Development of a Small Molecule Downmodulator for the Transcription Factor Brachyury

Abstract

Brachyury is an oncogenic transcription factor whose overexpression drives chordoma growth. The downmodulation of brachyury in chordoma cells has demonstrated therapeutic potential, however, as a transcription factor it is classically deemed “undruggable”. Given that direct pharmacological intervention against brachyury has proven difficult, attempts at intervention have instead targeted upstream kinases. Recently, afatinib, an FDA-approved kinase inhibitor, has been shown to modulate brachyury levels in multiple chordoma cell lines. Herein, we use afatinib as a lead to undertake a structure-based drug design approach, aided by mass-spectrometry and x-ray crystallography, to develop DHC-156, a small molecule that more selectively binds brachyury and downmodulates it as potently as afatinib. We eliminated kinase-inhibition from this novel scaffold while demonstrating that DHC-156 induces the post-translational downmodulation of brachyury that results in an irreversible impairment of chordoma tumor cell growth. In doing so, we demonstrate the feasibility of direct brachyury modulation, which may further be developed into more potent tool compounds and therapies.

Graphical Abstract

The small molecule brachyury downmodulator DHC-156 was developed through the optimization of afatinib, an FDA-approved inhibitor. A series of structure-activity relationships, guided by biochemical assays and crystallography, lead to the discovery of DHC-156, which was shown to spare all wild-type kinases and downmodulates brachyury in a post-translational manner.

Introduction

Chordoma is a malignant tumor localized to the spinal axis that has no approved targeted therapies.^{[1, 2]} Frontline therapy for chordoma remains surgery and radiation therapy; however, due to the tumor’s proximity to the central nervous system, these interventions are often ineffective and can lead to life-threatening complications.^{[1, 3]} Over the past decade, chordoma’s treatment with chemical therapies has proven challenging. The overexpression of receptor tyrosine kinases (RTKs) in chordoma called for further preclinical and clinical investigation of RTKs as potential therapeutic targets. The tyrosine kinase inhibitor imatinib^[4] and the RTK inhibitor lapatinib^[5] were assessed for their efficacy in clinical trials; however, neither demonstrated long-lasting antitumor activity. Despite its shortcomings in clinical trials, imatinib is nonetheless used as a last-line therapy when prior radiation and surgical interventions fail to stop tumor progression.^[6]

Chordoma is hypothesized to originate from remaining notochordal cells, which are typically absent after embryonic development.^[7],[8] The T-Box transcription factor, brachyury (gene symbol: T/TBXT), is a key regulator in notochordal development and its mesodermal specification and differentiation.^[9] Brachyury is highly expressed in chordoma cells but minimally expressed in healthy tissues,^[10],[11] making it a well-suited biomarker for chordoma.^[12] More recently, a genome-scale CRISPR-Cas9 screen identified TBXT as indispensable in chordomagenesis.^[13] Interestingly, brachyury has been found to also be widely expressed in non-chordoma cancer cell lines, including those derived from lung, breast, prostate, colon, and testicular cancer tumors.^[14] TBXT is located in a region heavily occupied by super-enhancers – regulators of key oncogene drivers in tumor cells^[15] – and is the most highly expressed super-enhancer-associated transcription factor.^[12] The therapeutic potential for brachyury has been explored showing TBXT silencing in chordoma cells results in cell growth inhibition and the development of senescence-like properties.^[16],[17] Accordingly, the use of an engineered cell line has demonstrated that brachyury degradation leads to an irreversible change in cell state by inducing cell senescence.^[17] Taken together, these data highlight degradation or downmodulation of brachyury as an attractive therapeutic approach for chordoma treatment.

