Intramolecular Orthogonal Reporters for Dissecting Oxidative Stress and Response

Yecan Pan Institute of Quality Standard and Testing Technology for Agro-Products, Chinese Academy of Agricultural Sciences Rui Weng Institute of Quality Standard and Testing Technology for Agro-Products, Chinese Academy of Agricultural Sciences Linghao Zhang Beijing University of Chemical Technology Jing Qiu Institute of Quality Standard and Testing Technology for Agro-Products, Chinese Academy of Agricultural Sciences Xinlu Wang Institute of Quality Standard and Testing Technology for Agro-Products, Chinese Academy of Agricultural Sciences Guangqin Liao Institute of Quality Standard and Testing Technology for Agro-Products, Chinese Academy of Agricultural Sciences Zhaohui Qin Beijing University of Chemical Technology Rui Liu Tsinghua University Xiaochuan Dai Tsinghua University Yongzhong Qian Institute of Quality Standard and Testing Technology for Agro-Products, Chinese Academy of Agricultural Sciences Xin Su (  pkusuxin@gmail.com ) Beijing University of Chemical Technology https://orcid.org/0000-0002-6629-9856


Introduction
Oxidative stress resulting from effects of reactive oxygen species (ROS) such as superoxide (O 2 •− ), hydrogen peroxide (H 2 O 2 ), and the hydroxyl radical (OH•) plays an important role in the pathogenesis of a variety of diseases and important biological processes 1,2 . Toxic effects of oxidative stress lead to cellular damage by oxidizing nucleic acids, proteins, and membrane lipids. In mammalian cells, oxidative stress and the corresponding molecular responses are thought to be linked to a variety of diseases such as cancer 3 , Parkinson's disease 4 , Alzheimer's disease 5 , heart failure 6 , and even cardiac complications in COVID-19 7 . Particularly, oxidative stress-induced cancer development is mainly attributed to the direct DNA damage caused by ROS, which is mutagenic, and may also suppress apoptosis and promote proliferation, invasiveness and metastasis 8 . DNA damage results in the accumulation of excessive apurinic/apyrimidic sites (AP sites) thereby activating base excision pathway for DNA repair 9 . On the other hand, ROS-induced DNA damage can be bene cial, as they are utilized for cancer therapeutics, such as photodynamic therapy, which employ the oxidative stress and response to elicit cell death 10,11 .
Therefore, it is of great signi cance to quantitatively map cellular ROS and the DNA repair enzymes for ROS administration and utilization. This motivates the development of advanced tools to observe ROS and its connected molecules with spatiotemporal accuracy in live cells.
Fluorescence imaging emerges as a powerful tool for directly observing the localizations and activities of labeled biomolecules despite the crowded intracellular environment. It is desired to integrate molecular probes into single nanostructures to simultaneously map multiple targets in live cells 13,14 . Such design enables to study the colocalization and abundance of multiple targets with spatiotemporal accuracy. However, uncontrolled integration of multiple uorescent reporters into a single assembly leads to unexpected interaction between uorophores resulting in low sensitivity and poor target speci city.
Employing the base-pairing language to construct nanostructures in a programmable manner, structural DNA nanotechnology has been widely applied in biosensors, molecular logic devices, and drug delivery vehicles [15][16][17] . The well-de ned geometry of DNA nanostructure makes them promising scaffolds to organize probe molecules in an addressable manner and integrate distinct functions in precise stoichiometries improving the speci city, sensitivity, and throughput of biosensors 15 . The high biocompatibility of DNA nanostructures enables them to enter cells autonomously 18,19 . DNA nanodevices are versatile chemical reporters that can quantitatively map multiple targets in living systems in real time 18,19 . However, it remains challenging to eliminate crosstalk between probes even by using DNA nanodevices because the exibility of DNA sequence and nanostructures may induce unexpected uorophore interaction.
