3 Methods
The methods described in this section cover, first, the principles and
steps of generating gene expression atlases; second, the protocol for
obtaining quality filtered and normalized scRNAseq count data;
and third, the bioinformatic approach for the matching the scRNA-
seq and ISH data together in its original spatial context.3.1 Generation of a
Gene Expression Atlas
The protocol described in this section makes use of high-
throughput microscopy and different image analysis steps to find,
for each gene, averaged expression patterns.
Gene expression atlases make use of corresponding landmarks
between different images to align them using image registration.
Alongside, the spatial profiles of different genes can be fitted to a
common reference, allowing coexpression analysis in silico. This
principle has been employed to build expression maps for highly
stereotypic tissues achieving cellular resolution [10], extended to
full-body atlases using universal markers [4] and more recently,
combined with image analysis to achieve a general method for
building single-cell resolution expression atlases for complex body
plans. We have used this protocol to generate a whole-body atlas for
the three-segmented wormPlatynereis dumerilii[5].3.1.1 High-Throughput
Microscopy
In order to build an ISH-based gene expression atlas, we need raw
microscopy data that consists of a series of stacks (virtual slices) of
the samples (e.g., 3D tissue or animal), each containing at least two
channels:- The first channel (DAPI) is common to all the stacks, and will
provide landmark information to calculate the spatial transfor-
mation (registration). This channel is referred to as the target
channel during the registration process. - The second channel (ISH signal) contains the specific gene-
expression information that will be registered, guided by the
target channel. - Image the samples including at least the target and gene expres-
sion channels. The acquisition resolution should be equal in x,
y, and z axes to generate isotropic stacks that facilitate the
subsequent 3D processing (seeNote 4). - For each genetic marker, depending on the desired final reso-
lution of the atlas, a minimum number of samples need to be
imaged (seeNote 4).
3.1.2 Generating
a Reference
In this step, we build a reference template, which will be used to
register all the individual samples to the same three-dimensional
framework. This reference ideally constitutes an average represen-
tation of the target channel (DAPI) and can be built by iterativeSpatial Transcriptomics 113