Application of Oleanolic Acid and Its Analogues in Combating Pathogenic Bacteria In Vitro/Vivo by a Two-Pronged Strategy of β-Lactamases and Hemolysins.

The rapid spread of β-lactamase-producing bacteria in clinical practice has increasingly deteriorated the performance of β-lactam antibiotics against such resistant strains. Thus, novel agents or strategies for the war against β-lactamase-producing bacteria, especially hypervirulent resistant bacteria (such as toxin-secreting Staphylococcus aureus) carrying complex β-lactamases, are urgently needed. In this study, we found that the natural compound oleanolic acid (OA) and its analogues (especially corosolic acid (CA)) significantly inhibited the activity of important β-lactamases (NDM-1, KPC-2, and VIM-1) in Enterobacteriaceae and β-lactamases (β-lactamase N1) in S. aureus. The results showed significant synergy with β-lactams against β-lactamase-positive bacteria (fractional inhibitory concentration (FIC) index <0.5). Additionally, OA treatment significantly inhibited the activity of hemolysin from various bacteria. In the mouse infection models, the combined therapy with OA and β-lactams exhibited a significant synergistic effect in the treatment of β-lactamase-producing bacteria, as evidenced by the survival rate of S. aureus- or Escherichia coli-infected mice, which increased from 25.0 to 75.0% or from 44.4 to 61.1% (CA increased to 77.8%), respectively, compared to treatment with individual β-lactams. Although OA treatment alone led to systemic protection against S. aureus-infected mice by directly targeting α-hemolysin (Hla), a relatively better therapeutic effect was observed for the combined therapy. To the best of our knowledge, this study is the first to find effective inhibitors against resistant bacterial infections with a two-pronged strategy by simultaneously targeting resistance enzymes and toxins, which may provide a promising therapeutic strategy for drug-resistant bacterial infections.

Introduction

The emergence of clinical isolates that are resistant to antibiotics has forced us to face the “postantibiotic” era.[1,2] For bacterial infection, β-lactams are the most commonly used antibiotics in clinical practice. However, the emergence and rapid global dissemination of β-lactam-antibiotic-resistant bacteria have increasingly reduced the therapeutic effect of such antibiotics and constituted a global crisis, especially for β-lactam-resistant Staphylococcus aureus and β-lactamase-positive Enterobacteriaceae; this could cause a variety of invasive diseases such as hospital-acquired pneumonia (HAP), prosthetic valve endocarditis (PVE), and chronic kidney disease (CKD), especially in patients in the intensive care unit (ICU) using ventilators for prolonged periods or subjected to blood transfusion.[3,4] Therefore, it is urgently necessary to develop novel antibiotics or propose anti-infective strategies to cope with this serious clinical threat.

The production of β-lactamase is one of the most important strategies employed by drug-resistant bacteria against survival pressure induced by β-lactam antibiotics in clinical practice and livestock. In Gram-negative organisms, Enterobacteriaceae (such as Escherichia coli and Klebsiella pneumoniae) mainly produce extended-spectrum β-lactamases (ESBLs), AmpC β-lactamases, and carbapenemases to hydrolyze β-lactams.[5−7] ESBLs, such as the TEM, CTX-M, and SHV enzyme families, can confer bacterial resistance to penicillins and most cephalosporins. AmpC β-lactamases are mainly associated with chromosomes that develop resistance following exposure to β-lactam antibiotics. Carbapenemases (mainly including NDMs, KPCs, and VIMs) have a high relevance ratio in Gram-negative pathogenic bacteria, especially in E. coli and K. pneumoniae, which further accelerated the severe global spread of this protein enzyme. Among the carbapenemases,[8] NDMs are the most widespread metallo-β-lactamases resistant to all clinically available β-lactam antibiotics, including carbapenem antibiotics.[9] KPCs are the most commonly encountered carbapenemases among Enterobacteriaceae isolates.[10] In Gram-positive organisms, such as methicillin-resistant S. aureus (MRSA), the production of β-lactamase is one of the main mechanisms for resistance to β-lactams but not carbapenem antibiotics.[4]

The successful establishment of infection mediated by virulence factors, which lyse host tissue cells for bacterial nutrition acquisition, colonization, and multiplication, is an indispensable premise for the production of enzymes that confer resistance to bacteria.[11]S. aureus, a common opportunistic bacterium in the clinic causing severe lethal infection,[12,13] is such a pathogen that produces hemolysin and β-lactamase, which contribute pathogenicity and resistance, respectively. A previous study has shown that α-hemolysin (Hla) plays an essential role in S. aureus infection, as strains lacking Hla are avirulent in a mouse infection model. Subsequently, targeting Hla with an inhibitor or vaccine provided systemic protection against S. aureus infection.[14] In addition, cholesterol-dependent toxins, another pore-forming toxin family, are crucial for the virulence of the associated bacteria.[15]

Therefore, the combination of antibiotics and inhibitors against resistance enzymes and essential virulence factors in targeted bacteria is a feasible and new strategy to fight infections by various pathogenic bacteria, especially polyinfection by resistant bacteria.[16,17] Oleanolic acid (OA), a pentacyclic triterpenoid compound widely found in medicinal herbs, the plant kingdom, and food products,[18,19] has been used as a dietary supplement and over-the-counter drug for the treatment of hepatitis for a long time. Here, we discovered OA as an effective inhibitor against both β-lactamase (mainly including carbapenemases) and bacterial hemolysin. To the best of our knowledge, this is the first study using combined therapy with antibiotics, drug resistance enzyme inhibitors, and virulence factor inhibitors against complex β-lactam-resistant pathogenic bacterial infections.

