Secondary structure assignment of mouse SOCS3 by NMR
defines the domain boundaries and identifies an
unstructured insertion in the SH2 domain
Jeffrey J. Babon
1
, Shenggen Yao
1
, David P. DeSouza
1,
*, Christopher F. Harrison
1,
*, Louis J. Fabri
2
,
Edvards Liepinsh
3
, Sergio D. Scrofani
2
, Manuel Baca
1,
† and Raymond S. Norton
1
1 Walter and Eliza Hall Institute, Parkville, Victoria, Australia
2 Amrad Corporation Ltd, Richmond, Victoria, Australia
3 Department of Medical Biochemistry and Biophysics, Karolinska Institute, Stockholm, Sweden
Cytokine signalling acts through membrane-bound,
multisubunit receptor complexes that are phosphoryl-
ated by activated Janus kinases (JAKs), leading to sub-
sequent activation and phosphorylation of members of
the signal transduction and activators of transcription
(STAT) family. The duration of the signalling response

mediated signal tranduction by binding to phosphorylated tyrosine residues
on intracellular subunits of various cytokine receptors, as well as possibly
the JAK proteins. SOCS3 consists of a short N-terminal sequence followed
by a kinase inhibitory region, an extended SH2 domain and a C-terminal
suppressor of cytokine signalling (SOCS) box. SOCS3 and the related pro-
tein, cytokine-inducible SH2-containing protein, are unique among the
SOCS family of proteins in containing a region of mostly low complexity
sequence, between the SH2 domain and the C-terminal SOCS box. Using
NMR, we assigned and determined the secondary structure of a murine
SOCS3 construct. The SH2 domain, unusually, consists of 140 residues,
including an unstructured insertion of 35 residues. This insertion fits the
criteria for a PEST sequence and is not required for phosphotyrosine bind-
ing, as shown by isothermal titration calorimetry. Instead, we propose that
the PEST sequence has a functional role unrelated to phosphotyrosine
binding, possibly mediating efficient proteolytic degradation of the protein.
The latter half of the kinase inhibitory region and the entire extended SH2
subdomain form a single a-helix. The mapping of the true SH2 domain,
and the location of its C terminus more than 50 residues further down-
stream than predicted by sequence homology, explains a number of previ-
ously unexpected results that have shown the importance of residues close
to the SOCS box for phosphotyrosine binding.
Abbreviations
CIS, cytokine-inducible SH2-containing protein; ESS, extended SH2 subdomain; IPTG, isopropyl thio-b-
D-galactoside; ITC, isothermal titration
calorimetry; JAK, Janus kinase; KIR, kinase inhibitory region; PtdIns, phosphatidylinositol; SOCS, suppressor of cytokine signalling; STAT,
signal transduction and activator of transcription.
6120 FEBS Journal 272 (2005) 6120–6130 ª 2005 The Authors Journal Compilation ª 2005 FEBS
themselves [2]. They act therefore by directly blocking
signal transduction or by interfering with STAT access
to the phosphorylated receptor subunits.

domain of murine SOCS3 has been mapped previously
by sequence comparison to residues 46–142 [10], but
mutagenesis experiments have shown that residues as
far away as Leu182 are important for phosphotyro-
sine–peptide binding [3]. There is therefore some uncer-
tainty about the extent of the SH2 domain, depending
on whether it is predicted by sequence homology or
functional analysis.
In addition to their role in blocking the activation
of downstream signalling intermediates, the SOCS pro-
teins may also act by directing the degradation of
bound signalling molecules [11]. As the C-terminal
SOCS box is capable of interacting with an E3–ubiqu-
itin ligase complex by binding directly to elongins B
and C [12], SOCS proteins can recruit bound signal
transduction proteins, such as activated kinases or the
cytokine receptors themselves, for proteasome-medi-
ated degradation [11,13,14]. Although there is no struc-
tural information on the SOCS box, sequence and
functional homologies suggest that it will adopt a sim-
ilar structure to the corresponding region in the VHL
protein [15], which is also responsible for binding to
elonginB ⁄ C. Reports differ as to whether the interac-
tion between elonginB ⁄ C and SOCS stabilizes [16,17]
or destabilizes [12,18] the SOCS proteins themselves.
Unambiguous secondary structure assignment, whe-
ther by NMR or other spectroscopic techniques, can
socs1
socs2
socs3

