Structural insights into the roles of the IcmS–IcmWcomplex in the type IVb secretion system ofLegionella pneumophilaJianpo Xua,1, Dandan Xua,1, Muyang Wana, Li Yina, Xiaofei Wanga, Lijie Wub, Yanhua Liuc,d, Xiaoyun Liuc,d, Yan Zhoua,2,and Yongqun Zhua,2

aLife Sciences Institute and Innovation Center for Cell Signaling Network, Zhejiang University, Hangzhou, Zhejiang 310058, China; bNational Center forProtein Science Shanghai, Institute of Biochemistry and Cell Biology, Shanghai Institutes for Biological Sciences, Shanghai Science Research Center, ChineseAcademy of Sciences, Shanghai 200031, China; cInstitute of Analytical Chemistry, College of Chemistry and Molecular Engineering, Peking University, Beijing100871, China; and dSynthetic and Functional Biomolecules Center, College of Chemistry and Molecular Engineering, Peking University, Beijing100871, China

Edited by Jorge E. Galán, Yale University School of Medicine, New Haven, CT, and approved November 8, 2017 (received for review April 25, 2017)

The type IVb secretion system (T4BSS) of Legionella pneumophila is amultiple-component apparatus that delivers ∼300 virulent effectorproteins into host cells. The injected effectors modulate host cellularprocesses to promote bacterial infection and proliferation. IcmS andIcmW are two conserved small, acidic adaptor proteins that form abinary complex to interact with many effectors and facilitate theirtranslocation. IcmS and IcmW can also interact with DotL, an ATPaseof the type IV coupling protein complex (T4CP). However, how IcmS–IcmW recognizes effectors, and what the roles of IcmS–IcmW are inT4BSSs are unclear. In this study, we found that IcmS and IcmW forma 1:1 heterodimeric complex to bind effector substrates. Both IcmSand IcmW adopt new structural folds and have no structural similar-ities with known effector chaperones. IcmS has a compact globalstructure with an α/β fold, while IcmW adopts a fully α-folded, rel-atively loose architecture. IcmS stabilizes IcmW by binding to its twoC-terminal α-helices. Photocrosslinking assays revealed that theIcmS–IcmW complex binds its cognate effectors via an extended hy-drophobic surface, which can also interact with the C terminus ofDotL. A crystal structure of the DotL–IcmS–IcmW complex revealsextensive and highly stable interactions between DotL and IcmS–IcmW. Moreover, IcmS–IcmW recruits LvgA to DotL and assemblesa unique T4CP. These data suggest that IcmS–IcmW also functions asan inseparable integral component of the DotL–T4CP complex in thebacterial inner membrane. This study provides molecular insights in-to the dual roles of the IcmS–IcmW complex in T4BSSs.

Legionella pneumophila | type IVb secretion system | type IV couplingcomplex | adaptor proteins

Type IVb secretion systems (T4BSSs) play key roles in thepathogenesis of many Gram-negative bacterial pathogens, in-

cluding Legionella pneumophila, the causative agent of a severeform of human pneumonia known as Legionnaires’ disease, andCoxiella burnetii, a zoonotic pathogen that causes human Q fever byinjecting a number of virulent effector proteins into host cells tomanipulate host signaling (1). The L. pneumophila T4BSS iscomposed of 27 Dot (defect in organelle trafficking) or Icm (in-tracellular multiplication) proteins (2, 3), which form two proteinsubcomplexes in the cell membrane. The first of these sub-complexes is the major core–transmembrane complex, which con-sists of five proteins, namely, DotC, DotD, DotF, DotG, and DotH(4, 5). DotC, DotD, and DotH form an outer membrane pore,while DotF and DotG are located in the inner membrane and in-teract with DotC, DotD, and DotH to form a secretion channel (4).The second subcomplex in the L. pneumophila T4BSS is the

type IV coupling protein complex (T4CP), which is located in thebacterial inner membrane and is assembled by DotL, DotM, andDotN (5). DotL has sequence similarities to a family of ATPasesincluding TraG, TrwB, and TraD, which are known as the cou-pling proteins in type IVa secretion systems (T4ASSs, also known

as the VirB/D4 systems) (6). However, DotL contains an addi-tional C-terminal extension region that has ∼200 residues (6). Thecoupling proteins in T4ASSs function as inner membrane recep-tors, linking substrates to the secretion machinery and providingenergy for substrate translocation via ATPase activity (5).IcmS and IcmW are small, acidic proteins in L. pneumophila and

are known as adaptor proteins for T4BSSs (7–9). Both IcmS andIcmW are conserved in the T4BSS-containing pathogens C. burnetiiand Rickettsiella gryll. It has been suggested that IcmS and IcmWare located in both the cytoplasm and the inner membrane inL. pneumophila. They interact with one another to form a binarycomplex. It has been reported that the IcmS–IcmW complex canfunction as type III secretion chaperones (7). This binary complexcan bind many important effectors, including WipA, SidJ, SidD,SidF, and SidG, and regulate their secretion. IcmS and IcmW alsointeract with the C-terminal extension region of DotL and are re-quired for DotL stability during the transition from the exponentialto stationary phase of L. pneumophila growth (10, 11). In addition,LvgA, another small, acidic protein in L. pneumophila, can directly

