Structure, Vol. 10, 1349–1361, October, 2002, 2002 Elsevier Science Ltd. All rights reserved.
PII S0969-2126(02)00857-2
Structure and Interactions of PAS Kinase
N-Terminal PAS Domain: Model
for Intramolecular Kinase Regulation
Carlos A. Amezcua, Shannon M. Harper,
Jared Rutter, and Kevin H. Gardner1
Department of Biochemistry
The University of Texas Southwestern
Medical Center
5323 Harry Hines Boulevard
Dallas, Texas 75390
Summary
PAS domains are sensory modules in signal-transducing proteins that control responses to various environmental stimuli. To examine how those domains can
regulate a eukaryotic kinase, we have studied the
structure and binding interactions of the N-terminal
PAS domain of human PAS kinase using solution NMR
methods. While this domain adopts a characteristic
PAS fold, two regions are unusually flexible in solution.
One of these serves as a portal that allows small organic compounds to enter into the core of the domain,
while the other binds and inhibits the kinase domain
within the same protein. Structural and functional
analyses of point mutants demonstrate that the compound and ligand binding regions are linked, suggesting that the PAS domain serves as a ligand-regulated switch for this eukaryotic signaling system.
Introduction
An essential element of biological adaptation to changing environmental conditions is the use of sensory proteins that transduce various stimuli into changes in the
activity of downstream signaling pathways. One particularly widespread solution to these sensory requirements
is provided by PAS (Per-Arnt-Sim) domains, identified
to date in over 1100 proteins from organisms in all three
kingdoms of life [1, 2]. PAS domains, as with many other
classes of signal transduction modules, are protein/protein interaction domains that are used for intra- and
intermolecular associations of macromolecules. To regulate such interactions in response to environmental
factors, many PAS domains have integrated organic cofactors within their hydrophobic cores as sensors.
Changes in the conformation or occupancy of these
cofactors signal changes in stimuli that cannot be adequately sensed by moieties composed solely of the
twenty natural amino acids. For example, PAS domains
have been demonstrated to use heme to sense O2 in
several bacterial enzymes [3], FMN to sense blue light
in plant phototropins [4], and 4-hydroxycinnamic acid
to sense blue light in photoactive yellow protein [5]. It
is intriguing to note that, from a structural standpoint,
these diverse ligands are tolerated within the cores of
different PAS domains, all of which retain the same
1
Correspondence: [email protected]
mixed ␣/␤ fold diagrammed in Figure 1A. Folding into
this structure is not dependent on ligand binding in all
cases, though, as demonstrated by the well packed,
ligand-free hydrophobic core observed in the crystal
structure of the human ether-a-go-go-related gene
(HERG) PAS domain [6].
PAS domains are found in several classes of eukaryotic signaling proteins, including transmembrane channels, transcriptional activators, and kinases [1]. To probe
the regulatory mechanisms of the last of these, we have
studied PAS kinase (PASK, otherwise referred to as
PASKIN), a PAS-containing serine/threonine kinase [7,
8]. Recent studies of the two yeast PASK genes have
shown that PASK controls two important cellular decisions [9]. PASK blocks the production of glycogen by
phosphorylating and inhibiting glycogen synthase. In
addition, genetic and biochemical experiments have implicated PASK in the control of protein synthesis. These
two cellular decisions, to store or utilize sugar and the
rate of protein synthesis, are both tightly coupled to
cellular energy status. The N-terminal PAS domain of
the human PASK (hPASK PAS A) directly inhibits the
kinase activity when added in trans to the kinase domain
[7], suggesting that it plays a role in this regulation.
The observation of ligand regulation of PAS domains
in other systems has suggested that hPASK PAS A might
directly serve as a regulatory sensor for this enzyme.
Here we combine biophysical and biochemical data to
establish several critical aspects of this model. We demonstrate that, although hPASK PAS A adopts a solution
structure similar to that of other members of this family,
it contains two unusually flexible segments. The locations of those segments suggest that they might serve as
functionally relevant binding interfaces. An NMR-based
screen of a library of over 750 organic compounds
shows that hPASK PAS A can specifically bind small
molecules within its hydrophobic core. Further, we show
that the hPASK kinase domain binds to a flexible loop
on the PAS domain surface, providing the first direct
biophysical observation of a protein interaction surface
on a PAS domain. We conclude that the ligand and
kinase binding segments are functionally and structurally linked through the analysis of several point mutants,
suggesting a regulatory pathway for small organic compounds to regulate kinase activity in this system.
Results
Solution Structure of hPASK PAS A
Sequence alignments of PAS domains exhibit relatively
poor sequence identity among members of this protein
family, leading to differing definitions of the length of
these domains, ranging from approximately 60 to 250
amino acids [1, 10]. Practically, this low level of conservation (Figure 1B) hampers the design of constructs that
produce well-folded domains for structural studies. To
Key words: PAS domain; PAS kinase; ligand binding; kinase regulation; NMR screening
Structure
1350
Figure 1. Structural Features of PAS Domains
(A) The typical PAS domain fold consists of a five-stranded antiparallel ␤ sheet flanked by several ␣ helices, labeled here with the nomenclature
initially proposed for FixL [13].
(B) Sequence alignment of hPASK PAS A with known PAS domain structures, with helical and sheet residues in red and blue, respectively.
The alignment was initially generated with the program MultAlin [45] and manually adjusted to match elements of secondary structure between
proteins. Aligned proteins and their RCSB accession codes (in parentheses) are as follows: hPASK PAS A (1LL8), PYP (2PHY), BjFixL (1DRM),
RmFixL (1D06), HERG (1BYW), and PHY3 (1G28).
address this, we used an NMR-based method to screen
directly for the folding of various fragments of the N-terminal PAS domain of hPASK while unpurified in E. coli
cell lysates [11, 12]. This approach allowed us to identify
a minimal fragment for this domain containing residues
131–237, designated hPASK PAS A, which was used for
all further studies.
Using standard double- and triple-resonance NMR
experiments conducted on uniformly 15N- and 15N, 13Clabeled protein samples, we determined the structure
of hPASK PAS A to high resolution (Figure 2). This structure is based on over 2500 geometric constraints, including measurements of interproton distances, dihedral
angles, and the alignment of hPASK PAS A within an
anisotropic medium (Table 1). All of these constraints
were well satisfied by the twenty lowest-energy structures that represent the ensemble that was used for
further analysis. Superimposition of these structures
shows that hPASK PAS A adopts the typical ␣/␤ PAS
domain fold, characterized by several ␣ helices flanking
a five-stranded antiparallel ␤ sheet scaffold. This structure is quite similar to that of other PAS domains, as
demonstrated by the low (1.2–1.4 Å) backbone rmsd
values produced by pairwise comparisons of all known
PAS structures with respect to hPASK PAS A (vide infra).
