High-Resolution Raman Imaging of >300 Patient-Derived Cells from Nine Different Leukemia Subtypes: A Global Clustering Approach

Abstract

Leukemia comprises a diverse group of bone marrow tumors marked by cell proliferation. Current diagnosis involves identifying leukemia subtypes through visual assessment of blood and bone marrow smears, a subjective and time-consuming method. Our study introduces the characterization of different leukemia subtypes using a global clustering approach of Raman hyperspectral maps of cells. We analyzed bone marrow samples from 19 patients, each presenting one of nine distinct leukemia subtypes, by conducting high spatial resolution Raman imaging on 319 cells, generating over 1.3 million spectra in total. An automated preprocessing pipeline followed by a single-step global clustering approach performed over the entire data set identified relevant cellular components (cytoplasm, nucleus, carotenoids, myeloperoxidase (MPO), and hemoglobin (HB)) enabling the unsupervised creation of high-quality pseudostained images at the single-cell level. Furthermore, this approach provided a semiquantitative analysis of cellular component distribution, and multivariate analysis of clustering results revealed the potential of Raman imaging in leukemia research, highlighting both advantages and challenges associated with global clustering.

Figure 1

Representative Raman images of leukemia cells and their associated clusters. (a) Selected representative pseudostained Raman images of the 9 different leukemia subtypes. The colors used for Raman images are associated with different clusters as shown in panels (b)–(g), representing the centroids spectra of each of the 17 clusters and used to virtually stain the Raman images. Panel (b) includes all the 17 centroids; panels (c)–(g) include centroids associated with specific cellular components, superimposed on the remaining centroids reported in light gray for comparison. (b–g) Report Raman shifts of the most intense peaks or those better identifying each cellular component. Raman shifts with the asterisks (*) are those used to automatically select the color intensity of specific clusters.

Figure 2

Pseudostained Raman images of 315 leukemia cells from 9 different leukemia subtypes. All Raman images were automatically preprocessed and produced by a global whole-data set cluster analysis. The colors used for Raman images are associated with different clusters as described by the legend and in Figure 1. For each subtype, cells originating from different patients are eventually separated by a gray dot, following the left-to-right, top-to-bottom direction.

Figure 3

Distribution of clusters for each leukemia subtype. (a) Bubble chart showing the proportion of cellular pixels (bubble size) associated with each cluster. The legend at the top indicates the color and corresponding subcellular component of each cluster (see Figures 1 and 2). (b) Bar plot showing the same distributions, highlighting the relative cluster frequency for each leukemia subtype (see the Supporting Information for more details).

Figure 4

Multivariate analysis of the global-clustering analysis results, using the cluster distributions for each cell as an input. Scatter plot of the first two scores of LDA of the cluster distribution in cells from (a) all nine leukemia subtypes (“AMLs+ALLs”); (b) AML leukemia subtypes (“AMLs”); c) ALL leukemia subtypes (“ALLs”). Kernel density estimate (KDE) plots are reported at the top/right of each scatter plot to better show the marginal distributions of, respectively, CV1 and CV2. Percentage values in parentheses represent the proportions of variance explained by the corresponding CV. The scatter plots of all the CVs are reported in Figures S7 and S8 for “AMLs+ALLs” and “AMLs”; the scalings for the three LDAs are reported in Figure S9. The KDEs relative to AML M6 in panels (a) and (b) have been rescaled for visualization purposes.