EM NSs were facilely prepared via the room-temperature ultrasonication of KMnO₄-treated E. coli cells followed by purification. As it was observed that the suspension turned from purple to red in the KMnO₄ incubation process, we considered that the structure of the bacterial cells was destroyed and the Mn element was completely or partially reduced. In the ultrasonication process, the Mn element might be further reduced and embedded in the oxidized bacterial fragments to form the final nanoparticles. As shown in the transmission electron microscopy (TEM) results (Figures 2(a) and 2(b), and Figure S1), the obtained EM NSs presented the spindle-like morphology with an average length of ~13.4 nm and an average width of ~1.7 nm, which were both smaller than the average hydrodynamic size (~31.0 nm, measured by dynamic light scattering (DLS); see Figure S2) due to the presence of hydration layers. The rod-shaped E. coli bacterial cells were chosen because they are Gram-negative bacteria with cell walls composed of thin peptidoglycan layers, which may be destructed after KMnO₄ and ultrasonication treatments, resulting in the formation of the spindle-like nanoparticles. We also selected the rod-shaped Bacillus subtilis (B. subtilis) and spherical Staphylococcus aureus (S. aureus) cells, the two kinds of Gram-positive bacteria with thick cell walls, and tried to prepare nanomaterials via similar KMnO₄ and ultrasonication treatments. Unfortunately, after these treatments, large bacterial cells were still observed in the obtained suspensions (Figures S3(a) and S3(b)), indicating that the short-time treatments could not completely destroy the Gram-positive bacteria. Additionally, when B. subtilis and S. aureus cells were preincubated with 1% Triton X-100 solution to increase the membrane permeability and then treated with the same KMnO₄ and ultrasonication, the spindle- and dot-like nanostructures with sizes of \text{~}13.5\times 2.2 nm and ~1.7 nm were obtained, respectively (Figures S3(c) and S3(d)). Because the rod-shaped bacteria (E. coli and B. subtilis) and the spherical bacteria (S. aureus) could be used to produce nanospindles and nanodots, respectively, we assume that the bacterial morphology affects the morphology of the synthesized nanostructures.

To date, several synthetic strategies have been proposed to prepare MnO{}_x-based nanomaterials with different sizes and morphologies, including the thermal decomposition method [61], exfoliation strategy [62], adsorption-oxidation method [63, 64], hydro-/solvothermal method [65], and permanganate reduction method [66, 67]. These methods show their intrinsic advantages and disadvantages [68]. For example, the thermal decomposition strategy can precisely regulate the particle size and morphology but the oily solvent-capping surface limits the biomedical applications of the products. The exfoliation strategy can successfully obtain two-dimensional materials, but its synthetic process is harsh and time-consuming. The particular preparation conditions of adsorption-oxidation and hydro-/solvothermal methods may also restrict their wide applications. In this work, the preparation process of EM NSs can be classified as the permanganate reduction method, which is fast, simple, and widely adopted to produce the MnO{}_x-based nanomaterials with good water dispersibility. Compared with the reported permanganate reduction methods, we creatively utilized the living bacterial cells as the reductant in this work, which has the following advantages: First, the bacterial cells as the raw materials are easily available and can be easily scaled up. Second, the whole bacterial cells are directly used as both the template and the reducing agent for the biosynthesis, without the extra steps such as the extraction of the cell walls/membranes. Third, this bacterial template synthesis is suitable for a variety of bacterial cells including Gram-negative and Gram-positive bacteria to generate the MnO{}_x-based nanomaterials with unique morphologies. Finally, the dose of KMnO₄ is low, which may ensure the good biocompatibility of the obtained nanoparticles.

