Synthesis of New Perhydropyrrolo[1,2-a]pyrazine Derivatives and Their Evaluation in Animal Models of Epilepsy

A series of novel stereochemically pure derivatives of the investigative broad-spectrum anticonvulsant ADD408003 was designed and synthesized. Five-center four-component (U-5C-4CR) and four-center three-component (U-4C-3CR) variants of Ugi reaction were used in the key step of the synthetic pathways. The compounds obtained were evaluated for the anticonvulsant activitiy in the maximal electroshock seizure (MES), subcutaneous Metrazole (scMET) and minimal clonic seizure (6 Hz) animal models of epilepsy. The efficacies of most derivatives in the 6 Hz model of pharmacoresistant partial seizures were markedly higher than in the ‘classical’ MES and scMET models. The most active compounds, (4R,8aR)-3a, and (4S,8aS)-6 displayed median effective doses (ED50) of 47.90 and 126.19 mg/kg, respectively, for the 6 Hz test.


Introduction
According to epidemiological studies, epilepsy affects approximately 1% of the world's population [1]. Although significant advances have been achieved in pharmacotherapy of this disorder, currently available anticonvulsant drugs (AEDs) produce satisfactory seizure control only in 60%-70% of patients. Moreover, their usage is often associated with disturbing side-effects which seriously limit quality of patients' life. Thus, there is a substantial need for further development of novel, safer and more efficient AEDs [2][3][4][5][6].
In our recent reports, we described a structurally novel class of investigative anticonvulsants, derivatives of perhydropyrrolo [1,2-a]pyrazines [7]. Among them, compound ADD408003 ( Figure 1) revealed high and broad activity in various preclinical animal models of epilepsy, including the models of pharmacoresistant epilepsy. A preliminary structure-activity relationship (SAR) study revealed that several structural patterns are necessary for the high anticonvulsant activity: the (S,S) absolute configuration on the stereogenic centers, the presence of the annulated pyrrolidine ring, the presence of imide moiety and the benzene ring in C-4 position of the pyrrolo [1,2-a]pyrazine core.  To complement our previous extensive SAR investigations in this group of active compounds [7][8][9][10], we synthesized and pharmalogically evaluated a new series of ADD408003 derivatives. Since the mechanism of action of the parent molecule remains unknown, the new compounds have been designed according to diverse classical medicinal chemistry methods. First, we decided to examine if increasing the distance of the benzene ring from the imide moiety by methylene insertion would influence the anticonvulsant activity of ADD408003. Next, we asked whether replacing the hydrogen atom at C-4 of the parent compound with a non-epimerizable substituent would prevent the putative metabolic conversion to the inactive (4R,8aS) isomer, thus prolonging the anticonvulsant activity. Further, we incorporated alkyl residues in the C-4 position. Finally, based on the common structural motifs and the stereochemistry of ADD408003 and the potent broad-spectrum AED Levetiracetam, we have designed a hybrid of both molecules ( Figure 2).
Since our previous studies on the stereochemistry-activity relationship of structurally related pyrido [1,2-a]pyrazines revealed interesting activity of (R,R) stereoisomers, we now focused on synthesizing both (R,R) and (S,S) enantiomers of the new ADD408003 derivatives [8].
The appropriate (S)-amino acids were condensed with aldehydes or aliphatic ketones, tert-butyl isocyanide and methanol in the presence of catalytic amount of FeCl 3 or TiCl 4 to give Ugi adducts 1a-g with chemical yields ranging from 17% to 64% (Table 1). When aldehydes or unsymmetrical ketones were employed as the starting materials, the reactions proceeded with the formation of the new stereocenters at carbons C-1 of the products 1a-d. In cases of adducts 1a-c the diastereoinduction was in favor of the (2S,1S) configurations, whereas equal amounts of (2S,1S)-1d and (2S,1R)-1d isomers were obtained for benzylmethyl ketone as the carbonyl component. In all cases the isomeric mixtures could be efficiently resolved by flash column chromatography on silica.
