Introduction

Isoxazolidines are heterocycles of great importance for organic chemists due to their utility in organic synthesis [1,2,3,4,5]. Many isoxazolidine derivatives are biologically active [6,7,8,9,10,11] and can lead, by simple and effective chemical transformations, to highly functionalized and important compounds such as aminoalcohols [12], alkaloids [13], aminoacids [14], and aziridines [15]. Since recent years, our research group has been interested in the design and the synthesis of enantiopure isoxazolidine derivatives [16,17,18,19,20,21] and biologically active natural and non-natural amino acids [22,23,24,25,26,27,28,29] via 1,3-dipolar cycloaddition between a nitrone derived from (‒)-menthone and various alkenes. The introduction of a fluorine atom or a fluorine group into organic molecules generally gives exceptional biological and physical properties, thereby facilitating lipid solubility, metabolic stability, and binding properties to biological targets [30]. In particular, various trifluoromethyl compounds have known to have very interesting biological activities [31]. Consequently, significant efforts have been dedicated to expeditious synthesis of fluorinated structures.

Nowadays, the discovery of new drugs relies specially on the interaction of an organic molecule with a pharmacological target. This target is often a protein, sugar, or nucleic acid that is involved in triggering the disease state. In addition, the spread of multidrug- as well as pandrug-resistant microorganisms became a very serious and concern problem in health. Based on the chemical structure, the desired compounds were characterized for the prediction of their pharmacokinetics properties such as absorption, distribution, metabolism, and elimination (ADME) as well as their pharmacodynamics potential and toxicity behavior, to avoid potential interactions of drugs with antitargets causing many side effects [32]. For this, successful drugs development requires screening for their bioavailability (efficacy index), absorption (oral), and permeability ability. Thereby, many computational methods have been developed and proposed by the QSAR (quantitative structure activity relationships)/QSPR (quantitative structure–property relationships) community to evaluate and find a promising drug that shows good potency and selectivity against selected targets [33].

In this study, we synthesized new 3,4,5-trisubstituted isoxazolidines molecules containing CF3 group. The obtained title compounds were examined for their in vitro antimicrobial activities, ADME, and GUSAR (General Unrestricted Structure–Activity Relationships) predictions using various cheminformatic studies. Great attention was offered to model validation, applicability domain assessment, interpretation of the selected molecular descriptors, and various types of toxicity profiles.

Results and discussion

The new chiral trifluoromethylated nitrone was synthesized, with a good yield, from the trifluoromethylated aldehyde as illustrated in Scheme 1. The latter was obtained using the diacetone glucose as starting material using the same procedure described by Ghannay et al. [31].

scheme 1

To evaluate the reactivity of trifluoromethylated nitrone 1 with monosubstituted alkenes, we have chosen to engage the nitrone 1 in a cycloaddition reaction with allyl bromide. After 12 h of reflux in toluene, TLC shows the complete disappearance of the trifluoromethylated nitrone 1 and the formation of two regioisomers, each one was in the form of two diastereoisomers (Scheme 2). Only isoxazolidine 2 could be isolated by chromatography on silica gel with 42% yield. The cycloaddition reaction between the trifluoromethylated nitrone 1 and the maleimide in refluxing toluene, for 12 h, leads to the formation of two cycloadducts 3 (37%) and 4 (60%) separable by chromatography on silica gel (Scheme 2). The trifluoromethylated nitrone derived from d-glucose reacts with propargyl acetate in refluxing toluene to afford the new aziridine 6. This latter was obtained following an intramolecular rearrangement of the isoxazolinic compound 5 in 55% yield over two steps (Scheme 2).

scheme 2

The stereochemistry of isoxazolidine 2 was determined from the NOE effects observed on the NOESY spectrum (Fig. 1). The analysis of this spectrum showed that both H-4 and H-8 presented an NOE with H-9′. This means that the protons H-4, H-8, and H-9′ pointed in same side of the molecule. We have also observed an NOE effect between H-9 and H-5. However, no correlations have been observed between H-5, H-9′, and H-4. This shows that H-5 pointed in the same direction of H-9. All these observations further corroborate the proposed stereochemistry for isoxazolidine 2 (Fig. 1).

Fig. 1
figure 1

Characteristic NOESY correlations of compound 2

Full size image

In addition, important information taken from the NOESY spectra of cycloadducts 3 and 4 is as follows:

For cycloadduct 3: (a) a NOE effect was observed between H-8 and H-9, this means that H-8 and H-9 were pointed in the same direction; (b) no NOE effect was observed between the proton H-4 and H-8 and H-9. This shows that the two protons H-8 and H-9 were pointed in the opposite direction of H-4; (c) a NOE effect was observed between H-5 and H-4. These observations confirm the proposed stereochemistry for the cycloadduct 3 (Fig. 2).

