Chaperonin GroEL accelerates protofibril formationand decorates fibrils of the Het-s prion proteinMarielle A. Wältia, Thomas Schmidta, Dylan T. Murraya, Huaibin Wangb, Jenny E. Hinshawb, and G. Marius Clorea,1

aLaboratory of Chemical Physics, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, MD 20892-0520;and bLaboratory of Cell and Molecular Biology, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda,MD 20892-0820

Contributed by G. Marius Clore, July 13, 2017 (sent for review June 30, 2017; reviewed by James G. Omichinski and David J. Weber)

We have studied the interaction of the prototypical chaperoninGroEL with the prion domain of the Het-s protein using solutionand solid-state NMR, electron and atomic force microscopies, andEPR. While GroEL accelerates Het-s protofibril formation by severalorders of magnitude, the rate of appearance of fibrils is reduced.GroEL remains bound to Het-s throughout the aggregation processand densely decorates the fibrils at a regular spacing of ∼200 Å.GroEL binds to the Het-s fibrils via its apical domain located at thetop of the large open ring. Thus, apo GroEL and bullet-shapedGroEL/GroES complexes in which only a single ring is capped byGroES interact with the Het-s fibrils; no evidence is seen for anyinteraction with football-shaped GroEL/GroES complexes in whichboth rings are capped by GroES. EPR spectroscopy shows that rota-tional motion of a nitroxide spin label, placed at the N-terminal endof the first β-strand of Het-s fibrils, is significantly reduced in bothHet-s/GroEL aggregates and Het-s fibrils, but virtually completelyeliminated in Het-s/GroEL fibrils, suggesting that in the latter, GroELmay come into close proximity to the nitroxide label. Solid-stateNMR measurements indicate that GroEL binds to the mobile regionsof the Het-s fibril comprising the N-terminal tail and a loop connect-ing β-strands 4 and 5, consistent with interactions involving GroELbinding consensus sequences located therein.

amyloid–chaperone interactions | NMR | EPR | electron microscopy |atomic force microscopy

The interaction of amyloids with chaperones, a group of pro-teins responsible for maintaining protein homeostasis (1–3),

has attracted considerable attention with regard to the etiol-ogy of amyloidoses and accumulation of amyloid plaques (4).Diseases associated with amyloidosis include type II diabetesand a number of neurodegenerative processes such as Alz-heimer’s disease, Huntington disease, and spongiform en-cephalopathies (5).Amyloids are unbranched, highly ordered protein aggregates

that contain a cross β-structure arranged perpendicular to thefibril axis (6). Amyloid fibrils, which can be readily identified byelectron microscopy, are formed by an ordered array of manycopies of the given amyloid protein. In addition to amyloids thatlead to pathology, there are amyloids where the fibrils contributea distinct biological function (7). An example of the latter isprovided by the Het-s prion protein from the filamentous fungusPodospora anserina (Fig. 1A) where Het-s fibril formation servesto prevent exchange of cytoplasmic material between geneticallydissimilar species (5, 7). Since the structure of the Het-s fibril iswell established (Fig. 1B) and not polymorphic at neutral pH (8,9), and the Het-s monomer is stable for several days at 4 °C, wehave used Het-s to study the interaction of a model amyloido-genic protein with a prototypical chaperone, GroEL, at themolecular and atomic levels.GroEL is a member of the Hsp60 chaperonin class of chap-

erones, characterized by two cylindrically stacked heptamericring structures, each enclosing a large interior cavity (or foldingchamber) that encapsulates substrate proteins (Fig. 1C) (10).Although GroEL is a bacterial protein, Hsp60 chaperonins are

structurally highly conserved throughout prokaryotes and eu-karyotes: for example, Escherichia coli GroEL and human mi-tochondrial Hsp60 are 55% sequence identical (11). In eukaryotes,Hsp60s exist not only intracellularly where they play a key role inprotein homeostasis, but also extracellularly where they function aspotent stimulators of the immune response (12). Each GroELsubunit comprises three domains: equatorial, intermediate, andapical (10). The latter forms the rim around the entrance to thecavity and is responsible for binding both the cochaperone GroESand protein substrates.Studying amyloid–chaperone interactions and their impact on

protofibril and fibril formation is challenging because the speciespresent during the early stages of fibril formation are heteroge-neous. Recent work has shown that the formation of toxic am-yloid species can be inhibited in a diverse manner by a range ofchaperones (2, 13–21). This may involve inhibition of aggregationand protofibril formation (22), disaggregation (23, 24), or target-ing misfolded proteins for rapid clearance (25). Biophysical data,however, on the interaction of chaperones with amyloid proto-fibrils and fibrils are relatively sparse, largely relying on fluores-cence measurements tracking fibril formation.Here, we make use of solution and solid-state NMR spectros-

copy, electron (EM) and atomic force (AFM) microscopies, andelectron paramagnetic resonance (EPR) spectroscopy to study theinteraction of the prion forming domain (residues 218–289) of Het-swith GroEL at atomic resolution during the course of the completeaggregation pathway from monomer to complete fibril formation.We show that GroEL accelerates the formation of Het-s aggregates/protofibrils by orders of magnitude, eventually leading to the for-mation of fibrils densely decorated at regular intervals with GroEL.The interaction, which is already present at the protofibril stage,

Significance

The interaction of amyloids with chaperones, a large group ofproteins responsible for proteostasis, is thought to play a sig-nificant role in the etiology of amyloidosis. Here, we study theinteraction of the model chaperonin GroEL with a model am-yloid protein, the prion domain of Het-s, by using NMR and EPRspectroscopies and electron and atomic force microscopies. Weshow that GroEL accelerates protofibril formation, eventuallyleading to the formation of fibrils densely decorated by GroEL.This type of chaperone–amyloid interaction may serve to re-duce the toxicity of amyloidogenic oligomers and target thefully formed fibrils for rapid elimination by facilitating invivo clearance.

Author contributions: M.A.W., T.S., D.T.M., J.E.H., and G.M.C. designed research; M.A.W.,T.S., D.T.M., and H.W. performed research; M.A.W., T.S., D.T.M., J.E.H., and G.M.C. ana-lyzed data; and M.A.W. and G.M.C. wrote the paper.

Reviewers: J.G.O., Université de Montréal; and D.J.W., University of Maryland Schoolof Medicine.

The authors declare no conflict of interest.1To whom correspondence should be addressed. Email: mariusc@intra.niddk.nih.gov.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1711645114/-/DCSupplemental.

9104–9109 | PNAS | August 22, 2017 | vol. 114 | no. 34 www.pnas.org/cgi/doi/10.1073/pnas.1711645114

occurs via the apical domain of GroEL and involves the mobileregions of Het-s.