Results

Identification of OA as a β-Lactamase Inhibitor

Following a culture with or without OA, the β-lactamase activities in bacterial culture supernatants were determined using an enzyme inhibition assay. As shown in Figure A–D, OA treatment exerted a significant inhibitory effect against β-lactamase activities in different clinically isolated strains carrying various types of β-lactamases. Consistent with these results, the β-lactamase activities in culture supernatants preincubated with OA were remarkably decreased (Figure A–D). These results indicated that OA is an effective inhibitor against β-lactamase. For the laboratory-constructed strains, E. coli BL21 carrying carbapenemases (NDM-1, KPC-2, and VIM-1), β-lactamase N1 in S. aureus, ESBLs (TEM-1 and OXA-1), or AmpC β-lactamase, OA co-culture or co-incubation treatment showed the strongest inhibitory effect against the activities of β-lactamase N1 (Figure E), NDM-1 (Figure F), KPC-2 (Figure G), and VIM-1 (Figure H) compared with ESBL-carrying laboratory-constructed strains (TEM-1 for Figure I and OXA-1 for Figure J). However, such an inhibitory effect was not observed for E. coli BL21 carrying AmpC β-lactamase (Figure K) or E. coli BL21 without β-lactamase (Figure L) following the co-culture or co-incubation treatment with OA, which suggested that the OA-mediated inhibitory effect seemed to be specific for main carbapenemases (such as NDM-1 and KPC-2) and β-lactamases in S. aureus. Taken together, our results established that OA was an effective inhibitor for carbapenemases (NDM-1, KPC-2, and VIM-1) and β-lactamases in S. aureus but had no activity against ESBLs (TEM-1 and OXA-1) or AmpC β-lactamase.

Figure 1

OA inhibited the activities of the β-lactamases in bacterial culture supernatants. A significant inhibitory effect was detected in the carbapenemase-positive isolates E. coli ZJ487 (NDM-1/MCR-1) (A), K. pneumoniae QD-KP2 (NDM-1), and (B) E. coli D3 (NDM-1/OXA-1); (C) β-lactamase-positive strain S. aureus USA300 (D); β-lactamase-positive laboratory strain E. coli BL21 (pET28a-β-lactamase N1) (E); and carbapenemase-positive laboratory strains E. coli BL21 (pET28a-SP-NDM-1) (F), E. coli BL21 (pET28a-KPC-2), and (G) E. coli BL21 (pET28a-VIM-1) (H) for both co-culture analysis and co-incubation analysis. For the extended-spectrum β-lactamase laboratory strains E. coli BL21 (pET28a-TEM-1) and (I) E. coli BL21 (pET28a-OXA-1) (J), a significant difference was observed only in the co-culture analysis. No significant inhibitory effect was found in the AmpC β-lactamases-positive laboratory strain E. coli BL21 (pET21a) (K) or β-lactamases-negative laboratory strain E. coli BL21 (pET28a) (L). ** Indicates P < 0.01; * indicates P < 0.05.

OA Restored the Antibacterial Activity of Different β-Lactam Antibiotics

The inhibition of β-lactamase activities by OA suggested that OA likely has a potential synergistic effect with β-lactam antibiotics. Consequently, the broth microdilution minimum inhibitory concentration (MIC) assay and time-killing assay were used to evaluate this hypothesis. As expected, the checkerboard broth microdilution MIC results of the representative strains (E. coli and S. aureus) showed that OA, at the concentrations of OA ≥32 μg/mL, led to the highest MIC fold change ≥8 (Figure A–C) for β-lactam antibiotics. The fractional inhibitory concentration (FIC) index values of this combination were all less than 0.5, suggesting that OA had an effective synergistic effect with β-lactam antibiotics. Specifically, OA in combination with β-lactam antibiotics showed an MIC fold change of 4–64 with FIC index values less than 0.33 ± 0.07 against all of the β-lactamase-positive S. aureus strains (including MRSA) (Table S1). For E. coli and K. pneumoniae strains carrying one or more β-lactamases, the combined therapy with OA and β-lactam antibiotics resulted in an MIC fold change of ≥4 with FIC index values less than 0.33 ± 0.07 (Table ). In agreement with the relatively lower inhibition of ESBL activities by OA (Figure I,J), OA combined with β-lactam antibiotics had no synergistic effect for the ESBL-positive laboratory-constructed strains E. coli BL21 (pET28a-TEM-1) and E. coli BL21 (pET28a-OXA-1) (Table ). Additionally, this synergistic effect was not observed in the strains without β-lactamases (S. aureus strain American Type Culture Collection (ATCC) 25923 and E. coli BL21 (pET28a)) or the β-lactamase-producing strains (S. aureus USA300 and E. coli ZJ487) treated with non-β-lactamases, such as erythromycin, tetracycline, chloramphenicol, streptomycin sulfate, kanamycin sulfate, and colistin (Tables S1 and 2). These results indicated that a synergistic effect was especially observed in clinical strains producing β-lactamases following the combined therapy with OA and β-lactam antibiotics.

Figure 2

OA restored the susceptibility of different β-lactamase-positive strains to classical β-lactam antibiotics without influencing bacterial growth. Microdilution checkerboard analysis showing a synergistic effect of OA and β-lactam antibiotics against the carbapenemase-positive laboratory strain E. coli BL21 (pET28a-SP-NDM-1) (A), carbapenemase-positive isolate E. coli ZJ487 (NDM-1/MCR-1) (B), and β-lactamase-positive strain S. aureus USA300 (C). The possibility of bacterial growth was represented by the color depth of the heat plots and a completely specified MIC fold change of ≥8 at the concentrations of OA ≥32 μg/mL. Growth curves for E. coli BL21 (pET28a-SP-NDM-1) (D), E. coli ZJ487 (NDM-1/MCR-1) (E), and S. aureus USA300 (F) cultured in the presence of various concentrations of OA (from 0 to 128 μg/mL). Time-killing assays of different combinations of β-lactam antibiotics and OA or a control treatment (only medium) against E. coli BL21 (pET28a-SP-NDM-1) (G), E. coli ZJ487 (H), and β-lactamase-positive S. aureus USA300 (I). Values represent the averages of three independent experiments. (J) Microscopic observation of the morphology of E. coli ZJ487 (NDM-1/MCR-1) under different treatments (0.5× MIC of 16 μg/mL meropenem (Mep), 32 μg/mL OA without antibacterial activity, 0.5× MIC of Mep combined with 32 μg/mL OA, no treatment as a negative control, and 2× MIC of Mep (64 μg/mL) as a positive control). Evident morphological changes were observed in the samples treated with 2× MIC of Mep (64 μg/mL) or a combined therapy with 0.5× MIC of Mep (16 μg/mL) and 32 μg/mL OA.