not required for phosphotyrosine binding but may
have an important functional role.
Results
SOCS3 phosphotyrosine peptide complex
Two initial constructs of mouse SOCS3 [SOCS3(22–
225) and SOCS3(22–185)] were cloned and expressed
in Escherichia coli. Both contain the KIR and the
extended SH2 domain, but SOCS3(22–185) lacks the
C-terminal SOCS box. Both constructs expressed in
inclusion bodies in E. coli and required refolding. A
phosphotyrosine peptide from gp130 (STASTV-
EpYSTVVHSG) has been shown previously to bind
with high affinity to mouse SOCS3 [19,20]. The addi-
tion of a molar excess of peptide significantly increased
the solubility in NaCl ⁄ P
i
from <<1mgÆmL
)1
to
 1mgÆmL
)1
for SOCS3(22–225) and to 3 mgÆmL
)1
for SOCS3(22–185).
As SOCS3(22–185) in the presence of the tyrosine-
phosphorylated peptide could not be concentrated
beyond  0.2 mm, seven constructs of shorter length
were expressed in E. coli and their solubility examined.
All constructs contained the SH2 domain, as defined by
sequence homology [10], but included differing lengths

peptide, no higher than the wild-type SOCS3(22–185)
construct. As SOCS3(22–225) was too poorly soluble
to obtain any meaningful structural data, the wild-type
SOCS3(22–185) construct was pursued.
NMR assignments for murine SOCS3(22–185)
After buffer optimization, SOCS3(22–185) was soluble
to  0.5 mm, but UV-visible spectra of the protein
showed that significant aggregation was occurring at
this concentration, indicated by a high apparent absorp-
tion at 320 nm as a result of scattering. Many NMR
experiments required for full protein assignment there-
fore did not yield acceptable results, in particular
HNCACB, HCCH-TOCSY and
13
C-NOESY-HSQC.
Nevertheless, near-complete backbone resonance assign-
ments were made for SOCS3(22–185). Apart from five
missing spin systems (Ser25–Ser28 and Gly170), 100%
of
1
H
N
, 100% of
15
N (excluding 18 proline residues),
96% of
13
C
a
, 84% of

15
N-edited NOESY-HSQC and by
using talos [21]. Assignments revealed that SOCS3 had
an aabbbbbabbb topology, with the ESS and the C-ter-
minal end of KIR forming the first a -helix (Fig. 2). Sig-
nificantly, there was a large unstructured region between
Met128 and Arg163 that contained a high proportion of
proline residues (12 out of 35). The chemical shifts of
mouse SOCS3 have been deposited in BioMagRes-Bank
(http://www.bmrb.wisc.edu) with accession number
6580.
Murine SOCS3 contains a PEST region
The sequence of the unstructured region of murine
SOCS3 is highly conserved in mammalian SOCS3, as
Domain characterization of SOCS3 J. J. Babon et al.
6122 FEBS Journal 272 (2005) 6120–6130 ª 2005 The Authors Journal Compilation ª 2005 FEBS
shown in Fig. 3. This region displays all of the com-
mon features of PEST sequences [22], namely a high
proportion of Pro, Glu, Ser and Thr residues, the
absence of Lys, His and Arg except at the termini, and
the fact that it is completely unstructured based on the
absence of medium- and long-range NOEs and the
observation of intense backbone amide peaks (Fig. S1).
The primary sequence of SOCS3 was analysed for the
presence of a PEST sequence by using the pestfind
program (http://www.at.embnet.org/embnet/tools/bio/
PESTfind) [23]. This analysis identified the likely pres-
ence (PESTfind score +11.11 [23]) of a single PEST
sequence in SOCS3 spanning residues His126–Lys162.
The unstructured region of SOCS3 spans Met128–

well with the seven C-terminal residues of a number of
SH2 domains, supporting this hypothesis. In agreement
with this scenario, deletion of residues 182–185 had
been shown previously to affect phosphotyrosine pep-
tide binding [3]. Although the sequence between Tyr165
and Pro175, which would form the ‘BG loop’, was not
significantly similar to other SH2 domains, the SHP-2
[26], grb7 [27] and, in particular, STAT3b [9] SH2
domains contain extended loops in this region that
A
B
Fig. 2. 15N-
1
H HSQC spectrum and secondary structure assign-
ment of SOCS3(22–185). (A) The
15
N-
1
H HSQC spectrum is shown
of 0.1 m
M SOCS3 at 500 MHz and 298 K in 50 mM sodium-phos-
phate buffer (pH 6.7) containing 2 m
M dithiothreitol. The assigned
residues are labelled with their residue number in the HSQC; some
assignment labels are omitted for clarity. (B) The secondary struc-
ture of SOCS3 was assigned by examining NOE patterns, analyses
of backbone and
13
C
b

