amyloid and essential cellular functions disrupted

8/7/2019 Amyloid and essential cellular functions disrupted

http://slidepdf.com/reader/full/amyloid-and-essential-cellular-functions-disrupted 1/24

Amyloid-like Aggregates Sequester

Numerous Metastable Proteinswith Essential Cellular FunctionsHeidi Olzscha,1,4 Sonya M. Schermann,1,4 Andreas C. Woerner,1 Stefan Pinkert,1 Michael H. Hecht,2 Gian G. Tartaglia,3,5

Michele Vendruscolo,3 Manajit Hayer-Hartl,1,* F. Ulrich Hartl,1,* and R. Martin Vabulas1,*1Department of Cellular Biochemistry, Max Planck Institute of Biochemistry, Am Klopferspitz 18, D-82159 Martinsried, Germany2Department of Chemistry, Princeton University, Princeton, NJ 08544, USA 3Department of Chemistry, University of Cambridge, Lensfield Road, Cambridge CB2 1EW, UK4These authors contributed equally to this work5Present address: Bioinformatics & Genomics Program, CRG Centre for Genomic Regulation, Dr. Aiguader 88, Barcelona 08003, Spain

*Correspondence: [email protected] (M.H-H.), [email protected] (F.U.H.), [email protected] (R.M.V.)

DOI 10.1016/j.cell.2010.11.050

SUMMARY

Protein aggregation is linked with neurodegeneration

and numerous other diseases by mechanisms that

are not well understood. Here, we have analyzed

the gain-of-function toxicity of artificial b sheet

proteins that were designed to form amyloid-like

fibrils. Using quantitative proteomics, we found that

the toxicity of these proteins in human cells corre-

lates with the capacity of their aggregates to promote

aberrant protein interactions and to deregulate the

cytosolic stress response. The endogenous proteins

that are sequestered by the aggregates sharedistinct physicochemical properties: They are rela-

tively large in size and significantly enriched in

predicted unstructured regions, features that are

strongly linked with multifunctionality. Many of the

interacting proteins occupy essential hub positions

in cellular protein networks, with key roles in

chromatin organization, transcription, translation,

maintenance of cell architecture and protein quality

control. We suggest that amyloidogenic aggregation

targets a metastable subproteome, thereby causing

multifactorial toxicity and, eventually, the collapse

of essential cellular functions.

INTRODUCTION

The majority of proteins must fold into well-defined three-dimen-

sional structures in order to fulfill their biological functions. This

fundamental process is aided by a complex cellular machinery

of molecular chaperones, which act to prevent misfolding and

aggregation ( Frydman, 2001; Hartl and Hayer-Hartl, 2002; Mori-

moto, 2008 ). Failure of a protein to fold properly, or to retain its

folded state, has emerged as the cause of numerous diseases.

Aberrant folding is often the result of destabilizing mutations

and may cause the loss of critical functions. However, in a

growing number of diseases, misfolding and aggregation results

predominantly in a toxic gain of function ( Stefani and Dobson,

2003; Winklhofer et al., 2008 ). In these disorders, specific

proteins, differing substantially in size and sequence, typically

self-assemble into amyloid-like fibrils with cross-b structure

which are deposited within or outside of cells. This phenomenon

underlies some of the most debilitating neurodegenerative

disorders, including Parkinson’s, Huntington’s, and Alzheimer’s

disease.

Amyloidogenic aggregation is observed with many protein

sequences ( Chiti and Dobson, 2006; Goldschmidt et al., 2010 )

and is often associated with the accumulation of soluble, oligo-

meric species that precede fibril formation and are thought to

be responsible for toxicity ( Campioni et al., 2010; Chiti and

Dobson, 2006; Jahn and Radford, 2008 ). The underlying mech-

anisms are only poorly understood but a prominent hypothesis

suggests that the aggregates, in particular the more heteroge-

neous oligomers, may expose flexible hydrophobic surfaces

that can mediate aberrant interactions with other proteins,

resulting in their functional impairment and sequestration

( Bolognesi et al., 2010; Chiti and Dobson, 2006 ). In another

model, misfolding proteins, by engaging the chaperone

machinery, are thought to interfere with central protein quality

control and clearance mechanisms, possibly resulting in a prop-

agation of folding defects ( Balch et al., 2008; Bence et al., 2001;

Gidalevitz et al., 2006 ). Finally, based on experiments with modelmembranes, oligomeric aggregation intermediates can compro-

mise the integrity of lipid membranes ( Lashuel and Lansbury,

2006 ). Importantly, these different routes of toxic action are not

mutually exclusive but may operate in parallel.

To investigate the toxicity mechanisms of amyloid-like aggre-

gation, we have established a cellular model based on the

expression of artificial proteins that were designed to form

b sheet structures, and shown previously to self-assemble into

fibrils in vitro ( West et al., 1999 ). Thesequences of these proteins

were explicitly designed to contain b strands with an alternating

pattern of polar and nonpolar residues, while the exact identities

of the side chains were varied combinatorially. Similar bipolar

Cell 144, 67–78, January 7, 2011 ª2011 Elsevier Inc. 67

mailto:[email protected]



http://dx.doi.org/10.1016/j.cell.2010.11.050

http://dx.doi.org/10.1016/j.cell.2010.11.050






segments occur in $30% of human proteins, including several

neurodegenerative disease proteins, but are usually buried

within the folded structure ( Tartaglia et al., 2008 ). Because the

model proteins were designed de novo, they are not biased by

the evolutionary burden of natural proteins and thus allowed usto study the gain-of-function toxicity caused by aggregation

without interference either from loss-of-function alterations or

from an augmentation of the biological activities of natural

disease proteins ( Cooper et al., 2006; Lam et al., 2006 ).

Here, we tested specifically the hypothesis that the aggre-

gates engage in widespread aberrant protein interactions. We

found that expression of the model proteins in human cells

results in aggregate formation and toxicity. Quantitative proteo-

mic analysis reveals that the aggregates interact with and

sequester multiple preexistent and newly synthesized proteins.

Interestingly, these interactions can be explained in terms of

specific sequence features of the coaggregating proteins, such

as their multidomain character and their enrichment in disor-

dered regions, properties that are strongly linked with multifunc-tionality and the occupancy of hub positions in the cellular

protein network. We suggest that aberrant interactions with

numerous proteins having key cellular functions contribute to

aggregate toxicity.

RESULTS

Designed b Sheet Proteins Are Cytotoxic

To investigate the gain-of-function cytotoxicity associated with

amyloid-like aggregation, we used several model polypeptides

from a combinatorial library rationally designed to form cross-

b fibrils ( West et al., 1999 ). These proteins, henceforth desig-

nated as b proteins, contain six b strands connected by 4-amino

acid linker segments, with each strand comprising seven amino

acids in a polar-nonpolar alternating pattern. An N-terminal

c-Myc-epitope was attached to facilitate detection ( Figure 1 A).

The three proteins chosen for analysis, b4, b17, and b23, differ

in sequence (pairwise identities of b strands $35%), with b23

having the highest hydrophobic volume and b sheet propensity,

due to its higher isoleucine content ( Figure 1 A) ( Tartaglia et al.,

2008 ). As a control, we used the designed a-helical protein,

a-S824, which is similar to the b proteins in amino acid compo-

sition but folds into a 4-helix bundle structure ( Wei et al., 2003 )

( Figure 1 A).

Upon dilution from denaturant into physiological buffer, the

purified b proteins adopted b sheet conformation as determined

by CD and rapidly assembled into aggregates detectable withtheamyloid-binding dyes thioflavin T (ThT) andNIAD-4 ( Nesterov

et al., 2005 ) ( Figures 1B, 1C, and Figures S1 A and S1B available

online). The intensity of ThT and NIAD-4 binding was highest for

b23, followed byb17 and b4 ( Figures 1B and 1C), consistent with

the relative b aggregation propensity of these proteins calcu-

lated with the sequence-based Zagg method ( Tartaglia et al.,

2008 ) (Zagg scores are: b4, 0.79; b17, 0.83; b23, 0.93; a-S824,

and 0.30). ANS fluorescence, a probe for exposed hydrophobic

regions, suggested the presence of hydrophobic surfaces on the

aggregates, in particular for b23 and b17 ( Figure 1D). As shown

via electron microscopy, the b proteins formed mostly relatively

short protofilaments ( $2–3 nm in diameter) as well as more

heterogeneous, globular aggregates ( Figure 1E). Formation of

globular species was most pronounced with b23 ( Figure 1E).

Similar prefibrillar aggregates are also observed with natural

amyloidogenic proteins and correlate with cytotoxicity ( Bolog-

nesi et al., 2010; Campioni et al., 2010; Chiti and Dobson,2006 ). In contrast, in low-salt buffer (pH 6), the b proteins formed

thicker (10–12 nm diameter) and longer fibrils ( Figure 1F) with

FTIR spectroscopic properties characteristic of amyloid ( Fig-

ure S1C). Thus, the model proteins undergo amyloid-like aggre-

gation in vitro, and in physiological buffer populate prefibrillar

species.

Upon expression in HEK293T cells for 3 days, the b proteins,

but not a-S824, reduced cell viability substantially, as measured

by the MTT assay ( Figure 2 A), and induced cell death in the order

b23 > b17 > b4 ( Figures S2 A and S2B). Cell viability was less

impaired after 24 hr of expression, without a significant differ-

ence in toxicity between the three b proteins ( Figure 2 A). Upon

cell fractionation, the b proteins were largely recovered in the

insoluble fraction, whereas a-S824 was soluble ( Figure 2B).Note that b4 and b17 migrated on SDS-PAGE more slowly

than expected, but this difference was not observed in urea/

SDS gels ( Figure S1 A). Confocal immunofluorescence micros-

copy with anti-Myc antibody showedthat theb-protein-express-

ingcells adopted a collapsedshapelackingfilopodia( Figure2C).

Aggregates accumulated mostly in the perinuclear space and

the nuclei were often deformed. The aggregates were NIAD-4

positive ( Figure 2D), suggesting the presence of amyloid-like

material. To detect oligomeric aggregation intermediates, cell

extracts were fractionated by size exclusion chromatography

followed by dot blot analysis with the A11 antibody, which was

raised against the Alzheimer A b peptide and preferentially

recognizes amyloid oligomers associated with cytotoxicity, inde-

pendent of amino acid sequence ( Kayed et al., 2003 ). b23

expression generated substantially higher levels of A11 reactive

material than expression of b4 and b17 ( Figure 2E and Fig-

ure S2C), consistent with the greater toxicity of b23 and its

pronounced tendency to form prefibrillar aggregates in vitro

( Figures 1E and 1F).

In summary, the designed b proteins resemble amyloidogenic

disease proteins in terms of aggregation properties and toxicity

and allow us to investigate the mechanism of gain-of-function

toxicity independently from evolved biological interactions.

Identification of the b Protein Interactome

Gain-of-function toxicity of aggregation may arise, at least in

part, from aberrant interactions of the aggregates with cellularproteins. To test this hypothesis, we performed a sensitive,

quantitative proteomic analysis of the b protein interactome

using SILAC (stable isotope labeling with amino acids in cell

culture) ( Ong and Mann, 2006 ) and peptide identification by

tandem mass spectrometry (LC-MS/MS). These experiments

were performed at 24 hr after b protein transfection when cell

viability was not yet severely impaired ( Figure 2 A). In one set of

experiments, cells labeled with light (L), medium (M) or heavy

(H) arginine and lysine isotopes were transfected with empty

vector, a-S824 and b23, respectively. In another set-up,

a-S824, b4, and b17 were expressed in L-, M-, and H-labeled

cells, respectively ( Figure 3 A). Preferential interactions with b4,

68 Cell 144, 67–78, January 7, 2011 ª2011 Elsevier Inc.



b17, orb23 were explored in a third type of experiment. Total cell

lysates were prepared essentially without removal of aggregate

material and combined 1:1:1 ( Figure 3 A and Figure S3 A). The

expressed proteins were quantitatively isolated using anti-Myc

antibody coupled to magnetic beads, followed by SDS-PAGE,

in-gel digestion, and LC-MS/MS analysis.

It seemed plausible that initial coaggregation may be driven by

relatively weak interactions, which might introduce a stochastic

element in the proteomic analysis. To overcome this problem

we based our analysis on extensive biological repetitions of the

experiments. Three proteomic experiments were performed,

each consisting of three biological repeats (independent trans-

fections). A protein was identified as b protein interactor

when its isotope-labeled peptides were either enriched relative

to the a-S824 control or relative to one of the other b proteins

with > 95% confidence in at least two of the three repeats of

a set (see Extended Experimental Procedures and Figures

S3B–S3D). A total of 94 interactors of b23, 73, of b17 and 57 of

b4 were identified in experiments of equivalent sampling size,

consistent with the relative toxicity of the proteins ( Figure 3B

Eβ4

100 nm 100 nm

β17

100 nm

β23

A

MCEQKLISEEDLGMQIS M DYQLEIEGNDNK V ELQL NDSGGE V KLQIR GPGGR V HF N V HSSGS NLE V NF N NDGGE V QFH M H

MCEQKLISEEDLGMQIS M DYEIKFHGDGD NFDL NLDDSGGDLQLQIR GPGGR V H V HIHSSSGK V DFH V N NDGGD V E V K M H

MCEQKLISEEDLGMQIS M DY NIQFH NNGNEIQFEIDDSGGDIEIEIR GPGGR V HIQL NDGHGHIK V DF N NDGGELQID M H

Myc epitope

β4:β17:β23:

1 80

MCEQKLISEEDLG M YGKL NDLLEDLQE VLKH V NQH WQGGQK N M NK V DHHLQN VIEDIHDFM QGG

GSGGKLQE MM KEFQQ VLDEIKQQLQGGDNSLHN V HENIKEIFHHLEELV HR

α-S824:1 64

65 105

B ThT

Wavelength (nm)460 480 500 520 540

F l u o r e s c e n c e ( A U )

0

5

10

15

20

β4 β17 β23α-S824

C

Wavelength (nm)


0

20

40

60

80

100

120

500 550 600 650 700

ANSD

Wavelength (nm)


0

2

4

6

8

10

400 450 500 550 600

NIAD-4

F

β4

100 nm 100 nm

β17

100 nm

β23

Figure 1. Amyloidogenic Aggregation of Model Proteins In Vitro

(A) Sequences of the model proteins, b4, b17, and b23, designed to form b-sheet fibrils, and a-S824 designed to form a 4-helical bundle. Polar and nonpolar

amino acids are indicated in grayand yellow, respectively,b strands anda helices by bluearrows and rods respectively. N-terminal c-Myc tagsare shown in red.

