Unmasking the roles of N- and C-terminal flankingsequences from exon 1 of huntingtin as modulatorsof polyglutamine aggregationScott L. Cricka, Kiersten M. Ruffa, Kanchan Garaia,b, Carl Friedenb,1, and Rohit V. Pappua,1

aDepartment of Biomedical Engineering and Center for Biological Systems Engineering, Washington University in St. Louis, St. Louis, MO 63130;and bDepartment of Biochemistry and Molecular Biophysics, Washington University School of Medicine, St. Louis, MO 63110

Contributed by Carl Frieden, November 1, 2013 (sent for review September 25, 2013)

Huntington disease is caused by mutational expansion of the CAGtrinucleotide within exon 1 of the huntingtin (Htt) gene. Exon 1spanning N-terminal fragments (NTFs) of the Htt protein resultfrom aberrant splicing of transcripts of mutant Htt. NTFs typicallyencompass a polyglutamine tract flanked by an N-terminal 17-residue amphipathic stretch (N17) and a C-terminal 38-residueproline-rich stretch (C38). We present results from in vitro bio-physical studies that quantify the driving forces for and mecha-nisms of polyglutamine aggregation as modulated by N17 andC38. Although N17 is highly soluble by itself, it lowers thesaturation concentration of soluble NTFs and increases the drivingforce, vis-à-vis homopolymeric polyglutamine, for forming insolu-ble aggregates. Kinetically, N17 accelerates fibril formation anddestabilizes nonfibrillar intermediates. C38 is also highly solubleby itself, and it lends its high intrinsic solubility to lower the driv-ing force for forming insoluble aggregates by increasing the satu-ration concentration of soluble NTFs. In NTFs with both modules,N17 and C38 act synergistically to destabilize nonfibrillar inter-mediates (N17 effect) and lower the driving force for forming in-soluble aggregates (C38 effect). Morphological studies show thatN17 and C38 promote the formation of ordered fibrils by NTFs.Homopolymeric polyglutamine forms a mixture of amorphousaggregates and fibrils, and its aggregation mechanisms involveearly formation of heterogeneous distributions of nonfibrillar spe-cies. We propose that N17 and C38 act as gatekeepers that controlthe intrinsic heterogeneities of polyglutamine aggregation. Thisprovides a biophysical explanation for the modulation of in vivoNTF toxicities by N17 and C38.

phase separation | subsaturation | supersaturation |tetramethyl rhodamine

Huntington disease (HD) is caused by mutational expansionof the CAG trinucleotide within exon 1 of the huntingtin

(Htt) gene (1). Mutations are translated as polyglutamine ex-pansions within the Htt protein. Neuronal intranuclear inclusionsare the pathological hallmarks of HD, and N-terminal fragments(NTFs) spanning exon 1 of the Htt protein are the main con-stituents of these inclusions (2). The Htt gene with expandedCAG tracts can undergo erroneous splicing, and the resultantaberrant messenger RNA is translated into a mutant exon 1version of Htt that is similar to toxic NTFs found in neuronalintranuclear inclusions (3). Exon 1 spanning NTFs typically in-clude a polyglutamine tract that is flanked on its N terminus byan amphipathic 17-residue stretch (MATLEKLMKAFESLKSF)denoted as N17 and by a 38-residue proline-rich stretch on itsC terminus (P11-QLPQPPPQAQPLLPQPQ-P10) denoted asC38. The N17 sequence is conserved among higher mammals (SIAppendix, Fig. S1), and mutations within N17 impact the prop-erties of NTFs (4, 5). N17 enhances the overall rate of aggre-gation, as measured by the rate of forming large insoluble speciesboth in vitro (6) and in yeast (7). The C-terminal proline-richregion of exon 1 modulates polyglutamine aggregation and reduces

the cellular toxicity of Htt exon 1 even when the polyglutaminetract is significantly expanded (8, 9).A molecular-level understanding of the synergy between the

length of polyglutamine tracts and its flanking sequences is es-sential for inferring the roles of N17 and C38 in vivo. This requiresa quantitative understanding of the driving forces, mechanisms,and morphologies for homopolymeric polyglutamine and theirmodulation by N17 and C38. Here, we report results from in vitrostudies that use solubility measurements to quantify drivingforces, kinetic assays to investigate aggregation mechanisms, andelectron microscopy (EM) to study aggregate morphologies.

