Synthesis and biological evaluation of novel N-substituted nipecotic acid derivatives with tricyclic cage structures in the lipophilic domain as GABA uptake inhibitors

A new class of GABA reuptake inhibitors with sterically demanding, highly rigid tricyclic cage structures as the lipophilic domain was synthesized and investigated in regard to their biological activity at the murine GABA transporters (mGAT1–mGAT4). The construction of these compounds, consisting of nipecotic acid, a symmetric tricyclic amine, and a plain hydrocarbon linker connecting the two subunits via their amino nitrogens, was accomplished via reductive amination of a nipecotic acid derivative with an N-alkyl substituent displaying a terminal aldehyde function with tricyclic secondary amines. The target compounds varied with regard to spacer length, the bridge size of one of the bridges, and the substituents of the tricyclic skeleton to study the impact of these changes on their potency. Among the tested compounds nipecotic acid ethyl ester derivates with phenyl residues attached to the cage subunit showed reasonable inhibitory potency and subtype selectivity in favor of mGAT3 and mGAT4, respectively.


Introduction
A balanced interplay between excitatory and inhibitory neurotransmission represents the fundamental basis for proper functioning of the central nervous system (CNS) in mammals. A disruption of this interplay due to, for example, an insufficient signaling of GABAergic neurons can lead to or intensify neurological disorders like Alzheimer's disease (AD) [1,2], depression [3], epilepsy [4,5], or Parkinson's disease (PD) [6][7][8]. One approach to influence the GABAergic neurotransmission and thus to treat the aforementioned diseases is to increase the release and the concentration of γ-aminobutyric acid 1 (GABA), representing the predominant inhibitory neurotransmitter in the CNS [9][10][11], in the synaptic cleft. As GABA is quickly removed from the synaptic cleft by reuptake into the presynaptic neurons and surrounding glia cells this may be achieved by inhibition of the GABA transporters (GATs) in charge of this process [12][13][14].
GATs are membrane-bound transport proteins of the solute carrier family 6. They consist of 12 transmembrane helices and translocate their substrate GABA through the cell membrane by cotransport of sodium and chloride ions [15,16]. Latest findings suggest a stoichiometry of 3:1:1 (Na + :Cl − :GABA) for sodium and chloride ions and GABA in this transport process [17]. For the GATs four different subtypes are known, which are denominated differently depending on the species they were cloned from [14,18]. When originating from mouse tissue they are termed mGAT1-mGAT4 [18][19][20]. For all other species including human, dog, or rat they are denominated as GAT-1 (≡ mGAT1), BGT-1 (≡ mGAT2), GAT-2 (≡ mGAT3), and GAT-3 (≡ mGAT4) whereby the individual transporter name is provided with a prefix such as h for human to indicate the individual species. This nomenclature has also been adopted by the Gene Nomenclature Committee of the Human Genome Organization (HUGO) but without any prefix as which it has also found use as a species independent nomenclature system [18,20,21]. As the biological test system applied in this study is based on GATs originating from mice, for the sake of consistency the corresponding nomenclature mGAT1-mGAT4 will be used throughout this paper.
Although they are structurally closely related, mGAT1-mGAT4 are expressed very differently. The predominating transporter subtype in the CNS is mGAT1, which is primarily located on the plasma membrane of presynaptic GABAergic neurons [14,18,22]. mGAT4 represents the second most abundant GAT in the brain, albeit with distinctly lower concentration than mGAT1, where it is responsible for the transport of GABA into glia cells which are neighboring the GABAergic neurons [21][22][23][24]. In contrast mGAT2 and mGAT3 are only weakly expressed in the brain and occur mainly in kidney and liver. As the level of mGAT2 and mGAT3 in the brain is too low for having a reasonable effect on the termination of GABAergic neurotransmission, mGAT1 and mGAT4 are the most promising targets amongst these proteins to be addressed for the treatment of above-mentioned diseases [22,25,26].
