Identification of potent pan-ephrin receptor kinase inhibitors using DNA-encoded chemistry technology

Given the prevalence of kinase inhibitors as FDA-approved drugs, we wondered whether certain kinases might be therapeutic targets in endometriosis. Accordingly, we performed an analysis of overexpressed kinases in endometriotic lesions from peritoneal endometriosis lesion (PeL, N = 71), deep infiltrating endometriosis (DiE, N = 167), and endometriomas (OMA, N = 25) relative to normal eutopic endometrial tissues (CE, N = 43) and the patient endometrium (PE, N = 104) from a previously published dataset ( GSE141549 ) ( 44 ). We identified EPHA2 and EPHA4 as potential nonhormonal therapeutic targets for endometriosis based on their significantly elevated expression levels in the PeL and DiE relative to CE ( Fig. 1 A and B and SI Appendix , Fig. S1 A ). In endometrioma tissues (OMA), EPHA4 was 1.74-fold higher ( FDR < 0.001) relative to control endometrium (CE); in peritoneal lesions (PeL), EPHA2 was 1.57-fold ( FDR < 0.001) and EPHA4 was 1.72-fold ( P < 0.001) higher than CE; while in deep infiltrating endometriotic lesions (DiE), EPHA2 was 1.34-fold ( FDR < 0.01), and EPHA4 was 1.75-fold ( FDR < 0.001) higher than in CE. Using this approach, we identified EPHA2 and EPHA4 as kinase receptors up-regulated in a large set of endometriotic lesions relative to CE. Identification of EPHA2 and EPHA4 as drug targets in endometriosis. ( A and B ) Expression levels of EPHA2 ( A ) and EPHA4 ( B ) were determined from GSE141549 and represent samples from CE (N = 43); PE (N = 104); PeL (N = 71); DiE (N = 167); and OMA (N = 25). Box and whisker plots represent log2Expression and are plotted as mean, with min-max expression. Lesion expression was evaluated using multiple testing ( FDR < 0.05 ) in CE versus PeL, OMA, and DiE. ( C – H ) EPHA2 immunohistochemistry (IHC) in CE obtained from proliferative phase ( C , ProEN , N=5), early secretory endometrium ( D, ESE , N = 4) and midsecretory endometrium ( E, MSE, N = 5). EPHA2 was also assessed in PeL (N = 3), DiE (N = 6), and OMA (N = 10). White boxes represent the areas presented with a higher magnification. ( I ) Quantification of the EPHA2 signal in the various lesions was determined using ImageJ and presented as the percent area analyzed with positive signal. Data are presented as box and whisker plots with min-max expression and analyzed using a Kruskal–Wallis test with a Dunnet’s multiple comparison posttest. ( J and K ) Analysis of 36,697 endometriosis cases and 116,071 controls from the FinnGen database identified 5 EPHA2 ( J ) and 210 EPHA4 ( K ) variants associated with endometriosis ( P < 0.001, in red). NOTE: 54 EPHA2 and 1049 EPHA4 variants are found with 0.001 ≤ P ≤ 0.01, in green. Variants within the EPHA2/EPHA4 gene boundaries are darker, and those up/downstream of the EPHA2/EPHA4 gene boundaries are displayed in lighter colors. To determine whether the levels of EPHA2 and EPHA4 changed throughout the menstrual cycle, we analyzed a gene expression dataset ( GSE51981 ) of endometrial biopsies staged according to the menstrual cycle phase ( 45 ). Levels of EPHA2 are significantly decreased in the early secretory (N = 6) and midsecretory (N = 8) endometrium relative to the proliferative endometrial biopsies (N = 20) ( SI Appendix , Fig. S1 B ). On the other hand, EPHA4 levels are increased in the early secretory and midsecretory endometrium relative to the proliferative phase endometrium ( SI Appendix , Fig. S1 C ). Thus, the levels of EPHA2 and EPHA4 dynamically change in the eutopic normal endometrium. To validate the expression levels and localization of EPHA2 in the normal endometrium and endometriotic lesions, we performed EPHA2 IHC in the normal eutopic endometrium ( CE, N = 15), peritoneal endometriotic lesions ( PeL, N = 3), deep infiltrating endometriosis ( DiE , N = 6), and in OMA (N = 10) ( Fig. 1 C – H ). IHC in the normal eutopic endometrium was performed using biopsies obtained during various phases of the menstrual cycle, using proliferative endometrium ( ProEn, N=5), early secretory endometrium ( ESE, N = 4), and midsecretory endometrium ( MSE, N = 5). We found that relative to the proliferative phase endometrium, EPHA2 levels decreased in the ESE ( Fig. 1 I ). Likewise, relative to the ESE normal endometrium, levels of EPHA2 were significantly elevated in the endometrioma tissues ( Fig. 1 I ). Compartment-specific analyses identified that EPHA2 was more frequently detected in the epithelial layers of the lesion; however, some lesions also displayed high EPHA2 expression in the stroma. Thus, our results here validate the microarray analyses, showing that EPHA2 is detected in the epithelium and stromal cell compartments of the endometriotic lesions and the normal eutopic endometrium. We also demonstrate that EPHA2 levels fluctuate in the normal eutopic endometrium. We also identified many EPHA2 and EPHA4 variants that were associated with endometriosis ( Fig. 1 J and K ). The analysis was performed in a cohort of 36,697 endometriosis cases and 116,071 controls annotated in the FinnGen database ( 46 ). We found that approximately five variants associated with EPHA2 and 210 variants associated with EPHA4 with a cutoff of P < 0.001 (displayed in red), while 54 variants associated with EPHA2 and 1,049 variants associated with