The identification of RNF213 as a host targeting protein of S. flexneri IpaH1.4
In order to investigate whether there are additional host proteins might be targeted by IpaH1.4 during the invasion of S. flexneri, we performed a biochemical affinity purification coupled with mass spectrometric analysis-based assay to identify novel IpaH1.4-binding proteins in HEK293F cells (Fig. 1a). The FLAG-tagged catalytically inactive IpaH1.4 C368A mutant and the substrate-binding LRR domain of IpaH1.4 were individually overexpressed in HEK293F cells, and then were enriched and pulled down via anti-FLAG beads. Subsequently, the related human proteins that can associate with the two IpaH1.4 proteins were detected by mass spectrometric-based analysis. One of the identified strongest candidates for potentially binding to IpaH1.4 was the giant human E3 ligase RNF213 (Fig. 1b), which was recently demonstrated to function as a key player in antibacterial immunity5,26. To confirm the potential interaction between IpaH1.4 and RNF213, we firstly performed relevant GST pull-down assays, which showed that the GST-tagged IpaH1.4 but not the GST tag can interact with the full-length RNF213 (Fig. 1c). Further co-immunoprecipitation (Co-IP) assays confirmed that the full-length RNF213 and the RNF213(371–5207) fragment that lacks the N-terminal highly un-conserved and intrinsically disordered region of RNF213 based on our sequence analyses (Supplementary Fig. 1a, b), can well interact with IpaH1.4 in co-transfected cells (Fig. 1d). To further confirm the direct interaction between RNF213 and IpaH1.4, we sought to purify full-length RNF213 proteins to conduct relevant biochemical assays. However, the purified full-length RNF213 proteins from HEK293F cells tend to form large aggregates and display poor homogeneity (Supplementary Fig. 2a, b), preventing further detailed biochemical and structural characterizations. Fortunately, the purified RNF213(371–5207) fragment, which can still well interact with intracellular IpaH1.4 (Fig. 1d), is homogeneous monomer in solution based on our multi-angle light scattering (MALS) analysis (Supplementary Fig. 2c). Importantly, using size exclusion chromatography (SEC)-based assays with purified RNF213(371–5207) and the full-length IpaH1.4 proteins, we demonstrated that IpaH1.4 can directly interact with RNF213 (Fig. 1e).
The ubiquitination and proteasomal degradation of RNF213 mediated by IpaH1.4
Given that the IpaH family members can induce the degradation of their targeting substrates via ubiquitin-proteasome system (UPS), we wondered whether IpaH1.4 could mediate the ubiquitination and proteasome degradation of RNF213. Using relevant in vitro ubiquitination assays with the purified catalytically inactive C4516A mutant of RNF213(371–5207), which loses the E3 enzymatic activity of its RZF domain and cannot undergo auto-ubiquitination, we revealed that IpaH1.4 can directly catalyze the ubiquitination of RNF213 in vitro (Fig. 2a, b). Importantly, the ubiquitination of RNF213 catalyzed by IpaH1.4 requires the presences of ATP, UBA1 (E1), UBE2L3 (E2) and IpaH1.4 (E3) (Fig. 2a), confirming that the ubiquitination of RNF213 by IpaH1.4 is through a canonical mode of ubiquitination modification. Further mass spectrometry analysis revealed that there are multiple ubiquitin modification sites on RNF213(371–5207) catalyzed by IpaH1.4 (Fig. 2c and Supplementary Fig. 2d), all of which located in the E3 module of RNF213 (Supplementary Fig. 2e). Notably, S. flexneri contains two strikingly similar IpaH family effectors IpaH1.4 and IpaH2.544,49, which have only two residues different in their substrate-binding LRR domains (Supplementary Fig. 3a). Not surprisingly, based on our in vitro ubiquitination assays, RNF213 can be effectively ubiquitinated by IpaH2.5 (Supplementary Fig. 3b), suggesting that both IpaH1.4 and IpaH2.5 can directly target host RNF213. Importantly, despite the E3 activity of RNF213 contributes to the host immune response during pathogenic infections5, the enzymatically active RNF213 was incapable to ubiquitinate IpaH1.4 conversely in vitro (Supplementary Fig. 4a, b). To further elucidate the type of ubiquitin modification on RNF213 catalyzed by IpaH1.4, we constructed 9 different ubiquitin mutants, which contains additional N-terminal extension sequences and/or specific mutations by replacing relevant lysine residues of ubiquitin with arginine residues (Supplementary Fig. 4c). Further in vitro ubiquitination assays revealed that IpaH1.4 mainly catalyzes the K48-linked ubiquitination on RNF213 in vitro (Supplementary Fig. 4d), similar to other IpaH family proteins47. In line with our biochemical results and the fact that K48-linked polyubiquitin is a signal for proteasomal degradation, when co-expression of Aequorea coerulescens GFP-tagged (AcGFP) RNF213(371–5207) with relevant IpaH1.4 proteins in HeLa cells, the wild type IpaH1.4 but not the enzymatically inactive C368A mutant or the NEL domain of IpaH1.4 can significantly decrease the cellular protein level of RNF213(371–5207) in the absence of proteasomal inhibitor MG132 (Fig. 2d). Importantly, expression of the wild type IpaH1.4 rather than the E3-dead C368A mutant of IpaH1.4 in HeLa cells can induce the degradation of endogenous RNF213 (Fig. 2e).
