Research ArticleMICROBIOLOGY

Mechanical penetration of β-lactam–resistant Gram-negative bacteria by programmable nanowires

See allHide authors and affiliations

Science Advances  03 Jul 2020:
Vol. 6, no. 27, eabb9593
DOI: 10.1126/sciadv.abb9593


β-Lactam–resistant (BLR) Gram-negative bacteria that are difficult or impossible to treat are causing a global health threat. However, the development of effective nanoantibiotics is limited by the poor understanding of changes in the physical nature of BLR Gram-negative bacteria. Here, we systematically explored the nanomechanical properties of a range of Gram-negative bacteria (Salmonella, Escherichia coli, Pseudomonas aeruginosa, and Klebsiella pneumoniae) with different degrees of β-lactam resistance. Our observations indicated that the BLR bacteria had cell stiffness values almost 10× lower than that of β-lactam–susceptible bacteria, caused by reduced peptidoglycan biosynthesis. With the aid of numerical modeling and experimental measurements, we demonstrated that these stiffness findings can be used to develop programmable, stiffness-mediated antimicrobial nanowires that mechanically penetrate the BLR bacterial cell envelope. We anticipate that these stiffness-related findings will aid in the discovery and development of novel treatment strategies for BLR Gram-negative bacterial infections.


The β-lactam antibiotics have been among the most successful drugs with regard to reducing human morbidity and mortality for the past 60 years (1) because of their excellent safety profile, breadth of spectrum of activity, and low cost (2). However, as a result of the intensive use of β-lactam antibiotics in human therapy and animal agriculture, many bacterial pathogens have acquired β-lactam resistance and can cause infections that are essentially untreatable, representing an increasing threat to public health (3). This situation is especially troubling with respect to β-lactam–resistant (BLR) Gram-negative bacteria, as they are harder to kill than Gram-positive bacteria. The difficulty in eradicating BLR Gram-negative bacteria is largely due to the permeability barrier that is provided by their unique cell envelope, in which the outer membrane is very challenging for small molecules to cross (4).

To combat BLR Gram-negative bacteria, numerous conventional antibiotics replacements including chemical antibacterial nanomaterials such as metal nanoparticles (Au and Ag) (5, 6), metal oxide nanoparticles (ZnO and TiO2) (7, 8), photothermal agents (9), and graphene oxide nanosheets (10) and physical antibacterial materials such as antimicrobial peptides (4) and cationic polyelectrolytes (11) have been reported to decrease the possibility of antibiotic resistance. However, all of these antibacterial strategies may fail to achieve effective killing of BLR Gram-negative bacteria because they rely only on material characteristics such as toxicity (12), photothermal conversion efficiency (13), photocatalysis performance (14, 15), and charge characteristics (11), ignoring the acquired mechanical properties of BLR bacteria. Therefore, understanding the scientific basis of β-lactam resistance is essential to find an effective strategy against BLR Gram-negative bacteria. Recent studies have demonstrated that Gram-negative bacteria can acquire or develop resistance to β-lactam via several mechanisms such as genic mutations, resistance gene transfer, overexpression of efflux pumps, and production of β-lactamase (16), but the physical nature of BLR Gram-negative bacteria remains largely unexplored.

Here, we systematically studied the mechanical properties of 22 clinical isolates of the Gram-negative bacteria Salmonella, Escherichia coli, Pseudomonas aeruginosa, and Klebsiella pneumoniae to investigate the association between cell envelope mechanical behavior and acquisition of β-lactam resistance. We found that BLR Gram-negative bacteria exhibited a significant decrease in stiffness compared to β-lactam–susceptible (BLS) bacteria, indicating that changes in mechanical properties were correlated with acquired β-lactam resistance. To identify the mechanism of reduced stiffness, we used a fluorescent probe for in situ visualizing bacterial peptidoglycan (PG) biosynthesis, which revealed that this reduced stiffness could be induced by reduced PG biosynthesis resulting from long-term exposure to subinhibitory concentrations of β-lactam antibiotics. This change was further evidenced by a thickness analysis of bacterial cell envelope through cryo–transmission electron microscopy (cryo-TEM) and TEM. On the basis of these stiffness findings, we first developed a cell membrane penetration model to explore the possibility of mechanically killing bacteria and established criteria of cell membrane penetration. We then validated this prediction experimentally in Salmonella strains by varying the tip diameters of NiCo(OH)2CO3 nanowires (NWs), showing that the soft BLR bacteria were more easily penetrated by sharp NWs. Our results provide insight into potential strategies for killing BLR Gram-negative bacteria with a combined focus on bacterial mechanical properties and performance of nanomaterials.


Nanomechanical properties of Gram-negative bacteria

We begin by asking whether the acquisition of β-lactam resistance is associated with changes in physical properties of the Gram-negative bacterial cell envelope. To systematically address this issue, we collected and characterized 22 clinical isolates of Salmonella, E. coli, P. aeruginosa, and K. pneumoniae (table S1), well-known pathogenic Gram-negative bacteria that usually cause a variety of life-threatening hospital-acquired infections (17). The selection and classification of BLS and BLR strains here are according to their degrees of β-lactam resistance. To this end, the antimicrobial resistance of these collected isolates was first identified using minimum inhibitory concentrations (MICs) to verify the selected bacterial populations including BLS and BLR bacteria (Fig. 1A and table S2). Several typical classifications of β-lactam antibiotics, the most important class of antibiotics currently in clinical use, have been tested, including penicillins, cephalosporins, carbapenems, and monobactams, which interfere with cell wall biosynthesis by covalent binding to the active site of penicillin-binding proteins (PBPs) (18). In addition, we used the acquired β-lactam resistance genes detected by using whole-genome sequencing and extended spectrum β-lactamase (ESBL) test to further evaluate the degree of β-lactam resistance (table S1). In combination of these results, the obtained Salmonella isolates (7 and 12), E. coli isolates (3, 17, and 15), P. aeruginosa isolates (84, 85, and 86), and K. pneumoniae isolates (3, 4, and 5) were deemed to be BLS bacteria, while the collected Salmonella isolates (44, 79, and 83), E. coli isolates (2 and 8), P. aeruginosa isolates (87, 88, and 89), and K. pneumoniae isolates (2, 12, and 25) were treated as BLR bacteria (see the details in Supplementary Discussion). Note that the collected P. aeruginosa isolates (87, 88, and 89) were resistant to all β-lactam antibiotics tested, indicating that such “superbugs” were very difficult to treat by classical β-lactam antibiotics.

Fig. 1 Nanomechanical properties of BLS and BLR Gram-negative bacteria.

