Cancer drug response and resistance: molecular mechanisms and combating strategies

Abstract

Despite remarkable advances in cancer drug treatment, including chemotherapy, targeted therapy, and immunotherapy, therapeutic resistance remains a formidable clinical barrier, limiting durable responses and long-term survival. Drug resistance can be broadly categorized as intrinsic, where tumors fail to respond to initial treatment, or acquired, which emerges during or after therapy due to adaptive or evolutionary processes. A comprehensive understanding of the multifactorial and dynamic nature of resistance is essential for improving treatment efficacy. In this review, we systematically examine the molecular and cellular determinants of drug response and resistance across 22 cancer types, highlighting key resistance mechanisms such as compensatory pathway activation, phenotypic plasticity, immune evasion, enhanced DNA damage repair, and the survival of drug-tolerant persister cells. These mechanisms are further contextualized across major therapeutic modalities, supported by clinical trials. We also present emerging strategies to overcome resistance, including rational drug combinations, novel agents, microbiome modulation, adaptive and intermittent therapies and advanced drug delivery systems, each illustrated with representative clinical studies. Moreover, we discuss cutting-edge tools that are revolutionizing resistance research, including single-cell and spatial multiomic profiling, patient-derived tumor organoid and xenograft (PDO/PDX) models, and artificial intelligence (AI)-powered predictive analytics. By integrating insights across molecular, cellular, and clinical dimensions, this review offers a strategic framework for understanding and tackling cancer drug resistance, with important translational implications for the future of precision oncology.

Introduction

Therapeutic resistance remains one of the most formidable challenges in oncology, limiting the long-term efficacy of anticancer therapies and contributing substantially to disease relapse and poor prognosis.¹ Despite remarkable advances in targeted therapies, immunotherapies, and combination regimens, a significant proportion of patients experience either intrinsic resistance, where tumors fail to respond from the outset, or acquired resistance, which emerges after an initial therapeutic response.^2,3 These resistance patterns underscore the dynamic, adaptive nature of cancer and highlight the pressing need to decipher the underlying mechanisms that enable tumor cells to evade therapeutic pressure.³

Resistance is not a singular or static phenomenon but rather the culmination of complex, multilevel processes involving both tumor-intrinsic and tumor-extrinsic factors.⁴ At the cellular level, tumor cells can rewire survival signaling pathways, upregulate drug efflux transporters, undergo epithelial-mesenchymal transition (EMT), and enter dormant states.³ EMT functions as a reversible cell-state program that endows tumor cells with increased motility, survival capacity, and tolerance to therapeutic stress, thereby contributing to invasion, metastasis, and treatment resistance.⁵ Epigenetic reprogramming and metabolic adaptation further enable phenotypic plasticity, while genetic alterations such as secondary mutations and gene amplifications sustain oncogenic signaling and confer drug tolerance.⁶ Beyond the cancer cell itself, the tumor microenvironment (TME), comprising immune cells, stromal components, and extracellular matrix, plays a pivotal role in shaping therapeutic outcomes by fostering immune evasion, altering drug availability, and facilitating cellular crosstalk that promotes resistance.^7,8 Tumor heterogeneity, both spatial and temporal, adds another layer of complexity, as subclonal populations with distinct genetic and epigenetic features coexist and evolve under selective therapeutic pressure.^7,9 The emergence and expansion of resistant clones are often driven by Darwinian evolutionary principles, whereby treatment selectively eliminates sensitive populations while allowing drug-tolerant persister (DTP) cells or preexisting resistant subclones to dominate.¹⁰ This evolutionary process not only fuels treatment failure but also underlies the transition from localized to metastatic disease.¹¹

Recent technological advances, particularly in single-cell multiomics, spatial profiling, liquid biopsy, and patient-derived models, have revolutionized our ability to dissect these mechanisms at unprecedented resolution.^12–17 These tools have illuminated the plasticity of drug-tolerant cells, the immunosuppressive reprogramming of the TME and the dynamic adaptation of cancer ecosystems throughout the course of treatment. Moreover, they have enabled real-time monitoring of resistance evolution and informed the development of personalized treatment strategies, including biomarker-guided therapy, adaptive dosing, and drug rechallenge protocols.

In this review, we provide a comprehensive and integrative overview of the multifactorial mechanisms that drive cancer drug resistance across 22 cancer types (Table 1). We begin by delineating intrinsic and acquired resistance pathways, including alterations in drug transport and metabolism, signaling rewiring, epigenetic remodeling, the DNA damage response, and tumor microenvironmental adaptation. We then explore how tumor cell plasticity (TCP), microbial influences, and evasion of regulated cell death contribute to resistance evolution. Subsequently, we discuss resistance patterns across different treatment modalities, including chemotherapy, targeted therapy, immunotherapy and hormonal therapy, and highlight emerging technologies and translational strategies to overcome resistance. Finally, we present future perspectives on how the integration of longitudinal omics, computational modeling, and innovative clinical trial designs may reshape the therapeutic landscape and improve long-term outcomes for patients with drug-resistant malignancies.

Table 1.

Drug resistance mechanisms and characteristics across cancer types

Drug resistance mechanisms	Resistance against therapy	Cancer types	Genes/proteins/pathways affected	Reference
Drug uptake	Platinum compounds	PDAC	Low expression of OCT2 reduces drug uptake and contributes to drug resistance.	138
		Ovarian cancer	Low level of OCT2 reduces drug uptake and leads to drug resistance.	139
	Anti-CTLA-4 and anti-PD-1 therapy	Skin cancer	SLC13A3 promotes PD-L1 stabilization through alkylation-mediated ITA transport	142
Drug efflux	Chemotherapy	Breast cancer	BCRP overexpression facilitates the efflux of chemotherapeutic agents.	160
	Chemotherapy	Colorectal cancer	DNA-damaging agents trigger Golgi dispersal in CSCs, which activates the PKCα/GSK3α/TFE3 axis to upregulate ABCG2-mediated drug efflux, thereby conferring chemoresistance.	51
	Paclitaxel, olaparib	Ovarian cancer	Paclitaxel-induced drug-resistant cells exhibit elevated P-gp expression.	108
Drug Pharmacokinetics and Metabolism	Sorafenib	HCC	Loss of CYP3A4 expression impairs the bioactivation of prodrugs, thereby leading to therapeutic failure.	164
Genome instability and mutations	Endocrine therapy	Breast cancer	Loss of ER expression contributes to therapeutic resistance in breast cancer.	110
			LBD mutations in ESR1 confer constitutive receptor activation and are associated with acquired resistance to endocrine therapy in breast cancer patients.	114
	RAS inhibitors	Gastroesophageal cancer	KRAS gene copy number amplification confers cross-resistance to RAS inhibitors.	97
	ADT	Prostate cancer	Selective splicing of AR-V7 drives resistance to ADT in prostate cancer.	120
	Imatinib	CML	Mutations within the BCR-ABL1 kinase domain render imatinib ineffective.	95
	MET inhibitors	NSCLC	Mutations in the MET can mediate resistance to targeted therapy in NSCLC.	96
DDR	PARPis	Prostate cancer	NSD3S enhances the stability of stalled replication forks by suppressing MRE11, thereby reducing the sensitivity of prostate cancer cells to PARPis.	103
	Radiotherapy	ESCC	Radiation-triggered release of HMGB1-containing exosomes from tumor cells activates the PI3K/AKT/FOXO3A axis in recipient cells, thereby enhancing DNA repair efficiency and evading apoptosis, collectively driving resistance to radiotherapy.	200
PI3K-AKT and mTOR signaling pathway	Tamoxifen	Breast cancer	CCL2 secreted by TAMs induces drug resistance via activation of the PI3K/AKT/mTOR signaling pathway.	215
	Endocrine therapy		Mutations in PIK3CA drive persistent activation of the PI3K/AKT/mTOR axis, promoting oncogenic signaling.	211
	Lenvatinib	HCC	Loss of EVA1A activates the PI3K/AKT signaling pathway, thereby driving drug resistance in hepatocellular carcinoma.	99
	EGFR-TKIs	NSCLC	Mutations in the PIK3CA gene lead to aberrant activation of the PI3K-AKT and mTOR signaling pathways, contributing to acquired resistance.	27
MAPK signaling pathway	Endocrine therapy	Breast cancer	NF1 deficiency results in sustained hyperactivation of the MAPK signaling pathway.	118
	PI3Kα inhibition		NF1 deficiency drives therapeutic resistance through dual mechanisms: activation of RAS–MAPK signaling and metabolic reprogramming.	223
	Anti-EGFR therapy	Colorectal cancer	Patients with elevated expression of EGFR ligands are more prone to EGFR/MAPK pathway activation and therapeutic resistance.	222
	BRAF inhibitor	Melanoma	Reactivation of the MAPK pathway and oncogenic bypass signaling contribute to therapy resistance.	221
NF-κB signaling pathway	Chemotherapy	Breast cancer	HSPB1 promotes drug resistance by inhibiting ferroptosis and modulating NF-κB signaling in cancer cells.	231
	Immunotherapy	Prostate cancer	RelB activates the NF-κB signaling pathway, upregulating PD-L1 expression and driving therapeutic resistance.	233
	5-FU	HCC	Overexpression of EIF5B promotes IκB phosphorylation and activates the NF-κB signaling pathway, thereby conferring resistance to 5-FU in hepatocellular carcinoma.	100
	BTK inhibitor	MCL	Compensatory activation of the non-canonical NF-κB signaling pathway contributes to therapeutic resistance in MCL.	232
	ICBs	Multiple cancers	Cancer cells upregulate RIPK1 to activate the NF-κB signaling pathway, thereby promoting both intrinsic and extrinsic resistance to immunotherapy.	234
TGF-β signaling pathway	Regorafenib	HCC	Aberrant activation of the TGF-β signaling pathway directly induces CSC properties, contributing to resistance against TKIs.	237
Wnt signaling pathway	Lenvatinib	HCC	CDK6 upregulation activates the Wnt/β-catenin signaling pathway, thereby driving lenvatinib resistance.	79
	Not specified	Melanoma	PDPN is significantly upregulated, driving activation of the Wnt/β-catenin signaling pathway and remodeling the TME, thereby promoting therapeutic resistance.	239
	Chemotherapy	Gastric cancer	The ADAR1/SCD1 axis regulates chemoresistance by enhancing lipid droplet formation to neutralize chemotherapy-induced endoplasmic reticulum stress.	250
	Gemcitabine	PDAC	LncRNA enhances pancreatic cancer resistance to gemcitabine by regulating Pygo2 and ATG14, thereby activating the Wnt/β-catenin and autophagy pathways.	242
	Chemotherapy	Glioblastoma	The Wnt/β-catenin signaling pathway induces stemness, mesenchymal transition, and drug efflux functions in tumor endothelial cells, thereby driving therapeutic resistance.	241
JAK/STAT signaling pathway	Docetaxel	Prostate cancer	The autocrine IL-11/IL-11Rα signaling loop confers resistance to docetaxel through activation of the JAK1/STAT4 pathway.	247
	AR-targeted therapy	CRPC	Loss of TP53/RB1 and overexpression of SOX2 establish a positive feedback loop that upregulates the JAK-STAT signaling pathway, promoting lineage plasticity and driving therapeutic resistance.	246
Notch signaling pathway	Doxorubicin	Breast cancer	Dll1⁺ quiescent CSCs drive the emergence of drug resistance via NF-κB signaling.	229
	Gemcitabine	PDAC	The TRIM59/RBPJ positive feedback circuit mediates drug resistance by activating the Notch signaling pathway.	253
	IFN-γ	Glioma	Glioma cells establish an immune-evasive barrier and develop resistance to IFN-γ therapy by downregulating the activity of the Notch1/2–RBPJ axis	256
Hippo signaling pathway	Chemotherapy	PDAC	The positive feedback loop between VCPIP1 and the Hippo signaling pathway promotes cancer progression and contributes to drug resistance.	263
GAS6/AXL signaling pathway	Chemotherapy	Breast cancer	Chemotherapy resistance is driven by Gas6/Axl-mediated activation of the Akt/GSK-3β/β-catenin signaling pathway.	272
	EGFR-TKIs	EGFR-mutant lung cancer	Tumor cells accelerate the acquisition of resistance-conferring mutations through the GAS6–AXL axis, ultimately leading to sustained resistance to EGFR-TKIs.	270
cGAS–STING signaling pathway	Endocrine therapy	Breast cancer	Hyperactivation of AKT1 suppresses cGAS–STING signaling, thereby promoting resistance to endocrine therapy.	119
Phenotypic and epigenetic reprogramming	MDR	Breast cancer	Epigenetic hypomethylation at the ABCB1 promoter results in increased ABCB1 expression, which augments drug efflux and accelerates MDR.	180
	AR-targeted therapy	Prostate cancer	KMT2C deficiency promotes the transdifferentiation of prostate cancer into DNPC, thereby contributing to therapeutic resistance.	106
	Castration		AcK609-AR promotes aberrant binding to key gene enhancers even in the presence of antagonists, and cooperates with the AR-ACK1 positive feedback loop to sustain oncogenic activity, thereby driving CRPC progression and therapeutic resistance.	121
	TKIs	Lung cancer	The SWI/SNF chromatin remodeling complex facilitates resistance to TKIs.	298
	PARPis	BRCA2-deficient tumors	Reduced EZH2 expression leads to insufficient H3K27 methylation, resulting in decreased recruitment of MUS81. This enhances replication fork stability, protecting cells from PARPis-induced cytotoxicity and ultimately conferring drug resistance.	102
	Cisplatin	Ovarian cancer	LncRNAs can promote therapeutic resistance by modulating the stability, translation, or activity of key signaling regulators.	297
Cell state plasticity	AR-targeted therapy	Prostate cancer	Upregulation of ZMYND8 promotes the transdifferentiation of anti-androgen therapy–induced prostate cancer into NEPC, leading to therapeutic resistance.	105
	Cisplatin and paclitaxel	ESCC	QSOX2 mediates disulfide bond modification of TSC2 to activate the mTOR/c-Myc signaling axis, enhancing tumor cell stemness and driving drug resistance.	319
	Chemotherapy	PDAC	DTP cells acquire a resistant phenotype by highly expressing CYP3A to mediate the metabolic detoxification of chemotherapeutic agents.	311
TME	Chemotherapy	Breast cancer	TSPAN8⁺ myCAFs promote chemoresistance by enhancing breast cancer stemness and activating the TSPAN8–MAPK11–RBBP6–SIRT6 signaling axis.	62
	Immunotherapy and chemotherapy	NSCLC	In the TME, COL11A1⁺ CAFs form a collagenous barrier that impedes CD8⁺ T cell infiltration and cooperate with SPP1⁺ TAMs to enhance fibrosis and immunosuppression, ultimately leading to T cell dysfunction and therapeutic resistance.	78
	Anti-PD-1 therapy		CircUSP7 induces CD8⁺ T cell dysfunction, driving immune evasion and resistance to anti-PD-1 therapy.	83
	Endocrine therapy	Prostate cancer	AR upregulates negative immune checkpoint molecules and pro-inflammatory factors, thereby establishing an immunosuppressive TME that contributes to therapeutic resistance.	122
	Immunotherapy	Cholangiocarcinoma	CXCL6 remodels lipid metabolism and induces NET formation, contributing to immunotherapy resistance.	81
	Chemotherapy		PDHA1 succinylation promotes chemoresistance by regulating metabolic reprogramming and suppressing macrophage MHC-II antigen presentation.	339
Microbiome	Trastuzumab	Breast cancer	Pseudomonas aeruginosa induces chemoresistance by secreting the quorum-sensing molecule 3OC, which activates key pro-survival signaling pathways, including TGF-β, MAPK, and PI3K–Akt.	348
	Docetaxel		Microbiota ETBF promotes drug resistance by secreting the toxin BFT-1, which relieves repression of the Notch signaling pathway.	351
	Gemcitabine	PDAC	The secretion of CDD-L by γ-proteobacteria metabolically inactivates gemcitabine, thereby conferring drug resistance.	344
	ICBs	Oral cancer	P. gingivalis infection upregulates PD-L1 expression on dendritic cells, suppresses CD8⁺ T cell responses, and induces therapeutic resistance.	354
Cell Death	MDR	Colorectal cancer	CBX3 promotes MDR by inhibiting ferroptosis through the CUL3/NRF2/GPX2 signaling axis.	359
	Chemotherapy		Histone lactylation enhances GCLC expression, thereby promoting chemoresistance of CSCs by suppressing ferroptosis.	360
	Sorafenib	HCC	USP18 impairs NCOA4-dependent ferritinophagy, thereby limiting sorafenib-induced ferroptotic lipid peroxidation and promoting acquired resistance.	101
	Chemotherapy	SCLC	METTL3 promotes chemoresistance in SCLC by inducing mitophagy.	366
JAK/STAT signaling pathway, TME	Immunotherapy	PD-L1-high lung cancer	Endogenous PD-L1 signaling promotes MDSCs-mediated immunosuppression via the IL-6/JAK/STAT3 pathway.	137
	Targeted therapy	CML	STAT3-mediated metabolic reprogramming contributes to targeted therapy resistance in CML.	249
	Immunotherapy	Multiple cancers	CXCR4 induces CD8⁺ T cell dysfunction via the JAK2/STAT3 pathway, leading to therapeutic resistance.	88
MAPK signaling pathway, TME	ICBs	Gastric cancer	CXCL1 induces the accumulation of PMN-MDSCs, promotes CD8⁺ T cell exhaustion, and contributes to therapeutic resistance.	225
Phenotypic and epigenetic reprogramming, Drug efflux	Chemotherapy	PDAC	PDAC cells downregulate miR-146a-5p, leading to the activation of NF-κB p65 transcriptional activity, which in turn induces P-gp expression and mediates chemoresistance.	292

PDAC pancreatic ductal adenocarcinoma, OCT the organic cation transporter, ITA itaconate, BCRP breast cancer resistance protein, CSCs cancer stem cells, P-gp P-glycoprotein, HCC hepatocellular carcinoma, ER estrogen receptor, LBD ligand-binding domain, ESR1 estrogen receptor 1, ADT androgen deprivation therapy, AR androgen receptor, NSCLC non-small cell lung cancer, KD kinase domain, DDR DNA damage repair, PARPis PARP inhibitors, TAMs tumor-associated macrophages, CML chronic myeloid leukemia, ESCC esophageal squamous cell carcinoma, EGFR-TKIs epidermal growth factor receptor tyrosine kinase inhibitors, 5-FU 5-fluorouracil, EIF5B eukaryotic initiation factor 5B, BTK Bruton’s tyrosine kinase, MCL mantle cell lymphoma, ICBs immune checkpoint blockades, TKIs tyrosine kinase inhibitors, PDPN podoplanin, MDR multidrug resistance, TME the tumor microenvironment, myCAFs myofibroblastic cancer-associated fibroblasts, NET neutrophil extracellular trap, 3OC N-(3-oxo-dodecanoyl) homoserine lactone, ETBF enterotoxigenic Bacteroides fragilis, CDD-L long-form cytidine deaminase, DNPC double-negative prostate cancer, CRPC castration-resistant prostate cancer, IFN-γ interferon-γ, AcK609-AR acetylation of AR at lysine 609, CAFs: cancer-associated fibroblasts, LncRNAs long non-coding RNAs, DTPs drug-tolerant persisters, NEPC neuroendocrine prostate cancer, SCLC small cell lung cancer, MDSCs myeloid-derived suppressor cells, PMN-MDSCs polymorphonuclear myeloid-derived suppressor cells

Overview and classification of resistance mechanisms

Cancer therapy resistance is a multifactorial and dynamic process that emerges from both tumor-intrinsic properties and tumor-extrinsic influences.¹ Understanding the classification of resistance mechanisms across diverse treatment modalities is crucial for guiding therapeutic choices and developing effective intervention strategies.¹⁸ Here, we provide an integrated overview that categorizes resistance based on cellular origin, molecular alterations, and therapy type (Fig. 1).

Fig. 1.

Overview and classification of resistance mechanisms across therapeutic modalities. a Resistance mechanisms to chemotherapy. b Resistance mechanisms to immunotherapy. c Resistance mechanisms to targeted therapy. d Resistance mechanisms to hormonal therapy. Each panel summarizes representative intrinsic and extrinsic resistance pathways associated with the respective therapeutic approach

Tumor-intrinsic determinants of drug resistance

Tumor-intrinsic mechanisms of resistance are rooted in the genetic, epigenetic, and phenotypic complexity of cancer cells. Genomic instability and somatic mutations, which are hallmarks of malignancy, not only shape oncogenic trajectories but also influence therapeutic responsiveness.¹⁹ High tumor mutational burden (TMB),^20,21 microsatellite instability (MSI-H),²² and mismatch repair deficiency (dMMR)²³ are associated with enhanced neoantigen presentation and improved responses to immune checkpoint inhibitors (ICIs).²⁴ Certain mutations paradoxically correlate with ICI resistance despite elevated TMB, highlighting the importance of antigen presentation, mutation type, and immune contexture.^25–27 In contrast, mutations in POLE/POLD1 enhance immunogenicity through hypermutation,²⁸ while excessive copy number alterations often drive immunosuppressive phenotypes and resistance.²⁹ Epigenetic dysregulation further contributes to immune evasion and therapy resistance.^30,31 Aberrant activity of histone-modifying enzymes (e.g., SETDB1, HDAC8, MLL4) silences genes involved in antigen processing, inflammation, and chemokine signaling, thereby dampening T-cell infiltration.^32,33 Moreover, immune checkpoint upregulation (PD-1, CTLA-4, TIGIT, LAG-3) and suppressive signaling cascades such as PTEN loss–PI3K/AKT hyperactivation exacerbate T-cell dysfunction and resistance to ICIs.^26,34,35 In parallel, TCP, defined as the ability of cancer cells to reversibly adopt alternative phenotypic states, including EMT, lineage transdifferentiation, and acquisition of stem-like features, confers phenotypic flexibility that enables drug tolerance.³⁶ Cancer stem-like cells, which thrive in immunosuppressive niches, evade immune detection by downregulating major histocompatibility complex class I (MHC-I) and resisting natural killer (NK) cell–mediated killing.³⁶ Tumor cell state plasticity drives therapeutic relapse by surviving cytotoxic insults and reshaping the TME to suppress effector immune responses.¹ Collectively, tumor-intrinsic mechanisms, spanning genetic mutation, epigenetic reprogramming, immune checkpoint activation, and cellular plasticity, converge to create a highly adaptable, treatment-refractory phenotype.

Tumor-extrinsic factors and the microenvironmental landscape

In addition to cell-autonomous mechanisms, tumor-extrinsic factors profoundly shape therapeutic outcomes. The TME, a complex assembly of stromal cells, immune infiltrates, vasculature, extracellular matrix (ECM), and soluble mediators, plays a decisive role in modulating response and resistance.³⁷ Immunosuppressive features such as tumor-associated macrophages (TAMs), myeloid-derived suppressor cell (MDSC)-mediated cytokine production, regulatory T-cell recruitment, and checkpoint ligand expression foster “cold” tumor states with poor T-cell infiltration.^38,39 Hypoxic and acidic conditions, driven by metabolic reprogramming, further dampen cytotoxic lymphocyte activity while promoting EMT and cancer stemness.⁴⁰ Crosstalk between tumor and stromal components, via exosomes, cytokines, or direct contact, transmits resistance signals across compartments, creating a cohesive barrier to therapy.^40–42 Pharmacokinetics (PK) and drug metabolism, often overlooked, also contribute to resistance.^43,44 Interpatient variability in absorption, distribution, metabolism, and excretion affects drug bioavailability and exposure.⁴⁵ Liver cytochrome P450 (CYP450) enzymes regulate the biotransformation of targeted agents (e.g., imatinib, erlotinib), and polymorphisms in these enzymes can result in therapeutic failure or toxicity.⁴⁶ Overexpression of efflux transporters such as P-glycoprotein (P-gp) and breast cancer resistance protein (BCRP) in tumors limits intracellular drug accumulation, while physical barriers such as the blood‒brain barrier constrain effective delivery to sites such as gliomas.^47,48 Moreover, active or inactive drug metabolites can influence both efficacy and adverse effects, as seen with prodrugs such as cyclophosphamide.⁴⁴ Thus, the extrinsic landscape, including immune architecture, stromal interactions, and pharmacologic variables, functions as a dynamic determinant of drug resistance.⁴³ Therapeutic strategies that remodel the TME, normalize metabolism or optimize PK profiles are critical for durable responses, particularly in the context of immune-based and precision therapies.

