Arginine (Arg)-based endoplasmic reticulum (ER)-localization signals are involved in the quality control of different heteromultimeric membrane protein complexes. ATP-sensitive potassium (KATP) channels are unique because each subunit in the heterooctamer contains an Arg-based ER-localization signal. We have dissected the inactivation events that override the ER-localization activity of the eight peptide-sorting motifs. Employing a 14-3-3-scavenger construct to lower the availability of 14-3-3 proteins, we found that 14-3-3 proteins promote the cell-surface expression of heterologously expressed and native KATP channels. 14-3-3 proteins were detected in physical association with KATP channels in a pancreatic β-cell line. Our results suggest that the Arg-based signal present in Kir6.2 is sterically masked by the SUR1 subunit. By contrast, 14-3-3 proteins functionally antagonized the Arg-based signal present in SUR1. The last ten amino acids were required for efficient 14-3-3 recruitment to multimeric forms of the Kir6.2 C-terminus. Channels containing a pore-forming subunit lacking these residues reached the cell surface inefficiently but were functionally indistinguishable from channels formed by the full-length subunits. In conclusion, 14-3-3 proteins promote the cell-surface transport of correctly assembled complexes but do not regulate the activity of KATP channels at the cell surface.
ATP-sensitive potassium (KATP) channels are metabolic sensors that couple the electrical excitability of many cell types to their metabolic status (Aguilar-Bryan and Bryan, 1999). They are hetero-octameric complexes formed by four inward rectifier K+ channel subunits (Kir6.1 or Kir6.2) and four sulfonylurea receptor subunits (SUR1 or SUR2). Both subunits are involved in the metabolic regulation of the channel, so it is essential that partially assembled KATP-channel complexes are excluded from the plasma membrane. Heteromultimeric assembly is controlled by a checkpoint mechanism based on endoplasmic reticulum (ER) localization motifs present in both Kir6.2 and SUR subunits (Zerangue et al., 1999). These arginine (Arg)-based peptide retention signals prevent unassembled subunits or partially assembled complexes from reaching the cell surface. By employing an affinity purification approach, we previously showed that the Arg-based ER-localization signal (RKR) present in the cytoplasmic C-terminus of Kir6.2 is recognized by the coat protein I (COPI) vesicle coat, as well as by 14-3-3 proteins (Yuan et al., 2003). COPI binding can explain the ER-localization activity of the RKR motif. Experiments with multimeric reporter proteins suggest that recruitment of 14-3-3 proteins is involved in the forward transport of KATP channel multimers to the cell surface (Yuan et al., 2003). It is therefore possible that 14-3-3 proteins are involved in inactivating the RKR motifs during the course of channel assembly.
All membrane proteins for which 14-3-3-dependent forward transport has been described to date also contain a COPI-interacting motif (Kuwana et al., 1998; Mrowiec and Schwappach, 2006; O'Kelly et al., 2002; Rajan et al., 2002; Shikano et al., 2005; Shikano et al., 2006; Yuan et al., 2003). Recent results from yeast imply that 14-3-3-dependent forward transport might depend on specific isoforms of 14-3-3 (Michelsen et al., 2006). For some complexes, e.g. the MHC class II complex and the two-pore-domain K+ channel KCNK3, recruitment of 14-3-3 required phosphorylation of a residue in the 14-3-3-binding motif (Anderson and Roche, 1998; Kuwana et al., 1998; O'Kelly et al., 2002; Rajan et al., 2002). Recruitment as a consequence of phosphorylation implies that 14-3-3-dependent forward transport will be regulated by activation of signal transduction cascades (Shikano et al., 2006). In other cases, such as Kir6.2, where 14-3-3 proteins are involved in monitoring the correct assembly of multimeric complexes, phosphorylation is not required.
In some proteins, such as KCNK3, the 14-3-3- and COPI-binding sites are physically distinct (O'Kelly et al., 2002). By contrast, the COPI-interacting motif and the 14-3-3-binding motif of Kir6.2 both lie in the C-terminus, and they physically overlap (Yuan et al., 2003). Although neither binding site has been fully mapped, functional studies indicate that the RKR motif contributes to both. However, although the Arg-based signal of Kir6.2 is necessary to recruit 14-3-3 to a multimeric reporter protein (Yuan et al., 2003), it is not known whether it is sufficient for 14-3-3 binding. The KATP-channel is unique in that it has COPI-interacting motifs in both subunits (Kir6.2 and SUR1) of the heteromultimer. Binding of 14-3-3 requires multiple copies of the Kir6.2 tail presented in close proximity to each other, which suggests that the spatial arrangement of the two peptide-binding grooves in the 14-3-3 dimer acts as a template that checks the assembly state of the membrane protein complex (Yuan et al., 2003).
During assembly of the KATP-channel complex, the individual Kir6.2 and SUR1 subunits transiently expose an Arg-based ER-localization signal. These signals are inactivated as a consequence of heteromultimeric assembly. Consequently, the assembly state and forward transport are coupled, and only properly assembled hetero-octamers reach the cell surface. Steric masking is a simple way of explaining this coupling in molecular terms: co-assembly might render the Arg-based signals present in Kir6.2 and SUR1 inaccessible to the COPI-vesicle coat. Unfortunately, currently available homology models of the assembled KATP channel do not provide structural insights into how the regions containing the Arg-based signal of either subunit are oriented with respect to the rest of the proteins in the complex. However, the finding that 14-3-3 proteins could be involved in the forward transport of KATP channels (Yuan et al., 2003) implies that steric masking by the other subunit is not sufficient to explain inactivation of all the Arg-based signals present in the complex.
