Characterization of a novel class A carbapenemase PAD-1 from Paramesorhizobium desertii A-3-ET, a strain highly resistant to β-lactam antibiotics

Although clinical antibiotic-resistant bacteria have attracted tremendous attention in the microbiology community, the resistant bacteria that persist in natural environments have been overlooked for a longtime. We previously proposed a new species Paramesorhizobium desertii, isolated from the soil of the Taklimakan Desert in China that is highly resistant to most β-lactam antibiotics. To identify potential β-lactamase(s) in this bacteria, we first confirmed the carbapenemase activity in the freeze–thawed supernatant of a P. desertii A-3-ET culture using the modified Hodge assay. We then identified a novel chromosome-encoded carbapenemase (PAD-1) in strain A-3-ET, using a shotgun proteomic analysis of the supernatant and genomic information. The bioinformatics analysis indicated that PAD-1 is a class A carbapenemase. Subsequent enzyme kinetic assays with purified PAD-1 confirmed its carbapenemase activity, which is similar to that of clinically significant class A carbapenemases, including BKC-1 and KPC-2. Because the location in which A-3-ET was isolated is not affected by human activity, PAD-1 is unlikely to be associated with the selection pressures exerted by modern antibiotics. This study confirmed the diversity of antibiotic-resistant determinants in the environmental resistome.

In our recent study, a new species of a novel genus, designated Paramesorhizobium desertii, was isolated from Taklimakan Desert soil samples, and shown to be highly resistant to most β-lactam antibiotics 9 . Here, we report the identification of a novel chromosome-encoded class A carbapenemase from the type strain of this species, A-3-E T , and the enzyme kinetic parameters of this carbapenemase.

Results
Identification of PAD-1. Figure 1 shows the results of a modified Hodge test for carbapenemase activity. On Mueller-Hinton (MH) plates containing meropenem or imipenem discs, carbapenem-sensitive Escherichia coli ATCC 25922 grew well along the grooves containing the A-3-E T supernatant, whereas it did not grow along the empty groove or the groove containing distilled water. This suggests that the carbapenemase-resistant phenotype of A-3-E T is unlikely to be associated with drug efflux or other mechanisms, but that the carbapenemase is present in the cell culture supernatant.
A subsequent shotgun proteomic analysis identified 20 proteins in the freeze-thawed supernatant (Table 1), among which ATN84_21655 was the only one in the A-3-E T genome to be annotated as a β-lactamase 10 . We designated it PAD-1, according to its species name Pa d ramesorhizobium esertii 9 .  A SWISS-MODEL analysis revealed that the class A carbapenemase KPC-2 in the database has a three-dimensional (3D) structure matching that of PAD-1(GQME value 0.67, QMEAN value −1.61), which is highly significant in carbapenem resistance, especially in clinical Enterobacteriaceae isolates 11 . Interestingly, PAD-1 also shares a similar 3D structure with an artificial class A β-lactamase, GNCA (GQME value 0.77, QMEAN value −0.13), which is the last common ancestor of the class A β-lactamases of the Gram-negative bacteria, predicted in a phylogenetic analysis 12 . Bayesian divergence estimates indicated that the ancestral GNCA gene was present on Earth about 2 billion years ago 12 . Figure 2 shows a phylogenetic tree of PAD-1 and 16 other class A β-lactamases. BKC-1 is the closest relative of PAD-1 (69% amino acid identity). BKC-1 is a plasmid-encoded carbapenemase from a Brazilian clinical K. pneumonia isolate 13 . An amino acid sequence alignment of PAD-1, BKC-1, and KPC-2 ( Fig. 3) shows that PAD-1 contains the four conserved structural elements of the class A serine carbapenemases: the motif 70-SXXK-73, where 70-S is the active serine of carbapenemase; the 130-SDN-133 loop; the single amino acid residue 166-E; and the motif 234-KTG-236 3,14,15 . PAD-1 also contains nearly all the reportedly important residues for class A carbapenemase activity (C69, S70, K73, H105, S130, R164, E166, N170, D179, R220, K234, S237, and C238), and only S237 and C238 are not conserved in PAD-1 3, 16-18 . Thus, the in silico bioinformatics analysis suggested that PAD-1 is a novel class A carbapenemase.
