Amyloid-like filaments and water-filled nanotubes formed by ... - Nature

91 downloads 0 Views 2MB Size Report
Jennifer Stine Elam1, Alexander B Taylor1, Richard Strange2, Svetlana Antonyuk2, ... Lawrence J Hayward4, Joan Selverstone Valentine3, Todd O Yeates3.
© 2003 Nature Publishing Group http://www.nature.com/naturestructuralbiology

ARTICLES

Amyloid-like filaments and water-filled nanotubes formed by SOD1 mutant proteins linked to familial ALS Jennifer Stine Elam1, Alexander B Taylor1, Richard Strange2, Svetlana Antonyuk2, Peter A Doucette3, Jorge A Rodriguez3, S Samar Hasnain2, Lawrence J Hayward4, Joan Selverstone Valentine3, Todd O Yeates3 & P John Hart1 Mutations in the SOD1 gene cause the autosomal dominant, neurodegenerative disorder familial amyotrophic lateral sclerosis (FALS). In spinal cord neurons of human FALS patients and in transgenic mice expressing these mutant proteins, aggregates containing FALS SOD1 are observed. Accumulation of SOD1 aggregates is believed to interfere with axonal transport, protein degradation and anti-apoptotic functions of the neuronal cellular machinery. Here we show that metal-deficient, pathogenic SOD1 mutant proteins crystallize in three different crystal forms, all of which reveal higher-order assemblies of aligned -sheets. Amyloid-like filaments and water-filled nanotubes arise through extensive interactions between loop and -barrel elements of neighboring mutant SOD1 molecules. In all cases, non-native conformational changes permit a gain of interaction between dimers that leads to higher-order arrays. Normal -sheet–containing proteins avoid such self-association by preventing their edge strands from making intermolecular interactions. Loss of this protection through conformational rearrangement in the metaldeficient enzyme could be a toxic property common to mutants of SOD1 linked to FALS.

Amyotrophic lateral sclerosis (ALS), known also as Lou Gehrig’s disease, is a neurodegenerative disorder characterized by the destruction of large motor neurons in the spinal cord and brain. The disease results in progressive paralysis, usually culminating in death within 2–5 years after the onset of symptoms1. Of all ALS cases, ∼10% are familial, and ∼20% of these familial ALS (FALS) cases are associated with dominantly inherited mutations in copper-zinc superoxide dismutase (SOD1)2,3, a 32-kDa homodimeric antioxidant enzyme4. SOD1-linked FALS was initially believed to result from oxidative damage caused by diminished SOD1 activity2,5, but Sod1-null mice develop normally and live to adulthood without developing motor neuron disease6. However, transgenic mice expressing human FALS SOD1 mutant proteins become paralyzed, despite possessing normal7,8 or elevated9,10 levels of SOD1 activity . Together, these observations indicate that pathogenic SOD1 molecules act through the gain of a cytotoxic property and not a loss of enzymatic function. The precise nature of this toxicity is unknown, with explanatory hypotheses ranging from aberrant copper-mediated catalysis11–14 to mutant SOD1 misfolding and aggregation8,15,16. Recent reports suggest that aggregates, composed in part of FALS SOD1, play a role in pathogenesis either by sequestering heat-shock proteins and molecular chaperones17,18 or by interfering with the neuronal axonal transport19,20 and protein degradation21 machineries. However, the molecular basis underlying the formation of these SOD1 aggregates

has remained undefined. Here we show that two members of a larger class of pathogenic SOD1 mutants termed ‘metal-binding region mutants’22 undergo conformational changes that facilitate their polymerization, via extensive non-native protein–protein interactions, into higher order filamentous assemblies. RESULTS Metal-deficient SOD1 and FALS To better understand how mutations in SOD1 can alter the wild-type protein and potentially lead to aggregation, we determined and refined the X-ray crystal structures of the human pathogenic SOD1 mutant proteins S134N, apo H46R and zinc-bound H46R (Zn–H46R) to resolutions of 1.3, 2.5 and 2.15 Å, respectively (Table 1). Both mutations result in defective metal binding22–24, a feature postulated to underlie the toxicity of many pathogenic SOD1 mutants14,23,25–27. The S134N and H46R mutant proteins studied here are all metal deficient to varying degrees when they are isolated from their respective yeast or baculovirus Sf21 expression systems (see Methods), and their low levels of metallation in these studies are likely representative of their states in vivo. The relevance of metal-deficient FALS SOD1 proteins to disease etiology is underscored by recent studies showing that they cause motor neuron disease. For example, transgenic mice overexpressing the FALS SOD1 double mutant H46R/H48Q, a protein that is incapable of binding copper ion, show disease onset and progression

1Department of Biochemistry and the Center for Biomolecular Structure Analysis, The University of Texas, Health Science Center at San Antonio, 7703 Floyd Curl Drive, San Antonio, Texas 78229-3900, USA. 2Molecular Biophysics Group, Department of Synchrotron Radiation, CCLRC Daresbury Laboratory, Warrington, Cheshire, WA44AD, UK. 3Department of Chemistry and Biochemistry, University of California, Los Angeles, California 90095, USA. 4Department of Neurology, University of Massachusetts Medical School, 55 Lake Avenue North, Worcester, Massachusetts 01655, USA. Correspondence should be addressed to P.J.H. ([email protected]).

