Self-catalyzed growth of S layers via an amorphous-to-crystalline transition limited by folding kineticsSungwook Chunga,b,1, Seong-Ho Shina,c,d,1, Carolyn R. Bertozzia,c,d,2, and James J. De Yoreoa,c,2

aMolecular Foundry, Lawrence Berkeley National Laboratory, Berkeley, CA 94720; bPhysical Biosciences Division, Lawrence Berkeley National Laboratory,Berkeley, CA 94720; cMaterials Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA 94720; and dDepartment of Chemistry, Universityof California, Berkeley, CA 94720

Edited by George M. Whitesides, Harvard University, Cambridge, MA, and approved August 4, 2010 (received for review June 10, 2010)

The importance of nonclassical, multistage crystallization path-ways is increasingly evident from theoretical studies on colloidalsystems and experimental investigations of proteins and biomin-eral phases. Although theoretical predictions suggest that proteinsfollow these pathways as a result of fluctuations that create un-stable dense-liquid states, microscopic studies indicate these statesare long-lived. Using in situ atomic force microscopy to follow 2Dassembly of S-layer proteins on supported lipid bilayers, we haveobtained amolecular-scale picture of multistage protein crystalliza-tion that reveals the importance of conformational transforma-tions in directing the pathway of assembly. We find that mono-mers with an extended conformation first form a mobile adsorbedphase, from which they condense into amorphous clusters. Theseclusters undergo a phase transition through S-layer folding intocrystalline clusters composed of compact tetramers. Growth thenproceeds by formation of new tetramers exclusively at clusteredges, implying tetramer formation is autocatalytic. Analysis ofthe growth kinetics leads to a quantitative model in which tetra-mer creation is rate limiting. However, the estimated barrier ismuch smaller than expected for folding of isolated S-layer proteins,suggesting an energetic rationale for this multistage pathway.

in situ atomic force microscopy imaging ∣ protein crystal growth ∣ two-stepcrystallization ∣ amorphous precursors ∣ assembly kinetics

Self-assembled protein architectures exhibit a range of struc-tural motifs (1) including particles (2), fibers (3), ribbons

(4), and sheets (5). Their functions include selective transport(5), structural scaffolding (6), mineral templating (4, 7), andpropagation of or protection from pathogenesis (3, 8). Althoughthe molecular structures of the isolated proteins dictate their gov-erning interactions, these functions emerge from the nanoscaleorganization that arises out of self-assembly. Recent theoreticalinvestigations have predicted that assembly pathways duringcrystallization of proteins can deviate from the classical picturein which order arises concomitantly with condensation (9). In-stead, dense-liquid droplets with little long-range order, whicharise through transient fluctuations, have been found to lie alongthe path of least steep ascent over the free energy barrier tonucleation of the ordered phase. After formation of these dense-liquid droplets, relaxation to the lower energy-ordered state oc-curs. Experimental studies of bulk protein crystallization (10, 11)have led to similar conclusions and provide convincing evidencefor a dense-liquid phase that precedes order. But due to experi-mental limitations associated with viewing assembly in threedimensions with molecular resolution, a molecular-scale pictureof multistage pathways has not been obtained. Moreover, in manyprotein systems, folding events and conformational transforma-tions to oligomeric forms are an inherent part of assembly, buttheir role in defining the assembly pathway is largely unexplored.

Among the myriad of self-assembled protein architecturesfound in nature, cell-surface layers (S layers), which form theoutermost cell envelope in many strains of bacteria and archaea(6, 12), present a distinct class. S layers form compact 2D crystal-line arrays out of a single protein or glycoprotein. They typically

overlie a lipid membrane or polymeric cell wall and exhibit crystalsymmetries ranging from p1 to p6. They play a role in all the func-tions described above and can be reconstituted in vitro into 2Darrays both in bulk solutions and at surfaces. Because they exhibitlarge-scale order and a periodicity commensurate with thedimensions of quantum dots and nanotubes, they were amongthe earliest self-assembled protein structures to be exploited asscaffolds for organizing nanostructures via a bottom-up approach(13, 14). However, a clear picture of the assembly mechanism,either in vivo or in vitro, has remained elusive due in part to alack of dynamic structural information (15). Here, we present re-sults from an in situ atomic force microscopy (AFM) investigationrevealing that S-layer assembly in 2D on supported lipid bilayers(SLBs) proceeds along a multistage pathway starting from mono-mers of extended conformation and passing through an amor-phous precursor phase before folding into the final crystallinearray of compact tetramers. Analysis of the assembly kinetics sug-gests that the energetics of tetramer formation drives the systemto follow this complex pathway.

