E-cadherin–driven adherens junction reinforcement promotes spheroid-mediated invasion and progression in ALK–rearranged lung cancer chemoresistance
Dawon Hong, Hyun Jung Kwon, Jiwon Jeong, Seokhyun Yoon, Jin-Haeng Chung, Sunjoo Jeong
TL;DR
This study shows how lung cancer cells become resistant to treatment by forming spheroids and strengthening cell adhesion through E-cadherin and EpCAM.
Contribution
The study identifies E-cadherin and EpCAM as key drivers of chemoresistance through spheroid formation and cytoskeletal remodeling in ALK-rearranged lung cancer.
Findings
Crizotinib-resistant H2228 cells form compact spheroids with upregulated E-cadherin and reinforced adherens junctions.
E-cadherin knockdown disrupts spheroid architecture and partially restores drug sensitivity.
Elevated E-cadherin and EpCAM expression in patient biopsies correlates with chemotherapy resistance.
Abstract
Metastasis of cancer cells is driven by morphogenic changes that involve cytoskeletal remodeling and adherens junction reorganization. These cytoskeletal dynamics enable cancer cells to modulate their shape, adhesion, and motility, contributing to invasion, metastasis, and therapeutic resistance. In non–small cell lung cancer, anaplastic lymphoma kinase (ALK) gene rearrangements represent key oncogenic drivers, and ALK tyrosine kinase inhibitors such as crizotinib have significantly improved clinical outcomes. However, resistance frequently develops, often through on-target mutations or poorly understood bypass mechanisms. To investigate resistance–associated morphogenic adaptation, we employed two-dimensional and three-dimensional culture systems using crizotinib–resistant H2228 cells, along with transcriptomic profiling. The resistant cells formed compact, highly organized spheroids…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7
Figure 8
Figure 9Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsWnt/β-catenin signaling in development and cancer · Microtubule and mitosis dynamics · Cancer Cells and Metastasis
INTRODUCTION
Metastasis in cancer is a highly coordinated process driven by morphogenic changes that involve cytoskeletal remodeling and dynamic regulation of adherens junctions (Yilmaz and Christofori, 2009). The cytoskeleton plays a central role in regulating cell shape, polarity, adhesion, and motility—all of which are essential for invasion and tumor progression. Through remodeling processes of the cytoskeleton and adherens junctions, cancer cells can transit between epithelial and mesenchymal states, acquire invasive phenotypes, and adapt to mechanical and chemical cues in the tumor microenvironment (Canel et al., 2010, Xin et al., 2023). These cytoskeletal adaptations, often coordinated with junctional remodeling, enable collective migration and spheroid formation, contributing to drug resistance and metastatic potential (Friedl and Mayor, 2017, Hamilton and Rath, 2019).
The discovery of anaplastic lymphoma kinase (ALK) gene rearrangements has significantly transformed the therapeutic landscape of non–small cell lung cancer (NSCLC). ALK fusions occur in approximately 4% to 6% of lung adenocarcinomas and typically involve the fusion of the 3′ kinase domain of ALK with a variety of 5′ partner genes. This leads to overexpression and constitutive activation of ALK, driving oncogenic signaling (Chen et al., 2014). ALK-rearranged tumors exhibit strong oncogene addiction and are initially highly responsive to ALK tyrosine kinase inhibitors (TKIs), such as crizotinib (Guo et al., 2015). Despite these therapeutic advances, resistance to ALK TKIs almost inevitably develops, limiting long–term treatment efficacy and post–progression therapeutic options. Resistance mechanisms are broadly classified into on-target and bypass pathways. On-target resistance includes secondary mutations or amplifications in ALK that preserve ALK dependency (Lin et al., 2017). In contrast, bypass mechanisms are ALK-independent and involve alternative survival pathways, lineage plasticity, histological transformation, or activation of other signaling axes such as EGFR, HER2/HER3, IGF-1R, or MET (Fischer et al., 2015, Schneider et al., 2023, Voena et al., 2016). Understanding these diverse resistance routes and the dynamic tumor evolution during therapy is essential for guiding subsequent treatments and improving patient outcomes. In this regard, longitudinal biopsy analyses provide critical insights into the adaptive changes occurring in tumors under therapeutic pressure.
Adherens junctions are key components of epithelial cell-cell adhesion and play a central role in tissue morphogenesis and structural integrity. E-cadherin, the primary component of adherens junctions, is classically regarded as a tumor suppressor and a gatekeeper of epithelial identity (Burandt et al., 2021, Wong et al., 2018). Its loss is associated with epithelial-to-mesenchymal transition (EMT) and poor prognosis in various carcinomas, including lung cancer, where reduced E-cadherin expression is seen in approximately 42% of cases (Kim et al., 2013a; Yang et al., 2014). However, recent studies suggest a more complex role for E-cadherin in advanced tumors. In some metastatic carcinomas, E-cadherin is maintained or re-expressed to support cell-cell adhesion, collective migration, and even proliferative signaling via cross-talk with receptor tyrosine kinases such as EGFR (Chao et al., 2012, Chao et al., 2010, Liu et al., 2017, Padmanaban et al., 2019, Rubtsova et al., 2022, Russo et al., 2024). These functions are often linked to changes in the cytoskeleton, particularly actin organization, which mediates junction stability and spheroid integrity. These paradoxical findings underscore the importance of understanding how E-cadherin and cytoskeletal remodeling cooperate to support phenotypic plasticity, invasion, and drug resistance in lung cancer. Three-dimensional (3D) spheroid cultures provide a more physiologically relevant model than two-dimensional (2D) monolayers, as they better recapitulate spatial organization, intercellular adhesion, and resistance mechanisms seen in vivo (Kunz-Schughart et al., 1998). Therefore, they are particularly well-suited for dissecting cytoskeleton-related adaptations in ALK-rearranged NSCLC under therapeutic pressure.
