Joint effects between cyanogenic toxicants, aldehydes
Before determining the joint effects between cyanogenic toxicants and aldehydes, the toxicity (EC50) of individual chemicals to Photobacterium phosphoreum was evaluated as -log EC50, as shown in Table 1. Based on these data, the joint effects (TU) between cyanogenic toxicants and aldehydes were determined, and the results are listed in Table 2. TU in the present study denotes the joint effects of equitoxic binary mixtures at median inhibition.
Table 1. Experimental results of the toxicity tests for single chemicals.
|Classification||No.||Single chemical||-log (EC50)a|
|Cyanogenic toxicants||4*||Allyl cyanide||2.06|
|Aldehyde toxicants||15**||p-methyl benzaldehyde||3.82|
Table 2. Results of joint effects between cyanogenic toxicants and aldehydes
|No.||Cyanogenic toxicants||Aldehyde toxicants||TUa||Ccyanogenic toxicant||Oaldehyde toxicant||Oaldehyde toxicant —Ccyanogenic toxicant|
Table 2 shows that different joint effects (additive, synergistic, or antagonistic) occur for different mixtures, which indicates that joint effects may result from the intracellular chemical reactions between cyanogenic toxicants and aldehydes.
QSAR model for joint effects
Initial QSAR model. Combining Equations 5 and 6, we obtain Equation 8. Then the relationships between TU and the gap (Oaldehyde toxicant−Ccyanogenic toxicant) (the gap between Oaldehyde toxicant and Ccyanogenic toxicant) can be shown in Figure 2.
Figure 2. The relationship between TU (the sum of toxic units) and the gap between Oaldehyde toxicant and Ccyanogenic toxicant. () denotes mixtures containing glycolonitrile and various aldehydes, (▴) α-hydroxy-isobutyronitrile and aldehydes, (▪) malononitrile and aldehydes, (★) allyl cyanide and aldehydes, (▾) 3-hydroxypropionitrile and aldehydes, (♦) acetonitrile and aldehydes. [Color figure can be seen in the online version of this article, available at wileyonlinelibrary.com]
Download figure to PowerPoint
Figure 2 shows, for each cyanogenic toxicant, a good relationship between TU and the gap (Oaldehyde toxicant −Ccyanogenic toxicant). The mixtures can be divided into four different sets according to the relationship. Group I and Group II are mixtures containing glycolonitrile and α-hydroxyisobutyronitrile, whose corresponding hydrolyzates are formic acid and acetone. The reactivity of some aldehyde toxicants (such as formaldehyde and acetaldehyde) is higher than that for these hydrolyzates (formic acid and acetone). Therefore, in the corresponding mixtures (an aldehyde toxicant of formaldehyde or acetaldehyde and a cyanogenic toxicant of glycolonitrile or α-hydroxy-isobutyronitril), the cyanogenic toxicants easily hydrolyze and release CN−, allowing chemical reactions to proceed and synergetic effects to occur (nos. 18–21, 26–28 in Table 2). For other aldehyde toxicants (such as p-methylbenzaldehyde, p-hydroxy-benzaldehyde, and p-dimethylaminobenzaldehyde), only additive or antagonistic effects occur because their reactivity with CN− is less than that of cyanogenic hydrolyzates (nos. 23–25, 37–40 in Table 2).
Group III are mixtures containing allyl cyanide, 3-hydroxypropionitrile, and acetonitrile, and only additive effects occur for these mixtures because the reactivity of aldehyde toxicants is less than their hydrolyzates, leading to a hindrance of the intracellular chemical reactions (nos. 41–76 in Table 2).
Group IV are mixtures containing malononitrile. The group is located at a special position because its two-step hydrolysis yields intermediates cyanoaldehyde and formic acid.
To further reveal the relationship between the joint effects and the gaps of reactivity between cyanogenic hydrolyzates and aldehyde toxicants, these toxicants are listed in Figure 3 based on their reactivity.