Significance

Type IVb secretion systems are crucial for the pathogenesis ofLegionella pneumophila and Coxiella burnetii. IcmS and IcmWare known as adaptor proteins for the Legionella T4BSS andregulate the translocation of many virulent effector proteinsinto host cells. However, the mechanism by which IcmS–IcmWrecognizes its substrates and facilitates their delivery is unclear.We performed structural and biochemical analyses of theIcmS–IcmW complex. We found that the IcmS–IcmW complexharbors a distinct structure and binds its cognate effectors viaan extended hydrophobic surface. IcmS–IcmW also functionsas an inseparable partner of DotL to assemble a unique typeIV coupling protein complex. Our results provide mechanisticinsights into the dual roles of the IcmS–IcmW complexin T4BSSs.

Author contributions: J.X., D.X., L.W., and Y. Zhu performed research; M.W., L.Y., X.W.,Y.L., and X.L. contributed new reagents/analytic tools; J.X., D.X., Y. Zhou, and Y. Zhuanalyzed data; and J.X., D.X., Y. Zhou, and Y. Zhu wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Published under the PNAS license.

Data deposition: The atomic coordinates and structure factors have been deposited in theProtein Data Bank, www.wwpdb.org (PDB ID code 5XNB).1J.X. and D.X. contributed equally to this work.2To whom correspondence may be addressed. Email: zhouyanlsi@zju.edu.cn orzhuyongqun@zju.edu.cn.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1706883115/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1706883115 PNAS | December 19, 2017 | vol. 114 | no. 51 | 13543–13548

MICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Apr

il 10

, 202

0

interact with IcmS and plays an important role in bacterial in-tracellular growth in mouse macrophages (12). However, it is un-clear how IcmS–IcmW recognizes effector substrates and what theroles of the DotL–IcmS–IcmW and IcmS–LvgA interactions inT4BSSs are. In this study, we performed a biochemical and struc-tural investigation of the IcmS–IcmW complex and revealed thedual roles of IcmS–IcmW in T4BSSs.

ResultsIcmS and IcmW Form a 1:1 Heterodimeric Complex. We confirmedprevious observations that show that IcmS and IcmW are localizedin both the cytoplasm and the inner membrane in L. pneumophila(8, 11) (Fig. S1A). To investigate how IcmS and IcmW form abinary complex in the cytoplasm, we coexpressed IcmS and IcmWin Escherichia coli and purified the complex to be homogenous.IcmS and IcmW have 114 and 151 amino acids, respectively. TheIcmS–IcmW complex rendered an elution peak between 19 kDaand 33 kDa in gel filtration, which suggested that IcmS formed a1:1 heterodimeric complex with IcmW in solution (Fig. 1A).Coomassie brilliant blue staining of the eluted samples followingSDS/PAGE also confirmed that IcmS and IcmW were at an equalmolar ratio in the complex (Fig. 1A). IcmS–IcmW has been sug-gested to bind effector substrates at multiple internal regions (7).To investigate whether IcmS–IcmW binds to effector substrates asa 1:1 heterodimer, we coexpressed the IcmS–IcmW complex withdifferent binding fragments of its effector substrates, includingSidF (residues 1–760), SidG (residues 300–600), SidJ (residues89–265), and SidJ (residues 596–703). Affinity purificationrevealed that the IcmS–IcmW complex was indeed bound to theseeffectors as a 1:1 heterodimer (Fig. 1B).The inner membrane localization of IcmS and IcmW is likely

due to interactions between IcmS–IcmW and the C-terminal ex-tension region of DotL (10). We further tested whether IcmS–IcmW also functions as a 1:1 heterodimer to bind DotL. First,we narrowed down the IcmS–IcmW binding region in DotL. Thefragment including residues 595–783 of DotL could interact withIcmS–IcmW. A truncated fragment including residues 661–773 still had the capability to bind IcmS–IcmW. However, furtherdeletion abolished the interactions between IcmS–IcmW andDotL. We copurified IcmS–IcmW with the truncated fragmentthat includes residues 661–773 of DotL (hereafter referred to asDotLc). Indeed, IcmS–IcmW acted as a 1:1 heterodimer to bindDotLc and formed a 1:1:1 ternary complex in gel filtration (Fig.S1B). Taken together, these data suggest that IcmS and IcmWform a 1:1 heterodimeric complex to interact with effector sub-strates and the T4CP component DotL.