EPHA4 with a cutoff of 0.001 ≤ P ≤ 0.01. Thus, our analysis reveals that EPHA2 and EPHA4 display increased gene expression and increased genetic variants that are associated with endometriosis. To identify potent and selective small molecules that inhibit the kinase activity of EPHA2 and EPHA4, 54 unique DECLs cumulatively containing 4.42 billion compounds were screened in our DEC-Tec platform with an N -terminal-His-tagged EPHA2 kinase domain (amino acids 596 to 900) or EPHA4 kinase domain (amino acids 601 to 892) that were produced in insect cells. During affinity selections, library pools were incubated with His-tagged EPHA2 or EPHA4 at a concentration of 0.2 μM. A no-target control (NTC) experiment without target protein was also performed in parallel as a negative control to rule out nonspecific bead binders. After three to four rounds of selection, NTC, EPHA2, and EPHA4 samples were PCR amplified and subjected to Illumina next-generation sequencing. Our informatics pipeline was used to decode the chemical structures bound to EPHA2 and EPHA4 from DNA sequences captured by the sequencer and to calculate the enrichment of these compounds through statistical analysis. Enrichment of the EPHA4 bound compounds is shown in Fig. 2 with their normalized z-score and number of counts. We developed an in-house normalized z-score to quantify the enrichment of n-synthons to address the bias of count data caused by factors such as sample size and library diversity ( 38 ). A higher z-score indicates a more significant enrichment in the selection. DECL selection data ( A ) qDOS28_1 and ( B ) qDOS26. The enrichment of each library series was shown as count/z-score in the box and graphed as z-score in the EPHA4 0.2 μM selection on the x axis versus z-score for the selection against no-target control on the y axis. For each library, the structures BB1 (in blue), BB2 (in red), and BB3 (in black) are shown from the Top to Bottom . The desired chemical structures of strongly enriched hit compounds CDD-2693 and CDD-2781 are shown on the Right . Analysis of the compounds from two libraries, qDOS28_1 and qDOS26, indicated that structurally similar molecules were enriched ( Fig. 2 ). In addition to normalized z-scores, the analysis of enriched compounds that are structurally similar gives further credence that the compounds are real binders, while also informing on structural features involved in target recognition. Hit compounds CDD-2693 (MW: 489) and CDD-2781 (MW: 513), identified from qDOS28_1 and qDOS26, respectively, were synthesized off-DNA and shown to be potent inhibitors of EPHA2 and EPHA4 in vitro as presented in the next sections. After DEC-Tec selection, we analyzed the BB combinations according to the most enriched sequences. This analysis revealed two compounds, CDD-2693 and CDD-2781 (encoded from libraries qDOS28_1 and qDOS26, respectively), having identical structures for BB1 and BB2 but differing in BB3. The structural similarity of the two compounds arising from two distinct libraries provided confidence for validating them by off-DNA synthesis. The chemical synthesis of DEC-Tec hit compounds is shown in Scheme 1. In brief, the synthesis of CDD-2693 initiated by amide coupling of readily available (1-(2-amino-4-fluoro-5-methylphenyl)-N-methylpiperidine-4-carboxamide ( 1 ) and 4-bromo-3-methylbenzoic acid ( 2 ) to provide key precursor ( 3 ). Subsequently, a Suzuki-Miyaura coupling was performed between arylbromomide intermediate 3 and 5-hydroxy-2-methylphenyl boronic acid to yield the desired target compound CDD-2693 (4a, 51%). Similarly, intermediate 3 was reacted with (5-methyl-1H-indazol-4-yl)boronic acid to furnish the hit compound CDD-2781 (4b) in 36% yield. Further confirmation of the binding of hit candidate CDD-2693 in the EPHA2 and EPHA4 kinase domain, we used a thermal shift assay (TSA) to measure the extent of protein stabilization induced by small-molecule binding ( SI Appendix , Fig. S2 ). To examine the inhibitory activity of the resulting hit candidates, we screened for EPHA4 and EPHA2 kinase domains using an in vitro biochemical assay ( Fig. 3 ). The CDD-2693 displayed picomolar/nanomolar potency [EPHA4 (K i : 0.8 nM), EPHA2 (K i : 4.0 nM)], and another hit candidate CDD-2781 [EPHA4 (K i : 3.3 nM), EPHA2 (K i : 7.3 nM)] displayed nanomolar potency. These biochemical data highlight that DEC-Tec can provide exceptionally potent compounds directly from the selection and without the need for extensive medicinal chemistry efforts. Off-DNA synthesis of DEC-Tec selection hits (CDD-2693 and CDD-2781). Inhibition of selected compounds against EPHA4 and EPHA2. ( A ) Chemical structures of potent compounds. ( B ) Dose–response curves of represented compounds against EPHA4. ( C ) Dose–response curves of represented compounds against EPHA2. To interrogate the kinome-wide selectivity of CDD-2693, we performed a KINOMEscan assay at Eurofins at 1 μM concentration toward a panel of 468 kinases ( 31 , 47 ). The KINOMEscan ( Fig. 4 ) revealed CDD-2693 to be selective across the human kinome, inhibiting only closely related EPH and SRC tyrosine kinase (TK) family members ( SI Appendix , Table S1 ). To further confirm this selectivity, we conducted a LanthaScreen kinase binding assay on selective EPH and SRC kinase targets using Thermo Fischer Scientific’s SelectScreenServices ( Table 1 ). CDD-2693 demonstrated nanomolar potency against EPHA2 (K i : 4.0 nM), EPHA4 (K i : 0.8 nM), EPHA8 (K i : 4.7 nM), and EPHB4 (K i : 4.8 nM). Additionally, CDD-2693 also displayed nanomolar potency toward a branch of nonreceptor SRC tyrosine protein kinases, including FGR (K i : 1.8 nM), YES1 (K i : 2.0 nM), BLK (K i : 8.8 nM), FRK (K i : 9.4 nM), LCK (K i : 10.6 nM), LYNA (K i : 11.8 nM), and SRC (K i : 11.9 nM). This underscores that in addition to generating potent compounds, DEL selections can also identify compounds with admirable selectivity. Selectivity profile of compound CDD-2693 was assayed at 1 μM against 468 kinases using the DiscoveRx KINOMEscan screen. Compound selectivity is represented in a TREEspot kinase dendrogram view of the human kinome phylogenetic tree. Inhibition was measured in the percentage of control; the lower the percent control and the larger the red circles indicate stronger inhibition against the corresponding kinases; all other kinases tested were inactive, as indicated by the green circles. Kinase selectivity data for compounds CDD-2693 and CDD-3167 NA, not active at 1 μM in the Z’-LYTE and Adapta inhibition data Having achieved excellent inhibition of EPHA2 and EPHA4 directly from the DEL selection, we next synthesized derivative compounds and performed computational modeling to explore SAR and understand the structural features required for activity. Additionally, we sought to produce compounds with further potency, greater off-target selectivity, and better drug-like properties such as solubility and metabolic stability. Given the modularity of DELs, off-DNA SAR could be efficiently determined through the incorporation of BBs having specific structural features without changing the synthetic pathway outlined in Scheme 1. We mainly focused on varying the substituents on the two aromatic rings (A-ring and B-ring) contained in the parent compounds CDD-2693 and CDD-2781 ( Table 2 ). To generate analogs with greater solubility for better cell penetration, we explored truncations at the A-ring R 3 site close to the former DNA attachment point. Removal of the entire methylpiperidine-4-carboxamide (R 3 ) group (CDD-2868) resulted in a complete loss of activity to both EPHA2 and EPHA4 as did truncation down to the dimethylamine (CDD-2903). Furthermore, eliminating just the N-methycarboxamide portion from the piperidine ring (CDD-2869 and CDD-2904) also ablated activity. These results suggested that the N-methycarboxamide forms critical interactions to EPHA4 and EPHA2. Our modeling results show that the nitrogen in the N-methycarboxamide moiety forms hydrogen bonds with Arg750 and Ser706 in EPHA4. Intriguingly, this serine is absent in EPHA2, being replaced by Ala599. This may partially explain why this series of compounds showed better binding affinity to EPHA4 in comparison to EPHA2 ( Fig. 5 ). We next prepared analogs with varying substituents at the R 1 and R 2 positions of the A-ring. Eliminating the R 2 fluoro group (CDD-2905) yielded similar potency but a turnover in selectivity given that better potency to EPHA2 (K i : 1.0 nM) over EPHA4 (K i : 5 nM) was observed. Substitution with the smaller N-methycarboxamide at R 3 of the defluorinated compounds (CDD-2969 and CDD-2970) diminished the EPH activities once again corroborating the importance of the entire N-methylpiperidine-4-carboxamide unit. CDD-2967 which lacks both the fluoro and methyl groups at R 1 and R 2 relative to the parent (CDD-2693), gave K i values of 0.6 nM and 0.25 nM to EPHA2 and EPHA4, respectively. This corresponded to a threefold increase in potency toward EPHA4 and a sevenfold increase toward EPHA2 relative to CDD-2693. Interestingly, removing the same fluoro and methyl substituents from the indazole parent compound (CDD-2781) resulted in CDD-2968 with K i values of 0.7 nM to EPHA2 and 9.7 nM to EPH4 and thus having a more than 10-fold selectivity toward EPHA4 over EPHA2. Structures and activities of hits (CDD-2693 and CDD-2781) and SAR analogs NA, not active at 250 nM in the biochemical assay. K i determined from Z’-LYTE and Adapta inhibition data. ( A ) Computational modeling of EPHA2/CDD-2693 interaction. CDD-2693 forms critical interactions with residues E663, M695, and R743. EPHA2 was shown in green cartoon; CDD-2693 was shown in gray stick. Residues involved in interaction were shown in green stick. ( B ) Computational modeling of EPHA4/CDD-2693 interaction. CDD-2693 forms critical interactions with residues E670, M702, S706, and R750. EPHA4 was shown in magenta cartoon; CDD-2693 was shown in gray stick. Residues involved in interaction were shown in magenta stick. We next turned our attention toward the B-ring and R 5 position which was the basis of the two original hits from the DEL selection. Removing the methyl (CDD-3163) or hydroxyl groups (CDD-3164) on the phenyl ring (compared to CDD-2967) reduced potency to EPHA2 and EPHA4 by more than two orders of magnitude, whereas converting the hydroxy to the methoxy group (CDD-3165) completely abrogated activity. Docking results indicate that the hydroxyl group located on ring 5 forms hydrogen bonds with Glu670 in EPHA4 and Glu663 in EPHA2. Substituting this hydroxyl group with a methyl group resulted in loss of these pivotal interactions, leading to a notable decrease in activity. Furthermore, the methyl group positioned para