Single-cell analysis is pivotal to deciphering complex phenomena like heterogeneity, bistability, and asynchronous oscillations, where a population ensemble cannot represent individual behaviors. Bulk cell-free systems, despite having unique advantages of manipulation and characterization of biochemical networks, lack the essential single-cell information to understand a class of out-of-steady-state dynamics including cell cycles. Here, by encapsulating Xenopus egg extracts in water-in-oil microemulsions, we developed artificial cells that are adjustable in sizes and periods, sustain mitotic oscillations for over 30 cycles, and function in forms from the simplest cytoplasmic-only to the more complicated ones involving nuclear dynamics, mimicking real cells. Such innate flexibility and robustness make it key to studying clock properties like tunability and stochasticity. Our results also highlight energy as an important regulator of cell cycles. We demonstrate a simple, powerful, and likely generalizable strategy of integrating strengths of single-cell approaches into conventional in vitro systems to study complex clock functions.https://doi.org/10.7554/eLife.33549.001
Spontaneous progression of cell cycles represents one of the most extensively studied biological oscillations. Cytoplasmic extracts predominantly from Xenopus eggs (Murray, 1991) have made major contributions to the initial discovery and characterization of the central cell-cycle regulators including the protein complex cyclin B1-Cdk1 (Masui and Markert, 1971; Murray et al., 1989; Lohka and Maller, 1985; Lohka et al., 1988) and the anaphase-promoting complex or cyclosome (APC/C) (Sudakin et al., 1995). Cell-free extracts have also been used to investigate downstream mitotic events such as spindle assembly and chromosome segregation (Hannak and Heald, 2006). Moreover, detailed dissections of the regulatory circuits in these extracts have revealed architecture of interlinked positive and negative feedbacks (Kumagai and Dunphy, 1992; Mueller et al., 1995; Yang and Ferrell, 2013; Chang and Ferrell, 2013; Trunnell et al., 2011; Kim and Ferrell, 2007; Pomerening et al., 2005; Pomerening et al., 2003; Novak and Tyson, 1993; Thron, 1996) (Figure 1A). Such interlinked feedback loops are also found in many other biological oscillators (Rust et al., 2007; Hoffmann et al., 2002; Cross, 2003; Lee et al., 2000) and have been shown computationally to play an important role in achieving the essential clock functions such as robustness and tunability (Tsai et al., 2008). These studies have stimulated major interests in quantitative characterization of clock functions, for which an experimental platform is still lacking.
Compared to in vivo systems, circuits reconstituted in cell-free extracts contain well-defined recombinant molecules and are more amenable to systematic design, manipulation and quantitative biochemical measurements. However, one major limitation for most in vitro reconstitutions up to date is that oscillations are generated in well-mixed bulk solutions, which tend to produce quickly damped oscillations (Pomerening et al., 2005; Nakajima et al., 2005). Additionally, these bulk reactions lack the similarity to the actual cell dimensions and the ability of mimicking spatial organization achieved by functional compartmentalization in real cells. These limitations make it impossible to retrieve the cellular heterogeneity to investigate important and challenging questions, such as stochasticity and tunability of an oscillator.
To overcome these challenges, we developed an artificial cell cycle system by encapsulating reaction mixtures containing cycling Xenopus egg cytoplasm (Murray, 1991) in cell-scale micro-emulsions. These droplet-based cells are stable for days and keep oscillating for dozens of cycles, offering large gains in high-throughput and long-term tracking of dynamical activities in individual droplets. In this system, we successfully reconstituted a series of mitotic events including chromosome condensation, nuclear envelope breakdown and destruction of anaphase substrates such as the proteins securin and cyclin B1. The oscillation profiles of the system such as period and number of cycles can be reliably regulated by the amount of cyclin B1 mRNAs or sizes of droplets. Additionally, we found that energy may be a critical factor for cell cycle behaviors.
