letters

570 nature structural biology • volume 9 number 8 • august 2002

Structure and catalyticmechanism of the E. colichemotaxis phosphataseCheZRui Zhao1,2, Edward J. Collins1, Robert B. Bourret1

and Ruth E. Silversmith1

1Department of Microbiology and Immunology, University of NorthCarolina, Chapel Hill, North Carolina, 27599-7290 USA. 2Current address:Department of Biochemistry and Molecular Genetics, University of ColoradoHealth Science Center, Denver, Colorado 80262, USA.

Published online: 24 June 2002, doi:10.1038/nsb816

The protein CheZ, which has the last unknown structure inthe Escherichia coli chemotaxis pathway, stimulates thedephosphorylation of the response regulator CheY by anunknown mechanism. Here we report the co-crystal struc-ture of CheZ with CheY, Mg2+ and the phosphoryl analog,BeF3

–. The predominant structural feature of the CheZ dimeris a long four-helix bundle composed of two helices fromeach monomer. The side chain of Gln 147 of CheZ inserts intothe CheY active site and is essential to the dephosphorylationactivity of CheZ. Gln 147 may orient a water molecule fornucleophilic attack, similar to the role of the conserved Glnresidue in the RAS family of GTPases. Similarities betweenthe CheY–CheZ and Spo0F–Spo0B structures suggest a gen-eral mode of interaction for modulation of response regula-tor phosphorylation chemistry.

Over the past 25 years, the molecular events governing chemo-taxis in Escherichia coli have emerged in detail1,2. As a con-sequence, bacterial chemotaxis is one of the most thoroughlycharacterized biological information processing networks, making it an excellent model for computer simulations in thefield of systems biology3,4. Bacterial chemotaxis has also servedas a prototype for two-component regulatory systems, a signal-ing transduction strategy widely used in prokaryotes, plants andfungi5 for response to a broad range of extracellular stimuli. Inchemotaxis, chemical attractants or repellants bound to trans-membrane receptors regulate the autophosphorylation of thesensor kinase CheA. The phosphoryl group is transferred from a histidyl residue on CheA to an aspartyl residue (Asp 57) on the response regulator CheY. Interaction of phospho-CheY (CheY-P) with the flagellar motor dictates cellular swimmingbehavior. Efficient removal of the phosphoryl group from CheY-P is essential for the continuous response to environmen-tal changes. Although CheY can catalyze its own dephosphoryla-tion, the protein CheZ in enteric bacteria such as E. colistimulates this rate and is critical for the rapid response of bacte-ria to stimuli6, which is essential for chemotaxis.

An invaluable component of our current level of understand-ing of chemotaxis is the atomic structures of six of the seven pro-teins in the pathway. Genetic and biochemical studies havegenerated considerable, but fragmented, data about the seventhprotein, CheZ, which could be put in context by the determina-tion of its structure. Protease sensitivity suggests that CheZ hasat least two structural domains7,8. Disabling single-site substitu-tions cluster in six regions of the CheZ primary structure8,9.Except for the C-terminus of CheZ, which is involved in binding

to the α4β5α5 surface of CheY-P10,11, the functional roles and spa-tial relationships of the CheZ regions identified by mutagenesisare unknown. Genetic studies suggest that α1

12 and the β5α5

loop13 of CheY bind to as yet unidentified portions of CheZ. Themechanism by which CheZ catalyzes the dephosphorylation ofCheY is also not known. CheY autodephosphorylation isbelieved to proceed through a direct reversal of the phosphoryla-tion reaction14, which involves an in-line substitution mecha-nism mediated by an active site Mg2+ (ref. 15). Whether CheZacts as an allosteric effector of CheY autodephosphorylationactivity or as a conventional phosphatase is not known. Possibleregulation of CheZ activity by CheY-P16 or a truncated form ofCheA17 have been reported but their molecular mechanismshave not been elucidated.

The lack of detailed structural or mechanistic informationconcerning CheZ has made our understanding of chemotaxisand, hence, its use as a model system, incomplete. Our goal wasto determine the three dimensional structure of CheZ with thehope that the structure would provide insight into its catalyticmechanism. The recent discovery that BeF3

– acts as a stable func-tional18 and structural analog19 of the phosphoryl group inCheY-P has allowed us to determine the co-crystal structure ofCheZ and CheY–BeF3

– –Mg2+. The structure completes the set ofatomic structures for the seven chemotaxis proteins and revealsthe fundamental mechanism of CheZ-stimulated dephosphory-lation of CheY.

