Steering H₂O₂ lysis pathway for ROS generation in Prussian blue nanozymes via alkali cation doping

Abstract

Prussian blue nanoparticles (PBNPs) have emerged as versatile nanozymes with reactive oxygen species (ROS)-scavenging capabilities, predominantly applied in antioxidant therapies. In this work, we present a combined theoretical and experimental study demonstrating that modulating Fe coordination environments can fundamentally reconfigure PBNPs’ catalytic properties, enabling ROS generation and pro-oxidative functionality. Ab initio molecular dynamics revealed different H₂O₂ lysis mechanisms at Fe sites with varying coordination numbers: Low-coordinated center (FeN₄) induced hydrogen atom transfer to form Fe=O species, while high-coordinated FeN₅ generated ·OH radicals via H⁺-assisted homolysis under acidic conditions. Guided by calculations, Cs-doped PBNPs (Cs-PBs) with elevated coordination numbers were synthesized via alkali cation stoichiometric control, leveraging high distribution coefficient and low hydration energy of Cs⁺. Experimental results confirmed radical generation in Cs-PBs aligned with theoretical predictions. The size-optimized Cs-PBs demonstrated ultrahigh peroxidase-like activity (1182.26 U·mg^-1) and outperformed ROS generating properties in both pollutant degradation and chemodynamic therapy. This work redefines PBNPs’ catalytic potential beyond conventional antioxidant roles, and lays the foundation for innovative environmental and therapeutic solutions.

Introduction

The development of highly active nanozymes has emerged as a promising alternative to address the challenges faced by natural enzymes in scalable manufacturing and environmental adaptability^1,2,3,4. Iron-based nanozymes, as pioneering nano catalysts, feature Fenton reaction as the main pathway for peroxidase (POD)-like activity⁵. Exceptionally, the POD-like process of Prussian blue nanoparticles (PBNPs) can scavenge hydroxyl radicals and has been heavily focused on antioxidant applications^6,7, of which the unique catalytic mechanism has been intensively studied. The function of PBNPs as POD mimetics could be traced to the initial work on exploring the role of efficient electron transporters due to the abundant redox potentials of different forms⁶. Current mechanistic understanding suggests POD-like activity of PBNPs operates through iron-oxo intermediates rather than free radical generation. Chen et al. proposed structural parallels between surface FeNₓ sites in PBNPs and macrocyclic iron complexes, where catalytic cycles proceed via Ferryl species (Fe=O)⁸. Recent work showed that Fe-OH and Fe=O moieties formed on the surface of PBNPs upon pre-oxidation by H₂O₂ can initiate self-enhanced catalysis via either the conduction or valence band pathways⁹. However, all these studies demonstrated conspicuously absent free reactive oxygen species (ROS) generation in PBNPs, different from the classical Fenton chemistry in most iron-based nanozymes.

This knowledge gap arises from the fast precipitation reaction kinetics (K_sp ca. 1 × 10⁻⁴¹) of the current preparation methods, especially the dual-precursor method^10,11,12, leading to numerous hexacyanoferrate vacancies in the neighboring sites of Fe^III centers in reported PBNPs¹³. Similar iron-coordination configurations have been extensively studied in single-atom nanozymes, where explicit first- or second-shell coordination and axial ligand alterations have a marked effect on catalytic performance^14,15,16,17. This result provokes us to explore the potential Fenton ability of PBNPs as iron-based nanozymes by ligand modulation to Fe sites. An important structural modulation strategy in this context is the addition of alkali metal ions during synthesis to reduce the hexacyanoferrate vacancies. The presence of alkali ions may contribute to the transformation of insoluble PBNPs (Fe[Fe(CN)6]_3/4□_1/4·xH₂O) with no or few alkali ions (A) to soluble PBNPs (AFe[Fe(CN)₆]·xH₂O) rich in alkali elements based on stoichiometry¹⁸. It is noteworthy that Cs⁺ has a high distribution coefficient (K_d) for PBNPs^19,20, and doping with Cs⁺ during synthesis could promote the crystallinity.

In this study, we applied theoretical predictions that the emergence of highly coordinated Fe sites in PBNPs can generate ROS under acidic conditions and successfully verified in Cs-doped highly crystalline PBNPs by electron paramagnetic resonance (EPR) spectroscopy. Constrained ab initio molecular dynamics (c-AIMD) calculations simulated the ·OH generation process in pristine PB surfaces and PB-H₂O interfaces with different degrees of Fe site coordination. Hydrogen atom transfer (HAT) was observed directly after H₂O₂ lysis at the low-coordinated (FeN₄) center, producing Fe=O species. And the high coordinated FeN₅ center, despite generating Fe=O species on the pristine (200) surface, has an unprecedented ·OH generating capacity under acidic conditions. H⁺ significantly lowers the energy barrier of H₂O₂ homolysis and facilitates the formation of free ·OH. Cs-doped PBNPs (Cs-PBs) synthesized by various methods have been shown to produce significant ·OH signals. High distribution coefficient and low hydration energy of Cs⁺ endow Cs-PBs with high crystallinity and surface highly coordinated FeN₅ sites, demonstrating superior pollutant degradation efficiency and tumor cytotoxicity in the proof-of-concept, which is different from PBNPs synthesized by conventional methods.

