Metagenome diversity illuminates origins of pathogen effectors

Abstract

Recent metagenome assembled genome (MAG) analyses have profoundly impacted Rickettsiology systematics. Discovery of basal lineages (Mitibacteraceae and Athabascaceae) with predicted extracellular lifestyles reveals an evolutionary timepoint for the transition to host dependency, which occurred independent of mitochondrial evolution. Notably, these basal rickettsiae carry the Rickettsiales vir homolog (rvh) type IV secretion system (T4SS) and purportedly use rvh to kill congener microbes rather than parasitize host cells as described for derived rickettsial pathogens. MAG analysis also substantially increased diversity for genus Rickettsia and delineated a basal lineage (Tisiphia) that stands to inform on the rise of human pathogens from protist and invertebrate endosymbionts. Herein, we probed Rickettsiales MAG and genomic diversity for the distribution of Rickettsia rvh effectors to ascertain their origins. A sparse distribution of most Rickettsia rvh effectors outside of Rickettsiaceae lineages indicates unique rvh evolution from basal extracellular species and other rickettsial families. Remarkably, nearly every effector was found in multiple divergent forms with variable architectures, illuminating profound roles for gene duplication and recombination in shaping effector repertoires in Rickettsia pathogens. Lateral gene transfer plays a prominent role shaping the rvh effector landscape, as evinced by the discover of many effectors on plasmids and conjugative transposons, as well as pervasive effector gene exchange between Rickettsia and Legionella species. Our study exemplifies how MAGs can provide incredible insight on the origins of pathogen effectors and how their architectural modifications become tailored to eukaryotic host cell biology.

INTRODUCTION

Until recently, Order Rickettsiales (Alphaproteobacteria) contained three families harboring diverse obligate intracellular parasites ¹. Rickettsiaceae and Anaplasmataceae are best studied and harbor invertebrate endosymbionts, human pathogens, and reproductive parasites ^2–7. Midichloriaceae contains some arthropod-associated bacteria of unknown vertebrate pathogenicity ⁸, but most species are described from protists ^9–14. Remarkably, Castelli and colleagues ¹⁵ described the first extracellular rickettsial species, “Candidatus Deianiraea vastatrix”, as a bacterium dependent on Paramecia and sharing many characteristics of the intracellular lifestyle. A new family, Deianiraeaceae, was proposed, calling into question the specific timepoint in rickettsial evolution wherein obligate intracellularity emerged from an obligate extracellular or facultative intracellular lifestyle.

Historically, Rickettsiales were widely considered a sister lineage to the mitochondria progenitor, with this assemblage representing a basal branch of the Alphaproteobacteria ^16–20. Pioneering work on rickettsial genomes supported this assertion, identifying decreased genome size and pseudogenization of genes within many metabolic pathways, conceptualized as “reductive genome evolution”, behind addiction to the eukaryotic cytosol ^18,21–25. This dogma held for two decades as hundreds of diverse Rickettsiales genomes were sequenced ^6,26. However, phylogenetic analysis of deep marine metagenome-assembled genomes (MAGs) illustrated that mitochondria likely originated outside of all described Alphaproteobacteria ²⁷, and recent phylogenomic description of certain novel MAGs established two basal rickettsial lineages, Mitibacteraceae and Athabascaceae, with features indicating an extracellular lifestyle not dependent on eukaryotic hosts ²⁸. These findings bolstered the growing trend for identifying mostly aquatic, protist-associated rickettsial species with traits (e.g., flagella, larger genome size, greater metabolic capacity, etc.) more characteristic of free-living and facultative intracellular bacteria but absent from the numerous genomes of well-characterized invertebrate- and animal associated rickettsial species ^{13,15,29–33}. Furthermore, they establish a Rickettsiales phylogenetic framework that allows for assessing the evolutionary trajectories within five derived lineages for innovations that emerged from transitions to host dependency.

Estimated to have arisen ~1.9 billion years ago ³⁴, the Alphaproteobacteria is highly diversified in form and function yet rife with convergence in morphology and lifestyle through common adaptation to numerous environments, including eukaryotic cells ³⁵. Order-level signatures are rare, making it curious that the genomes of Mitibacteraceae and Athabascaceae encode the Rickettsiales vir homolog (rvh) type IV secretion system (T4SS), a Rickettsiales signature that functions in overtaking eukaryotic cells ^36–40. The rvh T4SS is odd in its design ^41,42, with specialized duplications of some components hypothesized to autoregulate effector secretion ^43,44. Effectors have been characterized for species of Ehrlichia ^45–48, Anaplasma ^49–54, and Rickettsia ^55–57. Schön et al. proposed that species of Mitibacteraceae and Athabascaceae utilize the rvh T4SS for killing congener microbes ²⁸, provided that these genomes harbor candidate rvh effectors with characteristics similar to effectors in other T4SS and T6SS killing machines ^58,59. Thus, the five derived families purportedly repurposed the rvh T4SS to secrete effectors that commandeer host cellular processes to support intracellular replication (or epicellular parasitism in the case of Deianiraeaceae species).

The existence of an ancient secretion machine, yet independent gain of its effectors later in evolution, prompted us to poll the ever-growing MAG diversity for clues on rvh effector origins. We focus on known or candidate effectors from the genus Rickettsia, as recent studies have considerably expanded Rickettsiaceae diversity. Genome sequences from “environmental” Rickettsiaceae species (i.e. those from protists, apicomplexans, diplomonads, crustaceans, and insects) have illuminated basal lineages of Rickettsiaceae that are critical for inferring the emergence of genomic traits in Orientia and Rickettsia pathogens ^29,60–64. Further, phylogenetic analysis of genome sequences from novel genera “Candidatus Sarmatiella” (paramecium symbiont) ⁶⁵ and “Candidatus Megaira” (symbionts of algae and ciliates) ^31,66 indicate Orientia and Rickettsia species are more divergent than previously appreciated. Finally, a long-standing recognized basal lineage of Rickettsia termed “Torix Group”, which is highly diverse and widely present in non-blood-feeding arthropods ^67–71, was recently classified as a new genus, “Candidatus Tisiphia”, in a study that identified many new provisional Rickettsia (and Tisiphia) species from MAG analyses of diverse arthropods ⁷².

We present phylogenomic analyses that effectively demonstrate the utility of MAG data for not only inferring the origins of pathogen effectors, but for better understanding effector architectures through enhanced predictive power from greater sequence diversity. Provided that many MAGs come from environmental sampling or eukaryotes with no known human association, our approach stands to inform on the evolution of vertebrate pathogenesis.

RESULTS AND DISCUSSION

Mapping the Acquisition of Rickettsial Effectors.

We hypothesize that the transition to an intracellular lifestyle necessitated the acquisition of a more diverse effector repertoire. Thus, to gain an appreciation of the origins and conservation of Rickettsia rvh T4SS effectors, we performed a phylogenomics analysis encompassing the newly appreciated rickettsial diversity (FIG. 1A). This initially involved creating a matrix of taxa (genomes and metagenomes) determined to encode the rvh T4SS and the distribution of effectors. Six rvh effector molecules (REMs: RalF, Pat1, Pat2, Risk1, RT0527, and RARP-2) and 14 candidate REMs (cREMs) were evaluated based on prior studies implicating their secretion and/or interaction with the rvh coupling protein (RvhD4), or presence of motifs known to target either congener bacteria or eukaryotic molecules ^{55–57,73–80} (FIG. 1B). Our analyses expanded two REMs (Pat1 and RARP-2) and four cREMs based on identification of duplications (cREM-1, cREM-2, cREM-4), a partner protein (cREM-5), and a domain within the surface cell antigen (sca) Sca4 that we demonstrate to be widespread in non-Sca4 proteins. Collectively, a total of 26 proteins were analyzed within the phylogenomic framework (FIG. 1C; FIG. 2). A phylogeny was estimated from concatenated alignments of RvhB4-I and RvhB4-II proteins from 153 genome assemblies (FIG. 1C; FIG. 2; Table S1; FIG. S2). This collective matrix is an effective framework for mapping the earliest occurrence of these rvh effectors in the rickettsial tree, additionally identifying several likely origins for lateral gene transfer (LGT).

FIGURE 1. Probing Rickettsiales diversity for the evolution of Rickettsia type IV secretion system effectors.

