Structural and mechanistic basis for protein glutamylation by the kinase fold

Summary

The kinase domain transfers phosphate from ATP to substrates. However, the Legionella effector SidJ adopts a kinase fold yet catalyzes calmodulin (CaM)-dependent glutamylation to inactivate the SidE ubiquitin ligases. The structural and mechanistic basis in which the kinase domain catalyzes protein glutamylation is unknown. Here we present cryo-EM reconstructions of SidJ:CaM:SidE reaction intermediate complexes. We show that the kinase-like active site of SidJ adenylates an active site Glu in SidE resulting in the formation of a stable reaction intermediate complex. An insertion in the catalytic loop of the kinase domain positions the donor Glu near the acyl-adenylate for peptide bond formation. Our structural analysis led us to discover that the SidJ paralog SdjA is a glutamylase that differentially regulates the SidE-ligases during Legionella infection. Our results uncover the structural and mechanistic basis in which the kinase fold catalyzes non-ribosomal amino acid ligations and reveal an unappreciated level of SidE-family regulation.

Graphical Abstract

eTOC:

Osinski et al. report cryo-EM reconstructions of the SidJ:CaM:SidE reaction intermediate complex providing structural insights into protein glutamylation by a pseudokinase. Furthermore, they discover that SdjA is also a glutamylase that differentially inactivates the SidE ubiquitin ligases in vitro and during Legionella infection

Introduction

Legionella pneumophila, the causative agent of Legionnaires disease, translocates more than 330 effectors utilizing a Type 4 secretion system to establish a replicative niche known as the Legionella Containing Vacuole (LCV) (Cornejo et al., 2017; Isberg et al., 2009). Within the Legionella effector repertoire are protein domains with recognizable folds that occasionally catalyze unexpected reactions (Black et al., 2019; Mukherjee et al., 2011; Neunuebel et al., 2011; Qiu et al., 2016). For example, the Legionella effector SidJ adopts a kinase-like fold yet catalyzes calmodulin (CaM)-dependent glutamylation to inactivate the SidE ubiquitin (Ub) ligases (Bhogaraju et al., 2019; Black et al., 2019; Gan et al., 2019; Sulpizio et al., 2019). The SidE effectors SdeA, SdeB, SdeC and SidE (collectively referred to as SidE), employ ADP ribosyltransferase (ART) and phosphodiesterase (PDE) activities to catalyze ligation of Ub to proteins independent of E1 and E2 enzymes (Bhogaraju et al., 2016; Kotewicz et al., 2017; Qiu et al., 2016). The SidE ART domain uses NAD+ to ADP-ribosylate Ub on Arg42. ADP ribosylated Ub serves as a substrate for the PDE domain, which hydrolyzes the phosphodiester bond and transfers Ub to Ser residues on proteins forming a Ser-phosphoribosyl (pR)-Ub linkage.

SidE-family Ub ligases have been structurally and mechanistically analyzed (Akturk et al., 2018; Dong et al., 2018; Kalayil et al., 2018; Wang et al., 2018). Co-crystal structures of the SdeA/SidE catalytic core in complexes with Ub/Ub^R42A and NAD⁺/NADH provided insights into the molecular mechanism of ADP-ribosylation of Ub by the SidE ART domain (Dong et al., 2018; Wang et al., 2018). The PDE-Ub interactions prior to Ub transfer to substrates were also structurally analyzed, leading to a detailed reaction model (Akturk et al., 2018; Kalayil et al., 2018). All structural studies of SidE-family Ub-ligases to date have focused on interactions with Ub and NAD⁺ but have not involved interactions with SidJ.

SidE-family ubiquitination promotes infectivity of Legionella pneumophila and is required for proper formation of the LCV (Bardill et al., 2005). However, unrestrained SidE activity is harmful to the host. Therefore, SidE-family ubiquitination is regulated by Legionella deubiquitinases DupA and DupB, which reverse pR ubiquitination (Shin et al., 2020; Wan et al., 2019), and the pseudokinase SidJ, which glutamylates and inactivates the SidE effectors (Bhogaraju et al., 2019; Black et al., 2019; Gan et al., 2019; Sulpizio et al., 2019). Three of the four SidE-family members lie within a contiguous genomic locus that also includes dupA and sidJ (Figure 1A). The dupB gene neighbors the sidJ paralog, sdjA. Despite sharing 52% amino acid sequence identity, SidJ and SdjA are not functionally redundant (Qiu et al., 2017b). As such, the function of SdjA is unknown.

Figure 1. SidJ forms a stable acyl-adenylate intermediate complex with SdeA and SdeC.

(A) Organization of the sidE family (salmon), sidJ (green) and dupA (white), dupB, (grey) and sdjA (blue) effectors in the genome of L. pneumophila. SdeA, SdeB and SdeC lie in a genomic neighbourhood with SidJ and DupA, while SidE lies in a different locus (upper).

(B) Structure of SidJ (PDB: 6OQQ) depicting the N-terminal domain (NTD; orange), the kinase-like domain (green), the C-terminal domain (CTD; black) and CaM (purple). The kinase-like active site (red) and the migrated nucleotide binding pocket formed by an insertion in the catalytic loop (orange) are highlighted. Mg²⁺ ions are shown as green spheres.

(C) Schematic representation of SidJ-catalyzed glutamylation of the SidE effectors. The SidJ:CaM complex binds ATP/Mg²⁺, which is used to adenylate the SidE-family active site Glu (SdeC; Glu⁸⁵⁷). Adenylated SidE forms a stable reaction intermediate complex with SidJ, which facilitates Glu binding, positioning of the NH₂ group for attack of the acyl-adenylate and formation of the isopeptide bond. Note that under basic conditions, :OH⁻ can hydrolyze the acyl-adenylate.

(D) SEC trace (upper) and aligned SDS-PAGE and Coomassie staining (lower) of the SidJ:CaM:SdeC complex (blue) and the products resulting from the addition of Glu (dashed red) or increasing the pH to 7.5 (dashed blue). Note that the addition of Glu and increasing the pH dissociates the complex, consistent with the presence of an acyl-adenylate in the complex. Star indicates fraction used for Cryo-EM studies. The white dashed line on the Coomassie stained gel indicates that the samples are on the same gel.

(E) Cryo-EM density map representation of the SidJ:CaM:SdeA complex. SidJ is in green, SdeA^Core is in salmon and CaM is in purple. Front and back views are shown.

Pseudoenzymes contain a protein domain that resembles a catalytically active counterpart (Kwon et al., 2019; Ribeiro et al., 2019). However, pseudoenzymes lack key catalytic residues believed to be required for activity. Recent work on pseudokinases has revealed that different binding orientations of ATP and active site residue migration can repurpose the kinase scaffold to catalyze novel reactions (Black et al., 2019; Slanina et al., 2021; Sreelatha et al., 2018). These results suggest that the kinase fold is more versatile than previously appreciated and that pseudokinases should be reanalyzed for alternative transferase activities.

Structures of the SidJ:CaM complex uncovered an N-terminal domain (NTD) of unknown function (residues 136–313), a kinase-like domain with two ligand binding pockets (residues 336–758) and a C-terminal domain that binds CaM (residues 759–873) (Bhogaraju et al., 2019; Black et al., 2019; Gan et al., 2019; Sulpizio et al., 2019) (Figure 1B). The canonical kinase-like ATP binding pocket of SidJ contains residues that are required for glutamylation of the SidE ligases, including Lys367 (Lys72 using protein kinase A; PKA nomenclature), N534 (PKA; N171) and D542 (PKA; D184). An insertion in the catalytic loop of the kinase domain (residues 487–532) extends away from the kinase active site and forms a migrated nucleotide binding pocket, which is also required for glutamylation (Bhogaraju et al., 2019; Black et al., 2019; Gan et al., 2019; Sulpizio et al., 2019). From these studies, we proposed a catalytic mechanism, whereby CaM binding allows the kinase–like active site of SidJ to bind ATP and transfer AMP to the active site Glu on SidE forming a high energy acyl-adenylate intermediate. Adenylated SidE binds the migrated nucleotide-binding pocket, positioning the acyl-adenylate and donor Glu for glutamylation and inactivation of the SidE effectors (Black et al., 2019) (Figure 1C). However, others argue that the migrated nucleotide binding pocket is an allosteric site that facilitates glutamylation (Sulpizio et al., 2019).

Here, we present cryo-EM structures of the reaction intermediate complex formed between SidJ and SdeA, and SidJ and SdeC. We propose a model that explains how a pseudokinase catalyzes protein glutamylation. Furthermore, we show that the SidJ paralog, SdjA is an active glutamylase that differentially regulates SidE-family activity in vitro and during Legionella infection.

Results

SidJ and SdeA/C Form a Stable Reaction Intermediate Complex

In several amidoligases, including the ATP-grasp enzymes involved in tubulin glutamylation, a stable reaction intermediate complex facilitates the formation of the peptide bond (Garnham et al., 2015; Mahalingan et al., 2020). We hypothesized that SidJ:CaM would also form a stable reaction intermediate complex with the SidE ligases when Glu was excluded and when the reaction was performed under acidic conditions to prevent base-catalyzed hydrolysis of the acyl-adenylate (Figure 1C). We incubated SidJ^97–851:CaM (SidJ:CaM) with SdeA^231–1190 (SdeA^Core) or SdeC^231–1222 (SdeC^Core), initiated the reactions with ATP/Mg²⁺, and analyzed the reaction products by SDS-PAGE following size-exclusion chromatography (SEC) in a slightly acidic (pH 6.5) buffer. We detected the formation of stable ternary complexes consisting of SidJ, CaM, and SdeC^Core (Figure 1D) and SidJ, CaM and SdeA^Core (Figure S1A). Upon supplementation of the reaction with Glu, or when the complex was subjected to SEC using a mildly basic buffer (pH 7.5), the SidJ:CaM:SdeC^Core complex dissociated (Figure 1D). These results are consistent with the presence of an acyl-adenylate intermediate within the complex.