In vitro susceptibility tests. To investigate the in vitro susceptibility of BL21 (DE3) strain carrying plasmid pET28a-bla PAD-1 , we determined its MIC values against various β-lactams together with A-3-E T and BL21 (DE3) carrying pET28a. As shown in Table 2, A-3-E T strain exhibited high-level β-lactam resistance to most tested antibiotics, except for cephradine, cefoxitin and carbapenem. The BL21 (DE3) strain harboring pET28a-bla PAD-1 showed similar resistance spectrums to A-3-E T strain, while BL21 (DE3) harboring pET28a is sensitive to all tested antibiotics. Our results suggested that PAD-1 was responsible for the β-lactam resistance detected in A-3-E T strain. Interestingly, the MIC value of BL21 (DE3) harboring pET28a-bla PAD-1 for meropenem is four times higher than A-3-E T strain (1 μg/ml vs 0.25 μg/ml). If induced with 0.1 mM IPTG, the MIC value of BL21 (DE3) harboring pET28a-bla PAD-1 for meropenem is even higher (8 μg/ml), which implies the expression level of PAD-1 is vital for the resistance profile against carbapenems in the host strain.

Effects of antibiotics on the transcription of bla PAD-1 .
To determine whether the presence of antibiotics will influence the transcription of bla PAD-1 , we measured mRNA level of bla PAD-1 in A-3-E T strain grown in LB medium with or without ampicillin and meropenem by quantitative RT-PCR. As shown in Fig. 4, the expression levels of bla PAD-1 in A-3-E T strain with antibiotic pressure are almost identical to that of in normal LB and MH medium. This immediately suggested that the expression PAD-1 is unlikely induced by antibiotics.
Enzymatic kinetic parameters of PAD-1. The soluble expression and purification of PAD-1 was confirmed with sodium dodecyl sulfate-polyacrylamide gel electrophoreses (SDS-PAGE), and it was then subjected to enzyme kinetic assays. As shown in Table 3, PAD-1 hydrolysed penicillin, cephalosporins, carbapenems, and monobactams in vitro, but not cefoxitin. This is similar to BKC-1, which also does not hydrolyse cefoxitin 13 .
PAD-1 showed high affinity for cephalosporins (low K m values), except cephradine, whose K m value (347.9 μM) was three-fold higher than that of the other cephalosporins, implying that PAD-1 hydrolyses cephradine least  suggested that it has less affinity for these antibiotics to this enzyme. According to the k cat /K m values, cefoperazone (315.48 mM −1 s −1 ) is the best substrate for PAD-1 and meropenem ranks last (62.20 mM −1 s −1 ), although the difference is not significant. Table 3 lists the published kinetic data for BKC-1 and KPC-2. BKC-1 is very efficient in the hydrolysis of oxacillin, with a k cat /K m value 300 times higher than that of PAD-1, whereas the k cat /K m values for other similar β-lactams are all lower than that of PAD-1 13 . Unlike PAD-1 and BKC-1, KPC-2 hydrolyses cefoxitin, and it seems to hydrolyse meropenem and cefalotin more efficiently than PAD-1 11,19 .

Discussion
In response to the lethal selection pressures of antibiotics, pathogenic bacteria have developed various mechanisms to fight back, including impermeable barriers, multidrug-resistant efflux pumps, resistance mutations, the inactivation of the antibiotics, etc. 20 . Under the selection of antibiotics, spontaneous mutations (such as the GyrA83 mutation conferring fluoroquinolone resistance) are fixed in the clinical bacterial populations in a very short time 21 . In contrast, more delicate mechanisms, such as antibiotic-hydrolysing enzymes and efflux mechanisms, do not evolve quickly, but must be acquired from other sources via horizontal gene transfer 22 . However, the reservoirs of antibiotic-resistance determinants have been poorly understood for a long time 23 .