Resistance across treatment modalities

Chemotherapy resistance

Resistance to anticancer therapy manifests in distinct and recurring patterns that are characteristic of each therapeutic modality. Resistance to classical chemotherapeutic agents predominantly arises from the activation of multidrug resistance (MDR) mechanisms and a broad spectrum of molecular alterations within tumor cells (Fig. 1a).¹⁸ Given that many chemotherapeutics act by inducing DNA damage, enhanced DNA repair capacity constitutes a central resistance pathway.⁴⁹ For example, PD-L1 has been shown to stabilize mRNAs encoding key DNA repair proteins, thereby enhancing repair capacity and promoting chemoresistance.⁵⁰ Similarly, upregulation of thymidylate synthase (TYMS), a target of 5-fluorouracil (5-FU), contributes to resistance in colorectal cancer by augmenting DNA repair pathways.⁵¹ Dysregulation of programmed cell death also plays a critical role in chemoresistance.⁵² Mutations in TP53 or overexpression of antiapoptotic Bcl-2 family proteins impair apoptotic signaling, allowing tumor cells to evade drug-induced cytotoxicity.⁵³

In addition, EMT contributes to chemotherapy resistance by enhancing antiapoptotic properties, promoting DNA repair, and reprogramming drug metabolism.⁵⁴ Cancer stem cell (CSC)-like traits, characterized by robust DNA repair capacity and increased expression of drug efflux pumps, are tightly linked to chemoresistant phenotypes.⁵⁵ Epigenetic reprogramming, including DNA methylation and histone modifications, as well as noncoding RNAs (e.g., miRNAs, lncRNAs), further modulates transcriptional landscapes to maintain stemness and reshape the TME, thereby reinforcing resistance to chemotherapy.^56,57 Importantly, aberrant activation of oncogenic signaling pathways is another major driver of chemotherapy resistance. Dysregulated pathways such as PI3K/AKT,⁵⁸ MAPK,⁵⁹ NF-κB, and Wnt⁶⁰ not only promote tumor cell survival and proliferation but also modulate drug metabolism, DNA repair, and apoptotic thresholds.

A major mechanism involves the reduction of intracellular drug accumulation through downregulation of solute carrier (SLC) transporters or upregulation of ATP-binding cassette (ABC) transporters, resulting in decreased drug efficacy.⁶¹ The TME itself serves as a pivotal contributor; for instance, myofibroblastic cancer-associated fibroblasts (myCAFs) promote CSC maintenance and metabolic adaptation, collectively enhancing drug resistance in breast cancer.⁶²

Immunotherapy resistance

Although immune checkpoint blockade (ICB) therapies (e.g., anti-PD-1/PD-L1, anti-CTLA-4 antibodies) have revolutionized cancer treatment, both primary and acquired resistance remain major challenges (Fig. 1b).³ Tumor-intrinsic mechanisms of resistance represent a major layer of immune escape. Mechanistically, resistance to ICB can be broadly categorized into tumor-intrinsic determinants and tumor-extrinsic constraints, which together shape immune responsiveness.⁶³ One prevalent mechanism is the disruption of tumor antigen presentation. Loss or downregulation of β2-microglobulin (B2M), a critical component for MHC class I complex assembly, impairs surface antigen presentation and abrogates recognition by cytotoxic CD8⁺ T cells.^64,65 Tumors may also evade immune surveillance by downregulating tumor-associated antigens (TAAs), selecting for neoantigen-depleted clones, or upregulating alternative inhibitory checkpoints, as observed in EGFR-mutant non-small cell lung cancer (NSCLC) following targeted therapy.⁶⁶ Adoptive NK- or CAR-T-based approaches and CD3 bispecific antibodies (BsAbs) are currently being evaluated in preclinical studies and early-phase clinical trials.^64,65 Cytokine agents such as the CD122 agonist bempegaldesleukin have demonstrated immune-expanding activity in phase I-II trials and restored antitumor responses in B2M-null models when combined with PD-1 blockade in preclinical settings.⁶⁷ In parallel, oncolytic viruses and personalized cancer vaccines are under active preclinical investigation to re-establish tumor antigenicity.

From a clonal architecture perspective, tumor neoantigens can be classified as clonal or subclonal. Clonal neoantigens are shared by all cancer cells and represent optimal immune targets, being strongly associated with durable responses to ICB.⁶⁸ In contrast, subclonal neoantigens are confined to specific tumor subpopulations and are more likely to be retained under immune selection pressure, thereby serving as a major source of immune escape and relapse.⁶⁹ ICB preferentially eliminates tumor cells harboring dominant clonal neoantigens, promoting the outgrowth of subclones lacking these immunogenic epitopes, a process known as immune editing.⁷⁰ This evolutionary dynamic has been demonstrated in NSCLC and melanoma treated with PD-1 blockade and explains why tumors with high mutational burden may still acquire rapid resistance to immunotherapy. Furthermore, defects in IFN-γ signaling, often due to loss-of-function mutations in JAK1/2 or STAT1, impair human leukocyte antigen (HLA) upregulation, diminish PD-L1 expression and impede T-cell–mediated antitumor responses, thereby conferring resistance to ICB.⁷¹ Most available evidence supporting these approaches derives from preclinical models and early-phase clinical studies. Robust validation in large and randomized trials remains lacking.

Among tumor-associated antigens, claudin-18.2 (CLDN18.2), a tight junction protein aberrantly exposed on the surface of gastric and other gastrointestinal cancers, has emerged as a clinically actionable target for ex vivo chimeric antigen receptor (CAR)-T-cell therapy.⁷² CAR T-cell therapy directly targets tumor surface antigens via engineered CARs, completely bypassing the tumor cell-intrinsic MHC class I antigen presentation pathway, whereas CAR-NK cell therapy integrates CAR-directed cytotoxicity with the innate missing-self recognition of NK cells, enabling efficient elimination of tumor cells with downregulated or absent MHC class I expression and thereby directly addressing resistance mechanisms.⁷³ Therapeutic neoantigen vaccines are designed to prevent immune evasion and improve clinical outcomes by eliciting T-cell responses against tumor-specific antigens.⁷⁴

In addition to tumor-intrinsic alterations, tumor-extrinsic factors significantly modulate immunotherapy outcomes.⁷⁵ Immune-excluded tumors are characterized by T cells confined to the stromal compartment, a state frequently driven by TGF-β- or VEGF-dependent stromal programs.⁷⁶ Therapeutic strategies therefore aim to remodel the tumor microenvironment and convert these “cold” lesions into immune-inflamed states. Phase II–III clinical trials are evaluating checkpoint blockade in combination with TGF-β pathway inhibitors or antiangiogenic agents to normalize the vasculature and relieve T-cell exclusion.⁷⁷ CAFs, particularly the COL11A1⁺ subtype, induce desmoplastic remodeling and form physical barriers that exclude CD8⁺ T cells. Concurrently, SPP1⁺ TAMs amplify immunosuppressive signals, collectively fostering an immune-exclusion niche that blunts ICB efficacy.⁷⁸ Aberrant activation of oncogenic signaling pathways, such as WNT/β-catenin or PI3K-AKT, can suppress T-cell infiltration and contribute to a noninflamed tumor phenotype.^79,80 Additionally, upregulation of CXCL6 rewires lipid metabolism and promotes neutrophil extracellular trap (NET) formation, which impairs CD8⁺ T-cell function through ROS-mediated mechanisms.⁸¹ Combining immunotherapy with conventional treatments can eradicate tumor cells while concomitantly reversing the immunosuppressive tumor microenvironment.⁸²

At the epigenetic level, resistance is reinforced through reprogramming of immune effector cells. Tumor-derived extracellular vesicles, including exosomal circRNAs, disrupt miRNA networks in T cells, promoting exhaustion and loss of effector function.⁸³ Tumors may further upregulate compensatory inhibitory receptors such as TIM-3, LAG-3, VISTA, and TIGIT, especially under hypoxic conditions that exacerbate immune suppression and facilitate immune escape.⁸⁴ Chronic T-cell receptor (TCR) signaling and sustained antigen exposure contribute to CD8⁺ T-cell senescence, functional paralysis, and eventual deletion, which are hallmarks of acquired resistance.⁸⁵ The conflicting triangle framework highlights a context-dependent interplay in which tumor cell-intrinsic immune checkpoint signaling remodels the TME via EMT induction and autophagy-mediated immune regulation, contributing to immune evasion and resistance.⁸⁶ Importantly, recent studies have shown that the chemokine receptor CXCR4 is highly expressed in exhausted CD8⁺ and regulates CD4⁺ T cells.⁸⁷ Through activation of the JAK2/STAT3 axis and epigenetic remodeling, CXCR4 signaling sustains T-cell dysfunction. Pharmacologic inhibition of CXCR4 has been shown to rejuvenate T-cell function and restore sensitivity to ICB.⁸⁸

The central strategy for overcoming resistance to immunotherapy is to achieve synergistic efficacy through combination regimens, which primarily follow two directions: first, combining immune checkpoint inhibitors with other immunotherapies to activate the immune system at multiple levels; and second, integrating them with conventional therapies to remodel the immunosuppressive tumor microenvironment and enhance tumor immunogenicity, thereby reversing resistance.⁸⁹

Targeted therapy resistance

Molecularly targeted therapies have significantly improved clinical outcomes by selectively inhibiting oncogenic drivers. However, both primary and acquired resistance frequently emerge through diverse genetic, epigenetic, and signaling alterations (Fig. 1c).^90,91 Tumor-intrinsic mechanisms constitute the predominant drivers of resistance to targeted therapies. A canonical mechanism involves the emergence of secondary mutations within drug targets. For instance, in NSCLC, treatment with first- and second-generation EGFR tyrosine kinase inhibitors (TKIs) often leads to the development of the EGFR T790M mutation, while third-generation inhibitors subsequently select for the C797S mutation, both of which confer resistance.^92,93 These alterations are typically addressed by next-generation TKIs. Third-generation EGFR inhibitors, including osimertinib, have demonstrated efficacy against T790M-positive disease in phase III clinical trials, while newer ALK and ROS1 inhibitors have been developed to overcome solvent-front mutations.⁹⁴ Similarly, in CML, point mutations within the BCR-ABL1 kinase domain render imatinib ineffective.⁹⁵ In NSCLC, resistance to MET inhibitors can arise from acquired mutations within the MET kinase domain.⁹⁶

Gene amplification represents an additional mechanism of acquired resistance. Amplifications of MET, KRAS, or other oncogenes can activate parallel or downstream signaling pathways, effectively bypassing the inhibited target and restoring oncogenic signaling.^97,98 Gene amplification and bypass pathway activation represent closely related mechanisms of acquired resistance. Bypass pathway activation engages alternative drivers, including MET, AXL or HER2 amplification, as well as RAS/MAPK or PI3K mutations.⁹⁴ MET amplification occurs in approximately 15–20% of EGFR-TKI–resistant cases. Combinatorial strategies such as dual EGFR–MET inhibition have shown activity in early-phase clinical trials, and parallel co-targeting approaches (ALK–EGFR or MEK–BRAF) are under phase II evaluation in selected contexts. However, most supporting data derive from phase I–II studies or real-world analyses, and definitive phase III validation remains limited. For example, in hepatocellular carcinoma (HCC), activation of EVA1A-mediated PI3K/AKT/p53 signaling drives resistance to lenvatinib,⁹⁹ while NF-κB activation supports tumor survival through transcriptional induction of antiapoptotic genes.¹⁰⁰

Targeted therapy resistance is also linked to evasion of programmed cell death. For example, suppression of ferroptosis, a regulated form of iron-dependent lipid peroxidation, has been implicated in acquired resistance to sorafenib in HCC.¹⁰¹ Enhanced DNA damage repair mechanisms contribute similarly: in BRCA2-deficient tumors, reduced EZH2 expression stabilizes replication forks, diminishing sensitivity to PARP inhibitors (PARPis).¹⁰² Likewise, overexpression of NSD3S in prostate cancer strengthens replication fork stability, leading to PARPi resistance.¹⁰³

TCP presents another formidable challenge. EGFR-mutant NSCLC may undergo histological transformation into small cell lung cancer (SCLC) under TKI pressure, which is inherently resistant to EGFR-targeted therapy.¹⁰⁴ Similarly, prostate cancers treated with androgen receptor (AR)-targeted agents can transdifferentiate into neuroendocrine phenotypes, a lineage shift associated with aggressive progression and therapy resistance.^105–107

In addition to cell-intrinsic alterations, tumor-extrinsic mechanisms can also contribute to resistance. Altered drug transport mechanisms contribute to MDR. Upregulation of efflux transporters such as MDR1 (ABCB1) has been shown to mediate cross-resistance to chemotherapeutics and targeted agents, including paclitaxel and olaparib, in ovarian cancer.¹⁰⁸ In colorectal cancer, BRD4/SMAD3/4–regulated PAI-1 secreted by M2-TAMs binds to uPAR on cancer cells, activates JAK2/STAT3 signaling, and blocks oxaliplatin-induced apoptosis.¹⁰⁹ Targeted therapies, including AXL or TAM-family receptor inhibition, are currently undergoing early-phase clinical evaluation and remain largely supported by preclinical evidence.⁹⁴ However, the management of lineage-switched tumors largely relies on retrospective analyses and accumulated clinical experience, as prospective randomized data are still limited.

Hormonal therapy resistance

Primary endocrine resistance in breast cancer is often associated with the loss of estrogen receptor (ER) expression, while the aberrant activation of bypass signaling pathways and mutations in the gene encoding ERα constitute key mechanisms of acquired resistance (Fig. 1d).^110,111 Causes of ER expression loss include epigenetic regulation, hypoxia, overexpression of epidermal growth factor receptor (EGFR) or human epidermal growth factor receptor 2 (HER2), and dysregulation of tumor suppressors such as p53 and pRb2/p130.^112,113

In metastatic hormone receptor–positive breast cancer, mutations in the ligand-binding domain (LBD) of estrogen receptor 1 (ESR1), which encodes ERα, contribute to treatment resistance.¹¹⁴ The most common ESR1 mutations are Y537S and D538G.¹¹⁵ Prolonged aromatase inhibitor therapy can also induce point mutations in ESR1, resulting in constitutive ERα activation independent of estrogen ligands, thereby promoting resistance.¹¹⁶

In ER-positive breast cancer, PIK3CA mutations drive hormone-independent proliferation and enable cancer cells to bypass estrogen signaling. Meanwhile, aberrantly activated PI3K signaling upregulates ERα expression via sustained activation of the PI3K/AKT/mTOR pathway, forming a resistance-promoting feedback loop. This also reprograms glucose and glutamine metabolism to support tumor survival and undermines endocrine efficacy.¹¹⁷ Additionally, MAPK pathway hyperactivation, driven by NF1 loss, induces ligand-independent ER activation through persistent RAS signaling, contributing to endocrine resistance. This alteration is enriched in approximately 22% of endocrine-resistant advanced breast cancers.¹¹⁸ In ER⁺/HER2⁻ breast cancer, hyperactivated AKT1 binds to the kinase domain of TBK1, disrupting the STING-TBK1-IRF3 complex, thereby impairing downstream interferon signaling and blocking innate immune responses triggered by DNA damage. Inactivation of the cGAS–STING pathway in turn reinforces AKT1 activation, forming a self-sustaining loop that facilitates immune evasion and drives the persistence of resistance.¹¹⁹

In prostate cancer, point mutations and gene amplification of the AR, as well as the generation of splice variants such as AR-V7, underlie resistance to androgen deprivation therapy (ADT).¹²⁰ Acetylation of AR at lysine 609 (K609) enhances its binding to enhancer regions of both the AR gene and the ACK1 kinase gene, even in the presence of AR antagonists.¹²¹ This forms a positive feedback loop that continuously activates AR signaling and ACK1 expression. ACK1 further phosphorylates AR at Y267, sustaining the oncogenic activity of acK609-AR and promoting the growth of castration-resistant prostate cancer (CRPC).¹²¹ In the neuroendocrine subtype of prostate cancer (NEPC), the histone reader ZMYND8 is highly upregulated. ZMYND8 cooperates with the transcription factor FOXM1 to bind critical genomic loci and recruit the SWI/SNF chromatin remodeling complex, thereby activating neuroendocrine gene expression programs and promoting lineage plasticity that contributes to therapy resistance.¹⁰⁵

Beyond tumor cell–intrinsic alterations, endocrine resistance may also be shaped by the TME. In CRPC, AR signaling upregulates inhibitory immune checkpoints (e.g., B7-H3) and proinflammatory cytokines (e.g., IL-6, TNF-α), which recruit immunosuppressive cells and suppress T-cell function.¹²² Despite ongoing research, many mechanisms underlying endocrine resistance remain incompletely understood. Elucidating these mechanisms not only enhances the prediction of endocrine therapy response but also provides a theoretical foundation for the development of next-generation targeted and combination therapies. Moving forward, dynamic monitoring of resistance-related pathways, integration of multiomics analyses, and implementation of immune intervention strategies will be essential to develop personalized and mechanism-guided endocrine treatment approaches.

Cross-cutting features and implications

Despite the mechanistic heterogeneity of drug resistance across cancer types and treatment modalities, convergent themes emerge that bridge tumor-intrinsic and tumor-extrinsic origins.¹²³ A central paradigm is the evolutionary selection of resistant clones under therapeutic pressure.¹²⁴ Within heterogeneous tumor populations, preexisting or newly acquired genetic and epigenetic alterations confer survival advantages to specific subclones, which expand during treatment.¹²⁵ This is exemplified by immune evasion via antigen loss variants under immunotherapy or secondary mutations in driver oncogenes (e.g., EGFR-T790M, ALK-L1196M, BCR-ABL-T315I) in the context of targeted therapies.⁹⁰ Such clonal dynamics underscore the adaptive nature of cancer and the need for early combination or sequential regimens that preempt resistance emergence.³⁸ Furthermore, regardless of therapy type, immune escape is a universal barrier.¹²⁶ Resistant tumors often downregulate antigen-presenting machinery, upregulate inhibitory ligands, and attract immunosuppressive cells such as regulatory T cells (T_regs) and MDSCs.¹²⁷ Hypoxia and metabolic reprogramming within the TME additionally impair T-cell infiltration and function, reinforcing an immunologically “cold” phenotype.¹²⁸ These insights support the integration of immune-restorative strategies, including checkpoint inhibitors, innate immune activators or adoptive cell therapies, into standard treatment paradigms.

Another unifying feature of resistance is signaling pathway redundancy and compensatory network activation.¹²⁹ Oncogenic signaling is inherently plastic, and blockade of one pathway often triggers feedback reactivation or lateral bypass through alternative routes.¹¹⁷ For instance, inhibition of MAPK signaling frequently leads to compensatory activation of the PI3K-AKT or JAK/STAT pathways.¹³⁰ In endocrine-resistant breast and prostate cancers, growth factor receptor cross-talk circumvents hormonal dependence.² These rewiring events not only restore proliferative capacity but also reprogram tumor metabolism and immune evasion. Importantly, such adaptability creates synthetic vulnerabilities that may be exploited through rational drug combinations targeting parallel or downstream nodes.¹³¹ Recognizing these adaptive circuits enables the design of multiplexed therapeutic interventions to forestall resistance.

Finally, cancer cell plasticity and TME-mediated support form an interdependent axis of resilience.^123,132 In melanoma, BRAFi/MEKi-induced reactivation of MAPK signaling in tumor cells promotes the secretion of CXCL1, HGF, and TGF-β, thereby recruiting cancer-associated fibroblasts and immunosuppressive myeloid cells and remodeling the extracellular matrix to generate a drug-resistant microenvironment.¹³³ Tumor cells can undergo reversible phenotypic transitions, such as EMT, lineage switching, or entry into a DTP state, facilitated by epigenetic remodeling. These cell states exhibit reduced proliferation, increased resistance to apoptosis, and decreased immunogenicity, collectively enabling survival during cytotoxic or immune-based therapies.¹²³ Simultaneously, the TME provides biochemical and mechanical cues that sustain these plastic states.¹³⁴ CAFs, TAMs, and extracellular matrix components modulate drug accessibility, alter immune cell trafficking, and preserve CSC niches.¹³⁵ Cytokines such as TGF-β and IL-6 further enforce these adaptations by reshaping the immune and metabolic landscape.^136,137 Together, these features define resistance as a systems-level phenomenon, driven not by single mutations but by dynamic cellular and microenvironmental networks, necessitating integrated, multitargeted therapeutic strategies.

Mechanisms and strategies for combating cancer drug response and resistance

Altered drug transport and metabolism

One of the central barriers to effective chemotherapy is the development of MDR, wherein tumor cells acquire cross-resistance to structurally and mechanistically distinct agents (Fig. 2a).¹⁸ A hallmark of MDR is the dysregulation of intracellular drug availability, primarily through impaired drug uptake and enhanced drug efflux, involving the SLC and ABC transporter families.^138–140 These transport mechanisms, often coupled with vesicular trafficking and metabolic reprogramming, significantly reduce intracellular drug accumulation, attenuate target engagement and contribute to immune evasion, collectively undermining therapeutic efficacy.

Fig. 2.

Molecular mechanisms and combating strategies in cancer drug resistance: drug transport, metabolism, and signaling pathways. a Altered drug transport and metabolism, including decreased uptake mediated by solute carrier (SLC) transporters, enhanced efflux via ATP-binding cassette (ABC) transporters, and dysregulated pharmacokinetics/metabolic enzymes. b Alterations in DNA damage repair (DDR) mechanisms and corresponding therapeutic vulnerabilities. c Aberrant activation of major signaling pathways, including PI3K-AKT/mTOR, MAPK, NF-κB, TGF-β, Wnt, JAK/STAT, Notch, Hippo, and GAS6/AXL, along with potential targeted intervention strategies

Drug uptake: the role of SLC

SLC transporters, a large superfamily of membrane proteins responsible for nutrient and metabolite transport, also mediate the uptake of various chemotherapeutics (Fig. 2a).¹⁴¹ Their downregulation or functional loss has been implicated in primary and acquired resistance.¹⁴⁰ For instance, decreased expression of the organic cation transporter OCT2 (SLC22A2) confers resistance to platinum-based drugs in pancreatic and ovarian cancers by reducing cellular drug accumulation.^138,139 Similarly, the itaconate transporter SLC13A3 facilitates the uptake of itaconate (ITA), a TAM-derived immunometabolite. Elevated SLC13A3 expression in tumor cells stabilizes PD-L1 through ITA-mediated posttranslational modification, promoting immune escape and resistance to checkpoint blockade.¹⁴² Targeting SLC13A3 restores PD-L1 degradation and improves ICI responsiveness, highlighting a direct link between metabolite transport and immunotherapeutic resistance.¹⁴²

Enhanced drug efflux: ABC transporters and MDR

ABC transporters constitute the most prominent efflux systems driving MDR (Fig. 2a).¹⁴³ Powered by ATP hydrolysis, they actively extrude cytotoxic agents from tumor cells, thereby reducing intracellular drug concentrations below therapeutic thresholds.¹⁴⁴ Among them, P-gp/ABCB1,¹⁴⁵ MDR-associated proteins (MRPs/ABCCs),¹⁴⁶ and BCRP/ABCG2¹⁴⁷ are most extensively studied.

P-gp, encoded by ABCB1, is broadly expressed in various tumor types and correlates with disease progression, therapy resistance, and poor prognosis.¹⁴⁸ P-gp substrates are structurally diverse, including paclitaxel, doxorubicin, vincristine, imatinib, and olaparib.¹⁴⁹ Beyond tumors, P-gp is physiologically expressed in the gastrointestinal tract, renal tubules, liver, and blood–brain barrier, contributing to systemic drug clearance.¹⁵⁰ Notably, paclitaxel-induced P-gp overexpression in mitochondria has been implicated in acquired resistance in ovarian cancer.¹⁵¹ Elevated MDR1/MRP1 expression is frequently associated with tumor recurrence and metastasis, and metastatic lesions often display higher P-gp levels than corresponding primary tumors.¹⁵² Proteomic profiling further identifies MDR1 upregulation as a critical driver of both intrinsic and acquired resistance to proteolysis-targeting chimeras (PROTACs), extending its relevance to emerging targeted protein degradation strategies.¹⁵³

Complementing MDR1-mediated drug efflux, the multidrug resistance-associated protein (MRP) subfamily comprises nine members (MRP1–9) that confer resistance through partially distinct yet overlapping mechanisms.^154–156 MRPs preferentially export drug–glutathione conjugates and promote drug sequestration into perinuclear vesicles, thereby restricting intracellular and nuclear drug availability.¹⁵⁷ For example, MRP1 mediates resistance to methotrexate, cisplatin, and vinca alkaloids and is frequently overexpressed in lung, ovarian, and hematologic malignancies.¹⁵⁸ MRP4 and MRP5 facilitate the efflux of cyclophosphamide and 5-fluorouracil and are transcriptionally regulated by Nrf2 signaling and therapeutic stress.¹⁵⁴

BCRP, a half-transporter encoded by ABCG2, functions as a homodimer and mediates the efflux of topotecan, SN-38, mitoxantrone, and several TKIs, such as imatinib and osimertinib.^147,159 Elevated BCRP expression is associated with chemoresistance and unfavorable prognosis in leukemia, breast, and ovarian cancers.^160,161

Drug pharmacokinetics and metabolism

Pharmacokinetics, the study of drug absorption, distribution, metabolism, and excretion, is increasingly recognized as a determinant of therapeutic efficacy (Fig. 2a).⁴⁴ Interpatient variability in PK, influenced by genetic polymorphisms, organ function, and TME characteristics, profoundly affects drug response.⁴⁵ The lung resistance protein (LRP), encoded by the major vault protein (MVP) gene, contributes to MDR by mediating intracellular vesicular sequestration of drugs.¹⁶² By trafficking cytotoxic agents into exocytic vesicles or blocking nuclear entry, LRP prevents engagement with nuclear DNA or enzymes.¹⁶² The kinesin KIF4A regulates LRP trafficking, and elevated LRP expression has been associated with resistance in lung and ovarian cancers.¹⁶³ CYP450 enzymes in the liver regulate the metabolism of many anticancer drugs (e.g., imatinib and erlotinib), and altered activity can lead to reduced efficacy or increased toxicity.⁴⁶ High intratumoral expression of CYP3A4 can accelerate the metabolic inactivation of chemotherapeutic agents such as paclitaxel and docetaxel, thereby reducing local drug bioavailability, diminishing therapeutic efficacy, and ultimately contributing to drug resistance.¹⁶⁴ Conversely, the loss of CYP3A4 expression in hepatocellular carcinoma impairs the bioactivation of prodrugs, leading to treatment failure. In addition, tumor-associated systemic inflammation can suppress hepatic CYP3A4 activity, altering systemic drug pharmacokinetics. This disruption is further exacerbated by individual genetic polymorphisms, polypharmacy, and hepatic or renal dysfunction, collectively disturbing metabolic homeostasis.¹⁶⁴ These factors modulate drug exposure and the generation of active metabolites, ultimately fostering the emergence of therapeutic resistance.

