Biochem. J. (2012) 443, 39–47 (Printed in Great Britain)
39
doi:10.1042/BJ20111768
The crystal structure of the Sox4 HMG domain–DNA complex suggests a
mechanism for positional interdependence in DNA recognition
Ralf JAUCH*1 , Calista K. L. NG*†, Kamesh NARASIMHAN*‡ and Prasanna R. KOLATKAR*‡1
*Laboratory for Structural Biochemistry, Genome Institute of Singapore, 60 Biopolis St, Singapore 138672, †School of Biological Sciences, Nanyang Technological University,
50 Nanyang Avenue, Singapore 639798, and ‡Department of Biological Sciences, National University of Singapore, 14 Science Drive 4, Singapore 117543
It has recently been proposed that the sequence preferences of
DNA-binding TFs (transcription factors) can be well described
by models that include the positional interdependence of the
nucleotides of the target sites. Such binding models allow for
multiple motifs to be invoked, such as principal and secondary
motifs differing at two or more nucleotide positions. However,
the structural mechanisms underlying the accommodation of
such variant motifs by TFs remain elusive. In the present study
we examine the crystal structure of the HMG (high-mobility
group) domain of Sox4 [Sry (sex-determining region on the Y
chromosome)-related HMG box 4] bound to DNA. By comparing
this structure with previously solved structures of Sox17 and
Sox2, we observed subtle conformational differences at the DNAbinding interface. Furthermore, using quantitative electrophoretic
mobility-shift assays we validated the positional interdependence
of two nucleotides and the presence of a secondary Sox motif
in the affinity landscape of Sox4. These results suggest that a
concerted rearrangement of two interface amino acids enables
Sox4 to accommodate primary and secondary motifs. The
structural adaptations lead to altered dinucleotide preferences
that mutually reinforce each other. These analyses underline
the complexity of the DNA recognition by TFs and provide
an experimental validation for the conceptual framework of
positional interdependence and secondary binding motifs.
INTRODUCTION
the DNA may vary for different Sox family members and could
thus modulate their regulatory output [16]. Lastly, although the
core of the DNA-binding site is shared for all Sox family members,
minor variations in nucleotide preference at motif edges could be
detected [7]. These differences enabled a hierarchical clustering
of the Sox family by virtue of their sequence preferences. Yet,
whether these subtle variations within the DNA target sequences
alter the binding energy in a substantial manner and influence the
genomic occupancy profile remains to be explored.
Structural studies have greatly contributed to unravelling the
architecture and dynamics of the HMG domain and mechanisms
of DNA recognition by Sox proteins [15,17–24]. Importantly, the
kinked DNA conformation seems to depend crucially on highly
conserved amino acids and features of the C-terminal tail of the
domain [16,20]. Yet, features enabling individual Sox proteins
to discriminate between DNA sequences or to induce distinctive
DNA conformations have not been discovered. If such features
exist, the structural analysis of a larger set of Sox–DNA complexes
should reveal their existence.
In the present study, we addressed two fundamental questions:
(i) do different Sox proteins kink the DNA in a differential
manner; and (ii) are the subtly different DNA motifs suggested by
PBM studies supported by different DNA-interaction chemistries?
To tackle these problems we chose the HMG domain of Sox4
and Lama1 (laminin, subunit α1) enhancer DNA that was first
crystallized in complex with the HMG domain of Sox17 [21],
for several reasons. First, Sox4 belongs to the C-group of Sox
proteins [25] whose DNA recognition has not yet been structurally
investigated. Sox4 and other C-group proteins, Sox11 and Sox12,
are critical, partially redundant, regulators in important processes
The evolutionarily ancient [1] Sry(sex-determining region on the
Y chromosome)-box-containing TFs (transcription factors) {Sox
[Sry-related HMG (high-mobility group) box]} were named after
a highly conserved ∼ 80 amino acid HMG (high-mobility group)
DNA-binding domain first identified in the testis-determining
factor Sry [2]. The human genome encodes 20 Sox proteins, most
of which are critical regulators of gene expression programmes in
a diverse set of tissues and at various stages of embryonic as well
as adult development (reviewed in [3–6]).
The DNA-binding specificities of Sox proteins have been
comprehensively interrogated using high-throughput PBM
(protein-binding microarray) technologies [7]. These studies
reaffirmed previous findings on individual family members [8]
that all Sox proteins exhibit highly similar binding preferences to
a TTGT core sequence. However, the notion that all Sox proteins
bind similar and very short DNA sequences seems to contradict the
functional versatility of individual family members. Apparently,
different Sox proteins are recruited to distinctive genomic regions
by means of their affinity to DNA and regulate non-redundant
sets of genes to trigger particular developmental programmes.
Three mechanisms can be conceived to underlie the functional
uniqueness of individual Sox proteins. First, Sox proteins were
found to associate with a range of partner factors, and such
interactions may influence the selection of DNA target sites and
could modulate the cellular consequences of binding events [9–
13]. Secondly, Sox proteins have the peculiar property of binding
to the minor groove of DNA and cause a pronounced kink to the
double helix [15]. Such Sox-induced structural changes to
Key words: DNA binding, high-mobility group (HMG), positional
interdependence, protein–DNA recognition, Sox, transcription
factor.
Abbreviations used: Cy5, indodicarbocyanine; EMSA, electrophoretic mobility-shift assay; HMG, high-mobility group; Lama1, laminin, subunit α1; NSLS,
National Synchrotron Light Source; PBM, protein-binding microarray; PWM, position weight matrix; Sry, sex-determining region on the Y chromosome;
Sox, Sry-related HMG box; TF, transcription factor.
1
Correspondence may be addressed to either of these authors (email [email protected] or [email protected]).
The structural co-ordinates reported for the Sox4 HMG domain bound to DNA will appear in the PDB under accession code 3U2B.
c The Authors Journal compilation c 2012 Biochemical Society
40
R. Jauch and others
such as cardiac and neuronal development [8,26,27] and are
thought to be transcriptional triggers of malignancies [28].
