A single Legionella effector catalyzes a multi-step ubiquitination pathway to rearrange tubular endoplasmic reticulum for replication

Summary

Intracellular pathogens manipulate host organelles to support replication within cells. For Legionella pneumophila, the bacterium translocates proteins that establish an endoplasmic reticulum (ER)-associated replication compartment. We show here that the bacterial Sde proteins target host reticulon 4 (Rtn4) to control tubular ER dynamics, resulting in tubule rearrangements as well as alterations in Rtn4 associated with the replication compartment. These rearrangements are triggered via Sde-promoted ubiquitin transfer to Rtn4, occurring almost immediately after bacterial uptake. Ubiquitin transfer requires two sequential enzymatic activities from a single Sde polypeptide: an ADP-ribosyltransferase and a nucleotidase/phosphohydrolase. The ADP-ribosylated moiety of ubiquitin is a substrate for the nucleotidase/phosphohydrolase, resulting in either transfer of ubiquitin to Rtn4, or phosphoribosylation of ubiquitin in the absence of a ubiquitination target. Therefore, a single bacterial protein drives a multistep biochemical pathway to control ubiquitination and tubular ER function independently of the host ubiquitin machinery.

Introduction

Legionella pneumophila is an intravacuolar pathogen of both humans and amoebae (Rowbotham, 1980). As the agent of Legionnaires’ disease, infection is initiated by inhalation of contaminated water sources, followed by bacterial growth within alveolar macrophages (Copenhaver et al., 2014). L. pneumophila utilizes its main virulence factor, a type IVB secretion system (T4SS) known as Icm/Dot, to translocate ≥300 proteins into the host cytosol, establishing an endoplasmic reticulum (ER)-associated Legionella containing vacuole (LCV) (Swanson and Isberg, 1995). Bacteria lacking the T4SS are unable to form an LCV (Wiater et al., 1998). The formation of this intracellular Legionella ER compartment is evolutionarily conserved from amoebae to mammals (Abu Kwaik, 1996, Berger and Isberg, 1993).

The ER is from from membrane tubules and flattened sacs that can be classified into the perinuclear, ribosome-associated ER sheets and tubular ER (Voeltz et al., 2006). Recent high-resolution studies of the ER reveal that these classifications are an oversimplification of the breadth of ER structures, as architecture formerly described as peripheral sheets are instead composed of cross-linked ER tubules, termed ER matrices Nixon-Abel et al., 2016). The tubular ER is a vast network of elongated cylinders, enriched in structural ER membrane proteins such as Dp1/Yop1p and the reticulon family (English et al., 2009). Reticulons (Rtns) are evolutionarily conserved from yeast to humans (Yang and Strittmatter, 2007), with four subfamilies in mammalian cells (Yan et al., 2006). Reticulon 4 (Rtn4), also known as neurite outgrowth inhibitor (NOGO) is a highly abundant ER protein with three isoforms (Rtn4a/Nogo-A, Rtn4b/Nogo-B1, and Rtn4d/Nogo-B2), at least one of which is expressed in most mammalian cells (Yang and Strittmatter, 2007). Rtns generate ER curvature through their two conserved hydrophobic hairpins inserted in the cytoplasmic leaflet of the lipid bilayer (Zurek et al., 2011). Homo- and hetero-Rtn oligomers are believed to establish arc-like scaffolds (Zurek et al., 2011).

Previous studies hypothesized that the LCV acquires ER-associated markers by hijacking ER vesicles destined for the Golgi (Tilney et al., 2001, Kagan and Roy, 2002), but a recent report indicates that this may be preceded by association with tubular ER (Haenssler et al., 2015). The demonstration that the LCV acquires phosphoinositide-4-phosphate prior to acquisition of vesicular ER markers (Weber, et al., 2014), further argues for early association of peripheral ER. Therefore, interaction with peripheral ER tubules may represent the first step of ER association involved in LCV biogenesis. Many Icm/Dot translocated substrates (IDTS) control membrane trafficking and immune function by post-translational modifications (PTMs), which regulate protein stability, localization, and enzymatic activities (Ribet and Cossart, 2010, Zhou and Zhu, 2015). Enzymatic PTM by bacterial proteins is a commons strategy used by a wide swath of bacterial pathogens, including ADP-ribosylation (ADPr) and ubiquitination (Ribet and Cossart, 2010, Ravikumar et al., 2015, Michard and Doublet, 2015). Although there is deep insight into how PTM can regulate the activity and stability of targets, little is known regarding how bacterial-induced modifications could induce changes in quaternary interactions in the targeted proteins (Pieters et al., 2016).

The Sde family is a group of IDTS that edit host proteins (Sheedlo et al., 2015, Qiu et al., 2016). L. pneumophila encodes multiple gene paralogs, with the Philadelphia 1 strain having four, three of which are organized in a single contiguous locus (sdeA, sdeB, and sdeC) (Bardill et al., 2005). Members of the protein family are all 170+kDa proteins that contain an N-terminal deubiquitinase (DUB) and a central domain similar to mono ADP-ribosyltransferases (ART) (Sheedlo et al., 2015, Qiu et al., 2016). Loss of the entire Sde family results in defective L. pneumophila intracellular growth within amoebal hosts (Bardill et al., 2005, Jeong et al., 2015, Qiu et al., 2016). Sde proteins are able to ubiquitinate several ER-associated Rab proteins, dependent on their ART domain (Qiu et al., 2016), with ubiquitination occurring independently of host Ub machinery (Pickart and Eddins, 2004). Although there is a connection between the ART domain and ubiquitination, ADP-ribosylation of mammalian substrates by full-length Sde members is not observed. Mass spectrometry of reactions containing a SdeA ART domain fragment result in ADPr of Ub at residue 42, but the role of the ART domain in this particular ubiquitination mechanism remains cryptic, particularly because the full-length protein shows no evidence of this activity (Qiu et al., 2016).

