Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
[Preprint]. 2024 Jun 25:2024.01.18.576248.
doi: 10.1101/2024.01.18.576248.

Single-cell morphodynamical trajectories enable prediction of gene expression accompanying cell state change

Jeremy Copperman  1 , Ian C Mclean  2 , Sean M Gross  2   3   1 , Jalim Singh  3 , Young Hwan Chang  2   3 , Daniel M Zuckerman  2   3 , Laura M Heiser  2   3   1
Affiliations

Single-cell morphodynamical trajectories enable prediction of gene expression accompanying cell state change

Jeremy Copperman et al. bioRxiv. .

Abstract

Extracellular signals induce changes to molecular programs that modulate multiple cellular phenotypes, including proliferation, motility, and differentiation status. The connection between dynamically adapting phenotypic states and the molecular programs that define them is not well understood. Here we develop data-driven models of single-cell phenotypic responses to extracellular stimuli by linking gene transcription levels to "morphodynamics" - changes in cell morphology and motility observable in time-lapse image data. We adopt a dynamics-first view of cell state by grouping single-cell trajectories into states with shared morphodynamic responses. The single-cell trajectories enable development of a first-of-its-kind computational approach to map live-cell dynamics to snapshot gene transcript levels, which we term MMIST, Molecular and Morphodynamics-Integrated Single-cell Trajectories. The key conceptual advance of MMIST is that cell behavior can be quantified based on dynamically defined states and that extracellular signals alter the overall distribution of cell states by altering rates of switching between states. We find a cell state landscape that is bound by epithelial and mesenchymal endpoints, with distinct sequences of epithelial to mesenchymal transition (EMT) and mesenchymal to epithelial transition (MET) intermediates. The analysis yields predictions for gene expression changes consistent with curated EMT gene sets and provides a prediction of thousands of RNA transcripts through extracellular signal-induced EMT and MET with near-continuous time resolution. The MMIST framework leverages true single-cell dynamical behavior to generate molecular-level omics inferences and is broadly applicable to other biological domains, time-lapse imaging approaches and molecular snapshot data.

PubMed Disclaimer

Conflict of interest statement

Competing Interest The authors declare no competing interests.

Figures

Figure 1:
Figure 1:. MMIST approach to link live-cell imaging to molecular read-outs.
a) Live-cell imaging of MCF10A cells after treatment with a panel of microenvironmental ligands. Nuclei are identified using a convolutional neural network, and single-cells are featurized and tracked through time. b) Single-cell features are concatenated along single-cell trajectories to construct the morphodynamical trajectory embedding. c) Dynamical models learn cell states and cell state change sequences in the morphodynamical landscape. d) Cell state populations are used as a linear decomposition of bulk gene expression measurements to predict the gene expression programs underlying cell state change.
Figure 2:
Figure 2:. Data-driven models define morphodynamical cell states and state transition dynamics.
a.) The dynamical embedding landscape is visualized via UMAP from 200 microstates (dots) constructed from morphodynamical trajectories (trajectory snippet length = 10H), and average flows from each state (gray arrows), colored and labeled by cell state groupings, i.e., numbered cell “states”; large black arrows are guides to the eye for significant macroscopic flow paths. Lower panels show images from first and last frames of representative trajectory snippets (10H trajectory length) from each state with nuclear segmentations (red contours) and associated Voronoi segmentation (yellow contours). b.) Cell state flow (at t=24H) by ligand treatment. c.-f.) Cell morphology, motility, and cell cycle features by morphodynamical cell state. Panels (g) and (h) show violin-plot distributions of single-cell values, (i) shows average behavior with uncertainty based on single-cell variation, and (j) shows modeled cell-cycle phase durations averaged over single-cell behavior.
Figure 3:
Figure 3:. Morphodynamical cell states predict global gene expression patterns.
a.) Validation of model gene expression predictions: measured and model-reconstructed gene expression at 24hrs for every experimental condition, including training set (light gray) and test set conditions. b.) Correlation between measured and model-predicted gene expression (red diamonds), and null estimates using random state populations (gray violin plots). Horizontal lines are the mean, 5th, and 95th percentile of the null distribution.
Figure 4:
Figure 4:
Figure 5:
Figure 5:. Morphodynamical model predicts EGF+TGFB-induced EMT gene expression time evolution.
When predicted RNA levels of morphodynamic states are integrated with Markov model dynamics, an array of dynamical omics predictions can be made, shown here for the EGF+TGFB condition. a) Morphodynamical states, which are numbered 1–12 and color-coded (mesenchymal: green, epithelial: purple). Color labels for the states are consistent throughout figure. b) State probability time evolution, measured (grey dots) and model-derived (black lines), with y-axis limits set for each plot so small changes in state populations are visible. c) Prediction of gene expression over time at 30-minute intervals using morphodynamical state prediction and live-cell imaging measured state probabilities, with rows ordered identically to Figure 4c,and d) summarized to Hallmark gene sets. e) Model-predicted state probability time evolution over 96 hours, trained from live-cell imaging over 48 hours. f) Correlation between measured and model-predicted gene expression at t=48H (red diamond) based on training data from t=24H, relative to null models with random state probabilities (gray distribution). Also shown: correlation between t=24H and t=48H gene expression (black X). g) Correlation between predicted morphodynamical state gene signatures and PAMAF measurements out to 4 days.
See this image and copyright information in PMC

Similar articles

See all similar articles

References

    1. Thurman R. E. et al. The accessible chromatin landscape of the human genome. Nature 489, (2012). - PMC - PubMed
    1. Stergachis A. B. et al. XDevelopmental fate and cellular maturity encoded in human regulatory DNA landscapes. Cell 154, (2013). - PMC - PubMed
    1. Brunner A. et al. Ultra‐high sensitivity mass spectrometry quantifies single‐cell proteome changes upon perturbation. Mol Syst Biol 18, (2022). - PMC - PubMed
    1. Tang F. et al. mRNA-Seq whole-transcriptome analysis of a single cell. Nat Methods 6, (2009). - PubMed
    1. Lareau C. A. et al. Massively parallel single-cell mitochondrial DNA genotyping and chromatin profiling. Nat Biotechnol 39, (2021). - PMC - PubMed

Publication types