Hollow nanofactory with dotted interior surface produces potent hydroxyl radical to combat drug-resistant large tumors

Abstract

Ferroptosis, a novel form of programmed cell death, is highly dependent on intracellular hydrogen peroxide (H₂O₂) and Fe²⁺ levels. Numerous nanomaterials have been developed to co-deliver H₂O₂ and Fe²⁺ or their prodrugs into tumor cells, aiming to enhance hydroxyl radical (•OH) level via Fenton reaction. However, the •OH generation efficacy is frequently undermined by catalase (CAT) and ferritin in cytoplasm, which can respectively catalyze H₂O₂ to H₂O and oxidize Fe²⁺ to Fe³⁺. Herein, we designed and constructed a hollow nanofactory with dotted interior surface (HNDIS), which is composed of hollow mesoporous iron oxide nanoparticle (HMION, “house” of the nanofactory) with mesoporous channels (“window” of the nanofactory), and ultrasmall gold nanoparticles (AuNP) on the interior surface (“machine” of the nanofactory). The unique “window” allows pass for glucose to co-work with AuNP and HMION to drive H₂O₂ production and Fe²⁺ release, and then generate a huge amount of •OH. Macromolecules cannot pass through the “window”, which prevents the undesirable catalysis of ferritin and CAT. In vitro and in vivo experiments have demonstrated potent •OH production ability and powerful tumor suppression efficacy based on ferroptosis for HNDIS, offering a promising pathway to combat drug-resistant large tumors.

Graphical abstract

A hollow nanofactory with dotted interior surface (HNDIS) was designed and constructed with the function of glucose-driven H₂O₂ generation and self-supplied Fe²⁺ generation in the hollow core of HNDIS, which can produce potent hydroxyl radical to combat drug-resistant large tumors.

1. Introduction

Ferroptosis, a form of programmed cell death (PCD) initiated by iron-dependent lipid peroxidation (LPO), holds great promise for bypassing the drug-resistance disadvantage of tumors [[1], [2], [3]]. In tumor sites, Fe²⁺ can react with hydrogen peroxide (H₂O₂) to produce cytotoxic hydroxyl radicals (•OH) through the Fenton reaction (Fe²⁺ + H₂O₂ = Fe³⁺ + •OH + OH⁻). Therefore, co-delivery of H₂O₂- and Fe²⁺-generating prodrugs into tumors represents a straightforward strategy to amplify intracellular oxidative stress through Fenton reaction, and thereby trigger the ferroptosis of tumor cells [[4], [5], [6], [7]].

To achieve high concentrations of both H₂O₂ and Fe²⁺ within tumors, numerous drug delivery systems (DDS) have been developed to co-deliver H₂O₂-generating agents (e.g., glucose oxidase, GOx) and iron oxide nanoparticles into tumors, aiming to realize high efficacy ferroptosis therapy of tumors [[8], [9], [10]]. For example, Tian et al. prepared a core-shell nanoparticle (MU@P@GS-B), comprising upconversion nanoparticles (UCNPs) as the core, Fe(III)-based metal-organic frameworks (MOFs) as the intermediate layer, and glucose oxidase (GOx) pre-passivated with spiropyran as the outer shell. Under near-infrared (NIR) light irradiation, the MOFs network collapses, releasing free Fe^2+/3+ into the cytoplasm. Meanwhile, UV light activates GOx, enabling it to catalyze glucose to produce H₂O₂ in the cytoplasm. The co-delivery of Fe²⁺ and H₂O₂ jointly provide an efficient Fenton reaction to generate •OH [11]. In addition, Tu et al. fabricated an asymmetric Janus nanomotors (ZnO₂@PDA-Fe), consisting of a ZnO₂ nanoparticle core uniformly encapsulated by a polydopamine (PDA) shell, with Fe²⁺ coordinatively chelated to the PDA shell. Under the acidic tumor microenvironment (TME), the ZnO₂ core undergoes gradual degradation to release H₂O₂. Upon near-infrared (NIR) irradiation, the released Fe²⁺ reacts with H₂O₂ in cytoplasm via Fenton reaction to generate a large amount of •OH, inducing ferroptosis-based damage of tumor cells [6].

However, the cytoplasm contains a large amounts of enzymes, e.g., catalase (CAT) and ferritin. CAT can catalyze H₂O₂ to H₂O, and ferritin can catalyze Fe²⁺ to Fe³⁺ [12]. Moreover, Fe³⁺ has a very low efficiency of Fenton reaction (Fe³⁺ + H₂O₂ = Fe²⁺ + •OOH + H⁺, k ≈ 0.001–0.01 M⁻¹s⁻¹), which hinders the total Fenton reaction efficiency [13,14]. Therefore, re-engineering the structure of DDS to generate a secluded nano-compartment that precludes the catalytic interference of CAT and ferritin with H₂O₂ and Fe²⁺ is imperative to enhance the overall efficiency of the Fenton reaction in tumors.

Herein, inspired by the compartmentalized catalysis in natural organelles, we have engineered a hollow nanofactory with dotted interior surface (HNDIS) that can autonomously accomplish glucose-driven H₂O₂ production and Fe²⁺ release within a confined nano-space, and then generate a huge amount of •OH. Typically, ultrasmall gold nanoparticles (AuNP) are encapsulated inside the interior of hollow mesoporous iron oxide nanoparticles (HMION) by a layer-by-layer (LBL) synthesis strategy, using silicon dioxide as a sacrificial spacer. We chose the LBL strategy over in situ growth to ensure the precise spatial confinement of AuNP within the hollow cavity. Specifically, in situ growth strategies typically lack spatial selectivity, which inevitably leads to uncontrollable AuNP deposition on the exterior surface where they remain exposed to enzymatic interference. In contrast, the LBL approach physically anchors AuNP onto the inner surface prior to shell formation. After the etching of silicon dioxide, the AuNP are anchored onto the inner surface of the HMION, yielding a hollow nanofactory capable of producing •OH. The “house” of the nanofactory (i.e., HNDIS) is HMION. The “window” of the nanofactory is the mesoporous channels of HMION, through which small molecules (e.g., glucose) can pass, but macromolecules (e.g., CAT and ferritin) cannot. The “machine” of the nanofactory is AuNP that are distributed on the interior surface of HMION (Scheme 1a).

Scheme 1.

The schematic representation for the fabrication of hollow nanofactory with dotted interior surface (HNDIS) (a), and the production mechanism of a huge amount of •OH in the nanofactory to overcome drug resistance via ferroptosis therapy (b).

The HNDIS can accumulate in tumors based on the enhanced permeability and retention (EPR) effect. After the HNDIS is internalized into tumor cells, the high level of intracellular glucose can pass through the mesoporous channels, and then get into the HNDIS. The ultrasmall AuNP on the interior surface, functioning as GOx mimics, can catalyze the glucose in the nanofactory into H₂O₂ and gluconic acid (GA). The generated GA subsequently can react with the Fe₃O₄-based HMION, realizing the in situ release of Fe²⁺. The locally released Fe²⁺ can immediately react with the locally generated H₂O₂ through the Fenton reaction, yielding a huge amount of •OH. The potent •OH can trigger extensive LPO to destroy the cell membrane of tumors, resulting in ferroptosis of tumor cells (Scheme 1b).

This new strategy of ferroptosis therapy based on the nanofactory is promising to combat drug-resistant tumors because they are also sensitive to •OH [15,16].

2. Experimental section

2.1. Synthesis of HNDIS

10 mL of SiNP@AuNP@SiNP (5.0 mg/mL) was added into a mixed solution of acetone (14 mL) and H₂O₂ (1.0 mL, 30 wt%) under magnetic stirring (500 rpm). After that, 50 mg of ferrocene was added into the above mixture, followed by transfer into an autoclave under 230 °C for 24 h. After cooling, the above dispersion was washed thrice with Milli-Q water and collected by centrifugation (8000 g, 20 min). The above-prepared dispersion was re-dispersed in the 30 mL of Na₂CO₃ (2.0 M) under magnetic stirring (200 rpm) at 60 °C. After stirring for 6.0 h, the dispersion was washed thrice with Milli-Q water and collected by centrifugation (8000 g, 20 min) to obtain HNDIS. The resulting HNDIS dispersion was stored in a fridge for subsequent use.

