Nanosystem-Mediated Intercellular Crosstalk Modulation in the Tumor Microenvironment for Enhanced Cancer Immunotherapy

Abstract

Intercellular crosstalk in the tumor microenvironment (TME) is a central determinant of cancer progression and therapy response, in which malignant, immune, stromal and vascular compartments exchange information mainly through extracellular vesicles, signaling pathways, and metabolic cues. These communication circuits can enforce immune exclusion, sustain immunosuppressive cell states and enable adaptive resistance, thereby limiting the durability of otherwise effective therapies. Nanosystems have emerged as enabling tools to modulate this crosstalk across major tumor-centered interaction axes, including tumor cell-immune, tumor cell-stromal, and tumor cell-endothelial interactions, because their tunable physicochemical properties can support context-matched delivery, coordinated co-delivery and controlled exposure profiles that together facilitate multi-node intervention while aiming to reduce off-target perturbation. In this Review, we synthesize key mechanisms of intercellular crosstalk in the TME and summarize nanosystem strategies to reprogram these communication circuits for improved therapy, while outlining translational priorities including causal validation and regimen optimization. Collectively, nanosystem-mediated crosstalk modulation provides a unifying framework that may help shift the TME from a therapy-resistant ecosystem toward a more therapy-permissive state, with the potential to enhance the efficacy and durability of cancer immunotherapy.

Introduction

Over the past decade, oncology has made tangible progress through more precise surgery and radiotherapy, improved systemic regimens, and the expanding use of targeted therapy and immunotherapy. Nevertheless, durable control of advanced disease remains difficult at a population level, as reflected by the persistent global burden with close to 20.0 million new cases and 9.7 million deaths in 2022.1 Clinically, the key obstacle is not simply achieving an initial response, but preventing “escape” through residual disease, relapse, and especially metastatic progression, which is closely linked to cancer mortality.2 Immunotherapy highlights this challenge. Although immune checkpoint blockade can induce long-lasting remissions in selected patients, most patients either fail to respond or eventually develop resistance after an initial benefit.3 These realities underscore the need to target not only tumor-intrinsic vulnerabilities but also microenvironmental determinants of therapy response.

Determinants of therapy response are strongly shaped by the tumor microenvironment (TME), comprising immune and stromal compartments, aberrant vasculature, and extracellular matrix (ECM) architecture that collectively govern therapeutic penetration, immune-cell trafficking, and effector function.4 Mechanistically, intercellular crosstalk is recognized as the layer through which the TME acquires ecosystem-level robustness under therapy pressure. Tumor cells, myeloid and lymphoid populations, fibroblasts, and endothelial cells (ECs) exchange information through coupled paracrine and contact-dependent programs that coordinate recruitment, lineage plasticity, and functional states.5–7 Among lymphoid populations, NK cells contribute to antitumor surveillance through direct cytotoxicity and cytokine secretion, whereas B cells participate in antigen presentation, antibody-mediated responses, and immune regulation.8,9 These communication routes are highly interconnected and often assemble into self-reinforcing feedback circuits. Stromal and vascular programs limit immune infiltration and intratumoral drug exposure. In turn, immunosuppressive myeloid cells and dysfunctional T cells sustain stromal activation and tumor adaptability.10 Accordingly, targeting key communication nodes and circuits, rather than tumor cells alone, provides a principled route to shift the TME toward states that are more permissive for immunotherapy. However, conventional systemic approaches to TME modulation can be constrained by dose-limiting toxicity, suboptimal pharmacokinetics, and insufficient exposure within the relevant tumor compartments.11 These limitations highlight the need for delivery strategies that can enhance local drug exposure and improve access to specific cell populations.

Nanosystems have emerged as a central translational platform in cancer therapy. By packaging small molecules, proteins, or nucleic acids into engineered carriers, or using nanomaterials with intrinsic therapeutic properties, these systems are able to protect the drugs, increase their local concentration, and reduce side effects.12 Crucially, given that crosstalk spans multiple cell types and is stabilized by feedback circuitry, nanosystems offer distinctive advantages for modulating intercellular crosstalk. Nanosystems enable delivery to targeted cells, allowing intervention at network “hubs” while limiting systemic exposure.13 At the same time, they can target communication vehicles and interfaces (eg, extracellular vesicle biogenesis/uptake, cytokine–receptor axes, metabolic transporters, and mechanotransduction pathways), converting diffuse ecosystem interactions into druggable nodes.14–16 Moreover, nanosystems support rational co-delivery and sequencing by simultaneously dampening suppressive signals and providing activating cues, thereby matching the multi-node nature of crosstalk circuits.17 In this Review, we integrate an overview of major nanosystem platforms with current understanding of intercellular communication in the TME, and discuss how these systems can be used to modulate dominant tumor-immune and tumor-stroma circuits, with emphasis on representative preclinical studies, clinical progress, and translational challenges (Scheme 1).

Scheme 1.

Nanosystem-mediated intercellular crosstalk modulation in the tumor microenvironment. Spatially organized tumor niches, including the tumor core, stromal barrier, and perivascular niche, establish suppressive communication networks that promote immune exclusion and vascular gating. These networks involve tumor cells, T cells, DCs, TAMs, CAFs, and TECs and are mediated by extracellular vesicles, signaling pathways, and metabolites. Programmable nanosystems integrate diverse platform types, therapeutic cargos, and engineering features to guide precision TME remodeling, thereby enhancing cancer immunotherapy. Created using Adobe Illustrator 2026 (Adobe Inc., San Jose, CA, USA).

Abbreviations: DCs, dendritic cells; TAMs, tumor-associated macrophages; CAFs, cancer-associated fibroblasts; TECs, tumor endothelial cells.

Major Nanosystem Platforms in Cancer Therapy

Cancer nanomedicine encompasses a broad range of nanosystems that differ in their origin, composition, structure, and functional properties. These differences influence cargo compatibility, stability, biological interactions, and suitability for different therapeutic applications. For a general overview, these nanosystems can be broadly considered according to whether they are biologically derived, engineered, or designed to incorporate or mimic biological components.18,19

Lipid-based nanocarriers mainly include liposomes and lipid nanoparticles. Liposomes are engineered phospholipid vesicles with a bilayer structure resembling biological membranes, whereas lipid nanoparticles are engineered lipid assemblies widely used for the delivery of nucleic acids and other therapeutic cargos. Polymeric nanoparticles and micelles are constructed from polymeric materials whose physicochemical properties can be adjusted for drug loading and release. Inorganic nanoparticles, including metal- and silica-based materials, provide distinct physicochemical features that can be utilized for therapeutic applications.20 Protein-based nanostructures represent another major nanosystem platform. Self-assembled protein cages, for example, can be engineered for cargo loading and delivery.21

Extracellular vesicles (EVs), including exosomes, represent biologically derived nanosystems naturally released by cells and can be developed as therapeutic delivery vehicles.22 Their biological origin offers distinctive properties, although heterogeneity among EV populations and variability in isolation and characterization remain important challenges for standardization and therapeutic development.23 Biomimetic nanoparticles provide another approach by combining synthetic nanomaterials with biological components. Cell membrane-coated nanoparticles, for example, consist of a synthetic nanoparticle core surrounded by a naturally derived cell membrane.24 Lipoprotein-inspired carriers, particularly reconstituted high-density lipoprotein, reproduce key structural features of endogenous lipoproteins and have been investigated for tumor-targeted drug delivery.25 The main characteristics, advantages, limitations, and translational considerations of these nanosystems are summarized in Table 1.

Table 1.

Comparative Characteristics of Major Nanosystem Platforms

Platform	Main Characteristics	Major Advantages	Key Limitations	Translational Considerations	Ref.
Liposomes/LNPs	Lipid-based vesicular or nanoparticulate delivery systems	Compatible with diverse therapeutic cargos; lipid composition and surface properties can be adjusted	Stability, cargo retention/release, and biodistribution depend on formulation characteristics	Composition, particle attributes, stability, and manufacturing processes require consistent control	[20]
Polymeric nanoparticles/micelles	Polymer-based nanoparticles or self-assembled micellar structures	Tunable composition, surface properties, degradation, and cargo-loading characteristics	Performance depends on polymer composition, cargo properties, and formulation conditions	Polymer characteristics and manufacturing processes can influence product performance and reproducibility	[20]
Inorganic nanomaterials	Metal-, metal oxide-, silica-, or related inorganic nanoparticles	Material-dependent optical, thermal, magnetic, or other physical properties	Biodistribution and biological effects depend strongly on material composition and physicochemical properties	Material-specific safety, stability, and in vivo fate require careful characterization	[20]
Protein-based nanostructures	Self-assembled protein architectures, including protein cages	Structurally defined architectures amenable to engineering and cargo loading	Cargo capacity and delivery behavior depend on cage structure and engineering strategy	Structural integrity, cargo incorporation, and biological interactions require platform-specific characterization	[21]
EV-based systems	Naturally produced membrane-bound vesicles that can be adapted for therapeutic cargo delivery	Native membrane architecture and capacity to transport biological cargos	Heterogeneity among EV populations and variability in isolation and characterization methods	Standardized production, characterization, and quality control remain important for therapeutic development	[22,23]
Biomimetic/cell membrane-coated nanoparticles	Synthetic nanoparticle cores coated with naturally derived cell membranes	Combines core material properties with membrane-associated biological functions	Biological properties depend on membrane source and coating characteristics	Standardized membrane sourcing, coating procedures, and characterization are important for clinical development	[24]
Lipoprotein-inspired carriers	Lipid–protein assemblies modeled on endogenous lipoproteins, including rHDL	Suitable for drug incorporation and receptor-mediated cellular interactions	Carrier properties depend on lipid/protein composition, preparation method, and incorporated cargo	Reproducible formulation and characterization are important for further development	[25]

Mechanisms of Intercellular Crosstalk in the TME

Intercellular crosstalk in the TME is conveyed through recurring information routes that operate at distinct biological layers, yet coordinate multicellular behaviors. We review these mechanisms in sequence, covering cargo transfer by exosomes and other extracellular vesicles, classical signaling programs driven by ligands and receptors, and metabolic cues sensed through transporters and receptors. Figure 1 schematizes these three modalities using representative cell types, and Table 2 provides a more comprehensive summary of the channels and intervention-relevant nodes discussed in this section.

Figure 1.

Representative EV-, signaling pathway-, and metabolite-mediated intercellular crosstalk in the tumor microenvironment. (A) EV-mediated transfer of functional cargos from tumor cells to immune and stromal cells. (B) Ligand–receptor signaling circuits among tumor cells, TAMs, and CAFs. (C) Metabolic mediator-driven communication among tumor, immune, stromal, and endothelial compartments. Created using Adobe Illustrator 2026 (Adobe Inc., San Jose, CA, USA).

Abbreviations: TAMs, Tumor-associated macrophages; CAFs, Cancer-associated fibroblasts; TECs, Tumor endothelial cells; DCs, Dendritic cells.

Table 2.

