Summary
Plant root-associated microbes promote plant growth, in part by the induction of systemic resistance (ISR) to foliar pathogens. In an attempt to find novel growth-promoting and ISR inducing strains, we previously identified strains of root-associated Pseudomonas spp. that promote plant growth but unexpectedly induced systemic susceptibility (ISS) to foliar pathogens. Here we demonstrate that the ISS-inducing phenotype is common among root-associated Pseudomonas spp. and we identified the underlying genetic and molecular basis of ISS. Using comparative genomics we identified a single P. fluorescens locus containing a novel periplasmic spermidine biosynthesis gene speE2 that is unique to ISS strains. We generated a clean deletion of the speE2 gene in two ISS strains and found that speE2 is necessary for the ISS phenotype. Spermidine but not spermine is sufficient to phenocopy ISS strains. The ISS locus is present in diverse bacteria and has previously been implicated in pathogenesis in animals. Collectively these data show that a single bacterially derived molecule can modulate systemic plant immunity.
Introduction
Plant growth promotion by beneficial microbes has long been of interest because of the potential to improve crop yields. Individual root-associated microbial strains can promote plant growth by facilitating nutrient uptake, producing plant hormones, or improving resilience to both abiotic and biotic stresses [1]. In some cases, single bacterial loci underlie beneficial effects of microbes on plants while other traits appear to be complex and polygenic.
Pseudomonas fluorescens and related species are a model for beneficial host-associated microbes due to their genetic tractability and robust host-association across diverse eukaryotic hosts. Direct plant growth promotion (PGP) by Pseudomonas spp. can be mediated by bacterial production of the phytohormones auxin [2] or by the expression of 1-aminocyclopropane-1-carboxylate (ACC) deaminase that metabolizes plant-derived ethylene [1,3]. Indirect PGP through antimicrobial activity and pathogen suppression has been attributed to production of the antibiotic 2,4-diacetylphloroglucinol (DAPG) [4]. However, the molecular basis of many traits such as induced systemic resistance (ISR) has remained elusive, and multiple distinct bacterial traits including production of siderophores, LPS, and salicylic acid have all been implicated [5].
We previously reported two Pseudomonas spp. that induce systemic susceptibility (ISS) on Arabidopsis and can promote growth under nutrient limiting conditions [6,7]. These same Pseudomonas strains suppress a subset of SA-dependent responses and promote resistance to herbivores [7]. Although it is possible that these strains contain multiple genetic loci that affect plant growth and pathogen resistance, we were interested in investigating whether a single bacterial trait is primarily responsible for both growth and immunity phenotypes. Growth and immunity have a reciprocal relationship in plants, leading to growth-defense tradeoffs to the extent that plant stunting has been used as a proxy for autoimmunity [8]. As a result, we hypothesized that suppression of plant immunity by Pseudomonas strains that trigger ISS may be a consequence of PGP activity. The genomes of these strains do not contain genes for the ACC deaminase enzyme prevalent in other Pseudomonas PGP strains [3]; thus, we hypothesized that there may be a distinct mechanism of growth promotion in these strains.
Because of the high density of sampling and genome sequencing within P. fluorescens and related species, we reasoned that if ISS is an overlooked consequence of growth promotion then: 1) we should be able to identify additional ISS strains by sampling known PGP strains and additional root-associated strains, and 2) assuming a single unique locus was responsible, that a comparative genomics approach should reveal the underlying genetic basis of ISS.
Here we report that ISS is relatively common among Pseudomonas strains. We identified new ISS isolates including previously described PGP or environmental isolates and new isolates from Arabidopsis roots. Using comparative genomics, we identified a single bacterial locus containing a spermidine synthase gene (speE2) that is unique to Pseudomonas ISS strains. We show that speE2 is necessary to elicit ISS and that exogenous spermidine is sufficient to trigger ISS in the absence of bacteria. Collectively, these data indicate that a single monogenic microbial trait underlies a systemic immune response in a plant host.
