2000 Nature America Inc.

Transcription

1 letters Structure of the CO sensing transcription activator CooA William N. Lanzilotta 1, David J. Schuller 1, Marc V. Thorsteinsson 2, Robert L. Kerby 2, Gary P. Roberts 2 and Thomas L. Poulos 1 1 Departments of Molecular Biology and Biochemistry and Physiology and Biophysics and the Program in Macromolecular Structure, University of California, Irvine, California , USA. 2 Department of Bacteriology, University of Wisconsin, Madison, Wisconsin 53706, USA. CooA is a homodimeric transcription factor that belongs to the catabolite activator protein (CAP) family. Binding of CO to the heme groups of CooA leads to the transcription of genes involved in CO oxidation in Rhodospirillum rubrum. The 2.6 Å structure of reduced (Fe 2+ ) CooA reveals that His 77 in both subunits provides one heme ligand while the N-terminal nitrogen of Pro 2 from the opposite subunit provides the other ligand. A structural comparison of CooA in the absence of effector and DNA (off state) with that of CAP in the effector and DNA bound state (on state) leads to a plausible model for the mechanism of allosteric control in this class of proteins as well as the CO dependent activation of CooA. a Information regarding the structural details of interaction between transcription factors and specific DNA sequences is well advanced. This is especially true for the camp receptor (CRP) family of transcription activators, of which the catabolite gene activator protein (CAP) 1,2 and the CO sensing protein CooA are members. In both cases, binding of an effector molecule switches the structure of the protein from a conformation that does not bind DNA (off state) to one that does (on state). In CooA this requires displacement of one heme ligand to enable the effector molecule, CO, to coordinate the heme iron. For CAP the mechanism by which binding of camp alters the DNA binding domain more than 20 Å away is not understood, primarily because the only structures of CAP available are in the presence of camp alone or camp and DNA 3 6. To provide insights into the effector-mediated allosteric switch and the chemistry of CO sensing, we have solved the crystal structure of CooA in the effector-free, or off state. The structure of CooA provides the first opportunity to understand how effector binding is related to the allosteric switch leading to site-specific DNA recognition in this class of proteins. Moreover, CooA is also part of a growing family of proteins that have adapted the heme for regulatory functions. Examples include the NO receptor guanylate cyclase and the O 2 sensor FixL, with CooA being the most recent addition to this list CooA is also the first metalloprotein known to have a clear biological role in CO recognition. Overall structure and general comparison with CAP The structure of reduced (Fe 2+ ) CooA and a comparison with CAP d b c Fig. 1 Comparison of the overall fold and conformational differences between CooA and CAP with camp and DNA bound. a, Stereo view of CooA and CAP dimers. In both cases, the effector binding domain (residues for Cooa, for CAP), C helix (residues for CooA, for CAP), and DNA binding domain (residues for CooA, for CAP) of monomer A are colored dark blue, dark pink, and purple, respectively. For monomer B, the effector binding domain, C helix, and the DNA binding domain are colored dark green, light green, and cyan, respectively. Both the heme groups of CooA and the camp molecules of CAP are represented by red space-filling models. The structure of CAP was adapted from Shultz et al. 4. b, Superposition of the effector domains of monomer B for CooA (shown in blue) and CAP (shown in orange) showing the relative orientations of the C helices. The r.m.s.d. for the alignment of the backbone atoms in the core structural elements of the effector domain was 1.16 Å. Strictly conserved Pro and Leu residues are shown at the N-terminal and C-terminal ends of the C helices, respectively. While an individual effector domain of CooA (green) aligns well with one of CAP (orange), the orientation of the C helix in the second domain is completely different. Of particular interest is the movement of the C helix (yellow arrow) when the structure of CooA (no effector bound) is compared with that of CAP (effector bound). c, View looking down the C helices from the N-terminus toward the C-terminus. The alignment is the same as shown in (b), with the yellow arrow representing movement from the effector free (CooA) to the effector bound (CAP) state. d, Alignment of effector domains in monomer B for CooA and CAP with the DNA binding domains shown. As can be seen, the relative position of the DNA binding domain and subsequently the position of the recognition helix in CooA is rotated 180 away from the position observed in the structure of CAP with effector bound. 876 nature structural biology volume 7 number 10 october 2000