Overall structureThe co-crystal structure of CheZ and CheY–BeF3

––Mg2+ wasobtained at 2.9 Å resolution (Fig. 1a,b). The ribbon representa-tion (Fig. 1a) shows that, consistent with solution studies20,CheZ is a dimer (CheZ2) composed of two monomers (Z1 andZ2), which are related by a two-fold rotational axis. The predominant structural feature of CheZ2 is a long (∼ 105 Å)four-helix bundle (‘CheZcore’). Residues 35–168 from eachCheZ monomer form two amphipathic helices with a singlehairpin turn (residues 100–104) and assemble in a ‘head-to-head’ orientation to form the bundle. Many inter- andintrachain interactions within the bundle, including 33 pairs ofresidues involved in hydrophobic interactions and 26 hydrogenbonds, suggest a highly stable dimer interface. An additionalhelix (residues 5–34) extends from the four-helix bundle at anangle of ∼ 100° and contains a majority of the interdimer crystalcontacts. A 13-residue helix (‘CheZc’), believed to correspondto the extreme C-terminus of CheZ (residues 201–213; seebelow), is in proximity to the α4β5α5 region of CheY but isunattached to the rest of the visible CheZ. In all, coordinatesfor 177 out of the total 214 residues of CheZ were determined.The remaining residues (1–4 and 169–200) are disordered inthe crystal.

Intermolecular interactions between CheY and CheZEach CheY binds to CheZ through two distinct interaction sur-faces (Fig. 1a). The interface between CheY and CheZcore has aburied surface of 1,215 Å2, which is comparable to the surfaceburied between a typical antibody and antigen21. This interfaceinvolves residues from α1, the β5α5 and β4α4 loops, and the activesite surface of CheY and residues between 136–151 of Z1 and67–71 of Z2 of CheZ (Fig. 2a; Table 1), about halfway down thefour-helix bundle. Consistent with this binding interface,replacement of Asn 23 in α1 of CheY with Asp diminishes CheZbinding by 30–50-fold12,22. Likewise, the β5α5 loop, containingLys 109 of CheY, which interacts with CheZ (Fig. 2a; Table 1),

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://s

tru

ctb

io.n

atu

re.c

om

letters

nature structural biology • volume 9 number 8 • august 2002 571

has been implicated by mutagenesis studies to be involved in theCheZ interaction13.

The second interface between CheY and CheZ involves inter-action of the α4β5α5 face of CheY with CheZc (Fig. 1a). The elec-tron density for CheZc is poor (see Methods), but the followingevidence supports the identity of CheZc as the C-terminus ofCheZ. The tryptic fragment CheZ196–214 binds to CheY in a phosphorylation-dependent manner11, and CheY residues per-turbed by CheZ196– 214 binding in NMR experiments10 correlatewell with the CheY-binding surface for CheZc. Furthermore,CheZ196–214 shares significant sequence identity with FliM1–16, apeptide that corresponds to the N-terminus of the flagellarswitch protein FliM; the two peptides bind to CheY-P with near-ly identical affinities10 and FliM1–16 binds at a similar location onactivated CheY as CheZc

23. The assignment of the 169–200region of CheZ as a disordered linker is consistent with sec-ondary structure predictions using several programs (data notshown), proteolysis data7,8,11 and amino acid alignments of CheZsequences from different species8. With two large independentareas of interactions, the hinged CheZ molecule ‘clamps down’on the globular CheY molecule, consistent with the tight bindingconstant (50–250 nM) between CheY-P and CheZ22.

Locations of loss- and gain-of-function mutantsMapping the positions of loss-of-function CheZ mutants, whichfall predominantly in six clusters on the primary amino acidsequence of CheZ8, onto the three-dimensional structure ofCheZ allows recognition of functionally important regions(Fig. 1c). The largest two clusters of mutants correlate with thetwo regions of CheZcore that interact with CheY and a smallermutant cluster correlates with the CheZc domain. Therefore,CheY interactions with both CheZcore and CheZc are critical forCheZ function. Two additional mutant clusters are located oneither side of the hairpin region such that residues from onecluster interact with residues from the other within a monomer.This region may, therefore, be important in maintaining theconformational integrity of the helical hairpin. Finally, a smallcluster of mutants maps to the linker region of CheZ (data notshown), implying that although disordered in the crystal, thelinker has functional significance.

