Arrestins are key adaptor proteins that control the fate of cell-surface membrane proteins and modulate downstream signaling cascades. The Dictyostelium discoideum genome encodes six arrestin-related proteins, harboring additional modules besides the arrestin domain. Here, we studied AdcB and AdcC, two homologs that contain C2 and SAM domains. We showed that AdcC – in contrast to AdcB – responds to various stimuli (such as the chemoattractants cAMP and folate) known to induce an increase in cytosolic calcium by transiently translocating to the plasma membrane, and that calcium is a direct regulator of AdcC localization. This response requires the calcium-dependent membrane-targeting C2 domain and the double SAM domain involved in AdcC oligomerization, revealing a mode of membrane targeting and regulation unique among members of the arrestin clan. AdcB shares several biochemical properties with AdcC, including in vitro binding to anionic lipids in a calcium-dependent manner and auto-assembly as large homo-oligomers. AdcB can interact with AdcC; however, its intracellular localization is insensitive to calcium. Therefore, despite their high degree of homology and common characteristics, AdcB and AdcC are likely to fulfill distinct functions in amoebae.
In their environment, cells are constantly subjected to a variety of stimuli that orient their behavior and/or fate in terms of motility, growth and differentiation. Sensing of these environmental cues involves integral membrane proteins at the cell surface, among which G-protein-coupled receptors (GPCRs) are clearly the most studied example (Fredriksson et al., 2003; Lagerström and Schiöth, 2008). In mammals, β-arrestins play a central role in the regulation of GPCRs and associated signaling by impeding their coupling to heterotrimeric G proteins, modulating the activation of downstream effectors and controlling GPCR presence at the cell surface (Kendall and Luttrell, 2009; Gurevich and Gurevich, 2015; Tian et al., 2014; DeWire et al., 2007). These scaffolding proteins consist in a double crescent-shaped β-sandwich (arrestin N and C domains) that recognizes the receptor, and a short C-terminal tail (Han et al., 2001; Milano et al., 2002; Zhan et al., 2011; Vishnivetskiy et al., 2004; Gurevich and Gurevich, 2004). This tail contains recruitment sites for the endocytic machinery and becomes accessible upon β-arrestin binding to the activated – and usually phosphorylated – receptor, and subsequent destabilization of a central polar core, allowing the internalization of the receptor (Kim and Benovic, 2002; Laporte et al., 2002; Schmid et al., 2006). Besides the ubiquitous β-arrestins and the retina-specific visual arrestins, all seemingly restricted to higher eukaryotes of the animal kingdom, the arrestin family also comprises structurally related proteins called α-arrestins or arrestin-domain-containing proteins, discovered more recently and which are present from protists to human (Alvarez, 2008; Aubry et al., 2009). These novel members are all predicted to share the arrestin fold. However, the presence of a polar core is questioned, the C-terminal tail as found in β-arrestins is absent and novel extensions are present on either side of the arrestin domain, providing – or being likely to provide – different properties (Alvarez, 2008; Aubry and Klein, 2013; Becuwe et al., 2012a). Despite this structural diversity, the regulation of membrane cargo trafficking appears to be an evolutionarily conserved function of the arrestin clan, and the repertoire of known arrestin membrane targets now includes GPCRs, single-membrane span receptors, integrins, channels and transporters, as illustrated in mammals and yeast (Becuwe et al., 2012a; Kang et al., 2014; Kovacs et al., 2009; Lefkowitz et al., 2006; Lin et al., 2008).
In the soil amoeba Dictyostelium discoideum, six arrestin-related proteins (AdcA–AdcF) have been identified, but have so far been poorly studied (Aubry and Klein, 2013; Aubry et al., 2009). Recently, AdcC has been shown to localize to the plasma membrane in response to cAMP and to bind the cAMP receptor, cAR1 (Cao et al., 2014). The GPCR cAR1 is key in the starvation-induced development of the amoeba, especially during the aggregation phase that allows this unicellular organism to reach multicellularity and enter the differentiation program (Aubry and Firtel, 1999; Parent and Devreotes, 1996). AdcC and its close homolog, AdcB, both exhibit C2 and sterile alpha motif (SAM) domains surrounding the arrestin domain. Whereas SAM domains are mainly known to be protein–protein interaction modules, C2 domains are common phospholipid-binding domains, mostly present in proteins acting in membrane trafficking/fusion events and signal transduction (Corbalan-Garcia and Gómez-Fernández, 2014). In many C2 domain-containing proteins, lipid binding is conditioned by calcium binding, resulting in a calcium-dependent association with membranes. Because cAMP activation of cAR1 induces a large array of signaling events, among which a transient increase in calcium in the cytosol (Nebl and Fisher, 1997; Yumura et al., 1996), we investigated whether calcium might regulate AdcB and AdcC behavior/function. Our work revealed that, in contrast to AdcB, and despite shared biochemical properties, AdcC responds to various stimuli (calcium, cAMP and folate), all inducing a calcium elevation in the cytosol, by transiently associating with the plasma membrane. We uncovered a mode of membrane targeting/regulation unusual for members of the arrestin clan, requiring the C2 domain of AdcC and the SAM extension involved in the oligomerization of the protein. The distinct behaviors of these two homologs in response to internal calcium fluctuations are indicative of highly regulated and specific functions.
The protein AdcC, but not AdcB, translocates to the plasma membrane in response to an increase in intracellular calcium
An in silico analysis of AdcB/AdcC predicted C2 modules was performed using the homology modeling server PHYRE2 (http://www.sbg.bio.ic.ac.uk/phyre2/). The best hits were obtained with C2 domains from members of the PKC-C2 Pfam subfamily. This subfamily includes both calcium-dependent and -independent C2 domains with type I or type II topologies, depending on the orientation of the β-strands within the two four-stranded antiparallel β-sheets forming the C2 domain (Nalefski and Falke, 1996). For AdcB and AdcC, modeling and sequence alignment indicated a type II topology, and conservation of the acidic residues implicated in calcium binding in canonical calcium-dependent C2 domains (D20/D27/D76/D78 and D84 for AdcB; D20/D26/D72/E74 for AdcC), with the exception of one aspartate that is substituted by an arginine in AdcC (R80) (Fig. 1A,B). AdcB and AdcC C2 domains also exhibit an overall conservation of the amino acids involved in PKCα in anionic phospholipid binding (Fig. 1B). The C2 module of both AdcB and AdcC could therefore behave as a calcium-binding C2 domain with a preference for anionic membrane environments.
We investigated the in vivo dynamics of AdcB and AdcC in response to calcium by fluorescence imaging of live cells. Both proteins were tagged with GFP or RFP at the C-terminus and expressed in the parental strain KAx-3 and in adcB or adcC null cells (see Materials and Methods). To study the proteins at relevant stages of the D. discoideum life cycle, we first determined their expression profiles during development by western blot analysis (Fig. S1A,B). Both AdcB and AdcC were detected in growing conditions (vegetative cells) and during the multicellular phase triggered by starvation, with a decrease at later stages of development. Microscopy studies were thus carried out on vegetative cells and cells engaged in the early phase of development.
