To explore the diversity and distribution of phz bacteria, we developed the Identifier of Phenazine-Producing Bacteria (IPPB) bioinformatics pipeline using core biosynthesis genes phzA/BDEFG from 23 known phz bacterial strains as references (Extended Data Fig. 1). Applying this pipeline to 1,354,919 bacterial genomes from public repositories, we identified 3,258 phz bacterial isolates (Supplementary Data 1) distributed across 193 species in 34 families (Fig. 1a and Extended Data Fig. 2a). Notably, Pseudomonas dominated the phz strains (2,893/3,258, 88.8%), followed by Burkholderia (90/3,258, 2.7%), and Streptomyces (77/3,258, 2.4%) (Extended Data Fig. 2a). Importantly, we identified 109 previously unreported species with potential to produce phenazine compounds, including Spongiactinospora rosea. Verification of the S. rosea LHW63015 strain confirmed the presence of the typical phenazine biosynthetic gene cluster phzA/BCDEFG. After 14 days of incubation in M2 broth, PCA was detected in the ethyl acetate extract by high-resolution mass spectrometry (m/z 225.0750; Fig. 1b). PCA was further confirmed by nuclear magnetic resonance (NMR) and X-ray diffraction analysis (Extended Data Fig. 2b-e).

We further mapped the geographic distribution and habitat origins of 2,746 phz strains, revealing a global distribution, with strong representation from North America, Europe and Asia. Most strains were isolated from clinical settings, soil and plants, with a notable prevalence in clinical settings, probably due to the abundance of P. aeruginosa (Fig. 1c and Supplementary Data 2). Although phz strains are broadly distributed, it remains unclear whether their phenazine biosynthetic gene clusters are actively expressed in natural conditions, as microbial secondary metabolism often requires specific environmental cues for activation. To investigate in situ phenazine production, we randomly collected ten rhizosphere soil samples, a primary habitat of phz bacteria, and detected the presence of phenazines. Phenazines were detected in 7 of the 10 samples, including 1-hydroxyphenazine (1-OH-PHZ), PCN and pyocyanin (Supplementary Fig. 1 and Supplementary Data 3). These findings suggest that phz bacteria are widely distributed and actively produce phenazines in situ, particularly within the plant rhizosphere.

To understand the evolutionary relationships among phz bacteria, we constructed phylogenetic trees for representative strains from each species using concatenated sequences of core biosynthetic genes and 16S rRNA (Supplementary Fig. 2). The tree based on the phzA/BDEFG genes differs from that based on 16S rRNA, with Pseudomonas species clustering together, while strains from other genera, such as Burkholderia and Streptomyces, are dispersed across different branches. These results suggest that horizontal gene transfer has shaped phenazine biosynthesis evolution, supporting previous hypotheses on the evolution of phenazine biosynthetic gene clusters.

Phenazines are proposed to mediate microbial interactions due to antimicrobial activity, but their impact, along with that of phz bacteria, on bacterial communities remains unclear. To assess this, we first analysed the effects of P. chlororaphis ZJU60 (a phz strain) and its ΔphzA-H mutant (phenazine-deficient) on the wheat rhizosphere microbiome using 16S rRNA gene amplicon sequencing. Both strains significantly reduced bacterial richness and diversity, as indicated by lower observed amplicon sequence variants and Shannon indices compared with controls (P < 0.05; Extended Data Fig. 3a,b). Beta diversity analysis showed distinct clustering of ZJU60-treated soil at 3 days post inoculation, suggesting microbiota restructuring effects (Extended Data Fig. 3c). LefSe analysis identified ten bacterial classes with differential abundance, showing that ZJU60 induced stronger shifts than its mutant, promoting γ-proteobacteria and reducing Bacilli populations (Extended Data Fig. 3d and Supplementary Data 4). These findings indicated that phz bacteria can drive shifts in soil microbiomes. Next, we analysed 222 publicly available rhizosphere metagenome datasets (Supplementary Data 5) to identify potential correlations between phz Pseudomonas and other bacteria. phz Pseudomonas abundance was positively correlated with the overall abundance of phz taxa (Extended Data Fig. 4a), indicating Pseudomonas as the primary phenazine producer. At the phylum level, Pseudomonadota (Gram-negative bacteria) correlated with increased relative abundance of phz Pseudomonas (Fig. 2a and Supplementary Data 6). Specifically, γ-proteobacteria, and β-proteobacteria, along with the orders Pseudomonadales and Nitrosomonadales, showed positive associations, while Actinomycetota and Bacillota (Gram-positive bacteria) were negatively correlated with the abundance of phz Pseudomonas (Fig. 2a and Extended Data Fig. 4b-d). The most prevalent taxa in Actinomycetota (Actinomycetia, Thermoleophilia and Rubrobacteria) and Bacilli within the Bacillota consistently showed negative correlations (Supplementary Data 7). To further verify the negative correlation between phenazine producers and Bacilli, we analysed publicly available metatranscriptome datasets to assess the transcription of phenazine-related genes (Supplementary Data 8). Phenazine-related transcripts in both bulk soil and rhizosphere metatranscriptomes negatively correlated with Bacilli abundance (Extended Data Fig. 5a), indicating that Bacilli are affected by phenazine producers. In addition, phenazine-related transcript abundance analysis also revealed that Pseudomonas is the primary phenazine producer, with transcript levels positively correlating with phz Pseudomonas (Extended Data Fig. 5b). Overall, these findings suggest that phz strains and their phenazines negatively impact Gram-positive bacteria while benefiting Gram-negative bacteria in the microbiome.

