ZBP1 recognition of b-actin zipcode
induces RNA looping
Jeffrey A. Chao,1Yury Patskovsky,2Vivek Patel,1Matthew Levy,2Steven C. Almo,2
and Robert H. Singer1,3
1Department of Anatomy and Structural Biology, Albert Einstein College of Medicine, Bronx, New York 10461, USA;
2Department of Biochemistry, Albert Einstein College of Medicine, Bronx, New York 10461, USA
ZBP1 (zipcode-binding protein 1) was originally discovered as a trans-acting factor for the ‘‘zipcode’’ in the 39
untranslated region (UTR) of the b-actin mRNA that is important for its localization and translational regulation.
Subsequently, ZBP1 has been found to be a multifunctional regulator of RNA metabolism that controls aspects of
localization, stability, and translation for many mRNAs. To reveal how ZBP1 recognizes its RNA targets, we
biochemically characterized the interaction between ZBP1 and the b-actin zipcode. The third and fourth KH
(hnRNP K homology) domains of ZBP1 specifically recognize a bipartite RNA element located within the first 28
nucleotides of the zipcode. The spacing between the RNA sequences is consistent with the structure of IMP1
KH34, the human ortholog of ZBP1, that we solved by X-ray crystallography. The tandem KH domains are
arranged in an intramolecular anti-parallel pseudodimer conformation with the canonical RNA-binding surfaces at
opposite ends of the molecule. This orientation of the KH domains requires that the RNA backbone must undergo
an ~180° change in direction in order for both KH domains to contact the RNA simultaneously. The RNA looping
induced by ZBP1 binding provides a mechanism for specific recognition and may facilitate the assembly of post-
transcriptional regulatory complexes by remodeling the bound transcript.
[Keywords: ZBP1; RNA-binding protein; KH domain; RNA localization]
Supplemental material is available at http://www.genesdev.org.
Received September 10, 2009; revised version accepted November 23, 2009.
Localization of messenger RNA (mRNA) into distinct
subcellular compartments allows for the spatial regula-
tion of gene expression that is required for the establish-
ment and maintenance of cell polarity (for review, see
Martin and Ephrussi 2009). A global study using Dro-
sophila embryos found that >71% of all the mRNAs
characterized (3370 genes) were localized and, further-
more, these mRNA could be grouped into 35 unique
localization patterns (Lecuyer et al. 2007). A second
genome-wide study that isolated mRNAs from fibroblast
cell protrusions identified >50 mRNAs that were specif-
ically localized to pseudopodia (Mili et al. 2008). These
recent studies underscore the fundamental role of mRNA
localization in diverse cellular and developmental pro-
cesses and, while technological advances have increased
the number of mRNAs that been shown to localize, the
underlying mechanisms that give rise to these asymmet-
ric distributions have remained elusive.
The localization of b-actin mRNA to the leading edge
of chicken embryo fibroblasts was one of the earliest
transcripts identified to be subcellularly localized and has
served as a model system for understanding the process
(Lawrence and Singer 1986). Asymmetric sorting of the
b-actin transcript is achieved by transport along both
microtubule and actin microfilaments, and is delocal-
ized in myosin II-B knockout fibroblasts (Latham et al.
1994; Fusco et al. 2003; Oleynikov and Singer 2003). A
54-nucleotide (nt) cis-acting element, termed the zipcode,
positioned directly following the termination codon in
the 39 untranslated region (UTR) of b-actin mRNA was
shownto benecessary andsufficient for targetingreporter
RNA constructs to thecellular periphery (Kislauskis et al.
1994). The trans-acting factor Zipcode-binding protein
1 (ZBP1) was identified based on its ability to interact
with the zipcode, and its knockdown results in impaired
invadopodia formation, cytoplasmic spreading, and cell
adhesion (Ross et al. 1997; Vikesaa et al. 2006).
ZBP1 is the founding member of a highly conserved
family (termed VICKZ in reference to the founding
members:Vg1RBP/Vera, IMP1-3, CRD-BP, KOC, and
ZBP1) of RNA-binding proteins that have been impli-
cated in the post-transcriptional regulation of several
different RNAs (Yisraeli 2005). In Xenopus laevis,
Vg1RBP/Vera is required for the localization of Vg1
mRNA to the vegetal cortex of oocytes and also the
localization of b-actin mRNA in axons (Deshler et al.
E-MAIL firstname.lastname@example.org; FAX (718) 430-8697.
Articleis online at http://www.genesdev.org/cgi/doi/10.1101/gad.1862910.
148 GENES & DEVELOPMENT 24:148–158 ? 2010 by Cold Spring Harbor Laboratory Press ISSN 0890-9369/10; www.genesdev.org
1998; Havin et al. 1998; Leung et al. 2006; Yao et al. 2006).
Humans contain three paralogs (IMP1–3) that were orig-
inally identified because of their ability to regulate in-
sulin-like growth factor II (Igf-II) mRNA translation, but
have since been found to promote the localization of H19
and tau mRNAs as well as stabilize CD44 and b-TrCP1
mRNAs (Nielsen et al. 1999; Runge et al. 2000; Atlas
et al. 2004; Vikesaa et al. 2006; Elcheva et al. 2009). In
was shown to protect the c-myc mRNA from endonu-
cleolytic cleavage, thereby stabilizing the transcript
(Doyle et al. 1998). This family’s ability to broadly
regulate RNA metabolism can lead to adverse cellular
effects, as evidenced by their overexpression and correla-
tion with poor prognosis in several types of cancers
(Hammer et al. 2005; Dimitriadis et al. 2007; Jiang et al.
2008; Kobel et al. 2009).
VICKZ family members share a characteristic arrange-
ment of six canonical RNA-binding modules with two
RNA recognition motifs (RRM) followed by four hnRNP-
K homology (KH) domains. Sequence alignments of the
proteins show that conserved residues are clustered into
three didomains (RRM12, KH12, and KH34), which sug-
gests that these evolutionarily conserved regions may
function in concert (Fig. 1A; Git and Standart 2002).
Interestingly, ZBP1 and IMP1 share considerable se-
quence identity (>94%), yet studies of their interactions
with their respective RNA targets have failed to produce
a unified understanding of the requirements for specific
RNA recognition (Runge et al. 2000; Farina et al. 2003;
Nielsen et al. 2004; Patel and Bag 2006; Atlas et al. 2007;
Jonson et al. 2007). Models for RNA recognition differ
with regard to both the domains and oligomerization
state of ZBP1 required for binding as well as the pro-
posed RNA determinants, which range from a minimal
59-ACACCC-39 sequence to RNA binding being entirely
sequence-independent (Farina et al. 2003; Nielsen et al.
2004; Atlas et al. 2007; Oberman et al. 2007).
Here we present a biochemical characterization of
ZBP1 recognition of the b-actin zipcode RNA. The
ZBP1 KH34 monomer binds to two nonsequential
stretches of RNA located within the proximal portion
of the zipcode. This bipartite recognition element is
consistent with our crystal structure of IMP KH34 (98%
sequence identity with ZBP1 KH34), the first structural
data for this family of proteins, that positions the KH
domains in an anti-parallel arrangement with their puta-
tive RNA-binding surfaces located at opposite ends of the
molecule. This orientation of the KH domains explains
both the sequence and distance dependence of RNA
binding that were determined biochemically. The resi-
dues that link KH3 to KH4 were also shown to play a role
in RNA binding, demonstrating that the KH34 domain
functions as a single unit whose precisegeometry dictates
its interaction with RNA.