Figure 3. The relationship between the joint effects and the gaps of reactivity between cyanogenic hydrolyzates and aldehyde toxicants. [Color figure can be seen in the online version of this article, available at wileyonlinelibrary.com]
Download figure to PowerPoint
Figure 3 shows that malononitrile is located at a special position. This is because malononitrile has two hydrolyzates besides CN−: the intermediate cyanoacetaldehyde and the final formic acid. Considering the initial hydrolyzate (cyanoacetaldehyde), malononitrile should be classified into group III. Considering the other hydrolyzate (formic acid), malononitrile should be classified into group I. However, because of both hydrolyzates, malononitrile can be listed in neither group (group III or group I). We listed it as group IV. Therefore, the gap between malononitrile and formaldehyde (formaldehyde is the most active aldehyde in these aldehyde toxicants) is defined as a criterion to classify these mixtures, that is, Oformaldehyde −Cmalononitrile = −0.125. Then an initial QSAR model can be proposed as follows
The joint effects for group III can be described using the first formula in Equation 9 (n = 36). For groups I, II, and IV, the second formula is available (n = 40). To further reveal the relationship between TU and the atomic charges in groups I, II, and IV, Equation 10 is obtained by combining Equations 5, 6, and 7.
The relationship between TU and (Oaldehyde toxicant + 0.189 × Ccyanogenic toxicant + 0.300) can be developed using regression analysis of data from groups I, II, and IV, and the results are shown in both Figure 4 and Equation 11.
Figure 4. The correlation between TU (the sum of toxic units) and (Oaldehyde toxicant + 0.189 × Ccyanogenic toxicant + 0.300).
Download figure to PowerPoint
The value of r (0.332) shows a poor correlation. Figure 4 indicates that the outlier of mixtures containing malononitrile causes this poor correlation. The outlier is attributable to the difference of hydrolysis between malononitrile and other cyanogenic toxicants. Malononitrile hydrolysis is divided into two steps and yields two CN− and two hydrolyzates, whereas other cyanogenic toxicants only yield one CN− and one hydrolyzate. The deletion of the mixtures containing malononitrile from Equation 11 yields Equation 12.
where n = 23, r = 0.857, SE = 0.209, F = 57.9, p < 0.001, q2Loo = 0.660
The significant correlation (r = 0.857) indicates the developed model is successful in describing the relationship between the joint effects and the atomic charges of mixtures containing cyanogenic toxicants and aldehydes. Therefore, a QSAR model for the joint effects according to Oaldehyde toxicant and Ccyanogenic toxicant can be proposed as follows:
where n = 23, r = 0.857, SE = 0.209, F = 57.9, p < 0.001, q2Loo = 0.660.
Intervention analysis of the outliers
The QSAR model (Eqn. 13) has two outliers. One is the mixture containing allyl cyanide and p-dimethylaminobenzaldehyde (DMAB); the other is mixtures containing malononitrile.
The hydrolyzate of allyl cyanide is acrolein. Acrolein is the smallest unsaturated aldehyde, and its reactivity is stronger than formaldehyde 37. Based on the deduced criterion in the present study, only when the reactivity of the aldehyde toxicant is higher than that of the hydrolyzate of the cyanogenic toxicant, chemical reaction thus proceeds and results in joint effects. Obviously, the occurrence of chemical reaction between allyl cyanide and DMAB is impossible. Consequently, this mixture yields an additive effect. This conclusion is in agreement with our experimental data (no. 53 in Table 2).
However, Figure 3 shows that DMAB is the last in the homologous series of aldehyde toxicants and possesses far less reactivity than any other aldehyde toxicants because the value of Oaldehyde of DMAB is the least of all aldehyde toxicants (−0.309, no. 53 in Table 2). Furthermore, the hydrolyzate of allyl cyanide is acrolein. As the smallest unsaturated aldehyde, the reactivity of acrolein is stronger than that of formaldehyde. The gap of allyl cyanide and DMAB (ODMAB −Callyl cyanide) is so large that the mixture containing the two special chemicals does not obey this criterion, and the mixture is listed as an outlier.
Mixtures containing malononitrile are outliers of the prediction model (Eqn. 13) because malononitrile has two −CN, whereas other cyanogenic toxicants have only one. Malononitrile has two hydrolyzates as well as two Ohydrolyzate. The average Ohydrolyzate (the oxygen charge of the intermediate cyanoacetaldehyde in the first hydrolysis step and that of formic acid in the second hydrolysis step) was used to obtain Equation 13; thus, the unknown validity of Ohydrolyzate leads to the malononitrile outlier. If the Ohydrolyzate-validity (Ohydrolyzate-validity represents the validated value of Ohydrolyzate) can be found, mixtures containing malononitrile will cease to be outliers. Therefore, the relationships between Ohydrolyzate-validity and r were studied, and the results are shown in Figure 5.
