Antisense oligonucleotides selectively enter human- derived antibiotic-resistant bacteria through ATP- binding cassette transporter

Mingzhu Liu Soochow University Rong Sun Soochow university Jiali Ding Soochow university Han Ye Department of Ophthalmology and Vision Science, Eye, Ear, Nose and Throat Hospital, Fudan University Binbin Chu Soochow university Yunmin Yang Soochow university Yuqi Wu Soochow university Haoliang Shi Soochow university Bin Song Soochow university Jiaxu Hong Department of Ophthalmology and Vision Science, Eye, Ear, Nose and Throat Hospital, Fudan University Houyu Wang Soochow university https://orcid.org/0000-0002-5134-9881 Yao He (  yaohe@suda.edu.cn ) Soochow University https://orcid.org/0000-0003-1672-4057

bacteria robustly and selectively 'eat' their counterfeiting 'foods', that is, glucose polymer (GP) and antisense peptide nucleic acid (asPNA)-modi ed uorescent silicon nanoparticles (SiNPs) (GP-SiNPs-asPNA). Supplementary Fig. 1 showed the synthetic route for GP-SiNPs-asPNA. Typically, the aldehyde groups of the GP molecules (e.g. poly [4-O-(α-D-glucopyranosyl)-D-glucopyranose]) were reacted with the amino groups on the surface of SiNPs to form a stable structure (GP-SiNPs) based on the Schiff base reaction 29 . Subsequently, Cy7.5-labeled asPNA (see the detailed sequences in Supplementary Table 1 and related characterizations of HPLC and mass spectrometry in Supplementary Fig. 2) was covalently linked to GP-SiNPs via an EDC/NHS condensation reaction between the amino groups of Glu in asPNA and the carboxyl groups on the surface of the SiNPs. The GP-SiNPs-asPNA entered the bacterial intracellular volume through the bacteria-speci c ABC transporter pathway (Fig. 1a). Unlike CPP-conjugated asPNA (e.g. KFFKFFKFFK-asPNA ((KFF) 3  The GP-SiNPs-asPNA and free SiNPs ( Supplementary Fig. 3) in the transmission electron micrograph (TEM) images all displayed spherical shapes with good monodispersity. Dynamic light scattering (DLS) measurements ( Supplementary Fig. 4) revealed that the hydrodynamic diameter of GP-SiNPs-asPNA (e.g. 3.62 nm) was slightly larger than that of free SiNPs (e.g. ~5.62 nm). To assess whether the small-sized GP-SiNPs-asPNA could enter bacterial cells, we rst performed scanning electron microscopy (SEM) and high-angle annular dark eld-scanning TEM (HAADF-STEM). The MDR E. coli and MRSA cells isolated from patients with keratitis were treated with GP-SiNPs-asPNA at 37 °C for 2 h, followed by washing with PBS buffer several times. As revealed in the SEM images in Fig. 1c, the surface and morphology of the GP-SiNPs-asPNA treated bacteria were not distinctly different from those of the untreated bacteria. As further shown in the HAADF-STEM data (Fig. 1d), silicon elements existed only in the GP-SiNPs-asPNAtreated bacteria and not in the untreated bacteria. These results demonstrate that the bacteria indeed engulfed the GP-SiNPs-asPNA.
To facilitate gene therapy, asPNA consisted of two consequent blocks: Ec108acpP and Sau101fmhB. After being internalized into the intracellular volume of E. coli, the Ec108acpP block would complement its target, the messenger RNA of Ec108acpP (mEc108acpP), inhibiting the synthesis of fatty acids [35][36][37] . Analogously, when GP-SiNPs-asPNA entered the S. aureus cells, the Sau101fmhB block would combine with the messenger RNA of Sau101fmhB (mSau101fmhB), preventing the synthesis of peptidoglycan 38,39 . To con rm this antimicrobial mechanism, we extracted RNA from an equal number of MDR E. coli or MRSA incubated with GP-SiNPs-asPNA at various concentrations (i.e. 250 nM, 500 nM and 1 μM). The extracted RNA was analysed by semiquantitative PCR and qPCR; 16S cDNA ampli ed by primers was used as control. The corresponding primer sequences are shown in Supplementary Table  2. Expectedly, we found that GP-SiNPs-asPNA inhibited the expression of both acpP in MDR E. coli (Fig.  1e) and fmhB in MRSA (Fig. 1f). Moreover, the expression levels of both acpP and fmhB decreased as the GP-SiNPs-asPNA concentration increased, indicating a concentration-response strategy. The PCR results also indirectly demonstrated the internalization of GP-SiNPs-asPNA into bacterial cells.