In this study, we analyzed early events in the interaction between L. pneumophila and host cells. We show that Sde family members modulate tubular ER function by catalyzing a biochemical pathway in which ADPr of Ub provides a substrate for a nucleotidase/phosphohydrolase, promoting transfer of Ub to Rtn4, resulting in dramatic ER reorganization. An independent study has similarly identified this single peptide-catalyzed biochemical pathway (Bhogaraju, et al., 2016). That study argues that the modified Ub is involved in disrupting the host ubiquitination system

Results

Sde family promotes Rtn4 rearrangement

To probe association of tubular ER with the LCV, mammalian cells were challenged with L. pneumophila and analyzed by immunofluorescence microscopy. As previously shown (Haenssler et al., 2015), detergent-resistant reticulon 4 (Rtn4) formed a reticular network proximal to the LCV by 40 min post infection (hpi), which then condensed circumferentially by 8 hpi in Triton X100-permeabilized samples (Fig. 1A, B). Our inability to detect the tubular reticulon network throughout the cell was investigated by changing permeabilization conditions. Consistent with previous reports (Haenssler et al., 2015), Rtn4 colocalization with the LCV required a functional T4SS (Fig. 1C). Strikingly, staining of the entire Rtn4 network could only be detected in the absence of detergent extraction, while colocalization of Rtn4 with the LCV persisted even in the presence of 5% SDS. Therefore, the Rtn4 associated with the LCV resulted from structural changes that distinguish it from the cellular pool of Rtn4 (Fig. 1C). To determine if these structural changes could be detected on SDS gels, cells were challenged with the WT or dotA3 strains, and SDS extracts were analyzed. A high molecular weight (HMW) Rtn4 species was identified in extracts from infections with the WT strain, while another ER membrane protein, calnexin, showed no altered migration (Fig. 1D).

Fig. 1. Sde family members promote Rtn4 rearrangements in response to L. pneumophila challenge.

(A,B) Bone marrow derived macrophages (BMDMs) from A/J mice were challenged, followed by fixation, permeabilization with 0.1% Triton X100 and probing with α-Rtn4 (green), α-L. pneumophila (red), and Hoechst (blue). Scalebar: 5 µm. (C) BMDMs challenged for 1 h with LP02 (WT) or an icm/dot⁻(dotA3) mutant were fixed, permeabilized as noted, and probed. Scalebar: 5 µm. Arrows indicate location of bacterium within infected cells. (D,E) Altered electrophoretic migration of Rtn4 after L. pneumophila challenge. HeLa cells were challenged for 2 h, solubilized in SDS at room temperature, fractionated by SDS-PAGE, and probed. Lanes: Un, uninfected. (F) Sde family members result in Rtn4 electrophoretic variants. 40–46 h after transfection into Cos1 cells of noted plasmids, cells were extracted, gel fractionated and blots probed with α-Rtn4. (G) The chromosomal arrangement of the sde genes. (H) Rtn4 rearrangements in BMDM dependent on presence of Sde family members at 1 hpi (See Fig. S1A). Scalebar: 5µm. (I) Deletion of sde family (KK099) prevents Rtn4 rearrangements about the LCV. BMDM were challenged for 1 h prior to probing as in panel A. (J,K) Sde family members promote immediate Rtn4 rearrangements after host cell contact. Cos7 cells harboring Rtn4b-GFP were challenged with L. pneumophila and images from live cells were captured over a 10 min period (See Movies S1 & S2). Scalebar: 5µm. Images displayed at 1.15X the captured sizes. See Fig. S1; Tables S1 and S2.

To decipher which L. pneumophila T4SS substrate was responsible for altering Rtn4, cells were challenged for 2 h with L. pneumophila lacking Ceg9, a T4SS substrate that interacts with Rtn4 (Haenssler et al., 2015), or an L. pneumophila strain lacking 12.7% of the genome (Δpent; (O’Connor et al., 2011). Both mutant strains generated the Rtn4 HMW species (Fig. 1E). Based on this result, we performed a transfection screen to assay the effect of individual L. pneumophila proteins on Rtn4, by selectively screening gene candidates encoded by these mutants. Plasmids encoding individual GFP-tagged L. pneumophila T4SS substrates were transfected into mammalian cells, then cell extracts were fractionated by SDS-PAGE (Fig. 1F). Of the >60 L. pneumophila substrates examined (Table S2), three members of the Sde family, SdeC, SdeB, and SdeA, induced an HMW Rtn4 species. In addition, a modified form of Rtn4 that migrated just above the abundant Rtn4b/d monomer (~50kDa) was observed (Fig. 1F; modified). These three large L. pneumophila T4SS substrates are organized in a contiguous locus with lpg2154 and sidJ, a known regulator of SidE family function (Fig. 1G).

To determine if L. pneumophila lacking the sde family was capable of inducing colocalization of detergent-resistant Rtn4, cells were challenged for 1 h with L. pneumophila and analyzed by immunofluorescence microscopy (Fig. 1H). More than 70% of wild type (WT) LCVs were associated with Rtn4, while no colocalization was observed with dotA3. Both a complete sde family deletion (KK099, ΔsdeC ΔsdeB-A ΔsidE; Table S1) (Jeong et al., 2015), and a sde locus deletion (KK034, ΔsdeC-A; Table S1) were unable to induce Rtn4 association (Fig. 1H; Fig. S1A). Expression of plasmid-encoded SdeC or SdeB was able to completely restore Rtn4-LCV association to WT levels in either sde deletion backgrounds, while there was partial restoration with SdeA (Fig. 1I).

To survey the dynamics of Rtn4-LCV association in real-time, Cos7 cells were transfected with Rtn4b-GFP, then challenged with WT L. pneumophila or Δsde strains expressing mCherry and monitored (Movies S1, S2). The Rtn4b-GFP signal illustrates a high-resolution outline of the ER network, which strongly contrasts with the poor resolution of endogenous Rtn4 in micrographs after concentrated detergent extraction (Fig. 1J, K; Movies S1, S2). In response to the WT infection, an Rtn4 signal intensified around the vacuole membrane, then dramatically nucleated outward from the LCV in Rtn4-rich tubular protrusions, some of which formed junctions with other protrusions (Fig. 1J; Movie S1). In a Δsde challenge, there was no observable change in Rtn4 localization (Fig. 1K; Movie S2).

