A549, Calu-1, SK-LU-1, and HEK293T cells were obtained from ATCC. All cell lines were cultured under standard conditions (37°C in a humidified atmosphere with 5% CO₂) in DMEM containing 10% FBS (Nguyen et al., 2014; Nguyen et al., 2015; Yang et al., 2017; Igarashi et al., 2018; Kato et al., 2020). To establish erlotinib-resistant cells (ER cells), cells were cultured in the presence of 15–30 μM erlotinib (HY-50896, MedChemExpress) for 6 months. The culture medium containing erlotinib was replaced every 3–4 days. After 6 months, cells were frozen in a medium containing 90% FBS and 10% DMSO and stored in liquid nitrogen. ER cells were cultured in DMEM containing 10% FBS and 10 µM erlotinib at 37°C with 5% CO₂. Afuresertib (S7521) and Ipatasertib (S2808) were purchased from Selleckchem.

0.5–1.0 × 10⁴ cells were resuspended in 50 µL of Geltrex (A1413202, Thermo Fisher), placed in 8-well chambered coverglasses, and incubated at 37°C for 30 min to allow the Geltrex to solidify. Subsequently, 250 µL of DMEM containing 10% FBS was added. The cells were allowed to form spheroids for 7 days in a cell culture incubator at 37°C with 5% CO₂. After 7 days, the spheroids were collected using Corning Cell Recovery Solution (CLS354253, Millipore Sigma) and then resuspended at a concentration of 1 spheroid per 1 µL of Geltrex in 24-well plates coated with Sigmacote (SL2, Millipore Sigma). The plates were incubated at 37°C for 30 min to allow the Geltrex to solidify, after which 500 µL of DMEM containing 10% FBS was added. The culture medium was changed on day 3 or 4. Images of cells and spheroids were captured using phase-contrast microscopy (AXIO Observer Z1, Zeiss). The total spheroid area was calculated using NIH Fiji software.

Cells were lysed in lysis buffer (1% NP40, 10% glycerol, 50 mM NaCl, 20 mM Tris-HCl [pH 7.4]) supplemented with cOmplete, Mini, EDTA-free Protease Inhibitor Cocktail (11836170001, Sigma) on ice (Nguyen et al., 2014; Nguyen et al., 2015; Yang et al., 2017; Igarashi et al., 2018; Senoo et al., 2019; Kato et al., 2020; Senoo et al., 2021). The lysates were centrifuged at 16,000 g for 10 min at 4°C, and the supernatants were collected. Proteins were separated by SDS-PAGE and transferred onto Immobilon-FL Transfer Membranes (Millipore). The membranes were blocked in PBS-T (PBS containing 0.05% Tween 20) supplemented with 3% BSA at room temperature for 30 min and then incubated with primary antibodies in PBS-T supplemented with 3% BSA at 4°C overnight. The antibodies used are GRHL2 (ab271023, Abcam), HER3 (12,708, Cell Signaling), E-cadherin (14,472, Cell Signaling), β-catenin (ab32572, Abcam), EGFR (4267, Cell Signaling), AKT (9272, Cell Signaling), phospho-AKT at S473 (4060, Cell Signaling), p38 (8690, Cell Signaling), phospho-p38 (4511, Cell Signaling), ERK1/2 (9107, Cell Signaling), phospho-ERK1/2 (4370, Cell Signaling), GAPDH (MA5-15738, Invitrogen), PARP (9542, Cell Signaling), Caspase-3 (9665, Cell Signaling), and LC3 (PM036, MBL). The membranes were washed three times in PBS-T, followed by incubation with appropriate fluorescently labeled secondary antibodies at room temperature for 1 h. After washing the membranes three times in PBS-T, fluorescence signals were detected using a Typhoon biomolecular imager (Amersham). Images were analyzed using NIH Fiji software.

