P. endophytica X85 carrying the novel drug-resistance gene bla_PSZ-1 was isolated from the fecal swab of a rabbit at a livestock farm during a survey on the antimicrobial resistance of bacteria in Wenzhou, southern China. The genomes of the isolates were sequenced and their resistance profiles were determined. The relationship between the resistance genotypes and phenotypes was further analyzed. Species identification of the isolate was performed first by 16S rRNA gene homology and then by genome-wide average nucleotide identity (ANI) analyses. The bacterial strains and plasmids used in the study are listed in Table 1.

Following the Clinical and Laboratory Standards Institute (CLSI) guidelines, MICs were determined on Mueller-Hinton (M-H) agar using the standard agar dilution method, and susceptibility patterns were explained in accordance with CLSI M100 (31st Edition, 2021). The reference strain used for quality control in this study was E. coli ATCC 25922. The antimicrobials tested in this research included 13 β-lactams, 4 aminoglycosides, and 2 β-lactamase inhibitors (Table 2). All antibiotics were for human clinical use and purchased from pharmacies and hospitals, and MIC values were the mean values of three independent measures.

The bla_PSZ-1 gene and its upstream promoter region were amplified using 2× Phanta^® Max Master Mix (Nanjing Vazyme Biotech Co., Ltd., Nanjing, China), and P. endophytica X85 genomic DNA was extracted using the Generay Genomic DNA Miniprep kit (Shanghai Generay Biotech Co., Ltd., Shanghai, China) and used as the template for PCR. The primers with restriction endonuclease adaptors at the 5′-end (XbaI for the forward primer and HindIII for the reverse primer) are listed in Table 3. PCR products were digested with XbaI and HindIII and linked to the cloning vector pUCP24, which was also digested with XbaI and HindIII using a T4 DNA ligase cloning kit (Takara Bio, Inc., Dalian, China). The recombinant plasmid was transformed into E. coli DH5α by the calcium chloride method and the transformants were then screened on Luria-Bertani (LB) agar plates supplemented with gentamicin (40 μg/mL). Individual colonies were inoculated into LB mediums supplemented with the same antibiotic and cultured overnight, and then the inserts were verified by PCR and Sanger sequencing (Shanghai Sunny Biotechnology Co., Ltd., Shanghai, China).

The corresponding primers (Table 3) were used to amplify the ORF of bla_PSZ-1 containing the thrombin digestion site by PCR, which was inserted into the pCold I expression vector, and the recombinant plasmid (pCold I-bla_PSZ-1) was transformed into E. coli BL21 competent cells. For the expression of the PSZ-1 protein, E. coli BL21 cells with verified recombinant plasmids were cultured overnight in LB mediums supplemented with ampicillin (100 μg/mL), diluted 100 times in fresh mediums, and incubated at 37°C. After incubation in an ice bath for more than 30 min, the OD₆₀₀ of the culture reached 0.6 to 0.8 and protein expression was induced by the addition of 1 mM isopropyl-β-dithiogalactopyranoside (IPTG, Sigma Chemicals Co., St. Louis, MO, United States), and further incubated for approximately 20 h at 15–16°C. Cells were collected by centrifugation (5,000 × g, 10 min) at 4°C, resuspended in lysis buffer (20 mM Tris–HCl, 150 mM NaCl, 3 mM β-mercaptoethanol, 0.5% Nonidet-P-40; pH 8.0), and lysed by ultrasound. After the cell fragments were removed by centrifugation (12,000 × g, 30 min) at 4°C, the lysates were combined with pre-balanced nickel-nitrilotriacetic acid (Ni-NTA) agarose resin (Beyotime Biotechnology, Shanghai, China) at 4°C and shaken gently overnight. Then, the recombinant protein was purified by standard Ni-NTA affinity chromatography and digested with thrombin (Solarbio Science & Technology Co., Ltd., Beijing, China) at 37°C for 24 h to remove the His-tag. The purity of PSZ-1 protein was validated by SDS–PAGE with a 12% acrylamide separation gel and Coomassie blue G-250 staining, and the protein concentration was determined by a BCA protein assay kit (Thermo Fisher Scientific, Rockford, IL, United States).

