Engineering nanoscale order into a designedprotein fiberDavid Papapostolou†, Andrew M. Smith‡§, Edward D. T. Atkins¶, Seb J. Oliver�, Maxim G. Ryadnov†, Louise C. Serpell‡,and Derek N. Woolfson†,††‡‡

†School of Chemistry, University of Bristol, Bristol BS 8 1TS, United Kingdom; ‡Department of Biochemistry, School of Life Sciences, University of Sussex,Falmer BN1 9QG, United Kingdom; ¶Department of Physics, University of Bristol, Bristol BS8 1TL, United Kingdom; ††Department of Biochemistry,University of Bristol, Bristol BS8 1TD, United Kingdom; and �Department of Physics, University of Sussex, Falmer BN1 9QH, United Kingdom

Edited by Alan R. Fersht, University of Cambridge, Cambridge, United Kingdom, and approved May 21, 2007 (received for review January 30, 2007)

We have established a designed system comprising two pep-tides that coassemble to form long, thickened protein fibers inwater. This system can be rationally engineered to alter fiberassembly, stability, and morphology. Here, we show that ratio-nal mutations to our original peptide designs lead to structureswith a remarkable level of order on the nanoscale that mimicscertain natural fibrous assemblies. In the engineered system, thepeptides assemble into two-stranded �-helical coiled-coil rods,which pack in axial register in a 3D hexagonal lattice of size 1.824nm, and with a periodicity of 4.2 nm along the fiber axis. Thismodel is supported by both electron microscopy and x-raydiffraction. Specifically, the fibers display surface striationsseparated by nanoscale distances that precisely match the4.2-nm length expected for peptides configured as �-helices asdesigned. These patterns extend unbroken across the widths(>50 nm) and lengths (>10 �m) of the fibers. Furthermore, thespacing of the striations can be altered predictably by changingthe length of the peptides. These features reflect a high level ofinternal order within the fibers introduced by the peptide-design process. To our knowledge, this exceptional order, and itspersistence along and across the fibers, is unique in a biomimeticsystem. This work represents a step toward rational bottom-upassembly of nanostructured fibrous biomaterials for potentialapplications in synthetic biology and nanobiotechnology.

bionanoscience � nanofibers � peptide assembly � rational protein design �self-assembly

An ability to design nanostructured, self-organizing systemsfrom the bottom up would have an impact on nanoscale

science and technology. Natural peptide and protein-basedself-assembling systems are found extensively throughout biol-ogy (1). These natural systems offer inspiration in the effort toengineer novel assemblies through de novo design. Such studiesare in their infancy, but interest in the area is intense because anability to engineer or design water-soluble self-assembling sys-tems offers routes to novel biomaterials with potential applica-tions in synthetic biology and nanobiotechnology (2–8), forinstance, the preparation of biocompatible scaffolds for regen-erative biology (cell and tissue engineering) (9).

Self-assembling systems based on peptides, proteins, DNA,and RNA all are being explored (2, 4–8). Our focus has been onpeptide-based assemblies (7). Specifically, we are interested inmaking self-assembling fibers and fiber-based networks fromrelatively simple and synthetically accessible peptide buildingblocks. Fibrous structures are probably the most abundant andstraightforward self-assembling systems found in nature and,therefore, provide a sensible target for designing self-assemblingpeptides. In addition, biological fibrous structures perform awide variety of functions both within and outside cells (1). Forthese reasons, considerable work has been undertaken to mimicfibrous protein assemblies using self-assembling peptides. Pre-dominantly, such work has used �-structured peptides that formamyloid-like structures (3, 5, 6, 10). Relatively less has been

attempted with �-helix-based assemblies, however (7, 11–14).The development of �-helical systems would provide usefulcomparisons with the more-explored amyloid-like systems andalso allows the application of the considerable body of proteindesign and engineering knowledge for �-helical structures andassemblies (15–17). We have built fibrous assemblies based onthe �-helical coiled coil, which is one of the most widespread andbest understood protein–protein interaction motifs known (7,17–21). In particular, considerable attention has been paid toone type of coiled-coil architecture, namely the leucine-zippermotif, which is accepted as the archetypal two-stranded, parallelcoiled coil. Previously, we have combined established rules forleucine-zipper assembly in the first-generation design of a self-assembling-fiber (SAF) system (7, 11, 17).

The original SAF design is based on two complementaryleucine-zipper peptides, SAF-p1 and SAF-p2, of de novo design(11). However, unlike natural and foregoing designed leucinezippers, which are all blunt-ended structures (22, 23), the SAFpeptides are engineered to form offset dimers with complemen-tary sticky ends to promote longitudinal assembly into fibers(Fig. 1).