Results

Theoretical prediction of radical generation

Density functional theory (DFT) and c-AIMD calculations were applied here to study the complete dissociative chemisorption mechanisms of H₂O₂ for POD-like catalyses of PBNPs. To get insight into the effect of experimental conditions on the POD-like catalyses of PBNPs, we first considered the initial structures, denoted as pristine PB (200) surface with only one H₂O₂ molecule, which maintained a perfect FeN₅ coordination structure. After a relaxation process of 5 ps, as depicted in Supplementary Fig. 1, the system temperature and total energy exhibited that they reached equilibrium. We calculated the overall H₂O₂ dissociative process on the pristine PB surface using c-AIMD and the selected collective variable is the O−O distance in H₂O₂ (Fig. 1a). Initially, H₂O₂ bound to Fe atom on the PB surface (Supplementary Fig. 2). An 0.3 eV of energy barrier is required to dissociate H₂O₂ on the pristine PB surface (Fig. 1e). HAT was observed during the dissociative pathway, ultimately producing the Fe=O moiety and free H₂O as shown in Supplementary Movie 1. The results demonstrate that the only transformation pathways from H₂O₂ to adsorbed *O can occur on the pristine PB surface. In contrast, for the PB-H₂O interface, a higher 0.75 eV of energy barriers is required for the dissociation of H₂O₂ (Fig. 1b, f). The homolysis pathway was observed during the dissociation of H₂O₂ (Taken from Supplementary Movie 2). Due to the dynamic hydrogen bonding between H₂O₂ and water, a Fe-OH configuration was eventually formed, as shown in Fig. 1b, instead of a HAT process. The calculated free energy of the obtained Fe-OH structure is also higher than the initial structure, indicating the poor H₂O₂ activation performance. This is consistent with the generation of only rare free radicals by PBNPs under neutral conditions⁹. Furthermore, considering that POD-like catalysis of PBNPs generally occurred in the acidic environment, hydronium ions (H₃O⁺) are added to PB-H₂O interface. As shown in Fig. 1c (Taken from Supplementary Movie 3) and 1d, it is a barrierless process for the dissociation of H₂O₂ with the presence of H₃O⁺. The H₂O₂ molecule can be divided into two OH. One can capture H⁺ from H₃O⁺ to form ·OH (H⁺), i.e., H₂O^+·, while another absorbed on the surface to form Fe-OH configuration. The generated ·OH (H⁺) could act as a highly oxidative species extracting electrons from substrates such as 3, 3’, 5, 5’-Tetramethylbenzidine (TMB), completing the entire proton-coupled electron transfer process and driving the color reaction (Fig. 2a).

Fig. 1: Theoretical prediction of H₂O₂ lysis in different environments.

Structures of the initial, intermediate and final configurations of the dissociative process of H₂O₂ on a the pristine PB surface, b the PB-H₂O interface and c the PB-H₂O/H₃O⁺ interface during c-AIMD simulation. The gray balls represent C atoms, the blue balls represent N atoms, the red balls represent O atoms, the white balls represent H atoms, and the greyish blue balls represent Fe atoms (H₂O₂ adsorption sites). d Scheme of energy barriers for H₂O₂ lysis on different surfaces. The dissociative barriers for H₂O₂ on e the pristine PB surface and f the PB-H₂O interface.

Fig. 2: The different H₂O₂ lysis pathway in the PB surface.

a Without and b with V_CN.

We then further refined the interface model to assess the structure-activity relationship of PBNPs. The absence of ·OH in PBNPs synthesized by the dual-precursor method is attributed to an inherent ligand defect. For simplicity, we only consider a defect of cyano vacancy (V_CN) and theoretically investigate its impact on the kinetics of H₂O₂ dissociation. Supplementary Fig. 3a shows the dissociation of H₂O₂ on a pristine PB surface with V_CN (FeN₄ sites), as seen in Supplementary Movie 4. On the pristine PB surface with V_CN, the O–O bond cleavage presents an exceptionally low kinetic barrier (0.06 eV, Supplementary Fig. 3c). Here, the hydrogen atom directly transfers from the adsorbed *OH to the free ·OH, forming a configuration of Fe=O alongside H₂O. At the PB-H₂O interface with V_CN, the dissociation of H₂O₂ becomes a barrierless process (Supplementary Fig. 3b, taken from Supplementary Movie 5), independent of protons in the solvent. And the HAT process is likewise observed to form Fe=O. Under acidic experimental conditions, a similar process happens due to the barrierless characteristic of O–O bond breakage (Fig. 2b), suggesting that PBNPs with ligand defects, as typically produced in conventional synthesis, primarily utilize the Ferryl pathway for POD-like catalysis.

All the theoretical results demonstrated that experimental conditions seem to be dominated for POD-like catalysis of PBNPs. The generation of Fe=O species on the perfect pristine PB surface in vacuum seems to explain the non-radical mechanism of PBNPs. However, the addition of solvent significantly alters the active energy barrier and the H₂O₂ cleavage mechanism. The POD-like mechanism involving H₂O₂ primarily includes two fundamental reaction steps: homolytic cleavage of O-O bonds and oxidation processes accompanied by proton-coupled electron transfer to the hydroxyl group. As a critical elementary step, homolytic cleavage of the O-O bond occurs favorably only on defect-free surfaces under acidic conditions, facilitating ·OH generation. This finding reveals the potential Fenton-like activity of PBNPs as iron-based nanozymes, which appears contradictory to prior reports. The fallacy regarding non-radical pathways in PBNPs stems from overlooking the precise coordination environment of Fe sites. Regarding surface reaction processes, distinct proton donors lead to active species alterations from free radical to bound oxygen. At defect centers, O-O bond cleavage proceeds directly in aqueous, bypassing H⁺ assistance. Essentially, this transformation originates from charge redistribution at metal centers induced by electron-rich cyano defects. It is the prevalence of under-coordinated FeN₄ sites that dictates the non-radical pathway.

Coordination evolution via alkali cation doping

The generation of free radicals in PBNPs could be experimentally verified by synthesizing PBNPs with highly coordinated iron centers. Inspired by the stoichiometric formulas of soluble and insoluble PBNPs, we introduced alkali metal cations during precipitation to modulate crystallinity (Fig. 3a). The samples were denoted as A-x, where A represented alkali metal type, x referred to the molar amount relative to Fe³⁺. Alkali-free Prussian Blue (PB-0) featured cubic morphology with 60-nm edge length (Fig. 3b) and alkali-doped PBNPs maintained similar sizes (Supplementary Figs. 4–8). All PBNPs showed hydrodynamic sizes of ~100 nm (Fig. 3e) and ζ-potentials of −35 mV (Supplementary Fig. 9), indicating a good monodisperse and aqueous stability. The atomic alkali/iron ratios, determined by inductively coupled plasma mass spectrometry (ICP-MS) and energy dispersive spectrometry (EDS), increased with alkali ions feeding amount (Supplementary Table 1). High-angle angular dark-field scanning transmission electron microscopy (HAADF-STEM) and corresponding EDS mapping images of presentative doped PBNPs (Fig. 3c, d) revealed uniform iron and alkali metal distributions. More elemental mappings of K- and Cs-doped PBNPs (K-PBs and Cs-PBs) also proved rising alkali content (Supplementary Figs. 10 and 11). X-ray photoelectron spectroscopy (XPS) surveys detected negligible potassium in other alkali-doped samples (Supplementary Fig. 12), excluding the interference from K₄Fe(CN)₆. Fe 2p XPS deconvolution qualitatively estimated surface Fe valence distribution in PBNPs. Given the covalent characteristics of Fe-CN-Fe bonds²¹, Fe oxidation states are not precisely 3+ or 2+. The Fe^II/Fe^III ratio increased upon cations doping to maintain electroneutrality, particularly evident in Cs-PBs (Fig. 3f). The higher proportion of Fe^II also caused a redshift in the UV-vis spectrum (Supplementary Fig. 13), probably from the charge transfer of enhanced Fe_C-t_2g to Fe_N-t_2g band and suppressed Fe_C-t_2g to antibonding Fe_N-e_g band²². XPS results for N 1s showed broadened, asymmetric CN peaks, which were attributed to CN(···Cs)-Fe (398 eV) (Supplementary Fig. 14)^19,23.