To confirm these findings, we conducted pairwise interaction assays to evaluate the interplay between phz and non-phz strains. We selected P. chlororaphis ZJU60, a PCN-producing strain isolated from wheat microbiome, as the representative phz strain, and tested interactions with 70 other species isolated from the same microbiome. After 72 h incubation, a clear halo was observed between ZJU60 and 50 tested Gram-positive bacterial colonies, indicating inhibition. No or only weak inhibition zones were observed with 20 tested Gram-negative strains (Fig. 2b and Extended Data Fig. 6). The minimum inhibitory concentration (MIC) of PCN was also determined for all tested strains, revealing that Gram-positive bacteria were sensitive (MIC: 8-32 μg ml), while most Gram-negative bacteria had MICs > 128 μg ml (Supplementary Table 1).

We suspect that the insensitivity of Gram-negative bacteria to PCN is due to their outer membrane, which acts as a robust barrier preventing the penetration of external substances. To test this, we treated wild-type Escherichia coli (MIC > 128 μg ml) with PCN and CHIR-090, which inhibits lipopolysaccharide synthesis of outer membranes without affecting growth. Remarkably, the combination of CHIR-090 and PCN increased the sensitivity of E. coli to PCN, with an MIC of 32 μg ml (Fig. 2c). In addition, E. coli WO153, which has a compromised outer membrane, was highly sensitive to ZJU60 and PCN (MIC reduced from 128 μg ml to 8 μg ml) compared with wild-type E. coli (Fig. 2d-f). E. coli WO153 had reduced lipopolysaccharide expression and lacks the outer membrane porin TolC, crucial for docking multidrug resistance pumps. Similarly, the MIC of E. coli tolC deletion mutant was 32 μg ml, four times higher than that of WO153 but four times lower than that of the wild type. These findings suggest that both the outer membrane barrier and efflux pumps contribute to Gram-negative bacteria's resistance to phenazines. Taken together, phenazines substantially impact microbial communities, particularly affecting Gram-positive bacteria due to their antibacterial activity.

To explore the function of phenazines in interactions between phz and Gram-positive bacteria, we used P. chlororaphis ZJU60 and B. subtilis NCIB3610 (here referred to as 3610) as a model system. We assessed pairwise interactions between ZJU60, ΔphzA-H or ΔphzA-H supplemented with PCN and 3610-GFP (a 3610 derivative constitutively expressing GFP under the P promoter). Pronounced inhibition of 3610-GFP colony expansion was observed, with a clear halo separating the two colonies (Fig. 3a, left panels). Notably, GFP signal was absent in the interfacial region between ZJU60 and 3610-GFP or ΔphzA-H strains supplemented with PCN (Extended Data Fig. 7c), indicating that B. subtilis cells were killed during the interaction. Colony-forming units (c.f.u.s) of 3610-GFP progressively declined over time, while the ZJU60 population remained stable during the time-course experiment (Extended Data Fig. 7a). The antagonistic interactions were abolished when ΔphzA-H was used instead of ZJU60 (Fig. 3a, right panels, and Extended Data Fig. 7b). To further assess the role of PCN, we tested cell-free supernatants (CFS) from ZJU60, ΔphzA-H and ΔphzA-H supplemented with PCN (50 μg ml) on 3610 growth in liquid LBGM medium. ZJU60-CFS markedly inhibited 3610 growth, while ΔphzA-H-CFS had little effect (Extended Data Fig. 7d), and growth inhibition was restored upon PCN supplementation. These results confirm that PCN secreted by ZJU60 is crucial for inhibiting B. subtilis growth.

