CoChat

Evolution of cellular morpho-phenotypes in cancer metastasis.

Open access

The DOI matches its Crossref record; 57 of the 57 references it cites were found in a registry or an index.

Abstract

Intratumoral heterogeneity greatly complicates the study of molecular mechanisms driving cancer progression and our ability to predict patient outcomes. Here we have developed an automated high-throughput cell-imaging platform (htCIP) that allows us to extract high-content information about individual cells, including cell morphology, molecular content and local cell density at single-cell resolution. We further develop a comprehensive visually-aided morpho-phenotyping recognition (VAMPIRE) tool to analyze irregular cellular and nuclear shapes in both 2D and 3D microenvironments. VAMPIRE analysis of ~39,000 cells from 13 previously sequenced patient-derived pancreatic cancer samples indicate that metastasized cells present significantly lower heterogeneity than primary tumor cells. We found the same morphological signature for metastasis for a cohort of 10 breast cancer cell lines. We further decipher the relative contributions to heterogeneity from cell cycle, cell-cell contact, cell stochasticity and heritable morphological variations.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1038/srep18437

Cites

57 of 57 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 High-throughput cell imaging platform (htCIP) and morphology of PDAC cells. (A) Cell samples were seeded on 24-well glass plates and then fixed and stained. Images were acquired on an automated-stage epifluorescence microscope using a standardized scanning grid. Each fluorescence channel was subsequently processed using custom software and data was then extracted for analysis. (B) Sixteen randomly chosen, horizontally aligned cell and nuclear traces from each patient-derived pancreatic cancer cell line shown here for qualitative visual comparison. (C) Three-dimensional scatter plot showing the wide spectrum of conventional morphological descriptors, cell size, shape factor (SF) and cell aspect ratio, for the pancreatic cancer cell lines used in this study (n = 11 samples and 39,000 individual cells). (D) This panel shows that even for highly similar values of cell size, SF and cell aspect ratio represented by range of red spot in panel (C), the corresponding cells can still display a wide range of shapes not captured by these conventional morphological descriptors.
Figure 1 High-throughput cell imaging platform (htCIP) and morphology of PDAC cells. (A) Cell samples were seeded on 24-well glass plates and then fixed and stained. Images were acquired on an automated-stage epifluorescence microscope using a standardized scanning grid. Each fluorescence channel was subsequently processed using custom software and data was then extracted for analysis. (B) Sixteen randomly chosen, horizontally aligned cell and nuclear traces from each patient-derived pancreatic cancer cell line shown here for qualitative visual comparison. (C) Three-dimensional scatter plot showing the wide spectrum of conventional morphological descriptors, cell size, shape factor (SF) and cell aspect ratio, for the pancreatic cancer cell lines used in this study (n = 11 samples and 39,000 individual cells). (D) This panel shows that even for highly similar values of cell size, SF and cell aspect ratio represented by range of red spot in panel (C), the corresponding cells can still display a wide range of shapes not captured by these conventional morphological descriptors.CC BY · PubMed Central
Figure 2 Visually-aided morpho-phenotyping recognition (VAMPIRE) analysis. (A) Demonstration of VAMPIRE analytical processes. (B–D) Heat maps show the probability of cells in each nucleus and cell shape modes (P(NSk) or P(CSk)) for each different sample (B,C). Heat maps show the population distributions of different paired nuclear and cellular shapes modes (P(NSk&CSj)) for different samples (D). Color coding (blue to red) corresponds to low and high occurrence. (E) A panel of five heterogeneity properties, including CV of nucleus (CV(RN)) and cell size (CV(RC)), entropy of nucleus (S(NS)) and cell shape (S(CS)), and entropy of paired nucleus-cell shape (S(NS&CS)) was used to represent overall heterogeneity profiles of cell morphology of 9 PT and LM samples. The magnitude of heterogeneity of these cell populations is shown in a heat map where color from red to green indicates increasing degree of heterogeneity. (F) Images of PAC01 cells and PAC09 cells. Nuclei and F-actin are labeled in blue and green.
Figure 2 Visually-aided morpho-phenotyping recognition (VAMPIRE) analysis. (A) Demonstration of VAMPIRE analytical processes. (B–D) Heat maps show the probability of cells in each nucleus and cell shape modes (P(NSk) or P(CSk)) for each different sample (B,C). Heat maps show the population distributions of different paired nuclear and cellular shapes modes (P(NSk&CSj)) for different samples (D). Color coding (blue to red) corresponds to low and high occurrence. (E) A panel of five heterogeneity properties, including CV of nucleus (CV(RN)) and cell size (CV(RC)), entropy of nucleus (S(NS)) and cell shape (S(CS)), and entropy of paired nucleus-cell shape (S(NS&CS)) was used to represent overall heterogeneity profiles of cell morphology of 9 PT and LM samples. The magnitude of heterogeneity of these cell populations is shown in a heat map where color from red to green indicates increasing degree of heterogeneity. (F) Images of PAC01 cells and PAC09 cells. Nuclei and F-actin are labeled in blue and green.CC BY · PubMed Central
