| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
* Pharmacology and Toxicology Section, Institute of Pharmacy, and
Institute of Pharmaceutical Biology, Rheinische Friedrich-Wilhelms-University, Bonn, Germany;
Department of Pharmaceutical Chemistry, Institute of Pharmacy, Julius-Maximilians-University, Würzburg, Germany;
Institute of Medicinal and Toxicological Chemistry "Pietro Pratesi," University of Milan, Milan, Italy;
|| Department of Pharmacological, Biological and Applied Chemical Sciences, University of Parma, Parma, Italy;
¶ Institute of Pharmaceutical and Medicinal Chemistry, Heinrich-Heine-University, Düsseldorf, Germany; and
# Drug Discovery Biology Laboratory, Department of Pharmacology, Monash University, Victoria, Australia
2 Correspondence: Pharmacology and Toxicology Section, Institute of Pharmacy, Gerhard-Domagk-Str.3, D-53121 Bonn, Germany. E-mail: k.mohr{at}uni-bonn.de
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Key Words: agonism • allosterism • muscarinic acetylcholine receptor • signal trafficking • subtype selectivity • dynamic mass redistribution
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
, 2)
. Exogenous small-molecule GPCR activators with receptor subtype selectivity and signaling pathway specificity are highly desirable as tools for a targeted modulation of biological functions. A common approach in small-molecule development is to use the structure of the endogenous agonist as a template. In the organism, target selectivity of messenger compound action is fine-tuned by spatially restricted agonist release and inactivation. In contrast, exogenous activators that are distributed in the organism often suffer from a poor selectivity of action even if receptor subtypes are available. As a further complexity, a given GPCR may couple to a spectrum of intracellular signaling pathways. Such promiscuous agonism probably indicates that the three-dimensional (3-D) shape of an agonist-bound receptor is not stably fixed but free to adopt a subset of active conformations that differ in coupling specificity (e.g., ref. 3
).
The starting idea of the present work was that extending the interface area of ligand-receptor interactions should allow better exploitation of subtle differences in receptor subtype architecture and, additionally, restriction of conformational freedom of the active ligand-receptor complex. This should translate into subtype-specific binding and signaling pathway-selective activation. It is increasingly recognized that GPCRs contain allosteric ligand binding sites that lie outside the orthosteric binding area of the endogenous agonist (4)
. This leads to the concept of designing GPCR activators that attach simultaneously to the orthosteric and allosteric area, thereby providing selectivity of receptor binding and signaling pathway activation.
Muscarinic acetylcholine receptors are known as an excellent model system for studying allosteric/orthosteric interactions (5)
. Acetylcholine action is mediated by the five subtypes M1-M5 that control various autonomic and central nervous processes (6)
. M receptors belong to the rhodopsin-type subfamily of GPCRs, which have in common that the endogenous agonist binding site is located in the transmembrane region of the receptor (7)
. A high sequence homology between M-receptor subtypes in the orthosteric region precludes development of subtype selective orthosteric ligands (6)
. The allosteric site is located in the extracellular loop region that forms the entrance of the ligand binding crevice (8
, 9)
. Muscarinic allosteric agents commonly reveal pronounced subtype selectivity, often with the highest affinity for the M2 receptor and the lowest for M5 (10)
. Signaling of the M2 receptor is promiscuous in that G proteins of the Gi/o and of the Gs type are activated in a virtually identical pattern by various orthosteric agonists irrespective of their individual structure (11)
.
Here we report the design and chemical synthesis of allosteric/orthosteric receptor activators and show that these bind in a truly allosteric/orthosteric mode and engender both subtype-selective receptor activation and signaling pathway specificity. For distinction from orthosteric agonists and allosteric agonists, we propose to designate these conceptually novel small-molecule tools as dualsteric agonists.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
S and [3H]N-methylscopolamine chloride ([3H]NMS) (specific activity of 81 Ci/mmol) were from PerkinElmer Life and Analytical Sciences (Homburg, Germany). Naphmethonium and W84 are commercially available through AXXORA Deutschland (Grünberg, Germany).
