Structure of the Oxygen Sensor in Bacillus subtilis: Signal Transduction of Chemotaxis by Control of Symmetry

Transcription

1 Structure, Vol. 11, , September, 2003, 2003 Elsevier Science Ltd. All rights reserved. DOI /S (03) Structure of the Oxygen Sensor in Bacillus subtilis: Signal Transduction of Chemotaxis by Control of Symmetry Wei Zhang 1 and George N. Phillips, Jr. 1,2, * 1 Department of Biochemistry and Cell Biology Rice University Houston, Texas Department of Biochemistry and Computer Sciences University of Wisconsin-Madison Madison, Wisconsin Summary Introduction Bacterial chemotaxis utilizes a network of sensing and regulatory proteins to guide bacterial swimming behav- ior in response to fluctuations of environmental conditions. Nearly all of the major protein components in this signaling transduction system have been identified (Manson et al., 1998; Mowbray and Sandgren, 1998). Six chemotactic receptors in E. coli and ten methylaccepting chemotactic proteins in B. subtilis have been discovered, and all of them are transmembrane proteins, except HemAT (Hanlon and Ordal, 1994; Hou et al., 2000; Mowbray and Sandgren, 1998). Molecular structures of most of the protein components in the chemotaxis path- way have been determined at the atomic level. These include various periplasmic binding proteins, the ligand binding domain of Tar, the cytoplasmic domain of Tsr, CheA, CheB, CheR, CheY, and CheZ, and other protein components (Bilwes et al., 1999; Djordjevic and Stock, 1998; Kim et al., 1999; Phillips et al., 1976; Spurlino et al., 1991; Stock et al., 1989; West et al., 1995; Yeh et al., 1993; Zhao et al., 2002). This knowledge helps us develop specific models of interaction of signaling trans- duction pathways. Extensive studies of chemotaxis in E. coli have made it an excellent model for interpreting complex and coop- * Correspondence: phillips@biochem.wisc.edu Much is now known about chemotaxis signaling trans- duction for Escherichia coli and Salmonella typhimu- rium. The mechanism of chemotaxis of Bacillus subtilis is, in a sense, reversed. Attractant binding strengthens the activity of histidine kinase in B. subtilis, instead of an inhibition reaction. The HemAT from B. subtilis can detect oxygen and transmit the signal to regulatory proteins that control the direction of flagella rotation. We have determined the crystal structures of the HemAT sensor domain in liganded and unliganded forms at 2.15 Å and 2.7 Å resolution, respectively. The liganded structure reveals a highly symmetrical orga- nization. Tyrosine70 shows distinct conformational changes on one subunit when ligands are removed. Our study suggests that disruption of the symmetry of HemAT plays an important role in initiating the chemotaxis signaling transduction cascade. erative behaviors at the molecular level. A more complicated system has been observed and investigated in B. subtilis. Recent studies suggest that B. subtilis utilizes a somewhat similar chemotaxis framework to enteric bacteria. Homologs of CheA, CheW, CheY, CheR, and CheB have been identified in B. subtilis (Bischoff and Ordal, 1992; Fabret et al., 1999). Experimental evidence has demonstrated that these proteins can function in the chemotaxis pathway of B. subtilis (Bischoff et al., 1993; Fuhrer and Ordal, 1991; Garrity and Ordal, 1995). Furthermore, the genes from B. subtilis can complement the behavior of null mutants in E. coli, and vice versa. The methylesterase CheB and the methyltransferase CheR from either bacterial species are capable of functional catalytic activity in both organisms (Garrity and Ordal, 1995). However, the chemotaxis pathway in B. subtilis is significantly different from that of E. coli in several as- pects. Two chemotaxis-related proteins, CheC and CheD, in B. subtilis do not have homologs in E. coli (Rosario et al., 1995). Researchers have investigated and proposed that they regulate formation of the CheW:CheA:receptor complex in B. subtilis; however, their detailed biological roles in B. subtilis chemotaxis are still under investigation (Kirby et al., 2001; Kristich and Ordal, 2002). The most prominent differences between two organ- isms are their default modes of movement and the ways information flows along the pathway. The default behavior of E. coli is smooth swimming, which can be strength- ened by decreased levels of phosphorylated CheY. The default swimming behavior of B. subtilis is tumbling, which can be converted to smooth swimming by a higher concentration of phosphorylated CheY (Bischoff et al., 1993). For E. coli, binding of attractants inhibits histidine kinase activity, which lowers the phosphorylation level of the response regulator, CheY. In contrast, for B. subti- lis, binding of attractants augments histidine kinase ac- tivity and increases the phosphorylation level of CheY. Taking into account the genetic evidence that some chemotactic proteins in B. subtilis can rescue the corre- sponding mutants in E. coli, we assume that, because of the opposite default swimming behavior between B. subtilis and E. coli, these microbes must adapt different molecular strategies at the receptor level when detecting environmental stimuli. From an evolutionary point of view, the ancestors of E. coli might have adjusted their chemotactic behavior at some point in the pathway, reversing the final function of CheY. It might be favorable for bacteria to move smoothly as the default, rather than to tumble until death under adverse or conflicting conditions (Garrity and Or- dal, 1995). There is not yet much structural information on chemotactic receptors from B. subtilis. Characterization of the recently discovered heme-based aerotaxis transducer (HemAT) will provide insight into molecular mech- anism for the first step of signal transduction of chemotaxis from B. subtilis. The heme-containing domain of HemAT has weak, but recognizable, sequence homol-

