The Functional Interaction of 14-3-3 Proteins with the ERK1/2 Scaffold KSR1 Occurs in an Isoform-specific Manner*

Identifying 14-3-3 isoform-specific substrates and functions may be of broad relevance to cell signaling research because of the key role played by this family of proteins in many vital processes. A multitude of ligands have been identified, but the extent to which they are isoform-specific is a matter of debate. Herein we demonstrate, both in vitro and in vivo, a specific, functionally relevant interaction of human 14-3-3γ with the molecular scaffold KSR1, which is mediated by the C-terminal stretch of 14-3-3γ. Specific binding to 14-3-3γ protected KSR1 from epidermal growth factor-induced dephosphorylation and impaired its ability to activate ERK2 and facilitate Ras signaling in Xenopus oocytes. Furthermore, RNA interference-mediated inhibition of 14-3-3γ resulted in the accumulation of KSR1 in the plasma membrane, all in accordance with 14-3-3γ being the cytosolic anchor that keeps KSR1 inactive. We also provide evidence that KSR1-bound 14-3-3γ heterodimerized preferentially with selected isoforms and that KSR1 bound monomeric 14-3-3γ. In sum, we have demonstrated ligand discrimination among 14-3-3 isoforms and shed light on molecular mechanisms of 14-3-3 functional specificity and KSR1 regulation.

The 14-3-3 proteins comprise a large family of highly conserved, acidic polypeptides of 28 -33 kDa that are expressed ubiquitously in all eukaryotic species (1). Seven isoforms, each encoded by a distinct gene, have been described in mammals: ␤, ␥, ⑀, , , /, and . In plants, up to 15 isoforms have been identified, and in yeast, Drosophila melanogaster and Caenorhabditis elegans, only two isoforms have been reported (1). Initially described in 1967 (2), these proteins were characterized a decade ago as the first distinct phosphoserine (pSer) 6 -binding proteins (3). Since then, a varied multitude of interacting partners have been identified, participating in cellular processes as diverse as signal transduction, cell-cycle control, apoptosis, regulation of metabolism, protein trafficking, cell morphology, transcription, stress response, and oncogenic transformation (1,4), thereby highlighting 14-3-3 proteins as key mediators of intracellular signaling. Large scale analyses aimed at identifying potential 14-3-3 ligands have consistently resulted in long lists of proteins. Two laboratories independently have identified more than 200 interacting partners using and in vitro affinity chromatography protocol (5,6), and a recent direct proteomic analysis has identified as many as 170 specific 14-3-3-interacting proteins (7). Further, transgenic mouse proteomics allowed the identification of 147 brain proteins interacting with 14-3-3 (8). With the seven mammalian isoforms sharing a 70% identity, the question arises as to how they achieve specificity in regulating hundreds of different proteins.
On a structural level, 14-3-3 proteins form U-shaped dimers, each monomer containing nine anti-parallel ␣-helices, named A to I (9). Helices A to D are involved in dimer formation, and helices C, E, G, and I form a large amphipathic groove critically involved in binding to pSer/Thr-containing proteins (9). Screening of phosphopeptide libraries and structural analysis of 14-3-3/phosphopeptide complexes have identified two highaffinity binding motifs: RSXpSXP (mode 1) and RXXXpSXP (mode 2) (10). In addition, exoenzyme S is able to interact with 14-3-3 in a nonphosphorylated form (11), and recently a C-terminal mode 3-binding motif has been described in some 14-3-3 ligands having a general consensus of p(S/T)X 1-2 -COOH (12). Despite these exceptions, the majority of ligands bind to 14-3-3 proteins through the unique mode 1 or mode 2 sequence. This, in addition to extensive sequence conservation among 14-3-3 proteins, makes it difficult to uncover any specific role for each isoform and to understand the molecular determinants con-
Cell Culture-293 and COS-7 cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (Hyclone) and antibiotics. U2OS cells, kindly provided by Dr. M. A. Medina (Universidad de Málaga, Spain), were maintained in McCoy's 5A medium supplemented with 10% fetal bovine serum and antibiotics. Transfections were performed with the FuGENE 6 reagent (Roche Applied Science). Xenopus laevis oocytes were maintained in modified Barth saline (mBarth) medium as described (22).
GST Pulldown Assays-GST-14-3-3 proteins were expressed in Escherichia coli DH5␣ cells essentially as described (25). Lysates from 293 cells expressing FLAG-KSR1 (250 g) were incubated with recombinant GST-14-3-3 for 30 min (4°C). Complexes were recovered by adding 30 l of 50% (vol/vol) glutathione-Sepharose resin (GE Healthcare) for an additional 30 min (4°C). After three washes in NP-40 buffer, the amount of FLAG-KSR1 bound to GST-14-3-3 isoforms was determined by Western blot (WB) with a FLAG antibody and quantitated as the percent of input material by loading 12.5 g of the lysate in each gel. For the competition assays, 150 M of the phospho-or non-phosphopeptides were incubated with 1 g of GST-14-3-3 for 30 min at 4°C before the addition of 150 g of FLAG-KSR1 lysates. Endogenous B-KSR1 was pulled down by incubating 10 g of each GST-14-3-3 isoform with 1.5 mg of mouse brain lysates prepared as described (20).
