Bacterial strains, plasmids, and oligonucleotide primers used in this study are shown in Supplementary Table S1. All strains were grown at 37°C in Luria-Bertani (LB, Oxoid, United Kingdom) broth unless otherwise stated. Mueller-Hinton (MH, Beijing Land Bridge Technology, China) broth dilution was used to determine the minimal inhibitory concentration (MIC) of antibiotics. Antibiotic disk (Hangzhou Microbial Reagent, China) diffusion tests were used for the antibiotic susceptibility assay (cephalexin and cefazolin). Appropriate antibiotic agents were added when preparing the seed liquid of all bacteria possessing a plasmid. Then, the seed liquid was used directly in relevant experiments.

Whole-genome shotgun sequencing was performed using a Roche 454 Genome Sequencer FLX instrument as described previously (Xiong et al., 2011). Multiple amino acid sequence alignments of CzoA with representative Class A β-lactamases and CzoR with its homologs were conducted using Clustal Omega¹. The IscR-binding site was analyzed by the online program MEME² (Bailey and Elkan, 1994). The -35 and -10 sequences were predicted using Softberry BPROM Tool³ (Solovyev and Salamov, 2011).

The iscR-mutant strain iscR-280 and its complemented strain iscR-280C were generated in our previous study (Liu H.L. et al., 2015). To create a czoR-mutant strain, the suicide allelic exchange vector pCM184-Cm was used as previously described (Chen et al., 2015). The upstream and downstream regions of czoR were amplified with the primer pairs M-czoR-up-F/M-czoR-up-R and M-czoR-down-F/M-czoR-down-R, respectively. Subsequently, the upstream and downstream PCR fragments were cloned into the AatII–BsrGI and ApaI–SacI sites of pCM184-Cm, respectively. The resulting czoR allelic exchange vector pCM184-czoR was introduced into the strain S44 via biparental conjugation with the E. coli strain S17-1 (pir) (Simon et al., 1983), and the double crossover czoR mutant was selected using 50 μg/ml chloramphenicol and 25 μg/ml tetracycline. The tetracycline-sensitive and chloramphenicol-resistant mutant was then confirmed by PCR using primers czoR-inner-F/czoR-inner-R (Supplementary Table S1). For czoR complementation, the complete czoR-coding sequence was amplified via PCR and digested with XbaI and EcoRI. The fragment was subcloned into the broad host-range vector pCPP30, generating plasmid pCPP30::czoR. Then, the pCPP30::czoR plasmid was transferred into the ΔczoR strain by biparental conjugation (Simon et al., 1983) to yield the complemented strain ΔczoR-C.

Expression and purification of the recombinant His₆-CzoA, His₆-IscR, and His₆-CzoR proteins were conducted as described previously (Liu H.L. et al., 2015). Complete coding regions of CzoA, IscR, and CzoR were PCR amplified and subcloned into the His-tag expression vectors pET-28a(+) (CzoA and IscR) or pET-32a(+) (CzoR) (Novagen), yielding plasmids pET-28a(+)-CzoA, pET-28a(+)-IscR, and pET-32a(+)-CzoR (Supplementary Table S1). The recombinant plasmids were introduced via transformation into the E. coli strain BL21 (DE3). CzoA, IscR, and CzoR were overexpressed by adding 0.08 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) at an OD₆₀₀ of 0.2 and further culturing strains for 8 h at 28°C. Induced cells were then harvested by centrifugation and lysed in a French Press (JN-02C, JNBIO, China) in lysis buffer [50 mM Tris–HCl (pH 7.5) and 150 mM NaCl]. Soluble supernatant was mixed with 1 ml of nickel–nitrilotriacetic acid–agarose solution (Qiagen) and eluted in 1 ml of elution buffer [200 mM imidazole, 50 mM Tris–HCl (pH 7.5), and 150 mM NaCl]. After dialysis to remove imidazole, purified proteins were stored in 15% glycerol at -80°C. For in vitro enzymatic assay, His₆ tag of recombinant His₆-CzoA were excised by addition of 0.5 U/ml bovine thrombin (Sigma–Aldrich, Buchs, Switzerland) and incubated at 4°C for 8 h, followed by dialysis in buffer [50 mM Tris–HCl (pH7.5) and 150 mM NaCl] (Guan et al., 2017).

