Chemotactic Gold Nanozyme‐Powered Flasklike Pentosan Nanobots for Tumor‐Specific Drug Delivery

ABSTRACT

Enzyme‐powered nanobots represent a promising strategy for active and targeted drug delivery due to their autonomous propulsion and directional motion. However, the instability and environmental vulnerability of natural enzymes severely compromise their functionality in the tumor microenvironment (TME). Herein, we report a chemotactic nanobot system powered by gold nanozymes and constructed within a flasklike pentosan architecture for tumor‐specific drug delivery. Benefiting from the intrinsic glucose oxidase‐like and peroxidase‐like activities of gold nanozymes, these nanobots exploit endogenous glucose and elevated hydrogen peroxide (H₂O₂) levels in the TME to achieve sustained propulsion, chemotactic migration along H₂O₂ gradients, enhanced tumor penetration, and microenvironment‐responsive drug release. In vivo studies demonstrate a 4.7‐fold increase in tumor accumulation and a tumor growth inhibition rate of 75.8%, significantly outperforming free doxorubicin treatment. This study presents a nanozyme‐powered chemotactic nanobot platform enabling dynamic navigation of tumor biochemical gradients and precise cancer chemotherapy.

Au nanozyme‐powered, flasklike pentosan nanobots (AuFPNbots) as a robust and dynamic platform for precise cancer chemotherapy have been reported. The AuFPNbots utilize their intrinsic dual nanozyme activities to achieve catalytic self‐propulsion and chemotactic navigation within the TME, enabling deep tumor penetration and TME‐responsive drug release.

1. Introduction

The efficacy of cancer chemotherapy is critically limited by the lack of efficient and selective drug delivery to tumor tissues. Conventional nanocarriers rely predominantly on passive accumulation through the enhanced permeability and retention (EPR) effect, which often results in heterogeneous intratumoral distribution and insufficient penetration into dense tumor matrices [1, 2, 3]. To overcome these limitations, nanobots capable of autonomous motion have emerged as an attractive paradigm, enabling active transport, enhanced penetration, and improved spatial control of therapeutic payloads [4, 5, 6, 7, 8, 9, 10, 11, 12]. A variety of nanobot systems have been developed for tumor therapy, including magnetically actuated [13, 14, 15], light‐driven [16, 17, 18], chemically propelled [19, 20, 21], and enzyme‐powered [22, 23] nanobots. Among them, enzyme‐powered nanobots are particularly appealing for in vivo applications because they can continuously harvest energy from endogenous biological substrates such as glucose [24] and urea [25, 26, 27, 28], thereby enabling long‐term autonomous propulsion without external fields. However, the practical application of enzyme‐powered nanobots for tumor therapy remains constrained by the intrinsic instability of natural enzymes under complex pathological conditions [29, 30]. Particularly in the harsh catalytic environment of the tumor microenvironment (TME), acidic pH, oxidative stress, and elevated protease activity render protein‐based natural enzymes prone to structural denaturation and proteolytic degradation, thereby reducing their catalytic activity and impairing autonomous propulsion [31, 32]. These limitations highlight the need for more stable catalytic systems to power nanobots.

Nanozyme‐powered propulsion offers a promising alternative by combining enzyme‐mimicking catalytic activity with the physicochemical robustness of nanomaterials [33, 34, 35, 36]. Compared with natural enzymes, nanozymes generally exhibit greater tolerance to harsh catalytic conditions and are more amenable to structural and functional regulation, which facilitates their integration into multifunctional nanoplatforms [37]. Therefore, nanozyme‐powered nanobots may offer a more reliable and robust platform for active, tumor‐specific drug delivery. Among various nanozyme candidates, iron‐based nanozymes [38], carbon‐based nanozymes [39], and precious‐metal‐nanozymes [40] have been widely investigated for tumor catalytic therapy. Among these, gold (Au) nanozymes are particularly attractive due to their multiple enzyme‐mimicking activities, high catalytic efficiency, excellent biocompatibility, and exceptional stability under TME‐relevant conditions [41]. Accordingly, Au nanozyme‐powered nanobots hold strong potential for overcoming the limited stability and insufficient intratumoral delivery efficiency associated with conventional enzyme‐powered systems. Beyond propulsion, effective tumor therapy further requires precise tumor targeting and deep tissue penetration. Notably, solid tumors are characterized by elevated and spatially heterogeneous hydrogen peroxide (H₂O₂) concentrations, which serve as intrinsic biochemical gradients within the TME. Harnessing such endogenous gradients to achieve chemotactic navigation represents an appealing yet underexplored strategy for guiding nanobots toward tumor regions in a biomimetic manner.

In this work, we report Au nanozyme‐powered, self‐homing flasklike pentosan nanobots (AuFPNbots) that integrate catalytic propulsion, chemotactic targeting, and drug delivery within a single platform. Owing to the intrinsic glucose oxidase‐like (GOx‐like) and peroxidase‐like (POD‐like) activities of Au nanozymes, the AuFPNbots utilize endogenous glucose and elevated H₂O₂ in the TME to generate sustained self‐propulsion and directional chemotactic motion. The flasklike pentosan scaffold, prepared via a hydrothermal strategy, encapsulates Au nanozymes and doxorubicin (Dox) within an alginate‐based hydrogel core, enabling efficient drug loading and TME‐responsive release. Cloaking with a homologous tumor cell membrane further endows the nanobots with immune evasion, homologous recognition, and enhanced cellular internalization. The AuFPNbots migrate along H₂O₂ concentration gradients, mimicking immune cell chemotaxis, which allows them to overcome extracellular matrix (ECM) barriers and actively accumulate in deep tumor regions. Conceptually, this platform differs from previously reported nanozyme‐powered or chemotactic nanobots by integrating a single Au nanozyme engine with flasklike drug confinement, homologous membrane targeting, and TME‐responsive release in one coordinated design. As a result, propulsion is not used as an isolated function, but is coupled with chemotactic guidance and biomimetic targeting to promote barrier crossing, deep tumor penetration, and more effective intratumoral drug delivery. By integrating nanozyme‐powered propulsion, chemotactic navigation, and biomimetic targeting, this work establishes a synergistic nanobot platform for precise and efficient cancer chemotherapy.