Results
ISS is a common feature of growth-promoting Pseudomonas spp
We previously reported that two strains of Pseudomonas (CH229 and CH267) induce systemic susceptibility (ISS) to the foliar pathogen Pseudomonas syringae DC3000 (Pto DC3000) under conditions where a well characterized ISR strain (P. simiae WCS417 [9]) conferred resistance to Pto DC3000 [6,7]. To the best of our knowledge, descriptions of Pseudomonas-elicited ISS against bacterial pathogens are limited to Pseudomonas sp. CH229 and CH267, strains that were independently isolated from the rhizospheres of wild Arabidopsis plants in Massachusetts, USA. We reasoned that if ISS is common among Arabidopsis-associated Pseudomonas spp., we would be able to identify additional ISS strains from Arabidopsis roots from plants growing at distinct sites.
We isolated 25 new fluorescent pseudomonads from wild-growing Arabidopsis plants from additional sites in Massachusetts and in Vancouver, Canada. We generated ~800 bp sequences of a region of the 16S rRNA gene where strains CH229 and CH267 are 99.5% identical, but each shares only <96% identity to the well-characterized ISR strain WCS417. Reasoning that new ISS strains would be closely related to CH267 and CH229, we selected 3 new isolates [1 from Massachusetts (CH235) and 2 from British Columbia (PB101 and PB106)] that were >97% identical to CH267 by 16S rRNA sequence and another 3 (from British Columbia: PB100, PB105 and PB120) that were <97% identical to CH229 and CH267 (Figure 1A).
We tested these 6 new rhizosphere Pseudomonas isolates for their ability to trigger ISS. We found that 2 of the 3 strains that were most closely related to CH267 (CH235 and PB101) elicited ISS (Figure 1B). Two strains with <96% identity to CH267 failed to trigger ISS: PB105 triggered ISR and PB100 had no effect on systemic defenses (Figure 1). PB106 and PB120 consistently enhanced susceptibility in all experiments, but to a more moderate degree (*p<0.1). Collectively, these data indicate that the ability to elicit ISS on Arabidopsis ecotype Col-0 may be a common feature among some, but not all, closely-related strains of Pseudomonas spp. isolated from the Arabidopsis rhizosphere.
Because ISS seemed restricted to strains that were closely related to CH267, we obtained several additional isolates with similar 16S sequences including Pseudomonas sp. UW4, Pseudomonas sp. Pf0-1 and P. vancouverensis DhA-51 (Table 1) and a growth promoting strain, Pseudomonas sp. WCS365, that is more distantly related and to our knowledge has not been tested for ISR/ISS. We found that UW4 and DhA-51 elicited ISS while Pf0-1 and WCS365 did not (Figure 1B). Pseudomonas sp. UW4 [10] and WCS365 are well-characterized growth promoting strains. Pseudomonas sp. Pf0-1 [11] is an environmental isolate. Pseudomonas vancouverensis DhA-51 is also an environmental isolate [12] and was previously shown to be closely related to Pf0-1 [13]. Because DhA-51 is an environmental isolate that triggers ISS, these data show that the ability to trigger ISS is not specific to rhizosphere isolates.
To gain insights into the distinguishing features of ISS strains, we sequenced the genomes of the 6 new isolates (CH235, PB100, PB101, PB105, PB106 and PB120) from Arabidopsis roots as well as P. vancouverensis DhA-51 (UW4, WCS365, CH267 and CH229 have been sequenced previously). Whole genome shotgun sequencing was used to assemble draft genomes (Methods). We generated a phylogenetic tree using 122 conserved genes as described previously [7,14]. We found that all ISS strains are closely related to one another and fall within a monophyletic group which corresponds to the P. koreensis, P. jessenii, and P. mandelii subgroups of P. fluorescens identified in a recent phylogenomic survey of Pseudomonas spp. (Figure 2B; [15]). However, not every isolate in this clade is an ISS strain; notably Pf0-1, which has no effect on systemic immunity despite being closely related to CH229. We reasoned that the absence of the ISS phenotype in Pf0-1 should facilitate the use of comparative genomics by allowing us to separate the phylogenetic signature from the phenotypic signature of ISS.