In this study, we investigated the morphological and molecular adaptations associated with acquired resistance to crizotinib in ALK–rearranged lung cancer cells. Using both 2D and 3D culture systems, we found that crizotinib-resistant cells formed compact spheroids with elevated E-cadherin expression and reorganized adherens junctions, suggestive of cytoskeletal remodeling. Transcriptomic analysis revealed upregulation of morphogenesis– and cytoskeleton–related gene programs. Moreover, we identified elevated expression of EpCAM in invasive cell populations within spheroids. Importantly, longitudinal biopsy samples from patients with ALK–rearranged lung cancer showed increased expression of both E-cadherin and EpCAM during chemotherapy, underscoring the clinical relevance. Together, our findings reveal that cytoskeletal remodeling and junctional reinforcement contribute to a morphogenic program that supports drug resistance and tumor progression, suggesting novel therapeutic opportunities targeting adhesion and cytoskeletal dynamics in resistant lung cancer.
MATERIALS AND METHODS
Cell Culture and Reagents
The H2228 cell line was purchased from the American Type Culture Collection. All cells were incubated at 37°C with 5% CO_2_ injection in RPMI-1640 medium (Welgene Biotechnology) containing 4,500 mg/l glucose, 10% heat–inactivated fetal bovine serum, 100 U/ml penicillin, and 100 µg/ml streptomycin. All cell lines were routinely treated with anti–mycoplasma BM cyclin (Subramanian et al., 2012), and mycoplasma contamination was tested every 3 months using a Sartorius EZ-PCR Mycoplasma Detection Kit. Crizotinib was obtained from Selleck Chemicals Co Ltd.
Establishment of Crizotinib Resistance in H2228 Cells
Crizotinib–resistant H2228 cells were generated based on previously reported methods (Kim et al., 2013b). Briefly, parental H2228 cells were initially exposed to 1 nM crizotinib, with the concentration gradually increased up to 100 nM over an 8-month period. To further enrich for resistant populations, cells were continuously treated with crizotinib for an extended period. The sublines (H2228-CR1 and H2228-CR2) with acquired resistance to crizotinib were obtained through colony selection. For all subsequent experiments, resistant cells were cultured in drug-free medium for more than 1 week to eliminate residual effects of the crizotinib.
Spheroid Formation and 3D Embedding
To generate 3D spheroids, cells were seeded at a density of 1 × 10^4^ cells per well in PrimeSurface96U ultra-low attachment round-bottom plates (Sumitomo Bakelite), which promote uniform spheroid formation by preventing cell attachment to the plate surface (Vinci et al., 2012). The cells were cultured under standard conditions for 96 hours to allow spheroid formation. For 3D embedding, spheroids formed at 96 hours were gently processed by removing 200 μl of culture medium from each well. The spheroids were then embedded in either Matrigel Matrix (Corning, 356234) or in neutralized rat-tail collagen type I (3 mg/ml, Corning, 354236) adjusted to pH 7 to 7.5.
Characterization of Spheroids
Spheroids were imaged every 24 hours, and morphometric parameters, including circularity and solidity, were quantified using ImageJ/Fiji software, as previously described (Amaral et al., 2017, Gunay et al., 2020). Circularity was calculated based on standard ImageJ definitions, and solidity was defined as the ratio of the spheroid area to its convex hull area.
Transwell Invasion Assay Using Spheroids
After 96 hours of spheroid formation, spheroids were collected and transferred into the upper chamber of a Transwell insert (8.0 µm pore size, Corning) that was precoated with Matrigel in serum-free medium. The lower chamber contained medium with 10% FBS as a chemoattractant. After 24 hours, noninvaded spheroids and cells on the upper membrane were removed, and the invaded cells on the bottom surface were fixed in methanol and stained with hematoxylin and eosin.
siRNA-Mediated Knockdown (RNA Interference)
Cells were transfected with siRNA (1 µmol/l) using RNAiMAX (Invitrogen) and Opti-MEM (Invitrogen), following the manufacturer’s instructions. The following siRNA oligonucleotides were used: control siRNA-A (sc-37007), E-cadherin siRNA (sc-35242), and EpCAM siRNA (sc-43032), all purchased from Santa Cruz Biotechnology.
E-Cadherin Construct and Transfection
The FLAG-E-cadherin expression construct was generated by cloning the human CDH1 coding sequence into the p3xFLAG-CMV-10 vector using HindIII and KpnI restriction sites. Cells were transiently transfected with p3xFLAG-CMV-10 (control vector) or the FLAG-E-cadherin expression construct using Lipofectamine 3000 (Invitrogen) and Opti-MEM (Invitrogen), according to the manufacturer’s instructions.
Western Blot Analysis
Cells were lysed in RIPA buffer (25 mM Tris-HCl, pH 7.6; 150 mM NaCl; 1% Nonidet P-40; 0.5% sodium deoxycholate; 0.1% SDS; protease inhibitor cocktail) and centrifuged at 13,000 rpm, 4°C for 15 minutes. Samples (15-50 μg) were separated by SDS-PAGE and transferred to PVDF membranes (Millipore). Membranes were blocked in 5% non fat milk/TBST and incubated with primary antibodies: anti-E-cadherin (BD, 610181), anti-EpCAM (Abcam, ab71916), anti-FLAG (Invitrogen, MA1-91878), and anti-β-actin (Abcam, ab6276), followed by HRP-conjugated secondary antibodies. Signals were detected by chemiluminescence. Blots are representative of ≥2 independent experiments.
RNA Preparation, Reverse Transcription, Quantitative Real–Time PCR, and Sequencing
Total RNA was extracted from the cells using a Quick-RNA MiniPrep Plus kit (Zymo Research), according to the manufacturer’s instructions. cDNA was synthesized from the RNA samples (2.0 µg) using reverse transcriptase (Thermo Fisher Scientific). Quantitative real–time PCR was performed using PowerUp SYBR Green Master Mix (Thermo Fisher Scientific) and a StepOne Plus Real–Time PCR system (Applied Biosystems). Primer sequences used in qRT-PCR are provided in Supplementary Table S1. Sequencing was carried out at Theragen BIO on the Illumina NovaSeq 6000 platform. More detailed information about RNA-seq alignment is provided in the Supplementary Information.