Transcription factors, including brachyury, have traditionally been considered “undruggable”, based on their lack of ligandable pockets, active sites, and frequent sequestration into transcriptional complexes.^[18] Since no brachyury ligands have been reported, pharmacologically targeting brachyury directly for inhibition or degradation has proven challenging. Accordingly, previous attempts to target brachyury have sought to downregulate TBXT through inhibition of upstream kinases that drive its expression.^{[12, 19]} In 2018, Magnaghi et al. surveyed various small-molecule kinase inhibitors against several chordoma cell lines, identifying afatinib (BIBW 2992), a second-generation FDA-approved epidermal growth factor receptor (EGFR) inhibitor, as the most potent downmodulator of brachyury protein levels across all chordoma cell lines.^[20]

Based on these findings, we hypothesized that afatinib was directly interacting with brachyury, and thus could be used as a lead to design a novel selective brachyury ligand. As brachyury lacks kinase activity, we sought to eliminate afatinib’s “hinge-region” binding pharmacophore to retain brachyury potency but abrogate kinase binding. These compounds would represent the first reported chemical scaffold with activity and specificity for brachyury, and moreover as a proof-of-concept to validate pharmacologically targeting brachyury as an approach for therapeutic intervention against chordoma.

Herein, we report the development of a selective small molecule brachyury modulator. We designed and synthesized a scaffold that does not engage EGFR yet retains its ability to bind brachyury. We systematically explored structure-activity relationships (SAR), guided by structure determination of our scaffold in complex with brachyury DNA-binding domain (brachyury_DBD) to high-resolution, to improve the compound’s ability to induce brachyury downmodulation. Our efforts resulted in the development of DHC-156, a small molecule designed to modulate brachyury with potency equal to that of afatinib but with greater target selectivity. A kinase activity assay revealed that, unlike afatinib, DHC-156 does not inhibit EGFR or any of the 403 wild-type kinases tested. Interestingly, unlike PROTACs or molecular glues, DHC-156 modulates brachyury through a post-translational mechanism that does not appear to be proteasome- or lysosome-dependent, warranting further mechanistic studies. Employing chemical optimization, we demonstrate the feasibility of direct pharmacological intervention against brachyury and a new direction for future targeted chordoma therapies.

Results and Discussion

Afatinib Covalently Binds to Brachyury

In the Magnaghi et al. study, afatinib was one of the few kinase inhibitors tested that contained a Michael acceptor, in the form of an acrylamide (Figure 1a).^[20] Afatinib uses this reactive moiety to covalently bind a conserved cysteine residue in the EGFR active site, leading to its irreversible inhibition.^[21] We hypothesized that afatinib binds brachyury in a similar manner, and that this interaction causes the downmodulation of the transcription factor. We observed two solvent-exposed cysteines (Cys-186 and Cys-122) in a brachyury_DBD:DNA co-crystal structure (PDB ID: 6F58) that we predicted could participate in a Michael addition with afatinib.

Figure 1.

Afatinib binds directly to brachyury via a Michael addition with cysteine 122. (a) Covalent product formed from the binding of afatinib to brachyury cysteine. (b) LCMS spectra depicting the mass of brachyury_DBD apo (top) and afatinib bound (bottom). Brachyury expected mass: 21914 Da; apo brachyury mass: 21913 Da; afatinib-bound mass: 22399 Da. (c) The co-crystal structure of afatinib bound brachyury. Top: Afatinib and its interactions with the protein’s residues in the binding pocket. Water molecules were omitted. Bottom: Afatinib on the surface of brachyury. Nitrogen atoms are highlighted in blue, and oxygen atoms in red. Sulfur atoms are shown in yellow. (PDB ID: 6ZU8)

Using liquid chromatography-mass spectrometry (LCMS), we observed a mass-shift corresponding to the molecular weight of afatinib upon incubation of brachyury_DBD with the inhibitor (Figure 1b). Inspection of the crystal structure, as determined by x-ray crystallography, revealed an encouraging binding pocket containing Cys-122, leading us to hypothesize that this residue was responsible for covalent adduct formation. This was later confirmed by a co-crystal structure of brachyury_DBD and afatinib (PDB ID: 6ZU8) by the Gileadi lab at Oxford (Figure 1c). Our group, as well as the Gileadi lab, have demonstrated that afatinib covalently binds brachyury, leading to the discovery of the first validated ligand for the transcription factor brachyury.

