Báo cáo khoa học: Common mode of DNA binding to cold shock domains Crystal structure of hexathymidine bound to the domain-swapped form of a major cold shock protein from Bacillus caldolyticus pot

Common mode of DNA binding to cold shock domains
Crystal structure of hexathymidine bound to the domain-swapped
form of a major cold shock protein from Bacillus caldolyticus
Klaas E. A. Max
1
, Markus Zeeb
2,
*, Ralf Bienert
1
, Jochen Balbach
2,
and Udo Heinemann
1,3
1 Max-Delbru
¨
ck-Centrum fu
¨
r Molekulare Medizin Berlin-Buch, Germany
2 Lehrstuhl fu
¨
r Biochemie, Universita
¨
t Bayreuth, Germany
3 Institut fu
¨
r Chemie und Biochemie, Freie Universita
¨
t Berlin, Germany
When bacteria are subjected to a temperature decrease
of about 10 °C, they respond with an adaptive mech-
anism known as the cold shock response. Conse-

quently, the expression of most cellular genes is
downregulated, and the expression of some genes
involved in cellular adaptation to cold stress is upregu-
lated [1–3]. Although most genes involved in the cold
shock response vary between species, a conserved set
of genes encoding the major cold shock proteins (CSP)
has been found in more than 400 different bacteria,
including hyperthermophilic, thermophilic, mesophilic
and psychrophilic species. The CSP consist of 65–70
amino acids and bind to single-stranded nucleic acids
with micromolar to nanomolar dissociation constants
(K
D
).
The precise cellular function of the CSP is under
investigation. In vitro, Ec-CspA, a major CSP from
Escherichia coli, has been shown to prevent the forma-
tion of mispaired RNA duplex structures in a
sequence-unspecific manner [4]. Such structures are
expected to form preferentially at low temperatures
and may interfere with translation or cause mRNA
Keywords
cold shock response; domain swap;
OB-fold; protein–DNA complex;
single-stranded DNA
Correspondence
U. Heinemann, Max-Delbru
¨
ck-Centrum fu
¨

r
Molekulare Medizin, Robert-Ro
¨
ssle-Str. 10,
13125 Berlin, Germany
Fax: +49 30 9406 2548
Tel: +49 30 9406 3420
E-mail:
Present address
*Department of Molecular Biology, The
Scripps Research Institute, La Jolla, CA, USA
Fachgruppe Biophysik, Fachbereich Physik,
Martin-Luther-Universita
¨
t Halle-Wittenberg,
Germany
(Received 30 October 2006, revised 22
December 2006, accepted 22 December
2006)
doi:10.1111/j.1742-4658.2007.05672.x
Bacterial cold shock proteins (CSPs) regulate cellular adaptation to cold
stress. Functions ascribed to CSP include roles as RNA chaperones and in
transcription antitermination. We present the crystal structure of the Bacil-
lus caldolyticus CSP (Bc-Csp) in complex with hexathymidine (dT
6
)ata
resolution of 1.29 A
˚
. Bound to dT
6

, crystalline Bc-Csp forms a domain-
swapped dimer in which b strands 1–3 associate with strands 4 and 5 from
the other subunit to form a closed b barrel and vice versa. The globular
units of dimeric Bc-Csp closely resemble the well-known structure of
monomeric CSP. Structural reorganization from the monomer to the
domain-swapped dimer involves a strictly localized change in the peptide
bond linking Glu36 and Gly37 of Bc-Csp. Similar structural reorganiza-
tions have not been found in any other CSP or oligonucleotide ⁄ oligosac-
charide-binding fold structures. Each dT
6
ligand is bound to one globular
unit of Bc-Csp via an amphipathic protein surface. Individual binding sub-
sites interact with the DNA bases through stacking and hydrogen bonding.
The sugar–phosphate backbone remains solvent exposed. Based on crystal-
lographic and biochemical studies of deoxyoligonucleotide binding to CSP,
we suggest a common mode of binding of single-stranded heptanucleotide
motifs to proteins containing cold shock domains, including the eukaryotic
Y-box factors.
Abbreviations
CSD, cold shock domain; CSP, cold shock protein; OB-fold, oligonucleotide ⁄ oligosaccharide binding fold.
FEBS Journal 274 (2007) 1265–1279 ª 2007 The Authors Journal compilation ª 2007 FEBS 1265
degradation. Therefore, the CSP have been designated
as mRNA chaperones, which help to maintain protein
synthesis at low temperatures. In a different study,
Ec-CspA, Ec-CspC and Ec-CspE, whose synthesis is
upregulated during cold stress, were shown to function
as transcription antiterminators of cold-induced genes
both in vitro and in vivo [5]. Moreover, CSP have been
implicated in nucleoid condensation, the coupling of
transcription and translation, and regulation of trans-

lation [6–9].
Although it has been suggested that CSP bind to
Y-box sequences with high affinity [10], binding experi-
ments with oligonucleotides have shown that both
Ec-CspA and Bs -CspB from Bacillus subtilis preferen-
tially bind to pyrimidine-rich oligonucleotides [11,12].
Binding to the Y-box sequence, either alone or in a
poly(T) context, was negligible [13,14]; the highest
affinities of CSP were reported for thymine (T)- or
uracil (U)-rich sequences.
Several CSP structures have been determined using
X-ray crystallography [15–18] and NMR spectroscopy
[19,20], including Ec-CspA, Bs-CspB and the CSP
from Bacillus caldolyticus (Bc-Csp) and Thermotoga
maritima (Tm-Csp). The peptide chains of the CSP are
arranged as five antiparallel b strands, separated by
four loops and folded into a closed b barrel [21]. This
fold belongs to the oligonucleotide ⁄ oligosaccharide
binding (OB) fold [22]. It is conserved in eukaryotic Y-
box proteins [23], which contain structures in addition
to the cold shock domain (CSD) shared with the bac-
terial CSP.
A recently determined crystal structure of Bs-CspB
in complex with dT
6
[24] and data from solution
NMR experiments characterizing the binding of dT
7
to
Bs-CspB [25] has shown that T-rich DNA single

strands bind to an amphipathic protein surface. Sev-
eral residues participating in ligand binding are located
in regions designated as RNP motifs I and II, which
can also be found in other RNA-binding proteins
[26,27].
Here we present the high-resolution crystal structure
Bc-Csp in complex with dT
6
. Unexpectedly, in this
crystal structure, Bc-Csp is present in a domain-
swapped dimeric structure not observed in oligonucleo-
tide ⁄ oligosaccharide binding fold (OB-fold) proteins.
The domain swap pairs one half, b strands 1–3, of
Bc-Csp with the other half, b strands 4 and 5, of a
second molecule and serves as proof of an unantici-
pated structural plasticity in the CSD. In contrast to a
previously determined Bs-CspBÆdT
6
structure [24] in
which the DNA strands bridge adjacent protein mole-
cules in the crystal lattice, each hexathymidine strand
is associated with one CSD. Nevertheless, very similar
ligand-binding subsites are observed in Bs-CspB and
Bc-Csp, and the mode of DNA binding is dominated
by stacking interactions between nucleobases and aro-
matic protein side chains for both proteins. Based on
this observation and on binding assays using a set of
heptapyrimidines, a model of CSPÆheptanucleotide
binding is presented and a common mode of oligonu-
cleotide binding to CSD is proposed.

