All chemicals and solvents used in this study were purchased from Sigma-Aldrich (St. Louis, MO, USA). The centrifuges (room-temperature, Allegra X-30; low temperature, GS-15R) were acquired from Beckman Coulter, Inc. (Brea, CA, USA). The biological purification columns were purchased from GE Healthcare Life Sciences (Chalfont, UK). The fermentation broths of Cordyceps militaris, Haemophilus influenzae (H. influenzae) KW20, Streptococcus pneumoniae (S. pneumoniae) R6, Escherichia coli (E. coli) MG1655 and Staphylococcus aureus (S. aureus) RN4200 were provided by Hangzhou Vocational and Technical College (Hangzhou, China). An analytical HPLC system (Waters 600) with a 2487 dual wavelength UV detector and analytical software (Waters Corporation, Milford, MA, USA) was used for the detection and analysis of cordycepin. A prep-HPLC system (Dalian Elite Analytical Instruments Co., Ltd., Dalian, China) with dual P280 constant pressure pump, UV200 II variable wavelength UV detector and analytical software was used for the purification of cordycepin. The ¹H and ¹³C NMR spectra were recorded on a Bruker DPX 500 spectrometer (Bruker Corporation-Billerica, MA, USA). The interaction between cordycepin and different antibacterial targets was investigated using the CDOCKER protocol of Discovery Studio (version 3.1; Accelrys, Ltd., Cambridge, UK).

The fermentation broth of Cordyceps militaris with a cordycepin concentration of ~0.2 mg/ml was centrifuged at high speed (2,000 × g) for 15 min. The supernatant was concentrated under vacuum to a concentration of 0.6 mg/ml. Then, 1,800 ml of this sample solution was passed through 300 ml macroporous resin (HPD-100) for the preliminary removal of impurities. The volume flow rate of the loading sample was 2 BV/h, and the sample was eluted with ethanol at volume fraction of 25% at a volume flow of 3 bed volumes/h. The eluted sample was concentrated and freeze-dried to obtain the crude extract samples of cordycepin. The cordycepin contents of the crude samples were determined by HPLC. A solution was prepared by dissolving 400 mg crude sample in 100 ml mobile phase, and ultrasonic vibration was used to ensure that the sample was completely dissolved. This solution was subjected to further purification.

HPLC conditions were as follows: Column, symmetry shield C18 (250×4.6 mm; 5 µm); mobile phase, methanol:water (15:85, v/v); flow rate, 0.8 ml/min; UV detection, 260 nm; and injection quantity, 10 µl. The sample was filtered through a 0.45- µm membrane filter prior to injection. Quantitative analysis of cordycepin was conducted by evaluating the peak area on the basis of a standard curve. Peaks of cordycepin and other compounds in the sample were identified by their retention times and co-injection tests with corresponding standard compounds.

Prep-HPLC conditions were as follows: Column, Sinochrom ODS-BP (250×10 mm); mobile phase, methanol:water (15:85, v/v); flow rate, 1.0 ml/min; UV detection, 260 nm. A standard solution of cordycepin (1 ml) was injected into the prep-HPLC apparatus. The flow rate was 10 ml/min for observing the peak shape and retention time initially. Then, crude cordycepin solution (2 ml) was injected, with adjustment of the flow rate for complete separation and shortening of the sample peak time after observing the spectrum for crude cordycepin solution. The optimal flow rate was then determined and chromatographic fractions containing cordycepin were collected. The solvent was removed by freeze-drying. High purity cordycepin was obtained in crystalline form. In addition, the purified cordycepin was examined using NMR for comparison with the standard. These crystals were used for evaluating the inhibitory activity of cordycepin against various targets in the following experiments.