Here, we reported intramolecular Orthogonal Reporters (iOR) based on DNA nanostructure scaffolds in which uorescence reporters were precisely located with appropriate distance by the assistance of MD simulation. Spectrally separated uorophores in iOR respond to their corresponding targets with superior speci city. iOR functionalized with Nuclear targeting signal (NLS) peptide (NiOR) is capable of sur ng in both the cytoplasm and nucleus. NiOR quantitatively maps ROS and its induced overexpression of gene repair enzyme apurinic/apyrimidic (AP) endonuclease 1 (APE1) in living cells with spatiotemporal accuracy. Fluorescence imaging discloses that the expression level of APE1 is proportional to ROS level, and that their nuclear and mitochondrial accumulation are regulated by oxidative agents and natural antioxidants. In vivo uorescence imaging of tumor-bearing mice by NiOR also demonstrate the strong correlation between ROS and APE1 and the regulatory effect of those external stimuli. We anticipated that this simulation-guided iOR represented a new class of molecular probes for cellular imaging applications providing opportunities for deep understanding of the molecule interplay in live cells.

Results
MD simulation-guided design of iOR.
We chose DNA tetrahedron as the principal scaffold for iOR owing to its excellent biocompatibility and structural simplicity 21 Figure S1A and B). 22 . . APE1 speci cally cleaves AP site and therefore plays a signi cant role in BER pathway, particularly in the ROS-regulated BER 23 . Abnormal expression and subcellular distribution of APE1 have been linked to tumor development 24,25 . As our previous study, the simple uorescence dequenching of nucleic acid probes can be used for monitoring and quantifying the enzyme activity 26,27 . As shown in Scheme 1, the cleavage of AP site by APE1 results in the separation of Cy5 dye and BHQ3 quencher, leading to uorescence restoration. APE1 reporter exhibits a linearity to APE1 concentration ( Figure S2A and B), and gel electrophoresis also con rmed its correct function ( Figure S2C). Although DHE and Cy5 are separated in emission pro les, the overlap of the emission pro le of DHE and the excitation pro le of Cy5 would lead to Förster resonance energy transfer (FRET) between the two uorophores which results in poor sensitivity and speci city. Accordingly, it is desired to precisely control the distance between the uorophores to avoid FRET.
iOR was designed in silico by Tiamat software, and MD simulation were performed to optimize the structures ( Figure 2A). oxDNA, a coarse-gained model to incorporate thermodynamics as parameterized by the nearest-neighbor model into a description with continuous degrees of freedom that captures structural and mechanical properties of DNA 29 . oxDNA has previously been used to successfully assist the optimization of a variety of DNA nanostructures 29 . With oxDNA, the molecular behavior and structural exibility of iOR candidates were simulated, and the distances between the two uorophores in different candidates were evaluated ( Figure 2B), the simulation parameters were noted in Table S2. The average distances increase as a function of polyT spacer length. Due to the exibility of the DNA strand, the deviation of distance distribution also increases with the spacer length. The Förster distance of DHE and Cy5 was estimated as 5.4 nm which means the FRET e ciency can be lower than 10% if the distance between dyes exceeds 10 nm 33 . Therefore, based on the simulated dye distance distribution, the probabilities for 10% FRET of the iOR candidates can be obtained ( Figure 2B). S1-0T and S2-0T exhibits relatively low FRET probability among the candidates. By comprehensively considering the average distance of dyes and the FRET probability, S1-0T was chosen for potential iOR.
iOR assemble and characterization DHE dye was labeled on the carboxyl-modi ed ssDNA via NHS/EDC chemistry ( Figure S3A), the DHEmodi ed ssDNA shows slower electrophoresis speed indicating the successful modi cation ( Figure S3B). Next, DHE-modi ed strand (P7-DHE, see Table S1), APE1 reporter strands (P5 and P6, see Table S1), and DNA tetrahedron scaffold strands (P1, P2, P3, and P4, see Table S1) were assembled to formulate iOR. The stepwise strand assembly was veri ed by 6% non-denaturing polyacrylamide gel electrophoresis (PAGE) ( Figure 2C), the single band in lane 7 indicates the successful assembly and high purity of iOR.
