Helix mobility and recognition function of the rat thyroid
transcription factor 1 homeodomain – hints from
15
N-NMR
relaxation studies
Devrim Gu
¨
mral, Luana Nadalin, Alessandra Corazza, Federico Fogolari, Giuseppe Damante,
Paolo Viglino and Gennaro Esposito
Dipartimento di Scienze e Tecnologie Biomediche, Universita
`
di Udine, Italy
Homeodomains (HDs) comprise a very well-known
class of DNA-binding domains occurring in a large
family of transcription activators involved in the
determination of cell development [1–3]. The tertiary
structure of the HD of rat thyroid transcription fac-
tor 1 (TTF-1), a 67-residue domain, was determined
by NMR spectroscopy [4] (Brookhaven Protein Data
Bank ID code 1FTT). The whole TTF-1 protein
(378 residues) is responsible for transcriptional activa-
tion of genes expressed only in follicular thyroid cells
[5] and lung epithelial cells [6]. The structural fea-
tures of the TTF-1 HD are the typical ones observed
in HDs, i.e. three helices (Gln10–Gln22, Ala28–Ile38,
Thr43–Gln59) connected by a loose loop (Gln23–
Ser27) between helix I and helix II and by a tight
turn (His39–Pro42) between helix II and helix III
(helix–turn–helix motif; Fig. 1). The DNA recogni-
tion helix (helix III) is fairly ordered also in the
Keywords

steady-state {
1
H}–
15
N NOEs were measured at 11.7 T. These data were
analyzed by both the model-free formalism and the reduced spectral den-
sity mapping (RSDM) approaches. The global rotational correlation time,
s
m
, of the thyroid transcription factor 1 homeodomain in aqueous solution
at 286 K was found to be 10.51 ± 0.05 ns by model-free formalism and
9.85 ± 1.79 ns by RSDM calculation. The homogeneity of the values of
the overall correlation time calculated from the individual (R
2
⁄ R
1
) ratios
suggested a good degree of isotropy of the global molecular motion, consis-
tent with the similar global s
m
results obtained with the two different meth-
ods. Tyr25 was found to undergo slow conformational exchange by both
methods, whereas this contribution was identified also for Lys21, Gln22,
Ile38 and His52 only by RSDM. With both methods, the C-terminal frag-
ment of helix III was found to be more flexible than the preceding N-termi-
nal portion, with slightly different limits between rigid and mobile moieties.
Additionally, Arg53 appeared to be characterized by an intermediate
motional freedom between the very flexible N-terminal and C-terminal resi-
dues and the structured core of the molecule, suggesting the occurrence of
a hinge point. Finally, slow-time-scale motions observed at the end of

tional dynamics of the HD recognition helix upon
DNA binding [11].
In the following, we present a
15
N-NMR relaxation
study of the rat TTF-1 HD to address the backbone
dynamics in solution.
15
N-NMR as well as
13
C-NMR
relaxation studies can be usefully applied to determine
the dynamics of proteins [12,13]. In high magnetic
fields, the relaxation of these nuclei is mainly governed
by dipole–dipole and chemical shift anisotropy mecha-
nisms. For globular proteins, the analysis of the exper-
imental relaxation data by means of the model-free
(MF) approach [14,15] provides a description of the
motions in terms of global overall rotational correla-
tion time, s
m
, a generalized order parameter, S
2
, and
an effective internal correlation time, s
e
. For
15
N relax-
ation data, the generalized order parameter reflects the

Relaxation parameters
The individual R
1
, R
2
and NOE values of the back-
bone amide
15
N nuclei of the TTF-1 HD at 286 K are
given in supplementary Table S1, Table S2 and
Fig. S1. Side-chain nitrogens were not considered for
analysis, except for the indole nitrogen of Trp48,
which represents a convenient probe with which to
monitor the dynamics of the HD hydrophobic core
(supplementary Table S1).
The longitudinal relaxation rates range between 1.15
and 1.97 s
)1
. The lowest R
1
values are observed for
Lys24 and Met37 and the residues of the flexible termi-
nal segments, with a characteristic pattern of decreas-
ing values on approaching these latter segments from
the respective adjacent helices. The highest R
1
values
are observed for Ser27, Arg31, Glu32, Ser36, Ile38,
Val45, and Trp48. The transverse relaxation rate val-
ues, higher than the corresponding R