to the hydroxyl group on ring 5 contributes to binding by fitting into a hydrophobic pocket formed by alanine and valine amino acids. We also explored analogs focused around the indazole parent compound CDD-2968 and found that omitting the 5-methyl group from the indazole ring (CDD-3074) significantly attenuated activity to EPHA2 (K i : 53 nM) and EPHA4 (K i : 17.7 nM). The presence of this 5-methyl group aligns with the position of the methyl group in CDD-2967 within the predicted binding mode. This suggests that the absence of hydrophobic interactions in this pocket could negatively affect the binding affinity. Alternatively, switching the indazole heterocycle to the indole (CDD-3075), 7-azaindole (CDD-3076), or the pyridine (CDD-3170) abolished all activity. The additional nitrogen in the indazole moiety appears to establish a hydrogen bond with Ser763 in EPHA4 and Ser756 in EPHA2, suggesting the importance of this interaction in maintaining the binding pose. Removal of the methyl group at the R 4 (CDD-3166) resulted in a more than 10-fold loss in potency to EPHA2 and EPHA4 as did the introduction of an electron-donating methoxy group (CDD-3168). However, conversion of the phenyl ring to the pyridine (CDD-3167) gave excellent activity toward EPHA2 (K i : 0.13 nM) and EPHA4 (K i : 0.38 nM). Altogether, based on these SAR studies, CDD-2967 and CDD-3167 emerged as the potent inhibitors of EPHA2/EPHA4. During the lead optimization, we evaluated the metabolic stability of selected potent compounds in mouse/human liver microsomes (MLM/HLM) ( Table 3 ). Our DECL-generated hit compound CDD-2693 showed acceptable stability in MLM (t 1/2 = 37 min) and excellent stability in HLM (t 1/2 = 121 min). Additionally, CDD-2781 with the methyl-indazole (R 5 ) motif displayed diminished metabolic stability in MLM (t 1/2 = 27 min) but exceptional stability in HLM (t 1/2 = 518 min). Upon removal of the fluoro group (CDD-2905), the stability was reduced in both MLM (t 1/2 = 10 min) and HLM (t 1/2 = 114 min) compared to CDD-2693. Unfortunately, when the methyl and fluoro substituents were removed from the A-ring (CDD-2967 and CDD-2968), the half-lives dropped to less than 10 min in MLM and less than 100 min in HLM. When the N-methylpiperidine-4-carboxamide was replaced with the smaller N-methylcarboxamide (and CDD-2970), the microsomal stability dramatically improved for CDD-2969 (MLM: t 1/2 = 42 min; HLM: t 1/2 = 412 min) and CDD-2970 (MLM: t 1/2 = 38 min; HLM: t 1/2 = 467 min). The most potent pyridine analog, CDD-3167, was unstable in both MLM and HLM (MLM: t 1/2 = 9 min; HLM: t 1/2 = 29 min). In summary, all the active EPHA2/EPHA4 analogs tested had a t 1/2 in MLM of less than 45 min and were more stable in HLM, with some analogs being remarkably stable (t 1/2 of greater than 2 h). Metabolic stability of selected compounds in MLM and HLM * t 1/2 measured using the liver microsomal stability assay; MLM/HLM, mouse/human liver microsomes; CL, intrinsic clearance. To determine the cellular potency and selectivity, we evaluated a series of potent compounds in cell-based NanoBRET Intracellular Kinase assays (Promega). This assay determines the apparent affinity of test compounds by competitively dislodging the NanoBRET tracer that is reversibly attached to a NanoLuc luciferase kinase fusion in living cells. We evaluated our compounds against EPH receptors EPHA2, EPHA4, EPHA5, and EPHB2 and off-target kinases SRC, FGR, and YES1 ( Table 4 ). Dasatinib was used as a positive control in this assay. The results of the NanoBRET experiments demonstrated that remarkably, seven of the inhibitors (CDD-2693, CDD-2781, CDD-2967, CDD-2968, CDD-2905, and CDD-3167) exhibited IC 50 values below 500 nM (1.0 nM to 460 nM) toward these EPH family members, indicating their ability to enter cells and engage the EPH targets. The cellular potency of our qDOS28_1 hit compound CDD-2693 was below 100 nM cellular efficacy against EPHA4 (IC 50 : 39.6 nM), EPHA5 (IC 50 : 52.7 nM), and EPHB2 (IC 50 : 69.9 nM) but showed 460 nM potency against EPHA2 (i.e., 11-fold selective versus EPHA4). A similar selectivity pattern was observed for qDOS26 hit compound CDD-2781 EPHA4 (IC 50 : 40 nM), EPHA5 (IC 50 : 38.0 nM), EPHB2 (IC 50 : 60.0 nM), and EPHA2 (IC 50 : 159.0 nM) (i.e., fourfold selective). CDD-3167, our most potent analog in vitro, showed single-digit nanomolar potency in cells against EPHA2 (IC 50 : 8.0 nM), EPHA4 (IC 50 : 2.3 nM), EPHA5 (IC 50 : 3.0 nM), and EPHB2 (IC 50 : 5.2 nM). The cellular potency of CDD-2693 (and all analogs tested) against off-target kinases SRC, YES1, and FGR was in the micromolar range ( Table 4 ), unlike dasatinib, which showed IC 50 values of 19 nM, 16 nM, and 41 nM, respectively, against these three kinases. The NanoBRET target engagement results concluded that our developed inhibitors had tremendous cellular selectivity compared to dasatinib. NanoBRET target engagement data against EPH and SRC, YES1, and FGR kinases –, not tested; ND, not determined (IC 50 > 50,000 nM). Building upon our in vitro assays, we also characterized the cellular effects of our newly identified ephrin receptor inhibitors. We used an immortalized human endometriotic cell line ( 48 ), 12Z and an Ewing’s sarcoma cell line, A673, to examine the phosphorylation of EPHAs upon our inhibitor treatment. To further mimic the biological activation of the EPHAs, cells were treated with EFNA5, a potent ligand for both EPHA2 and EPHA4 ( 4 ), in the presence or absence of the inhibitors. We chose hit compound CDD-2693 and potent analog CDD-3167 for further experiments based on their potency in cell-based assays. 