To further investigate whether RNF213 is a physiological substrate of IpaH1.4/2.5 during S. flexneri invasion, we firstly generated an RNF213-knockout HeLa cell line using CRISPR-Cas9 technology (Supplementary Fig. 5a, b). Then, RNF213-knockout HeLa cells were transfected with relevant passer transposon system to generate a stable cell line expressing AcGFP-tagged wild-type RNF213 (Supplementary Fig. 5c, d). Subsequently, we infected the AcGFP-RNF213 stable HeLa cells with the wild-type S. flexneri M90T strain or the ipaH1.4/ipaH2.5 double knock-out (ΔipaH1.4/ΔipaH2.5) S. flexneri strain generated in our previous study49. As expected, intracellular AcGFP-RNF213 signals can be well detected in cells infected with the S. flexneri ΔipaH1.4/ΔipaH2.5 strain, especially on the surface of invaded bacteria, while they are barely detectable in cells infected with the wild-type S. flexneri (Fig. 2f-h). Consistently, obvious ubiquitin-coat on the surface of invading bacteria can be observed in cells infected with the S. flexneri ΔipaH1.4/ΔipaH2.5 strain but not the wild-type S. flexneri (Fig. 2f, g, and i). Given that RNF213 can directly mediate the ubiquitination of LPS on the surface of invading Salmonella5, these data suggested that S. flexneri can utilize its effector IpaH1.4 and IpaH2.5 to eliminate host RNF213 during invasion, thereby helping S. flexneri to escape from being marked by ubiquitination as well as related downstream antibacterial immune processes.
The cryo-EM structure of human RNF213
To elucidate the molecular basis underlying the specific recognition of RNF213 by IpaH1.4, we firstly purified the RNF213(371–5207)/IpaH1.4 complex by mixing excess amounts of IpaH1.4 with human RNF213(371–5207) followed by SEC-based separation (Supplementary Fig. 6a and b). The purified RNF213(371–5207)/IpaH1.4 complex exhibited reasonable homogeneity based on our negative staining electron microscopy analysis (Supplementary Fig. 6c). Subsequently, using the cryo-EM method, we obtained the final consensus density map of the RNF213(371–5207)/IpaH1.4 complex to an overall resolution of 3.46 Å (Supplementary Fig. 7a-c and Supplementary Table 1). Unfortunately, based on the solved cryo-EM density map, we could only build a high-confidence model of RNF213, which mainly contains five structural segments, the N-terminal stalk, the middle six AAA units-containing AAA ring, the C-terminal multi-domain E3 module, and the stalk/AAA ring-connecting linker as well as the AAA ring/E3 module-connecting hinge (Fig. 3a-c). Similar to previously determined cryo-EM structure of mouse RNF21350, human RNF213 folds into a compact zig-zag conformation from its N-terminal stalk to the C-terminal CTD (Fig. 3a-c). Notably, the E3 module of human RNF213 is composed of six sub-domains, namely back, RING, shell, RZF, core and CTD, and has direct contacts with the stalk and AAA ring modules through its back and CTD, respectively (Fig. 3a-c). The RING domain swinging between the E3 back and E3 shell was previously thought to be the only domain with E3 enzymatic activity just like other RING-type E3 ligases51,52, however recent studies well demonstrated that RNF213 can also perform its E3 ligase function in a RING-independent manner5. Particularly, RNF213 can directly ubiquitylate LPS on the surface of invading Salmonella through its active C4516 residue within the RZF domain5. Notably, the electron densities of the RING and RZF regions are relatively weak, likely owing to their conformational flexibilities (Fig. 3b). Further detailed structural analyses revealed that, among the six AAA units of RNF213, only AAA3 and AAA4 harbor all the essential catalytic residues and functional motifs for exerting ATP-binding and ATP-hydrolysis (Supplementary Fig. 8), while the AAA1, AAA5, and AAA6 are completely inactive due to the lack of relevant key structural elements (Supplementary Fig. 8a, e, f). Intriguingly, in the cryo-EM structure of human RNF213, the AAA2 of RNF213 contains a well defined ATP molecule, but