Tissue distribution also influences therapeutic penetration; for instance, limited blood‒brain barrier permeability hampers glioblastoma treatment.¹⁶⁵ Biotransformation pathways may generate active metabolites that enhance or reduce efficacy or cause off-target toxicity.¹⁶⁶ Cyclophosphamide, for example, requires hepatic activation, but excessive metabolism may cause marrow suppression.¹⁶⁷ Personalized dosing, PK monitoring, and targeted modulation of metabolic pathways are thus critical to optimizing cancer treatment and overcoming resistance.

Therapeutic targeting of transport-mediated resistance

Given the central role of efflux transporters in mediating MDR, extensive efforts have been dedicated to the development of pharmacological strategies aimed at overcoming transporter-driven drug efflux.¹⁶⁸ Over the past few decades, inhibitors targeting key transporters such as P-gp, MRPs, and BCRP have evolved through three major generations.¹⁶⁹ First-generation agents, including verapamil and cyclosporin A, showed in vitro efficacy but were hampered by off-target toxicity and limited clinical applicability. Second-generation inhibitors such as valspodar offered improved potency but interfered with CYP450 enzymes (e.g., CYP3A4), leading to problematic drug–drug interactions.¹⁷⁰ More recently, third-generation compounds such as tariquidar and elacridar demonstrated enhanced specificity and pharmacokinetic profiles, showing promise in preclinical models when combined with agents such as paclitaxel and irinotecan.¹⁷¹ However, their clinical translation has been modest, underscoring the complexity of transporter-targeted therapy.

To circumvent the limitations of direct transporter inhibition, alternative approaches have emerged. Nanoparticle-based delivery systems, including liposomes, polymeric micelles, PEGylated constructs, and PLGA nanoparticles, can bypass efflux recognition and enhance tumor-selective drug accumulation, thereby restoring therapeutic efficacy.¹⁷² Natural compounds such as curcumin, quercetin, and salvianolic acid B have also demonstrated transporter-inhibitory properties with relatively low toxicity, although issues with solubility and bioavailability have limited their clinical utility and prompted formulation optimization.¹⁷³ At the molecular level, gene modulation techniques, including siRNA- and shRNA-mediated silencing of efflux transporter genes (e.g., ABCB1, ABCG2), as well as CRISPR/Cas9-based gene editing, have successfully resensitized tumor cells to chemotherapy in preclinical models.¹⁷⁴ Similarly, endogenous microRNAs (e.g., miR-451 and miR-326) have been shown to negatively regulate MDR1 expression and restore drug sensitivity in breast and gastric cancers.^175,176 In parallel, novel therapeutic strategies are being explored to target functional domains of efflux transporters.¹⁴³ These include inhibitors that bind ABCs, transmembrane domain mimetics that disrupt substrate recognition, and competitive peptide antagonists.¹⁷⁷ One particularly promising avenue involves metabolic disruption: depletion of intracellular ATP, which powers transporter function, has been shown to impair drug efflux.¹⁷⁸ For example, hybrid nanodrugs (HREDs) derived from citrus extracellular vesicles reduce ATP levels and suppress efflux activity, demonstrating potent antitumor effects in drug-resistant ovarian cancer models.¹⁷²

Efflux-mediated resistance rarely occurs in isolation and often coexists with other resistance mechanisms, such as enhanced DNA repair, apoptosis evasion, or immune suppression.^179–182 This has led to the development of rational combination strategies aimed at targeting multiple resistance nodes simultaneously.¹⁸¹ Notably, the combination of MDR1 inhibitors with PARPis has been shown to restore olaparib sensitivity in resistant ovarian cancer.¹⁰⁸ Similarly, combined inhibition of ABCB1 (MDR1) and the use of PROTACs targeting the KRAS and MAPK signaling pathways enhances the treatment response in KRAS-mutant colorectal cancer xenograft models.¹⁵³ Despite challenges such as compensatory upregulation of alternative transporters and feedback activation of resistance pathways, the transportome remains a tractable and druggable vulnerability.¹⁸³ Moving forward, the integration of transporter profiling, nanotechnology-based delivery platforms and immunomodulatory interventions holds promise for overcoming transporter-mediated drug resistance and achieving more durable responses in refractory cancers.

Alterations in DNA damage repair (DDR)

Molecular mechanisms in DDR

DDR safeguards genomic stability under genotoxic stress, and its deficiency confers pronounced sensitivity to genotoxic therapies and PARP inhibitors while also driving resistance through repair restoration or pathway bypass.¹⁸⁴ The most prevalent DNA lesions include single-strand breaks (SSBs) and double-strand breaks (DSBs), which are sensed by key damage sensors such as ataxia telangiectasia mutated (ATM), ATM and Rad3-related (ATR), and poly (ADP-ribose) polymerase 1 (PARP1). Activation of these sensors initiates checkpoint cascades, most notably ATM–Chk2 and ATR–Chk1, which coordinate cell cycle arrest, DNA repair pathway selection and apoptosis or senescence when repair fails.^184,185 The downstream repair machineries comprise homologous recombination repair (HRR), nonhomologous end joining (NHEJ), base excision repair (BER), nucleotide excision repair (NER), and mismatch repair (MMR).¹⁸⁵

In normal cells, DDR is indispensable for maintaining genomic fidelity and preventing malignant transformation. In cancer, however, DDR exerts a paradoxical influence: while repair deficiencies promote mutagenesis and tumorigenesis, restoration or hyperactivation of DDR pathways in established tumors fosters resistance by enabling efficient repair of therapy-induced DNA damage.¹⁸⁶ These alterations underlie resistance not only to cytotoxic chemotherapy and radiotherapy but also to molecularly targeted agents and immunotherapies.^187,188 Deciphering the context-specific regulation and plasticity of DDR, as well as its interplay with oncogenic signaling and tumor evolution, is critical for the development of rational therapeutic interventions (Fig. 2b).

Tumors with inherited or somatic deficiencies in HRR genes, such as BRCA1, BRCA2, PALB2, and RAD51, exhibit synthetic lethality to PARPis, including olaparib, rucaparib, niraparib, and talazoparib, which have shown durable efficacy in BRCA-mutant breast, ovarian, and prostate cancers.^187,188 Nevertheless, clinical benefit is often limited by intrinsic and acquired resistance. For example, in pancreatic cancer, the inflammasome component NLRP4 mediates PARPi resistance by promoting NOXO1 expression, which inhibits nuclear translocation of SIRT7, thereby enhancing ROS-induced autophagy and reactivating DDR via reduced γH2AX accumulation and impaired BRCA1–RAD51 complex formation.¹⁸⁹ In parallel, error-prone DNA damage tolerance mechanisms, such as translesion synthesis (TLS), contribute to therapy evasion.¹⁹⁰ TLSs allow replicative bypass of alkylated or crosslinked lesions and prevent replication fork collapse. Upregulation of TLS has been implicated in resistance to combined olaparib and temozolomide in SCLC.¹⁹⁰ These insights highlight the importance of identifying predictive biomarkers and vulnerabilities associated with repair pathway rewiring.

Beyond PARPis, multiple DDR kinases, including ATM, ATR, CHK1/2, WEE1, and DNA-dependent protein kinase catalytic subunit (DNA-PKcs), have emerged as therapeutic targets.¹⁹¹ For instance, WEE1 inhibition (e.g., with AZD1775) abrogates G2/M checkpoints, thereby potentiating the cytotoxicity of DNA-damaging agents such as cisplatin.¹⁹² ATR inhibitors (e.g., BAY1895344) demonstrate promising activity in DDR-deficient tumors and are undergoing clinical evaluation across solid tumor types.¹⁹³ Furthermore, combining DDR inhibitors with ICB enhances antitumor immunity by promoting neoantigen exposure, micronuclei formation, and activation of the cGAS–STING pathway.¹⁹⁴ Conversely, restoration or upregulation of DDR components can drive resistance. Intriguingly, PD-L1, a key immunoregulatory molecule, can act as an RNA-binding protein to stabilize transcripts encoding DDR proteins such as BRCA1, MRE11, and RAD50, thereby enhancing DNA repair and promoting resistance to radiotherapy and chemotherapy.⁵⁰ D-mannose, by activating AMPK and impairing PD-L1 glycosylation, induces PD-L1 degradation and disrupts HRR, sensitizing tumors to ionizing radiation.¹⁹⁵

Epigenetic and chromatin remodeling regulators modulate DDR efficiency and replication fork stability. In BRCA2-deficient cells, EZH2-mediated H3K27 trimethylation at stalled forks facilitates MUS81 recruitment and fork restart. Loss of EZH2 impairs this mechanism, stabilizing forks and attenuating PARPi efficacy.¹⁰² Similarly, GRB2 stabilizes RAD51 at reversed forks by inhibiting its ATPase activity; GRB2 depletion leads to fork collapse, cytosolic DNA accumulation, cGAS–STING activation, and increased PARPi sensitivity.¹⁹⁶ Complex crosstalk between DDR and oncogenic survival pathways further complicates therapy. In RAS- or PI3K-driven tumors, elevated reactive oxygen species (ROS) levels induce oxidative DNA damage.¹⁹⁷ Tumor cells adapt by upregulating BER components (e.g., Polβ, APE1, FEN1) and scaffolding proteins such as CUX1/2 and SATB1 to promote repair and evade senescence.¹⁹⁸ Conversely, DNA-PK and ATM activate AKT in response to DNA damage, reinforcing DDR capacity and suppressing apoptosis via p53 inhibition.¹⁹⁹ Radiation-induced exosomes in esophageal squamous cell carcinoma deliver HMGB1, which activates PI3K–AKT–FOXO3A signaling and elevates γH2AX expression in recipient cells, promoting radioresistance.²⁰⁰ Genomic amplifications of DDR genes also contribute to broad-spectrum resistance.¹⁹⁹ Amplification of NBN, RAD51, or PARP1 has been associated with reduced sensitivity to more than 30 targeted agents across cancer types.²⁹ Despite these insights, DDR-targeted therapies remain challenged by nonselective toxicity, resistance heterogeneity, limited biomarker availability and modest clinical durability.

Combating strategies in DDR

To address DDR-mediated therapeutic resistance, multiple strategies have been proposed that exploit vulnerabilities in DNA repair pathways (Fig. 2b).¹⁸⁸ The synthetic lethality paradigm remains central, exemplified by the success of PARP inhibitors in HRR-deficient tumors.¹⁸⁵ To circumvent resistance and expand the scope of DDR targeting, next-generation agents such as ATR, CHK1, WEE1, and DNA-PK inhibitors are being investigated, both as monotherapies and in combination with DNA-damaging agents.^184,185,199 Rational combinations, such as PARPis plus ATR or WEE1 inhibitors, aim to exacerbate replication stress and collapse replication fork stability, particularly in tumors with partial or context-specific DDR competence.^188,191

Combination approaches that integrate DDR inhibition with immunotherapy are of increasing interest.²⁰¹ DDR-targeted agents promote immunogenic cell death by inducing micronuclei formation, enhancing TMB, and activating innate immune pathways such as cGAS–STING, thereby augmenting ICB efficacy.^185,188,202 Early-phase trials of PARPis combined with PD-1/PD-L1 inhibitors have shown promising responses, particularly in tumors refractory to ICB alone.²⁰¹ Targeting metabolic and chromatin remodeling pathways represents an additional strategy to sensitize DDR-proficient tumors. Modulating NAD⁺ metabolism, histone modifiers (e.g., EZH2), or ATP-dependent chromatin remodelers (e.g., BRG1, CHD4) may shift the balance of repair pathway utilization, destabilize replication forks, or impair lesion recognition.¹⁰² Furthermore, inhibiting DDR-related posttranslational modifications, such as ATM/ATR-dependent phosphorylation or ubiquitination cascades, can selectively impair tumor cell repair fidelity without disrupting normal cells.¹⁹⁹

Advances in precision oncology have also enabled real-time monitoring and individualized DDR targeting. Functional genomic screens, single-cell profiling, and liquid biopsy-based analyses of DDR mutations or gene expression signatures offer tools for patient stratification and adaptive treatment modulation.^12,13,203 These technologies are expected to refine predictive biomarker discovery and uncover resistance trajectories. Looking forward, a comprehensive understanding of DDR pathway rewiring, coupled with integrated omics-based profiling and network-level vulnerability mapping, will be pivotal.¹² Combinatorial approaches involving DDR inhibitors, epigenetic modulators, metabolic disruptors and immunotherapeutics hold promise for overcoming resistance and achieving durable clinical responses across malignancies.

Altered signaling pathways

At the systems level, diverse oncogenic signaling pathways intersect to drive common downstream programs that sustain tumor survival, plasticity, metabolic fitness, DNA repair capacity, and immune suppression, forming the mechanistic basis of therapeutic resistance (Fig. 2c).^{80,179,204,205}

PI3K/AKT and mTOR signaling

The PI3K-AKT-mTOR axis plays a central role in orchestrating tumor cell growth, survival, metabolic reprogramming, immune evasion, and apoptosis resistance.^204,205 Aberrant activation of this pathway is a well-established mechanism of resistance to targeted therapies, chemotherapy, and endocrine therapy.²⁰⁶ Activation is typically initiated by upstream receptor tyrosine kinases (RTKs) or G protein–coupled receptors (GPCRs), which trigger PI3K to convert PIP2 into PIP3, thereby recruiting and activating AKT.²⁰⁷ Activated AKT subsequently stimulates mTOR complexes (mTORC1/2), modulating a wide array of oncogenic processes.^208,209

This pathway is normally restrained by the tumor suppressor PTEN, which dephosphorylates PIP3 and prevents AKT overactivation.²¹⁰ In hormone receptor–positive/HER2-negative breast cancers, PI3K-AKT-mTOR is among the most frequently activated signaling axes.²¹¹ Hyperactivation commonly arises from mutations in PIK3CA (detected in ~35-40% of ER⁺/HER2⁻ cases), AKT1 mutations, or PTEN loss.^211,212 HER2 amplification further potentiates PI3K signaling.^213,214 Activation of this pathway also shapes the TME. For instance, tumor-associated macrophage (TAM)-derived CCL2 activates PI3K-AKT-mTOR signaling, promoting tamoxifen resistance in breast cancer via a prosurvival feedback loop.²¹⁵ In NSCLC, PIK3CA mutations are associated with acquired resistance to EGFR TKIs.²⁷ In HCC, loss of EVA1A leads to PI3K/AKT/MDM2 axis activation, resulting in destabilization of p53 and resistance to lenvatinib.⁹⁹

MAPK signaling pathway

The mitogen-activated protein kinase (MAPK) pathway is a highly conserved signaling module regulating cellular proliferation, differentiation, apoptosis, inflammation, and stress responses.²¹⁶ It comprises four primary branches: ERK, p38 MAPK, JNK, and ERK5. The canonical MAPK pathway involves the RAS–RAF–MEK–ERK cascade, typically activated by RTKs such as EGFR, FGFR and HER2. Sequential phosphorylation through this cascade culminates in the activation of nuclear transcription factors such as ELK-1, FOS, and MYC.^216,217

Acquired mutations within MAPK pathway components represent a major mechanism of resistance to targeted therapy.²¹⁸ Secondary mutations in BRAF, KRAS, NRAS, and MEK1/2 have been widely observed in patients treated with BRAF and MEK inhibitors, restoring ERK signaling and driving therapeutic failure.²¹⁹ These mutations enable tumor cells to escape pharmacologic inhibition while maintaining MAPK pathway dependency.

MAPK signaling contributes to therapeutic resistance by enhancing the expression of drug efflux transporters such as MDR1 and P-gp and by upregulating antiapoptotic genes such as Bcl-2.²²⁰ In melanoma, resistance to BRAF inhibitors frequently arises from reactivation of MAPK signaling or activation of alternative oncogenic pathways.²²¹ In colorectal cancer, elevated EGFR ligand expression (e.g., AREG, EREG) sustains EGFR–MAPK signaling and underlies resistance to EGFR-targeted therapies.²²² In breast cancer, NF1 loss drives resistance to PI3Kα inhibitors by activating RAS-MAPK signaling and promoting metabolic rewiring, including increased glycolysis and decreased ROS.²²³ In gastric cancer, downregulation of circMAPK1 decreases its encoded tumor-suppressive micropeptide MAPK1-109aa, disrupting competitive inhibition of the MEK1–MAPK1 interaction and resulting in sustained MAPK activation and immune evasion.²²⁴ The S100A8/A9–TLR4–p38 MAPK–NF-κB axis also contributes to immune resistance. This signaling cascade induces CXCL1 secretion by gastric cancer cells, driving recruitment of polymorphonuclear myeloid-derived suppressor cells (PMN-MDSCs), CD8⁺ T-cell exhaustion, and resistance to ICIs.²²⁵

NF-κB signaling pathway

NF-κB (nuclear factor kappa-light-chain-enhancer of activated B cells) represents a master transcription factor family governing immune and inflammatory responses, cell survival, DNA repair, metabolism, and stemness.²²⁶ It is activated through two major branches, the canonical (IκB-dependent) and noncanonical (NIK-dependent) pathways, both of which are frequently dysregulated in tumors and their microenvironments. In the canonical pathway, inflammatory cues activate the IκB kinase (IKK) complex, particularly IKKβ, which phosphorylates IκB proteins.²²⁷ This leads to their degradation and nuclear translocation of NF-κB dimers (typically p50/RelA), where they regulate gene transcription. The noncanonical pathway is stimulated by cytokines such as CD40L or BAFF, involving NIK-mediated processing of p100 to p52 and formation of RelB/p52 dimers that translocate into the nucleus.²²⁸

NF-κB confers resistance by upregulating antiapoptotic proteins (e.g., Bcl-2, Bcl-xL, XIAP, c-FLIP), stemness regulators, drug efflux transporters, and DNA repair enzymes.¹⁰⁰ In ER⁺ breast cancer, Dll1-expressing quiescent tumor stem cells activate Notch–NF-κB signaling to promote survival and resistance to doxorubicin.^229,230 The heat shock protein HSPB1 induces IκB-α ubiquitination and degradation, sustaining NF-κB activation and blocking ferroptosis. Nude mouse xenograft and lung metastasis models further confirmed the in vivo functional role of HSPB1.²³¹ In cisplatin-resistant bladder cancer, constitutive NF-κB activity upregulates EMT and stemness markers, as well as ABCB1, driving resistance and progression.¹⁷⁹ In mantle cell lymphoma, compensatory activation of noncanonical NF-κB supports survival pathways and mediates resistance to Bruton tyrosine kinase (BTK) inhibitors such as ibrutinib.²³²

In HCC, overexpression of EIF5B enhances IκB phosphorylation and activates NF-κB, contributing to EMT, stemness, and 5-FU resistance.¹⁰⁰ RelB, a key transcription factor of the noncanonical pathway, directly binds the CD274 promoter to upregulate PD-L1, promoting immune escape in prostate cancer.²³³ Moreover, IFN-γ stimulation induces RIPK1 expression, which sustains NF-κB activity. This enhances the transcription of antiapoptotic genes (e.g., CFLAR and TNFAIP3) and promotes the secretion of immunosuppressive cytokines (e.g., CCL2, CXCL1, and IL-10), recruiting ARG1⁺TREM2⁺-suppressive myeloid cells. Collectively, these events reduce T/NK cell infiltration and cytotoxicity, contributing to ICB resistance.²³⁴

TGF-β signaling pathway

The transforming growth factor-β (TGF-β) superfamily, including TGF-β1/2/3, bone morphogenetic proteins (BMPs), activins, and nodal, exerts pleiotropic effects on cancer biology, with context-dependent roles as both a tumor suppressor and promoter.²³⁵ In early stages, TGF-β suppresses tumorigenesis by inducing cell cycle arrest and apoptosis; however, in advanced cancers, it drives EMT, immune evasion, stemness, and therapy resistance.

TGF-β signaling operates through canonical (Smad-dependent) and noncanonical (Smad-independent) pathways.²³⁶ Upon ligand binding, TGF-β receptor II (TβRII) recruits and phosphorylates TβRI, which activates Smad2/3. The Smad2/3–Smad4 complex then translocates into the nucleus to regulate transcription. Alternatively, TGF-β can activate the PI3K/AKT, MAPK, Rho-GTPases, and JNK/p38 pathways via noncanonical signaling.²³⁵ A notable positive feedback loop amplifies TGF-β signaling: TGF-β-induced Smad2/3 activation upregulates TGFBRAP1, which competitively binds TβRI and prevents its degradation by Smurf1/2-mediated ubiquitination. Stabilization of TβRI sustains pathway activity, promotes CSC-like phenotypes, and confers resistance to TKIs such as regorafenib in HCC.²³⁷

Chemotherapy can also reinforce TGF-β–driven resistance. CXCL1 and CXCL5, secreted in response to therapy, recruit neutrophils that release neutrophil extracellular traps (NETs). These NETs bind latent TGF-β via integrin αvβ1 and activate it through matrix metalloproteinase 9 (MMP9). Active TGF-β subsequently phosphorylates Smad2, promoting EMT, CSC traits, and chemoresistance.¹³⁶ In BRAFV600E-mutant cancers, loss of the aryl hydrocarbon receptor (AhR) reduces expression of its repressor AhRR, allowing ARNT to accumulate in the nucleus. ARNT then interacts with Smad2/3, hyperactivating TGF-β/Smad signaling. This compensatory mechanism supports MAPK inhibitor resistance and promotes lineage plasticity and tumor progression.²³⁸ Platinum-based chemotherapy can induce tumor cell senescence, a state that is increasingly recognized as a contributor to acquired resistance. Senescent tumor cells secrete a TGF-β-enriched senescence-associated secretory phenotype (SASP), which activates AKT-mTOR signaling, reshapes the TME, and promotes malignant progression under therapeutic pressure.

