Impact of carbamazepine on SMARCA4 (BRG1) expression in colorectal cancer: modulation by KRAS mutation status

Two CRC cell lines, HCT116 and Hke3, were used in this study. HCT116 is a KRAS mutant cell line, harboring a G13D mutation, while Hke3 is KRAS WT. HCT116 and Hke3 are isogenic, and Hke3 was originally made by reverting the KRAS mutation in HCT116 [18]. HCT116 were purchased from the American Type Culture Collection (ATCC®). Hke3 cell lines were obtained from Dr. Takehiko Sasazuki (Medical Institute of Bioregulation, Kyushu University).

Cell lines were cultured in Roswell Park Memorial Institute (RPMI) 1640 media (Gibco™, Catalog #: 11875093), with 10% Fetal Bovine Serum (GemCell™, Catalog #: 100–500), 1% Non-Essential Amino Acids (Gibco™, Catalog #: 11140050), 2% HEPES buffer (Gibco™, Catalog #: 15630080), 1% Antibiotic–Antimycotic (Gibco™, Catalog #: 15240062), and 0.4% gentamicin (Gibco™, Catalog #: 15710064). The cells were maintained in an atmosphere of 5% CO2 at 37 °C and passaged according to ATCC®’s recommended protocol.

Carbamazepine powder was purchased from (Supelco™, Catalog #: PHR1067-1G). The CBZ powder was dissolved in absolute methanol at a concentration of approximately 2 mg/ml (8.5mMol) and placed into single-use aliquots at -20 °C. At the time of treatment one aliquot of the Carbamazepine solution was diluted with media to 500uM.

Cells were cultured until 70% confluency, trypsinized (Corning™, Catalog #: 25–053-CI), and spun down into cell pellets. Cells were then counted using the Countess™ II Automated Cell Counter (Invitrogen™, Catalog #: AMQAX1000) with Trypan Blue solution (Sigma-Aldrich™, Catalog #: T8154) as per the manufacturer’s protocol. Four plates of 5–7 million cells were made in a 100 mm plate (Denville™), two HCT116 and two Hke3, and allowed to remain for 24 h in 9 mL of cell culture media. After that time, one plate of each cell type was then treated with 1 mL of the 500uM Carbamazepine solution (final concentration in media = 50uM). 6 or 24 h after treatment, both the untreated and treated cell lines were trypsinized and harvested. 25% of the pellet was set aside for RNA extraction, while 75% was set aside for protein extraction.

RNA was extracted from the cell pellets using the Invitrogen™ PureLink™ RNA Mini Kit (Invitrogen™, Carlsbad, CA, USA, Catalog #: 12183018A) as per the manufacturer’s protocol. The purified RNA was then placed into single-use aliquots and stored at − 80 °C. The concentration of the extracted RNA was quantified using a Thermo Scientific™ NanoDrop 1000 (Thermo Scientific™, Catalog #: 2353–30-0010). The 260/280 of the RNA was checked and the RNA was only kept if the range of 260/280 was between 1.9 and 2.1.

A total of 1.5 μg of the extracted and quantified RNA was synthesized into cDNA using the iScript Reverse Transcription Supermix for RT-qPCR (Bio-Rad™, Catalog #: 1708841) as per the manufacturer’s protocol. A T100™ Thermal Cycler (Bio-Rad™, Catalog #: 1861096) was used to run the reaction. 100 μL of DEPC-treated water (Thermo Scientific™, Catalog #: R0601) was then added to each sample. The synthesized cDNA was then estimated using the Thermo Scientific™ NanoDrop 1000 and diluted to 25 ng/ul with DEPC-treated water and placed into single-use aliquots and stored at − 80 °C.

Primers were purchased from Sigma-Aldrich™ (Easy Oligo) and arrived pre-diluted in deionized water at a concentration of 100 µM. The primers were made into single-use aliquots upon arrival and stored at − 20 °C. The sequences of the primers used can be seen in Table 1.