Overall Structure of IcmS–IcmW in Complex with DotLc. To investi-gate the mechanism of the IcmS–IcmW complex in T4BSS, we

performed structural studies. We failed to crystallize the IcmS–IcmW dimer alone or its complexes with effector substrates. TheDotLc–IcmS–IcmW ternary complex was highly stable and wasresistant to trypsin digestion and high concentration of salt insolution (Fig. S1 C and D). We succeeded in crystallizing theternary complex and determined its structure at a resolution of2.6 Å (Table S1).The overall structure of the DotLc–IcmS–IcmW ternary complex

adopts an inverted “U” shape (Fig. 2A). IcmS and IcmW are locatedat the two bottom arms of the complex. DotLc covers the IcmS–IcmW dimer from the upper side. In the complex, the IcmS–IcmWdimer exhibits a dumbbell-shaped architecture (Figs. 2A and 3A).Both IcmS and IcmW adopt new structural folds. They do not showany structural similarities with known proteins. IcmS has a compactglobal structure with an α/β-fold, which contains five α-helices (α1–5)and two β-strands (β1–2) (Fig. 2B). The three relatively longer helicesof IcmS, α1, α3, and α4, stack obliquely. The two short helices, α2 andα5, bind to the three longer helices from the bottom side. Theβ1 and β2 strands form an antiparallel β-sheet, which stacks withα3 and α4 from the upper side (Fig. 2B). Unlike the α/β architecture

Fig. 1. IcmS and IcmW form a 1:1 heterodimeric complex. (A) IcmS and IcmW form a 1:1 heterodimer in gel filtration. IcmS and IcmW were copurified using aHiLoad Superdex 75 16/600 column. The samples in the elution peak were analyzed via SDS/PAGE with Coomassie blue staining. (B) IcmS and IcmW functionsas a 1:1 heterodimer to bind effector substrates. SidF (residues 1–760), SidG (residues 300–600), SidJ (residues 89–265), and SidJ (residues 596–703) werecoexpressed with IcmS and IcmW. After GST copurification using glutathione resins, the samples were analyzed via SDS/PAGE with Coomassie blue staining.

Fig. 2. Structure of IcmS–IcmW in complex with DotLc. (A) The overallstructure of the IcmS–IcmW–DotLc ternary complex. IcmS, IcmW, and DotLc arecolored in cyan, green, and red, respectively. The N and C termini of the threeproteins are labeled as indicated. (B) The structure of IcmS. The secondarystructures of IcmS are labeled as indicated. (C) The structure of IcmW.

13544 | www.pnas.org/cgi/doi/10.1073/pnas.1706883115 Xu et al.

Dow

nloa

ded

by g

uest

on

Apr

il 10

, 202

0

of IcmS, IcmW has a fully α-folded structure, which contains eighthelices (α1–8) (Fig. 2C). The most notable feature of the IcmWstructure is that the last two helices (α7–8) in the C terminus protrudeout from its main body and are bound by IcmS. In the IcmS–IcmW–

DotLc ternary complex, DotLc contains a long loop at its N ter-minus, followed by three α-helices (Fig. 2A). This long loop mainlywinds around the two N-terminal α-helices (α1 and α2) of IcmW.The two middle α-helices of DotLc interact with IcmS and with theC-terminal–protruding α7 and α8 helices of IcmW. The last α-helix(αc) of DotLc extends via a loop and turns vertically to insert into alarge groove between α3 and α4 of IcmW at the arm side of theternary complex (Fig. 2A).

IcmS Stabilizes IcmW by Binding to Its Protruded C Terminus. IcmS islargely negatively charged on the surface. The compact structureof IcmS suggests that it is stable in solution. Indeed, IcmS wassoluble when expressed alone in E. coli. However, IcmS wasprone to aggregation in gel filtration (Fig. S1E). The structure ofIcmW adopts a relatively loose architecture (Fig. 2C). The mainbody of the IcmW structure is formed by helices α1–6 and theN-terminal part of α7. The C terminus of α7 and the wholeα8 helix protrude from the main body. The α5 helix is located atthe center of the IcmW main body and is surrounded by helicesα1–4 and α6. The N-terminal part of α7 binds to a groove be-tween α5 and α6 and interacts extensively with α2, α4, α5, and α6(Fig. 2C). The extensive interactions of α7 with the main bodysuggest that α7 plays a key role in IcmW structural assembly.It has been reported that IcmW protein levels in L. pneumophila