Table 2

Binding Free Energy (kcal/mol) of WT-OA, D199A-OA, T201A-OA, and F240A-OA Systems Based on the Computational Method and the Values of the Binding Constants (KA) Based on Fluorescence Spectroscopy Quenching

	WT-NDM-1	D199A	T201A	F240A
computational method (kcal/mol)	–21.09 ± 2.70	–15.39 ± 2.70	–17.19 ± 2.68	–16.61 ± 2.34
KA (1 × 104) (L/mol)	10.50 ± 2.10	6.97 ± 1.12	7.80 ± 1.82	6.98 ± 1.73

Furthermore, OA, as an agent at concentrations no more than 128 μg/mL, had no visible influence on the growth of E. coli BL21 (pET28a-SP-NDM-1) (Figure D), E. coli ZJ487 (Figure E), or S. aureus USA300 (Figure F). However, the combination of OA (32 μg/mL) with Mep had an efficient bactericidal effect against NDM-1-positive E. coli, as evident by a complete elimination by 6 h postinoculation for E. coli BL21 (pET28a-SP-NDM-1) (Figure G) and by 3 h postinoculation for E. coli ZJ487 (Figure H). In addition, OA combined with penicillin G can significantly eliminate, but not thoroughly kill, S. aureus USA300 in the culture medium (Figure I). Furthermore, this synergistic bactericidal effect was examined for E. coli ZJ487 under microscopy. E. coli ZJ487 cells were rounded from short baculiform for the samples treated with OA (32 μg/mL) or Mep (1× MIC of 16 μg/mL), which was similar to cells without any treatment (Figure J). However, a significant morphological change was observed for the samples treated with Mep (2× MIC of 32 μg/mL) or a combined therapy with OA and Mep. Together, our results indicated that OA specifically restored the antibacterial activity of β-lactam antibiotics by targeting crucial β-lactamases.

Inhibition of β-Lactamase Activities by OA and Its Analogues

To further determine the specific inhibitory effect of OA and its analogues against β-lactamases, an enzyme inhibition assay was employed using purified β-lactamases. As expected, OA had a significant inhibitory effect on carbapenemases, such as NDM-1, KPC-2, and VIM-1, and β-lactamases in S. aureus, such as β-lactamase N1 (Figure A–D). The half-maximal inhibitory concentration (IC50) of OA for such inhibition ranged from 6.71 to 13.23 μg/mL (Figure A–D). However, the IC50 values for ESBLs, such as TEM-1 and OXA-1, were higher than 50 μg/mL (Figure E,F). Most of the OA analogues, especially corosolic acid (CA) and ursolic acid (an isomer of oleanolic acid), also showed robust inhibitory effects against NDM-1 instead of TEM-1 (Figure G,H).

Figure 3

OA and its analogues inhibited the activities of purified β-lactamases. Following incubation with the indicated concentrations of OA, the activities of NDM-1 (A), KPC-2 (B), VIM-1 (C), β-lactamase N1 (D), TEM-1 (E), and OXA-1 (F) were detected by enzyme inhibition assays. A significant inhibitory effect was observed for OA-treated β-lactamase activities in a concentration-dependent manner. All of the IC50 values are labeled in the upper-right corner. Additionally, the inhibitory effect of OA and its analogues against NDM-1 (G) and TEM-1 (H) was detected. ** Indicates P < 0.01; * indicates P < 0.05.

In addition, similar to the results of the enzyme inhibition assay, CA and ursolic acid showed significant synergistic effects with Mep for the NDM-1-producing strain E. coli ZJ487 (FIC index = 0.17 ± 0.04) (Table ). Together, our results indicated that OA and its analogues increased the antibacterial activity of β-lactam antibiotics against Enterobacteriaceae carrying influential carbapenemases and β-lactamase-positive S. aureus by specifically targeting β-lactamases.

Table 1

MIC Values of Analogues of OA and Meropenem Combination Therapy on NDM-1-Positive E. coli ZJ487a

		MIC of analogues of OA (μg/mL)		MIC of Mep (μg/mL)
species		alone	combination	alone	combination	FIC Index
E. coli ZJ487 (NDM-1/MCR-1)	corosolic acid	512(≥512)	32	32(16–32)	2(2–4)	0.17 ± 0.04
	ursolic acid	512(≥512)	32	32(16–32)	2(2–4)	0.17 ± 0.04
	maslinic acid	512(≥512)	32	32(16–32)	4(2–4)	0.38 ± 0.00
	glycyrrhizic acid	512(≥512)	32	32(16–32)	8(4–16)	0.40 ± 0.14
	α-boswellic acid	512(≥512)	32	16(16–32)	16(16)	0.73 ± 0.29
	arjunolic acid	512(≥512)	32	32(16–32)	8(4–16)	0.40 ± 0.14

All of the data of MIC values were the median (range for the data).

Identification of the Binding Mode of OA with NDM-1

As shown in Figure A, the expression of NDM-1 in the NDM-1-positive strains E. coli BL21(DE3) (pET28a-SP-NDM-1) or E. coli ZJ487 was not visibly affected by OA treatment for 4 h based on a western blot assay. This noninfluential phenomenon was also validated in culture supernatants of E. coli BL21(DE3) (pET28a-SP-NDM-1) co-cultured with OA for both 4 and 6 h (Figure B). These results suggested that OA inhibited β-lactamase activity without affecting bacterial growth or β-lactamase expression.

