Bioactive NIR-II gold clusters for three-dimensional imaging and acute inflammation inhibition

Abstract

Strong fluorescence and high catalytic activities cannot be achieved simultaneously due to conflicts in free electron utilization, resulting in a lack of bioactivity of most near-infrared-II (NIR-II) fluorophores. To circumvent this challenge, we developed atomically precise Au₂₂ clusters with strong NIR-II fluorescence ranging from 950 to 1300 nm exhibiting potent enzyme-mimetic activities through atomic engineering to create active Cu single-atom sites. The developed Au₂₁Cu₁ clusters show 18-fold higher antioxidant, 90-fold higher catalase-like, and 3-fold higher superoxide dismutase–like activities than Au₂₂ clusters, with negligible fluorescence loss. Doping with single Cu atoms decreases the bandgap from 1.33 to 1.28 eV by predominant contributions from Cu d states, and Cu with lost electron states effectuates high catalytic activities. The renal clearable clusters can monitor cisplatin-induced renal injury in the 20- to 120-minute window and visualize it in three dimensions using NIR-II light-sheet microscopy. Furthermore, the clusters inhibit oxidative stress and inflammation in the cisplatin-treated mouse model, particularly in the kidneys and brain.

A NIR-II fluorescent imaging probe with enzyme-mimicking activity can inhibit inflammation caused by cisplatin.

INTRODUCTION

Near-infrared-II (NIR-II, 1000 to 1700 nm) imaging with high tissue penetration depth and signal-to-noise ratio has shown great potential and gained wide attention in real-time monitoring, pathogenesis, and pathological evolution for cancer diagnosis as well as neuroscience (1–10). Meanwhile, the biocatalytic or bioactive NIR-II molecular agent is highly desired for real-time monitoring of pathological processes and the molecular mechanism of treatment (11–14). From a physical point of view, photoluminescence generally requires the sufficient electron-hole pair between the valance band and the conduction band in the semiconductor for radiative recombination (15, 16). However, infrared emission is always accompanied by lots of nonradiative recombination, causing inefficient luminescence quantum yield. Meanwhile, biocatalysts are always involved in electron transfer between surface-active atoms and substrate molecules, which will lead to losing free electrons continuously (11, 17–19), and the free electrons involved in the emission of light will be drastically reduced. Thus, this conflict caused failure to obtain the higher efficient fluorescence and catalytic activity synchronously.

Most oxidative stress–related diseases require early intervention and real-time monitoring of pathological evolution (20). However, most traditional noninvasive imaging tools, such as single-photon emission computed tomography, positron emission tomography, computed tomography, and magnetic resonance imaging, meet obstruction of low temporal-spacial resolution to achieve precise diagnosis and real-time monitoring. Oxidative stress and inflammation-related diseases are common in clinics, where the collaborative real-time monitoring and intervention is still an urgent unmet need. Chemotherapeutics often cause injury to normal tissues and organs during medical treatment. For example, cisplatin, which is widely used as an anticancer drug in clinics, can cross-link to DNA and induce oxidative stress and inflammation, which may cause severe side effects, especially nephrotoxicity and neurotoxicity (21–23). Cisplatin-induced acute kidney injury (AKI) (24–27), one of the most common side effects, presents as a sudden decrease in kidney function within a few hours or days. AKI is a typical disease that requires real-time monitoring for early diagnosis as well as early interventions to inhibit oxidative stress and acute inflammation (8, 28–30). Earlier studies shed light on the drug treatment and monitoring process for AKI (31). For instance, mitochondria targeting ceria nanoparticles with atorvastatin display anti-inflammatory and antioxidant effects, improving the therapeutic effect of AKI (32). Molybdenum-based polyoxometalate nanoclusters are developed as nano-antioxidants for the treatment of AKI (33). However, these materials lack the ability to monitor renal function in real time. Multiple kidney-targeting fluorescent probes enable renal monitoring functions but do not provide renal protection (34, 35). The present molecular agent cannot simultaneously both real-time monitor and decrease inflammation. Therefore, it is desirable to develop biocatalytic NIR-II molecular agents to address real-time monitoring and achieve the inhibition of oxidative stress and early inflammation synchronously (36, 37).

In this work, we developed atomic precision Au₂₂ clusters with NIR-II emission and biocatalytic enzyme–mimicking activity (Fig. 1). The single-atom engineering was used to create the Cu single-atom active site in the outer layer of Au₂₂ clusters, leading to 18 times higher antioxidant, 90 times higher catalase (CAT)–like, and 3 times higher superoxide dismutase (SOD)–like activity with negligible luminescence loss. Meanwhile, Au₂₂ clusters can distinguish and monitor the cisplatin-induced AKI after 20 to 120 min injection at the NIR-II window and achieve three-dimensional (3D) imaging under light-sheet microscopy. The Au₂₂ clusters show superior performance in inhibiting oxidative stress and inflammation partially in the injured kidney and brain.

Fig. 1. Schematic diagram of the properties and biomedical applications of Au₂₂ clusters.

We introduced Cu single-atom active sites to the atomic-precision Au₂₂ clusters having strong NIR-II fluorescence. This atom engineering procedure reduces the bandgap from 1.33 to 1.28 eV via contributions of Cu s and p states, and Cu atom with lost electron states contributes to potent enzyme-mimicking activities. Consequently, the bioactive NIR-II clusters exhibit good capacities for highly accurate monitoring of cisplatin-induced kidney injury and inhibition of oxidative stress and inflammation in multiple organs of the cisplatin-treated mouse model, particularly in the kidneys and brain.

RESULTS

Structural characterization of gold clusters

The Au₂₂ clusters are prepared and purified as per the previously reported method (38–40). Au₂₂(SG)₁₈ clusters with glutathione as the ligand are composed of a prolate Au₈ core, two trimeric, and two tetrameric motifs, which have been proved to be the stable structure model (Fig. 2A) (40). The transmission electron microscope (TEM) image and the size distribution histogram show that the core size of Au₂₂ clusters is about 1.5 nm (Fig. 2B), suggesting the ultrasmall size (41, 42). Dynamic light scattering (DLS) determines that the mean hydrodynamic size of Au₂₂ is 2.01 nm in phosphate-buffered saline (PBS) (Fig. 2C), slightly larger than the core size due to the stretching of the outer ligands. Since the renal filtration threshold value is 5.5 nm, the ultrasmall size further indicates its highly efficient renal clearance. Electrospray ionization mass spectra (ESI-MS) reveal a distinct mass/charge ratio (m/z) peak at ~1968.44, corresponding to the calculated molecular weight of [Au₂₂(SG)₁₈-5H]⁵⁻ (Fig. 2D) and indicating that atomic-precise Au₂₂ clusters have been successfully obtained. The characteristic absorption peak of Au₂₂ clusters is observed at 515 nm, with the emission center located at 925 nm and a broad tail extending to 1300 nm through the charge transfer between ligand and gold core (Fig. 2E) (43, 44). The excitation spectrum shows a very wide excitation range (fig. S1A). Besides, the characteristic absorption peak at 515 nm is still evident after the 100 K filtration (Fig. 2F), and the NIR-II signal exhibits negligible fluorescence loss (fig. S1B). The quantum yield of the Au₂₂ clusters is about 0.16%, which is slightly lower than Au₂₅(GSH)₁₈ and higher than other materials (44, 45). As illustrated in Fig. 2 (G and H), Au₂₂ clusters exhibit excellent photostability under continuous excitation at 808 nm and good fluorescence stability in deionized (DI) water and PBS at room temperature for a week. Meanwhile, Au₂₂ clusters demonstrate good stability without obvious signal loss from 300 to 900 nm for a week in DI water, PBS, and fetal bovine serum (FBS) (Fig. 2I and fig. S2). These results reveal that the Au₂₂ clusters with well-defined atomic precision hold ultrasmall size, great water solubility, and stable structure, which is favorable for further application (46). We also use atom engineering to modify Au₂₂ clusters with various metallic elements. The optical absorption spectra of doped clusters with different metals and different molar concentrations are shown in fig. S3. While the doping molar ratio increases to 15:7 or the highest ratio 11:11, the peaks of absorption spectra diminish or even disappear for Cu-, Pt-, Zn-, Cd-, Er-, and Ag-doped Au₂₂ clusters (fig S3). The density functional theory (DFT) simulations (fig. S4) reveal that the highest ratio can significantly induce the reduction of the Au sp and 5d states in the highest occupied molecular orbital (HOMO) and the energy levels below HOMO, which further affects the optical transition, leading to great changes in absorption peaks. In addition, an obvious distortion in cluster structure caused by heteroatom doping at high concentrations, especially for Zn, Cd, and Er, will also affect the electronic structure of doping clusters and thus have a distinct impact on optical properties. The TEM images demonstrate that Au₂₁Cu₁ clusters with ultrasmall size still have good homogeneity and dispersion, which is consistent with Au₂₂ clusters (fig. S5A). The full x-ray photoelectron spectroscopy (XPS) of Au₂₁Cu₁ clusters has more Cu 2p peaks than that of Au₂₂ clusters (fig. S6). XPS further confirms that the Cu atom substitution in Au₂₂ clusters induces a shift in the binding energy with the Au 4f region (Fig. 2J) (11, 47). Compared to pure Au₂₂ clusters, the full XPS spectrum of the binding energy of Au₂₁Cu₁ clusters shows a Cu signal, and the Cu 2p XPS spectrum can be found (Fig. 1K). To identify the doping of the Cu atom, the energy dispersive spectrometer (EDS) was performed and shows the signal of Cu, S, and Au (fig. S5B); meanwhile, the inductively coupled plasma mass spectrometry (ICP-MS) was recorded, revealing metal element percentages of 95.25 and 4.75% for Au and Cu, respectively (fig. S7). In addition, the extended x-ray absorption fine structure (EXAFS) spectra were recorded to demonstrate the formation of the Cu-S bonds (Fig. 2L). It can be seen that the dominant peak of Au₂₁Cu₁ clusters in the R-space is shorter than that of Cu foil, suggesting the successful single-atom substitution.