Sde-dependent ER rearrangements result in pseudovesicles

To perform high resolution probing for Sde-mediated ER changes, an Rtn4b-APEX2-GFP fusion protein was generated (Fig. S1B). This fusion allows Rtn4 localization to be determined by coupling the protein to an engineered peroxidase reporter, which can be detected by transmission electron microscopy (TEM) after addition of the substrate diaminobenzidine (DAB). Cells were transiently transfected with an Rtn4b-APEX2-GFP fusion (Fig. S1B) then challenged with the WT or Δsde strains expressing mCherry for 1 h, and analyzed for deposition of DAB by microscopy. Bright-field microscopy revealed strong DAB depositions associated with WT LCVs, mimicking the Rtn4b structures seen previously by fluorescence microscopy (Fig. S1B).

One hour after infection, TEM images of WT revealed vesicle-mimicking structures (termed pseudovesicles) with dense DAB deposition about their surface, as well as projections extending out from pseudovesicles (Fig. 2A,B). LCV membranes adjacent to the WT were darkly stained by DAB depositions, indicating high levels of Rtn4 contiguous with the vacuole membrane (Fig. 2A). In contrast, vacuole membranes encompassing the Δsde mutant had little evidence of pseudovesicles or DAB staining (Fig. 2C,D). To determine if pseudovesicular structures were generated by the fusion protein, TEM of bone marrow-derived macrophages (BMDMs) challenged with WT L. pneumophila was performed. Analogous pseudovesicular structures as well as linear projections from these structures were observed surrounding the LCV (Fig. 2E; Fig. S2). There is strong precedence for this observation, as pseudovesicular structures occurring immediately after infection have been observed numerous times (Table S3). There were no pseudovesicular structures in BMDMs challenged for 1 h with L. pneumophila mutants lacking sde. Instead, the LCV was associated with long membranous structures that resembled irregularly stacked rough ER sheets, indicating premature association of rough ER with the LCV (Fig. 2F; Fig. S3). Rough ER association with the LCV during the earliest phase of infection has rarely, if ever, been observed in the literature (Table S3), and was only occasionally observed in micrographs of cells challenged with WT (Fig. S2D).

Fig. 2. Sde-dependent ER rearrangements generate Rtn4-staining pseudovesicles or linear stacks.

(A–D): Cos7 cells harboring Rtn4b-GFP-APEX2 (See Fig. S1B) challenged for 1 h with either LP02 (A,B) WT or (C,D) the Δsde strain (KK099), subjected to DAB staining followed by TEM. Panels A,B are TEM images of different sections from the same cell. Panel B is a high magnification image of Rtn4-rich region abutting bacterium that can be seen in panel A. Arrows point to (A) membranes in direct apposition to the LCV or (B) projections of Rtn4-associated membranes. Panels E,F: BMDM challenged for 1 h with either (E) LP02 (WT) or (F) Δsde strain (KK034). (E(inset)) Boxed area at higher magnification. Arrowhead points to projections from round structure. (F) Stacks of ER surrounding Δsde strain (KK034). (See Fig. S1, S2, and S3; Table S3).

Sde family members induce Rtn4 ubiquitination

We next investigated the nature of the Rtn4 modification in response to Sde proteins (Fig. 1F, modified). GFP-SdeC was transiently transfected into cells to produce the modified Rtn4 species, and the modified Rtn4 species was excised from SDS gels for LC-MS/MS analysis (Fig. 3A). The modified Rtn4 sample showed almost complete coverage of Ub, although the classic Gly-Lys isopeptide diagnostic of ubiquitination could not be detected (Table S4). Several peptides present in the control Rtn4 monomer samples were absent from the modified Rtn4 species, consistent with those peptides containing residues targeted by the modification (Table S4). Furthermore, the migration of the modified Rtn4 was consistent with monoubiquitination (8.5 kDa). To confirm Ub modification of Rtn4, HA (hemagglutinin)-tagged Ub was transiently co-expressed with GFP-SdeC, then transfected cell extracts were subject to Rtn4 immunoprecipitation (IP). Eluates of immunoprecipitates from SdeC-transfected cells revealed two prominent HA positive bands above 50kDa (Fig. 3B, compare E lanes). The migrations of the two species were consistent with single and double Ub modification of Rtn4, although the higher MW species could not be detected by silver stain analysis. No ubiquitination of Rtn4 was observed in eluates from GFP control transfections (Fig. 3B, E lanes).

Fig. 3. Sde family members promote Rtn4 ubiquitination.

(A) GFP-SdeC or GFP (vector) were transiently expressed in HeLa cells for 24 h, followed by Rtn4 IP. Eluates were fractionated by SDS-PAGE and stained. (B) HA-Ub was transiently co-expressed with either GFP or GFP-SdeC in HeLa cells for 24 h, then subjected to IP with α-Rtn4 IP, fractionated by SDS-PAGE, and probed for Ub-modified Rtn4 with α-HA (E=eluate, FT=Flowthrough, T=Total). (C) HEK293T cells were transiently transfected with HA-Ub for 24 h, the cell culture medium was replaced with 10µM MG132 (Millipore) medium 30–60 min. prior to challenge with L. pneumophila and the infection was allowed to proceed (MPI, minutes post infection), prior to IP with α-Rtn4. (D) Domain structure of Sde family proteins, with endpoints noted for SdeC (Qiu et al., 2016). See Table S4.

