Download from membrane-embedded targets to amyloid fibrils

Signalling
Solid-state NMR spectroscopy as a tool for drug
design: from membrane-embedded targets to
amyloid fibrils
D.A. Middleton1
School of Biological Sciences, University of Liverpool, Crown Street, Liverpool L69 7ZB, U.K.
Abstract
Structure-based design has gained credibility as a valuable component of the modern drug discovery
process. The technique of SSNMR (solid-state NMR) promises to be a useful counterpart to the conventional
experimental techniques of X-ray crystallography and solution-state NMR for providing structural features
of drug targets that can guide medicinal chemistry towards drug candidates. This article highlights some
recent SSNMR approaches from our group for identifying active compounds, such as enzyme inhibitors,
receptor antagonists and peptide agents, that prevent the aggregation of amyloid proteins involved in
neurodegenerative diseases. It is anticipated that the use of SSNMR in drug discovery will become more
widespread in the wake of advances in hardware and methodological developments.
Introduction
There are now numerous examples of how the search for new
drug candidates has been hastened by structural information
provided by X-ray crystallography, NMR spectroscopy and
predictive in silico methods [1–4]. A general approach to
rational structure-based drug design begins with the structure
of a target protein that has been determined experimentally
or by homology modelling. Computational methods are then
used to position compounds or fragments available from a
library into selected regions of the structure. The compounds
predicted to have the most favourable interactions with the
target site are then screened for activity. Often a second cycle
of structure determination of the target in complex with
the best leads from the first iteration can suggest how the
compounds can be optimized to produce sub-micromolar
ligands, for example by linking together smaller fragments
[5].
A sizeable proportion of drug targets are not readily
amenable to structural analysis by conventional techniques
owing to the insolubility of the target in aqueous
environments. A prime example is found in the proteins that
function from within cellular membranes, most notably the
GPCRs (G-protein-coupled receptors), which constitute
over half of the targets for currently marketed drugs [6]. The
structural database of membrane proteins is growing at a
steady pace, but handling difficulties have hampered crystallographic studies and, until recently, have precluded
structural analysis of membrane proteins by solution NMR.
Alongside membrane-embedded proteins, amyloid fibrils
Key words: amyloid, cross-polarization magic-angle spinning (CP-MAS), drug discovery,
G-protein-coupled receptor (GPCR), NMR, solid-state NMR.
Abbreviations used: CP-MAS, cross-polarization magic-angle spinning; GPCR, G-protein-coupled
receptor; hIAPP, human islet amyloid polypeptide; SSNMR, solid-state NMR; SRE, self-recognition
element.
1
email [email protected]
represent another important class of disease targets that may
be classified as being insoluble in their disease-associated
forms. These protein aggregates are deposited as pathological
lesions within areas of the brain and other tissues [7] and
are associated with devastating and widespread diseases
including Alzheimer’s, Parkinson’s and Type 2 diabetes. An
attractive therapeutic strategy is to develop drugs that inhibit
the aggregation process and arrest or impede the formation
of the toxic amyloid species [8]. The search for such therapies
would again benefit from knowledge of the structural details
of the amyloid target.
SSNMR (solid-state NMR) is proving to be a powerful
tool for the structural analysis of membrane proteins and
amyloid fibrils [9] and is attracting attention as a potentially
valuable counterpart to the conventional methods for
structure-based drug design. SSNMR embraces a collection
of diverse methods that take advantage of the spatial and
dynamic properties of solid and semi-solid materials such as
biomembranes and protein assemblies to extract information
about structure and dynamics [10,11]. Some examples are
discussed here, from this research group, of how SSNMR can
contribute to the early phases of drug design and discovery.