Fig. 2. Structural characterization of gold clusters.

(A) The structure illustration of Au₂₂. Both blue and yellow represent Au atoms, and red represents S atoms. (B) Typical TEM image of Au₂₂ clusters with an average size of 1.5 nm. (C) The hydrodynamic size was measured to be 2.01 nm using DLS. (D) The ESI-MS of Au₂₂ clusters. The illustration shows the molecular weight of Au₂₂ in the m/z range from 1967 to 1970. (E) Ultraviolet-visible (UV-vis) absorption spectra and emission spectra of Au₂₂ clusters. The black line represents the absorption spectrum, and the red line represents the emission spectrum. (F) UV-vis absorption spectra and NIR-II fluorescence images of Au₂₂ cluster aqueous solution excited at 808 nm before and after filtration with the 100 K ultrafiltration tube (illustration: 1000 nm long pass filter, 100 ms). (G) Photostability of Au₂₂ clusters excited with an 808 nm laser for 30 min. (H) NIR-II fluorescence stability of Au₂₂ clusters in DI water and PBS before and after a week. (I) Stability of optical characteristic absorption peaks of Au₂₂ clusters of 7 days in water. (J) Au 4f region of Au₂₂ and Au₂₁Cu₁ clusters. (K) Cu 2p XPS spectrum of Au₂₁Cu₁ clusters. (L) Fourier-transformed magnitude of the Cu K-edge EXAFS spectra of Au₂₁Cu₁ and Cu foil, showing the radial distance of Cu-S bond and Cu-Cu bond.

Enzyme-like properties of gold clusters

Au₂₁Cu₁ clusters with a Cu single-atom active site show high catalytic activity, and the catalytic process is shown in Fig. 3A. The adsorption strength of reactant molecules and free radicals is a prerequisite for the consequent catalytic reaction. Thus, the adsorption energies (E_ads) of different species were tested in Fig. 3 to investigate the reactive oxygen and nitrogen species adsorption ability of Au₂₂ and Au₂₁Cu₁ by DFT. Meanwhile, we also assessed the antioxidative properties of gold clusters to pinpoint biocatalytic enzyme–mimicking activity using the 2,2'-azino-bis(3-ethylbenzthiazoline-6-sulfonic acid) (ABTS) method. The E_ads of ABTS^•+ is calculated to be −2.18 and −1.19 eV for Au₂₁Cu₁-trimer and Au₂₁Cu₁-tetramer, respectively (Fig. 3B), implying the relatively strong interaction between ABTS^•+ and the Au₂₁Cu₁ clusters, whereas it is difficult for ABTS^•+ to adsorb on the Au₂₂ cluster with a positive value E_ads of 1.59 eV. Furthermore, the antioxidant activity of Au₂₁M₁ (M: Ag, Zn, Cu, Pt, Cd, and Er) clusters was performed by controlling for different times and concentrations (Fig. 3C and fig. S8A). In particular, Au₂₁Cu₁ clusters display 18 times higher antioxidant activity than that of Au₂₂ clusters, indicative of fantabulous antioxidant activity (Fig. 3D). In addition, adsorbed O₂^•− and H₂O₂ on Au₂₁Cu₁ show a relatively high E_ads in the range of −0.79 to −0.50 eV; these negative adsorption energies also indicate the moderate interactions between the catalytic sites of Au₂₁Cu₁ clusters and the adsorbates O₂^•− and H₂O₂. (Fig. 3, E and H). Meanwhile, Au₂₁Cu₁ clusters show the strongest SOD-like activity among all the clusters (Fig. 3F and fig. S8B). The inhibition rate of pure Au₂₂ only reaches 26%, while that of Au₂₁Cu₁ clusters can be increased to 77% (Fig. 3G). Au₂₁Cu₁ clusters exhibit high catalytic degradation ability against H₂O₂, an illustration of CAT-like catalytic activity (Fig. 3I). Quantitative result shows that the CAT-like activity of Au₂₁Cu₁ clusters is 90 times higher than that of pure Au₂₂ clusters (Fig. 3J). For further verification, the free-radical scavenging capability for ONOO⁻ is also investigated, which is a highly toxic molecule generated in disease states. The ONOO⁻ absorbate on the Au₂₁Cu₁ clusters exhibits relatively lower E_ads of −0.23 and −0.14 eV with different Cu sites than Au₂₂ (Fig. 3K). For Au₂₂ clusters, the positive E_ads (0.16 eV) of ONOO⁻ indicates that the adsorption system is energetically unfavorable and thus results in the weak interaction between the ONOO⁻ absorbate and catalytic site on cluster. Furthermore, Au₂₁Cu₁ clusters still manifest a higher scavenging efficiency compared to pure Au₂₂ and other metal doped clusters (Fig. 3L). The quantitative result is presented in Fig. 3M. Thus, the increase of various unique enzyme-mimicking activities can contribute to the Cu single-atom active site in the Au₂₂ clusters. Metal clusters with catalytic properties have been widely reported in previous studies, but clusters with NIR-II properties are rarely mentioned, which play important and essential roles in bioscience (48–51). Thus, biocatalytic NIR-II molecular agents are urgent and of great significance to understand the mechanism.

Fig. 3. The enzyme-mimicking activity of gold clusters.

(A) Schematic illustration of the enzyme-mimicking activity of Au₂₂ clusters. (B) The adsorption energy of ABTS^•+ on Au₂₂, Au₂₁Cu₁-trimer, and Au₂₁Cu₁-tetramer clusters, respectively. Time-dependent curves (C) and quantitative results (D) of ABTS^•+ of Au₂₂ and Au₂₁M₁ (M: Ag, Zn, Cu, Pt, Cd, and Er) clusters. (E) The adsorption energy of O₂^•− on Au₂₂, Au₂₁Cu₁-trimer, and Au₂₁Cu₁-tetramer clusters, respectively. Time-dependent curves (F) and quantitative results (G) of O₂^•− of Au₂₂ and Au₂₁M₁ clusters. (H) The adsorption energy of H₂O₂ on Au₂₂, Au₂₁Cu₁-trimer, and Au₂₁Cu₁-tetramer clusters, respectively. Diagram (I) and quantitative results (J) of the CAT-like activity of Au₂₂ and Au₂₁M₁ clusters. (K) The adsorption energy of ONOO⁻ on Au₂₂, Au₂₁Cu₁-trimer, and Au₂₁Cu₁-tetramer clusters, respectively. Time-dependent curves (L) and quantitative results (M) of ONOO⁻ of Au₂₂ and Au₂₁M₁ clusters.

