Diastereoselective reduction of cyclic bioactive Mannich ketones

The reduction of cyclic Mannich ketones having antibacterial activity showed poor stereoselectivity with L-Selectride ® , sodium trimethoxyborohydride and diisobutylaluminium hydride, while lithium aluminium hydride and, in particular, sodium borohydride often yielded a single stereoisomer. The size of the ring strongly influenced the stereocomposition of the reaction mixtures. An increased preference for the trans isomer was attributed to a weak intramolecular hydrogen bond between the OH and the N, as demonstrated by X-ray crystallography. After reduction, the antibacterial activity of the product decreased dramatically (“S-form”) but was not completely abolished in deep rough mutants of Gram-negative bacteria.


Introduction
Studies on the mechanism of action for biologically active compounds often have been complex endeavours, with synthetic organic and medicinal chemistry often playing a crucial role in providing invaluable insight into the intricate and previously unrecognised details.Previously we reported the synthesis of some Mannich ketones of 2-arylidenecycloalkanones 1 1-9 (Scheme 1).Similar derivatives of fused bicyclic ketones such as 1-indanones and 1-tetralones 10-11 also were prepared 2 (Scheme 1).The majority of these compounds showed antibacterial activity both against standard Gram-positive (Staphylococcus saprophyticus, S. aureus, Micrococcus luteus, Bacillus subtilis) and a Gram-negative strain such as Escherichia coli.Such compounds have been considered thiol alkylators because under physiological conditions they give reactive methylene ketones (by 1,2-elimination) that can undergo addition reactions with thiol enzymes. 3ur previous examinations suggested however that the thiol depletion cannot be the only reason for the antibacterial activity. 1,2We showed that some compounds were highly active against bacteria, but caused no thiol depletion, others were not active although the free thiol pool decreased significantly when the compounds were administered at the minimum inhibitory concentration (MIC).

Ar
In order to investigate the importance of thiol reactivity in the mechanism of action, we prepared the corresponding amino alcohols from the Mannich ketones mentioned above.Unlike the Mannich ketones, the amino alcohols cannot yield very reactive methylene ketones.
First, we wanted to investigate the stereoselectivity of the reduction using different complex metal hydrides with organic groups.Sodium borohydride is a generally more selective agent for preparing allylic alcohols than lithium aluminium hydride. 4The hydride reduction of α,β-unsaturated Mannich ketones may, however, afford 1,2-or 1,4-addition products.We also aimed at studying the stereocomposition of reduction products depending on the structure of the substrate (size of the ketone ring and the nature of the amine substituent).The starting materials for the reductions were selected as both effective antibacterial agents and less effective ones (cf. 1, 3, 10 and 6-9).We also planned to analyse the effect of reduction on the antibacterial activity of our Mannich ketones.The antibacterial activity and the thiol depletion were examined in two E. coli strains, in an "S" type (ATCC 25292) and in a Re mutant one which has a defective cell wall biosynthesis.The Re mutants are therefore more permeable to some externally added chemicals.

Results and Discussion
In order to investigate the influence of the reducing agent on the stereocomposition, we explored the reduction of 3 with different complex metal hydrides.These reactions started from the Mannich base liberated from the corresponding hydrochloride 3 under mild conditions (Scheme 2).The complex hydrides applied for the reduction of 3 can be divided into two groups.The first one (L-Selectride ® , sodium trimethoxyborohydride, diisobutylaluminium hydride) represents reducing agents showing poor stereoselectivity, while the hydrides of the other group, such as lithium aluminium hydride, produced stereoselective reductions affording the trans isomer as a major isomer.In particular, reduction performed with sodium borohydride proved to be rather stereoselective, yielding mainly one stereoisomer (Scheme 2).

