusional motion is distinct.
The calculated proportions of the rapidly and slowly diffusing components, which would represent soluble (cleaved) forms and membrane-bound (not cleaved) forms of NRG β1, respectively, are shown in Fig. 6 (steady state).
Although the glass microbead method enabled us to monitor GFP-NRG β1 in 2–3 h after transfection, retrograde transport might bring intact or cleaved GFP-NRG β1 from the cell surface or the trans-Golgi network to the Golgi apparatus or from the Golgi apparatus to the ER. To avoid contribution of these retrograde transports, we examined the effect of Bafilomycin A1 (Baf). Baf inactivates endosomes and inhibits retrograde transport from the Golgi apparatus to the ER (Palokangas et al. 1998), enabling us to use FCS to detect newly synthesized NRG β1 and NRG β1 carried from the ER to the Golgi apparatus by anterograde transport. We confirmed the effects of this reagent on the retrograde transport from the Golgi apparatus to the ER by analysing the distribution of the YFP-tagged KDEL receptor introduced together with CFP-Golgi (Fig. 5D). When COS7 cells were treated with Baf (1 µm final concentration) for 1 h, the amount of KDEL receptor molecules was reduced in the ER (Fig. 5D, lower panel), indicating that retrograde transport from the Golgi apparatus to the ER was inhibited.
When GFP-NRG β1 was introduced together with Meltrin β (WT) into COS7 cells and they were treated with Baf for 1 h, results similar to those without Baf treatment were obtained; a significant difference (P < 0.005) was observed between the FCS in the ER and that in the Golgi apparatus in the same cells in Meltrin β (WT)-producing cells but not in empty vector- or Meltrin β (EQ)-introduced cells (Fig. 6, Baf treatment; Fig. S1, Baf treatment). Thus, these measurements further confirmed that cleavage of GFP-NRG β1 dependent on Meltrin β occurred in the Golgi apparatus.
To verify our hypothesis in another way, we used cells treated with BFA for 1 h. In cells treated with BFA for 1 h, anterograde transport from the ER to the Golgi apparatus is inhibited and part of the Golgi cisterna is fused to the ER (Schweizer et al. 1988; Klausner et al. 1992). If GFP-NRG β1 is shed in the Golgi apparatus by Meltrin β, rapid diffusional motion would be generated in the ER by the fusion of the Golgi apparatus with BFA treatment. We confirmed the effects of BFA by analysing the distribution of the YFP-tagged KDEL receptor introduced together with CFP-Golgi. COS7 cells treated with BFA (25 µm final concentration) for 1 h showed fusion of the Golgi apparatus and the ER, confirming the effect of the BFA (Fig. 5E). GFP-NRG β1, together with Meltrin β (WT), empty vector or Meltrin β (EQ), was expressed in the COS7 cells as above and diffusional motion in the ER was observed by FCS. After the initial measurement, cells were treated with BFA for 1 h and diffusional motion in the ER of the same cells was observed again (Fig. 6, BFA treatment; Fig. S1, BFA treatment). Treatment of cells that expressed Meltrin β (WT) with BFA resulted in the appearance of a higher proportion of molecules with fast diffusional motion than in steady-state cells. Because BFA causes fusion of the Golgi apparatus and the ER, we interpreted these results as evidence of enhanced movement of soluble GFP-NRG β1 generated in the Golgi apparatus or of enhanced movement of active Meltrin β into the ER or both, resulting in an increased number of soluble GFP-NRG β1 molecules in the fused ER.