ZBP1 recognition of zipcode RNA
We took the interaction between ZBP1 and the first 54 nt
of the 39 UTR of b-actin mRNA (zipcode[1–54]) as a start-
ing point to further investigate both the protein and RNA
contributions to specific recognition. A polyacrylamide
gel electrophoretic mobility shift assay (EMSA) was used
to resolve fluorescein-labeled zipcode[1–54] in complex
responsible for recognition of zipcode[1–54]
RNA. (A) Schematic diagram of ZBP1
showing conserved didomain organization.
(B) Representative EMSA results for full-
length ZBP1, RRM12, KH12, and KH34
binding to zipcode[1–54] RNA. The filled
triangle represents a 1:1 serial dilution of
recombinant protein. Free RNA (*) and
RNA–protein complexes (**) are labeled.
(C) Quantification of the fraction of RNA
bound in EMSA data for ZBP1 and KH34
were fit to the Hill equation to measure
the Kd, appand Hill coefficient.
The KH34 didomain of ZBP1 is
ZBP1 recognition of zipcode
GENES & DEVELOPMENT 149
with ZBP1 from free RNA. The fraction of bound RNA
was measured as a function of protein concentration, and
the affinity of recombinant ZBP1 for the zipcode[1–54]
was determined by fitting the data to the Hill equation
(Ryder et al. 2008). ZBP1 binds tightly to this RNA with
an apparent disassociation constant (Kd, app) of 7.3 6 3.0
nM and a Hill coefficient of 0.9 6 0.2, which is consistent
with previous studies of the ZBP1–zipcode[1–54] interac-
tion measured by nitrocellulose filter binding (Fig. 1B,C;
Farina et al. 2003).
In order to determine which of the putative RNA-
binding domainsof ZBP1 were responsible for recognition
of the zipcode[1–54], truncations of ZBP1 that contain
either RRM12, KH12, or KH34 were generated. Both
ZBP1 RRM12 and KH12 do not bind the zipcode[1–54]
with high affinity (Fig. 1B). ZBP1 KH34, however, binds
the zipcode[1–54] with a similar affinity as the full-length
protein (Kd, app= 4.7 6 2.0 nM, Hill coefficient of 0.9 6
0.2), indicating that recognition of the zipcode is de-
pendent on this domain. Recombinant ZBP1 constructs
that contained the individual KH3 or KH4 domains did
not bind the zipcode[1–54] with high affinity, which is
consistent with previous studies of isolated KH domains
having affinities for their targets in the micromolar range
(Supplemental Fig. 1; Valverde et al. 2008). These exper-
iments demonstrate that the two KH domains function
together for high-affinity RNA binding.
Interestingly, at higher protein concentrations of both
full-length ZBP1 and the KH34 domain alone, higher-
molecular-weight complexes can be resolved by EMSA
(Fig. 1B). While previous experiments performed with the
human and frog homologs of ZBP1 (IMP1 and Vg1RBP)
concluded that KH34 contains a dimerization motif, we
find that both full-length ZBP1 and the KH34 domain
exist predominantly as monomeric species in solution, as
assayed by size exclusion chromatography at concentra-
tions (>20 mM) above those used in the EMSA experi-
ments (Supplemental Fig. 2; Git and Standart 2002;
Nielsen et al. 2004; Oberman et al. 2007). Furthermore,
binding to the zipcode was not found to be cooperative
(Hill coefficient, ;1) for both full-length ZBP1 and the
KH34 domain, indicating that a potential ZBP1 dimer is
not stabilized by RNA binding. Based on these data, we
hypothesize that the higher-molecular-weight complexes
likely result from ZBP1 monomers binding indepen-
dently to alternative lower-affinity binding sites within
the zipcode[1–54]. Previous studies of isolated KH do-
mains have found that their RNA-binding sites are
usually comprised of four consecutive single-stranded
nucleotides, so it seems plausible that the 54-nt zipcode
element could contain multiple binding sites for ZBP1
KH34 (Auweter et al. 2006; Valverde et al. 2008).
Identification of the ZBP1 KH34 RNA-binding site
within the zipcode
Since the zipcode[1–54] was found to potentially contain
more than one binding site for ZBP1 KH34, fragments of
thisRNAwere synthesized to identify the positionsof the
nucleotides recognized by ZBP1. A fragment of the
zipcode containing nucleotides 1–28 (zipcode[1–28]) was
bound by ZBP1 KH34 with similar affinity (Kd, app= 3.6 6
0.2 nM, Hill coefficient of 0.9 6 0.2) as the entire
zipcode[1–54] sequence, indicating that the high-affinity
binding site is located within this region (Fig. 2A).
Importantly, even at the highest protein concentration
(500 nM), only a single RNA–protein complex was
observed, providing a more tractable system for further
biochemical characterization. A second ZBP1-binding
site within the zipcode[1–54], however, could not be
identified by EMSA. Experiments using a zipcode frag-
ment containing nucleotides 29–54 of the 39 UTR of
b-actin mRNA (zipcode[29–54]) did not result in a well-
resolved shifted complex (Fig. 2C). Perhaps in the context
ing to zipcode[1–28] RNAwith fit of data to the Hill equation. (B) Representative stoichiometry binding assay for ZBP1 KH34 binding to
zipcode[1–28] with fit to quadratic model of saturable ligand binding. (C–E) Representative EMSA results for ZBP1 KH34 binding
to zipcode[29–54], zipcode[5–44], and zipcode[1–21] RNAs. The filled triangle represents a 1:1 serial dilution or 3:1 serial dilution
(for stoichiometry experiment) of recombinant protein.
Binding site for ZBP1 KH34 is within the first 28 nt of the zipcode. (A) Representative EMSA results for ZBP1 KH34 bind-
Chao et al.
150GENES & DEVELOPMENT
of a shorter RNA, the lower-affinity binding site within
the zipcode is beyond the detection limit of the EMSA.
Further experiments will be required to determine if
these secondary interaction sites have any biological
significance, or if ZBP1’s function is mediated by only
the high-affinity binding site.
To further clarify the composition of the RNA–
protein complex, the stoichiometry of the ZBP1 KH34–
zipcode[1–28] interaction was determined by EMSA,
where the concentration of the unlabeled zipcode[1–28]
fragment (250 nM) was fixed well above the measured
Kd, app(3.6 nM) for this interaction. The data were fit to a
quadratic model of saturable ligand binding, resulting in
a stoichiometric equivalence point of 1.0 6 0.1 (Fig. 2B;
was also found to have a stoichiometry of 1:1 by size
exclusion chromatography (Supplemental Fig. 2). Taken
as a whole, the RNA-binding and stoichiometry experi-
ments, in conjunction with the size exclusion chroma-
tography data, demonstrate that ZBP1 KH34 can bind
RNA sequences within the zipcode as a monomer, and
that multimerization is not required for stable RNA
recognition, as suggested previously (Nielsen et al. 2004).
