Heat of supersaturation-limited amyloid burst directlymonitored by isothermal titration calorimetryTatsuya Ikenouea,1, Young-Ho Leea,1, József Kardosb, Hisashi Yagia, Takahisa Ikegamia, Hironobu Naikic,and Yuji Gotoa,2

aDivision of Protein Structural Biology, Institute for Protein Research, Osaka University, Osaka 565-0871, Japan; bDepartment of Biochemistry, Eötvös LorandUniversity, 1117, Budapest, Hungary; and cFaculty of Medical Sciences, University of Fukui, Fukui 910-1193, Japan

Edited* by David S. Eisenberg, University of California, Los Angeles, CA, and approved March 26, 2014 (received for review December 12, 2013)

Amyloid fibrils form in supersaturated solutions via a nucleationand growth mechanism. Although the structural features of amyloidfibrils have become increasingly clearer, knowledge on the thermo-dynamics of fibrillation is limited. Furthermore, protein aggregationis not a target of calorimetry, one of the most powerful approachesused to study proteins. Here, with β2-microglobulin, a protein re-sponsible for dialysis-related amyloidosis, we show direct heatmeasurements of the formation of amyloid fibrils using isothermaltitration calorimetry (ITC). The spontaneous fibrillation after a lagphase was accompanied by exothermic heat. The thermodynamicparameters of fibrillation obtained under various protein concentra-tions and temperatures were consistent with the main-chain domi-nated structural model of fibrils, in which overall packing was lessthan that of the native structures. We also characterized the ther-modynamics of amorphous aggregation, enabling the comparisonof protein folding, amyloid fibrillation, and amorphous aggregation.These results indicate that ITC will become a promising approach forclarifying comprehensively the thermodynamics of protein foldingand misfolding.

enthalpy change | metastability | solubility | thermodynamic stability |thioflavin T

Aggregation has often been an obstacle to studying the struc-ture, function, and physical properties of proteins. However,

a large number of aggregates associated with serious diseases,including Alzheimer’s, Parkinson, and prion diseases (1, 2)promoted the challenge of studying protein misfolding and ag-gregation. Researchers succeeded in distinguishing amyloid fibrilsand oligomers from other amorphous aggregates and character-ized the ordered structures present in amyloid fibrils or oligomers,which led to the development of the field of amyloid structuralbiology (3–8). These advances have been attributed to variousmethodologies that are also useful for studying the structuralproperties of globular proteins. Even X-ray crystallography hasbecome a powerful approach for studying amyloid microcrystals(5) or oligomers (9). The atomic details of amyloid fibrils arebecoming increasingly clearer, and a cross-β structure was shownto be the main structural component of fibrils (5, 6, 8). Althoughtightly packed core regions of amyloid fibrils have been reported,the overall structures were shown to be dominated by commoncross-β structures, which supported the argument for the main-chain dominated architecture in contrast to the side-chain dom-inated architecture of globular native states (10–12).These structural studies have been complemented by a series

of efforts to clarify the mechanism for the formation of amyloidfibrils (i.e., amyloid fibrillation). The presence of a long lag timein spontaneous fibrillation and rapid fibrillation by the additionof preformed fibrils represent a similarity with the supersaturation-limited crystallization of substances (13–18). We have revisited“supersaturation” and argued its critical role for amyloid fibrilla-tion (17–19). The role of supersaturation in neurodegenerativediseases at the proteome level has been reported recently (20).However, calorimetry, one of the most powerful methods

used to study the thermodynamic properties of globular proteins(21–24), has not played a significant role in understanding

protein aggregation. The aggregation of proteins following heatdenaturation as monitored by differential scanning calorimetryis an infamous example demonstrating how aggregation canprevent exact analyses (25, 26). To date, few studies have in-vestigated protein aggregation including amyloid fibrils withcalorimetry (27–32). Our previous study on the exothermic heateffects accompanying fibril growth was achieved by monitoringthe seed-dependent elongation of fibrils formed by β2-micro-globulin (β2m), a protein responsible for dialysis-related amy-loidosis, using isothermal titration calorimetry (ITC) (28).In the present study using β2m, we succeeded in characterizing

the total heat of spontaneous fibrillation and amorphous aggrega-tion. An analysis of the heat burst associated with fibrillation oramorphous aggregation under various temperatures clarified theirthermodynamic properties. The results obtained enabled the calo-rimetric characterization of amyloid fibrils and amorphous aggre-gates relative to that of the native globular structures, which opensa new field for the calorimetric study of protein aggregates.