12Z cells were pretreated with inhibitors for 15 min followed by the addition of 0.5 µg/mL of EFNA5 for a total of 30 min. Both CDD-2693 and CDD-3167 inhibited the EFNA5-induced phosphorylation of EPHA2/4 (Y772/Y779) in a dosage-dependent manner ( Fig. 6 A ). Further statistical analysis of the signal intensities revealed that CDD-2693 significantly inhibited the phosphorylation starting at 1 μM concentration, while a decreasing trend could be observed at 100 nM with a P value of 0.06. CDD-3167 is more potent, showing a significant inhibition effect starting at 1 nM concentration ( Fig. 6 B ). Such an inhibition effect was also observed in the A673 cells, starting at 1 μM (CDD-2693) and 100 nM (CDD-3167) ( Fig. 7 A and B ). Additionally, we tested the 12Z and A673 cell viability upon the inhibitor treatment. In 12Z cells, CDD-2693 and CDD-3167 both suppressed cell growth at a half-maximal inhibitory concentration (IC 50 ) of 2,831 and 456.3 nM, respectively ( Fig. 6 C ). In A673 cells, IC 50 for CDD-2693 is 8,755 nM, and CDD-3167 has an IC 50 of 5,085 nM ( Fig. 7 C ). Inhibited activation of EPHAs upon lead compound treatment in 12Z cells. ( A ) Western blot showing the inhibition of EPHA2/4 phosphorylation after treatment with 0.5 µg/mL Ephrin-A5 (EFNA5) for 15 min in the presence or absence CDD-2693 and CDD-3167. ( B ) Densitometric analysis of ( A ), fold changes were calculated toward vehicle treatment without EFNA5 stimulation, GAPDH intensities were used as normalizing internal control. (N = 3 biological replicates, one-way ANOVA with Tukey’s post hoc analysis, plotted as means ± SEM. **** P < 0.0001. *** P < 0.001. * P < 0.05). ( C ) Cell viability assay in 12Z cells upon CDD-2693 and CDD-3167 treatment after 72 h incubation. IC 50 was calculated by nonlinear fit analysis model using GraphPad Prism. ( D ) PTGS2 gene expression assay in 12Z cells after 3 h IL-1B challenging. Data were normalized toward vehicle treatment with IL-1B stimulation. IC 50 was calculated by the nonlinear fit analysis model using GraphPad Prism. ( E ) Endometrial organoid growth curve measured by diameter over a 3-d treatment of CDD-2693. ( F ) Bright-field images of endometrial organoids after 5 µM inhibitor treatment from day 1 ( Left ) to day 3 ( Right ). Inhibited activation of EPHAs upon active compound treatment in A673 cells. ( A ) Western blot showing the inhibition of EPHA2/4 phosphorylation after treatment with 0.5 µg/mL Ephrin-A5 (EFNA5) for 15 min in the presence or absence CDD-2693 and CDD-3167. ( B ) Densitometric analysis of ( A ); fold changes were calculated based on vehicle treatment without EFNA5 stimulation, and GAPDH intensities were used as a normalizing internal control. (N = 3 biological replicates, one-way ANOVA with Tukey’s post hoc analysis, plotted as means ± SEM. **** P < 0.0001. *** P < 0.001. * P < 0.05). ( C ) Cell viability assay in A673 cells upon CDD-2693 and CDD-3167 treatment after 72 h incubation (CDD-2693: N = 3 biological replicates, CDD-3167: N = 4 biological replicates). IC 50 was calculated by nonlinear fit analysis model using GraphPad Prism. One of the hallmarks of endometriosis is inducing a proinflammatory state in the microenvironment of the ectopic lesions ( 49 – 51 ) and previous studies in the ovary have shown that ephrin signaling controls expression of the proinflammatory gene, PTGS2 ( 18 ). To test the therapeutic potential of our inhibitors in the context of inflammation, we treated 12Z cells with IL-1β in the presence or absence of CDD-2693 and CDD-3167. After pretreatment of the CDD-2693 and CDD-3167 compounds for 30 min, IL-1β was added to the culturing medium to a final concentration of 0.05 ng/mL. RNA was collected after a 3-h incubation and subjected to reverse transcription quantitative real-time PCR (RT-qPCR) to evaluate the transcript level of PTGS2 , a known downstream target of IL-1β to initiate immunoinflammatory responses ( 52 ). As shown in Fig. 6 D , both CDD-2693 and CDD-3167 decreased the PTGS2 expression in the presence of IL-1β, with an IC 50 of 7.2 μM and 1.3 μM, respectively. Patient-derived organoids grown in three-dimensional cultures have become useful platforms to test the biological effects of drug-like molecules ( 53 ), especially in the field of endometrial biology ( 54 , 55 ). We established endometrial epithelial organoids obtained from hysterectomies of individuals with endometriosis following previously published conditions ( 54 , 56 ). Four days after plating, epithelial organoids were treated with 5 µM of CDD-2693 and CDD-3167 and allowed to grow over the course of 72 h ( Fig. 6 E and F ). Unlike the vehicle-treated organoids, which continued to expand in diameter over the course of the experiment, organoids treated with the ephrin receptor inhibitors, CDD-2693 and CDD-3167, showed a significant decrease in diameter after 72 h of treatment. These results show that CDD-2693 and CDD-3167 significantly decrease the average diameter and the expansion of patient-derived organoids over a 3-d treatment in a patient-derived model of endometriosis.