Chemical synthesis
The synthesis of hybrid 1 and hybrid 2 (H1 and H2) followed a pathway that has previously led to hybrid antagonists (12)
. A detailed description of the synthesis of H1 and H2 and of the new building block A2 is given in Supplemental Materials.
Mutagenesis and expression
The novel receptor mutant M2104Tyr
Ala was obtained by a procedure described previously (8
, 9)
, using the QuickChange site directed-mutagenesis kit (Stratagene, Amsterdam, The Netherlands). Successful mutation was verified by sequencing.
Cell culture, transient transfection, and membrane preparation
We used Chinese hamster ovary (CHO) cells that were stably transfected with the human M2 receptor gene (13)
or wild-type COS7 cells that were transiently transfected with wild-type or mutant M2 receptor genes as described previously (9)
. Receptor densities as determined by [3H]NMS-binding experiments amounted to 4–8 pmol/mg membrane protein in case of stably transfected cells and to 0.4–1.6 pmol/mg protein in case of transiently transfected cells.
[3H]NMS binding assays were carried out in a 5 mM Na,K,Pi-buffer, pH 7.4, at 23°C as described previously (8)
. Experiments with the mutant M2104TyrAla required a radioligand concentration of 1.0 nM instead of the normally used 0.2 nM to compensate for the diminished NMS affinity (Supplemental Table 2). Incubation times sufficient to attain binding equilibrium in the presence of allosteric agents were calculated as described previously (9)
. Dissociation binding assays were carried out as two-point [3H]NMS dissociation experiments (14)
.
[35S]GTPS binding experiments were carried out in a buffer (10 mM GDP, 10 mM HEPES, 10 mM MgCl2, and 100 mM NaCl, pH 7.4, 30°C) as described previously (15)
.
Dynamic mass redistribution measurements were carried out using the beta version of the Epic device (Corning, NY, USA). Flp-In-CHO cells, Hanks’ balanced salt solution (HBSS), HEPES buffer solution, and pertussis toxin (PTX) were obtained from Invitrogen (Carlsbad, CA, USA). Flp-In-CHO cells, empty or stably transfected with the hM2 gene, were seeded in a density of 12,500 cells per well in 384-well Epic microplates with 40 µl of growth medium (Ham’s F-12 medium, 10% FBS, 2 mM L-glutamine, and 1% penicillin/streptomycin) and cultured at 37°C in an atmosphere of 5% CO2 for 
20 h with or without PTX (100 ng/ml) to achieve confluent cell layers. After cell culture medium was removed and the cells twice were washed with 50 µl of assay buffer per well (HBSS with 20 mM HEPES, pH 7.0), cells were allowed to rest in 30 µl of assay buffer for 2 h in the Epic reader at a constant temperature of 28°C. After addition of 10 µl of test compound dissolved in assay buffer, dynamic mass redistribution (DMR) responses were monitored for 1 h.
Studies with isolated organs were performed on tissues excised from male guinea pigs and New Zealand White rabbits as described previously (16)
. The investigation conformed to the Rule for the Care and Use of Laboratory Animals of the European Community and is in accordance with Italian law (DL 116/92).
Molecular modeling
Three-dimensional modeling, docking, and molecular dynamics simulations were essentially performed as described recently (15)
. In this model of the inactive M2 receptor, the passage between the allosteric and the orthosteric site is closed in the core region of the allosteric site. As shown recently (15)
, this region rearranges during receptor activation. To open a channel between the orthosteric and the allosteric site and to check for hybrid receptor binding, the dihedral angles of the atoms C-CA-CB-CG of M2403Tyr and of M2426Tyr were rotated manually 54 and 17 degrees. The structure was then minimized by using the steepest descent algorithm. Afterward, it was embedded in a membrane environment and parameterized for the GROMACS ffgmx force field (17)
. The structure showed satisfying quality and stability as checked by a molecular dynamics simulation of the unoccupied receptor carried out over 5000 ps.
