Untangling Desmosomal Knots with Electron Tomography Wanzhong He, et al. Science 302, 109 (2003); DOI: 10.1126/science.

1086957 The following resources related to this article are available online at www.sciencemag.org (this information is current as of September 27, 2008 ):
Updated information and services, including high-resolution figures, can be found in the online version of this article at: http://www.sciencemag.org/cgi/content/full/302/5642/109 Supporting Online Material can be found at: http://www.sciencemag.org/cgi/content/full/302/5642/109/DC1 This article cites 45 articles, 20 of which can be accessed for free: http://www.sciencemag.org/cgi/content/full/302/5642/109#otherarticles This article has been cited by 84 article(s) on the ISI Web of Science. This article has been cited by 17 articles hosted by HighWire Press; see: http://www.sciencemag.org/cgi/content/full/302/5642/109#otherarticles This article appears in the following subject collections: Biochemistry http://www.sciencemag.org/cgi/collection/biochem Information about obtaining reprints of this article or about obtaining permission to reproduce this article in whole or in part can be found at: http://www.sciencemag.org/about/permissions.dtl Downloaded from www.sciencemag.org on September 27, 2008

Science (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by the American Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. Copyright 2003 by the American Association for the Advancement of Science; all rights reserved. The title Science is a registered trademark of AAAS.

REPORTS
during crystallization processes. This renders network assembly rather sensitive to inhomogeneities of the building blocks, which we experienced with respect to biotin-labeling of the cysteines of bR (18). Both problems may be ameliorated in future developments. Protein networks are well suited to placing functional proteins such as enzymes or membrane channels at defined relative positions, because such functions can be conveniently added by protein fusions (18). This may improve contemporary molecular lithography (25). Moreover, it permits the construction of spatially ordered biochemical teams, which may also include membrane components (26).
1. K. E. Drexler, Trends Biotechnol. 17, 5 (1999). 2. R. D. Piner, J. Zhu, F. Xu, S. Hong, C. A. Mirkin, Science 283, 661 (1999). 3. P. Morales et al., Mater. Sci. Eng. C 2, 173 (1995). 4. G. M. Whitesides, M. Boncheva, Proc. Natl. Acad. Sci. U.S.A. 99, 4769 (2002). 5. U. B. Sleytr et al., Progr. Colloid Polym. Sci. 121, 57 (2002). 6. R. Henderson, P. N. T. Unwin, Nature 257, 28 (1975). 7. J. P. Chalcroft, H. Engelhardt, W. Baumeister, FEBS Lett. 211, 53 (1987). 8. R. F. Service, Science 298, 2322 (2002). 9. J. E. Padilla, C. Colovos, T. O. Yeates, Proc. Natl. Acad. Sci. U.S.A. 98, 2217 (2001). 10. T. O. Yeates, J. E. Padilla, Curr. Opin. Struct. Biol. 12, 464 (2002). 11. N. Dotan, D. Arad, F. Frolow, A. Freeman, Angew. Chem. Int. Ed. 38, 2363 (1999). 12. S. Scheuring, D. J. Muller, P. Ringler, J. B. Heymann, A. Engel, J. Microsc. (Oxford) 193, 28 (1999). 13. I. Reviakine, W. Bergsma-Schutter, A. Brisson, J. Struct. Biol. 121, 356 (1998). 14. M. Kroemer, G. E. Schulz, Acta Crystallogr. D58, 824 (2002). 15. W. A. Hendrickson, et al., Proc. Natl. Acad. Sci. U.S.A. 86, 2190 (1989). 16. P. C. Weber, D. H. Ohlendorf, J. J. Wendoloski, F. R. Salemme, Science 243, 85 (1989). 17. The plasmid is available on request. 18. Experimental details are given as supporting material on Science Online. 19. N. Bischler et al., Biophys. J. 74, 1522 (1998). 20. C. Richard, F. Balavoine, P. Schultz, T. W. Ebbesen, C. Mioskowski, Science 300, 775 (2003). 21. U. Baumann, J. Mol. Biol. 242, 244 (1994). 22. C. Wiesmann, G. Beste, W. Hengstenberg, G. E. Schulz, Structure 3, 961 (1995). 23. J. Michl, T. F. Magnera, Proc. Natl. Acad. Sci. U.S.A. 99, 4788 (2002). 24. N. C. Seeman, Nature 421, 427 (2003). 25. K. Keren et al., Science 297, 72 (2002). 26. P. Ringler, G. E. Schulz, ChemBioChem, 463 (2002). 27. We thank P. Schultz (Institut de Genetique et de Biologie Moleculaire et Cellulaire, Illkirch, France) for a gift of the nickel-chelating lipid. The project was supported by the Deutsche Forschungsgemeinschaft under SFB-428. Supporting Online Material www.sciencemag.org/cgi/content/full/302/5642/106/ DC1 Materials and Methods Figs. S1 to S3 References and Notes 16 June 2003; accepted 20 August 2003

