(Received for publication, December 18, 1996)
From the Department of Microbiology and Immunology,
University of Illinois, Chicago, Illinois 60612 and the
¶ Division of Biology, California Institute of Technology,
Pasadena, California 91125
An aspartate to lysine mutation at position 13 of
the chemotaxis regulatory protein CheY causes a constitutive tumbly
phenotype when expressed at high copy number in vivo even
though the mutant protein is not phosphorylatable. These properties
suggest that the D13K mutant adopts the active, signaling conformation
of CheY independent of phosphorylation, so knowledge of its structure could explain the activation mechanism of CheY. The x-ray
crystallographic structure of the CheY D13K mutant has been solved and
refined at 2.3 Å resolution to an R-factor of 14.3%. The
mutant molecule shows no significant differences in backbone
conformation when compared with the wild-type, Mg2+-free
structure, but there are localized changes within the active site. The
side chain of lysine 13 blocks access to the active site, whereas its
-amino group has no bonding interactions with other groups in the
region. Also in the active site, the bond between lysine 109 and
aspartate 57 is weakened, and the solvent structure is perturbed.
Although the D13K mutant has the inactive conformation in the
crystalline form, rearrangements in the active site appear to weaken
the overall structure of that region, potentially creating a metastable
state of the molecule. If a conformational change is required for
signaling by CheY D13K, then it most likely proceeds dynamically, in
solution.
Cells continuously monitor various chemical and physical parameters in their surrounding environment and use this information to implement appropriate adaptive responses to changing conditions. This vital task is accomplished by a cascade of transient protein phosphorylation and dephosphorylation events arranged so that the flow of phosphoryl groups reflects environmental conditions. One ubiquitous class of signal transduction networks is the "two-component" regulatory systems that control diverse processes such as behavior, development, physiology, and virulence in bacteria, as well as adaptive processes in eukaryotes (for reviews, see Refs. 1-4). The first component is a histidine protein kinase, which autophosphorylates using ATP as the phosphodonor. The second component is a response regulator protein, which transfers the phosphoryl group from the histidine protein kinase to itself. Response regulators are also capable of autophosphorylation using small molecule phosphodonors such as acetyl phosphate (5, 6). Phosphorylation alters the activity of the response regulator, thereby generating a response (e.g. change in gene expression) to the environmental stimulus.
Our goal is to understand in atomic detail the effects of phosphorylation on response regulator structure and function. The best characterized response regulator is the bacterial chemotaxis protein CheY, which has been the subject of extensive genetic, biochemical, and biophysical analyses. Phosphorylated CheY (CheY-P)1 causes flagellar rotation to change from the default counterclockwise direction to clockwise (7-10). The coordination of changes in swimming behavior with the temporal changes in chemical concentrations experienced by cells swimming in chemical gradients results in chemotaxis (11, 12).
The three-dimensional structure of CheY has been solved by both x-ray crystallography (13-16) and multidimensional NMR (17-19). Response regulator active sites are formed from five highly conserved residues of the response regulator superfamily (20). In CheY, these are Asp12 and Asp13, involved in binding a Mg2+ ion essential for the phosphorylation and dephosphorylation reactions (15, 16, 21); Asp57, the site of phosphorylation (22); Thr87, implicated in the postphosphorylation signaling events (23); and Lys109, also required for postphosphorylation steps in the signaling pathway (24). The structure of CheY-P has not yet been determined due to the fact that the phosphoryl group has a half-life of only a few seconds on CheY (8, 9).
