Disruption of TWIST1-RELA binding by mutation and competitive inhibition to validate the TWIST1 WR domain as a therapeutic target

HEK-293 cells were grown in McCoy’s 5A medium supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin (P/S). Ovcar4 cells were grown in RPMI medium with 10% FBS and 1% P/S. All cells were maintained at 37 °C and 90% humidity in a tissue culture incubator with 5% CO₂ atmosphere. Cells were passaged every 2–4 days as they became confluent, using 0.25% trypsin. Where indicated, cells were transfected using 5 μL per well Lipofectamine 2000 (Life Technologies, Carlsbad, CA) in a total of 2 mL per well of OptiMEM low serum medium (Life Technologies). Cycloheximide (CHX) was obtained from Sigma Aldrich (St. Louis, MO) and used at a dose of 20 μg/ml. For CHX studies, cells were transfected using XtremeGene 9, also from Sigma Aldrich. For proteasome inhibition studies, MG132 was added to HEK-293 cells in normal medium four hours after transfection and left on overnight. A dose of 5 μM was used for fractionated western studies and 1 μM was used for luciferase assays.

The cloning of TWIST1 into the pcDNA4-MycHis vector has been described previously [18]. The wild type RELA gene was also cloned into pcDNA4-MycHis, including a stop codon at the C-terminus to prevent translation of the Myc-His tag. TWIST1 retained the tag. Amino acid substitution and truncation mutations were introduced using the QuikChange II site directed mutagenesis kit (Agilent, Santa Clara, CA) according to the manufacturer’s instructions and following their recommendations for primer design. Silent mutations were introduced in tandem with the desired mutations in order to create or eliminate restriction sites to facilitate screening for mutants. All mutations were confirmed by Sanger sequencing by the City of Hope Integrative Genomics Core.

In order to create a competitive inhibitor for TWIST1-RELA binding, the WR domain from TWIST1 was fused to eGFP. Briefly, PCR was used to amplify the final 63 nucleotides of the TWIST1 gene (including stop codon) and add 5’ XhoI and 3’ BamHI restriction sites. The PCR fragment and the pEGFP-C3 vector were subjected to XhoI-BamHI double digest (New England BioLabs, Ipswich, MA) and the two fragments ligated together. GFP lacking the WR domain was used as a control, and includes 21 residues at the C-terminus encoded by the multiple cloning site of the vector. As a result, the molecular weights of the two GFP proteins are indistinguishable on a western blot. To achieve equal expression of GFP-WR compared to unmodified GFP, it was necessary to transfect cells with three-fold more GFP-WR plasmid versus GFP. A one to one ratio was sufficient for CoIP illustrated in Fig. 4C. For all GFP-WR experiments, 4x refers to GFP-WR only, 3x to a 3:1 ratio of GFP-WR to GFP, 2x to equal amounts of both, 1x to a 1:3 ratio of GFP-WR to GFP, and 0 to GFP only.

HEK-293 cells were plated at 500,000 cells per well, in 2 mL normal medium, in a 6 well plate and allowed to adhere. The next day, medium was replaced with OptiMEM low serum medium (Life Technologies). Cells were transfected with various alleles of TWIST1, RELA, and GFP using Lipofectamine 2000 (Life Technologies). The following day, cells were detached using trypsin, washed with PBS, and pelleted. Cell pellets were lysed in RIPA buffer, and protein concentration was determined by BCA Protein Assay (Thermo Fisher, Waltham, MA). 50–100 μg total protein (equal between conditions) was pre-cleared by incubating with 1 μg normal rabbit IgG (Santa Cruz Biotechnology, Dallas, TX) and 20-30 μL Protein A/G Agarose beads (Santa Cruz Biotechnology, sc-2003) on a rocker at 4 °C for 1 h. Water was added to equalize volumes across conditions. Beads were centrifuged for 1 min at 3,000 rpm, and equal volumes of supernatant from each condition were transferred to new tubes, and incubated with 1 μg rabbit anti-RELA (Santa Cruz Biotechnology sc-109) or rabbit anti-GFP (Santa Cruz Biotechnology, sc-8334) antibodies on a rocker at 4 °C. After 1 h, 20–30 μL (equal between conditions) Protein A/G Agarose beads were added to each tube, and tubes were returned to the rocker at 4 °C overnight. The following day, unbound protein was removed and beads were washed five times with 1 mL PBS. Beads were boiled in 20 μL 2x loading dye to release bound protein. Equal masses of input and equal volumes of immunoprecipitated protein were used for western blotting.

