The PDB files of the EGFR–gefitinib complex structure, 4I22, and the CRBN–thalidomide complex structure, 6BN7, were downloaded from the Protein Data Bank (https://www.rcsb.org/). Water and ions were removed from the structures, and only the proteins and ligands were retained. CoDockPP was used to perform docking between the EGFR–gefitinib complex and the CRBN–thalidomide complex (Kong et al., 2019). Site-constraint search was used with a distance constraint of 20 Å between the tethering atoms from each ligand (Figure 1). A knowledge-based scoring function was used to rank the binding poses. Finally, the top 100 energetically favorable poses were retained.

HCC827 and H1975 cell lines were purchased from ATCC (United States) and cultured according to the manuals. Cells were grown in monolayers in appropriate media supplemented with 10% fetal bovine serum and 1% antibiotic/antimycotic solution. Ba/F3 cells stably expressing EGFR^L858R were obtained by infection with lentivirus produced by co-transfection of pLVX-EGFR^L858R, psPAX2, and pMD2.G into 293T cells, followed by selection with puromycin.

Cells (5 × 10³ for 48 h or 1 × 10³ for 96 h) were seeded in 96-well plates, allowed to attach overnight, and then incubated with various concentrations of compounds for 48 h or 96 h. The viability of cells was determined using Cell Counting Kit-8 (CCK-8; Dojindo), and IC₅₀ was calculated using GraphPad Prism (version 6.0). For colony formation assays, cells (1 × 10³ cells/well) seeded in 24-well plates were treated with different concentrations of compounds and replaced with fresh media with compounds every 4 days. After 7 days, the cells were washed with PBS, fixed with methanol, and stained with 0.1% crystal violet solution.

Cells at 70%–90% confluence were lysed in RIPA buffer with complete protease inhibitor cocktail (Roche Applied Science, Penzberg, Germany). Lysates were quantified, separated on SDS-PAGE gels, and transferred onto PVDF membranes, followed by the blockade with 5% milk in TBS+ 0.1% Tween-20 buffer for 1 h at room temperature. The following primary antibodies were used: anti-EGFR (1:50000, ABclonal Technology, Wuhan, China), anti-pEGFR (Tyr1068, 1, Cell Signaling Technology, MA, United States), anti-AKT (1:10000, ABclonal Technology, Wuhan, China), anti-pAKT (1:10000, ABclonal Technology, Wuhan, China), and anti-β-actin (1:10000, ABclonal Technology, Wuhan, China). Protein bands were detected using Tanon High-signal ECL Western Blotting Substrate (Tanon Science and Technology Co., Ltd., Shanghai, China). The band density of the protein of interest was quantified using ImageJ.

HCC827 cells incubated with control vehicle (DMSO), 10 nM of compound 14, or 10 nM of gefitinib for 48 h were collected, stained with FITC Annexin V and propidium iodide (PI), and analyzed using a BD FACSAria II Flow Cytometer. The percentage of apoptotic cells in total cells was designated as the apoptotic index.

Female BALB/c nude mice (5–6 weeks old) were purchased from Changzhou Cavens Laboratory Animal Co., Ltd. (China) and housed under pathogen-free conditions. HCC827 cells (5 × 10⁶) suspended in 150 μL PBS were inoculated subcutaneously on the left flank of the mouse. When tumor volume reached an average of 70–80 mm³, mice were randomly assigned to two groups (three mice in each group). Compound 14 (30 mg/kg, dissolved in 40% PEG 400 and 5% Tween-80) or vehicle (40% PEG 400 and 5% Tween-80) was administered to mice via intraperitoneal injection every 2 days for 21 days. The tumor volume was calculated as follows: tumor volume (mm³) = (length × width²)/2. All animal experiments were approved by the Ethics Committee for laboratory animals, The Nanjing Han & Zaenker Cancer Institute (File number: OGKQSPF/SQ-112).

All experiments were repeated at least three times independently with similar results, and the results are presented as the mean values ±standard deviation (SD). The significance of differences among groups was assessed using ordinary one-way ANOVA with multiple-comparisons tests using Prism 8 (*p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001). p < 0.05 was considered to indicate a statistically significant difference.