Pulldown/IP Depletion Assay-A FLAG-KSR1-expressing cell lysate (1.5 mg) was subjected to pulldown with 180 l of 50% (vol/vol) GST-14-3-3␥-prebound beads (10 g protein/30 l resin). One-third of the supernatant (ϳ500 g), containing unbound proteins, was immunoprecipitated with 30 l of an anti-FLAG resin, and the remaining (ϳ1000 g) was again pulled down with 120 l of GST-14-3-3␥ beads. After that, half of the supernatant (ϳ500 g) was subjected to a second round of immunoprecipitation with 30 l of FLAG resin. The activation state of FLAG-KSR1 was determined with an anti-pKSR1 antibody. Endogenous 14-3-3 still associated with FLAG-KSR1 after each pulldown was detected in the immunoprecipitates with a pan-14-3-3 antibody.
ELISA Format Binding Assay-We followed a protocol similar to a published assay (26). Briefly, different amounts of GST fusion proteins were diluted in coating buffer (PBS, 1 mM phenylmethylsulfonyl fluoride, and 4 mM dithiothreitol) and allowed to bind to the wells of a microtiter plate by overnight incubation at 4°C. Nonoccupied sites were blocked with blocking buffer (PBS, 0.3% bovine serum albumin, 0.1% Tween 20) for 1 h at 4°C. Lysates from 293 FLAG-KSR1-expresing cells were diluted in blocking buffer at a final concentration of 0.5 g/l and added to the wells (50 g/well) loaded with the GST-14-3-3 proteins. After 1 h (4°C), the wells were washed three times with PBS. A horseradish peroxidase-conjugated FLAG antibody diluted 1:3000 in PBST (PBS plus 0.2% Tween 20) was then added to the wells and incubated for 30 min at room temperature. Wells were washed twice with PBS and incubated with 200 l of solution D (0.04% o-phenylenediamine, 8 mM citric acid, 17 mM Na 2 HPO 4 , 0.012% H 2 O 2 ) at room temperature until color developed (usually 15 min). The reaction was stopped by adding 50 l/well 1 N H 2 SO 4 and the amount of FLAG-KSR1 recovered on each well was determined by measuring absorbance at 492 nm in a microplate reader. Data were normalized to values obtained from wells containing GST alone.
Oocyte Meiotic Maturation Assay-X. laevis oocytes were prepared essentially as described (22). Plasmids pFTX5-KSR1 and pFTX4-14-3-3 (2-3 g) were linearized and used as tem-plates for capped mRNA synthesis with the mMESSAGE mMACHINE in vitro transcription kit (Ambion Inc.). To avoid inhibition of Ras-induced maturation due to KSR1 scaffolding activity (17), the amount of injected KSR1 mRNA was titrated previously. Oocytes were microinjected with 50 nl of the in vitro transcribed mRNAs (2.5 ng for Myc-KSR1 and 35 ng for HA-14-3-3) and maintained at 18°C in mBarth medium. Five hours later the oocytes were reinjected with 50 ng of a mRNA encoding Myc-H-RasG12K and scored for germinal vesicle breakdown (GVBD) as a measure of meiotic maturation. GVBD was scored when 5-10% of the oocytes injected with Ras alone had undergone GVBD (usually 6 -8 h after injection) (17).
RNA Interference-Synthetic siRNAs specific for human 14-3-3␥ (NM_012479) or 14-3-3 (NM_006826) were purchased from Ambion Inc. as 19-mer complementary RNA duplexes with UU overhangs at their 3Ј-ends. A scrambled sequence with no homology in the human genomic data base was used as a negative control. U2OS cells were seeded in 6-well plates and transfected 24 h later (day 1) with the corresponding siRNAs (20 nM) by the calcium phosphate method. The cells were retransfected 24 h later (day 2) following the same protocol, and on day 3, the cells were transfected with 2.0 g of pCMV-FLAG-KSR1 using the FuGENE 6 reagent. On day 5, the cells were processed for Western blot and immunofluorescence.
Immunofluorescence-Transfected cells grown on coverslips in 6-well plates were formalin-fixed, permeabilized with PBS plus 0.2% Triton X-100, and stained either singly or doubly with a mouse monoclonal FLAG antibody and a rabbit polyclonal Myc antibody followed by an Alexa 488-conjugated goat antimouse antibody and a rhodamine-conjugated goat anti-rabbit antibody.
Isolation of 14-3-3 Dimers Bound to KSR1-Total cell lysates (3.5 mg) prepared from 293 cells coexpressing FLAG-KSR1 and Myc-14-3-3␥ were immunoprecipitated with anti-FLAG beads as described above and washed sequentially with NP-40 buffer (five times) and Tris-buffered saline (twice). Bound FLAG-KSR1 was eluted twice by incubation at 15°C for 15 min in 35 l of 0.1 mg/ml FLAG peptide (Sigma-Aldrich) diluted in Trisbuffered saline. The combined eluates were diluted 1:2 in NP-40 buffer. One-half was immunoprecipitated with 5.6 g of a Myc polyclonal antibody, and the other half was immunoprecipitated with 5.6 g of a non-immune rabbit IgG for 2 h at 4°C followed by a 1-h incubation with 30 l of protein A-Sepharose. Immunoprecipitates were washed three times with NP-40 buffer, resuspended in 70 l of Laemmli's sample buffer, and fractionated by SDS-PAGE followed by Western blot with specific anti-14-3-3 antibodies. A 10 l-aliquot was loaded per well. As a control, we also immunoprecipitated sequentially, with FLAG and Myc antibodies, a 293 lysate expressing FLAG-KSR1 alone.
