3-Fluoro-N-methyl-d-aspartic acid (3F-NMDA) Stereoisomers as Conformational Probes for Exploring Agonist Binding at NMDA Receptors

Poh Wai Chia,[a] Matthew R. Livesey,[b] Alexandra M. Z. Slawin,[a] Tanja van Mourik,[a] David J. A. Wyllie,[b] and David O’Hagan*[a]

Abstract: N-Methyl-d-aspartate (NMDA) is the prototypical agonist of the NMDA receptor subtype of iono- tropic glutamate receptors. Stereogenic placement of a C—F bond at the 3-position of (S)-NMDA generates either the (2S,3S)- or (2S,3R)- diastereoisomers of 3F-NMDA. The individual diastereo- isomers were prepared by synthesis in enantiomerically pure forms and it was found that (2S,3S)-3F-NMDA is an ag- onist with a comparable potency to whereas the (2S,3R)-3F-NMDA forces these bonds anti, losing electrostatic stabilisation, to achieve the required binding conformation. These observa- tions illustrate the utility of stereoselec- tive fluorination in influencing the mo- lecular conformation of b-fluorinated amino acids and thus probing the active conformations of bioactive com- pounds at receptors.

Introduction

Glutamate (1) is the principal excitatory neurotransmitter of the mammalian central nervous system where it acts on sev- eral subtypes of ligand-gated ion channels. N-Methyl-d-as- partate (NMDA) (2) is a synthetic analogue of aspartate[1] and is a prototypical agonist and widely used pharmacologi- cal tool that defines one of these subclasses. N-Methyl-d-as- partate receptors (NMDARs) mediate the slow component of glutamatergic synaptic currents but play key roles in brain development, learning and memory and are implicated in a range of neuro-developmental/degenerative diseases such as schizophrenia, Alzheimer’s, Parkinson’s, Hunting- ton’s and ischemia.[2] NMDARs are tetrameric proteins, the majority of which contain two GluN1 and two GluN2 NMDAR subunits. Four subtypes of the GluN2 subunit exist (A–D) and it is these that confer the majority of the sub- type-dependent pharmacological and biophysical properties of NMDARs.[2–4] Activation of NMDARs requires glycine (or d-serine) binding at the GluN1 subunit and glutamate binding at the GluN2 subunit.

Structural studies of the NMDAR by Furukawa et al. have resulted in co-crystals of glutamate (1) and most re- cently with NMDA (2) bound in the ligand binding domain of the GluN2D NMDAR subunit.[7,8] This has revealed a binding mode in which the carboxylate groups of NMDA
(2) arrange gauche to each other. The current study set out to explore the relative efficacies of the two 3-fluoro NMDA diastereoisomers (2S,3S)-3 and (2S,3R)-4 as glutamate agonists and as analogues of NMDA (2) and assays have been carried out on GluN2A and GluN2B NMDARs. Fluorine is the smallest atom after hydrogen that can form a covalent bond to carbon. It is also the most polar bond in organic chemistry due to the electronegativity of fluorine, the high- est on the Pauling scale.[12,13] Thus, fluorine can replace hy- drogen, with minimum steric impact, but the dipole of the
C—F bond has a significant electronic influence.The stereospecific incorporation of the C—F bond to gen- erate (2S,3S)-3 and (2S,3R)-4 is anticipated to introduce a conformational bias in solution due to a stabilising fluoro-g-aminobutyric acid (3F-GABA) (6) to probe the chirality of g-aminobutyric acid (GABA) (5) binding to both GABAA and GABAC receptors.[14,15,16] In those studies we observed significantly different efficacies for the two enantiomers (R)-6 and (S)-6 suggesting that the GABAA and GABAC receptors bind g-aminobutyric acid in different conformations. The conformational bias was attributed to the electrostatic charge–dipole interaction between the C—F and C—N+ bonds as illustrated in Figure 1, thus the different enantiomers of 3F-GABA can adopt preferred twisted con- formations of the opposite chiral sense, and a favoured bind- ing mode was deduced by comparing the efficacy of the two enantiomers of 3F-GABA (6). In this research programme the C—F bond is incorporated by synthesis into the two diastereotopic locations at C-3 of NMDA (2) to generate dia- stereoisomers (2S,3S)-3F-NMDA (3) and (2S,3R)-3F- NMDA (4). These diastereoisomers were then assayed in a glutamate receptor assay to compare their relative effica- cies as agonists.

