Simultaneous quantification of multiple endogenous biothiols in single living cells by plasmonic Raman probes

Three endogenous biothiols in single cells were simultaneously quantified by plasmonic Raman probes and quantitative principal component analysis (qPCA).

A regular PCA procedure has two main steps that alter the original data: zero-mean step and projection step. For zero-mean step, the column-means are subtracted from A to form a new matrix B whose mean values of each column are all zero. Thus, we have (eq. S5): where ̅ is a row vector whose value of each column is the mean of the same column within A. Please note that when only row size or column size of two matrixes is the same, the minus operation in this equation performs a subtraction that subtracts subtrahend with every columns or rows of the minuend respectively in turn, and the result is a matrix which is of the same size with the minuend. From eq. S3, eq. S4 and eq. S5, we have (eq. S6): Equation above clearly show that the original K still fits eq. 3 after C and M are zeromeant. The projection step calculates PC scores with the new matrix A which is zeromeant, we have (eq. S7): where SA is the matrix of PC scores and L is corresponding set of eigenvectors called loadings. CL (SC) and ML (SM) are PC scores of C and M after projection, respectively.
The brackets are used to form matrices. From eq. S3 we have (eq. S8): manuscript, we can easily have = ′ ′ (S16) = ′ ′ (S17) With auxiliary line AT//CC', the proportion of C within O can be calculated with To calculate a mixture containing n constituents, its PC1 to PCn-1 scores were needed to be mapped in an n-1 dimensional space and the proportion of each constituent could be calculated according to n equations similar to eq. S16 -eq. S20. For a system with n (n≥2) constituents, after spectra acquisition and PCA procedure, we can scratch scores of first n-1 PCs and calculate K with where e is an all-ones column vector with a same number of rows as SC and

×1
indicates a column vector with n rows. Every piece of data processing and calculation in this workflow is programmable, facilitating efficient proportion calculation without any spectra inspection or peak attribution. S-6

Strategy for discriminating and quantifying Cys, Hcy and GSH
For a reaction: + → , its reaction rate (r) could be given by: where k is the second-order reaction rate constant. If one reactant is of large excess over the other (e.g. [B]>>[A]), then eq. 22 could be written as:
Scheme S2 Schematic diagram explaining the workflow.  In our current work, PRPs can react with Cys, Hcy and GSH at the same time obtaining three different products with a wealth of fingerprint information in Raman spectra, but it is difficult to differentiate the one from the other by visual. In order to extract the most principal differences, we analyzed the data using PCA on the Raman spectra of S-17 the three products. Taking into consideration of the intracellular concentration of Cys, Hcy and GSH, we gave the initial concentrations of the three biothiols and then the given ratio of the three biothiols could be got in simulation experiment (Table S2).
Because the given initial concentration ratio of the three biothiols  (Table S2 and Fig. S12A), so we used spectral mixture of the three products to rectify the discrepancy. Based on the given ratio of the three biothiols and the Raman spectra of the three products, we could get a spectral mixture. The calibration curve for three biothiols was derived from a spectral mixture divided by a mixed product (Fig. S12B), so if giving a mixed product, we could calculate a corresponding spectral mixture, utilizing qPCA, thus the initial concentration ratio of the three biothiols could be calculated (Fig. S12C).
S-18  (Table S3 and Fig. S13). Based on the given ratio of the two biothiols and the Raman spectra of the two products, we could get a spectral mixture. The calibration curve for two biothiols was derived from a spectral mixture divided by a mixed product S-19 (Fig. S13B), so if giving a mixed product, we could calculate a corresponding spectral mixture, utilizing qPCA, thus the initial concentration ratio of the two biothiols could be calculated (Fig. S13C).
S-20 To examine the reaction selectivity, PRPs were treated with biologically related amino acids and possible necleophiles (Fig. S17). No observable changes of C≡N were observed with other amino acids and necleophiles (NO2 − , SO3 2− ), but a little change with S 2− and HS − . Considering the reactivity and the intracellular concentrations of Cys (30~200 μM), Hcy (0.1~1 mM) and GSH (1~10 mM) reported in previous literatures, 1-6 the average intracellular H2S level is in sub-micromolar range, 7-9 which is produced in cells through enzymatic, non-enzymatic processes and exogenous added. [7][8][9][10] Thus, in our current system, the sub-micromolar intracellular H2S seems not to have obvious influence on quantifying these three biothiols with either much more higher concentration (GSH) or much more higher reactivity (Cys and Hcy) than H2S.
S-24  (Table S1). Based on the above, we can say that the signal enhancement of Raman spectra is mainly derived from reporter, so the spectra of three product samples were quite similar with each other.