Primers were prepared for qPCR by adding 10 µL of forward primer, 10 µL of reverse primer, and 180 µL of Thermo Scientific TM DEPC treated water (Thermo Scientific, Catalog #: FERR0601) (5 µM final concentration of forward and reverse primer). A total of 1 µL of the prepared 5 µM primer mix, 4 µL of the synthesized cDNA, and 5 µL of the Applied Biosystems™ PowerUp™ SYBR™ Green Master Mix (Applied Biosystems™, Catalog #: A25918), were added to each well of a qPCR tube set (Bio Molecular Systems™, Catalog #: 71–107) (final reaction mix contained a 500 nM concentration of each primer and 100 ng of cDNA). Quantabio™ Q cycler was used to run the qPCR (Quantabio™, Catalog #: 95900-4C). All reactions were prepared in triplicates.

Data analysis was performed by the ΔΔCT method. ΔCT was first calculated by subtracting each sample’s GAPDH CT value from its average target gene CT value. ΔΔCT was calculated by subtracting ΔCT of the untreated cell from the ΔCT of the treated cell. The ΔΔCT values were then converted to fold values using 2^(-ΔΔCT).

Cell Extraction Buffer (Invitrogen™, Catalog #: FNN0011) and 10 µL of Thermo Scientific™ Halt™ Protease and Phosphatase Inhibitor Cocktail EDTA-free (100X) (Thermo Scientific™, Catalog #: 78445) was added to a microcentrifuge tube per 1 mL of cell Extraction Buffer. Each cell pellet was suspended in 250 µL of the freeze–thaw lysis buffer with protease and phosphatase inhibitor. The cell pellets were then dipped in liquid nitrogen for 10 s, allowed to thaw, and then vortexed. This was repeated three times. After the third freezing with liquid nitrogen, cells were placed on ice to thaw. The cell pellets were then placed in a microcentrifuge at max speed for 45 min at 4 °C. The supernatants were then placed into single-use aliquots at − 80 °C.

Protein was estimated using a Bradford assay by combining in each 1.5 mL tube: 500uL of H₂O, 495uL of Bradford, and 5uL of protein sample or BSA standard (or an additional 5uL Bradford for the plate blank). 600uL of this mixture was then plated in a 96 well plate (Corning™, Catalog #: 3598) at 200uL per well. The plate was then read using a SpectraMax Mini Multi-mode Microplate reader (Molecular Devices™, Catalog #: 76640–506) for absorbance at 595 nm. All values subtracted the plate blank, and the triplicate values were averaged together. A standard curve was made using BSA standards and protein was estimated using this curve.

40ug of protein from each sample was used. Protein was prepped by mixing 1 part protein and 1 part 2 × laemilli sample buffer (Bio-Rad™, Catalog #: 1610737) and placed in boiling water for 10 min. Each sample was loaded in a 4–20% gel (Bio-Rad™, Catalog #: 4561094) and 2 uL of Magic marker (Invitrogen™, Catalog #: LC5602) and 10uL of protein ladder (Thermo Scientific™, Catalog #: 26616) was loaded as well. Transfer was done using a wet transfer at 80 mV for 1 h to a nitrocellulose membrane (Bio-Rad™, Catalog #: 1620215). Membranes were blocked in 1% BSA in TBS for an hour and then then allowed to sit in antibody overnight. The primary antibody for BRG1 was Invitrogen™ BRG1 Monoclonal Antibody (GT2712) at a 1:1000 dilution (Invitrogen™, Catalog #: MA5-31550). The primary antibody for ULK1 was Invitrogen™ ULK1 Recombinant Rabbit Monoclonal Antibody (JA58-36) at a 1:1000 dilution (Invitrogen™, Catalog #: MA5-32699). The primary antibody for Beta actin was Abnova ACTB monoclonal antibody (M01) clone 3G4-F9 at a 1:750 dilution (Abnova, Catalog #: H00000060-M01). The antibody detection was done using the Pierce™ Fast Western Blot Kit (Thermo Scientific™, Catalog #: 35050) and followed the manufacturer’s protocol. The ChemiDoc MP imaging system (Bio Rad™, Catalog #: 1708280) was used to detect chemiluminescence.