2 letters a Fig. 2 Stereo view of the switch region demonstrating the positions of strictly conserved amino acids in the crystal structures of a, CooA and b, CAP. Monomer A and monomer B are colored in blue and green, respectively, in both (a and b). The residue numbering in (a) is for the primary sequence of CooA and the residue numbering in (b) corresponds to the primary sequence of CAP. b is shown in Fig. 1a. Both are homodimeric proteins composed of two distinct domains, a regulatory domain and a DNA binding domain. In CooA, residues form the regulatory domain with a single b-type heme bound in what would be the camp binding pocket of CAP. A sequence comparison between CooA and CAP reveals that the heme of CooA is accommodated by a deletion of residues in CAP. In addition, the N-terminus of CooA is longer to provide a ligand to the heme of the other subunit. A superposition of core structural elements in the effector domains of CAP and CooA results in a root mean square (r.m.s.) deviation of 1.16 Å for backbone atoms (Fig. 1b). Similar to CAP, the effector domains of CooA exhibit a β-roll architecture consisting of eight antiparallel β-strands and three helical segments 3,6. The overall structure of the DNA binding domain in CooA consists of the helix-turn-helix (HTH) core and four antiparallel β- strands. A superposition of the DNA binding domain of CooA with that of CAP results in an r.m.s. deviation of less than 1.0 Å for backbone atoms (data not shown). While the regulatory domains of CooA exhibit perfect two-fold symmetry, the DNA domains do not. This was also the case in early structures of CAP 3,6, although the deviation from perfect symmetry was not nearly as dramatic as seen in CooA. In CooA, one DNA domain is in the fully off state while the other, owing to crystal packing forces, has not completed the movement, but is still not in a position that can bind DNA as seen in CAP. This heterogeneity undoubtedly reflects flexibility in the hinge region (residues for CooA, Table 1) located at the end of the C helix between the regulatory and DNA domains. As has been observed in other reported structures of transcription activators 6,15,16, there is a striking variance in the B-factors of the effector domain compared to the DNA binding domain (Table 2), consistent with an inherent flexibility in the DNA binding domain. The dimer interface and allosteric switch Mutagenesis studies with CAP underscore the importance of the C helices along the dimer interface Mutating Ser 128 in the C helix can directly affect the cooperativity of ligand binding in CAP 17. In addition, Leu 127 in CAP is critical for the camp induced conformational changes that move the DNA binding domain into position for DNA interaction 18. While the C helix of a CooA monomer aligns well with CAP, the orientation of the symmetry related C helix at the dimer interface is notably different in CooA and CAP (Fig. 1b,c). The net result of effector binding thus appears to be bending and rotation of the C helices about the dimer interface. Compared to CooA, the C helices in CAP are significantly closer to the opposite subunit (Fig. 1b). This observation indicates that in order for CooA to adopt a DNA bound conformation, the entire heme and N-terminus must move upon binding of CO. Despite the close overall structural similarity between the separate domains of CAP and CooA, the relative positions of the regulatory domain and DNA binding domain are dramatically different for the two proteins (Fig. 1d). These differences can be attributed to structural changes in the highly conserved hinge or switch region at the carboxyl end of the C helix (Table 1). In monomer A of CooA, the C helix is the same length as seen in the CAP structure, due to a bend in the switch region near Met 131 of CooA or Ala 135 of CAP (Fig. 1). In contrast, helix C of CooA monomer B continues for another 13 residues. Stabilization of the off and on states The switch region consists of 10 residues centered at Phe 132 (Phe 136 in CAP) in the C helix (Table 1) that includes six strictly conserved amino acids, Leu-X-Phe 132 -X-Asp-X-X-X-Arg-Ile- Ala. The fulcrum for the switching mechanism is centered on a conserved Leu residue (130 in CooA, 134 in CAP). Consistent with this assigned role is the observation that, despite the very different orientations of the C helices in CooA and CAP (Fig. 1), the interaction of the symmetry-related Leu residues is not changed between the CooA and CAP structures (Fig. 2). The closest contact distance is 4 Å in both cases. Table 1 Sequence of CooA, CAP and FNR 1 1 Red indicates identity between two sequences and an underline indicates similarity. nature structural biology volume 7 number 10 october