Molecular Modeling-Interactions of KSR1 phosphopeptides with residues in the 14-3-3␥ basic pocket were modeled with the program Mutmodel (28) using the crystal structure of human 14-3-3␥ bound to mode 1 phosphopeptide RAIpSLP (Protein Data Bank ID: 2B05) as template. Water accessibility for 14-3-3␥ protein was calculated with the DSSP program (29). The DALI program (30) was used to superimpose the different 14-3-3 proteins complexes retrieved from the Protein Data Bank and the two modeled KSR1 phosphopeptides, using the 2B05 coordinates as the fixed structural reference. The threedimensional images were rendered with VMD (Visual Molecular Dynamics) (30).

KSR1 Interacts Preferentially with 14-3-3␥ in Vitro-
The seven human 14-3-3 isoforms are highly similar at the amino acid level, sharing an average identity of 70% and having strictly conserved residues directly involved in pSer recognition (9). To gain insight into the functional role of 14-3-3 proteins in KSR1 regulation, we first searched for any isotype-related difference in binding. The seven GST-tagged human 14-3-3 isoforms were purified to near homogeneity (supplemental Fig. 1A) and used as baits in pulldown experiments with FLAG-KSR1-expressing cell lysates. We could not perform direct interaction assays because production of full-length recombinant KSR1 was not possible in our laboratory. This problem has also been reported by other groups (31). As shown in Fig. 1, A and B, no KSR1 bound to recombinant 14-3-3, whereas only the highest concentrations of the ⑀ and isoforms were able to poorly associate with KSR1. On the contrary, KSR1 was efficiently pulled down, albeit to different extents, by isoforms ␤, ␥, , and . At any concentration tested, 14-3-3␥ consistently showed the strongest interaction with KSR1, suggesting a higher affinity of this isoform for KSR1. These differences cannot be attributed to altered binding properties of the GST fusion proteins, because difopein, a specific 14-3-3 ligand that does not possess isoform selectivity (32), was pulled down in similar amounts by all GST-14-3-3 proteins (supplemental Fig. 1B). In addition, none of them bound to the mutant ligand R18(Lys), indicating that the interaction occurred within the amphipathic groove (supplemental Fig. 1B) (32).
A combined pulldown/IP depletion assay confirmed that the amount of FLAG-KSR1 competent for binding to 14-3-3 (i.e. KSR1 phosphorylated on Ser-392) was not limiting in these assays (supplemental Fig. 1, C and D).
KSR1 forms a multiprotein complex inside the cells (18). Therefore, a legitimate caveat can be added that the interaction observed between KSR1 and 14-3-3␥ is mediated through a KSR1-associated protein and not by KSR1 itself. To exclude that possibility, we performed pulldown experiments with lysates expressing FLAG-KAA, a KSR1 mutant lacking the two consensus 14-3-3 binding sites, Ser-297 and Ser-392 (17). We did not detect binding of FLAG-KAA to GST-14-3-3␥ at any of the amounts tested (Fig. 1A). Next, we performed competition experiments with synthetic phosphopeptides based on amino acid sequences surrounding KSR1 residues Ser-297 and Ser-392. Preincubation of GST-14-3-3␥ with phosphopeptides PS-297 and PS-392, alone or in combination, provoked a marked reduction in KSR1 binding, whereas their nonphosphorylated counterparts had no effect (Fig. 1C). Collectively, these results confirm that the observed interaction between KSR1 and 14-3-3␥ is specific and is mediated by direct interaction with pSer-297 and pSer-392 in KSR1.
To allow for binding measurements at lower concentrations of the bait protein, we adapted the pulldown assay to a microtiter plate (ELISA) format. Again, 14-3-3␥ exhibited the highest affinity for KSR1 followed by the , ␤, and isoforms (supplemental Fig. 1E). At the low concentrations tested in the assay, we did not detect binding to the ⑀, , or isoform nor did we observe binding of 14-3-3␥ to the KAA mutant (data not shown).
KSR1 Interacts Preferentially with 14-3-3␥ in Vivo-To determine whether KSR1 also interacts preferentially with 14-3-3␥ in vivo, we performed coimmunoprecipitation (co-IP) experiments with lysates expressing FLAG-KSR1 and each of the seven Myc-tagged human 14-3-3 isoforms. Only 14-3-3␥ and 14-3-3 were detected in the FLAG immunoprecipitates ( Fig. 2A). In contrast, all seven isoforms associated similarly with EYFP-difopein, indicating that the Myc-tagged proteins are fully functional (supplemental Fig. 2A). Consistent with the results shown in Fig. 1, the amount of 14-3-3␥ associated with KSR1 was higher than that of 14-3-3, strongly confirming an intrinsic higher affinity of the ␥ isoform for KSR1. Using twice the amount of lysate gave the same result (data not shown). A 14-3-3␥ protein with a K50E mutation, which reportedly impairs ligand binding (34), did not associate with KSR1 (Fig.  2B), indicating that the interaction inside the cell is mediated by the phosphopeptide binding cleft. To further control for the specificity of the experiment we performed co-IP studies with C-Raf, which is the closest relative to KSR1 and displays overall sequence similarity and domain organization (13). FLAG-C-Raf immunocomplexes contained detectable levels of the ␥, , ⑀, , and isoforms, suggesting a more promiscuous interaction with 14-3-3 (Fig. 2C). In contrast to FLAG-KSR1, we consistently observed similar levels of the ␥ and isoforms bound to FLAG-C-Raf (Fig. 2C).