Results and Discussion
Bc-Csp, the major CSP from B. caldolyticus, was crys-
tallized in complex with dT
6
in space group P2
1
2
1
2,
and diffraction data were collected up to 1.29 A
˚
(Table 1). Initial phases were obtained by molecular
Table 1. Bc-CspÆdT
6
: data collection and refinement statistics.
Data collection
Wavelength (A
˚
) 0.9184
Resolution (A
˚
) 20.00–1.29
Last shell (A
˚
) 1.40–1.29
Space group P2
1
2
1
2

Temperature (K) 100
Detector MAR165 CCD
Unit-cell parameters
a(A
˚
) 74.34
b(A
˚
) 64.89
c(A
˚
) 31.20
Unique reflections (last shell) 37 691 (7838)
I ⁄ r(I) (last shell) 14.8 (5.2)
Data completeness (%) 97.0 (94.4)
R
meas
a
(%) 6.6 (32.0)
Refinement
Resolution (A
˚
) 19–1.29
Working set 34,220
Free set (5%) 1,908
R
work
⁄ R
free
b

(%) 12.9 ⁄ 16.2
Number of nonhydrogen atoms 1,614
Number of protein molecules 2
Number of dT
6
molecules 2
Number of water molecules 234
Mean B factor (A
˚
2
) 10.84
RMSD:
bond lengths (A
˚
) 0.019
bond angles (°) 1.41
torsion angles (°) 4.34
planarity (A
˚
) 0.008
Ramachandran statistics
Residues in allowed regions 95.3
Residues in add. allowed regions 4.7
a
R
meas
, redundancy independent R factor, which correlates intensi-
ties from symmetry related reflections [51].
b
R

work;free
¼
P
jF
obs
jÀjF
calc
j
jF
obs
j
, where the working and free R factors
are calculated using the working and free reflection sets, respect-
ively. The free reflections were held aside throughout refinement.
DNA single-strand binding to the cold shock domain K. E. A. Max et al.
1266 FEBS Journal 274 (2007) 1265–1279 ª 2007 The Authors Journal compilation ª 2007 FEBS
replacement using a crystal structure of Bc-Csp with-
out a ligand (1C9O). The crystal’s asymmetric unit
contains a swapped dimer of Bc-Csp (chains A and B)
in contact with two DNA molecules (chains C and D)
(Fig. 1). Two methylpentanediol molecules from the
crystallization setup, which are associated with the
protein–DNA complex, were included in the structural
model. The structure was refined using refmac
v. 5.1.24 to final R
work
⁄ R
free
values of 13.1 and 16.3%.
The electron density is well defined, all heavy atoms

from protein and ligand molecules could be placed.
Bc-Csp overall structure and domain swap
In the Bc-Csp structure the two protein chains form a
swapped dimer with two globular functional units
which are composed of residues 1–35 from one protein
chain and residues 38–65 from another protein chain
(Figs 1A, 2). The architecture of the functional units
closely resembles that of all other structural models of
CSP, featuring five highly curved antiparallel b strands
connected by four loops and a short 3
10
helix at the C-
terminus of b3. In the nonswapped (closed monomeric)
structures the respective b strands 1–3 and 4+5 are
arranged as two b sheets which form a closed b barrel.
In the domain-swapped structure, the first b sheet of
one chain assembles with a second b sheet from a dif-
ferent chain and vice versa. The swapped chains can
be interrelated by a noncrystallographic twofold rota-
tion axis. The functional units align well with the
structures of the two models of the closed protein
(1C9O) giving RMSD values of < 0.5 A
˚
for all a-car-
bon atoms. Both functional units superimpose with an
rmsd of 0.1 A
˚
, and the phosphorus atoms of the
sugar–phosphate backbone from both DNA chains
superimpose with an rmsd of 0.23 A

˚
. Formation of a
swapped dimer reduces the solvent-accessible surface
per subunit by 5.4% (493 A
˚
2
).
The Bc-Csp domain swap provides insight into
the folding and misfolding of the CSP
The Bc-Csp domain swap occurs in the middle of loop
L
34
and is promoted by a unique combination of
torsion angles Glu36w and Gly37/, compared with
closed monomers. Interestingly, crystal structures of
monomeric Bc-Csp show a two-state conformational
variability at these torsion angles: of 10 protein mod-
els from six different Bc-Csp crystal structures [17,18],
six models display Glu36w ⁄ Gly37/ mean torsion
angles of 162 ± 14°⁄99±5°, designated as state 1.
The other models show mean torsion angles of
A
B
Fig. 1. Crystal structure of The Bc-CspÆdT
6
complex. (A) The DNA strands (red ¼ back-
bone, beige ¼ bases) bind to globular units
of a swapped Bc-Csp dimer (green ¼ chain
A, blue ¼ chain B). Each globular unit is
composed of residues 1–35 and 38–66 of

two different protein chains. The base of
the terminal nucleotide T6 occupies two dif-
ferent positions in chain D (dotted arrow).
(B) The region of the domain swap in chain
B revealed by F
o
) F
c
difference electron
density calculated from a model devoid of
residues 35–38. The map (grey wire frame)
was contoured at 2.5r.
K. E. A. Max et al. DNA single-strand binding to the cold shock domain
FEBS Journal 274 (2007) 1265–1279 ª 2007 The Authors Journal compilation ª 2007 FEBS 1267
)15 ± 15°⁄ )65 ± 8°, designated as state 2
(Fig. 2A,B). In all crystal structures featuring two pro-
tein molecules, one molecule maintains Glu36w ⁄
Gly37/ torsion angles according to state 1, whereas
the other molecule adopts a conformation according to
state 2. All other torsion angles in Bc-Csp represent
single states with standard deviations of < 20°, except
for the terminal residues. The torsion angle differences
between states 1 and 2, 183° for Glu36w and )164° for
Gly37/, roughly compensate for each other and do
not result in noticeable tertiary structural deviations
between monomers in state 1 and state 2. In contrast
to closed Bc-Csp structures, the two polypeptide chains
of the swapped dimer show Glu36w ⁄ Gly37/ torsion
angle combinations of 141 ± 4° (similar to state 1)
and )85 ± 1° (similar to state 2). This results in

effective $ 180° rotations of their main chains at
Glu36w, causing the Bc-Csp structures to open up and
allowing them to re-associate as domain-swapped di-
mers. Apart from differences in the course of the pro-
tein backbone at the point of transition, in the
swapped dimer the overall structure of the functional
units is not altered significantly as compared to the
closed monomers in state 1 and 2.
In the open state, the monomer of Bc-Csp is parti-
tioned into two subdomains of similar length, which
are separated by a long loop. Subdomain 1 is a sheet
including b strands 1–3, subdomain 2 is a b ladder
comprising strands 4 and 5. In closed monomers, these
two subdomains are stabilized by 26 backbone hydro-
gen bonds; the interface between the subdomains con-
tains eight backbone hydrogen bonds (Fig. 3).
A
B
C
Fig. 2. Comparison of open (domain-
swapped) and closed states of Bc-Csp. (A)
Torsion angle distribution of Glu36w (left)
and Gly37/ (right) from 14 closed models of
Bc-Csp and Bs-CspB and two domain-
swapped Bc-Csp molecules (yellow trian-
gles). The closed structures feature a
two-state conformational variability involving
Glu36w ⁄ Gly37/ mean torsion angles of
either 162 ± 14° and 99 ± 5° (state 1, green
squares) or )15 ± 15° and )65 ± 8° (state