Previous studies using soluble protein-A fusion proteins and reporter membrane proteins have shown that the C-terminal tail of Kir6.2 can recruit 14-3-3 proteins when the 14-3-3-binding motif is present in multiple copies that lie in close proximity (Yuan et al., 2003). However, the role of 14-3-3 in the native KATP-channel complex, composed of full-length subunits, remains unclear. In particular, it is not known whether 14-3-3 proteins are involved in KATP-channel transport to the cell surface, whether they functionally inactivate the Arg-based signals present in Kir6.2 and SUR1, and whether recruitment of 14-3-3 alters the functional properties of KATP channels at the cell surface. To explore these questions, we have investigated the role of 14-3-3 proteins in the assembly and functional activity of native KATP channels.
It is difficult to manipulate cellular 14-3-3 proteins in order to assess their involvement in a given process of interest. Common approaches include the use of dominant-negative mutants (either incapable of ligand binding or of dimerization), downregulation by RNA-interference, and overexpression. There are seven isoforms of 14-3-3 expressed at high levels in most mammalian cell types. Applying any of the approaches listed above requires knowledge on the specific isoform involved and none of the manipulations will alter the total pool of 14-3-3 proteins. We therefore developed a novel 14-3-3-scavenger protein, which enabled us to show that 14-3-3 proteins promote the cell-surface expression of the native KATP-channel complex. Furthermore, our data offer mechanistic insight into 14-3-3 action, because they suggest that 14-3-3 proteins function by counteracting the Arg-based ER-localization signal present in the SUR1 subunit of the channel complex.
A 14-3-3 scavenger interferes with 14-3-3-dependent surface expression of membrane proteins
To study the role of 14-3-3 proteins in the assembly and surface expression of KATP channels, we manipulated the availability of endogenous 14-3-3 proteins using an artificial 14-3-3-scavenger protein. This was constructed (Fig. 1A) by fusion of protein G, the tetramer-forming coiled-coil domain pLI (Harbury et al., 1993) and the R18 peptide that binds 14-3-3 proteins with high affinity (Wang et al., 1999). We refer to this fusion protein as pGpLI-R18. To assess the efficacy of pGpLI-R18 in scavenging 14-3-3 proteins, homogenates of Xenopus oocytes expressing pGpLI-R18 were incubated with IgG-Sepharose (Fig. 1B). This resulted in a cellular extract depleted of 14-3-3 proteins, and an eluate containing both pGpLI-R18 and 14-3-3 proteins. This indicates that a large portion of cellular 14-3-3 proteins are bound to the 14-3-3 scavenger: consequently, there will be less 14-3-3 available for binding to other interaction partners, particularly those that bind with lower affinity.
We next tested whether pGpLI-R18 was able to prevent surface expression of KCNK3, which is known to require 14-3-3 for cell-surface expression (O'Kelly et al., 2002; Rajan et al., 2002). The β2-adrenergic G-protein-coupled receptor and the inward rectifier K+ channel Kir2.1 were used as negative controls because there is no evidence that they require 14-3-3 proteins to reach the cell surface. Surface expression was assessed by antibody labeling of extracellular epitopes combined with luminometry, as described previously (Zerangue et al., 1999). Surface expression of the β2-adrenergic receptor and Kir2.1 were unaffected by coexpression of pGpLI-R18, whereas that of KCNK3 was significantly reduced (Fig. 1C). We therefore conclude that the 14-3-3 scavenger can be used as a tool to probe the involvement of 14-3-3 proteins in the surface expression of membrane proteins.
14-3-3 proteins play a role in the cell-surface expression of KATP channels
We used pGpLI-R18 as a tool to assess whether the surface expression of KATP channels is sensitive to 14-3-3 proteins in Xenopus oocytes. Coexpression of pGpLI-R18 with the two KATP-channel subunits, Kir6.2 and SUR1, reduced surface expression of recombinant KATP-channels by more than 50% (Fig. 2A). Similar results were found for native KATP channels in the INS-1 pancreatic β cell line (Fig. 2B). INS-1 cells were infected with a retroviral construct encoding pGpLI-R18, and surface expression assessed by recording the magnitude of endogenous KATP (Kir6.2/SUR1) currents, using the standard whole-cell configuration of the patch-clamp method. Currents were initially small but increased with time after establishment of the whole-cell configuration, as ATP dialysed out of the cell (Trube et al., 1986). The amplitude of the steady-state washout current, which provides a measure of the maximal available KATP channels, was significantly smaller in cells expressing the 14-3-3 scavenger (49% of control; Fig. 2B). Steady-state currents were blocked by the KATP-channel inhibitor tolbutamide, indicating that they flow through KATP channels. Thus, these data provide evidence that scavenging of 14-3-3 proteins leads to a decrease in the surface expression of native KATP channels.
