The bacterial strains and plasmids used in this study are listed in Table 1, and the sequences of the primers are listed in Supplementary Table S1. The E. coli, P. aeruginosa WK172, and PAO1 strains were grown in Luria Bertani broth (LB) at 37°C unless otherwise indicated. WM3064 cells were grown in LB broth containing 0.3 mM DAP (2,6-diamino-pimelic acid). Carbenicillin (100 μg/ml) was used to maintain plasmids pMD19, pUT18C, and pMQ70, and kanamycin (50 μg/ml) was used to maintain plasmids pET28b, pKT25, and pHGR01. When needed, 1 mM isopropyl-β-d-thiogalactoside (IPTG) and 10 mM-arabinose were added as expression inducers.

To overexpress crlT, crlA, and crlTA in E. coli and P. aeruginosa PAO1 hosts, the full coding regions of crlT, crlA, and crlTA were amplified with the primer pairs listed in Supplementary Table S1 using WK172 genomic DNA as the template. PCR products were purified using a gel extraction kit (Qiagen, Hilden, Germany), digested with EcoRI (or NheI for pHGE-base plasmids) and HindIII, and purified with a PCR product purification kit (Qiagen). The purified PCR products were ligated into pMQ70 expression plasmids and transferred into E. coli cells and PAO1. The correct constructs were verified by DNA sequencing using the primer pair pMQ70-F/R.

Overnight cultures of E. coli BW25113 carrying the empty pMQ70, pMQ70-crlA, pMQ70-crlT, and pMQ70-crlTA plasmids were diluted with LB with carbenicillin (100 μg/ml) at an OD₆₀₀ of 0.1, and 10 mM arabinose was added at OD₆₀₀ ~ 0.5 to induce the production of proteins. Cells were collected at 3 h by centrifugation at 3000× g for 5 min, and the cells were resuspended in 300 μl of phosphate-buffered saline (PBS, pH 7.4). Cells were stained with the LIVE/DEAD^™ BacLight^™ Bacterial Viability Kit according to the manufacturer’s instructions. Cells were stained with 1 μl SYTO^®9 nucleic acid stain (300 μl, 3.34 mM in DMSO) and 1 μl propidium iodide (300 μl, 20 mM in DMSO) and incubated in the dark for 10 min. Then, the stained cells were collected and rinsed once with PBS before resuspending in 50 μl of PBS. The fluorescence emitted from the cells was observed using fluorescence microscopy (Zeiss, Germany).

Deoxyribonucleic acid fragments 300 bp upstream of the translational start of crlT were generated by PCR, digested with EcoRI and HindIII, and cloned into the promoter-less lacZ-fusion vector pHGR01 (Fu et al., 2013) to create plasmid pHGR01-PcrlTA-lacZ. The resulting plasmid was verified by sequencing, introduced into E. coli WM3064 strains for integration and prepared into competent cells. Then WM3064/pHGR01-PcrlTA-lacZ cells were transformed by plasmids pMQ70, pMQ70-crlA, and pMQ70-crlTA. Overnight cells were 1% diluted, and a final concentration of 10 mM arabinose was added as an expression inducer. The cells carrying the integrated reporter system were collected by centrifugation and washed with phosphate-buffered saline when they grew to mid-log phase (OD600 ∼ 0.7). The cell soluble protein and β-galactosidase activity were determined using previously described protocols (Wu et al., 2011).

For BACTH assays, the coding regions of crlA and crlT were cloned into pUT18C and pKT25. The constructed plasmids were co-transformed into E. coli BTH101 (cya-99) competent cells. The co-transformed cells were plated on LB agar plates supplemented with kanamycin (50 μg/ml), carbenicillin (100 μg/ml), and X-gal (20 μg/ml). The cells were cultivated at 30°C for 24 h. The cells harboring pKT25 (without an insert) and pUT18C-zip (fused with a leucine zipper protein) plasmids were used as negative controls, and the cells harboring pKT25 (without an insert) and the cells harboring pKT25-zip (fused with a leucine zipper protein) and pUT18C-zip plasmids were used as positive controls (Huihui et al., 2014).