To better define the interaction between ZBP1 KH34
and the zipcode[1–28] RNA, fragments that contained de-
letions at the 59 end (zipcode[5–44]) and 39 end (zipcode[1–
21]) of zipcode[1–28] were used in the EMSA. Both of
these RNAs failed to bind ZBP1 KH34 with high affin-
ity, demonstrating that nucleotides at both the 59 and 39
ends of the zipcode[1–28] are necessary for recognition
(Fig. 2D,E). Interestingly, nucleotides 16–21 and 22–27 of
the zipcode share an identical nucleotide sequence
(ACACCC), yet zipcode[1–21] could not bind to ZBP1
KH34. This suggests that the spacing between the nucle-
otides with which ZBP1 KH34 interacts at the 59 and 39
ends of zipcode[1–28] is an important factor for recogni-
tion. It should also be noted that zipcode[5–44] (40 nt) is
longer than the zipcode[1–28] (28 nt), indicating that
ZBP1 KH34 binding to RNA is not simply length-de-
pendent, but requires some component of RNA sequence
ACACCC ‘‘motif’’ is not sufficient for ZBP1 recognition,
although a subset of its nucleotides may form the
bindings site for one of the KH domains.
The previous experiments implicated nucleotides at
both the 59 and 39 ends of zipcode[1–28] as potentially
being recognized by ZBP1 KH34, but the identity of these
nucleotides could not be determined. To better under-
stand the sequence requirements for recognition, a selec-
tion for binding to ZBP1 KH34 was performed using
a degenerate RNA library where, at each position within
zipcode[1–28] sequence, the nucleotide’s identity had the
probability of being 85% wild-type and 5% of each non-
wild-type nucleotide. After three rounds of selection and
amplification, the RNA population was cloned and se-
quenced. This approach identified two stretches of nucle-
otides—59-GGACU-39 (4–8) and 59-ACA-39 (22–24)—that
were highly conserved compared with the initial popula-
tion of RNAs that retained binding to ZBP1 KH34 (Fig.
3A). The 39 UTRs of human b-actin mRNA also share
demonstrates that the
these sequence elements, but are spaced 2 nt further apart
than in the chicken zipcode. Mutation of either one of
these sequences resulted in a dramatic reduction in RNA
binding (Fig. 3B,C). In comparison, one cloned RNA that
harbors seven point mutations binds to ZBP1 KH34 as
well as the wild-type zipcode[1–28] sequence (Fig. 3D). In
this particular clone, the 59 RNA element is not mutated
and the 39 RNA element has been shifted 2 nt upstream.
While further biochemical characterization will be re-
quired to determine the precise nucleotides that form the
consensus RNA-binding site of ZBP1 KH34, the results of
this study support a model of recognition where the
individual KH3 and KH4 domains each bind to unique
nonsequential sequences within zipcode[1–28].
Overall structure of IMP1 KH34
The KH34 domain of IMP1, the human ortholog of ZBP1,
shares 98% sequence identity (four amino acid substitu-
tions: M407T, Q409H, I455V, and D533E) with ZBP1
KH34 and behaves identically in RNA-binding experi-
ments using the zipcode[1–54], suggesting that this do-
main has been structurally and functionally conserved in
vertebrates (Supplemental Fig. 3). Hexagonal crystals of
IMP1 KH34 were obtained that diffract to 2.75 A˚res-
olution, and the structure was determined by molecular
replacement (Table 1). There are three molecules of IMP1
identifies two RNA elements required for recognition. (A) Se-
quence of zipcode[1–28] that was used to generate a degenerate
RNA library that was designed to be 85% wild type. EMSA of
ZBP1 KH34 binding to RNA pool that resulted from three
rounds of selection. (B,C) Mutation of either conserved RNA
element reduces affinity for ZBP1 KH34. (D) One clone isolated
from the degenerate selection that harbors seven mutations
binds as well as the wild-type zipcode[1–28] RNA. The filled
triangle represents a 1:1 serial dilution of recombinant protein.
Selection for ZBP1 KH34 binding to doped library
ZBP1 recognition of zipcode
GENES & DEVELOPMENT 151
KH34 within the asymmetric unit and the independent
copies are almost identical, having a root-mean-square
deviation (RMSD) of 0.35 A˚for all 157 Ca. Interestingly,
two of the three IMP1 KH34 domains pack against
symmetry-related molecules via interactions between
their second and third a helices in a manner similar to
what was previously observed for Nova KH3 (Lewis et al.
1999). These arrangements, however, are likely an arti-
fact of crystallization and do not represent a stable di-
merization, based on calculations of the free energy of
disassociation (Krissinel and Henrick 2007). While we
cannot exclude the possibility that such an interaction
may form in the context of an RNA granule, the structure
of IMP1 KH34 is consistent with our biochemical data
showing that this domain recognizes RNA as a monomer.
In the structure of IMP1 KH34, both KH3 and KH4
adopt the b1a1a2b2b3a3topology seen in other type I
eukaryotic KH domains. The KH3 and KH4 domains have
the classical KH fold, with three a helices that pack
against one face of anti-parallel b sheet formed by the
three b strands (Fig. 4A). The conserved GXXG loop,
which connects a1and a2, is formed by residues 422–425
(GKKG) for KH3 and residues 504–507 (GKGG) for KH4,
and the variable loop, located between b2 and b3, is
formed by residues 442–450 for KH3 and residues 524–
543 for KH4. A short linker (residues 479–487) connects
a3of KH3 to b1of KH4. The structures of the KH3 and
KH4 domains relative to each other are similar, with an
RMSD of 1.0 A˚for 69 Ca, with the largest differences
located in the variable loops.
IMP1 KH34 forms an intramolecular anti-parallel
The KH3 and KH4 domains of IMP1 KH34 form an
intramolecular pseudodimer (Fig. 4A). The KH domains
are arranged in an anti-parallel manner that forms an
extended six-stranded b-sheet surface. This interaction is
stabilized by a number of residues located on the b1
strands and a3helices of both KH domains, and also by
residues located within the linker between KH3 and KH4
(Fig. 4B). Formation of the pseudodimer buries a large
combined buried surface area (;2250 A˚2), suggesting that
this arrangement is the functional conformation of IMP1
KH34 in vivo.
There are several large hydrophobic residues located
within this intramolecular interface that likely contrib-
ute to its stability (Fig. 4B). Alternating residues located
Crystallographic data and structure refinement
a, b, c
a, b, g
103.5 A˚, 103.5 A˚, 131.6 A˚
90°, 90°, 120°
53.0 A˚–2.75 A˚(2.90 A˚–2.75 A˚)
Number of reflections
Number of atoms
50.0 A˚–2.75 A˚
Values in parentheses are for highest-resolution shell.
dodimer. (A) KH3 (red) and KH4 (blue) with
N and C termini and secondary structures
labeled. The anti-parallel arrangement of
KH3 and KH4 places the putative RNA-
binding surfaces on opposite ends of the
molecule. (B) Detailed view of intramolec-
ular interface. The pseudodimer is stabi-
lized by the packing of several hydrophobic
residues as well as hydrogen bonds and
Structure of IMP1 KH34 pseu-
Chao et al.
152GENES & DEVELOPMENT
on the a3helices of KH3 (Phe464 and Tyr471) and KH4
(F543) form the crux of the interaction that packs against
a hydrophobic platform comprised of residues on the b1
strands of KH3 (Val8 and Val10) and KH4 (L488 and
Thr490). This hydrophobic surface is extended into the
linker by Val486 and Phe479, as well as a second phenyl-
alanine in the linker (Phe478) that packs against a1 of
While hydrophobic interactions likely dominate for-
mation of the pseudodimer, hydrogen bonds and electro-
static interactions may also contribute to its stability.