The finding that 14-3-3 proteins promote transport of KATP channels to the cell surface predicts that 14-3-3 proteins should physically associate with KATP channels. To test this idea, INS-1 cells were infected with a retroviral vector encoding a variant of Kir6.2 containing an extracellular hemagglutinin (HA)-epitope tag (Zerangue et al., 1999), and phogrin fused to enhanced green fluorescent protein (EGFP) (Pouli et al., 1998). The HA-tag was used to facilitate immunoprecipitation of KATP channels in complex with cytosolic-interacting proteins. The membrane protein phogrin is a well-characterized marker of secretory granules in pancreatic β cells (Wasmeier and Hutton, 1996). Since most KATP channels appear to localize to the insulin secretory granules in pancreatic β cells (Geng et al., 2003; Varadi et al., 2006), we used phogrin as a control in our immunoprecipitation experiments. INS-1 cells were permeabilized, treated with the crosslinker disuccinimidyl suberate (DSS) and homogenized. Total membranes were solubilized and subjected to immunoprecipitation with anti-HA or anti-GFP antibody (Fig. 2C). The anti-HA immunoprecipitate was probed with an anti-Kir6.2 antiserum (top panel) and the anti-GFP immunoprecipitate with the anti-GFP serum (middle panel). Both immunoprecipitates were tested for the presence of 14-3-3 proteins (bottom panel). Coimmunoprecipitation of 14-3-3 proteins was detected for the KATP-channel complex but not for the phogrin-EGFP fusion protein (Fig. 2C). Thus, we conclude that 14-3-3 proteins are physically associated with KATP channels in INS-1 β-cells.
The Arg-based signals of KATP channels are crucial for 14-3-3 effects on forward transport
Several reports suggest that 14-3-3 proteins override ER localization mediated by Arg-based signals (Kuwana et al., 1998; Michelsen et al., 2006; Shikano and Li, 2003; Yuan et al., 2003). Thus, we next asked whether 14-3-3 scavenging only affected the cell-surface transport of KATP channels when Arg-based signals are present in the channel complex. For these experiments, we used a SUR1-Kir6.2 fusion protein, in which the C-terminus of SUR1 was linked to the N-terminus of Kir6.2 (Fig. 3). This was necessary because mutation of the Arg-based signal allows SUR1 and Kir6.2 to reach the cell surface independently of each other (Zerangue et al., 1999), which complicates interpretation of the experiments. Since they contain both partner subunits, when SUR1-Kir6.2 fusion proteins are heterologously expressed in Xenopus oocytes they reach the cell surface irrespective of the presence of an Arg-based signal (black bars in Fig. 3).
We tested the effect of pGpLI-R18 on the surface expression of KATP-channel complexes containing either wild-type (WT) or mutated Arg-based trafficking signals in Xenopus oocytes (grey bars in Fig. 3). KATP channels formed from the WT fusion protein were highly sensitive to scavenging of 14-3-3, surface expression being reduced to 30% of control (Fig. 3). By contrast, sequestration of 14-3-3 proteins by pGpLI-R18 had no effect on the surface expression of a SUR1-Kir6.2 fusion protein in which both Arg-based ER-localization signals were mutated (Fig. 3). This result is consistent with previous studies of KATP-reporter proteins (Yuan et al., 2003) and further demonstrates that 14-3-3 proteins are involved in inactivating Arg-based ER-localization signals present in KATP-channel complexes assembled from full-length subunits. It also indicates that the sole role of 14-3-3 proteins is to counteract the RKR signal directly or indirectly. They do not appear to have a positive effect on forward trafficking per se, because they have no effect on surface expression when all eight of the Arg-based ER-localization signals present in KATP-channel complexes are mutated.
When expressed alone, Kir6.2 and SUR1 are retained within the ER because of their Arg-based signals (Zerangue et al., 1999). In the absence of SUR1, Kir6.2 can still form tetramers, but these are retained inside the cell, indicating that co-assembly with other Kir6.2 subunits cannot mask the Arg-based signals. Only when SUR1 is present is Kir6.2 able to reach the membrane, which implies that inactivation of the Arg-based signal is a consequence of co-assembly with SUR1. The simplest explanation is that the RKR motif of Kir6.2 is sterically masked in the presence of SUR1 (Zerangue et al., 1999), thereby preventing COPI binding. However, this is difficult to reconcile with the fact that scavenging of 14-3-3 proteins prevents surface expression only when Arg-based signals are present in the KATP-channel complex (Fig. 3). This implies that, 14-3-3 binding contributes to forward transport and, thus, the RKR motif on either Kir6.2 or SUR1 must be accessible rather than sterically masked by the partner subunit.
Previous results using SUR1-Kir6.2 fusion proteins with a single mutated Arg-based signal suggested that the Arg-based motif in SUR1, but not that in Kir6.2, is partially exposed in the fully assembled octameric channel (Zerangue et al., 1999). To distinguish which of the two Arg-based signals (Kir6.2 or SUR1) is responsible for the observed 14-3-3-dependence, we used SUR1-Kir6.2 fusion proteins in which the Arg-based signal was mutated in either SUR1 or Kir6.2 (Fig. 3). Fig. 3 shows that mutation of the Arg-based signal in Kir6.2 alone has no effect on the ability of 14-3-3 scavenging to prevent surface expression. By contrast, when the RKR motif of SUR1 was mutated, the ability of pGpL1-R18 to reduce surface expression was abolished. This is consistent with the idea that the Arg-based signal in Kir6.2 is not accessible in the fully assembled KATP-channel complex (Zerangue et al., 1999) and argues that 14-3-3 proteins reduce the efficacy of the Arg-based ER-localization signal present on SUR1.