Results and DiscussionInitial Aggregation. Addition of unlabeled GroEL (at naturalisotopic abundance) to 1H/15N-labeled Het-s(218–289) monomerresults in an immediate (within the first 5 min), GroEL concentration-dependent decrease in cross-peak intensities in the 1H-15N correlationspectrum of monomeric Het-s (Fig. 2 A and B), accompanied by theappearance of protein aggregates or protofibrils as seen by both EM(Fig. 2 C andD and Fig. S1) and AFM (Fig. S1). At a ratio of 4 Het-smonomers per GroEL 14-mer (corresponding to the sample at100 μM Het-s and 370 μM subunits of GroEL) only ∼4–5% Het-smonomer remains visible by NMR (Fig. 2B). Further, the con-centration of GroEL required to reduce the fraction of NMR-visible Het-s monomer by 50% (i.e., 50% aggregation) is ∼5 μMGroEL 14-mer (∼70 μM in subunits) (Fig. 2B). SDS/PAGE in-dicates that the aggregates comprise both Het-s and GroEL (Fig.S2). In the absence of GroEL, Het-s remains monomeric for atleast a day at room temperature with no EM evidence ofprotofibril formation.

Progression of Aggregation. We followed the disappearance of15N-labeled Het-s monomer as a function of time by solutionNMR, monitoring overall signal intensity from the first incrementof a 1H-15N correlation spectrum (Fig. 3A). Subsequent to theinitial rapid decrease in monomer concentration occurring withinthe first 5 min after addition of GroEL as a result of protofibrilformation (cf. Fig. 2 and Figs. S1 and S3B), there is a slow decreasein monomer concentration with a half-life of ∼2 d at GroEL con-centrations of 10 and 50 μM (in subunits) and ∼2.7 d at 100 μM (insubunits) GroEL concentration. In the absence of GroEL, the de-crease in Het-s monomer concentration is approximately mono-exponential with a half-life of approximately, 2.5 d (Fig. 3A). After11–14 d at room temperature under quiescent conditions (i.e., notshaken or stirred), both protofibrils and fibrils are clearly seen in theEM and AFM images of the Het-s/GroEL samples (Fig. 3B andFig. S3D); in the absence of GroEL only fibrils are seen by EM andAFM under the same conditions (Fig. 3C and Fig. S3A).

Characterization of Het-s/GroEL Fibrils by Electron Tomography.Electron tomograms taken 3 wk after addition of GroEL toHet-s reveals the presence of both multifilament fibril bundles(Fig. 4) and unifilament fibrils (Fig. 5) decorated in a regular,

MK I D A I VGRN S A K D I R T E E R A R VQ L G

N V V AT A A L HGG I R I S DQ T T N S V E

T V VGK GE S R V L I GN E YGGKG FWDN

PHHHH

P HHHH P HHH

Fibril axis

220 230 240

250 260

270 280

Het-s(218-289) GroEL

45 Å10 Å

9.4 Å

25 Å

β1 β2 β3

β4 β5

β6 β7 β8

A

B

C

Fig. 1. Het-s(218–289) and GroEL. (A) Primary and secondary structure of Het-s(218–289) fibrils. Sequences within the flexible tails and loop are colored inorange, and the sites of GroEL-binding consensus sequences (a polar residue, P, followed by four hydrophobic residues, H) are indicated in green. (Note theN-terminal methionine is not part of the natural Het-s sequence.) (B) Structure of the Het-s(218–289) fibril determined by solid-state NMR (PDB ID code 2KJ3; ref.9). (C) Structure of apo GroEL (PDB ID code 1XCK; ref. 41) showing a single heptameric ring viewed orthogonal to the long axis of the cavity. The structures inB and C are drawn to scale. Four to five Het-s(218–289) termini can potentially bind within each GroEL cavity.

9 8 7

110

115

120

125

7

BA

C

500 nm 100 nm

1H (ppm) 1H (ppm)

15N (ppm)

[GroEL] (μM)0

50

370100

10

789

[GroEL] (μM) 0

100

W287L276

A247R255 A228

N289

E280N226

S273

G224

G282G252G285

G283G242

H290F286/I254

% N

MR

vis

ible

mon

omer

[GroEL]total (μM)

100

80

60

40

20

0100 300

Fig. 2. GroEL concentration-dependent aggregation of Het-s(218–289) attime point zero. (A) 1H-15N HSQC spectra of 100 μM Het-s(218–289) alone(blue) and immediately after addition of 100 μM (in subunits) GroEL (red).(B) First increment of 1H-15N HSQC spectra acquired on 100 μM Het-s(218–289)in the presence of 0–370 μM (in subunits) GroEL at time point zero (i.e.,recorded immediately after addition of GroEL). Inset shows a plot of thefractional decrease in NMR visible monomer as a function of total GroELconcentration (in subunits). (C) Electron micrographs of the Het-s aggregates(protofibrils, 100 μM in monomer units) formed immediately after addition of100 μM (in subunits) GroEL at room temperature. In the same time frame, noaggregates/protofibrils are observed in the absence of GroEL addition.

Wälti et al. PNAS | August 22, 2017 | vol. 114 | no. 34 | 9105

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

densely packed manner by GroEL with a center-to-center spac-ing of ∼200 Å between adjacent bound GroEL molecules.GroEL binds on the lower (Fig. 4, Left), side (Fig. 4, Center) andupper (Fig. 4, Right) faces of the fibril bundles. The observationthat GroEL is present on both sides of a unifilament fibril sug-gests that the filament twists in a helical manner about the longaxis of the fibril (Fig. 5). Similar images are also observed whenGroEL is added to preformed Het-s fibrils (Fig. 6 and see Fig.S5). Further magnification (Fig. 5) also reveals that GroELbinding to Het-s fibrils occurs exclusively via the open ring faceof GroEL comprising a heptameric ring of apical domains (10).The binding site on GroEL was confirmed by addition of

GroEL/GroES complexes to preformed Het-s fibrils (Fig. 6A).GroES binds to the apical domain of GroEL, thereby cappingeither one or both GroEL rings (depending on conditions; seeExperimental Methods) to form bullet-shaped (Fig. S4A) orfootball-shaped (Fig. S4B) complexes, respectively. In theresulting electron micrographs, the majority of GroEL particlesbound to the Het-s fibrils is in the uncapped form (Fig. 6B, redboxes). Bullet-shaped GroEL/GroES complexes, however, arealso clearly seen bound to the fibrils (Fig. 6B, blue boxes), butwe see no evidence of any bound, football-shaped GroEL/GroES complexes (Fig. 6B, black boxes, and Fig. S5). Thus,binding of GroEL to Het-s fibrils requires the presence of atleast one uncapped ring.

