Systematic discovery of immunomodulatory plant-derived nanoparticles reveals RNA-mediated macrophage reprogramming Authors: Dokyung Jung, Ju-Hyun Bae, Il-Young Jung, Na-Eun Kim, Jung Min Sung, Kyung-Won Doh, Minju Kim, Jiwoong Choi, Eunji Ha, Hee Jin Cho, Byung Sun Min, Jong Hyun Kim, Sin-Hyeog Im, Moon-Chang Baek Published: 2026 ABSTRACT AbstractPlant-derived nanoparticles (PNPs) have emerged as a promising class of bioactive nanomaterials with intrinsic immunomodulatory properties, offering a sustainable and biocompatible alternative to synthetic delivery systems. However, their systematic exploration has been hindered by a fragmented “one-plant-one-function” research paradigm. Here, we establish a comprehensive bio-discovery platform by constructing a taxonomically diverse library of 110 PNPs. Through integrated physicochemical characterization and high-throughput functional screening, we identify Allium tuberosum-derived nanoparticles (ANPs) as a potent bioactive nanoparticle that effectively promotes pro-inflammatory macrophage polarization. Mechanistically, we demonstrate that the bioactivity of ANPs is largely mediated by their endogenous RNA cargo, specifically ANP-miR-488, which modulates the Bcl6 signaling axis to initiate macrophage reprogramming. Furthermore, ANPs significantly sensitize cold tumors to immune checkpoint blockade therapy in a syngeneic LLC1 lung carcinoma model. This study not only provides a generalizable strategy for the systematic discovery of bioactive nanomaterials but also highlights the clinical potential of RNA-mediated PNPs as a promising platform for enhancing cancer immunotherapy.Keywords: Plant-derived nanoparticles, Bioactive nanomaterials, RNA-Mediated immunomodulation, Macrophage reprogramming, Cancer immunotherapy Introduction Nanomaterials with defined nanoscale architectures have enabled diverse applications in drug delivery, imaging, and immunotherapy [[1], [2], [3], [4]]. In synthetic systems, rigorous control over chemical composition, structural geometry, and surface functionalization has enabled the engineering of specific functional properties into nanoscale carriers [[5], [6], [7]]. However, these design strategies have been largely restricted to engineered synthetic nanomaterials, while naturally occurring nanoscale systems have rarely been systematically explored as bioactive nanomaterial platforms [8,9]. Nature-derived nanomaterials offer a sophisticated alternative where nanoscale organization is shaped by biological evolution, providing structurally complex and highly biocompatible systems with inherent functional potential [10,11]. Among these, plant-derived nanoparticles (PNPs) have recently emerged as a promising class of bioactive nanomaterials, characterized by high biocompatibility, low toxicity, and a diverse repertoire of molecular cargo [8,[12], [13], [14]]. These PNPs are enriched with endogenous lipids, proteins, and nucleic acids and can be sustainably produced at high yields from renewable botanical sources [15,16]. Emerging evidence suggests that PNPs function not only as inert delivery vehicles but also active modulators capable of modulating complex biological processes, including targeted drug delivery [[17], [18], [19]], inflammation regulation [[20], [21], [22], [23]], and cancer therapy [[24], [25], [26], [27]]. Despite these rapid advances, current investigations into PNPs remain largely phenomenological and fragmented [28,29]. Most studies focus on a single plant species selected based on empirical or ethnopharmacological knowledge, following a restrictive “one-plant-one-function” paradigm reminiscent of early natural product research [30]. This conventional approach limits systematic comparison across diverse nanoparticle populations, obscures fundamental structure-function relationships, and reduces the probability of identifying rare but potent bioactive candidates. Consequently, PNPs have not yet been systematically categorized as a unified class of bioactive nanomaterials, and their underlying mechanistic principles remain poorly defined [31]. A critical challenge lies in the lack of standardized nanoparticle libraries and systematic discovery strategies comparable to those utilized in synthetic nanomaterials science [32]. Furthermore, although PNPs are frequently investigated as delivery systems, their intrinsic bioactivity, particularly that mediated by endogenous RNA cargo, remains insufficiently characterized [8,[33], [34], [35]]. It remains unclear whether the observed biological effects arise from generic plant metabolites or from the nanoscale organization and nanoparticle-associated cargo. In this study, we establish a systematic discovery platform for PNPs by constructing a taxonomically diverse library comprising 110 plant species using a standardized workflow. By integrating physicochemical characterization with functional immunological screening, this approach enables unbiased identification of PNPs beyond the conventional one-plant-one-function paradigm. Using this platform, we identify Allium tuberosum-derived nanoparticles (ANPs) as a potent immunomodulatory PNP that promotes macrophage polarization toward a pro-inflammatory phenotype. Mechanistic analyses demonstrate that this activity is mediated by RNA cargo, particularly ANP-miR-488, which regulates the Bcl6 signaling axis to drive macrophage reprogramming. Furthermore, we show that ANPs significantly sensitize the LLC1 tumor model to immune checkpoint blockade. Collectively, this study provides a generalizable framework for the systematic discovery of bioactive nanomaterials from natural sources and highlights the potential of PNPs as a promising platform for cancer immunotherapy. Results and discussion Library construction and physicochemical mapping of PNPs To enable systematic discovery of bioactive nanomaterials, we constructed a comprehensive library of PNPs encompassing 110 botanically diverse plant species across 34 taxonomic families (Fig. 1A and Table S1). Unlike conventional studies focusing on individual plants, we deliberately designed this comprehensive library as a comparative nanomaterial platform to systematically investigate complex structure-function relationships across a diverse range of species. PNPs were isolated from multiple plant tissues, including leaves, aerial parts, fruits, and roots, thereby capturing intrinsic variability in nanoscale architecture and molecular organization (Fig. S1). To ensure that functional differences arise from intrinsic nanoparticle properties rather than isolation artifacts, we implemented a standardized, low-shear extraction and differential centrifugation workflow (Fig. 1B). This protocol enabled reproducible fabrication of nanoscale carriers across all plant sources, independent of tissue texture or water content. Construction and physicochemical characterization of the PNP library. (A) Taxonomic analysis of the 110 plant species comprising the PNP library. The library includes 59 edible plants and 51 medicinal plants, spanning 34 families, with prominent representation from Asteraceae, Brassicaceae, and Fabaceae. Within the library, certain species include distinct cultivars (e.g., Capsicum annuum varieties) or specific plant tissues (e.g., Senna tora seed, stem, leaf). These entries are indicated as extended segments in the outer ring. (B) Schematic overview of the PNP isolation protocol applied uniformly across 110 plant species. The process consists of low-speed cold-press extraction to preserve nanoparticle integrity, followed by purification via differential centrifugation. All centrifugation steps were conducted at 4 °C. (C) Representative transmission electron microscopy (TEM) images of eight PNP species, randomly selected across plant tissue categories. The images show consistent vesicle-like nanoparticle structure across diverse tissue origins. Scale bar, 200 nm. (D, E) Nanoparticle tracking analysis (NTA) of the PNP library stratified by tissue origin. Violin plots illustrate the distribution of (D) mean particle diameter (nm) and (E) particle yield (particles per gram of plant tissue) measured by NTA. Data points represent individual plant species, with colors indicating medicinal (red) and edible (blue) plants. Comprehensive physicochemical characterization revealed that the resulting PNPs form a well-defined nanomaterial library. Transmission electron microscopy (TEM) confirmed the characteristic vesicle-like morphology across representative species (Fig. 1C). To accurately evaluate the structural integrity of the isolated vesicles, we performed high-resolution cryogenic electron microscopy (Cryo-EM) analysis on a representative sample (Allium tuberosum-derived nanoparticles, ANPs). As a result, the cryo-EM data clearly revealed an intact, liposome-like nanoscale structure with a distinct lipid bilayer (visible outer and inner leaflets, Fig. S2). In parallel, nanoparticle tracking analysis (NTA) demonstrated a consistent nanoscale size distribution with an average diameter of 133.5 ± 27.6 nm across the entire library (Fig. 1D and Fig. S3). Although particle yield varied among species, reflecting biological differences in tissue composition, the recovered quantities were sufficient for library-scale screening in all cases (Fig. 1E). The majority of PNPs exhibited narrow Span values ranging from 0.8 to 1.4, demonstrating high size