Synthesis of 3,4-dihydro-2H-1,2-benzothiazine-3-carboxylic acid 1,1-dioxides and their evaluation as ligands for NMDA receptor glycine binding site.

Abstract A series of 2-substituted 3,4-dihydro-2H-1,2-benzothiazine-3-carboxylic acid 1,1-dioxides were synthesized and evaluated for their affinity to the glycine binding site of the N-methyl-d-aspartate (NMDA) receptor. The binding affinity was determined by the displacement of radioligand [3H]MDL-105,519 from rat cortical membrane preparations. The most attractive structures in the search for prospective NMDA receptor ligands were identified to be 2-arylcarbonylmethyl substituted 3,4-dihydro-2H-1,2-benzothiazine-3-carboxylic acid 1,1-dioxides. It has been demonstrated for the first time that the replacement of NH group in the ligand by sp3 CH2 is tolerated. This finding may pave the way for previously unexplored approaches for designing new ligands of the NMDA receptor.


Introduction
The N-methyl-D-aspartate (NMDA) receptor is a member of the glutamate receptor superfamily and plays an important role in neurotransmission within the central nervous system (CNS), as well as in the synaptic plasticity that underlies learning processes and memory 1,2 . Dysfunction of this receptor is thought to be involved in various disorders including epilepsy, ischemic brain damage and neurodegenerative disorders, such as Parkinson's and Alzheimer's diseases 3,4 . Overstimulation of the NMDA receptor caused by excessive glutamate in the synaptic region is believed to contribute to the symptomatology of these disorders. The NMDA receptor, being organized as a ligand-gated ion channel, is activated by glutamate in the presence of a co-agonist -glycine or D-serine. The presence of a co-agonist, which binds to a specific strychnine-insensitive glycine binding site (glycine B) of the NMDA receptor, is necessary for ion channel activity 5 . Therefore, the glycine binding site was perceived as a unique pharmacological target and several highly potent glycine binding site antagonists have been discovered [6][7][8] . As, unsurprisingly, glycine B antagonist pharmacophore 9 does not look like a brain penetrant, this was one of the reasons for low efficacy in vivo. The NMDA receptors are present not only in the CNS but also in the periphery and there is growing interest in the potential use of glycine antagonist as treatment for different neuropathic pain conditions. An example of peripheral activity of a brain impermeant glycine selective antagonist is MRZ 2/596 (V) 4 , demonstrating efficacy in several pain models 10 . Successful optimisation of peripherally restricted NMDAR glycine binding site antagonists, based on the quinoxaline-2,3-dione structure VI, has already been reported 11 . As it targets peripheral NMDA receptors, the glycine antagonist pharmacophore model is much more promising for the optimisation of bioavailability of active compounds.
To search for new starting points in the design of effective glycine B antagonists based on structure II 12 , we focused on information from the early studies of 5,7-dichlorokynurenic acid (I, DCKA) 13 . It has been shown that 4-O-substitution dramatically reduced the activity of the compounds. The carbonyl group in structure I ensured the interaction with the receptor via water molecule hydrogen bonding 14 . Alternatively, the activity was improved by the introduction of acidic functional group. This occurred in our series of naphthalenes II, where the most active compounds contained a carboxyl group in the substituent R ( Figure 1).
Based on the speculations mentioned above, we expected that the H-bond acceptor in the position 4 of naphthalene II would allow to reduce the polarity of compounds by removing the carboxyl group, while not affecting the affinity for the NMDA receptor. The necessary H-bond acceptor can be included in structure II by using carboxamide or sulfonamide group. In both cases, the structure may be derivatised on nitrogen atom. However, compounds containing carboxamide group in this position (4-oxo-quinazolidine-2carboxylic acids) were inactive in [ 3 H]-Gly displacement at 100 mM concentration 15 . Surprisingly, sulfonamides were highly active in corresponding benzo-1,2,4-thiadiazine-3-carboxylic acid 1,1-dioxide series 14 . The role of the sulfonamide group in this particular class of compounds has never been explained. However, combination of sulfonamide and carboxy functions has already been known to be of great importance in this type of structures 15,16 . Both modifications described above have been designed based on a common pharmacophore model of 5,7-dichlorokynurenic acid (I, DCKA) 13 , including NH as a presumable H-donor and specific substituents as in structure 1. The surprising loss of activity, when the sulfonyl group was replaced by carbonyl 15 , prompted us to evaluate carba-analogs III possessing the same sulfonamide fragment and carboxyl group as crucial pharmacophores. The efficiency of sp 2 CH, rather than NH, for the development of new GlyB ligands has already been proved 12,17 . Altogether, this information provided a highly promising background to identify novel potent and less polar ligands for the NMDA receptor glycine binding site, using structure III and, possibly structure IV.

