Pseudomonads coordinate innate defense against viruses and bacteria with a single regulatory system

Summary

Bacterial cells live under constant existential threats imposed by other bacteria and viruses. Mechanisms for contending with these threats are well documented; however, the regulation of these diverse defense elements remains poorly understood. Here we describe a genome-wide, coordinated, and highly effective immune response, termed GUARD, that protects against bacterial and viral threats using a single regulatory pathway. Bioinformatic analyses reveal a Pseudomonas-wide form of the Gac/Rsm regulatory pathway (GRP), an established danger sensing system in P. aeruginosa. Proteomic studies of diverse Pseudomonas species show that the pathway regulates a large and variable suite of factors implicated in defense against both bacterial and phage threats. Focusing on P. protegens, we identify profound phenotypic consequences of these factors against multiple forms of bacterial antagonism and several phage. Together, our results reveal that bacteria, like multicellular eukaryotes, couple danger sensing to the activation of an immune response with antibacterial and antiviral arms.

In brief

Brinkley et al. describe a general immune response strategy in Pseudomonas species, termed GUARD, that protects against both bacterial and viral threats. This program, controlled by the Gac/Rsm global regulatory pathway, activates a variable suite of mechanisms that together provide robust defense in response to a general danger signal.

Graphical Abstract

Introduction

For most bacteria, survival is dependent on the ability to withstand myriad biotic threats, including antagonism from phage and other bacteria. The selection imposed by these threats has led to the evolution of diverse and widespread defensive mechanisms, exemplified by the many phage defense pathways uncovered in recent years¹. Though mechanistically distinct, these pathways employ common strategies, including blocking phage attachment, inhibiting the intracellular life cycle of the phage, or trigging programmed cell death².

Interbacterial antagonism can take many forms, ranging from the production and delivery of overtly toxic molecules to more passive competition strategies that restrict nutrients³. Correspondingly, defense against interbacterial antagonism can present as a diffuse collection of cellular functions. However, recent work studying discrete and mechanistically defined toxin delivery systems, in particular the type VI secretion system (T6SS), has led to the identification of factors with specialized functions in defense against interbacterial antagonism. These include the production of immunity proteins that inhibit the activity of incoming toxins, extracellular barriers that block toxin delivery, pathways that repair cellular damage caused by antagonism, and counter-attack strategies^4–7.

Defense factors are costly and the threats they protect against, though omnipresent, occur inconsistently; therefore, it stands to reason that cells might tightly regulate defense factor production⁸. The regulation of phage defense systems is perhaps most extensively documented at the posttranslational level. One common mechanism employs cyclic nucleotides, which are generated upon detection of infection or by active defense systems themselves, and then activate assorted responsive effector proteins⁹. An additional regulatory mechanism common to a number of phage defense system is transcriptional repression, including by WYL domain proteins and the CapH regulator of the CBASS phage defense system ^10–13. These proteins maintain low levels of expression of cell-death inducing defense pathways that appear to be sufficient to provide phage defense without need for further induction^10–13. Beyond these regulators, relatively few transcriptional and posttranscriptional mechanisms of phage defense system regulation have been identified, and it is generally thought that activation at the level of machinery biogenesis fails to keep pace with the rapid time scale of infection by many phage¹⁴. Indeed, quorum sensing systems are the most widely characterized class of transcriptional regulators for phage defense systems^15,16. By virtue of extrinsic sensing of cell density, these systems permit anticipatory behavior that links high cell density to an increased likelihood of encountering a phage.

Reflecting their multifaceted nature, the regulation of interbacterial antagonism defense factors is complex. Toxin delivery systems and the effectors they transport can disrupt cell wall integrity, and studies in V. cholerae and E. coli suggest that envelope stress sensing pathways could be important transcriptional activators of antagonism defense genes^17,18. Like phage infection, certain types of interbacterial antagonism can occur on a time scale too rapid for antagonized cells to respond directly. Therefore, it is not surprising that quorum sensing is also a frequently cited regulator of interbacterial defense functions³. Quorum sensing can activate the production of extracellular structures that serve as barriers, and it also generally promotes the production of antimicrobials¹⁹. However, because quorum sensing responds to cell density rather than damage, it is an imperfect predictor of danger.

Given that common signals may predict a multitude of threats, one can envision that coordinated regulation of defense factors would be prevalent. Such a strategy is suggested by the tendency of defense genes to aggregate, such as in mobile elements containing both phage defense gene clusters and T6SS toxin genes²⁰. However, outside of these situations, little is known about how genetically unlinked defense pathways are co-regulated. In P. aeruginosa, we previously reported the discovery of a pathway, termed the P. aeruginosa response to antagonism (PARA), which coordinates the induction of a broad array of interbacterial antagonism defense mechanisms in response to kin cell lysate. Together, these co-regulated factors provide multiple logs of fitness to populations confronted with highly antagonistic competitors (Figure 1A)^5,21. PARA is mediated by the Gac/Rsm posttranscriptional regulatory program, a pathway conserved across γ-proteobacteria and reported to regulate a wide range of traits^22–24. At the core of this pathway is the sensor kinase GacS, which phosphorylates the response regulator GacA²⁴. Phosphorylated GacA activates the transcription of small RNAs, which sequester the RNA binding protein RsmA (and orthologs) from target mRNAs, thereby increasing translation. Altogether, the Gac/Rsm pathway (GRP) controls expression levels of more than 200 proteins in P. aeruginosa²⁵. GRP targets in this species relevant to antagonism defense include the components of an antibacterial T6SS, three interbacterial antagonism defense gene clusters (antagonism resistance clusters 1–3; Arc1–3), and the extracellular polysaccharide Psl^5,25,26.

Figure 1. Pseudomonas species encode an expanded Gac/Rsm pathway.

A) Diagram depicting PARA (left) and the different configurations of the Gac/Rsm pathway described in the literature (right). Note that only one RsmA homolog and two small RNA molecules (RsmY and RsmZ) are depicted for simplicity; the number of homologs of these components varies across genomes. B) Concatenated marker gene-based phylogeny of Pseudomonas species and representatives from other γ-proteobacterial lineages indicating GRP components identified for each in our analyses. Also see Figure S1.

In this study, we show that the GRP of diverse Pseudomonas species regulates a variable suite of defenses against both interbacterial antagonism and phage infection. We experimentally demonstrate the central importance of the GRP in mediating defense against multiple forms of interbacterial antagonism and show that activation of the pathway provides potent protection against infection by phage. Our results support a model in which this defense program, which we term Gac/Rsm-regulated unified antagonism response and defense (GUARD), mediates a highly effective response for countering both bacterial and viral threats across Pseudomonas species.

Results

Pseudomonas species possess a unique form of the Gac/Rsm pathway

The Gac/Rsm signaling pathway (GRP) is widely distributed amongst γ-proteobacteria²⁴. Despite some level of characterization in many of these organisms, only in P. aeruginosa has the pathway been implicated in defense against interbacterial antagonism^5,21,27. Notably, the GRP of P. aeruginosa is augmented by the sensor kinases RetS, LadS, and RskA. We previously showed that RetS repression of GacS modulates the activity of the pathway during antagonism, and both LadS and RskA counter the activity of RetS (Figure 1A)^21,25,28,29. Given these observations, we sought to define the distribution of both the core components of the Gac/Rsm signaling system and the modulatory sensor kinases of P. aeruginosa.

We identified homologs of GacS, GacA, RsmA, RetS, LadS, and RskA in representative genomes across the major proteobacteria lineages. We then subjected homologs to phylogenetic analysis, allowing us to define candidate orthologous sequences across genomes (Figure S1). Finally, we compared the domain architecture of candidates to the corresponding P. aeruginosa proteins, permitting a confident determination of orthologs. As previously reported, we found the core Gac/Rsm proteins GacS, GacA, and RsmA are widely distributed across γ-proteobacteria and absent outside of this class (Figure 1B)³⁰. By contrast, our analysis showed that the accessory kinases RetS, LadS, and RskA are strictly confined to organisms belonging to the genus Pseudomonas. Although retS, ladS and rskA are not genetically linked, with only one exception (P. stutzeri), the genes universally co-occurred within the Pseudomonas spp we queried. Together, these data show that Pseudomonas spp broadly possess a unique version of the Gac/Rsm pathway that is potentially capable of mediating sensing of and response to antagonism.

Mass spectrometry of regulatory mutants sensitively defines the GRP regulon

We previously demonstrated that the critical role of the P. aeruginosa GRP in defense against interbacterial antagonism derives from the cumulative effect of multiple individual factors under its control⁵. Here, we sought to determine how the GRP regulon varies across a phylogenetically diverse cross-section of Pseudomonas species. We hypothesized that the identity of factors under GRP control in these species would provide general insights into its function beyond P. aeruginosa.

