Tumor cell-intrinsic mechanisms are important but insufficient to explain the full pattern of therapeutic failure. Genomic instability contributes to tumor progression and treatment resistance (5), yet several persistent clinical observations remain incompletely explained (1,2,6). Immune infiltration may occur without durable tumor clearance (7,8), technically achievable local control may not prevent pulmonary recurrence (1,2), and immunotherapies directed at distinct proximal targets have generally produced limited and inconsistent clinical benefits (3,6,9,10).

This interpretation does not diminish the importance of tumor cell biology. Rather, it suggests that assigning explanatory primacy to tumor cell-intrinsic mechanisms alone may be insufficient to account for immune dysfunction and delayed pulmonary recurrence (4).

The skeletal niche may contribute earlier and more actively than implied by purely cancer cell-centric models (17,24). Rather than providing a detailed structural account in this section, the niche is used as an explanatory bridge: Immune infiltration can coexist with ineffective tumor clearance, local control can be followed by pulmonary recurrence and different immune-based therapies can encounter convergent microenvironmental constraints. These observations suggest that the relevant failure may involve not only malignant cell programs but also the spatial and cellular context in which antitumor immunity is organized. The structural, myeloid, efferocytic, EV-related, metabolic and pulmonary modules of this context are therefore considered in the following sections (Fig. 1).

The bone niche not only provides the anatomical setting for OS growth and dissemination but also actively shapes the tumor microenvironment (TME), which may progressively acquire an immune-restricted and tumor-permissive state (4,24). Mineralized matrix and osteoid-rich architecture may restrict the spatial access of immune cells and affect drug distribution in certain contexts, whereas osteolytic remodeling may release matrix-bound growth factors that support malignant adaptation (4,24,25). Abnormal vasculature and mechanical stress may further promote invasion, repair-associated remodeling and immune tolerance (14,25). Thus, normal skeletal architecture may be converted into a structural barrier that limits early therapeutic efficacy.

Beyond these structural constraints, the bone marrow compartment adds an additional regulatory layer (26). Myeloid progenitors, macrophages, osteoclast precursors and immunomodulatory MSCs normally maintain skeletal homeostasis, tissue repair and regenerative remodeling (24). However, within the microenvironment of OS, these homeostatic programs may be co-opted to promote tumor persistence, immune evasion and stromal maintenance. Therefore, OS might exploit the intrinsic bone remodeling capacity rather than simply appear in defective tissue.

Such an interpretation is consistent with bone-mimetic experimental models (17). In co-culture systems comprising osteoid-like scaffolds and MSCs, the bone-like environment can promote OS cell proliferation, activate stemness-associated programs, stimulate matrix production and remodel the extracellular matrix (ECM) (17). These results suggest that the niche is not biologically passive. Instead, osteoid-like structures and stromal interactions may exert selective microenvironmental pressures that reinforce malignant adaptation and promote an invasive, immune-evasive tissue state (17).

Osteogenic and osteoclast-lineage cells coordinate bone modeling and remodeling via the receptor activator of nuclear factor κB ligand (RANKL)/RANK/osteoprotegerin system (48) under physiological conditions. This homeostatic program may be reprogrammed to support tumor signaling, immune regulation and metastatic progression in OS (24,25). This implies that biological processes that, under normal conditions, preserve skeletal integrity may, under malignant stress, contribute to tumor maintenance.

Bone niche remodeling is one of the more mechanistically well-supported anchors of the model (4,11–14,17,24,25) among the modules considered within the present review. It proposes a plausible link among structural constraints, myeloid accumulation, altered immunological handling of dying tumor cells, EV release, metabolic stress and conditioning of the pulmonary niche.

This integrative sequence is a synthesis of evidence across distinct modules rather than a single experimentally established cascade (17–19,21,33). Beyond a local anatomical context, the remodeled bone niche may connect tissue architecture to systemic relapse mechanisms, thereby linking skeletal turnover, immunosuppressive remodeling and metastatic preparation within a sustained pattern of treatment failure.