N-terminal and C-terminal residues and by Leu34,
Gln44, Arg53, Arg58 and Gln59. A unique value of
26.39 s
)1
, by far the highest one, is observed for
Tyr25, which strongly suggests the presence of a local,
low-frequency conformational exchange contribution.
The steady-state {
1
H}–
15
N NOEs span the interval
)1.70 ⁄ +0.89. Negative values are observed for the ter-
minal fragments, i.e. Arg1–Leu7 and Lys61–Gln67,
reflecting the local dynamics characterized by fast
motions. In particular, the sign inversion transitions of
NOEs, seen on approaching the helical tracts from
flexible terminal residues, parallel the similar trends
observed for relaxation rates, and reflect consistently
the changes in local motional properties. In the recog-
nition helix, lower NOE values are obtained for the
C-terminal moiety, confirming that it is more flexible
than the N-terminal one. The highest {
1
H}–
15
N NOEs
were measured for Glu17 in helix I, Ser27 and Leu34
in helix II, and Lys46 and Gln50 in the N-terminal
portion of helix III. For an isotropically tumbling

quent MF and RSDM analysis calculations. However,
the qualitative implication of a high {
1
H}–
15
N NOE
for Glu17, Leu34 and Lys46, i.e. low specific mobility,
is consistent with the NOE trend of the corresponding
adjacent residues and hence does not conflict with the
global interpretation of the data. With the exclusion of
the N-terminal octapeptidyl and C-terminal nonapept-
idyl fragments of Glu17, Leu34 and Lys46, the average
of the {
1
H}–
15
N NOEs is 0.68 ± 0.10 (supplementary
Table S2). This value can be reliably considered to be
the average NOE over the structured core of the inves-
tigated TTF-1 HD molecule.
MF motional parameters
Figure 2 shows the individual overall rotational corre-
lation time, s
mi
, calculated from the individual residue
R
2
⁄ R
1
ratios, the generalized order parameters, S

for prolines. The s
e
values of Ser27 and Gln50 are not reported,
because they were not optimized by MF analysis. The extension of
TTF-1 HD helical segments is depicted above the graphs.
D. Gu
¨
mral et al. Backbone dynamics of the rat TTF-1 homeodomain
FEBS Journal 275 (2008) 435–448 ª 2007 The Authors Journal compilation ª 2007 FEBS 437
that expected for Tyr25 (14.67 ± 2.35 ns) were found
from MF formalism calculations.
Rotational correlation time
From the estimates of s
mi
based on the individual
relaxation rate ratios (Fig. 2), an average value of
9.7 ± 0.4 ns is extracted for the overall tumbling by
considering only the parameters from the best defined
(and conceivably most rigid) regions of the TTF-1 HD
(Gln10–Gln22, Ala28–Ile38, Thr43–Gln50) as deter-
mined from the NMR structure of the molecule [4].
When averaging is extended over the whole s
mi
dataset, only a slight difference is obtained, i.e.
Æs
mi
æ = 9.5 ± 0.9 ns. The excellent agreement between
the averages shows that the local segmental mobility
differences, albeit remarkable as inferred from the
NOE data, have little effect on the value of the Æs

motion model entailing a single-frequency local fluctu-
ation superimposed on the global motion. The quality
of the fitting was statistically validated by v
2
test
against the corresponding parameter distribution of
Monte Carlo simulations. The individual generalized
order parameters and internal effective correlation
times are plotted in Fig. 2. Their values reflect, respec-
tively, the specific amplitude and the frequency of the
local fluctuations for the motion of each considered in-
ternuclear
15
N–
1
H vector. The lowest S
2
values and,
correspondingly, the shortest s
e
values are obtained
at the N-terminal and C-terminal fragments 1–7 and
60–67 of the TTF-1 HD. This pattern suggests wide
motional freedom of the
15
N–
1
H vectors, which is in
line with the disordered NMR structure observed for
the same regions [4]. The N-terminal flexibility starts

