Exploring Resistance of Fluoroquinolones in Uropathogenic E. Coli and Optimized Lead Prediction Through In-Silico

Introduction

Urinary tract infections (UTIs) are a significant public health issue, marked by the presence of microbial pathogens in the urinary system, including the kidneys, ureters, bladder, and urethra [1]. UTIs are among the most prevalent infectious diseases globally, with an estimated annual incidence of at least 250 million cases [2]. Escherichia coli (E. coli), the primary pathogen responsible for UTIs, accounts for approximately 80-90% of infections in children and is similarly prevalent among adult women [3, 4]. In Nepal, UTI prevalence is particularly concerning, with 20% of women experiencing at least one UTI in their lifetime and 3% experiencing recurrent infections annually [5]. Hospital-acquired UTIs contribute significantly to the burden, accounting for 23%-37% of infections, with a higher incidence rate of 37.84% during pregnancy [6, 7].

The impact of UTIs extends beyond infection, contributing to substantial morbidity and mortality, particularly among hospitalized patients, where they represent 25-50% of all infections [8]. This high incidence affects patient health and imposes a considerable economic burden on healthcare systems, leading to increased healthcare expenditures and lost productivity due to missed workdays [8].

A wide range of pathogens can cause UTIs, including Gram-negative and Gram-positive bacteria, and even fungi. However, uropathogenic E. coli (UPEC) is the most common cause of both uncomplicated and complicated UTIs, responsible for approximately 90% of community-acquired infections and a significant portion of hospital-acquired cases [9]. The virulence of UPEC is primarily due to specific adhesin-encoding operons that facilitate adherence to uroepithelial cells, enabling colonization and subsequent infection [10].

Fluoroquinolones have historically been a cornerstone in the empirical treatment of UTIs due to their broad-spectrum activity and favorable pharmacokinetic properties. However, rising resistance rates among uropathogens, particularly E. coli, have raised significant concerns regarding their continued efficacy. The emergence of fluoroquinolone resistance is attributed to factors such as over-prescription, inadequate treatment regimens, and the inherent genetic adaptability of bacteria.

Globally, fluoroquinolone resistance among uropathogenic E. coli is increasing, with rates as high as 76.2% in Iraq and concerning levels in the United States [11]. This issue is not confined to specific regions but is a global problem, particularly in developing countries where resistance rates often exceed 40% [12]. Numerous studies have documented the high prevalence of fluoroquinolone resistance among E. coli isolates from UTIs [13]. For example, a study in Spain reported that 22.9% to 30.8% of E. coli isolates were resistant to fluoroquinolones, consistent with other findings indicating a growing concern in healthcare settings [14].

The primary mechanism of action of fluoroquinolones involves the inhibition of two essential bacterial enzymes, DNA gyrase, and topoisomerase IV, which are responsible for the supercoiling and relaxation of DNA during replication and transcription [9]. The binding affinities of these drugs to their targets can significantly influence their efficacy and the development of resistance in bacterial populations [15]. Resistance to fluoroquinolones can develop through several mechanisms, including mutations in the genes encoding these target enzymes, alterations in membrane permeability, and the acquisition of efflux pumps that actively transport the drug out of the bacterial cell [15].

In response to the growing threat of antibiotic resistance, there is an urgent need for innovative approaches to predict and mitigate resistance mechanisms. In silico methods, leveraging machine learning and genomic data, have emerged as promising tools for the identification of resistance genes and the prediction of antimicrobial susceptibility profiles [16, 17]. These computational approaches can facilitate the rapid assessment of resistance determinants, enabling clinicians to make informed decisions regarding empirical therapy and potentially guiding the development of novel therapeutic agents [18].

Molecular docking studies can provide insights into the interactions between fluoroquinolones and their targets, aiding in the identification of structural features that enhance or diminish drug efficacy. By utilizing computational methods, researchers can predict how specific mutations in target enzymes affect drug binding and resistance development [19]. These studies can also facilitate the design of novel fluoroquinolone derivatives with improved binding affinities and reduced resistance potential. Tools like AutoDock 4.2 are instrumental in examining drug-receptor interactions and optimizing lead compounds for better therapeutic outcomes.