DFT calculations of gold clusters

To further explore the mechanism of higher CAT-like and SOD-like activities of Au₂₁Cu₁ clusters, the electronic properties of Au₂₁Cu₁ clusters were studied using DFT. To ensure the reliability and accuracy of this simulation, the structure of Au₂₂ and Au₂₁Cu₁ clusters protected by thiolate were adopted. Considering that the catalytic reaction is more likely to occur at the exposed active site of the clusters, two types of replacements on the staple motifs of the Au₂₂ cluster for the guest metallic atoms have been taken into consideration: trimer motif and tetramer motif. The DFT-optimized structures are shown in Fig. 4 (A, E, and I); the Cu doping causes the tetramer and trimer motifs to distort (Fig. 4, E and I). They are different from the normal S-Au-S chain, which aligns in a nearly straight line (Fig. 4A). The doped Cu shrinks the S-X-S chain; compared with the typical bond length of S-Au at 2.32 Å, the S-Cu bond is 2.20 Å (11, 52). The similarity between the Cu and Au atoms guarantees that the binding of S-Cu-S is so “tight” that the relative positions of Cu to S atoms can be hardly shifted by the dynamics during the catalytic processes. In addition, DFT optimization confirms the stability of the modeled Au₂₁Cu₁ clusters, and the doping formation energies of the two Cu-doped structures are calculated. As shown in Fig. 4 (E and I), the structures of Au₂₁Cu₁-tetramer and Au₂₁Cu₁-trimer have slightly positive doping formation energies (E_f-doping), and the tetramer motif-doping Cu is more stable (0.003 eV), demonstrating comparable structural stability to Au₂₂ clusters. Moreover, the formation energies (E_f) of Au₂₂, Au₂₁Cu₁-tetramer, and Au₂₁Cu₁-trimer are also calculated. As shown in fig. S9, the E_f of Au₂₂, Au₂₁Cu₁-tetramer, and Au₂₁Cu₁-trimer are all negative (about −15 eV), demonstrating that the three cluster configurations are all energetically favored. Furthermore, charge transfer and distributions on the undoped and two Cu-doped configurations were predicted by DFT. As shown in Fig. 4 (B, F, and J), the charge density differences between the metal atom (Au or Cu) and the surrounding S atoms are similar to some extent, and it can be seen that the metal atoms Au/Cu on motifs would donate electrons to the adjacent S atoms. Besides, the electron distribution around the active site on motifs in these configurations is further analyzed by the electron localization function (ELF). Since the ELF value describes the localization of electrons, a large ELF value indicates that there is a covalent bond or a lone pair. As seen from Fig. 4 (C, G, and K), the localization of electrons between Cu-S is distinctly lower than Au-S. Thus, it is easy for Cu to exchange with external electrons and form a bond with the small units in the catalytic process. Therefore, the catalyst efficiency of Au₂₁Cu₁ clusters may benefit from the Cu doping on the motifs. With the electronic structure predicted by DFT, energy levels of undoped and two Cu-doped Au₂₂ clusters were calculated. As shown in Fig. 4D, the HOMO–lowest unoccupied molecular orbital (LUMO) gap of the pure Au₂₂ cluster is predicted to be 1.33 eV. In the electronic structure of Au₂₂(SCH₃)₁₈, LUMO is mainly composed of Au 5d and Au 6p. The HOMO can be mainly assigned to Au 5d and Au 6s atomic orbitals. Both HOMO and LUMO energy levels have an obvious degree of S 3p states. With Cu doping at the site of tetramer or trimer motif, the HOMO-LUMO gap of the Au₂₂ cluster slightly decreases to around 1.27 eV; in Fig. 4 (H and L), Cu-doping can introduce the previously unidentified energy levels, which can be mainly attributed to Cu 3d atomic orbitals. The energy level diagram of Au₂₂ clusters and Au₂₁Cu₁ clusters is shown in Fig. 4M. In addition, the DFT-simulated optical properties of Au₂₂ and two types of Au₂₁Cu₁ clusters are also shown in Fig. 4. The imaginary part of the dielectric function is highly related to the absorption of the materials. Au₂₂ and two types of Au₂₁Cu₁ clusters show peaks at about 2.46 and 2.80 eV in the imaginary part of the complex dielectric functions (Fig. 4N), corresponding to the absorption peaks at about 505 nm and around 443 nm (Fig. 4O). It shows reasonable agreement with the experimental absorption peaks (515 and 450 nm) (40, 53).

Fig. 4. DFT calculations of gold clusters.

The geometrically optimized structures of (A) Au₂₂, (E) Au₂₁Cu₁-tetramer, and (I) Au₂₁Cu₁-trimer clusters. Orange, yellow, and blue balls represent Au, S, and Cu atoms, respectively. The locally enlarged version of charge density difference of (B) Au₂₂, (F), Au₂₁Cu₁-tetramer, and (J) Au₂₁Cu₁-trimer clusters. The ELF images of (C) Au₂₂, (G) Au₂₁Cu₁-tetramer, and (K) Au₂₁Cu₁-trimer clusters. The energy levels of (D) Au₂₂, (H) Au₂₁Cu₁-tetramer, and (L) Au₂₁Cu₁-trimer clusters were simulated by density DFT. (M) The energy level diagram of Au₂₂ and Au₂₁Cu₁ clusters. (N) The imaginary part of the dielectric function ɛ₂ (ω). (O) The absorption spectra of the Au₂₂, Au₂₁Cu₁-tetramer, and Au₂₁Cu₁-trimer system.

NIR-II light-sheet microscopy and real-time monitoring of kidney

Early diagnosis of kidney injury is extremely needed in clinics, ascribing to a remarkable capacity to withstand insults for an extended period, and needs a robust method for clinical diagnosis (54, 55). Sensitive and specific tests for early diagnosis rely heavily on ideal biomarker with high sensitivity and specificity (56, 57). Cisplatin is widely used in clinics, which is excreted through the kidney and could trigger AKI at high incidence (58). Herein, cisplatin is used for AKI mouse modeling to reveal the monitoring and treatment of Au₂₂ clusters. We conducted an ex vivo 3D imaging through NIR-II Airy beam–based light-sheet microscopy on the mouse kidney (59, 60) and a visualization study with wide-field bioimaging in vivo after being injected with different doses of cisplatin (Fig. 5A). We evaluated the existence of AKI based on a more direct way, which was 3D microimaging through NIR-II Airy beam–based light-sheet microscopy (Fig. 5, B to I). The homogeneity of the glomerulus for the control group (Fig. 5C) is much better than the cisplatin group, which shows some abnormal enlargements and nodules indicated by the arrows in Fig. 5F. Similarly, some tubules of the cisplatin group evolved to be abnormally coarse (Fig. 5G), while better homogeneity of the tubule morphology is found for the control group (Fig. 5D). Figure 5 (H and I) shows the statistical data from the arrows and cross lines in Fig. 5 (C, D, F, and G), suggesting quantitative differences between the two groups, confirming the micropathological feature of AKI. To further reveal the great potential of gold clusters for real-time monitoring of renal dysfunction, NIR-II imaging was performed in vivo on mice. As shown in Fig. 5J, the fluorophore intensity of the kidney and bladder for the normal group surpasses a peak between 5 and 20 min after the injection and then decreases rapidly because of the high renal clearance of the clusters. Signals in the kidneys and bladder of the mice were observed at 20 min after injection in the cisplatin group. However, signals decreased slowly and remained at a high level even 120 min after the injection (Fig. 5, K and L). In addition, the bladder-to-kidney intensity ratio of the cisplatin group was 2.4 times higher at 20 min than that of the normal group after intravenous injection. However, the bladder-to-kidney intensity ratio in the cisplatin group mice remained around 0.65 at 120 min, while no signals of the kidneys and bladder were observed in the normal group (Fig. 5M). This may be attributed to the severe impairment of kidney function in mice after cisplatin injection, and the filtering effect of the glomeruli is weakened, resulting in further accumulation of gold clusters in the kidney. These above results confirmed that Au₂₂ clusters can achieve 3D visualization under NIR-II light-sheet microscopy and distinguish and monitor the AKI after 20 to 120 min of injection due to its high sensitivity, high specificity, and high fluorescence intensity. Thus, Au₂₂ clusters show great potential to be used for the clinical diagnosis and monitoring of AKI-related diseases.

Fig. 5. NIR-II imaging of light-sheet microscopy and real-time monitoring process of gold clusters.

(A) Schematic illustration of the cisplatin-induced AKI disease model. The gold clusters can monitor cisplatin-induced kidney injury at the NIR-II window and realize 3D visualization using NIR-II light-sheet microscopy. 3D mouse kidney of the control group (B) and the AKI group (E). Scale bar, 200 μm. Glomerular markers (C and F) and renal tubule markers (D and G) for control mice and cisplatin mice, respectively. Scale bar, 100 μm. Quantitative analysis of (H) glomeruli and (I) renal tubules for the control group and the AKI group. (J) Real-time monitoring process of the kidney and bladder for the control group and the AKI group in the NIR-II window under an 808 nm laser. (K to M) Kidney and bladder fluorescence quantitative analysis results and kidney fluorescence value/bladder fluorescence value results.