To obtain further evidence of the cleavage of GFP-NRG β1 in the Golgi apparatus, we exploited isotetrandrine (ITD), an inhibitor of BFA-induced tubule formation from the Golgi complex and retrograde trafficking to the ER (Chan et al. 2004). Cells treated with ITD show inhibition of the transport of proteins from the Golgi apparatus to the ER induced by BFA. If the increase in diffusional motion in the ER of cells treated with BFA were caused by the fusion of part of the Golgi cisterna to the ER, ITD would inhibit this increase. To check the effect of ITD, COS7 cells were treated first with ITD (20 µm final concentration) for 1 h and then with ITD and BFA (25 µm final concentration) for an additional 1 h. This treatment inhibited fusion of the Golgi apparatus and the ER (compare Fig. 5E,F). GFP-NRG β1 together with Meltrin β (WT), empty vector or Meltrin β (EQ) was expressed in COS7 cells as above; the cells were cultured for 1 h in medium containing ITD and then diffusional motion in the ER was observed by FCS. After the initial observation, cells were treated with BFA and ITD and incubated for 1 h, and diffusional motion in the ER of the same cells was observed (Fig. S1, ITD + BFA). A much smaller proportion of rapid diffusional motion was observed in the ITD- and BFA-treated cells expressing Meltrin β (WT) than in the cells treated with BFA only (Fig. 6, ITD + BFA treatment). The FCS measurements in the Golgi apparatus of the cells treated with ITD and BFA are shown in Fig. S4. Results essentially opposite to those in the ER were obtained. That is, the fast population increased only in the WT Meltrin β-expressing cells treated with ITD + BFA, suggesting that ITD caused inhibition of constitutive recycling between the ER and Golgi in addition to the blockage of BFA-induced redistribution of Golgi components. The distribution of Meltrin β treated with BFA, ITD or ITD + BFA was monitored by ligating GFP-tag at the C-terminus in Fig. S3. Meltrin β-GFP was mainly localized in the Golgi apparatus and its vicinity. Meltrin β-GFP redistributed in the ER together with Ds-Red-Golgi-marker after the treatment of BFA, which was further blocked by the cotreatment of ITD. These results indicate that rapidly diffusing molecules appeared in the ER when the Golgi components of Meltrin β (WT)-producing cells, including matured Meltrin β itself, were delivered into the ER by treatment with BFA.
We identified the Golgi apparatus as one of the subcellular sites of ADAM protease-mediated ectodomain shedding of membrane-anchored ErbB ligands by monitoring the ligands in living cells for the first time. Although this study does not exclude roles of Meltrin β in other subcellular regions, we consider the Golgi apparatus to be one of the major subcellular sites that this protease acts in judged from its expression patterns in cells.
Although members of the ADAM family proteases are known to be key players in the ectodomain shedding of various membrane-anchored growth factors and adhesion proteins, it has been elusive at which subcellular sites ADAM proteases cleave these membrane proteins. In the case of Neuregulin 1, a number of researchers have roughly identified the subcellular regions where its membrane-anchored forms are proteolytically processed. However, determination of the exact subcellular sites of ectodomain shedding from experiments with cell lysates or immunostaining of fixed cells alone is extremely difficult. Although pulse-chase analysis can indicate whether the cleaved ectodomains accumulate inside the cell or how fast they are released in the culture medium, complicated anterograde and retrograde sorting of intracellular vesicles makes the subcellular regions of cleavage difficult to determine (Burgess et al. 1995; Loeb et al. 1998). In addition, the subcellular sites of protease-mediated ectodomain shedding cannot be specified simply from the distribution of proteases in cells. The finding that Meltrin β was localized mainly in the Golgi apparatus led us to expect that this protease acts on the substrate in the Golgi apparatus. The protease, however, could be in a latent form or it could be physically separated from its substrates or activators there. For example, ADAM proteases have proprotein domains, and proteolytic removal of these domains by furin-like proteases is required for protease activation (Schlondorff et al. 2000). For these reasons, we considered that monitoring the cleavage of substrates in living cells was essential for determining the subcellular sites of ectodomain shedding exactly and conclusively.
We chose NRG β1 as a protease substrate in this study for several reasons. Evidence suggests that the ectodomain shedding of NRG β1 occurs both at the cell surface and in intracellular compartments (Burgess et al. 1995). Moreover, NRG β1 is reported as a substrate of two different ADAM proteases. We found previously that Meltrin β is produced in developing neurons at similar stages as Neuregulin 1 and enhances the ectodomain shedding of NRG β1 in a manner dependent on the protease domain activity of Meltrin β (Shirakabe et al. 2001; Wakatsuki et al. 2004). On the other hand, TACE is involved in the PMA-induced ectodomain shedding of various ErbB ligands, including Neuregulin 1 (Peschon et al. 1998; Montero et al. 2000; Shirakabe et al. 2001; Hinkle et al. 2004; Blobel 2005; Horiuchi et al. 2005). Thus, we initially focused on the question of whether Meltrin β localized in the Golgi apparatus of cells such as the developing primary neurons was involved in intracellular ectodomain shedding of NRG β1. Then, we sought to determine whether Meltrin β and TACE shared subcellular sites of ectodomain shedding. The answer to the latter question is important, because the sites of ADAM protease-mediated ectodomain shedding should be closely associated with the biological roles of these proteases.