11 genes in a single genomic locus are unique to ISS strains and predicts ISS
To identify the potential genetic basis of the ISS phenotype, we used a previously described database of orthologous genes for Pseudomonas spp. [14] to identify genes that are present in ISS strains (CH229, CH235, CH267 and UW4) but are absent in the closely-related strain that has no effect on systemic defenses (Pf0-1). We used only the ISS strains with the most robust phenotypes for this analysis. We identified 29 predicted protein coding genes missing from Pf0-1 but present in all of the other strains. Of these, 12 were small (<100 aa) hypothetical proteins. The remaining 17 predicted protein coding genes were prioritized for further analysis and are shown in S1 Table. Intriguingly, 11 of the 17 ISS unique genes are found in a single genomic locus.
We surveyed the genomes of other Pseudomonas strains tested for ISS to determine if the presence of the 17 genes identified by our comparative genomics approach correlated with the ISS phenotype. We found that the 11 clustered genes were present in ISS strains (DhA-51 and PB101) and the strains with intermediate phenotypes (PB120 and PB106) but were absent in the non-ISS strain WCS365, WCS417 and PB105 (S1 Figure). The remaining 6 genes were all present in WCS365 and/or other non-ISS strains (S1 Figure). We chose to focus on the 11 ISS-unique genes (“ISS locus” hereafter) for further study.
We found that the 11 genes in the ISS locus are found at a single genomic locus in all 4 of the ISS strains (S2 Figure and Figure 2A). The flanking regions are conserved in the non-ISS strain Pf0-1 (Figure 2A), indicating a recent insertion or deletion event. Within this locus, there is a single gene that is conserved in Pf0-1 in addition to two genes that are unique to each individual strain suggesting multiple changes to this genomic region in recent evolutionary history. While all 11 genes are within the same genomic region in the ISS strains, the variability of this locus between closely related strains suggests it may be rapidly evolving.
We surveyed the genomes of sequenced isolates available in our collection for the presence of the ISS locus. We found a number of closely-related strains from various environmental sources that contained the ISS locus, as well as a more distantly related strain (Pf-5) (Figure 2B). We tested 2 of these newly identified strains that contain the ISS locus (Pf-5 and GW456-L13) as well as 2 that do not (FW300-N1B4 and FW300-N2C3) and found that the presence of the ISS locus correlated with the ISS phenotype, including the distantly related strain Pf-5 (Figure 2C). Collectively, these data show that the presence of the 11 candidate genes in the ISS locus identified by our comparative genomics approach is predictive of the ISS phenotype.
SpeE is necessary and spermidine is sufficient to trigger ISS
Because ISS strains CH267 and CH229 have previously been shown to promote growth [6], we examined the annotations of the 11 genes within the ISS locus for a gene that might be involved in biosynthesis of a growth-promoting or immuno-modulatory compound (Figure 2A). We identified a predicted polyamine synthase-encoding gene within the ISS locus annotated as speE2 (PputUW4_02826 and CP336_12795 in UW4 and CH267 respectively). CH267 speE2 has similarity to a characterized spermidine synthase gene speE in P. aeruginosa (25% predicted amino acid identity to P. aeruginosa PA1687 [16]). A second speE-like genes in the genomes of UW4 and CH267 annotated as spE1 is outside of the ISS locus (PputUW4_03691 and CP336_28780 in UW4 and CH267 respectively) and is highly similar to the P. aeruginosa speE gene (~84.0% predicted amino acid identity) [16]. Polyamines including spermidine have been implicated in a range of plant growth and defense-related phenotypes [17,18] and are known to be synthesized by diverse microbes [19].
To test if the gene cluster and the speE2 gene are necessary for ISS strains to induce systemic susceptibility, we deleted the entire 15 kB locus including the region spanning the 11 genes identified in our initial comparative genomics screen in strains CH267 and UW4 (Figure 2A). We also constructed an in-frame deletion of just the speE2 gene in both CH267 and UW4. We retested these deletion mutants for their ability to induce systemic susceptibility and found that deletion of the speE2 gene alone, or the entire 11-gene cluster, resulted in a loss of the ISS phenotype in both CH267 and UW4 (Figure 3A). This indicates that speE2 is necessary for ISS.