2. Results and Discussion

The fabrication of AuFPNbots was schematically illustrated in Figure 1A, following previously established protocols [42, 43, 44, 45, 46]. Briefly, polymerization of ribose oligomers was carried out at 160°C on the surface of a mixed nanoemulsion composed of P123 and sodium oleate, yielding flasklike pentosan nanoparticles (FPNPs). Scanning electron microscopy (SEM) imaging revealed that the prepared FPNPs had a uniform asymmetric flasklike structure with a body length of approximately 970 ± 85 nm (Figure S1). Transmission electron microscopy (TEM) imaging further showed a single‐opening hollow architecture with a channel diameter of 100 ± 30 nm and an internal cavity of approximately 230 ± 45 nm (Figure S2), providing sufficient space for cargo encapsulation. Before this step, Au nanozymes were successfully synthesized by reducing AuCl₄ ⁻ with sodium borohydride as the reducing agent [47]. The as‐prepared Au nanozymes exhibited good dispersibility, with an average particle size of 4 ± 0.3 nm (Figure S3A) and a hydrodynamic diameter of approximately 4 ± 0.4 nm (Figure S3B). Subsequently, a sodium alginate (SA) solution containing Au nanozymes was introduced into the cavities of the FPNPs under vacuum conditions. The resulting complexes were then cross‐linked in a CaCl₂ solution, forming FPNPs with Au nanozymes embedded within the hydrogel (designated as AuFPNPs). These AuFPNPs were immersed in a Dox solution to facilitate drug loading, resulting in Dox‐loaded AuFPNPs. To enhance targeting, the 4T1 tumor cell membrane was isolated and processed into nanovesicles, which were then fused with the surface of AuFPNPs via the ultrasonic method to yield AuFPNbots (Figure 1B). TEM images confirmed the successful incorporation of Au nanozymes throughout the bottom, mouth, and neck regions of the flasklike architecture (Figure 1C,D), while energy‐dispersive X‐ray spectroscopy (EDX) elemental mapping verified the presence of C, S, Ca, and Au, consistent with cell membrane coating, hydrogel encapsulation, and Au nanozymes loading (Figure 1E). To evaluate the encapsulation of alginate hydrogel within FPNPs, QDs were used to label the FPNPs with and without hydrogel loading (Figure S4 and Video S1). Confocal laser scanning microscopy (CLSM) images showed red fluorescence in hydrogel‐loaded cavities, while cavities without hydrogel showed no fluorescence, confirming effective hydrogel encapsulation within FPNPs. The phase composition of the AuFPNbots was analyzed by using X‐ray diffraction (XRD) analysis. As shown in Figure 1F, typical diffraction peaks at diffraction angles of 38.2°, 44.4°, 64.6°, and 77.6° corresponding to the face‐centered cubic phase (111), (200), (220), and (311) facets of the Au appeared, matching the characteristic peaks of Au (PDF#04‐0784) [48], confirming the successful loading of Au nanozymes. The UV–Vis absorption spectra of Dox‐loaded AuFPNbots showed a characteristic absorption band at 480 nm corresponding to Dox, confirming its successful loading (Figure 1G). The Dox loading amount in AuFPNbots was determined by UV–Vis absorption spectroscopy at 480 nm (Figure S5). According to the Dox standard curve, the amount of loaded Dox was calculated to be 0.567 mg, corresponding to a drug loading content of 36.2% (w/w). Zeta potential measurements revealed a surface charge of −31.6 ± 1.2 mV for AuFPNbots, closely matching that of the tumor cell membrane, indicating successful membrane fusion (Figure 1H). Sodium dodecyl sulfate‐polyacrylamide gel electrophoresis (SDS‐PAGE) analysis (Figure S6) confirmed that membrane proteins were preserved during synthesis. Fluorescence images (Figure S7) further revealed strong co‐localization of green fluorescence signal (DiO‐labeled 4T1 tumor cell membrane) and AuFPNbots, confirming successful membrane coating. Dox release studies demonstrated a pronounced pH‐responsive behavior, with drug release at pH 6.5 occurring markedly faster than that at physiological pH 7.4 (Figure 1I), attributable to protonation‐enhanced solubility of Dox and weakened hydrogel interactions.

FIGURE 1.

Construction and characterization of AuFPNbots. (A) Schematic illustration of the synthesis process of AuFPNbots. (B) TEM image of AuFPNbots. The inset highlights the tumor cell membrane coating on the nanobot surface. (C,D) The magnified TEM images of Au nanozymes in the bottle mouth (C) and neck (D) regions of the flasklike nanobots. (E) EDX elemental mapping of the distribution of C, S, Ca, and Au elements in AuFPNbot. (F) XRD patterns of FPNPs, Au nanozymes, and AuFPNbots. (G) UV–Vis absorption spectra of FPNPs and Dox‐loaded AuFPNbots. (H) Zeta potentials of FPNPs, AuFPNPs, cell membrane, and AuFPNbots. (n = 3) (I) Cumulative Dox release profiles of AuFPNPs and AuFPNbots under different pH conditions (pH 7.4 and pH 6.5), simulating physiological and TME. (J) Schematic illustration of the cascade catalytic activity of AuFPNbots. (K) UV–Vis absorption spectra of AuFPNbots with oxTMB. (L) The absorbance of oxTMB after different concentrations of AuFPNbots reacting with glucose. (M) UV–Vis absorption spectra of oxTMB by AuFPNbots under different glucose concentrations. (N) Zeta potential of Dox‐loaded AuFPNbots in PBS over 7 days. (n = 3). Data are presented as means ± SD.

