Pseudomonas aeruginosa and Klebsiella pneumoniae are gram-negative opportunistic pathogens that frequently colonize the human body and are major causes of infection. These bacteria are often co-isolated in polymicrobial urinary tract and lung infections, the latter of which is associated with increased disease severity and worse clinical outcomes. Despite their overlapping niches and clinical relevance, little is known about how these two pathogens interact and how those interactions influence human health. Given the growing recognition that microbial interactions are key drivers of disease, we investigated how P. aeruginosa and K. pneumoniae influence one another. We discovered an antagonistic interaction in which P. aeruginosa restricts the growth of K. pneumoniae. This inhibition is driven by phenazine production in P. aeruginosa, specifically the secondary metabolites pyocyanin and pyorubin, which are both necessary and sufficient to suppress K. pneumoniae growth. Using a diverse set of clinical isolates, we found that this antagonism is strain-dependent. Both the susceptibility of K. pneumoniae to phenazines and the ability of P. aeruginosa to restrict K. pneumoniae growth varies between strains. Moreover, the necessity of phenazine production is specific to the site of infection. Together, these findings demonstrate that strain background and environmental context are critical determinants of pathogen interactions. These findings reveal that both strain background and environmental redox conditions govern the ecological rules of pathogen interaction, providing a framework for predicting outcomes.

Many microbes use small molecules to compete. We found that phenazines, which are redox-active pigments made by Pseudomonas, stop Klebsiella from growing, but only in certain environments. This is because phenazines work by changing how electrons flow in cells, which depends on oxygen and other environmental factors. We show that phenazines are both necessary and sufficient to stop Klebsiella from growing and that different Klebsiella lineages vary in their susceptibility. Our results reveal a simple rule: the environment tunes chemical warfare between microbes. This principle helps explain why the same species interact differently in different human body sites and environments.

Pseudomonas aeruginosa restricts Klebsiella pneumoniae growth in a contact-independent manner

In the process of performing in vivo experiments unrelated to this study, we observed the presence of an indigenous gut bacterium associated with a reduction in Kp colonization from 24 to 48 hours after oral inoculation (S1A Fig). The goal of this experiment was to assess colonization across multiple mouse vendors and barriers (specific room in which the animals are housed), both of which are known to impact the gut microbial community structure [42]. Kp colonization was tested in three vendor backgrounds (The Jackson Labs, Taconic Farms, and Charles River Labs), from mice sourced from 13 barriers, with biological sex evenly distributed. For these experiments, we used an aggressive antibiotic selection strategy for our Kp strain, which is resistant to kanamycin and rifampicin. The unidentified gut bacterium was the only non-Kp bacteria that grew on selective media that we observed in any experiment and was only observed in a single vendor facility and two barriers (Taconic Farms, Germantown facility, barriers GT-20 and GT-1503, only Taconic Farms mice are shown in S1A Fig). This bacterium was only observed in three samples, all from male mice, and exhibited identical colony morphology. We selected three representative single colonies from these strains of this bacterium at random, which we named 145.1, 191.1, and 193.1 based on the mouse of origin, and identified them as Pa based on colony morphology and matrix-assisted laser desorption ionization–time of flight (MALDI-TOF) mass spectrometry. Whole-genome sequencing of these strains revealed that strain 145.1 was ST175 whereas 191.1 and 193.1 were unknown MLSTs, but matched 6/7 and 5/7 alleles to ST175 (as well as ST159, ST171, ST2798, ST2868, ST3748), respectively. These strains clustered with other ST175 Pa isolates, which grouped with Clade A strains (PAO1 major group, S1B Fig) from a previous study [43].

To determine if these wild Pa were able to restrict Kp growth, potentially explaining the observed reduction in Kp gut colonization density, we performed simple co-culture assays in Luria-Bertani (LB) broth. The growth of Kp strain KPPR1 was greatly restricted by Pa 145.1, 191.1, and 193.1 (Fig 1A). Given that the behavior of Pa is highly strain-dependent, we aimed to contextualize our finding using two well-characterized lab Pa strains, PAO1 and PA14. PAO1 exhibited comparatively less restriction despite its similarity of the wild Pa strains (Fig 1B), whereas PA14 was highly restrictive (Fig 1C). Next, we sought to determine if this phenotype was contact-dependent. To this end, we grew KPPR1 in filter-sterilized spent media of each Pa strain, using KPPR1 self-spent media as a control. As expected, KPPR1 growth was restricted in its own sterile spent media (Fig 1D–1F); however, growth was significantly more restricted in Pa 191.1-, and 193.1-spent LB broth compared to self-spent media (Fig 1D). To contextualize these findings with PAO1 and PA14, we repeated these assays using these strains. Similar to co-culture results, PAO1-spent media did not significantly restrict KPPR1 growth (Fig 1E), but PA14-spent media was highly restrictive (Fig 1F). Co-culture with KPPR1 had no significant effect on Pa growth (Fig 1G). Collectively, these data indicate that Pa restriction of Kp growth is contact-independent, but strain-dependent.

Next, we aimed to assess the specificity of the Kp growth restriction phenotype to Pa. First, we performed co-culture competitions with the Escherichia coli (Ec) strain MG1655. Notably, Ec and Kp have similar nutritional niches, which has been exploited to reduce Kp gut colonization in model colonization systems [44,45]. Ec MG1655 was unable to restrict KPPR1 (S2A Fig). Then, we tested our approach using the Kp strain 13F11, which is a KPPR1 mutant with an intergenic mariner transposon insertion. We reasoned that this strain should not restrict KPPR1 growth as they are nearly identical strains. As expected, Kp 13F11 did not significantly restrict KPPR1 growth (S2B Fig). Neither Ec MG1655 nor Kp 13F11 spent media significantly restricted growth of KPPR1 beyond self-spent media (S2C and S2D Fig). Next, we assayed the dependency of this phenotype on growth medium. To this end, we repeated our competition assays in M9 minimal medium supplemented with 1.0% casamino acids (“M9-cas”). Pa restriction of KPPR1 growth in M9-cas was comparable to LB broth (S3 Fig), and Ec MG1655 and Kp 13F11 were similarly unable to restrict KPPR1 growth (S4 Fig). These data indicate that growth restriction of KPPR1 was limited to Pa and was insensitive to the media we tested.

Nutrient depletion alone cannot explain Pseudomonas restriction of Klebsiella

Given that M9-cas and LB broth have similar available carbon sources, we hypothesized that the wild Pa strains and PA14 may be exhausting specific nutrients from these media more efficiently than that of Ec MG1655, Kp 13F11, or Pa PAO1. First, we measured the growth of these strains in both media types. KPPR1 had an early growth advantage over all Pa strains in LB broth that was mitigated by the end of the experiment, except for PA14, which reached a slightly lower final density than KPPR1 (S5A and S5B Fig). Ec MG1655 demonstrated a slight growth deficit compared to KPPR1, and as expected, Kp 13F11 growth was identical to KPPR1 (S5C and S5D Fig). Pa growth results in M9-cas were different than that of LB broth, wherein Pa grew to higher densities than KPPR1 by 24 hours(S5F and S5G Fig). Ec MG1655 and Kp 13F11 growth was identical to KPPR1 in M9-cas (S5H and S5I Fig). Given that Pa restriction of Kp was observed in both media, we concluded that the growth restriction phenotype was independent of growth characteristics, as Pa outgrew KPPR1 in M9-cas, but not LB broth.

Next, we aimed to determine if the metabolic potential of Pa and KPPR1 was similar. To this end, we assayed all five Pa strains, Ec MG1655, and KPPR1 for their ability to utilize different carbon sources using the BioLog PM1 and PM2 plates. No single carbon source was utilized more efficiently by the inhibitory Pa strains (PA14, 145.1, 191.1, 193.1) than the non-inhibitory strain PAO1 or KPPR1 (S6A Fig). Additionally, the Pa strains demonstrated a distinct carbon utilization profile compared to KPPR1 (S6B Fig), suggesting that differences in carbon source utilization were not explanatory for the restriction phenotype. Results with non-inhibitory Ec MG1655 were similar to the inhibitory Pa strains (S6 Fig), further indicating that differences in carbon source utilization did not explain Kp growth restriction.

