REVIEW ARTICLEpublished: 20 June 2013
doi: 10.3389/fneur.2013.00076
Trinucleotide repeats: a structural perspective
Bruno Almeida, Sara Fernandes†, Isabel A. Abreu† and Sandra Macedo-Ribeiro*
Instituto de Biologia Molecular e Celular, Universidade do Porto, Porto, Portugal
Edited by:Thomas M. Durcan, McGill University,Canada
Reviewed by:Denis Soulet, Laval University, CanadaThomas M. Durcan, McGill University,Canada
*Correspondence:Sandra Macedo-Ribeiro, Instituto deBiologia Molecular e Celular,Universidade do Porto, Rua do CampoAlegre 823, 4150-180 Porto, Portugale-mail: [email protected]†Present address:Sara Fernandes, Shannon ABC,Limerick Institute of Technology,Limerick, Ireland;Isabel A. Abreu, GplantS, Instituto deTecnologia Química e Biológica,Oeiras, Portugal.
Trinucleotide repeat (TNR) expansions are present in a wide range of genes involved in sev-eral neurological disorders, being directly involved in the molecular mechanisms underlyingpathogenesis through modulation of gene expression and/or the function of the RNA orprotein it encodes. Structural and functional information on the role of TNR sequences inRNA and protein is crucial to understand the effect of TNR expansions in neurodegenera-tion.Therefore, this review intends to provide to the reader a structural and functional viewofTNR and encoded homopeptide expansions, with a particular emphasis on polyQ expan-sions and its role at inducing the self-assembly, aggregation and functional alterations ofthe carrier protein, which culminates in neuronal toxicity and cell death. Detail will be givento the Machado-Joseph Disease-causative and polyQ-containing protein, ataxin-3, provid-ing clues for the impact of polyQ expansion and its flanking regions in the modulation ofataxin-3 molecular interactions, function, and aggregation.
Keywords: amino acid-repeats, microsatellites, protein complexes, protein aggregation, amyloid, protein structure
TRINUCLEOTIDE REPEATS AND HUMAN DISEASETrinucleotide repeat (TNR) expansions and their association withneurological disorders have been known for the past 20 years(La Spada et al., 1991). Expansion of CAG, GCG, CTG, CGG,and GAA repeats located in coding or non-coding sequences ofdifferent genes (summarized in Table 1; Figures 1 and 2) are asso-ciated with a diverse range of human monogenic diseases suchas Spinobulbar Muscular Atrophy (SBMA, a.k.a. Kennedy dis-ease), Huntington Disease (HD), Spinocerebellar Ataxias (SCAs),Oculopharyngeal Muscular Dystrophy (OPMD), Myotonic Type 1(DM1), Fragile X-Associated Tremor Ataxia Syndrome (FXTAS),and Friedreich Ataxia (FRDA) (for a review see Orr and Zoghbi,2007), with longer repeats being correlated with earlier age at onsetand increased disease severity. These TNR are highly unstableand the repeat tract length can change between affected indi-viduals within the same family and can be different in differenttissues (La Spada, 1997; Brouwer et al., 2009). More interestingly,in the brain of patients affected by CAG expansions, differencesin repeat instability have been found between specific cell types(Pearson et al., 2005; Gonitel et al., 2008; Lopez Castel et al.,2010). GCG repeats are usually shorter and reveal a higher sta-bility in different tissues and across generations than CAG repeats.The dynamic nature of these DNA repeat expansions is a con-sequence of their capability to form different secondary struc-tures, which interfere with the cellular mechanisms of replication,repair, recombination and transcription (for a recent review seeLopez Castel et al., 2010). The molecular mechanisms underly-ing pathogenesis in those disorders, either associated with mentalretardation, neuronal, or muscular degeneration, might resultfrom alterations in the levels of gene expression and/or the func-tion of the RNA or protein it encodes, mechanisms that likely
act in concert to influence the pattern of selective cell toxic-ity. Some of those toxicity mechanisms will be briefly discussedbelow.
TRINUCLEOTIDE REPEATS AND RNA STRUCTUREThe formation of hairpin structures within the TNR RNA is relatedto the gain in RNA toxic function, the major pathogenic mecha-nism associated with CUG and CGG repeat expansions in non-coding regions of DM1 and FXTAS transcripts, which was alsoshown to contribute to pathogenesis in CAG repeat disorders suchas HD and Machado-Joseph disease (MJD, a.k.a. SCA3) (reviewedin Krzyzosiak et al., 2012). These duplex structures, whose sta-bility is positively correlated with the repeat size (Napierala andKrzyzosiak, 1997), sequester dsRNA binding proteins involved inmRNA splicing such as CUG-binding protein (CUGBP) and mus-cleblind protein 1 (MBNL1) (Miller et al., 2000), inducing aber-rant splicing in affected cells, compromising multiple intracellularpathways, affecting cell-quality control regulation, and ultimatelyresulting in cell dysfunction (Li and Bonini, 2010). Structural stud-ies on model trinucleotide CUG, CAG, and CGG repeats formingdouble-stranded chains revealed the features induced by peri-odic U-U, A-A, and G-G mismatches, and provided hints into thestructural details of pathogenic RNAs that are recognized by RNA-binding proteins (Mooers et al., 2005; Kiliszek et al., 2010, 2011;Kumar et al., 2011; Parkesh et al., 2011). MBNL1 is composedof four zinc-containing RNA-binding domains arranged in twotandem segments, with the C-terminal zinc-finger pair displayinga GC-sequence recognition motif (Teplova and Patel, 2008) andinteracting with the stem region of expanded CUG RNAs (Yuanet al., 2007). Electron microscopy analysis of MBNL1:CUG136
complexes showed that the pathogenic dsRNA forms a scaffold
www.frontiersin.org June 2013 | Volume 4 | Article 76 | 1
http://www.frontiersin.org/Neurologyhttp://www.frontiersin.org/Neurology/editorialboardhttp://www.frontiersin.org/Neurology/editorialboardhttp://www.frontiersin.org/Neurology/editorialboardhttp://www.frontiersin.org/Neurology/abouthttp://www.frontiersin.org/Neurodegeneration/10.3389/fneur.2013.00076/abstracthttp://www.frontiersin.org/Community/WhosWhoActivity.aspx?sname=BrunoAlmeida&UID=83513http://www.frontiersin.org/Community/WhosWhoActivity.aspx?sname=SaraFernandes&UID=88667http://www.frontiersin.org/Community/WhosWhoActivity.aspx?sname=IsabelAbreu&UID=83538http://www.frontiersin.org/Community/WhosWhoActivity.aspx?sname=SandraMacedo-Ribeiro&UID=83310mailto:[email protected]://www.frontiersin.orghttp://www.frontiersin.org/Neurodegeneration/archive
Almeida et al. Structure and function of trinucleotide repeats
Tab
le1
|Hu
man
dis
ease
sas
soci
ated
wit
hn
ucl
eoti
de
rep
eat
exp
ansi
on
s(a
dap
ted
fro
mM
essa
edan
dR
ou
leau
,200
9;Lo
pez
Cas
tele
tal
.,20
10;M
ato
set
al.,
2011
).
Dis
ease
nam
eR
epea
t
typ
e
Rep
eat
loca
tio
n
Gen
eP
rote
in(U
niP
rot
iden
tifi
er,n
um
ber
of
resi
du
es)
Bio
log
ical
pro
cess
a
No
rmal
rep
eat
len
gth
Dis
ease
rep
eat
len
gth
Pro
tein
stru
ctu
red
eter
min
ed?
Spi
nala
ndbu
lbar
mus
cula
rat
roph
y
(SB
MA
)
CA
GPr
otei
nco
ding
regi
on(p
olyQ
)
AR
And
roge
nre
cept
or
(P10
275,
919
resi
dues
)
Tran
scrip
tion,
tran
scrip
tion
regu
latio
n
9–36
38–6
2R
esid
ues
20–3
0an
d67
1–91
9(P
DB
code
1xow
)
Hun
tingt
on’s
dise
ase
(HD
)
CA
GPr
otei
nco
ding
regi
on(p
olyQ
)
HTT
Hun
tingt
in(P
4285
8,31
42
resi
dues
)
Apo
ptos
is6–
3436
–121
Res
idue
s5–
18(3
lrh),
Res
idue
s1–
17
(2ld
0,2l
d2),
Res
idue
s1–
64(3
io4,
3io6
,3io
r,3i
ot,3
iou,
3iov
,3io
w)
Den
tato
rubr
al-
palli
douy
sian
atro
phy
(DR
PLA
)
CA
GPr
otei
nco
ding
regi
on(p
olyQ
)
ATN
1at
roph
in1
(P54
259,
1190
resi
dues
)
Tran
scrip
tion,
tran
scrip
tion
regu
latio
n
7–34
49–8
8N
ost
ruct
ural
info
rmat
ion
Spi
noce
rebe
llar
atax
ia1
(SC
A1)
CA
GPr
otei
nco
ding
regi
on(p
olyQ
)
ATX
N1
atax
in1
(P54
253,
815
resi
dues
)
Tran
scrip
tion,
tran
scrip
tion
regu
latio
n
6–39
40–8
2R
esid
ues
563–
693
(1oa
8)
Spi
noce
rebe
llar
atax
ia2
(SC
A2)
CA
GPr
otei
nco
ding
regi
on(p
olyQ
)
ATX
N2
atax
in2
(Q99
700,
1313
resi
dues
)
No
asso
ciat
edG
O
keyw
ords
for
biol
ogic
al
proc
ess
15–2
432
–200
Res
idue
s91
2–92
8(3
ktr)
Spi
noce
rebe
llar
atax
ia3
(SC
A3)
CA
GPr
otei
nco
ding
regi
on(p
olyQ
)
ATX
N3/
MJD
atax
in3
(P54
252,
364
resi
dues
)
Tran
scrip
tion,
tran
scrip
tion
regu
latio
n,U
blco
njug
atio
n
path
way
10–5
155
–87
Res
idue
s1–
182
(1yz
b),R
esid
ues
222–
263
(2kl
z)
Spi
noce
rebe
llar
atax
ia6
(SC
A6)
CA
GPr
otei
nco
ding
regi
on(p
olyQ
)
CA
CN
A1
AC
AC
NA
1 A,P
/Q-t
ype
α1A
calc
ium
chan
nels
ubun
it
(O00
555,
2505
resi
dues
)
Cal
cium
tran
spor
t,io
n
tran
spor
t,tr
ansp
ort
4–20
20–2
9R
esid
ues
1955
–197
5(3
bxk)
Spi
noce
rebe
llar
atax
ia7
(SC
A7)
CA
GPr
otei
nco
ding
regi
on(p
olyQ
)
ATX
N7
atax
in7
(O15
265,
892
resi
dues
)
Tran
scrip
tion,
tran
scrip
tion
regu
latio
n
4–35
37–3
06R
esid
ues
330–
401
(2kk
r)
Spi
noce
rebe
llar
atax
ia17
(SC
A17
)
CA
GPr
otei
nco
ding
regi
on(p
olyQ
)
ATX
N17
TATA
box
bind
ing
prot
ein
(TB
P)(
P20
226,
339
resi
dues
)
Tran
scrip
tion,
tran
scrip
tion
regu
latio
n,H
ost-
viru
s
inte
ract
ion
25–4
247
–63
Res
idue
s15
9–33
7(1
cdw
,1c9
b,1j
fi,
1nvp
,1tg
h)
Mul
tiple
skel
etal
dysp
lasi
as(C
OM
P)
GA
CPr
otei
nco
ding
regi
on
(pol
yasp
arta
te)
CO
MP
cart
ilage
olig
omer
icm
atrix
prot
ein
(a.k
.a
Thro
mbo
spon
din-
5)
(P49
747,
757
resi
dues
)
Apo
ptos
is,c
ella
dhes
ion
54,
6,7
Res
idue
s22
5–75
7(3
fby)
.
