This mechanistic integrative review was compiled through systematic searches of the PubMed and Web of Science databases for articles published between 2010 and 2026. The search terms covered three thematic domains: mechanical stress (e.g., “shear stress”, “matrix stiffness”, “compressive stress”, “mechanotransduction”); cytoskeleton and exosomes (e.g., “cytoskeleton”, “actin”, “microtubules”, “exosomes”, “multivesicular bodies”, “extracellular vesicles”); and autophagy/drug resistance (e.g., “autophagy”, “chemoresistance”, “drug resistance”, “tumor microenvironment”). Only peer-reviewed original research articles, reviews, and clinical studies published in English were considered. Additional relevant publications were identified through manual screening of reference lists of retrieved articles and key reviews. Priority was given to studies that provided mechanistic evidence - defined as those employing defined mechanical stimuli (e.g., tunable stiffness substrates, microfluidic shear devices, or compression setups), functional perturbation of cytoskeletal or exosomal components (e.g., pharmacological inhibition, siRNA/CRISPR-mediated knockdown), and quantitative readouts of drug sensitivity or autophagy flux. We also included selected studies on exosomal cargo and autophagy signaling that have not yet been directly linked to mechanical stress, as these were used to frame testable hypotheses and identify knowledge gaps. When multiple examples supported a similar mechanistic principle, we prioritized those with the strongest experimental evidence and clinical relevance rather than providing exhaustive listings.

Shear stress, a mechanical force that is generated by fluid flow, is prevalent in the TME and primarily originates from blood flow and interstitial fluid flow that are driven by abnormal vascular leakage^[9]. Owing to the chaotic architecture of tumor vessels, the stress distributions are highly heterogeneous, with frequent turbulence and eddies - in stark contrast to the stable laminar flow that is observed in normal tissues [Figure 1A]. This abnormal mechanical landscape not only influences tumor cell migration and aggregation but also imposes a selective pressure that drives the acquisition of malignant traits, among which drug resistance has emerged as a critical biological consequence.

In high-shear regions - such as tumor margins and vascular stenosis - mechanosensitive receptors (e.g., PIEZO1 and integrins) are activated^[10], triggering pro-survival signaling cascades [e.g., phosphoinositide 3-kinase (PI3K)/protein kinase B (Akt) and mitogen-activated protein kinase (MAPK)/extracellular signal-regulated kinase (ERK)]^[11,12]. These pathways consistently upregulate multidrug resistance-associated transporters, conferring resistant phenotypes and facilitating the spread of resistance across cell populations. Although the upstream receptors and downstream effectors may vary across contexts - for instance, PIEZO1 activation in ovarian cancer vs. its downregulation in breast cancer^[12,13] or epidermal growth factor receptor (EGFR)/mitogen-activated protein kinase kinase (MEK)/ERK activation in ovarian cancer and exosome-mediated insulin-like growth factor 2 (IGF2)/rat sarcoma virus oncogene (Ras)/rapidly accelerated fibrosarcoma kinase (Raf)/ERK activation in hepatocellular carcinoma^[14,15] - the net outcomes remain the same: enhanced survival and drug efflux. Collectively, these examples illustrate that high shear stress levels drive drug resistance primarily through the rapid activation of survival signals and efflux pumps, regardless of the specific receptor or pathway involved.

Conversely, low-shear regions, such as the tumor core and areas of stagnant flow, promote resistance through chronic adaptive responses. Mechanisms include the stabilization of caveolin-1 (Cav-1) to increase antiapoptotic capacity (e.g., in triple-negative breast cancer)^[16], the accumulation of autophagosomes and the release of autophagic extracellular vesicles, and the hypoxia-inducible factor 1-alpha (HIF-1α)-mediated upregulation of autophagy under nutrient deprivation^[17,18]. While the molecular details differ, these processes collectively foster a quiescent, tolerant state that shields tumor cells from chemotherapeutic insult. Thus, low shear stress consolidates resistance mainly via autophagy, metabolic adaptation, and apoptosis blockade, representing a distinct but equally potent route compared with high shear stress responses.