References and Notes

Untangling Desmosomal Knots with Electron Tomography

Wanzhong He,1 Pamela Cowin,2 David L. Stokes1,2,3*
Cell adhesion by adherens junctions and desmosomes relies on interactions between cadherin molecules. However, the molecular interfaces that dene molecular specicity and that mediate adhesion remain controversial. We used electron tomography of plastic sections from neonatal mouse skin to visualize the organization of desmosomes in situ. The resulting three-dimensional maps reveal individual cadherin molecules forming discrete groups and interacting through their tips. Fitting of an x-ray crystal structure for C-cadherin to these maps is consistent with a exible intermolecular interface mediated by an exchange of amino-terminal tryptophans. This exibility suggests a novel mechanism for generating both cis and trans interactions and for propagating these adhesive interactions along the junction. Two major types of intercellular adhesive junctions, desmosomes and adherens junctions, maintain cell shape and tissue integrity. Adherens junctions initiate cell contact and are found as adhesive belts, patches, and puncta between a wide variety of cells (13). Desmosomes reinforce and sustain adhesion and are particularly prevalent in organs subject to shear stress, such as heart and skin (4 ). Both enhance torsional strength by linking to the cytoskeleton. The physiological importance of cellular adhesion is illustrated by genetic and autoimmune diseases associated with various junctional components and by the phenotypes of their genetic knockout in mice. In
1 Skirball Institute of Biomolecular Medicine, 2Department of Cell Biology, New York University School of Medicine, 540 First Avenue, New York, NY 10016, USA. 3New York Structural Biology Center, 89 Convent Avenue, New York, NY 10027, USA.

*To whom correspondence should be addressed. Email: stokes@saturn.med.nyu.edu

particular, defects in desmosomal proteins disrupt heart, skin, and hair (4, 5), whereas those affecting proteins of the adherens junction increase cell proliferation and migration and are thus associated with carcinogenesis and metastasis (6 ). Despite a distinct molecular composition, desmosomes and adherens junctions adopt a similar architectural strategy. Both types of junctions use physical associations between members of the cadherin family of calcium-dependent cell adhesion molecules, whose cytoplasmic domains are tethered to the cytoskeleton via an electrondense plaque of intracellular proteins. Desmosomes contain two types of cadherins, desmoglein and desmocollin, whereas adherens junctions comprise a single type of classical cadherin (7, 8). Desmosomal and classical cadherins belong to the most closely related branches of the cadherin superfamily (9) and are characterized by the presence of five tandem, independently folding extracellular domains

(EC1 to EC5) that each contain the highly conserved calcium-binding motifs Asp-XAsp, Leu-Asp-Arg-Glu, and Asp-X-AsnAsp-Asn (10). The x-ray structures of several classical cadherins (E, N, and C) show that all EC domains adopt a sandwich related to the immunoglobulin fold, with calcium ions bound to the loops joining individual domains (11, 12). Desmosomal cadherins are very likely to have analogous structures given their high sequence similarity (13), including the conserved pattern of Pro, Gly, and hydrophobic residues that define the immunoglobulin fold (14 16). Considerable interest has focused on identifying molecular interfaces that mediate the association of classical cadherins. Mutational studies as well as adhesioninhibitory antibodies and peptides have implicated EC1 domains, in particular Trp2 and the region of the conserved His-AlaVal (HAV) sequence (1722). X-ray crystallography has shown this N-terminal Trp to insert into a pocket formed by hydrophobic residues in the vicinity of the HAV sequence. In the initial structure, a dimer resulted from an intermolecular exchange of the N-terminal residues, which suggested that Trp2 mediates the cis adhesive bond (23). Subsequent structures revealed either a disordered N terminus (12, 24) or the insertion of Trp2 into its own hydrophobic pocket (25), which was suggested to induce a conformational change at a distinct adhesion site. The most recent crystal structure, including all five domains of C-cadherin (26 ), again showed a symmetric Trp2 exchange between neighboring molecules, this time mediating a trans or intercellular contact. A number of alternative crystal contacts have also been proposed as adhesion interfaces, but no consensus has emerged for their physiological relevance.