Despite having distinct carbon utilization profiles, we did not exclude the possibility that Pa is exhausting specific nutrients from both LB broth and M9-cas. Thus, we reasoned that nutrient supplementation could fully restore KPPR1 growth. We tested the ability of KPPR1 to grow in spent LB broth supplemented with casamino acids using our single-strain spent media growth assay. We observed that growth was fully restored when self-spent LB broth was supplemented with casamino acids (Fig 2A). Interestingly, casamino acid supplementation was insufficient to restore KPPR1 growth to fresh LB broth levels in 145.1-, 191.1-, and 193.1-spent LB broth, or to the degree that supplementation restored growth in self-spent LB broth (Fig 2A). Similar results were observed for PAO1 and PA14; however, casamino acid supplementation not only failed to restore growth in either PAO1 or PA14 spent LB broth but also failed to significantly enhance KPPR1 growth beyond PAO1 or PA14 spent LB broth (Fig 2B).

Finally, we tested if we were able to completely restore KPPR1 growth in Pa-spent LB broth by supplementing a standard carbon and energy source: glucose. Supplementation of glucose to self-spent LB broth fully restored KPPR1 growth (Fig 2C). Like casamino acid supplementation, glucose supplementation of Pa 145.1-, 191.1-, and 193.1-spent LB broth was insufficient to restore Kp growth relative to fresh LB or glucose-supplemented self-spent LB broth (Fig 2C). Glucose supplementation was more restorative of Kp growth in PAO1 and PA14 than casamino acids; however, KPPR1 growth was still not restored to that of fresh media levels (Fig 2D). Collectively, these data suggest that an overlapping nutritional niche between Kp and Pa does not fully explain the restriction of Kp growth.

Phenazines are necessary and sufficient for Pseudomonas restriction of Klebsiella under defined redox conditions

Given that Pa inhibition of KPPR1 growth is contact-independent and is at least partially independent of the nutritional niche, we reasoned that there is a factor or set of factors that Pa secretes to inhibit Kp growth. To identify candidate factors, we screened the PA14NR transposon library. To this end, we generated a GFP-expressing KPPR1 strain using the pJL1-sfGFP plasmid and screened 5,708 Pa transposon mutants for their ability to restrict this strain, using fluorescent intensity as a proxy for KPPR1 growth (S7A Fig and S3 Data). This screen was repeated twice, and PA14 transposon mutants from co-cultures that had a fluorescent intensity |z-score| > 2.5 in both replicates were selected for validation. This yielded 18 candidate factors, 16 of which were unable to fully restrict KPPR1 growth and two that highly restricted KPPR1 growth.

The 18 transposon mutants identified as candidate factors involved in KPPR1 growth restriction were then validated for their ability to restrict KPPR1 growth. First, we determined if any of the non-restrictive phenotypes we observed were due to growth defects. Only one mutant, tpiA (PA14_62830), demonstrated a baseline growth defect, although this defect was modest (S7B and S7C Fig). Next, we tested the 18 mutants for their ability to restrict KPPR1 growth in co-culture and spent media assays (S7D and S7E Fig). 7/18 mutants displayed a phenotype analogous to our original screen in either the co-culture or spent media culture. Thus, this screen yielded 7 genes involved in Pa restriction of Kp growth (Table 1).

We observed that several mutants in our screen appeared to have modified phenazine production. Production of blue PYO and red PYR leads to coloration of spent culture media when in high enough concentrations. The psqA, rhlR, and tpiA mutants did not produce high amounts of pyocyanin (PYO, S7F Fig). As PYO has reported antimicrobial activity [46], we aimed to test their role in Kp restriction.

First, we confirmed our screen results using marker-less mutants to ensure that we were not observing unexpected effects of transposon mutagenesis. RhlR binds the autoinducer N-butanoyl-L-homoserine lactone, which is produced by RhlI, leading to expression of many genes, including the phenazine biosynthesis loci [47,48]. Both the marker-less ΔrhlR and ΔrhlI mutants were unable to restrict KPPR1 growth in our co-culture and spent media assays (Fig 3A and 3B), in both LB broth (Fig 3A and 3B), and M9-cas (S8A and S8B Fig) compared to their parental strain, PA14. Consistent with WT PA14 results (Fig 2B and 2D), supplementation of ΔrhlR and ΔrhlI spent media with casamino acids was insufficient to restore KPPR1 growth (S8C Fig), whereas glucose supplementation fully restored KPPR1 growth (S8D Fig). Importantly, the ΔrhlR and ΔrhlI mutant strains grow equivalently to PA14 in both LB and M9-cas (S5E and S5J Fig); thus, this phenotype is dependent on products of the RhlRI regulon, which includes phenazines.

Then, we repeated our co-culture and spent media assays with strains containing deletions in the phzA-G1-2, phzH, phzM, phzS, and phzHMS loci. PhzA-G are necessary to synthesize phenazine-1-carboxylic acid (PCA), which is converted to 5MPCA by PhzM and from 5MPCA to PYO by PhzS or directly converted from PCA to 1-hydroxyphenazine (1-HP) by PhzS [49]. Additionally, PhzH converts PCA to phenazine-1-carboxamide (PCN). Pyorubin (PYR), which collectively refers to 5-methylphenazine-1-carboxylate (5MPCA) and its spontaneously synthesized derivatives including 5-methyl-7-amino-1-carboxyphenazinium betaine (Aeruginosin A) and 5-methyl-7-amino-1-carboxy-3-sulfophenazinium betaine (Aeruginosin B) [50,51], is an intermediate metabolite in the PYO biosynthesis pathway that has previously been reported to have antimicrobial activities [52]. Pa encodes two phzA-G loci, and deletion of both loci renders Pa unable to synthesize PCA, PYO, 5MPCA (and thus, PYR), 1-HP, or PCN. Deletion of phzH ablates PCN production, deletion of phzS ablates PYO and 1-HP production, deletion of phzM ablates 5MPCA and PYO, and deletion of phzHMS ablates production of PYO, PYR, 1-HP, and PCN but maintains PCA production (Fig 3C). Interestingly, we observed significantly less Kp growth restriction by the ΔphzA-G, ΔphzM, ΔphzS, and ΔphzHMS mutants in co-culture, and no impact with the ΔphzH mutant (Fig 3D). Similar results were observed in our spent media assay, though the ΔphzH mutant displayed an entirely contact-dependent growth restriction of Kp which was not observed in other mutants, suggesting PYO and PYR were not responsible for the activity of this mutant (Fig 3E). Of note, we did not identify any phz mutant in our initial transposon screen. Given that one phzA-G locus is sufficient for phenazine production, we would not expect to identify these mutants as candidates. In fact, phzA1 (PA14_09480) was a candidate that inhibited Kp more than average, suggesting that there may be overexpression of phenazines; however, this observation failed to validate following the screen (S7D and S7E Fig). Additionally, the known phenazine biosynthesis regulator PqsE did not meet our significance criteria, but the data fit the general trend of a less inhibitory strain (trial 1 z-score = 2.13, trial 2 z-score = 1.19). We were surprised that the phzM or phzS mutants were not identified in our transposon screen, as these mutants should have disrupted phenazine biosynthesis; however, these transposon interruptions may not sufficiently inactivate gene function, whereas clean knockouts yielded higher-confidence data (Fig 3D). Collectively, this suggests that the phenazine PYO and/or PYR are necessary for the observed growth restriction phenotype.

Most studies investigating the antimicrobial mechanisms of phenazines have focused on PYO. The prevailing model for PYO’s antimicrobial activity involves the induction of oxidative stress: PYO readily diffuses across cellular membranes and promotes the formation of superoxide radicals, leading to cellular toxicity (reviewed in [53]). Thus, to determine if PYO is sufficient for restriction of Kp growth, we used a heterologous expression system in Ec MG1655 to constitutively produce PCA, PYR, and PYO, testing Kp (strain 13F11, Kan^R KPPR1 strain) in the resulting spent media. As expected, the empty vector and PCA over-producing strains failed to inhibit Kp growth beyond that of self-spent media (Fig 3F). Conversely, the spent media from the PYO over-producing strain inhibited Kp growth (Fig 3F) to a similar degree as WT PA14 (Fig 1F), demonstrating that PYO is sufficient to restrict Kp growth.