In vitro treatment of cisplatin-induced AKI with gold clusters

To pinpoint the biological activity of Au₂₂ and Au₂₁Cu₁ clusters, cytotoxicity tests for human proximal tubular cells (HK2) and mouse microglia (BV2) were performed, revealing almost no toxicity of gold clusters even at concentrations up to 200 ng/μl (fig. S10). Furthermore, the ability of Au₂₂ and Au₂₁Cu₁ clusters to regulate the level of oxidative stress was shown in vitro. As shown in Fig. 6 (B to E) and fig. S11, Au₂₂ and Au₂₁Cu₁ clusters markedly increased the survival of HK2 and BV2 cells treated with cisplatin. Meanwhile, fluorescence staining illustrates that the level of reactive oxygen species (ROS) increased after cisplatin and lipopolysaccharide (LPS) treatment in HK2 cells but decreased with the intervention of gold clusters (Fig. 6, F to H, and fig. S12). More precisely, the level of excess O₂^•− after cisplatin and LPS stimulation was significantly reduced after the intervention of Au₂₂ clusters (Fig. 6, F to H, and fig. S12). The corresponding quantitative results for the level of total ROS and O₂^•− in different treatment groups were detected by a flow cytometer (Fig. 6, I to N, and fig. S13), and the results showed the same trend as cytofluorescent staining, indicating that gold clusters can effectively reduce oxidative stress at the cellular level and exhibit favorable biological activity.

Fig. 6. In vitro treatment of cisplatin-induced AKI with gold clusters.

(A) Schematic diagram of cellular level experiments. HK2 cell viability in the presence of cisplatin with or without treatment of Au₂₂ clusters (B) and Au₂₁Cu₁ (C) determined by CCK8 assays (n = 3 per group). BV2 cell viability in the presence of cisplatin with or without treatment of Au₂₂ clusters (D) and Au₂₁Cu₁ (E) determined by CCK8 assays (n = 3 per group). (F) Fluorescence microscopic images of intracellular ROS and O₂^•− levels induced by cisplatin with or without gold clusters. Quantitative analysis of ROS (G) and O₂^•− (H) fluorescence intensity (n = 3 per group). (I to K) Fluorescence quantification of HK2 cells staining for ROS and O₂^•− by flow cytometry (n = 3 per group). (L to N) Fluorescence-activated cell sorting (FACS) results and fluorescence quantification of BV2 cells staining for ROS and O₂^•− by flow cytometry (n = 3 per group). The level of TNF-α in the cellular supernatant of HK2 cells (O) and BV2 cells (P) (n = 2 per group). (Q) Representative fluorescence image of TLR4 in HK2 cells. (R) Quantitative analysis of TLR4 fluorescence intensity. Data are presented as means ± SD; *P < 0.05, **P < 0.01, and ***P < 0.001 compared with the control group and cis group, analyzed by one-way analysis of variance (ANOVA) with the Tukey test.

To simulate and investigate the influence of AKI on the nervous system, HK2 cells were stimulated with cisplatin to simulate peripheral inflammatory conditions in vitro. The inflammatory mediators produced by HK2 cells were transferred into the medium of BV2 cells to explore the relationship between AKI and neuroinflammation as well as the therapeutic effects of Au₂₂ clusters. As shown in Fig. 6A, the content of inflammatory cytokines [tumor necrosis factor–α (TNF-α)] in the supernatant of BV2 cells was detected by enzyme-linked immunosorbent assay (ELISA). The results showed that the inflammatory cytokines that stimulated cisplatin in HK2 cells were reduced after treatment with Au₂₂ and Au₂₁Cu₁ clusters (Fig. 6O). Figure 6P demonstrates that hyperinflammatory HK2 cells induce a high production of inflammatory cytokines in BV2 cells, while the level of inflammatory cytokines in BV2 can be significantly reduced after intervention of HK2 cells with gold clusters. This suggests that peripheral inflammatory mediators produced by kidney cells can effectively mediate the inflammation of the nervous system and that bioactive gold clusters can effectively reduce neuroinflammation by eliminating peripheral inflammation, exhibiting excellent systemic inflammatory regulation.

Toll-like receptor 4 (TLR4) is important in inducing inflammatory factors and will overexpress and further induce inflammatory factors after external stimulation (61). After being stimulated by cisplatin, the TLR4 expression in HK2 cells is markedly elevated, as displayed by immunofluorescence. In contrast, gold clusters visibly down-regulated TLR4 levels, which, in turn, reduced inflammation at the cellular level (Fig. 6, Q and R). The combined cellular-level results indicate that cisplatin caused a large amount of peripheral inflammation by increasing the expression of TLR4, which induced microglia to produce large amounts of inflammatory factors. The bioactive gold clusters could alleviate peripheral inflammation by reducing the TLR4 expression, achieving the effect of modulating neuroinflammation.

Modulation of the oxidative stress and inflammation induced by cisplatin

Cisplatin, widely used for cancer treatment, is limited by strong side effects, especially nephrotoxicity. To reveal the biological activity of Au₂₂ clusters, AKI induced by cisplatin is used, which is characterized by renal-function decline, causing ischemic injury (Fig. 7A), and associated with severe ROS and inflammation, pronounced mortality, and comorbidities in other organs such as the brain (32, 62). High levels of serum creatinine (CREA) and blood urea nitrogen (BUN) are clinical manifestations of a build-up of nitrogenous wastes and can serve as effective indices of kidney excretory function (30, 62, 63). CREA and BUN levels are indeed increased in AKI groups over time, indicative of severe kidney injury (Fig. 7B and fig. S14). However, high levels of CREA and BUN can be inhibited by Au₂₂ and Au₂₁Cu₁ at 72 hours after intervention, suggesting the restoration of renal function. Proteinuria, which also exhibits progressive deterioration of kidney function (64), shows the same trend as CREA and BUN (fig. S15). Hematoxylin and eosin (H&E) staining further illustrated that cisplatin can induce severe nephrotoxicity over time and notable recovery after gold cluster intervention (Fig. 7C and fig. S16). As shown in Fig. 7D, the indicators of H₂O₂, malondialdehyde (MDA), SOD, and reduced glutathione/oxidized glutathione disulfide (GSH/GSSG) levels for the AKI group are increasingly serious over time.

Fig. 7. The oxidative stress and inflammation levels in kidney before and after treatment of gold clusters.

(A) Schematic diagram of AKI. (B) Serum CREA and BUN levels of mice at 6, 24, and 72 hours with or without gold cluster treatment (n = 3 per group). (C)Representative H&E staining in kidneys on day 3 after injury.(D) Levels of H₂O₂, MDA, SOD, and GSH/GSSG in kidneys at 6, 24, and 72 hours with or without cluster intervention (n = 3 per group). (E and F) Representative IL-6 and IL-1β staining in kidneys on day 3 after injury. (G) Levels of IL-6, IL-1β, CRP, G-CSF, KC, and MCP-1 in kidneys at 6, 24, and 72 hours with or without cluster intervention (n = 3 per group). (H) Representative terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling (TUNEL) staining in kidneys on day 3 after injury. (I to L) Relative quantification of H&E, IL-6, IL-1β, and TUNEL. Data are presented as means ± SEM; *P < 0.05, **P < 0.01, and ***P < 0.001 compared with the control group and AKI group, analyzed by ANOVA.

However, this decrease of GSH/GSSG and SOD can be recovered closer to the normal levels after Au₂₂ and Au₂₁Cu₁ cluster treatment. As the bioproducts of oxidative stress, lipid peroxides and H₂O₂ show higher accumulation in the kidney following AKI, indicating severe oxidative stress. Both clusters significantly inhibit the production of these harmful molecules, showing superior clearance of H₂O₂ and MDA after 72 hours, which is consistent with the enzyme-like activities in vitro (Fig. 3). Cisplatin-induced AKI can trigger immune cells to infiltrate the damaged kidney, recruiting large amounts of cytokines and chemokines and triggering a strong inflammatory response. The expression levels of inflammatory cytokines are significantly up-regulated following AKI at 6 hours after injury, indicative of strong local inflammation. Au₂₁Cu₁ can sharply down-regulate the inflammatory cytokines including monocyte chemoattractant protein-1 (MCP-1), granulocyte colony-stimulating factor (G-CSF), interleukin-6 (IL-6), IL-1β, keratinocyte-derived chemokine (KC), and complement-reactive protein (CRP), illustrative of anti-inflammation effects (Fig. 7, E to G, J, and K). AKI groups exhibit slight apoptosis in the kidney at 6 hours after injury and severe apoptosis at 24 hours even to 72 hours after injury (Fig. 7, C and I). However, the apoptosis of kidney can be rescued by Au₂₂ clusters with prominent recoveries at 24 hours (Fig. 7, H and L, and fig. S17). These results demonstrate that Au₂₁Cu₁ clusters have great effects on the local protection of the kidney via ROS scavenging and inflammation reduction.

