Structure of an EIIC sugar transporter trapped in aninward-facing conformationZhenning Rena,1, Jumin Leeb,1, Mahdi Muhammad Moosac, Yin Niana,d, Liya Hua, Zhichun Xua,d, Jason G. McCoya,Allan Chris M. Ferreonc,2, Wonpil Imb,2, and Ming Zhoua,d,2

aDepartment of Biochemistry and Molecular Biology, Baylor College of Medicine, Houston, TX 77030; bDepartment of Biological Sciences, LehighUniversity, Bethlehem, PA 18015; cDepartment of Pharmacology and Chemical Biology, Baylor College of Medicine, Houston, TX 77030; and dKunmingInstitute of Zoology, Chinese Academy of Sciences, Kunming 650223, China

Edited by Ernest M. Wright, David Geffen School of Medicine at UCLA, Los Angeles, CA, and approved April 30, 2018 (received for review January 12, 2018)

The phosphoenolpyruvate-dependent phosphotransferase system(PTS) transports sugar into bacteria and phosphorylates the sugarfor metabolic consumption. The PTS is important for the survival ofbacteria and thus a potential target for antibiotics, but itsmechanism of sugar uptake and phosphorylation remains unclear.The PTS is composed of multiple proteins, and the membrane-embedded Enzyme IIC (EIIC) component transports sugars acrossthe membrane. Crystal structures of two members of the glucosesuperfamily of EIICs, bcChbC and bcMalT, were solved in the inward-facing and outward-facing conformations, and the structures sug-gest that sugar translocation could be achieved by movement ofa structured domain that contains the sugar-binding site. How-ever, different conformations have not been captured on thesame transporter to allow precise description of the conforma-tional changes. Here we present a crystal structure of bcMalT trap-ped in an inward-facing conformation by a mercury ion thatbridges two strategically placed cysteine residues. The structureallows direct comparison of the outward- and inward-facing con-formations and reveals a large rigid-body motion of the sugar-binding domain and other conformational changes that accompanythe rigid-body motion. All-atom molecular dynamics simulationsshow that the inward-facing structure is stable with or withoutthe cross-linking. The conformational changes were further val-idated by single-molecule Föster resonance energy transfer(smFRET). Combined, these results establish the elevator-typemechanism of transport in the glucose superfamily of EIICtransporters.

bcMalT | double-cysteine cross-linking | inward-facing conformation |single-molecule FRET | elevator-type mechanism of transport

The phosphoenolpyruvate:carbohydrate phosphotransferasesystem (PTS) is ubiquitous in bacteria. The membrane-embedded component EIIC transports a sugar from the peri-plasmic side to the cytoplasmic side where the sugar is phos-phorylated by the EIIB protein and then released into thecytoplasm. Phosphorylation occurs while the sugar is still boundto EIIC, and the phosphorylation facilitates the release of thesugar (1). The phosphate group originates from phosphoenol-pyruvate and is transferred sequentially through four solublephosphocarrier proteins, enzyme I (EI), the histidine phospho-carrier protein (HPr), enzyme IIA (EIIA), and enzyme IIB(EIIB) and eventually reaches the incoming sugar while it is stillbound on EIIC (SI Appendix, Fig. S1). Phosphorylation of theincoming sugar helps maintain its concentration gradient tosustain the uptake, and the energy stored in the phosphate bondis recovered when the sugar is metabolized (2–5). Thus, EIIsystems are more efficient than other membrane-embeddedtransporters that either hydrolyze ATP or dissipate an ion gra-dient. EII systems have been categorized into four superfamilies,of which the largest is the glucose superfamily (6) and the focusof this research. Each EII has its own preference for a group ofsugar molecules and is composed of three proteins, EIIA, B, andC. In some of the EIIs, EIIB and C, or EIIA, B, and C form a

single polypeptide. Crystal structures of two EIICs from theglucose superfamily (7, 8) and one from the ascorbate andgalactitol superfamily (9) have been reported. Several electronmicroscopy projection maps of two EIICs from the glucose su-perfamily have also been reported (10–12). While members ofthe same superfamily share the same structural fold, those fromdifferent superfamilies do not.Structures of an N,N′-diacetylchitobiose EIIC transporter

bcChbC (7) and a maltose EIIC transporter bcMalT (8), bothfrom Bacillus cereus, have been reported. bcChbC and bcMalTshare 19% sequence identity and 50% similarity, and yet bothhave the same structural fold with almost all of the secondarystructural elements conserved (SI Appendix, Fig. S2 A–E). Bothproteins are homodimers, and each protomer has 10 trans-membrane (TM1-10) helices, two reentrant loops (HP1-2), andtwo amphipathic helices (AH1-2). These structural elementsfold into two distinctive structural domains. The dimerizationdomain (also referred to as the interface domain), which con-sists of TM1-5 and AH1, forms an expansive dimer interface.The substrate-binding domain (also referred to as the transportdomain), which is composed of TM6-10 and two reentrantloops (HP1-2), contains the sugar-binding site. In both struc-tures, the sugar substrate is coordinated by residues from TM6,TM7, HP1, and HP2 (SI Appendix, Fig. S2B). The two domains