HEK-293 cells were plated and transfected as described for co-immunoprecipitation above. The following day, cells were detatched using trypsin, washed with PBS, and pelleted. Pellets were resuspended in 100 μL hypotonic buffer (10 mM HEPES, 10 mM KCl, 0.1 mM EDTA, 1 mM Na₃VO₄, 1.25 mM NaF, 0.4% IGEPAL, 0.5 mM DTT) in the presence of protease inhibitor (Thermo Fisher, Waltham, MA). Cells were left on ice 15 min to swell, and then lysed by addition of NP-40 to a final concentration of 0.1%. Nuclei were separated from cytoplasmic lysate by centrifugation (3000 rpm, 10 min, 4 °C) and washed once in hypotonic buffer without NP-40. Nuclei were then resuspended in 50 μL high salt buffer (20 mM HEPES, 400 mM NaCl, 1 mM EDTA, 10% glycerol, 1 mM Na₃VO₄, 1.25 mM NaF, 0.5 mM DTT) plus protease inhibitor. Vials were shaken for 2 h at 250 rpm at 4 °C, and then centrifuged (5 min, 14,800 rpm, 4 °C). NaCl concentration was adjusted to 137 mM by addition of water prior to western blotting.

HEK-293 cells were plated at 150,000 or 250,000 per well in 12 well plates and allowed to adhere overnight. The following day, cells were transfected as described for the above procedures. On the third day, non-treated cells were harvested and cycloheximide was added to the remaining wells. Remaining treated cells were harvested at the indicated time points and used for western blotting.

Protein was run on 10% resolving polyacrylamide gels and transferred to PVDF membrane. Membranes were rinsed with PBS and blocked in 5–10% milk, 1 h at room temperature or overnight at 4 °C. Membranes were then incubated with mouse primary antibody in milk with Tween-20 (Ab Buffer) for 1 h at room temperature or overnight at 4 °C, and washed in PBS with 0.1% Tween-20 (PBST). Membranes were then incubated with anti-mouse secondary antibody in Ab Buffer for 1 h at room temperature, followed by an additional five PBST washes. Primary antibodies were: TWIST1, TWIST 2c1a (Santa Cruz Biotechnology sc-81417) 1:250-1:500; for RELA, NF-κB p65 F-6 (Santa Cruz Biotechnology sc-8008) 1:250-1:500; for GFP, GFP B-2 (Santa Cruz Biotechnology sc-9996) 1:1000; for actin, Sigma Aldrich A1978 or 2066. Secondary antibodies were HRP conjugated anti-mouse and anti-rabbit. Protein was detected using Blue Devil Film (Genesee) and ECL Plus (Thermo Fisher) or digital imaging. Quantitation of digital images was performed using the accompanying software from Syngene (Frederick, MD) or Carestream MI (Woodbridge, CT).

Ovcar3 and Ovcar4 cells were plated at 50,000 or 75,000 cells per well, in 500 μL RPMI, in a 24 well plate and allowed to adhere overnight. Ovcar4 cells were used for all luciferase assays except for that shown in Additional file 1: Figure S1. The following day, cells were switched to OptiMEM medium and transfected using Lipofectamine 2000 at 2 μL per well. Plasmids were: TWIST1 in pcDNA4, RELA in pcDNA4, Renilla luciferase, and firefly luciferase (FFluc) in pGL3. FFluc was under the control of the IL-8 promoter; construction of this vector has been described previously [18]. Empty pGL3 lacking a promoter was used as a negative control for FFluc expression. Each condition was tested in triplicate. The day after transfection, luciferase expression was quantified using the Dual Luciferase Assay kit (Promega, Madison, WI) according to the manufacturer instructions.