In the last step of the synthesis, amido esters 2 were converted to their cyclic derivatives 3 upon treatment with 1 eq. of NaOH in EtOH, with chemical yields ranging from 55% to 95%. The cyclocondensation of (2S,1S)-2a-c was accompanied by a slight degree of epimerization at carbon C-4. Pure diastereomers (4S,8aS)-3a-c were obtained by recrystallization.
It is well known that the U-5C-4CR reaction is highly versatile with regard to the substrate scope [11][12][13][14][15][16][17]. In this work and in our previous investigations, we have successfully coupled cyclic amino acids with aliphatic ketones [13]. However, to the best of our knowledge, aromatic ketones have not been employed as carbonyl components in this process. We were particularly interested in L-Proline and acetophenone adduct, as it could lead to the C-4 methylated ADD408003 derivative (4S,8aS)-3h. Propitiously, we observed by LC/MS that the U-5C-4CR product 1h was formed after 1 day, at room temperature, with the use of a TiCl 4 catalyst. The diastereoinduction slightly in favored (2S,1S)-1h (dr = 2:1, LC/MS). Unfortunately, attempts to optimize this particular reaction led to only a 13% yield after 7 days of stirring at room temperature.
Nevertheless, enough (4S,8aS)-3h was obtained for initial evaluation in animal models of epilepsy, following the synthetic route described for 3a-g (Scheme 2 and Figure 3). Similar to what had been observed for 1d-g cyclic imides 3h were formed upon treatment of the respective diastereomers of 1h with BF 3 •2CH 3 COOH complex. Interestingly, cyclocondensation of (2S,1S)-1h diastereomer to (4S,8aS)-3h was more facile than in the case of its epimer (2S,1R)-1h.   Encouraged by the results of U-5C-4CR of acetophenone, L-proline, tert-butyl isocyanide and MeOH, we tested if the biphenyl analog of ADD408003 could be obtained in the same manner as 3h (Scheme 3). However, the more bulky benzophenone failed to react and no traces of Ugi adduct were detected by LC/MS of the crude postcondensation mixture. On the other hand, when tert-butyl isocyanide was replaced by the linear n-butyl isocyanide, we observed formation of the product (2S)-1i in a modest 2% yield. This was in agreement with our previous observations that steric factors played the most important role for the outcome of U-5C-4CR of ketones [13]. As a further example of this relationship, the yield of acetophenone adduct 1j incresed from 13% to 32%, when less bulky n-butylisocyanide was used in lieu of the branched tert-butyl isomer. A loss of stereoselectivity was observed at the same time. The desired (4R,8aR) and (4S,8aS) diastereomers of 6, bearing oxygen atom at carbon C-6, were synthesized using the intramolecular Ugi five-center four-component reaction (U-4C-3CR) of methyl (S)-phenylglycinate, levulinic acid, tert-butyl isocyanide and MeOH as a key step of the synthetic sequence depicted in Scheme 4 [18][19][20]. The uncatalyzed reaction proceeded with a high yield and the resulting equimolar mixture of (2R,αS)-4 and (2S,αS)-4 was quantitatively separated by column chromatography on silica. The subsequent treatment of the respective diastereomers of 4 with BF 3 •2CH 3 COOH gave the dealkylated amido-esters 5 and cyclic imides 6. Similar to what had been observed for dealkylation reactions of 1h, acid-mediated cyclocondensation was more facile in case of the (2S,αS)-4 diastereomer. Finally, base-mediated cyclization of (2R,αS)-5 proceeded with a high degree of epimerization on C-4 of the products, resulting in the chromatographically separable mixture of (4S,8aR)-6 and the desired (4R,8aR)-6. No epimerization took place in the analogous cyclocondensation reaction of (2S,αS)-5.