Extensive Interface Interactions Facilitate Complex Formation Between SidJ:CaM and SdeA/C^Core

We determined cryo-EM structures of SidJ:CaM:SdeA^Core and SidJ:CaM:SdeC^Core complexes with resolution up to 2.5 Å and 2.8 Å, respectively (Figure 1E, and Figures S1, S2 and Table 1). The density maps reveal details of the interface between the proteins and their active sites, with clear densities for SidJ, CaM and the PDE and ART domains of SdeA^Core and SdeC^Core. Densities for the distal parts of the PDE domain and most of the C-terminal coiled coil are less clear, suggesting a higher degree of flexibility in these regions. We use the SidJ:CaM:SdeA^Core and the SidJ:CaM:SdeC^Core maps for description of the interface and the SidJ:CaM:SdeC^Core maps to describe interactions related to catalysis.

Table 1.

Data collection and refinement statistics.

	SidJ:CaM:SdeA (EMD-23862) (PDBID: 7MIR)	SidJ:CaM:SdeC (EMD-23863) (PDBID: 7MIS)
Data collection and processing
Magnification	81,000	81,000
Voltage (kV)	300	300
Electron exposure (e−/Å−2)	50	50
Defocus range (μm)	0.8 – 2.5	0.8 – 2.5
Pixel size (Å)	1.08	1.08
Symmetry imposed	C1	C1
Initial particle images (no.)	3,193,233	4,454,035
Final particle images (no.)	310,154	152,589
Map resolution (Å)	2.5	2.8
FSC threshold	0.143	0.143
Refinement
Initial model used (PDB code)	5YIM, 6S5T	SidJ-SdeA
Model resolution (Å)	2.6	2.9
FSC threshold	0.5	0.5
Map sharpening B factor (Å−2)	−37	−47
Model composition
Nonhydrogen atoms	12,997	13,039
Protein residues	1,603	1,616
Ligands	5	6
B factors (Å−2)
Protein	48.5	54.9
Ligands	36.9	30.8
R.m.s. deviations
Bond lengths (Å)	0.003	0.003
Bond angles (°)	0.485	0.468
Validation
MolProbity score	1.27	1.42
Clashscore	2.83	3.74
Poor rotamers (%)	0	0
Ramachandran plot
Favored (%)	96.8	96.2
Allowed (%)	3.2	3.8
Disallowed (%)	0	0

The overall reaction intermediate complexes are highly similar, with a root mean square deviation (RMSD) of 1.5 Å (Figures S2C and S3). The SidJ/CaM conformation remains similar to those previously determined by X-ray crystallography (Black et al., 2019) and cryo-EM (Bhogaraju et al., 2019). SidJ sits in a V-shaped cleft made up of the PDE and ART domains of SdeA/C^Core (Figure 2A). The interfaces between SidJ and SdeA/C^Core span ~2,200 Ų and are predominantly formed by the SidJ NTD (Figure 2B). Within the NTD of SidJ, there are three regions that form extensive contacts with SdeA^Core. 1) A helix-turn-helix (HTH) motif that forms electrostatic contacts with residues located in the back of the ART domain, 2) a β-hairpin wedged in between the PDE and ART domains and 3) a helical bundle that interacts with the PDE domain. To verify the relevance of these interactions for SidJ-catalyzed glutamylation of SdeA, we mutated D833, K569, D565 and D372 of SdeA^Core, which lie near the interfaces of the three regions. Reversing the charge on these residues in SdeA^Core markedly reduced its propensity to be glutamylated by SidJ, and abolished formation of the reaction intermediate complex as judged by SEC (Figure 2C and S2D). Our observation that SidJ interacts with both the SdeA/C ART and PDE domains reconciles previous data showing that SidJ-dependent suppression of SdeA-mediated yeast toxicity requires the PDE and ART domains of SdeA (Havey and Roy, 2015). As expected, SidJ did not glutamylate the isolated ART domain of SdeA (Figure 2D).

Figure 2. SidJ:CaM:SdeA/C^Core complex formation is mediated by extensive interface interactions.

(A) Overview of SidJ:CaM:SdeA^Core reaction intermediate complex. The SidJ NTD, kinase-like domain and CTD are in orange, green, and black, respectively. The SdeA PDE domain is in red and the ART domain is in blue. CaM is in purple. SidJ lies within a V-shaped cleft between SdeA PDE and ART domains.

(B) Back view of the SidJ:CaM:SdeA^Core interface. The SidJ NTD forms three distinct regions of interaction with SdeA (labelled 1–3). 1. A loop within the SdeA ART domain interacts with the helix-turn-helix (HTH) motif within the SidJ NTD. 2. A β-hairpin from the SidJ NTD is nestled between the PDE and ART domains of SdeA, forming electrostatic and hydrophobic interactions. A helix from the SidJ kinase-like domain also interacts with the SidJ β-hairpin and the PDE domain of SdeA. 3. A helical bundle from the SidJ NTD interacts with the PDE domain of SdeA. Zoomed in view depicting the residues involved in the interaction interfaces are shown in subpanels (lower) and colored as in A. Residues targeted for mutagenesis are marked with asterisks. Putative hydrogen bonds and salt bridges are depicted as dotted lines.

(C) Glutamylation activity of SidJ using SdeA^Core and mutations that disrupt interface interactions. Residues mutated are also marked with an asterisk in (B). [³H]Glu was used as a substrate and the reaction products were resolved by SDS PAGE and visualized by Coomassie staining. Radioactive gel bands were excised and ³H incorporation into SdeA^Core was quantified by scintillation counting, n=3.

(D) Incorporation of [¹⁴C]-Glu into full length SdeA, SdeA^PDE+ART, SdeA^PDE or SdeA^ART by SidJ. Reaction products were separated by SDS PAGE and visualized by Coomassie staining (upper) and autoradiography (lower).

Data in graphs shown is mean ± SD, ****P <0.0001.

Structural Comparison of the SidJ:SdeA and SidJ:SdeC Interfaces

Although the secondary structural elements that make up the interface between SidJ and SdeA^Core and SidJ and SdeC^Core are conserved, the interacting residues differ between the two complexes (Figure 3). Within the SidJ HTH, Arg262 forms a hydrogen bond with Asp833 within the α-25 helix of SdeA (Figure 3A; left). However, Asp833 is replaced with a Gly in SdeC (Gly830; Figure 3A, right). Nevertheless, Asp882 within the β12-β13 loop in SdeC forms a salt bridge with Arg262 in SidJ thus preserving the integrity of the interface (Figure 3A; right).

Figure 3. SidJ:CaM:SdeA^Core and SidJ:CaM:SdeC^Core complexes are formed by distinct interactions.

(A) Comparison of interactions between SidJ:CaM:SdeA^Core (left) and SidJ:CaM:SdeC^Core (right) in the Helix-loop-Helix region (orange). SidJ Arg262 forms a hydrogen bond with SdeA Asp833 (left), and SdeC Asp882 (right).

(B) Comparison of interactions in the hairpin region (orange). In complex with SdeA, a loop in SidJ extending from the α23 helix of the Kinase domain (green) forms electrostatic interactions with Asp565 and Lys569 in the α14 helix of SdeA (left). In complex with SdeC, the α13 helix of SdeC moves away, bringing the α1 helix closer to the SidJ hairpin, allowing for hydrogen bond formation between SidJ Arg293, and main chain carbonyl of Asn236.

(C) Comparison of interactions with the SidJ helical bundle region (orange). In complex with SdeA (left), SidJ Asn231 and Arg232 form hydrogen bonds with Asp238 and Asp372 on SdeA PDE domain, respectively (right). In complex with SdeC, the SdeC PDE domain moves away from the helical bundle (right).

SidJ^NTD – orange, SidJ^Kinase – green, SdeA – salmon, SdeC – purple. Dashed lines indicate interactions.

Within the C-lobe of the kinase-like domain of SidJ, a large loop between the α-23 and α-24 helices interacts with the α-14 helix within the PDE domain of SdeA^Core, adjacent to its interaction with the SidJ NTD β-hairpin (Figure 2B; inset 2 and Figure 3B). Glu707 and Gly704 within the α-23-α24 loop of SidJ interact with Lys569 and Asp565 in SdeA^Core. Interestingly, in the SidJ:SdeC^Core interface, the PDE domain is slightly shifted and the main chain carbonyl of Asn236 within the loop preceding the α-1 helix forms a hydrogen bond with Arg293 within the SidJ β-hairpin (Figure 3B).

Interactions with the helical bundle motif of SidJ are also not conserved between the two complexes. In SdeA^Core, Asp238 from the α-1 helix and Asp372 from the α-5 helix form electrostatic interactions with residues located within the SidJ helical bundle (Figure 3C, left). However, the SdeC PDE domain is shifted away from the SidJ helical bundle such that SidJ Asn231 is unable to interact with SdeC Asp238 (Asp238 in SdeA). Asp372 in SdeA is replaced by K381 in SdeC, making the interaction with SidJ R232 unfavorable (Figure 3C, right). The shift however is what allows for the interactions around the β-hairpin region of SidJ, bringing SidJ Arg293 and SdeC Asn236 closer, enabling hydrogen bonding (Figure 3B, right). Collectively, the differences between the SidJ:SdeA and SidJ:SdeC interfaces suggest that SidJ, rather than forming interactions with highly conserved residues, recognizes the overall fold of the SidE effectors by a few strong, and multiple weak interactions.

