The adapter protein Dok-4 (downstream of kinase-4) has been reported as both an activator and inhibitor of Erk and Elk-1, but lack of knowledge about the identity of its partner molecules has precluded any mechanistic insight into these seemingly conflicting properties. We report that Dok-4 interacts with the transactivation domain of Elk-4 through an atypical phosphotyrosine-binding domain-mediated interaction. Dok-4 possesses a nuclear export signal and can relocalize Elk-4 from nucleus to cytosol, whereas Elk-4 possesses two nuclear localization signals that restrict interaction with Dok-4. The Elk-4 protein, unlike Elk-1, is highly unstable in the presence of Dok-4, through both an interaction-dependent mechanism and a pleckstrin homology domain-dependent but interaction-independent mechanism. This is reversed by proteasome inhibition, depletion of endogenous Dok-4 or lysine-to-arginine mutation of putative Elk-4 ubiquitination sites. Finally, Elk-4 transactivation is potently inhibited by Dok-4 overexpression but enhanced by Dok-4 knockdown in MDCK renal tubular cells, which correlates with increased basal and EGF-induced expression of Egr-1, Fos and cylcinD1 mRNA, and cell proliferation despite reduced Erk activation. Thus, Dok-4 can target Elk-4 activity through multiple mechanisms, including binding of the transactivation domain, nuclear exclusion and protein destabilization, without a requirement for inhibition of Erk.
The Dok (downstream of kinase) family of adapter proteins consists of seven related proteins characterized by an N-terminal tandem of conserved pleckstrin homology (PH) and phosphotyrosine-binding (PTB) domains linked to more divergent C-terminal (CT) sequences . The first three members of the Dok family, Dok-1, 2 and 3, form a separate subfamily (Dok-A) and have been consistently found to inhibit Ras/Erk pathway activation downstream from tyrosine kinases. The inhibitory function of Dok-A family members derives from their interaction with multiple inhibitory effector molecules such as the Ras inhibitor p120RasGAP , the Src kinase inhibitory kinase Csk  or the lipid phosphatase Ship1 [3–5]. Expression of Dok-1, 2 and 3 is generally restricted to immune cells including monocyte/macrophages . Consistent with their role in inhibitory signaling, combined inactivation of Dok-1, 2 and 3 in mice results in aggressive histiocytic sarcoma . In contrast, Dok-4, 5 and 6 form a distinct subfamily (Dok-B) whose function is poorly understood. While expression of Dok-5 and 6 is essentially restricted to neural tissues [7–9], Dok-4 is expressed almost ubiquitously, with a preference for neural, endothelial and epithelial tissues . Dok-4 displays a striking level of phylogenetic conservation with 99.7% or more amino acid identity between human, mouse, rat, pig and dog sequences (data not shown). Dok-7 is more distant member of the Dok family expressed in muscles and involved in neuromuscular junction formation [10,11].
Activation of the Ras/Erk pathway is one of the most common events downstream from tyrosine kinase-initiated signaling and it serves as a nexus for regulation of key cellular functions . It is also a key target of regulation by Dok-A family members . Once activated, Erk can phosphorylate a wide array of substrates, both cytosolic and nuclear. While its cytosolic substrates post-translationally regulate cellular processes such as adhesion and motility, nuclear substrates of Erk tend to impart transcriptional responses and are particularly important for survival and proliferative responses [12,13]. The prototypical nuclear substrates of Erk are the three ternary complex factors (TCFs), Elk-1, Elk-3 (Net) and Elk-4 (Sap-1), of which Elk-1 is by far the best studied. TCFs form a unique subfamily within the larger Ets (E-twenty six) family of transcription factors . While they share a DNA-binding Ets domain with the other members of the Ets superfamily, TCFs are unique in their ability to heteromerize with another transcription factor, serum response factor (Srf), which binds DNA but lacks a transactivation domain. TCF/Srf complexes bind to serum response elements, which consist of paired Ets- and Srf-binding DNA sequences and are located in the promoter region of many immediate early genes (IEGs) such as egr-1 and c-fos [15–18]. Since the transcriptional activity of TCFs is dramatically increased upon phosphorylation of their CT transactivation domain by Erk , growth factor-induced TCF activation is a powerful and often essential stimulus for up-regulation of Fos and Egr-1 mRNA . In turn, Fos and Egr-1 proteins both lead to up-regulation of cyclin D1 [21–24], which is critical for G1/S cell cycle progression and mitogenesis . Hence, tyrosine kinase-induced TCF activation is critical for transcriptionally driven cell proliferation .
The role of Dok-4 in regulation of IEG transcription and cellular proliferation remains unknown. Whereas Dok-1 and Dok-2 consistently inhibit Erk activation and mitogenesis downstream from tyrosine kinase signaling [27–31], the impact of Dok-4 on Erk has been controversial since both positive and negative influences have been reported [8,9,32–34]. A major obstacle to our understanding of Dok-4 adapter protein function has been the extremely limited information available on the identity of its partner molecules. More than 15 years after the first report on Dok-4 , only three known partners of Dok-4 have been convincingly identified: two receptor tyrosine kinases, Ret and Tie2 , and the lipid phosphatase Ship1 . Unfortunately, none of these Dok-4 partners are expressed in epithelial cells, where Dok-4 expression predominates .
In the current study, we have identified Elk-4 as a novel binding partner to Dok4. Furthermore, we identified Elk-4 as the predominant TCF family member expressed in the kidney. We show that Dok4 can inhibit Elk-4 activity as well as cause its nuclear exclusion and destabilization. Knockdown of endogenous Dok4 in MDCK cells resulted in up-regulation of TCF target genes such as the IEGs Egr-1 and Fos, up-regulation of cyclin D1 and enhanced cell proliferation despite reduced Erk activation.
Materials and methods
Yeast two-hybrid screening
Yeast two-hybrid (Y2H) screening was performed by Hybrigenics (Paris, France) using full-length mouse Dok-4 fused C-terminally to LexA as bait, as recently described . The library consisted of random-primed mouse kidney cDNAs fused to the Gal4 activation domain. A total of 75 and 90 million possible interactions were screened with and without co-expression of Lyn tyrosine kinase, respectively.