lacks the catalytic arginine finger for ATP hydrolysis (Supplementary Fig. 8b and Fig. 3b, c), suggesting that it is catalytically incompetent, in line with a previous study50. Moreover, there is a unique insertion existed between the AAA2 and AAA3 of RNF213 (Fig. 3a-c). Further structural comparison analyses showed that the overall structure of human RNF213 is very similar to that of previously determined the apo-form mouse RNF21350, however the relative orientation and gap between the N-terminal segment of stalk and the E3 module of human RNF213 is very different from that of mouse RNF213 (Supplementary Fig. 9a-f). In addition, the density map of the RZF domain within the E3 module of human RNF213 is much better than that of mouse RNF213 (Supplementary Fig. 9g-i).
IpaH1.4 recognizes the RING domain of RNF213 through its LRR domain
After fitting the whole RNF213 structure model into the cryo-EM density map, we found an extra weak density around the RING domain of RNF213, which does not belong to RNF213 but can be partially matched with the IpaH1.4 LRR domain (Supplementary Fig. 10). Further local refinement focusing around the RNF213 E3 module resulted in a local density map with a much improved resolution for fitting the IpaH1.4 LRR domain (Fig. 3d, e and Supplementary Fig. 7). Based on the refined local density map, IpaH1.4 LRR can directly bind to the RING domain of RNF213 through its concave side (Fig. 3d and e). However, due to the low resolution of the local density map, we were unable to observe the detailed interactions between interface residues of RNF213 RING and IpaH1.4 LRR. Moreover, we also could not identify any density belonging to the IpaH1.4 NEL domain, likely due to its dynamic nature for facilitating its ubiquitination on multiple sites of RNF213. Importantly, consistent with our structural data, when mapping aforementioned MS-identified ubiquitination sites of RNF213 onto the cryo-EM structure of RNF213, we found that they are all adjacent to RNF213 RING and are reachable for the IpaH1.4 NEL domain (Supplementary Figs. 2d and 11a-c).
Using SEC-based biochemical assays, we further confirmed that the isolated RNF213 RING (residues 3990–4056) can directly interact with the IpaH1.4 LRR domain (residues 38–273) (Fig. 4a). In addition, quantitative isothermal titration calorimetry (ITC) analysis uncovered that IpaH1.4 LRR binds to RNF213 RING with a dissociation constant (KD) value of ∼0.18 µM (Fig. 4b). Consistent with our cryo-EM data, further analytical ultracentrifugation-based assay revealed that RNF213 RING and IpaH1.4 LRR can associate with each other to form a stable 1:1 stoichiometric complex in solution (Fig. 4c).
The crystal structure of RNF213 RING in complex with the LRR domain of IpaH1.4
In order to elucidate the detailed binding mode between RNF213 RING and IpaH1.4 LRR, we solved the complex structure of RNF213 RING and IpaH1.4 LRR to 1.70 Å resolution using X-ray crystallographic method (Supplementary Table 2). As expected, in the RNF213 RING/IpaH1.4 LRR complex structure, IpaH1.4 LRR forms a horseshoe-shape fold, and directly binds to RNF213 RING through its concave side (Fig. 4d). No significant structural changes can be observed in the IpaH1.4 LRR domain upon its binding to RNF213 RING (Supplementary Fig. 12a). Concurrently, in the complex structure, RNF213 RING that mainly consists of one N-terminal α-helix followed by two short C-terminal antiparallel β-strands, adopts a unique architecture containing two Zn2+ ions to engage with the β4 to β9 leucine-rich repeat region of IpaH1.4 LRR, burying a total surface area of ∼693 Å2 (Fig. 4d). Further structural comparison analyses revealed that the overall structure of human RNF213 RING in the RNF213 RING/IpaH1.4 LRR complex is similar to that of the apo-form mouse RNF213 RING (Supplementary Fig. 12b), and the overall binding mode of IpaH1.4 LRR with RNF213 RING resembles that of the IpaH1.4 LRR/HOIP RING1 and IpaH1.4 LRR/HOIL-1L UBL interactions (Supplementary Fig. 12c-e).