Next, we investigated how PCN inhibits B. subtilis growth. The median effective concentration of PCN against strain 3610 was 15 μg ml (Fig. 3b). PCN treatment caused pore formation and leakage of intracellular contents, leading to cell death (Fig. 3c). Notably, PCN-treated 3610 cells were approximately twice as long as untreated cells (Fig. 3d,e), a morphological change typical of DNA replication inhibitors. This elongation was also observed with PCN-treated Paenibacillus peoriae Ppeo-29, Kocuria rosea Kr-113, Rhodococcus erythropolis Re-123 and E. coli mutant strain WO153 (Extended Data Fig. 8). These results imply that PCN-induced cell elongation is a general phenomenon.

We conducted RNA-seq upon PCN treatment and found significant induction of the SOS repair system (SOS) in B. subtilis 3610, indicating DNA damage (Fig. 3f). The SOS response safeguards genome stability by activating RecA, which cleaves the LexA repressor upon DNA damage, thereby inducing SOS regulon genes for repair. To investigate the role of RecA in the interaction with ZJU60 and PCN sensitivity, we deleted the recA gene (ΔrecA), which notably increased sensitivity to PCN and ZJU60 in the pairwise interaction (Fig. 3g,h). Next, we detected LexA-GFP cleavage in 3610 upon PCN treatment. As expected, the cleavage of LexA-GFP was induced upon PCN treatment at all tested time points (Fig. 3i). Cleavage inactivated LexA, leading to a marked upregulation of 14 SOS regulon genes, particularly the cell division inhibitor gene yneA, which showed a 10-fold increase in expression (Fig. 3j and Supplementary Table 2). The increased expression of yneA upon PCN treatment was further validated using a gfp gene fusion to its promoter (3610::P-gfp) (Supplementary Fig. 3). Given that yneA encodes a cell division inhibitor, we hypothesized that its overexpression might contribute to PCN-induced cell elongation. Indeed, overproduction of YneA using an IPTG (isopropyl β-d-thiogalactopyranoside)-inducible P promoter led to cell elongation similar to that observed in PCN-treated cells (Fig. 3k,l). Furthermore, other natural phenazines, including pyocyanin, PCA and phenazine, also induced LexA cleavage in B. subtilis 3610 (Supplementary Fig. 4). Collectively, these findings suggest that most phenazine compounds induce DNA damage in B. subtilis, leading to cell death (Fig. 3m).

Phenazines, as electron shuttles, are proposed to induce cellular ROS in target cells, a key mode of action against microbes and mammalian cells. To determine whether ROS is responsible for DNA damage induced by PCN treatment, we measured ROS levels in B. subtilis 3610 using fluorescent probes after PCN exposure. Surprisingly, PCN treatments did not significantly increase intracellular ROS levels in strain 3610 cells, even at the EC (the concentration that inhibits bacterial growth by 90%) (30 μg ml), compared to the untreated control (Extended Data Fig. 9a,b). Transcriptome analysis further confirmed no significant upregulation of oxidative stress-responsive genes in response to PCN. In fact, the expression levels of the catalase-encoding genes katE and katX were significantly reduced (Extended Data Fig. 9c). In addition, we tested the susceptibility of seven gene deletion mutants to PCN, including mutants lacking superoxide dismutase (ΔsodA, ΔsodC and ΔsodF) and catalase (ΔkatA, ΔkatE, ΔkatX and ΔydbD). None exhibited altered susceptibility to PCN (Extended Data Fig. 9d). These results suggest that ROS is not the primary cause of DNA damage induced by PCN in B. subtilis 3610.