Characterization of Het-s Fibrils by Continuous Wave EPR. Continu-ous wave (CW) EPR spectroscopy provides a straightforwardapproach for probing the internal mobility of a nitroxide label[such as MTSL; (1-oxyl-2,2,5,5-tetramethyl-D3-pyrroline-3-methyl)methanethiosulfonate)]. We therefore MTSL-labeled Het-s(218–289) at position 227, located at the N terminus of strand β1, closeto the disordered N-terminal tail (Fig. 1A), via an engineered cys-teine mutation (S227C).The lineshape of the EPR spectrum is sensitive to the mobility

of the nitroxide label (26). At X-band, the EPR spectrum isdominated by the electron-nuclear hyperfine splitting tensor A.Rapid isotropic reorientation of the spin labels reduces the Atensor to a scalar value aiso = 1/3(Ax + Ay + Az), resulting in athree-line EPR spectrum with the two outer components sepa-rated by 2aiso; as rotational motion of the spin label is sloweddown (or motion of the label hindered), the EPR lineshapechanges, eventually reaching the powder limit in which the twoouter components are separated by 2Azz (27).The CW X-band EPR spectrum of the nitroxide labeled Het-

s(S227C) monomer is well described by an effective correlationtime (τeff) of ∼0.5 ns (Fig. 7A). The EPR spectrum of Het-s fi-brils formed in the absence of GroEL (Fig. 7B) and Het-s pro-tofibrils formed immediately upon addition of GroEL (Fig. 7C)are characterized by a two-species system comprising a slowcomponent with τeff ∼ 3.3 ns corresponding to the fibrils and

100 nm500 nm

B C

100 nm

Het-s aloneHet-s + GroELA

Time (days)5 10 15

rela

tive

in

ten

sity

(%

)

20

20

40

60

80

00

100[GroEL] (μM)

01050

100370

Fig. 3. Time dependence of Het-s(218–289) aggregation following addition of GroEL. (A) Disappearance of 100 μM Het-s(218–289) monomer over time as afunction of GroEL concentration (specified in subunits), measured by solution-state NMR. (B) Electron micrographs of GroEL-induced Het-s protofibrils (Left)and fibrils (arrow, Right) obtained 11 d following the addition of 370 μM (in subunits) GroEL to 100 μM monomeric Het-s(218–289) at room temperatureunder quiescent conditions. (C) Het-s(218–289) fibrils obtained with 100 μM Het-s(218–289) alone at room temperature after 11 d under quiescent conditions.

50 nm 50 nm 50 nm

Bottom slice Middle slice Top Slice

Fig. 4. Negative stain electron tomogram images of Het-s fibrils 3 wk after addition of 100 μM (in subunits) GroEL to 100 μM monomeric Het-s(218–289) atroom temperature under quiescent conditions. All three slices clearly show that GroEL binds via its apical domain to the Het-s(218–289) fibrils. Insets show azoom of a region of interest (black box).

9106 | www.pnas.org/cgi/doi/10.1073/pnas.1711645114 Wälti et al.

protofibrils, and a fast component τeff ∼ 0.5 ns corresponding toresidual Het-s monomer present in the sample. The reduction inmobility of the nitroxide label in the fibrils and protofibrils canbe attributed to the transition from a random coil to an orderedβ-sheet structure containing the label. Finally, the EPR spectrum ofthe fibrils of Het-s formed in the presence of GroEL can be wellreproduced by a three-species system comprising a powder (τeff >200 ns or severely restricted motion with a high order parameter)characterized by the 2Azz splitting (Fig. 7D), in addition to the twofaster components seen in the Het-s fibrils alone and the Het-s/GroEL protofibrils. We attribute the powder species of the EPRspectrum to the immobilization of the nitroxide label attached toHet-s through direct interaction with GroEL, presumably binding tothe GroEL-binding consensus sequences (28) located either in theN-terminal tail or the β4/β5 loop of Het-s fibrils (Fig. 1A). Many ofthe Het-s fibrils bound to GroEL consist of a multifilament bundle.Since GroEL binds to the outside of the fibrils, the nitroxide labels ofonly Het-s molecules in the outer filaments of the bundle can be fullyimmobilized by GroEL, accounting for the roughly equal populationof the τeff ∼ 3.3 and >200 ns species in the EPR spectrum.

Solid-State NMR of Het-s Fibrils in the Presence of GroEL. To furthercharacterize the interaction of GroEL with Het-s fibrils from theperspective of Het-s, we made use of solid-state NMR spectroscopy

on uniformly 15N/13C-labeled Het-s fibers grown in the absence andpresence of GroEL (Fig. 8 and Figs. S6 and S7).We first recorded a set of 1H-13C cross-polarization (CP)–based

magic angle spinning (MAS) spectra, both 1D 13C (Fig. 8A) and 2D13C/13C with dipolar assisted rotational resonance (29) transferbetween 13C nuclei (Fig. S6). Under CP-MAS conditions, only rigidsegments of the Het-s fibrils are probed (30). This is because mo-lecular motions in dynamic regions prevent the dipolar coupling-based magnetization transfer used in CP-MAS–based experiments.The CP-MAS spectra of Het-s fibrils in the presence and absence ofGroEL overlap perfectly, indicating that GroEL does not perturbthe structure of the rigid fibril core.Next, we recorded a set of 1D and 2D 1H/13CMAS spectra based

on 1H-13C INEPT (insensitive nuclei enhanced by polarizationtransfer) transfers (Fig. 8 B and C) to probe the mobile regions ofthe Het-s fibrils (30), specifically, the N- and C-terminal tails andthe long loop connecting β-strands 4 and 5. In INEPT-based ex-periments, magnetization is transferred from 1H to 13C via 1JCHscalar couplings, and 1H and 13C transverse relaxation rates mustbe longer than the time required for 1H-13C polarization transferto observe a signal. In rigid segments of the protein with slow oranisotropic reorientations, the 1H-1H and 1H-13C dipolarcouplings are not averaged to zero over time; in mobile re-gions, however, the dipolar interactions are time-averaged tozero, thereby eliminating the main mechanism of signal lossduring INEPT transfer. Comparison of both the 1D (Fig. 8B)and 2D (Fig. 8C) 1H/13C INEPT spectra shows that the pres-ence of GroEL results in significant differences. First, thereis an obvious decrease in signal intensity in the 1D 1H/13CINEPT spectrum of Het-s fibrils with GroEL (Fig. 8B), accom-panied by a corresponding decrease in the aliphatic 1H and 13Ctransverse relaxation times [from 24.7 ± 0.1 to 16.8 ± 0.1 ms,and from 4.8 ± 0.1 to 3.9 ± 0.2 ms, respectively (Fig. S7)]. Thus,one can conclude that GroEL binds to mobile and accessiblesegments of the Het-s fibrils. Further analysis of the 2D 1H/13CINEPT spectra reveals a few significant chemical shift differ-ences as well, and the shifted cross-peaks can be assigned residue/atom type probabilities (31) (Fig. 8C). The latter are consistentwith residues either in or close to the GroEL consensus-bindingsequences (28) located in the N-terminal tail or β4/β5 loop(Fig. 8C).

50 nm

Fig. 5. Regular spacing of GroEL bound via its apical domain to a Het-smonofilament. Shown is the middle slice of a negative stain electron to-mogram image of a Het-s fibril formed 3 wk after addition of 100 μM (insubunits) GroEL to 100 μM Het-s(218–289) at room temperature under qui-escent conditions. Also seen are three unbound GroEL molecules.