Design and Synthesis of EGFR-Sparing Scaffolds that Retain Brachyury Downmodulation

Having validated the binding of afatinib to brachyury, we sought to optimize the compound selectivity for brachyury over EGFR and other kinases by designing several unique scaffolds to abolish key interactions with EGFR while preserving other favorable interactions between the ligand and brachyury. The afatinib:EGFR binding mode, as previously determined by x-ray crystallography, reveals two critical binding interactions: the covalent bond between afatinib’s acrylamide with Cys-787, and a hydrogen bond (~2.9Å) between the kinase hinge region and afatinib’s quinazoline core.^[21] SAR studies on 4-aminoquinazoline scaffolds have reported the hydrogen bond formed between the 1-position nitrogen of afatinib (Figure 1a) and the backbone of Met-731 is indispensable for EGFR binding.^[22]

Comparing the afatinib:brachyury_DBD binding mode to that of afatinib and EGFR led to the identification of pharmacophores necessary for brachyury binding that do not participate in binding interactions with EGFR, and vice versa. (Figure 1c). As previously mentioned, the acrylamide was found to be critical for covalent engagement with brachyury at Cys-122. We observed favorable shape complementarity and hydrophobic contacts by the chloro-, fluoro-aniline at a pocket on brachyury comprised of Val-123, Val-173, the aliphatic region of the side chain of Arg-180, the backbone of His-171, Met-181, and Ile-182.

Lastly, substitution of the tertiary dimethyl amine with a larger piperidine group decreases brachyury degradation, likely due to steric hindrance that attenuates target-engagement.^[20]

The afatinib:brachyury_DBD binding mode also revealed several solvent-exposed features of the inhibitor that do not contribute to brachyury binding that could be modified or removed to enhance brachyury selectivity.

For example, the chiral tetrahydrofuran (THF) on afatinib was found to be completely solvent-exposed in the afatinib:brachyury_DBD co-crystal structure and does not contribute towards any molecular recognition event with brachyury. In addition, the 1-position nitrogen in the quinazoline core does not engage in any interactions directly with brachyury and instead only with a water molecule. Given the importance of the 1-position nitrogen to bind EGFR, we hypothesized that removal of this heteroatom would reduce affinity for EGFR without disrupting any interactions with brachyury.

First, substitutions were made to the chiral THF to reduce the complexity of synthesis and decrease molecular weight, leading to greater ligand efficiency and minimizing the entropic penalty paid upon binding. We explored the smaller methoxy and hydroxy substituents in addition to the THF enantiomer to verify the chiral center’s importance (Figure 2a). Afatinib and its epimer, (R)-afatinib, were commercially available, whereas the other methoxy- and hydroxy-substituents were synthesized (Scheme S1).

Figure 2.

