Functional diversification despite structural congruence in the HipBST toxin-antitoxin system of Legionella pneumophila

Abstract

Toxin-antitoxin (TA) systems are abundant genetic modules in bacterial chromosomes and on mobile elements. They are often patchily distributed, as a consequence of horizontal exchange, and the breadth of their functionality in bacterial physiology remains unknown. Here, we characterize a TA system in Legionella pneumophila that is highly conserved across Legionella species. This system is distantly related to Escherichia coli HipBST, and we demonstrate that it is a functional tripartite TA system (denoted HipBST_Lp). We identify HipBST_Lp homologs in diverse taxa, yet in the Gammaproteobacteria, these are almost exclusively found in Legionella species. Notably, the toxin HipT_Lp was previously reported to be a pathogenic effector protein that is translocated by L. pneumophila into its eukaryotic hosts. Contrary to this, we find no signal of HipT_Lp translocation beyond untranslocated control levels and make several observations consistent with a canonical role as a bacterial toxin. We present structural and biochemical insights into the regulation and neutralization of HipBST_Lp, and identify key variations between this system and HipBST_Ec. Finally, we show that the target of HipT_Lp is likely not conserved with any characterized HipA or HipT toxin. This work serves as a useful comparison of a TA system across bacterial species and illustrates the molecular diversity that exists within a single TA family.

IMPORTANCE

Toxin-antitoxin (TA) systems are parasitic genetic elements found in almost all bacterial genomes. They are exchanged horizontally between cells and are typically poorly conserved across closely related strains and species. Here, we report the characterization of a tripartite TA system in the bacterial pathogen Legionella pneumophila that is highly conserved across Legionella species genomes. This system (denoted HipBST_Lp) is a distant homolog of the recently discovered split-HipA system in Escherichia coli (HipBST_Ec). We present bioinformatic, molecular, and structural analyses of the divergence between these two systems and the functionality of this newly described TA system family. Furthermore, we provide evidence to refute previous claims that the toxin in this system (HipT_Lp) possesses bifunctionality as an L. pneumophila virulence protein. Overall, this work expands our understanding of the split-HipA system architecture and illustrates the potential for undiscovered biology in these abundant genetic elements.

INTRODUCTION

Bacterial genomes are a mosaic of genes that are highly conserved and often critical for growth (the core genome) or variable and typically non-essential (the accessory genome). The accessory genome is composed of genes frequently acquired through horizontal transfer, and the plasticity of this portion of the genome can promote organismal diversification while maintaining robustness to changing selective pressures (1). The contribution of accessory genomic elements to bacterial physiology is, therefore, of critical importance to understand, in particular with regard to genes that are considered accessory but demonstrate patterns of significant conservation across species (2).

Toxin-antitoxin (TA) systems are abundant and widely distributed components of the accessory genome (3). They are typically bipartite genetic modules found in nearly all bacterial chromosomes—often in multiple types and copies—as well as on mobile DNA elements (4, 5). Despite their widespread occurrence, these genetic systems are typically poorly conserved across related strains and species (5, 6). Several functions have been described for TAs, with substantial evidence for their role in phage defense and DNA stabilization (7), while considerable controversy surrounds their contribution to bacterial persistence (3, 8). TA systems are broadly divided into eight major groups based on antitoxin activity and molecule type (7, 9), with the most heavily studied being the type II systems. Type II TA systems are composed of a protein toxin that can reversibly arrest cellular growth and a protein antitoxin that neutralizes the toxin through direct physical interaction. The toxin is typically refractory to proteolytic degradation, whereas the antitoxin is often rapidly degraded by the cell and needs to be frequently replenished (10). This leads to cellular addiction to the TA system—a property that allows them to act as selfish elements and consequently, to stabilize mobile DNA (11).

Numerous mechanisms of cell toxification and toxin neutralization have been reported for type II systems (12). Genetic diversification has also been observed within the same TA system, such as the recent discovery of a split-kinase variant of the well-studied HipBA system in Escherichia coli (13). HipBA is composed of the HipA kinase toxin and HipB antitoxin and has been implicated in regulating bacterial persistence (14, 15). The newly described TA family HipBST is related to HipBA but displays a tripartite architecture: an N-terminal subdomain in the HipA kinase is encoded by a separate protein (HipS) and this functions as the antitoxin in the system (13). The catalytic core of HipA is expressed as a single protein (HipT) that retains toxin activity, while the canonical antitoxin HipB appears to transcriptionally regulate the locus. This unusual tripartite variant is, therefore, considerably divergent from the HipBA system despite their relatedness.

Most knowledge of TA biology has been gleaned from studying a small number of bacterial species, but given their modularity and portability, it remains unclear whether homologous systems function similarly within different bacterial hosts (7). To address this, we sought to investigate the TA landscape within the intracellular pathogen Legionella pneumophila. L. pneumophila is a Gram-negative environmental bacterium found within most global freshwater environments, where it parasitizes numerous protozoan species (16). To maintain this host range, L. pneumophila utilizes an arsenal of translocated virulence proteins (termed “effectors”) to transform the host cell into a replicative niche (17). The genome of L. pneumophila contains the largest known assemblage of effectors (~10% of genes), many of which contain homology to eukaryotic proteins and have likely been acquired via horizontal gene transfer from its hosts (18). Conversely, it encodes only a small number of predicted TA systems (19) and is largely devoid of mobile elements (such as prophage), which contribute to TA dissemination (7). We, therefore, wondered whether unique TA biology would be found in a species with the genomic constraints of an intracellular pathogen that has prioritized the acquisition of foreign genetic material from its hosts.

Here, we report the characterization of a TA system in L. pneumophila that is highly conserved across Legionella species genomes. This system is distantly related to the HipBST module in E. coli (herein referred to as HipBST_Lp and HipBST_Ec, respectively) and the toxin HipT_Lp was previously reported to be an effector that is translocated into the eukaryotic host (20). Recently, a bi-functional role for HipT_Lp was proposed with activity in both the host and the bacterial cell (21). Contrary to this, we demonstrate that HipT_Lp is not translocated at a level greater than a negative (non-effector) control and has no obvious role within the eukaryotic cell. Instead, we show that hipBST_Lp encodes a functional tripartite TA system with clear effects on bacterial growth. Interestingly, HipBST_Lp represents a heretofore undiscovered subclade of HipBST systems that are widely distributed outside of the Gammaproteobacteria class but are almost exclusively found in the Legionellales order within that group. We demonstrate that the toxin HipT_Lp is a kinase and report a survey of its phosphoproteome within intoxicated cells. We find that despite their shared architecture, HipT_Lp does not appear to target the same cellular substrate as HipT_Ec or any characterized HipA toxin. We additionally present structural and biochemical insights into the HipBST_Lp system, including its mechanism of neutralization and sites of divergence from HipBST_Ec. Overall, this work provides a comparison of a TA system across distant bacterial species and emphasizes that generalizations about system functionality should be made with caution. Instead, there appears to be considerable evolutionary diversity that remains underexplored within these abundant genetic elements.

RESULTS

Legionella pneumophila encodes a highly conserved HipBST locus

The genome of L. pneumophila has seven predicted TA loci in the Toxin-Antitoxin Database (19)—a comparatively small number relative to other studied bacterial species. We first investigated the distribution of these systems across 58 representative species of Legionella in the NCBI Refseq database (Table S4) to determine whether any showed a signature of high conservation. Surprisingly, despite the mobile nature of TA systems and their typically patchy distribution, we identified one system (lpg2368-70) that was present in nearly half of the Legionella species genomes (28/58 species; Fig. 1A). Additionally, eight other species had homologs of individual genes from this locus (Fig. 1A). This level of conservation is unusual for TA systems and was striking given the genomic diversity within this genus (22). Even effectors, which constitute a large portion of the Legionella accessory genome and are critical to pathogenicity, are poorly conserved across species (22, 23). Interestingly, lpg2368-70 is similar to the well-characterized HipBA module (24) but possesses an atypical tripartite architecture in which the HipA homolog is split into two separate open reading frames (ORFs) (Fig. 1B). Given its substantial conservation across Legionella species (>80% sequence similarity across homologs) and unusual genetic organization, we chose to investigate this putative TA system and its functionality in L. pneumophila.

FIG 1 Legionella species contain a highly conserved and taxonomically restricted HipBST TA system. (A) A core genome phylogeny of 58 representative species in the Legionella genus (Table S4) showing the conservation of Lpg2368-70 (HipBSTLp) homologs. The Legionella core genome was determined using OrthoMCL (25) to identify ORFs with conservation across all 58 species. Amino acid sequences were aligned with MUSCLE (26), a phylogenetic tree was constructed with RAxML (27), and the tree was rooted with Legionella geestiana as the outgroup based on previous phylogenies (22, 23). Tree annotation and visualization were performed with ggtree (28) and FlaGs (29). The scale bar denotes substitutions per site. Bootstrap values are indicated for each node. (B) Schematic of the HipBAEc, HipBSTLp, and HipBSTEc TA systems. Lpg2368 has homology to the canonical antitoxin HipB, whereas Lpg2369 and Lpg2370 align to the N and C termini of the HipA toxin, respectively. (C) An unrooted phylogeny of HipBA and HipBST homologs retrieved from the NCBI Refseq database. Homology searching was performed using cblaster (30) with seed sequences from five different TA systems: E. coli K12 HipBA (HipBAEc; NP_416025.1, NP_416024.1), Shewanella oneidensis MR-1 HipBA (HipBASo; AAN53783.2, AAN53784.1), Bacteroides uniformis HipBA (HipBABu; WP_149924066.1, WP_149924064.1), E. coli O127 HipBST (HipBSTEc; WP_000563102.1, WP_001346664.1, WP_001262465.1), and L. pneumophila HipBST (HipBSTLp; AAU28429, AAU28430, AAU28431). Retrieved sequences were clustered using MMseqs (31) to identify sequence-level representatives, aligned with MAFFT (32), and trimmed with trimAI (33). HipS/HipT sequences were concatenated prior to alignment with the HipA sequences, whereas HipB sequences were aligned separately. The phylogeny was constructed using IQ-TREE (34) and visualized with iTOL (35). The scale bar denotes substitutions per site. Bootstrap values are indicated for major nodes. Raw and sequence-level representative hits are provided in Table S5. (D) Distribution of HipBA and HipBST TA systems across diverse bacterial taxa. A bacterial phylogeny was retrieved from TimeTree (36) for species containing at least one system in our homology search (Table S6). These were filtered to remove species without a complete genome, to ensure that genome incompleteness did not influence system detection. The presence of a homolog for each system is indicated for each species and systems are ordered by similarity of taxonomic distribution. The Pseudomonadota phylum is colored light gray, the Legionella genus is colored purple, and L. pneumophila is indicated with an asterisk.

The lpg2368-70 locus encodes three genes within a predicted operon (Fig. 1B) (37), with each ORF overlapping by 4 bp. The latter two genes were originally annotated as having homology to domains of the HipA_Ec toxin; the proteins encoded by lpg2369 and lpg2370 align with the N-terminus and C-terminus of HipA_Ec, respectively. The third protein encoded by lpg2368 has sequence similarity (~30%) to the antitoxin HipB_Ec. These observations led us to hypothesize that Lpg2368-70 constituted a split-kinase TA system that was distantly related to HipBA. During the course of our studies, this was confirmed by the discovery of a similar system in E. coli (denoted HipBST_Ec) (13). From this point forward, we refer to Lpg2368-70 as HipBST_Lp. Despite this similar architecture, HipT_Lp bears only remote sequence similarity to either HipA_Ec (~28% identity) or the newly characterized HipT_Ec (~29% identity) (Fig. S1A), and HipB_Lp has no similarity to HipB_Ec of HipBST_Ec. This may explain why the HipBST systems in Legionella species were not detected in previous work (24). While they exhibit broad differences at the sequence level, the core catalytic residues in HipA_Ec and HipT_Ec are conserved in HipT_Lp (Fig. S1A). HipBST_Lp, therefore, provided the opportunity to compare two variants of the newly described HipBST family.

The HipBST_Lp system is phylogenetically and taxonomically distinct from HipBST_Ec

As HipBST_Lp was not identified previously, we sought to determine whether closely related modules were also undiscovered in other bacterial genomes and how they fit into the broader Hip phylogeny. We queried the NCBI Refseq database for HipBA and HipBST homologs using five diverse systems as seed sequences: HipBA from E. coli K-12 (HipBA_Ec), HipBA from Shewanella oneidensis (HipBA_So), HipBA from Bacteroides uniformis (HipBA_Bu), HipBST from E. coli O127 (HipBST_Ec), and HipBST from L. pneumophila (HipBST_Lp). While HipBA_Ec and HipBA_So have been characterized previously (24, 38), HipBA_Bu was included in order to expand our search space, as this system was found in close proximity to a HipBST_Lp homolog and was itself quite divergent from the other HipA/HipT sequences (Fig. S1B). From this search, we identified 164 HipBA_Ec, 37 HipBA_So, 184 HipBA_Bu, 31 HipBST_Ec, and 86 HipBST_Lp sequence-level representative systems (Table S5) and used these sequences to construct a HipBST/HipBA phylogeny. Despite their shared tripartite architecture, HipBST_Lp homologs cluster in a separate subclade from the HipBST_Ec sequences (Fig. 1C), indicating system-level divergence even within the HipBST family. We discovered numerous HipBST_Lp homologs that are only distantly related to the 48 previously reported systems (24), thereby expanding the sequence space of HipBST. In fact, the majority (60/86) of our HipT_Lp representative sequences share less than 60% identity with any of the 48 sequences previously identified. In order to investigate the contribution of the individual protein components to the concatenated system phylogeny, we further constructed individual phylogenies for the HipB and HipST/HipA homologs separately (Fig. S1C). For this purpose and owing to the size disparity between HipS and HipA sequences, in addition to the likely constrained coevolution of HipS-HipT pairs, we treated HipST as a single concatenated sequence. Interestingly, while both component phylogenies largely agree with the system phylogeny, some incongruence is observed in the HipB phylogeny with regard to the clustering of HipB_So-like sequences in the HipBST clade.