The nanostructure was characterized by AFM image showing the size is ~14 nm ( Figure 2D). Dynamic light scattering (DLS) also indicated the average hydrodynamic size of iOR is 16.2 nm ( Figure S4). Next, the ROS response of the optimized iOR candidate (S1-0T) was tested showing the greatly enhanced DHE uorescence as well as low FRET e ciency of 17.2% in the presence of ROS. In contrast, high FRET e ciency of 52.1% was found by using an undesired iOR candidate S3-0T resulting in weak uorescence of FRET donor DHE and poor sensitivity for ROS ( Figure 2E). These results highlight the accuracy of MD simulation of potential iOR structures.
The sensitivity and orthogonality were evaluated as a function of ROS and APE1. The two reporters show a similar linearity to their corresponding targets allowing for a detection limit of 330 nM and 5 pM for ROS and APE1, respectively. Across various concentrations of each target, iOR speci cally produced uorescence that responded only to respective targets, thereby enabling orthogonal dual detection of two targets ( Figure 1F). Non-speci c nucleases which may induce nonspeci c signal of iOR are commonly found in biological uids as well as in live cells 34,35 . As previously reported, DNA nanostructures may protect the carried nucleic acid probes from non-speci c degradation because the rigid scaffold may restrict enzyme activity 19 . As expected, there is no signi cant uorescence increase in the presence of other nucleases ( Figure S5A and B).

Nucleus targeting iOR and its internalization
Owing to its superior orthogonality, iOR was operated in living cells for simultaneously imaging ROS and APE1. DNA tetrahedrons can be internalized by in mammalian cells via a lipid-raft-/caveolin-mediated endocytic pathway 36 . Nevertheless, such nanostructures do not have the capability of nucleus entry. To study ROS and APE1 in nucleus, iOR was functionalized with NLS peptide which facilitated the cellular uptake of nanoparticles toward nucleus in a highly ordered manner with the help of molecular motors and reduced the possibility to the lysosomes 37,38 . iOR was conjugated with NLS by click chemistry. As shown in Figure 2A, azide-labeled oligonucleotide reacted with alkyne-labeled NLS at room temperature.
The successful conjugation is indicated in the MALDI mass spectrum by the appearance of a signal corresponding to a compound with a mass equal to the sum of the masses of the peptide and the oligonucleotide of 9496.53 ( Figure 2B). The conjugate product was integrated with iOR by ssDNA hybridization. As shown in Figure S6, gel electrophoresis indicates the successful modi cation of NLS on iOR (NiOR), with the bands at different locations. Both iOR and NiOR were singly labeled with Cy5 to study their cellular distribution. Hela cells incubated with NLS functionalized iOR (NiOR) exhibits signi cantly 5.1 times higher uorescence intensity in nucleus than that incubated with iOR. It indicates that NLS functionalization can effectively deliver iOR into nucleus ( Figure 2C). Both NiOR and iOR are distributed thoroughly in cytoplasm owing to the capability of DNA nanostructures for endo-lysosomal escape ( Figure 2C). Furthermore, the biocompatibility of probes was tested with the use of iOR and NiOR for 2-4 h. All groups show over 90% cellular viability indicating the high biocompatibility of these molecular probes.
NiOR was applied for the simultaneous imaging of ROS and APE1 in Hela cancer cells. The uorescence of DHE and Cy5 gradually increases as a function of time suggesting that NiOR can be internalized e ciently within 2 h, and the reactions between the reporters and their corresponding targets are su ciently sensitive to detect endogenous ROS and APE1 ( Figure 2D and E). Since ROS and APE1 reporters are internalized via the same DNA scaffold, they would be expected to co-distribute in nucleus and cytoplasm following internalization. Owing to the function of NLS, nuclear and cytoplasmic ROS and APE1 can be successfully detected by NiOR.
Dissecting the regulation and subcellular distribution of cellular ROS and APE1 by NiOR.