Notably, the molecular responses to high- and low-shear regions are nearly opposite - acute survival signaling vs. chronic autophagic quiescence - yet they converge on the same resistant phenotype. This suggests a biphasic or nonlinear regulatory mode, where different stress intensities achieve similar outcomes through entirely distinct mechanisms. Such a duality complicates therapeutic interventions: targeting a single mechanosensor, such as PIEZO1, in high-shear areas might inadvertently amplify autophagic tolerance in low-shear zones, rendering systemic blockade ineffective. Therefore, a comprehensive understanding of spatial mechanical heterogeneity is essential for designing context-specific strategies to overcome shear-induced drug resistance.

Extracellular matrix (ECM) stiffening is a hallmark of solid tumors, arising primarily from the accumulation and crosslinking of collagen fibers, a process that is driven by cancer-associated fibroblasts (CAFs) through the secretion of transforming growth factor-beta (TGF-β) and lysyl oxidase (LOX) [Figure 1B]^[19]. Importantly, elevated matrix stiffness is not merely a passive byproduct of tumor growth but also an active mechanical driver of drug resistance^[20]. Through mechanobiological coupling, stiffness promotes drug resistance through three interconnected routes: direct survival signaling, microenvironment remodeling, and intercellular communication.

At the cellular level, stiffness is sensed by integrins, which trigger cytoskeletal reorganization and activate a range of pro-survival pathways. Common downstream effectors include the PI3K/Akt, nuclear factor erythroid 2-related factor 2 (NRF2), and HIF-1α axes, which collectively increase cell survival and induce protective autophagy, allowing tumor cells to withstand chemotherapy-induced damage^[21,22]. Although the specific pathways involved vary across cancer types - e.g., integrin beta-1 (ITGB1)/PI3K/Akt in hepatocellular carcinoma and NRF2 in pancreatic cancer - the net effect is a consolidated drug-resistant phenotype. Moreover, stiffness promotes the intercellular spread of resistance through exosomes: in stiff microenvironments, increased exosome release [via focal adhesion kinase (FAK)-Akt-Rab8 in hepatocellular carcinoma] and the sorting of prometastatic miRNAs (via integrin-FAK-F-actin in prostate cancer) enable the transmission of mechanical information and resistance traits to distant recipient cells^[23,24]. Collectively, these findings indicate that the matrix stiffness directly reinforces intrinsic tumor cell survival and facilitates the horizontal dissemination of resistance, largely through cytoskeleton-dependent signaling and vesicle-mediated communication.

In addition to directly affecting tumor cells, stiffness also remodels the mechanical stress and immune microenvironments to indirectly promote drug resistance. Excessive ECM deposition generates a dense physical barrier that compresses tumor blood vessels and severely limits drug penetration^[25,26]. Pharmacological softening of the ECM - e.g., with losartan or LOX inhibitors - has been shown to improve vascular function and enhance the delivery of chemotherapeutics and immune cells in ovarian and cholangiocarcinoma models^[27,28]. Simultaneously, stiffness modulates the immune landscape by creating an immunosuppressive niche; for example, in triple-negative breast cancer, chemotherapy-induced collagen deposition results in the recruitment of immunosuppressive cells while excluding CD8⁺ T cells, thereby decreasing the efficacy of both oxaliplatin and programmed death-ligand 1 (PD-L1) blockade^[29]. Softening the ECM can reverse this immunosuppression by restoring the sensitivity to combined chemo-immunotherapy. Thus, the indirect route of stiffness-induced resistance operates primarily through the physical hindrance of drug delivery and immune exclusion, both of which can be partially alleviated by matrix-targeting agents.