(A) Heat map representation of antimicrobial resistance of five Salmonella isolates, five E. coli isolates, six P. aeruginosa isolates, and six K. pneumoniae isolates on a logarithmic color scale. Red represents a high level of antibiotic resistance (MIC ≥ 64 μg/ml), while blue represents a low level of antibiotic resistance (MIC ≤ 0.125 μg/ml). PEN, penicillin; AMP, ampicillin; PIP, piperacillin; AMC, amoxicillin-clavulanic acid; OXA, oxacillin; CAZ, ceftazidime; FEP, cefepime; CTX, cefotaxime; CRO, ceftriaxone; CXM, cefuroxime; CZO, cefazolin; IPM, imipenem; MEM, meropenem; BIPM, biapenem; ATM, aztreonam. (B) Strategy for measuring stiffness of Gram-negative bacteria using liquid AFM with a bacterium immobilized on a poly-l-lysine–coated glass substratum. (C) AFM topographic image recorded in PBS showing S.79 on a modified substratum. Scale bar, 500 nm. (D) The height profile along the white dotted line in the image (C). (E) Representative Derjaguin-Muller-Toporov Young’s modulus map of S.79. Scale bar, 1 μm. (F) A high-magnification elasticity map is shown, and the enlarged section is outlined in white in (E). Scale bar, 200 nm. (G to J) Quantification of the average Young’s modulus of the five Salmonella isolates (G), five E. coli isolates (H), six P. aeruginosa isolates (I), and six K. pneumoniae isolates (J). The BLS isolates are indicated with a pink band, while the BLR isolates are indicated with a blue band in (G to J). Thick horizontal lines and error bars represent mean and SD, respectively. Student’s t test, **P < 0.01.

To elucidate and correlate the respective nanomechanical profiles with bacterial resistance, we performed liquid atomic force microscope (AFM) analyses of Gram-negative bacteria in buffer corresponding to physiological conditions with the bacteria attached to poly-l-lysine–coated glass substrata to obtain high-resolution data (Fig. 1B). Figure 1C represents the topography of the chosen Salmonella isolate 79, which shows the expected rod-like shape roughly 700 nm in height (Fig. 1D), in good agreement with the typical morphology of Salmonella. Then, we directly detected the rigidity of all the obtained strains by measuring the Young’s modulus of bacteria. Figure 1E shows the representative AFM Young’s modulus mapping of Salmonella isolate 79 imaged in phosphate-buffered saline (PBS) in contact mode. To characterize the nanomechanical properties in detail, a high-magnification elasticity map was obtained from the location marked in Fig. 1E, which indicated that the bacterial stiffness was evenly distributed, with only minor variations in stiffness (Fig. 1F), likely because of internal metabolism. For better elaboration and comparison of the stiffness information from four different Gram-negative bacteria species, the Young’s modulus data were collected in histograms (figs. S1 to S4). The representative Salmonella isolates (7 and 12) exhibited a monomodal distribution with the average cell stiffness value of 3.31 ± 0.79 and 4.29 ± 0.86 MPa, respectively. By contrast, the average cell stiffness for the BLR Salmonella isolates (44, 79, and 83) showed significantly decreased average cellular elasticity, with values of 0.58 ± 0.13, 0.40 ± 0.16, and 0.48 ± 0.11 MPa, respectively (Fig. 1G). The finding that the BLR bacteria had cell stiffness almost 10× softer than that of BLS bacteria was consistent among other Gram-negative bacteria, such as E. coli, P. aeruginosa, and K. pneumoniae (Fig. 1, H to J). These modulus values measured under our experimental conditions were comparable to previously reported bacterial rigidity (19, 20), suggesting that our experimental setup allowed us to accurately assess the effects of acquired β-lactam resistance on the mechanical properties of the bacterial cell envelope. Collectively, our results strongly indicate that the changes in stiffness are associated with acquisition of β-lactam antibiotic resistance in Gram-negative bacteria.

Analysis of the mechanism of reduced stiffness

The Gram-negative bacterial cell envelope structure includes a PG layer that lies in a periplasmic space located between the outer membrane and inner membrane, providing mechanical strength and rigidity for the bacterium (Fig. 2A). The final stages of PG biosynthesis are catalyzed by PBPs, which have both transglycosylase and transpeptidase (Tpase) activities and are required to generate glycan strands and cross-linked adjacent glycan strands via stem peptides, respectively (21). β-Lactam antibiotics consisting of a four-member “β-lactam” ring act as a structural mimetic of the terminal d-Ala-d-Ala moiety of the donor stem peptide in transpeptidation. Therefore, β-lactam antibiotics inhibit PG biosynthesis by blocking the active site of Tpase (Fig. 2B). Gram-negative bacteria have developed different mechanisms to resist the action of β-lactams, including destroying the antibiotic molecule through the action of β-lactamases (Fig. 2C). This important role of β-lactam antibiotics in PG biosynthesis led us to hypothesize that the reduced bacterial stiffness resulted from exposing bacteria to subinhibitory concentrations of drugs during the selection of bacterial drug resistance, thus reducing PG biosynthesis. To test this, we performed an analysis of the mechanical properties of the typical BLS E. coli (ATCC25922) strain and New Delhi metallo-β-lactamase 1 (NDM-1)–producing wild-type E. coli strain after exposure to different concentrations of ceftazidime (CAZ) (Fig. 2D and fig. S5, A and B). First, it can be seen that the stiffness of NDM-1–producing wild-type strain was also nearly 10× lower than that of BLS strain before adding CAZ, which further confirmed that decrease in stiffness was associated with acquisition of β-lactam antibiotic resistance. Second, with increasing concentrations of CAZ, the stiffness of BLS E. coli gradually decreased while that of NDM-1–producing wild-type E. coli remained at approximately 0.7 MPa. This is because the BLS E. coli were susceptible to CAZ, and as a result, PG biosynthesis was inhibited. By contrast, NDM-1–producing wild-type E. coli isolate can produce β-lactamases, as confirmed by reaction with nitrocefin (fig. S5, C and D), which has the ability to hydrolyze β-lactam antibiotics (e.g., CAZ). These results revealed that the inhibition of PG biosynthesis by β-lactam antibiotics altered bacterial stiffness.

Fig. 2 Mechanism of reduction in stiffness for BLR Gram-negative bacteria.