Plasmids and cDNAs
The Dok-4 and Elk-4 constructs were generated in the following vectors: pcDNA3.1 (−) Myc/His version A (Invitrogen), pEGFP-N1 (Clontech), pEGFP-C1 (Clontech), pmCherry-C1 (Clontech) and pGEX2T (GE Healthcare). The following Dok-4 constructs were described previously [7,32]: pcDNA-Dok-4-Myc/His (1–325), pcDNA-Dok-4-Myc/His ΔPH (100–325), pcDNA-Dok-4-Myc/His CT (aa 131–325), pcDNA-Dok-4-Myc/His ΔCT (aa 1–233), pcDNA-Dok-4-Myc/His Δ100–233, pEGFP-N1-Dok-4 WT, pEGFPN1-Dok-4 Δ100–325, pGEX2T-Dok-4 ΔPH (aa 100–325).
The additional Dok-4 constructs, pGEX2T-Dok-4 PTB (aa 130–246), Dok-4-Myc/His (aa 100–325, G207A), Dok-4-Myc/His (aa 100–264), Dok-4-Myc/His (aa 100–246), Dok-4-Myc/His (G207A), pEGFP-N1-Dok-4 G12D, pEGFP-N1-Dok-4 G12D/L116,121A and W103A/L116,121A were generated by PCR cloning from the original Dok-4 expression vectors described above. The expression vector for activated (Y528F) Fyn was also described previously .
The mouse Elk-4 (accession #BC145795, in pCR4-TOPO) and the human Elk-4b cDNA (accession #BC063676, in pCMV-Sport6) were obtained from ImaGenes. For expression in mammalian cells, mElk-4 was subcloned to pcDNA3.1. The pCDNA3.1-Elk-4 ΔCT (aa 1–342) construct was created by truncation at the internal BamHI site of Elk-4. The Elk-4 NLS mutant was obtained by lysine-to-glutamic acid mutation within two putative nuclear localization signals (NLSs; K49,51E and K319,320E.) The Elk-4K/R mutant was obtained by lysine-to-arginine mutation at positions 69, 74, 92, 117, 124, 133, 177, 195 and 196. The Elk-4/Elk-1 CT substitution mutant was obtained by replacing aa 360–371 of Elk-4 with the corresponding aa in human Elk-1 (361–374). The EGFP-Elk-4 and mCherry-Elk-4 constructs were created by replacing the start codon of Elk-4 with an in-frame XhoI site by PCR followed by subcloning to pEGFP-C1 and pmCherry-C1, respectively. The pmCherry-C1-Elk-4 ΔCT (1–342) construct was generated by truncation at the BamHI internal site. The pEGFP-C1-Elk-4 CT (aa 317–430) and pmCherry-C1-Elk-4 CT constructs were created by deletion of an N-terminal BglII fragment from the respective pEGFP-Elk-4 and pmCherry-Elk-4 vectors. The Venus-N-Elk-4 constructs were obtained by fusing a Venus aa 1–157 sequence to the N-terminal end of Elk-4 in pCDNA3.1. The Dok-4-Venus-C construct was obtained by fusing a Venus aa 158–238 sequence to the C-terminus of Dok-4 in pCDNA 3.1. In Venus constructs, a flexible (GGGGS)x2 linker was included. The pFA2-Elk-1 plasmid (encoding aa 1–147 of the Gal4 DNA-binding domain fused to aa 307–427 of the Elk-1 transactivation domain) and the pFR-Luc plasmid (encoding Gal4-binding DNA sequences placed upstream of the firefly luciferase cDNA) were purchased from Stratagene/Agilent. Construct pFA2-Elk-4 CT was produced by creating an internal in-frame BclI site in Elk-4 to replace the Elk-1 sequence from pFA2-Elk-1 (removed by BamHI/EcoRI digestion) with aa 312–430 of Elk-4. The pFA2-Elk-4 ΔEts construct was generated using a similar strategy to fuse Gal4 to aa 90–430 of Elk-4. The pFA2-Elk-4 CT N395A construct was created by PCR mutagenesis of the pFA2-Elk-4 CT vector. All constructs created by PCR were fully sequenced.
Ponceau S (Sigma) was diluted at a concentration of 0.1% in 1% acetic acid solution. Epidermal growth factor (EGF)  was purchased from Biosource and used at a concentration of 100 ng/ml. Leptomycin B (LMB) (Sigma) was dissolved in 70% methanol and used at a concentration of 20 ng/ml for 4 h. MG-132 (Sigma) was dissolved in DMSO and used at a concentration of 25 µM unless otherwise specified. Chloroquine (CQ; Sigma) was dissolved in water and used at a final concentration of 10 µM. Bortezomib (EMD Millipore) was dissolved in DMSO and used at a concentration of 100 nM.
Cells and transfection
293HEK cells, COS-1 cells and Madine-Darby canine kidney (MDCK) II cells were cultured in DMEM-high glucose containing pyridoxine-HCl and sodium pyruvate (Invitrogen) with 10% fetal bovine serum (FBS) (Invitrogen). MDCK II EGFP-Elk-4 clone #6 was generated by transfection of pEGFP-N1-Elk-4 and selection in G418 (400 µg/ml). Plasmid transfections were done using the Lipofectamine 2000 reagent (Invitrogen) except for luciferase assays in 293HEK cells, which were performed using the TransIT reagent (Mirus).
As control siRNA, we used a mixture of siCONTROL #1 and #2 from Dharmacon. For Dok-4 knockdown we used a pool of two dog Dok4-specific siRNAs targeting separate sequences (ACTCAGAGCTGGAGGCAGA and ACGTGAGGCTGCTGAACAA). Dok-4 siRNA was designed using the online siRNA design tool from Dharmacon and was obtained from the same company. For most experiments, the siRNA were delivered into MDCK II cells using the Dharmafect #1 reagent. RNA was isolated 24 h after transfection. For luciferase assays, Lipofectamine 2000 was used to deliver siRNA and plasmids simultaneously.