Significance

The phosphoenolpyruvate-dependent phosphotransferase sys-tem (PTS) is a multiprotein system unique to bacteria. The PTStransports sugars into bacteria and then phosphorylates thesugars. Phosphorylation prevents sugars from escaping the celland primes them for metabolic consumption. As a major com-ponent of the PTS, Enzyme IIC (EIIC) transports sugar acrossthe membrane and assists the phosphorylation process, butthe molecular mechanism of EIIC-mediated sugar transport isunclear. Results from this study allow visualization of confor-mational changes during sugar transport and establish themechanism of transport at the atomic level. The knowledge willfacilitate development of inhibitors against EIIC and provide afoundation for understanding the phosphorylation process.

Author contributions: Z.R., J.L., M.M.M., Y.N., L.H., Z.X., J.G.M., A.C.M.F., W.I., and M.Z.designed research, collected and analyzed data, and 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 6BVG).1Z.R. and J.L. contributed equally to this work.2To whom correspondence may be addressed. Email: mzhou@bcm.edu, allan.ferreon@bcm.edu, or woi216@lehigh.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplemental.

Published online May 21, 2018.

5962–5967 | PNAS | June 5, 2018 | vol. 115 | no. 23 www.pnas.org/cgi/doi/10.1073/pnas.1800647115

Dow

nloa

ded

by g

uest

on

Apr

il 7,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://crossmark.crossref.org/dialog/?doi=10.1073/pnas.1800647115&domain=pdfhttp://www.pnas.org/site/aboutpnas/licenses.xhtmlhttp://www.wwpdb.orghttp://www.rcsb.org/pdb/explore/explore.do?structureId=6BVGmailto:mzhou@bcm.edumailto:allan.ferreon@bcm.edumailto:allan.ferreon@bcm.edumailto:woi216@lehigh.eduhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalwww.pnas.org/cgi/doi/10.1073/pnas.1800647115

are bridged by an amphipathic helix (AH2) (SI Appendix, Fig.S2 C and D).The bcChbC and bcMalT structures represent different con-

formations required to complete a transport cycle (13). Based onthe location of the substrate-binding site, bcChbC is in aninward-facing conformation, while bcMalT is in an outward-facing conformation (SI Appendix, Fig. S2A). When the twostructures are aligned by their dimerization domains, it seemsthat the substrate-binding domain could carry the substrateacross the membrane by a rigid-body motion. A similar elevator-type transport mechanism has also been reported in a number ofsecondary solute transporters including amino acid transporters(EAAT1 and GltPh) (14–16), bile acid transporters (ASBT) (17),

proton sodium exchangers (NhaA) (18, 19), concentrative nu-cleotide transporters (CNTNW) (20), and citrate transporters(vcINDY and SeCitS) (21, 22). These transporters have differentstructural folds, and yet they all transport substrates from oneside of the cell membrane to the other by rigid-body motions of asubstrate-binding domain.Although by comparing the inward-facing bcChbC structure

and the outward-facing bcMalT structure we can postulate thatthe glucose superfamily of EIICs have an elevator-type mecha-nism of transport, we need to visualize both conformations onthe same transporter to reveal details of the conformationalchanges. To achieve this, we first generated a structural model ofbcMalT in an inward-facing conformation by collective variable-based steered molecular dynamics (CVSMD) simulation (SIAppendix, Fig. S2F) using the bcChbC structure as a guide (23).During the simulation, the interface domain was kept static, andthe substrate-binding domain was steered toward the inward-facing conformation. Since the substrate-binding domain movesrelative to the interface domain, we expect distance changesbetween the two domains. Indeed, the CVSMD model showsthat residues that are far away from each other in the outward-facing structure become closer, for example, residues T280 andD55 and residues N284 and E54 (SI Appendix, Fig. S2G). Wethen showed that the pairs of residues predicted to becomecloser to each other can be cross-linked by a mercury ion whenmutated to cysteine residues and thus provide an experimentalvalidation to the CVSMD model and the elevator-type mecha-nism of transport (23).In this study, we solved the crystal structure of bcMalT cross-

linked in an inward-facing conformation. The structure providesdirect experimental evidence that the substrate-binding domaincan undergo a rigid-body rotation toward the intracellular side.The structure also shows conformational changes in otherregions of the transporter that accommodate the rigid-body