The following sequences were cloned into pLKO.1 puro (8453, Addgene) to generate shRNA plasmids (Adachi et al., 2020; Murata et al., 2024). Scramble: CCT​AAG​GTT​AAG​TCG​CCC​TCG​CTC​GAG​CGA​GGG​CGA​CTT​AAC​CTT​AGG; LEF1: CCA​CAC​TGA​CAG​TGA​CCT​AAT​CTC​GAG​ATT​AGG​TCA​CTG​TCA​GTG​TGG, CCA​TCA​GAT​GTC​AAC​TCC​AAA​CTC​GAG​TTT​GGA​GTT​GAC​ATC​TGA​TGG, GCA​CGG​AAA​GAA​AGA​CAG​CTA​CTC​GAG​TAG​CTG​TCT​TTC​TTT​CCG​TGC; GRHL2: GCC​GAT​TAC​AAG​GAG​AGC​TTT​CTC​GAG​AAA​GCT​CTC​CTT​GTA​ATC​GGC, CCT​TCA​AAG​CAG​ATG​AAA​GAA​CTC​GAG​TTC​TTT​CAT​CTG​CTT​TGA​AGG, GCT​GAA​GAT​TTC​ACA​CCA​GTT​CTC​GAG​AAC​TGG​TGT​GAA​ATC​TTC​AGC; EVX1: CGC​CTT​CTA​CAC​TTA​CAT​GAT​CTC​GAG​ATC​ATG​TAA​GTG​TAG​AAG​GCG, GTC​GGA​TTT​CTA​TGA​AGA​AAT​CTC​GAG​ATT​TCT​TCA​TAG​AAA​TCC​GAC, CCG​CCC​TAA​ACC​TGC​CGG​AAA​CTC​GAG​TTT​CCG​GCA​GGT​TTA​GGG​CGG; ZFP57: AGC​AAG​TCT​TTC​AGC​TCA​TTT​CTC​GAG​AAA​TGA​GCT​GAA​AGA​CTT​GCT, TGT​TAT​GTC​GGA​AAC​CTT​TAA​CTC​GAG​TTA​AAG​GTT​TCC​GAC​ATA​ACA, AGG​TCC​CAG​GAA​CCC​ATA​TTT​CTC​GAG​AAA​TAT​GGG​TTC​CTG​GGA​CCT; MEF2C: GCC​TAG​AAT​TTG​ATA​CGC​TTT​CTC​GAG​AAA​GCG​TAT​CAA​ATT​CTA​GGC, CGT​GGA​GAC​GTT​GAG​AAA​GAA​CTC​GAG​TTC​TTT​CTC​AAC​GTC​TCC​ACG, CCC​AAT​GAA​TTT​AGG​AAT​GAA​CTC​GAG​TTC​ATT​CCT​AAA​TTC​ATT​GGG; HER3: AGG​TTA​GGA​GTA​GAT​ATT​GAC​TCG​AGT​CAA​TAT​CTA​CTC​CTA​ACC​T, AAT​TCT​CTA​CTC​TAC​CAT​TGC​TCG​AGC​AAT​GGT​AGA​GTA​GAG​AATT,​ TAT​ATG​AAT​CGG​CAA​CGA​GAT​CTC​GAG​ATC​TCG​TTG​CCG​ATT​CAT​ATA; E-cadherin (CDH1): CCA​AGC​AGA​ATT​GCT​CAC​ATT​CTC​GAG​AAT​GTG​AGC​AAT​TCT​GCT​TGG, CGA​TTC​AAA​GTG​GGC​ACA​GAT​CTC​GAG​ATC​TGT​GCC​CAC​TTT​GAA​TCG, CCA​ACC​CAA​GAA​TCT​ATC​ATT​CTC​GAG​AAT​GAT​AGA​TTC​TTG​GGT​TGG.

To express GFP, pLentiCMV-GFP-Puro was generated by removing the Met gene from pLentiCMV-MetGFP-Puro (Addgene, 37,560). To express HER3, pLentiCMV-HER3-Puro was generated by replacing GFP with HER3, which was cloned from cDNA of A549 cells, in pLentiCMV-GFP-Puro. To express E-cadherin, pHAGE-CDH1 was used (Addgene, 116722).

HEK293T cells were seeded at 5 × 10⁵ cells per well in a 6-well dish and cultured for 24 h (Adachi et al., 2020; Murata et al., 2024). pLX313 carrying firefly luciferase (118,017, Addgene), pLKO.1 puro carrying shRNAs, pLentiCMV-GFP-Puro, pLentiCMV-HER3-Puro, or pHAGE-CDH1was co-transfected into HEK293T cells along with pCMV-VSVG and pHR-CMV8.2ΔR using polyethylenimine (23966-1, Polysciences, Inc.) (Adachi et al., 2020; Murata et al., 2024). The culture medium was replaced 3 h after transfection. Culture media containing virus particles were collected 48 h after transfection. The virus was aliquoted and stored at −80°C. For lentiviral transduction, cells were seeded at 5 × 10⁵ cells/well in a 6-well plate and cultured for 24 h. Cells were then incubated with lentivirus in DMEM containing 10% FBS and 8 μg/mL polybrene for 24 h.