A B

50 nm 20 nm

Het-s fibrils Het-s fibrils with GroEL 20 nm

Fig. 6. Negative stain electron micrographs of Het-s fibrils formed in the absence (A) and presence (B) of GroEL and GroES. Most of the GroEL bound to Het-sis in the uncapped form (red box); however, some bullet-shaped GroEL/GroES complexes are also bound to the Het-s fibrils (blue box). Right images show acollection of enlarged micrographs of differently prepared samples where uncapped bound GroEL (first and second rows, red), bullet-shaped bound GroEL/GroEScomplexes (third row, blue), and unbound football-shaped GroEL/GroES complexes (fourth row, black) are found. One hundred micromolar Het-s(218–289) wasfibrilized for 3 wk at room temperature under quiescent conditions. Bullet- or football-shaped GroEL/GroES complexes (prepared as described in ExperimentalMethods) were mixed separately with Het-s fibrils leading to micrographs shown in the blue or black boxes (B, Right), respectively (see also Fig. S5).

Wälti et al. PNAS | August 22, 2017 | vol. 114 | no. 34 | 9107

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Concluding RemarksUsing a range of biophysical and structural techniques, we haveshown that the chaperonin GroEL interacts directly with theHet-s prion protein, resulting in a two order of magnitude orlarger speed up of aggregation and protofibril formation. Therate of appearance of fully formed fibrils, however, is sloweddown in the presence of GroEL. Thus, after 11 d, both protofibilsand fibrils of Het-s are seen in the presence of GroEL (Fig. 3B),while only fibrils are observed for Het-s alone (Fig. 3C). After3 wk, however, the Het-s protofibrils obtained in the presence ofGroEL have largely been converted to fibrils. GroEL remainsbound to Het-s throughout the aggregation process and deco-rates Het-s fibrils with a regular spacing of ∼200 Å. Given thediameter of the GroEL cavity, we surmise that GroEL interactswith four to five Het-s units within a fibril, thereby resulting inhigh avidity.The interaction of GroEL with amyloid-type fibrils, in this

instance Het-s, requires that mobile regions within the fibril(containing one or more GroEL consensus binding sequences)be able to access the inner rim of the central cavity of GroELformed by the apical domain. Thus, amyloid fibrils without theseparticular characteristics would not be expected to interact withGroEL in the same manner as Het-s. For example, while GroELbinds to amyloid-β monomers via two GroEL-binding consensussequences (32), these sequences are largely located in orderedβ-sheet regions of the fibril (33–35). Hence, one would predictthat interaction between GroEL and amyloid-β fibrils would onlyoccur upon disruption of the fibril.In the context of the etiology of amyloidosis, the type of in-

teraction seen here between GroEL and Het-s protofibrils andfibrils may serve two functions. First, chaperone binding to toxicamyloid oligomers may render these nontoxic by sequestration

into large aggregates, as has been demonstrated in the case ofthe interaction between the small heat shock protein HspB1and amyloid-β protofibrils (36). Second, binding of chaperonesto fully formed fibrils may facilitate in vivo clearance (e.g.,in the liver) as was shown for the extracellular chaperoneclusterin (25).

Experimental MethodsExpression and Purification. Details of the expression and purification ofGroEL, GroES, and Het-s(218–289) are provided in SI Experimental Methods.GroEL/GroES bullet-shaped and football-shaped complexes in which one orboth GroEL rings are capped, respectively, were prepared essentially as de-scribed with some minor modifications (37). Fibrillization of Het-s(218–289)was carried out as described (38).

NMR and EPR Spectroscopy. Solution NMR measurements were carried out ona Bruker 600 MHz NMR spectrometer with a triple resonance z axis gradientcryoprobe at 10 °C. Further details are provided in SI Experimental Methods.Solid-state NMR was carried out on a 17.5 T spectrometer (1H frequency of746 MHz) using a Chemmagnetics console and a Black Fox magic angle spinningprobe with 12 kHz spinning. Details of the 1H-13C CP and INEPT experiments

Fibril axis

Fibril axis

100 50 0150

Het-s fibrils aloneHet-s fibrils + GroEL

100 50 0150200250 100 50 0150

[1H, 13C] CP-MAS [1H, 13C] INEPTA B

13C (ppm) 13C (ppm) 13C (ppm)

2030405060

1.0

2.0

3.0

4.0

13C (ppm)

1H (ppm)

[GroEL] (μM)0

100

10

Iδ

Vγ

Aβ

MK I D A I VGRN S A K D I R T E E R A R VQ L G

N V V AT A A L HGG I R I S DQ T T N S V E

T V VGK GE S R V L I GN E YGGKG FWDN

PHHHH

P HHHHP HHH

220 230 240

250 260

270 280

β1 β2 β3

β4 β5

β6 β7 β8

Nβ (47%)Yβ (26%)Fβ (19%)Dβ (7%)

SβSα (38%)Iα (12%)Tα(12%)

Kβ (37%)Mβ (31%)Rβ (20%)

Kγ (55%)Lγ (38%)

2D [1H, 13C] INEPTCHet-s(218-289)

Fig. 8. Solid-state NMR of Het-s fibrils formed in the presence and absenceof GroEL. Comparison of 1D [1H, 13C] CP-MAS (A) and [1H, 13C] INEPT-MAS(B) spectra of Het-s fibrils formed in the absence (blue) and presence (red) ofGroEL. Insets shows the structure of a Het-s(218–289) fibril (PDB ID code2RNM; ref. 8) with the rigid portions colored in dark gray (A) and the mobiletails and loops in orange (B). (C) Comparison of 2D [1H,13C] INEPT-MASspectra of Het-s fibrils formed in the absence (blue) and presence (red) ofGroEL. The 1D CP-based spectra (A) overlap perfectly showing that the rigidsegments (dark gray) of the Het-s(218–289) fibrils have the same overallstructure in presence and absence of GroEL, confirmed by the 2D [13C,13C]Dipolar Assisted Rotational Resonance (DARR) spectra shown in Fig. S6).There are large differences in peak intensities in the INEPT-based experi-ments (B and C) indicating differences in the mobile portions of the fibrilsupon GroEL binding. Cross-peaks of Het-s fibrils with significant chemicalshift differences in the presence and absence of GroEL, together with as-signments in terms of residue/atom type probabilities (31), are indicated.One hundred micromolar monomeric Het-s(218–289) was fibrilized at roomtemperature under quiescent conditions in the absence and presence of100 μM (in subunits) GroEL.

Experimental Simulated

-40 -20 0 20 40 -40 -20 0 20 40

4x

0 μM GroELt = 5 min

t = 2 weeks

4x t = 2 weeks

370 μM GroELt = 5 min

0 μM GroEL

370 μM GroEL

τc ~ 0.5 ns (100%)

τc ~ 0.5 ns (~ 1%)τc ~ 3.3 ns (~ 99%)

τc ~ 0.5 ns (~ 10%)τc ~ 3.3 ns (~ 90%)

τc ~ 0.5 ns (~ 1%)τc ~ 3.3 ns (~ 51%)τc > 200 ns (~ 48%)

ΔH0 (gauss)

A

B

C

D

2Azz

2aiso

Fig. 7. Comparison of experimental and simulated CW X-band EPR spectraof MTSL-labeled Het-s (S227C) in different states. (A) Monomeric Het-s.(B) Het-s fibrils were obtained after 2 wk at room temperature, centrifuged,and the pellet was resuspended in water (C) Het-s protofibrils obtainedwithin 5 min of addition of GroEL (the sample is not spun down). (D) Het-sfibrils obtained at room temperature 2 wk after the addition of GroEL,centrifuged, and the pellet was resuspended in water. The experimental andsimulated first derivative EPR spectra are shown in the left and right col-umns, respectively. The simulated spectra were calculated with differentcorrelation times and species populations as indicated (26). The derivativeEPR spectra are normalized to the double integral.

