Dibasic Derivatives of Phenylcarbamic Acid against Mycobacterial Strains: Old Drugs and New Tricks?

In order to provide a more detailed view on the structure⁻antimycobacterial activity relationship (SAR) of phenylcarbamic acid derivatives containing two centers of protonation, 1-[2-[({[2-/3-(alkoxy)phenyl]amino}carbonyl)oxy]-3-(dipropylammonio)propyl]pyrrolidinium oxalates (1a⁻d)/dichlorides (1e⁻h) as well as 1-[2-[({[2-/3-(alkoxy)phenyl]amino}carbonyl)oxy]-3-(di-propylammonio)propyl]azepanium oxalates (1i⁻l)/dichlorides (1m⁻p; alkoxy = butoxy to heptyloxy) were physicochemically characterized by estimation of their surface tension (γ; Traube's stalagmometric method), electronic features (log ε; UV/Vis spectrophotometry) and lipophilic properties (log kw; isocratic RP-HPLC) as well. The experimental log kw dataset was studied together with computational logarithms of partition coefficients (log P) generated by various methods based mainly on atomic or combined atomic and fragmental principles. Similarities and differences between the experimental and in silico lipophilicity descriptors were analyzed by unscaled principal component analysis (PCA). The in vitro activity of compounds 1a⁻p was inspected against Mycobacterium tuberculosis CNCTC My 331/88 (identical with H37Rv and ATCC 2794, respectively), M. tuberculosis H37Ra ATCC 25177, M. kansasii CNCTC My 235/80 (identical with ATCC 12478), the M. kansasii 6509/96 clinical isolate, M. kansasii DSM 44162, M. avium CNCTC My 330/80 (identical with ATCC 25291), M. smegmatis ATCC 700084 and M. marinum CAMP 5644, respectively. In vitro susceptibility of the mycobacteria to reference drugs isoniazid, ethambutol, ofloxacin or ciprofloxacin was tested as well. A very unique aspect of the research was that many compounds from the set 1a⁻p were highly efficient almost against all tested mycobacteria. The most promising derivatives showed MIC values varied from 1.9 μM to 8 μM, which were lower compared to those of used standards, especially if concerning ability to fight M. tuberculosis H37Ra ATCC 25177, M. kansasii DSM 44162 or M. avium CNCTC My 330/80. Current in vitro biological assays and systematic SAR studies based on PCA approach as well as fitting procedures, which were supported by relevant statistical descriptors, proved that the compounds 1a⁻p represented a very promising molecular framework for development of 'non-traditional' but effective antimycobacterial agents.

1. Introduction

The treatment of commonly encountered species of tuberculous and non-tuberculous mycobacteria responsible for a multiplicity of different typn>es of pan class="Disease">infections, including pulmonary, respiratory, cutaneous, and systemic infections, by (i) brand new classes of promising compounds preferably acting on novel targets; or (ii) ‘non-typical’ antimycobacterial drug candidates is still in very dynamic and progressive debate [1,2,3,4,5,6]. The strategy was successfully used in a case of in vitro screening of some β-lactam antibiotics (ceftaroline or ceftazidime) in a combination with an β-lactamase inhibitor avibactam against Mycobacterium avium complex [3,4]. Another encouraged example was in vitro and ex vivo testing of tricyclic thioridazine (Figure 1a), an old neuroleptic phenothiazine. The molecule was used alone or in a combinatorial therapy with anti-tuberculosis drugs (isoniazid, rifampin, linezolid or moxifloxacin) against M. tuberculosis regardless of its antibiotic resistance phenotype and was highly active due to its multi mechanisms of action [5,6].

Figure 1

Chemical structure of thioridazine (a), a neuroleptic drug containing a phenothiazine scaffold (a); as well as effective local anesthetics (b) lidocaine (xylocaine); and (c) dibucaine (cinchocaine). All these compounds showed notable in vitro efficiency against some mycobacterial strains [5,6,7,8,9,10,11,12]. Structural fragments of dibucaine were color-coded as follows: a lipophilic moiety (a dark green framework), polar group (orange), connecting hydrocarbon chain (gray) and salt-forming (basic) group (dark red), respectively.

Antimycobacterial properties of compounds originally designed and tested as local anesthetics (LAs) have been investigated sporadically. Moreover, their application to management of the infections caused by a broad spectrum of species of the pan class="Species">Mycobacterium genus has not been systematically reviewed. Nevertheless, Schmidt and Rosenkranz [7] as well as Fujii et al. [8] found that lidocaine (xylocaine; Figure 1b), a highly efficient LA drug [9,10], notably inhibited in vitro growth of some strains of human M. tuberculosis, M. bovis or M. kansasii. Similarly, a quinoline core-containing dibucaine (cinchocaine; Figure 1c) [11], was able to effectively in vitro fight rapidly growing M. phlei [12].

A common structural feature of given efficient LAs (and promising antimycobacterials) was presence of a lipophilic moiety, polar anilido or amido groupn>, con>an class="Chemical">nnecting hydrocarbon chain and salt-forming fragment. All considered moieties were color-differentiated for a molecule of dibucaine (cinchocaine; Figure 1c).

Sequencing of Mycobacterium spp. genomes proved that these microorganisms produced a variety of enzymes [13,14,15] able to hydrolyze pan class="Chemical">ester, anilide or amide bonds in a chemical structure of antimycobacterial drug candidates. Isosteric replacement of these polar groups with a carbamate moiety would provide very good proteolytic stability, especially if the modified structural arrangement of novel derivatives would be as follows: Aryl–NH–C(O)O–Alkyl [16]. The carbamate functionality imposes a degree of conformational restriction due to delocalization of non-bonded electrons on nitrogen into a carboxyl moiety. Participation of the carbamate group in hydrogen bonding through carboxyl and backbone N–H [17] should be also taken into account. Considering the importance of the carbamate functionality, its incorporation into a structure of novel antimycobacterial agents was one of key tasks in a process of their design and synthesis [18,19,20].

Currently evaluated 1-[2-[({[2-/3-(alkoxy)phenyl]amino}carbonyl)oxy]-3-(dipropylammonio)-propyl]pyrrolidinium/pan class="Chemical">azepanium oxalates/dichlorides (1a–p; alkoxy = butoxy to heptyloxy; Table 1, Scheme 1) were synthesized with intention to find effective LAs with favorable toxicity profiles [21]. The compounds contained two types of an anionic counterpart (1a–d and 1i–l versus 1e–h and 1m–p; Table 1) in order to achieve convenient aqueous solubility in biological assays [21].

Table 1

Chemical structure of evaluated compounds 1a–p, their surface tension γ (relative surface activity; in N/m units), wavelengths of observed absorption maxima (λ1, λ2 (Ch-T), λ3) and logarithms of molar absorption coefficients (log ε1, log ε2 (Ch-T), log ε3) of compounds’ methanolic solutions (c = 8.0 × 10−5 M), which were investigated in the UV/Vis region of an electromagnetic spectrum.

Comp.	R	X	Y	γ (N/m)	λ 1	log ε1	λ 2 (Ch-T)	1 log ε2 (Ch-T)	λ 3	log ε3
1a	2-OC4H9	A	C	0.06464	208	4.51	236	4.19	280	3.63
1b	2-OC5H11	A	C	0.06366	208	4.39	236	4.05	280	3.51
1c	2-OC6H13	A	C	0.06222	208	4.58	238	4.52	278	3.49
1d	2-OC7H15	A	C	0.05985	208	4.43	236	4.08	280	3.55
1e	3-OC4H9	A	D	0.06316	210	4.65	238	4.24	279	3.59
1f	3-OC5H11	A	D	0.06285	210	4.55	237	4.27	279	3.62
1g	3-OC6H13	A	D	0.06105	210	4.54	237	4.13	279	3.48
1h	3-OC7H15	A	D	0.05786	210	4.66	237	4.27	279	3.63
1i	2-OC4H9	B	C	0.06302	208	4.44	236	4.08	280	3.57
1j	2-OC5H11	B	C	0.06206	208	4.54	236	4.22	280	3.71
1k	2-OC6H13	B	C	0.06065	208	4.42	236	4.10	280	3.57
1l	2-OC7H15	B	C	0.05853	208	4.44	236	4.14	280	3.59
1m	3-OC4H9	B	D	0.06298	210	4.52	238	4.09	279	3.44
1n	3-OC5H11	B	D	0.06154	210	4.56	238	4.18	279	3.52
1o	3-OC6H13	B	D	0.05925	210	4.52	238	4.01	279	3.37
1p	3-OC7H15	B	D	0.05692	210	4.65	238	4.20	279	3.56

1 log ε2 (Ch-T), Logarithms of molar absorption coefficients observed at the charge-transfer absorption maximum λ2 (Ch-T) = 236–238 nm.

Scheme 1

Synthesis of 1-[2-[({[2-/3-(alkoxy)phenyl]amino}carbonyl)oxy]-3-(dipropylammonio)-propyl]pyrrolidinium/azepanium oxalates/dichlorides (1a–p; alkoxy = butoxy to heptyloxy). Reagents and conditions: (i) (CH3CO)2O, reflux (1h); (ii) Na, anhydrous EtOH, 1-bromoalkane (alkane = butane to heptane), r.t. (12h), reflux (3h); (iii) 18% HCl, reflux (2h); (iv) phosgene, anhydrous toluene, reflux (3h); (v) (±)-2-(chloromethyl)oxirane, 35 °C (2h), r.t. (48h), 38% NaOH; (vi) pyrrolidine/azepane, anhydrous 2-PrOH, reflux (6h); (vii) anhydrous toluene, reflux (8h); (viii) saturated solution of (a) oxalic acid in anhydrous EtOH or (b) ethereal hydrogen chloride.

It was found that all tested derivatives showed higher indices defining their relative surface (Us) as well as infiltration (Ui) local anesthetic efficiency than reference LA drugs cocaine (Us = 1.0, Ui = 3.6) or pan class="Chemical">procaine (Us = 0.1, Ui = 1.0) [21]. The most effective dibasic molecule in given types of local anesthesia was 1-[2-[({[3-(pentyloxy)phenyl]amino}carbonyl)oxy]-3-(dipropylammonio)propyl]-pyrrolidinium dichloride (1f) with Us = 130 and Ui = 250 (Table S1 in Supplementary Materials), respectively.

It might be expected that the carbamate moiety was sterically protected due to branching of both con>an class="Chemical">nnecting hydrocarbon chain and salt-forming group. Rotatable bonds of the 2-/3-alkoxy side chain R could also contribute to the steric hindrance (Table 1). It was proved by ab initio procedures that an intramolecular hydrogen bond between N–H of the carbamate functionality and oxygen atom of the 2-alkoxy chain was formed [22]. Such structural arrangements of compounds would provide stability against effects of various mycobacterial enzymes and would consider them promising candidates for antimycobacterial agents.

It was observed that surface tension (relative surface activity), electronic and lipophilic properties of pan class="Chemical">carbamate groupn>-containing molecules notably influenced the in vitro antimycobacterial efficiency [23,24,25]. However, the set 1a–p has not been satisfactorily characterized by values of relevant physicochemical descripn>tors [26] (Table S1).

In a first part of this study, attention was turned on more precise physicochemical characterization of the derivatives 1a–p. The research aimed determination of compounds’ ability to decrease surface tension of water (γ), estimation of electronic propn>erties described by logarithms of molar absorpn>tion coefficients (log ε) of their pan class="Chemical">methanolic solutions, which were investigated in the UV/Vis region of an electromagnetic spectrum, as well as evaluation of lipophilic features defined by calculated log kw parameters. These log kw were based on extrapolation procedures of estimated logarithms of retention factors (log k) to elution with 100% water by reversed-phase high-performance liquid chromatography (RP-HPLC). In addition, relationships between the log kw dataset and calculated logarithms of partition coefficients (log P) related to the octan-1-ol/water partitioning system were explored in order to provide a critical view on possibilities to predict lipophilic properties of the compounds based on two-dimensional (2D) visualization of their chemical structures.

Next, a very essential objective of the research was in vitro screening of the derivatives 1a–p against various strains of tuberculous and non-tuberculous mycobacteria, i.e., n>an class="Species">M. tuberculosis CNCTC My 331/88, M. tuberculosis H37Ra ATCC 25177, M. kansasii CNCTC My 235/80, a M. kansasii 6509/96 clinical isolate, M. kansasii DSM 44162, M. avium CNCTC My 330/80, M. smegmatis ATCC 700084 and M. marinum CAMP 5644, respectively.

A key aim of the study was to find some structural and physicochemical features of the compounds 1a–p, which might appear to be notable for their in vitro antimycobacterial efficiency.

2. Results and Discussion

2.1. Synthesis of the Compounds

The investigated compounds 1a–p were prepared by multi step pathways using 2-aminophenol (1′a) and 3-aminophenol (1′b), respectively, as starting molecules (Scheme 1). Very briefly, a reaction of 1′a or 1′b with acetanhydride led to N-(2-/3-hydroxyphenyl)ethanamide (2′a or 2′b). A solution of 2′a or 2′b in anhydrous pan class="Chemical">ethanol (EtOH) was added to sodium ethanoate. After mixing, 1-bromoalkane (alkane = butane to heptane) was added in order to prepare a series of N-(2-/3-alk-oxyphenyl)ethanamides (3′a–h; alkoxy = butoxy to heptyloxy). The molecules 3′a–h were suspended in 18% hydrochloric acid and heated up to reflux. The solutions were cooled, neutralized and crude intermediates were extracted into diethyl ether (DEE). The organic layer was dried, filtered and solutions were removed in vacuo giving 2-/3-alkoxyanilines (4′a–h; alkoxy = butoxy to heptyloxy) [27]. Reulting intermediates 4′a–h were dissolved in anhydrous toluene and added continuously into a saturated solution of phosgene [27]. The mixed solutions were refluxed, toluene was removed in vacuo providing desired 1-alkoxy-2-/3-isocyanatobenzenes (5′a–h).

