Reactive Oxygen Species (ROS)-Responsive Nanoprobe For Bioimaging And Targeted Therapy of Osteoarthritis

Stimulus-responsive therapy that allows precise imaging-guided therapy is limited for osteoarthritis (OA) therapy due to the selection of proper physiological markers as stimulus. Based on that the over-production of Reactive Oxygen Species (ROS) is one of the leading causes of OA, we selected ROS as markers and designed a cartilage-targeting and ROS-responsive theranostic nanoprobe that is highly specic for effective bioimaging and therapy of OA. This nanoprobe was fabricated by using PEG micelles modied with ROS-sensitive thioketal linkers (TK) and cartilage-targeting peptide, termed TKC, which was then encapsulated with Dexamethasone (DEX) to form TKC@DEX nanoparticles. Results showed that the nanoprobe can smartly “turn on” in response to excessive ROS and “turn off” in the normal joint. By applying different doses of ROS inducer and ROS inhibitor, this nanoprobe can emit ROS-dependent uorescence according to the degree of OA severity, helpful to precise disease classication in clinic. Specically targeting cartilage, TKC@DEX could effectively respond to ROS and sustained release DEX to remarkably reduce cartilage damage in the OA joints. This smart, sensitive and endogenously activated ROS-responsive nanoprobe is promising for OA theranostics. of OA catabolic biomarkers IL-6 and MMP-13 in chondrocytes was assessed by immunouorescence. Chondrocytes were xed with 95% ethanol for 30 min and permeabilized with 0.1% Triton X-100 for 10 min. Samples were incubated with primary antibody as follows: IL-6 (1:200, Boster, and MMP-13 (1:200, Boster) at 4°C overnight. Then the samples treated with the secondary antibodies FITC-anti-rabbit IgG (1:50, Boster) for 60 min at 37°C and counterstained with DAPI for 5 min. the were photographed a


Introduction
Osteoarthritis (OA) is a chronic disease characterized by the progressive degeneration of cartilage that leads to joint pain and even serious disabilities of patients around the world [1,2]. At the end-stage of OA, the joints gradually lose function and need prosthetic replacements surgery [3]. Thus, it is necessary to highlight the diagnosis and treatment of OA at the early-stage.
Nowadays, endogenous-based uorescent probes relying on certain physiological markers like NO [4] and MMP-13 [5][6][7][8] to generate uorescence have been explored for detection of early OA in vivo. However, these probes depending on pathological events are limited in clinical applications mainly due to the relatively low concentrations of endogenous biochemical markers [9][10][11][12], resulting in inaccurate and insensitive signals in the deep tissues. Moreover, most probes can only be used for diagnosis/monitoring and have little or no therapeutic effects on OA. Therefore, it is imperative to nd remarkable physiological markers and design a responsive theranostic probe that can penetrate through the deep tissues and simultaneously have therapeutic potential for OA.
It is generally accepted that Reactive Oxygen Species (ROS) are important causative factors during the development of OA. It has been reported that the levels of ROS, e.g. H 2 O 2 , O 2 − , HO − , and HOCl, which maintain at a low level in normal articulatory [13], are dramatically increased (may up to 1 mM) in the joints of OA patients [14]. The over-generation of ROS elicits hyper-peroxidation, protein carbonylation and DNA damage, which has been considered as the main mechanism of cartilage cells loss and tissue damage [15]. But until now, a speci c and biocompatible ROS-responsive system has not yet been moved forward to real-time monitoring and therapy of OA.
The drug delivery systems based on ROS-responsive functional moieties such as sul de[16], phenylboronic acids and esters [17][18][19], selenium-containing linkage [20], peroxalate [21][22][23] et., have been widely used for cancer treatment. Among various materials, nano-scaled ROS-responsive polyethylene glycol (PEG) polymers composed of thioketal linkers attracted most intention because of their sensitivity and responsiveness to endogenous ROS down to submicromolar concentrations (~ 100 µΜ) at deep physiological signaling levels [24][25][26]. PEG is usually employed to self-assemble micelles due to their excellent biocompatibility, which can shield from the mononuclear phagocyte system (MPS) and prevent host rejection after injected in vivo [27]. Moreover, the PEG micelles of su ciently small size from 20 to 200 nm facilitates entry into the dense cartilage. Further modi cation with a targeted biomolecular ligand may resist rapid clearance from the joint site [28,29]. Thus, the ROS-responsive PEG micelles based on thioketal linkers may hold promise for OA treatment in clinical.
Herein, we developed a multifunctional ROS-activatable theranostic polymer nanoparticles that are capable of loading hydrophobic drug and self-reporting the payload release upon ROS stimulation. The nanoparticles are formed by the amphiphilic block copolymers consisting of Cy5.5 modi ed cartilagetargeting peptide (CAP, DWRVIIPPRPSA) [30,31] and PEG modi ed an oxidation-responsive thioketal linkers (TK) hydrophobic block that contains Black Hole Quencher 3 (BHQ-3) as a quencher for Cy5.5, which was then encapsulated with Dexamethasone (DEX) to form TKC@DEX nanoparticles. DEX is a broad-spectrum synthetic corticosteroid medicine with long-lasting anti-in ammatory effect to potently reduce glycosaminoglycan loss in OA-affected cartilage site [32][33][34]. As show in Scheme 1, the smart TKC@DEX nanoparticles speci cally target on articular cartilage by CAP and respond to the high level of ROS since the thioketal linkages were cleaved by abundant ROS in in amed tissues, leading to gradual disassembly of the polymer to release Cy5.5 and drug. The increased distance between quencher (BHQ-3) and Cy5.5 enables stronger uorescence signal from Cy5.5, providing effective diagnosis and monitoring of OA condition. But in normal condition where ROS are minimal, the uorescence is turned off by BHQ-3.
This smart cartilage-targeting ROS responsive theranostic nanoprobe is promising for OA therapy.