Wnt signaling pathway

The Wnt signaling network is a master regulator of embryonic development, tissue regeneration, stem cell maintenance, metabolism, and immune homeostasis.⁸⁰ Dysregulation of this pathway is a hallmark of many cancers, contributing to tumor progression, metastasis, and therapeutic resistance. Wnt signaling operates through canonical (Wnt/β-catenin) and noncanonical branches (e.g., Wnt/PCP and Wnt/Ca²⁺ pathways).⁶⁰

In the canonical pathway, Wnt ligands bind Frizzled (FZD) receptors and LRP5/6 coreceptors, activating Dishevelled (DVL) and inhibiting the β-catenin destruction complex (Axin/APC/GSK-3β).¹²⁹ Stabilized β-catenin translocates to the nucleus and partners with TCF/LEF transcription factors to activate oncogenic targets, including MYC, Cyclin D1, LGR5, CD44, and ABCB1, genes that regulate proliferation, drug efflux, and stemness.¹²⁹ Wnt/β-catenin signaling is a key driver of EMT and CSC maintenance.¹²⁹ In melanoma, the transmembrane glycoprotein podoplanin (PDPN) activates β-catenin signaling via CLEC-2–mediated platelet activation.²³⁹ This cascade promotes the secretion of immunosuppressive cytokines such as TGF-β, recruitment of T_regs and M2 macrophages, and suppression of CD8⁺ T and NK cell cytotoxicity, ultimately facilitating metastasis and immune resistance.^80,239

Cross-talk with other oncogenic pathways further augments resistance.²⁴⁰ In glioblastoma, hepatocyte growth factor (HGF)/c-Met signaling enhances β-catenin nuclear localization and LEF1-dependent transcription of MDR–associated genes such as MRP-1, inducing CSC traits, endothelial–mesenchymal transition (EndMT), and temozolomide resistance.²⁴¹ In lenvatinib-resistant HCC, CDK6-mediated phosphorylation of GSK3β at Ser9 stabilizes β-catenin and promotes its nuclear translocation, reinforcing CSC phenotypes and immune evasion.⁷⁹ Similarly, in pancreatic cancer, β-catenin/TCF4/Pygo2 transcriptionally activates the lncRNA PVT1, which sponges miR-619-5p to derepress Pygo2 and ATG14. This positive feedback loop promotes Wnt target gene expression, autophagy activation, drug efflux, and gemcitabine resistance.²⁴²

JAK/STAT signaling pathway

The Janus kinase–signal transducer and activator of transcription (JAK/STAT) pathway mediates cellular responses to cytokines and growth factors, governing proliferation, apoptosis, differentiation, inflammation, and immune regulation.²⁴³ Upon ligand engagement, receptor dimerization triggers JAK autophosphorylation, followed by STAT recruitment and activation.⁸⁸ Phosphorylated STATs dimerize and translocate into the nucleus to regulate target gene expression.²⁴³

Persistent activation of the JAK/STAT axis is a hallmark of therapy resistance in multiple cancers.²⁴⁴ In high-grade serous ovarian cancer (HGSOC), single-cell transcriptomics reveals sustained JAK/STAT activity in both malignant epithelial cells and CAFs, driven by ascitic IL-6 and CAF-secreted cytokines, promoting platinum resistance via paracrine and autocrine loops.²⁴⁵ JAK2/STAT3 signaling also plays a pivotal role in CD8⁺ T-cell exhaustion.⁸⁸ STAT3 forms complexes with TOX, directly binding promoters of inhibitory receptors such as PD-1, TIM-3, and LAG-3, increasing chromatin accessibility and enforcing an exhausted phenotype.⁸⁸ This signaling axis simultaneously represses TCR signaling and interferon responses, dampening cytotoxic T-cell function.⁸⁸ Tumor-intrinsic PD-L1 amplifies JAK2/STAT3 signaling by sequestering PTP1B, a phosphatase that normally deactivates JAK2. Nuclear PD-L1 further complexes with pSTAT3, transcriptionally upregulating IL-6 and CXCL1 and enhancing MDSC recruitment and resistance to ICBs.¹³⁷ In metastatic CRPC, coloss of TP53/RB1 and overexpression of SOX2 activate JAK1/STAT1, promoting lineage plasticity and transition to a stem-like, multilineage-resistant state. This transition is reinforced by a SOX2–JAK–STAT feedback loop, which sustains chromatin accessibility and resistance to AR-targeted therapies such as enzalutamide.²⁴⁶

Other contexts include IL-11–mediated activation of JAK1/STAT4 in prostate cancer, where nuclear pSTAT4 enhances c-MYC transcription, conferring docetaxel resistance.²⁴⁷ In gastric cancer, defects in IFNγ-driven JAK/STAT signaling disrupt PD-1 expression, impairing responsiveness to PD-1 inhibitors.²⁴⁸ In CML, mitochondrial and nuclear forms of STAT3 mediate TKI resistance through metabolic rewiring.²⁴⁹ Nuclear STAT3-Y705 activates glycolysis and fatty acid oxidation (FAO) genes, while mitochondrial STAT3-S727 and acetylated STAT3-K685 impair oxidative phosphorylation by destabilizing electron transport chain (ETC) complexes, supporting leukemia stem cell dormancy and persistence.²⁴⁹ Furthermore, 5-FU/cisplatin–induced interferon responses upregulate ADAR1 via JAK/STAT, promoting A-to-I RNA editing that stabilizes SCD1 mRNA, reprograms lipid metabolism, activates Wnt/β-catenin signaling, and sustains stemness and chemoresistance in gastric cancer.²⁵⁰

Notch signaling pathway

The Notch signaling pathway is a highly conserved cell–cell communication system essential for cell fate determination, differentiation, and tissue homeostasis.²⁵¹ In cancer, Notch signaling exhibits context-dependent and often paradoxical roles, contributing to tumor progression, cancer stemness, immune evasion, and therapy resistance. The pathway operates through both canonical (RBP-Jκ/CBF-1–dependent) and noncanonical mechanisms.^251,252 In the canonical cascade, ligand binding (e.g., Jagged1/2, DLL1/3/4) to Notch receptors (Notch1–4) induces a two-step proteolytic cleavage, culminating in the release of the Notch intracellular domain (NICD), a cleaved cytoplasmic signaling domain that mediates intracellular signal transduction, by γ-secretase.²⁵² The NICD then translocates into the nucleus, where it forms a transcriptional complex with RBP-Jκ and Mastermind-like (MAML) coactivators, activating downstream targets such as Hes, Hey, c-Myc, and Bcl-2.²⁵² Noncanonical Notch signaling, in contrast, can function independently of ligand stimulation or involve nontraditional intracellular mediators.²⁵²

Aberrant Notch activation drives multiple resistance-related processes, including EMT, angiogenesis, metabolic rewiring, and maintenance of cancer stem-like cells.^251,252 In pancreatic cancer, for example, the E3 ubiquitin ligase TRIM59 stabilizes RBP-Jκ via site-specific ubiquitination. RBP-Jκ then transcriptionally enhances TRIM59 expression, forming a feedforward loop that drives gemcitabine resistance.²⁵³ In gastric cancer, 5-FU resistance is mediated by PRMT1-dependent arginine methylation of NUSAP1, promoting its interaction with the PEST domain of Notch2. This impairs Notch2 degradation and sustains oncogenic Notch2 signaling, which upregulates c-Myc and Cyclin D3 expression.²⁵⁴ This study established cell line–derived xenograft (CDX) models to evaluate the effects of NUSAP1 and related gene regulation on tumor growth. Mechanical cues also activate Notch signaling in drug-resistant tumors. Actomyosin tension transmitted via E-cadherin–α-catenin complexes facilitates Notch receptor cleavage, leading to NICD translocation and induction of MVP, a mediator of multidrug efflux.²⁵⁵ Conversely, in gliomas, suppression of the Notch1/2–RBP-Jκ axis impairs interferon-γ (IFN-γ) signaling and derepresses oncogenes such as MYC. This Notch downregulation reduces MHC-I expression and chemokine secretion, hindering T-cell infiltration and promoting immune escape.²⁵⁶ The concurrent expansion of glioma stem cells and remodeling of the TME further dampens the efficacy of IFN-γ–based therapies.²⁵⁶ Collectively, the Notch pathway undergoes dynamic rewiring in response to therapeutic pressure, reinforcing its role in resistance biology and validating its potential as a target for rational combinatorial interventions.

Hippo signaling pathway

The Hippo pathway is a conserved regulatory cascade that controls tissue size, cell proliferation, apoptosis, stemness, and regeneration by modulating the subcellular localization of Yes-associated protein (YAP) and transcriptional coactivator with PDZ-binding motif (TAZ).^257,258 The Hippo pathway also modulates apoptosis, as MST1/2 signaling can activate proapoptotic programs, whereas oncogenic RAS, PI3K/AKT, and RAF signaling suppress MST2 activity and promote cancer cell survival.²⁵⁸ In addition, ATR–RASSF1A-mediated mechanotransduction and YAP1–STAT1-dependent inflammatory regulation further support the concept that Hippo signaling integrates mechanical stress, immune remodeling, and apoptosis-related cell fate decisions during therapeutic resistance.^259,260 The core kinases MST1/2 and LATS1/2 phosphorylate YAP/TAZ, sequestering them in the cytoplasm or promoting their degradation. Inactivation of Hippo signaling leads to YAP/TAZ nuclear translocation, where they interact with TEAD family transcription factors to drive transcriptional programs linked to tumor growth and resistance.²⁵⁷

Persistent YAP/TAZ activation is frequently observed in cancers such as HCC, breast cancer, and NSCLC.²⁶¹ It promotes EMT, enhances DNA repair, maintains CSC phenotypes, and upregulates ABC transporters, collectively contributing to drug resistance.²⁵⁷ YAP/TAZ also serve as central integrators of upstream oncogenic signals (e.g., Wnt, TGF-β, and Notch) and mechanical cues from the TME, solidifying their role as resistance nodes.²⁶² In pancreatic ductal adenocarcinoma (PDAC), the deubiquitinase VCPIP1 prevents K48-linked polyubiquitination of YAP, thereby stabilizing its nuclear pool. Nuclear YAP transcriptionally activates VCPIP1, forming a positive feedback loop that reinforces the expression of prosurvival and multidrug resistance genes.²⁶³ Similarly, in HCC, the oncofetal protein CLDN6 competes with TJP2 for YAP1 binding via its PDZ domain, displacing YAP1 from tight junctions and facilitating its nuclear localization. Activated YAP1 induces a cholangiocyte-like lineage program characterized by dedifferentiation and resistance.²⁶⁴ In solid tumors, proteasome inhibitors can induce YAP/TAZ activation by inactivating the Hippo pathway, thereby promoting tumor cell proliferation and resistance to apoptosis and ultimately driving therapeutic resistance.²⁶⁵ YAP/TAZ also regulate immune evasion. In melanoma and breast cancers, they induce PD-L1 expression, attenuating T-cell-mediated cytotoxicity.²⁶⁶ These findings highlight the Hippo–YAP/TAZ axis as a pivotal regulator of therapeutic failure and a promising target for combined treatment strategies.

GAS6/AXL signaling pathway

The GAS6/AXL axis, a key member of the TAM (TYRO3, AXL, MERTK) receptor tyrosine kinase family, is frequently overactivated in solid tumors and is strongly associated with therapy resistance.²⁶⁷ GAS6, secreted by stromal cells such as CAFs and macrophages, binds to AXL, promoting its dimerization, autophosphorylation, and activation of downstream survival pathways, including the PI3K/AKT, MAPK/ERK, and NF-κB pathways. These cascades enhance tumor cell proliferation, migration, EMT, and stemness maintenance.²⁶⁸

AXL activation mediates both intrinsic and acquired resistance to targeted therapies (e.g., EGFR-TKIs in NSCLC, HER2 inhibitors in breast cancer), chemotherapy (e.g., cisplatin, gemcitabine), and antiangiogenic agents (e.g., bevacizumab).²⁶⁹ In EGFR-mutant NSCLC, AXL activation by GAS6 stabilizes the E3 ligase RAD18 via ubiquitin-like modification, promoting monoubiquitination of PCNA and facilitating error-prone translesion DNA synthesis, thus accelerating mutagenesis and resistance.²⁷⁰ AXL also drives MYC-dependent purine biosynthesis, creating nucleotide imbalance and genomic instability, further contributing to resistance.²⁷⁰ Additionally, AXL promotes immune evasion by upregulating PD-L1 and reducing T-cell cytotoxicity.²⁷¹ In breast cancer, ligand-induced activation of the AXL receptor tyrosine kinase initiates downstream signaling, leading to the phosphorylation and activation of AKT.²⁵² Activated AKT in turn phosphorylates and inactivates GSK3β. This cascade prevents β-catenin from being targeted for degradation, thereby stabilizing it and promoting its nuclear translocation. Nuclear β-catenin upregulates ZEB1, a key EMT transcription factor, which enhances invasiveness and suppresses DNA repair, thereby reducing chemosensitivity to agents such as doxorubicin.²⁷² AXL overexpression is a negative prognostic marker in several cancers.²⁷³ Pharmacological inhibitors targeting AXL (e.g., bemcentinib and gilteritinib) have shown promise in reversing resistance and enhancing treatment efficacy.^273,274 The GAS6/AXL pathway represents a critical driver of tumor progression and resistance and a valuable therapeutic target in combination strategies.

Aberrant activation of signaling pathways and therapeutic implications

Taken together, the aberrant activation of signaling pathways such as PI3K-AKT-mTOR, MAPK, NF-κB, TGF-β, Wnt/β-catenin, JAK/STAT, Notch, Hippo-YAP/TAZ, and GAS6/AXL contributes extensively to therapy resistance across cancer types (Fig. 2c). These networks orchestrate a wide array of resistance-enabling processes, including survival signaling, cell fate plasticity, immune evasion, metabolic rewiring, DNA repair, and maintenance of cancer stemness. Importantly, these pathways often engage in intricate crosstalk, exhibit functional redundancy, and display compensatory activation upon therapeutic inhibition, making monotherapies prone to failure.

This phenomenon underscores a critical, systems-level resistance mechanism: signaling network adaptation.²⁷⁵ Rather than functioning as isolated linear cascades, oncogenic pathways form dynamic, self-stabilizing systems capable of re-establishing essential survival outputs under therapeutic pressure. However, feedback loops alone are insufficient to fully restore signaling output. Kholodenko et al. demonstrated that complete signal recovery requires either a two‑pathway network architecture or drug‑induced target dimerization.²⁷⁶ Lauffenburger’s early modeling revealed how negative feedback quantitatively controls MAPK signal adaptation, and his concept of temporal collateral sensitivity shows that tumor evolution creates transient windows of drug susceptibility, supporting adaptive therapy.^277,278 Complementing these cell‑intrinsic models, Farrell extended network adaptation to the tumor microenvironment, arguing that resistance is ecologically driven by crosstalk among cancer cells, CAFs, and macrophages via paracrine signals and exosomes, enabling non‑cell‑autonomous drug tolerance.²⁷⁹

In parallel, many tumors exploit endoplasmic reticulum stress adaptation as a complementary systems-level mechanism.²⁸⁰ The accumulation of unfolded or misfolded proteins triggers the unfolded protein response (UPR), an adaptive signaling network coordinated by the endoplasmic reticulum sensors IRE1α, PERK, and ATF6, which functions to restore proteostasis and maintain cellular homeostasis.²⁸¹ In cancer cells, sustained or rewired UPR activation promotes survival under chemotherapy and targeted therapy by enhancing chaperone expression, attenuating global protein translation, limiting oxidative stress, and suppressing apoptosis through the ATF4–CHOP axis.²⁸¹ Importantly, while acute or overwhelming endoplasmic reticulum stress can induce cell death, chronic UPR activation favors drug tolerance and persistence, positioning endoplasmic reticulum stress adaptation as a systems-level resistance mechanism complementary to oncogenic signaling network adaptation.

Beyond discrete pathway alterations, tumor cells dynamically rewire their intracellular signaling networks through feedback loops and pathway crosstalk to maintain essential survival outputs under therapeutic pressure.²⁸² Experimentally, this is evidenced by rapid feedback reactivation and state switching.²⁸³ From a theoretical perspective, systems biology provides a quantitative framework to interpret these emergent behaviors. Computational approaches, including kinetic modeling and Boolean network simulations, demonstrate how robustness is intrinsically encoded within oncogenic signaling architectures, enabling networks to buffer perturbations and maintain functional stability.^284,285 Importantly, these models can delineate minimal intervention sets—combinatorial target nodes whose simultaneous inhibition is predicted to destabilize the network attractor state, thereby precluding adaptive escape and restoring therapeutic vulnerability.²⁸⁶ Kholodenko’s Modular Response Analysis (MRA) and Boolean network simulations can identify minimal intervention sets—combinatorial targets whose simultaneous inhibition destabilizes the adaptive network, precluding escape and restoring vulnerability.²⁸⁷ This theoretical perspective shifts the focus from static oncogenic drivers to the dynamic stability and plasticity of the signaling network as a whole, explaining why monotherapies often fail and providing a rationale for combinatorial strategies that target network vulnerabilities.

Beyond the core pathways discussed above, several additional signaling axes have been increasingly implicated in resistance phenotypes. These include the Hedgehog–GLI pathway, which sustains tumor stemness and contributes to chemotherapy resistance;²⁸⁸ the NRF2–KEAP1 antioxidant response pathway, which reprograms redox metabolism and drives multidrug resistance;²⁸⁹ and RTKs such as FGFR and MET, whose overactivation confers resistance to targeted therapies and immunotherapies. This growing complexity highlights the need for a comprehensive understanding of pathway interconnectivity and feedback dynamics in the resistant tumor state.

Therapeutically, this complexity necessitates the deployment of multitargeted and adaptive strategies with clinical trials (Table 2). These include (i) rational drug combinations targeting parallel or converging signaling cascades, (ii) the use of allosteric inhibitors or degraders that disrupt key signaling hubs, (iii) vertical inhibition along a single pathway to prevent upstream feedback reactivation, and (iv) the application of systems biology and network modeling to predict synthetic lethal vulnerabilities. Furthermore, the integration of temporal and spatial omics profiling may enable dynamic monitoring of signaling rewiring, informing the design of context specific, resistance-preemptive regimens.^12,13

Table 2.

Summary of clinical trials targeting signaling pathways

Study Identifier	Phase	Study type	Intervention/experimental agent	Targeted pathway	Cancer condition(s)	Study Status
NCT06530550	Ⅱ	Interventional	PI3K Inhibitors	PI3K-AKT and mTOR	Relapsed/refractory indolent T/NK-cell lymphomas	Recruiting
NCT02437318	Ⅲ	Interventional	Alpelisib		Advanced breast cancer	Completed
NCT03065062	Ⅰ	Interventional	Gedatolisib		Advanced squamous cell lung, pancreatic, head & neck and other solid tumors	Recruiting
NCT04305496	Ⅲ	Interventional	Capivasertib		Locally advanced (inoperable) or metastatic HR+/HER2- breast cancer	Active, not recruiting
NCT02014116	Ⅰ	Interventional	LY3009120	MAPK	Advanced and metastatic cancer	Terminated
NCT03454035	Ⅰ	Interventional	Ulixertinib/ Palbociclib		Advanced pancreatic and other solid tumors	Recruiting
NCT04330664	Ⅰ	Interventional	Adagrasib plus TNO155		Advanced cancer, metastatic cancer and malignant neoplastic disease	Terminated
NCT03900598	Ⅰ	Interventional	JNJ-67856633	NF-κB	Non-hodgkin’s lymphoma and chronic lymphocytic leukemia	Completed
NCT04876092	Ⅰ	Interventional	JNJ-67856633 and Ibrutinib		Non-hodgkin’s lymphoma and chronic lymphocytic leukemia	Completed
NCT04673942	Ⅱ	Interventional	AdAPT-001	TGF-β	Sarcoma and refractory solid tumors	Recruiting
NCT03732274	Ⅰb/Ⅱa	Interventional	Vactosertib		Advanced NSCLC	Completed
NCT03834662	Ⅰ	Interventional	AVID200		Malignant solid tumor	Completed
NCT02278133	Ⅰb/Ⅱ	Interventional	WNT974 + Encorafenib + Cetuximab	Wnt	BRAF-mutant mCRC and Wnt Pathway Mutations	Completed
NCT01973309	Ⅰ	Interventional	Vantictumab + Paclitaxel		Locally recurrent or metastatic breast cancer	Completed
NCT06519526	Ⅱ	Interventional	SHR-0302 and SHR-2554	JAK/STAT	Relapsed/refractory peripheral T cell lymphoma	Recruiting
NCT00674479	Ⅱ	Interventional	INCB018424		Advanced hematological malignancies	Completed
NCT01712659	Ⅰ/Ⅱ	Interventional	Ruxolitinib		Adult T-Cell leukemia	Terminated
NCT00106145	Ⅰ	Interventional	MK-0752	Notch	Advanced breast cancer	Completed
NCT03422679	Ⅰ/Ⅱ	Interventional	CB-103		Advanced or metastatic solid tumors and haematological malignancies	Terminated
NCT01189968	Ⅰ	Interventional	Carboplatin and Pemetrexed plus Demcizumab		Non-squamous NSCLC	Completed
NCT05228015	Ⅰ	Interventional	IK-930	Hippo	Advanced solid tumors	Terminated
NCT04665206	Ⅰ/Ⅱ	Interventional	VT3989		Metastatic solid tumors	Recruiting
NCT04857372	Ⅰ	Interventional	IAG933		Advanced mesothelioma and other solid tumors	Recruiting
NCT02424617	Ⅰ/Ⅱ	Interventional	Bemcentinib in combination with Erlotinib	GAS6/AXL	NSCLC	Completed
NCT03649321	Ⅰb/Ⅱ	Interventional	Chemotherapy and Bemcentinib		Metastatic pancreatic cancer	Terminated
NCT02988817	Ⅰ/Ⅱ	Interventional	Enapotamab Vedotin		Solid tumors	Completed

DLBCL diffuse large B cell lymphoma, NSCLC Non-small cell lung cancer, mCRC metastatic colorectal cancer

Multiple clinical trials have been launched to therapeutically target key oncogenic signaling pathways in drug-resistant cancers, underscoring the translational relevance of this axis (Table 2). For instance, a number of ongoing or completed trials are evaluating inhibitors of the PI3K-AKT-mTOR cascade, such as gedatolisib (NCT03065062) in advanced solid tumors, capivasertib (NCT04305496) in hormone receptor-positive breast cancer, and alpelisib (NCT02437318), which has shown efficacy in advanced breast cancer and is already approved for PIK3CA-mutated cases. Similarly, dual inhibition strategies targeting both PI3K and mTOR are being explored in relapsed/refractory T/NK-cell lymphomas (NCT06530550). In the MAPK pathway, clinical evaluation includes agents such as LY3009120 (NCT02014116) and ulixertinib in combination with palbociclib (NCT03454035) for patients with advanced pancreatic and other refractory solid tumors. Meanwhile, targeted inhibition of the TGF-β signaling axis is under active investigation with agents such as AdAPT-001 (NCT04673942) and vactosertib (NCT03732274) in sarcomas and NSCLC. Wnt pathway inhibitors, including WNT974 (NCT02278133) and vantictumab (NCT01973309), have been evaluated in BRAF-mutant colorectal cancer and metastatic breast cancer, respectively. Additional ongoing studies target JAK/STAT signaling with SHR-0302 and SHR-2554 (NCT06519526), as well as agents modulating the Notch (e.g., MK-0752, NCT00106145), Hippo (e.g., IK-930, NCT05228015), and GAS6/AXL (e.g., bemcentinib, NCT02424617) pathways in a variety of advanced or resistant tumors. These trials collectively demonstrate the increasing clinical momentum in disrupting compensatory or aberrant pathway activity and represent a critical component of precision strategies to surmount resistance in treatment-refractory malignancies.

Epigenetic reprogramming in cancer drug resistance and combating strategies

Phenotypic and epigenetic reprogramming

Epigenetic regulation, which governs heritable yet reversible changes in gene expression without altering the underlying DNA sequence, has emerged as a pivotal driver of therapeutic resistance across cancer types (Fig. 3a).¹ Major epigenetic mechanisms include DNA methylation, histone modifications (e.g., methylation, acetylation, lactylation), chromatin remodeling, and noncoding RNA-mediated regulation.³⁰ Together, these processes reshape transcriptional landscapes, sustain CSC plasticity, rewire metabolic networks, and remodel the immune microenvironment, thereby enabling tumor cells to evade therapeutic pressure through multifactorial routes.⁷⁵

Fig. 3.