3 letters Fig. 3 Stereo views of the heme environment in CooA. a, Back view and b, side view of the F o - F c omit map (green cage), contoured at 3σ by the simulated annealing protocol with Pro 2, Pro 3 and the heme omitted. c, Anomalous difference maps for data collected at 1.91 Å are shown contoured at 14 σ (purple cage) and 3 σ (green cage) for the heme iron and sulfur atom of Cys 75, respectively. The peak on the sulfur provides unambiguous confirmation on the correct orientation of the Cys 75 side chain. In going to the on state, one axial heme ligand must be displaced in order to allow CO to coordinate the heme iron. The N-Fe distance is 2.1 Å for both the His and Pro ligands with continuous electron density to the iron, so one bond cannot be judged to be more labile than the other based on the structure alone. Our model of the allosteric switch places the C helix of molecule B very close to the Pro in molecule A that coordinates heme B, suggesting that the allosteric switch involves displacement of the Pro ligand. However, if the entire heme were to move then it is possible that the Pro remains coordinated and that the His ligand is displaced. Our model cannot distinguish between these two possibilities. Phe 132 of CooA (Phe 136 of CAP) is located two residues after the Leu fulcrum and contacts made between Phe 132 and its local surroundings are totally different in CooA and CAP. In the fully off subunit of CooA, Phe 136 contacts the recognition helix (centered at Thr 182) in the DNA binding domain. In CAP, however, the homologous Phe 136 of one subunit contacts the hairpin turn consisting of residues in the effector domain of the other subunit, and vice versa. The primary interaction is stacking of the phenyl ring with the peptide backbone. The sequence for CAP in this region (K 52 DEEGK) is similar to that in CooA (V 57 GEERES) and the hairpin structure is conserved. Another important change involves a conserved Arg 138 residue (142 in CAP) in monomer A that interacts with Glu 59 (54 in CAP) in the hairpin loop of monomer B (Fig. 2). In going from the off to on states, this salt bridge ruptures and is replaced by an interaction between the conserved Arg and the DNA phosphate backbone. Again there is no net change in interactions, only a swapping of interacting partners. The movement of Arg 138 is of central importance because of the conserved hydrophobic residue that immediately follows. Ile 139 in CooA (Ile 141 in CAP) is part of a strictly conserved hydrophobic core found in many other transcription activators that contain the HTH motif 20. Hence, disruption of the salt bridge in going to the on state allows this Arg residue to adopt the DNA bound conformation and act as a lever that pulls the DNA binding domain into position (Fig. 2). In addition to these changes, the nonpolar interactions between the C helices and the β-strands of the hairpin loop tighten in the off state. Ile 127 in subunit B is 4.2 Å from Leu 56 and 5.7 Å from Ile 63 of subunit A while Ile 127 is 4.7 Å from its symmetry related mate. In the on state the homolog to Ile 127, Val 131 of CAP, is now only 3.5 Å from its symmetry related partner as well as 5.0 Å and 3.8 Å from Ile 51 and Met 59, respectively. Disruption and loss of some helical structure in the C helix in going from the off to on states is at least partially compensated for by an increase in nonbonded contacts between helix C and its neighboring effector subunit. The importance of the switch region just described is fully consistent with site a b c directed mutagenesis studies on CAP and the picture of the allosteric switch that has emerged from these earlier studies Heme ligands A major unknown in CooA has been the identity of the second heme ligand trans to His 77. This is of critical importance to CooA function since one of the axial heme ligands must be displaced in order to allow CO to bind in the allosteric switch from the off to the on state. An unprecedented finding is that the ligand for the heme in monomer A is provided by the nitrogen atom from the N-terminal Pro 2 in monomer B (Figs 1, 3). N- terminal sequencing of purified protein and from dissolved crystals confirmed the presence of an N-terminal Pro residue. This is the first