Kidney disease often exhibits multiple organ dysfunction (65) that is caused, in part, by marked connectivity between the kidney and other organs and tissues, especially the brain (Fig. 8A) (62, 66). Cisplatin-induced AKI produces harmful molecules that can act on the brain and consume amounts of endogenous antioxidants in the brain. Au₂₁Cu₁ clusters can lower H₂O₂ and MDA and restore the SOD and GSH/GSSG in the brain (Fig. 8, B to D, and fig. S18), indicating ROS scavenging ability of Au₂₁Cu₁ clusters. As neurons and glial cells are the main components in brains, studies with markers for astrocytes [glial fibrillary acidic protein (GFAP)], neurons (NeuN), and microglia (Iba-1) were performed (Fig. 8, E to G and N to P, and fig. S19). The results reveal that neurons are sharply depleted, and astrocytes and microglia are activated after AKI. In contrast, with the intervention of Au₂₂ and Au₂₁Cu₁ clusters, most of these nerve cells are retrieved. Besides, the morphology of AKI-activated astrocytes can be recovered to near-normal levels after intervention with clusters. Meanwhile, immunostaining illustrates that neuroinflammation is induced because of AKI, and Au₂₁Cu₁ clusters can relieve the inflammatory response (Fig. 8, H and Q, and fig. S20). ELISA further validates the immunostaining results that the Au₂₁Cu₁ clusters are capable of altering several inflammatory cytokines in brain tissues (Fig. 8, J to M, and fig. S20), therefore alleviating inflammation induced by AKI. H&E staining further confirmed that cisplatin could exhibit side effects on the brain, while Au₂₂ and Au₂₁Cu₁ clusters can inhibit the corresponding neuroinflammation (Fig. 8, I and R, and fig. S21). These results further confirmed that single atom–engineered Au₂₂ clusters with enzyme-mimicking activity show great effects on oxidative stress– and inflammation-related diseases.

Fig. 8. The oxidative stress and inflammation levels in the brain before and after treatment of gold clusters.

(A) Schematic diagram of oxidative stress and inflammation in the brain. (B to D) Levels of MDA, H₂O₂, and SOD in the brain at 6, 24, and 72 hours after AKI with or without cluster intervention (n = 3 per group). (E to I) Staining results of astrocytes (GFAP), neurons (NeuN), microglia (Iba-1), TNF-α, and H&E staining in the brain with or without gold clusters at 72 hours after AKI. (J to M) The levels of inflammatory cytokines including IL-1β, IL-6, KC, and CRP in brain tissues using ELISA kits at 6, 24, and 72 hours after AKI with or without gold cluster treatment (n = 3 per group). (N to R) Relative quantification of astrocytes (GFAP), neurons (NeuN), microglia (Iba-1), TNF-α, and H&E staining in the brain with or without gold clusters at 72 hours after AKI (n = 3 per group). Data are means ± SEM; *P < 0.05, **P < 0.01, and ***P < 0.001 compared with the control group and AKI group, as analyzed by ANOVA.

It is assumed that inflammatory response in the brain can be attributed to the amplification of peripheral inflammation, oxidative stress, and dysregulation of the tryptophan-kynurenine pathway related to neurological diseases (Fig. 9A). The accumulation of cisplatin triggers increased production of inflammatory cytokines and ROS, gradually stimulating inflammation, oxidative stress, vascular injury (65), and apoptotic pathways (67). In addition, cisplatin has been recognized as an activating ligand for TLR4, which has been implemented as a major contributor in driving pathogenesis and development of AKI through up-regulation of inflammation and subsequent renal dysfunction, renal injury, and nephrotoxicity (68). Western blots confirmed that TLR4 increased in AKI groups and decreased with Au₂₂ cluster intervention in the kidney (Fig. 9, B and C), which may contribute to brain-kidney interactions by activation of the innate immune system during AKI through TLR4 (61). Meanwhile, the activation of TLR4 can activate the downstream inflammation, resulting in peripheral inflammation, where peripheral inflammatory cytokines can cross the blood-brain barrier (BBB) and induce neuroinflammation (65, 69). Levels of inflammatory cytokines in serum including MCP-1, G-CSF, IL-6, IL-1β, KC, and CRP perform similar trends to the kidney and brains (Fig. 9, D to G, and fig. S22), further confirming that cisplatin-induced AKI could up-regulate proinflammatory cytokines and evoke neuroinflammation. However, with the intervention of Au₂₂ clusters, the levels of inflammatory cytokines in serum were down-regulated, demonstrating that clusters can alleviate local and global inflammation. Both clinical and experimental studies have shown that peripheral inflammation is associated with BBB disruption (70). To observe the integrity of the BBB, we studied the tight junction protein ZO-1, which acted as an important indicator for BBB. The results revealed that cisplatin-induced AKI can lead to the slight degradation of ZO-1 (Fig. 9H), resulting from the altered BBB permeability. Combined with brain uptake of clusters (fig. S23), it is rational to conclude that few clusters can enter into brains due to the ultrasmall size and slightly altered BBB permeability induced by high doses of cisplatin. Then, the Au₂₂ clusters scavenge ROS and alleviate inflammation by inhibiting glia activation owing to the enzyme-like activities of clusters. In addition, we hypothesized that metabolic small molecules would also be involved in kidney-brain interactions. The results demonstrated that cisplatin-induced AKI indeed influences metabolic small molecules such as 5-hydroxy-l-tryptophan and kynurenic acid (Fig. 9I and figs. S24 and S25), involved in the tryptophan kynurenine pathway, which can link the interaction of the brain and kidney (71, 72). Meanwhile, the Au₂₂ clusters can regulate the disorder of these metabolic small molecules to near-normal levels. Thus, the kidney-brain interaction can be attributed to peripheral inflammation and dysregulation of the tryptophan-kynurenine pathway.

Fig. 9. Kidney-brain interaction with gold clusters in AKI.

(A) Possible signal pathways in cisplatin-induced AKI leading to brain inflammation. (B) Western blots for TLR4 in the kidneys at 24 hours with cluster intervention after AKI and (C) the corresponding quantitative analysis. (D to G) ELISA quantitative analysis of proinflammatory cytokine levels in serum at 6, 24, and 72 hours with or without cluster intervention (n = 3 per group), respectively. (H) Immunofluorescence of tight junction protein ZO-1. (I) Comparison of differential metabolites in serum at 72 hours after cluster intervention among the control group, the AKI group, and the Au₂₂ group. Data are means ± SEM; *P < 0.05, **P < 0.01, and ***P < 0.001 compared with the control group and AKI group, as analyzed by ANOVA.

To further illustrate the effects of clusters on neuroinflammation, we used an LPS-induced brain injury model. LPS stimulation has been used as a useful model for the study of mechanisms underlying neuronal injury, which promotes the secretion of proinflammatory cytokines (68, 73). ROS has been proven as a causative factor in the damage of the BBB. Once BBB breakdown occurs, clusters could enter BBB. The clusters uptake in the brain from LPS-stimulated mice suggested that clusters can penetrate the brain across the BBB (fig. S26). Figures S27 to S29 reveal that LPS stimulation can induce oxidative stress and inflammation around the whole body. With the intervention of clusters, levels of indicators for oxidative stress and inflammation in serum, brain, and kidney can all be down-regulated. Glia activation and apoptosis of neurons can also be observed in the brain after LPS stimulation (fig. S29), indicative of strong neuroinflammation. Meanwhile, the Au₂₂ cluster–treated groups showed a decrease in the numbers of activated glia and apoptotic neurons. As astrocyte reactivity and glia activation are mediated by ROS, the mechanism of clusters for anti-inflammation can be ascribed to the antioxidative and enzyme-like activities of the Au₂₂ clusters.