Pyorubin exhibits bactericidal activity

Surprisingly, the spent media from the PYR over-producing strain was completely bactericidal (Fig 3F). Of note, PYR overproduction was a result of a spontaneous frameshift mutation at T998 of phzS (annotated as “pPhzA-GS*M”) that ablated the final step of PYO production, leading to rapid accumulation 5MPCA and its derivates (PYR). High-resolution mass spectrometry of the spent medium of this strain confirmed the presence of 5MPCA and that it was absent in the spent media of other heterologous expression strains (Fig 3G). A previous report of a 5MPCA-sensitive microbe indicated that the activity of 5MPCA is dependent on its oxidation state, wherein its reduced form is more active than its oxidized form [52]. To determine if this is true for the activity of PYR against Kp, we titrated a reducing agent into spent media from our PYR-producing Ec strain and measured its ability to kill Kp. Consistent with previous reports, the addition of a reducing agent (generating reduced 5MPCA) increased anti-Kp activity (S9 Fig). Given that mutation of PhzS led to PYR overexpression in our Ec system, we were curious as to why our PA14 phzS mutant was not equally inhibitory (Fig 3D–3F), as this mutant has the genes necessary to produce PYR. We performed high-resolution mass spectrometry of the spent media from this mutant, as well as a phzS transposon mutant from our PA14 transposon library screen. We did not detect substantial PYR levels from these strains (S10A Fig), suggesting that at least partially intact PhzS is required to produce PYR or that PA14 has alternative pathways to modify PYR. Finally, we did not detect 1-HP secretion from any of our Pa or Ec strains (S10B Fig). Taken together, these data show that specific phenazines, namely PYO and PYR, are both necessary and sufficient to restrict Kp growth.

Next, we aimed to assess the antimicrobial activity of phenazines against KPPR1. First, we measured the minimum inhibitory (MIC) and minimum bactericidal concentrations (MBC) of pure PYO, as it was unclear from our co-culture approach if this compound was bacteriostatic or bactericidal. To replicate our co-culture and spent media culture assays, we measured the MIC and MBC of PYO. We performed these assays in both fresh LB and PA14ΔphzA-G spent LB, the latter of which we used to mimic the nutrient conditions in which Kp is experiencing phenazine-induced stress, as nutrient availability will modulate phenazine sensitivity by changing the cellular redox state [54]. The MIC of PYO was 8- and 32-fold lower than the MBC (Fig 4A and 4B), in fresh and spent media, respectively, indicating that PYO is primarily bacteriostatic, which is consistent with the ability of PYO to arrest bacterial respiration [46], as opposed to kanamycin, which has an MIC only 2-fold lower than the MBC, indicating bactericidal activity (Fig 4A and 4B). Then, we sought to confirm our co-culture data that demonstrated that PCA and 1-HP had no activity against KPPR1 (Fig 3D). The MICs of both compounds in pure form were higher than that of PYO both fresh (Fig 4C) and spent (Fig 4D) media. Of note, the calculated PCA MIC is likely an overestimation of activity, as the large amount of DMSO (>10% volume at ≥250 μg/mL PCA) in culture medium used to maintain PCA solubility likely affected KPPR1 growth. Nonetheless, these data support our findings that PYO restricts Kp growth, whereas PCA and 1-HP have little or no effect.

We determined that PYR is primarily bactericidal, as the MIC is at most 2-fold lower (Fig 4E) than the concentration where bactericidal activity was detected (Fig 3F). Interestingly, PYO and PYR display concentration-dependent additive (0.5 < Fractional Inhibitory Concentration < 4, % MG1655pPhzA-GS*M spent media = 3.13%) and synergistic effects (Fractional Inhibitory Concentration > 0.5, % MG1655pPhzA-GS*M spent media = 6.25%, 12.5%, 25%, Fig 4F).

Finally, we estimated the amount of PYO and PYR for each of our WT Pa strains produced in conditions identical to our spent media assay. PYO results were concordant with our co-culture and spent media assays, wherein PAO1 produced the least PYO (0.272 μg/mL [1.29 μM]) and was the least restrictive strain, followed by 145.1 (4.02 μg/mL [19.1 μM]), 191.1 (15.5 μg/mL [73.7 μM]), 193.1 (17.4 μg/mL [82.8 μM]), and PA14 (20.7 μg/mL [98.5 μM], Fig 4G). All five strains displayed little PYR production (Fig 4H), measured at Abs₅₀₀ (S11A Fig), concordant with high-resolution mass spectrometry for PA14, which displayed no 5MPCA production (Fig 3G). Only 193.1 displayed an Abs₅₀₀ higher than PA14, suggesting limits to the analytical sensitivity of this assay. Nonetheless, these data demonstrate that phenazine production by the five Pa strains is explanatory for our spent media assay results, wherein the three inhibitory strains (191.1, 193.1, and PA14) produce phenazines at or near the MIC, and the two less-inhibitory strains (PAO1 and 145.1) produce phenazines well below the MIC.

Environmental oxygen tunes phenazine potency

Some studies have shown that PYO remains active under anoxic conditions [55–57]. These findings suggest that PYO can remain chemically redox-active in the absence of oxygen; however, antimicrobial effects may still be oxygen-linked if they require oxygen-mediated redox cycling/ROS generation or maintenance of PYO in its oxidized state. To determine if this is the case for KPPR1, we measured the MIC of PCA, 1-HP, and PYO under anaerobic conditions. We observed a significant reduction of PYO activity under anaerobiosis in both fresh (Fig 4C) and spent (Fig 4D) media. Additionally, we repeated our co-culture competition assays under anaerobic conditions. As with our MIC studies, we did not observe Pa growth restriction of KPPR1 (S12A Fig). We also repeated our spent media growth assays using the heterologous phenazine expression system. Like our co-culture results, we did not observe any growth restriction or bactericidal activity in the case of PYR under anaerobic conditions (S12B Fig).

Finally, to determine if phenazine activity is dependent on respiration, we tested the MIC of PYO in conditions that favor anaerobic respiration. To this end, we repeated our MIC experiments under anaerobic conditions but grew Kp in M9 minimal medium with 0.5% glucose (“M9-glu”), conditions in which Kp predominantly undergoes fermentation, and M9-glu supplemented with 10 mM NaNO₃, which permits anaerobic respiration by nitrate acting as an alternative electron acceptor [58,59]. Addition of nitrate had no effect on the PYO MIC (S12C Fig), indicating that nitrate-permitting anaerobic respiration does not restore PYO activity. Thus, in our system, PYO antimicrobial action may require oxygen-dependent redox cycling or maintenance of oxidized PYO, supporting dependency on environmental oxygen.

Species-level susceptibility landscape reveals consistent sensitivity

Next, we aimed to determine the translatability of these findings to a diverse set of human Pa and Kp clinical isolates derived from a variety of infection sources. First, we screened the ability of 198 clinical Kp isolates to grow in PA14 spent media. The growth of all 198 strains was highly restricted, like KPPR1 (Fig 5A), indicating a universal phenotype. Co-culture assays validated these findings (S13 Fig). We then screened a subset of these clinical Kp isolates (N = 41) for growth in the spent media of our PCA, 5MPCA, and PYO constitutively expressed Ec strains. Of note, to accommodate the growth of our clinical Kp isolates, phenazine-containing spent media was generated in the absence of antibiotic selection to maintain the expression plasmids, and correspondingly, phenazine yield was lower than that in Fig 3F. Similar to our results with Kp 13F11, we observed that PYR and PYO significantly restricted the growth of our clinical Kp isolates (Fig 5B); however, this phenotype was strain-dependent, wherein some Kp strains were more resistant to PYO and PYR growth restriction than others.