ResultsSaturation Concentrations. Aqueous milieus are poor solvents forpolyglutamine (10). In poor solvents, there exists a thresholdconcentration for polymers, c = cs, known as the saturation con-centration (11). For concentrations that exceed cs the polymerplus solvent system separates into insoluble and soluble phases.At c = cs the chemical potentials of soluble and insoluble phasesare equal providing the system is in thermodynamic equilibrium.The lower the value of cs, the stronger the driving force for ag-gregation and phase separation. Aggregation is a generic termthat refers to intermolecular associations that give rise to speciesas small as dimers or aggregates that are large enough to besedimentable. Conversely, phase separation refers to the phe-nomenon that leads to two distinct phases of distinct polymerdensities. We used a micro-Bicinchoninic Acid (BCA) assay (12)to estimate cs as a function of temperature for 10 different

Significance

How do N- and C-terminal flanking sequences from exon 1 ofthe huntingtin protein modulate the mechanisms of polyglut-amine aggregation? We answer this question using approachesthat combine distinct probes of aggregation mechanisms withmeasurements of solubility and aggregate morphologies. TheN- and C-terminal flanking sequence modules from exon 1 ofhuntingtin act as gatekeepers, whereby the N-terminal flankingsequence accelerates fibril formation while destabilizing non-fibrillar species, whereas the C-terminal flanking sequencereduces the overall driving force for aggregation. These re-sults provide a mechanistic underpinning for observations re-garding naturally occurring sequence contexts as modulatorsof polyglutamine toxicity.

Author contributions: S.L.C., K.M.R., K.G., C.F., and R.V.P. designed research; S.L.C., K.M.R.,and K.G. performed research; S.L.C., K.M.R., and K.G. contributed new reagents/analytictools; S.L.C., K.M.R., and R.V.P. analyzed data; and S.L.C., K.M.R., and R.V.P. wrote the paper.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.1To whom correspondence may be addressed. E-mail: pappu@wustl.edu or frieden@biochem.wustl.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1320626110/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1320626110 PNAS | December 10, 2013 | vol. 110 | no. 50 | 20075–20080

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

peptides. The sequence constructs for these cs measurementswere (Gln)n-(Lys)2, (Lys)2-(Gln)n-(Lys)2, N17-(Gln)n-(Lys)2,(Gln)n-C38, and N17-(Gln)n-C38, and these are abbreviated asQn-K2, K2-Qn-K2, N17-Qn-K2, Qn-C38, and N17-Qn-C38, re-spectively. In each of the constructs, the polyglutamine tract iseither 30 or 40 residues long; i.e., n = 30 or 40. The choice of nbrackets the mean threshold length that is relevant to the age-of-onset of HD. Each peptide was disaggregated using establishedprotocols (13), and different concentrations of the peptide werequiescently incubated at a specified temperature in 50 mMphosphate buffer at pH 7 for a period of 2 wk. The material wasthen centrifuged to separate it into a supernatant and pellet, andthe concentration in the supernatant was measured to estimate thecs for the peptide in question at the specified temperature. Usingthis method we can demonstrate that cs is a thermally reversiblequantity as shown in SI Appendix, Fig. S2. Additionally, akin tobrain-derived insoluble inclusions of Htt NTFs, the pellets stainpositively with the fluorescent dye thioflavin T (ThT) (SI Appendix,Fig. S3). The poor solubility of polyglutamine requires the use oflysine residues to facilitate in vitro measurements of syntheticpeptides (13). Data for K2-Qn-K2 are shown in the SI Appendix,Fig. S4, and they help quantify the effect of adding two additionallysine residues on cs values. Clearly, it is important to reduce thenumber of charged residues in synthetic constructs to ensure thattheir influence on the driving forces for aggregation is minimized(14, 15). Therefore, we selected Qn-K2 as the preferred mimic ofhomopolymeric polyglutamine.Fig. 1 shows the measured cs values for Qn-K2, N17-Qn-K2,

Qn-C38, and N17-Qn-C38 at 30 °C and 40 °C. There is a fourfoldreduction in cs for Q40-K2 versus Q30-K2 demonstrating that themagnitude of the driving force for forming insoluble aggregates

increases with polyglutamine length. Fig. 1 also shows the csvalues for N17-Qn-K2 and Qn-C38. The cs value for N17-Q40-K2is in the submicromolar range for T ≤ 40 °C. At 40 °C the cs valuefor N17-Q30-K2 is fourfold smaller than that of Q30-K2. Clearly,N17 lowers the cs of polyglutamine-containing peptides. Incontrast, we obtain a systematic increase in cs for Qn-C38 com-pared with the cs values for Qn-K2. The intrinsic solubilities ofthe N17 and C38 modules are in the millimolar range. Hence,the coupling between N17 and polyglutamine lowers the overallsolubility to be below that of the individual modules, whereasC38 acts as a solubilizing module by increasing the cs of Qn-C38vis-à-vis Qn-K2. Fig. 1 also shows measured cs values of N17-Qn-C38 peptides, which mimic exon 1 spanning NTFs. This con-struct combines the cs lowering effect of N17 and the cs elevatingeffect of C38. The cs values for N17-Qn-C38 increase relative toQn-K2, implying a stronger contribution from the solubilizingeffects of C38 compared with the cs diminishing effects of N17.Finally, in all of the constructs, cs decreases with increased pol-yglutamine length and increases with increased temperature (SIAppendix, Fig. S5).We have assumed that thermodynamic equilibrium between

soluble and insoluble phases can be established within a 2-wkincubation period. The following criteria justify this assumption.First, we test for reliability by assessing if similar values of cs arereproduced when we incubate different amounts of startingmaterial. Second, in all cases, visual inspection showed the de-velopment of precipitate well within the 2-wk incubation period.Third, all kinetics measurements (see below) show a plateauingof signals on time scales that are roughly an order of magnitudefaster than the 2-wk incubation period.