To determine if spermidine or a related polyamine is sufficient to trigger ISS, we watered soil-grown plants with purified spermidine or spermine at concentrations of 1, 10 and 100 µM. Spermidine application in this range has been previously shown to promote plant salt tolerance [20]. We found that 10 and 100 µM spermidine but not spermine could induce systemic susceptibility indicating that spermidine is sufficient to phenocopy the bacterial ISS phenotype (Figure 3B).
speE2 is a predicted periplasmic spermidine synthase that uses a novel source of dSAM
Because deletion of speE2 in CH267 and UW4 results in the specific loss of the ISS phenotype, this indicates that the speE1 and speE2 genes are not functionally redundant. SpeE1 and speE2 differ in length and predicted structure (Figure 4A). SpeE1 encodes a predicted 384-amino acid protein and contains a predicted polyamine synthase domain with a predicted decarboxylated S-adenosyl methionine (dSAM) binding motif. SpeE2 encodes a protein of a predicted 847 amino acids. The C-terminus of speE2 contains the same dSAM domain as speE1 but also contains predicted transmembrane domains at its N-terminus. Using a transmembrane domain prediction and protein localization tool, TMHMM, we found that the SpeE2 protein has a total of 13 predicted transmembrane helices [21]. We found that the spermidine synthase domain of SpeE2 is predicted to be in the periplasm (Figure 4B). This suggests that SpeE2 may represent a novel class of periplasmic spermidine synthase proteins
Because the cytoplasmic speE1 gene cannot complement ISS activity in a speE2 mutant of CH267 or UW4, we wondered if there are cognate periplasmic variants of the remainder of the canonically cytoplasmic spermidine synthase pathway. Spermidine biosynthesis can occur via two known pathways (Figure 4C). The first, by speE, occurs when the aminopropyl group of dSAM is transferred to putrescine. SAM decarboxylation to dSAM occurs via speD [16]. The second spermidine synthase pathway is independent of speDE and converts putrescine to spermidine by a carboxyspermidine intermediate via a dehydrogenase (CASDH) and a decarboxylase (CASDC) [22]. Because speE2 also contains a predicted dSAM binding motif, this suggests that a dSAM biosynthesis gene should also be present in the genomes of ISS strains.
We surveyed the genomes of ~3800 sequenced Pseudomonas isolates for the presence of speD, speE, CASDH and CASDC and found that the majority of strains contained genes for either the speD/E1 cytoplasmic spermidine biosynthetic pathway or the CASDH/C biosynthetic pathway (Figure 4E shows the same set of strains shown in Figure 2). We generated a Pearson correlation matrix of the co-occurrence of spermidine biosynthesis genes. We found a near perfect correlation of components within each of the two known biosynthetic pathways (speD/speE1 and CASDH/C). Interestingly, there was a nearly perfect anti-correlation between the presence of the speD/speE1 and CASDH/C pathways (Figure 4D). This suggests that nearly all Pseudomonas make spermidine cytoplasmically, but they do it either via speD/E or via CASDH/C and do not contain both pathways. These pathways are both polyphyletic, as there are examples of sister clades containing opposite pathways, both at the species level (S3 Figure) and within the P. fluorescens subclade (S4 Figure). dSAM is required as a precursor to synthesize spermidine via speE. We could not identify a dSAM biosynthesis gene in any of our strains that has a predicted periplasmic localization. Moreover, while most strains with speE2 contain speD/E1 genes, Pseudomonas spp. PB106, PB120 and Pf-5 do not (Figure 4E). Beyond the strains we characterized, there was a weak negative correlation at the genus level between speD/E1 and speE2 (Figure 4D) indicating that strains that contain an speE2 gene do not necessarily contain speD. Because speE2 has a predicted dSAM binding domain, the source of the dSAM needed to catalyze spermidine production in strains lacking speD is elusive. These data suggest that speE2 must work with a novel dSAM biosynthesis gene to synthesize spermidine, generate spermidine in a dSAM-independent manner, or use an external source of dSAM, either from a eukaryotic host, or from other members of a bacterial community.
