Structural basis for effector protein recognition by the Dot/Icm Type IVB coupling protein complex

The Legionella pneumophila Dot/Icm type IVB secretion system (T4BSS) is extremely versatile, translocating ~300 effector proteins into host cells. This specialized secretion system employs the Dot/Icm type IVB coupling protein (T4CP) complex, which includes IcmS, IcmW and LvgA, that are known to selectively assist the export of a subclass of effectors. Herein, the crystal structure of a four-subunit T4CP subcomplex bound to the effector protein VpdB reveals an interaction between LvgA and a linear motif in the C-terminus of VpdB. The same binding interface of LvgA also interacts with the C-terminal region of three additional effectors, SidH, SetA and PieA. Mutational analyses identified a FxxxLxxxK binding motif that is shared by VpdB and SidH, but not by SetA and PieA, showing that LvgA recognizes more than one type of binding motif. Together, this work provides a structural basis for how the Dot/Icm T4CP complex recognizes effectors, and highlights the multiple substrate-binding specificities of its adaptor subunit. The Legionella pneumophila Dot/Icm type IVB secretion system (T4BSS) translocates effector proteins into host cells, and the recognition of these effectors is mediated by the Dot/Icm type IV coupling protein (T4CP) complex. Here, the authors present the crystal structure of a four-subunit containing T4CP subcomplex bound to the effector protein VpdB, and identify a FxxxLxxxK binding motif that is present in a subset of the effectors and which is recognized by the T4CP adaptor subunit LvgA.

P rokaryotic pathogens subvert normal physiology of eukaryotic hosts by secreting effector proteins to achieve a successful infection. Legionella pneumophila, a Gram-negative intracellular pathogen causing Legionnaires' disease, secretes about 300 effector proteins into human macrophage cells through specialized protein-conducting channels. These proteins suppress immune defense and affect cellular homeostasis toward the intracellular survival, growth, and replication of the bacterium 1-3 . This type of protein secretion system involves three steps: (i) recognition of effector proteins by a coupling protein complex, (ii) unfolding of the effector proteins in the cytoplasm, and (iii) translocation of the effector proteins in an unfolded state through a transenvelope conduit [4][5][6][7] . L. pneumophila has a Dot/Icm type IVB secretion system (T4BSS) [7][8][9] , which is composed of 26 Dot/ Icm (defective for organelle trafficking/intracellular multiplication defect) proteins and the adaptor protein LvgA [10][11][12][13][14] . It is made up of two major complexes, the core transenvelope conduit and a Dot/Icm type IV coupling protein (T4CP) complex. The core transenvelope conduit, composed of~10 Dot/Icm proteins, serves as the channel for effector translocation 15,16 . In the Dot/ Icm T4CP complex, DotL/IcmO is the central subunit and belongs to the TraG family of coupling proteins that are classified as AAA+ type ATPases and function to link translocating substrates to the secretion conduit 4,17,18 . Unlike the prototypical T4CP TraG, DotL contains a~200 residue C-terminal extension. While a similar region exists in many T4CPs, they vary greatly in length and amino acid sequence 19 . The C-terminal extension (CTE) of DotL interacts with a heterodimer of IcmS and IcmW (called IcmSW), DotN (IcmJ), and LvgA 4,19 . In addition, DotL interacts with the inner membrane protein DotM (IcmP) through its transmembrane helices and its membrane-proximal ATPase domain 4,20 . Thus, this multi-protein T4CP complex, rather than the DotL T4CP alone, would function to recognize and process effector proteins 21 . We previously reported that a pentameric complex, composed of a fragment of the DotL CTE, DotN, IcmS, IcmW, and LvgA, interacted with a number of Legionella effector proteins, showing that this C-terminal assembly of DotL is the substrate-recognition module of the Dot/Icm T4CP complex 19 . Together with crystallographic and modeling studies, we put forth a pseudo-atomic model for the Dot/Icm T4CP holocomplex, which is a hexamer of an N-terminal ATPase domain of DotL and a substrate-recognition assembly built on the CTE of DotL 19 .
