A novel in silico framework for early mammalian embryo development

How animals develop from a single fertilised egg into a complex body structure made of trillions and trillions of cells is a fascinating question. A single fertilised egg must somehow divide over and over again to create many different tissue types and body parts, all in the correct arrangement.

During this first few days, before implantation, the fertilised egg repeatedly divides to make a ball-shaped structure called the blastocyst. This structure typically consists of a couple of hundred of cells and has a size of about a fifth of a millimetre.

The details of how this happens are still not well understood. In particular, despite decades of study, this area is full of unanswered questions. These include what happens if some cells are not positioned correctly and how cells communicate with each other.

Answering these kinds of question is of the upmost importance. If we understand the earliest stages of development, this will have great significance for global issues such as human fertility, IVF treatment (where most embryos fail before implantation), conservation of endangered species and food security (via animals such as pigs, cows and sheep).

Research in this area has traditionally required experimenting with large numbers (sometimes thousands) of animals, including mice, rabbits, cows, pigs and sheep. Using animals isn't needed just for studying the animals themselves, but also for studying human development. Since there are important ethical issues when using human embryos, animal models are normally used instead.

Any approach that leads to fewer animals being used would therefore be of great value. It is this vision that this project seeks to achieve. We believe that our approach holds the promise of reducing the number of animals that are used in this area of research by up 80%.

We will do this by using a combination of mathematics and computing. At first this may sound odd. Since embryo development is an area of biology, how can these non-biological subjects be of any use? However, experience has shown that combining disciplines and different ways of thinking can lead to quicker, cheaper progress, and to insight that simply could not be attained by using biology by itself.

Our idea is to design a freely-available computational tool (based on a mathematical approach) that developmental biologists can use to study the early embryo. Rather than starting with testing on animals, the biologist will instead first use our tool to investigate their question of interest.

This will narrow down options, and suggest answers to their question. This by itself may be enough to make progress. However, even if not and if it is still necessary to use animals, the insight gained from our computational tool is likely to mean that substantially fewer animals will need to be used.

There are two key design principles we will adopt, which will lead to a truly novel resource. First, we will make our tool as easy to use as possible. This will mean research groups can benefit from it even if they do not have access to mathematicians and programmers. Second, we will design a tool that works for many different animals. Changes will need to be made for each animal, but the underlying approach will remain the same.

The reason this resource has not been created until now is because of a lack of high-quality data. However, excitingly, last year, our collaborators at the University of Exeter managed to create the first realistic cellular model system for the human. This, along with data from mouse, rabbit and livestock provided by our other collaborators, means that, for the first time, we have the large quantities of data needed to make out approach a success.

Finally, to capitalise on the opportunity provided by our resource, it is important that we check it applies to humans as well as other animals. To do this, we will interact with IVF clinics and NHS Foundation Trusts who have agreed to be our project partners.

Understanding the earliest stages of embryogenesis in mammals is of great importance for boosting fertility, improving IVF success rates, conserving endangered species and ensuring global food security (via livestock such as pigs, cows and sheep).

Work in this area has traditionally required large numbers of animals, including mice, rats, rabbits, livestock and non-human primates. It is not unusual that a single large research project uses thousands of mice.

Mathematical modelling and computer simulation holds the promise of significantly reducing animal use. Modelling typically cannot reduce animal use to zero, but can quickly and cheaply identify underlying mechanisms, suggest experiments and make predictions that mean far fewer animals are needed.

A stumbling block to creating mathematical models of embryogenesis has been a lack of high-quality data. However, last year, a human embryo stem-cell model (the blastoid) was created by our collaborators. Along with data from our other collaborators on mice, rabbits and livestock, we can now build a general modelling framework of the mammalian embryo.

Based on our preliminary proof-of-principle models, we will design two classes of model: a 3D cellular Potts model (with the addition of cell division, differentiation and apoptosis) and a 2D vertex model (that also includes fluid dynamics of the blastocoel).

Models will be fit to the large imaging datasets provided by our collaborators using automated image segmentation and simulated annealing. Our IVF clinic project partners will also provide data, ensuring our framework applies equally to humans as to other animals.

Our approach will be unique in two ways. First, it will be as simple to implement as possible so that it can be used by groups without access to modellers. Second, it will work across animals. Although there is variation between animals, we believe this can be captured by changing model parameters rather than the underlying model itself.