To analyze if Rtn4 ubiquitination occurs during L. pneumophila infection, cells were transiently transfected with HA-Ub and challenged with L. pneumophila, then subjected to Rtn4 IP. Immunoprecipitates from WT challenge predominantly resulted in ubiquitination of the smaller Rtn4b isoform within 10 min of infection (Fig. 3C, WT). By 3 h, both Rtn4b and Rtn4d isoforms were robustly monoubiquitinated, with evidence of HMW ubiquitinated forms (Fig. 3C, WT, 180 MPI). The absence of the Sde family resulted in the complete loss of Rtn4 ubiquitination, similar to mock-infected cells (Fig. 3C, vector, mock). The Δsde strain harboring SdeC showed complementation, albeit inefficiently, with evidence of ubiquitinated HMW forms, whereas complementation with either SdeB or SdeA was robust, producing substantial Rtn4 mono- and multi-ubiquitination (Fig. 3C). The pattern of Rtn4 ubiquitination in these strains was broadly reminiscent of a WT infection, with an abundance of detectable mono- and di-Ub modified Rtn4.

Rtn4 reorganization requires ART activity

Sde family proteins have been reported to contain a conserved arginine-mono ADP-ribosyltransferase (ART) domain necessary for Rab ubiquitination, although the ART activity could not be detected in full length WT protein (Qiu et al., 2016). To probe the connection between the ART activity and ER reorganization, cells were challenged with a panel of Sde ART domain point mutants and analyzed for Rtn4-LCV association (Barth et al., 1998). The predicted enzymatic residues were mutated to alanine in each Sde member (Fig. 3D: for SdeB/C, E859A or R763A; for SdeA, E862A). Each Sde ART mutant was completely unable to restore Rtn4 association with the LCV, consistent with the ART being essential for Rtn4 reorganization (Fig. 4A,B; Fig. S4A).

Fig. 4. Sde family mono ADP-ribosyltransferase activity is required for Rtn4 restructuring and ubiquitination.

(A) A/J BMDMs were challenged for 1 h, fixed, permeabilized with 1% Triton X100 and probed with α-Rtn4 (green), α-L. pneumophila (red), and Hoescht (blue). 50 L. pneumophila vacuoles were assessed for Rtn4 colocalization per coverslip. (B) Representative micrographs of Rtn4 association with the LCV at 1hpi (hour post infection), scale bar=5µm. (C) HEK293T cells transiently transfected with HA-Ub were challenged with L. pneumophila, and extracts were subjected to IP with α-Rtn4. Eluates were analyzed for Ub by probing with α-HA. (D) HEK293T extracts were incubated at 37°C with 10nM recombinant SdeC, and 20µM recombinant human HA-Ub monomer. Reactions were separated by SDS-PAGE and probed for α-εAdo (ADPr), and α-HA (Ub). (E) SdeC was incubated with K63-linked Ub tetramers at 37°C. Reactions were fractionated by SDS-PAGE, and assayed for altered migration and ADPr of Ub, by silver staining and immunoblotting. Lanes; WT: WT SdeC; C118S: DUB mutant; E859A: ART mutant; ---: No SdeC; WT(No HA-UB): WT SdeC No HA-UB added (See Figs. S4 and S5).

We next evaluated the role of the ART in ubiquitination of Rtn4 after L. pneumophila challenge. WT L. pneumophila promoted Rtn4 ubiquitination after infection whereas the Δsde mutant was clearly defective (Fig. 3C, Fig. 4C). Expression of WT SdeB in a Δsde background was able to complement Rtn4 ubiquitination, whereas the ART mutant, SdeB R763A, was indistinguishable from a Δsde mutant infection (Fig. 4C, compare R763A to SdeB WT).

We hypothesized that any ADPr modifications were either inefficient or unstable. To address this problem, we devised an ADPr assay that exploited an analogue of β-NAD, ethenoNAD (εNAD) (Klebl et al., 1997), in which ADPr of substrates could be monitored by Western blotting with α-ethenoadenosine (α-εAdo) (Krebs et al., 2003). To assay for SdeC ART activity directed against mammalian proteins, while simultaneously monitoring cellular ubiquitination changes in response to Sde proteins, recombinant full-length SdeC was incubated with cell extracts and recombinant HA-Ub in the presence of εNAD. A 5 min reaction with either WT SdeC or a DUB-defective derivative resulted in robust laddering of ADPr-substrates, in a pattern reminiscent of polyubiquitin chain laddering (Fig. 4D, Fig. 5 min, WT and C118S). By 60min, evidence for the εAdo signal was greatly reduced, with the only remaining signal being above 250kDa (Fig. 4D, 60 min, WT and C118S). Probing with α-HA revealed that both SdeC_WT and SdeC_C118S induced robust HA-Ub polymerization, but the Ub polymerization was unchanged over time (Fig. 4D, lower panel). In contrast, there was no ADPr or HA-Ub polymerization by the SdeC_E859A ART mutant (Fig. 4D). In the absence of HA-Ub, a HMW species above 250kDa was recognized by α-εAdo and disappeared over time (Fig. 4D upper panel, WT-No HA-UB). These results indicate that SdeC promotes ubiquitination of host proteins dependent on the ART domain.

Fig. 5. Sde family NP domain is required for Rtn4 rearrangements, intracellular growth, and functions cooperatively with ART domain to conjugate Ub.

(A,B) A/J BMDM were challenged for 1 h followed by fixation, permeabilization with 1% Triton X100, and probed as in Fig. 4A–B. L. pneumophila vacuoles were assessed for Rtn4 colocalization (See Fig. S4A). (B) Representative micrographs of Rtn4 association with the LCV from part A, scale bar 5 µm. (C) Dictyostelium discoideum was challenged with WT (Lp02) or mutant L. pneumophila expressing luciferase (PahpC::lux). L. pneumophila intracellular growth (luminescence) was monitored hourly. Mean ± SEM for every 5 h increment; results representative of ≥3 replicate experiments (See Fig. S4B–D). (D) HEK293T extract was incubated at 37°C for the indicated time with recombinant SdeC, εNAD, and recombinant human HA-Ub. Rtn4 Ub and εADPr were assessed by immunoblot with indicated antibodies. (E) Recombinant Ub or poly-His-Ub monomers were incubated with εNAD and recombinant SdeC at 37°C for the indicated time. Ub ADPr was assessed as in panel D. (F) A four component system is sufficient to ubiquitinate Rtn4. Purified GST-Rtn4 was incubated with noted components and SdeC derivatives for 1 h. Proteins were fractionated by SDS-PAGE and visualized by silver stain (See Fig. S6). Lanes: WT, WT SdeC; E859A, ART mutant; H416A, NP mutant (See Figs. S4 and S6).