Then, we screened the ability of 194 Pa strains to restrict KPPR1 growth. Interestingly, KPPR1 restriction was highly strain dependent (Fig 5C), akin to our findings with PAO1 and PA14 (Fig 1B and 1C). Co-culture assays validated these findings (S14 Fig). Given that phenazines are necessary for PA14 restriction of KPPR1, we then tested phenazine production in our screening conditions (S11A Fig). The clinical Pa strains produced phenazines to varying degrees, with many producing very little (Figs 5D, S11B, and S11D). PYO and PYR levels as measured by absorbance were significantly correlated (S11F Fig). Of note, phenazine measurements in this assay are based on absorbance (S11A Fig), thus, there may be spectral overlap between phenazines or other non-phenazine compounds produced that are detected at the wavelengths we tested. Forty-seven Pa strains had Abs₅₀₀ values higher than PA14, suggesting that these strains may secrete PYR (S11D Fig). The ability of clinical Pa strains to restrict Kp significantly inversely correlated with phenazine production (Figs 5E, S11C, and S11E); however, the correlation between phenazine production and Kp growth restriction was weak (r = −0.48, r² = 0.23), wherein many Pa strains produce little or no phenazine but are highly restrictive and others less restrictive but appear to produce high levels of phenazine. The two strains that produced high levels of phenazines in Fig 5D were excluded from the analyses in Fig 5E, as these strains produced a pigment similar in appearance to pyomelanin, resulting in high absorbance values not due to phenazine production.

To test if other secreted Pa effectors may be a factor in Kp growth restriction as our data suggested, we selected two Pa strains that we validated as inhibitory in our co-culture assay (S14 Fig): JV46 and JV69. In our screen, JV46 was slightly more inhibitory than PA14 (mean normalized fluorescence = 26,524.17 versus 29,785.32) and JV69 was less inhibitory than PA14 (mean normalized fluorescence = 35,396.03 versus 29,785.32), yet these strains produced similar levels of phenazines (JV46 Abs₅₀₀ + Abs₆₉₅ = 0.154, JV69 Abs₅₀₀ + Abs₆₉₅ = 0.103, PA14 Abs₅₀₀ + Abs₆₉₅ = 0.154). The spent media of JV69 was inhibitory to a comparative level as PA14, where JV46 was bactericidal (S15A Fig). Of note, the Abs₅₀₀ measurement of JV46 spent media was lower than that of PA14 (0.07 versus 0.086), suggesting that the bactericidal phenotype is not due to increased PYR secretion, which we confirmed by mass spectrometry (S15B Fig). This indicates that phenazine production is likely only one mechanism by which Pa restricts Kp growth. Collectively, we concluded that at high concentrations, Kp is universally sensitive, but at low concentrations, sensitivity is strain dependent.

Phenazine-dependent restriction of Klebsiella pneumoniae is site-specific

There is growing interest in using live biotherapeutics to preventatively decolonize Kp and other gut colonizing pathogens to lower infection risk [44,45,60]; however, current approaches largely depend on manipulation of the metabolic niche, rather than leveraging direct growth inhibition. Given our isolation of the wild Pa strains from mouse gut samples with low Kp colonization and their ability to directly restrict Kp, we hypothesized that these Pa strains can exclude Kp from the gut. To test this hypothesis, we colonized C57Bl6/J mice with Pa 191.1. To open the gut niche for colonization, mice were treated with ampicillin prior to Pa colonization. 191.1 was able to stably colonize the gut (S16A Fig), though colonization density was low. One week following Pa colonization, mice were colonized with KPPR1. An antibiotic-treated group not colonized by 191.1 was included as a control. No difference in KPPR1 gut colonization density was observed between mono-colonized and co-colonized mice (S16B and S16C Fig), potentially due to low 191.1 colonization. Next, we hypothesized that the intact gut microbiome may be required to observe a restrictive phenotype, as the gut microbiome of the mice in our initial in vivo experiment was intact (S1A Fig). To this end, we used an ex vivo approach to test this hypothesis. Whole large intestinal contents of C57Bl6/J mice were collected and resuspended in sterile phosphate-buffered saline (PBS) to generate gut microbiota-replete competition media. KPPR1 was anaerobically competed in this media against 145.1, 191.1, and 193.1. All Pa strains were viable in these conditions (S17A Fig); however, no reduction in KPPR1 CFUs was observed (S16D Fig). Together, these results indicate that the wild Pa strains are unable to restrict Kp growth in intestinal contents. Of note, this corroborates our finding that phenazine-dependent Kp growth restriction is dependent on environmental oxygen.

Given our gut findings that were contradictory to our original observations, we turned to other relevant body sites for interrogating the interactions between Pa and Kp where phenazine production may be important. Interestingly, stratification of clinical Pa growth restriction data based on the clinical site of origin suggested site specificity (Fig 6A and S1 Table). Pa isolated from the blood and urine were more restrictive compared to those isolated from the respiratory tract or wounds. To determine if there is a causal relationship between phenazine production and site-specific restriction of Kp growth, we aimed to test phenazine-dependent Kp growth restriction in conditions that mimic respiratory and urine infections. To this end, co-culture competitions were performed in murine bronchoalveolar lavage fluid (BALF) and ex vivo bladder homogenate. Both ex vivo tissues sustained the growth of both Kp and Pa (S17B and S17C Fig). Interestingly, we did not detect phenazine-dependent growth restriction of Kp in murine BALF (Fig 6B), supporting a diminished role in Kp lung co-infection. Conversely, we detected a significant phenazine-dependent growth restriction phenotype in ex vivo bladder homogenate (Fig 6C), consistent with a more important role of phenazines during UTI. Collectively, these data demonstrate that the impact of phenazines on Pa-Kp interaction, and therein, infection outcomes, is site-specific.

Effects of phenazines on the growth of diverse bacteria

To assess the species-specificity of phenazine-dependent growth restriction, we repeated our heterologous expression spent media growth assays with a panel of different bacteria, specifically those that are important causes of UTI. We included two K. oxytoca strains (Ko, JV282, JV453) to test within the Klebsiella genus, two Ec strains (UTI89, CFT073) to test outside the Klebsiella genus, and an Enterococcus faecalis strain (JV25) isolated from a mouse gut to test a Gram-positive species. The effects of PCA, PYO, and PYR on gram-negative bacteria were similar to that of Kp, wherein they exhibited growth restriction by Pa PA14 and high sensitivity to PYR and PYO, with overall sensitivity varying by strain (Fig 7A–7D). The Ec strains and the Ko strain JV282 also exhibited sensitivity to PCA. Interestingly, our representative Gram-positive species, E. faecalis, did not exhibit a high degree of sensitivity to Pa PA14, PYR, or PYO, but modest sensitivity to PCA (Fig 7E). Collectively, this suggests that phenazines have greater effects on gram-negative bacteria than Gram-positive bacteria; however, a more extensive study is required to test this premise. Nonetheless, these data indicate the effects of phenazines are not specific to Kp and the data shown here are likely applicable to a wide array of bacteria.

Ethics statement

This study was performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals [83]. The Indiana University Institutional Animal Care and Use Committee approved this research (protocol #22114, PI: Jay Vornhagen). Wild Pa strains (145.1, 191.1, 193.1) were isolated from mouse experiments (S1A Fig) approved by The University of Michigan Institutional Animal Care and Use Committee (protocol #PRO00007474, PI: Michael A. Bachman). Clinical isolates were collected and de-identified by RFR in accordance with approval by the Indiana University Institutional Review Board (protocol #16139, PI: Ryan F. Relich).

Materials, media, and bacterial strains

All chemicals were purchased from Sigma-Aldrich (St. Louis, MO) or Fisher Scientific (Fairlawn, NJ) unless otherwise indicated. Bacterial strains used in this study are described in Table 2. Bacteria were cultured in Luria-Bertani (LB) broth, or in M9 minimal medium (M9 salts, 0.2 M MgSO₄, 0.01 M CaCl₂, with 1% casamino acids, “M9-cas” or with 0.5% glucose, “M9-glu”) at 37 °C with shaking at 220 rpm, or on LB agar at 27 °C (Kp) or 37 °C (Pa and Ec) supplemented with kanamycin (40 or 25 μg/mL) and/or rifampicin (30 μg/mL) as appropriate. pJL1-sfGFP was a gift from Michael Jewett (Addgene plasmid 102634). The PYR expression plasmid (pSB3K3_phzA-GS*M) was obtained spontaneously by serially plating on LB-agar, leading to the identification of a colony which displayed a red corona. The plasmid was subsequently purified and re-transformed into a fresh E. coli SJ102 background strain to avoid any genomic mutations. pSJ102 containing Ec was grown at 30 °C for optimal phenazine production. For experiments that required species-specific selection, KPPR1 was selected on LB agar with 30 μg/mL rifampicin grown at 27 °C and Pa was selected on Pseudomonas Isolation Agar (Becton, Dickinson and Company, Franklin Lakes, NJ). Anaerobic growth was performed in a Coy Lab’s Vinyl Anaerobic Chamber.