Mapping the ligand-binding sites of
membrane-embedded receptors
A convenient starting point in structure-based drug discovery
is to identify where the native ligand or an unoptimized
lead compound interacts with its target and to produce a
three-dimensional model of the ligand within the binding
pocket. The SSNMR method of CP-MAS (cross-polarization
magic-angle spinning) can be used to obtain high-resolution
spectra from which to detect weak interactions of ligands with
receptors embedded in native membrane or isolated and
incorporated into defined lipid bilayers [12–14]. When an
C The
C 2007 Biochemical Society
Authors Journal compilation 985
986
Biochemical Society Transactions (2007) Volume 35, part 5
Figure 1 Mapping the binding sites for ligands in membrane-embedded receptors using a combination of CP-MAS NMR and
site-directed mutagenesis
The example shown is for the substrate [1-13 C]uridine binding to the E. coli nucleoside transport protein NupC. (a) The
13 C-NMR
spectrum reveals a clear diagnostic peak for the substrate interacting with the protein. The position of the 13 C
label is marked by an asterisk (*). (b) Plots of peak intensities for the substrate measured at different NMR contact times
are sensitive to the dissociation constant K d for the complex. The intensity plots for substrate binding to wild-type NupC
and the mutant G146C (left) are compared with a range of simulated curves to find the value of K d in best agreement
with the experimental data (right). The substrate clearly has reduced affinity for the mutant, suggesting that the residue
may be situated close to the binding site.
isotopically labelled (e.g. 13 C, 15 N or 19 F) ligand is added to
a membrane sample, a peak from the ligand is observed in the
NMR spectrum only if the mobility of the ligand is restrained
by the membrane protein target [12]. The advantage of
CP-MAS is that, with appropriate control experiments,
specific ligand binding can be observed even when the target
receptor constitutes less than 20% of the total protein present
in the membrane sample which alleviates the requirement
for purification procedures. CP-MAS can also provide
quantitative information about ligand-binding affinities by
monitoring the build-up of peak intensities at different ligand
concentrations, NMR contact times or in the presence of
competitor ligands. Plots of NMR peak intensities for a ligand
can be compared with simulated curves to extract dissociation
constants. By measuring ligand-binding constants for
mutants of the target receptor, CP-MAS can also be used
to map the binding sites of ligands in membrane-embedded
receptors. This approach has been applied to the bacterial
nucleoside transport protein NupC for which a number of
C The
C 2007 Biochemical Society
Authors Journal compilation mutants (including G146C) have cross-polarization buildup curves that are consistent with reduced substrate-binding
affinity in comparison with the wild-type protein (Figure 1).
These results suggest that the the substituted residues are
situated close to the substrate-binding site.
A powerful feature of SSNMR is its ability to detect
direct interactions between specific residues in a target protein and the bound ligand. Smith and co-workers used
two-dimensional SSNMR experiments to identify couplings
between 13 C-labelled tyrosine, glycine, serine and threonine
residues around the binding pocket of the GPCR rhodopsin
and the covalently linked chromophore retinal [15]. It was
shown that the transition of rhodopsin to the MII state
involves the disruption of helix–helix interactions in the transmembrane domain as retinal moves some 5 Å (1 Å = 0.1 nm)
towards helix 5 and the C-20 methyl group of retinal
moves towards extracellular loop 2. We have recently applied
similar methods to identify interactions between 13 C-labelled
tryptophan residues in the sugar transport protein GalP from
Signalling
Escherichia coli and the bound substrate glucose. In this case
it was necessary to modify the methodology to overcome difficulties associated with weak non-covalent receptor–ligand
interactions, which now extends the range of this approach
into the realm of important drug targets such as GPCRs.
Structures of bound ligands
SSNMR can provide information about the molecular
conformations of native ligands, lead compounds and drugs
bound to target receptors. Such information can be exploited
to refine the structures of known ligands to improve binding
affinity or efficacy. This approach was first shown for flexible
inhibitors of the gastric H+ /K+ -ATPase, a proton pump that
is responsible for secretion of acid into gastric glands and is
thus a target for drugs in the treatment of gastric ulcer disease.
SSNMR measurements of conformationally diagnostic 13 C–
13
C and 13 C–19 F distances between labels placed within a
bound inhibitor revealed details about how it interacts with
its target even in the absence of a high-resolution structure
for this protein. The SSNMR measurements suggested how
the inhibitor could be chemically constrained in its active
conformation to produce an analogue that is 10 times more
active than the flexible inhibitor [16,17]. Similar experiments
have also given insights into how a homologous protein,
the cardiac Na+ /K+ -ATPase, is inhibited by high-affinity
cardiac glycoside inhibitors, collectively known as digitalis,
which have been used for over 200 years in the treatment of
congestive heart failure [18].