Scheme 2
We hypothesized that the relative thermodynamic stability of the respective hydride adducts (cis-14a-e and trans-14a-e) should explain the observed selectivity of the metal-hydride reduction of the Mannich base chosen.To prove this hypothesis, semi-empirical (PM5) quantum chemical calculations were performed to obtain heats of formation (∆H f °) for the hydride adducts (Scheme 3).As shown in Table 1, the relative stabilities of the isomeric cis-14a-e and trans-14a-e (expressed as ∆H f °cis -∆H f °trans ) correlated well with the observed stereocomposition of the end product 14, although a bias to yield trans-14 during the reduction also was apparent.Because of its propensity to produce the highest stereoselectivity based on the preliminary experiments, sodium borohydride was chosen as the reducing agent to study the influence of the substrate on the stereocomposition of the reaction mixture.The reductions were performed in methanolic solutions at room temperature.In all of the starting unsaturated Mannich ketones the configuration was E around the C=C bond.From the hydrochlorides of the Mannich ketones (1-9 and 10-11) the bases were liberated with equivalent NaOH at 0 °C and they were treated with NaBH 4 (Schemes 4-5).
The composition of the reaction mixture was monitored by 1 H NMR spectroscopy.The diastereoisomers were separated by column chromatography when it was possible.The conversion was always 100 %.The size of the ring strongly influenced the stereocomposition of the reaction mixtures giving only the trans isomers with the five membered and only the cis with the eight membered rings.Probably, the trans diastereoisomer is stabilized by the intramolecular hydrogen bond between OH and the nitrogen atom.
Cis and trans indicate the orientation of bonds C1-H1 and C2-H2 as well as of bonds C1-H1 and C7(8)-H7(8) each to other in compounds 12-20 and 21-22, respectively.The structure verification of our compounds was performed by FT-IR, 1 H-NMR, and 13 C-NMR spectroscopy (Fig. 1) * * and, in the case of trans-14, by crystal structure analysis (Table 2).The molecule shows an intramolecular H-bond O2-H4⋅⋅⋅N1 with an H4-N1 distance of 1.955 Å and an angle of 151.All distances and angles are within the experimental errors in the expected range. 5The C1-C6 as well as the hetero 6-ring are in chair conformations.Ring C1-C6 is slightly distorted from the ideal conformation due to the sp 2 hybridisation of C1.The conformation about the C1-C7 double bound is E (trans).The interaction between the molecules in the crystal is only of the van der Waals type.
The new compounds were screened against some standard Gram-positive and Gram-negative strains, among them some Re deep rough mutants strains (Table 3).The reduction dramatically decreased the antibacterial activity of the original Mannich ketones both against the Grampositive and -negative strains (e.g. 3 S. aureus MIC: 1.56 µg mL -1 ; cis-14 MIC: >200 µg mL -1 ). 1,2However, some of the compounds (trans-13, cis-17, trans-18, trans-20) showed slight activity to Grampositive strains.With the deep mutant strains, some of the amino alcohols tested (cis-14, trans-14, trans-16, cis-17, cis-19 and cis-20) showed moderate activity.The comparison of the antibacterial activity of Mannich ketones [1][2] and the corresponding amino alcohols proves that the keto group is required for the biological effect in most of the compounds.The slightly increased polarity of the compounds cannot explain their diminished antibacterial activity.The decreased antibacterial effect of the amino alcohols can partly support the thiol alkylation as a mechanism ISSN 1424-6376 Page 40 © ARKAT USA, Inc of action.However, the existing but weak antibacterial effect on M. luteus (trans-13, cis-20), S. saprophyticus (cis-20) and Re strains probably can be based on another mechanism of action.