Simple confocal microscopy cannot be used to observe the actual moment of the cleavage of N- and C-terminal-tagged proteins (such as YFP-NRG β1-CFP), because the distance between N- and C-termini is apparently below the limit of optical resolution. That is, N-terminal and C-terminal tags merges under the confocal microscope when segregation of these tags are below the resolution of an optical microscope even if they exist in separate polypeptides as a result of the cleavage. However, there are two alternative methods for analyzing proteolytic changes in a single molecule: fluorescence resonance energy transfer (FRET) and FCS. FRET is a powerful method for observing the interaction of two molecules or the interaction of the domains of a protein in living cells. The efficiency of FRET is critical for distinguishing between interacting and non-interacting states in a molecule, and this efficiency depends on the distance between the two chromophores and the orientation of their spins. For our purpose, the chromophores have to be sufficiently distant from the proteolytic site to avoid steric hindrance of the juxtamembrane proteolytic site within NRG β1. In addition, the orientation of the spins of the two chromophores must be arranged to obtain high FRET efficiency, and this requirement seriously restricts the choice of constructs for obtaining useful FRET results.
FCS provides information about the dynamics of molecular processes by means of measurements of spontaneous microscopic fluctuations at dilute molecular concentrations (about 102 molecules in the confocal volume). FCS detects fluctuations of fluorescence intensity in a confocally defined volume with a sharply focused laser and can measure translational and rotational diffusion coefficients of molecules in solution and in living cells (Widengren & Rigler 1998; Foldes-Papp et al. 2001; Hess et al. 2002; Larson et al. 2003; Gosch et al. 2004; Kamada et al. 2004; Bacia et al. 2006). Although FCS is not yet commonly used, several experiments have revealed the great potential of this method in living cells (Bacia et al. 2006; Kogure et al. 2006). In our experiment, the shed NRG β1 confined in the ER–Golgi lumen should show a 39%−60% reduction in diffusion from the theoretical diffusion constant estimated from the Einstein–Stokes equation (Olveczky & Verkman 1998); however, the effects of confinement on diffusional motion should be negligible compared with those of restriction by membrane-anchoring. Furthermore, although the ER and the Golgi apparatus are structurally and dynamically distinct compartments, their structural constraints on the diffusion of luminal proteins should be comparable, since the occlusion of space is comparable in both organelles (Voeltz et al. 2002). Hence, we reasoned that comparison of the time-correlated decay curves obtained by FCS in various compartments along the secretory pathway would enable us to identify the site of shedding. Although further improvements are required for more precise quantitative analysis, this study clearly showed FCS as a valuable tool to assess ectodomain shedding of membrane proteins in living cells.
We fused NRG β1 molecules with GFP at the N-terminus (GFP-NRG β1) to monitor the ectodomain shedding. FCS analysis with this probe was focused on the cleavage within and around the Golgi apparatus, where Meltrin β resides in the primary neurons during development. The results of the FCS analysis are summarized in Fig. 6. The following three sets of results indicated that Meltrin β plays a role in ectodomain shedding in the Golgi apparatus. First, a higher proportion of rapidly diffusing forms of GFP-NRG β1 was found in the Golgi apparatus than in the ER of Meltrin β-expressing cells. No such increase was observed by the introduction of protease-deficient Meltrin β (EQ) or the empty vector. Similar results were obtained with cells treated with Baf, which inactivates endosomes and inhibits retrograde transport from the Golgi apparatus to the ER (Palokangas et al. 1998). Baf is a specific inhibitor of vacuolar-type H + ATPase. According to a previous report, Golgi pH increases approximately 0.6 after the treatment of Baf (Llopis et al. 1998). Because the metalloproteases act at a relatively wide range of pH in general, a slight increase in pH may not affect protease activity of Meltrin β significantly. We therefore interpreted the data that the proportion of rapidly diffusing molecules slightly increased after Baf treatment in the Golgi apparatus of Meltrin β (WT)-expressing cells, probably because of the inhibition of the retrograde transport of shed molecules from the Golgi apparatus to the ER. The results also excluded the involvement of endosomal pathway in this process. Second, BFA treatment caused an increase in the proportion of rapidly diffusing molecules in the ER of Meltrin β-producing cells. Third, in contrast, treatment of the cells with ITD in addition to BFA resulted in a decrease in the proportion of rapidly diffusing molecules in the ER of Meltrin β-producing cells compared with