NATURE STRUCTURAL BIOLOGY VOLUME 10 NUMBER 6 JUNE 2003

461

ARTICLES

© 2003 Nature Publishing Group http://www.nature.com/naturestructuralbiology

a

regions deprotects the edges of β-strands 5 and 6, which together form a cleft located between the two β-sheets of the SOD1 β-barrel in these proteins. This deprotected depression serves as the molecular interface for non-native SOD1 protein–protein contacts. We refer to these new contacts as ‘gain-of-interaction’ (GOI) contacts to distinguish them from the naturally occurring interfaces between subunits of the native SOD1 dimer. When the GOI contacts are combined with the native dimeric interactions, the pathogenic SOD1 dimers assemble into either linear or helical filaments.

S134N and apo H46R linear amyloid-like filaments Amyloid-like filaments of pathogenic SOD1 dimers are linear and occur in two different crystalline environments: one in the S134N orthorhombic crystal form (S134N filament) and one in the apo H46R monoclinic crystal form (apo H46R filament 1). The two filaments are nearly identical, and their backbone atoms align with an r.m.s. deviation of 1.6 Å. The GOI interfaces are symmetric, and the ∼180° symmetry axes they produce are parallel to the symmetry axes of the natural dimers (Fig. 2a). Residues 125–131 of the electrostatic Figure 1 Disorder and conformational changes in pathogenic SOD1 molecules leading to GOI loop adopt an extended, non-native conforinterfaces and filamentous assembly. (a) Stereo view of structures of S134N and apo H46R SOD1 mation and participate in extensive hydrogen monomers (green) superimposed on a monomer of human wild-type SOD1 (gray). The disorder in the bonding and apolar interactions with the metal-deficient S134N and apo H46R proteins is nearly the same, so they are both represented by the green structure. Amino acid residues that are N- and C-terminal to the disordered regions are indicated deprotected surface depression on the β-barrel by red and black dots for the zinc and electrostatic loop elements, respectively. The electrostatic loop of a neighboring SOD1 dimer (Fig. 2b). These (blue) interacts with the exposed cleft (red) of an adjacent molecule in the crystal lattice. Yellow dots GOI interfaces are extensive, burying up to represent acidic residues in the wild-type protein that clash and prevent SOD1–SOD1 interactions from ∼640 Å2 of solvent-accessible surface area per occurring at the cleft (see text). (b) Structure of a Zn–H46R SOD1 monomer (black) superimposed on a polypeptide, which corresponds to approximonomer of human wild-type SOD1 (gray). Amino acid residues are colored as in a. The zinc ion is mately the same amount of surface area buried shown as a green sphere. The zinc loop (blue) interacts with deprotected β-strands 5 and 6 (red) of adjacent molecules in the crystal lattice. The yellow dots are as in a. in the highly stable native SOD1 homodimeric interface (∼660 Å2) (Fig. 2a). A small hydrophobic core is formed around the profiles nearly identical to those of mice expressing other FALS SOD1 pseudo dyad, with Leu42 and Leu126 from each subunit participating proteins that have mutations remote from the metal-binding sites16. (Fig. 2c). The overall arrangement of SOD1 dimeric building blocks is These mice demonstrate pathology in their motor neurons dominated similar to the arrangement recently proposed for transthyretin (TTR) by fibrillar (thioflavin-S–positive) inclusions similar to those observed amyloid fibrils29, in which the β-sheets of the protein molecules lie in mouse models that express other variants of human SOD1 parallel to the long axis of the fiber while their constituent β-strands protein16. Other transgenic mice that lack the copper chaperone for run perpendicular to this axis. An additional parallel to SOD1 and SOD1 (CCS), a protein that inserts copper specifically into nascent FALS is that ∼73 distinct point mutations of TTR have been identified SOD1, also become paralyzed when they express a variety of different that enhance the amyloidogenicity of the protein and cause another FALS SOD1 proteins that are thus forced to be copper deficient. The autosomal dominant neurodegenerative disease, familial amyloid pathology observed in these mice also includes aggregates composed polyneuropathy (FAP)30. in part of SOD1 (ref. 28). Thus, the metal-deficient pathogenic SOD1 The close packing of the β-barrels observed in these linear SOD1 proteins are not only relevant to FALS, they could provide key clues to filaments cannot occur when the proteins possess zinc and electrostatic how the aggregation of pathogenic SOD1 variants might occur loops in their well-ordered, wild-type conformations. Superposition of (see below). fully metallated native SOD1 dimers onto mutant dimers comprising the linear filaments show that steric and electrostatic repulsion from Gain of interaction in pathogenic SOD1 zinc loop residues Asp76 and Glu77 occur at the mutant GOI interface Although the eight-stranded Greek key β-barrel components of these (data not shown). Well-ordered SOD1 electrostatic and zinc loops are metal-deficient, pathogenic S134N and H46R SOD1 structures are thus critical to prevent spontaneous aggregation (data not shown), a preserved relative to the wild-type protein, they exhibit significant dis- mechanism consistent with the various strategies that natural order and conformational changes in their zinc and electrostatic loop β-sheet–containing proteins use to avoid unintended edge-to-edge, elements (Fig. 1). In all three structures, the disorder in these loop β-sheet–to–β-sheet association (reviewed in ref. 31).