CheZ dephosphorylation mechanismWhether CheZ acts as an allosteric activator of the intrinsicautodephosphorylation activity of CheY or as a true phos-phatase in its own right has long been debated. A striking feature

a b

c

d

Fig. 1 Overall structure of (CheY–BeF3––Mg2+)2CheZ2. a, Ribbon diagram

showing the topology of the CheY–CheZ structure. The CheZ2 chains arecyan and orange and the CheY molecules are gray. BeF3

– (green) andMg2+ (red) are in space-filling representation. The assignment of CheZc

with the nonproximal CheZ5–168 chain was arbitrary. b, Stereo representa-tion (BobScript38) of the electron density (2Fo – Fc map contoured at 1 σlevel) near the hairpin turn of the CheZ monomer. c, Positions of CheZloss-of-function (red) and gain-of-function (blue) mutants8 are mappedonto a GRASP39 surface representation of CheZ2. Only one CheY mol-ecule (magenta) is shown for clarity. d, Co-crystal structure of thephosphotransfer domain of the histidine phosphotransferase Spo0B andresponse regulator Spo0F29 (PDB entry 1F51). The two chains of theSpo0B dimer are cyan and orange. The Mg2+ ions are in red. (a) and (d)were created using RIBBONS40.

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://s

tru

ctb

io.n

atu

re.c

om

letters

572 nature structural biology • volume 9 number 8 • august 2002

of CheZcore–CheY interactions (Fig. 2a,b; Table 1) is the presenceof CheZ residues at the active site of CheY, suggesting thatresidue(s) from CheZ may participate directly in the catalysis ofCheY dephosphorylation. Most prominent is the side chain ofGln 147 of CheZ, which inserts directly into the CheY active site,filling one coordination site of the Mg2+ and making additionalvan der Waals interactions with Phe 14 of CheY and BeF3

–. Onehelical turn away, Asp 143 of CheZ forms a salt bridge with CheYactive site residue Lys 109. Gln 147 from CheZ is conservedamong all eight known CheZ sequences, and its central positionin the active site strongly suggests that it could have a catalyticrole. Cells containing a CheZ Q147A mutant display exclusivelyclockwise flagellar rotation, consistent with severely diminishedCheZ function. Although the Q147A mutant of CheZ still bindsCheY-P with moderate affinity (Fig. 2c) and is capable of forming co-crystals with CheY (data not shown), it displays nodetectable (<0.02% wild type) phosphatase activity (Fig. 2c).These data support an essential catalytic role for Gln 147 ofCheZ . A water molecule could be modeled into the active site(Fig. 2b), satisfying the geometry required for in-line attack ofthe beryllium atom (analog of the phosphoryl phosphorous)

and concomitant hydrogen bond formation with the amidenitrogen of Gln 147, suggesting a possible role for CheZ Gln 147in orienting the water for nucleophilic attack.

The current model of CheY autodephosphorylation involvesin-line attack of water on the phosphoryl phosphorous, generat-ing a pentavalent trigonal bipyramidal transition state that col-lapses into products14,24. The Mg2+, positioned in the active sitevia interactions with Asp 13, Asp 57, Asn 59 and the phosphorylgroup19, is required for catalysis14, perhaps for transition statestabilization25. There may also be other active site functionalgroups that contribute to catalysis which have not yet been iden-tified. The overall architecture of the CheY active site (includingthe positions of Asp 12, Asp 13, Asp 57, Asn 59, Lys 109, BeF3

–

and divalent cation) are similar in CheY–BeF3––Mg2+ bound to

CheZ and free CheY–BeF3––Mn2+ (ref. 19), suggesting that the

catalytic determinants for CheY autodephosphorylation are stillin place in the CheZ mechanism. Therefore, CheZ, with theinsertion of Gln 147 into the active site, seems to use the existingmechanism of CheY autodephosphorylation and render it moreefficient by positioning the attacking water molecule in theappropriate geometry for in-line attack. The proposed role does

Fig. 2 Interactions between CheY and CheZcore and potential catalyticmechanism. a, Residues involved in CheY–CheZcore interactions. Sidechains from CheY (gray), Z1 (cyan) and Z2 (orange) that are involved ininteractions are shown in ball-and-stick representation. BeF3