Immunofluorescence Staining and Fluorescence Microscopy
Immunofluorescence staining was performed as previously described with slight modifications (Bergdorf et al., 2021). Cells and spheroids were fixed with 3.7% formaldehyde in PBS for 15 minutes at room temperature, then permeabilized with either 0.1% Triton X-100 in PBS or methanol for 5 minutes, depending on antibody requirements. Samples were blocked with 5% goat serum in PBS for 1 hour at room temperature. Primary antibodies—anti-E-cadherin (BD, 610181) and anti-EpCAM (Abcam ab71916)—were applied overnight at 4°C. Samples were incubated with Alexa Fluor–conjugated secondary antibodies (1:500 dilution). Nuclei were counterstained with DAPI using VECTASHIELD Antifade Mounting Medium (Vector Laboratories, H-1200) and mounted on glass slides. Images were acquired with a FluoView FV300 confocal microscope (Olympus), and for spheroids, z-stack imaging and 3D reconstruction were performed using FV10-ASW software.
MTT Assay
Cell survival and viability were assessed using the Cell Proliferation Kit I (MTT; Roche, Merck) according to the manufacturer’s instructions. Briefly, cells were seeded into 96-well plates and treated as indicated. After the incubation period, 10 μl of MTT labeling reagent (final concentration, 0.5 mg/ml) was added to each well, followed by incubation for 4 hours at 37°C in a humidified atmosphere containing 5% CO_2_. Subsequently, 100 μl of solubilization buffer was added to each well, and the plates were incubated overnight at 37°C to allow complete solubilization of the formazan crystals. Absorbance was measured using a microplate reader at 650 nm. IC_50_ values were calculated by nonlinear regression analysis using GraphPad Prism.
Patients and Samples
A total of 23 patients with ALK–rearranged pulmonary adenocarcinoma who underwent multiple biopsies or surgeries were selected. Two sets of biopsy tissue samples were prepared for analysis from each patient. The first biopsy specimens were obtained for diagnosis before therapy. The second biopsy specimen was obtained after suspected progression in radiological studies or a poor response to treatment. Pathological diagnoses were based on the 2021 WHO Classification of Lung Tumors (Thoracic tumours. WHO Classification of Tumours, 2021). Clinical and pathological stages were determined based on the eighth edition of the AJCC staging manual (American Joint Committee on Cancer, 2017). Clinicopathological information, including survival data, was obtained through retrospective review of medical records. More detailed information about the patients and tissue samples is provided in Supplementary Information.
E-Cadherin and EpCAM Immunohistochemistry
Formalin–fixed paraffin-embedded slides of the 23 patients with ALK–rearranged lung cancer were stained with anti-E-cadherin antibody (Agilent M3612, Dako) and anti-EpCAM antibody (Abcam ab71916) using the Benchmark ULTRA system, following the manufacturer’s instructions. E-cadherin and EpCAM immunohistochemistry were evaluated in at least 100 viable tumor cells per specimen. Two-second biopsy specimens were unavailable for EpCAM immunohistochemistry due to insufficient tissue. Membranous staining was assessed using the H-score (McCarty et al., 1985), which summed each tumor area (%) multiplied by the intensity (scale of 0, 1, 2, and 3).
Quantification and Statistical Analysis
Data are expressed as the mean ± SEM. All statistical analyses of the cell line study, except for RNA-seq, were performed using GraphPad Prism software version 7.01. Patient data analyses were performed using SPSS Statistics version 25.0 (IBM Corp) and R software (R Foundation for Statistical Computing). Statistical significance with P < .05 was considered significant. Supplementary Information provides more detailed descriptions of the statistical analyses performed.
RESULTS
Spheroid Formation in Crizotinib-Resistant H2228 ALK–Rearranged Lung Cancer Cells
To investigate the mechanisms of acquired resistance to crizotinib in ALK–rearranged lung cancer, we used the H2228 cell line harboring the EML4-ALK rearrangement together with longitudinal tumor biopsy samples obtained at diagnosis (first biopsy) and at disease progression (second biopsy) (Fig. 1A). Building on the previous resistance model, we further extended crizotinib exposure and isolated 2 single-cell–derived clones through colony selection, designated H2228-CR1 and H2228-CR2, thereby establishing sublines with stable resistance (Kim et al., 2013b). This approach enabled a more refined analysis of resistance-associated phenotypes. Whole–exome sequencing analysis revealed no gatekeeper mutations in the ALK gene of resistant cells, suggesting that H2228 cells acquired crizotinib resistance through off-target mechanisms rather than direct ALK alterations (Supplementary Tables S2 and S3).Fig. 1. Acquired resistance to crizotinib in H2228 cells induces clustering morphology. (A) Experimental flowchart for investigating crizotinib resistance in ALK-rearranged NSCLC. (B) Microscopic images of H2228 and H2228-CR2 cell morphology. Scale bar, 10 µm. (C) Single-cell ratio, calculated as the percentage of isolated cells after 48 hours of 2D culture. (D) Quantification of average cell count per cluster. (E) Time-lapse images of cluster formation over 12 hours. (F) 3D morphology of H2228 and H2228-CR2 cells over 3 days. Scale bar, 50 µm. (G) Quantification of spheroid solidity using the ImageJ/Fiji software. (H) Spheroid circularity was measured using the ImageJ software. The circularity index ranges from 0 to 1, with values closer to 1 indicating a more uniform and spherical morphology.Fig. 1
Morphological analysis revealed distinct phenotypes of crizotinib–resistant H2228-CR1 and H2228-CR2 cells (Fig. 1B; Supplementary Fig. S1A). The crizotinib-resistant cells exhibited a cobblestone-like morphology with increased cell-cell contact, whereas H2228 cells displayed a dispersed and EMT-like morphology with reduced contact (Fig. 1A and B). Quantitative analysis confirmed that H2228-CR1 and H2228-CR2 cells formed clusters of approximately 8 to 9 cells, whereas parental H2228 cells formed smaller aggregates of approximately 3 cells (Fig. 1D). Time-lapse microscopy confirmed the increased clustering in crizotinib-resistant cells (Fig. 1E).