Fig. 3: Morphological and elemental characterizations.

a Scheme of the synthesis process of alkali-doped PBNPs. b SEM image of PB-0. TEM images and the corresponding EDS mapping images (Scale bar, 50 nm) for c K-32 and d Cs-5. e Hydrodynamic sizes of PBNPs doped with different alkali cations compared with PB-0. Error bars indicate the standard deviation for the hydrodynamic sizes of five PB groups doped with the same alkali cation. f Fe 2p XPS spectra and deconvolution results of PB-0, K-PBs and Cs-PBs.

More detailed coordination environment in alkali cation-doped PBNPs were investigated. The structural evolution, characterized by crystallinity, coordination number, and electronic configuration relative to PB-0, reveals significant dopant effects. X-ray diffraction (XRD) patterns of PB-0 and K-PBs matched the Fe₄^III[Fe^II(CN)₆]₃ phase (JCPDS 01-0239) (Supplementary Fig. 15). Potassium doping enhanced the (220)/(200) Bragg peak intensity ratio (I(220)/I(200)) from 0.403 to 0.589. Rietveld refinement revealed a cubic phase with identical unit cell parameters (10.17 Å) for both PB-0 and K-32, consistent with the reported value (~10.2 Å)²⁴ (Fig. 4a, Supplementary Fig. 16 and Supplementary Table 2). The iron coordination number increased progressively with K occupancy (0.360 to 0.576), corresponding to FeN_3.5-FeN_4.7 configurations (Fig. 4b). For Cs-PBs, I(220)/I(200) surged from 2.644 (Cs-0.5) to 14.28 (Cs-5) upon Cs occupancy in tetrahedral sites exceeded 40% (Supplementary Fig. 17), coinciding with lattice expansion to 10.26 Å (Supplementary Table 3). Refinement indicated Cs occupancy increased from 0.312 (Cs-0.5) to 0.528 (Cs-5), with iron adopting an average FeN₆ configuration (Fig. 4b). Rb-doped PBNPs (Rb-PBs) exhibited intermediate I(220)/I(200) values (0.764-1.397) between K and Cs (Supplementary Fig. 15). High-resolution transmission electron microscopy (HRTEM) images with corresponding fast Fourier transform (FFT) of PB-0 revealed nearly amorphous structures and weak diffraction signals (Fig. 4c). Li/Na/K-PBs exhibited similar amorphous morphology akin to PB-0 (Fig. 4c and Supplementary Figs. 18 and 19) while Rb/Cs-PBs showed more defined lattice patterns (Fig. 4c and Supplementary Fig. 20). Enhanced crystallinity in Cs-5 was further confirmed by stronger ν(Fe^II-CN-Fe^II) vibrational peaks in Raman spectra (Supplementary Fig. 21), reduced mass loss of interstitial water in thermogravimetric curves (Supplementary Fig. 22), and a broad, intense 77 K EPR signal at g = 2.02 (Supplementary Fig. 23).

Fig. 4: Structural characterization of alkali cation-doped PBNPs.

a XRD and Rietveld refinements for PB-0, K-32 and Cs-5. b Crystal structures of PBNPs with varying degrees of N-coordinated Fe sites and alkali metal contents. c HRTEM images of PB-0, K-32 and Cs-5. The images below are enlargement of the areas within the square above. d ⁵⁷Fe Mössbauer spectra of PB-0, K-32 and Cs-5. e Temperature-dependent magnetic susceptibility (χ_m-T) plots, the inset shows the effective magnetic moment (μ_eff) and unpaired electron number (n) of PB-0, K-32 and Cs-5. f Scheme of the structure regulation mechanism to PBNPs by different alkali cations (View from (200) direction).

The charge redistribution primarily dominates the spin state of Fe, thereby influencing the catalytic activity^25,26. C-coordinated Fe in PBNPs usually adopts a singlet state (S = 0) with negligible spin effects. N-coordinated Fe can adopt multiple states depending on the coordination environments. To probe the Fe species, ⁵⁷Fe Mössbauer spectroscopy was performed (Fig. 4d). The isomer shift (IS) for PB-0 revealed that 61.1% of the iron species adopted a low-spin Fe^II state (Supplementary Table 4), which also remained dominant in K-32. However, high-spin Fe^III increased to nearly match the low-spin Fe^II proportion in Cs-5. A three-doublet fitting approach was used to analyze Fe^IIIN_x sphere distributions (Supplementary Fig. 24)¹⁸. For PB-0, 22.93% of Fe^III adopted the (H₂O)FeN₅ or trans-(H₂O)₂FeN₄ configurations, 2.45% adopted the mer-(H₂O)₃FeN₃ or cis-(H₂O)₂FeN₄ configuration, and 13.47% adopted intact FeN₆ coordination. In K-32, Fe^III in (H₂O)FeN₅ or trans-(H₂O)₂FeN₄ configurations decreased, approaching intact FeN₆ levels. For Cs-PBs, Fe^III with the smallest quadrupole splitting increased to 33.50%, indicating enhanced FeN₆ coordination. The temperature-dependent magnetization demonstrated reduced unpaired electrons (n) from 4.61 (PB-0) to 4.06 (Cs-5), suggesting an electronic configuration transition from t_2g⁴e_g2 (high spin) to t_2g⁵e_g1 (medium spin) due to ligand field splitting from intact FeN₆ coordination (Fig. 4e).

