Cells sense the context in which they grow to adapt their phenotype and allow multicellular patterning by mechanisms of autocrine and paracrine signalling1,2. However, patterns also form in cell populations exposed to the same signalling molecules and substratum, which often correlate with specific features of the population context of single cells, such as local cell crowding3. Here we reveal a cell-intrinsic molecular mechanism that allows multicellular patterning without requiring specific communication between cells. It acts by sensing the local crowding of a single cell through its ability to spread and activate focal adhesion kinase (FAK, also known as PTK2), resulting in adaptation of genes controlling membrane homeostasis. In cells experiencing low crowding, FAK suppresses transcription of the ABC transporter A1 (ABCA1) by inhibiting FOXO3 and TAL1. Agent-based computational modelling and experimental confirmation identified membrane-based signalling and feedback control as crucial for the emergence of population patterns of ABCA1 expression, which adapts membrane lipid composition to cell crowding and affects multiple signalling activities, including the suppression of ABCA1 expression itself. The simple design of this cell-intrinsic system and its broad impact on the signalling state of mammalian single cells suggests a fundamental role for a tunable membrane lipid composition in collective cell behaviour.
Subscribe to Journal
Get full journal access for 1 year
only $3.90 per issue
All prices are NET prices.
VAT will be added later in the checkout.
Rent or Buy article
Get time limited or full article access on ReadCube.
All prices are NET prices.
Gene Expression Omnibus
Kicheva, A., Cohen, M. & Briscoe, J. Developmental pattern formation: insights from physics and biology. Science 338, 210–212 (2012).
Tabata, T. Genetics of morphogen gradients. Nature Rev. Genet. 2, 620–630 (2001).
Snijder, B. et al. Population context determines cell-to-cell variability in endocytosis and virus infection. Nature 461, 520–523 (2009).
Guan, J. L. & Shalloway, D. Regulation of focal adhesion-associated protein tyrosine kinase by both cellular adhesion and oncogenic transformation. Nature 358, 690–692 (1992).
Schaller, M. D. Cellular functions of FAK kinases: insight into molecular mechanisms and novel functions. J. Cell Sci. 123, 1007–1013 (2010).
Mitra, S. K., Hanson, D. A. & Schlaepfer, D. D. Focal adhesion kinase: in command and control of cell motility. Nature Rev. Mol. Cell Biol. 6, 56–68 (2005).
Puliafito, A. et al. Collective and single cell behavior in epithelial contact inhibition. Proc. Natl Acad. Sci. USA 109, 739–744 (2012).
Piccolo, S., Dupont, S. & Cordenonsi, M. The biology of YAP/TAZ: Hippo signaling and beyond. Physiol. Rev. 94, 1287–1312 (2014).
Tarling, E. J., Vallim, T. Q. D. A., Edwards, P. & a.. Role of ABC transporters in lipid transport and human disease. Trends Endocrinol. Metab. 24, 342–350 (2013).
Lawn, R., Wade, D. & Garvin, M. The Tangier disease gene product ABC1 controls the cellular apolipoprotein-mediated lipid removal pathway. J. Clin. Invest. 104, 25–31 (1999).
Battich, N., Stoeger, T. & Pelkmans, L. Image-based transcriptomics in thousands of single human cells at single-molecule resolution. Nature Methods 10, 1127–1133 (2013).
Costet, P., Luo, Y., Wang, N. & Tall, A. R. Sterol-dependent transactivation of the ABC1 promoter by the liver X receptor/retinoid X receptor. J. Biol. Chem. 275, 28240–28245 (2000).
Palamarchuk, A. et al. Akt phosphorylates Tal1 oncoprotein and inhibits its repressor activity. Cancer Res. 65, 4515–4519 (2005).
Brunet, A. et al. Akt promotes cell survival by phosphorylating and inhibiting a forkhead transcription factor. Cell 96, 857–868 (1999).
Holcombe, M. et al. Modelling complex biological systems using an agent-based approach. Integr. Biol. (Camb) 4, 53–64 (2012).
Zarubica, A. et al. Functional implications of the influence of ABCA1 on lipid microenvironment at the plasma membrane: a biophysical study. FASEB J. 23, 1775–1785 (2009).
Saffman, P. G. & Delbrück, M. Brownian motion in biological membranes. Proc. Natl Acad. Sci. USA 72, 3111–3113 (1975).
