The Sde DUB activity is required for efficient Rtn4 rearrangements in response to L. pneumophila.
To determine if there is a link between the Sde family K63 deubiquitinase (DUB) domain (Fig. 1A) and biogenesis of the Legionella-containing vacuole (LCV), the kinetics of polyubiquitin (polyUb) association and Rtn4 rearrangements about the LCVs were analyzed. A previous study had shown that the absence of the DUB domain resulted in a 2-fold increase in polyUb about the LCV32. Bone marrow-derived macrophages (BMDMs) were challenged with L. pneumophila derivatives lacking either the complete complement of sde family members (sidE, sdeABC) or missing sdeABC (ΔsdeC-A), which removes all proteins known to control Rtn4 dynamics19. After 5 min centrifugation onto BMDMs, followed by fixation and probing with a polyUb-specific antibody (FK1), there was no evidence of polyUb about the LCV of WT strains (Fig. 1B; 0 minutes post infection (MPI)). In contrast, challenge with L. pneumophila missing all sde family members (Δsde) resulted in over 30% of the LCVs displaying polyUb association. By 10 MPI, ~ 15% of the vacuoles containing WT strain positively stained with polyUb, approximately 4X lower than the Δsde strain at this timepoint, consistent with previous observations (Fig. 1B; 32). At 0 MPI, Δsde mutants were efficiently complemented by all three members of the Sde family known to have biological activity (SdeA, SdeB and SdeC; Fig. 1B), although complementation of ΔsdeC-A mutants retaining SidE was weaker (Fig. 1C). Surprisingly, for all complemented strains, polyUb quickly rose to the levels seen in the absence of Sde family function (Fig. 1B,C; 30 MPI).
The ability to prevent polyUb association with the LCV at the earliest timepoint was dependent on the activity of the DUB domain, as the SdeC-C118S active site mutation was unable to block polyUb association at the earliest timepoint (Fig. 1C; 0 MPI). We conclude that inhibition of the polyUb association with the LCV occurs transiently and is dependent on the Sde deubiquitinase activity. The fact that the inhibition is transient is consistent with the LCV-associated polyUb chains becoming resistant to the action of the DUB domain by 30 minutes post-infection (MPI).
The biological significance of having a K63 deubiquitinase activity is unclear, especially because the ART activity of Sde proteins directly blocks deubiquitination, arguing that the DUB may play some role other than reducing LCV polyubiquitination 22, 42. Alternatively, the DUB activity could modulate the dynamics of Rtn4 rearrangements about the LCV22. To test this model, BMDMs were challenged with L. pneumophila derivatives having defective DUB activity followed by probing for Rtn4 detergent-resistant structures about the LCV (Figs. 1D,E) 22. Surprisingly, the inactive DUB mutant (C118S) reduced Rtn4 colocalization with the LCV. In strains expressing either SdeA, SdeB or SdeC as the sole demonstrated drivers of Rtn4 rearrangement, the C118S mutation reduced detergent-resistant Rtn4 association with the LCV by approximately 50% (Figs. 1D,E). Plasmid-borne SdeA showed lowered complementation levels (Figs. 1D,E), but SdeA also drove detergent-resistant Rtn4 structures that were not LCV-associated, indicating that SdeA overproduction may result in leakage of the protein from the LCV, as has been documented previously19.
Consistent with data in Fig. 1E, Rtn4 association with the LCV was robust in strains expressing SdeB or SdeC as the sole plasmid-expressed isoforms (Figs. 1I,J) and indistinguishable from the WT strain (Fig. 1F). In contrast, loss of sdeC-A prevented Rtn4 association with the LCV, as did elimination of the T4SS (Figs. 1G, H). The presence of the C118S mutation resulted in two populations, with a mixture of robust Rtn4-colocalization with LCVs or RTN4 staining below the level of detection (Fig. 1K). Similar results were seen with the DUB D80A catalytic mutant, whereas a control Cys mutant C293S showed high levels of colocalization (Fig. S1). These results are consistent with the DUB domain playing a role in LCV biogenesis at the level of controlling Rtn4 dynamics.