Fig. 2
figure 2

Structures of compounds 3 and 4

Full size image

For the cycloadduct 4: (a) an NOE effect was observed between H-8 and H-9; (b) an NOE effect was observed between the proton H-4 and H-8 and H-9, this means that H-4, H-8, and H-9 are on the same side of the molecule; (c) no NOE effect was observed between H-5 and H-8. These information corroborate the proposed stereochemistry for cycloadduct 4 (Fig. 2).

The assignment of 1H NMR as well as 13C NMR signals of compound 6 was confirmed using COSY, HSQC, and DEPT experiments. The chemical shift values of aziridine 6 are in good agreement with the suggested structure. Indeed, the analysis of the proton NMR spectrum of aziridine 6 shows a triplet and a doublet with strong fields towards 2.42 ppm and 2.48 ppm, corresponding to protons H-5 (J5,6 = 6.8 Hz) and H-6 (J6,5 = 6.8 Hz), respectively. The analysis of the 13C NMR spectrum of aziridine 6 shows the existence of a weak magnetic field signal at 198.8 ppm, corresponding to the carbon of the ketone function. The carbon of the ester function resonates at 169.9 ppm, whereas the carbons C-5 and C-6 of the aziridine ring appear at 44.9 and 44.8 ppm, respectively. To further confirm the regioselectivity of the cycloaddition reaction, we carried out a spectral study in 2D NMR with the HMBC experiment. Indeed, the HMBC spectrum of aziridine 6 shows a correlation between the proton H-6 and the carbons C=O (of the ketone function) and C-5. Similarly, the proton H-4 correlates with the C-3 and C-5 carbons. Also, the H-8 proton correlates with the carbonyl of the ketone function C-8 and the C-9 ester. The methyl group is attached to a carbonyl correlating with C-9. These observations confirm that the oxygen of the trifluoromethylated nitrone 1 reacted on the most substituted carbon of the alkyne (Fig. 3).

Fig. 3
figure 3

Structure and HMBC spectrum of aziridine 6

Full size image

In silico evaluation of molecular properties

To better develop highly effective therapeutic agents and optimize the design and the biological applications of the target compounds, we are interested in evaluating their ADME and drug likeness properties and to predict their bioactivity score using molinspiration cheminformatics software [19, 20].

Pharmacokinetic parameters were estimated on the bases of Lipinski’s rule “rule of five”, in which target molecules meet the criteria of drug likeness if: (1) the molecular weight is under 500, (2) the calculated octanol/water partition coefficient (LogP) < 5, (3) there is fewer than five hydrogen-bond donors (NH and OH groups), and, (4) there are less than ten hydrogen-bond acceptors (notably N and O atoms). The computed molecular properties of the target compounds are depicted in Table 1. As it can be seen, compound 1 did not violate any of the Lipinski’s rule of five; however, one violation was observed for trifluoromethylated compounds 3, 4, and 6, and, therefore, may be a good candidate for good availability and, consequently, corroborating their oral drug likeness properties. Isoxazolidine 2 with two violations was expected to be orally inactive. TPSA parameter which is the sum of van der Waals surface areas of electronegative atoms (oxygen and nitrogen with their attached hydrogen) was used as a good descriptor to elucidate the absorption and the passive transportation properties through biological membranes, for the optimization of a drug’s ability to permeate cells for a better prediction of their transport in the intestines and permeability through Caco-2 cells and blood–brain barrier (BBB). Compounds 3 and 4 with TPSA values equal to in 95.58 Å2 displayed a good intestinal absorption; however, compounds 1, 2, and 6 with TPSA values ≤ 90 Å2 have BBB penetration and thus act on receptors in the central nervous system (CNS). Also, only compound 2 (TPSA < 60 Å2) will be completely absorbed, while the others (TPSA > 60 Å2) will be absorbed to a less degree. The percentage of absorption (%ABS) which is deduced from the TPSA values according to the following equation: %ABS = 109 − 0.345 × TPSA was also determined. All candidate compounds exhibited a great %ABS ranging from 76.02 to 91.95 suggesting their efficient oral absorption. The hydrophobicity of the desired compound which was described by milogP (octanol/water partition coefficient) is an excellent parameter that measures the drug potency and plays a crucial role in the distribution of the drug in the body after absorption. The milogP of all trifluoromethylated compounds were determined and found to be within acceptable range according to Lipinski’s rule except isoxazolidine 2 with milogP value of 5.53 (very lipophilic) which is in clear violation of the Lipinski’s rule of five, suggesting that its solvent solubility is higher than its membrane solubility and, consequently, showed poor permeability across the cell membrane. Compounds 1, 3, 4, and 6 were found to be in compliance with Lipinski’s rule displaying potent ADMET (absorption, distribution, metabolism, elimination, and toxicology) properties and, thus, were designed to be useful for good availability and particularly the BBB. Only compound 1 has molecular weight in the acceptable range (MW ≤ 500) suggesting its absorption, transportation, and diffusion through membrane, unlike those of 2, 3, 4, and 6. An increase in the MW (except certain limit) enhanced the bulkiness of the molecules [34]. The present tested trifluoromethylated molecules have a number of rotatable bonds equal to 7 (1, 3, and 4) and 8 (2) showing large conformational flexibility, and thereby might be administrated through oral route, while aziridine 6 with a number of rotatable bonds more than 10 (Nrotb. = 11) needed some modification in dosage form to be available orally. It is well known that flexibility increases with increasing the number of rotatable bonds leading to good binding affinity with binding pocket for the compound [35]. All compounds exhibited NON (H-bond acceptors) and NOHNH (H-bond donors) values in the acceptable range conferring them a remarkable intestinal absorption and oral bioavailability in rat. It can be concluded that among the tested derivatives, only trifluoromethylated compound 2 is likely to be orally inactive, as they did not obey Lipinski’s rule of five. The obtained results were correlated with standards.