9108 | www.pnas.org/cgi/doi/10.1073/pnas.1711645114 Wälti et al.

are provided in SI Experimental Methods. CW EPR spectra at X-band wererecorded on a Bruker EMX spectrometer (see SI Experimental Methods).

EM and AFM. Samples of Het-s/GroEL complexes were prepared in the samemanner as for the solution-state NMR experiments (SI ExperimentalMethods). Aliquots were taken at different time points and diluted to afinal GroEL concentration of 2.5 μM. For EM, 5-μL samples were blottedonto the carbon-coated copper EM grids (Ultrathin Carbon Film/HoleyCarbon; Ted Pella) for 1 min, quickly washed with deionized water, andstained with 2% uranyl acetate for 40 s. Images were taken with an FEITecnai T12 electron microscope operating at 120 kV and a Gatan US1000CCD camera.

For electron tomography, tilt series were collected using SerialEM (39) onan FEI Tecnai F20 electron microscope operating at 200 kV and equippedwith a Gatan K2 direct electron detector. The tilt series ranges from –60 to+60 degrees in increments of 2 degrees. Tomograms were reconstructedusing the ETomo routine in the IMOD package (40).

For AFM, the same samples were directly applied to freshly cleaved mica.The sample was incubated for 1 min, washed with deionized water, andfurther air dried. AFM images were taken in the tapping mode using aMultiMode AFMwith a Nanoscope IV controller (Veeco). This included a SPMprobemodel ACT silicon (Applied Nanostructures) with a 5- to 6-nm tip radius,40 N/m force constant, and oscillating at ≈300 kHz. Images were recorded ata scan rate of 0.7–1 Hz, 256 sampling points per line, and 256 lines.

ACKNOWLEDGMENTS. We thank Rob Tycko for useful discussions; JamesBaber, Dan Garrett, and Jinfa Ying for technical support; and Roland Riek forthe gift of the Het-s(218–289) clone. M.A.W. was supported by an EarlyPostdoc.Mobility Fellowship from the Swiss National Science Foundation.D.T.M. was supported by the Postdoctoral Research Associate Program ofthe National Institute of General Medical Sciences, NIH Award Fi2GM117604.This work was supported by the Intramural Program of the National Instituteof Diabetes and Digestive and Kidney Diseases, National Institutes of Health(G.M.C. and J.E.H.), and by the AIDS-Targeted Antiviral Program of the Of-fice of the Director of the National Institutes of Health (G.M.C.).

1. Hartl FU, Hayer-Hartl M (2002) Molecular chaperones in the cytosol: From nascentchain to folded protein. Science 295:1852–1858.

2. Balch WE, Morimoto RI, Dillin A, Kelly JW (2008) Adapting proteostasis for diseaseintervention. Science 319:916–919.

3. Hartl FU, Bracher A, Hayer-Hartl M (2011) Molecular chaperones in protein foldingand proteostasis. Nature 475:324–332.

4. Wyatt AR, Yerbury JJ, Dabbs RA, Wilson MR (2012) Roles of extracellular chaperonesin amyloidosis. J Mol Biol 421:499–516.

5. Chiti F, Dobson CM (2006) Protein misfolding, functional amyloid, and human disease.Annu Rev Biochem 75:333–366.

6. Eanes ED, Glenner GG (1968) X-ray diffraction studies on amyloid filaments.J Histochem Cytochem 16:673–677.

7. Fowler DM, Koulov AV, Balch WE, Kelly JW (2007) Functional amyloid––from bacteriato humans. Trends Biochem Sci 32:217–224.

8. Wasmer C, et al. (2008) Amyloid fibrils of the HET-s(218-289) prion form a β solenoidwith a triangular hydrophobic core. Science 319:1523–1526.

9. Van Melckebeke H, et al. (2010) Atomic-resolution three-dimensional structure ofHET-s(218-289) amyloid fibrils by solid-state NMR spectroscopy. J Am Chem Soc 132:13765–13775.

10. Braig K, et al. (1994) The crystal structure of the bacterial chaperonin GroEL at 2.8 A.Nature 371:578–586.

11. Zeilstra-Ryalls J, Fayet O, Georgopoulos C (1991) The universally conserved GroE(Hsp60) chaperonins. Annu Rev Microbiol 45:301–325.

12. Pei W, et al. (2016) Extracellular HSP60 triggers tissue regeneration and wound healingby regulating inflammation and cell proliferation. NPJ Regenerative Med, 16013.

13. Muchowski PJ, Wacker JL (2005) Modulation of neurodegeneration by molecularchaperones. Nat Rev Neurosci 6:11–22.

14. Jones GW, Tuite MF (2005) Chaperoning prions: The cellular machinery for propa-gating an infectious protein? BioEssays 27:823–832.

15. Dedmon MM, Christodoulou J, Wilson MR, Dobson CM (2005) Heat shock protein70 inhibits alpha-synuclein fibril formation via preferential binding to prefibrillarspecies. J Biol Chem 280:14733–14740.

16. Knowles TP, et al. (2007) Kinetics and thermodynamics of amyloid formation fromdirect measurements of fluctuations in fibril mass. Proc Natl Acad Sci USA 104:10016–10021.

17. Perrett S, Jones GW (2008) Insights into the mechanism of prion propagation. CurrOpin Struct Biol 18:52–59.

18. Zhang H, Xu LQ, Perrett S (2011) Studying the effects of chaperones on amyloid fibrilformation. Methods 53:285–294.

19. Shammas SL, et al. (2011) Binding of the molecular chaperone αB-crystallin to Aβamyloid fibrils inhibits fibril elongation. Biophys J 101:1681–1689.

20. Mannini B, et al. (2012) Molecular mechanisms used by chaperones to reduce thetoxicity of aberrant protein oligomers. Proc Natl Acad Sci USA 109:12479–12484.

21. Månsson C, et al. (2014) DNAJB6 is a peptide-binding chaperone which can suppressamyloid fibrillation of polyglutamine peptides at substoichiometric molar ratios. CellStress Chaperones 19:227–239.

22. Arosio P, et al. (2016) Kinetic analysis reveals the diversity of microscopic mechanismsthrough which molecular chaperones suppress amyloid formation. Nat Commun7:10948.

23. Mogk A, Bukau B (2004) Molecular chaperones: Structure of a protein disaggregase.Curr Biol 14:R78–R80.

24. Doyle SM, Genest O, Wickner S (2013) Protein rescue from aggregates by powerfulmolecular chaperone machines. Nat Rev Mol Cell Biol 14:617–629.