To simplify the εADPr assay, we excluded cell extracts, using recombinant polyubiquitin as a substrate. Recombinant K63-linked Ub tetramers and εNAD were incubated at 37°C with SdeC_WT for various times before termination. Immediate introduction of SDS buffer resulted in the appearance of a prominent ADP-ribosylated Ub (ADPr-Ub) tetramer (Fig. 4E, 0 min). This signal was dramatically reduced, however, if incubations were allowed to continue (Fig. 4E, α-εAdo; Fig. S5). This indicates that ADPr modification by SdeC is rapid and transient, explaining why previous studies were unable to detect the modification with full-length protein (Qiu et al., 2016). The ADPr signal required an intact ART domain as no εADPr signal was observed in SdeC_E859A reactions (Fig. 4E, α-εAdo E859A; Fig. S5). On further analysis of the SdeC-modified poly-Ub on silver stained gels, it was clear that loss of the εAdo signal was not due to total reversal of the modification. After initial appearance of an ADPr signal, a slower migrating Ub tetramer relative to an unmodified Ub tetramer predominated (Fig. 4E, WT 0), and this form persisted without detectable change in migration, even as the ADPr signal disappeared (Fig. 4E, WT 120). This is consistent with Ub chains being ADPr-modified followed by additional processing retaining an unknown modification (Fig. 4E, WT silver stain). Therefore, the Sde family generated ADPr-Ub as an intermediate reaction species prior to loss of the εAdo epitope.

Another striking aspect of the ART activity was observed in these assays: although SdeC has an amino terminal deubiquitinase (DUB) domain, there was minimal cleavage of poly-Ub when SdeC was incubated with εNAD (Fig. 4E, WT, silver stain). When the assay was repeated using the SdeC ART mutant, the DUB activity was restored, with 120 min incubation resulting in an accumulation of mono-Ub (Fig. 4E, E859A, silver stain; Fig. S5). Therefore, the ART activity strongly interfered with the DUB activity, and this inhibition continued after the ADPr was processed.

Sde family Nucleotidase/Phosphohydrolase domain is required for ubiquitin conjugation and biological function

The Sde proteins contain a region between the DUB and ART domains (Fig. 3D) with sequence homology to the Legionella IDTS Lem10, which has been crystalized, revealing structural similarities to nucleotidases and other phosphohydrolases (Wong et al., 2015, Morar et al., 2015). We hypothesized that this nucleotidase/phosphohydrolase domain could be responsible for processing of ADPr-Ub, resulting in the loss of εAdo, while leaving a modification that retarded Ub migration. To explore this possibility, several potential NP catalytic residues in SdeC were selected for site-directed alanine mutagenesis based on sequence similarity to the nucleotide-binding pocket in Lem10, and introduced on plasmids into a Δsde background. The SdeC NP mutant SdeC_H416A was completely incapable of generating Rtn4 structures associated with the LCV after 1 h challenge, in contrast to the behavior of the SdeC_WT derivative, which fully restored Rtn4-LCV association (Fig. 5A, B).

To determine if the Sde NP domain, similar to the ART domain (Fig. S4B–D), was important for promoting intracellular replication in natural hosts, WT or mutant L. pneumophila strains expressing luciferase were used to challenge the amoebal species Dictyostelium discoideum, and replication was monitored over 4–5 days. A plasmid harboring SdeC_H416A in a Δsde strain could not restore L. pneumophila intracellular growth to levels observed with either the WT or the deletion strain harboring SdeC_WT. Instead, expression of the SdeC NP mutant mimicked the poor intracellular growth observed with Δsde infection (Fig. 5C). Therefore, the Sde NP domain is required for promoting bacterial replication during amoebal challenge.

To probe the biochemical role of the NP domain, ADP-ribosylation and Rtn4 ubiquitination were simultaneously monitored in cell extracts in the presence of recombinant SdeC derivatives. Reactions with SdeC_WT produced several altered Rtn4 migration forms consisting of ~9kDa shifts, consistent with the addition of one or more Ub moieties (Fig. 5D, α-Rtn4). These modified Rtn4 species were not observed in the absence of SdeC, or with addition of the ART mutant SdeC_E859A or NP mutant SdeC_H416A (Fig. 5D, α-Rtn4), indicating that the ART and NP domains collaborate for ubiquitination. When probed with α-εAdo, a HMW ADP-ribosylated species was apparent that dissipated over 1 h in SdeC_WT reactions. At 5min post SdeC addition, weak laddering of ADPr proteins was also observed, presumably due to modification of endogenous poly-Ub (Fig. 5D, α-εAdo). The SdeC_E859A mutant phenotypically mimicked reactions lacking SdeC, with the residual ADPr signal dependent on endogenous enzymes from the extract (Fig. 5D, α-εAdo). Strikingly, the SdeC_H416A construct, which showed no evidence of Rtn4 ubiquitination (Fig. 5D, α-Rtn4), was able to produce robust ADPr of numerous cell extract proteins, including a protein that migrated at the size predicted for Ub (Fig. 5D, α-εAdo). Therefore, the presence of persistent ADPr modification in the NP mutant negatively correlated with Rtn4 laddering.