(B–D) Tinctorial properties of b protein aggregates. The purified proteins indicated (3mM) were analyzed in 25 mM HEPES buffer (pH 7.5), 150 mM KCl, 0.5 mM

MgCl2-containing 20 mM Thioflavin T (B), 1 mM NIAD-4 (C), or 20 mM ANS (D). Fluorescence was recorded as described in Experimental Procedures.

(E and F) Transmission electron microscopy of aggregates formed by b4, b17 and b23, as above, at pH 7.5 (E) or 10 mM potassium phosphate (pH 6.0) (F).

Proteins were negatively stained and observed at a magnification of 55,0003.

See also Figure S1.




and Tables S1–S3 ). Only four proteins were marginally enriched

on a-S824 relative to the vector only control, including two ribo-

somal proteins ( Figure S3B). Approximately 60% of the b4 and

b17 interactors were also found to interact with b23, indicating

a high degree of overlap in interaction profiles ( Figure 3B).

Western blotting of pulldowns and immunofluorescence analysis

of cells confirmed the results from SILAC/MS for several interac-

tors ( Figures S3E and S3F). Thus, interactions of the b protein

aggregates with multiple endogenous proteins precede the

strong decrease in cell viability observed at 72 hr after transfec-

tion ( Figure 2 A).

As summarized for b23, most of the proteins associated with

the aggregates have their primary location in the cytoplasm,

nucleus and mitochondria ( Figure 3C and Table S1 ). Proteins

involved in chromatin regulation, RNA processing, transcrip-

tion, translation, cytoskeletal function, vesicle transport, and

protein quality control were highly represented. These proteinsare generally of average cellular abundance ( Su et al., 2002 )

and for several of them between 10% and 45% of total was

associated with the aggregates, based on depletion from

supernatant fractions after pulldown as measured by SILAC/

MS ( Table S1 and Extended Experimental Procedures ). Note

that this analysis probably underestimates the extent of

sequestration, since coaggregates may partially dissociate

during isolation. Interestingly, 12 different translation initiation

factors interacted directly or indirectly with the aggregates,

including 9 of the 13 subunits of the eIF3 complex and 3

subunits of eIF4 ( Figure 3C). b17 aggregates contained 10

and b4 aggregates 9 of these proteins ( Tables S2 and S3 ).

Immunofluorescence analysis demonstrated extensive co-

localization of eIF3D and eIF4GII with the aggregates and

western blotting of pulldowns confirmed that at least $10%

of cellular eIF3D coaggregated with b23 and b17, compared

to $6% with b4 ( Figures S3E and S3F, and data not shown).

Indeed, labeling experiments showed that cells expressing

b23 for 24 hr had a $35% reduced protein synthesis capacity

( Figure S3G). Similarly, the altered morphology of b-protein-

expressing cells revealed by actin staining ( Figure S3H) may be

attributed to the association of filamins A, B, C, (FLNA, FLNB,

FLNC), and the giant protein plectin-1 (PLEC1) ( $500 kDa)

with the aggregates ( Figure 3C), proteins that are critical for

the formation and maintenance of cytoskeletal architecture.

These results show that many different proteins, involved in

a range of essential cellular functions, are affected by the

b protein aggregates.

Aberrant Stress Response inb-Protein-Expressing Cells

The proteomic analysis identified several cytosolic chaperones

and chaperone regulators to be associated with the aggregates,

including Hsc70 (Hsc71) and its cochaperones Hsp110

(Hsp105), Hdj1/2 and Bag2, as well as the nascent chain associ-

ated complex, NAC ( Figure 3C). Hsp110 was enriched in the

aggregates in a manner correlating with the relative toxicity of

b4, b17 and b23, as confirmed by western blotting and immuno-

fluorescence ( Figures 4 A and 4B). Indeed, overexpression of

Hsp110 ( Figure S4 A) partially suppressed b4 and b17 aggrega-

tion and toxicity but was inefficient in mitigating the toxic effects

of b23 ( Figures 4C and Figures S4B and S4C).

A

V i a b i l i t y ( % o f

c o n t r o l )

B

T T S P T S P T S P T S PkDa17

10

C β17β4 α-S824β23

A 1 1 r e a

c t i v i t y ( A U )

E

β4α-S824 β23

C

Myc

DIC

NIAD-4

D

α-S824 β4 β17 β23

72 h24 h

0

20

40

60

80

100

C β17β4 α-S824β23

0

1

2

3

4

5

β17

Figure 2. Cytotoxicity and Aggregation of

Model Proteins in HEK293T Cells

(A) Viability of HEK293T cells expressing b4, b17,

b23, and a-S824 measured by MTT assay 24 hr

and 72 hr after transfection. MTT reduction by

control cells transfectedwith empty vector,C, wasset to 100%. Standard deviations were derived

from at least three independent experiments.

(B) Solubility of b4, b17, b23,and a-S824analyzed

by fractionation of lysates from cells 24 hr after

transfection by centrifugation and immunoblotting

with anti-Myc antibody. T, total lysate; S, soluble

fraction;P, pellet fraction; C, emptyvectorcontrol.

Representative results from at least three inde-

pendent experiments.

(C and D) Protein distribution and aggregation in

intact cells expressing a-S824, b4, b17, and b23.

After 24 hr, proteins were detected by immuno-

fluorescence with anti-Myc antibodies (C). DIC,

differential interference contrast images. Amyloid-

like aggregates were detected by staining with

NIAD-4 (D) (see Experimental Procedures ). Nuclei

were counterstained with DAPI. The scale bar

represents 20 mm. Representative images of three

independent experiments.

(E) Quantification of A11 antibody reactivity in

extracts from b-protein-expressing cells 24 hr

after transfection. The cumulative dot blotsignal of

fractions from size exclusion chromatography was corrected for the cumulative anti-Myc signal, indicating the amounts of a-S824, b4, b17, and b23, and

expressed relative to theA11 reactivity ina-S824expressing cells (setto 1) (seeFigure S2C for originaldata). Averagesand standarddeviationsrepresent at least

three independent experiments.

See also Figure S2.




Remarkably, expression of the b proteins did not induce the

cytosolic stress response or heat-shock response (HSR), as no

increase in the levels of Hsp110, Hsp70, or Hsp27 was observed

( Figures 4 A and data not shown). A possible defect in the HSRwas further analyzed using a luciferase reporter gene under

control of the HSF1-dependent Hsp70 promoter ( Williams

et al., 1989 ). While inhibition of proteasome function by MG132

in control cells resulted in a 5-fold induction of the reporter,

this induction was completely abolished in cells expressing the

b proteins for 24 hr ( Figure 4D). The phorbol-12-myristate-13-

acetate (PMA)-mediated induction of a luciferase reporter under

the NF-kB promoter was also impaired, but to a lesser extent

than the inhibition of the stress response ( Figure 4E). Thus,

expression of the model b sheet proteins leads to a deficiency

of the normal cytosolic stress response, thereby limiting the

capacity of cells to mount an effective defense.

Structural Features of b Protein Interactors

A bioinformatic analysis of the physicochemical properties of the

b protein interactors was conducted to see whether these

proteins share certain structural features. We focused our initialanalysis on the interactors of b23 ( Table S1 ). Compared to a set

of 3055 control proteins identified by LC-MS/MS in a total cell

lysate( Table S4 ), theb23 interactors are shifted to higher molec-

ular weight, with a significantly greater fraction of proteins above

150 kDa (p < 0.005) ( Figure 5 A). In addition, the interactors have

a lower average hydrophobicity and a bimodal hydrophobicity

distribution ( Kyte and Doolittle, 1982 ) (p < 0.005) ( Figure 5B).

The occurrence of domain folds among the b23 interactors, as

classified in SCOP, was generally similar to that of lysate

proteins. However, the b23 interactors contained significantly

more proteins with all beta domains (SCOP class b) (p < 0.05)

( Figure S5 A), including the beta-barrel VDAC proteins of the

A L: Arg0,Lys0 M: Arg6,Lys4 H: Arg10,Lys8

β23mix

lysates1:1:1

LC-MS/MSIPExpt. II:

Vector only α-S824Expt. I:

Cytoskeleton

AKAP12

CCDC88A

MARCKS

DIAPH1

FLNA

FLNB

FLNC

KIF5B

PLEC1

TUBA1A

TUBB2C

VIM

ZYX

SEPT2

SEPT7

SEPT9

Chromatin structure

RUVBL1

SMARCA4

SMARCA5

SMARCC1

SMC4H1F0

CHD4

RNA processing

BAT1

GEMIN4

RBM8A

SMN1

GEMIN5 S R140

Unspecified function

HCFC1

BAT2D1

Transcription

CAND1

CNOT1

GTF2I

PNN

PURA

SND1

Nuclear transport

KPNA2

RANBP1

THOC2

DNA maintenance PRKDC

TNKS1BP1

PCNP

Mitochondria

SLC25A6

IMMT

CHCHD3

SSBP1

VDAC1

VDAC2

VDAC3

Molecular chaperones

NACB

BAG2

CHORDC1

DNAJA1

HSPA8

HSPH1

DNAJB1

NACA

Metabolism

HDLBP

ALDH18A1

CAD

CHERP

Ribosome biogenesis

NVL

PDCD11

DIMT1L

WDR3

Vesicle transport

AP1M1

CLTC

VAPA

A P1G1 S EC16A

AP1B1 tRNA synthetases

WARS

YARS

EPRS

Translation

EIF3A

EEF1A1

EIF3C

EIF3D

EIF3E

EIF3F

EIF3G

EIF3M

EIF3I

EIF3L

EIF4A1

EIF4A3

EIF4G3

IGF2BP3

LMNB1

NUMA1

AHNAK

Nuclear structure

C

RBBP4

RUVBL2

B

311

152347

21

β4(57)

β17

(73)β23(94)

Expt. III:

α-S824

m/z

I n t e n s i t y

Expt. I

C

β23

m/z

Expt. II

β17

β4α-S824

m/z

Expt. III

β17

β23

β4α-S824 β4 β17

β4 β17 β23

STRAP

4

ARHGDIA

Protein degradation

CACYBP

ERLIN2

UBC

UBR4

SUMO2

Miscellaneous

ABCE 1 OGT

SBDSFAM120A

Figure 3. Interactome Analysis of b proteins

(A) Design of SILAC experiments to identify b protein interactors by LC-MS/MS. L, M, and H, light, medium, and heavy isotope media. C, vector only control.

(B) Overlap between the interactor sets of b4, b17, and b23. Total numbers of identified interactors are given in parentheses.

(C) The b23 interacting proteins are grouped according to cellular location and function.

See also Figure S3 and Tables S1–S3.




outer mitochondrial membrane and the filamins which have

Ig-domain repeats ( Table S1 and Figure S3F).

The lower hydrophobicity of many interactors suggested that

these proteins may be rich in intrinsically unstructured regions

(IURs). Indeed, compared to lysate proteins, the b23 interactors

have a significantly greater fraction of total amino acids in IURs

(p < 0.05), based on the DisoDB database ( Pentony and Jones,

2009 ) and the DisEMBL and IUPred prediction tools for unstruc-

tured regions ( Dosztanyi et al., 2005; Linding et al., 2003 ) ( Fig-

ure 5C, Tables S1 and S4, and data not shown). Pronounceddifferences emerged when considering the fraction of proteins

with IURslongerthan30 or 50 residues( Figure 5D).For example,

$60% ofthe b23 interactors arepredicted to contain at least one

unstructured segment of 30 amino acids, compared to $45% in

lysate proteins or the complete proteome (p < 0.005) ( Figure 5D).

The corresponding numbers for IURs > 50 amino acids are

$40% and $25% (p < 0.005), respectively. Moreover, the b23

interactors contain on average $3 disordered segments of 30

amino acids (compared to $1.8 for the lysate proteins, p <

0.005). The predicted IURs of the interactors are shifted to

greater length (p < 0.005) ( Figure S5B and Tables S1 and S4 ),

with 21% of the proteins containing IURs > 80 amino acids

and 10% of 100 to 429 residues. The IURs are enriched in polar

amino acids and in amino acids that have a high propensity to

form coil and turn regions, such as M, K, R, E, S, Q, and P,

and are depleted in aromatic and hydrophobic amino acids W,

Y, F, C, I, and V ( Dunker et al., 2008 ). Such sequences are struc-

turally flexible and populate a range of conformational states

from extended disordered to collapsed, molten globule-like

structures ( Dunker et al., 2008; Pentony and Jones, 2009 ). Using

the Zagg algorithm to predict aggregation propensities, the b23

interactors have higher aggregation scores than lysate proteins(see Figure S6B below).

