letters to nature

................................................................. Eukaryotic type II chaperonin CCT interacts with actin through specific subunits Oscar Llorca*, Elizabeth A. McCormack†, Gillian Hynes†, Julie Grantham†, Jacqueline Cordell†, Jose´ L. Carrascosa*, Keith R. Willison†, Jose´ J. Fernandez‡ & Jose´ M. Valpuesta* * Centro Nacional de Biotecnologia, C.S.I.C., Campus Universidad Auto´noma de Madrid, 28049 Madrid, Spain † CRC Centre for Cell and Molecular Biology, Institute of Cancer Research, Chester Beatty Laboratories, 237 Fulham Road, Chelsea, London SW3 6JB, UK ‡ Departamento de Arquitectura de Computadores y Electro´nica, Universidad de Almerı´a, 04120 Almeria, Spain .................................. ......................... ......................... ......................... ......................... ........

Chaperonins assist the folding of other proteins1. Type II chaperonins, such as chaperonin containing TCP–1(CCT), are found in archaea and in the eukaryotic cytosol2. They are hexadecameric or nonadecameric oligomers composed of one to eight different polypeptides. Whereas type I chaperonins like GroEL are promiscuous, assisting in the folding of many other proteins1, only a small number of proteins, mainly actin and tubulin, have been described as natural substrates of CCT. This specificity may be related to the divergence of the eight CCT subunits3. Here we have obtained a three-dimensional reconstruction of the complex between CCT and a-actin by cryo-electron microscopy and image processing. This shows that a-actin interacts with the apical domains of either of two CCT subunits. Immunolabelling of CCT–substrate complexes with antibodies against two specific CCT subunits showed that actin binds to CCT using two specific and distinct interactions: the small domain of actin binds to CCTd and the large domain to CCTb or CCTe (both in position 1,4 with respect to d). These results indicate that the binding of actin to CCT is both subunit-specific and geometry-dependent. Thus, the substrate recognition mechanism of eukaryotic CCT may differ from that of prokaryotic GroEL. Chaperonins share an overall structure of two oligomeric rings, placed back-to-back. The atomic structure of the type I chaperonin GroEL from Escherichia coli4 and the type II chaperonin thermosome from Thermoplasma acidophilum5 have a similar organization in their three domains (apical, intermediate and equatorial) within the monomer. Type I chaperonins act with a cochaperonin, a small heptamer that seals the chaperonin cavity, whereas type II chaperonins seem to have a ‘built-in’ cochaperonin in the form of a helical protrusion5,6. Whereas GroEL uses a mechanism based on nonspecific hydrophobic interactions to interact with its substrate1, the recognition of chemically denatured actin by CCT requires the generation of folding intermediates which form slowly after removal of the denaturant. This contrasts with the rapid binding of actin to GroEL7, indicating that actin binds to CCT in a conformation different from the one that is recognized by GroEL. Furthermore, GroEL cannot productively fold actin7. Actin appears to interact with CCT through defined regions8. Therefore, we have determined the structure of actin trapped on CCT to investigate the molecular recognition mechanism of type II chaperonins. Native a-actin was chemically denatured and incubated with CCT in the absence of nucleotide, and aliquots of the solution were vitrified and observed at low temperature (−170 8C) under an electron microscope. Images were taken at 208 tilt and digitized, and 3,546 top views were extracted and processed (see Methods and Supplementary Information). From the several classification procedures performed, two main groups of particles were obtained: those with a mass in the cavity (2,480 particles) and those with an empty cavity (766 particles). Both sets of particles were used for NATURE | VOL 402 | 9 DECEMBER 1999 | www.nature.com