Secondly, PBM studies suggest a slightly altered DNA specificity
of the C-group as compared with the F-group represented by
Sox17 [7]. Thirdly, Sox17 and Sox4 were found to have converse
effects on β-catenin/TCF (T-cell factor) signalling and tumour
progression in colon carcinomas while sharing cellular interaction
partners [29]. This raises the possibility of an antagonistic
action of Sox4 and Sox17, possibly through eliciting different
effects on jointly targeted DNA sites. To be able to conduct
an unbiased comparison of DNA conformation and recognition
and to minimize artefacts introduced by altered DNA sequences,
we decided to use Lama1 DNA that was first crystallized in
complex with Sox17. Indeed, as a result of the present study we
discovered subtly different DNA-binding modes between Sox4
and Sox17. On the basis of these structural models, we propose
a general mechanism for the accommodation and discrimination
between primary and secondary binding sites by Sox proteins by a
concerted structural rearrangement at the DNA-contact interface.
MATERIALS AND METHODS
Cloning, protein expression and purification
Sox17 protein was purified as described previously [43]. The
region encoding 79 amino acids of the HMG domain of
mouse Sox4 was BP cloned (Invitrogen) from a cDNA clone
(UniProt code Q06831; IMAGE Consortium code 6822618) into
the pDONR221 vector. The resulting pENTR construct was
first verified by sequencing and recombined into the pETG40
expression plasmid using Gateway® LR technology (Invitrogen).
The destination plasmid was transformed into Escherichia
coli BL21(DE3) cells and plated on LB (Luria–Bertani)
plates supplemented with 100 μg/ml ampicillin. An overnight
innoculum was transferred into 6 litres of 1× modified Terrific
Broth and grown until the D600 reached ∼ 0.5–0.8 before inducing
with 0.5 mM IPTG (isopropyl-β-D-thiogalactopyranoside) at
18 ◦ C for ∼18 h. Cells were harvested, resuspended in lysis
buffer (20 mM Hepes, pH 7.0, 1 mM EDTA, 200 mM NaCl
and 10 mM 2-mercaptoethanol) and sonicated for 15 min at
4 ◦ C. Cell lysate was cleared by centrifugation at 48 000 g
for 30 min at 4 ◦ C. Pre-equilibrated amylose beads were added to
the filtered lysate and incubated at 4 ◦ C for 1–2 h. Fusion proteins
were eluted with lysis buffer containing 10 mM maltose. The
MBP (maltose-binding protein) tag was separated from Sox4
protein by digestion with TEV (tobacco etch virus) protease
(1:100, w/w) overnight at 4 ◦ C. The cleaved protein sample
was injected into a 6 ml Resource S column (GE Healthcare)
connected to an ÄKTA Express system equilibrated with buffer
A (20 mM Hepes, pH 7.0, and 100 mM NaCl) and was eluted by
increasing the salt concentration gradually to 1 M over a total of 25
column volumes. Peak fractions were pooled and subjected to gelfiltration chromatography using a HiLoadTM 16/60 SuperdexTM
75 prep grad column (GE Healthcare) equilibrated with buffer A.
Peak fractions were analysed by SDS/13.5 % PAGE, pooled and
concentrated to approximately 12.7 mg/ml and finally stored at
− 80 ◦ C until use. The protein concentration was measured with
a Thermo Scientific NanoDrop® ND-1000 spectrophotometer.
Crystallization, data collection and structure solution
PAGE-purified DNA purchased in liquid form from SigmaProligo at 1 mM was adjusted to pH 8.0 using Tris/HCl, mixed
and annealed by heating to 95 ◦ C and gradual cooling to ambient
temperature. The double-stranded DNA and the Sox4 HMG
c The Authors Journal compilation c 2012 Biochemical Society
Table 1
Crystallographic data
*Values for the highest-resolution shell are in parentheses. 1 Å = 0.1 nm.
Sox4-HMG–DNALama1
Data collection
Space group
Cell dimensions (Å)
a ( = b)
c
Resolution (Å)
R sym (%)
I /σ I
Completeness (%)
Multiplicity
Refinement
Resolution (Å)
No. reflections
R work /R free (%)
No. atoms (without hydrogens)
Protein
DNA
Water
Mean B -factors (isotropic equivalent)
Macromolecules
Water
Root mean square deviations from ideal
Bond lengths (Å)
Bond angles ( ◦ )
Ramachandran analysis
Favoured (%)
Additionally allowed (%)
Disallowed (%)
P32 21
69.94
63.20
50.0–2.4 (2.5–2.4)
8.1 (43.6)
39.3 (6.3)
98.4 (99.9)
22.2 (21.8)
42.0–2.4
6855
23.5/27.9
655
650
8
73.05
46.8
0.003
0.7
98.65
1.35
0
domain were mixed at a 1.2:1 molar ratio and the complex at a
concentration of ∼320 μM was subjected to crystallization trials.
The chromosome-shaped crystal used for data collection was
grown using the hanging-drop vapour diffusion technique after
combining equal volumes of protein solution and a reservoir buffer
containing 100 mM Hepes, pH 7.2, 20 % PEG [poly(ethylene
glycol)] 3350, and 50 mM MgCl2 .
Crystals were flash-frozen in liquid nitrogen. Data were
collected at beamline X29 of the NSLS (National Synchrotron
Light Source) and processed using HKL2000 [44] (Table 1).
Molecular replacement was performed using PHASER [45]
and a poly(alanine) model derived from the Sox17–HMG
DNA complex (PDB code 3F27). Model building was initiated
using Buccaneer [46] and finalized manually using Coot [47].
Refinement was carried out using phenix.refine [48].The first
refinement round using the Buccaneer model included a simulated
annealing with a starting temperature of 10 000 K to reduce model
bias. TLS (translation/libration/screw) groups were defined using
the TLSMD server [49]. The structure of the Sox4 HMG domain
bound to DNA has been deposited in the PDB under accession
code 3U2B.
EMSAs (electrophoretic mobility-shift assays)
EMSAs were essentially performed as described previously
[50] with the following modifications. Binding reactions
were set up containing a final concentration of 1 nM Cy5
(indodicarbocyanine)-labelled 16-mer Lama1 DNA with various
amounts of unlabelled DNA competitors and 50 nM Sox protein.