Molecular mechanisms and combating strategies in cancer drug resistance: cellular plasticity and the tumor microenvironment. a Epigenetic reprogramming and transcriptional adaptation that contribute to resistance and strategies targeting chromatin modifiers and transcriptional networks. b Phenotypic plasticity, including epithelial-mesenchymal transition (EMT), stemness acquisition, and dedifferentiation, and approaches to target plastic cell states. c Conceptual framework of drug-tolerant persister (DTP) states, depicting maintenance programs (epigenetic remodeling, stress signaling, metabolic rewiring), exit trajectories (resensitization versus mutation-driven stable resistance), and potential intervention windows to disrupt state stabilization or redirect evolutionary outcomes. d Tumor microenvironment (TME)-mediated resistance, involving stromal cells, immune modulation, and extracellular matrix remodeling, and corresponding therapeutic strategies

Epigenetic dysregulation influences the expression of drug transporters, metabolic enzymes, and apoptotic regulators, thereby altering drug uptake, biotransformation, and clearance.⁷⁵ In breast and ovarian cancers, for instance, hypomethylation of the ABCB1 (MDR1) promoter correlates with its overexpression, leading to elevated P-gp levels and enhanced efflux of chemotherapeutic agents, culminating in MDR.¹⁸⁰ Similar outcomes are achieved through histone acetylation or miRNA-mediated upregulation of ABC transporters.²⁹⁰ Additionally, histone methyltransferases such as G9a epigenetically silence CYP genes, impairing the bioactivation of prodrugs and reducing treatment efficacy.²⁹¹ Cancer cells frequently leverage epigenetic mechanisms to suppress drug target genes or activate compensatory signaling pathways, thereby bypassing pharmacological blockade.⁷⁵ For example, in prostate cancer, loss of the histone methyltransferase KMT2C represses ASPP2 via impaired enhancer activity.¹⁰⁶ This derepresses ΔNp63 expression, promoting lineage plasticity and the emergence of double-negative prostate cancer (DNPC) under androgen deprivation therapy. KMT2C deficiency simultaneously enhances fatty acid biosynthesis and HRAS palmitoylation, activating the MAPK/ERK pathway and stabilizing resistance phenotypes.¹⁰⁶ In PDAC, downregulation of miR-146a-5p leads to TRAF6 upregulation, activation of NF-κB signaling, and increased ABCB1 transcription by NF-κB p65. This drives gemcitabine efflux and chemoresistance.²⁹² In HCC, intracellular lactate triggers histone H3K18 lactylation, promoting HECTD2 transcription. As an E3 ligase, HECTD2 degrades KEAP1, thereby activating NRF2 and upregulating antioxidant defense programs to confer lenvatinib resistance.²⁹³ Multiple miRNAs modulate cisplatin resistance in bladder cancer by targeting critical pathways such as AKT/mTOR, JAK/STAT, Wnt/β-catenin, and p53, thereby regulating proliferation, apoptosis, autophagy, and glutathione metabolism.^294,295 Under treatment-induced stress, tumor cells activate survival responses, including antiapoptotic signaling, autophagy, and ferroptosis inhibition.²⁹⁶ Emerging evidence implicates other noncoding RNAs, including long noncoding RNAs (lncRNAs) and circular RNAs (circRNAs), in shaping platinum resistance by modulating RNA‒protein interactions, translational control, and downstream oncogenic signaling pathways.²⁹⁷ In particular, extracellular vesicle-mediated transfer of lncRNAs has been shown to activate MAPK signaling and reinforce chemoresistant states in HGSOC.

Epigenetic reprogramming also modulates DNA damage response pathways. For instance, platinum and radiation therapies induce DNA lesions, which are often counteracted by chromatin remodeling that enhances homologous recombination and nonhomologous end joining repair.¹⁸⁵ In EGFR-mutant lung cancer, resistance to osimertinib is tightly linked to epigenetic regulation by the ATPase subunit SMARCA4 of the mSWI/SNF complex. SMARCA4 maintains chromatin accessibility and upregulates stress response genes, facilitating survival and resistance.²⁹⁸

Within the TME, metabolic-epigenetic crosstalk sustains phenotypic heterogeneity and immune evasion.²⁹⁹ Lactate secreted by differentiated cancer cells (CDCs) is metabolized by CSCs into acetyl-CoA, promoting histone H3K27 acetylation, MYC activation, and CSC maintenance. This reciprocal reprogramming of CDCs into CSCs supports continuous intratumoural plasticity and therapeutic resistance.³⁰⁰ Furthermore, loss of MLL3/MLL4 in the TME silences enhancers and components of the RISC complex, inducing endogenous dsRNA stress and GSDMD-mediated pyroptosis, which can be harnessed to boost CD8⁺ T-cell–mediated responses.³⁰¹ Immune resistance is also epigenetically orchestrated through exosomal communication and metabolic rewiring. For example, exosomal circUSP7 from NSCLC cells inhibits miR-934 in CD8⁺ T cells, upregulates SHP2, suppresses effector cytokine secretion (e.g., TNF-α, IFN-γ), and induces T-cell exhaustion, thereby impairing anti-PD-1 efficacy.⁸³ Concurrently, drug-resistant tumor cells upregulate Fabp7 and suppress Lpcat3 via altered H3K27ac/H3K9ac marks, leading to ferroptosis resistance, enhanced FAO, and disrupted circadian signaling in T cells, all of which contribute to immune evasion and therapy failure.³⁰²

Combating strategies

The reversibility and plasticity of epigenetic modifications offer a compelling therapeutic window to counteract drug resistance (Fig. 3a). A variety of strategies have been developed to exploit these vulnerabilities. Epigenetic modulators, such as DNA methyltransferase inhibitors (DNMTis; e.g., decitabine, azacitidine), histone deacetylase inhibitors (HDACis; e.g., vorinostat, romidepsin), and bromodomain and extraterminal domain (BET) inhibitors (e.g., JQ1), can reverse the silencing of tumor suppressor genes, suppress oncogenic transcriptional programs, and resensitize tumors to chemotherapy, targeted agents, or immunotherapy.³⁰³ These compounds also reduce CSC populations and remodel the TME to enhance immune cell infiltration and activity.¹⁸³ In addition, rational combination regimens, such as HDACis with EGFR-TKIs, DNMTis with platinum agents, or BET inhibitors with anti–PD-1/PD-L1 antibodies, have demonstrated synergistic efficacy by simultaneously targeting multiple resistance pathways. For instance, BET inhibition downregulates PD-L1 and reverses T-cell exhaustion, thereby augmenting the immunotherapeutic response.³⁰³

Beyond pharmacological epigenetic inhibition, other innovative approaches have emerged. CRISPR-based functional genomic screens have uncovered synthetic lethal interactions involving epigenetic regulators such as KDM5A, EZH2, and SMARCA2 in resistant tumor cells, which can be therapeutically exploited with selective inhibitors.³⁰⁴ Moreover, targeting the metabolic and epigenetic axes by modulating cofactors such as acetyl-CoA, S-adenosylmethionine (SAM), or α-ketoglutarate (α-KG) offers a means to indirectly reprogram aberrant chromatin states.³⁰³ Inhibiting lactate production or FAO can disrupt histone acetylation patterns that sustain resistance.³⁰⁵ Advances in single-cell and spatial epigenomics now enable precise mapping of epigenetic heterogeneity across tumor ecosystems, informing adaptive treatment designs.^12,13 Furthermore, disrupting TME-derived epigenetic cues, including CAF-derived exosomes and lactate-induced histone lactylation, may help dismantle pro-resistance niches and restore antitumor immunity.³⁰⁶ Collectively, these multifaceted strategies reflect a paradigm shift in understanding epigenetic plasticity, not only as a mechanism of resistance but also as a therapeutic opportunity. Integrating epigenetic-targeted interventions into personalized regimens holds considerable promise for overcoming resistance and improving long-term clinical outcomes.

Several epigenetic agents targeting these pathways have progressed into clinical development. DNMTis, including azacitidine and decitabine, are approved for myeloid malignancies and are being actively evaluated in combination with chemotherapy and ICBs in solid tumors to reverse epigenetically mediated drug tolerance.³⁰³ HDACis such as vorinostat and entinostat have entered phase I–III clinical trials across multiple cancer types, where they are used as epigenetic priming agents to enhance tumor immunogenicity and resensitize tumors to targeted and immune-based therapies.³⁰³ In parallel, EZH2 inhibitors (EZH2is), exemplified by tazemetostat, have shown clinical activity in biomarker-defined populations, particularly in lymphomas and selected solid tumors, and are being tested in combination regimens to overcome lineage plasticity and resistance associated with chromatin reprogramming.^304,307

Cell state plasticity in cancer resistance and combating strategies

Cell state plasticity in cancer resistance

TCP endows cancer cells with the capacity to adapt dynamically to therapeutic pressures, evade treatment, and re-emerge after drug withdrawal, ultimately contributing to resistance and relapse (Fig. 3b).³⁰⁸ One prominent example of TCP is the ability of tumor cells to reversibly transition between proliferative and quiescent states, thereby avoiding the cytotoxic effects of chemotherapy or targeted therapies.³⁰⁹ Upon exposure to these treatments, a subset of cells enters a DTP state characterized by stem-like features, low proliferation, and heightened survival capacity.^132,309 DTP cells represent a rare subpopulation of tumor cells that survive initial anticancer therapy through reversible, nongenetic adaptive programs, often adopting a slow-cycling or quiescent state that enables transient drug tolerance and seeds subsequent resistance.¹³²

DTP cells constitute a transient, nongenetic adaptive state that enables a minor tumor cell subpopulation to survive otherwise lethal anticancer therapies and acts as a critical reservoir for resistance evolution within a dynamic balance between state maintenance and exit trajectories (Fig. 3c).³⁰⁸ Rather than harboring stable resistance mutations, DTP cells adopt a slow-cycling or quasiquiescent phenotype driven by extensive epigenetic remodeling, transcriptional plasticity and metabolic reprogramming, which collectively constitute core maintenance programs that stabilize the DTP state under therapeutic pressure.³¹⁰ Key mechanisms supporting DTP survival include chromatin reconfiguration mediated by histone-modifying enzymes, activation of stress-response and bypass signaling pathways such as IGF-1R, AXL and YAP/TAZ, and a shift toward oxidative or alternative nutrient metabolism, often coupled with enhanced mitochondrial ETC dependency, selective mitophagy and resistance to lipid peroxidation, which together mitigate therapy-induced apoptosis.³⁰⁸ Although this tolerant state is initially reversible upon drug withdrawal, representing a resensitization exit route, prolonged persistence under treatment pressure increases the probability of acquiring irreversible genetic alterations, driving an alternative exit toward stable, mutation-driven resistance and tumor relapse. DTP cells survive under chemotherapeutic pressure by actively engaging nongenetic adaptive mechanisms, such as CYP3A-mediated drug detoxification, thereby driving adaptive resistance and serving as a key cellular source of treatment failure and tumor recurrence in PDAC.³¹¹ Collectively, DTP cells represent an early, targetable node in resistance evolution and minimal residual disease, providing multiple intervention windows to disrupt state maintenance or bias exit trajectories toward therapeutic vulnerability.

A canonical manifestation of TCP is EMT, wherein epithelial tumor cells lose cell polarity and adhesion while gaining mesenchymal traits, including motility and resistance to apoptosis.³⁰⁹ EMT enhances DNA repair, alters drug metabolism, and facilitates immune evasion.⁵ For example, EMT-associated tumor cells upregulate RHOJ, a small GTPase that promotes DDR through nuclear actin polymerization, thus contributing to resistance against genotoxic agents.⁵⁴ EMT is frequently detected in therapy-resistant tumors across cancer types.⁵ In NSCLC, resistance is often driven by HGF/c-MET–induced EMT, which activates PI3K/AKT/mTOR signaling to enhance stem-like properties and immune escape.³¹² Similarly, EMT contributes to oxaliplatin and cisplatin resistance in gastric and colorectal cancers, mediated by factors such as Rab31 (via the Stat3/MUC-1/Twist1 axis) and THBS2⁺ CAFs (via COL8A1-driven PI3K/AKT signaling).^313,314 Notably, many tumors adopt a partial or hybrid EMT phenotype, retaining epithelial markers while acquiring mesenchymal traits, thereby enhancing adaptability, metastatic potential, and multidrug resistance.⁵ For instance, Rab31 promotes EMT and cisplatin resistance in stomach adenocarcinoma via the Stat3/MUC-1/Twist1 axis.³¹⁴ Likewise, THBS2⁺ CAFs mediate oxaliplatin resistance in colorectal cancer by activating COL8A1-driven PI3K/AKT signaling and EMT.³¹³ In NSCLC, hybrid EMT phenotypes in tumor-initiating cells correlate with immune escape via suppression of chemokines and increased expression of B7-H3, which resists NK cell-mediated killing.³¹⁵ These microenvironment-derived signals act primarily by reinforcing tumor cell intrinsic stemness, plasticity, and survival programs.

Lineage plasticity is another critical form of TCP, allowing tumor cells to undergo transdifferentiation under therapeutic stress.³¹⁶ In KRAS/LKB1-mutant NSCLC, resistance to KRAS inhibitors such as adagrasib involves adeno-to-squamous transdifferentiation (AST), driven by ΔNp63 and characterized by KRT6A expression.¹⁰⁷ In prostate cancer, anti-androgen therapies such as enzalutamide induce the histone reader ZMYND8, which cooperates with FOXM1 to activate neuroendocrine transcriptional programs (e.g., ASCL1), facilitating transdifferentiation into neuroendocrine prostate cancer (NEPC).¹⁰⁵ Similarly, loss of KMT2C promotes a transition to DNPC, further contributing to AR therapy resistance.¹⁰⁶

CSCs, known for their self-renewal and multipotent capabilities, are central players in intratumoral heterogeneity and treatment resistance.³¹⁷ CSCs may arise through dedifferentiation of non-stem-like tumor cells in response to therapeutic stress.³⁰⁹ They often express high levels of ABC transporters (e.g., ABCG2), ALDH activity, and robust DNA repair machinery, making them inherently resistant to chemotherapy and radiotherapy.^317,318 Additionally, resistant CSCs can secrete small extracellular vesicles (sEVs) carrying phosphorylated PKM2 (pY105-PKM2), which reprogram recipient cells to acquire stem-like traits and chemoresistance.⁵⁵ In esophageal squamous cell carcinoma (ESCC), QSOX2 promotes disulfide bond formation in TSC2, enhancing its phosphorylation by Akt and activating the mTOR/4E-BP1/c-Myc axis, thereby upregulating CSC markers (e.g., CD44, Notch1) and promoting platinum resistance.³¹⁹ Furthermore, CAF-derived IGF-1 activates the IGF1R/Akt/mTOR/c-Myc pathway to upregulate QSOX2, reinforcing CSC phenotypes and therapy resistance.³¹⁹

DTP cells are a transient, nongenetically distinct subpopulation of cancer cells that evade therapy by entering a quiescent or slow-cycling state.³⁰⁸ Their persistence relies on epigenetic remodeling, metabolic reprogramming, and altered signaling rather than stable genetic mutations.³⁰⁸ In EGFR-TKI–treated NSCLC, DTPs upregulate DPP4, activating the DPP4–CPT1A axis to increase FAO and mitochondrial respiration. Concurrently, NRF2 is engaged to mitigate oxidative stress and control cell cycle progression, enabling minimal residual disease and eventual relapse.³²⁰ Pharmacologic inhibition of DPP4 (e.g., with sitagliptin) restores EGFR-TKI sensitivity by disrupting this metabolic adaptation.^320,321 DTPs are also characterized by increased autophagy and altered redox metabolism. PINK1-mediated mitophagy maintains mitochondrial function and redox balance; blocking mitophagy with agents such as chloroquine sensitizes DTPs to MAPK inhibitors and delays recurrence.³²² Notably, DTPs can revert to a proliferative, drug-sensitive state upon treatment cessation, further complicating long-term disease control.³⁰⁸

Importantly, phenotypic plasticity and DTP-mediated resistance do not occur in isolation but intersect with other resistance mechanisms, including target mutations, bypass signaling, efflux transporters, and antiapoptotic pathways.^308,309 EMT, CSCs, and DTPs are interconnected via regulatory networks involving transcription factors (e.g., SNAIL, ZEB1),³¹⁸ signaling cascades (e.g., TGF-β), and cues from the TME (e.g., CAFs, hypoxia), collectively sustaining tumor heterogeneity and adaptive resistance.^{132,323–325}

Combating strategies

Given the central role of phenotypic plasticity and DTPs in therapeutic resistance, targeted strategies have been devised to disrupt these adaptive states and restore treatment sensitivity (Fig. 3b).^308,316 Suppressing EMT and lineage plasticity is a promising approach. For example, inhibition of epigenetic regulators such as ZMYND8 can prevent neuroendocrine transdifferentiation in prostate cancer, while restoration of KMT2C function suppresses the emergence of therapy-resistant DNPC phenotypes.^105,106,326 Targeting key signaling pathways, such as TGF-β, Wnt, and Notch, and EMT-related transcription factors (e.g., ZEB1, TWIST1) is another strategy under active investigation.^252,327 Moreover, certain natural compounds, including those from traditional Chinese medicine, exhibit potential in modulating EMT and suppressing plasticity-associated signaling, opening novel therapeutic avenues.³²⁸

Efforts to eliminate CSCs have focused on disrupting self-renewal and differentiation programs, such as those governed by the Hedgehog, Wnt, and Notch pathways.⁵⁵ Argeting CSC-specific markers (e.g., ALDH) and disrupting feedback loops (e.g., QSOX2–mTOR) can reduce stemness and hinder recurrence.³¹⁹ The metabolic dependencies of EMT-like cells and DTPs offer additional vulnerabilities. Therapeutic approaches include the inhibition of FAO, ROS detoxification pathways, and autophagy. For instance, DPP4 inhibition impairs oxidative metabolism and sensitizes DTPs to EGFR inhibition.³²² Likewise, blocking mitophagy through agents such as chloroquine compromises mitochondrial function, enhancing the efficacy of targeted therapies and preventing relapse^321,322 Recognizing the inherent heterogeneity and adaptability of plastic tumor cells, combination therapies are increasingly emphasized to achieve durable responses. Rationally designed regimens that integrate targeted inhibitors, epigenetic modulators, metabolic drugs and immunotherapies are being tested in preclinical models and clinical trials. Together, these multidimensional strategies hold promise in dismantling the plastic, drug-tolerant tumor cell states that underlie therapy resistance and disease progression.

TME-mediated cancer resistance and combating strategies

The TME in cancer resistance

The TME constitutes a critical layer of therapeutic resistance in which immune and stromal cells, extracellular matrix remodeling, and vascular–metabolic constraints jointly restrict drug and immune efficacy (Fig. 3d). Increasing evidence underscores the multifaceted roles of TME-resident stromal and immune components in fostering tumor survival, immune evasion, and therapeutic failure.³⁷

CAFs are key regulators within the TME that promote resistance through ECM remodeling, metabolic crosstalk, and paracrine signaling.³²⁹ In breast cancer, TSPAN8⁺ myCAFs secrete IL-6 and IL-8 through the SASP, enhancing chemoresistance. Moreover, they activate the MAPK11–RBBP6–SIRT6 axis, which upregulates GLS1 and PYCR1, reprogramming tumor metabolism and creating a prosurvival microenvironment through aspartate and proline secretion.⁶²

TAMs, especially those polarized toward the M2 phenotype, play an equally critical role in resistance.³³⁰ CAF-derived cytokines (e.g., IL-10, IL-6, CCL2) recruit and polarize macrophages into immunosuppressive M2-TAMs, which secrete TGF-β and IL-10, suppressing cytotoxic T-cell responses and facilitating immune escape.^330,331 Increased glucose uptake by TAMs triggers the hexosamine biosynthetic pathway (HBP), enhancing O-GlcNAcylation of Cathepsin B, which promotes lysosomal secretion and fosters metastasis and chemoresistance.³³²

Bidirectional crosstalk between CAFs and TAMs amplifies immune suppression.⁸ For example, in NSCLC, COL11A1⁺ CAFs accumulate at tumor margins, deposit dense ECM components (COL1A1, COL3A1), and physically restrict CD8⁺ T-cell infiltration by interacting with tumor DDR1.⁷⁸ These CAFs often colocalize with SPP1⁺ TAMs, enhancing fibrosis and immune exclusion. In gastric cancer peritoneal metastasis, SPP1⁺ TAMs and THBS2⁺ matrix CAFs (mCAFs) jointly construct a spatially dense immunosuppressive niche, further limiting T-cell access to tumor cores.³²⁹

T-cell dysfunction and exhaustion are additional hallmarks of TME-driven resistance.³³³ Tumor-derived extracellular vesicles carrying PD-L1 induce T-cell senescence by activating ATM/H2AX–CREB/STAT signaling, reprogramming lipid metabolism and suppressing effector function.³³⁴ TAMs deprived of TNFα signaling retain high expression of hematopoietic PGD2 synthase (HPGDS), continuously secreting prostaglandin D2 (PGD2), which reinforces macrophage immunosuppressive identity and directly inhibits CD8⁺ T-cell cytotoxicity.³³⁵ Additionally, lactate–HCAR1 signaling in tumor cells activates the 14-3-3ζ–STAT3 axis, promoting CCL2/CCL7 production and recruitment of CCR2⁺ PMN-MDSCs, which suppress T cells via ARG1 and ROS, thereby limiting PD-1 blockade efficacy.³⁰⁶ Intrinsic nuclear PD-L1 can also cooperate with p-STAT3 to upregulate IL-6, activating MDSCs through the IL-6/JAK/STAT3 axis in a PD-1–independent manner.¹³⁷ Elevated IL-6 levels are predictive of poor response to immunotherapy in PD-L1^high NSCLC patients.¹³⁷

The ECM not only provides physical scaffolding but also functions as a biochemical regulator of resistance.¹⁸¹ Overproduction of collagen and hyaluronic acid creates a dense ECM that impedes drug diffusion and immune infiltration.¹⁸¹ ECM components interact with integrins to activate FAK, DDR, and Rho/MRTF pathways, fostering apoptosis resistance, CSC enrichment, and metabolic rewiring.^336,337 ECM stiffness further modulates ABC transporter activity, affecting intracellular drug accumulation and compromising the efficacy of chemotherapeutics, targeted agents and ICIs.^181,182

Metabolic reprogramming within the TME is a key determinant of immune resistance.³³⁸ In cholangiocarcinoma, CXCL6–CXCR1/2–JAK/STAT/PI3K-AKT signaling promotes tumor growth, gemcitabine resistance, and lipid metabolism reprogramming while inducing NETs that impair CD8⁺ T-cell cytotoxicity via ROS.⁸¹ Additionally, DLST-mediated succinylation of PDHA1 at K83 enhances PDH activity and α-KG accumulation, which activates OXGR1–MAPK/ERK signaling in TAMs, suppressing MHC-II expression and T-cell priming.³³⁹ Hypoxia-induced HIF activation augments drug efflux, suppresses apoptosis, and enhances DNA repair mechanisms. HIF also promotes EMT, autophagy, and CSC phenotypes. For instance, USP9X-mediated stabilization of HIF-2α facilitates CSC maintenance and platinum resistance.³⁴⁰ Collectively, the TME orchestrates resistance through immune suppression, metabolic rewiring, altered signaling, and physical exclusion. CAFs, TAMs, and MDSCs form an interdependent network that reinforces CSC maintenance, impairs drug efficacy, and hinders durable responses (Fig. 3d).

Combating TME-mediated resistance

To counteract TME-driven resistance, a multipronged therapeutic approach is needed, one that combines tumor-intrinsic targeting with strategies that reprogram or neutralize the suppressive TME. Targeting CAFs and ECM remodeling is a promising strategy (Fig. 3d).^181,183 Inhibition of NOX4 in CAFs reduces ECM stiffness and enhances CD8⁺ T-cell infiltration, restoring immune surveillance and sensitizing tumors to immunotherapy.⁶⁴ Enzymatic degradation of ECM components (e.g., hyaluronidase) or blockade of ECM–integrin interactions (e.g., FAK inhibitors) can also improve drug penetration and reduce stemness-supportive niches.³⁴¹ Disrupting TAM and MDSC recruitment and reprogramming their polarization is another effective strategy.³³⁰ Blocking CSF1R, CCR2, or CXCR2 pathways can deplete or re-educate TAMs and MDSCs.³³⁰ For example, CCR2 inhibitors can prevent monocyte-derived TAM accumulation, enhancing the ICB response.³⁴² Simultaneously, targeting lactate signaling (e.g., HCAR1 antagonists) may prevent myeloid cell recruitment and immune evasion.³⁰⁶

Targeting immunometabolic adaptations is gaining traction. Inhibiting key metabolic nodes such as FAO, HBP, or α-KG signaling in TAMs can restore antigen presentation and enhance T-cell-mediated cytotoxicity.^332,339 Similarly, agents that reverse HIF stabilization or inhibit mitochondrial metabolism in CAFs and CSCs are being explored to disrupt metabolic dependencies that underlie resistance.³⁴⁰ Combination therapies incorporating ICIs, epigenetic drugs, metabolic inhibitors, and TME modulators hold particular promise. For example, cotargeting IL-6/JAK/STAT3 signaling and the PD-1/PD-L1 axis may overcome immune resistance in PD-L1^high tumors with elevated IL-6.¹³⁷ Nanoparticle-based delivery systems are also under investigation to codeliver anticancer agents and TME-targeted molecules in a spatiotemporally controlled manner.¹⁷²

In conclusion, tackling TME-mediated resistance necessitates integrated therapeutic strategies that dismantle the immunosuppressive, fibrotic, and metabolically reprogrammed tumor niche. Rationally designed combination regimens that simultaneously target tumor cells and the TME offer a transformative opportunity to improve outcomes in patients with therapy-resistant malignancies.

Microbiome in cancer resistance and combating strategies

Microbiome in cancer resistance

The tumor microbiome, encompassing bacteria, fungi, viruses, and mycoplasma, has emerged as a key modulator of tumor progression and therapeutic response (Fig. 4a). Increasing evidence suggests that microbial communities residing within tumors or at distant mucosal sites can drive resistance to anticancer therapies through diverse mechanisms, including modulation of drug metabolism and transport, induction of ROS, impairment of DDR, activation of oncogenic signaling, and suppression of antitumor immunity.^324,343

Fig. 4.