The majority of the hPASK PAS A structure was solved
with high precision, but portions of the F␣ helix and the
subsequent FG loop were poorly defined in the ensemble of our structures. This region is well ordered in the
other known PAS domain structures solved either by
NMR or X-ray methods and has been shown, in several
cases, to provide the side chains that form specific interactions in the hydrophobic core or with bound cofactors.
The exceptions to this are the crystal structures of the
PAS Domain Structure and Function in PAS Kinase
1351
Figure 2. Solution Structure of hPASK PAS A
(Left) Bundle of 20 lowest-energy structures
for hPASK PAS A, calculated as indicated in
the text.
(Right) Schematic ribbon diagram of the
structure closest to the mean.
Bradyrhizobium japonicum and Rhizobium meliloti FixL
PAS domains, both of which show elevated temperature
factors in the FG loop region consistent with increased
flexibility [13, 14]. Notably, this same region, called the
“regulatory loop” for FixL, is thought to be critical in the
PAS-mediated O2 signaling by this protein. As hPASK
PAS A can regulate the activity of its kinase domain [7],
we speculate that the FG loop region might be inherently
flexible for signal transduction purposes by facilitating
interactions with the kinase domain, a small organic
cofactor, or both.
Inherent Flexibility of F␣ Helix/FG Loop Region
To characterize more directly the dynamics of the F␣
helix/FG loop region in solution, we measured several
sets of 15N relaxation parameters (15N R1, R2 rates and
15
N{1H} steady-state NOE values) for the backbone amide
groups in hPASK PAS A. By examining how these fluctuate throughout the protein, we see that the F␣ helix, FG
loop, and H␤ strand are flexible, as manifested by high
R2 and low 15N{1H} NOE values for multiple sites in these
regions (Figure 3A). To deconvolute the dependence of
these parameters on motions with different timescales,
we analyzed those data using a reduced spectral density
function approach [15, 16]. With this technique, sites
that are flexible on a fast timescale (picosecond-nanosecond) typically exhibit elevated values of the spectral
density function J(␻) at high frequencies [J(0.87␻H)] and
correspondingly decreased values at lower frequencies
[Jeff(0)]. This is observed both in the FG loop (residues
185–194) and in the termini at either end of the PAS
domain (Figure 3B), indicating that these regions are
flexible on this fast timescale. In contrast, sites at both
ends of the FG loop show average J(0.87␻H) and elevated Jeff(0) values through residues 175–184 and 195–
200. Such behavior is commonly observed with regions
undergoing conformational changes on a slower (␮sms) timescale. This conclusion is further supported by
the strong dependence of 15N R2 values for these residues on the external magnetic field strength (data not
shown) [15]. The observation of such extensive flexibility
is unique among PAS domains studied by NMR to date,
not having been observed in the dark state of the ligandbound form of PYP [17], an unliganded form of the
Table 1. Statistics for hPASK PAS A Solution Structure Determination
List of Constraints
Number of Constraints
NOE distance restraints:
Unambiguous
Ambiguous
Hydrogen bond restraints
Dihedral angle (φ, ␺) restraints from TALOS
15
N-1H residual dipolar couplings:
2520
1683
837
48
116
61
Structural Analysis
Average number of NOE violations greater than 0.5 Å
Average number of dihedral violations greater than 5⬚
Residual dipolar coupling analysis
Geometric analysis of ordered residuesa
Rmsd to mean
Procheck
a
0
0
Q ⫽ 0.21, r ⫽ 0.977
0.44 ⫾ 0.08 Å (backbone)
0.95 ⫾ 0.12 Å (all heavy)
76.4% most-favored region
21.8% additionally allowed region
1.0% generously allowed region
0.9% disfavored
Residues used for geometric analysis are as follows: 133–169, 177–180, 200–217, and 227–235.
Structure
1352
Figure 3. Backbone Dynamics of hPASK
PAS A
Backbone amide 15N relaxation data is plotted
as a function of residue number. Secondary
structure elements identified in the solution
structure are shown schematically, with the
flexible F␣/FG region indicated by dashed
lines.
(A) 15N R1, R2, and 15N{1H} NOE values, recorded at 14.1 T (1H frequency, 600 MHz).
(B) Reduced spectral density function mapping analysis of data presented in (A).
N-terminal PAS domain of murine NPAS2 [18], or several
other PAS domains under study in our group (data not
shown). In sum, these data confirm that several regions
of hPASK PAS A are unusually flexible, as might be
observed for binding regions in the absence of their
small molecule or protein partners.
Identification of a Binding Site for Small Organic
Ligands within hPASK PAS A
Given the key role that organic cofactors and ligands
play in PAS domain signaling processes and the observation of unusual flexibility of hPASK PAS A near the
ligand binding sites of the FixL and Phy3 domains [13,
14, 19], we hypothesized that hPASK PAS A might bind
small organic compounds. As a way to test this model,
we used NMR-based methods to screen the ability of
hPASK PAS A to bind members of a library of over 750
organic compounds. This library is primarily composed
of small (average FW of 200 Da) commercially available
chemicals that contain structural elements commonly
found among many protein binding compounds, including pharmaceutical agents [20, 21]. This library was
screened by using 15N/1H HSQC spectra to monitor
changes in backbone amide chemical shifts in hPASK
PAS A upon the addition of compounds (Figure 4A),
allowing the rapid determination of both the affinity and
locations of ligand binding sites for this domain [22].
From this screen, we identified 9 compounds that
bound hPASK PAS A with equilibrium dissociation constants smaller than 100 ␮M and an additional 12 with
PAS Domain Structure and Function in PAS Kinase
1353
Figure 4. Identification of a Ligand Binding Pocket within the Core of hPASK PAS A
(A) Superposition of 15N/1H HSQC spectra of 800 ␮M 15N-labeled hPASK PAS A without (black) and with 900 ␮M compound 1 present (red).
(B) Structure-activity relationship information for compound 1, showing that significant changes in affinity result from minor structural modifications. Kd values for compounds 1 and 2 are determined from isothermal titration calorimetry as these compounds exchange between the free
and protein-bound states on an intermediate NMR timescale. In contrast, compounds 3–6 exhibit fast exchange on the NMR timescale in
their binding to hPASK PAS A, allowing a direct measurement of Kd from the concentration dependence of chemical shifts in a titration.
(C) Comparison of binding sites for aromatic compounds within hPASK PAS A (left) and heme within FixL (right) [14, 27]. Amide and methyl
groups significantly affected by the addition of saturating quantities of compound 1 are mapped onto hPASK PAS A, by the minimum chemical
shift criteria shown in the figure. No information is available for the amide groups of residues highlighted in gray because of solvent exchange
broadening.