2.2. Synthesis of HMION

10 mL of SiNP (5.0 mg/mL) was added into a mixed solution of acetone (14 mL) and H₂O₂ (1.0 mL, 30 wt%) under magnetic stirring (500 rpm). After that, 50 mg of ferrocene was added into the above mixture, followed by transfer into an autoclave under 230 °C for 24 h. After cooling, the above dispersion was washed thrice with Milli-Q water and collected by centrifugation (8000 g, 20 min). The above-prepared dispersion was re-dispersed in the 30 mL of Na₂CO₃ (2.0 M) under magnetic stirring (200 rpm) at 60 °C. After stirring for 6.0 h, the dispersion was washed thrice with Milli-Q water and collected by centrifugation (8000 g, 20 min) to obtain HMION for subsequent use.

2.3. Statistical analysis

Quantitative data are presented as mean ± standard deviation (SD) derived from at least three independent experiments (n ≥ 3). Statistical analysis was carried out using the Student's t-test for two groups, as well as a one-way analysis of variance for more than two groups. All statistical analyses were performed using GraphPad Prism software (Version 8.0, GraphPad Software, USA). The significance level was fixed as ∗ P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, or ∗∗∗∗P < 0.0001.

3. Results and discussion

3.1. Preparation and characterization of the nanofactory

The scheme of detailed step-wise synthesis procedure for HNDIS, illustrating the structural evolution from SiNP to the final HNDIS, is shown in Fig. S1. Fig. 1a–e presents photos of aqueous dispersions for silica nanoparticles (SiNP), AuNP, SiNP@AuNP, SiNP@AuNP@SiNP and HNDIS, showing well-dispersed states without aggregates. Fig. 1f–j shows schematic diagrams of structures for SiNP, AuNP, SiNP@AuNP, SiNP@AuNP@SiNP and HNDIS, and Fig. 1k–o exhibit the corresponding TEM images. SiNP has a monodispersed spherical morphology with ∼60 nm of diameter (Fig. 1k). AuNP display a uniform ultra-small morphology with ∼4 nm of diameter (Fig. 1l). SiNP@AuNP show uniformly distributed AuNP on the surface of SiNP (Fig. 1m), which is ascribed to the robust chelation between AuNP and the amino moieties from APTES [17]. SiNP@AuNP@SiNP exhibits a layered “sandwich” architecture, in which AuNP are encapsulated between inner and outer SiNP layers (Fig. 1n), thereby ensuring the chemical stability of AuNP during the subsequent synthesis processes. After deposition of HMION onto the surface of SiNP@AuNP@SiNP and the subsequent alkaline removal of SiNP, the resultant HNDIS exhibits a mesoporous and hollow structure with dotted interior surface (Fig. 1o), in which AuNP are primarily immobilized on the interior surface of HMION via robust chelation between carboxylic groups and AuNP/HMION. In addition, rather than investigating the influence of varying Au/Fe ratio, we adopted a “saturation AuNP loading strategy” to maximize the overall therapeutic efficacy. Given that the therapeutic efficiency relies on the catalytic efficiency of AuNP, our design priority is to maximize the loading of AuNP. To achieve this, an excess amount of ultrasmall AuNP are introduced to coordinate with the amino groups on the surface of the SiNP, followed by thorough washing to remove unbound particles. This process ensures that the interior surface of the final HNDIS is maximally occupied by AuNP, driving a saturated H₂O₂ generation rate to match the ample Fe supply from HNDIS shell. Zeta potentials of SiNP, AuNP, SiNP@AuNP, SiNP@AuNP@SiNP and HNDIS are measured to be −24.0 ± 5.8, −21.2 ± 4.1, 24.4 ± 5.8, −24.2 ± 6.2, and −28.2 ± 4.8 mV, respectively (Fig. S2). The positive zeta potential of SiNP@AuNP is attributed to the amino-functionalization of the SiNP using APTES. Specifically, the abundant amino groups (-NH₂) introduced onto the silica surface undergo protonation to form ammonium groups (-NH₃⁺), creating a positively charged surface. Despite the attachment of AuNP onto SiNP surface, the final SiNP@AuNP construct retains a net positive charge due to the high density of amino groups. The negative surface charge of HNDIS is expected to realize long blood-circulation after intravenous administration [18]. Hydrodynamic particle sizes of SiNP, AuNP, SiNP@AuNP, SiNP@AuNP@SiNP and HNDIS are respectively 144.9 ± 7.8, 9.8 ± 0.9, 151.3 ± 8.4, 174.3 ± 8.6 and 225.7 ± 6.1 nm measured by dynamic light scattering (DLS) (Fig. S3). After 72 h of incubation in PBS, DMEM or DMEM +10 % FBS, the hydrodynamic size of HNDIS remains respectively at 226.4 ± 15.7, 252.4 ± 15.4 and 284.7 ± 13.7 nm, confirming the excellent colloidal stability of HNDIS under physiological conditions (Fig. S4).

Fig. 1.

(a–o): The dispersion pictures (a–e), schematic structures (f–j), and TEM images (k–o) of SiNP (a, f, k), AuNP (b, g, l), SiNP@AuNP (c, h, m), SiNP@AuNP@SiNP (d, i, n), and HNDIS (e, j, o). (p, q): HAADF-STEM mapping (p) and the corresponding EDX spectrum (q) of HNDIS. (r, s): N₂ adsorption-desorption isotherm (r) and the corresponding pore size distribution (s) of HNDIS. (t): XRD spectra of HMION and HNDIS. (u, v): Full range XPS spectrum (u), and the corresponding high-resolution XPS spectra of Fe 2p (v) for HNDIS. (w): UV–Vis absorption spectra of AuNP and HNDIS dispersions.

The HAADF-STEM mapping of HNDIS reveals significant overlap between Au, Fe and O signals (Fig. 1p), indicating that AuNP are predominantly located on the interior surface of HMION. The EDS spectrum of HNDIS exhibits distinct peaks for C, O, Au and Fe (Fig. 1q). Notably, no Si peaks are detected, which confirms the complete etching of SiNP. The nitrogen adsorption-desorption isotherm of HNDIS exhibits a typical type-IV profile, a characteristic of mesoporous structure (Fig. 1r). From the pore size distribution (Fig. 1s), the average pore size of HNDIS is measured to be 2.1 nm. The specific dimensions (2.1 nm pore size and ∼4 nm AuNP) are rationally designed to create a size-selective barrier with three critical functions. Specifically, the 2.1 nm mesopores are larger than glucose (∼0.8 nm) to allow efficient fuel influx. Meanwhile, these pores are smaller than AuNP (∼4 nm) to physically confine the catalyst within the interior surface. Crucially, the narrow pore size effectively prevents the intrusion of large cytosolic enzymes, such as CAT and ferritin (hydrodynamic diameter >10 nm), thereby shielding the internal reaction from enzymatic interference. The XRD spectrum of HMION exhibits characteristic diffraction peaks of Fe₃O₄ at (220), (311), (400), (422), (511) and (440) (Fig. 1t), confirming the cubic spinel phase of Fe₃O₄ (JCPDS No. 19–0629) [19]. Meanwhile, the XRD pattern of HNDIS reveals distinct peaks of metallic gold at (111), (200), (220) and (311), which are consistent with the face-centered cubic structure of Au (JCPDS No. 04–0784) [20,21]. The saturation magnetization (Ms) of HNDIS is approximately 80 emu/g (Fig. S5), indicating that the nanofactory possesses high r₂ relaxivity and strong potential for T₂-weighted imaging contrast. Additionally, HNDIS exhibits rapid magnetic capture capability (Fig. S6), confirming the strong magnetism of the nanofactory. Because ∼11 nm of HMION shell exceeds the XPS detection depth, the XPS spectrum of HNDIS shows distinct O and Fe peaks, but lacks Au signals (Fig. 1u), which confirms that the majority of AuNP are encapsulated on the interior surface of HMION [22]. The Fe 2p XPS peak of HNDIS is deconvolved to Fe³⁺ (711.9 eV, 725.0 eV) and Fe²⁺ (709.9 eV, 723.0 eV) with a molar ratio of ∼2, which is consistent with the valence characteristics of Fe₃O₄ (Fig. 1v). In addition, AuNP exhibit a distinct UV–Vis absorbance peak at 520 nm, whereas HNDIS shows significantly reduced absorption at this wavelength (Fig. 1w). That's because AuNP are primarily encapsulated on the interior surface of HMION, thereby suppressing UV–Vis absorbance of AuNP [23].