Muscimol (2) and THPO (3) which are structurally related to GABA (Fig. 1), the natural substrate of GATs, were identified to be weak inhibitors at GATs. Structural alterations of the isoxazolol function of THPO (3) led to racemic nipecotic acid (rac-4) and guvacine (5) as the first reasonably potent inhibitors of the GATs. However, because of their zwitterionic character at physiological pH and their high polarity these compounds are not able to pass the blood-brain barrier (BBB). In order to increase lipophilicity and BBB permeation di-and triaryl residues were added via a linker to the amino nitrogen of the parent compounds. GAT inhibitors of this general structure have been synthesized in large numbers and broad structural diversity [27][28][29][30][31][32][33]. This includes the most prominent GAT inhibitor Tiagabine 6, a mGAT1 selective inhibitor, which is used in the treatment of epileptic seizures [34,35]. The Nlinked lipophilic aryl-alkyl side chain does not only improve the permeation of the BBB but often also mediates a substantial increase in potency and subtype selectivity of the GAT inhibitors. Modeling studies revealed the putative binding pose for mGAT1 inhibitors such as Tiagabine (6). According to that the amino acid subunit binds in the substrate-binding pocket (S1) whereas the lipophilic residue is accommodated in a binding site (S2) equipped with aliphatic residues and located in the vestibule oriented towards the extracellular space [36][37][38]. NO711 (7) and (S)-SNAP-5114 (8) represent two further well known GAT inhibitors ( Fig. 2) of which the former, 7, like Tiagabine (6) is highly selective for mGAT1 and can be considered like 6 as prototype for compounds exhibiting this subtype selectivity. A major difference of (S)-SNAP-5114 (8) as compared to these compounds is to be attributed to the lipophilic domain, which by comprising a triarylmethyl unit is distinctly larger than that of 6 and 7. It is this large steric demand of the lipophilic triarylmethyl subunit together with the (S)-configuration of nipecotic acid that is thought to mediate the subtype selectivity for mGAT4 of (S)-SNAP-5114 (8) [16,39].
Tiagabine (6) suffers from a series of adverse side effects and (S)-SNAP-5114 (8), though among the most potent mGAT4 inhibitors, of moderate potency [39,40]. Thus there is still a great need for GAT inhibitors with less adverse effects and higher potency. Structural changes to the aforementioned prototypic structures led to compounds with partially rigidized lipophilic domains at the terminal position of alkyl or heteroalkyl chains originating from the amino nitrogen of the polar subunit [41,42]. That way aryl groups present in the lipophilic domain were forced to adopt specific spatial orientations. Another option to achieve a well-defined orientation of substituents in the lipophilic domain is to use compounds with a polycyclic cage structure as central unit. The high rigidity of cage structures allows to reduce the flexibility of attached substituents and may lower the conformational entropy penalty resulting from target binding [43,44]. In addition, the high lipophilicity of cage-derived hydrocarbon rich structures may positively affect the pharmacokinetic and pharmacodynamic properties of drugs as it can facilitate the crossing of the BBB and the binding to lipophilic domains [45,46]. As a result of their inherent stability and steric bulk, polycyclic cage compounds also can slow down metabolic degradation [45][46][47]. Currently drugs with polycyclic cage structures are in use for the treatment of neurodegenerative diseases like γ-aminobutyric acid (GABA) 1 muscimol 2 3 THPO racemic nipecotic acid rac-4 guvacine 5 Fig. 1 Structures of GABA and some low molecular weight GAT inhibitors AD and PD [46]. The drug Deramciclane (rac-9) is a rare example for a GAT inhibitor albeit with moderate inhibitory activity at all four GAT subtypes in which a polycyclic cage serving as lipophilic residue is present [48]. Since no systematic study aiming at the development of GAT inhibitors with a polycyclic cage subunit as lipophilic domain has been presented so far, though this appears to be quite rewarding, we intended to carry out such a study. To this end, polycyclic cage structures based on a 2azabicyclo[2.2.2]octane scaffold should be used, as they are easily available by an efficient and straightforward synthesis recently reported by us [49,50]. For the present study the symmetric tricyclic imines 10 should be used (for general structure see Fig. 3). Though these polycyclic imines 10 display the same 2-azabicyclo[2.2.2]octane skeleton, the size of the bridge between the two substituted bridgehead atoms and thus the size of the tricyclic scaffold but also the orientation of the bridgehead substituents may be varied [51], thus allowing to study the impact of these two parameters on the inhibitory potency of the target compounds. As bridgehead substituents initially exclusively methyl and phenyl residues should be used as the synthesis of the respective symmetric tricyclic imines is known [51]. For the connection of these tricyclic cage units via their amino nitrogen, resulting from reduction of the imine function, with the amino nitrogen of racemic nipecotic acid (rac-4) a plain alky chain linker of varying length should be used. That way, the influence of the linker length on the biological activity should be explored as well.