(A) The atypical Rickettsiales vir homolog (rvh) T4SS is a hallmark of Rickettsiales, arising before the origin of host dependency ^28,197. Schema depicts recent genome-based phylogeny estimation ²⁸. Rvh characteristics ^37,44 are described at bottom and further in FIG. S1. (B) List of Rickettsia rvh effector molecules (REMs) and candidate REMs (cREMs). GBA, guilty by association. ‘Implied’ means analogous proteins are known to be secreted by other bacteria and/or the effector has strongly predicted eukaryotic molecular targets. Secretion, coimmunoprecipitation (CoIP), and bacterial 2-hybrid (B2H) data are compiled from prior reports ^{55–57,73–80}. SWA, Schuenke Walker Antigen domain. (C) Phylogenomic analysis of Rickettsia REMs and cREMs in non-Rickettsiaceae lineages. Cladogram summarizes a phylogeny estimated from concatenated alignments for RvhB4-I and RvhB4-II proteins from 153 rickettsial assemblies (full tree, FIG. S2; sequence information, Table S1). Non-Rickettsiaceae lineages are shown (see FIG. 2 for Rickettsiaceae). Dashed box for Pat1 proteins indicates the inability to discern Pat1a and Pat1b homology outside of Tisiphia and Rickettsia species (see FIG. 2; FIG. 4). SWAMP, SWA modular proteins. Information for all REMs and cREMs is provided in Table S2.

FIGURE 2. Phylogenomic analysis of Rickettsia REMs and cREMs in Rickettsiaceae.

Cladogram (continued from FIG. 1C) summarizes a phylogeny estimated from concatenated alignments for RvhB4-I and RvhB4-II proteins from 153 rickettsial assemblies (full tree, FIG. S2; sequence information, Table S1). “Candidatus Sneabacter namystus” (highlighted yellow) was manually added to the cladogram based on prior phylogeny estimation ^184,185 as this species lacks rvh genes but carries a type VI secretion system T6SS (see FIG. S3). Black boxes provide short names for 29 MAGs from Davison et al. ⁷² (NOTE: the clade colored green comprises genus Tisiphia though genus name Rickettsia reflects NCBI taxonomy as of Feb. 26^th, 2023). Asterisks depict multiple genome assemblies for a species. BG, Bellii Group; TRG, Transitional Group; TG, Typhus Group; TIG, Tamurae-Ixodes Group; SFG, Spotted Fever Group. Dashed box for Pat1 proteins indicates the inability to discern Pat1a and Pat1b homology outside of Tisiphia and Rickettsia species (see FIG. 4). Yellow boxes denote Risk2 proteins that are appended to C-terminal Schuenke-Walker antigen (SWA) domains (see FIG. 5). SWAMP, SWA modular proteins; all other REMs and cREMs are described in FIG. 1B and Table S2). Half circles for rCRCT-3a depict the presence of one or more antidotes but no toxin.

Origins of REMs.

An emerging diversity of bacterial Arf-GEFs.

Bacterial mimicry of eukaryotic-like Sec7 domains (S7D) to function as guanine nucleotide exchange factors (GEF) for host ADP-ribosylation factors (Arfs) was first described for Legionella pneumophila, which utilizes the dot/icm T4SS effector RalF to recruit and activate host Arf1 to the Legionella-containing vacuole (LCV) ^81,82. Certain Rickettsia genomes encode RalF proteins that are remarkably similar to Legionella counterparts across the S7D, as well as a Sec7 capping domain (SCD) that restricts access to the catalytic site ^83–85. The SCD has high specificity for host membranes and differentially regulates effector subcellular localization for Legionella (the LCV) and Rickettsia (the inside face of plasma membrane) RalF ⁸⁶. Rickettsia RalF was the first characterized REM; its secretion during host cell invasion activates host Arf6 at the plasma membrane, a process driven by a unique C-terminal extension, termed Variable with Pro-rich region (VPR), that interacts with host actin and phosphatidylinositol 4,5-biphosphate (PI(4,5)P₂) at entry foci ^55,87. The presence of ralF in the genomes of some Rickettsia pathogens but absence in non-pathogenic species led to speculation that this REM may be a lineage specific virulence factor ^55,87,88. Further, while species of Rickettsia and Legionella exchange genes in common intracellular environments ^89,90, the absence of ralF in any other known bacteria precluded insight on the origin of RalF and specifically how the diverse C-terminal architectures arose.

Our analyses provide clarity on RalF evolution by unearthing numerous bacterial analogs with novel S7D-containing architectures (FIG. 3; FIG. S4). First, an unusual Legionella RalF from L. clemsonensis was found to carry a conserved domain at its C-terminus that was also detected in a large ankyrin (ANK) repeat-containing protein of the Rickettsia endosymbiont of Graphium doson (Gdoso1) genome (FIG. 3A; FIG. S4A,B). This Gdoso1 protein also contains another conserved domain at its C-terminus that is widespread in Rickettsia genomes. These observations indicate frequent recombination in conjunction with LGT of these diverse genes. Second, while the most basal Rickettsia species (endosymbiont of Oxypoda opaca, “Oopac6”) harbors a RalF with the C-terminal VPR , two Tisiphia species (endosymbionts of Bryobia graminum, or “Moomin”, and Culicoides impunctatus, or “RiCimp”) are Legionella-like in C-terminal domain architecture (FIG. 3A,C; FIG. S4C–E). RiCimp RalF is encoded on a plasmid (pRiCimp001), the first for a ralF gene (FIG. 3D), supporting original speculation for RalF exchange between Legionella and Rickettsia species ⁸¹. Remarkably, pRiCimp001 also carries a toxin-antidote (TA) module highly similar to the plasmid-encoded TA module of Rickettsia felis str. LSU-Lb ⁹¹ that we previously described as part of the mobilome shuttling reproductive parasitism (RP) genes across Rickettsia and wolbachiae ⁹¹ (discussed below, section cREMs with Characterized Function).

FIGURE 3. MAG analysis divulges a greater diversity of Sec7-domain-containing proteins than previously appreciated.

Black boxes provide short names for MAGs from Davison et al. ⁷². These and additional newly discovered RalF-like proteins (highlighted yellow) substantially expand the prior RalF diversity. Structural models for proteins are found in FIG. S4D–F. (A,B) Novel insight from new (A) Legionella and rickettsial architectures and (B) new RalF-like proteins discovered in MAGs. Red shading and numbers indicate % aa identity across pairwise alignments (sequence information in Table S2). All protein domains are described in the gray inset. (C) Comparison of the Legionella pneumophila RalF structure (PDB 4C7P) ⁸⁵ with predicted structures of S7D-SCD regions of RiCimp RalF (LF885_07310) and R. typhi RalF (RT0362), and S7Ds of Mycobacteriaceae sp. co.spades.METABAT.1kb_110 (K2X97_15435) and Proteobacterium SZAS-39 (JSR17_09325). The delineation of the Sec7 domain (S7D, red) and Sec7-capping domain (SCD, green if present), is shown with an approximation of the active site Glu (asterisk). Additional eukaryotic-like domains for the non-rickettsial proteins are noted. Modeling done with Phyre2 ¹⁹². More detailed structural explanation is found in FIG. S4C. (E) RiCimp plasmid pRiCimp001 carries RalF and a TA module similar to those characterized in reproductive parasitism ⁹¹. Gene region drawn to scale using PATRIC compare region viewer tool ¹⁹⁸. Yellow, transposases and other mobile elements; skull-and-crossbones, pseudogenes; other domains described in gray inset at bottom. Plasmid map created with Proksee (https://proksee.ca/).

Finally, RalF proteins from three additional rickettsial species and two putative gammaproteobacterial species carry both the S7D and SCD but no C-terminal extensions (FIG. 3B; FIG. S4E,F). Four other S7D-containing proteins from non-rickettsial bacteria lack SCDs; however, two contain eukaryotic domains found in RHO GTPase-activating proteins (RHOGAP) that also target Arfs (FIG. 3B,C; FIG. S4F). All of these collective proteins have a highly conserved S7D and SCD (if present) and include most of the structural features that define RalF proteins (FIG. S4G). These collective characteristics attest to LGT disseminating the S7D-SCD architecture across divergent bacteria, with recurrent gains of additional domains tailored to eukaryotic cell functions (e.g., VPR, ANK, and RHOGAP). We posit that the acquisition of the Rickettsia-unique C-terminal VPR occurred early in Rickettsia evolution, with multiple losses of RalF in more than half of the sequenced species (FIG. 2).