Detailed structural analysis of the RNF213 RING/IpaH1.4 LRR complex unravelled that the specific interaction between IpaH1.4 LRR and RNF213 RING is mediated by both hydrophobic contacts and polar interactions (Fig. 4e and Supplementary Fig. 13a, b). Specifically, the hydrophobic side chains of I3999, W4024 and L4036 of RNF213 RING form hydrophobic contacts with the hydrophobic side chains of L175, V177, A197, A199, V217, M219, and F238 as well as the aliphatic side chain groups of R215 from IpaH1.4 LRR (Fig. 4e and Supplementary Fig. 13a). In parallel, the positively charged side chain of K100 in IpaH1.4 LRR forms a salt bridge with the negatively charged side chain of D4013 from RNF213 RING (Fig. 4e and Supplementary Fig. 13b). Concurrently, the side chains of D140, N200 and R215 in IpaH1.4 LRR form three hydrogen bonds with the side chains of H4014, Y4034 and Q4029 in RNF213 RING (Fig. 4e). In addition, the side chains of R157, S179, N200, and N220 residues of IpaH1.4 LRR couple with the backbone carboxyl groups of S3998, I3999, Y4034, C4035, and L4036 from RNF213 RING to form six specific hydrogen bonds (Fig. 4e). Using site-directed mutagenesis-based biochemical assays, we further validated the specific interactions between RNF213 RING and IpaH1.4 LRR found in the complex structure. In keeping with our structural data, mutations of key binding interface residues either from IpaH1.4 LRR or RNF213 RING, such as the R157A, A197R, R215E and R157A/R215E mutations of IpaH1.4 LRR or the D4013R, Q4029A, Y4034A, and L4036R mutations of RNF213 RING, all significantly reduced or essentially abolished the interaction of RNF213 RING with IpaH1.4 LRR in our ITC-based assays (Fig. 4f and Supplementary Fig. 14a-h). In agreement with our aforementioned biochemical and structural results, further in vitro ubiquitination assays proved that mutations of key interface residues of IpaH1.4 or RNF213, such as the A197R mutation of IpaH1.4 or the L4036R mutation of RNF213 C4516A, largely attenuate or completely disrupt the ubiquitination of RNF213 imposed by IpaH1.4 (Fig. 4g and h). Concomitantly, the A197R mutation of IpaH1.4 or the L4036R mutation of RNF213(371–5207) can effectively dampen the IpaH1.4-induced degradation of cellular AcGFP-RNF213(371–5207) in co-transfected cells (Fig. 4I). Notably, the key interface residues of IpaH1.4 LRR for recognizing RNF213 can be also found in IpaH2.5 but are absence in other S. flexneri IpaH family proteins or S. typhimurium effector SspH1 and SspH2 (Supplementary Fig. 15). Thus, IpaH1.4 and IpaH2.5 are the only two members in the IpaH family that can specifically target RNF213.