The anti-parallel arrangement of the KH3 and KH4 b1
strands forms two interdomain hydrogen bonds between
backbone amide and carbonyl groups (Gln409/His491
and Phe411/Glu489). Potential hydrogen bonds are also
formed by two opposing glutamine residues on the a3
helices of KH3 (Gln467) and KH4 (Gln550) and the e
amino group of Lys475 with the backbone carbonyl of
Glu484. Favorable electrostatic interactions may also be
mediated by the guanidino group of Arg554 with the side
chain carboxylate group of Glu406.
Implications of IMP1 KH34 pseudodimer for RNA
The pseudodimer arrangement of IMP KH34 positions
the putative RNA-binding surfaces of KH3 and KH4 at
opposing ends of the molecule, with a distance of ;50 A˚
between the GXXG loops of KH3 and KH4 (Fig. 5A).
Furthermore, the RNA-binding surfaces are oriented in
opposite directions, which would require the RNA back-
bone to undergo almost a 180° change in direction in
order to interact with both KH domains simultaneously
as predicted by the biochemical data. Modeling of the
KH34–RNA complex, based on structures of other KH
domains bound to their nucleic acid targets, positions the
59 and 39 ends of the RNA;35 A˚apart (Backe et al. 2005;
Du et al. 2005, 2007). This distance could be spanned by
a minimum of 5–6 nt; however, such a tight looping of the
RNA is likely to be energetically unfavorable and would
require the RNA elements recognized by the KH34
domain to be separated further.
To better determine the number of nucleotides that can
separate the two halves of the bipartite RNA recognition
element, nucleotides were systematically deleted from
the zipcode[1–28] and their affinity for ZBP1 KH34 was
measured by EMSA. Deletion of 2 and 4 ntdid not have an
appreciable effect on these RNAs’ ability to interact with
ZBP1 KH34 (Fig. 5B). However, when 6 nt were deleted
from zipcode[1–28], the affinity of this RNA for ZBP1
KH34 was reduced by almost an order of magnitude, and
deletion of 8 nt completely abolished RNA binding. This
precipitous drop in RNA-binding affinity is consistent
with the positioning of the KH domains in the structure
of IMP1 KH34. While an upper limit for binding has not
been determined, if the spacing becomes too large, the
RNA elements will likely function independently, which
will also lead to a reduction in apparent affinity.
Conservation of pseudodimer in VICKZ KH34 family
The high degree of sequence identity (>80%) and conser-
vation of hydrophobic residues at the intramolecular
interface between IMP1 KH34 and its homologs suggests
that all VICKZ family members adopt the same pseudo-
dimer arrangement. A crystal structure of IMP2 KH34
confirms that the organization of the KH domains is
conserved for this particular IMP1 paralog (J Chao,
Y Patskovsky, S Almo, and R Singer, in prep.). Conse-
quently, the bipartite RNA recognition element is also
likely to be conserved between VICKZ family members,
although it remains to be determined if the KH domains
of individual family members have distinct RNA se-
quence preferences. Although amino acid differences
between the homologs do not cluster around the putative
RNA-binding surfaces, preliminary data show that the
IMP1–3 paralogs do not recognize the zipcode[1–54] with
the same affinity, suggesting that the family members are
not functionally redundant (Supplemental Fig. 3).
Structural homology with other tandem KH domains
Despite the existence of numerous RNA-binding proteins
with multiple KH domains, most of the structural data
related to the function of these proteins have been
gleaned from studies of isolated KH domains. There are
only three structures of tandem type I KH domains that
have been reported to date: far-upstream element-binding
spacing between bound RNA sequences. (A) The puta-
tive RNA-binding surfaces of KH3 (red) and KH4 (blue)
are shown with modeled RNA tetranucleotide (ma-
genta). Anti-parallel arrangement of KH domains re-
quires RNA backbone to undergo an ;180° change in
direction to interact with both domains at the same
time, which induces looping of the RNA. (B) Systematic
deletion of nucleotides within zipcode[1–28]. Changes
in RNA-binding affinity are shown relative to zip-
Model of RNA binding by IMP1 KH34 and
ZBP1 recognition of zipcode
GENES & DEVELOPMENT153
protein (FBP) KH34 in complex with ssDNA, fragile X
mental retardation protein (FMRP) KH12, and poly-
C-binding protein 2 (PCBP2) KH12 (Braddock et al.
2002; Valverde et al. 2007; Du et al. 2008). In the case of
FBP KH34, nuclear magnetic resonance (NMR) experi-
ments found that the two KH domains, which are
separated by a 30-residue glycine-rich linker, function
independently (Braddock et al. 2002). A similar situation
was observed by NMR for the third and fourth KH
domains of K-homology splicing protein (KSRP), although
a full structure determination of the tandem domains was
not performed (Garcia-Mayoral et al. 2007). In the X-ray
crystal structure of FMRP KH12, which has a single
residue linking KH1 to KH2, the KH domains are posi-
tioned at ;60° relative to one another, resulting in a
limited intramolecular interaction surface between the
two KH domains (Valverde et al. 2007). PCBP2 KH12,
however,has a 14-amino-acid linker thatconnects its two
KH domains and adopts an anti-parallel pseudodimer
topology similar to IMP1 KH34 (Du et al. 2008).
IMP1 KH34 and PCBP2 KH12 share only 29% sequence
identity, yet adopt similar overall structures that have an
RMSD of 3.1 A˚for 149 Ca(Fig. 6A). Interestingly, only one
aromatic residue (Phe464-IMP1, Phe69-PCBP2) is con-
served between IMP1 KH34 and PCBP2 KH12 at the
pseudodimer interface, suggesting that this arrangement
can be stabilized by multiple complementary interac-
tions. Also, few residues involved in RNA binding are
conserved between IMP1 KH34 and PCBP2 KH12, and
only Arg57 that makes a cytosine-specific contact to the
second nucleotide in the PCBP2 KH1–RNA complex is
conserved in IMP1 KH3 (Arg452) (Du et al. 2007). In-
terestingly, this base-specific readout by an arginine
residue at this position is also found in the complexes
between hnRNP K KH3 and Nova-2 KH3 with their
nucleic acid-binding partners (Lewis et al. 2000; Backe
et al. 2005). This suggests that IMP1 KH3 may recognize
the 39 RNA element of zipcode[1–28], but the lack of
sequence conservation makes homology modeling diffi-
cult, and it is unclear if this particular interaction is
maintained between these structures.
Role of conserved linker between KH3 and KH4
While IMP1 KH34 and PCBP2 KH12 share a similar
organization of their respective KH domains, the linkers
that connect the domains adopt very different conforma-
tions (Fig. 6A). In the PCBP2 KH12 structure, the linker
contains a number of polar residues and its structure is
poorly defined due to lack of experimental NMR re-
straints, indicating that this region is more dynamic than
the rest of the molecule (Du et al. 2008). In contrast, the
linker in the IMP1 KH34 structure forms a compact
structure that helps to stabilize the pseudodimer inter-
face by the van der Waals interactions made by the two
large aromatic residues (Phe479 and Phe480). The linker
residues are highly conserved in the VICKZ family of
RNA-binding proteins, which has led previously to the
suggestion that this region may be functionally impor-
tant (Fig. 6B; Git and Standart 2002).