Synthesis of ROS-responsive polymer
Here, thioketal were synthesized via their reaction with 3-mercaptopropionic acid and acetone. The uorescent monomer (Cy5.5-CAP) was synthesized following a route as shown in Figure S1. BHQ-3 was chosen as a quencher and co-polymerized with a ROS-cleavable thioketal-containing linker to prepare the PEG-TK-BHQ-3 ( Figure S2-3). The detailed preparation protocols of ROS-responsive polymers are presented in the Supporting Information and veri ed by 1 H NMR spectra.

Characterization
The morphology of TKC-PEG and TKC@DEX was measured by transmission electron microscopy (TEM) (Bruker, Germany). The particle size distribution and zeta potential of the TKC-PEG and TKC@DEX were recorded using a dynamic light scattering (DLS) (Malvern, UK). The DLS was used to investigate the stability of TKC-PEG incubated in PBS or in different concentrations of KO 2 (0 µM, 50 µM and 100 µM) with or without ROS inhibitor (N-acetyl-L-cysteine, NAC) for 0, 1, 2, 4, 8 and 24 h. The Ultraviolet-visible (UV-VIS) absorbance spectra of Cy5.5, BHQ-3 and TKC-PEG were detected by a microplate reader (Thermo Fisher Scienti c, USA).

Fluorescent recovery of TKC-PEG
The TKC-PEG solution was added to different concentrations of KO 2 (0 µM, 50 µM and 100 µM) with or without ROS inhibitor (N-acetyl-L-cysteine, NAC) and then incubated at 37°C for 1, 2, 4, 8 and 24 h, respectively. The uorescence signals were captured by using In-vivo Multispectral Imaging Systems (Bruker, Germany).
DEX loading 20 mg of nanoparticles was dispersed in 0.5 mL tetrahydrofuran (THF), followed by the addition of 10 mg DEX (dissolved in 0.5 mL tetrahydrofuran). Then 10 mL deionized water with a syringe was slowly added in to the mixture. The mixture was moved to a dialysis bag (MWCO 3400, Sigma, USA) to dialyze with deionized water for 24 h to remove the unloaded drug. At last, the dried solid micelles were obtained by lyophilization. The drug loading and embedding ratio were measured by the High Performance Liquid Chromatography (HPLC) (Shimadzu, Japan), and then calculated by the following formula using: Drug loading (%) = weight of DEX entrapped/weight of nanoparticles × 100% (1) Embedding ratio (%) = weight of DEX entrapped/weight of DEX feeding × 100% (2) DEX release study The DEX release pro les of TKC@DEX NPs were determined by dialysis membrane method. Brie y, TKC@DEX NPs were placed in dialysis bags and immersed in 15 mL four different buffers: (1) PBS only; (2) PBS with 50 µM KO 2 ; (3) PBS with 50 µM KO 2 and its inhibitor; (4) PBS with 100 µM KO 2 . KO 2 was selected as a reagent to simulate the ROS microenvironment [24]. All solutions containing 1% tween 80 were shook constantly at 37°C. At each time point (0, 1, 2, 4, 12 and 24 h), 1 mL of aliquots was removed from the release media and 1 mL of the same buffer was supplemented. The concentration of DEX release from TKC@DEX NPs was measured by HPLC. The procedures were performed in triplicate.