3.2. •OH generation by the nanofactory

Fig. 2a illustrates the •OH generation mechanism of the nanofactory. (I) Glucose molecules diffuse into the hollow core of HNDIS through concentration gradient-driven diffusion, where they can be catalyzed into H₂O₂ and GA by the ultrasmall AuNP with GOx-mimicking activity on the interior surface [24,25]. (Ⅱ) The GA reacts with HMION, inducing in situ release of Fe²⁺ and Fe³⁺. (Ⅲ, Ⅳ) Fe²⁺ and Fe³⁺ react with H₂O₂ through the Fenton reaction to produce •OH. Fig. 2b shows the pH value changes of the glucose solution after incubation with AuNP, HMION or HNDIS. The pH of glucose solution decreases significantly only for AuNP co-incubation, but not for HMION (without AuNP) and HNDIS. That's because the generated GA catalyzed by the AuNP on the interior surface of HNDIS (the above-mentioned reaction I) is consumed by Fe₃O₄ of the HNDIS (the above-mentioned reaction II). To mimic the conditions of the TME, HNDIS is incubated in a simulated buffer solution (pH 6.8) supplemented with 10 mM glucose (Fig. S7). Specifically, at 24 h, the HNDIS nanoparticles show slight surface roughness but maintain a relatively intact spherical morphology. As the incubation extended to 48 h, visible signs of shell surface corrosion are observed. By 96 h, significant breakage and collapse of the hollow shell structure are detected. This degradation behavior of HNDIS in the TME is attributed to a synergistic effect, i.e., the mild acidity of the TME (pH 6.8) and the gluconic acid generated by AuNP jointly accelerate the dissolution and breakdown of the HNDIS shell. The cumulative release of Fe ions reaches 73.5 % at 96 h (Fig. S8a), confirming that the HMION shell can be effectively degraded within the TME to supply a sufficient amount of Fe ions for the Fenton reaction. In contrast, the release of Au nanoparticles remains relatively low, reaching only 13.6 % at 96 h (Fig. S8b). This stability is attributed to the strong chelation between the carboxyl groups on the surface of AuNP and the HNDIS framework, which ensures that the AuNP are tightly anchored within the nanofactory to continuously catalyze glucose oxidation. The concentration of generated H₂O₂ (measured by amplex red assay) in glucose solution increases significantly after incubation with AuNP (Fig. 2c), highlighting the strong GOx-mimicking ability of AuNP. In contrast, the H₂O₂ concentration in glucose solution decreases notably after incubation with HNDIS (Fig. 2c) because the generated H₂O₂ is consumed by Fe²⁺/Fe³⁺ released from HNDIS to produce •OH (the above-mentioned reaction III/IV).

Fig. 2.

(a): Chemical reaction for the catalysis of glucose to GA and H₂O₂ by AuNP (Ⅰ), the reaction between Fe₃O₄ and GA to generate Fe^2+/3+ (Ⅱ), and the Fenton reaction between H₂O₂ and Fe^2+/3+ (III, IⅤ). (b): The pH value of HMION, AuNP, or HNDIS dispersion after incubation with glucose for 0–72 h. The pH drop by AuNP indicates GA generation, while HNDIS maintains stable pH due to H⁺ consumption by the Fe₃O₄ shell. (c): The resorufin fluorescence emission spectra of glucose solutions after incubation without or with AuNP or HNDIS. (d): The fluorescence emission spectra of HPF solutions after incubation with HNDIS and/or glucose. (e): Schematic illustration of HNDIS-mediated •OH generation by using TMB and OPD assays. (f): UV–vis spectra of TMB in cell lysate with glucose after incubation with AuNP, AuNP + HMION, or HNDIS. (g, h): UV–vis spectra of TMB in cell lysate with glucose after incubation with HNDIS at different concentrations for 72 h (g), or at 5.0 mg/mL for different incubation times (h). (i): UV–vis spectra of OPD in cell lysate with glucose after incubation with AuNP, AuNP + HMION, or HNDIS. (j, k) UV–vis spectra of OPD in cell lysate with glucose after incubation with HNDIS at different concentrations for 72 h (j), or at 5.0 mg/mL for different incubation times (k).

As expected, the glucose solution exhibits a distinct fluorescence emission at 525 nm after incubation with HNDIS and the •OH indicator hydroxyphenyl fluorescein (HPF), highlighting the •OH generation ability of HNDIS (Fig. 2d). Furthermore, the •OH generation ability of HNDIS was assessed by a UV–Vis colorimetric method. The colorless 3,3′,5,5′-tetramethylbenzidine (TMB) or O-phenylenediamine (OPD) can be oxidized to oxidized TMB (oxTMB) or 2,3-diaminophenazine (DAP) [26,27], which shows UV–Vis absorption with a peak at 660 nm or 455 nm (Fig. 2e). In addition, HMION, which possesses a similar hollow mesoporous structure of HNDIS but lacks the incorporation of AuNP (Fig. S9), was synthesized to serve as a control group for assessing the •OH generation efficiency of HNDIS. The cell-lysis buffer containing glucose, ferritin and CAT display a higher oxTMB absorbance at 660 nm after incubation with HNDIS than after co-incubation with a mixture of AuNP and HMION (Fig. 2f), which demonstrates the more robust •OH generation ability of HNDIS than the mixture of AuNP and HMION. That's because HNDIS has mesoporous channels (i.e., “window” of the nanofactory) and secluded nanocompartment with dotted interior surface of AuNP (i.e., “machine” of the nanofactory). The unique “machine” of the nanofactory “operates” the reactions I-III as shown in Scheme 1b. The unique “window” of the nanofactory allows pass for small molecules (e.g., glucose) to initiate the cascade reactions I-III, but not for macromolecules (e.g., CAT and ferritin), which prevent the undesirable catalysis of Fe²⁺ to Fe³⁺ by ferritin and H₂O₂ to H₂O by CAT. Therefore, our nanofactory of HNDIS can preserve the Fenton reagents and sustain robust •OH generation.

Furthermore, HNDIS dispersion exhibits a higher oxTMB absorbance peak at 5.0 mg/mL compared to 1.0 mg/mL after incubation with glucose and TMB, indicating a concentration-dependent •OH generation capacity by the nanofactory (Fig. 2g). Additionally, the HNDIS dispersion exhibits the highest oxTMB absorbance peak after 72 h of incubation with glucose, and TMB, compared to other shorter incubation time (0–60 h) (Fig. 2h), demonstrating that the •OH generation ability of the nanofactory is time-dependent. The glucose influx rate is theoretically estimated based on the Fick's laws of diffusion. The average pore size of HNDIS is 2.1 nm, which is significantly larger than the hydrodynamic diameter of a glucose molecule (∼0.8 nm), allowing for effective diffusion. The characteristic diffusion time (t_diff) across the HMION shell using the equation t_diff ≈ L²/D_eff. Given a shell thickness (L) of approximately 11 nm, the diffusion time is estimated to be approximately 1.1 × 10⁻⁵ s. However, the glucose oxidation catalyzed by AuNP (functioning as GOx mimics) proceeds on a timescale of seconds to minutes, as evidenced by the slow pH decrease over hours in Fig. 2b. In addition, the Fenton reaction typically operates on a timescale of seconds (k ≈ 40-80 M⁻¹ s⁻¹). Consequently, the glucose influx rate (microseconds) is orders of magnitude faster than the catalytic reaction rate (minutes) and Fenton reaction rate (seconds), indicating that the glucose diffusion is not the rate-limiting step for the •OH generation.