Figure 4

Direct engagement of OA with the residues Asp199, Thr201, and Phe240 in NDM-1 inhibited the activity of this enzyme. The NDM-1-positive bacterial strains E. coli BL21 (pET28a-SP-NDM-1) or E. coli ZJ487 were cultured with various concentrations of OA for the indicated time, and the production of NDM-1 in bacteria (A) or culture supernatants (B) was determined by western blot assays. No visible influence of NDM-1 expression or secretion was observed in bacteria treated with OA. (C) Three-dimensional (3D) structure determination of NDM-1 with an OA complex by molecular modeling analysis. (D) Total binding energy on a per-residue basis in the binding sites of the NDM-1–OA complex. (E) Interaction between OA and the residues of the binding sites in NDM-1 is shown using a two-dimensional (2D) diagram by the LigPlus software. (F) Activities of NDM-1 and its mutants (NDM-1-Asp199Ala, NDM-1-Thr201Ala, and NDM-1-Phe240Ala) in the presence of various concentrations of OA. The sensitivity of OA for the NDM-1 mutants was much lower than that for NDM-1, as no significance was observed for the activities of NDM-1 mutants by OA at concentrations no greater than 32 μg/mL. ** Indicates P < 0.01.

Through the computational biology method, the potential binding mode of OA in the active site of NDM-1 was explored in this study. The binding modes of OA and NDM-1 are shown in Figure C. OA can bind to NDM-1 via hydrophobic interactions. During the time course of the simulation, OA could localize to the catalytic pocket of NDM-1. Specifically, the binding model of OA with NDM-1 revealed that the side chains of Thr168, Lys181, Gly197, Ile198, Asp199, Thr201, Ile203, Phe240, Lys242, and Ala243 could form strong interactions with OA (ΔEtotal of ≤0 kcal/mol).

To explore the energy contributions from the residues of the binding sites in the NDM-1–OA complex, the energy decomposition was calculated for the NDM-1 and OA complex systems. The residues of Asp199, Thr201, and Phe240 had a strong total binding energy contribution, with a ΔEtotal of ≤−1.5 kcal/mol (Figure D). These results suggested that these three residues are key residues for OA binding, and the interaction between residues in the binding sites and OA is shown in Figure E.

To confirm these theoretical results, the total binding free energy for the NDM-1–OA complex and their detailed energy contributions were calculated according to the molecular mechanics/Poisson–Boltzmann surface area (MM/PBSA) approach (Table ). According to the calculation results, the binding free energy, ΔGbind, of the interaction between OA and proteins decreases in the following order: WT-NDM-1 > mutants, which means that WT-NDM-1 has the strongest ability to bind with OA. By fluorescence spectroscopy quenching, we measured the ΔGbind and the number of binding sites between OA and the three mutants, and these results were highly consistent with those obtained by computational methods (Table ). Additionally, a significant inhibitory effect was not detected in the samples with NDM-1 mutants NDM-1-Asp199Ala, NDM-1-Thr201Ala, and NDM-1-Phe240Ala treated with OA by enzyme inhibition assays (Figure F). These results indicated that the information generated by the molecular dynamics (MD) simulation on the NDM-1–OA complex is reliable.

OA Combined with β-Lactam Antibiotics Synergistically Protects Mice from β-Lactamase-Producing Bacterial Infection

The synergistic effect of OA in combination with β-lactam antibiotics was further evaluated using mouse infection models with S. aureus or E. coli. For S. aureus pneumonia in mice, the survival rate of infected mice significantly increased from 6.3% (1/16 for the control group) and 25.0% (4/16 for penicillin G therapy) to 75.0% (12/16 for combination therapy) with OA treatment, suggesting that the synergistic effect of OA and β-lactam antibiotics also occurs in vivo (Figure A). Interestingly, treatment with OA also exhibited a certain therapeutic effect, as evidenced by a 43.8% (7/16) survival rate vs 6.3% (1/16) for the control group.

Figure 5

Combined therapy of OA and β-lactam antibiotics had a significant synergistic effect against β-lactamase bacterial pathogenicity in vivo. (A–H) Each mouse was nasally infected with 3.5 × 108 colony-forming units (CFU) of S. aureus USA300 for survival analysis or with 1.5 × 108 CFU of S. aureus USA300 for other analyses. Then, the mice were treated with the indicated therapy. The survival of infected mice was observed for 96 h (A). At 36 h postinfection, the lungs and bronchoalveolar lavage fluid from sacrificed mice were collected for pathological analysis and inflammatory response analysis, respectively. (B) Bacterial burden in the lungs was calculated by plating. The lung wet/dry weight ratio (C) and the activity of β-lactamase in the bronchoalveolar lavage fluid (D) of all mice following the indicated treatment. The gross pathological changes and histopathology of the lung tissue of mice are shown in (E). The production of the inflammatory mediators interleukin (IL)-1β (F) and IL-6 (G) and tumor necrosis factor (TNF)-α (H) in bronchoalveolar lavage fluid was detected using enzyme-linked immunosorbent assay (ELISA). (I–K) Each mouse was intraperitoneally infected with 2 × 108 CFU of E. coli ZJ478 for survival analysis or 1 × 108 CFU of E. coli ZJ478 for bacterial burden analysis. Then, the mice were treated with the indicated therapy. The survival of infected mice was observed for 120 h (I). At 36 h postinfection, the liver (J) and spleen (K) from sacrificed mice were collected for bacterial burden analysis by plating. ** Indicates P < 0.01; * indicates P < 0.05.

Consistent with the results of survival analysis, a significant reduction in the number of bacteria in the lungs was observed for the infected mice following the combination therapy with OA and penicillin G compared with the other groups (Figure B). In addition, the wet/dry weight ratio of the lung was reduced by monotherapy with OA or combined therapy (Figure C), suggesting decreased inflammation in these samples. Additionally, the activity of β-lactamases in the bronchoalveolar lavage fluid was inhibited at 36 h after treatment with OA or penicillin G combined with OA (Figure D). However, such inhibition was not observed in the samples treated with penicillin G (Figure D). Macroscopic and microscopic observations revealed that the lungs of the infected mice treated with monotherapy were kermesinus, as well as exhibited severe pulmonary tissue hyperemia, dropsy, tissue damage, and accumulated inflammatory cells, similar to the control group (Figure E). Conversely, the lung lesions in combination therapy mice were greatly alleviated and displayed the same pink color as the control group (Figure E). Then, we detected several typical inflammation-related factor levels in the bronchoalveolar lavage fluid of infected mice. The combination treatment showed significant decreases in IL-1β, IL-6, and TNF-α levels in the bronchoalveolar lavage fluid at 36 h postinfection (Figure F–H). In line with the results of survival analysis, OA monotherapy also inhibited the bacterial burden, inflammatory responses, and β-lactamases in the bronchoalveolar lavage fluid. The above results indicated that OA or OA combined with penicillin G showed an effective therapeutic effect against MRSA pneumonia in mice.