monodispersity and a uniform distribution profile (Fig. S4). Importantly, batch-to-batch analysis across multiple plant species confirmed high reproducibility in nanoparticle recovery (Fig. S5), establishing the robustness of the isolation protocol as a reproducible preparation method rather than a plant-specific biological extraction. Collectively, this physicochemical mapping defines the PNP library as a scalable bioactive nanomaterial platform, providing a reliable foundation for unbiased discovery of bioactive nanomaterials. By decoupling nanoparticle material properties from plant-specific anecdotal effects, this library enables systematic exploration of botanical diversity, analogous to synthetic nanomaterial libraries used in systematic materials discovery. Functional screening as materials discovery To enable systematic discovery of functional PNPs as active bio-nanomaterials, we employ interferon-γ (IFN-γ) production as a quantitative functional screening readout [36,37]. IFN-γ represents an emergent, integrative output of multicellular immune crosstalk and thus serves as a sensitive proxy for identifying nanomaterials capable of modulating complex biological responses, rather than merely interacting with isolated cell types. Primary murine splenocytes isolated from C57BL/6 mice, which preserve endogenous interactions among macrophages, dendritic cells, and lymphocytes, were employed as a systems-level screening matrix. Each PNP preparation from the 110-species library was evaluated under standardized conditions normalized to particle number to ensure comparability across botanically diverse sources (Fig. 2A). This design enabled quantitative mapping of functional performance across the nanoparticle library. Functional screening identifies ANPs as potent immunomodulatory nanomaterials. (A) Schematic illustration of the ex vivo screening assay using murine splenocytes. (B) Screening of 110 PNPs for IFN-γ induction. Top-performing PNPs (highlighted in blue) were selected for further analysis. (C) IL-10 concentrations in splenocyte supernatants following 24 h treatment (ELISA, n = 3). (D) Biocompatibility of PNPs in human PBMCs assessed by flow cytometry (Live/Dead staining) after 48 h incubation. Data are presented as the percentage of live cells (n = 4). (E) Frequency of IFN-γ+ human PBMCs after 24 h treatment (n = 4). (F) Comparative analysis of IFN-γ induction by ANPs relative to other Allium-derived nanoparticles. AcNPs, Allium cepa-NPs (onion); AfNPs, Allium fistulosum-NPs (spring onion); AsNPs, Allium sativum-NPs (garlic); AwNPs, Allium wakegi Araki-NPs (green onion). Representative flow cytometry histograms (left) and corresponding quantification (right, n = 3). (G) Representative flow cytometry plots (left) and corresponding quantification (right) of IFN-γ+ splenocytes treated with PBS, plant extract (extract), 10k pellet (10K-P), or ANPs for 24 h (n = 3). Data are presented as mean ± SEM, where n represents the number of independent biological replicates. Statistical significance was determined by one-way ANOVA followed by Dunnett's or Tukey's post-hoc test, as appropriate. ∗P < 0.05; ∗∗∗∗P < 0.0001; ns, not significant. Screening revealed a highly skewed activity distribution, in which only a small subset of PNPs induced robust IFN-γ secretion comparable to the positive control, lipopolysaccharide (LPS, 100 ng/mL), indicating that potent immunomodulatory activity is a rare intrinsic property rather than a general feature of PNPs (Fig. 2B). The initial screening identified five top-performing candidates: ANPs (Allium tuberosum), AgNPs (Apium graveolens), LNPs (Liriope muscari), SNPs (Solanum melongena), and ZNPs (Zingiber officinale). To systematically prioritize the most promising candidate for downstream antitumor immunotherapy, we applied a sequential filtering process based on counter-regulatory cytokine profiles, cellular toxicity, and cross-species reproducibility. First, to distinguish selectively functional nanomaterials from those inducing counter-regulatory or nonspecific responses, IFN-γ induction was evaluated in parallel with interleukin-10 (IL-10) secretion [38,39]. Candidates that induced relatively high levels of the immunosuppressive cytokine IL-10 alongside IFN-γ (LNPs, SNPs, and ZNPs) were deprioritized (Fig. 2C). Second, cytotoxicity analysis was performed to evaluate the safety profiles of the remaining candidates. Among them, ZNPs exhibited substantially higher toxicity, whereas ANPs maintained exceptional cell viability (Fig. 2D), supporting their classification as biocompatible and functionally active nanomaterials. Finally, reproducibility across biological systems was evaluated using BALB/c mouse splenocytes and human PBMCs. While AgNPs showed diminished IFN-γ induction in these validation models, ANP activity was consistently conserved across genetic backgrounds and species (Fig. 2E and Fig. S6). Through this multi-step selection process, ANPs were successfully identified as the single optimal candidate that uniquely combined potent immunostimulatory activity with minimal cytotoxicity. Subsequent endotoxin analysis and hemocompatibility assays further verified the biological compatibility of ANPs, confirming that the observed immunostimulatory activity originates from the PNPs themselves rather than from lipopolysaccharide contamination or hemolytic toxicity (Figs. S7 and S8). To assess whether this activity was conserved within the Allium genus, PNPs derived from closely related species were examined. Notably, PNPs from A. cepa, A. sativum, A. fistulosum, and A. wakegi failed to elicit comparable IFN-γ responses (Fig. 2F), indicating that functional activity is species-specific and not taxonomically conserved. Furthermore, comparison with juice extracts and larger particulate fractions (10K-P) demonstrated that IFN-γ induction was strictly dependent on the nanoscale vesicular structure of ANPs (Fig. 2G). Taken together, these results establish IFN-γ-based functional screening as a systematic discovery strategy, enabling unbiased identification of rare, highly functional PNPs. Rather than serving solely as an immunological readout, IFN-γ here functions as a systems-level reporter of nanoparticle-driven bioactivity arising from defined nanoscale structures. Through this framework, ANPs were identified as a uniquely potent functional nanomaterial within the plant-derived nanoparticle landscape. As this study represents the first functional screening of a large-scale PNP library, IFN-γ was selected as a primary readout due to its established role in antitumor immune activation. While this strategy enabled efficient identification of pro-inflammatory and M1-associated candidates relevant to cancer immunotherapy, it may underrepresent nanomaterials with immunoregulatory or tolerogenic functions. Future screening frameworks incorporating broader multi-parametric immune profiling may help expand the discovery of functionally diverse PNPs within the library. Nevertheless, given the primary objective of this study to identify potent antitumor agents, we selected ANPs as the most promising candidate from our screening for subsequent in vivo therapeutic evaluation. ANPs drive M1 macrophage-mediated antitumor efficacy in vivo To evaluate whether the immunostimulatory signature identified during functional screening translates into a therapeutic response in vivo, we next examined the antitumor performance of ANPs in syngeneic tumor models. In the B16F10 melanoma model, intravenous administration of ANPs resulted in a pronounced suppression of tumor growth compared with controls (Fig. 3A and B and Fig. S9). Immunophenotypic analysis of the tumor microenvironment revealed that ANPs induced a coordinated remodeling of myeloid and lymphoid compartments. Specifically, ANP treatment significantly increased the proportion of CD11b + macrophages expressing M1-associated markers (CD80 and TNF-α) while concomitantly enhancing the infiltration and effector activation of CD8+ T cells (Fig. 3C, Fig. S10, and Fig. S11A). These effects were not restricted to the tumor site but were also evident in secondary lymphoid organs, including the spleen and tumor-draining lymph nodes (Fig. 3D, Fig. S11B and Fig. S11C), indicating that ANPs act as systemically active immunomodulatory nanomaterials rather than locally confined agents. Immunohistochemical analysis further confirmed broad immune cell recruitment, including CD3ε+ T cells, F4/80+ macrophages, and CD11c + antigen-presenting cells, in ANP-treated tumors (Fig. S12). While proliferative activity, assessed by Ki-67 staining, remained unchanged, cleaved caspase-3 levels were significantly elevated, indicating enhanced apoptotic signaling within the tumor tissue. In vivo antitumor efficacy of ANPs mediated by macrophage reprogramming. (A) Schematic overview of the experimental design of the syngeneic cancer models. Five days after subcutaneous tumor implantation, PBS (control) or ANPs (2 × 1011 particles) were administered intravenously. (B) Growth curves of B16F10-luc melanoma tumors in C57BL/6 mice (left) and weights of excised tumors from each group (right, n = 5). (C, D) Flow cytometry analysis of T lymphocytes and macrophages in (C) tumors and (D) spleens (n = 5). (E) Ex vivo fluorescence imaging of major organs, including liver (Li), heart (He), kidney (Ki), lung (Lu), spleen (Sp), lymph