2, red circles). The domain-swapped struc-
tures show torsion angles of 141 ± 4° for
Glu36w (similar to state 1) and )85 ± 1° for
Gly37/ (similar to state 2). (B) Superposi-
tions of L
34
residues from Gln34 to Lys39
involving all backbone atoms. (Left) Super-
position of two models featuring
Glu36w ⁄ Gly37/ torsion angle combinations
of state 1 (green) and state 2 (red). (Centre)
Superposition of a model featuring
Glu36w ⁄ Gly37/ torsion angles of state 1
(green) and a domain-swapped structure
(yellow). Residues 34–37 were used for su-
perposition. (Right) Superposition of a model
featuring Glu36w ⁄ Gly37/ torsion angles of
state 2 (red) and a domain-swapped struc-
ture (yellow). Residues 36–39 were used for
superposition. (C) Comparison of an open
monomer CSP structure (yellow) (2HAX,
monomer A) with two closed monomeric
CSP structures (1HZA monomers A & B)
featuring Glu36w ⁄ Gly37/ torsion angles
according to state 1 (green) and state 2
(red).
DNA single-strand binding to the cold shock domain K. E. A. Max et al.
1268 FEBS Journal 274 (2007) 1265–1279 ª 2007 The Authors Journal compilation ª 2007 FEBS
Almost all of these interactions can also be found in
the swapped dimers. From

15
N relaxation and H ⁄ D-
exchange NMR experiments of Bs-CspB, which shares
86% sequence identity and a closely similar tertiary
fold with Bc-Csp, we expect that loop L
34
, which
divides the subdomains, remains flexible even in the
folded state of the protein [25]. A state similar to that
of the open monomer may reflect a substate in the
folding pathway of Bc-Csp. In such an arrangement,
the subdomains may form independently from an
unfolded chain (Fig. 2C). Formation of the subdo-
mains would contribute two thirds of all backbone
hydrogen bonds and stimulate the organization of a
bipartite hydrophobic core, which would be solvent
exposed at this stage. The association of subdomains
results in the formation of the closed b barrel burying
the hydrophobic core. At present there is no experi-
mental evidence for the occurrence of the domain-
swapped form of Bc-Csp or any other CSP in solution
or inside bacterial cells. However, the Bc-Csp crystal
structure reveals an unanticipated structural poly-
morphism. The domain-swapped form of the protein
must be close in energy to the globular monomeric
state, because otherwise these crystals could not have
formed. We cannot discount the possibility that its for-
mation has been overlooked in previous biochemical
studies of CSPs. Further studies are required to evalu-
ate its physiological relevance.

It has been shown that Ec-CspA, which shares 57%
sequence identity with Bc-Csp, aggregates forming
amyloid fibrils under acidic conditions [28]. Analysis of
this amyloid formation using NMR techniques has
revealed time-dependent changes in
15
N T2 relaxation
accompanying the exponential phase of polymeriza-
tion, which suggest that the first three b strands may
form association interfaces that promote aggregate
growth. In the late stage of amyloid formation, signals
from the N-terminal half of the molecules (equivalent
to residues 5–36 in Bc-Csp) appear to be more severely
broadened than those from the C-terminal half. This
may be relevant for folding, because Ec-CspA in this
experiment shows a bipartite organization resembling
that of the open state of Bs-CspB.
Domain swapping of a further b-sheet protein, ribo-
nuclease A, has recently been implicated in amyloido-
genesis [29]. For this enzyme, swapped dimers may be
formed in solution, which can be isolated from closed
monomers using chromatographic techniques [30,31].
A swapped dimer was formed from two different
defective ribonuclease A variants by complementation
involving the swapping of functional subdomains
[29,32]. Using a similar approach, amyloid fibres of
ribonuclease A were generated, for which enzymatic
activity could be demonstrated [29]. It has thus been
suggested that amyloid cross-b spines consisting of
extended b sheets may also be formed from domain-

swapped protein assemblies with retained native struc-
ture. This model of amyloid protofilament formation
may also be relevant for the CSP and may explain
why a bipartite organization can be observed in the
late stage of Ec-CspA amyloidogenesis [28]. To form
extended linear fibrils by swapping subdomains, an
arrangement different to that seen in the Bc-CspÆdT
6
crystal structure, in which two chains form a closed
dimeric arrangement, would be required.
Because the domain-swapped dimer of Bc-Csp has
only been observed in the presence of a bound dT
6
strand, the question remains whether DNA binding to
the CSP is responsible for domain swapping. To date
Fig. 3. Topology plot of Bc-Csp. b-Strands are depicted as blue
(subdomain 1) and green (subdomain 2) arrows. Intra- and intersub-
domain hydrogen bonds from the protein backbone are indicated as
black and grey arrows (donor acceptor). All hydrogen bonds can be
observed in the closed as well as domain-swapped Bc-CspB struc-
tural models. In the latter case, intersubdomain hydrogen bonds
are formed between different protein chains. Residues 36 and 37
are depicted in pink. Their backbone torsion angles display a two-
state conformational variability in closed CSP structures and enable
the domain swap observed in the Bc-CspBÆdT
6
structure.
K. E. A. Max et al. DNA single-strand binding to the cold shock domain
FEBS Journal 274 (2007) 1265–1279 ª 2007 The Authors Journal compilation ª 2007 FEBS 1269
solution data have only indicated the formation of

complexes consisting of one protein and one ligand
molecule for ligands in the range of hexa- or heptanu-
cleotides (data not shown). The Bs-CspBÆdT
6
structure
[24] clearly demonstrates that a domain swap is not
required for ligand binding to the CSP. Most protein–
ligand contacts of the two structures are in good agree-
ment (see below), suggesting that domain swapping
does not change the ligand interaction sites of the
CSP.
Preferential binding of different pyrimidine-based
oligonucleotides to Bc-Csp
It has been shown that homologous Bs-CspB has a
general binding preference for polypyrimidines over
polypurines, and its binding site has been suggested to
interact with 6–7 nucleotides [12–14]. In order to
further analyse the preferential binding of heptanucleo-
tides, we performed binding studies using deoxyhepta-
pyrimidines (Table 2). The selected oligonucleotides
differ from each other only by single pyrimidine bases
and thus allow us to determine the effect of thymine-
to-cytosine base changes at different sequence positions
in a heptanucleotide sequence by relating the K
D
val-
ues of the respective Bc-CspÆoligonucleotide complexes
to each other (Table 2; K
D
1 ⁄ K

D
2). To prevent slippage
of oligonucleotides within the binding site, caused by
single nucleotide changes within homogenous T-rich
surroundings, we used a less degenerate oligonucleo-
tide (CTCTTTC) as a scaffold, which is bound with
similar affinity as dT
7
.
The binding experiments show that there is only a
small preference for T over C at positions 1, 4, 5 and
7, associated with an increase in the K
D
value of up to
threefold. At positions 2 (Table 2, CT3, CT7) and 6
(Table 2, CT3, CT5) the decrease in affinity was signi-
ficantly stronger: When C was introduced at these
positions the K
D
increased 93- or 11-fold. By contrast,
at position 3 (Table 2, CT2, CT3) C was preferred
slightly over T, with an associated 2.5-fold decrease
in K
D
.
dT
6
is bound to a hydrophobic platform on the
protein surface
Globular functional units of Bc-Csp have a strongly

dipolar surface (Fig. 4B). One side has a prominent
negative surface potential which is derived from acidic
side chains. On the opposite side, several solvent-
exposed aromatic side chains form a hydrophobic plat-
form surrounded by basic and by polar groups. This
amphipathic interface associates with dT
6
via various
hydrophobic- and hydrogen-bonding interactions. In
the following description, the interactions between pro-
tein and ligand are observed in the ligand-binding
interfaces of both functional units unless stated other-
wise. Protein groups forming the ligand-binding sur-
face originate from the first three b strands and loop 3;
many are located within the RNP motifs RNP1
(Lys13–Val20) and RNP2 (Val26–Phe30), which are
conserved in various RNA-binding proteins [27,33].
Further groups participating in ligand binding are
located in b5, L
34
and L
45
(Fig. 5).
Oligonucleotide binding to Bc-Csp is dominated by
stacking interactions that involve single stacks between
the side chains of Trp8, Phe17 and Phe27, and the
nucleobases of T6, T5 and T4, respectively. An impres-
sive five-member stack is formed by successive side
chain and nucleobase groups from T3, His29, Phe30,
T1 and Phe38. T2 is contacted through an edge-on

