Increasing quantitation in spatial single-cell metabolomics by using fluorescence as ground truth

Abstract

Imaging mass spectrometry (MS) is becoming increasingly applied for single-cell analyses. Multiple methods for imaging MS-based single-cell metabolomics were proposed, including our recent method SpaceM. An important step in imaging MS-based single-cell metabolomics is the assignment of MS intensities from individual pixels to single cells. In this process, referred to as pixel-cell deconvolution, the MS intensities of regions sampled by the imaging MS laser are assigned to the segmented single cells. The complexity of the contributions from multiple cells and the background, as well as lack of full understanding of how input from molecularly-heterogeneous areas translates into mass spectrometry intensities make the cell-pixel deconvolution a challenging problem. Here, we propose a novel approach to evaluate pixel-cell deconvolution methods by using a molecule detectable both by mass spectrometry and fluorescent microscopy, namely fluorescein diacetate (FDA). FDA is a cell-permeable small molecule that becomes fluorescent after internalisation in the cell and subsequent cleavage of the acetate groups. Intracellular fluorescein can be easily imaged using fluorescence microscopy. Additionally, it is detectable by matrix-assisted laser desorption/ionisation (MALDI) imaging MS. The key idea of our approach is to use the fluorescent levels of fluorescein as the ground truth to evaluate the impact of using various pixel-cell deconvolution methods onto single-cell fluorescein intensities obtained by the SpaceM method. Following this approach, we evaluated multiple pixel-cell deconvolution methods, the 'weighted average' method originally proposed in the SpaceM method as well as the novel 'linear inverse modelling' method. Despite the potential of the latter method in resolving contributions from individual cells, this method was outperformed by the weighted average approach. Using the ground truth approach, we demonstrate the extent of the ion suppression effect which considerably worsens the pixel-cell deconvolution quality. For compensating the ion suppression individually for each analyte, we propose a novel data-driven approach. We show that compensating the ion suppression effect in a single-cell metabolomics dataset of co-cultured HeLa and NIH3T3 cells considerably improved the separation between both cell types. Finally, using the same ground truth, we evaluate the impact of drop-outs in the measurements and discuss the optimal filtering parameters of SpaceM processing steps before pixel-cell deconvolution.

Keywords: SpaceM; fluorescein diacetate (FDA); imaging mass spectrometry (imaging MS); ion suppression; pixel-cell deconvolution; spatial single-cell metabolomics.

FIGURE 1

Evaluating pixel-cell deconvolution in imaging MS-based spatial single-cell metabolomics by using fluorescein. (A) Procedure of SpaceM prior to pixel-cell deconvolution: (1) pre-MALDI microscopy, (2) MALDI-imaging MS acquisition, (3) post-MALDI microscopy, (4) image registration, ablated region selection and cell segmentation. (B) Pixel-cell deconvolution and the confounding factors. Single-cell pie charts illustrate the contributions of different cells or the extracellular area to the ablated regions sampling (i) multiple cells, (ii) a single cell, or (iii) both intra- and extracellular areas. Pixel-cell deconvolution methods aim to estimate the metabolite levels of the single cells depicted in magenta, red, blue and orange. (C) Overlay of the fluorescein fluorescence (magenta), fluorescein MALDI-signals (yellow), laser-ablated regions (grey) and segmented cells (blue). (D) Representative scatter plots (belonging to one replicate) of median fluorescein fluorescence versus assigned fluorescein MALDI-signal of single cells for the tested pixel-cell deconvolution methods: linear inverse modelling (LIM), weighted average (WA). The methods use sampling proportion cut-offs of 0.3, mass spectral intensities are integrated within 4 ppm m/z tolerance and TIC-normalised. (E) Boxplot with density-weighted Spearman correlations between fluorescent and MALDI intensities of fluorescein obtained for each of the pixel-cell deconvolution methods WA, LIM and the mixed model (MIX) for 6 replicates. (F) Illustration of the process of data simulation to evaluate the pixel-cell deconvolution methods. Cell signals are colour-scaled from white to red for cells (ground truth, top panel) and ablated regions (measured intensities, bottom panel). (G) Scatter plots with Spearman correlations of the ground truth versus assigned signals in the simulated data when using the pixel-cell deconvolution methods LIM (blue) or WA (red).