exhibit an opposite correla-
tion from what is expected. This casts substantial
doubts on the reliability of the picture emerging from
the application of MF formalism to TTF-1 HD relaxa-
tion data. In detail, the highest S
2
values are obtained
for Gln50 and Tyr54, two residues that are essential
for the DNA recognition specificity of the TTF-1 HD
[24,25]. The restriction in local motion amplitude,
implied by the values of S
2
, seems consistent with the
role of Gln50 and Tyr54, but the corresponding s
e
val-
ues are not easily rationalized. For Gln50, the optimi-
zation procedure fails to fit the experimental data with
s
e
£ 11 000 ps. A low frequency of the internal
motions could be considered to match the above-men-
tioned correlation between high S
2
and large s
e
values.
In contrast, for Tyr54 a very low value of the opti-
mized s
e

large error in S
2
may reflect some problems with the
available data quality, a reduced motional rate for the
Ser27 backbone appears to be plausible, considering its
involvement in the defective capping of helix II [4]. At
this stage, the results are better described by consider-
ing the average values observed in the different second-
ary structure elements as reported in Table 1.
The local dynamics of the three helical regions of
the TTF-1 HD look very similar when only the aver-
age generalized order parameters are considered. A
clear difference emerges, however, if the internal corre-
lation times are taken into account. Only the motion
of helix I appears quite uniform, as inferred from the
similar values of the mean and weighed mean s
e
.
Helix II shows the largest variability in local fluctua-
tion frequency, despite the fact that the relative mean
generalized order parameter and the standard devia-
tion are very close to the corresponding counterparts
of helix I. This result can be rationalized on a struc-
tural basis. Helix II, in fact, should be the least stable
among the TTF-1 HD helices, because of its incom-
plete hydrogen bond network, which is due to defec-
tive N-capping and distortions introduced by Pro29.
At the same time, the side chains of residues 34, 35
and 38 are tightly anchored in the hydrophobic core of
the molecule, whereas the Glu30 side chain is involved

comply with the implicit conditions imposed by the
MF approach. In most cases, an increase ⁄ decrease in
the generalized order parameter corresponds to an
increased ⁄ decreased s
e
, which calls for a motional
regime that appears to be inconsistent within the MF
framework. However, all attempts to fit the experimen-
tal data with the extended MF approach [26], which
uses a double-timescale model for internal motions,
were also unsuccessful. It is tempting to speculate that
the physically puzzling picture emerging from the MF-
based fitting of the majority of the TTF-1 HD relaxa-
tion data could be attributed to correlated local
dynamics that occur on a timescale similar to that of
the overall tumbling.
Graphical analysis of spectral densities
Spectral densities at three frequencies [J(0), J(x
N
) and
J(0.87x
H
)] were calculated according to the matrix
equation given in supplementary Doc. S1. The individ-
ual spectral density values along the sequence of the
TTF-1 HD are displayed in the bar graphs of Fig. 3,
and the corresponding numerical values are given in
supplementary Table S4. Linear correlations between
J(0) and J(x
N

2
æÆs
e
æÆs
e
æ
w
a
Helix I (10–22) 0.87 (0.04) 1983 (406) 1964 (37)
Helix II (28–38) 0.86 (0.04) 2493 (1374) 1561 (57)
Helix III (43–59) 0.85 (0.08) 1486 (780) 1008 (19)
Helix III (43–52) 0.87 (0.06) 1885 (710) 1596 (51)
Helix III (53–59) 0.82 (0.09) 1030 (613) 905 (21)
Helix III (42–56) 0.87 (0.06) 1630 (809) 1297 (37)
Helix III (51–56) 0.86 (0.06) 1345 (740) 1142 (47)
Loop (23–27) 0.84 (0.08) 1038 (542) 468 (47)
Tight turn (39–42) 0.92 (0.04) 1960 (1047) 1403 (106)
N-terminus (1–9) 0.63 (0.19) 805 (752) 276 (3)
C-terminus (60–67) 0.58 (0.09) 514 (277) 206 (3)
a
Weighted average calculated using the individual s
e
uncertainties
(r
i
) as weighting factors (1 ⁄ r
i
2
).
D. Gu

