MINIREVIEW
Neuronal growth-inhibitory factor (metallothionein-3):
structure–function relationships
Zhi-Chun Ding*, Feng-Yun Ni

and Zhong-Xian Huang
Chemical Biology Laboratory, Department of Chemistry, Fudan University, Shanghai, China
Introduction
Metallothioneins (MTs), first discovered in horse kidney
in 1957 by Margoshes & Vallee, are a family of small
( 7 kDa), cysteine-rich and metal-binding proteins [1].
In mammals, four subfamily MTs (i.e. MT1, MT2, MT3
and MT4), have been identified [2]. MT1 and MT2 are
ubiquitous isoforms found in most organs and play criti-
cal roles in essential metal homeostasis and heavy metal
ions detoxification [3]. By contrast, MT3 and MT4
are specifically expressed in the central nervous system
and the stratified squamous epithelia, respectively [2].
MT3, first isolated and identified as a neuronal
growth-inhibitory factor (GIF), has a distinct biologi-
cal activity of inhibiting the out-growth of rat embry-
onic cortical neurons in the presence of Alzheimer’s
disease (AD) brain extracts, a function not shared by
MT1 or MT2 [4]. As a member of the MT family, GIF
exhibits approximately 70% sequence similarity with
those well-studied mammalian MTs, including: a pre-
served array of 20 cysteine residues; and two domains,
each of which wrap around a metal–thiolate cluster
Keywords
metallothionein (MT); mutation; neuronal
growth-inhibitory factor (GIF); structure–

2912 FEBS Journal 277 (2010) 2912–2920 ª 2010 The Authors Journal compilation ª 2010 FEBS
(a three-metal cluster, M(II)
3
S
9
, in the N-terminal
b-domain, and a four-metal cluster, M(II)
4
S
11
, in the
C-terminal a-domain). However, there are two inserts
in GIF that are not present in MT1 or MT2: a threo-
nine at position 5 and a glutamate-rich hexapeptide
near the C-terminus. Additionally, all known GIF
sequences contain a conserved CPCP(6–9) motif, which
is absent in all other members of the MT family
(Fig. 1).
The neuronal growth-inhibitory activity is quite
unique in the mammalian MT family. We have long
pondered how nature could utilize such a simple pro-
tein (of only 68 amino acids) to fulfill such compli-
cated functions in the central nervous system.
However, the exact molecular basis of the bioactivity
of GIF remains elusive. It has been reported that the
neuronal growth-inhibitory activity of GIF is mainly
associated with its b-domain, and the single a-domain
does not show any growth-inhibitory activity [5,6].
As a new protein the structure is mostly concerned.
Consequently, much effort has been devoted towards

NOE cross-peaks, using NMR spectroscopy [9]. Most
interestingly, the hexapeptide insertion EAAEAE(55–
60), located near the C-terminus, was modeled using a
group of possible conformations because of the lack of
NOE signals in this region (Fig. 2). Spatially, this
insertion was far from the metal–thiolate cluster,
implying that it was less restricted and therefore had a
A
B
Insertion of EAAEAE(55–60)?
Insertion of EAAEAE(55–60)?
Insertion of Thr5 ?
Insertion of Thr5 ?
CPCP(6–9)?
CPCP(6–9)?
Fig. 1. (A) Amino acid sequence alignments
of human MT1a, MT1g, MT2a, MT4 and
some mammalian GIFs. Twenty conserved
cysteine residues are highlighted. The
distinctive sequence differences of hGIF
from other MT isoforms include an insertion
of Thr5, a conservative CPCP(6–9) sequence
and insertion of the charged hexapeptide
EAAEAE(55–60). (B) Crystal structure of
rlMT2 (PDB entry: 4MT2). The sequence
dissimilarities of hGIF from other MT
isoforms are located in the rlMT2 structure
and labeled with a question mark. The zinc
ions are shown as grey spheres, the
cadmium ions are shown as green spheres

