Abstract
Experimental studies of cell growth, inheritance and their associated processes by microscopy require accurate single-cell observations of sufficient duration to reconstruct the genealogy. However, cell tracking—assigning identical cells on consecutive images to a track—is often challenging, resulting in laborious manual verification. Here, we propose fingerprints to identify problematic assignments rapidly. A fingerprint distance compares the structural information contained in the low frequencies of a Fourier transform to measure the similarity between cells in two consecutive images. We show that fingerprints are broadly applicable across cell types and image modalities, provided the image has sufficient structural information. Our tracker (TracX) uses fingerprints to reject unlikely assignments, thereby increasing tracking performance on published and newly generated long-term data sets. For Saccharomyces cerevisiae, we propose a comprehensive model for cell size control at the single-cell and population level centered on the Whi5 regulator, demonstrating how precise tracking can help uncover previously undescribed single-cell biology.
This is a preview of subscription content, access via your institution
Access options
Access Nature and 54 other Nature Portfolio journals
Get Nature+, our best-value online-access subscription
$29.99 / 30 days
cancel any time
Subscribe to this journal
Receive 12 print issues and online access
$259.00 per year
only $21.58 per issue
Buy this article
- Purchase on Springer Link
- Instant access to full article PDF
Prices may be subject to local taxes which are calculated during checkout
Similar content being viewed by others
Data availability
The data used in this study as well the analysis scripts are available for download from the ETH Research Collection at https://doi.org/10.3929/ethz-b-000550509 (ref. 54). A summary table of external data used can be found in the Supplementary Material.
Code availability
The TracX software can be downloaded from https://gitlab.com/csb.ethz/tracx as well as demo data from https://gitlab.com/csb.ethz/tracx_demo_data. Further documentation can be found under https://tracx.readthedocs.io/en/latest/.
References
Luro, S., Potvin-Trottier, L., Okumus, B. & Paulsson, J. Isolating live cells after high-throughput, long-term, time-lapse microscopy. Nat. Methods 17, 93–100 (2020).
Camsund, D. et al. Time-resolved imaging-based crispri screening. Nat. Methods 17, 86–92 (2020).
Kuchen, E. E., Becker, N. B., Claudino, N. & Höfer, T. Hidden long-range memories of growth and cycle speed correlate cell cycles in lineage trees. eLife 9, e51002 (2020).
Loeffler, D. et al. Asymmetric lysosome inheritance predicts activation of haematopoietic stem cells. Nature 573, 426–429 (2019).
Moen, E. et al. Deep learning for cellular image analysis. Nat. Methods 16, 1233–1246 (2019).
Stringer, C., Wang, T., Michaelos, M. & Pachitariu, M. Cellpose: a generalist algorithm for cellular segmentation. Nat. Methods 18, 100–106 (2020).
Meijering, E. A bird’s-eye view of deep learning in bioimage analysis. Comput. Struct. Biotech. J. 18, 2312–2325 (2020).
Han, H., Wu, G. & Zi, Z. eDetect: a fast error detection and correction tool for live cell imaging data analysis. iScience 13, 1–8 (2019).
Sorokin, D. V. & Matula, P. Cell tracking accuracy measurement based on comparison of acyclic oriented graphs. PLoS ONE 10, e0144959 (2015).
Ulman, V. et al. An objective comparison of cell-tracking algorithms. Nat. Methods 14, 1141–1152 (2017).
Versari, C. et al. Long-term tracking of budding yeast cells in brightfield microscopy: cellstar and the evaluation platform. J. R. Soc. Interface 14, 20160705 (2017).
Hilsenbeck, O. et al. Software tools for single-cell tracking and quantification of cellular and molecular properties. Nat Biotech 34, 703–706 (2016).
Amat, F. et al. Fast, accurate reconstruction of cell lineages from large-scale fluorescence microscopy data. Nat. Methods 11, 951–958 (2014).
Bray, M.-A. & Carpenter, A. E. CellProfiler Tracer: exploring and validating high-throughput, time-lapse microscopy image data. BMC Bioinform. 16, 369 (2015).
Winter, M., Mankowski, W., Wait, E., Temple, S. & Cohen, A. R. LEVER: software tools for segmentation, tracking and lineaging of proliferating cells. Bioinformatics 32, 3530–3531 (2016).
Lugagne, J.-B., Lin, H. & Dunlop, M. J. Delta: automated cell segmentation, tracking, and lineage reconstruction using deep learning. PLoS Comput. Biol. 16, e1007673 (2020).