www.sciencemag.org SCIENCE VOL 302 3 OCTOBER 2003

109

Downloaded from www.sciencemag.org on September 27, 2008

REPORTS
Adding to these disparities, surface force measurements (27 ) and bead aggregation assays (28) have suggested that extensive interdigitation of cadherin molecules, rather than interactions at their tips, may be required for strong adhesion. To determine which of these many models is most physiologically relevant, we sought to visualize the molecular interactions directly within an intact junction. We used electron tomography (29, 30) to produce three-dimensional (3D) reconstructions of desmosomes in newborn mouse epidermis. Because samples of skin are relatively thick, we prepared conventional plastic sections (13) rather than using emerging methods for imaging isolated organelles and thin cultured cells in the unstained, frozen-hydrated state (31). We found that freeze-substitution preserved the smooth and regular appearance of delicate structures, such as cellular membranes and cytoskeletal filaments, and did not produce extraction and aggregation of soluble cytoplasmic components (Fig. 1). The membrane appears as two smooth, parallel black lines that are clearly separated by 3.5 nm of white space; the frequent desmosomes display distinct cross-bridges traversing a regular 28-nm intercellular space, which is interrupted by the electron-dense midline. The intracellular side of the junction comprises a very electron-dense plaque laminating the membrane, which has wispy connections to loose bundles of intermediate filaments. Although the general advantages of freeze-substitution have been described (32), we speculate that the dense molecular scaffold characterizing desmosomal architecture aided preservation by physically resisting molecular collapse and shrinkage (29). Similar resilience may explain the 2.5-nm resolution and minimal shrinkage obtained in plastic sections of myosin S1 crystals (33) and the ability to distinguish myosin conformations in tomographic reconstructions of insect flight muscle (34 ). We found that the best resolution came from the thinnest sections (e.g., R in Fig. 1, E and F, and Table 1), although the organization of cadherins was easier to evaluate in slightly thicker sections (e.g., P used for Fig. 2). Perhaps more important was the use of dual-axis tilt series, which produced more isotropic resolution and lower background noise than singleaxis tilt series, making molecular tracking more reliable (13). To evaluate the 3D organization of cadherins, we examined individual sections from the maps and manually assigned densities to particular protein components (Fig. 1, E and F). On the basis of this segmentation, finger-like densities are revealed within the intercellular space (Fig. 2A), which are 3 nm in diameter and have gently curved shapes that closely resemble the x-ray structure of the EC1-EC5 construct from C-cadherin (26 ). This consistency supports the hypothesis that the structure of extracellular domains of desmosomal cadherins is comparable to that of classical cadherins (14 ). These molecules, which are present at a density of 17,000/ m2, assemble into groups of 10 to 20 with their N-terminal domains forming a series of knots (Fig. 2B) at 20- to 25-nm intervals along the midline. Such grouping contrasts with the molecular packing in x-ray crystals, because interactions at the midline do not form a regular lattice and do not propagate along the length of the junction. Nonetheless, grouping facilitates extensive interactions between EC domains and is consistent with early freeze-fracture images showing clusters of 3-nm filaments connecting at the midline to form 5-nm particles (35). To characterize interactions between EC domains, we chose to focus on one particular desmosome, fitting the x-ray structure to 136 densities representing the majority of molecules present in this junction (Fig. 2A). We identified three distinct geometries for interacting molecules within the midline knots that resemble the shapes of the letters S, W, and (Fig. 2, C to E). The S- and W-shapes result from cadherin pairs interacting in trans across the intercellular space, whereas the -shape adds a third molecule and a cis interaction to the S-shape. After excluding molecules at the edge of the section that are likely to be missing their binding partners, we find that 43% engage in S-shapes, 23%

