DNA tetrahedron (Td) showed apparent superiority of crossing bacterial membrane with the help of LP2000 compared with linear DNA. They were internalized by bacteria effectively when mixed with 1/16 normal concentration of LP2000. When Td was linked with antisense oligonucleotides (ASOs) as a carrier, gene-knockdown efficiency of ASOs was influenced by the linkage mode between ASOs and Td, and the looped structure of ASOs linked to one side of the Td (Td-L) exhibited better gene-knockdown efficiency than the overhung structure (Td-O).

1. Introduction

Infections caused by antibiotic-resistant bacteria have raised public concerns worldwide. Among the assorted solutions, such as antibiotic combination(Brochado et al., 2018)and small-molecular compound screening,(Kim et al., 2018)the antisense antibacterial strategy has drawn considerable attention due to its convenient target selection, safety and decreased likelihood of inducing antibiotics resistance.(Hegarty and Stewart, 2018)By specifically blocking the expression of targeted genes in bacteria, antisense oligonucleotides (ASOs) have shown great potential to kill bacteria(Good et al., 2001)or reverse the resistance of bacteria.(Daly et al., 2017)However, without vectors, ASOs can hardly enter bacterial cells because of their high molecular weight(Frazier, 2015)and the complex structure of the bacterial cell wall,(Good et al., 2000)which are the main obstacles for ASOs in therapeutic field. Thus far, a multitude of materials have been studied for ASOs delivery, including cell-penetrating peptides (CPPs),(Good et al., 2001)vitamin B_{12(Rownicki et al., 2017)}and liposomes.(Meng et al., 2015)Among these materials, CPPs are the most widely used and have proven to be effective when covalently linked with ASOs, but their utilization is impeded by their cytotoxicity at high concentrations and potential immunogenicity to the host.(El-Andaloussi et al., 2007)Therefore, the development of alternative carriers is essential.

Recently, DNA nanomaterials have offered entirely new avenues for drug delivery systems.(Lee et al., 2016)Based on Watson-Crick base pairing, DNA nanoparticles exhibit excellent advantages over traditional nanoparticles, including precise manipulation of shape and size, biocompatibility, nontoxicity and increased likelihood of intelligent modification.(Linko et al., 2015)Particularly, DNA tetrahedra (Td) is widely used due to its simple preparation, rigid structure and flexible optimization.(Hu et al., 2017)In previous studies, it has been verified that Td can enter live HEK cells without transfection agents(Walsh et al., 2011)and deliver small-molecule compounds(Kim et al., 2013)or nucleic acid drugs(Lee et al., 2012)into eukaryotic cells. Furthermore, Td has great flexibility for structural modification by aptamers,(Charoenphol and Bermudez, 2014)folate acids(Lee et al., 2012)or tumor-penetrating peptides,(Xia et al., 2016)exhibiting considerable potential for facilitating versatile drug delivery.

Nevertheless, only a few studies have focused on the application of Td as a delivery system for antibacterial agents. Leong reported that Td intercalated with actinomycin D could be internalized efficiently byEscherichia coli (E. coli ) and Staphylococcus aureus (S. aureus ) and showed stronger antibiotic effects than free actinomycin D in vitro.(Setyawati et al., 2014)Other studies demonstrated that Td incorporated with peptide nucleic acids targetingbla _{CTX-M-group 1} in cefotaxime-resistant E. coli(Readman et al., 2017) or ftsZ in methicillin-resistantStaphylococcus aureus (MRSA)(Zhang et al., 2018)could enter bacteria to restore sensitivity to cefotaxime or to inhibit bacterial growth by inhibiting targeted genes. These results imply that Td could be a carrier to deliver ASOs into bacteria. However, the delivery efficiency of Td in different strains and the factors that influence Td into bacteria, as well as the type of Td structure or linkage modes with ASOs remain unclear.

In this study, we investigated the uptake characteristics and efficiency of Td by different bacterial strains, including S. aureus ,E. coli , Shigella flexneri (S. flexneri ),NDM1 -Klebsiella pneumoniae (K. pneumoniae ), multiple-drug resistantPseudomonas aeruginosa(P. aeruginosa ) andAcinetobacter baumannii(A. baumannii ). Next, we designed two types of linkages modes between Td and ASOs targeting gfp , encoding green fluorescent protein (GFP), or acpP , encoding the acyl carrier protein (Acp), and assessed the efficiency of delivery and gene knockdown in E. coli .