Representative Intercellular Crosstalk Nodes in the Tumor Microenvironment

Crosstalk Route	Communication Channels	Interacting Cells	Biological Impact	Cancer Type	Ref.
Exosome	Munc13-4–dependent secretion of PD‑L1 via sEVs	Tumor cell, CD8+ T cell	Promotes immune evasion and anti–PD‑1 resistance	Breast cancer	[26]
	Tumor EVs decoy anti‑PD‑L1 antibodies	Tumor cell, macrophage	Antibody sequestration/clearance reduces effective anti‑PD‑L1 activity (therapy resistance)	Colorectal Cancer	[27]
	PD‑1/CD80+ immunocyte‑derived sEVs	Immune cell, tumor cell	Induces adaptive PD‑L1 redistribution in tumor cells and strengthens adaptive immunosuppression	Colorectal Cancer	[28]
	CAF‑sEV lncRNA WEE2‑AS1 targeting MOB1A (Hippo)	CAF, CRC cell	Inhibits Hippo signaling (via MOB1A degradation) and promotes CRC progression	Colorectal Cancer	[29]
	HCC exosomal microRNAs reprogram lung niche glucose metabolism	Tumor cell, lung fibroblast; tumor cell, lung endothelial cell	Creates a glucose‑enriched premetastatic niche that supports lung metastasis	Hepatocellular Carcinoma	[30]
	Hypoxia‑inducible exosomal miR‑135a‑5p	CRC cell, Kupffer cell	Reprograms Kupffer cells and liver premetastatic niche to facilitate liver‑tropic metastasis	Colorectal Cancer	[31]
	BCSC exosomal lnc‑PDGFD	Breast cancer stem‑like cell, lung fibroblast	Activates fibroblasts (IL‑11 axis) to form a fibroblast niche and promote lung metastasis	Breast Cancer	[32]
	Exosomal ADAM17 (VE‑cadherin disruption)	CRC cell, vascular endothelial cell	Enhances vascular permeability and premetastatic niche formation, promoting hematogenous metastasis	Colorectal Cancer	[33]
Signaling pathway	IL‑6 signaling	Myeloid/stromal cell, T cell	Suppresses chemotherapy‑elicited antitumor immunity	Leukemia	[34]
	IL‑1R1–IL‑6–STAT3 axis	Myeloid cell, fibroblast, tumor cell	IL‑1R1 inhibition reduces stromal IL‑6 and tumor STAT3 signaling, suppressing PDAC growth	Pancreatic Cancer	[35]
	OSM–OSMR signaling	Macrophage, fibroblast	Reprograms fibroblasts to pro‑tumor states, promoting PDAC growth and metastasis	Pancreatic Cancer	[36]
	Cancer‑cell αV‑integrin activation of latent TGF‑β	Tumor cell, CD8+ T cell	Drives immune exclusion/dysfunction and reduces response to PD‑1 blockade	Melanoma	[37]
	TGF‑β blockade	Tumor/stromal cell, CD8+ T cell	Allows expansion and differentiation of stem‑like CD8 T cells in immune‑excluded tumors	Metastatic Urothelial Carcinoma	[38]
	TGF‑β/PD‑L1 targeting	Tumor/stromal cell, T cell	Reprograms the TME and synergizes with radiotherapy to overcome immune evasion	Breast cancer and lung metastases	[39]
	S100A9–CXCL12 axis	Myeloid cell, stromal cell, T cell	Establishes an immunosuppressive niche associated with immunotherapy resistance	Breast Cancer	[40]
	CXCR4–JAK2/STAT3 signaling	Stromal cell (CXCL12), CD8+ T cell	Orchestrates CD8 T‑cell exhaustion programs	Ovarian Cancer	[41]
	CXCR4 partial agonism	Neutrophil/granulocyte precursor	Targets immunosuppressive neutrophils and cancer‑driven granulopoiesis to restore antitumor immunity	Gastric Cancer	[42]
	CXCR4 blockade	Dendritic cell, T cell	Enhances dendritic‑cell activation and antitumor T‑cell responses in HCC	Hepatocellular Carcinoma	[43]
Metabolic molecule	Lactate uptake via MCT11 (SLC16A11)	Tumor cell, terminally exhausted CD8+ T cell	Lactate uptake sustains dysfunction; MCT11 targeting improves effector function and tumor control	Melanoma	[44]
	Lactate as conditioning cue (epigenetic Tcf7 program)	Tumor cell, CD8+ T cell	Reinforces stem‑like CD8 T‑cell state and improves persistence/function	Colorectal Cancer	[45]
	Lactate‑induced MOESIN lactylation in Tregs	Tumor cell, regulatory T cell	Enhances TGF‑β signaling and suppressive function of Tregs	Hepatocellular Carcinoma	[46]
	ENT1‑mediated intracellular adenosine checkpoint	Adenosine, activated T cell	Adenosine uptake suppresses de novo pyrimidine synthesis and antitumor T‑cell responses; ENT1 antagonism restores function	Breast Cancer	[47]
	PGE2 signaling in dendritic cells (EP2/EP4–cAMP–IRF8)	Tumor/stromal cell, cDC1	Imposes cDC1 dysfunction and limits effective anti‑tumor T‑cell priming	Melanoma	[48]
	Cancer‑derived arginine fueling TAM polyamine synthesis	Tumor cell, macrophage	Drives macrophage polyamine biosynthesis and immune evasion	Breast Cancer	[49]
	Succinate–SUCNR1 signaling	Tumor cell, macrophage	Succinate activates SUCNR1/HIF‑1α programs in macrophages and promotes tumor growth	Lung Carcinoma	[50]
	CAF glutamine synthesis	CAF, macrophage	Tumor‑instructed glutamine synthesis in CAFs promotes pro‑tumor macrophage programs	Melanoma	[51]
	Endothelial fatty‑acid delivery	Endothelial cell, tumor cell	Increases fatty‑acid supply to support outgrowth and survival of colon cancer metastases	Lung metastases	[52]

Extracellular Vesicle-Mediated Crosstalk

Extracellular vesicles (EVs), including small EVs often referred to as exosomes, provide a high-capacity route for tumors to transmit membrane proteins and regulatory cargos across the TME, thereby contributing to the remodeling of immune surveillance and stromal behaviors.22 Within the TME, EV-mediated communication contributes to immune evasion, stromal-tumor reprogramming, pre-metastatic niche formation, and vascular barrier disruption. In immune evasion, recent work shows that EV-associated immune checkpoints can function as mechanistic effectors rather than passive biomarkers. In particular, Munc13-4 was identified as a trafficking regulator that promotes PD-L1 sorting and secretion via exosomes. Genetic deletion of Munc13-4 in breast cancer enhanced T cell–mediated antitumor immunity, suppressed tumor growth, and improved the efficacy of immune checkpoint inhibitors.26 In addition, PD-L1⁺ tumor-derived extracellular vesicles (TEVs) can mediate resistance to anti–PD-L1 therapy by acting as antibody “decoys”: TEV-bound anti–PD-L1 is more readily phagocytosed and degraded by macrophages, resulting in insufficient blockade of PD-L1 on tumor cells and reduced therapeutic efficacy.27 Moreover, consistent with bidirectional checkpoint crosstalk, immunocyte-derived small extracellular vesicles (sEVs) carrying PD-1 and CD80 were shown to induce an adaptive redistribution of PD-L1 in tumor cells, highlighting that circulating sEV checkpoint components can remodel checkpoint organization rather than only transferring signals one-way.28 Such PD-1/CD80⁺ sEVs also induce downregulation of adhesion and antigen presentation-related molecules on tumor cells and impaired immune cell infiltration, thereby converting tumors to an immunologically cold phenotype. In addition to modulating tumor-immune interactions, EVs also mediate communication between stromal and tumor cells within the primary TME. Small EVs derived from cancer-associated fibroblasts (CAFs) carrying lncRNA WEE2-AS1 were shown to promote MOB1A degradation, inhibit Hippo signaling, and facilitate CRC progression, supporting a causal CAFs-EVs-tumor reprogramming axis.29

At distant sites, EVs can promote pre-metastatic niche formation by reprogramming resident stromal and immune cells. In hepatocellular carcinoma (HCC), matrix stiffness was shown to tune exosomal miRNA programs that contribute to a glucose-enriched lung pre-metastatic niche (PMN) by decreasing glucose uptake/consumption in lung fibroblasts while increasing angiogenesis and vascular permeability, which supports metastatic colonization and outgrowth.30 Complementarily, hypoxia in colorectal cancer (CRC) primary lesions was shown to boost exosome release and selectively facilitate liver-tropic PMN formation. Mechanistically, Kupffer cells phagocytosed circulating exosomes enriched in miR-135a-5p, supporting liver-specific niche programming and subsequent liver metastasis.31 Reinforcing the lung-fibroblast axis, a study in triple-negative breast cancer (TNBC) demonstrated that breast cancer stem cell (BCSC)–derived exosomes enriched in lnc-PDGFD activate lung fibroblasts through YBX1/NF-κB signaling, activated fibroblasts further secrete IL-11 to promote lung metastasis, and knocking out lnc-PDGFD in BCSC exosomes markedly reduced lung metastasis in vivo.32 At the vascular interface, EVs can promote metastatic dissemination by disrupting endothelial integrity. CRC-derived exosomal ADAM17 was reported to target endothelial cells, enhance vascular permeability by affecting VE-cadherin membrane localization, promote premetastatic niche formation, and accelerate metastasis in vivo.33 The results of these investigations indicate that EV-mediated crosstalk is not limited to the passive transfer of tumor-derived cargos, but represents a dynamic communication layer that reshapes multiple functional states across the tumor microenvironment and distant metastatic niches. These features make EV biogenesis, release, uptake and cargo function attractive targets for nanosystem-based interventions aimed at disrupting tumor-supportive communication and improving therapeutic efficacy.

Signaling Pathway-Mediated Crosstalk

Intercellular crosstalk in tumors is frequently hard-wired by ligand–receptor signaling networks that couple malignant cells to immune and stromal compartments, generating self-reinforcing circuits that shape immune exclusion, metastatic competency, and therapy response. The studies discussed below span different tumor types and experimental systems, with each example interpreted in the biological context in which the corresponding pathway was examined. Among these, inflammatory cytokine–JAK/STAT programs are recurrent “convergence hubs” because they can simultaneously reprogram myeloid cells, fibroblasts, and cancer cells. For instance, microenvironmental IL-6 can blunt therapy-induced antitumor immunity. In a mouse model of BCR-ABL⁺ B-cell acute lymphoblastic leukemia, tumor microenvironment–derived IL-6 suppressed chemotherapy-triggered CD8⁺ T cell–mediated antitumor responses, while doxorubicin became curative in IL-6–deficient hosts through the induction of CD8⁺ T cell immunity, supporting a causal role for IL-6 signaling in restraining immunogenic therapy effects in this leukemia model.34 Consistently, targeting upstream inflammatory wiring can downshift IL-6/STAT3 output at the tumor–stroma interface. IL-1R1 blockade with anakinra (an IL-1 receptor antagonist) significantly reduced stromal-derived IL-6, thereby suppressing IL-6-dependent STAT3 activation in pancreatic ductal adenocarcinoma (PDAC) cell lines and pancreatic tumors from the aggressive PKT genetic mouse model. Importantly, combining anakinra with cytotoxic chemotherapy significantly prolonged overall survival compared with vehicle control or anakinra monotherapy.35 Beyond IL-6 itself, heterocellular Oncostatin M (OSM)–OSMR signaling provides a clean experimental example of stromal “instruction”. In pancreatic cancer models, macrophage-secreted OSM drives inflammatory gene expression in CAFs, and these reprogrammed CAFs in turn build a pro-tumorigenic microenvironment and engage tumor-cell survival and migratory signaling pathways.36 What’s more, a major axis underlying immune-excluded tumors is TGF-β–centered signaling, which couples stromal remodeling to impaired CD8⁺ T-cell immunity. For example, Malenica et al showed that tumor-cell αV (ITGAV)–mediated activation of endogenous TGF-β shapes the immune infiltrate and modulates response to PD-1 blockade.37 Accordingly, in αV-knockout tumors, PD-1 therapy achieved markedly improved tumor control through increased infiltration by the activated tumor-specific CD103⁺CD8⁺ T cells. At the circuit level, combined PD-L1 and TGF-β blockade enables expansion and differentiation of intratumoral stem cell–like CD8⁺ T cells in immune-excluded tumors, providing a mechanistic rationale for converting boundary-trapped T cells into productive intratumoral effectors.38 Consistent with this rationale, a bifunctional PD-L1/TGF-β trap (bintrafusp alfa) synergized with radiotherapy in poorly immune-infiltrated, therapy-resistant murine models, reprogramming the TME and increasing tumor-infiltrating leukocytes with reconstitution of antitumor immunity.39

