## Introduction

Artificial enzyme engineering with sustainabilities emerges as a versatile methodology for creating biocatalysts1,2,3,4,5,6. The intrinsic feature of continuous electron transfers allows artificial enzymes to maintain high catalytic stability and arbitrarily tailor their characteristics7,8,9,10,11,12. The superior catalytic activity over the naturally-occurring enzymes is one of the major challenges existing in designing artificial enzymes13,14,15,16. A feasible strategy is to construct an artificial active center by imitating the biologically active center of enzymes or proteins2,17,18,19,20,21,22. As a fast growing category of artificial enzymes, the emerging nanozyme can actively tailor biocatalytic activities and selectivity via its flexible atomic structures and molecular engineering23,24,25,26. State-of-art nanozymes with an M-N4 center inspired by cytochrome P450 have superior reaction rates and high substrate affinities close to the natural horseradish peroxidase (HRP)27,28,29. Expanded Fe-N5 and Fe-N3-P were reported to have 17- and 2-fold higher catalytic efficiencies than Fe-N414,30, indicating their great potential in boosting bioactivities. However, the development of nanozymes with properties outperforming natural enzymes remains unfulfilled and highly challenging.

Other unresolved and high-profile questions are the underlying reaction mechanisms and detailed atomic coordination structures during biocatalytic reactions to endow catalytic selectivity13,31,32,33. Unlike the well-defined structures of single-atom catalysts16,34,35, traditional nanostructures without elucidated atomic structures on the surface cannot provide sufficient structure-function correlation to study the electron transfers at atomic levels36. For example, Pt and Fe3O4 exhibit peroxidase-like (POD-like) activities6,37; Au and V2O5 are more selective for the glutathione peroxidase-like (GPx-like) activity38,39,40; Cu demonstrates a superior superoxide dismutase-like (SOD-like) activity41. However, the selectivity of these biocatalysts merely relies on the specific atomic coordination and surface ligands of individual nanosystems, making it difficult to corroborate a universal catalytic mechanism explicitly. Besides, the complexity of reaction pathways also compromises catalytic reactions of nanozymes33, leaving the atomic configuration and bond morphology inadequately clarified during the dynamic reaction process.

In this work, we report structurally well-defined and atomically-precise biocatalysts of RhN4 and VN4 with high POD- and catalase (CAT)-like activities, surpassing corresponding natural enzymes and reported FeN442,43,44. Moreover, the RhN4 and VN4 favorably form the Rh/V-O-N4 structure as the active center of circular catalytic processes to decrease reaction energy barriers significantly, which is fundamentally different from FeN414,45,46,47 (Fig. 1). In particular, the catalytic mechanism of RhN4 and VN4 modulates a unique “two-sided oxygen-linked” catalytic reaction path, resulting in high catalytic efficiency. Meanwhile, the specific SOD activity is derived from Fe-Cu-N6, and the GPx-like activity of VN4 considerably exceeds that of the natural GPx. Furthermore, the high catalytic stability and recyclability enable these nanozymes to be used as medical sutures for sustained scalp healing from brain trauma via a serial biocatalytic process. RhN4 and VN4 in sutures play an essential role in the proactive regulation and immune control of macrophages and vascular endothelial growth factors (VEGF).

## Results

### Structural characterization of MNx

The MNx (M = Rh, V, Fe, Cu, and Fe-Cu; x = 4 or 6) nanozymes were synthesized via pyrolysis14,48,49. RhN4 and VN4 preserved the pristine polyhedral structure of zinc-imidazole frameworks (ZIF-8) as visualized by high-resolution transmission electron microscopy (HRTEM), while Fe-Cu-N6 exhibited an amorphous morphology with abundant micropores, similar to FeN450 (Fig. 2a, d, g). In addition, aberration-corrected high-angle annular dark-field scanning transmission electron microscopic (AC-HAADF-STEM) images showed single metal atoms sparsely dispersed in the ZIP-8 framework (Fig. 2b, e, h). The Fe-Cu diatomic pairs could be observed in the enlarged magnified AC-HAADF-STEM image of Fe-Cu-N6 (Supplementary Fig. 1a). The average distances between Fe and Cu atoms were calculated as 2.27 Å from the statistical intensity profiles (Supplementary Fig. 1b). For energy-dispersive X-ray spectroscopy (EDS) mapping analysis, an individual polyhedron was selected. Single metal atoms and C and N elements are uniformly distributed throughout the entire nanostructure (Fig. 2c, f, i, and Supplementary Fig. 2). Characteristic peaks of metal oxides and atomic crystallites were not detected in the ultraviolet-visible (UV-vis) absorption spectra, X-ray diffraction (XRD), and Raman spectra, indicating the absence of metal nanoparticles (Supplementary Figs. 35). X-ray photoelectron spectroscopy (XPS) identified major nitrogen species as pyridinic, graphitic, and pyrrolic N, while carbon species were primarily composed of graphitic C, C=C-N, and N-C=N (Supplementary Figs. 68). The elemental composition of metal-based active sites was investigated by inductively coupled plasma-mass spectrometry (ICP-MS). The mass percentages of Rh and V in Rh/VN4 nanozymes were only 0.19 and 0.22 %, while Fe, Cu, and Fe-Cu in FeN4, CuN4, Fe-Cu-N6 nanozymes were elevated at 1.4, 1.39, and 1.1 %, respectively (Supplementary Table 1).

Next, we characterized the atomic coordination structure between metal atoms and neighboring atoms by the X-ray absorption near-edge structure (XANES), along with the Fourier transformed (FT) extended X-ray absorption fine structure (EXAFS). By analyzing the XANES of RhN4 (Fig. 2j), we can observe that the Rh species become more positively charged than Rh(acac)3, which may originate from the Rh-N4 unit. Furthermore, VN4 showed a characteristic peak at 1.58 Å, shorter than the bulk V-V bond of 2.29 Å, suggesting the dispersed state of individual V atoms (Fig. 2k and Supplementary Fig. 9). Similarly, the major peaks of Fe and Cu atoms for Fe-Cu-N6, FeN4, and CuN4 were located at 1.47 and 1.48 Å (Fig. 2l, Supplementary Figs. 1013, and Supplementary Table 2), corresponding to Fe-N and Cu-N bonds, respectively51,52,53,54,55. To characterize the exact Fe-Cu coordination, we thermodynamically and kinetically predicted five possible structures of Fe-Cu-Nx based on the above findings, defined as Fe-Cu-N6, Fe-Cu-N8-I, Fe-Cu-N8-II, Fe-Cu-N8-III, and Fe-Cu-N8-IV (Supplementary Fig. 14). The simulated distance of Fe-Cu-N6 was 2.46 Å (Supplementary Table 3), which closely matched the experimental result of 2.27 Å. Fe-Cu-N6 was identified as the most thermodynamically favorable structure due to the lowest formation energy of −5.432 eV by the Density Functional Theory (DFT) simulation (Supplementary Table 3). Therefore, all of the subsequent calculations of the enzyme-like pathways were based on the Fe-Cu-N6 structure.