stack by Phe30. Shielding of Val26, Val28 and Tyr15
also may contribute to ligand binding, because the sol-
vent-exposed location of these side chains in the
absence of a ligand is expected to be thermodynamic-
ally unfavourable.
The polar contribution to ligand binding involves
eight hydrogen bonds and five water-mediated interac-
tions between protein and DNA groups. Two hydro-
gen bonds are formed between backbone groups of
Table 2. Relative increases in affinity associated with nucleobase exchanges at individual positions within a heptapyrimidine sequence.
Position oligo 1 K
D
1(nM) oligo 2 K
D
2(nM) K
D
2 ⁄ K
D
1
a
1dT
7
(TTTTTTT) 0.9 ± 0.2 CT1 (CTTTTTT) 2.8 ± 0.9 3.1 ± 0.3
2 CT3 (C
TCTTTC) 3.3 ± 0.2 CT7 (CCCTTTC) 307 ± 33 93 ± 4.4
3 CT2 (CT
TTTTC) 8.3 ± 0.2 CT3 (CTCTTTC) 3.3 ± 0.2 0.4 ± 0.01
4 CT3 (CTC
TTTC) 3.3 ± 0.2 CT6 (CTCCTTC) 3.5 ± 0.6 1.1 ± 0.1
5 CT3 (CTCT

TTC) 3.3 ± 0.2 CT4 (CTCTCTC) 5.2 ± 0.2 1.6 ± 0.1
6 CT3 (CTCTT
TC) 3.3 ± 0.2 CT5 (CTCTTCC) 36.4 ± 4.8 11 ± 0.4
7dT
7
(TTTTTTT) 0.9 ± 0.2 CT9 (TTTTTTC) 1.5 ± 0.2 1.7 ± 0.1
a
The binding propensities for thymine and cytosine at individual positions are compared by relating K
D
values of two different Bc-CspÆoligopyrim-
idine complexes, which differ only at the position of interest (underlined). Their errors were estimated by Gaussion propagation of mean errors.
DNA single-strand binding to the cold shock domain K. E. A. Max et al.
1270 FEBS Journal 274 (2007) 1265–1279 ª 2007 The Authors Journal compilation ª 2007 FEBS
Lys39 and the O
4
and N
3
atoms of T1. Asp25, Lys7
and Trp8 contact N
3
and O
2
of T5 by hydrogen bonds
in a similar way; O
4
of T5 is connected to the Asp25
backbone carbonyl group via a water-mediated inter-
action. The side chain of Gln59 contacts both nucleo-
bases of T3 and T4. Asn10 forms a hydrogen bond
with O

2
of T6. In chain D this interaction is observed
for both conformers of the terminal nucleobase. The
DNA backbone is contacted by a small number of
interactions. One direct hydrogen bond is formed
between the side chain of His29 and the sugar O
4¢
of
T2. In one functional unit, Arg56 interacts with the
phosphate group connecting T3 and T4. This side
chain shows great conformational flexibility in the set
of ligand-free CSP structures and interacts with the nu-
cleobase group in the structure of Bs-CspBÆdT
6
. Upon
ligand binding the solvent-accessible surface from a
functional unit is reduced by 16.2% (696 A
˚
2
), of which
65 and 35% can be assigned to hydrophobic and
hydrophilic areas, respectively.
Structural organization of the ligand
The dT
6
ligand adopts an extended, irregular confor-
mation. Looking from the protein surface towards the
ligand, the sugar–phosphate backbone appears curved
like a ‘C’ (Fig. 6), with the nucleobases pointing
towards the protein surface. The 5¢-to-3¢ polarity of

the ligand follows that of most other nucleic acid com-
plexes of OB-fold proteins, starting in the vicinity of
L
12
, proceeding along the N–C polarity of b2 and
pointing towards the kink in b1 (Fig. 4A). There is no
stacking between nucleobases, and all nucleosides are
in the anti conformation. The solvent-exposed sugar–
phosphate backbone shields the hydrophobic nucleo-
bases and the hydrophobic-binding platform of the
protein below them from the polar solvent (Fig. 2B).
The sugar of T1 from chain C maintains a C
4¢
-exo
pucker, whereas the remaining sugars adopt C
2¢
-endo
puckers, which are typical of double-stranded B-DNA.
In ligand chain D, the terminal nucleotides adopt a
C
3¢
-endo conformation, which is typical of double-
stranded A-DNA and RNA, whereas all other nucleo-
tides display C
2¢
-endo puckers. All sugar puckers
observed in the Bc-CspÆdT
6
structure are within ener-
getically favourable regions of the pentose pseudorota-

tion cycle [34] and the exocyclic angles of the sugar–
phosphate backbone are within limits observed in
tRNA structures [35].
Assignment of seven common interaction
subsites to Bc-Csp and Bs-CspB
In the Bc-CspÆdT
6
complex, each DNA molecule binds
to one globular functional unit of a swapped dimer. In
contrast to the swapped dimer complex of Bc-CspÆdT
6
,
closed protein monomers and ligand molecules
form an interspersed arrangement in the related
Bs-CspBÆdT
6
complex [24]. (Figures 7 and 8 give a
schematical and structural comparison of both hexa-
thymidine complex structures.) Many interactions
between protein and DNA ligand molecules are com-
mon to both structures, yet certain interactions can
only be observed with either Bs-CspB or Bc-Csp.
Based on the two structures, we can now define a com-
mon interaction interface that allows us to understand
A
B
Fig. 4. Binding of hexathymidine to amphipathic platforms of a Bc-
Csp swapped dimer. (A) Topological representation of a functional
unit of the swapped dimer associated with a single dT
6

molecule.
(B) Electrostatic surface potential of a Bc-Csp functional unit. All fig-
ures were drawn using
PYMOL [52], the electrostatic surface poten-
tial was calculated with
APBS [53] for pH 7 with a range from )10
(red) to +10 kT (blue).
K. E. A. Max et al. DNA single-strand binding to the cold shock domain
FEBS Journal 274 (2007) 1265–1279 ª 2007 The Authors Journal compilation ª 2007 FEBS 1271
how Bs-CspB and Bc-Csp interact with thymine-rich
heptanucleotides (see Fig. 5 for Bc-Csp).
In the Bc-CspÆdT
6
structure, interaction subsite 1
remains empty. In the Bs-CspBÆdT
6
complex, an edge-
on stack between Phe38 and a nucleobase is observed
at this subsite. In contact subsite 2, Phe30 and Phe38
form a three-membered stack with the T1 base. This
base is specifically bound via two hydrogen bonds to
A
B
Fig. 5. Hydrophobic and polar interactions
between dT
6
and Bc-Csp. (A) Stereoscopic
representation. The contact surface of Bc-
Csp is shown as a semitransparent grey
object, protein groups involved in stacking

interactions and hydrogen bonding are col-
oured according to CPK with the exception
of carbon which is green (monomer A) and
light blue (monomer B). Hydrogen bonds
between protein and DNA groups are depic-
ted as dotted lines. (B) Schematic overview
of intermolecular interactions: DNA (black)
and protein groups (grey) interact through
stacking interactions (solid lines) and hydro-
gen bonds (dashed lines). Some contacts
are mediated by water molecules (circles).
Interactions observed in only one functional
unit of the structure are in light grey,
whereas a common set of interactions also
observed in the Bs-CspB crystal structure is
highlighted in red. Nucleobase binding sub-
sites (numbers below the schemes) are
defined as discussed in the text. Subsites
not occupied by bases are parenthesized.
The numbers of the contact subsites for
individual nucleobases are given at the
bottom.
Fig. 6. DNA single strands adopt an irregular conformation upon binding to Bc-Csp. The sugar–phosphate backbone appears curved like a ‘C’
character. All nucleobases are unstacked with respect to each other. The nucleotides are in anti conformation. The stereo view is from the
protein surface towards DNA strand D. The DNA is surrounded by its 2F
o
) F
c
difference density contoured at 1.2 r.
DNA single-strand binding to the cold shock domain K. E. A. Max et al.