node (Ln), and tumor (Tu), in B16F10 tumor-bearing mice after intravenous injection of PBS or DiD-labeled ANPs at the indicated time points. (F) Representative flow cytometry plots showing DiD distribution among macrophages (MΦ), dendritic cells (DC), T cells (T), B cells (B), and NK cells (NK) in the spleen after 24 h (left) and corresponding quantitative analysis of the DiD + population (right, n = 3). (G) Schematic illustration of the co-culture system used to assess IFN-γ production. (H) IFN-γ levels in the supernatant after co-culture were quantified by ELISA (n = 3). (I) RT-qPCR analysis of Cd86, Tnf, Nos2, Il1b, Cd206, and Il10 mRNA expression in RAW 264.7 cells treated with control (PBS), LPS, or ANPs (n = 3). (J) Schematic overview of the syngeneic cancer model used for therapeutic evaluation (left). Five days after subcutaneous cancer cell implantation, tumor growth curves of B16F10 melanoma in C57BL/6 mice (middle, n = 5) and weights of excised tumors from each group (right, n = 5). Data are presented as mean ± SEM, where n represents the number of independent biological replicates. Statistical significance was determined by Student's t-test (two-tailed) and one-way or two-way ANOVA followed by Dunnett's or Sidak's post-hoc test, as appropriate. ∗P < 0.05; ∗∗P < 0.01; ∗∗∗P < 0.001; ∗∗∗∗P < 0.0001; ns, not significant. Importantly, the antitumor efficacy of ANPs extended beyond the B16F10 model to immune-refractory Lewis lung carcinoma (LLC1) [40] and highly aggressive 4T1 breast cancer models [41,42] (Fig. S13A and B). Notably, oral administration of ANPs achieved therapeutic outcomes comparable to intravenous delivery (Fig. S12C), highlighting their route-independent therapeutic potential. Collectively, these data demonstrate that ANPs induce robust systemic immune activation and functional reprogramming of the tumor microenvironment across multiple refractory tumor models. To ensure that this robust systemic immune activation did not induce systemic toxicity, we next evaluated the systemic safety and immunotoxicity of ANPs in vivo (Fig. S14). Enzyme-linked immunosorbent assay (ELISA) showed that plasma levels of major pro-inflammatory cytokines (TNF-α, IL-1β, and IL-6) were not significantly elevated compared with the PBS-treated group. Furthermore, ANP-treated mice exhibited no significant loss of body weight or hematological abnormalities in the complete blood count (CBC) analysis. Given that ANPs typically accumulate in the liver, liver function and tissue histology were further evaluated. Plasma aspartate aminotransferase (AST) and alanine aminotransferase (ALT) levels remained within normal ranges. Additionally, the hematoxylin and eosin (H&E) staining revealed no obvious histopathological damage or inflammatory infiltration in the liver. Collectively, these results demonstrate that ANPs possess a favorable systemic biocompatibility profile under the tested therapeutic conditions. To elucidate the in vivo biodistribution of ANPs, DiD-labeled ANPs were intravenously administered to tumor-bearing C57BL/6 mice (Fig. 3E). Fluorescence imaging revealed rapid accumulation in the liver, lung, and spleen within 1 h, followed by predominant hepatic signals at 6 h (Fig. S15). Strikingly, at 24 h post-injection, the spleen emerged as the primary site of ANP accumulation, contrasting with prior reports that identify the liver as the dominant sink for systemically administered nanoparticles [[43], [44], [45]]. Interestingly, ANPs exhibited preferential accumulation in the spleen rather than the liver at 24 h after administration. Although the precise mechanism underlying this biodistribution pattern remains unclear, physicochemical properties such as particle size, surface charge, lipid composition, and protein corona formation may contribute to preferential uptake by splenic macrophages. Given the efficient macrophage uptake observed in this study and the abundance of macrophages within the spleen, splenic retention may reflect enhanced recognition by the reticuloendothelial system. Further investigation of the surface and compositional properties of ANPs will be important for clarifying the mechanisms underlying this biodistribution profile. Given the spleen's central role in immune regulation [46,47], this observation prompted an analysis of cellular uptake of ANPs. Flow cytometric analysis of splenocytes demonstrated that ANPs were preferentially internalized by CD11b + F4/80+ macrophages, with substantially higher uptake compared to dendritic cells, T cells, B cells, or NK cells (Fig. 3F and Fig. S16). Confocal microscopy further confirmed efficient cytosolic delivery of ANPs into macrophages in vitro using RAW 264.7 cells (Fig. S17), establishing macrophages as the dominant cellular target of ANPs in vivo. Given that CD8+ T cells are the primary producers of IFN-γ [48], we employed a mechanistically defined co-culture system to determine whether ANPs directly activate CD8+ T cells or instead operate through macrophage-mediated signaling (Fig. 3G). ANP exposure failed to induce IFN-γ production in isolated CD8+ T cells; however, robust IFN-γ induction was observed when T cells were co-cultured with ANP-primed bone marrow-derived macrophages (BMDMs) (Fig. 3H). This result establishes that ANPs function as macrophage-targeting immunomodulatory nanomaterials, indirectly amplifying T-cell effector responses through macrophage reprogramming rather than direct lymphocyte stimulation. Consistent with this mechanism, transcriptional analysis confirmed that ANPs selectively upregulated pro-inflammatory, M1-associated genes (Tnf, Nos2, Cd86) in macrophages without inducing M2-associated markers (Il10, Mrc1) (Fig. 3I). This polarization effect was conserved in human THP-1-derived macrophages (Fig. S18), underscoring cross-species conservation of functional bioactivity. Importantly, systemic depletion of macrophages abrogated the antitumor efficacy of ANPs in vivo (Fig. 3J and Fig. S19), demonstrating that macrophages are indispensable mediators of ANP activity. ANP-RNA triggers M1 macrophage-mediated immunity via the ANP-miR-488/Bcl6 axis To more effectively resolve bioactive cargo encoded within ANPs, ultracentrifugation-isolated ANPs were further purified by density gradient purification (Fig. S20). Subsequently, functional contributions of distinct components were dissected at the nanomaterial level by reconstituting DNA, RNA, or metabolite fractions extracted from ANPs into liposomes, while lipid shells were applied directly. Protein-associated effects were examined independently using heat-inactivated ANPs (Fig. S21). Notably, in RAW 264.7 cells, treatment with the isolated RNA induced the most pronounced morphological changes (Fig. 4A) and a robust upregulation of M1-associated markers (CD80, CD86, MHC-I, iNOS, and TNF-α) along with IFN-γ, as revealed by flow cytometry (Fig. 4B and Fig. S22A). The preserved bioactivity of heat-inactivated ANPs indicates that the observed immunostimulatory function is not protein-mediated but instead primarily driven by non-protein components, excluding proteins as dominant functional mediators. In addition, treatment with isolated RNA cargo induced a clear dose-dependent increase in IFN-γ and CD80 expression (Fig. S22B and Fig. S23). RNA cargo mediates the immunomodulatory activity of ANPs. (A) Representative microscopy images of RAW 264.7 macrophages treated with PBS (Con), LPS, ANPs, commercial Lipofectamine 3000 alone (Lipo), ANP‐derived lipid shell (Shell), heat‐inactivated ANPs, ANP‐derived RNA-loaded liposomes (RNA), or metabolite‐loaded liposomes (Metabolite). Arrows indicate cells exhibiting activation-associated morphological changes. Scale bars, 100 μm (B) Representative flow cytometry histograms of CD86, MHC‐I, and TNF‐α expression in RAW 264.7 macrophages treated with individual ANP cargo components. (C) Schematic overview of ANP‐RNA generation. The lipid fraction and total RNA were independently isolated from ANPs and subsequently reconstituted into RNA‐reloaded nanoparticles using a standard encapsulation procedure. (D) RT-qPCR analysis of RAW 264.7 cells treated with increasing concentrations of ANP‐derived RNA-encapsulated lipid nanoparticles. Dose-dependent upregulation of M1-associated genes (Tnf, Nos2, Cd80), but not the M2 marker Cd206, was observed (n = 3). (E) Schematic illustration of the in vivo experimental design and treatment schedule. B16F10 melanoma cells were subcutaneously implanted into C57BL/6 mice (n = 5–7 per group), followed by intravenous administration of intact ANPs, ANP‐RNA, or ANP‐derived lipid‐only nanoparticles (Lipid) every other day starting at the early tumor stage. (F) Tumor growth curves of each treatment group. (G) Weights of excised tumors from each group at endpoint. Data are presented as mean ± SEM, where n represents the number of independent biological replicates. Statistical significance was determined by one‐way ANOVA followed by Dunnett's post‐hoc test, or two‐way ANOVA with Tukey's post‐hoc test, as appropriate. ∗P < 0.05; ∗∗P < 0.01; ∗∗∗∗P < 0.0001; ns, not significant. For orthogonal validation of this RNA-dependent mechanism, we enzymatically depleted the RNA cargo and assessed macrophage polarization at the transcriptional level via reverse transcription-quantitative polymerase chain reaction (RT-qPCR). Specifically, treatment with the total internal