Scheme 1. Synthesis of erythro (2S,3S)-3F-NMDA HCl (3).

Figure 1. GABA (5) and 3-F-GABA enantiomers (6). For enantiomers (6) their solution conformation is influenced by a charge dipole interac- tion indicated by the dashed bonds.

Results and Discussion

Synthesis and analysis of 3F-NMDA diastereoisomers: The (2S,3S)-3F-NMDA stereoisomer (3) was synthesised as illus- trated in Scheme 1. The synthesis started from (S,S)-epoxy succinic ester (7), which was prepared from diethyl ( ) d- tartrate as previously described.[17] This then involved ring opening of epoxide 7 with N-benzylmethylamine, followed by treatment of the resultant amino alcohol 8 with [bis(2- methoxyethyl)amino]sulfur trifluoride (Deoxo-Fluor) to generate a single diastereoisomer of the dehydroxyfluorinat- ed product 10. This fluorination reaction occurs with com- plete stereoselectivity, as evinced by production of a single diastereoisomer, and with an overall retention of configura- tion, proceeding by double inversion after fluoride ion attack of an aziridinium intermediate 9.[14,18,19] With the C—F bond now located, N-debenzylation by hydrogenation result- ed in diester 11 and then acidic hydrolysis gave (2S,3S)-3F- NMDA (3) as a single diastereoisomer (see the Supporting Information). The enantiomeric purity of (2S,3S)-3F-NMDA (3) is reasonably assumed based on its origin from tartarate 7 and the stereointegrity of the epoxide ring opening and fluorination steps, each of which gave rise to single diaste- reoisomers.

The relative stereochemistry required for the synthesis of (2S,3R)-3F-NMDA (4) necessitated that the route start from the achiral meso-epoxide 12.[20] Thus a resolution, to access the individual enantiomers, is required. It was envisaged that ring opening of meso-epoxide 12 with either (S)-N or (R)-N-a-dimethylbenzylamine (13) would generate a set of diastereoisomers that could be separated into single enantio- mers by chromatography. In the reaction, epoxide ring opening with the enantiomerically pure amines was relative- ly straightforward, however the diastereoisomeric products 14 and 15 could not be easily separated by chromatography. Instead the product mixture was treated with Deoxo-Fluor in a dehydroxyfluorination reaction, which following prece- dent,[14,18,19] is anticipated to proceed with an overall reten- tion of configuration. This gave a product mixture of diaste- reoisomers 16 or 17. The fluorohydrins could now be sepa- rated by careful chromatography. For practical purposes each enantiomer was secured by reaction of either (S)-N or (R)-N-a-dimethylbenzylamine (13), since the slower eluting diastereoisomer on chromatography could be cleanly sepa- rated, whereas the faster eluting diastereoisomer was always contaminated with the slower eluting isomer. Only the route using (S)-13 is shown in Scheme 2. Each stereoisomer, 16 or 17, which was enantiomerically pure as a consequence of Comparison of the 1H and 19F NMR spectra of the diaste- reoisomer (2S,3S)-3F-NMDA (3) prepared above (Scheme 1), showed very clear differences in chemical shifts and coupling constants consistent with their diastereoiso- meric relationship (see the Supporting Information). The NMR analysis provides some insight into the solution con- formations of both 3F-NMDAs (3) and (2S,3R)-3F-NMDA (4). Both the 3JHH the 3JHF coupling constants have an angu- lar dependence. Similar to the known Karplus relationship for vicinal hydrogen couplings, syn- and antiperiplanar vici- nal H-C-C-F relationships are typically about 30 Hz, where- as gauche relationships are around 10–15 Hz.[23,24,25] For isomer (2S,3S)-3F-NMDA (3) the 3JHH value is small at about 2.0 Hz suggesting vicinal C—H bonds a little larger than 608 and no significant contribution from conformer B, in which the vicinal hydrogen atoms are anti. The 3JHF cou- pling constant of 29 Hz suggests either an anti or syn H to F relationship (conformer A syn or conformer C anti).

Figure 3. Calculated minimum energy DFT structures for diastereoiso- mers 3 (left) and 4 (right). Each of the DFT calculated structures is con- sistent with the solution conformations (boxed Newman projections Figure 4).

Figure 4. The favoured solution conformers of 3F-NMDA isomers (2S,3S)-3F 3 and (2S,3R)-3F 4 as deduced by 1H and 19F NMR analysis are highlighted in the boxed Newman projections.