The catalytic functionality of AuFPNbots was subsequently investigated. Owing to the intrinsic GOx‐like and POD‐like activities of Au nanozymes, AuFPNbots catalyze a cascade reaction, in which glucose is converted into gluconic acid and H₂O₂, followed by H₂O₂‐mediated generation of hydroxyl radicals (•OH) (Figure 1J) [49, 50, 51]. This cascade reaction was confirmed using the 3,3',5,5'‐tetramethylbenzidine (TMB) oxidation assay, as evidenced by the formation of oxidized TMB (oxTMB), which displayed a characteristic absorbance peak at 652 nm. Only AuFPNbots exhibited a pronounced colorimetric response, verifying their intrinsic GOx‐like activity (Figure 1K). Moreover, increasing the concentration of AuFPNbots progressively intensified the blue color of the oxTMB solution, accompanied by a corresponding increase in absorbance at 652 nm (Figure S8 and Figure 1L). This can be attributed to higher concentrations of the AuFPNbots catalyzing the production of more •OH, thus leading to stronger oxidative behavior. Similarly, when the concentration of the AuFPNbots was 50 µg/mL, the oxTMB absorption signal showed a similar upward trend, increasing progressively with increasing glucose concentration (Figure 1M and Figure S9). To further analyze the catalytic kinetics of GOx‐like enzyme activity, AuFPNbots were incubated with different glucose concentrations. As shown in Figure S10A, the GOx‐like enzyme kinetics conform to the Michaelis‐Menten equation. And the K_m and V_max for GOx‐like activity of AuFPNbots were further calculated to be 10.3 mm and 8.3 × 10⁻⁸ M/s according to the Lineweaver–Burk plot (Figure S10B). Depletion of glucose suppresses tumor energy provision while enhancing POD‐like catalytic activity via increased H₂O₂ production. The POD‐like activity of AuFPNbots was performed using H₂O₂ and TMB as the substrate [52]. As shown in Figure S11, AuFPNbots can significantly drive the generation of oxTMB in the presence of H₂O₂, generating the characteristic blue color. By contrast, FPNPs failed to trigger this colorimetric change, thereby verifying that the POD‐like catalytic function originates from the Au nanozymes. AuFPNbots exhibited pronounced POD‐like activity under acidic conditions but only minimal oxTMB formation at neutral pH (Figure S12), demonstrating their capacity for selective TME activation while ensuring safety in normal tissues. Additional experiments evaluating the influence of AuFPNbots dosage on catalytic performance revealed a clear dependence of the POD‐like activity on both time and concentration (Figure S13). The synergistic GOx‐POD‐like cascade reaction was further confirmed by enhanced oxTMB generation upon simultaneous exposure to glucose and TMB (Figure S14). Collectively, these results provide compelling evidence that AuFPNbots possess GOx‐POD‐like enzyme catalytic activities.

Stability studies reveal that the zeta potential remained essentially unchanged after 7 days of storage in PBS (pH 7.4), confirming the stability of AuFPNbots under physiological conditions (Figure 1N). Next, the stability of the enzymatic activity of Au nanozymes within AuFPNbots was assessed. After 7 days of storage, AuFPNbots preserved about 90.7%, 85.0%, and 81.3% of their original catalytic activity at 4°C, 25°C, and 37°C, respectively (Figure S15A). Moreover, they exhibited robust catalytic activity over a wide pH range of 4.0–9.0 (Figure S15B) and a temperature range of 20°C–70°C (Figure S15C). These characteristics are advantageous for practical applications, as the catalytic performance of most natural enzymes is highly sensitive to pH and temperature. In addition, the recyclability study showed that AuFPNbots retained 93% of their cascade catalytic activity after five consecutive cycles (Figure S15D), highlighting a distinctive advantage over natural enzymes, which generally lack such reusability. To further evaluate their long‐term operational stability under tumor‐relevant conditions, both AuFPNbots and natural enzyme‐loaded nanobots were incubated in a simulated TME buffer (pH 6.5, 10% FBS, 37°C) for 10 days, and their catalytic activity was monitored throughout this period. As shown in Figure S16, AuFPNbots retained over 85% of their initial activity after 10 days, whereas the natural enzyme‐loaded nanobots preserved only about 70% under the same conditions. This result demonstrates the superior long‐term stability of the nanozyme‐powered system and highlights an important advantage of AuFPNbots for sustained operation in the TME.

Upon entering the bloodstream, nanoparticles interact with various blood components, leading to the formation of a “protein corona” on their surface [53]. To evaluate whether enveloping AuFPNbots with a tumor cell membrane could suppress the formation of this protein layer, we quantified the adsorbed proteins using SDS‐PAGE analysis. As illustrated in Figure S17, following a 4‐h incubation in plasma, AuFPNbots exhibited markedly reduced protein adsorption relative to the AuFPNPs group. These findings indicate that the tumor cell membrane coating strategy efficiently minimizes protein binding, thereby decreasing the overall protein content within the corona.

To elucidate the propulsion and chemotactic mechanisms of AuFPNbots, their dynamic behavior was systematically investigated under varying concentrations of glucose and H₂O₂, which serve as endogenous fuel sources in the TME. As illustrated in Figure 2A, Au nanozymes embedded within the flasklike architecture catalyze glucose oxidation and H₂O₂ decomposition, generating a cascade reaction that drives autonomous motion. Glucose molecules diffuse into the cavity through the bottle opening and are consumed, while the reaction products are expelled asymmetrically, establishing local concentration gradients along the longitudinal axis. Consistent with previous reports, this asymmetric solute flux induces self‐diffusiophoretic propulsion, directing the nanobot from the closed base toward the open neck as a consequence of momentum conservation and fluid imbalance [54, 55]. Representative trajectories of AuFPNbots in solutions with 5 mm glucose concentrations are presented in Figure 2B (captured from Video S2). Representative trajectories of AuFPNbots in solutions with different glucose concentrations are shown in Figure S18A. A clear correlation was observed between increasing glucose levels and enhanced nanobot displacement, indicating a significant improvement in active diffusion. This amplified diffusion behavior exhibited by autonomously powered nanobots can be quantitatively described by the following mathematical expression [56, 57]

$$\mathrm{MSD}=4\mathrm{D}\mathrm{\Delta }\mathrm{t}+\frac{V^2{\tau }_{\mathrm{R}}^2}{2}\left[\frac{2\mathrm{\Delta }\mathrm{t}}{{\tau }_{\mathrm{R}}}+e^{-\frac{2\mathrm{\Delta }\mathrm{t}}{{\tau }_{\mathrm{R}}}}-1\right]$$ (1)

FIGURE 2.