We extended these findings by culturing the cell lines in a 3D extracellular matrix microenvironment system. Crizotinib–resistant H2228 cells formed spheroids visible as hanging drops on day 4 (Supplementary Fig. S1B). In 3D Matrigel, H2228-CR1 and H2228-CR2 cells formed well-defined spheroids, whereas parental H2228 cells formed less organized aggregates (Fig. 1F; Supplementary Fig. S1C). By day 3, the solidity of H2228-CR1 and H2228-CR2 spheroids significantly increased to approximately 0.746 and 0.814, respectively, whereas parental H2228 spheroids exhibited a significantly lower solidity of 0.545 (Fig. 1G). Notably, spheroids formed by H2228-CR1 and H2228-CR2 exhibited higher circularity indices (∼0.7), whereas aggregates formed by parental H2228 cells showed a lower circularity (∼0.3), indicating irregular and loose morphology (Fig. 1H). These results suggest that crizotinib resistance is associated with spheroid formation in ALK–rearranged lung cancers.
Elevated E-Cadherin at the Adherens Junction in Crizotinib–Resistant H2228 Cells
The strengthened cell-cell contact and clustering observed in 2D and 3D cultures of crizotinib-resistant cells led us to investigate the involvement of E-cadherin, a key component of cell-cell adherens junctions, in spheroids. Quantitative real–time PCR and Western blot analysis revealed significantly elevated CDH1 (encoding E-cadherin) mRNA and E-cadherin protein levels in H2228-CR1 and H2228-CR2 cells compared to H2228 cells (Fig. 2A and B). Notably, this upregulation coincided with a decrease in EMT-inducing transcription factors, including ZEB1, ZEB2, SNAI2, and TWIST2 in RNA-seq and qRT-PCR (Supplementary Fig. S2A and B) (Hajra et al., 2002, Puisieux et al., 2014). However, increased SNAI1 expression implies a more complex regulatory mechanism in crizotinib-resistant cells (Supplementary Fig. S2A and B). In parental H2228 cells, CDH1 mRNA and E-cadherin protein levels increased in a dose-dependent manner after 24 hours of crizotinib treatment, whereas crizotinib-resistant cells consistently maintained high levels regardless of the dose (Fig. 2C and Supplementary Fig. S3A).Fig. 2E-cadherin localizes at the adherens junction in crizotinib-resistant spheroids. (A) CDH1 mRNA expression in H2228, H2228-CR1, and H2228-CR2 cells as determined using quantitative real–time PCR. (B, C) Western blot analysis of E-cadherin protein expression (B) and treatment with increasing concentrations of crizotinib for 24 hours (C) in H2228, H2228-CR1, and H2228-CR2 cells. β-actin was used as a loading control. (D) Immunofluorescence of E-cadherin (red) and β-catenin (green) proteins with DAPI staining of the nuclear in H2228 and H2228-CR2 cells. (E) Quantification of E-cadherin and β-catenin colocalization intensity ratio in H2228, H2228-CR1, and H2228-CR2 cells as measured using ImageJ. (F) Immunofluorescence staining of E-cadherin (green) in spheroids in H2228 and H2228-CR2 cells. (G) Quantification of E-cadherin fluorescence intensity per spheroid using ImageJ software.Fig. 2
We assessed whether elevated E-cadherin levels were localized at the adherens junctions using immunofluorescence analysis. In 2D culture, E-cadherin was strongly expressed at adherens junctions in H2228-CR1 and H2228-CR2 clusters, whereas low signal intensity was observed in parental H2228 cell aggregates (Fig. 2D and Supplementary Fig. S3B, S3C). Co-staining of β-catenin (green) and E-cadherin (red) showed colocalization at cell-cell junctions in crizotinib-resistant clusters, producing yellow merged signals (Fig. 2D and E). Immunoprecipitation analysis confirmed the interaction between E-cadherin and β-catenin in these cell clusters (Supplementary Fig. S3D and E).
Next, we explored E-cadherin expression in the 3D spheroids (Fig. 2F and Supplementary Fig. S3F). Unlike in the 2D culture, E-cadherin staining intensity did not significantly differ between parental and crizotinib-resistant spheroids (Supplementary Fig. S3G). However, the proportion of E-cadherin-positive cells differed markedly. In parental H2228 spheroids, only 3.9% of cells stained positive for E-cadherin, indicating heterogeneity, whereas over 90% of cells in crizotinib-resistant spheroids exhibited strong E-cadherin expression at cell-cell junctions (93.8% in H2228-CR1, 94.5% in H2228-CR2) (Fig. 2F and G). These findings indicate that elevated E-cadherin at adherens junctions in crizotinib-resistant spheroids contributes to cohesive cell cluster formation.
Regulation of Spheroid Formation by E-Cadherin
We hypothesized that enhanced E-cadherin expression would promote spheroid formation in crizotinib–resistant H2228 cells. To test this, E-cadherin expression was knocked down using siRNA in H2228, H2228-CR1, and H2228-CR2 cells (Fig. 3A and Supplementary Fig. S4A). Spheroids from H2228-CR1 and H2228-CR2 cells lost their structural integrity and showed reduced diameters upon E-cadherin knockdown, whereas parental H2228 cell aggregates remained largely unchanged (Fig. 3B and Supplementary Fig. S4B). H2228-CR1 and H2228-CR2 cell spheroids reached circularity indices of 0.84 and 0.79, respectively, compared to 0.45 in parental H2228 cell aggregates, implying that crizotinib-resistant cells form more rounded spheroids (Fig. 3B). E-cadherin knockdown reduced circularity indices to 0.37 and 0.34 in H2228-CR1 and H2228-CR2 spheroids, respectively, indicating disrupted spheroid formation, while parental H2228 cell aggregates remained unaffected (Fig. 3B). Phalloidin staining revealed that E-cadherin depletion disrupted multicellular clusters in crizotinib-resistant cells, whereas parental H2228 cells remained aggregated (Supplementary Fig. S4C and D).Fig. 3E-cadherin knockdown disrupts spheroids. (A) Bright-field images of spheroids formed by H2228 and H2228-CR2 cells after transfection with sicontrol (sicon) or siE-caderin (siE-cad). (B) Quantification of spheroid circularity indices in (A) using ImageJ software. (C) Western blot analysis confirming overexpression of FLAG-tagged E-cadherin (FLAG-E-cad). β-actin was used as a loading control. (D) Bright-field images of spheroids formed by parental H2228 cells following transfection with empty vector (vec) or FLAG-tagged E-cadherin (FLAG-E-cad). (E) Quantification of spheroid solidity in (C). (F) Immunofluorescence staining of E-cadherin (green) in spheroids formed by parental H2228 cells transfected with empty vector or FLAG-E-cadherin.Fig. 3
To orthogonally validate the role of E-cadherin in spheroid formation, E-cadherin was ectopically expressed in parental H2228 cells. E-cadherin overexpression promoted the formation of a compact spheroid-like structure compared with vector-transfected controls (Fig. 3C and D). These spheroids exhibited a high circularity index (0.87), indicating enhanced multicellular organization relative to the loosely aggregated clusters formed by parental H2228 cells (Fig. 3E). Immunofluorescence analysis further revealed that ectopically expressed E-cadherin localized predominantly at cell–cell adhesion junctions, consistent with enhanced cell-cell adhesion associated with spheroid formation in H2228 cells (Fig. 3F).