Molecular mechanisms and combating strategies in cancer drug resistance. a The role of the tumor-associated microbiome in modulating drug efficacy and resistance, with potential microbiota-targeted interventions. b Impairments in cell death pathways, including apoptosis, necroptosis, pyroptosis, and ferroptosis, and strategies to restore death sensitivity

In pancreatic and colorectal cancers, intratumoral γ-proteobacteria express a long isoform of cytidine deaminase (CDD-L) that catalyzes the deamination of gemcitabine into its inactive metabolite 2’,2’-difluorodeoxyuridine (dFdU), thereby conferring chemoresistance.^344,345 In cervical cancer, colonization by Lactobacillus iners promotes gemcitabine resistance by secreting L-lactate, which activates lactate signaling in tumor cells, triggering glycolytic reprogramming and enhancing nucleotide biosynthesis. Concurrently, L-lactate induces HIF-1α and ROS accumulation, disrupts the G2/M checkpoint, and impairs DDR mechanisms. Tumor-adapted L. iners strains also acquire lacG gene mutations that augment galactose metabolism, further driving glucose flux and lactate buildup.³⁴⁶

Microbial metabolites can also modulate therapeutic outcomes. For example, gut microbiota-derived indole-3-acetic acid (3-IAA), a tryptophan metabolite, accumulates in the TME and is oxidized by neutrophil myeloperoxidase (MPO), generating a burst of ROS. Simultaneously, 3-IAA downregulates the antioxidant enzymes GPX3 and GPX7 in tumor cells, enhancing ROS accumulation and inhibiting autophagy, ultimately sensitizing tumors to chemotherapy.³⁴⁷ In breast cancer, Pseudomonas aeruginosa secretes the quorum-sensing molecule N-(3-oxo-dodecanoyl)-L-homoserine lactone (3OC), which induces ligand-independent dimerization of TβRII, aberrantly activating TGF-β signaling. This cascade crosstalks with the ErbB2, PI3K/Akt, and MAPK pathways, thereby circumventing trastuzumab-mediated HER2 inhibition and promoting drug resistance.³⁴⁸ In colorectal cancer, Bacteroides fragilis interacts with tumor cells via the outer membrane proteins SusD and RagB, directly binding to Notch1 and activating its signaling. This engagement suppresses apoptosis induced by 5-FU and oxaliplatin while enhancing proliferative signaling, contributing to chemoresistance. Notably, the bacteriophage VA7, which targets B. fragilis, restores chemosensitivity in murine models.³⁴⁹ Similarly, Fusobacterium nucleatum activates TLR4/NF-κB signaling and upregulates the antiapoptotic protein BIRC3, conferring resistance to 5-FU.³⁵⁰ In breast cancer, enterotoxigenic B. fragilis (ETBF) secretes BFT-1, relieving the repression of Notch signaling and promoting the self-renewal and survival of breast cancer stem cells, thereby leading to taxane resistance.³⁵¹

Immunotherapy resistance is also influenced by the microbiome. In colorectal cancer, Porphyromonas gingivalis impairs cytotoxic T-cell and NK cell function, fostering immune escape.³⁵² Peptostreptococcus anaerobius binds integrin α2β1 on tumor cells, activating NF-κB and inducing CXCL1, which recruits MDSCs via CXCR2. Additionally, the bacterial lysozyme-like protein LytC_22 binds to Slamf4 on MDSCs, enhancing their immunosuppressive phenotype (increased Arg1 and iNOS), thereby suppressing CD8⁺ T-cell activation and resistance to ICB.³⁵³ In oral cancers, P. gingivalis activates the Akt–STAT3 pathway in dendritic cells through gingipain proteases, upregulating PD-L1 expression and suppressing CD8⁺ T-cell function. Clearance of P. gingivalis via antibiotics or knockout of the Kgp gene reverses ICB resistance.³⁵⁴

Combating strategies

Overcoming microbiome-mediated therapeutic resistance necessitates a multifaceted approach integrating microbial profiling, precision modulation, and host–microbiota–immune interplay (Fig. 4b).³²⁴ Targeted microbial clearance using selective antibiotics has demonstrated efficacy in restoring drug sensitivity; for example, depletion of P. gingivalis or F. nucleatum reverses resistance to ICB and chemotherapy, respectively.^351,354 However, nonselective antibiotic administration risks disrupting the commensal microbiota and weakening systemic immune responses. Precision antimicrobials such as bacteriophages (e.g., VA7 targeting B. fragilis) represent a promising alternative, allowing strain-specific depletion while preserving microbial diversity.^349,351

Another strategic axis involves microbiome-responsive drug design. Engineering drugs that resist microbial enzymatic degradation, such as CDD-L and resistant gemcitabine analogs, can circumvent intratumoral drug inactivation.^344,345 Moreover, prodrug conjugates with microbially cleavable linkers may enable site-specific drug release in tumor regions devoid of resistance-conferring microbes.³⁴⁶ In parallel, modulation of microbial metabolites offers therapeutic leverage: promoting beneficial compounds such as 3-IAA, which enhances ROS and chemosensitivity, or suppressing lactate-secreting strains such as Lactobacillus iners may reverse resistance-associated metabolic reprogramming.^346,347

Reprogramming the immune microenvironment is also critical.³⁵² Disrupting microbe–host immunosuppressive circuits, such as integrin α2β1–CXCL1–MDSC or Slamf4–Arg1/iNOS axes, restores antitumor T-cell activity and improves response to immunotherapy.^353,354 Microbial engineering approaches, including probiotic strains secreting immunostimulatory cytokines or locally delivering checkpoint inhibitors, are under active development. Finally, integrating metagenomic, metabolomic, and spatial transcriptomic analyses enables the identification of resistance-associated microbial signatures and host–microbiota interactions.^12,13,355 Collectively, these strategies underscore the microbiome not merely as a passive bystander but also as an active modulator of therapeutic outcomes. Therapeutically targeting the microbiome through eradication, modulation, or functional rewiring holds promise to overcome drug resistance, reinvigorate antitumor immunity, and enhance the durability of cancer therapies across diverse tumor contexts.

Cell death mechanism in cancer resistance and combating strategies

Cell death mechanism in cancer resistance

Programmed cell death is a fundamental mechanism for maintaining tissue homeostasis in multicellular organisms.³⁵⁶ One of the hallmarks of cancer is resistance to cell death, which contributes to tumor progression and therapeutic resistance.³ Dysregulated forms of cell death, including apoptosis, necroptosis, pyroptosis, ferroptosis, and cuproptosis, play distinct roles in shaping the drug response through unique signaling pathways.³⁵⁶

In DNA damage–induced conditions, frequent inactivation of the SLFN11 gene in tumor cells leads to the loss of its tRNAse activity, causing ribosome stalling and inhibition of protein translation. This disrupts activation of the ZAKα–MAPK–JNK apoptotic cascade and weakens GCN2-mediated global translation suppression, thereby abolishing p53-independent apoptosis and contributing to chemoresistance against DNA-damaging agents.⁵³ In pancreatic cancer, overexpression of MLKL triggers necroptosis, recruiting macrophages that form macrophage extracellular traps (METs).³⁵⁷ METs are web-like DNA–protein structures released by activated macrophages that can remodel the TME, dampen antitumor immunity, and thereby contribute to therapeutic resistance. METs cleave CXCL8 into active monomers to promote EMT and tumor–endothelial adhesion and release matrix metalloproteinases (MMPs) to degrade ECM, facilitating metastasis. Necroptosis also induces macrophage IL-6 production, which upregulates CD47 on tumor cells, allowing immune evasion and driving liver metastasis and therapeutic resistance.^357,358 In bladder cancer, gemcitabine induces caspase-1–dependent pyroptosis, releasing inflammatory mediators that activate the CCR6 signaling pathway in the TME. This reprograms αSMA⁺ CAFs into collagen-III⁺ inflammatory CAFs (iCAFs), forming a fibrotic niche that supports CD44⁺ CSCs, thereby fueling chemoresistance.³⁵⁸

In colorectal cancer, ferroptosis suppression via the CBX3/NRF2/GPX2 axis promotes multidrug resistance. CBX3 represses CUL3 transcription, leading to reduced NRF2 ubiquitination and degradation. Stabilized NRF2 translocates to the nucleus and upregulates antioxidant enzymes such as GPX2, enhancing lipid peroxide detoxification and preventing ferroptosis induced by irinotecan and oxaliplatin.³⁵⁹ Another study revealed that tumor acidosis and elevated LDHA activity promote lactate accumulation, which leads to histone H4K12 lactylation (H4K12la) via p300. This epigenetic modification activates GCLC transcription, enhances glutathione synthesis, and protects against lipid peroxidation and ferroptosis, thereby driving chemoresistance in colorectal CSCs.³⁶⁰ USP18 promotes resistance to sorafenib in HCC by removing ISGylation from nuclear receptor coactivator 4 (NCOA4), leading to its degradation and suppression of NCOA4-mediated ferritinophagy.¹⁰¹ This reduces iron-dependent lipid peroxidation and ferroptosis. Targeting USP18 with hypericin (HYP) restores NCOA4 stability and ferroptotic sensitivity, overcoming acquired resistance.^101,361

Cuproptosis, a recently discovered copper-dependent form of programmed cell death, is triggered by copper accumulation in mitochondria and preferentially occurs in cells with active tricarboxylic acid (TCA) cycles.³⁶² Excess copper binds PDK1 and enhances its interaction with AKT, activating the AKT/GSK3β/β-catenin cascade. This signaling promotes CSC traits, while the activated β-catenin/TCF4 complex transcriptionally upregulates ATP7B, a copper efflux transporter. Increased ATP7B expression facilitates copper export and reduces intracellular copper stress, thereby enabling CSCs to evade cuproptosis and acquire resistance.³⁶³ In the hypoxic TME, HIF-1α activation suppresses cuproptosis by inducing PDK1/3, which inhibits the expression of DLAT, a key cuproptosis mediator. Simultaneously, HIF-1α upregulates metallothionein MT2A, which chelates and sequesters mitochondrial copper, preventing copper accumulation beyond the cytotoxic threshold. Moreover, excess copper stabilizes HIF-1α by inhibiting its ubiquitination and degradation, forming a positive feedback loop that reinforces cuproptosis resistance.³⁶²

Autophagy mediates resistance to chemotherapeutic agents by promoting tumor cell survival and adaptation to stress.³⁶⁴ One study showed that circSEC24B induces oxaliplatin resistance in colorectal cancer by enhancing autophagic activity and thereby stabilizing the SRPX2 protein.³⁶⁵ METTL3 downregulates DCP2 expression through m6A modification, thereby enhancing Pink1-Parkin pathway-mediated mitophagy, alleviating mitochondrial damage, and ultimately conferring chemoresistance in SCLC.³⁶⁶ Reticulophagy is a receptor-mediated form of selective autophagy that promotes cell survival, tumor progression, and chemoresistance across multiple cancers by maintaining endoplasmic reticulum homeostasis and buffering therapy-induced stress.³⁶⁷ In bladder cancer, the highly expressed receptor CCPG1 drives reticulophagy to facilitate tumor progression and mediate cisplatin resistance, and inhibition of this pathway represents a potential therapeutic strategy to reverse resistance. Mitophagy constitutes a critical adaptive survival mechanism for tumor cells under therapeutic pressure by eliminating treatment-induced damaged mitochondria, limiting reactive oxygen species accumulation, and maintaining mitochondrial homeostasis, thereby providing a buffering capacity for the development of resistance.³⁶⁸ Recent studies further reveal that the mitophagy receptor NLRX1 is directly regulated by the metabolic signal cytosolic acetyl-coenzyme A (AcCoA) and mediates resistance to KRAS inhibitors, offering a novel molecular framework and actionable vulnerability for overcoming targeted therapy resistance.

Combating strategies

Overcoming cancer resistance mediated by dysregulated cell death mechanisms requires tailored strategies that restore or exploit specific forms of programmed cell death (Fig. 4b).³⁵⁶ For tumors deficient in apoptosis, such as those with SLFN11 inactivation or impaired p53-independent pathways, strategies to restore translational stalling (e.g., via tRNAse mimetics) or activate alternative stress responses may sensitize cells to DNA-damaging agents.⁵³ In the context of necroptosis, therapeutic modulation must be approached with caution. Although MLKL-mediated necroptosis can trigger immune activation, it may also promote metastasis and immune evasion via MET formation and CD47 upregulation.^357,358 Thus, combinatorial strategies that block downstream pro-metastatic cytokines (e.g., IL-6 or CXCL8) or macrophage reprogramming may uncouple necroptosis from its deleterious consequences.^357,358

In pyroptosis-mediated resistance, interventions targeting inflammasome components (e.g., caspase-1 inhibitors) or downstream fibroblast reprogramming (e.g., CCR6 antagonists) may limit fibrotic niche formation and CSC expansion.³⁵⁸ For ferroptosis, which represents a therapeutically actionable vulnerability, restoring iron-dependent lipid peroxidation is key. This may involve targeting the CBX3/NRF2/GPX2 axis with CBX3 or NRF2 inhibitors or impairing antioxidant buffering systems such as glutathione synthesis (e.g., via GCLC suppression).³⁵⁹ Epigenetic modulators that reverse histone lactylation, such as p300 inhibitors, may also resensitize CSCs to ferroptosis by limiting metabolic adaptation.³⁶⁰ In HCC, stabilizing ferritinophagy mediators such as NCOA4 using deISGylation inhibitors such as hypericin can restore ferroptotic sensitivity and overcome resistance to agents such as sorafenib.^101,264 Targeting cuproptosis presents a novel frontier. Strategies include inhibition of copper efflux transporters such as ATP7B or blockade of upstream β-catenin signaling to prevent metabolic escape in CSCs.³⁶³ In hypoxic tumors, antagonizing HIF-1α or its transcriptional targets (e.g., PDK1/3, MT2A) may prevent mitochondrial copper sequestration and promote lethal copper accumulation.³⁶² Additionally, copper ionophores or copper-ligand complexes (e.g., elesclomol) are being explored to selectively induce cuproptosis in tumor cells with active mitochondrial metabolism.³⁶⁹

Ultimately, combining cell death pathway modulators with standard therapies, ICIs, or epigenetic agents offers synergistic potential. Precision medicine approaches, guided by biomarkers such as GPX2, ATP7B, or H4K12la, may enable the stratification of patients for ferroptosis- or cuproptosis-inducing treatments. These integrative strategies hold promise to overcome resistance rooted in impaired cell death and improve clinical outcomes across malignancies.

Combating resistance: current and novel strategies

Overcoming drug resistance remains a central challenge in oncology.¹⁷⁸ Despite substantial progress in targeted therapies and immunotherapies, both intrinsic and acquired resistance continue to limit durable responses.³⁷⁰ Accordingly, multiple strategies are being developed to circumvent or delay resistance, including rational drug combinations, novel therapeutic agents, microbiome interventions, adaptive treatment regimens, and advanced delivery systems (Fig. 5). These approaches aim to target tumor plasticity, prevent escape mechanisms, and resensitize resistant cells.

Fig. 5.

Current and emerging strategies to overcome therapeutic resistance. Schematic overview of translational and therapeutic approaches aimed at counteracting molecular and microenvironmental drivers of cancer drug resistance. Illustrated strategies span both clinical and preclinical development, including rational drug combinations targeting pathway redundancy, next-generation agents such as PROTACs and antibody–drug conjugates, microbiome modulation strategies, adaptive and intermittent dosing regimens to limit evolutionary selection and advanced drug delivery platforms such as nanocarriers designed to enhance tumor targeting and therapeutic index

Rational drug combinations

Strategies to combat MDR include direct inhibition of transporter proteins, regulation of SLC function and expression pathways, development of nanocarrier systems, and combination with emerging targeted therapies.¹⁴⁶ Drugs that competitively bind transporter sites block chemotherapy efflux and increase intracellular drug concentration, such as inhibitors of ABCB1 (P-gp), ABCG2, and ABCC1. WS-917 binds ABCB1, stimulates its ATPase activity, and reverses resistance by increasing intracellular paclitaxel by 730-fold, with low toxicity and synergy with CD8⁺ T-cell responses.¹⁷⁴ Pancreatic cancer cells overexpress the bicarbonate transporter SLC4A4, promoting TME acidification, lactate secretion, and immunosuppression. Inhibiting SLC4A4 alleviates TME acidosis, restores CD8⁺ T-cell function, and sensitizes tumors to ICB.³⁷¹ Interventions upstream of transporters can downregulate expression and mitigate resistance at its root. CDK6 knockout downregulates PI3K110α/β, reduces ABCB1 mRNA splicing, and increases doxorubicin accumulation in tumors.³⁷² PLGA-PEG nanoparticle-based systems efficiently deliver HECTD2 inhibitors or KEAP1 stabilizers to inhibit HECTD2-mediated KEAP1 degradation, restoring KEAP1/NRF2 balance, suppressing antioxidant responses, and reversing lenvatinib resistance in HCC.²⁹³ Mirvetuximab soravtansine, the first FRα-targeting antibody‒drug conjugate (ADC), achieves OS benefits in platinum-resistant ovarian cancer.³⁷³

These mechanistic insights are now being translated into rationally designed combination regimens in clinical trials, marking a significant shift toward biology-driven therapeutic strategies (Table 3). Several early-phase studies are exploring synergistic pairings of epigenetic modulators, ICBs, DNA damage response agents, and targeted therapies across diverse cancer types. For example, epigenetic–chemotherapy combinations such as panobinostat plus epirubicin (NCT00878904) and vorinostat with trastuzumab (NCT00258349) aim to concurrently disrupt chromatin remodeling and HER2 signaling in metastatic tumors. A triple combination targeting epigenetic dysregulation and apoptotic resistance, comprising iadademstat (LSD1 inhibitor), azacitidine (a hypomethylating agent), and venetoclax (a BCL-2 inhibitor), is under investigation in a phase I trial for acute myeloid leukemia (NCT06357182). To overcome ICB resistance, combinations of decitabine or azacitidine with pembrolizumab are being tested in refractory Hodgkin’s lymphoma and solid tumors (NCT05355051, NCT03445858). Metabolic vulnerabilities are being exploited through combinations such as sirolimus (an mTOR inhibitor) and auranofin (a thioredoxin reductase inhibitor) in advanced NSCLC and SCLC (NCT01737502). Similarly, kinase–chemotherapy regimens, including lenvatinib with carboplatin/paclitaxel (NCT00832819) or with temozolomide (NCT00121680), are being evaluated for their multitargeted efficacy. Notably, the ongoing phase I/II trial of XNW5004, an EZH2 inhibitor, in combination with pembrolizumab (NCT06022757) exemplifies the effort to reprogram epigenetic states and restore ICB sensitivity in solid tumors resistant to conventional treatments. Collectively, these clinical trials reflect a paradigm shift toward mechanism-guided, biomarker-informed combination therapies. As our understanding of resistance evolution continues to advance, such rationally designed regimens, grounded in robust preclinical evidence, will be instrumental in achieving durable responses and delaying disease progression across multiple cancer types.

Table 3.

The summary of clinical trials about combating strategies based on resistance mechanisms

Study Identifier	Phase	Study type	Experimental drug	Mechanism	Cancer condition(s)	Study status
Clinical trials of rational drug combinations
NCT00878904	Ⅰ	Interventional	Panobinostat plus epirubicin	HDACis + anthracycline-based chemotherapy	Metastatic malignant solid tumors	Completed
NCT06357182	Ⅰ	Interventional	Iadademstat in combination with Azacitidine and Venetoclax	LSD1 inhibitors + HMAs + BCL-2 inhibitors	AML	Recruiting
NCT01896856	Ⅰ/Ⅱ	Interventional	Guadecitabine with Irinotecan	DNMT inhibitors + topoisomerase I inhibitors	Previously treated metastatic colorectal cancer	Completed
NCT00258349	Ⅰ/Ⅱ	Interventional	Vorinostat and Trastuzumab	HDACis + anti-HER2 monoclonal antibodies	Metastatic or locally recurrent breast cancer	Completed
NCT06022757	Ⅰ/Ⅱ	Interventional	XNW5004 in combination with Pembrolizumab	EZH2 inhibitors and PD-1 blockade	Advanced solid tumors that failed standard treatments	Recruiting
NCT05355051	Ⅱ	Interventional	Azacitidine and Pembrolizumab	HMAs + PD-1 inhibitors	Relapsed/refractory Hodgkin’s lymphoma	Recruiting
NCT03445858	Ⅰ	Interventional	Pembrolizumab in combination with Decitabine and hypofractionated index lesion radiation	HMAs + PD-1 inhibitors + radiation	Relapsed, refractory or progressive non-primary CNS solid tumors and lymphomas	Completed
NCT01737502	Ⅰ/Ⅱ	Interventional	Sirolimus and Auranofin	mTOR inhibitors + TrxR inhibitors	Advanced or Recurrent NSCLC or SCLC	Completed
NCT00832819	Ⅰ	Interventional	Lenvatinib in combination with Carboplatin and Paclitaxel	Multitarget TKIs + platinum and taxane-based chemotherapeutics	NSCLC	Completed
NCT00121680	Ⅰ/Ⅰb	Interventional	Lenvatinib and Temozolomide	Alkylating agents + multitarget TKIs	Advanced and/or metastatic melanoma	Completed
Clinical trials of novel agents
NCT02777710	Ⅰ	Interventional	DURVALUMAB combined with PEXIDARTINIB	Anti-PD-L1 antibody + CSF-1R TKI	Metastatic/advanced pancreatic or colorectal cancers	Completed
NCT03428217	Ⅱ	Interventional	CB-839 with Cabozantinib	Selective GLS inhibitors + multitarget TKIs	Metastatic RCC	Completed
NCT03934372	Ⅰ/Ⅱ	Interventional	Ponatinib	Third-generation BCR-ABL inhibitors	Pediatric recurrent or refractory leukemias, lymphomas or solid tumors	Recruiting
NCT04862780	Ⅰ	Interventional	BLU-945	Selective EGFR inhibitor	NSCLC	Terminated
NCT06881784	Ⅲ	Interventional	Daraxonrasib	RAS(ON) multi-selective, noncovalent tri-complex inhibitor	NSCLC	Recruiting
NCT06625320	Ⅲ	Interventional	Daraxonrasib	RAS(ON) multi-selective, noncovalent tri-complex inhibitor	PDAC	Active, not recruiting
NCT05573555	Ⅰ/Ⅱ	Interventional	ARV-471	Selective degradation of ERα via PROTACs	Advanced or metastatic breast cancer	Recruiting
NCT04404595	Ⅰ/Ⅱ	Interventional	CT041	Autologous CAR-T cell therapy targeting CLDN18.2	Gastric, pancreatic cancer, or other specified digestive cancers	Active, not recruiting
NCT04754191	Ⅱ	Interventional	Enfortumab Vedotin	ADCs targeting Nectin-4	Metastatic castration-resistant prostate cancer	Recruiting
NCT04077463	Ⅰ	Interventional	Amivantamab	BsAb targeting EGFR and c-MET	Advanced NSCLC	Active, not recruiting
Clinical trials of adaptive and intermittent therapies
NCT03810872	Ⅱ	Interventional	Afatinib	Irreversible TKIs targeting EGFR, HER2, and HER4	Advanced cancer carrying an EGFR, a HER2 or a HER3 mutation	Unknown status
NCT02415621	Ⅰ	Interventional	Abiraterone	Adaptive therapy restores tumor sensitivity to hormonal treatment	Metastatic castration resistant prostate cancer	Active, not recruiting
NCT03855020	/	Observational	Chemotherapy and radiotherapy	Plasma EBV DNA-guided adaptive therapy	Nasopharyngeal carcinoma	Unknown status
NCT01728311	Ⅰ	Interventional	BAY1082439	Intermittent administration of PI3Kα/β/δ inhibitors to overcome resistance	Advanced cancer	Completed
Clinical trials targeting microbiome, oncolytic viral/bacterial, and innate immune modulation
NCT04068181	Ⅱ	Interventional	T-VEC with Pembrolizumab	Selective infection and lysis of tumor cells coupled with the activation of systemic antitumor immune responses	Melanoma following progression on prior anti-PD-1-based therapy	Completed
NCT03003676	Ⅰ	Interventional	ONCOS-102 and Pembrolizumab	Engineered oncolytic adenoviruses expressing GM-CSF to activate dendritic cells	Advanced or unresectable melanoma progressing after PD-1 blockade	Completed
NCT04167137	Ⅰ	Interventional	SYNB1891	Delivery of STING agonists to activate dendritic cells and T Cells	Metastatic solid neoplasm	Terminated
NCT03775850	Ⅰ	Interventional	EDP1503	Induce antitumor responses via gut-mediated activation of both innate and adaptive immunity	Colorectal cancer, breast cancer, and checkpoint inhibitor relapsed tumors	Completed
NCT04130763	Ⅰ	Interventional	FMT capsule	Modulating the gut microbiota to enhance immunotherapy response	Gastrointestinal system cancer	Completed
NCT03435952	Ⅰ	Interventional	Pembrolizumab with intratumoral injection of Clostridium Novyi-NT	Intratumoral replication and oncolytic toxin expression	Malignant neoplasm	Active, not recruiting
Clinical trials of nanotechnology and smart drug delivery
NCT02213744	Ⅱ/Ⅲ	Interventional	MM-302 Plus Trastuzumab	Dual HER2 targeting with Trastuzumab and liposomal Doxorubicin demonstrates synergistic antitumor activity	HER2-Positive locally advanced/metastatic breast cancer	Terminated
NCT03531827	Ⅱ	Interventional	Combining CRLX101 with Enzalutamide	Nanoparticles formed by Camptothecin encapsulated in PEG-PLA	Progressive metastatic castration resistant prostate cancer	Terminated
NCT00583349	Ⅰ/Ⅱ	Interventional	Intravesicular Abraxane	Albumin-bound paclitaxel nanoparticles	Treatment-refractory bladder cancer	Completed
NCT04995536	Ⅰ	Interventional	CpG-STAT3 siRNA CAS3/ SS3 and localized radiation therapy	Targeted delivery of immunostimulatory agents and gene silencers via nucleic acid nanocomplexes for synergistic antitumor therapy	Relapsed/refractory B-Cell NHL	Withdrawn
NCT06271564	Ⅰ	Interventional	Topical magnetite Zinc oxide composite nanoparticles	Magnetic targeting to enhance drug penetration	Oral potentially malignant lesions	Recruiting