Supersaturation. We measured the kinetics of aggregation ofdifferent constructs using two different assays. Our goal was tocompare the mechanisms of aggregation for polyglutamineconstructs with and without the N17 and C38 modules. Thesestudies require that kinetics experiments be performed underconditions where the magnitudes of driving forces for aggrega-tion are equivalent for all constructs. The concept of supersat-uration is useful in this regard, and the measured values of cs canbe used to define the degree of supersaturation, S. For a givenconstruct, if we denote the bulk concentration of fully dis-

aggregated monomers as co, then S= ln�cocs

�. If co > cs, then the

uniformly mixed solution of monomers is metastable and su-persaturated with respect to the soluble phase. The parameter Squantifies the degree of metastability of the predominantly mo-nomeric solution and hence provides an estimate of the magni-tude of the driving force for aggregation and phase separation(16). If co > cs, then S > 0, and the soluble phase is supersatu-rated, whereas if co < cs, then S < 0, and the soluble phase issubsaturated. When comparing the aggregation mechanisms ofdifferent constructs, we performed experiments at equivalent su-persaturation values to ensure that comparisons between con-structs are made for equivalent magnitudes of driving forces.This helps unmask mechanistic differences that are otherwisedifficult to resolve if measurements are made at equivalent val-ues of co rather than equivalent S values.

Comparative Kinetics of Aggregation for Different Constructs Measuredat Similar Supersaturation Values. For a given construct, the csvalues decrease with increasing polyglutamine length. The focushere is on the effects of flanking sequence modules. Accordingly,we present results for Q30 peptides in four different sequencecontexts. Although the magnitudes of driving forces for aggre-gation increase, and the time scales for aggregation decrease,with increased polyglutamine length, our conclusions regardingcomparative differences in mechanisms for different constructsshould remain similar.

30ºC 40ºC0

5

10

15

20

25

30

35

c s (μ

M)

30ºC 40ºC0

2

4

6

8

10

12

14

16

18

20

c s (μ

M)

N17−Q30−K2Q30−K2N17−Q30−C38Q30−C38

N17−Q40−K2Q40−K2N17−Q40−C38Q40−C38

Fig. 1. Estimates of cs for Qn-K2, N17-Qn-K2, Qn-C38, and N17-Qn-C38 (n =30, 40) at 30 °C and 40 °C.

20076 | www.pnas.org/cgi/doi/10.1073/pnas.1320626110 Crick et al.

We used two peptides, Q30-K2K*G and N17-Q30-K2K*G, tomonitor the rate of loss of monomers into growing aggregates.Here, G denotes glycine and K* denotes a lysine residue thatwas modified by covalent attachment of tetramethyl rhodamine(TMR) through the amine. We assume that saturation concen-trations for Q30-K2K*G and N17-Q30-K2K*G do not deviateappreciably from those of the unlabeled molecules. This is rea-sonable because the free energy of TMR dimerization is ∼2.8kJ /mol (17), which is minimal compared with the strong inter-actions between polyglutamine molecules that give rise to low csvalues (SI Appendix, Fig. S6).A pair of TMR molecules can stack to form fluorescently dark

dimers. The percent probability of TMR forming a dimer isgoverned by the local concentration of other TMR molecules,which in micromolar aqueous solutions of TMR molecules willbe less than 1% based on the known dissociation constant.However, if aggregation prone molecules, each with a singleTMR label, were to form aggregates, then the probability ofmaking TMR dimers increases because of the increased localconcentration of TMR molecules that results from aggregation.Garai and Frieden (18) developed an assay that uses the loss ofTMR fluorescence of labeled molecules to monitor aggregation,thus obviating the requirement that aggregates have to be speciesthat are capable of binding ThT. This method affords highertemporal resolution of the aggregation process, especially withregard to the kinetic details of the early stages. Loss of fluores-cence requires TMR dimerization either on the surface or withinthe interior of aggregates. Our computational analysis indi-cates that aggregates with approximately 20 Q30-K2K*G mole-cules will be 65% as bright as an individual molecule (SI Appendix,Fig. S7).All measurements of the rate of change of TMR fluorescence

were initiated after disaggregation by dissolution of the peptidesin 100% formic acid. These peptides were then diluted intowater at twice the target concentration followed by further di-lution to pH 7 in a 50-mM phosphate buffer. The measurementsof TMR fluorescence were performed at 37 °C under continuousstirring. The disaggregation protocol had to be redesigned vis-à-vis the standard protocol (13), established for use with K2-Qn-K2peptides; the details are discussed in SI Appendix, section 2. SIAppendix, Fig. S8 shows the normalized TMR fluorescence forQ30-K2K*G measured as a function of time at 37 °C for threedifferent supersaturation values that range from S ∼ 0.1 to S ∼0.8. SI Appendix, Fig. S9 shows data for the rate of loss of TMR