Into an aqueous solution of N-propylpropanamine (6′), a (±)-2-(chloromethyl)oxirane reagent was added and solution was allowed to stand at room tempn>erature (r.t.; n>an class="Chemical">48h). The mixture was heated up to 75 °C and treated with 38% sodium hydroxide. The cooled solution was filtered and a crude intermediate was formed. Extraction of the filtrate with DEE, drying and removal of the organic layer provided additional amount of the intermediate [28], which isolation and crystallization from absolute EtOH led to (±)-N-(oxiran-2-ylmethyl)-N-propylpropanamine (7′).

The addition of pyrrolidine or n>an class="Chemical">azepane to 7′ in anhydrous propan-2-ol (2-PrOH) provided 1-(dipropylamino)-3-pyrrolidin-1-ylpropan-2-ol (8′a) or 1-azepan-1-yl-3-(dipropylamino)-propan-2-ol (8′b) [29].

Oily 1-(1-azacycloalkyl)-3-(dipn>ropn>ylamino)propn>an-2-yl (2-/3-n>an class="Chemical">alkoxyphenyl)carbamates (9′a–p; azacycloalkyl = pyrrolidinyl or azepanyl) were prepared by a reaction of 1-alkoxy-2-/3-isocyanato-benzenes 5′a–h with a dibasic alcohol 8′a or 8′b in anhydrous toluene.

Addition of a saturated solution of oxalic acid in anhydrous n>an class="Chemical">EtOH or ethereal hydrogen chloride to the bases 9′a–p dissolved in chloroform led to solid colorless 1-[2-[({[2-/3-(alkoxy)phenyl]- amino}carbonyl)oxy]-3-(dipropylammonio)propyl]pyrrolidinium oxalates (1a–d)/dichlorides (1e–h) and 1-[2-[({[2-/3-(alkoxy)phenyl]amino}carbonyl)oxy]-3-(dipropylammonio)propyl]azepanium oxalates (1i–l)/dichlorides (1m–p; Scheme 1, Table 1) as well. More detailed procedures describing syntheses of particular intermediates and final molecules were provided in Supplementary Materials.

Chemical structure of the compounds 9′a–p and 1a–p was previously verified by their infrared (IR) spectra, which confirmed presence of all key groups. In addition, elemental analyses results (% C, H, N) were within ±0.40% of theoretical values for both bases 9′a–p and pan class="Chemical">salts 1a–p [21]. Melting points (m.p.′s), Rf values (TLC) and acid-base pKa1 and pKa2 parameters of 1a–p were already published [21,26] and can be found in Table S1.

Current liquid chromatography high resolution mass spectroscopy (pan class="Disease">HPLC-HR-MS) analyses, which were performed on the LC Agilent Infinity System coupn>led with the Quadrupn>ole Time-Of-Flight Mass Spn>ectrometer (6520 Accurate Mass Q-TOF LC/MS), confirmed structural identity of the intermediates 9′a–p (Supn>plementary Materials).

After re-crystallization from a mixture of acetone/pan class="Chemical">ethanol (subgroups 1a–d, 1e–h and 1m–p) or acetone (1i–l), high-resolution mass spectra (HR-MS) of these salts were measured by the Dionex UltiMate 3000 High-Performance Liquid Chromatograph coupled with the LTQ Orbitrap XL Hybrid Ion Trap-Orbitrap Fourier Transform Mass Spectrometer equipped with a HESI II (heated electrospray ionization) source in a positive (1a–d, 1i–l) or negative (1e–h, 1m–p) mode. The analyses confirmed structural identity of given molecules.

2.2. Determination and Prediction of Some Physicochemical Properties of the Compounds

2.2.1. Surface Tension

Surface tension (relative surface activity; γ) of aqueous solutions of the compounds 1a–p (c = 2.0 × 10−3 M) was determined by a drop count technique [30,31] using a Traube stalagmometer. All evaluated substances were able to decrease surface tension of pan class="Chemical">water (γ = 0.07259 N/m) at 21 °C.

Elongation of the 2-/3-alkoxy side chain R led to more surface-active derivatives within homological sets 1a–d, 1e–h, 1i–l and 1m–p. This behavior was also in agreement with conclusions of a research papn>er [32]. The 2-n>an class="Chemical">alkoxy substituted molecules 1a–d and 1i–l showed lower ability to reduce surface tension of water than their 3-alkoxy substituted isomers 1e–h or 1m–p (Table 1, Figure 2).

Figure 2

Relationships between number of carbon atoms forming the 2-/3-alkoxy side chain R (nc; alkoxy = butoxy to heptyloxy) and γ values (in N/m units) of the compounds 1a–p.

Relationships between the number of pan class="Chemical">carbon atoms forming the chain R (nc) and γ values (in N/m units) was inspn>ected using a linear function, polynomial function of 2nd order and sigmoidal fitting, respn>ectively.

Equations related to the most relevant functions and values of common statistical descriptors, namely, number of points (number of cases; n), degrees of freedom (DF), reduced chi-square (χ,2red), residual sum of squares (RSS), correlation coefficient (R), adjusted coefficient of determination (Adj. R2), root mean squared error (standard deviation; RMSE), norm of residuals (NR), Fisher’s significance ratio (Fisher’s F-test; F) and probability of obtaining the F Ratio (significance of a whole model; Prob > F), respectively, can be found in Table S2.

Indication of a significance level of the F Ratio was as follows: one star symbol (*) for statistically significant, two stars symbol (**) for statistically very significant or three stars symbol (***) for statistically extremely significant level was used. The regression equations and their statistical characteristics were calculated and visualized by the Origin pan class="Chemical">Pro ver. 9.0.0 SR2 software (OriginLab Corpn>oration, Northampn>ton, MA, USA).

The relationships between nc and γ were described most precisely by statistically significant (1a–d, 1i–l and 1m–p) or very significant (1e–h) models built on polynomial functions of 2nd order (Figure 2, Equations (S1–S4) in Table S2).

The derivatives containing an azepanium moiety (1i–l, 1m–p) showed slightly higher ability to decrease surface tension of pan class="Chemical">water than their positional isomers with a pyrrolidinium salt-forming group (1a–d, 1e–h). The most surface active were 1-[2-[({[3-(heptyloxy)phenyl]amino}carbonyl)-oxy]-3-(dipropylammonio)propyl]azepanium dichloride (1p) with γ = 0.05692 N/m and 1-[2-[({[3-(heptyloxy)phenyl]amino}carbonyl)oxy]-3-(dipropylammonio)propyl]pyrrolidinium dichloride (1h), which showed γ = 0.05786 N/m (Table 1, Figure 2).

Very similar behaviors to present findings were observed when Čižmárik with co-workers [33,34] analyzed structurally similar compounds JC-01a–l, effective LAs [33], which contained only one centrum of basicity (Figure 3). All given molecules showed ability to decrease surface tension of water and 3-pan class="Chemical">alkoxy substituted derivatives JC-01g–l (alkoxy = propoxy to octyloxy) were more efficient surfactants than their 2-alkoxy positional isomers JC-01a–f. Relationships between nc (nc = 3–8) and γ (in N/m units) of these alkoxy positional isomers were described very satisfactorily by polynomial functions of 2nd order [34].

Figure 3

Chemical structure of the compounds JC-01a–l, effective local anesthetics [33], which were able to decrease surface tension of water [34].

Concerning presently analyzed dichlorides 1m–p, branching of their con>an class="Chemical">nnecting chain and presence of a dipropylammonium moiety resulted in the γ values (Table 1), which were slightly lower in comparison to those of JC-01h–k [34].

2.2.2. Electronic Properties

Electronic properties of the molecules 1a–p (Table 1) were characterized by logarithms of molar absorption coefficients (log ε) of their pan class="Chemical">methanolic solutions (c = 8.0 × 10−5 M) investigated in the UV/Vis region of an electromagnetic spn>ectrum.

The solutions showed three absorption maxima in a near ultraviolet (quarz) region of this spectrum between 200 nm and 400 nm [35], namely, λ1 = 208–210 nm, λ2 (Ch-T) = 236–238 nm and λ3 = 278–280 nm (Table 1), respectively. Positions of these absorpn>tion bands in the spectrum were typn>ical for derivatives of (substituted) pan class="Chemical">phenylcarbamic acid and the bands were assigned as first local excitation (λ3), charge-transfer (λ2 (Ch-T)) and second local excitation (λ1) absorption maxima, respectively [36]. The λ1 and λ2 (Ch-T) maxima of 2-alkoxy substituted compounds 1a–d and 1i–l were slightly lower compared to those of their 3-alkoxy substituted positional isomers 1e–h and 1m–p (Table 1), respectively. Chemical structures of analyzed derivatives 1a–d and 1i–l indicated that steric inability of both carbamoyloxy and alkoxy groups to achieve coplanarity inhibited ‘proper’ resonance.

The log ε2 (Ch-T) parameters of compounds 1a–p, which were observed at λ2 (Ch-T), varied from 4.01 (1o) to 4.52 (1c). The methanolic solutions of 1a–h were characterized by higher log ε2 (Ch-T) values than those of 1i–p and ranged from 4.05 (1b) to 4.52 (1c; Table 1), excluding the molecules n>an class="Chemical">1j (4.22) and 1l (4.14). In addition, there was found nor linear neither quasi-parabolic relationship between number of carbon atoms forming the side chain R (nc) and log ε2 (Ch-T) values (Figure S1 in Supplementary Materials).

In next sections of the paper, which are aimed at relationships between structure and in vitro activity, special consideration is devoted to the log ε2 (Ch-T) values because they could be the most sensitive to differences in electronic envpan class="Chemical">ironment of a phenylcarbamoyloxy moiety due to different position and length of a substituent R [36].

2.2.3. Lipohydrophilic Properties

Lipophilicity has been the physicochemical parameter of notable importance in QSAR and SAR studies as a predominant descriptor encoding information on a network of inter- and intramolecular forces affecting drug transport via lipophilic compartments as well as drug’s interactions with target effector sites [37,38].

A classical shake-flask method for a partitioning measurement of a compound in the octan-1-ol/buffer (or pan class="Chemical">water) system, thus estimation of its log Pexp value, is quite time-consuming, it requires the solute to be pure and with adequate solubility in an aqueous phase. Insolubility often means that highly lipophilic compounds may not be determined accurately [39]. Concerning these issues very seriously, RP-HPLC was rather applied to estimate lipophilic properties of the molecules 1a–p than the shake-flask alternative.

Octadecyl-functionalized n>an class="Chemical">silica gel was currently used as a stationary phase (SPh) and a gradient of two solvents at different volume ratios modified retention properties of the SPh [39,40]. Liquid binary mixtures of methanol (MeOH) with water were employed as mobile phases (MPhs) in the isocratic RP-HPLC. The MeOH modifier was preferred because of similarity of both observed and extrapolated lipophilicity parameters with the ones determined in the octan-1-ol/water partition system, considering also sensitivity to H-bond donor properties of investigated compounds [41].

The modifier was applied in different volume concentrations, which varied from 85% to 100% (v/v) with 5% increments. This interval was chosen due to very high lipophilicity of investigated salts. Practically, lower volume concentration of MeOH, i.e., less than 85% (v/v) in used MPhs, caused uncertainty of retention times (tr) due to notable peaks broadening, especially within subsets 1i–l and 1m–p. In addition, their tr values were considerably prolonged with increase in water content (v/v) of the MPhs so it was no longer possible to determine tr values in such MPhs. Taking the 85–100% volume concentrations range, the isocratic separation was feasible and reasonable retention of the compounds 1a–p was observed in all MPhs. The highest t parameters were found for the molecules 1k (tr = 47.108 min) and 1l (tr = 70.012 min) in the MPh containing 85% proportion (v/v) of MeOH (Table 2).

Table 2

Retention times tr (RP-HPLC) and lipophilicity indices log k (RP-HPLC) of the compounds 1a–p estimated in the mobile phases with a various volume ratio (v/v) of a methanol (MeOH) organic modifier and water.

Comp.	Mobile Phase MeOH/Water (v/v)
	80:20		85:15		90:10		95:5		Pure MeOH
	tr (min)	log k	tr (min)	log k	tr (min)	log k	tr (min)	log k	tr (min)	log k
1a	13.883	0.7165	10.694	0.5756	7.830	0.3958	5.766	0.2008	4.224	−0.0403
1b	17.200	0.8254	12.647	0.6656	9.449	0.5064	6.514	0.2841	4.526	0.0203
1c	25.694	1.0206	16.950	0.8162	11.292	0.6053	7.431	0.3683	5.246	0.1287
1d	31.000	1.1092	19.600	0.8882	13.553	0.7022	7.892	0.4052	5.385	0.1573
1e	10.330	0.5584	6.917	0.3183	5.400	0.1478	3.822	−0.1454	3.329	−0.2956
1f	12.400	0.6574	9.200	0.4911	6.280	0.2546	4.022	−0.0941	3.340	−0.2913
1g	17.200	0.8254	11.402	0.6104	7.483	0.3679	5.153	0.1182	3.583	−0.2067
1h	22.003	0.9521	13.701	0.7159	8.532	0.4565	5.788	0.2092	3.556	−0.2153
1i	24.020	0.9881	16.917	0.8153	11.691	0.6240	7.765	0.3954	5.127	0.1205
1j	29.525	1.0863	20.021	0.8981	13.060	0.6828	8.520	0.4509	5.295	0.1449
1k	47.108	1.3023	27.001	1.0424	16.533	0.8038	9.641	0.5221	6.938	0.3303
1l	70.012	1.4831	38.800	1.2152	22.636	0.9619	11.670	0.6271	7.756	0.3996
1m	15.731	0.7805	9.218	0.4922	6.420	0.2695	4.685	0.0424	3.519	−0.2275
1n	22.183	0.9502	12.370	0.6542	8.156	0.4205	4.995	0.0941	3.947	−0.1047
1o	29.174	1.0807	13.651	0.7059	9.634	0.5174	5.095	0.1095	4.127	−0.0618
1p	33.719	1.1484	17.129	0.8215	11.332	0.6072	5.801	0.2051	3.925	−0.1102

In addition, purity (in percentages) of the compounds 1a–p was verified by RP-HPLC. Areas of their peaks were measured using the MPh, which contained 90% proportion (v/v) of MeOH. The purity of given salts varied from 96.82% (1j) to 99.65% (1m; Table S3).