Kinetic parameters of the purified β-lactamase PSZ-1 against β-lactam antimicrobials were determined at 37°C, in a 10 mM phosphate buffer (pH 7.4), and a final reaction volume of 200 μL on a Synergy™ Neo2 Multi-Mode Microplate Reader (BioTek Instruments, Inc., United States). The steady-state kinetic parameters k_cat and K_m were determined by non-linear regression of the initial reaction rates with the Michaelis–Menten equation in Prism (version 8.0.2) software (GraphPad Software, CA, United States).

The concentrations of the β-lactamase inhibitors avibactam and tazobactam, leading to a 50% reduction in the hydrolysis of nitrocefin (IC₅₀), were measured after 5 min of preincubation of the enzymes with the inhibitors at 37°C, and nitrocefin (0.1 mM) was used as the substrate. The IC₅₀ values were determined by non-linear regression analysis (GraphPad Prism, version 8.0.2) using log (inhibitor) vs. response (three parameters). Values are the average of three independent measures.

The total bacterial DNA of P. endophytica X85 was extracted from an individual colony subcultured in LB at 37°C for approximately 16 h by using the Generay Genomic DNA Miniprep kit (Shanghai Generay Biotech Co., Ltd., Shanghai, China). Genomic DNA was sequenced by both the Illumina HiSeq 2,500 and PacBio RS II platforms, with a read length of PE150 and 10 kb, respectively, and a sequencing depth of 150× for both (Shanghai Personal Biotechnology Co., Ltd., China). Unicycler v0.4.8 was used to hybrid assemble the PacBio long reads and confirm the cyclization of the whole-genome assembly (Wick et al., 2017). Pilon improves the quality of genomic assembly sketches by mapping Illumina short reads onto the assembly to correct possible incorrect assembly (Walker et al., 2014). Genes were predicted and annotated by using Prokka v1.14.6 (Seemann, 2014); furthermore, DIAMOND v2.0.11 was used to search the predicted proteins in the NCBI non-redundant protein databases with an e-value threshold of 1e-5 (Buchfink et al., 2021). The 16S rRNA homology analysis was performed by comparing the 16S rRNA sequences extracted from the Prokka-annotated genome of P. endophytica X85 with those in the 16S ribosomal RNA sequence database in NCBI. Drug resistance gene identifier v5.2.02 and CARD were used to annotate drug resistance genes (McArthur et al., 2013). FastANI was used to calculate the ANI (Jain et al., 2018) and dDDH was calculated through the Type Strain Genome Server (TYGS) online database¹ (Meier-Kolthoff et al., 2022). Multiple sequence alignments of PSZ-1 and its relatives from the Beta-Lactamase Database² and UniProt/Swiss-Prot database³ were performed by MAFFT v7.490 (Katoh and Standley, 2013) and then a msa R package was used to visualize the comparison results and embellish the generated visual figure (Bodenhofer et al., 2015). MEGA11 was used to construct a neighbor-joining (N-J) phylogenetic tree including PSZ-1 and other functionally characterized β-lactamases (Kumar et al., 2018). The generated tree was visualized using the online tool iTol⁴ (Letunic and Bork, 2007). The figure depicting the genetic environments around the bla_PSZ-1 and bla_PSZ-1-like genes was generated by clinker v0.0.24 (Gilchrist and Chooi, 2021). GView was used to construct the basic genomic characteristics and comparative genomes of P. endophytica X85 (Stothard et al., 2019). ProtParam⁵ was used to predict the molecular weight and pI value of PSZ-1 (Wilkins et al., 1999).

The complete chromosome, plasmid, and bla_PSZ-1 gene sequences of P. endophytica X85 have been submitted to GenBank, and the accession numbers are CP121108, CP121109, and OQ725878, respectively.