FIGURE 2

Compensating for ion suppression in spatial single-cell metabolomics. (A) Representative scatter plot for ablated regions (belonging to one replicate), plotting the scaled ratio η (fluorescein ion [M-H]^- intensity divided by the fluorescence) against the ablated region sampling proportion; the ratios η are set to 1 for sampling proportions of 1; using TIC-normalised (red) and unnormalized (blue) MS intensities; both axes are log-transformed. (B,C) Boxplots with density-weighted Spearman correlations between fluorescence and MS intensities of fluorescein for either ablated regions (B) or cells (C) using the WA method (sampling proportion of 0.3, m/z tolerance 4 ppm) for 6 replicates, with different normalisation methods including the supervised and unsupervised ion suppression method (ISM). (D) Representative scatter plots of the ratio μ plotted against the sampling proportion for four ions [(C₂₀H₁₂O₅-H)^- corresponding to fluorescein, (C₁₂H₂₂O₁₁-H)^- for a disaccharide, (C₁₄H₂₅NO₁₁-H)^- for a polysaccharide, and (C₂₀H₃₂O₂-H)^- for arachidonic acid]; ratios are set to 1 for sampling proportions of 1; both axes are log-transformed. (E) UMAP visualisation of single-cell metabolomics data obtained from a co-culture of HeLa (blue) and NIH3T3 (red) cells (Rappez et al., 2021). Cells were WA-normalised with (upper panel) or without (bottom panel) applying the unsupervised method for compensating the ion suppression (ISM). (F) Quantification of the intermixing between the cell types from panel E, with barplots showing the mean and standard deviation of the mean fractions of all cell’s 10 nearest-neighbours with the opposite cell type, per normalisation approach.

FIGURE 3

Investigating mass spectrometry drop-outs. (A) Representative scatter plot (belonging to one replicate) showing relations between the fluorescence intensities and MS intensities for fluorescein, measured for the ablated regions. (B) Histogram of the fluorescence intensities for ablated regions with non-zero (blue) and zero (red) MS intensities for the fluorescein ion. (C,D) Boxplots with density-weighted Spearman correlations between fluorescence and MS intensities of fluorescein either for the ablated regions (C) or cells (D) using the WA method (sampling proportion of 0.3, m/z tolerance of 4 ppm) for 6 replicates, using all ablated regions (blue) or only those with non-zero MS intensities for fluorescein (red) from panels (A,B). (E,F) The effect of increasing peak integration m/z tolerance on the fraction of zeros in the input measurements (E) and correlation between fluorescence and MS intensity for fluorescein (F). Pixel-cell deconvolution methods were performed as described for panel (D). Data points indicate the mean values and standard deviation of the 6 replicates, the dashed lines show the 3 ppm (red, default in METASPACE) or 4 ppm tolerance found to be optimal for this experiment (grey).

FIGURE 4

Optimising parameters of the weighted average pixel-cell deconvolution. (A,B) The effect of increasing the sampling proportion cut-off (A) or sampling specificity cut-off (B) on the mean density-weighted Spearman correlation ρ (blue) between fluorescence and MS intensities of fluorescein (deconvolution method WA, m/z tolerance of 4 ppm) as well as the fraction of cells with assigned MS-signals (red). Data points indicate the mean values and standard deviation of the 6 replicates, dashed lines the optimal values. (C,D) Heatmaps of mean weighted Spearman correlation (C) or fraction of assigned cells (D) as function of sampling (x-axis) or specificity (y-axis) proportion cut-offs. White stars indicate the values maximising the mean weighted correlation ρ.