Contribution of E-Cadherin to Cell Survival in Crizotinib-Resistant Cells
We next investigated the role of E-cadherin in cell survival and crizotinib response. E-cadherin knockdown significantly reduced the survival of H2228-CR1 and H2228-CR2 cells, whereas the survival of parental H2228 cells was not markedly affected (Fig. 4A). We then tested whether E-cadherin depletion restores crizotinib sensitivity in resistant cells. Parental H2228 cells exhibited 50% reduced viability at 0.18 μM crizotinib after 48 hours, while crizotinib-resistant cells required up to approximately 5-fold higher concentrations (Supplementary Fig. S4E and F). E-cadherin knockdown did not alter the crizotinib IC_50_ value in parental H2228 cells. In contrast, in H2228-CR2 cells, E-cadherin knockdown reduced the IC_50_ value by approximately 0.27 μM from 1.01 μM in control cells (Fig. 4B). However, H2228-CR1 cells did not exhibit a significant restoration of crizotinib sensitivity upon E-cadherin depletion, suggesting clone-specific differences in adaptive dependency on E-cadherin–mediated survival mechanisms (Supplementary Fig. S4G).Fig. 4E-cadherin modulates crizotinib sensitivity in ALK–rearranged H2228 cells. (A) Cell survival of H2228, H2228-CR1, and H2228-CR2 cells following E-cadherin knockdown using siE-cad compared with control siRNA (siCon). (B) Half-maximal inhibitory concentration (IC_50_) values for crizotinib in H2228 and H2228-CR2 cells after transfection with siCon or siE-cad. (C) Cell survival of parental H2228 cells following expression of FLAG-tagged E-cadherin (FLAG-E-cad) compared with empty vector (vec). (D) IC_50_ values for crizotinib in parental H2228 cells expressing Vec or FLAG-E-cad.Fig. 4
We further examined the effects of ectopic E-cadherin expression in parental H2228 cells. E-cadherin overexpression increased cell survival under standard culture conditions compared with vector-transfected controls (Fig. 4C). However, ectopic E-cadherin expression had only a minor effect on crizotinib sensitivity in parental H2228 cells, suggesting that E-cadherin overexpression is insufficient to induce resistance (Fig. 4D). Taken together, E-cadherin contributes to cell survival in crizotinib-resistant cells.
Upregulation of Cell Morphogenic Genes Including EpCAM in Crizotinib–Resistant H2228 Cells
To comprehensively understand the alterations in gene expression associated with crizotinib resistance, we performed RNA-seq analysis of parental H2228 cells and crizotinib–resistant H2228-CR1 and H2228-CR2 cells (Supplementary Fig. S5A). A Venn diagram shows that up to 88% of differentially expressed genes (DEGs) were shared between crizotinib–resistant cell lines (Supplementary Fig. S5B). A volcano plot displays 3,114 significant DEGs, with 1,510 upregulated and 1,583 downregulated genes (Fig. 5A and Supplementary Fig. S5C and D). Gene ontology analysis of upregulated genes showed enrichment in tissue morphology and structure (Fig. 5B and Supplementary Table S4). In contrast, the downregulated genes were associated with cell death pathways, including ferroptosis and apoptosis (Supplementary Fig. S5E).Fig. 5RNA sequencing reveals the upregulation of tissue morphogenic transcripts in crizotinib–resistant H2228 cells. (A) Volcano plot showing significantly differentially expressed genes (DEGs) between parental H2228 and crizotinib–resistant H2228-CR1 and H2228-CR2 cells. Red and blue dots indicate upregulated and downregulated genes, respectively; arrow indicates CDH1. (B) Functional enrichment analysis of commonly upregulated DEGs in H2228-CR1 and H2228-CR2 cells using Metascape. (C) Heatmap showing the mean log2 fold change of selected DEGs in H2228-CR1 and H2228-CR2 cells. (D) Correlation analysis between CDH1 and selected epithelial junction–related genes (EpCAM, CLDNs, and GJB3) in lung adenocarcinoma (LUAD) patients using GEPIA2 base. (E) Quantitative real–time PCR analysis of EpCAM mRNA expression in parental H2228, H2228-CR1, and H2228-CR2 cells. (F) Western blot analysis of EpCAM protein levels in parental H2228, H2228-CR1, and H2228-CR2 cells. β-actin was used as a loading control.Fig. 5
Among the genes upregulated in crizotinib-resistant cells, we identified CDH1, CLDNs, and EpCAM, indicating a reinforcement of epithelial traits (Fig. 5C). Notably, MET and AXL were also upregulated, suggesting activation of alternative receptor tyrosine kinases independent of drug efflux, as ABCC family genes were downregulated (Fig. 5C). Consistent with this notion, crizotinib–resistant H2228 cells showed enhanced ERK phosphorylation and YAP/TAZ dephosphorylation, consistent with activation of bypass signaling pathways under ALK inhibition (Supplementary Fig. S5F). Analysis of lung adenocarcinoma patient data using GEPIA2 revealed a significant positive correlation (R = 0.68) between CDH1 and EpCAM mRNA expression, suggesting a potential cooperative role for EpCAM and E-cadherin in crizotinib resistance (Fig. 5D). In contrast, CLDNs, GJB3, and other morphogenesis-related genes showed only a weak correlation (Fig. 5D and Supplementary Fig. S5G). Based on this, we further investigated EpCAM as a contributor to spheroid formation in crizotinib-resistant cells. RT-qPCR confirmed that EpCAM mRNA expression was increased in resistant cells compared to parental H2228 cells (Fig. 4E). Consistently, EpCAM protein levels were significantly elevated in H2228-CR1 and H2228-CR2 cells compared to those in parental cells, reflecting the E-cadherin expression pattern (Fig. 5F). Western blot analysis revealed 2 distinct bands for EpCAM: a ∼39-kDa band present in both parental and resistant cells and an additional ∼35-kDa band observed exclusively in H2228-CR1 and H2228-CR2 cells (Fig. 5F). This smaller band may reflect altered post-translational modification, such as differential glycosylation or proteolytic processing, specifically associated with the resistant phenotype. These findings suggest that a modified form of EpCAM is enriched in crizotinib–resistant lung cancer cells.