Clinically, targeting matrix stiffness presents a fundamental paradox: softening the ECM may improve drug penetration but could also weaken the physical barrier that constrains tumor invasion. The failure of the Phase III HALO-301 trial of PEGPH20 suggests that simply reducing stiffness is insufficient to reverse established resistance^[30]. We speculate that this may be due to stiffnessinduced mechanical signals becoming “locked in” through epigenetic or metabolic reprogramming, a possibility consistent with the concept of mechanical memory in tumor cells. Therefore, matrix-softening strategies may be more effective when applied earlier in the disease course, before the drug-resistant phenotype is fully established, rather than as a late-stage intervention. A combined approach that integrates ECM modulation with conventional chemotherapy or immunotherapy, timed to the appropriate window, may hold great promise for overcoming this form of resistance.

In solid tumors, excessive ECM deposition and cross-linking not only increase tumor volumes and compress surrounding tissues - generating solid stress - but also increase interstitial fluid pressures because of blood and lymphatic vessel compression, which together produce compressive stress^[31]. This stress exhibits a “center-edge” gradient: central regions can reach pressures of 30-50 kPa, transitional zones 10-20 kPa, while normal tissues remain below 5 kPa [Figure 1C]^[32]. This spatial gradient creates intratumoral heterogeneity in drug resistance, with the cells in the high-stress core displaying greater tolerance to therapy. For instance, compressive stress inhibits proliferation and induces quiescence in tumor cells, thereby reducing the efficacy of cell cycle-specific agents (e.g., gemcitabine and docetaxel) - a broad-spectrum, mutation-independent mechanism that is particularly prominent in fibrotic tumors such as pancreatic ductal adenocarcinoma^[33].

At the molecular level, compressive stress is sensed by mechanosensory complexes at the plasma membrane. PIEZO1 channels undergo conformational changes to mediate Ca²⁺ influx, whereas integrin-ECM engagement activates downstream signaling^[34]. These inputs converge on the Ras homolog family member (RHO)-Rho-associated coiled-coil-containing protein kinase (ROCK) axis to increase cytoskeletal contractility and on the AMP-activated protein kinase (AMPK)-mammalian target of rapamycin (mTOR) pathway to induce autophagy^[35,36], which in this context functions to clear drug-induced damage and sustain cellular energetics. Concurrently, the upregulation of multidrug resistance proteins [e.g., multidrug resistance-associated protein 1 (MRP1)] accelerates drug efflux and suppresses apoptosis. In addition to these cell-autonomous effects, compressive stress also promotes the transmission of intercellular resistance. Yes-associated protein (YAP) nuclear translocation drives the secretion of exosomes that are enriched with resistance-related miRNAs and functional proteins [e.g., Annexin A3 (ANXA3)]^[37], while mitochondrial damage enables exosomal loading of mtDNA damage signals, which reprogram recipient cells via paracrine signaling^[38-40]. In breast cancer, compressive stress at the tumor-stroma interface induces YAP-dependent CD44/CD133 upregulation, conferring resistance to doxorubicin and tamoxifen^[41]. Additionally, a HIF-1α-ECM positive feedback loop further amplifies tumor cell stiffness, invasiveness, and stemness, exacerbating the resistant phenotype. Collectively, these findings demonstrate that compressive stress drives drug resistance through three core mechanisms, namely, direct activation of survival/autophagy pathways, upregulation of drug efflux pumps, and exosome-mediated propagation of resistance traits - which parallels the routes engaged by other mechanical stresses.

Crucially, the spatial interplay among the three stress types - compressive stress concentrated in the core, interstitial fluid pressure distributed internally, and matrix stiffness pervading the entire tumor - means that cells in different regions experience fundamentally distinct mechanical inputs. However, no study to date has investigated how this spatial combination shapes the heterogeneity of intratumoral drug resistance. The following questions remain: do the cells in each region independently adapt to their local mechanical environment, or are mechanical signals transmitted across regions via carriers such as exosomes? If the latter holds true, drug resistance would not simply be the sum of independent regional events but rather a population-level process orchestrated by mechanical gradients. Thus, while the three stress types differ in their physical forms and spatial distributions, they converge on common downstream pathways - efflux pump activation, exosome communication, and protective autophagy - to jointly initiate and sustain drug resistance. However, the precise intracellular mechanisms by which divergent mechanical inputs are integrated into unified intercellular signaling instructions remain incompletely understood. Key research cases are summarized in Table 1.