2 Structure 1098 ogy with other oxygen binding heme proteins, such as domain structure have been named according to the the dimeric hemoglobin from Vitreoscilla stercoraria and nomenclature of the globin classic fold. It has one helix sperm whale myoglobin (Hou et al., 2000, 2001). before the A helix, which we name the Z helix. According Here we report the crystal structures of the sensor to this naming convention, the sensor domain lacks a domain of HemAT in liganded and unliganded forms and D helix, so the helix arrangement in the HemAT sensor compare the structures with each other and with the domain is Z, A, B, C, E, F, G, and H. Partial electron structures of the liganded and unliganded states of the density of the N-terminal structure is observed in the A aspartate receptors from E. coli. subunit of our dimer, but not in the B subunit, because of disorder. The final liganded structure contains 169 residues for subunit A (residues ), 158 residues Results and Discussion for subunit B (residues ), 1 heme group, 1 cyanide ligand for each subunit, and 204 water molecules. The Overall Structural Organization unliganded structure has 168 residues for subunit A The crystal structure of the sensor domain of HemAT (residues ) and 156 residues for subunit B (resi- (amino acids 1 178) has been determined at 2.15 Å for dues ), with 186 water molecules and 1 heme a cyano-liganded form and at 2.7 Å for an unliganded group for each subunit. form. The multiple anomalous dispersion (MAD) diffraction from the iron atom was used to solve the phase problem of the liganded form, and the unliganded struc- Comparisons between Liganded ture was solved in the same crystal form after deliganding and Unliganded Structures the cyano-bound crystals (Zhang and Phillips, 2003). The overall structure of the unliganded form is very simi- There is 2-fold noncrystallographic symmetry for the lar to the liganded structure. The structural superpositwo subunits in one asymmetric unit. Two-fold symmetry tion was carried out with the ESCET program (Schneider, constraints were applied to the subunits during the initial 2002). The overall root-mean-square (rms) difference in refinement, but removed when the R factor plateaued the C atom position of the liganded and unliganded at 30%. The sensor domain is revealed to be a homodimeric forms is 0.42 Å when the two structures are superim- protein in which the dimerization interface forms posed with residues in subunit A and residues a four-helical bundle as a core, and this core is closely in subunit B. The magnitude of this difference packed with remaining helices (Figure 1). is greater than the individual coordinate errors estimated Like many isolated dimeric sensor domains in dilute from a Luzzati plot (mean rmsd of 0.22 Å for the liganded solution, HemAT is monomeric, but it is dimeric in crys- structure and 0.33 Å for the unliganded structure). The tals. The weaker dimerization is presumably due to the rms difference for the A subunits between the liganded loss of dimer-forming interactions from the methyl- and unliganded forms is smaller than that for the B accepting domain. For example, the sporulation response subunits. The rms difference between C atom positions regulatory protein Spo0B, the cytoplasmic do- in the A subunits of the liganded and unliganded struc- main of osmosensor protein EnvZ, and the E. coli PhoQ tures is 0.37 Å, versus 0.48 Å for the B subunits. This sensor domain show similar behavior (Hidaka et al., suggests that the two subunits of the HemAT sensor 1997; Lesley and Waldburger, 2001; Tomomori et al., domain undergo different rearrangements in the transi- 1999; Varughese et al., 1998; Zhou et al., 1997) (C. Bing- tion from the liganded to the unliganded states. We also man, personal communication). The dimerization interface see the heme group in the A subunit shift by 0.19 Å, of the sensor domain of HemAT has a buried sur- whereas the B subunit shifts by 0.49 Å. face area comparable to standard-size interfaces In the second analysis, a least-squares method was found in other protein complexes (Lo Conte et al., 1999). used to superpose subunits A and B within the dimer for Gel filtration chromatography of the sensor domain of each of the liganded and unliganded HemAT structures HemAT demonstrated that the higher-molecular weight (residues ) in order to reveal any intrasubunit dimeric form of the HemAT sensor domain is present, differences. The rms difference for C atom positions but not dominant, in dilute solution. for the liganded form is 0.39 Å, and the rms difference The structure-based sequence alignment of the for all atoms is 1.05 Å. The rms differences between the HemAT sensor domain, hemoglobin from V. stercoraria, two subunits in the unliganded form are larger than those and sperm whale myoglobin displays limited homology in the liganded structure, with a 0.47 Å difference for (Figure 2). The HemAT sensor domain has 25% sequence C atoms and a 1.22 Å difference for all atoms in the similarity to hemoglobin from V. stercoraria and only structure. This evidence indicates that the HemAT sen- 15% similarity to sperm whale myoglobin. The proximal sor undergoes small, but measurable, conformational residue (His) on the F helix of the three proteins is absolutely changes within the dimer when ligand is depleted from conserved. The distal pocket residues of HemAT the liganded form. and hemoglobin from V. stercoraria are similar and are To further characterize the details of the differences, shifted more toward the B helix and away from the E we carried out an objective comparison of the liganded helix than in myoglobins and vertebrate hemoglobins. and unliganded structures using an error-scaled differ- Despite these changes, the sensor domain subunit ence distance matrix method (Schneider, 2002). Comparison maintains a classic globin fold. There are eight helices was made not only between liganded and unlimaintains with one extended chain at the N terminus of each sub- ganded dimeric structures, but also between the two unit. In order to assist the comparison with other proteins individual subunits of each dimeric structure. in the globin family, helical segments in the sensor The plots produced by this method are advantageous,

3 Structure of the Sensor Domain of HemAT 1099 Figure 1. The Molecular Structure of the HemAT Sensor Domain Represented with Ribbon Diagrams (A) Stereo view of the structure. The signaling domain would be located further down on the page. (B) Top view showing the flanking of the core helices, G and H, by the rest of the molecule. The helices are labeled corresponding to the nomenclature of the globin fold. Subunit A, cyan; subunit B, yellow. ences between subunit A and subunit B. These results show that the liganded structure of the HemAT sensor has a more symmetrical organization than the unliganded structure. Moreover, significant differences are observed in the plot generated on the basis of the difference distance matrix of C atoms of the liganded and unliganded dimers (Figure 3C). The plot shows that differences be- tween the B subunits of each structure are larger than those in the A subunits. The largest displacements are observed in the B-C corner and the F and G helices. All of these variations occur near the heme pocket, which because they do not depend on the superposition of one structure upon another but compare internal distances between atoms. From the difference distance matrix plots comparing the A and B subunits, the liganded structure has smaller differences between its two subunits than does the unliganded form (Figures 3A and 3B). The plot of the liganded structure is basically featureless and suggests that its subunits have nearly identical conformation. However, the plot comparing subunits within the unliganded structure shows significant features. For instance, the residues from Leu73 to Ser77, which connect helices B and C, experience greater than 1 Å differ-