A comparison of the proteins that interact preferentially with

b4, b17, and b23 (18, 28, and 27 proteins, respectively), as

defined by the SILAC experiments ( Figure S3D and Extended

Experimental Procedures ), revealed that a gradual increase in

molecular weight and decrease in hydrophobicity of the interac-

tors, along with a slight increase in their disorder, correlated with

the differential cytotoxicity of the three b proteins ( Figure 5E).

This trend was also observed when comparing the complete

interactor sets of the three b proteins ( Tables S1–S3 ).

The prominent association of proteins with low hydropho-

bicity and high intrinsic disorder with the aggregates was

NF-κ B-Luc

C α-S824β4 β17 β23

A

IP

Lysate H s p 1 1 0

C α-S824β4 β17 β23

Immunoblot

Hsp110 overexpression

C

C α-S824β4 β17 β23 V i a b i l i t y ( % o f c o n t r o l )

0

20

40

60

80

100

120

+− − +− + −+ −+

C

α-S824

β4

β17

β23

B Myc Hsp110 Merge

HSP70-LucD

(

f o l d i n d u c t i o n )

P r o m o t e r a c t i v i t y

1

10

E

( f o l d i n d u c t i o n )

P r o m o t e r a c t i v i t y

C α-S824β4 β17 β23

1

10

100

Figur e 4. Impai rment of the Str ess

Response in b-Protein-Expressing Cells

(A and B) Association of Hsp110 with b protein

aggregates. HEK293T cells were transfected as

indicated (C, empty vector control). 24 hr after

transfection, Myc-tagged proteins b4, b17, b23,and a-S824 were immunoprecipitated from cell

lysates and analyzed by immunoblotting with anti-

HSP110 antibody (A). Lysate samples correspond

to 8% of input used for IP. b4, b17, and b23 in

pulldowns was associated with $5%, 9%, and

16% of total cellular Hsp110, respectively (also

see Figure S3F). For immunofluorescenceanalysis

(B), cells were fixed and costained with anti-Myc

antibodies and anti-Hsp110 antibodies. Nuclei

were stained with DAPI. Representative examples

of three independent experiments are shown.

(C) Partial rescue of b protein toxicity by Hsp110

overexpression. Cells were transfected with

empty vector or the expression vector for human

Hsp110. 24 hr later, cellswere electroporated with

empty vector,C, or expression vectors forb4, b17,

b23, and a-S824. Three days after the second

transfection, MTT assays were performed. Empty

vector control was set to 100% viability. Standard

deviations of three independent experiments are

shown.

(D and E) Inhibitory effect of b proteins on

cellular stress response pathways. Cells were

cotransfected with HSP70-luciferase reporter (D)

or NF-kB-luciferase reporter constructs (E) and

the b-protein-expressing plasmids. 6 hr later, 5mM

MG132 (D) or16 mM PMA (E)were addedto induce

the respective promoter. Luciferase activity was

measured 24 hr after transfection. The promoter

activity in cells transfected with control vector, C,

without inducer was set to 1. Standard deviations

of three independent experiments are shown.

See also Figure S4.




unexpected. To test whether such proteins are targeted more

generally by amyloid-like aggregation, we performed an initial

analysis of interactors of wild-type A b1-42 and its Arctic mutant

(E22G), which causes early-onset Alzheimer’s disease. The

latter was included because it is known to populate higher

levels of prefibrillar aggregates and toxic oligomers exposing

hydrophobic surfaces ( Bolognesi et al., 2010; Nilsberth et al.,

2001 ). To allow a comparison with the model b aggregates,

the A b proteins were also expressed in the cytosol, using

GFP fusions ( Kim et al., 2006 ). In contrast to the artificial

b proteins, the A b constructs were degraded but accumulated

upon partial proteasome inhibition with MG132 ( Figure S5C)

HydrophobicityErlin

C

A

Molecular weight (kDa)

% o

f i d e n t i f i e d p r

o t e i n s

0 4 0

8 0

1 2 0

1 6 0

2 0 0

2 4 0

2 8 0

3 2 0

4 8 0

5 2 0

5 6 0

6 0 0

6 4 0

B

% o

f i d e n t i f i e d p r o t e i n s

- 1 . 6

- 1 . 4

- 1 . 2

- 1 . 0

- 0 . 8

- 0 . 6

- 0 . 4

- 0 . 2

0 . 0

0 . 2

0 . 4

0 . 6

0 . 8

1 . 0

p<0.05

D i s o r d e r e d r e s i d u e s ( % )

0

4

8

12

16

20

24

>30a.a. >500

10

20

30

40

50

60

P r o t e i n s w i t h d i s o r d e r e d

s t r e t c h e s ( % )

D

Lysate proteins

β23 interactors

E

M o l e c u l a r w e i g h t ( k D a )

Preferred interactors

β17 interactors

Lysate proteins

β23 interactors

β4 interactors

0

10

20

30

Lysateproteins interactors

β230

50

100

150200

250

300

350

M o l e c u l a r w e

i g h t ( k D a )

p<0.005

0

2

4

6

8

10

12

14

Lysateproteins interactors

β23

H y d r o p h o b i c i t y

-1.2

-1.0

-0.8

-0.6

-0.4

-0.2

0.0

0.2p<0.005

**

**

0

10

20

30

40

50

60

70

-0.55-0.50

-0.45-0.40

-0.3520 40 60

80 100 H y d

r o p h

o b i c i t y

D i s o r d e r > 3 0 a .a . [ p r o t e i n s ( % ) ]

Figure 5. Structural Properties of the

b Protein Interactors

(A and B) Distribution of molecular weight (A) and

average hydrophobicity (B) of lysate proteins

(gray) andb23 interactors (red) ( TablesS1 andS4 ).

Box plots (insets) indicate the distribution of thedata. Dashed horizontal line indicates the median,

whisker caps and circles indicate 10th/90th and

5th/95th percentiles, respectively. P values are

based on Mann-Whitney test.

(C andD) Disorderanalysis of bprotein interactors.

Percentage of disordered residues in interactor

sequences (C). p value based on Mann-Whitney

test. Fractionof proteinswith disordered stretches

longer than 30 or 50 amino acids (D). **p < 0.005

based on a chi-square test. Disorder was deter-

mined using DisoDB.

(E) Structural properties of lysate proteins and of

proteins interacting preferentially with b4, b17, or

b23.

See also Figure S5 and Tables S1–S6.

( Lee et al., 2006 ). The Arctic mutant

formed visible aggregates more readily

and showed substantially greater toxicity

than WT A b1À42 ( Figures S5D and S5E).

Analysis by SILAC/MS revealed that

the A b interactome is comparable in

complexity to that of the b proteins,

with a direct overlap of $25%, promi-

nently including translation initiation

factors, chromatin regulators, RNA pro-

cessing proteins, mitochondrial mem-

brane proteins and chaperones ( Table

S5 ). We also identified 31 proteins which

were enriched on the Arctic mutant rela-

tive to the less toxic A b1À42 WT ( Table

S6 ). Notably, these proteins resemble

the b protein interactors in physico-

chemical properties and are significantly

enriched in IURs (p < 0.05) ( Figures

S5F–S5I).

From these results, we conclude that

cells contain a subpopulation of meta-

stable proteins that are prone to interact

with and potentially become sequesteredby toxic species populated in the process of amyloid-like

aggregation.

b Protein Interactors Include Pre-Existent and Newly

Synthesized Proteins

While theresults above suggestedthat structural flexibility is crit-

ical in facilitating the interaction of endogenous proteins with the

b aggregates, we noted that for $40% of the b23 interactors no

IURs > 30 amino acids are predicted ( Figure 5D and Table S1 ).

We therefore considered the possibility that some of these

proteins may succumb to coaggregation upon synthesis

before adopting stably folded structures. To test this idea, we




pulse-labeled HEK293T cells expressing b23 or a-S824 with35S-methionine, followed by immunoisolation of the proteins.

Around 7% of the proteins labeled within 15 min were coisolated

with b23, compared to only $1% with a-S824 ( Figure S6 A), sug-

gesting that a substantial fraction of newly synthesized polypep-

tides can interact with b23.

To identify such proteins, we performed pulse-SILAC experi-

ments. Cells were cultured with medium amino acid isotopes

(M) to label preexistent proteins, followed by transfection with

b23. The culture was divided and one half was immediately

shifted to media containing heavy amino acid isotopes (H).Control cells were cultured with light amino acids (L) and trans-

fected with a-S824. After 24 hr, the cells from the three condi-

tions were combined and subjected to anti-Myc pulldown and

LC-MS/MS analysis ( Figure 6 A). The H/M isotope ratio of the

b23 interactors in the pulldown relative to their H/M ratios in

the lysate was used to indicate whether they interact with b23

preferentially as newly synthesized (New) or pre-existent (Old)

proteins ( Figure 6 A). H/M labeling ratios were obtained for 50

b23 interactors, and a number of these showed a clear prefer-

ence for interaction soon after synthesis ( Figure 6B and Table

S7 ). In contrast, fewer proteins interacted preferentially as old

proteins. These interactors include Hsp110 as well as several

BA

Transfection :

m/z

I n t e n s i t y

β23 interactor: Log(H/Mβ23)/(H/Mlysate)

β23

α-S824 β23

Preexistent(Old)

α-S824 β23

m/z

β23

β23

α-S824

Newly-synthesized(New)

I n t e n s i t y

β23 interactors

New

Old

D i s o r d e r e d r e s i d u e s ( % )

0

5

10

15

20

25

30

35

C

>30a.a. >50

**

**

P r o t e i n s w i t h d i s o r d e r e d

s t r e t c h e s ( % )

D

Old β23 interactors

New β23 interactors

E

M o l e c u l a r w e i g h t ( k D a )

0

10

20

30

40

50

60

70

80

Lysate proteins

Old β23 interactors

New β23 interactors

L: Arg0,Lys0 M: Arg6,Lys4

H: Arg10,Lys8L: Arg0,Lys0 M: Arg6,Lys4

-0.6-0.3 0.0 0.3 0.6 0.9 1.2 1.5

5

10

15

20

25

30

35

40

45

50

p<0.05

0

10

20

30

40

50

60

70

-0.7-0.6

-0.5-0.4

-0.320 40 60

80 100D i s o r d e r > 3 0 a .a . [ p r o t e i n s ( % ) ]

H y d r

o p h o

b i c i t y

Figure 6. Newly Synthesized and Pre-Exis-

tent b Protein Interactors

(A) Design of SILAC based mass spectrometric

analysis to identify proteins preferentially inter-

acting with b23 as newly synthesized (new) or

preexistent (old) proteins (Pulse-SILAC).(B) Ratios of heavy to medium isotopes (H/M) of

b23 interactors relative to the H/M ratios for the

same proteins in thetotal cell lysate (see Extended

Experimental Procedures ). The log of this ratio of

ratios increases with the tendency of a protein to

interact with b23 as a new protein.

(C and D) Disorder analysis of new and old b23

interactors. Percentage of disordered residues

in interactor sequences (C). p value is based

on Mann-Whitney test. Fraction of proteins with

continuous disordered stretches > 30 or > 50

amino acids (aa) (D). **p < 0.005 for old interactors

relative to lysate (see Figure 5D), based on Chi-

square test.

(E) Molecular weight, disorder and hydrophobicity

of old and new b23 interactors relative to lysate

proteins.

See also Figure S6 and Tables S4 and S7.

translation initiation factors ( Table S7 ),

consistent with impairment of translation

efficiency being an early consequence

of b protein toxicity ( Figures S3E–S3G).

Interestingly, the old and new interac-

tors from the ends of the distribution (15

proteins each) differ markedly in their

structural properties. The old interactors

contain a significantly greater fraction of

amino acids in IURs than the new interac-

tors ( Figure 6C). They are strongly en-

riched in continuous disordered regions ( Figure 6D) and are of

low average hydrophobicity ( Figure 6E). In contrast, the new

proteins are similar to lysate proteins in terms of hydrophobicity,

but are lower in disorder and substantially larger in size ( Fig-

ure 6E). Their folding pathways may be complex and kinetically

slow, possibly resulting in the prolonged exposure of hydro-

phobic residues during folding. Based on analysis using the

Zagg algorithm, the new b23 interactors show high intrinsic

aggregation scores only in their unfolded states ( Figure S6B).

In contrast, the old interactors are highly flexible and are unable

to bury aggregation-prone regions in their native structures.Thus, these proteins have high aggregation scores both in their

unfolded and folded states ( Tartaglia et al., 2008 ) ( Figure S6B).

Some of them may coaggregate as preexistent or newly synthe-

sized proteins, consistent with their lower peak values in the

isotope labeling ratioscompared to thenew proteins ( Figure 6B).

In summary, the b23 interactors can be divided into two over-

lapping subsets of relatively aggregation-prone proteins: One

group is enriched in IURs, which would be prone to aggregate

even in their post-folding state. This group is highly represented

among the old proteins. The second group contains an abun-

danceof large and/or multidomain proteins, which require longer

times for synthesis and may fold slowly. Consequently, in




conditions of limited chaperonecapacity, they would be prone to

aggregate during and shortly after synthesis. This group is

enriched among the newly synthesized proteins. Finally, some

proteins occupy a transition zone, combining physicochemical

features of both groups.

