Results

Analysis of normalised reactivation rate constants

Initial analysis of the normalised first-order observed reactivation rate constant (_k_obs) for 115 compounds tested for BChE reactivation after inhibition by GA, GB, GF, and VX showed that the overall oxime reactivation efficacy followed the order GF > VX > GB > GA. Mean normalised _k_obs values with standard deviation are 1.05±1.17, 1.82±1.45, 2.06±1.32 and 3.87±1.26 for GF, GB, VX and GA reactivation, respectively (Figure 2).

Figure 2.: Statistical analysis of the normalised first-order observed reactivation rate constant (kobs), determined for 115 oximes in reactivation of BChE inhibited with sarin (GB), cyclosarin (GF), tabun (GA) and VX [9]. Due to the logarithmic normalisation, a negative logarithmic value for inactive compounds in BChE reactivation is set to 5 (kobs = 0 min-1). Black or red bars represent mean ± standard deviation of the normalised kobs

PC analysis of BChE reactivation efficacy

Principal component (PC) analysis of 115 compounds tested for BChE reactivation after inhibition with GA, GB, GF and VX, using a dataset comprising 27 variables, resulted in the selection of three PCs, each with eigenvector values higher than 1, and 72.04 % of cumulative proportion of variance (Table 1, Figure 3, Figure S1). Eigenvector values of the covariance matrix were 11.46, 4.18 and 3.8 for PC1, PC2, and PC3, respectively. Calculated loadings for 27 variables indicate the contribution of each variable to each PC; they are limited to the range ±1. The sign of the loading indicates a positive or negative contribution to the PC, while its magnitude provides a direct measure of the PC linear relationship. Principal component 1 (PC1) contains 42.4 % of the proportion of variance and has 8 positive PC loadings related to reactivation parameters, normalised _k_obs, and the time (at which Reactmax was achieved), for all four OP compounds. Oxime inhibition of control enzyme activity, and the maximal percentage of reactivation (Reactmax) parameters have negative loading values. Interestingly, all pharmacological parameters also have negative loading values. Molecular weight, number of rotational bonds, molecular volume, and molecular surface area have loadings below 0.9 (Table 1, Figure 4).

 | Loadings
PC1 (42.4 %) | PC2 (15.5 %) | PC3 (14.1 %)
Eigenvector value | 11.46 | 4.18 | 3.82
Time, h (GB) | 0.755 | 0.427 | -0.069
-log (kobs / min-1) (GB) | 0.741 | 0.523 | -0.218
-log (kobs / min-1) (GF) | 0.681 | 0.167 | -0.381
-log (kobs / min-1) (VX) | 0.678 | 0.489 | -0.373
Time, h (VX) | 0.663 | 0.433 | -0.375
Time, h (GF) | 0.643 | 0.365 | -0.416
-log (kobs / min-1) (GA) | 0.338 | 0.562 | 0.138
Molecular fractional polar surface area, nm2 | 0.301 | 0.429 | 0.304
Time, h (GA) | 0.084 | 0.458 | 0.189
ADMET Alog P98 | -0.189 | -0.333 | -0.870
Reactmax, % (GA) | -0.226 | -0.470 | -0.222
Alog P | -0.383 | -0.310 | -0.819
log D | -0.385 | -0.304 | -0.822
Dipole magnitude | -0.459 | 0.054 | -0.325
Oxime inhibition, % | -0.477 | 0.306 | -0.346
Reactmax, % (GF) | -0.542 | -0.249 | 0.184
Number of H bond donor atoms | -0.580 | 0.470 | 0.318
Reactmax, % (VX) | -0.731 | -0.306 | 0.276
Number of rings | -0.756 | 0.424 | -0.393
Number of aromatic rings | -0.773 | 0.399 | -0.388
Number of H bond acceptor atoms | -0.774 | 0.523 | 0.088
Reactmax, % (GB) | -0.793 | -0.357 | 0.265
ADMET PSA 2D | -0.814 | 0.491 | 0.151
Molecular weight | -0.903 | 0.382 | -0.122
Number of rotational bonds | -0.907 | 0.255 | 0.005
Molecular volume, nm3 | -0.911 | 0.346 | -0.095
Molecular surface area, nm2 | -0.915 | 0.343 | -0.031