predicted using molecular dynamics simulation [12].
It was found that the peptides near the N-terminus
(residues 1–13) in hGIF folded differently from those
in rlMT2; in particular, a characteristic conformation
of the TCPCP(5–9) sequence was formed in the b-
domain of hGIF, where both Pro7 and Pro9 faced out-
wards with their five-member rings arranged almost in
parallel, while Thr5 was at the opposite side of the
two rings. The specific folding of the TCPCP(5–9)
sequence, together with the constraints from the
metal–thiolate cluster, made the peptides at the
two ends of the TCPCP(5–9) sequence twisted
(Fig. 3A1,B1). This characteristic conformation around
the TCPCP(5–9) sequence in hGIF was suggested to
provide an interacting interface for protein–protein
interactions [12–14].
The other structural feature found in the predicted
structure of the b-domain of hGIF was the hydrogen-
bond network located around the first five N-terminal
residues and the fragment from residues 23 to 26,
which was different from that found in the simulated
structure of the b-domain of rlMT2 [12]. In rlMT2
there were two hydrogen bonds, one between Asp2
and Lys25 and one between Asn4 and Gln23 making
the whole structure compact (Fig. 3A2). However, the
insertion of Thr5 into hGIF interrupted these two
interactions, and Thr5 formed a hydrogen bond with
Asp2, pushing Lys26 (equivalent to Lys25 in rlMT2)
away from Asp2. Meanwhile, Lys26 in hGIF formed
hydrogen bonds with Glu4 and Gly24 (equivalent to

Fig. 2. Solution structure of the a-domain of hGIF (PDB entry:
2F5H). A group of minimized structures are superimposed to show
clearly that the EAAEAE(55–60) insertion is structurally disordered
and extending outwards. The cadmium ions are shown as green
spheres and the sulfur atoms are shown as yellow spheres.
Structure–reactivity–function study of GIF Z C. Ding et al.
2914 FEBS Journal 277 (2010) 2912–2920 ª 2010 The Authors Journal compilation ª 2010 FEBS
the results from simulation studies are consistent with
those from NMR studies.
The simulated structure of hGIF and its mutant [the
D55-60 hGIF mutant produced by the deletion of EA-
AEAE(55–60)] also disclosed the interdomain interac-
tion mode at the atomic level, which helped to
elucidate the structural basis of the relevance of
the hexapeptide insertion to the biological function of
hGIF [16]. The first point was that the interdomain
interaction modes were not exactly the same in hGIF
and rlMT2 (Fig. 4A,B). The common features were
that Lys32 in hGIF (equivalent to Lys31 in rlMT2)
faced into the interior of the b-domain to neutralize
the negative charge of the B-cluster, thus stabilizing
the b-domain; and Ser33 in hGIF (equivalent to Ser32
in rlMT2) formed a hydrogen bond with Cys38 (equiv-
alent to Cys37 in rlMT2) to make the a-domain more
stable. The differences lay in the fact that the addi-
tional hydrogen bond between Lys31 and Glu41 in
hGIF made the fragment around residue 41 closer to
the linker region, while Lys30 in rlMT2 (equivalent to
Lys31 in hGIF) had no direct interaction with Gly40
(equivalent to Glu41 in hGIF).

conformation between the fragment near
the N-terminus and the fragment from resi-
dues 23 to 26 in hGIF compared with that in
rlMT2. The cadmium ions are shown as
green spheres and the sulfur atoms are
shown as yellow spheres. The red regions
stress the differences.
Z C. Ding et al. Structure–reactivity–function study of GIF
FEBS Journal 277 (2010) 2912–2920 ª 2010 The Authors Journal compilation ª 2010 FEBS 2915
that Lys32 in this mutant would not induce the
structural re-arrangement in the b-domain that was
observed in the wild-type hGIF (Fig. 4D1,D2). These
results interpret structurally how the hexapeptide
EAAEAE(55–60) in hGIF exerts its function by affect-
ing the conformation of the b-domain.
Based upon the molecular structure of GIF, we have
conducted a systematic mutational study on the struc-
ture–property–reactivity–function relationship of GIF.
These results will provide us with valuable data to
understand the molecular mechanism of the neuronal
growth-inhibitory activity of GIF.
The role of the conserved TCPCP motif
The CPCP motif was the first segment demonstrated to
be indispensible for the neuronal growth-inhibitory
activity of GIF [6,17]. The bioactivity of GIF is com-
pletely abolished by a double mutation from Cys6-
Pro7-Cys8-Pro9 to either Cys6-Ser7-Cys8-Ala9 (the
P7S ⁄ P9A mutant) found in MT1 and MT2 or Cys6-
Thr7-Cys8-Thr9 (the P7T ⁄ P9T mutant) [6,17].
113