Aggregate Interactors Have Critical Network Functions

The structural flexibility and relatively large size of the aggregate

interactors suggests that these proteins may normally be

involved in numerous functional protein interactions. To address

this possibility, we analyzed how the b23 interactors are linked

with the cellular protein network. A query of the Human

Proteome Reference Data Base (HPRD) ( Keshava Prasad

et al., 2009 ) revealed that each of these proteins functionally

interacts with $12 different proteins on average, compared to

$7 per lysate protein and $7.5 per protein in HPRD (19,651

entries) ( Figure 7 A). Notably, most of the b23 interactors have

no or only few interactions with any of the other b23 interactors,

suggesting that coaggregation may disrupt their functionalcomplexes. For example, the microfilament protein vimentin

interacts with more than 100 different proteins according to

HPRD, but only three of those areamong the identifiedb23 inter-

actors, although 49 potential vimentin interactors were detected

in the lysate or background of the pulldowns (data not shown).

Essential proteins often occupy critical ‘‘hub’’ positions in the

network ( Hayneset al., 2006; Jeong etal.,2001 ).Eachb23 target

protein interacts on average with $5 different essential proteins,

compared to only $3 per lysate protein and $1.5 per entry in

HPRD ( Figure 7 A). Moreover, the b23 interactors are more

frequently linked than lysate proteins, through direct interac-

tions, with proteins that have been found in association with

neurodegenerative disease proteins ( Raychaudhuri et al., 2009 )

( Figure 7 A and Table S1 ).

Assuming that a disturbance of functional protein interactions

contributes critically to b aggregation toxicity, b23 would be

expected to differ in this regard from the less cytotoxic proteins

b4 and b17. We found that the b4, b17 and b23 interactors are

physically linked to a total of 600, 643, and912 different proteins,

including 216, 213, and340 essential proteins and53, 56, and84

proteins associated with neurodegenerative disease networks,

respectively ( Figure 7B). Thus, the capacity of the b protein

aggregates to interact with and sequester highly connected

cellular proteins correlates well with their relative cytotoxicity.

DISCUSSION

Widespread Coaggregation of Metastable Proteins

A key finding of this study is that amyloidogenic aggregation can

result in the sequestration of numerous proteins that share

distinct physicochemical properties: They are relatively large insize and exhibit high structural flexibility, with a significant

enrichment in disordered regions, features that are strongly

linked with multifunctionality ( Figure 7C).

The artificial b sheet proteins used as a model were designed

to assemble into fibrils ( West et al., 1999 ). Like natural amyloido-

genic proteins, they populate a range of prefibrillar aggregation

intermediates, which are likely to represent the primary toxic

agents in aggregation diseases ( Chiti and Dobson, 2006; Jahn

and Radford, 2008 ). Based on recent findings, the proteotoxicity

of such species correlates with the exposure of ANS-binding

hydrophobic surfaces ( Bolognesi et al., 2010; Campioni

et al., 2010 ) and reactivity with the A11 anti-oligomer antibody

C

Chaperones

Structuredproteins

protein complexesProteins or

containing

disordered regions

Amyloidogenicregions (red)

Proteins not yet foldedand/or assembled

Newly-synthesized

Chaperones

All interactions

Disease

Essential

B

co-aggregationco-aggregation

β-aggregates/ oligomers

Toxic

A

A v e r a g e i n t

e r a c t i o n s

/ P r o t e i n

HPRD Lysate β2301234567

89

10111213

0

20

40

60

80

550

700850

1000

200250

300350 I n

t e r a c t

o r s

E s s e n t i a l i n t e r a c t o r s

D i s e a s e p r o t e

i n s

β17β23

β47.5

1.6

0.4

7.2

2.8

0.8

12.4

4.9

1.5

Figure 7. Mechanism of b Aggregation

Toxicity

(A and B) Functional context of b protein inter-

actors within the protein interaction network.

Shown in (A) are the average number of functional

interactions of the b23 interactors in comparisonto proteins in the HPRD database and in the

experimentally determined cell lysate ( $3000

proteins). These functional interactions are cate-

gorized into total interactions, interactions with

essential proteins and interactions with proteins

involved in neurodegenerative disease ( Ray-

chaudhuri et al., 2009 ). In (B), the complete sets of

b4, b17, and b23 interactors are compared in

terms of these functional properties.

(C) Model for the interaction of b aggregates with

pre-existent and newly synthesized proteins. Pre-

existent proteins are structurally flexible in their

functional state and are involved in multiple

protein-protein interactions, which may be dis-

rupted by their association with the b aggregates.

The newly synthesized proteins are structurally

vulnerable to coaggregation during folding and

assembly. Interaction of both the preexistent

and newly synthesized proteins with the b aggre-

gates is facilitated by the limiting capacity of

chaperones to shield aggregate surfaces and by

the failure of the cells to mount an efficient stress

response.




( Kayed et al., 2003 ), properties that arereproduced by themodel

proteins. Flexible hydrophobic surfaces and unpaired backbone

structure that is not yet integrated into a stable cross-b core

( Mossuto et al., 2010 ) may endow oligomers and protofilaments

with the capacity to engage in widespread aberrant interactionswith metastable proteins. Whether oligomeric species with

similar interaction properties also occur during nonamyloido-

genic aggregation, leading to amorphous structures rather

than fibrils, remains to be determined.

The b protein aggregates were found to interact with preexis-

tent and newly synthesized polypeptides ( Figure 7C). The former

are strongly enriched in intrinsically unstructured regions (IURs)

and are of lower average hydrophobicity. A similar trend was

observed for interactors of the toxic aggregates of the Arctic

mutant of A b1-42, which is known to transiently populate high

concentrations of prefibrillar aggregates ( Bolognesi et al.,

2010 ) ( Figures S5C–S5I). Proteins rich in structural disorder are

considered to be adaptable to multiple interaction partners

( Dunker et al., 2008; Pentony and Jones, 2009 ). On the otherhand, local structural fluctuations in these proteins are expected

to give rise to the exposure of sequence elements with a higher

propensity to form aggregates, consistent with the relatively high

Zagg scores of the b protein interactors ( Tartaglia et al., 2008 ).

Indeed, some of the best known neurodegenerative disease

proteins, such as a-synuclein or tau, are thought to be almost

entirely unstructured. In contrast, the proteins that interact with

the aggregates during or soon after synthesis have average

hydrophobicity and disorder. These proteins tend to be large in

size and are likely to populate nonnative states which expose

hydrophobic surfaces during their folding, assembly or transport

that must be shielded by molecular chaperones ( Figure 7C). For

example, among the b protein interactors are mitochondrial

membrane proteins such as VDAC andthe ADP/ATP translocase

which require chaperone protection during post-translational

sorting ( Young et al., 2003 ).

By targeting flexible regions and hydrophobic surfaces of

preexistent and newly synthesized proteins, the b protein aggre-

gates may act in a ‘chaperone-like’ manner but cannot promote

folding through regulated release. Consequently, more and more

proteins are recruited, which in turn may generate new interac-

tion surfaces, thereby magnifying the toxic potential of the

aggregates (‘snowball effect’).

Interference with Multiple Key Cellular Functions

The b protein interactors include many proteins with key cellular

functions in transcription and translation, chromatin regulation,vesicular transport, cell motility and architecture, as well as

protein quality control ( Figure 3C). Similar proteins were also

found to interact with aggregates of A b ( Table S5 ), suggesting

that these pathwaysmay be more generally at risk in aggregation

disorders. Bioinformatic analysis showed that most of the coag-

gregating proteins have numerous functional interactors, consis-

tent with their preferential role as network hubs ( Haynes et al.,

2006; Jeong et al., 2001 ). The number of functional interactions

of the sequestered proteins correlates with the relative cytotox-

icity of the b protein aggregates ( Figure 7B). It is thus likely that

the aggregates compete for binding to disordered regions with

a protein’s normal interactors and the more toxic forms may be

able to compete more effectively. Based on our proteomic

and biochemical measurements, $10%–45% of total may be

sequestered for several of the interacting proteins ( Figures

S3E–S3F and Table S1 ). Moreover, certain proteins may misfold

upon interaction with the aggregates but remain in solution.Thus, dependent on the interaction strength of the aggregates,

an increasing number of key functions may be affected, eventu-

ally resulting in fatal network collapse. We estimate that the

human proteome contains $2000 proteins that structurally

resemble the experimentally identified b aggregate interactors.

Dependent on cell type and the exact structural properties of

the causative aggregate, different subsets of these proteins

may be affected, which may help to explain different patterns

of pathobiology. It will be interesting to see which of these

proteins are preferentially targeted by b aggregates in neuronal

cells.

Our results also lend support to the recent view that protein

misfolding and aggregation disturbs proteostasis by compro-

mising the cellular folding environment ( Morimoto, 2008 ). Wesuggest that the association of endogenous proteins with the

aggregates is facilitated by the failure of the affected cells to

mount an efficient stress response, a phenomenon that was

previously observed during prion infection ( Tatzelt et al., 1995 )

and may be particularly serious in postmitotic cells, such as

neurons. Inhibition of the stress response may be due to the

sequestration by the aggregates of multiple chromatin regula-

tors, which interact with numerous transcription factors,

including HSF1 ( Erkina et al., 2010; Sullivan et al., 2001 ). As

a consequence of limiting proteostasis capacity, newly synthe-

sizedpolypeptides with a high chaperone requirement for folding

may become increasingly vulnerableto sequestration by disease

protein aggregates ( Figure 7C). This fatal chain of events may be

further enhancedduring aging, which is associated with a decline

of proteostasis and thus would result in a reduced capacity of

cells to protect their more vulnerable proteins against coaggre-

gation ( Balch et al., 2008; Morimoto, 2008 ).

EXPERIMENTAL PROCEDURES

Protein Purification and In Vitro Analysis of Aggregates

Proteins a-S824, b4, b17 and b23 were expressed in E. coli BL21 cells and

purified as described in Extended Experimental Procedures. Fluorescence

analysis, circular dichroism, FTIR spectroscopy and negative stain electron

microscopy of the aggregates were performed using standard methods (see

Extended Experimental Procedures ).

Cell Culture, Immunoblotting and Reporter Assays

Human HEK293T cells werecultured understandard conditions(seeExtended

Experimental Procedures ). Transient transfectionswere performedby electro-

porationwith 30mg expression vector or by Lipofectamin (Invitrogen) transfec-

tion for overexpression of Hsp110. Immunoblots were developed using the

chemiluminescence kit Rodeo ECL (USB) and analyzed using a LAS-3000

image reader (Fujifilm) and the AIDA software (Raytest). For luciferase reporter

assays, cells were lysed in Lysis Buffer (Promega) and luciferase activity

measured using a Lumat LB9507 (EG&G Berthold).

Cell Viability

Cell viability was analyzed by measuring the capacity of cells to reduce 3-(4,5-

Dimethylthiazol-2-yl)-2,5-diphenyltetrazoliumbromide (MTT) to formazan

at different times after transfection with a-S824, b protein or A b42-GFP

constructs ( Shearman, 1999 ).




Solubility Analysis and Oligomer Quantification

Cells were lysed in Triton X-100/Na deoxycholate-containing PBS with

protease inhibitors. Benzonase was used to hydrolyze DNA. Raw debris was

removed at 20003 g for5 minand thesupernatant wasfractionated by centri-

fugation (100.000 3 g, 30 min) into pellet and soluble fractions or by gel filtra-

tion on a Superose 6 column (Amersham Bioscience), followed by dot blotanalysis with anti-oligomer antibody A11 ( Kayed et al., 2003 ) (see Extended

Experimental Procedures for details).

Immunofluorescence and Fluorescence Imaging

Transfected cells werefixed withparaformaldehyde, permeabilizedwith Triton

X-100 and stained with antibodies as indicated. Images were recorded with

a Leica TCS SP2 confocal laser scanning microscope. Protein aggregates

were analyzed by staining with NIAD-4 (ICX Nomadics) (see Extended Exper-

imental Procedures ).

SILAC and Sample Preparation for LC-MS/MS Analysis

Labeling of cells was performed in custom medium supplemented with light

(L), medium (M) or heavy(H) arginine andlysineisotopes(seeExtended Exper-

imental Procedures ). In pulse-SILAC experiments, M-labeled cells were

shiftedto H-medium,as indicated inFigure 6 A.Cells were lysedandcelldebris

removed by low-speed centrifugation(20003 g,5 min). Lysates fromL, M and

H cellswereadjusted to equal protein concentrationand mixed ata 1:1:1 ratio.

An aliquot of this mix was set aside as ‘‘lysate’’ control. Anti-Myc or anti-GFP

MicroBeads (Miltenyi Biotech) were used to isolate the Myc-tagged proteins

or GFP-fusion proteins and their interactors. The bound proteins were eluted

and processed as described ( Ong and Mann, 2006 ). The spectra were

interpreted using MaxQuant version 1.0.12.31 ( Cox and Mann, 2008 )

combined with Mascot version 2.2 (Matrix Science, www.matrixscience.

com ). See Extended Experimental Procedures for details. The raw MS data

along with a full list of identified proteins and quantitations is available at

https://proteomecommons.org/tranche, entering the following hash: +Ff0/

p8lSBrrzCKZfzAwYS3+Bqw5fonokB679f136te2iklhHtFMUpeT5SM/I3XuufTyr

Xj0ycVVC6G4Li/L02 dA4jcAAAAAAABVfg = =.