Design and development of EGFR-sparing brachyury downmodulators (a) Chemical structure of afatinib, (R)-afatinib (SJF-2504), SJF-1488, and SJF-2505 (b) Brachyury downmodulation profile for compounds (2a) in UM-Chor1 cells: western blots (left) and quantitation(right) depicting brachyury levels response to compounds at 1 μM to 10 μM, after 24 hr of incubation. Uncropped gel shown in Supporting Information. (c) Novel scaffolds of SJF-4601, DHC-7657 and DC-6142 designed from the modifications to the hinge-binding region of afatinib analog, SJF-1488, to disrupt affinity for EGFR. (d) LCMS spectra depicting the mass of brachyury (left) SJF-4601 bound, 88%, (middle) DHC-7657 bound, 95%, and (right) DC-6142 bound, 91%. Brachyury expected mass: 21914 Da; SJF-4601 bound mass: 22319 Da; DHC-7657 bound mass: 22340 Da; DC-6142 bound mass: 22341 Da (e) Corresponding EGFR inhibition activity plot (left) and quantitation of IC₅₀ values (right) of afatinib, SJF-4601, DC-6142, and DHC-7657. (n.d. not determined). Error bars indicate standard deviation. (f) Brachyury degradation profile of afatinib, SJF-4601, DC-6142, and DHC-7657 in UM-Chor1 cells after 24 hr, western blot (left) and quantitation(right). Uncropped gel shown in Supporting Information. (g) The co-crystal structure of brachyury-SJF-4601 complex (PDB ID: 8FMU). Top: Composite omit map (2F_o-F_c) at 1.5σ is shown for SJF-4601 and its covalent bond with Cys122 of the protein. Bottom: SJF-4601 is shown on the surface of brachyury. Nitrogen atoms are highlighted in blue, and oxygen atoms in red. Sulfur atoms shown in yellow.

The biological activity of these compounds was assessed by immunoblotting for brachyury levels in UM-Chor1 cells following treatments with increasing concentrations of each compound (Figure 2b). The afatinib enantiomer (SJF-2504) induced no change in brachyury levels relative to afatinib at 10 μM, suggesting that downmodulation is not dependent on the chirality of the THF. The phenol analog (SJF-2505), however, shows no modulation at the same concentration, demonstrating that an alkoxy substituent is required since the hydroxy is not tolerated. Interestingly, a methoxy substituent (SJF-1488) shows no change in brachyury degradation. These results suggest that brachyury degradation is dependent on the presence of an alkoxy substituent, although changes in on chirality or size are tolerated. Thus, we planned to use the smaller methoxy fragment for the synthesis of future analogs.

Next, we sought to abolish EGFR engagement by modifying the EGFR hinge-binding region, for which we designed three different scaffolds (Figure 2c). First, the benzamide scaffold of SJF-4601 was devised from opening of the pyrimidine ring, thereby removing the entire hinge-binding region while maintaining the halogenated aniline in an isosteric position. Conversely, we also sought to retain the bicyclic core, prompting us to disrupt the hinge-binding region by replacing the 1-positon or 3-position nitrogen with a carbon atom, yielding the isoquinoline scaffold (DHC-7657) and the quinoline scaffold (DC-6142), respectively. The three putative EGFR-sparing core scaffolds, SJF-4601, DHC-7657 and DC-6142, were synthesized as shown in Scheme 1a–c. Concise synthesis of the isoquinoline scaffold of DHC-7657 was enabled by the methodology recently reported by our lab.^[23]

Scheme 1.

Synthesis of SJF-1488 analogs. (a) Synthesis of benzamide scaffold, SJF-4601. (b) Synthesis of isoquinoline scaffold, DHC-7657. (c) Synthesis of quinoline scaffold, DC-6142.

We first sought to validate that our new scaffolds retained their ability to bind brachyury using LCMS (Figure 2d). Compounds were incubated with brachyury_DBD for 5 hours. We observed that all compounds presented comparable target-engagement relative to afatinib. We then sought to determine their ability to inhibit EGFR. Afatinib, SJF-4601, DHC-7657, and DC-6142 were profiled by a kinase assay in order to quantify their EGFR inhibition. Adenosine triphosphate (ATP) was present in the reaction at physiological concentrations (1 mM) to mimic physiological competition at the active site.

Afatinib and DC-6142 were found to possess half-maximal EGFR inhibition (IC₅₀) values of 18 nM and 1.1 μM, respectively (Figure 2e).

Excitingly, SJF-4601 and DHC-7657 were significantly less effective inhibitors of EGFR than afatinib, reducing activity to only 7% of the latter at the highest concentrations tested. Thus, we successfully designed two chemical scaffolds that selectively target brachyury while sparing EGFR inhibition.