In conclusion, a QSAR model can be successfully proposed using the adjusted Ccyanogenic toxicant (0.0529) of malononitrile derived from based on the formula ()
where n = 40, r = 0.887, variance inflation factor = 1.001, SE = 0.195, F = 140, p < 0.001, q2Loo = 0.748.
The r value of the QSAR model is 0.887, indicating a good goodness-of-fit. The variance inflation factor is 1.001, indicating no self-correlation between variables. The q2Loo of the QSAR is as high as 0.748, implying a good robustness of the model. The difference between r2 and q2Loo (r2 − q2Loo = 0.038) does not exceed 0.3, indicating no overfitting in the model. F = 140 and p < 0.001 indicate that the results are statistically significant at the significance level. Furthermore, the result of external validation demonstrates that the QSAR model has good predictive capability (see the result in the Supplemental Data, Table S2 and Fig. S3).
Discussion of the QSAR model
The QSAR model (Eqn. 14) provides a mechanism interpretation for the mixture toxicity between cyanogenic toxicants and aldehydes. It can be shown from the model that the charge of the carbon atom closest to −CN (Ccyanogenic toxicant) can be employed to describe the reactivity of cyanogenic toxicant in mixtures. This indicates that the charge of the carbon atom determines the capability of the cyanogenic hydrolysis and then the hydrolysis determines the contribution of the cyanogenic toxicants in mixture toxicity. For aldehyde toxicants, the charge of oxygen atom in −CHO (Oaldehyde toxicant) can be employed to describe the contribution of aldehydes in mixtures toxicity. That is, the charge determines the reactivity of aldehydes with cyanogenic toxicants. The Ccyanogenic toxicant and Oaldehyde toxicant determine the reaction between cyanogenic toxicants and aldehydes and then can be used to determine their joint effects.
Because the two atomic charges (Ccyanogenic toxicant and Oaldehyde toxicant) determine the reaction and joint effects between individual toxicants (cyanogenic toxicants and aldehydes), the atomic charge–based QSAR model should be a general model. In our previous studies, several QSAR models were developed using the Hammett constant (σp) to describe the contribution of aldehydes and the charge of the carbon atom in the carbon chain of cyanogenic toxicants (C*) to describe the contribution of cyanogenic toxicants (Table 3).
Table 3. The comparison of this model with our previous modelsa
|1b||TU = 0.788–0.882σp||8||Malononitrile and aromatic aldehydes||Aliphatic aldehydes|
|2b||TU = 0.0015–4.876σp||6||Malononitrile and aliphatic aldehydes||Aromatic aldehydes|
|3c||TU = 0.978–0.720σp||9||α-Hydroxyisobutyronitrile and aromatic aldehydes||Aliphatic aldehydes|
|4c||TU = 0.316–4.386σp||6||α-Hydroxyisobutyronitrile and aliphatic aldehydes||Aromatic aldehydes|
|5b||TU = 0.0824–0.237C||6||Cyanogenic toxicants and acetaldehyde||Malononitrile|
|6c||TU = -0.161–7.721C||7||Cyanogenic toxicants and p-nitrobenzaldehyde|| |
|7d||TU = 0.367–0.811σp-6.704C||40||Cyanogenic toxicants and aromatic aldehydes||Aliphatic aldehydes|
|8||The model in the present study||75||Cyanogenic toxicants and aldehydese|| |
These previous models can only be applied to mixtures containing individual cyanogenic toxicant and various aldehydes (nos. 1–4 in Table 3), or mixtures containing various cyanogenic toxicants and individual aldehyde (nos. 5–6 in Table 3). Another model, no. 7 in Table 3 (or Eqn. 1), can be applied to various cyanogenic toxicants and various aromatic aldehydes, but the model cannot be applied to aliphatic aldehydes because the employed descriptor (σp) of aromatic aldehydes is inherently different from that of aliphatic aldehydes.