Bioinformatic Analysis

Averagehydrophobicity was calculated accordingto Kyteand Doolittle(1982),protein disorder using the DisoDB database ( Pentony and Jones, 2009 ) and

aggregation propensities according to Tartaglia et al. (2008). Protein fold

prediction and the analysis of functional protein interactions are described in

Extended Experimental Procedures. Student’s t test and Mann-Whitney test

wereused to comparegroups.Chi-square tests wereused to determinesignif-

icant differences between categorical data.

SUPPLEMENTAL INFORMATION

Supplemental information includes Extended Experimental Procedures,

six figures, and seven tables and can be found with this article online at

doi:10.1016/j.cell.2010.11.050.

ACKNOWLEDGMENTS

We thank C.G. Glabe for providing the A11 antibody and H. Wagner for plas-

mids HSP70-Luc and NF-kB-Luc. We acknowledge the help of H. Engelhardt

withFTIR measurements, technical assistance by O. Mihalache and V. Marcus

with electron microscopy, and R. Zenke (MPIB core facility) for confocal

microscopy. Financial support from EU Framework 7 Integrated Project

PROSPECTS, the Deutsche Forschungsgemeinschaft (SFB 596), the Ernst-

Jung Foundation, and the Ko ¨ rber Foundation is acknowledged. H.O. and

A.W. have a fellowship from the Fonds der Chemischen Industrie, and H.O.

is supported by the Elite Graduate Network of Bavaria.

Received: April 15, 2010

Revised: September 6, 2010

Accepted: November 11, 2010

Published: January 6, 2011

REFERENCES

Balch, W.E., Morimoto, R.I., Dillin, A., and Kelly, J.W. (2008). Adapting proteo-

stasis for disease intervention. Science 319, 916–919.

Bence, N.F., Sampat, R.M., and Kopito, R.R. (2001). Impairment of the ubiqui-

tin-proteasome system by protein aggregation. Science 292, 1552–1555.

Bolognesi, B., Kumita, J.R., Barros, T.P., Esbjorner, E.K., Luheshi, L.M.,

Crowther, D.C., Wilson, M.R., Dobson, C.M., Favrin, G., and Yerbury, J.J.

(2010). ANS Binding RevealsCommon Featuresof Cytotoxic Amyloid Species.

ACS Chem. Biol. 5, 735–740.

Campioni,S., Mannini,B., Zampagni, M., Pensalfini, A., Parrini, C., Evangelisti,

E., Relini, A., Stefani, M., Dobson, C.M., Cecchi, C., et al. (2010). A causative

link between the structure of aberrant protein oligomers and their toxicity. Nat.

Chem. Biol. 6, 140–147.

Chiti, F., and Dobson, C.M. (2006). Protein misfolding, functional amyloid, and

human disease. Annu. Rev. Biochem. 75, 333–366.

Cooper, A.A., Gitler,A.D., Cashikar, A.,Haynes,C.M., Hill, K.J., Bhullar,B., Liu,

K.N., Xu, K.X., Strathearn, K.E., Liu, F., et al. (2006). Alpha-synuclein blocks

ER-Golgi traffic and Rab1 rescues neuron lossin Parkinson’s models. Science

313, 324–328.

Cox, J., and Mann, M. (2008). MaxQuant enables high peptide identification

rates, individualized p.p.b.-range mass accuracies and proteome-wide

protein quantification. Nat. Biotechnol. 26, 1367–1372.

Dosztanyi, Z., Csizmok, V., Tompa, P., and Simon, I. (2005). IUPred: web

server for the prediction of intrinsically unstructured regions of proteins based

on estimated energy content. Bioinformatics 21, 3433–3434.

Dunker, A.K., Silman, I., Uversky, V.N., and Sussman, J.L. (2008). Function

and structure of inherently disordered proteins. Curr. Opin. Struct. Biol. 18,

756–764.

Erkina, T.Y., Zou, Y., Freeling, S., Vorobyev, V.I., and Erkine, A.M. (2010).

Functional interplay between chromatin remodeling complexes RSC, SWI/

SNF and ISWI in regulation of yeast heat shock genes. Nucleic Acids Res.

38, 1441–1449.

Frydman, J. (2001). Folding of newly translated proteins in vivo: The role of

molecular chaperones. Annu. Rev. Biochem. 70, 603–647.

Gidalevitz, T., Ben-Zvi, A., Ho, K.H., Brignull, H.R., and Morimoto, R.I. (2006).

Progressive disruption of cellular protein folding in models of polyglutamine

diseases. Science 311, 1471–1474.

Goldschmidt, L., Teng, P.K., Riek, R., and Eisenberg, D. (2010). Identifying the

amylome, proteins capable of forming amyloid-like fibrils. Proc. Natl. Acad.

Sci. USA 107 , 3487–3492.

Hartl, F.U., and Hayer-Hartl, M. (2002). Molecular chaperones in the cytosol:

from nascent chain to folded protein. Science 295, 1852–1858.

Haynes,C., Oldfield,C.J., Ji,F., Klitgord,N., Cusick, M.E., Radivojac, P., Uver-

sky, V.N., Vidal, M., and Iakoucheva, L.M. (2006). Intrinsic disorder is

a common feature of hub proteins from four eukaryotic interactomes. PLoS

Comput. Biol. 2, e100.

Jahn, T.R., and Radford, S.E. (2008). Folding versus aggregation: Polypeptide

conformations on competing pathways. Arch. Biochem. Biophys. 469,

100–117.

Jeong, H., Mason, S.P., Barabasi, A.L., and Oltvai, Z.N. (2001). Lethality and

centrality in protein networks. Nature 411, 41–42.

Kayed, R., Head, E., Thompson, J.L., McIntire, T.M., Milton, S.C., Cotman,

C.W., and Glabe, C.G.(2003). Common structureof soluble amyloid oligomers

implies common mechanism of pathogenesis. Science 300, 486–489.

Keshava Prasad, T.S., Goel, R., Kandasamy, K., Keerthikumar, S., Kumar, S.,

Mathivanan, S., Telikicherla, D., Raju, R., Shafreen, B., Venugopal, A., et al.

(2009). Human Protein Reference Database–2009 update. Nucleic Acids

Res. 37 , D767–D772.

Kim, W., Kim, Y., Min, J., Kim, D.J., Chang, Y.T., and Hecht, M.H. (2006).

A high-throughput screen for compounds that inhibit aggregation of the

Alzheimer’s peptide. ACS Chem. Biol. 1, 461–469.


http://www.matrixscience.com/


https://proteomecommons.org/tranche

http://dx.doi.org/doi:10.1016/j.cell.2010.11.050

http://dx.doi.org/doi:10.1016/j.cell.2010.11.050






Kyte, J., and Doolittle, R.F. (1982). A simple method for displaying the hydro-

pathic character of a protein. J. Mol. Biol. 157 , 105–132.

Lam, Y.C., Bowman, A.B., Jafar-Nejad, P., Lim, J., Richman, R., Fryer, J.D.,

Hyun, E.D., Duvick, L.A., Orr, H.T., Botas, J., et al. (2006). ATAXIN-1 interacts

with the repressor Capicua in its native complex to cause SCA1 neuropa-

thology. Cell 127 , 1335–1347.

Lashuel, H.A., and Lansbury, P.T., Jr. (2006). Are amyloid diseases caused by

protein aggregates thatmimic bacterial pore-forming toxins? Q. Rev.Biophys.

39, 167–201.

Lee, E.K., Park, Y.W., Shin, D.Y., Mook-Jung, I., Yoo, Y.J., Lee, E.K., Park,

Y.W., Shin, D.Y., Mook-Jung, I., and Yoo, Y.J. (2006). Cytosolic amyloid-

beta peptide 42 escaping from degradation induces cell death. Biochem.

Biophys. Res. Commun. 344, 471–477.

Linding, R., Jensen, L.J., Diella, F., Bork, P., Gibson, T.J., and Russell, R.B.

(2003). Protein disorder prediction: implications for structural proteomics.

Structure 11, 1453–1459.

Morimoto, R.I. (2008). Proteotoxic stress and inducible chaperone networks in

neurodegenerative disease and aging. Genes Dev. 22, 1427–1438.

Mossuto, M.F., Dhulesia, A., Devlin, G., Frare, E., Kumita, J.R., de Laureto,

P.P., Dumoulin, M., Fontana, A., Dobson, C.M., and Salvatella, X. (2010).

The Non-Core Regions of Human Lysozyme Amyloid Fibrils Give Rise to

Cytotoxicity. J. Mol. Biol. 402, 783–796.

Nesterov, E.E., Skoch, J., Hyman, B.T., Klunk, W.E., Bacskai, B.J., and

Swager, T.M. (2005). In vivo optical imaging of amyloid aggregates in brain:

design of fluorescent markers. Angew. Chem. Int. Ed. 44, 5452–5456.

Nilsberth, C., Westlind-Danielsson, A., Eckman, C.B., Condron, M.M.,

Axelman, K., Forsell, C., Stenh, C., Luthman, J., Teplow, D.B., Younkin,

S.G., et al. (2001). The ‘Arctic’ APP mutation (E693G) causes Alzheimer’s

disease by enhanced Abeta protofibril formation. Nat. Neurosci. 4, 887–893.

Ong, S.E., and Mann, M. (2006). A practical recipe for stable isotope labeling

by amino acids in cell culture (SILAC). Nat. Protoc. 1, 2650–2660.

Pentony, M.M., and Jones, D.T. (2009). Modularity of intrinsic disorder in the

human proteome. Proteins 78, 212–221.

Raychaudhuri, S., Dey, S., Bhattacharyya, N.P., and Mukhopadhyay, D.

(2009). The role of intrinsically unstructured proteins in neurodegenerativediseases. PLoS ONE 4, e5566.

Shearman, M.S. (1999). Toxicity of protein aggregates in PC12 cells:

3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay. Methods

Enzymol. 309, 716–723.

Stefani, M., and Dobson, C.M. (2003). Protein aggregation and aggregate

toxicity: new insights into protein folding, misfolding diseases and biological

evolution. J. Mol. Med. 81, 678–699.

Su, A.I., Cooke, M.P., Ching, K.A., Hakak, Y., Walker, J.R., Wiltshire, T., Orth,

A.P., Vega, R.G., Sapinoso, L.M., Moqrich, A., et al. (2002). Large-scale anal-

ysis of the human and mouse transcriptomes. Proc. Natl. Acad. Sci. USA 99,

4465–4470.

Sullivan, E.K., Weirich, C.S., Guyon, J.R., Sif, S., and Kingston, R.E. (2001).

Transcriptional activation domains of human heatshock factor 1 recruithuman

SWI/SNF. Mol. Cell. Biol. 21, 5826–5837.

Tartaglia, G.G., Pawar, A.P., Campioni, S., Dobson, C.M., Chiti, F., and

Vendruscolo, M. (2008). Prediction of aggregation-prone regions in structured

proteins. J. Mol. Biol. 380, 425–436.

Tatzelt, J., Zuo, J., Voellmy, R., Scott, M., Hartl, F.U., Prusiner, S.B., and

Welch, W.J. (1995). Scrapie prions selectively modify the stress response in

neuroblastoma cells. Proc. Natl. Acad. Sci. USA 92, 2944–2948.

Wei, Y.N., Liu, T., Sazinsky, S.L., Moffet, D.A., Pelczer, I., and Hecht, M.H.

(2003). Stably folded de novo proteins from a designed combinatorial library.

Protein Sci. 12, 92–102.

West, M.W., Wang, W.X., Patterson, J., Mancias, J.D., Beasley, J.R., and

Hecht, M.H. (1999). De novo amyloid proteins from designed combinatorial

libraries. Proc. Natl. Acad. Sci. USA 96, 11211–11216.

Williams, G.T., McClanahan, T.K., and Morimoto, R.I. (1989). E1a transactiva-

tion of the human HSP70 promoter is mediated through the basal transcrip-

tional complex. Mol. Cell. Biol. 9, 2574–2587.

Winklhofer, K.F.,Tatzelt, J.,and Haass, C. (2008). Thetwo faces of proteinmis-

folding: gain- and loss-of-function in neurodegenerative diseases. EMBO J.

27 , 336–349.

Young, J.C., Hoogenraad, N.J., and Hartl, F.U. (2003). Molecular chaperones

Hsp90 and Hsp70 deliver preproteins to the mitochondrial import receptorTom70. Cell 112, 41–50.




Supplemental Information

EXTENDED EXPERIMENTAL PROCEDURES

Plasmids

For bacterial expression, the N-terminally Myc-tagged b proteins were cloned into the vector pTrcHis (Invitrogen), excluding the His-

tag of the vector. a-S824 protein was cloned into the pProEx HT vector. For mammalian expression, N-terminally Myc-taggedproteins were cloned into pcDNA3.1/Myc-His (Invitrogen), excluding the Myc- and His-tags of the vector. Human HSP110 was in-

serted into the vector pcDNA3.1 (Invitrogen). HSP70-luciferase ( Williams et al., 1989 ) and NF-kB-luciferase ( Heil et al., 2004 ) reporter

constructs were kindly provided by Dr. H. Wagner (Technical University Munich). GFP fusion constructs of wild-type (WT) A b1-42 and

its Arctic mutant (E22G) ( Wurth et al., 2002 ) were cloned into pcDNA3.1 using the restriction sites HindIII/NotI.