1272 FEBS Journal 274 (2007) 1265–1279 ª 2007 The Authors Journal compilation ª 2007 FEBS
the backbone of Lys39 in a geometry reminiscent of a
Watson–Crick TA base pair. The existence of a third
contact site has been hypothesized in the Bs-CspB
structure based on missing electron density for the 5¢
nucleotide, which was expected to be located at this
position. In Bc-CspÆdT
6
, a third subsite was found as
anticipated. It involves the side chain of His29, which
reveals an edge-on contact with the nucleobase and a
hydrogen bond with the deoxyribose ring oxygen of
the nucleoside in subsite 3. In contact subsites 4 and 5,
the side chains of His29 and Phe27 stack with bases
from adjacent nucleotides. Gln59 contacts their nucleo-
base head groups via hydrogen bonds. In the
Bc-CspÆdT
6
structure an additional hydrogen bond
provided by the backbone carbonyl group of Pro58
contacts the nucleobase. In subsite 6, a nucleobase
stacks against Phe17 while its head groups interact
with the side chains of Asp25, Lys7 and Trp8 via
hydrogen bonds. Interestingly, the orientations of the
nucleobases in this subsite differ between the oligo-
thymidine complexes of Bc-Csp and Bs-CspB. They
may be related by a 180° rotation. Consequently, O
2
is
contacted by Lys7 and Trp8 in the Bc-CspÆdT

6
complex structure instead of O
4
as observed in the
Bs-CspBÆdT
6
structure.
The most prominent difference between the two
CSPÆoligothymidine complexes involves interaction
subsite 7. In the Bc-Csp structure, the 3¢ nucleotide
stacks against Trp8, its O
2
is contacted by Asn10. In
the Bs-CspB structure, Trp8 is inaccessible due to a
crystal contact. An alternative seventh binding site was
attributed to a hydrogen bond between Arg56 and the
O
2
of the nucleobase. However, after evaluating both
crystal structures we conclude that the alternative ori-
entation of the base and sugar–phosphate backbone of
the nucleotide in subsite 6 and the formation of the
alternative subsite 7 are a consequence of the inaccessi-
bility of Trp8 in this crystal form.
In addition to their structures, Bc-Csp and Bs-CspB
also share functional similarities. Both proteins bind
dT
7
with a similar affinity of K
D

values of 0.9 ± 0.2
1.8 ± 0.2 nm (Table 2) [24]. In solution, their highest
preference for T was observed for positions 2 and 6 in
a heptanucleotide. Replacement of T by C at these
positions results in significantly decreased affinities as
observed by 93- and 11-fold increased K
D
values for
Bc-Csp. These distinct preferences are in good agree-
ment with the deduced binding mode, in which the
most specific contacts for thymine head groups are
formed at nucleobase subsites 2 and 6.
Fig. 7. Schematic overview of CSPÆoligonucleotide interactions. Protein molecules (grey) interact with bases from oligonucleotides at distinct
binding subsites. (A) In the Bc-CspÆdT
6
crystal structure, two protein chains (light and dark grey) form two functional units each of which
binds a DNA molecule. (B) In the Bs-CspBÆdT
6
crystal [24], a continuous arrangement of protein and DNA molecules is formed. A gap
between the 3’ nucleotide (bound to subsite 2) and the first structured 5’ nucleotide (bound to subsite 4) exists, which is expected to bind
the unstructured T1 nucleotide (grey). (C) In solution, all seven subsites are occupied by a single oligonucleotide molecule.
K. E. A. Max et al. DNA single-strand binding to the cold shock domain
FEBS Journal 274 (2007) 1265–1279 ª 2007 The Authors Journal compilation ª 2007 FEBS 1273
T-to-C changes at positions 4, 5 and 7 did not signi-
ficantly influence the affinity of dT
7
for Bc-Csp,
whereas in binding studies with Bs-CspB weak prefer-
ences for T were observed. At sequence position 3, a
cytosine base is preferred to thymine by both CSPs.

Although the preference of Bc-Csp for individual nu-
cleotides at most positions was not as pronounced as
with Bs-CspB, they clearly follow the same trend. The
smaller base discrimination by Bc-Csp may be related
to the fact that all fluorescence titration measurements
were performed at 15 °C to allow comparison.
Although for B. subtilis this temperature is close to its
growth conditions, the temperature optimum for
B. caldolyticus is more than 40 °C higher.
Functional implications
The common features of most nucleobase interaction
subsites suggest that both Bc-Csp and Bs-CspB share
identical ligand-binding interfaces (Fig. 8), whereas dif-
ferences in binding involving subsites 6 and 7, as well
as the arrangement of protein and DNA molecules,
appear to arise from different crystal-packing environ-
ments. Despite the fact that heptanucleotides rather
than hexanucleotides fully occupy the CSP binding site
[12,25], we have not yet found suitable crystallization
conditions for CSP in complex with heptadeoxynucleo-
tides. Likewise, attempts to grow crystals of Bs-CspB
in the presence of oligoribonucleotides have remained
unsuccessful. In contrast to the individual Bs-CspB
and Bc-CspÆhexathymidine complex structures, the
combined structural models allow us to understand
how both CSPs bind thymine-rich heptanucleotide
motifs and explain binding preferences seen in bio-
chemical binding studies in solution.
Although the CSPÆdT
6

crystal structures contain an
single-strand DNA ligand, they support the assump-
tion that single-strand RNA ligands bind the same
way, because the exposed sugar 2¢OH groups and the
missing methyl groups of uridines would not enhance
or impair ligand binding, and the backbone torsion
angles of the DNA ligands are compatible with data
obtained from tRNA crystal structures. The extended
irregular conformation of dT
6
oligonucleotides in the
binding site, the unstacking of bases with respect to
each other upon binding, and the shielding of the nu-
cleobase head groups by the protein suggest that the
CSP may counteract double-strand formation in
nucleic acids. CSP surface properties favour the bind-
ing of thymine-rich sequences, with the exception of
nucleobase binding subsite 2, which favours cytosine.
The biological functions of CSP are still under inves-
tigation. Certain CSP have initially been reported to
function as transcriptional activators of cold-induced
genes such as hns [36] and gyrA, encoding a subunit of
DNA gyrase [37]. The ability of CSP to bind to single-
stranded nucleic acids and prevent their association to
double strands in vitro led to the assumption that these
proteins may function as RNA chaperones [4], which
may prevent the formation of mRNA double strands
Fig. 8. Stereoview of the structurally con-
served CSP ligand-binding surface. Struc-
tures of the nucleobase ligands from the