Hydrolysis activities of 11 β-lactams (benzylpenicillin, ampicillin, cefalexin, cefazolin, cefuroxime, cefoxitin, ceftazidime, ceftriaxone, cefepime, meropenem, and imipenem) by CzoA without His₆ tag were determined. His₆-excised CzoA (0.01 μmol/l) and various concentrations (20–600 mmol/l) of each β-lactam were added to PBS buffer (pH 7.0), and incubated at 30°C for 30 min. The hydrolysis activities were evaluated through the changes in characteristic absorbance for the 11 β-lactams using a spectrophotometer (DU 800, Beckman, United States) (Lamoureaux et al., 2013). K_m values were determined by the Lineweaver–Burk plot (Bush and Sykes, 1986). Three technical and biological replicates were performed for each reaction.

Minimal inhibitory concentration profiles of the recombinant E. coli DH5α (Miller and Mekalanos, 1988) expressing czoA (pCT-Zori::czoA) and the isogenic strains of C. testosteroni S44 (S44, ΔczoR, ΔczoR-C, ΔiscR, and ΔiscR-C) were determined by the broth dilution method (Clinical and Laboratory Standards Institute, 2014) using 11 β-lactams. For the recombinant E. coli DH5α, two β-lactamase inhibitors, clavulanic acid, and tazobactam were also tested. E. coli DH5α (pCT-Zori) was used as a control. Clavulanic acid and tazobactam were generally fixed at a concentration of 2 and 4 μg/ml, respectively (Girlich et al., 2007; Naas et al., 2007; Lamoureaux et al., 2013; Clinical and Laboratory Standards Institute, 2014). Moreover, growth tendency of the isogenic strains of S44 under stress of 11 β-lactams with concentration below MIC was evaluated through OD₆₀₀ values. Isogenic strains of C. testosteroni S44 in exponential growth phase were inoculated with each of 11 β-lactams in MH medium at 37°C and OD₆₀₀ values were determined using a spectrophotometer after growth for 24 h.

The disk diffusion method was used for an antibiotic susceptibility assay (Jorgensen et al., 1999). Antimicrobial disks impregnated with cefalexin or cefazolin were separately placed onto the inoculated MH agar plates. After being incubated at 37°C for 36 h, an inhibition zones around each antibiotic disk was measured. In addition, a spotting dilution assay was performed to determine the susceptibility of strains to cefalexin and cefazolin. Overnight cultures of strains S44, iscR-280, iscR-280C, ΔczoR, and ΔczoR-C were grown in LB medium. Tenfold gradient dilutions of these strains (OD₆₀₀ = 1.0) were each plated (4 μl) onto solid LB medium containing 50 μg/ml of cefalexin or cefazolin. LB medium without antibiotics was used as a control. Agar plates were incubated at 37°C and photographed daily until colonies formed.

The intergenic region of czoR–czoA was PCR amplified with primers EMSA-CzoR-F and EMSA-CzoR-R (Supplementary Table S1). The primer EMSA-CzoR-F was labeled with the fluorophore 5-carboxyfluorescein (FAM, Tsingke Biological Technology Company, Wuhan, China). To identify exact binding sequences of IscR, a 30 bp FAM-labeled DNA was synthesized (Tsingke Biological Technology Company, Wuhan, China) and directly annealed in vitro. For EMSA, the purified His₆-CzoR and His₆-IscR were each incubated with FAM-labeled DNA in 30 μl of incubation buffer [100 mM HEPES, pH 7.6, 5 mM ethylene diamine tetra acetic acid (EDTA), 50 mM (NH₄)₂SO₄, 5 mM dithiothreitol (DTT), Tween 20, 1% (w/v), and 150 mM KCl] at 28°C for 30 min. After incubation, the mixtures were electrophoresed in an 8% native polyacrylamide gel in 1× Tris/Borate/EDTA (TBE) buffer for 1 h. The gels were then exposed using a phosphorimaging system (Fujifilm FLA-5100, United States).

To identify exact CzoR-binding sites within the intergenic region of czoR–czoA, a DNase I footprinting experiment was performed as described previously (Shi et al., 2017). The binding reaction was carried out in a 30 μl system containing 0 or 0.12 nM of purified CzoR and 100 ng of 5′-FAM labeled DNA fragment. After an incubation at 28°C for 30 min, DNase I (0.8 unit in 20 μl, Promega) was added to the binding mixture and incubated at 37°C for 10 min. Then, the reaction was terminated by adding 10 μl of 50 mM EDTA and an incubation in a water bath at 65°C for 10 min. Digested DNA fragments were purified using a NucleicSpin Gel and PCR Clean-up Kit (Macherey-Nagel, Germany) and analyzed with an Applied Biosystems 3730XL DNA Analyzer (Tsingke Biological Technology Company, Wuhan, China). Results were analyzed with GeneMarkerV1.6536 (Shi et al., 2017).

