Structural and functional study of Legionella pneumophila effector RavA

Abstract

Legionella pneumophila is an intracellular pathogen that causes Legionnaire's disease in humans. This bacterium can be found in freshwater environments as a free‐living organism, but it is also an intracellular parasite of protozoa. Human infection occurs when inhaled aerosolized pathogen comes into contact with the alveolar mucosa and replicates in alveolar macrophages. Legionella enters the host cell by phagocytosis and redirects the Legionella‐containing phagosomes from the phagocytic maturation pathway. These nascent phagosomes fuse with ER‐derived secretory vesicles and membranes forming the Legionella‐containing vacuole. Legionella subverts many host cellular processes by secreting over 300 effector proteins into the host cell via the Dot/Icm type IV secretion system. The cellular function for many Dot/Icm effectors is still unknown. Here, we present a structural and functional study of L. pneumophila effector RavA (Lpg0008). Structural analysis revealed that the RavA consists of four ~85 residue long α‐helical domains with similar folds, which show only a low level of structural similarity to other protein domains. The ~90 residues long C‐terminal segment is predicted to be natively unfolded. We show that during L. pneumophila infection of human cells, RavA localizes to the Golgi apparatus and to the plasma membrane. The same localization is observed when RavA is expressed in human cells. The localization signal resides within the C‐terminal sequence C⁴⁰⁹WTSFCGLF⁴¹⁷. Yeast‐two‐hybrid screen using RavA as bait identified RAB11A as a potential binding partner. RavA is present in L. pneumophila strains but only distant homologs are found in other Legionella species, where the number of repeats varies.

Short abstract

PDB Code(s): 6WO6;

1. INTRODUCTION

Legionella pneumophila is an intracellular pathogen, which in the natural environment replicates within amoebae or other protozoa, that are a group of vastly diverse unicellular eukaryotes. ¹ , ² Legionella gains access to the human lung through the inhalation of bacteria‐containing aerosols whereupon bacteria infect alveolar macrophages and establish an intracellular replicative niche. The spread of the infection leads to a severe pneumonia known as Legionnaires' disease. ³ L. pneumophila invades the host macrophages by phagocytosis. Instead of entering the phagocytic maturation pathway, nascent phagosomes fuse with ER‐derived secretory vesicles (ESVs); this results in the formation of a Legionella‐containing vacuole (LCV) wherein L. pneumophila replicates intracellularly. ⁴ In order to create a replicative niche, L. pneumophila subverts host cell processes through the action of more than 300 effector proteins that are secreted into the host cell via the Dot/Icm type IVB secretion system (T4SS). ⁵ The mechanism of establishing the LCV appears to be similar across eukaryotic hosts. ⁶ Rab GTPases serve as regulators of intracellular membrane trafficking ⁷ and several Legionella effectors have been shown to target Rabs. ⁸ The ability to modulate Rab GTPase activity garners L. pneumophila control over endocytic/secretory trafficking and protects the LCV from acidification and fusion with lysosomes.

The biochemical/cellular functions are known for only a fraction of more than 300 Legionella effectors, and for even fewer their host protein targets/interactors are known. ⁹ The difficulty in deciphering the function effectors in pathogenic bacteria is the redundancy of their actions, leading to none or to mild phenotypes upon deletion of a single effector gene. ¹⁰ This difficulty is substantially more pronounced in Legionella due to the sheer number of its effectors. Therefore, multiple approaches and methodologies are applied to decipher the effector functions. Here, we selected to investigate one of the L. pneumophila effectors with an unknown function, RavA (region allowing vacuole colocalization A, lpg0008). RavA is a 419 residues long protein that was first identified as a potential substrate of the Dot/Icm secretion system in a high‐throughput screen. ¹¹ Subsequently, a mass spectrometry query of proteins that are present within the purified LCVs from RAW264.7 macrophages and Dictyostelium discoideum identified RavA as one of them. ¹² Searching for clues to its cellular function, we applied protein crystallography to determine the crystal structure of RavA (without the C‐terminal ~90 residues) and discovered that it contains four structurally similar domains, which might function as protein–protein binding scaffolds. Moreover, we confirmed that RavA is secreted during an L. pneumophila infection and that it localizes to the Golgi apparatus. Next, we showed that the Golgi localization is directed by the C‐terminal segment of RavA, specifically by the 406–418 residues that contain the CX₃FCg/sGF motif. To identify potential host cell proteins interacting with RavA we employed a yeast‐two‐hybrid screen as well as mass spectrometry and found RAB11A to be a potential RavA‐interacting partner, co‐localizing with RavA to the Golgi.

2. RESULTS

2.1. RavA is secreted to the host cell during infection and localizes to the Golgi compartment

To establish that RavA is secreted during Legionella infection, HEK293T cells stably expressing Fcγ‐receptor (HEK‐FcγR) ¹³ were infected by the L. pneumophila strain Philadelphia‐1 harboring 4xHA‐tagged RavA. RavA was found inside the infected cells and was distributed in a punctate pattern. The cis‐Golgi marker GRASP65 was also observed in these punctate structures at 4 hr, 8 hr, and 24 hr after infection, co‐localizing with RavA (Figure 1a). While localization of 4xHA‐RavA was strongly concentrated at the Golgi at 4 hr, more dispersed staining was observed at 8 hr and 24 hr after infection (Figure 1a). This phenomenon might be due to either overexpression or be a genuine effect caused by the function of the protein.

FIGURE 1.

RavA localizes to Golgi compartment in mammalian cells. (a) HEK‐FcγR cells were infected by L. pneumophila strain Philadelphia‐1 harboring either the empty pICC562 vector or this vector containing p4HA‐RavA. Cells were fixed and permeabilized, followed by a one‐hour treatment with primary antibody cocktail containing mouse‐anti‐HA antibody and rabbit‐anti‐GRASP65 antibody. After washes, the cells were incubated with the secondary antibody mix consisting of Alexa Fluor 488‐anti mouse (green) and Alexa Fluor 568‐anti rabbit (red) for 30 min. Cell nuclei were stained with Hoechst and imaged by confocal microscope. Scale bar, 5 μm; (b) HEK293T (left panel) or HeLa (right panel) cells were transfected with plasmids encoding RavA‐GFP or GFP alone (green), fixed with 4% (wt/vol) of paraformaldehyde (PFA) 24 hr post‐transfection. Cell nuclei were stained with DAPI; (c) HEK293T that transfected with GFP or RavA‐GFP were incubated with 5 μM BODIPY™ TR‐ceramide in HBSS buffer at 4°C for 30 min. After wash, cells were incubated in a fresh medium at 37°C for 30 min and imaged live (left panel), or incubated with anti‐GM130 antibody, followed by Alexa Fluor 546 (red) secondary antibody (right panel); (d) HEK293T cells were transfected with RavA‐GFP or GFP alone for 24 hr. The cells were then incubated with Brefeldin A (BFA, 5 μg/ml) or DMSO (as a control) for 30 min, stained with 5 μM BODIPY™ TR‐ceramide and imaged by confocal microscopy. GFP, left panel; RavA‐GFP, right panel. Scale bars, 5 μm