Data analysis
[3H]NMS competition binding experiments were analyzed using a four-parameter logistic function yielding the IC50 and the slope factor n of the curve. If the observed slope factors did not differ significantly from unity (F test, >0.05) and in case of curve simulations, n was constrained to 1.
Irrespective of the slope factor IC50 values were converted to apparent binding constants K using the Cheng-Prusoff correction.
Unless indicated otherwise, [3H]NMS equilibrium binding in the presence of allosteric ligands was analyzed according to the common allosteric ternary complex model (18
, 19)
using Eq. 2 from Tränkle et al. (20)
that includes a slope factor. In the case of almost neutral cooperativity (
1) [3H]NMS equilibrium binding remains nearly unchanged under the influence of increasing concentrations of the allosteric modulator and curve fitting with Eq. 2 from Tränkle et al. (20)
does not work. Under this condition, a procedure described by Raasch et al. (21)
was applied that considers test compound binding affinity for radioligand occupied receptors as derived from dissociation binding experiments.
Hybrid compound effects on [3H]NMS-binding were simulated and analyzed according to the extended allosteric ternary complex model (22)
by means of the following equation:
 |
* is the cooperativity factor for the allosteric interaction between A and B. For sake of consistency throughout the manuscript, * from the equation above was converted to by = 1/*. Note that this model does not contain the slope factor as a variable.
Parallel curve shift analysis (modified Schild analysis) was performed globally as described previously (Eq. 4 in ref. 23
).
Data from dissociation binding experiments were analyzed as described previously (8)
. Nonlinear regression analysis was done using the software Prism 5.01 and Instat 3.0 (Graph Pad, San Diego, CA, USA). Values are means ± SE unless otherwise specified.
Shifts of [35S]GTPS binding curves in the presence of antagonists were analyzed according to Schild in case of competitive interactions and to Lanzafame et al. (24)
in case of allosteric interactions.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
; ref. 22
). The model differentiates between binding of an allosteric/orthosteric hybrid compound in the purely allosteric mode 4 and in the intended dual mode 5. Hybrid compounds (H1 and H2 in Fig. 1B
) were assumed to share an identical orthosteric agonist moiety (O) and differ in their allosteric building blocks (A1, A2). The orthosteric agonist will competitively inhibit radioligand binding. In the simulations, A1 was negatively cooperative with the radioligand but did not completely inhibit its binding, because ternary complexes can form that stabilize radioligand binding (mode 3 in Fig. 1A
). If the ternary complex is the predominant species, then orthosteric radioligand binding is elevated (A2, positive cooperativity). When a hybrid binds in the dual mode (mode 5), receptors will be drawn out of the quartet of states 1 to 4. Consequently, the hybrid will have a more pronounced inhibitive action on radioligand binding than its corresponding allosteric building block (Fig. 1B
; H1 vs. A1, H2 vs. A2).
|
For the design and synthesis of allosteric/orthosteric hybrid molecules, we chose bis(ammonio)alkane-type modulators to identify appropriate allosteric building blocks. Favorable properties of this class of muscarinic allosteric agents are high affinity for the allosteric site, pronounced M2 subtype selectivity, and a spaghetti-like flexible shape. Conformational flexibility should allow smooth adaptation to the unknown 3-D shape of the binding pocket of the active receptor in the allosteric/orthosteric interface region. Starting from the allosteric inhibitor W84 and the enhancer naphmethonium (ref. 9
; structures in Supplemental Fig. 1
), structure-activity-relationships of stepwise shortened derivatives yielded allosteric fragments (A1, A2 in Fig. 1C
) that confer similar M2 receptor subtype selectivity (relative to the low-affinity M5 receptor) as their parent molecules do. These allosteric building blocks were fused with the highly potent orthosteric receptor activator iperoxo (25
; see Supplemental Materials). M2-receptor binding of an orthosteric radioligand responded to the hybrids as predicted for a dual binding topography: stronger inhibition of radioligand binding by the hybrids H1 and H2 compared with the respective building blocks A1 and A2 (Fig. 1C
). The extended allosteric model yielded hybrid binding affinities for the dual mode (pKBdual H1: 7.73±0.04; H2: 8.53±0.02), which are more than one order of magnitude higher compared with the purely allosteric mode of the respective building blocks (pKBallo A1: 6.00±0.07; A2: 7.40±0.70).