ResultsHeat for the Formation of Amyloid Fibrils Monitored by ITC. At pH2.5, acid-denatured β2m formed amyloid fibrils in the presence ofmoderate concentrations of NaCl. As defined by the conforma-tion phase diagram, fibril formation is dependent on protein andNaCl concentrations (Fig. S1) (19, 33). Spontaneous fibrillationwas previously shown to be facilitated by various kinds of agi-tations such as stirring with a magnetic bar (34, 35) or ultra-sonication (19, 36–39), leading to a burst phase of fibrillationafter a lag phase. Under the conditions of persistent meta-stability of supersaturation, it is likely that these agitationsmay create seed-competent conformations. For instance, during

Significance

Although amyloid fibrils are associated with numerous pa-thologies, their conformational stability remains largely un-known. In particular, calorimetry, one of the most powerfulmethods used to study the thermodynamic properties ofglobular proteins, has not played a significant role in un-derstanding protein aggregation. Here, with β2-microglobulin,we established direct heat measurements of supersaturation-limited amyloid fibrillation using an isothermal titration calo-rimeter. We also revealed the thermodynamics of amorphousaggregation. By creating a totally new field of calorimetricstudy of protein misfolding, we can now comprehensivelyaddress the thermodynamics of protein folding and misfolding.

Author contributions: T. Ikenoue, Y.-H.L., J.K., T. Ikegami, H.N., and Y.G. designed re-search; T. Ikenoue, Y.-H.L., J.K., and H.Y. performed research; T. Ikegami contributednew reagents/analytic tools; T. Ikenoue, Y.-H.L., J.K., and H.Y. analyzed data; andT. Ikenoue, Y.-H.L., and Y.G. wrote the paper.

The authors declare no conflict of interest.

*This Direct Submission article had a prearranged editor.1T. Ikenoue and Y.-H.L. contributed equally to this work.2To whom correspondence should be addressed. E-mail: ygoto@protein.osaka-u.ac.jp.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1322602111/-/DCSupplemental.

6654–6659 | PNAS | May 6, 2014 | vol. 111 | no. 18 www.pnas.org/cgi/doi/10.1073/pnas.1322602111

Dow

nloa

ded

by g

uest

on

Dec

embe

r 17

, 202

0

air–water interface-dependent protein aggregation, a template-competent conformation is formed (39).In our studies, we used the ITC instrument to agitate the β2m

solution and monitor the heat response of fibrillation. To es-tablish supersaturation in the presence of various concentrationsof β2m at pH 2.5 in the cell at 37 °C, the NaCl concentration wasincreased to a final value of 0.1 M by stepwise injections ofa small volume of 1.0 M NaCl (Fig. 1A; Materials and Methods).After each injection, a sharp endothermic or exothermic spike,which represented the heat (q) of salt dilution, occurred and theheat flow (=∂q/∂t) returned back to the original reference powerlevel. Notably, a marked exothermic peak with a half-width of ∼2 hoccurred at 0.3 mg·mL−1 of β2m at ∼11 h (Fig. 1A). Similar exo-thermic peaks were observed at other concentrations of β2m. Thelag time for the major exothermic peak (Materials and Methods)shortened (Fig. S2A) and the exothermic peak became larger withan increase in the protein concentration. When 0.5 mg·mL−1 ofthe β2m solution at 0.1 M NaCl was prepared in a test tube, setin the ITC cell, and followed by stirring, a similar exothermic burstwith a lag time of 3.7 h was observed (Fig. S3). When we considerthe time for titration of salt (3.5 h; Fig. 3A), the observed lag timewas independent of the methods, although the titration in theITC cell was much simpler. The results suggested that the ti-tration inside the ITC cell did not bring any additional effects.The total heat calculated based on the peak area was nor-

malized by the protein concentration. The normalized heat didnot depend significantly on the protein concentration (Fig. 1E).Moreover, when the stirring speed was varied in the range of 200–1,000 rpm with a fixed protein concentration of 0.5 mg·mL−1, thelag time shortened with an increase in the speed (Fig. 2A and Fig.S2B). However, the total heat was independent of the stirringspeed (Fig. 2B). These results suggested that the observed heatrepresented the enthalpy change (ΔH) of the reaction triggeredby stirring. Assuming that the observed total heat was ΔH, theΔH value at 37 °C was estimated to be −77 kJ·mol−1 from the

dependence on stirring speed or −74 kJ·mol−1 from the de-pendence on protein concentration. The decrease in magnitudeof ΔH at high protein concentrations may have been linkedwith the partial and transient formation of amorphous aggre-gates with a smaller ΔH value (see below).After the exothermic peaks, all β2m solutions exhibited a far-

UV CD spectrum with a minimum at ∼218 nm, an atomic forcemicroscopy (AFM) image of fibrils with a height of 4.5–9.0 nmand various lengths up to 1 μm, and strong thioflavin T (ThT)fluorescence (Fig. 1 B–D). These results indicated that β2m sol-utions above 0.3 mg·mL−1 in 0.1 M NaCl at pH 2.5 were super-saturated (or metastable) and that agitation by stirring broke thissupersaturation, resulting in amyloid fibrillation. We considerthat the exothermic peak represents the formation of amyloidfibrils (“amyloid burst”) and the observed heat gives its ΔH value.Similar effects were expected for other salts, the effectiveness ofwhich follows the electroselectivity series (19, 33). One experi-ment with ammonium sulfate was shown in Fig. S4.