independent three-dimensional reconstructions. The reconstruction generated from the CCT particles with empty cavities (Fig. 1a) shows a structure similar to that of nucleotide-free CCT9. The reconstruction generated with the substrate-bound particles reveals a structure with some differences (Fig. 1b–d): whereas one of the rings (the bottom one in Fig. 1b, c) is similar to either of the rings of the substrate-free CCT9, the other reveals the presence of a rodshaped mass attached to the inner wall of the apical domains of two CCT subunits placed in a 1,4 arrangement (Fig. 1c, d). The atomic structure of actin has a ‘v’ shape formed by two arms, the large and small domains10, each of which is divided into two subdomains. The structure of the CCT-bound a-actin shown here can be seen as the tips of the two open arms of a-actin interacting, respectively, with each of two CCT subunits. In addition, the ˚ fit dimensions of the a-actin reconstructed here (,90 A˚ 3 30 A) well with the dimensions of an opened-up conformation of the ˚ so we surmise that aatomic structure of actin10 (,100 A˚ 3 35 A), actin could be substantially native-like before it interacts with CCT. To corroborate this model of interaction, a chimaeric protein was constructed with subdomain 4 of the large domain of b-actin linked to the carboxy terminus of the Ha-Ras protein, Ha-Ras–b-actin subdomain 4 (called b-actin.sub4 from now on). This hybrid protein interacts with CCT (Fig. 2Ab and d) in reticulocyte lysate in equilibrium between a free, unbound population and a CCTbound species; in contrast, b-actin is actively folded and processed to a non-CCT-interacting conformation in reticulocyte lysate (Fig. 2Aa and c). Recombinant b-actin.sub4 was chemically denatured and incubated with CCT. Images of the vitrified specimen were taken and digitized, and 3,361 208-tilted top views of CCT were processed. Again, we obtained two sets of CCT particles after classification: 1,220 substrate-free particles and 2,030 particles with a mass inside the cavity. Two independent three-dimensional reconstructions were carried out. The one performed with the substrate-free particles gave rise to a similar structure to that shown in Fig. 1a (results not shown), whereas that generated using the substrate-bound particles revealed a mass bound to just one of the CCT subunits (Fig. 3), probably mediated through the actin subdomain of the chimaeric protein, as Ha-Ras does not interact with CCT (Fig. 2Ba and b). An interesting conformational change takes place in the apical domains of the substrate-bound rings of both the CCT–a-actin and

Figure 1 Three-dimensional reconstructions of CCT and the CCT–a-actin complex. a, Side view of the reconstruction obtained from substrate-free top views. b–d, Views of the three-dimensional reconstruction obtained from top views containing a-actin. b, Side view; c, a cut along the longitudinal axis of the CCT–a-actin complex; d, top view of the CCT–a-actin complex showing the actin-containing ring.

© 1999 Macmillan Magazines Ltd

693

letters to nature CCT–b-actin.sub4 three-dimensional reconstructions. The upwards and outwards movements of the apical domains induced upon substrate binding, either to one (for b-actin.sub4) or two CCT subunits (for a-actin), are probably important in the mechanism of protein folding by CCT and may also allow signalling between the

two CCTrings so that, once unfolded substrate binds to either of the rings, binding of a second polypeptide is prevented. CCT particles with substrate bound to both rings have not been observed despite the high (10:1) substrate:CCT molar ratio used. The three-dimensional reconstruction of CCT–a-actin indicates

Figure 2 Interaction of b-actin and CCT. A, Analysis of the full-length b-actin (a, c) and b-actin.sub4 fusion protein (b, d) during an in vitro translation in rabbit retuculocyte lysate over 7–16-min. a and b show native polyacrylamide gel electrophoresis (PAGE) analysis of 35S-methionine-labelled proteins (full-length actin contains 17 methionine residues and b-actin.sub4 only six). c, d, Quantitative analysis of the gel data shown in a and b, respectively, expressed as a percentage of the total 35 S-labelled b-actin (c) or b-actin.sub4 (d) bound (open symbols) or not bound (closed symbols) to CCT. B, Native PAGE analysis of in vitro translation at 60-min time points of Ha-Ras (a) and Ha-Ras + sub.4 (b-actin.sub4) (b) showing absence of interaction between Ha-Ras and CCT. C, Antibodies used for immunolocalization. The anti-CCTd 8g monoclonal antibody used for electron microscope decoration is an IgG2b of high affinity which recognizes only a single polypeptide in western blots of total rabbit reticulocyte

lysate proteins. It does not recognize any other CCT subunit in western blot experiments using purified CCT (data not shown). D, Epitope localization. Solid-phase 15-mer peptide scanning arrays mapped the anti-CCTd 8g epitope to between residues P344–S358 of CCTd. The corresponding region is identifiably homologous in the thermosome5 and is located on the outer surface of the structure (shown here as a ribbon diagram of the a-thermosome subunit; PDB accession number 1A6D). The anti-CCTa 91A monoclonal antibody binds to the epitope A465–A469 (ref. 15), which is located in a clearly homologous region of the outer surface of the equatorial domain of the thermosome. The location of the yeast actin gene anc2-1 mutation in the CCT4p/CCTd subunit (residue G345D sequenced in this study (see Supplementary Information); Saccharomyces cerevisiae anc2 strain was a gift from D. Drubin) is indicated as a space-filled residue in the equivalent glycine of the a-thermosome subunit.