A combination of biophysical techniques confirms fiber for-mation (11). In particular, electron microscopy (EM) revealsthat the fibers are linear and extended for many microns.Curiously, however, the fibers measure �45 � 10 nm across,which is approximately 20 times thicker than expected forleucine-zipper-based dimers (24). Also, fiber thickening appearsto stop to give a finite distribution of fiber widths (11, 25, 26). Atpresent, we do not understand how fiber thickening occurs orhow it is arrested, although mechanisms proposed for theassembly of fibrous collagens (27) and �-tape structures (28) maybe applicable. Whatever the kinetic mechanism for thickening,this lateral assembly is likely driven by the free energy associatedwith the formation of many low-affinity interactions betweenfeatures that are brought together on neighboring componenttwo-stranded coiled coils (protofibrils) in the thickened fiber, ineffect, an avidity effect (11). As we have described (26), thechanges to the peptides further thicken the assembled fibers and

Author contributions: D.P. and A.M.S. contributed equally to this work; D.P., A.M.S., andD.N.W. designed research; D.P., A.M.S., and L.C.S. performed research; M.G.R. contributednew reagents/analytic tools; D.P., A.M.S., E.D.T.A., S.J.O., L.C.S., and D.N.W. analyzed data;and D.P., E.D.T.A., L.C.S., and D.N.W. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.

Abbreviations: AM, ammonium molybdate; SAF, self-assembling-fiber; EM, electronmicroscopy; TEM, transmission EM; UA, uranyl acetate; WAX, wide-angle x-ray diffraction.

§Present address: Material Science Centre, University of Manchester, P.O. Box 88, ManchesterM60 1QD, United Kingdom.

‡‡To whom correspondence should be addressed. E-mail: d.n.woolfson@bristol.ac.uk.

This article contains supporting information online at www.pnas.org/cgi/content/full/0700801104/DC1.

© 2007 by The National Academy of Sciences of the USA

www.pnas.org�cgi�doi�10.1073�pnas.0700801104 PNAS � June 26, 2007 � vol. 104 � no. 26 � 10853–10858

BIO

PHYS

ICS

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 17

, 202

1

also stabilize them with respect to thermal denaturation. Here,we describe detailed structural studies of the rationally rede-signed fibers, which illustrate that peptide assemblies withnanoscale order can be engineered from first principles and fromthe bottom up. Briefly, we used transmission EM (TEM) andx-ray fiber diffraction to study the nanoscale to mesoscalestructure and order in a series of rationally designed fiber-forming peptides. The resulting data were combined to producea molecular model for the folding and packing of the peptidesinto fibers. In turn, the model was used to back-calculate thex-ray fiber diffraction, linking the experimental and theoreticaldata.

Results and DiscussionTo discuss the work presented here and place it in context of ourpreviously published work (26), it is necessary to summarize thelatest work in the next two subsections. The previously publishedwork centers on improving the stability of the SAFs throughrational design. As described below, the design peptides are givenin ref. 26. However, here we present additional structural dataand focus on using this information to develop detailed molec-ular models for the SAF structure.

Rational Peptide Redesign. To further stabilize lateral fiber as-sembly in the SAFs, we sought to enhance potentially comple-mentary features presented on the surface of the leucine-zipperbuilding blocks of the SAF-p1/SAF-p2 design. Inspired by pro-posed mechanisms for the higher-order assembly of naturalfibrous proteins (29–32), we argued that this might best beachieved by optimizing potential interprotofibril Coulombic(charge–charge) interactions, that is, rather than by introducinghydrogen-bonding interactions, which would require specificorientation of the partnering protofibrils. To this end, wemodeled contiguous copies of the SAF-p1 and SAF-p2 se-quences (Table 1) as an extended dimeric coiled coil (26).[Leucine zippers, and coiled-coil proteins in general, share aseven-residue sequence repeat of hydrophobic (H) and polar (P)amino acids, HPPHPPP often designated abcdefg. This directsthe folding and subsequent assembly of amphipathic �-helices toform the so-called coiled coils. The first- and second-generationSAF peptides described herein have four heptads or 28 residues.]Inspection of the model suggested that two aspartic-acid sidechains at consecutive b sites in the heptad repeat of SAF-p1formed negatively charged pairs that would wind around thesurface of the protofibril (Table 1). Therefore, to introducepotentially complementary positively charged pairs, we rede-signed SAF-p2 to incorporate two arginine residues also spacedseven residues apart at two consecutive c sites. This resulted inpeptide SAF-p2a (Table 1), which should combine with SAF-p1to give protofibrils with matching acidic and basic patches ontheir surfaces (26). We refer to the SAF-p1:SAF-p2a combina-tion as the second-generation SAFs.

Biophysical Characterization. As judged by CD spectroscopy (spec-tra not shown) and EM (see below), SAF-2pa combined withSAF-p1 to form fibers that were at first sight similar to thoseoriginally observed for SAF-p1 and SAF-p2 (11). Consistent withan improved design, the critical concentration for fiber forma-tion by SAF-p1/SAF-p2a improved to �30 �M compared with�60 �M for the original design, and the second-generation fiberswere thicker (�70 � 20 nm) and better defined (Fig. 2A).Furthermore, the temperatures up to which fibers could beassembled also improved: previously, SAF-p1:SAF-p2 mixturesyielded fibers only below room temperature and assembly wasbest at 5°C; the new SAF-p1:SAF-p2a design, however, could beassembled at up to 22°C. The solution-phase biophysical exper-iments are described in detail in ref. 26.

