DOI: 10.1126/science.1218723, 929 (2012);335 Science

J. Murray GibsonTwo Pairs Beat One−−Solving Amorphous Structures

This copy is for your personal, non-commercial use only.

clicking here.colleagues, clients, or customers by , you can order high-quality copies for yourIf you wish to distribute this article to others

  here.following the guidelines

can be obtained byPermission to republish or repurpose articles or portions of articles

  ): March 3, 2012 www.sciencemag.org (this infomation is current as of

The following resources related to this article are available online at

http://www.sciencemag.org/content/335/6071/929.full.htmlversion of this article at:

including high-resolution figures, can be found in the onlineUpdated information and services,

http://www.sciencemag.org/content/335/6071/929.full.html#relatedfound at:

can berelated to this article A list of selected additional articles on the Science Web sites

http://www.sciencemag.org/content/335/6071/929.full.html#ref-list-1, 1 of which can be accessed free:cites 11 articlesThis article

http://www.sciencemag.org/cgi/collection/mat_sciMaterials Science

subject collections:This article appears in the following

registered trademark of AAAS. is aScience2012 by the American Association for the Advancement of Science; all rights reserved. The title

CopyrightAmerican Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by theScience

on

Mar

ch 3

, 201

2w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

www.sciencemag.org SCIENCE VOL 335 24 FEBRUARY 2012 929

PERsPEctiVEs

the specialized motifs required for binding to nerve terminals or cleaving SNARE proteins.

Gu et al. subjected NTNHA-A, BoNT/Ai and M-PTC to conditions resembling those relevant to the absorption of botuli-num neuro toxin in the gastrointestinal tract (7). Only M-PTC was stable in pepsin at pH 2.6 and in trypsin at pH 6. Neutral or alka-line pH promoted dissociation of the com-plex, resulting in loss of protection against trypsin inactivation. The essential residues involved in the interaction between BoNT/Ai and NTNHA-A were identified by site-directed mutagenesis. The results suggest that small molecule inhibitors could be developed to weaken this interaction and inactivate the neurotoxin early in intoxica-tion, when intervention would be most effec-tive. However, because there are no signs of botulism until SNARE protein cleavage is in progress (2), the inhibitors would need to be taken before exposure, requiring prior knowledge of an attack.

Examination of the crystal structure of the M-PTC complex helps to understand how NTNHA-A protects BoNT/Ai. The domain of botulinum neurotoxin most susceptible to proteolysis is the HC (8), and all three domains of NHTHA-A interact extensively with this domain, thereby protecting it against prote-olysis (see the figure, panel B). In contrast, the catalytically active LC domain of botu-linum neurotoxin does not appear to interact with NTNHA-A, suggesting that it is inher-ently resistant to proteolytic degradation in the M-PTC complex. These results suggest that it may be possible to orally deliver pro-tein-based therapeutics—both for botulism and for other conditions where a protein- or peptide-based drug is indicated—by cou-pling them to modified M-PTCs to protect the cargo from degradation.

The seminal findings of Gu et al. raise important questions for future studies. Is the domain homology of NTNHA unique to sero-type A, or is it the general pattern for all sero-

types? What are the roles of the other neuro-toxin-associated proteins, especially HA33, which resists proteolysis and enhances trans-cytosis of neurotoxin (9)? Finally, given that the 900 kD complex provides the best protec-tion against degradation of botulinum neu-rotoxin (5), how is this complex assembled, and how does each component contribute to keeping the toxin structurally intact and able to invade and inactivate cholinergic nerve cells in the host organism?

References 1. J.Sobel,Clin. Infect. Dis.41,1167(2005). 2. L.L.Simpson,Annu. Rev. Pharmacol. Toxicol.44,167

(2004). 3. S.Gu et al.,Science335,977(2012). 4. www.bt.cdc.gov/agent/agentlist-category.asp 5. K.-H.Eisele et al.,Toxicon57,555(2011). 6. L.W.Cheng et al.,Toxicology249,123(2008). 7. A.B.Maksymowych et al.,Infect. Immun.67,4708

(1999). 8. F.Chen et al.,Infect. Immun.65,1626(1997). 9. S.K.Sharma,B.R.Singh,J. Nat. Toxins7,239(1998).

