Toward Gene-Correlated Spatially Resolved Metabolomics with Fingerprint Coherent Raman Imaging

Abstract

Raman spectroscopy has long been known to provide sufficient information to discriminate distinct cell phenotypes. Underlying this discriminating capability is that Raman spectra provide an overall readout of the metabolic profiles that change with transcriptomic activity. Robustly associating Raman spectral changes with the regulation of specific signaling pathways may be possible, but the spectral signals of interest may be weak and vary somewhat among individuals. Establishing a Raman-to-transcriptome mapping will thus require tightly controlled and easily manipulated biological systems and high-throughput spectral acquisition. We attempt to meet these requirements using broadband coherent anti-Stokes Raman scattering (BCARS) microscopy to spatio-spectrally map the C. elegans hermaphrodite gonad in vivo at subcellular resolution. The C. elegans hermaphrodite gonad is an ideal model system with a sequential, continuous process of highly regulated spatiotemporal cellular events. We demonstrate that the BCARS spatio-spectral signatures correlate with gene expression profiles in the gonad, evincing that BCARS has potential as a spatially resolved omics surrogate.

Figure 1

BCARS imaging of C. elegans gonad and SPADE analysis. (a) Schematic drawing of a C. elegans hermaphrodite reproductive system. (b) 2925 cm^–1 BCARS image of two wild-type adult worms. Each pixel spectrum was acquired in 4 ms. The yellow line indicates the outline of the upper worm (“Up”). Scale bar: 20 μm. (c) SPADE tree representation of fingerprint BCARS spectra from the outlined area. Pixels 1–5 shown in (b) belong to the indicated nodes on the SPADE tree. (d) BCARS spectra corresponding to the 5 pixels, with their identities labeled in the legend. (e) Spatially averaged difference spectra of the 10 gonad sections of Up indicated in S1, with the mean spectrum of the entire Up gonad being subtracted. (f, g) Difference spectra corresponding to each node in branches A and B, respectively, labeled in (c). The mean spectrum of the full Up worm is subtracted. Note that the difference spectra are similar within a branch, but completely different across branches.

Figure 2

Clusters of nodes and their spatial distribution in the 2925 cm^–1 BCARS image and the corresponding BCARS spectra. (a–d) Different clusters of nodes (black squares) and the corresponding spatial distribution (green dots) in the BCARS image. Scale bar: 20 μm. (e) Raman spectra of clusters A–D shown in (a–d), and the assignment of related Raman peaks. d, DNA/RNA or nucleic acids; l, lipid; p, protein.

Figure 3

BCARS SPADE tree plot for gonad sections. The C. elegans gonad arm imaged by BCARS was divided into 10 sections with the same spatial segmentation method described in ref (28) (see the Materials and Methods section for details). For each section, the corresponding SPADE plot is shown.

Figure 4

Comparison between BCARS SPADE results and existing gene expression data obtained by Tzur et al. (a) Correlation of Gonad tree node frequency between each pair of 10 sections of the Up worm gonad. (b) The same process applied to expression of dynamically expressed genes (genes that show 2-fold change in mapped read count) in the hermaphrodite worm gonad. (c) Node frequency data from Figure 3 reordered according to the clustermap generated in Figure S2. The color indicates the z-score standardized node frequency. (d) Same process applied to the dynamically expressed gene data, with the clustermap shown in Figure S3.

Figure 5

Spatial distribution of gene-correlated nodes. Panels a, c, and e are the histograms of rme-1-, plp-1-, and cle-1-correlated nodes, respectively. The 10 nodes with the highest correlation to the respective gene were chosen for the next step. Panels b, d, and f are the spatial distribution of corresponding gene-correlated node pixels (green dots) in the 2925 cm^–1 BCARS image. Scale bar: 20 μm.