Patatins: divergent phospholipases are recurrent in Rickettsiales.

Rickettsiae (and other Rickettsiales species lysing the phagosome and/or lysing host cells) require membranolytic effectors throughout the intracellular lifestyle. Phospholipase D (PLD) is a highly conserved enzyme with demonstrated membranolytic activity in a surrogate expression system ⁹² though an unresolved function during Rickettsia infection of host cells ⁹³. PLD contains a N-terminal Sec signal ⁸⁸, yet other phospholipase A₂ (PLA₂) enzymes (patatins Pat1 and Pat2) have sequence characteristics of rvh substrates ⁷⁴ and Pat2 binds RvhD4 in coimmunoprecipitation assays ⁵⁶ (FIG. 1B). Studies on R. typhi have shown that both Pat1 and Pat2 are secreted during host cell infection, require host cofactors for activation, and function early in infection by facilitating phagosome escape ^74,75. Recent work on R. parkeri Pat1 also demonstrated a role in phagosome escape in addition to facilitating avoidance of host polyubiquitination and autophagosome maturation, as well as promotion of actin-based motility and intercellular spread ⁹⁴. R. parkeri lacks Pat2, which is slightly more restricted in Rickettsia genomes and possibly provides a function in host cell lysis for rickettsiae that do not spread intercellularly without host cell lysis (e.g., TG rickettsiae).

All patatins share a common active site architecture that is critical for PLA₂ activity (FIG. 4A). Despite this, Pat1 and Pat2 are highly divergent outside of the patatin domain and have different origins based on phylogeny estimation ⁷⁴. Further, Pat1 proteins form two distinct groups, Pat1A and Pat1B, with pat1B found on plasmids and often recombining with chromosomal pat1 loci ⁷⁴. Utilizing newly discovered rickettsial patatins from MAGs, we show that all three enzymes (Pat1A, Pat1B, and Pat2) have distinct sequence profiles, with Pat1B proteins having a highly length variable C-terminal region relative to Pat1A and Pat2 enzymes (FIG. 4A,B). Despite this, Pat1B and Pat2 proteins have cryptic similarity across their C-terminal regions to support robust modeling to the crystal structure of L. pneumophila dot/icm T4SS effector VipD ⁹⁵ (FIG. 4C). During L. pneumophila host cell infection, secreted VipD localizes to host endosomes, catalyzing the removal of phosphatidylinositol 3-phosphate (PI(3)P) from endosomal membranes (N-terminal patatin domain) and binding Rab5 or Rab25 (C-terminal domain), ultimately blocking endosome-LCV fusion ^95–97. As with RalF, it is likely that Rickettsia Pat1 and Pat2 proteins have rudimentary analogous functions to VipD (targeting host membranes and binding host Rabs) but spatiotemporal and biochemical differences provided that rickettsiae lyse the phagosome and seemingly do not interfere with early endosome trafficking on par with Legionella species.

FIGURE 4. Divergent patatin phospholipases are recurrent in rickettsial evolution.

Black boxes provide short names for MAGs from Davison et al. ⁷². (A) PLA₂ active site characteristics and divergent patatin forms. Green inset describes general patatin domain and active site architecture. HaloBlast results for Pat1A, Pat1B, and Pat2 (query sequences described at top) are shown, with top-scoring halos boxed (full results in TABLE S4). (B) Sequence logo ¹⁸⁷ showing conservation of the PLA₂ active site motifs across Tisiphia and Rickettsia patatins (sequence information provided in Table S2). Pat1A, Pat1B, and Pat2 sequences were aligned separately with MUSCLE ¹⁸⁶ (default parameters) with active site motifs compiled for conservation assessment. Features unique to each patatin are noted. (C) Legionella pneumophila VipD structure (PDBID: 4AKF) and modeling of three rickettsial patatins to VipD using Phyre2 ¹⁹². (D) Diverse architectures for select patatins. Red shading and numbers indicate % aa identity across pairwise alignments (sequence information in Table S2). All protein domains are described in the gray inset. Dark green indicates Pat1 domains not grouped into A or B. (E) Four Rickettsia plasmids carry pat1B (shaded orange). Plasmid maps created with Proksee (https://proksee.ca/).

We detected Pat1 and Pat2 proteins in several non-Rickettsiaceae genomes (FIG. 1C), with some genomes (e.g., novel sea urchin and cabbage root fly endosymbionts in the Anaplasmataceae ^98,99) harboring duplications (FIG. 4D). Pat1 proteins from Rickettsiales species outside of the genera Tisiphia and Rickettsia could not be confidently assigned to either Pat1A or Pat1B (FIG. 1C; FIG. 2; dark green domains in FIG. 4D). Most species of Tisiphia and Rickettsia carry either Pat1A or Pat1B and/or Pat2 (FIG. 2). The only two species carrying all three distinct enzymes (R. bellii and R. amblyommatis) have Pat1B encoded on a plasmid (FIG. 4E). Overall, the patchwork distribution of these divergent enzymes, evidence for modular domain diversification, and presence on plasmids indicate that PLA₂ activities for rickettsiae are lineage-specific and subject to continual patatin gene gain and loss throughout evolution. Further, certain pat gene profiles may confer advantages in particular hosts.

Domain repurposing is Risky business.

Bacterial pathogens can directly modify host membrane phosphaditylinositol (PI) composition by mimicking eukaryotic kinases, phosphatases, and phosphotransferases ^100–103. Secreted PI kinases from intracellular pathogens R. typhi (Risk1), L. pneumophila (LegA and LepB), and Francisella tularensis (OpiA) alter the PI composition on phagosomes to prohibit maturation and fusion with lysosomes ^56,104,105. Characterized as either PI3 (Risk1, LegA, and OpiA) or PI4 (LepB) kinases, these enzymes possess a similar PI3/4 active site architecture (pfam00454) analogous to eukaryotic PI kinases, as well as certain protein kinases, that function in a myriad of membrane-associated functions, including intracellular signalling and trafficking ¹⁰⁶ (FIG. 5A,B). Subverting these host cell functions is highly advantageous to intracellular parasitism; thus, the dearth of identified PI3/4 kinase effectors likely reflects the cryptic nature of the PI3/4 active site within these proteins, which lack similarity outside of the PI3/4 active site domain ^56,105.

FIGURE 5. Discovery of a novel rickettsial PI kinase that can associate with a widespread rickettsial surface antigen.

Black boxes provide short names for MAGs from Davison et al. ⁷². Amino acid coloring is described in the FIG. 3 legend. (A) Previous work (above dashed line) identified a Rickettsia PI kinase, Risk1, with a cryptic active site similar to human and other bacterial PI3/PI4 kinases and related protein kinases. Colored shapes depict characterized substrate specificity (see panel B). Our study (below dashed line; select proteins shown) identified new rickettsial Risk1 proteins, as well as a second PI kinase (Risk2) also prevalent in rickettsial genomes and MAGs (FIG. 1C; FIG. 2). All PI3/PI4 kinase domains were aligned using MUSCLE ¹⁸⁶ (default parameters). Sequence information is provided in Table S2. Yellow highlighting on end coordinates denotes Risk2 proteins fused to a C-terminal SWA domain (see panel G; FIG. S5). (B) Mechanisms of phosphorylation on the PI inositol ring at 3’, 4’ and 5’ positions. Data for R. typhi Risk1 is superimposed in red ⁵⁶. (C) HaloBlast results (R. typhi Risk1 and Risk2 as queries) broken down to illustrate the presence of Rickettsia-like PI kinases in MAGs and the similarity between Risk2 and Legionella PI kinases (full data in Table S2). (D) Risk1 threads with high confidence (90.7%, 72% coverage; Phyre2 ¹⁹²) to the Helicobacter pylori proinflammatory kinase CtkA ¹⁰⁸. (E) Risk2 threads with high confidence (85.1%, 9% coverage; Phyre2 ¹⁹²) to a limited region of LepB, a Rab GTPase-activating protein effector from L. pneumophila ¹⁹⁹. (F) Risk1 and Risk2 proteins have cryptic and distinct PI3/4 active sites, yet lack similarity outside of these regions. Logos depict individual alignments, which are summarized at top and were performed with MUSCLE ¹⁸⁶, default settings. (G) Rickettsiae utilize a diverse arsenal of PIK effectors, some of which are tethered to SWA domains. Six select species are shown with their full complement of PI3/4 kinase and SWA architectures. Red shading and numbers indicate % aa identity across pairwise alignments (sequence information in Table S2). R. peacockii pRRP proteins are shaded orange (see panel H). (H) Plasmid pRRP of R. peacockii str. Rustic ^112,200 carries a divergent Risk2 gene that is adjacent to an ORF encoding a SWAMP (orange highlighting). Plasmid map created with Proksee (https://proksee.ca/).