Chondrocytes isolation and culture
Chondrocytes were isolated from the knee joints of 3-day-old C57BL6/J mice (the Animal Experimental Center of Guangxi Medical University, Nanning, China) by enzymatic digestion in aseptic conditions according to previous report [35]. Firstly, articular cartilages were digested in trypsin (Gibco, USA) for 40 min at 37°C and then minced and digested with 2 mg/mL collagenase II for 3 h at 37°C. Secondly, the chondrocytes were centrifuged at 1000 rpm for 5 min and then suspended in DMEM medium containing 10% fetal bovine serum (FBS, Gibco, USA), 1% penicillin and streptomycin (Solarbio, China). Then, they were transferred into a culture ask and cultured at 37°C in a 5% CO 2 -humidi ed incubator. The third generation of cells was collected for further experiments.
In vitro cytotoxicity assay A n , and A p mean absorbencies of the sample, negative control and positive control, respectively.

Intracellular ROS detection and bioimaging
We used monosodium iodoacetate (MIA), an inhibitor of glyceraldehyde-3-phosphate dehydrogenase activity, to induce oxidative stress injure and the pathological OA symptoms of chondrocytes [37]. Chondrocytes were seeded into 6-well or 24-well plates and divided into four groups: (1)  and incubated with DCFH-DA (10 µM) for 20 min in the dark at 37°C. The chondrocytes were then washed three times with serum-free medium, and immediately detected by ow cytometer (BD, Biosciences, USA) [39]. The ROS generation of normal chondrocytes was also measured. Meanwhile, to reduce the production of ROS induced by MIA, chondrocytes were pretreated with 5 mM NAC for 1 h, follow by the addition of 3 µM MIA for 24 h, then intracellular ROS was detected.
In addition, cellular uptake and degradation of TKC-PEG NPs induced by endogenous ROS were investigated. After treated with MIA, the chondrocytes were incubated with TKC-PEG NPs for 4 h, and then they were washed with PBS and xed with 95% ethanol for 30 min. Meanwhile, MIA (3 µM) treated cells were also incubated with CAP-PEG and TK-PEG NPs as controls. Then the chondrocytes were stained with immuno uorescence of COL2A1 (1:200, Boster) and nuclei was counterstained with DAPI. Finally, the uorescence images were photographed using a uorescence inversion microscope (OLYMPUS, Japan).

MIA-induced chondrocytes and treatment
Chondrocytes were seeded into 24-well or 6-well plates and separated into ve groups: (1)  Quantitative real-time polymerase chain reaction (RT-qPCR) analysis The primer sequences for the OA-related genes are listed in Table 1. Total RNA was isolated using an RNA isolation kit (Tiangen Biotechnology, China). Then a reverse transcription kit (Takara, Japan) was used to reversely transcribe RNA to cDNA. Real-time PCR was conducted by a Light Cycle 96 system for 10 min at 95°C, 15 s at 95°C, and 60 s at 60°C. The relative gene expression levels were calculated using the 2 −ΔΔCT method with β-actin as the control.
The expression of OA catabolic biomarkers IL-6 and MMP-13 in chondrocytes was assessed by immuno uorescence. Chondrocytes were xed with 95% ethanol for 30 min and permeabilized with 0.1% Triton X-100 for 10 min. Samples were incubated with primary antibody as follows: TK@DEX (an equivalent DEX dose of 1 mg/kg) and TKC@DEX group (an equivalent DEX dose of 1 mg/kg). IA injections were performed twice a week. The rats in these groups were sacri ced for further analysis at 2 and 4 weeks after therapy.