Lasserre, R. et al. Raft nanodomains contribute to Akt/PKB plasma membrane recruitment and activation. Nature Chem. Biol. 4, 538–547 (2008).
Landry, Y. D. et al. ATP-binding cassette transporter A1 expression disrupts raft membrane microdomains through its ATPase-related functions. J. Biol. Chem. 281, 36091–36101 (2006).
Shah, M., Patel, K., Fried, V. A. & Sehgal, P. B. Interactions of STAT3 with caveolin-1 and heat shock protein 90 in plasma membrane raft and cytosolic complexes: preservation of cytokine signaling during fever. J. Biol. Chem. 277, 45662–45669 (2002).
del Pozo, M. A. et al. Integrins regulate Rac targeting by internalization of membrane domains. Science 303, 839–842 (2004).
Vlamakis, H., Chai, Y., Beauregard, P., Losick, R. & Kolter, R. Sticking together: building a biofilm the Bacillus subtilis way. Nature Rev. Microbiol. 11, 157–168 (2013).
Friedl, P. & Gilmour, D. Collective cell migration in morphogenesis, regeneration and cancer. Nature Rev. Mol. Cell Biol. 10, 445–457 (2009).
Yvan-Charvet, L. et al. ATP-binding cassette transporters and HDL suppress hematopoietic stem cell proliferation. Science 328, 1689–1693 (2010).
Bensinger, S. J. et al. LXR signaling couples sterol metabolism to proliferation in the acquired immune response. Cell 134, 97–111 (2008).
Lee, B. H. et al. Dysregulation of cholesterol homeostasis in human prostate cancer through loss of ABCA1. Cancer Res. 73, 1211–1218 (2013).
Carpenter, A. E. et al. CellProfiler: image analysis software for identifying and quantifying cell phenotypes. Genome Biol. 7, R100 (2006).
Wippich, F. et al. Dual specificity kinase DYRK3 couples stress granule condensation/dissolution to mTORC1 signaling. Cell 152, 791–805 (2013).
Rämö, P., Sacher, R., Snijder, B., Begemann, B. & Pelkmans, L. CellClassifier: supervised learning of cellular phenotypes. Bioinformatics 25, 3028–3030 (2009).
Gaus, K., Zech, T. & Harder, T. Visualizing membrane microdomains by Laurdan 2-photon microscopy. Mol. Membr. Biol. 23, 41–48 (2006).
Huang, D. W., Sherman, B. T. & Lempicki, R. A. Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nature Protocols 4, 44–57 (2009).
Huang, d. W., Sherman, B. T., Lempicki, R. & a.. Bioinformatics enrichment tools: paths toward the comprehensive functional analysis of large gene lists. Nucleic Acids Res. 37, 1–13 (2009).
Sandelin, A., Alkema, W., Engström, P., Wasserman, W. W. & Lenhard, B. JASPAR: an open-access database for eukaryotic transcription factor binding profiles. Nucleic Acids Res. 32, D91–D94 (2004).
Matyash, V., Liebisch, G., Kurzchalia, T. V., Shevchenko, A. & Schwudke, D. Lipid extraction by methyl-tert-butyl ether for high-throughput lipidomics. J. Lipid Res. 49, 1137–1146 (2008).
Clarke, N. G. & Dawson, R. M. Alkaline O→N-transacylation. 195, 301–306 (1981). A new method for the quantitative deacylation of phospholipids. Biochem. J. 195, 301–306 (1981).
Guan, X. L., Riezman, I., Wenk, M. R. & Riezman, H. Yeast lipid analysis and quantification by mass spectrometry. Methods Enzymol. 470, 369–391 (2010).
Epstein, S. et al. Activation of the unfolded protein response pathway causes ceramide accumulation in yeast and INS-1E insulinoma cells. J. Lipid Res. 53, 412–420 (2012).
Rouser, G., Fleischer, S. & Yamamoto, A. Two dimensional thin layer chromatographic separation of polar lipids and determination of phospholipids by phosphorus analysis of spots. Lipids 5, 494–496 (1970).
We thank B. Snijder for help with single-cell Laurdan quantification, P. Liberali for help with imaging of cell-to-cell variability, Y. Yakimovich for IT infrastructure support, and all members of the laboratory for discussions and support. M.F. was supported by an EMBO and a Marie Curie (301650) fellowship. E.-M.D. was supported by an Oncosuisse fellowship, L.S. was supported by a Bonizzi Theler fellowship. This work is supported by the University of Zurich and the SystemsX.ch RTD Project LipidX.
