Structural basis of rifampin inactivation byrifampin phosphotransferaseXiaofeng Qia,b, Wei Lina,1, Miaolian Maa, Chengyuan Wanga,b, Yang Hea,b, Nisha Heb,c, Jing Gaod, Hu Zhoud, Youli Xiaoc,Yong Wangc, and Peng Zhanga,2

aNational Key Laboratory of Plant Molecular Genetics, Chinese Academy of Sciences Center for Excellence in Molecular Plant Sciences, Institute of PlantPhysiology and Ecology, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, Shanghai 200032, China; bUniversity of Chinese Academyof Sciences, Beijing 100039, China; cChinese Academy of Sciences Key Laboratory of Synthetic Biology, Chinese Academy of Sciences Center for Excellence inMolecular Plant Sciences, Institute of Plant Physiology and Ecology, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences, Shanghai200032, China; and dChinese Academy of Sciences Key Laboratory of Receptor Research, Shanghai Institute of Materia Medica, Chinese Academy ofSciences, Shanghai 201203, China

Edited by Alexander Serganov, New York University, New York, NY, and accepted by the Editorial Board March 1, 2016 (received for review November30, 2015)

Rifampin (RIF) is a first-line drug used for the treatment oftuberculosis and other bacterial infections. Various RIF resistancemechanisms have been reported, and recently an RIF-inactivationenzyme, RIF phosphotransferase (RPH), was reported to phosphor-ylate RIF at its C21 hydroxyl at the cost of ATP. However, theunderlying molecular mechanism remained unknown. Here, wesolve the structures of RPH from Listeria monocytogenes (LmRPH)in different conformations. LmRPH comprises three domains: anATP-binding domain (AD), an RIF-binding domain (RD), and a cat-alytic His-containing domain (HD). Structural analyses reveal thatthe C-terminal HD can swing between the AD and RD, like a toggleswitch, to transfer phosphate. In addition to its catalytic role, theHD can bind to the AD and induce conformational changes thatstabilize ATP binding, and the binding of the HD to the RD is re-quired for the formation of the RIF-binding pocket. A line of hy-drophobic residues forms the RIF-binding pocket and interactswith the 1-amino, 2-naphthol, 4-sulfonic acid and naphthol moie-ties of RIF. The R group of RIF points toward the outside of thepocket, explaining the low substrate selectivity of RPH. Four resi-dues near the C21 hydroxyl of RIF, His825, Arg666, Lys670, andGln337, were found to play essential roles in the phosphorylationof RIF; among these the His825 residue may function as the phos-phate acceptor and donor. Our study reveals the molecular mech-anism of RIF phosphorylation catalyzed by RPH and will guide thedevelopment of a new generation of rifamycins.

antibiotic resistance | rifampin | phosphotransferase |molecular mechanism | toggle switch

Rifamycins are a group of natural or semisynthetic antibioticsused for treating a broad repertoire of bacterial infections.These compounds bind directly to the β-subunit of bacterialRNA polymerase (RNAP) at a highly conserved region, blockingthe exit tunnel for RNA elongation and thus inhibiting the processof transcription (1). The first member of the rifamycins to be de-scribed, rifamycin B, was extracted from the soil actinomyceteAmycolatopsis mediterranei (2). The natural product had modestantibiotic activity, but semisynthetic derivatives of the rifamycinfamily have proven highly successful in the clinic (3). The best-knownmember of the rifamycin family, rifampin (RIF), was introduced tothe clinic in 1968; it is highly effective against Mycobacterium tuber-culosis and greatly shortens the duration of tuberculosis therapy (4).At present, RIF continues to be a first-line drug for the treatment oftuberculosis (5). Through the years additional derivatives havebeen developed to treat a wider range of bacterial infections (3);for example, rifalazil serves as an effective antibiotic againstChlamydia-based persistent infections (6), and rifaximin is usedto treat travelers’ diarrhea and irritable bowel syndrome (7, 8).Extensive use of rifamycins has led to the development of

bacterial resistances (9). In M. tuberculosis and other mycobacteriathe most common resistance mechanisms are point mutations of

the target, the RNAP β-subunit; these mutations significantly de-crease the binding of rifamycins and thus neutralize the antibioticactivity (10). Another prevalent resistance strategy adopted bybacteria is modification of the rifamycins, such as ADP ribosy-lation, glycosylation, and phosphorylation (11–13). These covalentmodifications occur on the critical hydroxyls of the 1-amino,2-naphthol, 4-sulfonic acid (ansa) chain of rifamycins andthus make rifamycins unable to fit into the binding pocket onRNAP. Additional resistance mechanisms have been reportedalso (14–16).Antibiotic resistance is a great threat to the treatment of in-

fectious disease, and understanding the molecular mechanismsof resistance no doubt will help guide the development of a newgeneration of drugs (17, 18). A number of studies have beencarried out to understand rifamycin resistance caused by RNAPmutations (1, 19). However, the proteins and mechanisms in-volved in the covalent modifications of rifamycins remain largelyunknown. Recently, an antibiotic-resistance protein family, RIFphosphotransferase (RPH), was found to inactivate RIF byphosphorylating it at the hydroxyl attached to the C21 of its ansachain. RPHs in heterologous bacteria are able to inactivate diverse