We next examined whether the various HipBA/HipBST homologs displayed any pattern of taxonomic distribution. To this end, we mapped the presence or absence of each system type onto a phylogeny of species that contained at least one system from our search and had a complete genome assembly in the Refseq database (Table S6). From this, we observed that HipBST_Lp homologs in the Gammaproteobacteria are almost exclusively found within the Legionellales order, whereas most detected tripartite systems in this class are instead homologs of HipBST_Ec (Fig. 1D; Fig. S1D). In contrast to their restricted distribution within the Gammaproteobacteria, HipBST_Lp homologs are widely distributed within distant taxonomic groups, particularly the FCB (Fibrobacterota, Chlorobiota, and Bacteroidota) superphylum. Interestingly, while HipBA_Bu homologs are prevalent throughout the bacterial phylogeny, homologs of HipBA_Ec and HipBA_So appear to be mostly restricted to the Pseudomonadota phylum. In summary, we used the previously undetected HipBST_Lp system to broaden the sequence space of known HipBST homologs. We observed that HipBST_Lp and HipBST_Ec sequences constitute distinct HipBST subclades, revealing considerably more genetic and taxonomic diversity within the HipBST TA family.

Several lines of evidence highlight the pertinence of investigating the effects of HipT_Lp on the bacterial cell

It has been previously reported that HipT_Lp is translocated by L. pneumophila into its eukaryotic host via its Type IV secretion system (20). This observation has significant potential to direct the future investigation of HipT_Lp towards activities against the host and away from any role as a canonical bacterial toxin. As such, we began our own functional analysis by carefully quantifying the levels of HipT_Lp translocation into host cells. Notably, translocation of HipT_Lp was previously detected at very low efficiency using the TEM-1 assay (21), in which TEM-1 β-lactamase is fused to a protein of interest and expressed in L. pneumophila cells infecting monolayers of differentiated U937 cells (39). Translocation is determined by monitoring emission from a fluorescent substrate within the host cells, which shifts from emitting green fluorescence to blue fluorescence upon cleavage of an internal β-lactam ring by the TEM-1 fusion. Given the low level of translocation previously reported, we sought to confirm this observation. Using identical methodology, we observed that the translocation signal of HipT_Lp is equivalent to that of the negative control FabI—a cytosolic protein with no function in the host (Fig. 2A). We observed the same negative results at several different multiplicities of infection (MOIs) and confirmed fusion protein expression by immunoblotting (Fig. S2). To test whether the toxicity of HipT_Lp was impacting its observed level of translocation, we additionally tested two substitutions in predicted catalytic residues (D197Q, D219Q) that are conserved with HipA_Ec (Fig. S1A) (40). However, no translocation signal above control levels was detected for these mutants (Fig. 2A) despite their increased expression levels (Fig. S2). Further arguing against its role as an effector, we observe that overexpression of HipT_Lp does not display any growth inhibitory effects in yeast (Fig. 2B), which is a common phenotype of effector expression (41). This is consistent with our identification of HipBST_Lp homologs within non-pathogenic bacterial species (Fig. 1D; Fig. S1D) and argues that, while effects against the eukaryotic host remain a tantalizing possibility, it remains pertinent to thoroughly investigate HipT_Lp for canonical toxin activity within the bacterial cell.

FIG 2 HipTLp translocation is not detectable above the levels of established negative controls. (A) TEM-1 β-lactamase translocation assays were performed with Lp02 cells infecting monolayers of differentiated U937 cells. Uninfected cells were used to determine background fluorescence and the cytosolic protein FabI was used as a negative control. The positive control used was the L. pneumophila effector RalF. Both controls, in addition to wild-type and mutant HipTLp(D197Q, D219Q), were expressed from the pXDC61 vector (induced with 500 µM IPTG) as a fusion with the TEM-1 protein and infections were performed with the indicated MOIs. Quantified fluorescence is reported as the ratio of blue fluorescence (translocation) to green fluorescence (no translocation). The bar chart shows mean ± standard deviation of three biological replicates and is representative of three independent experiments. The dashed gray line indicates a blue/green fluorescence ratio of 1, which was used as a threshold for translocation. (B) Expression of HipTLp in S. cerevisiae BY4742 cells. Yeast cultures carrying HipTLp cloned into the pAG425GAL expression vector, or a vector-only control, were grown overnight and spotted in serial dilutions onto media containing either 2% glucose (repression) or 2% galactose (expression) and grown for 2 days. Spotting assays were performed in biological triplicate and images are representative of three independent experiments.

HipBST_Lp is a functional tripartite toxin-antitoxin system

We next wondered whether the HipBST_Lp module could serve as a functional TA system. To test this, we cloned HipT_Lp into an inducible plasmid and examined the impact of its expression on bacterial growth. HipT_Lp expression inhibited growth in both an L. pneumophila strain with the endogenous locus deleted (∆hipBST) and E. coli (Fig. 3A; Fig. S3A). This effect was neutralized in L. pneumophila by the co-expression of HipS_Lp but not HipB_Lp (Fig. 3B; Fig. S3B). No growth inhibitory effects were observed for the expression of HipB_Lp or HipS_Lp, either alone or in combination (Fig. S3C). These results are consistent with the reported functionality of the HipBST_Ec system (13) and confirm that HipBST_Lp is a functional TA module.

FIG 3 HipBSTLp is a functional tripartite TA system that restricts growth via HipTLp kinase activity. (A) Expression of HipTLp in L. pneumophila cells with the endogenous hipBST locus deleted (∆hipBST) and E. coli (TOP10) cells. Expression was induced with IPTG (100 µM) for the pJB1806 expression vector used in L. pneumophila and arabinose (0.2%) for the pBAD18 vector used in E. coli. (B) HipTLp co-expression with HipBLp and HipSLp in L. pneumophila ∆hipBST cells. HipTLp was expressed from the pJB1806 vector, and HipBLp and HipSLp were co-expressed from the pNT562 vector. Expression was induced with IPTG (100 µM). (C) ADP-Glo kinase assay with purified recombinant His6-SBP-tagged HipTLp. Both wild-type HipTLp and HipTLp with substitutions in two conserved catalytic residues (D197Q, D219Q) were assayed. Reactions contained 1 µg of protein and were incubated at 37°C for 30 min. (D) ADP-Glo kinase assay with purified recombinant His6-SBP-tagged HipBLp, HipSLp, and HipTLp. Reactions contained 1 µg of each protein and were incubated at 37°C for 30 min. (E) Expression of HipTLp with mutations in two conserved catalytic residues (D197Q, D219Q) in L. pneumophila ∆hipBST and E. coli (TOP10) cells. Expression was induced with IPTG (100 µM) for the pJB1806 expression vector used in L. pneumophila and arabinose (0.2%) for the pBAD18 vector used in E. coli. All growth curves show the mean ± the standard deviation of three biological replicates. Data are representative of three independent experiments. All kinase assays show the mean ± the standard deviation of three technical replicates. Data are representative of a minimum of two independent experiments. Statistical hypothesis testing for the kinase assays was performed with a two-tailed Student’s t-test and each sample was compared to HipTLp. ***P-value < 0.0001; α (0.05) was Bonferroni corrected for multiple hypothesis testing.

HipT_Ec has been shown to be a kinase and this activity is required for its toxicity to the cell (13). We, therefore, tested whether HipT_Lp also displayed kinase activity. We purified wild-type HipT_Lp, HipT_Lp(D197Q), and HipT_Lp(D219Q) (Fig. S3D) and tested for ATP hydrolytic activity using the ADP-Glo kinase assay. Wild-type HipT_Lp demonstrated substantial activity, whereas the activity of both mutants was abrogated (Fig. 3C). ATP hydrolysis was also reduced for wild-type HipT_Lp by the addition of the kinase inhibitor 5′-fluorosulfonylbenzoyl-5′-adenosine (FSBA) (Fig. S3E). Co-incubation of purified recombinant HipS_Lp (Fig. S3D) with HipT_Lp neutralized activity in vitro (Fig. 3D), whereas co-incubation with HipB_Lp (Fig. S3D) did not and instead an increase in activity was observed. The cause of this is unclear, though it may be a consequence of promiscuous phosphorylation of HipB_Lp by HipT_Lp. The D197Q and D219Q substitutions also eliminated the growth inhibitory phenotype of HipT_Lp in the ∆hipBST strain and E. coli (Fig. 3E; Fig. S3F and G), confirming that kinase activity is required for cellular toxicity. Taken together, these results demonstrate that HipBST_Lp is a functional TA system in which HipT_Lp toxicity results from its kinase activity, and neutralization is performed by HipS_Lp rather than HipB_Lp.

The HipBST_Lp system has the capacity for complex autoregulatory dynamics

Antitoxins of type II TA systems, such as HipB in the HipBA system, neutralize their cognate toxins via direct physical interaction (12). As HipS_Lp functions as the antitoxin in HipBST_Lp, we wondered if it physically interacts with the HipT_Lp toxin. To test this, we examined the binary protein-protein interactions of the HipBST_Lp system using the yeast two-hybrid (Y2H) assay. Our results showed that HipT_Lp could physically interact with both HipB_Lp and HipS_Lp independently, but these two proteins could not themselves physically interact (Fig. 4A; Fig. S4A). To test whether all three proteins could stably associate, we repeated the experiment with the addition of constitutively expressed HipT_Lp. In this setup, HipB_Lp and HipS_Lp came in close enough proximity to produce a detectable interaction (Fig. 4A; Fig. S4A), suggesting that HipT_Lp could interact with both proteins simultaneously and serve as a scaffold for a tripartite complex.

FIG 4 The HipBSTLp system has the capacity for complex autoregulatory dynamics. (A) Y2H experiments testing for binary physical interactions in the HipBSTLp tripartite proteins. Genes cloned into the pDEST-AD and pDEST-DB Y2H vectors are indicated, and representative images of S. cerevisiae Y8800 growth in the presence (+His) and absence (−His) of histidine are shown. Growth in the absence of histidine can only occur through a stable protein-protein interaction, due to the reconstitution of the GAL4 transcription factor (AD and DB domains) and subsequent expression of the HIS3 reporter gene downstream of the GAL1 promoter. Yeast two-hybrid experiments were also performed with a third protein constitutively expressed from the pAG416 vector (Y3H). Genes cloned into pAG416 are indicated. (B) Representative EMSA performed with recombinant purified HipBSTLp proteins and a 200-bp DNA fragment (10 nM) encompassing the promoter region upstream of the hipBSTLp locus. HipBLp was added to DNA at concentrations of 1, 2.5, 5, 10, and 20 nM, and HipSLp/HipTLp was added at a concentration of 20 nM. (C) Schematic of the operator sites upstream of hipBSTLp and consensus sequences for each site across 28 Legionella species. Nucleotide substitutions in O2 and O3 relative to O1 are colored red. Also displayed is a representative EMSA performed with 50-bp fragments from the 250-bp region upstream of hipBSTLp . HipBLp was added to DNA (10 nM) at a concentration of 2.5 nM. (D) Representative EMSA with 50-bp control DNA (10 nM) incubated with HipBLp (2.5 nM). Controls consisted of the 50-bp fragment containing O1 with a scrambled O1 sequence (OS), inversion of the downstream inverted repeat in O1 (OI; colored red), or a 50-bp random DNA control fragment (C) with the O1 sequence added (CO). The O1 sequence, where applicable, is displayed. (E) Representative EMSA showing increasing concentrations of HipSLp and HipTLp added to HipBLp-DNA complexes (10 nM HipBLp, 10 nM DNA). HipSLp and HipTLp were added at concentrations of 10, 40, 100, and 200 nM. All EMSA gels were stained with SYBR Green and protein-DNA complexes are indicated. Dashes indicate the absence of added protein.

In the HipBA system, HipB both neutralizes HipA and regulates transcription of the hipBA locus (42). We next asked whether toxin neutralization and autoregulation of hipBST_Lp transcription are decoupled, given the tripartite architecture of HipBST_Lp. We tested all three proteins for the binding of 200-bp upstream (hipBST_Lp promoter) sequence using the electrophoretic mobility shift assay (EMSA). Co-incubation with HipB_Lp slowed the migration of the DNA fragment (Fig. 4B), whereas HipS_Lp and HipT_Lp did not, indicating that transcriptional regulation of the hipBST_Lp locus is indeed distinct from toxin neutralization. Interestingly, increasing concentrations of HipB_Lp produced a gel shift pattern that was absent when compared to two unrelated promoter controls, which showed only nonspecific DNA binding at higher concentrations (Fig. S4B).

