P. diazotrophicus Pd1, E. coli DH5α and E. coli BL21(DE3) were preserved in our laboratory. E. coli DH5α served as the host for constructing recombinant plasmids. E. coli BL21(DE3) was used for the induced expression of the CAR-2 protein. E. coli S17-1(λpir) was purchased from Beyotime Biotechnology and was used for constructing suicide plasmid. P. diazotrophicus Pd1 was cultured on a blood agar plate at 37°C. The draft complete sequences of P. diazotrophicus Pd1 was deposited in NCBI GenBank under the accession number: PRJNA1020661.

To clarify the similarity between predicted CAR-2 and other β-lactamases, we conducted the analysis through the Beta-lactamase database (BLDB)-structure and function BLAST analysis (http://www.bldb.eu:4567/). The signal peptide was predicted through protein signal peptide prediction (https://www.novopro.cn/tools/signalp). The molecular mass and predicted isoelectric point of CAR-2 were estimated using ExPASy(https://web.expasy.org/compute_pi/).

A pHSG398::CAR-2 construct was generated to test the β-lactam substrate specificity of CAR-2, as described previously (Ranjan et al., 2021). High-fidelity PCR was used to amplify the complete coding frame sequence of bla_CAR-2 with the forward primer EcoRI-bla _CAR-2-F and the reverse primer PstI-bla _CAR-2-R. The primer design was carried out using Primer 5 software with the following parameter settings: annealing temperature is 58–62°C, primer length is 18–24 bp, and GC content is 40–60%. Primer sequences are shown in Supplementary Table S1 . High-fidelity PCR was purchased from Vazyme. The total reaction volume of 50 µL consisted of 20 µL ddH₂O, 25 µL 2× Phanta Max Master Mix, 2 µL primer F, 2 µL primer R, and 1 µL template DNA. The thermal cycle was programmed for 3 min at 95°C for pre-denaturation, followed by 32 cycles of 15 s at 95°C for denaturation, 15 s at 60°C for annealing, 60 s at 72°C for extension; a final extension was conducted for 5 min at 72°C. The PCR product and the pHSG398 plasmid (Takara) were double-digested with EcoRI and PstI, and then the PCR product was ligated to the pHSG398 plasmid using T4 ligase. The restriction enzymes EcoRI and PstI were purchased from Thermo Fisher Scientific. Using the forward primer EcoRI-bla _CAR-2-F and the reverse primer PstI-bla _CAR-2-R, PCR was performed to verify that the recombinant pHSG398 plasmid was successfully connected and transformed into E. coli DH5α, which was named E.coli DH5α(pHSG398::CAR-2). This recombinant strain expresses bla _CAR-2 under the control of the lac promoter. Similarly, using the forward primer Test-pHSG398-F and the rear primer Test-pHSG398-R, PCR was performed to confirm the successful transformation of the empty pHSG398 plasmid into E. coli DH5α, which was named E.coli DH5α(pHSG398). This strain was used as a negative control.

The minimum inhibitory concentration (MIC) values of susceptibility of strains to 14 kinds of β-lactam antibiotics including ampicillin, piperacillin, carbenicillin, cefalotin, cefoxitin, cefuroxime, ceftazidime, cefotaxime, ceftazidime/clavulanic acid, cefepime, aztreonam, imipenem, meropenem, and ertapenem were detected by the Mueller-Hinton (MH) agar plate dilution method. Antibiotics were purchased from Aladdin Biochemical Technology Co., Ltd. In brief, we first revived strains on a blood plate for 24 hours, then picked single clones and subcultured them for 24 hours. Subsequently we added each antibiotic with a serial dilution concentration to an empty plate, and then poured 25 mL of MH agar medium at about 50°C into each plate. Next, we adjusted the bacterial suspension concentration to 0.5 McFarland turbidity standard with a McFarland turbidimeter, and titrated 1 μL of the bacterial suspension onto the surface of the MH agarplates. After air-drying, finally we placed the plates in an incubator at 37°C, and observed the MIC values of each strain 20 hours later. E. coli ATCC8739 was selected as the quality control bacterium. Repeat the biological experiment three times.

For determination of bla _CAR-2 MBL activity, MIC values of 14 kinds of β-lactam antibiotics were also determined in the presence of 0.2 mM EDTA to demonstrate the reduction in the bla_CAR-2 activity due to metal chelation.