As the SdeC_H416A NP domain mutant appeared to cause accumulation of ADPr-Ub, the effect of the NP domain on the modification of Ub was analyzed in an in vitro system free of cell extract. SdeC_H416A was able to robustly ADP-ribosylate both polyhis-tagged and untagged Ub over a 60min reaction, with no loss of the εAdo signal (Fig. 5E). In contrast, WT SdeC showed a weak ADPr-Ub signal after only 1 min, and by 2 h the ADPr-Ub was undetectable, indicating that the NP activity efficiently removed εAdo (Fig. 5E, α-εAdo). These results indicate that even in the absence of a target to ubiquitinate, both the ART and NP domains collaborate to post-translationally modify ubiquitin, transitioning from ADPr-UB to a second modification which lacks the εAdo epitopes.

To demonstrate that Ub modification of Rtn4 by SdeC occurs catalytically on a natural substrate in the absence of any host components, SdeC derivatives were incubated with a 20X molar excess of GST-Rtn4 in the presence or absence of εNAD. Impressively, within 1 hour, nearly the entire Rtn4 population was mono- or multi-ubiquitinated, with resulting species migrating ~8k – 24kDa larger than GST-Rtn4, and no detectable modification of unfused GST (Fig. 5F; Fig. S6, α-Ub). Ubiquitination of Rtn4 required both the NP and ART domains, as neither SdeC_H416A nor SdeC_E859A could promote Rtn4 ubiquitination. The loss of either NP or ART activity, however, could be overcome by mixing the two mutant proteins together in the presence of GST-Rtn4, with extremely efficient ubiquitination after 1 h (Fig. 5F). These results are consistent with ADPr-Ub being a substrate of the SdeC NP domain, in which trimming of ADPr and transfer of Ub to Rtn4 requires the NP activity.

A SdeC-promoted biochemical pathway leads to ribose-monophosphate modified ubiquitin

Our results argue that Ub conjugation to Rtn4 by the Sde family is a consequence of covalent modification of Ub followed by enzymatic processing and transfer to targets. To understand the nature of the transient Ub-modified intermediate and its apparent trimming, monomeric Ub was incubated with SdeC derivatives in the presence of εNAD and analyzed by liquid chromatography-mass spectrometry (LC-MS). In the absence of SdeC, the molecular mass of monomeric Ub was 8564.57 Da, but after 1 h incubation with the SdeC_H416A mutant, ≥90% of Ub population increased by a mass of 565.05 to 9129.62 Da (Fig. 5E, Fig. 6A), consistent with a single εADP-ribose moiety added. Incubation of SdeC_WT with Ub, on the other hand, resulted in ≥90% of the Ub population converted to 8776.5 Da (Fig. 5E, Fig. 6A). This 212 Da mass increase is consistent with ribose-monophosphate modification of Ub, as a consequence of cleavage at the diphosphate bridge between adenosine and ribose (Fig. 6A,B). Therefore, in the absence of a Ub recipient, the ART domain recognizes and modifies Ub, followed by diphophohydrolase processing to ribose-monophosphate by the NP domain (Fig. 6B). The proposed reaction is similar to a subset of nucleotidases that show diphosphohydrolase activity toward ADPr-modified proteins (Daniels et al., 2015, Palazzo et al., 2015).

Fig. 6. ART and diphosphohydrolase-dependent ribose-monophosphate modification of Ub.

(A) SdeC in presence of εNAD results in a modification of 212 amu. Ub (black), Ub incubated with the NP mutant SdeC (H416A, gray) or SdeC (WT, red) were subjected to LC-MS analysis and the deconvoluted masses of the peaks for each sample displayed. (B) Proposed pathway to generated modification of 212 amu. (C,D) Trypsin/AspN treatment of modified Ub species followed by extracted ion chromatography (XIC) analysis reveals predicted modifications. Shown are XIC chromatograms of species having displayed m/z values for both major modifications displayed in panel B. (E) Treatment of species 3 with alkaline phosphatase results in a product predicted for ribosylated Ub. Ub was treated with noted enzymes, followed by LC-MS, and the deconvoluted masses of the peaks for each sample are displayed. (F) Likely products that lead to the generation of 132 amu modification. (G) Electrospray ionization MS/MS spectrum of trypsin/AspN Ub fragment having +212 amu modification resulting from SdeC treatment. The b-type ion fragments are displayed above the trypsin/AspN peptide that has an increase of 212.01 amu over the predicted size of the unmodified Ub peptide. Each predicted b-type ion was identified and displayed along with identified y-type ion fragments. Ions marked #212 denote fragment sizes that correspond to the predicted b-type ions having an added 212 amu.

To determine if the final reaction product is found on Ub residue R42 (Qiu et al., 2016), as predicted by diphosphodyrolase action on R42-εADP, the modified Ub species were gel extracted, subjected to trypsin/AspN double digestions and analyzed by liquid chromatography-tandom mass spectrometry (LC-MS/MS). The double digestion generated a 10 amino acid fragment with an expected m/z(+2) = 694.334 for the ribose-monophosphate modified form, and m/z(+3) = 580.91 for the εADPr modified form (Fig. 6C, D). When extracted ion chromatograms (XIC) were analyzed for the m/z(+2) expected for the ribose-monophosphate modified form, only Ub incubated with SdeC_WT could generate significant amounts (Fig. 6C). Similarly, XIC from the m/z(+3) predicted for the εADPr modified Ub fragment showed that only the diphosphohydrolase mutant SdeC_H416A could generate significant levels of this product (Fig. 6D).

To gain further evidence for a diphosphohydrolase activity, Ub was incubated with SdeC_WT for two hours, and then treated with alkaline phosphatase (AP) to remove the predicted phosphate group. The +212 Da modification by SdeC was reduced to a +132 Da modification (Fig. 6E, Ub+SdeC+AP), predicted for phosphatase processing to simple ribose (Fig. 6F). Therefore, in the absence of a ubiquination substrate, the ART and diphosphohydrolase collaborate to promote phosphoribosylation of Ub.