fluorescence for the N17-Q30-K2K*G peptide at four differentsupersaturation values ranging from S ∼ 0.8 to S ∼ 2.2. The ki-netic traces for the loss of TMR fluorescence of polyglutaminepeptides show the absence of a lag phase, and the profiles arerather distinct from those for amyloid beta peptides that werestudied using a similar approach (18).Fig. 2A compares the kinetics for the loss of TMR fluores-

cence measured for Q30-K2K*G and N17-Q30-K2K*G at equiv-alent supersaturation values, S ∼ 0.7–0.8. The time taken toachieve a normalized TMR fluorescence of 0.5 is approximatelyfourfold higher for N17-Q30-K2K*G compared with Q30-K2K*G.To further probe differences in aggregation mechanisms wecompared the kinetics for the loss of TMR fluorescence to thekinetics of the gain in ThT fluorescence. The results are shownin Fig. 2 B and C for Q30-K2K*G and N17- Q30-K2K*G, re-spectively. N17 clearly accelerates fibril formation. This accel-eration appears to derive from the destabilization of nonfibrillarintermediates—a conjecture that is supported by the compara-tive analysis of the kinetics of the loss of TMR fluorescence tothe gain of ThT fluorescence shown in Fig. 2C. This figure also

0 5 10 15 20 25 30 35 40 45 50

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

t (hrs)

Nor

mal

ized

TM

R fl

uore

scen

ce

Q30-K2K*GN17-Q30-K2K*G

0 5 10 15 20 25 30 35 40 45 50

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Exte

nt o

f rea

ctio

n

Q30-K2: ThTQ30-K2K*G: TMRComplement of TMR

0 5 10 15 20 25 30 35 40 45 50

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

N17-Q30-K2: ThTN17-Q30-K2K*G: TMRComplement of TMR

t (hrs) t (hrs)

Exte

nt o

f rea

ctio

n

A B C

Fig. 2. (A) Comparison of the kinetics of the loss of TMR fluorescence for Q30-K2K*G and N17-Q30-K2K*G. For each construct, we normalized the TMRfluorescence using the t = 0 value as a reference. (B) Comparison of extents of aggregation obtained by following the loss in TMR fluorescence and gain in ThTfluorescence for Q30-K2K*G and Q30-K2, respectively. The extents of reaction range from 0 to 1 and were calculated using the minimum and maximum valuesfor TMR and ThT fluorescence. The black points correspond to the complement of the TMR data, whereby we convert the loss of TMR fluorescence intoa growth curve to compare this with the kinetics of the gain in ThT fluorescence. (C) Comparison of extents of aggregation obtained by following the loss inTMR fluorescence and gain in ThT fluorescence for N17-Q30-K2K*G and N17-Q30-K2, respectively. All other details are similar to that of B.

0 5 10 15 20 25 30 35 40 45 500

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

t (hrs)

Exte

nt o

f rea

ctio

n

Q30-K2: co=10 μM, S≈0.8N17-Q30-K2: co=2 μM, S≈0.7Q30-C38: co=50 μM, S≈0.7N17-Q30-C38: co=12 μM, S≈0.7

Fig. 3. Comparative analysis of the kinetics of gain in ThT fluorescence forfour different Q30 constructs.

Crick et al. PNAS | December 10, 2013 | vol. 110 | no. 50 | 20077

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

shows the expected profile for the rate of growth of aggregatesthat is obtained by computing the complement of the normalizedprofile for the loss of TMR fluorescence. This computed curveessentially coincides with the profile for the gain in ThT fluo-rescence profile, thereby implying that the loss of TMR fluo-rescence results almost entirely from fibril formation. In directcontrast, the loss in TMR fluorescence for Q30-K2K*G reachesa plateau value before there is discernible increase in ThTfluorescence (Fig. 2B). This implies a separation of time scalesbetween two processes, namely, preequilibration of nonfibrillarspecies before their conversion into fibrillar aggregates. The timescales for the two processes, quantified as the times for theextents of distinct reactions to reach 50% completion, differ by atleast an order of magnitude for Q30-K2K*G.To measure the rate of loss of TMR fluorescence for Q30-C38

and N17-Q30-C38 peptides at equivalent supersaturation valueswe would need TMR-labeled material at concentrations whereinner filter effect confounds the interpretations of TMR fluo-rescence measurements (19). In addition, if we followed theprotocol used for N17-Qn-K2K*G and Qn-K2K*G by labelingthe C-terminal end, then we would be probing the likelihood ofassociations between the C38 modules as opposed to just thepolyglutamine-mediated associations. We need detailed simu-lations to help guide the optimal placement of the TMR labels inconstructs with the C38 module. These simulations are feasiblewith improvements to forcefields that describe proline-rich se-quences and will be pursued in future work. Here, we focus oncomparing the kinetics of changes in ThT fluorescence for allfour constructs at equivalent supersaturation values.Fig. 3 presents our comparative analysis of data obtained for