S-Layer Preparation and ImagingWe studied the S-layer SbpA from Lysinibacillus sphaericus(ATCC 4525, MW ≅ 132 kDa). Because SbpA readily forms2D crystals in vitro in the presence of Ca2þ, it has been utilizedin a broad range of applications both in suspensions and onnatural or synthetic surfaces (16–18). In particular, lipid layerscarrying zwitterionic groups such as phosphocholine or phos-phoethylamine have been adopted to support SbpA crystals ona matrix similar to a natural biological membrane (16, 19, 20).At least one previous investigation attempted to follow S-layerSbpA growth by AFM, but Si wafers were used as substratesand neither the process of nucleation nor the molecular-scale de-tails of the growth process were observed (15). In the presentstudy, we used the lipid 1-palmitoyl-2-oleoyl-sn-glycero-3-phos-phocholine (POPC) to make SLBs on bare mica. MonomericSbpA dissolved in pure water was mixed with 10 mM tris(hy-droxymethyl)aminomethane (Tris), pH 7.1, 100 mM NaCl,50 mM CaCl2 and injected into the fluid cell of an AFM.In situ imaging was then employed to investigate the dynamicsof S-layer assembly on the SLB for a range of protein concen-trations (Cp). (See SI Appendix for details.)

Author contributions: S.C., S.-H.S., and J.J.D.Y. designed research; S.C. and S.-H.S.performed research; S.C., S.-H.S., and J.J.D.Y. analyzed data; and S.C., S.-H.S., C.R.B.,and J.J.D.Y. 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.1S.C. and S.-H.S. contributed equally to this work.2To whom correspondence may be addressed. E-mail: jjdeyoreo@lbl.gov or crbertozzi@lbl.gov.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1008280107/-/DCSupplemental.

16536–16541 ∣ PNAS ∣ September 21, 2010 ∣ vol. 107 ∣ no. 38 www.pnas.org/cgi/doi/10.1073/pnas.1008280107

Dow

nloa

ded

by g

uest

on

Aug

ust 2

9, 2

020

Results and AnalysisPathway of S-Layer Assembly. Fig. 1 presents a set of in situ AFMimages and height profiles selected from a time sequence thatreveals multistage assembly comprised of four distinct processes:(i) adsorption of extended monomers onto the SLB, (ii) conden-sation into amorphous or liquid-like clusters, (iii) rearrangementand folding into crystalline arrays of tetramers, and (iv) growth bynew tetramer formation at edge sites of the crystalline clusters.

As Fig. 1A shows, following introduction of protein solution,S-layer monomers and/or small oligomers with an average heightof about 2 nm and an extended conformation began to adsorbonto the SLB. As the adsorbed protein (AP) coverage increased,two distinct morphological changes occurred over the time periodof ∼15–70 min. First, the SLB became densely populated withAPs (height profiles, Fig. 1 B–D). Second, amorphous clusters(ACs) of nearly uniform heights (∼10–12 nm) formed on theSLB (Fig. 1B), and these clusters appeared only after high cover-age on the SLB was established. Initially these clusters did notreveal any internal structure nor change their dimensions signif-icantly over a time period of ∼30–80 min. However, graduallyinternal structure and ordering began to emerge in each of theACs (Fig. 1C). Fig. 1D shows that, after this extended incubationperiod, nearly all of the ACs in Fig. 1C transformed into crystal-line clusters (CCs) within a short period of about 5–10 min.

The height profiles of the ACs and CCs show that, during thecourse of this transformation, not only did the individual lattice

units emerge (labeled with asterisks in the height profile ofFig. 1D), but the CCs became slightly shorter than their amor-phous precursors (9 nm vs. 11 nm). Separate volume measure-ments of ACs and CCs before and after the transition showthat the crystalline phase is about 10% more compact than theamorphous phase (See SI Appendix Figs. 3S-1 and 3S-2). Analysisof cluster size and position vs. time shows there is no correlationbetween either of these characteristics and our ability to resolvethe tetrameric lattice (See SI Appendix Figs. 6S-1 and 6S-2;SI Appendix Table 6S-1). Thus the appearance of order structureis not due to stabilization of cluster position but rather is a resultof the amorphous-to-crystalline transformation.