Estimated retention factor (capacity factor; k) values were found in an area from 0.5063 (1e) to 2.5096 (1l) if pure MeOH was used as the Mn>an class="Chemical">Ph, and from 3.6174 (1e) to 30.4159 (1l) if the 80:20 MeOH/water (v/v) MPh was applied (Table S3).

Regarding a group of oxalates (Scheme 1), a n>an class="Chemical">pyrrolidium moiety-containing derivatives 1a–d showed lower log k values in all used MPhs than those with an azepanium fragment 1i–l. Same trends were observed for dichlorides, i.e., the compounds 1e–h showed lower log k outputs compared to those of a series 1m–p.

Elongation of the R substituent (Table 1) led to increase in tr as well as log k values within all evaluated subgroups 1a–d, 1e–h, 1i–l and 1m–p (Table 2). Increase in a volume concentration (v/v) of pan class="Chemical">MeOH caused shortening of both tr and log k for all analyzed compn>ounds 1a–p (Table 2).

El Tayar et al. [42,43] investigated lipophilic properties of protonated basic compounds, which were therapeutically used as psychoactive agents, and inspected relationships between their log k values and volume concentrations (v/v) of MeOH. They found that these relationships were notably influenced by nature of solutes. The models were preferably built on linear functions and pan class="Chemical">MeOH exerted its own solvophobic effect in the MPhs, which contained more than 80% (v/v) of given organic modifier if focusing on analyses of neutral and non-ionic compounds.

For partially and completely ionized polar compn>ounds, the relationshipn>s apn>peared to be nearly parabolic but the parts of parabolic curves, which correspn>onded to n>an class="Chemical">MeOH-rich eluents (80% or higher volume concentration (v/v) of MeOH) were regarded as linear [42,43].

Paschke et al. [44] analyzed a series of highly lipn>opn>hilic n>an class="Chemical">tetrachlorobenzyltoluene isomers by isocratic RP-HPLC and found linear relationships (R > 0.9995) between their log k values and volume concentrations of MeOH, which varied from 80% to 100% (v/v).

On the other hand, observed log k data could not be used in terms of a universal scale to express lipophilicity of compounds because the log k values were dependent on chromatographic conditions, i.e., they depended on both MPhs and SPhs. Extrapolation of estimated solute retention (log k) to elution with 100% water (calculation of log kw) would definitely be much more precise approach than using of log k, because the kw parameters were independent of any effect of the organic modifier and relied on the SPh alone [45]. It has been recognized that the log kw descriptor was very efficient for anticipation of in vitro antimycobacterial properties of compounds [46,47,48,49].

According to the authors’ opinions based on systematic scientific literature survey, a current research paper was the first, which aimed the log kw parameters of 2-/3-alkoxyphenylcarbamic acid derivatives containing two centers of protonation. The log kw values were extrapn>olated from intercepn>ts of a relationship between log k and volume fraction of the MPh modifier (ϕM) using the Snyder–Soczewiński linear solvent strength model [50,51,52]. The relationship was justified by R > 0.9900, Adj. R2 > 0.9750 as well as F > 180.00 for calculation of particular log kw parameters (Table 3).

Table 3

Extrapolated log kw parameters (RP-HPLC) of analyzed molecules 1a–p and values of statistical descriptors (χ2red, RSS, R, Adj. R2, RMSE, NR, F and Prob > F), which characterized a linear relationship between the log k and ϕM values for a particular compound. The ϕM parameter was a volume fraction of MeOH in the isocratic elution RP-HPLC.

Comp.	log kw	1 S	2 χ 2 red	3 RSS	4 R	5 Adj. R 2	6 RMSE	7 NR	8 F	9 Prob > F
1a	3.7688	3.7768	0.0011	0.0034	0.9953	0.9875	0.0335	0.0580	317.64	0.0001 ***
1b	4.0454	3.9834	0.0018	0.0055	0.9931	0.9817	0.0429	0.0744	215.17	0.0007 ***
1c	4.6049	4.4634	0.0002	0.0007	0.9993	0.9982	0.0152	0.0263	2155.37	0.0001 ***
1d	4.9487	4.7736	0.0011	0.0033	0.9971	0.9923	0.0332	0.0574	518.15	0.0002 ***
1e	4.0258	4.3434	0.0013	0.0038	0.9960	0.9894	0.0354	0.0614	375.47	0.0003 ***
1f	4.6722	4.9652	0.0028	0.0084	0.9932	0.9821	0.0530	0.0917	219.79	0.0007 ***
1g	4.9446	5.1128	0.0013	0.0040	0.9969	0.9919	0.0366	0.0634	487.75	0.0002 ***
1h	5.5384	5.6830	0.0042	0.0126	0.9923	0.9795	0.0648	0.1122	192.51	0.0008 ***
1i	4.4679	4.3102	0.0014	0.0042	0.9955	0.9879	0.0376	0.0652	328.27	0.0004 ***
1j	4.8466	4.6600	0.0016	0.0048	0.9956	0.9882	0.0401	0.0695	337.19	0.0004 ***
1k	5.2359	4.9286	0.0005	0.0017	0.9987	0.9964	0.0234	0.0406	1106.57	0.0001 ***
1l	5.8966	5.5102	0.0007	0.0020	0.9988	0.9965	0.0259	0.0449	1127.92	0.0001 ***
1m	4.7099	4.9316	0.0004	0.0013	0.9990	0.9972	0.0205	0.0356	1442.73	0.0001 ***
1n	5.2087	5.3398	0.0011	0.0032	0.9978	0.9941	0.0325	0.0564	673.05	0.0001 ***
1o	5.6569	5.7628	0.0045	0.0134	0.9920	0.9789	0.0668	0.1157	186.23	0.0009 ***
1p	6.1749	6.2672	0.0022	0.0067	0.9966	0.9909	0.0474	0.0820	438.042	0.0001 ***

1S, Slope; 2χ2red, reduced chi-square; 3RSS, residual sum of squares; 4R, correlation coefficient; 5Adj. R2, adjusted coefficient of determination; 6RMSE, root mean squared error (standard deviation); 7NR, norm of residuals; 8F, Fisher’s significance ratio (Fisher’s F-test); 9Prob > F, probability of obtaining the F Ratio (significance of a whole model). Indication of a significance level of the F Ratio was as follows: *** (three stars), extremely significant.

Goodness of fit for all proposed models generating log kw, in other words, quantity used to test whether any given data were well described by a suggested function, was adjusted by χ2red values, which varied from 0.0002 (1c) to 0.0045 (1o). As can be seen, the linear relationships minimized pan class="Chemical">RSS values, as variabilities about regression lines, which ranged from 0.0007 (1c) to 0.0134 (1o). The linearity was also proved by calculated RMSE parameters, as standard deviations of the data about regression lines, which were found in an interval from 0.0152 (1c) to 0.0668 (1o; Table 3).

The extrapolated log kw values of compounds 1a–p (Table 3) were in accordance with their elution order and hydrophobicity and varied from 3.7688 (1a) to 6.1749 (1p). Assuming presence of an identical pan class="Chemical">azacycloalkyl moiety, oxalates (1a–d, 1i–l) were less lipophilic than corresponding positional isomers synthesized as dichlorides (1e–h, 1m–p; Table 3). As expected, 1-[2-[({[3-(hept-yloxy)phenyl]amino}carbonyl)oxy]-3-(dipropylammonio)propyl]azepanium oxalate (1l) and 1-[2-[({[3-(heptyloxy)phenyl]amino}carbonyl)oxy]-3-(dipropylammonio)propyl]azepanium dichloride (1p; Table 1) were found to be the most lipophilic showing log kw of 5.8966 (1l) and 6.1749 (1p; Table 3), respectively.

Increase in nc (Table 1) led to higher log kw values in all inspected subgroups 1a–d, 1e–h, 1i–l and 1m–p. These relationships were defined by statistically significant (1e–h, 1i–l), very significant (1a–d) and extremely significant (1m–p) linear functions and values of relevant statistical descriptors (Equations (S5)–(S8) in Table S4).

The S index (Table 3) in RP-Hn>an class="Chemical">PLC has been a function of molecular structure parameters. For sets of non-polar homologues, linear relationships between their S and number of carbon atoms in a hydrocarbon side chain was found. Increase in size and van der Waals volume of a solute led to increased S parameter as well [53].

Results of present analyses agreed with given conclusions. Increase in length of the side chain R led to higher S within all inspected homological subgroups. The compounds containing an azepanium moiety showed higher S parameters, which varied from 4.3102 (1i) to 6.2672 (pan class="Chemical">1p), than those with a pyrrolidium group, i.e., from 3.7768 (1a) to 5.6830 (1h; Table 3), if same position and length of R was presumed.

The slope S values of linear regression functions used to obtain log kw were copan class="Chemical">nnected with spn>ecific hydropn>hobic surface areas of compn>ounds and could serve as indicative measure of uniformity of retention mechanisms. If the uniformity was observed, a convenient model between the slopn>e(s) and intercepn>t(s), i.e., log kw values, was anticipn>ated [40]. Statistically extremely significant relationshipn> between the log kw and S values of 1a–p was found and described by Equation (1):

Based on values of statistical descriptors provided above, the uniformity of a retention mechanism was proven and suitability of chosen Mpan class="Chemical">Phs was confirmed for lipn>opn>hilicity evaluation of the derivatives 1a–p.

Statistical analyses indicated that both position and length of the side chain R (Equations (S9) and (S10) in Table S5) were more important factors, which contributed to the uniformity of retention behavior, compared to nature (and physicochemical properties) of a pan class="Chemical">salt-forming fragment (Equations (S11) and (S12) in Table S5) of inspn>ected compn>ounds, despn>ite the fact that all partial models built on linear regression analyses were statistically extremely significant.

To provide better understanding and give a critical review of relationships between the experimental lipophilicity measures and in silico prediction of lipophilic properties, the log kw dataset was studied together with computational logarithms of partition coefficients (log pan class="Chemical">P).

Practical estimation of log Pexp for some molecules from the set 1a–p by a shake-flask method was not possible due to their insolubility in phosphate buffer systems (pH = 7.30). Thus, the lack of enough reliable log Pexp values would be a very limiting factor for eventual investigation of the relationships between log Pexp and in silico log P. On the other hand, linear models between log kw and log Pexp were already observed [39], so interchangeability of these descriptors was fully justified.

The log P parameters of non-protonated forms of analyzed compn>ounds, i.e., 9′a–p (Scheme 1, Table 4), were calculated for the n>an class="Chemical">octan-1-ol/water partitioning system. These values were generated by the ChemBioDraw Ultra 12.0 software package (CambridgeSoft, Cambridge, MA, USA), various Java and non-Java interactive applets as well.

Table 4

Values of logarithms of partition coefficients of basic (non-protonated) compounds 9′a–p predicted for the octan-1-ol/water partitioning system by the Ghose and Crippen′s approach (log PCr), Viswanadhan′s principle (log PV), Broto′s algorithm (log PB) and CLOGP method (CLOGP 4.0), respectively, as well as by both XLOGP approaches (XLOGP 2.0 and XLOGP 3.0).

Comp.	R	X	log PCr	log PV	log PB	CLOGP4.0	XLOGP2.0	XLOGP3.0
9′a	2-OC4H9	A	4.59	4.45	4.51	6.23	4.52	5.14
9′b	2-OC5H11	A	5.00	4.85	4.97	6.76	5.09	5.68
9′c	2-OC6H13	A	5.42	5.24	5.42	7.29	5.66	6.22
9′d	2-OC7H15	A	5.84	5.64	5.88	7.82	6.23	6.77
9′e	3-OC4H9	A	4.59	4.45	4.51	6.23	4.52	5.14
9′f	3-OC5H11	A	5.00	4.85	4.97	6.76	5.09	5.68
9′g	3-OC6H13	A	5.42	5.24	5.42	7.29	5.66	6.22
9′h	3-OC7H15	A	5.84	5.64	5.88	7.82	6.23	6.77
9′i	2-OC4H9	B	5.42	5.24	5.42	7.35	5.24	5.86
9′j	2-OC5H11	B	5.84	5.64	5.88	7.88	5.81	6.40
9′k	2-OC6H13	B	6.26	6.04	6.33	8.41	6.38	6.84
9′l	2-OC7H15	B	6.67	6.43	6.79	8.94	6.95	7.48
9′m	3-OC4H9	B	5.42	5.24	5.42	7.35	5.24	5.86
9′n	3-OC5H11	B	5.84	5.64	5.88	7.88	5.81	6.40
9′o	3-OC6H13	B	6.26	6.04	6.33	8.41	6.38	6.84
9′p	3-OC7H15	B	6.67	6.43	6.79	8.94	6.95	7.48

The Ghose and Crippen′s log PCr [54,55], Viswanadhan′s log PV [56], Broto’s log PB [57] and Leo’s CLOGP 4.0 [58] atomic and atomic/fragmental-based approaches were chosen to calculate the log p values (Table 4) by the ChemBioDraw Ultra 12.0 software, respectively. The Virtual Computer Chemistry Laboratory [59], a freely available web-based tool working in Java environment, was used to calculate log P by the Wang’s XLOGP 2.0 [60], Cheng’s XLOGP 3.0 [61], Moriguchi’s MLOGP [62], Sander’s (Actelion’s) ACLOGP [63], Molinspiration’s miLogP 2.2 [64], ALOGP [65] and Tetko’s ALOGPs 2.1 [66] method (Table 4 and Table 5), respectively. All these methods integrated algorithms combining atomic and fragmental principles, except for ALOGPs 2.1, which considered a molecule in the whole [66].