### POD-like catalytic activities of MNx

We assessed the POD-like activities of MNx by monitoring the time-lapse reaction rates at equivalent concentrations using the 3,3′,5,5′-tetramethylbenzidine (TMB) colorimetric assay (Fig. 3a), which were correlated well with substrate concentrations (Supplementary Figs. 15, 16). The specific activities of MNx were quantified by referring to the activity curve of HRP standards. RhN4 and VN4 were 4.37 and 2.22 U/µmol in POD-like activities, about 8 times higher than the widely reported FeN4 (0.514 U/µmol) and CuN4 (0.286 U/µmol)47 (Fig. 3b). In addition, we continuously monitored the enzymatic activity for 1 month. All MNx displayed outstanding long-term stability with negligible decays, while the natural HRP rapidly lost its activity within 2 days (Fig. 3c). Besides, significant intrinsic kinetic differences between MNx were observed, including their Michaelis–Menten constants (Km) and catalytic constants (kcat). The catalytic reaction affinities of RhN4 (Km 55.7 µM) and VN4 (Km 65.6 µM) to the TMB substrate were 4–5 times better than those of FeN4 (Km 184.6 μM) and the natural HRP (Km 276.2 μM), denoting their excellent catalytic characteristics (Supplementary Table 4, Supplementary Figs. 17, 18). The kinetic parameters of MNx for H2O2 substrates were also quantified (Supplementary Figs. 19, 20, and Supplementary Table 5). Compared to the reported Km values of single-atom catalysts42,44,47,56, MNx showed a superior binding affinity for H2O2, confirming the outstanding catalytic properties. However, all single-atom nanozymes exhibited lower binding affinities for H2O2 substrates than HRP, which remains an unconquered challenge that would need further improvements.

It is worth noting that, unlike most previous studies27,45,46, the POD-mimicking reaction pathway simulated in this work is not accompanied by the generation of hydroxyl radicals (OH). Instead, the OH-producing catalytic reaction pathways were still simulated. As shown in Supplementary Fig. 23 and Supplementary Table 7, there are no significant differences in the energy barriers of FeN4 between the two paths at 0.539 eV and 0.405 eV, suggesting the existence of reaction pathways both with and without generated OH. However, all the other catalysts except FeN4 display significant differences in energy barriers between the two reaction paths, especially RhN4 and VN4, indicating the preference for the pathway without OH generation. We utilized electron spin resonance (ESR) spectroscopy to verify further whether MNx can scavenge OH in the inflammatory physiological environment (pH = 6.0–6.5). It turned out that MNx effectively scavenged OH (Supplementary Fig. 24). Therefore, exceedingly different from the widely reported FeN4, the catalysis of RhN4 and VN4 follows unique paths and mechanisms, forming the Rh/V-O-N4 active center by preference and then activating the circulatory catalytic reaction cascade.

### CAT-like activities of MNx

Meanwhile, the intrinsic CAT-like activities of MNx were evaluated by monitoring the decomposition of H2O2 at the concentration of 4.8 μM. Amongst all catalysts, RhN4 showed the optimal catalytic performance with a clearance rate of H2O2 up to 100 % within 40 min (Fig. 4a and Supplementary Fig. 25). Furthermore, the CAT-like activities of RhN4, Fe-Cu-N6, and VN4 were quantitatively determined to be 5177, 1510, and 621 U/µmol. In particular, the activity of RhN4 was 5 and 15 times higher than the widely reported FeN4 (1044 U/µmol) and CuN4 (313 U/µmol) (Fig. 4b). Likewise, the stability of the CAT-like activity for RhN4 was confirmed for over a month at room temperature, whereas the natural CAT activity was completely inactive after 14 days (Fig. 4c). Kinetic analysis revealed that RhN4 had a high affinity to the H2O2 substrate with a Km of 1.33 mM, about 20 times more superior than the natural CAT (Km 28.8 mM) (Supplementary Figs. 26, 27 and Supplementary Table 8). Therefore, MNx is potentially useful as an effective CAT mimetic to defend against oxidative stress in organisms during disease development.

Compared to the POD-like processes, simulated CAT-like processes are generally less complex (Fig. 4g and Supplementary Fig. 28). Without extra hydrogens, the initialization steps, including the “initialized” states and bilateral adsorption steps, are converged into the catalytic pathways of all catalysts except CuN4. The “initialized” Fe, V, and Rh in the MNx limit the loop within the bilateral adsorption step. In Fe-Cu-N6 with two metallic centers, the bilateral adsorption step is generalized into special adsorption on different metallic atoms. The first and second states with a single oxygen atom attached are directed to Fe (Fig. 4g row 2 col 3) and Cu (Fig. 4g row 2 col 7), respectively. FeN4 and Fe-Cu-N6 follow the “finalization” steps in which the catalysts are restored to their initial states. Besides, DFT simulations predict activation energies of RhN4, Fe-Cu-N6, VN4, FeN4, and CuN4 to be 0.213, 0.291, 1.10, 0.539 and 1.507 eV, respectively (Fig. 4d). The RDSs associated with the energy barriers of the activation energies disclose the reaction mechanisms in-depth: RhN4 breaks down the second H2O2 molecule with difficulty; VN4 and FeN4 have relatively low attractions towards the second H2O2 molecule; Fe-Cu-N6, on the other hand, has trouble in releasing the first water molecule; CuN4 can hardly extract another oxygen atom from an H2O2 molecule on the other side. The intermediate states of Rh/VN4 and Fe-Cu-N6 during catalytic cycles are consistent with those observed in the POD-like reaction pathways (Fig. 4e, f).

### SOD-like activities of MNx

We then studied the SOD-like activities of MNx under the same conditions, using the classical nitro-blue tetrazolium (NBT) chromogenic approach. CuN4 and Fe-Cu-N6 inhibited formazan production by scavenging O2•– (Fig. 5a). The inhibition rate of SOD was verified via the SOD-mimic reaction assay of MNx at various concentrations (Fig. 5b). The SOD-like activities of CuN4 and Fe-Cu-N6 were quantitatively calculated to be 1656 and 1344 U/µmol (Fig. 5c), which were attributed to the low energy barriers predicted by DFT simulations (Fig. 5d). Therefore, the high SOD-like activity of Fe-Cu-N6 can be ascribed to its Cu-N4 coordination structure, similar to that of CuN4. Overall, given that Fe-Cu-N6 possesses both POD-like and CAT-like activities of FeN4, along with the SOD-like activity of CuN4, the MNx can practically achieve multienzyme-mimetic activities by integrating different metal-based active sites.

Besides, we conducted the classic glutathione reductase (GR) coupling assay and found that only VN4 had substrate concentration-dependence for the GPx-like activity, which was quantified to be 47.9 U/µmol (Supplementary Fig. 30). The reaction affinity of VN4 (Km 1.31 mM) to glutathione (GSH) substrate was nearly 7-fold better than that of the natural GPx (Km 9.23 mM) (Supplementary Fig. 31 and Supplementary Table 9). Furthermore, the simulated GPx-mimetic reaction acquired an energy barrier of 3.28 eV during the GPx catalysis (Supplementary Fig. 32). Moreover, to exclude the interference from Zn atoms on the catalytic performance of MNx, the multiple enzyme-like activities of ZIF-N/C only anchored by Zn atoms were also tested (Supplementary Fig. 33). The ZIF-N/C showed a negligible catalytic activity, validating that the ultrahigh catalytic properties predominantly originate from the M-Nx active centers. In addition, the Donor Fukui Function (defined in Eq. 37) manifests that the electrophilic reactivities of the four central metals are significantly different, implying different charge-donating abilities of active sites on RhN4, VN4, and Fe-Cu-N6 (Supplementary Fig. 34)

### MNx-based sutures for scalp healing in traumatic brain injuries

Inspired by the enzyme-mimicking activities of MNx theoretically and experimentally demonstrated above, we developed MNx sutures by integrating MNx into polyglycolic acid (PGA) sutures and applied them to the scalps of living mice with traumatic brain injuries (TBI) to study their potential healing effects. The field-emission scanning electron microscopic (FE-SEM) images and ICP-MS showed that MNx was successfully incorporated into the suture structures (Supplementary Figs. 35, 36 and Supplementary Table 10). Meanwhile, the MNx sutures retained the original high enzyme-like activities, with consistent catalytic performances (Supplementary Fig. 37). Furthermore, as demonstrated by XRD and Raman spectroscopy, the characteristic peaks of graphite carbon were well preserved after modification of sutures with MNx, despite the interference from the inherent PGA background signal (Supplementary Figs. 38, 39). Additionally, signals of aggregated metal nanoparticles were not observed, suggesting the single-atomic distribution in the suture framework available for efficient catalysis.