of 74 ± 7 nm. The K
d
of wild-type
SOCS3(22–185) binding was 152 ± 25 nm.
PEST sequences in other SOCS family proteins
In order to determine whether other members of the
SOCS family contained PEST motifs, their sequences
were analysed using the PESTfind algorithm [23].
Of the eight members of the murine SOCS family,
SOCS1, -3, -5 and -7, and CIS, show a probable
PEST motif with a PESTfind score of > 5 (Table 1).
CIS and SOCS3 have the PEST motif within the
SH2 domain, while SOCS1, -5 and -7 contain PEST
motifs in the N-terminal domain. The PEST sequence
in the CIS SH2 domain is located eight residues
downstream from the terminus of the predicted aB
helix. Whether those eight residues are also unstruc-
tured, thus placing the unstructured insertion at an
identical position to the PEST sequence in SOCS3,
could not be determined. Secondary structure predic-
tion by sequence analysis gives no prediction for
those eight residues.
The KIR ⁄ ESS consists of a single a-helix
Based on observed NOEs, chemical shift deviations and
TALOS predictions, residues Glu29–Ser44 form a sin-
gle a-helix, whilst residues 22–28 are unstructured. The
helix encompasses the entire ESS and the four residues
at the C terminus of the KIR. The remaining residues
that comprise the KIR appear to be unstructured.
Discussion

modelled the structures of the ESS and KIR of
Fig. 3. PEST sequence conservation in SOCS3. Sequence alignment of the region of SOCS3 containing the PEST motif for a number of
mammalian species is shown, with conserved residues in the unstructured PEST motif shown hatched in grey. The numbering refers to
mouse SOCS3. The unstructured residues defined by this study are shown in bold.
Domain characterization of SOCS3 J. J. Babon et al.
6124 FEBS Journal 272 (2005) 6120–6130 ª 2005 The Authors Journal Compilation ª 2005 FEBS
SOCS1, based on the similarity of the ESS sequence to
a similar region in Stat1 [8] and Stat3b [9], and sugges-
ted that the ESS and KIR form two short orthogonal
helices. This differs from the single a-helix found in
SOCS3, but a comparison of the ESS sequences from
SOCS1 and SOCS3 shows that there are several diver-
gent residues in this region, including a leucine (Leu32)
in SOCS3 in place of an arginine (Arg67) in SOCS1,
predicted in their model to make a critical ion pair
with Asp76.
The identification, by deletion mutagenesis, of resi-
dues affecting phosphotyrosine binding but located
> 50 residues downstream of the predicted C terminus
of the SH2 domain suggested that the functional SH2
domain was longer than originally suggested by
sequence comparison [3]. However, subsequent
attempts to determine key residues important for the
binding specificity of SOCS3 by structural modelling
were hampered by the logical, yet incorrect, assump-
tion that the SH2 domain consisted of c. 100 contigu-
ous residues. In this report we have shown that the
true SH2 domain is disrupted in murine SOCS3 by a
35-residue unstructured insertion that is predicted to
form a PEST motif [22]. This results in residues 164–

proteolytic degradation [28,29]. This may be important
mechanistically, as the efficient turnover of SOCS pro-
teins, and their induction of degradation of associated
signalling molecules via the SOCS box, allows cells to
respond to cytokine stimulation, quickly inhibit any
prolonged activation and rapidly return to basal SOCS
levels, ready for another round of stimulation. There
appear to be a number of features important for effect-
ive degradation of SOCS proteins, even apart from
any role the SOCS box may play in this process. Sasaki
et al. [28] have shown that a naturally occurring
alternative transcript of SOCS3, lacking the first 11
Fig. 4. SOCS3 lacking the PEST motif binds
a gp130 peptide with high affinity. (A) Titra-
tion of 80 l
M gp130 peptide into 10 lM
SOCS3(D101–133). The integrated heats
from which the heat of dilution has been
subtracted are shown, as well as the fit to a
single site binding isotherm that yielded K
d
78 nM and DH )6.4 kcal mol
)1
. (B) Titration
of 160 l
M gp130 peptide into 13 lM wild
type SOCS3. The fit to a single site binding
isotherm is shown, yielding a K
d
of 168 nM