Reactions tubes were incubated at 4 ◦ C in the dark for 1 h
and separated on a 12 % native 1× Tris/glycine (25 mM
Tris, pH 8.3, and 192 mM glycine) polyacrylamide gel. Bands
were detected using a Typhoon 9140 PhosphorImager (GE
Structure of the Sox4 HMG domain
Figure 1
41
Overall structure and DNA binding by the Sox4 HMG
(A) Conservation score plotted against the sequence of the mouse Sox4 HMG domain used in the present study. The score was calculated using the ScoreCons server
(http://www.ebi.ac.uk/thornton-srv/databases/cgi-bin/valdar/scorecons_server.pl) [52] using an alignment of all 20 mouse Sox proteins. Residues were numbered according to HMG conventions
for easier comparision [3]. DNA-contact residues are marked with open circles, the three helices are marked with red bars and the FM wedge with an orange bar. (B) Schematic drawing of the Lama1
DNA sequence used in the present study. Prominent protein–DNA interactions are indicated. Dashed lines indicate base-specific interactions and dotted lines indicate phosphate or deoxyribose
interactions. Red fonts highlight alternative interactions seen for Arg18 and Asn30 in the Sox17 structure [21]. Red numbers label the base-pairs of the core of the Sox17 motif (1–7) discussed in
the text. (C) Cartoon drawing depicting the overall fold of the Sox4 HMG–DNA complex highlighting the strong overall bend of the helix. Positions of N-and C-termini are indicated, helices are
numbered H1–H3, and 5 and 3 DNA ends are labelled for both strand A and strand B. Arg18 and Asn30 are shown as sticks. The inset depicts a magnification of Arg18 and Asn30 rotated by 90 ◦
around the x -axis, and the 2F o − F c electron density is shown contoured at the 1.0 σ level.
Healthcare) and quantified using ImageQuant TL software
(GE Healthcare). The intensities of gel bands corresponding
to free or TF-bound labelled DNA can be used to calculate
the fraction of labelled DNA bound by the TF. Since the total
concentration of labelled DNA is known (1 nM), these fractional
concentrations can be converted into absolute estimates of the
amount of bound and free labelled DNA. The concentration of
unlabelled competitor DNA is also known, as is the total TF
concentration. We can therefore substitute these known values
into the equations that define the dissociation constants in terms
of reactant and product concentrations to calculate the K dmut /Kd value, which can be converted into PWMs (position weight
matrices) ([51], and W. Sun, C. Ng, R. Jauch and S. Prabhakar,
unpublished work). K dmut denotes the K d (dissociation constant)
for the competitor DNA containing systematic mutations and K d the dissociation constant for the labelled Lama1 DNA. EMSA
titrations to estimate absolute K d values were conducted and the
results analysed using equations, as described previously [50].
RESULTS AND DISCUSSION
Overall structure and DNA recognition
The Sox4 HMG domain contains characteristic sequence
signatures found in all 20 mammalian Sox paralogues and,
in particular, DNA contact residues are highly conserved
(Figures 1A and 1B). The protein resembles a prototypical HMG
scaffold with three helices forming an L-shaped fold (Figure 1C).
Previous biophysical studies divided the HMG domain into two
subdomains comprising a minor wing encompassing helices H1
and H2 and a major wing formed by the C-terminal part of helix
H3 and the extended N-terminus that runs antiparallel to this
helix [30]. In accordance with NMR studies of the Sox4 HMG
in the absence of DNA [23], the C-terminal end of the Sox4
HMG exhibits a high degree of structural flexibility as reflected
by high temperature factors and noisy electron density. Owing to
this structural disorder, the last three amino acids of the 79 amino
acid Sox4 HMG domain under study could not be traced and were
excluded from the model. Nevertheless, the C-terminal part of the
protein is crucial for DNA binding and the formation of the kink
[16,21].
A key question of this project was to compare DNA
architectures induced by variant members of the Sox family. To
this end, we crystallized the Sox4 HMG in complex with Lama1
DNA that we had earlier crystallized in complex with Sox17.
Lama1 is a cognate binding site of Sox17 during endodermal
differentiation [31], whereas Sox4 is not known to bind and
regulate this gene in vivo. It is possible that Sox proteins can only
deform their cognate binding sites in a manner compatible with
the assembly of regulation-competent complexes. We therefore
c The Authors Journal compilation c 2012 Biochemical Society
42
R. Jauch and others
carefully analysed the architecture of Sox4- and Sox17-bound
Lama1 DNA using Curves + [32]. Both Sox4 and Sox17 were
found to deform the Lama1 DNA by binding to the minor groove
in a virtually identical manner with overall bends of the helical
axis of 65.8 ◦ for Sox4 and 68.9 ◦ for Sox17. Likewise, intra- and
inter-base-pair parameters were also highly similar. We therefore
conclude that the differential deformation of DNA makes no major
contribution when Sox4 and Sox17 proteins select their target
genes. Accordingly, the majority of DNA contacts are identical
for both proteins (Figures 1A and 1B). As an exception, Arg18
and Asn30 show a different pattern of hydrogen bonding in Sox4
as compared with Sox17. The side-chain conformations of Arg18
and Asn30 are well supported by the 2F o − F c electron density and
were determined after automated model building using Buccaneer
and simulated annealing refinement (Figure 1C). Model bias is
unlikely since the poly(alanine) starting model used for molecular
replacement was derived from Sox17 co-ordinates, which exhibit
a different side-chain conformation in this region. In Sox4, Arg18
and Asn30 position their side chains to contact base-pairs 1 and
2 of the core motif. Since those nucleotides are positioned at the
edge of the Sox motif that was found to exhibit some variability
among Sox paralogues, we decided to study the DNA-binding
specificity of Sox4 in greater detail.
Sox4 binds primary and secondary DNA motifs in a
non-additive manner
We have recently developed a method to measure relative
dissociation constants (K d ) in a competitive EMSA (see
the Materials and methods section). This method enables the
quantification of the affinity ratio between a fluorescently labelled
reporter DNA as compared with an unlabelled competitor.