example of a Pro nitrogen atom being used as a ligand in a metalloprotein and is unusual considering the relatively high pk a of the Pro nitrogen (pk a = 10.4). An important observation is the location of Cys 75 relative to the His 77 heme ligand (Fig. 3c). There is conclusive evidence that in the oxidized (Fe 3+ ) state Cys 75 is the heme ligand, but is replaced by His 77 upon reduction to Fe 2+ (refs 22 25). Because the X-ray diffraction data were collected at longer wavelengths, anomalous scattering of the sulfur atoms in CooA was observed (Fig. 3c) and the position of the sulfur atom in Cys 75 can be accurately placed. Given the relative positions of the heme Fe and the side chains of His 77 and Cys 75 (4.8 Å, iron to sulfur distance), the inescapable conclusion is that the ligand change requires considerable movement of the protein and/or heme. The idea that the heme environment of CooA is very dynamic is consistent with recent work that investigated NO interaction with CooA nature structural biology volume 7 number 10 october 2000

4 letters Table 2 Data and refinement statistics Data statistics Wavelength (Å) Resolution range (Å) Unique reflections Completeness (%) R 1 sym , , , , ,881 (Outer shell I / σ = 1.9) Refinement statistics Protein nonhydrogen atoms 3,322 Solvent molecules 199 Resolution range (Å) Total reflections 18,474 Total reflections used in R free 985 R cryst R free R.m.s. deviations Bond distances (Å) Bond angles ( ) 1.26 Wilson B-factor (Å 2 ) 37.4 Average B-factors (Å 2 ) Solvent 64.5 Effector binding domains and hemes 56.3 DNA binding domain in monomer A 67.4 DNA binding domain of monomer B R sym = hkl [ I ( I hkl,i - I hkl )] / hkl,i I hkl, where I hkl is the intensity of an individual measurement of the reflection with indices hkl and I hkl is the mean intensity of that reflection. Contacts from the vinyl groups of the heme and side chains of the C helix are <4 Å and include Leu residues 112, 116, and 120. Trp 110 also makes a contact of <3.5 Å with the heme from the opposite subunit. The position of the hemes across the dimer interface may be crucial for eliciting conformational changes and important for the cooperative binding of CO. Supporting this conclusion are His 77 variants that still bind CO in a cooperative manner, but do not make CooA competent for DNA binding 26. This further implies that the structural changes responsible for communication between the hemes are insufficient to induce the full conformational shift required for DNA binding. Interestingly, His 77 also has considerable solvent exposure with the closest protein contact to this side chain from the Oδ atom of Asn 42 to the Nδ atom of His 77. Mutagenesis of Pro 2 and residues in the N-terminus is currently being performed to address the importance of this region in the CO induced conformational changes. Regardless of these insights, given the structural comparison presented here between the off state of CooA and on state of CAP coupled with the analysis of NO interaction with CooA 25, it is tempting to speculate that the CO induced conformational changes in CooA involve displacement of one ligand and movement of the entire heme along the dimer interface. Methods Protein purification and crystallization. CooA was purified essentially as described 14. Crystals of the reduced (Fe 2+ ) CooA protein were obtained by the following anaerobic procedure. A solution consisting of 0.1 M Tris ph 8.5, 0.6 M MgCl 2, 20% PEG 8 K (v/v), and 20% glycerol (v/v), and a solution containing purified CooA protein (10 mg ml -1 ) were made anaerobic by repeated exchange of the sample gases with O 2 -free argon on a vacuum manifold. Capillaries for crystal growth were 200 µl micropipets that were flame sealed at one end. Capillaries were set up and crystal growth was carried out in a soft-sided glove box (Coy, Grass Lake, Michigan) under an environment of 95% nitrogen and 5% hydrogen. Sodium dithionite was added to a final concentration of 10 mm in all solutions. The precipitation solution (100 µl) was placed in the bottom of the capillary and 80 µl of the protein solution was layered on top. Capillaries were sealed with sticky wax (Hampton Scientific) and allowed to equilibrate. After 5 days diffraction quality crystals were obtained. Given the evidence for a redox dependent ligand exchange in CooA and the time required to obtain crystals, it was important to establish the oxidation state of the heme iron prior to data collection. A singlecrystal absorption spectrum was used to confirm that CooA was in the ferrous form (data not shown). Prior to freezing and data collection, the