694

© 1999 Macmillan Magazines Ltd

NATURE | VOL 402 | 9 DECEMBER 1999 | www.nature.com

letters to nature that the tips of the arms of a-actin may interact with CCT through specific regions on the inner wall of the apical domains (apparently well below the helical protrusions5,6,9) of two CCT subunits placed in a 1,4 arrangement. To determine whether this interaction is subunit-specific or only geometry-dependent, we used two monoclonal antibodies, the first against the apical domain of the CCTd subunit (8g; Fig. 2C, D) and the second against the intermediate domain of the CCTa subunit (91A; Fig. 2D). The new anti-CCTd monoclonal antibody was produced because a mutation in the yeast CCT4p/CCTd subunit, anc2, rescues several mutations in subdomain 2 of yeast actin ACT1 (ref. 11 and Supplementary Information), indicating that this subunit may be involved in actin folding (Fig. 2D). Aliquots of the immunocomplexes were negatively stained (to contrast only one of the CCT rings) and images were recorded and digitized. We processed 970 untilted top views of the CCT–a-actin–8g complexes and obtained two main populations with substrate present in the CCT cavity (average images represented in Fig. 4a and b). In both average images, one

of the domains of actin interacts with CCTd (the subunit to which the antibody binds) and the other with a subunit that is in position 1,4 with respect to CCTd, either anticlockwise (Fig. 4a) or clockwise (Fig. 4b). According to the published model of CCT subunit arrangement12 (Fig. 4j), the two CCT subunits involved in the interaction with actin would be the CCTb and CCTe subunits. The processing of 1,069 untilted top views of CCT–b-actin.sub4– 8g complexes again revealed two main populations. Their average images (Fig. 4c and d) show that subdomain 4 of b-actin binds to either of the two subunits placed in a 1,4 arrangement (clockwise or anticlockwise) to CCTd (CCTb and CCTe, respectively). The results obtained with the 91A antibody reinforce those from the 8g antibody. Two populations were obtained from the processing of 2,054 untilted top views of CCT–a-actin–91A complexes (Fig. 4e and f). Antibody 91A binds to the CCTa subunit; the two average images generated show that one of the actin domains binds in both cases to a subunit that is in an anticlockwise 1,3 arrangement with respect to CCTa (CCTd, according to ref. 12), and the other to

Figure 3 Three-dimensional reconstruction of the CCT–b-actin.sub4 complex. a, A cut along the longitudinal axis. b, Top view showing the b-actin.sub4-containing ring. The conditions are as described in Fig. 1.

Substrate

Actin

Actin sub.4

a

b

c

d

e

f

g

h

Antibody

8g (CCTd)

91A (CCTa)

i

j sub.2

sub.4

sub.1

Large domain

α

L γ

δ

ε

θ

ζ γ

β

Small domain

Figure 4 Two-dimensional average images of immunocomplexes of substrate-bound CCT. a, b, Two-dimensional average images of two classes of CCT–a-actin–8g complex. c, d, Two-dimensional average images of CCT–b-actin.sub4–8g complexes. e, f, Two-dimensional average images of CCT–a-actin–91A complexes. g, h, Twodimensional average images of CCT–b-actin.sub4–91A complexes. No 8-fold symmetry NATURE | VOL 402 | 9 DECEMBER 1999 | www.nature.com

ε

S

θ η

sub.3

α

η δ

ζ β

Actin or η δ

S

α L

θ

ε ζ

γ

β

has been applied to any of these averages. i, Atomic structure of actin10 showing the large and small domains and their division into four subdomains (subs 1–4). j, The two possible modes of interaction of actin with CCT, using the CCT subunit orientation published in ref. 12, which is fully consistent with the results obtained here. L and S, large and small domains of actin, respectively10.