Given these findings, we searched the hipBST_Lp promoter region for putative DNA binding motifs and inverted repeats. This region does not contain either of the conserved operator sequences found in hipBA promoters (38, 43); however, we did detect an 18-nucleotide stretch containing an inverted repeat motif 24 nucleotides upstream of the HipB_Lp start codon (Fig. 4C). This sequence (referred to as O1) was almost perfectly conserved across all 28 Legionella species containing HipBST homologs (Fig. S4C) and overlaps with the predicted promoter of these systems. By comparing alignments of genomic sequence further upstream of hipBST systems, we detected two additional sites with asymmetrical inverted repeat sequences (termed O2 and O3) that are nearly identical to O1 except for two substitutions at the same position within both sequences (Fig. 4C; Fig. S4C). These core differences reflect the consensus sequences, however, and more variation is observed within O2 and O3 sequences across species relative to O1. The three putative operator sites are evenly spaced, with 53-bp inter-operator gaps in nearly all species, and no additional sites were detected further upstream or within the hipB gene sequence. To test whether HipB_Lp could bind each operator site, we performed EMSA experiments with 50-bp fragments of upstream DNA that either contained an operator site, did not contain one, or split one. Fragment migration was shifted only in the presence of a complete operator site (Fig. 4C), regardless of HipB_Lp concentration (Fig. S4D). We next examined HipB_Lp specificity for the O1 operator site—as it was the most conserved—by mutating the 50-bp fragment containing O1 to either completely scramble the O1 sequence or abolish its inverted repeat symmetry. Scrambling O1 was sufficient to abrogate the shift in fragment migration; however, this shift was still observed with abolished inverted repeat symmetry (Fig. 4D; Fig. S4E). Finally, we tested two additional 50-bp fragments containing either random 50-bp sequence or random sequence flanking an insertion of the O1 site. From this, we observed that the presence of the O1 site resulted in improved fragment shift efficiency in a concentration-dependent manner (Fig. 4D; Fig. S4E). In summary, we identified a conserved set of operator sites within the upstream promoter region of hipBST systems that are both necessary and sufficient for HipB binding.

We next wondered what effect co-incubation with HipS_Lp and HipT_Lp would have on the HipB_Lp binding of promoter DNA. To address this, we performed EMSAs with increasing concentrations of either protein alone or together added to the HipB_Lp-DNA complex. The addition of HipS_Lp to HipB_Lp had no impact on DNA migration, regardless of concentration, whereas co-incubation with HipT_Lp produced a secondary gel shift indicative of the formation of a HipT_Lp-HipB_Lp-DNA complex (Fig. 4E; Fig. S4F). Furthermore, the co-incubation of HipB_Lp with HipS_Lp and HipT_Lp resulted in a secondary gel shift relative to HipT_Lp alone. This additional shift in the presence of all three proteins also occurred at a lower concentration of HipS_Lp/HipT_Lp relative to HipT_Lp alone (Fig. 4E; Fig. S4F). Interestingly, the highest stoichiometric excess of either HipT_Lp or HipS_Lp/HipT_Lp relative to HipB_Lp also appeared to result in a decrease in unbound DNA in the reaction. These results demonstrate the in vitro formation of HipB_Lp-DNA and HipT_Lp-HipB_Lp-DNA complexes and are consistent with the formation of a HipS_Lp-HipT_Lp-HipB_Lp-DNA complex. Taken together, our findings reveal the potential for multiple binary and ternary interactions in the HipBST_Lp system and provide insights that distinguish transcriptional autoregulation by HipB_Lp from other HipB proteins. HipB_Ec has also been reported to regulate the transcription of the hipBST_Ec locus (44); however, our data demonstrate the occurrence of a HipB_Lp-HipT_Lp-DNA complex and suggest that interaction dynamics between binary and ternary protein combinations may serve as additional layers in a larger regulatory program.

HipBST_Lp neutralization exploits the P-loop ejection mechanism of HipA_Ec

To understand the structural basis for neutralization in the HipBST_Lp system, we solved the apo structure of HipB_Lp and the co-crystal structure of HipS_Lp-HipT_Lp (Table S7). HipB_Lp is highly similar to HipB_Ec (RMSD 1.6–2.2 Å over approximately 70 Cɑ atoms) and adopts a dimeric helix-turn-helix conformation, consistent with its capacity to bind DNA (Fig. 5A). Strikingly, the structural conformation of the HipS_Lp-HipT_Lp complex is nearly identical to HipA_Ec (RMSD 2.8 Å over 421 Cɑ atoms, when considering HipS_Lp and HipT_Lp as a single chain) (Fig. 5B), despite the functional divergence between the systems. Given that HipS_Lp serves the role of antitoxin, we wondered how neutralization could be achieved in this orientation. To address this, we compared the HipS_Lp-HipT_Lp structure to previously solved HipA_Ec structures. Independent of neutralization by HipB_Ec, HipA_Ec is capable of self-inhibition through intermolecular phosphorylation of its S150 residue (45). This site is found within the catalytic P-loop, and its phosphorylation or mutation to alanine yields a conformational shift whereby the P-loop becomes ejected and solvent exposed, thereby inactivating the toxin (Fig. 5C). Part of the P-loop in our HipS_Lp-HipT_Lp structure was unresolved—suggesting it was in a flexible or dynamic state—but the resolved portion (residues 50–56) was oriented in a manner highly similar to the ejected and inactive HipA_Ec P-loop (Fig. 5C). This suggests that the interaction between HipS_Lp and HipT_Lp facilitates P-loop ejection to neutralize the toxin. During our analyses, the structures of both HipT_Lp and HipS_Lp-HipT_Lp were reported by another group (21). These illustrate the orientation of the P-loop in its internalized and catalytically active state, and as in our HipS_Lp-HipT_Lp structure, this motif becomes unresolved and likely ejected upon HipS_Lp binding. In support of this, both HipS_Lp-HipT_Lp co-crystal structures show a high degree of concordance (HipS: RMSD 0.25 Å and TM-score 0.99 over 101 Cɑ atoms; HipT RMSD 0.25 Å and TM-score 0.96 over 261 Cɑ atoms) and are nearly identical in orientation, even with regard to the resolved portion of the P-loop. These findings are consistent with similar structural observations in the HipBST_Ec system (44), suggesting a conserved P-loop ejection model of HipBST neutralization.

FIG 5 HipTLp neutralization by HipSLp utilizes a P-loop ejection mechanism. (A) Alignment of the HipBLp structure with HipBEc (4YG1). (B) Comparison of the HipSLp-HipTLp co-crystal structure with HipAEc (3DNT) in which the HipAEc P-loop motif is in the internalized and active conformation. The HipAEc N-subdomain 1, corresponding to HipSLp, is indicated with a dashed box. The P-loops of both HipTLp (red) and HipAEc (black) are indicated. ATP in the HipAEc catalytic pocket is depicted with yellow spheres. Insets display the catalytic pocket of both kinases, along with two conserved residues involved in ATP coordination (D197, D309Q) and Mg2+ binding (D219, D332). (C) Alignment of the HipSLp-HipTLp structure and pHipAEc (3TPE), in which the P-loop motif is autophosphorylated and in the inactive, ejected conformation. Insets display two regions of variation between the HipSLp-HipTLp interaction interface and HipAEc: a shifted helix in HipSLp (inset on the left) and a loop in the N-terminal region of HipTLp (inset on the right). Key residues in each region are labeled. (D) Alignment of the HipSLp-HipTLp structure with the structure of HipBSTEc (7AB5). (E) Left: inset from (F) comparing the tryptophan residue in HipSEc (W65) with the methionine (M66) and isoleucine (I65) residues at the same position in HipSLp. Right: co-expression of HipTLp (pJB1806) with wild-type or mutant (I65A, M66A) HipSLp (pNT562) in L. pneumophila ∆hipBST cells. Expression was induced with IPTG (100 µM). Growth curves show the mean ± the standard deviation of three biological replicates. Data are representative of three independent experiments.

To better understand the evolution of this unique mechanism, we sought to identify sites of divergence between the HipS_Lp-HipT_Lp complex and HipA_Ec. Two major conformational changes are present in HipT_Lp loop 75–80 and HipS_Lp helix 65–66, relative to the equivalent regions in HipA_Ec (171–172 and 70–71), which shift these regions inward and impinge on the space where the P-loop would be internalized in its active state (Fig. 5C). We identified two residues in HipT_Lp (D77, Q78) that occupy this space and are not conserved in HipA_Ec (Fig. S1A). Substitution of either residue with alanine did not impair HipT_Lp toxicity (Fig. S5A); however, it also did not prevent neutralization by HipS_Lp (Fig. S5B). The HipT_Lp residue R154 is also not conserved in HipA_Ec or HipT_Ec and appears to contribute to the interaction between toxin and antitoxin through the formation of a hydrogen bond to the backbone of HipS_Lp(P62) and a salt bridge with HipS_Lp(E58) (Fig. 5C). However, mutation of this residue to alanine (R154A) was again not sufficient to impair the neutralization of HipT_Lp by HipS_Lp (Fig. S5A and B).

HipT_Lp neutralization by HipS_Lp does not rely on the tryptophan residue utilized by HipS_Ec

We next compared the L. pneumophila and E. coli HipBST systems directly, as a structure of HipBST_Ec was recently reported (44) (Fig. 5D). In the HipBST_Ec system, neutralization has been found to depend on a tryptophan residue in HipS_Ec (W65), which projects into the P-loop containing pocket of HipT_Ec (Fig. 5D and E). This residue appears critical for HipS_Ec function and is conserved across numerous HipS_Ec homologs, but is absent from HipS_Lp and instead, this site is occupied by much smaller methionine and isoleucine residues (I65, M66). We substituted both residues with alanine and observed a partial reduction in HipT_Lp neutralization by HipS_Lp(M66A), but no effect for HipS_Lp(I65A) (Fig. 5E; Fig. S5C). Consistent with these results, neither substitution impaired the physical interaction between HipS_Lp and HipT_Lp (Fig. S5D). In summary, the HipS_Lp-HipT_Lp P-loop ejection mechanism is broadly conserved with HipBST_Ec, yet is achieved through different motifs and interactions within the toxin-antitoxin interface.

HipT_Lp toxicity is not inhibited by autophosphorylation of its P-loop serine residue

The HipBST_Ec system was recently shown to contain a double serine motif in the HipT_Ec P-loop (S⁵⁷IS⁵⁹), which is proposed to allow for a dual autoregulatory dynamic in that system (44). This motif is absent from HipT_Lp, which contains only a single serine in this position (S54), similar to HipA_Ec (S150) (Fig. S1A). While many of the previously reported HipT homologs from the Gammaproteobacteria (24) contain either an SxS motif or some combination of double S/T residues, the corresponding motif in HipT_Lp (SVQ) is absent in these homologs but is either conserved or nearly identical (SIQ) across the HipT_Lp homologs in Legionella species. It has been shown previously that autophosphorylation or substitution of the single P-loop serine in HipA_Ec leads to loss of activity (40, 45). We therefore wondered what consequence modifying this residue would have on HipT_Lp toxicity. To test this, we constructed mutations that both ablate (S54A) and mimic (S54D) phosphorylation at this position. Neither mutation inhibited toxicity (Fig. S5E), which was surprising given that the corresponding mutation in HipT_Ec (S57A) renders the protein non-toxic (44), as does the S150A mutation in HipA_Ec (40), and autophosphorylation of HipA_Ec leaves it unable to bind ATP and retain catalytic activity (45). These findings are supported by recent structural work demonstrating that HipT_Lp can bind ATP despite being autophosphorylated (21) and highlight an unusual autoregulatory difference between these systems.

The cellular target of HipT_Lp is different from those of characterized HipT and HipA toxins

Despite their phylogenetic divergence, the HipBST_Lp system functions similarly to HipBST_Ec. Given the dearth of comparisons between TA homologs across distantly related bacteria, we chose to further explore the functional conservation between systems by testing whether they modified the same substrate. HipT_Ec was previously shown to phosphorylate tryptophan tRNA-ligase (TrpS) in order to arrest cellular growth (13), whereas the canonical target of HipA_Ec is glutamyl tRNA-synthetase (GltX) (46, 47). TrpS and lysine tRNA-ligase have also been reported as substrates for the HipBA systems in Caulobacter crescentus (15, 48). In E. coli, co-expression of TrpS or GltX with HipT_Ec or HipA_Ec, respectively, rescues the growth inhibitory effect of the toxins. In fact, E. coli TrpS was sufficient to rescue the toxicity of HipT homologs from Haemophilus influenzae and Tolumonas auensis as well (13). We therefore tested whether TrpS_Lp or GltX_Lp could rescue HipT_Lp-induced growth inhibition when co-expressed in L. pneumophila. Surprisingly, neither of these proteins could alleviate HipT_Lp toxicity (Fig. 6A; Fig. S6A), suggesting that HipT_Lp poisons the cell by targeting one or more previously undescribed substrates.

