2.2. Metabolite Identification and Transformation Pathway
In the following study, seven hepatic metabolites of miconazole were found and identified using high-resolution mass spectrometry. The fragmentation patterns of miconazole and its metabolites are summarized in
Table 1, while the fragmentation MS/MS spectra are presented in
Figure 1,
Figure 2,
Figure 3,
Figure 4,
Figure 5,
Figure 6,
Figure 7 and
Figure 8.
The spectrum of miconazole, presented in
Figure 1, is characterized by prominent fragments originating from the detachment of the dichlorobenzyl moiety, particularly at the oxygen linkage site. The most conspicuous peak in the spectrum originates from the detached fragment (158.9764
m/
z). The remaining structure is represented by a fragment at 255.0061
m/
z. The fragmentation process of this fragment involves the successive loss of oxygen (resulting in a fragment at 239.0141
m/
z), which in turn decomposes into a fragment without a chlorine atom (203.0364
m/
z) or, alternatively, undergoes fragmentation on the opposite side, leading to the removal of a carbon atom from the imidazole ring and subsequent rearrangement of the structure, as evidenced by the fragment at 227.0137
m/
z. In some instances, the entire imidazole ring may be lost, leading to the formation of a fragment at 172.9900
m/
z. The spectrum also contains visible fragments indicating the preservation of imidazole (69.0457
m/
z) and 2-(1-imidazolyl)ethanol (111.0566
m/
z). Another noteworthy fragment in the spectrum is observed at 379.0206
m/
z, which signifies the detachment of a chlorine atom.
M1 represents the derivative where hydroxylation and oxygenation occurred at the carbon atoms within the imidazole ring (
Figure 2). The primary peak in the M1 spectrum corresponds to 2-(2,4-dichlorophenyl)ethylamine (188.0034
m/
z), and the main peak from the parent spectrum (158.9760
m/
z) is also evident in the M1 spectrum. Notably, the spectrum lacks the fragment observed in the parent spectrum fragment resulting from the detachment of dichlorobenzyl. However, one of the main peaks is represented by a fragment with an
m/
z value of 273.0218, which indicates the presence of two hydroxyl groups and two hydrogen atoms within this fragment. The detachment of oxygen and carbon from the structure at 273.0218
m/
z results in the formation of the ion at 245.0246
m/
z. The fragment originated from the detachment of the imidazole moiety, with an ion at
m/
z 170.9801, indicating that redox reactions occurred outside of the phenyl ring. Additionally, the fragment with an
m/
z value of 86.0242 represents dihydroxyazetidine, providing direct evidence of a double oxidation within the imidazole structure. The fragment at 430.9880
m/
z resulting from dehydrogenation indicates that
N-oxidation within the imidazole moiety did not occur. While metabolic reactions that occur in the case of M1 are not frequent, there are documented cases in which CYP enzymes are responsible for such reactions. For instance, a comparable metabolite in humans involving an imidazole moiety has nafimidone [
20]. The double oxidation of the monosubstituted imidazole ring has been documented in the metabolism of econazole [
21], which is another representative imidazole antifungal. Based on these, we assume that oxygenation took place at the C2 position and hydroxylation occurred at the C4 position of the imidazole ring.
M2 has been identified as a product of
N-oxidation, primarily based on the presence of a distinct fragment in its spectrum with an
m/
z value of 85.0408, indicating oxidation within the imidazole moiety (
Figure 3). Notably, the primary peak in the spectrum corresponds to the detachment of the dichlorophenyl moiety (158.9765
m/
z), suggesting that oxidation did not occur within this moiety. Additionally, the spectrum reveals an ion peak from 2-(2,4-dichlorophenyl)ethylamine (188.0013
m/
z), indicating that similar processes occurred in these compounds. The nitrogen atom’s basicity being blocked resulted in a significant extension of the retention time in the applied chromatographic system. Notably, the MS/MS spectrum of M2 is notably deficient, a characteristic often observed in transient
N-oxides.