A simple explanation for this result could be that 14-3-3 binding to SUR1 sterically interferes with COPI binding. However, we were unable to consistently demonstrate direct binding of 14-3-3 proteins to the L1 loop of SUR1 in which the Arg-based signal is located (data not shown). The availability of 14-3-3 proteins clearly enhances the surface expression of the KATP-channel complex (Fig. 3), and we were able to detect binding of 14-3-3 to the KATP-channel complex (Fig. 2) but not the Kir6.2 tetramer in the absence of SUR1 (data not shown). In the context of the heteromeric complex, the Kir6.2-binding site for 14-3-3 must be distinct from the RKR motif because mutation of this motif does not affect the ability of the 14-3-3-scavenger pGpLI-R18 to prevent cell-surface expression (Fig. 3). This is not unreasonable, because some complexes between 14-3-3 proteins and ligand proteins have revealed an interplay between low-affinity and high-affinity sites as well as several low-affinity 14-3-3-binding sites (Giles et al., 2003; Wurtele et al., 2003).
14-3-3 binding to Kir6.2 is reduced when the distal C-terminus of Kir6.2 is truncated
We next tested this hypothesis by dissecting the roles of 14-3-3 and COPI in sorting of the KATP channel. To do this, we sought a mutation that abolishes 14-3-3 binding without disturbing the ER-localization activity of the Arg-based signal.
Binding of 14-3-3 to the C-terminal tail of Kir6.2 can only be detected in pull-down or coimmunoprecipitation assays when the tail is presented in multiple copies in close proximity (Michelsen et al., 2006; Yuan et al., 2003). For binding assays, we therefore employed tetrameric-affinity constructs consisting of protein A and the coiled-coil domain of pLI fused to the last 36 amino acid residues of Kir6.2, followed by a hexahistidine tag for purification of the full-length bait protein (Yuan et al., 2003). We then truncated the last ten (RKR ΔC10), or 16 (RKR ΔC16), residues of Kir6.2, which lie downstream of the Arg-based signal (Fig. 4A). The corresponding proteins were purified from E. coli, immobilized, and incubated with HeLa cytosol to pull-down 14-3-3 proteins (Fig. 4B). We probed the precipitates with an antibody recognizing all seven isoforms of 14-3-3 (anti-β antibody). Interestingly, both truncations reduced 14-3-3 binding to the same extent as mutation of the Arg-based signal in full-length Kir6.2 to KKK, despite the fact that neither truncation deleted the RKR motif (Fig. 4C). Mutation of the RKR signal to AAA abolished 14-3-3 binding to the WT and truncated variants of the distal Kir6.2 C-terminus. The data therefore suggest that C-terminal residues other than the RKR sequence contribute to 14-3-3 binding by Kir6.2. Since the last ten residues of Kir6.2 contain no known protein-interaction motif, we tested the sequence specificity of this effect by mutating the sequence KFSISPDSLS to KFSISADSLS (RKR_P-A), KFSISGDSGS (RKR_PL-GG), or KFAIAPDALA (RKR_S-A) (residues deviating from the WT sequence are indicated in bold). All three sequence variants show 14-3-3 binding very similar to the WT sequence (Fig. 4C), indicating that neither the Pro residue nor the Ser residues in this sequence were specifically required for 14-3-3 recruitment. Nevertheless, deleting these residues might decrease 14-3-3 recruitment without affecting the efficacy of the Arg-based signal as an ER-localization signal.
Reporter proteins distinguish between the effects of 14-3-3 and COPI binding
The Arg-based signal was present in both truncated Kir6.2 proteins; therefore, we first compared their ability to prevent a monomeric reporter protein (CD4) from reaching the plasma membrane. Since 14-3-3 proteins show no high-affinity interaction with monomeric Kir6.2 tails (Yuan et al., 2003), this provides an assay of COPI activity alone. We also tested whether the truncated Kir6.2 tails could support the forward transport of dimeric CD8-RKR and tetrameric CD4pLI-RKR to the cell surface (Fig. 5C,D) to assess 14-3-3-dependent forward transport.
As previously described (Yuan et al., 2003; Zerangue et al., 1999), monomeric CD4-RKR was unable to reach the plasma membrane because of the ER-localization activity of the Arg-based signal (Fig. 5B). By contrast, CD8-RKR and CD4pLI-RKR reached the cell surface efficiently (Fig. 5C,D). Mutation of the RKR motif resulted in marked surface expression of all proteins (Fig. 5B-D). These results are consistent with the earlier interpretation that 14-3-3 proteins interact with multimeric Kir6.2-reporter proteins and outcompete the activity of COPI (Michelsen et al., 2006; Yuan et al., 2003).
Deleting the last ten amino acids of the Kir6.2 tail had a minor effect on surface expression of the monomeric CD4-RKR ΔC10 (Fig. 5B) but strongly reduced that of dimeric CD8-RKR ΔC10 and tetrameric CD4pLI-RKR ΔC10 (Fig. 5C,D), which suggests that the ability of 14-3-3 proteins to enhance forward trafficking of multimeric Kir6.2 reporters is dependent on the last ten amino acids of the protein. This effect is not sequence-specific because all three mutated C-termini (RKR_P-A, RKR_PL-GG, and RKR_S-A) supported forward transport of CD4pLI-RKR to the same extent as the WT sequence (data not shown). It also argues that COPI binding is unaffected, because intracellular retention of monomeric CD4-RKR ΔC10 was unaffected. Thus deletion of the last ten amino acids of Kir6.2 appears to uncouple the effect of COPI binding (retention from the cell surface because of COPI-mediated ER retrieval) from the effect of 14-3-3 binding (expression at the cell surface).