Small Amyloid Burst and Excess Heat Immediately After Salt Titration.Careful inspection of the ITC thermograms indicated that, in allof the ITC profiles, a small exothermic peak, which we desig-nated “small amyloid burst,” appeared before the main amyloidburst (Figs. 1A, 2A, and 5A). To clarify the significance of thesesmall peaks, we performed CD and AFM measurements and aThT assay at several time points during the reaction at 1.0 mg·mL−1

β2m (Fig. 3 B and C). Neither the CD spectrum nor the AFMimage showed significant changes before and immediately after thesalt injection spikes, which indicated that the dominant molecularspecies were still monomers. When the small exothermic peakappeared at the ∼5.5-h time point, the AFM image revealed thepresence of short and thin fibrils with a height of 2.6–5.3 nm. Aslight change in the CD spectrum and small increase in ThT fluo-rescence were also observed. These results indicated that some fi-brillation, possibly the formation of protofibrils, started at the point

Fig. 1. Calorimetric observation of the amyloid burst of β2m at various protein concentrations at 37 °C. (A) Thermograms of the fibril formation of β2m at0.3–6.7 mg·mL−1 and pH 2.5 obtained using ITC. Inset shows a close-up view of exothermic heat at 0.3 mg·mL−1 β2m. The arrowheads indicate the locations of“small burst.” These also apply to the thermograms of Figs. 2, 3, and 5. (B–D) Characterization of β2m solutions after incubation in ITC cells by AFM images (B),far-UV CD spectra (C), and ThT fluorescence intensities (D). The scale bars on the AFM images indicate 1 μm, and the numbers under images are fibril height.The scale bar on the Right represents the height. These also apply to AFM images in Figs. 3–5. (E) Dependences of the observed heat of the small peak (blackinverted triangle), main peak (red circles), total heat including rapid heat effect (green circle), and amorphous aggregation (black circles) on the proteinconcentration. Inset in C shows the expansion of the heat of amorphous aggregation. The observed heats were normalized by the β2m concentration to givethe ΔH values.

Ikenoue et al. PNAS | May 6, 2014 | vol. 111 | no. 18 | 6655

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Dec

embe

r 17

, 202

0

of the small burst, and subsequent elongation coupled with thebreakage of fibrils to make new growing ends (i.e., secondary nu-cleation) caused the explosive amyloid burst (Fig. 3 B and C). Theexact position and size of the minor peaks were less dependent onthe experimental conditions than those of the major peaks (Figs.

1A, 2A, and 5A, and Fig. S2). The total heat accompanying thesmall exothermic peak was constant (−1.5 kJ·mol−1) and in-dependent of the protein concentration (Fig. 1E). Although theobserved heat contained information on the ΔH value of protofibrilformation, its small fraction precludes further analysis.We also recognized a small excess heat effect immediately

after each of the stepwise addition of 1.0 M NaCl (Fig. S5). Thissmall but notable heat effect increased with an increase in theconcentration of NaCl and β2m, suggesting that it represents theformation of amorphous aggregates. However, after the com-pletion of major amyloid burst, the formation of amorphousaggregates was evident neither from the CD spectra, ThT in-tensities, nor AFM images (Fig. 1 B–D). Thus, it is possible thata small amount of amorphous aggregates formed after the saltinjection finally transformed to the fibrils, although the exactkinetics is unknown. If this is a case, a total heat including thoseof rapid heat effect, small amyloid burst, and major amyloid burstshould represent the ΔH value for amyloid fibrillation. Indeed, thesum of these heat effects was constant (−78 kJ·mol−1) overa wide range of concentration, suggesting the validity of as-sumption (Fig. 1E).

Heat of Amorphous Aggregation. β2m formed amorphous aggre-gates at very high NaCl concentrations above 0.8 M at pH 2.5(Fig. S1) (19). In analogy with the crystallization of substances,amyloid fibrils and amorphous aggregates were shown to besimilar to crystals and glasses, respectively (19). In Yoshimuraet al. (19), we showed that, whereas crystalline amyloid fibrilsformed after a lag phase, glassy amorphous aggregates formedwithout a lag phase. The rapid and partial formation of amor-phous aggregates after the salt titration was consistent with thisview (Fig. S5).It was difficult to increase the NaCl concentration in the cell

up to ∼1.0 M by injecting the NaCl solution at a high concen-tration in the syringe. Thus, we performed an inverse titration:the β2m solution at a high concentration in the syringe wasinjected into the cell containing 1.0 M NaCl (Fig. 4). On thebases of the low CD signal, amorphous aggregates revealed by

Fig. 2. Dependencies of the observed heat of aggregation on the stirringspeed at pH 2.5 and 37 °C. (A) Thermograms of the fibril formation of β2m at0.5 mg·mL−1 at various stirring speeds. The arrowheads indicate the loca-tions of “small burst.” (B) Dependence on the stirring speed of the observedheat normalized by the protein concentration. (C and D) Characterization ofβ2m solutions after incubation in ITC cells by far-UV CD spectra (C) and ThTfluorescence intensities (D).