2.2. RavA localizes to Golgi when ectopically expressed in cells

We then expressed RavA in mammalian cells to confirm the subcellular localization observed during Legionella infection. A full‐length RavA was fused at the C‐terminus with the green fluorescent protein (RavA‐GFP) and this fusion protein was expressed in HEK293T and HeLa cells. While GFP alone was evenly distributed in the cytosol, RavA‐GFP was enriched at the plasma membrane and noticeably concentrated at subcellular structures in the perinuclear region (Figure 1b). To confirm that this localization was not an artifact of the placement of GFP, we also tested the localization of RavA tagged with 3xHA at the N‐terminus. The same distribution pattern was observed for 3xHA‐RavA (Figure S1). To identify the relevant subcellular compartment, we stained HEK293T cells expressing RavA‐GFP with the Golgi specific BODIPY™ TR ceramide in living cells or with anti‐GM130 antibody, a cis‐Golgi marker, in fixed cells. In both cases we observed co‐localization of RavA‐GFP with the Golgi apparatus (Figure 1c: left panel—BODIPY™ TR ceramide, right panel—anti‐GM130) but no co‐localization was apparent for GFP alone. To further confirm Golgi localization of RavA, HEK293T cells were transfected with either GFP or RavA‐GFP, and the structure of the Golgi apparatus was disrupted by treating the cells with brefeldin A (BFA) or DMSO (as a control). Indeed, after treating with BFA the juxtanuclear pattern was lost when stained for Golgi with BODIPY™ TR ceramide (Figure 1d). In BFA‐treated cells, RavA‐GFP was diffuse within the cytoplasm and no longer formed a punctate pattern (Figure 1d right panel), while localization of GFP was not affected by BFA treatment (Figure 1d, left panel). Taken together, these data suggest that RavA is localized to the Golgi apparatus during infection or when expressed ectopically by transfection.

2.3. Crystal structure of RavA

A three‐dimensional structure of an unknown protein can provide clues about its function from structural similarity to proteins with established activities or functions. For example, several effectors that contained cysteine protease‐like domains turned out to be either proteases (e.g., Salmonella Typhimurium GtgE ¹⁴ , ¹⁵ ) or deubiquitinases (e.g., Xanthomonas XopD and Rickettsia RickCE ¹⁶ ). Therefore, we determined the crystal structure of RavA to obtain clues about its function. The full‐length RavA protein (2–419) did not crystallize in our hands. This prompted us to examine the secondary structure and disorder predictions for RavA, which indicated that the C‐terminal ~90 residues segment has characteristics of an unstructured polypeptide (Figure S2). Hence, we subjected RavA to an overnight limited proteolysis with a set of proteases to identify a stable fragment devoid of flexible regions that might hinder crystal growth. The elastase‐treated sample resulted in a stable fragment represented by a band on SDS‐PAGE gel with an approximate molecular weight of 40 kDa (Figure S3). This sample was then analyzed by mass spectrometry whereby the accurate measurement of its molecular weight allowed us to identify the cleavage site as located between residues Gln329 and Ser330. We then cloned, expressed and purified the construct that included residues 2–329 of the wild type RavA. The produced protein yielded crystals, but they diffracted poorly. In an attempt to improve crystal quality we analyzed protein sequence for possible entropy reducing mutations using the SERp server. ¹⁷ One set of predicted mutations (E38A/K39A), when introduced to the previous construct, led to substantially better crystals than the ones obtained from the initial construct. These new crystals diffracted to 2.8 Å resolution and allowed us to determine the structure of RavA.

The structure was solved by using the Single‐wavelength Anomalous Diffraction (SAD) method from SeMet substituted protein crystals. These crystals contain two RavA molecules in the asymmetric unit. The two molecules are very similar and superimpose with a root‐mean‐square deviation (rmsd) of 0.9 Å. The main difference between the molecules is the conformation of the 64–74 loop interconnecting two α‐helices, and a slight difference in the bend of the elongated molecules. We attribute these conformational differences to differences in their crystal contacts. Overall, the RavA(2–329) molecule is shaped as a highly elongated and slightly bent tube with a length of ~110 Å and a diameter of ~32 Å (Figure 2a). Each molecule is made of four α‐helical domains (D1‐4), approximately ~85 residues in length, which are arranged like tightly packed beads on a string. Each of the four domains also displays the same basic fold consisting of two helical layers: the first, N‐terminal layer contains four short α‐helices arranged into a rectangular shape, with the second layer made of two antiparallel α‐helices (Figure 2b). The C‐terminal, longer α‐helix of the preceding domain packs on top of the N‐terminal four‐helix layer of the following domain, providing a tight interdomain connection and imposing rigidity on domain arrangement within the entire elongated structure. Domains D2 and D3 are the most similar and contain all elements of the fold. Domain D4 is missing the last helix, likely resulting from the construct design. Indeed, the secondary structure analysis predicted the residues 332–340 to form an α‐helix in the full‐length RavA (Figure S2). Domain D1 diverges the most from the other domains. It contains an additional α‐helix at the N‐terminus, the helices in the first layer forming a tighter circular arrangement while the loop connecting the last two helices is longer than in the other domains. The orientation of these two helices is somewhat different than in the other domains (Figure 2c). Indeed, while domains D2, D3 and D4 share a significant degree of sequence similarity ranging between 30% and 40%, the sequence of the D1 domain is more divergent from the other domains. The alignment of the sequences of all four domains based on their structural superposition (Figure 2d) shows that only two residues are strictly conserved in all four domains. However, seven residues are conserved across domains D2‐D3‐D4 and eight others constitute conservative replacements. Eleven of these residues are hydrophobic and contribute to the mini‐core between the two helical layers or contact the neighboring domains (Figure 2b), assuring tight packing of the domains. Despite a more divergent sequence of domain D1, a similar hydrophobic mini core is present in this domain as well.

FIGURE 2.