Td was prepared according to methods described previously(Goodman et al., 2005) (Figure 1a) and characterized by agarose gel, atomic force microscope (AFM) and dynamic light scattering (DLS). To verify the formation of Td , the four strands of the Td (S1, S2, S3, and S4) were added one-by-one, and the gradually formed complex presented distinct bands with slower mobility as one more strand was added, indicating the successful stepwise assembly of the Td (Figure 1b). AFM images showed that the Td exhibited an average diameter of approximately 10 nm, and a few aggregates were observed (Figure 1c). DLS analysis revealed that the Td had a hydrodynamic size of ~12 nm with a polydispersity index (PDI) of 0.3 (Figure 1d).

Then, we studied the uptake features of Td in different bacteria.E. coli or S. aureus bacterial cells were incubated with FAM-labeled Td at various concentrations (0.1, 0.5, and 1 μM) for 1.5 h, and the number of FAM-positive bacteria was analyzed by flow cytometry. As a result, the positive ratios of E. coli incubated with FAM-labeled Td at concentrations of 0.1, 0.5, and 1 μM were 5%, 35%, and 49%, respectively (Figure 2a), while the corresponding positive ratios of S. aureus were 5%, 24%, and 56% (Figure S1a). To determine whether the observed fluorescence signals represented the uptake of the Td into bacteria or simple adherence to the bacterial membrane, the bacterial cells were treated with DNase before flow cytometry analysis.(Walsh et al., 2011)We found that after DNase treatment, the positive ratios of both tested strains decreased to approximately 20% when treated with 1 μM Td (Figure 2a and S1a). However, the positive ratios of single-strand S1 labeled with FAM were lower than 5% in both tested bacteria whether treated with DNase or not (Figure 2a and S1a). Furthermore, confocal laser scanning microscopy (CLSM) also confirmed that the fluorescence intensity of the tested bacteria was reduced significantly after treatment with DNase (Figure 2b and S1b). Similar results were also observed in other bacterial strains, including S. flexneri , NDM1-K. pneumoniae ,multiple-drug resistant P. aeruginosa and A. baumannii(Figure S2). All the data demonstrated that Td had a tendency to bond with the bacterial membrane and only a fraction of Td entered into bacterial cells successfully.

Next, Lipofectamine 2000 (LP2000) was used to improve the uptake efficiency of the Td by bacteria because of the following reasons: 1) It is the cationic reagent most commonly used to neutralize the negative charge of DNA in order to reduce the electrostatic repulsion between DNA and the cellular membrane, and 2) it can electrostatically combine with the Td.(Garcia-Chaumont et al., 2000)Therefore, we explored the uptake efficiency of Td mixed with LP2000 (LP2000/Td: 0.0025, 0.0125, 0.025, and 0.125 μL/μg) for initial optimization. The Td mixed with LP2000 (LP-Td) showed the same size as the Td when the LP2000/Td ratio was 0.0025 (Figure S3a), while larger nanoparticles were formed when the ratios were greater than 0.0025. And the size increased as the ratio increased (Figure S3b and S3c). However, when the ratio of LP2000/Td was 0.125 μL/μg, the formed nanoparticles exhibited homogeneous sizes of approximately 30 nm (Figure S3d). Then we used this ratio (0.125 μL/μg) to assemble LP-Td, and characterized it by TEM and DLS (Figure 3a and 3b). To further investigate whether LP2000 at this concentration could protect Td from enzymatic hydrolysis, structural integrity of Td and LP-Td was monitored by FRET(Walsh et al., 2011). Strand S2 and S3 were labeled with cy3 and cy5 respectively, which were close enough for energy transfer from the donor cy3 to the acceptor cy5 when Td was intact. Once Td was degraded, a donor (cy3) was separated from an acceptor (cy5), then the fluorescence intensity of cy5 decreased while that of cy3 increased. Figure 3c showed that although Td and LP-Td were both degraded in 150 U/mL DNase, LP2000 protected Td from hydrolysis when DNase was 20 U/mL. This data indicated that LP-Td had higher enzyme stability than Td to some extent.