To our knowledge, there has not been an effort to apply consistent methodology to experimentally define the GRP regulon across multiple species of Pseudomonas. However, groups have undertaken various approaches to define the regulon in individual species. These include bioinformatic analyses to predict transcripts bound by RsmA, RsmA ChIP-seq and related methods, and transcriptome measurements ^25,31–36. As a post-transcriptional regulatory pathway, the GRP regulon is arguably most appropriately characterized at the protein level. Only a small number of studies have reported such data, and these compared protein expression between GRP-inactive (e.g. ΔgacS) and wild-type strains^27,34. However, our prior data suggest that activation of the pathway under in vitro growth conditions requires an antagonizing organism or deletion of a negative regulator (e.g. ΔretS)^21,37. We therefore reasoned that in the absence of antagonism, GRP-regulated factors might be most sensitively identified in a given Pseudomonas species by comparing the proteome of a GRP-inactivated strain to that of one in which a negative regulator of the pathway is removed. To test this, we compared whole cell proteomes of P. aeruginosa wild-type, ΔgacS, and ΔretS strains using principal component analysis^21,38. Along PC1 (42% of variance), ΔretS proteomes diverged substantially from wild-type and displayed a greater distance from ΔgacS than did the wild-type (Figure 2A). We conclude that GRP regulons of Pseudomonas spp can be most comprehensively defined by comparing protein abundances between ΔretS and ΔgacS genetic backgrounds.

Figure 2. Proteomic analysis of the GRP regulon of diverse Pseudomonas species.

A) Principal component analysis of MS-derived whole-cell proteomes (n=10) from the indicated strains of P. aeruginosa. B) Comparison of the fold change in abundance between P. aeruginosa ΔretS and ΔgacS strains for GRP targets as measured by transcript (X-axis, derived from Goodman et al.²⁵) or protein abundance (this study). C) Plot showing most differential previously undetected GRP targets in P. aeruginosa by rank fold change (left) and schematic illustrating antimicrobial and defensive functions associated with a subset of these targets. D-F) Volcano plots comparing differences in abundance of proteins detected by MS for ΔretS and ΔgacS strains of P. protegens (C), P. fluorescens (D) and P. protegens. Average fold-change and p-values were calculated from 5 biological replicates. Also see Tables S1 and S2 and Figure S2.

We began our investigation of the GRP regulon in Pseudomonas spp. by benchmarking our P. aeruginosa dataset against two prior transcriptomic studies^25,33. A strong correlation in abundance between significant targets was observed with both datasets, and consistent with expectations, a transcriptomic study comparing GRP-activated and -inactivated backgrounds more closely resembled our findings than did the study employing a wild-type reference (Figure 2B, Figure S2A). In total, our determination of the GRP regulon of P. aeruginosa supports prior data implicating the pathway in antimicrobial production and antagonism defense. Notable antimicrobial GRP targets confirmed by our data include the antibacterial H1-T6SS and 1-undecene biosynthesis proteins, whereas antagonism defense targets include Arc pathway components and Psl exopolysaccharide biosynthesis proteins²⁵.

In addition to the previously identified targets in our dataset, we found 72 GRP-regulated proteins not identified in prior genome-wide studies (Table S1, Table S2). Lending credence to the validity of these proteins as bona fide components of the GRP regulon, the group includes multiple proteins defined as GRP targets in focused investigations. Among these are the stationary phase sigma factor RpoS, the quorum sensing signal synthase LasI, and genes under the control of the transcription activator BexR^39–41. Ranking the previously undescribed GRP targets within our data by induction level revealed that a preponderance of those most strongly induced by the pathway represent additional antimicrobial and antagonism defense functions (Figure 2C). For instance, KynA, B and U generate an essential precursor of quinolone biosynthesis⁴². Several quinolones of P. aeruginosa are antimicrobial, and one of these, PQS (Pseudomonas quinolone signal), further stimulates production of the antimicrobial pyocyanin via quorum sensing^43,44,45. We also identified the enzyme responsible for biosynthesis of the siderophore 2,3-dihydroxybenzoic acid⁴⁶. Siderophores are well known to function in microbial competition by depleting accessible iron from neighboring microbes⁴⁷. Previously undescribed GRP targets we identified with defensive functions include two that counteract the effects of antibacterial toxins: a homolog of RtcB that repairs RNA molecules cleaved by RNase toxins and a NADAR (NAD- and ADP-ribose-associated) family protein, members of which remove ADP ribose moieties installed by ADP-ribosylating toxins on nucleic acids^48–51. Other defensive factors include a predicted restriction endonuclease (PA0820) and a homolog of the DNA replication inhibitor CspD, which promotes persister cell formation⁵². Notably, we identified the phage defense system Gabija among the most strongly induced, previously undescribed GRP targets of P. aeruginosa. This raised the intriguing possibility that the defensive function the GRP could extend to viruses. Together, these results validate our proteomics-based approach and significantly expand our understanding of the GRP regulon in P. aeruginosa. Furthermore, they strongly support the contention that the GRP of this bacterium controls factors that constitute a multifaceted response to a range of antagonistic threats.

Interbacterial antagonism and phage defense mechanisms dominate the GRP regulon in diverse Pseudomonas species

Given the insights garnered by our MS-based interrogation of the GRP regulon of P. aeruginosa, we applied the strategy to a diverse cross-section of other Pseudomonads. We selected P. protegens, P. fluorescens, and P. putida for these studies, primarily owing to their phylogenetic divergence from each other and from P. aeruginosa; however, these species also inhabit a range of distinct ecological niches, suggesting exposure to unique sets of biotic threats^53,54. Application of our proteomics-based approach to these species revealed extensive proteomic remodeling by the GRP pathway, similar in magnitude to that observed in P. aeruginosa, as well as substantial differences between the proteomes of each species (Figure 2D-F and Figure S2B-D). Relatively few proteins are under GRP control in all four species. Notably, all but one of these shared GRP targets represent functions related to interbacterial antagonism or defense. These include components of the H1-T6SS and the outer-membrane stress resistance protein LptF⁵⁵.

Beyond conserved GRP targets, we found extensive variability among regulated targets across the four species investigated. However, as in P. aeruginosa, the GRP regulon of each species encompassed a preponderance of functions related to antimicrobial production or defense against antagonism. For instance, the GRP regulons of both P. protegens and P. fluorescens include the biosynthetic machinery for production of numerous secondary metabolites, many of which have antimicrobial activity. Consistent with prior studies, we found the proteins responsible for production of 2,4-DAPG, rhizoxin, protegenin, pyoluteorin and orfamide in P. protegens are positively regulated by the GRP (Table S1, Table S2)⁵⁶. Expression of a candidate orfamide biosynthetic pathway is also activated by the GRP in P. fluorescens, as well as proteins belonging to an uncharacterized non-ribosomal peptide synthesis (NRPS) pathway (PFLU3_17930-PFLU3_17980) and the biosynthetic pathways for phenazine and azetidomonamide. Defensive functions under GRP control in these species include production of the extracellular polysaccharides Psl and Peb in P. protegens and P. putida, respectively, multiple universal stress protein homologs in P. putida, and periplasmic proteins with structural similarity to BepA in P. putida and P. protegens. This protein facilitates maturation of the BAM complex and degrades misfolded outer-membrane proteins during stress, promoting cell envelope integrity⁵⁷. An additional functional category we found is common to the GRP regulons of multiple species are proteins involved in catabolism of non-preferred carbon sources, such as methyl-branched compounds, branched chain α−keto acids and fumarate. We speculate that activation of such pathways during antagonism could indirectly contribute to fitness by allowing Pseudomonas species to utilize nutrients that are inaccessible to their competitors.

Interestingly, we identified phage defense systems within the GRP regulons of all four Pseudomonas spp analyzed. These achieved statistically significant association with the GRP regulons of P. protegens (p = 0.013), P. fluorescens (p = 0.012), and P. putida (p <0.001), where 4 of 7, 10 of 14, and 6 of 7 detected phage defense systems, respectively, are GRP-regulated (Figure 2D-F and Table S2). Components of several phage defense systems, including PARIS and BstA in P. protegens and Shango and Kiwa in P. fluorescens, are above the 85^th percentile of GRP induction level in these bacteria. This finding, taken together with the multitude of GRP-regulated antibacterial and antifungal pathways established by our MS-based approach, strongly suggests that the GRP regulon constitutes a central immune response with arms targeting each major category of biotic threat.

Pseudomonas sp. broadly employ Gac/Rsm in antagonism defense

The GRP regulons we defined in P. putida, P. protegens, and P. fluorescens are consistent with the pathway broadly mediating defense against microbial antagonism beyond P. aeruginosa. As a first experimental test of this, we measured the benefit of GRP activity in our panel of Pseudomonads under conditions of interbacterial antagonism by the T6SSs of Burkholderia thailandensis (B. thai) and Enterobacter cloacae. In competition with B. thai, ΔgacS strains of each species tested exhibited severe antagonism-dependent fitness defects (3–6 log, Figure 3A). With the exception of P. fluorescens, the ΔgacS strains of each species also displayed pronounced antagonism defense defects against E. cloacae. Importantly, the GRP-inactivated strains did not display a growth defect in isolation or in co-culture with antagonism-deficient competitors (Figure S3A,B). Prior work suggests that outside of Pseudomonas, strains bearing GRP-inactivating mutations can exhibit marked in vitro growth defects^58–61. We found an exception to this pattern is Vibrio parahaemolyticus, which allowed us to evaluate the contribution of the pathway to antagonism defense in a bacterium lacking the accessory kinases without this confounding variable (Figure S2C). Although antagonized effectively by the T6SSs of both B. thai and E. cloacae, we did not observe an antagonism-dependent fitness defect of V. parahaemolyticus ΔgacA against these competing organisms (Figure 3A, Figure S3D). Our experiments reveal that the specialized GRP of Pseudomonas is uniquely adapted for defense.

Figure 3. The GRP plays a general and central role in defense against interbacterial antagonism.