By measuring K d ratios for a DNA library where all bound
nucleotides are systematically mutated into the three other
possible nucleotides, a position-specific affinity matrix can be
constructed (Figures 2A–2C). We conducted this experiment
using a DNA library derived from the core of the Lama1 DNA.
As expected, we found that the central sequence T3 T4 G5 T6 is
critical for high-affinity binding. However, positions 1 and 2 of the
Sox4-binding site exhibit surprisingly low information contents
after converting the data into a PWM (Figure 2C). In contrast,
PBM studies reported a rather strong preference for CT at these
positions (Figure 2D). However, the PBM method measures a
large number of sites and is capable of detecting less preferred,
but significantly enriched, secondary motifs (Figure 2E). For
Sox4 and the other SoxC group proteins Sox11 and Sox12, the
secondary motif contains an AA at the beginning of the motif.
Two considerations led us to wonder whether our EMSA
method captures the true sequence preference of Sox4. First, the
affinity matrix was recorded on the basis of the Lama1 element
containing a starting TA which constitutes neither an ideal primary
nor an ideal secondary motif. Secondly, this approach assumes
positional independence. That is, nucleotides are changed one
by one, assuming that mutations at site 1 have no effect on the
nucleotide preference at site 2. As a consequence, positional
interdependence, in this case between site 1 and site 2, could
explain the discrepancy between the EMSA and the PBM-derived
PWMs. To test this hypothesis, we designed an additional series
of mutated Lama1 DNA elements and tested for the affinity ratio
compared with wild-type Lama1 DNA (Figure 2G). Indeed, the
CT element bound Sox4 more strongly than the TA-containing
Lama1 DNA. Overall, the T2 was preferred in instances when C1
was present, underlining the reliability of the primary PBM motif.
As expected, the secondary A1A2 motif has an affinity lower than
the primary C1T2 motif. However, when an A1 is present the
c The Authors Journal compilation c 2012 Biochemical Society
preference at position 2 changes from a T to an A. Taken together;
these experiments support the presence of primary and secondary
motifs in the affinity landscape of Sox4. Furthermore, there seems
to be a significant positional interdependence at binding positions
1 and 2: a C1 favours a T at position 2, whereas A1 enforces an
A at position 2.
Sox2, Sox17 and Sox4 adopt different conformations at the
DNA-interaction interface
Next we carefully inspected the environment in the structures of
Sox4 and Sox17 around base-pairs 1 and 2 exhibiting positional
interdependence in EMSA experiments. Intriguingly, the only
notable conformational differences at the DNA-contact interface
between Sox4 and Sox17 is seen for Arg18 and Asn30 , which
contact base-pairs 1 and 2 (Figures 1B and 1D and Figure 3A).
Specifically, in Sox4, Arg18 contacts the minor groove O2 carbonyl
of T2 and Asn30 contacts the minor groove N3 of A1 (numbering
according to base-pairs 1–7 of the core Sox4 motif shown in
Figures 2C–2E; N refers to the complementary nucleotide). This
conformational arrangement is further stabilized through direct
side chain contacts between Arg18 and Asn30 . In contrast, in
Sox17, the side chains of both amino acids are slightly re-oriented,
causing Arg18 to contact A3 and Asn30 to contact T2 , whereas
the intramolecular Arg18 –Asn30 interaction is lost.
To further investigate these subtle, but potentially consequential, conformational differences, we inspected the crystal structure
of Sox2 (PDB code 1GT0), which is bound to a sequence closely
resembling a primary motif beginning with C-G1;T-A2 base-pairs
(Figure 3B). Importantly, the binding pattern of the Arg18 –Asn30
dipeptide in Sox2 is reminiscent to the situation seen in Sox4.
Given that Sox2 is bound to a perfect primary sequence (C1T2), it
seems likely that the Arg18 –Asn30 conformation seen for Sox2 and
Sox4 represents the interaction mode characteristic for the higheraffinity primary Sox motifs. Apparently, the primary scenario
is facilitated by two consecutive purine N3 acceptors at motif
positions 1 and 2 within the reverse strand. This poses the question
of whether the conformation modelled for Sox17 represents a
conformational switch of Arg18 /Asn30 reminiscent of Sox4 bound
to a secondary motif (Figure 3C). If so, Arg18 and Asn30 may
undergo an interconnected rearrangement to accommodate the
different chemical environments encountered on primary and
secondary motifs (Figure 4).
To further assess this possibility, we converted the Sox17bound Lama1 sequence into a secondary motif and inspected
the potential binding pattern occurring on such an element. In
this modelled secondary motif a carbonyl is introduced by T1 at
position 1, whereas the primary motif as well as the Lama1 site
would provide a purine N3 as an H-bond acceptor. Both purine
N3 as well as the pyrimidine O2 carbonyl can act as H-bond
acceptors. However, a comprehensive analysis of protein–DNA
complexes revealed that the H-bonding geometries of amino acids
interacting with purines and pyrimidines are markedly different
[33]. For example, H-bonds around purine N3 were found to be
more compact and shifted by a different amplitude away from the
base plane and exhibit different binding angles. Although subtle,
such differences have enabled the design of sequence-specific
drugs that target the minor groove [34]. Thus replacing the purine
N3 with a pyrimidine O2 at position 1 would probably disturb
the H-bonding geometry and would be incompatible with the
Asn30 conformation seen for Sox4 and Sox2. As a consequence,
a conformation seen in Sox17 where the side chain flips and
the T2 carbonyl is contacted instead would be more favourable.
Indeed, we observed favourable H-bond distances when Sox17
was modelled on the secondary motif (Figure 3D and Figure 4).