According to the results of flow cytometry and CLSM, the positive ratio in E. coliimproved gradually as an increased amount of LP2000 was added, reaching 83% when the concentration of Td was 0.5 μM and the LP2000/Td ratio was 0.125 μL/μg (Figure S4a and S4b). However, increasing the LP2000/Td ratio only slightly improved the positive ratio of Td in S. aureus , reaching a peak value of only 40%, even at the highest LP2000/Td ratio (Figure S4c and S4d). This result may be attributed to the existence of more peptidoglycan in Gram-positive bacteria, which hindered the interaction between LP-Td and the lipid membrane. Strikingly, bacteria treated with single-strand DNA exhibited very low positive ratios (< 10%) regardless of how much LP2000 was added in those ratios. However, because the difference of molecular weights between Td and single-strand DNA led to distinct absolute amounts of LP2000, we added the same amounts of LP2000 (0.125 μL/μg to 1 μM Td) to 1 μM single-strand DNA and then incubated with E. coliand S. aureus , to eliminate this interference and further clarify the superiority of Td to enter bacteria. The positive ratios were 20% and 18% in E. coli and S. aureus , respectively (Figure S5a). Importantly, when LP2000 with a 0.125 μL/μg ratio to Td was added, DNase treatment had no influence on the ratio of FAM-positive bacteria, as demonstrated by the results of flow cytometry (Figure 4a and S6a) and CLSM image analysis (Figure 4b and S6b). Furthermore, Triton X-100 was used to disrupt LP-Td nanoparticles which may adhere to bacterial membrane, and the positive ratios did not decrease (Figure S5b). All results indicated that most of the Td crossed the membrane and entered bacteria with the help of LP2000.

Next, the influences of incubation temperature and time on the uptake efficiency of LP-Td were also investigated. As a result, the uptake efficiency did not decrease when E. coli and S. aureuswere treated with LP-Td at 4 °C compared to 37 °C, indicating that the uptake process in bacteria was energy-independent (Figure S7a). After incubation with 0.5 μM LP-Td for 10, 30, 60, 180 min, we found that the uptake efficiencies in E. coli and S. aureus reached the maximum value at the first time point (10 min) and remained at relatively constant values at longer incubation times, indicating that the process of LP-Td internalization in bacteria was fairly quick (less than 10 min) (Figure S7b). As a transfection reagent, LP2000 is commonly used in eukaryotic cells with an LP2000/DNA ratio of 2-3 μL/μg. In the present study, much less LP2000 (with an LP2000/Td ratio of 0.125) could significantly facilitate bacterial uptake of Td; by contrast, this amount of LP2000 had no effect on the uptake of single-strand DNA. Based on these results, we speculate that in this uptake process, the unique structure of the Td increases its binding with bacterial surface, while the hydrophobicity of LP2000 increases the transmembrane ability of the Td. Therefore, LP-Td could be successfully internalized by bacteria. Furthermore, the quantity of LP2000 used here (1 μM Td; 0.125 μL/μg LP2000/Td ratio, corresponding to 10 μL/mL LP2000) exerted no toxicity on E. coli and S. aureus . In fact, E. coli could endure higher concentrations of LP2000 (Figure S8).

To investigate whether the LP-Td complex could deliver ASOs into bacteria and exert gene inhibitory effects, two different strategies were designed to link functional ASOs to the Td: 1) Td-L, in which a single strand was protruding from the side of the Td as a loop (Td-L) (with 3 additional nucleotides at the end of both linking sites to expose antisense sequences), and 2) Td-O, in which the single strand was overhanging at one vertex of the Td (with 7 additional nucleotides added to the end of linking site to expose antisense sequences) (Figure 5a). To improve stability, ASOs were modified by phosphorothioate. The agarose gel results indicated the successful formation of both structures, which exhibited distinct bands with a slightly slower mobility than Td (Figure 5b). DLS demonstrated that the sizes of Td-L and Td-O were similar to that of Td (Figure 5c) and increased to ~30 nm when LP2000 was added (Figure S9).