(A) The cell envelope of Gram-negative bacteria consists of lipopolysaccharide (LPS) molecules, an outer membrane (OM), a thin single layer of PG, and an inner membrane (IM). (B) β-Lactam antibiotics inhibit PG synthesis by inhibiting the peptide bond formation reaction that is catalyzed by Tpase. (C) Breaking of the β-lactam ring by β-lactamase. (D) Quantification of the average Young’s modulus of the BLS and NDM-1–producing E. coli treated with different concentrations of CAZ. Thick horizontal lines and error bars represent mean and SD, respectively. (E and F) Pulse labeling of BLS [S.7 and S.12; (E)] and BLR [S.44, S.79, and S.83; (F)] strains with HADA to determine PG synthesis. Scale bars, 2 μm. (G and H) Schematic representations showing the change of periplasmic gel in BLS bacteria (G) and BLR bacteria (H) following dehydration. (I and J) Cryo-TEM images of BLS Salmonella isolate (S.7) at low magnification (I) and high magnification (J). (K) TEM image of S.7. (L and M) Cryo-TEM images of BLR Salmonella isolate (S.79) at low magnification (L) and high magnification (M). (N) TEM image of S.79. Regions of interest are indicated by yellow arrows in (I to N). Scale bars, 200 nm (I and L). 50 nm (J and M), and 100 nm (K and N). (O and P) Quantification (n = 20) of the cell envelope thickness in five Salmonella isolates measured from cryo-TEM images (O) and TEM images (P). Results are represented as means ± SD. Student’s t test, **P < 0.01; ns, not significant.

It is widely believed that the PG biosynthesis would be inhibited with the existence of β-lactam antibiotics (18). However, our mechanical results revealed that BLR Gram-negative bacteria exhibited a significant decrease in stiffness without adding β-lactam antibiotics, indicating that the alteration of stiffness was an intrinsic property of collected BLR isolates that can be expressed in the absence of β-lactam antibiotics. To examine this possibility, we used a published protocol for in situ labeling of PG of Salmonella cells with 7-hydroxycoumarin-3-carboxylic acid 3-amino-d-alanine (HADA), a fluorescent d-amino acid (FDAA) (22). These small-molecule probes incorporate at the terminus (fifth position) of the stem peptide of nascent PG, wherein FDAAs act as a surrogate acceptor strand in the transpeptidation reaction (23). When pulse-labeled for 3 hours with 500 μM HADA, both BLS isolates showed substantial incorporation of HADA at the peripheral region, while weaker peripheral HADA fluorescence was observed for BLR isolates (Fig. 2, E and F). This was presumably due to reduced PG biosynthesis, resulting in significantly lower fluorescence intensity compared with BLS isolates (fig. S6). This change was further evidenced by the difference in attenuated total refection Fourier transform infrared spectroscopy (ATR-FTIR) spectral profiles observed in the polysaccharides region (fig. S7).

Unlike the cell wall of Gram-positive bacteria, which have a thicker and multilayered PG layer (20 to 50 nm) sheath outside of the cytoplasmic membrane, the PG of Gram-negative bacteria consists of a single layer (approximately 5 nm) located in the periplasm. However, this classical view of the periplasm has been replaced by a concept of periplasmic gel consisting of cross-linked PG that fills most of the periplasmic space (24). Inspired by the swelling-shrinking properties of hydrogel, we next hypothesized that the periplasmic gel of BLS and BLR bacteria displayed different swelling-shrinking behaviors owing to decrease in PG biosynthesis (Fig. 2, G and H). To test this possibility, we compared the thickness of the cell envelope in two different states, frozen-hydrated and dehydrated, observed by cryo-TEM and TEM, respectively (Fig. 2, I to N, and fig. S8). There was no difference in the thickness of the cell envelope for BLS and BLR Salmonella isolates in a frozen-hydrated state (Fig. 2O). Since these frozen-hydrated cells retained their structures in the native environment, these measured thicknesses of the cell envelope should be accurate for bacteria. When the bacterial cells were dehydrated, the cell envelope thicknesses of BLS bacteria (isolates 7 and 12) changed from ~40 to 28.83 ± 1.63 nm and 30.32 ± 1.91 nm, respectively. However, with regard to the BLR bacteria (isolates 44, 79, and 83), the thickness of the cell envelope decreased to 14.46 ± 0.97, 14.73 ± 0.93, and 16.99 ± 1.50 nm, respectively, which were half that of BLS Salmonella (Fig. 2P). This difference was also observed in other Gram-negative bacteria, including E. coli, P. aeruginosa, and K. pneumoniae (figs. S9 to S11), suggesting that the decreased thicknesses of the dehydrated cell envelope could be attributed to the decrease in PG biosynthesis. Note that the size of BLS and BLR Gram-negative bacteria showed no marked differences after dehydration with ethanol (figs. S12 to S15). The combined results of cryo-TEM and TEM further indicated that a decrease in PG biosynthesis occurred in BLR Gram-negative bacteria, leading to reduced stiffness.

Computational modeling of cell membrane penetration

On the basis of the observation that BLR bacterial stiffness was altered from the native state and the knowledge that cell mechanics play a key role in cell membrane penetration events (25), we hypothesized that the stiff BLS and soft BLR bacteria may display different cell envelope penetration properties when prodded with an NW (Fig. 3, A and B). This penetration may provide the basis for an alternative antibacterial strategy in which NWs mechanically penetrate the BLR bacterial cell envelope, leading to cell death. Hence, gaining a deeper understanding of the bacteria penetration events is of great importance. However, the current criteria for cell penetration have been calculated from synthetic or membrane penetration experiments and are not suitable for the Gram-negative bacterial cell envelope because of its special structure. Therefore, to perform rigorous bacterial penetration studies, it is critical to establish consistent criteria to assess when (or whether) bacterial envelope penetration events occur.

Fig. 3 Computational modeling of cell membrane penetration.

(A and B) Schematic illustration of membrane penetration depending on bacterial stiffness. The single NW fails to penetrate stiff BLS bacteria (A) but is able to penetrate BLR bacteria (B). (C and D) The density distribution of bacterial free energy and membrane deformation as the bacteria deform on the three different NWs (d = 5, 50, and 100 nm) for stiff BLS bacteria (C) and soft BLR bacteria (D). The driving force was assumed to be 100 pN, and the deformation was amplified by a factor of 6. (E) The indentation depth of BLS and BLR bacteria at varying NW diameters, indicating that deformation, is highly dependent on NW tip diameter and bacterial stiffness. The inset shows that the cell envelope undergoes large-scale deformation due to NW indentation, while the bacteria maintain a constant volume. (F) The maximum value of bacterial free energy density produced by NW indentation with various NW tip diameters. (G) Finite element method (FEM) analysis of the relationship between Young’s modulus and the maximum value of bacterial free energy under a constant driving force of 100 pN (d = 5 nm). A penetration criterion (free energy density, 2.9 × 10−3 J/m2) separates the free energy density–tip diameter/Young’s modulus planes into pink (penetrated) and blue (intact) regions.