Chemistry
Commercial grade reagents and anhydrous solvents were used as received. 1

Binding experiments
For the evaluation of binding affinity of the test compounds on the glycine binding pocket of the NMDA receptor, [ 3 H]-MDL-105,519 (GE Healthcare, Freiburg, Germany) displacement studies were performed using a 96-well plate robotic platform. MDL-105,519 is a selective, high affinity antagonist at the NMDA receptor glycine binding site.

Preparation of cortical membranes
Tissue preparation was performed according to Foster & Wong 18 with some modifications 19 . For isolation of the cell membranes, rat cortices were homogenized in 20 volumes of ice-cold 0.32 M sucrose (Sigma-Aldrich, Taufkirchen, Germany) using a glass-Teflon homogenizer. The homogenate was centrifuged at 1000g for 10 min, the pellet was discarded and the supernatant centrifuged at 20 000g for 20 min. The resulting pellet was resuspended in 20 volumes of distilled water and centrifuged for 20 min at 8000g. The supernatant and the buffy coat were then centrifuged three times (48 000g for 20 min) in the presence of 50 mM Tris-HCl, pH 8.0 (assay buffer). All centrifugation steps were carried out at 4 C. After re-suspension in 5 volumes of 50 mM Tris-HCl, pH 7.5, the membrane suspension was frozen rapidly at À80 C. On the day of assay, the membranes were thawed and washed four times by re-suspension in 50 mM Tris-HCl, pH 7.5 and centrifugation at 48 000g for 20 min. The final pellet was suspended in assay buffer. The amount of protein in the final membrane preparation was determined according to the method of Lowry et al. 20 with some modifications 21 . The final protein concentration used for our studies was 400 mg/mL.

[ 3 H]MDL-105,519 displacement studies
A robotic system designed for binding assays (Tecan Deutschland GmbH, Crailsheim, Germany) was loaded with the membrane solution, solutions for bound control (buffer/DMSO 20%), unlabeled glycine (1 mM) for evaluation of non-specific binding, all compounds to be tested (at 20-fold concentrations), radioligand and respective 96-well plates. Before performing displacement studies, saturation experiments were performed to determine the equilibrium dissociation constant (K d ) of [ 3 H]-MDL-105,519, which is a parameter for the affinity of the radioligand to the binding site. The protein/receptor concentration was held constant, whereas the amount of specific bound radioligand was determined using increasing concentrations of ligand. On the basis of the saturation experiments, a final [ 3 H]-MDL-105,519 concentration of 2 nM was selected. First, the assay plates were loaded with membrane solution and were shaken at 4 C. The mother plates were then prepared by pipetting the compounds into assay buffer/20% DMSO to obtain the desired final concentrations (dose response curve with five different concentrations, e.g. 10, 3, 1, 0.3 and 0.1 mM). After transferring radioligand into the assay plates, the compounds were added (including the bound and the non-specific binding control). The final DMSO concentration was 1%. The assay plates were incubated and shaken at 4 C for 1 h, before the mixture was exhausted as rapidly as possible via a vacuum manifold using the Multiscreen HTS glass fibre (type B) filter plates (Millipore, Schwalbach, Germany) under a constant vacuum of 450 mbar. The membranes were washed four times with cold assay buffer (100 mL). A 50 mL of Ultima Gold scintillation cocktail (PerkinElmer, Rodgau-Jügesheim, Germany) was added to the wet filter plates and incubated at room temperature overnight before counting the disintegration per minutes using a liquid scintillation counter (MicroBeta, PerkinElmer, Rodgau-Jügesheim, Germany).