Recently, metagenomic analyses identified diverse genes encoding resistance to β-lactam, tetracycline, and glycopeptide antibiotics in 30,000-year-old Beringian permafrost sediments, which are far more ancient than the very first human antibiotic discovered 24 . Soil microbiomes have been shown to be very important evolutionary origins of ancient antibiotic-resistance genes, and are vast reservoirs of novel antibiotic-resistance genes, which are exchanged with clinical pathogens 25,26 .
However, the origins and diversity of antibiotic-resistance genes in soils throughout the world are still unclear. In previous studies, we isolated several β-lactam-resistant bacterial strains from soil samples from the Taklimakan Desert, the largest desert in the west of China, which is not affected by human activities such as farming or herding. One of these strains, A-3-E T , is extremely resistant to certain β-lactam antibiotics. For example, there   Table 3. Kinetic parameters of PAD-1, BKC-1, and KPC-2 against various β-lactam substrates. ND, no measurable hydrolysis detected; NA, data not available; data for BKC-1 from reference 13 ; data for KPC-2 from references 11, 19 . was no significant difference between its growth curves in medium containing 8 mg/ml ampicillinor in normal medium, and A-3-E T also grew well in medium containing 8 mg/ml carbenicillin, 1 mg/ml cefazolin, or 500 μg/ml cefotaxime. Based on systematic polyphasic taxonomic data, A-3-E T was proposed as the type strain of the novel species Paramesorhizobium desertii 9 . A nitrocefin assay of A-3-E T suggested that β-lactamase contributes to its β-lactam-resistance phenotype 9 .
To investigate the β-lactamase (s) in A-3-E T , we analysed its genome 10 . Among the 4946 annotated genes, we identified 26 genes encoding potential β-lactamases or proteins containing β-lactamase-like domains, using a bioinformatics analysis. When we performed a BLAST search for these proteins against public databases, all the top hits were in silico-annotated β-lactamases, although with no experimental functional support. We randomly selected six genes of the 26 candidates, amplified their intact ORFs, and cloned them into expression vectors. Unfortunately, we failed to detect β-lactamase activity in any of the six proteins.
We then used a novel strategy to identify the potential β-lactamase (s). We first confirmed the existence of active β-lactamase (carbapenemase) in the freeze-thawed supernatant of an A-3-E T culture using a modified Hodge assay with imipenem and meropenem (Fig. 1). We then performed a shotgun proteomic analysis of the supernatant, which immediately identified ATN84_21655 (bla PAD-1 ) as a candidate carbapenemase. In previous studies, to identify novel β-lactamases with low sequence similarity to known β-lactamases, researchers performed tedious time-consuming procedures, including purifying the lactamase from crude extracts, estimating the isoelectric point (pI) with an isoelectric-focusing-nitrocefin assay, fragment cloning and mapping, etc. 27,28 . The strategy developed in this study is straight forward and efficient, and should have great utility in identifying novel β-lactamases in other bacteria.
Of the 12 β-lactam antibiotics tested in the enzyme kinetic assays, cefoxitin was the only one not hydrolysed by PAD-1, which is consistent with the cefoxitin-sensitive phenotype of A-3-E T . In contrast, A-3-E T is sensitive to cephradine (MIC, 4 μg/ml), whereas purified PAD-1 hydrolysed cephradine in vitro with strong enzyme activity (k cat /K m 91.26 mM −1 s −1 ). Interestingly, this was also the case for meropenem, in that the results were positive for a modified Hodge assay of the A-3-E T supernatant on meropenem MH plates, and an enzyme kinetic assay indicated a high k cat /K m value (62.20 mM −1 s −1 ) for meropenem with purified PAD-1. Classical MIC tests defined A-3-E T as carbapenem-sensitive (imipenem 0.38 μg/ml, meropenem 0.25 μg/ml), while it grows well in Luria-Bertani (LB) broth containing 32 μg/ml imipenem or meropenem 9 .
