Synthesis of New C-3 Substituted Kynurenic Acid Derivatives

The application of kynurenic acid (KYNA) as an electron-rich aromatic system in the modified Mannich reaction has been examined. The extension possibility of the reaction was tested by using amines occurring in a number of bioactive products, such as morpholine, piperidine, or N-methylpiperazine and aldehydes of markedly different reactivities, like formaldehyde and benzaldehyde. The influence of substituents attached to position 3 on the aminoalkylation was also investigated. Thus, reactions of 3-carbamoyl-substituted precursors with tertiary amine containing side-chains were also tested to afford new KYNA derivatives with two potential cationic centers. By means of NMR spectroscopic measurements, supported by DFT calculations, the dominant tautomer form of KYNA derivatives was also determined.

Among the important features of KYNA, one is that it is one of the few known endogenous excitatory amino acid receptor blockers with a broad spectrum of antagonistic properties in supraphysiological concentrations. One of its confirmed sites of action is the α-7-nicotinic acetylcholine (α-7-nACh) receptor and, interestingly, the other, identified recently, is a higher-affinity positive modulatory binding site at the α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor [5].
Formally, KYNA can be considered to be a nitrogen-containing 1-naphthol derivative. In our previous studies, 1-naphthol and its N-containing analogues were successfully applied in the modified Mannich reaction (mMr) [22] leading to the corresponding aminonaphthols [23], aminoquinolinols or aminoisoquinolinols [24,25]. A similar transformation starting from xanthurenic acid has been described by Schmitt et al. [26]. They managed to perform regioselective aminoalkylation at position 3 on this substrate, by using benzyl-protection of 8-hydroxyl group. Our present aim was to examine the possibility of aminoalkylation of KYNA derivatives at position 3 via a mMr approach. Since it is postulated that the reactivity of KYNA derivatives is influenced by tautomerism, we also envisaged identifying the tautomeric form dominant in solution. Finally, the reactivity of representative KYNA amides, carrying N-dialkylaminoalkyl-type side-chain possibly protonated under the standard reaction conditions, was also planned to be tested as these substrates with slightly decreased electron density at C-3 might display a decreased reactivity compared to that produced by a KYNA ester.
To explore the scope and limitation of the transformation, 1 was first reacted with N,N-dimethylethane-1,2-diamine in the presence of benzaldehyde. The desired amino acid derivative 3 was isolated in a yield of 74%. Next, starting from dimethylamine or N-benzylmethylamine in the mMr, the insertion of cationic centers in one-carbon distance at position 3 can be achieved. The structures of the formed amino acids 4a and 4b are depicted in Scheme 1. Formally, KYNA can be considered to be a nitrogen-containing 1-naphthol derivative. In our previous studies, 1-naphthol and its N-containing analogues were successfully applied in the modified Mannich reaction (mMr) [22] leading to the corresponding aminonaphthols [23], aminoquinolinols or aminoisoquinolinols [24,25]. A similar transformation starting from xanthurenic acid has been described by Schmitt et al. [26]. They managed to perform regioselective aminoalkylation at position 3 on this substrate, by using benzyl-protection of 8-hydroxyl group. Our present aim was to examine the possibility of aminoalkylation of KYNA derivatives at position 3 via a mMr approach. Since it is postulated that the reactivity of KYNA derivatives is influenced by tautomerism, we also envisaged identifying the tautomeric form dominant in solution. Finally, the reactivity of representative KYNA amides, carrying N-dialkylaminoalkyl-type side-chain possibly protonated under the standard reaction conditions, was also planned to be tested as these substrates with slightly decreased electron density at C-3 might display a decreased reactivity compared to that produced by a KYNA ester.