Significance

Rifampin phosphotransferases (RPH) belong to a recently iden-tified antibiotic-resistance protein family that inactivates rifam-pin, the first-line drug against tuberculosis, by phosphorylation.By determining the structures of RPH from Listeria mono-cytogenes (LmRPH) in different conformations, we reveal atoggle-switch mechanism of the LmRPH catalytic process inwhich the C-terminal His domain swings between the ATP-binding domain and the rifampin-binding domain to transferphosphate from ATP to rifampin. These structures explain thelow substrate selectivity of RPH for the rifamycins. The residuesinvolved in rifampin phosphorylation are identified also. Thestructural mechanism revealed in this study will guide the de-velopment of a new generation of rifamycins.

Author contributions: X.Q. and P.Z. designed research; X.Q., W.L., M.M., C.W., Y.H., N.H.,J.G., and H.Z. performed research; X.Q., W.L., M.M., C.W., Y.H., N.H., J.G., H.Z., Y.X., Y.W.,and P.Z. analyzed data; and P.Z. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission. A.S. is a guest editor invited by the EditorialBoard.

Data deposition: The structural factors and coordinates reported in this paper have beendeposited in the Protein Data Bank (PDB) [PDB ID codes 5HV1 (LmRPH–ANP–RIF), 5HV2(LmRPHG527Y–apo), 5HV3 (LmRPHG527Y–ANP), and 5HV6 (LmRPH–AD)].1Present address: Waksman Institute of Microbiology, Rutgers, The State University ofNew Jersey, Piscataway, NJ 08854.

2To whom correspondence should be addressed. Email: pengzhang01@sibs.ac.cn.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1523614113/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1523614113 PNAS | April 5, 2016 | vol. 113 | no. 14 | 3803–3808

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

June

1, 2

021

http://crossmark.crossref.org/dialog/?doi=10.1073/pnas.1523614113&domain=pdfhttp://www.rcsb.org/pdb/explore/explore.do?structureId=5HV1http://www.rcsb.org/pdb/explore/explore.do?structureId=5HV2http://www.rcsb.org/pdb/explore/explore.do?structureId=5HV3http://www.rcsb.org/pdb/explore/explore.do?structureId=5HV6mailto:pengzhang01@sibs.ac.cnhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1523614113/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1523614113/-/DCSupplementalwww.pnas.org/cgi/doi/10.1073/pnas.1523614113

clinically used rifamycins with great efficiency (13). Bioinformaticanalyses suggest that RPHs are widespread in both pathogenicand nonpathogenic bacteria. The RPH protein contains threedomains (listed from the N terminus to the C terminus): theATP-binding domain (AD), the RIF-binding domain (RD), andthe His domain (HD), which contains a conserved His residueessential for phosphate transfer. This architecture is similar tothat of phosphoenolpyruvate (PEP) synthase, which also containsthree domains, an ATP-binding domain, a catalytic His domain, anda pyruvate-binding domain and catalyzes the reversible conversionof ATP, water, and pyruvate to AMP, inorganic phosphate (Pi), andPEP (20). Apart from this information, little is known about RPHs.Here we report the crystal structures of RPH from Listeria

monocytogenes (LmRPH) in different catalytic conformations.Structural and functional analyses reveal the molecular basis ofsubstrate binding, phosphate transfer, and RIF phosphorylationby LmRPH. This study identifies the molecular mechanism ofRIF phosphorylation and will guide strategies to overcome RPH-mediated rifamycin resistance.