4 Structure 1100 Figure 2. The Structure-Based Sequence Alignment of Sperm Whale Myoglobin (Mb_Sw), the HemAT Sensor Domain (HemAT_Bs), and Hemoglobin from Vitreoscilla stercoraria (Hb_Vs) (A) Ribbon diagrams of the three proteins in their monomeric forms and the nomenclature for the helical segments in each structure. The helical arrangement in the HemAT sensor domain is Z, A, B, C, E, F, G, and H. (B) Alignment of three amino acid sequences on the basis of their structures. Secondary structures are shown at the bottom. Identical residues between all three sequences, green; identical residues between two sequences, yellow. Similar residues are colored in cyan. The critical distal pocket Tyr and His residues and the proximal pocket His residues are shown in red letters. might be expected, since this is where the ligand is bound. Only modest domain motions are observed in going from the liganded to the unliganded states. The largest change is that the F helix (residues ) on subunit B shows a 14 rotation with a Å translation, which is detected with the help of the program DynDom (Hayward et al., 1997). This motion is likely responsible for the dramatic orientation change of proximal His123 residues (Figure 5). A more detailed examination has been carried out on the dimerization interface with a superposition of two structures. The G and H helices of the two subunits form an antiparallel four-helical bundle. The H helices are likely continuous with the extended helical structure of the signaling domain of HemAT. As stated above, the rms differences for A subunits between the two structures are smaller than those for B subunits, but both G helices show larger displacement than H helices in going from the unliganded to liganded structures. In Figure

5 Structure of the Sensor Domain of HemAT 1101 Figure 3. Error-Scaled Difference Distance Matrix (DDM) Plots of the HemAT Sensor Domain (A) DDM plot of subunit A versus subunit B within the liganded structure. (B) DDM plot of subunit A versus subunit B within the unliganded structure. The result shows that there is less dimeric symmetry in the unliganded structure. (C) DDM plot of the liganded versus unliganded structures. The changes in distances smaller than 0.7 Å are omitted. The color gradient indicates differences between 0.7 Å and 1.4 Å, where red represents expansion and blue represents contraction. The results show moredramatic changes in subunit B, representing asymmetrical effects of the ligand departure.

6 Structure A, the directions of displacements of the C atoms on the proximal side. The hydrophobic composition of larger than 0.25 Å are shown with arrows. Their directions this pocket is similar to the binding site in dimeric hemo- point from the unliganded to the liganded form, globin from V. stercoraria, but the cavity of HemAT is and their amplitudes are exaggerated by a factor of 25 larger. The heme plane is buried deeply in the cleft, so because of the small size of the shifts. that there are more contacts with the protein matrix. These difference vectors for the G ( ) and H The cyanide ligand is bound to heme iron and stabilized helices ( ) were summed on main chain atoms by the hydroxyl group of Tyr70 with a hydrogen bond. (N, C, O, and C ) and averaged per atom. The averaged The cyanide ligand shows similar geometry in both binddifference vectors of each helix were then projected ing sites of the liganded structure. Moreover, there is onto a plane parallel to the H helices and onto a plane one distal pocket water within each binding site of the perpendicular to these helices. The estimated standard liganded structure. The distal pocket water is stabilized deviations of the average atomic motions is 0.04 Å, by a hydrogen bond to the oxygen atom on the Thr95 which were calculated with the individual error estimates side chain. A hydrogen bond is also formed between a of each coordinate propagated on the assumption of water molecule and the cyanide ligand. independent and random uncertainties according to The most dramatic change in going from the liganded Taylor (Taylor, 1982, page 56). to unliganded form is that the side chain of Tyr70 of In the unliganded structure, both G helices exhibit subunit B moves by 100 around its C -C bond and downward movement relative to the H helices (away shows some signs of disorder. In contrast, both Tyr70 from the presumed location of the membrane). The mag- residues in the liganded structure point toward the heme nitude for the averaged difference vectors of 76 atoms group and are stabilized by hydrogen bonding to the of 19 residues is 0.17 Å 0.04 Å for helix G of subunit ligand (Figure 5). Space-filling models of the binding B and 0.12 Å 0.04 Å for helix G of subunit A on this site in the unliganded structure indicate that there is no projection. Helix H shows a small upward movement in space available for either a ligand or water molecule in subunit B (0.04 Å 0.04 Å) and nearly no upward or the pocket. Movements of side chains of Leu92, Leu96, downward movement in subunit A (Figure 4C). Overall, and Phe69 into the cavity center contribute to the collapse the displacement between the G and H helices in subunit of the pocket as the protein matrix relaxes to the B is more significant than that in subunit A, even though unliganded state. the average displacement of individual atoms is small. This is consistent with the observation of a larger dis- Subunit Interactions placement of the heme in the B subunit and the increase The dimerization interface of the HemAT sensor domain in asymmetry going from the liganded to unliganded comprises two long helices (G and H), part of the Z helix, forms. Not only do these helices show a small translaand the B-C corner from each subunit (Figure 1). The tional movement, but they also show a small rotational buried intrasubunit contact surface is about 1800 Å 2 movement, which is visualized with difference vectors between these substructures, more than 20% of the projected on a plane perpendicular to these helices (Figwhole surface area of a monomer. Two hydrophobic ure 4C). Subunit B has a tendency to have small rotations stretches of amino acids are observed on the G and H relative to the position of subunit A. These rotations are helices, from Leu140 to Ile145 on G and from Leu158 only 1 3 estimated standard deviations in magnitude, to Ile162 on H. The two hydrophobic patches make close but still measurable. contacts with each other and with the same regions The helical displacements of the HemAT sensor doon the partner subunit. These patches of hydrophobic main are small, but not much smaller than the those interactions make up roughly one-third of the area of obtained for the Tar ligand binding domain, which shows the homodimeric interface. There is also a large water average displacements of 0.5 Å (Milburn et al., 1991; cavity present at the interface. The cavity is about 293 Å 3 Yeh et al., 1996). Biochemical studies of the cytoplasmic and contains six water molecules, which form a hydrodomain of the Tar receptor also suggest that the conforgen network with Thr166, Lys167, and Asn170. Thus, mational changes involved are smaller than the flexibility hydrogen bonding appears to be another energetic force allowed by an artificially introduced disulfide bond (Bass that stabilizes the dimer. and Falke, 1999). These results indicate that displace- One can compare the results with the dimerization ments in the signaling process are small and varied, but interaction of Spo0B. Both the HemAT sensor domain they still produce a downstream signal. and Spo0B form a similar four-helix hairpin structure at the dimer interface. But the interface of Spo0B is made primarily of hydrophobic residues, which contribute to Ligand Binding Pockets the stability of the dimer in solution and crystal (Varu- The C and E helices form one side of the portion of ghese et al., 1998). ligand binding site, with the B helix completing the distal pocket (Figure 2). The F helix runs nearly parallel to the heme plane and, therefore, is close to the heme over Comparisons with Other Globins the large span. These features plus these close contacts The structure of the HemAT sensor domain has a some- of the G and H helices make the HemAT sensor domain what more compact conformation than other globin proteins, more compact than other eight-helix globin proteins. such as sperm whale myoglobin. The H and G The ligand binding site of the sensor domain com- helices run nearly parallel to each other, making up the prises Phe69, Tyr70, Ile83, Leu92, Thr95, and Leu96 on dimer interface. Furthermore, because of the lack of a the distal side and His123 as the covalent attachment D helix, the C and E helices are connected closely. Such