Table 1 In silico physicochemical properties of the synthesized compounds based on Lipinski’s rule
Full size table

In silico evaluation of bioactivity score

All our synthetic compounds were subjected for the calculation of their bioactivity score based on Molinspiration software against six different protein structures such as binding to G protein-coupled receptor (GPCR) and nuclear receptor ligand (NRL), ion channel modulation (ICM), kinase inhibition (KI), protease inhibition (PI), and enzyme activity inhibition (EI), and, therefore, compared to the standards (Table 2). The bioactivity score was estimated as follows: considerable, moderately biological activity, or inactive when bioactivity score is more than 0.00, between − 0.50 and 0.00, or less than − 0.50, respectively. The present results clearly revealed that all tested trifluoromethylated compounds exhibited moderately-to-potent bioactivity score but higher than aspirin for all drug targets. Trifluoromethylated compounds 2 and 6 having good bioactivity score against nuclear receptor ligand, protease inhibitor, and enzyme inhibitor, while trifluoromethylated isoxazolidines 3 and 4 are strongly active only against protease inhibitor and enzyme inhibitor. Their physiological actions might involve interaction with all the drug targets. Our results are in the standard’s level which encourages their antimicrobial behavior.

Table 2 Bioactivity score of the synthesized compounds according to molinspiration cheminformatics software
Full size table

Quantitative prediction of antitarget interaction profiles for synthesized compounds

To estimate the interaction of drug-like compounds with 32 antitarget end-points, we assessed their possible toxic effects during drug development using GUSAR (General Unrestricted Structure‒Activity Relationships) program [33] including three types of proteins: receptors, enzymes, and transporters. Data on the quantitative end-point values for the tested compounds (Table 3) show that they must be interact with the present antitarget proteins in different manner. The applicability domain (AD) evaluation is a guarantee for QSAR/QSPR models in predicting the oral availability of candidate drugs. Trifluoromethylated compounds 3 and 4 fall in the AD in order of 94%, followed by those of 2, 6, and 1 at 91%, 84%, and 81%, respectively. Consequently, the GUSAR predictions can help us to better know about the possible side effects, contraindications, and precautions.

Table 3 Quantitative prediction of antitarget interaction profiles for synthesized compounds
Full size table

In silico ADME properties of synthesized compounds

In this study, the synthesized compounds were screened using in silico Pre-ADMET software [32] to predict their overall ADME properties and toxicity hazards (Table 4), since they play a vital role in drug discovery and environmental riskiness. Different parameters have been screened such as:

Table 4 ADME properties of compounds according to pre-ADMET software
Full size table
  1. (a)

    Blood–brain barrier (BBB) penetration which served in reducing the side effects and toxicity or improving the efficacy of drugs whose pharmacological activity of the brain. Predicting BBB penetration means predicting whether the target pass across the Blood Brain Barrier. All tested targets showed positive values, indicating that they can easily cross the BBB, but their values are less than 1 (CBrain/Cblood <1), suggesting that they are inactive in the CNS (Central Nervous System), with the highest value was allowed to compound 1 (0.903).

  2. (b)

    Caco2 cell permeability (CCP) as a human colon epithelial cancer cell line generally used to estimate the in vitro human intestinal permeability of the drug in comparison to human enterocytes, and express the transporter and the efflux of proteins. All tested compounds displayed positive CCP justifying their middle permeability.

  3. (c)

    Cytochrome P450 isoforms (CYP_2C19, 2C9, 2D6, and 3A4) play a vital role in drug safety persistence and bioactivation. Among them, CYP3A4 which interact with more than a half of all clinically used drugs, and CYP2C9 which is involved in metabolizing NSAIDs (Non-Steroidal Anti-Inflammatory Drugs) and weakly acidic molecules with a hydrogen-bond acceptor drugs [36]. All compounds are an inhibitor of CYP3A4 and substrate of CYP3A4. For CYP2D6 (inhibition and substrate), a negative effect was observed with compounds 1, 2, 3, and 4; however, compound 6 has low probability to be a substrate with an ability to be an inhibitor. The behavior of both CYP2C19 and CYP2C9 inhibition is inversely proportional to that of CYP2D6.