The GEPIA online website was used to analyze the correlation of genes in CRC patients from The Cancer Genome Atlas (TCGA) database [19]. We assessed the correlation between gene expression of SMARCA4 and KRAS. This was done on the GEPIA website by selecting the Multiple Gene Analysis tab, selecting Correlation Analysis, and then selecting the following options: Gene A = SMARCA4, Gene B = KRAS, Normalized by gene = TUBA1A, Correlation Coefficient = Spearman, and Used Expression Datasets = COAD Tumor & READ Tumor, and then separately selected Used Expression Datasets = COAD Normal & READ Normal.

The Xena online website was used to analyze the expression of genes in CRC patients from TCGA database [20]. This was done by selecting the “GDC Pan-Cancer” dataset, limiting the “cancer type” to COAD or READ, and keeping only “primary tumor” or “solid tissue normal.” Three categories were created as follows: Solid Tissue Normal: Data with “sample_type” labeled as Solid Tissue Control. KRAS-wt CRC: Data with “sample_type” labeled as primary tumor and no mutation to the KRAS gene. KRAS-mut CRC: Data with “sample_type” labeled as primary tumor and one of the following mutations to the KRAS gene: G12A, G12C, G12D, G12R, G12S, G12V, G13C, G13D. Any sample without gene expression reported was removed. Cancer samples with “Null” for KRAS-mutation status were removed. The final sample size was n = 539 consisting of 51 solid tissue, 309 KRAS-wt CRC, and 179 KRAS-mut CRC.

We selected the three dots above the box containing the 3 subgroups of samples and selected differential expression. For our first calculation of mRNA changes in cancer we set subgroup A to include both cancer groups and subgroup B to include the Solid Tissue control. Our next calculation, of mRNA change in the presence of a KRAS mutation, we set subgroup A to include the KRAS-mut data and subgroup B to include the KRAS-wt data. The advanced settings were not changed. A file including Log(2)FC, p-value, and adjp is given, and fold change of the target gene is then calculated. In order to visualize these changes, the mRNA expression of the target genes were opened. The “view as chart” symbol was selected and “compare subgroups” was then selected. “Show data from” included the target gene and “subgroup samples by” included our three groups of samples.

The AI AlphaFold predicted structures for human KRAS (ID: AF-P01116-F1) and human SMARCA4 (ID: AF-P51532-F1) were downloaded from the Uniprot database in Protein Data Bank (PDB) format [21, 22]., Each file was uploaded separately to PyMol to visualize the 3D structures. Each protein file contained a single protein chain (Chain A). KRAS contained 189 residues and SMARCA4 contained 1647 residues. Chain A for both SMARCA4 was renamed “Chain B” to act as a “ligand” in a complex with Chain A of KRAS, serving as a “receptor.” To create the G13D mutation in KRAS, in PyMOL the Gly13 residue was mutated to Asp13 using protein mutagenesis. This was saved as a separate pdb file.

Wildtype KRAS and G13D were each docked to each other using the ClusPro server [23‐26]. Balanced structure “0” was downloaded as the complex structure for each docking. The proteins were again separated into separate chain files using PyMOL.

Using GROMACS, a topology file was created for KRAS-wt /G13D and SMARCA4. An OPLS-AA/L all atom force field was used. The SMARCA4 topology file was combined with the KRAS-wt topology to build the complex. A cubic box was generated and solvated. Ions were added to neutralize any charge in the complex if present. An energy minimization was run to reduce unfavorable sterics in the complex structure. NVT and NPT equilibrations were then used to respectively stabilize the temperature and pressure of the environment before running a 10 ns molecular dynamics simulation. Following the simulation, a trjconv command was used to correct any jumps of the protein around the box.