To examine how the Dot/Icm T4CP complex can recognize so many effector proteins, we extended our structural studies to examine the interaction between the Dot/Icm T4CP subcomplex and substrates. Herein, we report the crystal structure of a DotL (656-783)-IcmSW-LvgA complex bound to VpdB(461-590), a Cterminal fragment of the Legionella effector protein VpdB, which is sufficient for translocation in vivo. The structure provides the view of how the Dot/Icm T4CP complex recognizes its substrate protein and a glimpse at the molecular mechanisms underlying the recognition of a large number of effector proteins by this complex.
The four-helix bundle is involved in minor contacts with LvgA in comparison with that between α1 and LvgA (Fig. 1c). Indeed, in a native PAGE-based protein-binding assay, α1 of VpdB alone interacted with DotL(656-783)-IcmSW-LvgA, while the fourhelical bundle of VpdB did not (Fig. 1d), confirming that α1 of VpdB is necessary and sufficient for LvgA binding.
The other three substitutions affected the binding affinity, but to a lesser extent (Fig. 2d). Together, these mutational analyses highlight the importance of the hydrophobic pocket of LvgA, and identify the three most important residues on α1 in the recognition of VpdB by the adaptor subunit.
Secretion of VpdB and SidH depend on IcmSW and partially on LvgA. VpdB is one of the Legionella effector proteins whose secretion into host cells has been uncharacterized. We performed CyaA translocation assay 23 where the reporter Bordetella pertussis adenylate cyclase (CyaA) fused to VpdB was used, and the secretion of the fusion protein into the host Chinese hamster ovary cells overexpressing FcγRII (CHO FcγRII) was monitored by measuring cAMP production. We employed five strains: Lp01, a derivative of L. pneumophila (Philadelphia-1); Lp01 ΔdotA, an isogenic mutant lacking dotA and defective in the Dot/Icm transporter function; Lp01 ΔicmSW, a mutant lacking icmS and icmW; Lp01 ΔlvgA, a mutant lacking lvgA; Lp01 ΔicmSWΔlvgA, a mutant lacking icmS, icmW and lvgA. Like RalF, the translocation of VpdB from Lp01 ΔdotA was drastically affected, demonstrating that VpdB is a substrate of the Dot/Icm T4BSS (Fig. 4a, d). However, unlike RalF, whose translocation is independent of IcmSW 24,25 , translocation of VpdB from Lp01 ΔicmSW or Lp01 ΔlvgA was attenuated significantly, in comparison with that from the wild-type strain Lp01 (Fig. 4a, d). The absence of IcmS and IcmW (in Lp01 ΔicmSW) resulted in about 30-fold decrease in the cAMP level, while that of LvgA (in Lp01 ΔlvgA) resulted in about 10-fold decrease (Fig. 4a). The data demonstrate that VpdB is an IcmSW-dependent effector for its translocation, and that the VpdB translocation depends also on LvgA, but less heavily than it does on IcmSW. The C-terminal VpdB(461-590) fragment was sufficient for the translocation, since it exhibited nearly the same translocation pattern as full-length VpdB. In contrast, the N-terminal VpdB(1-343) fragment did not exhibit a demonstrable translocation (Fig. 4a). To know whether the intermolecular interaction between Phe476 and the hydrophobic pocket of LvgA observed in the crystal structure is important for VpdB translocation (Fig. 2b, c), the translocation assay was performed with VpdB(F476E) fused to CyaA. Notably, the translocation of this mutant from Lp01 (Fig. 4a, 4th set) was similarly attenuated as that of wild-type VpdB from Lp01 ΔlvgA (Fig. 4a, 1st set), supporting that the key intermolecular interaction observed in vitro is directly relevant for the translocation of VpdB in vivo.