© 1999 Macmillan Magazines Ltd

695

letters to nature another CCT subunit that is in either a clockwise 1,2 or 1,4 position with respect to CCTa (Fig. 4e and f; CCTe and CCTb subunits, respectively). The same two positions, CCTb and CCTe, are occupied when two populations are generated out of 851 untilted top views of processed CCT–b-actin.sub4–91A complexes (Fig. 4g and h, respectively). The apparent paradox of obtaining a single three-dimensional reconstruction of the CCT–a-actin complex despite there being two possible ways for actin to interact with CCT results from the low resolution of this reconstruction, in which the two arms of actin and all of the CCT subunits are indistinguishable. Thus, the two possible three-dimensional reconstructions (with actin interacting with the CCTd and CCTb subunits and with actin interacting with the CCTd and CCTe subunits) are the same if the first is rotated 1358 with respect to the second (Fig. 1b–d). Our results indicate that actin interacts with CCT by subunitspecific and geometry-dependent mechanism: the small domain of actin binds to the CCTd subunit and the large domain can bind to either CCTbor CCTe (see model in Fig. 4j). This type of interaction contrasts with the non-specific folding mechanism mediated by type I chaperonins like GroEL. Subjects for future study include the possible involvement of other CCT subunits in the interaction with tubulin3 and other substrates13, the conformational changes undergone by both substrate and CCT upon ATP binding and the hydrolysis needed to complete the CCT folding cycle. M

Methods Antibodies The anti-CCTd 8g monoclonal antibody was prepared by immunizing rats with an apical domain fragment of the CCTd subunit produced using recombinant E. coli (D219-N394 (6-His+23C epitope tagged at the C terminus) using a standard hybridoma protocol14. The anti-CCTa 91A monoclonal antibody was prepared as described14,15.

Preparation of complexes a-Actin (from rabbit muscle; Sigma) denatured in 7 M guanidinium chloride was incubated in a diluting buffer (100-fold) containing 0.4 mM murine CCT purified as described12. The Ha-Ras–b-actin fusion protein (b-actin.sub4) was constructed by linking residues 1–168 of human Ha-Ras to residues L178–F262 of human b-actin (subdomain 4). We prepared complexes of CCT–a-actin and CCT–b-actin.sub4 by incubation, for 15 min at room temperature, of CCT and either of the two substrates at a 1:10 molar ratio. In the case of the CCT–substrate–antibody complexes, preformed CCT– a-actin and CCT–b-actin.sub4 complexes were incubated with antibodies 8g or 91A (5:1 antibody: complex molar ratio) for 15 min at room temperature.

Multisubunit machinery assisting in protein folding and assembly in the eukaryotic cytosol. Eur. J. Biochem. 230, 3–16 (1995). 3. Willison, K. R. in Molecular Chaperones and Folding Catalysts (ed. Bukau, B.) 555–571 (Harwood Academic, Amsterdam, 1999). 4. Braig, K. et al. The crystal structure of the bacterial chaperonin GroEL at 2.8 A˚. Nature 371, 578–586 (1994). 5. Ditzel, L. et al. Crystal structure of the thermosome, the archaeal chaperonin and homologue of CCT. Cell 93, 125–138 (1998). 6. Klumpp, M., Baumeister, W. & Essen, L. O. Structure of the substrate binding domain of the thermosome, an archaeal group II chaperonin. Cell 91, 263–270 (1997). 7. Melki, R. & Cowan, N. J. Facilitated folding of actins and tubulins occurs via a nucleotide-dependent interaction between cytoplasmic chaperonin and distinctive folding intermediates. Mol. Cell. Biol. 14, 2895–2904 (1994). 8. Rommelaere, H., De Neve, M., Melki, R., Vandekerckhove, J. & Ampe, C. The cytosolic class II chaperonin CCT recognizes delineated hydrophobic sequences in its target proteins. Biochemistry 38, 3246–3257 (1999). 9. Llorca, O. et al. 3D reconstruction of the ATP-bound form of CCT reveals the asymmetric folding conformation of a type II chaperonin. Nature Struct. Biol. 6, 639–642 (1999). 10. Kabsch, W., Mannherz, H. G., Suck, D., Pai, E. F. & Holmes, K. C. Atomic structure of the actin:DNase I complex. Nature 347, 37–44 (1990). 11. Vinh, D. B.-N. & Drubin, D. G. A yeast TCP-1 like protein is required for actin function in vivo. Proc. Natl Acad. Sci. USA 91, 9116–9120 (1994). 12. Liou, A. K. F. & Willison, K. R. Elucidation of the subunit orientation in CCT (chaperonin containing TCP1) from the subunit composition of micro-complexes. EMBO J. 16, 4311–4316 (1997). 13. Kashuba, E., Pokrovskaja, K., Klein, G. & Szekely, L. Epstein-Barr virus-encoded nuclear protein EBNA-3 interacts with the e-subunit of the t-complex protein 1 chaperonin complex. J. Hum. Virol. 2, 33–37 (1999). 14. Willison, K. R. et al. The t complex polyptide 1 (TCP-1) is associated with the cytoplasmic aspect of Golgi membranes. Cell 57, 621–632 (1989). 15. Hynes, G., Kubota, H. & Willison, K. R. Antibody characterization of two distinct conformations of the chaperonin containing TCP-1 from mouse testis. FEBS Lett. 358, 129–132 (1995). 16. Llorca, O. et al. ATP binding induces large conformational changes in the apical and equatorial domains of the eukaryotic chaperonin containing TCP-1 complex. J. Biol. Chem. 273, 10091–10094 (1998). 17. Marabini, R. & Carazo, J. M. Pattern recognition and classification of images of biological macromolecules using artificial neural networks. Biophys. J. 66, 1804–1814 (1994). 18. Frank, J. et al. SPIDER and WEB: processing and visualization of images in 3D electron microscopy and related fields. J. Struct. Biol. 116, 190–199 (1996). 19. Marabini, R., Herman, G. T. & Carazo, J. M. 3D reconstruction in electron microscopy using ARTwith smooth spherically symmetric volume elements (blobs). Ultramicroscopy 72, 53–65 (1998).