Protein Purification

Protein expression in E. coli BL21 cells (DE3) was induced with 0.5 mM IPTG at an OD600 of 1.0 for 4 hr at 37C. The cells were

collected by centrifugation at 5000 x g for 10 min and resuspended in lysis buffer (25% sucrose, 50 mM Tris pH 8.0, 1 mM

EDTA), supplemented with complete protease inhibitor (Roche) and 1 mg/ml lysozyme (Sigma). The cells were lysed by sonication

or repeated freeze-thaw cycles and the DNA was digested by Benzonase (Novagen). Inclusion bodies were isolated by repeated

washing in0.5%TritonX-100,1 mMEDTAandcentrifugation(20.000 x g for 15 min). The pelletwas dissolved in8 M ureaand purified

on a MonoQ 10/100 HR16/10 column in 8 M urea, 25 mM Tris pH 7.5 using a gradient from 0 to 1 M NaCl. The b-protein-containing

fractions were further purified by size exclusion chromatography on Sephacryl S-300 HiPreP 26/60 in 8 M urea, 0.1 M NaCl, 25 mMTris pH 7.5. After dialysis overnight into 0.1 M NaCl, 25 mM sodium phosphate pH 7.5, the proteins formed soluble oligomers/aggre-

gates, which eluted in the excluded volume of Sephacryl S-300 size-exclusion chromatography. The protein concentration of the

b sheet proteins was determined by UV absorbance at 210 nm (Abs210 0.1% (= g/l) = 20). All proteins were stored at À80C in

25 mM sodium-phosphate pH 7.5, 100 mM NaCl.

To purify a-S824 protein, cell lysate was centrifuged at 20.000 x g for 15 min, and the supernatant was applied to Ni-NTA agarose

chromatography following standard procedures. The protein concentration of the purified a-S824 protein was determined by UV

absorbance at 280 nm ( 3280 = 12950 M-1 cm-1 ). The protein was stored atÀ80C in 25 mM sodium-phosphate pH 7.5, 100 mM NaCl.

Antibodies

A11 anti-oligomer antibody was kindly provided by Dr. C. Glabe (University of California, Davis). Anti-Myc (Cy3-conjugated), anti-

rabbit IgG (peroxidase-conjugated), and anti-mouse IgG (peroxidase-conjugated) were purchased from Sigma, anti-Hsp110, anti-

Myc, anti-VDAC1, anti-RanBP1, anti-eIF3D, anti-eIF4GII and anti-Vigilin from SantaCruz, anti-GAPDH from Chemicon, anti-GFP

from Roche, anti-Myc (FITC-conjugated) from Zymed, and anti-mouse IgG (FITC-conjugated) from Biosource and anti-mouse IgG(Cy3-conjugated) anti-rabbit (Cy3-conjugated) from Dianova.

Cell Culture, Immunoblotting, and Reporter Assays

HEK293T cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) (Biochrom KG), supplemented with 10% fetal calf

serum (FCS), 100 IU/ml penicillin G, 100 mg/ml streptomycin sulfate, 2 mM L-glutamine and nonessential amino acid cocktail (all

from GIBCO). Transient transfections were performed by electroporation with 30 mg expression vector. Lipofectamin (Invitrogen)

transfection was performed for overexpression of Hsp110. Immunoblots were developed using the chemiluminescence kit Rodeo

ECL (USB) and analyzed using a LAS-3000 image reader (Fujifilm) and the AIDA software (Raytest). For luciferase reporter assays,

cells were lysed in Lysis Buffer (Promega). After mixing 20 ml of lysate with 50 ml luciferin solution (Promega), luciferase activity was

measured using a Lumat LB9507 (EG&G Berthold).

Solubility and Detection of A11 Reactive OligomersHEK293T cells were lysed in PBScontaining 0.5% (v/v) Triton X-100, 0.5% (w/v) sodium deoxycholateand 10 mM MgCl2. In addition,

a protease inhibitor cocktail (without EDTA) containing 0.3 mg/ml Leupeptin (Roche), 0.3 mg/ml Aprotinin (Roche), 1 mM AEBSF

(Roche), 0.3 mg/ml Prestatin A (Sigma), and 1 mM PMSF (Serva) was added. After 20 min, 25 U/ ml benzonase was added, and

the samples incubated with agitation for a further 40 min at 4C.

For size fractionation, 300-450 mg total protein was loaded on a Superose 6 column using a SMART system (Amersham Biosci-

ence). The column was equilibrated and run using PBS. Lysate samples were separated into 22 fractions. 10 ml of each fraction

was dotted on a nitrocellulose membrane and allowed to dry for 10 min at room temperature. The membranes were blocked in

1% skimmed milk in Tris-buffered saline (TBS)/0.05% Tween-20 and subsequently incubated with A11 antibody (1:1000 dilution)

overnight at 4C. Bound A11 was detected with anti-rabbit-antibody (conjugated to HRP) followed by chemiluminescence detection.

The signal was visualized and quantified as described for immunoblotting. In parallel, a dot blot of the fractions with anti-Myc anti-

bodieswas performedto estimate theamount of bproteins. The cumulativeA11 signal was normalized relative to the cumulative anti-

Myc signal.

Cell 144, 67–78, January 7, 2011 ª2011 Elsevier Inc. S1



Immunofluorescence and Fluorescence Imaging

HEK293T cells were fixed in 4% paraformaldehyde/PBS solution for 1 hr, permeabilized with 0.1% (v/v) Triton X-100 for 5 min,

washed twice with PBS, blocked with 1% BSA solution (PBS-B) and then successively treated with primary and secondary anti-

bodies (1 hr each). Fluorescent dye-conjugated primary antibodies were diluted 1:200, other primary antibodies were diluted 1:50

to 1:300 in PBS/1% BSA. The secondary antibodies were diluted 1:200. During the last incubation step with secondary antibody,0.25 mg/ml DAPI (Invitrogen) was added. After the last incubation, cells were washed 3-times with PBS and mounted in Prolong Anti-

fade (Molecular Probes).

To detect amyloid species, cells were stained with 10mM NIAD-4 (dissolved in 10% DMSO/90% 1,2-propanediol) for 1 hr, washed

3-times with PBS, fixed, permeabilized, blocked, treated with DAPI and mounted as described for immunofluorescence. NIAD-4 fluo-

rescence was excited at 475 nm and emission recorded at 625 nm.

To visualize the actin cytoskeleton, cells were washed, fixed, and permeabilized as described above. Cells were incubated with

165 nM Rhodamine-conjugated Phalloidin in PBS for 1 hr, washed 3-times with PBS and mounted.

Detection of Apoptotic Cells

23 106 of the HEK293T cells were resuspended in 100 ml PBS and added to 0.5 ml of acridine buffer I (20 mM citrate-phosphate, pH

3.0/0.1 mM EDTA/0.2 M sucrose/0.1% Triton X-100). 0.5 ml acridine buffer II (10 mM citrate-phosphate pH 3.8/0.1 M NaCl) was

added together with 20 mg/ml acridine orange (Invitrogen). Cells were resuspended and analyzed by fluorescence microscopy.

TNF/cycloheximide/staurosporine treated cells were used as a reference for apoptoticmorphology. At least 200 cells were evaluated

per sample.

Radioactive Protein Labeling

Transfected HEK293T cellswere platedon 6-wellplates. 24 hr later,cellswere washedfirst with PBS, then with prewarmedFCS-free,

methionine-reduced DMEM (DMEM minus). Labeling was initiated by adding 50 mCi/ml 35S-Met (EasyTag, NEN Radiochemicals) in

DMEM minus. 15 min later, translation was stopped by discarding the radioactive medium and applying cold PBS. The cells were

lysed in 1% Triton X-100/PBS with Protease inhibitor cocktail (Roche). DNA was digested by adding 2.5 mM magnesium chloride

and Benzonase. The extracts were mixed with SDS gel loading buffer, heated and separated on a NuPAGE 4%–12% Bis-Tris gel.

For immunoprecipitation, total extracts were centrifuged at 400 x g for 5 min to remove raw cellular debris, and 15 ml anti-Myc Mi-

croBeads ( mMACS) were added for one hour. The unbound material was separated using MACS Separation columns (Miltenyi Bio-

tec). The beads and the associated proteins were washed three times with 0.1% Triton X-100/PBS, once with PBS and eluted with

120 ml SDS sample buffer without b-mercaptoethanol ( b-ME). Subsequently, b-ME was added to eluates, the eluates were heated

and analyzed together with total lysate fractions by SDS-PAGE, Coomassie blue staining and fluorography.

Fluorescence Spectroscopy

Fluorescence measurements with purified proteins were performed on a FluoroLog-3 spectrofluorometer. The fluorescent dyes ThT

and ANSwere adjusted to a final concentration of 20mM,NIAD-4to 1 mM.ThT fluorescence was excited at440 nm, NIAD-4 at475 nm

and ANS at 375 nm .

Circular Dichroism Spectroscopy

CD measurements were performed on a Jasco CD Spectrometer J-715 at 25C in 6 mM HEPES pH 7.5/25 mM KCl, at a protein

concentration of 0.1 mg/ml. Spectra were recorded in a 0.1 mm quartz cuvette between 197 and 250 nm (with a bandwidth of

1 nm and a scanning speed of 50 nm/min). Each single spectrum was averaged from 3 accumulative scans. Secondary structure

content was analyzed with the CDSSTR algorithm (Jasco).

Fourier Transform Infrared (FTIR) Spectroscopyb proteins (100mM in 25 mM sodium phosphatepH 7.5, 100 mM NaCl) were dialyzed overnight against 10 mM potassium phosphate

pH 6.0/10 mM NaCl;a-S824 wasdialyzedagainst10 mM potassium phosphatepH 7.5/10 mM NaCl. Infrared spectra were measured

in the Vertex 70 spectrophotometer (Bruker, Germany) equipped with a TGS detector using attenuated total reflection (ATR) with

parallelogram-shaped Germanium (Ge) crystals as internal reflection plates (Korth Kristalle, Germany). A thin film of 50-100 mg of

the respective protein was dried under N2 on one side of the Ge crystal that was placed in a home-made gas-tight chamber ( Heinz

et al., 2003 ). H-D exchange was performed by flushing D2O-saturated N2 through the chamber and monitored every 1 to 3 min until

the spectra were stable (usually after 45 min). Spectra were recorded before and after H-D exchange with a nominal resolution of

2 cm-1 in the double-sided, forward-backward mode, collecting 1024 scans per sample. Water vapor and CO 2 contributions were

corrected for using the atmospheric compensation of the OPUS software (version 6.5) from Bruker. The spectral region of the amide

I band (1705 to1595cm-1 ) was extracted, corrected forthe base line, andscaled so as to obtain a constant integral value forcompar-

ison. Peak positions of spectral components were analyzed using Fourier-self deconvolution and the second derivative of the

spectra.

S2 Cell 144, 67–78, January 7, 2011 ª2011 Elsevier Inc.



Electron Microscopy

Carbon-coated copper grids were covered with theb protein aggregates either in 25 mM HEPES pH7.5, 150 mM KCl, 0.5 mM MgCl2or 10 mM potassium phosphate pH 6.0 buffer for 1 min, washed 3 times, stained for 1 min in 2% uranyl acetate and air-dried. A

CM200 FEG transmission electron microscope (FEI, Eindhoven) with a CCD-camera was used at an accelerating voltage of 160

kV and magnification of 55,000 x.

Mass Spectrometry (MS)

SILAC Medium and Sample Preparation

Custom DMEM (GIBCO) was used for SILAC labeling: originally arginine- and lysine-free DMEM was supplemented with 10% dia-

lyzed FCS, 100 IU/ml penicillin G, 100 mg/ml streptomycin sulfate, 2 mM L-glutamine and nonessential amino acid cocktail (GIBCO).

To prepare L, M, and H media, the respective amino acids were added, for L: Arg0 and Lys0 (arginine and lysine, Sigma), for M: Arg6

and Lys4 (arginine-13C6 and lysine-4,4,5,5-d4, Isotec), for H: Arg10 and Lys8 (arginine-13C6,15N4 and lysine-13C6,15N2, Isotec).

107

SILAC labeled HEK293T cells were transfected. Live cells, attached to the support, were lysed 24 hr later. To this end, cells

were collected and washed by centrifugation. 1% Triton X-100/PBS with Protease Inhibitor Cocktail was added for 20 min. 8 ml ben-

zonase were added, and the cell lysate was incubated for further 40 min. To remove cell debris, the lysate was cleared by centrifu-

gation at 2000 x g for 5 min, the supernatant was removed, and its protein concentration was determined. The lysates labeled with

different isotopes were mixed at a 1:1:1 ratio. 50 ml mMACS anti-Myc (for b proteins) or anti-GFP (for A b1-42 and Arctic mutant

constructs) MicroBeads were used to isolate Myc-tagged and GFP-tagged proteins, respectively, and their interactome by rotating

the mixed lysates with the beads for 1 hr at 4C. Samples were applied onto MACS columns equilibrated with 200 ml lysis buffer. Thecolumns were washed 4 times with 200 ml 0.1% Triton X-100/PBS and once with PBS. Bound proteins were eluted with 120 ml LDS

samplebuffer (Invitrogen), followed by the addition of 2%b-mercaptoethanol and heating at 70C for10 min. Eluates were separated

on NuPAGE gradient gels, gels were fixed and stained with Colloidal Blue (Invitrogen), according to the manufacturer’s instructions.

Preparation of gel slices, reduction, alkylation, and in-gel protein digestion was carried out as described ( Ong and Mann, 2006 ).

Finally, peptides were desalted, filtered, and enriched on OMIX-C18 tips (Millipore).