14-3-3/KSR1 interaction in vitro.
A, increasing amounts of GST-14-3-3 isoforms were incubated with 250 g of a 293 cell lysate expressing FLAG-KSR1. The amount of pulled down FLAG-KSR1 in each sample was determined by WB with a FLAG antibody, and GST-14-3-3 loading was controlled by Coomassie staining. As a control, pulldowns were done with GST-14-3-3␥ and lysates expressing FLAG-KAA, a KSR1 mutant that cannot bind 14-3-3. Shown are blots representative of the results obtained in four independent experiments. B, quantification of the results shown in A (mean Ϯ S.D. of three independent experiments). C, pulldown assays were performed in the absence (Buffer) or presence (150 M) of the competing PS-297 and PS-392 phosphopeptides or their unphosphorylated equivalents (P-297, P-392). D, pulldown assays were performed with 10 g of GST-14-3-3 proteins and 1.5 mg of a mouse brain lysate (B.L.). The intensity of the B-KSR1 bands, as detected with a KSR1 antibody, was quantitated by densitometry and normalized to the value for ␥ ϭ 1.0.

14-3-3 Isoform-specific Binding to KSR1
CRD of KSR1 (named CA3) confers specificity to its interaction with 14-3-3␥. A KRK chimeric mutant was generated by swapping CA3 with the CRD domain of C-Raf. Co-IP experiments showed that, similar to KSR1, the KRK mutant only interacted with the ␥ and isoforms (supplemental Fig. 2C). No significant differences were observed in the binding of 14-3-3␥ to KSR1 or KRK as determined by co-IP (supplemental Fig. 2C,  compare lanes 1 and 4) or in our ELISA-based assay (supplemental Fig. 2D). On the contrary, a KSR1 mutant lacking a functional CA3 domain (CA3mut) was greatly impaired in its ability to interact with 14-3-3␥ (supplemental Fig. 2D). Taken together, these results suggests that the overall three-dimensional structure of CRDs may play a role in 14-3-3 substrate recognition, whereas variations in the sequence of different CRDs seem to be less relevant.
Finally, we used a mammalian two-hybrid system to quantitate the in vivo binding of each isoform to KSR1. Cotransfection of KSR1 and 14-3-3␥ induced the highest expression of CAT protein, indicative of a strong interaction between both pro-teins (Fig. 2D). CAT expression correlated well with the different affinities observed for each isoform in the co-IP assays, with 14-3-3 being the second preferred isoform after 14-3-3␥. The fact that some CAT was produced by coexpression of KSR1 and 14-3-3, which did not bind KSR1 in any of our assays, might be technically related; the two-hybrid system is designed so that expressed proteins are directed to the nucleus, where we cannot exclude a KSR1/14-3-3 association of unknown relevance. In sum, the above experiments confirmed in vivo the existence of ligand specificity among 14-3-3 isotypes and, in particular, that of 14-3-3␥ for KSR1.
Functional Relevance of the 14-3-3␥/KSR1 Interaction-Next, we wanted to determine whether the 14-3-3␥/KSR1 interaction was affected by growth factors. To that end, 293 cells cotransfected with both proteins were stimulated with EGF, and the amount of 14-3-3␥ bound to KSR1 was determined in co-IP experiments. In agreement with the current model of KSR1 activation (19), a 5-min EGF stimulation reduced the 14-3-3␥/KSR1 interaction concomitant with a dephosphorylation of KSR1 in residue Ser-392 (Fig. 3A), likely due to activation of the phosphatase PP2A. This effect was transient, as KSR1 and 14-3-3␥ began to reassociate 30 -60 min after stimulation, coincident with an increase in pSer-392 levels (Fig. 3A). Next, we coexpressed KSR1 with increasing amounts of 14-3-3␥ and determined the amount of pSer-392 after EGF addition. Stimulation of quiescent cells expressing FLAG-KSR1 alone provoked a reproducible dephosphorylation of pSer-392 (Fig. 3B). However, expression of Myc-14-3-3␥ inhibited dephosphorylation of KSR1 in a dose-dependent manner (Fig.  3B) indicating that 14-3-3␥ can regulate KSR1 signaling by protecting it from PP2A-mediated dephosphorylation (which likely explains the relatively modest effect of EGF on pSer-392 observed in Fig. 3A). These results demonstrate that KSR1 and 14-3-3␥ form functional complexes that are responsive to growth factor stimulation. KSR1 has been shown to facilitate oncogenic Ras signaling in Xenopus oocytes (17) and 14-3-3 proteins play an essential role in that process by sequestering KSR1 in the cytosol until Ras activation (19,21). We reasoned that KSR1 would fail to cooperate with Ras in oocytes overexpressing KSR1-interacting 14-3-3 isoforms because of its enhanced cytoplasmic retention. On the contrary, overexpression of those isoforms that do not interact with KSR1 should not interfere that function. KSR1 interacted with the same 14-3-3 isoforms in oocytes and in mammalian cells (supplemental Fig. 3A). We microinjected oocytes with mRNAs encoding Myc-KSR1 and HA-14-3-3 isoforms together with oncogenic RasG12K and scored the percentage of oocytes undergoing maturation (% GVBD). Overexpression of 14-3-3␥ or 14-3-3 completely blocked the ability of . Functional significance of the 14-3-3␥/KSR1 interaction. A, 293 quiescent cells, expressing the indicated proteins, were stimulated at different times with 50 ng/ml EGF, and co-IP experiments were performed as described for Fig. 2A. The