FIG 6 HipTLp has an unknown cellular target that is not conserved with characterized HipT or HipA toxins. (A) Co-expression of HipTLp (pJB1806) with GltXLp or TrpSLp (pNT562) in L. pneumophila cells. (B) Overview of phosphoproteomic experimental design. Left: cultures of L. pneumophila ∆hipBST carrying either pJB1806 or pJB1806::hipT Lp were grown to mid-log phase in the presence of 0.5% glucose, after which expression was induced by the addition of IPTG (100 µM). Prior to and 75 min postinduction (indicated by arrows on growth curves), cells were harvested for phosphoproteomic analysis. This experiment was performed independently twice, and growth curves show the mean ± the standard deviation from the two independent experiments. Right: cell lysates from cultures expressing HipTLp at both time points, and the vector control postinduction, were dimethyl labeled and multiplexed, enriched for phosphopeptides, and analyzed by mass spectrometry. The HipTLp-uninduced, vector-induced, and HipTLp-induced samples were labeled with the light, intermediate (med), and heavy channels, respectively. (C) Phosphopeptides detected in both HipTLp-uninduced and induced samples (left) and those enriched during HipTLp expression only (inset) are displayed. The dashed gray line is y = x for comparison between channels. In the inset, the suffix denotes the peptide number seen from a given ORF. The intermediate (vector-induced) channel is excluded from this plot for simplicity and because the results did not change the list of candidate substrates. The table contains candidate phosphoproteins that were observed across both replicates (excludes Lpg1968 and Lpg2511) along with their function, essentiality in broth, conservation in 58 Legionella species and E. coli (Ec), and detected phosphopeptide sequence. Candidates chosen for subsequent validation are highlighted in gray. The HipTLp phosphosite is underlined. (D) L. pneumophila cells co-expressing HipTLp (pJB1806) with phosphoproteomic candidates (pNT562) that were either essential or highly conserved in both Legionella and conserved in E. coli. (E) Co-expression of HipTLp with PheSTLp or phosphomimetic mutants of PheSLp(S50A) in L. pneumophila cells. Growth curves show the mean ± the standard deviation of three biological replicates. Expression was induced using IPTG (100 µM) for all constructs. Data are representative of three independent experiments.

GltX and TrpS were originally identified as the targets of HipA_Ec and HipT_Ec by screening E. coli genomic libraries for genes whose overexpression rescued toxin-induced growth inhibition (13, 46) or cold sensitivity (47). Given the observed toxicity of HipT_Lp in E. coli cells, we used the same approach to search for putative targets. We pooled and transformed the E. coli ASKA library into BL21 cells expressing HipT_Lp (pCDF1-b) and screened the resulting transformants. We observed robust growth inhibition upon HipT_Lp expression using an empty vector control (pCA24N), further demonstrating the toxicity of HipT_Lp in E. coli (Fig. S6B). When cells were transformed with the ASKA library pool, we recovered a small number of transformants across multiple screen replicates (Fig. S6B). Despite several clones exhibiting a stable growth rescue phenotype, sequencing revealed no clear enrichment of any genes or pathways, and subsequent attempts at validation in L. pneumophila were unsuccessful.

As rescue screening of the ASKA library did not produce any strong candidate HipT_Lp targets, we next pursued an orthogonal approach. Phosphoproteomic analysis was used previously to identify TrpS as a substrate of C. crescentus HipA2 (48) and to demonstrate the enrichment of GltX phosphorylation in the HipA_Ec phosphoproteome (49). We therefore performed a phosphoproteomic screen in L. pneumophila to detect substrates that were only modified under conditions of HipT_Lp overexpression, relative to both uninduced HipT_Lp and empty vector control (Fig. 6B). From this, we identified a small number of proteins that were enriched for phosphopeptides during HipT_Lp-induced growth inhibition (Fig. 6C; Table S8). We also observed that HipT_Lp is phosphorylated on S54, thereby demonstrating the occurrence of this modification in vivo. Of the eight candidate substrates identified, three are known to be essential for growth (50, 51), six are conserved across all Legionella species, and five have orthologs in E. coli (Fig. 6C). While HipA_Ec has been shown to phosphorylate a large pool of proteins, only co-expression with GltX is sufficient to rescue its growth inhibitory phenotype (49). In order to validate our candidate substrates, we cloned a subset of hits that were either essential or highly conserved and tested them by co-expression with HipT_Lp. Under these conditions, no putative substrate was able to rescue growth inhibition in L. pneumophila (Fig. 6D; Fig. S6C).

One of the hits we detected was the alpha subunit (PheS) of phenylalanine tRNA-ligase. Given that both GltX and TrpS are tRNA-ligases, this class of target would be consistent with other Hip toxins. Interestingly, phenylalanine tRNA-ligase is composed of two separate subunits, whereas GltX and TrpS are both single proteins. To test whether the complete phenylalanine tRNA-ligase (PheST_Lp) was required for growth rescue, we co-expressed it with HipT_Lp in L. pneumophila. However, we did not observe any rescue of growth inhibition (Fig. 6E; Fig. S6D). From our phosphoproteomic data, we also observed that PheS was phosphorylated on the S50 residue during HipT_Lp overexpression. We next tested whether mutation of this site to ablate phosphorylation (S50A) would prevent growth inhibition; however, neither PheS(S50A) nor PheST(S50A) was able to rescue L. pneumophila growth when co-expressed with HipT_Lp (Fig. 6E; Fig. S6D). While these approaches were unable to identify a definitive toxic substrate of HipT_Lp, they nevertheless suggest that its cellular target is likely not conserved with HipT_Ec or any characterized HipA homologs.

DISCUSSION

The ubiquity of toxin-antitoxin systems in bacterial genomes and their capacity for selfish maintenance have provided strong evidence for their role as parasitic elements. Given the plastic nature of the accessory genome and the patchy distribution of TA systems, the occurrence of highly conserved elements would suggest selection beyond mere addiction. Here, we characterize a HipBST system in L. pneumophila, which is highly conserved across Legionella species, despite the genomic variation within this genus. Indeed, more than half of the Legionella species we searched have either a complete HipBST module (28/58) or individual HipBST genes (8/58). This level of conservation is notable, in that it is higher than both the average accessory gene (26/58) and L. pneumophila effector (16/58).

In characterizing HipBST_Lp, we detected a large number of closely related homologs across diverse bacterial taxa. This greatly expands the sequence space of HipBST systems and reveals considerable diversity within this newly described TA family. By comparing the individual protein phylogenies (Fig. S1C), we observed an interesting pattern whereby HipB sequences from HipBA_So-like systems cluster with HipB sequences from HipBST systems. The structural similarities between HipBA_So and HipBST_Ec have been noted previously (44) and given the difference in HipB orientation between these systems and HipBA_Ec, this could suggest an evolutionary relationship between HipBA_So and the newly discovered HipBST systems.

We were surprised to discover that homologs of HipBST_Lp are almost entirely restricted to the Legionella genus within the Gammaproteobacteria but are widely distributed in other taxonomic groups. This could suggest a shared environmental niche, common functional role, or frequent DNA exchange between these taxa and Legionellales. Previous work has only compared HipBST systems that are closely related to HipBST_Ec (13), and as the sequence space of HipBST_Lp homologs appears to be both more diverse and taxonomically distributed (Fig. 1C and D), this offers the potential to reveal even greater breadth of TA biology within this family. One pattern that emerges from our initial data is an apparent exclusion between different HipBST clades (Fig. 1D). While it is tempting to speculate that some difference in biological function (e.g., persistence or chromosomal stabilization) or cellular target could be responsible for the taxonomic distribution of these two system types, it is important to note that further work is required to confidently establish this pattern. Importantly, our analyses focused on closely related rather than remote homologs of each TA system, and thus do not incorporate distant homologs that would be otherwise detected using more sensitive search methods or highly divergent systems comprising components from multiple system types. This was done to examine the taxonomic distribution of systems with possible shared ancestry and evolutionary history, and because our findings suggest that remote homology (such as between HipBST_Ec and HipBST_Lp) may mask important biological differences. However, we do note this methodological limitation of our analyses, and because of this, we cannot make assertions about the taxonomic distribution patterns across more distantly related HipBST homologs.

During our genomic search, we discovered a putative duplicate HipBST system downstream of HipBST_Lp. This locus (lpg2377-80) encodes five predicted ORFs and is only found in the L. pneumophila genome (Fig. S7A). Its unusual architecture is achieved through the splitting of both hipS and hipT, resulting in the separation of the P-loop and kinase core in HipT. Interestingly, all catalytically critical motifs remain conserved in split-HipT (Fig. S7A). To determine whether this system was functional, we expressed both split-HipT_Lp and split-HipS_Lp in L. pneumophila cells. Despite the preservation of the P-loop motif, the expression of either split protein pair, alone or in combination, did not exhibit a growth inhibitory effect (Fig. S7B). We also observed a similar organization in the HipBST system encoded by Legionella gormanii, where hipT is again split, but in this case, the upstream ORF is truncated due to a frameshift and the catalytic P-loop is no longer intact (Fig. S7C). Thus, this system appears to be in a state of decay or functional divergence. While it remains to be seen how these split-protein architectures affect system functionality, their growing diversity attests to the substantial modularity and evolvability of the Hip TA systems.

A striking feature of the HipBST_Lp system is the previous claim that HipT_Lp is an L. pneumophila effector (20, 21). This would represent an exceedingly rare example of toxin-antitoxin system/eukaryotic effector bifunctionality, as there is limited evidence of TA modules being repurposed as interdomain translocated substrates (7). However, we found no evidence of HipT_Lp translocation beyond levels of the negative control FabI. Instead, the broad taxonomic distribution of HipBST_Lp homologs likely extends to many species that do not deliver effectors into host cells, further arguing against this functionality. HipT_Lp was first hypothesized to be an effector due to the presence of a putative translocation motif (52); however, re-examination determined it was unlikely to encode a true secretion signal (53). Our inability to observe HipT_Lp translocation above control levels, therefore, argues for caution in ascribing functionality to HipT_Lp within the eukaryotic host. Given that cytosolic proteins (such as FabI) can be translocated at low frequencies (39, 54) or their localization potentially influenced by bacterial cell lysis events, basal level of secretion should be confirmed with orthologous methodologies, inactive mutants unlikely to cause cell lysis, careful quantification, and several controls. Incontrovertible evidence for HipT_Lp effector activity may come from the phenotypic observation of an impact on the host cell during infection. In the absence of this, exceeding a threshold ratio of quantified fluorescence is the necessary standard for effector validation (39, 55).

In this work, we demonstrate that HipBST_Lp is a TA system—the core functionality of which is conserved with HipBST_Ec. A recent report on HipBST_Lp (21) included the observation that HipT_Lp does not inhibit growth in E. coli. We routinely observed that overexpression of HipT_Lp inhibits E. coli growth, using multiple strains, vectors, and conditions (Fig. 3A; Fig. S3F, G, and S6B). Our detection of a physical interaction between HipB_Lp and HipT_Lp, by both yeast two-hybrid and EMSA assays, is also contradictory to the previous report (21), which found no such interaction. The absence of an interaction between HipB_Lp and HipT_Lp would suggest that only the neutralized HipS_Lp-HipT_Lp complex would participate in the autoregulatory activity of HipB_Lp. Instead, our observations are consistent with the formation of HipB_Lp-DNA, HipT_Lp-HipB_Lp-DNA, and HipS_Lp-HipT_Lp-HipB_Lp-DNA complexes (Fig. 4E; Fig. S4F). This raises the possibility of multi-layered regulation in the HipBST_Lp system, whereby dimeric, binary, and ternary protein interactions could influence system expression (Fig. 7). For example, the tripartite protein interaction dynamics may influence HipB_Lp binding affinity for promoter DNA or produce differential stability of protein-DNA complexes as a feedback mechanism for system repression/derepression. An emphasis of future work should, therefore, be to investigate the changes in HipB_Lp-DNA interactions and affinity in the presence of HipS_Lp and HipT_Lp, in addition to examining what impact, if any, the phosphorylation state of HipT_Lp has on these interactions (38).

FIG 7 Autoregulatory capacity of the HipBSTLp toxin-antitoxin system. HipTLp is neutralized by physical interaction with HipSLp and stable interactions are additionally observed between HipBLp-HipTLp and HipBLp-HipSLp-HipTLp. HipBLp can bind promoter DNA at three conserved operator sites upstream of hipBSTLp to regulate system expression and these sites contain unique sequence motifs relative to HipBA systems. Protein-DNA complexes formed by HipBLp-DNA, HipTLp-HipBLp-DNA, and HipSLp-HipTLp-HipBLp-DNA suggest the possibility for multiple layers of transcriptional regulation in the system that could be influenced by relative protein stoichiometry and complex stability.

While investigating the regulatory dynamics of HipBST_Lp, we discovered and characterized three conserved operator sites upstream of hipBST homologs across Legionella species (Fig. 7). The presence of three sites is concordant with the gel shift patterns we observed when incubating HipB_Lp with promoter DNA (Fig. 4B) and raises the intriguing possibility that HipB_Lp-DNA complexes with intermediate occupancy can be formed, in contrast to the cooperativity observed with HipB from HipBA_Ec (42). These are the first operators identified for a HipBST TA system and are notable in that they differ in both their inverted repeat motifs and overall structure relative to previously identified HipBA operators (38, 43). The Legionella HipBST inverted repeat sequences are composed of six nucleotides rather than the five in HipBA systems, possess larger inter-operator spacer distances, and the sequence between inverted repeats is largely conserved across operators rather than being variable. Importantly, O2 and O3 contain single nucleotide substitutions that abolish inverted repeat symmetry, yet are still bound by HipB_Lp in a concentration-dependent manner comparable with O1. This variation between operator sites within the same genome is, therefore, a unique aspect of HipBST autoregulation, with one possible explanation being that HipB binds each site with differing affinity. We did not observe HipB_Lp affinity differences in our EMSA assays (Fig. S4D); however, future work should address this and the potential for intermediate occupancy protein-DNA complexes with more sensitive methodologies, in addition to identifying which HipB_Lp residues and operator sites influence DNA binding and specificity. Interestingly, the HipBST operator motifs we identified are absent from the promoter region upstream of hipBST_Ec, which further supports the functional divergence between these two systems. We were not able to detect these motifs at other sites throughout the L. pneumophila genome, except for upstream of the pentapartite system, which is consistent with a putative duplication event for this locus.