Self‐propulsion motion and positive chemotaxis analysis of AuFPNbots. (A) Schematic illustration of the propulsion mechanism and motion direction of AuFPNbots powered by catalytic reactions. (B) Representative time‐lapse images showing the self‐propelled motion of AuFPNbots in 5 mM glucose. (C, D) MSD versus time interval of AuFPNbots at different glucose concentrations (C) and different H₂O₂ concentrations (D) over 5 s. (E) Schematic illustration of the chemotactic motion of AuFPNbots along the H₂O₂ concentration gradient (pH = 6.0). (F,G) Fluorescent images (F) and corresponding fluorescence quantification values (G) of individual chambers at different time points (n = 3). (H–K) Schematic (H) and time‐lapse images (I) and correspondingly red‐light intensity of the chemotactic behavior of AuFPNbots in response to PBS (J) and H₂O₂ (K) source in the vertical cuvettes. (L) Mean red‐light intensity in the upper and lower cuvette regions under different top chemotactic sources, quantified from Figure 2J,K (n = 3). Data are presented as means ± SD. (M) Time‐lapse images displaying the tracking trajectories of Brownian motion and positive chemotaxis movement of AuFPNbot under 4T1 MTS solution (12 s; the insertion time was the motion time of the nanobots). One‐way ANOVA followed by Tukey's multiple comparisons test, ^* p <0.05; ^** p <0.01; ^*** p <0.001; ^**** p <0.0001.

Here, MSD represents the mean squared displacement, D refers to the translational diffusion coefficient, Δt indicates the time interval, V corresponds to the velocity, and τ_R stands for the rotational time. The variation of MSD with respect to Δt is plotted in Figure 2C. In the absence of glucose, the MSD increases linearly, which is characteristic of Brownian motion. However, as glucose concentration rises, the MSD follows a parabolic trend that aligns with Equation (1), thereby validating the autonomous propulsion behavior of AuFPNbots. To further quantify this enhanced diffusion, the effective diffusion coefficient (D _eff) was calculated using the following formula [58, 59]

$${\mathrm{D}}_{\mathrm{eff}}=\mathrm{D}+1/4{\mathrm{V}}^2{\tau }_{\mathrm{R}}$$ (2)

Upon increasing the glucose concentration, both active diffusion and average speed of AuFPNbots showed a rapid increase, demonstrating fuel‐concentration‐dependent movement (Figure S18B). Time‐lapse images in Figure S19, taken from Video S3, show that the AuFPNbot moves from the rounded bottom toward the opening neck. The influence of H₂O₂ concentration on nanobot motility was similarly examined (Video S4). MSD analysis demonstrated a clear enhancement in motion capability with increasing H₂O₂ concentration at pH 6.0 (Figure 2D), with typical trajectories shown in Figure S20A. Both D _eff and velocity exhibited nonlinear increases, indicating a transition from passive Brownian diffusion to active enhanced diffusion as H₂O₂ levels increased (Figure S20B). Obviously, the movement of AuFPNbots is influenced by the concentrations of glucose and H₂O₂, enhancing their motility in the TME.

Beyond enhanced motility, directional chemotaxis toward TME‐associated H₂O₂ gradients is essential for self‐guided tumor targeting. To assess whether AuFPNbots can migrate along the H₂O₂ concentration gradient of the TME in vitro, a stable H₂O₂ concentration gradient was established in a Y‐shaped microfluidic channel using agarose gels containing 100 µm H₂O₂ as the chemotactic source, a concentration within the reported pathophysiological range of tumor‐associated H₂O₂ (typically 50–100 µm) [60]. Although this imposed in vitro gradient does not fully recapitulate the spatial and temporal complexity of real tumors, it provides a simplified and controllable model of a physiologically relevant H₂O₂‐enriched chemotactic cue. As illustrated in Figure 2E, AuFPNbots were introduced into the left reservoir, while an agarose gel containing 100 µm H₂O₂ was placed in reservoir ii, and the entire channel was filled with a homogeneous solution of 5 mm glucose. Reservoir i, which branched off and contained a low concentration of H₂O₂, served as the control. Time‐lapse imaging under the H₂O₂ concentration gradient (Figure 2F) showed marked accumulation of AuFPNbots in reservoir ii after 2 h, while only minimal changes were observed in reservoir i. For quantitative analysis, the mean fluorescence intensity (MFI) of each reservoir was measured using Image J, showing a significant increase only in reservoir ii and thereby confirming the positive chemotaxis of AuFPNbots toward the H₂O₂ source (Figure 2G). Chemotactic behavior under a vertical gradient was further validated using a cuvette‐based assay (Figure 2H). In the presence of a 100 µm H₂O₂‐containing agarose gel at the top, time‐lapse imaging revealed a progressive increase in red‐light scattering intensity near the gel interface (Figure 2I), reflecting upward migration of AuFPNbots. Red fluorescence intensity analysis further quantified this microscale chemotactic behavior (Figure 2J,K). The mean red‐light intensity at the top of the H₂O₂‐containing cuvettes increased by approximately 63% after 2 h, confirming the directional migration of AuFPNbots (Figure 2L). Collectively, these results demonstrate that AuFPNbots can actively migrate toward H₂O₂ chemoattractant sources in both horizontal and vertical configurations, highlighting the potential of their directional self‐propulsion for deep tumor penetration. The dynamic behavior of AuFPNbots within a tumor‐mimicking cellular microenvironment was further investigated using 4T1 three‐dimensional multicellular tumor spheroids (3D MTSs). Bright‐field microscopy revealed that, in the absence of H₂O₂ concentration gradients, AuFPNbots exhibited purely stochastic Brownian motion within the 3D MTS medium (Figure 2M and Video S5). In contrast, under an H₂O₂ concentration gradient characteristic of the TME, AuFPNbots displayed pronounced directional migration, indicating that their chemotactic responsiveness is preserved in complex, cell‐dense environments. In our design, endogenous glucose and H₂O₂ can both contribute as fuels for self‐propelled motion, whereas endogenous H₂O₂ gradients in the TME provide the directional cue for chemotactic migration. Under physiological conditions, H₂O₂ is typically maintained within the low nanomolar range [61], whereas H₂O₂ levels in the tumor microenvironment can increase to approximately 50–100 µm [60, 62]. Because of intratumoral metabolic heterogeneity, the relevant in vivo H₂O₂ gradients are expected to be local, dynamic, and spatially nonuniform. Therefore, in vitro gradient experiments were designed to simulate the local H₂O₂ concentration gradients in tumors in a simplified and controllable manner.