To test if the Sde DUB activity plays a direct role in promoting Rtn4 rearrangements, we determined if the catalytic domain contributed to the rate of Sde-dependent phosphoribosyl-ubiquitin (pR-Ub) modification of Rtn4. In a purified system, pR-Ub modification of Rtn4 is rapid and robust with the presence of excess mono-Ub22, 33. In contrast, infection provides multiple potential sources of Ub, including K63-linked polyUb, which the DUB activity could target to provide substrate for pR-Ub modification. To test this hypothesis, HEK293T cells transfected with HA-Ub were challenged by L. pneumophila harboring either WT or DUB-defective (C118S) derivatives as the only active Sde family members. At 10 or 180 MPI, cells were extracted and subjected to immunoprecipitation with anti-Rtn4, followed by immunoprobing with anti-HA to reveal Rtn4 ubiquitination (Fig. 1L). Detectable mono- and di-modification of the Rtn4B and Rtn4B2 isoforms were observed as early as 10 MPI with continued increase at 180 MPI in response to the WT strain (Fig. 1L, WT). In contrast, challenge with Δsde mutants showed HA-Ub modification that was indistinguishable from mock-infected controls (Fig. 1M; vector vs. Mock). Challenge with Δsde mutants harboring single SdeA or SdeB isoforms each showed rescue of the HA-Ub modification defect, and their rescue was largely eliminated by the presence of the C118S catalytic mutation (Fig. 1M; compare psdeA and psdeB to the respective C118S derivatives). Therefore, the absence of Sde DUB activity results in a strong reduction in both detergent-resistant Rtn4 and Ub modification of Rtn4, strongly arguing that Sde-mediated deubiquitination of K63-linked polyUb plays a role in LCV dynamics, possibly providing a pool of mono-Ub substrate for the ART domain.
ADPribosylation of ubiquitin by Sde results in hyper-ubiquitination of the LCV.
The data that polyUb association with the LCV occurs robustly immediately after bacterial contact in the absence of the DUB domain argues that the likely role of pR-modification is to stabilize polyUb (Fig. 1). Furthermore, both pR-modification and ADP-ribosylation by Sde block bacterial DUBs as well as a variety of host-encoded K63- and K48-specific DUBs (Fig. 2A)42, arguing that pR modification of Ub does not play a significant role in blocking polyubiquitination during LCV formation as proposed previously21.
The L. pneumophila sdeCH416A (PDE−) mutant results in ADPr modification of Ub, preventing pR-ubiquitination of target proteins or the generation of pR-Ub. The consequent ADPr-modified polyubiquitinated structures are resistant to DUB attack (Fig. 2A), allowing interrogation of Ub modification without confounding Rtn4 rearrangements. To analyze the consequences of L. pneumophila derivatives altered in Sde family function, BMDM monolayers were challenged for 20 min, and changes in polyUb about the LCV were analyzed by probing fixed samples with a polyUb-specific antibody (Fig. 2B). Neither sdeCC118S (DUB−) nor sdeCE859A (ART−) mutants showed significant increases in polyUb fluorescence intensity/vacuole if the analysis were restricted to LCVs having associated polyUb (Fig. 2B). In contrast, challenge with the L. pneumophila PDE− mutant resulted in dense accumulation of LCV-associated polyubiquitination in spite of the fact that Rtn4 rearrangements were blocked (Fig. 2B). This dense association was maintained over time (Fig. 2C). Most notable was the presence of unusual polyUb-modified structures after challenge with the PDE− mutant. While the Δsde and ART− mutants showed polyUb staining about individual LCVs (Figs. 2E,G), the PDE− mutant showed polyUb structures that wrapped around the LCV and extended into serpentine structures across the interior of the BMDM (Fig. 2H). We occasionally observed dense structures in response to challenge with the DUB− mutant as well (Fig. 2F), but the frequency was little changed from WT (Figs. 2B,C). We conclude that Sde-mediated modification of polyUb chains stabilizes LCV-associated Ub from the attack by both bacterial- and host derived DUBs. The observed deregulation of associated polyubiquitination in the absence of PDE activity indicates that direct modification of Ub is sufficient to stabilize polyUb. Furthermore, there is no evidence that pR- or ADPr-Ub modification plays a role in blocking polyubiquitination.