A) Mean relative fitness (±SD) of the wild-type strain for each species during T6SS-mediated antagonism by the indicated competitors, relative to the fitness of the corresponding GRP-inactivated mutants. Asterisks indicate strain pairings in which we observe a significant antagonism-dependent fitness difference between the wild-type and ΔgacS strains. (Welch’s T-test, p<0.05 with BH correction for multiple comparisons, n=3). B-E) Transposon library sequencing-based comparison of the fitness contribution of individual P. protegens (B-D) or V. parahemolyticus (E) genes during growth competition with wild-type versus antagonism deficient B. thai (B), E. cloacae (C, E) or L. enzymogenes (D). E. cloacae and B. thai ΔT6SS strains contain deletions of tssM^21,80; L. enyzomogenes ΔT4SS contains a deletion of virD4. GRP signaling components (GacA, GacS) indicated in B-E, and GRP-targets that contribute to defense against one or more antagonist are highlighted in B-D. Reads, normalized counts of transposon sequencing reads mapping to an given gene following growth competition assays with the indicated strains. F) Contribution of the indicated pathways to defense against different competitors. Average normalized transposon insertion count ratios (antagonism deficient competitor/WT competitor, ±SD ) were obtained using all genes within a given pathway, and replicate Tn-seq datasets when available. For Arc3 and PFL_5124, a single ratio measurement is provided for defense against L. enzymogenes, as a single Tn-seq screen was performed with that antagonist. Asterisks indicate ratios significantly different from 1 (1-sample t-test, p<0.05). (G) Representative LPS profiles of the indicated P. protegens strain, normalized by OD₆₀₀, run on an 8–16% gradient polyacrylamide gel, and visualized by silver stain. Due to the sensitivity of silver staining, OBC4 and Lipid A core regions shown are taken from separate biological replicates. (H) Densitometry analysis of OBC4 O-polysaccharide from silver-stained gels, normalized to wild-type levels (Welch’s t-test, n=3, p<0.05). (I) Relative survival of P. protegens ΔOBC4 in a mixed-recipient competition with the wild-type strain and the indicated strains of E. cloacae, B. thai, or L. enzymogenes, as determined by qPCR. Asterisks indicate significant changes in surviving mutant abundance (Welch’s t-test with BH correction, n=6–9, p<0.05). Also see Table S3 and Figure S3.

The GRP is the major defense determinant for P. protegens against multiple mechanisms of interbacterial antagonism

We next sought to contextualize the impact of GRP inactivation on defense against antagonism by comparing its contribution relative to other, potentially GRP-independent, defense mechanisms. To accomplish this in a comprehensive and unbiased manner, we conducted genome-wide transposon insertion sequencing (Tn-seq) screens wherein a P. protegens transposon insertion library was subjected to co-cultivation with wild-type and antagonism-deficient competitor bacteria. In these screens, clones harboring transposon insertions in genes important for fitness during antagonism are selectively depleted from the population during co-cultivation with antagonistic bacteria. Given our results above, selected competitor bacteria included B. thai, E. cloacae, and their corresponding T6SS-inactivated derivatives. To capture a broader range of antagonistic mechanisms, we also included Lysobacter enzymogenes and a derivative of this strain bearing an inactivated Xanthomonadales-like type IV secretion system (X-T4SS). Unlike conjugative T4SSs, this apparatus is specialized for the delivery of large cocktails of antibacterial proteins⁶². Notably, these toxins are distinct from those delivered by T6SSs⁶³. Strikingly, gacA and gacS, showed the strongest antagonism-dependent fitness contribution across all three competitors, with average selection ratios of 178 and 188, respectively (Figure 3B-D, Table S3). This finding is consistent with prior results showing that the P. aeruginosa GRP plays a similarly critical role in defense against T6SS-based antagonism^5,21. We expect the GRP is fully activated under the extended period of incubation at high population density employed in our screens, explaining why inactivation of the negative GRP regulators RetS and RsmA does not confer a fitness benefit. To compare the relative contribution of GRP and non-GRP defense mechanisms between Pseudomonads and a bacterium possessing only the basic GRP, we performed a similar selection using a transposon library of V. parahaemolyticus co-cultured with E. cloacae. Consistent with our findings using individual mutations, V. parahaemolyticus gacA and gacS mutations were not under selection duri ng antagonism (Figure 3E). Rather, genes belonging to the K-antigen biosynthesis pathway constituted the majority of those whose function was most highly selected by antagonism. Selection of these genes suggests a defensive strategy solely reliant on blocking the T6SS via modification of the cell surface rather than a coordinated regulatory response^7,64. Together, these genome-wide screens highlight the specialized and essential role that the extended GRP of Pseudomonads plays in antagonism defense.

In-line with the strong selection for the core GRP signaling elements in P. protegens, we found its GRP regulatory targets overrepresented among genes contributing to fitness during antagonism. Against all three antagonists, genes encoding essential components of the T6SS system constituted many of the most critical survival determinants (Figure 3F). Other GRP targets we identified as important for fitness during antagonism varied depending on the antagonizing organism, suggesting their functions may be specialized to provide protection from specific toxins or delivery mechanisms. For instance, we found that Arc3 provides P. protegens defense against B. thai and L. enzymongenes, whereas the pathway is dispensable during antagonism by E. cloacae (Figure 3F). Arc3 specifically grants defense against Tle3 family phospholipases, which B. thai, but not E. cloacae, possesses^5,65,66. Based on its genome sequence, L. enzymogenes encodes multiple antibacterial T4SS effectors, at least one of which we found is a phospholipase that could be a Tle3 family member⁶⁷.

Additional GRP targets we found to have antagonist-specific contributions to fitness include PFL_5124 and genes within the O-polysaccharide (O-PS) biosynthesis cluster 4 (OBC4), responsible for production of a long version of the polymer⁶⁸. The closest structural homolog of PFL_5124 is BepA (E-value, 1.8 × 10⁻⁷), a periplasmic protease that, as noted above, acts to promote membrane integrity by degrading misfolded proteins that accumulate in the periplasm during stress⁵⁷. Our Tn-seq screens indicated this protein specifically contributes to defense against B. thai, which encodes two T6SS effectors predicted to act in the periplasm⁶⁹. Pairwise competition assays employing an in-frame, unmarked deletion of PFL_5124 confirmed its contribution to defense against antagonism by B. thai and not E. cloacae (Figure S3E). Given that LPS production was not linked to the GRP prior to our proteomics analysis, we investigated the impact of pathway activation on long O-PS production. Comparison of LPS profiles of P. protegens ΔgacS and ΔretS strains revealed substantially less of the OBC4 polysaccharide in the GRP-inactive background (Figure 3G,H). To further evaluate the role of the OBC4 pathway in defense, we performed growth competition assays between wild-type and ΔOBC4 strains against E. cloacae, B. thai and L. enzymogenes. Consistent with our Tn-seq data, OBC4 contributed to P. protegens fitness specifically during antagonism by E. cloacae (Figure 3I).

We noted that the magnitude of the competitive defect derived from GRP-inactivating insertions is substantially greater than those that inactivate any single pathway in its regulon, and indeed greater than the predicted cumulative deficit of all the GRP-regulated pathways we hit (Figure 3F). One likely explanation for this finding is that, with many different GRP-targets acting together to promote fitness during antagonism, few individual factors make a substantial enough impact to be detected in our screens. However, we considered two additional factors that could be contributing. First, under the contact-promoting conditions of our screen, the strong fitness contribution of the T6SS could be masking other GRP-regulated defense functions. Additionally, a substantial component of the characterized GRP regulon across Pseudomonas species consists of antimicrobial secondary metabolite biosynthesis machineries. These pathways may well contribute to defense, but under the conditions of our screens, mutations within them would be complemented in trans by other clones in the population.

To evaluate whether the activity of the T6SS could be masking the importance of other defense pathways, we repeated our Tn-seq screen using a P. protegens strain lacking T6SS activity. For this experiment, we generated a transposon mutant library in P. protegens ΔtssM and selected this library, as previously, in the presence and absence of T6SS-based antagonism by E. cloacae. As expected for this genetic background, T6SS genes did not show selection during antagonism (Figure S3F). On the contrary, gacA and gacS genes remained the most critical defense determinants, underscoring the importance of non-T6SS-regulated functions in defense. Further analysis of these data revealed several GRP regulon defense factors not identified in our prior screens employing the wild-type P. protegens background (Figure S3G). These include maf_1 and an uncharacterized toxin-antitoxin module (PFL_0652–0653). Maf proteins exhibit diphosphatase activity against nucleotide triphosphate and are implicated in maintaining genome integrity through degradation of non-canonical, potentially mutagenic bases⁷⁰. Toxin-antitoxin systems are widely implicated in resistance to phage, yet to our knowledge these modules are not known to contribute to defense against interbacterial antagonism.

We next assessed the possibility that secondary metabolite production represents a second class of GRP targets important for defense that were not detected in our Tn-seq screen. Specifically, we asked whether production of the antibacterial and antifungal molecule 2,4-DAPG benefits P. protegens during antagonism. We inactivated the 2,4-DAPG biosynthetic pathway via in frame deletion of phlD, and subjected this strain to growth competition assays. P. protegens ΔphlD exhibited a significant fitness defect during antagonism with B. thai, reflecting the susceptibility of this species to synthetic 2,4-DAPG (Figure S3H,I). These results support our hypothesis that antimicrobial production constitutes an important arm of the GRP-regulated immune program. Together, our Tn-seq data underscore the central importance of the GRP in coordinating a multifaceted defense against interbacterial antagonism.