Figure 3 Cellular heterogeneity in 2-D and 3-D environments. (A) Strategy to analyze cell morphology in 3D collagen matrices. Z-stack images are obtained through sequential imaging at different z positions. A low-pass filter is then applied to individual images followed by maximum z-projection. Cell morphology in projected image is then obtained and subjected to VAMPIRE analysis. (B) Images of PAC01 cells and PAC09 cells in 3-D collagen matrices after z-projection. Nuclei and F-actin are labeled in blue and green. (C,D) Nucleus shapes modes and cell shape modes are identified by VAMPIRE analysis for cells in 3-D matrices. Histograms show nuclear and cellular shape mode distributions for PAC01, PAC03, PAC06 and PAC09 cells. More than 100 cells were analyzed for each sample. (E) Cell morphological heterogeneity properties in both 2D and 3D environments. Strong positive correlation is observed for CV of nucleus size (Spearman’s correlation coefficient ρ=0.8), CV of cell size (ρ=1.0), shape entropy of nucleus (ρ=0.6) and shape entropy of cell (ρ=0.4).
Figure 3 Cellular heterogeneity in 2-D and 3-D environments. (A) Strategy to analyze cell morphology in 3D collagen matrices. Z-stack images are obtained through sequential imaging at different z positions. A low-pass filter is then applied to individual images followed by maximum z-projection. Cell morphology in projected image is then obtained and subjected to VAMPIRE analysis. (B) Images of PAC01 cells and PAC09 cells in 3-D collagen matrices after z-projection. Nuclei and F-actin are labeled in blue and green. (C,D) Nucleus shapes modes and cell shape modes are identified by VAMPIRE analysis for cells in 3-D matrices. Histograms show nuclear and cellular shape mode distributions for PAC01, PAC03, PAC06 and PAC09 cells. More than 100 cells were analyzed for each sample. (E) Cell morphological heterogeneity properties in both 2D and 3D environments. Strong positive correlation is observed for CV of nucleus size (Spearman’s correlation coefficient ρ=0.8), CV of cell size (ρ=1.0), shape entropy of nucleus (ρ=0.6) and shape entropy of cell (ρ=0.4).CC BY · PubMed Central
Figure 4 Effects of cell cycle and local cell density on cell heterogeneity. (A,B) htCIP provides versatile single-cell measurements including cell DNA content (A) and local cell density (ρD) (B), for cells on 2D substrates. (C,D) Relation between cell morphology and progression of cell cycle (C) and increase in local cell density (D) for PAC01, PAC03, PAC06 and PAC09. Cells were sorted based on their DNA content and divided into 9 groups with equal sample size. Probability of cells found in different nucleus and cell shape modes (P(NSk), P(CSk)) at these groups are shown in heat maps. The averaged DNA contents, normalized nucleus size and normalized cell size for these groups are shown in the plot next to the heat maps. Nucleus and cell size are normalized by dividing the lowest value among all groups. The same procedures are used to show effect of local cell density on cell morphology. (E,F) Nucleus and cell shape modes, filled with colors to show correlation between DNA content and occurrence of individual shape modes (E). Color coding from red to green corresponds to Pearson’s correlation coefficient from −1 to 1. The same procedure is applied to show correlation between local cell density and occurrence of shape modes (F,G) Overall heterogeneity profiles of cell morphology of PAC samples in the G0/G1, S and G2/M phases. (H) Overall heterogeneity profiles of cell morphology of PAC samples under three local density conditions, including singlet (ρD = 0) semi-crowded (0 < ρD < 4), and crowded (ρD ≥ 4). The heterogeneity of these samples is shown in a heat map where colors from red to green indicate an increasing degree of heterogeneity.
Figure 4 Effects of cell cycle and local cell density on cell heterogeneity. (A,B) htCIP provides versatile single-cell measurements including cell DNA content (A) and local cell density (ρD) (B), for cells on 2D substrates. (C,D) Relation between cell morphology and progression of cell cycle (C) and increase in local cell density (D) for PAC01, PAC03, PAC06 and PAC09. Cells were sorted based on their DNA content and divided into 9 groups with equal sample size. Probability of cells found in different nucleus and cell shape modes (P(NSk), P(CSk)) at these groups are shown in heat maps. The averaged DNA contents, normalized nucleus size and normalized cell size for these groups are shown in the plot next to the heat maps. Nucleus and cell size are normalized by dividing the lowest value among all groups. The same procedures are used to show effect of local cell density on cell morphology. (E,F) Nucleus and cell shape modes, filled with colors to show correlation between DNA content and occurrence of individual shape modes (E). Color coding from red to green corresponds to Pearson’s correlation coefficient from −1 to 1. The same procedure is applied to show correlation between local cell density and occurrence of shape modes (F,G) Overall heterogeneity profiles of cell morphology of PAC samples in the G0/G1, S and G2/M phases. (H) Overall heterogeneity profiles of cell morphology of PAC samples under three local density conditions, including singlet (ρD = 0) semi-crowded (0 < ρD < 4), and crowded (ρD ≥ 4). The heterogeneity of these samples is shown in a heat map where colors from red to green indicate an increasing degree of heterogeneity.CC BY · PubMed Central