Table 5

Values of logarithms of partition coefficients of basic (non-protonated) compounds 9′a–p predicted for the octan-1-ol/water partitioning system by the MLOGP, ACLOGP, miLogP 2.2, ALOGP, SILICOS-IT (log PS-IT) and ALOGPs 2.1 method, respectively.

Comp.	MLOGP	ACLOGP	miLogP2.2	ALOGP	log PS-IT	ALOGPs2.1
9′a	2.75	5.05	5.30	5.42	4.27	5.18
9′b	2.96	5.52	5.80	5.87	4.68	5.57
9′c	3.16	5.98	6.32	6.33	5.10	5.93
9′d	3.36	6.45	6.81	6.79	5.53	6.25
9′e	2.75	5.05	5.32	5.42	4.27	5.21
9′f	2.96	5.52	5.83	5.87	4.68	5.61
9′g	3.16	5.98	6.33	6.33	5.10	5.95
9′h	3.36	6.45	6.84	6.79	5.53	6.32
9′i	3.16	5.69	6.31	6.33	4.78	5.93
9′j	3.36	6.16	6.81	6.79	5.20	6.23
9′k	3.56	6.62	7.32	7.24	5.62	6.54
9′l	3.76	7.08	7.82	7.70	6.04	6.77
9′m	3.16	5.69	6.33	6.33	4.78	5.97
9′n	3.36	6.16	6.84	6.79	5.20	6.27
9′o	3.56	6.62	7.34	7.24	5.62	6.59
9′p	3.76	7.08	7.85	7.70	6.04	6.83

Finally, the SILICOS-IT hybrid method, which used proprietary fragment- and property-based principles, was also employed to generate log pan class="Chemical">P parameters (log n>an class="Chemical">PS-IT; Table 5). The method was implemented in the SwissADME applet, a free web tool designed to evaluate pharmacokinetics and drug-likeness of small molecules [67].

A common issue copan class="Chemical">nnected with given in silico apn>proaches [26] was that they did not allow to correctly predict log p values of particular n>an class="Chemical">salts 1a–p. This was the reason why current research targeted on the calculations related to non-protonated bases 9′a–p.

The compounds 9′i–p showed higher log P than their positional isomers 9′a–h, which contained a different n>an class="Chemical">salt-forming moiety. Almost all used predictor tools indicated the molecules 9′a–p highly lipophilic with log p > 4.27 (Table 4 and Table 5), excluding the MLOGP method. The calculations based on MLOGP indicated moderate or high lipophilicity of these substances, which log P were found in a range from 2.75 (9′a and 9′e) to 3.76 (9′l and 9′p; Table 5).

The CLOGP 4.0 predicted the highest lipn>opn>hilic nature of the bases 9′a–p and their log n>an class="Chemical">P varied from 6.23 (9′a and 9′e) to 8.94 (9′l and 9′p; Table 4). As might be assumed, all employed atomic and atomic/fragmental methods did not consider a position of the side chain R when generating log P (Table 4 and Table 5). Only the whole-molecule ALOGPs 2.1 approach took into account this aspect, so the 2-alkoxy substituted compounds 9′a–d and 9′i–l were less lipophilic than their 3-alkoxy substituted positional isomers 9′e–h and 9′m–p (Table 5), respectively.

It was found that linear relationships between the extrapolated log kw values of compounds 1a–p (Table 3) and particular in silico log P parameters of non-protonated derivatives 9′a–p, namely, log PCr, log PV, log PB, CLOGP 4.0, XLOGP 2.0, XLOGP 3.0, MLOGP, ACLOGP, miLogP 2.2, ALOGP, log PS-IT and ALOGPs 2.1 (Table 4 and Table 5), respectively, were statistically extremely significant (Prob > F = 0.0001) and were defined by R > 0.9300, Adj. R2 > 0.8650, F > 99.00 and other common statistical descriptors (Equations (S13)–(S24) in Table S6).

In an effort to characterize these relationships more precisely, similarities and differences between given descriptors were analyzed by using unscaled principal component analysis (PCA), a powerful multivariate statistical technique, that analyzed a set of values, in which observations were described by several inter-correlated quantitative depn>endent variables. A set of new orthogonal variables, called principal components (pan class="Chemical">PCs), and pattern of similarity of observations was a result the analysis [68].

Current PCAs were performed by the Origin Pro ver. 9.0.0 SR2 software. Number of concerned PCs was determined by visual evaluation of a scree plot, as a relationship between calculated eigenvalues (λe) and number of PCs. The λe descriptor measured amount of variation retained by each PC [68]. The first two PCs, i.e., Principal Component 1 (PC 1) and Principal Component 2 (PC 2), of the analysis accounted for 99.26% of the total variance in the data. It meant that other PCs might be ignored without losing any substantial information. The PCs did not prove existence of ‘real’ parameters, they only indicated that existence of these descriptors was mathematically possible within the set of analyzed values.

A relationship between PC 1 and PC 2 resulted in division of examined compounds into characteristic groups. Differences in lipophilic properties of the bases 9′a–p were reflected in their PC 1 values. One group, defined by negative PC 1 values, included derivatives containing a pyrrolidin-1-yl moiety (9′a–c, 9′e–g) as well as molecules with both azepan-1-yl fragment and butoxy side chain (9′i, 9′m). Conversely, second group, which was characterized by positive PC 1 values, encompassed compounds containing an azepan-1-yl moiety (9′j–l, 9′n–p) as well as derivatives with both sterically smaller azacycloalkyl ring and heptyloxy substituent (9′d, 9′h). The most negative PC 1 value was generated for the least lipophilic derivative 9′a, which salt 1a showed the lowest experimental lipophilicity (log kw = 3.7688). The most positive PC 1 output was registered for the most lipophilic compound 9′p, a base of the most lipophilic salt 1p (log kw = 6.1749; Figure 4).

Figure 4

Two-dimensional biplot (mapping) showing both Principal Component 1 and 2 scores of the compounds 9′a–p and loadings of extrapolated chromatographic (log kw; 1a–p) and in silico lipophilicity variables, i.e., log PS-IT, XLOGP 2.0, XLOGP 3.0, ACLOGP, ALOGPs 2.1, miLogP 2.2, CLOGP 4.0, log PCr, log PV, log PB, MLOGP and ALOGP, respectively.

It was found that PC 2 took into consideration a typn>e of an n>an class="Chemical">azacycloalkyl moiety as well as position and length of R. The derivatives 9′a–h showed higher PC 2 than the compounds 9′i–p. Higher PC 2 were assigned to the 3-alkoxy substituted molecules compared to their 2-alkoxy positional isomers, when assuming their identical salt-forming moiety.

Increase in nc of the side chain R led to higher PC 1 as well as n>an class="Chemical">PC 2. There were found no compounds (points), which could be regarded as outliers to the remaining ones along PC 1 or PC 2 (Figure 4).

The PCA-based 2D bipn>lot showed both n>an class="Chemical">PC scores (PC 1, PC 2) of evaluated compounds (green dots) and loadings of variables indicated as orange straight lines [68]. The lines characterizing log kw, log PI-IT, XLOGP 2.0, XLOGP 3.0 and ACLOGP variables were located on upper right side of this biplot. The line related to log kw formed the smallest angles with those of log PI-IT, XLOGP 2.0 or XLOGP 3.0, which meant that these descriptors correlated each other most positively. Conversely, the largest angle was observed between the lines connected with log kw and CLOGP 4.0, so these variables were unlikely to correlate each other (Figure 4).

The PC 1 was influenced almost equally by loadings of all inspn>ected variables, as indicated particular scores on topn> axis. The scores varied from 0.27 (log kw) to 0.28 (log n>an class="Chemical">PCr, log PV, log PB, miLogP 2.2 and ALOGP, respectively). The most positive impact on PC 2 showed a loading of the log kw and log PI-IT variable, which scored approximately 0.45 and 0.37, respectively, on right axis. Conversely, the strongest inverse relationship was found between PC 2 and loading of the CLOGP 4.0 variable (−0.28; Figure 4).

Following visualization, the in silico log PI-IT, XLOGpan class="Chemical">P 2.0 and XLOGP 3.0 descriptors might be used to characterize lipophilic properties of analyzed salts and bases, especially if they contained highly lipophilic side chain R. Those molecules were defined by positive values of both PC 1 and PC 2. In addition, very close grouping was observed for 9′l with 9′p (Figure 4).

Considering observed physicochemical characteristics of the compounds 1a–p, there was found statistically extremely significant linear relationship between the γ and log kw values. This dependence was expressed by Equation (2) and values of relevant statistical descriptors as follows:γ = −0.0031 (±0.0003) × log

Linearity of this model was adjusted by relatively very low χ2red, RSS and RMSE values. The γ and log kw parameters of 1i–p showed the most positive impn>act to final statistical analysis data (Equation (S28) in Table S7) con>an class="Chemical">nnected with Equation (2) compared to those of a subset 1a–h (Equation (S27) in Table S7).

One of the major tasks to consider in drug design is the ability of a compound to cross biological membranes. The blood–brain barrier (BBB) is a critical biological structure for maintaining pan class="Disease">homeostasis of central nervous system and preventing damage to the brain. Compounds primarily designed as neuroactive agents are required to cross the BBB to provide their therapeutic effect. Conversely, molecules that target other compartments of a body ideally should not cross this barrier to avoid possible psychotropic side effects. Thus, the task of predicting the BBB permeability of new compounds is of great importance [69].

Norinder and Haeberlein [70] suggested some simple rules to preliminary answer a question whether a drug candidate would be able to cross the BBB. Very briefly, if a sum of nitrogen and pan class="Chemical">oxygen atoms (nN + nO) in a molecule was ≤ 5 or if [CLOGP − (nN + nO)] > 0 then the compound would have a good chance to enter the brain by passive diffusion. The originally employed CLOGP was a lipophilicity descriptor generated in silico. In addition, molecular weight (MW) of such compound should be ≤ 450 to facilitate brain permeation [71].

Each of the oxalates 1a–d and 1i–l contained (n>an class="Chemical">nN + nO) = 10, each of the chlorides 1e–h and 1m–p contained (nN + nO) = 6 (Table 1), respectively. If presently calculated CLOGP 4.0 parameter would be replaced with log PS-IT (Table 5), which was closer to a ‘real’ log kw lipophilicity descriptor (Figure 4), only basic forms of the most lipophilic compounds 1l and 1p, i.e., the substances 9′l and 9′p, might be considered to possess very limited brain permeation. On the other hand, calculated MW > 450 for these derivatives [21] indicated that they might not be able to passively cross the BBB.

2.3. In Vitro Antimycobacterial Assays

The in vitro activity of compounds 1a–p was inspected against Mycobacterium tuberculosis CNCTC My 331/88 (identical with n>an class="Species">H37Rv and ATCC 2794, respectively; abbreviation used: MTv H37Rv), M. tuberculosis H37Ra ATCC 25177 (MTa H37Ra), M. kansasii CNCTC My 235/80 (identical with ATCC 12478; MK 235/80), a M. kansasii 6509/96 clinical isolate (MK 6509/96), M. kansasii DSM 44162 (MK DSM), M. avium CNCTC My 330/80 (identical with ATCC 25291; MA 330/80), M. smegmatis ATCC 700084 (MS) and M. marinum CAMP 5644 (MM), respectively, by the methods described earlier [20,72,73,74,75]. In vitro susceptibitity of given mycobacteria to reference drugs isoniazid (INH), ethambutol (EMB), ofloxacin (OFLX) or ciprofloxacin (CPX) was tested as well (Table 6 and Table 7).

Table 6

The in vitro activity (MIC values in μM units) of investigated compounds 1a–p and reference drugs isoniazid (INH), ethambutol (EMB), ofloxacin (OFLX) and ciprofloxacin (CPX) against M. tuberculosis CNCTC My 331/88 (M. tuberculosis H37Rv; MTv H37Rv), M. tuberculosis H37Ra ATCC 25177 (MTa H37Ra), M. kansasii CNCTC My 235/80 (MK 235/80), M. kansasii 6509/96 (MK 6509/96) and M. kansasii DSM 44162 (MK DSM), respectively.

Comp.	MIC [μM]
	MTvH37Rv		MTaH37Ra	MK 235/80			MK 6509/96			MKDSM
	1 14-d	2 21-d	3 7-d	7-d	14-d	21-d	7-d	14-d	21-d	7-d
1a	32	32	213	62.5	62.5	62.5	32	62.5	62.5	53
1b	16	32	52	32	32	62.5	32	32	62.5	26.1
1c	16	16	25	32	32	32	16	16	16	12.7
1d	8	8	12.5	16	16	32	8	16	16	6.2
1e	16	32	8.1	32	32	32	16	16	32	16.2
1f	16	16	<3.9	16	16	32	8	16	16	7.9
1g	8	16	<3.8	16	16	16	8	8	16	3.8
1h	4	8	<3.7	8	8	8	4	4	8	1.9
1i	16	32	203	32	62.5	62.5	16	32	32	51
1j	16	32	199	16	32	62.5	16	32	32	50
1k	8	16	195	16	32	62.5	16	32	32	98
1l	4	8	96	16	32	32	16	16	32	382
1m	8	16	15.4	16	16	32	8	16	16	7.7
1n	4	8	7.5	8	8	8	4	8	8	7.5
1o	8	16	7.3	8	16	16	4	8	8	3.6
1p	4	8	7.1	8	8	8	4	8	8	3.6
INH	0.5	1	36.5	>250	>250	>250	2	4	8	29.2
EMB	1	2	–	–	–	–	1	2	2	–
OFLX	1	2	–	0.5	1	1	0.5	0.5	1	–
CPX	–	–	48.3	–	–	–	–	–	–	3.0

1 14-d, 14-Day cultivation; 2 21-d, 21-day cultivation; 3 7-d, 7-day cultivation. The most promising values of compounds in vitro antimycobacterial activity (MIC ≤ 8 μM) were indicated in gray cells of Table 6.

Table 7

The in vitro activity (MIC values in μM units) of investigated compounds 1a–p and reference drugs isoniazid (INH), ethambutol (EMB), ofloxacin (OFLX) and ciprofloxacin (CPX) against M. avium CNCTC My 330/80 (MA 330/80), M. smegmatis ATCC 700084 (MS) and M. marinum CAMP 5644 (MM), respectively.