The GRP of P. protegens provides defense against phage

The presence of multiple phage defense pathways within the GRP regulons of diverse Pseudomonas species led us to hypothesize that GRP innate immune capacity extends beyond interbacterial antagonism to include a second arm dedicated to countering viral threats. To test this, we isolated lytic phages from a variety of rhizosphere soil samples using the predicted phage-sensitive strain P. protegens ΔgacS as bait. Whole genome sequencing and phylogenetic analysis revealed that our collection included six distinct, tailed double-stranded DNA phages from the class Caudoviricetes, most closely related to the Pseudomonas phages vB_PpuP-Luke-3 and PollyC. We then assessed plaquing efficiency of these phages on P. protegens GRP activated versus inactivated strains. Strikingly, we found that activation of the GRP significantly restricted the growth of each phage (Figure 4A,B).

Figure 4. The GRP and defense factors under its control contribute to phage resistance.

A) Plaquing efficiency of the indicated phages on GRP inactive (ΔgacS) or activated (ΔretS) P. protegens. B) Ratio of pfu obtained from the indicated phages on P. protegens ΔgacS vs ΔretS. Asterisks indicate ratios significantly different from one (1-sample t-test, n=8, P<0.05). ND, not determined. C) Relative GRP activation level as measured by luminescence from a nanoluciferase reporter for rsmZ transcription in P. protegens after 2 hr incubation with Ppr_SeaP2 (MOI = 10⁻³) or a buffer control. 0 and 100% GRP activation defined by luminescence in ΔgacS and ΔretS backgrounds, respectively. Asterisks indicate significant differences in luminescence (one-tailed Welch’s t-test, n=6, p<0.05). D) Plaquing efficiency of Ppr_SeaP2 on the indicated strains of P. protegens. E) Quantification of the plaquing efficiency of Ppr_SeaP2 on the indicated mutants of P. protegens in a GRP activated (ΔretS), inactivated (ΔgacS), or wild-type genetic background. Asterisks indicate significant differences in PFU yield between strains (one-way ANOVA followed by Holm-Sidak post hoc test, n=5–12, p<0.05). F) Representative images of Ppr_SeaP2 plaques on the indicated P. protegens strains (left) and quantification of plaque areas on the indicated strains relative to a parental strain (right). Asterisks indicate significant difference in plaque sizes in the indicated strain relative to the wild-type (one-way ANOVA followed by Dunnett’s T3 test). G) Model of GUARD in Pseudomonas species. Two arbitrary Pseudomonas species are depicted to exemplify the diversity of possible threats and GRP-mediated defense responses. In each case, a common danger signal (lysate from neighboring kin cells) activates defenses that protect against multiple forms of interbacterial antagonism (e.g. T4SSs and T6SSs) and infection by different phages. Defenses display variable degrees of conservation across the genus. For example, the T6SS (orange) is a conserved GRP constituent, while GRP-regulated phage defenses (salmon, mauve, pink, dark green), and many defenses against interbacterial antagonism (blue, light green) are variable. Also see Figure S5.

If the anti-viral arm of a GRP innate immune program operates similarly to its antibacterial defenses, then we expect that phage infection of a subset of a population should lead to GRP induction in neighboring cells. To test this prediction, we generated a transcriptional reporter for GRP activation in which the gene encoding nanoluciferase was placed downstream of the rsmZ promoter and inserted at a neutral chromosomal location in P. protegens (Figure S4A). We confirmed sensitivity of this reporter to GRP activation by introducing it into RetS- and GacS-inactivated strains of P. protegens (Figure S4B). Upon infection of wild-type P. protegens with the highly GRP-sensitive phage Ppr_SeaP2, we observed reporter activity at levels comparable to the ΔretS strain, demonstrating that robust GRP activation occurs during phage infection (Figure 4C).

The phage defense pathways we identified as most strongly induced upon GRP activation in P. protegens are BstA and PARIS. To determine whether the restriction of phage replication we observed upon GRP activation is attributable to these pathways, we generated in-frame deletions of each pathway in both GRP-activated and inactivated backgrounds of P. protegens. We found that in P. protegens ΔretS, inactivation of BstA and PARIS significantly increased plaquing efficiency by phage Ppr_SeaP2 (Figure 4D,E). Notably, the contribution of BstA and PARIS to phage resistance requires GRP activation, as inactivation of these pathways did not affect sensitivity of P. protegens ΔgacS to Ppr_SeaP2 infection (Figure 4E). Additionally, resistance to Ppr_SeaP2 is restored to the P. protegens ΔretS defense pathway mutants when complemented by the relevant pathway under basal expression from the pBAD promoter at an ectopic location on the chromosome (Figure 4D,E). For the other phages tested, inactivation of these pathways did not affect the resistance conferred by RetS inactivation (Figure S4C). We speculate that for these phages, the GRP could negatively regulate a receptor, or that multiple defense pathways could be working in concert to provide resistance, similar to our observations with antibacterial defenses under GRP control.

In the experiments described above, we demonstrate a role for GRP-regulated phage defenses when the pathway is genetically activated by deletion of retS. However, we found previously that in wild-type populations, activation of the GRP requires killing and lysis of a subpopulation of Pseudomonas cells by bacterial antagonists. If this pattern extends to phage infection, as our rsmZ reporter data suggest, we would expect that GRP-regulated phage defenses would make little contribution to phage resistance upon initial infection of naïve, wild-type populations, but they would instead play a protective role at later timepoints during infection. Consistent with these predictions, we observe significantly increased Ppr_SeaP2 plaque size upon inactivation of PARIS or BstA in a wild-type P. protegens background (Figure 4F, Figure S4D), while the number of plaques formed is unaffected (Figure 4E). As a corollary, if induction of GRP-regulated phage defense systems requires an initial round of infection and lysis, constitutive expression should result in decreased plaque formation in a wild-type background. Indeed, expression of PARIS from the pBAD promoter with arabinose (0.01%) yields significantly reduced plating efficiency compared with a wild-type strain and reintroduction of bstA with arabinose induction abrogates plaque formation (Figure S4E). Arabinose-induced expression of PARIS also restricts plaque size to wild-type levels (Figure 4F); the lack of plaques formed upon BstA complementation precludes plaque size analysis of this strain. These results support a model in which phage infection of naïve P. protegens activates the GRP in neighboring cells, which restricts further phage propagation. Altogether, our findings are consistent with the GRP of P. protegens protecting against phage infection via coordinated regulation of phage defense systems, akin to its role in defense against interbacterial antagonism.

Discussion

In this study, we show that the GRP of diverse Pseudomonas species coordinates the regulation of pathways for defense against both interbacterial antagonism and phage. Although Pseudomonas species occupy diverse niches, encounter unique biotic threats, and possess distinct sets of defense factors, we show that the expanded GRP serves as a central mediator of innate defense across the genus. We term this defense program Gac/Rsm-regulated Unified Antagonism Response and Defense (GUARD, Figure 4G). The coordinated regulation of antibacterial and antiviral defenses has precedence in metazoa; type I interferon signaling in mammals, for example, controls the activation of a wide range of immune responses, including both antiviral and antibacterial mechanisms⁷¹. The convergent evolution of mechanisms for coordinating responses to disparate threats begs the question as to the adaptive value of such a coupling. Both interferon signaling and GUARD provide a mechanism by which a population of cells can become alerted to the presence of a wide range of threats, before directly encountering the threat itself. In the former case, this is achieved via the numerous receptors for the detection of diverse pathogen-associated molecular patterns that trigger secretion of a common signaling molecule⁷¹. The GRP, in contrast, is directly activated by the release of cell lysate, a signal that is generated in the course of both interbacterial antagonism and infection by lytic phages²¹. In both situations, by coupling activation of many pathways to a single, diffusible signal, populations gain the ability to anticipate a broad array of attacks and mount a general response, thus sidestepping the limitations of linking defense pathway activation to direct detection of incoming threats within an individual cell. An inevitable consequence of linking the regulation of defenses towards disparate threats under a single regulatory program is that a subset of activated defenses will not be effective against certain triggering threats (e.g activation of the T6SS will provide no direct benefit during phage infection). However, in the case of GUARD, this cost appears to be outweighed by the benefits of coordinate, anticipatory regulation of these pathways.

Our study revealed that the suite of defensive responses within the GRP regulon varies substantially across Pseudomonas species. Recognition of the full extent to which these regulons vary was facilitated by our development of a sensitive and comprehensive method for their characterization, which involves comparative proteomic analysis of GRP-activated and inactivated strains. A particular advantage of this approach over some earlier methods is its ability to capture both direct and indirect targets of the pathway^25,33. The variability we uncovered in the GRP regulon suggests that the specific threats encountered by a given species could be driving the diversification of defensive mechanisms it can produce. This is supported by the observation that different GRP targets in a given species can be important for defense against different antagonists, as observed in studies of P. aeruginosa^5,27.

The widespread conservation of the GRP across the Pseudomonas genus has long been at odds with theories put forth regarding its adaptive function in different contexts. In P. aeruginosa, for example, the GRP negatively regulates type III secretion, required for acute infection, and positively regulates traits associated with chronic infections, such as extracellular polysaccharide production^26,72. Accordingly, researchers argued it mediates the switch between acute and chronic infection-causing lifestyles²⁵. In contrast, in plant protective Pseudomonas species, where the GRP is linked to regulation of secondary metabolites important for biocontrol of plant pathogens, it was proposed to act as a quorum sensing pathway⁷³. Our finding that the GRP coordinates induction of diverse defensive mechanisms in response to antagonism suggests a clear adaptive benefit of the pathway that is relevant to the physiological conditions shaping the evolution of the genus.