Figure 5 Predictive signature for metastatic pancreatic cancer cells. (A) A 3-D scatter plot shows delineation between different types of samples using CV of nuclear size distribution among cell clusters, CV of cell size distribution among singlet cells and entropy of nucleus shape among single cells for each sample. Three distinct subgroups, respectively composed of PT, LM and NM samples, are readily observed. Two additional patient-derived pancreatic cancer cell lines, a primary tumor cell line (PAC21) and a lung metastases cell line (PAC20), were introduced as validation samples and these two parameters were measured. The location of these two parameters from the new primary tumor derived cell line overlayed well with the PT cluster previously obtained with the training set. The cell line derived from lung metastatic region co-clusters with LM cluster previously obtained with the training set. (B) Relationship of repeating somatic mutations between PT and LM is represented using a Circos plot that is generated using custom software in R. No distinct somatic mutation signature is identified for cells derived from the metastatic site (Interactive Circos plot can be viewed in supplementary materials).
Figure 5 Predictive signature for metastatic pancreatic cancer cells. (A) A 3-D scatter plot shows delineation between different types of samples using CV of nuclear size distribution among cell clusters, CV of cell size distribution among singlet cells and entropy of nucleus shape among single cells for each sample. Three distinct subgroups, respectively composed of PT, LM and NM samples, are readily observed. Two additional patient-derived pancreatic cancer cell lines, a primary tumor cell line (PAC21) and a lung metastases cell line (PAC20), were introduced as validation samples and these two parameters were measured. The location of these two parameters from the new primary tumor derived cell line overlayed well with the PT cluster previously obtained with the training set. The cell line derived from lung metastatic region co-clusters with LM cluster previously obtained with the training set. (B) Relationship of repeating somatic mutations between PT and LM is represented using a Circos plot that is generated using custom software in R. No distinct somatic mutation signature is identified for cells derived from the metastatic site (Interactive Circos plot can be viewed in supplementary materials).CC BY · PubMed Central
Figure 6 Hierarchy of cell heterogeneity. (A) Images of PAC01 and PAC09 cells after four-day growth from a sparse initial seeding density and show cells forms several spatially and morphologically distinct progenies. Detailed view of cell images in highlighted areas are shown in the bottom. (B) Paired correlation analysis of nucleus size and cell size for PAC01 and PAC09 cells. Elevated correlation for cells in proximity was found in both all population and cells in G0/G1 phase. After randomly permutation label cells, this correlation disappears. (C) Bar graphs show the average variances of nucleus size and cell size among different pancreatic cancer cells, including PAC01, PAC02, PAC03, PAC04, PAC06, PAC07, PAC08 and PAC09. Variance of cells depends on the underlying cellular conditions including singlet, crowding, in the G0/G1 phase, in the G2/M phase. Averaged nucleus size and cell size variation within a clone (<σp2>) for all cells and for cells in the G0/G1 phase is also shown. The variances are scaled by the variance among all populations. Great decrease in variations in both nucleus size and cell size were found for clonal cells in the G0/G1 phase. The P value is calculated using one-way ANOVA. (D) A plot illustrates that observed cell heterogeneity are combination of different effects including cell cycle, cell-cell contact, cell stochasticity and heritable cell variation. A proposed mathematical model to describe cellular heterogeneity is shown in the bottom. (E) CV of averaged nucleus size and averaged cell size among different progenies is used to measure heritable variation for both LM (PAC01~PAC04) and PT (PAC06 ~ PAC09). LM display in average lower heritable variation in both nucleus size (P < 0.05) and cell size (P > 0.05). The P value is calculated based on two-sample t-test.
Figure 6 Hierarchy of cell heterogeneity. (A) Images of PAC01 and PAC09 cells after four-day growth from a sparse initial seeding density and show cells forms several spatially and morphologically distinct progenies. Detailed view of cell images in highlighted areas are shown in the bottom. (B) Paired correlation analysis of nucleus size and cell size for PAC01 and PAC09 cells. Elevated correlation for cells in proximity was found in both all population and cells in G0/G1 phase. After randomly permutation label cells, this correlation disappears. (C) Bar graphs show the average variances of nucleus size and cell size among different pancreatic cancer cells, including PAC01, PAC02, PAC03, PAC04, PAC06, PAC07, PAC08 and PAC09. Variance of cells depends on the underlying cellular conditions including singlet, crowding, in the G0/G1 phase, in the G2/M phase. Averaged nucleus size and cell size variation within a clone (<σp2>) for all cells and for cells in the G0/G1 phase is also shown. The variances are scaled by the variance among all populations. Great decrease in variations in both nucleus size and cell size were found for clonal cells in the G0/G1 phase. The P value is calculated using one-way ANOVA. (D) A plot illustrates that observed cell heterogeneity are combination of different effects including cell cycle, cell-cell contact, cell stochasticity and heritable cell variation. A proposed mathematical model to describe cellular heterogeneity is shown in the bottom. (E) CV of averaged nucleus size and averaged cell size among different progenies is used to measure heritable variation for both LM (PAC01~PAC04) and PT (PAC06 ~ PAC09). LM display in average lower heritable variation in both nucleus size (P < 0.05) and cell size (P > 0.05). The P value is calculated based on two-sample t-test.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 57 references