Considering that the nanospindles prepared from E. coli have simpler synthetic procedures (which do not require the Triton X-100 treatment) compared with other bacterium-derived nanostructures, we chose the E. coli-derived nanospindles (namely, EM NSs) for further investigation. The X-ray diffraction (XRD) pattern of EM NSs demonstrated that the nanomaterials were in the amorphous state (Figure 2(c)). Figure 2(d) shows the Fourier transform infrared (FTIR) spectrum of EM NSs. The broad band at ~3300 cm⁻¹ originates from the stretching vibrations of O–H and N–H. The peaks at 1654, 1241, and 1077 cm⁻¹ represent the stretching vibrations of the C=O (accompanied by the contribution from the bending vibration of H₂O), C–O, and C–N groups, respectively. Besides, the peak at 1541 cm⁻¹ can be assigned to the bending vibration of N–H and/or the stretching vibration of C=C. Additionally, it was found that the characteristic FTIR peaks of EM NSs and freeze-dried E. coli (Figure S4) overlapped to a large extent, indicating that the main components of the organic materials in the EM NSs originated from the bacteria. Four elements can be observed in the X-ray photoelectron spectroscopy (XPS) total curve of EM NSs including C 1s (63.0%), O 1s (27.5%), N 1s (8.9%), and Mn 2p (0.6%) (Figure 2(e)). The valence states of key elements in EM NSs were determined according to the high-resolution curves (Figure 2(f) and Figure S5). The XPS high-resolution Mn 2p curve shows that Mn 2p_3/2 and Mn 2p_1/2 peaks center at 642.3 and 654.1 eV (Figure S5c), respectively, indicating that the main form of the Mn element is MnO₂ [69]. However, since the XPS signals were very weak due to the very low Mn content in the EM NSs, we could not rule out the presence of other manganese oxides such as MnO₃ and MnO in the NSs. In addition, the Mn element content in EM NSs was measured to be 8.6 wt% by inductively coupled plasma optical emission spectrometry (ICP-OES). The long-term stability of EM NSs was also tested in phosphate-buffered saline (PBS) or cell culture medium (Dulbecco’s modified Eagle’s medium (DMEM)), and no obvious changes of hydrodynamic size were observed within two weeks as measured by DLS (Figure S6).

Then, DOX, which can simultaneously act as the therapeutic agent and the FL probe, was loaded into EM NSs via electrostatic and hydrophobic interactions to obtain the EMD NSs. The TEM image shows the unchanged spindle-like structure of EMD NSs with an average length of ~14.2 nm and an average width of ~1.6 nm (Figures 3(a) and 3(b) and Figure S7). The change of the zeta potential value from -28.5\pm 1.6 to -23.4\pm 1.5 mV (Figure 3(c)) and the characteristic absorption peak of DOX at ~490 nm in the ultraviolet–visible (UV–vis) spectrum of EMD NSs (Figure S8) further confirmed the successful loading of DOX molecules. The nanospindles exhibited an ultrahigh DOX encapsulation efficiency of ~100% and loading efficiency of 11.1%.

Interestingly, the fluorescence of DOX was almost completely quenched in the formed EMD NSs via static quenching effect [50], and this MnO{}_x-induced quenching effect can be reversed upon the addition of GSH (Figure S9). To investigate the effect of GSH on EMD NSs, TEM analysis of EMD NSs after GSH incubation was conducted. As shown in Figures 3(d) and 3(e), it can be clearly observed that GSH could degrade the MnO{}_x-based nanospindles into smaller nanoparticles with an ultrasmall size of ~2.2 nm (which would promote deep tumor penetration and rapid renal clearance), accompanied by the release of DOX for cancer therapy. The GSH-responsive FL “turn on” process is schematically illustrated in Figure 3(f) and proven by the FL images (Figure 3(g)). Besides, the dissociation of EM NSs after GSH incubation demonstrated the destructive effect of GSH on the nanocarriers (EM NSs) (Figure S10), which could be due to the redox reaction of Mn element in high valence states in EM NSs, as shown in equation (1). We demonstrated that the “turn on” FL response of EMD NSs could be used for the sensitive and selective GSH detection. As shown in Figure S11, after the addition of varied concentrations of GSH, the FL intensity of EMD NSs increased under the 365 nm UV irradiation. The GSH concentration-dependent FL enhancement of EMD NSs was also shown in the FL spectra under 500 nm excitation (Figure 3(h)). Figure 3(i) shows the corresponding change of FL intensity (\Delta I/I_0) at 555 nm as a function of GSH concentration ([GSH]). The relationship between \Delta I/I_0 and [GSH] could be described by the following equation when [GSH] is in the range of 0–200  μM: \Delta I/I_0=0.2396+0.0470\times \left[\text{GSH}\right] (R^2=0.9925), with a detection limit of 0.28  μM (at a signal to noise ratio of 3). When [GSH] is higher than 200  μM, the \Delta I/I_0 value increased nonlinearly with increasing [GSH]. Importantly, the response of EMD NSs was highly selective towards GSH relative to other substances (common ions and small molecules) in physiological solutions. As shown in Figure S12, the FL intensity of EMD NSs markedly increased after GSH addition, while other substances such as ions (including K⁺, Na⁺, Ca²⁺, and Mg²⁺) and small molecules (including glucose, dithiothreitol (DTT), and amino acids like L-glutamic acid (Glu), glycine (Gly), L-serine (Ser), L-arginine (Arg), L-lysine (Lys), D,L-homocysteine (Hcy), and L-cysteine (Cys)) did not cause distinct FL enhancements, which was ascribed to the selective reducing ability of GSH towards the EMD NSs. It was worth noting that Hcy and Cys, the two biological thiols that may severely affect GSH detection, did not interfere with the GSH detection in our system.