Nanoscale Order in the Matured Fibers. The change in thermal-unfolding behavior between the first- and second-generationdesign fibers prompted us to look for differences in theirstructure and morphology by negative-stain TEM. Like thefirst-generation fibers (Fig. 2 A and B; ref. 11), the redesignedfibers were linear, nonbranched, and extended for many microns.However, the latter were thicker, which, along with the improvedthermal stability, is fully consistent with the redesign principles.

Fig. 1. Design principles of the SAF peptides. (A) Natural and previouslydesigned coiled coils are blunt-ended structures. (B) In the SAF system, com-plementary charges in companion peptides direct the formation of staggered,parallel, and codirectional heterodimers. The resulting ‘‘sticky ends’’ are alsocomplementary and promote longitudinal association into extended fibers.Asparagine residues (stars) at the coiled-coil interface preferentially interactwith each other, cementing the prescribed register further (11). (C) Comple-mentary charged pairs on the outer surfaces of the coiled-coil protofibrilspromotes protofibril–protofibril interactions, fiber assembly, and thickening.

Table 1. SAF peptide sequences

Peptide Sequence

Heptad repeat g abcdefg abcdefg abcdefg abcdefg abcdef

SAF-p1 K IAALKQK IASLKQE IDALEYE NDALEQ

SAF-p2 K IRALKAK NAHLKQE IAALEQE IAALEQ

SAF-p2a K IRRLKQK NARLKQE IAALEYE IAALEQ

Ac-SAF-p1-NH2 Ac-K IAALKQK IASLKQE IDALEYE NDALEQ-NH2

Ac-SAF-p2a-NH2 Ac-K IRRLKQK NARLKQE IAALEYE IAALEQ-NH2

SAF-p1-ext K IAALKQK IASLKQE IDALEYE NDALEQK IAALEQ

SAF-p2a-ext K IRRLKQK NARLKQK IAALEQE IAALEYE IAALEQ

Differences between successive designs are highlighted by bold italics.

10854 � www.pnas.org�cgi�doi�10.1073�pnas.0700801104 Papapostolou et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 17

, 202

1

In addition, and curiously, the second-generation fibers showedclear evidence of order above that in the first-generation struc-tures (compare Fig. 2 A and C). Specifically, the second-generation fibers showed well defined periodic light and darkbanding patterns (referred to here as ‘‘striations’’) perpendicularto the long fiber axis. These patterns persist along the entirelengths (approximately tens of microns) of the fibers and extendcompletely across their widths perpendicular to the long fiberaxes (Fig. 2C).

Like the reports for 2D crystals of truncated prion proteins

(33) and fibrous collagens (34), the dark striations probablyresult from the uranyl acetate (UA) acting as a positive stain,interacting with anionic moieties. This is supported by theobservation in our system that copious washing of the stainedfibers with water on the carbon TEM grids removed background(negative) stain, but failed to remove the striations. Consistentwith this finding, similar patterns with the same spacing ofstriations (see below) were observed for fibers grown in sodiumphosphate buffer and stained with a solution of phosphotungsticacid (a cationic stain, like UA) and also with the anionic stainsammonium molybdate (AM) and dimolybdate. Interestingly,similar patterns with comparable relative widths of the light anddark bands were observed for fibers stained with AM followedby washing with distilled water and a second staining with UA[see supporting information (SI) Appendix], that is, rather thantwo overlapping patterns or one with wider dark bands. Apossible explanation of this observation is that the anionic andcationic peptide moieties likely to be interacting with UA andAM are proximal.

The Striation Spacings Match the Lengths Expected for Folded Pep-tides. From our earlier studies (11), x-ray diffraction from drawnfibers of the first-generation SAFs indicates that the SAF peptideassemble as designed: that is, each peptide adopts an �–helicalconformation and these pair to form twisted two-strandedcoiled-coil structures as first postulated by Francis Crick (35). Inaddition, the axes, or long directions, of these coiled coils alignwithin the fiber along the long fiber axis itself. These architec-tural features are also evident in the diffraction patterns fromfibers prepared from the redesigned peptides as discussed below.

Therefore, it is reasonable that the striations observed by EMrelate to features that repeat in the peptides and/or along thecoiled-coil superhelical axis. Consistent with this concept, Fou-rier transform analysis of individual fibers from the positive-stainelectron micrographs revealed that the striations were spaced 4.2nm apart (SD 0.13 nm over 71 measurements). This experimen-tal measurement closely matches the distance that a 28-residueSAF peptide would cover in a fully folded �-helical coiled-coilconformation (i.e., 28 � 0.148 nm � 4.144 nm, as the rise perresidue in a coiled coil is 0.148 nm), which also corresponds to�1/4 of the supercoil pitch for a two-stranded coiled coil (36).