  4. (d)

    Human Intestinal Absorption (HIA) is the process through which the drug was administrated orally from the intestine. All compounds exhibited higher values in the range 70–100% belonging to the well-absorbed compounds and, therefore, may be assimilated through human intestine.

  5. (e)

    Madin–Darby canine kidney (MDCK) was taken as a model to assess the human intestine barrier and the uptake efficiency of drugs. We noticed that MDCK values are in the range cells system that may be employed as high tool for rapid permeability screening because of its the shorter growth period as compared to Caco-2 cell. All compounds displayed weaker MDCK values ranged from 0.019 to 0.124 nm/s, with the highest permeability was ascribed to compound 1.

  6. (f)

    P-glycoprotein (P-gp) is one of the main barriers for delivering drugs properly and is responsible for extruding toxins and xenobiotics out of cells. The inhibition of efflux pump is mainly done to improve the delivery of therapeutic agents. In general, P-gp can be inhibited through three mechanisms: blocking drug-binding site either competitively, non-competitively, or allosterically; than interfering with ATP hydrolysis; and finally, altering integrity of cell membrane lipids [37,38,39,40]. The results below indicate that only compound 1 was found to be ineffective.

  7. (g)

    Plasma protein binding (PPB) affects the time that a drug stays in the body and can also have an effect upon the drug’s efficiency. The degree of binding to plasma proteins dramatically influences the pharmacodynamic and pharmacokinetic behavior of a drug. Values of  % bound < 90 were classified as low and ≥ 90 as high. As shown, compounds 1, 2, 3, and 4 showed affinity for a plasmatic protein with the potent value close to 95% binding observed for compound 1; contrariwise, compound 6 present a lower binding. Likewise, it should be noted that the drug distribution process was highly affected by its ability of binding to protein plasma.

  8. (h)

    The skin permeability (SP) rate is an essential parameter for the transdermal delivery of drugs. The drug must diffuse into the intercellular lipid matrix, which is recognized as the major determinant of drug absorption by the skin [41]. All tested compounds showed negative values of skin permeability, ranging from − 2.29 (compounds 3 and 4) to − 1.74 (compound 2), meaning that it is not important that the compounds be administered via transdermal routes.

Toxicity prediction

Predictive toxicity by pre-ADMET software

The pre-ADMET program was used also to predict the toxicity risk parameters of the different targets assessed by Ames test (mutagenicity), carcinogenicity mouse and rate, and hERG (human Ether-à–go–go-Related Gene) inhibition. As shown in Table 5, all compounds act as mutagen. Only compound 2 had positive carcinogenicity for mouse, and only compounds 3 and 4 showed negative carcinogenicity for rats. Tested for their hERG inhibition, all candidates exhibited low risk.

Table 5 Predictive toxicity by pre-ADMET software
Full size table

Predictive toxicity by Lazar software

Lazar (lazy structure–activity relationships) is another tool for predicting toxicological properties. The results summarized in Table 6 indicate that all examined molecules showed mutagenic (Salmonella typhimurium) effects; however, their carcinogenicity remains inactive against mouse and other rodents. No carcinogenicity rate was observed with compounds 1 and 2, unlike that of 3, 4, and 6 with this model.

Table 6 Parameters of lazar toxicity prediction for different compounds
Full size table

Predictive toxicity by GUSAR software

GUSAR (General Unrestricted Structure–Activity Relationships) software was used to predict also the environmental toxicity. This tool creates QSAR/QSPR models on the basis based on chemical similarities between compounds with known toxic effects and the presence of toxic fragments represented as SDfile containing data about chemical structures and end-point in quantitative terms. We compared our results with well-known methods implemented in the T.E.S.T. (Toxicity Estimation Software Tool) program version 4.1 provided by the EPA (Environmental Protection Agency) [42]. The results depicted in Table 7 revealed that all molecules showed applicability domain of models for environmental toxicity testing. The bioaccumulation factor or bioconcentration factor (BCF) is defined as the ratio of the chemical concentration in biota as a result of absorption via the respiratory surface to that in water at steady state. The BCF values of our candidates ranged from 0.819 to 1.318 which is in the acceptable EPA model (− 1.7 to 5.694). The Daphnia magna LC50 end-point represents the concentration in water which kills half of a population of Daphnia magna (a water flea) in 48 h. All compounds exhibited very close LC50 values in the EPA set (0.117–10.064). The fathead minnow LC50 end-point represents the concentration in water which kills half of a population of fathead minnows (Pimephales promelas) in 4 days (96 h). The fathead minnow values off all compounds are very close and. The presented fathead minnow results vary from − 4.320 to − 3.280 which are very far as compared to the EPA range (0.037–9.261). Tetrahymena pyriformis IGC50 end-point represents the 50% growth inhibitory concentration of the T. pyriformis organism (a protozoan ciliate) after 40 h. Our results indicate that IGC50 values ranged from 0.819 to 1.387 and are in given EPA set (0.334–6.36).