The mouse intraperitoneal infection model by E. coli demonstrated that the survival rate of infection increased from 44.4% (8/18) to 61.1% (11/18) in the group of OA in combination with Mep when compared to only the Mep treatment group (Figure I). Interestingly, the combined therapy with CA and Mep showed more effective protection against infected mice in vivo, with a survival rate of 77.8% (14/18) (Figure I). For bacterial burden analysis, OA combined with Mep significantly inhibited E. coli colonization in the liver and spleen of infected mice (Figure J,K). However, OA monotherapy had no therapeutic effect against E. coli-infected mice. Taken together, our results revealed that the combination therapy provided potential protection against infection caused by pathogenic bacteria carrying β-lactamases (mainly including carbapenemases in Enterobacteriaceae and β-lactamases in S. aureus).

OA Inhibited Bacterial Hemolysin Activity

The in vivo mouse infection model assays revealed that OA monotherapy protected mice from only S. aureus infection but not E. coli infection, suggesting that other targets by OA may exist for S. aureus. Our previous studies showed that many natural compounds could significantly relieve S. aureus virulence by targeting Hla. Based on these data, we further examined the influence of OA on Hla activity and found that the activity of this toxin was significantly inhibited following preincubation with OA or its analogues in a dose-dependent manner (Figure A,E). Notably, CA and ursolic acid exhibited stronger inhibitory effects than OA (Figure A,E). This inhibitory effect was also confirmed by various cholesterol-dependent cytolysins (CDCs), such as pneumolysin (PLY) from Streptococcus pneumoniae (Figure B), suilysin (SLY) from Streptococcus suis (Figure C), and listeriolysin O (LLO) from Listeria monocytogenes (Figure D). In the co-culture system with Hla and A549 cells or MH-S cells, the addition of OA robustly protected the host cells from hemolysin-mediated cell injury at a relatively lower concentration ranging from 1 to 4 μg/mL (Figure F–I). The number of green fluorescently labeled cells was increased, and the released lactate dehydrogenase (LDH) was decreased as the concentration of OA increased. Taken together, these results indicated that OA is an effective inhibitor against bacterial hemolysin, including Hla and cholesterol-dependent cytolysins.

Figure 6

OA and its analogues inhibited the hemolytic activity of bacterial hemolysins. Hla (A), PLY (B), SLY (C), or LLO (D) was pretreated with various concentrations of OA, and the hemolytic activity of these pore-forming toxins was determined by a hemolysis assay. (E) Activity of Hla pretreated with the analogues of OA was examined using a hemolysis assay. All of the IC50 values are labeled in the upper-right corner. MH-S (F, G) or A549 cells (H, I) were incubated with culture supernatants of S. aureus USA300. Following the addition of various concentrations of OA (0–4 μg/mL) for 6 h, the cytotoxicity of each sample was detected by Live/Dead and LDH release assays. MH-S cells (F) or A549 cells (H) stained with green (live)/red (dead) were observed under a microscope following the indicated treatment. The LDH released into the supernatants was evaluated for MH-S cells (G) or A549 cells (I).

Engagement of OA with the Active Center of Hla Inhibited Hla Activity

In agreement with the action of OA against β-lactamases, the expression of Hla in S. aureus was not affected by OA treatment at concentrations required for hemolysis assays for both 4 and 6 h (Figure A). As shown in Figure B, the oligomerization of Hla was restricted following preincubation with OA, suggesting that a direct engagement of OA to Hla hinders the formation of Hla oligomers and may contribute to the inhibition of Hla activity. In the simulation, OA also bound to the active region of Hla through hydrophobic interactions similar to those of NDM-1 (Figure E). The residues Pro129, Phe179, Ser244, Leu245, Phe250, Ser251, Pro252, Asp253, and Phe254 of Hla were the closest to OA (Figure C,E). We then performed energy decomposition and found that the total contribution of van der Waals forces (ΔEvdw) of Phe250, Asp253, and Phe254 was less than −1.5 kcal/mol, which indicated that these three residues were important for the engagement of OA with Hla (Figure D). Consistent with these theoretical calculation results, the inhibition of Hla activity by OA was lost for the Hla mutants Hla-Phe250Ala, Hla-Asp253Ala, and Hla-Phe254Ala (Figure F). These findings further confirmed that the engagement of OA with the active center of Hla (Phe250, Asp253, and Phe254) hindered Hla oligomerization and, subsequently, inhibited Hla activity. Taken together, our results established that OA inhibited S. aureus virulence by simultaneously targeting β-lactamases and Hla.

Figure 7

Direct binding of OA with the Hla residues Phe250, Asp253, and Phe254 reduced Hla activity by inhibiting the oligomerization of this toxin. (A) Production of Hla in culture supernatants was determined by a western blot assay. (B) Inhibition of Hla oligomerization by OA treatment. (C) 3D structure determination of Hla in complex with OA by the molecular modeling method. (D) Total binding energy on a per-residue basis in the binding sites of the Hla–OA complex. (E) Interaction between OA and the residues of the binding sites in Hla using a 2D diagram by the LigPlus software. (F) Influence of OA on the activity of Hla and its mutants. ** Indicates P < 0.01.