EpCAM-Mediated Invasion in Crizotinib–Resistant H2228 Spheroids
Given the elevated EpCAM level and its correlation with E-cadherin in crizotinib-resistant cells, we explored its localization and functional roles in the spheroid structure. In 3D Matrigel cultures, EpCAM was predominantly localized in the peripheral cells of crizotinib-resistant spheroids, whereas E-cadherin was observed at adherens junctions throughout crizotinib-resistant spheroids (Fig. 6A and B). Based on these observations, we hypothesized that EpCAM contributes to the invasive behavior of peripheral cells in resistant spheroids. To assess this, spheroids were embedded in collagen I gels. Crizotinib-resistant spheroids exhibited extensive protrusive migration (white arrow) into collagen I gel, which was rarely observed in H2228 spheroids (Fig. 6C). EpCAM knockdown significantly reduced the number of protrusive cells in crizotinib-resistant spheroids, although spheroid size in Matrigel remained unchanged (Fig. 6C and Supplementary Fig. S6A, B). To further evaluate invasive potential, we performed Transwell invasion assays following EpCAM depletion in spheroids (Fig. 6D). Invasive capabilities were significantly higher in crizotinib-resistant spheroids than in H2228 cells but were dramatically reduced following EpCAM depletion.Fig. 6. EpCAM regulates migration and invasion in crizotinib-resistant spheroids. (A) Immunofluorescence analysis of EpCAM (green) and E-cadherin (red) localization in 3D Matrigel-cultured spheroids of parental H2228 and H2228-CR2 cells. (B) Fluorescence intensity mapping along the white dashed line shown in (A). Peri. region, peripheral protrusive region; AJ, adherens junction. (C) Bright-field images of collagen I-embedded spheroids derived from parental H2228 and H2228-CR2 cells following transfected with sicontrol (sicon) or siEpCAM. (D) Transwell invasion assays using spheroids from parental H2228 and H2228-CR2 cells transfected with sicon or siEpCAM.Fig. 6
We next examined the effects of EpCAM knockdown on cell survival and crizotinib sensitivity. EpCAM knockdown partially reduced cell survival in crizotinib-resistant cells, indicating a contributory role in survival maintenance (Supplementary Fig. S7A and S7B). Notably, EpCAM knockdown significantly increased crizotinib-induced cytotoxicity in parental H2228 cells (Supplementary Fig. S7C). However, EpCAM knockdown did not significantly alter crizotinib sensitivity in resistant clones (H2228-CR1 and H2228-CR2) (Supplementary Fig. S7C).
Taken together with its preferential localization to peripheral protrusive cells, these findings suggest that EpCAM primarily supports the survival and invasive behavior of crizotinib-resistant spheroids, rather than directly modulating crizotinib sensitivity.
Increased E-Cadherin and EpCAM Expression During Tumor Progression in Patients With ALK–Rearranged Pulmonary Adenocarcinoma
To further explore the potential role of E-cadherin and EpCAM in tumor progression, we investigated their expression in primary and progressed specimens from 23 patients with ALK–rearranged pulmonary adenocarcinoma (Fig. 7A). The clinicopathological characteristics of the patients are summarized in Supplementary Tables S5 and S6. Most of the patients were middle-aged men with a history of smoking. More than half were clinically stage 3 or 4 at the first biopsy. Before the second biopsy, 9 patients (39.1%) received cytotoxic chemotherapy, 7 (30.4%) received ALK-TKIs such as crizotinib or alectinib, and 1 (4.3%) received concurrent chemoradiotherapy. The remaining 6 patients (26.1%) were only closely observed clinically because of a low clinical stage or patient refusal of chemotherapy. The patients were followed up for a median of 157 months. During this period, 9 patients died of the disease. The median progression–free survival was 17.9 months, and the median overall survival was 69.9 months.Fig. 7. Immunohistochemistry of E-cadherin and EpCAM on longitudinal biopsy samples of patients with ALK–rearranged pulmonary adenocarcinoma (N = 23). (A) Workflow of pathological study. (B) Representative images of E-cadherin expression (×200). (C) Significantly higher E-cadherin expression was observed in the second biopsy specimen than in the first biopsy specimen. (D) Representative images of EpCAM expression (×20). (E) EpCAM expression increased significantly in the second biopsy specimen compared to that in the first biopsy specimen.Fig. 7
Immunohistochemistry for E-cadherin and EpCAM revealed membranous staining in the tumor cells (Fig. 7B and D). Although both E-cadherin and EpCAM showed variable intratumoral staining intensities, total loss of expression was observed only for E-cadherin. The second biopsy specimens exhibited significantly stronger expression of both E-cadherin and EpCAM than the first biopsy specimens (E-cadherin median: 190 [165-247] vs 160 [130-280], P = .026; EpCAM median: 240 [180-270] vs 190 [150-235], P = .022) (Fig. 7C and E). Moreover, in the 7 patients treated with TKIs, the second biopsy specimens also tended to express stronger expression of both markers, although the difference was not statistically significant (E-cadherin median: 200 [170-230] vs 180 [130-190], P = .156; EpCAM median: 250 [160-290] vs 180 [110-230], P = .125). This tendency was also observed in patients who did not receive ALK-TKI therapy (N = 16), although not statistically significant (E-cadherin median: 140 [130-172.5] vs 185 [152.5-280.0], P = .083; EpCAM median: 207.5 [155.0-253.75] vs 235.0 [188.75-267.5], P = .104).