Truncation of a further six residues (RKR ΔC16) resulted in significantly higher expression of the monomeric CD4 fusion protein at the cell surface (Fig. 5B), indicating it impairs COPI binding. Very high levels of the corresponding dimer and tetramer were observed at the cell surface (Fig. 5C,D), probably because of a combination of impaired recognition by COPI and reduced 14-3-3 binding. To confirm the correlation between 14-3-3 binding and cell-surface expression for the CD4pLI membrane proteins, we tested their ability to coimmunoprecipitate 14-3-3 proteins (Fig. 6). This was important because the fusion proteins employed in the in vitro binding assay (Fig. 4) contained a hexahistidine tag at the distal C-terminus to allow purification of the full-length proteins, whereas the CD4pLI reporter proteins expressed in mammalian cells lacked the six histidines. As described previously (Yuan et al., 2003), mild crosslinking (that converted only a small fraction of 14-3-3 proteins into crosslinked dimers, Fig. 6A) was required to observe coimmunoprecipitation with CD4pLI-RKR (Fig. 6B). Under these conditions CD4pLI-AAA, CD4pLI-dC10 and CD4pLI-ΔC16 coimmunoprecipitated much less 14-3-3 protein. This result further supports the hypothesis that recruitment of 14-3-3 proteins can override the activity of the Arg-based signal when the distal C-terminus of Kir6.2 is present in multiple copies. We conclude that the RKR ΔC10 variant, but not the RKR ΔC16 variant, can be used to study the effect of 14-3-3 binding to the C-terminal tail of Kir6.2 independently of mutating the Arg-based signal. This enables us to assess the consequences of reducing 14-3-3 binding to Kir6.2 in KATP channels assembled from full-length subunits without disturbing the efficacy of the Arg-based signal as an ER retention and/or retrieval signal.
14-3-3 binding to Kir6.2 antagonizes the Arg-based signal present on SUR1
Introducing the truncated tail (ΔC10) into an SUR1-Kir6.2 fusion protein, while leaving the Arg-based signals of both subunits intact, reduced surface expression of the KATP-channel complex (Fig. 7A) to approximately the same extent as 14-3-3 scavenging (Fig. 3). This is consistent with the idea that recruitment of 14-3-3 proteins through the tail of Kir6.2 is impaired by the truncation. Mutation of the Arg-based signal of SUR1 in the truncated fusion protein restored cell-surface expression (Fig. 7A). Thus, we conclude that 14-3-3 binding to Kir6.2 antagonizes the effect of the Arg-based signal present in SUR1.
Effects on channel activity
It is not known whether 14-3-3 proteins accompany KCNK3, Kir2.1-RKR-SWTY, or KATP channels to the cell surface (O'Kelly et al., 2002; Rajan et al., 2002; Shikano et al., 2005; Yuan et al., 2003). The fact that significant amounts of all KATP fusion proteins reached the plasma membrane allowed us to test whether impaired 14-3-3 recruitment would alter the functional properties of the KATP channel. Thus, we compared the ATP-sensitivity of the WT and ΔC10 fusion KATP channels in excised inside-out patches (Fig. 7B). Fusion of SUR1 to Kir6.2 produced a reduction in the ATP-sensitivity of the WT channel, as previously described (Shyng and Nichols, 1997): in the absence of Mg2+, the IC50 was 115±29 μM (n=5), compared with ∼10 μM for channels produced when Kir6.2 and SUR1 are expressed separately (Proks et al., 2004). Importantly, however, the ΔC10 deletion did not affect the ATP-sensitivity of the channel, either in the presence or absence of Mg2+ (Fig. 6B). In the absence of Mg2+, IC50 values were 115±29 μM (n=5) and 156±46 μM (n=5) for SUR1-Kir6.2 and SUR1-Kir6.2 ΔC10, respectively; in the presence of Mg2+, IC50 values were 576±230 μM (n=6) and 491±133 μM (n=6) for SUR1-Kir6.2 and SUR1-Kir6.2 ΔC10, respectively. Thus, 14-3-3 binding to KATP channels does not alter KATP-channel function.
A 14-3-3 scavenger as a strategy to reduce cellular 14-3-3 availability
Because of their high abundance, their diverse functions, and the presence of seven closely related isoforms it is a daunting task to test the role of 14-3-3 proteins in the cellular context. Since it is not known whether specific mammalian isoforms of 14-3-3 are involved in the forward transport of membrane proteins, we devised a novel way to reduce the availability of all 14-3-3 isoforms simultaneously: we designed an artificial 14-3-3 scavenger, and used it to investigate the role of 14-3-3 proteins in the cell-surface transport of membrane proteins in Xenopus oocytes and in the INS-1 β-cell line.
Control experiments showed that titrating out 14-3-3 proteins does not cause a general impairment of the surface expression of membrane proteins (Fig. 1C). However, it reduced the cell-surface expression of KCNK3, which is known to reach the cell surface with very low efficiency unless 14-3-3 proteins can bind to its C-terminus (O'Kelly et al., 2002; Rajan et al., 2002). The fact that cell-surface expression was not abolished completely could be due to a leak of KCNK3 not in complex with 14-3-3, but presumably means that a significant population of KCNK3 proteins was still able to recruit 14-3-3, despite the presence of the scavenger. This might be attributed to the fact that binding of 14-3-3 to the C-terminus of KCNK3 must occur with rather high-affinity (Coblitz et al., 2005; Shikano et al., 2005).