Bacterial penetration occurs when the needle tip of an NW touches the surface of a bacterium, deforms it, penetrates the outer membrane, and then penetrates the entire bacterial envelope. To simplify this issue, we regard the penetration of the bacterial cell membrane as the penetration of the bacterial envelope. Cell membrane penetration can be described by the activation energy theory (26). On the basis of this theory, thermally activated molecular-scale defects arise and disappear spontaneously in membranes, with the steady-state hole formation rate affecting the probability of rupture. It has been reported that the free energy of such deformed cells plays a critical role in the rupture of the membrane (25). According to previous research, cell membrane penetration occurs when the free energy per unit area is larger than a threshold value [(2.9~8.3) × 10−3 J/m2].

From a mechanical perspective, the bacterial envelope can be considered as a composite elastic shell (27, 28). The mechanical equilibrium of the bacterial shell is determined by its surface tension and its elastic deformation and turgor pressure. Hence, the free energy per unit area of a deformed bacterial envelope can be expressed in the form (28)f=Tr+EelasticThe tension (surface energy per unit area) is determined by Tr = Tin + Td, where Tin is the intrinsic tension at zero deformation caused by the bacterial turgor pressure and Td represents the indentation tension that arises from the deformation of the bacterial envelope. ∆Eelastic is the elastic energy induced by the deformation of the bacteria, which is often ignored in the bulge or vesicle indentation model (25) but plays a non-negligible role in the bacterial envelope model (27). Therefore, a useful method of quantifying membrane penetration is to simulate the free energy density distribution of deformed bacteria and compare its maximum value with threshold. On the basis of our experimental results, we chose the minimum value of threshold (2.9 × 10−3 J/m2) as the critical value in our following simulation to determine whether the bacteria were penetrated or not. To implement this approach, finite element method (FEM) analysis was performed to obtain the maximum free energy by using COMSOL Multiphysics version 5.2 on the basis of an inflated spherocylinder shell with isotropic elasticity model. Considering a realistic condition in which bacteria touch the NW surface from different directions, we started by analyzing the free energy of bacteria at different penetration angles (θ; fig. S16). The maximum value of the free energy reached a minimum when bacteria were perpendicular to NWs (θ = 0°). This suggests that if bacteria are penetrated from perpendicular angles, they are certain to be penetrated from other angles. Thus, we chose this case (θ = 0°) to determine whether the bacteria were penetrated for the subsequent FEM simulations.

Next, we analyzed the representative cases of NWs with tip diameters of 5-, 50-, and 100-nm indentations into a bacterium under a constant driving force of 100 pN (Fig. 3, C and D), which was calculated from our previously reported bendable NWs (29). As shown in Fig. 3, C and D, the free energy density is concentrated at the tip of the NW, and its maximum value decreases with increasing tip diameter. The soft BLR bacteria allow the generation of a maximum free energy density of 3.18 × 10−3 J/m2 when being indented with an NW with a tip diameter of 5 nm, which is greater than the corresponding value for the stiff BLS bacteria (0.86 × 10−3 J/m2). Moreover, the stiff BLS bacteria exhibited no significant change in shape according to NW tip diameter, while the soft BLR bacteria were markedly more distorted by NWs with sharper tips. This difference in deformation was further evidenced by their corresponding indentation depth results (Fig. 3E), confirming that the extent of this swelling was strongly dependent on the NW tip diameter and the bacterial stiffness. As expected, the soft BLR bacteria required a small critical tip diameter to induce membrane penetration and were triggered only when the NW tip diameter was 5 nm (Fig. 3F). By contrast, membrane penetration of stiff BLS bacteria did not occur in this range of NW tip diameter. These results suggested the possibility of directly penetrating only BLR bacteria using sharp NWs (d ≤ 5 nm). To determine the critical bacterial stiffness that can be penetrated by an NW with a diameter of 5 nm, bacteria with different stiffness were used for FEM analysis (Fig. 3G). Bacteria with low stiffness (E < 0.6 MPa) could be readily penetrated according to this model, which also provides a basis for detecting bacterial stiffness based on cell membrane penetration events.

The driving force for bacterial capture is generally focused on molecular recognition processes such as carbohydrate-protein interaction (30) and antigen-antibody binding (31), which vary in the order of piconewton. To test the role of driving force in cell membrane penetration, we examined the amount of vertical force required to achieve free energy required for critical failure within the membrane. With increasing driving force, the deformation of soft BLR bacteria became more apparent, while the stiff BLS bacteria retained their original rod-like shapes (fig. S17, A and B). The maximum free energy density at the NW tip increased with driving force, with penetration occurring at driving force ≥100 pN for BLR bacteria and driving force ≥300 pN for BLS bacteria when the NW tip diameter was 5 nm (fig. S17C). Notably, the minimum driving force required to induce membrane penetration increased with increasing tip diameter (fig. S17, D and E).

Experimental testing of simulation predictions by programmable NWs

To experimentally validate the predictions of our computational model, large-area high-density NiCo(OH)2CO3 NW arrays grown uniformly on fluorine-doped tin oxide (FTO) were used first as bacterial stiffness nanoprobes using a simple hydrothermal synthesis method (fig. S18, A and B). The obtained three-dimensional hierarchical NiCo(OH)2CO3 NWs were integrated and had sharp tips, which had a needle-like shape with a length of 5 ± 0.5 μm, root diameter of 160 ± 10 nm, and tip diameter of 20 ± 15 nm (fig. S18, C and D). The crystallographic structure of the NWs was investigated by x-ray diffraction, as shown in fig. S18E, and matched well with that of monoclinic binary NiCo(OH)2CO3 reported previously (32). The single-crystalline structure of NiCo(OH)2CO3 NWs was confirmed by high-resolution TEM (HRTEM), as well as the corresponding selected area electron diffraction pattern (fig. S18F), which had lattice fringes with interplanar spacing of 0.25 nm corresponding to the (040) plane of NiCo(OH)2CO3.

To test the prediction that the membrane penetration also depended on the NW tip diameter, we next aimed to alter the tip diameter of NWs through varying the aqua regia etching time to investigate the tip size effect (fig. S19A). The original NiCo(OH)2CO3 NWs with tip diameter of 5 ± 1.8 nm were converted to a series of NWs with increasing tip diameter controlled by etching time (fig. S19, B to J). Eventually, these NWs reached an average tip diameter of 120 ± 7.8 nm after exposure in aqua regia solution for 8 min (fig. S19I). However, it was difficult to obtain larger tip diameters of NWs with this etching method because further increases in etch time resulted in exposed NWs ultimately falling off the FTO substratum (fig. S20).