Consistent with the TMB assay results, the HNDIS dispersion shows an obvious DAP absorbance peak at 455 nm after incubation with glucose and OPD (Fig. 2i), which is higher than the mixture of AuNP and HMION. Furthermore, HNDIS dispersion demonstrates a more pronounced OPD absorbance peak after incubation at a high (5.0 mg/mL) than a low (1.0 mg/mL) concentration (Fig. 2j). The HNDIS dispersion also exhibits the highest OPD absorbance peak following 72 h of treatment with glucose (Fig. 2k).

3.3. Cellular internalization and lysosome escape behavior of the nanofactory

As observed by confocal laser scanning microscopy (CLSM), the cytoplasm of MCF-7/MDR cells exhibits a distinct R6G fluorescence, after incubation with R6G@HMION (Fig. 3a) or R6G@HNDIS (Fig. 3b), which indicates that MCF-7/MDR cells possess a robust capability for nanoparticle uptake. Moreover, MCF-7/MDR cells display the highest mean fluorescence intensity (MFI) of R6G at 8.0 h after incubation with R6G@HMION (Fig. 3c and d) or R6G@HNDIS (Fig. 3e and f) compared to short incubation times (0, 2.0, or 4.0 h), revealing a time-dependent nanoparticle uptake ability of MCF-7/MDR cells. CLSM images of MCF-7/MDR cells treated with R6G@HNDIS and stained with Lysotracker and 4′,6-Diamidino-2-Phenylindole (DAPI) (Fig. S10) shows extensive non-overlap of red and green fluorescence within the cytoplasm, indicating nanoparticle escape from lysosomes [28]. This is mainly because that Fe²⁺ released from HNDIS in lysosomes undergoes Fenton reaction with H₂O₂, which generates •OH and disrupts the lysosomal membrane. To investigate the long-term structural integrity of HNDIS under intracellular conditions, the nanoparticles are extracted from MCF-7/MDR cells after varying incubation periods (0, 24, 48, and 72 h) for TEM characterization (Fig. S11). HNDIS extracted from the cytosol maintains a relatively intact morphology at 24 h. However, visible signs of shell corrosion are detected at 48 h, and extensive structural collapse of the HMION shell is observed at 72 h. This time-dependent degradation is primarily attributed to the accumulation of GA produced by the encapsulated AuNP within the nanofactory. Overall, these results show that our HNDIS can be taken up by cancer cells and successfully escape from lysosomes to sustainably produce •OH [29,30].

Fig. 3.

(a, b): CLSM images of MCF-7/MDR cells after treatment with R6G@HMION (a) or R6G@HNDIS (b) for 4.0 h. DAPI for nuclei; FITC-Phalloidin for cytoskeleton; R6G for nanoparticles. (c–f): R6G fluorescence distributions (c, e) and the corresponding quantitative analysis (d, f) of MCF-7/MDR cells after treatment with R6G@HMION (c, d) or R6G@HNDIS (e, f) for 0, 2.0, 4.0 or 8.0 h. (g, h):The cell viability of MCF-7/MDR cells after treatment with PBS, AuNP, HMION, AuNP + HMION, or HNDIS at varying concentrations for 24 h (g), or at 50 μg/mL of Fe for different incubation times (6.0, 12, or 24 h) (h). (i): The cell viability of MCF-7/MDR cells after treatment with PBS, HNDIS, HNDIS + DFO, HNDIS + Fer-1, or HNDIS + NAC for 24 h. (j, k): CLSM images of MCF-7/MDR cells after treatment with PBS, AuNP, HMION, AuNP + HMION or HNDIS for 24 h, and staining with Calcium-AM and PI (j), or staining with EdU and DAPI (k). (l): The camera photos of MCF-7/MDR cells after treatment with PBS, AuNP, HMION, AuNP + HMION, or HNDIS for 144 h, and staining with crystal violet.

3.4. In vitro cytotoxicity of the nanofactory

The half maximal inhibitory concentration (IC₅₀) value of DOX against MCF-7 or MCF-7/MDR cells is calculated to be 1.0 or 12.4 μM, demonstrating the potent drug resistance ability of MCF-7/MDR cells (Fig. S12). MCF-7/MDR cells exhibit relatively high cell viability of 87.8 ± 5.7 or 77.5 ± 6.2 % after treatment with AuNP (C_Au = 9.6 μg/mL) or HMION (C_Fe = 50 μg/mL) (Fig. 3g), which indicates that AuNP or HMION alone have low toxicity to tumor cells. However, the viability of MCF-7/MDR cells after treatment with HNDIS (C_Fe = 50 μg/mL) was only 23.5 ± 5.7 %, significantly lower than 59.6 ± 6.3 % of cell viability after treatment with the mixture of HMION and AuNP, demonstrating the potent tumor cell killing ability of HNDIS against drug resistance tumor cells. The enhanced cytotoxicity of HNDIS can be attributed to its unique structure with mesoporous channels and dotted interior surface, which provides a relatively closed space that ensures the efficient production of •OH via the Fenton reaction. Additionally, MCF-7/MDR cells show cell viabilities of 67.5 ± 6.8, 46.5 ± 6.4 and 23.5 ± 5.7 % after treatment with HNDIS at 6.0, 12 and 24 h, suggesting the time-dependent cytotoxicity of the nanofactory (Fig. 3h). The viability of 21.4 ± 7.7 % for MCF-7/MDR cells after HNDIS treatment can be significantly rescued to 37.9 ± 7.8, 40.9 ± 7.7 and 37.9 ± 5.1 % by the introduction of ferroptosis antagonists Fer-1, DFO and NAC (Fig. 3i). This finding demonstrates that ferroptosis is the main pathway of cell death induced by HNDIS in MCF-7/MDR cells.

Fig. 3j and Fig. S13, 14 show CLSM images of MCF-7/MDR cells after staining with Calcein-AM and PI to visualize live/dead cells [31]. MCF-7/MDR cells display an abundance of live cells with green fluorescence for the treatment of PBS, AuNP, or HMION, but many dead cells with red fluorescence for the mixture of AuNP + HMION. However, MCF-7/MDR cells after treatment with HNDIS display more red fluorescence than the mixture of AuNP + HMION, which confirms the highest cytotoxicity of our HNDIS. The apoptosis status of tumor cells after different treatments was evaluated by flow cytometry using Annexin V-FITC/PI double staining as suggested (Fig. S15a). Quantitative analysis indicates that the total apoptosis ratios (early and late apoptosis) for the PBS, AuNP, and HMION groups are 0.11 %, 14.84 %, and 25.96 % (Fig. S15b), showing limited cytotoxicity. In contrast, the HNDIS group shows a remarkable apoptosis rate of 88.76 %, which is significantly higher than that of the mixture of AuNP and HMION (46.21 %), further confirming that HNDIS effectively triggers programmed cell death in drug-resistant tumor cells. Mechanistically, this pronounced apoptotic effect induced by HNDIS is attributed to the elevated intracellular ROS levels, which cause severe mitochondrial damage and subsequently activate the apoptosis pathway.

Fig. 3k illustrates the proliferative capacity of MCF-7/MDR cells after treatment with various nanoparticles, and co-staining with 5-Ethynyl-2′-deoxyuridine (EdU) and DAPI. Notably, MCF-7/MDR cells exhibit a large number of green fluorescence within nuclei after treatment with PBS, AuNP, or HMION, showing an active proliferation state. In contrast, MCF-7/MDR cells exhibit reduced or no green fluorescent-labeled nuclei after treatment with the mixture of AuNP + HMION, or HNDIS, which indicates that the proliferation activity of tumor cells is significantly or thoroughly suppressed by the mixture of AuNP + HMION, or our HNDIS.