Reinterpreting previous work in light of our findings, many prior observations hint at the importance of the GRP in defense against antagonism. For example, early studies on P. syringae showed that inactivation of GacS had no impact on its propagation when applied to leaf surfaces under greenhouse conditions, but in field studies, where successful plant colonization requires contending with many competitors, gacS mutants were highly impaired^74,75. Similarly, a previous study found GRP-deficient mutants of P. protegens have decreased fitness during interspecies competition, but grow as well as the wild-type in pure culture⁷⁶. Prior studies have also contributed data consistent with the involvement of the GRP in phage defense^77,78. For example, a group found that in the presence of exogenous putrescine, GRP-regulated polyamine uptake increases resistance of P. aeruginosa to several phage⁷⁷. Our results, taken together with these prior findings, strongly support the contention that a key adaptive function of the GRP is to coordinate defensive responses.

The clear benefits conferred by linking defensive mechanisms under a common regulatory pathway raises the question of whether this phenomenon occurs more broadly. While our work suggests that GUARD requires an expanded version of the GRP pathway that is limited to Pseudomonas species, the possibility that different accessory kinases could be performing an analogous role in other γ-proteobacteria remains unexplored. Indeed, a recent study employing Serratia and Pectobacterium strains demonstrated GRP-mediated regulation of CRISPR-Cas systems, suggesting the GRP may have an immune-related role in these organisms⁷⁹. We speculate that pathways coordinating defensive responses in other bacterial phyla may well await discovery.

Resource Availability

Lead Contact

Requests for additional information should be directed to the lead contact, Dr. Joseph Mougous, mougous@uw.edu.

Materials Availability

All strains, plasmids, and phages described in this paper are available upon request from the lead contact, Dr. Joseph Mougous. Phage sequences and NGS reads have been deposited under BioProject PRJNA1267186.

Data and Code Availability

All information required to reanalyze the data reported in this paper are available upon request from the lead contact, Dr. Joseph Mougous. This paper does not report original code.

STAR Methods

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Bacterial strains and culture conditions

All strains used in this study are described in Table S4 and are available upon request. Strains were derived from P. aeruginosa PAO1⁸¹, P. putida KT2440⁸², P. protegens Pf-5⁸³, P. fluorescens 2–79⁸⁴, E. cloacae ATCC 13047⁸⁵, B. thai E264⁸⁶, L. enzymogenes C3–1⁶⁷, V. parahaemolyticus RIMD 2210633⁸⁷, and E. coli MG1655⁸⁸. E. coli strains DH5α (Thermo Fisher Scientific) and EC100D pir+ (Lucigen) were used for cloning and maintenance of plasmids. E. coli strains S17–1 λpir⁸⁹, Sm10 λpir (Biomedal Lifescience, Cat# BS-3303), HB101⁹⁰, and RHO3⁹¹ were used for plasmid transfer. Routine growth was performed in Tryptic Soy Broth (TSB, L. enzymogenes), marine LB (LB containing 3% w/v NaCl, V. parahaemolyticus), Minimal Marine Medium (2% NaCl, 0.4% galactose, 5 mM MgSO₄, 7 mM K₂SO₄, 77 mM K₂HPO₄, 35 mM KH₂PO₄, and 2 mM NH₄Cl, V. parahaemolyticus), or Lysogeny broth (LB, all other strains), aerated, or on TSB, marine LB, minimal marine medium, or LB plates containing 1.5% w/v agar. E. coli, P. aeruginosa, E. cloacae, and B. thai were grown at 37°C, while all other species were maintained at 30°C. All strains were grown under aerobic conditions, with liquid cultures shaking at 200 rpm. Media were supplemented as needed with antibiotics at the following concentrations: gentamicin (15 μg ml⁻¹, E. coli and B. thai, 30 μg ml⁻¹, Pseudomonas and V. parahaemolyticus, 150 μg ml⁻¹, L. enzymogenes), irgasan (25 μg ml⁻¹, Pseudomonas), trimethoprim (200 μg ml⁻¹, Pseudomonas), streptomycin (50 μg ml⁻¹, E. cloacae and L. enzymogenes), kanamycin (50 μg ml⁻¹, L. enzymogenes), chloramphenicol (25 μg ml⁻¹, E. coli and V. parahaemolyticus), and carbenicillin (150 μg ml⁻¹, E. coli).

Phage isolation and propagation

Phages were isolated from fresh rhizosphere soil samples collected from Denny Park or the Picardo Farm P-Patch Community Garden, both located in Seattle, Washington. Soil, including pieces of plant roots, was resuspended in 15 mL PBS and shaken at 200 rpm at room temperature for one hour on an orbital shaker (Thermolyne). The resulting suspension was centrifuged 10 minutes at 2,000 × g, and supernatant was filtered sequentially through a 100 μm cell strainer (Corning) and a 0.45 μl syringe filter (Thermo Scientific). 300 μl filtrate was incubated 10 minutes at room temperature with 50 μl of a log phase culture of P. protegens ΔgacS, supplemented with 5 mM CaCl₂, mixed with 3 ml LB top agar containing 0.5% agar, poured evenly onto LB agar plates, and incubated overnight at 30°C. Plaques were picked with a 20 μl barrier pipet tip, resuspended in 500 μl SM buffer (50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 8 mM MgSO₄), and filtered through a 0.22 μm syringe filter (Thermo Scientific). Each phage was plaque purified in this manner at least twice to ensure stock purity. To generate phage stocks, 5 ml SM buffer was added to top agar following complete or near complete bacterial cell clearance. Plates were sealed with parafilm and rocked overnight at 4°C to release phage particles. Buffer was collected from the flooded plate, centrifuged 8 min at 3,000 × g, and supernatant was serially filtered through 0.45 μm and 0.22 μm syringe filters. Stocks were stored at 4°C.

METHOD DETAILS

Plasmid and strain construction

All plasmids used in this study are described in Table S4 and are available upon request. Plasmids were generated using the following vector backbones: pRE112 (allelic exchange, V. parahaemolyticus), pEXG2 (allelic exchange, all other species), pUC18-mini-Tn7T-Gm-AraE-AraC-pBad (insertion at attTn7 site for complementation, P. protegens), pUC18-mini-Tn7T-Gm (nanoluciferase reporter) and pBT20 (transposon library generation, P. protegens and V. parahaemolyticus^92–95. Constructs were designed using Geneious Prime software (v2025.0.3) and primers were synthesized by Integrated DNA Technologies. Plasmids were generated by Gibson assembly⁹⁶ (pUC18T-mini-Tn7T-Gm-AraE-AraC-pBad and pEXG2), in vivo homology cloning (pEXG2⁹⁷), or restriction cloning (pRE112). For deletion constructs, 500- or 1000-nucleotide sequences flanking the deleted gene, retaining 10–17 amino acids at both the N- and C-termini, were cloned into vector linearized with SacI-HF and XbaI (pRE112, pEXG2, New England Biosciences) or HindIII-HF and KpnI-HF (pUC18T-mini-Tn7T-Gm-AraE-AraC-pBad, New England Biosciences). Fragments were amplified by PCR from bacterial genomes or synthesized by Twist Biosciences. Complementation plasmids were designed with the coding sequence following the ribosome-binding site from P. aeruginosa hcp1 cloned into the multiple cloning site of pUC18T-mini-Tn7T-Gm-AraE-AraC-pBad⁹³. The nanoluciferase reporter for rsmZ transctiption employed a previously engineered luciferase gene^98,99. Assembled plasmids were transformed into CaCl2 competent DH5α (pEXG2) or EC100D pir+ (pRE112) E. coli by 50 second heat shock at 42°C, recovered for 1 hour in 2XYT, and plated on selective media. Plasmid identity was confirmed by Sanger sequencing (Azenta Life Sciences).

Strains bearing in-frame deletions were generated by allelic exchange, as previously described⁹⁵. Briefly, plasmids were transformed into E. coli strains S17–1 λpir (delivery to L. enzymogenes) or Sm10 λpir (all other species), then delivered to recipient strains by conjugation. Plate-grown donor and recipient cells were mixed on LB agar, incubated 6 hours at 37°C (P. aeruginosa) or 30°C (all other species), and plated on selective media. Transconjugant colonies were counterselected on LB containing 15 (V. parahaemolyticus) or 10% sucrose (all other species) at 30°C and deletions were confirmed by sequencing across the insertion junction.

Complementation plasmids (pUC18-mini-Tn7T-Gm-AraE-AraC-pBad) were delivered to recipient cells by tri-parental mating, as previously described^90,100. Briefly, 100 μL each overnight cultures of Sm10 E. coli with the complementation plasmid, recipient strain, E. coli HB101 with pRK2013, and E. coli Sm10 with pTNS3 were combined, pelleted by centrifugation at 7,000 × g, washed once with fresh LB, and resuspended in 30 μL LB. The mixture was spotted onto LB agar and incubated for 6–8 hours at 30°C. Following incubation, conjugation spots were harvested, resuspended in LB, and plated on selective media. Transconjugant colonies were streaked for isolation a second time, and individual colonies were screened by PCR for the desired insertion.