Unique Active Sites in SidJ Catalyze Adenylation and Glutamylation

Interactions also occur within the migrated-nucleotide binding pocket and the ARTT loop within the active site of the ART domain. Remarkably, we observed density that is consistent with a covalent bond between the phosphate of AMP and the γ-carboxyl group of E857 in SdeC^Core, which is consistent with an acyl-adenylate (Figure 4A). The acyl-adenylate is stabilized by Arg500, Asn733, His492 and Tyr506 within the SidJ migrated nucleotide binding pocket. Mutation of these residues to Ala markedly reduced glutamylation of SdeA by SidJ (Black et al., 2019). When comparing the active site of SdeA in the reaction intermediate complex to the apo structure of SdeA (Dong et al., 2018), significant rearrangement of the ARTT loop is observed, which facilitates SidJ binding and stabilization of the acyl-adenylate (Figure S4A).

Figure 4. Unique active sites facilitate SidJ-mediated glutamylation of the SidE effectors.

(A) Details of the migrated nucleotide binding pocket of SidJ (green) and the ARTT loop of SdeC^Core (blue) showing the interactions (dashed lines) involved in binding the acyl-adenylate reaction intermediate. The AMP is shown as sticks and the Coulomb potential map is shown in mesh. Note the formation of a covalent bond between AMP and the SdeC active site Glu857.

(B, C) Electrostatic surface representation (B) and molecular interactions (C) of the migrated nucleotide binding pocket of SidJ bound to SdeC^Core depicting a positively charged cleft adjacent to the acyl-adenylate reaction intermediate. A donor Glu was manually placed inside of the pocket, with its γ-carboxyl group interacting with Arg522 of SidJ. The donor Glu :NH₂ group is in position to attack the acyl-adenylate. Electrostatic surface is contoured at 7kT.

(D) Glutamylation activity of SidJ and the Arg522 mutant using SdeA^Core as substrate. Reaction products were analyzed as in Figure 2C, n=3.

(E) Details of the kinase-like active site of SidJ (green) depicting the molecular interactions (dashed lines) that facilitate ATP/Mg²⁺ binding. The ATP is in sticks, the Coulomb potential map is shown in mesh and the Mg²⁺ ions are shown as spheres.

(F) Glutamylation activity of SidJ and kinase-like active site mutants using SdeA^Core as a substrate. Reaction products were analyzed as in Figure 2C, n=3. Note that the other residues have been mutated and analyzed for glutamylation in our previous work (Black et al., 2019).

(G) Ball-and-stick representation of superimposed nucleotides from SidJ and protein kinase CK1 (left) and SidJ and SelO (right) as a result of superposition of the kinase active sites. The α, β, and γ-phosphates of SidJ, SelO and CK1 are highlighted. Note that the nucleotide orientation of SidJ is similar to SelO, which transfers AMP to proteins (Sreelatha et al., 2018).

(H) Quantification of acyl-adenylate formation following reactions with SidJ^59–851 (WT and mutants), SdeA^Core and [α-³²P]ATP. SdeA^Core E860Q was used to calculate the baseline signal in TCA precipitates. The reactions were terminated by the addition of TCA and the SdeA acyl-adenylate was detected by scintillation counting of the acid-insoluble material (red bars), n=4. Glu was also added to the reactions (green bars), which displaces the ³²P-AMP from SdeA into a TCA soluble fraction.

(I) Differential scanning fluorimetry (DSF) depicting the thermal stability profiles of SidJ (left) and SidJ:CaM (right) in the presence or absence of ATP/Mg²⁺. Protein denaturation was followed by monitoring the fluorescence of SYPRO Orange dye, which binds hydrophobic regions on proteins. The T_m values are shown in the insets. Graphs show a mean of separately normalized triplicates.

(J) Quantification of acyl-adenylate formation following reactions with SidJ^59–851, SdeA^Core, [α-³²P]ATP in the presence (+) or absence (−) of CaM. SdeA^Core E860Q was used to calculate the baseline signal in TCA precipitates. Reaction products were analyzed as in Figure 4H, n=3.

Data in graphs shown as mean ± SD, ****P <0.0001, ns >0.15.

During catalysis, SidJ cannot substitute Asp for Glu, suggesting a high degree of specificity (Black et al., 2019). Analysis of the electrostatic surface in the vicinity of the active site revealed a small positively charged pocket in SidJ that we predict binds Glu (Figure 4B). We modelled a donor Glu into this pocket, positioning the :NH₂ group near the acyl-adenylate for nucleophilic attack and formation of the Glu-Glu isopeptide bond. The γ-carboxyl was placed at a distance of ~3 Å from the SidJ Arg522 guanidinium, indicating a potential strong electrostatic interaction (Figure 4C). Mutation of Arg522 abolished SidJ-catalyzed glutamylation of SdeA (Figure 4D). Assuming the carboxyl group of a donor Asp side chain could enter the pocket and interact with Arg522, the distance between the :NH₂ group of Asp and the acyl-adenylate would be too long for it to perform a nucleophilic attack. This mode of specificity is analogous to the Glu specificity observed for the tubulin glutamylase, Tubulin-tyrosine ligase-like 6 (TTL6) (Mahalingan et al., 2020) (Figure S4B).

We also observed density in the kinase-like active site that corresponds to ATP and two Mg²⁺ ions (Figure 4E). The adenine is stabilized by stacking interaction with Ty532 and the phosphates and Mg²⁺ ions are bound by kinase-like active site residues including Lys367 (PKA; K72), Asn534 (PKA; N171) and Asp542 (PKA; D184). Mutation of these residues reduced glutamylation activity (Black et al., 2019) (Figure 4F). Similar to canonical kinases, the ATP sits in a pocket between the two lobes of the kinase-like domain; however, the nucleotide is positioned in a unique manner. The β- and γ-phosphates of ATP are buried in a positively charged pocket, which is reminiscent of the pseudokinase SelO that binds ATP in a flipped orientation and transfers AMP from ATP to hydroxyl-containing amino acids (Sreelatha et al., 2018) (Figure 4G).

To determine the role of the kinase-like active site in SidJ, we performed adenylation reactions by monitoring ³²P incorporation from [α-³²P]ATP in the absence of Glu. The reactions were terminated with trichloroacetic acid (TCA) and adenylated SdeA^Core was detected as ³²P-labelled protein in TCA-insoluble material. To account for any auto-modification of SidJ or non-specific modification of SdeA, we utilized a SdeA^Core E860Q mutant as a control and also added Glu to release the ³²P-AMP signal from SdeA^Core. Although the kinase-like active site residues are required for adenylation, His492 in the migrated nucleotide binding pocket is also required (Figure 4H), likely because it stabilizes the acyl-adenylate (Figure 4A). Notably, although the putative donor Glu-interacting Arg522 within the migrated nucleotide binding pocket is required for glutamylation (Figure 4D), the R522A mutant of SidJ still retains adenylation activity towards SdeA^Core (Figure 4H). Importantly, the addition of Glu to the R522A mutant reaction failed to decrease the ³²P-signal in TCA precipitates. These results provide evidence that Arg522 in SidJ plays a major role in positioning the donor Glu. Taken together, the kinase-like active site of SidJ adenylates the SidE active site Glu, which is stabilized by the migrated nucleotide binding pocket to position the donor Glu for attack of the acyl-adenylate intermediate and the formation of an isopeptide bond.

CaM Binding is Required for SidJ to Bind ATP and Adenylate SdeA

Several bacterial effectors and toxins are activated by eukaryote-specific proteins, such as CaM, to achieve spatial regulation (Guo et al., 2016; Guo et al., 2005; Leppla, 1984). CaM activates SidJ by binding to the IQ helix within the CTD (Black et al., 2019). However, the mechanism by which CaM activates SidJ is unknown. We monitored the thermal stability of SidJ and observed an ATP/Mg²⁺-dependent thermal shift that occurred only in the presence of CaM (Figure 4I). This thermal shift also occurred in the migrated active site mutant H492A, but not the kinase-like active site D542A mutant (Figure S4C–F). As expected, CaM was required for adenylation of SdeA^Core (Figure 4J). Thus, CaM binding renders the kinase-like active site of SidJ competent to bind ATP and adenylate the SidE-ligases.

The SidJ Paralog SdjA is an Active Glutamylating Enzyme

In L. pneumophila, the T4SS effector SdjA shares ~52% sequence identity to SidJ (Figure 5A); however, the function of SdjA is unknown. When overexpressed in yeast, SidJ, but not SdjA suppresses the growth inhibition phenotype induced by SdeA (Qiu et al., 2017a), suggesting different functions for the two effectors. Moreover, deletion of SidJ in L. pneumophila results in a growth inhibition phenotype that cannot be complemented by endogenous SdjA (Liu and Luo, 2007). Although the predicted kinase-like domains are 66% identical and the catalytic residues and CaM interacting IQ motif are well conserved, the NTDs are significantly different between SidJ and SdjA (Figure S5A and B). Because the majority of the interactions between SidJ and SdeA^Core lie within the SidJ NTD (Figure 2B), we asked whether SdjA could glutamylate the SidE-ligases. We incubated SdjA with CaM, ATP/Mg²⁺, [¹⁴C]-Glu and full-length SidE-family proteins and observed ¹⁴C incorporation into SdeB, SdeC and SidE, but not SdeA (Figure 5B). Consistent with our previous results (Black et al., 2019), SidJ glutamylated all 4 effectors. The SdjA reaction required CaM, ATP/Mg²⁺ and the residues in both the kinase-like active site and the migrated nucleotide binding pocket (Figure 5C and D). Similar to SidJ, SdjA glutamylated the active site Glu in SdeB, SdeC and SidE (Figure 5E). As expected, SdjA-mediated glutamylation completely abolished Ub ligase activity of SdeB, SdeC and SidE, in vitro (Figure S5C) and in cells (Figure 5F). Thus, SidJ and SdjA are CaM-dependent glutamylases that differentially inactivate the SidE-family of Ub ligases.