In vivo NIR bioimaging
For in vivo bioimaging, mice were anesthetized by iso urane. Each group of OA mice (n = 6) was IA injected with 50 µL of 400 µg/mL TKC-PEG, CAP-PEG or TK-PEG, and the normal mice (n = 6) were also IA injected with TKC as control. The images were captured by an In-vivo Multispectral Imaging Systems (Bruker, Germany) at 0, 1, 2, 4, 7 and 14 d.
In addition, the uorescence intensity of the TKC was also investigated. The mice were randomly sorted into three groups (n = 6): 0.05 mg MIA + inhibitor group, IA injection of 0.05 mg MIA concomitant with 5 mM NAC; 0.05 mg MIA group, IA injection of 0.05 mg MIA; 0.1 mg MIA group, IA injection of 0.1 mg MIA [43,44]. Finally, the mice were sacri ced and the macroscopic evaluations of joints were performed.

Macroscopic observation
After 2 weeks or 4 weeks of treatment, the knee joints of mice were harvested for macroscopic evaluation according to the macroscopic scoring system (scale of 0-4) by three independent observers [45].
The joints were xed in 4% paraformaldehyde and subsequently decalci ed with a 14% ethylenediaminetetraacetic acid (EDTA) solution for ten days. Next the joints were embedded in para n and cut into Statistical Analysis.
All data are presented as the mean ± SD, and p < 0.05 was considered statistically signi cant. The signi cant differences among groups were evaluated by one-way ANOVA. Statistical analyses were conducted using SPSS statistics (SPSS 19.0).

Synthesis and Fabrication of nanoreactors
The successful synthesis of ROS responsive monomer and functional moiety were demonstrated by 1 HNMR spectra ( Figure S1-S5). The transmission electron micrograph (TEM) showed that the amphiphilic polymer TKC-PEG and TKC@DEX could self-assemble into homogeneously spherical nanoparticles in aqueous solution (Fig. 1a&b). Furthermore, dynamic light scattering (DLS) analysis revealed that the average diameter of TKC-PEG NPs was ~ 60 nm (Fig. 1c). The zeta potential of TKC-PEG NPs was close to neutral charges at -0.13 mV (Fig. 1d). After loading with DEX, the mean diameter of TKC@DEX NPs increased up to ∼90 nm (Fig. 1b&1c) and the zeta potential changed to -0.43 mV (Fig. 1d). The UV-vis spectrum also con rmed the successful construction of TKC-PEG. As shown in Fig. 1e uorescence intensity was monitored. In the absence of KO 2 , the uorescence of TKC-PEG was extremely weak over time due to the short distance between Cy5.5 and BHQ-3 (Fig. 1f&1g). However, when KO 2 (50 µM) was added, the TKC-PEG showed gradual recovering of NIR uorescence over incubation time due to that the ROS cleaved the thioketal of TKC-PEG and increased the distance between Cy5.5 and BHQ-3. The recovery of uorescence intensity was effectively accelerated with the increasing concentration of KO 2 from 50 µM to 100 µM, but it was suppressed when an ROS inhibitor NAC was added together with the KO 2 . The TKC-PEG NPs had good stability in PBS, showing no appreciable change in size distribution. On the contrary, the size rapidly increased under different concentration of KO 2 (50 µM and 100 µM) as the inducer of ROS, suggesting that the disassembly of TKC-PEG NPs was triggered in response to oxidative milieu (Fig. 1h).
In vitro drug release at ROS-simulated levels As con rmed by the results of in vitro uorescent recovery study (Fig. 1g), the high level of ROS in OA intracellular could break thioketal linkers and also trigger the release of DEX. The drug loading and embedding ratio were 13% and 30% for the TKC-PEG NPs, respectively. The in vitro drug release showed that only 19% drug was released in the absence of KO 2 (ROS inducer) after incubation for 48 h (Fig. 1i).

Cytotoxicity analysis
The cytotoxicity of TKC-PEG and TKC@DEX NPs on chondrocytes was evaluated by CCK-8 assay. As show in Fig. 2a, no toxicity was observed with DEX at the range of 0 to 200 µg/mL after 24 h and 48 h incubation. After DEX loading, 25 µg/mL TKC@DEX (equivalent concentration of DEX was 3 µg/mL) promoted cell growth compared with untreated chondrocytes (Fig. 2b). Thus, 25 µg/mL TKC@DEX was selected for further study.
The effect of TKC@DEX on the cell viability of MIA (ROS inducer)-induced chondrocytes [47] was investigated by CCK-8 analysis. As shown in Fig. 2c, the cell viability of the chondrocytes treated with MIA decreased 40.7% compared with the normal cells. However, after treating with DEX, CAP@DEX, TK@DEX and TKC@DEX for 24 h, the cell viability was increased to 70.27%, 74.14%, 78.78% and 87.97% compared with the MIA group, respectively, suggesting potential function of protecting chondrocytes from catabolic activity of in ammatory factor and promoting chondrocytes proliferation.