The UV-vis absorption spectrum of GP-SiNPs-asPNA in Supplementary Fig. 5a showed two characteristic peaks at 320 nm (assigned to SiNPs) and 788 nm (assigned to Cy7.5). Accordingly, we observed two typical photoluminescence (PL) peaks located at 520 nm (excitation at 405 nm) and 808 nm (excitation at 808 nm), which were assigned to SiNPs and Cy7.5, respectively ( Supplementary Fig. 6). By leveraging the emission properties of GP-SiNPs-asPNA, we next used confocal laser scanning microscopy (CLSM) to image the bacteria treated with asPNA, SiNPs-asPNA, GP-SiNPs-asPNA and (KFF) 3 K-asPNA (Cy7.5 labeled) at equivalent doses at 37 °C for 2 h, followed by washing with PBS buffer several times. The detailed sequence of (KFF) 3 K-asPNA and related characterizations were shown in Supplementary Table 1 and Supplementary Fig. 7. To test whether GP-SiNPs-asPNA could speci cally target diverse bacteria, we rst selected four representative bacteria, i.e., Gram-negative bacteria of E. coli and P. aeruginosa and Gram-positive bacteria of S. aureus and M. luteus as targets. As revealed in Fig. 2a, green uorescent signals (assigned to SiNPs, rst column, λ ex = 405 nm, λ em = 500-550 nm) and red uorescent signals (assigned to Cy7.5, λ ex = 633 nm, λ em = 700-800 nm) were barely detectable in either the asPNA or SiNP-asPNA groups. Additionally, weak red uorescent signals were detectable in (KFF) 3 K-asPNAtreated E. coli and P. aeruginosa, while they were hardly detectable in (KFF) 3 K-asPNA-treated S. aureus and M. luteus, in agreement with the reported result that CPPs-asPNA more easily enter Gramnegative bacteria than Gram-positive bacteria 19,40,41 . In contrast, we observed very strong green and red uorescent signals in all GP-SiNPs-asPNA treated bacteria. Furthermore, green uorescence overlapped well with red uorescence in the merged channel in the GP-SiNPs-asPNA groups. Importantly, we observed the identical results in GP-SiNPs-asPNA treated MRSA and MDR E. coli isolated from patients with keratitis (Fig. 2b). The human-derived strains were obtained from patients with keratitis who were diagnosed and treated in the Shanghai Eye, Ear, Nose and Throat Hospital, Fudan University. The corresponding histograms of uorescent intensity in Fig. 2c further revealed that the red uorescent signals observed in GP-SiNPs-asPNA groups were signi cantly higher than those in (KFF) 3 K-asPNA groups (p< 0.001). These results primarily veri ed the bacteria-targeting ability of GP molecules.