Views of RavA structure. (a) RavA(2–329) displayed with a semitransparent molecular surface and a cartoon representation of the backbone. The structure is highly elongated and slightly curved. It consists of four structural repeats colored from N‐ to C‐direction in blue (repeat D1), cyan, green and red (repeat D4); (b) View of D3 rainbow‐colored from blue at the N‐terminus to red at the C‐terminus. Helices α1‐α4 are arranged in a semi‐rectangle and form the first layer, helices α5‐α6 form the second layer. The residues highly conserved among all domains are shown explicitly in a stick mode and colored by atom type. They contribute to a hydrophobic mini core. Phe201 and Leu204 participate in contacts to the C‐terminal helix (pink) of the preceding repeat (D2). The C‐terminus is marked with letter C; (c) Superposition of all four repeats shown in the same colors as in panel A. Repeats 2 and 3 superimpose very well, repeat 4 lacks the C‐terminal domain while N‐terminal half of repeat 1 is somewhat different from the other three repeats. Letters N and C mark the N‐ and C‐termini; (d) Structure‐based sequence alignment of RavA repeats. The highly conserved residues are boxed. Star (*) denotes conserved residues while dot (.) marks conservative substitutions. Letters N and C mark the N‐ and C‐termini; (e) The superposition of D3 (cyan—similar segment, blue—dissimilar segment) and the C‐terminal domain of REV1 (PDB code 3VU7, orange—similar segment, red—dissimilar segment); (f) The superposition of D3 (cyan—similar segment, blue—dissimilar segment) and the C‐terminal domain of RANGAP1 (PDB code 5D2M, orange—similar segment, red—dissimilar segment). Letters N and C mark the N‐ and C‐termini

2.4. RavA repeats show limited structural similarity to other proteins

To elucidate potential functional clues, we searched the Protein Data Bank (www.rcsb.org/pdb/) using the DALI server ¹⁸ for proteins containing domains that display structural similarity to the entire RavA(2–239). That search identified no significant structural homologs. We then performed searches with each individual RavA domain, D1 to D4 and identified distant structural homologs (Z‐scores ~6 or lower) to part of the domain. Not surprisingly, the similarity extended over several, but not all helices. In the structural superposition with the top hits, there was a relatively good overlap with the three C‐terminal helices of each RavA domain, but the first three N‐terminal helices of the domain did not have a good match. Moreover, the top matches for individual searches with each RavA domain were not the same, indicating that the N‐terminal layer of RavA domain fold is divergent from domains present in the Protein Data Bank. Nevertheless, we found distant structural homologs common to all domains and corresponding to small proteins of ~100 to 150 residues. The most similar protein to RavA domains was the C‐terminal domain of an error‐prone DNA polymerase REV1, pivotal in trans‐lesion DNA synthesis (PDB code 3VU7 with Z‐score ~5 for all domains) and the C‐terminal domain of the Ran GTPase‐activating enzyme RANGAP1 (PDB code 5D2M with a Z‐score between 5.9 for D1 [most similar] and 4.3–5.0 for D2 and D3). The C‐terminal domain of REV1 binds to the mitotic spindle assembly checkpoint protein MAD2B/REV7, ¹⁹ while the C‐terminal domain of Ran binds to the SUMO‐conjugating enzyme UBC9. ²⁰ REV1 and Ran interact with their binders through the segments of their structures that are the most similar to RavA domains. The superposition of RavA domain with these two proteins is shown in Figure 2e.

The surface of RavA does not have a deep cavity lined with conserved residues that would indicate an active site. As well, we could discern no arrangements of residues similar to known active sites. Taking these observations into account, we predict that RavA domains are protein–protein binding modules and that RavA function is to serve as a protein binding hub to sequester other proteins to the Golgi apparatus.

2.5. Golgi localization of RavA is determined by its C‐terminal segment

We have established that RavA localized to the Golgi apparatus during infection as well, when ectopically expressed. Therefore, we wanted to identify which section of RavA is responsible for this distinct localization. In silico analysis of the amino acid sequence of RavA revealed no transmembrane domain(s) or lipidation sequences. To interrogate if the localization‐responsible segment is contained within the four repeat domains, we generated a construct containing RavA(1–342) C‐terminally tagged with GFP and expressed it ectopically in HEK293T cells. The RavA(1–342)‐GFP showed diffused localization throughout the cell (Figure 3a) indicating that the Golgi localization signal was not present in this segment of RavA. Next, we tested the localization of an ectopically expressed construct containing the C‐terminal segment of RavA, absent in the crystallized protein, RavA(342–419)‐GFP. This protein showed a very similar cellular distribution pattern to the wild‐type protein and localized to the Golgi apparatus (Figure 3), indicating that this region was responsible for the localization of RavA (Figure 3). To confirm further that that RavA targets RavA to the Golgi apparatus, we fused RavA(342–419) to the C‐terminus of another L. pneumophila protein, LpiR1 (Lpg0634) which remain within the host cytosol. ²¹ The presence of the C‐terminal domain of RavA in the LpiR1‐RavA(342–419) fusion causes a punctate distribution when the fusion is expressed in the HEK293T cells, similar to that of the intact RavA (Figure 3b).

FIGURE 3.

Localization of RavA is determined by the C‐terminal region. (a) GFP‐tagged truncated constructs of RavA were transfected into HEK293T cells, cells were then fixed and imaged by confocal microscopy; (b) LpiR1 (Lpg0634) localizes to the cytosol (top row). Fusion of LpiR1with the RavA(342–419) C‐terminal domain has a punctate distribution, similar to that of the wild‐type RavA. Scale bars, 5 μm

To pinpoint the sequence directing Golgi localization we performed bioinformatic analysis of the C‐terminal region by the PSIPRED server. ²² No secondary structures were predicted within residues 341–400 and a low probability of an α‐helix was indicated for residues 401–415 (Figure S2). We generated six new RavA‐GFP‐tagged constructs with small deletions within the C‐terminal region, namely Δ351‐354, Δ361‐365, Δ381‐388, Δ401‐405, Δ406‐410 and Δ414‐418 (Figure 4a). Of these six deletion mutants, four were still localized to the Golgi in a manner similar to the wild‐type RavA while two, RavA(Δ406‐410)‐GFP and RavA(Δ414‐418)‐GFP, showed cytosolic distribution (Figure 4b). These results suggested that residues within the 406–418 region, within the weakly predicted C‐terminal α‐helix, were essential for localization of RavA to the Golgi apparatus, whereas the unstructured region (342–400) was not essential for this localization. The sequence E⁴⁰¹EEEE⁴⁰⁵ forms the so‐called E‐block motif. The E‐block motif, usually 5–9 residues long, is a signal sequence within some effectors recognized by component(s) of the Dot/Icm secretion system, which directs the effector for translocation into the host cell. ¹¹ The Golgi localization of the RavA(Δ401‐405)‐GFP indicated that the E‐block did not contribute to RavA localization. We further confirmed this by creating a mutant in which these five Glu residues were mutated to Ala and showing that this mutant still localized to the Golgi.

