In the previous post we introduced Smart-3SEQ, a new method just published in Genome Research that reduces the cost and preparation time for RNA-seq libraries and is also sensitive to single-cell inputs and degraded samples. Those are large degrees of improvement from existing methods, but Smart-3SEQ also enables a qualitatively new kind of study: gene-expression profiling on microdissected cells from clinical samples of formalin-fixed, paraffin-embedded (FFPE) tissue.
Since before anyone cared about RNA integrity, research hospitals around the world have been amassing millions of FFPE tissue blocks from biopsies and necropsies. These samples can be sectioned onto slides and stained for histology, but the DNA and especially the RNA are heavily degraded into small fragments, limiting FFPE tissues’s usefuless for sequencing. However, Smart-3SEQ’s robustness to both small samples and degraded samples unlocks the potential of FFPE tissue for histology-focused RNA-seq. Using laser-capture microdissection (LCM), we identified tumor cells by eye and physically cut them out of a single-cell-thick section of FFPE tissue from a case of ductal carcinoma in situ. Those tiny tissue samples became RNA-seq libraries, and while “bulk” samples of a few hundred tumor cells were easily distinguished from control macrophages, we also took LCM to its limits and cut out single cells. Those showed an interesting trend: half of the cells cut out of the tumor actually appeared to be macrophages rather than tumor cells.
Laser-capture microdissection of ductal carcinoma in situ.
A: Single cell within a duct involved by DCIS, targeted for dissection (green outline). B: Duct post-dissection and (inset) the captured cell on the LCM cap.
Gene-expression profiling on bulk and single-cell samples from FFPE tissue dissected by LCM
C: Expression (transcripts per million) of known marker genes for macrophage (CD68, CD163) and DCIS (EPCAM, KRT7, KRT18, ERBB2 (HER2)). Single cells from the DCIS tumor that lack the ERBB2 amplification are circled; we infer that these cells are intraductal macrophages. D: t-SNE analysis of all genes; same plotting scheme as C.
This is a view of tumor heterogeneity we would have missed if we only had the bulk tissue. High-throughput single-cell RNA-seq is a hot new technique, but it relies on reverse logic: homogenize all the cells, then categorize the cell types after sequencing. In contrast, LCM and Smart-3SEQ let you go in a more intuitive direction: first identify the cells of interest on a slide, even if they’re in small or irregularly shaped areas, then cut them out separately and prepare separate sequencing libraries from them, so you can look for differential gene expression between cells you’ve already categorized by histology. LCM and Smart-3SEQ may open up vast archives of old fixed tissue for new kinds of studies.
Conceptual diagram of different RNA-seq approaches.
Reverse RNA-seq: Cells are disaggregated from tissue, destroying information about histological context and organization. After single-cell RNA-seq, expression profiles are used to retroactively infer categories of cells. Forward RNA-seq: Cells are dissected in situ according to their histology, so these a priori classes can inform differential gene-expression analysis.