Phenazine synthesis plasmids

Plasmids for PCA (pSB3K3_phzA-G) and PYO (pSB3K3_phzA-GSM) expression were obtained from a previous study [84]. The empty vector control (pSB3K3_EVC) was made by performing a NotI digestion of pSB3K3_phzA-GSM, followed by ligation with a neutral oligonucleotide linker made by annealing the following primers: 5′-GGCCGCGTACGACTTACGC-3′ and 5′-GGCCGCGTAAGTCGTACGC-3′. All plasmids were transformed into E. coli SJ102 (MG1655 intC::λpR-YFP-Cmr [85]) by electroporation. The PYR expression plasmid (pSB3K3_phzA-GS*M) was obtained spontaneously by selecting a colony of pSB3K3_phzA-GSM cells which displayed a red corona on LB-agar plates. Plasmid sequences were confirmed by full plasmid sequencing.

Whole genome sequencing

Pa and Ec strains were cultured overnight in LB broth with appropriate antibiotics for 24 hours prior to DNA extraction. DNA was extracted using Monarch Genomic DNA Purification Kit (New England Biolabs, Ipswich, MA) according to the manufacturer’s instructions with the only deviation being the substitution of 60 °C elution buffer for 34 °C DNase/RNase-Free Water (Zymo Research, Irvine, CA). The extracted DNA quantity was analyzed using a Qubit 4 Fluorometer 1× dsDNA HS analysis. DNA extractions above a concentration of 8 ng/uL were prepared for sequencing with a Rapid Barcoding Kit 24 V14 (Oxford Nanopore Technologies, Oxford, England) utilizing their Rapid sequencing DNA V14 – barcoding protocol (RBK_9176_v114_revQ_27Dec2024). Bacterial genomes were sequenced with a Flo-Min114 flow cell on a MinION Mk1B device (Oxford Nanopore Technologies, Oxford, England) and base-called using the ONT MinKNOW software v24.06.5 (Oxford Nanopore Technologies, Oxford, England). The reads were assembled into scaffolds using Flye Assembler v2.9.5 [96].

Phylogenetic analysis

We determined that the in-house assembled genomes were most similar to ST175 by multi-locus sequence typing [97]. These assembled genomes were combined with genomic sequences of five ST175 Pa strains downloaded from the bacterial and viral bioinformatics resource center on December 9, 2023 [98]. The genomic sequences of an additional 65 Pa strains were downloaded from the National Center for Biotechnology Information on December 6, 2023 [99]. These 73 genomes, in addition to the three in-house assembled genomes, were annotated using Prokka v1.14.5 [100] and the core genome determined using Roary v3.13.0 [101]. The core genome in this study was defined as loci with 90% consistency between isolates. A maximum likelihood phylogenetic tree was assembled using FastTree v2.1.11 and visualized using iTOL v7.0 [102,103].

Interspecies competition assays

For co-culture competition assays, strains were cultured overnight LB or M9-cas. Strains were then co-inoculated 1:1,000 in 1 mL of the medium of interest and mixed. This input was serially diluted and selectively spot-plated for CFU quantification. After 24 hours of shaking incubation, the output was again serially diluted and plated. For spent-media competition assays, strains were cultured overnight in the medium of interest. Spent media was created by filtering the supernatant of centrifuged culture with a 0.22 μm syringe filter. The strain of interest was inoculated 1:1,000 into spent media and was then plated and incubated as described above. For MG1655pPhzA-GS*M spent media assays, except those in Fig 3F, dithiothreitol was added to final concentration of 5 mM.

Growth curves

Strains were cultured overnight in culture medium of interest (LB or M9-cas) then diluted to an OD₆₀₀ of 0.01 in the medium of interest the following day. Cultures were incubated at 37 °C with aeration and Abs₆₀₀ readings were taken every 15 min using a BioTek microplate reader with Gen5 software (Version 3.12.08, BioTek, Winooski, VT) for 24 hours. To simultaneously assess doubling time, growth rate, lag time, non-sigmoidal growth due to stress and bacterial density, area under the curve analysis was used to quantify differences in growth, as in [38].

BioLog Phenotype MicroArray analysis

BioLog Phenotype MicroArrays (Biolog, Hayward, CA) were performed according to manufacturer’s instructions with some modifications, as in [38]. KPPR1, PAO1, PA14, and the mouse-derived Pa strains (Pa 145.1, Pa 191.1, and Pa 193.1), were cultured overnight in LB, then bacteria were pelleted, washed once in sterile PBS, and re-suspended in sterile PBS to avoid aberrant transfer of xenonutrients. Each strain was diluted in IF-0 medium to a final OD₆₀₀ of 0.035 and 100 μL was plated onto plates PM1 and PM2 with gentle mixing. After inoculation, plates were statically incubated overnight at 37 °C. Following 24 hours of incubation, growth was measured at OD₅₉₅.

PA14NR library screen

Construction of the PA14NR transposon library used in this study has been extensively described elsewhere [104]. Library plates were pin-replicated onto black 96-well microplates containing 100 μL of LB agar. After 24 hours of static growth at 37 °C, liquid culture of KPPR1pJL1-sfGFP was diluted 1:1,000 in LB broth and 100 μL was added on top of the agar. After 24 hours of static 37 °C incubation, a fluorescence read was performed (Excitation: 479, Emission: 520, Optics: Bottom, Gain: 50).

High-resolution mass spectrometry

For high-resolution mass spectrometry (HR-MS) detection of phenazines in cell-free supernatant, bacteria were grown overnight in LB broth at 30 °C or 37 °C with cognate antibiotics. Cultures were centrifuged at 7,000g for 5 min and supernatant was filtered through a 0.22 μm syringe filter. This filtrate was then centrifuged at 16,000g for 10 min to remove debris, and 500 μL was diluted 1:1 with LC-grade methanol. Pure phenazines were also included. Sample (1 μL) was injected onto 1290 LC - 6545 QTOF. The ESI-HRMS analyses were performed in positive ion mode, utilizing nitrogen as the nebulizing and drying gas. The instrumental conditions were as follows: nebulizer pressure, 25 psi; drying temperature, 325 °C; drying gas, 8 L/min; sheath gas temperature, 400 °C; sheath gas flow, 12 L/min; fragmentor, 220 V; skimmer, 65 V; Oct 1 RF Vpp, 750 V. For the first HPLC condition, an Agilent ZORBAX Eclipse Plus C18 column (HD 2.1*5 0 mm 1.8-Micron) was used for separation. 0.1% formic acid in water (v:v) as mobile phase A and 0.1% formic acid acetonitrile as mobile phase B (v:v) were used. The flow rate is 0.6 mL/min. The gradient starts as 5% B and holds for 0.5 min, increased from 5%B to 95%B in 4.5 min and holds for 0.5-min, 95%B to 5%B in 0.5 min and holds for 0.5 min. For the second HPLC condition and ACQUITY UPLC BEH Amide column (150 mm × 2.1, 1.7 μm) was used for separation. The mobile phases were heated to 45 degrees. 0.1% formic acid in water (v:v) as mobile phase A and 0.1% formic acid acetonitrile as mobile phase B (v:v) were used. The flow rate is 0.6 mL/min. The gradient starts as 5% A and holds for 1 min, increased from 5% A to 95% A in 9 min and holds for 1 min, 95% A to 5% A in 1 min and holds for 1 min. Data were analyzed using mzmine [105]. Extracted ion chromatographs were generated for PCA ([M + H+] = 225.0586 m/z), 5MPCA ([M+] = 239.0815 m/z), and PYO ([M+] = 211.0866 m/z) with 10 ppm error.