To identify the true target of PCN, we generated spontaneous mutants resistant to PCN using the large-inoculum method. At 4× MIC (64 μg ml) in B. subtilis 3610, PCN resistance frequency was 5.2 × 10 to 1.0 × 10. We selected three PCN-resistant mutants that grew well on LBGM plates supplemented with PCN at 30 μg ml (Fig. 4a) and sequenced their genome. Resistance mutations mapped to a single amino acid change (E285Q) in ParC and two point mutations in the parEC promoter region: one in the LexA-binding site (C to G) and another in the -10 region (C to T) (Fig. 4b). ParE and ParC are subunits of bacterial topoisomerase IV, essential for resolving catenanes and separating daughter chromosomes after DNA replication during cell division.

We next confirmed whether these three mutations conferred resistance to PCN. We hypothesized that mutations in the parEC promoter region would alter the transcriptional levels of parE and parC. Using quantitative PCR with reverse transcription (RT-qPCR), we quantified mRNA levels of the parEC operon and found a 3- to 4-fold increase in transcription in resistant strains R-1 and R-2 compared with wild-type strain 3610 (Fig. 4c). This suggests that increased expression of the parEC operon reduces susceptibility to PCN in B. subtilis. To further test this, we replaced the native promoter of the parEC operon with variants harbouring 'C to G' or 'C to T' mutations, generating the mutants P-parEC and P-parEC. Both mutant strains exhibited increased resistance to PCN, similar to R-1 and R-2. In addition, strains with the native promoter replaced with an inducible P promoter in situ (P-parEC) or overexpressing parEC at the amyE locus under P control (OE-parEC; amyE:: P-parEC) also showed high resistance to PCN upon IPTG induction, both on LBGM agar plate and in liquid medium (Fig. 4d,e). To assess the functional importance of the E285Q mutation in ParC, we overproduced ParC by fusing the gene to an IPTG-inducible P promoter (OE-ParC; amyE::PparC). With IPTG induction, the OE-ParC strain exhibited dramatically increased resistance to PCN (30 μg ml) (Fig. 4d,e). In addition, we constructed parE and parC knockdown strains using CRISPR-interference, generating KD-ParE and KD-ParC strains, as both genes are essential in B. subtilis. Following xylose induction, both KD-ParE and KD-ParC strains were more sensitive to PCN than the wild-type strain 3610 (Fig. 4f,g). Altogether, these results suggest that the E285Q point mutation in ParC and the transcriptional regulation of the parEC operon contribute to increased resistance to PCN in B. subtilis, strongly implying topoisomerase IV as a direct target of PCN.

To further verify topoisomerase IV as the primary target of PCN in B. subtilis, we assessed the direct binding of PCN to purified topoisomerase IV of B. subtilis (Bs-Topo IV) using surface plasmon resonance (SPR) analysis. Results revealed that the K (equilibrium dissociation constant) of PCN binding to Bs-Topo IV was 6.35 μM, indicating a direct binding affinity (Fig. 5a). We also tested three other natural phenazines: PCA, pyocyanin and phenazine. All tested phenazines exhibited micromolar-level affinity for Bs-Topo IV, comparable to the affinity of the topoisomerase-targeting antibiotic levofloxacin (Fig. 5a).

Topoisomerase IV is responsible for decatenating interlinked double-stranded DNA molecules. To determine whether PCN inhibits Bs-Topo IV decatenation activity, we conducted a decatenation assay using purified Bs-Topo IV and kinetoplast DNA (kDNA) from Crithidia fasciculata as the substrate. The kDNA consists of a network of many minicircles and a few maxicircles. Topoisomerase IV decatenates the minicircles from the network, which then migrates into the gel and can be visualized after staining with ethidium bromide. Inhibition of topoisomerase IV prevents decatenation, causing the kDNA to remain in the wells after gel electrophoresis. As shown in Fig. 5b, Bs-Topo IV effectively decatenated kDNA, evidenced by clear minicircle bands. However, the decatenation activity of Bs-Topo IV was inhibited by increasing concentrations of PCN (1 to 30 μg ml). Increasing the amount of Bs-Topo IV neutralized the inhibitory effect of PCN (30 μg ml) (Fig. 5c). These results suggest that phenazines directly bind to topoisomerase IV and inhibit its decatenation activity. Thus, we conclude that topoisomerase IV is a direct target of phenazines in bacteria, blocking DNA synthesis, chromosome segregation, and subsequently causing DNA damage, leading to cell death (Fig. 5d).