To conclusively demonstrate the proposed biochemical pathway, the AspN/trypsin 10 amino acid fragment spanning R42 was subjected to b- and y-ion analysis after LC-MS/MS. If the ribose-monophosphate modification occurs on R42, beginning with the b₄ ion, each of the successive ions should have an increase in mass of +212.01 (Fig. 6G, noted as R^#). We were able to identify ions with high resolution that matched the predicted b₄ through b₉ ions, each with the expected mass increase (Fig. 6G). In addition, we were able to identify an ion predicted to be the intact peptide with neutral loss of the modification (Fig. 6G, m/z=588.330), which has been observed in ribose phosphate-modified peptides previously (Palazzo et al., 2015). Therefore, SdeC ART activity, followed by diphosphohydrolase processing of ADPr occurs on the R42 residue.

Discussion

Previous work has shown that as early as 10 min post-bacterial challenge, vesicle-like structures approximately 200 nM in diameter associate with the LCV (Abu Kwaik, 1996, Kagan and Roy, 2002, Tilney et al., 2001). These structures have been called ER-derived vesicles, based on the fact that ER vesicle-associated proteins rapidly associate with the LCV (Kagan and Roy, 2002, Tilney et al., 2001). The work described here argues that these circular structures are derived from tubular ER as a consequence of a biochemical pathway catalyzed by Sde family members (Fig. 2A).

We propose that Rtn4 ubiquitination by the Sde biochemical pathway promotes structural transformations of ER tubules, potentially through enhanced Rtn4 oligomerization or generation of a scaffold to form tubule matrix-like structures (Nixon-Abell et al., 2016). For this reason, we have called these structures “pseudovesicles,” which have Rtn4-rich appendages extending from their cytoplasmic face. These appendages have not been a focus of interest in the field, but have been observed previously at early time points (Robinson and Roy, 2006)(Fig. 2B). Surprisingly, in the absence of the Sde proteins, rough ER sheets prematurely associated with the LCV (Fig. 2F; Fig. S3). An examination of 54 manuscripts from the literature indicates that rough ER recruitment typically occurs 6h post-infection of cells, and is never observed as an early event (Table S3). In fact, the overwhelming consensus among these studies is that the earliest event observed is the formation of round compartments with identical morphology to the pseudovesicles (Table S3; Abu Kwaik, 1996). The observed ER transitions indicate that the bacterium engineers these ER transformations in a temporal process.

Our work shows that a single bacterial protein catalyzes a unique multi-step biochemical pathway in the absence of host proteins that leads to ubiquitination and rearrangements of Rtn4. Bacterial and viral pathogens are known to directly subvert the Ub system through mimicry of eukaryotic Ub editing proteins, such as the Legionella SidE family deubiquitinases or the E3 ligases LubX and SidC (Qiu et al., 2016, Hsu et al., 2014, Horenkamp et al., 2014). It was previously demonstrated that the unique feature of Sde ubiquitination is that it occurs independently of the host Ub conjugation system, and is dependent on an ART domain (Qiu et al., 2016). We argue here that ADPr of the Ub R42 residue is merely the first step in a pathway that leads to direct conjugation of Ub to recipient host proteins. The model we favor is that the Sde pathway is initiated by highly efficient ADPr modification of Ub, which occurs catalytically at 0°C using a ratio of 1:50 SdeC:Ub, arguing for rapid ADPr modification of Ub immediately after Sde translocation. The ADPr-Ub is then used as a substrate by the NP domain that can either trim the ADPr to ribose-monophosphate or promote transfer of Ub to Rtn4.

Consistent with this model, we have shown that in a purified system using only Rtn4, Ub, and small quantities of SdeC, there is rapid conjugation of Ub, arguing that ADPr-Ub is an intermediate that is acted on by the NP domain to promote ubiquitination of substrates. This model is supported by the fact that when an NP-deficient mutant protein is mixed with an ART-deficient mutant, ubiquitination of Rtn4 is extraordinarily efficient (Fig. 5G, Fig. S6). We believe that during intracellular growth, the amino terminal DUB domain also plays a role (Sheedlo et al., 2015), perhaps by preventing accumulation of K63-linked Ub about the replication vacuole, thereby making available a local pool of mono-Ub for the action of the Sde family.

In contrast to the specific transfer reaction described here, phosphoribosylated Ub has the potential to serve as a nonspecific reactive intermediate able to undergo a non-enzymatic Maillard reaction that would conjugate Ub to recipient proteins through an irreversible sugar crosslink. In Maillard reactions, the electrophilic carbonyl of a reactive sugar, such as glucose or ribose-phosphate, reacts with a free amino group of recipient proteins, generating advanced end products (AGEs) (Chuyen, 2006). The kinetics of AGE formation is typically quite slow, so a Maillard mechanism would require some accessary factors that could allow a biologically relevant reaction. We favor a model in which the primary role of Sde proteins is to use the combined action of the ART and NP domains to ubiquitinate high specificity targets. Therefore, it is more likely that the formation of phosphoribose-Ub downmodulates Sde-mediated Ub conjugation, reducing the concentration of ADPr-Ub that can act as a substrate for Sde-mediated transfer to host target proteins. Consistent with this role, Sde overexpression in either mammalian or yeast cells results in inhibition of the host Ub system (Bhogaraju et al., 2016). Our results argue, however, that an inhibitory role of Sde proteins during intracellular growth may be of secondary importance. Inhibition of the host Ub system by SdeA requires the ART activity, and is independent of the NP domain (Bhogaraju et al., 2016). We have shown, however, that a L. pneumophila NP mutant that is competent for interfering with the Ub system (Bhogaraju et al., 2016) is as defective for intracellular amoebal growth as a total Sde family deletion (Fig. 5C).

The unusual nature of Sde-mediated cellular effects combined with its mechanism of action indicates that the functions of these proteins have a broad range of consequences. It has already been shown that a subset of Rab proteins can be ubiquitinated by SdeA (Qiu et al., 2016) in addition to the structural ER membrane protein demonstrated in this study (Fig. 5F). These results argue for multiple pools of specific targets The spectrum of cellular functions controlled by this protein family is likely to be quite large, with tubular ER rearrangement being the most rapid and visually spectacular response, controlling a morphological change that had previously been a mystery for much of the past two decades.