FIGURE 4.

Effect of mutations on RavA localization. (a) Schematic representation of RavA truncations that were fused at the C‐terminus to GFP. (b) Cellular localization of these RavA‐GFP constructs in HEK293T cells. Scale bars, 5 μm; (c) Localization of RavA mutants with single amino acid replacement to an alanine

To identify which residues within the 406–410 and 414–418 regions are essential for localization, we individually substituted each of these residues with an alanine (Figure 4c). None of these mutations fully disrupted the localization of RavA to Golgi, however, C409A, C414A, L416A and F417A mutations resulted in an increased cytosolic localization of this protein (Figure 4c). These residues are located on one side of the predicted helix. Our data suggest that multiple residues within the 406–418 C‐terminal region of RavA are involved in recognition and binding to the Golgi.

2.6. RavA repeats are found in other Legionella species

DELTA‐BLAST search ²³ for sequence similarity to the RavA identified homologs in many Legionella species and in a few strains of Fluoribacter. No homologs were found in other species represented in the NCBI database. RavA orthologs in L. pneumophila species showed high sequence identity to RavA from strains Paris or Philadelphia (>93%) and contain all four domains as well as the C‐terminal region. Other Legionella and Fluoribacter species had more distant homologs that variously contained between one and four of the repeat domains that are present in RavA. The RavA C‐terminal extension, residues 330–419, is present and highly conserved in L. pneumophila. Distant homologs of this C‐terminal segment were found in other Legionella species with sequence identity below ~30%. However, it is interesting to note that they all share the C‐terminal sequence C(X)₃FCg/sLF, which is crucial for the Golgi localization in RavA.

2.7. Yeast‐two‐hybrid assay identified RAB11A as a potential interactor of RavA

To identify potential protein interactors of RavA, we performed a yeast‐two‐hybrid (Y2H) protein interaction screen using the universal human cDNA library. Several potential interacting proteins were identified (Table 1), including domains from RAB11A, ornithine decarboxylase antizyme 3, CUB and Sushi multiple domains 1 variant, Ubiquitin carboxyl‐terminal hydrolase MINDY‐2 and gamma‐butyrobetaine,2‐oxoglutarate dioxygenase 1.

TABLE 1.

Fragments of interaction partners for RavA identified using yeast two‐hybrid approach with cDNA human library

Prey no.	Interaction partners	NCBI protein accession no	aa sequence
1	Ornithine decarboxylase antizyme 3	NP_057262	152–235
2	CUB and Sushi multiple domains 1 variant	BAD92739	2,523–2,818
3	Ubiquitin carboxyl‐terminal hydrolase MINDY‐2	BAA86478	307–583
4	Gamma‐butyrobetaine,2‐oxoglutarate dioxygenase 1	NP_003977	1–255
5	RAB11A	NP_004654	118–216

RAB11A was the most interesting target since Rab GTPase family members are known interaction partners of several Legionella effectors. ²⁴ In order to determine which domains of RavA and RAB11A were essential for the interaction, full length and truncated versions of both RavA and RAB11A were expressed in pGBKT7 and pGADT7‐RecAB vectors, respectively (Clontech‐Takara, https://www.takarabio.com/). In addition to pGBKT7 encoding full length RavA and pGADT7‐RAB11A_118–216, the following constructs were tested: pGBKT7‐RavA_1–341; pGBKT7‐RavA_342–418 and pGADT7‐RAB11A_full‐length, pGADT7‐RAB11A_61–216; pGADT7‐RAB11A_147–216. The resulting plasmids based on pGBKT7 and pGADT7 backbones were transformed into PJ69‐4A and Y187 yeast strains respectively, yielding series of PJ69‐4A/pGBKT7‐bait and Y187/pGADT7‐prey strain series (Table 2). Each of the strains of PJ69‐4A/pGBKT7‐bait series was mated with every strain of Y187/pGADT7‐prey series. The diploids were assayed for histidine and adenine auxotrophy, as well as α‐galactosidase activation. Only full length RavA, but not the N‐terminal domain or the C‐terminal segments, supported yeast growth in the presence of full length or any N‐terminally truncated RAB11A constructs (i.e., 61–216, 118–216 and 147–216). This was confirmed by growth on the selective plates, as well as α‐galactosidase activity on the plates with a X‐α‐Gal substrate (Figure 5).

TABLE 2.

Growth on the selective plates and α‐galactosidase activity on the plates with X‐α‐Gal substrate of the diploids carrying respective bait and prey fragments

Bait/prey carrying strain	pGBKT7‐RavA_fl	pGBKT7‐RavA_1‐341	pGBKT7‐RavA_342‐418
pGADT7‐RAB11A full‐length	+	−	−
pGADT7‐RAB11A_61‐216	+	−	−
pGADT7‐RAB11A_118‐216	+	−	−
pGADT7‐RAB11A_147‐216	+	−	−

FIGURE 5.

Yeast two‐hybrid assay of full length RavA, its N‐ and C‐fragments with full length or truncated peptides of RAB11A. Diploid yeast strains were obtained after the mating of the respective haploid strains, carrying respective plasmid either from pGBKT7 or pGADT7 series, listed in Table 2. The cultures were spotted on the selective plates (SD medium without tryptophan, leucine, histidine, and adenine). Only diploids carrying pGBKT7‐RavA_full_length plasmid and the plasmid pGADT7 with any of RAB11A polypeptides, exhibited growth

2.8. RavA co‐localizes with RAB11A

Next, we questioned whether RAB11A co‐localized with RavA inside mammalian cells. For this purpose, we co‐transfected HEK293T cells with the plasmid encoding either RavA‐GFP or RavA(1–342)‐GFP together with plasmid encoding 3xHA‐RAB11A and followed their localization by immunofluorescence staining. Full length RavA co‐localized with RAB11A within punctate structures, however, no co‐localization was observed for the RavA(1–342)‐GFP construct (Figure 6a). To quantify the degree of co‐localization of RavA and RAB11A, we calculated the Pearson correlation coefficient for each construct, which was 0.72 ± 0.04 for full length RavA and RAB11A, and 0.32 ± 0.03 for RavA(1–342) and RAB11A (Figure 6b). Based on the Y2H results and the co‐localization of RavA and RAB11A, we considered that RavA may be a guanine nucleotide exchange factor (GEF) for RAB11A that activates the latter by promoting the exchange of GDP for GTP. However, no GEF activity for RavA was detected in an assay following the protocol described in ²⁵ (data not shown).

FIGURE 6.