After the transformation was complete, each CC grew by crea-tion of new tetramers at its edges until nearly all of the APs wereconsumed (Fig. 1 E and F). Fig. 2 follows the transformation andgrowth process for a single cluster and a detailed picture of thecomplete adsorption, condensation, transformation, and tetra-mer-by-tetramer growth stages can be seen in Movie S1. A sche-matic illustrating the sequence of events leading to formation ofcrystalline S layers as deduced from the above observations isgiven in Fig. 3.

Numerous aspects of the growth process are worth addressing.Each CC continued to grow without interruption as long as itwas unhindered by neighboring CCs and there were availableAPs around the cluster. The AP coverage dropped continuallyas the CCs grew until, near the end of the process, most of

Fig. 1. (A–F) Series of in situ AFM images and height profiles showing dynamics of S-layer organization into 2D crystals on SLBs. Height profiles were measuredalong the horizontal black lines in each image and are labeled to denote the heights of the lipid bilayer (LB), adsorbed proteins (APs), amorphous clusters (ACs),and crystalline clusters (CCs). Here, Δt indicates time elapsed since introduction of protein solution (concentration CP ≅ 16 μg∕mL). By Δt ¼ 15 min (A), S-layermonomers had started to adsorb onto the SLBs. During 15 min ≤ Δt ≤ 70 min, the AP coverage increased and by Δt ≅ 70 min (B), not only was the surfacewell covered with APs but ACs of nearly uniform heights had formed. By Δt ≅ 83 min (C), some of the ACs had begun to show the emergence of internalstructure, and by Δt ≅ 87 min (D), nearly all of ACs in C had transformed into CCs with a tetragonal lattice. The height profiles before and after the trans-formation (C vs. D) demonstrate that the ACs were slightly taller than the CCs and reflect the resulting periodic structure of the CCs. Detailed volume measure-ments show that the CCs were also more compact. At later times [e.g., Δt ≅ 108 min (E) and Δt ≅ 220 min (F)], each CC continued to grow by consumingavailable APs near the cluster until growth was physically hindered by neighboring CCs. A highly resolved image from mature CCs (G) revealed the tetramericarrangement and submolecular details (i.e., loop-like structure) of the four S-layer monomers that comprise each lattice unit.

Chung et al. PNAS ∣ September 21, 2010 ∣ vol. 107 ∣ no. 38 ∣ 16537

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Aug

ust 2

9, 2

020

the SLB was devoid of APs. This observation implies that, afterthe initial coverage had been established, adsorption of new pro-tein onto the SLB from the solution was a much slower processthan incorporation of protein into the CCs. Thus growth occurredalmost exclusively through surface transport and attachment.(See SI Appendix Fig. 2S-1 for further verification) We indirectlyverified that new tetramers did not attach to the clusters directly

from solution, but formed from the absorbed monomers on thesurface by using dynamic light scattering (DLS) to monitor as-sembly in bulk solution. The DLS data demonstrate that oligomerformation in solution was much slower than on SLBs (See SIAppendix Fig. 4S-1). In addition, to form crystalline clusters inthe absence of SLBs required solution protein concentrationsthat were 2–3 orders of magnitude larger.

The CCs neither moved nor reorientated relative to one an-other despite the fluidity of the SLB surface (see SI AppendixFigs. 6S-1–6S-3). This was in stark contrast to the APs; sequentialimages (shown in Movie S1), reveal a rapidly fluctuating AP dis-tribution. Given that SLBs can be highly fluidic, we cannot saywhether APs diffuse on the SLB or move with them, but theirmobility is critical for CC growth.

Once nucleation was over, new tetramers appeared exclusivelyat unpopulated lattice sites around the perimeter of the CCs. Notetramers appeared in the surrounding mobile phase of APs. Thefrequency of attachment to the CCs was somewhat higher at kinksites where the number of nearest neighbors was maximized.Consequently, newly formed lattice units tended to completeentire rows rather than appear at random edge sites (Fig. 1E).(See Fig. 2 and Movie S1). In addition, detachment of a tetramerfrom a CC was an extremely rare event. Once a tetramer formed,it remained a permanent part of the crystal. This irreversibility ismost readily seen in Movie S1.