Given that limited penetration across physiological barriers is a major obstacle to effective solid tumor therapy, we next examined whether the observed motility of AuFPNbots could be translated into functional advantages in overcoming such barriers. A Transwell coculture system was used to assess AuFPNbot transport across an endothelial barrier. Human umbilical vein endothelial cells (HUVECs) were cultured in the upper chamber to mimic the vascular endothelium, while 4T1 cells were cultured in the lower chamber to mimic tumor tissue. FPNPs, AuFPNPs, and AuFPNbots were then added to the upper chamber (Figure 3A). The formation of a confluent and functional endothelial barrier was confirmed by stabilization of trans‐endothelial electrical resistance (TEER) at approximately 160 Ω (Figure 3B–E). Following 8 h of incubation, HUVECs internalized large amounts of AuFPNbots, which were then delivered to 4T1 cells in the lower chamber. Compared with the FPNPs and AuFPNPs groups, the AuFPNbots group exhibited higher fluorescence intensity, suggesting more efficient trans‐endothelial transport (Figure 3F,G). Quantitative analysis showed that AuFPNPs achieved ∼3.7‐fold higher trans‐endothelial transport efficiency than FPNPs (2.108 vs 0.575), likely owing to nanozyme‐powered motility. Homologous membrane coating further increased the efficiency of AuFPNbots to 2.899 (∼1.4‐fold), while H₂O₂ stimulation boosted it to 4.885 (∼1.7‐fold), indicating additional contributions from homologous targeting and chemotactic propulsion. Fluorescence images and MFI (Figure S21) confirmed that under H₂O₂ conditions, after being captured by the upper HUVECs, more AuFPNbots could enter the lower chamber through the upper layer. These results demonstrate that TME‐activated motility substantially improves the ability of AuFPNbots to overcome endothelial barriers.

FIGURE 3.

Evaluation of the ability of AuFPNbots to cross the endothelial barrier and their lysosomal co‐localization in vitro. (A) Schematic illustration of the in vitro trans‐endothelial barrier models employed to assess the potential vascular penetration capabilities of FPNPs, AuFPNPs, AuFPNbots, and AuFPNbots+H₂O₂. (B,C) Schematic illustration of in vitro TEER measurement of endothelial barrier integrity (B) and its resistance principle(C). (D) Measured TEER values of the HUVECs monolayer cultured in the upper chamber of a 24‐well Transwell increased progressively over time until reaching saturation. (n = 3). Data are presented as means ± SD. (E) Image of detecting the resistance value of the endothelial barrier in vitro. (F,G) CLSM images of HUVEC layers (F) and 4T1 cells (G) after co‐incubation with different sample groups for 8 h. In the images, HUVEC and 4T1 cell nuclei were stained with DAPI, whereas FPNPs, AuFPNPs, and AuFPNbots were labeled with Dil. (H) CLSM images of 4T1 cells treated with the AuFPNbots (red) for 1 h, followed by staining with Lyso–Tracker (green) and DAPI (nucleus, blue). (I) Represent line scan profiles of fluorescence intensities at the white arrows in (H). (J) Co‐localization analysis (firework plots and Pearson's R‐value) of AuFPNbots and lysosomes in 4T1 cells at 1 h, performed using Image J. (K) CLSM images of 4T1 cells after incubation with AuFPNbots (red) for 4 h, followed by staining with Lyso‐Tracker (green) and DAPI (nucleus, blue). (L) Representative line scan profiles of fluorescence intensities along the white arrows indicated in (K). (M) Co‐localization analysis (firework plots and Pearson's R‐value) of AuFPNbots and lysosomes in 4T1 cells at 4 h, performed using Image J.

The cellular uptake of nanodrugs plays a crucial role in the realization of their therapeutic effects. Fluorescence imaging revealed that the cellular uptake of AuFPNbots was significantly higher than that of AuFPNPs at 1, 3, and 6 h (Figure S22A,B), suggesting a targeting advantage conferred by the tumor cell membrane coating. Quantitative analysis at 6 h showed that AuFPNbots exhibited ∼1.4‐fold higher cellular uptake than AuFPNPs (16.794 vs 11.96), indicating an additional contribution from membrane‐mediated homologous targeting. In the presence of H₂O₂, AuFPNbot uptake further increased to 31.741, corresponding to an additional ∼1.9‐fold enhancement and supporting a further role of H₂O₂‐induced chemotactic propulsion. Together, these results suggest that both homologous membrane targeting and active chemotactic motion contribute to the enhanced cellular internalization of AuFPNbots. Moreover, the corresponding red MFI data confirmed that cellular internalization increased with extended incubation time (Figure S22C,D).