These results suggest that alkali-rich PBNPs, especially Cs-PBs, significantly diminish vacancies and modulate spin states. Intact FeN_x coordination sphere increases a higher coordination number while lowering spin. Figure 4f revealed that the modulation mechanism of PBNPs crystallinity by alkali metal cations located in different periods was actually related to many variables like the ionic radius (Supplementary Table 5). The hydration radius of alkali metal ions decreased with increasing periodic numbers due to lower charge density²⁷. Cs⁺, with the smallest hydration radius (3.28 Å) and the largest ionic radius (1.67 Å), matches well with the size of the ion channel (~3.2 Å) in the PBNPs’ cells²⁸, whereas Rb⁺ (diameter 2.96 Å) and K⁺ (2.76 Å) are able to escape from the channel more flexibly. Furthermore, Cs⁺ is more susceptible to dehydrogenation for a loose hydration layer and the longest A⁺-O_w bond lengths (3.15 Å, 1.15 times longer than that of K⁺). For these reasons, Cs⁺ exhibits unique advantages in ion selectivity during the PB synthesis.

Activity measurement and reactive species identification

Significant changes in the Fe site coordination environment may modulate enzyme-like activities, as calculations predicted. The POD-like activities of PBNPs were then assessed by catalyzing the oxidation of TMB with H₂O₂. Quantitative comparison to the specific activity of PB-0 (a_nano = 78.1 U/mg) revealed gradual activity suppression in Li/Na/K-PBs. But the activities of Rb/Cs-PBs were conversely enhanced (Fig. 5a, a_nano is 260.5 U/mg for Cs-0.5 and 335.1 U/mg for Cs-5, representatively). The activity of the cation has firstly been excluded (Supplementary Fig. 25). Cs-5 exhibited superior POD-like activity at equivalent concentrations versus PB-0 or K-32 (Fig. 5b). We further demonstrated that the rapid electron transfer between H₂O₂ and different PBs. As shown in Supplementary Fig. 26, Cs-5 exhibited higher current density than PB-0 over a wide potential window of 0–0.6 V_RHE. The accelerated hydrogen peroxide reduction reaction (HPRR) also evidenced higher POD-like activity. Michaelis-Menten kinetics confirmed elevated activity in Cs-PBs with increasing Cs content, showing higher turnover numbers (k_cat) and substrate affinity for both H₂O₂ and TMB (Supplementary Fig. 27 and Supplementary Table 6). The fitted parameters revealed that Cs-5 had a 2.5-fold k_cat to H₂O₂ and a 4-fold higher affinity for H₂O₂ than PB-0 (Fig. 5c). In contrast, K-PBs displayed inhibited POD-like activity despite increased affinity for TMB substrates (Supplementary Fig. 27).

Fig. 5: POD-like enzymatic characterization and identification of active intermediates.

a Relative POD-like specific activity of alkali-doped PBNPs compared to PB-0. The lines represent the linear fits to the relationship between actual doping ratio and relative activity, and the shaded areas denote the 95% confidence band. b Time-dependent absorbance curves of the TMB colorimetric reaction catalyzed by representative PB-0, K-32 and Cs-5 in the same concentration ([Fe] = 2 μg/mL). c Steady-state kinetic for PB-0, K-32 ([Fe] = 5 μg/mL) and Cs-5 ([Fe] = 2 μg/mL) in the presence of varying H₂O₂ concentration. The curves denote the nonlinear fit using Michaelis-Menten equation. d Quantification results of PMSO and PMSO₂ in different PBNPs/H₂O₂ systems by HPLC. Error bars in a–d indicate the standard deviation of three independent experiments. EPR spectra of e DMPO-OH and f DMPO-O₂^·- catalyzed by PB-0, K-32 and Cs-5. g Time-dependent in situ Raman spectra catalyzed by PB-0 and Cs-5 at room temperature. h Overall catalytic process of PBNPs activated by H₂O₂. Cs-PB activates H₂O₂ mainly through the radical pathway from (i) to (v), while PB-0 or K-PB mainly produce non-radical species via the Ferryl pathway from (i) to (iv).

We then focused on validating the active species generated during the POD-like process. High-performance liquid chromatography (HPLC) was applied to track the generation of Fe=O, which could convert specifically sulfoxide (PMSO) to sulfone (PMSO₂)^29,30. All PBNPs showed evident depletion of PMSO versus controls (Fig. 5d). PB-0 and K-32 exhibited nearly identical yields of PMSO₂, indicating a transfer pathway to form Ferryl species. However, the minimal detection of PMSO₂ in Cs-5 suggests that most PMSO might be consumed through an alternative oxidative process. According to the computational prediction, fluorescence spectroscopy and EPR were employed to track ·OH generation, using p-terephthalic acid (PTA) and 5,5-Dimethyl-1-pyrroline N-oxide (DMPO) as ·OH trapping agents, respectively. Cs-5 exhibited rapid PTA oxidation (Supplementary Fig. 28) and a distinct DMPO-OH signal (Fig. 5e), which were neither observed in PB-0 or K-32. The ·OH generation of Cs-PB was inhibited by adding free radical inhibitors such as mannitol (an ·OH scavenger) or dimethyl sulfoxide (DMSO, an O-atom acceptor) (Supplementary Fig. 29). In addition, Cs-5 was capable of generating superoxide anion (·O₂⁻) (Fig. 5f) and dissolved oxygen (O₂) (Supplementary Fig. 30) in an acidic environment. This likely arises from the oxidation of H₂O₂ by ·OH to produce protonated ·O₂^- and the subsequent Haber-Weiss reaction (Reaction 1–3)³¹. Typically, iron-based nanozymes involved in the Fenton process are hindered by sluggish Fe(III) reduction by H₂O₂ (k = 0.01–0.02 M⁻¹ s⁻¹). But the reaction rate constant between Fe(III)-OH (*OH) and ·O₂⁻ (k = 5 × 10⁷M⁻¹ s⁻¹) is much higher than that of Fe(III) with H₂O₂, significantly expediting the regeneration of Fe(II) sites in Cs-PBs (Reaction 4). Time-dependent in situ Raman spectroscopy revealed a gradual intensity growth at ~430 cm⁻¹ for Cs-5 (Fig. 5g), demonstrating an accumulation of *OH species on Cs-5³². But the peak change was almost imperceptible for PB-0 and K-32 (Supplementary Fig. 31). Collectively, Cs-PBs enable enhanced H₂O₂ lysis through generating multiple ROS, boosting the catalytic rate beyond conventional PBNPs (Fig. 5h).

$${{{\rm{H}}}}_{2}{{{\rm{O}}}}_{2}\to*{{\rm{OH}}}+\cdot {{\rm{OH}}}$$

(1)

$$\cdot {{\rm{OH}}}+{{{\rm{H}}}}_{2}{{{\rm{O}}}}_{2}\to {{{\rm{H}}}}_{2}{{\rm{O}}}+{{\rm{H}}}{{{\rm{O}}}}_{2} \cdot$$