Upon the addition of phenol-sulfuric acid, the UV-vis absorption spectrum of GP-SiNPs-asPNA exhibited a new peak at 490 nm, assigned to furfural resin, which was produced by the reaction between GP and phenol ( Supplementary Fig. 5b). We determined the amounts of linked GP and Cy7.5-asPNA through the corresponding calibration absorption curves ( Supplementary Fig. 8). Typically, when the reaction concentration of asPNA was 1 μM, the loading rate reached ~53%. To investigate the effect of the amount of linked GPs and incubation time on the uptake e ciency of GP-SiNPs-asPNA by bacteria, we performed dose-and time-response experiments by using CLSM. Typically, ~1.0×10 9 CFU of E. coli or S. aureus were incubated with GP-SiNPs-asPNA containing GP at various concentrations (e.g., 0, 2.5, 5, 10 and 15 mg mL -1 ) for 0.5 h, 1 h, 1.5 h, 2 h and 2.5 h, respectively. As shown in Supplementary Fig. 9, the uorescent signal became stronger when GP concentration and incubation time increased, suggesting that the uptake of GP-SiNPs-asPNA by bacterial cells was dose-and time-dependent. Typically, when the GP concentration was 10 mg mL -1 and the incubation time was 2 h, the uorescence signal reached its peak. If further enhancing the GP concentration and the incubation time, the uorescence signal did not improve signi cantly. Quantitatively, the uptake rate of GP-SiNPs-asPNA by bacteria after 2 h of incubation was further analyzed by ow cytometry. Consistent with the results in Supplementary Fig. 9, the uptake rate of GP-SiNPs-asPNA (2.5 mg mL -1 GP) by E. coli was 15.2% and that by S. aureus was 16.9%, and the uptake rate of GP-SiNPs-asPNA (10 mg mL -1 GP) by E. coli climbed to 52.8% and that by S. aureus was 56.4% (Fig. 2d). When the GP concentration reached 15 mg mL -1 , the uptake rate of GP-SiNPs-asPNA by E. coli was 54.5% and that by S. aureus was 59.4%, with a tiny increase, indicating that a relative GP saturation state was achieved. As such, 10 mg/mL GP and 2-h incubation were employed in the following experiments. Comparatively, when loading the same amount of asPNA (e.g., 1 μM), the uptake rate of (KFF) 3 K-asPNA by E. coli was only 21.1% and that by S. aureus was only 9.97% after 2 h of incubation, which was much lower than that of GP-SiNPs-asPNA. Likewise, the uptake rate of GP-SiNPs-asPNA by MDR E. coli was 45.6% and that by MRSA was 51.6%; while the uptake rate of (KFF) 3 K-asPNA by MDR E. coli was only 14.7% and that by MRSA was only 17.5%.
To examine the selectivity of GP-SiNPs-asPNA towards bacteria over mammalian cells, human retinal pigment epithelial cells (ARPE-19) spiked with MDR E. coli or MRSA were incubated with GP-SiNPs-asPNA or (KFF) 3 K-asPNA for 2 h and then washed with PBS buffer. As shown in Fig. 2e, we observed green and red uorescent signals only in bacteria rather than in ARPE-19 cells in the GP-SiNPs-asPNA group. In contrast, in the (KFF) 3 K-asPNA-treated groups, red uorescence signals were detectable in both ARPE-19 cells and bacterial cells. In human blood samples spiked with E. coli or S. aureus under the same treatments, we observed similar phenomena ( Supplementary Fig. 10). Collectively, these results con rmed that the developed strategy featured higher selectivity towards bacteria over mammalian cells than (KFF) 3 K-asPNA.
Ultimately, the transport mechanism of GP-SiNPs-asPNA into bacteria was systematically investigated based on inhibition assays and competition assays. As shown in the inhibition assay in Fig. 2f, we did not observe any uorescent signals in sodium azide (NaN 3 )-and GP-SiNPs-asPNA treated bacteria in which NaN 3, serving as a bacterial respiratory inhibitor, could inhibit the work of ABC transporter 42 . As shown in the competition assay in Fig. 2g, the uorescent signals of GP-SiNPs-asPNA in E. coli gradually decreased when the bacteria were pretreated with GP at increasing concentrations (e.g., 0, 2 and 20 mg mL -1 ). These results together demonstrated that the internalization of nanoprobes into bacterial cells was based on the ABC transporter pathway.
