Man bites dog - Nature

1 downloads 0 Views 724KB Size Report
Concern about the use of anthrax as a bioweapon has increased over the last decade and has now become a reality, which makes the clarification of its toxic.
news and views

Man bites dog

© 2002 Nature Publishing Group http://structbio.nature.com

William E. Meador and Florante A. Quiocho The structure of the anthrax edema factor (EF) exotoxin reveals evidence of a new adenylate kinase mechanism and the first structure of an active calmodulin-dependent enzyme in complex with calmodulin (CaM). The CaM-free toxin opens to enfold CaM in a way that positions critical EF substrate-binding residues near the catalytic site.

Concern about the use of anthrax as a a b bioweapon has increased over the last decade and has now become a reality, which makes the clarification of its toxic action of paramount importance. The understanding of the pathology of the anthrax agent Bacillus anthracis has recently received a big boost with the publication of the structure of one of two major exotoxins, LF (lethal factor)1, as well as the cloning of the human receptor for the toxin cofactor PA (protective antigen)2. During the course of anthrax infection the bacteria release EF, LF and PA, which combine after PA has been bound by the PA receptor and modified by a protease. The PA–toxin complex Fig. 1 X-ray structures of the edema factor (EF) toxin. a, Free EF and b, EF complexed with CaM then enters the cell via the PA receptor and 3’deoxy-ATP. The EF helical domain (yellow) is moved by the binding of calmodulin (red), and and the toxins are released into the held between it and the catalytic domains CA and CB (green). Large conformational changes occur cytosol. Now Drum et al.3 have deter- in the interface domains switch A (blue), switch B (orange) and switch C (cyan), which complete an mined the structure of EF, a CaM-depen- active site in which 3′ deoxy-ATP and a metal are bound (purple). Figure courtesy of A. Bohm. dent adenylate cyclase, in both the free state and the EF–CaM complex with CaM gets clamped bound 3′ deoxy-ATP. complexes to date, and no calcium ions In the structure of ternary complex of EF, are present in the N-terminal domain Ready for action CaM and 3′ deoxy-ATP, dramatic confor- (N-domain). CaM is nearly surrounded The structure of free EF consists mational changes are seen to have occurred by the helical domain and the CA domain of the C-terminus (residues 291–800, (Fig 1b)3. Essentially, CaM is found of the EF molecule, burying 5,900 Å2 of Mr 58 kDa), which binds calmodulin and clamped between the CA domain and the solvent-accessible surface area. Contacts displays the adenylate cyclase activity of the helical domain in an extensive contact, between CaM and the EF helical domain intact toxin3. It comprises three domains, which is quite different from the view of apparently serve to stabilize the Ca2+-free CA (residues 294–349 and 490–622), CB CaM biting onto and moving an auto- form of the CaM N-domain (Fig. 1b). (350–489), and the ‘helical domain’ inhibitory domain during the activation of This Ca2+-free conformation has a good (660–800). In this structure the helical a kinase4. This interaction causes substan- overlap (root mean square deviation = domain folds against the CA domain in a tial movement of switch A, orders switch B, 1.2 Å) with a Ca2+–calmodulin NMR hydrophobic contact which buries and causes part of switch C to become a structure previously reported5. The CaM 3,600 Å2 of surface, with the CA domain helix. The net result of these movements is C-domain seen in the EF–CaM– and the CB domain conjoining to form the that switch A and switch B and the other 3′ deoxy-ATP structure is in a more stancatalytic site (Fig. 1a). In addition to the part of switch C are positioned near the dard conformation, a Ca2+-bound form catalytic site, three important regions are active site to complete the substrate bind- which is similar to all Ca2+-bound CaM identified, termed ‘switch A’ (502–551), ing pocket, with switch B providing a num- C-domain structures seen to date (the ‘switch B’ (578–591), and ‘switch C’ ber of significant contacts to the CaM–K-channel complex structure does (residues 630–659 from the loop linking CA 3′ deoxy-ATP, while a residue from switch not have calcium in the CaM C-domain, and the helical domain). Switch A and C stabilizes switch B, and switch A provides and is different)6 (Fig 2). In addition, the switch C form most of the interface with a hydrogen bond to the adenosine ring. CaM C-domain binds a helix of switch A the helical domain, while switch B is a dis- These changes create an active site channel in EF in a manner similar to that seen in ordered loop. Moreover, as the substrate open to solvent at both ends. other CaM–effector structures. However, analog 3′ deoxy ATP is not present in The calmodulin found in this complex in isolation this helix binds with very low CaM-free structure, the active site is open is in an unusual extended conformation affinity (8.8 µM), orders of magnitude to solvent. that has not been seen in CaM–effector lower than the nanomolar binding 156

nature structural biology • volume 9 number 3 • march 2002

news and views

© 2002 Nature Publishing Group http://structbio.nature.com

a

b

c

d

Fig. 2 Comparison of CaM–EF with other calmodulin structures aligned on a CaM C-domain helix. a, Ca2+–calmodulin, with the N-domain (orange) connected to the C-domain (red) by a rigid linker helix. b, Ca2+–CaM enclosing a target domain (blue) from myosin light chain kinase and c, a fragment of the Ca2+-activated K+ channel. Note that part of the linker helix of CaM has become unwound to allow close binding of these quite different targets. d, The CaM–EF structure. Calmodulin is held between EF domains, again with a partially unwound linker, which allows this unusual domain positioning. Also, note that the CaM–EF structure does not have Ca2+ in the CaM N-domain and the CaM–K+ channel structure does not have Ca2+ in the CaM C-domain. Figure courtesy of W. Tang.