contents (Total internal) effectively upregulated the M1-associated marker (Cd86) and suppressed the M2-associated marker (Cd206). This efficacy was moderately reduced compared to intact ANPs, likely due to partial cargo loss during extraction and differential intracellular delivery efficiency. Furthermore, this polarization activity was significantly abrogated by RNase treatment, whereas it remained completely unaffected by proteinase K digestion (Fig. S24). These contrasting enzymatic responses definitively rule out the contribution of residual proteins, confirming that intact RNA is indispensable for the phenotypic shift. Collectively, these results establish the internal RNA cargo as the primary driver of ANP-mediated macrophage reprogramming. To verify whether RNA cargo alone could recapitulate these effects, we generated ANP-RNA by encapsulating the ANP-derived RNA into empty ANP lipid shells (Fig. 4C). In vitro, ANP-RNA induced dose-dependent upregulation of Tnf and Nos2 without promoting M2-associated markers (Fig. 4D), demonstrating that immunomodulatory activity is intrinsically encoded in the RNA cargo rather than arising from vesicle-associated proteins or soluble plant components. Stability testing further showed that ANP-RNA retained immunostimulatory activity for up to 14 days at 4 °C, 20 °C, and −80 °C (Fig. S25), highlighting RNA as a stable functional component rather than a labile biological payload. Based on these results, we assessed the antitumor efficacy of ANP-RNA in immunotherapy-refractory B16F10 melanoma (Fig. 4E). Intravenous administration of ANP-RNA every other day suppressed tumor growth by 72.8% compared to controls (Fig. 4F, G and Fig. S26). In contrast, the empty lipid-only nanoparticles (Lipid) showed no therapeutic benefit, demonstrating that biological activity arises from RNA-mediated effects rather than nonspecific nanoparticle effects. However, it is important to note that the artificial reconstitution of ANP-RNA achieved an encapsulation efficiency of approximately 30%, as determined by the RiboGreen assay (Fig. S27). This partial encapsulation likely results in the co-delivery of both vesicle-encapsulated and unencapsulated RNA fractions. We speculate that this low encapsulation efficiency limits the actual amount of RNA successfully delivered into the cells, resulting in the lower immunomodulatory activity observed in vitro compared with native intact ANPs. Interestingly, despite this in vitro attenuation, ANP-RNA demonstrated robust in vivo antitumor efficacy comparable to that of native ANPs. This robust therapeutic performance further underscores that the RNA cargo serves as the primary driver of the antitumor response in vivo. To further evaluate the contribution of the intact nanoparticle structure to in vivo RNA delivery, the therapeutic efficacy of unencapsulated free ANP-derived RNA (Free RNA) was assessed in the melanoma model (Fig. S28). Unlike ANP-RNA, Free RNA exhibited substantially reduced antitumor efficacy, failing to significantly suppress tumor growth. Collectively, these findings demonstrate that while RNA is the primary active component driving the antitumor efficacy, the intact vesicular architecture is essential to stabilize the RNA and ensure its successful in vivo delivery. Guided by previous reports identifying small RNAs as key bioactive cargo [17,25,49], we next investigated whether the miRNA fraction of ANPs contributes to M1-like polarization. The miRNA fraction significantly induced an M1-like response, characterized by the upregulation of Tnf, Nos2, and Il1b mRNA (Fig. S29). RNA sequencing further revealed extensive transcriptional remodeling in macrophages exposed to ANPs, with 593 genes upregulated and 615 genes downregulated (Fig. S30). This global transcriptional reprogramming, marked by enrichment of M1-associated and interferon-response pathways (Fig. 5A–C), supports the concept that PNP-derived RNAs act as active regulatory signals rather than passive molecular cargo. ANP-miR-488 mediates macrophage transcriptional reprogramming via Bcl6 suppression. (A–C) Transcriptomic analysis of RAW 264.7 macrophages treated with ANPs or control (Con) by bulk RNA sequencing (RNA-seq, n = 3 biological replicates per group). (A) Heatmap of differentially expressed genes. (B) Gene set enrichment analysis (GSEA) showing top enriched hallmark gene sets. (C) GSEA showing enrichment of IFN-γ-associated gene sets. (D) Schematic overview of the three-step selection criteria used to define candidate miRNA-mRNA regulatory pairs. (E) Functional evaluation of candidate miRNAs in RAW 264.7 macrophages. Cells were transfected with individual miRNA mimics, and then Tnf and Il1b expression was quantified by RT-qPCR (n = 3). (F) Schematic illustration of the predicted ANP-miR-488 binding site in the Bcl6 3′ Untranslated Region (UTR), showing an 8-nt seed-matching region and additional 3′ supplementary base pairing. (G) HEK293T cells expressing Bcl6 WT or Bcl6 Mut luciferase reporters were treated with ANP-miR-488 mimic (ANP-miR-488), ANP-miR-488 mutant mimic (Mutant), or scrambled negative control (Scrambled) miRNA and relative luciferase activity was determined 24 h later using a dual-luciferase reporter assay (n = 3). (H, I) Quantitative analysis of Bcl6 (H) mRNA and (I) protein expression in macrophages treated with Con, ANPs, or ANP-RNA for 24 h (n = 3). (J, K) Quantitative analysis of Bcl6 (J) mRNA and (K) protein expression in macrophages treated with Con, scrambled miRNA, or ANP-miR-488 mimic for 24 h (n = 3). (L) Flow cytometric analysis of CD86 surface expression in macrophages treated with Con, scrambled miRNA, or ANP-miR-488 mimic for 24 h (n = 3). (M − O) RAW 264.7 cells were pre-transfected with a negative control inhibitor (NC) or a specific ANP-miR-488 inhibitor (Inhibitor) for 4 h, followed by ANP treatment for 24 h. The expression levels of (M) Bcl6 and (N) Tnf were quantified by RT-qPCR, and (O) surface CD86 was evaluated by flow cytometry (n = 3). Data are presented as mean ± SEM, where n represents the number of independent biological replicates. Statistical significance was determined by one-way ANOVA followed by Dunnett's or Tukey's post hoc test, as appropriate. ∗P < 0.05; ∗∗P < 0.01; ∗∗∗P < 0.001; ∗∗∗∗P < 0.0001; ns, not significant. To pinpoint specific miRNA regulators, we integrated small RNA sequencing of ANPs and AcNPs with macrophage transcriptomics. Using a rigorous integrative in silico analysis, we identified 1296 miRNAs in ANPs, with 688 sequences uniquely present compared to the non-stimulatory AcNP control. Using these datasets (Tables S2 and S3), we prioritized functional miRNA-mRNA pairs through a three-step filtering strategy (Fig. 5D): (1) selecting ANP-downregulated mRNAs known to negatively regulate M1 activation; (2) excluding miRNAs present in AcNPs to ensure ANP-specific cargo selection; and (3) prioritizing candidates based on miRNA copy number within ANPs as a proxy for functional relevance. Applying this strategy, we identified three potential target mRNAs (Irf4, Bcl6, and Klf4) downregulated in macrophages (Fig. S31A) and are known to restrain inflammatory macrophage polarization [[50], [51], [52], [53]]. Seed complementarity analysis between the 3′ untranslated regions (UTRs) of these transcripts and ANP miRNAs yielded 26 predicted pairs (Fig. S31B). After excluding candidates present in AcNPs and ranking by abundance, we narrowed the list to seven high-priority ANP-specific miRNAs for functional validation. Subsequent RT-qPCR analysis identified ANP-miR-488 as the lead candidate, which significantly upregulated the expression of M1-associated markers (including Tnf and Il1b) in macrophages (Fig. 5E). These results suggest that ANP-miR-488 is a major contributor to the macrophage polarization activity observed in ANPs. Importantly, RT-qPCR analysis of three independent ANP batches demonstrated less than 10% variation in ANP-miR-488 expression, indicating consistent cargo loading across independent preparations. In addition, ANPs exhibited significantly higher ANP-miR-488 levels compared with PNPs isolated from other Allium species that failed to induce IFN-γ, suggesting the specific enrichment of ANP-miR-488 within ANPs (Fig. S32). Structural modeling predicted a high-affinity interaction between ANP-miR-488 and the 3′ UTR of Bcl6, featuring an 8-nt seed match (Fig. 5F). To validate the direct interaction between ANP-miR-488 and its predicted target, Bcl6, we performed a dual-luciferase reporter assay (Fig. 5G). Co-transfection of ANP-miR-488 mimics with a WT Bcl6 3′ UTR reporter significantly reduced luciferase activity (by approximately 54.7%) in HEK293T cells. In contrast, this inhibitory effect was abolished when either the Bcl6 3′ UTR or the ANP-miR-488 mimic was mutated. These findings support that ANP-miR-488 directly regulates Bcl6 through sequence-specific interaction with its 3′ UTR. Consistently, treatment with ANPs or ANP-RNA significantly reduced Bcl6 mRNA and protein levels (Fig. 5H and I). To confirm the mechanism, we employed a gain-of-function strategy using an ANP-miR-488 mimic in RAW 264.7 cells. Treatment with the ANP-miR-488 mimic showed Bcl6 suppression at the transcript level and induction of CD86 at the protein level, demonstrating that this single miRNA treatment recapitulated the ANP-induced M1 phenotype independently (Fig. 5J–L