Since MnO₂ is known to be an excellent catalyst to trigger the decomposition of H₂O₂ into H₂O and O₂, we tested whether EM NSs and EMD NSs can generate O₂ in vitro upon reaction with H₂O₂ at endogenous levels (e.g., 100  μM). As shown in Figure 4(a), both EM NSs and EMD NSs could trigger rapid O₂ generation within the H₂O₂ solution, demonstrating the catalytic property of EM NSs and EMD NSs to react with H₂O₂ and trigger its decomposition for O₂ generation (equation (2)), relieving the tumor hypoxia.

Because the Mn element is an excellent T₁-shortening agent in MR imaging, we investigated the ability of EMD NSs as GSH-activated MR imaging contrast agents. The in vitro longitudinal relaxation rate (1/T_1) as a function of the Mn concentration of EMD NSs before and after GSH reduction was evaluated (Figures 4(b) and 4(c)). After GSH addition, the produced Mn²⁺ ions exhibited a stronger enhancement in T₁-weighted MR imaging than the EMD NSs. Longitudinal relaxivity r_1, obtained by measuring the relaxation rate as a function of Mn concentration, exhibited a 6-fold enhancement when EMD NSs were reduced to Mn²⁺ ions by GSH.

On the other hand, we selected methylene blue (MB), a dye that can be degraded by ^⋅OH, as an indicator of ^⋅OH generation. By comparing the results of groups 1–4 in Figures 4(d) and 4(e), we inferred that H₂O₂, Mn²⁺ ions, and carbonate/bicarbonate (CO₃²⁻/HCO₃⁻) buffer solution are all necessary to fade the blue MB solution, which was in accordance with the previous work [58]. Additionally, the results of groups 4 and 5 indicated that the added GSH eliminated the formed ^⋅OH and inhibited the MB degradation. Based on the above results, we investigated the chemodynamic activity of our nanoagents. When Mn²⁺ ions were replaced by EM NSs (group 6), partial MB molecules degraded due to the catalytic activity of EM NSs. More importantly, in the presence of GSH, EM NSs exhibited an even higher MB degradation efficiency (group 7), which could be explained by the GSH-stimulated Mn²⁺ release from EM NSs accompanied with GSH depletion. These results indicated that the EMD NSs consisting of DOX and EM NSs can serve as both chemotherapeutic and chemodynamic nanoagents.

Furthermore, the in vitro drug release properties of EMD NSs in the presence or absence of GSH displayed the sustained and GSH-responsive DOX release from EMD NSs (Figure 4(f)), which is determined by the electrostatic interaction between positively charged DOX molecules and negatively charged EM NSs. Besides, at 24 h, the slightly higher cumulative drug release of EMD NSs at pH 6.5 than that at pH 7.4 might be ascribed to the gradual protonation of DOX under the acidic condition. Based on the overexpressed chemical entities in the cancer TME, such as intracellular H₂O₂ and GSH, the above H₂O₂-/GSH-responsive characters of EMD NSs are very important in selective cancer therapy.