Molecular validation of dualsteric binding
To confirm that the hybrids were using both orthosteric and allosteric sites, we mutated key residues in either of these regions of the M2 receptor and transiently expressed each mutant in COS7 cells. For the orthosteric site, we mutated M2104Tyr to Ala, located next to M2103Asp, which is pivotal for the attraction of the positively charged nitrogen contained in acetylcholine and other muscarinic orthosteric agonists and antagonists (26)
. For the allosteric site, we used the M2177Tyr and M2423Thr double mutant, which markedly reduces M2 modulator affinity toward that of the low affinity M5 receptor subtype (8)
.
On mutation of the orthosteric site, we expected a loss of binding affinity of the hybrid orthosteric moiety, thus unmasking the actions of their allosteric components. In line with this, inhibition of radioligand binding by the hybrids at the M2104Tyr to Ala mutation was attenuated relative to wild-type M2 (Fig. 2A, B
; red vs. black curve). This is particularly evident for H2 (Fig. 2B
), where elevation of radioligand binding, a hallmark of the allosteric building block A2 (compare Fig. 1C
, right panel), is restored in the mutant receptor.
|
Conversely, mutation of the allosteric core region of the receptor (Fig. 2A, B
; blue vs. black curve) weakens the influence of the hybrid allosteric moiety, thereby promoting predominantly orthosteric inhibition of radioligand binding. This is clearly seen with H2, the weak inhibitory effect of which was converted into a nearly maximum suppression of radioligand binding.
Binding affinities were derived from the experimental data (compiled in Table 1
). Compared with the affinity for M2 wild type, the hybrids loose affinity both in the allosteric and the orthosteric receptor mutants. In contrast, a major loss of affinity is seen in case of the orthosteric agonists iperoxo and acetylcholine only in the orthosteric mutant and, in the case of the allosteric parent compounds W84 and naphmethonium, only in the allosteric receptor mutant.
|
The hybrids reveal M2/M5 selectivity (Fig. 2A, B
; Table 1
) amounting to 2 orders of magnitude as the allosteric parent compounds do, while the M2/M5-affinity ratios of iperoxo and of the physiological transmitter acetylcholine amount only to 1 log unit (Table 1)
. A similar extent of M2/M5 selectivity of the hybrids is found in receptors, the orthosteric site of which is blocked by the radioligand, thus allowing only for the allosteric binding mode. Also, in this mode, M2/M5 binding selectivity critically depends on the allosteric core epitopes M2177Tyr and M2423Thr (Supplemental Fig. 2).
In a 3-D model of the M2 receptor in the inactive state, the passage between the allosteric and the orthosteric site is closed, but recent findings (15)
suggest that rearrangement of this region may occur during receptor activation. When a channel between the orthosteric and the allosteric site is opened by rotating M2403Tyr and M2426Tyr manually, receptor geometry allows for dualsteric binding. The iperoxo moiety is located in the orthosteric site with a cation-
-interaction being possible between iperoxo’s quaternary nitrogen and M2104Tyr (Fig. 2C
). The second quaternary nitrogen of the hybrids is involved in cation- interactions with the allosteric epitopes M2177Tyr and M2422Trp. The phthalimido or naphthalimido moiety, respectively, is located in the adjacent areas of the allosteric binding site. Docking simulations also yielded some purely allosteric complex geometries, which, however, are less probable than the dualsteric binding mode (data not shown).