We adopted a mutational approach to determine the active conformation of CheY. Phosphorylation is not the only route to response regulator activation. Mutations that result in constitutive activity have been isolated for many response regulators. In the case of CheY, replacement of Asp13 with lysine (the D13K mutant) results in clockwise flagellar rotation (25, 26). Genetic evidence that the CheY D13K protein is activated in the absence of phosphorylation implies that the phosphoryl group does not play a direct role in the mechanism of CheY signal transduction, such as phosphotransfer to a downstream protein, or direct binding (25, 26). These results, plus biochemical evidence that CheY activation does not involve a change to a multimeric state (25, 27), lead to the conclusion that the indirect role of phosphorylation is to generate a conformational change that allows CheY to interact productively with the flagellar switch. NMR data show that CheY undergoes a conformational change upon phosphorylation (28, 29). Presumably, this conformational change is mimicked by constitutively active mutants such as CheY D13K. The two routes of activation (phosphorylation versus the D13K mutation) are thought to have the same mechanistic features because Lys109 is essential for both (24, 26). We therefore solved the x-ray crystal structure of the D13K CheY mutant to find the activated conformation of CheY.
CheY D13K was purified as described
previously (30), except on a larger scale. A 100 liter culture of
KO641recA/pRBB40.13DK (25) was grown in LB + 100 µg/ml ampicillin in
a fermentor at 37 °C. Expression of CheY D13K was induced at a cell
turbidity of A595 = 1 by addition of
3-indoleacrylic acid to 100 µg/ml final concentration. Cells were
harvested after 2 h. Approximately 11 liters of cell paste were
lysed by sonication and clarified by centrifugation. TEDG buffer (50 mM Tris-HCl, pH 7.5, 1 mM EDTA, 2 mM dithiothreitol, and 10% v/v glycerol) was used
throughout the purification procedure. The cell lysate was loaded onto
a 2.5 × 10 cm (50 ml) Affi-Gel Blue column, washed, and eluted
with a 0-0.6 M NaCl gradient. Two passes were required on
the Blue column to separate the D13K from the flow-through. The
fractions containing the D13K were pooled, desalted by dilution, and
concentrated in a stirred cell ultrafiltration unit with a PM10
membrane (Amicon, Inc.). The concentrated D13K was loaded on a 2.5 × 10 cm (50 ml) DEAE-Sepharose CL-6B column and immediately eluted
with 0.025 M NaCl. CheY D13K was found to bind to DEAE less
tightly than wild-type CheY, presumably due to the +2 charge change
resulting from the mutation. Finally, all fractions containing D13K
were pooled, concentrated with Centricon-10 concentrators (Amicon, Inc.), and applied to a Superose 12 HR16/50 fast protein liquid chromatography gel filtration column (Pharmacia Biotech Inc.). Fractions containing pure D13K were stored at
20 °C.
Approximately half of the CheY D13K (molecular mass = 14 kDa) passed through the PM10 filter (molecular mass cutoff 10 kDa) at the pre-DEAE step. Passage through the filter resulted in substantial purification, so this material was retrieved, concentrated in a stirred cell ultrafiltration unit with a YM5 membrane (Amicon, Inc.), and chromatographed separately on the DEAE column. The overall yield from both halves of the purification process was 28 mg of D13K.
Protein CrystallizationThe purified protein was precipitated in a solution of 3.2 M ammonium sulfate, centrifuged, redissolved in a minimal volume of 80 mM ammonium sulfate, 50 mM Tris-HCl, pH 8.3, and brought to a final concentration of 15 mg/ml. The protein was then loaded into 20-µl microdialysis chambers, each sealed with dialysis membrane. Ammonium sulfate was the precipitant used in the crystallization trials. The critical ammonium sulfate concentration for nucleation and crystallization was found to be 2.2 M. The pulsed-diffusion method (31) was used as described previously (32) for optimization of crystal size and quality, where graduated up- and down-pulses were explored over the range of 0.1-2.5 M ammonium sulfate. The crystal used for data collection grew to dimensions of 0.90 × 0.08 × 0.08 mm.
Mg2+ is required in the phosphorylation and dephosphorylation reactions of wild-type CheY (5, 21, 33). However, Mg2+ binding to CheY D13K is not detectable by either resonance shifts in 19F NMR spectra of 4-fluorophenylalanine-labeled protein or by quenching of Trp58 fluorescence (28, 34). Lack of metal ion binding to the mutant protein is consistent with the observation that in wild-type CheY, Asp13 is directly involved in coordination of the Mg2+ ion (15, 16). Furthermore, D13K is phosphorylated only very slowly in the presence of the histidine protein kinase CheA and ATP (26, 27), and apparently not at all in the presence of acetyl phosphate (34). Thus, crystallization of CheY D13K in the absence of both Mg2+ and phosphate is biologically relevant.