Rullan, M., Benzinger, D., Schmidt, G. W., Milias-Argeitis, A. & Khammash, M. An optogenetic platform for real-time, single-cell interrogation of stochastic transcriptional regulation. Molecular Cell 70, 745–756.e6 (2018).
Cox, I. J., Kilian, J., Leighton, F. T. & Shamoon, T. Secure spread spectrum watermarking for multimedia. IEEE Trans. Image Proc. 6, 1673–1687 (1997).
Izumi, I., Hitoshi, K. DCT sign-only correlation with application to image matching and the relationship with phase-only correlation. In Proc. 2007 IEEE International Conference on Acoustics, Speech and Signal Processing - ICASSP ’07 1237–1240 (2007).
Fei, M., Ju, Z., Zhen, X. & Li, J. Real-time visual tracking based on improved perceptual hashing. Multimed. Tools Appl. 76, 4617–4634 (2017).
Mayer, C., Dimopoulos, S., Rudolf, F. & Stelling, J. Using CellX to quantify intracellular events. Curr. Protoc. Mol. Biol. https://doi.org/10.1002/0471142727.mb1422s101 (2013).
Dimopoulos, S., Mayer, C. E., Rudolf, F. & Stelling, J. Accurate cell segmentation in microscopy images using membrane patterns. Bioinformatics 30, 2644–2651 (2014).
Ricicova, M. et al. Dissecting genealogy and cell cycle as sources of cell-to-cell variability in MAPK signaling using high-throughput lineage tracking. Proc. Natl Acad. Sci. USA 110, 11403–8 (2013).
Jonker, R. & Volgenant, A. A shortest augmenting path algorithm for dense and sparse linear assignment problems. Computing 38, 325–340 (1987).
Kuhn, W. H. The Hungarian method for the assignment problem. Nav. Res. Logist. Q. 2, 83–97 (1955).
Delgado-Gonzalo, R., Nicolas, D., Maerkl, S. & Unser, M. Multi-target tracking of packed yeast cells. In Proc. 7th IEEE International Symposium on Biomedical Imaging 544–547 (IEEE, 2010).
Wang, Q., Niemi, J., Tan, C. M., You, L. & West, M. Image segmentation and dynamic lineage analysis in single-cell fluorescence microscopy. Cytometry Part A. 77, 101–110 (2010).
Carpenter, A. E. et al. CellProfiler: image analysis software for identifying and quantifying cell phenotypes. Genome Biol. 7, R100 (2006).
Gordon, A. et al. Single-cell quantification of molecules and rates using open-source microscope-based cytometry. Nat. Methods 4, 175–181 (2007).
Ezgi Wood, N. & Doncic, A. A fully-automated, robust, and versatile algorithm for long-term budding yeast segmentation and tracking. PLoS ONE 14, e0206395 (2019).
Bi, E. et al. Involvement of an actomyosin contractile ring in Saccharomyces cerevisiae cytokinesis. J. Cell Biol. 142, 1301–1312 (1998).
Klein, J. et al. TLM-tracker: software for cell segmentation, tracking and lineage analysis in time-lapse microscopy movies. Bioinformatics 28, 2276–2277 (2012).
Maška, M. et al. A benchmark for comparison of cell tracking algorithms. Bioinformatics 30, 1609–1617 (2014).
Di Talia, S., Skotheim, J. M., Bean, J. M., Siggia, E. D. & Cross, F. R. The effects of molecular noise and size control on variability in the budding yeast cell cycle. Nature 448, 947–951 (2007).
Schmoller, K. M., Turner, J., Kõivomägi, M. & Skotheim, J. M. Dilution of the cell cycle inhibitor Whi5 controls budding-yeast cell size. Nature 526, 268–272 (2015).
Litsios, A. et al. Differential scaling between G1 protein production and cell size dynamics promotes commitment to the cell division cycle in budding yeast. Nature Cell Biol. 21, 1382–1392 (2019).
Chen, Y., Zhao, G., Zahumensky, J., Honey, S. & Futcher, B. Differential scaling of gene expression with cell size may explain size control in budding yeast. Mol. Cell 78, 359–370 (2020).
Qu, Y. et al. Cell cycle inhibitor Whi5 records environmental information to coordinate growth and division in yeast. Cell Reports 29, 987–994 (2019).