Synp
olyd
acty
ly
(HO
XD
13)
GC
GPr
otei
nco
ding
regi
on(p
olyA
)
HO
XD
13ho
meo
box
D13
(P35
453,
343
resi
dues
)
Tran
scrip
tion,
tran
scrip
tion
regu
latio
n
1522
–29
No
stru
ctur
alin
form
atio
n
(Con
tinue
d)
Frontiers in Neurology | Neurodegeneration June 2013 | Volume 4 | Article 76 | 2
http://www.frontiersin.org/Neurodegenerationhttp://www.frontiersin.org/Neurodegeneration/archive
Almeida et al. Structure and function of trinucleotide repeats
Tab
le1
|Co
nti
nu
ed
Dis
ease
nam
eR
epea
t
typ
e
Rep
eat
loca
tio
n
Gen
eP
rote
in(U
niP
rot
iden
tifi
er,n
um
ber
of
resi
du
es)
Bio
log
ical
pro
cess
a
No
rmal
rep
eat
len
gth
Dis
ease
rep
eat
len
gth
Pro
tein
stru
ctu
red
eter
min
ed?
Ocu
loph
aryn
geal
Mus
cula
rD
ystr
ophy
(OP
MD
)
GC
GPr
otei
nco
ding
regi
on(p
olyA
)
PAB
PN
1Po
lyad
enyl
ate-
bind
ing
prot
ein
2(Q
86U
42,3
06
resi
dues
)
mR
NA
proc
essi
ng10
12–1
7R
esid
ues
167–
254
(3b4
d,3b
4m,
3ucg
)
Cle
idoc
rani
al
dysp
lasi
a(C
BFA
1)
GC
GPr
otei
nco
ding
regi
on(p
olyA
)
RU
NX
2R
unt-
rela
ted
tran
scrip
tion
fact
or2
(Q13
950,
521
resi
dues
)
Tran
scrip
tion;
tran
scrip
tion
regu
latio
n
1727
No
stru
ctur
alin
form
atio
n
Hol
opro
senc
epha
ly
(ZIC
2)
GC
GPr
otei
nco
ding
regi
on(p
olyA
)
ZIC
2Zi
nc-fi
nger
prot
ein
ZIC
2
(O95
409,
532
resi
dues
)
Diff
eren
tiatio
n,
neur
ogen
esis
,
tran
scrip
tion,
tran
scrip
tion
regu
latio
n
1525
No
stru
ctur
alin
form
atio
n
Han
d-Fo
ot-G
enita
l
Synd
rom
e/H
OX
A13
)
GC
GPr
otei
nco
ding
regi
on(p
olyA
)
HO
XA
13ho
meo
box
A13
(P31
271,
388
resi
dues
)
Tran
scrip
tion,
tran
scrip
tion
regu
latio
n
1824
–26
No
stru
ctur
alin
form
atio
n
Ble
phar
ophi
mos
is/
ptos
is/e
pica
nthu
s
inve
rsus
synd
rom
e
type
II(F
OX
L2)
GC
GPr
otei
nco
ding
regi
on(p
olyA
)
FOX
L2Fo
rkhe
adbo
xlik
e2
(P58
012,
376
resi
dues
)
Diff
eren
tiatio
n,
tran
scrip
tion,
tran
scrip
tion
regu
latio
n
1422
–24
Res
idue
s32
2–32
8(2
l7z)
Infa
ntile
spas
m
synd
rom
e(A
RX
)
GC
GPr
otei
nco
ding
regi
on(p
olyA
)
AR
XA
rista
less
-rel
ated
hom
eobo
x(Q
96Q
S3,
562
resi
dues
)
Diff
eren
tiatio
n,
neur
ogen
esis
,
tran
scrip
tion,
tran
scrip
tion
regu
latio
n
10–1
617
–23
No
stru
ctur
alin
form
atio
n
Myo
toni
cdy
stro
phy
type
1(D
M1)
CTG
3′U
TRD
MP
KM
yoto
nic
dyst
roph
y
prot
ein
kina
se(D
MP
K)
(Q09
013,
639
resi
dues
)
No
asso
ciat
edG
O
keyw
ords
for
biol
ogic
al
proc
ess
5–37
90–6
500
Res
idue
s11
–420
(2vd
5),R
esid
ues
460–
537
(1w
t6)
Frie
drei
chat
axia
(FR
DA
)
GA
AIn
tron
FXN
Frat
axin
(Q16
595,
210
resi
dues
)
Hem
ebi
osyn
thes
is,I
on
tran
spor
t,Ir
onst
orag
e,
Iron
,tra
nspo
rt
6–32
>20
0R
esid
ues
88–2
10(1
ekg)
,Res
idue
s
91–2
10(1
ly7)
,Res
idue
s82
–210
(3s4
m,3
s5d,
3s5e
,3s5
f,3t
3j,3
t3k,
3t3l
,3t3
t,3t
3x)
Spi
noce
rebe
llar
atax
ia8
(SC
A8)
CTG
3′U
TRAT
XN
8A
taxi
n-8
(a.k
.apr
otei
n1C
2;
(Pre
sent
inS
CA
8-sp
ecifi
c
1C2-
posi
tive
intr
anuc
lear
incl
usio
ns)(
Q15
6A1,
80
resi
dues
)
Cel
ldea
th2–
130
>11
0N
ostr
uctu
rali
nfor
mat
ion
(Con
tinue
d)
www.frontiersin.org June 2013 | Volume 4 | Article 76 | 3
http://www.frontiersin.orghttp://www.frontiersin.org/Neurodegeneration/archive
Almeida et al. Structure and function of trinucleotide repeats
Tab
le1
|Co
nti
nu
ed
Dis
ease
nam
eR
epea
t
typ
e
Rep
eat
loca
tio
n
Gen
eP
rote
in(U
niP
rot
iden
tifi
er,n
um
ber
of
resi
du
es)
Bio
log
ical
pro
cess
a
No
rmal
rep
eat
len
gth
Dis
ease
rep
eat
len
gth
Pro
tein
stru
ctu
red
eter
min
ed?
Spi
noce
rebe
llar
atax
ia12
(SC
A12
)
CA
G5′
UTR
PP
P2R
2BS
erin
e/th
reon
ine-
prot
ein
phos
phat
ase
2A55
kDa
regu
lato
rysu
buni
tB
β
isof
orm
(Q00
005,
443
resi
dues
)
Apo
ptos
is7–
4555
–78
No
stru
ctur
alin
form
atio
n
Hun
tingt
on
dise
ase-
like
2(H
DL2
)
CA
GA
ltern
ativ
e
splic
eis
ofor
m
2–
poly
A-
expa
nsio
n
JPH
3Ju
ncto
phili
n3
(Q8W
XH
2,
748
resi
dues
)
No
asso
ciat
edG
O
keyw
ords
for
biol
ogic
al
proc
ess
6–27
51–5
7N
ost
ruct
ural
info
rmat
ion
FRA
XA
:fra
gile
X
synd
rom
e
CG
G5′
UTR
FMR
1Fr
agile
Xm
enta
l
reta
rdat
ion
1pr
otei
n
(Q06
787,
632
resi
dues
).
Tran
spor
t;m
RN
Atr
ansp
ort
6–52
230–
2000
Res
idue
s1–
134
(2bk
d),R
esid
ues
216–
280
(2fm
r),R
esid
ues
216–
425
(2qn
d),R
esid
ues
527–
541
(2la
5)
FXTA
S:f
ragi
leX
trem
or/a
taxi
a
synd
rom
e
CG
G5′
UTR
FMR
1Fr
agile
Xm
enta
l
reta
rdat
ion
1pr
otei
n
(Q06
787,
632
resi
dues
).
Tran
spor
t;m
RN
Atr
ansp
ort
6–52
59–2
30R
esid
ues
1–13
4(2
bkd)
,Res
idue
s
216–
280
(2fm
r),R
esid
ues
216–
425
(2qn
d),R
esid
ues
527–
541
(2la
5)
FRA
XE
:fra
gile
X
synd
rom
e
CG
G5′
UTR
FMR
2Fr
agile
Xm
enta
l
reta
rdat
ion
2pr
otei
n
(P51
816,
1311
resi
dues
)
mR
NA
proc
essi
ng,m
RN
A
splic
ing
4–39
200–
900
No
stru
ctur
alin
form
atio
n
UTR
,unt
rans
late
dre
gion
.aB
iolo
gica
lFun
ctio
nba
sed
onG
ene
Ont
olog
yas
anno
tate
din
Uni
Prot
.