All work involving animals was conducted according to the guidelines established by the Johns Hopkins University Committee on Animal Care and Use. Nude mice (strain number: 002019) were purchased from The Jackson Laboratory. The tumorigenicity assay was conducted as described previously (Pirazzoli and Politi, 2014; Ota et al., 2016; Du et al., 2017). Cells were infected with lentiviruses expressing luciferase and selected for 7 days with hygromycin (250 μg/mL). A total of 1 × 10⁶ cells were resuspended in 100 μL of PBS and injected subcutaneously into the dorsal skin of nude mice. For erlotinib treatment, mice were subjected to intraperitoneal injection at 25 mg/kg/day every Monday through Friday for 4 weeks (Politi et al., 2010; Wu et al., 2013). To monitor tumor development, mice were injected with 150 μL D-luciferin (30 mg/mL) for 10 min and analyzed using the In Vivo Imaging System (IVIS, Xenogen). Total photon counts were quantified using Living Image software 4.7.4. Tumors were dissected at 4 weeks, and their weight was measured.

The region spanning exons 18–21 of the EGFR gene was PCR amplified from the cDNA of ER cells using the primers TCG​GCC​TCT​TCA​TGC​GAA​GGC and CAA​TGC​CAT​CCA​CTT​GAT​AGG​CAC. The PCR products were then analyzed by DNA sequencing.

Total RNA isolation, RNA library preparation, and transcriptome sequencing were conducted by Novogene (Sacramento, CA). Genes with an adjusted p-value <0.01 and a log2 fold change >8 were considered differentially expressed genes (DEGs). Transcription factors were selected based on the RNA-seq analysis data provided by Novogene.

To prepare a low-adhesion culture plate, 24-well cell culture plates were treated with 500 μL Sigmacoat (SL2-25 mL, Sigma-Aldrich) according to the manufacturer’s protocol. Before using the coated plate, wells were washed twice with 500 μL PBS. A total of 5 × 10⁴ cells in 0.5 mL were added to the 24-well coated plates and incubated for 24 h on a rotary shaker at 125 rpm at 37°C with 8% CO₂. Images were obtained using EVOS Cell Imaging Systems (Thermo Fisher) and a 10x/0.30 NA EVOS objective lens (AMEP4981, Invitrogen). Images were analyzed using NIH Fiji software.

ER cells were plated at 1 × 10⁵ cells per well in a 12-well plate and cultured for 24 h esiRNAs targeting E-cadherin (EHU090371, Sigma) or a Renilla luciferase-targeting negative control (EHURLUC, Sigma) were transfected into the ER cells using Lipofectamine RNAiMAX (Invitrogen) for 3 days. The cells were then used for Western blotting and the cell aggregation assay.

Gene knockout of GRHL2 and E-cadherin (CDH1) was performed using the GeneArt CRISPR Nuclease (OFP reporter) vector kit (A21174; Thermo Fisher) according to the manufacturer’s instructions (Adachi et al., 2020; Murata et al., 2024). The following target gRNA sequences were used: GRHL2: CCG​AAG​AGC​CTA​CAC​CAG​TG, E-cadherin: AAG​ATT​GCA​CCG​GTC​GAC​AA.

Cells were plated in 24-well plates, and total RNA was extracted using the GeneJET RNA Purification Kit (K0732; Thermo Fisher). The RNA was reverse-transcribed into cDNA using the ReadyScript cDNA Synthesis Mix (RDRT, Sigma-Aldrich). Quantitative PCR (qPCR) was performed on a QuantStudio 3 Real-Time PCR System (Applied Biosystems) with PowerUp SYBR Green Master Mix (A25741, Thermo Fisher Scientific). GAPDH served as the reference gene. The following primers were used: HER3: CAA​CTC​TCA​GGC​AGT​GTG​TCC and CCA​TCA​CCA​CCT​CAC​ACC​TC, GAPDH: AAG​GTC​ATC​CCT​GAG​CTG​AA and CTG​CTT​CAC​CAC​CTT​CTT​GA.

Statistical analysis was performed using GraphPad Prism. Specific statistical tests and significance thresholds are detailed in the figure legends.