Antibodies and immunoblotting
Antibodies for c-Src, GFP, GST, Elk-4 (sc-13030), Fyn, Gal4, Myc tag and p38 were from Santa Cruz. For detection of endogenous Elk-4 antibodies from either Novus (H00002005-B01P) or Sigma (SAB1405753) were used, as indicated in the legends. Anti-α-tubulin antibody was from Sigma. Anti-non-phospho-Src (Y527), anti-phospho-ERK1/2, anti-ERK (total), anti-JNK, anti-phospho-JNK and anti-phospho-p38 antibodies were from Cell Signaling. Antibodies for GAPDH and His were from Genescript. Anti-phosphotyrosine (4G10) was from Millipore. Dok-4 rabbit antiserum was generated in our lab as previously described . Horseradish peroxidase-coupled secondary antibodies were from Jackson Immunoresearch and Rockland Immunochemicals. Cell lysis and immunoblotting were performed according to the standard protocol as detailed previously . In experiments not involving immunoprecipitation (IP) or pull-down assays, lysates were obtained directly in 1× Laemmli gel-loading buffer. Blots were revealed using ECL or ECL Plus reagents (GE Healthcare) or using SuperSignal West Pico (Thermo Scientific).
293HEK cells were transiently transfected with Lipofectamine 2000 (Invitrogen). Forty-eight hours after transfection, the cells were lysed on ice in standard IP buffer containing 1% Nonidet P-40 and protease inhibitors as previously described . IP was performed by the addition of either 1 or 2 µg of affinity-purified commercial antibody or 5 µl of Dok-4 antiserum followed by 8 µl of protein A-agarose beads (Santa Cruz). For monoclonal antibodies, protein G-agarose beads (Santa Cruz) were used. Immunoprecipitates were eluted in 1× Laemmli buffer, boiled and resolved on an SDS–polyacrylamide gel.
GST pull-down assays
The fusion proteins were induced in Escherichia coli by the addition of 1 mM isopropyl-1-thio-β-d-galactopyranoside for up to 4 h. After lysis and sonication in IP buffer containing protease and phosphatase inhibitors, GST or GST fusion protein was coupled to glutathione–agarose beads (Thermo Scientific) in the presence of 5 mM dithiothreitol. Protein concentration was estimated by resolving aliquots on SDS–polyacrylamide gels and staining with Coomassie Blue. The binding assays were performed by incubating lysates of transiently transfected COS-1 or 293HEK cells with 25 µl of GST or GST fusion protein-coupled beads for 3 h at 4°C. After five washes, proteins were eluted in 1× Laemmli buffer, boiled and loaded onto SDS–polyacrylamide gels. Immunoblots were carried out as described above. In the majority of experiments, comparable loading of GST fusion proteins was ascertained by Ponceau stain or anti-GST immunoblotting.
293HEK and MDCK cells were transiently transfected in triplicate on 24-well plates. In addition to Dok-4 and/or kinase expression constructs, the pFR-Luc vector (Gal-4/luciferase) was transfected in combination with pFA2-Elk-1 CT, pFA2-Elk-4 CT or pFA2-Elk-4 ΔEts (see above). Transfection with pFR-Luc and pFC-dbd (Gal4 DNA-binding domain alone) was used to determine background transactivation. The phRL-null Renilla luciferase vector (Promega) was included for normalization. Empty vector was added where necessary to equalize the amounts of transfected DNA. Cells were lysed 48 h after transfection and luciferase activity was determined using the Dual-Luciferase Reporter system (Promega) on a Lumat LB 9507 luminometer (Berthold) and expressed as the ratio of firefly to Renilla luciferase activities (measured in relative light units or RLU) ± standard deviation.
293HEK cells plated on fibronectin-coated coverslips were transfected with expression plasmids using LipofectAMINE 2000 and coverslips were processed 48 h after transfection. Coverslips were washed in PBS, fixed in 4% paraformaldehyde and permeabilized in 0.1% Triton X-100. Cells were then blocked using 3% BSA and incubated in anti-Gal4 primary antibody (Santa Cruz) for 1.5 h at a dilution of 1:100 in 3% BSA. Rhodamine red-anti-mouse secondary antibody (Jackson Immunoresearch) was diluted at a concentration of 1:500 in 3% BSA. Following three washes, cells were stained with the nuclear stain DAPI (Molecular Probes) and mounted onto slides using AquaMount (Thermo Scientific). For experiments with MDCK EGFP-Elk-4 clone, cells were plated on laminin-coated coverslips and transfected with either control or Dok-4 siRNA (as described above). Following transfection, the cells were fixed in 4% paraformaldehyde, permeabilized in 0.1% Triton X-100 and blocked in 3% BSA. Cells were then stained with the nuclear stain DAPI and mounted onto slides using AquaMount. Immunofluorescent images were captured using a Zeiss LSM780 laser scanning confocal microscope.
Cell proliferation assays
Cell proliferation was measured by the MTT assay. MDCK cells were plated into 35 mm plates and treated with either a pool of siCONTROL or Dok-4-specific siRNA as described above. Twenty-four hours following transfection, cells were trypsinized, counted and replated at 0.75 or 1.5 × 104 cells per well in triplicate in a 96-well plate. Cells were allowed to adhere overnight and were either serum-starved or given new media containing 10% FBS the following morning, as indicated. Addition of MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide from Sigma) was done 72 h after siRNA delivery. MTT was left to incubate for 2.5 h at 37°C, at which point purple formazan salts were visible. The formazan salts were dissolved in acidified isopropanol (0.8% 5N HCl in isopropanol). Absorbance was measured at 550 nm.
For experiments in MDCK cells RNA was extracted using the RNeasy Mini Kit (Qiagen). For the mouse tissues RNA was extracted with the TRIZOL reagent (Invitrogen). Reverse transcription was performed using Quantitect Reverse Transcription (Qiagen). All gene-specific primers were obtained from Integrated DNA Technologies (IDT) and selected based on the IDT PrimerQuest design tool. PCR reactions were prepared in triplicate using the iTaq SYBR Green Supermix (Bio-Rad) and were performed with the Applied Biosystems 7300 Real-Time PCR System. For normalization, we used β-actin as housekeeping control. The relative quantification (RQ) of mRNA expression was determined using the 2−ΔΔCt method . Briefly, Ct values of the target genes were normalized to the housekeeping gene signal by calculating the difference between Cttarget and Cthsk (ΔCt = Cttarget − Cthsk). Then, the ΔCt values were adjusted to an internal calibrator (ΔΔCt = ΔCtsample − ΔCtcalibrator) arbitrarily set at 1. Based on the assumption that the amplified DNA doubles at each PCR cycle, the amount of target gene expression relative to both an endogenous control and an internal calibrator was expressed as RQ = 2−ΔΔCt. The sequences of our PCR primers are presented in Supplementary Table S1.