Kd values between 100 and 500 ␮M. All of those compounds are hydrophobic in character, typically containing one or two aromatic rings substituted with polar
groups around their peripheries and separated by a
linker of variable length (0–2 atoms). Examination of the
structure-activity relationships of compounds with similar architectures reveals significant specificity in the protein-ligand interactions. This is demonstrated in Figure
4B, where minor structural changes to one of the tightest
binding compounds (1), such as replacing the two trifluoromethyl groups with methyl groups (4) or removing
them altogether (5), weaken the binding interaction by
two orders of magnitude.
All of the compounds identified in our screen interact
with hPASK PAS A in a similar manner, as revealed by
mapping the residues affected by ligand binding onto
Structure
1354
the solution structure (Figure 4C). This site is formed
between the E␣ and F␣ helices and the G␤, H␤, and I␤
strands of the main ␤ sheet, in a location strikingly similar
to the pockets used to bind heme and FMN in the FixL
and Phy3 PAS domains, respectively [13, 14, 19]. As
expected from the nature of the compounds that bind
therein, the interior of this site is rich in aliphatic and
aromatic residues and lined by a number of polar residues on the F␣ helix and G␤ strand. We note that this
binding site is not preformed in the absence of ligands,
as no sufficiently sized cavity is present in our solution
structure. The deformability of this site is further underscored by our observation that the number of residues
affected by ligand binding is proportional to the strength
of the interaction, indicating that tighter binding ligands
sequester themselves more deeply within the PAS domain (data not shown).
Identification of the hPASK PAS A
Kinase Binding Surface
Previous in vitro biochemical studies have demonstrated that the hPASK PAS A domain can specifically
inhibit its own kinase domain in trans, implying a direct
interaction between the two [7]. To map the regions of
hPASK PAS A involved in this interaction, we used 15N/
1
H HSQC spectra to monitor the effects of titrating unlabeled kinase into a sample of 15N-labeled PAS domain.
In contrast to the typical fast exchange behavior exhibited in similar titrations with organic compounds, peaks
in these spectra showed minimal chemical shift changes
and significant reductions in peak intensity upon the
addition of kinase (Figure 5A). Analysis of these changes
revealed two important trends. First, all peaks experienced a general intensity reduction because of the
slowed rotational tumbling of the larger 52.7 kDa PAS/
kinase complex compared with the 12.7 kDa isolated
PAS domain. Peak intensities decreased as increasing
amounts of kinase were added and saturated at an equimolar ratio of the two proteins, confirming that this
slowed tumbling is not caused by nonspecific increases
in sample viscosity. Second, peaks from a subset of
residues were preferentially attenuated upon the addition of even substoichiometric amounts of kinase (Figure 5A, labeled peaks). Such differential effects have
been observed in several other protein/protein complexes, allowing the identification of interfaces used by
15
N-labeled proteins to bind to complexes as large as
30S ribosomal subunits [23, 24]. Mapping these sites
onto the solution structure of hPASK PAS A shows that
the majority of these residues are in the FG loop, with
additional sites in the F␣ helix and HI loop (Figure 5B).
Interestingly, this region is adjacent to the ligand binding
cavity found by the library screen, suggesting that
changes in the occupancy or conformation of a bound
ligand could be used to regulate protein/protein interactions in this system.
Linkage between the Ligand and Kinase Binding
Sites of hPASK PAS A
To assess the elements necessary for kinase inhibition
by hPASK PAS A, we characterized the structural and
functional states of several proteins containing point
mutations at sites with high sequence or structural con-
servation among structurally characterized PAS domains. The folding of those proteins was rapidly assessed from 15N/1H HSQC spectra recorded on lysates
of 15N-labeled E. coli cultures overexpressing protein
[11, 12]. Functional status was assayed by the ability of
each protein to inhibit the isolated hPASK kinase domain
(hPASK⌬N) in trans [7].
On the basis of the results of these structural and
functional assays, these variant proteins (N148F, G156F,
F171A, G198I, I203F, G208V, and C228F) can be grouped
into three classes (Figure 6A). While members of two
classes had either no functional alteration or a complete
loss of structure, those in a third class all maintained
essentially wild-type structure but were unable to inhibit
kinase activity. All of the altered residues in this class
provide side chains that pack in the hydrophobic core
of the protein on the periphery of the ligand binding site
identified above. Detailed functional characterization of
two of these mutants, I203F and C228F, showed that
both were significantly impaired in their ability to inhibit
kinase activity in trans, with IC50 values at least four
times higher than the 50–100 ␮M value of the wild-type
PAS domain (Figure 6B). In addition, versions of the fulllength 142 kDa hPASK enzyme containing either the
I203F or C228F alterations were constitutively upregulated, indicative of a disrupted in cis regulatory mechanism.
To investigate the structural basis of the failure of
the I203F and C228F mutant hPASK PAS A domains to
inhibit kinase activity, we assigned the backbone chemical shifts of both variants and examined differences
in these as probes for structural changes. Changes in
secondary structure were determined by comparing the
13
C␣ and 13C␤ chemical shifts of the wild-type and both
mutant proteins to a database of random coil values
[25], showing that all three contain virtually the same
secondary structure elements and likely have similar
overall structures (Figure 6C). An analogous treatment
of the 13C shifts of wild-type protein saturated with compound 1 shows similar results (data not shown). Interestingly, I203F exhibited significant perturbations in the F␣
helix/FG loop region, suggesting that this region has
been drastically altered, to the point where it is no longer
able to inhibit the kinase. Tertiary structure changes
were probed by comparing backbone amide 15N and
1
H shifts for each of the mutants with wild-type values
(Figure 6D). For each of the mutants, chemical shift
changes are seen at the site of mutation and neighboring
residues, as expected. However, both mutations also
lead to significant changes in the F␣ helix, FG loop, and
H␤ strand, despite the fact that these are remote to the
altered sites. To compare these perturbations with those
introduced by ligand binding, we also examined the
changes in the 15N/1H shifts of wild-type protein caused
by saturation with compound 1 (Figure 4A). This analysis
shows a widespread distribution of chemical shift
changes within the protein core consistent with binding
within the domain. Notably, we also observed significant
line broadening at sites located primarily in the F␣ helix,
FG loop, and H␤ strand (Figure 6D, red bars). Such
observations are consistent with these regions rapidly
interconverting among multiple conformations. We conclude from these results that structural changes within
the hydrophobic core of the PAS domain—introduced
PAS Domain Structure and Function in PAS Kinase
1355
Figure 5. Identification of the Kinase Binding
Site on the Surface of hPASK PAS A
(A) Superimposed 15N/1H HSQC spectra of
200 ␮M 15N-labeled hPASK PAS A without
(black) or with 150 ␮M unlabeled kinase domain (red). Peaks originating from residues
at the PAS/kinase interface show selective
intensity reduction upon the addition of kinase because of chemical exchange broadening effects.