OS does not arise within an immunologically inert tissue bed (49). Compared with several soft-tissue malignancies, it develops in an actively remodeling bone marrow compartment enriched in myeloid progenitors, bone marrow macrophages, osteoclast precursors and immunomodulatory MSCs (28). Myeloid cell enrichment may therefore be embedded within the baseline cellular architecture of the OS niche rather than merely representing a late consequence of tumor progression (49).

Within this bone marrow niche, tumor-associated macrophages (TAMs) occupy a potentially consequential regulatory position (29). OS lesions are frequently macrophage-enriched, and TAMs participate in antigen processing, inflammatory regulation, matrix remodeling, angiogenic remodeling and adaptive immune regulation (30). The function of TAMs cannot, therefore, be reduced to the conventional label of ‘immunosuppressive cells’. More precisely, TAMs constitute a multifunctional regulatory population through which skeletal remodeling, impaired antitumor immunity, stromal adaptation and therapeutic failure may become biologically interconnected (29).

The binary M1/M2 dichotomy is also likely to be overly restrictive in this setting (50). Macrophages occupy a functional continuum, an observation that is particularly relevant in OS, where macrophage-rich and immune-excluded niches have been identified (8,11,13,27,31). TAM-fibroblast crosstalk, antigen-presenting macrophage states, and perivascular macrophage or endothelial-associated niches may alter T-cell localization, activation thresholds and exhaustion-like phenotypes (11,13,28,51,52). Macrophages may therefore reshape the spatial and functional organization of antitumor immunity rather than uniformly suppress it.

This interpretation has therapeutic implications (53). In OS, T-cell dysfunction may arise less from an autonomous defect in T cells than from a surrounding microenvironment dominated by suppressive myeloid, stromal, vascular and metabolic constraints (8). Even when effector T cells are present, their spatial access to tumor cells and their capacity to form productive cytotoxic contacts may remain limited (8). Impaired antigen presentation, myeloid-derived inhibitory signals, stromal immune exclusion, aberrant vasculature and metabolic deprivation may collectively reduce cytotoxic efficacy. Failure of T-cell-directed therapy may therefore reflect persistent myeloid-dominated tissue regulation and insufficient downstream immune activation (50).

TAMs may consequently represent a rational therapeutic entry point, particularly in myeloid-rich tumors (28,29,31,49). This rationale is supported by OS-specific datasets and by broader immuno-oncology evidence linking macrophage states to immune exclusion, treatment resistance and metastatic progression (50,51). The importance of the TAM axis lies in its ability to integrate local immunosuppression, dying cell clearance, extracellular communication, matrix remodeling and mechanisms of systemic recurrence (18,28,54). Targeting TAM recruitment or state may therefore provide a mechanistically coherent strategy when downstream T-cell activation remains constrained by a myeloid-dominated suppressive environment (49,50).

One candidate process through which myeloid remodeling may convert therapy-induced tumor cell death into immune tolerance is efferocytosis. Efferocytosis is the phagocytic clearance of apoptotic cells by macrophages and other phagocytes (55). Given that direct causal evidence in OS remains limited, this process is treated here as an emerging extension of the myeloid-remodeling framework rather than as an established mechanism.

Chemotherapy, checkpoint blockade and engineered cellular therapies encounter distinct but partly convergent microenvironmental limitations (3). Chemotherapy can reduce tumor burden while leaving regenerative and immunosuppressive networks intact. Checkpoint blockade can attenuate inhibitory signaling but may act too late in immune-excluded or myeloid cell-predominant lesions (8). Chimeric antigen receptor (CAR)-T and CAR-NK cells provide antigen-specific cytotoxicity, yet their homing, persistence and effector function remain constrained by transforming growth factor-β (TGF-β) signaling, suppressive myeloid circuits, structural barriers, and physical or metabolic stress within the TME (64). These shared barriers suggest that therapeutic failure may occur upstream of final cytotoxic engagement.