Membrane penetration involves two steps, bacterial capture and indentation, both of which require a driving force. To this end, concanavalin A (Con A), as a bacterium-binding molecule that offers a driving force, was introduced to the surface of substrata by sequential chemical covalent coupling (fig. S21). The driving force was produced by specific lectin-carbohydrate recognition between Con A and mannose of the bacterial surface lipopolysaccharide (29), which enabled the bacteria to adhere to the substrata. To optimize the incubation time for optimal bacterial capture, a set of bacteria capture experiments with different incubation times were carried out using Con A–coated NiCo(OH)2CO3 NW/FTO and FTO as substrata and Salmonella isolate 79 (BLR) as the target (fig. S22A). The maximal bacterial capture efficiency was reached after 60 min of incubation on the NW/FTO substratum, while the bacterial capture efficiency of smooth FTO showed no significant change with increasing incubation time, indicating the critical role of the NW topography in enhancing bacterial capture. For consistent comparison, 60 min was chosen as the incubation time in subsequent bacterial penetration experiments. Moreover, minimal nonspecific interactions were present between the BLR Salmonella (isolates 44, 79, and 83) and the NiCo(OH)2CO3 NW substrata in the absence of any surface modification (fig. S22B), suggesting that the bacterial binding molecules modified on the surface promoted bacterial capture.

On the basis of our modeling results, we then selected three kinds of NiCo(OH)2CO3 NWs (tip diameters: d = 5, 50, and 100 nm) to examine the roles of bacterial stiffness and NW geometry in cell envelope penetration (Fig. 4, A to C). Viability analysis of the bacteria in contact with NWs was performed to characterize the penetration efficiency using a LIVE/DEAD bacterial assay in which nonpenetrated bacteria with intact cell membranes were stained fluorescent blue while penetrated bacteria with ruptured membranes were stained fluorescent red. The BLR Salmonella (isolates 44, 79, and 83) captured on NWs (d = 5 nm) were efficiently penetrated (~95%), while the NWs exhibited lower penetration efficiency (~20%) to BLS Salmonella (isolates 7 and 12; Fig. 4A). The number of BLR bacteria penetrated decreased significantly with blunting of the NW substrata (d = 50 and 100 nm; ~38 and ~18%, respectively), whereas the percentage penetration of BLS bacteria showed little change (Fig. 4, B and C). It is notable that the penetration efficiency of each BLS or BLR Salmonella isolate did not significantly differ despite the change in substratum. Furthermore, compared to BLS bacteria, we found that the BLR bacteria were more prone to be captured on the substrata, especially in the case of sharpest NWs (Fig. 4D), which can be attributed to their different penetration efficiencies (Fig. 4, E and F). Given that the substrata were washed three times with deionized (DI) water during confocal laser scanning microscopy (CLSM) samples preparation, the penetrated BLR bacteria were still immobilized on the NWs, especially for sharp NWs (d = 5 nm), while intact BLS bacteria captured on the NWs were washed away.

Fig. 4 Experimental validation of cell membrane penetration.

(A to C) Left: TEM images of NiCo(OH)2CO3 NWs with tip diameters of 5 nm (A), 50 nm (B), and 100 nm (C). Right: CLSM overlay projections of nonpenetrated (blue fluorescent) and penetrated (red fluorescent) Salmonella (BLS isolates 7 and 12 and BLR isolates 44, 79, and 83) on three different substrata. Corresponding percentages of nonpenetrated and penetrated bacteria are plotted in the images in the right corner. Scale bars, 100 nm in TEM images and 50 μm in CLSM images. (D) The bacterial capture efficiencies of the three different NW substrata (d = 5, 50, and 100 nm) for BLS and BLR Salmonella were calculated as the number of bacteria captured per square micrometer of projected sample surface area (n = 3). (E) The difference in bacterial penetration efficiency of the three different substrata (d = 5, 50, and 100 nm) for BLS and BLR Salmonella (n = 3). (F) The penetration efficiencies of BLS (S.7) and BLR (S.79) Salmonella with different NW tip diameters (n = 3). (G) Representative SEM images of Salmonella captured on NW substrata (d = 5 nm) after incubation for 60 min. The BLS Salmonella (isolates 7 and 12) retained their intact morphology (indicated with blue arrows), whereas the BLR Salmonella (isolates 44, 79, and 83) appeared to be significantly penetrated through interaction with the sharp NWs (indicated with red arrows). Scale bars, 500 nm. Values are presented as means ± SD. Student’s t test, ***P < 0.001 and *P < 0.1.

Last, the process of bacterial penetration by sharp NWs was directly visualized by scanning electron microscopy (SEM; Fig. 4G and fig. S23). The BLR bacteria could be readily penetrated by the NWs with a tip diameter of 5 nm, while the BLS bacteria remained intact despite contact with NWs (Fig. 4G). Upon increasing NW tip diameter to 50 and 100 nm, it was difficult to observe membrane penetration of BLS or BLR bacteria (fig. S23). This penetration result was further substantiated by counting the number of bacteria that were impaled using SEM images (fig. S24), which was consistent with the bacterial viability test outlined above (Fig. 4, A to C). These findings confirmed that bacterial stiffness plays an important role in membrane penetration, with higher penetration efficiency for softer bacteria that are more sensitive to NW geometry.


NW-based platforms have been widely used for intracellular delivery of biomolecules (33), modulating stem cell differentiation and investigating basic cell physiology (e.g., division and migration) (34). Very different from mammalian cells, which can survive for as long as a week after physical penetration by NWs (35), BLR bacteria are more susceptible to cell death after mechanical penetration, which could be attributed to their intrinsic characteristics, e.g., leakage. Gram-negative bacteria are penetrated by an impaling mechanism, in which penetration driven by molecular force causes irreversible bacterial disruption, as confirmed by CLSM images (Fig. 4, A to C). In contrast, mammalian cells have larger size, lower stiffness, and a cytoskeleton structure, which enables an adhesion-mediated process (25), in which cells first adhere to the substrate and penetration is then induced by adhesive force. In this case, the integrity of the cellular membrane is preserved despite its penetration by NWs (36). Medical devices such as urinary catheters, hemodialysis equipment, intravenous infusion devices, and orthopedic implants are easily infected with pathogenic Gram-negative bacteria, which can result in serious illness, long hospital stays, or even death. According to our proof of concept of mechanical penetration of bacteria by NWs, NWs could be applied to one or more inner surfaces of a medical device by growing the NWs directly on one or more surfaces. In contrast to conventional physical methods that require other resources, such as power, light, and heat sources, the energy efficiency and strong efficacy against BLR bacteria offer enormous and untapped potential for use of NWs in medical devices for preventing bacterial infection.