Remarkably, using the median H-score as the cutoff, patients with stronger E-cadherin or EpCAM expression in all biopsy specimens tended to have a shorter median overall survival (Supplementary Table S7). Although both markers increased during treatment, no correlation was observed between the H-scores for E-cadherin and EpCAM (P = .542). In addition, E-cadherin or EpCAM expression showed no significant correlation with specific treatment modalities when stratified by no treatment, TKI received, and others (cytotoxic chemotherapy and concurrent chemoradiotherapy) (Kruskal-Wallis test of E-cadherin: P = .830 and EpCAM: P = .714).
DISCUSSION
This study demonstrates that crizotinib–resistant H2228 lung cancer cells with EML4-ALK translocation acquire distinct morphogenic and molecular features that contribute to treatment resistance. Notably, these resistant cells form large, cohesive spheroids with invasive potential, driven in part by elevated expression of E-cadherin and EpCAM (Fig. 8). Consistently, E-cadherin and EpCAM upregulation was also observed in post-treatment biopsy specimens from patients with ALK-rearranged NSCLC, underscoring the clinical relevance of our findings (Fig. 8).Fig. 8. Schematic diagram illustrating the role of E-cadherin and EpCAM in promoting spheroid and invasion in crizotinib–resistant ALK–rearranged lung cancer. In the upper panel, crizotinib–resistant H2228 lung cancer cells with EML4-ALK fusion exhibit enhanced morphogenic features, forming compact spheroids with elevated E-cadherin and EpCAM expression. RNA-seq analysis reveals enrichment of morphogenesis–related gene signatures. Reinforced adherens junctions (via E-cadherin) and the accumulation of EpCAM–marked protrusive cells contribute to invasion. In the lower panel, longitudinal biopsy samples from ALK–rearranged lung cancer patients show increased E-cadherin and EpCAM expression at the time of cancer progression, supporting the clinical relevance of adhesion-driven survival and structural remodeling in tumor progression.Fig. 8
Importantly, no secondary ALK mutations were detected in the resistant cells, suggesting that resistance in this model arises through ALK–independent bypass mechanisms. Given the striking increase in cell-cell adhesion and spheroid compactness, we focused on E-cadherin–dependent structural remodeling and its role in adhesion–driven survival mechanisms. During cancer progression, EMT is typically associated with enhanced invasion, loss of adhesion, and downregulation of epithelial markers like E-cadherin. Restoration of epithelial traits after treatment is often interpreted as EMT reversal and a favorable therapeutic response. However, our data challenge this notion: crizotinib-resistant cells exhibited increased E-cadherin expression and stronger spheroid cohesion compared to parental cells, suggesting a morphogenic adaptation that supports resistance rather than regression.
Moreover, the increased E-cadherin expression in treatment-resistant tumors may be a unique feature of ALK–rearranged lung cancer. In our cohort, E-cadherin upregulation was observed in resistant tumor biopsies irrespective of ALK TKI exposure, suggesting an association with tumor biology rather than treatment modality. However, as our functional experiments were limited to an ALK TKI–resistant cell line, this interpretation remains preliminary and requires validation using additional, appropriately selected models. In contrast, EGFR–mutated lung cancers resistant to EGFR TKIs have been reported to show reduced E-cadherin expression in metastatic lesions than primary cancers (Chung et al., 2011), supporting the possibility that E-cadherin regulation during progression and resistance may be genotype dependent.
Knockdown of E-cadherin disrupted spheroid integrity and reduced the viability of resistant—but not parental—cells, highlighting E-cadherin’s essential role in maintaining the resistant phenotype. These findings suggest that E-cadherin–mediated adherens junction reinforcement and cytoskeletal stabilization are not merely byproducts of therapeutic pressure but may be active mechanisms of resistance and survival. This phenotype aligns with the concept of epithelial plasticity, in which cancer cells retain or reacquire epithelial traits to promote collective migration, tissue invasion, and immune evasion (Cheung and Horne-Badovinac, 2025). Spheroids have been widely recognized as in vitro models that mimic the architecture, microenvironment, and drug resistance features of solid tumors (Nunes et al., 2019). Their increased resistance to chemotherapy, as compared to monolayer cultures, is attributed to multiple factors, including limited drug penetration, hypoxia, metabolic rewiring, and the presence of quiescent or stem-like cells (Oishi et al., 2025, Pease et al., 2012). Our results reinforce this model by showing that crizotinib-resistant spheroids were more compact and drug-resistant, with E-cadherin playing a key role in maintaining their structure and survival.
E-cadherin is known to be upregulated in 3D cultures across several cancer types and is critical for compact spheroid formation (Powan et al., 2017, Russo et al., 2024, Sommariva and Gagliano, 2020). In our model, E-cadherin contributes to both spheroid formation and crizotinib resistance, as shown by the partial restoration of drug sensitivity following its knockdown in both resistant clones. However, the magnitude of this effect differs between H2228-CR1 and H2228-CR2. Although both clones originated from the same parental line, they appear to have undergone somewhat divergent adaptive paths during long–term crizotinib exposure. This divergence may reflect the stabilization of distinct nongenetic states arising from rare–cell transcriptional variability under therapeutic pressure, which has been investigated using single-cell analyses in drug-resistant cancers, leading to differential dependence on E-cadherin–mediated survival mechanisms (Chen et al., 2025, Cheng et al., 2025, Shaffer et al., 2017). While additional studies will be needed to fully define the molecular basis of these differences, our findings suggest that clonal heterogeneity during resistance acquisition can influence the extent to which E-cadherin depletion restores drug sensitivity. These results align with growing evidence that E-cadherin can paradoxically promote cancer progression and resistance. For example, in colorectal and ovarian cancers, loss of E-cadherin has been associated with increased chemosensitivity, while high E-cadherin expression correlates with larger, more resilient spheroids and prolonged survival under chemotherapy (Druzhkova et al., 2019, Skarkova et al., 2021, Xu et al., 2014). Thus, E-cadherin may have context-dependent roles in lung cancer, acting either as a tumor suppressor or as a resistance driver depending on the cellular state. Therapeutically, targeting E-cadherin has shown promise. Studies have reported that antibodies against E-cadherin can disrupt spheroid cohesion, restore drug sensitivity, and enhance immune-mediated cytotoxicity (Green et al., 2004, Green et al., 2002). These approaches may offer new avenues for overcoming resistance in patients with ALK–rearranged lung cancer who exhibit epithelial redifferentiation under TKI pressure.