Supplementary information is available on Nature’s World-Wide Web site (http:// www.nature.com) or as hard copy from the London editorial office of Nature.

Acknowledgements This work was partially supported by grants from the DGICYT (J.M.V. and J.L.C.). O.L. is a fellow the Comunidad Auto´noma de Madrid. The UK laboratory is supported by the Cancer Research Campaign. Correspondence and requests for materials should be addressed to J.M.V. (e-mail: [email protected]).

Electron microscopy We applied aliquots of the immunocomplexes to carbon grids, negatively stained them with 2% uranyl acetate and recorded them at 08 tilt. Samples of the CCT–substrate solutions were applied to the grids for 1 min, blotted for 5 s and frozen quickly in liquid ethane at −180 8C. Images were recorded at 208 tilt in a JEOL 1200EX-II electron microscope equipped with a Gatan cold stage operated at 120 kV and recorded on Kodak SO-163 film at a 60,000× nominal magnification and ,2 mm underfocus.

Image processing and three-dimensional reconstruction In both negative stained and frozen-hydrated studies, top views were selected and twodimensional processing was carried out as described9. In the case of the immunocomplexes, the average images obtained were filtered at the resolution obtained16 (22 A˚). We used a self-organizing map algorithm17 to classify the particles into two main groups, those with no mass in the cavity of the particle and those with density in the cavity. With the two sets of particles, which have a large angular distribution, two independent threedimensional reconstructions were performed using angular refinement algorithms provided by SPIDER18. The apo-CCT volume was used as the initial model for refinement, and the volumes were generated using ART with blobs (Algebraic Reconstruction Techniques19). All of the two-dimensional processing carried out so far has revealed that the CCT particles have eightfold pseudosymmetry along the longitudinal axis9,16. However, the particles containing substrate do not have eightfold symmetry in the cavity. For this reason, no symmetrization was applied to the inner cylinder comprising the CCT cavity. The final resolution was calculated by Fourier ring correlation of two independent reconstructions and the value obtained was used to low-pass filter the volumes (30 A˚ for the CCT–a-actin complex and 35 A˚ for the CCT–b-actin.sub4 complex).

................................................................. Modified reaction centres oxidize tyrosine in reactions that mirror photosystem II L. Ka´lma´n*‡, R. LoBrutto†, J. P. Allen* & J. C. Williams* Center for the Study of Early Events in Photosynthesis and * Department of Chemistry and Biochemistry, and † Department of Plant Biology, Arizona State University, Tempe, Arizona 85287, USA .......................................... ......................... ......................... ......................... .........................

The participation of tyrosine in the oxidation of water by photosystem II, a multisubunit enzyme complex involved in plant photosynthesis, exemplifies the significant role amino-acid side chains play in oxidation/reduction reactions in proteins1. The influence of the surrounding protein on the properties of aminoacid radicals and the attributes necessary for electron transfer are

Received 15 July; accepted 29 September 1999. 1. Bukau, B. & Horwich, A. L. The hsp70 and hsp60 chaperone machines. Cell 92, 351–366 (1998). 2. Kubota, H., Hynes, G. & Willison, K. R. The chaperonin containing t-complex polypeptide 1 (TCP-1).

696

‡ Permanent address: Institute of Biophysics, Jo´zsef Attila University, Szeged H-6722 Hungary.

© 1999 Macmillan Magazines Ltd

NATURE | VOL 402 | 9 DECEMBER 1999 | www.nature.com