To explore the medical practicability of MNx, we employed mouse models of TBI with skin wounds to investigate the in vivo wound-healing effects. Wound healing is a complex physiological process divided into four stages: hemostasis, inflammation, proliferation, and tissue modeling, which have complicated and dynamic interactions associated with time. Figure 6a depicts the schematic illustration for the wound closure process modulated with MNx stitches. We hypothesized that the slow healing of scalp injuries using pristine PGA sutures could be accelerated by using MNx sutures by inhibiting the secretion of inflammatory cytokines and harmful molecules. As shown in Fig. 6b, c, and Supplementary Fig. 40, MNx-modified surgical sutures successfully seamed the wound with time. In addition, the wound size and area were significantly diminished and healed to healthy levels with MNx sutures, whereas the untreated TBI mice only showed a partial recovery of about 80%. As shown in Fig. 6d, Supplementary Figs. 41, 42, indicators of malondialdehyde (MDA), SOD, GSH, and H2O2 in the scalp tissues were relatively severe in the TBI group. By contrast, the decrease in SOD and GSH levels from scalp injuries could be rescued by MNx sutures with prominent recoveries on day 3 after treatment. Comparatively, VN4 induced a better recovery in GSH than RhN4 and Fe-Cu-N6, which correlated well with their in vitro GPx-like activities. Lipid peroxides and H2O2 can continuously accumulate in the scalp following injuries as the by-products of oxidative stress, resulting in severe damage. All RhN4, VN4, and Fe-Cu-N6 considerably inhibited the production of these harmful molecules and maintained their stable POD-like and CAT-like activities in the long term.

Excessive oxidative indicators, such as proinflammatory cytokines and macrophages, can prolong the inflammation duration and jeopardize the transition into the proliferative phase57,58. From enzyme-linked immunosorbent assays (ELISA), proinflammatory cytokines such as interleukin (IL)-6, IL-1β, and tumor necrosis factor-α (TNF-α) were significantly upregulated following wound injuries, indicative of potent local inflammation (Fig. 6e and Supplementary Fig. 43a–c). Fortunately, RhN4 and VN4 could sharply downregulate IL-6, IL-1β, and TNF-α levels, with a notable anti-inflammation effect. In comparison, Fe-Cu-N6 only showed negligible regulation of IL-6 and IL-1β due to its weaker catalytic activity. Although PGA sutures could gradually suppress the inflammatory cytokines through autoimmunity, MNx treatment essentially restored the cytokines to normal levels, suggesting its role in suppressing neuroinflammation. As shown in Fig. 6f, g, Supplementary Figs. 4447, the population of M1 macrophages decreased and was polarized into the M2 phenotype, further confirming the anti-inflammatory effects of MNx sutures. In addition, changes in other immune cells like natural killer cells, B cells, and T cells after various treatments in whole blood were also analyzed by flow cytometry. As shown in Supplementary Fig. 48, mice could induce systemic immune responses after brain injuries and gradually recovered to normal with MNx-based sutures. Besides, the tissue macrophage marker CD68 was upregulated after scalp injuries but recovered by MNx sutures, while the vascular epidermal growth factor (VEGF) was also ameliorated by MNx sutures (Fig. 6h–j, Supplementary Figs. 43d, 49). VEGF is known to benefit vascular endothelial cell migration and angiogenesis. VN4 and Fe-Cu-N6 exhibited marginally improved effects on VEGF than RhN4, with potentially higher healing efficacy. Moreover, the vascular endothelial-specific marker CD31 indicated neovascularization in the injured sites. Compared with other groups, more neovascularization was observed for VN4 and Fe-Cu-N6 than RhN4 by confocal microscopy, which directly correlated with the VEGF expressions (Fig. 6i, k, and Supplementary Fig. 50). In the early stages of wound healing, the transition from the inflammatory stage to the proliferative stage was accelerated by MNx sutures, with granulation tissues observed in the treatment groups, indicating MNx can competently mediate a beneficial wound healing process (Supplementary Fig. 51)59,60. Accordingly, we further investigated the effects of MNx sutures on oxidative stress and inflammation in TBI. All indicators exacerbated towards intense local inflammation and oxidative stress following TBI. However, MNx could sharply reduce harmful molecules and proinflammatory cytokines and oppositely rescue beneficial molecules in the brain (Supplementary Figs. 52, 53). Immunofluorescence (IF) staining of the cerebral cortex showed that MNx could remarkably decrease TBI-elevated expressions of IL-1β and TNF-α, inhibit microglial activation, rescue neurons (Supplementary Figs. 54, 55), and restore the body weight loss (Supplementary Fig. 56). In addition, no metal elements were detected in the blood of MNx-treated mice on day 6 after injuries, suggesting that the MNx-based sutures are biologically safe (Supplementary Fig. 57).

## Discussion

Rational design and practical exploration of nanozymes fulfill the unmet need for sustained and efficient enzymatic catalysis, a category that naturally derived enzymes have not yet covered. Currently, the inferior catalytic activities of nanozymes are primarily assigned to the low efficiency of atom utilization and scarcity of locally active sites. In the current work, MNx nanozymes enable precise modulation of the central coordination atom to improve overall activities catalytically outperforming natural enzymes. Besides, we identified a unique mechanism of “bilateral adsorption” from which the high enzyme-mimicking catalytic activity originated. During this process, the X*-M-X* (X = O, OH, O2•–) bilateral adsorption structure first formed and then reduced the activation energies, facilitating fine-tuning the coordination microenvironment of the central metal atom, which is necessary to boost the entire catalytic cycle. Hence, we deduced the X*-M-X* (X = O, OH, O2•–) intermediate state as the cyclic block of catalysis with a proposed “bilateral adsorption” mechanism.

As is known, the catalytic activities of naturally-occurring enzymes are unstable and would decay within 2 days, while MNx nanozymes behave in a comparatively stable manner without any evident decay after 1 month (Figs. 3c, 4c, and Supplementary Fig. 58). Another critical challenge is that natural enzymes can hardly be repeatedly used. They are efficient and versatile as biocatalysts in nature but suffer from environmental disturbance, poor catalytic stability, and low sustainability, making them limited for widespread use under current circumstances. By comparison, the catalytic activities of MNx nanozymes were barely attenuated during repetitive catalytic processes (Supplementary Fig. 59). All these advantageous features justified MNx as a promising enzyme substitute in the case of continuous biocatalysis. As shown in the schematic wound closure process (Fig. 6a), scalp injuries from brain trauma are physiologically complex, accompanied by complicated and dynamic interactions strongly associated with time, and can continuously trigger a cascade of molecular and biochemical events in vivo, resulting in severe inflammation and oxidative stress59,61. Typically, the scalp injuries with PGA sutures only recovered slowly from the inflammatory phase to the tissue modeling. Nevertheless, the as-developed MNx nanozymes with definitive molecular structures circumvent all such obstacles and accelerate the healing process with recyclable and catalytic features, remedying the limitations of natural enzymes.