The 16S rRNA gene homology analysis revealed that the 16S rRNA gene of the strain X85 had the highest homology with that of Pantoea endophytica 596 T (NR_178843.1); the coverage was 100.00%, and the identity was 100% (Supplementary Table S3). Furthermore, the ANI analysis between the genomes of P. endophytica X85 and all 863 genomes from Pantoea downloaded from the NCBI databases demonstrated that among them, 14 genomes with ≥95% ANI (the cutoff to define a classified bacterial species) were found (Jain et al., 2018), of which 3 were from the Pantoea endophytica species (Pantoea endophytica 596 T, Pantoea endophytica HN-23, and Pantoea endophytica RIT-836 with ANIs of 98.29, 98.28 and 95.68%, respectively), and the remaining 11 genomes were all species-undetermined ones of genus Pantoea (Supplementary Table S1). The digital DNA–DNA hybridization (dDDH) analysis results obtained by TYGS showed that P. endophytica X85 shared the highest identity (86.00%) with Pantoea endophytica 596 T (GCA_002858935.1), which was higher than the cutoff (70%) to classify a bacterial species (Goris et al., 2007; Supplementary Table S4). Therefore, based on the results mentioned above, the isolate X85 was finally designated P. endophytica X85.

The complete genome of P. endophytica X85 consisted of a chromosome and a plasmid named pPEX85. The size of the chromosome was approximately 4.22 Mb and encoded approximately 4,954 ORFs with an average GC content of 55.06%. The length of the plasmid was 771,939 bp, encoding approximately 720 ORFs, of which 540 (75.0%) were predicted to encode function-known proteins (Table 4).

Of the 17 antimicrobials tested, including 13 β-lactams and 4 aminoglycosides, the wild-type P. endophytica X85 exhibited the highest MIC level for amoxicillin (> 2048 μg/mL) and higher MIC levels for penicillin G (256 μg/mL), cefoxitin (64 μg/mL), ampicillin (64 μg/mL), cefazolin (32 μg/mL), and cephalothin (32 μg/mL) (Table 2). When analyzing the relationship between the drug resistance phenotype and genotype, especially for β-lactam antibiotics, we found that there was no functionally characterized β-lactam resistance gene annotated from the whole genome sequence. However, it had six predicted β-lactamase genes annotated in the genome, of which one gene showed the highest amino acid similarity of 75.13% (a coverage of 93.0% and an identity of 80.79%) with the function-characterized β-lactamase gene bla_ERH-1, which is described to confer resistance to some penicillin and cephalosporin antibiotics (Naas et al., 2004; Supplementary Table S2). The bla_ERH-1 homologous gene of this research was finally designated bla_PSZ-1.

To verify the resistance function of the gene, we cloned the ORF of bla_PSZ-1 and its promoter region into the clone vector pUCP24, and the recombinant plasmid was transformed into E. coli DH5α competent cells. The transformant (pUCP24-bla_PSZ-1/DH5α) conferred resistance to ampicillin, penicillin G, amoxicillin, cefazolin, cephalothin, cefoxitin, ceftazidime, cefotaxime, ceftriaxone, and aztreonam but not meropenem (Table 2). Compared with the control strain (pUCP24/DH5α), the MIC levels of the recombinant harboring bla_PSZ-1 increased more than 8-fold for most β-lactam antibiotics, especially cephalothin (>256-fold), amoxicillin (≥256-fold), penicillin G (128-fold), ceftazidime (64-fold), aztreonam (64-fold), ampicillin (32-fold), cefazolin (32-fold), and ceftriaxone (32-fold) (Table 2). However, the MIC level of carbapenem meropenem was not different for the recombinant strain compared to the control strain. Tazobactam, a classical class A β-lactamase inhibitor, had a poor inhibitory effect on the resistance activity of bla_PSZ-1, whereas the MIC level of bla_PSZ-1 against aztreonam was significantly reduced in the presence of avibactam. When comparing the antimicrobial resistance spectrum of bla_ERH-1 (ERH-1 shared the highest amino acid sequence similarity with PSZ-1) and bla_PSZ-1, bla_ERH-1 did not show resistance to the second-generation cephalosporin cefoxitin, whereas bla_PSZ-1 did. However, bla_ERH-1 demonstrated resistance to carbapenems (Naas et al., 2004) but bla_PSZ-1 did not.