LC-MS/MS

Peptides were eluted from OMIX tips using 35 ml of 80% methanol/0.1% TFA. The samples were dried in a vacuum centrifuge

concentrator at 45C until the volume was less than 5 ml (about 30 min). The volume was increased to 6 ml using 0.1% formic acid

(FA). Using an EasyLC system (Proxeon), 5 ml of sample were loaded at 0.5 ml/min in 0.1% FA onto a 15 cm long capillary column

(75 mm inner diameter) with a prepulled capillary tip (Proxeon) packed with Reprosil-Pur 3 mm C18 material (Dr. Maisch). Peptides

were eluted at 0.3 ml/min using a 120 min gradient (immunoprecipitate eluants) or 150 min gradient (lysate and flowthrough samples)

from 2%–64% acetonitrile, in 0.1% FA. In two cases, an Ultimate 3000 HPLC (Dionex) was used in place of the EasyLC; peptides

were separated as above, except that a C18 precolumn (Dionex) was used to preconcentrate the peptides before elution.

Peptides were directly injected into a Thermo LTQ-FT Ultra using a nano-electrospray ion source (Proxeon), with electrospray volt-

ages ranging from 1.5 to 2.5 kV. FT scans from m/z 400-2000 were taken at 100,000 resolution, followed by collision induced disso-

ciation (CID) scans in the LTQ of the 8 most intense ions with signal greater 2000 counts, and charge state larger than one. Dynamic

exclusion of parent masses already fragmented was enabled. CID settings were as follows: isolation width 3, normalized collision

energy 45 V, activation Q 0.220, and activation time 30 ms.

Analysis of MS Data

Generation of Ratios

MS data were analyzed with MaxQuant version 1.0.12.31 ( Cox and Mann, 2008 ) using the following parameters: Quant; SILAC Trip-

lets, Medium Labels: Arg6, Lys4, HeavyLabels: Arg10, Lys8. Maximum labeled amino acids: 3. Variable modifications: Oxidation(M),

Acetyl (Protein N-terminus). Fixed Modifications: Carbamidomethyl (C). Database: Human IPI, released March 3 2009 with contam-

inants anda decoy database added by the SequenceReverser.exe program released with MaxQuant v. 1.0.12.4. Enzyme: Trypsin/P.

MS/MS tolerance: 0.5 Da. Maximum missed cleavages: 2. Top MS/MS peaks/100 Da: 6. Mascot version 2.2 (Matrix Sciences,www.

matrixsciences.com ) was used to generate search results for MaxQuant. Identify; Peptide FDR: 0.01. Protein FDR: 0.01. MaximumPEP: 1. Minimum unique peptides: 1. Minimum peptide length: 6. Minimum peptides: 1. Protein Quantitation based on Razor and

Unique peptides. Minimum Ratio count: 2. ‘Requantify’ and ‘Keep low-scoring versions of identified peptides’ were both enabled.

Additionally, for those proteins for which no quantitation was available, but for which at least two unique peptides were found

only in one isotope but not another isotope, arbitrary ratios of 10 or 0.1 were assigned, and Significance B was assigned the value

of ‘0’. Unless noted otherwise, normalized ratios were used. Supplemental Tables show the leading protein for each protein group;

that is, the protein which best matched all of the peptides identified. The raw MS data along with a full list of identified proteins and

quantitations is available at https://proteomecommons.org/tranche, entering the following hash: +Ff0/p8lSBrrzCKZfzAwYS3

+Bqw5fonokB679f136te2iklhHtFMUpeT5SM/I3XuufTyrXj0ycVVC6G4Li/L02 dA4jcAAAAAAABVfg = =.

Determination of b Protein Interactors

Four sets of SILAC experiments were performed: I: empty vector (L) versus a-S824 (M) versus b23 (H); II: a-S824 (L) versus b4 (M)

versus b17 (H); III: b4 (L) versus b17 (M) versus b23 (H); IV: a-S824 (L) versus b23 (M) (see schematic in Figures 3 A). Experiment IV


http://www.matrixsciences.com/








is part of the pulse SILAC analysis described below and for the purpose of interactor identification is comparable to experiment I.

Each experiment was performed 3 times independently (biological replicates). A protein was considered a b protein interactor if

an elevated ratio (95% confidence) wasfoundin at least twoof thebiological replicates in anyexperiment. b23 interactors were deter-

mined from experiments I, III, and IV; for experiments I and IV, proteins with elevated b23/ a-S824 ratios with Significance B (gener-

ated by MaxQuant) less than 0.05 were selected. For experiment III, proteins with an elevated ratio for b23/ b17 or for b23/ b4 withSignificance B less than 0.05 were selected. This results in 105 interactors (Table S1) which were found to interact with b23in atleast

two replicates of any one experiment (I, III or IV). In order to define more stringently identified interactors, subsets of this list were

generated in which an interactor was found in at least 1, 2 or 3 additional replicates in either of the two remaining experiments.

This is shown in the ‘confidence’ column of Table S1. Note that only a minimal further increase in the number of newly identified in-

teractors was observed after 6 biological replicates.

b17 and b4 interactors were determined from experiments II andIII. Proteins which were enriched onb17or b4 againsta-S824 in at

least two replicates of experiment II were considered interactors. Additional b17andb4 interactors were determined from Experiment

III if they were enriched to Significance B < 0.05 against one of the other two b proteins in at least two replicates, as described above

for b23. Note that each replicate of experiment III was based on the same b23 transfection used in experiment I above, and thus the

replicates were coupled for the purpose of analysis. This allowed us to calculate ratios of b17/ a-S824 and b4/ a-S824. As a result, 73

b17 (Table S2) and 57b4 (Table S3) interactors were determined by requiring enrichment in at least two replicates in any one of three

experiments, analogous to what was done for b23. However, since ratios from both experiments I and III were needed to determine

b17/ a-S824 and b4/ a-S824 ratios, there was a reduced chance of finding b17 and b4 interactors than of finding b23 interactors from

experiment I. Therefore, in order to compare thenumbers of interactors for eachb protein, a list of b23 interactors was made in whichonly those interactors from experiment I were considered which also had a corresponding ratio in experiment III. This results in 94

interactors of b23; proteins from the list of 105 which are not included in this list of 94 are highlighted in yellow in Table S1. Note

that these 11 proteins were removed from thelist of b23 interactors only for thepurposesof comparison to theb17 and b4 interactors;

they are not identified with lower confidence than the other b23 interactors.

One of the proteins identified as the leading protein was IPI00179330.6 (RPS27A). This is a fusion protein between the small ribo-

somal protein 27A and ubiquitin, which is cleaved after synthesis. Peptides from both portions were identified, however closer

inspection showed that only the peptides corresponding to theubiquitin moiety were enriched onb23. We therefore decided to iden-

tify this protein according to the second protein in the protein group, IPI00743650.1 (UBC).

Determination of a-S824 Interactors

Three biological replicates of two independent experiments were performed in whicha-S824 (M) ora-S824 (H) was analyzed relative

to empty vector (L), respectively. a-S824 interactors were determined exactly as described above for b23, namely that the interactor

must be identified with 95% confidence, according to Significance B, in at least two of the three replicates in any experiment.

Determination of Preferred Interactors

Preferred b protein interactors were considered to be those which had an enriched ratio and Significance B less than 0.05 for one

b protein compared to at least one of the other two b proteins in at least two of the three replicates of experiment III (Tables S1–

S3; proteins in bold).

Depletion of Interactors

The depletion of interactors from the total cell lysate by binding to b23 aggregates was estimated by analyzing the supernatants

(technically flow-through fractions) from the IPs in experiment I. H/M ratios were used to determine the amount of each of the inter-

actors remaining in the cell lysate after IP. The depletion was then estimated by dividing by the approximate transfection efficiency

(50%).A depletionvalue is reported when an H/M ratio was measured in thesupernatant in at least twoout of thethree replicates and

when the H/M ratio is 0.95 or lower, corresponding to a mean depletion of greater than 10% (Table S1). Relative error in the SILAC

ratios prevents the calculation of reliable estimates for depletion when the depletion is very low. Therefore, for proteins where the

mean H/M ratio of at least two replicates is 0.95 or higher, a depletion value of % 10% has been entered in Table S1. Values of deple-

tion around 10% and higher were confirmed by quantitative western blotting in which protein amounts in IPs were compared with

amounts in total lysate (Figure S3F and Figure 4 A, and data not shown).

Pulse-SILAC

The design of this experiment (also see experiment IV above) is shown in Figure 6B. The M-labeled cells were transfected with b23.

After the transfection, half of the cells were shifted to H medium, while the other half continued in M medium. TheL-labeled cells were

transfected witha-S824, and half of them were left on L medium, as a control. Lysates were prepared, mixed and analyzed as above.

The measured H/Mlysate24h ratio of individual proteins in the lysate was used to calculate the expected H/Mlysate12h, the ratio at the

midpoint of the experiment, according to the equation:

H=M lysate12 h =H=M lysate24 h

2+H=M lysate24 h

:




In addition to the predominant contribution of protein interacting directly with b23, the H/MIP measured in the pulldown also

has a small contribution from background protein. Since background proteins will have the same isotope ratios as the lysate

(H/Mlysate24h ), the H/MIP was corrected according to the enrichment on b23 given by the M/LIP ratio. This generates the ratio H/

Mb23. Deviation of log of the ratio ((H/Mb23 )/(H/Mlysate12h )) from 0 indicates enrichment of newly synthesized or preexistent fraction

of a protein on b23. Nonnormalized H/M ratios were used for each of these calculations.

Determination of A b42 Interactors

A b42 interactors were determined from experiment V: GFP (L)versus A b42 WT (M) versus A b42 Arctic mutant (H); three replicates were

performed. For comparison of the A b42 interactors to the b protein interactors, all proteins which were enriched on either A b42 WT or

A b42 Arctic mutant compared to GFP to Significance B < 0.05 in at least two out of three of the replicates of experiment V were

included. Those proteins which are in common with the b protein interactors are shown in Table S5. In order to determine which

proteins interact preferentially with A b42 Arctic mutant, proteins which had an elevated A b42 Arctic/A b42 WT ratio and Significance

B < 0.01 in any replicate of experiment V were included (Table S6).

Bioinformatic Analysis

The SUPERFAMILY predictions ( http://supfam.cs.bris.ac.uk ) according to SCOP 1.73 classification ( http://scop.mrc-lmb.cam.ac.

uk/ ) where used to assign protein folds. Information on protein interactions is based on HPRD Release 9 ( Keshava Prasad et al.,

2009 ) ( www.hprd.org ). This data is combined with the information about gene essentiality in mouse from MGD ( Bult et al., 2008 )( http://www.informatics.jax.org/ ) and the compilation of proteins associated with neurodegenerative disease proteins ( Raychaudhuri

et al., 2009 ). The cellular abundance of the b protein interactors was estimated according to HEK293T mRNA levels measured by Su

et al. (2002) ( http://biogps.gnf.org ).

Prediction of Intrinsic Aggregation Propensities

The intrinsic aggregation propensity, p agg i , of an individual amino acid is defined as

p agg i =a h p

h i +a s p

s i +a hyd p

hyd i +ac p

c i

where p h i , p s

i , p hyd i , pc

i are the amino acid scales for a-helix and b sheet formation, hydrophobicity and charge and the a’s are the

corresponding weights (see below) ( Tartaglia et al., 2008 ). An aggregation propensity profile is then defined as

p

agg

i =

1

7X

3

j =À3 p

agg

i + j +

a pat I

pat

i +

a gk I

gk

i

where we considered theaggregation rate of a seven-residue segment of theprotein centered at position i . I pat i and I

gk i are included,

respectively, to account forthe presence of hydrophobic patterns andof gatekeeper residues. Theterm I pat i is 1 if residue i is included

in a hydrophobic pattern over five residues and 0 otherwise, while the term I gk i is defined as

I gk i =

X10

j =À10

c i + j

where the sum over the charges c i of individual amino acids is made over a sliding window of 21 residues; shorter windows are

considered at theN- andC-termini.The term I gk i is introduced to take into account thefact that, when a hydrophobic pattern is flanked

by charged residues, its contribution to the aggregation propensity is much reduced by electrostatic repulsions.

By normalizing p agg i , the intrinsic aggregation propensity score, Z

agg i , is obtained, which is used to compare the aggregation

propensity of different sequences

Z agg i =

p agg i À m agg

s agg

where m agg is the average value of Z agg i over a set of random polypeptides having the same length as the sequence of interest, and

s agg is the corresponding standard deviation from the average ( Tartaglia et al., 2008 ). The corresponding intrinsic propensity for

aggregation of the protein without structural correction, Z agg, is calculated as

Z agg=

1

N

XN

i =1

Z agg i

The weights a’s are determined by fitting the Z agg scores against a database of aggregation rates measured experimentally ( Tar-

taglia et al., 2008 ).


http://supfam.cs.bris.ac.uk/

http://scop.mrc-lmb.cam.ac.uk/


http://www.hprd.org/


http://www.informatics.jax.org/

http://biogps.gnf.org/

http://biogps.gnf.org/

http://www.informatics.jax.org/




http://supfam.cs.bris.ac.uk/



Prediction of Aggregation Propensities with Structural Corrections

The local stability of various regions in a protein can be investigated by using the CamP method, which uses the knowledge of the

amino acid sequence to provide a prediction of the protection factors from hydrogen exchange ( Tartaglia et al., 2007 ). By combining

the predictions of the intrinsic aggregation propensity profiles with those for the local stability in the folded state, we account for the

influence of the structural context on the aggregation propensities. We thus define ( Tartaglia et al., 2008 ) a sequence-dependentaggregation propensity score with the structural correction, ~ Z

agg

i , by modulating the intrinsic aggregation propensity profile with

the local stability score

~ Z agg

i = Z agg i ð1 À 3lnP i Þ

where 3 was fixed at 1/13 and lnP i, is the logarithm of the protection factor of residue i . The plot of ~ Z agg

i versus the residue number

represents the aggregation propensity profile calculated to account for the structural protection. The corresponding propensity for

aggregation of the protein, ~ Z agg

, is calculated as

~ Z agg

=

PN

i =1

~ Z agg

i wÀ~ Z

agg

i À ~ Z agg

0

Á

PN

i =1

wÀ~ Z

agg

i À ~ Z agg

0

Á

wherethe functionwð ~ Z agg i À ~ Z agg

0 Þ is1for ~ Z agg i À ~ Z agg

0 >0and0for~ Z agg i À ~ Z agg

0 <0; to predict ordered aggregation we use ~ Z agg0 =0, i.e.,

only positive contributions are taken into account; to estimate the propensity of a protein to be insoluble we use ~ Z agg

0 =~ Z min, i.e., both

positive and negative contributions are taken into account.