The authors declare no competing financial interests.
The microarray data set has been uploaded to the NCBI Gene Expression Omnibus as record GSE43873.
Extended data figures and tables
Related to Fig. 1. a, Immunofluorescence against phosphorylated FAK (Y397) in a population of A431 cells, corresponding curve shows single-cell phosphorylated FAK signals against local cell crowding (interquartile area is shown in grey, number of cells >104). b, Gene Ontology enrichment network of genes that are induced by FAK in cells experiencing low crowding. Greyscale indicates enrichment, node-size number of genes, edge width between nodes number of overlapping genes. c, Histogram of ABC transporters more expressed in cells lacking FAK compared to cells expressing FAK when facing low crowding. d, Single-cell transcript counts of Abca1 in 1.2 × 104 FAK-KO and 1.5 × 104 FAK-WT cells experiencing increasing levels of local crowding (interquartile area in grey). e, Control experiment of bDNA single-molecule FISH against bacterial dapB transcripts in FAK-KO or FAK-WT cells experiencing low crowding or high crowding. Representative of 104 cells. f, Real-time PCR measurements of Abca1 transcripts in cells at low and high local crowding in both FAK-expressing and FAK-KO cells in the presence of 10% FCS. Clearly, Abca1 mRNA levels are much higher in FAK-expressing cells facing high crowding than in the same cells facing low crowding (s.d., n = 4 biological replicates each made of 3 technical replicates, P < 10−15, t-test) but also in FAK-KO cells compared FAK-expressing cells (s.d., n = 4 biological replicates each made of 3 technical replicates, P < 10−10, t-test). This indicates that FAK-dependent adaptation of Abca1 transcription to cell crowding also operates in the presence of an abundant and homogeneous amount of growth factors and cytokines in the medium.
Extended Data Figure 2 FAK suppresses ABCA1 expression in cells at low crowding via TAL1 and FOXO3 in a cell-intrinsic way.
Related to Fig. 2. a, Percentage reduction of Abca1 mRNA in FAK-KO cells upon silencing of 19 potential transcription factors. b, Table of primers used for qRT–PCR amplification of Abca1 DNA and corresponding genomic position. c, Western blots of pFAK, pPI(3)K and pAKT levels in FAK-WT and FAK-KO MEFs, and A431 cells at low crowding, high crowding or low crowding + wortmannin. d, Real-time PCR quantification of Abca1 mRNA shows that treatment with LY-294002 alleviates the inhibitory effect of FAK on Abca1 transcription in cells (at low crowding) expressing FAK (s.d., n = 4 biological replicates each made of 3 technical replicates, P < 10−6, t-test), whereas this treatment has no significant effect on Abca1 transcription in cells that lack FAK (s.d., n = 4 biological replicates each made of 3 technical replicates, P > 0.1, t-test). e, Immunofluorescence imaging of ABCA1 over a population of A431 cells in the presence of Y15 FAK inhibitor and related projection of single cell measurements onto nuclear segmentations. f, Quantifications of Abca1 protein expression in FAK-WT cells adhering to micropatterned surfaces of large (10,000 µm2) or small (2,000 µm2) area (http://www.cytoo.com) at long distance from potentially secreting neighbouring cells. This shows that space constraints are sufficient to trigger differences in Abca1 expression (s.d., n = 100 cells, P < 10−4, t-test).
Extended Data Figure 3 Agent-based modelled single cells show characteristics similar to tracked cells.
a, Typical curve of the growth of the nucleus size of a single cell between two mitotic events (centre). Distribution of measured (number of tracks: 650) and agent-based modelled (number of tracks: 200) single-cell nucleus sizes (right histograms) and cell-cycle lengths (bottom histograms). Black, raw data, red, fitted Gaussian curve. Agent-based modelled cells and measured cells show similar distributions in cell-cycle length and nucleus size. b, Curve showing single-cell mean nuclear area against local cell crowding of measured (black, number of cells: >104) and agent-based modelled cells (red, number of cells: >103). c, Histograms of single-cell area distribution of measured (number of cells: >104) and agent-based modelled cells (number of cells: >103) showing that distribution of emerging cell areas of modelled cells are matching those of measured cells even for extreme values. d, Histograms of single-cell mean square displacement distribution of measured (number of tracks: 650) and agent-based modelled cells (number of tracks: 200). e, Timescales of information sensing and processing steps in the FAK–ABCA1 system. Absence of a capacitor does not allow gradual patterns to emerge (switch-like behaviour).