We were surprised to find that the HipS_Lp-HipT_Lp complex adopts a conformation that is almost unchanged from that of HipA_Ec, consistent with recently reported structures for both HipBST_Ec (44) and HipBST_Lp (21). Neutralization in both HipBST systems is achieved by exploiting an intrinsic property of HipA biology—namely its P-loop ejection mechanism of autoregulation. The HipBST_Ec and HipBST_Lp systems achieve this functionality through shared structural motifs, while also displaying several key differences. In particular, while the bulky W65 residue in HipS_Ec is both critical for toxin neutralization and conserved in many other HipS_Ec homologs (44), in HipS_Lp this position is occupied by the much smaller I65 and M66 residues. These do not impinge on the catalytic pocket of HipT_Lp to the same extent as W65 and are not necessary for HipS_Lp-HipT_Lp physical interaction or neutralization. The absence of an analogous residue to W65 in HipS_Lp, therefore indicates alternate biochemical means of achieving P-loop ejection may be present across HipBST system homologs.

A double serine motif (S⁵⁷IS⁵⁹) was recently characterized in HipT_Ec (44), with the differential autophosphorylation of either serine reported to affect toxin neutralization and activity. This motif is absent from HipT_Lp, which instead contains a single P-loop serine (S54) similar to the S150 residue in HipA_Ec. We observed that the phosphorylation state, or substitution of this residue with alanine, does not affect protein toxicity. This was unexpected, given that both pHipA_Ec and HipA_Ec(S150A) cannot bind ATP and are not toxic (45). In support of this unusual behavior, a structure of pHipT_Lp bound to an ATP analog was recently reported (21), further demonstrating the retention of HipT_Lp activity despite its phosphorylation state. Why HipT_Lp autoregulation differs in such a critical aspect from HipA_Ec is an outstanding question, as this would seemingly prohibit cell detoxification via trans autophosphorylation. Phosphomimetic mutation of either serine residue in HipT_Ec (S57D, S59D) does not inhibit toxicity (44); however, mutation of either residue to alanine appears to abrogate activity—though conflicting results have been reported (21). Regardless, our findings demonstrate that the autophosphorylation state of HipT_Lp is not required for its activity, indicating a shift in the autoregulatory functionality of the HipT_Lp kinase.

The most critical difference we observed between HipBST_Ec and HipBST_Lp is the failure of either GltX_Lp or TrpS_Lp to rescue HipT_Lp toxicity. This was further supported by the absence of both GltX_Lp and TrpS_Lp in our genomic rescue and phosphoproteomic screens. We did, however, observe a small set of candidate proteins that were only phosphorylated during HipT_Lp overexpression in L. pneumophila cells. In particular, we detected the phosphorylation of PheS, which is also a tRNA-ligase like GltX and TrpS, and thus a strong candidate for the target of HipT_Lp. While none of the putative targets were sufficient to rescue L. pneumophila growth when co-expressed with HipT_Lp, this does not eliminate them from consideration. Instead, this may be a consequence of some aspect of L. pneumophila or HipT_Lp biology that is not conducive to kinase target saturation for growth rescue, or possibly the requirement of multiple co-expressed substrates to overcome toxin activity. For example, while we observed no rescue upon co-expressing one promising phosphorylated candidate–the β′-subunit of RNA polymerase (RpoC)–with HipT_Lp, the complete enzyme is composed of three additional subunits (RpoA, RpoB, and RpoZ). It is possible that co-expression of each of these four subunits in stoichiometric amounts may be required to rescue growth inhibition by HipT_Lp. An aim of future work will, therefore, be to further test the candidate targets identified using alternative means of validation, in addition to co-expression tests that include functional interactors, such as the reconstitution of the entire RNA polymerase enzyme. In light of our results, it should also be considered whether HipT_Lp in fact phosphorylates a protein at all to poison the cell, as the true target may be a nucleic acid or other molecule. Conversely, the presence of both DotA and multiple effectors in our phosphoproteomic data may reflect either a previously undescribed mode of toxicity, such as disrupting the translocation apparatus (which is known to have effects on axenic growth) (56) or reveal a regulatory functionality of HipT_Lp with regard to effector secretion. As such, future work should focus on first identifying the overarching biological process (such as macromolecular biosynthesis), if any, that HipT_Lp perturbs. One approach might be to utilize metabolic labeling assays that monitor the incorporation of radiolabeled methionine, uridine, and thymidine, in order to determine whether HipT_Lp expression has a differential effect on rates of translation, transcription, or DNA synthesis, respectively—as has been found with other toxins (57, 58). Were this to be inconclusive, alternative strategies such as suppression mutant or gene knockout library screening could also be undertaken, in addition to testing the effect of HipT_Lp on Dot/Icm assembly or function. Regardless, we report that HipT_Lp likely has a different cellular substrate from characterized HipT and HipA toxins, further demonstrating its functional divergence from HipT_Ec.

Given the conservation of HipBST homologs across Legionella species, we wondered whether this system showed any pattern of association with chromosomal subregions or signatures of acquisition. The genome of L. pneumophila is organized into discrete clusters of essential and non-essential genes—the latter of which are enriched for effectors and mostly dispensable for broth growth and host infection (50, 51). The HipBST_Lp locus is found within one such cluster (Fig. S8A), and this led us to hypothesize that other HipBST_Lp systems might also be found within accessory genomic regions of non-essential genes. To test this, we predicted regions of non-essentiality across the 28 Legionella species’ genomes with HipBST systems. We found that HipBST is frequently associated with non-essential gene clusters (22/28) and that neither the local genetic neighborhood nor broader cluster composition is well conserved across HipBST_Lp-containing clusters (Fig. S8B). We further extended this search to incorporate remote homologs in the gene neighborhood of HipBST_Lp (Fig. S8C); however, despite the increased search sensitivity, we observed relatively poor conservation of most proximal elements (≤30% of clusters). Interestingly, one gene neighbor (WP_015444114.1) did co-occur with the majority of HipBST systems (75%) and was predicted to encode either a recombinase or integrase, which could indicate a possible origin for HipBST acquisition with Legionella species. The high conservation of HipBST modules, contrasted with the low overall cluster conservation, raises the possibility that these modules may serve to stabilize their unique genomic neighborhood in each species. Indeed, tripartite TA systems are more commonly found on mobile genetic elements and have been hypothesized to function in DNA stabilization (12). This would be advantageous for Legionella species, which encode vast and diverse effector repertoires—including some with fitness costs in specific hosts (51, 59). Given the broad host range of Legionella, maintaining effector diversity likely provides a selective benefit, and similar stabilizing effects have been demonstrated for other chromosomal TA systems (60 – 63).

Overall, our work provides a detailed characterization of the HipBST TA system in L. pneumophila. Given its small and unstudied TA system set, extreme genomic characteristics, and challenging host-associated lifestyle, investigating the TA systems in this species can be a source of new insights into TA biology and function. We find a strong signal of conservation for HipBST within the Legionella genus, which is unusual for a TA system and, in particular, one that is almost completely absent from all other taxa in the Gammaproteobacteria. Despite its frequent occurrence in accessory genomic clusters, this level of conservation suggests that some functionality beyond promiscuous acquisition and addiction is being selected for. We also show that considerable molecular divergence can be found between related or even the same TA system family, which underscores the potential for discovering new biology in the expansive and unexplored sequence space of bacterial genomes. These findings, therefore, justify caution when generalizing TA system functionality across different bacterial hosts and instead suggest that each system be considered within the context of its specific genomic niche.

MATERIALS AND METHODS

Strains and plasmids

All strains used in this study are listed in Table S1. All plasmids used in this study are listed in Table S2. All oligonucleotides used in this study are listed in Table S3. L. pneumophila strains used were derived from Lp01^JK (64), except for Lp02, which was used for the TEM-1 translocation experiments. E. coli XL1-Blue and TOP10 cells (Invitrogen) were used for cloning and plasmid maintenance, and BL21-GOLD (DE3) (Stratagene) cells were used for protein expression. TOP10 and BL21-GOLD (DE3) were additionally used for in vivo toxicity assays. Saccharomyces cerevisiae Y8800 (MATa leu2-3,112 trp1-901 his3-200 ura3-52 gal4∆ gal80∆ GAL2-ADE2 LYS2::GAL1-HIS3 MET2::GAL7-lacZ cyh2R) (65) was used for yeast two-hybrid assays and BY4742 (MATα his3Δ1 leu2Δ0 met15Δ0 ura3Δ0) (66) was used for in vivo toxicity assays. Bacterial expression plasmids were constructed using PCR products amplified from Lp01^JK genomic DNA with restriction cloning, or ligation-independent cloning for the pMCSG68-SBP-TEV protein purification constructs. PCR products were cloned into pDONR221-ccdB using BP clonase (Invitrogen). The resulting pDONR221 constructs were then used to clone the hipBST_Lp genes into the yeast expression vectors pAG425GAL-ccdB (for yeast toxicity tests), pAG416GPD-ccdB, pDEST-DB-ccdB, and pDEST-AD-ccdB (for the Y2H and Y3H assays) using LR clonase (Invitrogen). pAG425GAL-ccdB (Addgene plasmid # 14153; RRID:Addgene_14153) and pAG416GPD-ccdB (Addgene plasmid # 14148; RRID:Addgene_14148) (67) were gifts from Susan Lindquist. Vectors with the DNA-binding (DB) and transcription-activating (AD) domain of Gal4 (pDEST-DB, pDEST-AD) (68) were a kind gift from N. Yachie and F. Roth (University of Toronto, Canada). Genetic substitutions were introduced via site-directed mutagenesis using the Q5 site-directed mutagenesis kit (NEB). All generated constructs were confirmed by Sanger sequencing. Plasmids were introduced into E. coli via heat-shock transformation, into S. cerevisiae via the PEG/LiAc method (69), and into L. pneumophila via electroporation. The endogenous hipBST locus in L. pneumophila was deleted using a previously described genome editing method (70) with minor modifications (71). The resulting deletion mutant was screened by PCR using the oligos JL-P102 and JL-P103, and validated with Sanger sequencing using the oligo JL-P101.

Media and culture conditions

Bacterial experiments and routine strain maintenance were performed at 37°C. L. pneumophila strains were grown in N-(2-acetamido)-2-aminoethanesulfonic acid (ACES)-buffered yeast extract and on charcoal AYE agar plates supplemented with 0.4 g/L L-cysteine and 0.25 g/L ferric pyrophosphate. For liquid growth, cultures were inoculated from patches grown for 2 days. When required for selection or plasmid maintenance, media were supplemented with chloramphenicol (5 µg/mL), kanamycin (20 µg/mL), or thymidine (100 µg/mL). Ectopic gene expression was induced by isopropylthio-β-galactoside (IPTG; 100 µM) and repressed with glucose (1% vol/vol), unless otherwise indicated. E. coli strains were grown on lysogeny broth (LB, Miller) liquid media and agar. When required, media were supplemented with ampicillin (100 µg/mL), chloramphenicol (34 µg/mL), or kanamycin (40 µg/mL) for selection or plasmid maintenance. Ectopic gene expression was induced by arabinose (0.2% wt/vol) or IPTG (100 µM) and repressed with 1% glucose unless otherwise indicated. Yeast experiments and routine strain maintenance were performed at 30°C. S. cerevisiae strains were grown on yeast peptone adenine dextrose medium (2% bacto peptone wt/vol, 1% yeast extract wt/vol, 2% glucose vol/vol, 180 mg/L adenine sulfate), or synthetic defined (SD) medium composed of yeast nitrogen base with ammonium sulfate, supplemented with 2% glucose and all amino acids, lacking specific amino acids where necessary for selection or plasmid maintenance. When required, media lacking glucose were supplemented with galactose (2% vol/vol) to induce gene expression.

TEM-1 β-lactamase translocation assays

Protein translocation was tested as described previously (39). Briefly, Lp02 strains carrying pXDC61 constructs containing the TEM-1 β-lactamase fused to a gene of interest were grown overnight to mid-log phase (OD₆₀₀ = 1.5–2), and fusion protein expression was induced with IPTG (500 µM) for 3 h. The cultures were then used to infect monolayers of differentiated U937 cells (RPMI 1640 medium, Gibco) in triplicate at various MOIs (20, 50, and 125) for 1 h at 37°C with 5% CO₂. LiveBLAzer dye (ThermoFisher K1095) was subsequently added to the cells and incubated for another 2 h, followed by blue/green fluorescence detection using a TECAN Infinite M PLEX plate reader (blue fluorescence: 409 nm excitation, 460 nm emission, gain 140, integration time 20 μs; green fluorescence: 409 nm excitation, 530 nm emission, gain 140, integration time 20 μs). A blue/green fluorescence ratio >1 was used as the threshold to determine translocation (55). Each experiment was performed in triplicate and repeated three times. Immunoblotting to confirm TEM-1 fusion protein expression was performed with an anti-β lactamase antibody (Abcam 12251).

In vivo bacterial toxicity experiments

Growth assays were performed as follows: E. coli and L. pneumophila strains containing plasmids expressing the genes of interest were grown overnight in the presence of 1% glucose. Cultures were washed to remove glucose and adjusted to OD₆₀₀ = 0.1 in fresh media supplemented with either 1% glucose or arabinose/IPTG. Cultures were plated in triplicate in a flat bottom 96-well plate (100 µL volume), sealed with a Breathe-Easy sealing membrane (Diversified Biotech BEM-1), and optical density (600 nm) was monitored every 15 min for 24 h using an S&P growth curve robot. Each experiment was repeated a minimum of three times unless otherwise indicated.