MIC and MBC assays

MIC assays were performed as previously described [106]. For PCA, 1-HP, and PYO (1-phenazinecarboxylic acid, 1-phenazinol, 5-methyl-1(5H)-phenazinone, respectively, Cayman Chemical, Ann Arbor, Michigan) was resuspended to a concentration of 2–16 mg/mL in dimethylsulfoxide and then diluted into LB, M9-glu + /− 10 mM NaNO₃, or PA14ΔphzA-G spent media to 1 mg/mL. 100 μL of this solution was plated into U bottom 96-well plate in triplicate and then 2-fold serially diluted into LB or PA14ΔphzA-G spent media 10 times. Then, 50 μL of KPPR1 culture inoculated in LB or PA14ΔphzA-G spent media to final concentration of OD₆₀₀ = 0.02 was plated across each serial dilution, yielding a final OD₆₀₀ = 0.01, and incubated at 37 °C shaking for 24 hours. After 24 hours, MIC was determined to be the lowest concentration where growth was not visually observed.

For PYO MBC assays, the entire MIC assay described above was spot-plated onto LB agar at 24 hours and grown overnight at 37 °C. After overnight growth, MBC was determined to be the lowest concentration at which no Kp colonies were recovered.

For PYR, 100 μL of 100% MG1655pPhzA-GS*M spent media was plated into U bottom 96-well plate and serially diluted in MG1655pEmpty spent media with 10 mM DTT 7 times. Then, 50 μL of KPPR1 culture inoculated in MG1655pEmpty spent media to final concentration of Abs₆₀₀ = 0.02 was plated across each serial dilution, yielding a final DTT concentration of 5 mM and Abs₆₀₀ = 0.01. Additionally, KPPR1 culture was inoculated in 100 μL of 100% MG1655pPhzA-GS*M spent media with 5 mM DTT to final OD₆₀₀ = 0.01 to ensure a 100% MG1655pPhzA-GS*M spent media condition. All media were incubated at 37 °C shaking for 24 hours, and MIC was determined as above.

To assess additive and/or synergistic effects of PYO and PYR, a checkerboard assay was used. To this end, a 2-fold serial dilution of PYO, starting at 62.5 μg/mL was combined with 2-fold serial of MG1655pPhzA-GS*M spent media, starting at 100%. DTT was added to a final concentration of 5 mM and KPPR1 culture was inoculated to final OD₆₀₀ = 0.01. This array was incubated at 37 °C shaking for 24 hours. After 24 hours, MIC was determined to be lowest concentration where growth was not visually observed.

Phenazine quantification

The absorbance spectra of pure PYO and MG1655pPhzA-GS*M spent media was measured in both reducing (5 mM DTT) and oxidizing conditions (0.3% peroxide). We determined that oxidized MG1655pPhzA-GS*M spent media and untreated PYO yielded the greatest differences between these spectra at Abs₅₀₀ and Abs₆₉₅, respectively. For PYO quantification, Pa cells were removed from spent Pa media generated under conditions identical to spent media assays. The raw Abs₆₉₅ of spent media was measured or compared to a standard curve of pure PYO to estimate PYO concentrations. For PYR quantification, Pa spent media was oxidized, then the raw Abs₅₀₀ of spent media was measured or compared to a standard curve of oxidized MG1655pPhzA-GS*M spent media to interpolate estimate PYR concentrations.

Collection and screening of clinical isolates

The IU Health Division of Clinical Microbiology collected clinical isolates between May and July 2023 from IU Health facilities across the state of Indiana. Of these, 198 isolates were identified by MALDI-TOF (Bruker) as Kp, and 194 as Pa. These isolates were derived from a variety of infection sites and patient information was de-identified. Isolates were passaged once from primary plating media (sheep blood agar or MacConkey agar) prior to storage on brain-heart infusion agar slants (ThermoFisher) at room temperature.

To screen Kp growth in PA14 spent media, clinical Kp isolates and KPPR1 were cultured overnight in LB and inoculated 1:100 into sterile spent PA14 media or fresh LB. After 24 hours of shaking incubation at 37 °C, Kp growth was measured at OD₆₀₀.

To screen Kp growth against specific phenazines, a subset of 49 clinical Kp isolates and KPPR1 were cultured in sterile spent MG1655pEmpty media. Eight isolates were determined to be unable to grow in this spent media and excluded from future assays. The remaining 41 clinical Kp isolates were cultured in sterile spent MG1655pPhzA-G, MG1655pPhzA-GS*M, or MG1655pPhzA-GSM media generated in the absence of antibiotic selection for their respective expression plasmids to permit clinical isolate growth. After 24 hours of shaking incubation at 37 °C, Kp growth was measured at Abs₆₀₀.

To screen clinical Pa isolate restriction of KPPR1 growth, clinical Pa isolates and KPPR1pJL1-sfGFP were cultured overnight in LB and inoculated 1:1,000 in LB for a 1:1 competition of KPPR1pJL1-sfGFP in co-culture with one clinical isolate. After 24 hours of shaking incubation at 37 °C, an OD₆₀₀ reading was taken, as was a fluorescence read (Excitation: 479, Emission: 520, Optics: Top, Gain: 50).

Ex vivo interspecies competition assays

Strains for ex vivo interspecies competition assays were cultured overnight in LB, washed in sterile PBS and re-suspended in sterile PBS before inoculation to avoid aberrant transfer of xenonutrients.

Large intestinal contents: 6-to-8-week-old C57Bl6/J male mice were sourced from the Jackson Labs from barrier RB07. Following acclimation, mice were euthanized, the large intestine (cecum and colon) was removed, and its contents were gently homogenized in 2 mL sterile, pre-reduced PBS and divided into 400 μL aliquots. Immediately following collection, samples were transferred to an anaerobic chamber and inoculated to a final concentration of ~5 × 10⁷ CFU KPPR1 alone or an equal ratio of KPPR1 with each wild Pa strain. Samples were incubated at 37 °C for 48 hours, then dilution plated on species-selective media to quantify final bacterial densities. Only male mice were used to account for sex-derived microbiome differences.

BALF: 6-to-8-week-old C57Bl6/J male mice from the Jackson Labs’ barrier RB07 and C57Bl6 male mice from Charles River from barrier K61 were sourced for BALF collection. Two vendor-barrier sources were used to determine if there were vendor-specific effects of the lung microbiome, as has been previously reported [107]. Following acclimation, mice were euthanized, and BALF was collected as previously described [38]. If BALF recovery was >1 mL, sterile PBS was added to 1 mL. BALF was divided into 400 μL aliquots and inoculated 1:1,000 with a 1:1 mix of KPPR1 and WT PA14 or PA14ΔphzA-G. Input bacterial densities were determined by dilution plating on species-selective media. Samples were incubated aerobically with agitation (220 rpm) at 37 °C for 24 hours, then dilution plated on species-selective media to quantify final bacterial densities. Only male mice were used to account for sex-derived microbiome differences, and no vendor-based differences were detected.

Bladder homogenate: Bladders were collected from the mice used for BALF collection. Bladders were collected into 1 mL sterile PBS, then thoroughly homogenized. Homogenized tissue was centrifuged at 21,300g and resulting supernatant was collected. If homogenate recovery was 1 mL, sterile PBS was added to 1 mL. Bladder homogenate was divided into 400 μL aliquots and inoculated 1:1,000 with a 1:1 mix of KPPR1 and WT PA14 or PA14ΔphzA-G. Input bacterial densities were determined by dilution plating on species-selective media. Samples were incubated aerobically with agitation (220 rpm) at 37 °C for 24 hours, then dilution plated on species-selective media to quantify final bacterial densities. As above, no vendor-based differences were detected.

In vivo models

Gut colonization model: Gut colonization was performed as previously described [38,66], with some minor modifications. Briefly, 6-to-8-week-old C57Bl6/J (equal numbers of male and female) mice were sourced from the Jackson Labs from barrier RB07. Following acclimation, prior to colonization, mice were administered 0.5 g/L ampicillin via drinking water. Four days following antibiotic administration, mice were orally gavaged with ~10⁷ CFU stationary-phase Pa 191.1 suspended in 250 μL sterile PBS. Fecal pellets were collected at predetermined timepoints, homogenized in 300 μL sterile PBS, and Pa gut density was measured via dilution plating on selective medium. Seven days after 191.1 administration, mice were orally gavaged with ~10⁸ CFU stationary-phase KPPR1 suspended in 250 μL sterile PBS. Fecal pellets were collected at predetermined timepoints, homogenized in 300 μL sterile PBS, and Kp gut density was measured via dilution plating on selective medium. Mice were euthanized seven days after KPPR1 administration, ceca were collected and homogenized in 1 mL sterile PBS, and Kp density was measured via dilution plating on selective medium.