7 Structure of the Sensor Domain of HemAT 1103 Figure 4. Ca Trace of the G and H Helices at the Dimerization Interface The unliganded structure, green; the liganded structure, yellow. These difference vectors for the G and H helices are shown in red arrows with a threshold of 0.25 Å. (A) Stereograph of side view of the helices. (B) Stereo view looking down into the cell. (C) Schematic representation of helical motions. Left, the averaged difference vectors projected onto a line parallel to the H helices. Right, the same vectors projected onto a plane perpendicular to these helices. Helices of subunit A (GA and HA), yellow; those of subunit B (GB and HB), green. The numerical values for each helix are shown with the estimated standard deviation (0.04 Å) for these helices based on error propagation of individual coordinate errors.

8 Structure 1104 structural rearrangements make the angle formed be- subunit B of the unliganded structure, which results in tween the E helix and heme plane smaller and compress a dramatic disruption of the symmetry between the two the crevice for holding the heme group. subunits. Other increases in asymmetry of the side chain However, the dimerization of HemAT is unique among packing are also seen. these dimeric hemoglobins. Unlike any other, the dimer- The dimerization interface reveals small, but perceptible, ization interface of the HemAT sensor domain is composed conformational changes that may be related to cheization of the G and H helices with part of the Z helix and motaxis signal transduction. Both G helices of the unliganded the B-C corner. These dimeric contacts in hemoglobins structure have a downward movement with have been reviewed in a previous report, and the HemAT respect to the equivalent segments of the liganded is again distinctive (Hargrove et al., 2000). The buried structure. The movement has a greater extent for the G surface area in HemAT dimers is larger than the contact helix of subunit B than that for subunit A (Figures 4A surfaces of either homodimeric hemoglobins from Scapharca and 4C). For the H helix neither shows a significant inequivalvis or from Caudina arenicola, which movement. Therefore, the resultant displacements of are about 1000 Å 2 (Hargrove et al., 2000; Tarricone et the G helices of the unliganded structure show down- al., 1997). However, it is interesting that the protein is ward movement with respect to the H helices on each monomeric at low concentrations during its isolation. subunit, with the G helix on subunit B showing some- This may not be surprising, since much of the coiled- what more displacement. coil domain of the extended signaling region has been When the averaged difference distance vectors are deleted. projected onto the plane perpendicular to the middle The proximal pockets of the HemAT sensor domain, line of the dimerization interface, we can see the relative myoglobins, and hemoglobins, including those from V. movement of individual helices (Figures 4B and 4C). All stercoraria, are very similar. Histidine on the F helix acts motions are greater than the estimated standard devia- as an anchor to covalently bind the prosthetic heme tion of 0.04 Å, although sometimes not by much. The group in the pocket, and hydrophobic side chains interact helical core of the unliganded structure has a tendency with the porphyrin ring around its perimeter. How- of counterclockwise rotational movements if viewed ever, the heme group of the HemAT sensor domain is down the plane perpendicular to middle line of the di- buried deeper than the heme groups in sperm whale merization interface. myoglobin and hemoglobin from V. stercoraria (Figure So, when the HemAT sensor domain undergoes the 6). The calculated cavity has a volume of 995 Å 3 for the transition from the unliganded to liganded state, the HemAT sensor domain, 772 Å 3 for Mb, and 847 Å 3 for conformational changes at the dimerization interface the hemoglobin from V. stercoraria, respectively. The will cause a downward shift of the G helices. If one takes HemAT protein still has no trouble binding small gaseous the G helices as the reference, H helices will have an ligands; however, the result is that the heme group in upward movement. (Since the H helices continue with HemAT has less solvent-exposed area than in Mb and the elongated signaling domain, it is not obvious what hemoglobins (Hou et al., 2000). to consider the reference point.) The rotational movements of the four helices have different magnitudes, which could cause some unwinding of the coiled-coil Symmetry Breaking in Signal Transduction signaling domain. From our analysis of the liganded structure above, the Our structures contrast with the observations that two subunits of the HemAT sensor domain have similar conformational changes in the ligand binding domain structures on the basis of the distance difference matrix of the aspartate receptor occur only on one helix (Milplot (Figure 3). In contrast, large variations are observed burn et al., 1991; Yeh et al., 1996). However, on the basis between the two subunits of the unliganded structure. of the ribose receptor (Trg), some mutational substitu- This indicates that the liganded structure has a more tions in the periplasmic domain (R71H, S72L, I78T, or symmetrical organization than the unliganded structure. Q79L) can induce the transmembrane signaling stronger Besides the small helical shifts and rotations of the four- than does wild-type (Beel and Hazelbauer, 2001a, helical bundle in the dimerization interface, the most 2001b; Yaghmai and Hazelbauer, 1992). These mutasignificant changes between the liganded and unli- tions mimic the presence of two ligands bound to the ganded structures occur at the ligand binding pocket. receptor. This finding suggests that Trg and, perhaps, Subunit A maintains a similar arrangement of side chain other chemoreceptors are able, under extreme condipacking, but subunit B undergoes dramatic changes tions, to change the conformations of two helices (Beel when the HemAT sensor domain relaxes to an unli- and Hazelbauer, 2001a, 2001b). Thus, it is very possible ganded structure from the liganded state. that the unliganded structure of the HemAT sensor do- In the ligand binding pocket of subunit A of the unli- main represents a similar state to that of the occupancyganded structure, the side chains of Phe69, Leu92, mimicked Trg receptor described above. Thr95, and Leu96 move slightly into the distal pocket to From our structural investigations and those of others, fill the empty space left by the ligand. Except for subunit two general hypotheses can be advanced for relating B of the unliganded structure, these side chains not only conformational changes in the sensor domains to move into the ligand binding pocket to fill the space left changes in the signaling domains of these receptors by ligand, but, also, the heme plane tilts by an angle of and, hence, kinase activity regulation. In the first model, 9 toward the F helix. These structural rearrangements the asymmetry of the sensor domain subunit is propalead to a flipping of the side chain of Tyr70 up and out gated to the signaling domains. When HemAT is unliganded of the pocket. The side chain flipping only occurs in or the aspartate receptor is liganded, the asym-