Analysis of data
For the evaluation of binding affinity of the test compound to the glycine B binding site and its potency to displace [ 3 H]-MDL-105,519, the measured radioactivity of the radioligand alone is set as 100% bound control and the non-specific binding of the radioligand (which could not be displaced by glycine, 1 mM) represented the 0% control. The residual radioactivity after displacement of the test compound (n ¼ 5) is then corrected with respect to the set controls.

Chemistry
There are two main approaches for the synthesis of 2H-1,2benzothiazine-3-carboxylates 10-12. One of them is based on the diazotation of an appropriate aniline and subsequent Meerwein arylation of acrylonitrile 22 . Then sulfonation with chlorosulfonic acid gives the corresponding benzenesulfonyl chloride, which is cyclised to a derivative of 3,4-dihydro-2H-1,2-benzothiazine. First, we attempted the synthesis according to the procedure described above. Unfortunately, the Meerwein arylation of acrylonitrile with 3,5-dichlorobenzenediazonium chloride did not work well in our hands. All attempts to optimise the reaction conditions resulted in only a 9% isolated yield of corresponding 3-aryl-2-chloropropionitrile. Therefore, we focused on the second approach comprising a six-step synthesis from saccharin 23 . This involved chemical transformation of saccharin derivatives to sulfonamides 8 is shown in Scheme 1. Further on, the hydroxyl group of compound 8 was transformed to bromide which, in turn, was cyclised in the presence of a base to the corresponding 3,4dihydro-2H-1,2-benzothiazine. Finally, oxidation with NBS gave 2H-1,2-benzothiazine analogues of structure 10. Synthesis of necessary benzisothiazolones 5 was attempted in accordance with the previously described procedure 24 . As phenylsulfonyl chlorides 2a and 2b were not commercially available, they were prepared by chlorosulfonation of toluene derivatives 1a and 1b. In the next step, sulfonyl chlorides 2 were treated with t-butyl amine to generate corresponding sulfonamides 3. The N-tert-butyl group is compatible with the oxidative cyclisation of compounds 3 with H 5 IO 6 and CrO 3 in acetonitrile to obtain benzisothiazoles 4 in high yield 24 . The cleavage of N-tert-butyl group and subsequent N-protection with p-methoxybenzyl group gave benzisothiazoles 6.
The next step was crucial since the authors of this synthetic pathway for the reductive opening of the isothiazolone ring used lithium aluminum hydride, which gave the product only with 41% yield 24 . It appeared that the formation of byproducts could be overcome by using sodium borohydride in THF/water (1:1). Thus, benzenesulfonamides 7 were obtained in 94-98% yield. Alkylation of sulfonamides 7 with ethyl bromoacetate in the presence of base gave compounds 8. The reported three-step transformation of 2-(hydroxymethyl)-benzenesulfonamide to 2H-1,2-benzothiazines included the hydroxide conversion to bromide, ring closure, and finally the oxidation of the ring system 23 . To prepare the crucial intermediate 10 from compound 8c we used a two-step procedure as shown in Scheme 2. Oxidation of benzylic hydroxyl group with MnO 2 gave aldehyde 9, which was easily cyclised to benzothiazine 10 in the presence of base. The N-(pmethoxybenzyl) group of compound 17 was cleaved with trifluoroacetic acid to give benzothiazine 11 in good yield. Compound 11 was expected to be a versatile synthon for the parallel synthesis of N-substituted 2H-1,2-benzothiazine-3-carboxylic acids. Thus, alkylation of benzothiazine 11 with 3bromobenzylbromide gave the expected ester 12b in 95% yield. Finally, alkaline hydrolysis of esters 10 and 12 by NaOH or LiOH was used to afford the desired carboxylic acids 13.
Compounds 18 were synthesized according to the previously described approach 22 shown in Scheme 3. Benzyl alcohols 8 were treated with PBr 3 to give bromides 14, which were then cyclised to benzothiazines 15, using NaH for deprotonation. Cleavage of N-(p-methoxybenzyl) group with TFA gave 1,2-benzothiazines 16, which were used in N-alkylation reactions with different aryland heteroaryl-methyl halides. Then alkaline hydrolysis of ethyl esters 17 gave the desired carboxylic acids 18 in variable yields. The N-(p-methoxybenzyl) group containing compounds 18 was prepared by hydrolysis of the corresponding esters 15.