To determine if this linker has a functional role,
a mutant ZBP1 KH34 protein was generated that replaced
the linker residues with the amino acids that connect
KH1 and KH2 within ZBP1 (Fig. 6B). This mutation only
changes the identity of the residues in the linker, but not
its length. The RNA-binding ability of this ZBP1 KH3–
L12–KH4 mutant protein was measured by EMSA using
the zipcode[1–54] RNA. ZBP1 KH3–L12–KH4 mutant
binds to the zipcode[1–54] considerably weaker (Kd, app
; 500 nM) than ZBP1 KH34 (Kd, app= 3.6 nM) (Figs. 2A,
6C). This reduction in RNA-binding affinity could arise
from a destabilization of the KH34 pseudodimer, indicat-
ing that the precise orientation of the domains is required
for RNA recognition, or residues in the linker could make
direct interactions with the zipcode that were lost in the
ZBP1 KH3–L12–KH4 mutant.
In order to see if this linker mutation could affect the
physiology of ZBP1 in cells, the mutation was generated
within the full-length protein and fused to the C terminus
of green fluorescent protein (GFP). Wild-type ZBP1, when
fused to GFP, was found to form large cytoplasmic
granules when expressed in mouse embryonic fibroblasts
(MEFs), similar to what was observed in previous studies
(Fig. 6D; Nielsen et al. 2002; Farina et al. 2003). The GFP-
ZBP1 L34 linker mutant protein, however, exhibits
a diffuse cytoplasmic appearance and reduction of granule
formation, indicating that the impaired RNA-binding
affinity observed in vitro alters the subcellular localiza-
tion of ZBP1 in vivo (Fig. 6D). While the precise function
this linker has remains to be determined, the data
presented here show that residues located outside of the
in RNA binding and granule formation. (A) IMP1 KH34 and
PCBP2 KH12 adopt similar structures, with the largest de-
viation coming from the residues that link the KH domains.
(B) Sequence alignment of amino acids that connect KH3 and
KH4 from ZBP1, IMP1-3, and VgRBP1 shown with ZBP1 KH12
linker. (C) Mutation of KH34 linker to residues in KH12 linker
reduces RNA binding to zipcode[1–54]. The filled triangle
represents a 1:1 serial dilution of recombinant protein. (D)
Expression of GFP-ZBP1 with mutant KH34 linker results in
a change in subcellular localization of the fusion protein from
granular to diffuse cytoplasmic.
Conserved residues that link KH3 and KH4 function
Chao et al.
154 GENES & DEVELOPMENT
canonical KH domain RNA-binding surface can affect
RNA recognition, and that the ZBP1 KH34 domains acts
as a single functional unit.
The architecture of the KH domain only allows for the
recognition of short stretches of nucleic acid with rela-
tively weak binding affinity. Therefore, many nucleic
acid-binding proteins contain multiple copies of these
domains in order to increase both affinity and specificity.
Furthermore, these domains can be arranged in precise
orientations that can provide additional restrictions on
their interacting partnersand, also, insight into themech-
anism by which these proteins accomplish their biolog-
The third and fourth KH domains of ZBP1/IMP1
function as a single module by adopting an anti-parallel
pseudodimer arrangement that places the putative RNA-
binding surfaces on opposite ends of the molecule. This
structural organization of the KH domains allows the
protein to recognize its targets through sequence-specific
contacts distributed over two stretches of RNA that must
be separated by a certain number of nucleotides. By
fashioning the KH3 and KH4 domains together in this
orientation, highly specific RNA binding can be achieved
from two low-affinity interactions.
While the precise identity of the nucleotides that
comprise the consensus bipartite RNA-binding site re-
mains to be determined, a model describing the interac-
tion between KH34 and its target RNA can be proposed.
Due to the ambiguity of the orientation of the KH
domains to the RNA, two distinct models can be con-
structed that differ in the position of the extruded
nucleotide loop, but maintain the same polarity of the
RNA to the KH domains (Fig. 7A,B). Currently, it is
unclear which of the models is correct, and whether or
not residues outside of the canonical RNA-binding sur-
faces can participate in discriminating between the two
modes of binding. If both models are possible, then, in
principle, the position of the two RNA elements within
the transcript can also be swapped. Further experiments
will be required to clarify these features of the RNA–
protein interface, and whether or not the other domains
of ZBP1 can influence proper RNA target recognition.
A consequence, however, of both models is that bind-
ing of KH34 to the RNA will induce looping of the
transcript. This conformation of the RNA brings se-
quences that are distant from one another in the tran-
script into closer proximity, which may create the RNA-
binding sites for additional factors (Fig. 7C). In this way,
the VICKZ family of RNA-binding proteins may nucleate
the assembly of higher-order complexes without directly
interacting with other proteins. This result may help to
explain the lack of direct protein–protein partners for
IMP1 identified by immunopurification and mass spec-
trometry characterization of IMP1-containing granules
(Jonson et al. 2007; Weidensdorfer et al. 2009). RNA
looping has also been proposed to occur when polypy-
rimidine tract-binding protein 1 (PTB1) binds to the
g-aminobutyric acid–g2 pre-mRNA, and is thought to
contribute to PTB1’s function in repressing the splicing of
alternative exons (Oberstrass et al. 2005). We propose that
this feature may be a hallmark of RNA-binding proteins
that play a primary role in the regulation of RNA–protein
complexes by remodeling the transcript so that additional
factors can assemble.
Materials and methods
Full-length ZBP1 and truncations (RRM12 1–157, KH12 195–
354, KH34 404–561, KH3 404–478, KH4 487–576, and KH3–L12–
KH4 404–561 with 478–486 replaced by 268–276) were cloned by
PCR into a derivative of pMalc (New England BioLabs) that
contains a Tobacco Etch virus (TEV) protease site after the
maltose-binding protein (MBP). A C-terminal His6tag was added
by PCR to both ZBP1 full-length and KH34 constructs to ensure
purification of the intact fusion proteins. These constructs were
transformed into Escherichia coli strain Rosetta2 (EMD Bio-
sciences), and recombinant protein was induced with 1 mM
IPTG for 4 h at 37°C. Cell pellets were resuspended in lysis buffer
(50 mM Tris at pH 7.5, 1.5 M NaCl, 1 mM EDTA, 1 mM DTT)
supplemented with one Complete EDTA-free protease inhibitor
tablet (Roche), and were lysed by sonication. Cell debris was
removed by centrifugation, and the soluble fusion protein was
purified by amylose affinity chromatography (New England
BioLabs) followed by either TALON affinity (Clontech) or anion
exchange (GE Healthcare) chromatography. Protein concentra-
tions were calculated by measuring the absorbance at 280 nm
and using extinction coefficients determined by ProtParam
(Gasteiger et al. 2005).
IMP1 KH34 (404–566) was cloned by PCR into pET22HTas an
N-terminal His6tag followed by a TEV cleavage site (Chao et al.
2005). This construct was expressed and purified using condi-
tions identical to those described for MBP–ZBP1 except that
His6-IMP1 KH34 was lysed in a buffer compatible with IMAC
ZBP1 KH34 domain and the assembly of RNA–protein com-
plexes. (A,B) Current data allows two distinct modes of RNA
binding to be modeled based on which sequences within the
bipartite RNA element the individual KH domains interact
with. Polarity of the RNA to the KH domains is conserved in
both models. (C) Looping of the transcript may form the binding
sites for additional RNA-binding proteins and nucleate the
assembly of larger RNPs.
Model for RNA looping induced by binding of the
ZBP1 recognition of zipcode
GENES & DEVELOPMENT155
(50 mM sodium phosphate, 1.5 M NaCl, 10 mM imidazole).