To establish EGFR inhibition resistance in NSCLC cells, we cultured A549 NSCLC cells in the presence of the EGFR inhibitor erlotinib at 15–30 µM for 6 months (Figure 1A). This range of concentrations was chosen as sublethal doses based on our preliminary experiments, in which the proliferation of A549 cells was tested in the presence of 5–30 µM erlotinib. The prolonged erlotinib treatment was performed in three separate plates of cell cultures to obtain multiple independent resistant cell lines. After 6 months, A549 erlotinib-resistant cells (ER cells) from each culture showed resistance to erlotinib, with a dramatically increased half-maximal inhibitory concentration (IC₅₀) for cell proliferation, approximately 10-fold higher compared to the original A549 cells (Parental cells) (Figures 1B–D). In the absence of erlotinib, ER cells proliferated at a similar rate to Parental cells (Figure 1E). We also treated two other lung cancer cell lines, Calu-1 and SK-LU-1, with erlotinib for 6 months. We found that, similar to A549 cells, SK-LU-1 cells acquired resistance to erlotinib, whereas Calu-1 cells did not (Supplementary Figure S1). In this study, we focused on characterizing A549 ER cells.

To characterize ER cells under more physiological conditions, we cultured Parental and ER cells in Geltrex, an extracellular matrix composed of laminin, collagen IV, entactin, and heparin sulfate proteoglycans (El Harane et al., 2023). We found that ER spheroids grew much faster than Parental spheroids in this 3D culture condition (Figures 1F,G). Since Geltrex contains a mixture of laminin, collagen IV, entactin, and heparan sulfate proteoglycans, these ECM components may preferentially enhance the proliferation of ER cells. Alternatively, cells in Geltrex may experience limited oxygen availability, with resistant cells proliferating better under modestly hypoxic conditions. Consistent with this finding, when Parental and ER cells expressing luciferase were subcutaneously injected into nude mice, bioluminescence imaging showed that ER cells generated much larger tumors than Parental cells (Figure 1H). Similarly, ER cell-derived tumors were significantly heavier than those derived from Parental cells at 4 weeks after injection (Figures 1I, J). Notably, the development of tumors derived from Parental cells was significantly reduced when mice were treated with erlotinib, as indicated by bioluminescence imaging (Figure 1H) and tumor weight (Figures 1I, J). Conversely, erlotinib treatment did not affect the growth of tumors derived from ER cells (Figures 1H–J). These data demonstrate that ER tumors resist EGFR inhibition in vivo, which is consistent with the in vitro data.

To determine whether ER cells exhibit altered cell survival in response to erlotinib, we compared the effects of erlotinib on apoptosis and autophagy. Parental and ER cells were treated with 5 µM erlotinib for 24 h and analyzed by Western blotting using antibodies against apoptotic markers, including caspase-3 activation and PARP cleavage, as well as LC3 molecular species. We observed increased levels of caspase-3 but no evidence of its proteolytic activation or cleavage of its substrate, PARP, in either parental or ER cells treated with erlotinib (Figures 1K, L). Additionally, although LC3-I levels were decreased in ER cells compared to Parental cells, no conversion to active LC3-II was detected in either Parental or ER cells, regardless of erlotinib treatment (Figures 1K, L). These findings suggest that erlotinib does not induce apoptotic or autophagic cell death in either Parental or ER cells.

It has been reported that resistance to EGFR inhibition can be acquired through increased expression levels of EGFR, mutations in the EGFR gene, or partial deletions of the gene (Rosell et al., 2009; Yu et al., 2014; Vasan et al., 2019; Passaro et al., 2021; Cooper et al., 2022; Sattler et al., 2023; Tomuleasa et al., 2024). Accordingly, to investigate the mechanism underlying erlotinib resistance in ER cells, we tested the expression levels of EGFR by Western blotting. Surprisingly, we found that EGFR protein levels were decreased, rather than increased, in ER cells compared to Parental cells (Figures 2A, B). Therefore, the expression level of EGFR does not account for the resistance. We also analyzed the DNA sequence of the EGFR gene in ER cells. We found no resistance-associated mutations, such as T790M or L858R, nor any deletion of exon 19 in ER cells (Figure 2C). These data suggest that ER cells have established an EGFR-bypass drug resistance.

To decipher the EGFR-bypass mechanism underlying erlotinib resistance, we performed RNA-seq analysis in Parental and ER cells. We detected a total of 30,251 genes (Supplementary Table S1). Analysis of differentially expressed genes (DEGs) with a cut-off value of adjusted p-value <0.01 and log2 fold change >8 identified 2,629 DEGs. Gene ontology molecular function (GO:MF) analysis of these DEGs showed that the top enriched gene group is RNA polymerase II cis-regulatory region sequence-specific DNA binding (Figure 3A). Consistent with this, gene ontology analysis for biological processes (GO:BP) also revealed that the top enriched group is regulation of transcription by RNA polymerase II (Figure 3B). Since RNA polymerase II mediates mRNA transcription, these data suggest that alterations in transcription may play a vital role in the EGFR-bypass mechanism for erlotinib resistance in ER cells.