Biocompatibility and pharmacokinetics of gold clusters

Biosafety, especially excretion, is crucial for NIR-II in vivo imaging probes. The Au₂₂ clusters show rapid accumulation in the kidneys and bladder (figs. S30 and S31A). Time-course fluorescence imaging reveals that the signal in the kidney and bladder gradually decreases. No obvious signal of the kidney is observed at 72 hours after injection with ultrahigh concentration at 50 mg/ml (fig. S31B). The signal from isolated organs like the lung, liver, kidney, and spleen uncovers that Au₂₂ clusters can accumulate most in the kidney and can be excreted rapidly by renal clearance after injection (fig. S31C and D), consistent with their ultrasmall size (Fig. 2, B and C). The profiles exhibit that the size of the kidney is about 6.29 mm, close to the renal size of 5.1 mm, and further demonstrates the high spatial resolution of Au₂₂ clusters in NIR-II imaging (fig. S31E). Biodistribution confirmed that the kidney and bladder are the dominant concentrated organs of Au₂₂ clusters, indicative of renal clearance and good biosafety (fig. S31F). Au₂₂ and Au₂₁Cu₁ clusters show the blood half-time of 30.6 and 23.2 min, respectively, consistent with the results performed by fluorescence imaging of renal excretion and blood samples by analyzing the content of Au and Cu elements (fig. S32). In addition, the average urine excretion of Au₂₂ clusters is 85% (fig. S33), further illustrating the superior renal clearance. Meanwhile, no obvious toxicity is found at ultrahigh injection doses in organs, biochemistry, and hematology (figs. S34 to S36).

Noninvasive and hypersensitive NIR-II imaging combined with high-efficiency biocatalytic intervention is crucial for serious diseases, and it is urgent to develop biosafety diagnostics and therapeutics with nanotechnology. The gold clusters exhibited excellent NIR-II fluorescence with high temporal-spatial resolution and were used for imaging critical diseases such as tumors and traumatic brain injury (TBI) (44, 74). In addition, gold clusters showed potential for the biocatalytic treatment of diseases associated with oxidative stress and inflammation such as TBI and Parkinson’s disease, based on their high enzyme-like activities (11, 75). As a precise approach, manipulation at the single-atom level can minimally affect the recombination between electrons and holes, thereby ensuring no fluorescence loss. The present approach applies not only to cluster systems but also to other semiconductor quantum dots, polymers, rare earth up-conversion nanoparticles, and emerging 2D material systems (76–80). The current biocatalytic activity focuses on mimicking oxidoreductases, and atomic engineering can design a variety of non-oxidoreductases such as transferases and hydrolases in luminescence gold clusters by precisely creating catalytic active sites, which is crucial to chemical biology. At the same time, atomized clusters with biocatalysis can be used for precise tracking of intracellular molecular-level therapy and real-time molecular dynamic pathological detection via infrared emission. Therefore, single atom–engineered clusters can endow materials with biocatalysis and NIR-II luminescence, which has important scientific significance and clinical value in the molecular therapy and evolution of oxidative stress diseases.

DISCUSSION

In summary, we reported an atom engineering approach to create a Cu single-atom active site in atom-precision Au₂₂ clusters with ultrahigh enzyme-mimicking activity and without fluorescence loss. Compared with pure Au₂₂ clusters, the antioxidant activity of Au₂₁Cu₁ clusters is increased 18-fold by introducing the Cu single-atom active site. Besides, Au₂₁Cu₁ clusters show 90-fold and 3-fold higher CAT-like and higher SOD-like activity than those of Au₂₂ clusters, respectively. DFT calculations manifest that Au₂₁Cu₁ clusters have strong catalytic activity due to the lost electron state. Real-time fluorescence imaging shows that gold clusters can achieve early disease surveillance behavior. The biological experiment results show that the Au₂₂ clusters can inhabit the oxidative stress and inflammation levels in the injured kidney and brain. Last, ultrasmall gold clusters can be excreted rapidly by renal clearance, revealing nontoxicity at a high dose. In conclusion, the Au₂₂ clusters could be a biosafety, efficient, and multifunctional platform for developing multipurpose agents of high performance of NIR-II fluorescence and enzymatic catalytic activities via atom engineering. In addition, the Au₂₂ clusters show great potential for real-time imaging and early intervention of oxidative stress– and inflammation-related diseases like AKI and are promising for clinical translation.

MATERIALS AND METHODS

Experimental design

The primary objective of the experiment was to find out whether gold clusters with enzyme-like activity can have a mitigating effect on biological diseases. We designed an atom-precise Au₂₂ cluster with biocatalytic mimetic enzymatic activity and fluorescence emission in the NIR-II window. We found that the Au₂₂ cluster can be manipulated again at the atomic level to achieve multiple enzymatic activities using a series of physicochemical characterization and enzyme-like tests. The biological experiment results show that the Au₂₂ clusters can monitor the cisplatin-induced AKI and achieve 3D imaging under light-sheet microscopy, as well as inhibit the oxidative stress and inflammation levels partially in the injured kidney and brain.

Materials

All chemicals are commercially available with the highest purity and used without further treatment. Gold chloride (HAuCl₄·3H₂O) was purchased from Sigma-Aldrich; sodium hydroxide (NaOH), sodium borohydride (NaBH₄), silver nitrate (AgNO₃), zinc nitrate [Zn(NO₃)₂], copper nitrate [Cu(NO₃)₂], cadmium nitrate [Cd(NO₃)₂], erbium nitrate [Er(NO₃)₃], and GSH were purchased from Aladdin. Ultrapure water was used for all the experiments.

Materials preparation

Au₂₂(SG)₁₈ clusters were synthesized according to the reported method. In detail, aqueous solutions of HAuCl₄ (20 mM, 12.5 ml) and GSH (50 mM, 7.5 ml) were added to water (180 ml) and stirred vigorously for 2 min. Then, an aqueous NaOH solution (1 M) was added to adjust pH to 12.0, followed by the addition of 0.24 mg NaBH₄ dissolved in 0.1 ml solution, stirring at 500 rpm at room temperature for 30 min. An aqueous HCl solution (0.33 M) was added to adjust the pH to 2.5. The reaction solution was then sealed airtight and stirred at 200 rpm for 8 hours. For further purification of Au₂₂(SG)₁₈ clusters, we used ultrafiltration tubes of 3 K at 3500 rpm for ultrafiltration to remove impurities for further testing and application. Au₂₁M₁(SG)₁₈ clusters are synthesized on the basis of the same method. The only difference was that HAuCl₄ solution (20 mM, 12.5 ml) was mixed with Ag⁺, Zn²⁺, Cu²⁺, Pt²⁺, Cd²⁺, and Er³⁺ metal ions (20 mM, 0.57 ml), respectively.

Materials characterization

The clusters dispersed in ethanol solution were prepared on copper or molybdenum grids by drop-drying for TEM and EDS on a JEM-2100F TEM. Meanwhile, more than 100 clusters were analyzed to measure the average size of the clusters. DLS was measured by a Malvern Zetasizer nano ZS90 (UK) to test the hydrodynamic size of clusters. Ultraviolet-visible (UV-vis) absorption spectra were measured in aqueous solution in the range of 300 to 900 nm using a Shimadzu 3600 UV-vis-NIR spectrophotometer, and the background was corrected for contribution from water. ESI-MS was recorded on the Bruker microTOF-Q system. XPS of the gold cluster was obtained by a K-Alpha spectrometer with a monochromatic Al Kα x-ray source operating at 300 W (Thermo Fisher Scientific). The metallic elements were determined by using ICP-MS with Agilent 7900 ICP-MS. The EXAFS data were obtained at the 1W1B station (2.5 GeV, maximum current of 230 mA) of the Beijing Synchrotron Radiation Facility and analyzed by the ATHENA and ARTEMIS software.

Enzyme mimicking activity test

Total antioxidant capacity test (ABTS rapid method)

The total antioxidant capacity of clusters was determined by the rapid ABTS method using a prepared ABTS^•+ aqueous solution. The antioxidant capacity was analyzed by monitoring the absorption value at 414 nm using a UV-vis spectrophotometer. In the concentration-dependent experiments, we adjusted the concentration of clusters to 0.45, 0.9, 1.8, 2.4, and 3.5 μg/μl quantified by characteristic absorption value (the molar extinction coefficient of clusters: ɛ_{520 nm} = 4199 M⁻¹ cm⁻¹).

ONOO⁻ scavenging test

ONOO⁻ was prepared on the basis of methods from previous reports. Specifically, the cold aqueous solution of NaNO₂ (50 mM, 5 ml) and H₂O₂ (50 mM, 5 ml) was rapidly stirred under an ice bath. Then, HCl (1 M, 2.5 ml) and NaOH (1.5 mM, 2.5 ml) aqueous solutions were quickly added, and stirring continued for 5 min to obtain the yellowish ONOO⁻ solution. The scavenging capacity of ONOO⁻ was determined by monitoring the absorption value at 302 nm using a UV-vis spectrophotometer after the addition of 3.5 μg/μl clusters.