Anticonvulsant Evaluation
Compounds 3 and 6 were evaluated in the in vivo animal models of epilepsy within the Anticonvulsant Screening Program (ASP) of The National Institute of Neurological Disorders and Stroke (NINDS), at The National Institutes of Health according to the established decision schemes and using protocols described in the Experimental Section of this article [21]. Primary anticonvulsant studies involved two tests: maximal electroshock seizure (MES) and subcutaneous Metrazol (scMET), in mice [22]. It is emphasized that nearly all clinically significant AEDs are protective in at least one of these two models, thus making them very useful tools for initial high througput screening of candidate anticonvulsants. The MES test employs an electrical stimulus to induce generalized tonic-clonic seizures and identifies which compounds prevent the spread of seizures. The scMET model utilizes chemically induced myoclonic seizures and recognizes the agents that act by raising the seizure threshold. Selected compounds were subjected to evaluation of anticonvulsant activity in the minimal clonic seizure (6 Hz) model of pharmacoresistant partial seizures [23][24][25]. Recently, it was found that Levetiracetam, a novel broad-spectrum AED, is ineffective in most of the classic tests (e.g., MES and scMET) and that it suppresses seizures in the 6 Hz model at the same time. Therefore, an expanded approach employed by the ASP is to utilize this test as a screening procedure to examine the compounds that are inactive in conventional models, but which can possess novel activity spectra. In addition to the primary anticonvulsant evaluation in the MES, scMET and 6Hz models, the acute neurological impairment (TOX) was assessed in rotorod test [26]. The results are summarized in Tables 2 and 3. Table 2. Anticonvulsant activity and neurotoxicity of compounds in the MES and scMET models following intraperitoneal (ip.) administration in mice.   In contrast to the parent compound ADD408003, its opposite enantiomer (4R,8aR)-3a displayed only a weak activity in the 'classical' MES model (1/1 at 300 mg/kg, at 0.5 h). A considerable level of neurotoxicity was observed at the same dose and timepoint. An increase in the distance of the benzene ring from the imide moiety of ADD408003 and (4R,8aR)-3a by a methylene unit resulted in (4S,8aS)-3b and (4R,8aR)-3b, respectively. This structural modification did not improve the activity in MES model, when compared to parent molecules, although no signs of neurotoxicity were observed in the TOX test. The piperidine homologs of (4S,8aS)-3b and (4R,8aR)-3b, (4S,9aS)-3c and (4R,9aR)-3c, respectively, showed somewhat greater potency in the MES test (1/1 and 1/3 at 300 and 100 mg/kg, respectively, at 0.5 h). On the other hand, (4R,9aR)-3c isomer showed pronounced unwanted neurotoxic effect at 300 mg/kg. Replacing the hydrogen atom at C-4 with methyl group in (4S,8aS)-3b and (4R,8aR)-3b resulted in inactive (4S,8aS)-3d and (4R,8aR)-3d, respectively. Dibenzyl analogs (8aS)-3e and (8aR)-3e displayed no activity in MES model and no neurotoxicity in TOX test at a dose of 100 mg/kg. The alkyl derivatives (8aS)-3f and (8aS)-3g were devoid of desired anticonvulsant activity and the former produced severe neurotoxicity. The C-4 methylated analog of ADD408003, (4S,8aS)-3h, showed high levels of seizure protection in MES model (1/4 and 3/4 at 0.25 and 0.5 h, respectively, at 100 mg/kg) and lack of neurotoxic effects. However, the compound was not competitive when compared to the parent molecule, and due to the low synthetic accessibility it was not scheduled for secondary anticonvulsant evaluations. A comparable activity was observed for the oxo-derivatives (4S,8aS)-6 and (4R,8aR)-6.
The anticonvulsant activity of investigated compounds in the scMET model in mice was markedly lower than in the MES screen. Only (4R,8aR)-3a and (4S,9aS)-3c showed considerable levels of seizure protection (1/1 and 5/5, respectively, at 0.5 h) at dose of 300 mg/kg.
To conclude the above results, the investigated compounds showed only moderate anticonvulsant activity in the 'classical' MES and scMET screening. Encouragingly, the levels of seizure protection in the 6 Hz model of pharmacoresistant epilepsy were markedly higher (Table 3).