In parallel, CXCL12–CXCR4 signaling also contributes to antitumor immunity by mediating intercellular communication within the tumor microenvironment. In BRCA1-mutant breast cancer models, BRCA1 deficiency activated a S100A9–CXCL12 signaling program that promoted the expansion and accumulation of myeloid-derived suppressor cells, created an immunosuppressive microenvironment, and rendered tumors insensitive to immune checkpoint blockade.40 The oncogenic and immunosuppressive actions were effectively suppressed by combinatory treatment targeting the S100A9–CXCL12 axis together with anti-PD-1 therapy, illustrating a chemokine-mediated feedback circuit linking tumor-intrinsic alterations to immune remodeling. Complementing this stromal-centric view, CXCR4 signaling can also operate within immune lineages to enforce a dysfunctional state.41 Mechanistically, CXCR4 orchestrates a TOX-programmed exhausted phenotype in CD8⁺ T cells via the JAK2/STAT3 pathway, whereas CXCR4 blockade or CXCR4 deficiency in CD8⁺ T cells mitigates exhaustion-associated transcriptional/epigenetic transitions, linking a chemokine receptor to the enforcement of T-cell dysfunction. At the myeloid/systemic level, a recent study in gastric cancer models showed that a CXCR4 partial agonist restores anti–PD-1 sensitivity by targeting immunosuppressive neutrophils and cancer-driven granulopoiesis.42 These findings support a role for CXCR4 modulation in neutrophil-associated immunosuppression in gastric cancer. In hepatocellular carcinoma, combined CXCR4 and PD-1 blockade was reported to be safe and to significantly inhibit tumor growth and prolong survival across multiple murine models, with benefit mediated by reprogramming/activation of intratumoral dendritic cell subsets (including Batf3⁺ cDC1), supporting a practical, experimentally supported route to translate chemokine-axis interception into checkpoint sensitization.43 This provides additional evidence for CXCR4-mediated immune remodeling in HCC. In general, these examples highlight cytokine- and chemokine-mediated signaling as important modes of intercellular communication across distinct tumor contexts.

Metabolite-Mediated Crosstalk

Tumor metabolic reprogramming not only satisfies bioenergetic and biosynthetic demands, but also generates a shared extracellular metabolite milieu that functions as an information layer for intercellular communication.53 In exhausted T cells, Peralta et al showed that terminally exhausted T cells upregulate Slc16a11 encoding monocarboxylate transporter 11 (MCT11), which renders them sensitive to lactic acid present at high levels in the tumor microenvironment, and conditional deletion or antibody targeting of MCT11 reduced lactate uptake and improved exhausted T cell effector function with reduced tumor growth.44 In contrast, Feng et al demonstrated that lactate can be repurposed as a conditioning cue to reinforce stem-like properties of CD8⁺ T cells.45 This effect was linked to inhibition of histone deacetylase activity and increased H3K27 acetylation at the Tcf7 super-enhancer locus, leading to enhanced Tcf7/TCF-1 expression. Functionally, CD8⁺ T cells preconditioned with lactate in vitro exhibited stronger tumor growth inhibition after adoptive transfer into tumor-bearing mice. These contrasting findings underscore the context-dependent effects of lactate on T-cell states and caution against assuming a uniformly immunosuppressive role. In another context, Gu et al showed that tumor metabolite lactate promotes tumorigenesis by enhancing Treg cell stability and function.46 Mechanistically, lactate induces MOESIN lactylation at Lys72, strengthens its interaction with TGF-β receptor I, and activates downstream TGF-β/SMAD3 signaling, thereby linking a dominant tumor metabolite to Treg-mediated immunosuppression through post-translational modification. Beyond lactate, extracellular ATP catabolism in tumors promotes adenosine accumulation in the tumor microenvironment, constituting a well-established immunosuppressive axis. Under adenosine-rich conditions, adenosine dampens anticancer T-cell responses after equilibrative nucleoside transporter 1 (ENT1)-mediated uptake into activated T cells and subsequent inhibition of de novo pyrimidine nucleotide synthesis. ENT1 inhibition with EOS301984 restored pyrimidine nucleotide synthesis and augmented antitumor T-cell function. Combining EOS301984 with anti-PD-1 therapy showed synergistic tumor control in a humanized mouse model of triple-negative breast cancer.47 Tumor-derived PGE2, a lipid mediator, programmed dysfunction in mouse and human type 1 conventional dendritic cells (cDC1s) through EP2/EP4-dependent cAMP signaling and loss of IRF8. Blockade of the PGE2–EP2/EP4–cDC1 axis prevents intratumoral cDC1 dysfunction, locally reinvigorates anticancer CD8⁺ T-cell responses, and achieves cancer immune control.48

Metabolite exchange also coordinates pro-tumor macrophage programs, CAF nutrient synthesis, and endothelial gatekeeping, forming an integrated metabolic communication network in tumors. In the breast cancer tumor microenvironment, cancer cells serve as a major source of arginine, which fuels polyamine production in tumor-associated macrophages (TAMs) and drives a pro-tumor, suppressive TAM program that dampens CD8⁺ T-cell antitumor activity, thereby promoting immune evasion and restraining effective antitumor immunity.49 Additionally, tumor-derived succinate can act as an extracellular signaling metabolite by activating SUCNR1 to engage a PI3K–HIF-1α axis, thereby promoting macrophage polarization toward TAMs and facilitating cancer metastasis.50 Metabolic crosstalk is also organized through stromal and vascular relays, as cancer-associated fibroblasts can be instructed into a glutamine-synthesizing state. Tumor-derived palmitic acid activates a TLR4/Syk/NF-κB cascade in fibroblasts, driving inflammatory CAF polarization and IL-6–induced glutamine synthesis. In turn, CAF-derived glutamine promotes polarization of pro-tumorigenic TAMs and supports tumor growth by reshaping TAM composition.51 The endothelium adds another metabolic control layer by gating nutrient transfer from blood to tissue. Endothelial mTORC1 supports transendothelial delivery of long-chain fatty acids to early metastatic tumors, and endothelial-specific Raptor deletion reduces metastatic tumor burden with improved markers of T-cell cytotoxicity. Moreover, low-dose everolimus, which selectively inhibits endothelial mTORC1, enhances responses to immune checkpoint blockade (anti–PD-1) in metastatic disease models.52 Taken together, these studies establish tumor-derived metabolites as active signaling entities that orchestrate immune suppression, stromal reprogramming, and vascular gatekeeping through coordinated metabolic crosstalk, thereby creating integrated metabolic checkpoints that shape tumor progression and therapeutic responsiveness.

Nanosystem-Based Strategies for Crosstalk Modulation to Enhance Anticancer Efficacy

Nanosystem-mediated modulation of intercellular crosstalk requires nanosystem design to be tailored to the target cell populations and communication pathways within the TME. Accordingly, the following sections focus on several major forms of intercellular crosstalk between tumor cells and T cells/dendritic cells (DCs), TAMs, CAFs, or endothelial cells, and illustrate representative nanosystem strategies for modulating these interactions (Figure 2 and Table 3).

Figure 2.

Nanosystem-enabled reprogramming of tumor microenvironment crosstalk across major interaction axes. Representative nanosystem families (left) are mapped onto four dominant crosstalk axes (right): tumor-DC/T cell, tumor-TAM, tumor-CAF, and tumor-TEC. Tumor cues rewire recipient cells and reinforce feedback loops that create bottlenecks, including impaired priming, myeloid immunosuppression, CAF-driven barriers, and endothelial gating. Nanosystems intervene at these nodes to block, amplify, or redirect signals via DC activation, TAM reprogramming, CAF remodeling/ablation, and vascular normalization, thereby improving immune access and therapeutic delivery. Created using Adobe Illustrator 2026 (Adobe Inc., San Jose, CA, USA).

Table 3.

Nanosystem-Based Strategies of Crosstalk Modulation for Enhanced Anticancer Efficacy