SOD-like test

The SOD-like activity of clusters was determined according to the description in the SOD activity assay kit. The SOD-like activity was evaluated by monitoring the absorption value at 560 nm using a UV-vis spectrophotometer.

CAT-like test

The CAT-like activity of clusters was tested by monitoring the characteristic absorption value of H₂O₂ at 240 nm using a UV-vis spectrophotometer. The optical density decreases with the decomposition of H₂O₂ resulting from the CAT-like activity of clusters. The reaction solutions contained 40 mM H₂O₂ and 0.35 μg/μl of clusters in 200 μl of aqueous solution.

Computational details

The calculations were performed using the Vienna Ab initio Simulation Package code, which is based on the framework of DFT. The elemental core and valence electrons were represented using the projector augmented wave method, and the cutoff energy for the plane-wave basis was chosen as 450 eV. The Perdew-Burke-Ernzerhof functional, one of the most famous generalized gradient approximation functionals, is used to describe the exchange-correlation potential. The model was built by placing the undoped and Cu-doped Au₂₂ clusters in a 23 Å by 23 Å by 30 Å box to avoid the interaction between molecules due to periodicity. In calculations, a Gamma-centered 1 by 1 by 1 grid was used to sample the Brillouin zone of the system. The convergence criteria of the electronic energy and forces were set to 10⁻⁵ eV and −0.03 eV/Å, respectively. In this work, the doping formation energies (E_f) of two different Cu-doped Au₂₂ structures were calculated by the following expression

$$E_{\mathrm{f}}=E_{\mathrm{Cu}-\mathrm{doped}}-E_{\mathrm{undoped}}+E_{\mathrm{Au}}-E_{\mathrm{Cu}}$$ (1)

where E_Cu−doped and E_undoped are the total energies of Cu-doped and undoped Au₂₂ clusters, respectively. E_Au and E_Cu are energy per atom for the Au and Cu atoms, respectively. The negative value of E_f indicated the stability of the structure with doped impurity atoms.

To investigate the adsorption behaviors of various free radicals on Au₂₂ and Au₂₁Cu₁, it is necessary to determine their adsorption energy. The adsorption energy (E_ads) of free radicals on Au₂₁Cu₁ was calculated by the following expression

$$E_{\mathrm{ads}}=E_{\mathrm{total}}({\mathrm{Au}}_{22}/{\mathrm{Au}}_{21}{\mathrm{Cu}}_1+\mathrm{adsorbate})-E_{\mathrm{total}}({\mathrm{Au}}_{22}/{\mathrm{Au}}_{21}{\mathrm{Cu}}_1)-E_{\mathrm{total}}(\mathrm{adsorbate})$$ (2)

where E_total(Au₂₂/Au₂₁Cu₁ + adsorbate), E_total(Au₂₂/Au₂₁Cu₁), and E_total(adsorbate) represent the total energy of Au₂₂/Au₂₁Cu₁ with adsorbed species, the Au₂₂/Au₂₁Cu₁ nanocluster itself, and the adsorbed species, respectively.

In vitro cell experiments

Mouse microglia BV2 cells were obtained from Tianjin Huanhu Hospital. The human renal tubular epithelial cell line (HK-2) was purchased from Procell (Wuhan, China). BV2 cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) (Gibco), supplemented with 10% FBS (BI) at 37°C with 5% CO₂. HK2 cells were cultured in minimum essential medium, supplemented with 10% FBS (BI) at 37°C with 5% CO₂. Penicillin (100 U/ml) and streptomycin sulfate (100 mg/ml; Solomo) were applied according to the growth state.

Cytotoxicity assay

BV2 cells (3 × 10³) and HK2 cells (3 × 10³) were seeded in 96-well plates overnight. The culture medium was replaced by different doses of Au₂₂ and Au₂₁Cu₁ clusters dissolved in the DMEM, and then cells were incubated for another 24 hours. Cell toxicity was analyzed by CCK8 assay.

Cell viability

BV2 cells (3 × 10³) and HK2 cells (3 × 10³) were cultured in a 96-well plate in 100 μl of culture media. When roughly 60% confluence was reached in each well, cells were stimulated by 20 μM cisplatin or LPS (200 μg/ml) for 6 hours. Then, the culture media were substituted by fresh media containing Au₂₂ and Au₂₁Cu₁ clusters at different doses, and the cells were incubated overnight. Cell viability was determined by CCK8 assay.

Measurement of intracellular oxidative stress

BV2 cells (2 × 10⁵) and HK2 cells (2 × 10⁵) were cultured in a six-well plate in a 2 ml culture medium. Cells were grown to 60% confluence and treated for 6 hours under 20 μM cisplatin or 200 μg/ml LPS conditions. The solution was replaced by a fresh culture medium with 10% FBS containing Au₂₂ and Au₂₁Cu₁ clusters (100 ng/μl), and cells were incubated for another 18 hours. The intracellular oxidative stress was measured using various fluorescent probes, including 2′,7′-dichlorofluorescein diacetate for general ROS, and dihydroethidium for superoxide. According to the protocols from the manufacturer, cells were kept at 37°C for the appropriate period in the dark. Intracellular oxidative stress was captured by using a fluorescence microscope (EVOS, AMG), data were collected by a fluorescence-activated cell sorting flow cytometer (BD AccuriTM C6), and Flowjo 10.6.2 was used for quantitative analysis of free radicals.

Enzyme-linked immunosorbent assay

BV2 cells (2 × 10⁵) and HK2 cells (2 × 10⁵) were cultured in a six-well plate in a 2 ml culture medium. HK2 cells were grown to 60% confluence and treated for 6 hours under 20 μM cisplatin conditions. The solution was replaced by a fresh culture medium with 10% FBS containing Au₂₂ and Au₂₁Cu₁ clusters (100 ng/μl), and cells were incubated for another 24 hours. HK2 cell supernatant was removed, half of which was used for ELISA. The other half of the supernatant was used to replace the original culture medium of BV2 cells. After 24 hours, the BV2 cell supernatant was removed for ELISA.

Live-cell staining

HK2 cells (2 × 10⁵) were cultured in a 24-well plate with a coverslip on the bottom. HK2 cells were grown to 60% confluence and treated for 6 hours under 20 μM cisplatin conditions. The solution was replaced by a fresh culture medium with 10% FBS containing Au₂₂ and Au₂₁Cu₁ clusters (100 ng/μl), and cells were incubated for another 24 hours. Then, the cells were washed three times in PBS and fixed into 4% paraformaldehyde (PFA) for 30 min. Cells were washed three times in PBS again and permeabilized with 0.5% Triton X-100 followed by 10% goat serum for 20 min at room temperature. Subsequently, the sections were incubated in primary antibody (anti-TLR4, Bioss, BS-20594R) overnight at 4°C. After being washed with PBS three times, the cells were incubated with secondary antibodies for 2 hours at room temperature. Subsequently, the cell nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI) for 5 min, and cells were viewed under a fluorescence microscope.

In vivo and ex vivo animal experiments

All animal testing was authorized in accordance with the standards established by the Institute of Radiation Medicine, Chinese Academy of Medical Sciences. Male C57 mice (6 to 10 weeks) were used for all acute toxicity experiments. Male C57BL/6J mice (6 to 8 weeks) were purchased from Beijing Vital River Laboratory Animal Technology Co. Ltd. Mice were kept in a standard 12-hour light/12-hour dark cycle under specific pathogen–free conditions and were allowed food and water ad libitum.

Model construction and kidney collection of mice

Mice were treated with cisplatin to induce kidney injury and cerebral complications. The mice were divided into two groups. The nephritis group was injected intraperitoneally with 200 μl of cisplatin (20 mg/kg). The control mice were injected intraperitoneally with 200 μl of 0.9% saline. All mice were raised under 12-hour dark/light cycles and used for NIR-II imaging. Au₂₁Cu₁(SG)₁₈ was injected into the tail vein. Two minutes later, we dissected the mouse and took out the kidneys for fixation and solvent cleared.

The mice were intraperitoneally injected with LPS (10 mg/ml), followed by the intravenous injection of gold clusters for the mice in the LPS + cluster group at the dose of 5 mg/ml with 200 μl. The injection volume of LPS or nanozyme was 200 μl.