BlastP and HMMER ¹⁰⁷ analyses using only the Risk1 PI3/4 active site revealed nearly 300 Rickettsiales proteins, with many genomes having multiple divergent kinases. Further inspection revealed the presence of a second conserved protein harboring the PI3/4 active site, which we named Rickettsia intracellular secreted kinase-2 (Risk2) (FIG. 5A). Notably, the R. typhi protein (RT0527) was captured in the same RvhD4 coimmunoprecipitation assay that identified Risk1 as a REM ⁵⁶ (FIG. 1B). HaloBlast analyses of full length Risk1 and Risk2 proteins indicated distinct profiles, with Risk2 sharing low similarity to Legionellales kinases (FIG. 5C). Structural analyses corroborated this result, with limited regions of Risk1 and Risk2 modelling best to structures of Helicobacter pylori CtkA ¹⁰⁸ and L. pneumophila LepB ¹⁰⁴, respectively (FIG. 5E,F). Comparison of Risk1 and Risk2 PI3/4 active sites revealed 1) juxtapositioned aromatic residues in their catalytic loops, 2) the presence of a postively charged residue in the acivation loops of most Wortmannin-sensitive kinases (Risk1 and human class 1 and 2 PI3 kinases), and 3) greater sequence length between catalytic and activation loops in Risk2 proteins (FIG. 5F). Furthermore, only LepB and some rickettsial Risk2 proteins have the catalytic loop Asp replaced by Ser. These collective observations indicate two divergent PI3/4 kinases encoded in most rickettsial genomes (FIG. 2), leading us to posit that Risk2 is a PI4 kinase that complements the PI3 kinase activity of Risk1 ⁵⁶.

We determined a remarkable connection between Risk2 and another Rickettsia effector, Sca4, which is highly conserved in Rickettsia species and implicated in intercellular spread by reducing mechanotransduction at cell-cell junctions ^76,109. The Sca4 C-terminal region has eukaryotic-like vinculin binding sites (VBS) that reduce vinculin-α-catenin interactions, which facilitates neighboring cell engulfment of Rickettsia-induced protrusions. The N-terminal region, shown by Schuenke and Walker ¹¹⁰ to elicit anti-rickettsia antibodies (Pfam: 120_Rick_ant), has a novel structure posited to guide assembly of the VBSs at host plasma membranes ¹¹¹. Curiously, the plasmid pRPR of R. peacockii carries two adjacent genes encoding a divergent Risk2 and the N-terminal Sca4 domain ¹¹² (FIG. 5F,G). This led to discovering a cohort of Tisiphia species and other MAGs encoding chimeric Pat2-Sca4-NT proteins, as well as numerous other protein architectures with the Sca4-NT domain (FIG. 2; FIG. S5; FIG. 5G). We renamed this pervasive domain Schuenke Walker Antigen (SWA) and suggest using SWA-modular protein (SWAMP) or coupling “SWA” with another domain when function is apparent; e.g., SWA-PIK (for PI kinase) or SWA-VBD (vinculin-binding domain). In accord with prior observations on high rates of recombination shaping Sca architectures ^77,113,114, these results indicate the SWA domain is a recurrent module that may shuttle various types of effector functions to common host cell targets, particularly membranes.

From many forms, one descendent.

Early in host infection, pathogens R. typhi and R. rickettsii secrete the REM RARP-2, which traffics to the endoplasmic reticulum and Golgi apparatus, leading to trans-Golgi network (TGN) fragmentation and ultimately perturbed protein transport to the host cell surface ^57,115. Like RalF, RARP-2 has a C-terminal tail that binds RvhD4 (FIG. 1B); furthermore, the protein has well delineated N-terminal protease and C-terminal ANK repeat domains (FIG. 6A). The protease domain has minimal analogy to clan CD protease families (C13 legumain, C14 caspase 1, C11 clostripain, and C25 gingipain R), which share a common fold that arranges a His and Cys catalytic dyad ¹¹⁶. This active site is essential for RARP-2 fragmentation of TGN ¹¹⁵ and also contributes to the lytic plaque phenotype of virulent R. rickettsii strains ⁵⁷. The ANK repeat domain is atypical among most ANK repeat-containing proteins ¹¹⁷, as the composition of each repeat is highly similar in length and identity (FIG. 6B; FIG. S6A–C) despite a highly variable repeat number across orthologs, even at the strain level in most cases (FIG. 6C). RARP-2 active site mutants still traffic to perinuclear membranes, indicating that the ANK domain drives subcellular localization. However, shorter repeats (4 in attenuated strains Iowa R. rickettsii) do not disrupt TGN fragmentation relative to those of pathogenic R. rickettsii ¹¹⁵, suggesting a larger ANK domain is required for proper localization.

FIGURE 6. RARP-2 architecture is derived from multiple divergent forms.

Black boxes provide short names for MAGs from Davison et al. ⁷². Amino acid coloring is described in the FIG. 3 legend. Sequence logos constructed with WebLogo 3 ¹⁸⁷. Sequence information in Table S2. (A) General architecture of RARP-2 proteins deduced from an alignment of 53 non-redundant RARP-2 proteins using MUSCLE ¹⁸⁶ (default parameters). (B) Consensus sequence for the RARP-2 ANK repeat deduced from 404 repeats. (C) Depiction of the 53 non-redundant RARP-2 proteins with ANK domain repeat number provided. For brevity, some strain names are not shown for R. prowazekii: Chernikova, Katsinyian, Dachau, BuV67-CWPP, Rp22; R. japonica: YH, DT-1, HH-1, HH06154, HH07124, HH07167, MZ08014, Nakase, PO-1, Tsuneishi, HH-13, HH06116, HH06125, LON-151, M11012, M14012, M14024, SR1567, YH_M; R. heilongjiangensis: HCN-13; Sendai-29; Sendai-58. (D) RARP-2 and dRARP-2 proteins possess N-terminal domain clan CD cysteine protease-like active sites ¹¹⁶. Sequences were manually aligned to illustrate the conservation across all diverse protein groups. ‘Rick. endo. UWC8’, endosymbiont of Acanthamoeba str. UWC8 (not shown in FIG. 1C but closely related to endosymbiont of Acanthamoeba str. UWC36 in the Midichloriaceae). (E) Insight on RARP-2/dRARP-2 structure. Asterisks indicate proteins from panel D that were used in Phyre2 ¹⁹² searches to identify template structures for modeling ^118–120. A complete structure of R. typhi RARP-2 predicted with Alphafold ^194,195 corroborates these dRARP-2 models and indicates deviations on a common effector architecture (FIG. S6C).

Probing recently sequenced genomes and MAGs did not reveal RARP-2 sequences in the Bellii Group (BG) rickettsiae or other Rickettsiaceae genomes (FIG. 2), consistent with prior observations that RARP-2 is unique to derived Rickettsia lineages ⁸⁸. Yet, by focusing on the N-terminal protease domain, we discovered 56 divergent RARP-2 (dRARP-2) proteins that possess the clan CD active site architecture (FIG. 6D). These proteins were binned into six groups (FIG. S6D) that have very different ANK repeat domain identities (data not shown); further, several could be modeled to structures of eukaryotic ^118,119 and prokaryotic ¹²⁰ clan CD members (FIG. 6E). dRARP-2 proteins are predominantly found in BG rickettsiae and Tisiphia genomes, but likely shuttle in the intracellular mobilome given that one is carried by a Midichloriaceae species (endosymbiont of Acanthamoeba str. UWC8 ¹²¹). Based on the discordant genomic distribution of RARP-2 and dRARP-2 (FIG. 1C; FIG. 2) and the strong bias of RARP-2 in vertebrate-associated species, we speculate that RARP-2 and dRARP-2 may be tailored for similar functions related to TGN fragmentation yet well diverged to allow recognition of targets specific to disparate eukaryotic hosts. This is reminiscent of the recent discovery that R. parkeri utilizes different factors for apoptosis induction in ticks versus mammals ¹²².

cREMs with Unknown Function.

