PROJECT SUMMARY
ICCS2020-A1
This multi-PI project conducts experiments to study cell state instability in tumor cells, motivated by the theory of
“critical transitions” (CT). CTs are abrupt shifts of behavior of a complex non-linear system and are preceded by
system state destabilization. A cancer cell population represents a statistical ensemble of cells, each of which is
a nonlinear stochastic dynamical system. The latter is embodied by the gene regulatory network (GRN) and cells
are normally in stable attractor states. We hypothesize that cancer cells in small lesions can be poised between
either staying dormant or exiting dormancy (“escape”) and that this binary decision is a CT. This implies that to
be in such a poised state, the cell state has to be destabilized. Thus, detecting cell state instability, manifest in
the cell transcriptomes, can discern if a small tumor is safely in a stable state or poised in the above sense. Many
an observation suggests that cell density of the dormant tumor may be a “bifurcation parameter” that drives GRN
dynamics, via instability toward the CT, at which a cancer cell population can jump to the state of steady growth.
SPECIFIC AIMS. The proposed study is experimental but grounded in theory: Cell state instability is manifest in
an increase of the quantity IC that we derived from theory and requires single-cell (sc) transcriptomes in a popu-
lation to compute (=dynamics of a statistical ensemble of GRNs). Aim 1 (in vitro) uses large ensembles of micro-
cultures (=cancer cell populations) to quantitatively show destabilization and bifurcations of growth behaviors.
Aim 2 (in vivo) reevaluates old mouse tumor models in a new scheme that exposes the binary decision (dor-
mancy vs. “tumor-take”) to test the hypothesis that clinical dormancy escape is preceded by cell state instability.
APPROACH: In Aim 1, using massively-parallel micro-cultures, bulk RNASeq and scRNAseq, we examine hith-
erto undistinguished growth modes of cancer cells and measure bistability as a function of cell density (dormancy
vs. “take-off”). In Aim 2 we examine our intriguing observations in many mouse models: under specific condi-
tions, identified by titrating inoculum cell numbers in creating dormant tumors, some mice exhibit stable dor-
mancy and others a robust tumor-take despite same initial conditions. This finding suggests a poised state and
defines a bistable regime. Tumor models using cells studied in Aim 1 will be evaluated in our scheme to expose
bistable behaviors and Ic computed from scRNAseq data. We anticipate that tumors in unstable dormancy poised
to take-off display higher cell state instability (higher IC) than the stably dormant tumors. But sc-transcriptomes
will also reveal the genes that drive the CT and how they are linked to the risk of impending dormancy escape.
SIGNIFICANCE: While this first-in-its-class study analyzes abstract principles rather than specific molecules, its
potential impact is tangible: It predicts the fate trajectory of indolent tumors in a new way, complementing current
quest for molecular signatures to classify tumors by prognostic groups, by detecting in single-cell resolution cell
population data signs of destabilization that herald an approach to the CT or “tipping point” of dormancy escape.
This work also raises awareness of non-linear behaviors for the design of more relevant animal tumor models.