Frontiers in Neurology | Neurodegeneration June 2013 | Volume 4 | Article 76 | 4
http://www.frontiersin.org/Neurodegenerationhttp://www.frontiersin.org/Neurodegeneration/archive
Almeida et al. Structure and function of trinucleotide repeats
FIGURE 1 | Structural variability of proteins encoded byTNR-containinggenes. Illustrative domain graphics of the multi-domain structure of proteinsassociated with polyQ-expansion diseases. All proteins shown arereferenced by their name as annotated in UniProt. The protein domains forwhich information is annotated in the Pfam database are shown as coloredboxes with Pfam family accession code referenced above the domain box.Complete names of domains can be assessed by searching the specific
Pfam accession code at http://pfam.sanger.ac.uk/. Numbers below thedomain schemes represent amino acid residue numbers. Regionscontaining the amino acid repeats and with a prediction for formation ofcoiled-coils (as annotated in UniProt) are shown as well as regions withknown 3D structure (boxed in red, with PDB accession codes shown).Notice the predominant location of the repeat regions within the N-terminalregions of the proteins.
with tandem spaced MBNL1 binding sites were MBNL1 oligomerswith a ring-like structure can assemble, possibly leading to the for-mation of the ribonuclear foci identified in cell models of theseTNR diseases (Yuan et al., 2007; de Mezer et al., 2011). The struc-ture and stability of the TNR hairpin structures formed depends onthe presence of interruptions as well as on the nature of the flank-ing regions. This might be related with the ability of individual
repeats to participate in the RNA toxicity mechanisms (Krzyzosiaket al., 2012).
In FRDA and FXTAS, pathogenesis results predominantlyfrom decreased expression of the associated genes (FXN andFMR1/FMR2) caused by the expansion of GAA and CGG repeats,respectively, which results in loss of function of key proteinsinvolved in iron-sulfur cluster biogenesis and mRNA translation
www.frontiersin.org June 2013 | Volume 4 | Article 76 | 5
http://pfam.sanger.ac.uk/http://www.frontiersin.orghttp://www.frontiersin.org/Neurodegeneration/archive
Almeida et al. Structure and function of trinucleotide repeats
FIGURE 2 | Structural variability of proteins encoded byTNR-containing genes. Illustrative domain graphics of the multi-domainstructure of proteins associated with polyD- and polyA-expansion diseases.All proteins shown are referenced by their name as annotated in UniProt.The protein domains for which information is annotated in the Pfamdatabase are shown as colored boxes with Pfam family accession codereferenced above the domain box. Complete names of domains can beassessed by searching the specific Pfam accession code athttp://pfam.sanger.ac.uk/. Numbers below the domain schemes representamino acid residue numbers. Regions containing the amino acid repeatsand with a prediction for formation of coiled-coils (as annotated in UniProt)are shown as well as regions with known 3D structure (boxed in red, withPDB accession codes shown). Notice the predominant location of therepeat regions within the N-terminal regions of the proteins.
at synapses. Nevertheless, in FXTAS RNA toxicity is also proposedto play a role in pathogenesis (Li and Bonini, 2010). The recentlydiscovered mechanisms of pathogenesis in spinocerebellar ataxiatype 8 (SCA8) uncovered the extreme complexity of TNR disor-ders. In fact, SCA8 is caused by expansion of CTG/CAG repeatsin the affected gene, which are transcribed bi-directionally leading
to the generation of expanded CUG and CAG-containing tran-scripts further translated into homopolymeric proteins, so thatpathogenesis can be mediated by both RNA and protein toxicity(Merienne and Trottier, 2009). Curiously, recent data have high-lighted the possibility of non-ATG translation across expandedTNR in all possible reading frames, which might further con-tribute to the generation of novel toxic proteins and RNAs addingto the multi-parametric character of the pathogenic mechanismsassociated with TNR diseases (Li and Bonini, 2010; Pearson, 2011;Sicot et al., 2011).
TRINUCLEOTIDE REPEATS WITHIN PROTEIN CODING REGIONSOver 20 years ago, the finding that the expansion of CAG repeatswithin the coding sequence of the androgen receptor gene was thegenetic basis of SBMA (La Spada et al., 1991) represented a hall-mark in the discovery of these novel dynamic mutations and theirassociation with human disease. Some years later, the identifica-tion of intracellular inclusions containing the expanded proteins(Paulson et al., 1997) provided a clue to pathogenesis, directingresearch in the field into an extensive search for the mechanismsof polyQ-induced protein aggregation. The moderate expansionof GCG and CAG repeats, which are translated into polyA andpolyQ tracts in the affected proteins (Figures 1 and 2), results inprotein misfolding and aggregation, in accordance with a general,although not always unique, toxic gain of function mechanismof pathogenesis (Williams and Paulson, 2008). The appearanceof insoluble cytoplasmic or nuclear inclusions enriched in theexpanded polyA- or polyQ-containing protein constitutes a char-acteristic fingerprint of these diseases (Messaed and Rouleau, 2009;Orr, 2012a), regardless of their controversial role in pathogenesis.While the proteins containing polyA repeats are predominantlytranscription factors with a role in development (see Table 1and Amiel et al., 2004; Messaed and Rouleau, 2009), most ofthe proteins linked to polyQ-expansion diseases are involved inDNA-dependent regulation of transcription or neurogenesis andoften contain multiple intermolecular partners (Butland et al.,2007). Despite the overall lack of sequence or structural homol-ogy, both polyQ- and polyA-repeat expansions are associated withformation of ß-rich amyloid-like protein inclusions, and with thewider group of protein misfolding disorders. These inclusions areenriched in ubiquitin, proteasome subunits, and chaperones, andoften recruit macromolecules that are part of the macromolecularinteraction networks associated with the proteins’ native functions(Williams and Paulson, 2008). As an example, the poly(A)-bindingprotein PABNP1 forms insoluble inclusions upon alanine expan-sion, co-aggregating together with poly(A)-mRNA, proteasomesubunits, ubiquitin, heat-shock proteins, and SKIP, a transcrip-tion factor associated with muscle-specific gene expression (Brais,2003; Tavanez et al., 2009; Winter et al., 2013).
The simplistic view of the predominant role of the inclusionsin polyQ-induced pathogenesis was later challenged by the failureof this mechanism to explain the cell-specific vulnerability char-acteristic for each disease and by the identification of numerousexamples of neuronal toxicity in the absence of visible intracellu-lar inclusions (Arrasate et al., 2004). Indeed, the inclusions wereshown to be fibrillar and display amyloid-like properties bothin vivo and in vitro (Huang et al., 1998; Bevivino and Loll, 2001;
Frontiers in Neurology | Neurodegeneration June 2013 | Volume 4 | Article 76 | 6
http://pfam.sanger.ac.uk/http://www.frontiersin.org/Neurodegenerationhttp://www.frontiersin.org/Neurodegeneration/archive
Almeida et al. Structure and function of trinucleotide repeats
Sathasivam et al., 2010) and, in a mechanistic parallel with thepathogenic mechanisms proposed for “classical” amyloids, manystudies suggested that the insoluble inclusions played a protec-tive role, sequestering toxic, and misfolded protein conformers(Arrasate et al., 2004; Rub et al., 2006; Miller et al., 2010). Indeed,soluble intermediates in the aggregation pathway such as mis-folded β-sheet rich polyQ protein monomers and oligomers havelatter been identified and proposed to represent the major toxicspecies (Kayed et al., 2003; Gales et al., 2005; Nagai et al., 2007;Miller et al., 2011). Also, in OPMD, the primary toxic species areproposed to be the soluble variants of the expanded polyA-repeatprotein PABPN1 (Messaed et al., 2007). It is currently accepted thatin polyQ disorders the expanded region plays a role in inducingthe self-assembly of the carrier protein, which engages in patho-genic interactions and leads to the formation of toxic monomersor oligomers (Takahashi et al., 2008; Weiss et al., 2008) latterconverted to insoluble intracellular amyloid-like oligomers whereboth expanded and “normal” protein are sequestered along withother macromolecular partners (reviewed in Williams and Paul-son, 2008; Matos et al., 2011; Costa and Paulson, 2012). As morebiochemical data is gathered, more is understood about the role ofamino acid expansions in modulating the interaction with macro-molecular partners. As an example, expansion of the polyA tract inPABPN1 results in increased association with Hsp70 chaperonesand type I arginine methyl transferases (Tavanez et al., 2009). Thisindicates that the distinct neuropathological features arising fromthis amino acid-repeat expansion might at least partially resultfrom alterations on the native biological functions and macro-molecular interactions of the carrier protein, which might vary indifferent intracellular environments.
Recent data have shown that expansion of polyA repeats isfrequently associated with loss of normal function altering a mul-titude of cellular pathways with consequences in cell functionality(Amiel et al., 2004; Messaed and Rouleau, 2009), although proteinaggregation might also play a dominant role in some of the polyA-associated disorders (Messaed and Rouleau, 2009; Winter et al.,2013). Studies with polyQ proteins have shown that pathogene-sis might result from a subtle imbalance in the association of themutant protein with multiple cellular partners and that toxicityand neuronal death could result from a combination of proteinself-assembly and functional alterations (Friedman et al., 2007; Liet al., 2007b; Lim et al., 2008; Kratter and Finkbeiner, 2010; Orr,2012b; Pastore and Temussi, 2012). In fact, neuronal death as aresult of polyQ-expansion seems to resemble that of linker cell inC. elegans (Pilar and Landmesser, 1976; Chu-Wang and Oppen-heim, 1978; Blum et al., 2012, 2013) which involves the polyQprotein pqn-4, pointing for a common mechanism for linker celldeath, and neuronal death in polyQ diseases (Blum et al., 2013).