To create a cell-like in vitro mitotic system, we used a simple vortexing technique (Ho et al., 2017) to compartmentalize reactions of cycling Xenopus egg extracts (Murray, 1991) into microemulsion droplets, with radii ranging from 10 µm to 300 µm (Figure 1B, Materials and methods). The droplets were loaded into a Teflon-coated chamber and recorded using long-term time-lapse fluorescence microscopy. The fluorescence time courses of each droplet were analyzed to obtain information of period, amplitude, number of cycles, droplet size, etc.
To examine the functionality of the droplet mitotic system, we added de-membranated sperm chromatin, purified green fluorescent protein-nuclear localization signal (GFP-NLS), securin-mCherry mRNA and Hoechst 33342 dye to the cytoplasmic extracts. We demonstrated that the system is capable of reconstructing at least three mitotic processes in parallel that alternate between interphase and mitosis (Figure 1—video 1). An example artificial cell that undergoes these mitotic processes is shown in Figure 1C. The autonomous alternation of distinct cell-cycle phases is driven by a self-sustained oscillator, the activity of which was indicated by the periodic degradation of an anaphase substrate of APC/C, securin-mCherry. In interphase, the presence of sperm chromosomal DNA, labeled by Hoechst, initiated the self-assembly of a nucleus, upon which GFP-NLS protein was imported through the nuclear pores. The spatial distributions of Hoechst and GFP-NLS thus coincided in an interphase nucleus (Figure 1C columns 1, 3, and 5). As the artificial cell entered mitosis, the chromosome condensed resulting in a tighter distribution of Hoechst, while the nuclear envelope broke down and GFP-NLS quickly dispersed into a uniform distribution in the whole droplet (Figure 1C columns 2 and 4). The time courses for these processes were analyzed in Figure 1D, indicating that the chromosome condensation and nuclear envelop breakdown (NEB) happened almost at the same time, while securin degradation lagged behind these two processes at each cycle. All together, these experiments showed that the droplet system successfully reconstituted a cell-free mitotic oscillator centered on Cdk1 and APC/C that can reliably drive the periodic progression of downstream events including chromosome morphology change and nuclear envelope breakdown and re-assembly, like what occurs in vivo.
The ability to adjust frequency is an important feature shared by many oscillators (Tsai et al., 2008). Here, we demonstrated that the present system provides an effective experimental solution to the study of tunability of the clock. To avoid any interference from the complicated nuclear dynamics, we reconstituted a minimal mitotic oscillatory system which, in the absence of sperm chromatin, formed no nuclei. This simple, cytoplasmic-only oscillator produced highly robust, undamped, self-sustained oscillations up to 32 cycles over a lifetime of 4 days (Figure 2A,B, Figure 2—video 1), significantly better than many existing synthetic oscillators.
To modulate the speed of the oscillations, we supplied the system with various concentrations of purified mRNAs of full-length cyclin B1 fused to YFP (cyclin B1-YFP), which function both as a reporter of APC/C activity and as an activator of Cdk1. A droplet supplied with both cyclin B1-YFP and securin-mCherry mRNAs exhibited oscillations with highly correlated signals (Figure 2C, Figure 2—video 2), suggesting that both are reliable reporters for the oscillator activity. With an increased concentration of cyclin B1-YFP mRNAs added to the system, we observed a decrease in the average period (Figure 2D, Figure 2—source data 1), meaning that a higher cyclin B1 concentration tends to speed up the oscillations. However, the average number of cycles (Figure 2E, Figure 2—source data 1) was also reduced with increased cyclin B1 concentrations, resulting in a negative correlation between the lifetime of oscillations and the amount of cyclin B1 mRNAs. The extracts will eventually arrest at a mitotic phase in the presence of high concentrations of cyclin B1.