10.1126/science.1219602

Diffraction data can be used to deter-mine atomic structures by creating and refining a structural model. If

the calculated diffraction pattern is in suffi-ciently good statistical agreement with the data, we trust that a unique structure has been found. For single crystals, the data are well-defined diffraction spots that represent reflections off the repeating lattice of atoms. For amorphous materials, the data are trans-formed into a radial distribution function (RDF) that provides the distribution of inter-atomic distances, typically over the range from 1 to 10 Å. In such cases, structural models can be tested against the RDF by ana-lyzing the distribution of distances between random pairs of atoms. It is tempting to treat a model that has good agreement with the RDF as a unique structure, but on page 950 of this issue, Treacy and Borisenko (1) show that we cannot always rely on the RDF to fin-gerprint the correct medium-range structure (5 to 30 Å) of a material that is disordered on the nanoscale. They show that more than one

structure, including some that resemble crys-tals in their topology, can equally well fit the RDF of amorphous silicon (a-Si).

The limitation of the RDF arises because it depends only on two-atom (“pair”) corre-lations. New experimental data that revealed higher-order atom pair-pair correlations, used by Treacy and Borisenko to constrain modeling, points against the random net-work (see the figure) as a good model for most a-Si samples. Although the lack of

uniqueness of the RDF beyond very short-range structure (beyond 5 Å) has been rec-ognized [for example, see (2)], the RDF has still been repeatedly used to make structural inferences because it had been generally assumed that no competing nonrandom net-work model could fit the data equally well.

Because it has a pure, tetrahedrally bonded network of just one type of atom, a-Si is an excellent model of a disordered material. Interest accelerated when it was demonstrated in the 1960s that a-Si and amorphous germanium could be made semiconducting by preparation in the pres-ence of hydrogen, followed by demonstra-tion of doping by Spear and LeComber (3) and Anderson’s work on localization (4). Applications of a-Si now include photovol-taics and thin-film-transistor displays.

The structure of a-Si was assumed to be

Solving Amorphous Structures—Two Pairs Beat One

materials science

J. Murray Gibson

Consideration of atomic ordering beyond just pairs of atoms shows that amorphous silicon is better modeled as paracrystalline material than as a disordered network.

PhysicsDepartment,NortheasternUniversity,Boston,MA02115,USA.E-mail:m.gibson@neu.edu

Testing amorphous structure models.OneoftheatomicmodelsusedbyTreacyandBorisenkotofitthe radial distribution function and fluctuationmicroscopydata.Highlightedareregionsillustrativeofthecontinuousrandomnetwork(inblue)andtheparacrystallinecubicstructure(red).

on

Mar

ch 3

, 201

2w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

24 FEBRUARY 2012 VOL 335 SCIENCE www.sciencemag.org 930

PERsPEctiVEs

Mendelian genetic disorders, rare clinical phenotypes arising from a single-gene mutation, are

extremely diverse traits that affect every organ system, age group, and human popu-lation (1). Their cumulative incidence is rare (under 5%) because the clinical phenotypes are deleterious and affected individuals rarely reproduce. They persist in the population by de novo mutation in the past few generations, but some recessive mutations are an excep-tion because their effects can be sheltered in carriers for hundreds of generations. Identi-fying the genes and mutations for over 2500 Mendelian disorders—one of the early fruits of the Human Genome Project (2)—has been recently spectacularly advanced by sequenc-ing entire exomes (the protein-coding con-tent of the genome) (3). Nevertheless, we will need to closely examine gene-regulatory

sequences to understand the full spectrum of Mendelian phenotypic variation. Indeed, on page 966 of this issue, Lee et al. (4) demon-strate that a disorder called Joubert syndrome is caused by mutations in either of two differ-ent, adjacent genes that share a common reg-ulatory region (constituting a so-called cis-regulatory module). This is one example of how human genetics is maturing from a focus on single genes into a more genomic view.

Mutation analyses of single-gene defects have identified two puzzles: One is that not all individuals with a specific disorder have identifiable coding mutations; the other is that not all individuals with identical muta-tions, even in the same family, are equally affected, and some may be symptom-free. The first mystery has many suspected causes: The disorder may be due to another gene—even the adjacent one, as Lee et al. demon-strate—or arise from mutations in a gene’s regulatory sequences, or be a phenocopy (a trait that is not of genetic origin but is envi-ronmentally induced and mimics the pheno-type produced by a gene) (5). This is a persis-

Mendelian Puzzlesgenetics

Aravinda Chakravarti and Ashish Kapoor

Variations that lie outside of the coding region of a mutated gene can give rise to a range of clinical phenotypes for a Mendelian genetic disorder.