and Fig. S33). Conversely, to evaluate the specific requirement of ANP-miR-488 in ANP-mediated macrophage reprogramming, a targeted loss-of-function study was performed. RAW 264.7 cells were pre-transfected with a specific ANP-miR-488 inhibitor or NC inhibitor prior to ANP treatment (Fig. 5M − O). Subsequent analysis combining RT-qPCR and flow cytometry revealed that, compared to the ANP-treated NC group, intracellular blockade of ANP-miR-488 significantly restored Bcl6 expression and abrogated the upregulation of M1-associated markers (Tnf and CD86). These findings support ANP-miR-488 as an important contributor to the immunomodulatory activity of ANPs. While the molecular complexity of ANPs suggests that additional regulatory axes may operate in parallel, our findings highlight the ANP-miR-488/Bcl6 axis as a primary mechanism by which plant-derived miRNAs directly reprogram mammalian immune cells. Clinical relevance and synergistic αPD-1 efficacy of ANPs To explore the clinical relevance of macrophage state regulation associated with the ANP-miR-488/BCL6 axis, human single-cell and bulk transcriptomes were integrated. A macrophage-enriched signature was defined using lung adenocarcinoma (LUAD) single-cell data (GSE131907) (Fig. 6A). This signature was then applied to LUAD and lung squamous cell carcinoma cohorts to stratify macrophage-enriched patient samples, which were further classified into BCL6high and BCL6low groups (Fig. S34). The BCL6low group exhibited a pronounced immune-inflamed phenotype, characterized by elevated expression of M1-associated inflammatory markers (TNF, CD80, CD86, CXCL9, and CXCL10), whereas immunosuppressive markers (ARG1, CD163) were reduced or unchanged (Fig. 6B). Moreover, CD8+ T cell infiltration (CD8A, CD8B, CXCR3) and effector activation signatures (IFNG, GZMB, PRF1) were markedly upregulated in the BCL6low group (Fig. 6C), recapitulating the macrophage-T cell crosstalk observed in our experimental models. Clinical relevance of ANPs associated with BCL6-dependent immune remodeling and enhanced response to αPD-1 therapy. (A) Identification of a macrophage-enriched gene signature using the GSE131907 lung adenocarcinoma (LUAD) single-cell RNA-sequencing dataset. (B, C) Transcriptomic analysis of macrophage-enriched TCGA lung cancer cohorts (LUAD and LUSC) stratified into BCL6high and BCL6low groups. (B) Violin plots showing the expression of M1-associated inflammatory markers (TNF, CD80, CD86, CXCL9, CXCL10) and immunosuppressive markers (ARG1, CD163). (C) Violin plots showing genes associated with CD8+ T cell infiltration (CD8A, CD8B, CXCR3) and cytotoxic effector molecules (IFNG, GZMB, PRF1). (D) Association between BCL6 expression and clinical response to αPD-1 therapy in the GSE299686 cohort. Macrophage-enriched patients were stratified by BCL6 expression level, and the proportions of responders versus non-responders were compared. (E) Schematic overview of the treatment schedule and experimental design for in vivo evaluation of αPD-1 and ANPs combination therapy in an LLC1 subcutaneous tumor model. PBS or ANPs were administered intravenously every other day, and IgG or αPD-1 was administered intraperitoneally every three days starting at the early tumor stage. (F) Tumor weights measured at the study endpoint (day 15) in the indicated groups (n = 5 biological replicates per group). Data are presented as mean ± SEM. Statistical significance was determined by Student's t-test (two-tailed) or one-way ANOVA followed by Dunnett's post-hoc test, as appropriate. ∗P < 0.05; ∗∗P < 0.01; ∗∗∗P < 0.001; ∗∗∗∗P < 0.0001; ns, not significant. To further interrogate the clinical implications of this macrophage state, we analyzed an independent αPD-1-treated head and neck squamous cell carcinoma cohort (GSE299686). Patients within the BCL6low group exhibited a substantially higher objective response rate (55.0%, 33/60) compared with the BCL6high group (10.3%, 3/29; P = 5.87 × 10−5; Fig. 6D). A concordant trend was observed in an additional lung cancer immunotherapy cohort (GSE93157; Fig. S35), suggesting the broader clinical relevance of BCL6-associated macrophage states in modulating immune checkpoint responsiveness. While these human clinical cohort analyses provide valuable correlative evidence, applying these findings to clinical applications requires caution, as they reflect clinical associations rather than direct causal relationships. Therefore, future translational studies using human patient-derived macrophages, ex vivo tumor explants, and humanized mouse models are necessary to fully establish their clinical applicability. To experimentally validate this therapeutic rationale and evaluate its functional relevance in vivo, ANPs were tested in a syngeneic LLC1 tumor model in combination with αPD-1 therapy. Co-administration of ANPs with immune checkpoint blockade resulted in significantly enhanced tumor suppression compared with either treatment alone (Fig. 6E and F), suggesting that ANP-mediated macrophage modulation may improve responsiveness to immunotherapy in vivo. Collectively, these findings establish ANP-mediated miRNA delivery as a potent strategy to reprogram macrophage transcriptional states that are clinically associated with improved immunotherapy outcomes. By linking PNP-mediated immune modulation to patient-relevant immune phenotypes, this work provides a translational framework for leveraging plant-derived bio-nanomaterials as functional adjuvants to potentiate immune checkpoint therapies. Conclusion In this study, we establish a systematic discovery framework for bioactive plant-derived nanoparticles (PNPs) that moves beyond the conventional one-plant-one-function paradigm. By constructing and physicochemically mapping a taxonomically diverse library of 110 PNPs, we demonstrate that biological functionality can be quantitatively assessed through functional screening, enabling unbiased identification of functional bio-nanomaterials from natural diversity. Using this approach, ANPs were identified as a previously unrecognized functional PNP with pronounced immunomodulatory activity, emerging from systematic screening rather than predefined molecular design. Mechanistic analyses revealed that ANP bioactivity is primarily associated with RNA cargo carried by the PNPs. In particular, ANP-miR-488 suppresses the transcriptional regulator Bcl6, establishing a clear structure-function relationship that links nanoscale carrier architecture to immune modulation. These findings suggest that RNA functions as a key active component contributing to the immunomodulatory activity of ANPs. Collectively, this work positions PNPs as a promising class of bioactive nanomaterials and introduces a generalizable framework for the discovery of functional nanomaterials from natural sources. By integrating botanical biodiversity with functional screening approaches, this study provides a foundation for the development of RNA-associated immunomodulatory nanotherapeutics, with potential applications in cancer immunotherapy. Experimental section Cell culture B16F10, LLC1, 4T1, and RAW 264.7 cells were obtained from the American Type Culture Collection (ATCC) and maintained under standardized culture conditions at 37 °C in a humidified incubator with 5% CO2. B16F10-luc-g5 cells (Caliper) were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin (PS). All cell lines were used within 25 passages from the original stock. Human PBMCs and mouse splenocytes were cultured in RPMI-1640 supplemented with 10% FBS and 1% PS. Construction and isolation of the plant-derived nanoparticle (PNP) library To construct a taxonomically diverse and standardized materials library, 59 edible crops and 51 medicinal herbs were processed. Fresh plant tissues (100∼300 g) were washed and subjected to uniform mechanical extraction to obtain crude juice, which was diluted 1:1 with PBS and filtered through a 100 μm mesh. PNPs were isolated by stepwise differential centrifugation under temperature-controlled conditions. The filtrate was centrifuged at 500 ×g for 5 min and 2500 ×g for 15 min to remove large debris, followed by two centrifugation steps at 10,000 ×g for 30 min to eliminate micron-scale particles. The supernatant was clarified using a 40 μm cell strainer (#352340; Corning) and ultracentrifuged at 100,000 ×g for 60 min (SW32 Ti rotor, Beckman Coulter). Crude pellets were resuspended in PBS, filtered through a 0.22 μm syringe filter, and subjected to a second ultracentrifugation at 100,000 ×g for 60 min. Final PNP pellets were resuspended in PBS and stored until use. All centrifugation steps were performed at 4 °C to preserve nanoscale integrity. Physicochemical characterization The hydrodynamic size distribution and particle concentration of PNPs were quantified using a NanoSight LM10 (Malvern Panalytical, Malvern, UK) equipped with a 405 nm laser. Prior to analysis, samples were diluted in 0.22 μm-filtered PBS to achieve an optimal particle concentration for tracking. Measurements were performed by acquiring 30-60 s videos (30 fps, camera level 10), followed by analysis using NTA software (v3.2) with a detection threshold of 3. To quantify the physical homogeneity of the isolated PNPs, the Span value was calculated as (D90 - D10)/D50, where D10, D50, and D90 denote particle diameters at 10%, 50%, and 90% of the cumulative size distribution, respectively [[54], [55], [56]]. The zeta potential of PNPs was measured using a Zetasizer Nano ZS (Malvern Panalytical, Malvern, UK) at 25 °C. Morphological integrity was assessed by transmission electron microscopy (TEM). Briefly, PNP suspensions were fixed with 2% paraformaldehyde (PFA) for 5 min, deposited onto UV-treated pure carbon-coated copper grids, and allowed to adsorb for 2 min. Grids were then negatively stained with 2% (w/v) uranyl acetate for 1 min, air-dried at room temperature (RT), and imaged using a Hitachi HT7700 transmission electron microscope (Hitachi, Tokyo, Japan) operated at 120 kV. For Cryo-EM Analysis, carbon grids were glow-discharged for 60 s at 15 mA (negative polarity) using a GloQube system. 