Sde-mediated modification of polyUb blocks autophagy adaptor SQSTM1/p62 binding via UBA domain.
The Ub Arg42 residue is a key target for modulating ubiquitin dynamics. Autophagy adaptors such as p62 link ubiquitinated compartments to LC3-decorated phagophores via ubiquitin associated (UBA) domains that recognize a hydrophobic patch on the Ub surface50, 51. Integral to this patch are the residues I44, V70 and L71, directly proximal to R42, the target residue of ADPr modification by Sde (Fig. 3A, blue arrows). Structural analysis has shown that R42 can directly participate in the UBA interface (Fig. 3B) and is likely involved in p62 recognition based on NMR analysis52, 53. We therefore tested if modification of the Ub at Arg42 residue blocks recognition by the p62 autophagy adaptor.
His-p62-Ni-NTA resin was challenged with K48- or K63-linked Ub3 − 7 mixed polymers that had been pretreated with SdeC WT or its catalytic mutant derivatives (Materials and Methods). The unbound fraction was collected (FT; flow through), and the p62-associated Ub3 − 7 (Eluate) was detected after co-elution with imidazole, followed by SDS-PAGE fractionation and immunoprobing with anti-Ub (Fig. 3C). In the absence of SdeC, both K48- and K63-linked Ub3 − 7 were efficiently co-eluted with p62 (Fig. 3D; no SdeC preincubation). In contrast, pretreatment with the SdeC WT, DUB− (C118S) or PDE− (H416A) derivatives all reduced K63- or K48-linked Ub3 − 7 binding to p62 (Fig. 3D; E lanes). This indicates that either pR- or ADPr-modification of polyUb blocks p62 recognition. In the case of the K48 derivative, the loss of the ART domain restored Ub3 − 7-p62 coelution, however it caused depolymerization of K48-linked Ub3 − 7 into monomeric Ub, making the effects of this mutant difficult to evaluate in this assay (Fig. 3D; E859A).
To determine the effects of Ub Arg42 modification in a more quantitative assay, binding affinities to Ub4 were determined using surface plasmon resonance (SPR) in the presence of a DUB inhibitor (Ub-propargylamide, Ub-PA) to prevent degradation of the Ub4 polymer (Fig. S2A). K63-linked Ub4 was modified by SdeC WT or its catalytic mutant derivatives in the presence of Ub-PA (Materials and Methods), the modified K63-linked Ub4 was introduced onto flow chips with immobilized p62, and binding was measured by SPR over time (Fig. 3E). In the absence of SdeC, Ub4 bound His-p62 with apparent KD~1 X 10− 7 M, indicating that binding affinity was at least as efficient as previous SPR studies using diUb as a substrate (KD = 9.3 X 10− 8 M; 54) (Fig. 3F). Binding of p62 to mono-Ub was very poor, consistent with DUB activity interfering with recognition of p62 (Fig. 3G, Fig. S2C). Ub4 pretreated with SdeC, either in the presence of Ub-PA or having the C118S DUB mutation, resulted in blockade of p62 binding, consistent with phosphoribosylation blocking recognition of Ub by UBA domains (Figs. 3H,I). ADPr modification of Ub at Arg42 was sufficient to block binding, as the SdeCH416A PDE− mutant blocked binding as efficiently as WT (Fig. 3J). Elimination of the ART activity restored Ub4 binding to p62, with a similar binding affinity to untreated, consistent with PA blocking the DUB activity (Fig. 3K; KD = 1.6 X 10− 7 M). These results strongly argue that modification of the Ub at Arg42 residue by SdeC disrupts recognition by p62.
Sde-mediated modification of polyUb chains prevents association of p62 with the LCV.
To determine if Sde proteins enzymatically camouflage the replication vacuole from recognition by p62, BMDMs were challenged with L. pneumophila sde variants and probed for p62 localization about the LCV. A variety of p62 morphological variants were observed. After bacterial contact (5 min centrifugation) and immediate fixation, LCVs harboring L. pneumophila WT were free from the punctate localization of p62 found throughout the cytoplasm (Fig. 4A). In contrast, a large fraction of the vacuoles after challenge with the L. pneumophila Δsde strain were enveloped by p62 (Fig. 4B). As time progressed, a variety of morphologies became apparent, such as the alignment of puncta proximal to the LCV. As puncta associated with the LCV were difficult to distinguish from those in the general vicinity, we limited quantification to identifying vacuoles with enveloping p62 structures and determining total p62 intensity associated with individual vacuoles.