treatment of these cells with BFA alone. BFA inhibits anterograde transport from the ER to the Golgi apparatus but stimulates fusion of the Golgi apparatus to the ER (Schweizer et al. 1988; Klausner et al. 1992). ITD inhibits BFA-induced tubule formation from the Golgi complex and retrograde trafficking to the ER and also inhibits BFA-stimulated tubule formation from the Golgi apparatus to the ER (Chan et al. 2004). We therefore interpreted the results of the latter two experiments as indicating that retrograde traffic of Meltrin β from the Golgi apparatus to the ER as a result of treatment with BFA stimulated ectodomain shedding in the ER. BFA treatment might have also caused retrograde transport of the rapidly diffusing molecules generated in the Golgi apparatus to the ER. Indeed, inhibition of BFA-induced effects was completely reversed by addition of ITD. Although ITD seems to disturb the constitutive recycling, at least, we showed that accumulation of the fast component by over-expression of WT Meltrin β. Taken together, these results indicate that Meltrin β is involved in the ectodomain shedding in the Golgi apparatus.
We obtained supporting results from Western blotting of N1E115 and COS7 cell lysates after subcellular fractionation by Histodenz gradient centrifugation (Figs. 2 and 3). Meltrin β (WT), but not empty vector or Meltrin β (EQ), markedly enhanced generation of the shed ectodomain of NRG β1 in Golgi-enriched fractions. These biochemical data not only support the results of FCS but also indicate that the slowly and rapidly diffusing components of GFP-NRG β1 observed in the FCS measurements did in fact correspond to GFP-NRG β1 molecules before and after ectodomain shedding. Because all the experimental values of the autocorrelation function fitted well to the theoretical best-fit two-component simulation profile, the results indicate the reliability of the measurements in this study (Fig. S2).
In addition to determination of a subcellular site of ectodomain shedding that one of ADAM proteases mediates, our results provided evidence that different ADAM proteases act on the same substrate at different subcellular regions. First, subcellular fractionation suggested that Meltrin β and TACE cleaved the same growth factor at different subcellular fractions. Moreover, the increase in shed molecules of NRG β1 in the presence of Meltrin β in the Golgi fraction of TACE−/– MEFs showed clearly that ectodomain shedding in the Golgi apparatus in Meltrin β-treated cells required the protease activity of not TACE, but Meltrin β itself. Taken together, our results suggest that Meltrin β is involved in intracellular ectodomain shedding in and around the Golgi apparatus, and that this corresponds to the constitutive ectodomain shedding reported previously (Burgess et al. 1995). The ectodomain shedding of Neuregulin 1 is stimulated by PMA or by fetal bovine serum (Loeb et al. 1998). In our experiments, cells were cultured in 1% serum. Further investigation of whether ‘constitutive’ shedding in the presence of Meltrin β requires certain factors in serum is necessary.
TACE and Meltrin β play roles in valve formation during heart development (Jackson et al. 2003; Kurohara et al. 2004; Zhou et al. 2004). Double-knockout mice lacking the genes encoding these two proteins showed dilation of the cardiac muscle, which was not found in the respective single-knockout mice (Horiuchi et al. 2005). Thus, these two ADAM proteases have related functions in heart development. TACE is produced mainly in the endocardial epithelia and is considered to be involved in the ectodomain shedding of HB-EGF, which is also produced in these cells (Montero et al. 2000; Hinkle et al. 2004). Meltrin β is highly expressed not only in endocardial cell lineage but also in the cardiac neural crests adjacent to these epithelia. Although Neuregulin 1 is one of the major ErbB ligands produced in these cell lineages, further investigation is needed to determine whether this growth factor is the physiological substrate of Meltrin β in them. Thus, TACE and Meltrin β may shed different, but related, growth factors, both of which are required for valve formation and cardiac myogenesis. However, alternative mechanisms are also conceivable. We favor an idea that TACE and Meltrin β, and probably other ADAMs, play additive or synergistic roles through proteolytic processing of the same growth factors, including Neuregulin 1, but act in distinct manners. The idea comes from the distribution of the soluble NRGβ1 in the subcellular fractionation experiment. In the presence of Meltrin β, soluble NRG β1 was generated in the Golgi-enriched and ER–Golgi intermediate fractions (fractions 3–6). In contrast, when TACE was co-introduced, soluble NRGβ1 was found only in the light membrane fractions (fractions 1 and 2). A similar cleavage pattern of NRG β1 was found in N1E115 transfected with the EQ mutant (compare Figs. 2B and 3D). Soluble NRG β1 in the light membrane fractions could be generated by endogenous TACE in N1E115 cells transfected with the EQ mutant.