– is greenand Mg2+ is light purple. Labels for CheY residues are black and labels forCheZ residues are cyan (Z1) or orange (Z2). See Table 1 for details ofinteractions. b, Close-up view of the CheY active site in the co-crystal. Amodeled water molecule (2.0 Å from the Be atom) is in the correct geom-etry for in-line attack (narrow black stippled line) and within hydrogenbonding distance of the amide nitrogen of Gln 147 of CheZ (2.8 Å; thickblack stippled line) . Labels for CheY residues are black and CheZ residuesare cyan. Coordination of the active site Mg2+(magenta) is displayed asmagenta stippled lines. (a) and (b) were created using RIBBONS40. c, Biochemical data showing that CheZ Q147A is inactive and can stillbind to CheY-P. Upper panel: the effect of CheZ Q147A (circles) and wildtype CheZ (inset, squares) on the rate of Pi release from mixtures con-taining CheY and phosphoimidazole. Note the difference in scales of theabcissa. Rates were determined using the Enzchek Pi Assay (MolecularProbes)22. Lower panel: binding of wild type CheZ (squares), CheZ Q147A(circles) or CheZ1–181 (triangles) to CheY, assessed by their ability to com-pete with fluoresceinated wild type CheZ for CheY N59R-P.

a b

c

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://s

tru

ctb

io.n

atu

re.c

om

letters

nature structural biology • volume 9 number 8 • august 2002 573

not fall neatly into either the allosteric effector or phosphatasecategory but, rather, contains features of both.

Similarities to Ras GTPasesThe proposed mechanism for CheZ-mediated dephosphoryla-tion shares several similarities with the GTPase mechanism inthe Ras/Gα families. First, these GTPases also contain an activesite Gln residue, which has been proposed to orient a nucleo-philic water molecule for attack of GTP26, although in Ras it isthe amide carbonyl oxygen rather than the amide nitrogen thatdirectly interacts with the water. The codon for the conservedGln residue in p21ras is a common site of mutation, which resultsin cellular transformation27. Second, the general mechanism ofCheZ activity — stimulation of an existent hydrolysis mecha-nism by the insertion of a catalytic side chain — is directlyanalagous to the mechanism of the GAPs (GTPase activatingproteins), which insert an Arg residue into the Ras active site,assisting the existing GTPase mechanism through transitionstate stabilization28.

Possible generality of CheY–CheZcore interactionsAlthough CheZ does not show amino acid sequence similarity toany other proteins, the interactions between CheY and CheZcore

are strikingly similar to the interactions between the Bacillus subtilis histidine phosphotransferase Spo0B and response regulator Spo0F29 (Fig. 1d). Like CheZ, Spo0B contains a four-helix bundle composed of two helical hairpins assembled in ahead-to-head orientation. Although the helical bundle is much shorter in Spo0B than CheZ, the relative positions of the bundleand the response regulator are similar in the two complexes, as isthe surface of the response regulator, which interacts with thebundle. As with CheZ and CheY, two residues from Spo0B insertinto the active site of Spo0F. His 30 from Spo0B, the site of phos-phorylation, inserts into the Spo0F active site (resemblingGln 147 of CheZ) and Asn 34 from Spo0B forms a hydrogenbond with Lys 104 of Spo0F (resembling the CheZ Asp 143–CheYLys 109 interaction). A helical bundle structure is probably con-served among the phosphohistidine-containing domains inphosphotransferases and kinases of two-component systems24,although the overall topology can vary30. Thus, the nature of theinteractions observed in the CheY–CheZ and Spo0F–Spo0Bstructures may represent a general mode of interaction for modu-

lation of phosphotransfer reactions of response regulators,including His↔Asp or Asp→water phosphotransfer. Secondarystructure prediction suggests that each of the nine Rap proteins,which have phosphatase activity towards response regulatorsinvolved in B. subtilis sporulation31, are all-helical proteins (datanot shown). The Rap proteins may form helical bundles thatinteract with response regulators in a similar manner.

Roles of C-terminal and N-terminal helices of CheZWhereas the core domain of CheZ provides a catalytic residuefor the dephosphorylation reaction, CheZc is likely to providebinding and selectivity for activated CheY. CheZ196–214 binds toCheY-P and CheY with Kd values of 26 and 440 µM, respec-tively10. The structural basis for this selectivity is probably thesame as that for FliM1–16, whereby CheY activation induces rota-tion of Thr 87 and Tyr 106, removing Tyr 106 from the surface toan internal hydrophobic pocket and exposing the binding site23.In the CheY–CheZ structure, Tyr 106 from CheY is in the inter-nal conformation, and the external conformation is stericallyoccluded by CheZc. In light of the structural information, it issurprising that CheZ1–181 (Fig. 2c) and CheZ1–201 (ref. 11) cannotbind to CheY-P, and CheZ1–181 has no detectable phosphataseactivity (R.E.S., unpub. results). Thus, CheZc is critical for thebinding and activity of CheZ, possibly through one of the fol-lowing mechanisms. Binding of CheZc may induce a conforma-tional change on CheY-P, enhancing binding to CheZcore.Alternatively, binding of CheZc might bring CheZcore and CheY-Pto a high local concentration, which is critical for CheZ bindingand activity. Attempts to generate phosphatase activity in vitroby simultaneous addition of CheZ1–181 and CheZ196–214 wereunsuccessful (data not shown), arguing against the first model.