In summary, we developed single-atom nanozymes of RhN4, VN4, and Fe-Cu-N6 with M-N4 active centers and achieved efficient multienzyme-mimetic catalysis with good selectivity by maximizing atom utilization and tuning substitution atoms. RhN4 and VN4 preferentially exhibited high POD-like activities with Km of 55.7 and 65.6 μM to the TMB substrate, 5-fold lower than the natural enzyme counterpart HRP (Km 276.2 μM). In contrast, RhN4 (Km 1.33 mM) and Fe-Cu-N6 (Km 1.99 mM) showed 20 and 14 times higher affinities to H2O2 substrate than the natural CAT (Km 28.8 mM) during CAT-mimetic reactions. Further, Fe-Cu-N6 exhibited a high SOD-like activity of 1344 U/μmol, while VN4 possessed the unique GPx-like activity with an affinity to GSH (Km 1.31 mM), exceeding that of natural GPx (Km 9.23 mM). Meanwhile, DFT calculations pinpointed a unique bilateral reaction mechanism that reduced the reaction energy barriers and thus enhanced catalytic activities. Further, we medically applied MNx sutures that accelerated scalp healing from brain trauma via promoting VEGF and inhibiting oxidative stress and neuroinflammation.

## Methods

### Chemicals

Rhodium (III) 2,4-pentanedionate (Rh(acac)3, 98%) was purchased from Shanghai Xinbo Chemical Technology Co., Ltd. (China). Zinc nitrate hexahydrate (Zn(NO3)2·6H2O, 99%), hydrogen peroxide (30 wt % in H2O), Dimethyl sulfoxide (DMSO, >99.8%), and isopropanol (99.8%) were purchased from Shanghai Aladdin Biochemical Co., Ltd. (China). Vanadium (III)-2,4-pentanedionate (V(acac)3, 97%), 2-methylimidazole (99%), and 3,3′,5,5′-tetramethylbenzidine (TMB, 99%) were purchased from Meryer (Shanghai) Chemical Technology Co., Ltd. (China). Iron (III) chloride hexahydrate (FeCl3·6H2O, 99%), Cupric chloride (CuCl2, 98%), N, N-dimethylformamide (DMF, 99.5%), and HRP ( >300 U/mg), were obtained by Shanghai Macklin Biochemical Co., Ltd. (China). GPx (740 U/mg), CAT (4323 U/mg) were purchased from Sigma-Aldrich Company Ltd. NADPH tetracyclohexanamina (90%), GR (1164.8 U/mL) were purchased from Shanghai Yuanye Bio-Technology Co., Ltd. (China). GSH (97%) was purchased from Tianjin Xiensaopude Technology Co., Ltd. (China). SOD (5200 U/mg) was purchased from Beijing Solarbio Science & Technology Co., Ltd. (China). Deionized water (>99.5%), methanol (>99.5%), and ethanol absolute (>99.7%) were purified to analytical grade by Tianjin Yuanli Chemical Co., Ltd. (China). Except for special instructions, all experiments were carried out at room temperature.

### Preparation of MNx

The VN4 single-atom enzyme mimetics were synthesized by a one-pot method. Firstly, 0.8733 g of 2-methylimidazole was added to 20 mL of methanol to form a transparent solution. Next, the 77.3 mg of V(acac)3 and 0.8032 g of Zn(NO3)2·6H2O were dissolved in a beaker containing 10 mL of methanol to form the dark green solution. The two solutions were mixed and stirred vigorously. Then, the mixture was poured into a 50 mL high-pressure reactor lined with polytetrafluoroethylene, heated instantly to 120 °C, and kept for 4 h. The gray precipitate was collected and then naturally cooled to room temperature. It was then purified several times with DMF and anhydrous methanol, respectively, and V-doped ZIF-8 (V/ZIF-8) was obtained. Next, the dried V/ZIF-8 precursor powder was heated to 900 °C at a rate of 5 °C/min in a tube furnace under nitrogen gas flow and kept for 3 h. The resulting black powder was the VN4. Similarly, RhN4 follows similar procedures, and the difference is the use of 88 mg of Rh(acac)3 instead of 77.3 mg of V(acac)3.

The FeN4 was prepared by a two-step method. First, to synthesize the carbon-based supporting substrate, 5.909 g of 2-methylimidazole was dissolved in a beaker A containing 450 mL of methanol. Next, 5.085 g of Zn(NO3)2·6H2O was uniformly dissolved in another beaker with 500 mL of methanol and poured into beaker A. The mixture was rapidly stirred until it turned cloudy white and placed in a drying oven at 60 °C for 24 h. The white precipitate was collected by centrifugation and washed five times with absolute ethanol to obtain transparent ZIF-8 nanocrystals. After drying overnight in a vacuum, the nanocrystals were pyrolyzed at 900 °C for 2 h in an annealing furnace under nitrogen gas flow. The final product was the N-doped carbon framework with a rich pore structure, marked as ZIF-C/N.

The preparation of FeN4 and CuN4 was obtained by activating ZIF-C/N that was embedded with different metal ions at high temperatures. Typically, 40 mg ZIF-C/N powder was dispersed in isopropanol (4 mL) containing 2 mg FeCl3. The mixture was then sonicated and stirred vigorously for 2 h, resulting in Fe3+ adsorbed in the ZIF-C/N. Then, the ZIF-C/N with Fe3+ was activated in a tube-type furnace at 500 °C under a nitrogen atmosphere for 1 h, and the final product was the single-atom FeN4. CuN4 followed similar procedures, and the difference was the use of 1.5 mg CuCl2 instead of 2 mg FeCl3.

Similarly, the synthetic procedure of Fe-Cu-N6 was N2-annealing (500 °C) activation of ZIF-C/N co-embedded with Fe3+ and Cu2+ ions, in which the feeding of metal ions was 1 mg FeCl3 and 0.75 mg CuCl2.

### Preparation of catalyzed MNx sutures

The PGA sutures were purchased from Shandong Haidike Medical Products Co., Ltd. Each PGA suture was individually fixed on a homemade iron frame and then immersed in an isopropanol solution of MNx for 3–5 days so that MNx was uniformly loaded on the sutures. Meanwhile, the MNx solution was sonicated once every 4 h during the soaking procedure until the MNx was evenly dispersed. Last, the MNx sutures were dried overnight in a vacuum oven for subsequent characterization and activity testing.

### Characterization

The morphology and size of MNx were observed by bright-field TEM and corresponding EDS element mapping images on two field-emission transmission electron microscopes (JEM-F200 and JEM-2100F, JEOL, Japan) with an accelerating voltage of 200 kV. Next, the images were taken on a cooled emission spherical aberration-corrected atomic resolution microscope (JEM-ARM200F, JEOL, Japan) running at a resolution of 0.08 nm and a working voltage of 200 kV. Finally, the samples were scattered on the micro-grid carbon grid.