Figure 2.
Projection of trans-14 with anisotropic displacement ellipsoids (50% probability) for non-H atoms and an arbitrary small displacement parameter for H-atoms. 13close relationship can be deduced from our data between the permeability and the antibacterial activity of our compounds.We determined the MIC values for "S" form E. coli ATCC 25292 and "Re" form E. coli ReD31m4, Salmonella minnesota Re595, Shigella sonnei Re4350.7][8] Thus cis-14, trans-14, cis-17, cis-19 and cis-20, that are only mildly active or non active to "S" form, can kill the Re mutant Gram-negative bacteria.
Determination of the free intracellular thiol content of the "S" and Re mutant E. coli treated with these compounds at MIC concentration or at 200 µg mL -1 revealed that reduction completely abolished the thiol reactivity (Figs. 3 and 4).This supported our idea, that in some of these compounds the antibacterial activity cannot be explained by a thiol alkylation.Human cytotoxicity of our amino alcohols was evaluated against HeLa cell lines (Table 4).Similarly to the antibacterial effect, the cytotoxicity for most of the compounds diminished by the reduction, as demonstrated in the case of trans-13, trans-22, cis-14 and cis-20.On the contrary, the cytotoxicity of cis-17 increased and it is the most toxic in this series.It is worth mentioning that the cis-14 and trans-14 stereoisomers have very different IC 50 values, but very similar antibacterial activity.

Conclusions
In conclusion, reduction of 3 with various complex hydrides afforded reaction mixtures of very different stereocomposition.Reduction with sodium borohydride was found to be diasteroselective both for α,β-unsaturatedand fused Mannich ketones.The stereocomposition of the reaction mixtures was influenced by the size of the ring.The predominance of the trans isomers as trans-14 can be explained by the intramolecular hydrogen bond between the OH and the N as demonstrated by crystal structure analysis.Most of the amino alcohols almost completely lost their biological effect.Two compounds, however, retained or even increased their antibacterial activity on deep rough mutants (9/cis-20 MIC: from 200 to 50 µg mL -1 ) and thus emerged as lead compounds for antibacterial agents with a unique, yet currently unknown, activity profile.Since the amino alcohols did not cause thiol depletion, this finding points to a new mechanism of action other than alkylation.The MIC values for deep rough mutants show that the permeability of the outer membrane definitely can influence the antibacterial activity of our compounds.The cytotoxic activities of the amino alcohols also were evaluated using HeLa ISSN 1551-7004 Page 45 © ARKAT USA, Inc cells.Most of the reduced derivatives proved to be less toxic to the HeLa cells than the starting Mannich ketones, consistent with data obtained with bacteria.The effect of two amino alcohols, however, was significantly different in HeLa cells and in bacteria.Cis-17 increased its toxicity in HeLa but was not active against bacteria, cis-20 became nontoxic in HeLa but displayed antibacterial activity.