SJF-4601, DC-6142, and DHC-7657 were characterized by immunoblotting for their ability to modulate brachyury (Figure 2f). The quinoline scaffold of DC-6142 demonstrated the most downmodulation (91%), which is comparable to afatinib itself (92%). DHC-7657, which possesses the isoquinoline scaffold, demonstrated somewhat reduced activity to downmodulate brachyury (63%); however, of these three molecules DHC-7657 was far more selective for brachyury over EGFR. Interestingly, the benzamide scaffold of SJF-4601 demonstrated no modulation of brachyury even at 10 μM. These results suggest that the bicyclic core is critical for brachyury downmodulation. Taken together, we demonstrate that modifications to the hinge binding region, via the substitution of the 1-position nitrogen, as seen with DHC-7657, abrogates EGFR binding yet retains its ability to downmodulate brachyury.

Concurrent with our studies to quantify brachyury downmodulation, we sought to validate the binding modes between our new scaffolds and afatinib. To do so, we solved the structure of brachyury_DBD in complex to SJF-4601 to 2.0 Å (PDB ID: 8FMU) (Figure 2g). The co-crystal structure reveals electron density between Cys-122 and the acrylamide, validating the presence of a covalent adduct. Interestingly, the majority of ligand interactions with protein and solvent that were revealed in the afatinib:brachyury_DBD structure are preserved.

Despite extensive efforts to determine the structure of DHC-7657 in complex with brachyury_DBD, we were unable to produce crystals of the ligand bound to brachyury. Given the structural similarity between DHC-7657 to afatinib and SJF-4601, it is likely the former possesses a similar binding mode. To further support this, we performed docking studies with DHC-7657 and brachyury, using the afatinib:brachryury_DBD structure (PDB Code 6ZU8) as a template. We observed the expected binding conformation with the chloro-, fluoro-aniline moiety making interactions with the same pocket of brachyury as does afatinib (Figure S1).

Optimization of Isoquinoline Scaffold for Brachyury Degradation

Next, we sought to further optimize the EGFR-sparing scaffolds for brachyury downmodulation. Of our designed EGFR-sparing scaffolds, DHC-7657 demonstrated the most potent downmodulation (Figure 2e,f), thus we elected to perform a SAR campaign around this scaffold. The binding mode of SJF-4601 and afatinib to brachyury revealed that their common halogenated aniline moiety is pointed into a pocket consisting of His-171 and Val-173. Interestingly, the fluorine of both molecules is pointed towards His-171, possibly indicating a potential electrostatic interaction that is important in the recognition event. Based on the co-crystal structure we hypothesized that derivatization of the aniline could result in new interactions within the pocket, further increasing its activity.

A medicinal chemistry campaign was undertaken to develop more active brachyury modulators that incorporate the EGFR-sparing isoquinoline scaffold. Various analogs were synthesized through the use of different aniline building blocks (Figure 3a) and their efficacy, quantified by western blot as previously described, was used to guide the SAR campaign. We first set out to determine the importance of the chlorine and fluorine substituents by introducing bioisosteric substituents.

Figure 3.

Structure-based design of DHC-156, which downmodulates brachyury equipotently to afatinib. (a) Table of degradation profiles after 24 hr for compounds developed in SAR campaign. DG: Degradation. (b) Western blot depicting brachyury degradation profile in UM-Chor1 cells for afatinib and DHC-156 and quantitation of their respective DC₅₀ and D_Max values after 24 hr. Experiment was run in duplicate, replicates plotted separately. Uncropped gel shown in Supporting Information. (c) Immunofluorescence straining of CH-22 cells with (left) DMSO, (middle) 10μM afatinib and (right) 10μM DHC-156 after 24 hr. DAPI was used for nucleus staining. Scale bar indicates 50 μm. Brachyury fluorescence was quantified by average intensity per cell using Fiji. Data are plotted as means +/− SD (n > 16 cells). Statistical analysis was performed using a two-tailed unpaired t-test, ****P < 0.0001.