manner, as transplanting PEST sequences from unsta-
ble proteins into stable proteins has been shown to
reduce the half-life of the resulting chimaeras
[32,35,36]. The presence of a PEST sequence has been
shown to be important in the proteolysis ⁄ degradation
of a number of proteins with diverse functions, such as
the glutamate receptor [37], proto-Dbl [38], and c-Fos
[39]. Biophysical characterization of the NF-jb PEST
sequence [40] has shown it to be solvent-exposed and
probably unstructured.
The PEST sequence in SOCS3 is located between
two secondary structural elements, namely the aB
helix and the BG loop. In all SH2 domains these are
located on the opposite face of the protein to the
phosphotyrosine-binding site, so the PEST sequence is
not expected to interfere with phosphotyrosine binding
by the SH2 domain. Indeed, replacing the entire
35-residue PEST sequence with GSGSGSGS had little
effect upon binding a phosphorylated gp130 peptide,
as shown by ITC. In fact, the construct lacking the
PEST motif bound slightly more tightly to the phos-
photyrosine containing gp130 peptide than did wild-
type SOCS3(22–185). Whether the twofold change in
K
d
is significant is difficult to determine as the con-
struct lacking the PEST motif shows significantly less
aggregation than wild-type SOCS3, which could alter
the binding kinetics without representing a truly
enhanced K

score Sequence Location
SOCS1 14.2 22 RSEPSSSSSSSSPAAPVR 39 N terminus
SOCS2 NA NA NA
SOCS3 11.1 126 HYMPPPGTPSFSLPPTEPSSEVPEQPPAQALPGSTPK 162 SH2
SOCS4 NA NA NA
SOCS5 7.4 96 KDSDSGATPGTR 107 N terminus
11.3 243 HSTFFDTFDPSLVSTEDEEDR 263 N terminus
SOCS6 NA NA NA
SOCS7 11.9 74 KTAGGGCCP CPCPPQPPPPQPPPPAAAPQAGEDPTETSDALLVLEGLESEAESLETNSCSEEELSSPGR 142 N terminus
CIS 9.0 172 RSDSPDPAPTPALPMSK 188 SH2
Domain characterization of SOCS3 J. J. Babon et al.
6126 FEBS Journal 272 (2005) 6120–6130 ª 2005 The Authors Journal Compilation ª 2005 FEBS
cleavage site. For unlabelled protein, expression was per-
formed in baffled flasks, with cells grown to an attenuance
(D) at 600 nm of 0.6 in superbroth containing 50 lgÆmL
)1
kanamycin. Expression was induced with 1 mm isopropyl
thio-b-d-galactoside (IPTG). Cells were harvested, 3 h after
induction, by centrifugation (6200 g,4°C, 30 min). For
15
N labelling, cells were grown to a D at 600 nm of 0.6 in
Neidhardt’s medium [41] containing 1.0 gÆL
)115
NH
4
Cl as
the sole nitrogen source. For
15
N ⁄
13

derived phosphopeptide increased the solubility to
c.3mgÆmL
)1
. As this concentration was still too low for
structure determination by high-resolution NMR, a thor-
ough screen of buffer conditions was undertaken in an
attempt to improve the maximum solubility obtainable for
SOCS3(22–185). The buffer screen was performed in micro-
drop format [42] and studied the pH range from 4 to 9 in
0.5 unit intervals, the salt concentration from 0 to 500 mm
in 50 mm intervals, and temperatures of 4, 25 and 37 °C.
Both constructs of SOCS3 showed highest solubility in
buffers of low salt and high pH, and at low temperature.
The buffer conditions chosen for further additive screening
were 20 mm Tris, pH 8.5, 20 mm NaCl, at 25 °C. This
starting condition was used to test the effect of 14 different
additives, most at several concentrations. The additive
screen yielded promising results, and SOCS3 was shown to
be soluble to  10 mgÆmL
)1
( 0.5 mm) in buffers contain-
ing > 10% glycerol, > 0.5 m non detergent sulfobetaine
(NDSB), > 0.5 m trehalose or 50 mm arginine plus 50 mm
glutamate [43]. However, initial NMR analysis showed that
under these conditions, many amide cross-peaks were
missing from
15
N HSQC spectra. As the high pH was
judged to be the cause of this, the most promising additives
were screened again at pH 6.7. At this pH only the