To test the key active sites of the CAR-2 protein, we constructed six single mutants (H136A, H138A, H211A, D140A, H141A, and H276A) in E. coli DH5α. Site-directed mutagenesis was conducted using Mut Express II Fast Mutagenesis Kit V2 (Vazyme Biotech Co., Ltd). Briefly, the codons corresponding to the amino acid residues His136, His138, His211, Asp140, His141, and His276 were replaced with the codon GCC, corresponding to alanine(A), at the 5’ end of the primers. Primer sequences are listed in Supplementary Table S1 . Using pHSG398::CAR-2 plasmid DNA as a template, six single mutant plasmids—pHSG398::CAR-2(H136A), pHSG398::CAR-2(H138A), pHSG398::CAR-2(H211A), pHSG398::CAR-2(D140A), pHSG398::CAR-2(H141A), and pHSG398::CAR-2(H276A)—were amplified with the primer pairs mentioned above. Subsequently, these plasmids were individually transformed into E. coli DH5α. Sanger sequencing (first-generation sequencing) using the forward primer Sanger-F and the rear primer Sanger-R was performed to determine the ~650 bp region containing the mutation site. Sequence alignment using the Chromas software confirmed the successful construction of all six single mutant strains.

Firstly, we used high-fidelity PCR to amplify the coding frame sequence (removing the signal peptide sequence) of bla _CAR-2 with the forward primer NdeI-CAR-2-F and the reverse primer XhoI-CAR-2-R. The restriction enzymes NdeI and XhoI were purchased from Thermo Fisher Scientific. Primer sequences was showed in Supplementary Table S1 . The PCR product and pET-21b(+) were double-digested with NdeI and XhoI, and then PCR product ligated to the pET-21b(+) plasmid using T4 ligase. Using the forward primer NdeI-CAR-2-F and the reverse primer XhoI-CAR-2-R, PCR was performed to verify that the Recombinant pET-21b(+) plasmid was successfully transformed into E. coli BL21(DE3), which was named E. coli BL21(DE3) (pET-21b-CAR-2). A single clone of E. coli BL21(DE3) (pET-21b-CAR-2) was added to 500 mL of liquid LB medium and cultured at 37°C and 220 rpm/min for 9 hours. Subsequently, we added IPTG to the bacterial culture to a final concentration of 0.5 mM and then incubated it on a shaker overnight at 37°C.The bacterial culture was collected by centrifugation at 5000rpm/min for 5 minutes and resuspended in a Tris·HCl buffer solution containing 20 mM imidazole. The bacteria were lysed by sonication (power 30%, sonication time 45 min, on for 3 s and off for 3 s).After centrifuging the lysate at 12,000 rpm/min for 20 minutes, it was filtered through a sterilizing filter. In the affinity chromatography column, the filtrate was mixed with Ni²⁺-NTA Agarose 6FF (Shanghai Yuanye Bio - Technology Co., Ltd) for 30 minutes. A Tris·HCl buffer solution containing 20 mM imidazole was used as the washing buffer to elute the non-target proteins. A Tris·HCl buffer solution containing 50 mM, 100 mM, 200 mM, 300 mM or 400 mM imidazole was used as the elution buffer to elute the target protein bound to the nickel column. The purification effect was analyzed by SDS-PAGE. The His-tagged CAR-2 protein was dialyzed overnight in a PBS buffer solution (pH 7.4) using a dialysis bag. The concentration of the recombinant CAR-2 protein was determined by a Nano-300 Micro-Spectrophotometer.

Kinetic assays were conducted with a UV-1100 spectrophotometer as previously described (Stoczko et al., 2008).β-Lactam hydrolysis was detected by monitoring the variation in absorbance using the characteristic molecular extinction coefficient to substrates: cefalothin (Δϵ_260nm=-14,300 M^-1cm^-1); cefuroxime (Δϵ_274nm=-18,600 M^-1cm^-1); cefotaxime sodium (Δϵ_264nm=-11,625 M^-1cm^-1). The assays were conducted at 25°C with 3 ml of reaction mixture in 50mM HEPES buffer, pH 7.5 containing CAR-2 protein (0.634µM for cefalothin;1.27µM for cefuroxime; 0.634µM for cefotaxime sodium), 50µM ZnSO₄.The steady-state kinetic parameters (v _max and k _m) were calculated after direct fit of the Michaelis-Menten equation on the experimental data with the program GraphPad. The value of k _cat​ was obtained by dividing v _max by the CAR-2 protein concentration.