the extents of reaction derived from ThT fluorescence mea-surements as a function of time at 37 °C for the peptides, Q30-K2,N17-Q30-K2, Q30-C38, and N17-Q30-C38, respectively; S ∼ 0.7–0.8 for all peptides. Although C38 lowers the cs value and hencedecreases the magnitude of the driving force for aggregation,at equivalent supersaturation values the kinetic profile for the

change in ThT fluorescence for N17-Q30-C38 is essentiallyidentical to that of N17-Q30-K2. This suggests that even in thepresence of the solubilizing C38 module the mechanism of ag-gregation is controlled by N17, which accelerates fibril formationwhile destabilizing nonfibrillar intermediates. In contrast, theincrease in ThT fluorescence for Q30-C38 shows a significant lagphase, which is also present for Q30-K2.

Impact of Flanking Sequences on Morphologies. We asked if theobserved differences in aggregation kinetics at equivalent su-persaturation values lead to differences in the morphologies ofaggregates. Fig. 4 shows the morphologies obtained using EMfollowing long-time incubations of Q30-K2, N17-Q30-K2, Q30-C38, and N17-Q30-C38. All data were collected following ag-gregation in supersaturated solutions, with S ∼ 0.8. Additionalassessments of morphologies were obtained by imaging the endproducts of species obtained from TMR assays for Q30-K2K*Gand N17-Q30-K2K*G (SI Appendix, Fig. S10).Q30-K2 forms a dense mesh of aggregates with truncated fibrils

(Fig. 4). The rapid formation of nonfibrillar species by Q30-K2appears to contribute to the disordered morphologies of higher-order aggregates. In contrast, the representative images in Fig. 4show that N17-Q30-K2, Q30-C38, and N17-Q30-C38 form orderedfibrils with minimal nonspecific associations. Comparison of themorphologies for N17-Q30-K2 and Q30-C38 at equivalent magni-fications suggests that these modules get differently incorporatedinto the fibrils leading to differences in morphologies. These dif-ferences are consistent with published studies, which suggest thatC38 contributes to the formation of bottlebrush architectures (20).

Aggregates Form in Subsaturated Solutions. The homogeneousnucleation model (16) has been used as a conceptual frameworkfor explaining the kinetics of polyglutamine aggregation (21). Inthis model, an obligatory nucleus representing an embryo of thenew aggregated/insoluble phase has to form within the homo-geneously mixed, soluble phase. The free energy of nucleationdepends on the supersaturation value and the free-energy pen-alty associated with creating an interface between the new andold phases (16). Supersaturation is a prerequisite for a nucleatedprocess. It therefore follows that monomers have to be thepredominant species in subsaturated solutions either becauseaggregates never form or because they are only marginally stable.To assess the applicability of homogeneous nucleation theory,we quantified the kinetics of the loss of TMR fluorescence in

Q30-K210 M, S 0.8

N17-Q30-K22 M, S 0.7

N17-Q30-C3812 M, S 0.7

Q30-C3850 M, S 0.7

Fig. 4. Comparison of aggregate morphologies for different Q30 constructsobtained using EM. Additional comparisons using TMR-labeled constructsare shown in SI Appendix, Fig. S10. Visual inspection suggests that althoughN17 promotes long, ordered fibrils, these fibrils are thinner than thoseobtained for constructs with the C38 module.

0 5 10 15 20 25 30 35 40 450

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

t (hrs)

Nor

mal

ized

TM

R fl

uore

scen

ce

Q30−K2K*G, S ≈ −2.2

N17−Q30−K2K*G, S ≈ −0.8Q30−C38K*G, S ≈ −2.1

Q30−K2K*G, S ≈ −1.5Q30−K2K*G, S ≈ −0.6

Fig. 5. Comparing the kinetics of loss of TMR fluorescence in subsaturatedsolutions for Q30 constructs with and without N17 and C38.

20078 | www.pnas.org/cgi/doi/10.1073/pnas.1320626110 Crick et al.

subsaturated solutions. Fig. 5 shows the results of these mea-surements for Q30-K2K*G, N17-Q30-K2K*G, and Q30-C38K*G.The data in Fig. 5 highlight the presence of stable aggregates thatform without a lag phase on time scales that are comparable tothose seen in supersaturated solutions. Fig. 6 shows a represen-tative EM image of visible, nonfibrillar aggregates formed byQ30-K2K*G in subsaturated solutions. These aggregates arespherical and are roughly 10 nm in size. Increased concentrationof such aggregates leads to the increased loss of TMR fluores-cence that we observe for Q30-K2K*G as the solutions becomeless subsaturated, i.e., as S becomes less negative (Fig. 5). Takentogether, the data in Figs. 5 and 6 provide evidence against theapplicability of homogeneous nucleation for describing the ag-gregation of polyglutamine constructs. Instead the data pointto a heterogeneous process that involves nonfibrillar specieswhose stabilities vary depending on the presence or absence offlanking sequences.Fig. 5 points to the existence of multiple phases in solution