were analysed using xeasy (version 1.3.13) [45] or
nmrdraw [44]. Spectra were referenced to the H
2
O signal
at 4.77 p.p.m. (298 K) or a small impurity at 0.15 p.p.m.
C
a
,C
b
,H
a
,C¢, and N chemical shifts were used in the
program TALOS [21] to obtain backbone torsion angle pre-
dictions. Sequence-specific resonance assignments for the
backbone were accomplished using HNCA, HN(CO)CA,
CBCA(CO)NH, HN(CA)CO and HNCO experiments [46].
Side-chain assignments were accomplished by combining
the data from the following experiments:
15
N-edited
TOCSY-HSQC and NOESY-HSQC, HCCH-TOCSY and
HCCH-COSY [46].
ITC
Isothermal calorimetric titrations were performed using a
Microcal omega VP-ITC (MicroCal Inc., Northampton,
MA, USA). SOCS3(22–185) was dialysed against buffer
(50 mm NaCl, 50 mm arginine, 50 mm glutamate, 5 mm
2-mercaptoethanol, pH 6.7) and the dialysis buffer was
used to dissolve the tyrosine-phosphorylated gp130 peptide.
Experiments were performed at 298 K. Solutions of 10–

inhibits Janus tyrosine kinase by binding through the
N-terminal kinase inhibitory region as well as SH2
domain. Genes Cells 4, 339–351.
4 Sasaki A, Yasukawa H, Shouda T, Kitamura T, Dikic I
& Yoshimura A (2000) CIS3 ⁄ SOCS-3 suppresses ery-
thropoietin (EPO) signaling by binding the EPO recep-
tor and JAK2. J Biol Chem 275, 29338–29347.
5 Mohammadi M, Schlessinger J & Hubbard SR (1996)
Structure of the FGF receptor tyrosine kinase domain
reveals a novel autoinhibitory mechanism. Cell 86,
577–587.
6 Yasukawa H, Misawa H, Sakamoto H, Masuhara M,
Sasaki A, Wakioka T, Ohtsuka S, Imaizumi T, Matsuda
T, Ihle JN et al. (1999) The JAK-binding protein JAB
inhibits Janus tyrosine kinase activity through binding
in the activation loop. Embo J 18, 1309–1320.
7 Giordanetto F & Kroemer RT (2003) A three-dimen-
sional model of Suppressor of Cytokine Signalling 1
(SOCS-1). Protein Eng 16, 115–124.
8 Chen X, Vinkemeier U, Zhao Y, Jeruzalmi D, Darnell
JE Jr & Kuriyan J (1998) Crystal structure of a tyrosine
phosphorylated STAT-1 dimer bound to DNA. Cell 93,
827–839.
9 Becker S, Groner B & Muller CW (1998) Three-dimen-
sional structure of the Stat3beta homodimer bound to
DNA. Nature 394, 145–151.
10 Okazaki Y, Furuno M, Kasukawa T, Adachi J, Bono
H, Kondo S, Nikaido I, Osato N, Saito R, Suzuki H
et al. (2002) Analysis of the mouse transcriptome based
on functional annotation of 60,770 full-length cDNAs.