Additional roles for the ISS locus in host interactions
While speE2 is necessary for ISS, we wondered if the other 10 genes in the ISS locus are also involved in interactions with a plant host. We tested whether the pattern seen in Figure 4D, where speE2 is always associated with the same larger locus, would hold up across the genus Pseudomonas. When we analyzed our entire computational dataset of >3800 genomes from across Pseudomonas, we found that there was a strong correlation for the presence or absence of 9 of 11 genes (r > 0.9, Figure 5A). Moreover, we also found that these 9 co-occurring genes were frequently found in the same genomic region, as there were moderate to strong correlations for 9 of the 11 genes co-occurring in the same 50-kb genomic region (Figure 5B). From a phylogenomic standpoint, we found that these genes were broadly distributed throughout the Pseudomonas genus and co-occurred even in taxonomic groups far outside of the P. fluorescens clade (Figure 5C). Within the P. fluorescens clade, the ISS locus genes are frequently found in some clades, such as the koreensis and jessenii clades, which contain most of our isolates (Figure 5D). However, some clades are missing these genes entirely, such as the plant associated corrugata clade (Figure 5D). Together, these genomic data indicate that despite their polyphyletic distribution among divergent clades of Pseudomonas spp., these genes likely participate in conserved or similar functions.
Beyond the predicted role of SpeE2 as a spermidine synthase, the specific functional roles of genes in this locus are less clear. We identified an operon with 6 of the genes in the ISS locus with identical domain structure and organization that is involved in stress resistance and virulence in Francisella tularensis [23] (Figure 6). Another similar operon is associated with aerotolerance and virulence in Bacteroides fragilis [24]. Returning to our comparative genomics database, we found that these 6 genes comprise an operon broadly conserved in the Pseudomonas clade that is distinctly paralogous from the ISS operon (Figure 6). This raises the possibility that these six genes within the ISS locus contribute to host-bacterial interactions across diverse bacterial taxa and both plant and animal hosts (Figure 6).
Discussion
Plant root-associated (“rhizosphere”) microbes perform a diversity of functions that benefit their plant hosts including nutrient uptake and defense. Functional characterization of individual plant-associated bacterial and fungal strains of potential agronomic importance (i.e. growth promoters or nitrogen fixers) is widespread [5]. However, closely-related strains of bacteria can have very distinct effects on plant growth and defense [13], and these effects can be dependent on environmental context [1]. Lack of known correlations between microbial genotype and potential effects on plant hosts present a challenge when attempting to infer the effect that a microbe may have on its plant host from sequence identity alone.
Our use of comparative genomics and isolate phenotyping to identify the genetic basis of a complex microbial-derived trait indicates that this is an effective approach to identifying important microbial traits to improve plant health. For comparative genomics to be effective, traits should be controlled by single or limited genomic loci, and phylogeny should not be predictive of function. In this case, a close relative of ISS strains, Pseudomonas sp. Pf0-1 (>99% identical by full length 16S rRNA to the ISS strains) does not affect systemic defenses (Figure 1), which allowed us to use comparative genomic to identify the underlying basis. We previously used this approach to find the genomic basis of a pathogenic phenotype within a clade of commensals [14]. It has been previously observed that phylogeny is not predictive of function for ISR strains [13] suggesting that comparative genomics may be appropriate to find the basis of additional plant-associated traits.
Why spermidine applied to plant roots suppresses systemic immunity is not clear; the most direct role for polyamines in immunity is through oxidation and generation of reactive oxygen species (ROS) that promote immunity [25,26]. Spermine has been shown to modulate expression of redox and defense-related gene expression [27] and polyamines including putrescine, spermine and spermidine accumulate during pathogen infection [28]. Spermidine, spermine and related molecules can enhance resistance to pathogens in both plants and animals through breakdown via oxidases that result in generation of ROS [25,26]. P. syringae promotes generation of acetylated putrescine rather than spermidine, which is not readily oxidized and results in reduced ROS production and enhanced susceptibility [28]. We previously showed that the ISS strain CH267 suppresses a subset of SA-dependent gene expression and that ISS is SA-dependent [7]. As spermidine uptake should directly enhance resistance through accumulation of ROS, it seems more likely that spermidine applied to roots suppresses defense signaling through modulating local or systemic defense responses rather than direct uptake.