The XAFS spectra were obtained at the 1W1B station (2.5 GeV, maximum current of 250 mA) of the Beijing Synchrotron Radiation Facility. The ATHENA and ARTEMIS software were used to analyze and fit the XAFS results. All samples’ ultraviolet-visible (UV-vis) absorption spectra were tested on the UV-VIS NIR instrument (Shimadzu, Japan). The micro-quartz cuvette with a light path of 1 cm was utilized to contain the solution. The crystal structure of the precursor, the final MNx samples, and MNx sutures were researched by X-ray diffraction spectroscopy (XRD) scanning on an X-ray powder diffractometer (Smartlab, Rigaku, Japan) radiated by Cu Kα. The Raman spectra of carbon-based MNx and PGA sutures loaded with MNx were measured by an ultra-fast Raman imaging spectrometer (XploRA PLUS, Horiba-JY, France) with the scanning wavenumber range of 500  3800 cm−1. The valence states of different elements were detected by an X-ray photoelectron spectrometer (ESCALAB Xi+, Thermo Fisher Scientific, UK) excited by Al Kα, in which powder samples were pressed into discs. The original data were peak-fitted in the XPSPEAK41 software. SEM and elemental mapping images were collected using a field-emission scanning electron microscope for the MNx-based PGA sutures (S-4800, Hitachi, Japan). The ICP-MS measurement (7900 ICP-MS, Agilent, UK) was employed to characterize the content of metal elements in MNx and MNx -sutures.

### The POD-like activity test

The quantitative detection of the POD-like activities of the MNx was tested by colorimetric method using a TMB substrate color development kit (TELISA, SenBeiJia). During the experiment, the working solution was prepared according to the instructions. First, MNx samples (20 µL) and the working solution (180 µL) were added to the 96-well plate. Then, the absorbance changes at 652 nm were monitored over time through a microplate spectrophotometer (CMax Plus, Molecular Devices). Meanwhile, the different concentrations of HRP were selected as the standard. At last, the activity value of natural HRP and the corresponding A652 were plotted and fitted in the Origin software to obtain a standard curve. Then, based on the standard curve, the POD-like activity of the MNx was quantitatively calculated according to A652.

The kinetic test of the interaction between the MNx and TMB substrate was carried out using a UV-vis spectrophotometer under the kinetic mode to record the increase at A652 of the reaction solution over time. In a typical assay, the different concentrations of TMB (0–800 µM, dissolved in DMSO) as varying substrate, H2O2 solution (50 mM), and MNx (or HRP) were added to sodium acetate buffer (PH = 4.5) to maintain the volume of the mixture at 200 µL. Then, the increase at A652 was immediately recorded to measure the corresponding initial rate. The maximum reaction velocity (Vm) and Km were determined by plotting the initial rate and the substrate concentration using Microsoft Excel. Curves were fitted to the Michaelis–Menten curve on the Origin software. The kcat represents the catalytic efficiency of MNx per molar concentration, and its value was obtained by dividing Vm by the concentration of MNx (E) during the test. The concentration of MNx was expressed as the total concentration of active metal sites measured by ICP-MS.

In addition, kinetic analysis between MNx and H2O2 substrate was studied by measuring the changes at A652 in the reaction mixture with different concentrations of H2O2 (0–50 mM) at a fixed TMB (800 µM) concentration.

### The CAT-like activity assay

The CAT-like activity of MNx was assessed by measuring the decomposition rate of H2O2. In short, MNx of the same concentration (4.8 µM) was added to 40 µM H2O2 aqueous solution (300 µL), then the absorbance change at 240 nm over time was detected on a spectrophotometer. Moreover, a mixture of 100 µM H2O2 aqueous solution (380 µL) and MNx was incubated in a centrifuge tube for about 3 min. The picture containing abundant bubbles was captured, indicating that MNx catalyzed H2O2 to produce O2.

The quantitative detection of the CAT-like activity of the MNx was obtained by using the Catalase Assay kit (S0051, Biyuntian). One unit of the enzyme activity was defined as 1 µmol H2O2 decomposed within 1 min at 25 °C under neutral conditions (pH = 7).

Further, the kinetic analysis of MNx was performed using a Dissolved Oxygen Meter (HACH HQ40d, USA) with an LDO101 probe to measure the rate of oxygen generation catalyzed by MNx at different concentrations of H2O2 (0–5 mM). A Michaelis–Menten curve was established based on the O2 production rate and the H2O2 concentration and fitted to derive Km and Vm. For the kinetic test of the natural CAT, the decrease in absorbance at 240 nm was monitored at different concentrations of H2O2 (0–40 mM) in the presence of CAT using a UV-vis spectrophotometer in the kinetic mode. The molar extinction coefficient ε240 nm of H2O2 was calculated to be 39.4 M−1 cm−1.

### The SOD-like activity assay

The SOD-like activity of MNx was determined using the Total Superoxide Dismutase Assay Kit with NBT (R22261, Yuanye). In a typical assay, the SOD detection buffer (100 µL), NBT color developing solution (30 µL), an enzyme solution (30 µL), and different concentrations of MNx (20 µL) were added to a 96-well plate. Then, the reaction starter solution (20 µL) was added. The reaction mixture was irradiated under a fluorescent lamp for an appropriate amount of time, and the absorbance at 560 nm was measured immediately with a microplate spectrophotometer (CMax Plus, Molecular Devices). Simultaneously, a 20 µL SOD detection buffer was used instead of MNx as the blank control group (Acon1), and the other control group used an SOD detection buffer (20 µL) instead of MNx in the dark during the experiment as a light-proof control group (Acon2). The A560 was converted into the inhibition rate of the sample through the formula: inhibition rate (%) = (Acon1 − AMNx4) / (Acon1 − Acon2) × 100. The SOD-like activity of MNx was further calculated according to the formula: enzyme activity (units) = inhibition percentage / (1 − inhibition percentage), in which 50% of inhibition on NBT photochemical reduction was taken as an enzyme activity unit.

### The GPx -like assay

According to the previously reported glutathione reductase coupling reaction, the GPx-like activity of the MNx was evaluated by monitoring the decrease in NADPH content after MNx treatment. In the assay, the detection solution was composed of phosphate buffer (PBS, pH = 7.4), MNx solution (3.8 µM), reduced glutathione (GSH, 2 mM), glutathione reductase (GR, 1.7 Units/mL), NADPH (200 µM), and H2O2 (0.2 mM) at the total volume of 500 µL. When H2O2 was added, the change of DADPH at 340 nm (ε340 nm = 6.22 mM−1 cm−1) was immediately measured within 5 min using a UV-vis spectrophotometer operating in the kinetic mode. Furthermore, kinetic measurements of the interactions between MNx (or natural GPx) and GSH as a substrate were achieved at various concentrations of GSH (0–5 mM), while the H2O2 concentration was fixed at 0.2 mM. Conversely, the kinetic analysis of substrate H2O2 was conducted by changing the concentration of H2O2 with the fixed GSH (2 mM). Finally, the calculation process of Km and Vm is consistent with the method mentioned above.

The quantitative detection of GPx-like activities of MNx was carried out under strict conditions of 25 °C by using a total glutathione peroxidase detection kit (S0058, Beyuntian). One enzyme activity unit (1 unit) was defined as 1 µmol NADPH converted into NADP+ within 1 min catalyzed by MNx in the presence of GSH, GR, and organic peroxide (Cum-OOH) at 25 °C and pH 8.0.