Moreover, this system provides high flexibility in analyzing droplets with radii ranging from a few µm to 300 µm, enabling characterization of size-dependent behaviors of cell cycles. At the scale of a cell, the dynamics of biochemical reactions may become stochastic. Although stochastic phenomenon has been studied extensively in genetic expressions, studying a stochastic system that is out of steady-state can be challenging in living organism due to low throughput and complications from cell growth, divisions and other complex cellular environments. These limitations can be overcome by reconstitution of in vitro oscillators inside cell-scale droplets, which are in absence of cell growth and divisions. Parallel tracking of droplets also enables high-throughput data generation for statistical analysis. In vitro compartmentalization of molecules, especially rate-limiting molecules such as cyclin B1 mRNAs, into cell-sized droplets may have a major effect on the reaction kinetics of cell cycles. The smaller the size of a droplet, the smaller the copy number of molecules encapsulated inside the droplet and the larger the inherent stochasticity of biochemical reactions. Additionally, the partition errors of these molecule resulted from compartmentalization may further contribute to the variations of droplet behaviors in a size-dependent manner. Figure 2F (Figure 2—source data 2) showed that smaller droplets led to slower oscillations with a larger variance of the periods, consistent with the size effect reported on an in vitro transcriptional oscillator (Weitz et al., 2014). We also observed a reduced number of oscillations and a smaller variance of the cycle number in smaller droplets (Figure 2G, Figure 2—source data 2). Interestingly, these size effects become less dramatic for droplets with larger sizes or with higher concentrations of cyclin B1 mRNAs.
The results in Figure 2D–G indicated that the system is tunable by cyclin B1 mRNA concentration and droplet size in different manners. Although the period and number of cycles responded to varying droplet sizes in opposite directions, they followed the same trend when modulated by cyclin B1 mRNAs, resulting in a lifespan of the oscillatory system sensitive to cyclin B1 mRNA concentration. Moreover, we have observed that securin-mCherry and cyclin B1-YFP both exhibited oscillations of increased amplitude, baseline, and period over time (Figure 2C), of which the increasing period over time is evident by analysis in Figure 3—figure supplement 1A and B. These trends cannot be explained by existing cell cycle models (Yang and Ferrell, 2013; Tsai et al., 2014).
Unlike intact embryos, cell-free extracts lack yolk as an energy source and lack sufficient mitochondria for energy regeneration. We postulated that energy is an important regulator for a droplet system with a limited amount of energy source consumed over time. To gain insights into our experimental observations and better understand the in vitro oscillator system, we built a model to examine how energy consumption plays a role in modulating the oscillation behaviors. The energy depletion model is based on a well-established cell-cycle model (Yang and Ferrell, 2013; Tsai et al., 2014) modified by introducing ATP into all phosphorylation reactions (Figure 3A, Materials and methods 8 and 9).
In the cell cycle network, the activation of Cdk1 is co-regulated by a double positive feedback through a phosphatase Cdc25 and a double negative feedback through a kinase Wee1. The balance between Wee1 and Cdc25 activity was suggested to be crucial for the transition of cell cycle status during early embryo development (Tsai et al., 2014). In light of this, we defined the balance between Wee1 and Cdc25 by the ratio . We noted that ATP-dependent phosphorylation of Cdc25 and Wee1 can decrease R by activating Cdc25 and inhibiting Wee1 simultaneously, resulting in a high dependence of R on the ATP concentration (Figure 3B).
Using this model, we further investigated the relationship between ATP and the oscillation behaviors. We introduce a parameter r into our system to systematically change the ratio R (see Materials and methods 8). In Figure 3C, the phase plot of the two-ODE model shows that at a low r (e.g. 0.5), the system stays at a stable steady-state with low cyclin B concentration and at a high r (e.g. 2.5), the oscillation is arrested at a stable steady-state with high cyclin B concentration. At an intermediate value, increasing r can produce oscillations of increased amplitude, baseline and period (Figure 3C,D). If we assume that the available ATP concentration decreases over time, we can readily recapitulate the experimentally observed increment of amplitude, baseline, and period of the cyclin B time course (Figure 3E). The energy depletion model can also predict the experimental observations in Figure 2D and E by showing that both period and number of cycles decrease with increasing cyclin B concentration (Figure 3—figure supplement 1C, Figure 3—source data 1).
We noted that, besides phosphorylation, other processes, including protein synthesis and ubiquitination-mediated degradation, also consume ATPs and are sensitive to the energy level. However, the changes of synthesis and degradation rates yielded no obvious effects on the amplitude and baseline (Figure 3—figure supplement 1D).