Bs-CspBÆdT
6
(grey, black font) and Bc-
CspÆdT
6
(green) complexes have been shif-
ted (upper) to allow a better view on the
CSP binding site (lower). Lines schemati-
cally relate individual nucleobases to their
position in the complex structures. Protein
groups involved in nucleobase interactions
in the complex structures are shown as
sticks in corresponding colours. Their equiv-
alents from CSP structures without an oligo-
nucleotide ligand are displayed as blue lines.
DNA single-strand binding to the cold shock domain K. E. A. Max et al.
1274 FEBS Journal 274 (2007) 1265–1279 ª 2007 The Authors Journal compilation ª 2007 FEBS
at low temperatures. Whether mRNA base pairing
indeed impairs gene expression in the cold and whether
CSP binding can prevent this remains to be shown. It
is clear, however, that the CSP are required for cold
adaptation. Strains of B. subtilis, with two of three
CSP genes inactivated show severe restrictions in
growth at low temperatures [3]. Likewise, an Escheri-
chia coli strain with four (of nine) CSP genes inacti-
vated is cold sensitive and impaired in upregulation of
many cold-induced genes [38]. For RNA chaperone
activity, a low-affinity sequence-unspecific binding of
mRNA by CSP may be sufficient to prevent mispaired
mRNA double-strand formation in the cold. It

remains to be seen, however, which role the preferen-
tial binding of CSP to uracil-rich sequences may have
in this context.
In addition to the presumed RNA chaperone activ-
ity there is good evidence that some E. coli CSP func-
tion as transcriptional antiterminators of cold-
regulated polycistronic genes [5]. Some post-terminator
cistrons have been attributed to cold regulation, and
the amount of their mRNA was shown to increase
upon either a down-shift in temperature or the overex-
pression of certain CSP genes [39,40]. Transcription
termination occurs at terminator sites on the 3¢-end of
a mRNA transcript which contains two inverted,
guanine- and cytosine-rich sequence motifs that can
form a stem–loop. This stem–loop is followed on the
downstream side by five or more uridines. A current
model of transcription termination assumes the disso-
ciation of a nascent transcript from the DNA upon
stem–loop formation at the terminator site. When the
adjacent uracil-rich sequence is being transcribed, the
affinity of the dA–U base pairs is insufficient to stabil-
ize the DNA–RNA duplex, and the transcript,
together with the RNA polymerase, disassembles from
the template. It is expected that the obstruction of
stem–loop formation prevents termination and leads
to transcription of cistrons on the 3¢ side of termin-
ation sites. The ability of the CSP to prevent double-
strand formation in RNA may destabilize terminator
stem–loops and promote transcription antitermination
[5]. We suggest that the preference of CSP binding for

uracil-rich sequences may direct the CSP to terminator
sites.
Ligand-binding interfaces from CSP and Y-box
proteins are highly conserved
Remarkably, all side chains involved in ligand binding
in the CSPÆdT
6
crystal structures are located and orien-
ted almost invariantly in the absence of a ligand
(Fig. 8). The nucleotide-binding site of the CSP thus
appears to be a conserved preformed platform, which
does not undergo major reorientations upon ligand
binding. A similar degree of conservation can also be
found in the sequences of CSP homologues from other
bacteria (Fig. 9). Individual amino acids aligned to res-
idues from the binding interface are conserved at a
level of at least 75% similarity, whereas amino acids
aligned with residues from other surface areas do not
show a similar degree of conservation. This suggests
that binding modes and preferences of homologous
CSP in bacteria are very similar and may be attributed
to the same function, a finding that has already been
demonstrated for the CSP paralogues of B. subtilis and
E. coli [3,41].
Apart from eubacteria, proteins with CSD can also
be found in eukaryotic proteins. A high degree of
sequence identity (> 45%) was reported between the
CSP and the nucleic acid-binding domains of the
eukaryotic Y-box factors [10]. Y-Box proteins have
been implicated in transcriptional activation and

repression, regulation of alternative splicing, regula-
tion of mRNA stability, translational activation or
repression and RNA packaging. It was also shown
recently that a Y-box protein was required for
growth of cultured avian cells at low temperatures
[42]. Most residues involved in preferential ligand
binding in the two CSPÆdT
6
crystal structures can
also be found in the sequences of Y-box proteins,
with the exception of side chains forming nucleobase-
binding subsites 1 and 2 (Fig. 9B). We therefore
expect that the Y-box proteins and the CSP not only
have a common ancestor but also share a closely sim-
ilar DNA-binding mode.
Experimental procedures
Data collection and processing
Bc-Csp was purified, and the Bc-CspÆ dT
6
complex was
formed and crystallized as previously described [43]. A
crystal was frozen in liquid nitrogen, and X-ray diffrac-
tion data were collected at a wavelength of 0.9184 A
˚
at
the Protein Structure Factory beamline BL 14.1 [44] of
the Free University of Berlin at BESSY (Berlin, Ger-
many) using a MAR165 CCD camera. A complete data
set was collected to a maximal resolution of 1.29 A
˚

. The
xds package [45] was used to integrate reflection intensi-
ties. The quality of the data set is summarized in Table 1.
The Bc-CspÆdT
6
complex crystallized in space group
P2
1
2
1
2 with unit cell parameters of a ¼ 74.34 A
˚
, b ¼
64.89 A
˚
, c ¼ 31.20 A
˚
. Two protein and two DNA mole-
cules were found per asymmetric unit. The solvent con-
tent was 39%.
K. E. A. Max et al. DNA single-strand binding to the cold shock domain
FEBS Journal 274 (2007) 1265–1279 ª 2007 The Authors Journal compilation ª 2007 FEBS 1275
Structure determination, model building
and refinement
The phase problem could be solved via molecular replace-
ment using amore software [46] and free Bc-Csp (1C9O) as
a template. The structure was refined using refmac5 [47]
version 5.1.24. Five per cent of the reflections were set aside
for cross-validation, and R
free

was used to adjust the refine-
ment strategy and monitor progress. After some initial steps
of rigid-body, positional and B-factor refinement, negative
F
o
) F
c
difference density indicated an alternative progres-
sion of the backbone around residues 35–38. Additional
positive F
o
) F
c
density revealed the ligand molecules. The
model was progressively extended according to difference
density. Two hundred and thirty-six water molecules were
added using both automated methods [48] and manual
inspection of difference maps. The refinement converged at
R
free
and R
work
values of 13.0 and 16.2%, respectively. The
A
B
Fig. 9. The ligand-binding surfaces of CSP and Y-box proteins are highly conserved. (A) CSP sequence conservation mapped to the surface
of Bc-Csp bound to dT
6
. Most residues forming the ligand interaction site are conserved on the level of at least 75% identity (dark green)
and similarity (light green). Invariant regions which originate from the protein backbone are coloured light blue. The monomeric structure

shown in this figure represents a globular unit from a swapped Bc-Csp dimer. The model of heptanucleotide ligand (sticks) was assembled
by combining structural information of the Bs-CspBÆdT
6
and Bc-CspÆdT
6
complex structures. (B) Sequence alignment of CSP (upper) and Y-
box proteins (lower). Residues which are conserved at a level of at least 75% identity or similarity are highlighted in black or grey, respect-
ively. Side chains involved in DNA binding in the Bc-Csp crystal structure are marked by triangles (blue ¼ polar contacts, yellow ¼ hydropho-
bic contacts, green ¼ both).
DNA single-strand binding to the cold shock domain K. E. A. Max et al.
1276 FEBS Journal 274 (2007) 1265–1279 ª 2007 The Authors Journal compilation ª 2007 FEBS
resulting structural model and the experimental data of the
Bc-CspÆdT
6
complex has been deposited in the Protein Data
Bank under PDB ID 2HAX.
Fluorescence titration of Bc-Csp variants
with oligonucleotides
Fluorescence of Trp8 from Bc-Csp was measured using a
JASCO FP-6500 spectrofluorimeter (JASCO Ltd., Great
Dunmow, UK) equipped with a thermostat cell holder
attached to a water bath. All experiments were carried out
at 15 °Cin50mm cacodylate ⁄ HCl, 100 mm KCl, pH 7.0,
supplemented with 40 nm N-acetyl-tryptophan-amide. To
determine dissociation constants, a 20 nm protein solution
was titrated with increasing amounts of oligonucleotide in
the range 0.25 nm to 20 lm depending on complex affinity.
The samples were stirred gently during titration. After
2 min of equilibration, the fluorescence of Trp8 was excited
at 280 nm and measured at 343 nm. The initial sample vol-