### •OH scavenging test

The OH scavenging measurements were conducted on an ESR spectrometer (JES-FA200, JEOL, Japan). The nitrogen trap DMPO was used to trap OH to form DMPO-OH spin adducts. During the assay, OH was generated by UV illumination of a centrifuge tube containing DMPO (50 mM), H2O2 (2 mM), and acetate buffer (pH = 6.1) for about 20 min. Then, the mixture was transferred into a quartz tube for ESR measurement, and a clear 4-peak signal with an intensity of 1:2:2:1 was captured. Subsequently, MNx (100 ng/µL) was added, and the ESR test was performed again immediately.

### DFT simulation

Generally, a reaction is not only governed in the direction of energy minimization but is also significantly affected by the reaction rate. An exothermic reaction is energetically favored in nature but slow in practice. Therefore, stepping down in Gibbs free energy is more meaningful for chemical reactions. As we follow the changes in Gibbs free energies along the reaction path, the bottlenecks are always the ascending changes that catalytic materials could significantly reduce. The catalysts may involve many intermediate and transition states by adsorbing reactants or fragments. Many attempts were made to determine the reaction path and the catalytic mechanism to approach the reaction intermediates and transition states. The intermediates and transition states were simulated by DFT using the Gaussian 09 package62 and fukui function analysis using the Multifunctional Wavefunction Analyzer63. 6–31 G* and LAN2TZ54 were used to reduce the computation workload as the basis sets for light and heavy atoms64. Since DFT approximated the electron-electron interactions as a background functional of electronic density, the functional was generally split into 2 parts: the Hartree potential and the exchange-correlation potential. We chose the well-known B3LYP65 as the exchange-correlation potential. We did not assume close shells for different states but considered multiple possibilities of spins in this work. For all systems, the corresponding zero-point-energy (ZPE) and thermal corrections of frequency analyses were performed at the same level of theory to obtain Gibbs free energy.

Based on the experimental results, the model catalyst was constructed as a graphene flake with its central region replaced by a single metal atom surrounded by 4 N atoms52. The model catalyst was positively charged according to the experimental results and literature66. With the structure, the intermediate state could be obtained by fully relaxing the geometry of the flake with fragment/reactant attached. The transition state was the saddle point along the reaction coordinate.

Experimentally, different catalysts showed POD, CAT, and SOD-mimicking activities. Therefore, we focused on exploring POD-like, CAT-like, and SOD-like processes in the simulation.

### CAT-like process of MNx

With the simulated reaction pathways, the procedure of the CAT-like reaction can be summarized as:

$${H}_{2}{O}_{2}+{{MN}}_{x}={{MN}}_{x}-O+{H}_{2}O$$
(1)

which is the initialization of a catalytic process. The catalytic molecule is activated as MNxO complex. The reaction then behaves as

$${H}_{2}{O}_{2}+{{MN}}_{x}-O={{MN}}_{x}-O-O+{H}_{2}O$$
(2)

The MNxOO complex represents the “bilateral adsorptions” of the MNx with one oxygen atom on each side. Then another H2O2 molecule can be broken down into water and oxygen.

$${H}_{2}{O}_{2}+{{MN}}_{x}-O-O={{MN}}_{x}-O+{H}_{2}O+{O}_{2}$$
(3)

RhN4 and VN4 repeat Eqs. (2) and (3), but Fe-Cu-N6 continues to the finalization step:

$${H}_{2}{O}_{2}+{{MN}}_{x}-O={{MN}}_{x}+{H}_{2}O+{O}_{2}$$
(4)

Then, Fe-Cu-N6 returns to the initialization step (Eq. (1)) for the next cycle of catalytic procedures.

### The POD-like process of MNx

The POD-like mechanism is more complicated than the CAT-like for more units joint the reaction, and there are more possibilities in the intermediate/transition states. The total reaction of the POD-like mechanism is:

$$2H+{H}_{2}{O}_{2}=2{H}_{2}O$$
(5)

Due to the similarities between the CAT and POD processes, the initialization steps are the same as the CAT process (see Eq. (1)). Therefore, to be precise, we provide all the reaction steps by MNx.

RhN4 has the following POD-like procedure of reaction:

$${H}_{2}{O}_{2}+{{RhN}}_{4}={{RhN}}_{4}-O+{H}_{2}O$$
(6)
$${H}_{2}{O}_{2}+{{RhN}}_{4}-O={{RhN}}_{4}-O-O+{H}_{2}O$$
(7)
$$H+{{RhN}}_{4}-O-O={{RhN}}_{4}-O-{OH}$$
(8)
$$H+{{RhN}}_{4}-O-{OH}={{RhN}}_{4}-{OH}-{OH}$$
(9)
$$H+{{RhN}}_{4}-{OH}-{OH}={{RhN}}_{4}-{OH}+{H}_{2}O$$
(10)
$${H}_{2}{O}_{2}+{{RhN}}_{4}-{OH}={{RhN}}_{4}-{OH}-{OOHH}$$
(11)
$$H+{{RhN}}_{4}-{OH}-{OOHH}={{RhN}}_{4}-O+2{H}_{2}O$$
(12)

The main loop of the catalytic process is from Eqs. (7)–(12).

VN4 has the following POD-like procedure of reaction:

$${H}_{2}{O}_{2}+{{VN}}_{4}={{VN}}_{4}-O+{H}_{2}O$$
(13)
$$H+{{VN}}_{4}-O={{VN}}_{4}-{OH}$$
(14)
$${H}_{2}{O}_{2}+{{VN}}_{4}-{OH}={{VN}}_{4}-O-{OH}+{H}_{2}O$$
(15)
$$H+{{VN}}_{4}-O-{OH}={{VN}}_{4}-O+{H}_{2}O$$
(16)

The main loop of the catalytic process is from Eqs. (14)–(16).

Fe-Cu-N6 has the following POD-like procedure of reaction:

$${H}_{2}{O}_{2}+({{Fe}-{Cu}-N}_{6})=({{Fe}-{Cu}-N}_{6})-O+{H}_{2}O$$
(13)
$${H}_{2}{O}_{2}+\left({{Fe}-{Cu}-N}_{6}\right)-O=\left({{Fe}-{Cu}-N}_{6}\right)-O-O+{H}_{2}O$$
(14)
$$H+\left({{Fe}-{Cu}-N}_{6}\right)-O-O=\left({{Fe}-{Cu}-N}_{6}\right)-O-{OH}$$
(15)
$$H+\left({Fe}-{Cu}-N-C\right)-O-{OH}=\left({{Fe}-{Cu}-N}_{6}\right)-O+{H}_{2}O$$
(16)
$$H+\left({{Fe}-{Cu}-N}_{6}\right)-O=\left({{Fe}-{Cu}-N}_{6}\right)-{OH}$$
(17)
$$H+\left({{Fe}-{Cu}-N}_{6}\right)-{OH}=\left({{Fe}-{Cu}-N}_{6}\right)+{H}_{2}O$$
(18)

