3.1.4 Building the Gene
Expression Atlas
The alignment of different expression patterns to the same refer-
ence can be used to generate a gene expression atlas, which consists
of a matrix of determined three-dimensional points (voxels), each
containing binary (e.g., expressed or not expressed) information
for the expression of every included gene.
The script “ProSPr_6dpf_SuperVoxelPixCount.R” in [5] can
be used to generate the matrix which will be referred to instep 2of
Subheading3.3.3.3.2 Obtaining Single-
Cell RNAseq Data
- Prepare your single cell suspension (seeNote 5).
- Follow the Fluidigm company instructions for the chip
priming, loading of the cells and on-chip lysis, reverse tran-
scription and PCR. We recommend adding ERCC spike-in
molecules to the Clontech lysis mix. - Harvest the cDNA from the Fluidigm IFC and follow the
Illumina instructions for the DNA fragmentation, barcoding
and paired-end sequencing on Illumina platform. - Map the raw sequence data against reference transcriptome
usingbowtie2[6]. - Obtain the expression counts for each gene by running
HTseq1[7].
3.3 Spatial Mapping
to the Reference
Transcriptome
In this section, we will match the scRNAseq data and the ISH atlas,
to find the probable locations of the sequenced cells. We have used
this method to identify the probable locations of cells sequenced
from the 48 h post fertilizationPlatynereis dumeriliilarvae, demon-
strating that the spatial mapping successfully identified the locations
ofthe originalcellsin80%ofcases[8].Ourspatialmappingapproach
has several advantages: (a) it does not require a priori labeling or
known spatial landmarks, (b) using the specificity index (Subheading
3.3.4) rather than direct read count as a gene expression measure
effectivelyremoves technicalnoiseproblems and (c)a transcriptome-
wide correspondence score (Subheading3.3.5 ) provides an addi-
tional buffer for noise, as it means that the mapping cannot be driven
by a single landmark gene with a spatial expression profile that differs
from all other genes expressed in the cell of interest.
To start this protocol, we need, on the one hand, the spatially
referenced gene expression data from the ISH-based atlas (Sub-
heading3.1), and, on the other hand, the gene expression matrix
obtained by the scRNAseq method of choice (Subheading3.2),
and the spatial mappingRscripts (Subheading2.2).3.3.1 Quality Filtering
of RNA-Seq Data
- For quality filtering of the data, we removed the cells (a) where
more than 10% of the total read counts accounted for the spike-
in sequences, and (b) where less than 1200 unique transcripts
were detected. Depending on the cell capture platform, addi-
tional quality filters may need to be developed in order to
remove samples that contain multiple cells (Note 6).
Spatial Transcriptomics 115