To test the effect of substituent at 5-, 6-, 7-or 8-positions morpholine as a representative secondary cyclic amine was selected. First, the initial KYNA analogues (7)(8)(9)(10) were synthesised with the Conrad-Limpach method applying the following optimization: a) the intermediate was purified by column chromatography using n-hexane:EtOAc as eluent and b) ring closure was carried out utilizing 1,2-dichlorobenzene instead of diphenyl ether used in the literature [8]. Reactions leading to the formation of 11-14 were performed starting from the corresponding ethyl ester and morpholine in the presence of formaldehyde (Scheme 2). It can be concluded that aryl/alkyl substituents at position 6 or 8, and the halogen at position 5 or 7 have no significant influence on the substitution at position 3. To test the effect of substituent at 5-, 6-, 7-or 8-positions morpholine as a representative secondary cyclic amine was selected. First, the initial KYNA analogues (7)(8)(9)(10) were synthesised with the Conrad-Limpach method applying the following optimization: a) the intermediate was purified by column chromatography using n-hexane:EtOAc as eluent and b) ring closure was carried out utilizing 1,2-dichlorobenzene instead of diphenyl ether used in the literature [8]. Reactions leading to the formation of 11-14 were performed starting from the corresponding ethyl ester and morpholine in the presence of formaldehyde (Scheme 2). It can be concluded that aryl/alkyl substituents at position 6 or 8, and the halogen at position 5 or 7 have no significant influence on the substitution at position 3. Since KYNA amides 16 and 17-containing cationic side-chain-proved to be the most effective analogues [27,28], our attention focused on the substitution of these compounds at position 3 via the mMr. In this reaction, different secondary amines such as pyrrolidine, piperidine, and morpholine were tested. Substrates N- (2- (16) and 4-oxo-N-(2-(pyrrolidin-1-yl)ethyl)-1,4-dihydroquinoline-2-carboxamide (17), synthesised according to a literature method [21], were aminoalkylated with 15a-c in the presence of formaldehyde, resulting in aminoalkylated KYNA derivatives 18a-c and 19a-c, respectively (Scheme 3). Summarizing these results and those of a previous study [29], the synthesis of this type of KYNA derivatives containing an amide moiety at position 2 and an aminoalkyl group at position 3 can be achieved by applying two different synthetic pathways: amidation followed by aminoalkylation (route A) and a reverse reaction sequence (route B). Our results show that route A, that is amidation followed by aminoalkylation is more favorable, since it affords higher yields. For further investigation, the synthesis of 21 as representative aminoalkylated amide has been selected. Since KYNA amides 16 and 17-containing cationic side-chain-proved to be the most effective analogues [27,28], our attention focused on the substitution of these compounds at position 3 via the mMr. In this reaction, different secondary amines such as pyrrolidine, piperidine, and morpholine were tested. Substrates N-(2-(dimethylamino)ethyl)-4-oxo-1,4-dihydroquinoline-2-carboxamide (16) and 4-oxo-N-(2-(pyrrolidin-1-yl)ethyl)-1,4-dihydroquinoline-2-carboxamide (17), synthesised according to a literature method [21], were aminoalkylated with 15a-c in the presence of formaldehyde, resulting in aminoalkylated KYNA derivatives 18a-c and 19a-c, respectively (Scheme 3).

Molecules 2020, 25, x FOR PEER REVIEW 3 of 14
To test the effect of substituent at 5-, 6-, 7-or 8-positions morpholine as a representative secondary cyclic amine was selected. First, the initial KYNA analogues (7-10) were synthesised with the Conrad-Limpach method applying the following optimization: a) the intermediate was purified by column chromatography using n-hexane:EtOAc as eluent and b) ring closure was carried out utilizing 1,2-dichlorobenzene instead of diphenyl ether used in the literature [8]. Reactions leading to the formation of 11-14 were performed starting from the corresponding ethyl ester and morpholine in the presence of formaldehyde (Scheme 2). It can be concluded that aryl/alkyl substituents at position 6 or 8, and the halogen at position 5 or 7 have no significant influence on the substitution at position 3. Since KYNA amides 16 and 17-containing cationic side-chain-proved to be the most effective analogues [27,28], our attention focused on the substitution of these compounds at position 3 via the mMr. In this reaction, different secondary amines such as pyrrolidine, piperidine, and morpholine were tested. Substrates N- (2- (17), synthesised according to a literature method [21], were aminoalkylated with 15a-c in the presence of formaldehyde, resulting in aminoalkylated KYNA derivatives 18a-c and 19a-c, respectively (Scheme 3). Summarizing these results and those of a previous study [29], the synthesis of this type of KYNA derivatives containing an amide moiety at position 2 and an aminoalkyl group at position 3 can be achieved by applying two different synthetic pathways: amidation followed by aminoalkylation (route A) and a reverse reaction sequence (route B). Our results show that route A, that is amidation followed by aminoalkylation is more favorable, since it affords higher yields. For further investigation, the synthesis of 21 as representative aminoalkylated amide has been selected. Summarizing these results and those of a previous study [29], the synthesis of this type of KYNA derivatives containing an amide moiety at position 2 and an aminoalkyl group at position 3 can be achieved by applying two different synthetic pathways: amidation followed by aminoalkylation (route A) and a reverse reaction sequence (route B). Our results show that route A, that is amidation followed by aminoalkylation is more favorable, since it affords higher yields. For further investigation, the synthesis of 21 as representative aminoalkylated amide has been selected.