Following cellular uptake, nanocarrier‐mediated intracellular delivery is often restricted by poor endosomal escape, which limits cargo release into the cytosol and promotes subsequent lysosomal degradation [63]. To evaluate the lysosomal escape capability of AuFPNbots, the co‐localization of DiI‐labeled AuFPNbots (red) with LysoTracker‐stained lysosomes (green) in 4T1 cells was examined by CLSM. Lysosomal escape was assessed based on fluorescence intensity analysis (Figure 3I,L) and Pearson's R values (Figure 3J,M). As shown in Figure 3H, under H₂O₂ conditions, a pronounced yellow fluorescence signal was observed after 1 h of incubation, indicating that AuFPNbots were still largely confined within lysosomes at this stage. Consistently, the Pearson's R‐value at 1 h was 0.92. In contrast, after 4 h of incubation, the red fluorescence signal of AuFPNbots was clearly separated from the lysosomal green fluorescence (Figure 3K), indicating successful lysosomal escape. Correspondingly, the Pearson's R value decreased to 0.32 at 4 h. These findings demonstrate the efficient lysosomal escape capability of AuFPNbots, which may be facilitated by their unique motility after cellular internalization.

The highly dense and viscous ECM in tumors acts as a physical barrier that impedes drug delivery and reduces anti‐tumor treatment efficacy [64, 65]. Therefore, developing innovative strategies to enhance the penetration efficiency of drugs at tumor sites is of great significance. The AuFPNbots developed herein are designed to overcome both endothelial and ECM barriers through catalytic self‐propulsion and TME‐responsive drug release. Driven by endogenous H₂O₂ gradients, AuFPNbots exhibit directional motility and enhanced cellular uptake, enabling deep tumor infiltration (Figure 4A). To evaluate tumor penetration in vitro, 3D MTSs were employed as a physiologically relevant tumor model that recapitulates the architecture and transport limitations of solid tumors (Figure 4B). In contrast to two‐dimensional monolayer cell cultures, 3D MTSs more accurately recapitulate key features of solid tumors, including architecture, microenvironment, and three‐dimensional morphology [66]. As shown in Figure 4C, CLSM z‐stack images were collected at different depths from the top to the equatorial plane of the 3D MTSs to visualize fluorescence distribution and assess nanobot penetration in the spheroids. Fluorescence signals from FPNPs and AuFPNPs were largely confined to the peripheral regions of the spheroids, indicating limited penetration. In contrast, AuFPNbots exhibited strong fluorescence signals at depths up to 150 µm, covering most of the spheroid cross‐section (Figure 4D), which was further confirmed by three‐dimensional reconstructed images (Figure 4E). Notably, AuFPNbots penetrated deeper into the 3D MTSs in the presence of H₂O₂. Semi‐quantitative analysis at 150 µm further confirmed stronger and broader red fluorescence in the AuFPNbots+H₂O₂ group (Figure 4F,G). Quantitative analysis at 150 µm depth revealed that AuFPNPs achieved ∼1.7‐fold greater fluorescence intensity than FPNPs, likely reflecting the contribution of nanozyme‐powered motility. The incorporation of the homologous membrane coating in AuFPNbots yielded a further ∼2.4‐fold increase, suggesting an additional contribution from membrane‐mediated homologous targeting. In the presence of H₂O₂, AuFPNbot penetration rose by an additional ∼1.6‐fold, consistent with the further amplifying effect of H₂O₂‐induced chemotactic propulsion. These results suggest that the presence of H₂O₂ significantly enhances the penetration of AuFPNbots into 3D MTSs via self‐propulsion. The enhanced penetration capability enables efficient delivery of nanobots to deeper tumor regions, thereby augmenting their anti‐tumor efficacy.

FIGURE 4.

Penetration and antitumor efficacy of AuFPNbots in 3D MTSs. (A) Schematic illustrating the chemotaxis of AuFPNbots toward tumor tissue. (B) Schematic diagram of the culture and penetration of 3D MTSs. (C) Schematic diagram of the 3D MTS model observed along the Z axis, with the top surface of the MTSs defined as 0 µm, and z‐stack scanning performed from the top to the equatorial cross‐section at 25 µm thickness. (D) CLSM images depicting in vitro penetration of FPNPs, AuFPNPs, AuFPNbots, and AuFPNbots+H₂O₂ in 3D MTSs for 24 h. (E) 3D surface plot (Image J, from D) illustrating the distinct fluorescence intensity profiles between the nanobot and control groups across a 150 µm MTS section. (F,G) The corresponding fluorescence intensity (F) and MFI (G) along the yellow arrow of the 3D MTSs treated with different groups via Image J. (H) Microscopic images of 3D MTSs treated with Dox‐loaded nanobots were recorded over three days. (I) Live cell images of different treatment groups on MTSs after seven days. (J,K) Volume measurements (J) for different groups treated with nanobots were quantitatively analyzed by Image J alongside their corresponding inhibition rates (K). (n = 3). Data are presented as means ± SD. One‐way ANOVA followed by Tukey's multiple comparisons test, ^* p <0.05; ^** p <0.01; ^*** p <0.001; ^**** p <0.0001.

The functional consequence of enhanced penetration was further assessed by monitoring spheroid ablation following treatment with different Dox‐loaded formulations. While spheroids treated with PBS or FPNPs retained intact morphology after 3 days, and AuFPNPs induced only limited surface cell shedding, AuFPNbots in the presence of H₂O₂ caused pronounced spheroid disruption and extensive cellular disintegration (Figure 4H). Quantitative analysis revealed an inhibition rate of 86.1% for the AuFPNbots group, accompanied by the lowest number of viable cells as confirmed by Calcein‐AM staining (Figure 4I–K). These results demonstrate that AuFPNbots effectively penetrate tumor‐like tissues and achieve targeted drug release within the TME, offering a viable strategy to overcome intratumoral transport barriers and enhance antitumor efficacy.