The czoA promoter (P_czoA) region was amplified by PCR using the primers pLSP-czoA-F and pLSP-czoA-R (Supplementary Table S1). The PCR amplicon was then digested with EcoRI and BamHI and directionally cloned into the lacZ reporter plasmid pLSPkt2lacZ, which was transformed into E. coli S17-1(pir) (Supplementary Table S1). The resulting plasmid pLSP-czoA was introduced into strains S44, iscR-280, iscR-280C, ΔczoR, and ΔczoR-C via biparental conjugation (Simon et al., 1983). All strains were inoculated into LB medium with or without addition of β-lactams. After being incubated at 37°C for 8 h, β-galactosidase activities were measured as previously described (Liu H.L. et al., 2015).

Each strain of S44, iscR-280, iscR-280C, ΔczoR, and ΔczoR-C was each inoculated into LB medium and incubated at 37°C for 8 h. Next, 0 or 25 μg/ml of cefalexin or cefazolin was added to the culture. After 1 h of induction, bacterial cells were harvested for total RNA extraction using TRIzol Reagent (Invitrogen, Grand Island, NY, United States) according to the manufacturer’s instructions (Invitrogen, Grand Island, NY, United States). Real-time RT-PCR was carried out using an Applied Biosystems^® ViiA^TM 7 Real-Time PCR System (Life Technologies, Carlsbad, CA, United States) and primers listed in Supplementary Table S1. Gene expression was normalized by the ΔΔCT method with an iQ5 Real-Time PCR Detection System (Bio-Rad, United States) (Pfaffl, 2001). An ATP-binding subunit encoding gene clpX (CTS44_RS19450) was used as a reference (Caille et al., 2014) and three technical and biological replicates were performed for each reaction. Statistically significant difference between control and treated samples was performed using Student’s t-test with P < 0.01 as borderline and P < 0.01 as statistically significant level. Data are expressed as the average of three experiments.

In this study, a novel putative class A β-lactamase gene, we named czoA here (czo for cefazolin), was found from the draft genome of C. testosteroni S44 (ADVQ00000000.1, Xiong et al., 2011). CzoA consists of 301 deduced amino acids and shows high similarities with several established class A β-lactamases, PenA (AAB53622.1, 47%), NmcA (AOW71300.1, 43%), AST-1 (AAG44836.1, 43%), Sed-1 (WP_063864602.1, 43%), CdiA (CAA54738.1, 43%), BlaA (AIK22395.1, 43%), GIL-1 (WP_063860521.1, 41%), and HugA (AAL57765.1, 40%). A multiple sequences alignment showed that CzoA shares conserved residues (E¹⁶⁶ and R²²⁰) and motifs (S⁷⁰XXK⁷³, S¹³⁰DN, and K²³⁴TG) with other class A β-lactamases (Supplementary Figure S1). No obvious differences were detected among MICs of the wild-type strain S44 (Supplementary Table S2) and ΔczoR or ΔiscR (data not shown), possibly because there are two other putative β-lactamases of class B (WP_003070075.1) and class D (WP_034361410.1) in S44. Class B and Class D β-lactamases showed overlapping substrate profiles compared with Class A β-lactamase (Bush and Jacoby, 2010). Therefore, knock-out and complementation experiments of czoA were not performed.

A LysR family transcriptional regulator encoding gene, here named czoR, located immediately upstream of czoA was also identified (Figure 1). CzoR displays the highest amino acid identity (59%) with AmpR of Pseudomonas aeruginosa using BlastP analysis (Caille et al., 2014). Sequence alignments of CzoR with AmpR (ADB64523.1, 59%), PenR (AAB53621.1, 57%), SedR (AAK63224.1, 56%), CdiR (CAA54736.1, 55%), HugR (AAL57764.1, 49%), and NmcR (AOW71475.1, 45%) were performed (Supplementary Figure S2). No putative CzoR-binding motifs were predicted using online program MEME, however, we found a putative CzoR-binding box in the promoter region of czoA using EMSA and DNase I footprinting assays, suggesting that CzoR may regulate czoA expression (Figure 1). In addition, to investigate the effect of the Fe–S cluster assembly regulator encoding gene iscR on antibiotic susceptibility, the Fe–S cluster biosynthesis-related genes were also analyzed. The S44 genome contains only one isc system, which is composed of the iscRSUA-hscBA-fdx genes located in contig 61 (Figure 1). IscR-binding motif varies among bacteria originated from different taxa (Novichkov et al., 2013). Based on IscR-binding motifs (5′-WTAMYYRNSNVDWWYRVWMRRBWWH-3′) in C. testosteroni KF-1 obtained from the RegPrecise database (Novichkov et al., 2013), we found a putative IscR-binding site within the czoR–czoA intergenic region using online program MEME (Bailey and Elkan, 1994), suggesting that IscR may also be involved in the regulation of czoR/A.