Barber, F., Amir, A. & Murray, A. W. Cell-size regulation in budding yeast does not depend on linear accumulation of whi5. Proc. Natl Acad. Sci. USA 117, 14243–14250 (2020).
Garmendia-Torres, C., Tassy, O., Matifas, A., Molina, N. & Charvin, G. Multiple inputs ensure yeast cell size homeostasis during cell cycle progression. eLife 7, e34025 (2018).
Liu, X. et al. Reliable cell cycle commitment in budding yeast is ensured by signal integration. eLife 4, e03977 (2015).
Leitao, R. M. & Kellogg, D. R. The duration of mitosis and daughter cell size are modulated by nutrients in budding yeast. J. Cell Biol. 216, 3463–3470 (2017).
Soifer, I., Robert, L. & Amir, A. Single-cell analysis of growth in budding yeast and bacteria reveals a common size regulation strategy. Curr. Biol. 26, 356–361 (2016).
Mayhew, M. B., Iversen, E. S. & Hartemink, A. J. Characterization of dependencies between growth and division in budding yeast. J. R. Soc. Interface 14, 20160993 (2017).
Johnston, G., Ehrhardt, C., Lorincz, A. & Carter, B. Regulation of cell size in the yeast Saccharomyces cerevisiae. J. Bacteriology 137, 1–5 (1979).
Blank, H. M., Callahan, M., Pistikopoulos, I. P., Polymenis, A. O. & Polymenis, M. Scaling of G1 duration with population doubling time by a cyclin in Saccharomyces cerevisiae. Genetics 210, 895–906 (2018).
Stylianidou, S., Brennan, C., Nissen, S. B., Kuwada, N. J. & Wiggins, P. A. Supersegger: robust image segmentation, analysis and lineage tracking of bacterial cells. Mol. Microbiol. 102, 690–700 (2016).
Rudin, L. I., Osher, S. & Fatemi, E. Nonlinear total variation based noise removal algorithms. Physica D 60, 259–268 (1992).
Weigert, M., Schmidt, U., Haase, R., Sugawara, K. & Myers, G. Star-convex polyhedra for 3D object detection and segmentation in microscopy. In Proc. IEEE/CVF Winter Conference on Applications of Computer Vision 3666–3673 (IEEE, 2020).
Dietler, N. et al. A convolutional neural network segments yeast microscopy images with high accuracy. Nature Commun. 11, 5723 (2020).
Schmidt, G. W., Frey, O. & Rudolf, F. The cellclamper: a convenient microfluidic device for time-lapse imaging of yeast. Methods Mol. Biol. 1672, 537–555 (2018).
Schmidt, G. W., Cuny, A. P. & Rudolf, F. Preventing photomorbidity in long-term multi-color fluorescence imaging of Saccharomyces cerevisiae and S. pombe. G3: Genes Genom. Genet. 10, 4373–4385 (2020).
Lang, M., Rudolf, F. & Stelling, J. Use of youscope to implement systematic microscopy protocols. Curr. Protoc. Mol. Biol. 98, 14–21 (2012).
Cuny, A. P., Ponti, A. & Stelling, J. Cell Region Fingerprints Enable Highly Precise Single-Cell Tracking and Lineage Reconstruction: Data Collection (ETH Zurich Research Collection, 2022); https://doi.org/10.3929/ethz-b-000550509
Acknowledgements
We thank M. Dürr for initial implementation of Ricicova’ tracker in R and U. Küchler for its translation to MATLAB. We thank G. Schmidt for training of the CellClamper and sharing data sets.
Author information
Authors and Affiliations
Contributions
A.P.C., F.R. and J.S. conceptualized the project. A.P.C. developed the CRF and wrote the TracX software. A.P. wrote the Gaussian vector filtering. T.K. initiated and codeveloped the graphical user interface and wrote the amorphous cell lineage reconstruction. A.P.C. engineered the cell strains and performed the microscopy experiments. A.P.C. and J.S. validated the software and analyzed the data. A.P.C. and J.S. wrote the manuscript. All authors read and approved the final manuscript.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Peer review
Peer review information
Nature Methods thanks Beth Cimini, Ralf Mikut and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available. Primary Handling Editor: Rita Strack, in collaboration with the Nature Methods team.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Extended data
Extended Data Fig. 1 The effect of the re-size factor on CRF-based classification.
fr values are indicated per row and data presented as in Fig. 3b.