9 Structure of the Sensor Domain of HemAT 1105 Figure 5. Superposition of the Ligand Binding Sites of the Unliganded and Liganded Structures The unliganded structure, green; the liganded structure, yellow. (A) Left, side view of superposition of A sites; right, top view of A sites. (B) Left, side view of superposition of B sites; right, top view of B sites. The results show the dramatic motion of Tyr70 and the heme in the B subunit. (C) Left, electron density map of the B subunit of the liganded structure contoured at a level of 1 ; dashed lines represent hydrogen bonds. Right, electron density map of the B subunit of the unliganded structure contoured at a level of 0.7.

10 Structure 1106 Figure 6. Comparison of the Distal Pockets of the Liganded HemAT Sensor Domain and Sperm Whale Myoglobin (A) The distal pocket of the liganded HemAT sensor domain. (B) The distal pocket of the liganded sperm whale myoglobin. The result shows the deeper encapsulation of the heme in the sensor domain of HemAT. receptor level between two microorganisms, B. subtilis and E. coli. Six chemoreceptors found in E. coli are transmembrane proteins. On the basis of their sequence analysis and modeling studies, most of them very likely carry 260 Å-long cytoplasmic domains and form about 245 Å-long four-helical bundles (Kim et al., 1999). Unlike in E. coli, there are four types of chemoreceptors that have been discovered in B. subtilis. Type 1 chemorecep- tors have two transmembrane domains and a large periplasmic domain (about 230 residues). mcpa, tlpa, mcpb, tlpb, and mcpc are in this category. Type 2 chemoreceptors also have two transmembrane domains, but a small periplasmic domain (about 170 residues). TlpC and yvaq are in this category. Type 3 chemorecep- tors have only one transmembrane domain, but with about 150 residues in periplasmic space. YoaH is in this category. All of the proteins in these three categories have nearly the same length for the cytoplasmic domain, about 300 Å, which is 40 Å longer than that of E. coli chemoreceptors (Kim et al., 1999). Type 4 chemoreceptors are cytoplasmic proteins. HemAT and yfms are in this category. Taking this together with recent findings that chemoreceptors tend to form a planar network near metrical structures of both the sensor and signal domains are maintained by asymmetrical helix propagations, inhibiting kinase activity, perhaps by limiting kinase binding. The second model assumes that the asymmetry is present only in the sensing domains and that the signaling domains have symmetrical organization in all states. Any effects in the sensor domain would be transmitted through vectorial or rotational movements of symmetrical coiled coils to signaling domains. The structures of the isolated sensor or signaling domains cannot differentiate these two models. Receptor-Mediated Chemotaxis in Bacillus subtilis The chemoreceptors found in E. coli and B. subtilis have very high amino sequence similarity within their cyto- plasmic signaling domains, especially within the highly conserved domain (Hou et al., 2000; Le Moual and Kosh- land, 1996). In addition, they have the characteristic seven-residue repeat of coiled coils. This evidence sug- gests that these chemoreceptors probably adopt a simi- lar three-dimensional architecture. However, there are some significant differences at the

11 Structure of the Sensor Domain of HemAT 1107 Figure 7. A Model of the Four Known Types of Chemoreceptors Found in B. subtilis Type 1, purple; type 2, green; type 3, cyan; type 4, yellow. The lipid bilayer, blue and gray. The cytoplasmic domain of B. subtilis proteins is about 300 Å long, which is 40 Å longer than the cytoplasmic domain of E. coli chemoreceptors. (A) Side view of these chemoreceptors. (B) Bottom-up view from cytoplasm to periplasmic space showing how the type 4 chemoreceptors are likely to be arranged in the planar network of other receptors. the inside surface of the cell membranes to enhance Conclusions the signaling process, we have constructed a model The crystal structure of the HemAT sensor domain has for the relative positions of chemoreceptors of Bacillus the characteristic globin fold, though it only has very subtilis (Figure 7) (Ames et al., 2002; Bray et al., 1998; limited amino acids sequence homology to other hemo- Maddock and Shapiro, 1993). The implication is that the globin-like proteins. It is unlike the heme-based oxygen heme sensor domain of HemAT would sit about 40 Å sensors, FixL and Dos, whose heme-containing domain inside the membrane-sensing oxygen after it has dif- adopts a PAS fold (Fabret et al., 1999; Stock, 1997). fused through the membrane and over a small distance. The cavity-holding heme group is larger and deeper