3 μL of the ANP suspension was then applied onto the prepared grids. Sample vitrification was carried out using a FEI Vitrobot Mark IV (Thermo Fisher Scientific) at 4 °C and 100% humidity. Following a wait time of 10 s, the grids were blotted for 10 s with a blot force of 0, and immediately plunge-frozen into liquid ethane. The vitrified samples were subsequently imaged using a Titan Krios G4 transmission electron microscope (Thermo Fisher Scientific) operated at 300 kV. Ex vivo primary cell-based assays Mouse immune cells were obtained from C57BL/6 or BALB/c mice maintained under specific pathogen-free (SPF) conditions. For primary screening assays, single-cell suspensions of murine splenocytes were prepared from aseptically harvested spleens by mechanical dissociation and passage through 40 μm cell strainers (#352340; Corning), followed by red blood cell lysis using ACK lysis buffer. Human peripheral blood mononuclear cells (PBMCs) were isolated from whole blood using a density gradient medium (#17144002; Cytiva). For primary splenocyte-based screening, 2 × 105 splenocytes per well were seeded in a 96-well tissue culture plate in 200 μL of complete RPMI-1640 medium supplemented with 10% FBS and incubated with PNPs at a dose of 2.5 × 104 particles per cell for 48 h at 37 °C in a humidified atmosphere containing 5% CO2. Culture supernatants were collected and assayed for cytokines IFN-γ (#DY485; R&D systems) or IL-10 (#DY417; R&D systems) by ELISA. For intracellular cytokines analysis, 2 × 105 splenocytes or human PBMCs were seeded per well in 96-well plates (200 μL per well) and incubated with PNPs for 24 h at 37 °C and 5% CO2. Cells were subsequently stimulated with a cell stimulation cocktail (#00-4970-93; BD Biosciences) for 5 h at 37 °C, followed by staining with the Fixable Aqua Dead Cell Stain Kit (#L34957; Invitrogen) at 4 °C for 30 min. Cells were then fixed and permeabilized with IC Fixation Buffer (#00-8222-49; eBioscience) at 4 °C for 30 min and incubated with the indicated antibodies in permeabilization buffer (#00-8333-56; eBioscience) at 4 °C for 30 min. Flow cytometric data were acquired using a CytoFLEX and analyzed with CytExpert (v.2.6) or FlowJo (v.10.10). Flow cytometry-based immune profiling Tumors were excised, mechanically minced into fragments <4 mm in length, and enzymatically dissociated using a mouse tumor dissociation kit (Miltenyi Biotec) for 45 min at 37 °C. Spleens were processed into single-cell suspensions, and red blood cells were removed using Red Blood Cell Lysing Buffer (Sigma). Single-cell suspensions from tumors, tumor-draining lymph nodes (DLNs), and spleens were stained on ice for 30 min with a fixable viability dye to exclude dead cells. Cells were first stained for surface markers defining immune cell populations. For intracellular and transcription factor staining, the Foxp3/Transcription Factor Staining Buffer Set (#00-5523-00; eBioscience) was used according to the manufacturer's instructions. Prior to antibody staining, samples were incubated with Fc block (BD Pharmingen) for 10 min on ice to minimize nonspecific binding. Unless otherwise specified, antibodies were used at a 1:200 dilution. The following anti-mouse antibodies were obtained from BioLegend: CD19 (Clone QA17A27, PE, #159803), CD206 (Clone C068C2, APC, #141708), CD25 (Clone 3C7, PE, #101904), CD3 (Clone 145-2C11, PerCP-Cy5.5, #100328), CD4 (Clone GK1.5, APC, #100412), FoxP3 (Clone QA20A67, PE, #118904), IFN-γ (Clone XMG1.2, FITC, #505806; APC, #505810), and PD-1 (Clone RMP1-30, PE, #109104). Additional anti-mouse antibodies were purchased from Invitrogen, including CD11b (Clone M1/70, FITC, #11-0112-82), CD11c (Clone N418, PE, #12-0114-82), CD45 (Clone 30-F11, APC-eFluor780, #47-0451-82), CD80 (Clone 16-10A1, PE, #12-0801-82), CD8a (Clone 53-6.7, eFluor450, #48-0081-82), F4/80 (BM8, eFluor450, #48-4801-82), IL-10 (Clone JES5-16E3, PE, #12-7101-82), Tim3 (Clone 8B.2C12, APC, #17-5871-82), and TNF-α (Clone MP6-XT22, eFluor450, #48-7321-82). Anti-human IFN-γ antibody (Clone 4S.B3, APC, #17-7319-82) was obtained from Invitrogen. Flow cytometric acquisition was performed using a CytoFLEX system (Beckman Coulter). Single-color-stained samples were used for compensation, and isotype controls were included to assess nonspecific binding. Data were analyzed using CytExpert or FlowJo software as indicated. Total RNA isolation and RT-qPCR Total RNA was isolated from RAW 264.7 cells, THP-1-derived macrophages, and ANPs using the Direct-zol™ RNA MiniPrep Kit (#R2052; Zymo Research) according to the manufacturer's instructions. Complementary DNA (cDNA) was synthesized from total RNA using a Reverse Transcription Master Mix (EBT-1542; Elpisbiotech). RT-qPCR was performed using RbTaq™ SYBR Green qPCR PreMIX (RT530M; Enzynomics). All RT-qPCR reactions were carried out in triplicate in a total reaction volume of 20 μL using the primer sets listed in Table S4. Amplification and signal detection were conducted on a StepOnePlus™ Real-Time PCR System (Applied Biosystems). Relative mRNA expression levels were calculated using the comparative threshold cycle (Ct) method and normalized to Gapdh expression within the same samples. For ANP-miR-488 quantification in PNPs, Taq-man RT-qPCR analysis was performed. Briefly, total RNA containing small RNAs was reverse-transcribed using the TaqMan™ MicroRNA Reverse Transcription Kit (Thermo Fisher Scientific, Cat. No. 4366596) and miRNA-specific primers from the Custom TaqMan™ Small RNA Assay (Thermo Fisher Scientific, Cat. No. 4398987), according to the manufacturer's instructions. RT-qPCR was performed using the corresponding TaqMan small RNA assay and TaqMan™ Fast Advanced Master Mix (Thermo Fisher Scientific, Cat. No. 4444557). Ct values were obtained using a real-time PCR system. Immunohistochemistry Paraffin-embedded tumor tissues were sectioned at a thickness of 5 μm and fixed in 4% paraformaldehyde (PFA). Sections were deparaffinized in xylene, rehydrated through a graded ethanol series, and washed with phosphate-buffered saline (PBS). Antigen retrieval was performed by boiling the sections in antigen retrieval buffer (ab93678; Abcam) for 10 min, followed by cooling at room temperature for 2 h. Endogenous peroxidase activity was quenched using 3% hydrogen peroxide, and nonspecific binding was blocked with 5% normal goat serum (Vector Laboratories, S-1000) in PBS. Sections were incubated overnight at 4 °C with primary antibodies diluted in 5% normal goat serum (100 μL per slide). After washing, appropriate secondary antibodies were applied and incubated for 1 h at room temperature. Signal detection was carried out using 3,3′-diaminobenzidine (DAB; Vector Laboratories, SK-4100), followed by counterstaining with hematoxylin for 3 min. Stained sections were examined and imaged using a bright-field microscope. Animal studies Six-to seven-week-old BALB/c (BALB/cAnNCrl) and C57BL/6J (C57B6NCrl) mice were purchased from Orient Bio (Seongnam, Korea) and maintained under specific pathogen-free (SPF) conditions. For syngeneic tumor models, B16F10-luc melanoma cells or Lewis lung carcinoma (LLC1) cells (2 × 105 cells in 100 μL PBS) were subcutaneously injected into the right flank of male C57BL/6J mice. For the breast cancer model, 4T1 cells (2 × 105 cells in 100 μL PBS) were subcutaneously injected into the right flank of female BALB/c mice. Tumor growth was monitored every two days using a digital caliper, and tumor volume was calculated using the formula: volume (cm3) = (width2 × length) × 0.5. Once the average tumor volume reached <30 mm3, mice were randomized into treatment groups with comparable baseline tumor sizes. For systemic macrophage depletion, Clodrosome or control Encapsome (200 μL, intraperitoneal; Encapsula NanoSciences) was administered one day prior to tumor implantation and subsequently every three days throughout the study period [39]. Tumor growth was monitored until tumors reached the maximal size permitted by IACUC guidelines (1500 mm3), at which point mice were euthanized. At the study endpoint (day 15), mice were sacrificed, and tumors, spleens, and tumor-draining lymph nodes were harvested for downstream analyses. In vivo safety evaluation of ANPs To evaluate systemic safety following intravenous administration of ANPs, mice were intravenously injected with ANPs according to the treatment schedule described in the tumor model experiments. Body weight and general behavior were monitored throughout the study period. For cytokine analysis, blood samples were collected, and plasma levels of TNF-α (MTA00B), IL-1β (MLB00C), and IL-6 (M6000B) were quantified using ELISA kits according to the manufacturer's