Figure 3.: Principal component (PC) analysis of 115 compounds tested for BChE reactivation after inhibition with sarin (GB), cyclosarin (GF), tabun (GA), and VX using a dataset comprising 27 variables. (A) Correlation of PC1 and PC2 containing 42.4 and 15.5 % of proportion of the variance, and (B) correlation of PC2 and PC3 containing 15.5 and 14.1 % of proportion of the variance (Table 1). Dark green circles represent the most active oximes in the BChE reactivation, and dark red circles represent the inactive oximes in the reactivation

Figure 4.: Distribution of principal component (PC) loadings. The dataset includes 27 variables for 115 compounds tested for BChE reactivation after inhibition with sarin (GB), cyclosarin (GF), tabun (GA), and VX. Bars are ranked by the values of PC1 loadings. Top PC1 loadings are the time at which Reactmax was achieved and normalised first-order observed reactivation rate constant (kobs), for GB, GF and VX. The lowest PC1 loadings are molecular weight, number of rotational bonds, molecular volume, and molecular surface area. For further details about the distribution of variables, please see Table 1

Principal component 2 (PC2) accounts for 15.5 % of the variance and includes 20 positive loadings, encompassing the reactivation parameters and the majority of pharmacological descriptors. The exceptions were log D, Alog P and ADMET Alog _P_98, which are all related to describing a compound’s lipophilicity via the octanol-water partition coefficient. The reactivation parameter Reactmax shows negative loading values for all four OP compounds.

Principal component 3 (PC3) accounts for 14.1 % of the variance and has 10 positive loadings, including Reactmax for GB, GF, and VX, with the exception of GA. The reverse pattern is observed for the normalised _k_obs and the time at which Reactmax was achieved, where positive loadings were attributed to GA and negative loadings to GB, GF and VX.

Distribution of transformed dataset points after correlation of PC1 and PC2 (Figure 3) shows the position of the oxime 16C, with the highest reactivation efficacy according to normalised GF _k_obs, in the lower part of Figure 3A with coordinates (0.4, -4.1), and one of inactive oximes, 19D, in the far-right part of the figure with coordinates (7.7, 1.8). A distinct grouping of effective oximes can be observed in the top left part of the plot, corresponding to positive PC2 values. The PC3 axis values, in correlation with PC2, appear to discriminate oximes based on their reactivation efficacy, because most moderately effective and ineffective oximes are positioned in the lower part of the plot, characterized by negative PC3 values (Figure 3B). In contrast, highly effective oximes are associated with positive or near-zero PC3 values.

PC analysis of reactivation for individual organophosphorus compounds

Individual PC analysis of each OP compound reactivation dataset resulted in 18 PCs, but only three had covariance matrix eigenvectors larger than 1. On average, PC1 contains 49.8 %, PC2 20.0 % and PC3 10.9 % of variance, with matching standard deviations 1.96, 1.0 and 0.53 % (Supplementary material, Tables S2 to S5). This similarity in the corresponding proportions of variance is due to the relatively small number of kinetic variables (four) compared with the 14 pharmacological parameters shared across all four OP compounds. A similar trend is observed in the average values of the eigenvectors for PC1, PC2, and PC3. The values across all four OP compounds were 8.96±0.35, 3.60±0.18 and 1.96±0.09, respectively. The distribution of transformed dataset points is similar among GB, GF and VX data. However, the GA data show a different pattern (Figure 5). PC1 positioned ineffective oxime reactivators on the far right for all four OPs, with the highest positive values (>5). For the GA dataset, ineffective oximes have both PC1 and PC2 negative values. In the correlation of PC2 and PC3, PC3 discriminates oximes by reactivation efficacy. Most of the moderate and ineffective oximes are in the lower part of the graph, having negative PC3 values (Figure 5). Additionally, only PC3 of the GA dataset yielded positive values for ineffective oximes (Figure 5, bottom-right panel).

Figure 5.: Principal component analysis (PCA) on 115 compounds tested for BChE reactivation after inhibition with sarin (GB), cyclosarin (GF), tabun (GA), and VX. PCA was performed using a dataset comprising 18 variables (Tables S2 to S5). Left row panels represent PC1 vs. PC2 correlation, and right row panels represent PC2 vs. PC3 correlation. Dark green circles represent the most efficient oximes in the BChE reactivation, and dark red circles represent ineffective oximes in the reactivation.