While making these variations at R₂, we observed the greatest increase in activity relative to DHC-7657 by installing a larger methyl group or chlorine atoms to synthesize DHC-123 and DHC-180 respectively. When testing this trend on the R₁ position, we found that the presence of a methyl substituent in DHC-113 similarly enhanced brachyury downmodulation. Unfortunately, when attempting to combine these findings and add the more favorable methyl groups at both the R₁ and R₂ positions, DHC-236, we observed a decrease in activity indicating that the presence of a halogen at the R₁ or R₂ position was favored. Concordant with this notion, modifying both positions to hydrogens in DHC-202 resulted in a compound that is more active than DHC-236 but less effective than DHC-113. Modifying R₁ to a trifluoromethyl group, DHC-245, attempting to attenuate rotational entropy of the aniline moiety with the addition of a methyl at R₃ in DHC-244 did not yield a substantial improvement in their activity.

After the initial screen of bioisosteric substituents on the aniline, a variety of compounds showed enhanced degradation compared to DHC-7657, but the SAR could not be easily explained. So long as a halogen was present at R₁ or R₂, the compound was tolerant of the addition of hydrophobic electron-withdrawing groups at the other substituent. To probe the chemical space in the pocket proximal to R₁, we elected to move our SAR studies forward with the scaffold of DHC-113. We thus maintained the presence of the fluorine, which likely contributes to brachyury recognition through an interaction with His-171 and possesses a scaffold tolerable to modifications at the R₁ position. Accordingly, we sought to introduce larger than one-atom substituents, with varying physical characteristics, in order to maximize pocket occupancy on brachyury. The addition of electron-donating methoxy (DHC-115), isopropoxy (DHC-119) or trifluoromethoxy (DHC-121) groups to the aniline meta position did not increase activity; however, installation of the trifluoroethoxy (DHC-156) group notably improved downmodulation. On the other hand, the electron-withdrawing substituent of an alkyne (DHC-157) improved brachyury modulation, but not to the levels of DHC-156. Other electron-withdrawing groups such as cyano (DHC-144) and keto (DHC-157b) groups were ineffective.

Through the SAR campaign, we synthesized over twenty compounds and identified DHC-156 as the most potent brachyury downmodulator. Next, we sought to evaluate its activity compared to the parent molecule, afatinib (Figure 3b). We observed that DHC-156 (DC₅₀ = 4.1 μM) demonstrated no loss in potency relative to afatinib (DC₅₀ = 4.6 μM), with both compounds inducing complete downmodulation at 10 μM, D_Max > 99%. Meanwhile, we also attempted to visualize the brachyury downmodulation by immunofluorescent staining of brachyury in CH-22 cells (Figure S2, S3). Treatment with DHC-156 or afatinib at 10 μM significantly decreased the brachyury fluorescent intensity (Figure 3c), which confirmed the compound induced brachyury downmodulation. These results demonstrate the successful medicinal chemistry approach to optimize afatinib into DHC-156, an EGFR-sparing and equipotent brachyury modulator.

DHC-156 Lacks Appreciable Activity for All Wild-Type Kinases

Kinases possess highly conserved ATP-binding pockets, leading to difficulty in achieving selectivity when developing kinase inhibitors.^[24],[25] Accordingly, afatinib has been shown to inhibit numerous other kinases and their mutant isoforms in addition to EGFR.^[26] Having eliminated EGFR affinity from our brachyury modulators, we hypothesized that we may have also eliminated affinity for other kinases with which afatinib interacted. To test this, we subjected afatinib and DHC-156 to the KINOMEscan^™ scanMAX assay, which quantifies kinase interactions in a panel of over 480 kinases (Figure 4). The assay measured afatinib’s and DHC-156’s ability to bind specific kinase active sites, displacing these kinases from an immobilized pan-specific control ligand. Thus, compounds that can displace the control ligand will lead to lower quantities of the kinase captured on the immobilized support, signifying a stronger interaction of the test ligand with the kinase. For afatinib, the results were as expected, showing a strong interaction (~100% displacement of control) with EGFR, HER2, HER4, and IRAK1 at a concentration 10 μM. Afatinib was also found to target up to 70 other kinases (≥65% displacement of control), demonstrating poor selectivity. However, DHC-156 demonstrated no interactions against any of the tested wild-type kinases. Interestingly, DHC-156 shows engagement with a few EGFR mutants; EGFR(G719C), EGFR(L858R, T790M), and EGFR(861Q) (Figure S4). Taken together, we have designed a compound that lacks affinity for all wild-type kinases, resulting in a more selective brachyury downmodulator than afatinib.