Compounds (4R,8aR)-3a and (4S,8aS)-6 were most potent in the primary 6 Hz screening, at both early and late time points. Therefore they were further subjected to in vivo quantification in this model (Table 4). They displayed median effective dose (ED 50 ) values of 47.90 mg/kg and 126.19 mg/kg, respectively, at 0.25 h. Although the compounds proved less competitive than the parent ADD408003, their potencies and protective indices were comparable to those of standard AEDs. Table 4. Quantification studies of (4R,8aR)-3a and (4S,8aS)-6 in the 6 Hz and neurotoxicity tests following intraperitoneal (ip.) administration in mice.

Synthesis of Compounds 1 by U-5C-4CR Condensation
FeCl 3 (for 1a-c) or TiCl 4 (for 1d-g) (5 mol%) and isocyanide (1.0 eq.) were added to a stirred solution of appropriate α-amino acid (1.2 eq.) and carbonyl component (1.0 eq.) in MeOH (100 mL). The mixture was stirred at rt for 24 h (72 h for 1d-g) and the volatiles were removed under reduced pressure. The resulting crude products were purified by FC.
3.3.4. The minimal clonic seizure test (6 Hz) [23][24][25] The 6 Hz screen is an alternative electroshock paradigm that uses low-frequency (6 Hz), longduration (3 s) electrical stimulation. Mice were tested at time intervals between 0.25 and 4 h following intraperitoneal doses of 100 mg/kg of test compound. Corneal stimulation (0.2 ms-duration monopolar rectangular pulses at 6-Hz for 3 s) was delivered by a constant-current device. During the stimulation, mice were manually restrained and released into the observation cage immediately after the current application. The seizures manifested in 'stunned' posture associated with rearing, forelimb, automatic movements and clonus, twitching of the vibrissae and Straub-tail. The duration of the seizure activity ranged from 60 to 120 s in untreated animals. At the end of the seizure, animals resumed their normal exploratory behavior. The experimental end point was protection against the seizure. The animal was considered to be protected if it resumed its normal exploratory behavior within 10 s from the stimulation. The quantitative determination of the median effective (ED 50 ) and toxic doses (TD 50 ) was conducted at previously calculated time of peak effect (TPE) using the intraperitoneal route in mice. Groups of at least eight animals were tested using different doses of test compound until at least two points were determined between 100% and 0% protection and minimal motor impairment. The dose of the candidate substance required to produce the desired endpoint in 50% of the animals in each test, and 95% confidence interval were calculated by a computer program based on methods described by Finney [27].

Conclusions
We have synthesized a series of novel stereochemically pure pyrrolo [1,2-a]pyrazine derivatives by use of pathways based on the U-5C-4CR and the U-4C-3CR multicomponent reactions. The compounds displayed weak to good anticonvulsant activities in the MES model, while only few of them were active in the scMET screen. The efficacy of most of the new derivatives in the 6 Hz model of pharmacoresistant partial seizures was significantly higher than in the 'classical' models. As a supplement to our previous extensive SAR studies within this group of compounds, we learned that: (1) increasing the distance of the benzene ring from the imide moiety of ADD408003 lowered its potency in MES, scMET and 6 Hz models, (2) introduction of non-epimerizable residue in lieu of hydrogen atom at C-4 did not enhance the pharmacological properties of the parent compound, (3) introduction of alkyl residues at C-4 produced either inactive or strongly neurotoxic and proconvulsant compounds, (4) for most of the newly synthesized derivatives, the differences in pharmacological activities between (R,R) and (S,S) were only slight.
The opposite enantiomer of ADD408003 (4R,8aR)-3a was most active from the series with an ED 50 value of 47.90 mg/kg and a protective index (PI) of 5.8 in the 6 Hz model of pharmacoresistant epilepsy. The ADD408003-Levetiracetam hybrid (4S,8aS)-6 was markedly less potent in this test, with ED 50 value of 126.19 mg/kg and PI of 2.1. However, the overall synthetic feasibility of this compound was very high. Thus, manipulating the amino acid and keto acid component in the initial intramolecular U-4C-3CR condensation opens the access to chemical diversity of potentially interesting (4S,8aS)-6 analogs.