Root mean squared deviation (RMSD) was calculated for the carbon backbone to determine how much the complex moved from its original position, showing overall stability. Root mean square fluctuation (RMSF) was calculated for each protein in the complex in each simulation to determine the average displacement of their residues. Radius of gyration (Rg) was calculated for each protein in the complex as a measure of each of their overall structural compactness throughout the simulation while interacting.

gmx_MMPBSA software was used to calculate a per-residue decomposition analysis [27]. A MMGBSA (Generalized Born model) was used to produce energy values for the complex, the receptor KRAS /G13D), the ligand (SMARCA4), and the delta energy. The complex energy represented the bound state energy, and the receptor and ligand energies represented the unbound energies for KRAS /G13D and SMARCA4. The delta energy represented the overall binding strength. Delta energy (interaction energy) = Total Complex (bound state)—[Receptor + Ligand] (unbound states). Simulation energy values were represented in tables and graphs generated by the software.

In-silico protein-drug dockings were accomplished using the CB-Dock2 server [28, 29]. The KRAS-wt, G13D, and SMARCA4 pdb files generated for the simulations were reused. The model structure for carbamazepine (ID: N6W) was downloaded from the RCSB database. The model structures for trans-carbamazepine diol and carbamazepine-o-quinone were created using the CB-Dock2 server ligand drawing function. KRAS-wt, G13D, and SMARCA4 were each uploaded to the server with each form of carbamazepine for cavity detection and docking. The CB-Dock2 server generated five possible binding conformations for each protein-drug pair, ranked by their energetic favorability. The highest ranked conformation was chosen for analysis.

Statistical analysis of western blot and qPCR data was performed using Microsoft™ Office Excel. For fold change, a two-tailed one-sample t-test was used. When comparing fold changes, a two-tailed two-sample t-test was used. Outliers were determined and removed using Iglewicz and Hoaglin’s outlier test with modified z-scores using the outlier criterion of a modified z-score ≥ 3.5.

In order to better understand the method by which KRAS mutated CRC promotes cell survival, the TCGA COADREAD clinical patient dataset was analyzed with assistance from the UCSC Xena software. We first analyzed the difference in gene expression of normal tissue vs colorectal cancer data. We then focused on comparing the gene expression of KRAS-mut CRC vs KRAS-wt CRC. We found that SMARCA4 was indeed overexpressed in cancer patients by 59% (p = 1.81e-16, adj p = 1.71e-15). When comparing SMARCA4 expression in KRAS-mut vs KRAS-wt it was found that KRAS-mutant CRC has a 15% increase in SMARCA4 expression (p = 5.03e-5, adj p = 0.002) (Fig. 1A). We found that the homolog SMARCA2 decreased expression in cancer patients by 36% (p = 3.87e-10, adj p = 2.05e-09) (Fig. 1B). The gene expression can be expressed visually using Xena (Fig. 1C). This confirms previous literature that SMARCA4 uniquely acts as a tumor promoter in colorectal cancer [10, 16], and may present a new role of SMARCA4 in tumorigenesis in CRC with a KRAS mutation.

A plot of overall survival in KRAS-mutant CRC shows that high SMARCA4 expression significantly impacts survival probability in the first 1500 days compared to low SMARCA4 expression (p = 0.015). However, similar to the control sample, high SMARCA4 levels in KRAS-wt CRC do not significantly impact survival (Fig. 1D). This demonstrates a connection between KRAS-mutation and SMARCA4’s activity. Using GEPIA, we were able to analyze the correlation between KRAS and SMARCA4 in both COADREAD as well as normal tissue. In both the control as well as in cancer patients it was found a positive correlation between KRAS and SMARCA4 (p = 0, R = 0.95 and p = 0, R = 0.84 respectively) (Fig. 1E).