RavA co‐localizes with RAB11A. (a) Plasmids encoding RavA‐GFP or RavA(1–342)‐GPF and 3xHA‐RAB11A were co‐transfected into HEK293T cells, cells were fixed and permeabilized, then incubated with anti‐HA antibody, followed by Alexa Fluor 546 (red) secondary antibody. Then cells were imaged by confocal microscopy. Scale bars, 5 μm; (b) Quantitation of colocalization by Pearson coefficient. The Pearson correlation coefficient was determined with coloc 2 plugin of the ImageJ software and was calculated with at least 30 cells (mean and standard deviation, SD, of the mean from each experiment, ***p < .001)

2.9. Potential RavA interactors suggested by mass spectrometry

To further assess RavA roles in human cells, RavA fused with the 3xFlag tag was overexpressed in HEK293 cells and subjected to co‐immunoprecipitation followed by mass spectrometry (MS)‐based protein identification. Cells overexpressing 3xFlag were used as controls. Two or more unique peptides were recovered for 302 human proteins (Table S1). In addition to the 302 human proteins, as expected, RavA was recovered only in 3xFlag‐RavA samples. Of the 419 RavA amino acids, 384 were sequenced by MS (ca. 92% coverage). Two phosphorylation sites were identified with localization site p‐values < .01 per ptmRS (Thermo Fisher Scientific). One was at Ser400 with peptide p‐values ≤ .01, another at Ser368 with peptide probabilities ≤ .05. The same Ser368 and Ser400 were also recovered without modifications, which indicates that their phosphorylation is only partial and might modulate RavA activity.

Of the 302 human proteins, 301 were mapped to Gene Ontology (GO) categories. A majority of the identified proteins were intracellular (300 proteins, p‐value 2.34 × 10⁻⁴⁵, False Discovery Rate (FDR) 1.17 × 10⁻⁴⁸ (Table S2). Observed localization of RavA to Golgi and membrane is consistent with the over‐representation of specific GO Cellular Component categories (Table S2), including “membrane” (187 proteins, p‐value 3.57 × 10⁻⁷, FDR 1.14 × 10⁻⁵), “plasma membrane region” (30 proteins, p‐value 4.37 × 10⁻³, FDR 4.64 × 10⁻²), “clathrin coat of trans‐Golgi network vesicle” (3 proteins, p‐value 1.67 × 10⁻³, FDR 2.13 × 10⁻²), and “clathrin‐coated vesicle membrane” (7 proteins, p‐value 1.75 × 10⁻³, FDR 2.18 × 10⁻²). Clathrin participates in many clathrin‐mediated membrane trafficking processes, including endocytosis from the plasma membrane and budding from the Golgi. ²⁶ We also found the “myosin II filament” (2 proteins, p‐value 1.96 × 10⁻³, FDR 2.39 × 10⁻²) and the “myosin complex” (8 proteins, p‐value 4.30 × 10⁻⁶, FDR 1.09 × 10⁻⁴) enriched among proteins recovered with 3xFlag‐RavA. This is consistent with a non‐muscle myosin II association with the Golgi membranes. ²⁷ The localization of RavA is likely on the cytoplasmic side of the membranes as GO Cellular Component “cytosol” category is overrepresented (151 proteins, p‐value 2.01 × 10⁻²⁰, FDR 2.01 × 10⁻¹⁸).

Corresponding to these GO Cellular Component overrepresentations, a number of relevant GO Biological Process categories are overrepresented (Table S3). These include, for example, “transport along microtubule” (9 proteins, p‐value 6.69 × 10⁻⁴, FDR 2.40 × 10⁻²), “protein targeting to membrane” (10 proteins, p‐value 3.77 × 10⁻⁴, FDR 1.48 × 10⁻²), “establishment of protein localization to membrane” (17 proteins, p‐value 2.25 × 10⁻⁶, FDR 1.58 × 10⁻⁴), “organelle localization” (30 proteins, p‐value 9.66 × 10⁻⁹, FDR 9.29 × 10⁻⁷), and “multi‐organism transport” (20 proteins, p‐value 7.67 × 10⁻²⁰, FDR 1.59 × 10⁻¹⁷).

Proteins, identified by Y2H, were not recovered by MS. This is likely due to the high level of RavA overexpression and relatively low levels of potentially interacting proteins. Indeed, in a global study of proteins expressed in the HEK293T cell line, ²⁸ ornithine decarboxylase antizyme 3, CUB and Sushi multiple domains 1 variant, and gamma‐butyrobetaine,2‐oxoglutarate dioxygenase 1 were not detected, while RAB11A and Ubiquitin carboxyl‐terminal hydrolase MINDY‐2 were detected only at 150 and 81.9 ppm, respectively. ²⁹ The transient nature of interactions with RAB11A and Ubiquitin carboxyl‐terminal hydrolase MINDY‐2 may have also contributed to the failure to detect these proteins by MS. To determine if the list of proteins identified by MS was enriched for interactors of potential RavA partners detected in the yeast‐two‐hybrid experiments, we extracted the known RAB11A interactors from the IntAct database (version 4.2.14). ³⁰ RAB11A has a large number of known interacting partners (166 in IntAct). Among the proteins associating with RavA we recovered by mass spectrometry, RAB11A's partners were significantly overrepresented (chi square test p‐value < 1 × 10⁻²⁴), further supporting the link identified by Y2H between RavA and RAB11A.

3. DISCUSSION

RavA is one of ~350 proteins secreted by Legionella pneumophila, but its cellular function is unknown. RavA has previously been identified by mass spectrometry on purified LCVs from RAW264.7 macrophages and Dictyostelium discoideum. ¹² We showed here that RavA is indeed secreted during Legionella infection and localizes in human cells to the Golgi apparatus and also to the plasma membrane. A similar localization was observed for ectopically expressed RavA. To obtain functional clues, we determined the crystal structure of RavA(2–329), which excluded the ~90 C‐terminal residues predicted to be disordered by bioinformatic analysis. Structural analysis showed that RavA(2–329) contains four ~85 residues long domains that share the same fold. The domains are rigidly connected through a long C‐terminal helix of one domain binding to the N‐terminal helical layer of the following domain, resulting in an elongated shape of the molecule. The absence of a deep depression on the surface and no apparent active site configuration that we could discern, strongly suggested a protein interaction function. This functional clue was strengthened by identifying partial fold similarity to the domains of two other proteins that mediated protein–protein interactions. The highly elongated tubular shape of RavA with a slightly concave surface along one side of the tube suggests that the anchoring of RavA at the Golgi by its C‐terminal domain is the first step in association with a target host cell molecule.