SidH and its variants exhibited a translocation pattern quite similar to VpdB. Both full-length SidH and a C-terminal fragment SidH(1830-2200) depended on IcmSW and LvgA for translocation; but the effect of deletion of lvgA was less significant than that of icmSW (Fig. 4b), thus suggesting that VpdB and SidH share the same translocation mechanism.
Of note, translocation of VpdB and SidH from Lp01 ΔicmSW is similar to that from Lp01 ΔicmSWΔlvgA, indicating that the additional deletion of lvgA has no appreciable effect, and thus their LvgA-dependent translocation is subject to the presence of IcmSW (Fig. 4c). These data are consistent with the crystal structure of DotL(656-783)-IcmSW-LvgA-VpdB(461-590) (Fig. 1b), showing that the presence of IcmSW is required for LvgA to assemble into the T4CP complex and function in this complex.
LvgA-dependent secretion of PieA and SetA. Additionally, we performed the translocation assay for SetA and PieA, the latter of which is known to depend on IcmSW for its translocation 26 . In this assay, translocation of SetA was also attenuated by the NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-16397-0 ARTICLE NATURE COMMUNICATIONS | (2020) 11:2623 | https://doi.org/10.1038/s41467-020-16397-0 | www.nature.com/naturecommunications absence of icmSW or lvgA. The pattern of the attenuation was similar to that of VpdB: a greater attenuation displayed by Lp01 ΔicmSW than by Lp01 ΔlvgA (Fig. 4d). In comparison, the translocation of PieA appeared to depend specifically on LvgA, because the translocation from Lp01 ΔicmSW was as attenuated as that from Lp01 ΔlvgA (Fig. 4d).
A model for the Dot/Icm T4CP holocomplex bound to fulllength VpdB. In the structure of VpdB(461-590) bound to DotL (656-783)-IcmSW-LvgA, helix α1 is largely separated from the four-helical bundle, which is likely to result from a bindinginduced large conformational change. Conceivably, the Cterminal domain of VpdB is conformationally flexible in free   (Fig. 5). As expected, the junction between the N-and C-terminal domains of VpdB was heterogeneously modeled, and their relative orientation varied greatly among the five output homology models ( Supplementary Fig. 3). One of them exhibited no steric clash upon structural superposition onto the Dot/Icm T4CP holocomplex model, and showed that the C-terminal four-helix bundle of VpdB points toward the central chamber-like space of the holocomplex whereas the N-terminal domain of VpdB faces the outside of the holocomplex (Fig. 5). The constructed model shows that the bound substrate is far from the membraneproximal ATPase assembly, and suggests that a hinge-bending motion of the substrate-recognition assembly is required for the bound substrate to reach the ATPase assembly for processing (Fig. 5).

Discussion
LvgA, a virulence factor important for intracellular growth of Legionella 32,33 , is a newly identified subunit for the Dot/Icm T4CP complex 19 . This subunit interacts with at least four effector proteins (VpdB, SidH, SetA, and PieA). We define LvgA as an adaptor, because it is required for linking these effector proteins to the rest of the complex. Of note, DotL(656-783)-IcmSW-LvgA exhibited no detectable binding interaction with other effectors such as SidJ or Lpg0393 19 , indicating that a subset of Legionella effectors are recognized by the LvgA subunit. To serve as an adaptor, LvgA has a hydrophobic surface that binds to a concave surface of IcmSW 19 , whereas a shallow, wide surface containing a hydrophobic pocket on the opposite side serves as the interface for binding effector proteins (Figs. 1b, 2a-c, and 3b).
To date, two different translocation signals in Legionella effectors have been identified. One is the C-terminal extreme sequence containing a leucine or other hydrophobic residue, as observed for RalF whose secretion is independent of IcmSW 25 . The other is the E-block motif (EExxE) near the C-terminus, which was demonstrated with SidM 34 . The E-block motif was proposed to be recognized by the Dot/Icm T4CP complex via the DotM subunit, which contains basic patches that may serve as the binding platform for the E-block 20 .