For five R. typhi hypothetical proteins (HPs) previously shown to interact with RvhD4 (cREM-1–5; FIG. 1B), MAG analyses provided substantial clarity on the mechanisms of evolution shaping their architectures. Four of these proteins are described below in light of newfound gene fission/fusion and duplication events (cREM-1, cREM-2, and cREM-4), as well as a greater role of conjugative transposons shaping Rickettsia evolution (cREM-5). Unexpectedly, the small cREM-3 (~93 aa) was determined to have widespread conservation in Rickettsiales yet also exist in certain other Proteobacteria (FIG. 1C; FIG. 2). While likely not a REM, our analyses revealed a potential structure associated with this curious protein (FIG. S7C).

Cryptic gene fission and duplication obscured by rapid divergence.

For cREM-1 and cREM-2, we utilized phylogeny estimation in conjunction with sequence analysis to predict gene fission (cREM-1) and duplication (cREM-2) events behind the evolution of these proteins. Neither protein was found outside of Rickettsiaceae (FIG. 1C; FIG. 2). cREM-1 proteins are streamlined from larger modular Tisiphia proteins that harbor the entire cREM-1 sequence as a domain; accordingly, we named these divergent cREM-1 (cREM-1d) (FIG. 7A). Some rickettsiae carry cREM-1 tandem duplications, though most genomes have one conserved gene and the second pseudogenized (FIG. S7B; red clade). Curiously, cREM-1 proteins have similarity to a repeat region within the O. tsutsugamushi effector OtDUB (FIG. 7A; FIG. S7A). This region in OtDUB binds clathrin adaptor-protein complexes AP-1 and AP-2 and harbors a cryptic Rac1 GEF domain ^123–125. This indicates cREM-1 proteins may have evolved from larger modular proteins with functions tailored to the eukaryotic cytosol, with repeat regions of these large effectors streamlining to smaller proteins encoded by tandem genes (FIG. 7B; FIG. S7A).

FIGURE 7. Gene fission and duplication has shaped the architectures of four candidate REMs.

Black boxes provide short names for 29 MAGs from Davison et al. ⁷². Gene regions drawn to scale using PATRIC compare region viewer tool ¹⁹⁸. Sequence information is provided in Table S2. (A) Similarity between O. tsutsugamushi effector OtDUB (CAM80065), divergent cREM-1 (cREM-1d), and cREM-1 proteins. OtDUB characterized domains: deubiquitinase (light blue), ubiquitin-binding (pink), cryptic Rac 1-like guanine nucleotide exchange factor (yellow), clathrin adaptor-protein complexes AP-1 and AP-2 (orange), phosphatidylserine-binding (gray) ^5,123,201. Red shading and numbers indicate % aa identity across pairwise alignments. (B) Cladogram depicting phylogeny estimation of 102 cREM-1d and cREM-1 proteins (see FIG. S7A for phylogram and methods). Inset describes classification of cREM-1 proteins, with clade colors matching the protein colors in the schema in panel A. (C) Diversification of cREM-2 proteins via duplication. (D) Cladogram depicting phylogeny estimation of 158 cREM-2d, cREM-2a, and cREM-2b proteins (see FIG. S7B for phylogram and methods). Inset describes classification of cREM-2 proteins, with clade colors matching the protein colors in the schema in panel C. (E) Ancient gene duplication of cREM-4 and location of cREM-4 genes in select Rickettsiales species. The cREM-4 pentapeptide repeat domain is illustrated in FIG. S7D. (F) cREM-5/5p loci occur in variable genomic regions, including plasmids and RAGEs. The complete RAGE for RiCle is illustrated in FIG. S7F. (G) Genomic distribution of ProP genes in Rickettsiales genomes. OP, Rickettsia endosymbionts of Oxypoda opaca (Oopac6) and Pyrocoelia pectoralis (Ppec13); OACHA, Rickettsia endosymbionts of Omalisus fontisbellaquei (Ofont3) and Adalia bipunctata, R. canadensis, R. helvetica, and R. asiatica.

cREM-2 proteins belong to pfam17422 (DUF5410: specific to Rickettsia species). Our analyses identified a second DUF5410-like protein encoded adjacent to cREM-2 proteins in many Rickettsia genomes (FIG. 2). Neither of these tandem duplicates (designated cREM-2a and cREM2-b) contain Sec signal sequences or other predictable features (FIG. 7C). Further, some BG rickettsiae and Tisiphia species harbor a third divergent cREM-2 (cREM-2d) that is absent from derived Rickettsia lineages. With the assumption that cREM-2d are an ancestral form, phylogeny estimation indicates cREM-2b proteins are more divergent than cREM-2a proteins (FIG. 7D), though all three protein architectures share high conservation in dozens of residues within the central region of these proteins (FIG. S7B).

cREM-4 proteins also show evidence of an ancestral duplication (FIG. 7E), though no genomes contain a complete duplicate gene, indicating a consistent pseudogenization event that rapidly followed cREM-4 duplication (FIG. 2). Despite their large size (~950 aa) these proteins contain only one observable feature, a small internal pentapeptide repeat (PR). While widespread in diverse bacterial proteins, PR function is generally unknown, though some bacterial PR-containing proteins can interact with DNA binding proteins ¹²⁶ and contribute to virulence ¹²⁷ (FIG. S7D). cREM-4 proteins are encoded in certain other Rickettsiaceae genomes, and like dRARP-2, a single Midichloriaceae species (“Candidatus Jidaibacter acanthamoeba” ³⁰) encodes a cREM-4 protein. While cREM-4 of BG rickettsiae lack the PR (FIG. S7D), nearly all derived Rickettsia genomes encode a complete cREM-4 protein, indicating retention of a conserved function after an ancestral duplication.

LGT of cREM-5 as a two-gene module across select species.

cREM-5 proteins were previously noted for their restricted distribution in TG rickettsiae and R. felis (both in the chromosome and on plasmid pRF) ¹²⁸. Our analyses yielded several novel findings. First, while absent from any Spotted Fever Group (SFG) or Tamurae/Ixodes Group (TIG) rickettsiae, cREM-5 proteins are highly conserved in all BG rickettsiae genomes, as well as in a few Tisiphia genomes (FIG. 2). Second, most cREM-5 genes have an associated protein, cREM-5 partner (cREM-5p), encoded immediately downstream (FIG. 7F). Despite conserved regions (FIG. S7E), neither protein has detectable domains or similarity to proteins in other Rickettsiales (FIG. 2). Third, cREM-5/5p genes have a strong co-occurrence with PropP-2 genes (black, FIG. 7F). ProP (Proline Betain Transporters of the Major Facilitator Superfamily) function in osmoregulation ^129,130 are proliferated in Rickettsia genomes, with seven conserved groups (PropP1–7) containing species specific duplications ^90,131. Why specifically PropP-2 genes cluster near certain cREM-5/5p loci is unclear, though insight from MAGs illuminated a previously unrealized point in Rickettsiales evolution where ProP proliferation occurred (FIG. 7G).

Finally, cREM-5 modules are found in recombination hotspots and other less conserved genomic regions, indicating LGT behind their evolution in Rickettsia genomes. Aside from cREM-5/5p on plasmid pRF, one copy of cREM-5p from the RiClec (Endosymbiont of Cimex lectularius) genome is found on a conjugative transposon termed Rickettsiales Amplified Genetic Elements (RAGE) (FIG. 7F). RAGE are integrative and conjugative elements present on certain Rickettsia plasmids and chromosomes ^90,91,132, as well as proliferated and scattershot in O. tsutsugamushi genomes ^133,134. Cargo genes, or those piggybacking on RAGEs at indiscriminate insertion sites, have functions mostly related to the stringent response and metabolism, defense and resistance, and adaptation to host cells (e.g., ProP genes are shuttled by RAGE). The addition of cREM-5p and SWAMPs, as well as a myriad of TA modules, to the list of RAGE cargo (see FIG. S7F) indicates these mobile elements play a role in disseminating pathogenicity factors that was previously unappreciated.

cREMs with Characterized Function.

Several Rickettsia proteins that lack N-terminal Sec signals have either been well characterized for their roles in subverting host cell processes (e.g., Sca4 and RickA) or possess features that implicate them in targeting host molecules (e.g., VapC and other toxins within TA modules). Until secretory pathways for these molecules are characterized, we consider them cREMs (FIG. 1B). MAG analyses of these proteins have generated novel insight on the structure and evolution of domains targeting the host actin cytoskeleton. Furthermore, a greater appreciation for toxin architecture and distribution indicates the rvh T4SS may still function in congener killing despite the host-dependent lifestyle of the derived Rickettsiales species.