Figure 4.

DHC-156 lacks kinase inhibition activity, whereas afatinib is highly reactive with multiple kinases. Kinome treespot showing locations of targets for afatinib (left) and DHC-156 (right).

DHC-156 Modulates Brachyury Levels Post-Translationally

With an optimized brachyury modulator in hand, we sought to investigate the mechanism by which brachyury levels were being modulated. Magnaghi et al. reported brachyury downmodulation was occurring through a canonical protein degradation mechanism, as protein level rescue was seen with both proteasome and autophagy inhibitors.^[20] Encouraged by these results, we sought to validate that our new brachyury ligand, DHC-156, acts in a similar manner. To do so, we pretreated CH-22 cells with either the proteasome inhibitor MG-132 or the autophagy inhibitor bafilomycin A1 (Baf), as well as a co-treatment with the two (Figure S5, S6). To our surprise, we observed no brachyury rescue when subsequently challenged with 10 μM of afatinib or DHC-156. To further determine whether DHC-156 and afatinib are modulating brachyury pre- or post-translationally, we performed a time course experiment in CH-22 cells using the ribosome inhibitor, cycloheximide (CHX) (Figure 5). Brachyury is shown to be quite stable with the half-life of 20.6 hours, as determined by treatment with CHX (50 μg/mL). When CH-22 cells were co-treated with DHC-156 (10 μM) and CHX, we observed an additive decrease in brachyury levels when compared to cells treated with only CHX or compound. Considering there is no new protein generated after ribosome inhibition, we conclude that DHC-156 downmodulates brachyury post-translationally. Furthermore, we were able to demonstrate a similar effect when treated with afatinib (Figure S7). In conjunction, we performed RT-qPCR to quantify the mRNA level of TBXT gene after treating CH-22 cells with 10 μM compound for 12 hours. We observed a 50% decrease of mRNA levels when treated with afatinib or DHC-156 (Figure S8). This could be due to the impaired T transcription by brachyury degradation, as it has previously been reported that brachyury expression is autoregulated.^{[17, 27]} Thus, post-translational loss of brachyury by DHC-156 treatment ultimately downmodulates brachyury levels via a pre-translational mechanism as well.

Figure 5.

DHC-156 downmodulates brachyury post-translationally and displays loss of brachyury phenotype. (a) Western blots for the time course experiment depicting brachyury downmodulation in CH-22 cells with DHC-156 (10μM) in the presence of CHX (50 μg/mL). Experiments were performed in duplicate, and the representative data was shown. Uncropped gel shown in Supporting Information. (b) Quantitation and curve fitting of brachyury immunoreactivity for each of the tested conditions (replicates plotted separately); half-life for each treatment was calculated fitting to a one-phase exponential decay. (c) Cell viability for CH-22 cells treated siRNA-brachyury and DHC-156 (5μM and 10μM) at 0, 2, 4 and 6 days. Statistical analysis was performed using a two-tailed unpaired t-test, ****P < 0.0001 (n = 3).