Figure 5. SidJ and SdjA differentially regulate the SidE-Ub ligases in vitro.

(A) Cartoon representation of SidJ and SdjA depicting the NTD (orange), the kinase-like domain (green), and the CaM binding CTD (grey). The percent sequence identity and similarity are shown for each domain and for the overall protein.

(B) Glutamylation activity of SidJ (SidJ^59–851) and SdjA using full length SdeA, SdeB, SdeC and SidE as substrates. Reaction products were analyzed as in Figure 2D. DA denotes SidJ^D542A or SdjA^D480A. Note that SdjA does not glutamylate SdeA.

(C) Glutamylation activity of SidJ^59–851 and SdjA^36–789 using SdeC^Core as substrate. Reaction products were analyzed as in Figure 2D. DA denotes SidJ^D542A or SdjA^D480A. (D) Glutamylation activity of SdjA^36–789 and indicated mutants using SdeA^Core [³H]Glu as substrates. Reaction products were analyzed as in Figure 2C, n=3. The kinase-like active site residues and the residues in the migrated nucleotide binding pocket are highlighted. IQ mutant: Sdj^{AI776D; Q777D; R778E; R781E}.

(E) Glutamylation activity of SidJ^59–851 and SdjA^36–789 using full length SdeA, SdeB, SdeC and SidE as substrates. Reaction products were analyzed as in Figure 2D. DA denotes SidJ^D542A or SdjA^D480A. EQ denotes SdeA^E860Q, SdeB^E857Q, SdeC^E857Q, or SidE^E855Q.

(F) Protein immunoblotting of total extracts from HEK293A cells expressing HA-Ub^GG/AA, Myc-SdeA, Myc-SdeB, Myc-SdeC, or Myc-SidE and SidJ-V5, V5-SdjA or the indicated mutants. GAPDH is shown as a loading control. HA-Ub^GG/AA was used to specifically interrogate SidE-family activity because it cannot be used by the endogenous Ub machinery. DA denotes SidJ^D542A or SdjA^D480A.

Data in graphs shown as mean ± SD, ****P <0.0001.

The SidJ/SdjA NTDs Determine Specificity For the SidE Effectors

Because most of the interactions between SidJ/CaM with SdeA/C^Core come from the NTD, we hypothesized that the NTD could be the determining factor for the differential regulation of the SidE-family by SidJ and SdjA. We generated chimeric SidJ-SdjA proteins, in which the NTD from SidJ was replaced with the NTD from SdjA, and vice versa (Figure 6A). Replacing the HTH motif in SdjA with the corresponding HTH from SidJ resulted in an active chimeric protein that retained its ability to glutamylate SdeB, SdeC and SidE but also glutamylated SdeA (Figure 6B, lane 9). A SidJ chimera containing the SdjA NTD (SidJ^SdjA-NTD) lost most of its activity towards SdeA (Figure 6B, lane 13). Thus, the variable NTD appears to be the major determinant of SidJ and SdjA specificity for the SidE effectors.

Figure 6. The variable NTD is the major determinant of SidJ and SdjA specificity for the SidE effectors.

(A) Cartoon representations of SidJ^SdjA-NTD (left) and SdjA^SidJ-HTH (right) chimeras depicting the interchanged regions, presented on a crystal structure of SidJ (PDBID: 6OQQ). Portions from SidJ are in green and SdjA in salmon.

(B) Glutamylation activity of SidJ^59–851, SdjA^36–789, SdjA^SidJ-HTH and SidJ^SdjA-NTD using full length SdeA (A), SdeB (B), SdeC (C) and SidE (E) as substrates. Reaction products were analyzed as in Figure 2D. Compare lanes 1 and 13 (SidJ loss of activity), 5 and 9 (SdjA gain of activity).

Differential Regulation of the SidE-effectors by SidJ and SdjA During Legionella Infection.

SidJ is one of only a few T4SS effectors that when deleted from the Legionella genome causes a replication phenotype in infected cells, suggesting that SidJ and SdjA are not redundant (Jeong et al., 2015; Liu and Luo, 2007). We reasoned that in the background of a sdeA deletion, SidJ and SdjA would be functionally redundant (Figure 7A). We generated a ΔsdjAΔsidJΔsdeA Legionella strain and monitored its replication in the environmental host Acanthamoeba castellanii (Figure S6A). Remarkably, both SidJ and SdjA could complement the growth defect in the ΔsdjAΔsidJΔsdeA strain (Figure 7B). In contrast, SidJ, but not SdjA could rescue the growth defect in a ΔsdjAΔsidJ strain (Figure 7C). Collectively, our results suggest that the SidE-Ub ligases are differentially regulated by SidJ and SdjA during Legionella infection.

Figure 7. SidJ and SdjA differentially regulate the SidE-Ub ligases during Legionella infection.

(A) Schematic representation of SidJ/SdjA genetic interaction during Legionella infection. In a wild-type L. pneumophila (Lp02) strain, SdjA is unable to complement SidJ because it is inactive against SdeA (left). In a ΔsdeA background, SdjA is functionally redundant with SidJ (right).

(B, C) Replication of L. pneumophila strains in A. castellanii. Infected amoeba cells were lysed at the indicated timepoints and bacterial replication was quantified by plating serial dilutions of lysates. Results are representative of two independent experiments; error bars denote SD from 1 experiment performed in triplicate. ΔsdjAΔsidJΔsdeA (ΔΔΔ), ΔsidJΔsdeA (ΔΔ), EV; empty vector, DA denotes SidJ^D542A or SdjA^D480A, Lp03; ΔdotA (T4SS deficient). Error bars represent standard deviation.

(D) Model of SidJ-catalyzed glutamylation of the SidE effectors. (1) SidJ is translocated into host cells and binds CaM. (2) CaM binding renders the kinase-like active site competent to bind ATP/Mg²⁺ and (3) adenylate the active site Glu in the SidE effectors. (4) Adenylated SidE binds the migrated nucleotide binding pocket of SidJ, forming a stable reaction intermediate. Free Glu binds to a positively charged cleft in the migrated nucleotide binding pocket of SidJ which positions the NH₂ group for nucleophilic attack of the acyl-adenylate, (5) releasing AMP and forming an isopeptide bond between Glu and SidE.

Discussion

We propose a model for how SidJ glutamylates the SidE-family of Ub ligases (Figure 7D). During infection, SidJ is translocated into the host cell where it binds CaM (1), which allows the kinase-like active site to bind ATP/Mg²⁺ (2). Notably, because CaM is a eukaryote-specific protein, this mechanism allows for spatial regulation of the SidE-Ub ligases within the host cell and prevents premature inactivation of the SidE-Ub ligases within Legionella. SidJ then adenylates the active site Glu within the ARTT loop in the SidE effectors (3). Adenylated SidE binds the migrated nucleotide binding pocket (4), which positions free Glu near the acyl-adenylate intermediate for nucleophilic attack and subsequent formation of the Glu-Glu isopeptide bond (5).

Our structural and biochemical data suggest that the kinase-like active site performs the adenylation reaction and the migrated nucleotide binding pocket executes the glutamylation reaction. Although His492 in the migrated nucleotide binding pocket is required for adenylation (Figure 4H), the H492A mutant would be unable to stabilize the intermediate and therefore prevent us from detecting the acyl-adenylate in our assays. Consistent with this idea, Arg522 does not contact the acyl-adenylate and thus is not required for adenylation. Three pieces of evidence lead us to believe that Arg522 is required for Glu binding: 1) Arg522 is highly conserved (Figure S6B). 2) Arg522 is required for glutamylation but not adenylation of SidE (Figure 4D and H) and 3) the addition of Glu to adenylation reactions using the SidJ R522A mutant does not liberate AMP from adenylated SidE (Figure 4H).

SidJ homologs are found in taxonomically diverse groups of organisms including archaea and viruses, suggesting that SidJ is being spread by horizontal gene transfer (Figure S6C). Distant SidJ homologs retain strong conservation of residues in both the kinase-like active site and the migrated nucleotide binding pocket (Figure S6B). However, none of these organisms except for Legionella species have SidE homologs or the SidJ CaM-binding IQ motif and they show virtually no conservation within the SidE-interacting NTD (Figure S6B). Thus, their substrates and activation mechanisms will likely differ. Interestingly, SidJ homologs from crocodilepox viruses are located next to a cluster of proteins that differ from SidE but are involved in the modulation of host ubiquitination pathways (Afonso et al., 2006).