CenterforComplexDiseaseGenomics,McKusick-NathansInstituteofGeneticMedicine,JohnsHopkinsUniversitySchoolofMedicine,Baltimore,MD21205,USA.E-mail:aravinda@jhmi.edu

a random network. The continuous random network (CRN) model, originally introduced by Zachariasen (5) for silica glasses, was extended to amorphous silicon by Polk (6) and since has repeatedly been shown to fit the RDF well. Microcrystalline models, based on interconnecting regions of very small crystal grains, had earlier been dismissed as unable to fit the RDF, but it now appears that these models were too simple and did not properly include strains introduced between crystalline regions. The figure highlights examples of a random network, and a topologically cubic (“paracrystalline”) region, in one of Treacy and Borisenko’s models used to fit the experi-mental data for a-Si.

Problems with the CRN model were iden-tified many years ago. Early qualitative trans-mission EM (TEM) seemed to show that a more ordered microcrystalline structure may better explain the first high-resolution TEM images (7), yet it was soon realized that even a random model could have fluctuations that mimic ordering so that qualitative inspection of images could not distinguish these models (8). This limitation was taken as further evi-dence for the random network model.

In the 1990s, Treacy and Gibson (9) devel-oped a quantitative approach for examining scattering fluctuations on the atomic scale from TEM, which they named fluctuation EM (FEM). Their initial results showed high sen-sitivity of FEM to topological ordering and suggested experimentally that it would be hard to obtain a continuous random network in a-Si unless it has been well annealed.

Controversy has remained. Although most workers who have attempted FEM on a-Si have found similar results to Treacy and co-workers, some have assumed that the volume fraction of ordered regions is small (10). Gib-son et al. (11) demonstrated experimentally that the volume fraction of paracrystalline material is substantial (~50%).

The importance of including other con-straints together with the RDF in structural refinement has been recognized, for example, by Billinge (12). The blind spot in the RDF is particularly pronounced at the medium range; it is very insensitive to topology because it only examines the distribution of randomly selected pairs. The FEM data, which come from statistical studies of coherent nanodif-fraction, depend on higher-order correlations, such as the four-body pair-pair or “bond cor-relation” functions. Because this function starts with the local orientation of a bond and can reveal whether there are other bonds in the vicinity that are correlated in direction, it is much more sensitive to topology.

Treacy and Borisenko have reported

here a major step forward by carrying out an experimentally constrained relaxation of structural models with both RDF and FEM data, combined with exploration of the topo-logical characteristics for the structures that emerged. They show that previous studies that assumed only the CRN can fit the RDF data were misguided. Just as good a fit can be obtained from either a random or para-crystalline model. However, the FEM data can only be fitted by a structure with a sub-stantial fraction of paracrystallinity. Their result shows that the identification of an RDF with good fit to a CRN is not sufficient to con-clude that the CRN is a good model of the structure. Other data that are more sensitive to topological or medium-range order are necessary to constrain structures.

That a-Si does not readily form a random network is consistent with our knowledge that no glass transition from the liquid state to the solid state can occur for symmetry reasons, in contrast to silica or metallic glasses. It appears that the topologically paracrystalline state is

not thermally stable and that well-annealed a-Si approaches the random network struc-ture. Because the nature of defects in amor-phous networks would be controlled by local topology, experimentally constrained molecu-lar modeling should be an important line of study in understanding electrical, mechanical, and other properties of amorphous materials.

References 1. M.M.J.Treacy,K.B.Borisenko,Science335,950(2012). 2. R.A.Street,Hydrogenated Amorphous Silicon(Cambridge

Univ.Press,Cambridge,UK,1991),p.36. 3. W.E.Spear,P.G.LeComber,Solid State Commun.17,

1193(1975). 4. P.Anderson,Phys. Rev. Lett.34,953(1975). 5. W.H.Zachariasen,J. Am. Chem. Soc.54,3841(1932). 6. D.E.Polk,J. Non-Cryst. Solids5,365(1971). 7. M.L.Rudee,A.Howie,Philos. Mag.25,1001(1972). 8. J.F.Graczyk,P.Chaudhari,J.Non-Cryst.Solids17,299

(1975) 9. M.M.J.Treacy,J.M.Gibson,Acta Crystallogr.52,212

(1996).10. S.N.Bogleet al.,J. Phys. Condens. Matter19,455204

(2007).11. J.M.Gibson et al.,Phys. Rev. Lett.105,125504(2010).12. S.J.L.Billinge,Physics3,25(2010).

10.1126/science.1218723 on

Mar

ch 3

, 201

2w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from