We next sought to demonstrate DHC-156 engages brachyury in cells via activity-based protein profiling (ABPP). To that effect, we synthesized a derivative of DHC-156, that possesses an alkyne “click” handle in lieu of the solvent-exposed methoxy moiety, DHC-402 (Scheme S2). In accordance with the aforementioned SAR (Figure 2a–b), this chemical modification had no impact on brachyury downmodulation (Figure S9). ABBP assays were performed at a four-hour time point in order to minimize the effect of downmodulation. While slight downmodulation was observed for both conditions in the input, there was sufficient brachyury levels present (Figure S10). Competitive ABPP demonstrated that DHC-402 target engagement, as visualized by immunoblotting for brachyury (50 kDa), can be out-competed by pre-treatment with the parent compound, DHC-156 (Figure S10). Bands observed below 50 kDa in DHC-402 treated lysates are believed to be degradation products of brachyury. Together, these data are suggestive of brachyury engagement in cells.

Brachyury Dependent Phenotype Observed in DHC-156 Treated Cells

We wanted to confirm that the cellular phenotype upon treatment with DHC-156 was dependent on the loss of brachyury protein. As previously shown, brachyury degradation results in the irreversible impairment of cell growth, unending even when cellular brachyury levels return to normal.^[17] We sought to determine whether a similar effect was observed when brachyury was knocked down, using siRNA, and when treated with DHC-156. Both DHC-156 and siRNA treatment result in decreased cell viability (Figure 5c). 10 μM DHC-156 exhibited greater cell death than the siRNA knockdown of brachyury, which we believe is likely due to the off-target activity of the compound. Interestingly, we observed a recovery of brachyury levels at the late timepoints of both DHC-156 treatments (Figure S11); however, this does not rescue cell viability. These observations are consistent with the previous study demonstrating the irreversible impairment of cell growth even with the resurgence of brachyury protein levels. Additionally, DHC-156 downmodulates brachyury in a dose-dependent manner (Figure S12), which correlates with cell viability, furthering the notion that DHC-156 induced cellular phenotypes are due to brachyury downmodulation.

While the mechanism by which DHC-156 promotes brachyury downmodulation remains unknown, we have demonstrated that downmodulation requires the formation of a covalent bond between DHC-156 and brachyury, as the saturated ligand DHC-388 is not able to promote brachyury downmodulation (Scheme S3, Figure S13). We hypothesize that covalent modification of brachyury changes the stability of the protein, resulting in a non-proteasomal mechanism of clearance. Nevertheless, elucidating the precise mechanism of brachyury downmodulation remains of great interest and will be the subject of future studies.

Conclusion

In summary, we have described the discovery and development of DHC-156, a small-molecule brachyury downmodulator. We demonstrated that afatinib binds brachyury through direct engagement by Michael addition to Cys-122. We identified a chemical scaffold that abolishes EGFR inhibition while retaining its ability to bind and modulate brachyury. Further, we determined the co-crystal structure of brachyury_DBD in complex with SJF-4601, which, in tandem with biochemical characterization and previously reported co-crystal structures of afatinib in complex with EGFR and with brachyury_DBD, respectively, altogether revealed the structural basis of selectivity of our compounds for brachyury over EGFR. We subsequently undertook a medicinal chemistry campaign on the EGFR-sparing isoquinoline scaffold to increase compound potency, yielding DHC-156, a brachyury modulator that is equipotent to afatinib while sparing all 408 wild-type kinases. Finally, we demonstrate that DHC-156 modulates brachyury protein levels post-translationally in a proteasome- and lysosome-independent manner; and has a pre-translational suppressive effect on brachyury expression as well. Ultimately, this dual-mode mechanism of action results in an irreversible impairment of chordoma tumor cell growth. While DHC-156 constitutes a proof-of-concept small molecule modulator of brachyury, further work is required to develop DHC-156 beyond a chemical tool. By demonstrating the feasibility of small molecule interventions against the “undruggable” transcription factor, brachyury, we believe that DHC-156 presents a novel approach towards the treatment of chordomas.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon request.