Nicotinamide adenine dinucleotide (NADH), flavin adenine dinucleotide (FAD), tryptophan (TRY), vitamin B6 (VB6), etc., are domain endogenous fluorophores for monitoring cell metabolism. Specifically, NADH and FAD are found in eukaryotes and prokaryotes and are involved in the aerobic respiration of many important coenzymes. They are also the intermediate product of the mitochondrial electron transport chain, which is involved in cell material and energy metabolism. TRY is an essential amino acid that functions as a substrate in protein biosynthesis and regulates protein metabolism, thereby influencing protein deposition. TRY is also a regulator of endocrine hormone release and a signal molecule that regulates the activity of the tricarboxylic acid cycle and its downstream target proteins to promote protein synthesis. VB6 is a water-soluble vitamin that degrades slowly under alkaline conditions. It participates in the metabolism of amino acids, carbohydrates, and lipids, as well as the synthesis of hemoglobin. Therefore, qualitative and quantitative analysis of the contents of multiple endogenous fluorophore groups in cell metabolism is critical for assessing cell growth state and developing disease prediction models. In vitro cell culture is one of the most direct methods for studying the aforementioned issues as it can study the occurrence and development mechanisms of diseases at the cell level, as well as evaluate the drug resistance and chemotaxis of cells. Cell culture medium is a complex analysis medium. It contains various nutrients (vitamins, inorganic salts, fat, and nucleic acid derivatives), which causes serious interference to the qualitative and quantitative analysis of the fluorophores to be measured, which further becomes the difficulty and key point for the rapid and accurate determination of the content of metabolic fluorophore groups.

To achieve the rapid quantitative analysis of TRY, NADH, FAD, and VB6 in a cell culture medium, this study used a three-dimensional (3D) fluorescence spectrum separation method based on parallel factor analysis. In terms of sample preparation, the deployment and fluorophore concentrations were determined using an orthogonal experiment table with 6 factors, 5 levels, and 25 tests. Each sample’s three-dimensional (3D) excitation-emission contour was scanned using a 3D fluorescence spectrophotometer. Wavelength difference correction, the removal of Rayleigh scattering and Raman scattering peaks, inner filtration effect correction, light intensity standardization, and high leverage point removal were all performed sequentially. The core consistency function was used to determine that the optimal model’s factor was 4. Further, the parallel factor analysis model was established to decompose the 3D fluorescence spectrum dataset into two loading matrices and one score matrix. Two loadings could treat as the emission and excitation spectra of many pure chemical components (fluorophores). Finally, the parallel factor analysis model was validated by split-half analysis. Results showed that the emission and excitation spectra of all four pure fluorophores passed S4T6C3 validation and demonstrated a strong capability in predicting the concentrations of fluorophores.

The optimal excitation/emission wavelengths of TRY, NADH, FAD, and VB6 are 280/350, 350/450, 450/550, and 332/400 nm, as determined by the two loading matrices and one score matrix of the parallel factor analysis model (Fig. 5), which are consistent with the standard spectra in the database. In terms of quantitative analysis, the average recoveries of these four fluorophores are 100.2%±15.5%, 107.4%±37.1%, 100.5%±6.4%, and 99.5%±6.5%. Meanwhile, the root prediction errors (RMSEP) of each fluorophore are 0.124, 43.312, 1.601, and 0.639 μg/mL, and the relative mean deviations are 13.216%, 36.937%, 6.112%, and 6.331%. The limits of detection (LOD) are 0.013, 52.628, 0.003, and 0.012 μg/mL (Table 3). Nonetheless, TRY’s emission spectrum coincides with NADH’s excitation spectrum, and FAD’s excitation spectrum coincides with NADH’s emission spectrum. These two cases indicate that the presence of the fluorescence resonance energy transfer phenomenon limits the model’s quantitative analysis capability for NADH concentration, resulting in a large relative average deviation of NADH (Table 3 and Fig.8).

Because of the similar chemical structure of various fluorophore groups and cell culture medium, the spectra heavily overlap, making it difficult to conduct the qualitative and quantitative analysis of single components directly. To begin, we use Rhodamine B and a calibration standard baffle fluorescence spectrophotometer to solve the problem. Moreover, the water Raman unit (RU350) is used for fluorescence intensity normalization, which increases the dynamic detection range of the instrument, and then the parallel factor analysis algorithm, which inherits the advantage of "second order" or "mathematical separation, " decomposes the entire dataset into four fluorescent components (TRY, NADH, FAD, and VB6), without the complex "chemical separation." The average recoveries of these fluorophores are in the range of (99.5%±6.5%)-(100.5%±6.4%), and the detection limits are in the range of 0.003-0.013 μg/mL, indicating that the method reaches the detection level of trace analysis. This method can meet the demand for the quantitative detection of metabolic fluorescent components in a culture medium, indicating a significant advantage in cell metabolism and immune research, medical diagnosis in vitro, and cell sorting.