Ephrins signal through their corresponding EPH receptors to control cellular communication and influence cellular processes. Using our DEC-Tec platform, we identified a series of pan-ephrin receptor kinase inhibitors. Initially, we employed efficient on-DNA compatible chemistries to synthesize diverse structural chemical libraries, and after the affinity selection directly yielded two high count DECL selection hit compounds CDD-2693 and CDD-2781 from qDOS28_1 and qDOS26 libraries, respectively. These hits were synthesized off-DNA and their kinase potency against EPHA4 and EPHA2 isoforms was determined. These two compounds (CDD-2693 and CDD-2781) displayed effective biochemical potency on tested kinases with below 10 nM level of inhibition, especially CDD-2693, which showed picomolar efficacy with a K i value of 0.8 nM against EPHA4. Due to the excellent potency of these hits, we further studied the kinase selectivity more broadly. We determined the selectivity profile using a 468 human kinase panel and demonstrated that CDD-2693 showed excellent inhibition against the EPH family, especially outstanding potency on EPHA4, EPHA2, EPHA5, EPHB2, but no isoform selectivity, CDD-2693 also showed binding activities to SRC, FGR, and YES1 kinases. From these experiments, we conclude that DEC-Tec is an especially useful platform for generating potent and selective inhibitors of biologically validated kinase targets. To further optimize the hit compound CDD-2693, we synthesized several analogs to acquire SAR. During these studies, we found more potent inhibitors, including CDD-2905, CDD-3167, and CDD-2967, demonstrating potency below 5 nM. To determine the molecular effects of these compounds, we assessed the cellular potency of the most powerful selective compounds using the NanoBRET intracellular kinase assay against EPH and SRC family members. Remarkably, CDD-3167 and CDD-2693 selectively displayed significant cellular potency against EPHA2, EPHA4, EPHA5, and EPHB2, but did not show appreciable cellular activity on SRC, FGR, and YES1 off-targets. Taken together, we conclude that these potent inhibitors demonstrate exceptional selectivity toward EPH family members versus SRC family members. Additionally, we tested the metabolic stability of hit compounds, including their selected analogs, utilizing a human/mouse liver microsomal assay. The half-life of CDD-2693, CDD-2781, CDD-2969, and CDD-2970 had a significant improvement in metabolic stability over 120 min on HLMs and a notable stability over 25 min on MLMs. These inhibitors have the potential to target a major vulnerability in numerous pathologies – from benign conditions such as endometriosis to malignant cancers. We highlight the fact that EPHs are overexpressed and may serve as powerful therapeutic targets. Our study first reported the overexpression of EPHs in endometriosis and correlated the expression levels with the disease invasiveness. We also emphasized the specific niche for the advancement of nonhormonal therapeutic agents for managing pain syndrome and infertility issues in patients suffering from endometriosis. Additionally, in gynecological cancers, EPHs and their ligands are up-regulated in both endometrial and ovarian cancers ( 15 , 57 – 62 ). Higher EPHA2 expression level is correlated with more advanced disease stages and poorer survival rates in endometrial cancer patients ( 15 , 62 ). In ovarian cancer patients, EPHA1/2/4/8 have all been reported to be up-regulated and linked with adverse clinical prognosis ( 57 – 59 , 63 , 64 ). Prior literature on Ewing sarcoma also identifies EPHA2 as an oncogenic driver of therapeutic significance ( 23 , 24 ). In addition to cancers, EPHs are also indispensable in multiple physiological processes of reproduction, such as implantation ( 14 , 65 ) and placentation ( 66 , 67 ). Thus, perturbed ephrin-EPHs signaling pathways are observed in severe pregnancy complications, namely tubal pregnancies ( 68 , 69 ) and preeclampsia ( 70 – 72 ). Inhibitors targeting EPHs can also be valuable tools in advancing the mechanistic studies of ephrin-EPHs signaling pathways during developmental events. Given the profound detrimental effects of EPHs in both physiological and disease conditions, and the limited interventions or perturbation tools available, there is a dire need for the development of innovative compounds and therapies targeting EPHs. In this study, we refined the categories of EPH inhibitors in a high-throughput manner and bridged the gap of potent pan-ephrin receptor kinase inhibitors by developing two pan-EPH inhibitors and exemplified their efficacy in the biological context of endometriosis and Ewing sarcoma.