Tumor formation was confirmed seven days post‐inoculation, after which near‐infrared fluorescence imaging was employed to evaluate the in vivo fluorescence distribution and tumor‐associated signals. Therapeutic formulations labeled with the fluorescent probe 1,1′‐dioctadecyl‐3,3,3′,3′‐tetramethylindotricarbocyanine iodide (DiR) were intravenously administered via the tail vein, followed by time‐dependent imaging (Figure 5A). The AuFPNbots‐treated group showed significantly higher tumor‐associated fluorescence intensity than the FPNPs and AuFPNPs groups (Figure 5B). Quantitative fluorescence analysis revealed that the tumor fluorescence intensity peaked at 48 h post‐injection, with mice treated with AuFPNbots showing the strongest tumor fluorescence signal intensity among the three groups (Figure 5C). This enhanced fluorescence signal may be attributed, at least in part, to the chemotactic reaction mediated by internal Au nanozymes and the self‐homing effect of the tumor cell membrane. These results were further corroborated by ex vivo fluorescence imaging of excised tumors and major organs (Figure 5D). Quantitative fluorescence analysis revealed that tumor fluorescence intensity was 2.4‐ and 4.7‐fold higher in the AuFPNbots group than in the AuFPNPs and FPNPs groups, respectively (Figure 5E), indicating enhanced tumor‐associated localization of AuFPNbots. Time‐dependent analysis further demonstrated that maximal fluorescence intensity in the AuFPNbots group was achieved at 48 h post‐injection, suggesting that the tumor‐associated fluorescence signal reached its maximum at this time point (Figure 5F). It should be noted that near‐infrared fluorescence imaging provides semi‐quantitative biodistribution information, and the measured signal may be influenced by tissue attenuation, dye stability, and signal normalization. Therefore, the enhanced fluorescence signal observed for AuFPNbots reflects the combined effects of improved tumor‐associated localization, deeper intratumoral penetration, and prolonged retention. To comparatively evaluate tumor penetration and targeted accumulation, FPNPs, AuFPNPs, and AuFPNbots were labeled with QDs. Fluorescence imaging of tumor sections 48 h after intravenous administration revealed markedly distinct intratumoral distribution profiles (Figure 5G). Notably, tumors treated with AuFPNbots exhibited strong and widely distributed green fluorescence signals spanning nearly the entire tumor mass, suggesting enhanced intratumoral penetration. In contrast, FPNPs displayed only weak and localized fluorescence, suggesting limited tumor infiltration. Moreover, in the AuFPNbots group, Dox fluorescence was extensively distributed throughout the tumor and reached the highest intensity among all groups (Figure S23), further supporting the improved intratumoral distribution and drug delivery performance of AuFPNbots.

FIGURE 5.

Comprehensive in vivo evaluation of targeted enrichment and tumor penetration by chemotaxis‐based nanobots in tumor‐bearing mice. (A) A protocol for establishing a 4T1 tumor model in mice and conducting in vivo targeted experiments. (B,C) In vivo distribution (B) and subsequent quantitative analysis (C) of tumor‐site fluorescence intensity for DiR‐labeled FPNPs, AuFPNPs, and AuFPNbots in 4T1‐bearing mice at various time points (white circle in B: tumor site). (D,E) Representative ex vivo fluorescence imaging (D) and semi‐quantitative signal analysis (E) of excised major organs and tumors at 48 h post‐injection. (F) MFI of tumors at different time points (AuFPNbots group). (G) Representative fluorescence images of tumor tissue sections. In the images, FPNPs, AuFPNPs, and AuFPNbots were labeled with QDs (green), while Dox exhibits intrinsic autofluorescence properties (red), and the nuclei of 4T1 cells were stained with DAPI (blue). One‐way ANOVA followed by Tukey's multiple comparisons test, ^* p <0.05; ^** p <0.01; ^*** p <0.001; ^**** p <0.0001.

To further validate the therapeutic efficacy of AuFPNbot, in vivo antitumor experiments were conducted (Figure 6A). Tumor‐bearing mice were randomly assigned to five treatment groups (n = 3 per group), receiving PBS, free Dox, FPNPs, AuFPNPs, and AuFPNbots, respectively. Except for the PBS group, all formulations were administered at an equivalent Dox dose of 100 mg/kg. Tumor growth was monitored over time, revealing a pronounced suppression of tumor progression in the AuFPNbots‐treated group, which exhibited the smallest tumor volumes among all groups (Figure 6B,C). Tumor imaging further confirmed a significant inhibition of tumor growth after administration of AuFPNbots (Figure 6D). Quantitative analysis demonstrated that AuFPNbots elicited significantly enhanced antitumor efficacy compared with free Dox and non‐propelled nanocarrier controls (Figure 6E,F). Notably, no significant body weight loss was observed in the AuFPNbots‐treated mice, indicating minimal systemic side effects (Figure 6G). Histopathological examination of tumor tissues was conducted to further assess the antitumor efficacy of AuFPNbots. As presented in Figure 6H,I, H&E staining demonstrated extensive necrosis and structural disruption of tumor cells in the AuFPNbots treatment group. TUNEL staining analysis confirmed increased apoptosis of tumor cells induced by AuFPNbots (Figure 6H,J). To further evaluate the effect of AuFPNbots on tumor treatment, immunohistochemical staining analyses of CD31 and Ki67 at the tumor site were performed. CD31 expression, a representative marker of endothelial cells and vascular density, was markedly reduced in tumors treated with AuFPNbots, indicating effective suppression of tumor‐associated vasculature (Figure 6H,K). Similarly, the expression of Ki67, a widely used proliferation marker, was substantially decreased in tumors treated with AuFPNbots (Figure 6H,L). Collectively, these results confirm that AuFPNbots achieve superior antitumor efficacy in vivo. The superior therapeutic efficacy of AuFPNbots arises from the synergistic effects of self‐propelled chemotaxis, enhanced interstitial penetration, pH‐responsive Dox release, homologous membrane‐mediated tumor recognition, and nanozyme‐catalyzed ROS generation. Among these factors, self‐propelled chemotactic behavior is the dominant contributor, as it enables deeper tumor infiltration and more efficient intratumoral drug delivery.

FIGURE 6.