are severely reduced in the absence of IcmS (8). The relativelyloose structure of IcmW is consistent with the high instability ofIcmW. When overexpressed alone in E. coli, IcmW mainly pre-cipitated into inclusion bodies, which indicates that the IcmWprotein alone was not well folded. Coexpression with IcmS resultedin high solubility of IcmW (Fig. 1A), which suggests that IcmSstabilizes the IcmW structure and promotes its protein folding. Inthe complex structure, IcmS interacts with the C-terminal helicesα7 and α8 of IcmW (Fig. 3A). The interface includes 20 residuesfrom the α1, α3, and α4 helices of IcmS and from the α7 andα8 helices of IcmW. These interactions are mainly hydrophobicinteractions at the center of the interface. Five hydrogen bonds andsalt bridges strengthen the IcmS–IcmW interaction at the peripheryof the interface (Fig. 3B). We performed double mutations ofresidues C53 and L87 of IcmS and residues V134 and V143 ofIcmW, which are all located at the center of the interface, to serine

or aspartic acid residues. We tested the effects of these mutationson the IcmS–IcmW interaction and IcmW stability. The doublemutations of IcmS C53S/L87D or IcmW V134D/V143D com-pletely disrupted the IcmS–IcmW interaction and destabilizedIcmW, which resulted in the precipitation of IcmW into inclusionbodies during coexpression (Fig. 3C). Therefore, the interactions ofIcmS with the protruding C terminus of IcmW are necessary forthe structural stability of IcmW.

DotLc Is in an Extended Unfolded Conformation in Complex with IcmS–IcmW. In the complex with IcmS–IcmW, DotLc adopts an ex-tended unfolded conformation (Fig. S2A). The N-terminal loopand the middle α1 and α2 helices of DotLc assemble in a flatshape and are bound to the upper surface of IcmS–IcmW (Fig.S2B). The C-terminal αc helix of DotLc extends from the flat re-gion and turns vertically to be clamped by the α3 and α4 helices ofIcmW (Fig. 2A and Fig. S2C), which results in a horizontal “L”shape of DotLc. The secondary structural elements of DotLc donot have any internal interactions. The IcmS–IcmW complexbinds DotLc mainly via hydrophobic interactions. The 31 hydro-phobic interacting residues of DotLc constitute a continuous in-teraction network with IcmS and IcmW (Figs. S2 and S3). Thereare several hydrogen bonds and salt bridges surrounding thesehydrophobic interactions. The large interface between DotLc andthe IcmS–IcmW complex has an area of 3,074.5 Å2, which mostlycovers the entire upper surface of IcmS–IcmW. Because of thislarge interface, we generated double or triple mutations of theinteracting residues in IcmS, IcmW, or DotLc to test their effectson the interactions between DotLc and IcmS–IcmW. Consistentwith our observations of the ternary complex structure, the doublemutation of F27 and L128 in IcmW, which both interact with theN-terminal loop of DotLc, abolished the binding of IcmS–IcmWto DotLc (Fig. S3E). The residue L54 of IcmS and L49, V124, andL140 of IcmW interact with the middle and C-terminal α-helicesof DotLc. The combined mutations of these residues to alaninesalso disrupted the IcmS–IcmW interaction with DotLc (Fig. S3E).These mutations had no effect on the stability of the IcmS–IcmWcomplex. Correspondingly, the DotLc mutants L675A/F698A/L702A, F737A/I751A/L763A, and L675A/L763A could not bebound by IcmS–IcmW during copurification (Fig. S2D). Thesemutagenesis analyses suggest that these hydrophobic interactionsplay central roles in the interaction of DotLc with IcmS–IcmW.

The Effector Substrate-Binding Surface of IcmS–IcmW Overlaps withthe DotLc-Binding Surface. To investigate how IcmS–IcmW bindseffector substrates, we utilized a UV photocrosslinking assay toidentify the effector-binding surface on the IcmS–IcmW complex(Fig. S4A). Based on an orthogonal aminoacyl-tRNA synthetase/tRNA pair, the photocrosslinkable unnatural amino acid p-ben-zoyl-L-phenylalanine (pBpa) was site-specifically incorporated intoIcmS or IcmW at the position encoded by the amber codon (UAG)in E. coli (13). The incorporated pBpa around the effector-bindingsurface on IcmS or IcmW cross-linked with the effector protein,which was coexpressed with IcmS and IcmW, upon excitation byUV light at a wavelength of 365 nm. The cross-linked sites in IcmSor IcmW were determined by detecting cross-linked products ofIcmS or IcmW with the effector protein and were mapped onto thesurface of the IcmS–IcmW complex structure.IcmS–IcmW has many effector substrates and can bind one