Polyglutamine diseases constitute a representative and largelystudied group of neurodegenerative disorders where considerableamounts of data have been collected on the role of expandedpolyQ for disease pathogenesis. However, given the proposed func-tion of polyQ regions in mediating protein–protein interactions,which might be modulated by polyQ-expansion (Schaefer et al.,2012), the information on the role of these regions for native pro-tein function, structure, and dynamics is still limited. Structuraland functional information on the role of these repeat sequences
in protein function is crucial to better understand how expan-sion affects selected neuronal subpopulations. Below, we brieflydiscuss the current knowledge on the function and structure ofpolyQ repeats and their role on macromolecular interactions, andfinally focus on the known structural and functional informationon ataxin-3, the protein whose mutation causes MJD.
FUNCTION OF PolyQ ON PROTEIN–PROTEIN INTERACTIONSAND EVOLUTIONUntil recently, the function of many amino acid-repeat-containingproteins and the role of homopeptide regions were somewhatobscure. However, several global analysis studies on single aminoacid-repeat-containing proteins shed light onto their function andonto the biological significance of the repeated region, in particu-lar of polyQ, the most prevalent amino acid repetition in humans(Alba and Guigo, 2004). It is now accepted that TNR, particu-larly those located within protein-coding regions, are consideredimportant mutators providing the genetic variability required fordriving evolution (King, 1994; Kashi et al., 1997; Kashi and King,2006; Nithianantharajah and Hannan, 2007). In fact, simple orlow-complexity amino acid-repeats are rare within prokaryoticbut extremely abundant within eukaryotic proteins, particularlyover-represented in Plasmodium (49–90% of the total proteome),D. discoideum (52%), D. melanogaster (20%), C. elegans (9%),and H. sapiens (14%) (Haerty and Golding, 2010). Among allhomopolymeric repeats, the most common on eukaryotic pro-teins are glutamine, asparagine, alanine, and glutamate repeats(Faux et al., 2005). This seems to indicate that there has been astrong negative selection against the appearance of hydrophobicamino acid-repeats with high tendency to aggregate, such as poly-isoleucine, polyleucine, polyphenylalanine, and polyvaline (Omaet al., 2005, 2007).
The homopeptide regions seem to be particularly relevant forbrain development and function, since these repeated regions canbe found in various neurodevelopmental genes (Nithiananthara-jah and Hannan, 2007). Indeed, the sexual behavior of prairievoles (Hammock andYoung, 2005), as well as human pair-bonding(Walum et al., 2008), seems to be dependent on the repeat lengthin the vasopressin 1A receptor gene. A wide study of the distribu-tion and function of homopeptide-containing proteins could alsodemonstrate a clear trend in humans, D. melanogater, and C. ele-gans, with the majority of homopeptide-containing proteins per-forming roles in transcription/translation and signaling processesand to a less extend in transport and adhesion processes (Fauxet al., 2005). A similar profile was also found in a comparativeanalysis of proteins with amino acid-repeats in human and rodents(Alba and Guigo, 2004) and also on a comparative genomic studyin domestic dogs, which unveiled an association between mor-phological variations and the length of the repeated region in thetranscription factor-encoded genes ALX4 and RUNX2 (Fondonand Garner,2004). Analysis of the human genome also revealed theexistence of 64 CAG repeat-containing genes involved in biologicalprocesses such as regulation of transcription, binding of transcrip-tional co-activators and transcription factors, and in neurogenesisin general (Butland et al., 2007). Additionally, a detailed analy-sis of the human polyQ database (http://pxgrid.med.monash.edu.au/polyq/) (Robertson et al., 2011) also indicated that the
www.frontiersin.org June 2013 | Volume 4 | Article 76 | 7
http://pxgrid.med.monash.edu.au/polyq/http://pxgrid.med.monash.edu.au/polyq/http://www.frontiersin.orghttp://www.frontiersin.org/Neurodegeneration/archive
Almeida et al. Structure and function of trinucleotide repeats
majority of polyQ-containing proteins display domains involvedin development (Homeobox domain-containing proteins, Fibrob-last growth factor receptor), chromatin remodeling (Bromod-omain and PHD-containing proteins), and signal transduction(PDZ domain-containing proteins), all biological processes thatare highly dependent on protein–protein interactions and associ-ated with the formation of multicomponent protein complexes. Asfor humans, analysis of bovine polyQ proteins revealed an enrich-ment for large multi-domain transcriptional regulators (Whanet al., 2010).
It is currently accepted that the majority of repeat-containingproteins perform roles in processes that require the assembly oflarge multiprotein or protein/nucleic acid complexes (Faux et al.,2005; Hancock and Simon, 2005; Whan et al., 2010). Supportingthis notion is the fact that homopolymeric amino acid-repeats areconsidered to be unstructured (Gojobori and Ueda, 2011) andthat intrinsically unstructured regions are suggested to consti-tute macromolecular docking sites, which become structured onlywhen bound to cognate ligand partners (Huntley and Golding,2002; Simon and Hancock, 2009). In fact, “hub proteins” con-tain significantly longer and more frequent repeats or disorderedregions, which facilitate binding to multiple partners (Dosztanyiet al., 2006). Recently, Fiumara et al. (2010) found an overrep-resentation of coiled-coils domains in polyQ-containing proteinsand in their interaction partners, which are able to form α-helicalsupersecondary structures, often inducing protein oligomeriza-tion (Parry et al., 2008). Thus, polyQ tracts due to their intrinsicstructural flexibility, which is largely influenced by the flankingresidues (see PolyQ:A Simple Sequence Repeat with a PolymorphicStructure below), may act as stabilizers of intra- and intermole-cular protein interactions, possibly by extending a neighboringcoiled-coil region to promote its interaction with a coiled-coilregion in an interacting protein partner (Schaefer et al., 2012).A detailed analysis revealed heptad repeats typical of coiled-coilsin regions flanking or overlapping polyQ stretches, whose disrup-tion is sufficient to impair CHIP-huntingtin interaction, indicatingthat coiled-coils are crucial for polyQ-mediated protein contacts.Importantly, coiled-coils also seem to be important for the regula-tion of aggregation and insolubility of polyQ-containing proteins(see below and Fiumara et al., 2010) as recently proposed byPetrakis et al. (2012), which discovered a recurrent presence ofcoiled-coil domains in ataxin-1 misfolding enhancers, while suchdomains were not present in suppressors.
Based on the several observations on the function of polyQ-containing proteins it is suggested that a general function of polyQ,as for the majority of repeat sequences, is to aid in the assem-bly of macromolecular complexes, either through tethered distantdomains or through interactions with the polyQ itself (Gerberet al., 1994; Korschen et al., 1999; Faux et al., 2005). By affectingprotein interactions, and being present in particular functionalclasses such as transcription factors, polyQ is considered central tothe evolution of this type of proteins and consequently crucial tothe evolution of cellular signaling pathways (Hancock and Simon,2005).
A structural analysis of polyQ repeats and its flanking domainsas well as its role in protein aggregation will be discussed in greaterdetail in the next sections.
STRUCTURAL STUDIES ON PolyQ REPEATSSince the discovery that polyQ repeats are associated with humanneurodegenerative diseases that a huge effort has been made todetermine the structure of polyQ and to understand how expan-sion of the repeat affects the structure of the carrier protein and/orthe normal interaction with molecular partners. The first evidencefrom the aggregation-prone character of polyQ-rich proteins camefrom studies with glutamine-rich cereal storage proteins and syn-thetic glutamine polypeptides (Beckwith et al., 1965; Krull et al.,1965). After the discovery that a number of neurological disor-ders were triggered by expansion of a polyQ tract in different andunrelated proteins (La Spada et al., 1994), and before intracellularinclusions enriched in the polyQ-expanded protein were identi-fied as a major fingerprint in these diseases (Davies et al., 1997;Paulson et al., 1997), Perutz (1994) anticipated that the expandedpolyQ tract could mediate protein–protein interactions causingprotein aggregation in neurons and recruiting other polyQ-richproteins such as transcription factors leading to cellular dysfunc-tion. Below, the structural features and self-assembly properties ofpolyQ sequences are briefly discussed (for a detailed review on thebiophysical and structural features of polyQ, see Wetzel, 2012).
PolyQ: A SIMPLE SEQUENCE REPEAT WITH A POLYMORPHICSTRUCTUREIn order to elucidate the structure of the glutamine repeat andto uncover the structural changes induced by polyQ expansion,several strategies have been put forward including (a) the struc-tural analysis of polyQ-containing peptides of different lengths,(b) the characterization of proteins of well-known structure afterinsertion of an exogenous polyQ repeat, and structural determina-tion of (c) polyQ-antibody complexes, or (d) natural polyQ-richproteins.
Using synthetic peptides containing 15 glutamine repeats,Perutz and coworkers proposed that polyQ stretches could self-associate forming hydrogen bonds between their side-chain amidegroups and the main chain of a neighboring β-strand, to formcross-β structures (polar zippers) (Perutz, 1994). This study wasfollowed by many reports where synthetic polyQ peptides wereused as models of the biophysical properties of polyQ-rich pro-teins, which established that polyQ-containing peptides have atendency toward self-assembly into amyloid-like structures (Chenet al., 2002a). Moreover, the results obtained in vitro reflected dis-ease features observed in vivo such as the correlation betweenlarger polyQ size, increased protein aggregation, and earlier diseaseonset (Chen et al., 2002b; Kar et al., 2011). Circular dichroism stud-ies of polyQ peptides in solution have shown that their monomericforms lack regular secondary structure (Altschuler et al., 1997;Klein et al., 2007) and additional biophysical experiments pro-posed that these peptides can adopt collapsed (Crick et al., 2006;Dougan et al., 2009; Peters-Libeu et al., 2012) or extended (Singhand Lapidus, 2008) coils in solution whose compactness wasstrongly correlated with the polyQ size (Walters and Murphy,2009). The determination of the structure of monomeric polyQpeptides with atomic detail is however still lacking as a result oftheir intrinsic conformational flexibility and tendency to aggregateinto heterogeneously sized β-rich oligomers. From the combina-tion of experimental and theoretical methods a picture for polyQ
Frontiers in Neurology | Neurodegeneration June 2013 | Volume 4 | Article 76 | 8
http://www.frontiersin.org/Neurodegenerationhttp://www.frontiersin.org/Neurodegeneration/archive
Almeida et al. Structure and function of trinucleotide repeats
structure and aggregation is emerging, where the monomericpolyQ adopt an ensemble of conformations lacking regular sec-ondary structures that assemble into β-structures in a polyQ-length dependent fashion (Vitalis et al., 2009; Walters and Murphy,2009, 2011; Williamson et al., 2010; Kar et al., 2011). Divergentresults proposing the existence of predominantly extended or col-lapsed conformations or the minimum size for polyQ aggregationare likely due to the differences in the introduction of variableflanking residues (Kar et al., 2011). They might result from theinsertion of different polyQ tract interrupting residues (Waltersand Murphy, 2011), or be a consequence of the protocols used forthe preparation and disaggregation of the peptides used for thebiophysical studies (Jayaraman et al., 2011). Most results obtainedwith these peptides do not generally take into account the pos-sible effects of the protein context on the structural propertiesof the polyQ stretches, a particularly relevant feature consideringthat the role of non-polyQ domains in protein aggregation hasbeen reported for ataxin-1 (de Chiara et al., 2005), ataxin-3 (Galeset al., 2005), and huntingtin (Tam et al., 2009; Thakur et al., 2009;Liebman and Meredith, 2010).