activation state of KSR1 was determined by WB with an antibody specific for KSR1(pS392). B, cells were transfected with KSR1 alone or in combination with increasing amounts of 14-3-3␥, stimulated for 5 min with EGF, and processed for co-IP. Quantification of KSR1(pS392) normalized to total KSR1 in the immunoprecipitates is shown. C, oocytes were microinjected with mRNAs encoding for Myc-KSR1 and each of the seven HA-14-3-3 isoforms. They were reinjected 5 h later with a mRNA specific for Myc-RasG12K. GVBD was scored when 5-10% of the oocytes injected with Myc-RasG12K alone had initiated maturation. Translation of the mRNAs was assessed by WB with Myc and HA antibodies. Activation of ERK2 was detected by WB as a decrease in its electrophoretic mobility (due to phosphorylation), whereas activation of CDK1 resulted in a faster migrating band (due to dephosphorylation). D, U2OS cells grown in 6-well plates with or without coverslips were mock-transfected (C, control) or transfected twice with 20 nM siRNAs either scramble (S) or specific for 14-3-3 () or 14-3-3␥ (␥). On day 3, cells were transfected with FLAG-KSR1. Endogenous levels of 14-3-3, 14-3-3␥ and actin were determined on day 5 by WB with specific antibodies (upper panel). Cells grown on coverslips were processed for immunofluorescence with a FLAG antibody (lower panel) to detect membrane localization of FLAG-KSR1 (arrows). Selected areas of plasma membrane (bracketed) are shown enlarged. The graph shows the percentage of FLAG-positive cells transfected with S, ␥, or siRNAs showing membrane localization (mean Ϯ S.D.). At least 250 cells/condition were counted in two independent assays.

14-3-3 Isoform-specific Binding to KSR1
KSR1 to accelerate Ras-induced maturation, whereas overexpression of the ⑀, , and isoform had almost no effect (Fig. 3C). The ␤ and isoforms, which also interacted with KSR1 albeit with lower affinity, provoked a less marked inhibition in Rasinduced maturation. Importantly, the GVBD blockade observed in oocytes injected with 14-3-3␥ and 14-3-3 mRNAs was accompanied biochemically by a marked reduction in ERK2 activation (Fig. 3C, lower panel) indicating that the specific binding of KSR1 to the ␥ and isoforms inhibits its ability to potentiate the Ras-induced activation of ERK2 in oocytes.
To further confirm the functional relevance of the 14-3-3␥/ KSR1 interaction, we next inhibited the expression of endogenous 14-3-3␥ and 14-3-3 in human U2OS cells by using specific siRNAs and looked at the subcellular distribution of FLAG-KSR1. We were not able to silence expression of the ␤, ⑀, , , and isoforms by using either synthetic siRNAs or short hairpin RNA-expressing vectors (data not shown). The ␥ and isoforms were reduced by 60% at the protein level (Fig. 3D,  upper panel). Silencing of endogenous 14-3-3␥ provoked a marked accumulation of FLAG-KSR1 at the plasma membrane, whereas silencing of 14-3-3 had no effect (Fig. 3D, lower  panel). These results, as well as those from Fig. 3C, are in agreement with a specific role for 14-3-3␥ in preventing uninduced KSR1 translocation to the plasma membrane.
Molecular Modeling of the 14-3-3␥/KSR1 Interaction-To search for a structural basis underlying the specific binding of 14-3-3␥ to KSR1, we modeled the interaction of KSR1 phosphopeptides RSKpSHE (PS297) and RTEpSVP (PS392) with 14-3-3␥ using the published structure of 14-3-3␥ bound to a mode 1 phosphopeptide, RAIpSLP, as template (9). The resulting models (Fig.  4, A and B) showed that all 14-3-3␥ residues predicted to have at least 1 atom root mean square distance Ͻ5.0 Å from any atom in PS297 or PS392 are strictly conserved in the seven human isoforms, including residues Lys-50, Lys-57, Arg-132, and Tyr-133, which hold the pSer phosphate in all solved structures (9) (data not shown). These results suggest that interaction conformation specificity is determined mainly by the phosphopeptide sequence rather than by residues in the binding groove. In fact, superimposing the same 14-3-3 isoform bound to different phosphopeptides revealed changes in their spatial conformation, particularly toward their N and C termini (data not shown). The conserved Arg at pSer-3 showed a broad range of conformational freedom, making any conformation prediction at the N terminus too speculative. Interestingly, however, our model revealed that phosphopeptide structures with polar instead of hydrophobic residues in pSerϩ1 show a significant shift in the orientation of the conserved Pro in pSerϩ2. The predicted conformation and orientation of PS392 at the C terminus was similar to those in phosphopeptides with hydrophobic C-terminal sides (Fig. 4C). In marked contrast to PS392, PS297 has a highly polar C-terminal side (pSHE) and lacks the conserved Pro. As a result, PS297 cannot complement the conserved 14-3-3 hydrophobic patch (36) and is predicted by our model to be shifted away from the binding cleft (Fig. 4D), which might explain its reported lower binding affinity for 14-3-3 (17). In summary, our model indicates that direct side-chain interactions at the conserved ligand binding cleft do not seem to account for 14-3-3␥ specific binding to KSR1, whereas residue composition at the C terminus of its two consensus binding sites might determine their binding affinity.