concentration of glycerol in the mother liquor was increased to 30% (v/v) by soaking the crystals in fresh mother liquor with glycerol (5% increments). Data collection and processing. Data were collected with a charged coupled device (CCD) detector (ADSC) and image plate scanner (Mar Research) at Stanford Synchrotron Radiation Laboratory (SSRL) beamlines 1-5 and 7-1, respectively. Multiwavelength anomalous diffraction (MAD) data were collected at four wavelengths (Table 2) using the inverse beam mode. Data were indexed, integrated and scaled using Denzo and Scalepack 27. Subsequent data processing was carried out with the CCP4 software suite 28. Phasing and refinement. The crystal structure of CooA was solved by MAD methods from the anomalous signal of the heme iron. Reasonable solutions for the positions of two iron atoms were determined from the analysis of anomalous difference Patterson maps for the enantiomorphic space groups P and P The space group ambiguity was addressed by refinement of the iron positions and phases in both space groups (SHARP) 29. Only the space group P resulted in interpretable electron density maps. Within these preliminary maps the heme cofactors and C helices could be placed. Through several cycles of model building and nature structural biology volume 7 number 10 october

5 letters refinement a model was produced. The model was refined to 2.6 Å (Table 2) using the CNS 30 package of programs. Coordinates. The coordinates for this structure have been deposited in the Protein Data Bank (accession code 1FT9). Acknowledgments The authors would like to thank M. Sundaramoorthy and H. Li for help in the initial collection and processing of data. W.N.L. would also like to thank A.L. Pearcy for providing housing in Corona del Mar/Newport Beach during the completion of this work. The authors were supported by the United States Department of Agriculture (W.N.L.), National Science Foundation (T.L.P.), and National Institutes of Health (M.V.T. and G.P.R.). This work is based upon research conducted at the SSRL, which is funded by the Department of Energy (BES, BER) and the National Institutes of Health (NCRR, NIGMS). Correspondence should be addressed to T.P.L. poulos@uci.edu Received 18 May, 2000; accepted 21 August, Busby, S. & Ebright, R.H. J. Mol. Biol. 293, (1999). 2. Crasnier, M. Res. Microbiol. 147, (1996). 3. McKay, D.B., Weber, I.T. & Steitz, T.A. J. Biol. Chem. 257, (1982). 4. Schultz, S.C., Sheilds, G.C. & Steitz, T.A. Science 253, (1991). 5. Vaney, M.C., Gilliland, G.L., Harman, J.G., Peterkofsky, A. & Weber, I.T. Biochemistry 28, (1989). 6. Weber, I.T. & Steitz, T.A. J. Mol. Biol. 198, (1987). 7. Craven, P.A. & DeRubertis, F.R. Biochim. Biophys. Acta 745, (1983). 8. Ignarro, L.J., Degnan, J.N., Baricos, W.H., Kadowitz, P.J. & Wolin, M.S. Biochem. Biophys. Acta 718, (1982). 9. Yu, A.E., Hu, S.Z., Spiro, T.G. & Burstyn, J.N. J. Am. Chem. Soc. 116, (1994). 10. Gilles-Gonzalez, M.A., Gonzalez, G. & Perutz, M.P. Biochemistry 34, (1995). 11. Gilles-Gonzalez, M.A., Ditta, G.S. & Helinski, D.R. Nature 350, (1991). 12. Rodgers, K.R., Lukat-Rodgers, G.S. & Barron, J.A. Biochemistry 35, (1996). 13. Aono, S., H., N., Saito, K. & Okada, M. Biochem. Biophys. Res. Commun. 228, (1996). 14. Shelver, D., Kerby, R.L., He, Y. & Roberts, G.P. Proc. Natl. Acad. Sci. USA 94, (1997). 15. Parkinson, G., et al. J. Mol. Biol. 260, (1996). 16. Bell, C.E. & Lewis, M. Nature Struct. Biol. 7, (2000). 17. Cheng, X., Kovac, L. & Lee, J.C. Biochemistry 34, (1995). 18. Leu, S.F., Baker, C.H., Lee, E.J. & Harman, J.G. Biochemistry 38, (1999). 19. Cheng, X. & Lee, J.C. Biochemistry 37, (1998). 20. Shaw, D.J., Rice, D.W. & Guest, J.R. J. Mol. Biol. 166, (1983). 21. Kim, J., Adhya, S. & Garges, S. Proc. Natl. Acad. Sci. 89, (1992). 22. Ryu, S., Kim, J., Adhya, S. & Garges, S. Proc. Natl. Acad. Sci. 90, (1993). 23. Garges, S. & Adhya, S. Cell 41, (1985). 24. Aiba, H., Nakamura, T., Mitani, H. & Mori, H. EMBO J. 4, (1985). 25. Reynolds, M.F., et al. Biochemistry 39, (2000). 26. Thorsteinsson, M.V., Kerby, R.L. & Roberts, G.P. Biochemistry 39, (2000). 27. Otwinowski, Z. & Minor, W. Methods Enzymol. 276, (1997). 28. Collaborative Computational Project, Number 4. Acta Crystallogr. D 50, (1994). 29. Fortelle, E. & Bricogne, G. Methods Enzymol. 276, (1997). 30. Brünger, A.T. et al. Acta Crystallogr. D 54, (1998). 880 nature structural biology volume 7 number 10 october 2000

6

7 Copyright of Nature Structural Biology is the property of Nature Publishing Group and its content may not be copied or ed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or articles for individual use.