Quantification of Antiretroviral Drug Emtricitabine in Human Plasma by Surface Enhanced Raman Spectroscopy

Abstract

In this study, reproducible label-free detection and quantification of the antiretroviral drug emtricitabine (FTC) down to 78 ng/mL in human plasma by surface enhanced Raman spectroscopy (SERS) is presented. A novel plasma sample pretreatment method using silver nitrate and silver colloidal nanoparticles (Ag CNPs) was used to prepare the plasma samples for analysis. The pretreated plasma samples were evaporated to dryness on an aluminum surface and a computer-controlled Raman scanning system was used to collect spatially resolved SERS spectra of the entire surface. Calibration curves of commercial human plasma samples containing FTC in a concentration range of 5000 to 78 ng/mL were calculated using three different methods. First, a conventional approach was taken, where all the spectra collected for each concentration were averaged, then the SERS intensity of a known FTC peak (792 cm^-1) was used for calibrations (total population method). This approach was refined by utilizing a figure-of-merit (FOM) quality index (Q _i) to sample spectra from each concentration that contained the highest signal-to-noise (S/N), before averaging and calculating the SERS intensity of the 792 cm^-1 FTC peak (Q _i sample method). Finally, the distribution of all Q _i values for each concentration were modeled using cumulative distribution functions (CDFs) and were used for calibrations (CDF method). The CDF method exhibited the highest analytical sensitivity (slope = 3702.47) compared to the Q _i sample method (slope = 1591.05) and the total population method (slope = 754.21). The Q _i sample method exhibited the highest linearity (R ² = 0.99) compared to the CDF method (R ² = 0.95) and the total population average (R ² = 0.97). The CDF method exhibited the highest S/N in the concentration range of 5000 to 312 ng/mL (S/N range of 31.5-16.6). The Q _i sample method exhibited the highest S/N for concentrations 156 and 78 ng/mL (S/N = 9.7 and 7.4, respectively). These results show that the Q _i sample method is advantageous over all other methods when approaching the LOQ while the CDF method is advantageous over all methods at higher concentrations. The LOQ (78 ng/mL) was confirmed by principal component analysis (PCA). Together these results show that statistical treatment of a large population of SERS spectra, where the analyte signal intensity follows an exponential distribution, is superior to standard methods of averaging populations of spectra in terms of analytical sensitivity, linearity, and S/N. Additionally, it was found that the background signal had no interference with the quantitative data calculated for the total population and Q _i sample methods after repeating both analyses with baseline-subtracted spectra. The results and methodology presented in this study establish a framework for integrating SERS into drug adherence monitoring for FTC-based treatment and prevention of infections by demonstrating consistent SERS detection and quantification of FTC in human plasma at therapeutically relevant concentrations.

Figure 1

Workflow of the plasma sample treatment protocol used in this study.

Figure 2

Calibration curves prepared using the total population method for three replicate experiments. (A–C) Averaged SERS spectra for all concentrations and corresponding SERS intensity calibration curves beneath. Each spectrum shown is an average of 9030 spectra (1806 spectra from each concentration replicate). Each black data point represents the difference in SERS intensities at 792 and 723 cm^–1 for each concentration replicate. Linear regression lines were calculated using the average of all concentration replicates (red data points). The regression line, equation, and correlation coefficient for each replicate experiment are shown in red. Spectra were offset for clarity.

Figure 3

Calibration curves prepared using the Q_i sample method for three replicate experiments. (A–C) Averaged SERS spectra for all concentrations and corresponding SERS intensity calibration curves beneath. Each spectrum shown is an average of 100 spectra (20 spectra from each replicate corresponding to the highest 792 cm^–1Q_i). Each black data point represents the difference in SERS intensities at 792 and 723 cm^–1 for each concentration replicate. Linear regression lines were calculated using the average of all concentration replicates (red data points). The regression line, equation, and correlation coefficient for each replicate experiment are shown in red. Spectra were offset for clarity.

Figure 4

Calibration curves prepared using the CDF method for three replicate experiments. (A–C) Model CDFs of each FTC concentration and corresponding calibration curves beneath. The CDFs were constructed based on the Q_i of the 792 cm^–1 spectral region. A 4th order polynomial was fitted to each CDF in the probability range of 0.6–0.9. The Σ ΔQCDF was calculated for each concentration (see eq 2) and plotted as a function of the logarithm of FTC concentration. The Σ ΔQCDF values of the model CDFs (red data points) were used for linear regression. Black data points represent the Σ ΔQCDF of concentration replicates. Each data point in the calibration curves was increased by the absolute value of the smallest data point, ensuring all values are positive and maintain the same intervals between points. The regression line, equation, and correlation coefficient for each replicate experiment are shown in red.

Figure 5

PCA of the matrix blank (blue) and 78 ng/mL (red) SERS spectra used for calibrations in the (A–C) Q_i sample method and (D–F) total population method. The Python library sklearn was used for PCA. The spectra were preprocessed by first truncating the spectral region (585.48 to 1710.01 cm^–1) followed by applying an improved asymmetrically reweighted penalized least-squares (IarPLS) background correction algorithm^, (see Figure S10). The first two PC scores were plotted against each other, and the explained variance ratios for each PC are shown on corresponding axes. A 95% confidence ellipsoid of each group is shown. A detailed description of the PCA workflow is described in Figure S11.

Figure 6

SERS spectra of aqueous 1250 ng/mL FTC (blue), plasma containing 1250 ng/mL of FTC (black), and nonspiked plasma (red). Relevant peaks from each spectrum are noted by a vertical line with their wavenumber written at the top in the color corresponding to the sample type. Each spectrum shown is an average of 25 spectra, each acquired using 15 mW of laser power and 800 ms integration time.