HEK-293 cells were plated at 500,000 per well in glass bottom 35 mm cell culture dishes, and the next day were transfected with TWIST1, RELA, and GFP or GFP-WR as described above. After a further 24 h, cells were rinsed with PBS and stained for 15 min with DAPI. DAPI was then replaced with PBS. Images were captured using a Zeiss LSM700 Confocal Microscope and ZEN 2012 microscopy software (Zeiss AG, Oberkochen, Germany).

Western blots were quantified using GeneTools software. Data were graphed and analyzed in Microsoft Excel and GraphPad Prism 6, respectively. Luciferase assays were analyzed using one-way ANOVA with correction for multiple comparisons. For assay testing RELA mutants, all conditions were compared to all others. For assays testing TWIST1 mutants and GFP-WR inhibitor, positive control condition was compared to all others. Positive control conditions are indicated in each relevant figure. Cell counts were averages of four counts, and prior testing has demonstrated that the count is accurate to within 13%. All error bars represent standard deviation. *, p < .05; **, p < .01; ***, p < .001; ****, p < .0001 throughout.

Site-directed mutagenesis was used to generate mutations in the WR domain of TWIST1. On the basis of their high evolutionary conservation (Fig. 1c), we selected W190, R191, and E193 for mutation to alanine (W190A, R191A, E193A alleles, respectively). The ∆WR allele, in which all twenty amino acids of the WR domain have been deleted, was created previously as described elsewhere [18]. Mutants were screened by restriction digestion and confirmed by sequencing (data not shown). All alleles are shown schematically in Fig. 2a. In order to determine the contribution of individual amino acids in the WR domain to TWIST1-RELA binding, we transiently expressed RELA and all TWIST1 alleles in HEK293 cells and performed co-immunoprecipitation (CoIP). Following RELA pulldown, western blotting showed that as demonstrated previously, truncation of the entire WR domain reduced TWIST1 co-precipitation to basal levels. W190A, R191A, and E193A mutations reduced TWIST1 co-precipitation by 50-60%. A triple mutant with W190A, R191A, and E193A mutations also reduced RELA binding by 60%, with less variability (Fig. 2b-c).

We have previously established that formation of a TWIST1-RELA complex upregulates IL-8 expression by 2-2.5 fold over RELA alone, and that prevention of binding by truncating TWIST1 returns IL-8 expression to basal levels [18]. In order to determine the effect of W190A, R191A, and E193A mutations on IL-8 promoter activity, we performed a dual luciferase assay in which firefly luciferase (FFluc) was under the control of the IL-8 promoter. As expected, exogenous expression of RELA in Ovcar4 cells gave rise to a basal level of IL-8 driven FFluc, which was increased by co-expression of, and thus binding with, TWIST1 (Fig. 2d). Mirroring the phenotypes seen in our CoIP experiments above, W190A, R191A, and E193A mutations reduced expression of FFluc by 50%, and the triple mutant reduced FFluc expression a further 10–20% compared to the single point mutants (Fig. 2d). Similar results were obtained using the cell line Ovcar3 (Additional file 1: Figure S1), but a better range of IL-8 promoter induction was achieved in Ovcar4, and this line was selected for all subsequent functional assays.

While we have shown that the TWIST1 C-terminus is required for complex formation with RELA, the required residues of RELA remained unknown. In order to locate this site, we created a truncation mutant of RELA, ∆307 (Fig. 3a). Site directed mutagenesis was employed to insert a stop codon directly following the coding sequence for the REL homology domain, a well-conserved domain that has been structurally characterized [21]. CoIP of RELA revealed that truncation of RELA reduced co-precipitation of TWIST1 by approximately 90% (Fig. 3b-c). Truncating both proteins resulted in a greater loss of binding; under these conditions, only 1.86% of wild type levels of TWIST1 was detectable following CoIP (Fig. 3b).