The C Terminus of 14-3-3␥ Participates in Selective Binding to KSR1-A phylogenetic tree of the human 14-3-3 proteins shows that the closest isoforms to the related ␥ and are the ␤, , and , whereas the ⑀ and are classified in different clades (37). Interestingly, this ranking correlates fairly well with the different isoform affinities observed for KSR1 (Figs. 1 and 2). The highest level of variation among isoforms occurs in a C-terminal stretch (CTS) of acidic residues. Particularly, 14-3-3␥ and 14-3-3 share similar short CTSs, while the ⑀ and isoforms have the longest and most divergent CTSs (Fig. 5A). To investigate whether this region confers specificity to the 14-3-3␥/ KSR1 interaction, we deleted the CTS in each 14-3-3 isoform and PS392 to visualize their differences at positions pSerϩ1 and pSerϩ2, in the C terminus, below the conserved hydrophobic patch. All residues are colored as basic (blue), acidic (red), hydrophobic (green), or noncharged polar (white). PS392 residues are italicized.
Homo-and Heterodimeric Forms of 14-3-3␥ Are Bound to KSR1-We next evaluated the role played by dimerization in 14-3-3␥ ligand discrimination. The Myc-14-3-3␥ protein used in our studies was essentially able to homo-and heterodimerize with all overexpressed isoforms except 14-3-3 (Fig. 6A). Myc-14-3-3␥ also formed heterodimers with all endogenous isoforms except and (Fig. 6B). Therefore, the possibility exists that some of the interactions observed in our experiments were reflecting selective 14-3-3␥ heterodimers. To test this hypothesis, we developed an experimental protocol to specifically isolate the KSR1-bound pool of 14-3-3␥ molecules and determine  A, The C termini of the seven human 14-3-3 isoforms (residues shown in parentheses) were first aligned using a pairwise-based software and then manually according to the length of their divergent CTSs. White characters denote strict identity, and bold characters denote similarity within a group. Similarity across groups is highlighted by frames. B, Myc-14-3-3 deletion mutants (⌬C), each lacking a specific CTS, were coexpressed in COS-7 cells with FLAG-KSR1, and their interaction was determined by co-IP as described in Fig. 2A. C, binding of ␥-⌬C and -⌬C mutants to KSR1 was compared with that of their wild-type counterparts by co-IP. D, the C-terminal tails of 14-3-3␥ and 14-3-3⑀ were swapped to generate the chimeric constructs 14-3-3␥/⑀ and 14-3-3⑀/␥, and their binding to KSR1 was assessed by co-IP. E, the relevance of the CTS sequence for KSR1 binding to 14-3-3␥ and 14-3-3 was determined in co-IP assays using COS-7 cells lysates expressing FLAG-KSR1 together with the indicated 14-3-3␥ and 14-3-3 chimeric mutants.

hEta(215-246) S Y K D S T L I M Q L L R D N L T L W T S D D E E Q Q A G E G N hGamma(215-247) S Y K D S T L I M Q L L R D N L T L W T S D D D D Q Q G G E G N N hBeta(212-246) S Y K D S T L I M Q L L R D N L T L W T S E G D E N Q G D A G E G E N hZeta(210-245) S Y K D S T L I M Q L L R D N L T L W T S D G D E T Q A E A G E G G E N hTau(210-245) S Y K D S T L I M Q L L R D N L T L W T S D G E E S A C D A A E G A E N hSigma(212-248) S Y K D S T L I M Q L L R D N L T L W T A D G E E N A G G E A P Q E P Q S
its dimerization profile with other 14-3-3 isoforms (Fig. 6C). We detected ␥/␥ homodimers and ␥/ and ␥/ heterodimers associated with KSR1 (Fig. 6D). We did not detect the ⑀ and isoforms as part of the 14-3-3␥/KSR1 complex (Fig. 6D), in agreement with their lack of interaction or functional effect on KSR1 observed in different assays (Figs. 1-3). The absence of 14-3-3 in the complex was also expected, as this isoform preferentially forms homodimers (37). Interestingly, 14-3-3 and 14-3-3␤ were also absent from the 14-3-3␥/KSR1 complex (Fig.  6D), suggesting that the interaction observed in our binding assays reflects a direct and specific binding of these isoforms to KSR1. In the case of 14-3-3, we cannot exclude an interaction below the detection limit of the antibody because the signal in the control lysate was typically weak (Fig. 6D). The 14-3-3 antibodies used were fairly specific, with the exception of the anti-14-3-3, which cross-reacted with the ␥ and ⑀ isoforms (supplemental Fig. 3B). This nonspecificity lacks relevance here, as no WB signal could be detected in the 14-3-3␥/KSR1 complex using that antibody (Fig. 6D). Thus, 14-3-3␥ binds to KSR1 as a homodimer or as a heterodimer with selected isoforms. Role of 14-3-3␥ Dimerization-To further elucidate the role of 14-3-3␥ dimerization in KSR1 binding, we attempted to generate a 14-3-3␥ dimer-deficient mutant. Phosphorylation of 14-3-3 on Ser-58 or its phospho-mimic mutation to either Asp or Glu severely impairs dimer formation due to an increase in dimerization free energy (27,38). Ser-58 is located at the dimer interface and is conserved in all human 14-3-3 isoforms, except and . Therefore, it seemed plausible that substitution of the equivalent Ser-59 for an Asp would render a monomeric form of 14-3-3␥ by destabilizing the dimer. Cross-linking experiments confirmed that a 14-3-3␥-S59D mutant was mainly monomeric (Fig. 7A) and deficient in both hetero-and homodimerization (Fig. 7B). The small amount of 14-3-3␥ and 14-3-3⑀ found in the 14-3-3-␥-S59D immunoprecipitates likely reflects residual dimerization (Fig. 7A), magnified by the fact that 14-3-3␥ dimerized mainly with itself and with 14-3-3⑀ (Figs. 6A and 7B). Taken together, these results demonstrate that 14-3-3␥-S59D is a dimer-deficient mutant.