To test the effect of internal calcium variations on AdcB and AdcC, vegetative cells were stimulated by the addition of CaCl2 in the extracellular medium, a condition known to trigger a cytosolic calcium increase in most cell types, including Dictyostelium (Lombardi et al., 2008; Lusche et al., 2009). We recorded tagged-protein dynamics using spinning disk confocal microscopy over a period of 5–7 min. In untreated cells, both AdcBGFP and AdcCGFP were exclusively found in the cytosol (Fig. 2A,E). Addition of 2 mM CaCl2 led, within the next 1 min, to the enrichment of AdcCGFP at the plasma membrane, with a concomitant decrease of the fluorescent signal in the cytosol, in an oscillatory manner (Fig. 2B; Movies 1 and 2). The cytosol/total index (IC/T) (see Materials and Methods) was used as a measure of AdcC enrichment at the plasma membrane in response to calcium (Fig. 2A,B). Similar behavior was observed for AdcCRFP (Fig. 2C, lower panel) or an N-terminally GFP-tagged AdcC (data not shown). Cell response, in terms of AdcC translocation, was more homogeneous and prominent when cells were transferred in hyposmotic medium prior to calcium addition, possibly due to an increase in calcium entry consecutive to membrane stretching (Lombardi et al., 2008) caused by cell swelling. This treatment was therefore favored, with the perspective to evaluate the behavior of truncated mutants in conditions triggering efficient translocation. On average, in these conditions, the first oscillations were observed within the initial 30 s after stimulation and the number of oscillations varied from 0 to 13 with a mean of 7±3 (n=41) during the 7 min following calcium addition. These values are compatible with the work of Lombardi and colleagues on intracellular calcium fluctuations in similar conditions (Lombardi et al., 2008). Under prolonged incubation (>10–15 min), AdcC progressively re-adopted a stable cytosolic localization. To confirm that AdcC indeed responded to cytosolic calcium variations, AdcCRFP-expressing cells were electroporated with Calcium Green dextran (a calcium indicator) and membrane translocation of AdcCRFP was imaged simultaneously with the fluorescence variation of the calcium probe. As shown in Fig. 2C, AdcCRFP recruitment to the plasma membrane systematically paralleled intracellular calcium peaks. Cell treatment with the calcium-channel blockers Ruthenium Red (RR) and gadolinium (Gd3+) inhibited AdcC relocalization to the membrane, supporting a role for intracellular calcium in AdcC dynamics (Fig. 2D; Fig. S2A). A similar response to calcium was obtained in amoebae subjected to starvation in phosphate buffer, except that AdcCGFP translocation from the cytosol to the plasma membrane could be efficiently triggered with 10-fold lower concentrations of extracellular calcium (200–100 µM) compared with those of vegetative cells, owing to a higher sensitivity of starved cells to calcium (Movie 3). AdcC oscillations progressively attenuated with time as observed in vegetative cells.
Two chemoattractants, folic acid and cAMP, have been described to trigger a transient elevation of cytosolic calcium in vegetative and aggregation-competent amoebae, respectively (Nebl and Fisher, 1997; Yumura et al., 1996; Milne and Coukell, 1991). We therefore examined AdcC behavior upon folate and cAMP stimulation. As the chemoattractant-induced increase in calcium was shown to result primarily from a calcium influx across the plasma membrane, cells were first pre-incubated in phosphate buffer containing 100 µM calcium. In line with our above data, such a concentration of calcium induced some transient AdcC translocation to the plasma membrane (very weak in vegetative cells and marked in aggregation-competent cells; data not shown). Pre-incubation was thus prolonged until AdcC had regained a stable cytosolic localization (Fig. 3A,B; Time 0, top and middle rows). Subsequent stimulation of vegetative cells with folate, but not with buffer alone, led to a rapid and transient translocation of AdcC to the plasma membrane in a timescale consistent with the reported kinetics of the folate-triggered calcium rise (Fig. 3A, top and middle rows). AdcC similarly responded to cAMP when added to aggregation-competent cells (Fig. 3B, top and middle rows), in agreement with the data from Cao et al. (2014). As described, AdcC translocation in response to cAMP was lost in a mutant strain lacking the two cAMP receptors expressed during aggregation (car1/3 null cells), indicating that cAMP operates through a cAMP receptor-dependent pathway (Fig. 3C). In both conditions of stimulation, the chemoattractant-induced translocation of AdcC was not observed in the absence of calcium in the external medium (Fig. S2B,C), or in the presence of the calcium-channel inhibitors RR or gadolinium (Fig. 3A,B, bottom rows; Fig. S2D), supporting that the increase in calcium triggered by folate and cAMP is essential for the recruitment of AdcC to the plasma membrane in response to these chemoattractants. As car1/3 null cells are unable to induce a cytosolic calcium elevation in response to cAMP (Milne et al., 1997), we asked whether causing a calcium elevation in aggregation-competent car1/3 null cells with external calcium could drive AdcC to the membrane. As shown in Fig. 3D, AdcC responded equally well in the car1/3 null and parental Ax2 strains, suggesting that, in a wild-type context, the increase in cytosolic calcium resulting from cAMP receptor activation might be sufficient to recruit AdcC from the cytosol to the plasma membrane.
Taken together, our data indicate that the arrestin-related protein AdcC responds to a variety of stimuli known to elicit an elevation of cytosolic calcium, and that this calcium elevation is essential for AdcC membrane binding. In contrast to AdcC, AdcBGFP did not respond to the addition of calcium in the external medium. The protein remained cytosolic in the range of extracellular calcium concentrations tested (50 µM–40 mM) in vegetative- and aggregation-stage adcB null and KAx-3 cells (Fig. 2E; Movies 4–6). Tagging the protein at the N-terminus rather than C-terminally did not modify AdcB response. In addition, no translocation was observed in response to cAMP or folate (data not shown). Thus, despite their similar architecture, the two proteins AdcB and AdcC harbor distinct sensitivity to calcium in vivo.