Protein expression and purification for in vitro experiments

Protein expression and purification were performed as described previously (72). Briefly, BL21-GOLD (DE3) cells carrying the desired His₆-SBP-tagged constructs were grown at 37°C to OD₆₀₀ = 0.5, at which time protein expression was induced with IPTG (1 mM) for 5 h at 37°C. Cells were harvested by centrifugation (12,227 × g for 10 min at 4°C), resuspended in lysis buffer {300 mM NaCl, 5% glycerol, 5 mM imidazole, 50 mM HEPES [4-(2-hydroxyethyl)−1-piperazineethanesulfonic acid] pH 7.5}, lysed by sonication on ice (30% amplitude, 10 s on, 10 s off for 5 min) in the presence of 1 mM phenylmethylsulfonyl fluoride (PMSF), and the soluble fraction was obtained by high-speed centrifugation (34,957 × g for 20 min at 4°C). The tagged proteins were purified by immobilized metal affinity chromatography using nickel-nitrilotriacetic acid beads. Columns were washed with wash buffer (300 mM NaCl, 30 mM imidazole, 15 mM HEPES pH 7.5, 5% glycerol), and eluted in 300 mM NaCl, 300 mM imidazole, and 15 mM HEPES pH 7.5. Eluted protein was then dialyzed in 300 mM NaCl, 15 mM HEPES pH 7.5, 0.5 mM dithiothreitol (DTT), and concentrated using Vivaspin 5 kDa (HipB_Lp, HipS_Lp) and 10 kDa (HipT_Lp) cutoff columns (GE Healthcare). Protein concentration was quantified using the Bradford assay (BioShop). For additional purification (in vitro assays involving co-incubation), His₆-SBP-tagged HipB_Lp, HipS_Lp, and HipT_Lp were injected into a Superdex S75 10/300 GL size exclusion column (GE Healthcare) equilibrated in dialysis buffer, and the purified fractions were collected.

Crystallography and structure determination

The HipS_Lp-HipT_Lp complex was crystallized as selenomethionine (Se-Met)-derivatized proteins and as native proteins (for the higher resolution final structure). Cloned in pETDuet vector, HipS_Lp and His₆-HipT_Lp proteins were expressed in E. coli BL21(DE3)-Magic cells. For the (Se-Met)-derivatized protein complex, cells were grown in M9 minimal media (Shanghai Medicilon) with 1 mM IPTG induction at 20°C when the OD₆₀₀ reached 1.2. The native protein complex was expressed in ZYP-5052 auto-inducing complex medium (73) by incubating for a few hours at 37°C followed by overnight growth at 20°C. Overnight cell culture was then collected by centrifugation at 6,000 × g for 25 min at 4°C. Cells were resuspended in a binding buffer [100  mM HEPES (pH 7.5), 500  mM NaCl, 5  mM imidazole, and 5% glycerol (vol/vol)]. The purification was performed as described above, which resulted in the formation of the HipS_Lp-HipT_Lp complex as judged by SDS-PAGE. HipB_Lp was expressed as a native protein and purified as described above. Crystals were grown at room temperature (RT) using the vapor diffusion sitting drop method. For the HipS_Lp-HipT_Lp (Se-Met) complex, 17 mg/mL protein was mixed with reservoir solution [0.1 M potassium chloride, 10 mM Tris pH 7, 20% (vol/vol) PEG 4K]. For the HipS_Lp-HipT_Lp (native) complex, 8 mg/mL protein was mixed with reservoir solution [0.1 M HEPES pH 7.5 and 30% (vol/vol) PEG 1K]. For the HipB_Lp crystal, 20 mg/mL protein was mixed with reservoir solution [0.2 M ammonium sulfate, 0.1 M sodium acetate pH 4.6, and 25% (vol/vol) PEG 4K]. Crystals were cryoprotected with paratone oil. For the HipS_Lp-HipT_Lp complex crystal, diffraction data at 100 K were collected at beamline 19-ID of the Structural Biology Center at the Advanced Photon Source, Argonne National Laboratory. For the HipB_Lp crystal, diffraction data at 100 K were collected at a home source Rigaku HF-007 rotating anode with Rigaku R-AXIS IV image plate detector. All diffraction data were processed using HKL3000 (74). The HipS_Lp-HipT_Lp (Se-Met) complex was solved using Phenix.Autosol (75), which identified seven Se-Met sites out of seven total methionine residues in the asymmetric unit; the native complex was solved by molecular replacement using the Se-Met complex structure. The structure of HipB_Lp was solved by molecular replacement using the CCP4 online Balbes server and the structure of HipB_Ec (PDB 2WIU). All model building and refinement were performed using Phenix.refine and Coot (76).

In vitro kinase assays

Purified recombinant His₆-SBP-tagged HipB_Lp, HipS_Lp, and HipT_Lp (WT, D197Q, D219Q) were tested for ATP hydrolytic activity using the ADP-Glo kinase assay (Promega). Briefly, 1 µg of each protein was assayed in 25 µL reactions containing 40 mM Tris-HCl pH 7.5, 20 mM MgCl₂, 0.1 mg/mL bovine serum albumin, 1 mM DTT, and 25 µM ATP. When included, the kinase inhibitor FSBA was supplemented to a final concentration of 20 mM. Reactions were incubated at 37°C for 30 min and luminescence was measured using a TECAN Infinite M PLEX plate reader (500 ms integration time). Each experiment was replicated independently a minimum of two times.

Electrophoretic mobility shift assays

Varying concentrations of purified recombinant His₆-SBP tag-cleaved HipB_Lp, HipS_Lp, and HipT_Lp were incubated with double-stranded DNA substrates (gel purified PCR products or annealed oligos) in binding buffer (400 mM KCl, 150 mM HEPES pH 7.5, 10 mM EDTA, 50% glycerol, 5 mM DTT) for 30 min at RT. The protein-DNA mixtures were then run at 70 V on a 6% native polyacrylamide gel with 1× TAE (Tris-acetate-EDTA) running buffer, stained with SYBR Green (Invitrogen), and imaged. DNA substrates consisted of the promoter region of the hipBST_Lp operon (upstream 200 bp amplified with oligos JL-P336 and JL-P341) or two other predicted toxin-antitoxin operons used as controls (lpg1604-05 and lpg1934-35 upstream 200 bp amplified with oligos JL-P516/JL-P517 and JL-P518/JL-P519, respectively). The 50-bp fragments containing DNA sequence from upstream of the hipBST_Lp operon were generated by annealing complementary oligos (Table S3) in a 1:1 ratio in annealing buffer (100 mM KAc, 30 mM HEPES pH 7.5), heating at 94°C for 2 min, and cooling slowly to RT. Each experiment was repeated a minimum of three times.

Yeast two-hybrid assays

Yeast two-hybrid experiments were performed as described previously (77). Briefly, proteins of interest were fused to either the GAL4 transcriptional AD or DB domain, using the pDEST-AD-ccdB and pDEST-DB-ccdB constitutively active Gateway destination plasmids. Three independent clones of Y8800 containing pDEST-AD and pDEST-DB encoding gene fusions of interest were grown overnight at 30°C in liquid SD-Leu/Trp media supplemented with 2% glucose. These cultures were then stamped on plates containing (control) or lacking histidine (physical interaction selection). When required, a third protein was expressed constitutively from the pAG416GPD plasmid. Y8800 strains containing pDEST-AD, pDEST-DB, and pAG416GPD together were grown with SD-Leu/Trp/Ura media supplemented with 2% glucose. Plates were imaged after 2 days of growth at 30°C. Each experiment was repeated a minimum of two times.

Yeast spotting assays

Yeast spotting experiments were performed as described previously (77). Briefly, three independent clones of BY4742 containing the gene of interest in the pAG425GAL expression vector were grown in triplicate overnight at 30°C in liquid SD-Leu media supplemented with 2% glucose and adjusted to OD₆₀₀ = 1 the following day. Fivefold serial dilutions of each culture were then prepared in a 96-well plate at a volume of 120 µL and stamped onto solid SD-Leu media supplemented with either glucose (2%; suppression) or galactose (2%; expression) using a VP 407AH pin tool (VP Scientific). Plates were imaged after 2 days of growth at 30°C. Each experiment was repeated a minimum of three times.

ASKA genomic library rescue screen

The E. coli ASKA library (without GFP tags) (78) was pooled to a final concentration of 100 ng/µL, and 150 ng of the pool was electroporated into BL21-GOLD (DE3) cells containing HipT_Lp cloned into the pCDF1-b expression vector. Cells were recovered for 1 h at 37°C and transformations were plated on solid media containing either IPTG (100 µM) for gene expression or 1% glucose for repression and quantifying transformation efficiency. As a control, the empty vector pCA24N was transformed in an equivalent manner to the pooled library. Library transformations without gene expression yielded 10⁴–10⁵ cells and library transformations were performed a minimum of three times for each experiment. Individual screening experiments were repeated three times. Transformants that grew under selective conditions were re-struck on selective media to ensure a stable rescue phenotype. Plasmids from the resulting strains were then prepared and Sanger sequenced using the oligo JL-P206.

HipT_Lp in vivo expression for phosphoproteomic analysis

Cultures of L. pneumophila ∆hipBST carrying either pJB1806 or pJB1806::hipT _Lp were grown in the presence of 0.5% glucose until mid-log phase (OD₆₀₀ = 1.5–2). Expression was then induced with IPTG (100 µM) and 75 mL of cells was harvested prior to and 75 min postinduction. Lysates were prepared by resuspending cell pellets in 2 mL NP-40 buffer [50 mM Tris-HCl pH 7.8, 150 mM NaCl, 1% NP-40, 1 mM PMSF, 1× PhosSTOP (Roche)], sonication on ice (5 s on/10 s off for 2 min at 40% amplitude), and high-speed centrifugation (13,000 × g for 30 min at 4°C). Each experiment was repeated two times.

In-solution protein digestion

Cell lysates from the ∆hipBST cultures (vector-induced, HipT_Lp-uninduced, and HipT_Lp-induced) were quantified for total protein (Pierce BCA Protein Assay Kit, Thermo Scientific), and 2.5 mg of total protein from each sample was used. Samples were reduced with 5 mM DTT (37°C, 1 h) and alkylated with 10 mM iodoacetamide (25°C, 45 min). Samples were digested for 2 h at 25°C with constant agitation with a Trypsin/LysC mix (V5071, Promega) and overnight at 25°C with proteomics-grade porcine trypsin (T6567, Sigma-Aldrich). For both digestions, 6.25 µg of trypsin was used. Samples were desalted with 50-mg capacity Sep-Pak C18 cartridges (Waters Corporation). In brief, samples were acidified with trifluoroacetic acid (TFA) to a final concentration of 1% and centrifuged at 21,000 × g to remove precipitate. Columns were conditioned by passing one column volume (CV) of 100% acetonitrile (ACN), one CV of 50% ACN/0.1% formic acid (FA), and four CV of 0.1% TFA. Samples were loaded and desalted with 1% TFA and 1% FA, and washed peptides were eluted with 50% ACN/0.1% FA and concentrated by vacuum centrifugation.

Dimethyl labelling

Samples were labeled as described previously (79) with some modifications. Peptides were resuspended in 50 mM HEPES pH 8.0, and the HipT_Lp-uninduced, vector-induced, and HipT_Lp-induced samples were designated to be labeled with the light, intermediate, and heavy channels, respectively. Peptides were mixed with freshly prepared CH₂O (light, +28.0313 Da), CD₂O (intermediate, +32.0564 Da), and ¹³CD₂O (heavy, +36.0757 Da), and all solutions were made to 4% (vol/vol). NaBH₃CN (0.6 M) was added to the light and intermediate channels, and NaBD₃CN (0.6 M) was added to the heavy channel. Samples were labeled at RT for 2 h, quenched with FA (5% final), and desalted with Sep-Pak C18 Cartridges. Labeled peptides were quantified (Pierce Colorimetric Peptide Assay Kit, Thermo Scientific) and mixed 1:1:1 based on peptide amount.

Phosphopeptide enrichment

Labeled and mixed peptides were phosphoenriched with TiO₂ magnetic beads according to the manufacturer’s protocols (MagReSyn). In brief, beads (100 µL) were equilibrated by three washes with loading buffer (0.1 M lactic acid in 80% ACN/5% TFA) before beads were incubated with 480 µg of peptides for 20 min at RT with end-over-end mixing. Samples were washed with 100 µL loading buffer, 80% ACN/1% TFA, and 10% ACN/0.2% TFA. Phosphopeptides were eluted twice in 1% NH₄OH and acidified with FA to a final concentration of 2.5%. Samples were vacuum centrifuged and desalted with C18 stagetips. Stagetips were activated with 100% ACN/0.1% FA (buffer B) and washed twice with 5% FA. Samples resuspended in 5% FA were passed through each tip and washed with 0.1% FA (buffer A). Samples were eluted with a 2:1 mixture of buffer B and buffer A, and the eluate was vacuum centrifuged until dry.