Structure of the Sox4 HMG domain
Figure 2
43
Validation of positional interdependence in DNA-motif recognition by Sox4
Representative competition EMSA experiments performed using the Sox4 HMG (A) and Sox17 HMG (B). Proteins were mixed with 1 nM Cy5–DNALama1 and systematically mutated unlabelled
competitor DNA elements. Every position of the 7 bp core of the Lama1 core DNA element was mutated into all possible alternative base-pairs, producing a library of 21 elements. Lanes 1–3 contain
no competitor and the indicated concentration of protein. WT indicates wild-type competitor DNA, and the letters in lane 5–25 (A) or lanes 4–24 (B) and subscripts denote the mutant nucleotide and
its position within the core motif. The band intensity of the TF complexed with labelled DNA is a monotonic function of the dissociation constant K dmut of the unlabelled competitor DNA fragment. A
weak intensity indicates a low K dmut and a high affinity. This intensity can therefore be quantified and used to estimate K dmut /K d , where K d is the K d for the labelled reporter DNA (see the Materials
and methods section). (C) PWM derived from K dmut /K d ratios estimated from the competitive EMSA in (A) were visualized using the R tool seqLogo. Primary (D) and secondary (E) Sox4 motifs
as determined using the PBM method [7]. PBM PWMs were downloaded from the UniProbe database (http://the_brain.bwh.harvard.edu/uniprobe/) and visualized using seqLogo. (F) PWM derived
for the Sox17 HMG by EMSAs. To test for positional interdependence, double-mutant competitor DNA elements were analysed using Sox4 HMG (G) and Sox17 HMG (H). Positions 1 and 2 were
modified to resemble primary (CT) and secondary (AA) motifs. Note that the Lama1 element has a TA at these positions constituting neither a perfect primary nor secondary element. Base-pairs at
position 2 were changed into all of the three possible nucleotides. The y -axis shows the log2 -transformed K d ratios (K dmut /K d ) of the mutant DNA and the Lama1 DNA as means +
− S.D. Values
above zero correspond to an increased K d (decreased affinity) and values below zero to a decreased K d (increased affinity). The primary CT element has an even higher affinity than the WT element
to the Sox4 HMG, whereas the drop in affinity by an A at position 1 is partially rescued by a second A at position 2. However, the single-nucleotide exchange from CT to AT results in a pronounced
drop in affinity, suggesting positional interdependence. The positional interdependence is not observed for Sox17 (H), and the TA dinucleotide of the Lama1 DNA, which is a cognate Sox17 target
site, is indeed the highest-affinity binding site.
c The Authors Journal compilation c 2012 Biochemical Society
44
Figure 3
R. Jauch and others
Structural differences at the DNA-binding interface in Sox2, Sox4 and Sox17
Comparison of the binding to base-pair positions 1 and 2 as seen in (A) Sox4 (blue), (B) Sox2 (PDB code 1GT0; green), (C) Sox17 (PDB code 3F27; light pink) and (D) Sox17 on a model of a
secondary motif (light blue). The models correspond to a view rotated around the x -axis as compared with Figure 1(B) to visualize the complex from the top and are shown in wall-eyed stereo. The
side chains of Arg18 , His29 , Asn30 and Ser34 and the base-pairs − 1 to 3 (see Figures 1B and 2C) are shown as sticks. Note that Ser34 occupies the same position and binding characteristics in all
structures, whereas Arg18 and Asn30 show a different H-bonding geometry in Sox2 and Sox4 when compared with Sox17. In addition, His29 exhibits a flipped-out conformation in Sox2.
To facilitate the rearrangement from a primary to a secondary
binding mode, Arg18 has to give way for Asn30 to now contact
the N3 of A3 instead of the position 2 . Are these structural
rearrangements consistent with the positional interdependence?
Apparently, Sox4 favours a T2 when a T1 is present (Figure 2G).
It is conceivable that a purine with an N3 instead of a pyrimidine’s
c The Authors Journal compilation c 2012 Biochemical Society
O3 as at the 2 position would not install the energetically most
favourable H-bond geometry for the conformation seen in Sox17.
Likewise, if the pyrimidine O3 was provided by a C2 instead
of a T2 , the complementary G2 could disturb the H-bonding
geometry by inserting an additional amino group into the minor
groove which could, for example, interfere with the binding of
Structure of the Sox4 HMG domain
Figure 4
45
Model for how Sox4 structurally adapts to recognize primary and secondary binding sites
Only the side chains of Arg18 and Asn30 and base-pairs 1–3 of the reverse strand of the core Sox motif that are contacted by Arg18 and Asn30 in either scenario are displayed. The model assumes that
the change from an N3 to an O2 as the H-bond acceptor at position 1 enforces different H-bond acceptors at position 2 (the central nucleobase in this scheme): a thymidine O2 is enforced on the 2
position of the secondary motif with an O2 at position 1 , whereas a 2 purine N3 is preferred on the primary motif with an N3 at position 1 . If those conditions are met, a concerted re-orientation of
the side chains of Arg18 and Asn30 is triggered to accommodate the variant chemical environments. Although subtle, this rearrangement could rationalize the positional interdependence for changes
of base-pairs 1 and 2 and underlie the discrimination between primary and secondary motifs by SoxC TFs. The lower panels show representative binding isotherms recorded using EMSAs on primary
and secondary DNA elements. K d values are shown as means +
− S.D. for three different experiments. In the present assay, there is only a ∼ 3-fold K d difference, demonstrating that Sox4 is capable
of binding both primary and secondary motifs with high affinity.
Ser34 . We therefore propose that an A-T1 base-pair re-enforces
an A-T2 in order to have the less compact consecutive O2 Hbond acceptors at positions 1 and 2, which allow the side chains
of Arg18 /Asn30 to dislodge from their most favoured primary
position to adopt the secondary-like conformation seen in Sox17
(Figure 4). Anything except for an A-T2 would disfavour this
arrangement due to suboptimal H-bonding geometries, explaining
the positional interdependence.
Interestingly, Sox17 appears to have a stronger preference for
an A-T2 basepair using additive binding models (Figure 2F).