Molecular docking methods were used to predict the ternary complex structure of the POI-PROTAC-E3 ligase in previous studies (Pereira et al., 2023; Ignatov et al., 2023; Zaidman et al., 2020). In this study, computational methods were used for the rational design of PROTACs before conducting costly and time-consuming synthesis experiments. The complex structure of gefitinib with the L858R/T790M mutant EGFR (PDBID: 4I22) was chosen as the POI structure for docking (Gajiwala et al., 2013). By analyzing the solvent availability of the atoms in gefitinib, we selected the solvent-exposed methoxyl group as the attachment point (Figure 1A; Ribeiro et al., 2019). Considering synthesis feasibility, the carbon atom was removed, and the oxygen atom was used as a tethering atom to connect with potential linkers (Figure 1). For the E3 ligase part, the complex structure of thalidomide with CRBN (PDBID: 6BN7) was extracted and used for docking (Nowak et al., 2018). Accordingly, the solvent-available atoms from thalidomide were selected as the tethering atoms (Figure 1). Protein docking was performed using CoDockPP, setting a distance constraint of 20 Å between the tethering atoms from the warhead and the E3 ligand (Kong et al., 2019). The top 100 complex poses, ranked by a knowledge-based scoring function, were retained for further analysis (Supplementary Figure S1). For the top 100 poses, the distance distribution between the tethering atoms is shown in Supplementary Table S1. No pose satisfied with a distance of less than 7 Å, and the largest fraction of poses fell within an atom distance of 15 to 20 Å. We hypothesized that the ideal linker length should be tolerant, allowing as many energetically favorable poses as possible to stabilize the ternary structure. As the C–C bond length was approximately 1.5 Å, we designed and synthesized a series of compounds with different linker length composed of approximately 6–15 atoms, as shown in Table 1. Detailed information on the synthesis and characterization of the compounds is provided in the Supplementary Schemes.

The degradation rates of compounds 1–15 at 10 nM were evaluated in HCC827 (EGFR^Del19) cells, as shown in Table 1. To connect the tethering atoms with different lengths of alkyl chains, we synthesized compounds 1–3. The short linker with six carbon atoms in compound 1 exhibited no degradation effect, consistent with the distance distribution of the tethering atoms. As shown in Supplementary Table S1, when the distance between the tethering atoms was less than 11 Å, only a limited number of energetically favorable protein–protein poses satisfied this distance condition. With an increased linker length, compounds 2 and 3 showed weak degradation activities. To reduce linker flexibility, piperazine and carbonyl groups were introduced into the compound design. Among these, compounds 6, 7, and 9–13, which featured sufficiently long and relatively rigid linkers, demonstrated significantly enhanced degradation activities.

Based on the binding mode of gefitinib with EGFR (PDBID: 4I22), limited interactions were observed between the morpholine ring and the protein (Gajiwala et al., 2013). Therefore, the extended morpholine ring, which was intended to improve solubility, was removed from compound 12 to reduce molecular weight. The resulting compound 14 exhibited comparable degradation activity to compound 12 (66.3% ± 3.5% for 14 versus 60.0% ± 6.9% for 12, Table 1) with significantly decreased molecular weight. Compound 15, which incorporated an extra methyl group on the glutarimide moiety of thalidomide in compound 14, completely lost its degradation ability (Table 1). It is reported that methylation at this position entirely disrupts the binding of thalidomide to cereblon (Fischer et al., 2014), implying that E3 ligase binding is essential for EGFR degradation. The predicted ternary structure of the most active molecule, compound 14, is illustrated in Figure 1C. Compound 14 presents an appropriate conformation to bind to the top 1 docking pose of EGFR and CRBN, indicating its potential to induce ternary structure formation.

We further evaluated the degradation potencies of compounds 12 and 14, which were identified as the most potent degraders in the primary study. Both compounds could efficiently degrade EGFR^Del19 in HCC827 cell lines at 24 h treatment with a DC₅₀ value of 1.944 nM for 12 (D_max = 85.1%) and 0.261 nM for 14 (D_max = 91.2%) (Figure 2A). Concordantly, the levels of both pEGFR and pAKT also markedly reduced upon EGFR degradation. Compound 14 also efficiently degraded EGFR^L858R in Ba/F3 cells stably expressing the EGFR^L858R mutant with a DC₅₀ value of 20.57 nM (Figure 2B). However, neither compound degraded the EGFR^WT protein in H1299, H460, and HeLa cell lines nor the EGFR^L858R/T790M protein in the H1975 cell line, up to 1 μM (Figure 2C). We also compared the degradation efficacy of compound 14 and MS154 in both HCC827 and Ba/F3 cell lines. The results indicate that compound 14 showed remarkable improvements in the degradation of EGFR^Del19 and EGFR^L858R than MS154 (Supplementary Figure S2).

Next, the levels of the EGFR^Del19 protein in HCC827 cells incubated with compound 12 or 14 at a dose of 10 nM at multiple time points were examined (Figure 3). The degradation of EGFR^Del19 could be observed as early as 2 h after treatment. Compound 14 decreased the half-life of the EGFR^Del19 protein at 8.6 h and compound 12 at 6.8 h; however, compound 14 induced significantly greater degradation of EGFR, with a maximum degradation (D_max) of 85% after 24 h of incubation (Figures 3A,B). Meanwhile, compound 14 led to more marked decreases in pEGFR and pAKT compared with compound 12. In addition, compound 14 mediated a sustained degradation of the EGFR^Del19 protein over 96 h (Figure 3C). More interestingly, a reduction in the EGFR^Del19 protein induced by an 8 h-incubation with 10 nM of compound 14 was constantly maintained for over 72 h, even after compound removal (Figure 3D).