constants found for the target helix of kinases7. Perhaps the most intriguing question of the many raised by this unusual structure is how the calmodulin got to where it is. Somehow, the toxin has managed to capture a CaM molecule, presumably in an extended conformation, which then finds its way in between two large domains that have extensive contacts with each other prior to the capture. The authors discount the possibility that thermal motion exposes this cleft — because of these extensive contacts — and propose instead that calmodulin initially binds to a large positively charged area in the region of the helical domain, in particular Lys 525 of switch A. This then is thought to initiate a process that exposes residues of switch C, which are later bound by calmodulin. Certainly, since the interface in the CaM-free state is a mainly hydrophobic contact, some hydrophobic interactions could be expected during the course of this recognition process. In the CaM–EF complex, the hydrophobic pocket of the N-domain is not exposed because of the lack of calcium, and the pocket of the C-domain contacts an amphiphilic helix of switch C in the usual manner seen in kinase-derived CaM– peptide structures, except that the hydrophobic residues are said to be “out of register”. Whatever the real scenario is, this process will certainly be a fascinating and fruitful area for present and future studies of calmodulin. This structure does demonstrate two of the known functions in calmodulin’s bag of tricks — binding of a helix by a rigid calciumloaded C-domain and expansion of the linker helix to allow large changes of domain spacing and orientation when necessary to accommodate itself to a

target8,9. Of the new features revealed by this structure, the particulars of the binding of a Ca2+-free N-domain and the details of polar contacts made by the C-domain are of special interest. Too much signal Also of considerable interest is the mechanism proposed for the adenylate kinase activity of EF, which is 1,000-fold higher than those of mammalian adenylate cyclases and disrupts cell signaling by producing an excess of cyclic AMP3. The structure of the EF–CaM complexed with 3′ deoxy-ATP (Fig. 1b), a noncyclizable analog of the ATP substrate, reveals atomic details of substrate binding and a new catalytic mechanism. The analog is held in place by both polar and nonpolar interactions. The triphosphate oxygens are mainly engaged in charge-coupling interactions with two Lys and one Arg residues. Moreover, Yb3+, a crystallization additive and a surrogate of the Mg 2+ metal cofactor, is coordinated by the oxygens of the α- and β-phosphates and the carboxylates of two Asp residues, which are highly conserved among adenylate cyclases3. The adenosine ring makes polar interactions with main chain carbonyl oxygens, and the ribose ring is flanked on one side by a Phe and on the other side by a Leu. The interaction between aromatic or aliphatic residues and sugar ring faces has been previously found to be a nearly universal feature and to play a dominant role in the binding of sugars, especially pyranosides, to almost 50 proteins with known structures10. The proposed catalytic mechanism of the cyclization reaction by EF has two key elements: (i) the divalent metal and one of the Lys residues polarize the α-phosphate and stabilize the developing charge on the oxygen bridging the α- and β-phosphates;

nature structural biology • volume 9 number 3 • march 2002

and (ii) a His residue, which lies on the opposite side of the ribose from the bound metal, is in excellent orientation to act as a general base, abstracting a proton from the ribose 3′ hydroxyl for an in-line attack on the α-phosphate. This mechanism differs from the proposed twometal-ion mechanism of the mammalian adenylate cyclase11, whose structure shows essentially no similarity with EF and no conserved His in the active site. In the mammalian adenylate kinase mechanism, one of the two bound metals, whose location in the active site is not too far from that of the Yb3+ in the EF, is proposed to play a role in the activation of the 3′ hydroxyl attacking group. Interestingly, based on the observations that the ribose in the structure of the mammalian adenylate cyclase–ATP-αS complex11 is in a similar orientation to that of the ATP analog bound in the EF and that a water (or a hydroxide) molecule held in place by hydrogen bonds with the side chains of a Ser and a conserved Arg is positioned close to the ribose 3′ hydroxyl, Drum et al.3 propose a different mechanism for mammalian adenylate cyclases which is akin to that for EF. The mechanism assumes that a hydroxide anion, stabilized presumably by the Ser hydroxyl dipole and the positive charge of the Arg, functions as a general base in a manner similar to the active site His residue. This unusual and exciting set of structures of EF will be very useful not only for those designing anthrax antitoxins but also for those seeking to solidify the catalytic mechanism for this highly active adenylate cyclase. In addition, this toxin offers the opportunity to expand our knowledge of calmodulin activation of enzymes, since it employs a process different from the current kinase model. 157

news and views

© 2002 Nature Publishing Group http://structbio.nature.com

William E. Meador and Florante A. Quiocho are in the Howard Hughes Medical Institute and Department of Biochemistry and Molecular Biology, Baylor College of Medicine, Houston, Texas 77030, USA. Correspondence should be addressed to F.A.Q. email: [email protected]