M3 and M4 are isomers of metabolites in which hydroxylation has occurred within the phenyl ring. In the spectrum of M3, the main ion is observed at
m/
z 174.9709, replacing the fragment at 158.9764
m/
z from the parent compound’s MS/MS spectrum (
Figure 4). This indicates that hydroxylation took place in the dichlorobenzyl fragment. A prominent peak with an
m/
z value of 69.0453 represents imidazole. Additionally, the spectrum features a visible fragment at
m/
z value 110.9980, representing chlorophenyl.
In the spectrum of M4 (
Figure 5), two primary peaks are notable: one at 69.0451
m/
z and the other at 174.9708
m/
z. These peaks correspond to the imidazole and hydroxylated dichlorobenzyl fragments, respectively. The peak at 257.0237
m/
z represents the remaining structure after the detachment of the 174.9708
m/
z fragment, indicating a structure in which oxidation took place. Further fragmentation of this fragment leads to the appearance of a peak at 188.9897
m/
z, which occurs following the detachment of the imidazole moiety. The spectrum also includes the fragment at 110.9980
m/
z, as observed in the M3 spectrum, indicating that hydroxylation of the phenyl ring results in different fragmentation behavior, making it more prone to the loss of a chlorine atom during fragmentation. Drawing from the insights provided by Rietjens et al., who presented a hypothesis for predicting the regioselectivity of the cytochrome P450 hydroxylation of halogenated benzenes, we can infer that in the case of M3, hydroxylation occurred at the C4 position of the benzene ring, while in the case of M4, hydroxylation took place in the meta position relative to the chloro substituents [
22]. De Souza et al. have previously identified a product in which hydroxylation took place in the benzene ring of miconazole [
18]. However, this was not a miconazole metabolism product, but a transformation product formed during the ozonation and electro-oxidation of miconazole.
M5 represents another mono-oxygenated metabolite, with hydroxylation occurring within the imidazole moiety. Consistent with the spectra of other metabolites, a peak with an
m/
z value of 158.9754 was the most prominent in its spectrum (
Figure 6). The presence of a peak at 85.0398, which replaces the fragment representing imidazole (69.0457
m/
z), provides direct evidence that hydroxylation took place within the imidazole moiety. Additionally, the spectrum features a noticeable peak at 243.3078, indicating that the fragment visible in the parent spectrum at 227.0137
m/
z underwent hydroxylation. Based on the metabolism of econazole, we assume that in the case of M6, hydroxylation takes place at the carbon atom positioned between the two nitrogen atoms within the imidazole moiety [
21].
The last two metabolites, M7 and M6, were formed by the successive breakdown of the imidazole moiety of M1. Initially, M1 underwent a transformation resulting in the formation of M7. Subsequently, M7 underwent further degradation, leading to the complete degradation of the imidazole moiety and the formation of the free amine M6. The primary peak in the spectrum of M6 (
Figure 7) corresponds to a fragment (188.0025
m/
z) that results from the detachment of the dichlorobenzyl fragment and the loss of an amine group. The spectrum of M6 prominently features a peak originating from dichlorobenzyl (158.9744
m/
z). Another fragmentation pathway involves the detachment of the dichlorobenzoxy moiety, as well as the loss of one chlorine atom from the left benzene ring, resulting in the rearrangement of a structure in a fragment with an
m/
z of 153.0352. Subsequently, the second chlorine atom is lost from the benzene ring, leading to the formation of a fragment at 117.0563
m/
z. The rearrangement of the structure likely occurred due to the presence of the free amine group, which served as an electron donor to the carbon atom in the benzene ring after the detachment of the chlorine atom. The formation of this product in degradation studies was also reported by de Souza et al. [
18].
