Triazophos (TAP) is a widely used pesticide that is easily accumulated in the environment due to its relatively high stability: this accumulation from agricultural runoff results in potential hazards to aquatic ecosystems. Constructed wetlands are generally considered to be an effective technology for treating TAP polluted surface water. However, knowledge about the biological mechanisms of TAP removal is still lacking. This study investigates the responses of a wetland plant (Canna indica), substrate enzymes and microbial communities in bench-scale horizontal subsurface-flow constructed wetlands (HSCWs) loaded with different TAP concentrations (0, 0.1, 0.5 and 5mg·L−1). The results indicate that TAP stimulated the activities of superoxide dismutase (SOD) and peroxidase (POD) in the roots of C. indica. The highest TAP concentrations significantly inhibited photosynthetic activities, as shown by a reduced effective quantum yield of PS II (ΦPS II) and lower electron transport rates (ETR). However, interestingly, the lower TAP loadings exhibited some favorable effects on these two variables, suggesting that C. indica is a suitable species for use in wetlands designed for treatment of low TAP concentrations. Urease and alkaline phosphatase (ALP) in the wetland substrate were activated by TAP. Two-way ANOVA demonstrated that urease activity was influenced by both the TAP concentrations and season, while acidphosphatase (ACP) only responded to seasonal variations. Analysis of high throughput sequencing of 16S rRNA revealed seasonal variations in the microbial community structure of the wetland substrate at the phylum and family levels. In addition, urease activity had a greater correlation with the relative abundance of some functional microbial groups, such as the Bacillaceae family, and the ALP and ACP may be influenced by the plant more than substrate microbial communities.