25. Wyatt AR, et al. (2011) Clusterin facilitates in vivo clearance of extracellular misfoldedproteins. Cell Mol Life Sci 68:3919–3931.

26. Stoll S, Schweiger A (2006) EasySpin, a comprehensive software package for spectralsimulation and analysis in EPR. J Magn Reson 178:42–55.

27. Nesmelov YE, Thomas DD (2010) Protein structural dynamics revealed by site-directedspin labeling and multifrequency EPR. Biophys Rev 2:91–99.

28. Stan G, Brooks BR, Lorimer GH, Thirumalai D (2006) Residues in substrate proteinsthat interact with GroEL in the capture process are buried in the native state. ProcNatl Acad Sci USA 103:4433–4438.

29. Takegoshi K, Nakamura S, Terao T (2003) 13C-1H dipolar-driven 13C-13C recouplingwithout 13C rf irradiation in nuclear magnetic resonance of rotating solids. J ChemPhys 118:2325–2341.

30. Sackewitz M, et al. (2008) Structural and dynamical characterization of fibrils from adisease-associated alanine expansion domain using proteolysis and solid-state NMRspectroscopy. J Am Chem Soc 130:7172–7173.

31. Vranken WF, et al. (2005) The CCPN data model for NMR spectroscopy: Developmentof a software pipeline. Proteins 59:687–696.

32. Libich DS, Fawzi NL, Ying J, Clore GM (2013) Probing the transient dark state ofsubstrate binding to GroEL by relaxation-based solution NMR. Proc Natl Acad Sci USA110:11361–11366.

33. Lührs T, et al. (2005) 3D structure of Alzheimer’s amyloid-β(1-42) fibrils. Proc Natl AcadSci USA 102:17342–17347.

34. Paravastu AK, Leapman RD, Yau WM, Tycko R (2008) Molecular structural basis forpolymorphism in Alzheimer’s β-amyloid fibrils. Proc Natl Acad Sci USA 105:18349–18354.

35. Wälti MA, et al. (2016) Atomic-resolution structure of a disease-relevant Aβ(1-42)amyloid fibril. Proc Natl Acad Sci USA 113:E4976–E4984.

36. Ojha J, Masilamoni G, Dunlap D, Udoff RA, Cashikar AG (2011) Sequestration of toxicoligomers by HspB1 as a cytoprotective mechanism. Mol Cell Biol 31:3146–3157.

37. Taguchi H, Tsukuda K, Motojima F, Koike-Takeshita A, Yoshida M (2004) BeF(x) stopsthe chaperonin cycle of GroEL-GroES and generates a complex with double foldingchambers. J Biol Chem 279:45737–45743.

38. Daskalov A, et al. (2014) Contribution of specific residues of the β-solenoid fold toHET-s prion function, amyloid structure and stability. PLoS Pathog 10:e1004158.

39. Mastronarde DN (2005) Automated electron microscope tomography using robustprediction of specimen movements. J Struct Biol 152:36–51.

40. Kremer JR, Mastronarde DN, McIntosh JR (1996) Computer visualization of three-dimensional image data using IMOD. J Struct Biol 116:71–76.

41. Bartolucci C, Lamba D, Grazulis S, Manakova E, Heumann H (2005) Crystal structure ofwild-type chaperonin GroEL. J Mol Biol 354:940–951.

42. Grason JP, Gresham JS, Widjaja L, Wehri SC, Lorimer GH (2008) Setting the chaperonintimer: The effects of K+ and substrate protein on ATP hydrolysis. Proc Natl Acad SciUSA 105:17334–17338.

43. Todd MJ, Lorimer GH (1998) Criteria for assessing the purity and quality of GroEL.Methods Enzymol 290:135–141.

44. Ritter C, et al. (2005) Correlation of structural elements and infectivity of the HET-sprion. Nature 435:844–848.

45. Chaudhry C, Horwich AL, Brunger AT, Adams PD (2004) Exploring the structural dy-namics of the E.coli chaperonin GroEL using translation-libration-screw crystallo-graphic refinement of intermediate states. J Mol Biol 342:229–245.

46. Fei X, Ye X, LaRonde NA, Lorimer GH (2014) Formation and structures of GroEL:GroES2 chaperonin footballs, the protein-folding functional form. Proc Natl AcadSci USA 111:12775–12780.

Wälti et al. PNAS | August 22, 2017 | vol. 114 | no. 34 | 9109

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Supporting InformationWälti et al. 10.1073/pnas.1711645114SI Experimental MethodsExpression and Purification of GroEL. GroEL was purified as de-scribed (42). Briefly, GroEL was expressed in E. coli (BL21DE3)in Luria Bertini (LB) medium at 37 °C and induced at OD600 =0.8 with 0.5 mM isopropyl β-D-1-thiogalactopyranoside (IPTG)and grown overnight at 20 °C. After cell lysis, nucleic acids wereremoved by streptomycin sulfate precipitation. The supernatantwas applied to a self-packed 60-mL ion exchange column (QSepharose FF; GE Healthcare). The GroEL-containing fractionswere pooled and subjected to ammonium sulfate precipitation.The precipitate was dissolved and injected onto a HiPrepSephacryl S-300 gel filtration column (GE Healthcare). GroELwas further purified with an acetone precipitation step as pub-lished (42). The purity was confirmed by mass spectrometry (MS),SDS polyacrylamide gel electrophoresis (SDS/PAGE), and theabsence of tryptophan fluorescence (43). Since GroEL has notryptophans, all fluorescence signal can be attributed to proteinimpurities. The GroEL assembly was confirmed by blue nativepolyacrylamide gel electrophoresis (BN-PAGE) and stored at 4 °C.

Expression and Purification of GroES. GroES expressed in E. coli(BL21 DE3) was induced in LB medium at 37 °C at an OD600 of0.8 with 0.5 mM IPTG. The protein was expressed for 5–6 h at37 °C. The harvested cells were lysed in buffer A (50 mM Tris,pH 8, 300 mM NaCl, and 10 mM imidazole), and streptomycinsulfate was added to eliminate nucleic acids. The cells werecentrifuged, and the supernatant was loaded onto a 5 mL nickelSepharose column (HisTrap HP; GE Healthcare). The proteinwas eluted in buffer A containing in addition 250 mM imidazole.The GroES containing fractions as visualized by SDS/PAGEwere pooled and dialyzed against the thrombin cleavage buffer(20 mM Tris, pH 8, 150 mM NaCl, 2.5 mM MgCl2) overnight.The next day, the buffer was changed, and thrombin was addedto the dialysis cassette and left for another night. The proteinwas then loaded onto a 5-mL nickel Sepharose column, equili-brated in buffer B (50 mM Tris, pH 8 and 100 mM NaCl) toremove the uncleaved fraction. The flow-through was collectedand applied to a benzamidine column (HiTrap Benzamidine FF;GE Healthcare) equilibrated with buffer B to remove thrombinfrom GroES. The purity of the protein was verified by SDS/PAGE,mass spectrometry, and solution-state NMR spectroscopy, and theassembly was affirmed by BN-PAGE. GroES was flash frozen inliquid N2 and stored at −20 °C.