Furthermore, HCC827 cells were incubated with cycloheximide (CHX) with or without compound 14, and the levels of the EGFR^Del19 protein were determined. As a result, after 4 h and 8 h of cycloheximide treatment, control DMSO-treated samples still retained approximately 70% and 55% of EGFR proteins, whereas, in the presence of compound 14, only 34.5% and 12.5% remained (Figure 4).

To clarify whether these degradations were dependent on E3 ligase complex recruitment, HCC827 cells were pretreated with either the proteasome inhibitor MG132 or the NEDD8-activating enzyme inhibitor MLN4924 for 2 h and continuously incubated with compound 14 for an additional 8 h. As shown in Figures 5A,B, both MG132 and MLN4924 restored EGFR^Del19 degradation induced by compound 14, implying that EGFR degradation by compound 14 was dependent on the ubiquitination and neddylation system.

Moreover, competitive binding experiments were conducted with either the CRBN ligand thalidomide (Figure 6A) or the EGFR ligand gefitinib (Figure 6B). Both thalidomide and gefitinib could completely rescue the EGFR^Del19 reduction induced by compound 12 or 14 at 500 times the degrader dose, whereas a slight decrease in EGFR^Del19 was still observed at 100 times the dose. These results suggested that compound 14 directly targeted EGFR^Del19 and mediated EGFR^Del19 degradation through the CRBN-dependent mechanism.

To study the biological effects of the decrease in PROTAC-mediated EGFR^Del19 on cancer cell survival, HCC827 cells were treated with compound 12, compound 14, gefitinib, or thalidomide for 48 h or 96 h, and the number of viable cells was calculated using CCK-8 (Figure 7). With a 48 h-incubation, the IC₅₀ values of compound 12, compound 14, and gefitinib were approximately 30.68 ± 6.15 nM, 8.29 ± 3.31 nM, and 4.74 ± 1.19 nM, respectively (Figure 7A). With a 96 h-incubation, the IC₅₀ values of compound 12, compound 14, and gefitinib were approximately 25.64 ± 3.8 nM, 4.91 ± 0.78 nM, and 3.93 ± 1.4 nM, respectively (Figure 7A). In particular, compound 14 exhibited cytotoxic activity similar to that of gefitinib. Thalidomide had no influence on the viability of HCC827 cells up to 200 nM. To explore the prolonged efficacy, HCC827 cells were treated with 10 nM of compound 14 or gefitinib for 8 h, washed with PBS, and cultured with fresh media for an additional 24, 48, 72, or 96 h. Intriguingly, gefitinib-pretreated cells gradually continued to proliferate after the removal of compounds, whereas compound 14-pretreated cells almost completely stopped proliferating (Figure 7B), illustrating the distinct pharmacological mechanisms of compound 14 and gefitinib.

In addition, compound 12, compound 14, and gefitinib did not affect the viability of H1299, HeLa, and H1975 cells up to 1 μM, suggesting high specificity of PROTACs toward the EGFR^Del19 mutant (Figure 7C). Concordantly, compound 14 markedly attenuated the proliferation of HCC827 cells compared with control vehicle-treated cells (Figure 7D) and formed fewer and smaller colonies in HCC827 cells than the control in colony formation assays (Figure 7E).

In the above study, numerous apoptotic cells were observed under the microscope following treatment with compound 14. Hence, we performed FACS analysis to further calculate the apoptotic cell population upon treatment with compound 14 or gefitinib in HCC827 cells. As expected, both gefitinib and compound 14 led to at least a two-fold increase in the number of apoptotic cells, and there was little difference between compound 14 and gefitinib (Figures 8A,B).

We evaluated the efficacy of compound 14 to inhibit the growth of HCC827 cells using a xenograft immunodeficient mouse model. As shown in Figure 9A, compound 14 (30 mg/kg) markedly suppressed the growth of implanted tumors. The mean volume of HCC827 tumors treated with compound 14 (56.58 ± 10.91 mm³) was noticeably decreased compared with that of control vehicle mice (350.86 ± 191.41 mm³) (Figure 9A, p < 0.05). Meanwhile, compound 14 effectively degraded the EGFR^Del19 protein in vivo (Figure 9C). Furthermore, mice-bearing HCC827 xenografts were administered a one-time intraperitoneal injection of compound 14 at 30 mg/kg. Intriguingly, as shown in Figures 9D,E, one dose of compound 14 led to a mean tumor volume inhibition of 79% (P = 5.4E-05 versus vehicle control) 18 days after treatment, indicating the long-term effect for this compound. None of the mice showed signs of wasting (Figures 9B,F).