AdcC membrane targeting is dependent on its calcium-binding type C2 domain
To assess the contribution of the C2 domain to AdcC response to calcium, GFP-tagged truncated forms of the protein were expressed in the adcC null strain. Deletion of the whole region upstream of the arrestin core (AdcCΔNtGFP), or of the first 101 amino acids, thus solely impairing the C2 domain (AdcCΔC2GFP), completely abrogated AdcC response to calcium stimulation, indicating that the C2 domain is required for AdcC translocation to the plasma membrane (Fig. 4A). Point mutations in the C2 domain of AdcC were also generated, substituting D20 and D26 for asparagines, alone or together with D72 and E74 (AdcCD20N/D26NGFP or AdcCN4GFP) (Figs 1B and 4B). Although they interfer with calcium binding, such substitutions are expected to mimic the charge-neutralizing effect of calcium by removing the electrostatic repulsion provided by the aspartate residues. As reported for similar mutations in DOC2B C2A or copine 2, 6 and 7 C2B domains (Friedrich et al., 2008; Groffen et al., 2006; Perestenko et al., 2015), AdcCD20N/D26NGFP and AdcCN4GFP were found constitutively located at the plasma membrane, further indicating that a functional calcium-sensing C2 domain is required for the detection of intracellular calcium variations and appropriate trafficking of AdcC (Fig. 4B, left).
In contrast to several isolated calcium-binding C2 domains (Ananthanarayanan et al., 2002; Manna et al., 2008; Oancea and Meyer, 1998; Perisic et al., 1999), the C2 domain of AdcC is not sufficient to fully support membrane association of the protein on its own, as neither AdcC-C2GFP nor AdcC-NtGFP, limited to the C2 and Nt domains, respectively, were able to translocate to the plasma membrane in response to calcium (Fig. 4A). Mutation in AdcC-Nt of the four acidic residues involved in calcium binding (AdcC-NtN4GFP) led to the constitutive membrane localization of the mutated protein, but with reduced efficiency compared with the equivalent mutation in full-length AdcC, AdcCN4GFP (Fig. 4B, left). The value of the IC/T was significantly higher for AdcC-NtN4GFP (0.82±0.05, n=25) compared with AdcCN4GFP (0.48±0.14, n=25) and AdcCD20N/D26NGFP (0.39±0.08, n=25) (Fig. 4B, right). Together, these results suggest that, besides the C2 domain, an additional region of the AdcC protein likely contributes to its membrane recruitment.
AdcC and AdcB bind directly to phospholipids in vitro in a calcium-dependent manner
To characterize the lipid-binding properties of AdcC, liposome-binding assays were performed in the presence of calcium or EGTA. Full-length AdcC and the N-terminal domain were expressed as MBP fusion proteins (MBP-AdcC and MBP-AdcC-Nt) and tested for their ability to bind phospholipid vesicles containing phosphatidylcholine (PC) alone, or a mix of PC and phosphatidylserine (PS). Whereas MBP exhibited no binding on any liposomes, MBP-AdcC was found to bind to PC/PS liposomes in a calcium-sensitive PS-dependent manner, suggesting, as expected, a selectivity towards anionic lipids (Fig. 5A,B). Calcium titration experiments using nonlimiting PS concentrations indicated an apparent dissociation constant (Kd) for AdcC of ∼16 µM. In contrast to MBP-AdcC, and in agreement with the constitutive association of AdcCN4GFP to the plasma membrane, MBP-AdcCN4 fully bound to PC/PS liposomes in the presence of EGTA (Fig. 5C). In saturating conditions of calcium and PS, MBP-AdcC-Nt displayed no binding at all (Fig. 5A), further supporting that this domain is not sufficient for efficient membrane binding. However, when expressed as a GST fusion protein, AdcC-Nt was able to bind PC/PS liposomes in a calcium-dependent manner. In contrast to MBP, GST is known to form dimers (Maru et al., 1996; Tudyka and Skerra, 1997), and crosslinking experiments using bis(sulfosuccinimidyl) suberate (BS3) indeed suggested a different oligomeric state for GST- and MBP-tagged AdcC-Nt proteins (Fig. S3A,B). This GST-promoted oligomerization could increase the avidity of the construct for anionic lipids, thereby favoring binding. These results, and the distinct behavior of the full-length versus N-terminal alone constructs in cellulo, further support the hypothesis that an additional domain of AdcC participates directly or indirectly in membrane association, together with the C2 domain.
Similar experiments on AdcB showed, unexpectedly, that full-length MBP-AdcB, as well as MBP-AdcB-Nt (but to a lower extent), can associate with PS-containing liposomes upon calcium addition (Fig. 5D,E). Similar to observations for MBP-AdcC, half-maximal binding of MBP-AdcB to 50% PC/50% PS liposomes was observed with a free calcium concentration of ∼18.5 µM. As observed for AdcC-Nt, use of a GST tag also improved calcium- and PS-dependent binding of the truncated AdcB-Nt construct (Fig. 5D), again likely to be caused by the oligomerizing property of the GST (Fig. S3C). Therefore, the C2 domain of AdcB can bind anionic lipids in a calcium-dependent manner in vitro and qualifies as a bona fide calcium-sensitive C2 domain. However, swapping the N-terminal domain of AdcC in the AdcCGFP protein for the N-terminal domain of AdcB (AdcB/AdcCGFP chimera) inhibited AdcC translocation to the plasma membrane in vivo in response to 2 mM extracellular CaCl2 (Fig. 5F, right). In contrast, AdcC/AdcBGFP chimera, in which the Nt domain of AdcC replaced that of AdcB, responded with a massive and oscillatory membrane association similarly to AdcC (Fig. 5F, left), indicating distinct properties for AdcB and AdcC C2 domains in vivo.
The SAM domain-containing region of AdcC contributes to its plasma membrane association
A preliminary analysis of AdcC and AdcB using SMART (http://smart.embl-heidelberg.de/) had revealed the presence of a putative SAM domain at the C-terminal end of the proteins (Aubry and Klein, 2013; Aubry et al., 2009). SAM domains are ∼70 amino acid modules that exhibit limited sequence homologies but a well-conserved structure (Fig. 6A). They are most often found as single units, but sometimes as repeats, such as in liprins, CASKIN and AIDA proteins (Kurabi et al., 2009; Stafford et al., 2011; Wei et al., 2011). Based on the structure of liprin-α2 (PDB c3tadB), PHYRE2 predicted the presence of a second SAM domain for AdcB and AdcC, just downstream of the arrestin domain. Fig. 6B and C show the three-dimensional structure models obtained from PHYRE2 under intensive mode using a set of SAM-containing templates selected to maximize coverage and confidence. In contrast to proteins containing SAMs in tandem, the linker separating the two SAM domains is shorter in AdcB and AdcC, implying a possibly different spatial organization of the domains relative to each other. To investigate a possible contribution of the SAM domain-containing region to AdcC response, C-terminally truncated forms were generated, removing the last (AdcCΔSAM2GFP) or both SAMs (AdcCΔSAM1/2GFP) (Fig. 6D). When expressed in adcC null or KAx-3 cells, these two cytosolic proteins responded to calcium by some visible, but limited, membrane translocation (Fig. 6D in adcC null cells). Recruitment to the plasma membrane was significantly impaired compared with full-length AdcC (Fig. 6E), indicating that this domain is required for efficient AdcC membrane translocation.