Together, the x-ray and TEM data suggest that the striationshighlight the longitudinal repeat of the peptides along the fibermain axis. Furthermore, as individual striations cover the entirewidth of the fibers without interruption, it is likely that the helicalpeptides are also in register across the widths of the fiber.

To test the correspondence between peptide length and theseparation between striations further, a fifth heptad was addedto the second-generation design (Table 1) extending thepeptides to 35 aa, which should increase the length of eachindividual helix to �5.2 nm. Fibers assembled from thesethird-generation peptides were imaged by TEM and striationswere observed (Fig. 2 E and F). Consistent with our design andprediction, Fourier transform analysis returned a striationlength of 5.2 nm (SD 0.07 nm over 23 measurements).

Other groups have described longitudinal stripes on negativelystained images of designed fibers (37, 38) and natural proteins(39). It is important to emphasize the second- and third-generation SAFs show identical striation patterns upon positiveand negative staining. Thus, we believe that any signal fromlongitudinal stripes of negatively stained SAFs, if any exist on theSAFs at all, is too weak to compare with the lateral striations.

Internal Nanoscale Order. The wide-angle x-ray diffraction (WAX)pattern from an oriented, partly dried fiber of second-generationSAFs is shown in Fig. 3. The overall x-ray diffraction features aresimilar to those predicted (35), and seen (40) for �–helical coiled-coil conformations, and, in particular, double-stranded coiled coils.

Fig. 2. Imaging the SAFs. Transmission electron micrographs of UA-stainedfirst-generation fibers (SAF-p1:SAF-p2; A and B), second-generation fibers(SAF-p1:SAF-p2a; B and C), and third-generation fibers (SAF-p1-ext:SAF-p2a-ext; E and F), matured for 12 h at 5°C (first generation), 22°C (second gener-ation), and 37°C (third generation). A, C, and E are high-magnification images.(Scale bars: 50 nm.) B, D, and F are at low magnification. (Scale bars: 2 �m.)

Papapostolou et al. PNAS � June 26, 2007 � vol. 104 � no. 26 � 10855

BIO

PHYS

ICS

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 17

, 202

1

The prominent second-order layer line (marked by L in Fig. 3) bearsa relationship to the relatively strong 0.54-nm layer line typical ofthe undistorted �–helical conformation. However, it is important tonote that there is no expected meridional (c* axis) diffraction signalon this layer line for an undistorted �–helix (35, 41). Indeed, the firstmeridional diffraction signal expected for an �–helical coiled coilis at 0.148 nm (35), but it is not typically observed in standard WAXas it falls in the ultra-wide-angle region. The observed sharpmeridional diffraction arc at 1.036 and 0.518 nm (error limits �0.003 nm) (labeled M1 and M2, respectively, in Fig. 3) emanatefrom the signature seven-residue repeat of side chains of the coiledcoil (35). These are, therefore, related to the aforementioned 0.148signal: 7 � 0.148 nm � 1.036 nm. Thus, the x-ray diffractionfingerprint is wholly consistent with double-stranded �–helicalcoiled coil; moreover, these structures are aligned parallel to thelong fiber axis, consistent with the original design hypothesis (11).

The relative sharpness of the meridional arcs indicate coher-ent length orders of magnitude longer than the lengths of themolecular axial repeat alone, the value of which is 28 � 0.148nm � 4.144 nm plus a small addition for the terminating endgroups. [It does not matter that the two strands in the SAFcoiled-coil design are staggered relative to each other (11); the4.144-nm repeat value remains the same.] Thus, the coiled-coilpeptides must connect and self-assemble (along the axial direc-tion) in a precise, integer fashion as multiples of the meridionalrepeat of 1.036 nm. For the second-generation SAFs this will be4 � 1.036 � 4.144 nm. This value is consistent with the spacingof 4.2 nm observed by TEM, thus linking the two sets ofexperimental measurements (TEM and WAX).

A series of relatively sharp equatorial (a*b* plane) diffractionsignals (labeled E1, E2, E3, and E4 in Fig. 3) were observed atd-spacings: 1.578, 0.912, 0.790, and 0.595 nm (error limits � 0.003nm), respectively. These signals represent the packing arrangementof the double-strand coiled coils in the ab plane, perpendicular tothe fiber axis (c axis). They are, within experimental error, in theratio 1/�1 (d100):1/�3 (d110):1/�4 (d200):1/�7 (d210) and index ona hexagonal net of side a � 1.824 � 0.003 nm. Thus, the observed

diffraction signals can be indexed on a hexagonal unit cell withparameters: a � b � 1.824 nm, c (molecular axis) � 1.036 nm, � �� � 90°, � � 120°.