We have developed here a novel artificial cell system that enables highly robust and tunable mitotic oscillations. The system is amenable to high throughput, quantitative manipulation and analysis of both cytoplasmic and nuclear processes. Given cell cycles share common topologies with many biological oscillators, the system may be valuable to investigate fundamental principles of oscillator theory.
Our energy depletion model suggested an interesting mechanism to modulate oscillations with a single control parameter r that depends on the energy-tunable balance of two positive feedback loops. Considering that the rapid, synchronous cleavages of an early embryo require a large amount of energy that remains unchanged for the first cleavage stages before rapidly dropping until the mid-blastula stage when cell cycles slow down (Zotin et al., 1967), this energy-dependent control may function as a ‘checkpoint’ to regulate cell cycles when r becomes large.
Cycling Xenopus extracts were prepared as described (Murray, 1991), except that eggs were activated with calcium ionophore A23187 (200 ng/µL) rather than electric shock. Freshly prepared extracts were kept on ice while applied with de-membranated sperm chromatin (to approximately 250 per µl of extract), GFP-NLS (10 µM) and recombinant mRNAs of securin-mCherry (10 ng/µL) and cyclin B1-YFP (ranging from 0 to 10 ng/µL). The extracts were mixed with surfactant oil 2% PFPE-PEG to generate droplets.
GFP-NLS protein was expressed in BL21 (DE3)-T-1 competent cells (Sigma Aldrich, B2935) that were induced by 0.1 mM IPTG (Isopropyl β-D-1-thiogalactopyranoside, Sigma Aldrich, I6758) overnight. Cells were broken down to release protein through sonication. GE Healthcare Glutathione Sepharose 4B beads (Sigma Aldrich, GE17-0756-01) and PD-10 column (Sigma Aldrich, GE17-0851-01) were used to purify and elute GFP-NLS protein. 200 mg/ml Hoechst 33342 (Sigma Aldrich, B2261) was added to stain chromosomes. Securin-mCherry and cyclin B1-YFP plasmids were constructed using Gibson assembly method (Gibson et al., 2009). All mRNAs were transcribed in vitro and purified using mMESSAGE mMACHINE SP6 Transcription Kit (Ambion, AM1340).
VitroCom miniature hollow glass tubing with height of 100 µm (VitroCom, 5012) was cut into pieces with lengths of 3–5 mm. A heating block was heated up to 95°C in a Fisher Scientific Isotemp digital incubator and then it was placed into a Bel-art F42025-0000 polycarbonate vacuum desiccator with white polypropylene bottom. The cut glass tubes and a 1.5 ml Eppendorf tube containing 30 µl Trichloro (1 hr,1H,2H,2H-perfluorooctyl) silane (Sigma Aldrich, 448931) were placed in the heating block. Vacuum was applied to the desiccator and the tubes was left incubated overnight.
To generate droplets, we used a Fisher Scientific vortex mixer to mix 20 µl cycling extract reaction mix, and 200 µl 2% PFPE-PEG surfactant (Ran Biotechnologies, 008-FluoroSurfactant-2wtH-50G) at speed level 10 for 3 s. By adjusting the vibration speed and ratio between aqueous and oil phase appropriately, we can obtain droplets with various sizes, ranging from 10 µm to 300 µm.
All imaging was conducted on an Olympus FV1200 confocal microscope under MATL mode (multiple area time lapse) and Olympus IX83 microscope equipped with a motorized x-y stage, at room temperature. Time-lapse images were recorded in bright-field and multiple fluorescence channels at a time interval of 6–9 min for at least 12 hr up to four days.
We used Imaris 8.1.2 (Bitplane Inc.) for image processing. Level-set method on bright-field images was used for droplet segmentation, and autoregressive motion algorithm was used for tracking. Tracks that had less than two oscillations were discarded. Results were then manually curated for accuracy. Means and standard deviations of fluorescence intensities as well as areas of each droplet were calculated for further analysis. The volume of a droplet was calculated using the formula proposed by a previous study (Good et al., 2013). To compensate for intensity drift over time, fluorescence intensity in droplets were normalized by average intensity of the background. For period calculation, Matlab (Mathworks Inc.) was used to detect peaks and troughs over the signal of mean intensity for cyclin B-YFP and securin-mCherry. All peaks were manually curated and edited to ensure reliability.