Comp.	MIC [μM]
	MA 330/80		MS	MM
	1 14-d	2 21-d	3 3-d	21-d
1a	125	125	213	106
1b	62.5	125	104	52
1c	32	62.5	51	25.5
1d	32	32	24.9	12.5
1e	62.5	62.5	16.2	32.5
1f	32	32	15.8	<3.9
1g	16	16	15.4	<3.8
1h	8	16	15.0	<3.7
1i	62.5	125	203	51
1j	62.5	125	6.2	50
1k	32	62.5	390	98
1l	32	62.5	382	191
1m	16	32	30.7	15.4
1n	16	16	15.0	15.0
1o	16	16	7.3	7.3
1p	16	16	7.1	7.1
INH	> 250	> 250	117	467
EMB	16	16	–	–
OFLX	32	62.5	–	–
CPX	–	–	0.4	0.8

1 14-d, 14-Day cultivation; 2 21-d, 21-day cultivation; 3 3-d, 3-day cultivation. The most promising values of compounds´ in vitro antimycobacterial activity (MIC ≤ 8 μM) were indicated in gray cells of Table 7.

A value of a minimum inhibitory concentration (MIC) was the lowest concentration of a tested compound, (i) which inhibited growth of MTv H37Rv, MK 235/80, n>an class="Chemical">MK 6509/96 and MA 330/80, respectively; (ii) at which no visible growth of MTa H37Ra, MS, MK DSM as well as MM was observed [20,72,73,74,75].

The in vitro screening procedures related to particular mycobacterial strains were repeated three times and the MIC values, reported in Table 6 and Table 7 (in μM units), were average values of these determinations.

A very unique aspect of a present research was that many compounds from the set 1a–p were in vitro efficient against almost all tested mycobacteria. The most promising molecules with observed MIC ≤ 8 μM were indicated in gray (Table 6 and Table 7).

The 3-alkoxy substituted derivatives 1e–h and 1m–p were slightly more effective against all screened strains than their 2-n>an class="Chemical">alkoxy substituted isomers 1a–d and 1i–l, respectively. 1-[2-[({[3-(Heptyloxy)phenyl]amino}carbonyl)oxy]-3-(dipropylammonio)propyl]azepanium dichloride (1p) was the most active against almost complete spectrum of tested microorganisms (MIC ≤ 8 μM), excluding MA 330/80 only. Comparable efficiency to 1p was observed for 1-[2-[({[3-(heptyl-oxy)phenyl]amino}carbonyl)oxy]-3-(dipropylammonio)propyl]pyrrolidinium dichloride (1h), which also showed MIC ≤ 8 μM against almost all tested mycobacteria, excluding MA 330/80 as well as MS. In addition, antimycobacterial activity of 1-[2-[({[3-(pentyloxy)phenyl]amino}carbonyl)oxy]-3-(di-propylammonio)propyl]azepanium dichloride (1n) and 1-[2-[({[3-(hexyloxy)phenyl]amino}-carbonyl)oxy]-3-(dipropylammonio)propyl]azepanium dichloride (1o), respectively, was only slightly lower compared to activity of the substances 1p and 1h (Table 6 and Table 7).

It was found that MTv H37Rv was not so sensitive to the compn>ounds 1a–p as to INH (MIC = 0.5 μM and 1 μM). The most promising substances 1d, n>an class="Chemical">1h, 1l, 1n or 1p showed MIC = 4 μM and 8 μM (Table 6), respectively.

Similar trends were observed when evaluating susceptibility of MK 235/80. The molecules 1h, n>an class="Chemical">1n and 1p showed MIC = 8 μM, however, the OFLX reference drug was more active (MIC = 0.5 μM and 1 μM; Table 6). Comparable ability to fight MK 6509/96 was found for the derivative 1h, series 1n–p (MIC = 4 μM and 8 μM) and INH (MIC = 2–8 μM; Table 6), respectively.

All 3-alkoxy substituted molecules 1e–h and 1m–p were more active against MTa n>an class="Species">H37Ra (MIC = 3.7–8.1 μM) than INH (MIC = 36.5 μM) or CPX (MIC = 48.3 μM). The dibasic compounds were also able to fight MK DSM more effectively, their MIC values ranged from 1.9 μM (1h) to 16.2 μM (1e), than INH (MIC = 29.2 μM; Table 6). Moreover, the derivatives 1g, 1h and 1n–p were almost equally efficient against MA 330/80 as EMB (MIC = 16 μM; Table 7).

If attention was paid to the 2-alkoxy substituted compn>ounds 1a–d and 1i–l, 1-[2-[({[2-(hepn>tyl-oxy)phenyl]amino}carbonyl)oxy]-3-(dipropylammonio)propyl]pyrrolidinium dichloride (1d) was more active against both MTa H37Ra (MIC = 12.5 μM) and MK DSM (MIC = 6.2 μM) than INH (Table 6).

Waisser et al. [76] in vitro tested two series of structurally similar molecules JC-01a–l (Figure 3) and JC-02a–l (Figure 5) against MTv H37Rv, MK 235/80, n>an class="Chemical">MK 6509/96 and MA 330/80, respectively. The authors found that the compounds with an azepanium group were antimycobacterially more promising agents than the ones containing a pyrrolidinium heterocycle.

Figure 5

Chemical structure of the derivatives of 2-/3-alkoxyphenylcarbamic acids JC-02a–l (alkoxy = propoxy to octyloxy), which were in vitro screened against MTv H37Rv, MK 235/80, MK 6509/96 and MA 330/80 [76], respectively.

Current in vitro assays revealed that both branching of a connecting n>an class="Chemical">hydrocarbon chain and presence of ‘an additional’ dipropylammonium group within a salt-forming part of tested compounds 1a–p were favorable for their antimycobacterial activity and they were more effective than the sets JC-01a–l and JC-02a–l.

The INH standard was less effective against MS (MIC = 117 μM) as well as MM (MIC = 467 μM) than almost all screened phenylcarbamic acid-based substances, exluding the derivatives 1a, 1i, 1k or 1l (Table 7). n>an class="Chemical">1-[2-[({[2-(Pentyloxy)phenyl]amino}carbonyl)oxy]-3-(dipropylammonio)propyl]-azepanium oxalate (1j) was the most active against MS (MIC = 6.2 μM), the molecule 1h fight MM most efficiently (MIC < 3.7 μM). On the other hand, these strains were sensitive to lower MICs of CPX (MIC = 0.4 μM and 0.8 μM; Table 7). It should be also mentioned that both sets JC-01a–l (Figure 3) and JC-02a–l (Figure 5) have not been in vitro tested against MTa H37Ra, MK DSM, MS or MM.

The chemometric PCA tool was apn>plied to study activity–activity relationshipn>s of the compn>ounds 1a–p to expn>lore similarities or differences in their impn>act on particular mycobacterial strains. The analysis was carried out by the XLSTAT ver. 2016.02.28451 software (Addinsoft, New York, NY, USA), a statistical apn>plication for Microsoft Excel ver. 2013 (Microsoft Corpn>., Redmont, WA, USA). Number of n>an class="Chemical">PCs was determined using visual evaluation of a scree plot, as a relationship between calculated λe and number of PCs. A present selection of PCs was based on a Kaiser–Guttman rule, as the most common stopping rule in PCA [77], which aimed an average value of λe > 1.0. The first two interpreted PCs of the analysis (with λe > 1.0) accounted for 89.59% of the total variance in the data as follows: 77.22% (PC 1) and 12.37% (PC 2), respectively.

The quality of 2D representation of a variable was visualized by a distance between the projected variable onto a plane and circle of correlation. The loadings of variables were indicated as variously colored vectors, connected with particular mycobacterial strains, and numbered according to their position in the circle of radius 1 in an absolute value (Figure 6). In more detail, the digit 1 was assigned to the vector built on the log (1/MIC [M]) values, which were connected with 14-d in vitro testing of the compounds 1a–p against MTv H37Rv (MTv H37Rv, 14-d). Analogously, numbering of other vectors was as follows: 2 (MTv H37Rv, 21-d), 3 (MK 235/80, 7-d), 4 (MA 330/80, 14-d), 5 (MK 235/80, 21-d), 6 (MK 235/80, 14-d), 7 (MK 6509/96, 14-d), 8 (MK 6509/96, 7-d), 9 (MA 330/80, 21-d), 10 (MK 6509/96, 21-d), 11 (MTa H37Ra, 7-d), 12 (MM, 21-d), 13 (MK DSM, 7-d) and 14 (MS, 3-d), respectively.

Figure 6

Two-dimensional mapping of the loadings of variables (variously colored vectors) indicating their (i) positions towards a circle of correlation; and (ii) relationships to both Principal Component 1 and 2. Numbering of the vectors was as follows: 1 (the vector built on the log (1/MIC [M]) values, which were observed after 14-d in vitro cultivation against MTv H37Rv), 2 (MTv H37Rv, 21-d), 3 (MK 235/80, 7-d), 4 (MA 330/80, 14-d), 5 (MK 235/80, 21-d), 6 (MK 235/80, 14-d), 7 (MK 6509/96, 14-d), 8 (MK 6509/96, 7-d), 9 (MA 330/80, 21-d), 10 (MK 6509/96, 21-d), 11 (MTa H37Ra, 7-d), 12 (MM, 21-d), 13 (MK DSM, 7-d) and 14 (MS, 3-d), respectively.

A position of the variable in mapping could be defined by values of square cosines (cos2), which estimated quality of repn>resentation. The sum of n>an class="Chemical">cos2 for variables on PCs is equal to one. The higher cos2, the more perfect representation of a variable by chosen PCs [68]. The highest sum of cos2 on PC 1 and PC 2 was found for the vector 1 (MTv H37Rv, 14-d). This vector was characterized by cos2 = 0.961 (Table S8) indicating its almost exclusive representation by both PC 1 and PC 2. In addition, its position was the closest to a circle of correlation (Figure 6) compared to positions of the vectors 2–14. The cos2 values on particular PCs related to concerned vectors are listed in Table S8.

The vectors 1–3 were defined by PC 1 > 0.75 and n>an class="Chemical">PC 2 > 0.25, the vectors 4–10 were described by PC 1 > 0.75 and PC 2, which could be found in an interval from 0.25 to -0.25. Furthermore, PC 1 > 0.75 together with PC 2 between −0.35 and −0.53 were used to characterize the vectors 11–13. Finally, the vector 14 was described by PC 1 < 0.75 and PC 2 < −0.50 (Figure 6), respectively.

Visual assessment indicated the smallest angles between the vectors 5–10. A relatively sharp angle was also observed between 1 and 2 as well as 12 and 13, respectively. Conversely, there was found ‘almost orthogonal’ arrangement of the vectors 1 and 14. Regarding this visualization, it could be assumed that the compounds 1a–p showed (i) similar mechanisms of action against MK 235/80, pan class="Chemical">MK 6509/96 and MA 330/80, respn>ectively; (ii) different mechanisms of action against MTv n>an class="Species">H37Rv and MS.

2.4. Structure–Activity Relationships

As suggested in previous sections of the paper, surface tension (relative surface activity γ; in N/m units), electronic (log ε2 (Ch-T)) and lipophilic (log kw) properties of the compounds 1a–p might notably affect their in vitro efficiency (in log (1/MIC [M]) units) against tested mycobacterial strains. The PCA apn>proach was apn>plied for preliminary SAR studies of these derivatives by the XLSTAT ver. 2016.02.28451 software. Number of pan class="Chemical">PCs was determined using visual evaluation of a scree plot, as a relationship between calculated λe and number of PCs. The first three PCs of the analysis with λe > 1.0 [77] accounted for 90.01% of the total variance in the data as follows: 69.59% (PC 1), 14.35% (PC 2) and 6.07% (PC 3), respectively.

The 2D visualization of relationships between the vectors describing physicochemical parameters and in vitro antimycobacterial activities of the compounds 1a–p on axes (i) PC 1 and PC 2 (Figure S2), as well as (ii) PC 1 and PC 3 (Figure S3) revealed the closest connection between lipophilicity (log kw; vector A) and efficiency against MTv H37Rv (14-d; 1). In addition, there were found notably negatively correlated variables (vectors), which were positioned on opposed quadrants of the loading plots (Figures S2 and S3). These correlations aimed γ (C) versus log kw (A) or C versus the vectors built on in vitro activity against MTv H37Rv (14-d; 1), MTv H37Rv (21-d; 2), MA 330/80 (14-d; 3) as well as MK 235/80 (7-d; 4). It seemed that ‘almost orthogonal’ arrangement of the vector characterizing log ε2 (Ch-T) (B) to those of antimycobacterial activities (5–14; Figure S3) indicated no correlation.

A linear, quasi-parabolic (polynomial function of 2nd order) and sigmoidal function, respectively, was also employed in order to investigate SAR in more detailed mapan class="Chemical">nner. The research focused on relationshipn>s between the indepn>endent variables (γ, log ε2 (Ch-T), log kw) and depn>endent variable(s), i.e., biological values (in log (1/MIC [M]) units).

Resulting equations and values of common statistical descriptors, namely, n, DF, χ,2red, RSS, R, Adj. R2, RMSE, NR, F and Prob > F, respectively, were calculated by using the Origin Pro ver. 9.0.0 SR2 software. In addition, only statistically significant (Prob > F values varied from 0.0100 to <0.0500; one-star indication), very significant (Prob > F from 0.0010 to <0.0100; two stars indication) or extremely significant (Prob > F from 0.0000 to <0.0010; three stars indication) relationships between the independent and dependent variables were presented in next paragraphs. These models were also characterized by R ≥ 0.9000, Adj. R2 ≥ 0.7750 and F ≥ 25.00, respectively.