In a pioneer work, Stott et al. (1995) inserted a G-Q10-G peptide into the inhibitory loop of chymotrypsin inhibitor2 (CI2), a soluble small protein from barley seeds, showingthat this CI2-polyQ chimera has an increased tendency for self-assembly. Even though a CI2 variant with four glutamines crys-tallized, the structure of the CI2-Q4 dimer showed that thepolyQ region was disordered and that oligomerization was medi-ated by domain swapping (Figure 3A) and not by direct polyQassociation (Chen et al., 1999). A structure resembling the pro-posed polar zipper was later observed between two asparaginesin the hinge loop of the major domain swapped dimer ofbovine pancreatic ribonuclease A (Liu et al., 2001) (Figure 3B).Insertion of a 10 glutamine repeat within this hinge loop ofribonuclease A, resulted in domain swapping, oligomerization,and amyloid-like fiber formation, but strikingly the enzymewithin the fibers was catalytically active, retaining its nativefold (Sambashivan et al., 2005). However, although the struc-ture of the domain swapped dimer was solved by X-ray crys-tallography, the repeat region was not visible in the electrondensity maps.
FIGURE 3 | Structure of proteins/protein domains containing polyQregions. (A) Cartoon representation of the domain swapped dimer ofchymotrypsin inhibitor 2 with a 4 glutamine insertion [(Chen et al., 1999); PDBaccession code 1cq4], dotted lines represent the polyQ linker not visible inthe X-ray crystal structure. (B) Cartoon representation of domain swappedmajor dimers of ribonuclease A. Inset shows a short segment resembling thepolar zipper formed by asparagine residues in the linker region [(Liu et al.,2001); PDB accession code 1f0v]. (C) Surface representation Fv fragment of amonoclonal antibody in complex with a polyQ peptide shown as sticks [(Li
et al., 2007a), PDB accession code 2otu]. (D) Cartoon representation of theglutamine-rich domain from HDAC4 showing details of the polar interactions(dotted lines) at the oligomer interfaces involving glutamine residues [(Guoet al., 2007), PDB accession code 2o94]. (E) Cartoon representation of thecrystal structures of huntingtin exon-1 fragments observed in different crystalforms, highlighting the different orientations of the C-terminal polyQ residuesshown as sticks. The 17 glutamine stretch adopts variable conformations inthe structures: α helix, random coil, and extended loop. [(Kim et al., 2009),PDB accession codes 3io4, 3iow, 3iov, 3iou, 3iot, 3ior, 3io6].
www.frontiersin.org June 2013 | Volume 4 | Article 76 | 9
http://www.frontiersin.orghttp://www.frontiersin.org/Neurodegeneration/archive
Almeida et al. Structure and function of trinucleotide repeats
A first overview of a short polyQ stretch at atomic resolu-tion resulted from the structure of a polyQ10 peptide (GQ10G)(Figure 3C) bound to MW1, an antibody against polyQ. Thisstructure reveals that polyQ adopts an extended, coil-like struc-ture in which contacts are made between side chains and/or mainchain atoms of all 10 glutamines and the antibody-combining site(Li et al., 2007a). The peculiar structural features of these repeat-containing regions were also revealed by the crystallographicstructure of a glutamine-rich domain of human histone deacety-lase (HDAC4), that folds into a tetramer-forming straight α-helix(Figure 3D). The protein interfaces consist of multiple hydropho-bic patches separated by polar interaction networks, in whichclusters of glutamines engage in extensive intra- and interheli-cal interactions (Guo et al., 2007). Further details on the structureof polyQ were unveiled by the high-resolution crystal structuresof huntingtin (HD) exon 1, containing 17 glutamines (Htt17Q)(Kim et al., 2009). Htt17Q in fusion with maltose-binding pro-tein (MBP) folds into an amino-terminal α-helix followed by apolyQ17 region that adopts multiple conformations in the differ-ent crystal forms, including α-helix, random coil, and extendedloop, and a polyproline helix formed by the polyP11 and mixedP/Q regions (Figure 3E). The authors suggested that the shallowequilibrium between α-helical, random coil, and extended confor-mations can be subtly altered by the size of polyQ sequence, theneighboring protein context, protein interactions, or by changesin cellular environment, and that this polymorphic behavior isa common characteristic of many amyloidogenic proteins (Kimet al., 2009).
SELF-ASSEMBLY AND AGGREGATION OF PolyQ REPEATSThe first approaches to characterize polyQ-induced protein aggre-gation and pathogenesis in the context of a full-length proteinincluded the insertion of the polyQ peptides into well-known non-pathogenic protein carriers such as hypoxanthinephosphoribosyltransferase (HPRT), which resulted in a neurological phenotypemimicking that observed in mice expressing the mutant HD trun-cated protein (Ordway et al., 1997). In vitro studies aiming at bettercharacterizing the structure and function of polyQ repeats in thecontext of full-length soluble proteins, included the insertion ofectopic polyQ stretches into well-characterized and soluble pro-teins such as CI2 (Stott et al., 1995; Chen et al., 1999), myoglobin(Mb) (Tanaka et al., 2001; Tobelmann and Murphy, 2011), glu-tathione S transferase (GST) (Masino et al., 2002; Bulone et al.,2006) and the B domain from Staphylococcus aureus Protein A(SpA) (Saunders et al., 2011). Fusion of the polyQ sequenceswith stable and soluble proteins moderates the intrinsic polyQpeptide aggregation propensity, but induces the self-assembly ofcarrier proteins into fibrillar amyloid-like structures, a nucleation-dependent process whose kinetics is directly proportional to thesize of the inserted polyQ repeat. Likewise, polyQ peptides are ableto seed the aggregation of intracellular soluble polyQ-containingproteins when added to cell cultures, conferring a heritable pheno-type of self-sustaining seeding, resembling a prion-like mechanism(Ren et al., 2009), reviewed in Cushman et al. (2010).
The impact of the polyQ tract and its expansion on the per-turbation of the structure of flanking sequences and domains is
critically dependent on the location of the amino acid-repeats,revealing impressive location-dependent changes in structural sta-bility, and fibril morphology of the host proteins (Robertsonet al., 2008; Saunders et al., 2011; Tobelmann and Murphy, 2011).Curiously, the studies with these model proteins showed that sta-bility and structure of the carrier protein remained unalteredby polyQ expansion when the repeat was inserted at the N- orC-terminus of the structured domain (Robertson et al., 2008),mimicking the location of polyQ tracts in most disease-relatedproteins (Figure 1).
The role of the flanking regions in modulating protein fibrilformation in polyQ disease proteins is well supported by experi-mental data (de Chiara et al., 2005; Gales et al., 2005; Bhattacharyyaet al., 2006; Saunders and Bottomley,2009; Tam et al., 2009; Thakuret al., 2009; Liebman and Meredith, 2010), in agreement with theknowledge that different polyQ-containing proteins have a diversethreshold for aggregation. For example, addition of a polyprolineextension after the polyQ repeat slows down aggregation (Bhat-tacharyya et al., 2006), while protein domains outside the polyQtract [e.g., Josephin domain (JD) of ataxin-3 and AHX domainof ataxin-1] have been shown to contribute to protein aggre-gation (Masino et al., 2004; de Chiara et al., 2005; Gales et al.,2005; Ellisdon et al., 2006, 2007). The multitude of data on thepolyQ-induced aggregation of disease and non-disease-proteinshighlights the complex interplay between the polyQ region andthe adjacent protein domains. In light of the polymorphic natureof the polyQ and the modulation of its structural features bythe protein context, two general mechanisms have been proposedfor polyQ-mediated toxicity (Kim et al., 2009): (a) the expandedpolyQ stretch adopts a novel conformation that mediates toxicityor is the precursor to toxic species; (b) intra- or intermolecularprotein interactions mediated by expanded polyQ in the randomcoil conformation are sufficient to result in pathological effects. Inboth cases the affinity of the interactions involving the expandedpolyQ region could be higher with selected target proteins, lead-ing to a preference of the disease proteins for some of the proteinpartners, a fact that is in agreement with the hypothesis raisedby Zuchner and Brundin (2008), which postulate that resistanceto NMDA receptor-mediated excitotoxicity occurring in somemouse models for HD is a consequence of a differential bind-ing of partner proteins, in a polyQ tract size dependent manner, tothe proline-rich domain of huntingtin. In this context, differencesin molecular interactions occurring in a cell- and tissue-specificmanner would result in different toxicities according to particularcellular environments.
Given the above mentioned studies, it is nowadays clear that thepolyQ region influences aggregation of proteins, but this process ishighly dependent on the surrounding protein context. Therefore,even though the structural information on peptides and proteinswith polyQ expansions is a useful guideline for the investiga-tion of the pathogenic effects of polyQ expansion, each of theproteins involved in polyQ diseases shows distinctive characteris-tics, cellular roles, and structural properties causing difficulties inthe formulation of structural hypothesis that could explain howdifferent monomeric conformations of polyQ leads to variousaggregated species and how they contribute to neurotoxicity.