In summary, our findings are notable because they demonstrate that a decrease in BLR Gram-negative bacterial stiffness is associated with the acquisition of β-lactam antibiotic resistance, which is attributed to reduced PG synthesis. Theoretical work revealed that, with their decreased stiffness, BLR bacteria could be more efficiently penetrated by NWs, and that this penetrating effect can be controlled by the NW tip diameter and driving force. Experiments confirmed that ~95% of BLR bacteria with stiffness values below 0.5 MPa are penetrated by sharp NWs with a tip diameter of 5 nm compared to ~20% of stiff BLS bacteria with a Young’s modulus of 5 MPa, thus providing an alternative therapeutic strategy for attacking BLR Gram-negative bacteria. A fundamental understanding of BLR Gram-negative bacteria combined with our demonstration of the effectiveness of NWs open a viable route for addressing the emerging threat of BLR bacteria over the long term.


Bacterial strains

The bacterial strains in this study are listed in table S1. A total of 10 Salmonella and E. coli isolates were obtained from the laboratory of B.Y. at Northwest A&F University (Yangling, Shaanxi, China). The clinically collected strains (P. aeruginosa and K. pneumoniae) were obtained from the group of Y.Y. at Zhejiang University (Hangzhou, Zhejiang, China). To detect acquired β-lactam resistance genes, genomic DNA was extracted, fragmented, and tagged for multiplexing with Nextera XT DNA Sample Preparation Kits, followed by whole-genome sequencing on an Illumina HiSeq 2000 platform (Illumina, San Diego, CA, USA). Comparator strains consisted of two reference strains: E. coli (ATCC25922) and NDM-1–producing E. coli (07HAE27). Polymerase chain reaction–based analysis was used to screen for the presence of blaNDM-1 in E. coli isolate 07HAE27.

Bacterial culture conditions

Typically, the bacterial strains were stored at −80°C in glycerol/Difco nutrient (20% v/v) and reactivated by inoculation in 50 ml of sterile Luria-Bertani broth medium at 37°C. After being grown to exponential growth phase [optical density at 600 nm (OD600), 0.5 to 0.6], bacterial cells were harvested by centrifugation (5000 rpm, 5 min) and washed with sterile normal saline three times. Then, the cells were resuspended in sterile normal saline and diluted to predetermined volumes as stock solutions. The bacterial concentration could be monitored photometrically by measuring the OD600. Before performing the bacterial penetration experiments, the OD600 values of bacterial stock solutions were readjusted to 0.1, which corresponded to a concentration of ~108 colony-forming units/ml.

Antimicrobial susceptibility test

All collected strains were tested for their susceptibility to antimicrobial agents. The MICs of the antimicrobial agents were determined by the agar dilution method using Mueller-Hinton agar according to the guidelines recommended by the Clinical and Laboratory Standards Institute (CLSI). The β-lactam antibiotics included penicillin, ampicillin, piperacillin, amoxicillin-clavulanic acid, oxacillin, ceftazidime, cefepime, cefotaxime, ceftriaxone, cefuroxime, cefazolin, imipenem, meropenem, biapenem, and aztreonam. E. coli ATCC25922 and ATCC35218 and Enterococcus faecalis ATCC29212 were used as quality control organisms in MIC determination. The breakpoints for the antibiotics were as per the interpretive standards provided by CLSI. ESBL detection was performed according to CLSI guidelines. A zone of inhibition for cefotaxime/clavulanic acid discs (Oxoid) 5 mm greater than cefotaxime discs (Oxoid) was interpreted as a positive result for ESBLs.

Sample preparations for AFM observation

An important requirement for liquid AFM investigations is that the sample must be immobilized on a surface. For this purpose, an aliquot of the bacterial suspension of ~105 cells per ml was allowed to adhere through electrostatic interaction to a poly-l-lysine–coated glass substratum that was prepared as previously described (37). The excess liquid was then drained after 10 min to allow the bacteria to adhere; fresh, sterile PBS (200 μl) was refilled for the glass, and the sample was used for AFM observation.

Nanomechanical measurements of the bacterial cell envelope with AFM

AFM measurements in liquid were conducted using a Bruker Dimension FastScan AFM, equipped with the NanoScope V controller and a small scanner. For imaging live bacteria, cells were imaged in PBS at room temperature (22° to 24°C) in PeakForce quantitative nanomechanical mapping mode using a silicon nitride tip (ScanAsyst-Fluid, Bruker). The spring constant of the cantilevers was 0.3 N/m, as determined by the thermal tune method for each experiment. The radius of the standard AFM probe was found to be 20 nm. Force-displacement curves were recorded at 0.8 Hz for determination of Young’s modulus. Young’s modulus was calculated by converting the force curves into force-indentation curves and fitting with the Derjaguin-Muller-Toporov model, which describes the indentation of an elastic sample using a stiff conical indenter, as described elsewhere. The half opening angle of the AFM tip was 18°, and the Poisson ratio of the cell was taken to be 0.5, as is typical for soft biological materials. To avoid the influence of the glass substrate on the measured bacterial stiffness, we used experimental conditions in which the cantilever indentation was 10 nm, far less than the thickness of the entire cell (~700 nm). Cantilever spring constants and sensitivity were calibrated before and after each experiment. Data processing was performed using the commercial NanoScope Analysis software (Bruker AXS Corporation).

β-Lactamase assay

β-Lactamase production was triggered for 2 hours (AMP, 100 μg/ml). Cells were then lysed on ice by sonication (2 × 2 min, 70%) and pelleted using a centrifuge. Then, 50 ml of nitrocefin (0.5 mg/ml) was added to 500 ml of the supernatant. Samples were observed after 30 min of incubation in the dark, and a red color indicated β-lactamase activity.