To test the role of mRNA transcription in erlotinib resistance, we identified 1,349 transcription factors in our RNA-seq analysis and focused on the top five whose mRNA levels were highly upregulated in ER cells (Figure 4A; Supplementary Table S1). These include LEF1, GRHL2, EVX1, ZFP57, and MEF2C (Figure 4A). We individually knocked them down in ER cells using shRNAs and determined the impact on the IC₅₀ of erlotinib. We found that the knockdown of GRHL2 (grainyhead-like 2), but not the other four transcription factors, significantly decreased the IC₅₀ relative to a scramble control (Figure 4B). These data indicate that GRHL2 is vital for resistance to EGFR inhibition in ER cells.

GRHL2 is a member of the highly conserved grainyhead-like transcription factor family, which plays an important role in the development, function, and integrity of epithelial cells, controlling processes such as neural tube closure, epithelial barrier function, and wound healing (Mlacki et al., 2015; Frisch et al., 2017; Kotarba et al., 2020). Additionally, GRHL2 has been implicated in tumorigenesis, particularly in epithelial-mesenchymal transition (Mlacki et al., 2015; Frisch et al., 2017; Kotarba et al., 2020). However, the role of GRHL2 in drug resistance against EGFR inhibition has not been previously reported. Western blotting confirmed that GRHL2 protein abundance is dramatically upregulated in ER cells (Figures 4C, D). To further test the role of GRHL2 in ER cells, we knocked it out using the CRISPR/Cas9 genome editing system (Figure 4E; Supplementary Figure S2). The complete loss of the GRHL2 protein was confirmed by Western blotting (Figures 4F, G). Supporting the data from shRNA knockdown, the knockout of GRHL2 significantly reversed the IC₅₀ for erlotinib in ER-GRHL2-KO cells (Figure 4H). Taken together, these data demonstrate that increased expression of GRHL2 mediates EGFR-bypass erlotinib resistance.

To identify critical downstream components of GRHL2 in erlotinib resistance, we hypothesized that other receptor tyrosine kinases (RTKs) might be involved in this mechanism when EGFR is inhibited. Among the 53 RTKs identified in our RNA-seq data, HER3 (human epidermal growth factor receptor 3) is known to be regulated by GRHL2 (Saini et al., 2023), leading us to hypothesize that increased levels of GRHL2 promote the expression of HER3 in ER cells. Supporting this hypothesis, Western blot analysis showed that HER3 protein levels are significantly higher in ER cells compared to Parental cells (Figures 5A, B). To determine whether the increased level of HER3 depends on GRHL2, we compared HER3 protein levels in ER cells and ER-GRHL2-KO cells. Indeed, the loss of GRHL2 significantly decreased HER3 abundance (Figures 5C, D). To understand how GRHL2 controls HER3 levels, we analyzed the expression of HER3 transcripts using qPCR. We found that while HER3 transcript levels are increased in ER cells compared to Parental cells, they were only partially reduced in ER-GRHL2-KO cells (Figures 5E, F). These data suggest that GRHL2 regulates HER3 protein levels through transcription as well as additional mechanisms, such as translation or protein stability. Furthermore, to assess the functional contribution of HER3 to drug resistance, we knocked down HER3 using shRNAs in ER cells (Figure 5G). We found that HER3 knockdown significantly decreased the IC₅₀ of erlotinib (Figure 5H). These data suggest that EGFR-bypass resistance to erlotinib is driven by GRHL2-dependent upregulation of HER3.

During cell culture, we observed ER cells adhering to each other more than Parental cells (Figure 6A). This observation suggested that ER cells may have increased cell-cell adhesion. To test this hypothesis, we performed a cell-cell adhesion assay by incubating dissociated cells under continuous gentle shaking for 24 h, followed by measuring the size of the cell aggregates as an indicator of cell-cell adhesion. Results showed that while Parental cells remained largely dissociated, ER cells were associated with each other and formed large aggregates (Figures 6B, C). These data suggest that ER cells have an increased ability for cell-cell adhesion. Our findings are consistent with the RNA-seq data showing enrichment of cell-cell adhesion mediator activity (GO:MF) (Figure 3A) and homophilic cell adhesion via plasma membrane adhesion molecules (GO:BP) (Figure 3B).