14-3-3 interaction with Kir6.2
Experiments employing the 14-3-3 scavenger not only revealed that 14-3-3 proteins facilitate the surface expression of the KATP channel, but also imply that this role is mechanistically related to the inactivation of the Arg-based signals present in the complex (Fig. 8). There are eight of these signals in the fully assembled complex – four contributed by Kir6.2 and four by SUR1. It is important to bear in mind that although Kir6.2 can assemble into functional tetramers in the absence of SUR (Tucker et al., 1997), these tetramers do not reach the cell surface unless the Arg-based signal is mutated (Zerangue et al., 1999). We have been unable to coimmunoprecipitate 14-3-3 proteins with the Kir6.2 tetramer under any conditions (H.Y. and B.S., unpublished results). Similarly, the tail of Kir6.2 does not recruit 14-3-3 proteins to the homologous tetrameric Kir2.1 channel in the Kir2.1-RKR-reporter fusion (Shikano et al., 2005). However, 14-3-3 does interact with the tetrameric CD4pLI-RKR fusion protein (Fig. 6) (Yuan et al., 2003) and a very similar reporter membrane protein in yeast (Michelsen et al., 2006), presumably because the Kir6.2 tails are tethered in closer proximity than in a Kir6.2 tetramer lacking SUR1. Furthermore, we found that 14-3-3 proteins are recruited to the full KATP-channel complex (Fig. 2). Thus, we have to assume that the tails of Kir6.2 are brought into closer proximity to each other upon co-assembly with SUR1, or that SUR1 contributes to 14-3-3 recruitment by adding an additional low-affinity binding site (Fig. 8). This additional binding site could reside in the vicinity of the RKR signal of SUR1 or in a completely different region of the protein, but would have to be within 30 Å of the distal tail of Kir6.2 in the context of the assembled complex because this is the distance between the two ligand-binding grooves in the 14-3-3 dimer.
14-3-3 and SUR1
Our data argue that it is not the Arg-based signal present in Kir6.2 but that present in SUR1 that requires 14-3-3 recruitment in order to be inactivated and thus relieve COPI-mediated retention (Fig. 8). First, inactivation of the Arg-based signal present in SUR1 rendered the whole complex insensitive to 14-3-3 scavenging, whereas a fusion protein in which only the Arg-based signal in Kir6.2 was mutated remained sensitive to 14-3-3 scavenging (Fig. 3). Second, a truncation of the Kir6.2 tail that decreased its ability to recruit 14-3-3 in the context of reporter proteins (Figs 4, 6) only resulted in reduced surface expression of the KATP-channel complex when the Arg-based signal in SUR1 was present (Fig. 7). Therefore, steric masking by the partner subunit seems to apply to the Arg-based signal of Kir6.2, whereas 14-3-3 proteins appear to be required to inactivate the Arg-based signal of the SUR subunit (Fig. 8). This is entirely consistent with the earlier conclusion that the RKR motif in SUR1, but not in Kir6.2, is partially exposed in the fully assembled octamer (Zerangue et al., 1999).
We have not been able to obtain evidence for a 14-3-3-binding site on SUR1 but cannot exclude the possibility of a low-affinity site that contributes to 14-3-3 binding once 14-3-3 has been recruited to the complex through the tail of Kir6.2. Based on our results, it seems unlikely that 14-3-3 proteins have a scaffolding role in bringing or keeping the different subunits in close proximity to each other during primary assembly in the ER. Otherwise, surface expression of the independently expressed subunits would be expected to be more sensitive to 14-3-3 scavenging than surface expression of the physically linked subunits. Instead, 14-3-3 binding to Kir6.2 seems to impinge on the forward transport of the KATP-channel complex either by directly shielding the Arg-based signal of SUR1 from COPI (steric masking by 14-3-3) or by indirect activation, which could be because of the recruitment of additional proteins to the complex (Fig. 8). These could be proteins that affect the functional state of the COPI coat or that couple the protein to the forward-transport machinery, such as components of the COPII coat (Mrowiec and Schwappach, 2006; Shikano et al., 2006).
Effects on KATP-channel activity
It is not clear whether 14-3-3 proteins remain associated with KATP channels at the plasma membrane. Our results do, however, suggest that even if they remain associated, 14-3-3 proteins do not alter the ATP-sensitivity of KATP channels at the cell surface. A fusion protein of SUR1 and a truncated Kir6.2 subunit gave rise to KATP channels with an ATP-sensitivity indistinguishable from that of the corresponding WT channel. This was true both in the presence and absence of Mg2+. In the absence of Mg2+, ATP interacts only with the Kir6.2 subunit. In the presence of Mg2+, however, ATP is also hydrolysed by the nucleotide-binding domains of SUR1, resulting in a stimulation of channel activity that produces an apparent reduction in ATP-sensitivity. Our data therefore indicate that neither the ATP-binding site on Kir6.2, nor the ability of MgATP binding and/or hydrolysis at SUR1 to influence channel activity, are affected by the ΔC10 truncation. This is consistent with the observation that truncations of up to 36 amino acids from the C-terminus of Kir6.2 do not influence KATP-channel function when Kir6.2 and SUR1 subunits are coexpressed (Tucker et al., 1997).