Insight on Rickettsia interactions with host actin.

With their RalF proteins lacking VPRs and their SWAMPs lacking VBDs (Sca4), Tisiphia species may interact with host cell actin cytoskeleton differently than Rickettsia species (FIG. 8A). We analyzed another host actin-associated protein, RickA, which some Rickettsia species use for intracellular motility and possibly intercellular spread ^135–138. While no association with the rvh T4SS has been characterized, RickA localizes to the bacterial surface in the absence of a Sec signal peptide ¹³⁸. RickA directly activates host Arp2/3 complexes through an architecture that mimics host nucleation promoting factors (NPF) ^78,139. Surprisingly, we discovered several RickA proteins from Tisiphia MAGs that differ in their C-terminal architectures relative to SFG rickettsiae RickA proteins characterized in actin polymerization (FIG. 8B; FIG. S8A). While the functional relevance of these differences is unclear, we gained further insight on the N-terminal structure of all RickA proteins. The substantial increase in diversity from MAGs illuminated a large (~95 aa) repeat region enclosing the G actin-binding domain, with each repeat highly conserved in hydrophobicity and predicted structure (FIG. 8B; FIG. S8B). We envisage that this conserved region may facilitate docking of G actin to the G actin-binding domain and overall positioning of the N-terminus to Wiskott-Aldrich syndrome protein homology 2 (WH2) motifs at the C-terminal region.

FIGURE 8. MAG analyses lend insight on Rickettsia interactions with host actin cytoskeleton and rvh T4SS function.

Black boxes provide short names for 29 MAGs from Davison et al. ⁷². (A) MAGs shed light on the evolution of Rickettsia factors behind host actin polymerization and invasion. Tisiphia and BG rickettsiae taxa, as well as SWAMP, SWA-VBD, RickA, and RalF info, is from FIG 2. The passenger domains of R. conorii Sca2 (AAL02648), R. typhi dSca2 (AAU03539), R. bellii Sca2–6 (ABE05361), R. conorii Sca0 (AAL03811), R. typhi Sca3 (AAU03915), R. typhi Sca1 (AAU03504) and R. typhi Sca5 (AAU04158) were used in BlastP searches directly against Tisiphia and BG rickettsiae genomes. Passenger domains and linker sequences were delineated as previously shown ⁸⁸. ReBt, Rickettsia spp. MEAM1, wb, and wq (endosymbionts of Bemisia tabaci). (B) Some Rickettsia genomes encode one or more host actin nucleation proteins. Top: Tisiphia and Rickettsia RickA proteins share a large N-terminal repeat domain but diverge at their C-termini. WH2, Wiskott-Aldrich syndrome protein homology 2 domain. Further details on RickA architecture are provided in FIG. S8. Bottom: Sca2, d-Sca2 and Sca2–6 proteins have a common autotransporter domain (β) but divergent passenger domains. FH2, formin homology 2. Sca2 mimics host formin actin nucleators ¹⁴³ to recruit and polymerize actin for intracellular motility and intercellular spread ^141,142. The functions of dSca2 and Sca2/6 are unknown. (C) The mobile nature of CDI-like/Rhs-like C-terminal toxin/antidote (CRCT/CRCA) modules across diverse rickettsial genomes. Schema shows integration of CRCT/CRCA modules into larger polymorphic toxins (hemagluttinin-like toxins, LysM-like peptidoglycan/chitin-targeting toxins, etc.), as well as CRCT/CRCA modules independent of larger toxins. The toxin warhead for RZI45292 is unknown. Further details are provided in FIG. S9.

Despite activation by RickA, the specific role of host Arp2/3 complexes during rickettsial infection is unclear, perhaps due to different species utilized across studies garnering contrasting results ^137,138,140. Specifically, actin-based motility in certain Rickettsia species is carried out by a second NPF, Sca2, that polymerizes actin independent of Arp2/3 complexes ^141,142. The passenger domains of Sca2 mimic eukaryotic formins by elaborating ring-like structures to elongate actin, with intervening Pro-rich regions and WH2 domains incorporating profilin-actin for elongation and recruiting actin monomers for nucleation, respectively ¹⁴³. Thus, at least for species carrying both RickA and Sca2 (most SFG rickettsiae), RickA-mediated Arp2/3 activation may play a greater role early in infection, possibly in filopodia formation during invasion ^{79,142,144,145}. Still, few Rickettsia species outside of SFG rickettsiae encode Sca2 proteins with intact formin-like passenger domains (FIG. 8B) and some of these species lack RickA genes as well ^90,112,146 (FIG. 8A; FIG. 2). This implies that host actin polymerization for motility is an expendable trait for most Rickettsia species, and that Arp2/3 recruitment and activation during invasion can be instigated by other bacteria-driven processes, i.e. Arf recruitment to the plasma membrane for inducing PI shifts required for filopodia formation ¹⁴⁷.

MAG analyses indicate that, barring acquisitions via LGT, RickA and RalF were likely present before the diversification of major Rickettsia lineages, whereas Sca2 appeared later in Rickettsia evolution (FIG. 8A). Further, we polled MAGs for the presence of genes encoding the four major autotransporters (Sca0, Sca1, Sca3, and Sca5) with known (or anticipated) functions in host cell binding and/or invasion ^148–152. Remarkably, Sca3 is predominant in Tisiphia genomes despite a very limited distribution in Rickettsia species (restricted mostly to TG and TRG rickettsiae ⁸⁸). Further, BG rickettsiae are counter to most other rickettsiae in lacking both Sca0 and Sca5, the dominant proteins of the characterized Rickettsia S-layer ¹⁵³. Collectively, these analyses show that Rickettsia factors described in host cell invasion and actin cytoskeleton subversion are sporadically encoded across genomes, indicating host specificity and/or expendability in their contribution to the intracellular lifestyle.

A repurposed or multi-purposed rvh T4SS?

Aside from secreting effectors that target host cellular processes, evidence is mounting for intracellular bacteria utilizing large contact dependent growth inhibition (CDI) and Recombination hotspot (Rhs) toxins for interbacterial antagonism ^154,155. We recently identified a few rickettsial genomes encoding specialized TA modules that some bacteria integrate into CDI and Rhs toxins to expand toxic activities ^156–159. Widespread in bacteria, these CDI-like/Rhs-like C-terminal toxin and antidote (CRCT/A) modules are extremely polymorphic, variable at the species- and strain-level, and found either associated with larger toxins or alone as small TA modules ¹⁶⁰. The two types of Rickettsia CRCT/A (rCRCT/A) modules we identified, rCRCT/A-1 and rCRCT/A-3a, were once associated with large Rhs-like toxins that have mostly degraded ⁷³. rCRCT/A-1 modules are highly divergent from other characterized CRCT/A modules and predominantly occur in Actinomycetia and Cyanobacteria genomes. Only two Rickettsia species, R. tamurae and R. buchneri, harbor rCRCT/A-1 modules; however, the “Cand. J. acanthamoeba” genome encodes one as an independent module and one integrated into a large modular hemagglutinin toxin with nuclease and peptidase domains. MAG analyses herein discovered several more rCRCT/A-1 modules mostly in Tisiphia genomes associated with pseudogenized hemagglutinin-like toxins (FIG. 1C; FIG. 2).

rCRCT/A-3a modules resemble the prototype CDI TA module (CdiA-CT/CdiI), wherein the nuclease CdiA-CT targets tRNAs in recipient cytosol ¹⁵⁶. CdiA-CT/CdiI is associated with a large modular protein (CdiA) that joins with a second protein (CdiB) as a type Vb secretion system (T5bSS) to deliver the toxin into neighboring bacteria ^158,160,161. However, rCRCT/A-3a modules (and all Rickettsiales genomes) lack CdiB genes. This type of CRCT/A module is widespread in proteobacterial genomes ¹⁵⁸. MAG analyses also revealed more rCRCT/A-3a modules in Rickettsia genomes and a much higher presence of single antidotes (FIG. 1C; FIG. 2), possibly indicating greater selection for defense against toxins versus toxin secretion. Additionally, a rCRCT/A-3a module was found integrated into a large hemagglutinin-like toxin in the Rickettsiales endosymbiont Ac37b, an early-branching Rickettsiaceae species that can co-infect ticks with SFG rickettsiae (FIG. 8B; FIG. S9). Remarkably, this species also carries an identical hemagglutinin-like toxin but with a divergent warhead of the HNH/ENDO VII nuclease superfamily, illustrating the integrative nature of diverse CRCT/A modules. Further, two toxins in the genome of the Rickettsiales endosymbiont of Peranema trichophorum (Midichloriaceae) carry different C-terminal toxins, as well as lysin motifs (LysM) that often occur in cell wall degrading enzymes ¹⁶². Collectively, our analyses illuminate diverse CRCT/A modules in the Rickettsiales mobilome that equip bacteria with weapons for interbacterial antagonism.