Identification of novel Dok-4 partners by Y2H screening
To identify new partners of Dok-4, we performed a Y2H screen using full-length Dok-4 (Figure 1A) as bait in a mouse kidney cDNA library. In two independent Y2H screens (with and without co-expressed Lyn), the top two candidate genes included the TCF Elk-4/Sap1 (Figure 1B). Elk-4 was represented by up to six independent fragments, all mapping to the CT region and including its transactivation domain. The shortest fragment recovered represented amino acids 348–430 of mouse Elk-4 (Figure 1C) with strong homology to the corresponding sequences of human Elk-4, partial homology to Elk-1 and 3 but little to no homology with hElk-4b (Figure 1D).
Dok-4 directly interacts with Elk-4.
Interaction of Dok-4 with Elk-4 was verified by co-immunoprecipitation (co-IP) in transiently transfected 293HEK cells (Figure 1E). Anti-Dok-4 antiserum could immunoprecipitate a Gal4 DNA-binding domain–Elk-4 CT fusion protein (Gal4-Elk-4 CT) in the presence (lane 2) but not in the absence (lane 1) of Dok-4, whereas normal rabbit serum could not immunoprecipitate Gal4-Elk-4 CT in either condition (lanes 3 and 4). This was further validated by the GST pull-down assay (Supplementary Figure S1A), where GST-Dok-4(100–325) but not GST alone could bind full-length Elk-4, while neither GST construct could bind Elk-4 with a deleted CT region (Elk-4 ΔCT).
Structural basis of Dok4-Elk-4 interaction
To more precisely determine the structural basis of this interaction, we tested the ability of various GST- or Myc/His-tagged Dok-4 constructs to associate with the CT region of Elk-4 (Elk-4 CT) in both GST pull-down assays and co-IP experiments. Confirming the results obtained with GST-Dok-4, a GST-Elk-4 CT fusion protein was able to bind full-length wild-type (WT) Dok-4-Myc, whereas GST alone was not (Figure 2A, lanes 3–4). The Dok-4/Elk-4 binding was even more striking when the PH domain was deleted from Dok-4 (lanes 5–6). In contrast, binding was greatly diminished when the CT region of Dok-4 and a portion of the PTB domain were deleted (lanes 7–8), and it was abolished when a wider deletion involving the PTB domain of Dok-4 was introduced (lanes 9–10).
Binding of Dok-4 to Elk-4 involves an unconventional PTB domain interaction.
Because our recent studies had indicated that the Dok-4 PTB domain extends C-terminally beyond its previously ascribed boundaries , we tested the interaction of Elk-4 with a more rigorously defined Dok-4 PTB domain construct. Interestingly, as shown in Figure 2B, while GST-Dok-4 PTB(131–246), which encompasses the entire PTB domain, could bind to Elk-4 (lane 3), it did so much less efficiently than a construct comprising amino acids 100–325 of Dok-4 (lane 2). Similar results were obtained in co-IP experiments in the presence of activated Fyn (Supplementary Figure S1B). This was further confirmed with GST-Elk-4 CT pull-down assays, which showed equal binding of Elk-4 to Dok-4(100–325), Dok-4(100–264) and Dok-4(131–325) but reduced binding to Dok-4(100–246) (Figure 2C). Taken together, these results suggest that sequences located C-terminally to the PTB domain (within aa 247–264) contribute to the interaction with Elk-4. This contrasts with our recent findings regarding PTB-mediated binding of Dok-4 to Ret and Ship1 where amino acids 131–246 confer optimal binding .
To determine whether this atypical PTB domain binding might still involve a canonical PTB-binding motif in Elk-4, we looked for candidate sequences in the Elk-4 CT region (aa 348–430). The absence of any tyrosine residue within that sequence ruled out canonical NPxY motifs or other tyrosine-based motifs as targets of PTB binding. However, an alternate PTB binding has been identified as NxxF in the case of the Numb PTB domain [38,39]. One such sequence, NTLF (aa 395–398), exists in Elk-4 (Figure 1D). However, an N395A mutation of this putative binding motif had minimal impact on binding to GST-Dok-4 (Figure 2D). Compared with Elk-4, Elk-1 differs substantially in the primary sequence of its CT region and it lacks any NxxF or NxxY motif (Figure 1D). Nevertheless, weak binding of Gal4-Elk-1 to Dok-4 could be observed (Figure 2D). Thus, Dok-4 does not bind Elk-4 (or Elk-1) through a canonical PTB motif.
We have recently shown that mutation of a highly conserved glycine residue in the Dok-4 PTB domain (G207A) prevents Dok-4 phosphorylation by Ret, suggesting that it disrupts PTB domain function . To determine if structural integrity of the PTB domain was needed to bind Elk-4, we introduced the G207A mutation in the Dok-4(100–325) construct. As shown in Figure 2E, the mutation had no impact on Elk-4 binding. Similar results were obtained by co-IP using either full-length Dok-4 G207A or truncated Dok-4(100–325)/G207A (Supplementary Figure S1C). The G207A mutation also failed to reduce Elk-4 binding when introduced into a suboptimally binding Dok-4(100–246) construct (Supplementary Figure S1D). Thus, while binding of Dok-4 to Elk-4 appears to require sequences within the PTB domain as well as the adjacent CT region, it does not require a canonical PTB domain fold.
The CT region of Elk-4 promotes nuclear exclusion in the presence of Dok-4
Since Dok-4 is constitutively associated with the cell membrane, whereas Elk-4 is a nuclear transcription factor, we wondered how Dok-4 might influence Elk-4 localization. To examine this, the Elk-4 CT region was fused to the Gal4 DNA-binding domain and the NLS. The resulting construct was expressed in 293HEK cells with either enhanced green fluorescent protein alone (EGFP) or with Dok-4-EGFP and analyzed by confocal microscopy. As shown in Figure 3A, Gal4-Elk-4 CT was concentrated in the nucleus in the presence of EGFP. However, in the presence of Dok-4-EGFP, Gal4-Elk-4 CT was redistributed away from the nucleus and co-localized with the punctate membrane distribution of Dok-4 (Figure 3B,C). In contrast, Dok-4 caused no nuclear exclusion of Gal4 alone or of Gal4-Elk-1 CT (Supplementary Figure S2). Thus, the Dok-4-interacting CT region of Elk-4 is essential for its Dok-4-mediated nuclear exclusion.