The preceding sections have established that all three mechanical stresses, despite their distinct physical forms and spatial distributions, converge on common downstream mediators - most notably, the cytoskeleton and exosome pathways. This convergence naturally raises a critical question: how does the cytoskeleton, as the central hub of mechanotransduction, translate these diverse mechanical inputs into coordinated regulation of exosome biogenesis, transport, release, and uptake? In the next section, we address this question by systematically dissecting the role of cytoskeletal dynamics in governing the entire exosome lifecycle under mechanical regulation.

Exosome biogenesis begins with the invagination of the plasma membrane, which captures cell surface and extracellular proteins to form early-stage sorting endosomes (ESEs). The trans-Golgi network (TGN) and endoplasmic reticulum can also contribute to ESE formation and the cargo supply. ESEs migrate toward the perinuclear region and mature into late sorting endosomes (LSEs), which subsequently generate multivesicular bodies (MVBs) through inward budding of the endosomal membrane, producing intraluminal vesicles (ILVs, ~40-160 nm) - the exosome precursors. MVBs then either fuse with lysosomes or autophagosomes for degradation or anchor and fuse with the plasma membrane to release ILVs as mature exosomes^[45]. Key molecular players include Ras-related protein (Rab) GTPases; endosomal sorting complexes required for transport (ESCRT) proteins [tumor susceptibility gene 101 protein (TSG101) and ALG-2-interacting protein X (ALIX)]; tetraspanins [cluster of differentiation 63 (CD63), CD81, and CD9)]; ceramide, sphingomyelinase; and soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) complexes. Once released, exosomes diffuse through tissue fluids or the circulation; adhere to target cell surface receptors; and are internalized via endocytosis or membrane fusion, delivering biological information that modulates proliferation, differentiation, metabolism, and stress responses^[46]. This regulation contributes to drug resistance by altering target cell phenotypes - e.g., through the activation of resistance-related molecules (e.g., non-coding RNAs, proteins, and signaling molecules) - or by modulating immune and stromal cell functions in the TME^[47]. In essence, exosomes serve as vehicles for the intercellular transfer of resistance traits, and their production and functions are governed by a tightly regulated vesicle trafficking network. Exosomes are thus key vehicles for the intercellular transfer of resistance traits. However, they are not the only extracellular vesicle (EV) population with functional relevance in this context. The tumor secretome also contains microvesicles, mitochondria-derived vesicles (MDVs), and apoptotic bodies - all of which participate in intercellular communication and contribute to drug resistance. These EV subtypes are functionally informative in their own right. MDVs, for instance, can transfer functional mitochondria or damaged organelles between tumor cells, propagating metabolic reprogramming and resistance phenotypes, while microvesicles enriched in drug efflux pumps similarly influence recipient cell sensitivity. We therefore include selected examples of these broader EV populations where they illuminate mechanistic principles - such as cargo transfer or metabolic regulation - that parallel or extend the exosome biology central to this review. Where the cited studies do not definitively establish endosomal origin, we refer to them as EVs rather than exosomes to maintain terminological clarity. Throughout this review, we reserve “exosomes” for endosome-derived vesicles and use the broader term “EVs” when the endosomal origin of a vesicle cannot be confirmed from the cited literature.

The cytoskeleton, which consists of microtubules, microfilaments, and intermediate filaments, undergoes reorganization that is driven by motor proteins (e.g., dynein, kinesin, and myosin), providing mechanical support and energy for the entire exosome life cycle. Dysregulation of this process can amplify drug resistance signals through exosomal pathways, accelerating resistance development.