effector at multiple sites. To map the effector-binding surface ascompletely as possible, a stable and large effector was needed toserve as the substrate for IcmS–IcmW in photocrosslinking assays.SidF is a large effector substrate of IcmS–IcmW (7). A fragment ofSidF including its N-terminal 760 residues (residues 1–760) couldinteract with IcmS–IcmW and function as its substrate. SidF (1–760) was soluble and stable when expressed alone or when coex-pressed with IcmS–IcmW. We coexpressed SidF (1–760) withIcmS–IcmW to map the effector-binding surface. In total, we

Fig. 3. Detailed interactions between IcmS and IcmW. (A) The structure ofIcmS–IcmW in the complex with DotLc. (B) Detailed interactions between IcmSand the α7 and α8 helices of IcmW. IcmS, and IcmW are colored in cyan andgreen, respectively. The interacting residues are shown as sticks. The hydrogenbonds and salt bridges are shown as dashed lines. (C) Mutation effects of theinteracting residues in IcmS and IcmW on the IcmS–IcmW interactions. His6–SUMO-IcmS and GST-IcmW were coexpressed in E. coli. The interactions be-tween IcmS and IcmW were examined by Ni-NTA resin affinity copurification.

Xu et al. PNAS | December 19, 2017 | vol. 114 | no. 51 | 13545

MICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Apr

il 10

, 202

0

mutated the codons of 62 surface residues in the IcmS–IcmWcomplex, including 45 residues from IcmW and 17 residues fromIcmS, to TAG codons for pBpa incorporation (Table S2). Theseresidues are distributed on the entire surface of the IcmS–IcmWcomplex. Detection of the IcmW–SidF and IcmS–SidF cross-linked products by SDS/PAGE revealed that 5 sites from IcmSand 12 sites from IcmW cross-linked with SidF (1–760) upon UVexcitation (Fig. 4 and Table S2). Among these sites, E66, K120,F27, V31, and R127 of IcmW and L54 of IcmS had the strongestcross-linking abilities with SidF. We mapped all 17 cross-linkablesites onto the IcmS–IcmW structure. Other than S98 of IcmW,which is located on the arm side of the IcmS–IcmW complex andshowed an extremely weak cross-linking ability with SidF, theother 16 sites are all located at the edge of the DotLc-bindingsurface (Fig. 4D), which indicates an overlapping binding surfacebetween the effectors and DotLc on IcmS–IcmW. Consistently,SidF could not form a tetramer with DotLc during coexpressionwith IcmS–IcmW. We also used SidD as the substrate of IcmS–IcmW in the photocrosslinking assays. The SidD-binding surfaceon IcmS–IcmW is similar to the surface that SidF binds to (Figs.S5 and S6). Validation of the DotLc interactions with IcmS–IcmWin photocrosslinking assays further confirmed that the DotLc-binding surface on IcmS–IcmW is the same surface that is ob-served in the ternary complex structure (Fig. S4 B–E). Therefore,IcmS–IcmW binds to effector substrates and DotLc through thesame hydrophobic surface.

IcmS–IcmW Recruits LvgA to DotL and Assembles a Unique T4CP. It wasreported that IcmS could interact with LvgA (12). We examinedwhether IcmS–IcmW and the DotLc–IcmS–IcmW complex couldinteract with LvgA. When coexpressed with LvgA in E. coli, IcmS–IcmW indeed interacted with LvgA (Fig. S7A). However, theIcmS–IcmW–LvgA complex could not interact with SidD or SidFsubstrate (Fig. S7 B–D), suggesting that LvgA is not a componentof the IcmS–IcmW adaptor complex in the cytoplasm. Whencoexpressed with the DotLc, IcmS, and IcmW in E. coli, LvgAformed a highly stable tetramer with DotLc, IcmS, and IcmW atan equal molar ratio in gel filtration (Fig. 5A). In contrast, theIcmS–IcmW–LvgA complex could not interact with DotD, anouter membrane component of the L. pneumophila T4BSS (Fig.5B). We further tested whether the DotLc–IcmS–IcmW–LvgAcomplex could bind the intracellular component, DotN, or thecytoplasmic domain of the DotM subunit of the T4CP. We found

that a longer C-terminal fragment (residue 595–783) of DotL, butnot DotLc, could interact with IcmS, IcmW, LvgA, and DotN, andform a stable pentamer in gel filtration (Fig. 5C).The findings that the DotLc–IcmS–IcmW complex was resistant