Fig. 1. Images of desmosomes from neonatal mouse epidermis. (A) Low-magnication image showing an irregular border between keratinocytes coupled by frequent desmosomes. This region of the cell contains many ribosomes but, if the opaque discs are construed as en face views of desmosomes, lacks organelles. (B to D) Higher magnication images reveal the typical lamellar structure of desmosomes. The membrane appears as a narrow white zone; cadherin molecules appear as strands crossing the extracellular space, which is bisected by an electron-dense midline. Individual cadherins are difcult to identify because of extensive superposition of these densely packed molecules within the section; individual molecules are more readily seen in ultrathin sections that are

unsuitable for tomography but are included in (13). A very dense plaque abuts the intracellular face of the membrane and leads to a looser network of brous densities that ultimately connect to bundles of intermediate laments. (E and F) Sections through the tomographic reconstruction of desmosome R (see Table 1) cut parallel (E) and perpendicular (F) to the untilted sample [e.g., (B)]. The membrane is outlined in red, cadherin molecules in blue, two zones of the cytoplasmic plaque in orange and light green, and intermediate laments in dark green. The perpendicular section in (F) reveals the thickness of the plastic section and illustrates that the resolution was quite isotropic [see also (13)]. Scale bars, 500 nm (A), 100 nm [(B) to (D)], 30 nm [(E) and (F)].

110

3 OCTOBER 2003 VOL 302 SCIENCE www.sciencemag.org

Downloaded from www.sciencemag.org on September 27, 2008

REPORTS
engage in W-shapes, and 40% function as the cis-related molecule in -interactions (some molecules have been counted more than once because they engage in multiple interactions, and 9% of molecules lack these particular molecular contacts). Our next step was to fit the x-ray structure of C-cadherin to these shapes, using its characteristic curvature to uniquely define the

Fig. 2. Delineation and tting of cadherin molecules to the desmosome. With the C-cadherin x-ray structure as a template, 136 cadherin molecules were delineated in the region of desmosome P (see Table 1). (A) Densities from the map, with individual cadherin molecules in various colors and the membrane in cyan. (B) A representative group of cadherin molecules clustering at the midline and interacting predominantly at their tips. (C to E) Three recurrent molecular interactions within the molecular groups, referred to as W, S, and , respectively. The x-ray structure for C-cadherin was tted as a rigid body with no changes within the structure itself. (G to I) The resulting juxtaposition of EC1 domains, where each molecule has a distinct ribbon color, calcium ions are brown, the space-lling representation of the Trp2 side chain is dark orange, and the HAV sequence has a light orange ribbon with stick-like side chains.

The W-shape is consistent with symmetric insertion of Trp2 into the hydrophobic pocket formed by the hydrophobic pocket of the neighboring molecule as seen in the C-cadherin x-ray structure (26). In the S-shape, the lower molecule is rotated such that its Trp2 is available for a secondary interaction, although the purple molecules Trp2 is still near the green molecules hydrophobic pocket. In the -shape, the green Trp2 is near the cyan hydrophobic pocket; the cyan Trp2 is near the purple hydrophobic pocket, and the purple Trp2 is free. (F and J) A four-way molecular network, which can be thought of as a -shape plus one additional molecule in cis. Such interactions produce networks of up to six molecules with an apparently stochastic arrangement. These networks appear to be interwoven at the midline to form molecular tangles such as that in (B).