b

462

VOLUME 10 NUMBER 6 JUNE 2003 NATURE STRUCTURAL BIOLOGY

ARTICLES

© 2003 Nature Publishing Group http://www.nature.com/naturestructuralbiology

a

c

Figure 2 GOI interfaces in pathogenic SOD1 give rise to cross-β fibrils in two different crystal systems. (a) Orthogonal views of the linear, amyloid-like filaments represented by three dimers shown from top to bottom in green, gold and blue, respectively. Both the S134N and apo H46R linear filaments are represented by the single filament shown in panels i–iv. The GOI interface is red in the filament in i and boxed in ii–iv. In iv, which is rotated 90° relative to ii and iii, β-strands 1, 2, 3 and 6, comprising one-half of each SOD1 β-barrel, are shown in red. The ‘cross-β’ structure observed in amyloid fibrils is shown schematically in v. (b) Stereo view of the GOI interface in the S134N filament. Residues 125–131 of the electrostatic loop from one S134N dimer (orange) interact with a depression in the β-barrel of a neighboring S134N dimer (green) in the crystal lattice. Water molecules are represented as black spheres. The 1.3 Å σA55-weighted electron density, with coefficients 2mFo – DFc, is contoured at 1.0 σ. (c) Small hydrophobic core formed at the GOI interface in the linear, pathogenic SOD1 filaments (see text). The image is an enlargement of the region boxed in image iii of panel a.

b

These crystal structures provide a first glimpse of how pathogenic SOD1 mutant proteins might assemble into amyloid filaments. In the absence of constraints imposed by the crystal repeat, SOD1 filaments may adopt arrangements different in detail from the arrangements reported here. For example, X-ray diffraction from most amyloid fibers suggest that the β-stands tend to run roughly perpendicular to the filament axis and seem to have a spacing of 4.7 Å over a length of many strands. The linear models of SOD1 reported here are consistent with currently accepted ideas of amyloid structure in the broad sense. At a finer level of detail, however, some structural rearrangements between crystalline and noncrystalline forms of SOD1 filaments would be necessary to bring the observed filaments in line with models proposed for other amyloid fibrils on the basis of their X-ray fiber diffraction. The nature of the aggregated mutant SOD1 protein in vivo is complex and unlikely to represent only the ‘clean’ associations we observe in our in vitro studies with purified recombinant pathogenic SOD1 proteins. For example, the SOD1-containing protein aggregates are ubiquitinated in murine models of the disease, and other modifications also occur, such as crosslinking, protein oxidation and the association of heat-shock proteins16,32. Further study of the aggregates formed in vivo is required to address this issue. Apo H46R zigzag filament The second filament in the monoclinic crystal system, apo H46R filament 2, shares some features with the two amyloid-like linear filaments described above, but it also shows some significant differences.

NATURE STRUCTURAL BIOLOGY VOLUME 10 NUMBER 6 JUNE 2003

Here, the rotation axes at the GOI interfaces are nearly parallel to the dimeric symmetry axes, but the rotation angles are 145° rather than 180°. Thus, the H46R filament 2 ‘zigzags’ (Fig. 3) compared with the linear filaments (Fig. 2). This zigzag arrangement occurs because one of the dimers possesses fully ordered electrostatic and zinc loops, thereby preventing the close approach of β-barrels from neighboring molecules observed in the more robust, linear amyloid-like filaments. Although the amino acids participating in the interdimer contacts are the same as in the linear fibrils, they interact in a different orientation, burying only 270 Å2 of solvent-accessible surface area per polypeptide. However, this interaction is not reciprocal across the GOI interface as is observed for the linear filaments. The significance of the zigzag filament to FALS is uncertain because of the tenuous nature of the contacts between SOD1 dimers and also because a nearly identical zigzag filament has recently been observed for the human wild-type apo SOD1 protein33. Together, these structures suggest that an equilibrium exists in solution between the metal-bound loop conformations and disordered conformations of the apo proteins. Because the wild-type enzyme has presumably evolved to avoid nonproductive assembly when it is metal free, this equilibrium may be shifted to favor the disordered conformations in pathogenic SOD1 proteins known to be defective in metal binding. Zn–H46R water-filled nanotubes In contrast to the linear amyloid-like filaments described above for the S134N and apo H46R SOD1 proteins, a distinct but related GOI inter-

463

ARTICLES

© 2003 Nature Publishing Group http://www.nature.com/naturestructuralbiology

α-synuclein and β-amyloid pores exhibit dimensions similar to those observed here for the Zn–H46R nanotubes (external diameter = 95 Å; internal diameter = 30 Å). There is growing support for the hypothesis that the toxic species in neurodegenerative diseases such as Alzheimer’s and Parkinson’s diseases are not the long, insoluble fibrils (which may in fact be cytoprotective) but rather the ‘amyloid oligomers’ or ‘protofibrils’ formed by their soluble precursors36.