Table 7 Environmental toxicity predictions
Full size table

Predictive of rat acute toxicity by GUSAR software

GUSAR software was used for quantitative in silico toxicity prediction of LD50 values of the designed compounds for rats with four types of administration (oral, intravenous, intraperitoneal, and subcutaneous). As shown in Table 8, difference in LD50 values obtained for different routes indicates that availability of compound for metabolism by liver is a major factor in their toxicity. The LD50 values, for intraperitoneal, oral and subcutaneous routes of administration, vary in the following order: compounds 3 and 4 < compound 6 < compound 2 < compound 1; however, it is of intravenous follow the next order: compound 1 < compound 2 < compound 6 < compounds 3 and 4. All title compounds fall in applicability domain for intraperitoneal, oral, and intravenous administration routes, but the subcutaneous one falls out of the applicability domain.

Table 8 Rat acute toxicity predicted by GUSAR software
Full size table

Assessment of in vitro antimicrobial activities

The antimicrobial activities of the newly synthesized target were evaluated against a panel of pathogenic tested organisms by the agar-well diffusion method. The depicted results in Tables 9 and 10 were measured quantitatively in terms of IZD, MIC, MBC, and MFC values, and were compared to the standard drugs (chloramphenicol for bacteria and cyclodextrine for fungi). As shown, the experimental investigations revealed that all tested substances generated antimicrobial activities against all strains with different spectrum activity (except compound 2 against F. oxysporum). Nitrone 1 itself was found to be the most active in the series against all the tested microorganisms, especially against S. aureus with IZD = 19.00 ± 0.33 mm, MIC = 1.58 mg/cm3, and MBC = 3.17 mg/cm3 followed by compounds 3 and 4, in comparison to the standard, chloramphenicol. Compound 6 displayed moderately-to-higher inhibitory potential and that of 2 exhibited moderately-to-weakly sensitivity toward all the tested strains. All the tested compounds were also screened for their in vitro antifungal activity. The results revealed that F. oxysporum was the most sensitive to all compounds when compared to F. phyllophilum with the significant level of activity was observed for compound 1 (IZD = 18.00 ± 0.00 mm, MIC = 12.50 mg/cm3, MFC = 50.00 mg/cm3) followed, respectively, by those of 3 and 4 (IZD = 16.00 ± 0.88 mm, MIC = 6.25 mg/cm3, MFC = 25.00 mg/cm3), 6 (IZD = 13.00 ± 0.33 mm, MIC = 12.50 mg/cm3, MFC = 50.00 mg/cm3), and 2 (IZD = 8.00 ± 0.00 mm, MIC = 12.50 mg/cm3, MFC = 100.00 mg/cm3).

Table 9 In vitro antimicrobial activity of the synthesized compounds
Full size table
Table 10 Determination of MIC and MBC of synthesized compounds
Full size table

Interestingly, we attempted to determine the bactericidal/fungicidal or bacteriostatic/fungistatic behavior of our compounds. The results (Table 10) revealed that compound 1 exerts bactericidal and fungicidal effects higher than compounds 3 and 4. Compound 6 was found to be bactericidal against Gram-positive strains, bacteriostatic against Gram-negative strains, fungicidal against F. oxysporum, and fungistatic against F. phyllophilum; however, the analogue 2 was bacteriostatic (except for E. coli) and fungistatic against all strains.

As shown, Gram-positive bacteria were the most susceptible to this class of compounds. The higher resistance of Gram-negative bacteria as compared to Gram-positive bacteria is due to the differences in their cell wall structure. The presence and richness of thick cell wall with many layers of peptidoglycan and teichoic acids in Gram-positive bacteria facilitate the permeability behavior of tested compounds compared to Gram-negative bacteria possessing relatively thin cell wall with a few layers of peptidoglycan enclosed by a second lipid membrane rich in lipopolysaccharides and lipoproteins.

In addition, the antimicrobial activity was associated with the lipophilicity character of these compounds. Obviously, as shown above, the effectiveness antimicrobial activity towards compound 1 possesses the highest lipophilicity (milogP = 2.99) followed by compounds 3 and 4 (milogP = 3.18), 6 (milogP = 2.99) and 2 (milogP = 2.99), respectively. Likewise, these molecules contain sugar fragment that enhance and participate to their biological activities and, therefore, are well expected to be a good candidate of oral drugs.

Conclusion

The reaction of the trifluoromethyl nitrone derived from d-glucose with the propargyl acetate gave access, after rearrangement of the 4-isoxazoline, to aziridine in good yield. The 1,3-dipolar cycloaddition reaction of the same nitrone with allyl bromide leads to the formation of two regioisomers, each one in the form of two diastereoisomers. With the maleimide, the cycloaddition reaction leads to two diastereoisomers with the creation of three contiguous asymmetric centers. The antimicrobial investigation revealed the higher bactericidal and fungicidal activity of compound 1 followed by 3 and 4, against reference strains. Using in silico approaches, the present studies revealed that all tested compounds exhibited moderately-to-potent bioactivity score with higher level than aspirin. Only compound 2 do not satisfy the Lipinski’s rule of five. Analyzing the interaction of drug-like compounds with 32 antitarget end-points outlines that they must be interact with the examined antitarget proteins with different manner. Pre-ADME prediction properties provide a large spectrum of biologically for each target in good agreement with the in vitro antimicrobial properties. In addition, their toxicity prediction using pre-ADMET, Lazar, and GUSAR software showed various properties with good applicability domain in general suggesting their future testing as a potential candidate for drug discovery possessing several different end-points.