Crosstalk Type	Nanosystem	Particle Size	Type of Drug Loading	Administration Route	Cancer Type	Crosstalk Modulation Node	Targeting Ligand	Delivery Strategy	Combination Therapy	Clinical Stage	Key Limitations	Improved Property	Ref.
Tumor cells and T cells/ dendritic cells	PEIM@OVA	125 nm	OVA/MSA-2	S.C.	Melanoma/colorectal tumor	STING-mediated antigen cross-presentation	None	Tumor-draining lymph node-directed delivery	Immune checkpoint blockade	Preclinical: syngeneic mouse model	Complex fabrication may hinder clinical translation	Augment STING cascade activation and antigen cross-presentation for B16-OVA melanoma and immunotherapy	[54]
	PC7A NP	29 nm	Antigen	S.C.	Melanoma, colon carcinoma, HPV-associated tumor	STING-mediated antigen cross-presentation	None	Cytosolic antigen delivery to DCs	Immune checkpoint blockade	Preclinical: syngeneic mouse model	Risk of particle aggregation and protein corona formation	Potent tumour growth inhibition; 100% survival over 60 days with anti-PD-1 antibody; boosting anti-tumour immunity	[55]
	PCD vaccines	75 nm	Antigen	S.C.	Melanoma	cGAS-STING-mediated antigen cross-presentation	None	APC uptake of self-assembled antigen condensates	Alone	Preclinical: syngeneic mouse model	Need for scalable manufacturing processes for clinical translation	Induce robust antigen-specific CTL responses and humoral immunity; induces mitochondrial DNA leakage	[56]
	ACT-DC (AC-NPs + cDC1s)	160 nm	Polyinosinic: polycytidylic acid	I.T.	Colon cancer, melanoma, and glioma	Antigen capture and cross-presentation	None	In situ antigen capture and cDC1-mediated lymph-node relay	Immune checkpoint blockade	Preclinical: syngeneic mouse model	Limited availability of autologous cDC1s and ACT-DC complexity may challenge clinical translation	Eliminate primary tumors in 50–100% and effectively reject tumor rechallenges; capture antigens directly from the tumor and facilitate their delivery to cDC1s	[57]
	NLCs	Not reported	mRNA	i.p.	Melanoma; lymphoma	DC activation and T-cell recruitment	None	NLC-mediated mRNA delivery	Immune checkpoint blockade	Preclinical: syngeneic mouse model	Prevention of cancer recurrence requiring combinatory immunotherapy.	Intratumoral remodelling of the innate and adaptive immunity; increase in the gene of chemokines (Cxcl10, Cxcl11, Cxcl9) involved in CD8+ T cell attraction	[58]
Tumor cells and tumor-associated macrophages	R848@M2pep-MPsAFP	352.9 nm	Resiquimod	i.v.	Hepatocellular carcinoma; Melanoma	TAM polarization and antigen cross-presentation	M2pep	Ligand-mediated TAM targeting	Immune checkpoint blockade	Preclinical: syngeneic mouse model	Concerns about translating this platform from the bench to human clinical trials	Generate strong antitumor immune memory and stem-like CD8+ T cell proliferation for long-term immune surveillance in HCC models	[59]
	PCPA&PPM@TR	136 nm	Temozolomide and Resiquimod	i.v.	Glioblastoma	TAM polarization and immunosuppression	Angiopep-2; mannose analogue	BBB-penetrating sequential tumor/TAM delivery	Chemotherapy	Preclinical: syngeneic mouse model	Limited improvement in survival	Enhanced BBB penetration, dual glioblastoma/TAM targeting, M2-to-M1 repolarization, and anti-glioblastoma efficacy	[60]
	HMMDN-Met@PM	276.7 ± 12.8 nm	Metformin	i.v.	Breast cancer	TAM polarization	M2pep	Biomimetic membrane camouflage with TAM-targeted delivery	Alone	Preclinical: syngeneic mouse model	Limited consideration of the complexity and multifunctionality of TAMs interactions	Re-polarization of M2 macrophages and finally realizing the inhibition of tumor growth and loaded Mn2+ as magnetic resonance imaging for tracking of tumor.	[61]
	P/T@MM NPs	159.6 nm	TMP195	i.v.	Breast cancer	Post-ablation TAM polarization	None	Macrophage membrane-mediated biomimetic delivery	Photothermal therapy	Preclinical: syngeneic mouse model	Potential damage to macrophages caused by PTT in vivo	Elevated the levels of M1-like TAMs, ultimately resulting in a tumor-elimination rate of 60%, increased from 10% after photothermal therapy	[62]
	MG5-S-IMDQ	14.9 nm	Imidazoquinoline	i.v.	Glioblastoma	TAM antigen presentation	Mannose	Receptor-mediated TAM delivery	Alone	Preclinical: syngeneic mouse model	The risk of particle aggregation, and protein corona formation	Capacity of penetrating the BBB, selectively targeting M2-TAMs, improved tumor antigen cross-presentation capability.	[63]
	ME@C NPs	98 nm	E64	i.v.	Melanoma; breast cancer	TAM antigen cross-presentation	Galactose ligand	Ligand-mediated TAM-targeted delivery	Immune checkpoint blockade	Preclinical: syngeneic mouse model	Lack of clinical assessment in patients	Activated CD8+ T cells and suppressing melanoma growth, enhanced the therapeutic efficacy of breast tumor with anti–PD-L1 antibodies	[64]
Tumor cells and cancer-associated fibroblasts	SHK-GW	93.89 nm	Shikonin, GW4869	i.v.	Lung cancer	Tumor exosome-CAF signaling	EpCAM antibody	Active tumor-cell-targeted delivery	Immune checkpoint blockade	Preclinical: syngeneic mouse model	Clinical translation requires the development of more selective inhibitors of the molecular targets.	Reduced exosome-mediated CAF activation and enhanced anti-PD-L1 efficacy	[65]
	AEAA-MSNs@RA/GA & MSNs@Ce6	80 nm	Retinoic acid, gambogic acid; Ce6	i.v.	Hepatocellular carcinoma	CAF activation and vascular normalization	Aminoethyl anisamide	CAF-directed sequential nanodrug/PDT delivery	Photodynamic therapy	Preclinical: syngeneic mouse model	The difficulty of PDT for primary liver cancer.	Improved tumor accumulation of photosensitizers and alleviated hypoxia in the TME, which potentiated the PDT-induced immune response	[66]
	IEVs-PFD/138	163.2 nm	miR-138-5p; pirfenidone	i.v.	Pancreatic cancer	CAF TGF-β signaling and stromal remodeling	Integrin α5-targeting peptide	CAF-targeted engineered EV delivery	Chemotherapy	Preclinical: stroma-rich and patient-derived xenograft mouse models	Clinical translation requires the development of more selective inhibitors of the molecular targets.	Synergistically promoted CAF reprogramming and decreased tumor pressure, enhanced gemcitabine perfusion, tumor hypoxia amelioration	[67]
	αFAP-Z@FRT	16.8 ± 2.1 nm	ZnF16Pc	i.v.	Breast cancer	FAP+ CAF-mediated stromal immunosuppression	FAP-specific scFv	CAF-targeted nanocage delivery	Photodynamic therapy and immune checkpoint blockade	Preclinical: syngeneic mouse model	The risk of particle aggregation, and protein corona formation	Enhanced anti-cancer immunity, causing suppression of both primary and distant tumors	[68]
	LNCs@si-CXCL12	80.5 nm	CXCL12 siRNA	i.p.	Esophageal squamous cell carcinoma	CAF-derived CXCL12 signaling	Anti-FAPα	CAF-targeted siRNA delivery	Alone	Preclinical: spontaneous esophageal squamous cell carcinoma mouse model	Clinical translation requires the development of more selective inhibitors of the molecular targets.	CD8+T cell migration and reduced tumor growth, enhanced CD8+T cell-mediated tumoricidal activity, and improved overall survival.	[69]
Tumor cells and endothelial cells	bsMINM	60 nm	None (intrinsic bispecific molecular recognition)	i.v.	Breast cancer	VEGF/VEGFR and DLL4/Notch signaling	None	Ligand-free bispecific molecular recognition	Alone	Preclinical: MCF-7 xenograft mouse model	Scale-up of nanoparticle preparation	Enhances anti-tumor angiogenesis and growth inhibition	[70]
	FLG and MAR/MPA NPs	140.63 nm	Gambogic acid; maraviroc	i.v.	Melanoma; pancreatic cancer	VEGF/VEGFR2 and CCL5/CCR5 signaling	F3 peptide; anisamide	Dual vascular-endothelial and tumor-cell targeting	Alone	Preclinical: mouse tumor models	Scale-up of nanoparticle preparation	Induced vascular normalization window lasting 9 days and restored vascular permeability and oxygen supply in Panc-1 tumor.	[71]
	PEI-PEG-cRGD	150 nm	MYC siRNA/miR-218	i.v.	Melanoma; Lung Carcinoma	Endothelial MYC/Notch signaling	cRGD	Endothelial-targeted siRNA/miRNA delivery	Chemotherapy and immune checkpoint blockade	Preclinical: syngeneic mouse model	Potential limitations related to siRNA stability	Normalized tumor vessels, increased T-cell infiltration, and enhanced chemotherapy and anti-PD-1 responses	[72]
	E-selectin-targeted NS	163 nm	None	N/A (in vitro)	Breast cancer	E-selectin-mediated tumor-endothelial adhesion	E-selectin antibody	Active endothelial targeting	Alone	Preclinical: in vitro	Evaluation of the approach in more dynamic conditions that mimic in vivo scenarios	Reduced MDA-MB-231 binding to HMVEC-Ls by up to 41%	[73]
	P-selectin-targeted NPs	80 ± 10 nm	Vismodegib	i.p.	Medulloblastoma	P-selectin-mediated endothelial barrier transport	Fucoidan	Endothelial transcytosis across the BBB	Radiation treatment	Preclinical: genetic Sonic Hedgehog medulloblastoma mouse model	Further clinical applicability of this P-selectin-targeting approach to other intracranial tumour	Enhanced tumor-selective BBB penetration with antitumor efficacy, reduced bone toxicity, and reduced drug exposure to healthy brain tissue	[74]
	NI@I-NPs	128 nm	Nitazoxanide and ibrutinib	i.v.	Breast cancer brain metastases	WNT/pericyte-mediated blood-tumor barrier regulation	ICAM-1-targeting ligand	Endothelial targeting with blood-tumor barrier modulation	Doxorubicin/etoposide	Preclinical: breast cancer brain-metastasis xenograft mouse model	Further clinical diagnosis and systemic treatments of brain metastases	Improved blood-tumor barrier permeability and extended median survival in breast cancer brain metastasis models	[75]

The Crosstalk Between Tumor Cells and T Cells/Dendritic Cells

Tumor cells and T cells/ dendritic cells engage in a tightly coupled, bidirectional crosstalk that governs the balance between effective immune surveillance and immune evasion. Tumor cell turnover releases neoantigens and tumor associated antigens, which are taken up by DCs, processed through MHC-I and MHC-II pathways, and presented to CD8⁺ and CD4⁺ T cells. These processes, including direct presentation, classical cross presentation and MHC-I cross dressing, initiate and strengthen antitumor T cell responses.10,76,77 Activated CD8⁺ T cells then enter the tumor, form immune synapses with cancer cells, and kill them through perforin and granzyme or death receptor signaling.78 Immunogenic tumor cell death provides additional antigens and danger signals, allowing DCs to continue priming new T cells.79 However, persistent antigen stimulation and inflammatory stress in the tumor microenvironment lead to T cell exhaustion, marked by increased PD-1, CTLA-4, TIM-3 and LAG-3 expression and reduced effector function.80 Tumor cells further weaken immunity by upregulating PD-L1 and altering dendritic cell behavior. Factors such as VEGF, IL-6, IL-10, PGE2 and lactic acid impair dendritic cell maturation and cross priming and promote tolerogenic DCs that tend to induce regulatory T cells rather than effective CD8⁺ T cells.81 Overall, these interactions shape key checkpoints of the cancer immunity cycle.