inactive mouse MT1 (the S6P ⁄ S8P mutant), and
examined its inhibitory activity [18]. Quite unexpect-
edly, the S6P ⁄ S8P mutant of mouse MT1 did not
show any inhibitory activity. However, introduction
of a unique Thr5 insert before the S6P ⁄ S8P motif in
the modified mouse MT1 restored the neuronal bio-
activity [18]. The neuronal assay results undoubtedly
reflect that the CPCP motif alone is insufficient for
the bioactivity of GIF, and that both the Thr5 insert
and the CPCP motif are necessary for the neuronal
bioactivity of GIF. The
113
Cd NMR results showed
that the acquisition of GIF bioactivity in the
A
B
C
D1 D2
Fig. 4. Simulated structure of rlMT2 (A), hGIF (B), the D55-60
mutant of hGIF (C), the b-domain of hGIF (D1) and the b-domain of
the D55-60 mutant of hGIF (D2). The EAAEAE(55–60) insertion in
hGIF is shown in red. Comparison between panels A and B clearly
shows that the interdomain interaction modes are different in rlMT2
and hGIF. Comparison between panels B and C shows that the
deletion of EAAEAE(55–60) in hGIF would change the interdomain
interaction modes in hGIF. Comparison between panels D1 and D2
shows that the deletion of EAAEAE(55–60) in hGIF would affect the
hydrogen-bond network in the b-domain of hGIF. The cadmium
ions are shown as green spheres and the sulfur atoms are
shown as yellow spheres. The red regions stress the differences.

which has an ACPCP(5–9) motif instead of the
TCPCP(5–9) motif present in hGIF, still retains about
60% of bioactivity [19]. However, a previous study
clearly showed that the bioactivity of GIF is almost
completely abolished by replacing Thr5 with Ala5
[15,18]. This contradiction raises the question of why
the bioactivity of sGIF is unaffected by replacement of
Thr5 with Ala5. After comparison of the amino acid
sequence of sGIF with that of hGIF, it was found that
sGIF appears to contain three fewer cysteine residues,
owing to deletion of the sequence Ser-Cys-Cys (nor-
mally found at positions 33-35 of hGIF) and the
replacement of a cysteine residue (normally found at
position 30) with serine. However, both the Cd
2+
titra-
tion and ESI-MS results showed that sGIF binds seven
metal ions with the overall metal-to-thiolate ratio of
Cd
7
S
17
. These seven metal ions were wrapped into two
separate metal–thiolate clusters by the polypeptide
chain of sGIF: one M
3
cluster and one M
4
cluster, in
which the M

molar basis [6,23]. Hence, it is suggested that the
a-domain might play some important roles in the neu-
ronal growth-inhibitory activity of hGIF. To confirm
this assumption, Ding et al. constructed two domain-
hybrid mutants, in which the a-domain of hGIF was
replaced with either the b-domain of hGIF [the
b(MT3)-b(MT3) mutant] or the a-domain of hMT1g
[the b(MT3)-a(MT1) mutant] [23]. It was found that
the metal-binding ability and solvent accessibility of the
Cd
3
S
9
cluster of the b-domain of the b(MT3)-b(MT3)
mutant decreased significantly compared with those of
hGIF, while the b(MT3)-a(MT1) mutant showed
biochemical properties similar to those of hGIF [23].
Interestingly, bioassay data showed that the b(MT3)-
b(MT3) mutant exhibited reduced activity, while the
b(MT3)-a(MT1) mutant had similar activity, confirm-
ing that the a-domain is not dispensable for the neuro-
nal growth-inhibitory activity of hGIF [23]. Therefore,
we suggest that although the single a-domain does not
exhibit any neuronal growth-inhibitory activity, it does
play an important role in modulating the stability of
the metal–thiolate cluster and conformation of
the b-domain by domain–domain interactions, thus
altering zinc homeostasis in the brain and influencing
the bioactivity [23].
The role of the EAAEAE insert