We used the agar plate assay to intuitively evaluate the antimicrobial activity of the proposed strategy in vitro (Fig. 3a). In order to highlight the superiority of this strategy, the direct comparison with the clinically used antibiotics was performed, including cipro oxacin, nor oxacin and ampicillin. As expected, compared to PBS groups, we found slightly fewer MDR E. coli colonies, and almost no P. aeruginosa or M. luteus colonies in the cipro oxacin, nor oxacin or ampicillin groups. This suggested that even broad-spectrum antibiotics at the high dose of 15 µg mL -1 featured limited therapeutic effects against the human-derived MDR E. coli. Although they were not effective at killing MDR E. coli, they indiscriminately killed the non-resistant bacteria, e.g., P. aeruginosa or M. luteus. Also, we observed numerous MRSA colonies in ampicillin treated group. In the GP-SiNPs-asPNA groups, we observed nearly no MDR E. coli or MRSA colonies, corresponding to a much lower bacterial count; while numerous P. aeruginosa or M. luteus colonies, corresponding to a large amount of bacterial count (Fig. 3b). As further revealed in Fig. 3c, the presented strategy showed dominant antibacterial rates against MRSA (~99.99%) and MDR E. coli (~99.99%) but displayed almost no therapeutic effects against P. aeruginosa or M.
luteus. On the contrary, other antibiotics even at 15 µg mL -1 displayed inferior antibacterial rates against MDR E. coli (e.g., cipro oxacin: ~60.51%, nor oxacin: ~75.92%; ampicillin: 56.04%) but distinct therapeutic effects against P. aeruginosa or M. luteus (e.g., >99%). The results of semiquantitative PCR and qPCR further demonstrated that GP-SiNPs-asPNA speci cally inhibited the expression of target genes ( Supplementary Fig. 11). These results demonstrated that the developed strategy could selectively and e ciently kill antibiotic-resistant bacteria. We also employed the established methylthiazole tetrazolium (MTT) method to evaluate the cytotoxicity of GP-SiNPs-asPNA towards mammalian cells, including mouse retinal endothelial cells (MRECs) and ARPE-19 cells. As shown in Supplementary Fig. 12, the cell morphology of both MRECs and ARPE-19 cells did not change signi cantly, and the cell viability remained above 90% even when the concentration of GP-SiNPs-asPNA was up to 1 μM. These results indicated the negligible cytotoxicity of GP-SiNPs-asPNA towards ocular cells. Additionally, GP-SiNPs-asPNA showed good stability after 7 days of storage in PBS or 10% fetal bovine serum (FBS) ( Supplementary Fig. 13), implying its promising applications in vivo.
As for in vivo applications, we chose ocular bacterial infections as targets. Due to tear clearance and frequent blinking, the e ciency of conventional antibiotics for eye infections by using topical administration is usually less than 5% [43][44][45] . Therefore, the administration of antibiotics at high doses is generally performed multiple times a day for severe cases. Intense antibiotics therapy is prone to the generation of an increasing number of clinically resistant pathogens, especially superbugs. We established bacterial keratitis and endophthalmitis models in mice. To construct bacterial keratitis models, Sprague-Dawley (SD) mice (female, 6-8 weeks, n=5) were rst anesthetized, followed by scratching 4 scars with the same depth and size on the cornea by using a sterile knife. Then, 200 µL of E. coli, S. aureus or the mixture of E. coli and S. aureus (E. coli+S. aureus) with various concentrations was dropped onto the corneas. After one day of infection with bacteria, the corneas were dropped with SiNPs, SiNPs-asPNA, vancomycin-modi ed SiNPs (SiNPs-Van) or GP-SiNPs-asPNA at the equivalent dose for 5 days (one drop per time (~20 µL), three times per day) (Fig. 4a). The dropped agents were kept on the corneal surface for 30 min and then washed with PBS. Supplementary Fig. 14 illustrated the construction of bacterial endophthalmitis models (E. coli + S. aureus, ~3.0×10 6 CFU) in mice. For treating the mice with bacterial endophthalmitis, we added poloxamer 407 (P407) to GP-SiNPs-asPNA solution, followed by dropping on corneas. Gelation of the mixed solution was rapidly achieved, prolonging drug retention on the ocular surface. As revealed in the drug release curves in Supplementary Fig. 15, the 25% gel formulation displayed the slowest drug release (e.g., 50% drug release rate after 40 min, 71% drug release rate after 2 hours). Supplementary Fig. 16 revealed the red uorescence from GP-SiNPs-asPNA could be observed in the external retina, suggesting that GP-SiNPs-asPNA could enter the inside of the eyeball.