These immune-evasion mechanisms define several tractable points for nanosystem intervention, including enhancement of antigen capture and cross-presentation, promotion of DC activation, and reinforcement of downstream T-cell priming. Activation of the STING pathway enhances dendritic cell antigen presentation and promotes potent T-cell-mediated antitumor responses, and growing evidence shows that nanoadjuvant platforms can further amplify this process by engaging the STING or cGAS–STING axis to boost downstream cytotoxic T-cell activation. An acid-ionizable iron nanoadjuvant, named PEIM@OVA, was constructed from ionizable polymer-coated iron oxide nanoparticles co-loaded with the small-molecule STING agonist MSA-2.54 This formulation enables controlled endolysosomal disruption and targeted delivery to tumor-draining lymph nodes, resulting in enhanced antigen uptake by CD169⁺ antigen-presenting cells (APCs) and markedly improved antigen cross-presentation. In B16-OVA melanoma and MC38 colorectal cancer models, this system enables efficient delivery of autologous tumor antigens and, when combined with immune checkpoint inhibition, markedly suppresses postoperative tumor relapse and distant spread. Beyond nanoadjuvants, several nanovaccine systems have employed STING activation as an intrinsic immunomodulatory mechanism. A representative example is the PC7A polymeric nanovaccine, a minimalist formulation in which tumor antigen is physically mixed with synthetic PC7A nanoparticles (approximately 20–50 nm).55 This system enables efficient cytosolic antigen delivery and cross-presentation in DCs while selectively activating STING, not TLR or MAVS pathways, thereby inducing type I interferon responses and promoting potent tumor-specific CD8⁺ T-cell priming. In melanoma, colon carcinoma, and HPV-associated tumor models, PC7A vaccination suppressed tumor growth and synergized with PD-1 blockade to induce complete tumor clearance and durable immune memory. Notably, later work demonstrated that intratumoral administration markedly enhanced therapeutic efficacy compared with subcutaneous injection by activating STING in tumor-resident myeloid cells, inducing CXCL9 expression, and recruiting IFNγ-producing CD8⁺ T cells to establish a CXCL9–IFNγ positive feedback loop essential for tumor eradication.82 Additionally, condensate nanovaccine adjuvants formed by the self-assembly of protein antigens into nanoscale condensates have been shown to promote mitochondrial DNA leakage within DCs, thereby activating cGAS–STING signaling.56 This activation leads to DC maturation, increased type I interferon and IL-12 production, enhanced cross-presentation, and potent CD8⁺ T-cell-dependent antitumor immunity. Together, these studies establish that nanoadjuvant and nanovaccine platforms capable of engaging the STING or cGAS–STING pathway represent a validated strategy for amplifying tumor–APC–T cell communication and achieving durable antitumor immunity. Besides this, nanotechnology has also been applied to directly enhancing antigen acquisition, activation, and T-cell priming functions of DCs, thereby promoting potent antitumor immunity. To illustrate the efficacy of this approach, one strategy integrates polymer-based antigen capturing nanoparticles synthesized from acid-ended poly(lactic-co-glycolic) acid and polyethylenimine with adoptively transferred migratory CD103⁺ cDC1s, thereby establishing an in situ antigen relay vaccination system termed Antigen Capturing nanoparticle Transformed Dendritic Cell therapy (ACT-DC).57 Following intratumoral administration, antigen-capturing nanoparticles bind endogenous tumor proteins and danger-associated signals and are selectively internalized by migratory cDC1s. These cDC1s subsequently migrate to tumor-draining lymph nodes, where they markedly enhance antigen cross-presentation and prime polyclonal CD8⁺ T-cell responses, while also contributing to improved T-cell infiltration within tumors. In multiple solid tumor models including colon cancer, melanoma, and glioma, this ACT-DC vaccination strategy demonstrates durable tumor regression, protection against tumor rechallenge, and enhanced therapeutic benefit when combined with radiotherapy or PD-1 blockade. Complementing this approach, a chitosan nanoparticle-coated dendritic cell vaccine has been shown to enhance DC maturation markers, and significantly increase CD4⁺ T-cell activation in vivo compared to unmodified DCs.83 Nanostructured lipid carrier-based mRNA vaccines have demonstrated the ability to activate DCs through TLR4/TLR8- and ROS-dependent pathways, induce strong antigen-specific CD4⁺ and CD8⁺ T-cell responses, and create a T-cell-inflamed tumor microenvironment that improves responsiveness to PD-1 blockade and supports long-term immune memory.58 Collectively, these studies suggest that dendritic-cell-targeted nanosystems enhance antitumor immunity by reinforcing multiple steps of the tumor–DCs–T cell communication axis, including antigen capture, dendritic cell maturation, cross-presentation, lymph node priming, and effector T-cell infiltration. Rather than acting only as antigen carriers, these platforms convert tumor-derived antigens and danger cues into sustained CD4⁺ and CD8⁺ T-cell responses, promote immune memory, and establish a T-cell-inflamed microenvironment that may improve the efficacy of immune checkpoint blockade.

The Crosstalk Between Tumor Cells and Tumor-Associated Macrophages

As integral elements of the tumor microenvironment, TAMs interact reciprocally with tumor cells to form a bidirectional regulatory circuit that strongly influences tumor progression.84 Tumor cells secrete chemokines and growth factors such as CCL2, CCL5, CXCL12, colony-stimulating factor 1 and VEGF to recruit circulating monocytes into the TME and drive their differentiation toward an M2-like, pro-tumoral TAM phenotype.85 These reprogrammed TAMs, in turn, promote tumor growth, epithelial–mesenchymal transition, angiogenesis, and metastasis through the release of VEGF, EGF, TGF-β, CXCL12, and matrix-remodeling enzymes.86 TAMs also dampen antitumor immunity by expressing PD-L1, secreting immunosuppressive cytokines such as IL-10 and TGF-β, and impairing cytotoxic T-cell activation and infiltration.87 Extracellular vesicles and non-coding RNAs further refine this communication, establishing positive feedback loops that reinforce immunosuppression and therapy resistance.88 Meanwhile, emerging evidence demonstrates that TAMs are not exclusively tumor-promoting; instead, they can also exhibit antitumor activity under specific microenvironmental or therapeutic cues. Inflammatory or early-stage TAMs may display M1-like characteristics, including production of IL-12, TNF-α, reactive oxygen species (ROS), and inducible nitric oxide synthase (iNOS), contributing to direct tumoricidal activity.89 These TAM subsets can also present antigens via MHC-II or CD80/CD86 to support T-cell priming and enhance antitumor immune responses.90 These findings indicate that TAMs function as a “double-edged sword” in cancer. Their functional plasticity supports nanosystem strategies beyond M2-to-M1 repolarization, including depletion of tumor-supportive macrophages, phenotypic reprogramming, and restoration of phagocytic or antigen-presenting functions.

A first class of nanosystem-based strategies has therefore focused on selective delivery of immunomodulators to M2-like TAMs, aiming to reprogram their phenotype and weaken their tumor-supportive functions. Engineered microparticles (MPs) derived from alpha-fetoprotein (AFP)-overexpressing macrophages carrying resiquimod (R848@M2pep-MPsAFP) can be efficiently internalized by TAMs, driving robust M2-to-M1 repolarization, enhancing antigen cross-presentation, and promoting the expansion of stem-like CD8⁺ T cells, thereby markedly boosting anti-PD-1 therapy in hepatocellular carcinoma models.59 A related but more integrated strategy employed a pH-responsive hierarchical micelleplex designed to sequentially target glioblastoma cells and TAMs. In this system, temozolomide delivery induced tumor cell damage and antigen release, while co-delivered resiquimod selectively reprogrammed M2-like TAMs.60 This dual intervention weakened TAM-mediated immunosuppression, alleviated chemoresistance, and reshaped tumor-TAM communication in glioblastoma models. In another investigation, hollow mesoporous MnO₂ nanoparticles loaded with metformin and cloaked with macrophage membranes, together with M2-targeting ligands, showed preferential accumulation in M2-like TAMs.61 These nanoparticles induced TAM repolarization toward an M1-like phenotype while simultaneously enabling magnetic resonance imaging of TAM-rich tumor regions, thereby reducing TAM-driven tumor support in vivo. In the context of local tumor ablation, biomimetic polydopamine-based nanoparticles carrying TMP195 acted as both photothermal agents and TAM-repolarization carriers.62 Following photothermal therapy, these nanoparticles targeted infiltrating TAMs, reversed their immunosuppressive phenotype, and remodeled the post-ablation inflammatory microenvironment that otherwise favors tumor recurrence and metastasis. Repolarization may be particularly suitable when tumor-supportive TAMs retain sufficient plasticity for functional reprogramming.

In addition to repolarization, depletion of tumor-supportive TAMs represents an alternative nanosystem-based strategy. For example, clodronate-containing liposomes have been used to deplete TAMs in primary and metastatic melanoma models, resulting in anti-angiogenic and antitumor effects.91 However, because TAMs are functionally heterogeneous, depletion may also remove macrophage populations with potentially beneficial immune functions, which should be considered when selecting this strategy.

Beyond altering TAMs abundance or polarization, several nanosystems have leveraged tumor-TAMs interactions as an entry point to initiate broader immune remodeling. Wang et al introduced dendrimer-based nano-reprogrammers functionalized with mannose and conjugated to a TLR7/8 agonist.63 These nano-reprogrammers selectively accumulated in M2-like TAMs within orthotopic glioblastoma, enhanced their phagocytic and antigen-presenting capacity, and initiated a cascade of immune activation described as a “gear effect”, linking TAM reprogramming to broader antitumor immunity. Surgical tumor-derived galactose ligand-modified cancer cell membrane-coated cysteine protease inhibitor (E64)-loaded mesoporous silica nanoparticles further extend this concept by restoring antigen cross-presentation in M2-like macrophages without altering the macrophage phenotype, thereby facilitating the activation and proliferation of antigen-specific CD8⁺ T cells, leading to effective suppression of postsurgical tumor recurrence.64 Together, these studies suggest that enhancing TAM phagocytosis and antigen presentation can improve antitumor immunity even without complete macrophage repolarization. Such functional restoration may be advantageous when preserving macrophage populations is desirable, rather than broadly depleting TAMs or requiring complete phenotypic repolarization.

Instead of just targeting TAMs solely as therapeutic targets, TAMs-derived biological components can also be incorporated into nanosystems to modulate tumor-TAMs crosstalk and reshape macrophage function. Chen et al developed TAMs-membrane-coated upconversion nanoparticles (NPR@TAMM), in which the TAMs membrane depleted tumor-derived CSF1 and disrupted tumor cell-TAMs communication.92 Combined with photodynamic therapy, this platform shifted macrophage activation from an immunosuppressive M2-like phenotype toward a more inflammatory M1-like state and enhanced antitumor immunity through activation of antigen-presenting cells and tumor-specific effector T cells. In another TAMs-membrane-based strategy, a ROS-sensitive camptothecin/cinnamaldehyde prodrug nanoparticle was camouflaged with TAM membranes to form DCC@M2. The resulting nanodecoy targeted both primary tumors and lung metastatic lesions and scavenged tumor-derived CSF1, thereby disturbing tumor-TAMs interactions while promoting TAMs depletion and immunogenic cell death.93 These studies highlight the potential of TAMs-derived membranes to provide both tumor-targeting and cytokine-scavenging functions, extending their role beyond conventional nanoparticle coating. Macrophage-derived extracellular vesicles represent another biologically derived nanosystem format. Gunassekaran et al engineered M1 macrophage-derived exosomes with NF-κB p50 siRNA and miR-511-3p and surface-modified them with an IL4R-binding peptide.94 The resulting IL4R-targeted exosomes facilitated selective delivery to IL4R-expressing M2-like macrophages and promoted TAM reprogramming toward an M1-like phenotype, thereby enhancing antitumor immunity. Overall, these studies show that nanosystems can remodel tumor-macrophage communication by altering macrophage state, abundance, or function, while macrophage-derived membranes and extracellular vesicles provide additional biomimetic platforms for disrupting or redirecting tumor-TAMs communication.

The Crosstalk Between Tumor Cells and Cancer-Associated Fibroblasts

CAFs are a major stromal compartment and are particularly prominent in desmoplastic tumors, where extensive reciprocal signaling with cancer cells shapes progression and therapy response.95 In these tumors, cancer cells can drive CAF emergence and functional heterogeneity through paracrine signaling. In PDAC, CAFs are not a uniform population but comprise distinct states, including myofibroblastic CAFs and inflammatory CAFs, reflecting spatially organized stromal niches.96 Mechanistically, IL-1 induces LIF expression and activates downstream JAK/STAT signaling to generate inflammatory CAFs, whereas TGF-β antagonizes this process and shifts fibroblasts toward a myofibroblastic program by downregulating IL-1R1, thereby illustrating how tumor-derived cues “program” distinct CAF niches within the PDAC microenvironment.97 Conversely, CAFs promote malignant tumor phenotypes through multiple mechanisms. In breast cancer, carcinoma-associated fibroblasts with elevated CXCL12/SDF-1 were shown to enhance tumor growth and angiogenesis, including through recruitment of endothelial progenitor cells.98 In colorectal cancer, TGF-β activity on stromal cells induces CAF-derived IL-11, which activates tumor-cell GP130/STAT3 signaling and confers a survival advantage that facilitates metastatic colonization, establishing a defined tumor–CAF paracrine circuit driving metastasis.99 In cholangiocarcinoma, CAFs are described as being recruited and persistently activated by tumor-derived factors including PDGF-D and FGF/TGF-β–related signals and, reciprocally, CAFs enhance cholangiocarcinoma cell proliferation and invasiveness.100 Furthermore, in scirrhous gastric carcinoma, conditioned media from scirrhous gastric cancer cells increased α-SMA in CAFs via TGF-β/Smad2-dependent mechanisms, providing direct tumor-to-CAF activation evidence in this clinically fibrotic gastric cancer subtype.101 All these studies support tumor–CAF crosstalk as a multi-layered communication network governing tumor growth, metastasis, immune evasion, and treatment failure.