Similarly, the adaptive role of spermidine production by host-associated Pseudomonas remains to be determined. We and others previously showed that the polyamine putrescine promotes biofilm production in Pseudomonas and bacterial intracellular accumulation of putrescine inhibits rhizosphere fitness [29]. Spermidine can promote biofilm formation in Bacillus [30]. Our data show that predicted spermidine biosynthesis is perfectly conserved through the genus Pseudomonas, although there is a clear anti-correlation between the speD/E and CASDH/C pathways. The speE2 gene shows polyphyletic inheritance and occurs in diverse Pseudomonas spp. (Figure 5C). The evolutionary and ecological pressures that result in speE2 promoting bacterial fitness have yet to be elucidated.
The SpeE2 enzyme affects the plant host presumably by producing secreted spermidine, and the protein is predicted to lie within the periplasm (Figure 4A). A periplasmic localization might position the enzyme to utilize exogenous dSAM and convert either an internal or external supply of putrescine to spermidine. Complementarity of biosynthetic processes has been shown in other systems. For example, although most bacteria use cobalamins as enzyme cofactors, the majority do not have the ability to synthesize them [31]. While it is possible that a novel speD gene is present in the genomes of speE2-containing strains, it is possible that spermidine biosynthesis via speE2 might occur with an exogenous supply of dSAM, such as from a plant or another member of the microbial community.
While enhancement of systemic susceptibility is not an obviously agronomically useful plant trait, spermidine has been studied for its role in improving drought tolerance [17,20]. Additionally, several ISS strains promote growth and enhance resistance to insect pests [6,7]. Using ISS strains or spermidine might be beneficial for crops where drought or insects are the primary pressure on crop productivity. However, ISS illustrates the complexity of host microbe interactions and should be considered when engineering the microbiome.
Materials and Methods
Plant growth conditions
For all experiments, plants were grown in Jiffy-7 peat pellets (Jiffy Products) under a 12 h light/12 h dark at 22 °C temperature regime. Seeds were surface sterilized by washing with 70% ethanol for 2 minutes followed by 5 minutes in 10% bleach and 3 washes in sterile water. Seeds were stored at 4° C until use. Unless otherwise indicated, seeds were sowed in Peat pellets (Jiffy 7) and placed in a growth chamber under 12-hour days and 75 μM cool white fluorescent lights at 23° C.
Bacterial growth and 16S rRNA sequencing
Pseudomonas strains were cultured in LB or King’s B at 28 °C. New Pseudomonas strains were isolated from the roots of wild-grown Arabidopsis plants around eastern Massachusetts, USA and British Columbia, Canada as described [6]. New Pseudomonas isolates were preliminary identified based on fluorescence on King’s B and confirmed by 16S rRNA sequencing.
ISS assays
ISS and ISR assays were performed as described [7]. Briefly, Pseudomonas rhizosphere isolates were grown at 28 °C in LB medium. For inoculation of plant roots for ISR and ISS assays, overnight cultures were pelleted, washed with 10 mM MgSO4 and resuspended to a final OD600 of 0.02. Jiffy pellets were inoculated 9 days after seed germination with 2 mls of the indicated bacterial strains at a final OD600 of 0.02 (5x105 CFU g-1 Jiffy pellet). For spermidine and spermine treatments, 10 mM stocks were kept frozen until just before use. Just prior to use, they were diluted to the indicated concentration in water and 2 mLs were applied to the soil surrounding 3-week old plants.
For infections, the leaves of 5-week old plants were infiltrated with Pto DC3000 at an OD600 = 0.0002 (starting inoculum ~103 CFU/cm2 leaf tissue). Plants were maintained under low light (<75 µM) and high humidity for 48 hours. Leaf punches were harvested, ground, and plated to determine CFU counts.
16S rRNA sequencing, bacterial genome sequencing, assembly and phylogenomics
Bacterial DNA preps were performed using Qiagen Purgene Kit A. 16S rRNA was amplified using 8F and 1391R and sequenced using 907R.
Bacterial genomic library prep and genome sequence was performed as described [7]. Briefly, bacterial DNA was isolated using Qiagen Purgene Kit A and sonicated into ~500 bp fragments. Library construction was performed as described [7], individually indexed and sequenced using MiSeq V3 paired end 300 bp reads.
After barcode splitting, approximately 500,000 to 1 million reads were used for each sample to assemble draft genomes of the ISS strains Pseudomonas sp. CH235, PB100, PB101, PB103, PB105, PB106, PB120 and P. vancouverensis DhA-51. Genome assembly was carried out as previously described [7] and draft genomes are available from NCBI (see below).