In Figure 2D–E, error bars indicate the median absolute deviations (MAD) of measurements pooled across multiple loading replicates. Each loading replicate refers to droplets encapsulated with extracts prepared from the same batch of Xenopus eggs that were loaded into a distinct individual Teflon-coated micro-chamber. For cyclin B mRNA concentrations of 0, 3, 5, 8, 10 ng/µL, respectively two, two, one, two and three loading replicates were performed, resulting in sample sizes of 373, 443, 227, 430 and 554 droplets that were analyzed. Extracts with cyclin B mRNA concentrations above 10 ng/µL did not exhibit sustained oscillations before arresting at a mitotic phase.
In Figure 2F–G, droplets generated with three cyclin B mRNA concentrations (0, 3, 5 ng/µL) were analyzed, with samples sizes of 246, 475, and 177 respectively. Experiments with cyclin B mRNA concentrations higher than 5 ng/µL showed similar behaviors as that of the 5 ng/µL cyclin B mRNA experiment and therefore are not included in the figures. Colored areas are defined by first and third quantiles after LOWESS (Locally Weighted Scatterplot Smoothing) with a window size of 20. The quartiles are calculated using running percentile with a binning size of 50. Volumes of droplets were first estimated using the compressed diameters (Good et al., 2013). The equivalent diameter was then calculated by the diameter of a sphere with the same volume as the estimated volume of a droplet.
In Figure 3B and D, we performed simulations at chosen ranges of ATP percentage (Figure 3B) and r (Figure 3D) to yield self-sustained oscillations. Error bars indicate the ranges of 3 independent simulations. The time courses in the figure inserts are randomly selected from the simulation replicates as examples.
In Figure 3—figure supplement 1A and B, measurements from experiments with 0 ng/µL cyclin B mRNA (from two loading replicates, 373 droplets, 5820 cycles) were pooled to analyze the change of period over time. In Figure 3—figure supplements 1C, 50 independent simulations were performed for each cyclin B mRNA concentration. Error bars indicate the standard deviations of the simulated data. In Figure 3—figure supplement 1D, the ranges of r and the degradation/synthesis rate were chosen to ensure all points in grids have self-sustained oscillations. All simulations in the supplementary figures were performed in a long enough time window to ensure stable statistics (period, amplitude) while the ATP concentration remained positive.
Complicated models have been proposed to describe the embryonic cell cycle oscillation (Novak and Tyson, 1993a; Ciliberto et al., 2003; Pomerening et al., 2005; Tsai et al., 2008). However, simple two-ODE models with fewer parameters are more amenable to analysis, while still capturing the general property of the feedback loops. We described the net productions of cyclin B1 and active cyclinB-Cdk1 complex using the following two equations (Yang and Ferrell, 2013; Tsai et al., 2014):
The parameters for the model are listed below:
Here, and refer to the concentrations of cyclin B1 and active cyclin B1-Cdk1 complex. is the concentration of active Wee1, while is the concentration of active Cdc25. We assumed that Cyclin B1 is synthesized at a constant rate. Its degradation rate is dependent on Cdk1 activity in the form of a Hill function with an exponent of 17 (Yang and Ferrell, 2013). Active cyclin B1-Cdk1 complex can also be eliminated through cyclin degradation. In addition, we considered the concentration of Cdk1 to be high compared to the peak concentration of cyclin B1 (Hochegger et al., 2001; Kobayashi et al., 1991) and the affinity of these cyclins for Cdk1 to be high (Kobayashi et al., 1994). Thus, there is no unbound form of cyclin B1, and the newly synthesized cyclin B1 is converted to cyclin-Cdk1 complexes, which are rapidly phosphorylated by the Cdk-activating kinase CAK to produce active Cdk1. According to previous studies, these complexes can then be inactivated by Wee1 and re-activated by Cdc25, via the double-negative and positive feedback loops, with Hill exponent of as 3.5 and as 11 (Kim and Ferrell, 2007; Trunnell et al., 2011).
