Identification of the Cellular Tipping Point in the Inflammation Model of LPS-Induced RAW264.7 Macrophages Through Raman Spectroscopy and the Dynamical Network Biomarker Theory

Abstract

Raman spectroscopy is a non-destructive spectroscopic technique that provides complex molecular information. It is used to examine the physiological and pathological responses of living cells, such as differentiation, malignancy, and inflammation. The responses of two cellular states, initial and full-blown inflammation, have mainly been investigated using a comparative analysis with Raman spectra. However, the tipping point of the inflammatory state transition remains unclear. Therefore, the present study attempted to identify the tipping point of inflammation using a cell model. We stimulated RAW264.7 mouse macrophages with lipopolysaccharide (LPS) and continuously collected Raman spectra every 2 h for 24 h from the initial and full-blown inflammation states. A Partial Least Squares analysis and Principal Component Analysis-Linear Discriminant Analysis predicted the tipping point as 14 h after the LPS stimulation. In addition, a Dynamical Network Biomarker (DNB) analysis, identifying the tipping point of a state transition in various phenomena, indicated that the tipping point was 14 h and identified tryptophan as a biomarker. The results of a multivariate analysis and DNB analysis show the cellular tipping point.

Keywords: DNB theory; Raman spectroscopy; inflammation model; tipping point; tryptophan.

Figure 1

Raman spectra of LPS stimulation and non-stimulation groups: (A) the LPS stimulation group; (B) the non-stimulation group; (C) the subtracted Raman spectra of (A,B).

Figure 2

PLS analysis of Raman spectra −1 h and 24 h after the LPS stimulation. (A): PLS score plot of Factor-1 and Factor-2; (B): loading plot of Factor-1 and Factor-2; (C): box plots of 2 to 22 h predictions between −1 h and 24 h using the PLS model. Boxes represent the interquartile range (IQR). Whiskers extended 1.5× the IQR. Circle plots indicate outliers, while red cross plots represent averages. Orange horizontal lines within boxes show the median.

Figure 3

PCA-LDA analysis of discriminant results from 2 to 22 h after the stimulation using −1 h and 24 h models. (A): Results of model creation for −1 and 24 h; (B): results of adaptation of the discriminant model from 2 to 22 h; (C): number of PCs used and model error rates; (D): score plots for model application results for each hour.

Figure 4

Results of the DNB analysis of the LPS stimulation: (A): a dendrogram of highly fluctuating Raman shifts; (B): (left) DNB scores, (center) the average standard deviation, and (right) the average correlation strength; (C): a heat map of correlations among the DNB group (cyan-colored part).

Figure 5

Time series of the relative intensity of the tryptophan band. (A): LPS-stimulation groups; (B): non-LPS stimulation groups. Boxes represent the IQR. Whiskers extended 1.5× the IQR. Circle plots indicate outliers, while red cross plots represent averages. Orange horizontal lines within the boxes show the median.

Figure 6

The Raw264.7 cell culture and Raman measurement timeline. Thirteen samples each from the LPS-stimulated and non-stimulated groups were prepared and passaged 24 h prior to the start of the stimulation. Raman measurements were taken 1 h before the LPS stimulation and then at 2-h intervals for 24 h after the stimulation, for a total of 13 measurements.