We observed that the second layer line (L) is sampled: forexample (at the head of the arrow labeled L in Fig. 3; and listedin table 2 in SI Appendix), the relatively strong diffraction at0.488 � 0.005 nm is 102 diffraction signal with a calculatedd-spacing of 0.492 nm; other weaker diffraction signals are alsodiscrete sampling peaks and not Bessel function maxima. Thisfinding indicates that the molecules are in 3D register with eachother; i.e., there is no relative axial slippage between themolecules. Indeed, if slippage had occurred we would expectlayer line streaking, which is not observed. These x-ray diffrac-tion results are wholly consistent with the observed 0.42-nmorthogonal striations observed in the fibers by using real-spaceEM imaging.

For completeness, we note that the true crystallographicperiodicity of a slowly twisting coiled-coil structure of this kindis in the region 15–18 nm (35). Furthermore, in our case thecoiled coil is not continuous, but will have scattering dislocationsapproximately every 4 � 1.036 � 4.144 nm from the termini ofthe peptides that make up the fiber. In both cases, however, therelatively weak scattering expected would require more-crystalline samples and would have to be sought in the ultra-lowto low-angle x-ray diffraction regions.

A Molecular Model for the SAF Structure. Putting the above obser-vations and analyses together, we derived a model for the structureof the SAFs in which canonical two-stranded �-helical coiled coilsare aligned along the long axis of the fibers and packed hexagonallyaround this axis (Fig. 4). In addition, from other experiments thatindicate the fibers are polar, the coiled coils are aligned parallel, i.e.,in one direction along the fiber (42).

To compare this model and the observed experimental data,a simulated diffraction pattern (Fig. 3) was generated by usingthe program CLEARER (43) and the above unit cell dimensionsand model coordinates for a unit cell. The positions of thediffraction signals in the simulated and experimental diffractionpatterns match very closely (Fig. 3). However, the experimentaldata do not allow a detailed molecular description of the coiledcoils with accurate positions of side chains; therefore, theintensities in the experimental and calculated patterns will notnecessarily compare well. Nonetheless, this simulation stronglysupports the overall arrangement of coiled coils in the model,namely, longitudinal registered coiled-coil protofibrils arrangedlaterally on a hexagonal lattice (Fig. 4).

Origin of the Positive Staining and the Resulting Striations. Theapparent colocation of the anionic and cationic stains observed inthe TEM experiments (even with the inherent low resolution of �2nm with these methods) possibly suggests that they are highlightingproximal oppositely charged moieties in the fibers. One possibilityis the N- and C-terminal extremities of each peptide that abut in thedesign; i.e., UA could be interacting with the carboxylates of the Ctermini, and AM could be interacting with the primary amines ofthe N termini. To probe this possibility, we synthesized cappedversions of SAF-p1 and SAF-p2a, Ac-SAF-p1-NH2, and Ac-SAF-p2a-NH2, in which the terminal charges were replaced by neutralamide bonds (Table 1). Unfortunately, the mixture of both cappedpeptides did not render fibers. When the individual capped peptideswere mixed with their complementary noncapped peptide (i.e.,Ac-SAF-p1-NH2 with SAF-p2a, and Ac-SAF-p2a-NH2 with SAF-p1) fibers were observed, but only in preparations incubated at 5°C.This destabilization is fully consistent with our design principles; i.e.,that the ends of the (uncapped) peptides are close enough in spaceto abut and form stabilizing intermolecular salt bridges (Fig. 1) (11).Perhaps surprisingly, however, the fibers containing one cappedpeptide were still striated (see SI Appendix), and the distances

Fig. 3. Intermolecular organization in the redesigned fibers. WAX patternobtained from an oriented, partly dried fiber prepared from a 1:1 mixture ofSAF-p1/SAF-p2a molecules, incident x-ray beam (wavelength, � � 0.15418 nm)orthogonal to the fiber axis (vertical). Note: the first strong hyperbolic layer lineL at the top (and bottom) of the pattern [this corresponds to the fifth layer linein the undistorted �–helix diffraction pattern and the strongest contributingBessel function J1 (41)]; the sharp meridional diffraction arcs M1 and M2 (thefundamental and second order of 1/1.026 � 0.003 nm); the relatively sharpequatorial arcs E1, E2, E3, E4, with d-spacings in the ratio 1/�1: 1/�3: 1/�4: 1/�7.The simulated pattern (top right) closely matches the experimental data.

10856 � www.pnas.org�cgi�doi�10.1073�pnas.0700801104 Papapostolou et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 17

, 202

1

between striations was the same as that in normal second-generation fibers. Thus, it is unlikely, that the stains are interactingwith the peptide extremities, at least not with the N- and C-terminalcharges.