Extended Data Figure 4 Alternative models do not lead to the emergence of gradual patterns in ABCA1 expression, and the full model recapitulates experimentally observed dynamics of reduction in ABCA1 expression in scratch assays.
Conclusions are parameter-independent, for details see mathematical appendix in the Supplementary Information. a, A FAK activation model without autophosphorylation does not result in a pFAK pattern in an agent-based modelled cell population. b, A FAK–ABCA1 model based on free diffusion of signalling molecules without or with c, addition of a putative direct inhibitory effect of ABCA1 on its own suppression does not result in a patterning of ABCA1 expression. d, Introduction of a membrane relay for AKT activation without ABCA1 feedback on the membrane relay does not result in a patterning of ABCA1 expression. e, Simulated single-cell ABCA1 variability over local crowding is similar to the variability seen in our experiments (see Fig. 2d). f, Scratch assays, at which cells at high crowding suddenly become exposed to free space to spread and followed over time, show that reduction of ABCA1 levels in these cells has a half-maximum effect at ∼50 min, and full effect at ∼200 min. g, This is in agreement with simulations of scratch assays using our cell-intrinsic Agent-based model of the FAK–ABCA1 system. The process was iterated thousands of times with random starting levels of ABCA1 similar to the variability seen in the experimental scratch assay. 20 representative curves are shown. In the simulations, it takes ∼150 min for the disappearance of half of ABCA1. h, Distributions of pixel GP values of FAK-KO cells stained with Laurdan at different time-points after treatment with glyburide. After just 20 min of drug treatment, the membranes of these cells become more ordered (P < 10−100, t-test, pixel distributions at each time point are made from 2 × 103 cells).
a, Heat map representing Euclidian distance between modelled and measured levels of pFAK in single cells as a function of local crowding when autophosphorylation constant k1 and removal rate RR varies. Stars represent the values used for further modelling; any pair of k1-RR values with the same low Euclidian distance will lead to the proper pFAK pattern. b–d, Same analysis for k1 and the FAK-independent phosphorylation of FAK rate k2 for a fixed RR value shows that FAK-independent phosphorylation of FAK has no effect on the formation of a pFAK pattern even if k2 is bigger than k1 by several orders of magnitude.
a, Heat map representing the slope of ABCA1 expression against local cell crowding when k3 and 3′ and HL1 and 1′ vary over an extreme range of values for model A. This demonstrates that such topology cannot lead to emergence of gradual expression patterns ABCA1 expression as a function of local cell crowding. b, Mean relative ABCA1 expression in agent-based modelled cells as a function of its inhibition power (Ip) in model B, where ABCA1 would be able to directly inhibit activation of AKT (or PI(3)K). This demonstrates that such direct feedback only leads to switch-like behaviour where ABCA1 is either expressed or not in all cells of the population, independent of local cell crowding. Inhibition power represents the ABCA1 competitive inhibitory power. c, Heat map representing Euclidian distance between modelled and measured levels of ABCA1 in single cells as a function of local crowding when Trsh1 and Trsh2 vary in model C. d, The capacity of model C to generate a gradual expression pattern (low Euclidian distance is black) does not depend on k3 and 3′, and HL1 and 1′, demonstrating the central role of the membrane relay for gradual patterns to emerge.
Extended Data Figure 7 The FAK–ABCA1 system adapts membrane lipid composition, ordering and signalling to local crowding.