Dualsteric agonists activate G proteins via the orthosteric site
Having confirmed a true dualsteric binding mode for the hybrids, we next measured M2-receptor-mediated G protein activation via monitoring the binding of [35S]GTPS in membranes from CHO cells stably expressing the human M2 receptor. On their own, the hybrids achieved levels of maximum G protein activation that come close to those of the orthosteric agonists iperoxo and acetylcholine (Fig. 3A
). The potency of the hybrids (concentration for half-maximum G protein activation) is reduced relative to iperoxo but is still clearly higher than the potency of acetylcholine.
|
Atropine is an orthosteric competitive antagonist that would induce a parallel rightward shift of the concentration-effect curve of an orthosteric agonist. With the hybrids, atropine behaved differently (Fig. 3B
). First, atropine diminishes maximum receptor activation by H1; to the best of our knowledge, such a type of antagonism by atropine has not been reported before. Second, the rightward shift of the hybrid curves by atropine approaches an upper limit that is indicative of negative binding cooperativity. The characteristics of atropine interaction with H1 correspond with a mixed inhibition in enzymology. The atropine pIC50 (6.05±0.06) for suppression of receptor activation matches the atropine binding constant at hybrid-occupied receptors (6.25, calculated as the sum of pKA atropine=8.44±0.04 and patropine/hybrid1=–2.19±0.14). Obviously, both effects result from the same underlying event, namely atropine binding to the orthosteric site that displaces the hybrid into the purely allosteric binding mode. We conclude that receptor activation by the hybrids is caused in the dualsteric mode through the orthosteric site.
When the hybrid allosteric parent compound W84 was used as an antagonist (Fig. 3C
), there was no significant reduction of the maximum effect of H1. This observation and the pattern of W84-induced rightward curve shift are compatible with a competitive interplay (Schild slope not different from unity, F test, P>0.05). Thus, binding of W84 to the allosteric site (pKB= 6.70±0.07) completely hinders the hybrid from attaching to the receptor and vice versa.
Dualsteric agonists induce pathway-specific signaling
Having shown receptor activation by the dualsteric agonists, we probed receptor-induced signaling using a real-time biophysical whole cell assay. Signaling-induced modification of cell function involves relocation of cellular constituents (27)
. The resulting local change of optical density affects cellular light diffraction properties. This was detected by means of the Epic system in cells grown to confluence in microplates with resonant waveguide grating optical biosensors in each well (28)
. Cells are illuminated from the underside with a light penetration depth of 150 nm. DMR is recorded as a shift of wavelength of reflected light. Notably, DMR signatures are dependent on and specific for the signaling pathways engaged by a given receptor (29)
.
The conventional orthosteric agonist-bound M2 receptor stimulates the Gi/o pathway and additionally the Gs pathway irrespective of differences in agonist structure (11)
. CHO cells expressing the hM2 receptor responded to receptor saturating agonist concentrations with signatures of similar shape independent of whether orthosteric or dualsteric agonists were applied (Fig. 4A, C
). Cells that had been pretreated with the irreversible Gi/o inhibitor PTX, however, no longer responded to the dualsteric agonists (Fig. 4B, D
). In contrast, PTX-pretreated cells were still reactive to iperoxo and acetylcholine, indicating that the orthosteric agonists activate additional PTX-insensitive signaling pathways. The downward deflected DMR signal is compatible with a Gs signature (29)
. Control experiments in CHO cells having been transfected with the empty Flp-In vector showed no response to any of the applied agonists (data not shown). Taken together, the findings demonstrate signaling specificity for dualsteric but not for orthosteric agonists.
|
Validation of selective agonism engendered in native tissues
To confirm that our novel mode of engendering selective agonism is valid in a native tissue context, we investigated hybrid actions in the electrically stimulated guinea pig left atrium (M2 model), the rabbit vas deferens preparation (M1 model), and the guinea pig ileum (M3 model). Compared with iperoxo, both hybrids induced nearly the same maximum effect in all three models (Fig. 5A
). In contrast to the nonselective iperoxo, however, the potency of the hybrid depended on the tissue preparation, with a preference for cardiac M2 receptors that was especially pronounced with H1 (Fig. 5B
). If the cellular signaling cascade augments the stimulus resulting from receptor-mediated G protein activation, stimulation of a part of the receptor population may be sufficient for a maximum effect. Under this condition, concentration-effect curves do not reflect receptor binding curves. We reduced the number of functional receptors by pretreating the preparations with the alkylating agent phenoxybenzamine to eliminate receptor reserve and to estimate agonist binding affinity (pKD in Fig. 5C
). The rank order of affinity was M2 > M1 > M3 for both hybrids. The level of affinities is higher for H2 compared with H1. Taken together, dualsteric agonism yields M2 selectivity of action in living native tissues.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
) and from receptor activation by allosteric agonists (31
, 32)
. Such compounds are also subtype selective but interact solely with the allosteric site. Until now, however, the highest binding affinities, but not selectivity, have been achieved with exclusively orthosteric ligands. The dualsteric ligands point out the orthosteric site as an affinity providing epitope. Here we show that the dualsteric strategy in fact allows linkage of subtype selectivity with high affinity and high efficacy, as both hybrids have a higher affinity and a comparable maximum effect, when compared with acetylcholine. Notably, signaling specificity is introduced by dualsteric receptor binding in addition to binding selectivity.