Protein CrlTA complexes with a hexahistidine tag at the N-terminus of the CrlTA complex without any tag were purified using E. coli BL21 with pET28b-His-crlTA and pET28b-crlTA. Strains were grown in LB with kanamycin (50 μg/ml) and were induced with 1 mM IPTG at OD₆₀₀ ~ 0.1 for 5 h. Cells were collected and resuspended in lysis buffer [50 mM potassium phosphate buffer (pH 8.0), 300 mM NaCl, and protease inhibitor cocktail (Sigma-Aldrich, United States)]. Samples were sonicated using a Sonic Dismembrator (Ningbo Dongzhi, China) at level 2 for 5 min on ice. Ni-NTA resin (Qiagen) was used according to the manufacturer’s protocol. Purified proteins were desalted by passage on disposable Sephadex G-25 prepacked PD-10 columns pre-equilibrated in 20 mM Tris–HCl buffer (pH 8.0), and the protein concentration was measured by the Bi Yuntian BCA assay kit (Haimen, Jiangsu, China). Tricine–SDS-PAGE was performed as previously described (Baba et al., 2006). A total of 20 μg of protein from each sample was loaded for Tricine–SDS-PAGE.

The experiment was conducted as described previously (Jia et al., 2018). The T7 RNA polymerase promoter sequence was designed in the forward primer and the PCR products were transcribed in vitro with the HiScribe^™ T7 Quick High Yield RNA Synthesis Kit following the instructions (New England Biolabs, Ipswich, MA, United States).

Total RNA was isolated using bacterial total RNA isolation kit (Tiangen, Beijing, China). cDNA synthesis was conducted using reverse transcription (Promega, Madison, WI, United States). Total cDNA (50 ng) was used for qRT-PCR using the Step One Real-Time PCR System (Applied Biosystems StepOne Real-Time PCR System, United States). The primers used for qRT-PCR are listed in Supplementary Table S1. The level of 16S rRNA gene transcript was used to normalize the gene expression data, and the fold change of each gene was calculated as described previously (Pfaffl, 2001).

Three independent overnight cultures of pMQ70, pMQ70-crlA, and pMQ70-crlTA were diluted to OD₆₀₀ 0.05, and a final concentration of 10 mM arabinose and 100 μg/ml carbenicillin were added. The tested phages were serially diluted 10-fold in LB medium. The same volume (100 μl) of cells and phages were mixed with MOI = 1:100 in 96-well plates, and the final volume of each well was 200 μl. The plate was then cultured at 37°C, and optical density measurements at a wavelength of 600 nm were taken every 2 h using an Infinite M200 Pro NanoQuant with the lid open (Goldfarb et al., 2015). The one-step growth curves of these phage were determined as previously recently (Xuan et al., 2022b).

The whole genome of clinically isolated Pseudomonas aeruginosa WK172 was sequenced and submitted to the NCBI database (Accession No. CP060004; Duan et al., 2021). The potential TA systems were predicted using the web-based tool TADB database (Xie et al., 2018), and two neighboring genes overlapping eight nucleotides, H5022_22855 and H5022_22850, were predicted to be a putative TA pair (Figure 1A; Supplementary Figure S1). H5022_22855 (CrlT) is 121 aa in length, and it is predicted to be an uncharacterized RelE-like toxin. H5022_22850 (CrlA) is a member of XRE family transcriptional regulator 107 aa in length. To probe whether the two-gene cassette is a bona fide TA system, three constructs were constructed using the plasmid pMQ70 under the control of the pBAD promoter to enable arabinose-dependent induction. When the pMQ70-based plasmids were transformed into the E. coli K-12 BW25113 host, cells harboring pMQ70-crlT exhibited notable growth inhibition as shown by the slower increase in turbidity (OD₆₀₀) compared to the empty vector pMQ70, while no reduction was observed based on colony-forming units (CFUs; Figure 1B). In contrast, the expression of crlA via pMQ70-crlA did not affect cell growth. CrlA neutralized the inhibitory effect of CrlT when the two genes were co-expressed via the pMQ70-crlTA plasmid (Figure 1B). We thus hypothesized that the CrlT toxin should inhibit cell growth rather than result in cell death. To further check this, we first plated BW25113/pMQ70-crlT cells on plates containing arabinose to induce the production of CrlT, and the cells grew slower than those cells harboring the empty plasmid pMQ70 (Figure 1C; Supplementary Figure S2). In contrast, CrlA restored the inhibitory effect of CrlT on cell growth (Figure 1C). Next, live/dead staining was used to check whether the production of CrlT induces cell death. The production of CrlT was induced with 10 mM arabinose for 5 h, and cells were co-stained with SYTO^®9 and propidium iodide (PI). As expected, green fluorescent nuclear and chromosome counterstains entered the cell membranes, and the majority of cells emitted green fluorescence when CrlT was overproduced. Only a few cells had damaged cell membranes and were able to be stained with PI and emitted red fluorescence, suggesting that most cells were alive (Figure 1D). In addition, no difference was obtained among the cells harboring empty vector or CrlA or CrlTA. These results suggest that the CrlT toxin reduces the growth of the bacteria rather than killing them. Since CrlT is a RelE family protein, it may be able to cleave mRNA. To test this, we purified Chis-tagged CrlT (Figure 2A) and used it to cleave in vitro transcribed ompA mRNA. As shown in Figure 2B, CrlT-CHis cleaves the ompA mRNA in dose-dependent manner (lanes 1–5), and heat inactivation of the toxin totally abolished the cleavage activity of CrlT-Chis (lane 6). Thus, CrlA constitute a bona fide TA system in P. aeruginosa, in which CrlT is the bacteriostatic toxin that inhibits cell growth by cleaving mRNA and CrlA neutralizes its toxicity.