Fig. 1. Cross-linking of bcMalT(T280C-E54C). (A) SDS/PAGE analysis ofbcMalT(T280C-E54C) under various conditions. Lanes: 1, molecular weightstandard; 2, bcMalT(T280C-E54C) without any treatment; 3, bcMalT(T280C-E54C) treated with Hg2+. Cross-linked bcMalT migrates faster, and the bandis marked as X-linked; 4, bcMalT(T280C-E54C), first treated with 2 mM N-ethyl maleimide (NEM) and then incubated with 1:1 molar ratio of Hg2+; 5,bcMalT(T280C-E54C) in the presence of 2 mM β-mercaptoethanol (β-ME); 6,bcMalT(T280C-E54C) treated with Hg2+ and then incubated with 2 mM β-ME.(B) Time course of Hg2+-induced cross-linking of bcMalT(T280C-E54C).

Fig. 2. Side-by-side comparison of bcMalT and bcMalT-X. Crystal structures of bcMalT wild type (Left, PDB accession number 5IWS) and bcMalT-X (Right) areshown in cartoon (A) and surface (B) representations. In A, the two protomers are colored blue and green, respectively. Within each protomer, the di-merization domain is colored darker than the substrate-binding domain. In B, the electrostatic surface was calculated by the Adaptive Poisson–BoltzmannSolver plugin in Pymol (34); the bar represents electrostatic potential from −61.1 to +61.1 kBT/eC. (C) Location of T280 and E54 on a single protomer in theoutward-facing conformation (Left) and inward-facing conformation (Right). TM1 is removed for clarity. Cβ of T280 (magenta) and E54 (orange) are shown asspheres. (D) Superposition of the dimerization domain (Left) and substrate-binding domain (Right). Green, bcMalT wild type; blue, bcMalT-X.

Ren et al. PNAS | June 5, 2018 | vol. 115 | no. 23 | 5963

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Apr

il 7,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplemental

movement of the substrate-binding domain. Since cross-linkingof a pair of residues could potentially trap the structure into aconformation that may not be native, we did two experiments toexamine whether the structure is distorted. First, we applied all-atom molecular dynamics simulations to the structure and foundthat the inward-facing conformation is stable even when the cross-linking constraint is released. Second, we estimated the distancebetween a pair of symmetry-related residues on the substrate-binding domain by single-molecule Föster resonance energytransfer (smFRET) and found that the distances between thetwo residues are consistent with those measured from the in-ward- and outward-facing structures.

ResultsTrapping bcMalT in an Inward-Facing Conformation. CVSMD wasused to show that residues T280 on the substrate binding domainand D55 on the dimerization domain can move to within 12.4 Å ofeach other as opposed to 23.7 Å in the outward facing crystalstructure (SI Appendix, Fig. S2G). Similarly, N284 on the substrate-binding domain and E54 on the dimerization domain are 26.7 Åaway in the crystal structure and 13.1 Å in the CVSMD model.Both pairs can be cross-linked by a mercury ion when mutated tocysteines, thus confirming their proximity (23). Although we werenot able to crystallize the cross-linked proteins, we were encour-aged by the predictive power of the inward-facing model and testedadditional pairs in the vicinity of the previous successful ones. TheCα atoms of the T280-E54 pair are 19.3 Å apart in outward-facingbcMalT and 9.6 Å apart in the inward-facing model. When bothresidues were mutated to cysteines (T280C-E54C) they can becross-linked at low (micromolar) concentrations of Hg2+ and at aprotein-to-mercury molar ratio of ∼1:1. The cross-linked proteinmigrates faster on SDS/PAGE (Fig. 1A, lanes 2 and 3). The cross-linking is complete within seconds (Fig. 1B). These results suggestthat the two residues have a high probability of reaching proximity.

We then showed that the cross-linking is specific to the two cys-teine mutations because the reaction can be reversed by addition ofreducing reagent β-mercaptoethanol (Fig. 1A, lane 6), prevented bypretreatment with N-ethyl maleimide that alkylates free cysteines(Fig. 1A, lane 4), and did not happen on single-cysteine mutations(SI Appendix, Fig. S3A).We then assessed the functions of the T280C-E54C double-

cysteine mutation in bcMalT. The double mutant transportsmaltose through facilitated diffusion when reconstituted into li-posomes, suggesting that the cysteine mutations do not affect thefunction of the protein significantly (SI Appendix, Fig. S3B). Wealso found that the double-cysteine mutant binds to maltose withsimilar affinity before and after cross-linking (SI Appendix, Fig.S3C), suggesting that cross-linking the two residues does notsignificantly distort the sugar-binding site.