(2)

$${{\rm{H}}}{{{\rm{O}}}}_{2} \cdot+{{{\rm{H}}}}_{2}{{{\rm{O}}}}_{2} \to {{{\rm{H}}}}_{2}{{\rm{O}}}+{{{\rm{O}}}}_{2}+\cdot {{\rm{OH}}}$$

(3)

$${{\rm{H}}}{{{\rm{O}}}}_{2} \cdot+\ast {{\rm{OH}}} \to {{{\rm{H}}}}_{2}{{\rm{O}}}+{{{\rm{O}}}}_{2}$$

(4)

The experimental validation of the free radical pathway can be fundamentally attributed to the subtle refinements to the iron coordination sphere. Nucleophilic cyanide vacancies induce charge redistribution at Fe sites, polarizing O–H bond and promoting subsequent Ferryl formation. Highly coordinated iron centers (FeN₅) exhibit intermediate spin states and weakened oxidative capacity, which causes the H₂O₂ homolysis to be dependent on H⁺. The catalytic activity of Cs-PBs is enhanced at lower pH (Supplementary Fig. 32). Elevated pH values markedly suppress the EPR signal (Supplementary Fig. 33), where diminished proton availability impedes the velocity of H₂O₂ dissociation, rendering the process kinetically unfavorable. Cs⁺ doping primarily modulates crystallinity, substantially increasing the proportion of radical-dominant FeN₅ sites while maintaining partial FeN₄ sites in Cs-PBs. In situ Raman spectroscopy detected a Fe=O vibrational signal in both PB-0 and Cs-5 (Supplementary Fig. 34), providing direct evidence for the coexistence of dual catalytic pathways in Cs-PBs. Critically, this crystallinity modulation strategy relies on synthesis-stage alkali cation incorporation, not post-synthetic adsorption. Although Cs⁺ incorporation into PB frameworks occurred after dispersed in CsCl solution (Supplementary Fig. 35), PB-0 adsorbing Cs⁺ exhibited no significant changes in hydrodynamic properties, nor enzymatic activity and detectable ·OH generation (Supplementary Fig. 36). Small cation-doping (Li, Na, and K) paradoxically suppressed activity. While inducing minor structural refinement, K⁺ altered negligible Fe coordination number and spin states overall. Typically, defect suppression in highly amorphous conventional PBNPs impedes catalysis³³. Given that K-PBs retain amorphous surface characteristics, the observed activity inhibition likely stems from bulk electronic modifications (e.g., band structure alterations) rather than surface coordination environment.

Applications in environmental and therapeutic solutions

The explicit ROS generation in PBNPs has modified the application landscape significantly, from environment to therapeutics. Before displaying the application potential, we first verified the universality of the mechanism in various methods to optimize Cs-PBs with higher POD-like activity. Three Cs-PBs were synthesized by different methods. CsPB-1, 2, and 3 are denoted as citrate-modified Cs-PBs synthesized by the dual-precursor method, PVP-modified Cs-PB synthesized by the dual-precursor method, and PVP-modified Cs-PB synthesized by the single-precursor method, respectively, featuring different particle sizes (Fig. 6a). The specific activities exhibited size-dependent properties, in which CsPB-1 at about 40 nm had a maximum a_nano of 1182.26 U mg⁻¹ (Fig. 6b). EPR spectra indicated that Cs doping strategy was capable of inducing the ·OH generation in Cs-PBs with different modifications and synthesis methods. These results suggested that in addition to Cs⁺ doping, size effects and surface modifications also modulate the POD-like activity of Cs-PBs. Moreover, CsPB-1 with good dispersion and consistent catalytic activity was obtained in different batches and progressively expanded reaction systems (Supplementary Figs. 37 and 38), proving that the simple precipitation method exhibited good stability and scalability and has potential for large-scale catalytic applications.

Fig. 6: Potential applications of Cs-PBs for contaminant degradation and cancer treatment.

a TEM graphs (Scale bar, 100 nm), and hydrodynamic sizes of Cs-PBs synthesized with different methods. b Upper figure represents the specific activities of different Cs-PBs. Solid lines represent the linear fit of each specific activity, and the shaded areas denote the 95% confidence band. Lower figure refers corresponding hydroxyl radical generation capacity. c The degradation curves of MB and d the calculated k_obs according to the pseudo-first-order kinetics in a series of iron-based catalytic systems (Conditions: [MB]₀ = 0.2 mg/L, [H₂O₂]₀ = 500 mM, [Catalyst] = 2 mg/L, pH = 4.0). e MALDI-TOF-MS results of MB degradation in different iron-based catalytic systems. f Cyclic catalysis results of CsPB-1 towards MB degradation in ambient temperature. g Cell viability of 4T1 and 3T3 cells treated with CsPB-1 for 24 h. The curves represent a nonlinear fit based on the dose-response function. h Laser scanning confocal microscopy images of ROS levels for 4T1 cells treated with PB-0 and CsPB-1 (Scale bar, 20 μm). PBS-treated 4T1 cells were used as control. i Quantitative results of mean fluorescence intensity for ROS levels of 4T1 and 3T3 cells. Error bars in b, c, f, g indicate the standard deviation of three independent experiments, and bars in i denote four independent experiments.

