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 HipBSTLp). We identify HipBSTLp homologs in diverse taxa, yet in the Gammaproteobacteria, these are almost exclusively found in Legionella species. Notably, the toxin HipTLp 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 HipTLp 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 HipBSTLp, and identify key variations between this system and HipBSTEc. Finally, we show that the target of HipTLp 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 HipBSTLp) is a distant homolog of the recently discovered split-HipA system in Escherichia coli (HipBSTEc). 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 (HipTLp) 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.

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 wellstudied 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.

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 spe cies 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-charac terized 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.
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  S4) showing the conservation of Lpg2368-70 (HipBST Lp ) 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 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. pneumo phila (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 representa tive 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 concaten ated 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.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 HipBA Ec , HipBST Lp , and HipBST Ec 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 (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).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.

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.

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.
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 inhib itor 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 promiscu ous 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.
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 HipT Lp was expressed from the pJB1806 vector, and HipB Lp and HipS Lp were co-expressed from the pNT562 vector.Expression was induced with IPTG (100 µM).
(C) ADP-Glo kinase assay with purified recombinant His 6 -SBP-tagged HipT Lp .Both wild-type HipT Lp and HipT Lp 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 His 6 -SBP-tagged HipB Lp , HipS Lp , and HipT Lp .Reactions contained 1 µg of each protein and were incubated at 37°C for 30 min.(E) Expression of HipT Lp 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 HipT Lp .***P-value < 0.0001; α (0.05) was Bonferroni corrected for multiple hypothesis testing.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-oper ator 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, regard less 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 secon dary 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 1.6-2.2Å 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.
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 (S 57 IS 59 ), 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 influen zae 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.
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 toxininduced 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 coexpression 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 tRNAligase.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 homo logs 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 inter action 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).
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 (S 57 IS 59 ) 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, respec tively-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)(61)(62)(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 charac teristics, 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.

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-gal actoside (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), chloram phenicol (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 (OD 600 = 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% CO 2 .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 OD 600 = 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.

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 His 6 -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 OD 600 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 cryopro tected 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.refineand Coot (76).

In vitro kinase assays
Purified recombinant His 6 -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 MgCl 2 , 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 His 6 -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 OD 600 = 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 transforma tion efficiency.As a control, the empty vector pCA24N was transformed in an equiva lent manner to the pooled library.Library transformations without gene expression yielded 10 4 -10 5 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.

In-solution protein digestion
Cell lysates from the ∆hipBST cultures (vector-induced, HipT Lp -uninduced, and HipT Lpinduced) 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 Tryp sin/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.Pepti des were resuspended in 50 mM HEPES pH 8.0, and the HipT Lp -uninduced, vectorinduced, and HipT Lp -induced samples were designated to be labeled with the light, intermediate, and heavy channels, respectively.Peptides were mixed with freshly prepared CH 2 O (light, +28.0313Da), CD 2 O (intermediate, +32.0564Da), and 13 CD 2 O (heavy, +36.0757Da), and all solutions were made to 4% (vol/vol).NaBH 3 CN (0.6 M) was added to the light and intermediate channels, and NaBD 3 CN (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 TiO 2 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% NH 4 OH 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 spectrome try (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 reversedphase 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 Dimeth Lys0/Nter0, intermediate labels with DimethLys4/Nter4, and heavy labels with Dimeth Lys8/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 PROTGAM MAAUTO) (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-essen tial 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 assign ment occurred and that each retrieved hit corresponded to only one seed.The result ing 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 Nterminus/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).

FIG 1
FIG 1Legionella species contain a highly conserved and taxonomically restricted HipBST TA system.(A) A core genome phylogeny of 58 representative species in the Legionella genus (TableS4) showing the conservation of Lpg2368-70 (HipBST Lp ) homologs.The Legionella core genome was determined using OrthoMCL

FIG 2
FIG 2 HipT Lp 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 HipT Lp (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 HipT Lp in S. cerevisiae BY4742 cells.Yeast cultures carrying HipT Lp 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.