In the present study, the binding mode of the original ligand (4-(2-amino-1, 3-thiazol-4-yl)pyrimidin-2-amine) and cordycepin were studied using molecular docking. Molecular docking of cordycepin into the three dimensional X-ray structure of NAD⁺-dependent DNA ligase (LigA) and other potential antibacterial targets was carried out using Discovery Studio, using the DS-CDOCKER protocol. The crystal structures, including that of LigA [Protein Data Bank (PDB) code: 3PN1] complex, were retrieved from the Research Collaboratory for Structural Bioinformatics PDB (http://www.rcsb.org/pdb/home/home.do). All bound water molecules and ligands were eliminated from the protein and polar hydrogens were added to the protein. Images showing the binding were prepared using Pymol software version 8.2.0.tar.bz2 (Slashdot Media Inc., San Diego, CA, USA).

Purification of DNA ligases from H. influenzae, S. pneumoniae, E. coli and S. aureus was conducted by the same procedure. In brief, 100 ml lysis buffer (Tiangen Biotech (Beijing) Co., Ltd., Beijing, China) was used to suspend the frozen cell paste (~1×10⁶ cells). The cells were disrupted using a French press (model 1548-294US; Bodum USA Inc., New York, NY, USA) at 18,000 psi at 4°C, and the crude extract was centrifuged at 12,000 × g for 30 min at 4°C. A 20-ml Q-Sepharose HP (HR 16/10) column was pre-equilibrated with Buffer A [50 mM Tris-HCl, pH 7.5, 5 mM dithiothreitol (DTT), 2 mM ethylenediamine tetra-acetic acid (EDTA), 10% glycerol], and then the supernatant was added at a rate of 1.5 ml/min. After washing the column with Buffer A, the protein was eluted with a linear gradient (0 to 1 M NaCl in Buffer A). The ligase-containing eluents were combined, and dialyzed against 2 l Buffer A at 4°C overnight. The dialyzed sample was purified using a 20-ml Heparin Sepharose CL-6B (HR 16/10) column as described above. The fractions were pooled, and proteins were purified by the addition of 0.4 g/ml (NH₄)₂SO₄. The mixture was kept on ice for 1 h. The sample was centrifuged at 5,000 × g for 30 min at 4°C. The sample was dispersed in 8 ml Buffer A, and loaded onto 320-ml Sephacryl S200 (HR 26/60) at a rate of 1 ml/min, which was pre-equilibrated with Buffer B (50 mM Tris/HCl, pH 7.5, 1 mM EDTA, 5 mM DTT, 10% glycerol, 150 mM NaCl). The ligase-containing eluent was pooled and dialyzed against 1 l Storage Buffer (50 mM Tris/HCl, pH 7.5, 1 mM EDTA, 100 mM KCl, 2 mM DTT, 20% glycerol) overnight at 4°C. The protein was characterized by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and analytical liquid chromatography-mass spectrometry. The protein was stored at −80°C.

A fluorescence resonance energy transfer (FRET) assay was used for the analysis of LigA. The method was similar to that previously described (20,21). For the assays for H. influenzae LigA, 96-well black polystyrene flat-bottomed plates were used, and 100 µl reaction mixture (20% glycerol, 30 mM KCl, 30 mM (NH₄)₂SO₄, 10 mM DTT, 1 mM EDTA, 0.002% Brij 35, 50 mM morpholinepropanesulfonic acid, 100 nM bovine serum albumin, 1 µM NAD⁺, 40 nM DNA substrate, 16 mM MgCl₂ and 0.15 nM H. influenzae LigA) was added to the plates. The mixture was reacted at room temperature for 20 min, then 30 µl Quench reagent (8 M urea, 1 M Tris and 20 mM EDTA in water) was added as the reaction terminating agent. Plates were read using a Tecan Ultra plate reader (Tecan Group Ltd., Männedorf, Switzerland), with monitoring of the ratio of fluorescence intensities at two emission wavelengths (the TAMRA acceptor at 595 nm and the FAM donor at 535 nm) upon excitation of the FAM donor at 485 nm. The larger the ratio of 595 to 535 nm emission values, the larger the fraction of ligated DNA in the reaction. The ratio of the emission values at 595 and 535 nm was recorded. Control samples were also tested, which were 0.2% dimethyl sulfoxide (no compound) serving as the 0% inhibition control, and a mixture containing 50 mM EDTA as the 100% inhibition control. The inhibitory activity of the compound [cordycepin or a broad-spectrum adenosine analog (BCX4430; cat. no. HY-18649A; Medchemexpress LLC, Princeton, NJ, USA), used as a control] was measured at 10 different compound concentrations in the reaction system in order to determine the 50% inhibitory concentration (IC₅₀). Assays for the other bacterial LigA isozymes were performed as described above for H. influenzae LigA, with the following changes in order to optimize enzyme performance, as previously described (22,23). S. aureus LigA reaction mixtures contained 0.15 nM enzyme, 72 nM DNA and 94 µM NAD⁺; S. pneumoniae LigA reaction mixtures contained 0.1 nM enzyme, 24 nM DNA and 6.8 µM NAD⁺; E. coli LigA reaction mixtures contained 0.1 nM enzyme, 22 nM DNA and 5.3 µM NAD+, with KCl and (NH₄)₂SO₄ concentrations adjusted to 40 mM and 5 mM, respectively.