These methods are ideally suited to determining the structures of GPCR ligands in their active sites, which remains
a major goal for this group and for others in academia and
industry. Progress has been slow, however, owing to technical
difficulties in the synthesis of suitably isotope-labelled ligands
to provide sufficient geometric constraints on molecular conformation, compounded by the lack of availability of GPCRs
in high enough quantities for NMR analysis. We are taking
measures in this laboratory to alleviate the first problem, and
the second issue is being overcome by collaborations with scientists in the pharmaceutical industry who have the resources
and expertise to express and purify receptors on a large scale.
Our developments in these areas are described below.
Most SSNMR experiments on membrane proteins, and
their bound ligands, substrates and prosthetic groups, rely on
direct observation of couplings between naturally rare nuclei
(13 C, 15 N, 2 H and 19 F) in isotope-enriched samples rather
than direct measurements of proton–proton couplings,
which present certain technical difficulties. Unfortunately
the number of sites in a ligand that can be isotopically labelled
is generally rather low owing to the limited availability of
labelled precursors and by restrictions in the synthetic route
[19]. Consequently, for a typical ligand having three or more
degrees of conformational freedom, the number of geometric
constraints (interatomic distances and angles) is insufficient
to determine the ligand conformation unambiguously. A
good example of this is the histamine H2 GPCR antagonist
cimetidine. Cimetidine can be prepared by the commercial
synthetic route with a 13 C label and a 15 N label at the two ends
of the molecule (Figure 2a) but the introduction of additional
labels would require a significant investment of resource and
expense. The 13 C–13 C distance in the solid crystalline form is
approx. 3.7 Å, but a conformation grid search varying each of
the seven degrees of freedom in 30◦ increments identifies over
5000 ‘hits’ consistent with this constraint. We have therefore
devised a strategy in which the bonded protons are utilized
indirectly to provide three additional constraints from the
same double-labelled compound [20]. By measuring C-N-H,
H-C-N and H-C-N-H angles, the number of hits can be
reduced to less than 20, all of which fall within just two
closely related families of conformations (Figure 2a). The H2
receptor is not yet available to us in sufficient quantitites to
repeat these experiments for cimetidine bound to its target,
but this strategy is instead being applied to determine the
conformation of β 2 -adrenergic receptor ligands salbutamol
and salmeterol in their binding site. In collaboration with
pharmaceutical companies, the β 2 -adrenergic receptor has
been prepared retaining ligand binding in its native membrane
and reconstituted into lipid bilayers suitable for NMR
analysis.
Structure-based design of amyloid
inhibitors
All amyloid fibrils contain repeat arrays of β-strands arranged
with interstrand hydrogen bonds parallel to the fibril long
axis. The three-dimensional architecture of amyloid fibrils
varies from protein to protein, but involves a complex
network of electrostatic, van der Waals and π –π interactions
between amino acid side groups, which bring together laminae of β-sheets arranged with water-exposed faces and waterexcluded cores. SSNMR is the method of choice for probing
polypeptide structures and intermolecular contacts in intact
amyloid fibrils and could therefore be useful in drug design.
An attractive therapeutic strategy for preventing amyloid
deposition in brain and other tissues is to identify agents
that interfere with the fibril growth process by arresting the
onset or propagation of fibril growth [8,21,22]. Information
about the structural features of an amyloid species could help
to identify inhibitors of amyloidogenic proteins. SSNMR
can assist in this process by identifying specific interactions
between amino acid side groups that could be targeted by
agents aimed at destabilizing the growing fibrils or preventing
the lateral association of β-sheet layers. For example, SSNMR
measurements of interactions between 2 H- and 19 F-labelled
phenylalanine side groups in fibrils formed by a fragment
of the hIAPP (human islet amyloid polypeptide) reveals that
aromatic π –π interactions might contribute to the overall
architecture that holds the fibrils together [23] (Figure 3a).