GroEL and GroES Assembly. GroES was added to GroEL to finalconcentrations of 400 and 100 μM (in subunits), respectively. Thebuffer was adjusted to 50 mM Tris, pH 7.4, 100 mM KCl, and10 mMMgCl2. The final solution was incubated with ≈10 mMATPfor at least 15 min at room temperature. Two millimolar AlF3(formed by premixing 2 mM AlCl3 and 40 mM NaF) was addedwith either ATP or ADP to form bullet-or football-shapedGroEL/GroES complexes (37), respectively (Fig. S4). The un-bound GroES was removed by washing and filtering through a100-kDa cutoff centrifugal filter (Amicon; Millipore). Thissolution was diluted 1,000 times and then added to 100 μMHet-s fibrils.

Expression, Purification, and Sample Preparation of Het-s(218–289).Het-s(218–289) with a C-terminal His6 tag and a N-terminalmethionine was purified under denaturing conditions as de-scribed (44). Briefly, 15N-labeled Het-s(218–289) was expressedin standard M9 minimal medium. Expression was induced with

0.5 mM IPTG at an OD600 of 0.8 and further grown at 37 °Covernight. The cells were harvested and resuspended in buffer C(6M guanidinium chloride, 300 mMNaCl, and 50 mMTris at pH 8)and stirred overnight at 4 °C. Cell lysis was achieved by heating to60 °C overnight. The lysate was centrifuged, and the supernatant wasbatch bound to nickel Sepharose resin (GE Healthcare), preequili-brated with 10 mM imidazole, in buffer C overnight. Het-s(218–289)was eluted into buffer C containing additionally 250 mM imidazole.The Het-s(218–289)–containing fractions were pooled and desaltedin 10% acetic acid using a HiTrap column (Amersham). The proteinwas lyophilized and stored at −20 °C. The purity of the sample wasconfirmed by SDS/PAGE and mass spectrometry.For the S227C mutation, the codon encoding the Ser at position

227 was replaced by one for Cys using the QuikChange mutagenesiskit, and the plasmid was transformed into E. coli BL21 Star (DE3)competent cells (Life Technologies). Het-s (S227C) was purified asdescribed above. For site-specific spin labeling, Het-s(S227C) wasincubated with a 10-fold excess of either paramagnetic MTSL(Toronto Chemical Research) or diamagnetic (1-acetoxy-2,2,5,5-tetramethyl-d3-pyrroline-3-methyl)methanethiosulfonate (TorontoResearch Chemicals), overnight at room temperature in buffer Cbefore buffer exchange to 10% acetic acid. Spin-labeling was con-firmed by mass spectrometry.Fibrillization of Het-s(218–289) was carried out as published

(38). In summary, 1 mg of lyophilized protein was dissolved in200 μL of 45 mM HCl, filtered through a 0.2-μm filter, and thepH was adjusted to 7.4 by addition of 1 M Tris, pH 7.4, resultingin a final buffer composition of 230 mM Tris/45 mM HCl, pH7.4. The protein was diluted to a concentration of ≈200 μM asmeasured by UV absorption at 280 nm, and then further dilutedby a factor of two either by the addition of buffer or GroEL.

SDS/PAGE Analysis of Het-s Protofibrils. One hundred micromolar(in subunits) GroEL was added to Het-s(218–289) in a 1:1 ratioand incubated overnight. The turbid sample was centrifuged for5 min at 14,000 × g on a tabletop centrifuge. The supernatantwas removed, and an aliquot of 5 μL was mixed with the SDSprotein loading buffer from Quality Biological [containing Trisbase, 4% (wt/vol) SDS, 0.2% (vol/vol) bromophenol blue, 20%(vol/vol) glycerol, and molecular biology grade water], and loadedon a SDS/PAGE gel (4–12% wt/vol). The pellet was washed threetimes with water, dissolved with the SDS protein loading buffer(Quality Biological), and diluted to the final concentration givenin Fig. S2.

Solution NMR Spectroscopy. Het-s(218–289) was treated as de-scribed above except for the addition of D2O (5% vol/vol) to thesamples before measurements. Several microliters of either a∼1 mM (in subunits) GroEL stock solution or the GroEL buffer(10 mM Tris, pH 7.4 and 10 mMMgCl2) were added to ≈200 μMHet-s(218-298) and further diluted to reach a final concentrationof 100 μM Het-s and 0–370 μM (in subunits) GroEL. Fibriliza-tion was achieved under quiescent conditions at room temper-ature for approximately 3 wk. Solution-state NMR experimentswere performed in Shigemi tubes in a volume of 250 μL on aBruker 600 MHz spectrometer with a triple resonance z axisgradient cryoprobe at 10 °C. Data processing was performed byusing NMRPipe and analyzed either in Topspin3.5 or using theccpNMR analysis suite (31). Standard [1H,15N] HSQC spectra wererecorded to study the aggregation kinetics. Backbone resonanceassignments of Het-s(218–289) were carried out using standard3D triple resonance NMR experiments (HNCA, HNCACB, and

Wälti et al. www.pnas.org/cgi/content/short/1711645114 1 of 5

HNCOCA) recorded on a 215 μM [15N,13C]-labeled Het-s(218–289) sample.

Solid-State NMR Spectroscopy. The Het-s(218–289) fibrils with andwithout GroEL were packed into a thin-walled 3.2-mm Varianstyle NMR rotor by centrifugation at 5,000 × g. All experimentswere performed on a 17.5 T NMR spectrometer operating at a1H frequency of 746.0548 MHz using a Chemmagnetics con-sole and a 3.2 mm Black Fox magic angle spinning (MAS) probe(Black Fox) with 12 kHz spinning and the variable temperaturegas set to −15 °C to maintain the sample temperature at ap-proximately +15 °C. For the 1D 1H-13C CP-based experiments,the length of the 1H 90° pulse was 4 μs and the radio-frequency(RF) field strength for 1H two-pulse-phase modulation (TPPM)decoupling was 85 kHz. During the 1.5-ms CP period, the RF fieldstrengths used for 1H and 13C irradiation were ∼58 and ∼44 kHz,respectively. One thousand twenty-four complex points were ac-quired with a dwell time of 15 μs and a recycle delay of 1.75 s. The13C-13C DARR experiment was obtained under the same condi-tions as the 1D CP spectra. The length of the 13C 90° pulse was5 μs, the DARR mixing time was 30 ms, and the RF field strengthof the DARR 1H pulse was ∼12 kHz. Five hundred t1 points wereacquired with a 22.4-μs increment. Data were processed in“States” mode. One-dimensional and 2D INEPT spectra wereobtained with 4-μs 1H 90°, 8-μs 1H 180°, 5-μs 13C 90°, and 10-μs13C 180° pulses. The first and second J-evolution times were setto 0.9 and 1.4 ms, respectively. Ten kilohertz RF 1H decoupling

was applied during the acquisition period comprising 2,048 complexpoints with a dwell time of 15 μs. For the 2D INEPT, the t1 in-crement was 133 μs for both samples. For the Het-s sample con-taining GroEL, 90 t1 points were collected with an evolution time of12 ms; for the Het-s sample without GroEL, 135 t1 points werecollected for an evolution time of 18 ms. All data were processedusing a 60-Hz Gaussian apodisation window function and zero-filledto four times the original data size.

