bioRxiv2025
Per- and polyfluoroalkyl substances (PFAS) are man-made compounds that bioaccumulate in environments. Current PFAS detection technologies encounter difficulty in detecting trace concentrations and require complex data processing, limiting their on-site applicability. By leveraging biological chemical sensing systems (insect olfaction) we can detect broad ranges of PFAS. Insects advanced combinatorial coding mechanism at the level of olfactory sensory neurons enables highly sensitive and specific odor detection. Here, we harness the locust olfactory system to differentiate several PFAS. In-vivo extracellular neural recordings displayed unique odor-evoked responses for multiple PFAS at environmental concentrations. Using population neuronal response, we classified multiple PFAS with an average 87% accuracy. Machine learning algorithms incorporated separate training and testing datasets, reaching a 61% accuracy. Overall, our study demonstrates the first biological olfaction based broad PFAS detection system. Structured AbstractO_ST_ABSIntroductionC_ST_ABSPer- and polyfluoroalkyl substances (PFAS) pose a significant environmental threat due to their widespread presence in consumer waste and resistance to degradation. These "forever chemicals" persist in various ecosystems and exhibit bio accumulative behavior. Increased human exposure to PFAS has been linked to numerous health issues. Despite their growing relevance, current detection methods often lack the sensitivity and efficiency needed for comprehensive environmental monitoring and are unable to simultaneously detect multiple PFAS at environmental concentrations. RationaleThe locust (Schistocerca americana) possesses a highly developed olfactory system that has been extensively studied and is accessible for physiological recordings across multiple brain regions. Utilizing olfactory receptor neurons and combinatorial coding strategies, locusts can generate distinct neural fingerprints for trillions of odorants across a wide range of concentrations. Through spatiotemporal neural coding at the antennal lobe (AL) neurons, they can detect chemicals at parts per trillion levels, functioning as a biological chemical sensor with exceptional sensitivity and broad specificity. In this study, we aimed to directly harness the locusts olfactory neural circuitry to develop a next-generation, brain-based cyborg sensor capable of (1) simultaneously identifying multiple PFAS and (2) determining their concentration ranges, addressing key limitations of existing detection technologies with the integration of machine learning algorithms. ResultsIn-vivo extracellular neural recordings from the locust AL revealed that individual neurons exhibited distinct response profiles to different PFAS and their varying concentrations, indicating that neuronal activity if modulated by both chemical identity and concentration. By incorporating both spatial and temporal dimensions of neural activity, the recorded neuronal populations produced unique and reproducible response patterns corresponding to specific PFAS and concentration levels. Using this approach, the cyborg sensor demonstrated an overall detection accuracy of 87% across a panel of seven PFAS, with high sensitivity and specificity for both individual analytes and broader chemical groupings. Notably, classification of PFAS concentration ranges down to parts per trillion achieved 84% accuracy, with PFOS concentrations reaching 100% detection rate. A machine learning algorithm trained on high concentration data and tested on low concentration data achieved a 61% accuracy. These results underscore the potential of biologically integrated cyborg sensors for real-time, high-resolution environmental monitoring of several PFAS. ConclusionThis study demonstrates, for the first time, the locust olfactory neural network harnessed as a highly effective cyborg sensor for detecting and classifying various PFAS across concentration ranges. This sensor can accurately distinguish between multiple PFAS and controls with high sensitivity and specificity. Through combinatorial coding and spatiotemporal neural dynamics, the locust neural computation encodes distinct activity patterns in response to PFAS and their concentrations. These neural signatures serve as unique "fingerprints" for individual PFAS and concentrations, enabling precise identification. In-vivo electrophysiological recordings revealed clear, compound-specific differences in neural activity, with high classification accuracy. Real-time and machine learning analysis further addressed key limitations of conventional PFAS sensors. This novel approach represents a significant step toward the development of compact, real-time, brain-based PFAS detection sensor capable of discriminating multiple compounds and concentrations simultaneously.