One approach that is gathering interest is to synthesize
short peptides that correspond to a SRE (self-recognition
element) of the native amyloid sequence, but containing
key modifications so that the peptides bind to the complementary sequence of the aggregating parent protein and
prevent further aggregation. A convenient and useful way
to modify an SRE peptide is to introduce N-methylated
C The
C 2007 Biochemical Society
Authors Journal compilation 987
988
Biochemical Society Transactions (2007) Volume 35, part 5
Figure 2 Experiments toward the conformational analysis of isotopically labelled ligands when bound to GPCRs
(a) The histamine H2 receptor antagonist cimetidine can be labelled with 15 N and 13 C at the sites denoted by asterisks
(top). A SSNMR strategy to measure multiple distance and angle constraints for [13 C,15 C]cimetidine in the crystalline state
obtained by SSNMR (centre) restricts the conformation of the molecule to two similar families (bottom), which are close to
the true conformation determined by X-ray crystallography. This approach is now being adapted for the structural analysis of
salbutamol and salmeterol bound to the β 2 -adrenergic receptor β 2 AR (b). Salbutamol (top) is amenable to extensive isotope
labelling via existing synthetic routes. The β 2 -adrenergic receptor has been cloned and expressed with an N-terminal His6
tag, an N-terminal FLAG tag and mutated at two glycosylation sites using the Bac-to-BacR baculovirus expression system
(Invitrogen). An SDS/PAGE gel of the purified receptor after both FLAG- and His-affinity chromatography steps shows bands
from the receptor which were identified using MS analysis and immunoblotting using a horseradish-peroxidase-conjugated
anti-FLAG M2 antibody (middle). The detergent-solubilized receptor retains the ability to bind salmeterol (bottom). BAP,
bacterial alkaline phosphatase.
backbone amide groups so that the peptide is unable to
hydrogen-bond on one side [24–27]. SSNMR can guide the
design of modified SRE peptides by first identifying which
amino residues of the unmodified peptides are essential
for aggregation and which are redundant. This is done by
constructing an amyloidogenicity profile for the peptide
C The
C 2007 Biochemical Society
Authors Journal compilation based on chemical shifts and NMR peak widths. The flanking
sequences of residues that are redundant in the aggregation
of the peptide can then be eliminated leaving only the core
residues, usually six to eight in number. One or more of
these remaining residues can then be modified (e.g. by
N-methylation) and screened for their effects on aggregation
Signalling
Figure 3 Structure determination of amyloid fibrils provides information that can guide the design of inhibitors of protein aggregation
(a) Analysis of fibrils formed by a fragment of IAPP (islet amyloid polypeptide) reveals interactions between aromatic side
groups of adjacent β-sheets, which could be targeted by compounds to prevent fibril assembly or destabilize pre-existing
protofilaments. (b) A strategy for designing peptide inhibitors of amyloid formation. Fibrils of peptide fragments corresponding
to SREs of the parent protein are analysed by SSNMR to provide an amyloidogenicity profile which pinpoints which amino
acid residues are necessary for fibril formation and which are redundant. The peptide fragment is truncated to leave five
or six essential core amino acids, which are modified (e.g. with N-methyl groups) and tested for inhibition. In successful
cases, the modified peptides bind to the complementary sequence of the aggregating parent protein and prevent further
elongation of the fibrils.
of the parent amyloid protein (Figure 3b). This approach has
recently been used successfully to design an N-methylated
peptide which prevents the aggregation of α-synuclein, the
main protein component of Lewy bodies associated with
Parkinson’s disease and other neurodegenerative disorders.
Outlook
Examples of how SSNMR can assist structure-based design
and drug discovery are, at present, few and far between and
the technique is underexploited as a tool for drug design.
With advances in instrumentation and magnet technology,
sample preparation and methodology, the future of SSNMR
in drug discovery looks positive. Entire structures of
proteins in the solid state, including membrane proteins and
amyloid fibrils, are now being solved by NMR, and this
technique promises to offer a realistic alternative to X-ray
crystallography in providing the structures of experimentally
challenging drug targets as the starting point for drug design.
The following are acknowledged as collaborators on the work
described here: Professor S.A. Baldwin, Professor P.J.F. Henderson,
Dr S. Patching and Dr R.G. Herbert (University of Leeds) for bacterial
transport proteins; Dr J. Madine (University of Liverpool) and
Professor A. Doig (University of Manchester) for amyloid inhibitors;
and Dr S.J. Dowell and Dr M. Koglin (GlaxoSmithKline) for GPCRs.