(DTNB) and S-nitrosocysteine (SNOC), and increases
the stability of the metal–thiolate in the b-domain
[11,16]. Then, we predicted the structure of intact hGIF
and the D55-60 mutant. Unlike MT2, hGIF has a par-
ticular strategy for its own domain–domain interaction
and regulation mechanism. It was found that the Ser33-
Cys38 and Lys31-Glu41 hydrogen bonds observed in
hGIF fell apart in the D55-60 mutant, resulting in the
movement of the a-domain away from the b-domain.
These conformational alterations around the linker
region of the D55-60 mutant also affected the interac-
tion between Lys32 and the b-domain, leading to a con-
formational change of the b-domain [16]. This agreed
well with the results of previous SNOC and DTNB
experiments [11,16]. Thus, it was concluded that dele-
tion of the EAAEAE(55–60) insert in hGIF would
make the A-cluster wrap more tightly and alter the
domain–domain interactions in hGIF through intramo-
lecular interactions, and eventually affect the structural
and dynamic properties of the b-domain by domain–
domain interactions, which would be a reason for the
reduced bioactivity of the D55-60 mutant.
The role of the linker
It was found that the linker connecting the two
domains of MTs is so conserved that in all mammalian
MTs it exists as a KKS sequence (except for MT4, in
which the conservative substitution linker RKS is
found) (Fig. 1). Hence, we constructed three mutants
of hGIF in the link region (the K31 ⁄ 32A mutant, the
K31 ⁄ 32E mutant and the KKS-SP mutant) and

cluster in the
b-domain through domain–domain interactions, and is
thus indispensable for the biological activity of hGIF.
The role of the acid–base catalysis site
of hGIF
It was reported that nitric oxide (NO) reacts with the
thiolate group of MTs under pseudo-first-order condi-
tions, leading to the release of zinc ions [25]. However, it
was found that GIF is significantly more reactive than
MT1 and MT2 towards S-nitrosothiols [26]. Chen et al.
attributed the high activity of GIF towards SNOC to
the unique acid–basic catalysis motif in the b-domain:
KCE(21–23) [16]. To understand the role of the acid–
basic catalysis motif in S-nitrosylation, we constructed
an E23K mutant protein of hGIF by comparing the pri-
mary sequence between hGIF and hMT1g [a human
MT1 isoform where the segment is KCK(20–22)] [27].
Interestingly, it was found that the reaction of the E23K
mutant with SNOC exhibits biphasic kinetics, and the
reaction is much faster than that of hGIF at the initial
step [27]. This result was not anticipated, indicating
that the acid–base motif might not be the only factor
Structure–reactivity–function study of GIF Z C. Ding et al.
2918 FEBS Journal 277 (2010) 2912–2920 ª 2010 The Authors Journal compilation ª 2010 FEBS
contributing to the high activity of hGIF towards
SNOC. Based on the results of CD spectral study, reac-
tion with EDTA and pH titration, the structure and sta-
bility of the metal–thiolate clusters of the E23K mutant,
compared with those of hGIF, are not very different.
The differences between hGIF and its E23K mutant lie

activity of GIF. Other factors include solvent accessibil-
ity and dynamic properties of the metal–thiolate cluster
(especially the metal–thiolate cluster in the b-domain),
which are closely associated with the mutual accessibil-
ity of metal–thiolate clusters with biologically sensitive
small molecules such as NO, thus influencing zinc
homeostasis in the brain. Another factor identified is
domain–domain interactions, which might play impor-
tant roles in modulating the stability of the metal–
thiolate cluster and the conformation of the b-domain.
Comments
The particular reactivity of GIF related to the specific
metal–thiolate cluster has been reviewed by Peter
Faller in the paper entitled ‘Reactivity and structure of
metal-thiolate clusters in growth inhibitory factor’.
Furthermore, Roger S. Chung and coworkers summa-
rized current understandings on the biological func-
tions of GIF in the article entitled ‘A current
evaluation of the biological function of growth inhibi-
tory factor in the injured and neurodegenerative
brain’.
Recently, the putative key role of b-amyloid (Ab)
peptide in the pathogenesis of AD led to a promising
outlook for the treatment of AD. However, the
AN1792 trail report on AD patients showed that
‘although immunization with Ab
42
(AN1792) resulted
in clearance of amyloid plaques in patients with
Alzheimer’s disease, this clearance did not prevent