FIG 3
FIG 3 HipBST Lp is a functional tripartite TA system that restricts growth via HipT Lp kinase activity.(A) Expression of HipT Lp 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) HipT Lp co-expression with HipB Lp and HipS Lp in L. pneumophila ∆hipBST cells.

FIG 4
FIG 4 The HipBST Lp system has the capacity for complex autoregulatory dynamics.(A) Y2H experiments testing for binary physical interactions in the HipBST Lp 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 HipBST Lp proteins and a 200-bp DNA fragment (10 nM) encompassing the promoter region upstream of the hipBST Lp locus.HipB Lp was added to DNA at concentrations of 1, 2.5, 5, 10, and 20 nM, and HipS Lp /HipT Lp was added at a concentration of 20 nM.(C) Schematic of the operator sites upstream of hipBST Lp 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 hipBST Lp .HipB Lp 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 HipB Lp (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 HipS Lp and HipT Lp added to HipB Lp -DNA complexes (10 nM HipB Lp , 10 nM DNA).HipS Lp and HipT Lp 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.

FIG 5
FIG 5 HipT Lp neutralization by HipS Lp utilizes a P-loop ejection mechanism.(A) Alignment of the HipB Lp structure with HipB Ec (4YG1).(B) Comparison of the HipS Lp -HipT Lp co-crystal structure with HipA Ec (3DNT) in which the HipA Ec P-loop motif is in the internalized and active conformation.The HipA Ec N-subdomain 1, corresponding to HipS Lp , is indicated with a dashed box.The P-loops of both HipT Lp (red) and HipA Ec (black) are indicated.ATP in the HipA Ec 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 Mg 2+ binding (D219, D332).(C) Alignment of the HipS Lp -HipT Lp structure and pHipA Ec (3TPE), in which the P-loop motif is autophosphorylated and in the inactive, ejected conformation.Insets display two regions of variation between the HipS Lp -HipT Lp interaction interface and HipA Ec : a shifted helix in HipS Lp (inset on the left) and a loop in the N-terminal region of HipT Lp (inset on the right).Key residues in each region are labeled.(D) Alignment of the HipS Lp -HipT Lp structure with the structure of HipBST Ec (7AB5).(E) Left: inset from (F) comparing the tryptophan residue in HipS Ec (W65) with the methionine (M66) and isoleucine (I65) residues at the same position in HipS Lp .Right: co-expression of HipT Lp (pJB1806) with wild-type or mutant (I65A, M66A) HipS Lp (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.

FIG 6
FIG6 HipT Lp has an unknown cellular target that is not conserved with characterized HipT or HipA toxins.(A) Co-expression of HipT Lp (pJB1806) with GltX Lp or TrpS Lp (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 HipT Lp at both time points, and the vector control postinduction, were dimethyl labeled and multiplexed, enriched for phosphopeptides, and analyzed by mass spectrometry.The HipT Lp -uninduced, vector-induced, and HipT Lp -induced samples were labeled with the light, intermediate (med), and heavy channels, respectively.(C) Phosphopeptides detected in both HipT Lp -uninduced and induced samples (left) and those enriched during HipT Lp 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 HipT Lp phosphosite is underlined.(D) L. pneumophila cells co-expressing HipT Lp (pJB1806) with phosphoproteomic candidates (pNT562) that were either essential or highly conserved in both Legionella and conserved in E. coli.(E) Co-expression of HipT Lp with PheST Lp or phosphomimetic mutants of PheS Lp (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.

FIG 7
FIG7 Autoregulatory capacity of the HipBST Lp toxin-antitoxin system.HipT Lp is neutralized by physical interaction with HipS Lp and stable interactions are additionally observed between HipB Lp -HipT Lp and HipB Lp -HipS Lp -HipT Lp .HipB Lp can bind promoter DNA at three conserved operator sites upstream of hipBST Lp to regulate system expression and these sites contain unique sequence motifs relative to HipBA systems.Protein-DNA complexes formed by HipB Lp -DNA, HipT Lp -HipB Lp -DNA, and HipS Lp -HipT Lp -HipB Lp -DNA suggest the possibility for multiple layers of transcriptional regulation in the system that could be influenced by relative protein stoichiometry and complex stability.