to trypsin digestion and high salt concentrations and that there areextensive hydrophobic interactions between IcmS–IcmW andDotLc suggest that IcmS–IcmW is inseparable from DotL. Theidentification of the stable DotL–IcmS–IcmW–LvgA tetramer ingel filtration experiments (Fig. 5A) indicates that IcmS–IcmW cannot only function as a secretion adaptor in the bacterial cytoplasm,but is also an inseparable, stable binding partner of DotL to recruitLvgA and assemble a unique T4CP in the bacterial inner mem-brane. This finding is consistent with the previously reportedmembrane localization of LvgA (12). As expected, the DotLc–IcmS–IcmW–LvgA complex could not bind IcmS/IcmW-dependent effectors, including SidD, SidJ, SidG, and SidF(Fig. S7E). Given that LvgA is also a small acidic protein, similarto type III chaperones, it is possible that the DotLc–IcmS–IcmW–

LvgA complex can recruit LvgA-dependent or other effector proteins.

DiscussionT4BSSs play key roles in the pathogenesis of many bacterialpathogens (1). IcmS and IcmW are important adaptor proteins inT4BSSs. In this study, we demonstrated that IcmS and IcmW inthe Legionella T4BSS can function as a 1:1 heterodimeric adaptorto bind effector substrates in the bacterial cytoplasm and are alsoinseparable partners of DotL that recruit LvgA and assemble aunique T4CP in the inner membrane for T4BSSs. IcmS and IcmWhave no structural similarities with known export chaperones.DotLc in complex with IcmS–IcmW is in an extended unfoldedconformation. The extensive hydrophobic interactions betweenIcmS–IcmW and DotLc stabilize DotL and prevent its degrada-tion by the bacterial cytoplasmic protease ClpP (5).Our photocrosslinking assays revealed that the effector

substrate-binding surface on IcmS–IcmW overlaps with theDotLc-binding surface, suggesting that IcmS–IcmW utilizes thesame binding surface to bind effectors and the T4CP componentDotL. Many large effectors in Legionella, including SidJ, SidF,and SidG, have been identified as substrates of IcmS–IcmW (7, 8).These effectors have long sequences of amino acids and containmultiple domains. IcmS–IcmW has multiple binding regions inthese effectors. The multiple-binding mode of IcmS–IcmW likely

Fig. 4. Photocrosslinking analysis of the effector-binding surface in IcmS–IcmW. (A and B) Photocrosslinking analysis of the residues in IcmW. The codons ofthe analyzed residues in IcmW were mutated to TAG for pBpa incorporation during coexpression with IcmS and SidF (residues 1–760). The cross-linkingreaction was excited by UV light (365 nm). The cross-linked products of IcmW with SidF were analyzed by SDS/PAGE with anti-HA and anti-Flag antibodies.IcmW in the samples was immunoblotted with the anti-Flag antibody. (C) Photocrosslinking analysis of the residues in IcmS. The cross-linked products of IcmSwith SidF (residues 1–760) were analyzed via SDS/PAGE and immunoblotting with anti-HA and anti-Myc antibodies. IcmS in the samples was immunoblottedwith the anti-Myc antibody. (D) Mapping the cross-linkable residues on the surface of IcmS–IcmW. The cross-linkable residues of IcmS and IcmW are high-lighted in green on the surface of the IcmS–IcmW complex.

13546 | www.pnas.org/cgi/doi/10.1073/pnas.1706883115 Xu et al.

Dow

nloa

ded

by g

uest

on

Apr

il 10

, 202

0

provides an efficient way to maintain the large type IV effectorproteins in unfolded conformations and prevent their aggregationduring translocation. DotL and effector proteins bind to the samesurface on IcmS–IcmW, which indicates that a similar extendedunfolded conformation is likely adopted by effectors when they arein complex with IcmS–IcmW.We found that the effector SidJ couldnot compete with SidF on the purified SidF–IcmS–IcmW complex(Fig. S6), suggesting that an unknown factor is required to releaseeffectors from IcmS–IcmW for effector translocation and adaptorcycling.Based on our results, we propose a working model for IcmS–