EPR Spectroscopy.MTSL-labeled Het-s (S227C) samples (100 μM)with or without 370 μM (in subunits) GroEL were loaded intoquartz capillaries (0.6-mm inner diameter × 0.84-mm outer di-ameter, VitroCom). Room temperature CW EPR spectra wererecorded on a Bruker EMX spectrometer at X-Band with a scanwidth of 100 Gauss using an HS cavity and an incident micro-wave power of 20 mW. EPR spectra of Het-s fibrils were alsorecorded on a paramagnetically diluted sample comprising a1:9 ratio of paramagnetic MTSL to diamagnetic MTS-labeledHet-s to determine the impact of dipolar coupling arising fromneighboring Het-s units within the cross β-sheet fibril; for thepurposes of the qualitative analysis presented in Fig. 7, no sig-nificant difference in effective correlation time for the spin labelwas seen in the paramagnetically diluted sample (τeff ∼3.1 nsversus ∼3.3 ns in the 100% MTSL-labeled Het-s fibrils) [i.e.,dipolar couplings between spin labels spaced at 10-Å intervalsalong the fibril (cf. Fig. 1B) in the 100% MTSL-labeled sampleresult in only a small (<10%) increase in the fitted τeff value].

A

B

C

D

mn 001mn 002 1µm

200 nm 100 nm

200 nm

200 nm

100 nm

100 nm

1µm

1µm

1µm

+ 10 µM GroEL

+ 50 µM GroEL

+ 100 µM GroEL

+ 370 µM GroEL

Fig. S1. Negative stain electron micrographs (Left and Center) and amplitude AFM images (Right) of Het-s protofibrils obtained immediately after addition ofGroEL at different concentrations. Ten (A), 50 (B), 100 (C), and 370 (D) micromolar (in subunits) GroEL was added to 100 μM Het-s(218–289) monomer.

Wälti et al. www.pnas.org/cgi/content/short/1711645114 2 of 5

Gro

EL

Het-sHet-s + GroEL

protofibrilsMW

1701309672

(kDa)

55

43

3426

17

101 2 3 4 5 6 7 8

Fig. S2. Characterization of GroEL, Het-s, and Het-s/GroEL protofibrils by Coomassie-stained SDS/PAGE (4–12% wt/vol). Lane 1, molecular mass standards; lane2, 10 μM (in subunits) GroEL (subunit theoretical molecular mass = 57.1 kDa); lanes 3 and 4, 10 and 20 μMHet-s(218–289) (theoretical molecular mass = 8.7 kDa),respectively; lane 5, supernatant after spinning down protofibrils obtained 5 min after addition of 100 μM (in subunits) GroEL to 100 μM Het-s; lanes 6–8,pelleted protofibrils (obtained 5 min after addition of 100 μM in subunits GroEL to 100 μM Het-s) dissolved in SDS loading buffer at concentrations of 7, 4, and2 μM, respectively, in GroEL (subunit concentration) and Het-s in a 1:1 ratio.

A B C D

1 µm 1 µ 1m µ 1m µm

Het-s t = 2 weeks

Het-s + GroEL t ≤ 5 min

Het-s + GroEL t = 18 hrs

Het-s + GroEL t = 2 weeks

Fig. S3. Amplitude AFM images of Het-s protofibril/fibril formation over time in the absence and presence of GroEL. (A) Het-s(218-29) fibrils (100 μM inmonomer) formed after 2 wk at room temperature. Het-s(218–289) protofibrils and fibrils (100 μM in monomer) obtained immediately (B), after 18 h (C), andafter 2 wk (D) following the addition of 370 μM (in subunits) GroEL. In B and C, only protofibils are present; but after 2 wk (C), both fibrils and profibrils areclearly observed. Fibrilization was carried out at room temperature under quiescent conditions.

GroES

GroES GroESGroEL GroEL

BA

20 nm20 nm

Fig. S4. Negative stain electron micrographs of 100 μM (in subunits) GroEL capped with 100 μM (in subunits) GroES. (A) Football-shaped GroEL/GroEScomplexes (both ends capped by GroES) formed upon addition of ATP. (B) Bullet-shaped GroEL/GroES complexes (one end capped by GroES) formed uponaddition of ADP. Typical examples are indicated by arrows in both A and B. Insets in A and B show the X-ray structures of the football-shaped (PDB ID code4PKN; refs. 45 and 46) and bullet-shaped (PDB ID code 1SX4; ref. 45) GroEL/GroES complexes, respectively. Details regarding the assembly of the two GroEL/GroES complexes are given in Experimental Methods.

Wälti et al. www.pnas.org/cgi/content/short/1711645114 3 of 5

20 nm

Fig. S5. Negative stain EM of Het-s fibrils in the presence of football-shaped GroEL/ES complexes. Most of the GroEL is still uncapped and bound to the fibrils;football-shaped GroEL/ES complexes can be clearly seen (typical examples are indicated by arrows) and are never observed bound to the fibrils. Het-s (100 μMmonomer) was fibrilized at room temperature under quiescent conditions. Football-shaped GroEL/GroES complex (prepared as described in ExperimentalMethods) were added directly to the fibrils.

203040506070

20

30

40

50

60

70

13C (ppm

)

13C (ppm)

Het-s aloneHet-s + GroEL

Fig. S6. Comparison of 2D CP-MAS 13C/13C DARR correlation spectra of Het-s fibrils obtained in the absence (blue) and presence (red) of GroEL. During the30-ms DARR mixing time, magnetization is transferred between adjacent 13C atoms by spin diffusion arising from 13C-13C dipolar couplings. Fibrilization of100 μM Het-s(218–289) in the absence or presence of 100 μM (in subunits) GroEL was performed over a period of 3 wk at room temperature under quiescentconditions.

Wälti et al. www.pnas.org/cgi/content/short/1711645114 4 of 5

A B

+ 100 µM GroEL

6420Echo Time (ms)

1.0

0.8

0.6

0.4

0.2

0.0In

tegr

al

3020100Echo Time (ms)

aliphatic 1H-T2 13C-T2

Het-s fibrils

0 µM GroEL

Fig. S7. Global aliphatic 1H (A) and 13C transverse (B) relaxation rates (T2) of Het-s(218–289) fibrils formed in the absence (blue) and presence (red) of GroEL.1H-T2 values are 24.7 ± 0.1 and 16.8 ± 0.1 ms for fibrils formed in the absence and presence of GroEL, respectively; the corresponding 13C-T2 values are 4.8 ± 0.1 and3.9 ± 0.2 ms, respectively. Het-s fibrils were obtained by fibrilization of 100 μMHet-s monomer at room temperature for 3 wk under quiescent conditions in the absenceand presence of 100 μM (in subunits) GroEL.

Wälti et al. www.pnas.org/cgi/content/short/1711645114 5 of 5