Fig. 3. Monitoring the kinetics of the amyloidburst of β2m using various approaches. (A) ITCprofile of 1.1 mg·mL−1 β2m at pH 2.5 and 37 °C. Thelag time (red dot) is determined by a baseline anda tangent line at the middle of the major peak. (Band C) Conformational changes in β2m during in-cubation in an ITC cell characterized using AFMimages and ThT fluorescence intensities (B) and thefar-UV CD spectra (C) at the four time points: “Ini-tial state (0 h),” “After salt injection (∼3.5 h),”“Small burst (∼5.5 h),” and “After main burst (∼12 h).”Conformations of β2m based on AFM, ThT fluores-cence, and CD are illustrated above the AFM images:monomers (blue curves), oligomers (magenta curves),and fibrils (red rectangles). The CD spectra at the re-spective time points are shown by red solid curves.The spectra of monomers (black dotted curves) andmature fibrils (red dotted curves) are shown forcomparison.

6656 | www.pnas.org/cgi/doi/10.1073/pnas.1322602111 Ikenoue et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 17

, 202

0

the AFM image, and low ThT fluorescence, we confirmed thatβ2m formed amorphous aggregates. There was no lag phase inamorphous aggregation, which was consistent with our previousresults (19). Careful subtraction of the heat for the control ex-periment revealed the heat of amorphous aggregation. At 37 °C,the control heat effect without β2m was ∼1,250 μcal, whereas thatof β2 was around ∼1,190 μcal, with the excess heat of aggregationaround ∼5% of the basal heat effects. Again, there was no proteinconcentration dependence in the range of 0.1–0.7 mg·mL−1,which suggested that the heats represented the ΔH of amorphousaggregation (Fig. 4B). The ΔH value for amorphous aggregationwas estimated to be −43 kJ·mol−1 at 37 °C and assumed to beindependent of the protein concentration as indicated by thedotted line in Fig. 1E.

Temperature Dependency of Aggregation Heat. ITC measurementsof the amyloid burst at 1.0 mg·mL−1 were performed at varioustemperatures between 31 and 43 °C (Fig. 5A). The lag timeshortened (Fig. S2C) and the exothermic peak became largerwith an increase in temperature. The ΔH value increased inmagnitude from −41.3 to −101.1 kJ·mol−1 (Fig. 6A and TableS1). We confirmed using far-UV CD, AFM, and ThT assays thatthe products observed after heat burst at all of the temperatureswere amyloid fibrils (Fig. 5 B–D).Assuming that the observed heat effect represented ΔH,

temperature dependence provided a heat capacity change (ΔCp)of fibrillation based on the relationship of ΔCp = ∂ΔH/∂T. Theplot of ΔH against temperature was linear, providing a ΔCp valueof −5.0 kJ·mol−1·K−1 (Fig. 6A and Table S1). We previouslyobtained the ΔH value and temperature dependence for theseed-dependent elongation of amyloid fibrils of β2m monitoredby ITC (28) (Fig. 6A). Although the ΔH values for the sponta-neous fibrillation obtained here were slightly smaller than thoseof seed-dependent elongation, the current ΔCp value was similarto that (−4.8 kJ·mol−1·K−1) of seed-dependent elongation (28).We also measured temperature dependence of the heat effects

of amorphous aggregation (Figs. 4B and 6A, and Table S1).Although the ΔH values for amorphous aggregates at 1.0 M andvarious temperatures were smaller in intensity than those ofmature fibrils, ΔH changed linearly against temperature, pro-viding a ΔCp value (−3.5 kJ·mol−1·K−1) that was slightly smallerthan that of amyloid fibrils.