Rickettsiales species may also utilize filamentation induced by cAMP (FIC) proteins and type II TA modules for interbacterial antagonism. Some intracellular bacteria harbor FIC domain-containing proteins ¹⁵⁵ and several human pathogens secrete effectors with FIC domains into host cells to subvert cellular processes ¹⁶³. Furthermore, a recent report illustrated that Yersinia pseudotuberculosis utilizes a FIC domain effector, CccR, that alters conspecific gene expression and inhibits congener growth ¹⁶⁴. Many Rickettsia genomes encode multiple divergent FIC proteins (TABLE S2); however, to our knowledge, none of these proteins are known to be secreted by rickettsiae. Similarly, Rickettsia species also harbor a myriad of diverse type II TA modules, with many found on RAGE (e.g., FIG. S7F; FIG. S7F) or plasmids ^91,165. Only one module, VapBC of R. felis, has been characterized. Structural analysis revealed the nature of antidote (VapB) binding to toxin (VapC) ¹⁶⁶, and VapC possesses toxic RNase activity when expressed in bacterial or eukaryotic host cells . We previously showed that Rickettsia genomes encode VapBC and/or a divergent module (VapBC-d) ⁸⁸, and MAG analysis confirmed this observation (FIG. 2). Further, in light of the new genomic diversity, more discrete VapBC loci are encoded in many genomes (data not shown), as well as other type II TA modules (e.g., those encoding ParE, BrnT, and RatA toxins) that have yet been characterized (TABLE S2).

Finally, MAG analyses doubled the number of Rickettsiaceae proteins harboring domains found in RP toxins, particularly those of Wolbachia cytoplasmic incompatibility inducing nucleases (CinB) and deubiquitinases (CidB), as well as the Spiroplasma male killer toxin deubiquitinase (Spaid) ^167–171 (FIG. S10). Many of these toxins are substantially large and modular, encoding numerous domains with uncharacterized effects on host cells ^5,172,173 (e.g., see FIG. 3D). The increasing number of RP toxins (and antidotes when present) in rickettsiae, particularly in species associated with male-killing and parthenogenesis phenotypes, indicates undiscovered molecular mechanisms underpinning these modes of RP. Like the rCRCT/A modules, FIC toxins, and type II TA modules, these RP toxins all lack characterized secretion pathways. Thus, the rvh T4SS cannot be ruled out as a secretion pathway for any of these potential effectors, though a T1SS conserved in all Rickettsiales ⁸⁸ (and possibly other unappreciated routes) could be involved.

Power and Efficacy of MAG Diversity.

The inclusion of diverse MAGs in the assessment of rvh effector evolution has provided several key insights. First, like the rvh T4SS ⁴¹, many REMs and cREMs contain gene duplication. However, unlike the rvh machine, effector duplication seems to define basal lineages (Tisiphia, BG rickettsiae, and other Rickettsiaceae) and tends to lead to retention of only one protein in the derived Rickettsia groups. Still, divergent forms arising from duplication stand to inform on effector function, particularly if derived proteins are utilized for vertebrate cell infection.

Second, the sparse distribution of polled effectors outside of Rickettsia genomes indicates they originated after the divergence of rickettsial families. In some cases, analyses strongly implicate LGT for acquisition of effectors, with a particular bias from Legionellales (e.g., RalF, patatins and PIKs) and other aquatic microbes. This supports the “intracellular arena” hypothesis for the gain of similar effectors in divergent pathogens that occupy common hosts (i.e. protists and arthropods) ¹⁷⁴. It also corroborates our earlier observations that LGT, particular by RAGEs and plasmids, offsets reductive genome evolution in rickettsiae ^26,90. A more recent study reached a similar conclusion via discovery for gene gain shaping Chlamydiae genome architecture, despite the reduced size of most chlamydial genomes ¹⁷⁵. MAGs have also provided a greater appreciation for Legionellales diversity and revealed that the major host-adaptive features (i.e. the dot/icm T4SS and a few conserved effectors) were established in the last common Legionellales ancestor ¹⁷⁶. This is consistent with the discovery by Schön et al. of the rvh T4SS in ancestral Rickettsiales ²⁸; however, we caution that strict rvh repurposing from congener killing to host parasitism is premature until the secretory pathways of the numerous toxins described above are experimentally determined.

Third, MAGs help bridge the gap between research on microbial ecology and human pathogenesis, revealing genome evolutionary and architectural traits that are underappreciated due to biases of clinical isolates or more common environmental strains on public databases ^71,72. Our discovery here of REMs and cREMS on novel plasmids and RAGEs accentuates this point, indicating that such genetic elements may be underestimated for roles in rickettsial biology due to the strong bias of high passage clinical isolates on databases. This is particularly relevant in light of the recent demonstration that the Rickettsia regulator of actin motility (roaM) is often pseudogenized in highly passaged laboratory strains, suggesting serial passage in cell culture can eliminate essential genes lacking environmental selective pressure (in this case the arthropod cytosol) ¹⁷⁷.

Fourth, the most profound insight gained from our work shows how MAG analyses often illuminate novel architectures for well-studied virulence factors. Unearthing new effector designs provides clues on how general foundations are tailored to different hosts and host cell processes. This is epitomized by our discovery of novel RalF-like proteins with SCDs substituted for ARF-interacting domains, which not only fortifies the literature on Legionella and Rickettsia RalF-mediated host ARF recruitment ^{55,81,84,86,87}, but also pinpoints the rise of the actin-targeting VPR regions in Rickettsia RalF proteins after the divergence from Tisiphia species. Combined with numerous other novel effector architectures identified herein, this highlights remarkable recapitulation on mechanisms for mimicking eukaryotic functions that exist beyond Rickettsia and other human pathogens and are widespread in the environment. We assert that widening the comparative genomics lens will allow evolution, which has already matched effector form and function to host environments, to guide experimental designs and reinvigorate pathogen effector research.

Finally, as the landscape of Rickettsia pathogenesis undergoes gradual change due to virulence factor characterization and immunological studies ^1,2, the traditional designation of SFG and TG rickettsiae as the major lineages defining the genus has become grossly outdated. A substantial spike in TRG rickettsiae diversity ^72,178,179, coupled with robust genome-based phylogeny estimations and phylogenomic analyses ^28,72, make the common ancestry of TG and TRG rickettsiae incontrovertible. Prior bias in genome sequences for SFG rickettsiae portrayed TG rickettsiae as unique by smaller genome size and greater pseudogenization relative to all other rickettsiae. However, our focus on rvh effectors across a highly diverse and unbiased genomic sampling shows that all of the major Rickettsia groups (BG, TRG, TG, TIG, and SFG rickettsiae) have distinct evolutionary trajectories of gene gain, loss, and modification (FIG. 2). Thus, MAGs have exposed far greater Rickettsia diversity than previously realized, though long ago conjectured by environmental sampling ¹⁸⁰. These data, as well as careful dissection of the attributes distinguishing Tisiphia and Rickettsia species, will be paramount for deciphering how human pathogens have emerged, possibly multiple times, from this veritable bevy of endosymbiont diversity. Furthermore, an understanding of environmental genomic richness, particular in mobile element diversity, may help forecast the next serious rickettsial diseases to emerge.

CONCLUSION

Discovery and analyses of MAGs has greatly impacted the landscape of Rickettsiology, adding substantial diversity and dispelling the long-held dogma for an ancestral link to the mitochondrial ancestor. Despite predicted extracellular lifestyles, basal rickettsial species carry the rvh T4SS and likely use it as a congener killing machine. Our study coupled a robust evolutionary framework with the inspection of over two dozen known or predicted Rickettsia rvh effectors to provide insight on the origin of mechanisms for host cell subversion and obligate intracellular parasitism. Though focused in taxonomic scope, this experimental design is amenable to probe the origins of virulence factors in any human pathogen with representation in the diverse treasure trove of MAG data. At bare minimum, our work demonstrates that utilizing MAGs in comparative approaches greatly enlightens dialogue on mechanisms of pathogenesis.