Thus Sox17, unlike Sox4, may be unable to discriminate
between primary and secondary motifs. Indeed, a C1T2type motif has not been found for Sox17 in PBM studies
(http://the_brain.bwh.harvard.edu/uniprobe/). In fact, according
to this dataset, the distinction between C1T2- and A1A2-type
motifs appears to be characteristic for SoxC-type HMG domains
and may not apply to the whole Sox family. Additional structures
of Sox HMG domains in complex with motif variants, in
conjunction with molecular dynamics simulations, are needed
to further explore differences between Sox family members and
the discrimination between motif variants by HMG domains. In
addition, although the present model emphasizes the H-bonding
geometry as a key contributor to the Sox4 interaction with primary
and secondary DNA motifs, other factors such as base stacking
energies can influence sequence preferences and shape the binding
landscape of TFs [35].
Conclusion
It has long been recognized that protein–DNA recognition
is a highly complex process and does not follow a simple
recognition code [36,37]. That is, for most TF families the DNA
target sequence cannot be straightforwardly predicted from the
amino acid sequence. The goal of many high-throughput TF–
DNA-interaction studies, such as HiTS-FLIP (high-throughput
sequencing fluorescent ligand interaction profile), PBM and HTSELEX (high-throughput systematic evolution of ligands by
exponential enrichment), and microfluidic experiments, such
as MITOMI (mechanically induced trapping of molecular
interactions), is to arrive at a minimal model that would
adequately describe the DNA-interaction landscape for TFs
[7,38–40]. Accurate models for the TF–DNA interactions are
critical to predict the genomic binding of a TF in cells and to
understand gene regulatory networks. In addition, these insights
can help to re-engineer TF proteins with novel specificities. The
issue has been complicated by high-throughput studies which
contend that simple additive PWM-based models do no suffice to
comprehensively capture the affinity landscape of TFs [7,40].
Rather, individual TFs seem capable of recognizing multiple
motifs with positionally interdependent base substitutions.
However, the prevalence of such secondary motifs in available
datasets is under some debate and method development to analyse
TF binding data is far from complete [41,42]. Nevertheless, the
structural pre-requisite for binding to disparate DNA motifs is that
a TF must be sufficiently flexible to adapt to alternative chemical
environments and to engage in new intermolecular contacts.
In the present study we explored the positional interdependence
and the structural basis for the discrimination between primary
and secondary motifs by SoxC group TFs. By definition, a
primary motif describes the highest-affinity consensus motif of
a TF. In contrast, a secondary motif describes a population of
lower-affinity binding sites that significantly deviates by more
than 1 bp from the primary consensus. Additive PWM-based
binding models are not expected to be able to fully describe the
affinity landscape of a protein if a noticeable secondary binding
mode is present. We propose a model that links recognition
of primary versus secondary DNA motifs with the structural
c The Authors Journal compilation c 2012 Biochemical Society
46
R. Jauch and others
rearrangements at the DNA-binding interface of SoxC proteins.
In the light of the present study, we propose that the SoxC proteins
are capable of high-affinity binding to primary as well as to
secondary binding sites by the concerted re-adjustment of the
H-bonding pattern of two amino acid side chains. High-resolution
structures of primary and secondary binding sites are needed
to put this model to a rigorous test. Although the positional
interdependence appears to be significant, the resultant affinity
gain is rather moderate, and it remains to be investigated how
these differences affect the targeting of genomic loci. It is possible
that primary and secondary motifs represent populations of highand medium-affinity binding that are sensitive to even slightly
different concentration thresholds of a TF. A TF’s ability to
be inherently flexible and to structurally adapt in order to bind
primary and secondary motifs with graduated affinity windows
may therefore expand its repertoire of binding sites. It will be
interesting to explore whether primary and secondary motifs of
Sox4 can be functionally set apart for carrying out distinctive
regulatory roles.
AUTHOR CONTRIBUTION
Ralf Jauch conceived and designed the study, collected and assembled data, wrote the
paper and gave final approval of the paper. Calista Ng collected, analysed and interpreted
data and gave final approval of the paper. Kamesh Narasimhan analysed and interpreted data
and wrote the paper. Prasanna Kolatkar obtained financial support, provided administrative
support and gave final approval of the paper.
ACKNOWLEDGEMENTS
Bob Robinson [Institute of Molecular and Cell Biology, Singapore Agency for
Science, Technology and Research (A*STAR), Singapore] generously provided access
to crystallization and X-ray diffraction equipment. We thank Howard Robinson for data
collection and processing at beamline X29 of the NSLS, and Shyam Prabhakar and Wenjie
Sun for sharing analysis methods and critical comments on the paper prior to publication.
FUNDING
This work was supported by the A*STAR. The NSLS is supported by the Office of Basic
Energy Sciences, Office of Science, U.S. Department of Energy.
REFERENCES
1 Jager, M., Queinnec, E., Houliston, E. and Manuel, M. (2006) Expansion of the SOX gene
family predated the emergence of the Bilateria. Mol. Phylogenet. Evol. 39, 468–477
2 Sinclair, A. H., Berta, P., Palmer, M. S., Hawkins, J. R., Griffiths, B. L., Smith, M. J., Foster,
J. W., Frischauf, A. M., Lovell-Badge, R. and Goodfellow, P. N. (1990) A gene from the
human sex-determining region encodes a protein with homology to a conserved
DNA-binding motif. Nature 346, 240–244
3 Bowles, J., Schepers, G. and Koopman, P. (2000) Phylogeny of the SOX family of
developmental transcription factors based on sequence and structural indicators.
Dev. Biol. 227, 239–255
4 Kiefer, J. C. (2007) Back to basics: Sox genes. Dev. Dyn. 236, 2356–2366
5 Lefebvre, V., Dumitriu, B., Penzo-Mendez, A., Han, Y. and Pallavi, B. (2007) Control of cell
fate and differentiation by Sry-related high-mobility-group box (Sox) transcription factors.