Given the high resistance rates and the need for more effective treatment strategies, the empirical use of fluoroquinolones in UTI treatment may no longer be justified, especially in light of the high resistance rates reported in various studies [20]. A study conducted in Nepal highlights that irrational use of antimicrobials significantly contributes to drug therapy-related problems [21]. This issue is particularly concerning due to the high rates of antibiotic resistance observed in uropathogenic E. coli, which is the primary causative agent of urinary tract infections (UTIs) [22, 23]. The increasing prevalence of extended-spectrum beta-lactamase (ESBL) producing strains has raised concerns about treatment efficacy, as common antibiotics like ampicillin and ciprofloxacin show reduced effectiveness against these pathogens [22]. This resistance pattern presents a significant challenge for healthcare providers in Nepal, underscoring the need for advanced computational methods to predict and develop optimized treatment strategies against these resistant strains.

This study hypothesizes that newly synthesized fluoroquinolone derivatives can offer improved inhibition of DNA gyrase enzymes, providing a more effective treatment for resistant UTIs. By exploring the resistance of fluoroquinolones in uropathogenic E. coli at a Nepali tertiary care hospital and employing in silico approaches to predict optimized lead compounds, this research aims to pave the way for next-generation antimicrobials that are both effective against resistant strains and optimized for safety and efficacy.

Materials and Methods

Ethics approval

The ethical approval for the study protocol was obtained from the Institutional Review Committee of Manmohan Memorial Institute of Health Sciences College (MMIHS-IRC Ref.no.: NEHCO/IRC/080/24). Approval for the data collection from the Manmohan Memorial Teaching Hospital (MMTH), Kathmandu, was obtained from the Hospital board (Ref. No. 241).

Procedure of fluoroquinolones resistance prevalence determination

A retrospective observational study was conducted at Manmohan Memorial Teaching Hospital (MMTH) in Kathmandu. The study involved retrospective collecting data over six months from both inpatient (IPD) and outpatient (OPD) departments, which was obtained from the medical record department of the Hospital. We focused on patients who had undergone urine culture and sensitivity testing in association with urinary tract infections (UTIs). Patients whose cultures did not result in bacterial isolation were excluded from the study. The collected data were then systematically entered into Excel 2019. Descriptive statistics were employed to analyze socio-demographic variables, and the percentage of fluoroquinolone resistance among isolated microorganisms was calculated.

The procedure of lead optimization through in-silico methods

Molecular docking

AutoDock 4.2 [24] was employed to predict how small molecules, including fluoroquinolone derivatives, bind to the DNA gyrase of E. coli (PDB ID: 4KFG), a known target for urinary tract infections (UTIs) [25]. Potential leads from novel fluoroquinolone derivatives were identified through in-silico methods.

Preparation of ligand

Ligands were prepared using various computational tools. PubChem [26] was utilized to obtain SDF files of existing fluoroquinolones, while Marvin Sketch [27] was used to construct novel fluoroquinolone derivatives in SDF format. These files were then converted into PDB format using Discovery Studio [28] and subsequently into PDBQT format via Autodock software [29]. The ligands were composed of a fused ring system (benzene and pyridine) with various substituents such as ethyl, cyclopropane, and 2,4-difluorobenzene rings at the N-1 position, along with functional groups like carboxylic acid at C-3, a keto group at C-4, an amino group at C-5, a fluoro atom at C-6, and various scaffolds [R] at C-7 as shown in Figure 1.

Preparation of protein molecule

The protein structure of DNA gyrase from E. coli was retrieved from the Protein Data Bank (PDB) ((https://www.rcsb.org/) using the ID 4KFG [30]. The structure, matching required parameters such as x-ray diffraction resolution and no mutations and Ramachandran plot, was optimized by cleaning and removing irrelevant residues, fixing structural errors, and adding polar hydrogen bonds. The final structure was saved in PDB format for subsequent docking studies.