Statistical analysis

All in vitro experimental replicates represent biological replicates. For in vitro studies two-tailed Student t test or ANOVA followed by Tukey’s multiple comparisons post-hoc test on log₁₀ transformed data (transformation performed to fit data to normal distribution) was used to determine significant differences between groups. For ex vivo and in vivo studies, all experiments were replicated at least twice, accounting for sex as a biological variable when appropriate. A p-value of less than 0.05 was considered statistically significant for the above experiments, and analysis was performed using base R version 4.5.0. Data were processed, analyzed, and visualized using R packages “readxl,” “tidyverse,” “dplyr,” “vegan,” “outliers,” “EnvStats,” “ggplot2,” “ggrepel,” and “ggtext.”

Use of AI

ChatGPT-4o was used to edit the R scripts associated with this manuscript to enhance their accessibility following human-driven drafting and analysis. Functionality of all edited scripts was confirmed prior to publication.

S1 Fig. Detection and phylogenetic characterization of wild Pa strains in Kp-colonized mice.

6- to 8-week-old C57 mice from Taconic Farms were orally inoculated with 10⁸ CFU WT KPPR1 or 13F11. Kp was enumerated from feces at 24 hours and large intestinal contents (LI) 48 hours post-inoculation (A). Pa was isolated from mouse 145, 191, and 193 and subjected to whole genome sequencing. These strains and subset of Pa isolates from a previous study (31173069) were used to build an approximately-maximum-likelihood phylogenetic tree based on a core genome alignment of these strains to determine if strains 145.1, 191.1, and 193.1 group with Clade A (PAO1 clade), Clade B (PA14 clade), or Clade C (split Pa7 clade). Select Pa strains absent in the previous study that represent ST175 were also included, as strains 145.1, 191.1, and 193.1 were predicted to be most like ST175 using multi-locus sequence typing. The ST175 representative and mouse strains grouped with Clade A, thus they are labeled “Proposed Clade A” (B). The data underlying this Figure can be found in S2 Data and the raw phylogenetic tree data can be found in S4 Data.

https://doi.org/10.1371/journal.pbio.3003809.s001

(TIF)

S2 Fig. Ec MG1655 and an isogenic neutral Tn mutant do not restrict KPPR1 growth in LB.

Kp KPPR1 was grown alone or in co-culture in LB with Ec MG1655 (A), Kp 13F11 (B), or in filter sterilized spent media of KPPR1, Ec MG1655 (C), or Kp 13F11 strain (D). For A–D, “Log fold change_{24 hours}” = log₁₀(output KPPR1 CFU at 24 hours/input KPPR1 CFU). p-values represent Tukey multiple comparison correction following one-way ANOVA. Each data point is a biological replicate, and horizontal lines indicate the mean of each dataset. The data underlying this Figure can be found in S2 Data.

https://doi.org/10.1371/journal.pbio.3003809.s002

(TIF)

S3 Fig. Pa restricts Kp growth in M9 medium supplemented with 1.0% casamino acids in a strain-dependent, contact-independent manner.

Kp KPPR1 was grown alone or in co-culture in M9 medium supplemented with 1.0% casamino acids with mouse-derived wild Pa (A), PAO1, PA14 (B) or in filter-sterilized spent media of KPPR1 or each Pa strain (C–D). For A–D, “Log fold change_{24 hours}” = log₁₀(output KPPR1 CFU at 24 hours/input KPPR1 CFU). p-values represent Tukey multiple comparison correction following one-way ANOVA. Each data point is a biological replicate, and horizontal lines indicate the mean of each dataset. The data underlying this Figure can be found in S2 Data.

https://doi.org/10.1371/journal.pbio.3003809.s003

(TIF)

S4 Fig. Ec MG1655 and an isogenic neutral Tn mutant do not restrict KPPR1 growth in M9 medium supplemented with 1.0% casamino acids.

Kp KPPR1 was grown alone or in co-culture in M9 medium supplemented with 1.0% casamino acids with Ec MG1655 (A), Kp 13F11 (B), or in filter-sterilized spent media of KPPR1, Ec MG1655 (C), or Kp 13F11 strain (D). For A-D, “Log fold change_{24 hours}” = log₁₀(output KPPR1 CFU at 24 hours/input KPPR1 CFU). p-values represent Tukey multiple comparison correction following one-way ANOVA. Each data point is a biological replicate, and horizontal lines indicate the mean of each dataset. The data underlying this Figure can be found in S2 Data.

https://doi.org/10.1371/journal.pbio.3003809.s004

(TIF)

S5 Fig. Growth dynamics of select Kp, Pa, and Ec strains in LB and M9 medium supplemented with 1.0% casamino acids.

KPPR1, PAO1, PA14, Pa 145.1, Pa 191.1, Pa 193.1, MG1655, 13F11, PA14ΔrhlR, PA14ΔrhlI were grown in LB (A-E) or in M9 medium supplemented with 1.0% casamino acids (F-J). Area under the curve (AUC) analysis was used to quantify growth at early (0–6 hours, A-B, F-G) and late (0–24 hours, A-J) stages of growth. p-values represent Tukey multiple comparison correction following one-way ANOVA. For growth curves, each data point represents the mean, and vertical bars represent the standard error of the mean. For AUC analysis, each data point is a biological replicate, and horizontal lines indicate the mean of each dataset. The data underlying this Figure can be found in S2 Data.

https://doi.org/10.1371/journal.pbio.3003809.s005

(TIF)

S6 Fig. WT Pa and Ec strains are less metabolically flexible than Kp.

KPPR1, PAO1, PA14, Pa 145.1, Pa 191.1, Pa 193.1, and Ec MG1655pEmpty were grown in BioLog Phenotype Microarray plates PM1 and PM2 (mean of three biological replicates displayed, each row is an individual carbon source, A). Euclidean distance was used to measure the dissimilarity between the growth phenotypes of each strain (B). The data underlying this Figure can be found in S2 Data.

https://doi.org/10.1371/journal.pbio.3003809.s006

(TIF)

S7 Fig. PA14NR library screen and validation.

To identify candidate factors involved in Kp growth restriction, the PA14NR library was replicate plated onto LB-agar, grown overnight at 37 °C, and inoculated with GFP-expressing KPPR1 (A). After 24 hours of co-culture, the fluorescence of each co-culture was measured, yielding 18 candidate transposon mutants, 16 of which exhibited reduced restriction, and 2 that exhibited enhanced restriction. PA14 and the 18 candidate transposon mutants identified to have a role in Kp growth restriction were grown in LB (B) and area under the curve (AUC) analysis was used to quantify growth (C). In panel B, each data point represents the mean, and vertical bars represent the standard error of the mean. Kp KPPR1 was grown alone or in co-culture in LB with PA14 or the 18 transposon mutants (D) or in filter-sterilized spent media of KPPR1, PA14, or the 18 transposon mutants (E). PYO (F) and PYR (G) were measured at 695 and 500 nm, respectively, from select transposon mutants. For D–E, “Log fold change from WT PA14” = log₁₀(output KPPR1 CFU at 24 hours/input KPPR1 CFU) in transposon mutant co-culture or spent media culture/ log₁₀(output KPPR1 CFU at 24 hours/input KPPR1 CFU) in WT PA14 co-culture or spent media culture. For C, p-values represent Tukey multiple comparison correction following one-way ANOVA, and for D–E, p-values represent one-sample t test from a hypothetical mean of 0. For C–G, each data point is a biological replicate, horizontal lines indicate the mean of each dataset, and in C–E, red datasets are statistically significant from their relative comparisons. Panel A was Created in BioRender. Tilston-lunel, N. (2026) https://BioRender.com/6b5vwn8. The data underlying this Figure can be found in S2 Data.

https://doi.org/10.1371/journal.pbio.3003809.s007

(TIF)

S8 Fig. rhlRI is required for Kp growth restriction by Pa in M9 medium supplemented with 1.0% casamino acids.