Figure 3 Two orthogonal views of apo H46R filament 2, represented by three SOD1 dimers in the same orientation as in Fig. 2a. Although the residues involved in this GOI interface are nearly identical to those of the other linear filaments, they touch each other in a different way, resulting in a substantially less extensive interface and a zigzag arrangement of subunits.

face between Zn–H46R dimers results in the assembly of helical filaments consisting of four Zn–H46R dimers per turn. A helical filament rather than a linear filament arises because the 180° symmetrical GOI is offset 45° from the 180° symmetrical interaction between SOD1 subunits in the homodimer (Fig. 4a). The GOI interface in this tetragonal crystal system also occurs between β-strands 5 and 6 of the SOD1 β-barrel, but the intermolecular contact surface is shifted by several amino acids relative to that of the linear filaments and, in this case, residues of the zinc loop rather than the electrostatic loop participate (Fig. 1b). Zinc loop residues 78–81 adopt a non-native conformation and hydrogen bond to the deprotected edge of β-strand 6, extending the β-sheet by one strand (Fig. 4a). Repetition of this GOI in the Zn–H46R crystal results in the formation of hollow ‘nanotubes’ with an overall diameter of ∼95 Å and an inner water-filled cavity with a diameter of ∼30 Å. The reciprocal addition of β-strands to the β-sheet from the neighboring SOD1 molecules around the GOI dyad gives rise to continuously hydrogen-bonded β-sheets that spiral around the long axis of the nanotube. This architecture represents a variation on a theme recently proposed for polyglutamine amyloid fibrils34, except that in the case of Zn–H46R, the plane of the continuous β-sheets is perpendicular to the long axis of the nanotube rather than parallel, as proposed for the polyglutamine case. The GOI interface in Zn–H46R buries ∼550 Å2 per polypeptide. As with the GOI interfaces in the linear filaments, the extensive contacts are a combination of apolar and hydrogen-bonding interactions (Fig. 4b) that are prevented in the normally folded and metallated wild-type protein. The observation of the hollow, pore-like structure formed by Zn–H46R is particularly intriguing because ‘amyloid pores’ are observed in mutants of α-synuclein and β-amyloid, proteins that cause Parkinson’s and Alzheimer’s disease, respectively35. For α-synuclein, electron microscopy reveals annular α-synuclein oligomeric species with an external diameter of 80–120 Å and an inner diameter of 20–25 Å. For β-amyloid protein, electron microscopy reveals annular oligomeric species with an external diameter of 70–100 Å and an inner diameter of 15–20 Å (ref. 35). Thus, both the

464

DISCUSSION Significance of GOI interfaces Several lines of evidence point to the significance of the GOI interfaces between pathogenic SOD1 molecules. First, they bury approximately the same amount of solvent-accessible surface area as the stable, naturally occurring homodimer interface. Second, the FALS SOD1 filaments are not always generated by crystallographic symmetry operators alone, but instead involve noncrystallographic symmetry operations in the monoclinic and tetragonal crystal forms. This means that the formation of filaments is not simply a necessary consequence of crystallization. Third, the GOI that forms the linear, amyloid-like FALS SOD1 filaments are observed multiple times independently—with two different pathogenic SOD1 mutants in two different crystal systems and in crystallographically distinct environments within and between the asymmetric units of a single crystal form of the protein. Similarly, the GOI forming the helical FALS SOD1 filament appears independently in crystallographically distinct environments within and between the asymmetric units of the tetragonal crystal form. Fourth, each GOI interface arises through subunits of SOD1 that are metal deficient and possess disorder in both the zinc and electrostatic loop elements. Defective metal binding is a wellcharacterized feature common to multiple pathogenic SOD1 mutants14,16,22–27 that could all exhibit similar disorder. Fifth, highmolecular-weight FALS SOD1 aggregates are observed in humans37 and mouse models of the disease8,15,32,38. The pathology of many of these transgenic mice is dominated by fibrillar inclusions of pathogenic SOD1 protein16. Thus, the crystallographic observation of fibrillar, amyloid-like SOD1 filaments is generally consistent with in vivo observations. Sixth, a survey of molecular packing interactions in crystal structures of fully metallated wild-type SOD1 from a variety of organisms, including human, bovine, yeast and spinach, reveals that the structurally intact zinc and electrostatic loop elements protect against the formation of the GOI interfaces and filamentous assemblies reported here (data not shown). Finally, preliminary analyses of two additional metal-deficient pathogenic SOD1 molecules, G85R and E133∆, reveal that they crystallize isomorphously with the orthorhombic S134N crystal form and possess a linear filamentous arrangement of SOD1 dimers similar to that described above (X. Cao, L.J. Whitson and P.J.H., unpublished data). Implications for familial ALS Neuronal damage mediated by toxic, aggregation-prone proteins is a common theme emerging from studies on a broad range of neurodegenerative disorders, including Alzheimer’s disease, Parkinson’s disease, prion disease and polyglutamine disease39. It has become clear that a major effect of mutations in the genes associated with these disorders is the abnormal assembly of misfolded mutant protein into neuronal inclusions, plaques and, more recently, protofibril pores. Our studies suggest that FALS-causing mutant SOD1 may be the latest addition to an ever increasing list39. Finally, we suggest that the two most prominent theories of FALS etiology—aberrant copper chemistry and mutant protein aggregation—are not necessarily mutually exclusive. The ‘metal-binding region’ mutant pathogenic

VOLUME 10 NUMBER 6 JUNE 2003 NATURE STRUCTURAL BIOLOGY

ARTICLES

© 2003 Nature Publishing Group http://www.nature.com/naturestructuralbiology

a

b

c

SOD1 proteins, such as those reported here, are poorly metallated in vivo and seem to be ready for aggregation without modification. The ‘wild type–like’ mutants, such as A4V and G93A, are metallated in vivo and seem to be similar to the wild-type protein in activity and structure. Increased rates of nonspecific peroxidation are observed for several FALS SOD1 proteins of the ‘wild type–like’ class11,12. This activity, exerted on the pathogenic SOD1 proteins themselves, is known to cause damage to the metal-binding ligands, resulting in metal loss39–42, which in turn could contribute to the modes of filamentous assembly suggested by our studies. Thus, the ‘wild type–like’ and ‘metal-binding region’ classes of pathogenic SOD1 mutants may become fused into a single class of molecules with an increased propensity to oligomerize. In summary, the filamentous arrangement of mutant, metal-deficient SOD1 proteins provides a specific and testable hypothesis linking various pathogenic SOD1 mutations to deleterious protein aggregation, thereby making it possible to draw a parallel between ALS and other established amyloid diseases.