slots are for residues 29 and 42 (prolines) and Glu17, Leu34 and
Lys46, which were excluded because of the extensive overlap
affecting the corresponding signals. Correlations were calculated by
means of
MATHEMATICA 5.2 software, using the relaxation dataset
given in supplementary Table S2. Relaxation data obtained from lin-
ear prediction were used for calculation only when the error intro-
duced by the procedure was acceptable, as discussed in
supplementary Doc. S1. The extension of TTF-1 HD helical seg-
ments is depicted above the graphs.
Fig. 4. J(x
N
)–J(0) correlation for the individual residues of the TTF-
1 HD from
15
N relaxation measurements. Different colors are used
to indicate the distinct groups of residues along the sequence, i.e.
N-terminal (orange), C-terminal (violet), helix I (yellow), helix II
(pink), helix III (green), loop (cyan), tight turn (brown), and residues
that undergo conformational exchange motions (blue). The fit (dark
solid line) was obtained by linear regression with the exclusion of
Arg1 and Gln67 (which exhibit strong negative NOE values) and
Lys21, Gln22, Tyr25, Ile38 and His52 [which make conformational
exchange contributions to J(0)]. The dashed curve (theoretical
curve) was calculated for J(0) and J(x
N
) as a function of s
,
using a
simple Lorentzian function. The left-hand inset shows an overview

N–
1
H vector is defined by a single Lorentz-
ian function with a fast s that is interpreted as general-
ized internal correlation time, s
gi
. For any point
between the upper and lower intercepts of the theoreti-
cal curve with the fitting line, the spectral density func-
tion can be expressed as a linear combination of the
two Lorentzian functions defined by s
m
and s
gi
, respec-
tively. The proximity to one of the intercepts between
the theoretical and fitting curves reflects the relative
contribution of each component Lorentzian function
to the specific spectral density of each experimental
point. Therefore, according to the RSDM analysis
[21], most of the
15
N–
1
H vectors of the TTF-1 HD
core move at the rate of the overall rotational correla-
tion frequency, and relaxation mainly occurs as a
result of overall rotational diffusion. Among all the
TTF-1 HD backbone
15

contributions in the ls-to-ms timescale, as mentioned
above. A similar situation is observed for the pairs
Glu30–Arg31 and Gln44–Val45, with the first residues
exibiting faster motions (ps-to-ns timescale), and the
latter residues slower motions on the nanosecond
timescale.
Detailed analysis of the spectral density functions
can be performed using the bar charts of Fig. 3 to
obtain the individual dynamic properties of each
15
N–
1
H vector. It can be seen that the
15
N–
1
H vectors
of the N-terminal and C-terminal residues undergo the
most rapid motions as compared to the rest of the
TTF-1 HD backbone. This is highlighted by low J(0)
and J(x
N
) values and correspondingly high J(0.87x
H
)
values, a pattern that is typically expected when the
considered internuclear vectors reorient on a fast (ps-
to-ns) timescale.
In the loop between helix I and helix II, Tyr25
shows a J(0) value that is much higher than that of

H
)
for Leu16] and are more dispersed along helix II. For
the latter, this indicates segmental mobility being
adopted with a less defined secondary sturucture, prob-
ably resulting from the lack of a complete hydrogen
bond network [4]. The dynamics of helix III can be
divided into two different regions, with a border occur-
ing at His52–Arg53 for all spectral density values. The
C-terminal segment of helix III (Arg53–Gln59) has
lower J(0) values than the adjacent N-terminal moiety
and the whole core of the TTF-1 HD, reflecting mobil-
ity related to poorly defined secondary structure [4].
Conversely, at the N-terminal segment of helix III,
higher J(0) values are inferred from analysis, consistent
with the better defined and more stable secondary
structure.
Overall, apart from the singularity at His52 that
results from an exchange contribution due to slow aro-
matic ring motion, as previously described, J(0) values
are seen to vary along helix III with some regularity
within the two identified moieties, i.e. a slight decrease
along segment 47–53, followed by a slight increase
in segment 53–55. The J(0) minimum is reached at
Arg53, where the low-frequency motion profile shows
similar characteristics as found at Arg58 for Gln59,
the frayed extremity of the recognition helix. The pat-
tern described for J(0) is observed also for J(x
N
) along

N
), an indication of slow local motion consistent
with the presence of a hydrogen bond network that
restricts the excursion of the Glu50 backbone. Other
relevant details of the spectral density analysis are seen
for Lys24 and Met37 amides, where increased values
of J(0) are coupled to low J(x
N
) values. Although
significantly low values of J(x
N
) are considered to be
evidence for fast motions, the corresponding J(0.87x
H
)
of the same residues rather suggests more complex
dynamics, i.e. other than the dual low-frequency and
high-frequency motional regime that appears to govern
local dynamics elsewhere, e.g. Leu26.
Table 2 lists the mean J(x) values together with
the corresponding standard deviations for the differ-
ent secondary structure elements of the TTF-1 HD.
It is readily seen that the
15
N–
1
H vectors of helix I,
helix II and the N-terminal segment of helix III do
not show major differences in the J(x) values. Con-
versely, the C-terminal fragment of helix III has