Crystal Structure of the Cross-Linked bcMalT EIIC. The T280C-E54CbcMalT EIIC cross-linked by Hg2+ (termed bcMalT-X hereafter)can be crystallized. After extensive refinement of crystallizationconditions, the crystals diffracted to a resolution of 3.2 Å, and afull dataset was collected and processed (SI Appendix, TableS1). Molecular replacement using individual interface andtransport domains led to a clear solution (SI Appendix, Fig. S4 Aand B). A nonprotein density is resolved at the substrate-bindingsite into which we can unambiguously build a maltose (SI Ap-pendix, Fig. S4C). Since the dataset was collected at a wavelengthclose to the mercury edge, we calculated an anomalous differ-ence Fourier map and it shows a strong peak (>9σ) that iden-tifies the Hg2+ position between cysteines 280 and 54 (SIAppendix, Fig. S4D). The distance between the Cα atoms ofcysteine 54 and 280 is 7.7 Å, and the distances between the Hgand the two sulfur atoms are ∼2.3 Å. These distances are con-sistent with those reported in previous crystal structures of Hg2+

-mediated cysteine cross-links (14, 24). The final refined modelcontains a homodimer of bcMalT, and within each protomer,residues 5–451, one Hg2+ and one maltose are resolved.When bcMalT and bcMalT-X structures were shown side-by-

side with their AH1 and AH2 aligned to either side of themembrane (Fig. 2 A and B), the bound substrate on bcMalT-X iscloser to the intracellular side, and its substrate-binding domainassumes an inward-facing conformation (Fig. 2A). The distancebetween residues 280 and 54 is smaller in the bcMalT-X struc-ture due to the cross-linking (Fig. 2C). However, the root-mean-square deviation (rmsd) of all Cα atoms in the dimerizationdomain between the two structures is 0.64 Å and that of thesubstrate-binding domain is 0.43 Å, indicating that the cross-linking does not change the arrangement of the secondarystructural elements within each domain (Fig. 2D). The structureprovides direct evidence that a rigid-body elevator-type motionof the substrate-binding domain carries a substrate from one sideof the membrane to another (Movie S1).

Substrate-Binding Site. Similar to the bcMalT structure, thebcMalT-X structure also has a nonprotein electron density cor-responding to a bound maltose on each protomer (Fig. 3 and SIAppendix, Fig. S4C). Since there is very little change in thesubstrate-binding domain, the coordination of the bound malt-ose is essentially identical in the two structures. The C6 hydroxylon the nonreducing end is known to be phosphorylated by EIIB,and the hydroxyl is coordinated by E355 and H240 from HP1band TM7, respectively (Fig. 3). These two residues are impli-cated in the phosphotransfer reaction and are highly conservedin the glucose EIIC superfamily (25, 26) (SI Appendix, Fig. S5).The bound maltose has one additional interaction with theprotein in the bcMalT-X structure. Y249 on TM7 likely forms aweak hydrogen bond with the C1 hydroxyl on the reducing end(Fig. 3). This additional hydrogen bond does not seem to affectbinding affinity (SI Appendix, Fig. S3B). The distance between

Fig. 3. Stereoview of the maltose-binding site in bcMalT and bcMalT-X.(A) bcMalT wild type; (B) bcMalT-X. Maltose is shown as sticks with carbonin yellow and oxygen in red. Protein–maltose interactions are marked asdotted lines between the sugar and residues from HP1 (blue), HP2 (raspberry),and TM6,7,8 (light green). The C6-OH on the nonreducing end of maltosewhere phosphorylation happens is marked with a star.

5964 | www.pnas.org/cgi/doi/10.1073/pnas.1800647115 Ren et al.

Dow

nloa

ded

by g

uest

on

Apr

il 7,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://movie-usa.glencoesoftware.com/video/10.1073/pnas.1800647115/video-1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalwww.pnas.org/cgi/doi/10.1073/pnas.1800647115

the hydroxyl on Y249 and the hydroxyl on the C1 is 3.1 Å inbcMalT-X and 4.5 Å in bcMalT, and the difference is due toslight movement of TM7. Coincidentally, in a previous moleculardynamics simulation study performed on the outward-facingbcMalT, Y249 and TM7 were found to be more mobile than theneighboring structural element, moving away from the substrateto open the entrance to the substrate-binding site (8).