Figure 5. Agonist activities of the NMDA (2), (2S,3S)-3, (2S,3R)-4 and (2R,3S)-4 stereoisomers at NMDARs. Example TEVC currents recorded from an oocyte expressing GluN2A NMDARs. Note that both NMDA (2) (a) and (2S,3S)-3 (b) each elicit large inward currents. However (+)-(2S,3R)-4 or (2R,3S)-4 (c) and ( )-(2R,3S)-4 or (2S,3R)-4 (d) evoke only modest currents which are only resolvable on a higher gain setting (light grey traces).

Figure 6. a) Summary of the relative efficacies for NMDA (2), (2S,3S)-3, (2S,3R)-4 and (2R,3S)-4. Currents have been normalized to the response
produced by NMDA (2). b) Same data as illustrated in (a) but on an ex- panded scale to indicate the small responses evoked by(+)-(2S,3R)-4 or (2R,3S)-4 and (—)-(2R,3S)-4 or (2S,3R)-4.

GluN2A and GluN2B NMDARs the concentrations with which (2S,3S)-3 and NMDA (2) produce their respective half maximal responses are similar. Isomer (2S,3S)-3 can thus be considered to act as a partial agonist at both GluN2A and GluN2B NMDARs. These findings suggest that (2S,3S)-3 is the sole stereoisomer that can access a bind- ing conformation similar to NMDA (2) to act as an agonist at GluN2A and GluN2B NMDARs.The data also indicated that neither of the enantiomers (2S,3R)-4 and (2R,3S)-4 are effective agonists at the concentrations tested. These observations are consistent with the solution conformation of 3 (conformer A in Figure 4), most closely representing the receptor binding conformation. Ste- reoisomers (2S,3R)-4 and (2R,3S)-4 would have to adopt a higher energy conformation in which the C—F bond is ori- ented anti to the C—N+ bond, losing electrostatic stabilisa-
tion. These observations are consistent with the recent X-ray crystallography study by Furukawa et al., in which NMDA (2) was bound in a co-crystal with the GluN2D pro- tein as illustrated in Figure 7, although the assays were car- ried out, in our study, on GluN2A and GluN2B and not GluN2D NMDARs. There is however high homology be- tween these receptors, with all binding residues conserved, suggesting similar binding modes.

Figure 7. The X-ray structure of Furukawa et al., showing NMDA (2) bound to the GluN2D subunit, and a flattened distorted view of key hy- drogen bonding interactions. The lower structures illustrate the bound conformation of NMDA (2) as a Newman projection and then as it is ori- entated on the receptor and the relationship between the C—F bond and the C—N+ bonds in the stereoisomers of (2S,3S)-3 and (2S,3R)-4. Isomer (2S,3S)-3 bound in this conformation would have a stabilising charge– dipole gauche relationship between the C—F and C—N+ bonds, whereas as for (2S,3R)-4 there is no such stabilising interaction as these bonds are antiperiplanar to each other.

Conclusion

An electrostatic effect of a fluorine atom, stereospecifically located at the C-3 position of the glutamate receptor agonist NMDA has been used to influence binding efficiency to the glutamate receptor. Stereogenic placement of the C—F bond to generate (2S,3S)-3 and (2S,3R)-4 3F-NMDAs gave rise to very different levels of agonist activity on glutamate recep- tors. Indeed, (2S,3S)-3 showed similar activity to NMDA (2), whereas (2S,3R)-4 was almost inactive. NMR, X-ray structural analysis and DFT theory analyses was used to assess the favoured conformations of these diastereoisomers. For (2S,3S)-3 the favoured conformer in solution appears to be similar to that required for binding to the receptor. On the other hand the (2S,3R)-4 isomer would have to orientate the C—F and C—N+ bonds antiperiplanar in a less favoured orientation, to adopt the same agonist binding mode, thus as
predicted this isomer is a very poor agonist. The deduced binding mode of NMDA to GluN2A and GluN2B subunits gets some support from a recent structural biology study on the GluN2D receptor subunit. A referee raised the reasona- ble possibility that the C—F bond in 4 might interact in a re- pulsive manner with the surface of the agonist binding site, and that the poorer agonist activity could be due to that. This cannot be excluded per se from our data, although careful examination of the co-crystal structure of NMDA and the GluN2D NMDAR subunit does not reveal a candi- date residue for such a repulsive interaction if (4) is bound into that structure.