IcmW in the T4BSS of L. pneumophila. In the cytoplasm, IcmSand IcmW form a 1:1 heterodimer to interact with effector sub-strates. The interactions maintain effectors in unfolded or partiallyunfolded states. Through the help of the ATPase DotB or otherunknown factors, these effector substrates are released from IcmS–IcmW and are transferred to the protein coupling complex forT4BSSs. IcmS–IcmW also functions as an inseparable partner ofDotL and an integral component of T4CP. IcmS–IcmW stabilizesDotL and recruits LvgA to assemble a unique T4CP with DotL andDotN in the inner membrane. The DotL–IcmS–IcmW–LvgAcomplex possibly functions as a receptor to recruit LvgA-dependentor other effectors for the Legionella T4BSS. Effector release fromthe DotL–IcmS–IcmW–LvgA complex is likely carried out by DotL.Recently, Kwak et al. (14) reported crystal structures of the

DotL (residues 590–659)–DotN complex and the DotL (residues656–783)–IcmS–IcmW–LvgA complex. These two structureshighlighted the roles of IcmS, IcmW, and LvgA as inseparableintegral components of the DotL–T4CP subcomplex in the innermembrane, which is consistent with our findings in this study.They also observed direct interactions of the DotL–IcmS–IcmW–LvgA complex with two effectors, VpdB and SetA (14).These two effectors share no sequence homology with each other

and do not contain classical secretion signals at their C terminifor the Dot/Icm T4BSS. There was no evidence that the trans-location of VpdB and SetA via the T4BSS was dependent onIcmS or IcmW. These data suggest that VpdB and SetA belongto another group of Leginonella effectors. In our studies, wefound that the IcmS–IcmW heterodimeric complex, but not theIcmS–IcmW–LvgA, IcmS–IcmW–DotL, or IcmS–IcmW–DotL–LvgA complexes, could directly interact with and functioned asan adaptor for IcmS/IcmW-dependent effectors, including SidG,SidJ, SidD, and SidF in the cytoplasm. We also found that theeffector-binding surface on the IcmS–IcmW complex overlapswith the DotL-binding surface, which suggests that IcmS–IcmWutilizes the same surface to recognize IcmS/IcmW-dependenteffectors and interact with the T4CP component DotL. Ourstudies, together with the discoveries by Kwak et al. (14), revealthe dual roles for the IcmS–IcmW complex in the type IV cou-pling complex and provide molecular insights into the mecha-nism of the IcmS–IcmW complex in T4BSS effector secretion.

Materials and MethodsPlasmids, Reagents, Strains, and Protein Purification. DNA of IcmS, IcmW,LvgA, DotL, DotN, and the effectors SidF, SidD, SidJ, and SidG was amplifiedfrom the L. pneumophila Lp02 strain. All recombinant proteins wereexpressed in the E. coli BL21 (DE3) strain. Plasmids and primers used in thisstudy are listed in Dataset S1. For details, see SI Materials and Methods.

Pull-Down, GST Copurification, and Photocrosslinking Assays. Ni-NTA resin,anti-Flag beads, or glutathione resin were used for Ni-NTA, Flag pull-down,and GST copurification assays, respectively. The photocrosslinking assaysamples were analyzed by SDS/PAGE and immunoblotting. For details, see SIMaterials and Methods.

Crystallization and Structure Determination. The crystal diffraction data of theIcmS–IcmW–DotLc complex were processed with the XDS package (15). The

Fig. 5. IcmS and IcmW recruits LvgA to DotL and assembles a unique T4CP. (A) IcmS–IcmW forms a stable tetramer with DotLc and LvgA. The DotLc–IcmS–IcmW–LvgA (residues 1–186) complex was purified using a HiLoad Superdex 75 16/600 column. The samples in the elution peak were analyzed by SDS/PAGEwith Coomassie blue staining. (B) IcmS–IcmW and LvgA specifically interacts with DotL, but not DotD. The interaction of DotL or DotD with IcmS–IcmW andLvgA was detected by GST pull-down assay by using GST-fused LvgA as bait after coexpression. (C) Interaction of DotN with the DotL–IcmS–IcmW–LvgAcomplex. A longer fragment (residues 595–783) of DotL was coexpressed with IcmS–IcmW, LvgA, and DotN in E. coli. The DotN–DotL–IcmS–IcmW–LvgApentamer was purified using a HiLoad Superdex 200 16/600 column.

Xu et al. PNAS | December 19, 2017 | vol. 114 | no. 51 | 13547

MICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Apr

il 10

, 202

0

structure was solved in Phenix (16). Model building was carried out in Coot(17). Structure refinement was carried out in Phenix and Refmac5 in CCP4 (18).The final structural model was checked with PROCHECK (19). The data col-lection and refinement statistics are listed in Table S1. For details, see SI Ma-terials and Methods.