Research works printed but not linked: 1

  1. Wirtz, D., Konstantopoulos, K. & Searson, P. C. The physics of cancer: the role of physical interactions and mechanical forces in metastasis. Nat Rev Cancer 11, 512–522 (2011). · DOI · Paper page

Cited by

Items 101–115 of 115

Metadata

Authors
Wu, P.-H., Phillip, J. M., Khatau, S. B., Chen, W.-C., Stirman, J., Rosseel, S., Tschudi, K., Van Patten, J., Wong, M., Gupta, S., Baras, A. S., Leek, J. T., Maitra, A., & Wirtz, D.
Affiliations
  • Johns Hopkins University (US)
  • The University of Texas MD Anderson Cancer Center (US)
Corresponding author
  • Wu, Pei-Hsun
  • Wirtz, Denis
Venue
Scientific Reports
Published
2015-12-17
Volume
5
Issue
1
Pages
18437-18437
ISSN
2045-2322 (ISSN-L)
OpenAlex
W2198136410
Cited by
117
References
57
Citations by year
2026: 6 · 2025: 15 · 2024: 9 · 2023: 14 · 2022: 16 · 2021: 19 · 2020: 17 · 2019: 7 · 2018: 5 · 2017: 6 · 2016: 3
FWCI
2.82
Type
journal article
Language
en
Open access
Licence
Locations
  • Scientific Reports (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
Funders
  • American Heart Association (12POST12050638)
  • National Institutes of Health (UL1TR001079)
  • National Cancer Institute (U54CA143868, R01CA174388)
  • National Center for Advancing Translational Sciences
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Biophysics › Cell Image Analysis Techniques
  • Health Sciences › Medicine › Oncology › Cancer Cells and Metastasis
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Cell Biology › Cellular Mechanics and Interactions
Keywords
metastasis, intratumoral heterogeneity, cell morphology, vampires, nuclear shape, pancreatic cancer, breast cancer, cell cycle, cell-cell contact, randomness, single-cell analysis
MeSH terms
Adult, Aged, Aged, 80 and over, Breast Neoplasms, Cell Count, Female, Humans, Male, Microscopy, Fluorescence, Middle Aged, Neoplasm Metastasis, Neoplasms, Pancreatic Neoplasms, Reproducibility of Results, Computational Biology, Cell Line, Tumor, Cell Shape, Young Adult, Single-Cell Analysis, Tumor Microenvironment

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.