In order to verify that loss of binding between RELA ∆307 and TWIST1 impacted IL-8 expression, we again utilized a dual luciferase assay. As expected, RELA truncation was able to reduce FFluc expression (Fig. 3d). However, this phenotype was independent of TWIST1; in the absence of TWIST1, RELA ∆307 produced only 30% of wild-type IL-8 promoter activity. TWIST1 expression upregulated IL-8-driven FFluc approximately two-fold, regardless of RELA status. As seen previously, co-expression of triple mutant TWIST1 with RELA led to an intermediate phenotype, for both WT and ∆307 alleles of RELA (Fig. 3d). Thus, we conclude that the domains required for both IL-8 transactivation and complexing with TWIST1 are contained within the relatively uncharacterized C-terminus of RELA.

Given the demonstrated role for the WR domain in RELA binding, as well as in the transcription factor activity of TWIST1 [17, 20], we propose that this domain is an attractive target for therapeutic intervention. To test whether the WR domain could act as a competitive inhibitor of TWIST1-RELA binding, the WR domain was fused to GFP in the pEGFP-C3 vector (Fig. 4a). Empty pEGFP-C3 encodes GFP followed by 21 residues encoded by the multiple cloning site. We therefore used this vector as a negative control, since its protein product would be of the same size as GFP-WR (Fig. 4a). Both forms of GFP could be expressed to similar degrees in HEK293 cells (Fig. 4b).

To determine the effect of GFP-WR on TWIST1-RELA binding, we performed CoIP analyses. Total GFP expression in transfected cells was held constant across all conditions by supplementing GFP-WR with control GFP. RELA pulldown revealed that levels of TWIST1 co-precipitated were reduced in a dose dependent fashion with increasing GFP-WR expression (Fig. 4c). GFP pulldown revealed that TWIST1 was co-precipitated in a dose-dependent manner with increasing GFP-WR expression (Fig. 4d). These findings suggest that GFP-WR is interacting with TWIST1 via WR-WR binding, an interaction illustrated by recent studies of higher order TWIST1 complexes [17]. In order to determine whether GFP-WR-mediated inhibition of TWIST1-RELA binding impacted downstream signaling, we again employed a dual luciferase assay to quantify IL-8 promoter activity. As expected, GFP-WR expression led to a dose-dependent reduction in FFluc expression (Fig. 4e). Thus, the TWIST1-driven IL-8 pathway can be inhibited by direct competition using the WR domain.

As GFP-WR was primarily expressed in the cytoplasm of transfected cells (Additional file 1: Figure S2), we hypothesized that GFP-WR was sequestering TWIST1 in the cytoplasm. In order to test this hypothesis, we isolated cytoplasmic and nuclear cell fractions and analyzed the levels of TWIST1 found in each. Western blot of fractionated cells showed that in both cytoplasmic and nuclear fractions, the protein levels of TWIST1 and GFP decreased as the proportion of GFP-WR transfected was increased (Fig. 5a). This suggested that rather than sequestration, the interactions between these proteins may lead to their degradation, as seen previously following altered binding of TWIST1 to partner proteins [19, 22]. In order to test this hypothesis, we transfected HEK-293 cells with TWIST1, RELA, and either GFP or GFP-WR and after 24 h, treated with cycloheximide (CHX) to prevent further protein production. TWIST1 was degraded more quickly in cells expressing GFP-WR than in those expressing GFP (Fig. 5b), suggesting that GFP-WR leads to TWIST1 turnover. To determine if this process was dependent on proteasomal activity, we transefcted HEK-293 cells and treated them with the proteasome inhibitor MG132 overnight. Western blots show that MG132 was able to increase the levels of TWIST1 and GFP by up to two fold in the cytoplasmic fraction of these cells (Fig. 5c). Finally, in order to determine the effect of proteasome inhibition on IL-8 promoter activity, a dual luciferase assay was once again employed. Treatment with MG132 following GFP-WR expression increased IL-8 promoter activity two fold, correlating with increased TWIST1 expression observed following MG132 treatment (Fig. 5d).