ResultsCharacterization of LmRPH. The gene encoding RPH fromLmRPH was cloned, expressed in Escherichia coli, and purified.The enzymatic activity of the recombinant LmRPH was tested ina reaction system containing the substrates RIF and ATP, andthe products were separated by HPLC. As the reaction proceeded,the amount of RIF gradually decreased, accompanied by the in-crease of a subsequent product peak (Fig. 1A), which was identi-fied as phosphorylated RIF (RIF-P) by LC-MS (Fig. 1B and Fig.S1). The other substrate, ATP, was converted into AMP ratherthan ADP until RIF was phosphorylated completely (Fig. 1C).To examine ability of LmRPH to inactivate RIF in vivo, E. coli

BL21 (DE3) cells transformed with pQE80L-LmRPH were cul-tured on solid LB medium containing a gradient of RIF concen-trations. The results show that E. coli growth is strongly inhibited by10 μg/mL RIF, but the introduction of LmRPH at concentrationsgreater than 1,000 μg/mL confers resistance to RIF (Fig. 1D). Thesedata suggest that LmRPH catalyzes the conversion of RIF to RIF-P

at the cost of ATP and confers the bacteria with high-level re-sistance to RIF.

Structures of LmRPH at Different Conformations. The LmRPHprotein was purified further using gel filtration before crystalli-zation, and the two major peaks (peaks 1 and 2) observed wereboth confirmed to be LmRPH proteins with similar molecularradius/mass by dynamic light scattering (DLS) (Fig. S2). Thisresult indicates that LmRPH might have different conformationsin solution. However, we could obtain diffractable crystals onlywith LmRPH protein from peak 1. The LmRPH structure wassolved in an AMP–PNP (ANP)-, Mg2+- and RIF-bound state(LmRPH–ANP–RIF) by the single-wavelength anomalous dispersion(SAD) method. The overall structure adopts a saddle-like shape, withthe AD (residues 1–315) and the RD (residues 323–748) forming twoflaps of the saddle. The C-terminal HD (residues 771–867) binds tothe RD from the concave side of the “saddle” (Fig. 2A). The ATPanalog, ANP, binds in a cleft of the AD from the concave side of thesaddle, and RIF binds in a pocket of the RD from the convex side.The two substrate-binding sites are about 49 Å apart, leaving ampleroom for the HD to play an indispensable role in catalysis. In-triguingly, the HD is linked to the RD by a long, flexible linker(residues 749–770) through which the HD might swing betweenthe AD and RD to transfer phosphate from ATP to RIF.In the LmRPH–ANP–RIF structure, the HD contacts the RD

mainly through hydrophobic interactions (Fig. S3). We in-troduced mutations at this interface to disrupt these interactionsand found that LmRPH proteins containing these mutations hadgel-filtration profiles different from those of wild-type proteins(Fig. 2B), i.e., two major peaks for wild-type proteins vs. onemajor peak for mutants. Accordingly, LmRPH-G527A, LmRPH-G527S, and LmRPH-G527Y mutants have much decreased orno RIF-phosphorylation activity in vitro and reduced or no RIFresistance in vivo (Fig. 2 C and D). These data suggest that, in-stead of two conformations, the LmRPH-G527A, LmRPH-G527S, and LmRPH-G527Y mutants tend to adopt one con-formation in solution. Indeed, the structures of LmRPHG527Y inboth the apo form (LmRPHG527Y–apo) and the ANP-boundform (LmRPHG527Y–ANP) were in a conformation differentfrom that of the LmRPH–ANP–RIF structure (Fig. 2 E and F),with the HD binding to the AD from the concave side. Notably,even though the interactions between the HD and the RD andbetween the HD and the AD have been observed in differentLmRPH conformations, we could not quantify the interactionaffinities between the individually purified HD and RD or AD inisothermal titration calorimetry (ITC) experiments; the interac-tions are too weak to be detected by ITC (Fig. S4), suggestingthat the interactions between the HD and the RD and betweenthe HD and the AD are dynamic. Our structural data demon-strate that the HD can swing between the RD and the AD, as isrequired for LmRPH catalysis, more specifically, for the transferof phosphate from ATP to RIF.

ATP Binding with the AD Is Stabilized by the HD. Although theLmRPHG527Y–ANP and LmRPH–ANP–RIF structures adoptdifferent conformations, both can bind with ANP, prompting usto determine their ATP-binding affinities. The results show thatthe ATP-binding affinity of LmRPHG527Y (Kd = 0.43 μM) ishigher than that of the wild-type protein (Kd = 1.11 μM), but theseparated AD itself cannot bind with ATP (the affinity is too lowto be detected by ITC) (Fig. 3A), suggesting that the HD maycontribute to the ATP binding. To resolve this mystery, we solvedthe structure of the AD in apo state (LmRPH–AD) and comparedit with the LmRPHG527Y–apo structure (Fig. 3B). The AD containstwo subdomains, subdomain I (residues 1–183) and subdomain II(residues 190–315), which are connected by a flexible linker (L13,residues 184–189) to form a hinge-like conformation. The bindingof the HD with the AD induces significant conformational changes

Fig. 1. Activity of LmRPH. (A) In vitro LmRPH reaction products analyzed byHPLC with the RIF detection program. (B) Identification of the two peaks in Aby LC-MS. (Left) Peak 1. (Right) Peak 2. (C) The samples in A were analyzedby HPLC with a nucleotide-detection program. (D) E. coli growth assay.Bacteria transformed with LmRPH or vector were cultured in solid LB me-dium complemented with 0, 10, 100, or 1,000 μg/mL RIF.