Next, the gene-inhibiting effects of Td-L and Td-O carrying anti-gfp ASOs were studied in E. coli expressing GFP (GFP-E. coli ). Mismatched ASOs linked to Td (Td-L_mis and Td-O_mis) were used as controls to confirm the specific gene-inhibiting effect. A confocal study showed that Td-L_anti-gfp , Td and Td-L_mis at a concentration of 1 μM all had no influence on the fluorescent intensity of GFP-E. coli (Figure 6a and 6b). In contrast, when treated with LP-Td-L_anti-gfp (1 μM), a 75% reduction in GFP fluorescence intensity was observed in GFP-E. coli , while LP-Td and LP-Td-L_mis also had no effect (Figure 6a and 6b). However, unlike Td-L_anti-gfp , Td-O_anti-gfp did not affect GFP fluorescence regardless of whether LP2000 was used (Figure 7a and 7b), implying the importance of the specific structure of Td-L for the gene inhibitory effect.

Then, we tested whether the observed decrease in fluorescence intensity in GFP-E. coli resulted from targeted gene inhibition. After incubating for 3 h, bacteria were collected, and total RNA was extracted to detect the mRNA level of gfp . The results in Figure 6c show that when treated with LP-Td-L_anti-gfp at concentrations of 0.5 and 1 μM, gfp expression decreased by 41.5% and 60.5%, respectively, while treatment with 0.1 μM LP-Td-L_anti-gfp did not lead to an observable decrease in gfp expression. Moreover, no significant decreases ingfp mRNA level were observed in the other groups at all concentrations. In addition, Figure 7c indicates that Td-O_anti-gfp and LP-Td-O_anti-gfp did not inhibit the gfpexpression level, consistent with the confocal results. However, flow cytometry analysis demonstrated that the bacterial uptake efficiencies of Td-L and Td-O were comparable (Figure 5d), indicating that the loop structure in Td-L facilitated its gene inhibition activity. The loop design may help ASOs combine with RNA or increase the sensitivity of the ASOs/RNA complex to RNase. Further studies to uncover the underlying mechanisms of the antisense effect of Td-L can provide potential strategies to design more effective linkage modes between Td and ASOs.

Finally, we investigated the antibacterial activity of Td-L carrying anti-acpP ASOs (Td-L_anti-acpP ). Here,acpP is a gene encoding the acyl carrier protein Acp, which is critical for fatty acid biosynthesis in E. coli . As shown in Figure 8a, 0.1 μM LP-Td-L_anti-acpP did not influence the growth ofE. coli . When the concentration was increased to 0.5 μM, LP-Td-L_anti-acpP significantly inhibited the growth of E. coli at 5 h after treatment, as highlighted by the reduced colony-forming units (CFU) in the LP-Td-L_anti-acpP group compared to the other groups. Furthermore, 1 μM LP-Td-L_anti-acpP exhibited inhibitory effects on bacterial growth at 3 h and stronger effects at 5 h. Then, we further analyzed the mRNA level of acpPto examine whether the LP-Td-L_anti-acpP -induced bacterial inhibition was mediated by targeting acpP. As a result, the mRNA level of acpP was found to significantly decrease in the LP-Td-L_anti-acpP group compared to the other groups. The mRNA expression level of acpP decreased by 23% and 43% after treatment with 0.5 μM and 1 μM LP-Td-L_anti-acpP , respectively (Figure 8b). Collectively, these results demonstrated that LP-Td-L_anti-acpP exhibited antibacterial activity via its gene inhibitory effects targeting acpP and that LP2000 played a fairly essential role in the process.

In this study, we investigated the uptake efficiency of DNA tetrahedron (Td) in different bacterial strains. Interestingly, the Td adhered to the surface of the bacterial membrane efficiently, but a few could penetrate the membrane, probably due to its hydrophilicity. However, a very low ratio of LP2000 to Td facilitated the penetration of the Td into bacteria within a few minutes in an energy-independent manner. Compared with single strand DNA, Td showed superiority of crossing bacterial membrane with the help of LP2000, the reason of which may be its characteristic structure. In addition, based on the optimized LP-Td system, the modes of ASOs linked with the Td had a significant impact on gene knockdown efficiency, with the looped structure of ASOs linked to one side of Td exhibiting excellent gene-inhibiting activity. Importantly, LP-Td-L_anti-acpP showed gene inhibitory effects targeting the acpP gene and eventually inhibited bacterial growth. Therefore, this study provides novel insights into the uptake features of Td in bacteria and highlights the importance of linkage strategy between ASOs and Td. In further research, multivalent ASO-containing particles could be designed based on the superior structure of Td and the selectivity of ASOs linking modes, which will open novel opportunities for developing effective antisense delivery systems.