instructions (R&D Systems). Additionally, plasma levels of AST and ALT were measured to assess liver function using specific ELISA kits (ab325368 and ab282882, Abcam) following the manufacturer's protocols. CBC analysis (Vet ABC, SCIL) was performed using whole blood samples collected in EDTA-containing tubes. For H&E staining, liver tissues were collected from mice treated with PBS or ANPs administered every 2 days for a total of five injections. The tissues were fixed in 4% paraformaldehyde (PFA), paraffin-embedded, and sectioned at a thickness of 5 μm. After deparaffinization in xylene and rehydration through a graded ethanol series, the sections were stained with hematoxylin (CS700, Dako) for 5 min at RT and washed with distilled water. The sections were then stained with eosin (CS701, Dako) for 10 min at RT, followed by washing with distilled water. After dehydration and mounting, histological images were acquired using a bright-field microscope. In vivo biodistribution of ANPs For fluorescence labeling, ANPs were incubated with Vybrant™ DiD Cell-Labeling Solution at room temperature for 1 h. Unincorporated DiD was removed using a Zeba spin desalting column according to the manufacturer's instructions. Labeled ANPs (DiD-ANPs) were administered intravenously to male C57BL/6 mice (6-8 weeks old) at a dose of 2 × 1011 particles per injection, with PBS used as a control. Whole-body in vivo fluorescence imaging was performed at predefined time points (0, 1, 3, 6, 12, 24, and 48 h post-injection) to monitor the temporal biodistribution of DiD signals. At the indicated endpoints, mice were euthanized, and major organs, including heart, liver, spleen, lung, kidney, lymph nodes, and tumors, were harvested for ex vivo fluorescence imaging. For cellular-level biodistribution analysis, spleens were collected 24 h after injection and processed into single-cell suspensions for flow cytometry. The relative uptake of DiD-ANPs by immune cell populations was quantified based on DiD fluorescence intensity in dendritic cells (CD45+CD11c+), macrophages (CD45+CD11b+), T cells (CD45+CD3+CD8+), B cells (CD45+CD19+), and natural killer (NK) cells (CD45+NK1.1+). Cells were stained with CD11b (Clone M1/70, FITC, Invitrogen, #11-0112-82), CD11c (Clone N418, PE, Invitrogen, #12-0114-82), CD19 (Clone QA17A27, PE, BioLegend, #159803), CD3 (Clone 145-2C11, PerCP-Cy5.5, BioLegend, #100328), CD4 (Clone GK1.5, APC, BioLegend, #100412), CD45 (Clone 30-F11, APC-eFluor780, Invitrogen, #47-0451-82), NK1.1 (Clone PK136, eFluor450, Invitrogen, #48-5941-82). Flow cytometric analysis enabled quantitative comparison of ANP accumulation across immune cell subsets in the spleen. Co-culture assay Bone marrow cells were isolated from the femurs and tibias of C57BL/6 mice, passed through 70-μm cell strainers, and subjected to red blood cell lysis. Cells were cultured in complete RPMI medium supplemented with 10% FBS and 1% antibiotic-antimycotic, and differentiated into bone marrow-derived macrophages (BMDMs) by supplementation with macrophage colony-stimulating factor (M-CSF, 20 ng mL−1) for 7 days at 37 °C in a humidified atmosphere containing 5% CO2, with a half-medium change on days 3-4. Differentiated adherent BMDMs were gently detached, seeded at 2 × 104 cells per well in 96-well plates, and treated with ANPs at the indicated concentrations for 24 h. After treatment, supernatants were removed and cells were washed twice with prewarmed medium to eliminate residual ANPs. Naïve CD8+ T cells were isolated from spleens and peripheral lymph nodes (inguinal and axillary) by mechanical dissociation to single-cell suspensions, followed by red blood cell lysis and magnetic-activated cell sorting (MACS)-based negative selection (Miltenyi Biotec). Purified naïve CD8+ T cells were added to ANP-primed BMDMs at a density of 1 × 105 cells per well (T cell:BMDM ratio of 5:1) in complete medium. Where indicated, soluble anti-CD3 antibody (0.01 μg mL−1; BD Pharmingen) and interleukin-2 (IL-2, 10 U mL−1; R&D Systems) were added at the time of T-cell seeding. Co-cultures were maintained for 72 h at 37 °C under 5% CO2 prior to downstream analysis. Supernatants were collected and analyzed for IFN-γ (#DY485; R&D Systems) using ELISA. Lipid extraction and ANP-RNA encapsulation ANP-derived lipids were extracted from purified ANPs (1 × 1011 particles) using a modified methyl tert-butyl ether (MTBE)-based extraction protocol. Briefly, phase separation was induced by aqueous addition, followed by centrifugation at 10,000 × g for 10 min. The organic phase was collected and subjected to a second extraction to maximize lipid recovery. Combined organic fractions were evaporated under reduced pressure at 50 °C to obtain dried lipid films. For RNA encapsulation, lipid amounts corresponding to 1 × 1010 ANPs were rehydrated with ANP-derived RNA in a final volume of 200 μL of buffer and subjected to bath sonication (40 kHz) for 30 min at room temperature to generate RNA-reconstituted nanoparticles (ANP-RNA). Encapsulation efficiency was quantified using a RiboGreen assay and was consistently ∼30%. RNA sequencing and data processing Total RNA was extracted from RAW 264.7 macrophages using the RNeasy Mini Kit (Qiagen) with on-column DNase digestion according to the manufacturer's instructions. RNA integrity and quality were assessed using an Agilent Bioanalyzer (RNA Small RNA chip). Sequencing libraries were prepared with the TruSeq Stranded Total RNA Library Prep Gold Kit (Illumina) and sequenced on an Illumina NovaSeq X platform to generate paired-end reads (2 × 101 bp). Raw sequencing reads were processed by trimming adapter sequences and low-quality bases using Trimmomatic (v0.39). Filtered reads were aligned to the mouse reference genome (mm10) using STAR (v2.7.9a). Gene-level expression was quantified using TPM Calculator (v0.0.3), and differential expression analysis was performed on raw count data using the DESeq2 R package (v1.46.0). Differentially expressed genes (DEGs) were defined by a Benjamini-Hochberg adjusted P value < 0.05 and an absolute log2 fold change >1. Functional enrichment analyses were conducted using Gene Set Enrichment Analysis (GSEA) with Hallmark gene sets, and Gene Ontology (GO) network analysis was applied to identify immune-related regulatory modules. Heatmaps were generated using the ComplexHeatmap R package (v2.22.0) based on Z-score-normalized expression values. Small RNA sequencing and identification of miRNAs Total RNA was extracted from purified ANPs using TRIzol reagent (Invitrogen) according to the manufacturer's protocol. Small RNA libraries were prepared using the TruSeq Small RNA Library Prep Kit (Illumina) and sequenced on an Illumina platform to generate 151 bp single-end reads. Raw sequencing reads were processed using Cutadapt for 3′ adapter trimming and quality filtering. The resulting high-quality reads were aligned to mature miRNA reference sequences from miRBase (v22.1) using Bowtie (v1.3.1). Reads mapped to known miRNAs were quantified to determine relative abundance. Unmapped reads with a minimum read count of ≥50 were further analyzed by clustering based on seed sequence identity (nucleotides 2-8). These clusters were assigned provisional identifiers with arbitrary numbering (e.g., ANP-miR-1) to distinguish them from annotated miRNAs in miRBase [57]. Single-cell RNA sequencing analysis and signature definition Publicly available single-cell RNA sequencing (scRNA-seq) data from lung adenocarcinoma (LUAD) patients were obtained from the Gene Expression Omnibus (GEO; GSE131907). Data preprocessing, normalization, dimensionality reduction, and clustering were performed using the Seurat R package (v5.0.0). Macrophage clusters were identified based on the expression of canonical macrophage markers, including CD68, CD163, and LYZ. To derive a robust macrophage-enriched gene signature, differentially expressed genes (DEGs) were identified using the FindMarkers function, with selection criteria designed to capture genes exhibiting high, selective expression in macrophage clusters while maintaining low basal expression in other immune cell types (NK cells, dendritic cells, T cells, and B cells). The top 30 genes ranked by log2 fold change were selected to define the final macrophage signature, which was subsequently used for stratification of bulk transcriptomic datasets. Prediction of ANP miRNA targeting in RAW 264.7 macrophages The 3′ UTR sequences of candidate target mRNAs (Irf4, Bcl6, and Klf4) were retrieved from the Ensembl database using the biomaRt R package (v2.62.1). Putative interactions between ANP-derived miRNAs and these 3′ UTR sequences were predicted using the miRanda algorithm (v3.3a). High-confidence miRNA-mRNA interactions were defined using stringent criteria, including an alignment score ≥150 and a minimum free energy (MFE) ≤ −25 kcal mol−1. To ensure specificity to ANPs, miRNAs detected in the AcNP dataset (read count >0) were excluded from further consideration. The remaining ANP-specific miRNAs were ranked according to normalized abundance, and the seven most abundant candidates were selected for subsequent experimental validation. Dual-luciferase reporter assay Oligonucleotide pairs containing the predicted ANP-miR-488 binding site within the Bcl6 3′ UTR were synthesized (Macrogen Inc.) and cloned into the pmirGLO Dual-Luciferase miRNA Target Expression Vector (Promega). The oligonucleotide sequences used were as follows: Bcl6 WT, 5′-GTTTAAACGTGGGGGTGGGGTTTGCGGGACCGGGAGCCAAGGCAGCTCCCCTTCCCACACTGCCATAAAACATTAATCTAGA-3′; and Bcl6 Mut, 5′-GTTTAAACGTGGGGGTGGGGTTTG CGGGACCGGCTATGTCGGCAGCTCCCCTTCCCACACTGCCATAAAACATTAATCTAGA-3′. HEK293T cells harboring the WT or Mut Bcl6 3′ UTR luciferase reporter construct were seeded into 96-well plates and transfected with ANP-miR-488 mimic, ANP-miR-488 mutant mimic, or scrambled negative control miRNA using Lipofectamine™ RNAiMAX Transfection Reagent (13778150, Invitrogen) according to the manufacturer's protocol. After 24 h, Firefly and Renilla luciferase activities were measured using the Dual-Luciferase® Reporter Assay System (E1910, Promega). Firefly luciferase activity was normalized to Renilla luciferase activity, and luminescence was quantified using a SpectraMax® L Microplate Reader (Molecular Devices). Transfection of miRNA mimics and inhibitors RAW 264.7 macrophages were seeded in 12-well plates at a density of 2.0 × 105 cells per well in DMEM supplemented with 10% fetal bovine serum (FBS) and allowed to adhere overnight. For gain-of-function studies, cells were transfected with individual miRNA mimics or a scrambled control miRNA (Bioneer, Korea) at a final concentration of 100 nM using Lipofectamine RNAiMAX (13778150, Invitrogen), following the manufacturer's protocol. Transfection complexes were prepared in Opti-MEM (Thermo Fisher Scientific), added to the cells, and incubated at 37 °C in a humidified atmosphere containing 5% CO2 for 24 h. For loss-of-function studies, cells were transfected with either an ANP-miR-488 inhibitor or a negative control (NC) inhibitor utilizing Lipofectamine RNAiMAX according to the manufacturer's protocol. After 4 h of initial incubation, the transfection medium was replaced, and the cells were subsequently treated with ANPs for an additional 24 h. Following the respective treatments, cells were harvested for downstream analyses, including RT-qPCR, immunoblotting, and flow cytometry. TCGA and clinical cohort analysis RSEM-normalized transcriptomic datasets for lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC) were obtained from the Broad Institute GDAC Firehose repository (firebrowse.org). To assess macrophage-associated transcriptional features, single-sample gene set enrichment analysis (ssGSEA) was performed using the macrophage gene signature defined from single-cell RNA sequencing analysis. ssGSEA scores were calculated using the GSVA R package (v2.0.7). Tumors with ssGSEA scores above the cohort median were classified as macrophage-enriched. Macrophage-enriched tumors were subsequently stratified into BCL6high and BCL6low groups based on the median BCL6 expression level. Differential expression of immune-related genes between groups was evaluated using an unpaired two-tailed Wilcoxon rank-sum test. For analysis of immunotherapy response, transcriptomic profiles and corresponding clinical metadata were obtained from the GSE299686 cohort. Macrophage-enriched patients were identified using the same ssGSEA-based stratification approach, and clinical responses were categorized as responders (partial or complete response) or non-responders (stable or progressive disease). Differences in response rates between BCL6high and BCL6low groups were assessed using Fisher's exact test. Western blotting For immunoblot analysis, total cellular proteins were extracted from RAW 264.7 macrophages, and equal amounts of protein (20 μg per sample) were separated by SDS-PAGE (Invitrogen) and transferred onto nitrocellulose membranes (Invitrogen). Membranes were blocked for 1 h at room temperature in Tris-buffered saline containing 0.1% Tween 20 (TBS-T) supplemented with 5% skim milk, followed by incubation with primary antibodies overnight at 4 °C. Primary antibodies used were anti-Bcl6 (1:1000; Clone D65C10; #5650S; Cell Signaling Technology) and horseradish peroxidase (HRP)-conjugated anti-β-actin (1:20,000; Clone C4; #sc-47778; Santa Cruz Biotechnology). Membranes were washed three times with TBS-T for 10 min each. For Bcl6 detection, membranes were subsequently incubated with an HRP-conjugated secondary antibody (anti-rabbit IgG; #7074S; Cell Signaling Technology) for 1 h at room temperature. Protein bands were visualized using enhanced chemiluminescence detection reagents (#34580; Thermo Fisher Scientific) and imaged with a FUSION Solo S imaging system (Vilber). Data analysis Appropriate statistical tests were selected for each experiment, and data are presented as mean ± standard error of the mean (SEM) or standard deviation (SD), as indicated. Comparisons between two groups of normally distributed data were performed using an unpaired two-tailed Student's t-test, whereas non-normally distributed data were analyzed using the Mann-Whitney U test. Comparisons among multiple groups were conducted using one-way or two-way analysis of variance (ANOVA), followed by Dunnett's, Sidak's, or Tukey's multiple-comparison post hoc tests, as appropriate. A P value < 0.05 was considered statistically significant. Statistical significance is denoted as P < 0.05 (∗), < 0.01 (∗∗), < 0.001 (∗∗∗), and <0.0001 (∗∗∗∗); ns indicates not significant. Exact P values and the statistical tests used are specified in the corresponding figure legends. All statistical analyses were performed using GraphPad Prism software (v9.5.1; GraphPad, USA). All experiments were conducted with at least three independent biological replicates, and measurements were performed in triplicate or more unless otherwise stated. Data availability statement The data that support the findings of this study are available from the corresponding author upon reasonable request. Ethics approval and consent to participate All animal procedures were conducted in accordance with protocols approved by the Kyungpook National University Institutional Animal Care and Use Committee (IACUC; approval number: KNU-2025-0783). PBMCs from healthy donors were obtained from Kyungpook National University Medical Center (KNUMC). Blood collection was approved by the Institutional Review Board of KNUMC (Approval No. 2022-02-011), and written informed consent was obtained from all donors. All authors are in compliance with all relevant ethical regulation. CRediT authorship contribution statement Dokyung Jung: Data curation, Formal analysis, Investigation, Methodology, Software, Validation, Visualization, Writing – original draft, Writing – review & editing. Ju-Hyun Bae: Data curation, Formal analysis, Investigation, Methodology, Software, Validation, Visualization, Writing – original draft, Writing – review & editing. Il-Young Jung: Data curation, Formal analysis, Investigation, Methodology, Validation, Writing – original draft, Writing – review & editing. Na-Eun Kim: Data curation, Formal analysis, Investigation, Methodology, Validation, Writing – original draft, Writing – review & editing. Jung Min Sung: Data curation, Methodology, Software, Visualization, Writing – review & editing. Kyung-Won Doh: Data curation, Investigation, Methodology. Minju Kim: Data curation, Investigation, Methodology, Visualization. Jiwoong Choi: Methodology, Supervision, Writing – review & editing. Eunji Ha: Methodology, Software, Supervision, Writing – review & editing. Hee Jin Cho: Data curation, Methodology, Software, Supervision, Writing – review & editing. Byung Sun Min: Resources, Supervision, Writing – review & editing. Jong Hyun Kim: Methodology, Resources, Supervision, Writing – review & editing. Sin-Hyeog Im: Methodology, Resources, Supervision, Writing – review & editing. Moon-Chang Baek: Conceptualization, Funding acquisition, Methodology, Project administration, Supervision, Writing – original draft, Writing – review & editing. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Acknowledgements We thank Professor Byungheon Lee (Kyungpook National University) for kindly providing the dual-luciferase reporter vector. This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Ministry of Science and ICT (MSIT) (RS-2023-NR077266, RS-2021-NR060094, RS-2024-00353479). Footnotes Peer review under the responsibility of editorial board of Bioactive Materials. Supplementary data to this article can be found online at https://doi.org/10.1016/j.bioactmat.2026.06.034. Contributor Information Dokyung Jung, Email: christion3@knu.ac.kr. Ju-Hyun Bae, Email: jhbae89@knu.ac.kr. Il-Young Jung, Email: metis0327@gmail.com. Na-Eun Kim, Email: silverly23@naver.com. Jung Min Sung, Email: tjdwjdals00@knu.ac.kr. Kyung-Won Doh, Email: gregdoh97@gmail.com. Minju Kim, Email: minzuz01@naver.com. Jiwoong Choi, Email: wooong0110@knu.ac.kr. Eunji Ha, Email: eha@knu.ac.kr. Hee Jin Cho, Email: heejincho@knu.ac.kr. Byung Sun Min, Email: bsmin@cu.ac.kr. Jong Hyun Kim, Email: kimjohn@cu.ac.kr. Sin-Hyeog Im, Email: iimsh@postech.ac.kr. Moon-Chang Baek, Email: mcbaek@knu.ac.kr. Supplementary data The following are the Supplementary data to this article: Multimedia component 1 Multimedia component 2 Multimedia component 3 Multimedia component 4 Multimedia component 5 References Associated Data Supplementary Materials Multimedia component 1 Multimedia component 2 Multimedia component 3 Multimedia component 4 Multimedia component 5 Data Availability Statement The data that support the findings of this study are available from the corresponding author upon reasonable request.