Advanced oxidation process (AOP) based on the Fenton reaction is an effective strategy for contaminant degradation. CsPB-1 was applied for the degradation of methylene blue (MB) due to high POD-like activity and ·OH generation capability. In contrast to the excellent degradation performance of the CsPB-1/H₂O₂ system, the removal percentages of MB in H₂O₂ system within 30 min remained below 2%, indicating negligible activation of H₂O₂ alone (Fig. 6c). Neither PB-0 nor Fe₃O₄ NPs showed significant degradation capacity, stemming from non-radical catalysis and weaker POD-like activity, respectively, emphasizing the important role of Cs-PBs in MB degradation. Specifically, the degradation rates of MB in CsPB-1/H₂O₂ system were 0.17 min⁻¹, one order of magnitude higher than Fe²⁺ and two orders of magnitude higher than PB-0 (Fig. 6d). The CsPB-1/H₂O₂ system exhibited outstanding capabilities to those of Fe²⁺ for multi-dyes, with removal efficiencies exceeding 95% (Supplementary Fig. 39). The products after 24 h MB degradation in different systems were monitored by Matrix-assisted laser desorption/ionization time of flight mass spectrometry (MALDI-TOF MS). PB-0 mainly oxidized the S atom in MB through the Ferryl pathway, generating sulfoxide products. Whereas, in the case of producing ·OH, CsPB-1 and Fe²⁺ exhibit similar degradation products, including bilateral dimethylamine demethylation and central aromatic heterocyclic ring breakage (Fig. 6e). The long-term stability of CsPB-1 was also investigated under the same test conditions by centrifugal recovery (Supplementary Fig. 40). The system maintained a high decontamination activity over four consecutive catalytic cycles (more than 95% removal rate in 20 min, Fig. 6f) with a slight decrease. The catalytic process was accompanied by only slight Cs⁺ leaching (Supplementary Fig. 41), which is negligible regarding environmental pollution. In addition, incubating CsPB-1 with H₂O₂ under acidic conditions for 24 h to thoroughly oxidize the surface resulted in a further decreased catalytic capacity. Persistent H₂O₂ exposure during cyclic catalysis induces gradual surface oxidation and etching of Cs-PBs³⁴. This structural degradation reduces Fe coordination numbers and elevates metal valency, consequently shifting the dominant reactive pathway from ·OH generation to Fe=O formation—directly evidenced by diminished ·OH signal intensity in EPR results (Supplementary Fig. 42). The loss of highly coordinated Fe sites may cause a shift to the Ferryl pathway to maintain a certain level of catalytic activity.

Cs-PBs demonstrated robust ROS generation capacity under tumor environment. CsPB-1 was potently cytotoxic to 4T1 cells, with a viability of less than 20% under treatment at a concentration of 40 μg/mL, whereas it maintained high biocompatibility (>80% survival) with 3T3 cells under the same conditions (Fig. 6g). Note that sole Cs⁺ has negligible cytotoxicity for 4T1 cells. The slight effect on 3T3 at high concentrations stems from the sensitivity of fibroblasts to heavy metal ions, which can be effectively inhibited by Cs⁺ encapsulation in PB (Supplementary Fig. 43). Confocal microscopy revealed that CsPB-1 elevates intracellular ROS levels by 2-fold (p < 0.001) in 4T1 cells, attributable to greater H₂O₂ accumulation within tumor cells (Fig. 6h, i). This effect was attenuated in normal 3T3 fibroblasts (Supplementary Fig. 44). Control experiments with PB-0 showed the opposite function: the PB scavenged endogenous ROS in 4T1 cells (−51% vs. untreated, p < 0.0001), associated with negligible cytotoxicity (Supplementary Fig. 43). This suggests that cesium doping fundamentally reversed the redox-regulated behavior of PB nanozymes.

Discussion

Through a combined theoretical-experimental approach, we confirm explicit ROS generation in the POD-like process catalyzed by highly crystalline PBNPs. This activity originates from the highly coordinated FeN₅ sites in Cs-PBs. AIMD analyses revealed distinct mechanistic bifurcations: low-coordination FeN₄ centers promote HAT via enhanced O-H polarization at cyano-deficient sites, generating Fe=O intermediates. While FeN₅ configurations facilitate to yield ·OH radicals under acidic conditions. The successful synthesis of high-crystallinity Cs-PBs through alkali cation-mediated coordination engineering validated our computational models. The kinetic perspective overcomes the limitations of conventional transition-state heuristic methods, providing mechanistic clarity through molecular dynamics simulations. Notably, the ·OH-dominant Cs-PB system exhibits fundamentally distinct behavior from low-coordination amorphous analogs. It maintains high radical concentration despite irreversible erosion of FeN₅ sites, partially weakening catalytic persistence. The differential oxidizing capacities of adsorbed and free oxygen species break the notion of PBNPs as antioxidants only and confer mild oxidizing properties to normal PB and excellent ROS generation characteristic to Cs-PB. The shift and trade-off between the two identities offer potential for continuous catalytic applications.

Methods

Chemicals

All the chemicals were of analytical grade and were used without further purification. Polyvinyl pyrrolidone (PVP-K30), potassium hexacyanoferrate (II) trihydrate (K₄Fe(CN)₆·3H₂O), sodium chloride (NaCl), potassium chloride (KCl), iron (III) chloride hexahydrate (FeCl₃·6H₂O), 30% hydrogen peroxide (30% H₂O₂), methanol, acetic acid (HAc) and sodium acetate (NaAc) were obtained from Sinopharm Chemical Reagent Co., Ltd (Shanghai, China). Lithium chloride (LiCl), rubidium chloride (RbCl), cesium chloride (CsCl), TMB, 2,2’-Azinobis-(3-ethylbenzthiazoline-6-sulfonate) (ABTS), PTA, MB, rhodamine B (RhoB), methyl orange (MO) and DMSO were purchased from Aladdin Industrial Inc. (Shanghai, China). DMPO and Nafion 117 containing solution were acquired from Sigma-Aldrich (Shanghai, China). PMSO were purchased from Shanghai Yuanye Bio-Technology Co., Ltd. Hoechst 33342 staining solution and reactive oxygen species assay kit were purchased from Beyotime (Shanghai, China).

Characterization

The morphology of PBNPs was observed by scanning electron microscopy (SEM, Ultra Plus, Carl Zeiss). The elemental composition was analyzed by EDS, ICP-MS (Agilent Technologies 7700 Series) and XPS (Thermo ESCALAB 250Xi). The crystal structures of PBNPs were explored by HRTEM (Talos F200X, Thermo Fisher) and XRD (D8 Advance, Bruker). The hydrodynamic diameters and ζ-potentials of PBNPs in deionized water were analyzed by dynamic light scattering (NanoZS90, Malvern). The thermo gravimetric analysis (TGA) was conducted by TherMax 500 (Thermo Cahn). Raman spectra of PBNPs were collected by Xplora Plus (HORIBA) with 638 nm excitation. Ambient and 77 K EPR spectra were obtained from Bruker A300. The ⁵⁷Fe Mössbauer spectra were recorded by WissEl MR-2500 Mössbauer drive unit in transmission geometry, with a ⁵⁷Co/Rh source, which is equipped with a helium cryostat (Advanced Research Systems, Inc., 4 K). MALDI-TOF-MS results were recorded on Bruker autoflex speed.