(B) Location of the kinase binding loop as
identified by residues (red sphere) with selectively broadened amide resonances. No information was obtained from residues colored
in gray, signals from which are broadened
in the absence of kinase, because of rapid
solvent exchange at the high pH (pH 7.5)
needed to stabilize the kinase domain.
by mutation or ligand binding—are converted into
changes in the structure and/or dynamics of the F␣ helix,
H␤ strand, and adjacent kinase binding FG loop that
disrupt inhibitory PAS/kinase interactions.
From CD spectroscopy, we found that the thermal
stabilities of the I203F and C228F mutants were perturbed. The apo wild-type hPASK PAS A domain is inherently very stable, with a melting temperature of approximately 84⬚C (Figure 6E). The I203F mutation is
significantly destabilizing, dropping the Tm by 9⬚C, to
75⬚C. Such destabilization is commonly found among
point mutations of most proteins and is also consistent
with the distortion observed in the F␣ helix in I203F
(Figure 6C). In contrast, the C228F mutant is stabilized
to the point where its melting transition begins near the
boiling point of water (Tm ⬎ 95⬚C). This data suggests
that the phenylalanine side chain introduced by this mutation may serve as a surrogate ligand for this domain,
stabilizing the protein as do many other ligands upon
binding to their protein receptors.
Discussion
Ligand Binding in the Hydrophobic
Core of hPASK PAS A
The solution NMR structure of hPASK PAS A shows that
this domain has a typical PAS fold that is flexible enough
to accommodate small organic ligands within its hydrophobic core. Compounds appear to enter the core
of the molecule through a dynamic portal composed of
Structure
1356
Figure 6. Structural and Functional Effects of Point Mutations within hPASK PAS A
(A) Map of point mutants within the hPASK PAS A domain, categorizing the effects of these on overall folding and kinase inhibitory activity.
(B) Effects of the I203F and C228F mutations on the ability of hPASK PAS A to inhibit kinase activity, as assessed both in trans (left, with 400
␮M of each PAS domain) and in cis (right).
(C) Secondary structure analysis based on 13C␣/13C␤ chemical shifts for two point mutants (I203F and C228F) and wild-type hPASK PAS A.
The plotted function is the difference of the secondary shifts of both nuclei [⌬␦(13C␣) ⫺ ⌬␦(13C␤)] and has been treated with a three-residue
smoothing window. Red dashed lines at ⫾1.4 ppm inddicate the amplitudes of significant chemical shift deviations from random cell values
[25]. The shaded region indicates the residues most affected by these mutations, which also corresponds to the flexible region identified in
Figure 3.
(D) Residues perturbed by mutation (I203F and C228F) or ligand binding (compound 1), as identified by changes in backbone amide 15N/1H
shifts from unliganded, wild-type hPASK PAS A. Changes were quantitated by [⌬␦(1H)2 ⫹ 0.1 ⫻ ⌬␦(15N)2 ]1/2. Residues with significant line
broadening upon the addition of compound 1 are identified with pink bars.
(E) Altered stability of I203F and C228F point mutants, as assayed by thermal melts monitored by circular dichroism at 222 nm. Melting curves
were produced by a five-point running average of experimentally observed ellipticity data.
residues from the flexible F␣ helix and G␤ strand. This
region includes residues that appear essential for discrimination against differently sized ligands by various
PAS domains [26]. The cavity that binds those compounds is not formed a priori, avoiding the energetic
costs of preforming a hydrophobic pocket within a rela-
tively small domain. While such flexibility might suggest
a lack of specificity in hPASK PAS A/ligand interactions,
this domain discriminates between different ligands of
a very similar chemical architecture (Figure 4B). Given
those features of the hPASK PAS A structure and the
roles that ligands are known to play in PAS-mediated
PAS Domain Structure and Function in PAS Kinase
1357
Figure 7. Model for Kinase Regulation by
Small-Molecule PAS Ligands
In the nonliganded state, the hPASK PAS A
is capable of inhibiting the activity of its own
kinase domain through interactions mediated
by the FG loop. The introduction of aromatic
groups (purple) within the hydrophobic core
of the domain, provided either by the binding
of small molecules or via site-directed mutagenesis, leads to conformational changes
that propagate to the F␣ helix, FG loop, and
H␤ strand and relieves the kinase inhibition.
signal transduction in other proteins, we propose that
hPASK PAS A ligand binding is linked to kinase regulation.
An essential element of this model is the identity of
the physiological ligand of hPASK PAS A, which remains
unknown at present. This ligand is unlikely to be one
previously identified for other PAS domains, such as
heme, ATP, or FMN, since none of those compounds
bound to hPASK PAS A with significant affinity. NMR
and UV/visible spectroscopy have failed to reveal any
compounds bound to bacterially expressed hPASK PAS
A. Because PAS domains can bind a variety of cofactors,
it is likely that the physiologically relevant compound
(or class of compounds) is novel among characterized
PAS binding ligands and not abundant in E. coli. Clues
to the identities of such compounds are most likely to
come from ongoing genetic and biochemical studies
showing that close relatives of human PASK in yeast
regulate carbohydrate metabolism and translation [9].
Those regulatory pathways provide a wide range of potential candidate ligands in the form of metabolites having chemical properties similar to the compounds identified from our screen.
and 2 in Figure 4B) also inhibit isolated forms of the
hPASK kinase domain at the high concentrations required for saturation. To address this, we are using the
SAR data obtained from the screen to guide the synthesis of structurally related compounds with optimized
binding affinities for hPASK PAS A. This will facilitate
studies of kinase regulation in vitro and in vivo.
In the absence of a natural ligand that binds the PAS
domain and activates the hPASK activity, we emphasize
that the model presented in Figure 7 is speculative at
present. While alternatives can be generated from the
other kinds of interactions in which PAS domains engage, such as activation via the binding of additional
protein partners, our data provides strong circumstantial
support for small molecule ligands playing a role in the
activation process. In particular, we demonstrated that
small compounds specifically bound to hPASK PAS A
at the same regulatory site used by bona fide ligands
for other PAS domains (heme for FixL and FMN for Phy3).
Our site-directed mutagenesis results link this site to
the ability of hPASK PAS A to inhibit kinase activity,
suggesting that modest structural changes can switch
the regulatory ability of this domain.
Regulation of Kinase Activity by PAS Ligands
This work provides the first direct biophysical identification of the site used by any PAS domain to bind to
another protein. In hPASK, the flexible FG loop is the
primary interface for binding and inhibiting the neighboring kinase domain. The proximity of this loop to the
ligand binding site supports our model that small molecules may regulate enzymatic activity using the PAS
domain as a transducer to convert chemical signals into
a switch of kinase activity (Figure 7). Site-directed mutagenesis establishes the critical linkage between these
two sites, since the insertion of aromatic residues at
two sites within the ligand binding pocket (I203F and
C228F) abrogates kinase inhibition with selective structural perturbations in the F␣, FG, and H␤ regions.