Evaluation of Thermodynamic Parameters. To comprehensivelyunderstand the thermodynamics of aggregation, we have to knowthe changes in free energy (ΔG) and entropy (ΔS) in addition tothe ΔH and ΔCp terms, which are directly determined by calo-rimetry (24). Although the detailed mechanical models of fibrilformation remain elusive (16), a simplified model will still bevalid for describing the equilibrium between monomers (M) andfibrils (P) (13–15, 28):

ðPÞ+ ðMÞk1!k−1 �

ðPÞ; [1]

where k1 and k−1 are the apparent rate constants for polymer-ization and depolymerization, respectively. The elongation offibrils is defined by the equilibrium association constant (K)as follows:

K =½P�½P�½M�= k1=k−1; [2]

where [P] is the concentration of fibrils and [M] is the concentra-tion of monomers. The equilibrium is clearly independent of [P].Hence, we obtain the equilibrium monomer concentration [M]eas follows:

½M�e = k−1=k1 = 1=K : [3]

[M]e is referred to as the “critical concentration” (13, 15) becausefibrils form when the concentration of monomers exceeds [M]e. Bydetermining [M]e, we can calculate the apparent free energy changeof fibrillation (ΔGapp) by the following: ΔGapp = −RTlnK =RTln[M]e, where R and T are the gas constant and temperature,respectively. Combined with the ΔH value directly obtainedfrom the ITC measurements, we can obtain the ΔS value byΔGapp=ΔH – TΔS. Althoughmechanism 1might not be exactlytrue for amorphous aggregation, we assumed that it is also a re-versible process determined by solubility and thus is approxi-mated by mechanism 1.We used an ELISA (SI Text) to determine the [M]e value

under various conditions (Table S2). We then estimated the

Fig. 4. Calorimetric observation of amorphous aggregation of β2m at 37 °C.(A) Thermogram of amorphous aggregation revealed by titrating 1.0 M NaClin the ITC cell with 3.6 mg·mL−1 β2m (red) or solvent (black) in the syringe.Titration was repeated 13 times to increase the β2m concentration. Theexpanded thermogram shows the second titration peak. (B) After subtract-ing the control, the excess heat was plotted against the final β2m concen-tration. The results of measurements at various temperatures are shown. (Cand D) CD spectrum, ThT fluorescence intensity, and AFM image of theamorphous aggregates formed in the ITC cell at 1.0 M NaCl. CD spectra offibrils and monomers are also shown.

Fig. 5. Temperature dependence of the amyloid burst of β2m monitored bycalorimetry. (A) Thermograms at various temperatures between 31 and 43 °C.The arrowheads indicate the locations of “small burst.” (B–D) Characteriza-tion of the β2m solution at 1.1 mg·mL−1 and 0.1 M NaCl after the heat burstby the far-UV CD (B), ThT fluorescence (C), and AFM (D).

Ikenoue et al. PNAS | May 6, 2014 | vol. 111 | no. 18 | 6657

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Dec

embe

r 17

, 202

0

ΔGapp and TΔS for fibrillation and amorphous aggregation. Wealso estimated the temperature dependencies of these param-eters as well as those of ΔH, in which we used ΔGapp values at37 °C to link the ΔH and TΔS functions. These functions werecompared with those for folding to the native state (Fig. 6).The ΔGapp value of fibrillation (−45.0 kJ·mol−1) at 37 °C

and pH 2.5 was the same as that of amorphous aggregation(−45.4 kJ·mol−1) under the same conditions (Fig. 6C). These valueswere significantly larger in intensity than that (−21.0 kJ·mol−1) ofthe native state at pH 7.0 (28), although distinct pH values precludea direct comparison. Although a small range of temperatures usedfor the experiments makes the extrapolation less accurate at thisstage, separation of ΔGapp into the enthalpy and entropy termsindicated that both amyloid fibrils and amorphous aggregates arestabilized enthalpically above 40 °C, whereas they are stabilizedentropically below 20 °C.

DiscussionAmyloid formation occurs in supersaturated solutions via anucleation-dependent manner (18, 40, 41), analogous to crys-tallization of substances (17, 42). Under the conditions ofpersistent metastability, nucleation does not occur in practice(14, 18). However, various kinds of agitations can break su-persaturation, leading to the formation of fibrils. We used ITCfor stirring the solution and for monitoring the accompanyingheat effects. The results showed that we can perform calori-metric measurements of amyloid fibrillation of β2m as well asamorphous aggregation revealing the ΔH and ΔCp values. Bycombining these values with ΔG obtained from the solubility ofβ2m monomers, we can address the thermodynamics of proteinaggregation (Fig. 6 and Tables S1 and S2). The methodology isstraightforward and can be applied to study various amyloid fibrilsas well as amorphous aggregates.The heat capacity change upon protein unfolding has been

primarily determined by the hydration of polar and apolargroups and to a much lesser extent by the disruption of internalnoncovalent interactions such as van der Waals interactions,H bonds, and ionic interactions (21, 24). Considering the mor-phological difference between the intramolecularly folded na-tive state and intermolecularly associated amyloid fibrils andassuming the same packing densities, the extent of burial shouldbe higher for fibrils assuming the same packing densities. The ΔCpvalues for the native, amyloid, and amorphous conformationswere −5.6, −5.0, and −3.5 kJ·mol−1·K−1, respectively (Fig. 6Aand Table S1). The similar values of ΔCp upon protein foldingand amyloid formation suggest a similar overall burial of sur-faces in the two forms. We consider that the tightly packed coreregions, as observed in amyloid microcrystals (5, 6), coexist withthe less densely packed noncore regions with cavities accessible

to bulk water (3), leading to an overall similar extent of burial ofsurfaces (12, 28). In contrast, the smaller ΔCp value of amor-phous aggregation suggests looser packing, which is consistentwith the absence of notable ordered structures.Two main effects have been shown to be responsible for the