The role of the N-terminal helix in CheZ function remains tobe determined. It is striking, however, that the ∼ 20 known gain-of-function missense CheZ mutants contain amino acid substi-tutions that are concentrated on one face of this helix, as well ason regions in the 50s and 160s of CheZcore

9 (Fig. 1c). This sug-gests that, in the absence of CheY-P, the N-terminal helix mayfold against CheZcore, blocking access to the Gln 147 region.Gain-of-function substitutions may destabilize this interaction,thereby enhancing activity. This model would account for theapparent weak affinity of CheZ1–201 for CheY-P11. With the N-terminal helix precluding access to CheZ, a high local concen-tration of CheZcore, such as brought about by the CheZc tether,may be required to overcome the occlusion of the Gln 147region. Manipulation of the conformation of the N-terminalhelix — for example, by other chemotaxis proteins17 — couldprovide a means to regulate CheZ activity in the cell.

MethodsProtein purification, mutagenesis and activity assays. CheYand CheZ were purified using published protocols22,32. CheZ E134K,which confers a wild type Che+ phenotype and displays 80% of wildtype CheZ activity in vitro8, was used for crystallization becausewild type CheZ did not crystallize. Selenomethionine (SeMet)-CheYwas purified from cultures containing pRBB40 (ref. 33) trans-formed into the methionine auxotroph E. coli strain B834(DE3)(Novagen) and grown in the presence of SeMet (Sigma). Themutant cheZ Q147A was made by PCR mutagenesis. CheZ phos-phatase activity was determined by spectroscopic measurement ofthe rate of Pi release in the presence of CheY and phospho-imidazole22. The magnitude of catalytic activity of CheZ Q147Acompared to wild type CheZ was estimated by dividing the concen-tration of wild type CheZ required to obtain a minimal significantincrease in Pi release rate by the maximum concentration of CheZQ147A used. Binding of Che Z Q147A to CheY was assessed by mea-

Table 1 CheY–CheZcore interactions1

Location of residueZ1 Z2 YMet 136 Asn 23Met 137 Ile 20Asp 140 Met 17, Ile 20Gln 142 Thr 16Asp 143 Asp 12, Phe 14, Met 17, Lys 109Gln 147 Mg2+, BeF3

–, Phe 14Arg 1512 Glu 892

Glu 67 Arg 19Asn 71 Arg 19

1Interactions were defined as a maximum distance of 3.6 Å betweenatoms.2There is no electron density for the CheY Glu 89 side chain and the Cγand Cδ atoms of the CheZ Arg 151 side chain in the 2Fo – Fc map. TheCheY Glu 89 side chain was modeled using the Glu 89 conformation inactivated CheY (see Methods). The rotamer for CheZ Arg 151 was chosento best fit the rest of the side chain density. Thus, this interaction is con-sidered tentative.

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://s

tru

ctb

io.n

atu

re.c

om

letters

574 nature structural biology • volume 9 number 8 • august 2002

suring its ability to compete with fluoresceinated wild type CheZusing fluorescence anisotropy22. Fluoresceinated CheZ (0.20 µM)was mixed with varying amounts of nonfluoresceinated CheZ (wildtype, CheZ Q147A or CheZ1–181) in the presence of 17 mM acetylphosphate. CheY N59R (0.20 µM) was added, and the change inanisotropy was determined. CheY N59R binds nearly quantitativelyto CheZ under these conditions22.