To demonstrate that CARR binds IR111 (the 111- bp sequence of the interval between CARR and bla _CAR-2) and inhibits the expression of the chloramphenicol resistance gene Cm ^R, we constructed E.coli DH5α (pBluescript/IR111-Cm^R) and E.coli DH5α (pBluescript/CARR-IR111-Cm^R). In brief, the IR111 fragment and IR111-CARR fragment were amplified from P. diazotrophicus Pd1 genomic DNA using primers IR111-F, IR111-R and IR111-F, IR111-CARR-R, respectively. The Cm ^R coding frame sequence was amplified from the pDM4 plasmid using primers Cm^R-F and Cm^R-R. IR111 and Cm^R were ligated to pBluescript double-digested with KpnI and SacII by ligase independent cloning (LIC) with ClonExpress MultiS One Step Cloning Kit (Vazyme Biotech Co., Ltd).Likewise, IR111-CARR and Cm^R were ligated into the pBluescript vector. The two recombinant plasmids were subsequently introduced into E.coli DH5α by heat shock. Finally, the MIC values of chloramphenicol in E. coli DH5α (pBluescript/IR111-Cm^R) and E. coli DH5α (pBluescript/CARR-IR111-Cm^R) were determined by the plate agar dilution method.

The construction of ΔCARR was carried out according to homologous recombination. In brief, a upstream 583 bp fragment and a downstream 635 bp fragment of CARR were amplified from P. diazotrophicus Pd1 genomic DNA with primers CARR-up-F, CARR-up-R, and primers CARR-down-F, CARR-down-R, respectively. These two fragments and a tetracycline resistance cassette, which was amplified with CARR-TC-F and CARR-TC-R, were ligated into Suicide pDM4 plasmid by LIC with ClonExpress MultiS One Step Cloning Kit. The recombinant plasmids were subsequently introduced into P. diazotrophicus Pd1 strains by electrotransformation. The bacterial solution was subsequently spread on LB agar plates containing tetracycline (20µg/mL) and piperacillin (24µg/mL). A single colony was subcultured for 5 generations in liquid LB medium. Then, 20µl of bacterial culture was spread on LB agar plates containing tetracycline (20µg/mL). Chromosomal DNA of the transformants was checked by PCR with primers CARR-Delete-F and CARR-Delete-R. The primers used in this study are listed in Supplementary Table S1 .

RNA extraction was performed using the FreeZol Reagent (Vazyme Biotech Co., Ltd) according to the Manufacturer’ instructions. The RNA concentration was analyzed using a Nano-300Micro-Spectrophotometer. Subsequently, 500 ng of total RNA was reverse transcribed into cDNA with the HiScript III RT SuperMix for qPCR(+gDNA wiper) kit (Vazyme Biotech Co., Ltd). Real time quantitative PCR (qPCR) was performed using a fully-automated real-time fluorescent quantitative PCR analysis system Gentier 96 (Tianlong Technology Co., Ltd, Xi’an, China) with ChamQ Blue Universal SYBR qPCR Master Mix (Vazyme Biotech Co., Ltd) in 50 ng of cDNA. The primers used are detailed in Supplementary Table S1 . Data normalization was performed using 16S rRNA (a housekeeping gene), and relative gene expression was calculated using the 2^–ΔΔCt approach.

Student’s t-tests was used to evaluate the statistical significance of differences. A P-value < 0.05 was considered to indicate a statistically significant difference.

The 990-bp bla _CAR-2 encodes a 330-residue protein, which the highest sequence identity (60.82%) with subclass B3 MBLs CAR-1, as shown in Figure 1 . Compared with that in CAR-1, the N-terminus of CAR-2 lacks a segment of 10 residues, LPSQGTETKG. CAR-2 shows lower identity to other subclass B3 MBLs, with identity scores ranging from 27.99% (AIM-1) to 34.85% (BJP-1). The novel CAR-2 was predicted to have two zinc-binding motifs: His136, His138, His211, Asp140,His141, and His276, as shown in Figure 1 . The estimated theoretical isoelectric point and molecular weight of CAR-2 without the predicted signal peptide were 7.28 and 33.717 kDa, respectively. CAR-2 is speculated to possess three external loops (eLs), namely eL1- eL3 (Stoczko et al., 2008; Yun et al., 2022), as shown in Figure 1 .