Nanosystem strategies can intervene at different points in this tumor–CAF circuit, either by limiting tumor-derived signals that activate CAFs or by directly modulating CAF state, abundance, and stromal output. Cancer cell–targeted dual-payload nanoliposomes were engineered by decorating the liposome surface with an anti-EpCAM antibody and co-encapsulating shikonin plus GW4869 to inhibit sequential steps of exosome release and biogenesis, thereby weakening tumor-derived exosome signaling and limiting fibroblast activation and differentiation into CAFs.65 This strategy illustrates how EV-mediated communication can provide a mechanistic basis for nanosystem design. In parallel, CAF-directed nanoliposomes were designed to target a GPR77 and CD10 double-positive CAF subset using dual-antibody recognition and to co-deliver TAS-120 and ICG-001, which block FGFR signaling and Wnt/β-catenin transcriptional activity, respectively, thereby shifting activated CAFs toward a more quiescent fibroblast-like state. Co-administration of the two formulations was reported to increase intratumoral cytotoxic T-cell infiltration and improve the efficacy of anti–PD-L1 therapy in immunocompetent lung cancer models.65 Building on the same microenvironment-priming paradigm, Fei et al developed a sequential regimen in which a CAF-targeting nanodrug co-delivering retinoic acid and gambogic acid reprograms activated CAFs and normalizes tumor vasculature, thereby reducing stromal barriers, alleviating hypoxia, and increasing intratumoral accumulation of chlorin e6 nanophotosensitizers to potentiate subsequent photodynamic therapy in hepatocellular carcinoma.66 In addition to pharmacologic rewiring, CAF phenotypes can be pushed toward a less tumor-supportive program using material-driven reprogramming, as exemplified by gold nanoparticles reported to transform activated CAFs toward quiescence and to provide mechanistic insights into this phenotypic reversal.102 In a related approach, in pancreatic cancer, engineered extracellular vesicles loaded with miR-138-5p and the antifibrotic agent pirfenidone and surface-modified with integrin α5–targeting peptides were designed for CAF targeting and reprogramming, resulting in decreased tumor pressure, improved gemcitabine perfusion, hypoxia amelioration, and increased chemosensitivity in stroma-rich and patient-derived xenograft models.67

In contrast to phenotypic reprogramming approaches that push activated CAFs toward a less tumor-supportive state, selective CAF depletion provides a more direct route to break CAF-driven crosstalk circuits and remodel the tumor microenvironment. A representative example is a site-specific CAF-ablation photodynamic therapy (PDT) using ZnF₁₆Pc (a photosensitizer)-loaded apoferritin nanocages decorated with a FAP-specific single chain variable fragment (αFAP-Z@FRT), which accumulates in intratumoral FAP⁺ CAFs and eliminates them upon local irradiation with no detectable systemic toxicity in mice.68 Mechanistically, in 4T1 models, this CAF-targeted PDT induces CD8⁺ T cell–dependent tumor control and an abscopal effect, and it further enhances anti–PD-1 therapy while eliciting T-cell responses against both cancer cells and CAFs. In a complementary vein, an aptamer-modified PLGA nanoemulsion co-delivering doxorubicin and small interfering RNA (siRNA) against hepatocyte growth factor (HGF) to CAFs was described to induce CAF apoptosis and reduce ECM deposition, while lowering HGF expression in residual CAFs, thereby inhibiting tumor proliferation, migration, and invasion and improving tumor permeability in colorectal cancer.103 In addition, Cys-Arg-Glu-Lys-Ala peptide-conjugated, DOX-loaded hydroxyethyl starch–IR780 nanoparticles that target fibronectin overexpressed on CAFs were reported to combine light-triggered photothermal heating with chemotherapy, which decreased CAFs, remodeled the tumor mechanical microenvironment, disrupted the cancer stem cell niche, and ultimately suppressed 4T1 triple-negative breast tumor growth.104 Notably, CAF-associated chemokine cues can also propagate extended crosstalk through immune remodeling. CXCL12 was identified as an ESCC-CAF-associated chemokine. CAF-derived CXCL12 was shown to suppress CD8⁺ T-cell tumor-killing activity via the CXCL12-CXCR4 axis, and nanoparticle-mediated CXCL12 silencing targeting CAFs in a spontaneous ESCC mouse model reduced tumor growth, enhanced CD8⁺ T-cell–mediated tumoricidal activity, and improved overall survival.69 These findings shift the interpretation of CAF-directed nanosystems from stromal remodeling tools to regulators of intercellular communication within the TME. By altering how CAFs receive tumor-derived cues and transmit stromal, vascular, and immunosuppressive signals, nanosystems can weaken the communication loops that maintain immune-excluded and therapy-resistant niches. This crosstalk-oriented mechanism provides a rationale for integrating CAF modulation with cancer immunotherapy, particularly in tumors where poor T-cell entry and CAF-driven suppression limit the efficacy of immune checkpoint blockade.

The Crosstalk Between Tumor Cells and Endothelial Cells

Tumor cells and ECs form a central bidirectional crosstalk axis in the tumor microenvironment because ECs line the blood vessels that physically connect tumors to the systemic circulation.105 This vascular network therefore functions as both a conduit and an interface that regulates the transport of oxygen, nutrients, and therapeutics into tumors, and it also provides the endothelial barrier that tumor cells must interact with to enter the bloodstream and to exit it at distant sites during metastasis.106 Tumor-driven pro-angiogenic and inflammatory programs activate and remodel ECs, and the resulting vascular abnormalities feed back to reinforce hypoxia, acidosis, and therapeutic resistance.107 On the tumor-to-endothelium side, tumors often sustain angiogenic signaling through factors such as VEGF. During sprouting angiogenesis, VEGF cooperates with Dll4/Notch signaling to regulate tip-stalk cell specification and branching behavior.108 Sustained angiogenic stimulation typically generates a disorganized and inefficient vascular network with heterogeneous flow, poor perfusion, and increased leakiness. These abnormalities can elevate interstitial fluid pressure and contribute to spatially heterogeneous oxygenation within tumors.109 On the endothelium-to-tumor side, tumor endothelial cells (TECs) are not merely passive conduits. They can deliver membrane-bound and secreted angiocrine signals that have been linked to tumor progression, therapy resistance, and metastatic phenotypes in experimental frameworks.110,111 The endothelial barrier and junctional integrity also act as key checkpoints in the metastatic cascade. TEVs can weaken endothelial junctions, and exosomal miR-105 has been shown to target the tight junction-associated protein ZO-1 and disrupt endothelial barriers to promote metastasis.112 At the immune interface, the tumor vasculature can impose selective immune exclusion. FasL expression in tumor endothelium has been associated with scarce CD8⁺ T-cell infiltration and a predominance of regulatory T cells in multiple human and mouse solid tumors.113 Together, these vascular abnormalities provide the rationale for vascular normalization. Rather than simply eliminating tumor vessels, vascular normalization aims to restore abnormal tumor vasculature to a more organized, stable, and functional state, with improved perfusion and endothelial barrier integrity. Importantly, improved perfusion does not necessarily translate into accelerated tumor growth. Simultaneous Ang2 inhibition and Tie2 activation normalized tumor vessels and enhanced perfusion and drug delivery while reducing tumor growth and metastasis.114 Endothelial normalization induced by PHD2 haplodeficiency likewise improved tumor perfusion and oxygenation without increasing primary tumor growth, while suppressing tumor-cell invasion, intravasation, and metastasis.115 Vascular normalization is also closely linked to antitumor immunity: disruption of vessel normalization reduces T-cell infiltration, while CD4⁺ TH1-cell activity can reciprocally promote vascular normalization.116

The tumor endothelium therefore provides several actionable targets for nanosystem intervention, including VEGF-VEGFR2, DLL4-Notch, and transcriptional programs that maintain abnormal TEC states. Attenuating VEGF-VEGFR2 signaling in TECs represents a practical entry point to induce vessel normalization and improve responses to subsequent therapies. An arginyl-glycyl-aspartic acid-modified lipid nanoparticle platform based on a pH-sensitive, biodegradable ssPalm lipid delivered siVEGFR2 to TECs, achieved VEGFR2 knockdown, and was reported to induce vascular normalization. Importantly, this endothelial silencing strategy enhanced anti–PD-1 efficacy by accelerating normalization and increasing T-cell infiltration in tumors.117 In addition to VEGFR2-focused silencing, multi-node endothelial pathway blockade has been explored to suppress compensatory angiogenic circuits. A bispecific molecularly imprinted nanomissile simultaneously targeting VEGF and DLL4 was reported to inhibit VEGF–VEGFR signaling in ECs and DLL4–Notch signaling across endothelial and tumor compartments, thereby reducing VEGF/DLL4 feedback and restraining tumor angiogenesis and growth in vivo.70 To mitigate TME-driven negative feedback that can limit anti-VEGF vascular normalization, Deng et al developed a combined nanosystem composed of FLG and MAR/MPA nanodrugs to co-regulate tumor vasculature and the TME.71 FLG was built by connecting low-molecular-weight heparin and gambogic acid, and was decorated with an F3 peptide to directly act on vascular ECs and induce vascular normalization. In parallel, MAR/MPA encapsulated the CCL5/CCR5 blocker maraviroc and was designed to restrict cytokine signaling linked to angiogenesis and TME deterioration, thereby supporting vasculature repair and TME reconstruction. With intravenous co-administration, this combination was reported to synergistically extend the vascular normalization window to 9 days and restore vascular permeability and oxygen supply in Panc-1 tumors, while increasing CD4⁺ and CD8⁺ T-cell infiltration in melanoma with a remodeled microenvironment. Beyond VEGFR2, TEC programs can be disrupted by targeting transcriptional dependencies enriched in tumor angiogenesis. p5RHH-based nanoparticles delivering Etv2/Er71 siRNA potently inhibited tumor angiogenesis and growth, and the Etv2 siRNA nanoparticle regimen did not elicit cardiovascular side effects under their experimental conditions.118 More recently, single-cell transcriptomic analyses of TECs revealed that canonical Notch signaling normalizes tumor vessels primarily by transcriptionally restraining a MYC-driven, highly proliferative TEC state (~30% of TECs). Endothelial Notch activation suppresses MYC programs and depletes this cycling subset, thereby shifting TEC composition toward more functional, stabilized vasculature. Building on this Notch–MYC regulatory axis, PEI-PEG-cRGD nanoparticles were engineered for endothelial delivery of MYC siRNA (EC-siMYC) or the Notch-downstream MYC-suppressing miRNA miR-218 (EC-miR-218), which recapitulated Notch activation–induced tumor vessel normalization and thereby sensitized tumors to cisplatin chemotherapy and anti-PD-1 immunotherapy in mouse models.72