Crystallization and data collection. Both the unsubstituted andSeMet-containing crystals were grown at 4 ºC using the hangingdrop vapor diffusion method. Drops contained 1 µl of the proteincomplex (264 µM CheY, 3.6 mM BeCl2, 10 mM NaF, 10 mM MgCl2 and264 µM CheZ E134K, assembled in the stated order) and 1 µl of wellsolution (0.1 M bicine, pH 8.5, 0.2 M ammonium acetate and 30%(v/v) isopropanol). Crystals grew for 2–3 weeks before reachingmaximal dimensions (0.5 × 0.1 × 0.1 mm). Only the combination ofSeMet-substituted CheY and unsubstituted CheZ produced usefulSeMet-containing crystals. Crystals were transferred into 2 µl of sta-bilization solution (0.1 M bicine, 30% (v/v) isopropanol, 1.8 mMBeCl2, 5 mM NaF and 5 mM MgCl2), followed by the sequential addi-tion of 2, 4 and 8 µl of cryoprotectant solution (stabilization solu-tion plus 50% (w/v) sucrose) at 5 min intervals. The crystals were

finally transferred to pure cryoprotectant solution and flash frozenin liquid nitrogen. Tb3+-derivitized crystals were prepared by growing the crystals using 10 mM TbCl3 in place of the MgCl2.Native and MAD data were collected at beam lines X4A and X25,respectively, of the National Synchrotron Light Source (NSLS) at theBrookhaven National Laboratory. All data were processed usingDENZO and SCALEPACK34. Data statistics are shown in Table 2.

Structure determination and refinement. Initial phases weredetermined using SOLVE35 with the combination of the SeMet MADdata and the TbCl3 MAD data. The SOLVE results clearly show thatP43212 is the correct space group instead of P41212. One Tb3+ site andfive Se sites were found by SOLVE (Table 2). The Tb3+ ion, surpris-ingly, binds to CheZ instead of at the expected Mg2+-binding site inCheY. The five Se sites were further refined in CNS36, and the SeMetMAD data alone were used to generate the initial map (addition ofthe Tb3+ data did not improve the phases), which was furtherimproved by solvent flattening with CNS. This experimental phasemap revealed a clear four-helical bundle structure for the CheZdimer but had poor density for CheY. Activated CheY (PDB entry1FQW)19 was fit in the experimental phase map based on the five Sesites found in SOLVE (r.m.s. deviation between the found Se sites

Table 2 Data collection and refinement

Data collectionNative SeMet TbCl3

inflection peak remote inflection peak remoteSpace group P43212Unit cell1 parameters (Å)

a = b 163.33c 54.17

Resolution (Å) 2.9 3.5 3.5 3.5 3.5 3.5 3.5Wavelength (Å) 1.0087 0.9789 0.9786 0.9562 1.6490 1.6486 1.5853Redundancy2 7.3 (6.8) 6.1 (4.9) 6.4 (5.8) 5.5 (3.3) 6.5 (4.2) 6.8 (5.2) 6.7 (6.1)Completeness (%)2 88.0 (89.6) 99.4 (99.5) 99.0 (99.9) 97.8 (91.9) 88.8 (85.2) 92.0 (84.9) 97.7 (92.9)Rmerge

2,3 7.4 (45.7) 8.4 (38.5) 8.9 (27.3) 9.6 (37.3) 6.2 (37.5) 5.9 (18.9) 8.1 (30.9)I / σ (I)2 22.3 (3.1) 19.4 (5.5) 20.7 (9.8) 15.1 (3.3) 16.3 (2.3) 18.6 (4.6) 19.5 (5.8)

Phasing statisticsResolution shell (Å) 11.16 7.54 6.04 5.18 4.61 4.19 3.87 3.61Figure of merit 0.68 0.67 0.62 0.55 0.54 0.48 0.44 0.33Mean figure of merit 0.53

Refinement statistics (|F| > 2 σ)Resolution (Å) 20–2.9Rworking (%)4 27.9 (10,767)Rfree (%)4,5 29.8 (563)Number of atoms

Protein 2,370Nonprotein (1 BeF3

–, 1 Mg2+, 1 bicine) 16R.m.s. deviation from ideal

Bond lengths (Å) 0.009Bond angles (o) 1.54

Ramachandran statistics6

Most favored 244Additionally allowed 32Generously allowed 0Disallowed 1

1One CheY–CheZ complex per asymmetric unit, 73% solvent content.2Numbers in parenthesis represent values for the highest resolution bin, which are 3.0–2.9 Å for the native data set and 3.62–3.50 Å for the otherdata sets.3Rmerge = Σ| Iobs – Iavg | / ΣIavg.4Number in parentheses is the number of reflections.5Rfree was calculated with 5% of reflections.6The residue in the disallowed region is CheY Asn 62, which is at the same conformation in the 2.37 Å CheY–BeF3

––Mn2+ structure19 and the 2.22 ÅCheY–BeF3

––Mg2+–FliM–peptide structure23.