Int. J. Biochem. Cell Biol. 39, 2195–2214
6 Wegner, M. (1999) From head to toes: the multiple facets of Sox proteins. Nucleic Acids
Res. 27, 1409–1420
7 Badis, G., Berger, M. F., Philippakis, A. A., Talukder, S., Gehrke, A. R., Jaeger, S. A., Chan,
E. T., Metzler, G., Vedenko, A., Chen, X. et al. (2009) Diversity and complexity in DNA
recognition by transcription factors. Science 324, 1720–1723
8 van de Wetering, M., Oosterwegel, M., van Norren, K. and Clevers, H. (1993) Sox-4, an
Sry-like HMG box protein, is a transcriptional activator in lymphocytes. EMBO J. 12,
3847–3854
9 Ambrosetti, D. C., Basilico, C. and Dailey, L. (1997) Synergistic activation of the fibroblast
growth factor 4 enhancer by Sox2 and Oct-3 depends on protein–protein interactions
facilitated by a specific spatial arrangement of factor binding sites. Mol. Cell. Biol. 17,
6321–6329
c The Authors Journal compilation c 2012 Biochemical Society
10 Jauch, R., Aksoy, I., Hutchins, A. P., Ng, C. K., Tian, X. F., Chen, J., Palasingam, P.,
Robson, P., Stanton, L. W. and Kolatkar, P. R. (2011) Conversion of Sox17 into a
pluripotency reprogramming factor by reengineering its association with Oct4 on DNA.
Stem Cells 29, 940–951
11 Kamachi, Y., Uchikawa, M., Tanouchi, A., Sekido, R. and Kondoh, H. (2001) Pax6 and
SOX2 form a co-DNA-binding partner complex that regulates initiation of lens
development. Genes Dev. 15, 1272–1286
12 Kuhlbrodt, K., Herbarth, B., Sock, E., Enderich, J., Hermans-Borgmeyer, I. and Wegner, M.
(1998) Cooperative function of POU proteins and SOX proteins in glial cells. J. Biol.
Chem. 273, 16050–16057
13 Nishimoto, M., Fukushima, A., Okuda, A. and Muramatsu, M. (1999) The gene for the
embryonic stem cell coactivator UTF1 carries a regulatory element which selectively
interacts with a complex composed of Oct-3/4 and Sox-2. Mol. Cell. Biol. 19, 5453–5465
14 Ng, C. K., Li, N., Chee, S., Prabhakar, S., Kolatkar, P. R. and Jauch, R. (2012) Deciphering
the Sox-Oct partner code by quantitative cooperativity measurements. Nucleic Acids Res.,
in the press
15 Werner, M. H., Huth, J. R., Gronenborn, A. M. and Clore, G. M. (1995) Molecular basis of
human 46X,Y sex reversal revealed from the three-dimensional solution structure of the
human SRY–DNA complex. Cell 81, 705–714
16 Dragan, A. I., Read, C. M., Makeyeva, E. N., Milgotina, E. I., Churchill, M. E.,
Crane-Robinson, C. and Privalov, P. L. (2004) DNA binding and bending by HMG boxes:
energetic determinants of specificity. J. Mol. Biol. 343, 371–393
17 Cary, P. D., Read, C. M., Davis, B., Driscoll, P. C. and Crane-Robinson, C. (2001)
Solution structure and backbone dynamics of the DNA-binding domain of mouse Sox-5.
Protein Sci. 10, 83–98
18 Love, J. J., Li, X., Case, D. A., Giese, K., Grosschedl, R. and Wright, P. E. (1995)
Structural basis for DNA bending by the architectural transcription factor LEF-1. Nature
376, 791–795
19 Murphy, IV, F. V., Sweet, R. M. and Churchill, M. E. (1999) The structure of a chromosomal
high mobility group protein–DNA complex reveals sequence-neutral mechanisms
important for non-sequence-specific DNA recognition. EMBO J. 18, 6610–6618
20 Murphy, E. C., Zhurkin, V. B., Louis, J. M., Cornilescu, G. and Clore, G. M. (2001)
Structural basis for SRY-dependent 46-X,Y sex reversal: modulation of DNA bending by a
naturally occurring point mutation. J. Mol. Biol. 312, 481–499
21 Palasingam, P., Jauch, R., Ng, C. K. and Kolatkar, P. R. (2009) The structure of Sox17
bound to DNA reveals a conserved bending topology but selective protein interaction
platforms. J. Mol. Biol. 388, 619–630
22 Remenyi, A., Lins, K., Nissen, L. J., Reinbold, R., Scholer, H. R. and Wilmanns, M. (2003)
Crystal structure of a POU/HMG/DNA ternary complex suggests differential assembly of
Oct4 and Sox2 on two enhancers. Genes Dev. 17, 2048–2059
23 van Houte, L. P., Chuprina, V. P., van der Wetering, M., Boelens, R., Kaptein, R. and
Clevers, H. (1995) Solution structure of the sequence-specific HMG box of the
lymphocyte transcriptional activator Sox-4. J. Biol. Chem. 270, 30516–30524
24 Williams, Jr, D. C., Cai, M. and Clore, G. M. (2004) Molecular basis for synergistic
transcriptional activation by Oct1 and Sox2 revealed from the solution structure of the
42-kDa Oct1 · Sox2 · Hoxb1-DNA ternary transcription factor complex. J. Biol. Chem.
279, 1449–1457
25 Penzo-Mendez, A. I. (2009) Critical roles for SoxC transcription factors in development
and cancer. Int. J. Biochem. Cell Biol. 42, 425–428
26 Dy, P., Penzo-Mendez, A., Wang, H., Pedraza, C. E., Macklin, W. B. and Lefebvre, V.
(2008) The three SoxC proteins – Sox4, Sox11 and Sox12 – exhibit overlapping
expression patterns and molecular properties. Nucleic Acids Res. 36, 3101–3117
27 Potzner, M. R., Tsarovina, K., Binder, E., Penzo-Mendez, A., Lefebvre, V., Rohrer, H.,
Wegner, M. and Sock, E. (2010) Sequential requirement of Sox4 and Sox11 during
development of the sympathetic nervous system. Development 137, 775–784
28 Scharer, C. D., McCabe, C. D., Ali-Seyed, M., Berger, M. F., Bulyk, M. L. and Moreno,
C. S. (2009) Genome-wide promoter analysis of the SOX4 transcriptional network in
prostate cancer cells. Cancer Res. 69, 709–717
29 Sinner, D., Kordich, J. J., Spence, J. R., Opoka, R., Rankin, S., Lin, S. C., Jonatan, D.,
Zorn, A. M. and Wells, J. M. (2007) Sox17 and Sox4 differentially regulate
β-catenin/T-cell factor activity and proliferation of colon carcinoma cells. Mol. Cell. Biol.