Phylogenomic tree building
To generate the 29-taxon species tree used in Figures 2B and 4E, we made use of an alignment of 122 single-copy genes we previously found to be conserved in all bacteria [14]. From this amino acid alignment, we extracted 40,000 positions ignoring sites where >20% of the taxa had gaps. Using RAxMLv8.2.9, we inferred 20 independent trees under the JTT substitution model using empirical amino acid frequencies and selected the one with the highest likelihood. Support values were calculated through 100 independent bootstrap replicates under the same parameters.
To build the 3,886-taxon phylogeny of the Pseudomonas genus in Figures 5C and S1, the same 122-gene alignment was used. For computational feasibility, the alignment was randomly subsampled to 10,000 amino acid positions, again ignoring sites that were highly gapped (>20%). FastTree v2.1.9 was used to build the phylogeny using default parameters. The phylogeny was rooted to a clade of Pseudomonas identified as an outgroup to all other Pseudomonas spp. as previously described [14]. To more easily visualize this tree, we collapsed monophyletic clades with strong support (as determined by FastTree’s local Shimodaira-Hasegawa test) that correspond with major taxonomic divisions identified by Hesse et al. (2018).
To build the tree for the Pseudomonas fluorescens (Pfl) subclade seen in Figures 5D and S2, we identified 1,873 orthologs specific to the Pfl clade found in >99% of all strains in the clade and then aligned them all to the hidden Markov models generated by PyParanoid using hmmalign, prior to concatenation. This alignment had 581,023 amino acid positions, which we trimmed to 575,629 positions after masking sites with >10% of taxa with gaps. From this alignment, we randomly subsampled 120,000 sites for our final phylogenomic dataset. Using RAxMLv8.2.9, we inferred 20 independent trees under the JTT substitution model using empirical amino acid frequencies and selected the one with the highest likelihood. Support values were calculated through 100 independent bootstrap replicates under the same parameters.
Comparative Genomics
Comparative genomics analyses were performed by using our previously described framework for identifying PyParanoid pipeline and the database we built for over 3800 genomes of Pseudomonas spp. Briefly, we had previously used PyParanoid to identify 24,066 discrete groups of homologous proteins which covered >94% of the genes in the original database. Using these homolog groups, we annotated each protein-coding sequence in the newly sequenced and merged the resulting data with the existing database, generating presence-absence data for each of the 24,066 groups for 3,886 total Pseudomonas genomes.
To identify the groups associated with induction of systemic susceptibility, we compared the presence-absence data for 4 strains with ISS activity (Pseudomonas spp. CH229, CH235, CH267, and UW-4) and 1 strain with no activity (Pseudomonas sp. Pf0-1). We initially suspected that ISS activity was due to the presence of a gene or pathway (i.e. not the absence of a gene) and thus initially focused on genes present only in Pf0-1. We identified 29 groups that were present in the 4 ISS strains but not in Pf0-1. To obtain the correlation coefficients in Figs. 4D and 5A, we coded group presence or absence as a binary variable and calculated Pearson coefficients across all 3,886 genomes. To calculate the correlation coefficients in Fig. 5B, we split the genomic database into 50-kb contiguous regions and assessed group presence or absence within each region. Because this dataset is heavily zero-inflated, we ignored regions that had none of the 11 groups, taking the Pearson coefficient of the 11 genes over the remaining regions.
Initial annotation of the ISS groups was based on generic annotations from GenBank Further annotation of the 11 groups specific to the ISS locus was carried out using the TMHMM v2.0 server, the SignalP 4.1 server and a local Pfam search using the Pfam-A database from Pfam v31.0. To identify homologous genes in the genomes of Francisella tularensis subsp. holarctica and Bacteroides fragilis YCH46, we relied on locus tags reported in the literature which we confirmed using annotation based on another Pfam-A domain search.