To clarify the substrate specificity of CAR-2, we constructed E.coli DH5α(pHSG398::CAR-2), which expressed CAR-2, and tested the MIC values for 14 β-lactam antibiotics. Compared with those for E.coli DH5α and E. coli DH5α(pHSG398), the MIC values for E.coli DH5α(pHSG398::CAR-2) increased significantly only for cefalothin, cefuroxime and cefotaxime sodium by 2-, 16- and 32- fold, respectively ( Table 1 ). This result indicates that cefalothin, cefuroxime and cefotaxime are substrates of CAR-2. To confirm that CAR-2 is a MBL, we measured MICs in the presence of EDTA. Compared with those for E. coli DH5α(pHSG398::CAR-2) without EDTA, the MIC values for cefalothin, cefuroxime and ceftriaxone sodium in the presence of EDTA dropped to baseline levels ( Table 1 ). This confirms that EDTA inhibited CAR-2 catalytic activity, indicating that CAR-2 is a MBL.

To determine the catalytic active site of CAR-2, we constructed six single mutants: E.coli DH5α[pHSG398::CAR-2(H136A)], E.coli DH5α[pHSG398::CAR-2(H138A)], E. coli DH5α [pHSG398::CAR-2(H211A)], E. coli DH5α [pHSG398::CAR-2(D140A)], E. coli DH5α[pHSG398::CAR-2(H141A)], and E. coli DH5α[pHSG398::CAR-2(H276A)], as shown in Supplementary Figure S1 . Subsequently, we tested the MICs for cefuroxime and cefotaxime sodium. Cefuroxime susceptibility results revealed that compared with E. coli DH5α(pHSG398::CAR-2), mutants H136A, D140A and H211A showed the greatest MIC reduction- dropping the baseline levels, H138A and H276A showed moderate decreases, and H141A the least ( Table 2 ). Cefotaxime MICs mirrored this trend. We conclude that these six residues are active sites, with H136, D140, and H211 being the primary zinc-binding sites.

We successfully induced the recombinant CAR-2 protein, and the size of this protein is consistent with the expected molecular weight of approximately 35 kDa, as shown in Supplementary Figure S2 . To demonstrate that CAR-2 acted as a MBL can hydrolyze cefalothin, cefuroxime and cefotaxime, we performed CAR-2 enzymatic kinetic experiments. k _cat was 0.84S^-1 and k _cat/k _m was 1.32×10⁴ M^-1·S^-1 for cefalothin; k _cat was 0.21 S^-1 and k _cat/k _m was 4.26×10³ M^-1·S^-1 for cefuroxime; and k _cat was 0.61 S^-1, k _cat/k _m was 1.87×10⁴ M^-1·S^-1 for cefotaxime sodium, as shown in Table 3 . When 1 mM of EDTA was added to the above reaction solution, no activity was observed, indicating that the CAR-2 enzyme requires zinc ions.

Through alignment analysis using DNAMAN software, the amino acid sequence of this CARR showed 63.76% similarity to that of the LysR-type transcriptional regulator (ECA2848)) located directly upstream of CAR-1. However, the CARR protein was poorly conserved with β-lactamase transcriptional regulators such as AmpR, VarR, and NmcR. The antibiotic sensitivity results showed that the MIC values ​of chloramphenicol for E.coli DH5α(pBluescript/IR111-Cm^R) and E.coli DH5α(pBluescript/CARR-IR111-Cm^R) were 120 µg/mL and 30 µg/mL, respectively, as shown in Figure 2 . In the presence of CARR, the MIC of chloramphenicol decreased 4-fold (P < 0.001). These results indicate that CARR likely inhibits bla _CAR-2 expression through the IR111 promoter. We successfully constructed the CARR deletion mutant strain, as shown in Supplementary Figure S3 . To further prove that CARR inhibits the expression of bla _CAR-2, we used qPCR to detect the mRNA expression of 16S rRNA and bla _CAR-2 in P. diazotrophicus Pd1 wild-type strain andΔCARR, as shown in Supplementary Figure S4 . Compared with that of the wild-type strain, the mRNA expression of bla _CAR-2 in the ΔCARR strain increased approximately 10-fold (P < 0.01), as shown in Figure 3 , indicating that CARR inhibited bla _CAR-2 expression.

In order to clarify the association between CARR and cefalothin, cefuroxime and cefotaxime sodium, the MIC values of cefalothin, cefuroxime and cefotaxime sodium of P. diazotrophicus wild-type strain and ΔCARR strain were determined. Compared with those of the wild-type strain, the MIC values of the ΔCARR strain for cefalothin, cefuroxime and cefotaxime sodium increased 8-,8- and 16-fold, respectively, as shown in Table 4 . These results demonstrated that CARR reduced the resistance to cefalothin, cefuroxime and cefotaxime sodium. Therefore, we conclude that CARR inhibits the expression of bla _CAR-2, which reduces the resistance of P. diazotrophicus to cefalothin, cefuroxime and cefotaxime.