The Elk-4 CT region promotes nuclear exclusion of Elk-4 in the presence of Dok-4.
The PH domain and a nuclear export signal co-operate to exclude Dok-4 from the nucleus
We have previously shown that Dok-4 is constitutively localized at the cell membrane [7,33], so the physiological relevance of its interaction with a transcription factor primarily localized in the nucleus seemed questionable at first. However, we had noted that a point mutation at a highly conserved glycine residue in the Dok-4 (G12D), in addition to disrupting the usual punctate membrane distribution (Supplementary Figure S3B), resulted in weak nuclear expression of Dok-4 in 293HEK cells (Figure 3A). This was even more prominent in MDCK II cells (Supplementary Figure S3A).
Dok-1 and IRS-1 have been reported to translocate to the nucleus under specific conditions [40,41]. Dok-1 contains a nuclear export signal (NES) that normally maintains it in the cytosol . NESs are in most cases leucine-rich sequences that bind the Crm1 exportin in a RanGTP-dependent manner . Analysis of the Dok-4 amino acid sequence with the NetNES prediction tool (http://www.cbs.dtu.dk/services/NetNES/)  suggested the presence of a NES in the PH/PTB linker region (Figure 4B). To confirm its functionality, we mutated two of its four hydrophobic residues to alanine (L116,121A) in the Dok-4-EGFP G12D construct. When expressed in 293HEK cells, the resulting G12D/L116,121A construct was expressed exclusively in nuclei (Figure 4C). To confirm that nuclear localization was due in part to PH domain disruption rather than to an aberrant behavior unique to the G12D mutation, we mutated instead another highly conserved residue of the PH domain, tryptophan 103. The resulting W103A/L116,121A mutant was also localized almost exclusively in nuclei (Figure 4D). Similar levels of nuclear accumulation of Dok-4 were observed when Dok-4 G12D but not Dok-4 WT was treated with the Crm1 inhibitor LMB (Supplementary Figure S3B–E). This suggests that the PH domain and NES co-operate to exclude Dok-4 from the nucleus. Interestingly, the Dok-4 PH domain and NES also co-operated in promoting relocalization of mCherry-tagged Elk-4 CT (Figure 4E–G). In contrast with the truncated Gal4-Elk-4 CT and mCherry-Elk-4 CT constructs, full-length mCherry-Elk-4 showed minimal to no co-localization with Dok-4 WT (Figure 5A), suggesting that N-terminal sequences in Elk-4 limit either the interaction or nucleocytoplasmic shuttling. However, bimolecular fluorescence complementation, which more directly detects interactions and tends to stabilize those that are transient , demonstrated interaction of full-length Dok-4 and Elk-4 proteins in a punctate membrane compartment (Supplementary Figure S4A). However, when a nuclear-localized Dok-4 mutant was used, complementation occurred exclusively in the nucleus (Supplementary Figure S4B) and this was abolished when the Elk-4 C-terminus was deleted (Supplementary Figure S4C).
The PH domain co-operates with a NES to exclude Dok-4 from the nucleus.
NLSs in Elk-4 restrict co-localization with Dok-4.
Two putative lysine-rich NLSs were identified in the N-terminal region of Elk-4 using a combination of NLStradamus , cNLS Mapper  and the ELM database (http://elm.eu.org/) as predictive tools (Figure 5B). When key basic residues were mutated to glutamic acid in these two putative NLSs (K49,51,319,320E), a large portion of Elk-4 indeed relocalized to the cytosol in a homogeneous pattern (Figure 5C). However, when Dok-4-EGFP was co-expressed, the mCherry-Elk-4 NLS mutant adopted a more coarse distribution partially overlapping with that of Dok-4 (Figure 5D).
Dok-4 destabilizes Elk-4 protein
To our surprise, attempts to perform co-localization experiments and co-IP of Dok-4 WT with non-tagged Elk-4 WT consistently failed to produce any Elk-4 signal (data not shown). We eventually realized that, in 293HEK cells, non-tagged Elk-4 WT expression was dramatically reduced in the presence of Dok-4 but that this was completely reversed by treatment with MG-132 and not CQ (Figure 6A), suggesting that it involved proteasome-mediated degradation and not autophagy. This was confirmed with a more specific proteasome inhibitor, bortezomib (Supplementary Figure S5A). In sharp contrast with Elk-4 WT, an Elk-4 construct lacking the Dok-4-interacting CT region was either unaffected or increased in the presence of co-expressed Dok-4 (Figure 6B). Identical results were obtained with the human splice variant Elk-4b, which possesses a distinct CT sequence (see Figure 1D). Results were even more striking in MDCK II cells, which, unlike 293HEK cells, express endogenous Dok-4. Repeated attempts to stably overexpress Elk-4 or even EGFP-Elk-4 in these cells failed to yield adequate signal in immunoblotting or fluorescence localization experiments, but when stable clones were treated with MG-132, a dramatic expression of both Elk-4 and EGFP-Elk-4 was seen (Figure 6C).
Dok-4 promotes degradation of Elk-4 through the ubiquitin/proteasome system.
To examine if Elk-4 was subject to ubiquitination, we used an MDCK Elk-4 clone and immunoprecipitated Elk-4 followed by anti-ubiquitin immunoblotting. Under conditions of treatment with MG-132, a strong ubiquitin signal precipitated with anti-Elk-4 but not with control anti-Tie2 antibody (Figure 6D). Ubiquitination takes place on specific lysine residues that can be very difficult to predict. Moreover, lysine residues may also be subject to acetylation, which may exert complex interactions with ubiquitination that can modulate protein stability [47,48]. Indeed, results available from the PhosphoSitePlus database (http://www.phosphosite.org) suggested that Elk-4 contains acetylated and ubiquitinated lysines (data not shown). Taking these results into consideration and combining them with several available online predictive tools for ubiquitination (including in particular UbPred , UbiPred ) and acetylation , we selected nine lysine residues with a high probability of ubiquitination and low probability of acetylation and we mutated them to arginine. Strikingly, the resulting Elk-4K/R mutant, though yielding a lower basal signal than Elk-4 WT in 293HEK cells, showed a dramatic and dose-dependent increase in its expression in the presence of co-transfected Dok-4 (Figure 6E). In contrast, expression of Elk-4 WT was dose-dependently reduced in the same conditions.