Fig. 3l presents photos of MCF-7/MDR cells after various treatments and staining with crystal violet to assess the single-cell proliferation of tumor cells. MCF-7/MDR cells exhibit much fewer colonies after treatment with HNDIS compared to AuNP, HMION, or the mixture of AuNP and HMION, which reinforces the high cytotoxicity of HNDIS toward MCF-7/MDR cells. The surviving fraction of MCF-7/MDR cells in group I-V is calculated to be 1.0, 0.42, 0.38, 0.14, or 0.04 (Fig. S16), further confirming the superior toxicity of HNDIS in suppressing tumor cell proliferation compared to the other groups. The observed slight difference between the cell viability data (Fig. 3g and h) and the colony formation results (Fig. 3l, Fig. S16) is attributable to the synergistic impact of treatment duration and initial cell density. Notably, the colony formation assay spans 144 h, which is significantly longer than the short-term (24 h) MTT assay. Given that AuNP act as GOx mimics, this prolonged exposure time leads to a continuous and cumulative buildup of H₂O₂. Moreover, unlike the high-density populations in the MTT assay that benefit from protective cell-cell interactions, the colony formation assay starts from isolated single cells. All these findings jointly demonstrate that our HNDIS is more effective to kill MCF-7/MDR cells than the mixture of AuNP and HMION without structural rearrangement.

3.5. In vitro ROS production by the nanofactory

Fig. 4a presents CLSM images of MCF-7/MDR cells following treatment with various nanoparticles and staining with BBoxiProbeTMA (H₂O₂ probe). MCF-7/MDR cells display prominent red fluorescence upon AuNP treatment compared with the control group, indicating that AuNP retain their GOx-mimicking activity within the cytoplasm to produce H₂O₂. Notably, MCF-7/MDR cells exhibit stronger intense red fluorescence after treatment with HNDIS compared with AuNP alone. This is attributed to the secluded nanocompartment of HNDIS with dotted interior surface, which protects the H₂O₂ generated by AuNP from being broken down by CAT. Quantitative H₂O₂ assay kit data further reveal that the intracellular H₂O₂ concentration in MCF-7/MDR cells is much higher after treatment with HNDIS than the AuNP alone (Fig. 4b). Fig. 4c presents CLSM images of MCF-7/MDR cells after various treatments and staining with BCECF-AM (H⁺ probe). MCF-7/MDR cells display weak green fluorescence upon AuNP treatment compared with the control group, which suggests that GA is generated by the catalysis of AuNP. Furthermore, compared to the treatment with AuNP alone, the green fluorescence within MCF-7/MDR cells is significantly increased after treatment with the mixture of AuNP and HMION, and HNDIS. That's because the generated GA is consumed by Fe₃O₄ of HMION and HNDIS. Flow cytometry (FCM) data reveal that the average percentage of BCECF-AM positive-staining cells after treatment with the group Ⅰ-Ⅳ are respectively 50, 3.4, 55.5, 24.6 and 26.4 % (Fig. 4d), which reinforces that GA can be produced by AuNP and consumed by Fe₃O₄ of HMION and HNDIS.

Fig. 4.

(a): CLSM images of MCF-7/MDR cells after treatment with PBS, AuNP, HMION, AuNP + HMION, or HNDIS for 24 h, followed by staining with BBoxiProbeTMA (H₂O₂ probe) and DAPI. (b): The intracellular H₂O₂ concentrations in MCF-7/MDR cells after exposure to various treatments for 24 h, quantified using an H₂O₂ assay kit. (c, d): CLSM images (c) of MCF-7/MDR cells after treatment with PBS, AuNP, HMION, AuNP + HMION, or HNDIS for 24 h, followed by staining with BCECF-AM (H⁺ probe), and the corresponding flow cytometry fluorescence distributions (d) of MCF-7/MDR cells. (e–g): CLSM images (e) of MCF-7/MDR cells after treatment with PBS, AuNP, HMION, AuNP + HMION, or HNDIS for 24 h, followed by staining with RhoNox-1 (Fe²⁺ probe), and the corresponding flow cytometry fluorescence distributions (f) and quantitative analysis (g) of MCF-7/MDR cells. (h–j): CLSM images (h) of MCF-7/MDR cells after treatment with PBS, AuNP, HMION, AuNP + HMION, or HNDIS for 24 h, followed by staining with HPF (•OH probe), and the corresponding fluorescence distributions (i) and statistical analysis (j) of MCF-7/MDR cells.

Fig. 4e presents CLSM images of MCF-7/MDR cells after treatment with the group Ⅰ-Ⅳ and staining with FeRhoNox-1 (Fe²⁺ probe). MCF-7/MDR cells treated with HNDIS exhibit stronger red fluorescence than the mixture of AuNP and HMION because the “window” of the nanofactory prevents ferritin from entering the nanofactory and oxidizing Fe²⁺ to Fe³⁺. Additionally, the fluorescence distributions (Fig. 4f) and mean fluorescence intensity (MFI, Fig. 4g) of FeRhoNox-1 measured by flow cytometry confirm that MCF-7/MDR cells treated with HNDIS exhibit the strongest Fe²⁺ fluorescence compared to other treatments. Fig. 4h presents CLSM images of MCF-7/MDR cells after treatment with the group Ⅰ-Ⅳ and staining with hydroxyphenyl fluorescein (HPF, •OH probe). MCF-7/MDR cells treated with HNDIS display stronger green •OH fluorescence than the mixture of AuNP and HMION. Moreover, the fluorescence distributions (Fig. 4i) and MFI (Fig. 4j) of HPF further confirm that HNDIS can trigger the most intense •OH generation compared to other treatments. Similarly, Fig. S17 presents CLSM images of MCF-7/MDR cells after treatment with the group Ⅰ-Ⅳ and staining with DCFH-DA (ROS probe). The cells show green ROS fluorescence after treatment with the group Ⅱ-Ⅴ, which indicates that ROS are appreciably generated by AuNP and HMION. Notably, MCF-7/MDR cells treated with HNDIS display strongest green ROS fluorescence measured by flow cytometry than the group I-IV (Fig. S18 and 19). Collectively, these results demonstrate that AuNP within the secluded nanocompartment of HNDIS can catalyze the conversion of glucose to H₂O₂ and GA (the reaction I in Scheme 1b), and the generated GA can react with Fe₃O₄ of HMION and HNDIS to produce Fe²⁺ (the reaction II in Scheme 1b), which participates in the Fenton reaction with H₂O₂ to generate •OH (the reaction III in Scheme 1b). The therapeutic efficacy of HNDIS relies on the local availability of glucose and oxygen, which serve as the catalytic reaction substrate of AuNP. Although the tumor microenvironment is often hypoxic, the cytoplasm retains dissolved oxygen essential for the GOx-mimicking catalytic reaction. Importantly, the tumor cells exhibit significantly upregulated glucose uptake due to the Warburg effect, ensuring a continuous and abundant influx of glucose into the cytoplasm. Therefore, the high intracellular glucose concentration plays a critical compensatory role against hypoxia, driving the reaction kinetics to maximize H₂O₂ production.

3.6. The damage of mitochondria by the nanofactory

Fig. 5a presents CLSM images of MCF-7/MDR cells after various treatments and staining with JC-1 (a mitochondrial membrane potential probe) [32]. MCF-7/MDR cells treated with PBS display bright red JC-1 aggregates in the mitochondria, while those treated with AuNP, HMION, or the mixture of AuNP + HMION show a notable dispersion of green JC-1 monomers in the cytoplasm. The cells treated with HNDIS exhibit the strongest green fluorescence, indicating a significantly lower mitochondrial membrane potential (MMP) [33]. Furthermore, the proportion of JC-1 monomer in MCF-7/MDR cells is measured to be 0, 13.5, 19.4, 40.8 and 100 % after treatment with group Ⅰ-Ⅴ (Fig. 5b). Collectively, these results indicate that HNDIS with dotted interior surface exhibits a significantly stronger capability to damage mitochondria compared to the mixture of AuNP and HMION.

Fig. 5.