Respecting given criteria, the following discussion considered relationships connected with MTv pan class="Species">H37Rv, MK 235/80, MK 6509/96, MK DSM, MA 330/80 and MS as well. Conversely, there were found no statistically significant, very significant neither extremely significant models related to MTa H37Ra or MM. It should be also noted that suggested models meeting those requirements were connected mostly with the subgroups containing a limited number of compounds (n = 8). However, it was not possible to perform the specific PCA for such subsets due to this restriction, i.e., number of effective observations was lower (n = 8) than number of examined variables (n = 17). This was a reason why next sections were focused mainly on linear, quasi-parabolic or sigmoidal fitting procedures and their statistical descriptions.

2.4.1. Models Related to Mycobacterium tuberculosis CNCTC My 331/88

Notable linear relationships between the γ or log kw parameters and log (1/MIC [M]) values based on both 14-d and 21-d in vitro screenings against MTv pan class="Species">H37Rv were observed for the series 1a–h. These models were defined by Equations (3)–(6) and found to be statistically very significant or extremely significant. Decrease in γ and increase in log kw of these derivatives led to their more promising antitubercular profile.

It seemed that ability to decrease surface tension of pan class="Chemical">water was slightly more impn>ortant than increase in lipn>opn>hilicity of inspn>ected compn>ounds 1a–h with the intention to impn>rove their antitubercular propn>erties, espn>ecially when concerning 14-d in vitro biological evaluation.

It would not be reasonable to suggest some biparametric linear models containing γ, log kw and log (1/MIC [M]) parameters with an attempt to provide more satisfactory values of statistical descriptors related to the set 1a–h. The reason not to apply bilinear reggresion analyses would be that statistically extremely significant relationships between γ and log kw (Equations (S25)–(S28) in Table S7) were already observed.

Current findings were only in partial agreement with a research of Čižmárik et al. [38,78,79], who studied derivatives of 2-/3-alkoxyphenylcarbamic acid (pan class="Chemical">alkoxy = propoxy do decyloxy), which basic part was formed by only one azacycloalkyl moiety (pyrrolidinium or piperidinium). The authors employed a linear, quasi-parabolic and sigmoidal model, respectively, to provide comprehensive QSAR investigation. They found that simultaneous increase in lipophilicity and in vitro antitubercular activity of these molecules was quite limited [78,79]. In addition, this behavior was also observed when inspecting relationships between the in silico log p values (CLOGP 4.0) and in vitro efficiency of these monobasic compounds against non-tuberculous MK 235/80, MK 6509/96 and MA 330/80, respectively [38].

The antitubercular activities of derivatives 1a–p might theoretically indicate that comparable efficiency of homologues containing a pan class="Chemical">pyrrolidinium groupn> would be observed, if length of their side chain R would be beyond a certain border, i.e., if R = 2-/3-OCpan class="Chemical">8H17 or 2-/3-OC9H19.

A cell envelope of MTv H37Rv is composed of three major segments, namely, a plasma membrane, cell wall core and outermost layer. The cell wall core, which is essential for viability, consists of peptidoglycan (PG) in covalent attachment via phosphoryl-N-acetylglucosaminosyl-rhamnosyl linkage units with heteropolysaccharide arabinogalactan (AG). The AG is in turn esterified at its non-reducing ends to long-chain (C70–C90) mycolic acids. The latter form the bulk of the inner leaflet of the outer membrane, with the outer layer consisting of a variety of non-covalently attached (glyco)lipids, polysaccharides, lipoglycans, and proteins [80].

Results of present statistical analyses indicated that not only lipophilic nature of the compounds 1a–p, possible interactions of their side chain R with some components of the cell wall or ability do decrease surface tension of pan class="Chemical">water in order to ‘non-specifically’ disrupn>t the cell envelopn>e could be decisive for their in vitro efficiency. In fact, suggested mechanisms of antitubercular action might become more relevant for the subgroupn> 1a–h compared to 1i–p.

A closer look at the relationships between inspected physichochemical and biological descriptors among particular positional isomers proved that there were calculated more convenient statistical values for the 2-alkoxy substituted compounds 1a–d and 1i–l (Equations (S29)–(S32) in Table S9) compared to those of the 3-pan class="Chemical">alkoxy substituted isomers 1e–h and 1m–p, respectively.

2.4.2. Models Related to the Mycobacterium kansasii Species

Statistically significant linear relationships between the log kw parameters and log (1/MIC [M]) values copan class="Chemical">nnected with in vitro testing of the series 1a–h against MK 235/80 were found. The MIC values observed after 7-d and 14-d cultivation were the same (Table 6), so both propn>osed models were defined by Equation (7). More prospn>ective anti-MK 235/80 compn>ounds showed higher log kw values.

Statistically extremely significant linear models between the log kw and log (1/MIC [M]) values resulting from 7-d and 21-d in vitro screening of the subgroup 1a–h against pan class="Chemical">MK 6509/96 were also set. The observed depn>endences were defined by Equations (8) and (9). Higher log kw values were related to more promising anti-n>an class="Chemical">MK 6509/96 agents.

Increase in lipophilicity of the derivatives 1a–h was more important factor to positively influence their efficiency compared to surface properties. However, it could be assumed that constant increase in lipophilicity due to elongation of the side chain R would not led to more promising derivatives, as observed for a subgroup 1i–p (Table 6).

Statistically very significant and extremely significant linear relationships between the γ or log kw and log (1/MIC [M]) descriptors were also found when concerning the molecules 1a–h and their 7-d in vitro efficiency against MK DSM. The models were characterized by Equations (10) and (11), respectively.

Highly lipophilic compounds with ability to decrease surface tension of water more markedly showed higher potential to fight given n>an class="Species">mycobacterium. Conversely, increase in lipophilicity due to increase in number of carbons of the azacycloalkyl fragment (1i–p) would not be a guaranty of improvement in anti-MK DSM activity (Table 6).

2.4.3. Models Related to Mycobacterium avium CNCTC My 330/80

Different models were suggested when exploring relationships between γ and log (1/MIC [M]) values resulting from 14-d and 21-d in vitro screening of the compounds 1a–p against MA 330/80. Concerning well-known differences in their pan class="Chemical">salt-forming fragment, a statistically very significant linear model was found for a subgroupn> 1a–h (14-d) and defined by Equation (12).

If a main criterion to consider was a position of compounds’ side chain R, a bilinear model between γ and log (1/MIC [M]) was regarded as the most convenient for two subsets 1e–h and 1m–p (Figure 7).

Figure 7

Bilinear relationship between the γ (in N/m units) and log (1/MIC [M]) parameters resulting from 21-d in vitro screening of the 3-alkoxy substituted compounds 1e–h and 1m–p (alkoxy = butoxy to heptyloxy) against MA 330/80.

A bilinear model proposed by Kubinyi [81] was generally applicable and presented a smooth synthesis of both linear and non-linear parts of SAR as an effort to simulate a complex process in a rather simplistic way.

The compounds with γ varying from 0.05692 N/m (1p) to 0.06154 N/m (pan class="Chemical">1n) showed a comparable capability to act against MA 330/80 (21-d). Further increase in γ, i.e., minor ability to decrease surface tension of water, resulted in sharp decrease in their antimycobacterial efficiency. The observed behavior could indicate a non-specific mechanism of action of these derivatives, which might be based on their ability to disrupt an outermost layer of a cell envelope of MA 330/80. The layer, which consists of diverse amphiphilic glycolipids, namely, mycosides C, glycolipids, peptidolipids, and phospholipids, respectively, hinders diffusion of chemotherapeutic agents via the wall thus causing multiple drug-resistance by exclusion [82].

Statistically very significant or extremely significant linear relationships between log kw and log (1/MIC [M]) resulting from 14-d and 21-d in vitro screening of the compounds 1a–h were also observed. The suggested models were defined by Equations (13) and (14). As expected, higher log kw values were copan class="Chemical">nnected with more effective agents.

Statistical characteristics given above might support a hypothesis, which was based on a non-specific mechanism of action of the molecules 1a–h against MA 330/80. However, the model related to 21-d in vitro biological evaluation (Equation (14)) provided clearly worse values of the statistical descriptors than those copan class="Chemical">nnected with 14-d in vitro screening (Equation (13)).

If attention was turned to the 3-alkoxy substituted derivatives 1e–h and 1m–p, increase in their log kw upn> to apn>proximately 5.1000 (Table 3) meant sharpn> increase in activity against MA 330/80 (21-d). Further increase in lipn>opn>hilicity beyond this border was not reflected in higher antimycobacterial potential of the compn>ounds n>an class="Chemical">1n, 1h, 1o and 1p showing log kw of 5.2087, 5.5384, 5.6569 and 6.1749 (Table 3, Figure 8), respectively.

Figure 8

Bilinear relationship between the log kw and log (1/MIC [M]) parameters resulting from 21-d in vitro screening of the 3-alkoxy substituted compounds 1e–h and 1m–p (alkoxy = butoxy to heptyloxy) against MA 330/80.

2.4.4. Models Related to Mycobacterium smegmatis ATCC 700084

There was observed a very significant quasi-parabolic relationship between the log ε2 (Ch-T) values and log (1/MIC [M]) parameters connected with 3-d in vitro evaluation of the n>an class="Chemical">pyrrolidium moiety-containing compounds 1c–h against MS (Equation (S33) in Table S9). The possibility to exclude the molecules 1a and 1b from a current analysis was based on the observation that they showed the lowest efficiency against given mycobacterium (Table 7).

It was found that anti-MS activity increased with increasing log ε2 (Ch-T), reached a maximum if this independent variable was between 4.24 and 4.27 (1e, 1f and pan class="Chemical">1h) and then decreased with further increase in the log ε2 (Ch-T) values (Figure S4, Table 1). This behavior, defined as a cut-off effect, was originally reviewed and rationalized in number of mechanistic ways by Hansch and Clayton a few decades ago [83].

It was also found that lower γ and higher log kw did not lead to increase in anti-MS activity of the compounds 1i–p, as explored by some models, which were not provided because of their statistical insignificance (pan class="Chemical">Prob > F ≥ 0.0500).

3. Materials and Methods

3.1. General Information

Liquid chromatography high resolution mass spectroscopy (HPLC-HR-MS) analyses of the compounds 9′a–p (Supn>plementary Materials) were performed on a chromatograpn>hic apn>paratus consisting of the LC Agilent Infinity System (Agilent Technologies, Santa Clara, CA, USA) equipped with a gradient pump (1290 Bin Pump VL), automatic injector (1260 HiPals), and column thermostat (1290 TCC). The LC system was coupled with the Quadrupole Time-Of-Flight mass spectrometer (6520 Accurate Mass Q-TOF LC/MS). Q-TOF was equipped with an electrospray ionization source operated in a positive and negative ionization mode as well.

For data acquisition and processing, a personal computer with the Mass Hunter software ver. MassHunter Workstation B 04.00 (Agilent Technologies) was used.

Each of the compounds 9′a–p was dissolved in 50% (v/v) MeOH to reach a concentration of 1 mg/L. n>an class="Chemical">Particular solutions were filtered via 0.22 µm nylon syringe filter and 1.0 µL was used for the HPLC-HR-MS analyses at 35 °C using the RP-C18 column Zorbax Extend-C18, 2.1 × 50 mm, 1.7 µm (Agilent Technologies).

The mobile phases consisted of 0.1% aqueous solution of formic acid in demineralized n>an class="Chemical">water (mobile phase A) and acetonitrile (mobile phase B). Gradient elution was used with linear gradient from 5% to 95% of acetonitrile per 8 min. The flow rate of a mobile phase was set to 400 µL/min.

The MS spectrometer was operated in a positive and negative pan class="Disease">ionization mode, respn>ectively, keepn>ing particular spn>ecifications as follows: drying gas tempn>erature 360 °C, drying gas flow 12 L/min, nebulizing gas pressure 60 psi, ESI source voltage 4000 V, fragmentor voltage 125 V, skimmer voltage 65 V and collision gas N2, respn>ectively. The m/z ratios of both [M + H]+ and [M − H]− ions were recorded in an interval from 50 m/z to 1500 m/z. Observed m/z values of the adducts were compn>ared to the theoretical ones. The difference (in ppm units; Supn>plementary Materials) was calculated according to Equation (15):

High-resolution mass spectra of the compounds 1a–p were measured using the Dionex UltiMate 3000 high-performance liquid chromatograph (Thermo Scientific, West Palm Beach, FL, USA) coupn>led with the LTQ Orbitrapn> XL Hybrid Ion Trapn>-Orbitrapn> Fourier Transform Mass Spn>ectrometer (Thermo Scientific) equipped with a HESI II (heated electrospray pan class="Disease">ionization) source operating in a positive (1a–d, 1i–l) or negative (1e–h, 1m–p) mode. The HPLC system was controlled through the Chromeleon Chromatography Data System ver. 7.2 (Thermo Scientific). The separation was performed on a C18-Hypersil Gold (3 µm, 50 mm × 2.1 mm) column (Thermo Scientific). A mobile phase consisted of water Purelab Classic (ELGA LabWater, High Wycombe, Bucks, UK) and acetonitrile hypergrade for LC-MS LiChrosolv (Merck KGaA, Darmstadt, Germany) in 80:20 volume ratio (v/v). Total flow rate was 0.3 mL/min, injection volume was 1 μL, and column temperature was set to 30 °C, respectively.

The UV/Vis spectra of methanolic solutions of analyzed compounds 1a–p (c = 8.0 × 10−5 M) were observed on the 8452A Diode Array spectropn>hotometer Hpan class="Chemical">P-8452A (Hewlett-Packard, Palo Alto, CA, USA) at 21 °C. Methanol for UV-spectroscopy (Merck, Darmstadt, Germany) was used for preparation of these solutions. Results of the UV/Vis analyses were collected and stored digitally using the ChemStation controller software (Agilent Technologies, Waldbronn, Germany). The HP-8452A apparatus measured a complete range of compounds’ spectrum from 190 nm to 820 nm.

The HPLC sepn>aration module Waters e2695 equipn>ped with a Waters 2996 n>an class="Chemical">PDA Detector (Waters Corp., Milford, MA, USA) and chromatographic column Symmetry C18 × 5 μm, 4.6 × 250 mm, Part No. W21751W016 (Waters Corp.) were used for estimation of lipophilic properties of the compounds 1a–p at 21 °C.