We use a free parameter r, representing the ratio of the double negative and double positive feedback strengths, to permute the balance between the two feedbacks. This balance is suggested to be critical for oscillatory properties (Tsai et al., 2014). Note that this r is a parameter while R in the main text is a measurement that changes over a simulation.
In droplets that have small volumes and contain small numbers of molecules, the stochastic nature of the underlying biochemical reactions must be considered. We adapted a stochastic two-ODE model (Yang and Ferrell, 2013), and converted our two-ODE model to the corresponding chemical master equations (Kampen, 1992) and carried out numerical simulations using the Gillespie algorithm (Gillespie, 1977). The reaction rates and molecular stoichiometry are shown in Table 1.
To explore how energy consumption could affect the oscillations, we took ATP into account for phosphorylation and dephosphorylation of Wee1 (Tuck et al., 2013), such that:
In our model, we assumed Wee1 is in equilibrium with the activity of Cdk1 due to fast reactions between Cdk1 and Wee1. Using the reaction coefficients for Wee1 phosphorylation as and that for Wee1-Pi dephosphorylation as , along with the steady-state approximation, we have:
All above modifications for Wee1 reactions were also applied to Cdc25. After normalizing [ATP] and [ADP] by, we have the updated reaction rates summarized in Table 2. Here the [wee1]0 and [cdc25-Pi]0 represent the steady-state concentration of active Wee1 and Cdc25 when ATP is not considered in reaction. The ratios of the steady-state to total concentrations of Wee1 and Cdc25 can be calculated as a function of active Cdk1 using the parameters from previous work (Novak and Tyson, 1993).
Mathematical model of the morphogenesis checkpoint in budding yeastThe Journal of Cell Biology 163:1243–1254.https://doi.org/10.1083/jcb.200306139
Two redundant oscillatory mechanisms in the yeast cell cycleDevelopmental Cell 4:741–752.https://doi.org/10.1016/S1534-5807(03)00119-9
Exact stochastic simulation of coupled chemical reactionsThe Journal of Physical Chemistry 81:2340–2361.https://doi.org/10.1021/j100540a008
Stochastic Processes in Physics and ChemistryNew York: North-Holland.
Cyclins and their partners during Xenopus oocyte maturationCold Spring Harbor Symposia on Quantitative Biology 56:437–447.https://doi.org/10.1101/SQB.1991.056.01.051
Induction of nuclear envelope breakdown, chromosome condensation, and spindle formation in cell-free extractsThe Journal of Cell Biology 101:518–523.https://doi.org/10.1083/jcb.101.2.518
Cytoplasmic control of nuclear behavior during meiotic maturation of frog oocytesJournal of Experimental Zoology 177:129–145.https://doi.org/10.1002/jez.1401770202
Cell cycle regulation of a Xenopus Wee1-like kinaseMolecular Biology of the Cell 6:119–134.https://doi.org/10.1091/mbc.6.1.119
Modeling the Cell Division Cycle: M-phase Trigger, Oscillations, and Size ControlJournal of Theoretical Biology 165:101–134.https://doi.org/10.1006/jtbi.1993.1179
Building a cell cycle oscillator: hysteresis and bistability in the activation of Cdc2Nature Cell Biology 5:346–351.https://doi.org/10.1038/ncb954
A model for a bistable biochemical trigger of mitosisBiophysical Chemistry 57:239–251.https://doi.org/10.1016/0301-4622(95)00075-5
Ultrasensitivity in the Regulation of Cdc25C by Cdk1Molecular Cell 41:263–274.https://doi.org/10.1016/j.molcel.2011.01.012
Robust mitotic entry is ensured by a latching switchBiology Open 2:924–931.https://doi.org/10.1242/bio.20135199
Naama BarkaiReviewing Editor; Weizmann Institute of Science, Israel
In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.
Thank you for submitting your article "A robust and tunable mitotic oscillator in artificial cells" for consideration by eLife. Your article has been reviewed by two peer reviewers, and the evaluation has been overseen by Naama Barkai as the Senior and Reviewing Editor. The reviewers have opted to remain anonymous.
The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission. Please address in full all comments of reviewer #2.