We showed that Golgi localization is fully determined by the C‐terminal segment of RavA and in particular by the region 406–418. Individual alanine mutations within this region showed that only Cys409, Cys414, Leu416, and Phe417 disrupted Golgi localization and lead to a broad distribution within the cytosol. The residues C⁴⁰⁹WTSFCGLF⁴¹⁷ are fully conserved in all known L. pneumophila RavA sequences present in the NCBI database. Several other Legionella species contain more distant homologs of RavA with the C‐terminal region sharing between 21 and 31% sequence identity. In all of these homologs, the aforementioned motif conforms to a somewhat more relaxed C(X)₃FCg/sLF pattern, which nevertheless contains all residues we established to be essential for Golgi targeting, suggesting that these homologs have the same localization as RavA. Both cysteines are present in this consensus sequence. The importance of the two proximal cysteine residues for localization could be that they form a disulfide bond, which rigidifies this segment into a conformation essential for binding a target molecule on the Golgi membrane. Another possibility is that they are posttranslationally modified, for example, by palmitoylation or prenylation, which would provide a membrane anchor.

Y2H screening of a human cDNA library identified several potential interaction partners of RavA, of which RAB11A was the most interesting due to the observed localization of RavA. RAB11A has been implicated in regulating vesicular trafficking through the recycling of the endosomal compartment and early endosomes to the trans‐Golgi network (TGN) and plasma membrane. ³¹ Several other Legionella effectors are known to interact with RABs and influence vesicle trafficking. ²⁴ For example, LidA localizes to the LCV membrane and interacts with both activated and non‐activated RAB1 protein to regulate fusion of ER‐derived vesicles to the LCV ³² The co‐localization of RavA with RAB11A in HEK293 cells suggests a possible role for RavA in intracellular vesicle trafficking. Our inability to detect RavA‐RAB11A interaction by pull‐down or co‐IP suggests that the interaction might be weak and/or transient. Moreover, RavA does not appear to act as a GEF for RAB11A but might be involved in binding to (?) other cellular proteins involved in trafficking.

In conclusion, the N‐terminal region of RavA carries four structural repeats that show only partial similarity to other protein domains and are likely protein–protein interaction modules. RAB11A appears to be one of the RavA interacting proteins. The Golgi localization signal in RavA conform to a C(X)₃FCg/sLF pattern not previously associated with Golgi localization.

4. EXPERIMENTAL PROCEDURES

4.1. Cloning

The construct used in the initial protein crystallization trials encompasses residues 2–329 of RavA from Legionella pneumophila strain Paris 1 (lpp0008, https://www.ncbi.nlm.nih.gov/protein/CAH11156). This construct was cloned into the vector pMCSG7, a derivative of the vector pET‐21a, adapted for ligation‐independent cloning. ³³ The plasmid was then transformed into BL21(DE3) for protein over‐expression. The expressed protein contained a TEV‐cleavable hexa‐histidine tag at the N‐terminus. Mutations E38A and K39A, which were found essential for successful crystallization, were introduced using KOD DNA polymerase (Toyobo, Tokyo, Japan) with mutagenic primers (5′‐AAA TTA ACC AAT GAT GCA GCA TTA GTT TTA GCC AGC 3′ and 5′‐GCT GGC TAA AAC TAA TGC TGC ATC ATT GGT TAA TTT‐3′).

To obtain RavA‐GFP fusion‐containing plasmid for expression in mammalian cells, the full length of RavA was PCR‐amplified from the genomic DNA of L. pneumophila strain Paris using the primers listed in Table 3. Fragment was digested with Xho I and Kpn I and ligated to the same restriction sites of pEGFP‐N1 (Clontech Laboratories, Mountain View, CA) expression vector. Fragments of RavA(1–342) and RavA(342–419) were PCR‐amplified from pEGFP‐RavA vector and cloned into pEGFP‐N1. Site‐directed mutagenesis was used to create truncated RavA and single substitution mutants using primers listed in Table S4. Constructs with N‐terminal 3xFLAG‐RavA was obtained by cloning 3xFLAG tag into HindIII and KpnI site of pcDNA5/FRT‐TO vector, then inserted full length RavA to KpnI and XhoI to create this fusion gene. Restriction enzymes and T4 DNA ligase were purchased from New England Biolabs (NEB, Ipswich, MA).

TABLE 3.

Data collection and refinement statistics

	SeMet RavA	Native RavA
Data collection statistics
Space group	P2 2 21	P2 2 21
a,b,c (Å)	51.8 62.4 285.6	51.4 62.67 286.93
α, β, γ (°)	90.0 90.0 90.0	90.0 90.0 90.0
Wavelength (Å)	0.9794	0.9793
Resolution (Å)	48.67–3.2 (3.27–3.2)	47.82–2.8 (2.86–2.8)
Total reflections	334,348 (25584)	83,666 (6396)
Unique reflections	29,478 (2223)	22,566 (1679)
R meas	0.15 (1.38)	0.097 (0.623)
CC1/2	99.8 (89.8)	99.7 (83.5)
Completeness (%)	100.0 (100.0)	94.5 (95.4)
Redundancy	20.8 (21.62)	3.7 (3.82)
(I)/σ (I)	16.09 (3.79)	9.96 (2.52)
Wilson B (Å2)	86.0	63.3
Refinement statistics
Resolution (Å)		47.82–2.8 (2.86–2.8)
R work/R free (%)		0.203/0.230
Rmsd on angles (°)		0.535
Rmsd on bonds (Å)		0.003
Ramachandran plot
Favored (%)		97.24
Allowed (%)		2.3
Outliers (%)		0.46
Average B (Å2)		67.6
Protein (atoms)		5,070
Solvent (atoms)		36
PDB code		6WO6

Note: R _meas is an indicator of data quality. ⁴⁵ CC1/2 is defined at https://strucbio.biologie.uni‐konstanz.de/xdswiki/index.php/CC1/2. R _work = ∑|F_obs−F_calc|/∑|F_obs| summed over reflections included in the refinement. R _free = ∑|F_obs−F_calc|/∑|F_obs| summed over ~5% of reflections excluded from the refinement.

For infection studies in HEK293T cells, a construct expressing 4xHA‐RavA (p4HA‐RavA) was obtained by inserting full length RavA into BamHI and HindIII site of pICC562 vector. ³⁴ The plasmid was then electrotransformed into L. pneumophila strain Philadelphia 1.

4.2. Protein expression

For large scale expression, a 15 ml overnight culture in LB was inoculated into each liter of terrific broth media. Each preparation consists of 6 L of cultures. The inoculated culture was grown at 37°C until it reached OD₆₀₀ of 1.0 before it was induced with 1 mM isopropyl β‐D‐1‐thiogalactopyranoside (IPTG). The culture was induced at 18°C overnight and was then harvested by centrifugation at 9,110 × g for 15 min.