Binding results
The affinity of benzothiazine compounds 13a, b, c to the glycine binding site of the NMDA receptor was determined by the displacement of radioligand [ 3 H]MDL-105,519 25-27 from rat cortical membrane preparations. Surprisingly, all 2H-benzothiazine-3-carboxylic acids 13a, b, c were found completely inactive at a 10 mM concentration. Only compound 13b displayed weak affinity (K i ¼ 90 mM) for the glycine binding site.
The in vitro activity data of compounds 18 and 19 are shown in Table 1. Inhibition of radioligand binding at 10 mM concentration of substrate (I, %) is given for less potent compounds in cases when a correct dose-response curve has not been obtained. Roughly, figures close to 50% of radioligand displacement correspond to IC 50 $ 10 mM in Table 1. Our first synthesized 3,4-dihydrobenzothiazine 18b showed a considerable activity at 10 mM concentration, replacing ca 50% of radioligand. It should be noted that the reference compound of kynurenic acid with two halogen atoms in positions 6 and 8 (DCKA), showed higher activity than its 6-monochloro analogue 13 . In our case, 6,8dichlorobenzothiazine 18b and monochloro derivative 18a displayed comparable activities. Therefore, the study was continued mainly with 6-chlorosubstituted benzothiazines. An argument in favour of this decision was based on the dramatic solubility diminishing effect after the introduction of the second halogen in position 8. Surprisingly, 6,7-dichlorobenzothiazines were better soluble, allowing the screening to be performed at concentration of at least 10 mM. We also tested 6,7-dichloro-1,2-benzothiazine-3-carboxylic acids 18h and 18m and both showed significantly lower activity, displacing only 45% and 20% of radioligand, respectively at 10 mM concentration. It is important to note that even 2-unsubstituted benzothiazine 19 displays moderate affinity for Gly B with K i $ 10 mM. N-benzyl derivatives displayed higher affinity if the substituents in the phenyl ring were in the meta position (MeO, 18c; Cl, 18e and Br, 18f), regardless of the electronic properties of the substituent. Replacement of N-benzyl with N-phenethyl group resulted in lowered Gly B activity (18i, R 4 ¼ 3-MeOPh; K i ¼ 2.5 mM and 18j, R 4 ¼ 3-MeOBn; displacing only $40% of radioligand at 10 mM). However, in the case of unsubstituted N-benzyl and N-phenethyl derivatives (18g and 18i, respectively), this effect is not marked because of the reduced activity of both compounds. Actually, compound 18i is more active than 18g (62% and 41% displacement, respectively). The group of acetophenones (18n-18q) possessing additional hydrogen bond acceptor displayed significantly higher activity than N-phenethyl-benzothiazines. It is possible that the oxo group is   involved in additional hydrogen bonding with the receptor. In this series of compounds the highest Gly B affinity was observed for compounds with methoxy group in meta and ortho positions of the N-benzyl group (18n and 18o, K i ¼ 1.1 mM and K i ¼ 1.7 mM, respectively). Considering the above-mentioned higher activity of the phenone group containing benzothiazines, one could assume that N-substitution with heteroaryl group, which may act as a hydrogen bond acceptor, should also result in potentially active compounds. Nonetheless, N-heteroarylmethyl benzothiazines 18r, 18s and 18t were less active. Low activity of 2-pyridinyl derivative 18r may point to different hydrogen bonding vectors in comparison to the phenone group, if such an interaction occurs. To investigate if there is any functional importance of phenyl group, ester 18u was screened. Compound 18u was inactive, even though it possessed an oxygen functionality at the same distance from the benzothiazine core as corresponding acetophenone 18o.

Conclusions
In summary, a series of 2-substituted 3,4-dihydro-2H-1,2benzothiazine-3-carboxylic acid 1,1-dioxides were synthesized and evaluated for their affinity to the glycine binding site of the NMDA receptor. The most active compounds displayed micromolar level of activity. The most attractive structures in the search for prospective NMDA receptor ligands were identified to be 2arylcarbonylmethyl substituted 3,4-dihydro-2H-1,2-benzothiazine-3-carboxylic acid 1,1-dioxides. It has been demonstrated for the first time that the replacement of NH group in the ligand by sp 3 CH 2 is tolerated. This finding may pave the way for previously unexplored approaches for designing new ligands of the NMDA receptor.