Materials and methods
Anhydrous reactions were performed under an argon atmosphere in vacuum-dried glassware. All solvents were distilled prior to use and dry 1,4-dioxane and CH 2 Cl 2 were prepared under a nitrogen atmosphere according to standard procedures [52]. The CH 2 Cl 2 employed as solvent in reactions was stabilized with amylene, the CH 2 Cl 2 used for workups was stabilized with ethanol. All purchased chemicals were used without further purification. TLC was performed with plates from Merck KGaA (silica gel 60 F 254 ). For purification via flash chromatography (FC) silica gel 60 (40-63 µm mesh size) from Merck KGaA was employed. Purification by preparative RP-MPLC was performed using an Büchi instrument (C-605 binary pump system, C-630 UV detector at 254 nm and C-660 fraction collector) and a Sepacore glass column B-685 (26 × 230 mm) equipped with YMC Gel Triart Prep C18-S (12 nm, 5-20 µm). Melting points were determined with a BÜCHI 510 melting point apparatus and are uncorrected. Infrared spectra were recorded with a Perkin Elmer Paragon 1000 and a Jasco FT/IR-410. Solid substances were 10 rac-11 Fig. 3 General structure of the polycycles 10 to be used as starting material for the construction of the desired GAT inhibitors rac-11 Tiagabine (6) pIC 50  ). The purity of the biologically tested compounds was determined by quantitative 1 H NMR (qH NMR) according to a method described by Pauli et al. with internal calibration [54]. The qH NMR measurements were carried out under conditions allowing complete relaxation to assure the exact determination of peak area ratios. Used internal standards were benzyl benzoate (LOT# BCBN 6347V; purity 99.43%) and 1,3,5-trimethoxy benzene (LOT# BCBW 3670; purity 99.96%) in CDCl 3 , CD 2 Cl 2 , CD 3 OD or CD 3 OD + 1M NaOD in D 2 O (6:1). All tested esters had a purity >95%. The tested carboxylic acids contained varying amounts of water which was not considered an impurity as the acids were dissolved in aqueous media later on to perform the assays. The amount of water was identified by qH NMR and calculated from the change of the peak area ratio of the exchangeable protons (water peak) to the solvent residual protons compared to the same peak area ratio determined for pure solvent. In due consideration of the amount of water contained, the purity of all carboxylic acids was >95% with exception of the biologically inactive acids rac-18b and rac-11m, for which no purity was determined.

General procedures
Synthesis of ethyl nipecotate precursors rac-15a-15f (general procedure/GP1) Potassium carbonate and sodium iodide were added to a solution of racemic ethyl nipecotate rac-16 (1.0 equiv) in the solvent stated. The organic halide was added to this mixture that was stirred for the time period and at the temperature indicated in the respective experiment. The mixture was concentrated under vacuum, dissolved in ethyl acetate, and washed with water. Drying of the organic phase (Na 2 SO 4 ) and removal of the solvent under vacuum afforded the crude product which was purified by FC.
Part B: In the meantime, the dimethoxy acetal (2.0 equiv) was dissolved in CH 2 Cl 2 (16 mL/mmol), and FeCl 3 • 6 H 2 O was added. The acetal/salt suspension was rotated on a rotary evaporator at 45°C (no vacuum) for 20 min. In doing so, the total volume was maintained by regular solvent addition. The suspension was quenched with concentrated aqueous NaHCO 3 , extracted with CH 2 Cl 2 (stabilized with amylene) for three times, dried (Na 2 SO 4 ), and concentrated under vacuum. The remaining crude aldehyde was dissolved in CH 2 Cl 2 (6.25 mL/mmol Acetal ), added to the imine/triacetoxyborohydride solution and stirred for the time period and at the temperature stated in the experiment. The reaction was quenched with potassium carbonate solution (1 mol/L), extracted with CH 2 Cl 2 for three times, dried (Na 2 SO 4 ), and concentrated under vacuum to afford the crude product which was finally purified by FC (SiO 2 , EtOAc/MeOH/NEt 3 88:10:2) and, if denoted, by RP-MPLC (DCM/MeOH 1:1).
Deprotection and reductive amination of the dimethoxy protected aldehydes rac-15e-15f with tricyclic imines 10e-10f (general procedure/GP3) Part A: The tricyclic imine was dissolved in MeOH (13.3 mL/mmol) and sodium cyanoborohydride (5 equiv) and hydrochloric acid (1 mol/L in Et 2 O, 10 equiv) were added. The solution was stirred at 20°C for 3 h. The reaction was quenched with water, adjusted to pH = 11 with K 2 CO 3 and the crude amine was extracted with CH 2 Cl 2 for three times. After drying (Na 2 SO 4 ) and removal of the solvent under vacuum the crude amine was dissolved in CH 2 Cl 2 (15 mL/mmol) again and sodium triacetoxyborohydride (2.5 equiv) and acetic acid (2.1 equiv) were added.