IpaH1.4 can be recruited to the RNF213-coat decorated on cellular lipid droplets through the specific RNF213 RING/IpaH1.4 LRR interaction
Previous studies well demonstrated that RNF213 directly participates in the lipid droplet (LD) formation, and can form toroidal patterns decorated on the surface of intracellular LDs, when cells are treated with oleic acid (OA)30. Consistently, the OA-treatment of the AcGFP-RNF213 stable HeLa cells can induce the intracellular toroidal pattern formation of RNF213 surrounded LDs (Supplementary Fig. 16). To further validate whether IpaH1.4 can directly target RNF213 under physiological conditions, we also generated a specific cell line from RNF213-knockout HeLa cells with a stable expression of the RNF213 L4036R mutant, which was proved to be unable to interact with IpaH1.4 (Fig. 4f, h, and I). Then, we over-expressed the mCherry-tagged E3-dead IpaH1.4 C368A mutant, relevant IpaH1.4 variants, or the control mCherry tag in the AcGFP-RNF213 or AcGFP-RNF213 L4036R stable cell lines, which were further treated with OA to induce cellular LD formations (Fig. 5a-e). As expected, the mCherry-tagged IpaH1.4 C368A but not the mCherry tag in the transfected AcGFP-RNF213 stable cells exhibited an obvious enrichment on the RNF213-coated LDs in the cytoplasm (Fig. 5a and f), suggesting that RNF213 can effectively recruit IpaH1.4 to cellular LDs. In contrast, the R157A/R215E/C368A and A197R/C368A mutants of IpaH1.4, both of which were demonstrated to lose their abilities to interact with RNF213 (Fig. 4f and g), were essentially unable to co-localize with the RNF213-coat on the surface of LDs in the transfected AcGFP-RNF213 stable cells (Fig. 5b, c and f). Conversely, the RNF213 L4036R-coat surrounded LDs was unable to recruit IpaH1.4 C368A in the transfected AcGFP-RNF213 L4036R stable cells (Fig. 5d and f). Taken together, these data clearly demonstrated that the specific RNF213 RING/IpaH1.4 LRR interaction is essential for the cellular targeting of IpaH1.4 to the RNF213-coat decorated on lipid droplets.
Targeting of RNF213 by IpaH1.4 facilitates the proliferation of S. flexneri in host cells
To further investigate the impact of the RNF213 RING/IpaH1.4 LRR interaction towards bacteria proliferation during S. flexneri invasion, we utilized the ipaH1.4/ipaH2.5 double-knockout (ΔipaH1.4/ΔipaH2.5) S. flexneri strain and rescued it with the wild type ipaH1.4 gene or relevant different ipaH1.4 mutant gene (the R157A/R215E, A197R or C368A mutant of IpaH1.4). Then, we performed infection experiments in the wild type or RNF213-knockout HeLa cells using these S. flexneri strains. The results showed that the lack of RNF213 facilitates the invasion of wild type S. flexneri but not S. flexneri ΔipaH1.4/ΔipaH2.5 (Supplementary Fig. 17a), suggesting that RNF213 can restrain the infection of S. flexneri and is likely one of the targets of IpaH1.4/2.5 in host cells. Importantly, comparing with that of the wild type S. flexneri, the proliferation of S. flexneri ΔipaH1.4/ΔipaH2.5 stain in the wild type HeLa cells was largely attenuated (Fig. 5g). However, the deficiency of RNF213 facilitates the replication of both wild type S. flexneri and S. flexneri ΔipaH1.4/ΔipaH2.5 in infected RNF213-knockout cells (Fig. 5g). These findings strongly suggested that S. flexneri relies on IpaH1.4/2.5 to target RNF213 thereby promoting its proliferation in infected host cells. Consistently, although S. flexneri ΔipaH1.4/ΔipaH2.5 strain rescued with the ipaH1.4 R157A/R215E or A197R mutant has a similar infective activity towards the wild type HeLa cells as that of S. flexneri rescued with the wild type ipaH1.4 (Supplementary Fig. 17b), the proliferation of S. flexneri rescued with ipaH1.4 R157A/R215E, or A197R in the infected cells was largely reduced compared with that of S. flexneri rescued with the wild type ipaH1.4 (Fig. 5h), confirming that the targeting of RNF213 by IpaH1.4 is critical for promoting the proliferation of S. flexneri in host cells. Intriguingly, S. flexneri rescued with E3-dead ipaH1.4 C368A shows a much lower infective activity than that of S. flexneri rescued with the wild type ipaH1.4, the ipaH1.4 R157A/R215E mutant, or the ipaH1.4 A197R mutant (Supplementary Fig. 17b), implying that the E3 activity of IpaH1.4 contributes to the infection of S. flexneri and there are other host targets of IpaH1.4 except for RNF213, in line with previous studies44,49,53. Notably, our quantitative RT-PCR analyses revealed that the transcription levels of RNF213 in host cells are not significantly changed during the infection of S. flexneri (Supplementary Fig. 17c and d). Taken together, all these data well demonstrated that RNF213 can restrain S. flexneri proliferation in infected host cells, but S. flexneri can secrete E3 effector IpaH1.4/2.5 to specifically target and subvert RNF213, thereby facilitating its proliferation during invasion.