Abundance, detailed subcellular distribution, and corresponding regulation of ROS and APE1 were further investigated ( Figure 3). The expression level of DNA repair enzyme APE1 is expected to relate to oxidative stress because they protect DNA from oxidation damage. Since cancer cells are typically growing vigorously, they have an elevated demand for energy production and, thus, generate a high level of line HEK-293 ( Figure 3A and Figure S8). The abundance of cytoplasmic and nuclear APE1 of cancer lines is signi cantly higher than that of normal cell line as ROS shows the same trend. More speci cally, the expression of APE1/ROS in HeLa cytoplasm was 2.6/1.8 times that of HEK-293, and about 3.1/1.7 times for A549. The expression of APE1/ROS in HeLa nucleus was 2.8 / 2.5 times as much as HEK-293, and 3.6 / 4.1 were due to A549 ( Figure S8B). These results indicate that the expression level of APE1 is strongly correlated to ROS abundance. In these cancer cells, due to changes in angiogenesis, there are cycles of hypoxia/re-oxygenation that contribute to a high ROS level. Responses to increased ROS include the elevated level of APE1, which might be responsible to maintain DNA integrity and function 40,41 . Mitochondria is responsible to process an extra-nuclear genome such as mtDNA, which is highly susceptible to attack by ROS. In the cells with high metabolic or proliferative rates, APE1 was expected to display co-localization with mitochondria 42,43 . To test this hypothesis, we used mitochondrial tracker to label mitochondria. APE1 and ROS exhibit Pearson's correlation coe cient (PCC) of 0.68 and 0.65 for mitochondria, respectively, and PCC of APE1 and ROS is 0.69, suggesting that three targets are slightly correlated in localization in the absence of external stimuli ( Figure 3B).
The regulation of ROS and APE1 by external stimuli was studied. Phorbol myristate acetate (PMA) is believed to stimulate oxidative burst and release of ROS in various types of mammalian cell resulting in imbalance and abnormality at molecular level [44][45][46] . HeLa cells were stimulated to produce O 2 •− by PMA in advance, and the uorescence of DHE and Cy5 of NiOR were enhanced simultaneously, implying the level of cytoplasmic and nuclear APE1/ROS were 1.7/1.7 and 2.0/4.3 times elevated with the simulation of PMA ( Figure 3A and 3C). PCC value of ROS and APE1 increased to 0.91 yielding a strong spatial correlation. Meanwhile, the colocalization of ROS/mitochondria and APE1/mitochondria was also enhanced ( Figure 3B). These results suggest that nuclear APE1 is overexpressed and cytoplasmic APE1 is accumulated under elevated oxidative stress to maintain the integrity of genomic DNA and mtDNA. To further con rm this mechanism, we extracted cellular DNA and measured the degree of guanine oxidation by ELISA ( Figure S9A), and found that oxidized guanine (8-oxo-dG) signi cantly increased by 29% as PMA stimulation ( Figure S9B), which supports the proposed mechanism. Allicin [S-(2-propenyl)-2propene-1-sul nothioate], the most biologically active sulfur-containing compound of garlic, was reported to be capable of modulating cellular ROS and increasing cellular antioxidant enzymes 47 . Allicin an organosulfur antioxidant obtained from garlic was proposed to be capable of effectively protecting cells from the adverse effects of oxidative stress 48,49 . As shown in Figure 3A (row 3) and 3C, the level of ROS and APE1 was reduced when cells were treated with allicin. Moreover, allicin can effectively weaken the uorescence of ROS and APE1 in PMA stimulated cells, and allicin pre-treated cell is less sensitive to PMA simulation (row 4 and row 5 in Figure 3A and 3C). Allicin also down regulated the level of 8-oxo-dG in the untreated cells and the PMA-stimulated cells ( Figure S9B). All these results indicate that allicin effectively mitigates cellular oxidative stress and protects cells from oxidative simulation, which is consistent with the previous observations.
The western blots performed show good agreement with the results of NiOR and validate the expression level of APE1 under different conditions. ( Figure S10). Commercially available ROS probe (DCFH-DA) was used to test ROS level, also indicating PMA and allicin can up and down regulate ROS, respectively ( Figure S11). All these results demonstrate that NiOR correctly re ects the subcellular distribution of ROS and APE1 as well as their regulation by external stimuli.