TALON affinity chromatography was used as the first step of
purification followed by TEV cleavage. A second TALON affinity
chromatography step was performed to remove His6-TEV pro-
tease and uncleaved recombinant His6-IMP1 KH34.
The complexes between recombinant ZBP1 constructs and
zipcode RNA fragments were monitored by an EMSA. Zipcode
RNA fragments with fluorescein modifications were prepared by
chemical synthesis (Dharmacon) and deprotected, lyophilized,
and stored according to the manufacturer’s protocol. The se-
quences of the RNAs used in these experimentswere 1–54, 59-AC
AAAACCCAUAAAUGC-Fl-39; 1–28, 59-Fl-ACCGGACUGUUA
CCAACACCCACACCCC-39; 29–54, 59-Fl-UGUGAUGAAACA
AAACCCAUAAAUGC-39; 1–21, 59-Fl-ACCGGACUGUUACC
AACACCC-39; and 5–44, 59-Fl-GACUGUUACCAACACCCAC
ACCCCUGUGAUGAAACAAAAC-39. RNAs (100 pM) were
equilibrated with a twofold serial dilution of recombinant pro-
teins in a buffer containing 10 mM Tris (pH 7.5), 100 mM NaCl,
0.1 mM EDTA, 0.01 mg mL?1tRNA, 50 mg mL?1heparin, and
0.01% IGEPAL CA630 for approximately 3 h in order to ensure
the binding reaction reached equilibrium. Protein–RNA com-
plexes were resolved from unbound RNA by native poly-
acrylamide gel electrophoresis (5% [w/v] 29:1 acrylamide/Bis-
acrylamide, 0.53 TBE) run at 80 V at 4°C. Gels were scanned
using a fluorescent gel imager (Typhoon 9400, GE Healthcare)
with 488 nm excitation and 520 BP 40 filter. The fraction of
bound RNA was determined by either the disappearance of the
free RNA or a ratio of the free and bound RNA, and then fit to
a modified version of the Hill equation as described previously to
determine the apparent dissociation constant (Ryder et al. 2008).
Stoichiometry experiments were performed similarly as de-
scribed above except that the binding reaction contained un-
labeled zipcode[1–28] RNA at 250 nM, and MBP–ZBP1 KH34
was serial-diluted 3:1 in order to obtain more data points in the
transition from free RNA to bound. The data were fit to
a quadratic model of saturable ligand binding resulting in de-
termination of the stoichiometric equivalence point (Rambo and
Analytical size exclusion chromatography
The apparent molecular weights of ZBP1, ZBP1 KH34, and
the ZBP1 KH34–zipcode[1–28] complex were determined using
a HiLoad 16/60 Superdex 200 column (GE Healthcare) gel
filtration column. Recombinant protein samples (;20 mM) were
loaded onto a column equilibrated in a buffer of 50 mM Tris (pH
7.5), 300 mM NaCl, and 1 mM DTT, and run at 1 mL min?1. The
ZBP1 KH34–zipcode[1–28] complex (;100 mM) was loaded onto
a column equilibrated in a buffer containing 50 mM Tris (pH 7.2)
and 150 mM NaCl and run at 1 mL min?1. The A280 absorbance
was monitored to determine the retention time of the protein,
and the apparent molecular weight was estimated by comparison
with known protein standards: g-globulin (158 kDa), ovalbumin
(44 kDa), myoglobin (17 kDa), and vitamin B12(1.35 kDa) (Bio-
RNA selection using degenerate zipcode[1–28] library
An antisense degenerate zipcode[1–28] library with the sequence
was prepared by chemical synthesis (IDT). Italicized nucleotides
indicate positions that were doped at 85% wild-type and 5% of
each non-wild-type nucleotide. The initial DNA library was gel-
purified by denaturing polyacrylamide gel electrophoresis (8%
[w/v] 29:1 acrylamide/Bis-acrylamide, 0.53 TBE), and was eluted
by crush and soak in 300 mM NaCl. The library (200 pmol) was
converted to dsDNA by reverse transcription (Invitrogen) using
the 41.30 primer (59-GATAATACGACTCACTATAGGGAATG
GATCCACATCTACGA-39). The RNA pool used for selection
was generated by T7 transcription (Ambion) of the DNA library,
gel-purified, and resuspended in TE. In the first round of se-
lection, MBP–ZBP1 KH34 (;10 nM) was bound to amylose resin
and then equilibrated with the RNA pool (;75 nM) for 1 h in
a buffer containing 10 mM Tris (pH 7.5), 200 mM NaCl, 0.1 mM
EDTA, and 0.01 mg mL?1tRNA. The binding reaction was
transferred to a 0.45-mm centrifugal filter tube (Millipore) and
washed extensively (10 mM Tris at pH 7.5, 200 mM NaCl) before
elution with wash buffer supplemented with 10 mM maltose.
After phenol/chloroform extraction, the RNA was converted to
cDNA by reverse transcription using the 24.30 primer (59-AAGC
TTCGTCAAGTCTGCAGTGAA-39). The resulting cDNA li-
brary was amplified by PCR using the 41.30 and 24.30 primers
and then transcribed into RNA for the next round of selection.
Subsequent rounds of selection were performed similarly, except
that a negative selection step was included by incubating the
RNA pool with amylose resin in the absence of protein to
remove any RNAs with nonspecific affinity for the amylose
resin. RNA from the third round of selection was fluorescein-
labeled, and its affinity for MBP–ZBP1 KH34 was quantified by
EMSA (Pagano et al. 2007).
Cell culture, transfection, and microscopy
GFP-ZBP1 with the mutant L12 linker between KH3 and KH4
was generated by PCR by ligating overlapping fragments con-
taining the mutation. This GFP-ZBP1 L34 mutant and GFP-
ZBP1 wild type were transfected into MEFs that were seeded
onto coverslips using FuGENE 6 (Roche). After transfection, the
MEFs were incubated at 37°C in Dulbecco’s modified Eagle’s
medium (DMEM; HyClone) supplemented with 10% fetal bo-
vine serum (FBS) for 24 h prior to washing with PBS and fixation
using 4% paraformaldehyde (PFA) in PBS for 20 min. Cells were
washed twice with PBS prior to mounting onto glass slides using
ProLong Gold Anti-fade reagent (Invitrogen). Cells were visual-
ized on a BX61 microscope (Olympus) with an UPlanApo 1003,
1.35NA objective (Olympus) coupled to an X-Cite 120 PC metal-
halide light source (EXFO Life Science) and filter sets for
GFP (Chroma Technology). Digital images were collected with
a CoolSNAP HQ camera (Roper Scientific) using IPLab (Win-
dows version 4.0, BD Biosciences).
Crystallization and structure determination of IMP1 KH34
IMP1 KH34 (0.5 mM) was crystallized using sitting-drop vapor
diffusion at 22°C by mixing equal volumes of the protein and
reservoir solution (1.65 M ammonium citrate, 100 mM Tris at pH
7.8). Crystals were cryoprotected by soaking in reservoir solution
supplemented with 25% glycerol before flash-cooling in liquid
nitrogen. Data were collected to 2.75 A˚resolution from a single
crystal at the National Synchotron Light Source X29a beamline at
a wavelength of 0.979 A˚. The diffraction data were indexed,
integrated, and scaled using MOSFLM and the CCP4 suite of
programs (Collaborative Computational Project, Number 4 1994).