Within the complexity of the cell, there is potential for KRAS interaction with SMARCA4. If the two proteins interacted, an MDS details the structural and energetic components of the interaction. A comparison of KRAS-wt and G13D interactions with SMARCA4 demonstrates the effect of the mutation. The RMSD of each complex indicated that both systems were well-equilibrated (Supplementary Fig. S1A). Interestingly, the G13D complex initially had a higher deviation than the KRAS-wt complex from 0.5 ns—3 ns. It again had slightly higher deviation from 4 ns—5 ns. Between 6 ns—7 ns the KRAS-wt complex actually had a slightly higher deviation. The KRAS-wt complex ultimately equilibrated at 2.12 nm, which was higher than the equilibrium of the G13D complex at 1.97 nm. A higher RMSD value indicates lesser stability at those points in the simulation. The RMSF of SMARCA4 with each KRAS-wt and G13D are compared (Supplementary Fig. S1B). SMARCA4 also had slightly more fluctuation when interacting with KRAS-wt, except around residues 300–400, where there was significantly more fluctuation with KRAS-wt. In the Rg of both complexes SMARCA4 increased in compactness throughout the simulation, indicated by their negative slopes. However, SMARCA4 with KRAS-wt had lower Rg values than SMARCA4 with G13D, signifying the former had higher compactness. This trend was also observed for KRAS-wt and G13D themselves in which the former had lower Rg values throughout the simulation (Supplementary Fig. S1C). Energetically, both simulations produced stable binding patterns with large negative values. Line plots of the total complex energies throughout the simulations indicate that the KRAS-wt complex has less total energy than the G13D complex (Supplementary Fig. S1D). The KRAS-wt complex has a gradual decrease in energy, signifying an increase in favorable energy with a large standard deviation. However, the G13D complex remains more consistent with less of a standard deviation (Supplementary Table S1). Line plots of the delta energies indicate the overall strength of the binding energy in each simulation. The KRAS-wt complex has a gradual increase in favorable energy from ~ -80 kcal/mol to ~ -200 kcal/mol over the course of the simulation with an average energy of -133.5 kcal/mol. The G13D complex has a stronger binding energy of -150.6 kcal/mol and a much greater fluctuation in binding strength throughout the simulation (Supplementary Fig. S1E). The amount of individual residues involved in the delta energy reduced from 95 in the KRAS-wt complex to 84 in the G13D Complex.

In order to understand the possible effect of CBZ on SMARCA4 expression in KRAS-mut and KRAS-wt tumors, CRC cell lines were treated with 50uM of CBZ for 6 or 24 h. The fold change between the untreated and treated showed that in KRAS-mut cells, BRG1 expression lowered at 24 h (p < 0.05). In KRAS-wt, BRG1 expression was significantly raised at 6 h (p < 0.01) but not at 24. The fold change between KRAS-mut and KRAS-wt was statistically significant at both 6 and 24 h (p < 0.01 for both). This demonstrates the opposite effect of CBZ treatment on KRAS-mut and KRAS-wt CRC BRG1 expression (Fig. 2).

Having shown that BRG1 expression for CRC treated with CBZ is dependent on the presence of a KRAS mutation, we performed qPCR analysis to determine mRNA expression of SMARCA4 after 6 and 24 h of CBZ treatment. Our qPCR analysis found results consistent with BRG1 protein expression. Results showed that in KRAS-mut cells, SMARCA4 expression lowered at 24 h, resulting in a fold change of 0.69 (p < 0.01). In KRAS-wt, SMARCA4 expression was raised at both 6 and 24 h after treatment, resulting in a fold change of 1.62 and 1.31 respectively (p < 0.01 and p < 0.001 respectively). The difference in fold change between KRAS-mut and KRAS-wt was statistically significant at both time points as well (p < 0.01 and p < 0.001 respectively). This supports our findings and demonstrates the opposite effect CBZ has on KRAS-mut vs KRAS-wt CRC on SMARCA4 mRNA expression (Fig. 3).