In vivo uorescence imaging of tumor-bearing mice Encouraged by the successful uorescence imaging of NiOR in living cells, we further evaluated the ability of NiOR to function in animals. The in vivo imaging was performed through in situ injection of NiOR into BALB/c mice bearing subcutaneous HeLa cervical cancer xenografts ( Figure 4A). After 4h of injection, uorescence imaging was performed. As shown in Figure 4B, uorescence signal of DHE and Cy5 accumulated at the tumor site, implying that NiOR can re ect the level of ROS and APE1 accordingly.
The DHE and Cy5 signal of all groups reach maximum at 1 h and 2 h after injection, respectively ( Figure  4D). After euthanizing the mice at 24 h post-injection, the major organs and tumors were harvested for ex vivo imaging suggesting that NiOR was massively accumulated at the tumor site ( Figure 4C, 4E and Figure S12). Lipopolysaccharides (LPS) was commonly used to elevate the ROS level in experimental mice 50 . As expected, under LPS stimulation, the uorescence of DHE and Cy5 enhanced simultaneously attributed to the increased ROS and induced overexpressed APE1 ( Figure 4B), ex vivo imaging of harvested tumor also supported this effect. The function of allicin to mitigate oxidative stress was also validated on both the control mice and LPS simulated mice ( Figure 4B and 4C). The uorescence of DHE and Cy5 of allicin-treated mice was lower than the corresponding untreated groups ( Figure 4D). Collectively, the molecular probe NiOR can be successfully deployed in experimental mice, and the results are consistent with cell imaging. In addition, immunohistochemistry validated the fact that LPS and allicin are capable of up-and down-regulating APE1 on mice, respectively, which was the important veri cation evidence for NiOR to detect the expression of APE1 in vivo ( Figure S13). TUNEL staining revealed that all groups exhibit a low level of nucleus necrosis and apoptosis of tumor tissue ( Figure  S14). H&E staining analysis of major organs and tumors revealed the high biocompatibility of NiOR ( Figure S15).

Discussion
Arising from an imbalance between generation and elimination of ROS, oxidative stress has drawn increasing attention owing to its contribution to the development of various diseases. It is now recognized that redox regulation involving ROS is crucial to the modulation of critical cellular functions 51 .
Underlying molecular response is complicated, but useful to understand critical pathology. "Off-line" methods such as western blotting and ELISA are available to analyze the changes under oxidative stress at the molecule level. However, such approaches cannot provide spatiotemporal information about the changes in living cells. Fluorescence imaging is an emerging "on-line" method to re ect the molecules in situ. Nanoparticle-based probes are desired to study the related targets in the uorescence imaging of living because such probes are capable of carrying multiple reporters in a single assembly. Nevertheless, the unexpected interaction between reporters results in poor sensitivity and speci city. To data, there is no effective approach to administrate the molecular reporters.
Taking advantages of structural DNA nanotechnology, we integrated ROS reporter and APE1 into a single DNA scaffold (iOR). In iOR, the reporter distance was precisely controlled by MD simulation tool to enable minimum interaction. The relation between ROS and APE1 was disclosed in living cells and experimental mice by iOR and NiOR,. By analyzing the level of 8-oxo-dG, we found the molecular mechanism that elevated ROS level induces the overexpression of APE1 mainly because of the cumulative damaged bases in genomic DNA and mtDNA. Figure 5 shows the abundance of multiple targets under different oxidative stress implying APE1 and 8-oxo-dG are both proportional to ROS level, as well as nuclear and cytoplasmic APE1 are equally sensitive to ROS change. Moreover, we for the rst time demonstrate that allicin a nature product from garlic exhibit great potential to mitigate oxidative stress, by spatiotemporal imaging of active small molecules and biological enzyme regulated oxidative stress.
In summary, a DNA scaffold-based molecular probe (iOR) was rationally designed by molecular dynamics simulation tool. It enables orthogonally reporting cellular ROS and its induced gene repair enzyme APE1.