Beyond vascular normalization, tumor-endothelial crosstalk can be intercepted by targeting endothelial adhesion and barrier-transport functions, which together define both tumor cell–vessel interactions and therapeutic entry into protected lesions. For example, E-selectin antibody–functionalized gold nanoshells were designed to bind human lung microvascular ECs under baseline and TNF-α-inflamed conditions and reduced MDA-MB-231 cell binding in vitro, supporting an endothelial selectin-blocking approach to interfere with the vascular arrest step preceding extravasation.73 In intracranial disease, fucoidan-based P-selectin–targeted nanocarriers were developed to target endothelial P-selectin and induce caveolin-1-dependent transcytosis, enabling selective and active transport into the brain tumor microenvironment; in a Sonic hedgehog medulloblastoma model, fucoidan-based nanoparticles encapsulating vismodegib showed striking efficacy with reduced bone toxicity and reduced drug exposure to healthy brain tissue, and radiation further increased transport efficiency.74 Along the same barrier-gating theme, ICAM-1–targeted NI@I-NPs were reported to open the blood–tumor barrier by targeting endothelial WNT signaling and tumor pericytes, improving chemotherapeutic efficiency and extending survival when combined with doxorubicin/etoposide in breast cancer brain metastasis models.75 Taken together, nanosystem-based modulation of tumor–endothelial crosstalk converges on two complementary principles: reprogramming TEC states to normalize vessel function and directly tuning endothelial gatekeeping (adhesion and barrier transport) to control extravasation and therapeutic access. These strategies are most impactful when implemented as time- and context-aware combinations, leveraging a normalization “window” and endothelial addressability to coordinate drug delivery with immune engagement, thereby offering a coherent framework for improving both local tumor control and metastasis prevention.

Nanosystem Modulation of Stromal and Metastatic Microenvironments

Beyond direct cell–cell crosstalk, the physical and spatial microenvironments of tumors also contribute to tumor progression and metastatic dissemination. The ECM regulates immune-cell trafficking and therapeutic access, whereas circulating tumor-cell clusters and distant metastatic sites present distinct biological and physical constraints during metastasis. These microenvironmental features therefore provide additional opportunities for nanosystem-based intervention.

Extracellular Matrix Interactions and Nanosystem Modulation

The ECM comprises diverse stromal matrix elements, including structural and adhesive proteins such as collagen and fibronectin, proteoglycans such as versican, and glycosaminoglycans, including hyaluronan and heparan sulfate. Rather than serving solely as a structural scaffold, these components regulate tumor-stromal and tumor-immune interactions through changes in matrix organization, mechanical properties, cell adhesion, growth-factor availability, and immune-cell trafficking. Tumor-cell-derived discoidin domain receptor 1 (DDR1), for example, promotes collagen-fiber alignment and contributes to T-cell exclusion, whereas disruption of DDR1-dependent collagen organization improves immune-cell access and suppresses tumor growth.119 Fibronectin matrix assembly also affects immune trafficking; inhibition of α5β1-dependent fibronectin organization enhances CD8⁺ T-cell transendothelial migration and improves the response to PD-L1 blockade in preclinical breast cancer models.120 Stromal versican accumulation is associated with CD8⁺ T-cell exclusion, whereas greater versican proteolysis is associated with increased epithelial infiltration of CD8⁺ T cells in breast cancer.121 Thus, matrix proteins, proteoglycans, and glycosaminoglycan-rich networks participate directly in the regulation of immune access and tumor-stromal interactions rather than functioning only as physical barriers.

Nanosystems can modulate these matrix-dependent processes through different approaches. One strategy is to selectively remodel matrix structures that restrict immune or drug access. An inhalable lipid nanoparticle co-delivering mRNA encoding an anti-DDR1 single-chain variable fragment and siRNA targeting PD-L1 disrupted DDR1-dependent collagen-fiber alignment, reduced tumor stiffness, increased immune-cell infiltration, and improved tumor control in orthotopic and metastatic lung cancer models.122 Glycosaminoglycan-rich matrices can also be remodeled to facilitate intratumoral delivery. Hyaluronidase-mediated degradation of hyaluronan increased the penetration of pH-responsive shPD-L1-loaded nanoparticles and enhanced PD-L1 silencing in a melanoma model.123 A second strategy is to exploit specific ECM components as targeting or retention sites rather than removing them. A protease-activated bioinspired lipoprotein with affinity for fibronectin and tenascin C showed prolonged intratumoral retention and enhanced antitumor immunity.124 Given the context-dependent functions of the ECM, selective modulation of defined matrix components or interactions may be more appropriate than broad stromal depletion.

Circulating Tumor-Cell Clusters and Metastatic-Site Targeting

Metastatic dissemination extends tumor-associated interactions beyond the primary lesion into the circulation and distant organs. Circulating tumor cells (CTCs) may disseminate as single cells or as multicellular clusters. These clusters can be homotypic, consisting predominantly of tumor cells, or heterotypic, incorporating host cells such as neutrophils and platelets. Such CTC-host cell interactions can enhance survival in the circulation and facilitate metastatic seeding. In breast cancer, CTC-neutrophil clusters exhibit increased proliferative activity and metastatic potential, demonstrating that heterotypic interactions in the bloodstream can actively promote dissemination.125 The ECM can also contribute to CTC clustering. Tumor-derived hyaluronan promotes CTC aggregation through hyaluronan-CD44-dependent cell-ECM-cell interactions and facilitates the recruitment of non-tumor cells into heterotypic clusters.126 Together with the EV-mediated pre-metastatic niche formation discussed above, these circulatory interactions connect tumor-cell dissemination with subsequent colonization of distant organs.

Nanosystems can be designed to interfere with these processes at different stages of metastatic dissemination. Activated neutrophil membrane-coated nanoparticles have been developed as nanodecoys that interfere with neutrophil recruitment and adhesion to CTCs and vascular endothelium, thereby disrupting CTC-neutrophil cluster formation and reducing multiorgan metastasis in mouse models.127 Other platforms act simultaneously on circulating CTC-associated aggregates and pre-metastatic sites. The H@CaPP nanotherapeutic, incorporating piceatannol and low-molecular-weight heparin, inhibits platelet-associated CTC microthrombi and preferentially accumulates in P-selectin-rich pulmonary pre-metastatic sites, thereby reducing metastatic colonization.128 At established metastatic sites, nanosystem delivery is further influenced by the local microenvironment and may differ substantially from that in primary tumors. In a 4T1 metastatic mouse model, clinical-stage core-crosslinked polymeric micelles accumulated in metastatic lesions, and docetaxel-loaded micelles showed greater therapeutic efficacy than conventional docetaxel. Nevertheless, nanoparticle accumulation was lower in metastases than in primary tumors, with increased collagen crosslinking contributing to reduced delivery efficiency.129 These differences indicate that circulating CTC clusters, pre-metastatic sites, and established metastatic lesions present distinct targets and delivery barriers that should be considered separately in the design of antimetastatic nanosystems.

Clinical Translation of Nanosystem-Enabled Intercellular Crosstalk Modulation

Clinical Progress of Crosstalk-Modulating Nanosystems

Clinical translation of crosstalk-modulating nanosystems is beginning to take shape, most visibly through platforms that re-encode or redirect information flow between tumors and the immune system, including systemic nanoformulated vaccines and locally administered nanoinjectables. In the randomized phase 2b KEYNOTE-942 study, adjuvant V940/mRNA-4157, an mRNA-based individualized neoantigen therapy encoding up to 34 patient-specific neoantigens in a lipid nanoparticle formulation, plus pembrolizumab prolonged recurrence-free survival versus pembrolizumab alone in resected high-risk cutaneous melanoma and showed a clinically meaningful improvement in distant metastasis-free survival, a prespecified key secondary endpoint, consistent with enhanced APC–T cell crosstalk driven by endogenous antigen processing and presentation and the resulting increase in endogenous and de novo antitumor T-cell responses (NCT03897881).130 Similarly, an individualized uridine mRNA–lipoplex neoantigen vaccine (autogene cevumeran) was evaluated in an investigator-initiated adjuvant Phase I trial in resected PDAC with sequential atezolizumab followed by vaccination and a modified version of a four-drug chemotherapy regimen (NCT04161755).131 The vaccine induced de novo high-magnitude neoantigen-specific T cells in 8/16 patients, with vaccine-expanded clones reaching up to 10% of circulating T cells and re-expanding after a booster, and these responses may correlate with delayed recurrence. In addition, a related systemic RNA-lipoplex concept has also advanced in melanoma as a first-in-human phase I program (NCT02410733), in which liposomal RNA vaccination targeting non-mutant shared tumor antigens elicited robust and durable antigen-specific cytotoxic T-cell responses in checkpoint-inhibitor-treated patients, in some responders reaching magnitudes typically reported for adoptive T-cell therapy, thereby underscoring the broader feasibility of systemic RNA-lipoplex vaccination to re-engage antitumor T-cell circuits in patients.132

Beyond vaccines, other nanomedicine formats are being evaluated clinically as important tools to modulate intercellular crosstalk for antitumor therapy. In the non-randomized, single-arm iEXPLORE Phase I study, an engineered exosome delivering Kras^G12D-specific siRNA was tested using Phase Ia 3+3 dose escalation followed by Phase Ib accelerated titration (NCT03608631). The treatment was well tolerated, with no treatment-related adverse events, no dose-limiting toxicities, and no maximum tolerated dose reached. Importantly, paired on-treatment biopsies showed evidence of target engagement, including reduced KRAS^G12D DNA and suppressed phospho-ERK, together with increased intratumoral CD8⁺ T-cell infiltration, and these immunological correlates informed subsequent combination testing with immune checkpoint blockade (notably anti-CTLA-4) in aligned PDAC models.133 L-MTP-PE is a liposomal formulation of muramyl tripeptide phosphatidylethanolamine, designed to enable targeted delivery to monocytes/macrophages and to activate these cells toward tumoricidal function. In the metastatic cohort of the randomized Phase III INT-0133 trial (newly diagnosed metastatic osteosarcoma), adding L-MTP-PE to multi-agent chemotherapy produced numerically higher survival estimates, with 5-year event-free survival 42% vs 26% and overall survival 53% vs 40%, although these differences did not reach statistical significance in the metastatic subgroup analysis.134 These clinical studies, summarized in Table 4, demonstrate the emerging clinical feasibility of crosstalk-modulating nanosystems, while also indicating that therapeutic benefit may vary across disease settings. From a translational perspective, current clinical experience indicates that this strategy has gained the clearest momentum in immune-activating platforms, especially RNA-based nanovaccines, whereas EV-, myeloid-, stromal- and endothelial-directed nanosystems remain at an earlier clinical stage, with evidence mainly supporting feasibility, safety, target engagement or immune-context remodeling. Since most clinically tested examples have been developed in combination with immune checkpoint blockade, chemotherapy or other standard therapies, crosstalk-modulating nanosystems may be better positioned as immune-priming or microenvironment-sensitizing components of combination regimens rather than as stand-alone agents. The uneven clinical maturity of these platforms also highlights the importance of addressing in vivo delivery, tracking, manufacturing reproducibility, and regulatory assessment during further translation.

Table 4.