Part B: Identical with Part B from GP2.
Hydrolysis of the N-substituted nipecotic acid ethyl esters (general procedure/GP4)

MS binding assays
For the MS binding assays mGAT1 membrane preparations, obtained from a stable HEK293 cell line, and NO711 as native MS marker were employed in competitive binding experiments as described earlier [56].

Synthesis
As direct precursors for the preparation of the target compounds rac-11 their carboxylic acid esters rac-19 should be employed. Their synthesis should be accomplished by linking of the tricyclic amines 14 with suitable N-substituted nipecotic acid derivatives via reductive amination (Fig. 4). Accordingly, besides the tricyclic amines 14, which should be accessible from the tricyclic imines 10 by reduction, nipecotic acid derivatives carrying N-alkyl substituents with an aldehyde function at the terminal position of the N-alkyl chain were needed. These nipecotic acid derivatives with Nalkyl chains of different lengths between the amino nitrogen and the terminal aldehyde function, rac-12 and rac-13, should be generated from suitable precursors, rac-15, in which the aldehyde function is present in masked form, for instance as alcohol or acetal group.
Preparation of the aldehyde precursors rac-15a-f and generation of the aldehydes rac-12-13 The required nipecotic acid derivatives with an N-alkyl residue with a terminal alcohol or acetal function, Fig. 4 Retrosynthetic analysis of the targeted N-substituted nipecotic acid derivatives rac-11 rac-15a-f, were obtained by N-alkylation of racemic ethyl nipecotate rac-16 with ω-hydroxy and ω-dimethoxy substituted n-propyland n-butlyhalides 17a-b and 17e-f and the ω-(1,3-dioxolane-2-yl) substituted ethyl-and n-propylhalides 17c-d, respectively, in good to excellent yields ( Table 1, entries 1-6). The synthesis of alcohol rac-15a was performed according to a procedure described by Dhar et al. [39], which method was also used for the construction of rac-15b-f. As besides the aldehyde precursors rac-15a-f also the corresponding free carboxylic acids rac-18a-f should be evaluated for their inhibitory potency at mGAT1-mGAT4 the later were synthesized as well. This was accomplished by treating rac-15a-f with Ba(OH) 2 • 8 H 2 O in analogy to a literature procedure [33], which led to rac-18a-f in moderate to excellent yields (43-92%, Table 1, entries 1-6).
With the aldehyde precursors rac-15a-f in hand, the synthesis of the aldehydes rac-12-13 was studied. Attempts to access the aldehydes rac-12-13 by oxidation of the alcohols rac-15a-b showed, that even using mild oxidation conditions, e.g. Swern-, Parikh-Doering or Dess-Martin periodinane oxidation, the desired aldehydes were not formed or only in traces. As, in addition, the starting material had been completely consumed and a multitude of side products appeared, this approach was dismissed. Instead attempts to deprotect the acetals rac- Table 1 Synthesis of the nipecotic acid derived aldehyde precursors rac-15a-f and their hydrolysis to the carboxylic acids rac-18a-f Reagents and conditions: (a) K 2 CO 3 , NaI, neat, acetone or 1,4-dioxane; (b) Ba(OH) 2 • 8 H 2 O, MeOH/H 2 O; (c) various conditions tested, for rac-15e-f: FeCl 3 • 6 H 2 O, CH 2 Cl 2 a Synthesis according to literature [39] 15c-f were undertaken. In this regard, only reaction conditions that should allow to deprotect the acetals without affecting the ester function were taken into account.
However, the crude aldehyde rac-13 was contaminated with unknown side products, resulting from decomposition most likely, which in addition to the dimethyl acetal rac-15f could not be separated from the desired compound rac-13. A similar situation was observed when the deprotection of dimethyl acetal rac-15e to aldehyde rac-12 was attempted. In consequence, the crude aldehydes rac-12-13 should be directly used for the subsequent reductive amination without prior chromatographic purification and without any delay.
Reduction of the imines 10a-f and synthesis of the target compounds rac-11a-m The amines 14a-f, required for the reductive amination of rac-12 and rac-13, were synthesized by reduction of the  [50]. Indeed, when imines 10a-f were treated with NaBH 3 CN and HCl in methanol the corresponding amines 14a-f were formed. Unfortunately, amines 14a-b (bridge size m = 0) were found to be instable and to decompose quickly, whereas amines 14c-f did not show such a behavior. Hence, in addition to the aldehydes rac-12-13, it seemed best to use also amines 14a-f directly after their formation without prior purification and isolation.