including one that should give rise to a distinct saturation con-centration for soluble aggregates. These could either be liquid-like oligomers that arise from liquid–liquid demixing (22) ormicellar structures that arise due to the possible existence ofa critical micelle concentration. We used the plateau values forthe TMR fluorescence in subsaturated solutions to estimatesaturation concentrations for oligomers of Q30-K2K*G, N17-Q30-K2K*G, and Q30-C38K*G, respectively (SI Appendix, sec-tion 2). We denote the oligomer saturation concentration as cc.It is appropriate to use the C-terminally labeled Q30-C38K*Gpeptide to estimate cc from the plateau value of the TMRfluorescence. This value should be indicative of molecules thatremain unincorporated into aggregates, irrespective of where theTMR label is situated within the construct.The values we obtain for cc at 37 °C in 50 mM phosphate

buffers at pH 7 are 250 nM, 165 nM, and 3 μM for Q30-K2K*G,N17-Q30-K2K*G, and Q30-C38K*G, respectively. Clearly, theoligomer saturation concentration is distinct from and lower

than cs. The ratiocscc quantifies the gap between cc and cs. For T ∼

37 °C this ratio is 18.4, 6.5, and 8.3 for Q30-K2, N17-Q30-K2, andQ30-C38, respectively.

DiscussionFig. 7 shows schematic representations of our proposals for theaggregation mechanisms of homopolymeric polyglutamine andthe modulation of these mechanisms by N17 and C38. Thescheme shown in Fig. 7A is in accord with the data presentedhere for Q30-K2 and Q30-K2K*G and a recent two-stage nucle-ation model for polyglutamine aggregation (23). This modelpostulates the existence of heterogeneous distributions of oligom-ers that arise due to liquid–liquid demixing of intrinsically dis-ordered polyglutamine. For homopolymers in poor solvents,intermolecular interfaces are favored over chain-solvent inter-faces. Accordingly, disordered globules associate in a thermody-namically downhill fashion to form a heterogeneous distributionof liquid-like oligomers (23). The model assumes a slow con-version of oligomers into fibrillar species, providing co > cs. Afinite concentration of suitably large oligomers is required tosupport a nucleated conformational conversion to generate atemplate for fibril elongation (24, 25). In the model, the size ofthe smallest fibril-competent oligomer is denoted as imin. Thedata presented here for Q30-K2 and Q30-K2K*G provide directevidence in support of equilibrium distributions of oligomers thatform in sub- and supersaturated solutions. This step precedesfibril formation in supersaturated solutions, and Fig. 7A proposesthat flux into fibrils is decreased due to the collection of non-fibrillar species that form in the absence of flanking sequences.The model (23) also showed that if imin is decreased, whichimplies destabilization of oligomers that cannot convert to fibrils,then fibril formation is accelerated. In such a scenario, if ag-gregation is modeled as homogeneous nucleation, then the ap-parent nucleus size can become less than or equal to zero (23) aswas shown by Thakur et al. (6) who applied a variant of homo-geneous nucleation to model the kinetics of the loss of solublespecies for N17-Qn-K2 peptides.The schematic in Fig. 7B summarizes the effects of N17 and

C38 on the aggregation of Htt NTFs. We propose that N17 andC38 act as gatekeepers to destabilize nonfibrillar intermediatesthus ensuring that the fluxes are distributed across fewer routes.This schematic is consistent with data presented here and withresults from atomistic simulations (15) which show that N17destabilizes nonspecific associations vis-à-vis homopolymericpolyglutamine. In vitro (6) and cellular measurements (7) showed

100 nm

Fig. 6. Representative EM image of oligomers that form in subsaturatedsolutions of Q30-K2. Here, co = 1 μM and S ∼ –1.5. The black arrows point tosmall oligomers, and the red arrows point to larger oligomers that form viathe aggregation of smaller oligomers. The sizes of oligomers are expected tobe in the 10–50-nm range.

M1

MCS MC

I

FS F I M1

MCS

FS F I

ear aggreg

aggregat

n-K2 n-K2 and Ex1A Schematic for Q B Schematic for N17-Q

Fig. 7. Schematic representations for the distinct aggregation mechanismsexpected for (A) Qn-K2, (B) N17-Qn-K2 and N17-Qn-C38.M1,MS

C ,MIC , F

S, and FI

denote monomers, soluble oligomers, insoluble nonfibrillar aggregates,soluble fibrils, and insoluble fibrils, respectively. The boxed regions delineateinsoluble aggregates. Small arrows within the boxed regions depict equi-libria between insoluble and soluble species as well as equilibria betweendifferent forms of insoluble species. The proposal is that imin, i.e., the size ofsmallest oligomer that can convert to fibrils, is considerably larger for Qn-K2than for polyglutamine constructs with the N17 and C38 flanking sequences.In B, the decrease in imin is depicted as a decrease in the stability of MS

C .Glutamine, hydrophobic residues, positive and negatively charged residues,and proline are denoted using spheres that are orange, yellow, blue, red,and purple, respectively (B).