Synthesis of HADA

Preparation of HADA was carried out following the procedure described by Kuru and colleagues (23). The purity was determined using a Waters ACQUITY UPLC system (Waters Corp., Milford, MA, USA) with ACQUITY UPLC HSS T3 column (2.1 mm by 100 mm, 1.8 μm). Elution was performed with a gradient of water/acetonitrile from 95/5 to 5/95 for 10 min and maintained at 5/95 for another 10 min. The flow rate was 200 μl/min. Peaks were detected at 254 nm [high-performance liquid chromatography (HPLC), tR = 2.58 min]. The purification, as analyzed by HPLC, was 93% [1H nuclear magnetic resonance (NMR) (400 MHz, DMSO-d6): δ = 3.41 (s, 1H), 3.48 (s, 1H), 3.69 to 3.77 (d, 2H), 3.82 to 3.96 (d, 2H), 6.87 (s, 1H), 6.89 (s, 1H), 7.82 (s, 1H), 8.76 (s, 1H), 9.12 (t, 1H), and 11.51 (br s, 1H); 13C NMR (250 MHz, DMSO-d6): δ = 39.11, 51.0, 101.4, 110.5, 112.5, 114.1, 131.3, 148.0, 155.9, 160.3, 162.5, 163.9, and 168.6; liquid chromatography–mass spectrometry mass/charge ratio calculated for C13H11O6N2 ([M-H]): 291.08, found 291.08].

PG labeling

We labeled the PG of Salmonella isolates with HADA as previously described (22). Cells were grown to exponential phase and incubated with HADA (final concentration, 0.5 mM) for 3 hours at 37°C. Cells were then fixed in 70% ethanol for 10 min to prevent potential cell stress resulting from the washing steps. Cells were collected by centrifugation (5000 rpm, 5 min) and washed three times with PBS (pH 7.4; 137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, and 1.8 mM KH2PO4) to remove excess dye. Approximately 20 μl of the samples was fixed on glass slides and dried at room temperature in the dark. The stained cells were observed using superresolution confocal microscopy (Leica TCS SP8 STED 3X, Leica Microsystems, Wetzlar, Germany). The samples were excited with a laser at 405 nm, and the emission was detected through a 419- to 465-nm emission filter.

Structure analysis of PG

ATR-FTIR was performed to clarify the structural differences of PG among the BLS and BLR strains. ATR-FTIR spectra were recorded using a spectrometer (VERTEX 70, Bruker) equipped with a KBr beam splitter and a deuterated triglycine sulfate detector. The sampling station was equipped with an overhead ATR accessory, which included transfer optics within the chamber through which infrared radiation was directed to a detachable ATR zinc selenide crystal mounted in a shallow trough for sample containment. Distilled water was used as the background spectra, and 256 scans were taken for each sample in the mid-infrared region of 4000 to 400 cm−1 at a resolution of 4 cm−1.

Cryo-TEM imaging

Salmonella cells were grown to exponential phase and harvested by centrifugation (5000 rpm, 5 min). The resulting pellet was washed with sterile normal saline three times and resuspended in sterile normal saline. All cell suspensions (5 μl) were placed on Quantifoil holey carbon film–coated 200-mesh copper grids (Quantifoil Micro Tools GmbH, Jena, Germany) in the chamber of a Vitrobot (FEI, Hillsboro, OR, USA). All the processes, blotting, and plunge-freezing into liquid ethane were performed by the Vitrobot. Images were acquired on a JEM-3200FSC transmission electron microscope (JEOL, Tokyo, Japan) incorporating a liquid helium stage and an omega-type energy filter operating at 300 kV. The stage was cooled with liquid nitrogen to 80 K during acquisition of all datasets.

TEM analysis of bacteria

The ultrastructure of the bacterial cells was evaluated using TEM. Bacterial suspensions were prepared in the same manner as described above for the cryo-TEM samples placed in a cold fixative solution composed of 2.5% glutaraldehyde in PBS (0.1 M, pH 7.2) at 4°C until use. After rinsing with PBS for 5, 10, 15, 20, and 30 min, the specimens were postfixed in 1% osmium tetroxide (in 0.2 M PBS, pH 7.2) at 4°C for 1 to 2 hours. The samples were then rinsed again with PBS for 5, 10, 15, 20, and 30 min. The samples were dehydrated for 15 min in a series of ethanol solutions (30, 50, 70, 80, 90, and 100%), and infiltrated overnight in a mixture of LR white resin (London Resin Company, Reading, UK) and alcohol (1:1, v/v), followed by infiltration with pure LR white resin twice (for 1 and 2 hours) at room temperature. Pure LR white resin was then used for embedding, and the samples were incubated at 60°C for 48 hours. Sections (50 nm) were obtained with a diamond knife on the Leica EM UC7 Ultramicrotome (Leica, Nussloch, Germany) and picked up on the coated grids. The ultrathin sections were floated on 3% aqueous solution of uranyl acetate for 10 to 15 min and refloated in 4% Pb solution for 8 to 10 min. The prepared bacterial samples were examined with TEM (JEOL, JEM-1011).

Bacterial morphology

Bacterial morphology was investigated by field emission SEM (FE-SEM; Hitachi, S-4800). After centrifugation, the bacteria were first fixed with 2.5% glutaraldehyde for 4 hours at room temperature. The bacteria were then washed with sterile normal saline followed by dehydration with increasing concentrations of ethanol (25, 50, 75, 90, 95, and 100%) for 10 min in each step and air-dried overnight. Before imaging, the bacteria on substrata were sputter-coated with platinum and imaged by FE-SEM.

Computational methods

An FEM environment (COMSOL Multiphysics version 5.2) was used to explore cell membrane penetration, of which the model of inflated spherocylinder shell with isotropic elasticity was carried out to simulate the scenario of a bacterium indented by an NW. In our simulations, the cell envelope was modeled as a single isotropic elastic sheet, where the Young’s modulus (E) encompasses the entire cell envelope. Therefore, the modulus of BLS bacteria is assumed to be 5 and 0.5 MPa for BLR bacteria. The bacterium body was assumed to initially be a spherocylindrical shape with a radius of 0.25 μm and a length of 0.9 μm, which were calculated from the SEM results. The net gravitational force on a cell or particle herein was determined from the difference between gravitational and buoyant forces and may be written asFG=Vcg(pppw)where Vc is the volume of bacteria, g is the gravitational constant, pw indicates the water density, and pp indicates the particle density. Since most microorganisms are small and their density is similar to that of water, the net gravitational force is negligible (38). Therefore, the driving force for bacterial deformation is assumed to be the molecule force. Under a driving force of 100 pN, the bacterium is axisymmetrically indented by a truncated cone NW whose diameter varied from 100 to 5 nm and length remained constant (5 μm). The bacterial turgor pressure can be calculated as follows (39)P=E/100where ∆P is the bacterial turgor pressure and E is the bacterial stiffness.