Acknowledgements

We are grateful to the University of Malaysia Terengganu and the Minis- try of Higher Education of Malaysia for a PhD Studentship and we thank Scott Badin and Kin Chow for performing initial experiments with 3F-NMDA isomers. We also thank EaStCHEM via the EaStCHEM Re- search Computing Facility and DO’H acknowledges an ERC Advanced Investigator grant.

[1] J. C. Watkins, J. Med. Pharm. Chem. 1962, 5, 1187 – 1199.
[2] S. F. Traynelis, L. P. Wollmuth, C. J. McBain, F. S. Menniti, K. M. Vance, K. K. Ogden, K. B. Hansen, H. Yuan, S. J. Myers, R. Dingle- dine, Pharmacol. Rev. 2010, 62, 405 – 496.
[3] K. Erreger, P. E. Chen, D. J. A. Wyllie, S. F. Traynelis, Crit. Rev. Neurobiol. 2004, 16, 187 –224.
[4] P. E. Chen, D. J. A. Wyllie, Br. J. Pharmacol. 2006, 147, 839 –853.
[5] N. Armstrong, E. Gouaux, Neuron 2000, 28, 165 – 181.
[6] S.. Cull-Candy, S. Brickley, M. Farrant, Curr. Opin. Neurobiol. 2001,11, 327 –335.
[7] H. Furukawa, S. K. Singh, R. Mancusso, E. Gouaux, Nature 2005,438, 185 –192.
[8] K. M. Vance, N. Simorowski, S. F. Traynelis, H. Furukawa, Nat. Commun. 2011, 2, 294.
[9] J. P. Snyder, N. S. Chandrakumar, S. N. Rao, D. P. Spangler, D. C. Lankin, J. Am. Chem. Soc. 1993, 115, 3356 – 3387.
[10] C. R. Briggs, M. J. Allen, D. O’Hagan, A. M. Z. Slawin, A. E. Goeta, J. A. K. Havard, Org. Biomol. Chem. 2004, 2, 732 –740.
[11] N. E. J. Gooseman, D. O’Hagan, M. J. G. Peach, A. M. Z. Slawin, D. J. Tozer, R. J. Young, Angew. Chem. 2007, 119, 6008 –6012.
[12] L. Hunter, Beilstein J. Org. Chem. 2010, 6, 38. [13] D. O’Hagan, Chem. Soc. Rev. 2008, 37, 308 – 319.
[14] G. P. Deniau, A. M. Z. Slawi, T. Lebl, F. Chorki, J. P. Issberner, T. van Mourik, J. M. Heygate, J. J. Lambert, L. A. Etherington, K. T. Sillar, D. O’Hagan, ChemBioChem 2007, 8, 2265 –2274.
[15] I. Yamamoto, G. P. Deniau, N. Gavande, M. Chebib, G. A. R. John- ston, D. O’Hagan, Chem. Commun. 2011, 47, 7956 – 7958.
[16] D. O’Hagan, Future Med. Chem. 2011, 3, 139 – 140.
[17] A. Korn, S. R. Böhner, L. Moroder, Tetrahedron 1994, 50, 8381 – 8392.
[18] F. B. Charvillon, R. Amouroux, Tetrahedron Lett. 1996, 37, 5103 – 5106.
[19] J. de Villiers, L. Koekemoer, E. Strauss, Chem. Eur. J. 2010, 16, 10030 – 10041.
[20] R. Cysewski, M. Kwit, B. Warzajtis, U. Rychlewska, J. Gawron´ski, J. Org. Chem. 2009, 74, 4573 –4583.
[21] Y. Zhao, D. G. Truhlar, Theor. Chem. Acc. 2008, 120, 215 – 241.
[22] E. Cancès, B. Mennucci, J. Math. Chem. 1998, 23, 309 – 326.
[23] D. O’Hagan, H. S. Rzepa, M. Schüler, A. M. Z. Slawin, Beilstein J. Org. Chem. 2006, 2, 19.
[24] M. Schüler, D. O’Hagan, A. M. Z. Slawin, Chem. Commun. 2005, 4324 – 4326.
[25] P. E. Chen, M. T. Geballe, P. J. Stansfeld, A. R. Johnston, H. Yuan,
A. L. Jacob, J. P. Snyder, S. F. Traynelis, D. J. A. Wyllie, Mol. Phar- macol. 2005, 67, 1470 – 1484.
[26] D. J. A. Wyllie, A. R. Johnston, D. L. Lipscombe, P. E. Chen, J. Physiol. 2006, 574, 477 –489.