Experimental

Allyl bromide, N-benzylhydroxylamine, and trimethyl(trifluoromethyl)silane were used as purchased from Aldrich. TLC plates were inspected under UV light and developed by spraying with 5% H2SO4 in EtOH, followed by charring. The NMR spectra were registered on a Bruker instrument, operating at 400 MHz for 1H NMR, 100 MHz for 13C NMR, and 188 MHz for 19F NMR with the residual solvent as the internal standard. NMR solvent (CDCl3) was purchased from Aldrich. HRMS (LSIMS) data were registered in the positive mode (unless stated otherwise) using a Thermo Finnigan Mat 95 XL spectrometer.

(Z)-N-[[(3aR,5R,6S,6aR)-6-(Benzyloxy)-2,2-dimethyl-6-(trifluoromethyl)- tetrahydrofuro[2,3-d][1,3]dioxol-5-yl]methylene]-1-phenylmethanamine oxide (1, C23H24F3NNaO5)

To a solution of 340 mg 1,2-O-isopropylidene-3-O-methyl-3-C-trifluoromethyl-α-d-allofuranose [1] (0.89 mmol) in 10 cm3 H2O/EtOH (1:1) was added 192.3 mg NaIO4 (0.89 mmol). The reaction mixture is kept at room temperature for 1 h. The ethanol was evaporated and the residue was extracted with dichloromethane (3 × 10 cm3). The combined organic phases are dried over Na2SO4 and then evaporated to give the aldehyde sugar (300 mg, 72%). The latter is treated with N-benzylhydroxylamine hydrochloride in buffered medium (NaOAc) in 7.5 cm3 EtOH/H2O (3:2). The reaction mixture is stirred at room temperature for 12 h. Ethanol was evaporated and the crude was then extracted with dichloromethane (3 × 10 cm3). The combined organic phases are dried over Na2SO4, filtered, and then evaporated. The crude was then chromatographed on silica gel (PE/EtOAc 7:3) to give the trifluoromethylated nitrone 1 (330 mg, 85%). Colorless oil; Rf = 0.46 (PE/EtOAc (7:3); 1H NMR (400 MHz, CDCl3): δ = 1.24 (s, 3H, CH3), 1.28 (s, 3H, CH3), 4.66 (d, 1H, J2,1 = 4.0 Hz, H-2), 4.68 (d, 1H, Jgem = 10.8 Hz, OCH2Ph), 4.72 (d, 1H, Jgem = 10.8 Hz, OCH2Ph), 4.80 (s, 2H, NCH2Ph), 5.66 (br d, 1H, J = 7.6 Hz, H-4), 5.72 (d, 1H, J1,2= 4.0 Hz, H-1), 6.67 (br dd, 1H, J = 2.0 Hz, 7.6 Hz, H-5), 7.20 (m, 10H, Ar–H) ppm; 13C NMR (100 MHz, CDCl3): δ = 26.7 (CH3), 26.8 (CH3), 70.6 (OCH2Ph), 70.6 (NCH2Ph), 73.4 (C-4), 79.7 (C-2), 87.2 (q, 1JC,F = 24.8 Hz, C-3), 104,2 (C-1), 114.1 (C(CH3)2), 127.5, 127.6, 128.1, 128.9, 129.2, 129.3, 131.1(C-5), 131.9, 137.3 ppm; 19F NMR (188 MHz, CDCl3): δ = − 76.65 (s, 3F, CF3) ppm; HRMS (ESI): m/z calcd for C23H24F3NNaO5 ([M+Na]+) 474.1504, found 474.1493.

(3S,5S)-2-Benzyl-3-[(3aR,5R,6R,6aR)-6-(benzyloxy)-2,2-dimethyl-6-(trifluoromethyl)tetrahydrofuro[2,3-d][1,3]dioxol-5-yl]-5-(bromomethyl)isoxazolidine (2, C26H29BrF3NO5)