Considering that both, the aldehydes rac-12-13 and amines 14a-f had appeared to be labile to some extent, we intended to generate and directly subject them to the next reaction step, the reductive amination to give the respective esters rac-19. Thus, for the overall reaction sequence first acetals rac-15e-f should be cleaved by treatment with FeCl 3 • 6 H 2 O in CH 2 Cl 2 . Then the respective aldehyde should be added to a mixture of imine and reducing agent, which was premixed to mediate imine reduction and to allow subsequent reductive amination of the aldehyde function of rac-12 or rac-13 with the formed amine. When in a test reaction aldehyde 13 was added to a mixture of an imine, structurally similar to imine 10c but with one of the methyl residues substituted by hydrogen (for a depiction of the structure see compound rac-14a in [49]), and NaBH 3 CN, that had proven well suited for the reduction of the imines 10 to the corresponding amines 14, besides the reductive amination product also the alcohol rac-15b resulting from the reduction of aldehyde rac-13 was obtained. However, when the mild reducing agent NaBH(OAc) 3 [62,63] in combination with acetic acid was used instead of NaBH 3 CN no such unfavorable reaction occurred. Thus, starting from dimethyl acetal rac-15f and the imines 10a-d the esters rac-19g-k were obtained in moderate to good yields (Table 2, entries 7-10). This method could also successfully be applied to the reductive coupling of dimethyl acetal rac-15e -via the corresponding aldehyde rac-12-with the imines 10a-d to give the desired esters rac-19a-d. However, in these cases the yields were poor ( Table 2, entries 1-4), which is likely to be attributed to the instability of the intermediate aldehyde rac-12 and its propensity to undergo a retro-Michael addition leading to further side reactions.
Unfortunately, the reaction of imines 10e-f with in situ generated aldehydes rac-12 and rac-13 did not lead to the desired products, the nipecotic acid esters rac-19e-f and rac-19l-m under the aforementioned reaction conditions. Actually, despite treatment with NaBH(OAc) 3 imines 10e-f remained unchanged, indicating that they are less reactive than compounds 10a-d. This is likely to be due to a more severe shielding of the imine function by the adjacent R 1 groups as a result of the larger "upper" bridge (m = 2) in 10e-f as it was claimed before in cycloaddition reactions performed with these compounds [51]. To overcome this problem the aforementioned procedure was changed as follows: Instead of NaBH (OAc) 3 NaBH 3 CN was employed for the reduction of imines 10e-f to the amines 14e-f. Then, when the conversion to the amines 14e-f had gone to completion according to TLC, excess reducing agent was removed by basic-aqueous workup and the crude amines were reacted with NaBH(OAc) 3 and the aldehydes rac-12-13 in analogy to the original procedure. That way, the remaining esters rac-19e-f and rac-19l-m could finally be obtained in yields of 34-37% (Table 2, entries 5-6 and 11-12). Basic hydrolysis of the esters rac-19a-m with Ba (OH) 2 • 8 H 2 O according to a literature procedure [33] provided finally the desired carboxylic acids rac-11a-m in moderate to excellent yields (53-98%).

Biological evaluation
For the evaluation of the inhibitory potencies of the nipecotic acid derivatives rac-11a-m exhibiting a free carboxylic acid function and a fully established lipophilic domain, as well as of rac-18a-f possessing only small Nsubstituents and their corresponding esters rac-15a-f and rac-19a-m at the different GAT subtypes mGAT1-mGAT4 a standardized [ 3 H]GABA uptake assay was used [55]. HEK293 cell lines, each stably expressing one individual subtype of the GATs, represent the basis of this assay. Additionally, with a MS Binding Assay the binding affinities towards mGAT1 were determined using NO711 as native MS marker. If the tested compounds did not reduce the [ 3 H]GABA uptake or NO711 marker binding significantly below 50% in preliminary experiments at a concentration of 100 µM, which corresponds to a pIC 50 of ≤ 4.0 and a pK i of ≤ 4.0 respectively, only percent values of the remaining [ 3 H]GABA uptake or NO711 marker binding are given. In case of a significant reduction of the [ 3 H]GABA uptake or NO711 marker binding below 50% at an inhibitor concentration of 100 µM, the inhibitory potency (pIC 50 ) and the binding affinity (pK i ), respectively, were determined in a single experiment performed in triplicates.
As Tiagabine (6), NO711 (7), (S)-SNAP-5114 (8), or Deramciclane (rac-9) represent prototypic GAT inhibitors, they provide important reference values for the estimation of the biological activities of the newly synthesized and tested compounds described in this paper, despite the marked differences in their chemical structures. When considering the values of these reference compounds (see Fig. 2), it must be noted that these were partially obtained for enantiomerically pure [Tiagabine, (S)-SNAP-5114] or achiral (NO711) GAT inhibitors, whereas the substances displayed in this work are racemic mixtures.
