QSAR study of β‐lactam antibiotic efflux by the bacterial multidrug resistance pump AcrB

Abstract

AcrAB-TolC is the most important efflux pump system of Gram-negative bacteria, responsible for their resistance to lipophilic and amphilic drugs. In this work, HCA-PCA studies were performed to investigate the relationship between efflux activities (negative logarithm of minial inhibitor concentration, pMIC) of three strains of S. thypimurium with respect to b-lactams, and to analyze the relationship between lipophilicity parameters calculated by different methods. The analyses demonstrate that pMICs strongly depend on properties of both bacterial strains and drug molecules, and that lipophilicity parameters do not necessarily contain the same information about the drugs. QSAR studies have shown that the calculated lipophilicities, in some cases, are non linearly related to the pMICs originated by active AcrAB-TolC bacterial pumps, due to the existence of b-lactams with nitrogen-and sulfur-rich substituents. Among the most important b-lactam molecular properties quantitatively related to pMICs are lipophilicity and electronic and hydrogen-bonding properties. Parameters describing these properties were included in the QSAR study to obtain parsimonius regression models for MICs. b-Lactams were classified as good, moderately good and poor AcrABTolC substrates. Their stereoelectronic molecular properties, especially the Y-component of the molecular dipole moment and hydrogen binding properties, reflected this classification.

Nonequivalent Effects of Diverse LogP Algorithms in Three QSAR Studies

“…In a previous work from our group [25], the problem of the most relevant lipophilicity descriptor(s) in 3 regression models for b-lactam inhibitors of 3 strains of Salmonella thypimurium was posed, and solved by exploratory and PLS analyses. b-Lactam antibiotics belong to a specific class of organic compounds for which lipophilicity is an essential determinant of variations in antibacterial activity.…”

“…Is it not possible that better results could be obtained if the algorithm B is used instead of A, leading to a more robust model? In a previous study by Ferreira and Kiralj [25] it has been shown that various algorithms for LogP calculation result in values encoding different structural information and this, consequently, lead to different QSAR models. In this work, the relevance of the algorithm selection for performing LogP calculation in QSA(P)R studies is revised, extended and better explored.…”

Extensive Chemometric Investigations of the Multidrug Resistance in Strains of the Phytopathogenic Fungus Penicillium Digitatum

“…I -IV are DMIs, used as fungicides against the cytochrome enzyme P450 sterol 14a-demethylase (P450 14DM ) [12,13]. V is an antibiotic (Atb) that interferes in the microbial protein synthesis [43], and VI and VII are mutagens (Mut) which act as effective nucleic acid intercalators and modifiers [44 -46] It has been shown in a previous work [47] that bacterial activities, when treated by PCA and HCA, could distinguish strains in accordance with their baseline resistance. Similarly, cluster analysis methods and PCA are used today extensively in taxonomy of living beings, taking into account functional, morphological, structural, phylogenetic, physiological, biochemical, and other characterizations of species and strains [48 -53].…”

Chemometric modeling of core-electron binding energies

“…Calculation methods employed for generation of descriptors are (a) previously developed a priori approach based on known atomic properties [34][35][36][37]; (b) some original or modified literature data for atomic or molecular properties; (c) modeling and HF 6-31G * optimization of molecular geometries, and calculation of properties by using the Titan program [38]; (d) calculation of molecular polarizabilities using these optimized geometries in PM3 single point calculation (to be consistent with HF calculations) by the MOPAC 6.0 program [39,40]; (e) combination of two or more of these approaches. The descriptors were generated for the whole data set (59 atom types) and for the four particular data sets (27 C, 9 N, 14 O1, and 8 F atom types).…”