= 9.85 ± 1.79 ns and
s
gi
= 0.28 ± 0.11 ns. J(0.87x
H
)–J(0) correlation
yielded three roots, one for s
m
(9.84 ± 0.20 ns) and
two for s
gi
(0.26 ± 0.03 ns and 0.55 ± 0.06 ns) (sup-
plementary Doc. S1 and Fig. S2).
Comparison of results from MF and RSDM
The results for s
m
obtained by the MF and RSDM
approaches are in fairly good agreement, especially if
the comparison is drawn using the average value esti-
mated from R
2
⁄ R
1
ratios. Therefore, the assumption of
isotropic overall rotational diffusion for the TTF-1
HD proves to be convincingly appropriate.
The generalized order parameter values obtained
from the MF approach are consistent with the results
of RSDM. Lower generalized order parameters are
obtained for N-terminal and C-terminal residues, for

and the corresponding rela-
tively low s
e
values (sub-nanoseconds) for Arg53,
Arg58 and Gln59 suggest less restrictive and faster local
motions that are consistent with the reduced spectral
density results. In the ls-to-ms timescale, only Tyr25
Table 2. Mean spectral density values (ns) and corresponding
standard deviations (in parentheses) for the secondary structure
elements of the TTF-1 HD at 286 K.
Structural unit J(0) J(x
N
) J(0.87x
H
)
Helix I (10–22) 3.78 (0.30) 0.354 (0.008) 0.008 (0.001)
Helix II (28–38) 3.84 (0.37) 0.355 (0.029) 0.008 (0.001)
Helix III (43–59) 3.63 (0.34) 0.343 (0.020) 0.011 (0.005)
Helix III (43–52) 3.83 (0.26) 0.354 (0.017) 0.008 (0.002)
Helix III (53–59) 3.38 (0.25) 0.329 (0.014) 0.014 (0.006)
Loop (23–27) 3.79 (0.12)
a
0.340 (0.024) 0.010 (0.003)
Tight turn (39–42) 3.69 (0.11) 0.342 (0.010) 0.008 (0.001)
N-terminus (1–9) 2.83 (0.52) 0.271 (0.065) 0.027 (0.013)
C-terminus (60–67) 2.62 (0.30) 0.262 (0.052) 0.030 (0.006)
a
Tyr25 was excluded to avoid a significant bias on the average
from the slow exchange contribution (see text).
Backbone dynamics of the rat TTF-1 homeodomain D. Gu

was defined as the average of the position vector scalar
product
~
r
NÀH
ðtÞÁ
~
r
NÀH
ðt þ mDtÞ over the trajectory for
residue i. The root mean square of the quantity
[1 ) C(i,m)] was thus indicative of the deviation of the
vector N–H of residue i from the global behaviour.
This procedure is solely motivated by the inadequate
time sampling provided by a 10 ns MD simulation.
The largest deviations from global behavior are
observed at the N-terminus and C-terminus, with a
transition from disordered to more ordered vectors
between Phe8 and Ser9, and between Gln59 and
Arg58. Interestingly, this analysis highlights local
motions at Gln10–Val13, Gln22–Ser27 and Met37–
Leu40 and in the second part of helix III. As could be
expected, the analysis does not reproduce exactly the
experimental findings, but it is consistent with them
overall. In particular, the long loop involving Gln22–
Ser27 appears to be rather unconstrained, resulting in
large conformational motions in its central part. Simi-
larly, the second part of helix III appears to be less
restrained than the first part, starting from Tyr54.
Arg53 appears to be more mobile than the preceding

