Investigating plant uptake of organic contaminants through transpiration stream concentration factor and neural network models

Uptake of seven organic contaminants including bisphenol A, estriol, 2,4-dinitrotoluene, N,N-diethyl-meta-toluamide (DEET), carbamazepine, acetaminophen, and lincomycin by tomato (Solanum lycopersicum L.), corn (Zea mays L.), and wheat (Triticum aestivum L.) was measured. The plants were grown in a growth chamber under recommended conditions and dosed by these chemicals for 19 days. The plant samples (stem transpiration stream) and solution in the exposure media were taken to measure transpiration stream concentration factor (TSCF). The plant samples were analyzed by a freeze-thaw centrifugation technique followed by high performance liquid chromatography-tandem mass spectrometry detection. Measured average TSCF values were used to test a neural network (NN) model previously developed for predicting plant uptake based on physicochemical properties. The results indicated that moderately hydrophobic compounds including carbamazepine and lincomycin have average TSCF values of 0.43 and 0.79, respectively. The average uptake of DEET, estriol, acetaminophen, and bisphenol A was also measured as 0.34, 0.29, 0.22, and 0.1, respectively. The 2,4-dinitrotoluene was not detected in the stem transpiration stream and it was shown to degrade in the root zone. Based on these results together with plant physiology measurements, we concluded that physicochemical properties of the chemicals did predict uptake, however, the role of other factors should be considered in the prediction of TSCF. While NN model could predict TSCF based on physicochemical properties with acceptable accuracies (mean squared error less than 0.25), the results for 2,4-dinitrotoluene and other compounds confirm the needs for considering other parameters related to both chemicals (stability) and plant species (role of lipids, lignin, and cellulose).