Frontiers in Neurology | Neurodegeneration June 2013 | Volume 4 | Article 76 | 10
http://www.frontiersin.org/Neurodegenerationhttp://www.frontiersin.org/Neurodegeneration/archive
Almeida et al. Structure and function of trinucleotide repeats
PolyQ REPEATS IN ATAXIN-3 FUNCTION AND DYSFUNCTIONMachado-Joseph disease is an inherited neurodegenerative disor-der of adult onset originally described in people of PortugueseAzorean descent but later shown to be the most common auto-somal dominant spinocerebellar ataxia worldwide. Clinically, it ischaracterized by ataxia, ophthalmoplegia, and pyramidal signs,associated in variable degree with dystonia, spasticity, periph-eral neuropathy, and amyotrophy (Coutinho and Andrade, 1978).Pathologically, the disorder is associated with degeneration ofthe deep nuclei of the cerebellum, pontine nuclei, subthalamicnuclei, substantia nigra, and spinocerebellar nuclei (Coutinhoet al., 1982; Rosenberg, 1992; Margolis and Ross, 2001). It is causedby an expansion of a repetitive CAG tract within the ATXN3 gene(Kawaguchi et al., 1994). While in the healthy population the num-ber of CAG repeats ranges between 10 and 51, in MJD patients thelength of ataxin-3 polyQ tract exceeds 55 consecutive residues.Ataxin-3 is a modular protein, located both in the nucleus and thecytoplasm (Perez et al., 1999; Antony et al., 2009; Macedo-Ribeiroet al., 2009), encompassing an N-terminal globular JD, with struc-tural similarity to cysteine proteases (Scheel et al., 2003; Albrechtet al., 2004), followed by an extended tail composed of two ubiq-uitin interaction motifs (UIMs), the expandable polyQ tract, anda C-terminal region (Matos et al., 2011). The C-terminal region ofataxin-3 may contain a third UIM, depending on the splice vari-ant (Goto et al., 1997), with the 3UIM isoform of ataxin-3 beingpredominantly found in the brain (Harris et al., 2010). Currently,the physiological function of ataxin-3, as well as the molecularmechanism by which expanded polyQ sequences causes selectiveneurodegeneration remain mostly unknown. However, since itis ubiquitously expressed and cell death is region specific, neu-rodegeneration is currently viewed as depending on sequence andstructural features outside the ataxin-3 polyQ tract [reviewed inMatos et al. (2011) and references therein].
ATAXIN-3 BIOLOGICAL ROLESATXN3 orthologs have been identified in eukaryotic organismsincluding protozoans, plants, fungi, and animals (Albrecht et al.,2004; Costa et al., 2004; Rodrigues et al., 2007). Several functionshave been ascertained to ataxin-3 based on studies with orthologs.Specifically, a role in cell structure and/or motility was proposedfor mouse ataxin-3 as it is highly abundant in all types of muscleand in ciliated epithelial cells (Costa et al., 2004). In fact, ataxin-3is able to interact with tubulin through its JD domain (Figure 4),with nM affinity (Mazzucchelli et al., 2009), which supports arole in cell structure. Interestingly, data on ataxin-3 C. elegansortholog not only reinforces a function in structure/motility andsignal transduction (Rodrigues et al., 2007), but also indicate afunction in development as absence of ATXN3 strongly modifiesexpression of several development-related genes. ATXN3 knock-out animals showed no obvious deleterious phenotype, probablydue to a putative redundant function between ataxin-3 and otherJD-encoding proteins, such as ataxin-3-like protein, Josephin 1 andJosephin 2, all containing a typical cysteine protease catalytic triad.However the studies with ATXN3 knock-out animals revealed anoverall increase in the levels of ubiquitinated proteins (Schmittet al., 2007) and signs of altered expression of core sets of genesassociated with the ubiquitin-proteasome and signal transduction
pathways (Rodrigues et al., 2007), pointing to a dual function ofataxin-3 in the ubiquitin-proteasome system and transcriptionalregulation (Matos et al., 2011; Orr, 2012a).
Ataxin-3 function as transcriptional regulatorThe putative role of ataxin-3 in transcriptional regulation isproposed to entail the modulation of histone acetylation anddeacetylation at selected promoters. Ataxin-3 interacts with themajor histone acetyltransferases cAMP-response-element bindingprotein (CREB)-binding protein (CBP), p300, and p300/CREB-binding protein-associated factor (KAT2B/PCAF, Figures 4 and5), and is proposed to inhibit transcription in specific promot-ers (e.g., MMP-2 promoter) either by blocking access to histoneacetylation sites or through recruitment of histone deacetylase 3(HDAC3) and nuclear receptor co-repressor (NCOR1; Figures 4and 5) (Li et al., 2002; Evert et al., 2006). Although, the interac-tion sites have not been mapped in detail for all these proteins,co-immunoprecipitation experiments showed that KAT2B/PCAF,p300, and CBP bind exclusively to the polyQ-containing C-terminal region of ataxin-3 (Figure 4), apparently in a polyQ-sizedependent manner (Li et al., 2002). Experimental evidence alsoindicates that ataxin-3 forms part of a CREB-containing complex,although no direct interaction has been observed between the twoproteins (Li et al., 2002). In contrast, the N-terminal region ofataxin-3 directly binds histones H3 and H4 (Table 2; Figure 4)(Li et al., 2002). Of note, p300 and CBP, as well as NCOR1,also encompass amino acid repetitions in its sequence. Interest-ingly, in huntingtin and in ataxin-1, polyQ interferes with CBP-activated gene transcription via interaction of their glutamine-rich domains (Shimohata et al., 2000; Nucifora et al., 2001) andmutant huntingtin targets specific components of the core tran-scriptional machinery, in a glutamine-tract length-sensitive man-ner (Zhai et al., 2005), pinpointing once again the role of theamino acid-repeat region in the establishment of protein–proteininteractions.
Ataxin-3 molecular function: ubiquitin hydrolaseA role for ataxin-3 in ubiquitin-dependent pathways was pro-posed by bioinformatic analysis (Scheel et al., 2003; Albrecht et al.,2004), and its ability to bind and cleave poly-ubiquitin chainsand polyubiquitinated proteins was later demonstrated experi-mentally (Burnett et al., 2003; Chai et al., 2004). Importantly,inhibition of ataxin-3 catalytic activity results in the increaseof polyubiquitinated proteins, resembling the effects of protea-some inhibition (Berke et al., 2005), indicating that ataxin-3 isinvolved with proteins targeted for proteasomal degradation. Thefunction of ataxin-3 in the ubiquitin-proteasome system was fur-ther supported by the identification of its association with theubiquitin-like domain of the human homologs of the yeast DNArepair protein Rad23, HHR23A, and HHR23B (Wang et al., 2000;Doss-Pepe et al., 2003; Nicastro et al., 2005, 2009), with valosin-containing protein (VCP)/p97 (Hirabayashi et al., 2001; Doss-Pepeet al., 2003; Boeddrich et al., 2006; Zhong and Pittman, 2006), andwith the ubiquitin ligase E4B (Matsumoto et al., 2004) (Figures 4and 5). Strikingly, the weak direct association between ataxin-3and E4B is strongly reinforced by the addition of VCP/p97, indicat-ing that these proteins form part of a higher order macromolecular
www.frontiersin.org June 2013 | Volume 4 | Article 76 | 11
http://www.frontiersin.orghttp://www.frontiersin.org/Neurodegeneration/archive
Almeida et al. Structure and function of trinucleotide repeats
FIGURE 4 | Overview of ataxin-3 structural information. Schematicillustration of ataxin-3 (isoform 2; a.k.a. 3UIM isoform) domain structurehighlighting the regions involved in protein–protein interactions. The solutionstructures of the Josephin domain (PDB accession code 1yzb) and UIMs1-2(PDB accession code 2klz) are shown colored from N-(blue) to C- terminus(red). JD-, UIM-, NLS-, and polyQ-mediated interactions are represented byblue, red, green, and purple arrows, respectively; blue arrows indicate thelocation of post-translational modification sites, resulting from the interactionand phosphorylation by CK2 and GSK3. Representative multi-subunitcomplexes where ataxin-3 participates are boxed (Li et al., 2002; Matsumoto
et al., 2004; Scaglione et al., 2011; Durcan et al., 2012). One of the mainquestions in the quest for ataxin-3 interacting proteins is whetherpolyQ-expansion of the disease-protein modulates the binding affinities.Current data indicates that polyQ-expansion increments the ataxin-3 affinityfor CHIP (Scaglione et al., 2011), VCP/p97 (Matsumoto et al., 2004; Boeddrichet al., 2006; Zhong and Pittman, 2006), and the transcription regulators p300,CBP, and PCAF (Li et al., 2002) (interactions represented by broken lines).Strikingly, all these interactions are mediated by ataxin-3 flexible tail, whichincludes the polyQ tract. Moreover the transcriptional regulators p300, CBP,and NCOR all contain amino acid repeats.
complex to regulate the degradation of misfolded ER proteins(Matsumoto et al., 2004; Zhong and Pittman, 2006) (Figure 5).