27, 7802–7815
30 Crane-Robinson, C., Read, C. M., Cary, P. D., Driscoll, P. C., Dragan, A. I. and Privalov,
P. L. (1998) The energetics of HMG box interactions with DNA. Thermodynamic
description of the box from mouse Sox-5. J. Mol. Biol. 281, 705–717
31 Niimi, T., Hayashi, Y., Futaki, S. and Sekiguchi, K. (2004) SOX7 and SOX17 regulate the
parietal endoderm-specific enhancer activity of mouse laminin α1 gene. J. Biol. Chem.
279, 38055–38061
32 Lavery, R., Moakher, M., Maddocks, J. H., Petkeviciute, D. and Zakrzewska, K. (2009)
Conformational analysis of nucleic acids revisited: Curves + . Nucleic Acids Res. 37,
5917–5929
33 Moravek, Z., Neidle, S. and Schneider, B. (2002) Protein and drug interactions in the
minor groove of DNA. Nucleic Acids Res. 30, 1182–1191
Structure of the Sox4 HMG domain
34 Spitzer, G. M., Wellenzohn, B., Markt, P., Kirchmair, J., Langer, T. and Liedl, K. R. (2009)
Hydrogen-bonding patterns of minor groove-binder–DNA complexes reveal criteria for
discovery of new scaffolds. J. Chem. Inf. Model. 49, 1063–1069
35 Hunter, C. A. and Lu, X. J. (1997) DNA base-stacking interactions: a comparison of
theoretical calculations with oligonucleotide X-ray crystal structures. J. Mol. Biol. 265,
603–619
36 Benos, P. V., Lapedes, A. S. and Stormo, G. D. (2002) Is there a code for protein–DNA
recognition? Probab(ilistical)ly. Bioessays 24, 466–475
37 Luscombe, N. M., Laskowski, R. A. and Thornton, J. M. (2001) Amino acid–base
interactions: a three-dimensional analysis of protein–DNA interactions at an atomic level.
Nucleic Acids Res. 29, 2860–2874
38 Jolma, A., Kivioja, T., Toivonen, J., Cheng, L., Wei, G., Enge, M., Taipale, M., Vaquerizas,
J. M., Yan, J., Sillanpaa, M. J. et al. (2010) Multiplexed massively parallel SELEX for
characterization of human transcription factor binding specificities. Genome Res. 20,
861–873
39 Maerkl, S. J. and Quake, S. R. (2007) A systems approach to measuring the binding
energy landscapes of transcription factors. Science 315, 233–237
40 Nutiu, R., Friedman, R. C., Luo, S., Khrebtukova, I., Silva, D., Li, R., Zhang, L., Schroth,
G. P. and Burge, C. B. (2011) Direct measurement of DNA affinity landscapes on a
high-throughput sequencing instrument. Nat. Biotechnol. 29, 659–664
41 Morris, Q., Bulyk, M. L. and Hughes, T. R. (2011) Jury remains out on simple models of
transcription factor specificity. Nat. Biotechnol. 29, 483–484
42 Zhao, Y. and Stormo, G. D. (2011) Quantitative analysis demonstrates most
transcription factors require only simple models of specificity. Nat. Biotechnol. 29,
480–483
47
43 Ng, C. K., Palasingamn, P., Venkatachalam, R., Baburajendran, N., Cheng, J., Jauch, R.
and Kolatkar, P. R. (2008) Purification, crystallization and preliminary X-ray diffraction
analysis of the HMG domain of the Sox17 in complex with DNA. Acta Crystallogr., Sect. F:
Struct. Biol. Cryst. Commun. 64, 1184–1187
44 Otwinowski, Z. and Minor, W. (1997) Processing of X-ray diffraction data collected in
oscillation mode. Methods Enzymol. 276, 307–326
45 McCoy, A. J., Grosse-Kunstleve, R. W., Storoni, L. C. and Read, R. J. (2005)
Likelihood-enhanced fast translation functions. Acta Crystallogr., Sect. D: Biol.
Crystallogr. 61, 458–464
46 Cowtan, K. (2006) The Buccaneer software for automated model building. 1. Tracing
protein chains. Acta Crystallogr., Sect. D: Biol. Crystallogr. 62, 1002–1011
47 Emsley, P. and Cowtan, K. (2004) Coot: model-building tools for molecular graphics. Acta
Crystallogr., Sect. D: Biol. Crystallogr. 60, 2126–2132
48 Afonine, P. V., Grosse-Kunstleve, R. W. and Adams, P. D. (2005) The Phenix refinement
framework. CCP4 Newsl. 42 (http://www.ccp4.ac.uk/newsletters/newsletter42/
articles/Afonine_GrosseKunstleve_Adams_18JUL2005.doc)
49 Painter, J. and Merritt, E. A. (2006) Optimal description of a protein structure in terms of
multiple groups undergoing TLS motion. Acta Crystallogr., Sect. D: Biol. Crystallogr. 62,
439–450
50 BabuRajendran, N., Palasingam, P., Narasimhan, K., Sun, W., Prabhakar, S., Jauch, R. and
Kolatkar, P. R. (2010) Structure of Smad1 MH1/DNA complex reveals distinctive
rearrangements of BMP and TGF-β effectors. Nucleic Acids Res. 38, 3477–3488
51 Zhao, Y., Granas, D. and Stormo, G. D. (2009) Inferring binding energies from selected
binding sites. PLoS Comput. Biol. 5, e1000590
52 Valdar, W. S. (2002) Scoring residue conservation. Proteins 48, 227–241
Received 4 October 2011/19 December 2011; accepted 19 December 2011
Published as BJ Immediate Publication 19 December 2011, doi:10.1042/BJ20111768
c The Authors Journal compilation c 2012 Biochemical Society