Establishing the function between the mean peak area (Y) and the sample concentration (X) by HPLC, the linear regression equation was determined to be as follows: Y = 2.23 × 104X + 5,643 (R2=0.9998). Peak area and mass concentration exhibited a good linear relationship in the range of 0.112–56 µg/ml. The solution of crude cordycepin (0.1 ml) used for separation and preparation was diluted 100-fold for HPLC injection. The purity of the crude cordycepin sample was 12.1%, and the concentration of cordycepin following purification by prep-HPLC was 99.6%. The HPLC chromatograms of the crude sample and purified cordycepin are shown in Fig. 1B and C, respectively.

The purified cordycepin was analyzed by NMR. The NMR results indicated that the purified cordycepin was consistent with the standard: ¹H NMR (500 MHz, d⁶-DMSO): δ 8.34 [singlet (s), 1H], 8.13 (s, 1H), 7.26 (s, 2H), 5.86 [doublet (d), J=2.2 Hz, 1H], 5.64 (d, J=4.0 Hz, 1H), 5.14 [triplet (t), J=5.5 Hz, 1H], 4.57 (s, 1H), 4.34 [double doublet (dd), J=8.4, 6.1 Hz, 1H], 3.74-3.62 [multiplet (m), 1H], 3.56-3.45 (m, 1H), 2.32-2.18 (m, 1H), 1.91 (doublet of doubled of doublets, J=12.9, 6.2, 3.2 Hz, 1H); ¹³C NMR (125 MHz, d⁶-DMSO) δ 157.93, 154.31, 150.74, 140.97, 120.98, 92.69, 82.56, 76.46, 64.52, 35.96.

The CDOCKER protocol, which is used to precisely model molecular docking, was used to analyze the interaction between cordycepin and some common antibacterial targets. The results are shown in Table I.

The interaction energy between cordycepin and LigA was the lowest, at −35.6055 kcal/mol, and thus it was hypothesized that cordycepin may achieve an antibacterial effect by binding to bacterial LigA. Therefore, the activity of cordycepin against different kinds of LigA was further tested and its activity was compared with that of a broad-spectrum adenosine analog (Fig. 1D) (20).

The activity results against LigA enzyme from S. pneumoniae, H. influenzae, E. coli and S. aureus are shown in Table II. The results show that the inhibitory activity of cordycepin against LigA is similar to that of the broad-spectrum adenosine analog. This may be because they have similar adenosine-based chemical structures.

The binding modes of the ligand (4-(2-amino-1,3-thiazol-4-yl)pyrimidin-2-amine) and cordycepin were investigated using molecular docking. A comparison of the bonding of the two compounds with the LigA protein crystal is shown in Fig. 2. The results indicated that both compounds possess the same binding mode with LigA (PDB Code: 3PN1). Clearly, the two residues LEU-80 and SER-81 play a key role in the affinity of the small molecules with the target protein. The surface created by the residues within 4 Å close to the original ligand provide a deep binding pocket, and each of the two molecules can inserted into this cavity. These data indicate that cordycepin acts as a broad-spectrum antimicrobial agent by exerting an inhibitory effect against the LigA enzyme in different bacteria.