Based on these findings, we propose that Meltrin β is one determinant of the fate of NRG β1 in cells (Fig. 7). Meltrin β might regulate generation of soluble growth factors temporally or spatially through the ectodomain shedding in the Golgi apparatus. Compared to ectodomain shedding in response to some stimuli at the cell surface or in the exosomes, continuous generation of soluble ligands in the Golgi apparatus should be important for the maintenance of growth factor signaling. Alternatively, because the Golgi apparatus plays a central role in sorting various membrane and soluble proteins, the soluble and the membrane-anchored forms of the growth factor in the Golgi apparatus might be sorted through different pathways to different directions from each other. Thus, production of soluble growth factors at different subcellular sites would modulate delivery of growth factors spatially. Further studies will clarify the biological significance of the ectodomain shedding in the Golgi apparatus mediated by Meltrin β.
Expression vectors, materials, cell lines, and transfection
Bafilomycin A1 and brefeldin A were purchased from Sigma-Aldrich (Baf: B1793, BFA: B7651, PMA: 79346-1MG, St. Louis, MO, USA). Isotetrandrine was purchased from BIOMOL International (G-520, Plymouth Meeting, PA, USA). The construction methods for wild-type Meltrin β (Meltrin β (WT)), E347Q protease-deficient Meltrin β (Meltrin β (EQ)) and hemagglutinin (HA)-Neuregulin β1 (HA-NRG β1) were previously described (Shirakabe et al. 2001). The construction of ERD-like protein (ELP-1)-YFP (KDEL receptor-YFP) was previously described (Nagaya et al. 2002). DSRed-Monomer Golgi and pCFP-Golgi were purchased from BD Biosciences (Clontech, Mountain View, CA, USA).
GFP-Neuregulin β1 (GFP-NRG β1) was constructed by ligating Acrosin-GFP (kindly provided by G. Kondo (Kondoh et al. 1999)) to the N-terminus of NRG β1. The TACE vector was kindly provided by S. Nagata (the TACE vector we used is described as mTACE in Itai et al. (2001)).
The establishment and maintenance of TACE−/– MEFs have been previously described (Buxbaum et al. 1998). COS7 and N1E115 neuroblaststoma cells were propagated every 1 or 2 days by culturing in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal bovine serum. Except for FCS, cells were plated at a confluence of 70%−80% for 12 h before transfection. For FCS, vectors were introduced directly into the cells using siliconized glass microbeads (Nagaya et al. 2002), and the cells were then cultured in DMEM containing 1% fetal bovine serum. The total amount of introduced DNA was 5 µg, and the DNA concentration was 1 µg/µL. For subcellular fractionation, the cells were transfected by the lipofection method using FuGENE6 (Roche Applied Science, Mannheim, Germany).
The antibodies used and their sources were as follows: rabbit polyclonal anti-C-terminus of Meltrin β antibody (Shirakabe et al. 2001), mouse anti-HA epitope antibody (16B12, Babco, Richmond, CA, USA), anti-ERp72 (Stressgen, Victoria, BC, Canada), anti-Golgi 58K (Sigma), anti-β COP antibody (G6160, Sigma), HRP-conjugated secondary antibody (PI-1000, Vector Laboratories, Burlingame, CA, USA), Alexa488-labeled secondary antibody (A-11029, Invitrogen) and Cy3-labeled secondary antibody (A-11029, Invitrogen).