To explore the possible role of endogenous Dok-4 in determining the dramatic instability of Elk-4 in MDCK II cells, we treated an EGFP-Elk-4 clone with control or dog Dok-4-specific siRNA. Dok-4 knockdown was accompanied by a dramatic increase in EGFP-Elk-4 expression (Figure 6F). When these cells were examined under confocal microscopy, nuclear expression of EGFP-Elk-4 was obvious in a large proportion of Dok-4 siRNA-treated cells, whereas in control siRNA-treated cells EGFP-Elk-4 was weak and limited to cytosolic granules or vesicles (Figure 6G and Supplementary Figure S5B–G).
The relevance of stably overexpressing Elk-4 for studies in MDCK cells was supported by the finding that epithelial tissues express high levels of Elk-4 protein, although corresponding cultured cell lines do not (Supplementary Figure S6 and data not shown).
Structural basis of Dok-4-mediated degradation of Elk-4
To determine the structural basis of Dok-4-mediated degradation of Elk-4, we first examined the role of the Dok-4/Elk-4 interaction. We identified a 12 aa Elk-4-to-Elk-1 CT substitution (E4/E1 mutant; see Materials and methods) that disrupted the interaction of Gal4-Elk-4 CT with Dok-4 (Supplementary Figure S7A). When this mutation was introduced in full-length Elk-4, it rendered it virtually insensitive to degradation induced by Dok-4 (Figure 7A). Similarly, both full-length Elk-1 and the Elk-4 splice variant Elk-4b, which contains an alternate C-terminus (Figure 1D), were insensitive to degradation induced by Dok-4 (Supplementary Figure S7B,C). Thus, interaction with Dok-4 contributes to degradation of Elk-4.
Structural basis of Elk-4 destabilization.
Next, we examined the role of the Dok-4 PH domain in Elk-4 destabilization. We had previously shown that inhibition of Elk-1 transactivation of Dok-4 in Caco-2 cells requires the intact PH domain. However, we had never examined whether the PH domain by itself exerts any downstream effect on signaling. Surprisingly, in 293HEK cells, suppression of Elk-4 expression was strikingly enhanced in the presence of the isolated Dok-4 PH domain (1–115) compared with Dok-4 WT. In contrast, a Dok-4 PH domain deletion mutant (100–325) was much weaker in suppressing Elk-4 expression and it mostly reduced expression of higher molecular mass forms of Elk-4. Similarly, the PH domain point mutant of full-length Dok-4, G12D, either had no effect at all or caused an increase in Elk-4 expression (Figure 7B). Notably, the destabilizing effect of the isolated Dok-4 PH domain on Elk-4 was dose-dependent and could be completely reversed by the G12D point mutation (Figure 7C). The effects of Dok-4 constructs were specific to Elk-4 because expression of hElk-4b and Elk-1 was unaffected by any of these destabilizing Dok-4 constructs (Supplementary Figure S7). Using EGFP-tagged Dok-4 constructs, the G12D PH domain mutation again abolished Dok-4-mediated degradation of Elk-4 and the addition of a NES mutation (L116,121A), which restricts Dok-4 to the nucleus (Figure 4), did not enhance the destabilizing activity of Dok-4 on Elk-4. In contrast, EGFP-Dok-4(100–325), which lacks the PH domain, caused disappearance of high molecular mass forms of Elk-4 and lower overall Elk-4 expression, but expression of Dok-4(247–325), lacking both the PH and PTB domain, had no impact on Elk-4 expression (Figure 7D). Thus, the PH domain of Dok-4 is sufficient to promote degradation of Elk-4 despite its apparent inability to directly bind Elk-4. However, the Dok-4-interacting Elk-4 CT region is also essential for Elk-4 degradation and its interaction with a strongly binding Dok-4 PH deletion mutant can also promote partial degradation. Therefore, Dok-4 contributes to Elk-4 degradation through both an interaction-dependent and an interaction-independent, PH domain-mediated effect.
Although virtually nothing is known about the regulation of TCF expression and its mechanisms, it has recently been shown that Elk-1 is rapidly degraded by the proteasome in the cytosol but not in the nucleus . To determine if Elk-4 is subject to similar regulation, we first examined the role of Elk-4 cytoplasmic relocalization in its degradation. As shown in Figure 7E, in contrast with Elk-4 WT, which was highly expressed but was subject to destabilization in the presence of Dok-4, the non-tagged Elk-4 NLS mutant was expressed only very weakly and was insensitive to further destabilization by Dok-4. Furthermore, whereas Elk-4 WT was unaffected by proteasome inhibition in the absence of Dok-4, expression of the Elk-4 NLS mutant was greatly increased following proteasome inhibition (Figure 7F). These results suggest that, like Elk-1, Elk-4 is highly sensitive to proteasomal degradation, independently from Dok-4. However, the co-ordinated action of the Dok-4 PH domain and PTB/CT region can destabilize nuclear Elk-4, probably by promoting its nuclear exclusion.
Dok4 represses IEG transcription
In view of the above findings, and since our antibodies were unable to detect endogenous dog Elk-4 in MDCK cells, we wondered if knockdown of endogenous Dok-4 would impact Elk-4/TCF-mediated transcription. We therefore used qPCR to assess mRNA expression of the above two IEGs as well as cyclin d1 and the cyclin-dependent kinase inhibitor p27waf1 in parental MDCK II cells treated with control or Dok-4-specific siRNA and stimulated with EGF. As shown in Figure 8, Dok-4 siRNA was accompanied by an at least 60% reduction in endogenous Dok-4 mRNA. This knockdown was associated with a marked increase in expression of Fos and Egr-1 mRNA both basally and 1 h after EGF stimulation (Figure 8A,B). Similarly, Cyclin D1 mRNA was dramatically up-regulated by Dok-4 knockdown, especially under non-stimulated conditions (Figure 8D). In contrast, basal and EGF-induced expression of p21Waf1 mRNA was inhibited by Dok-4 knockdown (Figure 8E).