Considering that the GSH concentration in cancer cells is much higher than that in normal cells [30, 31], which was also verified by our cellular total GSH assay results (Table S1), we would like to investigate whether the EMD NSs respond to the GSH in the two types of cells differently. In the cancer (human breast cancer MCF-7)/normal (human normal mammary epithelial MCF-10A) cell experiments (Figures 4(g) and 4(h)), it can be seen that after treatment with EMD NSs, stronger FL signals were observed in the cell nuclei of MCF-7 cells compared with those in the cell nuclei of MCF-10A cells. These results revealed the excellent cancer/normal cell discrimination ability of EMD NSs, which mainly resulted from their GSH-responsive DOX release. Based on the different DOX release behaviors of EMD NSs in cancer/normal cells, we further assessed the anticancer efficacy of EMD NSs towards three different cancer/normal cell groups by a 3-(4,5-dimethylthiazol-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide (MTT) assay. As shown in Figure 5(a), EMD NSs exhibited a stronger inhibitory/killing effect towards cancerous MCF-7 cells than normal MCF-10A cells. According to the dose-dependent cell growth curves, the half maximal inhibitory concentration (IC₅₀) value of EMD NSs for MCF-7 cells was calculated to be 2.6\pm 0.1  μg mL⁻¹, which was much smaller than that (30.9\pm 2.8  μg mL⁻¹) for MCF-10A cells (Figure 5(d)). Additionally, we also observed the stronger cancer cell inhibitory/killing effects in two other pairs of cancer/normal cells (Figures 5(b) and 5(c)), and the differences were quantified by the IC₅₀ values of EMD NSs: human lung cancer A549 cells: 14.1\pm 1.2  μg mL⁻¹, normal human pulmonary alveolar epithelial HPAEpiC cells: 84.7\pm 21.4  μg mL⁻¹, human liver cancer HepG2 cells: 0.7\pm 0.1  μg mL⁻¹, and human normal liver L02 cells: 2.5\pm 0.2  μg mL⁻¹ (Figures 5(e) and 5(f)).

On the basis of the above results, we eventually selected MCF-7 cells for our subsequent experiments to study the anticancer effects of EMD NSs. First, the cytotoxicity of the TME-responsive nanocarriers, EM NSs, was proven to be low by the MTT assay (Figure 5(g)). The viability of MCF-7 cells was concentration-dependent and remained >80% in the EM NS concentration range of 0–160  μg mL⁻¹. Importantly, using a total GSH detection kit, it was observed that the total GSH content decreased from 1.75 mM in the cells without treatment (Table S1) to 0.58 mM in the EM NS-treated cells, suggesting the GSH-depleting ability of EM NSs to enhance the CDT efficiency. Next, at low DOX concentrations (0–5  μg mL⁻¹), EMD NSs exhibited the similar cytotoxicity against MCF-7 cells compared with free DOX (Figure 5(h)). We also carried out additional experiments to explore the anticancer mechanisms of EMD NSs. After the MCF-7 cells were treated with free DOX, EM NSs, or EMD NSs for 24 h, the caspase-3 levels are shown in Figure 5(i). The results showed that the apoptosis level induced by EMD NSs was higher than that in the control group and close to that in the “free DOX” group at the same DOX concentration. According to the cell apoptosis detection results (Figure 5(j)), compared with the control group, the cells treated with EMD NSs showed significant early apoptosis (39.25%) and late apoptosis/necrosis (54.29%), which were similar to those treated with free DOX (early apoptosis: 43.14%; late apoptosis/necrosis: 49.52%). Additionally, the evident increase of G2/M phase in cell cycle distribution was observed in both “free DOX” and “EMD NSs” groups (Figure 5(k)). Therefore, the anticancer effect of EMD NSs mainly resulted from the released DOX molecules (evidenced by the confocal imaging results in Figure 4(g)) in this TME-responsive nanosystem. However, at high DOX concentrations (10–20  μg mL⁻¹), EMD NSs showed the relatively higher cytotoxicity than free DOX (Figure 5(h)), so we assumed that some other factors may also play roles in the EMD NS-induced cell death. Although EM NSs caused negligible changes in the cell viabilities, the substantial accumulation of ROS was observed in the EM NS- or EMD NS-treated cells (Figure 5(l) and Figure S13), which could be attributed to the conversion of endogenous H₂O₂ to highly reactive ^⋅OH (one of the most harmful forms of ROS) induced by the Mn²⁺-mediated Fenton-like reaction [58]. Importantly, based on the in vitro EM NS-assisted O₂ generation results (Figure 4(a)), EM NSs showed great promise in downregulating the P-gp expression by increasing the O₂ level under hypoxia. Therefore, P-gp expression was also measured by western blot analysis after different treatments. As shown in Figure 5(m), the untreated hypoxic MCF-7 cells exhibited high P-gp expression, which could promote the efflux of DOX. After 24 h incubation of EM NSs, the P-gp expression in hypoxic cells decreased to a relatively low level similar to the expression level in normoxic cells, which indicated that the EM NS-based nanoplatform could inhibit P-gp expression under hypoxia, possessing the potential to reduce drug efflux and thus overcome hypoxia-induced chemoresistance. Taken together, all these observations proved that EMD NSs can efficiently kill cancer cells by the GSH consumption- and O₂ generation-enhanced chemo-chemodynamic combination therapy.