M7 was formed through the detachment of the hydroxyl group along with a carbon atom from the imidazole ring. In the spectrum of M7 (
Figure 8), there is also a peak at an
m/
z value of 158.9758, which, as with other metabolites, excludes metabolic changes within the structure of dichlorobenzyl. The primary peak in the spectrum (245.0226
m/
z) corresponds to the structure formed after the detachment at the above-mentioned 158.9758
m/
z. The fragment at 245.0246
m/
z was also visible in the MS/MS spectrum for M1, suggesting that M7 is formed following a partial breakdown of the imidazole moiety in the M1 structure. Additionally, there is a visible prominent fragment with an
m/
z of 188.0033, which corresponds to 2-(2,4-dichlorophenyl)ethylamine. The metabolic pathway of miconazole is presented in
Figure 9.
2.5. Electrochemical Experiments
The primary electrochemical products observed during this study were M2 and M6 (
Figure S3). The product formed from the decomposition of miconazole, M6, was readily formed in electrochemical reactions, and its formation demonstrated an increase as the applied potential was raised. Its peak formation potential occurs at 2.4 V and it was the sole metabolite formed in a substantial quantity under these conditions. M2, recognized as the
N-oxide metabolite, was the most easily formed oxidation product and the main metabolite formed at low potentials, with its potential range of 0.8 to 1.2 V. Specifically, the peak formation for M2 was situated at a potential of 0.8 V, although its formation declined at higher potentials. Nonetheless, an optimal condition for its synthesis was observed at a potential of 2.0 V. In a manner similar to the
N-oxide, in the case of M3, M4, and M5, two distinct potential optima were also observed. The first peak formation potential for M3 occurs at 0.8 V, while for M4 and M5, it was at 1.2 V. Nevertheless, a significantly greater amount of these metabolites was formed at the second optimum of 2.0 V. At this potential, M1 was also created with maximum effectiveness, even though it was formed only in small quantities at low potentials. However, as the potential increased further, the efficiency of their formation began to decrease, likely due to increased decomposition of the resulting electrochemical products. As a consequence of the inefficient formation of M1, the formation of M7 using electrochemical methods could not be obtained; however, its formation in the biological experiment was also very slight. It should be noted that electrochemical methods helped in the synthesis of minor metabolites, for example, M6, with greater efficiency, which was an advantage in terms of its identification and structural characterization. The high efficiency in the formation of M2 by electrochemical methods has assisted in the analysis of its MS/MS spectra.
These findings suggest the existence of a distinct reaction mechanism at potentials within the range of 0.8–1.2 V in contrast to those around 2.0 V. The anodic oxidation of organic compounds can take place through two distinct mechanisms: direct electron transfer (DET) and oxidation mediated by hydroxyl radicals. DET is a process where electrons are directly transferred between an electrode and a redox-active molecule or species in solution, occurring without the intervention of redox mediators. On the other hand, at high anodic potentials in the region of water discharge, organic compounds can also be oxidized due to the participation of intermediates of oxygen evolution.
2.6. Toxicity
In this study, the toxicity assessment was performed using the in silico approach. Although the in vivo methods (utilizing living organisms) are still considered the most reliable, they possess several serious drawbacks as they are expensive, time-consuming, and, above all, questionable from an ethical point of view. Therefore, it is considered important not to use them unless it is necessary. Thus, other approaches, such as methods based on computational modelling (in silico) have gained great popularity recently. Obviously, their reliability varies depending on the calculation method or the applicability domain, but it is widely accepted that if they are properly used, they can be successfully applied at the preliminary stage of research [
23,
24,
25].
In order to maximize the reliability of the estimations, we applied more than one model for each toxicity category (unless only one model was available). In cases where a large number of models were used, the outcomes were submitted to PCA in order to facilitate interpretation of the results as well as to graphically present the relationships between the metabolites and applied QSAR models. It should be noted that scores of PCA represent the identified metabolites and the parent compound, while loadings represent the toxicity models.
2.6.1. Aquatic Toxicity
The descriptions of twenty-three models used to calculate the toxicity of miconazole metabolites towards aquatic organisms are summarized in
Section S1. Due to a high number of applied toxicity models, the obtained results were divided into three subgroups:
D. magna, Fish (including zebrafish and
F. minnow), and algae, and subsequently submitted to PCA (when interpreting the results of chemometric analysis, it should be remembered that toxicity decreases in parallel with increasing values such as LD
50, LC
50, EC
50, etc.). All raw data concerning aquatic toxicity are presented in the
Supplementary Material (Tables S3–S5).