be masked by an overestimated s
m
. For the experimen-
tal data of the TTF-1 HD presented here, it was con-
cluded that only the two latter causes of deviation may
contribute to the erroneous estimates obtained from
MF analysis, although the possible uniform conforma-
tional exchange does not involve the whole molecule,
but rather specific regions. We could infer this conclu-
sion from the simultaneous analysis of the data
obtained using the RSDM approach. The fitting
obtained from the correlation plots among the differ-
ent spectral densities ensures that the assumption of
isotropic overall tumbling is correct within the experi-
mental error. This is consistent with previous evidence
obtained for the vnd ⁄ NK-2 HD [28], which is very clo-
sely related to the TTF-1 HD, as well as with explicit
anisotropy calculations that rule out anisotropic
motion (supplementary Doc. S1). The increase in the
refined overall correlation time with respect to the
average value obtained from relaxation rate ratios of
single residues, within the MF context, most likely
arose from inclusion in the dataset of the relaxation
rates with slow exchange contributions (namely those
from Lys21, Gln22, Ile38, and His52). The ensuing
overestimated s
m
, in turn, obscured the detection of
exchange contributions other than those of Tyr25
(which, in fact, was excluded from the dataset for

lar dimensions to account for the large s
m
value. In
addition to the details that are discussed in supplemen-
tary Doc. S1, one could mention that, as some 20 resi-
dues of the TTF-1 HD appear to be statistically
disordered, the increment of the average hydrodynamic
radius is well beyond 0.05 nm, which is expected to
increase by 10% the overall s
m
[30]. In fact, the
Stokes–Einstein relationship gives a hydrodynamic
radius of 1.98 nm for the TTF-1 HD under the condi-
tions of this study, i.e. very close to the mean radius of
the NMR structure of the molecule (1.94 nm) [4]. The
conclusions inferred here may be much more intrigu-
ingly challenged if one wonders whether the dynamic
properties of an isolated HD at 286 K can be extended
to the whole TTF-1 molecule under physiological con-
ditions. The temperature increase at 310 K and the
molecular size of the entire transcription factor should
lead to an overall tumbling rate of 20–22 ns
)1
. Besides
noting that the selected experimental conditions for
characterizing the dynamics of the TTF-1 HD are not
completely unrelated to the dynamic regime within the
whole protein, it is clear that the local mobility trends
that may influence HD function should still apply, and
may possibly be elicited, under physiological condi-

for the
Arg31
15
N–
1
H vector. Thus, the result for individual
s
e
> s
mi
obtained for Ser27, which is involved in
N-capping with Arg31 N–H and Glu30 N–H is, at least
qualitatively, justified, and suggests an interpretation
based on the compensation between the amplitude and
frequency of local fluctuations. In other words, a wider
motion amplitude is accompanied by a slower motion
rate because of the increased mechanical inertia.
In the context of RSDM, the detailed analysis of the
three spectral densities J(0), J(x
N
) and J(0.87x
H
)
allowed us to obtain a rather complete description of
the dynamics of the TTF-1 HD over a large range of
timescales. The current observations are in agreement
with our previously published structural characteriza-
tion of the TTF-1 HD [4]. As we concluded before, the
C-terminal segment of helix III, which is involved in
the DNA recognition process, displays higher mobility

¨
mral et al.
444 FEBS Journal 275 (2008) 435–448 ª 2007 The Authors Journal compilation ª 2007 FEBS
an aromatic side chain from the same or a nearby resi-
due. Slow motion of aromatic side chains creates local
field gradients at the neighboring residues, which may
provide very efficient relaxation pathways, because of
the well-known effects of ring currents on chemical
shifts. These contributions could be recognized as the
collective slow motions that appear to occur along the
helical backbone, as inferred from MF analysis failure.
The process seems particularly effective at the C-end
of the helices, and could be regarded as helix–coil tran-
sition on a slow timescale [29]. For the recognition
helix, this behavior appears to correspond with the
kink at Asn51–His52–Arg53; residues 51 and 53 are
nearly invariant in all eukaryotic HDs, i.e. are an early
feature in HD evolution, and thus could represent a
conserved determinant for the local dynamics [11]. The
resulting abrupt change of the recognition helix regis-
ter introduced by the 51–53 kink, as confirmed by
NMR evidence [4,11], should affect the amide bond
vector dynamics of Tyr54, an important recognition
determinant for the NK-2 HD subfamily [31], within
the flexible joint between the N-terminal and C-termi-
nal moieties of the recognition helix. The implication
for DNA binding that may be envisaged from the
available conformational options within the recogni-
tion helix [2–4,7–9] is that the latter helix, firmly ori-
ented within the helix–turn–helix motif, may undergo a