AdcB and AdcC homo-oligomerize in a SAM-dependent manner
Because SAM domains can mediate homo- or hetero-oligomerization of SAM domain-containing proteins (Kim and Bowie, 2003; Qiao and Bowie, 2005), we tested the possibility that AdcB and AdcC might exist as stable oligomers. Analysis of the purified proteins in native conditions by blue-native (BN)-polyacrylamide gel electrophoresis (PAGE) indicated that MBP-AdcB and MBP-AdcC (∼110 kDa) migrate as high-molecular-weight species (Fig. 7A). Their molecular weight, estimated at ∼1000–1200 kDa, suggested that the purified proteins self-interact in complexes of ≥10 subunits. To further characterize these complexes, MBP-AdcB and MBP-AdcC were examined by negative-staining electron microscopy. This revealed a monodisperse preparation, with visible ring-shaped structures (∼20–30 nm diameter), which were absent from the MBP alone preparation, providing additional evidence that MBP-AdcB and MBP-AdcC auto-assemble in organized structures in vitro (Fig. 7B). Treatment of MBP-AdcB and MBP-AdcC with calcium or EGTA did not affect the overall appearance of the oligomers in electron microscopy (Fig. S4), or the size of the complexes on nondenaturing gels (data not shown).
AdcB-AdcB and AdcC-AdcC interactions were confirmed in vivo in yeast two-hybrid assays (Fig. 7C), as well as in co-immunoprecipitation experiments using KAx-3 cells expressing AdcCGFP or adcB null cells expressing AdcBRFP/AdcBGFP (Fig. 7D,E). Endogenous AdcC or AdcBGFP were specifically pulled down with immunoprecipitated AdcCGFP and AdcBRFP, respectively. Calcium addition to the lysis buffer had no effect on AdcB or AdcC homotypic interactions (data not shown). Whereas deletion of the N-terminal domain of AdcB or AdcC (ΔNt constructs) did not noticeably affect pulldown of the full-length proteins, truncation of the C-terminal SAM (ΔSAM2 constructs) or of the SAM tandem (ΔSAM1/2 constructs) in AdcB or AdcC markedly interfered with binding (Fig. 7C–E). Accordingly, constructs limited to the Nt domains also failed to bind the full-length proteins (Fig. 7D,E). These results therefore support a role for the SAM domain-containing region in the oligomerization of the proteins.
When expressed in Dictyostelium, the constructs limited to the SAM-containing region of AdcB and AdcC (AdcB-SAM1/2 and AdcC-SAM1/2) could not be recovered in the soluble fraction after cell lysis, precluding co-immunoprecipitation approaches to test the implication of a SAM–SAM interaction in the oligomerization. As an alternative, a C-terminally S-tagged version of AdcC SAM tandem (AdcC-SAM1/2-Stag) was expressed in bacteria together with an N-terminally His-tagged version (His-AdcC-SAM1/2). Purification of His-AdcC-SAM1/2 on Ni-NTA beads allowed the co-purification of the S-tagged protein (Fig. 7F), suggesting that AdcC homo-oligomerization might involve a SAM–SAM interaction.
Analysis by BN-PAGE of a cytosolic fraction from vegetative KAx-3 indicated that, in nondenaturing conditions, cytosolic AdcB and AdcC were exclusively found in high-molecular-weight complexes of ∼750 kDa (Fig. 7G). Their exact composition is currently unknown, and we cannot exclude a possible association with partners. However, in the hypothesis of homo-oligomeric complexes, the molecular weight of ∼750 kDa estimated from the gel would again suggest the presence of ∼10 subunits, as obtained for the in vitro complex with the MBP-tagged counterparts. Similar analyses were conducted on a plasma membrane-enriched fraction from KAx-3 cells after calcium stimulation to examine whether membrane-associated AdcC is also part of a complex. Prior to electrophoretic separation, the fraction was treated with 0.5% dodecylmaltoside, allowing the solubilization of ∼50% of the protein. This pool of AdcC was mostly found associated with a high-molecular-weight complex, similar in size to the one present in the cytosol, and occasionally with much larger species, migrating as smears, that could correspond to precipitated protein (Fig. 7G). To test whether the membrane complexes contain several AdcC entities, we analyzed the ability of AdcCRFP to bring to the plasma membrane the Nt-deleted construct, AdcCΔNtGFP, which is unable to reach the membrane on its own (Fig. 4A). Although GFP alone was insensitive to the presence of AdcCRFP at the membrane and remained cytosolic (Fig. 7H, right), expression of AdcCRFP restored some AdcCΔNtGFP translocation to the plasma membrane upon calcium stimulation, which paralleled that of AdcCRFP (Fig. 7H, left), indicating that AdcC associates with the membrane as an oligomeric complex.
AdcB is a partner of AdcC
Given the presence of SAM domains in both AdcB and AdcC, we tested a possible interaction between the two proteins. In KAx-3 cells, the endogenous proteins were found to co-immunoprecipitate, independently of the antibody used for immunoprecipitation (anti-AdcB or anti-AdcC) (Fig. 8A). Binding of AdcB to AdcC is likely to be direct, as AdcB and AdcC also strongly interacted with each other in yeast two-hybrid assays (Fig. 8B). The interaction was maintained upon truncation of the AdcB-Nt domain, but was completely abolished by deletions removing the SAM domains of either protein (Fig. 8B,C), indicating that the SAM-containing regions are essential for the formation of the hetero-oligomer. Despite this interaction, co-expression of AdcCRFP together with AdcBGFP was not sufficient to efficiently recruit AdcB to the plasma membrane upon calcium stimulation (Fig. S5). Barely visible membrane staining, if any, was observed with AdcBGFP in cells harboring membrane-associated AdcCRFP, indicating that AdcB is not part of the AdcC membrane complex, or is in a very limited amount compared with AdcC. BN-PAGE analyses of Dictyostelium cytosolic fractions showed that deletion of adcB or adcC did not visibly alter the size or the abundance of the AdcC- and AdcB-positive complexes, respectively (Fig. 7G), suggesting that the AdcB- and AdcC-positive complexes detected around 750 kDa are unlikely heteromeric. It cannot be excluded, however, that in the single-null background, the remaining protein substitutes for the disrupted one, allowing the formation of complexes of a similar size to those in the wild-type context.