These substitutions may mimick the regulatory ability
of aromatic compounds bound within the core of hPASK
PAS A to modulate the kinase activity of this protein.
We are not yet able to evaluate directly the functional
effects of ligands on the kinase, as the highest affinity
compounds identified from our PAS binding screen (1
Parallels with the Prokaryotic
Histidine Kinase, FixL
Despite the large evolutionary distance between PASK
and FixL, it is apparent that there are parallels in the
PAS-based regulatory mechanisms of the two proteins.
Both proteins bind hydrophobic compounds within their
PAS domains, using changes in the occupancy or conformations of those compounds to alter the structure of
the FG loop and surrounding regions. In turn, those
changes modulate the activity of a kinase domain within
the same protein.
For FixL, this pathway starts with a heme bound within
the PAS domain that directly contacts the FG loop via
the propionic acid moieties on this cofactor. Binding of
oxygen or other ligands to the distal side of the heme
results in structural changes in the porphyrin ring that
are relayed to the FG loop, as shown by crystal structures of several FixL-ligand complexes [13, 27]. Notably,
residues in this FG loop switch region have relatively
high temperature factors, suggesting that this loop is
inherently flexible [14, 27]. This parallels our observa-
Structure
1358
tions of extensive dynamics throughout the analogous
region of hPASK PAS A, suggesting that both proteins
may use this flexibility as an essential element of the
switching function.
A key difference between these two systems seems
to be the nature of the triggering signal. The triggering
signal is likely to be a diffusible organic ligand in hPASK,
compared with the diffusible gases that interact with
the heme iron in FixL. In light of the large structural
differences between the serine/threonine and histidine
kinases regulated by hPASK and FixL, respectively, it
is intriguing that PAS domains serve as regulatory elements for both.
Sensing Via PAS Domains
On the basis of limited characterization of the unliganded forms of PAS domain proteins, it is apparent
that many of those apoproteins are predominantly unstructured in solution. This is supported by studies of
the apo and holo forms of PYP, which show that the
removal of the hydroxycinnamic acid chromophore renders the protein significantly more accessible to deuterium exchange and prone to precipitation [28, 29]. Further evidence that certain PAS domains unfold in the
absence of their cognate ligands is provided by the association between hsp90 and the apo form of the ligand
binding PAS B domain of the aryl hydrocarbon receptor
(AHR) [30]. A variety of aromatic compounds that activate the AHR response pathway displace hsp90 from
this complex, suggesting that those chemicals stabilize
a folded form of the PAS domain. These data suggest
that a significant number of PAS domains fold in a ligand-dependent manner, opening the prospects for this
mechanism to be used as a regulated step in PASmediated signaling. This hypothesis also relates to a
proposed mechanism for the photosensory action of
PYP, where a light-triggered rearrangement of the chromophore within the hydrophobic core of this protein
destabilizes its fold [28, 31].
In contrast to the above cases, some PAS domains
fold stably in a ligand-free state, as demonstrated by
the structures of hPASK PAS A and the N-terminal PAS
domain of the HERG potassium channel [6]. Both of
these domains contain well-packed hydrophobic cores
that lack any obvious cavities for binding small organic
compounds. This stability does not preclude the use of
such domains as sensors, as we have shown by the
specific incorporation of organic compounds within the
core of hPASK PAS A.
Our observations imply several points that may be
relevant to the widespread use of PAS domains as biological sensory modules. First, a very broad range of
PAS domains, including those that do not copurify with
ligands when isolated from natural sources, may serve
sensory roles in vivo. Second, the apo forms of various
PAS domains have a broad range of stabilities. This
provides a mechanism to balance needs for affinity and
specificity in ligand binding by altering the energetic
linkage of this process with protein folding [32]. Understanding this tuning may provide a key to understanding
why PAS domains are well suited to be sensors for an
extremely wide range of stimuli.
Biological Implications
PAS domains are widely used through biology as protein/protein interaction modules in signal transduction
proteins. Many of these domains serve as environmental
sensors, binding cofactors within their cores to sense
diverse stimuli. Changes in these stimuli alter the cofactor conformation or occupancy, altering the structure
and dynamics of the surrounding PAS domain and thus
switching its ability to interact with protein targets.
This study focuses on the N-terminal PAS domain
from the human PAS-containing kinase, PAS kinase
(hPASK PAS A). Prior work demonstrated that hPASK
PAS A is capable of specifically inhibiting its own kinase
activity in cis [7]. We used solution NMR methods to
determine the structure of hPASK PAS A, demonstrating
that it contained two unusually flexible segments. One
of these was near the ligand binding site of other kinaseregulating PAS domains, leading us to examine whether
hPASK PAS A could bind small organic compounds by
screening a chemical library for their ability to bind this
domain. Several compounds bound specifically within
the core of hPASK PAS A, suggesting that biological
cofactors could also do so. We further determined the
location of the kinase binding surface on hPASK PAS
A, identifying that the second flexible region is involved
in this interaction. By introducing point mutations within
the ligand binding area, we demonstrated that small
structural perturbations there resulted in the loss of kinase inhibition. Taken together, these data suggest a
model where small organic cofactors could control the
activity of hPASK by acting through the PAS domain.
This study sets the foundation for the discovery of artificial ligands of hPASK PAS A that will aid investigations
of the biological role of PAS kinase.
Experimental Procedures
Protein Expression and Purification
To identify the minimal region containing the hPASK PAS A domain,
DNA sequences encoding residues 131–285 and several C-terminal
truncations were amplified by PCR and subcloned into a modified
form of the T7-based pHis-Parallel1 expression vector [33], where
the His6 tag was replaced by the streptococcal protein G ␤1 domain
(G␤1). E. coli BL21(DE3) cells were transformed with these plasmids
and grown in 100 ml M9 media cultures containing 1 g/l
15
N-ammonium chloride (Cambridge Isotope Labs). Protein overexpression was induced by adding 0.5 mM isopropyl-1-thio-␤-Dgalactopyranoside (IPTG) and allowed to proceed overnight at 20⬚C.
These cultures were harvested, resuspended in 50 mM Tris (pH 6.5)
and 100 mM NaCl, lysed by high-pressure extrusion, and centrifuged, and supernatants were concentrated to 0.5 ml. By examining
the amide proton chemical shift dispersion of 15N/1H HSQC spectra
recorded on these cell lysates [11], we determined that residues
131–237 encoded a minimal folded fragment of this PAS domain
and used it for all further studies.