Hemolysis test
The biocompatibility of TKC@DEX was measured by the hemolysis test. As shown in Figure 2d, the TKC@DEX at various concentrations induced lower than 4.0% hemolysis rate by contacting erythrocytes at 37 o C for 1 h. These results indicated that TKC@DEX NPs were considered as excellent biocompatible and nonhemolytic materials, showing a great potential of biomaterials for clinical applications.

Cell uptake and in situ uorescent release
To investigate the chondrocytes targeting property of TKC-PEG NPs, cellular uptake of TKC-PEG and TK-PEG (without chondrocyte-a nity CAP peptide) NPs were compared. Treatment with 3 μM and 6 μM MIA caused a 1.1-fold and 2.7-fold increase of red uorescence (46.40% and 82.51%) in the cellular uptake compared with untreated cells (22.19%), respectively (Figure 2e&2f), which revealed that thioketal linkages exhibited e cient ROS-dependent degradation. the uorescence intensity of TKC-PEG group was observed to be much stronger than that of TK-PEG group after treatment with 3 μM MIA (Figure 2e&2g). However, the uorescence intensity declined to 30.55% after treating with NAC, which was similar to that of normal cell environment due to the elimination of ROS (Figure 2e&2f). These results indicated that ROS could trigger the cleavage of thioketal linkages to release Cy5.5 from TKC-PEG according to the level of ROS at simulated pathological conditions, and the targeting effect of CAP moieties in TKC-PEG NPs could facilitate high a nity to CEM-rich chondrocytes compared those with non-targeting modality. But in normal condition, Cy5.5 was hardly released and activated. The ow cytometer was used to monitor the ROS production in chondrocytes (Figure 2h). At OA microenvironment induced by MIA (the concentration was 3 μM), an elevated level of ROS (24.10%) was produced compared with normal cells (12.75%). And the ROS was up to 42.76% at the concentration of 6 μM MIA, indicating a signi cant increase of ROS production dependent on the concentration of MIA. Furthermore, the high level of ROS could decrease down to 17.48% after treated with NAC. Thus TKC-PEG may potentiate smart real-time monitoring of ROS levels.

In vitro anti-in ammatory activity
We then evaluated the e cacy of treatment groups to induce crucial OA catabolic biomarkers (MMP13, IL-6, and MMP-3) and chondrogenic markers genes (Col2a1) by qRT-PCR (Fig. 3a). It showed that the MMP13, IL-6 and MMP-3 expression of TKC@DEX group was signi cantly decreased than that of the MIA-treated group, down to 80.85%, 91.30% and 64.19%, respectively. Compared with DEX, CAP@DEX and TK@DEX group, there were 56%, 36% and 18% decreases of MMP13, IL-6 and MMP-3 expression in TKC@DEX group respectively, indicating its pronounced anti-in ammatory effect. TKC@DEX group also signi cantly up-regulated the expression of Col2a1, one key component of the cartilage matrix compared with other groups after 24 h of culture (Fig. 3a), suggesting that TKC@DEX NPs inhibited the degradation of collagen and protected chondrocytes under the pathological conditions of OA. The expression of OA biomarkers by immuno uorescent staining further con rmed the PCR results (Fig. 3b&3c). The results showed the expressions of MMP-13 and IL-6 were both decreased in the TKC@DEX group, almost 25%~45% lower than other groups (MIA, DEX, CAP@DEX and TK@DEX groups).