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://s

tru

ctb

io.n

atu

re.c

om

letters

nature structural biology • volume 9 number 8 • august 2002 575

and the actual Sδ atoms of Met was 1.1 Å). A poly-Ala model forCheZ was built into the map and went through one round of posi-tional refinement. Side chains of CheZ were constructed based onthe CheZ Trp 94 and Trp 97 reference points, which are clear in the2Fo – Fc and Fo – Fc maps generated with the refined poly-Ala model.Further refinement against a native data set (20–2.9 Å) was carriedout in CNS interspersed with rounds of manual re-building in O37

based on the 2Fo – Fc and Fo – Fc maps. The final Rfree is 29.8%(Table 2). Fig. 1b shows the quality of the CheZ electron densityafter refinement. The side chains of the following CheZ residues arenot visible: 5, 8–10, 26, 33, 39, 46, 68, 77, 81, 108, 150, 151, 158, 168,201–213. The CheY density after refinement was still poor. Sidechains at the CheY–CheZ interface were well defined, but many ofthe side chains in the other regions of CheY were not clear. The sidechains of the following residues were visible for CheY: 4–6, 8, 12, 13,14, 16–25, 27, 30, 35, 37, 44, 46, 51, 53, 57, 59–61, 78, 81, 106,108–112, 116 and 119–121. Side chain conformations from activatedCheY (PDB entry 1FQW) were used when there was no clear densityfor CheY in the current structure. Main chain densities for the fivehelices of CheY were clear, but densities for the five β-strands werenot well separated. The poor density of CheY probably reflectsCheY flexibility in the crystal because it is not involved in any crystalcontacts and interacts only with CheZ. As a consequence, we do notdiscuss the structural differences between CheZ-bound CheY andfree CheY. Similarly, the side chain densities for the CheZc peptidewere visible but poor. There is strong evidence that this peptide cor-responds to the extreme C-terminus of CheZ (see main text). Thispeptide was tentatively modeled as CheZ 201–213, which loweredthe Rfree by almost 1%. However, we cannot preclude the possibilitythat this sequence could be misassigned by one or two residues ateither end of the peptide.

Coordinates. The structural coordinates have been deposited inthe Protein Data Bank (accession code 1KM1).

AcknowledgmentsWe thank J. Snyder and M. Kimple (UNC Chapel Hill) for help collecting thenative data set; D. Wemmer and S.-Y. Lee (UC Berkeley) for sharing the BeF3

–

parameter and topology files and the coordinates for the CheY–BeF3––FliM

structure before they were publicly accessible; G. Zhang (National Jewish Medicaland Research Center) and H. Ke, D. Worthylake and M. Redinbo (UNC ChapelHill) for helpful discussion; L. Betts (UNC X-ray facility) for technical support; X. Chen, M. Churchill and the X-ray Core Facility at the University of ColoradoHealth Science Center (UCHSC) for their support for the completion of R.Z.’sproject at UCHSC; and the Brookhaven beamline staff for help with datacollection.

Competing interests statementThe authors declare that they have no competing financial interests.

Correspondence should be addressed to R.E.S. email: silversr@med.unc.edu

Received 11 April, 2002; accepted 29 May, 2002.

1. Bren, A. & Eisenbach, M. J. Bacteriol. 182, 6865–6873 (2000).2. Bourret, R.B. & Stock, A.M. J. Biol. Chem. 277, 9625–9628 (2002).3. Yi, T.M., Huang, Y., Simon, M.I. & Doyle, J. Proc. Natl. Acad. Sci. USA 97,

4649–4653 (2000).4. Duke, T.A., Novere, N.L. & Bray, D. J. Mol. Biol. 308, 541–553 (2001).5. Stock, A.M., Robinson, V.L. & Goudreau, P.N. Annu. Rev. Biochem. 69, 183–215

(2000).6. Segall, J.E., Manson, M.D. & Berg, H.C. Nature 296, 855–857 (1982).7. Stock, A.M. & Stock, J.B. J. Bacteriol. 169, 3301–3311 (1987).8. Boesch, K.C., Silversmith, R.E. & Bourret, R.B. J. Bacteriol. 182, 3544–3552 (2000).9. Sanna, M.G. & Simon, M.I. J. Bacteriol. 178, 6275–6280 (1996).