C The
C 2007 Biochemical Society
Authors Journal compilation 989
990
Biochemical Society Transactions (2007) Volume 35, part 5
References
1 Laurie, A.T.R. and Jackson, R.M. (2006) Curr. Protein Pept. Sci. 7, 395–406
2 Reggio, P.H. (2006) AAPS J. 8, E322–E336
3 Patny, A., Desai, P.V. and Avery, M.A. (2006) Curr. Med. Chem. 13,
1667–1691
4 Williams, S.P., Kuyper, L.F. and Pearce, K.H. (2005) Curr. Opin. Chem. Biol.
9, 371–380
5 Hajduk, P.J. and Greer, J. (2007) Nat. Rev. Drug Discov. 6, 211–219
6 Drews, J. (2000) Science 287, 1960–1964
7 Chiti, F. and Dobson, C.M. (2006) Annu. Rev. Biochem. 75, 333–366
8 Estrada, L.D. and Soto, C. (2007) Curr. Top. Med. Chem. 7, 115–126
9 Tycko, R. (2006) in Amyloid, Prions, and Other Protein Aggregates, Pt C.,
pp. 103–122, Elsevier, Amsterdam
10 Hughes, C.E. and Baldus, M. (2005) Annu. Rep. NMR Spectrosc. 55,
121–158
11 Watts, A. (2005) Nat. Rev. Drug Discov. 4, 555–568
12 Boland, M.P. and Middleton, D.A. (2004) Magn. Reson. Chem. 42,
204–211
13 Patching, S.G., Brough, A.R., Herbert, R.B., Rajakarier, J.A.,
Henderson, P.J.F. and Middleton, D.A. (2004) J. Am. Chem. Soc. 126,
3072–3080
14 Spooner, P.J.R., Rutherford, N.G., Watts, A. and Henderson, P.J.F. (1994)
Proc. Natl. Acad. Sci. U.S.A. 91, 3877–3881
15 Patel, A.B., Crocker, E., Eilers, M., Hirshfeld, A., Sheves, M. and Smith,
S.O. (2004) Proc. Natl. Acad. Sci. U.S.A. 101, 10048–10053
C The
C 2007 Biochemical Society
Authors Journal compilation 16 Watts, J.A., Watts, A. and Middleton, D.A. (2001) J. Biol. Chem. 276,
43197–43204
17 Middleton, D.A., Robins, R., Feng, X.L., Levitt, M.H., Spiers, I.D., Schwalbe,
C.H., Reid, D.G. and Watts, A. (1997) FEBS Lett. 410, 269–274
18 Middleton, D.A., Rankin, S., Esmann, M. and Watts, A. (2000) Proc. Natl.
Acad. Sci. U.S.A. 97, 13602–13607
19 Patching, S.G., Middleton, D.A., Henderson, P.J.F. and Herbert, R.B.
(2003) Org. Biomol. Chem. 1, 2057–2062
20 Madine, J. and Middleton, D.A. (2006) Phys. Chem. Chem. Phys. 8,
5223–5228
21 Gilead, S. and Gazit, E. (2004) Angew. Chem. Int. Ed. 43, 4041–4044
22 Sciarretta, K.L., Gordon, D.J. and Meredith, S.C. (2006) in Amyloid, Prions,
and Other Protein Aggregates, Pt C., pp. 273–312, Elsevier, Amsterdam
23 Jack, E., Newsome, M., Stockley, P.G., Radford, S.E. and Middleton, D.A.
(2006) J. Am. Chem. Soc. 128, 8098–8099
24 Gordon, D.J., Sciaretta, K.L. and Meredith, S.C. (2001) Biochemistry 40,
8237–8245
25 Gordon, D.J., Tappe, R. and Meredith, S.C. (2002) J. Pept. Res. 60, 37–55
26 Kapurniotu, A., Schmauder, A. and Tenidis, K. (2002) J. Mol. Biol. 315,
339–350
27 Kokkoni, N., Stott, K., Amijee, H., Mason, J.M. and Doig, A.J. (2006)
Biochemistry 45, 9906–9918
Received 25 June 2007
doi:10.1042/BST0350985