These observations can be organized into a staged translational logic rather than a catalogue of experimental strategies. The first stage, niche reprogramming, aims to relieve dominant myeloid, stromal, EV-related or metabolic constraints before or alongside immune activation. Representative approaches include TAM recruitment blockade or state reprogramming, Gas6-MerTK/AXL-axis inhibition, TGF-β or stromal modulation, metabolic immunomodulation, and exploratory EV-directed strategies (18,19,21,28,31,35–37,49,58,65). These interventions should be interpreted as candidate approaches for testing whether niche-directed modulation can improve immune trafficking, antigen presentation and effector-cell function, rather than as established therapeutic requirements.

The second stage, immune activation, would deploy checkpoint blockade, B7 homolog 3 (B7-H3)-directed therapy, CAR-T/CAR-NK platforms, organoid-guided T-cell reconstruction or other engineered cellular approaches after at least partial relief of upstream constraints (3,64,66,67). This timing is intended to reduce the mismatch between downstream cytotoxic activation and an upstream immune-restrictive niche, although antigen heterogeneity, inefficient homing, limited persistence, stromal exclusion, metabolic stress and suppressive myeloid populations remain major barriers (64,67,68).

The third stage, pulmonary niche maintenance, focuses on high-risk relapse settings after local treatment or during minimal residual disease surveillance. EV/chemokine monitoring, lung-directed immune support, CAR macrophages generated in situ using EV-delivered mRNA, inhaled NK cell-supportive platforms and postoperative immunomodulatory strategies may be considered as investigational approaches, but this stage remains predominantly preclinical or conceptual (19,33,63,68,69). The preferred endpoint should shift from short-term tumor shrinkage alone to suppression of micrometastatic persistence and improvement in pulmonary recurrence-free survival.

Clinical implementation would require careful patient selection, explicit timing and predefined biological readouts. Candidate populations may include patients at high risk of perioperative pulmonary recurrence, patients with primary resistance to T-cell-directed immunotherapy, and patients whose tumors show myeloid enrichment, abnormal EV or chemokine profiles, or spatially organized suppressive niches (8,11,19,33,61,62,70). EV cargo, CCL5/CXCL10 profiles, macrophage-distribution patterns and spatial immune features may help identify candidates for niche reprogramming or pulmonary niche maintenance, but these biomarkers require prospective validation.

Accordingly, the proposed sequence should be regarded as a trial-design framework rather than a clinically established algorithm. Future studies should map candidate interventions to the phase in which they are most biologically plausible: Niche reprogramming before or during cytotoxic injury; immune activation after partial relief of upstream constraints; and pulmonary niche maintenance during postoperative surveillance or in high-risk minimal residual disease. Myeloid-enrichment biomarkers, spatial immune profiling, EV/chemokine monitoring, pulmonary recurrence endpoints and safety outcomes should be incorporated rather than relying solely on tumor-size response (Fig. 3; Table II).

Efferocytosis should be explored as a potential proximal cause of immune suppression in specific OS cases (18,35). The impact of blocking Gas6-MERTK/AXL or related pathways on macrophage status, antigen presentation, PD-L1 expression and T-cell activation should be studied. Additionally, metabolism-associated decreases in immune recognition may exist in certain types of tumors; however, this is one of the least established sections of the model. Studies investigating mitophagy, BNIP3-mediated adaptability, ER stress or metabolic buffering should directly assess immune recognition, rather than merely tumor proliferation (44–47).

Third, the risk of pulmonary recurrence may potentially be classified prior to radiographic development. EV cargo, CCL5, CXCL10, spatial macrophage patterns and other niche-related traits are possible biomarkers. Prospective validation in cohorts with specific pulmonary outcomes is needed to evaluate their predictive value (19,33,61,62).