under physiological conditions, when the HD is part
of a much larger transcription factor, and determine
the extent of the conformational changes and, hence,
the energetics of the interaction with DNA
[2,3,8,11,32].
Experimental procedures
Sample preparation
Uniformly
15
N-labeled (U-
15
N) TTF-1 HD (68 residues
including the segment 160–226 of the whole rat thyroid
transcription factor, plus an extra methionyl residue at the
N-terminus, numbered Met0) was obtained from overex-
pression in Escherichia coli strain BL21, by growth in a
minimal medium containing
15
NH
4
Cl as a source of nitro-
gen. Expression and subsequent purification were per-
formed as described previously [5,33]. NMR samples were
prepared by dissolving the lyophilized powder in H
2
O ⁄ D
2
O
(95 : 5, v ⁄ v) and adjusting the pH (uncorrected pHmeter
reading) to 4.3 by microadditions of 1 m HCl. The labeled

N relaxation data analysis
The longitudinal and transverse rate constants were calcu-
lated from peak heights of the
1
H–
15
N correlation data ser-
ies. Under the typical conditions employed for protein
NMR relaxation studies, peak heights have been proven to
be more accurate than the corresponding volumes [35]. To
determine the R
1
and R
2
values, a three-parameter and
two-parameter, respectively, nonlinear least-square fit of the
equations
I sðÞ¼I
1
À I
1
À I
0
ðÞexp ÀR
1
sðÞ ð1Þ
and
I sðÞ¼I
0
exp ÀR

¼ NOE À 1ðÞ
c
N
c
H
R
1
ð3Þ
The details concerning the relaxation data analysis per-
formed with the MF approach [14,15] and the RSDM
approach [18–22] are given in supplementary Doc. S1.
Error estimations
In order to establish the errors on individual peak height
values, the reproducibility of the experimental R
1
and R
2
data was assessed by measurement duplication over a series
of arbitrarily selected relaxation delays (at least three;
see supplementary Doc. S1). The average uncertainties
obtained for R
1
constants were 1.4% for resolved reso-
nances and 1.0% for partially overlapping ones, whereas
the corresponding quantities for R
2
were 14% and 17%.
This difference reflects the inherent accuracy limit diversity
of R
1

This work was financially supported by AIRC, MIUR
(2006058958, RBNE03PX83) and EU (LSHM-CT-
2005-037525). The suggestions of Dr A. Makek are
acknowledged.
References
1 Gehring WJ (1987) Homeo boxes in the study of devel-
opment. Science 236, 1245–1252.
2 Pabo CO & Sauer RT (1992) Transcription factors:
structural families and principles of DNA recognition.
Annu Rev Biochem 61, 1053–1095.
3 Gehring WJ, Qian YQ, Billeter M, Furukubo-Tokuna-
ga K, Schier AF, Resendez-Perez D, Affolter M, Otting
G&Wu
¨
thrich K (1994) Homeodomain–DNA recogni-
tion. Cell 78, 211–223.
4 Esposito G, Fogolari F, Damante G, Formisano S, Tell
G, Leonardi A, Di Lauro R & Viglino P (1996) Analy-
sis of the solution structure of the homeodomain of rat
thyroid transcription factor 1 by
1
H-NMR spectroscopy
and restrained molecular mechanics. Eur J Biochem
241, 101–113.
5 Guazzi S, Price M, De Felice M, Damante G, Mattei
MG & Di Lauro R (1990) Thyroid nuclear factor 1
(TTF-1) contains a homeodomain and displays a novel
DNA binding specificity. EMBO J 9, 3631–3639.
6 Bohinski RJ, Di Lauro R & Whitsett JA (1994) The
lung-specific surfactant protein B gene promoter is a