Clinical Translation of Nanosystem-Enabled Intercellular Crosstalk Modulation

Intervention	Crosstalk Node	Disease Setting	Study Design	Key Findings	Ref.
V940/mRNA-4157 (LNP) plus pembrolizumab	Neoantigen presentation to boost APC–T priming	Resected high-risk melanoma, adjuvant	Randomised, open-label phase 2b (KEYNOTE-942)	Improved RFS vs pembrolizumab alone; manageable safety profile.	[130]
Autogene cevumeran (uridine mRNA–lipoplex) with atezolizumab and mFOLFIRINOX	De novo neoantigen-specific T-cell induction	Resected PDAC, adjuvant	Phase I; sequential atezolizumab, vaccine, then mFOLFIRINOX	Tolerable; de novo neoantigen T cells in 8/16; expanded clones up to ~10% of blood T cells and re-expanded after booster; responders had longer RFS in follow-up.	[131]
FixVac (BNT111) intravenous RNA-lipoplex, alone or with PD-1 blockade	Shared-antigen vaccination under CPI pressure	Advanced melanoma, CPI-experienced	First-in-human phase I dose escalation (interim)	Durable objective responses with strong CD4+ and CD8+ immunity to vaccine antigens; in some responders, cytotoxic T-cell magnitudes approached adoptive T-cell therapy.	[132]
iExoKrasG12D engineered exosomes carrying KRASG12D siRNA	KRAS pathway knockdown with immune-context changes	Metastatic PDAC, heavily pretreated	Single-arm phase I dose escalation with accelerated titration	Well tolerated; no DLT and no MTD; reduced KRASG12D DNA and phospho-ERK, with increased intratumoural CD8+ T cells post-treatment.	[133]
L-MTP-PE (mifamurtide) plus chemotherapy	Monocyte/macrophage activation	Newly diagnosed metastatic osteosarcoma	Randomised phase III metastatic cohort analysis	Numerically higher 5-year EFS and OS, but not statistically significant in this subgroup.	[134]

In vivo Behavior, Tracking, and Clinical Translation of Nanosystems

Following systemic administration, nanosystems encounter biological environments that can substantially alter their intended behavior. Adsorption of plasma proteins can generate a protein corona that changes nanoparticle biological identity, cellular recognition, biodistribution, and tumor delivery.135 Nanoparticles may also be recognized and cleared by the mononuclear phagocyte system, resulting in off-target accumulation, particularly in the liver and spleen, and reducing the fraction available for tumor delivery.136 Moreover, tumor accumulation is not uniform across tumors, patients, or even individual lesions. Clinical positron emission tomography (PET) imaging with ⁶⁴Cu-labeled MM-302 revealed marked variability in nanoparticle deposition across patients and among lesions within patients, illustrating the heterogeneity of the enhanced permeability and retention effect (EPR effect).137 These limitations are especially relevant to crosstalk-modulating nanosystems, whose biological effects depend not only on reaching the tumor, but also on gaining sufficient access to the specific cellular compartment in which the intended communication pathway operates.

This variability also makes reliable in vivo tracking important for determining where a nanosystem distributes and whether it reaches its intended target. Fluorescence-based imaging, Förster resonance energy transfer (FRET) approaches, and radiolabeled PET/CT have been used to assess different aspects of nanoparticle biodistribution and integrity. However, the detected signal does not always faithfully represent the behavior of an intact nanosystem. Fluorescent labeling itself can alter nanoparticle biodistribution,138 while FRET-based imaging has demonstrated that nanoparticles may dissociate after systemic administration, such that signals derived from individual labels do not necessarily indicate intact particles.139 In patients, PET/CT with ⁸⁹Zr-labeled CPC634 enabled quantitative assessment of nanoparticle tumor accumulation and revealed substantial heterogeneity among tumor lesions.140 For crosstalk-modulating nanosystems, whole-body biodistribution should therefore be complemented, where feasible, by tissue- or cell-level measurements confirming delivery to the intended target population, together with appropriate assessment of nanosystem integrity and target engagement.

Clinical translation further depends on reproducible dose-exposure-effect relationships, well-defined safety windows, and consistent manufacturing. Exposure and dosing regimens should be characterized sufficiently to assess the consequences of repeated administration, while key formulation attributes should be linked to biological performance to minimize batch-dependent variability. Crosstalk-modulating nanosystems may incorporate immune-active cargos, targeting components, or multiple functional modules, increasing the demands on manufacturing comparability and regulatory assessment of nano-specific attributes.141 Current regulatory guidance for drug products containing nanomaterials emphasizes appropriate physicochemical characterization, manufacturing controls, stability assessment, and evaluation of how nanomaterial attributes relate to product quality, safety, and efficacy.142 A practical translational approach is therefore to build on scalable carriers with established manufacturability while introducing well-characterized crosstalk-modulating components whose biological effects can be measured. Clinical studies should also incorporate mechanism-aligned pharmacodynamic readouts and, where appropriate, biomarker- or imaging-informed patient selection to determine whether adequate nanosystems exposure and modulation of the intended communication axis are achieved in patients.

Conclusion and Future Perspectives

Intercellular communication in the TME operates as a coupled network coordinating tumor, immune, stromal and vascular compartments.143 These interactions can skew APC programs, reshape T-cell priming and effector function, and reinforce macrophage–CAF circuits that remodel extracellular matrix, alter drug access and sustain therapy-adapted states.144 Tumor endothelium further regulates vascular permeability, immune-cell trafficking and nutrient delivery, thereby shaping both immune context and tumor growth constraints.145 Beyond these primary-tumor cellular circuits, matrix architecture and metastasis-associated microenvironments, including circulating tumor-cell clusters and distant metastatic niches, further shape immune access, therapeutic delivery, and metastatic dissemination. Given the unique features of the TME, nanosystems are well suited to intervene at this level because they can concentrate modulators in the niches where these interactions occur, align exposure across the relevant cell populations, and tune release kinetics so that communication reprogramming happens within a therapeutically useful window.146 Accordingly, nanosystem selection should be matched to the therapeutic cargo, target-cell population, and communication context. In this Review, we distil the key principles of intercellular communication in the TME and highlight how nanosystems can target specific cellular compartments to perturb critical communication nodes, thereby influencing tumor progression. The biological and translational barriers to these strategies, including heterogeneous in vivo delivery, tracking challenges, manufacturing reproducibility, and regulatory requirements, also remain important considerations. By reframing the TME as a communication network rather than a collection of isolated cell types, this perspective provides a rational basis for designing interventions that overcome microenvironmental barriers and improve the durability of anticancer responses.

Recent advances in single-cell RNA sequencing (scRNA-seq) and spatial transcriptomics provide complementary approaches for resolving intercellular communication in the TME. scRNA-seq resolves cellular heterogeneity and cell-state-specific gene-expression programs, whereas spatial transcriptomics preserves tissue organization and helps determine whether putative sender and receiver populations occupy the same microenvironment. Computational frameworks such as CellChat and CellPhoneDB can use these data to infer and compare candidate ligand-receptor communication networks, with recent versions extending comparative analysis and incorporating spatial or multimodal information.147,148 Integrative frameworks such as LIANA+ and FlowSig further support communication inference and link extracellular signals to downstream cellular programs.149,150 However, inferred interactions remain sensitive to the computational method, interaction database, and transcript-level assumptions, and systematic comparisons have shown substantial variation among communication-inference approaches.147 These tools should therefore be used primarily to identify and prioritize candidate communication nodes rather than as direct evidence of mechanism.

Establishing that a nanosystem actually modulates such a candidate communication node requires causal validation. A rigorous framework should establish a sequential relationship between nanosystem action, target-cell engagement, modulation of a defined communication node, and the resulting immunotherapeutic response. Once delivery to the intended cell population has been demonstrated, causal evidence should directly establish modulation of the relevant signaling or communication pathway, followed by tests of necessity using pharmacological blockade or genetic perturbation and, where feasible, functional rescue. Spatial and longitudinal profiling can further determine where and when this causal sequence operates during treatment rather than relying on changes in cell abundance or cytokine levels at a single endpoint. Human single-cell and spatial profiling has already revealed early divergence between treatment-sensitive and resistant tumor neighborhoods during immunotherapy.151 Such validation would help distinguish genuine nanosystem-mediated crosstalk reprogramming from secondary changes in cellular composition or inflammatory state.

These experimentally validated communication nodes can then be translated into biological design requirements for nanosystems, defining which cell population should be reached, which communication signal should be modulated, and when the intervention is needed. Artificial intelligence (AI)-assisted approaches could facilitate this process by accelerating formulation screening and optimization for cell-selective delivery; deep-learning-guided lipid nanoparticle development, for example, has been used to screen large ionizable-lipid libraries and identify formulations with distinct cell-specific transfection preferences.152 Spatial omics may further inform nanosystem design by identifying tumor regions or patient subgroups in which specific communication circuits are most relevant. Integrating these mechanistic and spatial features with data-driven formulation design may therefore support precision nanomedicine and more personalized immunotherapy. Prospective studies will be needed to determine whether such omics-guided and AI-based approaches can improve nanosystem performance and therapeutic outcomes in clinically relevant settings.

Funding Statement

This work was financially supported through grants from the Zhejiang Provincial Natural Science Foundation of China under Grant No. ZCLQN26H1802, the medical and health research project of Zhejiang Province (2025KY955 and 2024KY435), Zhejiang Vanguard and Leading Goose + X Science and Technology Program Project (2025C02090), Jiaxing Public Welfare Research Program Project (2024AY30004), Zhejiang Provincial Clinical Key Specialty Development Project (2024-ZJZK-002).

Abbreviations

AC-NPs, antigen-capturing nanoparticles; AI: artificial intelligence; ACT-DC, antigen capuring nanoparticle transformed dendritic cell therapy; AFP, alpha-fetoprotein; APCs, antigen-presenting cells; BBB, blood-brain barrier; BCSC, breast cancer stem cell; CAFs, cancer-associated fibroblasts; cDC1s, type 1 conventional dendritic cells; CRC, colorectal cancer; CTCs, circulating tumor cells; DCs, dendritic cells; DDR1, discoidin domain receptor 1; ECM, extracellular matrix; ECs, endothelial cells; ENT1, equilibrative nucleoside transporter 1; EVs, extracellular vesicles; FAP, fibroblast activation protein; FRET, Förster resonance energy transfer; HCC, hepatocellular carcinoma; HGF, hepatocyte growth factor; i.p., intraperitoneal injection; I.T., intratumoral injection; i.v., intravenous injection; LNPs, lipid nanoparticles; MCT11, monocarboxylate transporter 11; MPs, microparticles; NLCs, nanostructured lipid carriers; NPs, nanoparticles; OSM, oncostatin M; OVA, ovalbumin; PCD, nanoscale protein condensate; PDAC, pancreatic ductal adenocarcinoma; PDT, photodynamic therapy; PET/CT, positron emission tomography/computed tomography; PMN, pre-metastatic niche; ROS, reactive oxygen species; S.C., subcutaneous injection; scRNA-seq, single-cell RNA sequencing; siRNA, small interfering RNA; sEVs, small extracellular vesicles; TAMs, tumor-associated macrophages; TECs, tumor endothelial cells; TEVs, tumor-derived extracellular vesicles; TME, tumor microenvironment.

Ethics Approval and Consent to Participate

The ethics statement is not applicable.

Author Contributions

All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

The authors declare that they have no competing interests.