The structure of IMP1 KH34 was determined by molecular
replacement with Phaser using the structure of IMP2 KH34
(J Chao, Y Patskovsky, S Almo, and R Singer, in prep.) as a search
Chao et al.
156GENES & DEVELOPMENT
model (Read 2001). The crystal contained three independent
copies of IMP1 KH34 in the asymmetric unit. Rounds of re-
finement and model building were carried out with Phenix and
Coot (Adams et al. 2002; Emsley and Cowtan 2004). Noncrys-
tallographic symmetry (NCS) restraints were applied during all
stages of refinement, and TLS-refinement was performed after
initial rounds of refinement with domains corresponding to KH3
and KH4. Protein stereochemistry was checked using Molpro-
bity (Davis et al. 2007). The final model contained residues 405–
562 for chain A, 405–481; 484–562 for chain B and 405–482; 484–
562 for chain C. The side chain of K450 of all three IMP1 KH34
copies was truncated to the Cbdue to lack of electron density.
Protein Data Bank: Coordinates and structure factors were
deposited with accession codes 3KRM (IMP1 KH34).
We thank the staff at the NSLS X29a beamline for assistance
with data collection, and G. Arenas, U. Meier, S. Nguyen,
D. Rueda, and S. Ryder for helpful discussions. This work
was supported by the U.S. National Institutes of Health
(GM084364 to R.H.S.) and NRSA individual fellowship support
(F32GM083430 to J.A.C.) and the Albert Einstein Cancer Center.
Adams PD, Grosse-Kunstleve RW, Hung LW, Ioerger TR, McCoy
AJ, Moriarty NW, Read RJ, Sacchettini JC, Sauter NK,
Terwilliger TC. 2002. PHENIX: Building new software for
automated crystallographic structure determination. Acta
Crystallogr D Biol Crystallogr 58: 1948–1954.
Atlas R, Behar L, Elliott E, Ginzburg I. 2004. The insulin-like
growth factor mRNA binding-protein IMP-1 and the Ras-
regulatory protein G3BP associate with tau mRNA and HuD
protein in differentiated P19 neuronal cells. J Neurochem 89:
Atlas R, Behar L, Sapoznik S, Ginzburg I. 2007. Dynamic
association with polysomes during P19 neuronal differenti-
ation and an untranslated-region-dependent translation reg-
ulation of the tau mRNA by the tau mRNA-associated
proteins IMP1, HuD, and G3BP1. J Neurosci Res 85: 173–183.
Auweter SD, Oberstrass FC, Allain FH. 2006. Sequence-specific
binding of single-stranded RNA: Is there a code for recogni-
tion? Nucleic Acids Res 34: 4943–4959.
Backe PH, Messias AC, Ravelli RB, Sattler M, Cusack S. 2005.
X-ray crystallographic and NMR studies of the third KH
domain of hnRNP K in complex with single-stranded nucleic
acids. Structure 13: 1055–1067.
Braddock DT, Louis JM, Baber JL, Levens D, Clore GM. 2002.
Structure and dynamics of KH domains from FBP bound to
single-stranded DNA. Nature 415: 1051–1056.
Chao JA, Lee JH, Chapados BR, Debler EW, Schneemann A,
Williamson JR. 2005. Dual modes of RNA-silencing suppres-
sion by Flock House virus protein B2. Nat Struct Mol Biol 12:
Collaborative Computational Project, Number 4. 1994. The
CCP4 suite: Programs for protein crystallography. Acta
Crystallogr D Biol Crystallogr 50: 760–763.
Davis IW, Leaver-Fay A, Chen VB, Block JN, Kapral GJ, Wang X,
Murray LW, Arendall WB III, Snoeyink J, Richardson JS, et al.
2007. MolProbity: All-atom contacts and structure valida-
tion for proteins and nucleic acids. Nucleic Acids Res 35:
W375–W383. doi: 10.1093/nar/gkm216.
Deshler JO, Highett MI, Abramson T, Schnapp BJ. 1998. A
highly conserved RNA-binding protein for cytoplasmic
mRNA localization in vertebrates. Curr Biol 8: 489–496.
Dimitriadis E, Trangas T, Milatos S, Foukas PG, Gioulbasanis I,
Courtis N, Nielsen FC, Pandis N, Dafni U, Bardi G, et al.
2007. Expression of oncofetal RNA-binding protein CRD-BP/
IMP1 predicts clinical outcome in colon cancer. Int J Cancer
Doyle GA, Betz NA, Leeds PF, Fleisig AJ, Prokipcak RD, Ross J.
1998. The c-myc coding region determinant-binding protein:
A member of a family of KH domain RNA-binding proteins.
Nucleic Acids Res 26: 5036–5044.
Du Z, Lee JK, Tjhen R, Li S, Pan H, Stroud RM, James TL. 2005.
Crystal structure of the first KH domain of human poly(C)-
binding protein-2 in complex with a C-rich strand of human
telomeric DNA at 1.7 A. J Biol Chem 280: 38823–38830.
Du Z, Lee JK, Fenn S, Tjhen R, Stroud RM, James TL. 2007.
X-ray crystallographic and NMR studies of protein–protein
and protein–nucleic acid interactions involving the KH
domains from human poly(C)-binding protein-2. RNA 13:
Du Z, Fenn S, Tjhen R, James TL. 2008. Structure of a construct
of a human poly(C)-binding protein containing the first and
second KH domains reveals insights into its regulatory
mechanisms. J Biol Chem 283: 28757–28766.
Elcheva I, Goswami S, Noubissi FK, Spiegelman VS. 2009. CRD-
BP protects the coding region of bTrCP1 mRNA from miR-
183-mediated degradation. Mol Cell 35: 240–246.
Emsley P, Cowtan K. 2004. Coot: Model-building tools for
molecular graphics. Acta Crystallogr D Biol Crystallogr 60:
Farina KL, Huttelmaier S, Musunuru K, Darnell R, Singer RH.
2003. Two ZBP1 KH domains facilitate b-actin mRNA
localization, granule formation, and cytoskeletal attach-
ment. J Cell Biol 160: 77–87.
Fusco D, Accornero N, Lavoie B, Shenoy SM, Blanchard JM,
Singer RH, Bertrand E. 2003. Single mRNA molecules
demonstrate probabilistic movement in living mammalian
cells. Curr Biol 13: 161–167.
Garcia-Mayoral MF, Hollingworth D, Masino L, Diaz-Moreno I,
Kelly G, Gherzi R, Chou CF, Chen CY, Ramos A. 2007. The
structure of the C-terminal KH domains of KSRP reveals
a noncanonical motif important for mRNA degradation.
Structure 15: 485–498.
Gasteiger E, Hoogland C, Gattiker A, Duvaud S, Wilkins MR,
Appel RD, Bairoch A. 2005. Protein identification and
analysis tools on the ExPASy server. In The proteomics
protocols handbook (ed. JM Walker), pp. 571–607. Humana
Press, Totawa, NJ.
Git A, Standart N. 2002. The KH domains of Xenopus Vg1RBP
Hammer NA, Hansen TO, Byskov AG, Rajpert-De Meyts E,
Grondahl ML, Bredkjaer HE, Wewer UM, Christiansen J,
Nielsen FC. 2005. Expression of IGF-II mRNA-binding pro-
teins (IMPs) in gonads and testicular cancer. Reproduction
Havin L, Git A, Elisha Z, Oberman F, Yaniv K, Schwartz SP,
Standart N, Yisraeli JK. 1998. RNA-binding protein con-
served in both microtubule- and microfilament-based RNA
localization. Genes & Dev 12: 1593–1598.