ΔH of protein unfolding: the hydration of the buried hydro-phobic and polar groups that become exposed in the unfoldedstate, and the disruption of internal interactions such as van derWaals interactions, and H bonds (21, 24). The magnitude of theΔH of amyloid fibrils (normalized by protein concentration) wassignificantly less than that of the folding of native β2m (Fig. 6and Table S1). The ΔH values for amorphous aggregation wereeven smaller in intensity. From the observed similarity of theΔCp values, we assumed a similar contribution of the hydrationof the buried groups between native and fibril conformations.Therefore, the observed decrease in ΔH appeared to be the re-sult of different internal interactions (28). It is generally ac-cepted that there is a stronger and more persistent backboneH-bond network in the amyloid structure than there is in theglobular fold of proteins, leading to an increase in the β-sheetcontent (3, 12, 43). However, H bonds should increase themagnitude of the ΔH value, which is inconsistent with the results.Thus, a reasonable explanation for the ΔH order in magnitude

of “native structure > amyloid fibril > amorphous aggregate” isthat it dominantly represents side-chain packing in folded ormisfolded structures (Fig. 6B). The overall side-chain packing inthe amyloid form cannot be as optimal as that in the native statebecause the structure is determined by extensively H-bondedβ-structured backbones (11, 12). The loss of tight packing may bemore serious for amorphous aggregation.The separation of overall stability of amyloid fibrils (ΔG) into

the ΔH and TΔS terms illustrates that the contributions of thetwo terms vary depending on temperature. Fibrillation is de-termined by the favorable entropic term at ∼20 °C at which ΔH isclose to zero. ΔG is minimal at ∼35 °C, at which the fibrils exhibitmaximal stability and, thus, ΔS is zero because ΔS = ∂ΔG/∂T.Fibrillation is then determined by the favorable enthalpy term at35 °C. Thus, the temperature-dependent enthalpy–entropy in-terplay determines the stability of amyloid fibrils. To understandthis interplay, we have to estimate amyloid-specific factors suchas the entropy loss resulting from a rigid H bonding of backbonesand a reduction in the number of monomers as well as the en-thalpy gain obtained from numerous molecular contacts.In conclusion, we showed that quantitative calorimetric anal-

ysis with ITC was indeed possible for the supersaturation-limitedamyloid fibrillations. Stirring inside the ITC cell can break per-sistent supersaturation, which triggers fibrillation. Comparedwith the single crystals of substances, amyloid fibrils retain a thinand linear morphology. Moreover, the shear forces of stirring

Fig. 6. Thermodynamic characterization of the folding and misfolding of β2m. (A) Temperature dependencies of ΔH for amorphous aggregation (bluecircles), spontaneous (red triangles) and seed-dependent fibrillations (black squares), and folding (black circles). (B) The difference in ΔH at 60 °C among thedifferent conformational states is illustrated. (C) Temperature dependencies of ΔG (curves), ΔH (solid straight lines), and −TΔS (dotted straight lines) forfolding to the native state (black lines and circles), spontaneous amyloid fibrillation (red lines and triangles), and amorphous aggregation (blue lines andcircles). The ΔG values at 37 °C are shown with the closed symbols. The sign of the ordinate is opposite to that in A to compare the profiles with the standardprofile of protein unfolding.

6658 | www.pnas.org/cgi/doi/10.1073/pnas.1322602111 Ikenoue et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 17

, 202

0

keep fibrils dispersed in solution and fragment fibrils, whichaccelerate seed-dependent propagation. These enabled accu-rate calorimetric measurements of the amyloid burst, makingthe thermodynamic characterization of fibrillation possible.By carefully adjusting these conditions, we can also monitorthe heat of amorphous aggregation. Accordingly, ITC willbecome a promising approach for clarifying the thermody-namic properties of protein aggregates.

Materials and MethodsAssays of Amyloid Fibrils. Expression and purification of human β2m aredescribed in SI Text. The formation of fibrils and amorphous aggregateswas characterized by various methods including ThT fluorescence, AFM,CD, and ELISAs. The details are described in SI Text.

ITC Measurements. ITC measurements for the spontaneous fibrillization ofβ2m at 0.3–6.7 mg·mL−1 dissolved in 10 mM HCl solution (pH 2.5) wereperformed with a VP-ITC instrument (GE Healthcare) at the desired tem-peratures (31–43 °C). The consecutive injections of 20 μL of the 10 mM HClsolution containing 1 M NaCl in the syringe into the β2m solution in the cellwere conducted following a 60-min initial delay for complete equilibration.To minimize the heat effects caused by the difference in temperature, theconsecutive injections were required because the temperature of the solu-tion inside injection syringe was not controlled except 20 μL in the needle.