Analysis of Assembly Kinetics. These observations suggest a simplemodel for the growth phase that leads to a specific prediction ofCC size vs. time. The rate of increase in the number of tetramersin a cluster dNT∕dt should be proportional to the number of lat-tice sites around its perimeter (4L∕a), the surface concentrationof adsorbed proteins nðtÞ, and the rate coefficient for tetramerformation β. For a square island, this is

dNT

dt¼

�4La

�·�nðtÞ4

�· β; [1]

where L is the island width and a is the lattice parameter. Eq. 1implies that the CC growth rate should be greater for larger initialcluster size and increase as the clusters grow. But as the APs areconsumed and nðtÞ decreases, the growth rate should approachzero. As shown in SI Appendix Eq. S1-7, nðtÞ can be expressed

Fig. 2. Sequential in situ AFM height images andsurface plots showing the S-layer adsorption,followed by condensation and phase transition ofa single cluster as well as its subsequent growth(CP ≅ 16 μg∕mL). Here, Δt indicates time elapsedsince collection of the image in A. (A) At Δt ¼ 0,monomers or small oligomers in an extended confor-mation have adsorbed to the SLB. (B) By Δt ∼ 62 minthe amorphous nucleus has formed but shows signsof emerging order. Images at C and D show the phasetransition from the amorphous C to crystalline Dstate, at which time each individual lattice unit isclearly discernible. After the transformation [(D) Δt ∼70.3 to (H) 116.8 min), the crystal grows by formationof new lattice units at—and only at—unpopulatedlattice sites along the perimeter of the crystal. For ex-ample, between D and E, newly formed lattice unitscomplete the bottom and third rows of the cluster.Times are (A) 0, (B) 62, (C) 65.8, (D) 70.3, (E) 88.5,(F) 95.5, (G) 102.6, and (H) 116.8 min. (SeeMovie S1 of the growth process.)

Fig. 3. (A) Proposed mechanistic scheme of 2D crystallization of S layer fromsolution on SLBs. Step I: adsorption of S-layer monomers in extended confor-mations on the surface, step II: condensation of the amorphous cluster, stepIII: relaxation to the crystalline nucleus, and step IV: self-catalyzed crystalgrowth. (B) Animated representation of a single crystal growth precededby the amorphous-to-crystal transition (S-layer monomers on SLBs shownin light blue; oligomers and newly formed lattice units in pink). Adsorbedmonomers of extended conformation (1) condense into an amorphous clus-ter (2). This amorphous precursor then restructures to form a crystal of foldedtetramers (dark blue) composed of 5 lattice units (3). The crystalline clusterthen grows by formation of new tetramers preferentially at kink sites (4).

16538 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1008280107 Chung et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 2

9, 2

020

in terms of the average island area hLðtÞ2i vs. time, the surfacedensity of clusters m, and the initial coverage n0, all of which canbe determined from the experiments. Integrating Eq. 1 with theexpression for nðtÞ gives

NTðtÞ ¼ ½βn0f ðtÞ þ N1∕2T;0 �2; [2]

where NT;0 is NTðtÞ at t ¼ 0 and f ðtÞ is a function that approachesa constant at large t and depends on m, L0, and n0. (See SIAppendix Eqs. S1–10 for the detailed expression.) Eq. 2 predictsthat larger initial cluster size should correspond to higher growthrate and that the rate should be stronger than linear at short timesand eventually decline to zero. Because n0 should increase withbulk solution concentration Cp, Eq. 2 also predicts that thegrowth rate will increase with Cp.

Fig. 4A shows the measured dependence of CC size on time fora range of initial cluster sizes at Cp ≅ 16 μg∕mL. Fig. 4B showsthe growth curves for CCs of similar initial sizes at three differentsolution concentrations. The data exhibit all of the featurespredicted by the model as demonstrated by the curves, whichare fits according to Eq. 2.

The fits to the data in Fig. 4 give a value for β∕a2 of2.0 × 10−3 s−1 at Cp ≅ 16 μg∕mL. As Table 1 shows, the valuesof β derived from experiments with other protein concentrationsare similar (See SI Appendix Tables 1S-1–1S-3). To better under-stand this parameter, we first assume that, like proteins in solu-tion or the lipids themselves, the adsorbed proteins move bydiffusive motion in a viscous media. Then β∕a is the productof the diffusive collision rate (D∕d), where D is the diffusivityand d is the typical jump distance of a few water molecules,and a Boltzmann factor expð−EA∕kTÞ in which the activation en-ergy EA is associated with the creation of new tetramers. Fromthis analysis we obtain EA ∼ 51 kJ∕mol (Table 1). (For details,see SI Appendix Tables 1S-1, 1S-2, and 1S-3 and SI AppendixEq. S1-17.) To put this value in context, we compare it to thatexpected from empirical scaling laws that correlate folding timeswith protein size and activation energies (21). Based on that lawthe measured EA is about 50 kJ∕mol lower than would be ex-pected for folding of an isolated S-layer protein and the observedtetramer formation time is smaller by more than 15 orders ofmagnitude. Though we cannot say a priori whether the SLB itselfraises or lowers EA, this reduction in the folding time and activa-tion barrier is reminiscent of that induced by a potent foldingcatalyst, such as the Pro region found as part of α-lytic protease,the extracellular bacterial protease, which lowers the folding en-ergy by ∼76 kJ∕mol (22). The importance of catalysis in tetramer