To a solution of 80 mg of the trifluoromethylated nitrone 1 (0.177 mmol) in 5 cm3 of toluene was added 123.5 mm3 allyl bromide (1.41 mmol). The mixture was stirred at reflux of toluene for 6 h. A TLC shows the total disappearance of the nitrone and the formation of four new cycloadducts. Only the major compound 2 (44 mg, 42%) was isolated by silica gel chromatography. Viscous clear liquid; Rf = 0.32 (EtOAc/PE 1:9); 1H NMR (400 MHz, CDCl3): δ = 1.40 (s, 3H, CH3), 1.56 (s, 3H, CH3), 2.18 (dq, 1H, J = 8.4 Hz, 12.4 Hz), 2.70 (dt, 1H, J = 6.8 Hz, 12.4 Hz), 3.36 (dd, 1H, J = 6.8 Hz, 10.4 Hz, CH2Br), 3.47 (dd, 1H, J = 4.8 Hz, 10.4 Hz, CH2Br), 3.51 (m, 1H, H-5), 3.93 (d, 1H, J = 14.0 Hz, NCH2Ph), 4.04 (d, 1H, J = 14.0 Hz, NCH2Ph), 4.30 (m, 2H), 4.76 (d, 1H, J = 10.8 Hz, OCH2Ph), 4.81 (d, 1H, J = 4.0 Hz, H-2), 4.90 (d, 1H, J = 10.4 Hz, OCH2Ph), 5.88 (d, 1H, J = 4.0 Hz, H-1), 7.16 (m, 3H, Ar–H), 7.24 (m, 2H, Ar–H), 7.33 (m, 5H, Ar–H) ppm; 13C NMR (100 MHz, CDCl3): δ = 26.8 (2 CH3), 33.8 (CH2Br), 34.2 (CH2), 61.6 (NCH2Ph), 63.1 (C-5), 70.3 (OCH2Ph), 76.9 (C-2), 78.5 (C-4), 79.8 (CH), 104.3 (C-1), 113.7 (C(CH3)2), 127.2, 127.8, 128.1, 128.2, 129.2, 136.8, 137.4, 174.7 ppm; HRMS (ESI): m/z calcd for C26H29BrF3NNaO5 ([M+Na]+) 594.1079, found 594.1062.

Compounds 3 and 4

To a solution of 93 mg of the trifluoromethylated nitrone 1 (0.21 mmol) in 10 cm3 of toluene was added 100 mg maleimide (1.03 mmol). The reaction mixture was refluxed with toluene for 6 h. TLC shows the total disappearance of the nitrone and the formation of two new compounds. After evaporation of the solvent, the residual oil was purified by flash chromatography (PE/EtOAc 6:4) to give 42 mg 3 (37%) and 68 mg 4 (60%).

(3R,3aS,6aR)-2-Benzyl-3-[(3aR,5R,6S,6aR)-6-(benzyloxy)-2,2-dimethyl-6-(trifluoromethyl)tetrahydrofuro[2,3-d][1,3]dioxol-5-yl]dihydro-2H-pyrrolo[3,4-d]isoxazole-4,6(5H,6aH)-dione (3, C27H27F3N2O7)

Colorless oil; Rf= 0.35 (EtOAc/PE 4:6); 1H NMR (400 MHz, CDCl3): δ = 1.37 (s, 3H, CH3), 1.61 (s, 3H, CH3), 3.93 (d, 1H, J = 9.6 Hz), 4.03 (m, 2H, H-5, NCH2Ph), 4.15 (d, 1H, J = 14.8 Hz, NCH2Ph), 4.41 (m, 2H), 4.74 (d, 1H, J = 10.4 Hz, OCH2Ph), 4.83 (d, 1H, J = 4.0 Hz,H-2), 4.91 (d, 1H, J = 10.4 Hz, OCH2Ph), 5.84 (d, 1H, J = 4.0 Hz, H-1), 7.19 (m, 5H, Ar–H), 7.29 (m, 5H, Ar–H), 8.33 (s, 1H, NH) ppm; 13C NMR (100 MHz, CDCl3): δ = 26.7 (CH3), 26.72 (CH3), 50.4 (C-5), 57.4 (NCH2Ph), 65.3, 70.8 (OCH2Ph), 76.2, 76.3, 77.9 (C-2), 85.1 (q, 1JC,F = 25.8 Hz, C-3), 104.1 (C-1), 114.0 (C(CH3)2), 127.1, 127.6, 128.0, 128.1, 128.2, 128.5, 135, 137, 174.2 (C=O), 175.2 (C=O) ppm; HRMS (ESI): m/z calcd for C27H27F3N2NaO7 ([M+Na]+) 571.1668, found 571.1663.