EPHA2 (PDB ID: 5I9Y) and EPHA4 (PDB ID: 2Y6O) crystal structures were retrieved from the protein data bank (PDB). The proteins were prepared using Schrodinger Suite Release 2022-1 ( 73 ) with default settings. A grid for each protein was generated at the site of the ligand in the crystal structure. CDD-2693 and its derivative compounds were prepared using the LigPrep program ( 74 ). 3D conformations and protonation states were generated with the Epik program ( 75 ). The processed compounds were then docked into the ATP-binding pockets of the two proteins using Glide in the extra precision mode ( 76 ). The acquired poses were visualized and analyzed using Maestro and PyMOL program ( 77 ). The normal endometrium (proliferative phase, ProEN, N = 5; early secretory, ESE, N = 4; midsecretory, MSE, N = 5) and endometriotic lesions (peritoneal, PeL, N = 3; deep infiltrating, DiE, N = 6; endometrioma, OMA, N = 10) were provided by a board-certified gynecological pathologist who also reviewed EPHA2 staining results. Fixed paraffin-embedded tissues underwent antigen retrieval in 10 mM citrate buffer with 0.5% Tween, pH 6.0, in a microwave for 20 min. After cooling on ice for 30 min while submerged in antigen retrieval buffer, endogenous peroxides were neutralized with a 10 min incubation in hydrogen peroxide, blocked with avidin and biotin, and blocked in 3% BSA for 60 min at room temperature. Sections were incubated with primary antibody (EPHA2, R&D Cat. # AF3035, 1:20 dilution) diluted in 3% BSA and incubated overnight at 4 °C. The following morning, the primary antibody was detected with a biotinylated secondary antibody followed by incubation with a signal amplification avidin/biotin complex (Vector Labs, PK-6100) and developed with DAB peroxidase substrate (Vector Labs, SK-4100). Sections were counterstained with hematoxylin (Sigma), dehydrated, and mounted using Permount mounting medium (VWR). Peroxidase-labeled and H&E-stained slides were imaged using the Aperio AT2 slide scanner (Thomas Huynh) in the Department of Veterinary Medicine and Surgery at the University of Texas MD Anderson Cancer Center. EPHA2 signal was quantified using ImageJ following a previously published protocol ( 78 ) and analyzed using a Kruskal–Wallis test for nonparametric datapoints with a Dunn’s multiple comparison posttest to assess the difference between the disease subtypes and endometrial phases. 12Z cells were acquired from ATCC and maintained in DMEM supplemented with 10% FBS. A673 cells were acquired as a kind gift from Jason Yustein and Atreyi Dasgupta (Baylor College of Medicine, TX) and maintained in DMEM supplemented with 2 mM glutamine and 10% FBS. For small molecules tested with EPHrin-A5 stimulation, 0.6 × 10 5 cells were seeded on 12-well plates and allowed to attach overnight and change the medium to DMEM supplemented with 2% charcoal-stripped FBS 24 h after seeding. Eighteen hours after medium change, cells were pretreated for 15 min with small-molecule inhibitors at 2× concentration and then supplemented with Ephrin-A5 (R&D Systems, #374-EA-200) to a final concentration of 0.5 µg/mL for 15 min. Cells were harvested on ice, lysed with 1× LDS sample buffer (Invitrogen, #NP0007), and denatured at 95 °C for 15 min. Then, lysates were electrophoresed by SDS-PAGE gel and probed with pEPHA2/3/4/5 (Cell Signaling, #D10H1), EPHA4 (Invitrogen, #37-1600), and GAPDH (Proteintech, HRP-60004) antibodies. Densitometry analysis was performed using ImageJ ( 79 ) and analyzed using GraphPad Prism. For cell viability assays, cells were plated in the 96-well at the density of 4,000 cells per well to incubate overnight and change the medium to DMEM supplemented with 2% charcoal-stripped FBS 24 h after seeding. Eighteen hours after medium change, cells were treated with corresponding small molecules. Seventy-two hours after the small-molecule treatment, cell viability was assessed by the CellTiter-Glo Luminescent Assay following the manufacturer’s protocol (Promega, #G7570). For the gene expression assay, 1 × 10 5 12Z cells were seeded on 12-well plates and allowed to attach overnight. The medium was changed to DMEM supplemented with 2% charcoal-stripped FBS 24 h after seeding. Eighteen hours after medium change, cells were pretreated for 30 min with small-molecule inhibitors at 2× concentration and then supplemented with IL-1β (R&D systems, #201-LB-005/CF) to the final concentration to 0.05 ng/mL for 3 h. Total RNAs were subsequently harvested by the RNeasy Kit (QIAGEN, #74004) following the manufacturer’s protocol. Reverse transcription was performed using 500 ng of total RNA as