Uniformly 15N-labeled samples of hPASK PAS A were purified from
E. coli cell lysates by SOURCE 15Q anion exchange chromatography
(Amersham Pharmacia Biotech), digestion with TEV protease, IgGSepharose flow-through chromatography, and high-resolution
MonoQ anion exchange chromatography. The molecular weight of
the purified protein, containing eight vector-derived residues, was
determined by ESI-MS to be 12796.0 Da (12796.5 Da calculated
from sequence, assuming 15N labeling of ⬎98%). NMR samples
typically contained 0.7–1.2 mM protein in 20 mM sodium phosphate
buffer (pH 6.5), 50 mM NaCl, 5 mM DTT, 6 mM NaN3, and a protease
inhibitor cocktail (Sigma) in 90% H2O/10% D2O, unless otherwise
noted. Typical yields from this approach were 20 mg/l.
PAS Domain Structure and Function in PAS Kinase
1359
Uniformly 15N, 13C-labeled samples were produced in a similar
manner as described above, supplementing the standard M9 media
with 3 g/l 13C6 glucose and 10 ml/l U-15N/13C Bioexpress 1000 growth
media (Cambridge Isotope Labs).
Mutagenesis
Point mutants of full-length hPASK were created from the previously
described KIAA0135 probe [7] containing the wild-type DNA and
primers, including the desired mutation, with the Quick Change SiteDirected Mutagenesis Kit (Stratagene), as directed by the manufacturer. HPASK PAS A mutants were obtained by PCR amplification
of the corresponding full-length sequence and subcloned into the
G␤1-parallel vector (see above). Transformations, protein induction,
and purification were performed as described above.
NMR Spectroscopy
All NMR data were recorded at 25⬚C with Varian Inova 500 and
600 MHz spectrometers, with nmrPipe for data processing [34] and
NMRview for analysis [35]. Backbone chemical shift assignments
were obtained from 3D HNCACB, CBCA(CO)NH, HNCO, and HNHA
spectra. Aliphatic side chain resonances were assigned from C(CO)NH TOCSY, H(CO)-NH TOCSY, and HCCH-TOCSY experiments, with
constant time 1H/13C HSQC spectra recorded on a uniformly 10%
13
C-labeled sample to obtain stereospecific chemical shift assignments of valine and leucine methyl groups [36]. Aromatic proton
chemical shift were assigned via 2D DQF-COSY, TOCSY (␶m ⫽ 30
ms), and NOESY (␶m ⫽ 120 ms) spectra of an unlabeled sample in
99.9% D2O.
15
N relaxation analyses were based on measurements of 15N T1,
T2, and 15N{1H} NOE recorded at 500 and 600 MHz by modified 15N/
1
H HSQC experiments [37]. For the 600 MHz data, 15N T1 (T2) values
were determined from spectra with relaxation delays of 11, 77, 165,
253, 374, 516, 725, and 1066 ms (16, 33, 49, 66, 82, 99, 115, and
132 ms). Comparable delays were used for these measurements at
500 MHz. Heteronuclear 15N{1H} NOE values were obtained from
spectra recorded with and without a 3 s proton saturation delay.
Structure Determination
Distance constraints were obtained from 3D 15N-edited NOESY (␶m ⫽
150 ms) and 15N, 13C-edited NOESY (␶m ⫽ 100 ms) spectra recorded
on a uniformly 15N, 13C-labeled sample and a 2D NOESY (␶m ⫽ 120
ms) spectrum recorded on an unlabeled sample. Peak intensities
from these spectra were used to classify the upper bounds of distance constraints in three categories, strong (2.8 Å), medium (3.5 Å),
or weak (5.0 Å), and corrected for multiplicity [38]. All lower bounds
for these constraints were set to 1.8 Å. Hydrogen bond constraints
were set for amide protons protected from exchange with D2O solvent for more than 6 hr (25⬚C, pH 6.5) with 1.3 Å ⬍ dNH-O ⬍ 2.5 Å
and 2.3 Å ⬍ dN-O ⬍ 3.5 Å constraints. Constraints for the backbone
dihedral angles φ and ␺ were generated from an analysis of 1H␣,
13
C␣, 13C␤, 13CO, and 15N chemical shifts with the program TALOS
[39]. Error bounds on these constraints were set to two times the
standard deviation of the TALOS predictions, with a minimum bound
of ⫾30⬚.
Initial structures were determined by a standard simulatedannealing protocol in torsion angle space as implemented in CNS
0.9 [40]. Constraints for these calculations consisted of 515 manually
assigned NOE-based distance constraints, 116 torsion angle constraints, and 48 hydrogen bond constraints. Sixty low-resolution
structures were generated in this manner. The five lowest-energy
structures of this ensemble were further refined with ARIA [41],
which identified a total of 2520 NOE-based constraints (1683 unambiguous, 837 ambiguous) from the three NOESY spectra listed
above. These constraints were combined with 61 15N-1H residual
dipolar coupling constraints obtained from a sample aligned in
buffer containing 20 mg/ml Pf1 filamentous bacteriophage. Out of
the 100 structures generated in ARIA with these constraints, the
twenty lowest-energy structures were analyzed with the programs
MOLMOL [42] and PROCHECK-NMR [43] to generate the statistics
provided in Table 1.
The closest structure to the mean out of the ensemble of 20 lowenergy structures was superimposed against other PAS domains
with the Swiss-PDB Viewer program [44] with the backbone atoms
of residues in secondary structure elements (A␤, B␤, C␣, D␣, G␤, H␤,
and I␤). F␣ was not used for superposition because of its disordered
nature in hPASK PAS A and the variability in orientation it assumes
in different PAS domains [26]. The calculated rmsds were 1.40 Å,
1.28 Å, 1.26 Å, and 1.30 Å for PYP, FixL, HERG, and Phy3, respectively.
Kinase Assays
Kinase domain (hPASK⌬N; residues 950–1323) was prepared and
assayed as previously described [7]. In brief, assay conditions contained 40 mM HEPES (pH 7.0), 5 mM MgCl2, 5 mM KCl, 200 ␮M
synthetic peptide substrate AMARA, and 18 nM kinase. The reaction
was started with the addition of 2 mM ␥-[32P]ATP and allowed to
proceed 4–8 min at 30⬚C. Reactions were spotted on P81 paper and
then washed with 85% H3PO4 five times. The papers were dried and
radioactivity measured with a scintillation counter.
Ligand Screening
The ligand search was conducted in three steps. First, mixtures of
five compounds dissolved in d6-DMSO were added to 15N-labeled
protein solutions (250 ␮M) where the final concentration for each
ligand was 1 mM. Potential hits were identified by comparing the
chemical shift changes of peaks in 1H/15N HSQC experiments for
each sample with a control containing only protein and d6-DMSO.
Second, identification of compounds binding to the protein was
achieved by screening the individual components of each hit selected before. Finally, determination of dissociation constants was
carried out by titration of increasing amounts of ligand to the protein
solution [22]. Assuming fast exchange in the NMR timescale (corresponding to Kd values of greater than ⵑ20–50 ␮M), a nonlinear
regression fit with equation 1 was used to obtain the corresponding Kd.