Extended Data Fig. 2 Sensitivity of cell region fingerprinting to cell rotations.
a: Schematics of rotation analysis. To evaluate the sensitivity of the df to cell rotations within a cell neighborhood, images were rotated between -90 and 90 degrees and then cropped with windows of sizes fl ∈ [25; 40; 100] px. b-f: The df is computed between the non-rotated and rotated image crops for each cell and we plot the fraction of correct assignments (defined as the sum of all occurrences where the df was smallest over all possible assignments) as a function of the rotation angle in degrees. Data sets used were: b: TTS-SC8-BF; c: TTS-60mrnaCropped; d: TTS-SC9-BF; e: TTS-SC9-FL; f: TTS-Fluo-N2DH-GOWT1. a-f n = 7/ 25/ 610/ 610/ 35 cells (repetitions) each from an independent experiment. c-f: scale bar, 32px.
Extended Data Fig. 3 Evaluation of the effect of the neighborhood fraction threshold (τf) for the different cell types and image modalities in Fig. 3.
For each data set, we chose a random parameter set yielding a high F-score according to the data in Fig. 3. Window sizes (fl), frequencies (fq) and resizing (fr) were [30, 8, 32], [300, 12, 40], [300, 10, 25], and [200, 20, 32]. We permuted assignments by the fractions indicated. The very conservative value τf = 0 can be relaxed to ≈ 0.2 while still accurately detecting matching consecutive (cellular) regions on all images and for all permutation frequencies.
Extended Data Fig. 4 Flow chart of the TracX core software architecture.
A tracking project defines all the required parameters and links to input files such as raw images and segmentation masks. Tracking is then started in a frame-by-frame manner. First, the CRF is calculated for each segmented centroid on each image frame. A first assignment is obtained from the LAP tracker process. These assignments are evaluated for correctness using the fingerprint distance (df), before refinements to handle unexpected motion where the image frequency was too low for the df to be informative enough. When all frames were tracked, results are saved and displayed to the user. Boxes encode for manual user input (trapezoid), documents (waved rectangle), process (rectangle), decision (square standing on the tip), start and end of the flow (ellipse), data display (ellipse with the tip to the right), complying with the commonly used ISO 5807:1985.
Extended Data Fig. 5 Effect of imaging frequency on the false positive rate.
The relative rate of false positives (FPs) as a function of neighborhood fraction thresholds (τf) evaluated for different imaging frequencies (colored dots). a False color image overlays of the first image frame with the one after 2, 6, 10, 14, 18, 22, 26, 30 minutes to simulate different imaging frequencies. Data set TTS-SC9; n = 8 representative image pairings out of 1980 possible pairings. Scale bar: 10μm. b Fraction of false positives as a function of neighborhood fraction threshold (τf). Neighborhood window size was 60 px or 8.80 ± 0.14 neighbors (left panel). Neighborhood window size was 80 px or 13.70 ± 0.31 neighbors (right panel). n = 520 assignments.
Extended Data Fig. 6 Effects of mis-segmentation on cell region fingerprint performance.
a: Schematic depicting randomly introduced erroneous merges (left) and splits (right) of cells on an image frame of test set TTS-SC7. Original segmentation (green or white outlines) and randomly merged cells (left, pink) or splits (right, pink). For better visibility, only half of the split cell is plotted. b: Erroneous merges (purple) and splits (orange) of cells were generated by randomly selecting 10% of the tracks and merging the segmentation masks of two neighboring cells or splitting single cells. For a selected track, we propagated the error for one to five consecutive frames (5-25 min). A true positive ratio (tpr) of one indicates that all introduced errors are captured after tracking with identical parameters compared to the ground truth by Ff > 0. c: Same as in b for five consecutive frames at different error rates. In both analyses, note the slightly higher scores for erroneous merges. We reason that the assignment score reflects erroneous merges better because the new cell center will be between the cells at a new position, which shifts the cell neighborhood relevant for the cell region fingerprint quite substantially. For erroneous splits, shifts of new cell centers relative to the correct cell center, and thereby changes to the neighborhood, are less pronounced.
Extended Data Fig. 7 The effect of the edge sensitivity threshold on asymmetric lineage reconstruction performance.
a Left: total variation regularization (ROF47) filtered false colored fluorescence image channel depicting the bud neck (green) with cell segmentation outlines (white). Middle and right: bud necks (magenta) detected for edge sensitivity thresholds of 0.09 (middle) and 0.225 (right). The data set used was TTS-SC7 (see Table S3). Three representative images shown out of n = 115/ 59 independent images from two independent data sets. Scale bar: 10μm. b Assignment counts as a function of edge sensitivity threshold classified into true positive (TP, bud to mother assignment correct), true negative (TN, no mother for bud), false positive (FP, wrong bud to mother assignment), and false negative (FN, missing bud to mother assignment).