HDACis histone deacetylase inhibitors, LSD1 lysine-specific demethylase 1, HMAs hypomethylating agents, AML acute myeloid leukemia, DNMTis DNA methyltransferase inhibitors, EZH2 enhancer of zeste homolog 2, NSCLC non-small cell lung cancer, SCLC small cell lung cancer, TrxR thioredoxin reductase, TKIs tyrosine kinase inhibitors, GLS glutaminase, RCC renal cell carcinoma, ERα estrogen receptor α, PROTACs proteolysis-targeting chimeras, CLDN Claudin, ADCs antibody-drug conjugates, BsAb bispecific antibody, T-VEC Talimogene Laherparepvec, GM-CSF granulocyte-macrophage colony-stimulating factor; FMT: fecal microbiota transplant; PEG-PLA: polyethylene glycol-polylactic acid

Novel agents targeting resistance mechanisms

Innovative drug design strategies have emerged to limit the development of drug resistance. These include nonclassical binding drugs, PROTACs, allosteric drugs, and multisite targeting compounds.^116,153 For example, the third-generation BCR-ABL inhibitor ponatinib forms novel hydrophobic and van der Waals interactions with the T315I-mutated ABL kinase, providing precise inhibition of resistance in acute lymphoblastic leukemia.³⁷⁴ PROTACs targeting the N-terminal domain (NTD) of AR degrade AR variants, effectively blocking persistent AR signaling and overcoming anti-androgen resistance in advanced prostate cancer.³⁷⁵ However, PROTACs face translational challenges due to poor membrane permeability, instability, suboptimal pharmacokinetics, uneven tissue distribution, and limited specificity. To address these issues, nanotechnology has been employed for PROTACs delivery.³⁷⁶ NPs enhance PROTAC stability, prevent premature degradation in circulation, and improve tumor accumulation and cellular uptake via passive or active targeting mechanisms, ultimately enhancing efficacy and reducing systemic toxicity.³⁷⁷ Allosteric site targeting enhances treatment specificity and reduces side effects. Sotorasib binds covalently to the switch II pocket formed by the KRAS G12C mutation, inhibiting KRAS activity and overcoming resistance.³⁷⁸ Amivantamab, a bispecific antibody targeting EGFR and MET, addresses the limitations of monotargeted inhibitors and is approved for advanced or metastatic NSCLC with EGFR exon 20 insertions.³⁷⁹

A growing number of early-phase clinical trials are now translating these novel mechanisms into therapeutic strategies to overcome resistance (Table 3). For instance, CB-839 (a glutaminase inhibitor) paired with cabozantinib (a multitarget TKI) (NCT03428217) is being tested in metastatic renal cell carcinoma, leveraging metabolic reprogramming to enhance TKI efficacy. Ponatinib, a third-generation BCR-ABL inhibitor with potent activity against T315I-mutant leukemia, is under investigation in pediatric relapsed or refractory hematologic malignancies (NCT03934372). Several trials are exploring next-generation mutant-selective agents. BLU-945, a selective EGFR inhibitor, was tested in EGFR-mutated NSCLC (NCT04862780), while daraxonrasib, a RAS(ON) multiselective noncovalent tri-complex inhibitor, is being evaluated in both NSCLC (NCT06881784) and PDAC (NCT06625320). PROTAC-based therapies are also entering clinical development; ARV-471, which selectively degrades estrogen receptor α, is in phase I/II trials for advanced breast cancer (NCT05573555). In the realm of cell-based therapies and antibody–drug conjugates (ADCs), CT041, a CLDN18.2-targeted CAR-T product, is being studied in gastrointestinal cancers (NCT04404595), while enfortumab vedotin, targeting Nectin-4, is under investigation for metastatic castration-resistant prostate cancer (NCT04754191). Amivantamab, the bispecific EGFR–MET antibody already approved for EGFR exon 20 insertion–positive NSCLC, continues to be evaluated in ongoing clinical trials (NCT04077463). Collectively, these trials reflect a strategic shift toward precision therapeutics that are specifically engineered to circumvent or dismantle key resistance mechanisms, whether through selective degradation, allosteric blockade, metabolic disruption or immunologic remodeling.

Microbiome modulation

The microbiome plays a pivotal role in modulating tumor behavior and treatment outcomes.³⁸⁰ Fecal microbiota transplantation (FMT) shows promise in overcoming PD-1/PD-L1 resistance.³⁸¹ Several clinical trials have demonstrated that FMT can reshape the gut microbiota, activate immune cells within the TME, and increase objective response rates by approximately 30–60%, with manageable adverse effects. Probiotic/prebiotic strategies enhance beneficial bacteria, improve gut diversity, and produce immunoactive metabolites that help restore immune efficacy. For example, orally administered probiotic Lactobacillus rhamnosus GG (LGG) targets infiltrated PDAC via the gut–pancreas axis and, using a surface-engineered gallium-polyphenol (Ga-poly) network, modulates intratumoral microbiota and reshapes the immune microenvironment. This strategy synergizes with chemotherapy and ICB to overcome PDAC resistance.³⁸² However, individual variability remains significant, and personalized microbiota combinations may be required for optimal therapy.³⁸⁰ Targeted microbiome interventions to eliminate resistance-associated microbes offer a more precise strategy to improve treatment outcomes. For example, Sang et al. designed a ginger-derived nanovesicle (GDNV)-based targeted nanotherapy (P-GDNVs), which selectively eliminates Porphyromonas gingivalis enriched in oral squamous cell carcinoma (OSCC) tissues. This approach significantly reduces P-gp expression and inhibits the IL-6/pSTAT3 inflammatory pathway, thereby reversing resistance in the TME.³⁸³

Clinical exploration of microbiome-targeting strategies is accelerating, with several early-phase trials demonstrating feasibility and immunomodulatory efficacy (Table 3). In melanoma patients refractory to anti–PD-1 therapy, the combination of Talimogene laherparepvec (T-VEC), an oncolytic herpesvirus, with pembrolizumab (NCT04068181) showed enhanced systemic immune activation and tumor response. Similarly, ONCOS-102, an engineered adenovirus expressing GM-CSF, has been tested with pembrolizumab in advanced melanoma (NCT03003676), aiming to activate dendritic cells and potentiate T-cell–mediated responses. Synthetic biology-based approaches, such as SYNB1891, a live microbial vector delivering STING agonists, have been evaluated for intratumoral immune activation in metastatic tumors (NCT04167137). Gut-directed agents such as EDP1503, derived from a commensal bacterium, modulate both innate and adaptive immunity via the microbiota and were tested in colorectal and breast cancers (NCT03775850). FMT in capsule form has also demonstrated safety and the potential to enhance ICB responsiveness in gastrointestinal cancers (NCT04130763). Additionally, intratumoral delivery of Clostridium novyi-NT spores is being explored as a novel oncolytic and immunogenic strategy in solid malignancies (NCT03435952). These clinical studies support the growing notion that microbiome modulation, whether through FMT, probiotics, or engineered bacterial vectors, offers a compelling adjunct to immunotherapy, particularly in resistant tumors. However, interpatient variability and strain-specific effects underscore the need for precision microbiome approaches tailored to individual host–microbe–tumor interactions.

Adaptive and intermittent therapies

Another strategy involves targeting specific resistance mechanisms to resensitize tumors before rechallenging them with previously ineffective drugs.⁴ Lomitapide blocks the RFWD3/PHGDH interaction, restores PHGDH stability, disrupts NAD⁺/nucleotide metabolism, and synergizes with cisplatin in osteosarcoma.³⁸⁴ Anlotinib plus sotorasib downregulates c-Myc/ORC2, induces apoptosis, and reverses resistance in KRAS-G12C NSCLC.³⁸⁵ Dual PD-1/LAG-3 blockade synergistically reactivated CD8⁺ T cells, repressed TOX, and promoted effector gene expression, reversing exhaustion and enhancing immunity.^386,387 Intermittent dosing, which reduces selection pressure, allows sensitive clones to repopulate and delays resistance. Intermittent PI3K inhibition reshapes the immune TME in PTEN-null prostate cancer, converts “cold” tumors to “hot” tumors, suppresses T_regs while preserving CD8⁺ T-cell function, and maintains inflammatory states posttreatment.³⁸⁸ Adaptive therapy uses biomarkers to tailor treatment intensity. The Moffitt Cancer Center’s trial adjusted abiraterone dosing based on PSA/testosterone dynamics, delaying resistance and reducing toxicity.³⁸⁹ The EP-STAR study used ctEBV DNA levels to stratify nasopharyngeal carcinoma patients, improving outcomes (EP-SEASON study).³⁹⁰ MRD monitoring by circulating tumor DNA (ctDNA) allows real-time adjustment of therapy. Escalation during ctDNA increase and de-escalation during decline maintains sensitive clone dominance and delays resistance.^1,391

Clinical trials of adaptive and intermittent therapies have begun to validate these strategies in real-world settings (Table 3). A phase I trial (NCT02415621) evaluating adaptive dosing of abiraterone in metastatic castration-resistant prostate cancer demonstrated that intermittent androgen pathway suppression based on PSA dynamics could delay resistance and reduce cumulative toxicity. Notably, adaptive therapy prolonged the time to progression to at least 27 months compared with approximately 16.5 months under standard dosing, while reducing cumulative drug exposure by nearly half.³⁹² Similarly, the EP-STAR observational study (NCT03855020) in nasopharyngeal carcinoma implemented a plasma EBV DNA-guided treatment adjustment strategy, allowing escalation or de-escalation based on real-time viral load monitoring. In the realm of targeted therapy, the intermittent use of the PI3Kα/β/δ inhibitor BAY1082439 (NCT01728311) was shown to reprogram the immune microenvironment and sustain antitumor responses, particularly in PTEN-deficient cancers. Furthermore, NCT03810872 explores the utility of intermittent afatinib in patients with EGFR, HER2, or HER3 mutations to mitigate resistance via on-off EGFR/HER blockade. These trials highlight the clinical feasibility and therapeutic benefit of dynamic treatment modulation, offering a promising paradigm to delay or overcome acquired resistance while minimizing toxicity.

Nanotechnology and smart drug delivery

Nanoparticle-based drug delivery systems have emerged as promising tools in cancer therapy due to their capacity to improve pharmacokinetic profiles, enable tumor-targeted delivery, and facilitate site-specific drug release in response to the TME.¹⁷⁰ These features contribute to enhanced antitumor efficacy and hold significant clinical potential. For instance, the HAZD system, comprising gold nanoparticles (AuNPs) and ZIF-8, combines passive targeting via nanoparticles, active targeting via hyaluronic acid, and pH-responsive drug release via ZIF-8. This system enables multistage precision delivery at the tissue, cellular, and even subcellular levels. Moreover, it allows for spatiotemporal control of photothermal therapy (PTT), chemotherapy, and chemodynamic therapy (CDT) by tuning the laser power, effectively overcoming resistance barriers seen in conventional therapies and demonstrating potent in vivo antitumor activity.³⁹³ Another example involves Fe₂O₃-based hollow multishelled structures (HoMS), which enable temporally controlled sequential release of cisplatin and pemetrexed (MTA) within the TME. This dual-release strategy induces both apoptosis and ferroptosis, leading to synergistic killing of lung cancer cells, enhanced chemotherapeutic efficacy, and reduced toxicity to normal tissues.³⁹⁴ Similarly, a pH-responsive niosomal platform codelivering doxorubicin and curcumin achieved enhanced cytotoxicity via tumor-selective drug release and synergistic activity, underscoring the value of smart codelivery systems in overcoming chemoresistance.³⁹⁵

Multifunctional nanocarriers coloaded with a STING agonist and PD-L1 siRNA not only activate the STING pathway to boost innate immunity but also inhibit PD-L1 expression, enabling potent immune activation and broad-spectrum inhibition in “cold tumors”. These systems demonstrate the potential to overcome immune resistance and broaden immunotherapeutic efficacy.³⁹⁶ Other systems using CuO₂ and cisplatin coloaded silica nanoparticles remodel the TME and induce cuproptosis, reversing platinum resistance.³⁹⁷ These biocompatible, targeted, and responsive delivery systems represent promising avenues for overcoming resistance and improving clinical outcomes. Nanocarriers overcome MDR by bypassing efflux pumps, enhancing tumor penetration, and enabling controlled or combinatorial drug delivery.³⁹⁸ Validation in 2D and 3D tumor models further supports the rational optimization of nanomedicines to reverse MDR in physiologically relevant settings.

Clinical trials of nanotechnology and smart drug delivery have begun to evaluate the translational feasibility of these advanced systems (Table 3). For instance, in HER2-positive breast cancer, a phase II/III trial (NCT02213744) investigated MM-302, liposomal doxorubicin targeted to HER2, combined with trastuzumab, aiming to achieve dual HER2 blockade and improved tumor targeting. Although the study was terminated, it provided insight into the pharmacokinetics and safety of antibody-guided nanoparticle delivery. In metastatic castration-resistant prostate cancer, the combination of CRLX101, a nanoparticle camptothecin formulation, with enzalutamide (NCT03531827) explored synergistic cytotoxic and anti-androgenic effects but faced early discontinuation due to limited efficacy. Other trials have tested intravesicular administration of albumin-bound paclitaxel nanoparticles (NCT00583349) in treatment-refractory bladder cancer, which demonstrated tolerability and potential local tumor control. Novel nucleic acid nanocomplexes, such as CpG-STAT3 siRNA conjugates (NCT04995536), were evaluated for their ability to silence oncogenic pathways and stimulate immune responses in relapsed B-cell non-Hodgkin lymphoma (NHL), although the study was withdrawn before completion. Emerging approaches include magnetically guided topical delivery systems (NCT06271564) using magnetite-zinc oxide nanoparticles to enhance mucosal penetration in oral premalignant lesions. These trials collectively reflect the challenges and promise of smart nanocarriers in overcoming drug resistance and improving spatiotemporal control in cancer therapy.

Emerging tools for dissecting drug resistance

Single-cell omics

Single-cell omics technologies enable high-throughput profiling of the genome, transcriptome, epigenome, proteome, and metabolome at the individual cell level.^12,399 These tools have revolutionized our ability to dissect the biological features of malignant tumor cells, uncover intratumoral heterogeneity and its dynamic evolution, elucidate tumor-immune cell interactions, and uncover mechanisms of drug resistance and relapse, paving the way for personalized treatment strategies to improve patient outcomes (Fig. 6a).¹²

Fig. 6.

Emerging tools for dissecting drug resistance and clinical translation. a Technological advances enabling resistance dissection, including single-cell omics, spatial transcriptomics, organoids and patient-derived xenograft (PDX) models, and AI-driven predictive modeling. b Clinical applications and biomarker development: liquid biopsy-based monitoring (e.g., ctDNA, cTCs, and exosomes), predictive biomarkers for patient stratification, personalized treatment and rechallenge strategies, and innovative clinical trial designs

For example, Lambrechts et al. utilized paired tumor samples from head and neck squamous cell carcinoma (HNSCC) patients before and after treatment, combined with tumor-draining lymph node and peripheral blood samples. Through single-cell RNA-seq (scRNA-seq) and single-cell T-cell receptor/B-cell receptor sequencing (scTCR/BCR-seq), they dynamically captured real-time remodeling of the TME during ICI therapy.⁴⁰⁰ They found that the specific expansion of CD4⁺ T cells was a biomarker of early response to anti-PD-L1 plus anti-CTLA-4 therapy. TCR lineage tracking revealed that naive/central memory CD4⁺ T cells were recruited from draining lymph nodes and differentiated into effector/Th1 cells in the tumor.⁴⁰¹ The advent of scRNA-seq has also enabled the detection of rare cellular subpopulations and intercellular communication. In osteosarcoma, T_regs were significantly enriched in the TME and aberrantly activated oxidative phosphorylation, mTORC1 signaling, and angiogenesis pathways, all contributing to osteosarcoma progression.⁴⁰² Cell–cell communication analysis revealed that T_regs interacted through a highly active CXCR4–CXCL12 axis with other TME cells. Moreover, extensive immune checkpoint interactions and metabolic reprogramming of T_regs synergistically promoted chemoresistance in osteosarcoma.^88,402

Single-cell DNA sequencing (scDNA-seq) is mainly used to infer clonal lineages and track resistant subclones harboring mutations. In acute myeloid leukemia (AML), treatment pressure selectively amplifies preexisting minor subclones or enables escape via bypass signaling.¹¹ Using scDNA-seq, researchers tracked the real-time evolution of resistant clones under therapy, identifying key intervention windows.¹¹ Epigenomic analysis, focusing on DNA methylation and histone modifications, when combined with scRNA-seq, allows integration of gene expression with regulatory dynamics. Valemetostat, a dual EZH1/EZH2 inhibitor approved for relapsed/refractory adult T-cell leukemia/lymphoma (ATL), acts by removing H3K27me3-mediated chromatin condensation to restore tumor suppressor gene expression. However, Yamagishi et al. reported that ~50% of long-term treated patients acquired EZH2 or EED mutations, disrupting drug binding and reaccumulating H3K27me3.⁴⁰³ In nonmutated cases, TET2 loss or DNMT3A overexpression created an alternative silencing mechanism via de novo DNA methylation, mimicking the repressive H3K27me3 state.⁴⁰⁴ Joint scRNA-seq and single-cell assay for transposase-accessible chromatin sequencing (scATAC-seq) analyses captured the synchronized collapse of gene expression and chromatin accessibility during adaptive resistance.⁸⁸

Single-cell proteomics, which quantifies protein expression and pathway activation at single-cell resolution, reveals functional heterogeneity, dynamic regulation of resistance-associated proteins, and their interactions with the TME.^405,406 In prostate cancer, proteomic and transcriptomic analyses revealed that drug-resistant cells (DRCs) exhibited polyploidy, enlarged morphology, and elevated total protein abundance. Multidimensional proteomic clustering identified three DRC subpopulations with distinct functional features: translation/metabolism cluster, stress response cluster, and high invasiveness cluster.⁴⁰⁵ Differential expression analysis showed downregulation of adhesion/metabolic proteins and upregulation of immune evasion markers, motility proteins, and chromatin regulators, collectively driving invasive and resistant phenotypes.

Spatial omics

A suite of emerging spatial omics technologies is revolutionizing our understanding of tumor spatial heterogeneity and its relationship with acquired therapeutic resistance (Fig. 6a).¹³ By integrating molecular data with the precise spatial localization of cells within tissue architecture, spatial omics offers a powerful approach to uncover the complex interplay between drug-resistant cancer cells and their surrounding microenvironment.^13,407 These techniques reveal how the spatial distribution of resistant cells and their molecular programs contribute to treatment failure.⁴⁰⁷

In PDAC, spatial transcriptomics combined with SCOTIA-based spatial interaction analysis demonstrated that neoadjuvant chemotherapy reshaped ligand‒receptor (LR) interactions between CAFs and tumor cells.⁴⁰⁸ Although overall CAF-tumor interactions declined post-treatment, residual interactions showed specific enrichment and activation of IL-6 family signaling. IL-6/JAK/STAT signaling delivered via spatial proximity from CAFs to tumor cells functionally promoted EMT, invasion, and chemoresistance.⁴⁰⁸ Spatial metabolomics, built on mass spectrometry imaging, enables high-resolution mapping of metabolites across histological structures, providing insights into spatially regulated metabolic reprogramming that underlies therapeutic resistance.³⁵⁵ Ji et al. integrated spatial metabolomics (SM) and spatial transcriptomics (ST) in nasopharyngeal carcinoma, revealing enhanced branched-chain amino acid (BCAA) metabolic activity enriched in tumor regions and colocalized with immune cells. Regulation of the metabolite α-ketoisovaleric acid (KMV) promoted cancer cell proliferation, invasion, and cell cycle progression, contributing to therapy resistance.⁴⁰⁹

The advent of spatial proteomics has further extended spatial omics into the proteome. By combining multiplexed immunolabeling and imaging platforms such as PhenoCycler (formerly CODEX) and multiplexed fluorescence in situ hybridization (mFISH), dozens to hundreds of proteins can be visualized and quantified in situ while preserving tissue structure.^13,407 Integrative analysis using scRNA-seq, ST, and spatial proteomics in melanoma identified the enrichment of mesenchymal-like (MES-like) cancer cells in ICI-nonresponsive lesions.¹³ These cells exhibited TCF4-driven transcriptional programs that suppressed melanocytic differentiation and antigen presentation, impairing T-cell recognition and enabling immune escape. Targeting TCF4 expression restored sensitivity to both ICIs and targeted therapies.⁴¹⁰

Despite their transformative potential, spatial omics technologies face notable barriers to clinical translation.¹³ High costs, technical complexity, and limited throughput restrict their accessibility. Furthermore, robust data analysis relies on specialized pipelines for image segmentation, spatial modeling, and multimodal integration, which remain under active development. The standardization of platforms, reproducibility of results, and harmonization with traditional pathology are ongoing challenges.^13,407 Nonetheless, spatial omics represents a next-generation tool for dissecting mechanisms of acquired drug resistance in cancer. Resolving the spatial determinants of treatment failure at the cellular and molecular levels offers a critical perspective for uncovering new therapeutic vulnerabilities. Future integration of spatial omics with genomics, transcriptomics, and epigenomics is expected to yield robust biomarkers and enable precision strategies to overcome resistance.

Organoids and patient-derived models

Organoids and patient-derived models have emerged as powerful tools for dissecting drug resistance mechanisms and facilitating translational research (Fig. 6a). These models retain the genetic, histological, and functional heterogeneity of primary tumors, offering superior physiological relevance compared to traditional cell lines.^411,412

Patient-derived tumor organoids (PDOs) are three-dimensional (3D) microtissue models developed in vitro from normal or malignant tissues.⁴¹² These organoids faithfully recapitulate the histological architecture and genetic landscape of the original tumors while retaining the stem cell potential for proliferation and differentiation. These models have been used to test combination therapies and unravel mechanisms. As such, they have emerged as a powerful platform for investigating the molecular mechanisms underlying cancer drug resistance.⁴¹³ For instance, Yang et al. successfully established 20 gastric cancer PDOs from freshly resected tumors. Using this platform, they demonstrated that Aurora kinase inhibitors (AURKi) induce senescence in diffuse-type gastric cancer (DGC) cells and promote secretion of MCP-1/CCL2, which remodels the immune microenvironment and skews macrophages toward an M2 phenotype, contributing to therapeutic resistance.⁴¹² PDOs can also serve as models for acquired drug resistance. In a study on NSCLC, investigators generated PDOs from drug-resistant tumor tissues to replicate acquired resistance in vitro. They showed that the CDK9 inhibitor Z11 effectively suppressed tumor growth in osimertinib-resistant models, thereby capturing clinical resistance mechanisms.⁴¹⁴ Despite their utility, PDOs face certain limitations.¹⁷ They are primarily applicable to epithelial solid tumors and typically lack components of the TME. However, this shortcoming can be partially addressed by coculturing PDOs with stromal or immune cells. As an emerging clinical model, PDOs hold great promise not only for elucidating resistance mechanisms but also for drug screening and guiding individualized therapeutic strategies.

