Rapid and Quantitative Determination of S-Adenosyl-L-Methionine in the Fermentation Process by Surface-Enhanced Raman Scattering

Concentrations of S-Adenosyl-L-Methionine (SAM) in aqueous solution and fermentation liquids were quantitatively determined by surface-enhanced Raman scattering (SERS) and verified by high-pressure liquid chromatography (HPLC). The Ag nanoparticle/silicon nanowire array substrate was fabricated and employed as an active SERS substrate to indirectly measure the SAM concentration. The linear relationship between the integrated intensity of peak centered at ~2920 cm−1 in SERS spectra and the SAM concentration was established, and the limit of detections of SAM concentrations was analyzed to be ~0.1 g/L. The concentration of SAM in real solution could be predicted by the linear relationship and verified by the HPLC detection method. The relative deviations (δ) of the predicted SAM concentration are less than 13% and the correlation coefficient is 0.9998. Rolling-Circle Filter was utilized to subtract fluorescence background and the optimal results were obtained when the radius of the analyzing circle is 650 cm−1.

RCF algorithm: A rolling-circle spectral filter (RCF) method proposed by Mikhailyuk et al was utilized to remove the background of Raman spectra. The principle of this method is that the curvature radii of the Raman peaks and the backgrounds are different. Therefore, a certain radius R which is significantly greater than the Raman linewidth and less than the curvature radius of the background should be determined. Eight radius values were used as the candidates and 650 cm -1 was selected as the optimized R. The input array contains the coordinates of points of the original spectrum. In the beginning, the output array is identical to the input array. A circle is constructed for each point of the input array starting from the first point. The abscissa of the center of the circle coincides with the abscissa of the corresponding point. The spectrum and the circle have at least one common point, while the ordinates of the remaining points of the circle are less than the ordinates in the input array. Thus, the circle lies below the spectrum. The differences between the ordinates in the input array and the ordinates of the upper arc of the circle are compared to the corresponding ordinates of the output array. The minimum of two values becomes the new ordinate of the output array. Then, the procedure is repeated for the next point, so that the output array is modified many times. Thus, a circle rolls under the spectrum and subtracts the fragments of the curve whose radius of curvature is greater than R. At a certain radius of the circle, the background is effectively subtracted, whereas the Raman lines remain virtually unchanged. The key advantage of RCF is that we only use a single parameter (radius) in the RCF process. Besides, we should normalize the spectrum before applying the RCF in next step.
Where X is the spectrum intensity, N is the number of all spectral points. MATLAB 2014 is used to implement and build the algorithm of RCF.
EF algorithm: In order to understand the ability for the Ag/Si nanowire array (Ag/Si NWA) to enhance the Raman signal of SAM, several tests are carried out. Because the Raman signals of the SAM solutions were not observed distinct, the R6G was carried out to detecte the Raman signals. To determine the enhancement factor (EF) of the Ag/Si NWA composite structure, a Si NWA without Ag NPs is used as a reference along with the following standard equation: Where the parameters P and T represented the laser power and the acquisition time, respectively.
In the measurement, we carefully kept the same experimental conditions (laser power 9 mW, acquisition time 15 s, laser spot size 1 μm in diameter) to make the ratios of PRaman/PSERS and TRaman/TSERS equal 1. ISERS and IRaman are the integrated Raman intensities of a specific SERS peak observed for the solutions adsorbed on Ag/Si substrate and Si substrate, respectively; NSERS and NRaman are the numbers of molecules adsorbed on the surface of the corresponding substrates.
Here the Raman band centered at 773 cm -1 was chosen to calculate I values. The conventional Raman spectrum was collected when a 10 μl 10 -3 M R6G loaded on a 1×1cm 2 Si substrate and the integrated peak intensity IRaman was measured to be ~3004.75. NRaman was determined based on the 10 -3 M R6G solution and the illuminated volume (Villu) of our Raman system. For our Raman setup, the illumination focus has a diameter of the laser spot size ~1 μm and the penetration depth of 633 nm laser beam is ~3mm. As a result, Villu is 2.36×10 3 μm 3 and NRaman is 1.42×10 9 . When determining NSERS in the illuminated volume of our Raman setup, we assumed that R6G molecules were absorbed as a monolayer on the surface of Ag/Si substrate. The surface area of one R6G molecule is ~2.0 nm 2 , which was calculated from the geometric area of length (1.37 nm) × width (1.43 nm) of one R6G molecule. As show in figure X, the corresponding integrated peak intensity (ISERS) of the 773cm -1 Raman band was significantly enhanced with increasing R6G concentration from 10 -5 M to 10 -3 M, but the ISERS values of 10 -2 M and 10 -1 M are very close to that of 10 -3 M. Consequently, the surface of the Ag/Si substrate is presumed to be fully surrounded with R6G molecule when its concentration reaches 10 -3 M, thus yielding NSERS is

Fig. S2
The results of RCF processing at different radius (a). The different values between the results of RCF processing and the spectra with the background subtracted manually (b).

Fig. S3
SERS spectra of deionized water and two SAM samples (0.1 g l -1 and 0.01 g l -1 ), and the Raman spectra of SAM samples of 0.1 g 1 -1 .