NATURE STRUCTURAL BIOLOGY VOLUME 10 NUMBER 6 JUNE 2003

Figure 4 Gain-of-interaction in Zn–H46R SOD1 giving rise to water-filled helical filaments. (a) One-half of the helical Zn–H46R filament, shown in i and ii, is represented by the two dimers shown from top to bottom in green and gold. Image ii is related to the left half of iii by a rotation of 90°. A schematic diagram of the tubular filament is shown in iv. In iii, the approximate location of the crystallographic two-fold axis that runs along the diagonal in the tetragonal unit cell is indicated by a black line without arrows and a 180° rotation symbol. Application of this two-fold operator generates one complete turn of the helical filament. The double-headed black arrow indicates the diameter of the helical filament, and the blue arrow indicates the diameter of the central cavity. The GOI interface between Zn–H46R dimers is boxed. In iii, β-strands 1, 2, 3 and 6, which form one-half of each SOD1 β-barrel, are shown in red. The zinc loop forms a short β-strand (blue) that reciprocally adds to this β-sheet in neighboring Zn–H46R dimers, stabilizing the GOI interface. Zinc ions are shown as purple spheres. (b) Stereo view of the GOI interface in the Zn–H46R helical filament. Residues 78–81 of the zinc loop from one Zn–H46R dimer (orange) interact with an exposed edge of a β-strand in a neighboring Zn–H46R dimer (green) in the crystal and vice versa. Zinc ions are represented as purple spheres. The 2.15 Å σA55-weighted electron density, with coefficients 2mFo – DFc, is contoured at 1.0 σ. Water molecules have been omitted for clarity. (c) Stereo view of two turns of the helical filament generated by repetition of the GOIs in the Zn–H46R structure. This view of the helical filament is rotated 90° around a horizontal axis relative to images iii and vi of panel a. Successive Zn–H46R dimers (green, yellow, blue and red) comprise one turn of the helical filament with a pitch of ∼35 Å.

METHODS Protein expression and crystallization. Recombinant human proteins were produced in Saccharomyces cerevisiae (S134N and Zn–H46R) and baculovirus (apo H46R) expression systems and purified as described22,40. Crystals were grown by the hanging drop vapor diffusion method. S134N at 20 mg ml–1 in 2.25 mM potassium phosphate, pH 7.0, was mixed with an equal volume of reservoir solution containing 2.0 M ammonium sulfate at 25 °C. apo H46R at a concentration of 13.8 mg ml–1 in 100 mM potassium phosphate, pH 7.2, and 50 mM KCl was mixed with an equal volume of reservoir solution containing 0.2 M magnesium acetate, 0.1 M sodium cacodylate, pH 6.5, and 20% (w/v) PEG 8000 at 4 °C. Zn–H46R at a concentration of 15 mg ml–1 in 10 mM MES buffer, pH 6.5, was mixed with an equal volume of reservoir solution containing 0.2 M sodium sulfate decahydrate and 20% (w/v) PEG 3350 at 25 °C. Data collection, structure determination and refinement. All crystals were flash-cooled in liquid nitrogen before X-ray data collection. Diffraction data were taken at beamline X8-C at the NSLS at the Brookhaven National Laboratory in Upton, New York (S134N and apo H46R) and station 14.2 at the SRS, Daresbury Laboratory (Zn–H46R), and were processed with the

465

ARTICLES Table 1 Crystallographic data and refinement statistics S134N

apo H46R

Zn–H46R

Data collection Space group

P212121

P21

P412121

a (Å)

40.1

79.6

190.8

b (Å)

56.5

70.0

190.8

c (Å)

105.4

113.8

34.6

© 2003 Nature Publishing Group http://www.nature.com/naturestructuralbiology

Unit cell

β (°)

110.4

X-ray data λ (Å)

1.100

0.989

0.978

Total

270,095

167,752

652,196

Unique

Number of observations 58,507

40,758

35,432

Resolution range (Å)a

50–1.3 (1.38–1.30)

100–2.5 (2.56–2.50)

20–2.15 (2.21–2.15)

Completeness (%)a

97.9 (87.1)

99.8 (100.0)

99.3 (98.6)

Rsym (on I) (%)a,b

5.0 (49.8)

8.6 (40.8)

9.0 (48.0)

Refinement Number of dimers

1

4

2

Resolution range (Å)

21.1–1.3

42.4–2.5

20–2.15

Rcryst (%)c

18.8

22.2

20.6

Rfree (%)d

21.1

27.0

23.6

Raniso (%)e

17.7

Rfree_aniso (%)e

19.3

F/σF

>0

>0

>0

Bonds (Å)