Major Differences Between the Inward- and Outward-FacingStructures. The rigid-body motion of the transport domain re-sults in different interactions between the dimerization domainand transport domain. In the previous bcMalT structure, the twodomains have a buried surface area of 1,317 Å2, while in thecurrent structure of bcMalT-X the buried surface area is slightlyless at 1,217 Å2 (SI Appendix, Fig. S6 A and B). The buriedsurface is largely hydrophobic, with several hydrogen bonds. Ofparticular interest are residues S343 and T351 on the HP1 motif.The backbone carbonyl oxygen of A169 from the interface do-main forms a hydrogen bond with the side-chain hydroxyl ofS343 in the outward-facing conformation, and the same carbonyloxygen atom forms a hydrogen bond with the hydroxyl fromT351 in the inward-facing structure (SI Appendix, Fig. S6C). Wealso found similar exchange of hydrogen bond partners at theextracellular side, for example, K34-E397 in the outward-facingconformation and K34-N284 in the inward-facing conformation(SI Appendix, Fig. S6D). These interactions between the in-terface and substrate-binding domain likely stabilize the relativeposition of the two domains. Residues that form hydrogen bondsin the inward-facing state, S343 and T351, are both highly con-served in the glucose subfamily of EIICs such as MalT and PtsG(SI Appendix, Fig. S5). These observations also suggest that thecross-linking likely stabilizes an inward-facing state that is notsignificantly different from the native state.The rigid-body motion of the substrate-binding domain is ac-

companied by conformational changes at both the N- and C-ter-minal ends of the amphipathic helix AH2 (Fig. 4). In the bcMalToutward-facing structure, AH2 becomes an extension of TM5,extending 10 Å into the extracellular space. The angle betweenAH2 and TM5 is 137.5°. In the bcMalT-X inward-facing structure,AH2 and TM5 forms a bend, and the angle between these twohelices becomes 117.0°, and, as a result, AH2 resides horizontallyat the membrane surface (Fig. 4). The bend occurs at Pro199 atthe N-terminal end of AH2, which is highly conserved in theglucose subfamily of EIIs (SI Appendix, Fig. S5). This movementallows the substrate-binding domain to descend vertically towardthe intracellular side. Conformational changes at the C-terminalend of AH2 are equally important for accommodating movementsof the substrate-binding domain. In the bcMalT structure, sixresidues (215–220) at the C-terminal end of AH2 form a loop thatconnects TM6, and the two helices form an angle of 25°. In the

bcMalT-X structure, residues S215, K216, and D217 wind into ahelical turn, and the angle between AH2 and TM6 is 70° (Fig. 4).The length of AH2 changes from 17.1 Å in the outward-facingbcMalT to 20.3 Å in bcMalT-X. These changes seem to allow therotation of the substrate-binding domain.

Structure of bcMalT-X (T280C-E54CHg) Resembles the Native Inward-Facing State. Although evidence from the cross-linking experi-ment suggests that the two cysteine residues are in close prox-imity, and hence the cross-linked bcMalT is likely not distortedinto a nonnative state, we further examined the effect of cross-linking on bcMalT. We examined the stability of the bcMalT-Xstructure using molecular dynamics (MD) simulations. MDsimulations of bcMalT-X embedded in a fully hydrated lipidbilayer showed that the structure is stable for 1 μs with orwithout cross-linking the two cysteines. The inward-facing con-formation was well maintained in all of the simulations withbackbone rmsd of 2–3 Å against the bcMalT-X crystal structure(Fig. 5A) and backbone rmsd of 7–8 Å against the outward-facing bcMalT crystal structure (Fig. 5B). The backbone rmsd ofeach domain is also stable, fluctuating around 1–3 Å (SI Ap-pendix, Fig. S7 A–C). These results indicate that bcMalT-X isclose to an energy minimum and likely not significantly distortedfrom the native conformation. In addition, we compared thebcMalT-X structure with the previous CVSMD model of bcMalT.We performed 1-μs MD simulation using the final structurefrom our previous bcMalT CVSMD simulation (23). The 1-μsequilibrated bcMalT CVSMD model is well aligned to thebcMalT-X structure with a backbone rmsd of 2.41 Å (SI Appen-dix, Fig. S7D), reinforcing the predictive power of the previouscomputational approach.The proposed large motion of the substrate-binding domain

would cause large distance changes between certain residues.Single-molecule Föster resonance energy transfer (smFRET)provides us a tool to visualize individual protein conformationswithout the need for ensemble averaging, therefore providing amethod to directly observe minor protein populations that are indynamic equilibrium. We made a mutation N288C at the ex-tracellular side on the substrate-binding domain of bcMalT andlabeled it with Alexa Fluor 488 and Alexa Fluor 594 maleimide(Fig. 6A). Similarly, M340 at the intracellular side of the sub-strate-binding domain was mutated and labeled with the twodyes as well. Since bcMalT is a homodimer, only the ones labeledwith both dyes will register a FRET signal. The labeled proteins,when reconstituted into liposomes, transport maltose with simi-lar efficiency as wild-type bcMalT (SI Appendix, Fig. S3D), sug-gesting that the mutation and the subsequent labeling do notsignificantly affect the motions of the protein. smFRET histo-grams of each mutant show two populations. For N288C, weobserved one population of low FRET efficiency (EFRET) at

Fig. 4. Conformational changes of AH2. One protomer of bcMalT (A) andbcMalT-X (B) is shown as cartoon representation. The dimerization domain iscolored dark blue and the substrate-binding domain light blue. TM5 is col-ored green; AH2 is colored orange and TM6 is colored red. The hinge pointbetween TM5 and AH2, P199, is marked as a magenta sphere.