### SOD-like process of MNx

The equation of reactions of SOD-like activities for Rh/VN4 can be summarized in the following steps:

$${O}_{2}^{\cdot -}+{{{{{{\rm{Rh}}}}}}/{{{{{\rm{VN}}}}}}}_{4}({III})={{{{{{\rm{Rh}}}}}}/{{{{{\rm{VN}}}}}}}_{4}({II})-{OO}$$
(19)
$${O}_{2}^{\cdot -}+{{{{{{\rm{Rh}}}}}}/{{{{{\rm{VN}}}}}}}_{4}\left({II}\right)-{OO}={{{{{{\rm{Rh}}}}}}/{{{{{\rm{VN}}}}}}}_{4}\left(I\right)-{OO}-{OO}$$
(20)
$${H}^{+}+{{{{{{\rm{Rh}}}}}}/{{{{{\rm{VN}}}}}}}_{4}\left(I\right)-{OO}-{OO}={{{{{{\rm{Rh}}}}}}/{{{{{\rm{VN}}}}}}}_{4}\left({II}\right)-{OO}-{OOH}$$
(21)
$${H}^{+}+{{{{{{\rm{Rh}}}}}}/{{{{{\rm{VN}}}}}}}_{4}\left({II}\right)-{OO}-{OOH}={{{{{{\rm{Rh}}}}}}/{{{{{\rm{VN}}}}}}}_{4}\left({III}\right)-{OO}+{H}_{2}{O}_{2}$$
(22)
$${O}_{2}^{\cdot -}+{{{{{{\rm{Rh}}}}}}/{{{{{\rm{VN}}}}}}}_{4}\left({III}\right)-{OO}={{{{{{\rm{Rh}}}}}}/{{{{{\rm{VN}}}}}}}_{4}\left({II}\right)-{OO}+{O}_{2}$$
(23)

The Fe-Cu-N6 has a SOD-like process with the initialization stage which can be summarized as the following equations:

$${O}_{2}^{\cdot -}+({{Fe}-{Cu-N}}_{6})(I)=({{Fe}-{Cu-N}}_{6})(0)-{OO}$$
(24)
$${H}^{+}+({{Fe}-{Cu-N}}_{6})(0)-{OO}=({{Fe}-{Cu-N}}_{6})(I)-{OOH}$$
(25)
$${H}^{+}+({{Fe}-{Cu-N}}_{6})(I)-{OOH}=({{Fe}-{Cu-N}}_{6})({II})-{OOHH}$$
(26)

$${O}_{2}^{\cdot -}+\left({{Fe}-{Cu-N}}_{6}\right)\left({II}\right)-{OOHH}=\left({{Fe}-{Cu-N}}_{6}\right)\left(I\right)-{OO}-{OOHH}$$
(27)
$${O}_{2}^{\cdot -}+\left({{Fe}-{Cu-N}}_{6}\right)\left(I\right)-{OO}-{OOHH}=\left({{Fe}-{Cu-N}}_{6}\right)\left(0\right)-{OO}-{OO}-{OOHH}$$
(28)
$${H}^{+}+\left({{Fe}-{Cu-N}}_{6}\right)\left(0\right)-{OO}-{OO}-{OOHH}=\left({{Fe}-{Cu-N}}_{6}\right)\left(I\right)-{OO}-{OOH}-{OOHH}$$
(29)
$${H}^{+} +\left({{Fe}-{Cu-N}}_{6}\right)\left(I\right)-{OO}-{OOH}-{OOHH} \\ =\left({{Fe}-{Cu-N}}_{6}\right)\left({II}\right)-{OO}-{OOHH}+{H}_{2}{O}_{2}$$
(30)
$$\left({{Fe}-{Cu-N}}_{6}\right)\left({II}\right)-{OO}-{OOHH}=\left({{Fe}-{Cu-N}}_{6}\right)\left({II}\right)-{OOHH}+{O}_{2}$$
(31)

### GPx-like process of VN4

The GPx-like activity is similar to POD activities. The GSH provides a hydrogen atom to H2O2 via catalyst. The total reaction is

$$2{GSH}+{H}_{2}{O}_{2}={GSSG}+2{H}_{2}O$$
(32)

The GPx reaction can be decomposed in the following steps with catalyst involved:

$${H}_{2}{O}_{2}+{{VN}}_{4}={{VN}}_{4}-{OOHH}$$
(33)
$${GSH}+{{VN}}_{4}-{OOHH}={{VN}}_{4}-{OH}-{GS}$$
(34)
$${GSH}+{{VN}}_{4}-{OH}-{GS}={{VN}}_{4}-{H}_{2}O+{GSSG}$$
(35)

### Fukui function

Fukui function67 is an important concept and a valid approach for analysis of DFT, which can be used in predicting the reactivity of electrophilic and nucleophilic at different sites68. In general, a site with an isosurface of Fukui function indicates its better reactivity. Fukui function is defined as

$$f\left(\vec{{{{{{\bf{r}}}}}}}\right)={\left[\frac{\delta \mu }{\delta v\left(\vec{{{{{{\bf{r}}}}}}}\right)}\right]}_{N}={\left[\frac{\partial n\left(\vec{{{{{{\bf{r}}}}}}}\right)}{\partial N}\right]}_{v(\vec{{{{{{\bf{r}}}}}}})}$$
(36)

μ is the electronic chemical potential $$v(\vec{{{{{{\bf{r}}}}}}})$$ is the external potential, N is the number of electrons in the present system, and $$n(\vec{{{{{{\bf{r}}}}}}})$$ is the electron density. In practice, the donor Fukui function can be evaluated in its finite difference form:

$${f}^{-}\left(\vec{{{{{{\bf{r}}}}}}}\right)={n}_{N}\left(\vec{{{{{{\bf{r}}}}}}}\right)-{n}_{N-1}\left(\vec{{{{{{\bf{r}}}}}}}\right)$$
(37)

nN and nN-1 represent the electron densities of the system with N electrons and N-1 electrons with identical structures, respectively. The donor Fukui function $${f}^{-}(\mathop{{{{{{\bf{r}}}}}}}\limits^{ \rightharpoonup })$$ shows the change of electron density at each position when the system is attacked by electrophilic reagents and then donates charge to electrophiles, reflecting the ability of electrons in different regions to participate in electrophilic reactions.

The donor Fukui functions of RhN4, VN4, Fe-Cu-N6, FeN4 and CuN4 are shown in Fig. S27. The isosurface near the Fe site of FeN4 is broader than that of Cu in CuN4. Thus, the electrophilic reactivity of the Fe sites is much higher than the Cu sites (Fig. S27c–e). In the Fe-Cu-N6 system, the two dual centers of Fe and Cu atoms can enhance their common neighboring N atoms’ electrophilic reactivity. Therefore, the Fukui function suggests that the two-center N atoms may also contribute to catalytic processes (Fig. S27e). For RhN4 and VN4, the active region around Rh is wider than V, corresponding to Rh’s higher electrophilic reactivity.