In mammals, tubulin glutamylation, tyrosination and glycylation are performed by TTLL enzymes, members of the ATP-grasp superfamily (Song and Brady, 2015; Yu et al., 2015). TTLLs catalyze amino acid ligation via the formation of a high-energy acyl-phosphate intermediate, which is subsequently attacked by the amine group of the donor amino acid (Szyk et al., 2011). Interestingly, both the protein kinase and ATP-grasp fold enzymes share a common topology (Grishin, 1999). In contrast to most protein kinases, SidJ adenylates a carboxyl containing amino acid as an intermediate step in peptide bond formation. We propose that the protein kinase fold can phosphorylate or adenylate carboxylic groups to mediate non-ribosomal amino-acid ligation reactions. Notably, a recent study of “the hidden phosphoproteome” revealed that a vast number of Glu and Asp residues within HeLa cells are phosphorylated (Hardman et al., 2019).

In summary, our work underscores the catalytic versatility of the kinase fold and reveals a previously unappreciated level of regulation of the SidE Ub ligases.

Limitations of the Study

Our biochemical and mutagenesis data suggests that the kinase-like active site performs the adenylation reaction and the migrated nucleotide binding pocket catalyzes the glutamylation reaction. The structures presented here have captured the reaction intermediate complex that facilitates acyl-adenylate stabilization and subsequent glutamylation. However, we lack structural information on the first step of the reaction: adenylation of the SidE active site Glu by the kinase-like active site in SidJ. Likewise, the structural rearrangements that need to occur following the adenylation reaction are unknown at this time.

Furthermore, our discovery that SidJ and SdjA differentially regulate the SidE effectors suggests that SdeB, SdeC and SidE require an additional layer of regulation and may have distinct substrate specificities. Thus, the reason why L. pneumophila has evolved two glutamylases to differentially inactivate the SidE effectors is unknown and warrants further investigation.

STAR METHODS

Resource Availability

Lead Contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Vincent S. Tagliabracci (vincent.tagliabracci@utsouthwestern.edu).

Materials Availability

Plasmids, primers, recombinant proteins, experimental strains, and any other research reagents generated by the authors will be distributed upon request to other research investigators under a Material Transfer Agreement.

Data and Code Availability

• Cryo-EM maps and final refined coordinates have been deposited in EMDB (The Electron Microscopy Data Bank) and PDB (Protein Data Bank), respectively, and are publicly available as of the date of publication. Accession numbers are listed in the key resources table. Raw data for protein immunoblots, SDS-PAGE, and autoradiography have been deposited at Mendeley Data (https://dx.doi.org/10.17632/sgnknvxz3j.1).

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit anti-SidJ	(Black et al., 2019)	N/A
Rabbit anti-SdjA	This study	N/A
Rabbit anti-SdeA	(Black et al., 2019)	N/A
Mouse anti-GAPDH	ThermoFisher	Cat#MA5-15738
Mouse anti-HA antibody	Sigma-Aldrich	Cat# H3663
Donkey anti-Mouse IgG, HRP-Linked Whole Ab	VWR	Cat#95017-554
Donkey anti-Rabbit IgG, HRP-Linked Whole Ab	VWR	Cat# 95017-556
Bacterial Strains
Rosetta DE3	Novagen	Cat#70954
Rosetta DH5α	This study	N/A
Chemicals, Peptides, and Recombinant Proteins
L. pneumophila SidJ 59-851 (E. coli)	This study	N/A
L. pneumophila SidJ 97-851 (E. coli)	This study	N/A
L. pneumophila SdjA (E. coli)	This study	N/A
L. pneumophila SdjA 36-789 (E. coli)	This study	N/A
L. pneumophila SdeA (E. coli)	This study	N/A
L. pneumophila SdeB (E. coli)	This study	N/A
L. pneumophila SdeC (E. coli)	This study	N/A
L. pneumophila SidE (E. coli)	This study	N/A
L. pneumophila SdeA 231-1190 (E. coli)	This study	N/A
L. pneumophila SdeC 231-1222 (E. coli)	This study	N/A
Human CALM2 (E. coli)	This study	N/A
Human HA-Ubb (E. coli)	This study	N/A
Ulp1 S. cerevisiae SUMO protease	This study	N/A
PfuTurbo DNA Polymerase	Thermo	Cat# 50-125-946
Q5 High-Fidelity 2X Master Mix	NEB	M0492L
Adenosine 5’-triphosphate disodium salt hydrate	Sigma	Cat# A2383
ATP, [α-32P]- 3000Ci/mmol 10mCi/ml, 250 μCi	PerkinElmer	BLU003H250UC
L-[14C(U)]-Glutamic Acid	PerkinElmer	NEC290E050UC
L-[3,4-3H]-Glutamic Acid	PerkinElmer	NET490001MC
Budget-Solve scintillation cocktail	RPI	Cat# 111167
EN3HANCE Autoradiography Enhancer	PerkinElmer	Cat# 6NE9701
Tris-(2-Carboxyethyl)phosphine, Hydrochloride (TCEP)	Fisher Scientific	Cat# PI20490
Hydrogen peroxide	Sigma	Cat# H1009
L-glutamic acid, Sodium salt	Sigma	Cat# G1626
Sodium Chloride	Fisher Scientific	BP358-10
Tris base	Fisher Scientific	BP152-5
Kanamycin sulfate	Millipore Sigma	K1377
Ampicillin sodium salt	Millipore Sigma	A9518
Chloramphenicol	Millipore Sigma	C0378
PolyJet In Vitro DNA Transfection Reagent	SignaGen Laboratories	SL100688
Protease Inhibitor Cocktail Tablets (Complete)	Millipore Sigma	11697498001
Dithiothreitol (DTT)	Millipore Sigma	D0632
DMEM, High Glucose	Thermo Fisher Scientific	11-965-118
Fetal bovine serum	Millipore Sigma	F2442
Penicillin-Streptomycin	Millipore Sigma	P0781
Magnesium Chloride hexahydrate	Millipore Sigma	M2670
Deposited Data
SidJ:CALM2:SdeA	This study	EMDB: EMD-23862 PDB: 7MIR
SidJ:CALM2:SdeC	This study	EMDB: EMD-23863 PDB: 7MIS
Experimental Models: Organisms/Strains
Legionella pneumophila Philadelphia-1 (Lp02)	Dr. Ralph Isberg	N/A
Legionella pneumophila Philadelphia-1 ΔdotA (Lp03)	Dr. Ralph Isberg	N/A
Lp02 ΔsidJ	(Black et al., 2019)	N/A
Lp02 ΔsdjA	This study	N/A
Lp02 ΔsdeA	This study	N/A
Lp02 ΔsidJΔsdjA	This study	N/A
Lp02 ΔsidJΔsdjA (pJB908 sidJ)	This study	N/A
Lp02 ΔsidJΔsdjA (pJB908 sidJ D542A)	This study	N/A
Lp02 ΔsidJΔsdjA (pJB908 sdjA)	This study	N/A
Lp02 ΔsidJΔsdjA (pJB908 sdjA D480A)	This study	N/A
Lp02 ΔsidJΔsdjA (pJB908)	This study	N/A
Lp02 ΔsidJΔsdjAΔsdeA	This study	N/A
Lp02 ΔsidJΔsdjAΔsdeA (pJB908 sidJ)	This study	N/A
Lp02 ΔsidJΔsdjAΔsdeA (pJB908 sidJ D542A)	This study	N/A
Lp02 ΔsidJΔsdjAΔsdeA (pJB908 sdjA)	This study	N/A
Lp02 ΔsidJΔsdjAΔsdeA (pJB908 sdjA D480A)	This study	N/A
Lp02 ΔsidJΔsdjAΔsdeA (pJB908)	This study	N/A
HEK293a	Invitrogen	R70507
Acanthamoeba castellanii	Dr. Kim Orth	N/A
Oligonucleotides
See Table S1 for primers used in this study
Recombinant DNA
ppSUMO L. pneumophila SidJ 59-851	(Black et al., 2019)	N/A
ppSUMO L. pneumophila SidJ 97-851	This study	N/A
ppSUMO L. pneumophila SdjA 36-789	This study	N/A
ppSUMO CALM2	(Black et al., 2019)
ppSUMO L. pneumophila SdeA	(Black et al., 2019)	N/A
ppSUMO L. pneumophila SdeB	(Black et al., 2019)	N/A
ppSUMO L. pneumophila SdeC	(Black et al., 2019)	N/A
ppSUMO L. pneumophila SidE	(Black et al., 2019)	N/A
ppSUMO L. pneumophila SdeA 231-1190	This study	N/A
ppSUMO L. pneumophila SdeC 231-1222	This study	N/A
ppSUMO L. pneumophila SidJSdjA-NTD	This study	N/A
ppSUMO L. pneumophila SdjASidJ-HTH	This study	N/A
pETDuet1 L. pneumophila SdjA CALM2	This study	N/A
pProEx2 L. pneumophila SdeA 231-1190	This study	N/A
pProEx2 L. pneumophila SdeC 231-1222	This study	N/A
pSR47s	Dr. Shaeri Mukherjee	N/A
pSR47s sdeA flanks	This study	N/A
pSR47s sdjA flanks	This study	N/A
pSR47s sidJ flanks	(Black et al., 2019)	N/A
pJB908 sidJ	(Black et al., 2019)	N/A
pJB908 sidJ D542A	(Black et al., 2019)	N/A
pJB908 SdjA	This study	N/A
pJB908 SdjA D480A	This study	N/A
pJB908	Dr. Ralph Isberg	N/A
pcDNA-HA-Ubb	Dr. Jenna Jewell	N/A
pcDNA-HA-Ubb GG/AA	(Black et al., 2019)	N/A
pcDNA-SidJ-V5	(Black et al., 2019)	N/A
pcDNA-SidJ-V5 D542A	(Black et al., 2019)	N/A
pCCF-V5-SdjA	This study	N/A
pCCF-V5-SdjA D480A	This study	N/A
pCDNA-Myc-SdeA (c. opt)	(Black et al., 2019)	N/A
pCCF-Myc-SdeB	(Black et al., 2019)	N/A
pCCF-Myc-SdeC (c. opt)	(Black et al., 2019)	N/A
pCCF-Myc-SidE	(Black et al., 2019)	N/A
Software and Algorithms
Pymol	Schrödinger, Inc.	N/A
WebLogo server	(Crooks et al., 2004)	http://weblogo.threeplusone.com/
Dalilite server	(Holm and Rosenstrom, 2010)	http://ekhidna2.biocenter.helsinki.fi/dali/
Jackhmmer server	(Finn et al., 2015)	https://www.ebi.ac.uk/Tools/hmmer/search/jackhmmer
BLAST server	(Altschul et al., 1990)	https://blast.ncbi.nlm.nih.gov/Blast.cgi
MAFFT server	(Katoh and Standley, 2013)	https://mafft.cbrc.jp/alignment/software/
Pfam database	(Finn et al., 2016)	http://pfam.xfam.org/
RELION	(Zivanov et al., 2018)	https://relion.readthedocs.io/en/latest/index.html
crYOLO	(Wagner et al., 2019)	https://cryolo.readthedocs.io/en/stable/
Gctf	(Zhang, 2016)	https://www2.mrc-lmb.cam.ac.uk/research/locally-developed-software/zhang-software/
DeepEMhancer	(Sanchez-Garcia et al., 2020)	https://github.com/rsanchezgarc/deepEMhancer
Chimera	(Pettersen et al., 2004)	https://www.cgl.ucsf.edu/chimera/
Phenix	(Adams et al., 2010)	https://www.phenix-online.org
Coot	(Emsley et al., 2010)	https://www2.mrc-lmb.cam.ac.uk/Personal/pemsley/coot
CCP4 7.1	(Winn et al., 2011)	https://www.ccp4.ac.uk/?page_id=1417
GraphPad Prism	N/A	https://www.graphpad.com/scientific-software/prism/
ESPript	(Robert and Gouet, 2014)	http://espript.ibcp.frESPript/ESPript/