Having shown that SMARCA4 expression is affected by both the presence of a KRAS mutation as well as treatment by carbamazepine, we sought to understand the connection between these components. In patient datasets, mRNA of ULK1 and mRNA of SMARCA4 correlated positively (p = 0, R = 0.84) (Fig. 4A). qPCR data reveals that ULK1 mRNA expression decreased in CBZ treated KRAS-mut CRC at both 6 and 24 h, with a fold change of 0.50 and 0.60 (p < 0.01 and p < 0.001 respectively) and was not changed in KRAS-wt CRC. The difference between KRAS-mut CRC and KRAS-wt CRC was significant at both time points as well. (p < 0.05 and p < 0.01 respectively) (Fig. 4B). This decrease in ULK1 corresponds to a decrease in SMARCA4 levels and may suggest a method by which ULK1 affects SMARCA4 in a KRAS-mut environment. CBZ can facilitate KRAS-mutated inhibition of ULK1 which may lead to a decrease in autophagy and downstream decrease in SMARCA4 levels. In non-tumor settings, carbamazepine is known to affect ULK1 phosphorylation and thus induce autophagy [12]. Our findings suggests that CBZ may affect autophagy differently in cancers, or that the affect it has on ULK1 phosphorylation may be slightly counteracted by its effect on ULK1 mRNA levels.

Docking carbamazepine (CBZ) and two metabolite forms, trans-carbamazepine diol (t-CBZ) and carbamazepine-o-quinone (CBZ-q) with KRAS-wt and the G13D mutant indicated the most favorable binding form of CBZ (Table 2) and which residues participated in interaction for each (Table 3). The binding between KRAS-wt and G13D with each CBZ form, respectively, was visualized (Fig. 5A and B).

The more negative that the vina docking score calculated by CB-Dock2 was, the stronger the binding through weak interactions between KRAS-wt and G13D and the CBZ form (Table 2). Binding energy was favored for G13D in CBZ and CBZ-q. However, it is notable that CBZ-q had the strongest binding, followed by CBZ, and lastly t-CBZ. There was a difference between cavity volume in KRAS-wt and G13D for the CBZ form binding. In KRAS-wt the cavity volume was 491 ų, while in G13D the cavity volume was 500 ų. These values were consistent with each CBZ form. Thus, the G13D mutation slightly increased the binding space.

A total of twenty residues were involved in the binding of KRAS-wt and G13D with the CBZ forms (Table 3). However, only fourteen of the twenty residues were consistent between the six dockings. Notably, in KRAS-wt the Gly13 residue only is involved in the binding with t-CBZ but not with CBZ or CBZ-q. However, when the Gly13 is mutated to Asp13, it is involved with all three forms.

Several residues appeared to bind specifically with certain CBZ forms but not others. In all instances the residues appeared in both KRAS-wt and G13D. The residues Leu19 and Thr144 only were involved in the binding of CBZ-q. The Ser145 residue was involved in the binding of CBZ and CBZ-q but not with t-CBZ. Only two residues besides Gly13 presented variability between KRAS-wt and G13D. Interestingly, Asn33 only participated in binding in KRAS-wt with t-CBZ. Another residue with variable results was Asn85. It participated in binding in KRAS-wt and G13D with CBZ, only in G13D with t-CBZ, and neither in KRAS-wt or G13D with CBZ-q.

CBZ, t-CBZ, and CBZ-q were also each docked with SMARCA4 to determine which CBZ form bound best with SMARCA4 (Supplementary Table S2) and which residues participated in interaction for each (Supplementary Table S3). The binding between SMARCA4 with each CBZ form was visualized as well (Fig. 5C). Binding energy was similar for each CBZ form with SMARCA4. The cavity volume was the same between the three, measured at 8709 ų.

A total of thirty-one residues were involved in the binding of SMARCA4 with the CBZ forms (Supplementary Table S3). There were no consistent residues between the three dockings. CBZ-q only bound to ten residues, none of which were similar to the CBZ or t-CBZ dockings. CBZ and t-CBZ shared sixteen residues in common. Three residues, Thr910, Gly911, and Gln1185, were involved in binding with CBZ but not t-CBZ. Two residues, Ile187 and Glu821, were involved in binding with t-CBZ but not CBZ.