In vivo ROS -responsive activity
We then evaluated the e cacy of treatment groups to induce crucial OA catabolic biomarkers (MMP13, IL-6, and MMP-3) and chondrogenic markers genes (Col2a1) by qRT-PCR (Figure 3a). It showed that the MMP13, IL-6 and MMP-3 expression of TKC@DEX group was signi cantly decreased than that of the MIA-treated group, down to 80.85%, 91.30% and 64.19%, respectively. Compared with DEX, CAP@DEX and TK@DEX group, there were 56%, 36% and 18% decreases of MMP13, IL-6 and MMP-3 expression in TKC@DEX group respectively, indicating its pronounced anti-in ammatory effect. TKC@DEX group also signi cantly up-regulated the expression of Col2a1, one key component of the cartilage matrix compared with other groups after 24 h of culture (Figure 3a), suggesting that TKC@DEX NPs inhibited the degradation of collagen and protected chondrocytes under the pathological conditions of OA. The expression of OA biomarkers by immuno uorescent staining further con rmed the PCR results ( Figure  3b&3c). The results showed the expressions of MMP-13 and IL-6 were both decreased in the TKC@DEX group, almost 25%~45% lower than other groups (MIA, DEX, CAP@DEX and TK@DEX groups).

In vivo ROS -responsive activity
To detect the ROS activity at the early stage of OA, we established the OA model by IA injection of MIA (ROS inducer). In the normal group (healthy joints) treatment, IA injection of TKC-PEG probes resulted in weak uorescent signal in the normal group (healthy joints) and the relative uorescence intensity was 1.20 at 1 day (Fig. 4a&4b). By contrast, a strong uorescent signal was detected in the OA joints, indicating signi cantly up-regulated ROS activity in the OA joints compared to healthy joints. The relative uorescence intensity of TKC-PEG group in OA sites reached peak intensity at 1 d, and still maintained obvious uorescence at 14 d. Moreover, the relative uorescence intensity of TKC-PEG NPs treated mice was nearly ~ 1.34-fold higher than that of CAP-PEG NPs, and ~ 0.98-fold higher than TK NPs (Fig. 4a&4b) at 7d, indicating that the decoration of CAP-PEG endowed the potential enrichment effect of TKC-PEG NPs and thioketal linkage could be effectively cut off by ROS in cartilage. Thus, TKC-PEG NPs exhibited cartilage targeting property and prolonged retention time in joints, which was suitable for effective bioimaging to detect the development of OA in vivo.
Compared with the OA mice (0.05 mg MIA) injected with the TKC-PEG probes, those injected with the TKC-PEG + ROS inhibitor group produced undetectable uorescence signal (Fig. 4c&4d). In addition, when mice injected with 0.1 mg MIA, a higher ROS circumstance was induced, leading to more serious joint damage (Fig. 4c&4d). And the relative uorescence intensities could be increased at higher ROS circumstance (IA with 0.1 mg MIA), increasing up to 30% higher than that of IA with 0.05 mg MIA after treated with TKC-PEG probes. This result indicated that the intensity of uorescent signals of TKC-PEG probe could respond to ROS related OA severity, favorable for precise disease classi cation.
In vivo treatment effect We further carried out in vivo experiments to evaluate the OA therapeutic e cacy of TKC@DEX. The mice were divided into six groups, including PBS, OA, DEX, CAP@DEX, TK@DEX and TKC@DEX. In the PBStreated group, general characteristic OA features, such as surface irregular and large cartilage erosion, were observed. The macroscopic score of OA group was signi cantly higher than that of control group over time, which indicated mild or moderate cartilage destruction based on pathological alteration of joint morphology from 2 to 4 weeks (Fig. 5a). After treatment with DEX, there was a 31.82% and 30.67% reduction in the joint score in 2 weeks and 4 weeks, respectively, as compared to that of treatment with PBS alone. Meanwhile, both CAP@DEX and TK@DEX treatment signi cantly decreased the severity of OA. Furthermore, there was a 64.55% and 57.87% reduction in TKC@DEX group after treatment for 2 and 4 weeks, respectively, compared to that OA group, which showed its effective therapeutic repair effect of OA (Fig. 5b).
The HE and safranin O/fast green staining were used to evaluate the therapy e cacy of TKC@DEX in vivo (Fig. 6a&6b). In the OA group, OA characteristics, including cartilage damage, brillated lesions and loss of aggrecan worsened over time, were showed in the progression of OA. In contrast, TKC@DEX treatment resulted in a signi cant improvement in matrix arrangement, tide line maintenance or cartilage lesion, and the OARSI score exhibited improvement, with a reduce of 86.7% and 86% compared with OA group at 2 and 4 weeks, respectively (Fig. 6c). The scores for the TKC@DEX group were markedly lower than that of other treatment groups, exhibiting positive effects on the restoration of cartilage matrix expression.
We also assessed the expression of MMP-13 in OA cartilage by immunohistochemistrical analysis, which is a degradation product of type II collagen and corresponds to articular cartilage damage in the early stages of OA. The positive staining of MMP-13 could hardly be detected in the normal cartilage. In contrast, it showed high expression in the super cial and middle zones of the cartilage in OA group. Following treatment with DEX, CAP@DEX or TK@DEX, osteoarthritic changes of MMP-13 over-expression was reduced (Fig. 6d). These results indicated that TKC@DEX signi cantly slowed the progression of early OA and prevented the severe damage to articular cartilage in the OA.