Experimental Model and Subject Details

Legionella pneumophila

Legionella pneumophila subsp. pneumophila strain Philadelphia 1 derived strain Lp02 and Lp03 and mutants were cultured at 37°C in AYE liquid medium, or on CYA solid medium, prepared according to ATCC: Medium 1099 CYE, i.e. ACES [N-(2-acetamido)-2- aminoethanesulfonic acid]-buffered charcoal yeast extract (CYE) agar plates or grown in ACES buffered yeast extract (AYE) liquid cultures supplemented immediately before use with 100x concentrated sterile filtered solutions of ferric nitrate (final 0.135 g/L) and cysteine (final 0.4 g/L). Thymidine was added to a final concentration of 100 μg/mL for maintenance of the thymidine auxotrophic strains. Additional supplements used were used where applicable: kanamycin (30–50 μg/mL), sucrose (5%). Protein knockouts were confirmed by PCR and immunoblotting.

Acanthamoeba castellanii

Acanthamoeba castellanii was maintained as a monolayer culture in ATCC 712 PYG medium (20 g/L protease peptone, 1 g/L yeast extract, 100 mM glucose, 4 mM Mg₂SO₄, 0.4 mM CaCl₂, 0.1% (w/v) sodium citrate dihydrate, 0.05 mM Fe(NH₄)₂(SO₄)₂·6H₂O, 2.5 mM NaH₂PO₄, 2.5 mM K₂HPO₄, final pH adjusted to 6.5) in tissue culture flasks at 23°C.

HEK293A

HEK293A cells were grown in DMEM containing 10% (vol/vol) FBS with 100 μg/mL penicillin/streptomycin (GIBCO) at 37 °C with 5% CO₂. Cells were routinely tested for Mycoplasma contamination by PCR based methods.

Method Details

Antibodies

Mouse anti-GAPDH antibodies (CB1001) were from EMD Millipore. Mouse anti-HA antibodies were from Sigma (H3663) and anti-Legionella pneumophila antibodies (ab20943) were from Abcam. Rabbit anti-SdeA, SidJ and SdjA were raised by Cocalico and purified in-house as previously described in (Sreelatha et al., 2018), and in “Production of SdjA Antibodies” section.

Generation of Plasmids and Strains

SidJ, SdjA, SdeA, SdeB, SdeC and SidE coding sequences (CDS) were amplified by PCR using Legionella pneumophila Philadelphia-1 strain genomic DNA (gDNA) as a template. SidJ, SdjA, SdeA, SdeC CDS were also codon-optimized for mammalian expression and synthesized as gBlocks (SidJ, SdjA, SdeC; Integrative DNA Technologies, SdeA; Genscript). CALM2 and FcγRIIa were amplified from the Ultimate^™ ORF Lite human cDNA collection (Life Technologies) and cloned into mammalian or bacterial expression vectors. Amino acid mutations were introduced via Quick Change site-directed mutagenesis. Briefly, primers were designed manually, or using the Agilent Quick Change Primer design tool: https://www.genomics.agilent.com and used in PCR reactions to generate the desired mutation using PfuTurbo DNA polymerase. Reaction products were digested with Dpn1 restriction endonuclease and mutations were confirmed by Sanger sequencing. For mammalian cell expression in HEK293A cells, codon-optimized SidJ, SdjA, SdeA and SdeC were amplified from gBlocks by PCR and cloned in frame with a C-terminal V5 tag (SidJ), N-terminal V5 tag (SdjA) or amplified with an N-terminal Myc tag (SdeA, SdeC) into the CMV promoter-driven vector pcDNA (Invitrogen). pcDNA-HA-tagged Ubiquitin B (UbB) was a generous gift from Dr. Jenna Jewell and used as a template to generate the Ub^GG/AA mutant.

L. pneumophila strains Lp02, Lp03 (Lp02 ΔdotA), and thymidine auxotrophic derivatives used in this study were derived from Legionella pneumophila Philadelphia-1 strain and were generous gifts from Dr. Ralph Isberg. Legionella bacteria were maintained on ACES [N-(2-acetamido)-2- aminoethanesulfonic acid]-buffered charcoal yeast extract (CYE) agar plates or grown in ACES buffered yeast extract (AYE) liquid cultures supplemented with ferric nitrate (0.135 g/L) and cysteine (0.4 g/L). Thymidine was added to a final concentration of 100 μg/mL for maintenance of the thymidine auxotrophic strains.

SidJ, SdjA and SdeA knockout strains were generated using the R6K suicide vector pSR47s (KanR, sacB), a generous gift from Dr. Shaeri Mukherjee, UCSF. Briefly, ~800bp regions flanking the SidJ ORF were amplified and cloned using Gibson assembly into pSR47s to generate pSR47s-ΔsdeA, pSR47s-ΔsidJ, and pSR47s-ΔsdjA, which was maintained in S17-1 λpir E. coli. pSR47s vectors with flanks were introduced by electroporation into Thy− strain Lp02, or ΔsidJ, or ΔsidJ/ΔsdeA, and colonies having undergone homologous recombination were selected with kanamycin (30 μg/mL). Bacteria were streaked on CYA Thy+ plate. Merodiploids were resolved on 5% sucrose, and the resulting colonies were screened for loss of respective genes by PCR and protein immunoblotting. Complementing strains were generated using the RSF1010 cloning vector pJB908 (AmpR tdΔi), a gift from Dr. Ralph Isberg. Mutants were cloned into pJB908 and transformed by electroporation into the appropriate parental strain. Transformants were selected for on CYE medium without thymidine and complementation was verified by PCR and protein immunoblotting. For intracellular replication assays and HEK293 cell challenge, Legionella were scraped from a 2/3-day heavy patch to inoculate 2 mL AYE broth. Serial dilutions of the inoculum were grown 24–48 hours at 37°C with shaking to an OD600 of 2.5 to 4, at which time the cultures acquired a brown pigmentation.

Intracellular replication in amoeba

18 hours prior to infection, confluent amoeba monolayers were washed with A. castellani buffer ACB (4 mM Mg₂SO₄, 0.4 mM CaCl₂, 0.1% (w/v) sodium citrate dihydrate, 0.05 mM Fe(NH₄)₂ (SO₄)₂·6H₂O, 2.5 mM NaH₂PO₄, 2.5 mM K₂HPO₄, final pH adjusted to 6.5), to remove non-adherent cells, and collected by scraping. Resuspended in fresh ACB, amoeba were counted, and 6×10⁵ cells were seeded into individual wells of 24-well plates 1–2 h prior to infection. Plates were equilibrated to reach 37°C. All subsequent incubations were performed at 37°C. Legionella cultures at post-exponential phase were diluted in A. castellanii buffer and ~6×10⁴ bacteria were added to each well for a multiplicity of infection (MOI) of 0.1 (assuming 1₆₀₀OD=10⁹ Legionella CFU). Infections were synchronized by centrifugation at 230 × g for 5 minutes. Infections were allowed to proceed for 1 hour, then extracellular bacteria were removed by washing each well 3 times in ACB, before adding ACB buffer to a final volume of 0.5 mL/well. Timepoint zero (1 HPI) was lysed immediately after washing, and subsequent timepoints were taken after 24h, and 48h. Amoeba cells were lysed by addition of saponin to a final concentration of 0.05%, and scraped off the plate with a mini cell scraper. Wells were subsequently washed with 500 μL of Milli-Q H₂O, and combined with saponin lysates. Serial dilutions of the infectious inoculum and the amoeba lysates were plated on CYE plates to confirm the MOI, quantify infection efficiency, and assess bacterial growth. All solutions used during the infection were filtered using sterile 0.22 μm syringe filters.