In conclusion, our data support the following model for trafficking of KATP channels to the surface membrane (Fig. 8). First, both Kir6.2 and SUR1 possess RKR motifs that bind COPI, thereby producing ER localization and preventing trafficking to the cell surface. This ensures that individual subunits, Kir6.2 tetramers or partially assembled complexes are prevented from reaching the plasma membrane. Second, the RKR motif on Kir6.2 is masked by the presence of SUR, preventing COPI binding. This is consistent with recent structural studies that show that the Kir6.2 tetramer is surrounded by SUR1 subunits in the channel complex (Mikhailov et al., 2005). Third, the RKR motif on SUR1 is inactivated directly (steric masking) or indirectly, by 14-3-3 initially recruited through the C-terminal tail of Kir6.2 and stabilized on the KATP channels by a combination of low-affinity binding sites. This model thus provides a mechanistic explanation for how both SUR1 and Kir6.2 partner subunits contribute to surface expression of KATP channels.
Finally, our results provide the first example of 14-3-3-dependent cell-surface expression in the context of a native, heteromultimeric membrane protein complex. They further dissect the signal-inactivation events that allow the forward transport of fully assembled KATP channels to the plasma membrane and thus contribute to our understanding of cellular quality-control mechanisms.
Materials and Methods
For oocyte expression, constructs were in pGemHE (Liman et al., 1992). For retroviral gene transduction, constructs were engineered in the pBI-GFP vector (Liu et al., 2000) for Kir6.2-11HA (Zerangue et al., 1999) and in pBICD4 (Liu et al., 2000) for phogrin-EGFP (Pouli et al., 1998). The pGpLI-R18 fusion is analogous to the previously described pApLI-RKR fusion (Yuan et al., 2003). For KCNK3, an extracellular protein C (PC)-epitope tag (EDQVDPRLIDGK) was inserted at amino acid position 213, as described (O'Kelly et al., 2002). The SUR1-Kir6.2 fusion construct was made by inserting the following linker between the C-terminus of SUR1 and the N-terminus of Kir6.2: AAAGGGGGGGSA. All other constructs used to assess surface expression in Xenopus oocytes have been described previously (Zerangue et al., 1999). Fusion constructs of the last 36 amino acids of Kir6.2 to CD4, CD8, or CD4pLI, and the respective truncated variants (RKR ΔC10 and RKR ΔC16) were in pcDNA3. All protein-A fusion constructs used for recombinant protein expression in E. coli were in pQE60 (Yuan et al., 2003) and consist of two copies of the IgG-binding domain of protein A, the coiled-coil-forming domain pLI (Harbury et al., 1993) and the C-terminal residues from Kir6.2 as described in the text and Figure legends. A C-terminal hexahistidine tag ensured the purification of full-length proteins. For expression of SUR1-Kir6.2 fusion constructs (Zerangue et al., 1999) in mammalian cells, the cDNAs were engineered in pcDNA3.
Preparation and surface labeling of oocytes
Xenopus oocytes were prepared as described (Collins et al., 1997). Oocytes were injected with 0.1 ng cRNA for Kir6.2-11HA, 2 ng SUR1, 5 ng KCNK3-PC, 5 ng SUR1-Kir 6.2-11HA fusion constructs, 5 ng Kir 2.1-HA, 5 ng HA-β2 adrenergic receptor and 5 ng pGpLI-R18. Surface assays were performed 2-3 days after injection as described (Zerangue et al., 1999). Oocytes were homogenized in PBS (50 mM phosphate, 150 mM NaCl, pH 8.0) supplemented with a protease inhibitor cocktail (complete; Roche) by five passages through an 18-gauge and one passage through a 27-gauge needle (Tucker et al., 1996). Homogenates were cleared by repeated centrifugation at 1000 g. The supernatants were supplemented with SDS-PAGE-loading buffer and subjected to gel electrophoresis.
INS-1 cell culture and retroviral gene transduction
INS-1 cells (Asfari et al., 1992) were cultured in RPMI 1640 medium containing L-glutamine (Gibco) supplemented with 10% fetal bovine serum, 50 μM 2-mercaptoethanol, 100 iu/ml penicillin, and 100 μg/ml streptomycin, at 37°C, 5% CO2, absolute humidity. INS-1 cells were infected by incubation with retroviral particles produced by co-transfection of HEK-293T cells with pVPack-Eco and pVPack-GP (Stratagene) with either pBI-Kir6.2-11HA-IRES-GFP, pBI-phogrin-EGFP-IRES-CD4, or pBI pGpLI-R18-IRES-CD4. Populations of infected cells were purified by fluorescence-activated cell sorting. Expression of the epitope-tagged variant of Kir6.2 was low compared with the amount of endogenous channels expressed (Fig. 2C).
INS-1 cells expressing 14-3-3-scavenger protein were identified using phycoerythrin-conjugated anti-CD4 monoclonal antibody (EDU-2, Diatec). Whole-cell currents were recorded from CD4+ cells using the standard whole-cell configuration of the patch-clamp method. Currents were elicited by ±10 mV voltage steps from a holding potential of –70 mV. Currents were normalized to cell capacitance to correct for differences in cell size. The pipette was filled with intracellular solution (in mM): 140 KCl, 2 MgCl2, 1 CaCl2, 10 EGTA, 0.3 ATP, 10 HEPES (pH 7.2 with KOH). The bath contained extracellular solution (in mM): 137 NaCl, 5.6 KCl, 2.6 CaCl2, 1.1 MgCl2, 10 HEPES (pH 7.4 with NaOH). Tolbutamide (0.5 mM) was added as indicated. All experiments were performed at 21-23°C. Data are expressed as the mean ± s.e.m.