Owing to precisely controlled distance of the two reporters, cellular ROS and APE1 were analyzed with high speci city and sensitivity and spatiotemporal accuracy in living cells. Their cellular localization and abundance have been dissected, revealing that elevating oxidative stress can induce the expression level of APE1, the colocalization of ROS and APE1, and APE1's mitochondrial accumulation. This work demonstrates a generalizable strategy to construct orthogonal molecular reporters in an integrated nanostructure for dissecting related molecular targets in living cells and experimental animals.

Methods
Reagents and materials.
All the oligonucleotides used in this work were synthesized and puri ed by high-performance liquid chromatography (HPLC), (Sangon Co., Shanghai, China), their sequences are summarized in Table S1 and the use of these strands is explained in the denotation. Dihydroethidium (DHE) and 1-Ethyl-3-(3- oxDNA simulation of iOR. The distance between uorophores on DNA scaffold was analyzed by oxDNA, which is a coarse-grained MD simulation software program. In our cases, PolyT sequences of different lengths were used to construct different iOR candidates. The initial DNA structure was generated from oxView (https://sulcgroup.github.io/oxdna-viewer/), then two simulation les "example.top" and "example.dat" were used for oxDNA running program. The above-mentioned simulation les need to be "relax" in oxDNA to produce initial structure simulation les for Virtual Movement Monte Carlo (VMMC) model. After each initial structure was generated as expected, the molecular behavior of the DNA nanostructures was simulated by sequence-dependent Virtual move Monte Carlo conditions without mutual trapping. The speci c parameters for VMMC simulation are shown in Table S2. During the simulation, DNA conformations over time were written down in "trajectory.dat". In these cases, the conformations were recorded every 10 steps and total simulation step was 100000. Therefore, 10000 data points can be extracted from trajectory through a python script named "distance.py" in oxdna-analysis tool. The distance from the base modi ed Cy5 to the base modi ed DHE was measured in each recorded conformation. Obtained data were used to chart and analyze, and the best structure of iOR can be obtained.
Preparation of iOR and NiOR.
To assemble iOR, 1 μM of oligonucleotides (P1, P2, P3, P4, P5, P6, P7 Table S1) were mixed in 1×TM buffer (20 mM Tris, 50 mM MgCl 2 , pH=8.0), and denatured at 80 °C for 10 min followed by cooling to 4 °C in 30 min by using PCR thermocycler. DHE was conjugated to carboxyl-labeled P7 strand by EDC chemistry, and the conjugates were puri ed by centrifugation at 12000 rpm for 5 min by ultra ltration devices (3k, 0.5 mL) to remove unreacted DHE. To obtain NiOR, Alkyne-labeled NLS peptide (N-DDEATADSQHSTPPKKKRKVEDPKDFPSEPra-C) was coupled to azide-labeled P8 strand using click chemistry. P7-HE and P8-NLS were characterized by 12.5% native PAGE which was operated at 4 °C for 45 min at a constant voltage of 120 V, and the gel was subsequently stained with SYBR Gold. P8-NLS was also characterized by positive ion MALDI mass spectroscopy.
Characterization of iOR and NiOR by gel electrophoresis, DLS and AFM.
Native PAGE (6%) was used to characterize the assembly process of DNA nanostructures. The gel was subsequently stained with SYBR Gold.  , then the uorescence emission pro le of Cy5 was recorded (λex: 625 nm). Enzymatic kinetics of multiple enzymes (5 U/mL APE1, 1 U/mL DNase I, 0.4 U/mL Exo I, 2 U/mL λ Exo, 2.5 U/mLT5 Exo, 2 U/mLT7 Exo) was measured by uorescence kinetics at 37 °C on a real-time PCR thermocycler (Rotor-Gene Q, QIAGEN, Germany). 10 mM EDTA ( nal concentration) was used to stop the reaction. The products were characterized by 6.0% native PAGE.
Cell culture and uorescence imaging.