One possible explanation is that the conflict results from MIC assays and LB broth is caused by certain inducible carbapenemases. If induced with 0.1 mM IPTG, the MIC value of BL21 (DE3) harboring pET28a-bla PAD-1 is 8 times higher than that of BL21(DE3) not induced by IPTG (8 μg/ml vs 1 μg/ml, Table 2). As shown in Table 3, the k cat /K m of PAD-1 in the meropenem assay was nearly 30 times higher than that of BKC-1 (62.20 mM −1 s −1 vs 2.25 mM −1 s −1 , respectively), whereas the MIC of meropenem for A-3-E T was much lower than that of a clinical K. pneumonia isolate containing plasmid-borne bla BKC-1 (0.25 μg/ml vs 32 μg/ml) 13 . These results implies that the protein level of PAD-1 do infect the MIC values.
However, our qRT-PCR assays showed that the transcription level of PAD-1 in A-3-E T is not affected by ampicillin (100 μg/ml) or meropenem (32 μg/ml). Figure 4 revealed that neither ampicillin/meropenem nor components in LB broth will induce the expression of PAD-1. Notably, the supernatant used in the modified Hodge assay was concentrated and the enzyme kinetic assays were performed with purified PAD-1. In standard MIC assays and both experiments confirm its activity to hydrolyse carbapenems. In microdilution broth MIC assay, A-3-E T was cultured in MH medium for 24 hours without shaking, while it is able to grow in LB or MH broth containing meropenem with shaking. It is known that lots of factors (inoculum size, type medium, incubation time, etc.) can influence MIC values [29][30][31] . A most likely explanation is that the low level constitutive expression PAD-1 (chromosome-encoded) will lead to low values in MIC assays (MH medium, without shaking), while shaking culture in LB medium might favour the growth of A-3-E T and the accumulation of PAD-1 help it to resist high concentration of meropenem.
PAD-1 is a class A serine carbapenemase with an amino acid sequence similar to those of clinically identified enzymes, such as BKC-1 (66%) and KPC-2 (47%) 11,13 . Unlike KPC-2, neither PAD-1 nor BKC-1 hydrolyses cefoxitin. Although these three proteins share several common motifs and key amino acids essential for class A serine carbapenemases (70-SXXK-73, 130-SDN-133, 234-KTG-236, etc.) 3,14 , they also contain several variant amino acid. It has been reported that residues C69 and C238 of KPC-2 form a disulfide bond 32 , whereas amino acid 238 in PAD-1 and BKC-1 is asparagine (N). This missing disulfide bond may contribute to the differences in the hydrolytic efficiency for certain β-lactams between PAD-1 and KPC-2. In another class A serine carbapenemase, SME-1, the disruption of the C69-C238 disulfide bond causes the loss of hydrolytic activity against imipenem and cefotaxime 33 . The relationship between the amino acid diversity and the enzymatic activities of the carbapenemases warrants further study.
In the genomic analysis, we identified no plasmid in the genome of A-3-E T . The G+C content of the bla PAD-1 gene (63.87%) is similar to that of the A-3-E T strain (60.93%) and there is no identifiable transfer structure near bla PAD-1 10 . Because there are no mobile elements neighbouring bla PAD-1 , it would be difficult for bla PAD-1 to move into other bacteria. Therefore, we assume that bla PAD-1 is inherent to this P. desertii strain rather than acquired from other bacterial species, or that the acquisition event occurred sufficiently long ago for any evidence of chromosomal recombination to have been smoothed away. It is noteworthy that the modelled 3D structure of PAD-1 is similar to that of GNCA (Global Model Quality Estimation, GMQE 0.73). GNCA is a laboratory-resurrected class A β-lactamase based on a comprehensive phylogenetic analysis, which is considered to be the last common ancestor of the class A β-lactamases of various Gram-negative bacteria. The divergence time of GNCA and modern class A β-lactamase is estimated to have been 2 billion years ago (Precambrian). Despite its extensive sequence differences from modern enzymes (∼100 amino acid differences), the catalytic efficiency of GNCA for various antibiotics is similar to theirs 12 .