1
H nuclear mag-
netic resonance spectroscopy. J Mol Biol 214 , 183–197.
10 Gu
¨
ntert P, Qian YQ, Otting G, Mu
¨
ller M, Gehring W
&Wu
¨
thrich K (1991) Structure determination of the
Antp (C39-S) homeodomain from nuclear magnetic
resonance data in solution using a novel strategy for the
structure calculation with the programs DIANA, CALI-
BA, HABAS and GLOMSA. J Mol Biol 217, 531–540.
11 Esposito G, Fogolari F, Damante G, Formisano S, Tell
G, Leonardi A, Di Lauro R & Viglino P (1997) Hydro-
gen–deuterium exchange studies of rat thyroid tran-
scription factor 1 homeodomain. J Biomol NMR 9,
397–407.
12 Nirmala NR & Wagner G (1988) Measurement of
13
C
relaxation times in proteins by two-dimensional hetero-
Backbone dynamics of the rat TTF-1 homeodomain D. Gu
¨
mral et al.
446 FEBS Journal 275 (2008) 435–448 ª 2007 The Authors Journal compilation ª 2007 FEBS
nuclear
1

19 Ishima R & Nagayama K (1995) Quasi-spectral density
function analysis for nitrogen-15 nuclei in proteins.
J Magn Reson B 108 , 73–76.
20 Ishima R & Nagayama K (1995) Protein backbone
dynamics revealed by quasi spectral density function
analysis of amide N-15 nuclei. Biochemistry 34, 3162–
3171.
21 Lefe
`
vre JF, Dayie KT, Peng JW & Wagner G (1996)
Internal mobility in the partially folded DNA binding
and dimerization domains of GAL4: NMR analysis of
the N–H spectral density functions. Biochemistry 35,
2674–2686.
22 Dayie KT, Wagner G & Lefe
`
vre JF (1996) Theory and
practice of nuclear spin relaxation in proteins. Annu
Rev Phys Chem 47, 243–282.
23 Mandel AM, Akke M & Palmer AG (1995) Backbone
dynamics of Escherichia coli ribonuclease HI: correla-
tions with structure and function in an active enzyme.
J Mol Biol 246, 144–163.
24 Fogolari F, Esposito G, Viglino P, Damante G &
Pastore A (1993) Homology model building of the
thyroid transcription factor 1 homeodomain. Protein
Eng 6, 513–519.
25 Damante G, Pellizzari L, Esposito G, Fogolari F,
Viglino P, Fabbro D, Tell G, Formisano S & Di Lauro
R (1996) A molecular code dictates sequence-specific

31 Kim Y & Nirenberg M (1989) Drosophila NK-homeo-
box genes. Proc Natl Acad Sci USA 86, 7717–7720.
32 Dragan AI, Li Z, Makeyeva EN, Milgotino EI, Liu Y,
Crane-Robinson C & Privalov PL (2006) Forces driving
the binding of homeodomains to DNA. Biochemistry
45, 141–151.
33 Viglino P, Fogolari F, Formisano S, Bortolotti N,
Damante G, Di Lauro R & Esposito G (1993)
Structural study of rat thyroid transcription factor 1
homeodomain (TTF-1 HD) by nuclear magnetic
resonance. FEBS Lett 336, 397–402.
34 Stone JM, Fairbrother WJ, Palmer AG III, Reizer J,
Saier MH Jr & Wright PE (1992) Backbone dynamics
of the Bacillus subtilis glucose permease IIA domain
determined from
15
N NMR relaxation measurements.
Biochemistry 31, 4394–4406.
35 Palmer AG, Rance M & Wright PE (1991) Intramolecu-
lar motions of a zinc finger DNA-binding domain from
Xfin characterized by proton-detected natural abun-
dance
13
C heteronuclear NMR spectroscopy. JAm
Chem Soc 113, 4371–4380.
36 MacKerell AD Jr, Bashford D, Bellott M, Dunbrack
RL Jr, Evanseck JD, Field MJ, Fischer S, Gao J, Guo
H, Ha S et al. (1998) All-atom empirical potential for
molecular modeling and dynamics studies of proteins.
J Phys Chem B 102 , 3586–3616.

ex
values.
Table S4. Spectral density function values.
Fig. S1. Bar graph of R
1
, R
2
and {
1
H}–
15
N NOE val-
ues along the sequence of the TTF1 HD at 11.7 T and
286 K.
Fig. S2. Plot of J(0.87x
N
)–J(0) correlation from
15
N
relaxation measurements of the TTF-1 HD.
This material is available as part of the online article
from
Please note: Blackwell Publishing are not responsible
for the content or functionality of any supplementary
materials supplied by the authors. Any queries (other
than missing material) should be directed to the corre-
sponding author for the article.
Backbone dynamics of the rat TTF-1 homeodomain D. Gu
¨
mral et al.