The first titration of 2 μL was adopted to minimize the influence of residualbubbles and imperfect solution filling the syringe. Nine salt titrations intotal, spaced at intervals of 900 s, were performed with a duration of 4 s forthe first titration and 40 s for the others to reach the final NaCl concen-tration of 100 mM. Changes in the heat flow in microcalories per secondwere monitored in real time with 10 μcal·s−1 of reference power. The re-action cell was continuously stirred at 600 rpm. Lag time was defined bya period between the time starting the measurement under stirring andthe time of major heat effect occurred as shown in Fig. 3A. To examinethe effects of the stirring speed on fibrillation, the stirring speedwas changed from 200 to 1,000 rpm. To monitor amorphous aggrega-tion, 3.5 mg·mL−1 β2m in 10 mM HCl solution without salt was inverselytitrated into 10 mM HCl solution containing 1.0 M NaCl at the desiredtemperatures (31–43 °C). The parameters for ITC measurements exceptfor shortening the initial delay to 30 min were identical to those used forfibril formation. The total heat effects, which were shown to be equal tothe ΔH values, were calculated using peak areas after subtracting theheat of dilution and baseline corrections.

ACKNOWLEDGMENTS. We thank Ms. Kyoko Kigawa for the expression andpurification of β2m. This work was supported by the Japanese Ministry ofEducation, Culture, Sports, Science and Technology. J.K. is supported byBolyai János fellowship of the Hungarian Academy of Sciences and HungarianScientific Research Fund (OTKA; Grant 81950).

1. Sipe JD, et al. (2012) Amyloid fibril protein nomenclature: 2012 recommendationsfrom the Nomenclature Committee of the International Society of Amyloidosis.Amyloid 19(4):167–170.

2. Jucker M, Walker LC (2013) Self-propagation of pathogenic protein aggregates inneurodegenerative diseases. Nature 501(7465):45–51.

3. Hoshino M, et al. (2002) Mapping the core of the β2-microglobulin amyloid fibril by H/Dexchange. Nat Struct Biol 9(5):332–336.

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

5. Sawaya MR, et al. (2007) Atomic structures of amyloid cross-β spines reveal variedsteric zippers. Nature 447(7143):453–457.

6. Eisenberg D, Jucker M (2012) The amyloid state of proteins in human diseases. Cell148(6):1188–1203.

7. Bemporad F, Chiti F (2012) Protein misfolded oligomers: Experimental approaches,mechanism of formation, and structure-toxicity relationships. Chem Biol 19(3):315–327.

8. Tycko R, Wickner RB (2013) Molecular structures of amyloid and prion fibrils: Con-sensus versus controversy. Acc Chem Res 46(7):1487–1496.

9. Laganowsky A, et al. (2012) Atomic view of a toxic amyloid small oligomer. Science335(6073):1228–1231.

10. Fändrich M, Dobson CM (2002) The behaviour of polyamino acids reveals an inverseside chain effect in amyloid structure formation. EMBO J 21(21):5682–5690.

11. Chatani E, Kato M, Kawai T, Naiki H, Goto Y (2005) Main-chain dominated amyloidstructures demonstrated by the effect of high pressure. J Mol Biol 352(4):941–951.

12. Lee YH, Chatani E, Sasahara K, Naiki H, Goto Y (2009) A comprehensive model forpacking and hydration for amyloid fibrils of β2-microglobulin. J Biol Chem 284(4):2169–2175.

13. Jarrett JT, Lansbury PT, Jr. (1993) Seeding “one-dimensional crystallization” ofamyloid: A pathogenic mechanism in Alzheimer’s disease and scrapie? Cell 73(6):1055–1058.

14. Naiki H, et al. (1997) Establishment of a kinetic model of dialysis-related amyloid fibrilextension in vitro. Amyloid 4:223–232.

15. Wetzel R (2006) Kinetics and thermodynamics of amyloid fibril assembly. Acc ChemRes 39(9):671–679.

16. Morris AM, Watzky MA, Finke RG (2009) Protein aggregation kinetics, mechanism,and curve-fitting: A review of the literature. Biochim Biophys Acta 1794(3):375–397.

17. Kitayama H, et al. (2013) A common mechanism underlying amyloid fibrillation andprotein crystallization revealed by the effects of ultrasonication. Biochim BiophysActa 1834(12):2640–2646.

18. Lin Y, Lee YH, Yoshimura Y, Yagi H, Goto Y (2014) Solubility and supersatura-tion-dependent protein misfolding revealed by ultrasonication. Langmuir 30(7):1845–1854.