Deletion of the speE2 gene and gene clusters
Deletions in the CH267 and UW4 strains were constructed by a two-step allelic exchange as described [29]. The flanking regions directly upstream and downstream of the spermidine biosynthesis cluster or the speE gene were amplified and joined by overlapping polymerase chain reaction (PCR) using genomic DNA as template and primers listed in Table 2. Following digest, the product was ligated into the pEXG2 suicide vector that contains the sacB gene for counter-selection on sucrose [36]. The recombinant plasmid was then transformed into calcium-competent E. coli DH5α by heat shock. After confirmation of correct insert by PCR and sequencing, the plasmid was transformed into WM3064 [37]. Conjugation of plasmid into CH267 and UW4 from WM3064 was performed by biparental mating on King’s B media supplemented with diaminopimelic acid, and transconjugants were selected using 10 µg/mL gentamicin and 15 µg/mL nalidixic acid. The second recombination leading to plasmid and target DNA excision was selected for by using sucrose counter-selection. Gene deletions in CH267 and UW4 were confirmed by PCR amplification of the flanking regions with primers listed in Table 2, agarose gel electrophoresis and Sanger sequencing.
Author Contributions
C.H., R.A.M., and P.B. designed experiments. P.B. Y.S. and C.H.H. performed experiments. C.H., R.A.M., X. L. analyzed data. and R.A.M. performed genome assembly, annotation, phylogenetic analysis and comparative genomics. C.H.H., P.B. and R.A.M. wrote the manuscript with input from all.
Data Availability
Data for the Whole Genome Shotgun project has been deposited at DDBJ/ENA/GenBank under the accessions RRZJ00000000 (CH235), RRZK00000000 (DhA-51), RWIL00000000 (PB103), RWIM00000000 (PB106), RWIN00000000 (PB120), RWIO0000000 (PB105), RWIQ00000000 (PB100), and RWIR00000000 (PB101). The versions described in this paper are versions RRZJ01000000 (CH235), RRZK01000000 (DhA-51), RWIL01000000 (PB103), RWIM01000000 (PB106), RWIN01000000 (PB120), RWIO0100000 (PB105), RWIQ01000000 (PB100), and RWIR01000000 (PB101).
Declaration of interests: The authors declare no competing interests.
S1 Table. Unique loci identified in comparative genomics. The genome content of 4 ISS strains (CH267, CH235, UW4 and CH229) was compared with the closely-related non-ISS strain Pf0-1. 17 predicted protein-coding genes were identified.
S1 Figure. Distribution of loci identified by comparative genomics ISS loci across Pseudomonas strains. Comparative genomics between ISS strains UW4, CH229, CH235 and CH267 (black arrows) and non-ISS strain Pf0-1 (red arrow) identified 17 predicted protein-coding genes that were absent in Pf0-1 and present in the ISS strains. 11 of these genes were found in a single cluster (box) and were absent in the non-ISS strain WCS365.
S2 Figure. The ISS locus is highly variable between closely-related strains The 11 genes in the ISS locus are present in the ISS strains Pf0-1, CH235, CH267 and CH299 but absent in Pf0-1. Genes in the ISS locus are colored as in the key at the bottom of the figure and in Figure 2. Conserved genes not unique to the ISS strains are colored similarly among strains; genes in gray are not conserved between strains at this locus. In CH229, Pf0-1 and CH267 the genes flanking the ISS locus are conserved in the same orientation suggesting a recent insertion or deletion event.
S3 Figure. Distribution of speD/E1 and the CASDH/C spermidine biosynthetic pathway across Pseudomonas species. While some species contain both pathways (such as P. fluorescens) many others (such as P. putida or P. aeruginosa) contain exclusively the speD/E1 or CASH/C pathway. The number of strains included in each genus is shown in parentheses.
S4 Figure. Distribution of speD/E1 and the CASDH/C spermidine biosynthetic pathway within subgroups in the P. fluorescens clade. With the exception of Pf-5 which is within P. protegens, all other ISS strains described fall within P. madelii, P. jessenii and P. koreensis clades as defined by Hesse et al., (2018). The number of strains included in each group is shown in parentheses.
Acknowledgements
This work was supported by an NSERC Discovery Grant (NSERC-RGPIN-2016-04121) awarded to C.H.H., a Life Sciences Research Foundation Fellowship from the Simons Foundation awarded to R.A.M., a fellowship from China Postdoctoral Science Foundation awarded to Y.S., and an NSERC CGS-M award to Z. L.