Fig. 5. Backbone rmsd of bcMalT-X structure during all-atom MD simula-tions. (A) Average backbone rmsd between the simulated structure andbcMalT-X structure. (B) Average backbone rmsd between the simulatedstructure and bcMalT structure. Each line is the average of three in-dependent simulations. Red, bcMalT-X; blue, bcMalT-X with the cross-linkingbetween T280 and E54 removed; black, bcMalT CVSMD (23).

Ren et al. PNAS | June 5, 2018 | vol. 115 | no. 23 | 5965

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Apr

il 7,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplemental

0.22 and one of high EFRET at 0.50 (Fig. 6B). The estimateddistances for these two populations are 74.1 and 60.0 Å, re-spectively (SI Appendix, Table S2). For M340, EFRET segregatesinto two well-resolved populations centered around 0.37 and0.84 (Fig. 6C), which correspond to an estimated distance of65.4 and 45.3 Å, respectively (SI Appendix, Table S2). As anegative control, we mutated and labeled a residue (D123) onthe dimerization domain, which is predicted to remain anchoredin the membrane during substrate translocation. Its EFRET showsa uniform population at 0.59 (SI Appendix, Fig. S8), which cor-responds to a distance of 56.5 Å. These distances are consistentwith the two structures within the error of smFRET measure-ments (SI Appendix, Fig. S8 and Table S2).smFRET also provides an estimate to the dynamics of bcMalT

(27). In both N288C and M340C, bcMalT seems to prefer theoutward- to the inward-facing conformation by a ratio of 5:1 and7:3, respectively. This fractional population ratio corresponds to afree energy difference of approximately 2–4 kJ/mol. Similar energydifference was observed in conformational changes of Gltphmeasured by the atomic force microscopy (28). A more stableoutward-facing conformation seems consistent with a slightly moreextensive buried interface between the dimerization and substrate-binding domains and may explain why it was difficult to crystallizethe wild-type bcMalT in an inward-facing conformation.

Summary and DiscussionThe structures of bcMalT and bcMalT-X provide direct visuali-zation of the outward- and inward-facing conformations of anEIIC from the glucose superfamily (Fig. 2A and Movie S1). Thestructures show that the substrate-binding domain moves as a rigidbody that keeps the substrate-binding site intact while moving it toface the alternate side of the membrane. An amphipathic helix,AH2, which connects the dimerization and the substrate-bindingdomains, bends at a conserved hinge point and changes its lengthto accommodate the movement of the substrate-binding domain.The bcMalT-X structure shows that the substrate-binding

domain has a 9-Å vertical translation and a 45° rotation. Suchlarge-scale motions are common among transporters using theelevator-type mechanism (14–22). In at least three of the mul-timeric transporters that have an elevator-type mechanism, thesubstrate-binding domains can move independently of eachother. Both smFRET and atomic force microscopy studies havevisualized the unsynchronized subunit motion in a single tri-meric Gltph (19, 28). In crystal structures of CNTNW, onetransport domain of the trimeric transporter was found in dif-ferent states than the other two (20). In addition, the structureof sodium citrate symporter SeCitS captured an asymmetricstate of the dimer, with one protomer in the outward-facingstate and the other in the inward-facing state (22). It is not clearwhether in bcMalT the two substrate-binding domains can moveindependently. If there exists a conformation in bcMalT where

one protomer is outward-facing and the other inward-facing,a population of EFRET at 0.34 for N288 and 0.63 for M340would be expected on the smFRET histogram. However, wewere not able to resolve either of these states under the currentexperimental conditions.The substrate in bcMalT-X is now accessible to solvent from

the cytoplasm. Among six independent MD simulations, we haveobserved two sugar release events: one in the bcMalT CVSMDsystem and the other in the bcMalT-X without the disulfide bondsystem (SI Appendix, Fig. S9A). The sugar molecules were re-leased without any significant conformational change duringthe simulations, suggesting that the bcMalT-X structure hasan inward-facing open conformation. The comparison of thebcMalT-X with bcChbC (inward-occluded) crystal structures alsosupports this observation, as the intracellular exit of the sugar-binding site is blocked by the TM4-5 loop in bcChbC, while thesame loop has a different orientation in bcMalT-X and no longercovers the sugar-binding site (SI Appendix, Fig. S9 B and C).Thus, bcMalT-X structure likely represents a conformation thatis compatible for phosphate transfer from the EIIB. These ob-servations also provide an explanation for facilitated sugartransport activity of bcMalT when no phosphorylation occurs.In summary, the combined approach of structural biology, MD

simulations, and smFRET revealed conformational changes re-quired for substrate translocation in an elevator-type mechanism oftransport. The inward-facing open conformation also provides astarting point for understanding how EIIB could interact with EIICand how a phosphate group is transferred to the bound maltose.