A direct role for calcium in AdcC targeting to the plasma membrane
Arrestin proteins respond to a variety of external signals, among which a large panel of GPCR ligands. Although cytosolic in resting conditions, these adaptor proteins translocate to their GPCR targets at the plasma membrane upon receptor ligand activation, and modulate their fate and associated signaling. For most GPCRs, arrestin binding involves recognition, by different regions of the arrestin core, of the activated conformation of the receptor and its phosphorylated state, even though requirement for receptor phosphorylation is not absolute (Zhou et al., 2017; Gurevich and Gurevich, 2013; Tobin, 2008). In this work, we established that the arrestin-related protein AdcC of Dictyostelium is transiently targeted to the plasma membrane in response to various stimuli (external calcium, cAMP and folate), all inducing a cytosolic calcium elevation, and we provide evidence that calcium is a direct regulator of AdcC localization. In contrast to mammalian canonical arrestins, AdcC displays a C2 domain located at the N-terminal extremity of the protein and a C-terminal SAM domain-containing extension. We showed here that the C2 domain confers the protein calcium-sensing and lipid-binding properties, and that a functional C2 domain is essential for the translocation of AdcC to the plasma membrane: (1) deletion of the C2 domain prevents AdcC membrane translocation in response to external calcium; (2) recombinant AdcC, as well as its N-terminal domain alone, can bind anionic lipids in a calcium-dependent manner in liposome sedimentation assays, provided that the N-terminal domain is expressed with the dimerizing GST tag that partially substitutes for the oligomerization-mediating SAM domain in the full-length protein; and (3) interfering with calcium binding through point mutations in the C2 domain (AdcCN4) modifies AdcC properties both in vivo and in vitro.
So far, the only known membrane target of AdcC is the high-affinity cAMP receptor cAR1 (Cao et al., 2014). This GPCR-type receptor is essential during the early stages of Dictyostelium development. At this stage, pulses of cAMP in the nM range, released by starving cells, orchestrate their chemotactic aggregation. As cells reach mound stage, they are subjected to higher concentrations of cAMP, leading to the activation of cell-type differentiation-specific pathways. The repertoire of cAMP receptors present at the plasma membrane is modified, with a partial replacement of cAR1 by lower-affinity members of the cAMP receptor family (Kim et al., 1998; Ginsburg et al., 1995; Sergé et al., 2011). Stimulation of cAR1 by cAMP (as well as other cARs) has been shown to trigger a transient calcium influx, the extent of which depends on ligand concentration (Nebl and Fisher, 1997; Yumura et al., 1996; Milne and Devreotes, 1993). In this work, we established that calcium entry and subsequent cytosolic elevation play a key role in the cAMP-induced recruitment of AdcC to the plasma membrane, as no translocation is observed in the absence of external calcium or in the presence of calcium channel inhibitors. Whether calcium-dependent membrane binding (via the C2 domain) occurs simultaneously to, or precedes, receptor recognition (likely through the arrestin domain) is not yet clarified. Our results showing the calcium-dependent lipid-binding properties of recombinant AdcC in vitro, and the fact that an increase in external calcium, and thus in cytosolic calcium, is sufficient to trigger transient AdcC membrane binding in the absence of cAMP, or in a car1/3 null mutant, argue in favor of a model in which the calcium increase caused by the chemoattractant is the primary signal driving AdcC recruitment from the cytosol to the plasma membrane. This hypothesis is also supported by data from Cao et al. (2014), showing that although AdcC preferentially interacts with the cAMP-activated phosphorylated form of cAR1 in co-immunoprecipitation experiments, its translocation to the plasma membrane in response to cAMP is not visibly perturbed in a car1/3 null strain expressing the cm1234 version of cAR1, which is nonphosphorylatable but is still capable of causing calcium influx (Milne et al., 1995). Nonetheless, in mammals, receptor phosphorylation is not a systematic prerequisite to arrestin recruitment (Tobin, 2008). In addition, besides GPCRs, members of the arrestin clan are known to regulate a variety of other receptor and nonreceptor membrane proteins (Lefkowitz et al., 2006; Becuwe et al., 2012a). Stimulation of car1/3 null cells (and parental strains as well) with external calcium could not only increase internal calcium but also ‘activate’ some other membrane targets of AdcC, which could contribute to the recruitment of the arrestin to the plasma membrane, together with the C2 domain/lipid interaction. Identification of the complete repertoire of AdcC targets and further analysis of the chronology of the events following cAMP stimulation are needed to fully understand the modalities of recruitment and the functioning of this unconventional arrestin in this cAMP context.
The involvement, in Dictyostelium, of a calcium-dependent component in the recruitment of the arrestin AdcC at the plasma membrane raises the question of the gain of such a calcium-dependent regulation, compared with the situation in mammals. In the context of cAMP signaling, membrane docking of AdcC in response to the increase in calcium triggered by cAR1 stimulation could contribute to amplify the effect of cAMP and facilitate receptor recognition of AdcC by concentrating the arrestin in the receptor-close environment. As this calcium influx is dependent on the cAMP concentration, we propose that intracellular calcium, by reflecting the level of external cAMP, allows a spatiotemporal control of AdcC activities by setting the conditions (such as time and amount) of recruitment of the arrestin to the plasma membrane as a function of the stimulus intensity, thereby adjusting the effect of the arrestin on its membrane target fate and/or associated signaling differently depending on the developmental stage. Cao et al. (2014) have established that disruption of adcC and of its homolog adcB (see below) affects the development of the double-null strain, with altered Erk2 phosphorylation kinetics during aggregation and reduced internalization of cAR1 postaggregation. To pursue the functional characterization of AdcC, it would have been interesting to test the ability of truncated and point mutants of AdcC to substitute for the wild-type protein and complement the adcC/B null defects. In our hands, the KAx-3 cells devoid of adcC or adcB/C (this work), and the double null strain (Ax2 background) from Cao and colleagues obtained from the DictyStock Center, displayed no obvious defect regarding development, Erk2 phosphorylation kinetics and internalization of cAR1 (L.A. and L.M., unpublished), precluding complementation assays. The reasons for the discrepancy between our observations and their published data are currently unclear, but they could – in part – reside in differences in culture conditions or experimental procedures.
Given that AdcC also responds to folate, it is possible that the recently identified GPCR-type folate receptor fAR1 (Pan et al., 2016) is also a target of AdcC. This receptor is required for both cell chemotaxis towards folate-secreting bacteria and their phagocytosis (Pan et al., 2016). As in the case of cAMP, cell stimulation with folate has been shown to induce an increase in intracellular calcium that depends on folate concentration in the extracellular medium (Nebl and Fisher, 1997; Yumura et al., 1996). Depending on fAR1 activation level, and on the intracellular calcium signal generated, AdcC could intervene in the fate or functions of the receptor, and adjust the cell response to the chemotactic and/or phagocytic contexts.
In addition to the direct role of calcium on AdcC localization, we identified three calcium-binding proteins among the putative interactors of AdcC in a yeast two-hybrid screen (L.M. and L.A., unpublished), suggesting that calcium might additionally impact AdcC-dependent signaling by modulating its interaction with partners and/or the resulting outcomes.