12 Structure 1108 Table 1. Statistics of Data Collection and Refinement Liganded Unliganded Space group P P Unit cell parameters (Å) a 49.99, b 80.12, c a 49.45, b 79.72, c Resolution (Å) Observations 83,301 55,718 Unique reflections 18,722 9,226 Completeness (%) a 96.0 (83.7) 95.1 (74.0) R sym (%) a 6.3 (21.8) 9.8 (21.6) I/ a (5.35) (6.80) Refinement Statistics R crystal (%) R free (%) Model Statistics Protein atoms 2,666 2,621 Solvent atoms Rmsd bond lengths (Å) Rmsd bond angles ( o ) a Highest resolution shell parameters in parentheses. than other globins, and nearly all of the heme group play an important role in desensitizing partially liganded is buried within the protein scaffold. The homodimeric HemAT, like the methylation and demethylation pro- interface is composed of the entire G and H helices with cesses. The design of the HemAT structure seems partial structures from the Z helix and the B-C corner. elegantly to satisfy both sensitivity and robustness re- Thus, the dimerization interface is among the larger ones quirements for chemotaxis. seen for homodimeric hemoglobins. This also further Our structural evidence also readily explains the dif- demonstrates that hemoglobin-like proteins have evolved ferences between the signaling transduction pathways in such a way that the dimerization interface can be between E. coli and B. subtilis. The E. coli sensor goes located in a variety of regions on the protein surface from a symmetrical to an asymmetrical structure, inhibiting (Hargrove et al., 2000). the downstream histidine kinase upon ligand bind- The results of the kinetics and equilibrium binding ing (Biemann and Koshland, 1994; Milburn et al., 1991; experiments that the ligand binding is biphasic for Yeh et al., 1993; Yu and Koshland, 2001). In contrast, HemAT (W.Z., J. Olson, and G.N.P., unpublished data) the B. subtilis sensor goes from an asymmetrical to a suggest that HemAT might utilize heterogeneity or negative more symmetrical form on ligand binding, activating the cooperativity to expand the dynamic range for de- downstream histidine kinase. Thus, although the con- tecting the diatomic oxygen and transferring the structural nection between structural symmetry of the receptor information to the downstream histidine kinase. and kinase activation is conserved, the information This speculation includes the evidence obtained for switch is reversed at the level of the receptor, consistent other chemoreceptors, like Tar and Tsr, where it has with previous null mutation rescue studies. been suggested that negative cooperativity and heterogeneity are important parts of the molecular mechanism Experimental Procedures for the ligand binding reactions (Biemann and Koshland, 1994; Koshland, 1996; Lin et al., 1994). Expression and Purification of the HemAT Sensor Domain The expression and purification procedure has been described else- Along with the observation that the liganded form of where (Zhang and Phillips, 2003). Briefly, His 6 tag and factor Xa HemAT is more symmetrical than the unliganded form, cleavage site (Ile-Glu-Gly-Arg) sequences were added before the we propose the following model for signaling transduc- first residue of the sensor domain of B. subtilis HemAT (residues tion by HemAT. The unliganded form of HemAT is in an 1 178). E. coli [BL21(DE3)] cells bearing HemAT plasmid were grown energetically unfavorable form, not unlike the T state overnight with addition of 30 g/ml kanamycin. Cells were broken in hemoglobin. When environmental oxygen molecules either by a French cell press or lysozyme with a small amount of DNase1. The clear red supernatant was loaded onto a Ni affinity diffuse into the cytoplasm of B. subtilis, one subunit of column (Qiagen). The factor Xa enzyme (Novagen) digestion was the unliganded HemAT will bind one O 2 molecule. But, carried out to remove the amino-terminal residues. The protein was unlike in hemoglobin, the ligand binding affinity of the further purified with anion exchange and hydroxyapatite chromatography second subunit is decreased because of the structural (Pharmacia and BioRad). arrangement of first binding event. Thus, the partially liganded HemAT will exist often. As organisms move Crystallization, Data Collection, Structure Determination, along the oxygen gradient, they would meet more and and Refinement more oxygen molecules. Because of negative coopera- The protein used in crystallization was prepared by oxidizing purified HemAT with an excess of potassium cyanide. Crystals of the HemAT tivity or heterogeneity, it could take several orders of sensor domain were grown from 10% 18% sodium citrate and magnitude of increasing O 2 concentrations to saturate 100 mm KH 2 PO 4 (ph 7.0) with a 0.1% concentration of the detergent the second subunit, thus allowing a wide dynamic range n-octyl- -D-glucoside at room temperature. of responses. Other regulatory mechanisms may also Multiple-wavelength anomalous dispersion (MAD) data were col-