Crick et al. PNAS | December 10, 2013 | vol. 110 | no. 50 | 20079

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

that N17 accelerates fibril formation. This was taken to meanthat the “aggregation propensity” of homopolymeric polyglut-amine is considerably weaker than that of NTF constructs con-taining both polyglutamine and N17. The data presented hereprovide a more nuanced explanation that connects the simula-tion results to previous experimental observations. Fibrils formslowly and inefficiently for homopolymeric polyglutamine be-cause the gap between cc and cs is largest for these constructs.This gap gives rise to kinetic competition with and thermody-namic destabilization of ordered fibrillar aggregates and increa-ses the overall heterogeneity due to the increased thermodynamicand kinetic stability of nonfibrillar species as sketched in Fig. 7A(26). The use of two separate assays that are designed to probedistinct features of the kinetics of aggregation combined withsolubility and morphological measurements lead to a more com-plete picture of the heterogeneities inherent to polyglutamineaggregation and the modulation of these heterogeneities by N17and C38. The data presented here reconcile results from theory(23), simulations (15), in vitro (6), and in-cell experiments (7).Both N17 and C38 contribute to narrowing the gap between cc andcs thereby destabilizing nonfibrillar intermediates. C38 contrib-utes to increasing cs thus weakening the overall driving force foraggregation whereas N17 accelerates fibril formation and sta-bilizes these species while destabilizing nonfibrillar species.The aggregation mechanisms for polyglutamine-containing

peptides are distinct from mechanisms reported for the amyloidbeta (Aβ) system (18). For Aβ peptides, it has been argued thataggregation can be modeled using a quasi-homogeneous nucle-ation model (11) if co > cs and co < cc (27). Our results show thatthe combination of co > cs and co < cc cannot be achieved forpolyglutamine-containing systems. Even in the presence offlanking sequences the ratio cs

cc remains greater than 1. As a re-sult, in supersaturated solutions, the mechanisms are likely to

always involve heterogeneities such as nonfibrillar oligomers (23,25) that form before or concomitantly with fibrils. It is possiblethat de novo sequence design can be used to generate sequencevariants of Aβ for which cc < cs, as is the case with polyglutamine-containing systems.

Implications for Htt NTFs in Vivo.We propose that N17 and C38 actas natural gatekeepers to minimize the deleterious effects ofheterogeneities that are intrinsic to polyglutamine aggregation.Time-resolved microscopy data from neurons show that diffuseaggregates and smaller species of mutant Htt directly correlatewith neuronal death, whereas the formation of large insolubleinclusions diminishes the levels of diffuse, heterogeneous aggre-gates and the risk of striatal neuron death (28). It is conceivablethat variations in NTFs that result from aberrant splicing of exon 1or proteolytic processing of mutant Htt (29) contribute to varia-tions in the relative levels of small diffuse aggregates versus largeinsoluble inclusions, leading to the observed cell-to-cell variationsin phenotype (28).

Materials and MethodsPeptides were purchased in crude form from Yale University’s Keck Bio-technology center. In addition to the unlabeled peptides used for cs and ThTmeasurements, the labeled peptides were synthesized with the red TMR dyeattached through the lysine amine. SI Appendix, section 2, provides all of thedetails of the preparation and handling of each peptide system. In addition,this section also details the protocols used for measuring cs values, TMR andThT fluorescence, and imaging of aggregates using EM.

ACKNOWLEDGMENTS. We thank Dorothy Beckett, Nicholas Corsepius, MarcDiamond, Tyler Harmon, Alex Holehouse, George Lorimer, M. Muthukumar,Dev Thirumalai, and Andreas Vitalis for insights and helpful discussions. Thiswork was supported by Grant 5R01NS056114 from the National Institutes ofHealth (to R.V.P.).

1. Walker FO (2007) Huntington’s disease. Lancet 369(9557):218–228.2. Becher MW, et al. (1998) Intranuclear neuronal inclusions in Huntington’s disease and

dentatorubral and pallidoluysian atrophy: Correlation between the density of in-clusions and IT15 CAG triplet repeat length. Neurobiol Dis 4(6):387–397.

3. Sathasivam K, et al. (2013) Aberrant splicing of HTT generates the pathogenic exon 1protein in Huntington disease. Proc Natl Acad Sci USA 110(6):2366–2370.

4. Zheng ZQ, Li AM, Holmes BB, Marasa JC, Diamond MI (2013) An N-terminal nuclearexport signal regulates trafficking and aggregation of Huntingtin (Htt) proteinexon 1. J Biol Chem 288(9):6063–6071.