Our study also implicates EpCAM in the invasive behavior of resistant spheroids. Unlike E-cadherin, which was uniformly expressed throughout the spheroid, EpCAM was concentrated in peripheral, protrusive cells—suggesting a specialized role in invasion and migration. This spatial localization aligns with the concept of “leader cells” at the invasive front (Cheung and Ewald, 2014). EpCAM is a membrane-spanning protein that undergoes regulated intramembrane proteolysis to release its extracellular domain (EpEX), which activates EGFR-mediated signaling (Brown et al., 2021). We detected a 35-kDa fragment likely corresponding to EpEX in resistant cells, supporting its role in bypass signaling. In other cancers, such as head and neck tumors, soluble EpEX has been shown to activate ERK pathways and contribute to resistance against anti-EGFR therapies (Umemori et al., 2023). In lung cancer models, EpCAM knockdown impairs proliferation, invasion, and survival, reinforcing its oncogenic potential. While the prognostic significance of EpCAM in NSCLC remains debated (Pak et al., 2012), its functional role in resistance and invasiveness supports further exploration as a therapeutic target.
Despite the insights gained, this study has several limitations. First, we used crizotinib—a first–generation ALK TKI—for resistance modeling, whereas second- and third-generation inhibitors such as alectinib and lorlatinib are now standard in clinical practice. Further studies using models of resistance to these newer agents are necessary to validate our findings. Second, while we identified the involvement of E-cadherin and EpCAM in resistance, the underlying signaling pathways remain to be fully elucidated. Future work should explore relevant pathways such as Wnt/β-catenin, Hippo-YAP, and others implicated in cell adhesion and drug resistance. Third, due to limitations in biopsy sampling techniques (eg, small specimens from transbronchial or core needle biopsies), direct assessment of spheroid structures in clinical samples was not feasible. Finally, the small and heterogeneous patient cohort limits the interpretation of the prognostic relevance of E-cadherin and EpCAM expression. Larger studies are needed to validate their roles in resistance and metastasis.
In conclusion, our study provides evidence that crizotinib resistance in ALK–rearranged lung cancer involves a morphogenic adaptation characterized by adherens junction reinforcement and cytoskeletal remodeling. E-cadherin supports spheroid cohesion and survival, while EpCAM contributes to invasive behavior, together promoting a phenotype that facilitates resistance and metastasis. These findings underscore the therapeutic potential of targeting adhesion- and cytoskeleton-related pathways to overcome drug resistance in ALK-rearranged NSCLC.
Ethical Compliance
This study was approved by the Institutional Review Board of Seoul National University Bundang Hospital (Seongnam, Korea) (approval no. B-2411-936-101), and the requirement for informed consent was waived.
Author Contributions
Seokhyun Yoon: Writing – review & editing, Software, Resources. Dawon Hong: Writing – original draft, Investigation. Hyun Jung Kwon: Writing – original draft, Validation, Investigation. Sunjoo Jeong: Writing – review & editing, Supervision, Conceptualization. Jiwon Jeong: Investigation. Jin-Haeng Chung: Writing – review & editing, Visualization, Conceptualization.
Declaration of Competing Interests
The authors declare no potential conflicts of interest.
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1Amaral R.L.F.Miranda M.Marcato P.D.Swiech K.Comparative analysis of 3D bladder tumor spheroids obtained by forced floating and hanging drop methods for drug screening Front. Physiol.8201760510.3389/fphys.2017.0060528878686 PMC 5572239 · doi ↗ · pubmed ↗
- 2Bergdorf K.N.Phifer C.J.Bechard M.E.Lee M.A.Mc Donald O.G.Lee E.Weiss V.L.Immunofluorescent staining of cancer spheroids and fine-needle aspiration-derived organoids STAR Protoc.22202110057810.1016/j.xpro.2021.100578 PMC 818210634136836 · doi ↗ · pubmed ↗
- 3Brown T.C.Sankpal N.V.Gillanders W.E.Functional implications of the dynamic regulation of Ep CAM during epithelial-to-mesenchymal transition Biomolecules 117202195610.3390/biom 1107095634209658 PMC 8301972 · doi ↗ · pubmed ↗
- 4Burandt E.Lubbersmeyer F.Gorbokon N.Buscheck F.Luebke A.M.Menz A.Kluth M.Hube-Magg C.Hinsch A.Hoflmayer D.E-Cadherin expression in human tumors: a tissue microarray study on 10,851 tumors Biomark. Res.9120214410.1186/s 40364-021-00299-434090526 PMC 8180156 · doi ↗ · pubmed ↗
- 5Canel M.Serrels A.Miller D.Timpson P.Serrels B.Frame M.C.Brunton V.G.Quantitative in vivo imaging of the effects of inhibiting integrin signaling via Src and FAK on cancer cell movement: effects on E-cadherin dynamics Cancer Res.702220109413942210.1158/0008-5472.CAN-10-145421045155 PMC 3079905 · doi ↗ · pubmed ↗
- 6Chao Y.Wu Q.Shepard C.Wells A.Hepatocyte induced re-expression of E-cadherin in breast and prostate cancer cells increases chemoresistance Clin. Exp. Metastasis 2912012395010.1007/s 10585-011-9427-321964676 PMC 3991430 · doi ↗ · pubmed ↗
- 7Chao Y.L.Shepard C.R.Wells A.Breast carcinoma cells re-express E-cadherin during mesenchymal to epithelial reverting transition Mol. Cancer 9201017910.1186/1476-4598-9-17920609236 PMC 2907333 · doi ↗ · pubmed ↗
- 8Chen Z.Fillmore C.M.Hammerman P.S.Kim C.F.Wong K.K.Non-small-cell lung cancers: a heterogeneous set of diseases Nat. Rev. Cancer 148201453554610.1038/nrc 377525056707 PMC 5712844 · doi ↗ · pubmed ↗