Fabrication of programmable NWs

In a typical synthesis process of NiCo(OH)2CO3 NWs, 2 mmol of CoCl2·6H2O, 1.25 mmol of NiCl2·6H2O, and 3 mmol of urea were dissolved in 15 ml of water to form a transparent pink solution. Following the addition of a piece of FTO (1.0 cm by 1.0 cm), the solution was transferred to a 25-ml Teflon-lined stainless steel autoclave and kept at 120°C for 6 hours. After hydrothermal growth, the NiCo(OH)2CO3/FTO was carefully washed with DI water and ethanol several times to remove the excess surfactant and dissociative ions and air-dried.

To obtain NWs with different tip diameters, the NiCo(OH)2CO3/FTO substratum was immersed into freshly prepared aqua regia (HCl:HNO3, 3:1 ratio by volume) at room temperature, and the etching time was varied to control the tip diameter. The substrata were then immersed in DI water to rinse their surfaces and dried by natural convection under ambient conditions.

Materials characterization

The morphology of the samples was characterized by FE-SEM (Hitachi, S-4800) and TEM (JEOL, JEM-1011). The HRTEM images were recorded using a JEM-2100F microscope. The crystalline structures of the samples were characterized using a PANalytical x-ray diffraction instrument with Cu Kα radiation and ranging from 5° to 70° at room temperature.

Surface modification with Con A

To bind Con A onto substrata, the substrata of NiCo(OH)2CO3/FTO were first treated with oxygen plasma for 30 min (EXTRON25). The treated substratum surface was first incubated in 4% (v/v) 3-aminopropyltriethocysilane in ethanol at room temperature for 60 min. After incubation, the surface was rinsed with ethanol and dried under argon. The substrata were then immersed into a 10 mM bis(N-succinimidyl) carbonate solution in acetonitrile at room temperature for 10 min. After the solution was removed, the surface was washed with acetonitrile and dried with argon. Last, the substratum was treated with bacteria binding molecule solutions (50 μg/ml) of Con A in PBS (0.02 M, pH 7.2) at room temperature for 60 min. After the solution was removed, the substratum was immersed in 5 ml of PBS buffer and shaken for 10 min on a shaker to remove noncovalently attached Con A. After the PBS buffer was removed, the substratum modified with Con A was obtained and stored at 4°C for later use.

Analysis of bacterial penetration

For bacterial capture and penetration tests, functionalized NiCo(OH)2CO3/FTO substrata with an exposed area of 1 cm2 was immersed in 1 ml of bacteria solution for 60 min for the bacteria to adhere to the substrata surface. After the test bacteria solution was removed, the substrata were washed three times with DI water to completely remove the nonattached bacteria. Then, the substrata were fixed, dehydrated, and dried before SEM observations.

CLSM measurement of bacterial viability was carried out using an Olympus FV1000-IX81 laser confocal microscope. The postwashed substrata were immediately stained with propidium iodide (PI; 5 μg/ml) for 15 min and then counterstained with 4′-6-diamidino-2-phenylindole (DAPI; 5 μg/ml) for 15 min in the dark. DAPI permeates all cells, binding to nucleic acids and fluorescing blue when excited by a 405-nm-wavelength laser. PI only enters cells with membrane damage, which were considered to be penetrated, and binds to nucleic acids with a higher affinity than DAPI. The postwashed substrata were then imaged by using CLSM. On the basis of LIVE/DEAD staining image analysis using ImageJ software (version 1.44), the penetration efficiency (P%) of the functionalized substrata can be quantified by the following equationP%=Fr/FrsFb/Fbswhere Fr and Fb represent the total bacteria fluorescence intensity in red (PI) and blue (DAPI) channels, respectively. Frs and Fbs represent the fluorescence intensity per single bacterium in red and blue channels, respectively. The number (Nc) of bacteria captured on substrata can be calculated using the following equation: NC = Fb/Fbs.

Statistical analysis

In Figs. 1 (G to J) and 2D, a total of 20 bacteria (Salmonella, n = 72, 94, 106, 75, and 84; E. coli, n = 124, 183, 128, 176, and 155; P. aeruginosa, n = 102, 123, 123, 147, 127, and 152; K. pneumoniae, n = 101, 107, 98, 123, 160, and 134; BLS E. coli, n = 130, 129, 148, and 141; NDM-1–producing E. coli, n = 124, 100, 106, and 115; listed from left to right in each graph) were used to generate the dot plot. Each box chart in Fig. 2 (O and P) includes the minimum, lower quartile (lower horizontal line), median (middle horizontal line), mean (solid circle), upper quartile (upper horizontal line), and maximum. The number of captured bacteria and penetration efficiency for substrata were displayed as mean proportions ± SD from three independent experiments. Significance was calculated using a two-way analysis of variance (ANOVA) with a Tukey post hoc test using Minitab software. P < 0.05 was considered statistically significant (*P < 0.05, **P < 0.01, and ***P < 0.001). Results with P > 0.05 were considered as not significant.

Correction (1 September 2020): In Fig. 4C, the CLSM images of S. 44 (d=100 nm) and S. 83 (d=100 nm) were erroneously combined. In fig. S23, the SEM images of S. 12 (d= 50 nm) and S. 44 (d= 100 nm) were erroneously a duplicate of S. 7 (d= 50 nm) and S. 7 (d= 100 nm), respectively. The two figures have been updated in both HTML and PDF versions.


Supplementary material for this article is available at

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: We thank Y. Hao at Instrument Analysis Center of Xi’an Jiaotong University for technical support in confocal imaging. Funding: Jianlong Wang acknowledges the support from the National Natural Science Foundation of China (21675127) and the Shaanxi Provincial Science Fund for Distinguished Young Scholars (2018JC-011). T.W. acknowledges the support from the National Natural Science Foundation of China (21925405, 201874005, and 21635002), the National Key Research and Development Program of China (grant no. 2018YFA0208800), and the Chinese Academy of Sciences (XDA23030106 and YJKYYQ20180044). Author contributions: L.L., Jianlong Wang, and T.W. conceived and designed the project. L.L. carried out the synthesis and performed all experiments. S.C. performed the FEM simulations. Z.X., S.C., and Y.L. assisted in the AFM test. C.L. and Jing Wang assisted in TEM and HRTEM characterization. X.H., Y.Y., T.Y., Z.Z., and B.Y. offered the bacterial strains. X.Z. and H.Y. assisted in the ATR-FTIR test and PG labeling. W.Y. performed the HADA synthesis. F.W. assisted in cryo-TEM characterization. W.X. and V.-P.L. provided technical support and insights. L.L., Jianlong Wang, and T.W. cowrote the manuscript. All authors discussed the results and commented on the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors.

Stay Connected to Science Advances

Navigate This Article