A comparison of the overall yields to obtain 21 by using the two approaches shows that amidation followed by aminoalkylation (route A) resulted in the formation of the desired compounds in slightly higher yield (Scheme 4).

Molecules 2020, 25, x FOR PEER REVIEW 4 of 14
A comparison of the overall yields to obtain 21 by using the two approaches shows that amidation followed by aminoalkylation (route A) resulted in the formation of the desired compounds in slightly higher yield (Scheme 4). In the frame of our previous work, KYNA derivatives containing cationic centers (5a, 16, 18c, 17, 19c) have been surveyed in blood-brain-barrier models. It was concluded that 5a, 18c, and 19c have higher permeability towards the brain compared with amides 16 and 17. In the case of the permeability measured from the brain to the outer compartment, compounds 5a and 18c showed the most promising results [P1]. Further investigation of the transport mechanism is still in progress.
Finally, on two selected aminoalkylated KYNA amides 19a and 19b with closely related structures we studied the possible oxo-enol tautomerisation associated with the intramolecular hydrogen-bond framework that might exert significant influence on the intermolecular binding properties thus, on the drug-like character of these highly fuctionalised heterocycles. On the basis of the cross-peak correlations discernible in the 2D-COSY-, 2D-HSQC-and 2D-HMBC spectra, complete assignment of 1 H-and 13 C-NMR data was performed for the two investigated compounds assumed first to have oxo structures (Figure 1). The highly similar 1 H-and 13 C-NMR spectra of these compounds registered in DMSO-d6 solution refer to their practically identical structural features also including oxo-enol tautomerism. One of the two downfield broadened signals could be identified as 10-NH amide resonance coupled with the quartet originated from the adjacent CH2 group as disclosed by a COSY measurement carried out for 19b that gives somewhat sharper signals than does 19a. Regarding tautomerisation, it is of pronounced importance that in the NOESY spectrum of 19b weak, but clearly discernible cross peaks connecting H-8 doublet with the most downfielded broadened signal (separated from the amide 10-NH signal by ca. 0.4 ppm) strongly suggest the presence of the oxo tautomer. Thus, this signal must be originated from the skeletal 1-NH proton being in the proximity of H-8. In order to get an additional support for the exclusive presence or dominance of the oxo tautomers over the enol forms, theoretical 1 H-and 13 C-NMR chemical shifts were calculated for both 19a,b and their enol forms ( Figure 1, Table 1) by GIAO method [30] using B3LYP fuctional [31][32][33] and the extended 6-311++G(2d,p) basis set [34]. The calculations were carried out on the structures optimised by the same functional with 6-31+G(d,p) basis set [35] and IEFPCM solvent model [36] employing dielectric constant of DMSO (ε = 46.7) that represents the conditions of NMR measurements. The comparision of the diagnostic chemical shifts of the skeletal and directly attached atoms calculated for the tautomer pairs with the corresponding experimental 1 H-and 13 C-NMR chemical shifts (Table 1)   In the frame of our previous work, KYNA derivatives containing cationic centers (5a, 16, 18c, 17, 19c) have been surveyed in blood-brain-barrier models. It was concluded that 5a, 18c, and 19c have higher permeability towards the brain compared with amides 16 and 17. In the case of the permeability measured from the brain to the outer compartment, compounds 5a and 18c showed the most promising results [P1]. Further investigation of the transport mechanism is still in progress.