In vivo antitumor efficacy of nanobots in 4T1 tumor‐bearing mice. (A) Schematic overview of the therapeutic mechanism of the nanobots and the treatment regimen in the 4T1 tumor‐bearing mouse model. (B) Tumor growth curve obtained by dynamic measurement of the tumor volume over a two‐week treatment period. (C) The tumor volume of tumor‐bearing mice following intravenous administration of PBS, Dox, FPNPs, AuFPNPs, and AuFPNbots via the tail vein throughout the treatment process. (D,E) Representative tumor images (D) and weights (E) of excised tumors from mice at the conclusion of the treatment with nanobots and other control groups. (F) Quantitative analysis of tumor inhibition rates for each treatment group, calculated using the mean tumor weight of the PBS group and the respective treatment groups at the end of the therapy. (G) Body weight variation in tumor‐bearing mice during different treatments. (H–L) Representative images of H&E, TUNEL, CD31, and Ki67 staining of resected tumor tissues from mice administered with AuFPNbots and other control groups (H), along with the fluorescence intensity quantification results corresponding to H&E (I), TUNEL (J), CD31 (K), and Ki67 (L) staining. One‐way ANOVA followed by Tukey's multiple comparisons test, ^* p <0.05; ^** p <0.01; ^*** p <0.001; ^**** p <0.0001.

Given the importance of biosafety for translational nanomedicine, the biocompatibility of AuFPNbots was systematically evaluated. CCK‐8 assays performed on HUVECs revealed negligible cytotoxicity for FPNPs, AuFPNPs, and AuFPNbots across a broad concentration range (0–300 µg/mL), confirming their favorable biocompatibility (Figure S24). Hemolysis assays revealed minimal hemolytic activity of AuFPNbots (<2%), which was significantly lower than that of the positive control (DI H₂O group) (Figure S25). Furthermore, blood smear analysis after 4 h of incubation showed no morphological alterations in erythrocytes, confirming excellent hemocompatibility (Figure S26). The viability of 4T1 cells decreased with increasing Dox concentration in the Dox‐loaded AuFPNbots group, reaching 40% at 10 µg/mL, likely due to Dox inhibiting DNA replication and RNA transcription (Figure S27). The killing effect of AuFPNbots on 4T1 cells was further confirmed by fluorescence staining, showing increased red fluorescence in the PI channel, especially in the AuFPNbots+H₂O₂ group, indicating enhanced cell death due to enhanced movement behavior of nanobots induced by H₂O₂, leading to increased Dox release. Semi‐quantitative analysis further supported these findings with reduced green and elevated red signals, validating effective chemotherapy (Figure S28). Notably, degradability studies from our prior work demonstrated that the polysaccharide‐based FPNPs framework undergoes near‐complete lysozyme‐mediated degradation within 14 days under physiologically simulated conditions [67]. Upon framework disassembly, the released ultrasmall Au nanozymes (∼4 nm) are anticipated to undergo renal clearance, consistent with the established biodistribution profile of sub‐6 nm Au nanoparticles [68].

To further assess the biosafety of AuFPNbots in vivo, mice were administered PBS, FPNPs, AuFPNPs, or AuFPNbots, respectively. Major organs were harvested and subjected to H&E staining to evaluate potential systemic toxicity. No evident histopathological abnormalities were observed in any of the examined organs across all treatment groups (Figure S29), indicating the favorable in vivo biocompatibility of AuFPNbots. Collectively, these findings demonstrate that the rationally engineered self‐homing nanobot platform exerts potent antitumor activity while causing minimal systemic toxicity, highlighting its potential for tumor‐targeted therapy.

3. Conclusion

In conclusion, we have developed a gold nanozyme‐powered flasklike pentosan nanobot platform for active and tumor‐specific drug delivery. The asymmetric hollow architecture enables efficient integration of catalytic propulsion and therapeutic payloads, allowing the nanobots to exploit endogenous glucose and elevated H₂O₂ in the TME for sustained self‐propulsion and microenvironment‐responsive drug release. Driven by robust nanozyme catalysis and chemotactic self‐homing behavior, the AuFPNbots exhibit markedly enhanced tumor penetration and accumulation, leading to superior antitumor efficacy. Both in vitro and in vivo biocompatibility assessment results consistently show that the system has good biological safety and extremely low systemic toxicity, providing strong support for future clinical translation. More broadly, this study presents a feasible strategy to address some of the limitations associated with natural enzyme‐powered nanobots, particularly in terms of enzymatic instability and limited therapeutic performance. The nanozyme‐powered nanobot system, despite its notable performance, faces critical challenges toward further development and clinical translation, including systematic assessment of long‐term biosafety, immunocompatibility, and in vivo biodistribution, alongside the necessity for robust and precise navigation in complex biological milieus. In future work, we aim to establish a comprehensive preclinical safety evaluation pipeline and validate its therapeutic efficacy in large animal models. The proposed nanozyme‐powered propulsion and chemotactic navigation strategy is readily adaptable to diverse nanozyme catalytic systems and endogenous biochemical gradients, thereby holding promise for broad applications in active drug delivery and precision nanomedicine.

Ethics Statement

This research complies with all relevant ethical regulations. All animal experiments were approved by the Animal Experimentation Ethics Committee of Harbin Institute of Technology (approval number: IACUC‐2024045). The mice were housed in a barrier facility under a 12 h light/12 h dark cycle at 24°C and 50% relative humidity, with free access to food and water.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File 1: adhm71361‐sup‐0001‐SuppMat.docx.

Supporting File 2: adhm71361‐sup‐0002‐MovieS1.mp4.

Supporting File 3: adhm71361‐sup‐0003‐MovieS2.mp4.

Supporting File 4: adhm71361‐sup‐0004‐MovieS3.mp4.

Supporting File 5: adhm71361‐sup‐0005‐MovieS4.mp4.

Supporting File 6: adhm71361‐sup‐0006‐MovieS5.mp4.

Data Availability Statement

The data that supports the findings of this study are available in the supplementary material of this article.