MethodsDetailed materials and methods are found in SI Appendix. Protein purifi-cation and cross-linking experiments followed established protocols repor-ted previously (8, 23). bcMalT-X was crystallized using the sitting-drop vapordiffusion method. The dataset was collected close to the mercury edge, andthe phase problem was solved by molecular replacement. Maltose bindingand transport were measured using the scintillation proximity assay and li-posome uptake assay, respectively, as described previously (8, 23). MD sim-ulations were conducted in CHARMM-GUI (29, 30), following optimizedprotocols (31, 32). For smFRET experiments, labeling conditions were opti-mized to maximize the amount of protein that contained both dyes. Dif-fusion-based smFRET was conducted using a custom-built ISS Alba confocallaser microscopy system, as described previously (33).

ACKNOWLEDGMENTS. We thank K. Rajashankar for advice on X-ray crystal-lography and K. Rajashankar, L. Keefe, E. Zoellner, K. Battaile, and A. Mulichakfor beamline support. We thank R. Bruni and B. Kloss, both from the Centeron Membrane Protein Production and Analysis in New York City, for thecloning and initial screen of EIIC genes. This work was supported by NationalKey Basic Research Program of China 2014CB910301 (to M.Z.); National In-stitutes of Health Grants U54GM087519 (to W.I. and M.Z.) and GM098878,DK088057, and HL086392 (toM.Z.); and Grant 12EIA8850017 from the AmericanHeart Association and Cancer Prevention and Research Institute of TexasGrant R12MZ (both to M.Z.).

Fig. 6. smFRET on bcMalT. (A) bcMalT (Left) and bcMalT-X (Right) are shown as cartoon and colored according to Fig. 2A. Cβ of N288 and M340 are shown asspheres and colored red and purple respectively. The distances between the two symmetry-related residues are marked on the dotted lines. Histograms ofsmFRET efficiency are shown for N288C (B) and M340C (C). Solid lines are fitting of the histograms with two Gaussian functions, and the mean value of eachGaussian function (appEFRET) is tabulated in SI Appendix, Table S2.

5966 | www.pnas.org/cgi/doi/10.1073/pnas.1800647115 Ren et al.

Dow

nloa

ded

by g

uest

on

Apr

il 7,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://movie-usa.glencoesoftware.com/video/10.1073/pnas.1800647115/video-1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1800647115/-/DCSupplementalwww.pnas.org/cgi/doi/10.1073/pnas.1800647115

1. Kaback HR (1968) The role of the phosphoenolpyruvate-phosphotransferase system inthe transport of sugars by isolated membrane preparations of Escherichia coli. J BiolChem 243:3711–3724.

2. Deutscher J, et al. (2014) The bacterial phosphoenolpyruvate:carbohydrate phos-photransferase system: Regulation by protein phosphorylation and phosphorylation-dependent protein-protein interactions. Microbiol Mol Biol Rev 78:231–256.

3. Kundig W, Ghosh S, Roseman S (1964) Phosphate bound to histidine in a protein as anintermediate in a novel phospho-transferase system. Proc Natl Acad Sci USA 52:1067–1074.

4. Kundig W, Kundig FD, Anderson B, Roseman S (1966) Restoration of active transportof glycosides in Escherichia coli by a component of a phosphotransferase system. J BiolChem 241:3243–3246.

5. McCoy JG, Levin EJ, Zhou M (2015) Structural insight into the PTS sugar transporterEIIC. Biochim Biophys Acta 1850:577–585.

6. Saier MH, Hvorup RN, Barabote RD (2005) Evolution of the bacterial phospho-transferase system: From carriers and enzymes to group translocators. Biochem SocTrans 33:220–224.

7. Cao Y, et al. (2011) Crystal structure of a phosphorylation-coupled saccharide trans-porter. Nature 473:50–54.

8. McCoy JG, et al. (2016) The structure of a sugar transporter of the glucose EIIC su-perfamily provides insight into the elevator mechanism of membrane transport.Structure 24:956–964.

9. Luo P, et al. (2015) Crystal structure of a phosphorylation-coupled vitamin C trans-porter. Nat Struct Mol Biol 22:238–241.