Discussion

Bacterial resistance mediated by resistance enzymes and pathogenicity mediated by virulence factors have always been the two main concerns of scientists.[20] Here, we found that OA simultaneously exhibited a robust inhibitory effect against various β-lactamases, including carbapenemase (such as NDM-1 and KPC-2) and β-lactamase (N1) carried by S. aureus and bacterial hemolysins (Hla and CDC family hemolysins). This inhibition effect was fully reflected in a mouse lung infection model compared with a mouse intraperitoneal infection model. OA itself showed obvious antibacterial activity against β-lactamase-negative S. aureus and weak antibacterial activity against low-β-lactamase-resistant S. aureus (such as S. aureus 8325-4 and S. aureus ATCC 29213, as shown in Table S1), which may be due to high functional similarity between penicillin-binding proteins (PBPs) and β-lactamases.[21] Interestingly, such a synergistic effect was observed for both clinic-isolated and lab strains, suggesting that OA possesses the potential for clinical application to fight the resistance of bacterial infection with antibiotics. Furthermore, an effective inhibition of Hla by OA provided robust protection of infected mice against S. aureus by targeting Hla and β-lactamases when combined with β-lactam antibiotics, suggesting that with a dual strategy, OA could treat S. aureus infection.

However, a relatively lower inhibitory effect on extended-spectrum β-lactamases, such as TEM-1 and OXA-1, or no inhibitory effect on AmpC β-lactamases in Gram-negative bacteria by OA was observed. The fact that β-lactamases negatively affect the hydrolytic activity of carbapenems may contribute to this observation.[22]

For the analysis of the inhibition of the activity of carbapenemases and hemolysins by OA, we found that OA could localize to the catalytic pocket of NDM-1 or Hla, which is critical for binding with the substrates. Due to the binding of OA with NDM-1 or Hla, the binding of substrates with these two proteins was blocked, leading to the loss of the biological activity of NDM-1 or Hla. In addition, both KPC-2 and Hla lack zinc ions, suggesting that OA-mediated inhibition was not caused by chelation of zinc ions.[17]

The inhibitory effect of OA analogues on β-lactamase and hemolysin is highly variable. For example, CA inhibited the hemolysis by SLY significantly better than OA with an IC50 of less than 1 μg/mL (data not shown). The isomer of OA, ursolic acid, also showed excellent biological activity on Hla. Besides, we found that CA inhibited different factors of bacteria and reduced the mortality of mice infected with carbapenem-positive bacteria better than OA. Therefore, modification of the structure of OA or its analogues may lead to a more effective dual target inhibitor that simultaneously targets bacterial β-lactamases and hemolysins. The therapeutic effect of OA combined with β-lactam antibiotics at concentrations of 25, 50, and 100 mg/kg was detected in the pre-experiment. The dosage of 50 mg/kg OA was used in the formal experiment after comprehensive consideration.

All OA and its analogues, such as CA, ursolic acid, and maslinic acid, exist widely in more than 200 types of plants and have various biological effects, including antiviral, anti-inflammatory, antioxidant, antibacterial, and antidiabetic properties.[18,23−25] OA and its analogues all belong to pentacyclic triterpenoids with similar basic chemical structures, which facilitates further analysis of the pharmacodynamic effects based on this large class of compounds. Thus, our results laid the foundation for the development of pentacyclic triterpenoids as agents for fighting infections of resistant bacteria by targeting resistance enzymes and pore-forming toxins.[26]

Conclusions

In summary, the synergy of OA or its analogues with β-lactam antibiotics may represent a new strategy or agents for the treatment of infections mediated by multidrug-resistant bacteria by targeting both hemolysins and single or multiple β-lactamases.

Experimental Section

Bacterial Strains and Chemicals

All bacterial strains are listed in Tables S1 and S2. S. aureus USA300; S. aureus USA400, MRSA 252; S. aureus ATCC 29213; S. aureus ATCC 25904; and S. aureus ATCC 25923 were purchased from American Type Culture Collection (ATCC). Animal- and human-origin clinical isolates, including MRSA, E. coli, and K. pneumoniae, were collected in Shandong and Jilin, China; these isolates carried one or more β-lactamases.[27]S. aureus 8325-4 was obtained from Prof. Timothy J. Foster.[15]E. coli BL21(DE3)(pET28a-SP-NDM-1) carried an NDM-1 gene originating from K. pneumoniae QD-KP2. E. coli BL21(DE3)(pET21a) was used as a type C β-lactamase-positive strain. In addition, E. coli BL21(DE3)(pET28a) and S. aureus ATCC 25923 were used as negative control strains.

OA and its analogues (corosolic acid (CA), ursolic acid, maslinic acid, glycyrrhizic acid, α-boswellic acid, and arjunolic acid) (Figure S1) were purchased from Sigma-Aldrich, St. Louis, MO. All antibiotics were purchased from the National Institute for the Control of Pharmaceutical and Biological Products (Beijing, China).

Expression and Purification of Recombinant β-Lactamases and Hemolysins

The interesting genes were obtained from associated strains using the polymerase chain reaction (PCR) or synthesized and cloned into pET28a to generate expression vectors. The ndm-1 gene was obtained from E. coli ZC-YN3. β-Lactamase N1 (gene name, β-lactamase; GenBank, CP000730.1; protein ID, ABX29221.1) was obtained from S. aureus USA300.[28] The oxa-1 gene was obtained from E. coli D3. The kpc-2 gene (NCBI reference sequence: NG_049253.1), vim-1 gene (NCBI reference sequence: NC_014368.1), and tem-1 gene (NCBI reference sequence: NG_050145.1) were synthesized according to the sequences reported on NCBI. All primers with endonucleases BamHI and XhoI are listed in Table S3. Following transfer into E. coli BL21(DE3), all recombinant vectors were identified by sequencing.

An overnight culture of E. coli BL21(DE3) containing the recombinant vectors was supplied in 1000 mL of Luria–Bertani (LB) medium for culture expansion and grown to the midlogarithmic phase with OD values of 0.6–0.8. Following an overnight induction with 1 mM isopropyl-β-d-thiogalactopyranoside (IPTG) at 15 °C, the E. coli cells were harvested by centrifugation (6000 rpm for 30 min at 4 °C) and resuspended in lysis buffer (pH = 7.4) for sonication. The supernatant was loaded through a Ni column three times prior to centrifugation (6000 rpm for 30 min at 4 °C). After three washes with lysis buffer (pH = 7.4) to remove the unbound protein, the His-tagged protein was eluted by imidazole ranging from 10 to 200 mM in lysis buffer. Following dialysis, the purified protein was stored at −80 °C.