AdcB and AdcC: alike but different
Despite high structural similarities, the present study unveiled a major difference between AdcC and its homolog, AdcB. In contrast to AdcC, AdcB was never observed at the plasma membrane in response to a calcium-induced elevation of cytosolic calcium. This result is coherent with AdcB insensitivity to folate and cAMP (our own observations; Cao et al., 2014). Our data using chimeric proteins indicate that the inability of AdcB to bind membranes in response to calcium likely resides in the N-terminal domain. However, in vitro characterization of the AdcB C2 domain in liposome-based sedimentation assays revealed a genuine calcium-dependent C2 module. For many C2 domain-containing proteins with a preference for anionic lipids, phosphatidylinositol 4,5-bisphosphate [PI(4,5)P2] contributes to their membrane association, together with PS, by binding to a distinct surface, the basic β-groove region (Cho and Stahelin, 2006; Guerrero-Valero et al., 2009; Guillén et al., 2013). Some preliminary experiments indicated that the addition of 5% PI(4,5)P2 to 20% PS-containing liposomes significantly increases binding of both AdcC and AdcB in a similar manner, without modifying their calcium affinity, suggesting that the distinct behavior of the two proteins is not a consequence of a different response to that particular phosphoinositide. The opposite response of the two proteins in vivo could rather result from distinct activatory/inhibitory post-translational modifications or partners targeting the C2 domain, thereby impacting calcium and/or membrane binding as shown for several members of the synaptotagmin and PKC protein families (Lee et al., 2004; Pepio and Sossin, 2001; Roggero et al., 2005). Independently of the mechanism(s) responsible for this behavioral difference, our data argue against a full functional redundancy between AdcB and AdcC, at least for the functions fulfilled by AdcC at the plasma membrane. In this context, identification of AdcB as a partner of AdcC raises the question of the functional role of such interaction. Additional experiments will be required to establish whether the association of AdcB with AdcC allows a specific set of functions or plays some regulatory role, for example in controlling a pool of cytosolic AdcC, available for membrane targeting.
AdcC and AdcB: large polymeric platforms
We have shown here that AdcB and AdcC exist mainly as part of high-molecular-weight complexes in vivo. Their composition, which could include several partners, is currently unknown, but our results support the presence of several monomers per complex: (1) purified proteins form large size homo-oligomers in vitro; (2) two-hybrid and co-immunoprecipitation experiments established that AdcB and AdcC are able to self-interact in cellulo; and, (3) in the case of AdcC, the calcium-triggered membrane-bound form contains several copies of AdcC. Our data indicate that the SAM module is involved in oligomerization, and a direct SAM-SAM interaction, as shown for AdcC on truncated domains, could very well account for the organized structures observed by electron microscopy with recombinant full-length proteins. Many isolated SAM domains can produce helical polymeric assemblies (fibers) through head-to-tail interactions (Harada et al., 2008; Kim et al., 2002; Knight et al., 2011; Thanos et al., 1999). Such fibers were not present in our preparations of MBP-tagged full-length proteins, but were sometimes observed with the isolated MBP-tagged SAM region of AdcC (C.D. and L.A., unpublished). In the full-length proteins, polymerization might be restrained by a specific arrangement of the SAM tandem and/or the presence of the long N-terminal extension. Besides the SAM region, oligomer formation could also involve the arrestin core, as described for mammalian β-arrestins and visual arrestin-1 (Chen et al., 2014). β-arrestins, which are devoid of SAM domains, are able to self-associate as dimers or tetramers through their N and C domains in an IP6-dependent manner, and their oligomeric status has been shown to impact their subcellular localization and functions (Chen et al., 2014; Hanson et al., 2008; Milano et al., 2006). AdcB and AdcC arrestin domains alone were found to be poorly expressed in Dictyostelium and bacteria, but their capacity to multimerize and their regulation by IP6 would certainly be worth investigating, all the more because this inositol polyphosphate, which is highly abundant in the amoeba (Martin et al., 1987), might modulate lipid binding by the C2 domain, as shown for the mammalian synaptotagmin 1 C2B domain (Joung et al., 2012; Lu et al., 2002).
Oligomerization could serve different purposes. In the case of AdcC, we proposed that oligomerization increases the avidity of the complex for membrane lipids by gathering multiple C2 domains, thereby indirectly participating in AdcC membrane binding. Given the multimodular organization of the protein, it could also directly contribute to the assembly of signaling platforms or specific architectures in the vicinity of activated membrane targets, and thereby modulate/facilitate receptor downstream signaling or internalization in the endocytic pathway.
To conclude, our study characterized the AdcC protein of the Dictyostelium amoeba as a novel calcium sensor, showing that calcium plays a key role in AdcC recruitment to the plasma membrane in response to various stimuli, such as the chemoattractants cAMP and folate. It revealed an unusual mode of regulation for a protein of the arrestin clan that involves a C2 domain with calcium-dependent membrane-targeting properties, and a double SAM domain involved in AdcC oligomerization, which is necessary for its association with the membrane and interaction with the calcium nonresponsive homolog AdcB. Despite high homology and common properties, our data suggest distinct roles and specific regulatory mechanisms for these two arrestin-related scaffolding proteins.
MATERIALS AND METHODS
Strains and cell culture
Most experiments were conducted on the D. discoideum parental strain KAx-3 from the R. Firtel laboratory (University of California San Diego, La Jolla, CA, USA) and derived knockout or overexpressing strains. The car1/3 null cells and their parent Ax2 cells were obtained from P. Devreotes (Johns Hopkins University School of Medicine, Baltimore, MD, USA). All cells were grown in Petri dishes or in shaking culture at 21°C in maltose-containing HL5 medium. Overexpressors were selected by addition of G418 (20 µg/ml) or hygromycin (40 µg/ml), depending on the expression vector. The adcB and adcC null mutants were selected in the presence of blasticidin (7.5 µg/ml). When required, cells were washed and resuspended in 12 mM NaK-phosphate buffer pH 6.2 (PB), then starved for 2–4 h (starved cells) or pulsed with 100 nM cAMP every 6 min for 4 h (aggregation-competent cells).