The length of the bla_PSZ-1 gene was 1,086 bp, and it encoded an AmpC β-lactamase of 361 amino acids. The predicted molecular weight of mature β-lactamase PSZ-1 was 39.68 kDa, and the pI was 9.14. The kinetic parameters of the purified PSZ-1 demonstrated different degrees of hydrolytic activities against penicillins and narrow-spectrum cephalosporins. Among them, the strongest hydrolytic activity (k_cat/K_m 1,088.21 ± 140.96 mM⁻¹·s⁻¹) was observed for cephalothin. PSZ-1 showed moderate hydrolytic activities for penicillin G (k_cat/K_m 176.08 ± 4.90 mM⁻¹·s⁻¹) and the first-generation cephalosporin cefazolin (k_cat/K_m 123.82 ± 5.75 mM⁻¹·s⁻¹). However, PSZ-1 showed almost no hydrolytic activity against the third-generation cephalosporin cefotaxime or monoamide ring beta-lactam class antibiotic aztreonam (Table 5). Notably, the kinetic hydrolytic activity results of the enzyme were not completely consistent with the MIC levels of the recombinant (pUCP24-bla_PSZ-1/DH5α) compared with the control strain. The hydrolytic activity of the PSZ-1 β-lactamase for cefotaxime or aztreonam was not detectable, which contrasted with the increased MIC levels (increased MIC levels of 16-and 64-fold, respectively) of the recombinant with cloned bla_PSZ-1 to the two antimicrobials. This discrepancy between in vivo and in vitro results might be attributed to the fact that, despite several attempts, for weak substrates (e.g., cefotaxime or aztreonam), β-lactamase production may be excessively induced in vivo (even up to 260-fold), whereas in vitro experiments were unable to enrich a sufficient amount of β-lactamase production for us to observe the detectable hydrolysis of the enzyme against the antibiotics (Lakaye et al., 1999). A similar phenomenon was reported in a study on the AmpC enzyme CDA-1 (Ammenouche et al., 2014). When comparing the hydrolytic activity of PSZ-1 with CDA-1, which is another AmpC β-lactamase that shares an amino acid similarity of 65.65% with PSZ-1, the highest among the β-lactamases that had the kinetic parameters characterized, PSZ-1 showed lower hydrolytic activity than CDA-1 toward cefoxitin (k_cat/K_m of 19.21 vs. 840 mM⁻¹·s⁻¹) and cephalothin (k_cat/K_m of 1,088.21 vs. 14,500 mM⁻¹·s⁻¹). In accordance with the kinetic parameters, the MIC result of the recombinant strain with the cloned bla_PSZ-1 against cefoxitin was also significantly lower than that of bla_CDA-1 (increased 8-fold vs. >256-fold). However, the MIC values of bla_PSZ-1 and bla_CDA-1 to cephalothin did not seem to vary much, and both of them increased >256-fold to cephalothin compared to the controls.

When analyzing the inhibitory effect of the β-lactamase inhibitors on PSZ-1, avibactam had a strong inhibitory effect on PSZ-1 (IC₅₀: 0.0555 μM), whereas tazobactam had a weaker inhibitory effect (IC₅₀: 7.96 μM) when the concentration of PSZ-1 was 0.00214 μM and the nitrocefin substrate concentration was 100 μM. This result is consistent with the properties of AmpC β-lactamases for inhibitors.

The evolutionary relationship analysis revealed that PSZ-1 formed a new branch in the phylogenetic tree of the function-characterized AmpC β-lactamases (Figure 1). Sequence comparison analysis between PSZ-1 and these function-characterized β-lactamases in the Beta-Lactamase DataBase and UniProt/Swiss-Prot database together revealed that PSZ-1 shared higher amino acid sequence similarities with ERH-1 (75.13%), CDA-1 (65.65%), EC-152 (65.65%), EC-2045 (65.37%), EC-96 (65.10%), EC-329 (65.10%), EC-106 (64.82%), and EC-P49 (64.82%; Figure 2). Within the deduced amino acid sequence of the protein, a serine-valine-serine-lysine tetrad (S-V-S-K) with the conserved and characteristic serine and lysine amino acid residues of β-lactamases possessing a serine active site was found at positions 65 to 68. Additionally, three motifs characteristic of class C β-lactamases (cephalosporinases) were also found: YAN (tryptophan-alanine-asparagine) at positions 151 to 153, DAEX (aspartic acid-alanine-glutamic acid-xaa) at positions 218 to 221, and KTG (lysine-threonine-glycine) at positions 315 to 317 (Joris et al., 1988; Mack et al., 2020; Figure 2).