(a, b): CLSM images (a), and the corresponding fluorescence distributions (b) of MCF-7/MDR cells after treatment with PBS, AuNP, HMION, AuNP + HMION, or HNDIS for 24 h, and staining with JC-1 (mitochondrial membrane potential probe). (c–h): 2D (c, e, g) and corresponding 3D (d. f, h) CLSM images of MCF-7/MDR cells after treatment with PBS (c, d), AuNP + HMION (e, f) or HNDIS (g, h) for 24 h, and staining with MitoTracker (mitochondria probe). (i–k): TEM images of MCF-7/MDR cells after treatment with PBS (i), AuNP + HMION (j) or HNDIS (k) for 24 h.

Fig. 5c–h presents CLSM images of MCF-7/MDR cells after treatment with PBS (Fig. 5c and d), the mixture of AuNP + HMION (Fig. 5e and f), or HNDIS (Fig. 5g and h), and staining with MitoTracker (a mitochondrial probe) to visualize mitochondrial number and state. MCF-7/MDR cells display intense red fluorescence following incubation with PBS or the mixture of AuNP + HMION, corresponding to high mitochondrial number. In contrast, MCF-7/MDR cells display markedly reduced red fluorescence following incubation with HNDIS, which indicates that HNDIS can effectively reduce the mitochondrial number.

As observed from the TEM images of MCF-7/MDR cells, the mitochondria cristae structures of MCF-7/MDR cells are almost intact after treatment with PBS (Fig. 5i), slightly damaged after treatment with the mixture of AuNP and HMION (Fig. 5j), and severely damaged after treatment with HNDIS (Fig. 5k).

As expected, the relative ATP levels of MCF-7/MDR cells treated with the group I-V are respectively measured to be 100 ± 11, 85.4 ± 6.4, 84.9 ± 7.4, 56.0 ± 4.8 and 36.2 ± 4.8 % using an ATP kit (Fig. S20), which indicates that HNDIS has the strongest ability to damage mitochondria and down-regulate ATP production [34,35]. Furthermore, Fig. S21 presents CLSM images of MCF-7/MDR cells after treatment with the group I-V, followed by staining with a fluorescence-labeled γ-H2AX secondary antibody, a nuclear DNA damage probe [36]. MCF-7/MDR cells treated with HNDIS show the most intense green fluorescence in the nucleus compared to other groups, which suggests that the •OH generated by HNDIS causes severe nuclear DNA damage.

Collectively, the above results reinforce that HNDIS with dotted interior surface has the strongest ability to damage mitochondria and nuclei.

3.7. The ferroptosis efficacy induced by the nanofactory

Fig. 6a and Fig. S22 show CLSM images of MCF-7/MDR cells after treatment with the group Ⅰ-Ⅴ, and staining of C11-BODIPY^581/591 (a LPO probe). MCF-7/MDR cells treated with PBS, AuNP, or HMION display bright red fluorescence, indicating low LPO accumulation. However, MCF-7/MDR cells treated with HNDIS exhibit the most intense green fluorescence, which is markedly higher than treatment with the mixture of AuNP and HMION. Furthermore, the fluorescence distributions (Fig. 6b) and MFI (Fig. 6c) of C11-BODIPY^581/591 confirm that MCF-7/MDR cells treated with HNDIS have the strongest green fluorescence among all groups. This result is attributed to the fact that HNDIS can produce more •OH than the mixture of AuNP and HMION, thereby generating more LPO within tumor cells [29].

Fig. 6.

(a–c): CLSM images of MCF-7/MDR cells after treatment with PBS, AuNP, HMION, AuNP + HMION, or HNDIS for 24 h, followed by staining with C11-BODIPY^581/591 (LPO probe), and the corresponding fluorescence distributions (b) and quantitative analysis (c). (d–f): CLSM images of MCF-7/MDR cells after treatment with PBS, AuNP, HMION, AuNP + HMION, or HNDIS for 24 h, followed by staining with DIO (cell membrane probe), and the corresponding flow cytometry fluorescence distributions (e) and quantitative analysis (f). (g): CLSM images of MCF-7/MDR cells after treatment with PBS, AuNP, HMION, AuNP + HMION, or HNDIS for 24 h, followed by staining with TRITC- phalloidin (F-actin probe) and DAPI.

In addition, the intracellular concentration of malondialdehyde (MDA), a byproduct of LPO, is quantified as 8.3 ± 2.3, 23.9 ± 4.8, 30.0 ± 3.0, 49.6 ± 5.4, or 86.2 ± 10.9 nmol per 10⁴ cells after treatments with the group Ⅰ-Ⅴ (Fig. S23), highlighting the potent LPO generating capacity of the nanofactory [37].

Fig. 6d presents CLSM images of MCF-7/MDR cells after treatment with the group Ⅰ-Ⅴ, followed by staining with DIO (a cytomembrane probe). MCF-7/MDR cells in the group I-IV display evident green fluorescence surrounding the cell nucleus, indicating an intact cell membrane structure [38]. However, the green fluorescence in MCF-7/MDR cells is significantly reduced after treatment with HNDIS, which indicates that HNDIS can cause the most severe cell membrane damage. Moreover, the DIO fluorescence distribution (Fig. 6e) and corresponding MFI (Fig. 6f) reinforce that HNDIS induces the most significant cell membrane damage among the group Ⅰ-V.

Fig. 6g presents CLSM images of MCF-7/MDR cells treated with the group Ⅰ-Ⅴ and stained with TRITC-Phalloidin (an F-actin probe) to intuitively reveal cellular morphology. The F-actin in MCF-7/MDR cells maintains a uniformly distributed three-dimensional network architecture characterized by densely packed filaments after treatment with PBS, AuNP, or HMION. However, the F-actin filaments show slight disassembly after treatment with the mixture of AuNP and HMION. Notably, MCF-7/MDR cells exhibit significantly morphological changes, including cellular shrinkage and F-actin aggregation after treatment with HNDIS. To explore the cell death mechanisms of MCF-7/MDR cells at different exposures times, the expression of key ferroptosis and apoptosis markers was evaluated after treatment with HNDIS (C_Fe = 50 μg/mL) for 0, 24, or 48 h (Fig. S24a). Specifically, the GPX4 expression is significantly downregulated (up to ∼ 57 %) by HNDIS treatment at 24 h (Fig. S24b). This is because the GSH, as an essential cofactor of GPX4, is consumed by the Fe³⁺ released from the nanofactory via a redox reaction (Fe³⁺ + GSH → Fe²⁺ + GSSG). Simultaneously, SLC7A11 expression is suppressed up to ∼76 % at 48 h (Fig. S24c). This is likely because the elevated ROS activate the transcription factor p53, which in turn inhibits SLC7A11 transcription to promote ferroptosis [39]. In contrast, the pro-ferroptotic enzyme ACSL4 is upregulated up to ∼144 % after 48 h of incubation (Fig. S24d). This upregulation is induced by the elevated ROS, which activate stress-responsive transcription factors to promote ACSL4 transcription [40]. Notably, the cleaved caspase-3 expression is markedly enhanced by HNDIS as the incubation time increased (Fig. S24e). This is due to the overproduction of ROS generated by HNDIS causing mitochondrial damage and subsequent cytochrome c release, thereby activating caspase-3 [41]. However, combined with the inhibitor rescue experiments in Fig. 3i where ferroptosis inhibitors (Fer-1 and DFO) significantly restored cell viability, these results collectively demonstrate that ferroptosis is the dominant pathway, while apoptosis serves as a secondary outcome of ROS-induced damage. Collectively, the above results demonstrate that the nanofactory can effectively induce accumulation of LPO within cells, and then trigger robust ferroptosis-based damage for tumor cells [42,43].