The initially tested nipecotic acid esters rac-15a-f, that had been synthesized to serve as synthetic intermediates for the introduction of the tricyclic cage unit, and the corresponding carboxylic acids rac-18a-f displayed only very weak to negligible inhibitory potency and affinity. Only the dimethoxy substituted nipecotic acid derivatives rac-18e and rac-18f showed weak inhibitory potency at mGAT1 the remaining [ 3 H]GABA uptake amounting to 50% and 46%, respectively, at a test compound concentration of 100 µM. In addition, these compounds displayed inhibitory potency at mGAT3 and mGAT4, though this was even lower than that at mGAT1 with values for the remaining [ 3 H]GABA uptake in the range of 61-75% (Table 3, entries 10 and 12).
Due to their structural similarity it seemed appropriate to compare the test results for the synthesized carboxylic acids rac-11a-m and carboxylic acid esters rac-19a-m exhibiting a tricyclic residue as lipophilic domain among each other as this should provide insight on the influence of the spacer length (n), the bridge size (m) and the residues (R) on the biological activity. The comparison of test results of carboxylic acids of identical structure varying only in their spacer lengths (n = 0 or n = 1) among each other showed no significant impact of the spacer length on the biological activity for most structures. Only for the two nipecotic acid derivatives rac-11g and rac-11k with a butyl spacer improved inhibitory potencies were observed compared to their analogs with a propyl spacer rac-11a and rac-11d. For compound rac-11g a pIC 50 of 4.25 at mGAT1 was determined, whereas the structurally related carboxylic acid rac-11a with a propyl spacer could only reduce the [ 3 H]GABA uptake to 66%. Even more pronounced was the effect for carboxylic acid rac-11k for which a pIC 50 of 4.40 at mGAT1 and a remaining [ 3 H]GABA uptake of 45% at mGAT2 was found. The corresponding nipecotic acid derivate rac-11d with a propyl spacer merely reduced the [ 3 H]GABA uptake to 66% at mGAT1 and to 79% at mGAT2.
A comparative analysis of the biological activity of carboxylic acid esters rac-19a-m among each other to study the influence of the spacer length led to diverging results. For some esters of otherwise identical structure the variation of the spacer length did not seem to affect the results of the biological testing (compare: rac-19a and rac-19g; rac-19e and rac-19l). However, most nipecotic acid ester derivatives showed differences in the biological activity at the different GAT subtypes when the spacer length was altered. The carboxylic ester rac-19b substituted with phenyl residues and equipped with a methylene bridge (m = 0) and a propyl spacer (n = 0) exhibited higher inhibitory potencies at mGAT2 and mGAT3 with pIC 50 values of 4.53 and 4.43, respectively, compared to its analog rac-19h with a butyl spacer. Yet this analog rac-19h displayed a higher inhibitory potency at mGAT4 with a pIC 50 value of 4.89, whereas the potencies at mGAT1 were almost identical. Nipecotic acid ester derivative rac-19c with a C 3spacer reached lower remaining [ 3 H]GABA uptake with values ranging from 50 to 60% at mGAT2-mGAT4 as compared to its structural analog rac-19j with a C 4 -spacer. For the phenyl substituted ester rac-19k with a butyl spacer (n = 1) and a C 2 -bridge (m = 1) at mGAT1-mGAT3 inhibitory potencies with pIC 50 values ranging from 4.60 to 4.65 were observed, whereas the related ester rac-19d with a propyl spacer proofed to be less biologically active at mGAT1-mGAT3 and to have an identical activity at mGAT4. Finally, compound rac-19f displaying phenyl residues, a C 3 -bridge (m = 2) and a propyl spacer (n = 0) had a considerably higher activity at mGAT2 and mGAT3 with pIC 50 values of 4.28 and 4.97, respectively, but also a lower one at mGAT4 as compared to the analogous ester rac-19m with a butyl spacer, who had pIC 50 of 4.33 at mGAT4. Unfortunately, these results did not indicate a universal trend for the inhibitory potency at mGAT1-mGAT4 when the spacer length was altered.