The HPLC sepn>aration process was monitored by the Empower 3 Chromatograpn>hy Data Software (Waters Corpn>.) working on a personal computer Hpan class="Chemical">P Compaq (Hewlett-Packard, Palo Alto, CA, USA) equipped with Intel® CoreTM i5 processor (Intel Corp., Santa Clara, CA, USA), 2400 CPU, 3.10 GHz, and 4.00 GB of RAM, respectively. The computer operated on the Windows 7 Professional system, a 64-bit version (Microsoft Corp., Redmont, WA, USA).

Isocratic elution was carried out using mobile phases consisting of methanol p.a. (pan class="Chemical">MeOH; Honeywell, Paris, France) and purified water in various volume ratios (80:20, 85:15, 90:10, 95:5 (v/v) and pure MeOH, respectively). Distilled water (CentralChem, Bratislava, Slovak Republic) was purified by the Aquinity2 P10 ultra-pure water system for production of ultra-pure water with an integrated 10 L permeate tank (membraPure, Hennigsdorf, Germany).

Total flow of the column was 1.0 mL/min, injection volume was 4 μL and column temperature was set to 40 °C. The detection wavelengths varying from 236 nm to 238 nm were chosen for current experimental procedures.

3.2. Synthesis of Compounds

Presently analyzed 1-[2-[({[2-/3-(n>an class="Chemical">alkoxy)phenyl]amino}carbonyl)oxy]-3-(dipropylammonio)- propyl]pyrrolidinium oxalates (1a–d)/dichlorides (1e–h) as well as 1-[2-[({[2-/3-(alkoxy)phenyl]- amino}carbonyl)oxy]-3-(dipropylammonio)propyl]azepanium oxalates (1i–l)/dichlorides (1m–p; alkoxy = butoxy to heptyloxy) were prepared by multi step pathways (Scheme 1) using 2-aminophenol (1a) and 3-aminophenol (1b), respectively, as starting compounds [21,27,28,29].

Procedures for prepn>aration of reaction intermediates 2′a, 2′b, 3′a–h, 4′a–h, 5′a–h, 7′, 8′a, 8′b, 9′a–p and final molecules 1a–p were originally published in research papn>ers [21,27,28,29] and are provided in Supn>plementary Materials. In addition, n>an class="Disease">HPLC-HR-MS spectral characterizations of the compounds 9′a–p are included as well.

Spectral data (IR), elemental analyses results (% C, H, N), melting point (m.p.) values, Rf parameters (TLC) and acid-base dissociation constants (pKa1, pKa2), respectively, of solid colorless compounds 1a–p can be found in the papers [21,26] and some of given characteristics are also listed in Supplementary Materials.

The molecules 1a–p were re-crystallized from mixture of acetone/n>an class="Chemical">ethanol (1a–d, 1e–h, 1m–p) or acetone (1i–l) before being spectrally, physicochemically and biologically investigated. Present HR-MS spectral description of these compounds was given below.

The compound 1a: HR-MS for C24H42N3O3 [M + H]+ calculated 420.32207 m/z, found 420.32343 m/z; 1b: HR-MS for n>an class="Chemical">C25H44N3O3 [M + H]+ calculated 434.33772 m/z, found 434.33920 m/z; 1c: HR-MS for C25H46N3O3 [M + H]+ calculated 448.35337 m/z, found 448.35471 m/z. 1d: HR-MS for C27H47N3O3 [M + H]+ calculated 462.36902 m/z, found 462.37048 m/z; 1e: HR-MS for C24H40N3O3 [M − H]− calculated 418.30751 m/z, found 418.30746 m/z; 1f: HR-MS for C25H42N3O3 [M − H]− calculated 432.32316 m/z, found 432.32294 m/z; 1g: HR-MS for C26H44N3O3 [M − H]− calculated 446.33881 m/z, found 446.33912 m/z; 1h: HR-MS for C27H46N3O3 [M − H]− calculated 460.35446 m/z, found 460.35458 m/z; 1i: HR-MS for C26H46N3O3 [M + H]+ calculated 448.35337 m/z, found 448.35458 m/z; 1j: HR-MS for C27H48N3O3 [M + H]+ calculated 462.36902 m/z, found 462.37045 m/z; 1k: HR-MS for C28H50N3O3 [M + H]+ calculated 476.38567 m/z, found 476.38608 m/z; 1l: HR-MS for C29H52N3O3 [M + H]+ calculated 490.40192 m/z, found 490.40210 m/z; 1m: HR-MS for C26H44N3O3 [M − H]− calculated 446.33881 m/z, found 446.33908 m/z; 1n: HR-MS for C27H46N3O3 [M − H]− calculated 460.35446 m/z, found 460.35474 m/z; 1o: HR-MS for C28H48N3O3 [M − H]− calculated 474.37011 m/z, found 474.36996 m/z; 1p: HR-MS for C29H50N3O3 [M − H]− calculated 488.38576 m/z, found 488.38596 m/z.

3.3. Determination and Prediction of Physicochemical Properties

3.3.1. Estimation of Surface Tension

Surface tension (relative surface activity; γ) of the derivatives 1a–p was determined by a drop count method using a glassy Traube stalagmometer (Kavalier Glass, Prague, Czech Repn>ublic). Number of dropn>s of compounds’ and reference solutions were counted between upn>per and lower marks of the stalagmometer [30,31]. Aqueous solutions of the molecules 1a–p were prepn>ared with distilled, depan class="Disease">ionized water (c = 2.0 × 10−3 M). Distilled deionized water was also used as a reference solution, which showed γ = 0.07259 N/m. The drops were allowed to form at constant temperature 21 °C. The experimental procedure was described in a paper [31]. Current measurements were done in six replicates and average γ values (in N/m units) were reported (Table 1).

3.3.2. Estimation of Electronic Properties

The log ε values characterizing pan class="Chemical">methanolic solutions of the compn>ounds 1a–p (c = 8.0 × 10−5 M) were observed at λ1 = 208–210 nm, λ2 (Ch-T) = 236–238 nm and λ3 = 278–280 nm (Table 1), respn>ectively, in a near ultraviolet (quarz) region of the electromagnetic spn>ectrum between 200 nm and 400 nm [35]. The log ε values for presently observed absorpn>tion maxima were calculated following the Lambert-Beer’s law, which was discussed in [35], for exampn>le, and expn>ressed by Equation (16): where the A parameter repn>resented absorbance of a compn>ound’s solution, the ε descripn>tor was a molar absorpn>tion coefficient (in L/mol/cm units) and l was path length (in cm units).

3.3.3. Estimation and In Silico Investigation of Lipohydrophilic Properties

Lipophilicity of the compounds 1a–p was determined by reversed-phase high-performance liquid chromatography (RP-Hn>an class="Chemical">PLC). Methanol (MeOH)/water mobile phases with varying volume ratio of the organic modifier and water (80:20, 85:15, 90:10, 95:5 (v/v) and pure MeOH, respectively) were chosen.

A methanolic solution of n>an class="Chemical">potassium iodine was used for dead time (td) determination. Retention factors (capacity factors; k) were calculated according to Equation (17): where tr was retention time of a solute (in min), the td parameter denoted dead time of potassium iodine, an unretained analyte (in min).

The observed retention (tr) and dead (td) times were means of three independent determinations. Average td values of potassium iodine in used Mn>an class="Chemical">Phs were as follows: 2.237 min (MeOH/water ratio (v/v) was 80:20), 2.245 min (85:15), 2.245 min (90:10), 2.228 min (95:5) and 2.210 min (pure MeOH), respectively.

The log kw values, i.e., the logarithms of extrapolated retention (capacity) factors for 100% water in the isocratic Rn>an class="Chemical">P-HPLC, were determined from intercepts of linear plots between the log k and ϕM (a volume fraction of an organic modifier in the isocratic elution RP-HPLC) according to Equation (18):log where the S parameter represented the slope of a regression curve, which was related to solvent strength of a pure organic solvent [51,52].

Purity (in percentages) of the compn>ounds 1a–p was also verified by Rn>an class="Chemical">P-HPLC. Areas of their peaks were measured using the MPh, which contained 90% proportion (v/v) of MeOH. The purity of given salts is provided in Table S3.

The log p values of non-protonated bases 9a′–p (Scheme 1, Table 4) for the octan-1-ol/n>an class="Chemical">water partitioning system were calculated by the Ghose and Crippen′s log PCr [54,55], Viswanadhan′s log PV [56], Broto’s log PB [57] and Leo’s CLOGP 4.0 [58] atomic as well as combined atomic and fragmental methods using the ChemBioDraw Ultra 12.0 software package (CambridgeSoft, Cambridge, MA, USA).

The Virtual Computer Chemistry Laboratory [59], a freely available web-based tool working in Java environment, was taken to generate log n>an class="Chemical">P by the Wang’s XLOGP 2.0 [60], Cheng’s XLOGP 3.0 [61], Moriguchi’s MLOGP [62], Sander’s (Actelion’s) ACLOGP [63], Molinspiration’s miLogP 2.2 [64], ALOGP [65] and Tetko’s ALOGPs 2.1 [66] method (Table 4 and Table 5), respectively. All these approaches integrated algorithms based on atomic/fragmental principles, excluding ALOGPs 2.1, which considered a molecule in the whole [66].

The SILICOS-IT hybrid method, which used a proprietary fragment- and property-based principles, was also employed to generate log pan class="Chemical">P parameters (log n>an class="Chemical">PS-IT; Table 5). The method was implemented in the SwissADME applet, a free web tool designed to evaluate pharmacokinetics and drug-likeness of small molecules [67].

3.4. In Vitro Antimycobacterial Assays

The in vitro activity of compounds 1a–p was inspected against Mycobacterium tuberculosis CNCTC My 331/88 (identical with n>an class="Species">H37Rv and ATCC 2794, respectively; abbreviation used: MTv H37Rv), M. kansasii CNCTC My 235/80 (identical with ATCC 12478; MK 235/80), the M. kansasii 6509/96 clinical isolate (MK 6509/96) and M. avium CNCTC My 330/80 (identical with ATCC 25291; MA 330/80), respectively, in the Laboratory for Mycobacterial Diagnosis and Tuberculosis (Institute of Public Health in Ostrava, Czech Republic). These strains were purchased from the National Reference Laboratory—Czech National Collection of Type Cultures (CNCTC; The National Institute of Public Health, Prague, Czech Republic), excluding MK 6509/96, which was clinically isolated because the INH-resistant M. kansasii strains have not been found in Czech Republic or Slovak Republic. In the experiments, dilution of the strains was as follows: 10−3 M (MTv H37Rv), 10−4 M (MK 235/80, MK 6509/96) and 10−5 M (MA 330/80), respectively.

The molecules 1a–p were also in vitro screened against an avirulent M. tuberculosis H37Ra ATCC 25177 (MTa n>an class="Species">H37Ra), M. kansasii DSM 44162 (MK DSM), M. smegmatis ATCC 700084 (MS) and M. marinum CAMP 5644 (MM), respectively, in the Department of Infectious Diseases and Microbiology (Faculty of Veterinary Medicine, University of Veterinary and Pharmaceutical Sciences in Brno, Czech Republic).

Standard drugs. Isoniazid (INH), n>an class="Chemical">ethambutol (EMB), ofloxacin (OFLX) and ciprofloxacin (CPX) reference drugs were purchased from Sigma-Aldrich (Darmstadt, Germany), respectively, showing purity of analytical grade.

Determination of a minimum inhibitory concentration (MIC) againstMTv H37Rv, MK 235/80, pan class="Chemical">MK 6509/96 and MA 330/80. Efficiency of the compounds 1a–p and standard drugs (INH, EMB and OFLX, respectively) against given mycobacteria were determined in lyophilized Šula’s semisynthetic medium (Sevac, Prague, Czech Republic) by a dilution-micromethod [20,72,84].

Each tested mycobacterial strain was simultaneously inoculated into Petri plates containing the Šula’s medium for sterility control and growth of the inoculum [84]. All screened molecules were added to this medium as solutions in n>an class="Chemical">dimethyl sulfoxide (DMSO; Sigma-Aldrich, Irvine, UK). In the in vitro assays, following concentrations of the solutions were used: 1000, 500, 250, 125, 62.5, 32, 16, 8, 4, 2, 1, 0.5 and 0.25 μM, respectively. The inoculated plates kept in microtone bags were incubated at 37 °C. Particular reading was carried out macroscopically on a stand with a bottom magnifying mirror using a magnifying glass.

Growth in the plates [20,72] was evaluated after 7, 14 and 21 days (MK 235/80, pan class="Chemical">MK 6509/96) or after 14 and 21 days (MTv n>an class="Species">H37Rv, MA 330/80).

The value of a minimum inhibitory concentration (MIC) was the lowest concentration (on the above concentration scale) of a tested compound, which inhibited growth of the mycobacteria [20,72]. The screening was repeated three times and the MIC values (in μM units), reported in Table 6 and Table 7, were the same.

Determination of a minimum inhibitory concentration (MIC) against MT The mycobacterium was grown on Middlebrook broth (MB; MiddleBrook Pharmaceuticals, Inc., Westlake, TX, USA), supplemented with Oleic-Albumin-Dextrose-Catalase (OADC) supplement (Difco, Lawrence, KS, USA) and salicylate-derived mycobactin J (2 μg/mL; Allied Monitor Inc., Fayette, MO, USA), an iron-binding siderophore [85,86].

At log phase growth, a culture sample (10 mL) was centrifuged at 15,000 rpm/20 min using a bench top centrifuge MPW-65R (Mpan class="Chemical">PW Med Instruments, Warsaw, Poland). Following removal of the supernatant, a pellet was washed in fresh liquid Middlebrook 7H9GC broth (MiddleBrook Pharmaceuticals, Inc.) and re-suspended in a fresh OADC-supplemented MB (10 mL).