Finally, on two selected aminoalkylated KYNA amides 19a and 19b with closely related structures we studied the possible oxo-enol tautomerisation associated with the intramolecular hydrogen-bond framework that might exert significant influence on the intermolecular binding properties thus, on the drug-like character of these highly fuctionalised heterocycles. On the basis of the cross-peak correlations discernible in the 2D-COSY-, 2D-HSQC-and 2D-HMBC spectra, complete assignment of 1 H-and 13 C-NMR data was performed for the two investigated compounds assumed first to have oxo structures ( Figure 1). The highly similar 1 H-and 13 C-NMR spectra of these compounds registered in DMSO-d 6 solution refer to their practically identical structural features also including oxo-enol tautomerism. One of the two downfield broadened signals could be identified as 10-NH amide resonance coupled with the quartet originated from the adjacent CH 2 group as disclosed by a COSY measurement carried out for 19b that gives somewhat sharper signals than does 19a. Regarding tautomerisation, it is of pronounced importance that in the NOESY spectrum of 19b weak, but clearly discernible cross peaks connecting H-8 doublet with the most downfielded broadened signal (separated from the amide 10-NH signal by ca. 0.4 ppm) strongly suggest the presence of the oxo tautomer. Thus, this signal must be originated from the skeletal 1-NH proton being in the proximity of H-8. In order to get an additional support for the exclusive presence or dominance of the oxo tautomers over the enol forms, theoretical 1 H-and 13 C-NMR chemical shifts were calculated for both 19a,b and their enol forms (Figure 1, Table 1) by GIAO method [30] using B3LYP fuctional [31][32][33] and the extended 6-311++G(2d,p) basis set [34]. The calculations were carried out on the structures optimised by the same functional with 6-31+G(d,p) basis set [35] and IEFPCM solvent model [36] employing dielectric constant of DMSO (ε = 46.7) that represents the conditions of NMR measurements. The comparision of the diagnostic chemical shifts of the skeletal and directly attached atoms calculated for the tautomer pairs with the corresponding experimental 1 H-and 13 C-NMR chemical shifts (Table 1) supports at least the dominance of the oxo tautomers over the enol forms. Finally, the relative thermodynamic stability obtained for the tautomeric pairs by B3LYP/6-31 + G(d,p) method also confirm the dominance of the oxo tautomers over the enol counterparts [∆G(19a/enol − 19a/oxo) = +0.57 kcal/mol, ∆G(19b/enol − 19b/oxo) = +0.59 kcal/mol].

Conclusions
A series of novel C-3 substituted kynurenic acid derivatives were synthesised by using the modified Mannich reaction. Aminoalkylations were achieved starting from primary or secondary amines, using formaldehyde as the representative aldehyde component. In addition, the applicability of benzaldehyde was also proven. Among the amine components, cyclic secondary amines such as morpholine, piperidine, or N-methylpiperazine were used. The limitations of the reaction were examined on the synthesis of different KYNA derivatives containing phenyl, chloro, or methyl substituents in ring B. The modified Mannich reaction was further tested starting from KYNA amides containing cationic center at the side-chain. In the case of 21 both sequences proceeding via aminoalkylation followed by amidation or amidation followed by aminoalkylation, were investigated. Amidation followed by aminoalkylation resulting in 21 was found to give a somewhat higher yield. By means of 1 H-and 13 C-NMR spectroscopy including NOESY measurement it was established that two representative aminoalkylated products feature quinolin-4(1H)-one skeletal structure rather than quinolin-4-ol tautomer form. The dominance of the oxo form over the enol counterpart, also supported by DFT modelling studies, can certainly be stated for all the KYNA amides presented in this contribution.

Materials and Methods
The 1 H and 13 C-NMR spectra were recorded in DMSO-d 6 , CDCl 3 and D 2 O solutions in 5 mm tubes at room temperature (RT), on a Bruker DRX-500 spectrometer (Bruker Biospin, Karlsruhe, Baden Württemberg, Germany) at 500 (1H) and 125 (13C) MHz, with the deuterium signal of the solvent as the lock and TMS as internal standard (1H, 13C). The 2D-COSY, NOESY, HSQC, and HMBC spectra of 19a and 19b were obtained by using the standard Bruker pulse programs (cosygpppqf (2D COSY with gradient pulses for selection and purge pulses before relaxation delay d1) for COSY, noesygpphpp (2D phase sensitive NOESY with gradient pulses in mixing time and purge pulses before relaxation delay d1 for NOESY), hsqcetgp (2D phase sensitive HSQC using Echo/Antiecho-TPPI gradient selection with decoupling during acquisition and using trim pulses in inept transfer) for HSQC and hmbcgpndqf (2D H-1/X HMBC optimised on long range couplings, no decoupling during acquisition using gradient pulses for selection) for HMBC, Bruker Biospin, Karlsruhe, Baden Württemberg, Germany). All calculations were carried out by using Gaussian 09 software package [37]. The optimised structures are available from the authors.
Melting points were determined on a Hinotek X-4 melting point apparatus. Elemental analyses were performed with a Perkin-Elmer 2400 CHNS elemental analyser. Merck Kieselgel 60F 254 plates were used for TLC.