Root Vivo V2024

[Paulette Dzurilla]

Thankyou for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

The final size, shape and function of tissues in multicellular organisms hinge upon the precise control of cell division4. Owing to intrinsic and extrinsic cell polarity, a 90 rotation of the division plane determines whether a cell will divide formatively (producing daughter cells with different fates) or proliferatively5,6 (producing daughter cells with similar fates). A wrong choice can lead to over-proliferation of cells, resulting in aberrant morphogenesis or tumorigenesis7,8. Developmental regulators that specify cell fate and interface directly with the cell cycle machinery9,10,11 are likely arbiters of this decision. However, we have limited knowledge about how these regulators dynamically control cell division in situ.

Cruz-Ramrez et al.3 proposed a bistable model to explain both how and where SHR and SCR trigger the decision to divide. According to the model, two positive feedback loops generate high stable steady states of SCR and nuclear SHR, triggering formative division (Fig. 1b). Bistability arising from positive feedback is at the heart of mathematical models of decision making in many systems16. However, positive feedback does not always lead to bistability17, and alternative decision-making mechanisms exist. For example, the simple presence of a factor at the right place and time can alter the cell cycle programme and lead to a different cell fate11.

Quantitative time-lapse imaging of transcription factor dynamics has provided key insights into gene-regulatory network function in single cell organisms and mammalian cell lines18,19,20. Assays of multiple transcription factors in tandem on a long timescale can enable examination of their regulatory relationships21. However, many technical challenges have made studies of network dynamics in vivo difficult22. Phototoxicity and photobleaching, in particular, restrict studies using confocal microscopy to short timescales or infrequent sampling and limit the number of fluorophores that can be imaged simultaneously. Owing to its lower phototoxicity, light sheet microscopy provides the means for longer-term multi-colour imaging of protein dynamics in vivo. This potential has been extolled for nearly two decades, but the technology has been used primarily for observation of cellular dynamics and morphology changes during development23,24,25.

Therefore, we hypothesized that a threshold amount of SHR and SCR triggers formative division at an earlier timepoint. To test this, we determined the accuracy of predicting formative division across a range of SHR and SCR thresholds (Fig. 3a and Supplementary Methods). Optimal thresholds were low relative to the full range of SHR and SCR levels and were able to accurately predict formative division 80% and 73% of the time, respectively. A similar analysis of the SHR confocal data found a maximum prediction accuracy of 87% (Fig. 3a).

To improve these predictions, we considered the possibility that dynamic features of the SHR and SCR trajectories or position in the cell cycle may contribute to the decision to divide formatively. We took a simple machine learning approach (Supplementary Methods) to determine whether we could predict which cells divide using a set of features describing various aspects of the dynamics of the SHR, SCR and nuclear size trajectories (for example, maximum rate, mean SHR and area under the curve; Supplementary Table 1). We used nuclear size as a proxy for position in the cell cycle11 (Fig. 3b and Supplementary Methods). Our learning model was able to predict whether a cell divides formatively 89% and 92% of the time for the light sheet and confocal data, respectively.

To determine the most predictive features, we assessed the ability of each individual feature to discriminate between formatively dividing and non-dividing cells (Supplementary Methods). In addition to features associated with SHR levels, features relating to nuclear size were significant predictors of formative division (Supplementary Tables 2 and 3), suggesting that threshold levels of SHR might be required during a specific window of the cell cycle for formative division to occur.

a, Threshold levels of SHR and SCR specify formative division only when present during G1 or early S. b, The presence of SHR and SCR during G1 and early S activates CYCD6 to specify the orientation of the division plane, whereas other cyclins and developmental cues commit the cell to division. CYCD6 and other cyclins along with their associated kinases phosphorylate RBR, committing the cell to formative division. The two positive feedback loops (SCR autoregulatory loop and RBR release of SCR after phosphorylation by CYCD6) have a smaller role in the decision to divide formatively than previously predicted3.

The pre-processed image files are available in the Duke Digital Research Data Repository ( ; dataman...@duke.edu). Owing to their large size, original image files are available upon request (please contact researchd...@duke.edu for the first 6 years from publication. To inquire about the availability of this dataset beyond 6 years, please contact car...@gmail.com). Complete trajectory data and all metadata needed to run the code are included in the Supplementary material. Source data for figures that were not generated by the code are provided in Excel files. Source data are provided with this paper.