Fourth, spatial organization may be more informative than mean expression levels. 3D spatial transcriptomics and multimodal imaging can uncover compact suppressive niches of macrophages, fibroblasts, endothelial cells and defective lymphocytes (11,52,71). Such microenvironments may be better predictors of the response to therapy than bulk transcriptome averages; however, prospective predictive validation is yet to be shown. This approach can be operationalized using a claim evidence gap validation matrix (Table III) (72–77). A minimal validation strategy for a bone niche-driven framework should include paired pre- and post-treatment primary tumor specimens, circulating EV and chemokine profiling during neoadjuvant therapy and postoperative surveillance, matched pulmonary metastatic or recurrent specimens when available, and spatial mapping of myeloid, stromal, endothelial and T-cell compartments. In parallel, candidate pathways such as TAM recruitment, MerTK/AXL signaling, EV cargo transfer and metabolic stress adaptation should be tested through cell type-resolved perturbation models. The clinical relevance of this framework would be strengthened by predefined pulmonary recurrence endpoints and external validation in an independent cohort. Evidence would be particularly compelling if niche-directed disruption altered immune trafficking, antigen presentation, effector-cell function and pulmonary recurrence risk beyond changes in tumor burden alone.

The present review argues that therapeutic resistance in OS and pulmonary relapse requires a tissue-specific explanation that extends beyond tumor cell intrinsic mechanisms. The bone niche provides a testable architecture in which structural barriers, marrow-derived myeloid enrichment, treatment-induced repair, extracellular communication, metabolic stress and pulmonary surveillance converge.

The evidence hierarchy is deliberately separated from therapeutic maturity. Bone niche remodeling and myeloid cell enrichment are the most mature OS-supported anchors; pulmonary niche conditioning, efferocytosis and EV-mediated bone-lung communication have growing but incomplete longitudinal or functional support, and physical/metabolic adaptation remains a lower-tier extension.

Accordingly, the framework should be read as a falsifiable model, not a fixed cascade; its modules may operate in parallel or only in selected molecular, immune, age-related or metastatic contexts, with Table III defining the claim evidence validation boundaries.

OS is biologically and clinically heterogeneous; therefore, the proposed bone niche-driven framework should not be assumed to operate uniformly across all patients. Molecular subtypes, transcriptional states, immune infiltration patterns, age-related skeletal and immune contexts, chemotherapy response and metastatic trajectory may all influence the contribution of the bone niche to antitumor immune failure (11,13,27,70,78). Tumors with myeloid-enriched, immune-excluded or stromal-activated phenotypes may depend more on macrophage-centered regulation and niche-mediated immune restraint, whereas other tumors may be driven predominantly by tumor cell intrinsic genomic instability or therapy-resistant clonal evolution.

Pediatric, adolescent and adult patients may also differ in marrow composition, bone-remodeling activity, immune competence and treatment tolerance, which may modify the strength and timing of niche-mediated immune regulation. Similarly, chemotherapy-sensitive and -resistant tumors may differ in the immunological consequences of therapy-induced cell death, with dying tumor cells being processed through more immunogenic pathways in some contexts and through tolerogenic repair or efferocytosis-associated programs in others. Finally, localized disease, synchronous pulmonary metastasis and delayed pulmonary recurrence may reflect different degrees of primary bone lesion-lung communication. Therefore, the present framework should be viewed as a context-dependent and testable model that helps organize heterogeneous immune-failure patterns rather than as a universal mechanism shared by all patients with OS.

Bone niche reprogramming can be viewed as a translational extension of the proposed framework. In the present review, bone niche reprogramming refers to the attempt to shift the OS microenvironment from a structurally constrained, myeloid-enriched, stromal-protective and metabolically suppressive state toward a more immune-permissive and therapeutically responsive state. This concept does not imply that niche-directed intervention should replace established or emerging immunotherapies. Rather, it suggests that the efficacy of immune activation strategies may depend, at least in part, on whether upstream bone niche constraints have been relieved.

This concept is particularly relevant to current immunotherapy approaches in OS. Immune checkpoint blockade may be insufficient when effector T cells remain spatially excluded, functionally exhausted or exposed to persistent myeloid and stromal inhibitory signals (3,6,8). B7-H3-targeted therapy and CAR-T or CAR-NK strategies may also be limited by poor homing, reduced persistence, antigen heterogeneity, TGF-β-rich stromal niches, metabolic stress and suppressive macrophage or MDSC populations (64,67). Bone niche reprogramming may therefore provide a rational preparatory step before immune activation. Candidate approaches include TAM repolarization or recruitment inhibition, Gas6-MERTK/AXL blockade, TGF-β mitigation, matrix or vascular normalization, EV-directed intervention and metabolic monosensitization (18,21,35–37,49,58,65). These strategies may help convert an immune-restricted bone lesion into a state in which immune-activating treatments can act more effectively, but most remain preclinical or early translational.