Since both CrlT and CrlA are proteins and CrlT is a RelE-like family protein, they may act together as a type II TA system. To test whether crlT and crlA are co-transcribed, we first designed a primer pair covering the crlTA region, in which the forward primer crlT-F is in the crlT gene and the reverse primer crlA-R is in the second gene crlA. Then, a polymerase chain reaction (PCR) assay was conducted with three templates, including genomic DNA, cDNA, and total RNA. As shown in Figure 3A, a single band of 682 bp was amplified using cDNA synthesized from total RNA as the template, and the PCR product was sequenced to be the region among the two primers, indicating that crlT and crlA form a single operon (Figure 3A, lane 3). As controls, the same band was detected using genomic DNA (Figure 3A, lane 2) as the template but not for total RNA (Figure 3A, lane 4). Then the products were purified and ligated into the pMD19 plasmid and sequenced, and they were indeed the crlTA region (data not shown). In addition, to exclude the DNA contamination in the RNA, primer pair non-transcribed region-F/R of a non-transcribed region was also amplified in the three templates, and the region was only amplified in the genomic DNA (Figure 3A, lane 6). These results showed that crlT and crlA are co-transcribed.

One typical feature of type II TA systems is that the antitoxin itself or in the context of the TA complex inhibits the expression of TA by binding to its promoter region. To test this hypothesis, we PCR amplified the promoter region of crlTA and cloned it into the pHGR01 plasmid to create the WM3064/pHGR01-PcrlTA-lacZ reporter system. We then overexpressed crlA using pMQ70-crlA in the reporter system, and the β-galactosidase activity was determined. The β-galactosidase activity was reduced from 1928.4 ± 218.1 MU to 1116.8 ± 59.5 MU after overexpressing crlA (Figure 3B). However, overproduction of the TA complex CrlTA resulted in 2397.0 ± 2.9 MU β-galactosidase activity, which implies that CrlT abolishes the repression of CrlA on the expression of the crlTA TA system. In addition, a palindrome sequence 5'-TCNGGC GCCNGA-3′ was identified in the promoter region, and it may be the binding site of CrlA to autoregulate the expression of the crlTA operon (Supplementary Figure S1). To confirm this, we mutated the palindrome sequence to 5'-TCNGGC AGGNAC-3′ in the reporter system. Both CrlA and the CrlTA complex were produced in the cells as above, and the β-galactosidase activity was affected by neither CrlA nor the CrlTA complex (Figure 3B). Collectively, these results showed that CrlA indeed repressed the expression of crlTA by binding to the palindrome in the promoter region.