As detailed above, mechanical stress regulates exosome biogenesis and cargo sorting at multiple levels through cytoskeletal remodeling. Previous studies have confirmed that tumor cells, via exosome biogenesis and secretion, directly reduce effective intracellular drug concentrations and eliminate drug-sensitizing molecules (cell-autonomous resistance) while promoting the dissemination of resistant phenotypes to entire cell populations via paracrine signaling^[82].

Within this network, autophagy occupies a unique nodal position. Unlike linear pathways, autophagy is a homeostatic system that integrates multiple stress signals. Notably, other TME stressors, namely, chemotherapy, hypoxia, and nutrient deprivation, also activate autophagy, making it a critical bridge connecting microenvironmental cues to cellular adaptation^[83]. Autophagy's functional output is pleiotropic: it directly protects tumor cells by clearing damaged organelles and reactive oxygen species (ROS), synergizes with drug efflux by regulating the membrane localization of multidrug resistance proteins, and inhibits apoptosis through the degradation of apoptotic regulators [Figure 3]. Moreover, autophagy and exosome pathways are tightly intertwined: core autophagy proteins directly regulate exosome biogenesis via noncanonical pathways; exosomes transport autophagy-related molecules to transmit autophagic signals between cells; and autophagy influences MVB fate through amphisome formation - induction of autophagy promotes MVB degradation (reducing exosome release), while inhibition may increase secretion^[84,85]. Thus, in the specific context of mechanical stress-driven adaptation - which is the focus of this review - autophagy functions not only as a core executor of drug resistance but also as a signal integrator driving paracrine dissemination. This pro-survival role, however, represents one facet of the broader functional spectrum of autophagy; its tumor-suppressive functions under different contexts (e.g., early tumorigenesis or specific genetic backgrounds) are well documented elsewhere^[86].

Autophagy in tumors is context-dependent, acting as a “double-edged sword” - it is tumor-suppressive in the early stages (clearing damaged organelles, maintaining genomic stability, limiting inflammation) but promotes survival in later stages, particularly under stress (clearing ROS, providing metabolic fuel, and inhibiting apoptosis)^[87]. In the context of drug resistance, autophagy predominantly has protumor and prosurvival functions. This review therefore focuses on the pro-survival role of protective autophagy that is activated by the mechanical stress-exosome axis.

In terms of the scope and dissemination mechanisms, mechanical signal-regulated autophagic resistance can be categorized into two types: (1) cell-autonomous autophagic resistance, in which mechanical signals directly initiate autophagy within stressed cells, enabling tolerance through damage clearance and metabolic reprogramming; and (2) non-cell-autonomous (paracrine) autophagic resistance, in which stressed cells release autophagy-regulatory molecules via exosomes, inducing autophagy in recipient cells and achieving transcellular dissemination. This section systematically elaborates the molecular mechanisms of these two modes.

The mechanical properties of the TME (e.g., matrix stiffness and compressive stress) activate downstream signaling via mechanosensors, promoting drug resistance. Intervening at this level can follow three approaches. First, inhibiting mechanosensor function prevents signal transmission into cells. For example, pharmacological or genetic blockade of integrins disrupts integrin-ECM signaling, inhibiting matrix stiffness-induced autophagy and exosome secretion and thereby reducing therapy resistance^[23,108-110]. Second, modulating ECM remodeling alleviates mechanical burden: degrading excessive ECM components (e.g., via MMP inhibitors or hyaluronidase) not only reduces stress-induced resistance but also enhances drug penetration^[111,112]. Third, targeting downstream transcriptional effectors such as YAP/transcriptional coactivator with PDZ-binding motif (TAZ) - whose stress-induced nuclear translocation drives autophagy-related genes (ATG5 and LC3) and exosome secretion genes (Rab27A) - can simultaneously suppress both autophagy activation and exosome-mediated resistance dissemination, showing promise in colorectal and pancreatic cancer models^[113,114].