13 Structure of the Sensor Domain of HemAT 1109 lected at or near the iron edge at the APS BioCARS beamline. The plasmic domain of low-abundance chemoreceptor Trg that induce same crystal was used to collect three MAD datasets to 2.8 Å resolution or reduce transmembrane singalling: kinase activation and context and one native data set, which was complete to 2.15 Å resolu- effects. J. Bacteriol. 183, tion, with a 1 Å X-ray beam. The unliganded HemAT crystal was Biemann, H.P., and Koshland, D.E., Jr. (1994). Aspartate receptors obtained by reducing the liganded crystal with an excessive amount of Escherichia coli and Salmonella typhimurium bind ligand with of sodium dithionite and diffracted to 2.7 Å resolution. The time negative and half-of-the-sites cooperativity. Biochemistry 33, courses for the reduction reaction were tested to get a fully reduced form. The crystal color change indicated the transition from the liganded form to the unliganded form, as has been used for other Bilwes, A.M., Alex, L.A., Crane, B.R., and Simon, M.I. (1999). Struc- hemeproteins (Brucker et al., 1996; Hao et al., 2002; Quillin et al., ture of CheA, a signal-transducing histidine kinase. Cell 96, ). All data were processed on site with Denzo and Scalepack Bischoff, D.S., and Ordal, G.W. (1992). Bacillus subtilis chemotaxis: (Otwinowski and Minor, 1997). The MAD data were used to locate a deviation from the Escherichia coli paradigm. Mol. Microbiol. 6, the iron atoms but were not adequate to produce an immediately interpretable map (the correlation between peak and inflection Bischoff, D.S., Bourret, R.B., Kirsch, M.L., and Ordal, G.W. (1993). wavelengths was about Å resolution). For solving the phase Purification and characterization of Bacillus subtilis CheY. Biochemproblem, the identification of a 2-fold noncrystallographic symmetry istry 32, was crucial to give a traceable electron density map after density Bray, D., Levin, M.D., and Morton-Firth, C.J. (1998). Receptor clusmodification (see details in Zhang and Phillips [2003]). Refinement tering as a cellular mechanism to control sensitivity. Nature 393, was carried out with SHELXL for the liganded structure (Sheldrick et al., 1997). For the unliganded structure, the model from the liganded structure was adjusted to fit a simulated annealing omit map pre- Brucker, E.A., Olson, J.S., Phillips, G.N., Jr., Dou, Y., and Ikeda- pared with phases from the liganded structure. It was refined initially Saito, M. (1996). High resolution crystal structures of the deoxy, with simulated annealing protocols implemented in CNS (Brünger oxy, and aquomet forms of cobalt myoglobin. J. Biol. Chem. 271, et al., 1998). The final several cycles of refinement of the unliganded structure were also completed with SHELXL. The refinement meth- Brünger, A.T., Adams, P.D., Clore, G.M., DeLano, W.L., Gros, P., ods yielded nearly identical structures. There are no distance re- Grosse-Kunstleve, R.W., Jiang, J.S., Kuszewski, J., Nilges, M., straints applied between the ligand and ligand binding pocket heme Pannu, N.S., et al. (1998). Crystallography and NMR system: a new group during the refinement. Model building was performed with software suite for macromolecular structure determination. Acta XtalView (McRee, 1999). It should be noted that the distal pocket Crystallogr. D Biol. Crystallogr. 54, of subunit A of the unliganded structure is completely devoid of density at the ligand binding site. Subunit B, however, shows a small Carson, M. (1997). Ribbons. Methods Enzymol. 277, amount of residual density that lies too close to the relaxed position Djordjevic, S., and Stock, A.M. (1998). Chemotaxis receptor recogni- of Leu92 to represent significant ligand occupancy. The summary tion by protein methyltransferase CheR. Nat. Struct. Biol. 5, of the data collection and refinement statistics is listed in Table 1. Fabret, C., Feher, V.A., and Hoch, J.A. (1999). Two-component signal transduction in Bacillus subtilis: how one organism sees its world. Structural Analysis J. Bacteriol. 181, The cavity calculation was carried out for three proteins, the HemAT Fuhrer, D.K., and Ordal, G.W. (1991). Bacillus subtilis CheN, a homosensor domain, sperm whale Mb (Protein Data Bank code 1A6M), log of CheA, the central regulator of chemotaxis in Escherichia coli. and hemoglobin from V. stercoraria (PDB code 1VHB) with an algo- J. Bacteriol. 173, rithm presented in Liang et al. (1998). The buried area of the dimerization interface was calculated with the Protein-Protein Interaction Garrity, L.F., and Ordal, G.W. (1995). Chemotaxis in Bacillus subtilis: Server ( All of the how bacteria monitor environmental signals. Pharmacol. Ther. 68, molecular representation diagrams were produced with Ribbons (Carson, 1997). Hanlon, D.W., and Ordal, G.W. (1994). Cloning and characterization of genes encoding methyl-accepting chemotaxis proteins in Bacillus Acknowledgments subtilis. J. Biol. Chem. 269, Hao, B., Isaza, C., Arndt, J., Soltis, M., and Chan, M.K. (2002). Struc- We wish to thank Dr. John S. Olson at Rice University for helpful ture-based mechanism of O2 sensing and ligand discrimination by discussions. We also wish to thank staff at the BioCARS beamline the FixL heme domain of Bradyrhizobium japonicum. Biochemistry of the Advanced Photo Source for help collecting the diffraction 41, data. This work was supported by the Robert A. Welch Foundation Hargrove, M.S., Brucker, E.A., Stec, B., Sarath, G., Arredondo-Peter, [grant C-1142 (GNP)], the W.M. Keck Center for Computational Biol- R., Klucas, R.V., Olson, J.S., and Phillips, G.N., Jr. (2000). Crystal ogy, and the Wisconsin Alumni Research Foundation. structure of a nonsymbiotic plant hemoglobin. Structure 8, Received: December 19, 2002 Revised: June 3, 2003 Hayward, S., Kitao, A., and Berendsen, H.J. (1997). Model-free meth- Accepted: June 10, 2003 ods of analyzing domain motions in proteins from simulation: a Published: September 2, 2003 comparison of normal mode analysis and molecular dynamics simulation of lysozyme. Proteins 27, References Hidaka, Y., Park, H., and Inouye, M. (1997). Demonstration of dimer formation of the cytoplasmic domain of a transmembrane osmosensor Ames, P., Studdert, C.A., Reiser, R.H., and Parkinson, J.S. (2002). protein, EnvZ, of Escherichia coli using Ni-histidine tag affin- Collaborative signaling by mixed chemoreceptor teams in Escherichia ity chromatography. FEBS Lett. 400, coli. Proc. Natl. Acad. Sci. USA 99, Hou, S., Larsen, R.W., Boudko, D., Riley, C.W., Karatan, E., Zimmer, Bass, R.B., and Falke, J.J. (1999). The aspartate receptor cytoducers M., Ordal, G.W., and Alam, M. (2000). Myoglobin-like aerotaxis transplasmic domain: in situ chemical analysis of structure, mechanism in Archaea and Bacteria. Nature 403, and dynamics. Structure 7, Hou, S., Freitas, T., Larsen, R.W., Piatibratov, M., Sivozhelezov, V., Beel, B.D., and Hazelbauer, G.L. (2001a). Signalling substitutions in Yamamoto, A., Meleshkevitch, E.A., Zimmer, M., Ordal, G.W., and the periplasmic domain of chemoreceptor Trg induce or reduce Alam, M. (2001). Globin-coupled sensors: a class of heme-con- helical sliding in the transmembrane domain. Mol. Microbiol. 40, taining sensors in Archaea and Bacteria. Proc. Natl. Acad. Sci. USA , Beel, B.D., and Hazelbauer, G.L. (2001b). Substitutions in the peri- Kim, K.K., Yokota, H., and Kim, S.H. (1999). Four-helical-bundle