Dorsal root ganglia preparation
Dorsal root ganglia were dissected from mice at E16.5 and dissociated by treatment with 0.125% collagenase (Worthington Biomedical, Freehold, NJ, USA; CLS-1) and 40 U/mL Dispase (Godo Shusei Corporation, Tokyo, Japan) for 30 min at 37 °C. Cells were plated on glass-bottomed dishes coated with poly d-lysine and laminin and were cultured in DMEM supplemented with 2% B27 (Invitrogen) and 50 ng/mL 2.5S nerve growth factor (NGF) (Invitrogen) for 24–48 h.
Confocal microscopy and FCS analysis
FCS analysis with living cells using confocal microscopy has been previously described (Kamada et al. 2004). FCS measurements were performed 2.5 h after DNA transfection into COS7 cells. The culture temperature of cells on the microscope stage was controlled using an objective heater for a planapochromat lens × 63 (Bioptechs, Butler, PA, USA) or a silicon heater (Cell MicroControls, Norfolk, VA, USA) for a C-Apochromat × 40 lens in combination with a stage heater (Kitazato Supply, Fujinomiya, Japan). To monitor the temperature of the cells under observation, a CT820 infrared thermometer (Citizen Co., Tokyo, Japan) was used. For confocal microscopy and FCS analysis, a ConfoCor2 instrument (Carl Zeiss Microscopy, Jena, Germany) was used. Confocal images were taken with the laser scanning microscopy module. The excitation light of an argon ion laser at a wavelength of 514 nm was reflected by a dichroic mirror (HFT 514) and focused through a C-Apochromat × 40, NA 1.2 water-immersion objective (pinhole width, 70 µm). A 530- to 560-nm band-pass filter was used to filter out the remaining scattered laser light. In all the measurements, the minimum laser power available was used. The fluorescence signal was recorded for three consecutive periods of 15 s (time resolution, 200 ns). The autocorrelation function and data fitting were performed with the software provided with the ConfoCor2. Indirect immunofluorescence of fixed cells and time-lapse analysis of fluorescent molecules in living cells were carried out and results were processed as previously described by Nagaya et al. (2002). The confocal images in Fig. 5 were taken with an LSM 510 META laser scanning microscopy module (Carl Zeiss Microscopy, Jena, Germany).
Subcellular fractionation of COS7 cells and N1E115 cells
After 24 h of transfection, cells cultured on 100-mm plates (one plate per sample) were washed twice with ice-cold phosphate-buffered saline, harvested in homogenization buffer (10 mm Tris–HCl, pH 7.5; 250 mm sucrose) at 0.6 mL per plate and homogenized by passage 10 times through a 27-gauge needle on a 1-mL syringe. All subsequent steps were performed at 4 °C. Unbroken cells and nuclei were removed by centrifugation at 1200 g for 5 min. The post-nuclear supernatant was loaded on preformed Nycodenz (Histodenz, Sigma) gradients (Okumura et al. 2006). These preformed gradients were prepared for the Beckman MLS-50 rotor from initial discontinuous gradients (24%, 19%, 15% and 10% Nycodenz in 10 mm Tris–HCl, pH 7.5; 3 mm KCl; and 1 mm EDTA) that were allowed to diffuse in a horizontal position for 45 min at room temperature and then centrifuged for 4 h at 100 000 g in an Optima Max ultracentrifuge (Beckman Coulter, Fullerton, CA, USA) to generate a non-linear density gradient profile. The post-nuclear supernatant was loaded on top of the gradient and centrifuged for 3 h at 100 000 g. Ten fractions were collected from the top, and the proteins in aliquots of the fractions were resolved by SDS-PAGE. The distributions of HA-NRG β1, Meltrin β, ERp72 and Golgi 58K were determined by immunoblotting using ECL Plus (Amersham Biosciences, GE Healthcare, Piscataway, NJ, USA).
This work was supported in part by Grants-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology of Japan, by a research grant from CREST (Core Research for Evolutional Science and Technology) within the Japan Science and Technology Agency and by a grant from Japan Space Forum, JAXA (Japan Aerospace Exploration Agency). T. Yokozeki was supported by a research fellowship from the Japan Society for the Promotion of Science (JSPS). We thank Dr G. Kondo for his kind provision of Acrosin-GFP vector, and Dr S. Nagata for his kind provision of TACE vector. We also thank Dr E. Nishi for technical advice for cell culture, and Mr A. Yaguchi of Carl Zeiss Co., Osaka, Japan, for his help with analysis of the FCS data.