The major obstacle to targeting mechanical signals is not a lack of targets but avoiding systemic toxicity. Integrins and PIEZO1 perform essential physiological functions in the cardiovascular, skeletal, and immune systems; systemic inhibition may cause unacceptable side effects. Selective targeting of tumor-enriched subtypes [e.g., Integrin subunit beta-like 1 (ITGBL1)] offers a potential way forward, but such subtype-specific inhibitors remain scarce. ECM-remodeling strategies (e.g., PEGPH20) avoid direct receptor targeting, yet their phase III failure suggests that matrix modulation alone is insufficient to reverse established resistance programs. In essence, single-node mechanical stress blockade has limited efficacy, primarily because of compensatory pathway activation; therefore, it requires combination with downstream exosome and autophagy targets to achieve durable resistance reversal while precisely controlling the intervention intensity to avoid disrupting the tumor-suppressive functions of the stromal barrier.

Cancer cells adapt and reorganize their cytoskeletons to cope with mechanical constraints, while cytoskeletal dynamics support drug resistance through exosome biogenesis/transport and autophagosome formation/trafficking. Intervention strategies can target three levels. First, modulating microtubule and microfilament polymerization kinetics: microtubule stabilizers (e.g., paclitaxel derivatives) increase cytoskeletal stiffness and have been suggested to interfere with extracellular vesicle release, potentially contributing to reduced dissemination of resistance-related signals^[115,116]; ROCK inhibitors disrupt cytoskeletal dynamics, impair autophagosome clearance and contribute to cell death in cancer cells^[117]. Second, targeting cytoskeleton-associated proteins [e.g., tubulin beta-3 chain (TUBB3) and Annexin A2 (ANXA2)], which are essential for exosomal cargo sorting - the use of siRNA or small-molecule inhibitors - disrupts exosome-mediated resistance signal transmission and autophagic flux integrity^[118,119]. Third, interfering with the interaction between the cytoskeleton and autophagosomes/MVBs: Rab GTPases (e.g., Rab7 and Rab11) regulate fusion and cytoskeletal integrity^[120]; inhibiting their activity blocks both autophagic flux and the propagation of exosome-mediated resistance^[67,121,122].

Cytoskeletal-targeted interventions can theoretically block both exosome transport and autophagosome formation, but their clinical translational challenges are often underestimated. ROCK inhibitors, while suppressing tumor cell motility, may paradoxically promote the survival of cancer stem cells and dormant cells - a net effect that is highly dependent on cell type and context^[123,124]. Microtubule-targeting drugs cause rapid resistance through efflux pump upregulation, tubulin mutation, or compensatory protein activation^[125]. Moreover, the effects of cytoskeletal drugs generally suffer from off-target toxicity: the targeting of microtubules interferes with normal mitosis, whereas the targeting of actin may affect immune cell migration and phagocytosis^[126]. These challenges suggest that the functional homogeneity of the cytoskeleton across physiological and pathological processes constrains the therapeutic window of broad-spectrum strategies. We propose a shift from “broad-spectrum blockade” to “precision targeting” - i.e., targeting cytoskeleton-associated regulatory proteins that are aberrantly overexpressed in tumor cells (e.g., TUBB3, ANXA2, and STMN1) or utilizing stress-responsive nanocarriers to specifically deliver drugs to regions with high mechanical stress, thereby disrupting resistance-related transport with minimal systemic toxicity.