10. Jeckelmann J-M, et al. (2011) Structure and function of the glucose PTS transporterfrom Escherichia coli. J Struct Biol 176:395–403.

11. Koning RI, et al. (1999) The 5 Å projection structure of the transmembrane domain ofthe mannitol transporter enzyme II. J Mol Biol 287:845–851.

12. Kalbermatter D, et al. (2017) Electron crystallography reveals that substrate releasefrom the PTS IIC glucose transporter is coupled to a subtle conformational change.J Struct Biol 199:39–45.

13. Mitchell P (1957) A general theory of membrane transport from studies of bacteria.Nature 180:134–136.

14. Reyes N, Ginter C, Boudker O (2009) Transport mechanism of a bacterial homologueof glutamate transporters. Nature 462:880–885.

15. Canul-Tec JC, et al. (2017) Structure and allosteric inhibition of excitatory amino acidtransporter 1. Nature 544:446–451.

16. Yernool D, Boudker O, Jin Y, Gouaux E (2004) Structure of a glutamate transporterhomologue from Pyrococcus horikoshii. Nature 431:811–818.

17. Zhou X, et al. (2014) Structural basis of the alternating-access mechanism in a bile acidtransporter. Nature 505:569–573.

18. Lee C, et al. (2013) A two-domain elevator mechanism for sodium/proton antiport.Nature 501:573–577.

19. Erkens GB, Hänelt I, Goudsmits JMH, Slotboom DJ, van Oijen AM (2013) Un-synchronised subunit motion in single trimeric sodium-coupled aspartate trans-porters. Nature 502:119–123.

20. Hirschi M, Johnson ZL, Lee SY (2017) Visualizing multistep elevator-like transitions ofa nucleoside transporter. Nature 545:66–70.

21. Mancusso R, Gregorio GG, Liu Q, Wang D-N (2012) Structure and mechanism of abacterial sodium-dependent dicarboxylate transporter. Nature 491:622–626.

22. Wöhlert D, Grötzinger MJ, Kühlbrandt W, Yildiz Ö (2015) Mechanism of Na(+)-de-pendent citrate transport from the structure of an asymmetrical CitS dimer. eLife 4:e09375.

23. Lee J, Ren Z, Zhou M, Im W (2017) Molecular simulation and biochemical studiessupport an elevator-type transport mechanism in EIIC. Biophys J 112:2249–2252.

24. Verdon G, Oh S, Serio RN, Boudker O (2014) Coupled ion binding and structuraltransitions along the transport cycle of glutamate transporters. eLife 3:e02283.

25. Saraceni-Richards CA, Jacobson GR (1997) A conserved glutamate residue, Glu-257, isimportant for substrate binding and transport by the Escherichia coli mannitol per-mease. J Bacteriol 179:1135–1142.

26. Ruijter GJ, van Meurs G, Verwey MA, Postma PW, van Dam K (1992) Analysis ofmutations that uncouple transport from phosphorylation in enzyme IIGlc of theEscherichia coli phosphoenolpyruvate-dependent phosphotransferase system. J Bacteriol174:2843–2850.

27. Ferreon ACM, Ferreon JC, Wright PE, Deniz AA (2013) Modulation of allostery byprotein intrinsic disorder. Nature 498:390–394.

28. Ruan Y, et al. (2017) Direct visualization of glutamate transporter elevator mecha-nism by high-speed AFM. Proc Natl Acad Sci USA 114:1584–1588.

29. Jo S, Kim T, Iyer VG, Im W (2008) CHARMM-GUI: A web-based graphical user interfacefor CHARMM. J Comput Chem 29:1859–1865.

30. Lee J, et al. (2016) CHARMM-GUI input generator for NAMD, GROMACS, AMBER,OpenMM, and CHARMM/OpenMM simulations using the CHARMM36 additive forcefield. J Chem Theory Comput 12:405–413.

31. Wu EL, et al. (2014) CHARMM-GUI Membrane Builder toward realistic biologicalmembrane simulations. J Comput Chem 35:1997–2004.

32. Jo S, Lim JB, Klauda JB, Im W (2009) CHARMM-GUI Membrane Builder for mixed bi-layers and its application to yeast membranes. Biophys J 97:50–58.

33. Tsoi PS, et al. (2017) The N-terminal domain of ALS-linked TDP-43 assembles withoutmisfolding. Angew Chem Int Ed Engl 56:12590–12593.

34. Baker NA, Sept D, Joseph S, Holst MJ, McCammon JA (2001) Electrostatics of nano-systems: Application to microtubules and the ribosome. Proc Natl Acad Sci USA 98:10037–10041.

Ren et al. PNAS | June 5, 2018 | vol. 115 | no. 23 | 5967

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Apr

il 7,

202

1