involved in the EMT regulation. Some regulated only the motility.
Others regulated the invasive process, and still others regulated the
de-adhesion component. No single transcription factor was
involved in all of the EMT behaviors we scored. Some of those
behaviors were most interesting. For example, we found that
FoxN2/3 was involved in acquisition of motility. FoxN2/3 KO
cells remained in the epithelium but continued to remodel the
basement membrane as part of the invasive function. Without
motility, however, the FoxN2/3 KO cells failed to take advantage
of their own remodeled matrix.
With at least a partial understanding of the EMT transcriptional
control, the next goal was to ask what genes were controlled by
each contributing transcription factor. Those downstream effector
genes, we hypothesized, encoded the proteins that initiated and
conducted the EMT mechanics. We knew this strategy would not
yield all proteins involved in the EMT. For example, actin and
myosin are present constitutively in the cell and surely are participants in the motility component, but other proteins, controlled by
the motility transcriptional sub-circuit, we hypothesized, were the
drivers of actin and myosin cytoskeletal contractions. Those proteins were our targets. We decided to first use RNA-seq and did a
temporal profile starting with sampling 2 h before, then during and
after the EMT. We also did the same profile with embryos in which
twist or snail had been knocked down. The database was huge, but
we could eliminate constitutively expressed RNAs since we wanted
to find the genes activated by twist and snail. We were able to
segment that population of genes into clusters that were activated
at the predicted times relative to the EMT, and we could identify
which of those genes were putative targets of Twist or Snail regulation. That allowed us to narrow the search from over 16,000 genes
to fewer than 50, a manageable group to then work with. Upon
further narrowing we arrived at a small number of proteins that we
could functionally confirm as participants in the EMT.
While we were successful in pulling out a few genes involved in
the de-adhesion phase of the EMT, and those studies continue, we
wanted to access proteins in all component processes. A handicap
with the RNA-seq profiles, however, is noise. Even though the
skeletogenic cells went through EMT at a similar time, they were
not perfectly synchronous, and also there were a number of cells in
the database that were not involved in the EMT. With improvements in single cell-sequencing (sc-RNAseq) it is now possible to
interrogate each cell of an embryo. For that reason, we launched a
sc-RNAseq project, the methods of which are described in this
Methods book. It still is too early to fill in the details about that
approach, but it offers a way to identify a substantial fraction of the
RNAs present in a cell at any given time and has the very nice
advantage of being able to computationally project a temporal
10
David R. McClay
Others regulated the invasive process, and still others regulated the
de-adhesion component. No single transcription factor was
involved in all of the EMT behaviors we scored. Some of those
behaviors were most interesting. For example, we found that
FoxN2/3 was involved in acquisition of motility. FoxN2/3 KO
cells remained in the epithelium but continued to remodel the
basement membrane as part of the invasive function. Without
motility, however, the FoxN2/3 KO cells failed to take advantage
of their own remodeled matrix.
With at least a partial understanding of the EMT transcriptional
control, the next goal was to ask what genes were controlled by
each contributing transcription factor. Those downstream effector
genes, we hypothesized, encoded the proteins that initiated and
conducted the EMT mechanics. We knew this strategy would not
yield all proteins involved in the EMT. For example, actin and
myosin are present constitutively in the cell and surely are participants in the motility component, but other proteins, controlled by
the motility transcriptional sub-circuit, we hypothesized, were the
drivers of actin and myosin cytoskeletal contractions. Those proteins were our targets. We decided to first use RNA-seq and did a
temporal profile starting with sampling 2 h before, then during and
after the EMT. We also did the same profile with embryos in which
twist or snail had been knocked down. The database was huge, but
we could eliminate constitutively expressed RNAs since we wanted
to find the genes activated by twist and snail. We were able to
segment that population of genes into clusters that were activated
at the predicted times relative to the EMT, and we could identify
which of those genes were putative targets of Twist or Snail regulation. That allowed us to narrow the search from over 16,000 genes
to fewer than 50, a manageable group to then work with. Upon
further narrowing we arrived at a small number of proteins that we
could functionally confirm as participants in the EMT.
While we were successful in pulling out a few genes involved in
the de-adhesion phase of the EMT, and those studies continue, we
wanted to access proteins in all component processes. A handicap
with the RNA-seq profiles, however, is noise. Even though the
skeletogenic cells went through EMT at a similar time, they were
not perfectly synchronous, and also there were a number of cells in
the database that were not involved in the EMT. With improvements in single cell-sequencing (sc-RNAseq) it is now possible to
interrogate each cell of an embryo. For that reason, we launched a
sc-RNAseq project, the methods of which are described in this
Methods book. It still is too early to fill in the details about that
approach, but it offers a way to identify a substantial fraction of the
RNAs present in a cell at any given time and has the very nice
advantage of being able to computationally project a temporal
10
David R. McClay