### Formation energy

To evaluate the stability of catalysts, we investigated the formation energy with the Vienna ab initio Simulation Package (VASP)69,70. The elemental core and valence electrons were represented using the projector augmented wave (PAW) method71. The Perdew−Burke−Ernzerhof generalized gradient approximation (GGA-PBE)72 functional was employed to estimate the exchange−correlation potential energy. For simulation of five different Fe-Cu-Nx bimetallic structures, a 12.78 Å × 12.30 Å supercell containing 58 atoms is sufficient. To avoid the interaction between layers due to periodicity and improve the reliability and accuracy of calculations, a 15 Å vacuum distance was taken along the z-direction perpendicular to the surface of catalysts. We ensured the convergence of calculations by setting the cutoff energy for the plane-wave basis at 500 eV. In addition, all Brillouin zone integrations were performed according to the Monkhorst Pack scheme5, and 4 × 4 × 1 k-mesh was used for relaxation calculations. For structural relaxation, the convergence criteria for energy and force were set to 1 × 10−5 eV and −0.01 eV/Å, respectively. In this work, the formation energies (Ef) of different Fe-Cu-Nx bimetallic structures were calculated by the following equation:

$${E}_{f}={E}_{{FeCu}{N}_{x}C}-{E}_{{N}_{x}C}-{E}_{{Fe}}-{E}_{{Cu}}$$
(38)

Where $${E}_{{FeCu}{N}_{x}C}$$ and $${E}_{{N}_{x}C}$$ are the total energies of Fe-Cu-Nx systems and NxC pore substrate, respectively. EFe and ECu are energy per atom for the Fe and Cu atoms, respectively.

### Animal experiment

All animal procedures were approved by the Institute of Radiation Medicine, Chinese Academy of Medical Science, and Peking Union Medical College, following the ethics and ethics rules of the Animal Committee and complying with the principles of animal protection, animal welfare, and ethics (IRM-DWLI-2021107). The animal models were established in Tianjin Medical University General Hospital.

### Fluid percussion injury (FPI) models

Male C57BL/6 J mice (6–8 weeks) were purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd. and reared for 1 week in a constant temperature and humidity condition to adapt to the environment. Mice were housed in a constant temperature (21–23 °C) and animal humidity (45–60%) environment on a 12-h light-dark cycle, with food and water available ad libitum. Fluid percussion injury (FPI) models were conducted on male C57BL/6 J mice (7–9 weeks). Mice were intraperitoneally injected with 10% chloral hydrate (10 mg/kg), and the anesthesia state was judged through skin pinching reaction and toe stimulation reaction. Mice were fixed in a stereotaxic frame and depilated, and the scalp was cut with a 1.5–2 cm length wound. Craniotomy was performed by drilling the skull on the right side of the parietal-temporal cortex in a circle of 4 mm. FPI was established with a hydraulic impact operated at 1.9 atm. The scalp of injured mice was sutured with PGA sutures modified with MNx randomly and divided into RhN4, VN4, FeN4, CuN4, and Fe-Cu-N6 groups as the administration groups (n = 17 per group). Notably, the MNx-sutures were treated with a high-temperature sterilization procedure before use, and the entire animal experiments were conducted under a specific pathogen-free (SPF) environment. TBI groups were sutured with pure PGA suture with FPI as the vehicle group. The control groups were only depilated as wild animals without FPI. All mice were marked, classified, and put into cages under the SPF level environment. The scalp injury and body weight were tracked throughout the experiments.

### Histological analysis

Mice were anesthetized, and all organs were taken out on day 3 or day 4 post-injury. Scalp and brain tissues were fixed in neutral formalin solution for 48 h, embedded in paraffin, and mounted on slides (2 µm). Tissues were deparaffinized twice with xylene, dehydrated with gradient alcohol (100%, 95%, 80%, and 70%) for 5 minutes every time, and washed with phosphate-buffered saline (PBS) twice. Then, the slices were covered with a 3% H2O2 solution and kept at 37 °C for 15 minutes to inactivate endogenous peroxidase. Subsequently, the slides were immersed in a dye vat filled with a citrate buffer solution at 95 °C for 10 min to repair the antigen. After cooling naturally, slices were blocked with 10% goat serum albumin for 40 min at 37 °C. For immunofluorescence staining, the primary antibody was dropped onto the slices and incubated at 4 °C overnight. The primary antibody information is as followed: CD31 Polyclonal antibody (1:100, Proteintech, 28083-1-AP), CD68 Polyclonal antibody (1:100, Proteintech, 28058-1-AP), anti-TNFα antibody (1:100, Abcam, ab183218), anti-IL-1β antibody (1:100, Affinity, AF5103), anti-Iba1 antibody (1:300, Abcam, ab48004), and anti-NeuN antibody (1:100, Proteintech, 26975-1-AP). After the primary antibody was cleaned, CoraLite488-conjugated Affinipure Goat Anti-Rabbit IgG (H+L) (1:100, Proteintech, SA00006-2) was utilized as the secondary antibody in the dark for 1–1.5 h. Then slices were mounted with an anti-fade mounting medium with DAPI (S2110, Solarbio, China) and photographed with a fluorescence microscope (EVOS, AMG). Granulation tissue was examined by staining scalp sections for 3 days post-injury with hematoxylin-eosin (H&E) staining.

### Kit detection

Mice were anesthetized, and all organs were taken out and washed once in neutral PBS buffer on day 3 post-injury (n = 3 per group). Homogenates of scalp and brain tissues were centrifuged at 10,000 g for 10 min, and the supernatant was stored at −80 °C for preparation. An Enhanced BCA protein assay kit (Beyotime, P0010) was used to detect the protein concentration. Typical oxidative stress-related factors MDA, SOD, GSH/GSSG, and H2O2 for scalp and brain tissues were detected with a hydrogen peroxide assay kit (Beyotime, S0038), total SOD assay kit with WST-8 (Beyotime, S0101M), GSH and GSSG assay kit (Beyotime, S0053), and lipid peroxidation MDA assay kit (Beyotime, S0131S). In addition, inflammatory cytokines for scalp and brain tissues along with VEGF for scalp tissues were detected with IL-6 (Abcam, ab100713), IL-1β (Abcam, ab197742), TNF-α (Abcam, ab208348), and VEGF (Abcam, ab209882). These assays were conducted following the instructions provided by the manufacturers. GraphPad software was used for plotting.

### Flow cytometry

To evaluate the changes of immune cells in the wound and blood, we collected the scalp and blood in different groups on day 6 post brain injuries. The fresh wound tissues were digested with collagenase and hyaluronidase (Stemcell, 07919) overnight at 37 °C to obtain single cells for flow cytometry analysis (BD FACSCanto II). Lymphocytes in the blood were collected with Ficoll kits (Solarbio, P8620). Lymphocytes were stained with PE anti-mouse F4/80 (Biolegend, 123110), APC anti-mouse CD80 (Biolegend, 104714), FITC anti-mouse CD206 (MMR) (Biolegend, 141704), APC/Cyanine7 anti-mouse CD3 (Biolegend, 100222), PE/Cyanine7 anti-mouse CD8a (Biolegend, 100722), FITC anti-mouse CD4 (Biolegend, 100406), PE anti-mouse NK-1.1 (Biolegend, 156504), PerCp anti-mouse CD45 (Biolegend, 103130), APC anti-mouse CD19 (Biolegend, 152410), APC anti-mouse CD25 (Biolegend, 101909), and PE-Foxp3 (FJK-16s) (Thermo, 12-5773-82) antibodies, PE Rat IgG2a, ĸ Isotype Ctrl Antibody (Biolegend, 400508), FOXP3 Monoclonal Antibody (FJK-16s), PE (eBioscienceTM InvitrogenTM, 12-5773-82).

### Statistic methods

Data are presented as mean ± standard deviation (SD) or mean ± standard error of the mean. For multiple comparisons, the one-way analysis of variance (ANOVA) with one-sided Tukey’s multiple comparisons test was used to assess the difference in means among groups.

### Reporting summary

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