A Single-Organelle Optical Omics Platform for Cell Science and Biomarker Discovery

Abstract

Research in fundamental cell biology and pathology could be revolutionized by developing the capacity for quantitative molecular analysis of subcellular structures. To that end, we introduce the Ramanomics platform, based on confocal Raman microspectrometry coupled to a biomolecular component analysis algorithm, which together enable us to molecularly profile single organelles in a live-cell environment. This emerging omics approach categorizes the entire molecular makeup of a sample into about a dozen of general classes and subclasses of biomolecules and quantifies their amounts in submicrometer volumes. A major contribution of our study is an attempt to bridge Raman spectrometry with big-data analysis in order to identify complex patterns of biomolecules in a single cellular organelle and leverage discovery of disease biomarkers. Our data reveal significant variations in organellar composition between different cell lines. We also demonstrate the merits of Ramanomics for identifying diseased cells by using prostate cancer as an example. We report large-scale molecular transformations in the mitochondria, Golgi apparatus, and endoplasmic reticulum that accompany the development of prostate cancer. Based on these findings, we propose that Ramanomics datasets in distinct organelles constitute signatures of cellular metabolism in healthy and diseased states.

Figure 1:. Quantification of major biomolecular classes in single organelles by Raman-BCA.

(A) Raman spectra were acquired in organelles located by fluorescent probes; the endoplasmic reticulum (ER) is shown as an example. The spectral acquisition site is shown by the red dot, inside a punctuated white outline. (B) A preprocessed Raman spectrum contains contributions from all biomolecules present at the spectral acquisition site. (C) The BCA algorithm selectively identifies the contributions of different types of biomolecules to the acquired spectrum in (B). (D) The concentrations of major biomolecule groups in ER were derived by BCA and arranged into an R-dataset. The table shows concentrations of proteins (Prot), DNA, RNA, glycogen (Gly), lipids (Lip), saturated phospholipids (L-St), unsaturated phospholipids (L-USt), phosphatidylcholine sphingomyelin (PC), cholesterol and cholesteryl esters (CL), and triglycerides (TG) in mg/ml. In addition, the lipid unsaturation parameter (LSU), average number of double bonds per unsaturated phospholipid (LSU-n) and trans/cis ratio (TCP) were determined. All resolvable molecular parameters are collectively referred to as the R-dataset. (E) Single-cell R-datasets are then compiled into a cross-omics R-matrix characterizing a cellular population or a cell line.

Figure 2:. Ramanomics parameters measured in endoplasmic reticulum (ER) and Golgi apparatus (GA) of HeLa, BMEC, and U251/R132 cells.

(A,D) Biomolecular profiles measured in ER and GA, respectively. The R value shows concentrations of proteins (Prot.), DNA, RNA, glycogen (Gly), lipids (Lip.), phosphatidylcholine (PC), and cholesterol (CL), in mg/ml. The lipid unsaturation parameter (LSU) and ratio of trans isomers to the total number of unsaturated phospholipids (TCP) are shown as well. Brackets indicate a statistically significant difference (p < 0.05). (B,C) Scatterplots represent the LSU and PC concentration vs the total lipid weights in ER for the studied cell lines. (E,F) Scatterplots represent PC, LSU vs the total lipid weights in GA for the studied cell lines. For A-E the values were determined from the n=61 (HeLa), n=23 (BMEC) and n=36 (U251/R132). For D-F n=60 (HeLa), n= 29 (BMEC) and n= 52 (U251/R132).

Figure 3:. Heatmap representation and hierarchical clustering analysis for the single-organelle datasets in HPN and HPC cells.

The Ramanomics data are split by organelle (GA, ER, mitochondria, LD) and cell line (HPN vs HPC). Within a cohort, the samples are clustered according to their similarity along the Raman variables, and the cohorts are clustered according to their similarity. Color and intensity of the boxes represents absolute values for Raman parameters according to the color key. Number of measurements: n= 64 (HPC GA), n= 73 (HPN GA); n= 72 (HPC ER), n=71 (HPN ER); n= 81 (HPC mito), n=54 (HPN mito); n=8 (HPC LD) n=25 (HPN LD).

Figure 4:. Raman molecular parameters measured in murine prostate epithelium and TRAMP tumors.

Results of one-way ANOVA (p < 0.01) for R-matrices obtained in normal wild-type prostate (W.T.), prostate malignant tumors (Tm), and tumors at terminal castrate-resistant prostate cancer (CRPC) stage. (A) Spectra were obtained at random sites in cytoplasm. (B) Spectra were acquired in lipid droplets (LD). For (A) the values were determined from n= 65 (WT), n= 85 (Tm) and n= 74 (CRPC). For (B) n= 24 (WT), n=37 (Tm) and n=30 (CRPC).