31972–31979.
17 Kamura T, Sato S, Haque D, Liu L, Kaelin WG Jr,
Conaway RC & Conaway JW (1998) The Elongin BC
complex interacts with the conserved SOCS-box motif
present in members of the SOCS, ras, WD-40 repeat,
and ankyrin repeat families. Genes Dev 12, 3872–3881.
18 Chen XP, Losman JA, Cowan S, Donahue E, Fay S,
Vuong BQ, Nawijn MC, Capece D, Cohan VL & Roth-
man P (2002) Pim serine ⁄ threonine kinases regulate the
stability of Socs-1 protein. Proc Natl Acad Sci U S A
99, 2175–2180.
19 Nicholson SE, De Souza D, Fabri LJ, Corbin J, Willson
TA, Zhang JG, Silva A, Asimakis M, Farley A, Nash
AD et al. (2000) Suppressor of cytokine signaling-3
preferentially binds to the SHP-2-binding site on the
shared cytokine receptor subunit gp130. Proc Natl Acad
Sci USA 97, 6493–6498.
20 Schmitz J, Weissenbach M, Haan S, Heinrich PC &
Schaper F (2000) SOCS3 exerts its inhibitory function
on interleukin-6 signal transduction through the SHP2
recruitment site of gp130. J Biol Chem 275, 12848–
12856.
Domain characterization of SOCS3 J. J. Babon et al.
6128 FEBS Journal 272 (2005) 6120–6130 ª 2005 The Authors Journal Compilation ª 2005 FEBS
21 Cornilescu G, Delaglio F & Bax A (1999) Protein back-
bone angle restraints from searching a database for che-
mical shift and sequence homology. J Biomol NMR 13,
289–302.
22 Rogers S, Wells R & Rechsteiner M (1986) Amino acid
sequences common to rapidly degraded proteins: the

SOCS1 deficiency causes a lymphocyte-dependent peri-
natal lethality. Cell 98, 609–616.
30 Peltola KJ, Paukku K, Aho TL, Ruuska M, Silvennoi-
nen O & Koskinen PJ (2004) Pim-1 kinase inhibits
STAT5-dependent transcription via its interactions with
SOCS1 and SOCS3. Blood 103, 3744–3750.
31 Prakash S, Tian L, Ratliff KS, Lehotzky RE &
Matouschek A (2004) An unstructured initiation site is
required for efficient proteasome-mediated degradation.
Nat Struct Mol Biol 11, 830–837.
32 Loetscher P, Pratt G & Rechsteiner M (1991) The C
terminus of mouse ornithine decarboxylase confers
rapid degradation on dihydrofolate reductase. Support
for the pest hypothesis. J Biol Chem 266, 11213–11220.
33 Huxford T, Huang DB, Malek S & Ghosh G (1998)
The crystal structure of the IkappaBalpha ⁄ NF-kappaB
complex reveals mechanisms of NF-kappaB inactiva-
tion. Cell 95, 759–770.
34 Almrud JJ, Oliveira MA, Kern AD, Grishin NV, Phil-
lips MA & Hackert ML (2000) Crystal structure of
human ornithine decarboxylase at 2.1 A resolution:
structural insights to antizyme binding. J Mol Biol 295,
7–16.
35 Ghoda L, Phillips MA, Bass KE, Wang CC & Coffino
P (1990) Trypanosome ornithine decarboxylase is stable
because it lacks sequences found in the carboxyl termi-
nus of the mouse enzyme which target the latter for
intracellular degradation. J Biol Chem 265, 11823–
11826.
36 Salama SR, Hendricks KB & Thorner J (1994) G1

Bax A (1995) NMRPipe: a multidimensional spectral
processing system based on UNIX pipes. J Biomol
NMR 6, 277–293.
45 Bartels C, Xia TH, Billeter M, Guntert P & Wuthrich
K (1995) The Program Xeasy for Computer-Supported
Nmr Spectral-Analysis of Biological Macromolecules.
J Biomol NMR 6, 1–10.
46 Sattler M, Schleucher J & Griesinger C (1999) Hetero-
nuclear multidimensional NMR experiments for the
structure determination of proteins in solution employ-
ing pulsed field gradients. Prog Nuclear Magn Reson
Spectrosc 34, 93–158.
J. J. Babon et al. Domain characterization of SOCS3
FEBS Journal 272 (2005) 6120–6130 ª 2005 The Authors Journal Compilation ª 2005 FEBS 6129
Supplementary material
The following supplementary material is available for
this article online:
Fig. S1. Chemical shifts and peak intensities for
mSOCS3(22–185). The chemical shift differences from
random coil are shown for the (A) CO, (B) C
a
, (C)
H
a
, and (D) C
b
atoms of mSOCS3 plotted against resi-
due number. (D) The average chemical shift index
values for each residue are shown. (E)
1

helix and b-sheet topology determination. Pairs of resi-
dues with either NH–NH or NH–H
a
NOEs are listed.
Domain characterization of SOCS3 J. J. Babon et al.
6130 FEBS Journal 272 (2005) 6120–6130 ª 2005 The Authors Journal Compilation ª 2005 FEBS