In conclusion,we have identified a novel chromosome-encoded class A carbapenemase, PAD-1, in P. desertii strain A-3-E T with unusual β-lactam-resistance characteristics. Because this strain was isolated from soil samples collected in the Taklimakan Desert, a natural environment unaffected by human activities, PAD-1 should extend our understanding of the diversity and evolutionary scenarios of environmental carbapenemases.

Materials and Methods
Bacterial strains and plasmids. The A-3-E T strain used in this study was isolated from Taklimakan Desert, China, and proposed as the type strain for P. desertii, the type species of the novel genus Paramesorhizobium 9 . Plasmid pET28a and E. coli BL21(DE3) were used to express the proteins in vitro. Both strain A-3-E T and E. coli were cultured in LB medium (1% tryptone, 0.5% yeast extract, 1% NaCl) at 37 °C. A carbapenem-sensitive E. coli (American Type Culture Collection [ATCC]25922) was used in the modified Hodge assay to visualize the zones of antibiotics inhibition.
Whole-genome sequencing. To detect the antibiotic resistance gene, we sequenced the whole genome of strain A-3-E T with the Illumina HiSeq. 2000 platform using a paired-end strategy. The putative resistance genes were predicted with GeneMarkS and then annotated by searching public databases (KEGG, COG, and NR) with BLAST 34,35 , as has been previously described in detail 10 .
Detection of carbapenemase activity. The A-3-E T strain was cultured in LB medium to stationary phase, and 20 ml bacteria were harvested and resuspended in 1 ml of distilled water. The bacterial culture was then frozen at −70 °C for 1.5 h, thawed at room temperature, and the freeze-thaw procedure was repeated five times to lyse the bacteria. After centrifugation, the supernatant was collected and sterilized with 0.22 μM filters (Millipore Inc., Massachusetts, USA), then a cloverleaf test (modified Hodge test) was used to detect carbapenemase production 36 . Briefly, E. coli ATCC 25922 was plated onto MH agar and allowed to dry for 5 min. Meropenem and imipenem discs (Biomerieux Inc., France) were placed in the centres of the agar plates and three grooves were dug around the discs. The A-3-E T supernatant (200 μl) was added to one groove, 200 μl of distilled water was added to another as the control, and the last groove was left empty. The MH agar plates were incubated at 37 °C for 24 h.

Proteomic analysis.
To identify the carbapenemase in the supernatant, we used a brief proteomic analysis.
The supernatant was mixed with loading buffer in a 100 °C water-bath for 10 min, and separated with 12% (w/v) SDS-PAGE. The gel was stained with Coomassie Brilliant Blue R-250. The visible bands in the gel were digested with trypsin and sent for proteomic analysis. All the digested peptides were analysed with a SYNAPT G2 Mass Spectrometer (Waters Inc., USA). The results were processed with PLGS 2.3 and the resulting peaklists were identified with the annotated genome of strain A-3-E T , as previously described 37 .
Bioinformatics analysis of PAD-1. The amino acid sequence of PAD-1 was compared with the Conserved Domains Database at the National Center for Biotechnology Information 38 . The 3D protein structure of PAD-1 was then predicted at the SWISS-MODEL server 39 . An amino-acid-based phylogenetic tree containing 16 typical class A β-lactamases (KPC-2, BKC-1, BIC-1, TEM-1, CTX-M-2, etc.) was constructed with MEGA 6.0 by using the maximum likelihood algorithm 40 . Expression and purification of carbapenemase PAD-1. To purify PAD-1, we cloned the bla PAD-1 gene into the expression vector pET28a, under the control of the T7 promoter 41 . The intact bla PAD-1 gene was amplified from the DNA of strain A-3-E T with primers 5088_F (5′-CTAGCTAGCATGACGATATCCCTTTC-3′) and 5088_R (5′-CCGGAATTCTTAGACCCGCGAAGC-3′), containing NheI and EcoRI restriction sites (underlined), respectively. The PCR product was purified with the QIAquick PCR Purification Kit (Qiagen Inc., USA). Both the purified PCR product and the expression vector pET28a were digested with restriction endonuclease NheI and EcoRI (New England Biolab Inc.) and ligated with T4 DNA ligase. The recombinant plasmid pET28a-bla PAD-1 was then introduced into E. coli BL21(DE3).