formation at the edge sites around existing S-layer crystals onSLBs was already clear from the lack of any such events in regionsaway from the crystals. What this comparison to the scaling lawsand the example of α-lytic protease provides is an energeticrationale for the accelerated rate.

DiscussionNonclassical Behavior. Many of the observations reported heredemonstrate the nonclassical nucleation behavior of this system.For example, if tetramer attachment were reversible, there wouldbe a critical size for cluster growth, some clusters would growwhile others disappeared, and growth would begin immediatelyupon condensation of the clusters. None of these phenomenaare observed in this system. Instead, there is a substantial lag timeof tens of minutes between cluster condensation and clustergrowth, with growth beginning only after transformation to thecrystalline state. Moreover, all clusters grow and none disappearregardless of size, tetramers never form anywhere on the surfaceprior to transformation of the clusters to the crystalline state, andthe tetramers do not leave the clusters. In short, the usual as-sumption of rapid exchange between the nuclei and the surround-ing thermodynamic reservoir, which is a prerequisite for applyingclassical nucleation theory, is not valid in this system.

Relationship to Other Crystal Systems. Since ten Wolde andFrenkel (9) first used simulations to predict a two-step processfor protein crystallization, the importance of nonclassical, multi-stage crystallization pathways has become increasingly evident inexperimental studies of proteins (10) and viruses (11), as well asbiominerals (23–26) and other inorganic (27) phases. In the simu-lations (9), critical point fluctuations near a liquid-liquid coexis-tence line led to formation of unstable dense-liquid droplets inwhich crystalline nuclei could eventually form. More recent simu-lations suggest a mix of isotropic non-specific interactions andspecific directional bonds can lead to two-step crystallization(28). Using in situ AFM and optical microscopy, Galkin et al.(10) and Kuznetsov et al. (11) then showed that both proteinand virus crystals could form via a two-step process during bothhomoepitaxy (11) and bulk crystallization (10). Subsequently, thesignificance of amorphous or dense-liquid precursors was ex-tended to inorganic systems, particularly biomineral phases, suchas calcium carbonate and calcium phosphate generated both invivo (24, 27) and in vitro (25), as well as other inorganic (27)phases. Most recently, the existence of stable prenucleation clus-ters (26) in bulk solutions of calcium carbonate has been reported.

Fig. 4. Growth rates of crystalline clusters. (A) The main plot shows growth of individual clusters for a range of initial sizes NT;0 (Cp ≅ 16 μg∕mL). It demon-strates that growth rate for an individual cluster depends on its size at t ¼ 0, with larger clusters exhibiting faster growth rates. t ¼ 0 is time when the crystalphase appeared. Dotted lines give the fits to experimental data where the fitting was based on the model described in the text according to Eq. 2. The legendrelates the data for each curve to the initial number of lattice units in the crystal nuclei. (B) The second plot presents three growth curves and the correspondingtheoretical fits obtained from independent experiments at three different concentrations. Three similar sizes of nuclei were compared to emphasize accel-erated growth rate as the concentration was increased.

Chung et al. PNAS ∣ September 21, 2010 ∣ vol. 107 ∣ no. 38 ∣ 16539

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Aug

ust 2

9, 2

020

The discoveries in our current study are significant in tworegards. They allow us to directly follow at a molecular level themorphological evolution and dynamics that occur during multi-stage crystallization. But more importantly, they reveal thepreviously unrecognized importance of conformational transfor-mation in these nonclassical pathways. The final crystalline stateconsists of folded tetramers, but apparently, to get to the tetra-meric state, monomers require cooperative interactions withclosely packed neighbors. This process occurs only at surface con-centrations sufficient to allow condensation of amorphous clus-ters. As a consequence, the unfolded amorphous precursor doesnot merely provide a potential route to the crystalline state, butmust always precede formation of crystalline nuclei. Once thesenuclei are formed, the precipitation process continues exclusivelythrough tetramer formation out of monomers in an extendedconformation, which evidently requires interaction between themonomers and the crystalline clusters. Hence the crystal servesa catalytic role, lowering the barrier to formation of folded tet-ramers.