www.sciencemag.org SCIENCE VOL 302 3 OCTOBER 2003

111

Downloaded from www.sciencemag.org on September 27, 2008

REPORTS
rotational orientation of molecules within the density and thus the interface between interacting molecules (Fig. 2, G to J). The Wshape corresponds closely to the dimer seen in the x-ray structure of C-cadherin, in which Trp2 is inserted into the hydrophobic pocket in the EC1 domain of a partner molecule (Fig. 2G) (26). This Trp exchange is plausible for desmosomal cadherins, because both Trp2 and the critical alanine at the base of the hydrophobic acceptor pocket are strictly conserved (24) and, like classical cadherins, peptides corresponding to the HAV region inhibit desmosomal adhesion (36). This W-shape represents a dead-end trans dimer because of the mutual engagement of N termini and hydrophobic pockets, but S- and -shapes provide a mechanism for propagating this interaction. In particular, both S- and -shapes are also compatible with Trp2 exchange, if one takes into account the flexibility of the N terminus implied by its ability to adopt multiple configurations in x-ray structures (12, 2326). In fact, the S-shape is simply related to the W-shape by a rotation of the lower molecule relative to the upper one (Fig. 2, C and D). Although a flexible N terminus would allow one Trp2 to remain in its partners hydrophobic pocket, the other Trp2 would be pulled out of the pocket and become free for interaction with additional cadherins (Fig. 2H). Indeed, the -shape in Fig. 2E consists of an S-shaped, trans interaction with the free Trp2 forming a cis interaction via the hydrophobic pocket of a third molecule (Fig. 2I). Alternatively, Trp2 of the third molecule can engage with the free hydrophobic pocket of the S-shaped pair. Thus, the Trp2 exchange is responsible for both cis and trans interactions and for propagating these interactions among networks comprising four to six molecules (e.g., Fig. 2, F and J) (13). These networks have a wide variety of configurations and therefore appear to result from stochastic addition of molecules in either cis or trans; they also represent a distinct level of organization from the larger midline knots, which generally contain two or more networks interacting through other, undefined interfaces. The binding geometries that we have observed are highly reminiscent of interactions between E-cadherin coupled to pentameric assemblies of cartilage outer matrix protein (ECAD-COMP) (25, 37 ). In these studies, electron micrographs also revealed finger-like EC domains interacting at their distal tips. Cis interactions were postulated for molecules within a given pentameric assembly and are consistent with our -shapes, whereas trans interactions were postulated to interconnect pentamers and are consistent with our S-shapes. These similarities support the idea that desmosomal and classical cadherins use analogous adhesive interactions involving EC1 domains. Although these images fundamentally conflict with interdigitation of cadherin molecules (27, 28), the physical proximity of molecules within these groups suggests that complementary faces of various EC domains may interact to produce a tangle of interwoven molecules at the midline, and that full adhesion may result from the sum of a large number of diverse, weak interactions. Thus, a possible model for assembly begins with a recognition step, involving Trp2 and the hydrophobic pocket of neighboring molecules, followed by a compaction step, in which networks of molecules coalesce into larger adhesive knots to produce full adhesion. Flexibility of cadherin molecules may be an important property for adhesion. In particular, the molecular shapes delineated in our map show structural variations that suggest flexibility in the extracellular domain. A likely source of flexibility is the linker regions between successive EC domains, which represent hinges that determine the angle between domains. These angles are variable in x-ray structures (24, 26 ), and calcium binding by the linkers has been shown to modulate overall flexibility (25, 38). Although glycosylation and nonhomogeneous staining may contribute to the structural variations in our map, desmosomal cadherins may be especially flexible, given that the EC4/5 linker of desmoglein 1 has lost all calcium-binding ligands and other desmoglein and desmocollin isoforms have lost one of three calcium sites in EC3/4 and EC4/5 linkers (16, 39). With regard to adhesion, such flexibility might be useful in maximizing interactions between cadherins forming the junction, just as it is used in building a viral shell (40) and in bundling actin filaments within microvilli (41). Although EC5 domains are far from the site of intermolecular bonds, these domains play an important role in adhesion, as suggested by studies of the effects of dithiothreitol on conserved cysteines in EC5 (20), inhibitory antibodies associated with the EC4/5 loop (17 ), and EC5 truncation in desmoglein 1 (9). Our map indicates that EC5 domains adopt two alternative orientations relative to the membrane: either perpendicular or parallel to the membrane plane (Fig. 2). The former is consistent with previous models of cadherin packing within a junction (23, 25, 26), but the latter swivels the whole extracellular region relative to the membrane surface and implies flexibility in the linkage of EC5 to the transmembrane domain. EC5 domains also tend to associate either in pairs or triplets on the membrane surface (13). Because 80% of paired molecules are incorporated into different groups when followed to the midline, it is possible that the knotting of EC1 domains may form the physical bonds between cells, whereas the intracellular domain may be responsible for holding the groups together laterally. This idea is consistent with the results of studies that investigated catenins as a general determinant of adhesive strength (42), a role for the juxtamembrane portion of the intracellular domain in cadherin dimerization (43 45), and the formation of coiled coils by the leucine-rich transmembrane helices of Ecadherin (46 ). These studies all indicate that tailless cadherins produce only weak adhesion and that their intracellular domains are required for strong adhesion. Whatever the mechanism, our structure suggests that although tailless cadherin constructs would be able to form a small group of molecules with interacting EC1