For over‐expression of the Seleno‐methionine derivative, a cell pellet from 100 ml of overnight culture grown in LB media was inoculated into each liter of M9 Minimal media. Each preparation consists of 2 L of cultures. After shaking at 37°C to reach an OD₆₀₀ of 0.6, a mixture of L‐amino acids (100 mg of lysine, phenylalanine, and threonine; 50 mg of isoleucine, leucine, and valine) and 60 mg of selenomethionine were added to each liter of culture. The cultures were induced with 1 mM of IPTG after 15 min. The induced cultures were grown overnight at 18°C and were then harvested by centrifugation at 9,110g for 15 min. The Seleno‐methionine labeled protein was purified the same way as the native protein.

4.3. Protein purification

The cell pellet was completely re‐suspended in lysis buffer (50 mM Tris–HCl buffer pH 8.0, 10% glycerol and 0.1% Triton X). The re‐suspended cell pellet was then lysed using a cell disruptor (Constant Systems Ltd., Northants, UK). The cell debris was removed by centrifugation at 28,965g for 30 min. The resulting supernatant was applied to 3 ml of nickel‐NTA agarose resin (Qiagen). The column was washed with 5 column volumes of standard buffer (20 mM Tris pH 8.0 and 50 mM NaCl). A step gradient containing 100 mM, 200 mM and 300 mM of imidazole in standard buffer was used to elute the his‐tagged protein. Fractions positive for RavA were pooled and the TEV‐cleavable his‐tag was removed by incubating the samples overnight with the TEV protease at room temperature. The treated sample was then applied to 3 ml of Q‐sepharose fast flow resin (GE healthcare). The column was washed with 5 column volumes of standard buffer (20 mM Tris pH 8.0 and 50 mM NaCl). A step gradient containing 100 mM, 300 mM and 500 mM of NaCl in standard buffer was used to elute the protein. The 300 mM and 500 mM NaCl fractions were then applied to Superdex200 (Pharmacia) equilibrated with crystallization buffer (15 mM Tris–HCl pH 8.0, 100 mM NaCl). The flow of the buffer was controlled by the ÄKTA Prime™ system (Pharmacia) and was pumped at a flow rate of 0.5 ml/min. Fractions with pure RavA were pooled and concentrated to 5 mg/ml using the Millipore centrifugal filter with a molecular weight cut‐off of 10,000 for crystallization trials. The concentration was measured using the Nanodrop UV Spectrophotometer (Thermo Scientific) using extinction coefficient of 3,230 as calculated by the ProtParam server. ³⁵

4.4. Protein crystallization

Initial crystals were obtained by screening using commercial and in‐house screens in a 96‐well plate format. The crystallization was setup using Gryphon crystallization robot (Art Robbins Instruments, Sunnyvale, CA). The best crystals were obtained using hanging‐drop vapor diffusion method at 20°C. 1 μL of protein solution was mixed with 1 μL of reservoir solution containing 0.1 M Ammonium citrate tribasic pH 7.0, 20% w/v Polyethylene glycol 8,000 and 15% glycerol and suspended over 0.5 ml of reservoir.

4.5. Data collection

For data collection, the crystal was soaked briefly in a cryo‐protecting solution containing 0.1 M Ammonium citrate tribasic pH 7.0, 20% wt/vol Polyethylene glycol 8,000 and 20% glycerol. The same procedure was followed for the Seleno‐methionine labeled derivative. Data collection was carried out at the Canadian Light Source (CLS) on beam line 08ID‐1 for the native crystal and on 08B1‐1 for the Selenomethionine labeled derivative. The dataset from the Selenomethionine derivative was indexed, integrated and scaled using XDS program ³⁶ with AutoProcess, script. ³⁷ The structure was solved by single‐wavelength anomalous dispersion (SAD) using SHELX ³⁸ on the CCP4 online server. ³⁹ The native dataset was processed using XDS using the AutoProcess script. Molecular replacement on the partial model from SHELX was carried out using Phaser in PHENIX. ⁴⁰ The rest of the model was built manually using Coot. ⁴¹ The data processing and refinement statistics are shown in Table 3. The structure factors and coordinates have been deposited in the Protein Data Bank with the accession number 6WO6.

4.6. Transient transfection and stable cell line

Human embryonic kidney cell line 293 T (HEK293T) was cultured in Dulbecco's Modified Eagle Medium (Sigma‐Aldrich, St. Louis, MO) supplemented with 10% Fetal Bovine Serum (FBS) (Sigma‐Aldrich, St. Louis, MO) at 37°C with 5% CO₂. DNA constructs were transfected into cells using the X‐treme GENE™ HP DNA Transfection Reagent (Roche, Cat. 06366236001) according to the manufacturer's instructions.

4.7. Fluorescence cell imaging

For Staining the Golgi complex in living cells, HEK293T cells were seeded in a glass‐bottom dish (Thermo Fisher Scientific) at 3 × 10⁵ cells/ml in DMEM and were cultured for 24 hr. Then cells were incubated with 5 μM BODIPY™ TR‐ceramide (Thermo Fisher Scientific) in HBSS buffer at 4°C for 30 min. After washing with ice‐cold medium, cells were incubated in a fresh medium at 37°C for a further 30 min and then imaged live.

For immunofluorescent labeling, cells were cultured on glass coverslip, fixed 24 hr post‐transfection in 4% paraformaldehyde for 30 min and then permeabilized with 0.1% Triton X‐100 in phosphate buffered saline (PBS) for 10 min. Cells were then blocked using 5% FBS and 0.05% Tween‐20 in PBS. Endogenous GM130 and HA‐tag was probed using anti‐GM130 antibody (BD Biosciences) or anti‐HA antibody (Santa Cruz), followed by Alexa Fluor 546 anti‐rabbit(red) or Alexa Fluor 488 anti‐mouse (green) secondary antibody (ThermoFisher Scientific). Slides were mounted and visualized on a Laser Scanner Confocal Microscope (Zeiss LSM700).