The same logic may apply in preventing pulmonary relapses. If the primary bone lesion releases EVs, chemokines and myeloid-recruiting signals that contribute to pulmonary niche conditioning, bone niche reprogramming may need to be considered together with pulmonary niche maintenance in selected high-risk patients (19,33). Local or regional lung-directed immune strategies, including CAR macrophages generated through EV-delivered mRNA and inhaled NK cell-supportive platforms, warrant investigation as relapse-prevention strategies (68,69). However, these approaches remain preclinical or early translational and should be evaluated in biomarker-guided perioperative studies, window-of-opportunity trials and longitudinal cohorts incorporating spatial immune readouts, EV/chemokine monitoring and pulmonary recurrence endpoints.

Prespecified studies should test whether niche-directed perturbation improves immune trafficking, antigen presentation, effector-cell function or pulmonary niche quiescence beyond tumor shrinkage alone. Failure to change these readouts after adequate pathway perturbation would restrict the model to specific phenotypes, windows or metastatic contexts.

A perioperative window study is the most direct first test: Paired pre- and post-neoadjuvant specimens should be profiled for TAM state, MerTK/AXL activity, PD-L1 induction, antigen-presentation markers, T cell-tumor spatial contact, stromal barriers and circulating EV/chemokine signals.

A primary-lung paired cohort should integrate primary tumor, serial blood EV/chemokine data, and pulmonary metastatic or recurrent tissue to determine whether primary bone niche states predict pulmonary myeloid recruitment, stromal activation, micrometastatic persistence and pulmonary recurrence-free survival.

Functional perturbation models should pair TAM modulation, Gas6-MERTK/AXL blockade, EV cargo perturbation, TGF-β or matrix modulation, and metabolic-stress intervention with antigen-presentation assays, CD8⁺ T-cell and NK cell function, killing assays, and in vivo lung recurrence endpoints.

Table IV (72–77) translates the framework into three practical settings. In poor histological responders after neoadjuvant chemotherapy, perioperative profiling of TAM state, MerTK/AXL activation, PD-L1, tumor cell MHC and CD8⁺ T-cell proximity could determine whether niche reprogramming should precede or accompany immune activation.

In high-risk localized disease after definitive surgery, serial EV cargo, CCL5/CXCL10, myeloid signatures and spatial features from the resected tumor could support pulmonary recurrence-prevention studies that use pulmonary recurrence-free survival rather than short-term shrinkage as the primary endpoint.

In pulmonary micrometastatic disease or early lung recurrence, the lung immune state, systemic EV/chemokine communication and cytotoxic cell function should be assessed alongside lung-directed immune support or engineered cell approaches, with predefined monitoring for pulmonary toxicity, immune overactivation, infection and impaired repair.

The present review has limitations. As a mechanistic narrative review rather than a systematic review or meta-analysis, it was designed to construct and stress-test a conceptual framework, not to capture every eligible study, pool effects or formally grade risk of bias. Selection and evidence weighting, therefore, involve author judgment, and the proposed hierarchy should be viewed as a practical organizational tool rather than a validated grading system. The maturity of evidence is also uneven: Bone niche remodeling and myeloid cell enrichment have broader OS support, while efferocytosis, EV-mediated bone-lung communication and metabolic immune adaptation are more reliant on transcriptomic, computational, spatial-inference or limited functional data. Heterogeneity between patients, models, ages, treatment settings and illness stages further limits generalizability. Future validation will need a longitudinal sample, coupled with primary lung specimens, independent replication, lineage-resolved perturbation, clinically relevant immune readouts and predefined pulmonary recurrence endpoints.