Introduction

Complex systems are especially well suited to be studied by computer models as their behaviour depends on many variables (or parameters) which interact in a non-linear manner. What this means is that minor differences in a single variable can produce drastic effects on the system’s behaviour. At their best, computer models can produce accurate predictions on the behaviour of such systems given a set of initial conditions. Every day we rely on computer models: when we open our weather forecasting app and decide we should take the raincoat, we are trusting the results of heavily intensive computer simulations that predicted that it will rain today.

Developmental biology is a field where the use of computational modelling has become increasingly popular in recent years. This is totally justified, as developmental processes are perfect examples of complex non-linear systems: their outputs depend on multiple parameters at different levels (biochemical interactions, gene regulatory networks, cell-cell interactions, cell behaviours, physical constraints, etc) that interact in a non-linear manner and that feed back into each other. Multiple types of models are used in developmental biology, but they all share the same purpose: “to treat mechanistic ideas in a more rigorous manner than we are capable with our own minds” (Sharpe, 2017). One type of computational model is the one that allows for hypothesis testing against quantitative data (Sharpe, 2017). In this case, the researcher has a specific hypothesis from the beginning that he/she wants to test. What is needed are quantitative data on the parameters that he/she considers relevant to the model.

About the preprint

In the preprint Matsuda et al. use such a modelling approach that allows hypothesis testing against quantitative data. In their case the hypothesis put to test is the following: Can differences in the speed of biochemical reactions between mouse and human cells explain the timing difference of their segmentation clocks? In vertebrates, the segmentation clock refers to the oscillatory gene expression that regulates the timing of sequential formation of body segments (somites).

To answer this, Matsuda et al. used cultures of pre-somitic mesodermal cells (PSM) of both mouse and human, making sure that the oscillation periods (i.e., the duration of one cycle) of each species were recapitulated in vitro. Using the temporal expression pattern of the gene Hes7 to quantify the oscillation period (Fig. 1), they showed that it was ~2hrs and ~5hrs in mouse and human, recapitulating the 2-3 fold period difference observed in vivo.

After first demonstrating that the timing differences are cell-autonomous and that these do not depend on the sequence differences between the orthologous genes (see Teresa’s preLight) they focused on the intracellular network that drives gene expression oscillation (Fig. 2). This intracellular network consists of an auto-inhibitory feedback loop of HES7. This means that when HES7 is expressed, it inhibits its own expression (going from ON to OFF); when HES7 is not expressed any more, there is no inhibition so expression starts again (from OFF to ON).

Matsuda et al. used a previously proposed mathematical model (Lewis, 2003) that theoretically showed that auto-repression of a gene by its own product (like HES7) can generate oscillations, with its period determined by transcriptional and translational delays, and degradation rates. For an experimental validation of this model, they carefully quantified the relevant biochemical parameters: the degradation rate of HES7 protein, the transcription and translation delay of HES7, the delay produced by intron processing, etc.

They fed these experimentally determined biochemical parameters into the computational model, generating in this way a mouse and a human version of it that could be directly compared. The simulations showed that the mouse and human in silico periods were ~150min and ~300min, reproducing the observed 2-3 fold period difference in their in vitro experiments. This finally proved that the slower biochemical reactions of HES7 in the human PSM can indeed explain its longer oscillation period, as compared to the mouse.

Why I chose this preprint:

I liked very much this preprint because it is a remarkable example of model testing against quantitative data. The authors carefully quantified the parameters needed to build a more realistic segmentation clock model and were able to prove a hypothesis that has been previously made based on theoretical computational modelling. In doing that, Matsuda et al. built a bridge between theoretical modelling and quantitative developmental biology. Let’s hope we see more studies like this.

 

References:

Lewis, J. (2003). Autoinhibition with transcriptional delay: A simple mechanism for the zebrafish somitogenesis oscillator. Current Biology, 13(16), 1398–1408. https://doi.org/10.1016/S0960-9822(03)00534-7
Sharpe, J. (2017). Computer modeling in developmental biology: growing today, essential tomorrow. Development (Cambridge, England), 144(23), 4214–4225. https://doi.org/10.1242/dev.151274

Irepan and Teresa’s questions to the authors:

Q1: Since the size and number of somites differ between mouse and human, it would be interesting to know if the identified temporal mechanisms play a role in counting the number or measuring the size of somites. Do the authors think that the in vitro protocol in mouse and human generates the appropriate number of somites? Does the oscillation period vary over time in vitro?

Q2: For the comparative analysis between species, reporter activity has to be normalized. I wonder if the amplitude of Hes7 is different between mouse and human, or if the levels of expression are compensated between species.

Q3: Since the differences in biochemical reaction parameters determine the tempo of the species, do the authors think that each parameter is independent?

Q4: Since the time of development is species-specific, do the authors think that this has any evolutionary advantage for the species?

Q5: Have you considered perturbation analyses (e.g. introduce large introns to increase intron delay or alteration of mRNA turnover rate) as a further validation of the model?

 

doi: https://doi.org/10.1242/prelights.11311

Read preprint (No Ratings Yet)