0.010

0.011

0.016

Angles (°)

1.5

1.7

1.7

per asymmetric unit

R.m.s. deviations

Number of atoms Protein

2,005

7,966

3,657

Water

247

235

277

Metal ions

3

0

4

Sulfate anions

2

0

4

number in parentheses is for the last shell. bRsym = Σ|I – |/ΣI, where I is the observed intensity and is the average intensity of multiple symmetry-related observations of that reflection. cRcryst = Σ||Fop| – |Fcp|| / Σ|Fop|. dRfree = Σ||Fop| – |Fcp|| / Σ|Fop| where |Fop| is from a test set not used in the structural refinement. eRaniso and Rfree_aniso are as Rcryst and Rfree, except with anisotropic atomic displacement parameter refinement. aThe

DENZO/SCALEPACK suite41. The structure of SOD1 mutant G37R (PDB entry 1AZV)42 was used as the search model for molecular replacement in AMoRe43 (S134N) and EPMR44 (apo H46R), whereas the structure of native human SOD1 (ref. 33) was used as the search model for molecular replacement in MOLREP45 (Zn–H46R). The structures were iteratively refined using REFMAC5 (ref. 46) (S134N and Zn–H46R) and CNS47 (apo H46R). The models were manually adjusted in O48. For S134N, anisotropic atomic displacement thermal parameters were refined late in the process49. For Zn–H46R, TLS refinement was implemented in the end stages. The final models were evaluated using WHATCHECK50 and PROCHECK51. No residues of the S134N and H46R structures fall outside of the allowable regions of their respective Ramachandran plots. Metal deficiency. All three pathogenic SOD1 proteins are metal deficient, as indicated by inductively coupled plasma mass spectroscopy, when purified from their respective expression systems22,24. The S134N subunits are designated as A–B, where the (–) symbol represents the naturally occurring homodimer interface. X-ray data collected at the copper and zinc absorption edges coupled with anomalous difference Fourier calculations reveal that in S134N

466

subunit A, there is a mixture of copper and zinc ions in the Cu-binding site and zinc alone is bound in the Zn-binding site. S134N subunit B contains almost exclusively zinc in the Cu-binding site and no metal in the Zn-binding site. The apo H46R subunits are designated as G–H, I–J, K–L and M–N in the four dimers, and all are completely devoid of both copper and zinc ions in their binding sites. The apo H46R filaments 1 and 2 are packed in layers that alternate and are approximately normal to the ac plane of the monoclinic crystal. The Zn–H46R subunits are designated as W–X and Y–Z, respectively, and all have zinc bound in their Zn-binding sites but no metal in their Cu-binding sites. Structure analysis and figure preparation. Solventaccessible surface areas were calculated in CNS47 using a probe radius of 1.4 Å. Structural alignments were performed using ALIGN52. Figures were prepared using MolScript53, BobScript54 and POV-Ray (http://www.povray.org). Coordinates. Coordinates and structure factors have been deposited in the Protein Data Bank (accession codes 1OZU, 1OZT and 1OEZ for S134N, apo H46R and Zn–H46R SOD1 structures, respectively). ACKNOWLEDGMENTS We thank L. Flaks, and J. Berendzen for support at beamline X8-C at the NSLS, Brookhaven National Laboratory; D. Cascio and M. Hough for their interest and valuable discussions; S. Holloway for assistance with the illustrations and our colleagues who have offered comments during the preparation of this manuscript. This work was supported by the National Institutes of Health (L.J.H., J.S.V. and P.J.H.), the Robert A. Welch Foundation (P.J.H.), the ALS Association (L.J.H., J.S.V and P.J.H.), the MND Association (S.S.H.) and a predoctoral fellowship from the Association for the Advancement of Aging Research (J.S.E.). Funding from CCLRC and resources at Daresbury are also gratefully acknowledged. COMPETING INTERESTS STATEMENT The authors declare that they have no competing financial interests. Received 6 December 2002; accepted 22 April 2003 Published online 19 May 2003; doi:10.1038/nsb935

1. Haverkamp, L.J., Appel, V. & Appel, S.H. Natural history of amyotrophic lateral sclerosis in a database population. Validation of a scoring system and a model for survival prediction. Brain 118, 707–719 (1995). 2. Deng, H.X. et al. Amyotrophic lateral sclerosis and structural defects in Cu,Zn superoxide dismutase. Science 261, 1047–1051 (1993). 3. Rosen, D.R. et al. Mutations in Cu/Zn superoxide dismutase gene are associated with familial amyotrophic lateral sclerosis. Nature 362, 59–62 (1993). 4. Fridovich, I. Superoxide dismutases. An adaptation to a paramagnetic gas. J. Biol. Chem. 264, 7761–7764 (1989). 5. Bowling, A.C., Schulz, J.B., Brown, R.H. Jr. & Beal, M.F. Superoxide dismutase activity, oxidative damage, and mitochondrial energy metabolism in familial and sporadic amyotrophic lateral sclerosis. J. Neurochem. 61, 2322–2325 (1993). 6. Reaume, A.G. et al. Motor neurons in Cu/Zn superoxide dismutase-deficient mice develop normally but exhibit enhanced cell death after axonal injury. Nat. Genet. 13, 43–47 (1996). 7. Ripps, M.E., Huntley, G.W., Hof, P.R., Morrison, J.H. & Gordon, J.W. Transgenic mice expressing an altered murine superoxide dismutase gene provide an animal model of amyotrophic lateral sclerosis. Proc. Natl. Acad. Sci. USA 92, 689–693 (1995). 8. Bruijn, L.I. et al. ALS-linked SOD1 mutant G85R mediates damage to astrocytes and promotes rapidly progressive disease with SOD1-containing inclusions. Neuron 18, 327–338 (1997). 9. Gurney, M.E. et al. Motor neuron degeneration in mice that express a human Cu,Zn superoxide dismutase mutation. Science 264, 1772–1775 (1994).