The expression and purification of Hla from S. aureus, listeriolysin O (LLO) from L. monocytogenes, pneumolysin (PLY) from S. pneumonia D39, and suilysin (SLY) from S. suis were performed according to our previous reports.[29−32]

Site-directed mutagenesis of NDM-1 or Hla was performed using a QuikChange site-directed mutagenesis kit (Stratagene) (TransGen Biotech, Beijing, China) based on pET28a-NDM-1 or pET28a-Hla.

Enzyme Inhibition Assay

Nitrocefin was used to accurately detect all β-lactamase activity.[33] In this study, the effect of OA on the activity of different β-lactamases during bacterial growth and in unprocessed culture supernatants was determined. Briefly, bacteria were cultured with different concentrations of OA (co-culture) for 6 h at 37 °C, and the culture supernatants were collected by centrifugation at 12 000 rpm for 10 min at 4 °C. Then, 100 μL of the supernatant was mixed with 75 μL of phosphate-buffered saline (PBS) and 25 μL of nitrocefin and incubated at 37 °C for 15–45 min. The change in absorbance at 492 nm was determined in 96-well plates at room temperature. Moreover, the activities of β-lactamases in a culture supernatant from bacteria without any OA treatment or purified β-lactamases were also determined as described above. Additionally, the effect of the OA analogue on the activity of NDM-1 and TEM-1 was determined with a similar method.

MIC and FIC Index Determination

Minimum inhibitory concentrations (MICs) for all tested strains were determined with the twofold checkerboard microdilution method by following the Clinical and Laboratory Standards Institute (CLSI) guidelines as previously described.[34,35] In brief, 5 × 105 CFU/mL bacteria cells were added to each well (96-well plates) in 200 μL of Luria–Bertani (LB) broth supplemented with various concentrations of OA (each column) and β-lactam antibiotics (each row). Then, the 96-well plates were incubated statically at 37 °C for 16–24 h, and the MIC was detected by visual observation. The selection of antibiotics against different bacteria according to their clinical use was as follows: penicillins and cephalosporins were used with S. aureus, and carbapenems and cephalosporins were used to test Gram-negative bacteria.

The fractional inhibitory concentration (FIC) index values were calculated as follows: FIC index = (MICcompounds used alone/MICcompounds used in combination) + (MICantibiotics used alone/MICantibiotics used in combination). FIC index ≤0.5 was defined as synergy, FIC index >1 was defined as the addition effect, and 0.5 < FIC index ≤ 1 was considered invalid.

Growth Curves

E. coli BL21(DE3) (pET28a-SP-NDM-1), E. coli ZJ487, and S. aureus USA300 were used to determine the influence of OA on the growth of β-lactamase-positive bacteria. Briefly, overnight cultured bacteria cells normalized to OD600 = 0.3 were transferred to 50 mL Erlenmeyer flasks with different concentrations of OA (from 0 to 128 μg/mL). Then, the bacteria in each Erlenmeyer flask were cultured with shaking, and the growth was continuously detected by measuring the OD600 value of each sample every 30 min.

Time-Killing Assays

Bacterial strains (E. coli BL21(DE3) (pET28a-SP-NDM-1), E. coli ZJ487, and S. aureus USA300) were cultured to the exponential phase, diluted to 5 × 105 CFU/mL in LB medium in 96-well plates, and statically cultured continuously at 37 °C with the addition of OA (32 μg/mL), β-lactam antibiotics (8 or 4 μg/mL meropenem (Mep) for E. coli, 8 μg/mL penicillin for S. aureus USA300), β-lactam antibiotics plus OA, or the solvent control. Tenfold serial dilutions of the different samples were coated onto LB agar plates containing antibiotics (kanamycin for E. coli BL21(DE3) (pET28a-SP-NDM-1), colistin for E. coli ZJ487, and methicillin for S. aureus USA300) to detect the number of bacteria in each group at the indicated time points. The number of colonies on the agar plates was recorded after incubation at 37 °C for 24 h.[36]

Microscopic Observation

The NDM-1-positive strain E. coli ZJ487 was cultured overnight to the exponential phase (OD600 = 0.3) and diluted to OD600 = 0.1 in 2 mL of LB broth supplemented with 16 μg/mL Mep, 32 μg/mL OA, or 16 μg/mL Mep in combination with 32 μg/mL OA at 37 °C for 3 h. Samples treated with 64 μg/mL Mep or LB broth were used as positive controls or negative controls, respectively.[37] A 3 μL volume of the sample was dropped onto a coverslip and simply stained with a bacterial staining kit (Solarbio Science & Biotechnology Co., Ltd.). A thin glass piece was covered on the area where bacteria were spotted after drying, and the bacteria were observed and imaged under an Olympus IX71 microscope, with an oil objective of 600×, with image acquisition software (cellSens Dimension).

Molecular Dynamics Simulation for NDM-1–OA and Hla–OA

The initial structures of NDM-1 or Hla were derived from the 3D structure of the X-rays of the protein data bank (PDB codes: 5JQJ and 3ZQ4, respectively).[38] The initial structure of the molecular dynamics (MD) simulation complex with OA and the targeted protein was obtained using the AutoDock 4 package for standard docking procedures.[29,39] The Gromacs 4.5.1 software package was used during the simulation. After the 100 ns simulation, the free energy decomposition was performed using the Amber 10 package and was divided into the van der Waals contribution (ΔEvdw), the electrostatic contribution (ΔEele), and the salvation contribution (ΔEsol) by the molecular mechanics/Poisson–Boltzmann surface area (MM/PBSA) method.[40] The binding constant (KA) of OA to wild-type or mutant NDM-1 was calculated by fluorescence quenching with the formula ΔGbind = RT ln KA.[41]

Mouse Model of Intranasal Lung Infection

Six- to eight-week-old C57BL/6J male mice were purchased from Liaoning Changsheng Biotechnology Co., Ltd. (Shenyang, China). The animal experiments were approved by and conducted strictly following the guidelines of the Animal Care and Use Committee of Jilin University. S. aureus strain USA300 was grown in tryptic soy broth (TSB) overnight, transferred to 100 mL of TSB medium for culture expansion, and grown to an OD600 of 0.5 at 37 °C with shaking. Bacteria were concentrated by centrifugation and resuspended in PBS for intranasal lung infection. All mice were transferred to a jar filled with ether to be anesthetized (loss of consciousness), and then 40 μL of 3.5 × 108 CFU of S. aureus USA300 was dropped into the lung by the left nasal cavity for survival studies. For mortality analysis, each mouse in every group (n = 16) was treated with the control solvent, penicillin G (50 mg/kg), OA (50 mg/kg), or penicillin G (50 mg/kg) in combination with OA (50 mg/kg) by subcutaneous administration every 8 h. The number of alive or dead mice was recorded until day 4 postinfection.