Synthesis of PBNPs

Alkali metal-doped PBNPs were synthesized by three methods according the previous reports^6,8,35. PVP-modified PBNPs by dual-precursor methods: 1.11 g of PVP-K30 and 27.0 mg (0.1 mmol) of FeCl₃·6H₂O were dissolved in 80 mL purified water and continuously stirred at ambient temperature until dissolved. A 20 mL K₄Fe(CN)₆·3H₂O (42.2 mg, 0.1 mmol) aqueous solution was then slowly drop to the mixture of FeCl₃ and PVP for 30 min at 60 °C with vigorously stirring. The use of an injection pump allows for precise control of dosing at a uniform speed. After aging for an additional 30 min at 60 °C, the as-obtained precipitates were centrifugally rinsed for several times to remove excess reactants. Finally, the concentrate after centrifugation was filtrated with a 220 nm filter to remove the aggregates. For the A-x samples, the synthesis requires the addition of alkali metal chloride in x times the molar amount of FeCl₃, dissolving in 20 ml of potassium ferricyanide solution. Particularly, CsPB-2 was synthesized with 336.7 mg (2 mmol) addition of CsCl in order to obtain PVP-modified PBNPs with a sufficiently high activity.

CA-modified PBNPs by dual-precursor methods: 406 mg FeCl₃·6H₂O (1.5 mmol) and 505 mg CsCl (3 mmol) were dissolved in 50 mL of citric acid solution (0.25 M) to form solution A. 126.7 mg K₄Fe(CN)₆·3H₂O (0.3 mmol) was dissolved in 550 mL of citric acid solution (0.25 M) to form solution B. Solution A was then dropped into solution B in 10 min with vigorously stirring. The obtained blue solution was then aged at 60 °C for extra 20 min after mixing. The precipitates were collected by centrifuging and washed with DI water three times. CsPB-1 was synthesized by this method.

PVP-modified PBNPs by single-precursor methods: 3 g PVP-K30, 90 mg K₃Fe(CN)₆ (0.27 mmol) and 101 mg CsCl (0.6 mmol) were evenly dispersed in 40 mL of HCl (0.01 M). The mixed solution was then kept at 80 °C for 2 h to obtain a blue solution. The precipitates were collected by centrifuging and washed with DI water three times. CsPB-3 was synthesized by this method.

Computational methods

Both geometric optimization and electronic property calculations were performed using DFT as implemented in VASP^36,37,38. The Perdew-Burke-Ernzerhof (PBE) approach was employed to describe the exchange-correlation function^39,40 while the projector-augmented wave (PAW) method^41,42 accounted for core-valence electron interactions. To ensure accuracy during structural relaxation, a plane-wave basis set with a kinetic energy cutoff of 450 eV was applied, along with a total energy convergence tolerance of 10⁻⁵eV and a force convergence tolerance of 0.02 eV/Å. The van der Waals interactions were treated using the DFT-D3 method⁴³. The surface was modeled with four Fe-CN-Fe layers; the bottom two layers were fixed to represent the bulk structure, and the top two layers were allowed to fully relax. The dynamic barriers were calculated via c-AIMD simulations and the slow-growth sampling method (SG-AIMD)^44,45. For SG-AIMD, the O–O bond length in H₂O₂ was used as the collective variable to drive reaction progression. The canonical ensemble was used, with temperature maintained at 300 K by a Nosé-Hoover thermostat. A time step of 0.5 fs was used, and each simulation was propagated for 10 ps. A vacuum layer of approximately 15 Å was consistently maintained to avoid interactions between periodic images. The reaction free energies were evaluated using the constant-potential hybrid solvation dynamics model (CP-HS-DM), as developed by Liu et al.^46,47,48,49. The atomic coordinates for the lysis process in Fig. 1 and Supplementary Fig. 3 are provided in Supplementary Data 1. Furthermore, the adsorption energies (E_ads) of H₂O₂ on the (001) and (220) surfaces are −0.945 eV and −0.643 eV. The results show that H₂O₂ exhibits significantly stronger binding affinity on the (001) surface compared to the (220) surface, indicating that the (001) plane provides a more favorable thermodynamic environment for H₂O₂ activation. Therefore, the (001) surface was selected as the adsorption surface.

Zero-field cooling temperature-dependent magnetic susceptibility measurement

Physical Property Measurement System (PPMS, Quantum Design) was used to conduct temperature-dependent zero-field cooling (ZFC-T) magnetic susceptibility measurements. Samples were cooled down to 5 K via the cooling system at zero fields. Then an external field of 100 Oe was applied and the samples were heated to 300 K at the rate of 2 K/min. The effective magnetic moment (μ_eff) was obtained according to the equation, ${\mu }_{{\mbox{eff}}}=\sqrt{8{\chi }_{{\mbox{m}}}T}=\sqrt{n(n+2)}$, to calculate the number of unpaired electrons (n).

Measurement of the POD-like activity

The POD-like activity assays of PBNPs were performed at room temperature using H₂O₂ and TMB as substrates in HAc-NaAc buffer solution (0.2 M, pH 3.6). The absorbance at 650 nm of the color reaction was measured at specific time intervals over a 5-min period by a microplate reader (TECAN, Infinite M200) to determine the POD-like activity. The reaction mixture consisted of PBNPs solution, TMB (10 mg/mL), H₂O₂ (30%), and pH 3.6 buffer solutions in a volume ratio of 1:2:4:20. The specific activity (a_nano) was calculated according to ${a}_{{\mathrm{nano}}}=\frac{V\times \Delta A/\Delta t}{{m}_{{Fe}}\times \varepsilon \times l}$, where V is the total volume of the reaction system, ΔA/Δt is the initial slope of absorbance change at 650 nm, ε is the molar absorption factor of TMB (39,000 M⁻¹ cm⁻¹), and l is the optical path of light traveling in the liquid (~0.73 cm), m_Fe is the total mass of Fe (mg) contained in the reaction system.

Steady-state kinetic assays

Kinetic data were obtained using the same conditions in the POD-like activity measurement, but the concentration of H₂O₂ or TMB was varied. The kinetic parameters were calculated from the Michaelis-Menten equation: $v=\frac{{v}_{\max }\left[S\right]}{{K}_{m}+\left[S\right]}$, where v is the reaction velocity at the beginning, calculated by equation $v=\frac{\Delta A/\Delta t}{\varepsilon \times l}$, [S] is the substrate concentration, K_m is the Michaelis-Menten constant, and v_max is the maximum reaction velocity. k_cat was calculated by the equation ${k}_{{cat}}=\frac{{v}_{\max }}{[{Fe}]}$, where [Fe] (mM) was the concentration of PB catalysts in the reaction system.