14 Structure 1110 structure of the cytoplasmic domain of a serine chemotaxis receptor. Nature 400, Three-dimensional structure of CheY, the response regulator of bacterial chemotaxis. Nature 337, Tarricone, C., Galizzi, A., Coda, A., Ascenzi, P., and Bolognesi, M. (1997). Unusual structure of the oxygen-binding site in the dimeric bacterial hemoglobin from Vitreoscilla sp. Structure 5, Kirby, J.R., Kristich, C.J., Saulmon, M.M., Zimmer, M.A., Garrity, L.F., Zhulin, I.B., and Ordal, G.W. (2001). CheC is related to the family of flagellar switch proteins and acts independently from CheD to control chemotaxis in Bacillus subtilis. Mol. Microbiol. 42, Taylor, J.R. (1982). An Introduction to Error Analysis: The Study of Uncertainties in Physical Measurements (Oxford University Press). Koshland, D.E., Jr. (1996). The structural basis of negative coopera- Tomomori, C., Tanaka, T., Dutta, R., Park, H., Saha, S.K., Zhu, Y., tivity: receptors and enzymes. Curr. Opin. Struct. Biol. 6, Ishima, R., Liu, D., Tong, K.I., Kurokawa, H., et al. (1999). Solution Kristich, C.J., and Ordal, G.W. (2002). Bacillus subtilis CheD is a structure of the homodimeric core domain of Escherichia coli histichemoreceptor modification enzyme required for chemotaxis. J. dine kinase EnvZ. Nat. Struct. Biol. 6, Biol. Chem. 277, Varughese, K.I., Madhusudan, Zhou, X.Z., Whiteley, J.M., and Hoch, Le Moual, H., and Koshland, D.E., Jr. (1996). Molecular evolution of J.A. (1998). Formation of a novel four-helix bundle and molecular the C-terminal cytoplasmic domain of a superfamily of bacterial recognition sites by dimerization of a response regulator phospho- receptors involved in taxis. J. Mol. Biol. 261, transferase. Mol. Cell 2, Lesley, J.A., and Waldburger, C.D. (2001). Comparison of the Pseustructure of the catalytic domain of the chemotaxis receptor meth- West, A.H., Martinez-Hackert, E., and Stock, A.M. (1995). Crystal domonas aeruginosa and Escherichia coli PhoQ sensor domains: evidence for distinct mechanisms of signal detection. J. Biol. Chem. ylesterase, CheB. J. Mol. Biol. 250, , Yaghmai, R., and Hazelbauer, G.L. (1992). Ligand occupancy mim- Liang, J., Edelsbrunner, H., and Woodward, C. (1998). Anatomy of icked by single residue substitutions in a receptor: transmembrane protein pockets and cavities: measurement of binding site geometry signaling induced by mutation. Proc. Natl. Acad. Sci. USA 89, 7890 and implications for ligand design. Protein Sci. 7, Yeh, J.I., Biemann, H.P., Pandit, J., Koshland, D.E., and Kim, S.H. Lin, L.N., Li, J., Brandts, J.F., and Weis, R.M. (1994). The serine (1993). The three-dimensional structure of the ligand-binding doreceptor of bacterial chemotaxis exhibits half-site saturation for main of a wild-type bacterial chemotaxis receptor. Structural comserine binding. Biochemistry 33, parison to the cross-linked mutant forms and conformational Lo Conte, L., Chothia, C., and Janin, J. (1999). The atomic structure changes upon ligand binding. J. Biol. Chem. 268, of protein-protein recognition sites. J. Mol. Biol. 285, Yeh, J.I., Biemann, H.P., Prive, G.G., Pandit, J., Koshland, D.E., Jr., Maddock, J.R., and Shapiro, L. (1993). Polar location of the chemore- and Kim, S.H. (1996). High-resolution structures of the ligand binding ceptor complex in the Escherichia coli cell. Science 259, domain of the wild-type bacterial aspartate receptor. J. Mol. Biol. Manson, M.D., Armitage, J.P., Hoch, J.A., and Macnab, R.M. (1998). 262, Bacterial locomotion and signal transduction. J. Bacteriol. 180, Yu, E.W., and Koshland, D.E., Jr. (2001). Propagating conformational changes over long (and short) distances in proteins. Proc. Natl. McRee, D.E. (1999). XtalView/Xfit a versatile program for manipulating atomic coordinates and electron density. J. Struct. Biol. 125, Zhang, W., and Phillips, G.N., Jr. (2003). Crystallization and X-ray Acad. Sci. USA 98, diffraction analysis of the sensor domain of the HemAT aerotactic Milburn, M.V., Prive, G.G., Milligan, D.L., Scott, W.G., Yeh, J., Jancarik, J., Koshland, D.E., Jr., and Kim, S.H. (1991). Three-dimensional Zhao, R., Collins, E.J., Bourret, R.B., and Silversmith, R.E. (2002). receptor. Acta Crystallogr. D Biol. Crystallogr. 59, structures of the ligand-binding domain of the bacterial aspartate Structure and catalytic mechanism of the E. coli chemotaxis phos- receptor with and without a ligand. Science 254, phatase CheZ. Nat. Struct. Biol. 9, Mowbray, S.L., and Sandgren, M.O. (1998). Chemotaxis receptors: Zhou, X.Z., Madhusudan, Whiteley, J.M., Hoch, J.A., and Varughese, a progress report on structure and function. J. Struct. Biol. 124, K.I. (1997). Purification and preliminary crystallographic studies on the sporulation response regulatory phosphotransferase protein, Otwinowski, Z., and Minor, W. (1997). Processing of X-ray diffraction Spo0B, from Bacillus subtilis. Proteins 27, data collected in oscillation mode. Methods Enzymol. 276, Accession Numbers Phillips, G.N., Jr., Mahajan, V.K., Siu, A.K., and Quiocho, F.A. (1976). Structure of L-arabinose-binding protein from Escherichia coli at The coordinates have been deposited in the Protein Data Bank 5Åresolution and preliminary results at 3.5 Å. Proc. Natl. Acad. Sci. under codes 1OR4 for the liganded HemAT sensor domain and USA 73, OR6 for the unliganded structure. Quillin, M.L., Arduini, R.M., Olson, J.S., and Phillips, G.N., Jr. (1993). High-resolution crystal structures of distal histidine mutants of sperm whale myoglobin. J. Mol. Biol. 234, Rosario, M.M., Kirby, J.R., Bochar, D.A., and Ordal, G.W. (1995). Chemotactic methylation and behavior in Bacillus subtilis: role of two unique proteins, CheC and CheD. Biochemistry 34, Schneider, T.R. (2002). A genetic algorithm for the identification of conformationally invariant regions in protein molecules. Acta Crystallogr. D Biol. Crystallogr. 58, Sheldrick, G.M., Schneider, T.R., and Diederichs, K. (1997). SHELXL: high-resolution refinement. Methods Enzymol. 277, Spurlino, J.C., Lu, G.Y., and Quiocho, F.A. (1991). The 2.3-Å resolution structure of the maltose- or maltodextrin-binding protein, a primary receptor of bacterial active transport and chemotaxis. J. Biol. Chem. 266, Stock, A.M. (1997). Energy sensors for aerotaxis in Escherichia coli: something old, something new. Proc. Natl. Acad. Sci. USA 94, Stock, A.M., Mottonen, J.M., Stock, J.B., and Schutt, C.E. (1989).