3804 | www.pnas.org/cgi/doi/10.1073/pnas.1523614113 Qi et al.

Dow

nloa

ded

by g

uest

on

June

1, 2

021

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1523614113/-/DCSupplemental/pnas.201523614SI.pdf?targetid=nameddest=SF1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1523614113/-/DCSupplemental/pnas.201523614SI.pdf?targetid=nameddest=SF1http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1523614113/-/DCSupplemental/pnas.201523614SI.pdf?targetid=nameddest=SF2http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1523614113/-/DCSupplemental/pnas.201523614SI.pdf?targetid=nameddest=SF3http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1523614113/-/DCSupplemental/pnas.201523614SI.pdf?targetid=nameddest=SF4www.pnas.org/cgi/doi/10.1073/pnas.1523614113

in both subdomain I and II: helices α4, α5, and α8 of subdomain Iundergo a dramatic shift toward the HD, leading to hydrophobicinteractions between α8 (subdomain I) and α31 (HD), and helixα9 unwinds to bind with the HD, also through hydrophobic inter-actions (Fig. 3B). As a result, the conformations of subdomains Iand II of the AD are stabilized by the binding of the HD, as is theATP-binding cleft between these two subdomains. These findingsexplain why the HD is required for tight binding of ATP. Inthe LmRPHG527Y mutant, the HD is restricted from binding withthe RD; therefore the binding affinity of LmRPHG527Y is higherthan that of the wild-type protein (Fig. 3A).The binding of ANP to the cleft of the AD results in further

conformational changes in the surrounding structural elements,as can be seen clearly by comparing the LmRPHG527Y–apo andLmRPHG527Y–ANP structures (Fig. 3C). After ANP binding,β3–β4, which adopts a loop conformation in the LmRPHG527Y–apo structure, forms a five-stranded antiparallel β-sheet withβ1–β2–β5, as do β10–β11 in subdomain II. In addition, loop L9

(residues 123–134) from subdomain I, which is disordered in theabsence of ANP, can be seen clearly after ANP binding. ANPbinding also induces a conversion of the α9 from subdomain II.The formation of L9 and α9 after ANP binding generates stericrepulsions of the HD, thereby weakening the interaction be-tween the AD and at HD, as reflected by the poor electrondensity of the HD in the LmRPHG527Y–ANP structure (Fig. S5).The structural rearrangements of the AD described above

accommodate the tight binding of ANP to the cleft through anumber of conserved residues in addition to an Mg2+ (Fig. 3D).Specifically, the adenine ring of ANP forms three hydrogenbonds with the guanidine group of Arg117, the carbonyl oxygenof Gln184, and the side chain of Gln183; the 2′-hydroxyl group ofthe ANP ribose forms a hydrogen bond with the side chain ofGlu297; the α, γ-phosphates of ANP form hydrogen-bondinginteractions with residues Arg117, Thr136, Lys22, Arg311, andGly132; and the β, γ-phosphates of ANP are coordinated withresidues Glu297 and Gln309 through the Mg2+. The importance

Fig. 2. Overall structures of LmRPH at different states. (A) Structure of wild-type LmRPH in complex with RIF, ANP, and Mg2+. The AD, RD, and HD are coloredlemon, light blue, and orange, respectively. ANP and RIF are shown as sticks and are colored green and magenta, respectively. Mg2+ is shown as a sphere.(B) Gel-filtration profiles of wild-type LmRPH (red) and the G527A (green), G527S (cyan), and G527Y (blue) mutants. (C) In vitro catalytic activity of Gly527mutants detected by HPLC. The reaction time of G527A, G527S, and G527Y is 1 h, and that of wild-type LmRPH is 5 min. Proteins were used at 0.5 mg/mL.(D) E. coli growth assay for Gly527 mutants. Bacteria transformed with wild-type LmRPH, vector, G527A, G527S, or G527Y were cultured in solid LB mediumcomplemented with 0, 10, 40, or 80 μg/mL RIF. (E) Structure of LmRPHG527Y in apo form. (F) Structure of LmRPHG527Y in complex with ANP and Mg2+. Colorcodes in E and F are as A.