Fish
In order to facilitate the interpretation of the fish toxicity results predicted by 10 models, PCA was performed. As shown in
Figure 12A, there are four pairs of highly correlated (giving similar results) variables. The first two pairs, Zebrafish Embryo AC
50 IRFMN/Coral Fish Acute LC
50 KNN/Read-Across and Fish Acute LC
50 IRFMN–Fish Acute LC
50 IRFMN/Combase, predicted the lowest toxicity for the products of aromatic hydroxylation (M4 and M3) and relatively low toxicity for the products of imidazole oxidation and its total decomposition (M2, M5, and M6). On the other hand, the parent compound and M7 were defined as generally the most toxic. Interestingly, the
F. minnow 96 h T.E.S.T. model gave opposing results. The third pair of correlated variables consists of the ECOSAR predictions: Fish 96 h NO SAR–Fish ChV (predictions of Fish Acute LC
50 NIC were quite close to this pair). Those models predicted the lowest toxicity for M1, moderate toxicity for M2 and M6, and the highest toxicity for M5 and miconazole. The results predicted by the last pair of correlated variables (
F. minnow LC
50 96 h EPA–Fish Chronic NOEC IRFMN) were generally close to the above-mentioned lowest-toxicity estimations for M1 and highest estimations for M5 and the parent compound. Although it may be difficult to unambiguously indicate the most and the least harmful compounds due to the diversity of the applied models’ predictions, it can be assumed that the metabolites are generally less toxic than the parent ion, and hydroxylation of the imidazole ring (M5) changes its properties to the least extent. On the other hand, the toxicity of the major metabolite (M1) should be considered as relatively low.
D. magna
In contrast to fish, the majority of the models predicting toxicity towards daphnia gave highly correlated results (
Figure 12B), indicating the lowest toxicity for M1 and generally the highest for miconazole. Predictions for the remaining metabolites were diverse; however, the outcomes were generally closer to the parent compound (higher toxicity) than to M1 (similarly as in the other aquatic toxicity categories and, in this case, the higher loading values correspond to decreased toxicity, which is justified by increasing LC
50 or EC
50 values). Interestingly, one pair of correlated models (
D. magna LC
50 48 h DEMETRA–
D. magna LC
50 48 h EPA) indicated that aromatic hydroxylation (M3 and M4), and, to a lesser extent,
N-oxidation (M2) results in a significant decrease of the toxic properties. On the other hand, the most outlying model (
D. magna Acute EC
50 Toxicity model IRFMN/Combase) gave opposite results, also predicting miconazole as the least toxic.
Algae
The results obtained for algal toxicity were the most divergent (
Figure 12C). Only two models (both provided by ECOSAR) gave similar results, predicting M1 as the least toxic, while miconazole, along with M5, was the most harmful. The lowest toxicity of all hydroxylated metabolites (M3–M5) was predicted by the Algae Acute EC
50 IRFMN model (conversely, the opposite results were generated by the Algae Acute EC
50 ProtoQSAR/Combase model, which predicted the lowest toxicity for M6).
2.6.2. Acute Toxicity to Rodents
Toxicity to mice and rats was calculated using seven models: six provided by Percepta (rat and mouse oral, mouse intravenous, intraperitoneal and subcutaneous, and rat intraperitoneal) and one provided by T.E.S.T. (rat oral, nearest neighbor model with non-relaxed fragment constraint). All outcomes were expressed as LD
50, using log
(mg/kg) values (raw data are presented in
Table S6). As can be seen in the PCA plot (
Figure 12D), there are two groups of correlated variables: the first one consisting of Mouse SC, IP, and IV models, and the second one grouping Rat IP, OR, and Mouse OR (a slightly more outlying vector) together. However, these two groups of variables are weakly correlated with each other, as the angle between them is close to 90°. On the other hand, the variable representing the T.E.S.T. model (Rat OR) is inversely correlated with the second group of variables (which means that this model predicted contrasting results). Considering the compounds, it can be seen that miconazole, M3, and M5 possess rather similar toxic properties. M1 is significantly less toxic according to the second group of variables (and more toxic according to the Rat OR T.E.S.T. model. The results should be interpreted in the same way as in the case of aquatic toxicity). Inverse results were obtained for M6. On the other hand, the first group of models predicted M2 as the least toxic and M4 as the most harmful compound. Based on the obtained results, it can be seen that there is no apparent dependence between the structure of the metabolite and its toxicity to rodents.