Custom code was central to the conclusions of the paper. The RootTracker (for microscope hardware control), image processing and quantification pipeline, and trajectory data analysis pipeline code are available at

We express profound gratitude for the mentorship and unwavering support of Philip Benfey, who recently passed away. He was a guiding light whose visionary thinking shaped the lives and careers of many scientists. We are thankful that he was able to read the final version of this manuscript, and that he could share in the excitement of our findings. We also thank S. DiTalia, L. You, J. Socolar, R. Shahan, R. Sozzani, E. Pierre-Jerome, I. Taylor, T. Nolan, M. Zhu, S. Van Dierdonck, Q. Zhou and O. Szekely for critical reading and discussions of the manuscript; O. Szekely for help with graphics; D. Holland, F. Cutrale and J. Choi for contributing to the light sheet microscope design and construction; and L. Cameron and the Light Microscopy Core Facility at Duke for providing the workstations and support for Imaris image analysis. This work was funded by the US National Institutes of Health (NRSA 5F32GM106690-02 and MIRA 1R35GM131725) to C.M.W. and P.N.B., and by the Howard Hughes Medical Institute to P.N.B. as an Investigator. M.J., S.E.F. and T.V.T. were supported by the Translational Imaging Center, Bridge Institute, University of Southern California.

C.M.W. and P.N.B. conceived the project and designed the experiments. C.M.W. and H.B. conducted experiments and generated transgenic plants. T.V.T. developed the light sheet imaging platform, with contributions from M.J. and S.E.F. V.P. developed the microscope control code and the image data extraction pipeline. C.M.W., H.B. and R.C. performed image analysis. P.S. developed computational tools and performed data analysis of the trajectories. C.M.W. and P.S. interpreted the results and wrote the paper with comments from all authors.

a, Imaging chamber. b, Capillary tube containing growing root mounted onto custom holder. The holder is lowered into the imaging chamber for imaging. c, Image acquisition and analysis pipeline to produce SHR and SCR trajectories for confocal and light sheet imaging.

Registered median slices of inducible SHR timecourse. Registered median longitudinal z-slices from a confocal time course of a growing SHR:GAL4-GR UAS:SHR-GFP 35S:H2B-RFP shr2 root after induction with 10 μM dex. Magenta, H2B-RFP; Green, SHR-GFP. The white box highlights cell 3 in the left cell file up to an formative division.

Registered median slices of low dex inducible SHR timecourse. Registered median longitudinal z-slice of a confocal time course of a growing SHR:GAL4-GR UAS:SHR-GFP 35S:H2B-RFP shr2 root after low dex (0.02uM) induction. The SHR levels peak at a low level, go back down, and after several hours the cell divides proliferatively. Corresponds to images shown in Figure 1e. Green, SHR-GFP; Magenta, H2B-RFP.

Registered maximum projection of inducible SHR and SCR timecourse. Registered 3D reconstruction of a light sheet time course of a growing SHR:GAL4-GR UAS:SHR-GFP SCR:SCR-mKATE2 UBQ10:H2B-CFP shr2 root after induction with 10 μM dex. The spheres toward the end of the video show the nuclei detected in Imaris that were used for quantification of SHR, SCR, and H2B fluorescence intensity. Cyan, H2B-CFP used for normalization; Green, SHR-GFP; Magenta, SCR-mKate2.

Registered maximum projection of PlaCCI timecourse. Maximum projection of a light sheet time course of a PlaCCI root used to correlate nuclear size with position in the cell cycle. Blue: CDT1a-CFP (G1 marker); Red: H3.1-mCHERRY; Green: CYCB1;1-GFP.

Registered median slices of inducible SHR timecourses in cell cycle synchronized roots. Registered median longitudinal z-slices of confocal time courses of growing SHR:GAL4-GR EN7:H2B-RFP shr2 roots after induction with 10 μM dex. Roots were pre-treated for 17 hours with 10 μM hydroxyurea (synchronizes cells at G1/S of the cell cycle), 2 μM oryzalin (synchronizes cells at G2/M of the cell cycle) treatment, or a control treatment (transfer to 1/2 MS only plates). The EN7 promoter is active only in the ground tissue, so H2B-RFP is expressed in the mutant ground tissue layer and in the endodermis and cortex after division. Green, SHR-GFP; Magenta, SCR-mKATE2.

3a8082e126