Further analysis of the biological activity of carboxylic acids rac-11a-m by comparing structures deviating only in their attached residues R 1 , being either methyl or phenyl residues, showed that for most of the carboxylic acids the residue had a very small to negligible effect on the inhibitory potency at mGAT1-mGAT4 (compare rac-11a and rac-11b; rac-11c and rac-11d; rac-11e and rac-11f; rac-11l and rac-11m). Exceptions are the methyl-substituted nipecotic acid derivative rac-11g with a butyl spacer (n = 1) and a methylene bridge (m = 0), which had an improved inhibitory potency at mGAT1 with a pIC 50 of 4.25 as compared to its phenyl substituted analog rac-11h, and the phenyl substituted nipecotic acid derivative rac-11k with a butyl spacer (n = 1) and a C 2 -bridge (m = 1), that had a higher biological activity at mGAT1 and mGAT2 with a pIC 50 of 4.40 and a remaining [ 3 H]GABA uptake of 45%, respectively, as compared to its related methyl-substituted carboxylic acid rac-11j.
When taking a look at the carboxylic acid esters rac-19a-m it became evident that almost always the phenyl substituted esters had higher inhibitory potencies at mGAT1-mGAT4 than their otherwise identical methylsubstituted analogs. This observation is nicely highlighted by ester rac-19k with pIC 50 values in a range of 4.60-4.65 at mGAT1-mGAT4, which are in strong contrast to the biological results obtained for the related, basically inactive methyl-substituted ester rac-19j. Obviously, the aromatic phenyl residue in the nipecotic acid ester derived GAT inhibitors seems to be necessary as structural element to achieve a reasonable activity at all GAT subtypes.
The examination of the influence of the bridge size (m) on the biological activity of the carboxylic acids rac-11a-m at mGAT1-mGAT4 led to contradictory results. For the methyl-substituted nipecotic acid derivatives rac-11a, rac-11c, and rac-11e with a propyl spacer (n = 0) no significant effect of the bridge size on the inhibitory potencies at mGAT1-mGAT4 could be observed. Also, the methylsubstituted carboxylic acids rac-11g, rac-11j, and rac-11l with a butyl spacer (n = 1) showed similar biological activities at mGAT2-mGAT4 despite their varying bridge size (m = 0-2) and only at mGAT1 a preference for the carboxylic acid rac-11g with the smallest bridge size (m = 0), for which a pIC 50 of 4.25 was found, could be noticed. The comparison of the phenyl substituted carboxylic acids rac-11b, rac-11d, and rac-11f with a propyl spacer (n = 0) among each other indicated a preference for structures with smaller bridge sizes with regard to biological activity at mGAT1 and mGAT3 as the remaining [ 3 H]GABA uptake declined from 82 to 52% at mGAT1 and from 90 to 67% at mGAT3 with decreasing bridge size. For these structures at mGAT4 no influence of the bridge size on the biological activity was observed and at mGAT2 only a weak preference for carboxylic acid rac-11f with a C 3 -bridge was recognized. The comparative analysis of the phenyl substituted carboxylic acids rac-11h, rac-11k, and rac-11m with a butyl spacer (n = 1), in contrast, showed a preference for the medium-sized bridge (m = 1) for the biological activity at mGAT1-mGAT3, as the best inhibitory potencies with a pIC 50 value of 4.40 at mGAT1 and remaining [ 3 H]GABA uptakes of 45-58% at mGAT2-mGAT3 were determined for carboxylic acid rac-11k.
In addition, the influence of the bridge size (m) on the biological activity was studied for the carboxylic acid esters rac-19a-m. When the esters rac-19a, rac-19c, and rac-19e, all equipped with methyl residues and a propyl spacer (n =  a All values were determined in one experiment performed in triplicate. The results of the MS binding assay are given as pK i , the results of the [ 3 H] GABA uptake assay as pIC 50 . Percent values indicate remaining specific NO711 binding or remaining [ 3 H]GABA uptake, respectively, in presence of 100 µM test compound 0), were compared among each other, only for the inhibitory potency at mGAT4 the bridge size seemed to be important to some extent. Here ester rac-19c with a medium-sized C 2bridge (m = 1) reducing the remaining [ 3 H]GABA uptake to 54% at a test compound concentration of 100 µM proofed to be best. Also, the biological activity of the methylsubstituted esters rac-19g, rac-19j, and rac-19l with a butyl spacer (n = 1) at mGAT2-mGAT4 appeared to be rather unaffected by the bridge size of these compounds. Solely, according to the results of the inhibitory potencies at mGAT1, a methylene bridge (m = 0) mediates a slightly higher potency at this GAT subtype (see ester rac-19g). The comparative analysis of the test results of the phenyl substituted esters rac-19b, rac-19d, and rac-19f with a C 3spacer (n = 0) showed, that ester rac-19f with the largest bridge size (m = 2) turned out best to address