2.6.3. Mutagenicity and Developmental Toxicity
The mutagenic potential (expressed as a probability of the positive outcome of the Ames test) of the miconazole metabolites was estimated using two models provided by Percepta and T.E.S.T. software. According to the Percepta model, none of the discussed compounds possess mutagenic potential. However, M1, M2, and M7 are marked with a slightly higher probability of such action. On the other hand, T.E.S.T. software predicted significantly higher values. According to this model, the parent compound possesses a borderline probability of mutagenic action (0.5 but defined as ‘mutagenic positive’ by the software). Moreover, the predicted mutagenicity of miconazole N-oxide (M2) was exceptionally high (0.82). The remaining metabolites were defined as less toxic than the parent compound.
Developmental toxicity was calculated using one model provided by the T.E.S.T. software. The obtained outcomes indicate that all the metabolites may possess greater toxic potential than miconazole itself (which is in fact the only compound defined as non-developmentally toxic), especially in the cases of M1 (the most abundant metabolite) and M7.
2.6.4. Receptor-Mediated Toxicity
The endocrine-disruption potential (receptor-mediated toxicity) was estimated using the Endocrine Disruptome platform. The following target receptors were chosen: androgen (AR), estrogen alpha (ERα) and beta (ERβ), progesterone (PR), glucocorticoid (GR), mineralocorticoid (MR), liver X alpha (LXRα), retinoid X alpha (RXRα), peroxisome proliferator-activated receptor gamma (PPARγ), thyroid alpha (TRα), and beta (TRβ). In the cases of AR, ERα, ERβ, and GR receptors, the antagonistic interactions were also studied.
The obtained results are shown in
Table S8. Numerical values presented in the table correspond to the binding probability (increasing in parallel with decreasing numbers); however, they should be compared only column-wise, e.g., a value of −9.8 represents a high binding probability in the case of MR, but is still low in the case RXRα. Therefore, each receptor should be considered separately. Miconazole and the majority of its metabolites possess a high potential of binding to MR (two exceptions are M5 and M7, where moderate-high and moderate-low probabilities were observed, respectively). All the analyzed compounds can also bind to GR (moderate-low probability) and AR as the antagonists of this receptor (also moderate-low probability, except M7, where a moderate-high probability was predicted). Most of the compounds may also bind to TRβ, albeit at a moderate-low level of probability. Among the metabolites, M1, M2, and M5 can be viewed as the most harmful. M1 binds to four receptors at moderate-low levels (the same number as miconazole); however, it is the only compound possessing high probability in the case of two proteins, MR and ERβ (as its antagonist). M2 and M5 bind to six and seven receptors at moderate-low levels (both can strongly bind to MR). M7 can also be considered as more toxic than the parent compound (moderate-low binding to five receptors, moderate-high to one, and high to one).
Based on the obtained results, it can be seen that the majority of the metabolites possess higher endocrine-disrupting potential than the parent compound (only M4 and M6 have definitively lower binding probabilities). Noticeably, the most harmful compounds were formed as a result of imidazole ring hydroxylation/oxidation, and even partial decomposition of this moiety did not reduce such properties (M7). On the other hand, the total degradation of imidazole, resulting in the formation of primary amine, did not significantly alter the studied properties. In addition, hydroxylation of the dichlorophenyl ring did not increase the toxicity, and in the case of M4, a decrease was observed.