3.8. MRI performance and in vivo biodistribution of the nanofactory

HMION, with high saturation magnetization, serves as a T₂-MRI contrast agent to track the accumulation of the nanofactory in tumors [44]. As the concentration of HNDIS solutions increases, the T₂-MRI signal progressively shifts from bright to dark (Fig. 7a). In contrast, the T₁-MRI signal remains dark even at high concentrations of HNDIS solutions (Fig. 7b). In addition, the transverse relaxivity (r₂) (Fig. 7c) and longitudinal relaxivity (r₁) (Fig. 7d) of HNDIS are quantified as 241.7 and 0.31 mM⁻¹ s⁻¹, which confirms that our HNDIS itself can serve as an excellent T₂-MRI contrast agent for monitoring its in vivo metabolism.

Fig. 7.

(a–d): T₂-weighted (a) or T₁-weighted (b) MR images of HNDIS dispersion with various Fe concentration (0–100 mM), and the corresponding T₂ relaxation rate (1/T₂) (c) or T₁ relaxation rate (1/T₁) (d) plotted as a function of C_Fe under a 7.0 T MRI scanner. (e): The percent injected dose per gram of tissue (%I.D./g) of Au in blood at different times after i.v. injection of AuNP or HNDIS (Fe dosage = 10 mg/kg) into mice. (f, g): The %I.D./g of Au (f) or Fe (g) in major organs and tumors at different time points after i.v. injection of HNDIS (Fe dosage = 10 mg/kg) into mice. (h): Schematic illustration for tracing the nanofactory accumulation in tumors using an MRI scanner. (i, j): Axial T₂-weighted MR images (i) for MCF-7/MDR tumor-bearing mice pre-injection and post-injection of HNDIS (Fe dosage = 10 mg/kg) acquired on a 7.0 T MRI scanner at 12, 24, 48, or 72 h, and the corresponding relative signal intensity (j).

The blood circulation half-life (t_1/2) of AuNP and HNDIS in the bloodstream after intravenously injection is measured to be 2.3 h and 8.3 h, respectively (Fig. 7e). Compared to AuNP, the prolonged t_1/2 of HNDIS can enhance its accumulation in tumors. The extended blood circulation half-life of HNDIS (t_1/2 = 8.3 h) can be attributed to the synergistic effect of its specific physicochemical properties. Primarily, the hydrodynamic diameter of HNDIS is approximately 225.7 nm, which significantly exceeds the renal filtration cut-off, thereby effectively preventing rapid clearance by the kidneys. Furthermore, HNDIS exhibits a negative surface charge (−28.2 mV), which generally reduces the adsorption of plasma proteins compared to cationic nanoparticles, facilitating the evasion of rapid recognition and phagocytosis by the reticuloendothelial system (RES). Furthermore, the in vivo biodistribution of the nanofactory was evaluated by measuring the percent injected Au (Fig. 7f) and Fe (Fig. 7g) per gram of tumors and major organs (heart, liver, spleen, lungs, and kidneys). After injection of HNDIS, the content of Au or Fe in tumors reaches maximum at 24 h followed by a slight decline, which indicates the effective accumulation of HNDIS in tumors. Based on the biodistribution profiles in Fig. 7f and g, HNDIS exhibits primary enrichment in the liver and spleen with a peak accumulation at 24 h, followed by a consistent decline attributed to a coordinated hepatobiliary and renal excretion mechanism. Specifically, HNDIS possesses a hydrodynamic diameter of approximately 225.7 nm, resulting in its initial capture by the RES. Following biodegradation within the RES (e.g., by Kupffer cells), the HMION shell can be degraded into biocompatible iron ions that enter the host's metabolic pool. Concurrently, the released ultrasmall AuNP (∼4 nm), possessing a particle size below the renal filtration threshold (6–8 nm), can be efficiently eliminated from the body via renal clearance. This dual clearance mechanism significantly minimizes the risk of long-term retention and potential toxicity [45].

The accumulation of HNDIS in tumors can also be tracked using a 3.0 T MRI scanner owing to its strong T₂-weighted MRI signal (Fig. 7h). The tumor region exhibits a progressive decrease in brightness after injection, and reaches the minimum brightness at 24 h (Fig. 7i), which indicates the maximal accumulation of the nanofactory. Consistently, the average relative brightness signal intensity in tumors at 12, 24, 48, and 72 h after injection of HNDIS is measured to be 48.7, 31.7, 45.4 and 59.4 % (Fig. 7j), which reinforces the high accumulation of our HNDIS in tumors.

3.9. In vivo antitumor efficacy of the nanofactory

The ferroptosis therapy protocol conducted in 4T1 tumor-bearing mice is schematically outlined in Fig. 8a. This protocol involves the intravenous administration of saline (Ⅰ), HMION (Ⅱ), HMION + AuNP (Ⅲ), or HNDIS (Ⅳ). The tumors exhibit rapid growth rate following the injection of saline (Fig. 8b and c), HMION (Fig. 8d), or the mixture of AuNP + HMION (Fig. 8e), but display markedly reduced growth rate after HNDIS treatment (Fig. 8f). Furthermore, the average tumor inhibition rate is calculated to be 0, 22.3, 63.1 or 87.5 % for the group Ⅰ-Ⅳ at day 16 (Fig. 8g). The average tumor mass in the group Ⅰ-Ⅳ is respectively 1.97, 1.51, 0.78 and 0.21 g at day 16, confirming the excellent anti-tumor efficacy of the nanofactory (Fig. 8h). The mice in the group Ⅰ-Ⅲ were first sacrificed at day 16, 18, or 24 (Fig. 8i), whereas those in the group Ⅳ survived beyond day 36. In addition, the weight of mice shows no significant alteration during the treatment period, which suggests that our HNDIS has no long-term toxicity (Fig. 8j).

Fig. 8.

(a): Schematic illustration of the treatment protocol for evaluating the anti-tumor efficacy of HNDIS in MCF-7/MDR tumor-bearing mice after intravenous injection with saline (Ⅰ), HMION (II, Fe dosage = 10 mg/kg), AuNP + HMION (III, Au or Fe dosage = 10 mg/kg), or HNDIS (IV, Fe dosage = 10 mg/kg) on day 0, 7 and 14. (b–f): Relative tumor volume curves (b) and individual tumor volume curves after treatment with saline (c), HMION (d), AuNP + HMION (e), HNDIS (f). (g, h): Tumor inhibition rate (g) and tumor mass (h) of the mice in different treatment groups on day 16. (i, j): Survival ratio (i) and body weight (j) of the mice in different groups during treatment period. (k–o): Representative images of tumor tissues from the five groups after staining with HE (k), Ki67 (l), TUNEL (m), DCFH-DA (n), or C11-BODIPY^581/591 (o) (scale bar: 100 μm).

Moreover, as observed in H&E staining images (Fig. 8k), the tumor cells treated with HNDIS show marked nuclear shrinkage and fragmentation, indicating extensive cellular damage caused by the nanofactory. The tumor cells treated with HNDIS show the lowest Ki67 expression (a proliferation marker) and highest TUNEL level (an apoptosis indicator) (Fig. 8l and m), demonstrating the superior antitumor efficacy of the nanofactory. Tumors in the group IV exhibit the strongest fluorescence of DCFH-DA (Fig. 8n) and C11-BODIPY^581/591 (Fig. 8o), which collectively confirms the robust ROS production and LPO accumulation induced by the nanofactory. Consistently, the relative tumor MDA levels in the group Ⅰ-Ⅳ are respectively 100 ± 8.7, 99.6 ± 6.4, 140.8 ± 11.4 and 251.4 ± 18.7 % (Fig. S25), which confirms the powerful ferroptosis-inducing ability of our HNDIS in tumors.