Another possibility is that the stains are interacting with thecharged side chains on the surfaces of the folded peptides. Twogroups of such side chains can be considered: first, there arethose at the e and g positions of the heptad repeats (Table 1),which help direct and cement the SAF-p1:SAF-p2 heterodimers.However, these are common to both the first-, second-, andthird-generation SAF, and striations are observed only for thesecond and third generations. Second, there is the additionalquartet of aspartic acid and arginine residues introduced be-tween the first- and second-generation designs. In this case,cationic UA stain could interact with one or both of the asparticacid residues of SAF-p1 and the anionic AM stain could interactwith the complementary arginine residues of SAF-p2a. Thesegroups of charged residues are within 1 nm of each other alongthe coiled-coil axis. Thus, within the resolution of the stainedTEM, it is reasonable to consider these groups of residues as onecluster of negative and positive charge repeated every peptidelength along the fibers. We favor this model for the origin ofpositive staining to give lateral striations. However, the removalof this cluster, i.e., effectively partially reversing the first- tosecond-generation experiment, drastically reduces the stabilityof and order in the fibers, and it is difficult to see how thishypothesis could be tested by mutagenesis alone.

ConclusionTo summarize our studies and results, we have built on our previousdesign of a SAF system considerably. Specifically, after noting thatthe first-generation fibers were thickened we sought to explore andpromote this by engineering improved protofibril–protofibril inter-

actions. This rational redesign results in second-generation fibersthat are indeed thickened further and stabilized. Moreover, theseredesigned fibers show intriguing nanoscale features in positive-stain TEM and considerable internal nanoscale order as judged byWAX. The features observed by TEM are striations orthogonal tothe long fiber axis. These extend coherently along the whole length(�10 �m) of the fibers and run uninterrupted across the widths ofthe fibers (�50 nm). The separation between the striations preciselymatches the lengths of the peptide building blocks as originallydesigned (11), that is, configured as �-helices and assembled end toend along the length of the fiber. In further support of thisarrangement, WAX reveals contiguous coiled coils along the longaxis of the fibers, which are packed hexagonally across the widthsof the fibers. A resulting molecular model for the structure has beenderived and validated by back-calculating the WAX pattern.

We propose that the structure and order in the SAFs arisesthrough the aforementioned protofibril–protofibril interactions in-troduced in the peptide design process: weak, complementaryinteractions are repeated with helical symmetry along the long axisof the fibers; these combine cooperatively to cement protofibril–protofibril interactions throughout the fibers. As a result, the fibersare thickened, stabilized, and highly ordered.

To our knowledge, the level of organization that we observe hasnot been reported for any other biomimetic fibrous peptide orprotein system. The ultrastructure observed is comparable, at leastqualitatively, to that seen in natural protein fibers, for instance, incertain natural protein fibers, including some collagens, fibrins, andlamins (27, 34, 44, 45). In these cases, this external order reflectshigh internal organization within these highly evolved systems. Ourfibers, however, are only the second and third iterations in a rationaldesign process. This unexpected mimicry in a relatively simple anddesigned binary peptide system is intriguing and may shed light onboth natural and synthetic self-assembly processes in general. Incomparison to other designed �-helix-based fibers, the SAFs displayconsiderably more order in their assembled structures (7, 17, 46);although we note longitudinal stripes are observed by TEM in onesystem (38), and hexagonal packing of helical fibrils has also beendemonstrated by WAX in another (12). In addition, because oursis a dual-peptide system, whereas the others rely on just one peptideto self-assemble, we have more control over the assembly process(42). This ability to control assembly brings further utility to thesystem as it allows additional functionalized peptides to be added toself-assembly mixtures. In turn, it allows fibers morphology andfunction to be engineered (7, 25, 42, 47–49). In view of currentattention being given to the generation of bioinspired materials forapplications in nanobiotechnology (9, 46, 50), our findings offerpromise for nanoscale engineering of self-assembling systems fromthe bottom up in water; for example, the directed incorporation offunctionalized peptides regularly spaced along the fibers can beenvisaged from the observations presented here.

Materials and MethodsPeptide Synthesis. Peptides were made chemically and purified asdescribed (11, 42). The final products were identified byMALDI-TOF MS (Micromass, Manchester, U.K.). MS [M�H]�: SAF-p1, m/z 3174 (calc), 3175 (found); SAF-p2, m/z 3128(calc), 3129 (found); SAF-p2a, m/z 3325 (calc), 3326 (found);SAF-p1-ext, m/z 3927 (calc), 3928 (found); SAF-p2a-ext, m/z4079 (calc), 4080 (found); Ac-SAF-p1-NH2, m/z 3215 (calc), 3214(found); and Ac-SAF-p2a-NH2, m/z 3367 (calc), 3367 (found).

EM. Samples for TEM were prepared as described (11), exceptphosphotungstic acid-stained fibers that were grown in phos-phate buffer and stained with a 1% solution of the stain in waterwere used. Fiber samples for the third-generation peptides andthe capped versions of the second-generation peptides wereprepared similarly except that fibrollogenesis was performedovernight at 37°C and 5°C, respectively. In all cases, the final

Fig. 4. Hexagonal-packing model for the designed SAF coiled coils asdeduced from the x-ray diffraction pattern (Fig. 3). (A) Schematic view of afiber cross-section. The coiled-coils are packed in an hexagonal lattice of size1.824 nm (a-axis). (B) Computer model showing two dimensions of the hex-agonal lattice. (C) Additional 3D schematic section illustrating the proposedorganization of protofibrils in the fibers. The black sections orthogonal to thefiber axis represent an EM stain interacting with the fibers with a periodicityof 4.2 nm, as seen with the 2-nm resolution inherent to the technique.