Motivated by the in vitro results, we then performed in vivo theranostic experiments using tumor-bearing mice via the tail vein injection with free DOX or EMD NSs. It was observed that free DOX was distributed nonspecifically and rapidly to various organs and tissues after intravenous (i.v.) administration, leading to a rapid elimination from the circulation (Figure 6(a)). Meanwhile, the FL intensity of EMD NSs in the tumor areas showed a continuous increase and reached the maximum at 12 h postinjection followed by a slight decrease, whereas free DOX exhibited very weak FL signals within the whole observation time period (Figure 6(b)). Such a direct FL comparison verifies the excellent tumor targeting ability of EMD NSs, which can be attributed to their proper size (~14 nm in length) that endows them with excellent passive tumor targeting ability via the enhanced permeability and retention effect. Importantly, since the tumor tissues show at least 4-fold higher concentrations of GSH levels compared with normal tissues in the tumor-bearing mice [70], the GSH-triggered “off–on” DOX release of EMD NSs was demonstrated in vivo. As shown in Figure S14, a healthy tissue and a tumor tissue in a mouse were injected with EMD NSs, respectively. The negligible FL intensity in the normal tissue and the strong FL intensity in the tumor region provided a solid proof for the selective tumor imaging capability of the nanoagents. Apart from the in vivo FL imaging, the MR images of a tumor-bearing mouse were also acquired before and after the i.v. injection of EMD NSs. As shown in Figure 6(c), a detectable T₁-weighted contrast enhancement was observed in the tumor area at 12 h post i.v. injection, which proved that the obtained EMD NSs can be used as promising contrast agents for MR imaging application. In addition, the mice intravenously injected with EMD NSs were sacrificed at 1, 3, and 7 d postinjection, respectively. Ex vivo images and quantified FL intensities revealed that EMD NSs were distributed mainly in tumors and partially in livers and kidneys at 1 and 3 d postinjection and were still retained in the tumor at 7 d postinjection without evident distributions in other organs (Figures 6(d) and 6(e)), which indicated their excellent tumor accumulation, prolonged tumor retention, and good biosafety. Besides, the FL intensities in the kidneys were higher than those in the livers, implying that EMD NSs might be mainly captured and metabolized by kidneys due to their small size.

Subsequently, inspired by the high tumor accumulation of EMD NSs, the in vivo tumor suppression activity of the TME-responsive nanoagents was evaluated using tumor-bearing mice. As shown in Figures 7(a) and 7(c), the tumor growth of the mice intravenously injected with EMD NSs was almost completely suppressed during the period of our observation (14 days), whereas free DOX did not exhibit noticeable therapeutic efficacy. Additionally, the tumor growth of mice treated with EM NSs was also not effectively suppressed, which might be due to the inadequate administration of EM NSs. The hematoxylin and eosin- (H&E-) stained tumor slices from the EMD NS-treated group showed that most cells were severely damaged after treatment (Figure 7(b)), further confirming the satisfactory in vivo therapeutic outcome of the nanoagents. Furthermore, the tumor lung metastasis was examined on the 30th day after various treatments. As revealed by the photographs of India ink-stained whole lungs and the H&E-stained lung slices (Figures 7(e) and 7(f)), a large number of tumor nodules were observed in the lungs of untreated, DOX-treated, and EM NS-treated tumor-bearing mice, while no obvious lung metastasis was observed in the EMD NS-treated group, suggesting the excellent inhibitory effect of EMD NSs on tumor metastasis.