For other analyses, 12 mice per group were inoculated with 40 μL of 1.5 × 108 CFU of S. aureus USA300 (6 mice for bronchoalveolar lavage fluid experiments and 6 mice for bacterial load assays and histopathology) and treated as described above. The bronchoalveolar lavage fluid was collected by intratracheal instillation of 1 mL of sterile PBS and centrifugation at 4 °C with 1000 rpm for 5 min for enzyme inhibition detection as described above. The cytokines in the bronchoalveolar lavage fluid were measured using IL-1β, IL-6, TNF-α, and INF-γ eBioscience mouse ELISA kits (10255 Science Center Dr., San Diego, CA). The typical lesions of the left lungs of the mice were photographed or used to prepare paraffin sections with hematoxylin–eosin staining for histopathological analysis. The remaining left lobes were weighed and dried for wet/dry ratio analysis or weighed, homogenized, and plated to calculate the bacterial load.

Mouse Intraperitoneal Infection Model

Six- to eight-week-old C57BL/6J male mice were intraperitoneally infected with a lethal dose of E. coli ZJ478 (2 × 108 CFU) to cause a systemic infection.[42,43] The infected mice were subcutaneously administered with Mep (5 mg/kg), OA (50 mg/kg), a combination of Mep (5 mg/kg) and OA (50 mg/kg), or the control solvent every 8 h. The number of alive or dead mice was recorded until day 5 postinfection. For bacterial burden analysis, each mouse in every group was intraperitoneally infected with 1 × 108 CFU of E. coli ZJ478 and treated as described above. Then, the liver and spleen from sacrificed mice were weighed, homogenized, and plated to calculate the bacterial load.

Hemolysis Test

The hemolysis test was described by our previous studies.[44] Erythrocytes (rabbit erythrocytes for Hla and sheep erythrocytes for cholesterol-dependent cytolysin hemolysins) were mixed with different concentrations of OA or its analogues (from 0 to 32 μg/mL) in PBS buffer (LPBS buffer containing sodium phosphate, sodium chloride, and bovine serum albumin at pH = 5.5 for LLO) in a final volume of 1 mL and incubated for 30 min at 37 °C.[30] Then, the OD600 of the supernatants of the mixture was detected after centrifugation (5000 rpm, 5 min) to determine the hemolytic activities of each sample.

Western Blot Assay

The NDM-1-producing strains E. coli BL21(DE3) (pET28a-SP-NDM-1) and E. coli ZJ487 were cultured in LB broth supplemented with different concentrations of OA (0, 8, 16, and 32 μg/mL) at 37 °C with shaking for 4–6 h. Following centrifugation at 12 000 rpm for 1 min, the supernatant or precipitate of the bacterial cultures was mixed with 5× protein loading buffer, boiled at 100 °C for 10 min, separated by 12% sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (PAGE) gels, and transferred onto poly(vinylidene fluoride) (PVDF) membranes. The membranes were incubated with an anti-NDM-1 mouse polyclonal antibody (sera from mice were immunized with purified NDM-1) as a primary antibody and HRP-conjugated goat anti-mouse antiserum as a secondary antibody after a block with 5% nonfat milk for 2 h at room temperature. Then, the blots were developed using the Amersham ECL western blot detection reagent (GE Healthcare, U.K.).

The influence of OA on the production of Hla in the supernatants of the hemolysin-producing strain USA300 was detected as described above using an anti-Hla primary polyclonal antibody (Sigma-Aldrich) and goat anti-rabbit antiserum.

Oligomerization Analysis

A total of 50 μL of 1 μg/mL recombinant Hla was incubated with 5 mM deoxycholate in the presence or absence of various concentrations of OA (0–32 μg/mL) at 37 °C for 30 min. Then, the samples were incubated in loading buffer without β-mercaptoethanol at 55 °C for 10 min, and the Hla oligomerization was determined using a western blot assay as previously described.[29]

Live/Dead Detection and Cytotoxicity Analysis

Lung-tissue-related human lung epithelial cells (A549, ATCC) and mouse alveolar macrophages (MH-S, ATCC) were cultured in Dulbecco’s modified Eagle’s medium (DMEM), and approximately 2 × 105 cells per well were inoculated in a 96-well plate in CO2 incubators overnight. S. aureus USA300 was grown in TSB at 37 °C with OA at concentrations of 0–4 μg/mL until the postexponential growth phase was reached at an OD600 value of 2.5. Bacterial culture supernatants were harvested and filtered with a 0.22 μm filter. Next, the cells were incubated with 100 μL of the above culture supernatants of S. aureus USA300 for 6 h at 37 °C. Following centrifugation (1000 rpm, 10 min), the LDH released in the supernatant was examined using a lactate dehydrogenase (LDH) test kit (Roche, Mannheim, Germany). The cells in 96-well plates were treated with a live/dead reagent (Invitrogen, Carlsbad, CA) to qualitatively evaluate cell viability. Green fluorescently labeled cells were viable, and red fluorescently labeled cells were dead.

Statistical Analysis

The data were expressed as the mean ± standard deviation, and MICs were expressed as the median value. Significant differences were determined by Student’s t-test with the software Statistical Program for Social Sciences (SPSS). *, P values ≤0.05; **, P values ≤0.01.