Western blotting showed that the cell adhesion molecule E-cadherin and its functional partner, β-catenin, are considerably upregulated in ER cells (Figures 6D, E). To test the role of E-cadherin in increased cell adhesion in ER cells, we knocked down E-cadherin using esiRNAs (Figures 6F, G). E-cadherin knockdown greatly decreased cell-cell adhesion in our cell aggregation assay (Figures 6H, I). To assess the role of E-cadherin in drug resistance, we knocked E-cadherin using CRISPR/Cas9 in ER cells (Figures 7A–C). The loss of E-cadherin was confirmed by Western blotting (Figure 7B). Strikingly, ER-Ecad-KO cells exhibited a significantly reduced IC₅₀ for erlotinib (Figure 7D). These data demonstrate that E-cadherin is critical for EGFR-bypass drug resistance.

To investigate the relationship between the roles of E-cadherin and GRHL2-HER3, we probed GRHL2 and HER3 levels in ER-Ecad-KO cells using Western blotting. The results showed that the loss of E-cadherin does not affect the protein levels of GRHL2 (Figures 7B, C) or HER3 (Figures 7E, F) in ER-Ecad-KO cells. Therefore, the change in IC₅₀ in ER-Ecad-KO cells is not due to the loss of GRHL2 or HER3.

To determine whether GRHL2 regulates E-cadherin expression, we performed Western blotting on ER-GRHL2-KO cells and found that E-cadherin levels were slightly decreased in ER-GRHL2-KO cells compared to ER cells (Figures 7G, H). Consistent with the largely remaining expression of E-cadherin, ER-GRHL2-KO cells maintained enhanced cell-cell adhesion and formed large cell aggregates in the cell adhesion assay (Figure 7I). In contrast, ER-Ecad-KO cells exhibited a significant decrease in cell-cell adhesion, as expected (Figure 7I). These data indicate that the increased levels of E-cadherin are primarily independent of GRHL2 and that E-cadherin and GRHL2 mediate EGFR-bypass drug resistance through distinct pathways.

To further investigate the mechanisms underlying EGFR-bypass drug resistance in ER cells, we analyzed intracellular signal transduction pathways involving the phosphorylation of AKT, ERK1/2, and p38 using Western blotting, since these proteins can function downstream of HER3. First, we found that the level of AKT S473 phosphorylation was elevated in ER cells (Figures 8A,B, pAKT). Since the protein level of AKT was decreased in ER cells (Figures 8A, B, AKT), the proportion of phosphorylated AKT was much higher than in Parental cells (Figures 8A, B, pAKT/AKT). Second, while ERK1/2 phosphorylation levels were not changed in ER cells (Figures 8A,B, pERK), the protein levels of ERK1/2 were increased (Figures 8A, B, ERK), leading to a reduced proportion of phosphorylated ERK1/2 in these cells (Figures 8A, B, pERK/ERK). Third, p38 phosphorylation levels were decreased in ER cells (Figures 8A, B, pp38), while their protein levels were increased (Figures 8A, B, p38). These data suggest that the AKT pathway plays a crucial role in conferring drug resistance to erlotinib in ER cells. Supporting the critical role of AKT in ER cells, we found that ER cells were more sensitive to AKT inhibitors compared to Parental cells: the IC₅₀ of two different AKT inhibitors (afuresertib and ipatasertib) was lower in ER cells (Figure 8C).

Interestingly, we found that AKT phosphorylation levels remained elevated in ER-GRHL2-KO and ER-Ecad-KO cells (Figures 8A, B, pAKT), while p38 phosphorylation decreased further in these knockout cells (Figures 8A, B, pp38). These data suggest that drug resistance involves increased AKT phosphorylation and that the loss of drug resistance in ER-GRHL2-KO and ER-Ecad-KO cells is accompanied by decreased p38 phosphorylation. If this is the case, increased AKT phosphorylation alone may not be sufficient for drug resistance. To test this hypothesis, we overexpressed HER3 and E-cadherin individually and in combination in Parental cells using lentiviruses and measured the IC₅₀ for erlotinib. The results showed that while HER3 overexpression or its combination with E-cadherin overexpression enhanced AKT phosphorylation (Figures 9A, B, pAKT), they did not induce drug resistance (Figure 9C). Therefore, although HER3 and E-cadherin are necessary for drug resistance, their overexpression alone is not sufficient, likely requiring the involvement of other components.