Experimental Procedures

Bacterial/Eukaryotic Culture, Antibodies, and Media

L. pneumophila derivatives used in this study were derived from Legionella pneumophila Philadelphia-1 strain (Berger and Isberg, 1993). Bone marrow-derived macrophages (BMDMs) were isolated from A/J mice (Swanson and Isberg, 1995). Bacterial strains, primers, plasmid construction, and challenge of mammalian cells are detailed in SI and Table S1. Animal protocols were approved by the Institutional Animal Care and Use Committee of Tufts University. Antibody sources/concentrations are detailed in supplemental information (SI).

Screen for Rtn4 altered electrophoretic mobility

Cos1 or HeLa cells were transiently transfected with pDEST53 or pDEST53 harboring GFP fusions to L. pneumophila proteins (Losick et al., 2010) using Lipofectamine 2000 (Life technologies) for 40–46h. Cells were collected in PBS and sample buffer (SB), lacking reducing agent, at room temperature. A list of the transfected Legionella GFP fusions is in Table S2.

Rtn4 Colocalization

Rtn4 colocalization with the LCV was assayed by immunofluorescence microscopy (Haenssler et al., 2015). Infected BMDMs were PFA-fixed, Triton X-100 permeabilized, stained with α-L. pneumophila rat serum and α-Rtn4. and detected with α-rat IgG Alexa Flour 594 and goat α-rabbit Alexa Fluor 488. Hoescht 33342 was used to label DNA.

Immunoprecipitations (IP)

Prior to Rtn4 IP, HEK293T cells transiently transfected with pMT123 (HA-Ub, gift of D. Bohmann and S. Lippard (Treier et al., 1994)) and 24h later, the transfection medium was replaced with DMEM+10%FBS with 10µM proteasome inhibitor MG132 (Millipore) 30–60min prior to L. pneumophila challenge at an MOI=10. Then cells were collected and washed in PBS and stored at −80°C until IP.

See SI for details of α-Rtn4 resin generation. For IPs, samples were lysed in 1% Triton X-100 for 20 min at 4°C. Cell debris was removed by centrifugation at 4°C, cleared lysates were diluted with equal volume detergent-free buffer and incubated with resin. Resin binding was allowed to proceed at 4°C for ≥4h, then washed ≥5X in buffer containing 0.1% Triton X-100 at 4°C. For elution, Rtn4 resin was incubated with 0.1M glycine (pH = 2.8) with 0.2% TritonX-100 for 5min in a spin column, centrifuged, then repeated, and neutralized with 0.5M Tris (pH =10.55).

In vitro DUB and ART/NP Assays

In DUB assays, 1µM recombinant Ub tetramers were incubated with 20nM SdeC and 100µM nicotinamide 1,N6-ε-adenine dinucleotide (εNAD) at 37°C in 1X ART buffer (see protein purification) for the indicated time. Reactions were terminated by the addition of reducing SB, and heated to 50–55°C for 20min. ADPr assays in 1X ART buffer included 10µM HA-Ub, Ub, or polyHis-Ub, 20nM SdeC, and 100µM εNAD unless otherwise indicated. Reactions were terminated by addition of reducing SB and boiling.

For ART/NP assays with cell extracts, HEK293T cells were harvested in PBS then lysed by Dounce homogenization and nuclear material removed by 3000 × rcf spins. Soluble extract was quantified and 100µg was added to reactions, which were terminated by freezing on liquid N₂, lyophilization, resuspension in 8M urea, and addition of reducing SB.

Rtn4 Ubiquitination in vitro

In 1X ART buffer, 400nM GST-HA-Rtn4 or GST, 10µM recombinant Ub, 20nM recombinant SdeC, and 100µM εNAD were combined and incubated for 1h at 37°C. then terminated by addition of reducing SB and boiling.

Intracellular Growth in Dictyostelium discoideum

D. discoideum strain Ax4 was cultured (Solomon and Isberg, 2000) and challenged with L. pneumophila lux derivatives at MOI=0.5 for 2h, then intracellular replication was monitored by luciferase production for ≥100h in a microtiter lumimometer at 25.5°C; see SI for details

Electron Microscopy

For analysis of Rtn4b localization at the LCV (Lam et al., 2015), Cos7 cells were plated in 35mm glass bottom No.2 uncoated gridded dishes (MatTek). The following day, cells were transfected with Rtn4b-APEX2-GFP. Cells were then challenged with L. pneumophila mCherry or GFP for designated times, washed with PBS, fixed with 2% gluteraldehyde solution ≥1h in cold PBS, then washed in PBS before further processing for transmission electron microscopy (TEM), detailed in SI.

Mass Spectrometry

Identification of Rtn4b/d-modification was performed after IP of Rtn4b/d from extracts of HeLa cells, fractionation on SDS-PAGE followed by silver staining, excision of Rtn4b/d electrophoretic variants that were trypsin-digested and subjected to liquid chromatography tandem mass spectrometry (LC-MS/MS); see SI methods. To determine molecular masses of Ub deriviatives, 10µM Ub monomer was incubated with 20nM SdeC_WT or SdeC_H416A and 100µM εNAD for 2h and 1h, respectively in ART buffer. Reactions were terminated in liquid N₂ and subjected to LC-MS analysis; see SI for details.

Tandem Liquid Chromatography Mass Spectrometry

LC-MS/MS analysis was performed by the Taplin Biological Mass Spectrometry Facility at Harvard Medical School, method details in SI.

Live Microscopy

Cos7 cells were seeded in 35mm glass bottom No.1.5 uncoated dishes (MatTek) and the next day cells were transiently transfected with Rtn4b-APEX2-GFP then placed on the stage of a Zeiss AxioObserver fitted with environmental and temperature controls set at 37°C, 5% CO₂. After 15 min equilibration, mCherry expressing L. pneumophila derivatives were introduced at MOI = 20. Imaging was initiated immediately on a single cell and grabbed every ~15 s for 10 min, see SI for details.