Qi et al. PNAS | April 5, 2016 | vol. 113 | no. 14 | 3805

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

June

1, 2

021

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1523614113/-/DCSupplemental/pnas.201523614SI.pdf?targetid=nameddest=SF5

of these residues was validated by an in vitro enzymatic activityassay. The results show that the activity of Q183A is slightlylower than that of the wild-type LmRPH, the activities of T136A,Q309A, and R311A mutants are significantly reduced, and thoseof other mutants (K22A, R117A, and E297A) are extremely low(Fig. 3E). Accordingly, the RIF-resistance levels of these mu-tants are reduced to different extents, except for the Q183Amutant, in which resistance is comparable to that in wild-typeLmRPH (Fig. 3F).

Both the RD and HD Are Involved in Rif Binding. The structure of theRD can be divided further into three subdomains: subdomain I(α12–16, 28–30, and β14–18), II (α17–20 and 26–27), and III(α21–25) (Fig. 4A). Searches of the Protein Data Bank failed toidentify any entry that is structurally homologous to the RD,suggesting that the RD represents a previously unidentifiedstructural fold related to RIF binding. In the LmRPH–ANP–RIFstructure, the HD binds with all three subdomains of the RDfrom the concave side and forms the RIF-binding pocketwith subdomains I and II of RD. Distinct from the AD ATP-binding cleft, which faces the concave side of LmRPH, theopening of the RIF-binding pocket faces the convex side (Figs.2A and 4A). Structural comparison of LmRPH–ANP–RIF withLmRPHG527Y-ANP reveals significant conformational changesat the RIF-binding pocket (Fig. 4B). Specifically, the binding ofthe HD with the RD pushes away α14–α16 and connecting loopsof the RD, leading to a rearrangement of the surrounding in-ward-facing residues that create an RIF-binding pocket (Fig. 4B–D). These structural observations suggest that both the RDand HD are involved in RIF binding. Consistently, we found thatthe RD alone is not sufficient to bind with RIF, but the RD andHD together can bind RIF with high affinity (Kd = 79.4 μM)(Fig. 4E). (The binding affinity of RIF with full-length LmRPH

changes over time; therefore we used the RD and HD for thedetection of RIF binding.).The RIF-binding pocket is comprised mainly of hydrophobic

residues. Residues Val333, Met359, and Val368 constitute ahydrophobic patch and contact the naphthol ring of RIF throughvan der Waals forces; residues Ile331, Ile370, Ile394, Met383,Leu387, Met823, Met491, Met488, Leu478, and Met673 stabilizethe ansa chain of RIF through hydrophobic interactions (Fig. 4 Fand G). Mutations V333A, V368A, M383A, or M673A increasethe size of the pocket and reduce the phosphorylation activity ofLmRPH, whereas V333W or V368W causes steric conflict andalmost abolishes the activity (Fig. 4H). The R group of RIFpoints toward the opening of the pocket and packs against res-idues Pro356 and Phe479; replacement of either of these tworesidues with alanine has only minor effects on the phosphory-lation activity and RIF binding (Fig. 4F and Table S1). Thisfinding likely explains why RPH can phosphorylate variousmembers of the rifamycin family that differ primarily at the Rgroup (13).

Phosphorylation of RIF. The LmRPH–ANP–RIF complex struc-ture allows us to examine the phosphorylation site of RIF. Thepreviously identified phosphorylation site of RIF, C21 hydroxyl,is about 6.7 Å away from residue His825 of the HD (Fig. 5A). Awater molecule between His825 and C21 hydroxyl forms a hy-drogen bond with the C21 hydroxyl. When we modeled theproduct RIF-P into the structure (Fig. 5B), we found that thephosphate group of RIF-P could form four hydrogen bonds withLys670, Arg666, and Gln337 and that the distance between RIF-Pand residue His825 is about 4 Å. These structural observationssuggest that the HD residue His825 and residues Lys670, Arg666,and Gln337 are involved in RIF phosphorylation. To verify thispossibility, we mutated these four residues to alanine and