Considering C. testosteroni S44 contains multiple β-lactamases, we determined MICs of various β-lactam antibiotics against E. coli DH5α (pCT-Zori::czoA) expressing CzoA. MIC tests using a conventional broth dilution method indicated that the recombinant strain produced the β-lactamase and its activity was inhibited by clavulanic acid or tazobactam. Determination of MICs of β-lactams for E. coli DH5α (pCT-Zori::czoA) showed that it was resistant to benzylpenicillin, ampicillin, and some cephalosporins such as cefalexin, cefazolin, cefuroxime, ceftriaxone, cefepime, and imipenem, but it remained susceptible to cefoxitin, ceftazidime, and meropenem (Table 1). The MICs of benzylpenicillin, ampicillin, ceftriaxone, and cefepime were significantly reduced by both tazobactam and clavulanic acid, while these of cefalexin, cefazolin, and cefuroxime were merely reduced by clavulanic acid (Table 1), and the others were not inhibited by tazobactam or clavulanic acid (data not shown in Table 1). It is interesting that MIC of ampicillin–tazobactam was 16-fold of that of ampicillin–clavulanic acid (Table 1), which indicated clavulanic acid is a sound inhibitor for class A β-lactamases, while tazobactam is a good inhibitor for other class of β-lactamase.

Kinetic parameters of the CzoA β-lactamase obtained with the purified enzyme (without His₆ tag) (Supplementary Figure S3) showed that CzoA had strong activities (k_cat values of 99–1085 s^-1) to degrade benzylpenicillin, ampicillin, and cefazolin. Cefalexin, cefuroxime, ceftriaxone, and cefepime were hydrolyzed at low levels (k_cat values of 2–12 s^-1), whereas hydrolysis of cefoxitin, ceftazidime, meropenem, and imipenem was not detectable (Table 2). These results are in agreement with antibiotic susceptibility testing that CzoA did not show obvious hydrolyzing activity for cefoxitin, ceftazidime, and meropenem. However, for imipenem, the MIC difference was observed (1 vs. 0.125 μg/ml), which indicates imipenem may be a substrate of CzoA (Table 1).

Expression of czoA::lacZ was significantly induced by all 11 β-lactams tested (Figure 2A), although CzoA did not show obvious hydrolysis activity of cefoxitin, ceftazidime, meropenem, and imipenem in vitro (Table 2). Substrate induction experiments indicated that CzoA was an inducible class A β-lactamase in presence of β-lactams. Transcription level of czoR was significantly induced by nine β-lactams, but not inducible by cefuroxime and cefoxitin (Figure 2B); however, czoR still showed constitutive transcriptions without and with addition of cefuroxime and cefoxitin (Figure 2B).

To further investigate effects of CzoR and IscR on cephalosporin susceptibility, a mutant strain ΔczoR and its complemented strain ΔczoR-C were constructed. Successful deletion and complementation of czoR were confirmed by diagnostic PCR, as shown in Supplementary Figure S4. An iscR-mutant strain iscR-280 (ΔiscR) and a complemented strain iscR-280C (ΔiscR-C) were obtained from our previous study (Liu H.L. et al., 2015). Both mutant strains ΔczoR and ΔiscR showed significantly inhibited growth compared to S44 under a certain concentration of different β-lactams (Figure 2C). Antibiotic susceptibility phenotype of complemented strains ΔczoR-C and ΔiscR-C was mostly recovered (data not shown).