Jiang Z, Lohse CM, Chu PG, Wu CL, Woda BA, Rock KL, Kwon
ED. 2008. Oncofetal protein IMP3: A novel molecular
marker that predicts metastasis of papillary and chromo-
phobe renal cell carcinomas. Cancer 112: 2676–2682.
Jonson L, Vikesaa J, Krogh A, Nielsen LK, Hansen T, Borup R,
Johnsen AH, Christiansen J, Nielsen FC. 2007. Molecular
ZBP1 recognition of zipcode
GENES & DEVELOPMENT157
composition of IMP1 ribonucleoprotein granules. Mol Cell
Proteomics 6: 798–811.
Kislauskis EH, Zhu X, Singer RH. 1994. Sequences responsible
for intracellular localization of b-actin messenger RNA also
affect cell phenotype. J Cell Biol 127: 441–451.
Kobel M, Xu H, Bourne PA, Spaulding BO, Shih Ie M, Mao TL,
Soslow RA, Ewanowich CA, Kalloger SE, Mehl E, et al. 2009.
IGF2BP3 (IMP3) expression is a marker of unfavorable
prognosis in ovarian carcinoma of clear cell subtype. Mod
Pathol 22: 469–475.
Krissinel E, Henrick K. 2007. Inference of macromolecular
assemblies from crystalline state. J Mol Biol 372: 774–797.
Latham VM Jr, Kislauskis EH, Singer RH, Ross AF. 1994.
b-Actin mRNA localization is regulated by signal trans-
duction mechanisms. J Cell Biol 126: 1211–1219.
Lawrence JB, Singer RH. 1986. Intracellular localization of
messenger RNAs for cytoskeletal proteins. Cell 45: 407–415.
Lecuyer E, Yoshida H, Parthasarathy N, Alm C, Babak T,
Cerovina T, Hughes TR, Tomancak P, Krause HM. 2007.
Global analysis of mRNA localization reveals a prominent
role in organizing cellular architecture and function. Cell
Leung KM, van Horck FP, Lin AC, Allison R, Standart N, Holt
CE. 2006. Asymmetrical b-actin mRNA translation in
growth cones mediates attractive turning to netrin-1. Nat
Neurosci 9: 1247–1256.
Lewis HA, Chen H, Edo C, Buckanovich RJ, Yang YY, Musunuru
K, Zhong R, Darnell RB, Burley SK. 1999. Crystal structures
of Nova-1 and Nova-2 K-homology RNA-binding domains.
Structure 7: 191–203.
Lewis HA, Musunuru K, Jensen KB, Edo C, Chen H, Darnell RB,
Burley SK. 2000. Sequence-specific RNA binding by a Nova
KH domain: Implications for paraneoplastic disease and the
fragile X syndrome. Cell 100: 323–332.
Martin KC, Ephrussi A. 2009. mRNA localization: Gene expres-
sion in the spatial dimension. Cell 136: 719–730.
Mili S, Moissoglu K, Macara IG. 2008. Genome-wide screen
reveals APC-associated RNAs enriched in cell protrusions.
Nature 453: 115–119.
Nielsen J, Christiansen J, Lykke-Andersen J, Johnsen AH,
Wewer UM, Nielsen FC. 1999. A family of insulin-like
growth factor II mRNA-binding proteins represses transla-
tion in late development. Mol Cell Biol 19: 1262–1270.
Nielsen FC, Nielsen J, Kristensen MA, Koch G, Christiansen J.
2002. Cytoplasmic trafficking of IGF-II mRNA-binding pro-
tein by conserved KH domains. J Cell Sci 115: 2087–2097.
Nielsen J, Kristensen MA, Willemoes M, Nielsen FC, Christiansen
J. 2004. Sequential dimerization of human zipcode-binding
protein IMP1 on RNA: A cooperative mechanism providing
RNP stability. Nucleic Acids Res 32: 4368–4376.
Oberman F, Rand K, Maizels Y, Rubinstein AM, Yisraeli JK.
2007. VICKZ proteins mediate cell migration via their RNA
binding activity. RNA 13: 1558–1569.
Oberstrass FC, Auweter SD, Erat M, Hargous Y, Henning A,
Wenter P, Reymond L, Amir-Ahmady B, Pitsch S, Black DL,
et al. 2005. Structure of PTB bound to RNA: Specific binding
and implications for splicing regulation. Science 309: 2054–
Oleynikov Y, Singer RH. 2003. Real-time visualization of ZBP1
association with b-actin mRNA during transcription and
localization. Curr Biol 13: 199–207.
Pagano JM, Farley BM, McCoig LM, Ryder SP. 2007. Molecular
basis of RNA recognition by the embryonic polarity de-
terminant MEX-5. J Biol Chem 282: 8883–8894.
Patel GP, Bag J. 2006. IMP1 interacts with poly(A)-binding
protein (PABP) and the autoregulatory translational control
element of PABP-mRNA through the KH III-IV domain.
FEBS J 273: 5678–5690.
Rambo RP, Doudna JA. 2004. Assembly of an active group II
intron-maturase complex by protein dimerization. Biochem-
istry 43: 6486–6497.
Read RJ. 2001. Pushing the boundaries of molecular replace-
ment with maximum likelihood. Acta Crystallogr D Biol
Crystallogr 57: 1373–1382.
Ross AF, Oleynikov Y, Kislauskis EH, Taneja KL, Singer RH.
1997. Characterization of a b-actin mRNA zipcode-binding
protein. Mol Cell Biol 17: 2158–2165.
Runge S, Nielsen FC, Nielsen J, Lykke-Andersen J, Wewer UM,
Christiansen J. 2000. H19 RNA binds four molecules of
insulin-like growth factor II mRNA-binding protein. J Biol
Chem 275: 29562–29569.
Ryder SP, Recht MI, Williamson JR. 2008. Quantitative analysis
of protein-RNA interactions by gel mobility shift. Methods
Mol Biol 488: 99–115.
Valverde R, Pozdnyakova I, Kajander T, Venkatraman J, Regan L.
2007. Fragile X mental retardation syndrome: Structure of
the KH1–KH2 domains of fragile X mental retardation pro-
tein. Structure 15: 1090–1098.
Valverde R, Edwards L, Regan L. 2008. Structure and function of
KH domains. FEBS J 275: 2712–2726.
Vikesaa J, Hansen TV, Jonson L, Borup R, Wewer UM, Christiansen
J, Nielsen FC. 2006. RNA-binding IMPs promote cell adhe-
sion and invadopodia formation. EMBO J 25: 1456–1468.
Weidensdorfer D, Stohr N, Baude A, Lederer M, Kohn M,
Schierhorn A, Buchmeier S, Wahle E, Huttelmaier S. 2009.
Control of c-myc mRNA stability by IGF2BP1-associated
cytoplasmic RNPs. RNA 15: 104–115.
Yao J, Sasaki Y, Wen Z, Bassell GJ, Zheng JQ. 2006. An essential
role for b-actin mRNA localization and translation in
Ca2+-dependent growth cone guidance. Nat Neurosci 9:
Yisraeli JK. 2005. VICKZ proteins: A multi-talented family of
regulatory RNA-binding proteins. Biol Cell 97: 87–96.
Chao et al.
158 GENES & DEVELOPMENT