In this study, we identified the FxxxLxxxK motif on a Cterminal α-helix of VpdB and SidH that is recognized by the adaptor subunit LvgA in the Dot/Icm T4CP complex (Figs. 1c, d,  2d, and 3d, e). 257 out of a total of 2930 proteins of the L. pneumophila strain Philadelphia-1 (Proteome ID: UP000000609) contain the FxxxLxxxK sequence motif. Of these, 46 proteins, including SidH and VpdB, belong to the 280 known Legionella effectors 35 , indicating that this sequence is enriched in the effector proteins (16.4% of effectors vs. 8.0% of non-effectors). However, the FxxxLxxxK motif is not present in the C-terminal regions of SetA and PieA, which can also bind to LvgA. The Cterminal regions of SetA and PieA, which are composed mostly of α-helices, do not share sequence homology with each other (Supplementary Fig. 1a). Therefore, they likely present an alternative motif(s) that interacts with the binding interface of LvgA.
Since VpdB(461-590) interacts only with LvgA in the crystal structure of DotL(656-783)-IcmSW-LvgA-VpdB(461-590), VpdB translocation from the Lp01 ΔlvgA strain was expected to be as attenuated as that from the Lp01 ΔicmSW strain. However, export was more diminished in the Lp01 ΔicmSW strain compared to the Lp01 ΔlvgA strain (Fig. 4a), indicating that VpdB translocation is partially dependent on LvgA. We hypothesize that additional, as-yet-unidentified, adaptor proteins may exist that compensate in the absence of LvgA. In contrast, strains lacking IcmSW would not display any adaptor, thus explaining the greater reduction of VpdB translocation observed in the Lp01 ΔicmSW mutant. The same scenario would be true for SetA translocation, but not for PieA translocation. LvgA appears to be the sole adaptor for PieA, because its translocation from Lp01 ΔicmSW was as attenuated as that from Lp01 ΔlvgA (Fig. 4b).
We further postulate that the Dot/Icm T4CP complex is likely to be heterogeneous in terms of the adaptor displayed, where the six copies of IcmSW bound to the DotL hexamer are either unoccupied, occupied by LvgA or by an unidentified adaptor protein(s). This heterogeneity in adaptors would mediate an expanded specificity of the Dot/Icm T4CP complex and allow recognition of a wide range of substrate proteins. Unoccupied IcmSW would recruit effector proteins that interact directly with IcmSW, such as SdeA, which contains a 200-residue domain near its C-terminus capable of binding to DotL(656-798)-IcmSW 36 . On the other hand, IcmSW bound to LvgA or other adaptors would recruit effector proteins that are able to interact with their cognate adaptor(s). In this scenario, IcmSW is of greater importance than LvgA or other adaptor protein(s) in effector translocation, which explains why the Legionella strains lacking icmS or icmW are more defective than the strain lacking lvgA in the intracellular growth of Legionella 33 . The adaptor proteins would have multiple specificity of binding, as observed for LvgA, in order to recognize not a single effector but a subset of effector proteins. Based on the large number of effectors exported by the Dot/Icm system, we suspect that Legionella may encode an Fig. 3 SetA, PieA, and SidH interaction with DotL(656-783)-IcmSW-LvgA. a Native PAGE analysis. Not the enzymatic domain but the C-terminal domain of SetA, PieA, and SidH interacted with (His) 10 -MBP-DotL(656-783)-IcmSW-LvgA, as indicated by the newly formed protein bands (red arrowheads). Schematic drawings of the three effectors are shown at the top. Black brackets indicate diffusive input protein bands. b Importance of the hydrophobic pocket on LvgA. Native PAGE analyses were performed with the indicated complex containing an I153E mutation. Its interactions with SetA, PieA, and SidH were disrupted or significantly reduced. c Quantification by bio-layer interferometry. The I153E substitution on the hydrophobic pocket of LvgA decreases the binding affinity drastically for SetA and significantly for PieA and SidH (~13-fold and~11-fold increase in K D , respectively). Each quantification was repeated 2 or more times. d Identification of the binding motif in SidH. A C-terminal segment of SidH (residues 2183-2200) bears sequence similarity with α1 helix of VpdB, and is predicted to be an α-helix (left). The three indicated mutations in SidH(1830-2200) reduced the binding affinity for DotL(656-783)-IcmSW-LvgA in a varying degree (middle and right). The measurement was repeated 2 or more times for each variant. e FxxxLxxxK motif containing α-helix of SidH as a major binding fragment. (His) 10 -YFP-SidH(2183-2200) (10 μM) was incubated with (His) 10 -MBP-DotL (656-783)-IcmSW-LvgA at a 1:1 molar ratio, and visualized on a native gel. The red arrowhead indicates the formation of a new protein band. All the native PAGE analyses were repeated more than 3 times. NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-16397-0 ARTICLE NATURE COMMUNICATIONS | (2020) 11:2623 | https://doi.org/10.1038/s41467-020-16397-0 | www.nature.com/naturecommunications arsenal of adaptors similar to LvgA that are essential to the regulation of export of T4BSS substrates by this pathogen.