SUPPLEMENTAL REFERENCES

Bult, C.J., Eppig, J.T., Kadin, J.A., Richardson, J.E., and Blake, J.A. (2008). The Mouse Genome Database (MGD): mouse biology and model systems. Nucleic

Acids Res. 36, D724–D728.

Heil, F., Hemmi, H., Hochrein,H., Ampenberger, F., Kirschning, C., Akira, S., Lipford, G., Wagner, H., and Bauer, S. (2004). Species-specificrecognition of single-

stranded RNA via toll-like receptor 7 and 8. Science 303, 1526–1529.

Heinz, C., Engelhardt, H., Niederweis, M., Heinz, C., Engelhardt, H., and Niederweis, M. (2003). The core of the tetrameric mycobacterial porin MspA is an

extremely stable beta-sheet domain. J. Biol. Chem. 278, 8678–8685.

Tartaglia, G.G., Cavalli, A., and Vendruscolo, M. (2007). Prediction of local structural stabilities of proteins from their amino acid sequences. Structure 15, 139–

143.

Wurth, C., Guimard, N.K., and Hecht, M.H. (2002). Mutations that reduce aggregation of the Alzheimer’s A beta42 peptide: an unbiased search for the sequencedeterminants of A beta amyloidogenesis. J. Mol. Biol. 319, 1279–1290.

Zandomeneghi, G., Krebs, M.R., McCammon, M.G., Fandrich, M. (2004). FTIR reveals structural differences between native beta-sheet proteins and amyloid

fibrils. Protein Sci. 13, 3314–3321.




Figure S1. Characterization of b Proteins, Related to Figure 1

(A) Purified proteins b4, b17, b23, and a-S824. Left panel: 13.5% SDS-PAGE, Coomassie blue staining. Middle panel: Immunoblotting with anti-Myc antibodies.

Right panel: 13.5% SDS-PAGE with 4 M urea, Coomassie blue staining.

(B) Circular dichroism spectra of the purified model proteins. Spectra of 0.1 mg/ml protein solutions were recorded in 6 mM HEPES (pH 7.5)/25 mM KCl, as

detailed in Extended Experimental Procedures. The spectra of b4 (green), b17 (blue), b23 (red), and a-S824 proteins (gray) are representative scans of three

independent experiments. The table shows secondary structure content calculated from three independent experiments using the CDSSTR algorithm (Jasco).

(C) FTIR spectra of the purified model proteins. Spectra of $100 mg protein were recorded as detailed in Extended Experimental Procedures. Shown are

representative scans of 3 independent experiments. The peak at $1647 cm-1 originates from dominant a-helical structure and the band at $1624 cm-1 ischaracteristic of extendedb sheet structure in amyloid fibrils ( Zandomeneghiet al.,2004 ).The smallpeak at1695–1690 cm-1 seenfor theb proteinsis indicative of

anti-parallel b sheet structure. b4 possesses a significant spectral component at 1651 cm-1, suggesting the presence of some a-helical structure.




Figure S2. Cellular Effects of b Protein Expression in HEK293T Cells, Related to Figure 2

(A and B) b Protein expression induces cell death. HEK293T cells were transfected with the indicated proteins. After 3 days, surviving cells were counted by

Trypan Blue exclusion assay (A). The number of surviving cells after transfection with empty control vector, C, is set to 100%. The fraction of apoptotic cells was

determined by Acridine Orange staining and counting of cells displaying characteristic apoptotic morphology(B). Averagesand standarddeviations from at least

three experiments are shown. Statistical significance was estimated by t test: *, p < 0.05, **, p < 0.005.

(C) A11 antibody reactivity of extracts of b-protein-expressing cells. HEK293T cells were transfected with a-S824 or b23. After 24 hr, cell extracts were prepared

and separated by gel filtration on a Superose 6 column. Fractions were dot blotted and analyzed with the A11 anti-oligomer antibody. The A11 signal was

normalized to protein amounts, based on the signal obtained with anti-Myc antibody. Fraction numbers and positions of molecular weight markers are shown. A

quantification of these results is shown in Figure 2E.




Figure S3. Interactome Analysis of b Proteins, Related to Figure 3

(A) To control the amounts of isolated aggregates, anti-Myc immunoblotting analysis of the mixed lysates was performed before and after immunoprecipitation

(IP). Note that different mobility of the proteins in SDS-PAGE allows all three b proteins to be visualized on the same gel lane. The amount of lysate analyzed

corresponds to 12% of input for IP. Approximately 85% of the b proteins was immunoisolated.

(B) b23 interactors. HEK293T cells were SILAC-labeled and transfected with b23, a-S824, or vector only and processed as detailed in Extended ExperimentalProcedures and Figure 3 A. Normalized protein ratios fromtwo replicates are plotted as indicated. Red circlesindicateproteins which met thecriteria for inclusion

on theb23 interactor list( Table S1 ); whitecircles showproteins whichare not includedon the interactor list, becausethey did not reproducibly fulfill the criteriafor

enrichment. Proteins which had intensity only in one isotope were assigned the artificial value of 0.1 or 10, resulting in proteins with a b23/ a-S824 ratio of exactly

10.

(C and D) b4, b17 and b23 interactors. HEK293T cells were SILAC-labeled, transfected with b proteins and processed as detailed in Extended Experimental

Procedures and Figure 3 A. Normalized protein ratios are plotted as in (B); green circles show b4 interactors, blue circles show b17 interactors. As above, red

circles indicate b23 interactors and white circles show proteins not meeting the interactor criteria.

(Eand F) Validationof b protein interactors ( Tables S1–S3 ) by immunofluorescencemicroscopy (E) and western blotting of pulldowns(F). The interaction of eIF3D

(preferred interactor of b23 and b17) and Vigilin (preferred interactor of b17) was analyzed by both methods. HEK293T cells were transfected as indicated (C,

vector only control). 24 hr after transfection, cells were fixed and costained with anti-Myc, anti-eIF3D or anti-Vigilin antibodies. Nuclei were stained with DAPI.

Representative examples of two independent experiments are shown. No interaction of GFP and of an actin-GFP fusion protein with the aggregates was

observed(data not shown). Western blots werealso analyzedwith antibodies againstVDAC1 (preferred interactor of b17),RanBP1 (interactor of b17and b23)and

eIF3E (interactor of b4, b17 and b23) (F). Western blots are representative examples of two to three independent experiments. Quantitative western blotting in

comparison to lysate samples revealed that b4, b17, and b23 interacted with 6%, 10%, and 11% of VDAC1; 0%, 1.3%,and 1.8% of RanBP1; 6%, 9%, and 10%

of eIF3D; and 7%, 8%, and 8% of eIF3E, respectively. This analysis was based on averages from two to three experiments and took into account that the

transfection efficiency for the b proteins was $50%.(G) Impairment of protein synthesis uponb23 expression. 24 hr after transfection,HEK293T cells expressinga-S824 orb23 were pulsed-labeled with35S-Met for

15 min to determine the amount of newly synthesized proteins (New). Radioactive proteins were normalized to total Coomassie blue stained protein (Total). The

amount of newly synthesized protein in a-S824 expressing cells was set to 100%. b23 expression inhibited protein synthesis to 63± 19% (standard deviation of

seven experiments).

(H) Phalloidin-rhodamine staining of the actin cytoskeleton of C, a-S824, b4, b17, and b23 expressing cells 24 hr after transfection.




Figure S4. Hsp110 Expression in b-Protein-Expressing Cells, Related to Figure 4

(A) Levels of overexpression of Hsp110 in HEK293T cells detected by immunoblotting. Cells were transfected with human Hsp110 expression plasmid using

Lipofectamine.24 hr later theywere transfected with30 mgofthe b protein expression plasmidsor empty control vector by electroporation. Three dayslater,cells

were lysed and Hsp110 expression levels determined by immunoblotting with anti-Hsp110 antibody. GAPDH was detected with anti-GAPDH antibody asa loading control.

(B and C) Changes in cell morphology and b protein solubility upon overexpression of Hsp110. Transfection of Hsp110 and b proteins was performed as in (A),

except that 5 mg DNA was used for b protein expression. 24 hr hours later, cell morphology and aggregation of the model proteins were evaluated by confocal

microscopy using anti-Myc and anti-Hsp110 immunostaining. Nuclei were visualized by DAPI staining. An example of a cell with Hsp110 overexpression

containing soluble b17 and having preserved normal cell shape is indicated by a white arrow (B). The fraction of cells withb protein aggregates, soluble b protein

and normal morphology was determined by evaluating at least 900 cells per condition (C). In the absence of Hsp110 overexpression (solid bars), the number of

cells expressing the b proteins was set to 100%. In case of Hsp110 overexpression (hatched bars), cells containing elevated levels of Hsp110 were set to 100%.

Averages and standard deviations from three independent experiments are shown.




Figure S5. Structural Properties of interactors of b23 and A b42-GFP, Related to Figure 5

(A) Domain fold distribution of lysate proteins and b23 interactors. The frequency of domain folds (SCOP fold classes) among the proteins with identified fold

topology (72 b23 interactors and 2261 lysate proteins, corresponding to 129 and 3331 folds, respectively) is shown. Repetitive domain folds are counted only

once per protein. SCOP classification ( http://scop.mrc-lmb.cam.ac.uk/ ):a, allalpha; b, allbeta; c, alpha/betaproteins (a/b);d, alpha plus beta proteins (a + b);e,

multidomain proteins (alpha and beta); f, membrane and cell surface proteins and peptides; g: small proteins; h: coiled-coil proteins. *p < 0.05 forb23 interactorsrelative to lysate proteins, based on Chi-square test.

(B) Length distribution in amino acids (aa) of intrinsically unstructured regions (IURs) in lysate proteins and b23 interactors (p < 0.005 according to the Mann-

Whitney test).

(C–E) Expression of A b1À42 -GFP constructs and cell viability. (C) Anti-GFP western blot of HEK293T cells expressing GFP, WT A b1À42 or the Arctic mutant of

A b1À42 for 24 hr in the absence or presence of 1 mM MG132 (left panel). Asterisk indicates a proteolytic fragment of WT A b1À42 and the Arctic mutant. Cells

expressing b4, b17 or b23 in the absence or presence of 5 mM MG132 were analyzed by anti-Myc western blotting for comparison (right panel). (D) GFP fluo-

rescence microscopy of HEK293T cells expressing GFP, WT A b1À42 or Arctic mutant in the presence of 1 mM MG132. (E) Viability of cells expressing GFP, WT

A b1À42 or Arctic mutant in the absence or presence of 1 mM MG132, as measured by the MTT assay.

(F–I) Physicochemical properties of proteins that interact preferentially with the Arctic mutant of A b1À42. Interactors that were enriched on the Arctic mutant

relative to WT A b1À42 (31 proteins) were compared to lysate proteins in terms of molecular weight (F), hydrophobicity (G), fraction of amino acid residues in

disordered regions (H), and the presence of disordered stretches longer than 30 or 50 amino acids (I). Box plots indicate the distribution of the data. Dashed

horizontal lineindicates the median, whiskercaps and circlesindicate 10th/90thand 5th/95thpercentiles, respectively. *p < 0.05based on theMann-Whitneytest

(H) or Chi-square test (I); **p < 0.005 based on a Chi-square test.






Figure S6. Aggregation Propensities of the b23 Interactome, Related to Figure 6

(A) Newly synthesized proteins are susceptible to b23 coaggregation. HEK293T cells were transfected and newly synthesized proteins were labeled as in Fig-

ureS3G (Total). Subsequently,an anti-Mycimmunoprecipitation wasperformed to determine thefractionof new polypeptidesinteracting withthe modelproteins

(Bound). The average ± standard deviation from 5 independent experiments was 1 ± 0.5% for a-S824 and 7 ± 1.8% for b23. The amount of total newly

synthesized protein was set to 100%.

(B) Aggregation propensity of the unfolded polypeptides (Zagg predictor) is shown in solid bars; and aggregation propensity after correction for the local stabilities

of thefoldedstate,Zagg + lnPpredictoris shownin hatched bars ( Tartagliaet al.,2008 ) (seeExtended Experimental Proceduresfor details).Average scores for the

sets of proteins as indicated. Left: lysate proteins compared to b23 interactors ( Tables S1 and S4 ); right: newly synthesized b23 interactors (New) compared to

preexistentb23 interactors (Old) (top 15 versus bottom 15 proteins in Figure 6B and Table S7 ).

amyloid and essential cellular functions disrupted

Documents