⌬␦ ⫽ ⌬␦max ⫻ [(L ⫹ PT ⫹ Kd) ⫺ ((L ⫹ PT ⫹ Kd)2
(1)
⫺ (4 ⫻ L ⫻ PT))1/2]/[2 ⫻ PT]
where ⌬␦ is the observed chemical shift change at the ligand concentration L, ⌬␦max is the change in chemical shift at saturation, and
PT is the total protein concentration.
Isothermal titration calorimetry was performed with a VP-ITC microcalorimeter (MicroCal, Northhampton, MA) for those compounds
exhibiting peak broadening because of intermediate timescale exchange during the titration (compounds 1 and 2). Briefly, a 1 mM
solution of compound dissolved in 50 mM HEPES (pH 6.5), 50 mM
NaCl, and 2% DMSO buffer was titrated into a 45 ␮M protein solution
in the same buffer. Dissociation constants were obtained by analysis
of the data with the vendor-supplied software.
CD Spectroscopy
All CD experiments were recorded with an Aviv 62DS spectropolarimeter with 1 mm path length cells. Samples consisted of 15 ␮M
protein in 50 mM Tris (pH 7.0) buffer. Thermal denaturation experiments were obtained by monitoring the CD signal at 222 nm with a
bandwidth of 1.5 nm. Data were collected in 1⬚C steps with three
repeats for each point, an averaging time of 3 s, and an equilibration
time of 12 s.
Acknowledgments
We thank John Falck for providing an initial set of compounds from
his laboratory for preliminary screening studies, Lewis Kay, Michael
Rosen, Ronald Estabrook, and Marie-Alda Gilles-Gonzales for constructive criticism on this manuscript, and Steve McKnight for ongoing discussions and support of this work. We gratefully acknowledge
financial support from the NIH (CA-90601), Searle Scholars Program,
Robert A. Welch Foundation, and The University of Texas Southwestern Endowed Scholar Program.
Received: April 29, 2002
Revised: June 12, 2002
Accepted: August 16, 2002
Structure
1360
References
1. Taylor, B.L., and Zhulin, I.B. (1999). PAS domains: internal sensors of oxygen, redox potential and light. Microbiol. Mol. Biol.
Rev. 63, 479–506.
2. Gu, Y.Z., Hogenesch, J.B., and Bradfield, C.A. (2000). The PAS
superfamily: sensors of environmental and developmental signals. Ann. Rev. Pharmacol. Toxicol. 40, 519–561.
3. Gilles-Gonzalez, M.A. (2001). Oxygen signal transduction.
IUBMB Life 51, 165–173.
4. Christie, J.M., Reymond, P., Powell, G.K., Bernasconi, P., Raibekas, A.A., Liscum, E., and Briggs, W.R. (1998). Arabidopsis
NPH1: a flavoprotein with the properties of a photoreceptor for
phototropism. Science 282, 1698–1701.
5. Borgstahl, G.E.O., Williams, D.R., and Getzoff, E.D. (1995). 1.4 Å
structure of photoactive yellow protein, a cytosolic photoreceptor: unusual fold, active site and chromophore. Biochemistry
34, 6278–6287.
6. Cabral, J.H.M., Lee, A., Cohen, S.L., Chait, B.T., Li, M., and
Mackinnon, R. (1998). Crystal structure and functional analysis
of the HERG potassium channel N terminus: a eukaryotic PAS
domain. Cell 95, 649–655.
7. Rutter, J., Michnoff, C., Harper, S.M., Gardner, K.H., and
McKnight, S.L. (2001). PAS kinase: an evolutionarily conserved
serine/threonine kinase regulated in cis by PAS domains. Proc.
Natl. Acad. Sci. USA 98, 8991–8996.
8. Hofer, T., Spielmann, P., Stengel, P., Stier, B., Katschinski, D.M.,
Desbaillets, I., Gassman, M., and Wenger, R.H. (2001). Mammalian PASKIN: a PAS-serine/threonine kinase related to bacterial
oxygen sensors. Biochem. Biophys. Res. Commun. 288,
757–764.
9. Rutter, J., Probst, B.L., and McKnight, S.L. (2002). Coordinate
regulation of sugar flux and translation by PAS kinase. Cell 111,
in press.
10. Zhulin, I.B., Taylor, B.L., and Dixon, R. (1997). PAS domain
S-boxes in archaea, bacteria and sensors for oxygen and redox.
Trends Biol. Sci. 22, 331–333.
11. Huth, J.R., Bewley, C.A., Jackson, B.M., Hinnebusch, A.G.,
Clore, G.M., and Gronenborn, A.M. (1997). Design of an expression system for detecting folded protein domains and mapping
macromolecular interactions by NMR. Protein Sci. 6, 2359–2364.
12. Gronenborn, A.M., and Clore, G.M. (1996). Rapid screening for
structural integrity of expressed proteins by heteronuclear NMR
spectroscopy. Protein Sci. 5, 174–177.
13. Gong, W., Hao, B., Mansy, S.S., Gonzalez, G., Gilles-Gonzalez,
M.A., and Chan, M.K. (1998). Structure of a biological oxygen
sensor: a new mechanism for heme-driven signal transduction.
Proc. Natl. Acad. Sci. USA 95, 15177–15182.
14. Miyatake, H., Mukai, M., Park, S.Y., Adachi, S., Tamura, K.,
Nakamura, H., Nakamura, K., Tsuchiya, T., Iizuka, T., and Shiro,
Y. (2000). Sensory mechanism of oxygen sensor FixL from Rhizobium meliloti: crystallographic, mutagenesis and resonance
Raman spectroscopic studies. J. Mol. Biol. 301, 415–431.
15. Farrow, N.A., Zhang, O., Szabo, A., Torchia, D.A., and Kay, L.E.
(1995). Spectral density function mapping using 15N relaxation
data exclusively. J. Biomol. NMR 6, 153–162.
16. Bracken, C., Carr, P.A., Cavanagh, J., and Palmer, A.G. (1999).
Temperature dependence of intramolecular dynamics of the
basic leucine zipper of GCN4: implications for the entropy of
association with DNA. J. Mol. Biol. 285, 2133–2146.
17. Düx, P., Rubinstenn, G., Vuister, G.W., Boelens, R., Mulder,
F.A.A., Hard, K., Hoff, W.D., Kroon, A.R., Crielaard, W., Hellingwerf, K.J., et al. (1998). Solution structure and backbone dynamics of the photoactive yellow protein. Biochemistry 37, 12689–
12699.
18. Holdeman, T.C., and Gardner, K.H. (2001). 1H, 13C and 15N chemical shift assignments for the N-terminal PAS domain of
mNPAS2. J. Biomol. NMR 21, 383–384.
19. Crosson, S., and Moffat, K. (2001). Structure of a flavin-binding
plant photoreceptor domain: insights into light-mediated signal
transduction. Proc. Natl. Acad. Sci. USA 98, 2995–3000.