Patient-derived xenograft (PDX) models involve the direct implantation of primary human tumor tissues or cells into immunodeficient mice, preserving the original tumor’s histopathological architecture, genomic background, and molecular characteristics within an in vivo context.^415,416 PDX models maintain tumor heterogeneity and recapitulate tumor–microenvironment interactions, making them widely used for studying drug resistance mechanisms. In the context of acquired resistance, PDX models can be subjected to repeated drug exposure to mimic disease progression observed in patients transitioning from initial sensitivity to resistance.⁴¹⁶ Comparative multiomics analyses between drug-sensitive and drug-resistant PDXs can identify key genetic mutations, pathway rewiring, and phenotypic transitions associated with resistance.⁴¹⁶ For example, Ma et al. used MYC-amplified breast cancer PDXs and employed Mini-PDX techniques to evaluate treatment responses in vivo. They found that MYC amplification promotes transcription of the E3 ligase KLHL42, leading to pRB1 ubiquitination and degradation, thereby conferring resistance to CDK4/6 inhibitors.⁴¹⁷ Similarly, Yao et al. established a large PDX cohort for HNSCC by orthotopically transplanting surgical tumor samples into highly immunodeficient mice. Longitudinal sampling and molecular profiling before, during, and after cetuximab therapy revealed that acquired resistance was driven by RAS pathway activation.⁴¹⁶ While PDX models offer high fidelity and translational relevance, they are not without limitations.⁴¹⁵ The success rate of PDX establishment varies across cancer types.¹⁶ Furthermore, as these models rely on immunocompromised hosts, they cannot replicate immune-related resistance mechanisms.⁴¹⁶ PDX generation is also time-consuming, costly, and technically demanding, posing challenges for standardization and large-scale clinical implementation.⁴¹⁶

Artificial intelligence (AI)-driven prediction and model integration

AI, particularly machine learning (ML) and deep learning (DL), is revolutionizing the study of cancer drug resistance by enabling predictive modeling, multiomics integration, and dynamic treatment optimization (Fig. 6a).⁴¹⁸ In modeling tumor resistance, appropriate AI algorithms must be selected according to specific experimental contexts to effectively uncover latent patterns and key associations within resistance-related data.⁴¹⁹ Currently, a spectrum of approaches ranging from traditional machine learning methods to deep learning models has been applied to integrate pathological imaging, molecular subtyping, and clinical information, enabling accurate prediction of resistance outcomes and tumor recurrence risk.

One of the key applications of AI is in predicting drug response and resistance.⁴²⁰ ML algorithms trained on genomic, transcriptomic, proteomic, and imaging data can identify patterns associated with resistance phenotypes. These models can stratify patients based on their likelihood of treatment benefit or resistance, thus facilitating precision therapy. For instance, ADAM Substudy Luik 2 leverages a digital patient model based on real-world data to use AI for predicting lung cancer treatment outcomes and resistance evolution, thereby optimizing clinical decision-making (NCT05783024). DL-based convolutional neural networks have shown high accuracy in predicting resistance from histopathological images or radiomic signatures.⁴²¹ By providing high-confidence protein conformations in the absence of experimental structures, AlphaFold enables a closed-loop pipeline in which generative AI supports rapid optimization from target modeling to oral candidate compounds, representing a new paradigm for AI-driven drug discovery targeting tumor immune resistance.⁴²² AI also excels in multimodal data integration.⁴²³ By simultaneously processing large-scale omics, spatial, and clinical data, AI models can capture the complex interplay between genetic alterations, epigenetic states, microenvironmental cues, and therapeutic responses.^423,424 Integrative models help identify combinatorial biomarkers and uncover latent resistance pathways that might be overlooked in single-omics analyses.⁴²³ In addition, AI is being leveraged to optimize therapeutic strategies.⁴²⁵ Reinforcement learning and digital twin models can simulate tumor evolution and therapy adaptation, enabling the design of dynamic and individualized treatment regimens. AI can also assist in drug repurposing and virtual screening of potential resistance-overcoming agents by mining pharmacogenomic databases.⁴²⁶

In contrast to AI/ML approaches that derive statistical correlations from data, mechanistic modeling builds mathematical representations based on underlying biological processes to explain and predict drug resistance.⁴²⁷ Kholodenko et al. developed MRA to infer causal signaling network topologies from perturbation data, revealing that resistance often arises from network architecture and kinase dimerization rather than individual feedback loops.^276,427 Quantitative systems pharmacology (QSP) models integrate drug mechanisms, pathway dynamics, and patient factors to predict combination therapy outcomes and have progressed to informing clinical trial design.⁴²⁸ The early mechanistic models of MAPK pathway dynamics by Doug Lauffenburger quantified how negative feedback controls signal adaptation, and his concept of temporal collateral sensitivity provides a theoretical basis for adaptive therapy.⁴²⁹ Importantly, several mechanism‑informed models have now entered clinical validation, guiding trial design for targeted therapies in non‑small cell lung cancer and other malignancies.^48,430

However, challenges remain in terms of interpretability, model generalizability across cohorts, and the need for high-quality annotated data.^421,423 Interdisciplinary collaboration among clinicians, biologists, and data scientists is crucial to refine these models and translate AI-driven insights into clinical interventions. In summary, AI offers transformative potential in predicting resistance, uncovering hidden vulnerabilities, and personalizing therapeutic decisions, thus serving as a critical enabler in the era of data-driven oncology.

Clinical implications and biomarker development

Liquid biopsy for dynamic monitoring: ctDNA, circulating tumor cells (CTCs), and exosomes

Liquid biopsy, through the analysis of ctDNA, CTCs, and extracellular vesicles (EVs) in peripheral blood or other body fluids, provides a minimally invasive and real-time approach for early cancer detection, treatment monitoring, and resistance mechanism identification (Fig. 6b).^14,431,432 Changes in ctDNA levels often precede radiographic assessments and offer more precise reflections of molecular responses. ctDNA-based monitoring of clonal evolution can provide earlier warning of ovarian cancer recurrence than imaging modalities and offers critical evidence for the timely adjustment of precision treatment strategies.⁴³³ This is especially valuable for differentiating pseudoprogression from true progression during immunotherapy, potentially prompting revisions to existing RECIST criteria.^434,435

CTCs have demonstrated robust clinical utility in prognostication, surveillance for recurrence and metastasis, treatment response evaluation, and even early diagnosis.⁴³⁶ In a longitudinal study of 135 patients with hormone receptor–positive metastatic breast cancer, CTC counts ≥5 were significantly associated with shorter progression-free survival (PFS) and overall survival (OS); moreover, dynamic changes in CTC burden served as predictors of disease progression.⁴³⁷ Liquid biopsy also facilitates the early detection of resistance-driving subclones and emergent mutations in key oncogenes, enabling timely therapeutic adaptation.¹⁵ Phenotypic and molecular profiling of CTCs has uncovered significant heterogeneity and resistance traits. For instance, CTC-derived organoids from breast cancer patients exhibited elevated FGFR1 expression in advanced disease, promoting resistance via the NRG1–HER3 signaling axis.⁴³⁸ Similarly, sequencing of estrogen pathway-related genes in CTCs revealed resistance mutations distinct from those in the primary tumor, underscoring their value in guiding endocrine therapy.⁴³⁷ In cisplatin-sensitive oral squamous cell carcinoma, the autophagy-associated protein LC3B-II was found to be enriched in EVs, potentially modulating drug sensitivity in neighboring cells via miRNA transfer, highlighting its promise as a noninvasive biomarker for cisplatin resistance.⁴³⁹

Liquid biopsy provides a dynamic window for monitoring tumor resistance, but its clinical application faces four major bottlenecks.¹⁴ First, there is an inherent trade-off between sensitivity and specificity. Low-frequency resistance mutations are difficult to detect accurately, while background interference, such as clonal hematopoiesis, can readily lead to false-positive results.⁴⁴⁰ Second, tracing the origins of resistance mechanisms remains challenging.¹⁴ Current assays have limited capacity to distinguish among distinct resistance pathways, such as target mutations, bypass pathway activation, or phenotypic transitions. Third, clinical validation remains insufficient. The survival benefit of adjusting treatment strategies based on early warnings from liquid biopsy still requires confirmation in prospective studies. Finally, a standardized framework has yet to be established.⁴⁴¹ The lack of uniform standards for analytical methods and result interpretation hampers cross-platform comparisons and large-scale clinical implementation.

Altogether, liquid biopsy enables longitudinal and dynamic tumor surveillance, empowering early recognition of therapeutic response and resistance evolution while advancing personalized oncology.⁴⁴² Future progress will require coordinated breakthroughs in high-specificity clonal tracking, multiomics–based mechanistic analysis, prospective clinical validation, and standardized workflow development for liquid biopsy to become a key tool in guiding strategies to overcome therapeutic resistance.⁴²³

Predictive biomarkers and patient stratification

In the precision medicine era, predictive biomarkers are indispensable for optimizing therapy selection, improving efficacy, and minimizing treatment-related toxicity (Table 4). These biomarkers elucidate underlying resistance mechanisms by identifying resistance-prone subpopulations, metabolic adaptations, and key microenvironmental cues.^313,443 Clinically, they help stratify patients who are more likely to benefit from specific treatments, facilitating rational therapeutic allocation. For example, ALK fusions in NSCLC represent actionable alterations, with multiple ALK inhibitors demonstrating efficacy in ALK-positive tumors.⁴⁴⁴ T-cell exhaustion has also emerged as a cross-cutting resistance signature that compromises the response to ICIs across multiple tumor types.^445,446 Beyond static prediction, predictive biomarkers offer dynamic insights during treatment. Liquid biopsy enables real-time, noninvasive monitoring of therapeutic response and resistance evolution. Stratifying patients by molecular features not only enables resource optimization and efficacy maximization but also reduces overtreatment-related toxicity.⁴⁴⁷

Table 4.

Molecular classes of cancer therapy resistance and associated biomarkers, therapeutic strategies, and clinical development status

Drug resistance mechanisms	Biomarkers	Representative interventions	Clinical development strategies
Drug transport and metabolism	ABCB1 (P-gp), ABCC1 (MRP1), ABCG2 overexpression, CYP450 or UGT enzyme upregulation	P-gp inhibitors, nanoparticle delivery, Standard chemotherapeutics that are P-gp substrates	Avoidance of transporter substrates and rational combination with efflux-modulating agents
DDR	BRCA1/2, PALB2, ATM, ATR mutations or epigenetic silencing, MSI/MMR status	PARPis, ATR/CHK1 inhibitors	Biomarker-enriched and synthetic-lethality–driven trial designs
Altered signaling pathways	PI3K/AKT, MAPK, Wnt/β-catenin activation	Targeted kinase inhibitors, PI3K/mTOR inhibitors	Vertical or horizontal pathway blockade strategies
Epigenetic reprogramming	DNA methylation, Histone modification, Chromatin regulator mutations	DNMTi, HDAC inhibitors, EZH2 inhibitors	Epigenetic priming–based trial designs
Cell state plasticity	EMT/ stemness markers, DTP state	AXL inhibitors, Wnt/TGF-β/Notch inhibitors	Adaptive and lineage-informed trial designs
TME	Immune biomarkers, Myeloid/ fibroblast markers	PD-1/PD-L1 inhibitors, CTLA-4 inhibitors, TME-targeted: anti-VEGF, CSF-1R inhibitors, TGF-β inhibitors	Microenvironment-remodeling combination strategies
Microbiome	Gut microbiota, Tumor-associated bacteria	FMT, probiotics, antibiotics adjustment	Microbiome-stratified or intervention-based trial designs
Cell death mechanism	Ferroptosis regulators, Antiapoptotic proteins	BCL-2/BCL-XL inhibitors, erastin, RSL3	Cell-death–restoration–oriented combination trials

P-gp P-glycoprotein, MRP MDR-associated protein, CYP450 cytochrome P450, DDR DNA damage repair, ATM ataxia telangiectasia mutated, ATR ATM and Rad3-related, MSI microsatellite instability, MMR mismatch repair, PARPis poly (ADP-ribose) polymerase, DNMTis DNA methyltransferase inhibitors, EMT epithelial-mesenchymal transition, DTP drug-tolerant persister, TME Tumor microenvironment, FMT fecal microbiota transplantation

Predictive biomarkers prospectively assess the risk of primary resistance to therapy by identifying specific molecular features prior to treatment, thereby guiding precise patient stratification and treatment selection in clinical trials.³⁸ In the CheckMate 915 study, synergistic integration of dynamic ctDNA monitoring with static analysis of tumor immune microenvironment biomarkers enabled early and accurate prediction of resistance risk to adjuvant immunotherapy in melanoma.⁴⁴⁸ In another study, a composite biomarker model incorporating PD-L1 expression, CD4 gene expression, TMB, and TGF-β–related features successfully predicted responses to nivolumab in patients with urothelial carcinoma, allowing precise identification of patient subgroups most likely to benefit.⁴⁴⁹ Nevertheless, the clinical implementation of predictive biomarkers faces challenges, including assay standardization, sampling heterogeneity, detection sensitivity/specificity, and incomplete mechanistic understanding.¹² Future efforts should prioritize clinical validation, standardized testing, and integration with multiomics and AI approaches to enhance precision and feasibility.

Personalized treatment and drug rechallenge

Therapeutic resistance continues to hinder durable cancer control, especially in the context of targeted therapies and immunotherapies (Fig. 6b). In response, personalized treatment and drug rechallenge strategies have emerged as adaptive precision approaches to address resistance dynamics.⁴⁵⁰

Personalized treatment leverages comprehensive profiling of tumor genomics, mutational landscapes, immune status, and evolutionary trajectories to tailor therapy selection and minimize resistance.⁴¹⁸ Rechallenge strategies, by contrast, operate under the premise of clonal evolution reversibility, whereby reintroducing previously used agents, often after a drug holiday, can resensitize tumors due to regrowth of drug-sensitive clones.⁴⁵¹ For example, in advanced soft tissue sarcoma, rechallenging with anlotinib following resistance, particularly in combination with chemotherapy, has shown promise in sustaining VEGFR pathway inhibition and disease control, especially in non-ASPS subtypes that are more chemotherapy-responsive.⁴⁵⁰ EGFR-TKI rechallenge restores sensitivity to targeted therapy by disrupting the dominance of resistant clones through chemotherapy-mediated clonal selection and population remodeling.⁴⁵²

These strategies depend heavily on dynamic resistance monitoring via liquid biopsy, coupled with AI-assisted data interpretation, integrated multiomics analysis, and functional testing using patient-derived organoids. Together, these approaches represent a paradigm shift toward dynamic, long-term, and evolution-informed cancer management.

Innovative trial designs

To overcome the limitations of traditional therapeutic development in the context of drug resistance, innovative clinical trial designs are being increasingly adopted (Fig. 6b).⁴⁵³ Novel agents, including antibodies, ADCs, DDR inhibitors, epigenetic therapies, PROTACs, and nanomedicine delivery platforms, offer new opportunities for resistance-targeting interventions.^453,454 Conventional randomized controlled trials (RCTs), while the gold standard, are often inefficient in studying rare mutations or highly personalized interventions. Thus, master protocol designs, such as umbrella trials, basket trials, and adaptive platform trials, are gaining prominence.⁴⁵⁵

Umbrella trials test multiple targeted agents in a single cancer type, stratified by molecular subtypes.⁴⁵⁶ The I-SPY 2 umbrella trial precisely matches multiple novel agents or combination therapies to breast cancer molecular subtypes, enabling rapid assessment of early efficacy signals and adaptive optimization, thereby accelerating the development of precision treatments for high-risk breast cancer patients (NCT01042379). In contrast, basket trials enroll patients with shared oncogenic alterations across various tumor types, increasing efficiency and molecular precision.⁴⁵⁷ The basket trial of KY100001K uses IDH mutation as the core enrollment criterion, overcoming tumor tissue-of-origin constraints to efficiently evaluate its efficacy and safety across multiple cancer types, thereby accelerating the precise development of rare molecular subtypes and providing a rationale for pancancer applications.⁴⁵⁸ Adaptive platform trials enable flexible modifications, such as adjusting sample size, endpoints, or treatment arms, based on interim data, thereby optimizing resource allocation and accelerating trial progression.^456,459 The STAMPEDE platform trial employs a multiarm, parallel, adaptive design with a unified master protocol to efficiently evaluate multiple prostate cancer treatment strategies, accelerating clinical evidence generation and optimizing precision therapy for high-risk patients.⁴⁶⁰ These innovative designs not only enhance clinical trial efficiency but also better reflect the evolving nature of resistance and individualized therapeutic responses, ultimately facilitating the transition of resistance-targeting strategies from bench to bedside.

Conclusions and future perspectives

Therapeutic resistance remains one of the major obstacles in cancer treatment, limiting the long-term efficacy of standard and targeted therapies.¹ As summarized in this review, resistance is not solely dictated by tumor-intrinsic genetic alterations but is shaped by a dynamic interplay among clonal evolution, signaling rewiring, epigenetic remodeling, altered DNA damage repair, metabolic adaptation, dysregulated cell death, TCP, microenvironmental remodeling, and host–microbe–tumor interactions. Tumor heterogeneity is a fundamental hallmark of cancer and a central driver of this process.⁹ The diversity of tumor cells at the genetic, transcriptomic, epigenetic, and phenotypic levels gives rise to intratumoral heterogeneity, which underlies the wide variability in treatment responses. Acquired resistance often originates from preexisting heterogeneity within the tumor and is further shaped by continuous diversification during therapy, enabling certain subclones to survive therapeutic pressure and eventually emerge as dominate resistant phenotypes.⁴⁶¹ Primary early-stage tumors and advanced metastatic lesions also exhibit marked differences in tumor heterogeneity, drug resistance, and TME composition.^399,462 A pancancer whole-genome sequencing (WGS) study revealed that metastatic tumors frequently evolve from dominant subclones in the primary lesion, displaying higher clonality and lower intratumoral heterogeneity.⁴⁶³ These observations underscore the need to understand resistance not as a single molecular event but as an evolving, systems-level property of cancer.

Beyond genetic diversity, accumulating evidence indicates that therapeutic resistance is increasingly driven by non-genetic adaptive processes. Phenotypic plasticity, including EMT, lineage switching, CSC reprogramming, and DTP states, enables tumor cells to survive therapeutic pressure in the absence of stable genetic alterations.^308,309 These transient adaptive states may subsequently serve as reservoirs from which genetically resistant clones emerge. Importantly, such plasticity is tightly regulated by epigenetic remodeling, metabolic adaptation, stress-response programs, and microenvironmental cues, highlighting the dynamic and reversible nature of resistance.^308,316 To address this evolutionary challenge, ecology- and evolution-informed treatment paradigms have been proposed. For example, adaptive therapy dynamically adjusts drug dosing to maintain a population of drug-sensitive cells, thereby constraining the competitive expansion of resistant clones.⁴⁶¹ These model-driven, personalized dosing strategies aim to match the evolving tumor landscape in real time. Future therapeutic strategies should therefore move beyond targeting individual mutations and instead focus on disrupting the adaptive networks that sustain tumor evolution, cellular state transitions, and minimal residual disease.

Another major challenge lies in the increasing recognition that resistance is not solely a tumor cell–intrinsic phenomenon. CAFs, TAMs, MDSCs, exhausted T cells, extracellular matrix remodeling, and microbial communities collectively shape a permissive ecosystem that protects tumor cells from therapeutic eradication.^135,136 The TME promotes resistance by restricting drug penetration, suppressing antitumor immunity, providing survival cytokines, maintaining CSC niches, and reinforcing immune-excluded or immune-cold states.^123,135 Similarly, the intratumoral and gut microbiome can modulate therapeutic outcomes by altering drug metabolism, reshaping host immunity, producing immunomodulatory metabolites, and activating oncogenic signaling pathways.^324,352 These findings highlight the importance of viewing resistance through an ecological lens, in which tumor cells, stromal cells, immune cells, metabolites, extracellular matrix components, and microbes form spatially organized and functionally interconnected resistance niches. Advances in spatial transcriptomics, spatial proteomics, metabolomics, and multimodal imaging have begun to reveal the complex organization of these niches. However, translating these discoveries into clinically actionable interventions remains difficult because tumor ecosystems are highly dynamic, context-dependent, and spatially heterogeneous. Developing strategies that simultaneously target tumor cells and their supportive microenvironment will therefore represent an important direction for future research.

Despite significant advances, several challenges continue to limit the clinical translation of resistance-reversal strategies. First, the spatial and temporal complexity of resistance remains incompletely captured by current sampling strategies. Single biopsies often fail to reflect the full spectrum of resistant subclones across metastatic sites and treatment stages. Emerging technologies such as liquid biopsy, next-generation sequencing (NGS), and single-cell and spatial sequencing have enabled dynamic and comprehensive monitoring of tumor evolution and heterogeneity.^12,13,37 However, current single-cell technologie still struggle to fully capture the native tissue architecture, while the sensitivity and specificity of liquid biopsies require further optimization.³⁷⁰ Second, suitable experimental models remain limited. Traditional cell lines cannot adequately recapitulate intratumoral heterogeneity, whereas patient-derived organoids, xenografts, and ex vivo co-culture systems only partially reproduce immune, stromal, vascular, and microbial interactions. Moreover, the lack of robust animal models limits the in vivo simulation of long-term tumor evolution. What’s more, many biomarkers associated with resistance have not yet been sufficiently validated for routine clinical use, and assay standardization, sampling heterogeneity, and data interpretation remain unresolved. Translating laboratory-discovered resistance-reversal strategies into clinical practice faces multiple barriers, including biological complexity, economic constraints, data integration challenges, ethical concerns, and unequal access to advanced diagnostics and personalized therapies.^464,465 Overcoming these limitations will require multidisciplinary collaboration, prospective validation, standardized analytical workflows, and long-term real-world data collection.

Looking forward, the integration of longitudinal monitoring, multiomic profiling, patient-derived models, and computational analytics is expected to reshape resistance research and precision oncology. Multitimepoint sampling and liquid biopsy can enable early detection of temporal resistance signals and facilitate timely treatment adjustments. Next-generation sequencing, single-cell omics, spatial sequencing, proteomics, metabolomics, and microbiome profiling may together provide a more comprehensive view of resistance evolution. Future research should focus on integrating these technologies with computational modeling to identify upstream regulators, convergent resistance mechanisms, and actionable therapeutic vulnerabilities.³⁷⁰ Recently, digital twin frameworks have been proposed that integrate patient-specific molecular data with multiphysics, multiscale mathematical models to simulate individualized drug responses.⁴⁶⁶ These computational platforms can incorporate genomic, signaling, imaging, and cellular dynamic data to support adaptive precision oncology. The rapid development of AI, machine learning, and systems biology approaches is expected to accelerate the transition from descriptive resistance profiling to predictive resistance modeling.⁴²⁴ By integrating longitudinal molecular data, clinical characteristics, treatment history, and imaging information, AI-driven frameworks may help predict resistance trajectories, identify patient-specific vulnerabilities, and optimize treatment sequencing. In parallel, emerging patient-derived organoid, xenograft, and ex vivo immune co-culture platforms provide increasingly realistic experimental systems for validating resistance mechanisms and testing personalized therapeutic strategies.^16,17 The combination of these technologies may ultimately facilitate the development of truly adaptive and individualized treatment paradigms.

Future clinical development should increasingly emphasize biomarker-guided treatment selection, longitudinal response monitoring, and adaptive therapeutic interventions. Innovative clinical trial designs, including basket, umbrella, platform, and adaptive trials, may accelerate the evaluation of personalized treatment strategies while accommodating the biological complexity of resistance evolution. Rational combination therapies that simultaneously address multiple resistance pathways are likely to become increasingly important, particularly those integrating targeted agents, epigenetic modulators, metabolic inhibitors, immunotherapies, microbiome-directed interventions, and TME-reprogramming strategies. Furthermore, integration of liquid biopsy technologies with real-time molecular monitoring may enable earlier detection of resistance emergence and facilitate proactive therapeutic adjustments before overt clinical progression occurs.¹⁵ Combating therapeutic resistance thus requires an integrative, precision-driven approach that includes novel drug combinations, adaptive treatment regimens, microenvironmental reprogramming and rational modulation of host–microbe–tumor crosstalk. As cancer treatment paradigms shift toward personalization and immunomodulation, the integration of molecular, cellular and ecological perspectives will be critical to overcoming resistance and improving durable responses in diverse cancer types. Ultimately, the future of cancer therapy will depend on a paradigm shift from reactive management of established resistance to proactive prediction, monitoring, and prevention of resistance development.

Author contributions

C.C., M.X. and X.C. conceived the whole article and completed the original draft preparation. J.W. and C.C. performed major reviews and edited the structure of the article. The primary drawing effort was performed by C.C., M.X. and X.C. During subsequent revisions of the article, J.W. and C.C. were primarily responsible for refining the manuscript’s structure and enhancing the clarity of the content. All authors reviewed and approved the manuscript.

Data availability

Not applicable.

Competing interests

The authors declare no competing interests.