ume was 1.75 mL and the increase in volume after succes-
sive addition of highly concentrated oligonucleotide
solution was below 4%. The fluorescence was corrected for
inner-filter effects, buffer fluorescence, oligonucleotide auto-
fluorescence and dilution. The changes in the fluorescence
reveal binding isotherms, which were analysed according to
the binding equation [49,50]:
Q ¼ Q
max
Á
A À
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
A
2
À 4n Á½P
0
Á½L
0
p
2 Á½P
0
;
with
A ¼ K
D
þ½P
0
þ n Á½N
0
where Q is the quenching of the intrinsic fluorescence inten-

sity of Trp8 after each addition of oligonucleotide. Q
max
represents the maximal quenching upon complete satura-
tion of the protein with ssDNA. [P]
0
and [Y]
0
are the over-
all Bc-Csp and oligonucleotide concentration, respectively.
n is the number of oligonucleotide strands bound to one
Bc-Csp molecule, and K
D
is the equilibrium dissociation
constant of the complex. The binding model does not take
the degeneracy of multiple binding sites within one single-
strand DNA into account and therefore the obtained
results represent apparent equilibrium constants.
Alignment of CSD sequences
A BLAST search was performed to find homologous
sequences to Bc-Csp. The closest 250 hits were aligned and
used to define a profile, which was trained by two further
cycles of searching ⁄ alignment. The 250 hits reported after
the fourth search cycle were used for a multiple sequence
alignment. Thirteen sequences with the highest degree of
divergence from the profile were added to the sequences of
Bc-Csp and Bs-CspB and arranged for display in Fig. 9B.
The level of conservation was calculated for each sequence
position and shaded at a level of 75% sequence identity
(black) or similarity (grey) using bioedit (Ibis Therapeutics,
Carlsbad, CA, USA).

Acknowledgements
We are grateful to A. Feske (MDC, Berlin, Germany)
for help with protein crystallization, and to the staff at
BL 14.1 ⁄ BESSY (Berlin, Germany) for help with
X-ray diffraction experiments. This work was supported
by a Kekule
´
fellowship of the Fonds der Chemischen
Industrie to KM and the Deutsche Forschungsge-
meinschaft.
References
1 Jones PG, VanBogelen RA & Neidhardt FC (1987)
Induction of proteins in response to low temperature in
Escherichia coli. J Bacteriol 169, 2092–2095.
2 Graumann P, Schroder K, Schmid R & Marahiel MA
(1996) Cold shock stress-induced proteins in Bacillus
subtilis. J Bacteriol 178, 4611–4619.
3 Graumann P, Wendrich TM, Weber MH, Schroder K
& Marahiel MA (1997) A family of cold shock proteins
in Bacillus subtilis is essential for cellular growth and
for efficient protein synthesis at optimal and low tem-
peratures. Mol Microbiol 25, 741–756.
4 Jiang W, Hou Y & Inouye M (1997) CspA, the major
cold-shock protein of Escherichia coli, is an RNA cha-
perone. J Biol Chem 272, 196–202.
5 Bae W, Xia B, Inouye M & Severinov K (2000)
Escherichia coli CspA-family RNA chaperones are
transcription antiterminators. Proc Natl Acad Sci USA
97, 7784–7789.
6 Graumann PL & Marahiel MA (1998) A superfamily of

proteins that contain the cold-shock domain. Trends
Biochem Sci 23 , 286–290.
7 Phadtare S & Inouye M (1999) Sequence-selective inter-
actions with RNA by CspB, CspC and CspE, members
of the CspA family of Escherichia coli. Mol Microbiol
33, 1004–1014.
8 Weber MH, Volkov AV, Fricke I, Marahiel MA &
Graumann PL (2001) Localization of cold shock pro-
teins to cytosolic spaces surrounding nucleoids in Bacil-
lus subtilis depends on active transcription. J Bacteriol
183, 6435–6443.
9 Mascarenhas J, Weber MH & Graumann PL (2001)
Specific polar localization of ribosomes in Bacillus
subtilis depends on active transcription. EMBO Rep 2,
685–689.
10 Graumann P & Marahiel MA (1994) The major cold
shock protein of Bacillus subtilis CspB binds with
high affinity to the ATTGG- and CCAAT sequences
K. E. A. Max et al. DNA single-strand binding to the cold shock domain
FEBS Journal 274 (2007) 1265–1279 ª 2007 The Authors Journal compilation ª 2007 FEBS 1277
in single stranded oligonucleotides. FEBS Lett 338,
157–160.
11 Lopez MM & Makhatadze GI (2000) Major cold
shock proteins, CspA from Escherichia coli and CspB
from Bacillus subtilis, interact differently with single-
stranded DNA templates. Biochim Biophys Acta 1479,
196–202.
12 Lopez MM, Yutani K & Makhatadze GI (1999) Inter-
actions of the major cold shock protein of Bacillus
subtilis CspB with single-stranded DNA templates of

different base composition. J Biol Chem 274, 33601–
33608.
13 Lopez MM, Yutani K & Makhatadze GI (2001) Inter-
actions of the cold shock protein CspB from Bacillus
subtilis with single-stranded DNA. Importance of the T
base content and position within the template. J Biol
Chem 276, 15511–15518.
14 Zeeb M & Balbach J (2003) Single-stranded DNA
binding of the cold-shock protein CspB from Bacillus
subtilis: NMR mapping and mutational characteri-
zation. Protein Sci 12, 112–123.
15 Schindelin H, Marahiel MA & Heinemann U (1993)
Universal nucleic acid-binding domain revealed by crys-
tal structure of the B. subtilis major cold-shock protein.
Nature 364, 164–168.
16 Schindelin H, Jiang W, Inouye M & Heinemann U
(1994) Crystal structure of CspA, the major cold shock
protein of Escherichia coli. Proc Natl Acad Sci USA 91,
5119–5123.
17 Mueller U, Perl D, Schmid FX & Heinemann U (2000)
Thermal stability and atomic-resolution crystal structure
of the Bacillus caldolyticus cold shock protein. J Mol
Biol 297, 975–988.
18 Delbruck H, Mueller U, Perl D, Schmid FX & Heine-
mann U (2001) Crystal structures of mutant forms of
the Bacillus caldolyticus cold shock protein differing in
thermal stability. J Mol Biol 313, 359–369.
19 Schnuchel A, Wiltscheck R, Czisch M, Herrler M, Wil-
limsky G, Graumann P, Marahiel MA & Holak TA
(1993) Structure in solution of the major cold-shock