DISCUSSION
The identification of 14-3-3 isoform-specific ligands is of cardinal importance in the field of signal transduction, as it might help to discriminate functional roles in this highly conserved family of proteins (1). 14-3-3 proteins have evolved so that unicellular eukaryotes have only a few isoforms, whereas multicellular ones have many (39), likely reflecting the need for isoform-specific functions in complex organisms. Here, we have reported a detailed analysis, both in vitro and in vivo, of functional specificity in 14-3-3 proteins, including all human isoforms. In a variety of experimental settings, the molecular scaffold KSR1 consistently showed a marked preference for binding to 14-3-3␥, although it was able also to interact with

α-Flag
Flag-14-3-3γ FIGURE 6. 14-3-3␥ binds KSR1 in both homo-and heterodimeric forms. A, COS-7 cells transfected with FLAG-14-3-3␥, and the indicated Myc-14-3-3 isoforms were processed by co-IP. B, 293 cells expressing FLAG-14-3-3␥ (Lysate) were processed for co-IP with a FLAG antibody or a nonimmune IgG, and binding to endogenous human 14-3-3 isoforms was detected by WB with isoform-specific antibodies. C, overview of the protocol followed in D. D, 293 cells expressing FLAG-KSR1 and Myc-14-3-3␥ (Lysate) were processed as outlined in C to isolate the FLAG-KSR1-bound pool of Myc-14-3-3␥ molecules. Shown are the results of the second immunoprecipitation (IP2) including two controls in parallel: a co-IP with a nonimmune IgG and a co-IP with a Myc antibody using a lysate expressing FLAG-KSR1 alone. Recovery of the myc-14-3-3␥/FLAG-KSR1 complex was assessed by WB with Myc and FLAG antibodies (two upper panels). Aliquots (10 l) of IP2 were fractionated by SDS-PAGE, and the presence of endogenous 14-3-3 isoforms was detected by WB as described in B. Note that a 14-3-3␥ antibody detects both the endogenous and the Myc-tagged species.
some but not all 14-3-3 isoforms ( Ͼ ␤, Ͼ ). In vivo KSR1 interacted almost exclusively with 14-3-3␥ and 14-3-3 (␥ Ͼ ). Notably, this ligand specificity correlated with sequence similarity among the seven human isoforms, which, given their conserved structural features (9), strongly suggests that regions of sequence variation among 14-3-3 isoforms contribute to ligand discrimination. In particular, the short stretch of acidic residues (CTS) located at their C termini is the most divergent region among isoforms, and we have demonstrated here its critical role in KSR1 recognition. The in vitro assays showed that the longer the CTS, the weaker the binding to KSR1. Also, KSR1 associated in vivo exclusively with 14-3-3␥ and 14-3-3-3, which share a short and almost identical CTS. Therefore, both the sequence and the length of the CTS seem to be important for substrate specificity. Previous work form Truong et al. (40) had shown that a ⌬C form of 14-3-3, lacking its CTS, could bind to Raf1 and BAD with higher affinity than the full-length protein, supporting the notion of an autoinhibitory role for that region. These data could reflect a specific feature of 14-3-3 binding to Raf1 and BAD or be due to alleviation of the reported phosphorylation-induced conformational change in the 14-3-3 CTS, which inhibits ligand binding (41). In our hands, however, deletion of the CTS impaired KSR1 binding to the ␥ and isoforms and did not increase binding to the other isoforms. Notably, the CTSs of these two related isoform were interchangeable, whereas swapping the CTSs between two dissimilar isoforms like 14-3-3␥ and 14-3-3⑀ did not allow full binding to KSR1. A 14-3-3␥/ chimera was intermediate between 14-3-3␥/⑀ and 14-3-3␥/ in terms of KSR1 binding, further indicating that sequence similarity at the CTS also correlates with isoform specificity. According to Obsil and co-workers (42), the flexible CTS folds back into the ligand binding pocket and can be displaced only by specific phosphopeptide sequences.