1. Pannifer, A.D. et al. Nature 414, 229–233. (2001). 2. Bradley, K.A., Mogridge, J., Mourez, M., Collier, R.J. & Young, J.A.T. Nature 414, 225–229. (2001). 3. Drum, C.L. et al. Nature 415, 396–402 (2002). 4. Krueger, J.K. et al. Biochemistry 36, 6017–6023 (1997). 5. Zhang, M., Tanaka, T. & Ikura, M. Nature Struct. Biol. 2, 758–767. (1995). 6. Schumacher, M.A., Rivard, A.F., Bachinger, H.P. & Adelman J.P. Nature 410, 1120–1124. (2001).

7. O’Neal, K.T. & DeGrado, W.F. Trends Biochem. Sci. 15, 59–64 (1990). 8. Meador, W.E., Means, A.R., & Quiocho, F.A. Science 262, 1718–1721 (1993). 9. Meador, W.E., George, S.E., Means, A.R., & Quiocho, F.A. Nature Struct. Biol. 2, 943–945 (1995). 10. Duan, X., & Quiocho, F.A. Biochemistry 41, 706–712 (2002). 11. Tesmer, J.J. et al. Science 285, 756–760. (1999).

How a rotavirus hijacks the human protein synthesis machinery Gabriele Varani and Frédéric H.-T. Allain The NSP3 protein from rotaviruses recognizes a unique sequence at the 3′ end of the rotaviral mRNA. By doing so, it promotes translation of viral proteins while repressing host protein synthesis. The structure of the NSP3 protein bound to a viral 3′ end sequence reveals how this occurs and suggests how it might be possible to design a new class of antiviral drugs.

Rotaviruses are a class of double-stranded RNA viruses responsible for pediatric diarrhea, which causes the death of approximately one million children worldwide each year1. The viral genome is composed of 11 segments of doublestranded RNA that encode six structural proteins forming the viral capsids (named VP for viral proteins) and six nonstructural proteins (NSPs)2. The 36 kDa nonstructural protein 3 (NSP3) is essential for promoting the synthesis of viral proteins. Its N-terminal domain binds the 3′ end of viral mRNA and its C-terminal domain interacts with translation initiation factor eIF4G. These interactions are functionally equivalent to the interactions between human poly(A) binding protein (PABP), and both the poly(A) tail at the 3′ end of eukaryotic mRNAs and eIF4G. Through these interactions, the rotavirus achieves circularization of its mRNA and selectively boosts the efficiency with which the host translational machinery synthesizes viral proteins3 (Fig. 1). In a recent issue of Cell, Burley and colleagues4 report the high resolution structure of the N-terminal domain of an NSP3 protein (from group A simian agent rotavirus) bound to a consensus sequence found at the 3′ end of the viral mRNA. The structure reveals how the virus hijacks the human translation machinery to enhance synthesis of viral proteins and simultaneously shut down production of host proteins. It also reveals a new and surprising structure for RNA recognition and 158

a

b

Fig. 1 The 5′ and 3′ ends of viral and eukaryotic mRNA synergistically stimulate initiation of protein synthesis. a, In host mRNAs, poly(A) binding protein (PABP) binds the 3′ poly(A) tail and eIF4E binds the 5′ cap; together, they interact with eIF4G, which delivers the mRNA to the ribosome. b, Rotaviral mRNAs have no poly(A) tail, but the viral NSP3 protein binds the consensus 3′ sequence 5′-GACC-3′, which also interacts with eIF4G, similar to the mechanism of action of human PABP. The atomic resolution structure of the eIF4E–eIF4G peptide–7methyl GDP ternary complex22 (1EJF), of the PABP–A8 complex21 (1CVJ) and of the recently determined NSP3 dimer– 5′-UGACC-3′ complex4 (PDB accession code 1KNZ) were used to prepare this schematic figure. Unfortunately, we do not yet know exactly how the pieces are assembled to form the translation apparatus. Figs 1 and 2 were generated using MOLMOL25.

explains the sequence-specific recognition of the 3′ end of the viral mRNA by NSP3. The deep burial of RNA bases within a tunnel formed by the interaction of two NSP3 monomers suggests a possible means for designing new antivirals that selectively target NSP3 and inhibit its activity.

RNA genome. The innermost capsid consists of two structural proteins VP1 and VP3. VP1 is the RNA-dependent RNA polymerase (RdRP) and VP3 is the mRNA capping enzyme. These proteins allow viral RNA replication and 5′ capping to occur within the virus, but synthesis of viral proteins is dependent upon the host Translating rotaviral mRNA translational apparatus. The rotaviral Rotaviruses are icosahedral viruses with mRNAs are capped at their 5′ ends but three capsids enclosing a double-stranded their 3′ ends are not polyadenylated; nature structural biology • volume 9 number 3 • march 2002