Quinolone antibiotics target two essential bacterial type II topoisomerases, DNA gyrase and topoisomerase IV. We have shown that PCN and other natural phenazines inhibit topoisomerase IV activity. To investigate whether PCN also targets DNA gyrase, similar to quinolones, we constructed knockdown strains for gyrA and gyrB, generating KD-GyrA and KD-GyrB, respectively, and assessed their susceptibility to PCN. Both strains showed no appreciable changes in susceptibility to PCN compared to the wild type (Fig. 5e,f). These data indicated that PCN specifically targets topoisomerase IV. Since both PCN and quinolones can target topoisomerase IV, we tested for cross-resistance between PCN and levofloxacin, a representative quinolone antibiotic. We evaluated susceptibility to PCN and levofloxacin in Streptococcus pneumoniae wild-type strain R6WT and the derived levofloxacin-resistant strain R6M-3 (ref. ), which carried mutations in GyrA (S81F, E85K), ParC (S79Y) and ParE (P454S). We found that these mutations did not affect PCN susceptibility (Fig. 5g). Unexpectedly, PCN-resistant B. subtilis strains, including R-3 (ParC), OE-ParC and OE-ParEC, did not show cross-resistance to levofloxacin; in fact, they were more sensitive than the wild-type strain (Fig. 5h). These findings suggest that phenazines bind to topoisomerase IV and inhibit its decatenation activity through a mechanism distinct from that of quinolone antibiotics.

phz Pseudomonas and Bacillus are effective biocontrol agents for plant diseases, but their antagonistic interactions can limit synergistic effects in natural environments (Fig. 2). We have elucidated the MOA of phenazines against Bacillus and identified point mutations in topoisomerase IV of B. subtilis that confer resistance to phenazines. This raises the question: can a compatible synthetic community of phenazine-resistant Bacillus and phz Pseudomonas improve biocontrol efficacy against plant diseases? To test this, we developed two consortia: one comprising P. chlororaphis ZJU60 and wild-type B. subtilis strain 3610 (ZJU60&3610) and another with phenazine-resistant strain 3610R-3 (ZJU60&3610R-3). These consortia were evaluated for biocontrol efficacy against Fusarium crown rot (FCR) in wheat.

Both ZJU60 and 3610 inhibited the mycelial growth of the FCR pathogen, Fusarium pseudograminearum, in co-cultures on Warkingsman's agar plates. The phenazine-resistant strain 3610R-3 demonstrated similar antagonistic activity as 3610. Notably, the ZJU60&3610R-3 combination showed more compatibility when co-grown and stronger antagonistic activity than ZJU60 or 3610 alone (Fig. 6a). We then assessed the population dynamics of 3610 and 3610R-3 in the wheat rhizosphere during the biocontrol process. In single-strain inoculations, their populations were comparable. However, in the two-species consortium, the population of 3610R-3 in the ZJU60&3610R-3 consortium was significantly higher than that of 3610 in ZJU60&3610 (Fig. 6b). This suggested that the population of 3610 in ZJU60&3610 was inhibited by phenazine compounds secreted by ZJU60, whereas 3610R-3's phenazine resistance allowed it to coexist with ZJU60 in the rhizosphere. As expected, the ZJU60&3610R-3 consortium significantly enhanced FCR control efficacy. Individual bacterial treatments of ZJU60, 3610 and 3610R-3 resulted in control efficacies of 42%, 28% and 34%, respectively. The ZJU60&3610R-3 consortium improved biocontrol efficacy to 70%, a significant enhancement compared with ZJU60 or 3610R-3 alone (Fig. 6c). Conversely, the ZJU60&3610 consortium showed reduced efficacy (27%) due to antagonistic interactions between the strains. Collectively, these results indicate that compatible two-species consortia of phz Pseudomonas and Bacillus can have synergistic effects in controlling plant fungal diseases.