Biochemical studies showed that ataxin-3 displays a strongpreference for chains containing four or more ubiquitins (Chaiet al., 2004) and that full-length ataxin-3 and its JD both displayproteolytic activity toward either linear substrates containing asingle ubiquitin molecule (Burnett et al., 2003; Chow et al., 2004b;Weeks et al., 2011) or K48/K63-linked poly-ubiquitin chains (Win-born et al., 2008; Todi et al., 2009), displaying also the capacity tobind the ubiquitin-like protein NEED8 in a substrate-like fashion(Ferro et al., 2007). Moreover, ataxin-3-like protein, Josephin 1 andJosephin 2, also display ubiquitin protease activity (Tzvetkov andBreuer, 2007; Weeks et al., 2011), although the relative activities arehighly variable in spite of their high sequence similarity. Charac-terization of ataxin-3 ubiquitin hydrolase activity has also revealedthat the full-length protein preferentially cleaves Lys-63-linkedand mixed-linkage chains with more than four ubiquitins (Bur-nett et al., 2003; Winborn et al., 2008). This specificity is dictated
by the UIMs, as the isolated JD shows a preference toward thedisassembly of Lys-48-linked chains (Nicastro et al., 2009, 2010).Altogether, this indicates that ataxin-3 ubiquitin hydrolase activ-ity is likely to be associated with delivery of target substrates tothe proteasome rather than with their rescue from degradation,as it happens with most of the other deubiquitinases (Ventii andWilkinson, 2008; Matos et al., 2011; Scaglione et al., 2011). Inter-estingly, ubiquitin hydrolase activity of ataxin-3 is not affectedby polyQ expansion and both normal and expanded ataxin-3 areable to increase the cellular levels of a short-lived GFP normallydegraded by the ubiquitin-proteasome pathway (Burnett et al.,2003).
The 3D structures for JD alone or in the presence of ubiquitin aswell as that of the tandem UIM1-UIM2 have already been deter-mined (Mao et al., 2005; Nicastro et al., 2005, 2009; Song et al.,2010), giving a structural perspective on the ubiquitin hydrolasefunction of ataxin-3. The JD contains two ubiquitin binding sites,both of hydrophobic nature, with site 1 being negatively charged to
Frontiers in Neurology | Neurodegeneration June 2013 | Volume 4 | Article 76 | 12
http://www.frontiersin.org/Neurodegenerationhttp://www.frontiersin.org/Neurodegeneration/archive
Almeida et al. Structure and function of trinucleotide repeats
Tab
le2
|Hu
man
atax
in-3
asso
ciat
edp
rote
ins.
Ata
xin
-3in
tera
ctin
gp
rote
in
(Un
iPro
tac
cess
ion
cod
e)
Pro
tein
nam
eD
irec
tin
tera
ctio
n?
Inte
ract
ion
do
mai
ns
Ref
eren
ce
Ata
xin
-3Pa
rtn
erp
rote
in
CE
LL-Q
UA
LITY
CO
NT
RO
L(P
RO
TE
INH
OM
EO
STA
SIS
)
HH
R23
A/B
(P54
725/
P54
727)
UV
exci
sion
repa
irpr
otei
n
RA
D23
hom
olog
A/B
Yes,
kD(J
D:U
bl)=
12µ
MJD
Ubi
quiti
n-lik
e(U
bl)
N-t
erm
inal
dom
ain
Wan
get
al.(
2000
),D
oss-
Pepe
etal
.
(200
3),N
icas
tro
etal
.(20
05,2
009)
Poly
-ubi
quiti
n(P
0CG
48/P
0CG
47)
Poly
ubiq
uitin
-
C/P
olyu
biqu
itin-
B
Yes,
kD(a
txn3
:K48
-
tetr
aUb)=
0.2
µM
,kD
(atx
n3:U
b)=
50µ
M
UIM
s,JD
K48
-and
K63
-link
edU
b
(≥4
Ub)
,K48
-link
eddi
Ub
Bur
nett
etal
.(20
03),
Dos
s-Pe
pe
etal
.(20
03),
Cha
iet
al.(
2004
),
Nic
astr
oet
al.(
2009
,201
0)
Ubi
quili
n-1
(Q9U
MX
0)Pr
otei
nlin
king
IAP
with
cyto
skel
eton
1
n.d.
n.d.
n.d.
Hei
ret
al.(
2006
)
NE
DD
8(Q
1584
3)U
biqu
itin-
like
prot
ein
Ned
d8
Yes
JDN
ED
D8
Ferr
oet
al.(
2007
)
Park
in(O
6026
0)E
3ub
iqui
tin-p
rote
inlig
ase
park
in
Yes
JD,U
IMs
IBR
dom
ain,
Ubi
quiti
n-lik
e
(Ubl
)dom
ain
Dur
can
etal
.(20
11,2
012)
Ubc
7(P
6225
3)U
biqu
itin-
conj
ugat
ing
enzy
me
E2
G1
Yes
(tra
nsie
ntin
tera
ctio
n
dete
cted
usin
g
cros
s-lin
king
reag
ents
)
n.d.
n.d.
Dur
can
etal
.(20
12)
p45
(P62
195)
26S
prot
easo
me
regu
lato
rysu
buni
t8
Yes
N-t
erm
inal
atxn
3re
gion
(res
idue
s1–
133)
n.d.
Wan
get
al.(
2007
)
20S
Prot
easo
me
(P25
786,
P25
787,
P25
788,
P25
789,
P28
066,
P60
900,
O14
818,
P20
618,
P49
721,
P49
720,
P28
070,
P28
074,
P28
072,
Q99
436)
Prot
easo
me
subu
nits
α
type
s1-
7an
dβ
type
s1-
7
n.d.
N-t
erm
inal
atxn
3re
gion
(res
idue
s1–
150)
n.d.
Dos
s-Pe
peet
al.(
2003
)
CH
IP(Q
9UN
E7)
E3
ubiq
uitin
-pro
tein
ligas
e
CH
IP
Yes,
kD
(atx
n3:C
HIP
)=2.
2µ
M,k
D
(atx
n3:U
b-C
HIP
)=0.
1µ
M
Atx
n3C
-ter
min
us
(res
idue
s13
3–35
7)
CH
IPN
-ter
min
usJa
naet
al.(
2005
),S
cagl
ione
etal
.
(201
1)
VCP
/p97
(P55
072)
Tran
sitio
nale
ndop
lasm
ic
retic
ulum
ATPa
se
Yes
Res
idue
s27
7–28
1
(incl
udes
argi
nine
/lysi
ne-r
ich
NLS
)
Ndo
mai
n,re
sidu
es1-
199
Hira
baya
shie
tal
.(20
01),
Dos
s-Pe
pe
etal
.(20
03),
Mat
sum
oto
etal
.
(200
4,?)
Boe
ddric
het
al.(
2006
),an
d
Zhon
gan
dP
ittm
an(2
006)
E4B
(O95
155)
Ubi
quiti
nco
njug
atio
n
fact
orE
4B
Yes
(with
79Q
-ata
xin-
3)n.
d.n.
d.M
atsu
mot
oet
al.(
2004
)
(Con
tinue
d)
www.frontiersin.org June 2013 | Volume 4 | Article 76 | 13
http://www.frontiersin.orghttp://www.frontiersin.org/Neurodegeneration/archive
Almeida et al. Structure and function of trinucleotide repeats
Tab
le2
|Co
nti
nu
ed
Ata
xin
-3in
tera
ctin
gp
rote
in
(Un
iPro
tac
cess
ion
cod
e)
Pro
tein
nam
eD
irec
tin
tera
ctio
n?
Inte
ract
ion
do
mai
ns
Ref
eren
ce
Ata
xin
-3Pa
rtn
erp
rote
in
OTU
B2
(Q96
DC
9)U
biqu
itin
thio
este
rase
OTU
B2
n.d.
n.d.
n.d.
Sow
aet
al.(
2009
)
US
P13
(Q92
995)
Ubi
quiti
nca
rbox
yl-t
erm
inal
hydr
olas
e13
n.d.
n.d.
n.d.
Sow
aet
al.(
2009
)
KC
TD10
(Q9H
3F6)
BTB
/PO
Z
dom
ain-
cont
aini
ngad
apte
r
for
CU
L3-m
edia
ted
Rho
A
degr
adat
ion
prot
ein
3
n.d.
n.d.
n.d.
Sow
aet
al.(
2009
)
Tubu
lindi
mer
(Q71
U36
/P68
363)
Tubu
linα-1
A,T
ubul
inβ-2
BYe
s,kD
(atx
n3:tu
bulin
)=50
–70
nM
JDn.
d.M
azzu
cche
lliet
al.(
2009
)
Dyn
ein
(Q9Y
6G9)
Cyt
opla
smic
dyne
in1
light
inte
rmed
iate
chai
n1
n.d.
n.d
n.d.
Bur
nett
and
Pitt
man
(200
5)
HD
AC
6(Q
9UB
N7)
His
tone
deac
etyl
ase
6n.
d.n.
d.n.
d.B
urne
ttan
dP
ittm
an(2
005)
TR
AN
SC
RIP
TIO
NA
LR
EG
ULA
TIO
N
p300
(Q09
472)
His
tone
acet
yltr
ansf
eras
e
p300
Yes
Poly
Q-c
onta
inin
gC
term
inus
ofat
xn3
(res
idue
s28
8–35
4)
n.d.
Liet
al.(
2002
)
CB
P(Q
9279
3)cA
MP-
resp
onse
-ele
men
t
bind
ing
prot
ein
(CR
EB
)-bin
ding
prot
ein
Yes
Poly
Q-c
onta
inin
gC
term
inus
ofat
xn3
(res
idue
s28
8–35
4)
n.d.
Liet
al.(
2002
)
PC
AF
(Q92
831)
p300
/CR
EB
-bin
ding
prot
ein-
asso
ciat
edfa
ctor
:
hist
one
acet
yltr
ansf
eras
e
KAT
2B
Yes
Poly
Q-c
onta
inin
gC
term
inus
ofat
xn3
(res
idue
s28
8–35
4)
n.d.
Liet
al.(
2002
)
His
tone
H3/
H4
(P68
431/
P62
805)
His
tone
Yes
JD+
UIM
1an
d2
(res
idue
s1–
288)
n.d.
Liet
al.(
2002
)
HD
AC
3(O
1537
9)hi
ston
ede
acet
ylas
e3
Yes
n.d.
n.d.
Eve
rtet
al.(
2006
)
NC
OR
1(O
7537
6)N
ucle
arre
cept
or
core
pres
sor
1
n.d.
n.d.
n.d.