5. Maiuri T, Woloshansky T, Xia J, Truant R (2013) The huntingtin N17 domain isa multifunctional CRM1 and Ran-dependent nuclear and cilial export signal. HumMolGenet 22(7):1383–1394.

6. Thakur AK, et al. (2009) Polyglutamine disruption of the huntingtin exon 1 N terminustriggers a complex aggregation mechanism. Nat Struct Mol Biol 16(4):380–389.

7. Tam S, et al. (2009) The chaperonin TRiC blocks a huntingtin sequence element thatpromotes the conformational switch to aggregation. Nat Struct Mol Biol 16(12):1279–1285.

8. Duennwald ML, Jagadish S, Muchowski PJ, Lindquist S (2006) Flanking sequencesprofoundly alter polyglutamine toxicity in yeast. Proc Natl Acad Sci USA 103(29):11045–11050.

9. Park S-H, et al. (2013) PolyQ proteins interfere with nuclear degradation of cytosolicproteins by sequestering the Sis1p chaperone. Cell 154(1):134–145.

10. Crick SL, Jayaraman M, Frieden C, Wetzel R, Pappu RV (2006) Fluorescence correlationspectroscopy shows that monomeric polyglutamine molecules form collapsed struc-tures in aqueous solutions. Proc Natl Acad Sci USA 103(45):16764–16769.

11. Garai K, Sahoo B, Sengupta P, Maiti S (2008) Quasihomogeneous nucleation of am-yloid beta yields numerical bounds for the critical radius, the surface tension, and thefree energy barrier for nucleus formation. J Chem Phys 128(4):045102.

12. Smith PK, et al. (1985) Measurement of protein using bicinchoninic acid. Anal Bio-chem 150(1):76–85.

13. Chen SM, Wetzel R (2001) Solubilization and disaggregation of polyglutamine pep-tides. Protein Sci 10(4):887–891.

14. Walters RH, Murphy RM (2009) Examining polyglutamine peptide length: A connec-tion between collapsed conformations and increased aggregation. J Mol Biol 393(4):978–992.

15. Williamson TE, Vitalis A, Crick SL, Pappu RV (2010) Modulation of polyglutamineconformations and dimer formation by the N-terminus of huntingtin. J Mol Biol396(5):1295–1309.

16. Kashchiev D, van Rosmalen GM (2003) Review: Nucleation in solutions revisited. CrystRes Technol 38(7-8):555–574.

17. Corsepius NC, Lorimer GH (2013) Measuring how much work the chaperone GroELcan do. Proc Natl Acad Sci USA 110(27):E2451–E2459.

18. Garai K, Frieden C (2013) Quantitative analysis of the time course of Aβ oligomeri-zation and subsequent growth steps using tetramethylrhodamine-labeled Aβ. ProcNatl Acad Sci USA 110(9):3321–3326.

19. Fanget B, Devos O, Draye M (2003) Correction of inner filter effect in mirror coatingcells for trace level fluorescence measurements. Anal Chem 75(11):2790–2795.

20. Bugg CW, Isas JM, Fischer T, Patterson PH, Langen R (2012) Structural features anddomain organization of huntingtin fibrils. J Biol Chem 287(38):31739–31746.

21. Chen SM, Ferrone FA, Wetzel R (2002) Huntington’s disease age-of-onset linked topolyglutamine aggregation nucleation. Proc Natl Acad Sci USA 99(18):11884–11889.

22. Ganazzoli F, Raos G, Allegra G (1999) Polymer association in poor solvents: frommonomolecular micelles to clusters of chains and phase separation. Macromol. The-ory Simul 8(1):65–84.

23. Vitalis A, Pappu RV (2011) Assessing the contribution of heterogeneous distributionsof oligomers to aggregation mechanisms of polyglutamine peptides. Biophys Chem159(1):14–23.

24. Serio TR, et al. (2000) Nucleated conformational conversion and the replication ofconformational information by a prion determinant. Science 289(5483):1317–1321.

25. Lee J, Culyba EK, Powers ET, Kelly JW (2011) Amyloid-β forms fibrils by nucleatedconformational conversion of oligomers. Nat Chem Biol 7(9):602–609.

26. Cheng SZD, Keller A (1998) The role of metastable states in polymer phase transitions:Concepts, principles, and experimental observations. Annu Rev Mater Sci 28:533–562.

27. Auer S, Ricchiuto P, Kashchiev D (2012) Two-step nucleation of amyloid fibrils: om-nipresent or not? J Mol Biol 422(5):723–730.

28. Arrasate M, Finkbeiner S (2012) Protein aggregates in Huntington’s disease. ExpNeurol 238(1):1–11.

29. Landles C, et al. (2010) Proteolysis of mutant huntingtin produces an exon 1 fragmentthat accumulates as an aggregated protein in neuronal nuclei in Huntington disease.J Biol Chem 285(12):8808–8823.

20080 | www.pnas.org/cgi/doi/10.1073/pnas.1320626110 Crick et al.