Guan et al. present a highly interesting manuscript showing that a few simple components in the cell cycle machinery can generate oscillations in vitro in droplets. They showed robust oscillatory behaviors and how the oscillations can be tuned by the size of the droplets as well as by concentrations of cyclin mRNAs. The theoretical work showed that the ratio of Cdc25 and Wee1 affect the oscillation timing. They further showed that the energy depletion model is consistent with the experimental results. The experiments are elegantly carried out and the results are beautiful. I would recommend for publication in eLife.
This paper reports on the development of a droplet assay in which cytoplasmic extracts can be compartmentalized to sizes in the micrometer range. This allows the authors to generate artificial mitotic oscillations using an in vitro setup that combines the ease of manipulation and control of bulk cell-free extracts with the realism of cell-size compartmentalization (in terms of heterogeneity and sustained dynamics). The observed sustained oscillations (Figure 2B) are certainly noteworthy, and the assay does allow for a valuable study of how the period of the oscillator depends on "cell" size and on the availability of one of the clock components (cyclin B1). The setup also enables the authors to conjecture that the amplitude and period of the oscillator depend on the energy available to the cell in terms of ATP, a fact that is studied computationally but not experimentally. I would like to raise the following points about this work:
1) The oscillations shown in Figure 1C, D are really interesting, given that they exist even in the presence of nuclear self-assembly and disassembly. I was wondering if the extra compartmentalization provided by the nucleus would have an impact on the oscillations reported in Figure 2. Would the authors be able to obtain self-sustained oscillations for long periods in the presence of a nucleus? This would bring the results closer to real-life situations.
2) Figures 2F and G show that the period and lifetime of the oscillations are rather independent of "cell" sizes for large sizes. I think it'd be interesting to relate these results to real cells, by taking into account the range of sizes that real cells exhibit (keeping in mind again that these "cells" don't have a nucleus, see my preceding comment).
3) The way the modeling results are being presented makes it difficult to understand what is being done. The title of the caption of Figure 3 refers to a "stochastic model", but as far as I understand the results of Figure 3C have been obtained with an ODE model, in the absence of fluctuations, and probably those of Figure 3E too. Figure 3B, on the other hand, has been presumably obtained with a stochastic model (since this seems to be the way in which ATP can be included in the model described in the Materials and methods section), before the stochastic description is mentioned in the text. On the other hand, the time traces in the insets look very clean, seemingly obtained from an ODE model. A similar comment can be made regarding Figure 3D. Can the authors clarify what is being shown in Figure 3?
4) Figure 3C is really hard to follow. I would ask the authors to plot the steady states (at least the stable ones, and preferably all) as symbols, and to represent the trajectories in colors different from those of the corresponding nullclines.
Finally, I have two suggestions that are not essential to support the major conclusions of the paper, but would (in my opinion) enrich it:
5) I'm wondering if the range of droplet sizes that has been studied has been selected on purpose by the authors, or if there is some limitation resulting from the experimental process. I'm asking because it would be interesting, in my opinion, to see what happens with the oscillations as the size of the droplets increases. For what droplet size does the oscillation dampening typical of bulk extracts is recovered in these experiments?
6) It would also be desirable to validate the modeling results of Figure 3 experimentally by controlling experimentally the amount of available ATP and measure the corresponding effect on the oscillation characteristics.https://doi.org/10.7554/eLife.33549.029
- Allen P Liu
- Qiong Yang
- Qiong Yang
- Qiong Yang
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
We thank Madeleine Lu for constructing securin-mCherry plasmid, Lap Man Lee and Kenneth Ho for discussions about droplet generation, Neha Bidthanapally and Zheng Yang for helping image processing, Jeremy B. Chang and James E. Ferrell Jr for providing GFP-NLS construct. QY is funded by the National Science Foundation (Early CAREER Grant #1553031), the National Institutes of Health (MIRA #GM119688), and a Sloan Research Fellowship. AP is funded by the National Science Foundation (MCB #1612917).
- Naama Barkai, Reviewing Editor, Weizmann Institute of Science, Israel
© 2018, Guan et al.
This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.