Trace Detection of Tetrahydrocannabinol in Body Fluid via Surface-Enhanced Raman Scattering and Principal Component Analysis

Tetrahydrocannabinol (THC) is the main active component in marijuana and the rapid detection of THC in human body fluid plays a critical role in forensic analysis and public health. Surface-enhanced Raman scattering (SERS) sensing has been increasingly used to detect illicit drugs; however, only limited SERS sensing results of THC in methanol solution have been reported, while its presence in body fluids, such as saliva or plasma, has yet to be investigated. In this article, we demonstrate the trace detection of THC in human plasma and saliva solution using a SERS-active substrate formed by in situ growth of silver nanoparticles (Ag NPs) on diatom frustules. THC at extremely low concentration of 1 pM in plasma and purified saliva solutions were adequately distinguished with good reproducibility. The SERS peak at 1603 cm-1 with standard deviation of 3.4 cm-1 was used for the evaluation of THC concentration in a methanol solution. Our SERS measurement also shows that this signature peak experiences a noticeable wavenumber shift and a slightly wider variation in the plasma and saliva solution. Additionally, we observed that THC in plasma or saliva samples produces a strong SERS peak at 1621 cm-1 due to the stretching mode of O-C═O, which is related to the metabolic change of THC structures in body fluid. To conduct a quantitative analysis, principal component analysis (PCA) was applied to analyze the SERS spectra of 1 pM THC in methanol solution, plasma, and purified saliva samples. The maximum variability of the first three principal components was achieved at 71%, which clearly denotes the impact of different biological background signals. Similarly, the SERS spectra of THC in raw saliva solution under various metabolic times were studied using PCA and 98% of the variability is accounted for in the first three principal components. The clear separation of samples measured at different THC resident times can provide time-dependent information on the THC metabolic process in body fluids. A linear regression model was used to estimate the metabolic rate of THC in raw saliva and the predicted metabolic time in the testing data set matched well with the training data set. In summary, the hybrid plasmonic-biosilica SERS substrate can achieve ultrasensitive, near-quantitative detection of trace levels of THC in complex body fluids, which can potentially transform forensic sensing techniques to detect marijuana abuse.