The most efficient oxime for GB reactivation is 16C, and it is positioned in the top part of the PC1 vs. PC2 correlation plot at coordinates (1.63, 3.07). In contrast, oxime 19D, which is ineffective, not only for GB reactivation but also for other OPs, is located at coordinates (5.3, 0.05). Additionally, oxime 16C is the fastest in VX reactivation, and the PC1 vs. PC2 correlation positioned it at (1.9, 2.4), while inactive 19D is found at (5.1, 0.4) coordinates. In the case of GF reactivation, the best reactivator is oxime 4B, and the PC1 vs. PC2 correlation positioned it at (1.83, 1.98), while 19D is at coordinates (5.65, 0.84). For GA reactivation, the most efficient oxime 5C is positioned at (1.46, 1.04), while 19D is located at (4.4, - 0.1).

PC loading analysis reveals consistent patterns after ranking PC1 loadings for each OP compound (Figure 6, Table S2 to S5). The top three loadings are time, h; –log (_k_obs / min-1) and molecular fractional polar surface area. These are followed by ADMET Alog _P_98, Alog P and log D variables, which maintain the same order across all four OP compounds. Negative PC1 loadings also display a consistent order for all four OP compounds: number of H-bond acceptor atoms, ADMET PSA 2D, number of rings, number of aromatic rings, number of rotational bonds, molecular surface area, molecular volume, and molecular weight. Four variable loadings, which are also grouped but whose rank is interchangeable, are dipole magnitude, oxime inhibition of control enzyme activity, %, number of H-bond donor atoms and Reactmax, %. The similarity in PC1 loading ranks among OP compounds is attributed to the relatively low number of kinetic variables (four) compared to the fourteen pharmacological parameters shared across all four OP compound analyses.

Figure 6.: Distribution of loadings from individual PCA using a dataset comprising 18 variables for 115 compounds tested for reactivation of BChE inhibited by sarin (A), cyclosarin (B), VX (C) and tabun (D). Bars are ranked according to the values of PC1 loadings. On average, PC1 contains 46.9 %, PC2 19.1 % and PC3 12.4 % of variance. For further details about the distribution of variables, please see Tables S2 to S5

ADME results of studied oximes

The studied oximes were designed and synthesized as potential antidotes for the treatment of OP poisoning and therefore must fulfil the criteria required of viable drug candidates. In addition to sufficient activity toward the therapeutic target, a promising drug candidate must exhibit appropriate absorption, distribution, metabolism, and excretion (ADME) properties. Among these, BBB penetration is one of the most crucial factors, as OP compounds are lipophilic and readily cross biological barriers, including the BBB, thereby inhibiting synaptic AChE [18]. Therefore, the design of effective antidotes should incorporate BBB permeability assessment to identify oximes capable of reactivating ChEs in the CNS. Analyses of numerous approved CNS drugs have led to the establishment of general guidelines describing the physicochemical and ADME properties associated with successful CNS penetration [22]. Favourable physicochemical and ADME properties include adequate oral absorption and sufficient BBB permeability to enable therapeutic concentrations in the brain [23,24]. In this study, the predicted BBB penetration of the oximes was evaluated using a correlation plot of lipophilicity (Alog _P_98) and polar surface area (PSA 2D) (Figure 7). The results show that the majority of the compounds, with the exception of 15 (13 %), fall within the BBB-99 ellipse, indicating increased oxime concentration and residence time in the brain after oral absorption.

Figure 7.: Blood-brain barrier (BBB) permeability plot. ADME (absorption, distribution, metabolism and excretion) properties of the studied compounds were correlated to calculated lipophilicity (Alog P98) and polar surface area (PSA 2D) (Table S1). Standard oximes known for efficacy in reactivation of OP-inhibited AChE are 2-PAM, HI-6, obidoxime and TMB-4. The area inside magenta (BBB-95) and red (Absorption-95) ellipses represents compounds with good BBB permeability and human intestinal absorption [24]

A plot of six physicochemical properties of the studied oximes (Supplementary material, Figure S2) was compared to the recommended values for CNS-active drugs. Recommendations include lower molecular weight (MW < 450), moderate hydrophobicity (log P < 5), fewer hydrogen bond donors and acceptors (HBD < 3 and HBA < 7), fewer rotatable bonds (RB < 8), and are less polar (polar surface area PSA < 0 .70 nm2 (70 Å2) [22]. All oximes exhibited optimal values for molecular weight, hydrophobicity, and hydrogen-bond donor and acceptor atom counts. However, 50 compounds (45 %) exceeded the recommended number of rotatable bonds, and 68 compounds (62 %) surpassed the recommended polar surface area threshold (< 0.70 nm2). It is generally accepted that a molecule may still exhibit CNS activity if it violates one recommendation criterion. However, it was shown that 47 compounds in the studied library (41 %) do not meet two of the recommendation criteria (Table S1, Figure S2).