Extracellular antibiotic resistance genes (eARGs) exacerbate the propagation of antimicrobial resistance in the environment and pose great threats to human health. On-site identification and quantification of multiple eARGs at trace level are an urgent need. Herein, we designed a core-satellite platform based on surface enhanced Raman scattering (SERS) to simultaneously detect multiple eARGs (sul1, tetA, and bla_TEM) in the field without PCR amplification. The detection limits of 0.66 (sul1), 0.66 (tetA), and 0.13 aM (bla_TEM) were obtained in one SERS measurement which evidenced its ultra-high sensitivity. The multivariate machine learning model with random forest algorithm exhibited good regressions between the actual and predicted concentrations of eARGs fragments (R_sul1² =0.985, R_tetA² =0.993, R_blaTEM² = 0.994, p < 0.05). The whole process is within 40 min from sample collection to the acquisition of the SERS spectra, realizing an efficient analysis. This SERS strategy successfully quantified the eARGs in wastewater treatment plants, cattle and aquaculture farms, and groundwater, and the results are comparable to droplet digital PCR assays. Our study opens a new avenue for multiple eARGs surveillance on-site.

细胞外抗生素耐药基因（eARGs）加剧了环境中抗生素耐药性的传播，对人类健康构成了巨大威胁。迫切需要在痕量水平上对多种 eARG 进行现场识别和量化。在此，我们设计了一个基于表面增强拉曼散射 (SERS) 的核心卫星平台，无需 PCR 扩增即可在现场同时检测多个 eARG（ sul1、tetA和bla _{TEM ）。}0.66（ sul1）、0.66（tetA）和0.13 aM（bla _TEM ）的检测限) 是在一次 SERS 测量中获得的，这证明了它的超高灵敏度。采用随机森林算法的多元机器学习模型在 eARGs 片段的实际浓度和预测浓度之间表现出良好的回归 ( R _sul1 ² =0.985, R _tetA ² =0.993, R _blaTEM ² = 0.994, p < 0.05)。从样品采集到SERS光谱采集，整个过程在40分钟内，实现了高效的分析。这种 SERS 策略成功地量化了污水处理厂、养牛场和水产养殖场以及地下水中的 eARG，其结果与液滴数字 PCR 测定相当。我们的研究为现场多种 eARG 监测开辟了一条新途径。