We also investigated the potential side effects of EMD NSs. First, no evident body weight loss was observed during the treatments (Figure 7(d)), suggesting the negligible toxic effect of EMD NSs on the animals. Next, according to the tissue slices from the EMD NS-treated mice at 14 d postinjection (Figure S15), compared with the control group, no obvious cell apoptosis/necrosis was observed in the major organs, indicating the good biosafety of the nanoagents, which may result from the efficient excretion of the nanoagents by the mice. Besides, even at the high DOX concentration of 200  μg mL⁻¹, EMD NSs did not induce hemolysis (Figures S16(a) and S16(b)), and no abnormalities in the detected blood indices were seen in the blood routine analysis for the mice sacrificed on the 14th day after EMD NS treatment (Figures S16(c–n)), showing the good hemocompatibility of EMD NSs. Collectively, the above results demonstrated that EMD NSs can be used as safe antitumor nanoagents.

E. coli and S. aureus were purchased from China Center of Industrial Culture Collection (CICC, Beijing, China). KMnO₄ was purchased from Chengdu Kelong Chemical Inc. (Chengdu, China). Doxorubicin hydrochloride (DOX) was purchased from Beijing Huafeng United Technology Co., Ltd. (China). L-Glutathione (GSH) was purchased from Sigma-Aldrich (St. Louis, MO). Dialysis membranes (Spectra/Por® 6, regenerated cellulose) with the MWCO of 10 kDa were purchased from SpectrumLabs, Inc. (Rancho Dominguez, CA). All solutions/suspensions were prepared with deionized (DI) water (18.2 MΩ·cm) purified by a Milli-Q system (Millipore, Billerica, MA). The Annexin V-Fluorescein Isothiocyanate (FITC) Apoptosis Detection Kit containing annexin V/propidium iodide (PI), Caspase-3 Colorimetric Assay Kit, Annexin V-FITC Apoptosis Detection Kit, Cell Cycle Detection Kit, and ROS Detection Kit were all purchased from Nanjing KeyGen Biotech Co. Ltd.

The TEM experiment was performed using a transmission electron microscope (JEM-2100, JEOL Ltd., Japan). Before measurements, the samples were prepared by dropping the sample suspensions onto the ultrathin carbon-coated copper grids and air-drying. The size distribution was obtained by counting at least 100 particles from one TEM image per sample using an image analysis software (Nano Measurer 1.2). For XRD analysis, the fine EM NS powder was first obtained by freeze-drying and grinding. Then, the powder was placed into the sample holder and gently pressed using a glass slide to make the surface smooth. XRD patterns were then recorded on an XRD system (D8 Discover, Bruker, Germany). The FTIR spectra of the freeze-dried EM NS powder and bacteria (E. coli) were collected with a Nicolet iS50 FTIR spectrometer (Thermo Fisher Scientific Co., Waltham, MA, USA). Additionally, the EM NS powder was sprinkled onto the surface of sticky carbon conductive tape for XPS analysis using an X-ray photoelectron spectrometer (Quantera 2000, Ulvac-Phi Inc., Japan). UV–vis spectra were obtained using a UV–vis spectrophotometer (UV-2600, Shimadzu, Japan). The FL spectra were obtained on a spectrophotometer (RF-5301PC, Shimadzu, Japan). The hydrodynamic diameter and zeta potential measurements were conducted on a Zetasizer instrument (Nano ZS, Malvern Instruments, UK). To determine the Mn concentration in the EM NS suspension, the sample was first digested with aqua regia at 60°C overnight and then diluted with DI water for analysis using ICP-OES (Perkin-Elmer Optima 2100DV, USA).

Female BALB/c nude mice were obtained from Yangzhou University Medical Center (Yangzhou, China) and used under protocols approved by the Southeast University Laboratory Animal Center. 100  μL of PBS containing 4T1 cells (2\times {10}^6) was subcutaneously inoculated onto the back of each mouse to build the xenograft tumor model.