4814.
2 Binz PA & Kagi JHR (1999) Metallothionein IV,
Metallothionein: Molecular Evolution and Classification.
Birkhause, Baser.
Z C. Ding et al. Structure–reactivity–function study of GIF
FEBS Journal 277 (2010) 2912–2920 ª 2010 The Authors Journal compilation ª 2010 FEBS 2919
3 Palmiter RD (1987) Molecular biology of metallothion-
ein gene expression. Experientia Suppl 52, 63–80.
4 Uchida Y & Tomonaga M (1989) Neurotrophic action
of Alzheimer’s disease brain extract is due to the loss of
inhibitory factors for survival and neurite formation of
cerebral cortical neurons. Brain Res 481, 190–193.
5 Uchida Y & Ihara Y (1995) The N-terminal portion of
growth inhibitory factor is sufficient for biological activ-
ity. J Biol Chem 270, 3365–3369.
6 Sewell AK, Jensen LT, Erickson JC, Palmiter RD &
Winge DR (1995) Bioactivity of metallothionein-3 cor-
relates with its novel beta domain sequence rather than
metal binding properties. Biochemistry 34, 4740–4747.
7 Robbins AH, McRee DE, Williamson M, Collett SA,
Xuong NH, Furey WF, Wang BC & Stout CD (1991)
Refined crystal structure of Cd, Zn metallothionein at
2.0 A resolution. J Mol Biol 221, 1269–1293.
8 Braun W, Vasak M, Robbins AH, Stout CD, Wagner
G, Kagi JH & Wuthrich K (1992) Comparison of the
NMR solution structure and the x-ray crystal structure
of rat metallothionein-2. Proc Natl Acad Sci USA 89,
10124–10128.
9 Wang H, Zhang Q, Cai B, Li H, Sze KH, Huang ZX,
Wu HM & Sun H (2006) Solution structure and

3558.
17 Hasler DW, Jensen LT, Zerbe O, Winge DR & Vasak
M (2000) Effect of the two conserved prolines of human
growth inhibitory factor (metallothionein-3) on its bio-
logical activity and structure fluctuation: comparison
with a mutant protein. Biochemistry 39, 14567–14575.
18 Romero-Isart N, Jensen LT, Zerbe O, Winge DR &
Vasak M (2002) Engineering of metallothionein-3
neuroinhibitory activity into the inactive isoform
metallothionein-1. J Biol Chem 277, 37023–37028.
19 Chung RS, Holloway AF, Eckhardt BL, Harris JA,
Vickers JC, Chuah MI & West AK (2002) Sheep have
an unusual variant of the brain-specific metallothionein,
metallothionein-III. Biochem J 365 , 323–328.
20 Faller P, Hasler DW, Zerbe O, Klauser S, Winge DR &
Vasak M (1999) Evidence for a dynamic structure of
human neuronal growth inhibitory factor and for major
rearrangements of its metal-thiolate clusters. Biochemis-
try 38, 10158–10167.
21 Faller P & Vasak M (1997) Distinct metal-thiolate clus-
ters in the N-terminal domain of neuronal growth
inhibitory factor. Biochemistry 36, 13341–13348.
22 Jiang LJ, Vasak M, Vallee BL & Maret W (2000) Zinc
transfer potentials of the alpha - and beta-clusters of
metallothionein are affected by domain interactions in
the whole molecule. Proc Natl Acad Sci USA 97, 2503–
2508.
23 Ding ZC, Zheng Q, Cai B, Yu WH, Teng XC, Wang
Y, Zhou GM, Wu HM, Sun HZ, Zhang MJ et al.
(2007) Effect of alpha-domain substitution on the struc-

2920 FEBS Journal 277 (2010) 2912–2920 ª 2010 The Authors Journal compilation ª 2010 FEBS