Papapostolou et al. PNAS � June 26, 2007 � vol. 104 � no. 26 � 10857

BIO

PHYS

ICS

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 17

, 202

1

concentration of each SAF peptide was 100 �M. Micrographswere taken at �1,000 and �40,000 by using software from Gatan(Pleasanton, CA). Images were converted to the Image2000Software format (51) and then examined with XIMDISP (52).Fourier transforms of selected boxed, padded, and floatedregions were interpolated and calculated interactively to searchfor diffraction spacings. A catalase calibration specimen (AgarScientific, Essex, U.K.) was processed in the same way as thesamples to confirm the calibration of the microscope at allmagnifications.

X-Ray Fiber Diffraction. The x-ray fiber diffraction pattern for theSAF-p1:SAF-p2a assembly was recorded for a sample preparedunder the stretched frame procedure (53). A droplet of the freshsolution containing 700 �M of each peptide was placed betweenthe ends of two wax-filled capillaries distant from 1.5 mm fromeach other and allowed to dry slowly at 22°C, yielding a dried rodcomposed of partially aligned fibers. X-ray fiber diffractionimages were collected using a CuK rotating anode source(wavelength 0.15418 nm) and a R-AXISIV detector (Rigaku,Tokyo). The sample was kept at room temperature and placedat a distance of 300 mm to the detector, then exposed for 10 minover a rotation of 0.5° angle.

Model Building. Atomic coordinates for a SAF-p1:SAP-p2a two-stranded coiled were generated by using the programMAKECCSC kindly provided by Gerald Offer (personal com-munication) with standard parameters for a double-strandedcoiled coil (24): pitch, 15 nm; radius of the supercoil, 0.47 nm;

helical rise per amino acid unit, 0.1495 nm. Successive side chainswere placed in their favored conformations as observed in aside-chain rotamer library created from a database of dimericcoiled coils of known 3D structure (ref. 22 and D.P., GeraldOffer, and D.N.W., unpublished work). The output pdb fileconsisted of a 140-residue dimeric coiled coil with the amino acidsequences of SAF-p1 (chain A) and SAF-p2a (chain B) repeatedfive times each. PyMOL (www.pymol.org) was used to generatethe symmetry related objects (hexagonally packed coiled coils)based on the experimentally measured unit-cell dimensions.

Calculation of Simulated X-Ray Diffraction Pattern from Model Coor-dinates. A simulated diffraction pattern was calculated fromthe modeled coordinate file by using the program CLEARER(43). The coordinates were arranged within a unit cell withdimensions a � b � 1.824 nm, c � 1.036 nm, � � � � 90°, � �120°, and the long axis of the fiber was oriented parallel to c.The simulated diffraction pattern was calculated by adding theintensities from all reciprocal lattice points, taking account ofthe fiber-axis direction, crystallite dimensions, disorder, anddiffraction geometry. Diffraction settings were identical to theexperimental settings.

We thank Gerald Offer, Sam MacDonald, and John Squire for valuablediscussions. A.M.S. received studentship support from the Engineeringand Physical Sciences Research Council. D.N.W. and L.C.S. weresupported by Biotechnology and Biological Sciences Research CouncilGrant B04676. S.J.O. was supported by Medical Research Council GrantG0300584.

1. Pollard TD, Earnshaw WC (2002) Cell Biology (Saunders, London).2. Yeates TO, Padilla JE (2002) Curr Opin Struct Biol 12:464–470.3. Zhang SG, Marini DM, Hwang W, Santoso S (2002) Curr Opin Chem Biol

6:865–871.4. Seeman NC (2003) Biochemistry 42:7259–7269.5. MacPhee CE, Woolfson DN (2004) Curr Opin Solid State Mat Sci 8:141–149.6. Rajagopal K, Schneider JP (2004) Curr Opin Struct Biol 14:480–486.7. Woolfson DN, Ryadnov MG (2006) Curr Opin Chem Biol 10:559–567.8. Jaeger L, Chworos A (2006) Curr Opin Struct Biol 16:531–543.9. Holmes TC (2002) Trends Biotechnol 20:16–21.

10. Scheibel T, Parthasarathy R, Sawicki G, Lin XM, Jaeger H, Lindquist SL (2003)Proc Natl Acad Sci USA 100:4527–4532.

11. Pandya MJ, Spooner GM, Sunde M, Thorpe JR, Rodger A, Woolfson DN(2000) Biochemistry 39:8728–8734.

12. Potekhin SA, Melnik TN, Popov V, Lanina NF, Vazina AA, Rigler P, VerdiniAS, Corradin G, Kajava AV (2001) Chem Biol 8:1025–1032.