Our results are compatible with this model, although, rather than a strictly autoinhibitory role, they suggest a positive role for the CTS in ligand discrimination by stabilizing specific interactions engaged by the core of the 14-3-3 protein. They are also in accordance with recent data from optimal docking area studies suggesting that, secondary to the phosphopeptide binding, general protein-protein interactions at the so-called "desolvation patches" and subsequent chain/loop rearrangements allow for isoform-specific ligand binding (36). In sum, our results suggest that contacts with the core of each 14-3-3 isoform mainly determine substrate specificity. Once the primary interaction has occurred, the isoform-specific CTSs seems to recognize bona fide ligands and stabilize the complex. Computational modeling of the 14-3-3␥ interaction with KSR1 phosphopeptides did not reveal any specific feature that could account for isoform specificity. However, we cannot exclude ligand-or phosphorylation-induced structural rearrangements at the 14-3-3␥ binding groove to specifically accommodate KSR1. Still, our model could serve a structural basis to explain the reported evidence that some 14-3-3 ligands, including KSR1, use their two binding sites hierarchically with one site being dominant over the other (4). Thus, in BAD, RSRpS136AP is dominant over RHSpS112YP (43), in C-Raf RSTpS259TP binds more strongly than RSApS621EP (44), and in KSR1 RTEpS392VP is dominant over RSKpS297HE (17). Interestingly, pSer-112 in BAD, pSer-621 in C-Raf, and pSer-297 in KSR1 are all within motifs with non-hydrophobic residues at the pSerϩ1 position (Tyr, Glu, and His, respectively). Their inability to complement a conserved hydrophobic patch in the roof of the binding cleft (36) might be the reason for their lower binding affinity to 14-3-3 proteins. In addition, KSR1 also lacks the conserved Pro at pSerϩ2, a situation that reduces phosphopeptide binding (10).
Dominant sites have been proposed to function as "gatekeepers" that bind one monomer in the 14-3-3 dimer and bring the other subunit closer to the secondary, low affinity site, which then acts to stabilize the dimer (45). This mechanism allows for simultaneous binding of two identical or different substrates, thereby amplifying the range of 14-3-3 specific functional interactions. We have shown here that, despite being able to homodimerize and heterodimerize with all isoforms except , 14-3-3␥ binds to KSR1 preferentially as ␥/␥, ␥/, and ␥/ dimers, suggesting that KSR1 function can be regulated by 14-3-3 heterodimerization. We did not detect ␥/ or ␥/␤ . Role of 14-3-3␥ dimerization in KSR1 binding. A, immunoprecipitates of FLAG-14-3-3␥ or FLAG-14-3-3␥-S59D were treated or not with the chemical cross-linker bis(sulfosuccinimidyl) suberate (BS 3 ), fractionated by SDS-PAGE, and immunoblotted with a 14-3-3␥ antibody to reveal its monomeric and dimeric forms. An aliquot of the lysate shows the endogenous and the FLAG-tagged 14-3-3␥ proteins. B, FLAG-14-3-3␥ and FLAG-14-3-3␥-S59D were coexpressed with the seven Myc-tagged human 14-3-3 isoforms to determine their ability to homo-and heterodimerize by co-IP experiments. Both short and long exposures of the blots are shown. C, co-IP assays were performed with lysates expressing FLAG-KSR1 or the indicated mutants together with Myc-tagged 14-3-3␥ or 14-3-3␥-S59D.
dimers, which indicates that the binding of KSR1 to 14-3-3 and, to a lesser extend, 14-3-3␤ observed in our in vivo assays is direct and specific. Further, the KSR1-bound pool of 14-3-3␥ molecules did not dimerize (at least to a detectable level) with endogenous 14-3-3⑀, although, in agreement with previous data (36, 46), 14-3-3␥ formed dimers preferentially with that isoform (Figs. 6A and 7B). The role played by different 14-3-3 dimers in the regulation of KSR1 function awaits further investigation, but it could be relevant in coupling to other 14-3-3binding proteins, altering its subcellular localization, and/or inducing conformational changes. Nevertheless, our data stress the notion of isoform specificity in ligand discrimination and also highlight the complex mechanisms that 14-3-3 proteins may use to achieve functional specificity.
Several reports have demonstrated that dimerization-deficient 14-3-3 mutants can bind their targets as strongly as wildtype 14-3-3 (47-50), and we have also shown here that KSR1 can bind a phospho-mimic, dimer-deficient 14-3-3␥-S59D mutant. Although Ser-297 phosphorylation is constitutive and does not change after Ras activation, the dominant site Ser-392 gets dephosphorylated by PP2A, resulting in dissociation of the KSR1/14-3-3 complex, exposure of the CA3 domain, and translocation of KSR1 to the plasma membrane (19). An unsolved question is how the PP2A catalytic subunit gains access to pSer-392, given its buried position in the phosphopeptide binding cleft. In fact, protecting ligands from dephosphorylation by phosphatases is one important 14-3-3 function (51) (see also Fig. 3B). Interestingly, the S59D mutation in 14-3-3␥ mimics an agonist-induced phosphorylation event that reportedly impairs dimer formation in vivo (27,38,52). Of potential relevance, we found that a KSR1-S392A mutant retained binding to 14-3-3␥-S59D, whereas a KSR1-S297A did not, suggesting that KSR1 bound monomeric 14-3-3␥ through the low affinity site Ser-297. These results raise the intriguing possibility that Ras-induced phosphorylation of 14-3-3␥ might be an early step in KSR1 activation. Relaxation of the KSR1 molecule, because of 14-3-3␥ dimer disruption, might then facilitate access to pSer-392 by the catalytic subunit of PP2A. As PP2A has been shown to be specific for pSer-392 (19), the remaining pSer-297 might sustain low affinity binding to one monomer as part of the mechanism by which KSR1 translocates to the plasma membrane. We are currently investigating this possibility.
The identification of isoform-specific 14-3-3 interactions will help in the design of more physiologically relevant experiments, whereas elucidating the molecular determinants of specificity will help in designing more selective inhibitors. In particular, the combination of the in vitro and in vivo assays shown here has allowed the identification of 14-3-3␥ as the likely isoform that regulates KSR1 translocation to the plasma membrane.