19. Yoshimura Y, et al. (2012) Distinguishing crystal-like amyloid fibrils and glass-likeamorphous aggregates from their kinetics of formation. Proc Natl Acad Sci USA109(36):14446–14451.

20. Ciryam P, Tartaglia GG, Morimoto RI, Dobson CM, Vendruscolo M (2013) Widespreadaggregation and neurodegenerative diseases are associated with supersaturatedproteins. Cell Rep 5(3):781–790.

21. Makhatadze GI, Privalov PL (1995) Energetics of protein structure. Adv Protein Chem47:307–425.

22. Leavitt S, Freire E (2001) Direct measurement of protein binding energetics by iso-

thermal titration calorimetry. Curr Opin Struct Biol 11(5):560–566.23. Ladbury JE, Williams MA (2004) The extended interface: Measuring non-local effects

in biomolecular interactions. Curr Opin Struct Biol 14(5):562–569.24. Privalov PL (2007) Thermodynamic problems in structural molecular biology. Pure

Appl Chem 79(8):1445–1462.25. Vermeer AW, Norde W (2000) The thermal stability of immunoglobulin: Unfolding

and aggregation of a multi-domain protein. Biophys J 78(1):394–404.26. Benjwal S, Verma S, Röhm KH, Gursky O (2006) Monitoring protein aggregation

during thermal unfolding in circular dichroism experiments. Protein Sci 15(3):635–639.27. Dzwolak W, Ravindra R, Lendermann J, Winter R (2003) Aggregation of bovine in-

sulin probed by DSC/PPC calorimetry and FTIR spectroscopy. Biochemistry 42(38):

11347–11355.28. Kardos J, Yamamoto K, Hasegawa K, Naiki H, Goto Y (2004) Direct measurement of

the thermodynamic parameters of amyloid formation by isothermal titration calo-

rimetry. J Biol Chem 279(53):55308–55314.29. Sasahara K, Naiki H, Goto Y (2005) Kinetically controlled thermal response of β2-

microglobulin amyloid fibrils. J Mol Biol 352(3):700–711.30. Sasahara K, Yagi H, Naiki H, Goto Y (2009) Thermal response with exothermic effects

of β2-microglobulin amyloid fibrils and fibrillation. J Mol Biol 389(3):584–594.31. Jeppesen MD, Hein K, Nissen P, Westh P, Otzen DE (2010) A thermodynamic analysis

of fibrillar polymorphism. Biophys Chem 149(1-2):40–46.32. Morel B, Varela L, Conejero-Lara F (2010) The thermodynamic stability of amyloid

fibrils studied by differential scanning calorimetry. J Phys Chem B 114(11):4010–4019.33. Raman B, et al. (2005) Critical balance of electrostatic and hydrophobic interactions is

required for β2-microglobulin amyloid fibril growth and stability. Biochemistry 44(4):

1288–1299.34. Kad NM, et al. (2003) Hierarchical assembly of β2-microglobulin amyloid in vitro re-

vealed by atomic force microscopy. J Mol Biol 330(4):785–797.35. Giehm L, Otzen DE (2010) Strategies to increase the reproducibility of protein fi-

brillization in plate reader assays. Anal Biochem 400(2):270–281.36. Ohhashi Y, Kihara M, Naiki H, Goto Y (2005) Ultrasonication-induced amyloid fibril

formation of β2-microglobulin. J Biol Chem 280(38):32843–32848.37. Chatani E, et al. (2009) Ultrasonication-dependent production and breakdown lead to

minimum-sized amyloid fibrils. Proc Natl Acad Sci USA 106(27):11119–11124.38. So M, et al. (2011) Ultrasonication-dependent acceleration of amyloid fibril forma-

tion. J Mol Biol 412(4):568–577.39. Yoshimura Y, So M, Yagi H, Goto Y (2013) Ultrasonication: An efficient agitation for

accelerating the supersaturation-limited amyloid fibrillation of proteins. Jpn J Appl

Phys 52:07HA01-01–07HA01-08.40. Come JH, Fraser PE, Lansbury PT, Jr. (1993) A kinetic model for amyloid formation in

the prion diseases: Importance of seeding. Proc Natl Acad Sci USA 90(13):5959–5963.41. Cabriolu R, Kashchiev D, Auer S (2010) Atomistic theory of amyloid fibril nucleation.

J Chem Phys 133(22):225101.42. Durbin SD, Feher G (1996) Protein crystallization. Annu Rev Phys Chem 47:171–204.43. Lee YH, Goto Y (2012) Kinetic intermediates of amyloid fibrillation studied by hy-

drogen exchange methods with nuclear magnetic resonance. Biochim Biophys Acta

1824(12):1307–1323.

Ikenoue et al. PNAS | May 6, 2014 | vol. 111 | no. 18 | 6659

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Dec

embe

r 17

, 202

0