Electrochemical HPRR measurements

Linear sweep voltammetry performance was conducted in a standard three-electrode system. Ag/AgCl, Pt wire, and the modified glassy carbon electrode (GCE) were applied as a reference, counter, and working electrodes, respectively. The catalyst ink was prepared by dispersing 100 μL PBNPs (15 mg/mL) and 100 μL 5% Nafion solution in 800 μL Ethanol. Then the 8 μL ink was drop-casted onto a polished GCE surface and dried in air. HPRR were carried out in 0.2 M HAc-NaAc buffer (pH 3.6) at a scan rate of 20 mV/s. Before the test, N₂ was pumped into the electrolyte for 10 min to remove O₂. Then 500 μL H₂O₂ (30%) was added to the electrolyte to start the test.

High-performance liquid chromatography analysis

The concentration of PMSO and PMSO₂ was quantified by Shimadzu LC-20A high-performance liquid chromatography. PMSO and PMSO₂ were separated by a Water C18 column (4.6 × 250 mm, 5 µm; XBridge), with a flow rate of 0.2 mL min⁻¹ (Detection wavelength: 230 nm for PMSO and 215 nm for PMSO₂).

Radical identification

PTA was used as a ·OH probe. The ·OH oxidation products of PTA emit a fluorescence signal at 450 nm when excited at 315 nm. A mixture of 800 μL HAc-NaAc Buffer (pH 3.6), 40 μL PBNPs (5 μg/mL), 80 μL H₂O₂ (30%), and 80 μL PTA (10 mM) was incubated at 37 °C for 4 h. Fluorescence spectrometer was used to measure the fluorescence signals after the reaction. The EPR spectra were also applied to explore the ROS generation during the activation of H₂O₂ by the catalysts. DMPO was used as a spin trapping reagent to detect ·OH and ·O₂⁻. The ·OH measurements were conducted as follows: 27.3 µL HAc-NaAc buffer (0.2 M, pH 3.6), 2.7 µL PBNPs (250 µg/mL) and 30 µL DMPO aqueous solution (0.2 M) were mixed. Ten µL of H₂O₂ (15%) was added to initiate the reaction. The mixture was immediately aspirated into a capillary tube for measurement. The conditions of ·O₂^- measurements were similar to ·OH except that the buffer and DMPO dissolution reagent were replaced with methanol.

Dissolved oxygen detection

Oxygen generation capacity of PBNPs was detected by measuring dissolved oxygen concentration at room temperature on Multi-Parameter Analyzer (DZS-708, Cany). 0.1 mL 50 μg/mL PBNPs solution was mixed with 8 mL HAc-NaAc buffer (0.2 M, pH 3.6). Then 1 mL 30% H₂O₂ solution was added to the above mixture to trigger the reaction. Oxygen concentration was recorded every 20 s, continuing for 400 s.

In situ Raman measurements

The in situ Raman test was performed in a 50 μL reaction cuvette under a dark field (Excitation wavelength: 638 nm, integration time: 20 s). The reaction conditions were as follows: 32 μL HAc-NaAc buffer (0.2 M, pH 3.6), 4 μL PBNPs (0.4 mg/mL), and 4 μL ABTS solution (20 mg/mL) were added to the cuvette and mixed uniformly. Finally, the reaction was initiated by 4 μL of H₂O₂ (3%) and the spectra signals were acquired every minute for 5 min.

Cesium adsorption experiment

Two mL of PB-0 (250 μg/mL) were mixed with 50 mL of CsCl solution (1 mM). Leaching solutions were sampled at 1, 2, 3, 6, 9, 12, 24, 36, 48, and 60 h for the detection of K⁺ and Cs⁺ concentrations by ICP-MS. The collected liquid was centrifuged three times and filtered by 0.22 μm membrane to completely remove the nanoparticles. PB-0 after Cs adsorption were centrifuged and washed three times by DI water for subsequent activity detection.

Measurement of contaminant degradation

MB degradation experiments were conducted at room temperature in HCl solution (pH 4 ± 0.2). The initial concentration of each component in the reaction system is: [Dye] = 0.2 mg/L, [H₂O₂] = 500 mM, [Catalyst] = 2 mg/L. For MB, RhoB and MO, absorbance changes at 665 nm, 554 nm and 500 nm were measured to determine concentration changes for different times. The C/C₀ value was obtained by normalizing to the initial absorbance after deducting the background. k_obs was calculated according to the equation $-{\mathrm{ln}}\left(\frac{C}{{C}_{0}}\right)={k}_{{\mbox{obs}}}t$, where t (min) is the reaction time. After the reaction, the PBNPs were removed by centrifugation and the MB degradation products were tested by MALDI-TOF-MS. During each catalytic run, 1 mL solution was collected at 20th, 40th, and 60th min. Nanoparticles were removed by centrifugation, and the resulting supernatant was diluted 10-fold with 1% nitric acid prior to measuring leached Cs concentrations via ICP-MS.

Cells

4T1 and 3T3 cell lines were obtained from the Cell Bank of Shanghai Institute of Biochemistry and Cell Biology, Chinese Academy of Sciences (Shanghai, China), and cultured with high-glucose Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 1% penicillin-streptomycin and 10% fetal bovine serum (FBS) at the condition of 37 °C with 5% CO₂ and humidified atmosphere.

In vitro cytotoxicity assay

The 4T1 or 3T3 cell lines were seeded in a 96-well plate, cultured in a standard state for 24 h, and replaced with fresh medium and NPs with the corresponding concentration gradient. The final concentration gradients were 10, 20, 30, 40, 50 μg/mL for PB-0 and CsPB-1 (determined by Fe) and 15, 30, 45, 60, 75 μg/mL for CsCl (determined by Cs). Medium was then pipetted, washing with PBS was performed, and 200 μL serum-free medium and 10 μL CCK-8 solution were added. Culture was continued for an additional 1 h, and the absorbance of each well was measured at 450 nm with a microplate reader.

ROS detection

For the observation of intracellular ROS yield, 4T1 and 3T3 cell lines were seeded in a confocal dish, cultured in a standard state for 12 h, and replaced with fresh medium and NPs. The final concentration was 30 μg/mL (determined by Fe). Incubation continued for 12 h. The cells were then stained with DCFH-DA reactive oxygen species probe and Hoechst 33342, and the yield of reactive oxygen species (ROS) was observed under confocal microscopy. The mean fluorescence intensity was calculated by ImageJ. The significance between the two groups was analyzed by two-tailed Student’s t test.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.