Crystallographic Data Collection and ProcessingCrystals of
the CheY D13K mutant were indistinguishable from those of wild-type
CheY, belonging to the orthorhombic space group
P212121, with unit cell dimensions
of a = 45.9 Å, b = 47.0 Å, and
c = 54.0 Å. Diffraction data were collected from one
crystal at the area detector facility at Argonne National Laboratory. The crystal was exposed for a total of ~42 h in graphite
monochromatized Cu K radiation from a rotating anode
x-ray generator (50 kV, 100 mA) passing through a 0.3-mm collimator.
Data were recorded at a crystal to detector distance of 100 mm on a
Siemens multiwire area detector, with a swing angle of 15° in 2
for a theoretical upper resolution limit of 2.26 Å. Data processing
and reduction was done using the XENGEN program package (35). The cell
parameters determined by XENGEN deviated from wild-type by only 0.26%.
The final data set included a total of 12,828 observations of 5,820 unique reflections ranging from infinity to 2.26 Å. The unweighted absolute R-factor on intensities for merging all reflections
was 6.02%. Statistics for the final data set used in the
structural solution are given in Table I.
|
Because the CheY D13K
crystals were isomorphous with those of wild-type apoCheY, phasing of
the D13K structure was initiated with a phase calculation using the
wild-type structure with all solvent molecules excluded, and residue 13 was modeled as an alanine. Refinement of atomic positions was done
using the restrained least squares method (36). The R-factor
smoothly converged to a final value of 14.3% after a total of 110 cycles of refinement interspersed with 10 partial rebuilds. 4,432 reflections greater than 2 in the resolution range 10-2.26 Å were
used in the final refinement. The final refinement status is presented
in Table II.
|
Calculations of electron density maps and other data processing were carried out with the XTAL software package (37) and several locally developed programs. The least squares refinement was done with the software packages PROTIN and PROFFT (36, 38). Visualization of the electron density maps and model rebuilding were done using the graphics package FRODO (39) on an Evans and Sutherland PS300 system.
The electron
density maps from the final |2Fo Fc|
calc and
|Fo
Fc|
calc
difference Fourier calculations are clear and well defined for the
entire backbone of the molecule. The average thermal parameters for all protein atoms is 13.8 Å2 (Fig. 1). The
final coordinate set consists of 1108 atoms, of which 1003 are protein
and 105 are solvent molecules with unit occupancies.
Clearly defined electron density for the entire lysine residue at
position 13 confirms the amino acid substitution (Fig.
2). The amino group of the Lys13 side chain
makes no interactions with other groups of the active site. Instead,
the side chain is in a fairly extended conformation (1-4 = 70°,
156°, 152°, and 123°,
respectively), where the
-amino group has weak hydrogen bonding
contacts with external solvent molecules. The non-ideal value of
4 is due to a close, non-bonding contact of the terminal
group with the side chain of Asn59. Although the lysine
side chain at position 13 appears capable of assuming an alternate
rotameric position (t) away from the active site with no
steric restrictions, its observed position (g
)
in the structure blocks access to the active site, especially the
Mg2+ binding region. This is consistent with the fact that
the CheY D13K mutant does not bind Mg2+ (34) and exhibits
essentially no phosphorylation ability (26, 27, 34).
Comparison of Wild-type and CheY D13K Structures
A least
squares superposition of the Mg2+-free, wild-type, and D13K
mutant CheY structures shows a root mean square difference in positions
of equivalent -carbon atoms of just 0.21 Å and reveals only one
significant difference in the position of backbone atoms, 0.71 Å for
the
-carbon of Lys13, the site of the mutational change
(Fig. 3).