Identification of binding pocket

The active binding site of the protein was identified using the Cast-P server ((http://cast.engr.uic.edu), utilizing the amino acid sequence as the binding pocket for blind docking [31]. The amino acid residues involved in the active site were thoroughly analyzed and listed for further study.

Pharmacokinetic and toxicity prediction

The pharmacokinetic properties, including GI absorption, distribution, metabolism, and excretion (ADME), of all ligands were predicted using the SWISS ADME lab web server (http://www.swissadme.ch/) [32]. Additionally, the toxicity of these compounds was predicted using the ProTox-II web server (https://toxnew.charite.de/) [33], ensuring the identification of safe and effective drug candidates.

Biological activity prediction

To validate the docking results, the PASS web server (https://www.way2drug.com/passonline/) [34] was used to predict the biological activity of the bioactive compounds, focusing on antibacterial activity. The results indicated that the probability of these compounds being active (Pa) was greater than the probability of being inactive (Pi), suggesting potential antibacterial properties.

Docking procedure

The docking studies were carried out using Autodock Vina. The binding energy between ligand-receptor interactions was calculated using the formula:

Here, ΔG(Gauss) represents the dispersion of two Gaussian functions, ΔG(Repulsion) accounts for repulsion beyond a threshold distance, ΔG(H−Bond) models hydrogen bond interactions, ΔG(Hydrophobic) is a ramp function for hydrophobic interactions, and ΔG(Tors) is proportional to the number of rotatable bonds.

The protein structure was loaded into Autodock 4.2, and the format was converted to PDBQT. The ligands were uploaded, their geometrical energies minimized to global maxima and converted into PDBQT format. The docking grid parameters were set with the following values: x: 16.332, y: 18.653, z: -11.116, with dimensions of x: 69.99, y: 47.320, z: 69.00. The obtained conformations were further analyzed using Discovery Studio 2023.

Results and Discussion

Out of 2306 patients diagnosed with UTI, 167 samples showed significant growth of microorganisms, and all the samples consisted of a single microorganism. The gender distribution indicated a higher prevalence among females (69.5%) compared to males (30.5%). Age-wise, the majority of participants were between 31-60 years (43.11%), followed by those aged 1-30 years (37.12%), and a smaller portion of the sample aged 61-90 years (19.76%). Regarding bacterial classification, gram-negative bacteria were predominant, accounting for 88.6% of cases, whereas gram-positive bacteria were found in 11.4% of cases.

Twelve Pathogenic bacteria were isolated from 167 patients' urine samples, of which E. coli (108, 64.7%) was the most common isolate.

Out of 108 E-coli isolates tested for Ciprofloxacin, 68 were sensitive, while 33 were resistant, yielding a resistance rate of 32.67%. Levofloxacin, however, exhibited a much higher resistance rate. Among the 6 E. coli isolates tested, only 1 was sensitive, and 5 were resistant, leading to an 83.33% resistance rate, the highest among the three antibiotics. Ofloxacin showed intermediate results, with 3 of the 7 isolates being sensitive and 4 being resistant, resulting in a resistance rate of 57.14%. These findings indicate a significant level of resistance of E. coli to fluoroquinolones, particularly to Levofloxacin, suggesting that the effectiveness of these antibiotics may be limited in treating E. coli-related infections.

Molecular docking result

In our study, 42 novel fluoroquinolone ligands were docked with 4KFG (DNA gyrase of E. coli protein to predict the potential anti-bacterial activity. The binding energy, number of hydrogen bonds, bond distance, and amino acid responsible for interaction were presented in Table 1. The ligand showing the lowest binding affinity, a greater number of hydrogen bonds, a shorter bond distance, and a greater number of amino acid interactions was taken as the best ligand for further investigation.