Fig. 3. ATP-binding site. (A) ATP-binding affinity ofthe AD (green isotherm), wild-type LmRPH (blueisotherm), and the G527Y mutant (red isotherm)measured by ITC. (B) Conformational changes of theAD induced by HD binding. The LmRPH–AD struc-ture (gray) is superposed with the AD of theLmRPHG527Y–apo structure (lemon). The L13 loopconnecting subdomains I and II is highlighted in red.The interaction interfaces between the AD and HD(orange) are shown in zoom-in views, and residuesconstituting the interface are shown with side chains.(C) Conformational changes induced in the AD andHD by ANP binding. The AD (lemon) and HD (orange)of the LmRPHG527Y–apo structure are superposedwith those of the LmRPHG527Y-ANP structure (lightblue). Structural elements undergoing conforma-tional changes after ANP binding are colored in red.(D) Residues constituting the ATP-binding site. ANP(green) and residues (lemon) are shown as sticks,and Mg2+ is shown as a lemon sphere. Coordinationand hydrogen bonds are shown as dashed lines.(E) In vitro catalytic activity of ATP-binding sitemutants detected by HPLC. The amounts of en-zymes used in the assays of the K22A, R117A, E297A,T136A, Q309A, and R311A mutants are 10× those usedin assays of wild-type LmRPH. (F) E. coli growth assayfor ATP-binding site mutants. Bacteria transformedwith wild-type LmRPH, vector, or mutants were culturedin solid LB medium complemented with 0, 10, 40, 80,or 320 μg/mL RIF.

3806 | www.pnas.org/cgi/doi/10.1073/pnas.1523614113 Qi et al.

Dow

nloa

ded

by g

uest

on

June

1, 2

021

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1523614113/-/DCSupplemental/pnas.201523614SI.pdf?targetid=nameddest=ST1www.pnas.org/cgi/doi/10.1073/pnas.1523614113

determined their activities. The results show that the H825A,R666A, and K670A mutants lose the ability to phosphorylateRIF both in vitro and in vivo, and Q337A has significantly re-duced activity (Fig. 5 C and D). Notably, residue His825 is highlyconserved among RPHs, reminiscent of the catalytic His residuein PEP synthase (20). Using the Phos-tag SDS/PAGE experiment,we found that LmRPH is phosphorylated in the presence of ATPbut the H825A mutant is not (Fig. 5E). The phosphorylation ofresidue His825 is confirmed in an MS analysis (Fig. S6). Theseresults suggest that this conserved His residue also may function asa phosphate acceptor and donor in the phosphorylation of RIF. In

addition, positively charged residues often function as catalyticbases to abstract a proton at the reaction centers of phosphate-transfer enzymes and other enzymes, including MAPK, phos-phothreonine lyase, IMP dehydrogenase, pectate/pectin lyases,fumarate reductase, and L-aspartate oxidase (21–23), suggestingthat Lys670 and Arg666 might be candidate residues for the cat-alytic bases of LmRPH.

DiscussionIn this work we captured two major conformational states of theLmRPH catalytic process: a conformation in which the HD binds

Fig. 4. The RIF-binding pocket. (A) Structure of the RD (in LmRPH–ANP–RIF). The gray dashed lines separate three subdomains (I, II, and III) of the RD. RIF isshown as magenta sticks. (B) Conformational changes of the RD induced by HD binding. The RD (light blue) and HD (orange) of the LmRPH–ANP–RIF structurewas superposed with that of the LmRPHG527Y–apo structure (gray). α14–16, which undergo conformational changes after HD binding, are highlighted in red.(C) Surface view of the RIF-binding pocket in the LmRPHG527Y–apo structure. (D) Surface view of the RIF-binding pocket in the LmRPH–ANP–RIF structure. TheRD, HD, and α14–16 are colored light blue, orange, and red, respectively. (E) The RIF-binding affinity of the RD and RD–HDmeasured by ITC. Binding isothermsfor RD and RD–HD are colored green and red, respectively. (F) Residues at the RIF-binding site. RIF (magenta) and interacting residues (light blue) are shownas sticks. (G) Chemical structure of RIF. The naphthol ring and ansa chain of RIF are shown in pink and blue, respectively, and the R group is outlined by adashed box. (H) In vitro catalytic activity of RIF-binding mutants detected by HPLC.

Fig. 5. Catalytic center of RIF phosphorylation. (A) Catalytic site of RIF phosphorylation. Residues His825, Lys670, Arg666, and Gln337 and RIF are shown as sticks; thewater molecule is shown as a red sphere. Distances between the water molecule and surrounding residues are shown as dashed lines. (B) Catalytic site with a modeledRIF-P. Distances between the phosphate group and residues are shown as dashed lines. (C) In vitro catalytic activity of H825A, K670A, R666A, and Q337A detected byHPLC. (D) E. coli growth assay of the mutants in C. Bacteria transformed with wild-type LmRPH, vector, or mutants were cultured in solid LB medium complementedwith 0 or 10 μg/mL RIF. (E) Phosphorylation analysis of wild-type LmRPH and the H825A mutant using Phos-tag SDS PAGE. LmRPH-P, phosphorylated LmRPH.