MATERIALS AND METHODS

Rickettsiales Phylogeny Estimation.

Robust genome-based phylogeny estimations for Rickettsiales ²⁸ and Rickettsia-Tisiphia ⁷² were used as benchmarks to evaluate our estimated phylogenies based on single or concatenated rvh proteins. We polled the rich MAG diversity on the NCBI database for the presence of vir-like T4SS genes possessing rvh hallmarks ^37,41,44 (i.e., RvhB8, RvhB9 and RvhB4 duplication, multicopy RvhB6, no VirB5 analog; FIG. S1). Provided that many MAGs and certain genome assemblies are likely incomplete, we limited our dataset to assemblies containing both RvhB4-I and RvhB-II, except for a few cases where strong evidence from other rvh genes indicated a Rickettsiales assembly. A total of 153 genome assemblies were retained for further analyses: 1) 93 Rickettsiaceae genome assemblies (including the 28 MAGs from Davison et al. ⁷² and another 15 previously unanalyzed MAGs), 2) 14 and 9 genome assemblies from Anaplasmataceae and Midichloriaceae, respectively, 3) the “Candidatus Deianiraea vastatrix” (Deianiraeaceae) genome assembly, and 4) 33 environmental MAGs likely comprising Deianiraeaceae, Athabascaceae or Mitibacteraceae (nine previously analyzed by Schön et al. ²⁸) (Table S1; FIG. S2).

Only RvhB4-I and RvhB4-II proteins were included in phylogeny estimation, as alignments of other Rvh proteins were extraordinarily variable across the selected taxa (data not shown). RvhB4-I and RvhB4-II proteins were separately aligned using MUSCLE (default parameters). Each alignment included Agrobacterium tumefaciens str. F4 VirB4, which was used as an outgroup to root estimated trees. RvhB4-I and RvhB4-II protein alignments were subsequently concatenated (1974 total positions, “unmasked alignment”). TRIMAL ¹⁸¹ was used to create a second alignment with less conserved positions masked (1613 total positions, “masked alignment”).

For both unmasked and masked alignments, a maximum likelihood-based phylogeny was estimated with PhyML ¹⁸², using the Smart Model Selection ¹⁸³ tool to determine the best substitution matrix and model for rates across aa sites (LG (G+I+F) for both alignments). Branch support was assessed with 1,000 pseudo-replications. Trees were drawn using FigTree (https://github.com/rambaut/figtree/) and manually modified using Adobe Illustrator Ô. Final trees were manually adjusted to place “Candidatus Sneabacter namystus” (which lacks the rvh T4SS) in a position on the phylogram suggested by prior phylogeny estimation ^184,185. All terminal taxa were assigned names based on NCBI database taxonomy (as of Feb. 26^th, 2023), with some “short-names” taken from Davison et al. ⁷² (these are provided in black boxes throughout the figures). Rickettsial classification scheme (STG, Scrub Typhus Group; BG, Bellii Group; TRG, Transitional Group; TG, Typhus Group; TIG, Tamurae-Ixodes Group; SFG, Spotted Fever Group) follows our prior reports ^73,165.

Phylogenomics Analysis.

The RvhB4-based estimated phylogeny was used as a scaffold to complete a distribution matrix for REMs and cREMs. It was not our goal to assess the relative completeness of each MAG included in the matrix, yet to only assess if MAGs and other genome assemblies possessing a rvh T4SSs also include counterparts (homologs or analogs) to Rickettsia REMs and cREMs. REM/cREM assignment is based on prior studies implicating their secretion and/or interaction with RvhD4 (by bacterial 2-hybrid and/or coimmunoprecipitation assays) or presence of motifs known to target either congener bacteria or eukaryotic molecules ^{55–57,73–80}. Analyses of some REMs and cREMs illuminated more complex gene structures (duplications, gene streamlining, and gene fusions) that prompted expansion of the total effector dataset. A total of 26 proteins were analyzed within the final phylogenomic framework (Table S2).

In silico Protein Characterization.

Analyses of each REM and cREM dataset contained discrete workflows tailored to the level of effector conservation in Rickettsiales (and in some cases other bacteria), prior studies that included bioinformatics analyses, and identification of gene duplication, streamlining, or gene fusion. All individual workflows are described in the pertinent figure legends and/or supplemental figure legends. Only general bioinformatics analyses are described below.

Dataset compilation.

Rickettsia REMs and cREMs were used as queries in Blastp searches to compile and analyze diverse proteins harboring significant similarity across the entire lengths of the queries. Analyses utilized our HaloBlast method, which is a combinatorial Blastp-based approach originally designed to interrogate proteins for LGT ²⁶. HaloBlasting compiles Blastp subjects from restricted taxonomic searches that theoretically decrease in similarity by sampling lower levels of bacterial classification. A general search strategy for rickettsiae entails individual Blastp searches against six distinct taxonomic databases: 1) “Rickettsia” (NCBI taxid 780)”; 2) “Rickettsiales” (taxid: 766) excluding “Rickettsia”; 3) “Alphaproteobacteria” (taxid: 28211) excluding “Rickettsiales”; 4) “Proteobacteria” (taxid: 1224) excluding “Alphaproteobacteria”; 5) “Bacteria” (taxid: 2) excluding Proteobacteria”; and 6) “minus bacteria”. Data subsets were constructed strictly using NCBI taxid and following the NCBI taxid hierarchy to identify “daughter” taxonomic groups. Typically, 500 subjects (if available) are retained per search. All subjects from each search were separately ranked by Sm score (= b * I * Q, where b is the bitscore of the match, I is the percent identity, and Q is the percent length of the query that aligned), a comparative sequence similarity score designed to de-emphasize highly significant matches to short stretches of the query in favor of longer stretches of similarity ²⁶. The “halos” or separate database searches are then compared to one another to determine the taxon with the strongest similarity to the query sequences. These analyses usually make LGT apparent when divergent datasets contain top-ranking proteins more similar to the Rickettsia queries than more closely related datasets.

Protein characterization.

Select proteins or domains (again, context-dependent) are typically compiled and aligned with MUSCLE using default parameters ¹⁸⁶. To identify conserved regions, alignments are then visualized as sequence logos using WebLogo ¹⁸⁷. Domain analyses are performed by cross-checking predictions from the NCBI Conserved Domains Database (CDD) and EMBL’s Simple Modular Architecture Research Tool (SMART) ¹⁸⁸. In some cases, proteins were evaluated for N-terminal signal peptides ¹⁸⁹ and transmembrane-spanning regions ¹⁹⁰. Alignments shown in the figures and supplemental figures are manually assessed for conservation, typically considering 80% of a position conserved (alignment size-dependent), with amino acid coloring scheme and assignment as follows: black, hydrophobic (Ala, Val, Iso, Leu, Pro, Met, Gly); gray, less hydrophobic (can include a minority of Try, Phe, Tyr); red, negatively charged (Glu, Asp); green, hydrophilic (Cys, Asn, Gln, Ser, Thr); purple, aromatic (Try, Phe, Tyr); blue, positively charged (His, Lys, Arg). Individual protein schemas were generated using Illustrator of Biological Sequences ¹⁹¹ with manual adjustment in Adobe Illustrator ^™.

Protein structures were predicted using the Protein Homology/analogY Recognition Engine V 2.0 (Phyre2) ¹⁹²; in some cases, published structures were retrieved from the Protein Data Bank ¹⁹³ and used in one-to-one threading mode with Phyre2. For some effectors, we also evaluated structures generated with Alphafold ^194,195. Finally, some regions of proteins were analyzed for predicted secondary structure using JPred ¹⁹⁶.

Phylogenies were estimated for some REM and cREM datasets, which were compiled uniquely for each case and utilized HaloBlast to obtain non-rickettsial taxa (if available). Alignments were not masked since masking eliminated too many informative positions. Maximum likelihood-based phylogenies were estimated with PhyML ¹⁸², using the Smart Model Selection ¹⁸³ tool to determine the best substitution matrix and model for rates across aa sites. Branch support was assessed with 1,000 pseudo-replications. Trees were visualized and drawn as described above.