ConclusionsMany questions concerning the detailed pathway of S-layer crys-tallization on SLBs remain. First, our results provide little insightinto the adsorption of proteins onto or diffusion over the lipids,nor do they tell us whether the new tetramers form through fourrapid attachments of monomers to a site or through a simulta-neous and cooperative reorganization of four monomers intothe tetrameric state at that site. Also, as discussed above, duringthe amorphous-to-crystalline transition, a reduction in clusterheight corresponding to ∼10% reduction in volume was apparent(see SI Appendix Fig. 3S-2). In a natural system, such heightvariations on membranes sometimes occur through structural re-arrangements. For example, a cytolysin prepore on a membraneextends its transmembrane domain through the bilayer to resultin a functional complex with height reduced by 4 nm (29). In ourcase, the reduced height may be due to enhanced interaction ofthe lipid head group with the crystal (16), or simply tighter pro-tein-protein contacts. However, we cannot exclude the possibilitythat resulting changes in the local charge distribution on the clus-ter surface alter the electrostatic interaction with the AFM tip,which can be enhanced in the presence of divalent ions (30).

A vast number of bacteria and archae have S layers as outermembranes, and these layers overlie lipid or peptidoglycan mem-branes (12). As with all membranes, S layers mediate the inter-actions of these microbes with their surrounding environment.Moreover, there are many examples of pathogenic protein assem-bly on or within membranes (31). Thus, although we have studieda particular protein system that crystallizes on an SLB, the insightit provides into this 2D world is relevant to a wide range of naturalsystems. The results are also likely to be relevant to protein as-sembly in far different contexts. This is because phase diagramsand kinetic pathways for 2D systems are not inherently differentfrom those for 1D or 3D systems. Regardless of dimensionality,proteins often need to transform into an oligomeric form in orderto create the final ordered structure. For example, prions and

beta-amyloids, which crystallize into quasi-1D fibrillar structures,also exhibit folding transformations that produce oligomericgrowth units (3, 8). What is special about the 2D nature of theS-layer system is that its ability to assemble on a lipid surface ren-ders it amenable to investigation using an in situ tool such asAFM that provides molecular resolution of the assembly process.

Despite the broad relevance of the S-layer-on-SLB system toprotein assembly, the connections between the results presentedhere and the in vivo mechanism of S-layer assembly, its fate dur-ing cell division, and the symmetry of the cell surface (29, 32, 33)is unclear. But given the central role of the S layer in cellular func-tions ranging from environmental mineralization (7) to antibioticresistance (12), as well as its potential for nanomaterials organi-zation (13), developing an in vitro platform in which assembly canbe manipulated is an important goal. Understanding the detailedphysical mechanism of assembly is a significant step towardthat goal.

Materials and MethodsProtein Purification. The protein was purified as described with minor mod-ifications (16). Lysinibacillus sphaericus (ATCC 4525) was grown at 32 °C inSVIII medium (50 mM Hepes, pH 7.2, 7 mM K2HPO4, 10 g∕L peptone,5 g∕L yeast extract, 5 g∕L meat extract, 0.2 mM MgSO4, 1.8 mM sucrose,17 mM glucose) shaking at 250 rpm. The cells were suspended in cold50 mM Tris, pH 7.2 and sonicated (Misonix Sonicator 3000) in ice. The lysateswere centrifuged at 16;000 × g for 15 min. The separated cell wall in thewhite portion on the pellet was transferred and resuspended in 0.75% TritonX-100 (Thermo Scientific), 50mMTris, pH 7.2 for 10min at room temperature.After the centrifugation at 28;000 × g for 10 min, the cell walls were washedthree more times with the same buffer. The unfolded protein was releasedwith 50 mM Tris, pH 7.2, 5 M guanidine hydrochloride. The undissolved cellwalls were removed by centrifugation at 100;000 × g for 45 min. Thesupernatant was dialyzed against double distilled water at 4 °C overnight,and insoluble aggregates were removed by centrifugation at 100;000 × gfor 30 min.