Discussion
The abnormal over-production of endogenous ROS within cartilage tissue is a key hallmark of OA[48], providing a disease-speci c triggering mechanism for drug control and release systems. Herein, we fabricated an advance ROS responsive and cartilage targeting TKC@DEX nanoprobe with loaded drug for imaging and effective therapy of OA, which may provide reference for clinical application.
As a promising domain, ROS-stimuli responsive theranostic nanoprobes have been widely applied in disease theranostic owing to their unique all-in-one features [49][50][51]. Chung et al. proposed ultrasensitive ROS-responsive carriers that contained agents to release the drug to the disease site [48,52,53]. Compared with unresponsive delivery systems, these ROS-responsive systems remain stable in normal tissue, preventing their release to non-in amed tissues, which suggested their speci city in in ammation site. But these ROS responsive nano systems can only control the release of drugs, but cannot visually real-time monitor the progression of disease and most of them have no therapeutic effect. In our study, TKC@DEX is superior to unresponsive CAP@DEX, as evidenced by uorescent recovery studies simulated by abundant of ROS in vitro and vivo (Fig. 1f, 1g, 2e&4a). The uorescent signal is strong in the OA microenvironment, but is extremely weak in normal chondrocytes and joints, indicating that the level of uorescence signal correlates with ROS content. TKC@DEX not only exhibits smart drug release potential in response to high levels of ROS as demonstrated by drug release behavior test (Fig. 1i), but also can real-time track ROS activity since it emits signal of uorescence varied with ROS, promising for monitoring and on-demand therapy of OA.
On the other hand, cartilage targeting is of importance for drug therapy. Most in situ injected particles may be quickly cleared by the joint uid because of the inability of homing to lesion sites, leading to application limitations in OA diagnosis and treatment. Liu et al. engineered a chondrocyte-a nity peptide (CAP, DWRVIIPPRPSA) [31,54,55] by phage display technology. In our study, the nanoprobe TKC with cartilage targeting CAP showed stronger uorescence than TK (without CAP), as observed in vitro cellular uptake tests (Fig. 2a) and in vivo NIR imaging (Fig. 4). Moreover, the CAP-modi ed TKC@DEX prolonged the retention time of nanoparticles in the joint (Fig. 4a&4c), which could improve therapeutic outcomes of OA. These results indicated that cartilage targeting ensured the real-time monitoring and speci c treatment of OA.
In our studies, we observed that the TKC@DEX NPs showed stronger therapeutic e cacy than that of delivery systems without conjugation of CAP and TK. The TKC@DEX NPs play a key role in relieving the in ammation in vitro and in vivo (Fig. 3a, 3b, 3c&Figure 5). The results of macroscopic and histological analysis further evaluated that IA injection of TKC@DEX showed the better therapeutic effect than other formulations (TK@DEX, CAP@DEX or free DEX) after 2 and 4 weeks of treatment (Fig. 5-6). In addition, hemolysis test further assessed the excellent biocompatibility of the TKC@DEX NPs platform (Fig. 2d). The above results revealed that the drug delivery system targeted to the cartilage and responded well to low concentrations of ROS, exhibiting highly sensitive imaging and effective anti-in ammatory activity for therapy of OA.

Conclusions
In summary, we have successfully fabricated a novel cartilage-targeting and ROS-responsive delivery platform for diagnosis and therapy of OA, which enables real-time imaging to monitor the severity of OA and on-demand drug release at the site of abnormal ROS milieu. We expect that this theranostic system may be applied in clinic for treatment of OA.

Declarations Supporting Information
All data generated or analysed during this study are included in this published article and its supplementary information les. Tables Table 1 is not available with