The standard ligands Ciprofloxacin and Ofloxacin showed the minimum binding energy with the DNA-gyrase of E-coli protein 4KFG. The minimum binding energy ΔG = -7.5 kcal/mol was exhibited by ciprofloxacin with two H-bonds LYS-189 and ILE-186 with a bond distance of 2.19 and 2.75 angstroms, whereas minimum binding energy ΔG = -7.7 kcal/mol was exhibited by ofloxacin with two H-bonds LYS-189 and ILE-186 with a bond distance of 2.53 and 2.25 angstroms respectively.

The ligand having Fluro-Indole [SP12], Fluoro-Benzimidazole [SP9], Fluoro-Benzothiazole [SP13, SP15], Morpholino [SP42], Amino-Azetidine [SP23, SP24] and Diamino Pyrrolidine [SP25, SP26, SP27] substituted derivatives showed the maximum binding energy and highest number of hydrogen bond with the protein 4KFG. The maximum binding energy was exhibited by fluoro-benzimidazole, fluoro-indole, and morpholino-substituted derivatives, and the maximum hydrogen bond was exhibited by diamino-pyrrolidine substituted derivatives.

Fluoro-benzimidazole [SP9] exhibited maximum binding affinity ΔG = -11.9 Kcal/mole having three hydrogen bonds ASP-49, ARG-76, GLY-77 with bond length 2.3, 2.1, 2.1 Angstroms. Fluoro-indole [SP12] exhibited maximum binding affinity ΔG = -11.7 Kcal/mole having four hydrogen bonds ASP-49, GLU-50, ARG-76, GLY-77 with bond length 2.1, 2.3, 2.1, 2.9 Angstroms and morpholino [SP42] exhibited maximum binding affinity ΔG = -11.6 Kcal/mole having two hydrogen bonds GLY-77, ARG-76 with bond length 2.7, 2.9 Angstroms. Diamino pyrrolidine [SP25] exhibited a maximum number of hydrogen bonds [five hydrogen bonds] ARG-76, GLY-77, ILE-90, VAL-93, LEU-94 with bond distance 2.8, 2.2, 2.2, 2.2, 2.3 Angstroms. The 2D and 3D interaction of CIP, OFL, SP-9, SP-12, SP-25, SP-42 was represented in Figure 2.

Biological activity prediction

Docking parameters are further supported by the biological activity prediction tools, which predict the anti-bacterial and DNA-gyrase inhibitors, such as the activity of forty-two novel fluoroquinolones derivatives. The probability to be active (Pa) value of all the bioactive compounds was greater than their probability to be inactive (Pi) values. This finding indicated that all these compounds, including potential molecules SP9, SP12, SP25, and SP42, showed good biological activity against UPEC bacteria, as shown in Table 2.

Lipinski’s rule of five and ADME properties analyses

We analyzed all the derivatives and found that none of them violated Lipinski's rule of five, as shown in Table 3. This rule helps filter candidates assess their drug-likeness and durability. Absorption, distribution, metabolism, and excretion (ADME) properties of all derivatives were determined by the swissADME web server, and the result is shown in Table 4. Positive findings from the examination of the prospective compounds SP9, SP12, SP25, and SP42's ADME characteristics and Lipinski's rule of five supported their further research and development.

Toxicity prediction

The toxicity of all 42-novel fluoroquinolone derivatives was predicted, and the result of selected molecules is shown in Table 5. Most of the derivatives exhibit hepatotoxicity and mutagenicity, with fewer cases showing cytotoxicity, carcinogenicity, and immunotoxicity. Compounds SP-9 and SP-12 demonstrate hepatotoxicity, mutagenicity, immunotoxicity, and carcinogenicity. SP-25 shows mutagenicity, while SP-42 is free from all of these toxicities. Different classes of toxicity are indicated as follows: Class I: fatal if swallowed, Class II: fatal if swallowed, Class III: toxic if swallowed, Class IV: harmful if swallowed, Class V: may be harmful if swallowed, and Class VI: non-toxic.