4.8. L. pneumophila infection microscopy

HEK‐FcγR were seeded on 24‐well tissue culture plates (Corning) and 16–24 hr later infected with L. pneumophila harboring either the empty vector pICC562 or p4HA‐RavA that had been grown in media containing 1 mM IPTG overnight. For 4 hr and 8 hr infection, a multiplicity of infection (MOI) of 1 was used, whereas for 24 h infection, MOI of 0.5 was used to prevent too much cell death. The infection was then synchronized by centrifugation at 1100g for 5–10 min. At the appropriate timepoints, infected cells were fixed with 4% (wt/vol) PFA in PBS for 12 min on ice. Fixation was quenched with the addition of 50 mM ammonium chloride (NH₄Cl) for 20 min. Cells were then permeabilized using 0.1% (vol/vol) Triton X‐100 (Sigma–Aldrich) for 3 min and blocked using 3% (wt/vol) Bovine Serum Albumin (BSA, Sigma–Aldrich) for 30 min at room temperature. Cells were then incubated with the primary antibodies, consisting of 1:200 mouse anti‐HA (Covalence) and 1:1000 rabbit anti‐GRASP65 (Abcam) for 1 hr in the dark at room temperature. After 3 × 5‐min washes with PBS, to detect the bound primary antibodies, cells were incubated for 30 min in a secondary antibody cocktail containing 1:2,000 Alexa Fluor 488‐conjugated anti‐mouse and 1:2,000 Alexa Fluor 568‐conjugated anti‐rabbit antibodies. To stain the cells and bacterial nuclei, 1:4,000 Hoechst was also added to the secondary antibody mixture. Slides were then mounted using the ProLong Gold Antifade Mounting Medium (ThermoFisher Scientific) and visualized using Olympus FV1200 confocal microscope. Image processing was done using FIJI.

4.9. Yeast two‐hybrid

In order to identify host cell proteins that interact with RavA, the yeast two‐hybrid screening with normalized universal human library (Clontech‐Takara, Cat. No. 630480) was performed. The construct pGBKT7‐RavA, encoding RavA fused to the DNA binding domain of GAL4 (GAL4BD‐RavA), was transformed into the strain PJ69‐4A of S. cerevisiae yeast. The auto‐activation of the resulting fusion was checked by mating of the PJ69‐4A/pGBKT7‐RavA strain and Y187 strain, transformed by empty vector pGADT7‐RecAB. No growth of the diploids on the selective medium without histidine and adenine, as well as no activity of MEL1 reporter gene of α‐galactosidase on the appropriate medium with X‐α‐Gal substrate were observed, which confirmed the absence of auto‐activation. To identify the interaction partners for RavA the bait‐carrying strain PJ69‐4A/pGBKT7‐RavA was mated with Y187 strain carrying normalized human cDNA library cloned into pGADT7‐RecAB plasmid according to the manufacturer's protocol (Clontech‐Takara). The diploids were screened for double (histidine and adenine) auxotrophy and for MEL1 gene activation on selective plates with X‐α‐Gal substrate. DNA sequences fused to the activating domain of GAL4 of the selected diploids were PCR amplified, sequenced and analyzed. Some of the perspective preys‐candidate pairs, listed in Table 2, were tested for the false positive interactions. The plasmids pGADT7‐RecAB with the respective DNA insert corresponding to the listed peptides, were rescued from the selected diploids. Y187 strain was transformed with these plasmids yielding reporter strains Y187/pGADT7‐prey that carried the DNA fragment corresponding to each of the peptides listed in Table 3. The resulting reporter strains were mated to PJ69‐4A strain carrying empty pGBKT9 plasmid. The diploids were tested for the reporter genes (HIS3, ADE2, MEL1) activation. They neither exhibit growth on the selective plates nor showed α‐galactosidase activity on the plates with X‐α‐Gal substrate suggesting the absence of the reporter genes activity and accordingly confirmed true positive interaction with RavA.

4.10. Mass spectrometry

The 3xFLAG‐RavA was overexpressed in HEK293T cells as described above. Protein complexes were cross‐linked with dithiobis(succinimidylpropionate) (DSP) and affinity‐purified (AP) for analysis by mass spectrometry (MS) essentially as described previously. ⁴² Control (3xFLAG) and experimental (3xFLAG‐RavA) AP‐MS experiments were independently performed three times. Purified complexes were divided into two parts and proteolyzed by trypsin or LysN (Thermo Fisher Scientific). The Orbitrap Elite raw files were searched against Homo sapiens proteins (taxonomy ID 9606; 21,044 Uniprot entries, retrieved on 1 December, 2016) and 3xFLAG‐RavA sequence using SequestHT algorithm in Proteome Discoverer (Version 2.1.1.21; Thermo Fisher Scientific). Default search parameters were used and included cysteine carbamidomethylation (+57.0214 Da), methionine oxidation (+15.9949 Da), and serine, threonine, and tyrosine phosphorylation (+79.996 Da). Searches were performed with full digestion setting and allowed a maximum of two missed cleavages. Percolator node (Proteome Discoverer Version 2.1.1.21; Thermo Fisher Scientific) was used to filter the data with maximum Delta Cn of 0.05; FDR 0.01 for high‐confidence peptides, FDR 0.05 for medium‐confidence peptides, and q‐value‐based validation. Peptides with FDR > 0.05 were filtered out. Phosphorylation sites were localized using ptmRS (implemented in Proteome Discoverer Version 2.1.1.21; Thermo Fisher Scientific). Proteins, present in controls and frequent contaminants, identified in all purifications were eliminated from analyses. List of unique proteins was derived by retaining only a single isoform of each protein, summing up the number of unique peptides, and eliminating any proteins whose identification was supported by fewer than two unique peptides. Statistical overrepresentation test was performed using PANTHER ⁴³ version 15.0 with default settings against release 23 March 2020 of Gene Ontology database. ⁴⁴

AUTHOR CONTRIBUTIONS

Ivy Y. W. Chung: Data curation; investigation; resources; writing‐original draft; writing‐review & editing. Lei Li: Conceptualization; data curation; formal analysis; investigation; resources; visualization; writing‐original draft; writing‐review & editing. Oleg Tyurin: Data curation; investigation; methodology; resources; validation; writing‐original draft; writing‐review & editing. Alla Gagarinova: Data curation; investigation; validation; writing‐original draft; writing‐review & editing. Raissa WIbawa: Data curation; investigation; validation; writing‐original draft. Pengfei Li: Investigation; validation. Elizabeth L. Hartland: Conceptualization; methodology; project administration; resources; supervision; validation; writing‐original draft; writing‐review & editing. Miroslaw Cygler: Conceptualization; funding acquisition; project administration; resources; supervision; validation; writing‐original draft; writing‐review & editing.

DATA AVAILABILITY STATEMENT

Coordinates and structure factors are deposited in the Protein Data Bank with PDB ID: 6WO6.

Supporting information

Appendix S1: Supporting information

Table S1 Supporting information

Table S2 Supporting information

Table S3 Supporting information

Chung IYW, Li L, Tyurin O, et al. Structural and functional study of Legionella pneumophila effector RavA. Protein Science. 2021;30:940–955. 10.1002/pro.4057