In the bacterial keratitis models, we used a uorescence stereoscopic microscope (Olympus, SZX16, λ ex = 488 nm) to capture the uorescent images of infected corneas after rst drug administration (Fig.  4b). The actual count of E. coli or S. aureus at the infected cornea during imaging was determined by tissue harvesting, homogenization and culturing with CFU count. As expected, no uorescent signals could be detectable in the PBS, asPNA or SiNPs-asPNA groups. Indeed, SiNPs-Van could image S. aureus keratitis, while they could not image E. coli keratitis owing to the strong a nity between vancomycin and the D-Ala-D-Ala moiety in the cell wall of S. aureus 46 . As further supported by the corresponding histograms, S. aureus-infected corneas treated with SiNPs-Van had an ~16.4-fold increase in uorescence intensity compared with E. coli-infected corneas treated with SiNPs-Van. In contrast, we observed green uorescence signals in both E. coliand S. aureus-infected corneas, indicating that the proposed hitchhiking strategy allowed imaging of diverse bacterial keratitis. Additionally, we were eager to image keratitis caused by diverse bacteria at various concentrations to determine the detection threshold of the developed strategy. Typically, when the E. coli or S. aureus concentration during imaging decreased from ~1.0 × 10 6 to ~1.0 × 10 4 CFU, the corresponding uorescence signal became weak, suggesting the bacterial concentration-dependent manner of the hitchhiking strategy. As revealed in the corresponding histograms, even though the S. aureus concentration during imaging was as low as ~1.0 × 10 4 CFU, the GP-SiNPs-asPNA treated cornea had an ~3.62-fold enhancement in uorescence intensity compared with the SiNPs-Van group. In this case, SiNPs-Van targeted the bacterial cell wall and was unable to access the bacterial intracellular volume. As a consequence, the payloads of GP-SiNPs-asPNA internalized into the intracellular volume of bacteria were relatively higher than those of SiNPs-Van on the bacterial cell wall, leading to a better detection sensitivity. Such ultrahigh sensitivity was approximately three orders of magnitude higher than most contrast agents, which should be su cient for many in vivo scenarios 47 .
We used a slit-lamp microscope to observe the daily corneal changes during the treatment. As shown in Fig. 4c, clearly visible corneas were observed at the beginning (0 d). After one day of infection, each group had mild corneal edema, mild iris hyperaemia and a small amount of secretion. After ve days of treatment, these symptoms were increasingly relieved in GP-SiNPs-asPNA groups and S. aureus keratitis treated with SiNPs-Van. In contrast, these signs worsened in the other groups. Speci cally, on the fth day of treatment, corneas returned to normal in the GP-SiNPs-asPNA groups, while the corneas in the PBS, SiNPs and SiNPs-asPNA groups were completely cloudy and severely congested, featuring severe corneal stroma edema and unclear iris texture. As further evaluated by slit-lamp microscope in Fig. 4d, on average, the corneal opacity scores in the GP-SiNPs-asPNA group gradually decreased. Similarly, the scores in S. aureus keratitis mice treated with SiNPs-Van gradually decreased. In contrast, the scores in other groups gradually increased. In the bacterial endophthalmitis models, as shown in Supplementary  Fig. 17, one day after infection, the pupils in all groups appeared opaque, and the eyes were congested and swollen. After one day of treatment, the pupils only in the GP-SiNPs-asPNA gel system were slightly relieved, while the in ammation symptoms worsened in the other groups. Speci cally, after four days of treatment, we observed only mild in ammation in the GP-SiNPs-asPNA gel system, while the pupils were yellow and cloudy, with considerable hyperaemia and secretion, and the eyeballs were severely damaged in the other groups. We also used the slit-lamp microscope to evaluate the effects of GP-SiNPs-asPNA on the corneas of healthy mice. After 5-day topical administration, the corneas in each group were clearly visible, without swelling or unknown secretions ( Supplementary Fig. 18).