Preparation of Small Unilamellar Vesicles and SLBs on Mica. The lipid POPC inchloroform (Avanti Polar Lipids Inc., #850457C) was dried under vacuum over-night. The dried lipid was hydrated with water to 2 mg∕mL by vortexing thesuspension to maximize the hydration. The suspended lipid was allowed toequilibrate for several hours. The lipid solution was passed through 100-nmpolycarbonate membranes (Avanti Polar Lipids Inc., #610005) more than20 times and mixed with the same amount of PBS buffer. The solution con-taining unilamellar vesicles (1 mg∕mL) was placed on fleshly cleaved mica for10 min. SLBs on mica were carefully washed in degassed double distilledwater and kept in 10 mM Tris pH 7.1, 100 mM NaCl, 50 mM CaCl2 at 4 °C.

In Situ AFM. In situ force microscopy was performed using an AFM fluid cellwithin a multimode AFM (Nanoscope V controller, Veeco Metrology, Inc.,Santa Barbara, CA) equipped with a liquid-resistant, vertical engagement160-μm “JV” scanner. The AFM probe consisted of a sharp silicon tip on asilicon nitride cantilever (HYDRA probe, length: 200 μm, spring constantk ¼ ca. 0.035 N∕m, average tip diameter ≤15 nm, AppNano, Santa Clara,CA). Freshly peeled Mica discs (Grade V-1 Muscovite, Structural Probe, Inc)glued on metal disks using 20-min epoxy glue were used as substrates forSLBs of POPC. For typical imaging conditions, AFM images were collectedat scan frequencies of 1–8 Hz while applying a minimum loading force of∼150 pN or less using optimized feedback and setpoint parameters for stableimaging conditions. Time evolution in situ AFM images were analyzed to de-termine the average height and domain size of ACs and CCs by using imageprocessing and statistical analysis software such as SPIP (Scanning Probe Im-age Processor, NanoScience Instruments, Inc.) and IGOR Pro (WaveMetrics,Inc.) with custom programmed IGOR scripts. Three-dimensional perspectivesurface plots were generated with Image J software (http://rsbweb.nih.gov/ij/index.html).

ACKNOWLEDGMENTS. We gratefully acknowledge the assistance of BabakSanii in fabricating supported lipid bilayers and Julie Norville for providingadvice on protein purification. This work was performed at the MolecularFoundry, Lawrence Berkeley National Laboratory, with support from theOffice of Science, Office of Basic Energy Sciences, of the U.S. Departmentof Energy under Contract DE-AC02-05CH11231.

Table 1. Energy barrier (EA) associated with the creation andattachment of new tetramers at different protein concentration(Cp)

Cp, μg∕mL Ave. β∕a2, ×10−3 s−1 Ave. EA*, kJ∕mol

16 2.0 5123 2.2 5031 3.0 48

*The slight variations are likely due to either the tetramer formation ratecoefficient having a slight dependence on coverage or errors in ignoringcluster-cluster competition and/or monomer deposition from the bulkduring growth.

16540 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1008280107 Chung et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 2

9, 2

020

1. Mann S (2008) Life as a nanoscale phenomenon. Angew Chem Int Edit47:5306–5320.

2. Johnson JE (2008) Multi-disciplinary studies of viruses: The role of structure in shapingthe questions and answers. J Struct Biol 163:246–253.

3. Rambaran RN, Serpell LC (2008) Amyloid fibrils: Abnormal protein assembly. Prion2:112–117.

4. Du C, Falini G, Fermani S, Abbott C, Moradian-Oldak J (2005) Supramolecular assemblyof amelogenin nanospheres into birefringent microribbons. Science 307:1450–1454.

5. Tanaka S, et al. (2008) Atomic-level models of the bacterial carboxysome shell. Science319:1083–1086.

6. Engelhardt H (2007) Are S-layers exoskeletons? The basic function of protein surfacelayers revisited. J Struct Biol 160:115–124.

7. Schultzelam S, Beveridge TJ (1994) Nucleation of celestite and strontianite on acyanobacterial S-layer. Appl Environ Microb 60:447–453.

8. Cherny I, Gazit E (2008) Amyloids: Not only pathological agents but also orderednanomaterials. Angew Chem Int Edit 47:4062–4069.

9. ten Wolde PR, Frenkel D (1997) Enhancement of protein crystal nucleation by criticaldensity fluctuations. Science 277:1975–1978.

10. Galkin O, Chen K, Nagel RL, Hirsch RE, Vekilov PG (2002) Liquid-liquid separation insolutions of normal and sickle cell hemoglobin. Proc Natl Acad Sci USA 99:8479–8483.

11. Kuznetsov YG, Malkin AJ, McPherson A (1998) Atomic-force-microscopy studies ofphase separations in macromolecular systems. Phys Rev B 58:6097–6103.