LC-MS/MS

For data-dependent acquisition (DDA) liquid chromatography-tandem mass spectrometry (LC-MS/MS), labeled, phoshoenriched peptides were analyzed using a nano-HPLC coupled to MS. Coated nano-spray emitters were generated from fused silica capillary tubing (75 µm ID, 365 µm OD) with a 5–8 µm tip opening, using a laser puller (Sutter Instrument Co., model P-2000). Nano-spray emitters were packed with C18 reversed-phase material (Reprosil-Pur 120 C18-AQ, 3 µm) and resuspended in methanol using a pressure injection cell. The sample in 5% FA was directly loaded at 400 nL/min for 20 min onto a 75 µm × 15 cm nano-spray emitter. Peptides were eluted from the column with an ACN gradient generated by an Eksigent ekspert nanoLC 425 and analyzed on an Orbitrap Fusion Lumos Tribrid mass spectrometer (ThermoFisher). The gradient was delivered at 200 nL/min from 2.5% ACN with 0.1% FA to 35% ACN with 0.1% FA using a linear gradient of 120 min. This was followed by an 8-min gradient from 35% ACN with 0.1% FA to 80% ACN with 0.1% FA. Afterward, there was an 8-min wash with 80% ACN with 0.1% FA and equilibration for another 23 min to 2.5% ACN with 0.1% FA. The total DDA protocol was 180 min. The MS1 scan had an accumulation time of 50 ms within a mass range of 400–1500 Da, using an orbitrap resolution of 120,000, 60% RF lens, AGC target of 125%, and 2400 volts. This was followed by MS/MS scans with a total cycle time of 3 s. Accumulation time of 50 ms and 33% HCD collision energy were used for each MS/MS scan. Each candidate ion was required to have a charge state from 2-7 and an AGC target of 400%, isolated using an orbitrap resolution of 15,000. Previously analyzed candidate ions were dynamically excluded for 9 s.

MS data processing and analysis

Data files were processed and searched using MaxQuant (version 2.1.4.0) querying against Legionella pneumophila sequences (RefSeq accession: GCF_001941585.1) that were modified to include common contaminants and reverse sequences (FragPipe v.18.0). Search parameters were set to search for tryptic cleavages allowing two missed cleavages. Variable modifications included methionine oxidation, protein N-terminal acetylation, asparagine deamidation, serine, threonine, and tyrosine phosphorylation, and lysine dimethylation. Fixed modifications included carbamidomethylation. A maximum of five modifications per peptide was allowed. Other settings were left as default parameters. For dimethyl labeling, light labels were processed with DimethLys0/Nter0, intermediate labels with DimethLys4/Nter4, and heavy labels with DimethLys8/Nter8. Biological duplicates were processed separately by assigning different parameter groups with posttranslational modifications (PTM) set to TRUE. Peptides that were flagged as contaminants, not labeled as phosphorylated or only found in one biological replicate were filtered out, and the intensity columns (e.g., Intensity.H) were used for analysis. Data analysis and plotting were conducted in R (version 4.2.0).

HipBST conservation and accessory genome search across Legionella species

Genome assemblies for 58 representative Legionella species were retrieved in July 2020 from the NCBI RefSeq database (Table S4). When possible, genome assemblies with higher sequencing coverage (23) were used instead. For Legionella pneumophila, a recently reannotated genome (Refseq ID: GCF_001941585.1) was used. ORFs were predicted using Prodigal (80) and homologs were determined using OrthoMCL (25). A core genome phylogeny was constructed by aligning all conserved ORFs with MUSCLE and building a phylogenetic tree with RAxML (JTTDCMUT model selected by PROTGAMMAAUTO) (27), consistent with previous work (22). The resulting tree was annotated using the R package ggtree (28). All software was run with default parameters. To identify non-essential gene clusters, genes in L. pneumophila were first annotated as essential if they were reported to demonstrate a growth defect in broth (50, 51) or replication within a host (51) when deleted or were conserved across all 58 species. These annotations were then applied to homologous groups across all Legionella genomes. To predict non-essential clusters, a 10-gene sliding window analysis was performed on each genome and used to define regions where the local maximum of essential genes never exceeded 10%. In addition, each region was required to be larger than 1% of the total genome size. Gene neighborhood analysis across Legionella species with HipBST systems was performed using FlaGs (29) with default parameters to retrieve 10 genes upstream and downstream of hipS in each genome.

Phylogenetic analysis of HipBST systems across bacterial species

The NCBI RefSeq database was queried using cblaster (default parameters) (30) in November 2021 to retrieve homologous sequences of HipBA or HipBST systems. We excluded systems that have an intergenic distance of above 150 bp and/or contained partial protein sequences. Five seed templates were used for this search: E. coli K12 HipBA (HipBA_Ec; NP_416025.1, NP_416024.1), Shewanella oneidensis MR-1 HipBA (HipBA_So; AAN53783.2, AAN53784.1), Bacteroides uniformis HipBA (HipBA_Bu; WP_149924066.1, WP_149924064.1), E. coli O127 HipBST (HipBST_Ec; WP_000563102.1, WP_001346664.1, WP_001262465.1), and L. pneumophila HipBST (HipBST_Lp; AAU28429, AAU28430, AAU28431). We first confirmed that no hit accession or sequence was retrieved by multiple system seeds in order to ensure that no ambiguity in hit assignment occurred and that each retrieved hit corresponded to only one seed. The resulting protein sequences were clustered using MMseqs2 (31) to identify sequence-level representatives (80% identity, 80% coverage, coverage mode = 1), aligned with MAFFT (E-INS-i) (32), and trimmed with trimAl (automated1 for parameter selection) (33). HipB sequences were aligned separately and the HipS/HipT sequences were concatenated prior to alignment with the HipA sequences. A HipBA/HipBST phylogeny was inferred via the maximum likelihood method using IQ-TREE (34). We used the edge-proportional partition model and performed partitioned phylogenetic analysis by segregating the protein sequence alignments into three regions: the HipB alignments, the HipA N-terminus/HipS alignments, and the HipA C-terminus/HipT alignments. Branch confidence was assessed using an ultrafast bootstrap with 1,000 replications. The final tree was then manually inspected to ensure that all representative hits clustered with their respective seeds and that there was no ambiguity in the assignment of a hit to a given seed system. As a check for congruence, individual phylogenies were constructed for HipB and HipA/HipST (concatenated) homologs using the same protocol for alignment and trimming (see above). Individual maximum likelihood phylogenetic trees for HipB and HipA/HipST were constructed as above and ModelFinder (81) was used to determine the best partitioning scheme. To explore the phylogenetic distribution of bacteria harboring each system, species-level representatives of every taxon containing a system homolog from the initial search results were used to retrieve a phylogenetic tree from TimeTree (36). A second tree was retrieved after filtering for species with complete genome assemblies according to the NCBI Microbial Genome database, to ensure the absence of HipBA/HipBST systems is not due to genome incompleteness. Presence or absence was then compared for each system type across species and represented using the R package ggtree (28).

ACKNOWLEDGMENTS

We thank Rosa Di Leo for assistance in cloning the protein purification constructs and Dylan Valleau for help with protein purification for in vitro experiments. We thank Cassandra Wong and Brett Larsen for their help with the phosphoproteomic methodology and data analysis. We thank Dr. Malene Urbanus, Dr. Beth Nicholson, and Emma Brownlie for critical feedback on the manuscript.

J.D.L. was supported by an Ontario Graduate Scholarship. K.T.A. was supported by a CIHR CGS-D scholarship. This work was supported by the Natural Sciences and Engineering Research Council of Canada, Grants RGPIN-2020–06636 and RGPAS-2020–00014 to A.W.E.

Crystal structures solved in this work were funded in whole or in part with US federal funds from the National Institute of Allergy and Infectious Diseases, National Institutes of Health, Department of Health and Human Services, under contract No. HHSN272201700060C [Center for Structural Genomics of Infectious Diseases (CSGID); http://csgid.org]. Proteomics was performed at the Network Biology Collaborative Centre at the Lunenfeld-Tanenbaum Research Institute, a facility supported by Canada Foundation for Innovation funding, the Ontario Government, and Genome Canada and Ontario Genomics (OGI-139), and by a Canadian Institutes for Health Research Foundation grant to A.C.G.

References

Jackson RW, Vinatzer B, Arnold DL, Dorus S, Murillo J. 2011. The influence of the accessory genome on bacterial pathogen evolution. Mob Genet Elements 1:55–65.

McNally A, Oren Y, Kelly D, Pascoe B, Dunn S, Sreecharan T, Vehkala M, Välimäki N, Prentice MB, Ashour A, Avram O, Pupko T, Dobrindt U, Literak I, Guenther S, Schaufler K, Wieler LH, Zhiyong Z, Sheppard SK, McInerney JO, Corander J. 2016. Combined analysis of variation in core, accessory and regulatory genome regions provides a super-resolution view into the evolution of bacterial population. PLoS Genet 12:e1006280.

Fraikin N, Goormaghtigh F, Van Melderen L. 2020. Type II toxin-antitoxin systems. J Bacteriol 202.

Leplae R, Geeraerts D, Hallez R, Guglielmini J, Drèze P, Van Melderen L. 2011. Diversity of bacterial type II toxin-antitoxin systems: a comprehensive search and functional analysis of novel families. Nucleic Acids Res 39:5513–5525.

Makarova KS, Wolf YI, Koonin EV. 2009. Comprehensive comparative-genomic analysis of type 2 toxin-antitoxin systems and related mobile stress response systems in prokaryotes. Biol Direct 4:19.

Fiedoruk K, Daniluk T, Swiecicka I, Sciepuk M, Leszczynska K. 2015. Type II toxin-antitoxin systems are unevenly distributed among Escherichia coli phylogroups. Microbiology (Reading) 161:158–167.

Jurėnas D, Fraikin N, Goormaghtigh F, Van Melderen L. 2022. Biology and evolution of bacterial toxin-antitoxin systems. Nat Rev Microbiol 20:335–350.

Ronneau S, Helaine S. 2019. Clarifying the link between toxin-antitoxin modules and bacterial persistence. J Mol Biol 431:3462–3471.

Song S, Wood TK. 2020. Toxin/Antitoxin system paradigms: toxins bound to antitoxins are not likely activated by preferential antitoxin degradation. Adv Biosyst 4:e1900290.

LeRoux M, Culviner PH, Liu YJ, Littlehale ML, Laub MT. 2020. Stress can induce transcription of toxin-antitoxin systems without activating toxin. Mol Cell 79:280–292.

Van Melderen L, Saavedra De Bast M. 2009. Bacterial toxin-antitoxin systems: more than selfish entities PLoS Genet 5:e1000437.

Harms A, Brodersen DE, Mitarai N, Gerdes K. 2018. Toxins, targets, and triggers: an overview of toxin-antitoxin biology. Mol Cell 70:768–784.

Vang Nielsen S, Turnbull KJ, Roghanian M, Bærentsen R, Semanjski M, Brodersen DE, Macek B, Gerdes K. 2019. Serine-threonine kinases encoded by split homologs inhibit tryptophanyl-tRNA synthetase. mBio 10:e01138-19.

Keren I, Shah D, Spoering A, Kaldalu N, Lewis K. 2004. Specialized persister cells and the mechanism of multidrug tolerance in Escherichia coli. J Bacteriol 186:8172–8180.

Huang CY, Gonzalez-Lopez C, Henry C, Mijakovic I, Ryan KR. 2020. HipBA toxin-antitoxin systems mediate persistence in Caulobacter crescentus. Sci Rep 10:2865.

Boamah DK, Zhou G, Ensminger AW, O’Connor TJ. 2017. From many hosts, one accidental pathogen: the diverse protozoan hosts of Front Cell Infect Microbiol 7:477.

Ensminger AW. 2016. Legionella pneumophila, armed to the hilt: justifying the largest arsenal of effectors in the bacterial world. Curr Opin Microbiol 29:74–80.

Gomez-Valero L, Buchrieser C. 2019. Intracellular parasitism, the driving force of evolution of Legionella pneumophila and the genus Legionella. Genes Immun 20:394–402.

Xie Y, Wei Y, Shen Y, Li X, Zhou H, Tai C, Deng Z, Ou H-Y. 2018. TADB 2.0: an updated database of bacterial type II toxin-antitoxin loci. Nucleic Acids Res 46:D749–D753.

Zhu W, Banga S, Tan Y, Zheng C, Stephenson R, Gately J, Luo Z-Q, Kwaik YA. 2011. Comprehensive identification of protein substrates of the Dot/Icm type IV transporter of Legionella pneumophila. PLoS ONE 6:e17638.

Zhen X, Wu Y, Ge J, Fu J, Ye L, Lin N, Huang Z, Liu Z, Luo Z-Q, Qiu J, Ouyang S. 2022. Molecular mechanism of toxin neutralization in the HipBST toxin-antitoxin system of Legionella pneumophila. Nat Commun 13:4333.

Gomez-Valero L, Rusniok C, Carson D, Mondino S, Pérez-Cobas AE, Rolando M, Pasricha S, Reuter S, Demirtas J, Crumbach J, Descorps-Declere S, Hartland EL, Jarraud S, Dougan G, Schroeder GN, Frankel G, Buchrieser C. 2019. More than 18,000 effectors in the genus genome provide multiple, independent combinations for replication in human cells. Proc Natl Acad Sci U S A 116:2265–2273.

Burstein D, Amaro F, Zusman T, Lifshitz Z, Cohen O, Gilbert JA, Pupko T, Shuman HA, Segal G. 2016. Genomic analysis of 38 Legionella species identifies large and diverse effector repertoires. Nat Genet 48:167–175.

Gerdes K, Bærentsen R, Brodersen DE. 2021. Phylogeny reveals novel HipA-homologous kinase families and toxin-antitoxin gene organizations. mBio 12:e0105821.