The daily corneal colony of each group was given in Figs. 4e & 4f. Typically, the number of bacterial colonies in the GP-SiNPs-asPNA group continued to decrease over time. Additionally, the number of S. aureus colonies in the SiNPs-Van group decreased over time. In contrast, the number of bacterial colonies in other groups increased over time. Accordingly, the in vivo bacteriostatic rate of GP-SiNPs-asPNA on keratitis was calculated to be 99% against E. coli and 99% against S. aureus, 98.6% against E. coli + S. aureus. The bacteriostatic rate of SiNPs-Van on keratitis was calculated to be 98% against S. aureus. In the bacterial endophthalmitis models, the bacteriostatic rate of GP-SiNPs-asPNA on endophthalmitis caused by E. coli + S. aureus was calculated to be 98.8% supported by agar plate assays, in which eyeball homogenates were cultured to obtain bacterial colonies ( Supplementary Fig. 19). Ultimately, to histologically analyze the in ammation of the infected cornea after the treatment, hematoxylin-eosin (H&E) staining was performed (Fig. 4g). Consistently, we observed almost no in ammatory cell in ltrates in bacteria-infected corneal tissues treated with GP-SiNPs-asPNA and S. aureus-infected corneal tissues treated with SiNPs-Van. In contrast, we found a large number of in ammatory cell in ltrates in other groups. We also observed the similar results in H&E staining of the retinal membrane in the bacterial endophthalmitis models (Supplementary Fig. 20). As shown by H&E staining of eyeball sections of healthy mice treated with GP-SiNPs-asPNA in Supplementary Fig. 21, we observed no signi cant changes in the morphology of corneal tissues and retinal tissues among the groups, suggesting negligible toxicity of GP-SiNPs-asPNA in vivo. Collectively, these evidences fully proved that the hitchhiking strategy featured excellent therapeutic effects on ocular infections in mice caused by diverse bacteria. of EDC solution (50 µL, 50 mg mL -1 ) and NHS solution (12.5 µL, 50 mg mL -1 ), followed by followed by centrifugation for 15 min (6000 rpm) and ultra ltration (MWCO, 1000, Spectra/Pro) to remove unreacted vancomycin. A transmission electronic microscopy (TEM) (Philips CM 200, 200kV) was employed to characterize the morphology and size of as-prepared nanoagents. The Delsa™ nano submicron particle size analyzer (Beckman Coulter, Inc.) was used to analyze the dynamic light scattering (DLS) of nanoagents. A UV-vis spectrophotometer (Perkin-Elmer lambda) was utilized to measure the UV-vis absorption spectra of nanoagents. A spectro-uorimeter (HORIBA JOBIN YVON FLUORMAX-4) was used to record photoluminescence (PL) spectra of nanoagents.

Methods
In vitro imaging of diverse bacteria via the developed strategy. The lyophilized powders of E. coli, S. aureus, M. luteus and P. aeruginosa were respectively dissolved in LB medium. Next, the bacteria dissolved in LB medium were cultured on the LB plate medium at 37 o C for 12 hours to pick a single colony. After that, the picked colony was cultured in LB liquid medium at 250 rpm and 37 o C. Afterwards, the suspensions of bacteria in the exponential growth phase were washed twice and re-suspended in PBS buffer. The optical density (OD) at 600 nm was measured to determine the concentration of bacteria. A colony counting instrument (Czone 8) was used to determine the number of bacterial colonies. MRECs cells were cultured in the Dulbecco's modi ed Eagle's medium (DMEM) containing high glucose in a humidi ed incubator with 5% CO 2 at 37 °C. ARPE-19 cells were cultured in Roswell Park Memorial Institute (RPMI)-1640 medium supplemented with 10% heat-inactivated fetal bovine serum (FBS),100 μg mL -1 streptomycin and 100 U mL -1 penicillin in a humidi ed incubator with 5% CO 2 at 37 °C. The ~1.0×10 9 CFU E. coli, S. aureus, M. luteus or P. aeruginosa was incubated with GP-SiNPs-asPNA (1µM, 200 µL) in a shaking incubator at 200 rpm and 37 o C for 2 hours. Next, the treated bacteria were collected by centrifugation at 6000 rpm for 5 min. Then the collected bacteria were re-suspended and washed with PBS buffer for several times, followed by imaging by a confocal laser scanning microscope (Leica, TCSSP5 II) equipped with 30% power of diode laser. All confocal images were captured under the same brightness and contrast. The commercial image analysis software (Leica Application Suite Advanced Fluorescence Lite) was used to analyze region of interest (ROI). The scanning electron microscopy (SEM) and high-angle annular dark eld-scanning TEM (HAADF-STEM) were performed to verify the internalization of GP-SiNPs-asPNA into bacterial cells.