One might expect that the combination of two different moieties that target a common receptor simultaneously would yield an additive gain of affinity. That this was not observed in our study is not unexpected because the building blocks on their own stabilize functionally opposite receptor conformations. The orthosteric full agonist favors an active conformation, whereas the allosteric bis(ammonio)alkane-type agent behaves as an inverse agonist stabilizing an inactive conformation (33)
. In line with this, we previously found a pronounced mutual inhibition of receptor binding (negative binding cooperativity) between orthosteric agonists and bis(ammonio)alkane-type compounds (15)
. Future studies will have to show whether a gain of dualsteric agonist binding affinity can be achieved by using allosteric building blocks that are tailor-made for orthosteric agonist-bound receptors. In any case, dualsteric compounds containing allosteric building blocks that are highly negative cooperative with orthosteric agonists are unlikely to form ternary complexes (mode 3 in Fig. 1A
) with receptors being occupied by the endogenous acetylcholine (in contrast to the radioantagonist used in our study).
Rational design of dualsteric agonists will additionally have to address that allosteric agents may impair receptor activation by orthosteric agonists ("negative activation cooperativity"; ref. 15
). Fusion of some oxotremorine-like full orthosteric agonists (other than iperoxo used here) with allosteric fragments resulted in antagonist-like M2-receptor ligands (12)
. In the case of the weak partial M2-receptor activator McN-A-343, which was formerly classified as an allosteric agonist, there is recent evidence (34)
that the compound may partially interact with the orthosteric site, whereas the allosteric moiety counteracts orthosteric receptor activation. Taken together, the dualsteric agonists introduced here show for the first time that allosteric and orthosteric interactions including high efficacy receptor activation can favorably be linked in one ligand molecule.
Most interestingly, the hybrid-bound receptor conformation appears to be unique in that it directs signaling in a selective fashion, whereas signaling is promiscuous with the common orthosteric agonist-bound conformation.
Meanwhile an impressive body of evidence shows that GPCR activation may result in the activation of multiple signaling pathways (e.g., ref. 3
). Agonist-induced GPCR signaling, irrespective of whether being G protein dependent or independent, is mediated by the cytosolic face of the receptor. The interdependence between active receptor conformations and signaling pathway selectivity are far from being understood. The dualsteric agonists provide new insight into this important aspect of GPCR structural biology. Signaling specificity of dualsteric activators depends on the allosteric building block, as the orthosteric parent compound iperoxo is not signaling specific. The allosteric building block interacts with the extracellular region of the receptor protein. Thus our findings disclose that the extracellular region of a rhodopsin like GPCR has critical influence on ligand-induced signal trafficking.
In conclusion, dualsteric GPCR targeting enabled us to combine high affinity and subtype-selective receptor binding with signaling-specific receptor activation. Dualsteric agonists represent a novel type of small molecules with considerable potential to better understand GPCR structural biology and to improve the specificity of therapeutic GPCR activation.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Received for publication June 26, 2008. Accepted for publication September 11, 2008.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  C. Valant, P. M. Sexton, and A. Christopoulos Orthosteric/Allosteric Bitopic Ligands: Going Hybrid at GPCRs Mol. Interv., June 1, 2009; 9(3): 125 - 135. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