Experimental Section
General Procedures.The diethyl ether used for the reductions was dried by distillation 12 from Na under Ar.Our starting Mannich ketones (1-9 and 10-11) are known compounds synthesised according to our methods 1,2 or literature methods. 9In all of the unsaturated Mannich ketones the configuration is E around the C=C double bond. 1  Å 3 (determined from 17676 reflections, 3 º < θ <30 º ), Z = 2, d calc = 1.204 g cm -3 and µ x (λ = 0.49592 Å) = 0.18 cm -1 .Reflections were recorded at 100 K using a CCD-detector (MarCCD 165) and monochromatic synchrotron radiation (λ = 0.49592 Å).Reflections were indexed, integrated, corrected and scaled using XDS. 10 The same program was used to refine the lattice parameters.
A long and a short exposure time series each with 360 º of total oscillation range and with 0.5 º per frame were recorded.Data were integrated for -11 ≤ h ≤ 11, -20 ≤ k ≤ 21 and -29 ≤ l ≤ 29 with (sinθ/λ) max = 1.04.Total number of observations was 87736 with 13034 unique (R merge = 2.1%).All unique reflections were used for further analysis.The structure was solved by direct methods 11 .All non-H atoms were refined anisotropically using full matrix least-square 12 based on F 2 with weights 1/[σ 2 (F o 2 )+(0.0633*P) 2 +0.07*P]where P = (Max(F o 2 ,0)+2*F c 2 )/3.For each H-atom the co-ordinates and an isotropic displacement parameter were refined.The shifts (∆/σ) max of the final least square cycle were smaller than 0.002.R-values for F 2 > 2σ(F 2 ) were R = 0.0393 (based on |F|) and R w = 0.1159 (based on F 2 ).Goodness of fit was 1.074.The final difference Fourier was featureless with the highest maxima at the positions of the binding electrons (∆ρ min = -0.32,∆ρ max = 0.67).All relevant crystallographic data have been deposited with the Cambridge Crystallographic Data Center under the number CCDC 197256.Theoretical calculations.CAChe Worksystem Pro (version 5.04, Fujitsu America, Inc., Beaverton, OR), a personal computer-based molecular modeling program, was used in the computational studies.Geometry-optimized molecular models (in vacuo, gradient norm below 0.1 kcal/Å as the criterion for convergence) and ∆H f ° values for cis/trans-3a-e were obtained by a semi-empirical quantum mechanical method (PM5 parameters).Biological procedure.Determination of antibacterial activity.MIC values were determined by the broth macrodilution method 14 as described previously. 1 The water soluble hydrochlorides of the aminoalcohols were screened.The test bacteria were Pseudomonas aeruginosa NIH Hungary 170000, E. coli ATCC 25922, E. coli ReD31m4, S. minnesota Re595, S. sonnei Re4350, S. saprophyticus NIH Hungary 120008, S. aureus NIH Hungary 118003, M. luteus ATCC 9341 and B. subtilis ATCC 6633.The tube dilution method started with making twofold dilution series of test compounds in nutrient broth.The test bacteria were in the logarithmic phase; their final concentration was ca.5x10 5 colony-forming units mL -1 .The MIC values were determined after one night incubation (37 °C) as the lowest concentration of compounds producing no visible bacterial growth.All experiments were performed in triplicate.In vitro effect of Mannich ketones on HeLa cell line. 1 The HeLa cells were cultured in RPMI-1640 culture medium containing 10% heat-inactivated fetal calf serum with antibiotics streptomycin (25 µg mL -1 ) and penicillin (50 µg mL -1 ).The cells were grown as a monolayer, in the logarithmic phase, had viability better than 98 % as proved by Tripan blue exclusion.The growing cells were plated out in 100 µL of medium at concentration of 4-5 x 10 3 cells/flatbottomed well in 96-well microtiter plates.Plates were incubated for 24 hours at 37 °C in a humidified atmosphere of 5% CO 2 .100 µL of media containing the drug dissolved in appropriate solvent were added to each well and incubated for a further 48 hours.MTT was filtered through a 0.22 µm filter and diluted in RPMI-1640 containing antibiotics to 1 mg mL -1 .100 µl of this solution added to each well and incubated for 4 hours.All untransformed MTT (medium) was removed by careful aspiration from the wells.The formazan crystals were dissolved in 100 µL of isopropanol-1 N HCl (24/1 v/v).Then the plate was shaken in order to ensure solubilization of the blue formazan crystals.The absorbance was recorded in an enzyme-linked immunosorbent assay plate reader at a 560 nm test wavelength and a 690 nm reference wavelength.Determination of free thiol content.The method was published previously. 1 Shortly, the bacterial cells were cultivated in a shaker incubator (37 °C) to the mid log phase and treated with the different compounds at the MIC concentration or at 200 µg ml -1 in cases of ineffective substances (MIC > 200 µg mL -1 ) for one hour.The treated cells were centrifuged (+4 °C), washed with physiological saline solution and stored at -80 °C as pellets.These cell pellets were resuspended in a buffer containing 50 mM TRIS-HCl, 0.1 % SDS of pH 7.4, and lysed by sonication.The cell lysates were clarified by centrifugation and used immediately for the determination of free thiol by the DTNB assay. 15The protein concentration was determined by the BCA (bicinchoninic acid) assay. 16The assay was repeated using eight selected compounds in three parallel cultures.Values represent the average of three independent determinations ± SD.

Table 3 .
Comparison of the in vitro antibacterial activity of the aminoalcohols (12-22 in the first line, in italics) and starting Mannich ketones (1-11 in the second line), expressed as minimum inhibitory concentration values (MIC, µg mL -1 )