protein from Bacillus subtilis. Nature 364, 169–171.
20 Kremer W, Schuler B, Harrieder S, Geyer M, Gronwald
W, Welker C, Jaenicke R & Kalbitzer HR (2001) Solu-
tion NMR structure of the cold-shock protein from the
hyperthermophilic bacterium Thermotoga maritima. Eur
J Biochem 268, 2527–2539.
21 Agrawal V & Kishan KV (2003) OB-fold: growing
bigger with functional consistency. Curr Protein Peptide
Sci 4, 195–206.
22 Murzin AG (1993) OB (oligonucleotide
⁄ oligosaccharide
binding)-fold: common structural and functional solu-
tion for non-homologous sequences. EMBO J 12, 861–
867.
23 Wolffe AP, Tafuri S, Ranjan M & Familari M (1992)
The Y-box factors: a family of nucleic acid binding pro-
teins conserved from Escherichia coli to man. New Biol
4, 290–298.
24 Max KEA, Zeeb M, Bienert R, Balbach J & Heinemann
U (2006) T-Rich DNA single strands bind to a pre-
formed site on the bacterial cold shock protein Bs-
CspB. J Mol Biol 360, 702–714.
25 Zeeb M, Max KEA, Weininger U, Lo
¨
w C, Sticht H &
Balbach J (2006) Recognition of T-rich single-stranded
DNA by the cold shock protein Bs-CspB in solution.
Nucleic Acids Res 34, 4561–4571.
26 Messias AC & Sattler M (2004) Structural basis of sin-
gle-stranded RNA recognition. Acc Chem Res 37, 279–

287.
27 Schroder K, Graumann P, Schnuchel A, Holak TA &
Marahiel MA (1995) Mutational analysis of the puta-
tive nucleic acid-binding surface of the cold-shock
domain, CspB, revealed an essential role of aromatic
and basic residues in binding of single-stranded DNA
containing the Y-box motif. Mol Microbiol 16, 699–
708.
28 Alexandrescu AT & Rathgeb-Szabo K (1999) An NMR
investigation of solution aggregation reactions preceding
the misassembly of acid-denatured cold shock protein A
into fibrils. J Mol Biol 291, 1191–1206.
29 Sambashivan S, Liu Y, Sawaya MR, Gingery M &
Eisenberg D (2005) Amyloid-like fibrils of ribonuclease
A with three-dimensional domain-swapped and native-
like structure. Nature 437, 266–269.
30 Fruchter RG & Crestfield AM (1965) Preparation and
properties of two active forms of ribonuclease dimer.
J Biol Chem 240, 3868–3874.
31 Gotte G, Bertoldi M & Libonati M (1999) Structural
versatility of bovine ribonuclease A. Distinct conformers
of trimeric and tetrameric aggregates of the enzyme. Eur
J Biochem 265, 680–687.
32 Crestfield AM, Stein WH & Moore S (1963) Properties
and conformation of the histidine residues at the active
site of ribonuclease. J Biol Chem 238, 2421–2428.
33 Schindler T, Perl D, Graumann P, Sieber V, Marahiel
MA & Schmid FX (1998) Surface-exposed phenylala-
nines in the RNP1 ⁄ RNP2 motif stabilize the cold-shock
protein CspB from Bacillus subtilis. Proteins 30, 401–

406.
34 Levitt M & Warshel A (1978) Extreme conformational
flexibility of the furanose ring in DNA and RNA. JAm
Chem Soc 100, 2607–2613.
35 de Leeuw HPM, Haasnoot CAG & Altona C (1980)
Empirical correlations between conformational para-
meters in a-d-furanoside fragments derived from a sta-
tistical survey of crystal structures of nucleic acid
constituents. Full description of nucleoside molecular
geometries in terms of four parameters. Isr J Chem 20
,
108–126.
36 Graumann P & Marahiel MA (1997) Effects of hetero-
logous expression of CspB, the major cold shock pro-
DNA single-strand binding to the cold shock domain K. E. A. Max et al.
1278 FEBS Journal 274 (2007) 1265–1279 ª 2007 The Authors Journal compilation ª 2007 FEBS
tein of Bacillus subtilis, on protein synthesis in Escheri-
chia coli. Mol Gen Genet 253, 745–752.
37 Jones PG, Krah R, Tafuri SR & Wolffe AP (1992)
DNA gyrase, CS7.4, and the cold shock response in
Escherichia coli. J Bacteriol 174, 5798–5802.
38 Phadtare S & Inouye M (2004) Genome-wide transcrip-
tional analysis of the cold shock response in wild-type
and cold-sensitive, quadruple-csp-deletion strains of
Escherichia coli. J Bacteriol 186, 7007–7014.
39 Jones PG & Inouye M (1994) The cold-shock response
– hot topic. Mol Microbiol 11, 811–818.
40 Sands JF, Regnier P, Cummings HS, Grunberg-Manago
M & Hershey JW (1988) The existence of two genes
between infB and rpsO in the Escherichia coli genome:

DNA sequencing and S1 nuclease mapping. Nucleic
Acids Res 16 , 10803–10816.
41 Xia B, Ke H & Inouye M (2001) Acquirement of cold
sensitivity by quadruple deletion of the cspA family and
its suppression by PNPase S1 domain in Escherichia
coli. Mol Microbiol 40, 179–188.
42 Matsumoto K, Tanaka KJ & Tsujimoto M (2005) An
acidic protein, YBAP1, mediates the release of YB-1
from mRNA and relieves the translational repression
activity of YB-1. Mol Cell Biol 25, 1779–1792.
43 Bienert R , Zeeb M, Dostal L, Feske A, Max KEA,
Welfle H, Balbach J & Heinemann U (2004) Single-
stranded DNA bound to bacterial cold-shock proteins:
preliminary crystallographic and Raman analysis. Acta
Crystallogr D Biol Crystallogr 60, 755–757.
44 Heinemann U, Bussow K, Mueller U & Umbach P
(2003) Facilities and methods for the high-throughput
crystal structural analysis of human proteins. Acc Chem
Res 36, 157–163.
45 Kabsch W (1993) Automatic processing of rotation dif-
fraction data from crystals of initially unknown sym-
metry and cell constants. J Appl Crystallogr 26, 795–
800.
46 Navaza J (2001) Implementation of molecular replace-
ment in AMoRe. Acta Crystallogr D Biol Crystallogr
57, 1367–1372.
47 Murshudov GN, Vagin AA & Dodson EJ (1997)
Refinement of macromolecular structures by the
maximum-likelihood method. Acta Crystallogr D Biol
Crystallogr 53, 240–255.

48 Lamzin VS, Perrakis A & Wilson KS (2001) The
ARP ⁄ wARP suite for automated construction and
refinement of protein models. Int Tables Crystallogr F
720–722.
49 Lohman TM & Bujalowski W (1991) Thermodynamic
methods for model-independent determination of equili-
brium binding isotherms for protein–DNA interactions:
spectroscopic approaches to monitor binding. Methods
Enzymol 208, 258–290.
50 Eftink MR (1997) Fluorescence methods for studying
equilibrium macromolecule–ligand interactions. Methods
Enzymol 278, 221–257.
51 Diederichs K & Karplus PA (1997) Improved R-factors
for diffraction data analysis in macromolecular crystal-
lography. Nat Struct Biol 4, 269–275.
52 DeLano WL (2003) The PyMOL Molecular Graphics
System. DeLano Scientific LLC, San Carlos, CA. http://
www.pymol.org.
53 Baker NA, Sept D, Joseph S, Holst MJ & McCammon
JA (2001) Electrostatics of nanosystems: application to
microtubules and the ribosome. Proc Natl Acad Sci
USA 98, 10037–10041.
K. E. A. Max et al. DNA single-strand binding to the cold shock domain
FEBS Journal 274 (2007) 1265–1279 ª 2007 The Authors Journal compilation ª 2007 FEBS 1279