KPPR1 was grown alone or in co-culture in M9 medium supplemented with 1.0% casamino acids with WT PA14, PA14ΔrhlR, or PA14ΔrhlI (A) or in filter-sterilized spent media of KPPR1 or each Pa strain (B), supplemented with water, 1% casamino acids (“Cas,” C) or 0.4% glucose (“Glu,” D). For A-D, “Log fold change_{24 hours}” = log₁₀(output KPPR1 CFU at 24 hours/input KPPR1 CFU). p-values represent Tukey multiple comparison correction following one-way ANOVA. p-values over columns indicate comparison to “Fresh media” condition. Each data point is a biological replicate, and horizontal lines indicate the mean of each dataset. The data underlying this Figure can be found in S2 Data.

https://doi.org/10.1371/journal.pbio.3003809.s008

(TIF)

S9 Fig. PYR activity is redox-dependent.

13F11 (Kan^R KPPR1 variant) was grown in fresh LB broth or spent media from MG1655 constitutively expressing PYR (pPhzA-GS*M). Dithiothreitol (DTT) was titrated into both media. “Log₁₀(Fold change from no DTT)” = log₁₀(output 13F11 CFU at 24 hours/input 13F11 CFU) in fresh or spent media + DTT/ log₁₀(output 13F11 CFU at 24 hours/input 13F11 CFU) in fresh or spent media without DTT. Each data point is a biological replicate, and horizontal lines indicate the mean of each dataset. The data underlying this Figure can be found in S2 Data.

https://doi.org/10.1371/journal.pbio.3003809.s009

(TIF)

S10 Fig. Pa phzS mutants do not produce PYR and PA14, and heterologous phenazine-expressing Ec do not produce 1-HP.

LC-MS was used to quantify PYR secretion from WT PA14, MG1655pPhzA-GS*M, PA14ΔphzS, and PA14_09400, which is a phzS transposon mutant from PA14NR library (A) or 1-HP secretion from WT PA14, MG1655pEmpty, MG1655pPhzA-G, MG1655pPhzA-GS*M, or MG1655pPhzA-GSM in comparison to pure 1-HP (B). The data underlying this Figure can be found in S2 Data.

https://doi.org/10.1371/journal.pbio.3003809.s010

(TIF)

S11 Fig. Clinical Pa strain phenazine production.

The spectra of pure PYO and the spent media MG1655 constitutively expressing PYR (pPhzA-GS*M) was measured under native, oxidizing (0.3% hydrogen peroxide) and reducing (5 mM dithiothreitol [DTT]) conditions (A) to identify wavelengths at which those phenazines can be differentiated. We determined that PYO and PYR in oxidizing conditions can be differentiated at 695 and 500 nm, respectively (red vertical bars). Clinical Pa strains (N = 194), WT PA14, and PA14ΔphzA-G were grown in LB broth and the Abs₆₉₅ of spent media was measured after 24 hours (B). Abs₆₉₅ results were correlated to growth restriction results from Fig 5C (Spearman correlation test, C). Clinical Pa strains (N = 192), WT PA14, PA14ΔphzA-G, and MG1655 constitutively expressing PYR (pPhzA-GS*M) and PYO (pPhzA-GSM) were grown in LB broth and the Abs₅₀₀ of spent media was measured after 24 hours following oxidation by 0.3% hydrogen peroxide (D). Abs₅₀₀ results were correlated to growth restriction results from Fig 5C (Spearman correlation test, E). Abs₅₀₀ and Abs₆₉₅ results were correlated with one another (Spearman correlation test, F). Each data point represents the mean read of each culture. The data underlying this Figure can be found in S2 Data.

https://doi.org/10.1371/journal.pbio.3003809.s011

(TIF)

S12 Fig. Phenazines do not restrict Kp growth under anaerobic conditions.

KPPR1 was grown anaerobically alone or in co-culture in LB with mouse-derived wild Pa, PAO1, and PA14 (A). 13F11 (Kan^R KPPR1 variant) was grown anaerobically in filter-sterilized spent media of MG1655 containing an empty vector (pEmpty) or constitutively expressing PCA (pPhzA-G), PYR (pPhzA-GS*M), and PYO (pPhzA-GSM, B). The minimum inhibitory concentration (MIC) and minimum bactericidal concentrations (MBC) were determined for KPPR1 for pure PYO in M9 minimal media with 0.5% glucose and 10 mM NaNO₃ (C). For A-B, “Log fold change_{24 hours}” = log₁₀(output Kp CFU at 24 hours/input Kp CFU). Each data point is a biological replicate; horizontal lines indicate the mean of each dataset. For C, p-values represent Tukey multiple comparison correction following one-way ANOVA, comparing aerobic versus anaerobic MICs within each phenazine. Note that 13F11 growth in MG1655 spent media is no different than self-spent media in anaerobic conditions, whereas growth is restricted in the presence of PYR and PYO in aerobic conditions (see Fig 3F). The data underlying this Figure can be found in S2 Data.

https://doi.org/10.1371/journal.pbio.3003809.s012

(TIF)

S13 Fig. Validation of clinical Kp screen results.

Clinical Kp strains (N = 6) were selected for validation (A). KPPR1 or select Kp strains were grown alone or co-cultured with select clinical Pa strains (B). For B, “Log fold change_{24 hours}” = log₁₀(output KPPR1 CFU at 24 hours/input KPPR1 CFU). p-values represent Tukey multiple comparison correction following one-way ANOVA compared to the “PA14 + KPPR1” condition. Each data point is a biological replicate, and horizontal lines indicate the mean of each dataset. The data underlying this Figure can be found in S2 Data.

https://doi.org/10.1371/journal.pbio.3003809.s013

(TIF)

S14 Fig. Validation of clinical Pa screen results.

Clinical Pa strains (N = 10) were selected for validation (A). KPPR1 was grown alone or in co-culture with select clinical Pa strains (B). For B, “Log fold change_{24 hours}” = log₁₀(output KPPR1 CFU at 24 hours/input KPPR1 CFU). p-values represent Tukey multiple comparison correction following one-way ANOVA compared to the “KPPR1 alone” condition. Each data point is a biological replicate, and horizontal lines indicate the mean of each dataset. The red line represents the mean normalized fluorescence for KPPR1, and the orange line represents the mean normalized fluorescence for KPPR1 + PA14. The data underlying this Figure can be found in S2 Data.

https://doi.org/10.1371/journal.pbio.3003809.s014

(TIF)

S15 Fig. Kp growth restriction in clinical Pa spent media is variable.

KPPR1 was grown in fresh LB medium or filter-sterilized spent media of KPPR1, PA14, JV46, or JV69 (A). “Log fold change_{24 hours}” = log₁₀(output KPPR1 CFU at 24 hours/input KPPR1 CFU). Datapoints outlined in red are below the limit of detection (200 CFU/mL). p-values represent Tukey multiple comparison correction following one-way ANOVA. Each data point is a biological replicate, and horizontal lines indicate the mean of each dataset. LC-MS was used to quantify phenazine secretion from JV46 (B). The data underlying this Figure can be found in S2 Data.

https://doi.org/10.1371/journal.pbio.3003809.s015

(TIF)

S16 Fig. Pa does not exclude Kp from the gut.

C57Bl6/J mice were treated for 4 days with 0.5 g/L ampicillin, then orally gavaged with ~10⁷ CFU 191.1. Seven days post-191.1 colonization, mice were orally gavaged a second time with ~10⁸ CFU KPPR1 (N = 10). Antibiotic-treated KPPR1 mono-colonized mice served as a control (N = 10). 191.1 fecal loads were monitored post-colonization (A), and KPPR1 fecal (B) and cecal loads (C) were measured were monitored post-colonization (after 7 days 191.1 colonization) or at the end of the experiment, respectively. Large intestinal contents from C57Bl6/J (no antibiotic treatment, N = 5) mice were collected and resuspended in sterile PBS. ~5 × 10⁷ CFU KPPR1 alone or an equal ratio of KPPR1 with each wild Pa, was inoculated into large intestinal contents and grown anaerobically. KPPR1 density was measured at 48 hours (D). The data underlying this Figure can be found in S2 Data.

https://doi.org/10.1371/journal.pbio.3003809.s016

(TIF)