Table 1. Parameters for tomographic reconstruction of desmosomes. Independent samples were imaged and used for 3D reconstruction. Separate tilt ranges apply to perpendicular tilt axes and were used for dual-axis tomograms. Width of junction is the intercellular distance between extracellular surfaces of apposing plasma membranes. Theoretical resolution is based on the formula given by McEwen and Marko (30). Alignment errors are those reported during tomographic reconstruction by IMOD (47). Sample R P H K G C Tilt range 78 68 76 75 62 61 58 62 62 62 56 73 75 73 75 57 59 53 57 58 60 53 No. of images* 150 145 150 151 118 118 112 120 121 123 110 Magnication 66,000 50,000 38,000 38,000 38,000 38,000 Section thickness (nm) 29.0 47.2 57.7 41.9 56.8 57.7 Width of junction (nm) 26.7 29.1 28.2 28.3 28.9 28.2 Theoretical resolution (nm) 1.96 1.41 3.18 3.29 1.98 2.01 1.29 1.44 1.98 2.04 1.73 Alignment error (nm) 0.785 1.169 0.834 0.697 0.641 0.388

*Generally reect images at 1 tilt angles.

Measured from tomogram.

112

3 OCTOBER 2003 VOL 302 SCIENCE www.sciencemag.org

Downloaded from www.sciencemag.org on September 27, 2008

REPORTS
domains, intracellular domain associations would be required to bring these groups together into an adhesive patch with a sufficient number of intercellular bonds to resist shear forces.
References and Notes
47. J. R. Kremer, D. N. Mastronarde, J. R. McIntosh, J. Struct. Biol. 116, 71 (1996). 48. We thank Y. Ji, M. Auer, and K. MacDonald for help in preparing samples; F. Macaluso and J. Ault for assistance with high-pressure freezing; and K. Taylor, E. Branlund, N. Kisseberth, and D. Mastronarde for providing facilities and support for tomographic data collection and image reconstruction. Coordinates for Fig. 2, G to I, have been deposited in the Protein Data Bank with accession codes 1Q55, 1Q5A, 1Q5B, and 1Q5C. Tomographic reconstructions have been submitted to the electron microscopy database of the European Bioinformatics Institute with accession codes EMD-1051, EMD1052, and EMD-1053. Supported by NIH grant R01 GM47429 (P.C.). Supporting Online Material www.sciencemag.org/cgi/content/full/302/5642/109/ DC1 Materials and Methods SOM Text Figs. S1 to S5 Tables S1 and S2 Movies S1 to S3 References 19 May 2003; accepted 26 August 2003

