磷酸氯喹胁迫污泥与餐厨垃圾厌氧共消化机制及 SHAP 分析

This study systematically investigated the effects of concentration gradients (0 – 25 mg/L) of the active pharmaceutical ingredient chloroquine phosphate (CQ) on anaerobic digestion performance and microbial response mechanisms, with a focus on its environmental residue risk. As CQ is an antimalarial and antiviral drug, its global consumption has increased in recent years, leading to its more frequent detection in wastewater, sewage sludge, and food waste. While the persistence and bioaccumulation potential of CQ are well documented, its effects on anaerobic co-digestion systems and the underlying microbial regulatory mechanisms remain unclear. Existing studies have primarily focused on conventional antibiotics, resulting in a knowledge gap regarding emerging pharmaceuticals such as CQ. To address this gap, sequencing batch reactors were established using sewage sludge and food waste as substrates, with the inoculum acclimated at (37.5 ± 0.5) °C. Six CQ concentrations (0, 2.5, 5.0, 10.0, 15.0, and 25.0 mg/L) were applied across two phases: Phase I (0 – 38 d) to evaluate immediate effects, and Phase II (39 – 75 d) to assess system recovery following CQ stress removal. The results demonstrated a concentration-dependent dual effect of CQ on methane production. Low concentrations (≤ 5 mg/L) enhanced methane yield (up to 5.2%) and organic matter removal, whereas high concentrations (≥ 15 mg/L) significantly inhibited methane production (by 25% – 34%) and substrate utilization. After CQ removal (Phase Ⅱ), the system function fully recovered, with cumulative methane yield reaching 251.45 mL/g COD and SCOD removal stabilizing at 85.06% ± 2.79%, indicating strong microbial metabolic resilience. Volatile fatty acid (VFA) analysis revealed significant changes in acetic and propionic acid profiles under CQ stress, with high CQ concentrations suppressing their production via inhibition of hydrolytic microbial communities. Extracellular polymeric substance (EPS) secretion increased at low CQ concentrations (e.g., TB-EPS reached 39.76 mg/g VS at 5 mg/L) but declined at higher concentrations, exhibiting a defense–collapse pattern. Enzyme activity assays showed that antioxidant enzymes (SOD) were activated to counteract CQ-induced reactive oxygen species (ROS), while coenzyme F₄₂₀ peaked at 10 mg/L, suggesting a shift toward acetoclastic methanogenesis. Lactate dehydrogenase (LDH) activity was stimulated at intermediate CQ concentrations (5 – 10 mg/L) but was insufficient to reverse metabolic dysfunction at higher doses. Mechanistic analysis further indicated that CQ stress altered dissolved organic matter (DOM) composition and reduced the humification index (HIX), reflecting impaired organic matter stabilization. Given the complexity of nonlinear interactions among multiple parameters, interpretable machine learning was employed to identify key drivers. SHAP analysis revealed that dissolved chemical oxygen demand (SCOD) was the dominant factor influencing methane yield among 17 input variables, with the highest mean absolute SHAP value, far exceeding those of TCOD and CQ concentration. These findings confirm that CQ-induced organic matter accumulation is the primary cause of methane yield suppression. Spearman correlation analysis further supported these results, showing a significant positive correlation between CQ concentration and SCOD/TCOD, and a significant negative correlation with HIX. Overall, these findings indicate that CQ suppresses methane production primarily through an indirect inhibition mechanism—impairing substrate utilization and promoting SCOD accumulation—rather than through direct toxicity, highlighting the value of interpretable machine learning for assessing pharmaceutical residue risks in anaerobic digestion systems.