Extended Data Fig. 8 Cell cycle regulation by Whi5 in S. cerevisiae for the selected track in Fig. 6a.
Merge of bright field image channel with the two fluorescent channels depicting Whi5 (magenta) and Myo1 (cyan; see Methods for details). Contrast adjusted to brightest pixels for both fluorescent channels. The selected cell from Fig. 6a is centered in each tile, with overlay of the outline of its segmentation mask (green). Scale bar: 10μm. The exemplary cell shown is one out of more than n =700 cells (repetitions) originating from 6 independently acquired data sets.
Extended Data Fig. 9 Cell cycle regulation by Whi5 in S. cerevisiae.
a,b Identification of threshold for nuclear concentrations of Whi5 by a two-component Gaussian mixture model (see Methods for details). a Example of empirical (bars) and fitted (lines) probability densities as well as inferred threshold (red line). b Ranking of cluster membership scores based on posterior probability indicates good cluster separation. c-e Cell cycle characteristics as a function of cell age and glucose concentration complementing Fig. 6b-e for growth rate in G1 (c), daughter volume at next division (d), and nuclear Whi5 in the mother at division (e). Sample sizes and statistical analysis are identical to Fig. 6b-e; see also Table S4. f Correlation plot as in Fig. 6f-i for growth rate in G1 and cellular Whi5 in mothers at the next division. g Effects plot for linear models for G1, G2/M, and total cell cycle duration (log-scaled response variables; see Fig. 6j). h Single-cell data vs model predictions for cell cycle duration as in Fig. 6k,l.
Extended Data Fig. 10 Whi5 concentration estimation based on cell volume.
Panels correspond to those for concentration estimation based on cell area as follows: a: Fig. 6d; b: Extended Data Fig. 9e; c: Fig. 6f; d: Fig. 6g; e: Fig. 6i; f: Extended Data Fig. 9f; g: Fig. 6j; h: Fig. 6k,l; i: Extended Data Fig. 9g; j: Extended Data Fig. 9h. Sample sizes, statistical analysis, and data presentation in a-b are identical to the referenced panels; see also Table S4 for details.
Supplementary information
Supplementary Information
Supplementary Figs. 1–4, Tables 1–3 and 5–7 and References.
Supplementary Video 1
Lineage results for asymmetric cell division (S. cerevisiae). Cells colored by lineage tree of seeding cells from first frame. Details about data set TTS-SC7 are given in Supplementary Table 3.
Supplementary Video 2
Lineage results for symmetric cell division (S. pombe). Cells colored by lineage tree of seeding cells from first frame. Details about data set TTS-SP4 are given in Supplementary Table 3.
Supplementary Video 3
Lineage results for symmetric cell division (B. megaterium). Cells colored by lineage tree of seeding cells from first frame. Details about data set TTS-Bmeg are given in Supplementary Table 3.
Supplementary Video 4
Lineage results for symmetric cell division (HeLa cells). Subset of TTS-Fluo-N2DL-HeLa. Cells colored by lineage tree of seeding cells from first frame. Details about the data set are given in Supplementary Table 3.
Supplementary Video 5
Missing parent assignment errors for symmetric cell division (HeLa cells). Full data set (TTS-Fluo-N2DL-HeLa). Tracks that were not assigned to a parent are colored red in the first frame that they appear in. Details about the data set are given in Supplementary Table 3.
Supplementary Table 4
Detailed statistics (sample sizes, P values and confidence intervals) for Fig. 6 and Extended Data Figs. 9 and 10.
Rights and permissions
Springer Nature or its licensor holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Cuny, A.P., Ponti, A., Kündig, T. et al. Cell region fingerprints enable highly precise single-cell tracking and lineage reconstruction. Nat Methods 19, 1276–1285 (2022). https://doi.org/10.1038/s41592-022-01603-2
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1038/s41592-022-01603-2
This article is cited by
-
The CellPhe toolkit for cell phenotyping using time-lapse imaging and pattern recognition
Nature Communications (2023)
-
Bridging live-cell imaging and next-generation cancer treatment
Nature Reviews Cancer (2023)
-
Improved tracking via cell region fingerprints
Nature Methods (2022)