Exosomes are critical for disseminating drug-resistant phenotypes along this axis, and their secretion, cargo sorting, and uptake by target cells all offer intervention nodes. First, by inhibiting exosome biogenesis, the nSMase2 inhibitor GW4869 reduces the release of drug resistance related exosomes^[127], reversing CAF-mediated 5-fluorouracil/oxaliplatin resistance in colorectal cancer^[128]. Similarly, disrupting the ESCRT complex function (e.g., TSG101 and ALIX) impairs the packaging of specific cargo into exosomes, including receptors that can modulate chemoresistance^[129-131]. Second, preventing exosome uptake by disrupting lipid rafts with methyl-β-cyclodextrin impairs exosome internalization by target cells, blocking signal transmission^[132]. Third, the direct elimination of circulating exosomes by antibodies or aptamers targeting surface tetraspanins (e.g., CD63 and CD81) enables selective in vivo clearance, reducing TME remodeling and resistance dissemination^[133].

GW4869 is the most widely used exosome inhibitor in research, but its clinical limitations are often underestimated. It blocks exosome biogenesis via nSMase2 inhibition; however, nSMase2 is involved in multiple aspects of sphingolipid metabolism, and the off-target effects of long-term inhibition remain unknown^[134]. More importantly, compensatory mechanisms raise concern - when one secretory pathway is blocked, do cells upregulate alternative routes (e.g., microvesicle release) to maintain signal output? This question remains completely unstudied. In the absence of compensatory safety data, the use of exosome inhibitors in combination chemotherapy carries unforeseeable risks.

Given the pivotal role of autophagy in mechanical stress-exosome-mediated drug resistance, targeting autophagic flux has become a feasible strategy. However, autophagy has a dual effect - providing metabolic adaptation, while excessive activation may trigger autophagic cell death. Thus, regulation should depend on the autophagic activity status of tumor cells. With respect to chemoresistant tumors that are dependent on protective autophagy, the inhibition of autophagic flux is the main strategy: classic inhibitors (e.g., chloroquine and hydroxychloroquine) prevent autophagosome-lysosome fusion, impairing drug degradation and DNA damage repair^[135], and sensitize colorectal and pancreatic cancers to gemcitabine when combined with chemotherapy^[136,137]; silencing ATG5/ATG7 directly blocks autophagosome formation^[138]. For chemoresistant subtypes with low autophagic activity, activating lethal autophagy may be effective: mTOR inhibitors (e.g., rapamycin) or AMPK activators drive excessive autophagy exceeding adaptive capacity, inducing autophagic cell death^[139,140]. Additionally, targeting the autophagy-exosome interface - such as blocking the ATG12-ATG3-ALIX interaction or the LC3-dependent exosome loading (LDELS) pathway - can achieve dual inhibition of both autophagy-mediated survival and the propagation of exosome-mediated resistance^[141,142].

Despite autophagy inhibitors having entered numerous clinical trials, their translation faces several structural challenges. Foremost among these is the inherent functional duality of autophagy itself - it can exert either tumor-suppressive or pro-survival effects depending on tumor type, treatment context, timing, and microenvironmental conditions. However, most clinical trials have not stratified patients on the basis of autophagy-dependent status, indicating that inhibiting autophagy in the context of its tumor-suppressive function could theoretically accelerate rather than retard tumor progression. The sensitizing effects observed preclinically have also not been stably reproduced clinically; in metastatic pancreatic ductal adenocarcinoma, hydroxychloroquine plus MEK inhibition achieved only a 5% disease control rate^[143]; in platinum-sensitive recurrent ovarian cancer and small cell lung cancer, no significant survival benefit was observed^[144,145]. Additionally, whether clinical doses achieve sufficient intratumoral concentrations to fully inhibit autophagy lacks definitive evidence^[146], and tumor cells can upregulate HLTF to repair hydroxychloroquine-induced ROS-mediated DNA damage, establishing resistance to the inhibitors themselves^[147]. Together, these challenges point to a critical conclusion: autophagy should not be viewed as a single target that can be simply “switched on or off” but rather as a dynamic regulatory network that is intimately coupled with tumor context. Future strategies need to shift from broad-spectrum inhibition to precision modulation - selectively targeting aberrantly activated autophagic nodes in tumor cells based on defined autophagy-dependent status rather than simply suppressing autophagy in all cells.