Escherichia coli BL21(DE3) carrying the plasmid pET28a-bla PAD-1 was grown in LB medium containing kanamycin (50 μg/ml) at 37 °C to an optical density at 620 nm (OD 620 ) of 0.6. Isopropyl β-d-thiogalactopyranoside (IPTG; final concentration 0.1 mM) was added and incubated at 20 °C with shaking at 100 rpm for 5 h. The bacterial cells were harvested by centrifugation, resuspended in 10 ml of lysis buffer (300 mM NaCl, 50 mM NaH 2 PO 3 , 10 mM imidazole, pH 8.5), and then disrupted by sonication. The lysate was centrifuged and the supernatant was collected. The protein was isolated from the supernatant with a flow column containing Ni-NTA Agarose (Qiagen Inc., Germany). The purity of the protein was estimated with SDS-PAGE and the concentration of the protein was measured with a commercial BCA assay kit (Thermo Scientific, USA) 42,43 . The purified PAD-1 protein was stored at −20 °C.
In vitro susceptibility tests. The MIC values of A-3-E T strain, E. coli BL21 (DE3) harboring pET28a and pET28a-bla PAD-1 were determined by the broth dilution method using Mueller-Hinton (MH) broth. All the bacterial isolates were grown on the MH agar plates at 37 °C overnight to reach stationary phase and the colonies were resuspended in MH broth to demanded concentration. Then the cell cultures were inoculated into the cell plates containing MH broth with a range of β-lactam concentrations. The MIC values were determined after 24 h incubation at 37 °C.
Quantitative RT-PCR of bla PAD-1 . Quantitative RT-PCR (qRT-PCR) was performed to compare the expression levels of the bla PAD-1 transcript in the A-3-E T strain with and witout antibiotics. The A-3-E T strains were firstly grown in LB medium (normal, ampicillin 100 μg/ml, meropenem 32 μg/ml) and MH medium, and all the RNA samples were extracted with the Pure Link TM RNA Mini Kit (Invitrogen, Carlsbad, CA, USA). The cDNA was synthesized from the RNA samples with the Thermo Scrip RT-PCR System (Invitrogen, Carlsbad, CA, USA). The qRT-PCR reactions were performed in duplicate with 20 ng cDNA template on the LightCycler ® 480 II Real-Time PCR System (Roche, Burgess Hill, UK) using SYBR ® Premix Ex Taq ™ II (Takara, Japan).
Enzyme kinetic assays of PAD-1. The enzyme kinetic parameters (k cat and K m ) of purified PAD-1 were assayed spectrophotometrically in sterile phosphate-buffered saline (50 mM, pH 7.0) at 37 °C. The purified carbapenemase PAD-1 was added to 80 μl solutions of various antibiotics and the initial hydrolysis rates were determined with a SpectraMax M2 Microplate Reader (Molecular Devices., USA), as previously reported 11 . The absorption wavelengths used to measure the kinetic parameters for the different antibiotics were determined by spectrum scanning: oxacillin, 220 nm; cefazolin, 265 nm; cefalotin, 265 nm; cephradine, 265 nm; cefoxitin, 235 nm; ceftazidime, 255 nm; cefoperazone, 266 nm; ceftriaxone, 255 nm; cefotaxime, 259 nm; cefepime, 265 nm; meropenem, 290 nm; and aztreonam, 310 nm. All antibiotics were purchased from the National Institutes for Food and Drug Control, China. The values of the kinetic parameters (k cat and K m ) were estimated with Lineweaver-Burk linearization of the Michaelis-Menten equation 13 .