Bacterial growth assays

Cultures were grown in liquid LB from isolated colonies, shaking at 200 rpm, at 37°C (P. aeruginosa) or 30°C (all other strains) until turbid. Cultures were then diluted to OD₆₀₀ = 0.01, transferred to a CellStar clear bottom 96-well plate (Greinier), with 5–10 replicates per genotype, and sealed with a BreatheEasy membrane (Diversified Biotech). Plates were incubated in a LogPhase 600 Microbiology Reader (BioTek) set to 37°C (P. aeruginosa) or 30°C (all other strains), shaking at 800 rpm, with OD₆₀₀ measurements taken every 5 minutes for a total of 18 hours. Maximum growth rate was determined by the LogPhase software using the change in OD₆₀₀ reading in successive measurements during log phase growth. Maximum growth rates were compared using a Welch’s t-test with Benjamini-Hochberg (BH) correction for multiple comparisons.

Competitive growth assays

Strains were grown overnight in liquid media, shaking at 200 rpm, inoculated from single colonies from freshly streaked plates. Cultures were collected with recipient strains (Pseudomonas, V. parahaemolyticus) in mid-log phase growth and donor strains (E. cloacae, L. enzymogenes, B. thai, and E. coli) in early stationary phase, pelleted by centrifugation at 4,000 × g for 10 minutes, resuspended at high cell density (OD₆₀₀ = 100) in fresh LB media, and mixed together at defined OD ratios. For gacS mutant competitions, cultures were mixed with E. cloacae at a donor:recipient OD ratio of 8:1 (P. aeruginosa and V. parahaemolyticus) or 4:1 (P. putida, P. putida, and P. fluorescens), or were mixed with B. thai at a donor:recipient ratio of 1:4 (P. fluorescens), 50:1 (V. parahaemolyticus), or 4:1 (all other strains). For the PFL_5124 mutant competition, cultures were mixed at a donor:recipient ratio of 10:1 (E. cloacae) or 20:1 (B. thai). 5 μl spots were plated on LB with 3% (w/v) agar (V. parahaemolyticus) or no-salt LB with 3% (w/v) agar (all other recipients) and were incubated at 6 hours at 37°C (P. aeruginosa) or 30°C (all other recipients). Surviving CFUs were enumerated by serial dilution on selective media. Contribution of genes to antagonism-dependent fitness was determined by the relative antagonism-dependent competitive index ([CI_{WT_vs_WT_donor} / CI_{WT_vs_ΔT6SS_donor}] / [CI_{mutant_vs_WT_donor} / CI_{mutant_vs_ΔT6SS_donor}]) of each strain. Competitive index is defined by [recipient_final / donor_final] / [recipient_initial / donor_initial] using recovered CFU counts. Data were analyzed using R v4.4.0 ¹⁰¹ and plotted using Prism (GraphPad, v10.4.1). Statistical significance was determined by Welch’s t-test comparing log transformed T6SS-dependent competitive index between wild-type and mutant recipients, with BH correction for multiple comparisons.

Competitive growth assays with OBC4 mutant strains were performed as described above, with the following modifications. P. protegens recipient cells were resuspended directly from LB agar plates into 10 mL LB. Donor and recipient cultures were normalized to OD₆₀₀ = 1 and 0.1, respectively. Following competition, genomic DNA of recovered cells was purified using a QIAGEN Blood and Tissue gDNA prep kit and relative proportions of wild-type P. protegens and P. protegens ΔOBC4 were determined by qPCR on a CFX Connect Real-Time System (Bio-Rad) using SSoAdvanced Universal SYBR Green Supermix (Bio-Rad). Significant differences in mutant fraction change were determined by Welch’s t-test, with BH correction for multiple comparisons.

Proteomics

Pseudomonas cultures were grown in 2 ml liquid LB media, shaking at 200 rpm, at 30°C (P. protegens and P. fluorescens) or 37°C (P. aeruginosa) until late log phase growth, determined by an OD₆₀₀ between 1.2 and 1.5. Cells were pelleted by centrifugation at 9,000 × g for 3 minutes, washed twice in phosphate buffered saline (PBS), resuspended in 100 μl lysis buffer (8 M urea, 75 mM NaCl, 50 mM Tris-HCl, pH 8.2), lysed by three freeze-thaw cycles interspersed with 2 minute sonication in a Cole-Parmer ultrasonic cleaner, and centrifuged at 20,000 × g for 10 minutes. Protein content in the supernatant was quantified using a Pierce BCA Protein Assay Kit (Thermo Fisher Scientific). Samples were reduced in 5 mM 1,4-dithioreitol (DTT) for 25 minutes at 55°C, cooled to room temperature, and alkylated using 14 mM iodoacetamide for 30 minutes in the dark. Alkylation was quenched by addition of DTT to 10 mM. Samples were diluted in 5 volumes of 25 mM Tris-HCl, pH 8.2, CaCl2 was added to 1 mM, and protein was digested at 37°C overnight with 4 μg/ml trypsin. Digestion was halted with trifluoroacetic acid (TFA, 0.4% final volume), samples were centrifuged 10 minutes at 20,000 × g, and pellets were discarded, retaining supernatant. P. aeruginosa and P. protegens samples were loaded onto BioPureSPN Mini RPC desalting columns (Nest Group) preconditioned with 100% acetonitrile (ACN), LC-MC grade water, and 0.1% formic acid (FA), washed with 0.1% FA, and eluted in 80% ACN with 0.1% FA. For P. fluorescens, 0.4% TFA was added to samples, which were then applied to MCX StageTips¹⁰² that had been preconditioned twice with 100% MeOH, once each with 100% ACN, 75% ACN with 5% NH₄OH, and 75% ACN with 0.5% acetic acid, and twice with 0.1% TFA. StageTips were washed with 0.1% TFA, twice with 75% ACN containing 0.5% acetic acid, and once with 0.5% acetic acid in LC-MS grade water. Samples were eluted in 75% ACN with 5% NH₄OH. All samples were dried using an SPD1030 SpeedVac (Thermo Fisher Scientific) and resuspended in 5% acetonitrile with 0.1% formic acid (FA) to a final protein concentration of 0.25 μg/ml. Samples were analyzed by LC-MS/MS as previously described⁵ using a Lumos Fusion Orbitrap Mass Spectrometer (Thermo Fisher Scientific). Spectra were mapped to reference proteomes and analyzed using the MaxLFQ algorithm in MaxQuant software^103,104, using 1% FDR cutoffs at the peptide, protein, and site levels. The unannotated ORF downstream of PFL_2023 predicted as phage defense gene ariB was not included in the initial peptide matching. Peptide abundance for this protein was determined by a second MaxQuant run. Missing values were imputed from a normal distribution of log-transformed data defined by a mean at (data median – 1.4 × std. dev) and a standard deviation of (data std. dev * 0.5), akin to previously described methods¹⁰⁵. Differentially regulated proteins were identified using a Welch’s t-test with BH correction for multiple comparisons and a filter of >2-fold difference between genotypes. Principal component analysis was performed using the ggbiplot package in R¹⁰⁶. Identification of phage defense systems was performed using web tools PADLOC and DefenseFinder^107,108.

Transposon library generation

A Himar1 transposon insertion library was generated in P. protegens Pf-5, Pf-5 ΔtssM, and V. parahaemolyticus RIMD 2210633 using established methods⁵. Briefly, the suicide vector pBT20, containing a Himar1 mariner transposon and transposase, was conjugated into recipient strains from E. coli Sm10 λpir for 6 hours on LB agar (P. protegens) or from RHO3 overnight on LB agar containing 400 μg/ml diaminopimelic acid (DAP) (V. parahaemolyticus). Transconjugant colonies were selected on LB agar containing 30 μg/ml gentamicin and 25 μg/ml irgasan (P. protegens) or on marine LB agar containing 30 μg/ml gentamicin (V. parahaemolyticus), then pooled in LB containing 15% dimethylsulfoxide (DMSO, P. protegens) or 15% glycerol (V. parahaemolyticus). Libraries were aliquoted, flash frozen in liquid nitrogen, and stored at −80°C. Transposon insertion sequencing, as described below, revealed library complexities of 148,362 unique insertion sites in the wild-type P. protegens library, 170,203 in the P. protegens ΔtssM library, and 304,000 in the V. parahaemolyticus library.