There are only two side chains that differ in position by any
significant extent between the two structures, Asp12 and
Asp57 (Fig. 4). In the CheY D13K structure,
Asp12 is rotated by 50° about its
2
bond, where its O
1 atom establishes a hydrogen bond with
the amide nitrogen of Phe14. Similarly, Asp57
is rotated by
60° about its
2 bond, so the
Asp57 O
2 atom loses its hydrogen bond with
the amide nitrogen of Asn59. The new position of the
Asp57 side chain is 1.1 Å closer to the Mg2+
binding site, a site now unoccupied because of the occlusion by the
Lys13 side chain. This repositioning of Asp57
eliminates its strong interaction with the
-amino group of
Lys109 seen in the wild-type CheY structure. The only other
change in the CheY D13K mutant is the pronounced lack of order in the
solvent structure of the active site. In wild-type CheY, the active
site contains some of the best ordered solvent molecules, whereas in the D13K structure, the solvent is poorly ordered and with much higher
temperature factors and weaker hydrogen bonding geometries (Fig.
4).
The presumably inactive, nonphosphorylated, wild-type CheY and the presumably activated D13K mutant CheY have backbone conformations in the crystalline state that are indistinguishable within experimental error. There are three possibilities that could account for this result. First, activation of CheY might not involve a conformational change. There is at least one example of lack of conformational changes in an enzyme following phosphorylation (40, 41), concerning the regulation of isocitrate dehydrogenase. However, this explanation does not apply to CheY because it is inconsistent with all previous experimental data (25, 27-29) as described in the Introduction. Second, both proteins might have crystallized in the active conformation. This is unlikely because the conformation observed in crystalline CheY D13K is the same as wild-type, apoCheY, both crystalline (13-15) and in solution (17-19), and the conformation of wild-type CheY has been shown to change upon Mg2+ binding (15, 16, 18, 28, 42) and phosphorylation (28, 29). Thus, these results strongly suggest the third possibility that the conformation of D13K observed here does not represent the activated state of CheY.
There are three possible explanations why the D13K CheY mutant might
have crystallized in an inactive conformation. First, it might assume
the active conformation only upon binding to the flagellar switch. This
seems improbable, and also not readily testable. Second, in solution,
D13K might truly mimic the phosphorylation-induced conformational
changes that are critical for activation, but lattice forces prevent
the solution-state conformational change from occurring in the crystal.
This seems unlikely in view of the facts that the crystallization
conditions, crystal morphology, space group, and unit cell dimensions
were identical for Mg2+-free, wild-type CheY and D13K.
Finally, the CheY molecule may normally exist in equilibrium between
the active and inactive conformations, and either phosphorylation or
the D13K mutation could serve to shift a sufficient fraction of the
population into the activated state to generate clockwise flagellar
rotation in vivo (Fig. 5). With this
explanation, the observed, crystalline conformation of CheY D13K would
be a metastable state, the conformation of the inactive majority of the
population, which could transform to the activated form in solution.
This possibility is consistent with results from 1H- and
15N-NMR spectroscopy on CheY D13K, which allow for a small
percent of the D13K mutant molecule to be in an alternate conformation, rapidly interconverting with its resting
state.2
Comparison of Wild-type and CheY D13K Activities
CheY binds specifically to three different proteins: CheA, which donates the phosphoryl group to CheY (43-46); CheZ, which stimulates dephosphorylation of CheY-P (43, 47-49); and FliM, the recipient of the activating signal of CheY, located in the flagellar switch complex (34, 50). Phosphorylation of wild-type CheY reduces binding to CheA but enhances binding to CheZ and FliM. Measurement of the binding affinities between CheY D13K and each of these three proteins may therefore indicate the degree of structural similarity between CheY D13K and CheY-P. D13K binds to CheA with about 2-fold lower affinity than nonphosphorylated wild-type CheY (46, 51), but its binding affinities for CheZ (47) and FliM (34) are the same as nonphosphorylated wild-type rather than the enhanced affinities expected for an activated conformation of CheY-P. Thus, the in vitro protein binding assays suggest that CheY D13K only partially assumes the activated conformation in solution.