Related to Fig. 4. a, Histogram of transcript copy number (number of spots) per cell determined with bDNA single-molecule FISH against endogenous Abca1 in cells at high crowding, or against ABCA1–GFP transcripts in cells at low crowding transfected with the pEGFP-N1-ABCA1 construct. This shows that plasmid-driven ABCA1–GFP expression in cells at low crowding does not exceed that of endogenous Abca1 levels in cells at high crowding. b, Hierarchical clustering of lipid profiles of mouse embryonic fibroblasts grown at high crowding or low crowding conditions and transiently expressing ABCA1 from a plasmid (+ABCA1) or not. The clustergram shows the 48 lipid species that represent 80% of the total lipid amount. Colours correspond to pmol/pmol total lipid z-scored over the four conditions, colours of lipid names refer to their clusters. For complete lipid mass spectrometry data, see Supplementary Table 3. c, Histograms displaying the quantity of free cholesterol in nmol per cell (n = 4 biological replicates, each the mean of 4 technical replicates, s.d.). d, P values related to the bar graphs in Fig. 4c. e, Pie charts representing the percentage of saturated, monounsaturated and polyunsaturated lipids for the four different conditions. f, Fluorescence imaging using Bodipy 493/503 dye of lipid droplets in low crowding (n = 5 × 104 single cells) or high crowding conditions (n = 5 × 104 single cells). This confirms that cells at low crowding contain a larger amount of cholesteryl-esters, which are stored in lipid droplets. g, Diagram summarizing the method to measure membrane ordering of a formaldehyde fixed population of cells at the single-cell level (left flow chart). Distributions of single-cell GP values for groups of cells that are the top 20, 100, 200, 300 ABCA1–GFP expressing cells compared to all cells (top right distributions, n = 500 cells) and curve showing the relationship between single-cell ABCA1 expression and scGP value (bottom right curve, n = 500 cells). h, Image-based quantification of free cholesterol (filipin), GM1 content (cholera toxin B binding or anti-GM1 antibody) and lipid ordering (Laurdan, as in panel d) in single MEFs with (FAK-WT) or without FAK (FAK-KO). n = 4 experiments, each >104 cells. *P values (t-test) < 10−4. i, Because some GM1 may not be accessible in formaldehyde-fixed cells, we performed dot blot analysis of lipid extracts from FAK-KO and FAK-WT cells using HRP-conjugated cholera toxin B. This indicates that FAK-WT cells have higher levels of GM1 than FAK-KO cells. j, pAKT and pPDK1 immunostaining in cells without FAK (FAK-KO) exogenously loaded with GM1 and cholesterol (FAK-KO + GM1 + Chol.), treated with DMSO, or with 10 and 25 μM glyburide in DMSO (n = 3 experiments, each 104 cells, s.d., *P values (t-test) <10−4.
Extended Data Figure 8 Phosphorylation of STAT3 and PAK1/2 are sensitive to ABCA1-mediated membrane perturbation.
a, Curve showing the relationship between ABCA1–GFP expression and phosphorylated STAT3 (T705) and PAK1/2 (T423/T402) amounts in single cells. b, Quantification of immunostaining of phosphorylated STAT3 (T705) and PAK1/2 (T423/T402) amounts in FAK-KO cells after exogenous loading of the plasma membrane with cholesterol and GM1 (s.d., n = 4 experiments, each with 104 cells, t-test).
Extended Data Figure 9 Hierarchical clustering of human ABC transporters according to 118 transcription factor binding profiles from the ENCODE database.
a, Diagram of the algorithm used to generate ABC transporter clusters. b, Heat map of the cluster of ABC transporters containing ABCA1, A9, A6 and G1 that share Tal1 binding (see bar graph representation of Tal1 binding on the right). These 4 ABC transporters are the same 4 ABC transporters that were found higher expressed in cells lacking FAK (FAK-KO) (see Extended Data Fig. 1c).
This file contains Supplementary Text and Data 1-3, a Supplementary Discussion, additional references and Supplementary Tables 2 and 4-6 (see separate excel files for Supplementary Tables 1 and 3). (PDF 1403 kb)
This file contains the Microarray data. (XLS 6222 kb)
This file contains the Lipidomics data. (XLS 604 kb)
TIRF live imaging of GFP-FAK in HeLla cells over 3.5 days. (MP4 21601 kb)
Principles and dynamics of the agent-based modelling of a population of cells (MP4 20598 kb)
Principles and dynamics of the mathematical modelling of the FAK-ABCA1 system. (MP4 18621 kb)
About this article
Cite this article
Frechin, M., Stoeger, T., Daetwyler, S. et al. Cell-intrinsic adaptation of lipid composition to local crowding drives social behaviour. Nature 523, 88–91 (2015). https://doi.org/10.1038/nature14429
Journal of The Royal Society Interface (2020)
Molecular & Cellular Proteomics (2020)
Current Biology (2019)
Scientific Reports (2019)
The FEBS Journal (2019)