1. C. L. Adams, W. J. Nelson, S. J. Smith, J. Cell Biol. 135, 1899 (1996). 2. B. Angres, A. Barth, W. J. Nelson, J. Cell Biol. 134, 549 (1996). 3. V. Vasioukhin, C. Bauer, M. Yin, E. Fuchs, Cell 100, 209 (2000). 4. D. R. Garrod, A. J. Merritt, Z. Nie, Curr. Opin. Cell Biol. 14, 537 (2002). 5. S. Hatsell, P. Cowin, Nature Cell Biol. 3, E270 (2001). 6. C. Birchmeier, W. Birchmeier, B. Brand-Saberi, Acta Anat. 156, 217 (1996). 7. B. D. Angst, C. Marcozzi, A. I. Magee, J. Cell Sci. 114, 629 (2001). 8. B. M. Gumbiner, Cell 84, 345 (1996). 9. F. Nollet, P. Kools, F. van Roy, J. Mol. Biol. 299, 551 (2000). 10. M. Ozawa, J. Engel, R. Kemler, Cell 63, 1033 (1990). 11. G. Wagner, Science 267, 342 (1995). 12. B. Nagar, M. Overduin, M. Ikura, J. M. Rini, Nature 380, 360 (1996). 13. See supporting data on Science Online. 14. L. Shapiro, P. D. Kwong, A. M. Fannon, D. R. Colman, W. A. Hendrickson, Proc. Natl. Acad. Sci. U.S.A. 92, 6793 (1995). 15. M. Overduin et al., Science 267, 386 (1995). 16. S. Mechanic, K. Raynor, J. E. Hill, P. Cowin, Proc. Natl. Acad. Sci. U.S.A. 88, 4476 (1991). 17. D. Vestweber, R. Kemler, EMBO J. 4, 3393 (1985). 18. B. Geiger, D. Ginsberg, D. Salomon, T. Volberg, Cell Differ. Dev. 32, 343 (1990). 19. A. Nose, K. Tsuji, M. Takeichi, Cell 61, 147 (1990). 20. M. Ozawa, H. Hoschutzky, K. Herrenknecht, R. Kemler, Mech. Dev. 33, 49 (1990). 21. M. Amagai, S. Karpati, R. Prussick, V. Klaus-Kovtun, J. R. Stanley, J. Clin. Invest. 90, 919 (1992). 22. O. W. Blaschuk, R. Sullivan, S. David, Y. Pouliot, Dev. Biol. 139, 227 (1990). 23. L. Shapiro et al., Nature 374, 327 (1995). 24. K. Tamura, W. S. Shan, W. A. Hendrickson, D. R. Colman, L. Shapiro, Neuron 20, 1153 (1998). 25. O. Pertz et al., EMBO J. 18, 1738 (1999). 26. T. J. Boggon et al., Science 296, 1308 (2002). 27. S. Sivasankar, B. Gumbiner, D. Leckband, Biophys. J. 80, 1758 (2001). 28. S. Chappuis-Flament, E. Wong, L. D. Hicks, C. M. Kay, B. M. Gumbiner, J. Cell Biol. 154, 231 (2001). 29. W. Baumeister, Curr. Opin. Struct. Biol. 12, 679 (2002). 30. B. F. McEwen, M. Marko, J. Histochem. Cytochem. 49, 553 (2001). 31. O. Medalia et al., Science 298, 1209 (2002). 32. K. McDonald, Methods Mol. Biol. 117, 77 (1999). 33. D. A. Winkelmann, T. S. Baker, I. Rayment, J. Cell Biol. 114, 701 (1991). 34. K. A. Taylor et al., Cell 99, 421 (1999). 35. R. LeLoup, L. Laurent, M. F. Ronveaux, P. Droch-Mans, J.-C. Wanson, Biol. Cell 34, 137 (1979). 36. C. Tselepis, M. Chidgey, A. North, D. Garrod, Proc. Natl. Acad. Sci. U.S.A. 95, 8064 (1998). 37. A. Tomschy, C. Fauser, R. Landwehr, J. Engel, EMBO J. 15, 3507 (1996). 38. M. Takeichi, J. Cell. Biol. 75, 464 (1977). 39. J. E. Collins et al., J. Cell Biol. 113, 381 (1991). 40. D. L. D. Caspar, A. Klug, Cold Spring Harbor Symp. Quant. Biol. 27, 1 (1962). 41. D. J. DeRosier, L. G. Tilney, E. Egelman, Nature 287, 291 (1980). 42. M. Ozawa, H. Baribault, R. Kemler, EMBO J. 8, 1711 (1989). 43. M. Ozawa, R. Kemler, J. Cell Biol. 142, 1605 (1998). 44. A. S. Yap, C. M. Niessen, B. M. Gumbiner, J. Cell Biol. 141, 779 (1998). 45. J. Y. Roh, J. R. Stanley, J. Cell Biol. 128, 939 (1995). 46. O. Huber, R. Kemler, D. Langosch, J. Cell Sci. 112, 4415 (1999).