13. Zimenkov Y, Conticello VP, Guo L, Thiyagarajan P (2004) Tetrahedron60:7237–7246.

14. Wagner DE, Phillips CL, Ali WM, Nybakken GE, Crawford ED, Schwab AD,Smith WF, Fairman R (2005) Proc Natl Acad Sci USA 102:12656–12661.

15. Schneider JP, Lombardi A, DeGrado WF (1998) Folding Des 3:R29–R40.16. Hill RB, Raleigh DP, Lombardi A, Degrado NF (2000) Acc Chem Res

33:745–754.17. Woolfson DN (2005) Adv Protein Chem 70:79–112.18. Lupas A (1996) Trends Biochem Sci 21:375–382.19. Kohn WD, Hodges RS (1998) Trends Biotechnol 16:379–389.20. Burkhard P, Stetefeld J, Strelkov SV (2001) Trends Cell Biol 11:82–88.21. Mason J, Arndt K (2004) Chem Biol Chem 5:170–176.22. Walshaw J, Woolfson DN (2001) J Mol Biol 307:1427–1450.23. Lupas AN, Gruber M (2005) Adv Protein Chem 70:37–78.24. O’Shea EK, Klemm JD, Kim PS, Alber T (1991) Science 254:539–544.25. Ryadnov MG, Woolfson DN (2003) Nat Mater 2:329–332.26. Smith AM, Banwell EF, Edwards WR, Pandya MJ, Woolfson DN (2006) Adv

Funct Mat 16:1022–1030.27. Prockop DJ, Fertala A (1998) J Struct Biol 122:111–118.

28. Aggeli A, Nyrkova IA, Bell M, Harding R, Carrick L, McLeish TCB, SemenovAN, Boden N (2001) Proc Natl Acad Sci USA 98:11857–11862.

29. McLachlan AD, Stewart M (1982) J Mol Biol 162:693–698.30. Meng JJ, Khan S, Ip W (1994) J Biol Chem 269:18679–18685.31. Parry DAD (1999) in Guidebook to the Cytoskeletal and Motor Proteins, eds

Kreis T, Vale R (Oxford Univ Press, Oxford), pp 285–291.32. Herrmann H, Aebi U (2004) Annu Rev Biochem 73:749–789.33. Wille H, Michelitsch MD, Guenebaut V, Supattapone S, Serban A, Cohen FE,

Agard DA, Prusiner SB (2002) Proc Natl Acad Sci USA 99:3563–3568.34. Holmes DF, Gilpin CJ, Baldock C, Ziese U, Koster AJ, Kadler KE (2001) Proc

Natl Acad Sci USA 98:7307–7312.35. Crick FHC (1953) Acta Crystallogr 6:689–697.36. Seo J, Cohen C (1993) Proteins Struct Funct Genet 15:223–234.37. Lopez de la Paz M, Goldie K, Zurdo J, Lacroix E, Dobson CM, Hoenger A,

Serrano L (2002) Proc Natl Acad Sci USA 99:16052–16057.38. Zimenkov Y, Dublin SN, Ni R, Tu RS, Breedveld V, Apkarian RP, Conticello

VP (2006) J Am Chem Soc 128:6770–6771.39. Quinlan RA, Stewart M (1987) J Cell Biol 105:403–415.40. Fraser RDB, Macrae TP (1973) Polymer 14:61–67.41. Cochran W, Crick FHC, Vand V (1952) Acta Crystallogr 5:581–586.42. Smith AM, Acquah SFA, Bone N, Kroto HW, Ryadnov MG, Stevens MSP,

Walton DRM, Woolfson DN (2004) Angew Chem Int Ed 44:325–328.43. Makin OS, Sikorski P, Serpell LC (2007) J Appl Crystallogr, in press.44. Mosesson MW, Di Orio JP, Hernandez I, Hainfeld JF, Wall JS, Grieninger G

(2004) Biophys Chem 112:209–214.45. Weisel JW (2005) Adv Protein Chem 70:247–299.46. Fairman R, Akerfeldt KS (2005) Curr Opin Struct Biol 15:453–463.47. Ryadnov MG, Woolfson DN (2005) J Am Chem Soc 127:12407–12415.48. Ryadnov MG, Woolfson DN (2004) J Am Chem Soc 126:7454–7455.49. Ryadnov MG, Woolfson DN (2003) Angew Chem Int Ed 42:3021–3023.50. Scheibel T (2005) Curr Opin Biotechnol 16:427–433.51. Crowther RA, Henderson R, Smith JM (1996) J Struct Biol 116:9–16.52. Smith JM (1999) J Struct Biol 125:223–228.53. Serpell LC, Fraser PE, Sunde M (1999) Methods Enzymol 309:526–536.

10858 � www.pnas.org�cgi�doi�10.1073�pnas.0700801104 Papapostolou et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 17

, 202

1