VOLUME 10 NUMBER 6 JUNE 2003 NATURE STRUCTURAL BIOLOGY

© 2003 Nature Publishing Group http://www.nature.com/naturestructuralbiology

ARTICLES 10. Wong, P.C. et al. An adverse property of a familial ALS-linked SOD1 mutation causes motor neuron disease characterized by vacuolar degeneration of mitochondria. Neuron 14, 1105–1116 (1995). 11. Wiedau-Pazos, M. et al. Altered reactivity of superoxide dismutase in familial amyotrophic lateral sclerosis. Science 271, 515–518 (1996). 12. Yim, M.B. et al. A gain-of-function of an amyotrophic lateral sclerosis-associated Cu,Zn-superoxide dismutase mutant: an enhancement of free radical formation due to a decrease in Km for hydrogen peroxide. Proc. Natl. Acad. Sci. USA 93, 5709–5714 (1996). 13. Beckman, J.S., Chen, J., Crow, J.P. & Ye, Y.Z. Reactions of nitric oxide, superoxide and peroxynitrite with superoxide dismutase in neurodegeneration. Prog. Brain Res. 103, 371–380 (1994). 14. Estevez, A.G. et al. Induction of nitric oxide-dependent apoptosis in motor neurons by zinc-deficient superoxide dismutase. Science 286, 2498–2500 (1999). 15. Bruijn, L.I. et al. Aggregation and motor neuron toxicity of an ALS-linked SOD1 mutant independent from wild-type SOD1. Science 281, 1851–1854 (1998). 16. Wang, J. et al. Fibrillar inclusions and motor neuron degeneration in transgenic mice expressing superoxide dismutase 1 with a disrupted copper-binding site. Neurobiol. Dis. 10, 128–138 (2002). 17. Bruening, W. et al. Up-regulation of protein chaperones preserves viability of cells expressing toxic Cu/Zn-superoxide dismutase mutants associated with amyotrophic lateral sclerosis. J. Neurochem. 72, 693–699 (1999). 18. Okado-Matsumoto, A. & Fridovich, I. Amyotrophic lateral sclerosis: a proposed mechanism. Proc. Natl. Acad. Sci. USA 99, 9010–9014 (2002). 19. Borchelt, D.R. et al. Axonal transport of mutant superoxide dismutase 1 and focal axonal abnormalities in the proximal axons of transgenic mice. Neurobiol. Dis. 5, 27–35 (1998). 20. Williamson, T.L. & Cleveland, D.W. Slowing of axonal transport is a very early event in the toxicity of ALS-linked SOD1 mutants to motor neurons. Nat. Neurosci. 2, 50–56 (1999). 21. Johnston, J.A., Dalton, M.J., Gurney, M.E. & Kopito, R.R. Formation of high molecular weight complexes of mutant Cu, Zn-superoxide dismutase in a mouse model for familial amyotrophic lateral sclerosis. Proc. Natl. Acad. Sci. USA 97, 12571–12576 (2000). 22. Hayward, L.J. et al. Decreased metallation and activity in subsets of mutant superoxide dismutases associated with familial amyotrophic lateral sclerosis. J. Biol. Chem. 277, 15923–15931 (2002). 23. Liu, H. et al. Copper2+ binding to the surface residue cysteine 111 of His46Arg human copper-zinc superoxide dismutase, a familial amyotrophic lateral sclerosis mutant. Biochemistry 39, 8125–8132 (2000). 24. Rodriguez, J.A. et al. Familial ALS-associated mutations decrease the thermal stability of distinctly metallated species of human copper-zinc superoxide dismutase. J. Biol. Chem. 277, 15932–15937 (2002). 25. Lyons, T.J. et al. Mutations in copper-zinc superoxide dismutase that cause amyotrophic lateral sclerosis alter the zinc binding site and the redox behavior of the protein. Proc. Natl. Acad. Sci. USA 93, 12240–12244 (1996). 26. Crow, J.P., Sampson, J.B., Zhuang, Y., Thompson, J.A. & Beckman, J.S. Decreased zinc affinity of amyotrophic lateral sclerosis-associated superoxide dismutase mutants leads to enhanced catalysis of tyrosine nitration by peroxynitrite. J. Neurochem. 69, 1936–1944 (1997). 27. Lyons, T.J. et al. The metal binding properties of the zinc site of yeast copperzinc superoxide dismutase: implications for amyotrophic lateral sclerosis. J. Biol. Inorg. Chem. 5, 189–203 (2000). 28. Subramaniam, J.R. et al. Mutant SOD1 causes motor neuron disease independent of copper chaperone-mediated copper loading. Nat. Neurosci. 5, 301–307 (2002). 29. Serag, A.A., Altenbach, C., Gingery, M., Hubbell, W.L. & Yeates, T.O. Arrangement of subunits and ordering of β-strands in an amyloid sheet. Nat. Struct. Biol. 9, 734–739 (2002). 30. Connors, L.H., Richardson, A.M., Theberge, R. & Costello, C.E. Tabulation of transthyretin (TTR) variants as of 1/1/2000. Amyloid 7, 54–69 (2000). 31. Richardson, J.S. & Richardson, D.C. Natural β-sheet proteins use negative