Eve
rtet
al.(
2006
)
MA
ML3
(Q96
JK9)
Mas
term
ind-
like
prot
ein
3n.
d.n.
d.n.
d.R
avas
iet
al.(
2010
)
EW
SR
1(Q
0184
4)R
NA
-bin
ding
prot
ein
EW
Sn.
d.n.
d.Vi
naya
gam
etal
.(20
11)
(Con
tinue
d)
Frontiers in Neurology | Neurodegeneration June 2013 | Volume 4 | Article 76 | 14
http://www.frontiersin.org/Neurodegenerationhttp://www.frontiersin.org/Neurodegeneration/archive
Almeida et al. Structure and function of trinucleotide repeats
Tab
le2
|Co
nti
nu
ed
Ata
xin
-3in
tera
ctin
gp
rote
in
(Un
iPro
tac
cess
ion
cod
e)
Pro
tein
nam
eD
irec
tin
tera
ctio
n?
Inte
ract
ion
do
mai
ns
Ref
eren
ce
Ata
xin
-3Pa
rtn
erp
rote
in
SIG
NA
LT
RA
NS
DU
CT
ION
CK
2(P
1978
4)C
asei
nki
nase
IIsu
buni
tα
Yes
n.d.
n.d.
Tao
etal
.(20
08),
Mue
ller
etal
.
(200
9)
GS
K3B
(P49
841)
Gly
coge
nsy
ntha
se
kina
se-3
β
Yes
n.d
n.d
Feie
tal
.(20
07),
Vina
yaga
met
al.
(201
1)
DN
M2
(P50
570)
Dyn
amin
-2n.
d.n.
d.n.
d.Vi
naya
gam
etal
.(20
11)
CD
KN
1A(P
3893
6)C
yclin
-dep
ende
ntki
nase
inhi
bito
r1
n.d.
n.d.
n.d.
Vina
yaga
met
al.(
2011
)
AN
XA
7(P
2007
3)A
nnex
inA
7n.
d.n.
d.n.
d.Vi
naya
gam
etal
.(20
11)
RP
S6A
K1
(Q15
418)
Rib
osom
alpr
otei
nS
6
kina
seα-1
n.d.
n.d.
n.d.
Vina
yaga
met
al.(
2011
)
TK1
(P04
183)
Thym
idin
eki
nase
,cyt
osol
icn.
d.n.
d.n.
d.Vi
naya
gam
etal
.(20
11)
MK
NK
1(Q
9BU
B5)
MA
Pki
nase
-inte
ract
ing
serin
e/th
reon
ine-
prot
ein
kina
se1
n.d.
n.d.
n.d.
Vina
yaga
met
al.(
2011
)
ATA
XIO
ME
TEX
11(Q
8IY
F3)
Test
is-e
xpre
ssed
sequ
ence
11pr
otei
n
n.d.
n.d.
n.d.
Lim
etal
.(20
06)
C16
orf7
0(Q
9BS
U1)
UP
F018
3pr
otei
nC
16or
f70
n.d.
n.d.
n.d.
Lim
etal
.(20
06)
AR
HG
AP
19(Q
14C
B8)
Rho
GTP
ase-
activ
atin
g
prot
ein
19
n.d.
n.d.
n.d.
Lim
etal
.(20
06)
PIC
K1
(Q9N
RD
5)P
RK
CA
-bin
ding
prot
ein
n.d.
n.d.
n.d.
Lim
etal
.(20
06)
Box
essh
aded
ingr
ayre
pres
ent
asso
ciat
ions
iden
tified
inhi
gh-t
hrou
ghpu
tin
tera
ctom
esc
reen
ings
.
Atx
n3,
atax
in-3
;IB
R,
InB
etw
een
Rin
gfin
gers
;JD
,Jo
seph
indo
mai
n;n.
d.,
not
dete
rmin
ed;
NLS
,nu
clea
rlo
caliz
atio
nse
quen
ce;
Ub,
ubiq
uitin
;U
BA
,ub
iqui
tinas
soci
ated
dom
ain;
Ubl
,ub
iqui
tin-li
kedo
mai
n;U
IM,
ubiq
uitin
-inte
ract
ing
mot
ifs.
www.frontiersin.org June 2013 | Volume 4 | Article 76 | 15
http://www.frontiersin.orghttp://www.frontiersin.org/Neurodegeneration/archive
Almeida et al. Structure and function of trinucleotide repeats
FIGURE 5 | Overview of ataxin-3 protein interaction network. Data onthe ataxin-3 interactors was obtained by analysis of Interactome3D (Moscaet al., 2012), MINT (Ceol et al., 2010), and Dr. PIAS (Sugaya and Furuya,2011) protein interaction databases, and completed with data compiledfrom current literature on ataxin-3 protein associations obtained with adiverse set of experimental approaches (see complete information on
Table 2). Red arrows indicate interactions for which structural data has beenobtained, while orange arrows indicate that biophysical data on interactionaffinity in vitro is known (Table 2). Broken arrows represent interactions thatresult from high-throughput interactome analysis that still require detailedbiochemical and functional analysis. Proteins are grouped according to theirbiological role.
facilitate docking of the positively charged ubiquitin C-terminusclose to the catalytic site. Binding of ubiquitin to site 1 is of crucialimportance for both JD and full-length ataxin-3 activity as ubiqui-tin hydrolase (Nicastro et al., 2010). Site 2 confers ubiquitin-chainlinkage preference to ataxin-3 and it overlaps with the surface forinteraction of the ubiquitin-like domain in HHR23B (Nicastroet al., 2005, 2010). Solution structure for the two UIMs (UIM1and UIM2), which are separated by a short 2 amino acid spacer,revealed that they fold into two α-helices separated by a flexiblelinker (Song et al., 2010). Upon ubiquitin binding, this structureadopts a typical helix-loop-helix folding pattern, where hydropho-bic interactions dominate the complex formation (Song et al.,2010). When in tandem, UIM1 and UIM2 show higher bindingaffinity for mono- or poly-ubiquitin than individual UIMs (Songet al., 2010), suggesting a cooperative binding mechanism (Songet al., 2010). The effect of the presence of UIM3 in ataxin-3 bindingaffinity for ubiquitin has not been shown, but its role in ubiqui-tin chain binding and recognition is unlikely to be of relevance toataxin-3 activity, since no differences in proteolytic activity wereidentified when the 2UIM and 3UIM isoforms were compared. Inthe model proposed for ataxin-3 ubiquitin chain proteolysis, theUIMs (UIM1-UIM2) select and recruit poly-ubiquitin substrates,presenting them to the catalytic JD for cleavage (Mao et al., 2005).
Even though ataxin-3 functions as ubiquitin hydrolase, its pro-teolytic activity is rather low, indicating that either ataxin-3/JD
requires additional factors (post-translational modifications,cofactors, intracellular interactions) to exhibit significant prote-olytic activity or the substrates used in vitro so far are not optimal.Interestingly, only three amino acid mutations are sufficient tosignificantly increase the proteolytic activity of ataxin-3, to avalue close to that of ataxin-3-like protein (Weeks et al., 2011).Under physiological conditions, one candidate for an activatingsignal is mono-ubiquitination at K117, which has been shownto increase the enzyme’s rate of cleavage of Lys-63 linked sub-strates (Todi et al., 2009). However, the molecular mechanism bywhich ubiquitination increases enzyme activity is not still clear,nor is it known whether other cellular signals (e.g., phospho-rylation by CK2 or GSK3b; Fei et al., 2007; Tao et al., 2008)may also modulate the activity of ataxin-3. Interestingly the JD-containing protein, Josephin 1 was also demonstrated to cleaveubiquitin chains only after it is mono-ubiquitinated (Seki et al.,2013). The regulation of ataxin-3 activity through ubiquitinationmight depend on the interaction of ataxin-3 with several E3 ubiq-uitin ligases (Durcan and Fon, 2013), such as the C-terminus of70 kDa heat-shock protein (Hsp70)-interacting protein (CHIP),parkin, and E4B (Figure 5), since all were shown to promoteataxin-3 ubiquitination and regulate its degradation by the pro-teasome (Matsumoto et al., 2004; Jana et al., 2005; Miller et al.,2005). Association of ataxin-3 with CHIP is a multistep processregulated by mono-ubiquitination of the N-terminal region of
Frontiers in Neurology | Neurodegeneration June 2013 | Volume 4 | Article 76 | 16
http://www.frontiersin.org/Neurodegenerationhttp://www.frontiersin.org/Neurodegeneration/archive
Almeida et al. Structure and function of trinucleotide repeats
CHIP by the E2-conjugating enzyme Ube2w, and occurs throughthe region encompassing polyQ and UIM1 and 2 (Jana et al.,2005) (Figure 4). As observed for other interactions involvingthe C-terminal region of ataxin-3, the ataxin-3-CHIP complex isaffected by polyQ expansion and the polyQ-expanded protein dis-plays a sixfold increase in binding affinity (Scaglione et al., 2011).The presence of ataxin-3 in multicomponent E3-ligase complexesis also supported by the identification of a direct interactionwith parkin, an association that stabilizes the interaction betweenparkin and the E2-conjugating enzyme Ubc7 (Durcan et al., 2011).In contrast with what is observed in the ataxin-3:CHIP com-plex, ataxin-3 association with parkin remains unaltered by polyQexpansion (Durcan et al., 2012) (Figure 4). However, we still donot understand the mechanisms that regulate shuttling of ataxin-3 between these functional complexes or how its distribution ismodulated by polyQ expansion. Further biochemical studies arerequired to establish the correlation between these macromolec-ular interactions and their relevance for ataxin-3 aggregation andneurodegeneration in MJD patients
ATAXIN-3 AGGREGATION: A MULTISTEP PATHWAY MODULATED BYTHE PROTEIN CONTEXTA characteristic hallmark of MJD and other polyQ-expansion dis-eases is the appearance of intracellular inclusions enriched inthe disease protein and containing components from the cell-quality control machinery (e.g., ubiquitin, proteasome subunits,and chaperones), indicating that these diseases form part of thelarger family of protein misfolding disorders (Williams and Paul-son, 2008). Early in vitro studies showed that expansion of thepolyQ tract within the pathological range induced formationof insoluble β-rich fibrils with the capacity to bind amyloid-specific dyes (Bevivino and Loll, 2001). Later it was demonstratedthat non-pathological ataxin-3 could also form insoluble fibrillaraggregates upon destabilization of its structure by temperature,pressure or denaturing agents (Marchal et