input using qScript cDNA SuperMix (Quantabio, #95048-025). The PTGS2 gene was amplified using primers: F-5′CGGTGAAACTCTGGCTAGACAG3′, R-5′GCAAACCGTAGATGCTCAGGGA3′. GAPDH was used as the internal control for normalization, with primers: F-5′GTCTCCTCTGACTTCAACAGCG3′, R-5′ACCACCCTGTTGCTGTAGCCAA3′. The final fold-change for plotting was calculated compared to vehicle treatment with IL-1β stimulation. Studies using human specimens were conducted under the H-21138 protocol approved by the Institutional Review Board at Baylor College of Medicine. Human endometrial epithelial organoids from endometrial tissues of patients with confirmed endometriosis were established as previously described ( 80 ). Organoids were plated in Matrigel (Corning Catalog #354230) in a 24-well plate and allowed to grow for 5 d in complete medium (Advanced DMEM/F12 supplemented with 1× N2 supplement (Life Technologies, #17502048), 1× B27 supplement (Life Technologies, #12587010), 100 μg/mL Primocin (Invivogen, #ant-pm-1), 2 mM L-glutamine (Life Technologies, #25030024), 500 nM A83-01 (Tocris, #2939), 10% R-spodin-1 conditioned media (obtained from the Center for Digestive Diseases, Baylor College of Medicine), 10 nM Nicotinamide (Sigma, #N0636-100G), 1.25 mM N-acetyl-L-cysteine (Sigma, #A9165-5G), 10% NOGGIN conditioned media (obtained from the Center for Digestive Diseases, Baylor College of Medicine), 100 ng/mL FGF10 (Peprotech, #100-26), 50 ng/mL HGF (Peprotech, #120-38), 50 ng/mL EGF (Peprotech, #AF100-15), and 10% WNT3a conditioned media (obtained from the Center for Digestive Diseases, Baylor College of Medicine). Fully formed organoids were then treated with 5 µM of CDD-2693 and CDD-3167 for a total of 72 h. For the small-molecule treatments, organoids were cultured in minimal differentiation medium [Advanced DMEM/F12 supplemented with 1× N2 supplement (Life Technologies, #17502048), 1× B27 supplement (Life Technologies, #12587010), and 100 μg/mL Primocin (Invivogen, #ant-pm-1)]. Organoid diameters were analyzed by ImageJ ( 79 ). The KINOMEscan TM Profiling Service from Eurofins/DiscoverX provided the TREEspot kinase dendrogram showing the kinase selectivity profile of the hit compound (CDD-2693) against 468 kinases ( 31 , 47 ).

Methods and procedures for the synthesis of compounds, DEC-Tec affinity selection, biochemical assays, NanoBRET™ kinase target engagement intracellular assay, thermal shift assay, and metabolic stability assay are available in SI Appendix . We assessed gene expression in GSE141549 to evaluate expression of genes encoding kinases that were up-regulated in human endometriotic lesion specimens compared to normal and patient eutopic endometrium. The dataset represents specimens from CE (N = 43); PE, (N = 104); PeL, (N = 71); DiE (N = 167); and OMA (N = 25). Datasets are presented as box and whisker plots of the log2Expression (mean, with min-max expression). Lesion expression of EPHA2 and EPHA4 was evaluated using multiple testing with an FDR < 0.05 of each condition (PeL, OMA, DiE) versus CE. Analysis of EPHA2 and EPHA4 in endometrial biopsies from the normal eutopic endometrium was performed in a previously published dataset ( GSE51981 ). The tissues represent specimens obtained from proliferative endometrium (PE, N = 20); early secretory endometrium ( ES, N = 6), and midsecretory endometrium ( MS, N = 8). Log2 expression data for EPHA2 and EPHA4 were plotted as mean, with min-max expression, and were analyzed by a one-way ANOVA with a Tukey’s postcomparison test. EPHA2 and EPHA4 genetic variants associated with endometriosis were identified by mining the FinnGen database ( 46 ). Datasets associated with the following phencodes were obtained: N14_ENDOMETRIOSIS (N = 10,029, control cases, N = 81,593); N14_ENDOMETRIOSIS ASRM_STAGE1_2 (N = 3,712, control cases, N = 143,349); N14_ENDOMETRIOSIS_ASRM_STAGE3_4 (N = 5,028, control cases, N = 142,033); N14_ENDOMETRIOSIS_DEEP (N = 1,841, control cases, N = 142,033); N14_ENDOMETRIOSIS_FALLOPIAN_TUBE (N = 135, control cases, N = 81,593); N14_ENDOMETRIOSIS_INTESTINE (N = 226, control cases, N = 81,593); N14_ENDOMETRIOSIS_NOS (N = 1,769, control cases, N = 81,593); N14_ENDOMETRIOSIS_OVARY (N = 3,889, control cases, N = 81,593); N14_ENDOMETRIOSIS_PELVICPERITONEUM (N = 3,620, control cases, N = 81,593); N14_ENDOMETRIOSIS_RECTPVAGSEPT_VAGINA (N = 1,626, control cases, N = 81,593); N14_ENDOMETRIOSIS_UTERUS (N = 2,866, control cases, N = 81,593); and N14_ENDOMET_INFERT (N = 1,956, control cases, N = 83,912). EPHA2 and EPHA4 variants associated with endometriosis were calculated using P < 0.001 and P < 0.01.

Appendix 01 (PDF)