Qi et al. PNAS | April 5, 2016 | vol. 113 | no. 14 | 3807

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

June

1, 2

021

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1523614113/-/DCSupplemental/pnas.201523614SI.pdf?targetid=nameddest=SF6

to the AD, and a conformation in which the HD binds to theRD. Structural-based analysis confirmed that the HD functionsas a toggle switch, swinging between the two distant domains.When binding to the AD, the HD facilitates ATP binding andhydrolysis, grabbing a phosphate by residue His825. Then theHD swings over to the RD, facilitating RIF binding and initiatingRIF phosphorylation. The dynamic nature of the LmRPH proteinenables the smooth transition between these two conformationalstates (Fig. 6). This mechanism resembles that of three-domainpyruvate orthophosphate dikinase (PPDK) enzymes in which theHis domain swivels between the nucleotide-binding domain and

the pyruvate-binding domain to transfer phosphate from ATP topyruvate (Fig. S7) (24–26).Based on the structure of RIF bound to the Thermus aquaticus

RNAP core enzyme, the C21 hydroxyl of RIF points toward theinside of the RIF-binding pocket and forms hydrogen-bonding in-teractions with nearby residues (1). Phosphorylation of this hydroxylmay lead to steric clash, thereby weakening or abolishing thebinding of RIF to RNAP and ultimately resulting in resistance toRIF. RPHs are widespread among Bacillales, Actinomycetales,and Clostridiales, which include many human pathogens. Searchesof the pathogenic bacterial genomes of Bacillus anthracis, En-terococcus faecalis, Nocardia brasiliensis, and Listeria mono-cytogenes all reveal RPH genes, which may limit the clinical use ofrifamycins against these organisms. Our mechanistic study ofLmRPH provides feasible strategies, such as developing high-affinity RPH inhibitors or new RIF derivatives that are notsusceptible to RPH, to overcome RPH-mediated resistance.

Materials and MethodsSee SI Materials and Methods for details. In general, LmRPH protein wasexpressed in E. coli and purified to homogeneity for crystallization. All datawere collected and processed with HKL3000 (27). The structures were de-termined using programs in Phenix (28), and structural models were builtwith Coot (29). The products of the enzymatic assay were detected withHPLC/MS. The substrate-binding affinity was measured with ITC. Data col-lection and refinement statistics are summarized in Table S2.

ACKNOWLEDGMENTS. We thank the staff members at the BL19U beamlineof the National Center for Protein Science Shanghai and the BL17U beamlineof the Shanghai Synchrotron Radiation Facility for technical assistance indata collection and the staff at the core facility center of the Institute ofPlant Physiology and Ecology for MS experiments and analysis. This workwas supported by National Natural Science Foundation of China Grant31322016 and National Program on Key Basic Research Projects Grant2015CB910900 and by funding from the National Key Laboratory of PlantMolecular Genetics, CAS Center for Excellence in Molecular Plant Sciences,Institute of Plant Physiology and Ecology, Shanghai Institutes for BiologicalSciences, CAS.

1. Campbell EA, et al. (2001) Structural mechanism for rifampicin inhibition of bacterialrna polymerase. Cell 104(6):901–912.

2. Sensi P, Margalith P, Timbal MT (1959) Rifomycin, a new antibiotic; preliminary re-port. Farmaco, Sci 14(2):146–147.

3. Aristoff PA, Garcia GA, Kirchhoff PD, Showalter HD (2010) Rifamycins–obstacles andopportunities. Tuberculosis (Edinb) 90(2):94–118.

4. Sensi P (1983) History of the development of rifampin. Rev Infect Dis 5(Suppl 3):S402–S406.

5. Getahun H, Matteelli A, Chaisson RE, Raviglione M (2015) Latent Mycobacteriumtuberculosis infection. N Engl J Med 372(22):2127–2135.

6. Rothstein DM, van Duzer J, Sternlicht A, Gilman SC (2007) Rifalazil and other ben-zoxazinorifamycins in the treatment of chlamydia-based persistent infections. ArchPharm (Weinheim) 340(10):517–529.

7. Huang DB, DuPont HL (2005) Rifaximin–a novel antimicrobial for enteric infections.J Infect 50(2):97–106.

8. Schoenfeld P, et al. (2014) Safety and tolerability of rifaximin for the treatment ofirritable bowel syndrome without constipation: A pooled analysis of randomised,double-blind, placebo-controlled trials. Aliment Pharmacol Ther 39(10):1161–1168.

9. Dorman SE, Chaisson RE (2007) From magic bullets back to the magic mountain: Therise of extensively drug-resistant tuberculosis. Nat Med 13(3):295–298.