The single-null adcB−, adcC− and double-null adcB−/adcC− mutants were generated by homologous recombination using the Cre-Lox pLPBLP vector (blasticidin resistance BsR cassette) (Faix et al., 2004). Genomic DNA fragments corresponding to bp 223–545 (ClaI-HindIII) and 814–1160 (BamHI-SpeI) for AdcB and bp 58–426 (ClaI-HindIII)/451–951 (BamH1-EcoRI/SpeI) for AdcC were amplified by PCR with oligonucleotides containing the mentioned restriction sites and subcloned in pLPBLP on each side of the BsR cassette. The resulting plasmids were linearized with ClaI and ClaI/EcoRI prior to electroporation in KAx-3 cells. After selection, cells were cloned by plating on SM agar plates in association with Klebsiella aerogenes. To obtain the adcB−/adcC− mutant, the adcC− null mutant was transformed with the pDEX-NLS-Cre vector for BsR cassette extraction (Faix et al., 2013), and used to introduce the second disruption construct as described above. Gene knockouts were confirmed by PCR, Southern blot and/or western blot analyses (Fig. S1A) of individual clones. For overexpression purposes, the following AdcB- and AdcC-derived constructs were generated by subcloning PCR fragments in the vectors pExp4+ (G418R), pDM1045 and/or pDM1043 (HygromycinR): AdcB [amino acids (a.a.) 1–617], AdcB-Nt (a.a. 1–174), AdcBΔC2 (a.a. 106–617), AdcBΔNt (a.a. 166–617), AdcBΔSAM2 (a.a. 1–547), AdcBΔSAM1/2 (a.a. 1–472), AdcC (a.a. 1–654), AdcCD20N/D26N, AdcCD20N/D26N/D70N/E74N (AdcCN4), AdcC-C2 (a.a. 1–120), AdcC-Nt (a.a. 1–169), AdcC-NtN4, AdcCΔC2 (a.a. 101–654), AdcCΔNt (a.a. 160–657), AdcCΔSAM1/2 (a.a. 1–485) and AdcCΔSAM2 (a.a. 1–567), and the chimeric proteins AdcB (a.a. 1–168)/AdcC (a.a. 162–654) and AdcC (a.a. 1–161)/AdcB (a.a. 169–617). For biochemical analyses, AdcB, AdcB-Nt, AdcC, AdcC-Nt and AdcCN4 were expressed as GST and/or MBP fusion proteins from pGEX-KG or pMAL-C2 plasmids. AdcC-SAM1/2 (a.a. 485–654) and AdcC-Nt were expressed as His6-tagged proteins using the plasmids pET-duet1 and pET22, respectively. AdcC-SAM1/2 corresponding cDNA was also subcloned in the second site of pET-duet1 to co-express His-tagged and S-tagged versions of the domain. Constructs requiring PCR amplification were verified by sequencing (Beckman Coulter).
Recombinant protein expression and purification
Recombinant proteins were expressed in Bl21-DE3 Escherichia coli as described (Becuwe et al., 2012b), but with a 2% ethanol treatment. For GST- and MBP-tagged proteins, bacterial pellets were resuspended in PBS (pH 7.4), 5 mM EDTA-containing protease inhibitors and 1 mg/ml lyzozyme. Bacteria were sonicated after addition of 1% Triton X-100 and three volumes of lysis buffer A (PBS, 10 mM EDTA, 1 mM EGTA, 3 mM DTT, protease inhibitors). After clearing of the lysate (30 min, 16,000 g), recombinant proteins were purified on glutathione sepharose (GE Healthcare) or amylose resin (New England Biolabs) according to the manufacturer’s instructions. After several washes in buffer A containing 1% Triton X-100 and a wash in PBS, GST- and MBP-tagged proteins were eluted in 10 mM glutathione, 50 mM Tris pH 8.0 or 10 mM maltose, 20 mM Hepes pH 7.4, 100 mM NaCl, 1 mM EGTA, respectively. Proteins were dialyzed against 20 mM Hepes pH 7.4, 100 mM NaCl and quantified before use. The His-tagged proteins were purified on Ni-NTA beads in the presence of 20 mM imidazole as described in Guetta et al. (2010) and eluted in 150 mM imidazole.
Production of antibodies and western blot analysis
Antibodies against AdcB and AdcC were raised in Dunkin Hartley guinea pigs and New Zealand White rabbits, respectively (Covalab, Villeurbanne, France). Purified GST-AdcB-Nt and His6-AdcC-Nt proteins were used as antigens, and antibodies were, respectively, purified on immobilized GST-AdcB-Nt (after removal of anti-GST antibodies) and GST-AdcC-Nt as described (Harlow and Lane, 1988). Antibody specificity was assessed by western blotting using protein extracts from adcB and adcC null mutants as controls. Anti-GFP (7.1/13.1, 1:1000), anti-RFP (5F8 or 3F5, 1:1000), anti-His (27471001, 1:1500), anti-Stag (71549-3, 1:5000) and anti-MBP (E8032, 1:10,000) were from Roche, ChromoTek, GE-Healthcare, Millipore and New England Biolabs, respectively. Western blots were performed on polyvinylidene difluoride (PVDF) membranes blocked in 1% bovine serum albumin, or as suggested by the manufacturers.
Approximately 1–2×107 Dictyostelium cells were resuspended in lysis buffer (10 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.5 mM EDTA, 0.5% NP40 and protease inhibitors). After 10 min on ice, the lysates were centrifuged at 16,000 g for 10 min. The appropriate antibodies (anti-GFP 7.1/13.1, 1:500; anti-RFP 5F8, 1:500; anti-AdcC, 1:200; or anti-AdcB, 1:200) and protein A or G agarose (from Roche) were added to the supernatants. After 1 h of incubation and washes in lysis buffer, the proteins were recovered by addition of denaturing buffer. Alternatively, immunoprecipitation was performed using the GFP- or RFP-Trap MA kit (ChromoTek) on whole-cell NP40 extracts as described by the manufacturer.
Liposome binding assay
For co-sedimentation assays, sucrose-loaded liposomes were generated from mixtures of L-α-phosphatidylcholine (PC) and/or L-α-phosphatidylserine (PS) (P3556 and P7769 from Sigma-Aldrich) as described (Mosior and Epand, 1993), except that lipid rehydration was performed in 20 mM Hepes pH 7.4, 180 mM sucrose, and liposomes were centrifuged for 30 min at 20,000 g (22°C), and then washed and resuspended in 20 mM Hepes pH 7.4, 100 mM NaCl (Hepes/NaCl buffer). Proteins (∼0.5 µM) were incubated with liposomes (0.9 mg/ml) and EGTA (1 mM) or CaCl2 at the desired concentration, for 30 min at room temperature. Liposomes were pelleted as above, generating a supernatant and a pellet that was washed and resuspended in Hepes/NaCl buffer with EGTA or CaCl2 prior to transfer to a new tube. Equivalent amounts of supernatant (free protein) and pellet (bound protein) were analyzed by sodium dodecyl sulfate (SDS)-PAGE. Coomassie-stained proteins were quantified using ImageJ software (https://imagej.nih.gov/ij/). Data were fitted to the Hill equation y=a[Ca2+]n/(Kdn+[Ca2+]n), where ‘n’ is the Hill coefficient, ‘a’ an arbitrary normalization constant and ‘Kd’ the apparent dissociation constant. Ca2+/EGTA buffers with free calcium in the range of 0.1–1000 µM were prepared according to MaxChelator (http://maxchelator.stanford.edu).