Immediately after bacterial contact with BMDMs, both the WT and the T4SS− strains were devoid of p62 staining (Fig. 4C). In contrast, approximately 30% of the L. pneumophila Δsde bacteria were found to be enveloped by p62 (Fig. 4C). Colocalization was largely lost when either sdeA, sdeB, or sdeC expressed on plasmids was introduced into the Δsde strain. Quantification of p62 signal associated with the LCV showed that there were two populations of bacteria after challenge with the Δsde strain, with a population showing p62 recruitment levels that were similar to WT and a second population with 8-10X higher fluorescence intensity (Fig. 4D). In contrast, the WT and Δsde strains harboring Sde isoforms in trans showed indistinguishable low levels of p62 recruitment (Fig. 4D).
As the infection proceeded, colocalization of p62 with the Δsde strain continued to increase (Fig. 4E, > 50% at 20 MPI) with the population of LCVs strongly skewing toward hyperaccumulation of p62 (Fig. 4F). Concurrently, a second phenomenon became apparent in strains harboring plasmids that produced excess Sde isoforms. There was clear complementation of the Δsde defect (Fig. 4E), but accumulation of p62 about the LCV at 20 MPI was observed that distinguished these constructs from the WT strain (Fig. 4E). Accumulation could also be observed when the fluorescence intensity of individual LCVs was determined, with a population emerging with increased p62 density (Fig. 4F). This phenomenon appeared to be a premature version of what occurs with the WT strain, as 20% of the LCVs harboring the WT showed accumulation of p62 by 60 MPI, with evidence of an emerging population showing increased p62 density about the LCV (Figs. 4G, H). There was robust accumulation of p62 about individual LCVs harboring WT strain at this timepoint that appeared indistinguishable from individual LCVs harboring the Δsde strain (Fig. 4I). Furthermore, at 180MPI, accumulation of p62 around the LCV in the Δsde strain was indistinguishable from p62 accumulation around WT LCVs (Figs. 4J, K). When the percentage of LCVs associated with p62 was plotted as a function of time post-infection (Fig. 4L), there was early accumulation of p62 that was blocked by the Sde family (0–20 MPI). This was followed by decay of p62 accumulation after infection of the Δsde strain, while the WT showed increasing p62 between 20 and 60 MPI, resulting in a convergence of the phenotypes (180 MPI). The loss of Sde protection of the LCV is consistent with reversal of its activity by known metaeffectors SidJ and DupA/B 24, 25, 26, 55.
To determine the role of the individual Sde activities in preventing p62 accumulation, point mutations in each of the three catalytic domains were analyzed at 0 MPI after macrophage challenge, when the Sde family exerts its most striking effects. Mutations in sdeB were introduced into the Δsde strain, as this isoform showed the most effective complementation (Fig. 4C), and these mutants were then used to challenge macrophages. Point mutations in the catalytic sites of the DUB or PDE domains, as well as a double mutant missing the catalytic activity in both of these domains, still retained the ability to block p62 colocalization after association (Figs. 5A-C). In contrast, a mutation in the ART domain caused a large increase in p62 recruitment to LCVs, indicating that ADP ribosylation of Ub Arg42 by the ART domain blocks p62 recognition of the LCVs (Fig. 5A). It should be noted that there appeared to be some residual interference of p62 in the single ART mutant, but this was eliminated by simultaneously introducing a DUB mutation, indicating that the DUB domain may play a minor role in lowering p62 recognition of the LCV (Fig. 5A). As seen with the Δsde strain, the ART mutants resulted in a clear subpopulation showing approximately 10-fold increase in p62 accumulations (Figs. 5B,C). These results support the biochemical data that either ADPribosylation or phosphoribosylation of Ub at Arg42 is sufficient to block p62 recognition. It also argues against models that propose pR-linked Ub modification of a peripheral factor plays a role56, as ADPribosylation of Ub was sufficient to interfere with p62 recruitment.