3.10. In vivo therapeutic effect of the nanofactory on large tumors

Fig. 9a shows the schematic diagram for the therapy protocol on large tumor-bearing mice (>500 mm³) after in situ injection of saline, HMION, the mixture of AuNP + HMION, or HNDIS. Treating large solid tumors is often hindered by severe physiological barriers (e.g., dense extracellular matrix and high interstitial fluid pressure), which can restrict the penetration of chemotherapeutic agents. In addition, an upregulated antioxidant enzyme system (e.g., CAT and ferritin) can resist oxidative stress. Our HNDIS utilizes glucose, a small molecule capable of permeating dense tumor tissues as fuel. The mesopores of HNDIS exclude CAT and ferritin to protect Fenton reagents (H₂O₂/Fe²⁺), thereby sustaining potent LPO generation and ferroptosis in large tumors. The tumors grow rapidly after injection of saline (Fig. 9b and c), HMION (Fig. 9d), or the mixture of AuNP + HMION (Fig. 9e), but grow slowly after injection of HNDIS (Fig. 9f). In addition, tumors cells in the HNDIS group exhibit most pronounced nuclear fragmentation (Fig. 9g), the lowest Ki67 expression (Fig. 9h), the highest TUNEL staining (Fig. 9i), and the maximum LPO accumulation (Fig. 9j). These results collectively demonstrate that HNDIS exhibits superior therapeutic efficacy compared to other groups including the mixture of AuNP and HMION. The superior anti-tumor efficacy can be mainly attributable to the unique architecture of HNDIS, in which AuNP are situated on the interior surface of HMION. The mesopore channels of HMION sterically hinder CAT and ferritin enzymes from catalyzing H₂O₂ to H₂O and Fe²⁺ to Fe³⁺, respectively. Consequently, HNDIS generates markedly higher levels of •OH and LPO within the tumor cells than the non-integrated mixture of AuNP and HMION.

Fig. 9.

(a): Schematic illustration for evaluation of the anti-tumor efficacy of HNDIS on large 4T1 tumor-bearing (over 500 cm³) mice via intravenous administration of saline (Ⅰ), HMION (II, Fe dosage = 10 mg/kg), HMION + AuNP (III, Fe or Au dosage = 10 mg/kg), or HNDIS (IV, Fe dosage = 10 mg/kg) on day 7, 10 and 13. (b–f): Tumor volume curves (b) and individual tumor volume curves after treatments with saline (c), HMION (d), HMION + AuNP (e), HNDIS (f). (g–j): Tumor tissues from the groups I-IV after staining with HE (g), Ki67 (h), TUNEL (i), or C11-BODIPY^581/591 (j). (k): Schematic representation for assessment of the ferroptosis efficacy of HNDIS on drug-resistant mice via intravenous injection of saline (I), DOX (II, drug dosage = 10 mg/kg), CPT (III, drug dosage = 10 mg/kg), or HNDIS (IV, Fe dosage = 10 mg/kg) on day 0, 7 and 14. (l–p): Relative tumor volume curves (l) and individual tumor volume curves after treatments with saline (m), DOX (n), CPT (o), HNDIS (p). (q–t): Tumor tissues from the groups I-IV after staining with HE (q), Ki67 (r), TUNEL (s), or C11-BODIPY^581/591 (t). The scale bar is 100 μm.

3.11. Therapeutic efficacy of the nanofactory on drug-resistant tumors

The multidrug-resistant MCF-7/MDR cell line, which is resistant to clinically approved chemotherapeutics doxorubicin (DOX) and camptothecin (CPT) [46], was utilized to assess the therapeutic potential of our HNDIS. The relatively high dosage of 10 mg/kg for DOX or CPT was selected to rigorously confirm the robust drug resistance of the MCF-7/MDR tumors, guided by established protocols in literature [47,48]. Fig. 9k present the therapeutic schedule for evaluating the anti-tumor efficacy of the nanofactory in nude mice bearing drug-resistant MCF-7/MDR tumors. The tumors exhibit a rapid growth rate following treatment with saline (Fig. 9l and m), DOX (Fig. 9n), or CPT (Fig. 9o), but show significantly slow growth rate after treatment with our HNDIS (Fig. 9p). The experimental results highlight HNDIS's superior ability to overcome the specific tumor resistance where even high-dose chemotherapy failed. In addition, tumors exhibit the most significant nuclear shrinkage (Fig. 9q), the lowest Ki67 expression (Fig. 9r), and the greatest extent of TUNEL staining following treatment with HNDIS (Fig. 9s). Notably, the tumor tissues exhibit intense red C11-BODIPY^581/591 fluorescence after treatment with saline, DOX or CPT. In contrast, the tumor tissues display significantly green C11-BODIPY^581/591 fluorescence after HNDIS treatment (Fig. 9t), which indicates that the nanofactory can potently induce LPO accumulation and thereby promoting ferroptosis-based damage in drug-resistant tumors.

3.12. In vivo biosafety of the nanofactory

The hemolysis ratio of HNDIS remains below 5.0 % at a high concentration of Fe (C_Fe = 200 μg/mL), indicating the favorable blood compatibility of the nanofactory (Fig. S26). The routine blood indices, including RBC, HGB, HCT, MCV, MCH, MCHC, WBC, and PLT, stay within standard levels even at HNDIS doses of 10 mg/kg or 15 mg/kg (Fig. S27). In addition, even at HNDIS doses of 10 mg/kg or 15 mg/kg, the serum biochemistry indices (ALT, AST, BUN, CRE, and CK-MB) show no significant deviation from normal ranges (Fig. S28). In addition, the H&E-stained sections of major organs (heart, liver, spleen, lung, and kidney) reveal no inflammatory or degenerative changes following treatment with HNDIS (15 mg/kg Fe dose) (Fig. S29). The above results collectively confirm the favorable in vivo biosafety of the nanofactory.

4. Conclusion

In summary, a hollow nanofactory with dotted interior surface (HNDIS) was designed and successfully constructed. The “house” of HNDIS is HMION. The “window” of HNDIS is the mesoporous channels of HMION, through which small molecules (e.g., glucose) can pass, but macromolecules (e.g., CAT and ferritin) cannot. The “machine” of HNDIS is AuNP that are distributed on the interior surface of HMION. The unique “window” of the nanofactory allows pass for glucose to co-work with AuNP and HMION to drive H₂O₂ production and Fe²⁺ release, and then generate a huge amount of •OH. Macromolecules cannot pass through the “window”, which prevents the undesirable catalysis of Fe²⁺ to Fe³⁺ by ferritin and H₂O₂ to H₂O by CAT. The data of TEM, XPS, HAADF-STEM, DLS, zeta potential measurements, XRD, and N₂ adsorption and desorption analysis collectively demonstrate the successful fabrication of the nanofactory. The nanofactory can generate an abundance of •OH, thereby inducing significant LPO accumulation and powerful ferroptosis within tumor cells. The average 4T1-tumor inhibition rate reaches 87.5 % after treatment with the nanofactory, and the mice survived beyond 36 days without significant alteration of body weights, indicating the excellent anti-tumor efficacy and low long-term toxicity of the nanofactory. The nanofactory also achieves a superior therapeutic effect to traditional chemotherapeutics (e.g., DOX, or CPT) and possesses effective tumor inhibition on large tumor-bearing mice. Meanwhile, the nanofactory enables visualization of nanoparticle distribution and metabolism via its T₂-weighted MRI capability in vivo. Finally, the biosafety of the nanofactory is confirmed by the minimal hemolysis ratio, the normal hematological and serum biochemical parameters, and the histopathological findings.

CRediT authorship contribution statement

Jing Yang: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Software, Validation, Visualization, Writing – original draft. Tianshui Bai: Conceptualization, Data curation, Investigation, Methodology, Software, Visualization. Haobin Cai: Investigation, Methodology, Software, Visualization. Haolong Ma: Conceptualization, Methodology, Software, Visualization. Bin Ren: Conceptualization, Methodology, Software, Visualization. Zongheng Li: Formal analysis, Methodology, Software, Visualization. Qingdeng Fan: Formal analysis, Methodology, Software, Visualization. Lin Huang: Resources, Visualization, Writing – review & editing. Gang Liu: Conceptualization, Funding acquisition, Project administration, Resources, Writing – review & editing. Guobin Hong: Conceptualization, Methodology, Resources, Visualization, Writing – review & editing. Zheyu Shen: Conceptualization, Funding acquisition, Methodology, Project administration, Resources, Supervision, Visualization, Writing – review & editing.

Declaration of competing interest

The authors declare no conflict of interest.

Appendix A. Supplementary data

The following is/are the supplementary data to this article.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.