NATURE STRUCTURAL BIOLOGY VOLUME 10 NUMBER 6 JUNE 2003

design to avoid edge-to-edge aggregation. Proc. Natl. Acad. Sci. USA 99, 2754–2759 (2002). 32. Watanabe, M. et al. Histological evidence of protein aggregation in mutant SOD1 transgenic mice and in amyotrophic lateral sclerosis neural tissues. Neurobiol. Dis. 8, 933–941 (2001). 33. Strange, R.W. et al. The structure of metal deficient wild type human Cu,Zn superoxide dismutase and its relevance to familial amyotrophic lateral sclerosis. J. Mol. Biol. 328, 877–891 (2003). 34. Perutz, M.F., Finch, J.T., Berriman, J. & Lesk, A. Amyloid fibers are water-filled nanotubes. Proc. Natl. Acad. Sci. USA 99, 5591–5595 (2002). 35. Lashuel, H.A., Hartley, D., Petre, B.M., Walz, T. & Lansbury, P.T. Jr. Neurodegenerative disease: amyloid pores from pathogenic mutations. Nature 418, 291 (2002). 36. Sherman, M.Y. & Goldberg, A.L. Cellular defenses against unfolded proteins: a cell biologist thinks about neurodegenerative diseases. Neuron 29, 15–32 (2001). 37. Shibata, N. et al. Intense superoxide dismutase-1 immunoreactivity in intracytoplasmic hyaline inclusions of familial amyotrophic lateral sclerosis with posterior column involvement. J. Neuropathol. Exp. Neurol. 55, 481–490 (1996). 38. Wang, J., Xu, G. & Borchelt, D.R. High molecular weight complexes of mutant superoxide dismutase 1: age-dependent and tissue-specific accumulation. Neurobiol. Dis. 9, 139–148 (2002). 39. Taylor, J.P., Hardy, J. & Fischbeck, K.H. Toxic proteins in neurodegenerative disease. Science 296, 1991–1995 (2002). 40. Goto, J.J. et al. Loss of in vitro metal ion binding specificity in mutant copperzinc superoxide dismutases associated with familial amyotrophic lateral sclerosis. J. Biol. Chem. 275, 1007–1014 (2000). 41. Otwinowski, Z. & Minor, W. Processing of X-ray diffraction data collected in oscillation mode. Methods Enzymol. 276, 306–326 (1997). 42. Hart, P.J. et al. Subunit asymmetry in the three-dimensional structure of a human CuZnSOD mutant found in familial amyotrophic lateral sclerosis. Protein Sci. 7, 545–555 (1998). 43. Navaza, J. & Saludjian, P. AMoRe: an automated molecular replacement program package. Methods Enzymol. 276, 581–594 (1997). 44. Kissinger, C.R., Gehlhaar, D.K. & Fogel, D.B. Rapid automated molecular replacement by evolutionary search. Acta Crystallogr. D 55, 484–491 (1999). 45. Vagin, A.A. & Teplyakov, A. MOLREP: an automated program for molecular replacement. J. Appl. Crystallogr. 30, 1022–1025 (1997). 46. Murshudov, G.N., Vagin, A.A., Lebedev, A., Wilson, K.S. & Dodson, E.J. Efficient anisotropic refinement of macromolecular structures using FFT. Acta Crystallogr. D 55, 247–255 (1999). 47. Brunger, A.T. et al. Crystallography & NMR system: a new software suite for macromolecular structure determination. Acta Crystallogr. D 54, 905–921 (1998). 48. Jones, T.A., Zou, J.Y., Cowan, S.W. & Kjeldgaard, M. Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Crystallogr. A 47, 110–119 (1991). 49. Winn, M.D., Isupov, M.N. & Murshudov, G.N. Use of TLS parameters to model anisotropic displacements in macromolecular refinement. Acta Crystallogr. D 57, 122–133 (2001). 50. Hooft, R.W., Vriend, G., Sander, C. & Abola, E.E. Errors in protein structures. Nature 381, 272 (1996). 51. Laskowski, R.A., McArthur, M.W., Moss, D.S. & Thornton, J.M. PROCHECK: a program to check the stereochemical quality of protein structures. J. Appl. Crystallogr. 26, 283–291 (1993). 52. Cohen, G.E. ALIGN: a program to superimpose protein coordinates, accounting for insertions and deletions. J. Appl. Crystallogr. 30, 1160–1161 (1997). 53. Kraulis, P.J. MOLSCRIPT: a program to produce both detailed and schematic plots of protein structures. J. Appl. Crystallogr. 24, 946–950 (1991). 54. Esnouf, R.M. Further additions to MolScript version 1.4, including reading and contouring of electron-density maps. Acta Crystallogr. D 55, 938–940 (1999). 55. Read, R.J. Improved Fourier coefficients for maps using phases from partial structures with errors. Acta Crystallogr. A 42, 140–149 (1986).

467