To analyze the structure of the bla_PSZ-1-related sequences, comparative analyses of the bla_PSZ-1-encoding plasmid and the gene context were carried out. We found that five sequences sharing ≥45.0% similarities with P. endophytica X85 plasmid (CP121109) were present in the NCBI nucleotide database, all of which were complete plasmids from the genus Pantoea. Among them, the plasmid from Pantoea sp. SOD02 (CP102605, 926,844 bp in length approximately 180 kb larger than pPEX85) shared significantly higher similarities, of 62.20%, with pPEX85, while the other four, namely, the plasmid from Pantoea dispersa YSD_J2 (CP074351.1, 710,238 bp in length nearly the same size as pPEX85), the plasmid from P. dispersa Lsch (CP082347.1, 689,940 bp in length approximately 80 kb smaller than pPEX85), the unnamed plasmid from P. dispersa AHKW2b (CP082342.1, 653,898 bp in length approximately 120 kb smaller than pPEX85), and the plasmid from Pantoea sp. SO10 (NZ_CP040096, 744,154 bp in length, nearly the same size as pPEX85), shared similar sequence similarities of, respectively, 49.80, 49.70, 49.50, and 48.30% with pPEX85 (Figure 3). These sequences did not contain the bla_PSZ-1 gene, except for Pantoea sp. SO10 and Pantoea sp. SOD02.

To analyze the genetic environment of bla_PSZ-1, we first searched the NCBI nucleotide database with the bla_PSZ-1 gene as a query and collected sequences carrying a bla_PSZ-1-like gene that shared a nucleotide sequence similarity higher than 75.0% with bla_PSZ-1. Among them, only the sequences longer than 20 kb with a bla_PSZ-1-like gene at the center were kept, and finally, a total of four sequences were left for further analysis. When the five 20 kb sequences (including one of the present study) were analyzed, it was found that two of them shared >75.0% nucleotide sequence similarities with the 20 kb sequence of P. endophytica X85, whereas the similarities between the sequence of P. endophytica X85 and any of the remaining two were less than 50.0%.

Among the five sequences, three (including one of this study) were from the same genus Pantoea, and only one of the three, namely, the one in the present study, has been classified as a definite species. The 20 kb sequence from Pantoea sp. SO10 plasmid (NZ_CP040096) was particularly similar (99% coverage and 97.94% identity) to that of the present study, while the 20 kb sequence from Pantoea sp. SOD02 plasmid (CP102605) showed a lower similarity (88% coverage and 87.30% identity) with the sequence in this study. The ANI between P. endophytica X85 and Pantoea sp. SO10 (GCA_005281435.1) was 98.29%, and the two might belong to the same species, whereas the ANI between P. endophytica X85 and Pantoea sp. SOD02 (GCA_024707505.1), or between Pantoea sp. SO10 (GCA_005281435.1) and Pantoea sp. SOD02 (GCA_024707505.1) was less than 95.0 (93.49 and 93.51%, respectively). This result indicated that Pantoea sp. SOD02 was not of the same species as either of the other two.

The other two 20 kb sequences were from the genus Erwinia: Erwinia rhapontici BY21311 (NZ_CP085627) and Erwinia rhapontici MAFF 311155 (NZ_AP024333), and the upstream and downstream regions of the bla_PSZ-1-like genes were almost completely different from those of P. endophytica X85, except for the ampR-ampC fragment. No mobile genetic element was identified in these 20 kb sequences. The upstream regions of all five sequences, including the one from the present study, had an ampR gene that encoded an AmpR protein (Figure 4). AmpR activates AmpC by binding anhydrous forms of cell wall precursor muropeptides, which are believed to act as cofactors for AmpC induction (Girlich et al., 2000). These findings suggested that the sequences from phylogenetically closer species had higher sequence identity and were more conserved in the related species.