Dok-4 knockdown enhances IEG expression and cellular proliferation.
Dok-4 inhibits cell proliferation while stimulating Erk activity
To determine if up-regulated IEG and cyclin d1 expression induced by Dok-4 knockdown had an impact on cellular proliferation, we performed MTT assays on MDCK II cells in both serum-free and FBS-treated conditions. As shown in Figure 9A, cell proliferation was enhanced under both conditions. To determine if the observed enhancement of this proliferative response and the upstream enhancement of IEG expression could be explained by an increase in Erk activity, we analyzed mitogen-activate protein (MAP) kinase activity by immunoblotting with phospho-specific antibodies. Unexpectedly, Dok-4 knockdown was associated with decreased basal and EGF-induced activity of Erk1/2 and c-Jun kinase but with increased activation of p38 (Figure 9B). Thus, Dok-4 appears to exert opposing effects on Erk activation and downstream Elk-4/TCF-mediated transcription.
Dok-4 knockdown enhances cell proliferation and modulates MAP kinase activation.
Dok 4 inhibits Elk-4 transactivation
We have previously shown that, in Caco-2 cells, Dok4 can inhibit the transactivation of Elk-1 , in part through inhibition of Erk activation . To determine if Dok4 modulates transactivation by Elk-4, we used a gal4-luciferase reporter and compared its transcriptional response to Gal4-Elk-1 CT and Gal4-Elk-4 CT chimeras following expression of activated Fyn. Notably, the N-terminal boundary of the Elk-4 chimera was chosen to match exactly that of the Elk-1 chimera, based on sequence alignment (data not shown). In 293HEK cells, Gal4-Elk-4 CT displayed extremely low basal transactivation and dramatic activation in response to Fyn (Figure 10A). This was potently inhibited in a dose-dependent manner by Dok-4. By comparison, Gal4-Elk-1 showed slightly higher basal transactivation and significantly lower induction in response to Fyn. Thus, although absolute levels of Elk-1 and Elk-4 transactivation were comparable in the presence of Dok-4, in relative terms, a much more robust inhibition of Elk-4 than that of Elk-1 was seen. Interestingly, compared with Dok-4 WT, nuclear-localized Dok-4 G12D/L116,121A had a reduced impact on Gal4-Elk-1 transaction but an increased inhibitory effect on Gal4-Elk-4. Since our Gal4-Elk-4 reporter constructs were insensitive to degradation induced by Dok-4 (Supplementary Figure S8) and since Dok-4 G12D/L112,116A resides almost exclusively in the nucleus and cannot displace Elk-4 from this compartment, this suggests that binding of Dok-4 to the Elk-4 transactivation domain probably contributes to the inhibitory action of Elk-4, independently of other inhibitory mechanisms involving Elk-4 destabilization or relocalization.
Dok-4 inhibits Elk-4 transactivation.
We extended transactivation assays in MDCK II cells, which express endogenous Dok-4. As shown in Figure 10B, basal transactivation of Elk-4 was slightly higher than transactivation of Elk-1, but the response to activated Fyn was again more pronounced compared with Elk-1. This was completely inhibited by overexpression of Dok-4. To ensure that the observed inhibition of Elk-4 transactivation by Dok-4 was not limited to a heavily truncated transactivation domain chimera, we constructed a Gal4-Elk-4 chimera where only the DNA-binding Ets domain was deleted. When expressed in MDCK II cells, this Gal4-Elk-4 ΔEts construct displayed lower overall transactivation compared with Gal4-Elk-4 CT, but it was similarly activated by Fyn and repressed by Dok-4 (Figure 10C).
Since MDCK cells express endogenous Dok-4, we examined the impact of Dok-4 knockdown on transactivation of a transiently transfected Elk-4 reporter. Dok-4 siRNA dramatically enhanced both basal and Fyn-induced transactivation of Elk-4 (Figure 10D). Therefore, Dok-4 is a powerful inhibitor of tyrosine kinase-induced Elk-4 transactivation in both 293HEK and MDCK cells and this effect correlates with transcriptional repression of TCF target genes in MDCK cells.
In the current study, we show that a sequence overlapping the Dok-4 PTB domain and CT region binds the transactivation domain of Elk-4. We also demonstrate that Dok-4 contains a NES that, together with the PH domain, mediates nuclear exclusion of Dok-4 and Dok-4-interacting Elk-4. Although Dok-4 interaction with Elk-4 is restricted by two Elk-4 NLSs, Dok-4 strongly promotes destabilization of the Elk-4 protein through both an interaction-dependent mechanism and a PH domain-dependent but interaction-independent mechanism. Elk-4 destabilization involves ubiquitination and proteasome-mediated degradation. In MDCK cells, endogenous Dok-4 not only precludes forced overexpression of Elk-4 but also inhibits Egr1, Fos and cyclin D1 expression as well as cell proliferation despite promoting Erk activation. In short, Dok-4 has evolved an array of mechanisms that directly target Elk-4 activity, thus potentially uncoupling it from Erk activity. This could explain in part why Dok-4 seemingly exerts opposite effects on Erk signaling depending on specific cellular contexts and readout methodology.
Although direct interaction allows Dok-4 to more directly target Elk-4, the effects of endogenous Dok-4 on gene expression and cell proliferation observed in our study may be mediated through molecules other than Elk-4, such as Elk-1 [7,32], which is highly expressed in MDCK cells (data not shown). Indeed, Dok-4 inhibits Elk-1 transactivation — albeit more modestly than it does for Elk-4 (Figure 10A,B) — a phenomenon that our current results cannot yet explain, since Dok-4 interacts only weakly with Elk-1 and does not impair its stability. Thus, additional mechanisms of Dok-4-mediated inhibition remain to be discovered. This further highlights the multilayered nature of Dok adapter function .
Unconventional PTB domain binding by Dok-4
While the interaction between Dok-4 and the transactivation domain of Elk-4 requires sequences contained in the PTB domain, it is enhanced by sequences contained in the Dok-4 CT region and it is virtually unaffected by a PTB-disrupting point mutation (Dok-4 G207A) or by mutation of the only recognizable PTB-binding motif in the Elk-4 C-terminus (N395A). This suggests that Elk-4 does not bind within the usual PTB domain pocket, but instead makes contacts with a distinct interface. This would be consistent with the notion that the β-sheet sandwich structure of PTB and PH domains is poised to bind targets using multiple interfaces . Indeed a similar, unconventional PTB interaction has recently been shown to explain the binding of the Tensin2 PTB domain to DLC1 .