Inhibition of target gene expression in bacteria. GP-SiNPs-asPNA (200 µL, 1 µM) incubated with ~1.0×10 9 CFU E. coli or S. aureus in a shaking incubator at 200 rpm and 37 o C for 12 hours. After incubation, the bacteria were collected by centrifugation at 8000 rpm for 10 min, then were re-suspended and washed with PBS buffer for several times. We used the RNeasy Plus Mini kit with a genomic DNA eliminator to extract total RNA from the treated bacteria. Then we used the RNase-free water to dissolve the asextracted RNA, followed by cDNA preparation by using the cDNA synthesis kit. Afterwards, we used the qPCR to amplify each target mRNA. The whole qPCR procedure began with an initial activation step of 94°C for 3 min, followed by 40 cycles of 94 °C for 5 sec and 60 °C for 34 sec. The primer dimer formation and incorrect priming was examined by a melting curve. 16S cDNA ampli ed by primers in this case was used as the control. Agarose gel electrophoresis was running at 90 V to analyze the ampli ed nucleic acids.
In vivo imaging of diverse bacteria in mice corneas via the developed strategy. Approximately 150-200 g Sprague-Dawley (SD) female mice were purchased from Changzhou Kavins Experimental Animal Co., Ltd. All animal experiments were following the protocol approved by the animal care committee of Soochow University. Sprague-Dawley (SD) mice (female, SPF grade, 6-8 weeks old) were feed at 25 o C and 65% humidity adjusted by the ventilation equipment and air ltration system. Mice were anesthetized by intraperitoneal injection of 150 μL of 5% pentobarbital sodium solution, and then intramuscular injection of 50 μL of 2% xylazine hydrochloride. The eyelid of mice was dragged and the scratched cornea were topically administrated with one drop of 20 μL of PBS, asPNA (1 µM), SiNPs-asPNA (1 µM), SiNPs-Van (10 mg/mL) or GP-SiNPs-asPNA (1 µM). After that, the cornea was washed with PBS buffer. The infected corneas were imaged by using a uorescence stereoscopic microscope (Olympus, SZX16, λ ex = 488 nm), and the image was analyzed by the ImageJ software. The actual amount of bacteria at the infection cornea during imaging was determined via tissue harvesting, homogenization and culturing with CFU count. The number of bacterial colonies was counted by a colony counting instrument (Czone 8).
Therapy of ocular bacterial infections. The bacterial keratitis model in mice was established by using the scratch method. In details, the scars with the same depth and size were cut on the deep layer of corneal stroma of anesthetized mice by using a sterile scalpel. Afterwards, 200 μL of bacterial solution (~1.0×10 9 CFU) was dropped onto the cornea and held for 15 min. The mice were divided into ve groups, and were topically administrated with one drop of 20 μL of PBS, asPNA (1 µM), SiNPs-asPNA (1 µM), SiNPs-Van (10 mg/mL) or GP-SiNPs-asPNA (1 µM), respectively (one drop per time, three times per day). The severity of keratitis was scored based on the cornea images captured by the slit lamp. At the last day of treatment, the bacteria were extracted from the infected cornea and were cultured on the agar plates. The number of bacterial colonies was counted by a colony counting instrument (Czone 8). Meanwhile, the infected tissues from each group were xed in eyeball xative and then embedded in para n solution for H&E staining. The bacterial endophthalmitis model in mice was established by intravitreal injection.