ACKNOWLEDGMENTS. We thank Drs. Peng Chen and Zengyi Chang forproviding plasmids and reagents for photocrosslinking assays and the staff

at beamlines BL17U1 and BL18U1 of the Shanghai Synchrotron RadiationFacility and the National Center for Protein Science Shanghai forassistance with diffraction data collection. This research was supportedby National Natural Science Foundation of China Grants 81530068,81322024, 31370722, 81561130162 (to Y. Zhu) and 81501717 (toY. Zhou) and the Fundamental Research Funds for the Central Universi-ties. Y. Zhu was awarded the Newton Advanced Fellowship by the RoyalSociety.

1. Nagai H, Kubori T (2011) Type IVB secretion systems of Legionella and other Gram-negative bacteria. Front Microbiol 2:136.

2. Segal G, Purcell M, Shuman HA (1998) Host cell killing and bacterial conjugation re-quire overlapping sets of genes within a 22-kb region of the Legionella pneumophilagenome. Proc Natl Acad Sci USA 95:1669–1674.

3. Vogel JP, Andrews HL, Wong SK, Isberg RR (1998) Conjugative transfer by the viru-lence system of Legionella pneumophila. Science 279:873–876.

4. Kubori T, et al. (2014) Native structure of a type IV secretion system core complexessential for Legionella pathogenesis. Proc Natl Acad Sci USA 111:11804–11809.

5. Vincent CD, et al. (2006) Identification of the core transmembrane complex of theLegionella Dot/Icm type IV secretion system. Mol Microbiol 62:1278–1291.

6. Buscher BA, et al. (2005) The DotL protein, a member of the TraG-coupling proteinfamily, is essential for viability of Legionella pneumophila strain Lp02. J Bacteriol 187:2927–2938.

7. Cambronne ED, Roy CR (2007) The Legionella pneumophila IcmSW complex interactswith multiple Dot/Icm effectors to facilitate type IV translocation. PLoS Pathog 3:e188.

8. Ninio S, Zuckman-Cholon DM, Cambronne ED, Roy CR (2005) The Legionella IcmS-IcmW protein complex is important for Dot/Icm-mediated protein translocation. MolMicrobiol 55:912–926.

9. Zamboni DS, McGrath S, Rabinovitch M, Roy CR (2003) Coxiella burnetii express typeIV secretion system proteins that function similarly to components of the Legionellapneumophila Dot/Icm system. Mol Microbiol 49:965–976.

10. Sutherland MC, Nguyen TL, Tseng V, Vogel JP (2012) The Legionella IcmSW complexdirectly interacts with DotL to mediate translocation of adaptor-dependent sub-strates. PLoS Pathog 8:e1002910.

11. Vincent CD, Friedman JR, Jeong KC, Sutherland MC, Vogel JP (2012) Identification of

the DotL coupling protein subcomplex of the Legionella Dot/Icm type IV secretion

system. Mol Microbiol 85:378–391.12. Vincent CD, Vogel JP (2006) The Legionella pneumophila IcmS-LvgA protein complex

is important for Dot/Icm-dependent intracellular growth. Mol Microbiol 61:596–613.13. Ryu Y, Schultz PG (2006) Efficient incorporation of unnatural amino acids into pro-

teins in Escherichia coli. Nat Methods 3:263–265.14. Kwak MJ, et al. (2017) Architecture of the type IV coupling protein complex of Le-

gionella pneumophila. Nat Microbiol 2:17114.15. Kabsch W (2010) XDS. Acta Crystallogr D Biol Crystallogr 66:125–132.16. Adams PD, et al. (2010) PHENIX: A comprehensive Python-based system for macro-

molecular structure solution. Acta Crystallogr D Biol Crystallogr 66:213–221.17. Emsley P, Cowtan K (2004) Coot: Model-building tools for molecular graphics. Acta

Crystallogr D Biol Crystallogr 60:2126–2132.18. Collaborative Computational Project, Number 4 (1994) The CCP4 suite: Programs for

protein crystallography. Acta Crystallogr D Biol Crystallogr 50:760–763.19. Laskowski AR, MacArthur WM, Moss SD, Thornton MJ (1993) PROCHECK: A program

to check the stereochemical quality of protein structures. J Appl Crystallogr 26:

283–291.20. Dormán G, Prestwich GD (1994) Benzophenone photophores in biochemistry.

Biochemistry 33:5661–5673.21. Wittelsberger A, Mierke DF, Rosenblatt M (2008) Mapping ligand-receptor interfaces:

Approaching the resolution limit of benzophenone-based photoaffinity scanning.

Chem Biol Drug Des 71:380–383.

13548 | www.pnas.org/cgi/doi/10.1073/pnas.1706883115 Xu et al.

Dow

nloa

ded

by g

uest

on

Apr

il 10

, 202

0