(3S,3aR,6aS)-2-Benzyl-3-[(3aR,5R,6S,6aR)-6-(benzyloxy)-2,2-dimethyl-6-(trifluoromethyl)tetrahydrofuro[2,3-d][1,3]dioxol-5-yl]dihydro-2H-pyrrolo[3,4-d]isoxazole-4,6(5H,6aH)-dione (4, C27H27F3N2O7)

Colorless oil; Rf = 0.33 (EtOAc/PE 4:6); 1H NMR (400 MHz, CDCl3): δ = 1.42 (s, 3H, CH3), 1.64 (s, 3H, CH3), 3.96 (m, 2H, NCH2Ph), 4.03 (dl, 1H, J = 8.0 Hz), 4.17 (dl, 1H, J = 5.2 Hz, H-5), 4.51 (m, 1H, H-4), 4.76 (m, 2H), 4.88 (m, 2H, H-2, OCH2Ph), 5.90 (d, 1H, J = 4.0 Hz,H-1), 7.12 (m, 5H, Ar–H), 7.32 (m, 5H, Ar–H), 8.57 (s, 1H, NH) ppm; 13C NMR (100 MHz, CDCl3): δ = 26.6 (CH3), 26.7 (CH3), 52.6, 59.5 (NCH2Ph), 64.8 (C-5), 70.2 (OCH2Ph), 78.2, 78.4 (C-4), 78.8 (C-2), 85.1 (q, 1JC,F = 25.3 Hz, C-3), 104.2 (C-1), 113.9 (C(CH3)2), 127.4, 127.8, 128.2, 128.3, 128.6, 135.8, 137.1, 174.7 (C=O), 176.0 (C=O) ppm; HRMS (ESI): m/z calcd for C27H27F3N2NaO7 ([M+Na]+) 571.1668, found 571.1659.

2-[1-Benzyl-3-[(3aR,5R,6S,6aR)-6-(benzyloxy)-2,2-dimethyl-6-(trifluoromethyl)tetrahydrofuro[2,3-d][1,3]dioxol-5-yl]aziridin-2-yl]-2-oxoethyl acetate (6, C28H30F3NO7)

To a solution of 54 mg of the trifluoromethylated nitrone 1 (0.12 mmol) in 5 cm3 of toluene was added 15.4 mm3 propargyl acetate (0.16 mmol). The mixture was stirred at reflux of toluene for 6 h. A CCM shows the total disappearance of the nitrone and the formation of a new stain. The crude reaction product was chromatographed on a silica gel (EtOAc/PE 3:7) to give 6 (37 mg, 55%).Viscous clear liquid; Rf = 0.46 (EtOAc/PE 3:7); 1H NMR (400 MHz, CDCl3): δ = 1.36 (s, 3H, CH3), 1.57 (s, 3H, CH3), 2.10 (s, 3H, CH3), 2.42 (t, 1H, J = 6.8 Hz, H-5), 2.48 (d, 1H, J = 6.8 Hz), 3.22 (d, 1H, J = 13.2 Hz, NCH2Ph), 4.00 (d, 1H, J = 13.2 Hz, NCH2Ph), 4.42 (dl, 1H, J = 7.2 Hz Hz, H-4), 4.56 (d, 1H, J = 17.2 Hz), 4.78 (d, 1H, J = 4.0 Hz, H-2), 4.85 (m, 2H, OCH2Ph), 4.96 (d, 1H, J = 17.2 Hz), 5.79 (d, 1H, J = 4.0 Hz, H-1), 7.34 (m, 10H, Ar–H) ppm; 13C NMR (100 MHz, CDCl3): δ = 20.4 (CH3), 26.7 (CH3), 26.9 (CH3), 44.8, 44.9 (C-5), 63.1 (NCH2Ph), 67.9 (CH2), 70.3 (OCH2Ph), 78.0 (C-4), 79.4 (C-2), 85.3 (q, 1JC,F = 26.4 Hz, C-3), 104.2 (C-1), 114.0 (C(CH3)2), 127.5, 127.7, 127.8, 127.9, 128.1, 128.2, 128.4, 136.8, 137.5, 169.9 (O-C=O), 198.8 (C=O) ppm; 19F NMR (188 MHz, CDCl3): δ = − 81.81 (s, 3F, CF3) ppm; HRMS (ESI): m/z calcd for C28H30F3NNaO7 ([M+Na]+) 572.1872, found 572.1865.

Physicochemical and ADMET properties

The prediction of physicochemical properties such as LogP, TPSA, bioactivity score, and drug likeness of the synthesized molecules were assessed using Molinspiration software [43]. ADMET (Absorption, Distribution, Metabolism, Excretion, and Toxicity) properties of the title compounds such as permeability for Caco-2 cell, MDCK cell, and BBB (blood–brain barrier), HIA (human intestinal absorption), skin permeability, and plasma protein binding as well as toxicological properties from chemical structures, such as mutagenicity and carcinogenicity were analyzed using Pre-ADMET software [32].

Antimicrobial activities

The antimicrobial activities were tested against seven microorganisms including three Gram-positive (Bacillus subtilis JN 934392, Bacillus cereus JN 934390, and Staphylococcus aureus ATCC 6538) and two Gram-negative (Salmonella entericsero type Enteritidis ATCC43972 and Escherichia coli ATCC 25922) bacterial strains, and two fungal strains (Fusarium oxysporum and Fusarium phyllophilum AB587006). The determination of the inhibition zone diameter (IZ), minimum inhibitory concentration (MIC), minimum bactericidal concentration (MBC), and minimum fungicidal concentration (MFC) were based on the same protocol as described by Felhi et al. [44] and Ghannay et al. [20].