Li L, Stoeckert CJ, Roos DS. 2003. OrthoMCL: identification of ortholog groups for eukaryotic genomes. Genome Res 13:2178–2189.

Edgar RC. 2004. MUSCLE: multiple sequence alignment with high accuracy and high throughput. Nucleic Acids Res 32:1792–1797.

Stamatakis A. 2014. RAxML version 8: a tool for phylogenetic analysis and post-analysis of large Phylogenies. Bioinformatics 30:1312–1313.

Yu G, Smith DK, Zhu H, Guan Y, Lam TT. 2017. ggtree: an R package for visualization and annotation of phylogenetic trees with their covariates and other associated data. Methods Ecol Evol.

Saha CK, Sanches Pires R, Brolin H, Delannoy M, Atkinson GC. 2021. FlaGs and webFlaGs: discovering novel biology through the analysis of gene neighbourhood conservation. Bioinformatics 37:1312–1314.

Gilchrist CLM, Booth TJ, van Wersch B, van Grieken L, Medema MH, Chooi Y-H. 2021. cblaster: a remote search tool for rapid identification and visualization of homologous gene clusters. Bioinform Adv 1:vbab016.

Steinegger M, Söding J. 2017. MMseqs2 enables sensitive protein sequence searching for the analysis of massive data sets. Nat Biotechnol 35:1026–1028.

Katoh K, Standley DM. 2013. MAFFT multiple sequence alignment software version 7: improvements in performance and usability. Mol Biol Evol 30:772–780.

Capella-Gutiérrez S, Silla-Martínez JM, Gabaldón T. 2009. trimAl: a tool for automated alignment trimming in large-scale phylogenetic analyses. Bioinformatics 25:1972–1973.

Nguyen L-T, Schmidt HA, von Haeseler A, Minh BQ. 2015. IQ-TREE: a fast and effective stochastic algorithm for estimating maximum-likelihood phylogenies. Mol Biol Evol 32:268–274.

Letunic I, Bork P. 2021. Interactive tree of life (iTOL) v5: an online tool for phylogenetic tree display and annotation. Nucleic Acids Res 49:W293–W296.

Kumar S, Stecher G, Suleski M, Hedges SB. 2017. TimeTree: a resource for timelines timetrees, and divergence times. Mol Biol Evol 34:1812–1819.

Mao F, Dam P, Chou J, Olman V, Xu Y. 2009. DOOR: a database for prokaryotic operons. Nucleic Acids Res 37:D459–D463.

Wen Y, Behiels E, Felix J, Elegheert J, Vergauwen B, Devreese B, Savvides SN. 2014. The bacterial antitoxin HipB establishes a ternary complex with operator DNA and phosphorylated toxin HipA to regulate bacterial persistence. Nucleic Acids Res 42:10134–10147.

Allombert J, Vianney A, Charpentier X. 2017. Monitoring effector translocation using the TEM-1 beta-lactamase reporter system. Methods Mol Biol 1615:489–499.

Correia FF, D’Onofrio A, Rejtar T, Li L, Karger BL, Makarova K, Koonin EV, Lewis K. 2006. Kinase activity of overexpressed HipA is required for growth arrest and multidrug tolerance in Escherichia coli. J Bacteriol 188:8360–8367.

Heidtman M, Chen EJ, Moy M-Y, Isberg RR. 2009. Large-scale identification of Legionella pneumophila Dot/Icm substrates that modulate host cell vesicle trafficking pathways. Cell Microbiol 11:230–248.

Black DS, Irwin B, Moyed HS. 1994. Autoregulation of hip, an operon that affects lethality due to inhibition of peptidoglycan or DNA synthesis. J Bacteriol 176:4081–4091.

Schumacher MA, Balani P, Min J, Chinnam NB, Hansen S, Vulić M, Lewis K, Brennan RG. 2015. HipBA-promoter structures reveal the basis of heritable multidrug tolerance. Nature 524:59–64.

Bærentsen RL, Nielsen SV, Skjerning RB, Lyngsø J, Bisiak F, Pedersen JS, Gerdes K, Sørensen MA, Brodersen DE. 2023. Structural basis for kinase inhibition in the tripartite E. Coli HipBST toxin-antitoxin system. eLife 12:RP90400.

Schumacher MA, Min J, Link TM, Guan Z, Xu W, Ahn Y-H, Soderblom EJ, Kurie JM, Evdokimov A, Moseley MA, Lewis K, Brennan RG. 2012. Role of unusual P loop ejection and autophosphorylation in HipA-mediated persistence and multidrug tolerance. Cell Rep 2:518–525.

Germain E, Castro-Roa D, Zenkin N, Gerdes K. 2013. Molecular mechanism of bacterial persistence by HipA. Mol Cell 52:248–254.

Kaspy I, Rotem E, Weiss N, Ronin I, Balaban NQ, Glaser G. 2013. HipA-mediated antibiotic persistence via phosphorylation of the glutamyl-tRNA-synthetase. Nat Commun 4:3001.

Zhou X, Eckart MR, Shapiro L. 2021. A bacterial toxin perturbs intracellular amino acid balance to induce persistence. mBio 12:e03020-20.

Semanjski M, Germain E, Bratl K, Kiessling A, Gerdes K, Macek B. 2018. The kinases HipA and HipA7 phosphorylate different substrate pools in Escherichia coli to promote multidrug tolerance. Sci Signal 11:eaat5750.

O’Connor TJ, Adepoju Y, Boyd D, Isberg RR. 2011. Minimization of the Legionella pneumophila genome reveals chromosomal regions involved in host range expansion. Proc Natl Acad Sci U S A 108:14733–14740.

Park JM, Ghosh S, O’Connor TJ. 2020. Combinatorial selection in amoebal hosts drives the evolution of the human pathogen Legionella pneumophila. Nat Microbiol 5:599–609.

Huang L, Boyd D, Amyot WM, Hempstead AD, Luo Z-Q, O’Connor TJ, Chen C, Machner M, Montminy T, Isberg RR. 2011. The E block motif is associated with Legionella pneumophila translocated substrates. Cell Microbiol 13:227–245.

Lifshitz Z, Burstein D, Peeri M, Zusman T, Schwartz K, Shuman HA, Pupko T, Segal G. 2013. Computational modeling and experimental validation of the Legionella and Coxiella virulence-related type-IVB secretion signal. Proc Natl Acad Sci U S A 110:E707–E715.

Charpentier X, Gabay JE, Reyes M, Zhu JW, Weiss A, Shuman HA. 2009. Chemical genetics reveals bacterial and host cell functions critical for type IV effector translocation by Legionella pneumophila. PLoS Pathog 5:e1000501.

de Felipe KS, Glover RT, Charpentier X, Anderson OR, Reyes M, Pericone CD, Shuman HA. 2008. Legionella eukaryotic-like type IV substrates interfere with organelle trafficking. PLoS Pathog 4:e1000117.

Conover GM, Derré I, Vogel JP, Isberg RR. 2003. The Legionella pneumophila LidA protein: a translocated substrate of the Dot/Icm system associated with maintenance of bacterial integrity. Mol Microbiol 48:305–321.

Jørgensen MG, Pandey DP, Jaskolska M, Gerdes K. 2009. HicA of Escherichia coli defines a novel family of translation-independent mRNA Interferases in bacteria and archaea. J Bacteriol 191:1191–1199.

Kurata T, Saha CK, Buttress JA, Mets T, Brodiazhenko T, Turnbull KJ, Awoyomi OF, Oliveira SRA, Jimmy S, Ernits K, Delannoy M, Persson K, Tenson T, Strahl H, Hauryliuk V, Atkinson GC. 2022. A hyperpromiscuous antitoxin protein domain for the neutralization of diverse toxin domains. Proc Natl Acad Sci U S A 119:e2102212119.

Shames SR, Liu L, Havey JC, Schofield WB, Goodman AL, Roy CR. 2017. Multiple effector virulence phenotypes revealed through high-throughput analysis of targeted mutant libraries. Proc Natl Acad Sci U S A 114:E10446–E10454.

Szekeres S, Dauti M, Wilde C, Mazel D, Rowe-Magnus DA. 2007. Chromosomal toxin-antitoxin loci can diminish large-scale genome reductions in the absence of selection. Mol Microbiol 63:1588–1605.

Wozniak RAF, Waldor MK. 2009. A toxin-antitoxin system promotes the maintenance of an integrative conjugative element. PLoS Genet 5:e1000439.

Huguet KT, Gonnet M, Doublet B, Cloeckaert A. 2016. A toxin antitoxin system promotes the maintenance of the Inca/C-mobilizable Salmonella genomic Island 1. Sci Rep 6:32285.

Zhao Y, Wang W, Yao J, Wang X, Liu D, Wang P. 2022. The HipAB toxin-antitoxin system stabilizes a composite genomic Island in CN-32. Front. Microbiol 13:858857.

Rao C, Benhabib H, Ensminger AW. 2013. Phylogenetic reconstruction of the Legionella pneumophila Philadelphia-1 laboratory strains through comparative genomics. PLoS One 8:e64129.

Yu H, Braun P, Yildirim MA, Lemmens I, Venkatesan K, Sahalie J, Hirozane-Kishikawa T, Gebreab F, Li N, Simonis N, Hao T, Rual J-F, Dricot A, Vazquez A, Murray RR, Simon C, Tardivo L, Tam S, Svrzikapa N, Fan C, de Smet A-S, Motyl A, Hudson ME, Park J, Xin X, Cusick ME, Moore T, Boone C, Snyder M, Roth FP, Barabási A-L, Tavernier J, Hill DE, Vidal M. 2008. High-quality binary protein interaction map of the yeast interactome network. Science 322:104–110.

Brachmann CB, Davies A, Cost GJ, Caputo E, Li J, Hieter P, Boeke JD. 1998. Designer deletion strains derived from Saccharomyces cerevisiae S288C: a useful set of strains and plasmids for PCR-mediated gene disruption and other applications. Yeast 14:115–132.

Alberti S, Gitler AD, Lindquist S. 2007. A suite of gateway cloning vectors for high-throughput genetic analysis in Saccharomyces cerevisiae. Yeast 24:913–919.

Dreze M, Monachello D, Lurin C, Cusick ME, Hill DE, Vidal M, Braun P. 2010. High-quality binary interactome mapping. Methods Enzymol 470:281–315.

Gietz RD, Schiestl RH. 2007. Large-scale high-efficiency yeast transformation using the LiAc/SS carrier DNA/PEG method. Nat Protoc 2:38–41.

Bailo N, Kanaan H, Kay E, Charpentier X, Doublet P, Gilbert C. 2019. Scar-free genome editing in Legionella pneumophila. Methods Mol Biol 1921:93–105.

Nicholson B, Deecker SR, Ensminger AW. 2022. Mobile element integration reveals a chromosome dimer resolution system in Legionellales. mBio 13:e0217122.

Valleau D, Quaile AT, Cui H, Xu X, Evdokimova E, Chang C, Cuff ME, Urbanus ML, Houliston S, Arrowsmith CH, Ensminger AW, Savchenko A. 2018. Discovery of ubiquitin deamidases in the pathogenic arsenal of Legionella pneumophila. Cell Rep 23:568–583.

Studier FW. 2005. Protein production by auto-induction in high density shaking cultures. Protein Expr Purif 41:207–234.

Minor W, Cymborowski M, Otwinowski Z, Chruszcz M. 2006. HKL-3000: the integration of data reduction and structure solution--from diffraction images to an initial model in minutes. Acta Crystallogr D Biol Crystallogr 62:859–866.

Afonine PV, Grosse-Kunstleve RW, Echols N, Headd JJ, Moriarty NW, Mustyakimov M, Terwilliger TC, Urzhumtsev A, Zwart PH, Adams PD. 2012. Towards automated crystallographic structure refinement with phenix.refine. Acta Crystallogr D Biol Crystallogr 68:352–367.

Emsley P, Lohkamp B, Scott WG, Cowtan K. 2010. Features and development of Coot. Acta Crystallogr D Biol Crystallogr 66:486–501.

Urbanus ML, Quaile AT, Stogios PJ, Morar M, Rao C, Di Leo R, Evdokimova E, Lam M, Oatway C, Cuff ME, Osipiuk J, Michalska K, Nocek BP, Taipale M, Savchenko A, Ensminger AW. 2016. Diverse mechanisms of metaeffector activity in an intracellular bacterial pathogen, Legionella pneumophila. Mol Syst Biol 12:893.

Kitagawa M, Ara T, Arifuzzaman M, Ioka-Nakamichi T, Inamoto E, Toyonaga H, Mori H. 2005. Complete set of ORF clones of Escherichia coli ASKA library (a complete set of E. coli K-12 ORF archive): unique resources for biological research. DNA Res 12:291–299.

Song C, Wang F, Ye M, Cheng K, Chen R, Zhu J, Tan Y, Wang H, Figeys D, Zou H. 2011. Improvement of the quantification accuracy and throughput for phosphoproteome analysis by a pseudo triplex stable isotope dimethyl labeling approach. Anal Chem 83:7755–7762.

Hyatt D, Chen G-L, Locascio PF, Land ML, Larimer FW, Hauser LJ. 2010. Prodigal: prokaryotic gene recognition and translation initiation site identification. BMC Bioinformatics 11:119.

Kalyaanamoorthy S, Minh BQ, Wong TKF, von Haeseler A, Jermiin LS. 2017. ModelFinder: fast model selection for accurate phylogenetic estimates. Nat Methods 14:587–589.