Protein purification

L. pneumophila SidJ residues 59–851 or 97–851, SdjA FL and 36–789, human CaM (CALM2), and the SidE effectors (and indicated truncations) were cloned into a modified pET28a bacterial expression vector (ppSumo), containing an N-terminal 6X-His tag followed by the yeast Sumo (smt3) CDS. Ubiquitin, CaM, SdeA 231–1190 and SdeC 231–1222 were also cloned into pProEX2 containing an N-terminal 6X-His tag followed by a TEV protease cleavage site. For purification from E. coli, plasmids were transformed into Rosetta DE3 cells. Cells were grown in Miller modification of Luria Bertani (Miller LB) broth with 100 mg/L Ampicillin or 50 mg/L Kanamycin, in presence of 34 mg/L of Chloramphenicol. At the OD₆₀₀ of 0.6–1.0, protein expression was induced with 0.4 mM IPTG for 16–18 hours at 18 °C. Cells were harvested by centrifugation and lysed in 50 mM Tris-HCl pH 8, 300 mM NaCl, 15 mM Imidazole, 1 mM PMSF, (containing 5 mM β-ME) by sonication. Cell lysates were cleared by centrifugation at 35,000 × g for 30 minutes. The cleared lysate was incubated with washed Ni-NTA beads for a minimum of one hour at 4°C. Beads were passed over a gravity column and washed with 20 column volumes of 50 mM Tris-HCl pH 8, 300 mM NaCl, 25 mM imidazole, 5 mM β-ME. Proteins were eluted with 50 mM Tris-HCl pH 8, 300 mM NaCl, 300 mM imidazole, 1 mM DTT. SdjA purifications were performed in 500 mM NaCl to aid stability. Full-length SdjA cloned into ppSumo was coexpressed with human 6X-His-CALM2 in pProEX2 and grown under triple selection (Kan, Amp, Cm), and was purified in one step (Ni-NTA). Proteins were cut overnight at 4 °C with 6X-His tagged Ulp1 Sumo protease or TEV protease, followed by gel filtration chromatography using a Superdex 200 gel filtration column attached to an AKTA Pure FPLC chromatography system (GE Healthcare) in 25 mM Tris 7.5 or 8.0, 150 mM or 300 mM of NaCl, 1 mM DTT. Proteins were concentrated (~15–30 mg/mL SidJ, ~30 mg/mL CALM2, ~30 mg/mL SdeA^Core, 60 mg/mL SdeC^Core, ~10 mg/mL full-length SidE effectors), aliquoted, flash frozen in liquid N₂ and stored at −80 °C until use.

Size-exclusion chromatography for SidJ:CaM:SdeC and SdeA complexes

For preparation of the reaction intermediate complex sample, tag-less proteins were preincubated for 15 minutes at 37 °C: 2.5 mg SidJ 97–851, 0.83 mg CALM2 and 2.5 mg SdeA/C^Core. Conditions within the preincubation were: 50 mM Tris 7.5, 50–75 mM NaCl, 10 mM MgCl₂, 4 mM ATP, 1 mM DTT, 300 μL. The reaction was spun at 22,000 × g for 10 minutes at 4 °C, diluted to 400 μL with size-exclusion buffer (25 mM Bis-tris pH 6.5, 100 mM NaCl, 1 mM TCEP, 2 mM MgCl₂, 1 mM ATP), and subjected to SEC on a Superdex S200 increase 10/300 column.

For SEC experiments, the same conditions were used, with less protein: 0.5 mg of each SidJ and SdeC and + 0.17 mg CALM2. ATP was omitted from the SEC buffer in SdeC samples. Preincubation was supplemented with 5 mM L-Glu for experiment with Glu treatment. Tris pH 7.5 was used for basic condition size-exclusion experiment shown in Figure 1C.

Cryo-EM grid preparation

The reaction intermediate complex sample was prepared as above, in size-exclusion buffer. All proteins were tag-less. Concentration of proteins within the complex fractions were assessed according to Bradford reagent (Bio-rad protein assay), and were ~1–1.5 mg/mL. The peak fraction was diluted to 0.35 mg/mL using size-exclusion buffer, and spun at 22,000 × g for 5–10 minutes at 4 °C. The sample used for grid preparation was never frozen and used within 1 hour after preparation. Grids were prepared using a FEI Vitrobot mark IV (ThermoFisher Scientific) at 4 °C with 95% relative humidity and plunged into liquid ethane. 3.5 μL of sample was applied to a Quantifoil 300-mesh R1.2/1.3 grid (Quantifoil, EMS), which was glow discharged using Pelco EasiGlo. The sample was preincubated with grid for 10 seconds before blotting.

Cryo-EM Data collection and processing

Prior to data collection, sample grids were screened on a Talos Artica microscope, and a pilot dataset of SidJ-SdeC-CaM was collected on a Titan Krios microscope at the Cryo Electron Microscopy Facility at UT Southwestern. Datasets of both SidJ-SdeA-CaM and SidJ-SdeC-CaM complexes used for final data processing were collected at the Pacific Northwest Cryo-EM Center (PNCC) on a Titan Krios microscope operating at 300 kV, with the post-column energy filter (Gatan) and a K3 direct detection camera (Gatan), using SerialEM(Mastronarde, 2005). Movies were acquired at a pixel size of 0.54 Å; in super-resolution counting mode, with an accumulated total dose of 50 e-/Ų over 78 frames. The defocus range of the images was set to be −0.8 to −2.5 μm. 11,694 and 9,409 movies were collected for SidJ-SdeA-CaM complex and SidJ-SdeC-CaM complex, respectively.

Image processing and 3D reconstruction

Unless described otherwise, all datasets were processed with Relion (Scheres, 2012). For the pilot dataset, the motion corrected micrographs from Relion were preprocessed with MicAssess (Li et al., 2020). Particles were picked with crYOLO (Wagner et al., 2019). A 4.5 Å map of SidJ:CaM:SdeC was obtained from the pilot dataset. This map was low pass filtered and used as the initial model for data processing as described below. For both datasets collected at PNCC, movies were aligned and summed with MotionCor2 (Zheng et al., 2017), with a downsampled pixel size of 1.08 Å. MicAssess (Li et al., 2020) was used to preprocess all motion corrected images. The CTF parameters were calculated using Gctf (Zhang, 2016), and images with estimated CTF max resolution better than 5 Å were selected for further processing. For the SidJ:CaM:SdeA dataset, 3,193,233 particles were picked using crYOLO (Wagner et al., 2019) from 9,839 images. 2D and 3D classifications identified two subclasses, which were combined with 386,460 particles. After further 3D refinement, CTF refinement and particle polishing, an additional step of alignment-free 3D classification with eight classes was performed. 310,154 particles from three subclasses with more complete maps for SdeA were combined to calculate the final map at 2.5 Å resolution. For the SidJ:CaM:SdeC dataset, 4,454,035 particles were picked using crYOLO5 from 8,432 images. 212,768 particles were selected after 2D and 3D classifications for subsequent 3D refinement, CTF refinement and particle polishing, followed by alignment-free 3D classification. 152,589 particles from the two best classes were used to calculate the final map at 2.8 Å resolution. Map resolution values were calculated by Relion with the gold standard FSC method.

Model building and refinement

For initial model building, the existing crystal structure of SdeA (PDBID: 5YIM), and cryo-EM structure of SidJ:CALM2 (6S5T) were fit into experimental density of SidJ:CaM:SdeA using Chimera. Models were manually adjusted in Coot to account for main chain rearrangements, compared to the crystal structure. Fit was further enhanced by rigid fit and morph routines in Phenix real space refine. Resultant model was split into SidJ:CaM and SdeA, and fit into SidJ:CaM:SdeC map using chimera. Sequence of SdeA was mutated in coot, to match SdeC sequence. Rearranged areas were manually adjusted. AceDRG from CCP4 7.1 suite (Winn et al., 2011) was used to generate AMPGlu link restraints. DeepEMhancer (Sanchez-Garcia et al., 2020) was used to generate post-processed maps, with the two unfiltered half maps from Relion, for further manual model building and inspection in Coot(Emsley et al., 2010). The models were then subjected to real space refinement in Phenix(Adams et al., 2010), with secondary structure restraints and non-crystallographic symmetry restraints. Model geometries were assessed with MolProbity(Chen et al., 2010) as a part of the Phenix validation tools (Table 1). Figures were rendered in PyMOL (The PyMOL Molecular Graphics System, Schrödinger), Coot or Chimera (Pettersen et al., 2004).