Real-time RT-PCR assays were performed in S44 with or without addition of cefalexin and cefazolin. Results showed that, similar to czoA and czoR, iscR expression was also significantly induced by cefalexin and cefazolin (Figure 3A), indicating iscR is also involved in cefalexin and cefazolin susceptibility. Kirby–Bauer disk diffusion assays showed that growth inhibition zones of ΔczoR- and iscR-280-mutant strains were significantly larger than S44 in presence of cefalexin and cefazolin (Figure 3B). Antibiotic susceptibility phenotypes of complemented strains were restored (Figure 3B). In addition, spotting assays showed that the iscR-280 strain, and to a greater extent for the ΔczoR strain, was more susceptible to cefalexin and cefazolin relative to the wild-type strain S44 (Figure 3C). Phenotypes of the complemented strains were restored, and all strains showed a similar growth trend on LB plates without antibiotics (Figure 3C). These results suggested that both IscR and CzoR are essential for cephalosporin susceptibility and that CzoR may play a more important role.

To examine interactions between CzoR and PczoA, EMSA was performed with a 259 bp fragment of PczoA and purified CzoR (Supplementary Figure S5). With an increasing CzoR concentration, free DNA substrates gradually disappeared, while intensity of shifted DNA bands increased (Figure 4A). Reactions using heat-denatured CzoR and a non-specific DNA probe did not show any lagging bands. Moreover, unlabeled czoA DNA substrates could competitively inhibit CzoR binding to the FAM-labeled czoA (Figure 4A). These results indicated that CzoR could specifically bind to the czoA regulatory region.

Subsequently, a DNase I footprinting assay was conducted to determine exact binding sites of CzoR. Results showed that the -88 to -72 region in PczoA was obviously protected from DNase I digestion (Figure 4B), indicating that a 17 bp fragment (5′-TCTCAATCAAGATAAAA-3′) upstream of czoA was a CzoR-binding box in S44 (Figure 4C). For further confirmation, interactions between CzoR and a 211 bp fragment of PczoA without CzoR-binding box were tested. EMSA results showed that there was no band shift without CzoR-binding box (Figure 4D). These experiments demonstrated that the LysR family regulator CzoR can regulate the class A β-lactamase gene czoA in S44.

Based on IscR-binding motifs in C. testosteroni KF-1 obtained from the RegPrecise database (Novichkov et al., 2013), we found a putative IscR-binding site within czoR–czoA intergenic region. A putative IscR motif (TTTTCTAATGGATGGTGTCAATTAT) was located in the sense strand adjacent to the CzoR box (Figure 5A). The interaction between IscR and czoR–czoA intergenic region was examined by EMSA. With an increasing IscR concentration, lagging bands were clearly observed. In contrast, negative controls (non-specific DNA probe or heat-denatured IscR) did not show any lagging bands (Figure 5B). In addition, IscR was capable of binding to substrates containing a 30 bp sequence with a refined IscR-binding motif (Figure 5C). These data suggest that IscR can directly bind to the czoR–czoA promoter region, and may regulate the expression of both czoR and czoA.

To investigate how iscR and czoR influence each other and further affect czoA expression, real-time RT-PCR transcription analyses were performed using C. testosteroni S44 isogenic strains with or without addition of cefalexin or cefazolin. Deletion of iscR significantly decreased czoR expression, and czoR was not induced by cefalexin and cefazolin in iscR-280-mutant strain (Figure 6A). czoR deletion did not affect transcription level of iscR (Figure 6B) indicating that czoR induction by cefalexin and cefazolin was depend on IscR expression. Phenotypes of the complemented strain iscR-280C were recovered.

Subsequent efforts focused on czoA expression in the isogenic strains of C. testosteroni S44 using a lacZ reporter gene assay. Results showed that expression of czoA::lacZ was significantly induced by cefalexin and cefazolin in S44, consistent with induction expressions of czoR and iscR in presence of cefalexin and cefazolin (Figure 3A). Deletion of iscR (iscR-280) decreased expression of czoA::lacZ compared to S44 with addition of cefalexin and cefazolin (Figure 6C), possibly because transcription level of czoR was decreased in iscR-280 (Figure 6A). However, in ΔczoR, β-gal activity of CzoA::LacZ was almost non-existent, even at higher concentrations of cefalexin and cefazolin (50–100 μg/ml) or using different induction times (data not shown), indicating that czoR is essential for czoA expression. Although iscR expression was also induced by cefalexin and cefazolin in ΔczoR, IscR could not directly regulate czoA expression without CzoR. Higher β-gal activities in iscR-280C and ΔczoR-C compared to S44 may be attributable to a multicopy-based complementation used for both iscR and czoR. These results suggest that IscR may positively regulate czoR expression of and affect czoA expression, which provided a link between IscR and β-lactam susceptibility regulation.