A. M. Clement,1,3,4* M. D. Nguyen,5 E. A. Roberts,2,3 M. L. Garcia,1,3,4 S. Boillee,1,3,4 M. Rule,6 A. P. McMahon,6 W. Doucette,7 D. Siwek,8 R. J. Ferrante,8 R. H. Brown Jr.,7 J.-P. Julien,5 L. S. B. Goldstein,2,3 D. W. Cleveland1,3,4
The most common form of amyotrophic lateral sclerosis (ALS), a neurodegenerative disease affecting adult motor neurons, is caused by dominant mutations in the ubiquitously expressed Cu-Zn superoxide dismutase (SOD1). In chimeric mice that are mixtures of normal and SOD1 mutant expressing cells, toxicity to motor neurons is shown to require damage from mutant SOD1 acting within nonneuronal cells. Normal motor neurons in SOD1 mutant chimeras develop aspects of ALS pathology. Most important, nonneuronal cells that do not express mutant SOD1 delay degeneration and signicantly extend survival of mutant-expressing motor neurons. Amyotrophic lateral sclerosis (ALS) is a progressive neurodegenerative disorder in which motor neurons die beginning in midadult life. About 10% of cases are dominantly inherited; about 20% of these arise from mutations in the gene for Cu-Zn superoxide dismutase (SOD1) (1). Transgenic mice (24) and rats (5, 6) that
1 Ludwig Institute for Cancer Research, 2Howard Hughes Medical Institute and 3Department of Cellular and Molecular Medicine and 4Department of Neurosciences, University of California, 9500 Gilman Drive, La Jolla, CA 92093 0670, USA. 5Centre for Research in Neurosciences, Research Institute of the McGill University Health Care Centre, Montreal General Hospital, 1650 Cedar Avenue, Montreal, Quebec, Canada H3G 1A4. 6 Department of Molecular and Cellular Biology, Harvard University, 16 Divinity Avenue, Cambridge, MA 02138, USA. 7Day Neuromuscular Research Laboratory, Massachusetts General HospitalEast, Charlestown, MA 02139, USA. 8Departments of Neurology, Pathology, and Psychiatry, Boston University School of Medicine, Bedford VA Medical Center, Geriatric Research Education Clinical Center, Bedford, MA 01730, USA.

*Present address: Institute for Physiological Chemistry and Pathobiochemistry, Johannes-Gutenberg-University, Duesbergweg 6, 55099 Mainz, Germany. Present address: Department of Pathology, Harvard Medical School, 200 Longwood Avenue, Boston, MA 02115, USA. Present address: Centre de Recherche de lUniversite Laval (CHUL), Quebec, QC, Canada G1V 4G2. To whom correspondence should be addressed. Email: dcleveland@ucsd.edu

express mutant SOD1 develop a progressive motor neuron disease that shares many features with human ALS; the complete absence of SOD1 in mice does not cause such disease (7). Because toxicity is neither accelerated nor ameliorated by reducing wild-type SOD1 activity (8) and is either unaffected (8) or enhanced (9) by increasing wild-type SOD1 activity, mutant SOD1 must cause disease through acquisition of toxic properties. These may include aberrant oxidative chemistry catalyzed by SOD1-bound copper (1014) or poisoning of a cellular process (or processes) by abundant SOD1 protein aggregates (1517). This triggers a cell death pathway in motor neurons that includes activation of caspase 3 (18, 19). Damage to nonneuronal cells may be involved in toxicity. Before onset of disease in SOD1 mutant mice, there is an inflammatory response, including activation of microglia (2022) and astrocytosis (3, 20); the anti-inflammatory compound minocycline extends survival in mouse models of ALS (2325), although whether this reflects action on microglia, astrocytes, or more directly on motor neurons is not established (24). Although accumulation of mutant SOD1 damages motor neurons in culture (26), SOD1 mutant expression only in neurons (27, 28) or glia (29) has not provoked disease in mice. Thus, fundamental unanswered questions are whether motor neuron

www.sciencemag.org SCIENCE VOL 302 3 OCTOBER 2003

113

Downloaded from www.sciencemag.org on September 27, 2008

Wild-Type Nonneuronal Cells Extend Survival of SOD1 Mutant Motor Neurons in ALS Mice