20. Fejzo, J., Lepre, C.A., Peng, J.W., Bemis, G.W., Ajay, Murcko
M.A., and Moore, J.M. (1999). The SHAPES strategy: an NMR-
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
based approach for lead generation in drug discovery. Chem.
Biol. 6, 755–769.
Hajduk, P.J., Bures, M., Praestgaard, J., and Fesik, S.W. (2000).
Priveleged molecules for protein binding identified from NMRbased screening. J. Med. Chem. 43, 3443–3447.
Shuker, S.B., Hajduk, P.J., Meadows, R.P., and Fesik, S.W.
(1996). Discovering high-affinity ligands for proteins: SAR by
NMR. Science 274, 1531–1534.
Matsuo, H., Walters, K.J., Teruya, K., Tanaka, T., Gassner, G.T.,
Lippard, S.J., Kyogoku, Y., and Wagner, G. (1999). Identification
by NMR spectroscopy of residues at contact surfaces in large,
slowly exchanging macromolecular complexes. J. Am. Chem.
Soc. 121, 9903–9904.
Shekhtman, A., Ghose, R., Wang, D., Cole, P.A., and Cowburn,
D. (2001). Novel mechanism of regulation of the non-receptor
protein tyrosine kinase Csk: insights from NMR mapping studies
and site-directed mutagenesis. J. Mol. Biol. 314, 129–138.
Wishart, D.S., and Sykes, B.D. (1994). The 13C chemical-shift
index: a simple method for the identification of protein secondary structure using 13C chemical-shift data. J. Biomol. NMR 4,
171–180.
Pellequer, J.L., Brudler, R., and Getzoff, E.D. (1999). Biological
sensors: more than one way to sense oxygen. Curr. Biol. 9,
R416–418.
Gong, W.M., Hao, B., and Chan, M.K. (2000). New mechanistic
insights from structural studies of the oxygen-sensing domain of
Bradyrhizobium japonicum FixL. Biochemistry 39, 3955–3962.
Hoff, W.D., Xie, A., Van Stokkum, I.H.M., Tang, X.J., Gural, J.,
Kroon, A.R., and Hellingwerf, K.J. (1999). Global conformational
changes upon receptor stimulation in photoactive yellow protein. Biochemistry 38, 1009–1017.
Kroon, A.R., Hoff, W.D., Fenneman, H.P.M., Gijzen, J., Koomen,
G.J., Verhoeven, J.W., Crielaard, W., and Hellingwerf, K.J.
(1996). Spectral tuning, fluorescence, and photoactivity in hybrids of photoactive yellow protein, reconstituted with native
or modified chromophores. J. Biol. Chem. 271, 31949–31956.
Coumailleau, P., Poellinger, L., Gustafsson, J.-A., and Whitelaw,
M.L. (1995). Definition of a minimal domain of the dioxin receptor
that is associated with hsp90 and maintains wild type ligand
binding affinity and specificity. J. Biol. Chem. 270, 25291–25300.
Ohishi, S., Shimizu, N., Mihara, K., Imamoto, Y., and Kataoka,
M. (2001). Light induces destabilization of photoactive yellow
protein. Biochemistry 40, 2854–2859.
Dyson, H.J., and Wright, P.E. (2002). Coupling of folding and
binding for unstructured proteins. Curr. Opin. Struct. Biol. 12,
54–60.
Sheffield, P., Garrard, S., and Derewenda, Z. (1999). Overcoming
expression and purification problems of RhoGDI using a family
of “parallel” expression vectors. Protein Expr. Purif. 15, 34–39.
Delaglio, F., Grzesiek, S., Vuister, G.W., Zhu, G., Pfeifer, J., and
Bax, A. (1995). NMRPipe: a multidimensional spectral processing system based on UNIX pipes. J. Biomol. NMR 6,
277–293.
Johnson, B.A., and Blevins, R.A. (1994). NMRView: a computer
program for the visualization and analysis of NMR data. J. Biomol. NMR 4, 603–614.
Neri, D., Szyperski, T., Otting, G., Senn, H., and Wüthrich, K.
(1989). Stereospecific nuclear magnetic resonance assignments of the methyl groups of valine and leucine in the DNAbinding domain of the 434 repressor by biosynthetically directed
fractional 13C labeling. Biochemistry 28, 7510–7516.
Farrow, N.A., Muhandiram, D.R., Singer, A.U., Pascal, S.M., Kay,
C.M., Gish, G., Shoelson, S.E., Pawson, T., Forman-Kay, J.D.,
and Kay, L.E. (1994). Backbone dynamics of a free and a phosphopeptide-complexed Src homology 2 domain studied by 15N
NMR relaxation. Biochemistry 33, 5984–6003.
Fletcher, C.M., Jones, D.N.M., Diamond, R., and Neuhaus, D.
(1996). Treatment of NOE constraints involving equivalent or
nonstereoassigned protons in calculations of biomacromolecular structures. J. Biomol. NMR 8, 292–310.
Cornilescu, G., Delaglio, F., and Bax, A. (1999). Protein backbone angle restraints from searching a database for chemical
shift and sequence homology. J. Biomol. NMR 13, 289–302.
Brunger, A.T., Adams, P.D., Clore, G.M., DeLano, W.L., Gros,
PAS Domain Structure and Function in PAS Kinase
1361
41.
42.
43.
44.
45.
P., Grosse-Kunstleve, R.W., Jiang, J.S., Kuszewski, J., Nilges,
M., Pannu, N.S., et al. (1998). Crystallography and NMR system
(CNS): a new software system for macromolecular structure
determination. Acta Crystallogr. D Biol. Crystallogr. 54, 905–921.
Nilges, M., and O’Donoghue, S.I. (1998). Ambiguous NOEs and
automated NOE assignment. Prog. Nucleic Magn. Reson. Spectrosc. 32, 107–139.
Koradi, R., Billeter, M., and Wüthrich, K. (1996). MOLMOL: a
program for display and analysis of macromolecular structures.
J. Mol. Graph. 14, 51–55.
Laskowski, R.A., Rullman, J.A.C., MacArthur, M.W., Kaptein,
R., and Thornton, J.M. (1996). AQUA and PROCHECK-NMR:
programs for checking the quality of protein structures solved
by NMR. J. Biomol. NMR 8, 477–486.
Guex, N., and Peitsch, M.C. (1997). SWISS-MODEL and the
Swiss-PdbViewer: an environment for comparative protein
modeling. Electrophoresis 18, 2714–2723.
Corpet, F. (1988). Multiple sequence alignment with hierarchical
clustering. Nucleic Acids Res. 16, 10881–10890.
Accession Numbers
The coordinates have been deposited in the Protein Data Bank
with the accession number 1LL8, and chemical shifts have been
deposited at BioMagResBank with the accession number 5354.