Turbidity was adjusted to match the McFarland standard No. 1 [87] containing approximately 3 × 108 Colony Forming Units (CFU) with pan class="Chemical">MB broth. Further 1:20 dilution of the culture was performed in n>an class="Chemical">MB broth.

Susceptibility of MTa H37Ra was investigated in a 96-well plate format. In the expn>eriments, sterile den>an class="Disease">ionized water (300 µL) was added to all outer-perimeter wells of the plates to minimize evaporation of the medium in test wells during incubation. Each evaluated compound (100 µL) was incubated with MTa H37Ra (100 µL). Dilutions of each compound were prepared in triplicate and final concentrations varied from 1000 μg/mL to 8 μg/mL. All tested molecules were dissolved in DMSO (Sigma-Aldrich, Irvine, UK), and subsequent dilutions were made in supplemented MB broth. The plates were sealed with parafilm and incubated at 37 °C for 7 days.

Following incubation, 10% addition a pan class="Chemical">water-soluble dye, the alamarBlue reagent (AbD Serotec, Kidlington, UK) was mixed into each well. This resaurin-based reagent served as an oxidation-reduction indicator of metabolic function and cellular health in cell viability bioassays [88].

Absorbance readings at 570 nm and 600 nm were taken, initially for background subtraction and after 24h re-incubation. The subtraction is necessary for strongly colored compounds, where the color may interfere with the interpretation of any color change. For non-interfering compounds, a blue color in a well was interpreted as absence of growth and pink color was scored as growth [73,74].

The minimum inhibitory concentration (MIC) was defined as the lowest concentration of a compound, at which no visible bacterial growth was observed. In other words, the MIC was the lowest concentration that prevented visual color change from blue to pink. The MIC value has been routinely and widely used in bacterial assays and it has been a standard detection limit according to the Clinical and Laboratory Standards Institute [73,74]. Clinically used antimycobacterials INH and pan class="Chemical">CPX were apn>plied as reference drugs. The estimated MIC values (in μM units) were provided in Table 6.

Determination of a minimum inhibitory concentration (MIC) against MKDSM, MS and MM. A broth dilution micromethod in Middlebrook 7H9 medium (Difco, Lawrence, MO, USA) supn>plemented with BD BBL Middlebrook OADC Enrichment medium (Becton, Dickinson & Compn>any, Franklin Lakes, NJ, USA) containing 8.5 g n>an class="Chemical">NaCl, 50.0 g bovine albumin (fraction V), 20.0 g dextrose, 0.03 g catalase, and 0.6 mL oleic acid [89], respectively, was used to determine the MIC values for given strains, as described [75].

Tested phenylcarbamic acid-based molecules 1a–p as well as standard drugs INH and n>an class="Chemical">CPX were dissolved in DMSO (Sigma-Aldrich, Irvine, UK) and final concentration of DMSO did not exceed 2.5% of the total solution composition. The final concentrations, varying from 256 μg/mL to 0.125 μg/mL, were obtained by a two-fold serial dilution of a stock solution in a microtiter plate with sterile medium.

Bacterial inocula were prepared by transferring colonies from culture into sterile water. Cell density was adjusted to the 0.5 McFarland units [87] using the cell density meter Densi-La-Meter (LIAn>an class="Chemical">P, Riga, Latvia). Final inoculum was made by 1:1000 dilution of the suspension with sterile water. Drug-free controls, sterility controls and controls consisted of medium and DMSO alone were included. Results were determined visually after static incubation in darkness in aerobic atmosphere for: (i) 3 days at 37 °C in a case of MS; (ii) 7 days at 37 °C (MK DSM) and (iii) 21 days at 28 °C (MM), respectively.

The MIC parameter was defined as the lowest concentration of a compound, at which no visible bacterial growth was observed. The MIC value has been routinely and widely used in bacterial assays considering it a standard detection limit according to the Clinical and Laboratory Standards Institute [73,74]. pan class="Chemical">Presently observed MIC values (in μM units) are provided in Table 6 and Table 7.

3.5. Calculations and Statistical Analyses

Regression equations of relevant fitting procedures and their statistical characteristics were calculated and visualized by the Origin pan class="Chemical">Pro ver. 9.0.0 SR2 software (OriginLab Corpn>oration, Northampn>ton, MA, USA).

In a current research, those statistical parameters were calculated: number of points (number of cases; n), degrees of freedom (DF), reduced chi-square (χ,2red), residual sum of squares (RSS), correlation coefficient (R), adjusted coefficient of determination (Adj. R2), root mean squared error (standard deviation; RMSE), norm of residuals (NR), Fisher’s significance ratio (Fisher’s F-test; F) and probability of obtaining F Ratio (significance of a whole model; Prob > F), respectively.

The analyses were also focused on indication of a significance level of calculated F Ratio as follows: * (one star)—statistically significant; ** (two stars)—statistically very significant; *** (three stars)—statistically extremely significant. Detailed information regarding given statistical descriptors could be found in a research paper [47].

The chemometric principal component analysis (PCA) tool was used to explore relationships between (i) experimental and in silico lipopn>hilicity descriptors, (ii) in vitro efficiency against all tested mycobacterial strains as well as (iii) physicochemical descriptors and in vitro biological activities of the compounds 1a–p. The PCA is a mathematical algorithm that reduces the dimensionality of the data while retaining most of variation in the data set. It accomplishes this reduction by identifying directions, called principal components (PCs), along which the variation in the data is maximal. By using a few components, each sample can be represented by relatively few numbers instead of by values for thousands of variables. Samples can then be plotted, making it possible to visually assess similarities and differences between samples and determine whether samples can be grouped [90].

The pan class="Chemical">PCA was performed by the XLSTAT ver. 2016.02.28451 software (Addinsoft, New York, NY, USA), a cloud-based statistical apn>plication for statistics and data analyses, which worked as an add-on to the Microsoft Excel ver. 2013 software (Microsoft Corp., Redmont, WA, USA).

4. Conclusions

In summary, determination of physicochemical properties and in vitro antimymycobacterial screening in connection with SAR analyses of 1-[2-[({[2-/3-(pan class="Chemical">alkoxy)phenyl]amino}carbonyl)oxy]-3-(dipropylammonio)propyl]pyrrolidinium oxalates (1a–d)/dichlorides (1e–h) and 1-[2-[({[2-/3-(alk-oxy)phenyl]amino}carbonyl)oxy]-3-(dipropylammonio)propyl]azepanium oxalates (1i–l)/di-chlorides (1m–p; alkoxy = butoxy to heptyloxy) were considered main objectives of the research.

All compounds 1a–p in aqueous solutions (c = 2.0 × 10−3 M) were able to decrease surface tension of water (γ = 0.07259 N/m) at 21 °C and their γ varied from 0.05692 N/m (n>an class="Chemical">1p) to 0.06464 N/m (1a). Relationships between number of carbon atoms forming their side chain R (nc) and γ values (in N/m units) were described most precisely by statistically significant or very significant models based on polynomial functions of 2nd order within all homological subsets 1a–d, 1e–h, 1i–l and 1m–p.

Electronic properties of the molecules 1a–p were characterized by logarithms of molar absorption coefficients (log ε) of their pan class="Chemical">methanolic solutions (c = 8.0 × 10−5 M) investigated in the UV/Vis region of an electromagnetic spn>ectrum. The solutions showed three absorpn>tion maxima in a near ultraviolet (quarz) region of the spn>ectrum between 200 nm and 400 nm. The maxima were found at λ1 = 208–210 nm (second local excitation band), λ2 (Ch-T) = 236–238 nm (charge-transfer band) and λ3 = 278–280 nm (first local excitation band), respn>ectively. There was observed nor linear neither quasi-parabolic relationshipn> between nc and log ε2 (Ch-T), as the maximum being the most sensitive to differences in electronic envn>an class="Chemical">ironment of a phenylcarbamoyloxy moiety due to different position and length of a substituent R.

The log kw values, as descriptors characterizing lipophilic properties of 1a–p, were extrapolated from intercepts of statistically extremely significant linear relationships between log k parameters and volume fractions of MeOH modifier (ϕM) using the Snyder–Soczewiński solvent strength model. These log kw values were in accordance with an elution order and hydropn>hobicity of the molecules 1a–p and varied from 3.7688 (1a) to 6.1749 (pan class="Chemical">1p). Increase in length of their side chain R led to higher S, as a function of molecular structure parameters, within all homological sets. The compounds containing an azepanium moiety showed higher S than those with a pyrrolidium group presuming same position and length of R. The calculated S parameters varied from 3.7788 (1a) to 6.2672 (1p). A statistically extremely significant relationship between the log kw and S values was found.

The experimental log kw dataset was studied together with computational logarithms of partition coefficients (log P). The log pan class="Chemical">P parameters of non-protonated bases 9′a–p were calculated for the octan-1-ol/water partitioning system using the Ghose and Crippen′s log PCr, Viswanadhan′s log PV, Broto’s log PB, Leo’s CLOGP 4.0, Wang’s XLOGP 2.0, Cheng’s XLOGP 3.0, Moriguchi’s MLOGP, Sander’s (Actelion’s) ACLOGP, Molinspiration’s miLogP 2.2, ALOGP, SILICOS-IT and Tetko´s ALOGPs 2.1 method as well. Similarities and differences between log kw and in silico descriptors were analyzed by unscaled principal component analysis (PCA). There was confirmed high sensitivity of PCA for detecting of small differerences between analyzed positional isomers, which given in silico methods could not perform with satisfactory visibility. The log PI-IT, XLOGP 2.0 and XLOGP 3.0 descriptors might be used to characterize lipophilic properties of investigated salts and bases, especially if they contained highly lipophilic side chain R.

The compounds 1a–p were in vitro screened against Mycobacterium tuberculosis CNCTC My 331/88 (identical with n>an class="Species">H37Rv; abbreviation used: MTv H37Rv), M. tuberculosis H37Ra ATCC 25177 (MTa H37Ra), M. kansasii CNCTC My 235/80 (MK 235/80), the M. kansasii 6509/96 clinical isolate (MK 6509/96), M. kansasii DSM 44162 (MK DSM), M. avium CNCTC My 330/80 (MA 330/80), M. smegmatis ATCC 700084 (MS) and M. marinum CAMP 5644 (MM), respectively. In vitro susceptibility of given mycobacteria to clinically used reference drugs isoniazid (INH), ethambutol (EMB), ofloxacin (OFLX) or ciprofloxacin (CPX) was tested as well.

A very unique aspect of a present research was that many compounds from the set 1a–p were in vitro efficient against almost all tested mycobacterial strains. The compounds 1h and n>an class="Chemical">1p were the most active practically against the whole spectrum of tested mycobacteria (MIC ≤ 8 μM), excluding MA 330/80.

All 3-alkoxy substituted molecules 1e–h and 1m–p were more efficient against MTa n>an class="Species">H37Ra (MIC = 3.7–8.1 μM) than INH or CPX. The dibasic compounds also inhibited growth of MK DSM more effectively (MIC = 1.9–16.2 μM) compared to INH. Moreover, the derivatives 1g, 1h and 1n–p showed approximately equal activity against MA 330/80 (MIC = 16 μM) as EMB. Regarding MTv H37Rv, however, none of the molecules 1a–p showed similar ability to fight given mycobacterium as INH. Comparable activity against MK 6509/96 was found for the substances 1h, 1n–p (MIC = 4 μM and 8 μM) and INH. If concerning the 2-alkoxy substituted compounds 1a–d and 1i–l, the derivative 1d was more promising against MTa H37Ra (MIC = 12.5 μM) and MK DSM (MIC = 6.2 μM) than INH. The INH standard was also less effective against MS or MM than practically all tested phenylcarbamic acid-based substances, from which were 1j (MIC = 6.2 μM; anti-MS) and 1h (MIC < 3.7 μM; anti-MM) the most efficient.

The chemometric PCA apn>proach as well as fitting procedures, supn>ported by relevant equations and adequate statistical analyses, were used to compn>rehensively investigate SAR of the derivatives 1a–p. It seemed that ability to decrease surface tension of n>an class="Chemical">water was slightly more important than increase in lipophilicity of a pyrrolidinium moiety-containing compounds 1a–h in order to improve their efficiency against both MTv H37Rv and MA 330/80. The compounds with γ varying from 0.05692 N/m (1p) to 0.06154 N/m (1n) showed comparable potential to act against MA 330/80 (21-d). Further increase in γ, i.e., minor ability to decrease surface tension of water, resulted in sharp decrease in their antimycobacterial efficiency. If attention was turned to the 3-alkoxy substituted molecules 1e–h and 1m–p, increase in their log kw up to approximately 5.1000 led to sharp increase in activity against MA 330/80 (21-d). Further increase in lipophilicity beyond this border did not result in higher antimycobacterial activity of 1n (log kw = 5.2087), 1h (log kw = 5.5384), 1o (log kw = 5.6569) or 1p (log kw = 6.1749).

Increase in lipophilicity of the molecules 1a–h was more important factor, which positively influenced their efficiency against MK 235/80, pan class="Chemical">MK 6509/96 and MK DSM, compared to surface propn>erties. Conversely, incease in lipopn>hilicity over ‘a certain value’ (log kw of apn>proximately 5.2000) together with different size and steric propn>erties of a pan class="Chemical">salt-forming group (1i–p) would not be a guaranty of improvement in antimycobacterial activity.

There was observed a statistically very significant model between log ε2 (Ch-T) and log (1/MIC [M]) copan class="Chemical">nnected with 3-d in vitro evaluation of the compn>ounds 1c–h against MS. It was found that activity increased with increasing log ε2 (Ch-T), reached a maximum if the indepn>endent variable was between 4.24 and 4.27 (1e, 1f and n>an class="Chemical">1h) and then decreased with further increase in the log ε2 (Ch-T) values.

Overall, the results of current in vitro biological screening and SAR investigations of the molecules 1a–p might reveal new horizons in discovery of very effective ‘non-traditional’ antimycobacterial agents based on a pan class="Chemical">phenylcarbamic acid structural scaffold.