10. McEvoy, M.M., Bren, A., Eisenbach, M. & Dahlquist, F.W. J. Mol. Biol. 289,1423–1433 (1999).

11. Blat, Y. & Eisenbach, M. Biochemistry 35, 5679–5683 (1996).12. Sanna, M.G., Swanson, R.V., Bourret, R.B. & Simon, M.I. Mol. Microbiol. 15,

1069–1079 (1995).13. Zhu, X., Volz, K. & Matsumura, P. J. Biol. Chem. 272, 23758–23764 (1997).14. Lukat, G.S., Stock, A.M. & Stock, J.B. Biochemistry 29, 5436–5442 (1990).15. Stock, A.M. et al. Biochemistry 32, 13375–13380 (1993).16. Blat, Y., Gillespie, B., Bren, A., Dahlquist, F.W. & Eisenbach, M. J. Mol. Biol. 284,

1191–1199 (1998).17. Wang, H. & Matsumura, P. Mol. Microbiol. 19, 695–703 (1996).18. Yan, D. et al. Proc. Natl. Acad. Sci. USA 96, 14789–14794 (1999).19. Lee, S.Y. et al. J. Biol. Chem. 276, 16425–16431 (2001).20. Blat, Y. & Eisenbach, M. J. Biol. Chem. 271, 1226–1231 (1996).21. Saul, F.A. & Alzari, P.M. Methods Mol. Biol. 66, 11–23 (1996).22. Silversmith, R.E., Smith, J.G., Guanga, G.P., Les, J.T. & Bourret, R.B. J. Biol. Chem.

276, 18478–18484 (2001).23. Lee, S.Y. et al. Nature Struct. Biol. 8, 52–56 (2001).24. Robinson, V.L., Buckler, D. & Stock, A.M. Nature Struct. Biol. 7, 626–633 (2000).25. Lukat, G.S., Lee, B.H., Mottonen, J.M., Stock, A.M. & Stock, J.B. J. Biol. Chem. 266,

8348–8354 (1991).26. Maegley, K.A., Admiraal, S.J. & Herschlag, D. Proc. Natl. Acad. Sci. USA 93,

8160–8166 (1996).27. Lowy, D.R. & Willumsen, B.M. Annu. Rev. Biochem. 62, 851–891 (1993).28. Scheffzek, K. et al. Science 277, 333–338 (1997).29. Zapf, J., Sen, U., Madhusudan, Hoch, J.A. & Varughese, K.I. Structure 8, 851–862

(2000).30. Mourey, L. et al. J. Biol. Chem. 276, 31074–31082 (2001).31. Reizer, J., Reizer, A., Perego, M. & Saier, M.H.J. Microb. Comp. Genomics 2,

103–111 (1997).32. Hess, J.F., Bourret, R.B. & Simon, M.I. Methods Enzymol. 200, 188–204 (1991).33. Bourret, R.B., Hess, J.F. & Simon, M.I. Proc. Natl. Acad. Sci. USA 87, 41–45 (1990).34. Otwinowski, Z. & Minor, W. Methods Enzymol. 276, 307–326 (1996).35. Terwilliger, T.C. & Berendzen, J. Acta Crystallogr. D 55, 849–861 (1999).36. Brünger, A.T. et al. Acta Crystallogr. D 54, 905–911 (1998).37. Jones, T.A., Zou, J.-Y., Cowan, S.W. & Kjeldgaard, M. Acta Crystallogr. A 47,

110–119 (1991).38. Esnouf, R.M. Acta Crystallogr. D 55, 938–940 (1999).39. Nicholls, A., Sharp, K. & Honig, B. Proteins 11, 281–296 (1991).40. Carson, M. J. Appl. Crystallogr. 24, 958–961 (1991).

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://s

tru

ctb

io.n

atu

re.c

om

corrigendum

erratum

nature structural biology • volume 9 number 9 • september 2002 711

Structure and catalytic mechanism of the E. colichemotaxis phosphatase CheZRui Zhao, Edward J. Collins, Robert B. Bourret and Ruth E. Silversmith

Nature Structural Biology 9, 570–575 (2002).

The PDB accession code for the structure of CheZ–CheY-BeF3– was incorrectly reported in the ‘Coordinates’ section of this paper.

The correct PDB accession code is 1KMI. We apologize for any inconvenience this may have caused.

Intrinsic metal binding by a spliceosomal RNASaba Valadkhan and James L. Manley

Nature Structural Biology 9, 498–499 (2002).

A mistake occurred during the production of this News and Views report. The observation stated at the bottom of the second col-umn on page 499 was incorrectly referenced to ref. 14. The correct reference for this observation is ref. 12, and the correct sentenceis printed as follows: “… the block to splicing due to thio substitution of either Sp oxygen in the domain V loop could not be rescuedby Mn2+ (ref. 12).” We apologize for any inconvenience this may have caused.

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

rest

ruct

ura

lbio

log

y