10. Goldstein BP (2014) Resistance to rifampicin: A review. J Antibiot (Tokyo) 67(9):625–630.

11. Baysarowich J, et al. (2008) Rifamycin antibiotic resistance by ADP-ribosylation:Structure and diversity of Arr. Proc Natl Acad Sci USA 105(12):4886–4891.

12. Spanogiannopoulos P, Thaker M, Koteva K, Waglechner N, Wright GD (2012) Char-acterization of a rifampin-inactivating glycosyltransferase from a screen of environ-mental actinomycetes. Antimicrob Agents Chemother 56(10):5061–5069.

13. Spanogiannopoulos P, Waglechner N, Koteva K, Wright GD (2014) A rifamycin in-activating phosphotransferase family shared by environmental and pathogenic bac-teria. Proc Natl Acad Sci USA 111(19):7102–7107.

14. Tupin A, et al. (2010) Resistance to rifampicin: At the crossroads between ecological,genomic and medical concerns. Int J Antimicrob Agents 35(6):519–523.

15. Wright GD (2005) Bacterial resistance to antibiotics: Enzymatic degradation andmodification. Adv Drug Deliv Rev 57(10):1451–1470.

16. Hoshino Y, et al. (2010) Monooxygenation of rifampicin catalyzed by the rox geneproduct of Nocardia farcinica: Structure elucidation, gene identification and role indrug resistance. J Antibiot (Tokyo) 63(1):23–28.

17. Fischbach MA, Walsh CT (2009) Antibiotics for emerging pathogens. Science325(5944):1089–1093.

18. Wright GD (2007) The antibiotic resistome: The nexus of chemical and genetic di-versity. Nat Rev Microbiol 5(3):175–186.

19. Ramaswamy S, Musser JM (1998) Molecular genetic basis of antimicrobial agentresistance in Mycobacterium tuberculosis: 1998 update. Tuber Lung Dis 79(1):3–29.

20. Narindrasorasak S, Bridger WA (1977) Phosphoenolypyruvate synthetase of Escher-ichia coli: Molecular weight, subunit composition, and identification of phosphohis-tidine in phosphoenzyme intermediate. J Biol Chem 252(10):3121–3127.

21. Li J, et al. (2014) Palladium-triggered deprotection chemistry for protein activation inliving cells. Nat Chem 6(4):352–361.

22. Zhu Y, et al. (2007) Structural insights into the enzymatic mechanism of the patho-genic MAPK phosphothreonine lyase. Mol Cell 28(5):899–913.

23. Guillén Schlippe YV, Hedstrom L (2005) A twisted base? The role of arginine inenzyme-catalyzed proton abstractions. Arch Biochem Biophys 433(1):266–278.

24. Herzberg O, et al. (1996) Swiveling-domain mechanism for enzymatic phosphotransferbetween remote reaction sites. Proc Natl Acad Sci USA 93(7):2652–2657.

25. Nakanishi T, Nakatsu T, Matsuoka M, Sakata K, Kato H (2005) Crystal structures ofpyruvate phosphate dikinase from maize revealed an alternative conformation in theswiveling-domain motion. Biochemistry 44(4):1136–1144.

26. Lim K, et al. (2007) Swiveling domain mechanism in pyruvate phosphate dikinase.Biochemistry 46(51):14845–14853.

27. Minor W, Cymborowski M, Otwinowski Z, Chruszcz M (2006) HKL-3000: The in-tegration of data reduction and structure solution–from diffraction images to aninitial model in minutes. Acta Crystallogr D Biol Crystallogr 62(Pt 8):859–866.

28. Adams PD, et al. (2010) PHENIX: A comprehensive Python-based system for macro-molecular structure solution. Acta Crystallogr D Biol Crystallogr 66(Pt 2):213–221.

29. Emsley P, Lohkamp B, Scott WG, Cowtan K (2010) Features and development of Coot.Acta Crystallogr D Biol Crystallogr 66(Pt 4):486–501.

Fig. 6. Catalytic process of LmRPH-mediated RIF phosphorylation.

3808 | www.pnas.org/cgi/doi/10.1073/pnas.1523614113 Qi et al.

Dow

nloa

ded

by g

uest

on

June

1, 2

021

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1523614113/-/DCSupplemental/pnas.201523614SI.pdf?targetid=nameddest=SF7http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1523614113/-/DCSupplemental/pnas.201523614SI.pdf?targetid=nameddest=STXThttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1523614113/-/DCSupplemental/pnas.201523614SI.pdf?targetid=nameddest=ST2www.pnas.org/cgi/doi/10.1073/pnas.1523614113