12. Sleytr UB, Messner P, Pum D, Sara M (1999) Crystalline bacterial cell surface layers(S layers): From supramolecular cell structure to biomimetics and nanotechnology.Angew Chem Int Edit 38:1035–1054.

13. Shenton W, Pum D, Sleytr UB, Mann S (1997) Synthesis of cadmium sulphide super-lattices using self-assembled bacterial S-layers. Nature 389:585–587.

14. Moll D, et al. (2002) S-layer-streptavidin fusion proteins as template for nanopat-terned molecular arrays. Proc Natl Acad Sci USA 99:14646–14651.

15. Gyorvary ES, Stein O, Pum D, Sleytr UB (2003) Self-assembly and recrystallization ofbacterial S-layer proteins at silicon supports imaged in real time by atomic forcemicroscopy. J Microsc (Oxford) 212:300–306.

16. Norville JE, Kelly DF, Knight TF, Belcher AM, Walz T (2007) 7 angstrom projection mapof the S-layer protein SbpA obtained with trehalose-embedded monolayer crystals.J Struct Biol 160:313–323.

17. Ilk N, et al. (2002) Molecular characterization of the S-layer gene, SbpA, of Bacillussphaericus CCM 2177 and production of a functional S-layer fusion protein with

the ability to recrystallize in a defined orientationwhile presenting the fused allergen.Appl Environ Microb 68:3251–3260.

18. Wetzer B, Pum D, Sleytr UB (1997) S-layer stabilized solid supported lipid bilayers.J Struct Biol 119:123–128.

19. WeygandM, et al. (2002) Structural reorganization of phospholipid headgroups uponrecrystallization of an S-layer lattice. J Phys Chem B 106:5793–5799.

20. Gyorvary E, et al. (1999) Lateral diffusion of lipids in silane-, dextran-, and S-layer-supported mono- and bilayers. Langmuir 15:1337–1347.

21. NaganathanAN,Munoz V (2005) Scaling of folding times with protein size. J Am ChemSoc 127:480–481.

22. Sauter NK, Mau T, Rader SD, Agard DA (1998) Structure of alpha-lytic proteasecomplexed with its pro region. Nat Struct Biol 5:945–950.

23. Politi Y, Arad T, Klein E, Weiner S, Addadi L (2004) Sea urchin spine calcite forms via atransient amorphous calcium carbonate phase. Science 306:1161–1164.

24. Tsuji T, Onuma K, Yamamoto A, Iijima M, Shiba K (2008) Direct transformationfrom amorphous to crystalline calcium phosphate facilitated by motif-programmedartificial proteins. Proc Natl Acad Sci USA 105:16866–16870.

25. Lee JRI, et al. (2007) Structural development of mercaptophenol self-assembledmonolayers and the overlying mineral phase during templated CaCO3 crystallizationfrom a transient amorphous film. J Am Chem Soc 129:10370–10381.

26. Gebauer D, Völkel A, Cölfen H (2008) Stable prenucleation calcium carbonate clusters.Science 322:1819–1822.

27. Chen XB, Samia ACS, Lou YB, Burda C (2005) Investigation of the crystallization processin 2 nm CdSe quantum dots. J Am Chem Soc 127:4372–4375.

28. Whitelam S (2010) Control of pathways and yield of protein crystallization throughthe interplay of nonspecific and specific attractions. Phys Rev Lett 105:088102.

29. Howard LV, Dalton DD, Mccoubrey WK (1982) Expansion of the tetragonally arrayedcell-wall protein layer during growth of Bacillus sphaericus. J Bacteriol 149:748–757.

30. Muller DJ, Engel A (1997) The height of biomolecules measured with the atomic forcemicroscope depends on electrostatic interactions. Biophys J 73:1633–1644.

31. Brundin P, Melki R, Kopito R (2010) Prion-like transmission of protein aggregates inneurodegenerative diseases. Nat Rev Mol Cell Bio 11:301–307.

32. Sleytr UB, Glauert AM (1976) Ultrastructure of cell-walls of 2 closely related clostridiathat possess different regular arrays of surface subunits. J Bacteriol 126:869–882.

33. Pum D, Messner P, Sleytr UB (1991) Role of the S-layer in morphogenesis and cell-divi-sion of the archaebacteriumMethanocorpusculum sinense. J Bacteriol 173:6865–6873.

Chung et al. PNAS ∣ September 21, 2010 ∣ vol. 107 ∣ no. 38 ∣ 16541

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Aug

ust 2

9, 2

020