Dok-4 nucleocytoplasmic shuttling
The presence in Dok-4 of a functional NES adds to the emerging concept of nucleocytoplasmic shuttling by Dok and IRS family adapter proteins. A NES has previously been identified in Dok-1 and its function appears to be regulated by Src family kinases . Nuclear localization of Dok-1 may be a regulatory mechanism that prevents it from exerting its inhibitory functions in the cytosol . In contrast, IRS-1 translocates to the nucleus where it associates with the promoters of myc and cyclin d1 and enhances transcription of the corresponding mRNAs [41,55]. In the case of Dok-4, the mechanisms of nuclear entry and the exact role of NCS remain to be determined. Since nuclear Dok-4 G12D/L116,121A is more potent than Dok-4 WT in suppressing Gal4-Elk-4 transactivation (Figure 10A), nuclear entry of Dok-4 probably facilitates interaction with nuclear Elk-4. However, physiological conditions under which detectable amounts of WT Dok-4 might accumulate in the nucleus remain undefined.
Regulation of Dok-4 action
The present study was focused on the impact of altered Dok-4 expression levels on Elk-4 and related downstream events. However, we expect that post-translational modification events yet to be defined and characterized will be found to influence either Dok-4 expression, conformation, interactions or subcellular localization.
Regulation of Elk-4 expression and activity
Very little is known about the mechanisms regulating TCF expression. Indeed, the current study is, to our knowledge, the first to demonstrate and characterize the highly unstable nature of Elk-4 protein. However, a recent study has shown that cytosolic Elk-1 is rapidly degraded by the proteasome unless it can undergo dimerization-mediated stabilization . Elk-4 could be subject to similar regulation since mutation of its two putative NLSs resulted in a dramatically unstable protein. More studies will be needed to confirm this possibility and to determine how the Dok-4 PH domain might influence this process, independently of any direct interaction.
Fusion to an N-terminal GFP or mCherry tag dramatically stabilized Elk-4 in 293HEK cells. The reason for this remains unclear but an N-terminal tag might influence the way in which Elk-4 degrons are exposed, especially if they involve the N-end rule pathway . The EGFP tag could also impair nuclear export of Elk-4, which might indirectly retard Elk-4 degradation. However, the instability of EGFP-Elk-4 in MDCK cells suggests that cell context-specific factors contribute to the mechanisms regulating Elk-4 stability.
In any event, regulation of Elk-4 protein stability represents a novel component of a multilayered system geared to attenuate TCF and Elk-4 activity. In addition to repression of TCF DNA-binding through intramolecular interactions  and interaction with Id family proteins , reversible SUMOylation can sequester TCFs in the cytosol [59,60]. Elk-4 mRNA expression may also be regulated by antisense transcripts of the overlapping gene Mfsd4 . Finally, several heavily truncated Elk-4 mRNA transcripts have been shown to exist , which might possess dominant negative actions.
Antiproliferative action of Dok-4
Knockdown of endogenous Dok-4 in MDCK cells resulted in enhanced basal and EGF-induced mRNA levels of the IEGs Egr-1 and Fos together with Cyclin D1, a known target of both Egr-1 and Fos [22,62]. In contrast, basal levels of the cdk inhibitor p21Waf1 were reduced in cells treated with Dok-4 siRNA. The observed suppression p21Waf1 expression was somewhat surprising, given that Egr-1 has been reported to contribute to p21Waf1 induction in some contexts [63,64], but it nevertheless probably contributed to enhanced proliferative response . The mechanism of this differential regulation of cyclin D1 and p21Waf1 remains to be defined, but signals emanating from RhoA have previously been shown to exert opposing effects on cyclin D1 and p21Waf1 , a possibility that will require direct attention in future studies of Dok-4.
We have shown a novel mechanism for the inhibition of tyrosine kinase signaling by a Dok family adapter protein involving direct interaction of Dok-4 with the transcription factor Elk-4. Elk-4 is exquisitely sensitive to activation by Src kinase signaling, and Dok-4 potently inhibits this event and its downstream consequences. We suggest that inhibition of Elk-4 by Dok-4 occurs through a combination of interaction-dependent and interaction-independent mechanisms that include relocalization, destabilization and, possibly, direct interference with transactivation domain function. We speculate that interaction of Dok-4 with the Elk-4 C-terminus induces a conformational change that exposes N-terminal degrons, while cytosolic relocalization exposes Elk-4 to ubiquitylating enzymes. These mechanisms allow uncoupling of cytosolic Erk functions from the downstream Elk-4-mediated proliferative response and may represent an important regulatory mechanism to establish a balance between the Erk-mediated proliferative transcriptional response in the nucleus and the differentiation-promoting effects exerted by Erk in the cytosol. One of the main challenges ahead will be to identify signaling events or physiological conditions that modulate the action of Dok-4 on Elk-4 through altered expression, subcellular localization, Elk-4 binding or formation of complexes with additional partners.
downstream of kinase
epidermal growth factor
enhanced green fluorescent protein
fetal bovine serum
Integrated DNA Technologies
immediate early genes
Madine-Darby canine kidney
nuclear export signal
nuclear localization signal
serum response factor
ternary complex factor
E.H. performed the experiments, analyzed data and co-wrote the manuscript. C.B. performed the experiments. A.B. performed the experiments. N.N.I. performed the experiments. T.T. obtained funding, analyzed data and co-wrote the manuscript. S.L. obtained funding, designed the experiments, analyzed data and co-wrote the manuscript.
This work was supported by grants to S.L. from the Kidney Foundation of Canada and to T.T. from the Canadian Institutes of Health Research [MOP-126180].
We thank Drs Marcela Nunez, Jean-Marie Tantôt and the rest of the Hybrigenics team for Y2H screening service; Dr Min Fu for assistance with confocal microscopy; and Drs Ursula Stochaj and Edward Fon for helpful discussions.
The Authors declare that there are no competing interests associated with the manuscript.