In conclusion, the presented work provides the atomicresolution view of how the Dot/Icm T4CP complex recognizes effector proteins to be translocated. The adaptor subunit LvgA utilizes a wide and shallow binding surface with a hydrophobic pocket to recognize a subset of cognate effectors. This surface exhibits multiple substrate-binding specificity, as it binds as-yetunknown motif(s) as well as the FxxxLxxxK motif on an α-helix, both located near the C-terminus of the four effector proteins tested in this study. We suggest that the Dot/Icm T4CP complex is equipped with a number of different mechanisms to recognizẽ 300 effector proteins, including compositional heterogeneity involving more adaptor proteins other than LvgA.    CyaA-RalF CyaA-SetA CyaA-PieA CyaA-VpdB CyaA-SidH Fig. 4 CyaA translocation assay. a Translocation of VpdB. CHO-FcγRII cells were infected with the Legionella strains expressing the indicated VpdB constructs fused to CyaA, and the cAMP production was measured. Shown at the top is protein immunoblotting of CyaA-VpdB expressed in each Legionella strain using anti-CyaA IgG. The expression level of the tested VpdB constructs was similar. Each experiment was biologically duplicated and technically triplicated. b Translocation of SidH. Translocation patterns of SidH and its variants are similar to those of VpdB. Translocation assay and immunoblotting were performed similarly as in (a). c Dependency of LvgA-mediated translocation on IcmSW. Lp01 ΔicmSW and Lp01 ΔicmSWΔlvgA strains exhibited a similar level of attenuation in the transport of VpdB and SidH. d LvgA-dependent translocation of SetA and PieA. Both effectors depended on LvgA to be translocated. CyaA-RalF was used as a control. Each experiment was triplicated, including biological duplication. For immunoblotting, 10 milliOD 600 units were loaded. Statistical differences compared with the results obtained with Lp01 ΔicmSW were determined by the two-sided t-test for the other strains. The error bars represent mean values ± SD. ***p < 0.001; **p < 0.01; *p < 0.05; n.s., not significant. Exact p-values are provided in Supplementary Fig. 4 The junction loop between the ATPase assembly and the substrate recognition assembly is indicated, where a hinge bending motion might take place. A model for full-length VpdB built by homology modeling is shown. Each subunit is color coded. The holocomplex model does not include the membrane segment of DotL and the DotM subunit. The Legionella pneumophila Dot/Icm type IVB secretion system (T4BSS) translocates effector proteins into host cells and the recognition of these effectors is mediated by the Dot/Icm type IV coupling protein (T4CP) complex. Here, the authors present the crystal structure of a four-subunit containing T4CP subcomplex bound to the effector protein VpdB and identify a FxxxLxxxK binding motif that is present in a subset of the effectors and which is recognized by the T4CP adaptor subunit LvgA.