Owing to the extensive cross-regulation within the mechanical stress-cytoskeleton-exosome-autophagy axis, single-node interventions are often insufficient to disrupt the resistance network. Consequently, combination strategies that target multiple pathways are gaining attention. One approach combines ECM-remodeling agents with autophagy inhibitors to first mitigate mechanical stress-induced autophagy activation and then directly inhibit cytoprotective autophagy, potentiating chemotherapy efficacy^[148]. Conceptually, a dual strategy that simultaneously targets cytoskeleton-mediated exosome transport and exosome biogenesis could, in theory, inhibit both the propagation and establishment of resistance signals, although this combinatorial approach remains to be experimentally validated. A triple-combination strategy adds sunitinib and autophagy modulators (e.g., hydroxychloroquine) to standard chemotherapy, showing promising efficacy in pancreatic cancer models^[149]. Another focus is the development of multitarget inhibitors that target key cross-molecules - for example, given that YAP promotes exosome secretion and drug resistance through transcriptional upregulation of RAB27A and other exosomerelated genes, dual inhibition of YAP and Rab27A represents a conceptually attractive strategy that warrants further investigation, though it remains purely theoretical at present^[37].

The elucidation of this regulatory axis provides a mechanics-based mechanistic explanation for the substantial primary resistance that is observed in desmoplastic tumors, such as pancreatic cancer and triple-negative breast cancer - resistance that cannot be readily explained by genetic mutations alone. Multiple core nodes within this axis, namely. autophagic flux, microtubule dynamics, and actomyosin contractility, can be targeted by approved drugs (e.g., hydroxychloroquine, metformin, and microtubule-targeting agents). Through a “drug repurposing” strategy, translational risk can be significantly reduced, and the clinical development cycle can be shortened. Looking forward, combining in vivo imaging of matrix stiffness, exosomal liquid biopsy, and pharmacodynamic biomarkers to enable patient stratification and adaptive therapy based on the activity of this regulatory axis holds promise for overcoming the precision intervention bottleneck in refractory solid tumors and substantially improving patient survival benefits [Table 2].

Future research should prioritize the mechanistic dissection of cytoskeletal remodeling and exosomal cargo sorting, particularly the identification of cell-type-specific regulatory factors that could serve as highly selective therapeutic targets. Achieving this goal will require experimental models that faithfully recapitulate the mechanical landscape of the TME. Organoid models, especially when cultured in tunable-stiffness hydrogels, offer a powerful platform to interrogate how matrix stiffness and compressive stress regulate exosome biogenesis and autophagic resistance in a three-dimensional, patient-derived context^[20]. Microfluidic shear-stress systems complement these models by enabling precise control over fluid dynamics, allowing researchers to dissect how high- and low-shear regions differentially modulate exosome release and intercellular resistance transmission^[14,156]. On the analytical front, single-cell sequencing and spatial transcriptomics will be instrumental in mapping stress drug resistance gradients within intact tumors, revealing how mechanical heterogeneity shapes intratumoral drug resistance at the cellular and regional levels^[157]. Integrating these approaches - mechanically tunable culture systems, microfluidic devices, and spatial multi-omics - will bridge the gap between in vitro mechanotransduction studies and in vivo tumor biology. From a translational standpoint, the development of dual-target or multitarget inhibitors (e.g., YAP/TAZ-Rab27A dual inhibitors) and the validation of predictive biomarkers (e.g., exosomal autophagy-related miRNAs and cytoskeletal protein expression levels) should be prioritized, as both will accelerate the clinical translation of strategies targeting this axis. Ultimately, combining in vivo imaging of matrix stiffness, exosomal liquid biopsy, and pharmacodynamic biomarkers to enable patient stratification and adaptive therapy based on the activity of this regulatory axis holds promise for overcoming the precision intervention bottleneck in refractory solid tumors and substantially improving patient survival benefits.