Hela, A549, HEK-293 were cultured in the complete medium at 37 °C in a humidi ed atmosphere of 5% CO 2 /95% air (for details see Table S3). Cells were transferred to a laser confocal culture dish with an appropriate density, incubated with mito-tracker (200 nM) and nuclear staining dye Hoechst33324 (10 μg/mL). After removing the above dyes, the cells were incubated with 100 nM of NiOR. Highly inclined and laminated optical sheet (HILO) uorescence microscopy was assembled by a Nikon inverted microscopy (ECLIPSE, Ti−U) carried with 100×magni cation, 1.49 numerical aperture (NA) TIRFM objective (Nikon), an EMCCD cameras (Ixon 897, Andor). For HILO illumination, the lasers of 488 and 647 nm were coupled into a homemade illuminator to excite DHE and Cy5, respectively. Optical lters (Semrock) of 590 nm and 670 nm were used to detect the corresponding emission. The uorescence images were given to 0.15 μm per pixel, and the images were acquired with an exposure time of 50 ms (gain 3). Mito-tracker and nuclear staining dye were excited by mercury light with appropriate lters.
External stimuli were used to regulate ROS and APE1 (see the legend in Figure 3).
Evaluation of cellular APE1 by western blotting.
Protein quantitative analyses were performed according to supplier's protocol (ProteinSimple, USA). Cells were lysed on ice and then centrifuged. BCA Protein Assay Kit was used to determined protein concentrations. Western blotting was performed using a Wes Simple Western system, an automated capillary-based size sorting system (ProteinSimple, USA). Five micrograms of total protein from each group was loaded into WES 25-well plates for separation (Protein Simple, USA). The antibody of -Tubulin (dilution ratio was 1/100) and APE1 (dilution ratio was 1/100) were used as primary antibody. The electrophoretic separation and immunological detection parts were taken place in the capillary system automatically.
Detection of human 8-oxo-dG in cell extracts using an enzyme-linked immunosorbent assay kit (ELISA).
All of the cells were collected and broken by high-throughput ultrasonic disruptor in 1×PBS. The supernatant was collected, and centrifugation with 12000 rpm was performed at 4 °C for 20 min to remove bio-macromolecules. The supernatants were detected by 8-oxo-dG ELISA kit (MMBIO, China). The absorbance of 450 nm was recorded to re ect the 8-oxo-dG abundance.
Animals experiments and in vivo uorescence imaging.
BALB/c nude mice were from Servicebio Co., Ltd. (Wuhan, China). They were raised in a speci c pathogen-free animal facility and had access to water and food. The animal experiment protocols were approved by the Animal Ethics Committee of the Beijing University of Chemical Technology. HeLa tumor cells were inoculated under the armpits of mice until the tumor volume reached 100 mm 3 . Table S4 shows the feed administration of each group in Figure 4. NiOR (10 μM, 25 μL) was injected into the tumor site of mice by using a micro syringe. Mice were imaged at predetermined time intervals by in vivo uorescence imaging system (IVIS). After injection of the probe for 24 h, The tumors and major organs of mice were isolated and xed in 4% paraformaldehyde solution. They were para n-embedded, depara nized, rehydrated, and then Hematoxylin and Eosin (H&E) staining and TdT-mediated dUTP Nick-End Labeling (TUNEL) was performed. Immuno uorescence was performed for APE1 with a dilution ratio of antibody of 1/100.

Data Processing and statistical Analysis
The cell uorescence images were analyzed by ImageJ 8.0. The colocation of uorescence is analyzed by 'Colocalization' and 'plot pro le' module, while 'Measure' module is used to analyze mean uorescence intensity of cells. Compass software (ProteinSimple, USA) was used to analyze the relative amount of each protein in western blotting, through the areas under peaks from the chemiluminescence chromatograms. Living Image Software 4.4 for IVIS Lumina XR Series was used to analyze the uorescence imaging results of mice. Bar charts, line plots and heat map were made by GraphPad 8.0.6. Radar Diagrams ( Figure 5) were made by Excel 2016. One-way ANOVA was conducted using IBM SPSS Statistics 23 according to Duncan's test (p < 0.05).

Competing interests
There is no con ict to declare.

Data availability
The data that supports the plots within this paper and other nding of this study are available from the corresponding author upon reasonable request.

Supplementary information
Supplementary Figure S1-