mGAT3, with a pIC 50 value of 4.97, whereas, in order to address mGAT2, ester rac-19b with the smallest bridge size (m = 0) led to the best result (pIC 50 of 4.53). The esters rac-19b, rac-19d with a small or medium-sized bridge were equally suited to address mGAT4. As the esters rac-19b, rac-19d, and rac-19f displayed almost equal inhibitory potencies at mGAT1, no effect of the bridge size on the biological activity at this GAT subtype could be noticed. Finally, the structurally related phenyl substituted esters rac-19h, rac-19k, and rac-19m with a butyl spacer (n = 1) were compared among each other to study the influence of the bridge size on the biological activity for these compounds. Ester rac-19k with a medium-sized bridge (m = 1) demonstrated to be superior as compared to esters rac-19h and rac-19m with regard to inhibitory activities at mGAT1-mGAT3. Since at mGAT4 the inhibitory potency of esters rac-19h, rac-19k, and rac-19m was decreasing with an increase in bridge size, the ester rac-19h led with a pIC 50 value of 4.89 to the best result. However, by the above obtained results no general correlation between the biological activity at a certain GAT subtype and the bridge size (m) in the lipophilic domain of the tested carboxylic acids rac-11a-m or their corresponding esters rac-19a-m could be concluded. Interestingly, all phenyl substituted nipecotic acid ester derivatives, i.e., rac-19b, rac-19d, rac-19f, rac-19h, rac-19k, and rac-19m exhibited higher inhibitory potencies at mGAT1-mGAT4 than their corresponding carboxylic acids. For the methyl-substituted nipecotic acid ester derivatives no such universal effect was observed. The former phenyl substituted nipecotic acid derivatives showed rather equal inhibitory potencies at all four GAT subtypes ( Table  4, see entries for compounds rac-19b, rac-19d, rac-19k, and rac-19m), but also a weak subtype selectivity for mGAT3 and for mGAT4 was achieved with ester rac-19f (pIC 50 value of 4.97 at mGAT3; Table 4, entry 11) and ester rac-19h (pIC 50 value of 4.89 at mGAT4; Table 4, entry 15), respectively. These esters, rac-19f and rac-19h, represent the first subtype selective GAT inhibitors carrying a cage unit in the lipophilic domain.
Still to be mentioned is the fact, that the binding affinities at mGAT1 determined in binding assays often do not correlate with the inhibitory potencies from mGAT1 uptake assays. This phenomenon, the cause of which is still to be clarified, can be seen for example in case of ester rac-19k. This compound, rac-19k, exhibits a pIC 50 value of 4.60 at mGAT1 in the uptake assay, but a reduction of remaining NO711 marker binding in the binding assay to 84% only (at a test compound concentration of 100 µM).

Conclusion
Inspired by the drug Deramciclane (rac-9), a new class of GABA uptake inhibitors with bulky and highly rigid tricyclic subunits in the lipophilic domain delineated from the 2-azabicyclo[2.2.2]octane scaffold by the presence of an additional carbon bridge was developed. The polycyclic subunits are connected via a plain hydrocarbon spacer with the amino nitrogen of nipecotic acid or that of the corresponding ethyl ester. For the synthesis of the new compounds, nipecotic acid derivates with an N-alkyl residue displaying a terminal aldehyde function, were connected with symmetric tricyclic amines by reductive amination. The tricyclic amines used were either generated in situ from tricyclic imines serving as precursors directly before the reductive amination by the same reducing agent or they were generated from the tricyclic imines in a separate reaction step. The new GAT inhibitors varied in regard to the spacer length, the size of one of the bridges of the tricyclic skeleton of the lipophilic domain and the substituents attached to the latter. Whereas the nipecotic acid derived GAT inhibitors displayed only weak inhibitory potencies and binding affinities at the four different GAT subtypes, all phenyl substituted nipecotic acid ethyl ester derivatives exhibited moderate biological activity at mGAT1-mGAT4. The structure activity relationship of these GAT inhibitors demonstrated the importance of the phenyl residues and the ester function for the biological activity. Two of the phenyl substituted nipecotic acid ethyl ester derivates, rac-19f and rac-19h, being equipped with either a propyl spacer and a C 3 -bridge (rac-19f) or a butyl spacer and a methylene bridge (rac-19h), showed even moderate subtype selectivity at mGAT3 and mGAT4 respectively. As demonstrated by the obtained results tricyclic cage structures represent promising subunits for the construction of novel GAT inhibitors.

Compliance with ethical standards
Conflict of interest The authors declare that they have no conflict of interest.
Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons. org/licenses/by/4.0/.