Coimmunoprecipitation and binding assay
Permeabilization of retrovirally transduced INS-1 cells and transfected HEK293T cells was performed according to Jadot et al. (Jadot et al., 1995). The crosslinker DSS was then added to a final concentration of 200 μM from a 10 mM stock in dimethyl sulfoxide (DMSO) and the reaction allowed to proceed for 15 minutes at 25°C. Cells were homogenized in K-Hops buffer (130 mM KCl, 25 mM Tris-HCl; pH 7.5) by sequential repeated passage through 22- and 27-gauge needles. After clearing of the lysate, total membranes were prepared by centrifugation. Membranes were solubilized in lysis buffer (50 mM Tris/HCl, pH 8.5, 100 mM NaCl, 1% Triton X-100, 0.1% SDS) and up to 5 μg of the respective antibody was added overnight. Antibody–antigen complexes were immobilized on protein G-Sepharose (Roche), washed, and eluted in SDS sample buffer. Recombinant protein expression, purification through the C-terminal hexahistidine tag and the binding assay were performed as described (Yuan et al., 2003).
Western blot analysis
Blots were blocked in Tris-buffered saline (TBS) containing 5% milk powder and 0.02% NP-40. Primary antibodies were: anti-Kir6.2 guinea pig polyclonal, 1:2000; anti-GFP rabbit polyclonal, 1:1000; anti-CD4 rabbit polyclonal (H-370, Santa Cruz Biotechnology), 1:1000; anti-CD8 (H-160, Santa Cruz Biotechnology), 1:1000; anti-14-3-3 β mouse monoclonal (H-8, Santa Cruz Biotechnology), 1 μg/ml, anti-14-3-3 β rabbit polyclonal (K-19 and FL-246, Santa Cruz Biotechnology); 80 ng/ml for detection of coimmunoprecipitated 14-3-3 in Fig. 6. Secondary antibodies were: horseradish peroxidase (HRP)-conjugated anti-mouse, anti-guinea pig and anti-rabbit (Jackson ImmunoResearch Laboratories), at 1:4000 and 1:50,000 when using SuperSignal West Femto Maximum Sensitivity Substrate (Pierce Biotechnology) diluted in TBS-blocking solution. Washes were in TBS-blocking solution and then in TBS, 0.02% NP-40. Detection was performed using the ECL system (Amersham) and with SuperSignal West Femto Maximum Sensitivity Substrate (Pierce) for the 14-3-3 coimmunoprecipitation shown in Fig. 6.
Luminometric surface-expression assay for mammalian cells
COS1 cells plated in 35-mm tissue-culture dishes were transfected and surface expression of all constructs was assayed 24 hours after transfection. Cells were fixed with 4% formaldehyde in PBS (20 minutes), blocked in PBS with 1% fetal bovine serum (30 minutes), and then labeled with primary antibody for 1 hour, and with an appropriate HRP-coupled secondary antibody for 20 minutes. Chemiluminescence of the whole 35-mm dish was quantitated in a TD-20/20 luminometer (Turner Designs) after 15 seconds of incubation in SuperSignal ELISA Femto Maximum Sensitivity Substrate (Pierce). Extensive washing was performed between steps and before chemiluminescence measurement; all steps were performed at room temperature. Surface expression of CD4- or CD8-fusion proteins was detected by mouse monoclonal anti-CD4 antibody EDU-2 (Dianova) diluted 1:5000, or mouse monoclonal anti-CD8 antibody UCHT-4 (Dianova) diluted 1:5000 and goat anti-mouse HRP-conjugated IgG (Jackson) diluted 1:2000. The cell-surface expression of SUR1-Kir6.2 fusion constructs was detected by monoclonal mouse anti-HA antibody (clone 16B12; BAbCO) using an HA-epitope engineered into the Kir6.2 portion of the fusion protein (Kir6.2-11HA) (Zerangue et al., 1999). For each construct, surface expression was assayed in three 35-mm dishes, and each experiment was repeated several times (see figure legends for specific details), with nearly identical results when normalized to a given construct.
Data are reported as the mean ± s.e.m. For the cell-surface expression assay in oocytes the values obtained for a given construct in the absence of the scavenger pGpLI-R18 were averaged in each batch and the value obtained for each oocyte in this batch was divided by the mean to normalize the data. Bar diagrams represent the mean values of these normalized values across all batches (see figure legends for pertinent numbers), and error bars depict the corresponding s.e.m. Statistical significance was determined using Student's t-test. The significance values are given in the relevant figure legend.
We thank Patrik Rorsman for the INS-1 cell line, Guy Rutter for the cDNA encoding phogrin-EGFP, and Walter Nickel for the anti-GFP antibody. We acknowledge Martina Götzmann and Jutta Metz for expert technical assistance as well as Kai Michelsen for valuable discussions and comments on the manuscript. We are indebted to Thomas Ruppert from the ZMBH Biomolecular Facility for mass spectrometry. This work was supported by grants from the DFG (SFB638) and European Union (BioSim, EuroDia) and a Royal Society Joint Project award. I.N. was the recipient of a DAAD fellowship. B.S. is an EMBO Young Investigator and F.M.A. is a Royal Society Research Professor.