Antitoxin in the type II TA system neutralizes the toxicity of toxins by direct protein–protein interactions. We then conducted a bacterial two-hybrid assay with crlT and crlA overexpressed using pUT18C and pKT25 separately, and they showed a strong interaction, similar to the positive control to overexpress zip in both vectors (Figure 3C). To further confirm this, we exchanged the two vectors to overexpress crlT and crlA, and the results consistently showed that CrlT and CrlA interact with each other. We further performed a pull-down assay to determine the in vivo interaction between CrlT and CrlA. We used pET28b-His-crlTA to express a N-terminal hexahistidine-tagged (His-tagged) CrlT along with an untagged antitoxin CrlA. After induction, CrlT and CrlA were both induced to produce successfully (Figure 3D, lanes 2 and 3). Affinity purification of NHis-CrlT using Ni-NTA agarose beads and subsequent Tricine–SDS-PAGE revealed that a small protein could be pulled down together with His-tagged CrlT, this small protein should be CrlA, and the ratio between CrlA and CrlT should be 1:1 (Figure 3D, lane 4). To exclude the non-specific binding of protein to the Ni-NTA beads during the purification process, we also constructed a pET28b-crlTA to express untagged CrlT and untagged CrlA, and the two proteins were also induced successfully (Figure 3D, lanes 5 and 6), and neither of them could bind to Ni-NTA beads (Figure 3D, lane 7). The antitoxin protein of the type II TA system is usually unstable, and the degradation of antitoxin can free the toxin. Here, we purified the N-terminal His-tagged CrlA via pET28b-crlA-His (Figure 3E, lanes 1–3) and monitored the stability of CrlA-His daily after purification. As shown in Figure 3E, lanes 4–9, smaller degraded products appeared after day 1, and the full-length CrlA disappeared by day 5. Therefore, CrlA and CrlT comprise a type II TA system in which CrlT is a toxin and unstable antitoxin CrlA neutralizes its toxicity by protein–protein interactions in vivo.

To explore whether the expression of the crlTA TA system is physiologically related, we determined the mRNA levels of crlA and crlT at both the exponential and stationary growth phases using qRT-PCR method. As shown in Figure 4, both crlA and crlT were induced significantly at stationary phase, especially crlA. Furthermore, the mRNA levels of crlA were significantly higher than those of crlT at the stationary phase (Figure 4), indicating that there may be more CrlA in cells and that there might be a separate promoter for the crlA gene. In addition to the promoter of the crlTA operon (Supplementary Figure S1), a separate promoter with its -10 and -35 regions located in the crlT open reading frame was identified similarly to that reported previously for the type II antitoxin HigA (Guo et al., 2019) and HicB (Turnbull and Gerdes, 2017).

To uncover the potential function of CrlA, we first BLASTed the protein using the web-based Universal Protein Resource (UniProt). Among the 250 aligned XRE family regulators and helix-turn-helix (HTH) domain containing proteins, the Cro-like protein in Pseudomonas sp. BAY1663 showed the highest similarity (89.5% identity), and several other Cro-like proteins were also identified (Figure 5A). Then eight Cro-like proteins in different bacteria were selected and aligned, and highly conserved amino acid sites were obtained (Figure 5B). In temperate phages, such as lambda and P22 phages, Cro proteins compete with prophage repressors to bind to the region controlling the prophage lysogenic/lytic switch, and they were involved in the phage infection process (Harvey and Prell, 1981; Albright and Matthews, 1998). Here, we reasoned that CrlA may confer an antiphage protein. To test this, several available Pseudomonas lytic phages were mixed with cells overexpressing crlA and crlTA with a certain multiplicity of infection (MOI) of 1:100, and the growth of cells was determined in 96-well plates. These phages are environmentally isolated lytic phages, including PAP-L5 (NCBI Accession NO. OL754589), PAOP5 (GenBank: KU297675.1), PAP8 (NCBI Accession NO. OL754588), and QDWS (Xuan et al., 2022a). Phages PAP-L5 and PAOP5 were resisted by CrlA only rather than CrlTA complex (Figure 5C). PAP8 and QDWS were resisted by both CrlA and the CrlTA complex. Besides, crlTA TA operon palindrome and nearby sequences were identified in the genome of these phages, and CrlA may inhibits phage replication by binding to these sequences (Supplementary Table S3). We further analyzed the diversity of CrlA in 9,696 Pseudomonas genomes using the Pseudomonas genome database with 50% identity and 50% coverage cut off, and 3,536 hits were obtained, indicating that 36.1% of Pseudomonas harbor this protein and that 2,880 hits showed > 90% identities with 100% coverage (Supplementary Table S2). Thus, CrlAs may protect hosts against phage infection.