Tn-seq screen and analysis

Himar1 mariner transposon mutant libraries were grown to log phase in LB broth, shaking at 200 rpm, at 30°C. Cells were pelleted by centrifugation at 4,000 × g for 10 minutes, washed in fresh LB and resuspended at a defined OD₆₀₀ to yield the donor:recipient ratio described below when mixed with a donor strain at OD₆₀₀ = 100. 196 replicate spots of 5 μl of each mixture were plated on marine LB with 3% (w/v) agar (V. parahaemolyticus) or on no-salt LB with 3% (w/v) agar (P. protegens) and incubated at 30°C for a defined time period described below. The wild-type P. protegens library was grown with E. cloacae (wild-type or ΔtssM) at a 100:12.5 donor:recipient ratio for 18 hours, B. thai (wild-type or ΔtssM) at 100:2 for 18 hours, and L. enzymogenes C3 (wild-type or ΔvirD4) at 100:12.5 for 6 hours. The P. protegens ΔtssM library was grown with E. cloacae (WT or ΔtssM) at a 100:25 donor:recipient ratio for 18 hours. The V. parahaemolyticus library was grown with E. cloacae (WT or ΔtssM) at a 100:12.5 donor:recipient ratio for 4 hours. Following incubation, cell mixtures were resuspended in LB and immediately plated for surviving Pseudomonas and Vibrio cells on LB agar containing 25 μg/ml irgasan (P. protegens) or on LB agar containing 30 μg/ml gentamicin (V. parahaemolyticus). After overnight incubation at 30°C (P. protegens) or 20°C (V. parahaemolyticus), cells were harvested using a sterile cell scraper from plates flooded twice with 10 mL LB, and genomic DNA was purified using the QIAGEN Blood & Tissue gDNA prep kit. Tn-seq libraries were generated by DNA shearing, C-tailing, and repeated PCR amplification, as previously described¹⁰⁹. Transposon insertion libraries were quantified using a KAPA Library Quantification Kit (KAPA Biosystems) for each experiment were pooled, and multiplexed samples were sequenced as 51-base single-end reads on an Illumina MiniSeq with 30–40% phiX DNA spiked in.

Illumina sequencing reads were analyzed using a previously described custom Python script⁹⁹. Briefly, sequences were filtered for those with the first 6 bases matching the transposon end and were mapped to the relevant genome using a BWA aligner. Reads per insertion site were tallied, and the sum of all insertion sites within 5 and 90% of a coding sequence was recorded for each gene. Transposon insertion site reads for a library competed against different donor strains were normalized by the median fold-difference between all sequenced genes. Missing values were imputed from a normal distribution of log-transformed data defined by a mean at (data median – 1.4 × std. dev) and a standard deviation of (data std. dev * 0.5). Conditional gene essentiality was determined using a Mann-Whitney U-test comparing the fold-difference in reads between donor strains for each transposon insertion site in a gene with the same ratio across the whole genome, using BH correction for multiple comparisons.

Gac/Rsm system ortholog identification

Orthologous gene groups (orthogroups) were identified by integration of two complementary methods. First, orthologous relationships between proteins were identified across diverse bacterial genomes using OrthoFinder v2.2.3¹¹⁰. In parallel, an independent all-against-all BLASTP search¹¹¹, with a significance threshold of 0.005, was performed. Only reciprocal top hits from different genomes were grouped into orthogroups. Results from both methods were integrated, and false positives were further removed by domain architecture analysis, gene neighborhood analysis, and phylogenetic inference. For domain architecture analysis, hmmscan¹¹² was used to search both the Pfam database¹¹³ and a custom in-house domain profile database. Gene neighborhood was assessed by examining the ten genes upstream and downstream of each target gene. Phylogenetic relationships were inferred using PhyML¹¹⁴.

Generation of GTDB-based phylogeny

The phylogenetic tree underlying the Genome Taxonomy Database (GTDB) v220 was downloaded from the GTDB^115–118 website (https://gtdb.ecogenomic.org/) and converted to Newick format in Geneious Prime (v2025.0.3) Select reference species were pruned from the tree using the prune function in ETE software v3.1.3¹¹⁹, preserving branch length. Trees were visualized using FigTree (v1.4.4). Species were labeled using NCBI taxonomy.

LPS extraction and quantification

LPS extraction and visualization was performed using established methods^120,121. Briefly, overnight cultures were pelleted by centrifugation at 9,000 × g for 3 minutes, washed twice in PBS, and resuspended at a concentration of OD₆₀₀ = 0.75. Serial dilutions were plated on LB agar to ensure equal cell concentrations. 1 mL of resuspended cells was centrifuged 10 minutes at 10,600 × g, discarding supernatant. Pellets were optionally stored at −20°C. Cells were gently resuspended in 200 μl SDS buffer, freshly diluted in distilled H₂O from a 2x stock (4% β-mercaptoethanol, 4% SDS, 20% glycerol in 0.1 M Tris-HCl, pH 6.8) and boiled at 95°C for 10 minutes. Samples were cooled to room temperature for 10 minutes, then 5 μl of RNase A (10 mg/ml) and 2 μl of Turbo DNase (Thermo Fisher Scientific) were added. Samples were gently mixed and were incubated at 37°C for 30 minutes. Following nuclease treatment, 10 μl of proteinase K (10 mg/ml) was added, and samples were incubated at 59°C for 3–4 hours, shaking at 350 rpm. 10 μl of each sample was loaded onto an 8–16% gradient Criterion TGX stain-free precast polyacrylamide gel (Bio-Rad), and a 100V current was applied until the dye front reached nearly to the bottom of the gel. Gels were briefly rinsed in dH₂O, incubated 15 minutes on an orbital shaker with fixing-oxidizing solution (40% ethanol, 5% acetic acid, 1% sodium (meta)periodate), then silver stained with a SilverQuest staining kit (Invitrogen) according to the manufacturer’s protocol, beginning with the sensitization step. Gels were imaged using a GelDoc Go imager (Bio-Rad). Due to the sensitivity of silver staining and the differential abundance of OBC4 O-polysaccharide and Lipid-A core, the two gels shown in Figure S4B derive from distinct biological replicates.

OBC4 was quantified by densitometry using FIJI¹²², according to established methods^123,124. Briefly, a consistently sized box was drawn to one-third the width of each lane at the OBC4 region of the gel and mean signal intensity inside the box was measured. The lane containing P. protegens ΔOBC4 was set as background. Mean signal intensity of background was subtracted, and samples were normalized to the wild-type lane. Statistical significance was determined by a Welch’s t-test.

2,4-DAPG sensitivity assay

Bacterial sensitivity to 2,4-DAPG was determined using established methods¹²⁵. Briefly, cultures were grown to early log phase from freshly streaked colonies, diluted to OD₆₀₀ = 0.001, and pipetted into a clear-bottom 96-well plate (150 μl/well). 4 μl of 2,4-DAPG, serially diluted two-fold in DMSO, was added to each well to the indicated final concentration. Cultures were grown in triplicate for 24 hours in a LogPhase 600 Microbiology Reader (BioTek) set to 30°C, shaking at 800 rpm, with OD₆₀₀ measurements collected at 5 minute intervals. OD₆₀₀ readings were corrected by subtraction of background from a blank LB well. Sensitivity was determined by the OD₆₀₀ reading relative to a DMSO control as the DMSO control was exiting log phase growth (10.08 hours). IC50 was calculated based on a sigmoidal fit of the relative OD₆₀₀ readings.

Efficiency of plating and plaque size assays

Plaquing efficiency for each phage was determined by spotting 10-fold serial dilutions of a single phage stock onto top agar lawns containing different P. protegens mutants. P. protegens strains were grown overnight on LB agar, resuspended in LB at OD₆₀₀ = 1, mixed with LB top agar containing 5 mM CaCl₂ (112.5 μl cell suspension in 6.75 ml top agar), and poured evenly onto 15 cm LB agar plates. Top agar lawns were dried 15 minutes in a biosafety cabinet, then 3 μl spots of phage stock, serially diluted in SM buffer, were dispensed using a Rainin BenchSmart 96 pipettor (Mettler Toledo). Plates were incubated at 30°C for 6–15 hours, until plaques appeared. Differential plaquing efficiency of each phage on P. protegens ΔgacS versus ΔretS was determined using a one-sample t-test. Differential plaquing of SeaP2 on phage defense system mutant strains was determined by one-way ANOVA followed by Holm-Sidak post hoc test. Plaque images were captured on a GelDoc Go imager (Bio-Rad) under Coomassie settings. Images were inverted, scaled, and cropped in FIJI.

For plaque area assays, 350 μl of log phase cells normalized to OD₆₀₀ = 1 were combined with an estimated 250 pfu of Ppr_SeaP2, 5 mM CaCl₂, and 9 ml LB top agar and gently poured onto a 15 cm LB agar plate. Complementation strains were supplemented as necessary with 0.01% arabinose in water or a water control. After overnight growth, plates were imaged using a GelDoc Go imager (Bio-Rad) under Coomassie gel settings. Plaque areas were determined manually using the oval tool in FIJI. Statistical significance was determined by one-way ANOVA followed by a Dunnett’s T3 multiple comparisons test with the parental strain.

Phage sequencing and classification

Phage DNA was purified for sequencing from high titer stocks. DNA was released from capsids by sequential incubation with nucleases (100 μg/ml DNase I and RNase I, 37°C for 30–45 minutes), 10 mM EDTA (37°C for 15 minutes), and 200 μg/ml proteinase K (50°C for 30 minutes). Samples were centrifuged at 21,000 × g for 2 minutes. DNA was precipitated from supernatant by addition of 1/10 volume sodium acetate, pH 5.5, 1/100 volume GlycoBlue, and 2.5× volume 100% ethanol, followed by incubation for 1 hour at −20°C. DNA was pelleted by centrifugation at 21,000 × g for 15 minutes and washed sequentially with 100% isopropanol and 70% ethanol. Pellets were dried 20–30 minutes and gently resuspended in water. DNA concentration and integrity were determined by Qubit fluorometry and gel electrophoresis, respectively.