Fluorescent labeling and tissue collection

Two groups of C57 mice (nephritis and control groups) were first injected with 200 μl of Au₂₁Cu₁(SG)₁₈ (5 mg/ml) through the tail vein. Mice were euthanized 2 min after injection, and kidneys were taken out. The kidneys were cleaned twice in PBS before being preserved at room temperature in 10% neutral-buffered formalin. After fixation in formalin overnight at 4°C, the kidneys were solvent cleared using 3DISCO. The 3DISCO method is convenient and fast, and has little effect on the fluorescence signal. Tetrahydrofuran was used to dehydrate the sample and used in concentrations of 50% (v/v), 70% (v/v), 80% (v/v), and 100% (v/v). Dichloromethane was used as the lipid-solving agent. Dibenzyl ether (DBE) was used to match the refractive index. The sample can be stored in the DBE after being processed. During imaging, DBE was used as an immersion solvent.

Imaging in vitro

The effective numerical aperture (NA) of the imaging system was NA = nsinα = nsin[arctan (D/2f)], where n is the refractive index, α is half of the aperture angle, D is the illumination pupil diameter of the light sheet [adjusted by holographically controlling of spatial light modulator (SLM)] as it exits the illumination objective, and f is the focal length of the illumination objective. D can be measured by placing scattered paper close to the aperture of the illumination objective when given a mask on SLM. Furthermore, it can be changed by adjusting the coordinate factor u. The experimentally measured beam parameters (light waist and Rayleigh length) were consistent with the theoretical values based on effective NA values. For all imaging, the effective NA of the illumination objective was 0.27, the coordinate factor u corresponding to SLM is 1, α = 7, and the resulting imaging field of view is 632 μm. Uncleared tissue is imaged using 1X PBS immersion, and cleared tissue is imaged using DBE-matched refractive index. When switching between different wavelengths and beam types for imaging, we kept the beam power at the back aperture between 3.5 and 12.2 mW and kept it constant. The displacement stage is linked to the camera exposure time to scan biological samples in steps of 500 nm. The exposure time of the biological samples was 20 ms.

Imaging in vivo

For real-time NIR-II imaging, Au₂₂ clusters of 200 μl with a concentration of 10 mg/ml were intravenously injected after using 180 μl of 5% chloral hydrate solution. The mouse was mounted on the platform below the 808 nm diode laser for real-time monitoring. A 2D InGaAs array (Princeton Instruments, 640 × 512 pixels) was operated to collect images in NIR-II regions. Excitation light was provided by an 808 nm diode laser coupled to a 4.5 mm collimator. The fluorescence signal in the NIR-II region was collected by the detector using a 1000 nm long pass filter.

Kit detection

Mice were anesthetized with isoflurane and cleaned with 10 ml of PBS perfusion on day 3 after AKI. Kidney and brain samples were harvested quickly from the body. Homogenates were centrifuged at 6000g for 15 min three times, and the supernatant was stored at −80°C for preparation. Protein concentration was determined with an enhanced bicinchoninic acid (BCA) protein assay kit (Beyotime, P0010). Indicators for oxidative stress were detected with a lipid peroxidation MDA assay kit (Beyotime, S0131S), a hydrogen peroxide assay kit (Beyotime, S0038), a total SOD assay kit with WST-8 (Beyotime, S0101M), and a GSH and GSSG assay kit (Beyotime, S0053). Inflammatory cytokines were determined by ELISA kits (Abcam, ab100712, ab197742, and ab208348; Proteintech, KE10007, KE10003, KE1006, and KE10023, respectively). The assays were performed according to the instructions provided by the manufacturer.

Immunostaining

All organs were harvested from the mice and fixed into 4% PFA for 24 to 48 hours, embedded in paraffin, and mounted on slides (4 μm coronal sections). The slides were dewaxed in xylene and dehydrated in gradient alcohol. Then, endogenous peroxidase was extinguished, and antigen retrieval was performed in a citrate antigen retrieval solution (C1032, Solarbio, China). Immediately afterward, the slides were blocked with serum, and the primary antibody was added to the slides and incubated overnight at 4°C. The information about the primary antibody is as follows: anti–TNF-α antibody (1:150, Abcam, ab183218), anti–IL-6 antibody (1:100, Proteintech, 66146-1-lg), anti–IL-1β (1:100, Affinity, AF5130), anti–ZO-1 (1:100, Proteintech, 21773-I-AP), anti-GFAP (1:100, Proteintech, 16823-1-AP), and anti–Iba-1 antibody (1:100, Proteintech, 10904-1-AP). After being rinsed three times, the secondary antibody was incubated with slides in the dark for 1 to 1.5 hours. The information about the secondary antibody is as follows: goat anti–rabbit immunoglobulin G (IgG) H&L (Alexa Fluor488) (1:300, Abcam, ab150077), CoraLite488-conjugated Affinipure goat anti–mouse IgG (H + L) (1:100, Proteintech, SA00013-1), and Cy3-conjugated Affinipure goat anti–rabbit IgG (H + L) (1:100, Proteintech, SA00009-2). Next, the slides were mounted with an anti-fade mounting medium with DAPI (S2110, Solarbio, China) and captured with a fluorescence microscope (EVOS, AMG). Terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling (TUNEL) assay was carried out using the TUNEL apoptosis assay kit (C1089, Beyotime, China). Slides were also stained with H&E for toxicological analysis and pathological observation. The quantification was analyzed with ImageJ and assessed by the target expression levels.

Western blotting

Tissue samples (n = 3 per group per point) were lysed and operated on ice, and protein was extracted from brain homogenates in radioimmunoprecipitation assay lysis buffer (strong) (CWBIO, CW2333) containing protease inhibitors at 95°C for 5 min. Tissue extract supernatant protein concentrations were determined by a BCA assay (Beyotime, P0010). SDS–polyacrylamide gel electrophoresis was performed to resolve protein lysates (50 μg) before protein lysates were transferred onto nitrocellulose membranes (CWBIO, CW0022S). Antibodies specific for TLR4 (Abcam, ab217274) were used.

BBB penetration

Mice in deep anesthesia were perfused with 0.01 M PBS to clean out blood in blood vessels at the point of 1, 4, 6, 12, 24, and 72 hours (n = 3 per group per point). Brain tissues were harvested and weighed, and the Au element was detected on ICP-MS to evaluate the BBB penetration.

Targeted analysis of metabolites by UHPLC-Q/Orbitrap MS

The targeted metabolomics analysis of serum was carried out with the help of the Tianjin University of Traditional Chinese Medicine Analytical Testing Center, based on ultrahigh-performance liquid chromatography tandem quadrupole Orbitrap MS, (UHPLC-Q/Orbitrap MS). The raw data consisted of 12 quality control samples and 30 experimental serum samples. A total of 260 compounds were targeted for peak extraction, and 75% of the detectable compounds in each group were taken to build a matrix. A total of 103 compound peak areas were extracted from plasma, and the normalized areas obtained on the basis of the peak areas of internal standard compounds were used as raw data. After that, the missing values were filled by one-half of the minimum value for simulation (missing value recoding), and the data were normalized. Normalization was performed using the sum of peak areas for each sample. After obtaining the collated data, we performed principal components analysis on the multivariate variables. The results were then analyzed using the statistical method of orthogonal partial least squares–discriminant analysis. Last, the differential metabolites were obtained by permutation test and multivariate statistical analysis.

Pharmacokinetic and toxicological studies

Healthy control mice were intravenously injected with Au₂₂ (10 mg/kg in 200 μl volume) to conduct pharmacokinetic and toxicological studies and excretion. The blood was collected at different time points by the orbital venous plexus blood collection method. The element Au was detected using ICP-MS and fluorescence imaging for half-time studies. Some of the blood samples and all organs were harvested for hematology, biochemistry, and pathological studies on day 7 after injection. The blood was collected into anticoagulation tubes, and the serum was separated by centrifugation for biochemistry analysis. All organs were weighed and detected with Au content to determine the biodistribution using ICP-MS. Urine was collected at different time points within 24 hours, and the Au element was determined with ICP-MS as well to analyze the excretion of Au₂₂ clusters. The pathological studies were conducted with H&E staining and observed with a digital light microscope. For serum CREA and BUN in AKI mice on day 3 after AKI, the method is the same as above.

Statistical analysis

Data are presented as means ± SD (standard deviation) or SEM (standard error of mean). One-way analysis of variance (ANOVA) with the Tukey test was used to assess differences among groups. Asterisks indicate significant differences (*P < 0.05, **P < 0.01, ***P < 0.001).

Supplementary Materials

This PDF file includes:

Figs. S1 to S36