Several other observations support the hypothesis that CheY D13K exists in a metastable state, in a conformation that can easily convert to the activated form. First, the hyper-signaling activity of CheY D13K is observed in a strain containing a multicopy plasmid (25) but is substantially reduced when the mutant protein is expressed from a single copy chromosomal gene (26). Second, the in vivo activity of wild-type CheY is increased by simple overexpression (52), and there is evidence that nonphosphorylated wild-type CheY possesses a low level of clockwise-generating activity (10). Finally, in the case of another response regulator, OmpR, overproduction of the non-phosphorylatable mutant D55Q (position 55 being the site of phosphorylation of OmpR) generates the phenotype characteristic of wild-type OmpR-P (53).
19F-NMR Results Regarding the Conformational State of CheY D13KThe CheY molecule contains six phenylalanine residues. Four of these (Phe8, Phe30, Phe53, and Phe124) are clustered far from the active site, Phe14 is solvent exposed, and Phe111 is buried underneath Lys109. 19F-NMR spectra obtained with 4-fluorophenylalanine-labeled CheY D13K did not show any differences in the environments of Phe8, Phe30, Phe53, and Phe124 between wild-type CheY and CheY D13K (26), suggesting that the D13K substitution does not induce a conformational change in this remote region. This is consistent with the conformation observed in the CheY D13K crystal structure. The resonance for 4-fluorophenylalanine-labeled Phe111 was perturbed, perhaps reflecting an effect of the D13K substitution on the position of Lys109. However, the environment of Phe111 is essentially identical in the crystal structures of wild-type CheY and CheY D13K. The only structural differences in the crystallographic results involve the two aspartate residues, Asp12 and Asp57, discussed above, which are more than 7 Å away from Phe111. Thus, the 19F-NMR work also suggests that CheY D13K adopts a conformation in solution different than that observed in the crystal structure presented here.
Conclusions and Possible Mechanisms for CheY D13K ActivationIn summary, the crystal structure of the constitutively active D13K mutant of CheY surprisingly has no conformational differences when compared with the Mg2+-free form of the wild-type CheY molecule. Thus the CheY D13K structure presented here is the inactive form of the molecule, and the interpretation is that only in solution does this form proceed to a conformation resembling activated CheY.
In all CheY structures published to date, (14-16, 23, 54, 56), the
Asp57 side chain assumes one of two general positions,
depending upon the presence or absence of a divalent cation
(Mg2+ or Ca2+) (Fig. 6). In the
apo forms, Asp57 and Lys109 have strong bonding
interactions, but introduction of Mg2+ or Ca2+
eliminates this interaction (15, 16, 54, 56). Phosphorylation of
Asp57 is certain to rearrange the active site even further,
requiring a repositioning of the Lys109 side chain (14).
Lys109 is known to be essential for stimulation of
clockwise flagellar rotation but not for phosphorylation or
autodephosphorylation of CheY (24, 26), and thus, it has a presumed
role in the propagation of a conformational change that causes
activation. The unusual position of the Asp57 side chain in
CheY D13K may reflect a predisposition of this mutant form of CheY to
readily assume the activated conformation in the absence of
phosphorylation.
An alternate activation mechanism is that CheY D13K might mimic a part of the CheY-P structure that is important for clockwise signal generation but lacks the conformation needed for efficient FliM binding since CheY D13K does not exhibit the enhanced binding to FliM in vitro observed with wild-type CheY-P (34). The fact that we found D13K in an "inactive" conformation is consistent with this, again suggesting that the activity resides in only a minor sub-population of the molecules. Thus it is possible that FliM binding and promotion of clockwise flagellar rotation may be separable events. It has been proposed elsewhere (10, 34, 55) that binding of CheY-P to the flagellar motor is a necessary, but not sufficient, event for signal transduction.
The atomic coordinates and structure factors (code pdb1ehc.ent and r1ehc.ss) have been deposited in the Protein Data Bank, Brookhaven National Laboratory, Upton, NY.
We thank E. Westbrook and M. Westbrook for help and generosity in data collection at Argonne National Laboratory.