Typically, 3 μL of ~1.0×10 9 CFU E. coli and S. aureus was injected into the vitreous cavity of the left eye of mice with a microsyringe, and the needle was removed after holding for 15 seconds. Likewise, the mice were divided into ve groups, and the cornea in each group were dropped with one drop of 20 μL of 25% poloxamer 407 (P407) gel solutions containing PBS, asPNA (1 µM), SiNPs-asPNA (1 µM), SiNPs-Van (10 mg/mL) or GP-SiNPs-asPNA (1 µM), respectively (one drop per time, three times per day). Similarly, the mice anterior segment was observed by a slit lamp. Accordingly, the severity of in ammation was scored based on the mice anterior segment photos. At the last day of treatment, the mice in each group were sacri ced and their eyeballs were removed, followed by homogenization and culturing with CFU count.
The bacterial colonies were determined by using a colony counting instrument (Czone 8). Other harvested eyeballs were treated with Davidson's solution for 30 min and 30% sucrose overnight for dehydration.
DAPI-dyed frozen sections were observed under an confocal laser scanning microscope (Leica, TCSSP5 II). Meanwhile, the infected tissues of eyeballs from each group were xed in eyeball xative and then embedded in para n solution for H&E staining.

Statistical analysis.
We used one-way ANOVA or the paired two-tailed t test (* means p < 0.05, ** means p < 0.01, *** means p < 0.001, **** means p<0.0001, ns means no signi cance) for statistical signi cance testing. We used Origin or GraphPad Prism software to perform the statistical analysis. We used commercial image analysis software (Leica Application Suite Advanced Fluorescence Lite, LAS AF Lite) and ImageJ software (NIH Image; http;//rsbweb.nih.gov/ij/) to process the region of interest (ROI) in uorescence images.
Life Science Reporting Summary. Further information on the experimental design is available in the Life Science Reporting Summary.

Figure 3
Agar plate assays to evaluate the in vitro antimicrobial activity of the developed strategy. a, Photographs of agar plates of MDR E. Coli, MRSA, P. aeruginosa and P. aeruginosa treated by GP-SiNPs-asPNA (1 μM), cipro oxacin, nor oxacin and ampicillin with various concentrations for 12 h. b, Corresponding histograms of bacterial amounts. Statistical analysis was performed using one-way ANOVA. Error bars represent the standard deviation obtained from three independent measurements (***means p < 0.001, n=3). c, Corresponding antibacterial rates. Evaluation of the hitchhiking strategy in the therapy of bacterial keratitis. a, Scheme illustrating the construction of the model of bacterial keratitis in mice and the corresponding treatment procedures. b, Fluorescent images of bacterial keratitis induced by E. coli or S. aureus at various concentrations based on the proposed strategy. The infected corneas and healthy corneas were treated with PBS, asPNA (1 μM), SiNPs-asPNA (1 μM), SiNPs-Van (10 mg mL -1 ) or GP-SiNPs-asPNA (1 μM). The actual count of E. coli or S. aureus at the infected cornea during imaging was determined by tissue harvesting, homogenization and culturing with CFU count. All imaging experiments were repeated three times with similar results. c-g, Daily slit lamp microscopic images of 2.0×10 8 CFU E. coli, S. aureus or E. coli + S. aureus -infected corneas with different treatments (e.g., PBS, asPNA (1 μM), SiNPs-asPNA (1 μM), SiNPs-Van (10 mg mL -1 ) or GP-SiNPs-asPNA (1 μM) (1 drop per time, 3 times per day)) (c), corresponding histograms of silt-lamp examination scores (d), homogenates of infected corneas with different treatments for 0, 1, 2, 3, 4 and 5 d cultured on solid LB agar (n=3) (e), corresponding quanti cation of bacterial colonization (f), and H&E staining images of infected corneas with different treatments for 5 d (g). Scale bars: 25 μm. The black arrows in slit lamp images indicate the position of the pupil. The yellow arrows in H&E staining indicate granulocyte in ltration. Statistical analysis was performed using one-way