Sequencing libraries were prepared from 100 – 300 ng purified DNA using Illumina DNA Prep kit. Libraries were pooled and sequenced in multiplex as 2 × 150-bp paired-end reads using an Illumina iSeq. Reads were trimmed using Trimmomatic (Galaxy Version 0.39+galaxy2) with Illuminaclip, Trailing (minimum quality 25–30), AvgQual (minimum 25–30) and MinLen (120 bp) operations¹²⁶. Trimmed paired reads were assembled using SPAdes (Galaxy version 3.15.5+galaxy2) with either automatic or user-specified (55, 77, 99, 101) k-mer size values to obtain a single large contig of >40kb for each purified phage¹²⁷. Phage were taxonomically classified by BLASTP searches of their tail protein sequences against the NCBI nr database. The closest matches corresponded to tail proteins from Pseudomonas phages vB_PpuP-Luke-3 (95.7% aa ID to the B2 tail protein) and PollyC (99.6% aa ID to the Ppr_SeaP2 tail protein). VIRIDIC analysis with the full sequences of these phages indicated that they belonged to the same two genera as those of the phage we isolated (SeaD1–2 with vB_PpuP-Luke-3 and Ppr_SeaP2–5 with PollyC)¹²⁸.

Nanoluciferase assay

For nanoluciferase assays without phage, log phase cells were diluted to an OD₆₀₀ of 0.5. For nanolicuferase induction during phage infection, log phase cells were washed once in fresh LB, resuspended to an OD₆₀₀ of 2, and mixed with either Ppr_SeaP2 at an MOI of 10⁻³ or SM buffer, supplemented with 5 mM CaCl₂. 50 μl was spotted on LB agar, allowed to dry, and incubated at 30°C for 2 hours. Spots were recovered with a sterile cell scraper in 1 ml LB and normalized by OD₆₀₀. For all nanoluciferase assays, 30 μl of culture was mixed with 30 μl Nano-Glo Luciferase assay reagent (Promega N1110) in a 96-well plate. Luminescence signal was detected using a Cytation 2 plate reader following a 3 minute incubation at room temperature. Genomic DNA was prepared using Instagene Matrix (Bio-Rad 732–6030) according to the manufacturer’s protocol and DNA content was determined by qPCR. For luminescence values reported as percent GRP activation, 0 and 100% were set as the normalized luminescence values of ΔgacS and ΔretS strains, respectively. Significance was determined using a one-tailed Welch’s t-test, with BH correction where appropriate.

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical details for all experiments can be found in the figure legends. Statistical analysis was performed using R (v4.4.0) or Prism (v10.4.1) software. All graphs were generated using Prism and were formatted using Adobe Illustrator (v29.5). Detailed descriptions of statistical methods used for proteomics and Tn-seq experiments are described in the relevant Method Details section.

Supplementary Material

2

Table S1. Whole-cell proteomics data for GRP regulon identification, related to Figure 2

3

Table S3. Tn-Seq competition data, related to Figure 3

4

Table S4: Strains, phages, plasmids, and oligonucleotides used in this study, related to STAR methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Bacterial and virus strains
All bacterial strains and phages used in this study are listed in Table S4	N/A	N/A
Chemicals, peptides, and recombinant proteins
Instagene Matrix	Bio-Rad	Cat#732–6030
Sequencing Grade Modified Trypsin	Promega	Cat#V5111
Turbo DNase	Thermo Fisher Scientific	Cat#AM2239
Gentamicin sulfate	Amresco	Cat#304
Irgasan	Millipore Sigma	Cat#72779
Trimethoprim	Sigma-Aldrich	Cat#T7883
Streptomycin sulfate	Cepham Life Sciences	Cat#3810–74-0
Kanamycin monosulfate	Goldbio	Cat#K-120–50
Chloramphenicol	Research Products International	Cat#C61000
Carbenicillin disodium salt	Goldbio	Cat#C-103–5
Trifluoroacetic acid	Sigma-Aldrich	Cat#80457
Dimethylsulfoxide	G Biosciences	Cat#786–1388
Glycerol	Fisher Scientific	Cat#56–81-5
2-propanol	Sigma-Aldrich	Cat#I9516
Formic acid	Thermo Scientific	Cat#85178
Acetonitrile	Thermo Fisher Scientific	Cat#51101
Ammonium hydroxide	Sigma-Aldrich	Cat#221228
Acetic acid	Thermo Fisher Scientific	Cat#A38S-212
Calcium chloride	Sigma-Aldrich	Cat#C7902
Sucrose	VWR	Cat#VWRV0335
2,6-Diaminopimelic acid	Sigma-Aldrich	Cat#D1377
Proteinase K	Qiagen	Cat#19133
Urea	Sigma-Aldrich	Cat#U5378
Sodium chloride	Fisher Scientific	Cat#S271–3
1,4-Dithioreitol (DTT)	Research Products International	Cat#D11000
Iodoacetamide	Thermo Scientific	Cat#A39271
Sodium (meta)periodate	Sigma-Aldrich	Cat#S1878
2,4-Diacetylphloroglucinol	Santa Cruz Biotechnology	Cat#sc-206518
Galactose	Sigma-Aldrich	Cat#G0750
Water, LC-MS grade	Thermo Fisher Scientific	Cat#047146-K7
LB Broth, Miller	BD	Cat#214906
Tryptic soy broth	Thermo Fisher Scientific	Cat#R455056
Tryptone	Sigma-Aldrich	Cat#T3938
Yeast extract, technical	BD	Cat#288620
Ethylenediaminetetraacetic acid disodium salt dihydrate	Sigma-Aldrich	Cat#ED2SS
GlycoBlue Coprecipitant	Gift from Meeske Lab	N/A
Critical commercial assays
DNeasy Blood & Tissue Kit	Qiagen	Cat#69506
Nano-Glo Luciferase Assay System	Promega	Cat#N1110
Pierce BCA Protein Assay Kit	Thermo Fisher Scientific	Cat#23225
BioPureSPN Mini	The Nest Group	Cat#HUM S18V
8–16% Criterion TGX polyacrylamide gel	Bio-Rad	Cat#5678104
SilverQuest Staining Kit	Invitrogen	Cat#45–1001
Qubit 1X dsDNA HS Assay Kit	Thermo Fisher Scientific	Cat#Q33231
SSoAdvanced Universal Supermix	Bio-Rad	Cat#1725272
KAPA Library Quantification Kit	Roche	Cat#07960140001
Deposited data
Genome Taxonomy Database v220	Parks et al. (114)	https://gtdb.ecogenomic.org
ETE v3.1.3	Huerta-Cepas et al. (116)	N/A
FigTree v1.4.4	https://tree.bio.ed.ac.uk/software/figtree/	N/A
Oligonucleotides
All oligonucleotides used in this study are listed in Table S4	N/A	N/A
Recombinant DNA
All plasmids used in this study are listed in Table S4	N/A	N/A
Software and algorithms
Geneious Prime v2025.0.3	Geneious, Software, Newark, New Jersey, USA	https://www.geneious.com; RRID:SCR_010519
Prism v10.4.1	GraphPad, Software, La Jolla, California, USA	https://www.graphpad.com; RRID:SCR_022798
Adobe Illustrator v29.5	Adobe Systems Incorporated, San Jose, California, USA	https://www.adobe.com/products/illustrator; RRID:SCR_010279
R v4.4.0	R Foundation	https://www.r-project.org/; RRID:SCR_001905
MaxQuant v2.0.3.0	Tyanova et al. (100)	RRID:SCR_014485
MaxLFQ	Cox et al. (99)	N/A
PADLOC v2.0.0	Payne et al. (103)	https://padloc.otago.ac.nz/
DefenseFinder v0.1.0	Tesson et al. (104)	https://defensefinder.mdmlab.fr/
Custom script for Tn-seq analysis	Wang et al. (106)	https://github.com/lg9/Tn-seq
OrthoFinder v2.2.3	Emms and Kelly (107)	https://github.com/davidemms/OrthoFinder
BLASTP	Altschul et al. (108)	RRID:SCR_001010
hmmscan	Potter et al. (109)	https://github.com/EddyRivasLab/hmmer
PhyML	Guindon et al. (111)	https://github.com/stephaneguindon/phyml
FIJI	Schindelin et al. (119)	https://fiji.sc/
Trimmomatic	Bolger et al. (123)	http://www.usadellab.org/cms/?page=trimmomatic
SPAdes	Bankevich (124)	http://bioinf.spbau.ru/spades
VIRIDIC	Moraru et al. (125)	http://viridic.icbm.de/
Other
MiniSeq System	Illumina	RRID:SCR_016378
CFX Connect Real-Time System	Bio-Rad	RRID:SCR_026760
SpeedVac SPD1030 Integrated Vacuum Concentrator	Thermo Fisher Scientific	Cat#SPD1030
Lumos Fusion Orbitrap Mass Spectrometer	Thermo Fisher Scientific	Cat# FETD2–10002
GelDoc Go	Bio-Rad	Cat#12009077
LogPhase 600 Microbiology Reader	BioTek	N/A
Rainin BenchSmart 96	Mettler Toledo International	Cat#30296706
iSeq 100 System	Illumina	RRID:SCR_016377
Cytation 1 Plate Reader	BioTek	RRID:SCR_019730
CellStar clear bottom 96-well plate	Greiner Bio-One	Cat#655185
Breathe-Easy membrane	Diversified Biotech	Cat#BEM-1

Highlights.

• Pseudomonas spp. possess an expanded Gac/Rsm global regulatory pathway (GRP)

• The GRP regulates defenses against assorted biotic threats in diverse pseudomonads

• Efficacy of individual GRP-regulated factors depends on the antagonizing species

• GRP-regulated defense systems protect against phage infection