Cells were allowed to adhere to eight-well Labtek chambered coverglasses in HL5 culture medium (vegetative cells) or in PB (starved or pulsed cells). Prior to imaging, the medium was exchanged for PB or distilled water (vegetative cells) or PB (starved or pulsed cells). In the case of vegetative cells, water or PB was used, as the cell response to calcium was stronger in hypo-osmotic conditions. After several minutes, cells were treated with CaCl2 in the same medium and imaged within the next 6–8 s on a confocal spinning disk inverted microscope (Nikon TI-E Eclipse) equipped with a Yokogawa motorized confocal head CSUX1-A1 and an Evolve EMCCD camera (1 frame/2.5 s). For folate and cAMP stimulation, vegetative and pulsed cells, respectively, were first placed in PB containing 100 µM CaCl2, as external calcium was shown to be required for the chemoattractant-induced increase in calcium. Because calcium induces a transient phase of oscillations of AdcC at the plasma membrane (see Results), cells were left in buffer until AdcC regained a stabilized cytosolic localization. Cells were then stimulated with 50 µM folate or 10 µM cAMP and imaged. For Ruthenium Red (RR) treatment, cells were exposed to 20 µM RR for several minutes prior to the addition of the appropriate stimulus. Image acquisition was performed at a median z-plan every 2.5 s for 5–7 min using Metamorph software. Fluorescence quantification was performed using ImageJ software. Mean total (T) and cytosolic (C) fluorescence density values were measured on a median z-plan at time points of maximal membrane labeling on regions of interest (ROIs) corresponding to the whole cell or to the cytoplasm, respectively. Nuclei and large-size cytoplasmic vesicles visible on the z-plan were purposely removed from the selections to restrain the measurement to the cytosol contribution only. The mean background value obtained from ROIs taken outside of the cells was subtracted from the T and C values, leading to corrected C and corrected T values. The IC/T corresponding to the ratio (corrected C/corrected T) was used as an estimation of membrane translocation (or membrane binding in the case of constitutive binding) efficiency of the constructs. AdcC-binding sites at the plasma membrane were visibly not saturated in our conditions of protein expression.
For Calcium Green-1 dextran (CG-1) imaging, 6×106 cells were resuspended in 50 µl PB containing 50 mM sucrose and 8 mg/ml CG-1 and electroporated (1 pulse, 500 V, 3 µF, 0.2 cm gap cuvette) before transfer to HL5 medium. After cell adhesion, the medium was replaced by fresh HL5 and by PB 1 h later. Imaging was performed within the next 1–2 h. To compare CG-1 variations of fluorescence intensity and AdcCRFP behavior, corrected mean fluorescence density values were measured as a function of time on ROIs corresponding to the cytoplasm only for CG-1, or to the cytoplasm and to the plasma membrane for AdcCRFP.
Yeast two-hybrid interaction and β-galatosidase assays
Two-hybrid interaction assays were performed using the LexA two-hybrid system (Gyuris et al., 1993) in the EGY48 S. cerevisiae strain containing the LacZ reporter construct pSH18-34. The vectors pEG202 and pJG4-5 (GAL1 promoter) were used to express the bait and prey constructs, respectively. Interactions were examined on plates under high stringency selection conditions [synthetic dropout (SD) medium containing 2% galactose (Gal) instead of glucose and lacking Ura, His, Trp and Leu] and confirmed by re-streaking the colonies on Gal-SD lacking Ura, His and Trp, supplemented with 5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside (X-gal). The β-galactosidase activity was quantitatively assayed in liquid medium using O-nitrophenyl β-D-galactopyranose as a substrate. The background activity was obtained from the strain expressing the bait constructs and containing empty pJG4-5.
Transmission electron microscopy
For negative staining, 2 µl containing 30 ng purified MBP-AdcB or MBP-AdcC was loaded in a carbon-mica interface. The carbon layer was floated on 2% uranyl acetate solution, recovered with a 400-mesh copper grid (Agar Scientific), air dried and observed with a JEOL 1200EX transmission electron microscope at 80 kV. Images were taken with a digital camera (Veleta, Olympus) at 100,000× magnification.
Approximately 5×107 cells were resuspended in 500 µl 50 mM Tris, pH 7.5, 50 mM NaCl plus protease inhibitors and stimulated with or without 2 mM CaCl2 for 5 min at 21°C in shaking conditions. Cells were directly broken through a 3-µm pore polycarbonate filter in the presence of CaCl2 or 2 mM EGTA for nonstimulated cells. Lysates were centrifuged for 10 min at 1000 g to recover the AdcC- and plasma membrane-enriched fraction (pellet, P) in the CaCl2 condition, or for 30 min at 100,000 g to obtain the AdcC-enriched soluble fraction (supernatant, S) in the EGTA condition. The pellet was resuspended in 500 µl of the same buffer containing 0.5% dodecylmaltoside, left for 1 h on ice with regular vortexing, and centrifuged at 16,000 g for 20 min to remove nonsolubilized material. All samples were stored at −20°C in 15% glycerol. Soluble and membrane samples corresponding to 2×105 and 4×105 cells, respectively, were analyzed by BN-PAGE on 3–12% Bis-Tris native gels (Novex, Thermo Fisher Scientific) according to the manufacturer’s instructions. BN-PAGE gels were soaked in transfer buffer (25 mM Tris, 192 mM glycine, 0.1% SDS, 20% EtOH pH 8.3) for 20 min prior to protein transfer to PVDF membranes. For analysis of purified MBP-tagged proteins, 1–5 ng of material was loaded on the native gel and separated as above.
We thank F. Letourneur and A. Bouron for critical reading of the manuscript; the group of C. Picart for advice on liposome binding assays; T. Rabilloud, E. Faudry, R. Dumas and F. Parcy for discussions; T. Soldati for pDM1045 and pDM1043 plasmids; R. Firtel for the KAx-3 cells; and P. Devreotes for the car1/3− and parent Ax2 strains. We also acknowledge the µLife cell imaging platform of the Biosciences and Biotechnology Institute of Grenoble and the Electron Microscopy Facility of Grenoble (MEC).
Conceptualization: L.A.; Methodology: L.M., C.D., A.J., L.A.; Validation: L.M., A.C., C.D., A.J., L.A.; Formal analysis: L.M., A.J., L.A.; Investigation: L.M., A.C., C.D., A.J., L.A.; Resources: C.D.; Writing - original draft: C.D., A.J., L.A.; Writing - review & editing: L.M., A.C., C.D., A.J., L.A.; Visualization: L.M., C.D., A.J., L.A.; Supervision: L.A.; Project administration: L.A.; Funding acquisition: L.A.
This work was supported by the Commissariat à l'Énergie Atomique et aux Énergies Alternatives, Institut National de la Santé et de la Recherche Médicale, Université Grenoble Alpes, Centre National de la Recherche Scientifique and Agence Nationale de la Recherche [DYNOTEP ANR-12-BSV6-0016-01]. L.M. was the recipient of a fellowship from the Ministère de l'Education Nationale, de l'Enseignement Supérieur et de la Recherche/Université Grenoble-Alpes.
The authors declare no competing or financial interests.