Production of SdjA antibodies

L. pneumophila SdjA 36–789 D480A was purified as above as a 6X-His-SUMO fusion and was cleaved off of Ni-NTA resin with Ulp1 protease (on-column cleavage). The protein was further purified by Superdex 200 16/600 gel filtration chromatography and used to inoculate rabbits for generation of rabbit anti-SdjA anti-serum (Cocalico Biologicals). Total IgG was partially purified by ammonium sulfate precipitation as described (Kent, 1999). In short, 1 mL NHS-activated HP columns (Cytiva, 17071601) were washed with 1 mM HCl, and crosslinked with the antigen by repeatedly passing 2 mL of 5 mg/mL protein in coupling buffer (0.2 M NaHCO₃, 0.5 M NaCl, pH 8.3) for 30 minutes, at ~1 ml/min. 15 mL of rabbit serum was centrifugated at 10,000 × g for 30 minutes. To 15 mL of serum, 35 mL of ice cold, saturated Ammonium Sulfate solution in BBS (Borate Buffered Saline) was added to precipitate the antibodies for 2 hrs at 4°C (400 g of Ammonium Sulfate dissolved in 500 mL of hot BBS; BBS: 0.2 M Sodium Borate, 160 mM NaCl, pH 8.0). The precipitate was spun at 10,000 × g for 30 min, and resuspended in 15 mL of BBS, and spun for 10 minutes at 10,000 × g to remove insoluble particles. Antibodies from solution were affinity-purified by passing it through a Hi Trap NHS-activated HP column, crosslinked with WT untagged SdjA 36–789, or SdeA 519–1100 (SdeA^ART domain).

Antibodies were eluted with 9 mL of 100 mM Glycine pH 2.5, into a tube with 1 mL of 1 M Tris 8.0 for neutralization, concentrated to ~1 mg/mL, aliquoted and stored at −20°C until use.

Glutamylation assays

Glutamylation reactions were performed in a reaction mixture containing 50 mM Tris-HCl pH 7.5, 50 or 75 mM NaCl, 1 mM DTT, 1 mM MgCl₂, 0.3 mM ATP, and 100 μM [U-¹⁴C] or [³H] L-Glu (specific radioactivity 200 cpm/pmol for ¹⁴C, 1500 cpm/pmol for ³H). For standard endpoint reactions, a typical 20 μL reaction contained 10 μg SidE effector, 1 ug SidJ/SdjA, and 2 μg CaM. Reactions were initiated by adding ATP/Mg²⁺, Glu and CaM, and were allowed to proceed at 37 °C for 60 minutes, after which they were terminated by addition of 7 uL of STOP mix (0.5 M EDTA, 5X SDS-PAGE loading dye; 2:5 ratio). Reaction products were resolved by SDS-PAGE and visualized by Coomassie staining. Gels were then soaked in EN3HANCE reagent for 30 minutes, followed by a 30-minute soak in H₂O, then dried. ¹⁴C incorporation was detected by autoradiography.

For kinetic assays, 0.1 ug of SidJ was incubated with 10 ug SdeA^FL for 15 minutes at room temperature (For SdjA, 0.6 ug of protein was incubated for 20 min with 5 ug of SdeC^Core). 100 μM [³H] L-Glu (specific activity of 1500 cpm/pmol). Triplicate 20 uL reactions were incubated for 15 min at room temperature and terminated as above. After SDS-PAGE separation, Coomassie stained SidE bands were excised, placed in glass scintillation vials and digested in 1 mL of 30% H₂O₂ at 60 – 75 °C overnight. After digestion was complete, vials were cooled down to ambient temperature, mixed with 12 mL of Budget-Solve scintillation cocktail (RPI, 111167) and ³H incorporation quantified by scintillation counting.

Adenylation assays

SdeA adenylation end-point assays using [α³²P]ATP were performed in a 20 uL reaction volume for 1 h at room temperature using 15 μg SidJ, 15 μg SdeA and 4.5 μg CALM2. Proteins were prediluted in dilution buffer (50 mM Bis-tris 6.5, 50 mM NaCl, 1 mM DTT). Reactions were initiated with ATP/Mg²⁺. The final reaction conditions were: 50 mM Bis-tris pH 6.5, 50 mM NaCl, 1 mM DTT, 150 μM [α³²P]ATP (SA: 1000 cpm/pmol), 1 mM MgCl₂, 1 mg/mL fatty acid-free BSA, −/+ 1 mM Glu. Reactions were stopped by addition of 500 μL of 30 mM ATP in 20% Trichloroacetic acid (TCA). Reactions were incubated on ice for 40 minutes, spun at 20,000 × g for 10 minutes at 4 °C and washed twice with 250 μL of 20% TCA solution. Radioactivity in the pellets was quantified by scintillation counting.

Differential Scanning Fluorimetry/thermal shift assay

Protein thermal stability in solution was monitored by measurement of fluorescence of SYPRO Orange dye (Thermo). Triplicate 25 μL reactions contained 5.5 μM SidJ^59–851 and 5.5 μM CALM2. Reaction conditions were: 10 mM Tris 7.5, 50 mM NaCl, 5 mM MgCl₂, 1 mM ATP and 625x dilution of SYPRO Orange dye (Invitrogen, S6650).

Reactions were performed in domed PCR tubes in a BioRad CFX96 thermocycler. Samples were subjected to a gradient of temperature 25–85 °C at a rate of +1 °C/minute. Fluorescent signal was read using FRET mode of the thermocycler. Melting curves were normalized within each replicate, averaged and plotted in Prism 9. The temperature at the derivative peak as reported by BioRad CFX manager, was used as the melting temperature.

SidE Ubiquitination Assay in HEK293A Cells

HEK293A cells were collected and plated into individual wells of a 24-well dish at near confluency. The following day, individual wells were transfected with pcDNA-HA-Ub, pcDNA-HA-Ub^GG/AA, pcDNA-SidJ-V5 (or mutants), pcDNA-Myc-SdeA -SdeB, -SdeC, or -SidE (or mutants), or empty vector using PolyJet transfection reagent (SignaGen) with a total of ~1 μg plasmid DNA per well: HA-Ub (0.5 μg): SidJ-V5 or V5-SdjA (0.5μg): myc-SdeA (0.02 μg), myc-SdeB/C/SidE (0.04 μg). DNA was premixed separately for each well and diluted with 25 μL of serum free (S.F.), antibiotic free DMEM. 25 μL of PolyJet solution in S.F. DMEM containing ~3.1 μL of PolyJet transfection reagent per well was mixed with DNA solutions and added dropwise to cells. The medium was replaced 5 hours after transfection with DMEM containing 10% FBS and Pen/Strep. ~16–20 hrs after transfection, cells were washed twice with ice cold PBS and lysed directly on the plate with 2X SDS-PAGE loading buffer (25 mM Tris-PO₄ pH 6.8, 20% (w/v) glycerol, 2.5% (w/v) SDS, 0.04% (w/v) bromophenol blue and 2% β-mercaptoethanol (β-ME). Samples were boiled, resolved by SDS-PAGE, transferred to a nitrocellulose membrane, and immunoblotted with the indicated antibodies.

SidE effector in-vitro Ubiquitin laddering and inactivation assays

Full-length SdeA, SdeB, SdeC and SidE were glutamylated in 20μL reactions containing 0.2 mg/mL SidE effector, 0.2 mg/mL, SidJ 59–851 (WT or D542A, D545A) or SdjA 36–789 (WT or D480A) and 0.1 mg/mL CALM2 in 50 mM Tris HCl pH 7.5, 50 mM NaCl, 1 mM DTT, containing 1 mM glutamic acid. Some reactions contained no SidJ/SdjA or contained catalytic mutants of each SidE effector to inactive ART activity (SdeA^E860Q, SdeB^E857Q, SdeC^E857Q and SidE^E855Q). Reactions were initiated by addition of ATP to 1 mM and MgCl₂ to 5 mM and incubated for 60 min at 37°C. Reactions were then diluted 1:10 (for a final concentration of 0.02 mg/mL SidE effector) into fresh tubes containing 50 mM Tris HCl pH 7.5, 150 mM NaCl, 1 mM DTT, 5 mM EDTA and 0.15 mg/mL HA-UbB and 100 μM NAD+ or 0.15 mg/mL ADPR-HA-UbB. The phosphoribosyl-ubiquitination assay was conducted at 37°C for 10 min (for SdeA, SdeB, and SdeC) or 15 min (for SidE). Reactions were terminated by addition of SDS loading buffer with β-ME and boiling.

Bioinformatics

For exploration of sequence diversity among SidJ homologs, a homolog set was collected using five iterations of PSI-Blast on the NR database. In the resulting sequence set, redundancy was removed at 70% sequence identity threshold using CD-hit (Huang et al., 2010). The sequences were aligned with Mafft (Katoh et al., 2019), and in-house script removed columns with gaps in the original SidJ (L. pneumophila) sequence. The sequence logos were built using the Weblogo server (Crooks et al., 2004). A phylogenetic tree was built for selected SidJ-like kinase domains, using the PhyML Maximum Likelihood method (Dereeper et al., 2008) with an Approximate Likelihood Ratio Test for branch support, and visualized with iToL (Letunic and Bork, 2016).

Quantification and Statistical Analysis

Quantification of biochemical data was performed by scintillation counting of triplicate or quadruplicate samples, as described in individual methods for respective assays. GraphPad Prism 9 was used to calculate the means, normalize the data, plot the graphs, and to conduct T-tests. All error bars represent standard deviation. Value of n (number of replicates) for each experiment is indicated in the figure legends. No methods were used to verify statistical assumptions.

Supplementary Material

1

Table S1. List of primers and DNA oligonucleotides used in the study (Related to STAR Methods).

Highlights:

• SidJ forms a stable reaction intermediate complex with the SidE effectors

• Structures of two reaction intermediate complexes were determined by cryo-EM

• Two active sites in SidJ facilitate adenylation and glutamylation

• SdjA is a glutamylase that differentially regulates the SidE effectors.