A Tridimensional Atlas of the Developing Human Head

Featured here is work originating from the research group of Alain Chédotal a member of the HDBI Blood and Immune development theme.

Alain’s group, at the Vision Institute in Paris, has expertise in amazing imaging techniques showcased in the recent paper we are featuring here. Their membership in HDBI has allowed other researchers in the consortium to travel to Alain’s lab, train with his team to apply these methods to their own work, as well as make use of the facilities and resources there.

This work used cutting edge microscopy techniques to provide perhaps the most significant update to the characterisation of the developmental anatomy of the Human head in 75 years. Generating new insights in to the timing of these processes and their variability between individuals. Using innovative virtual reality methods to provide annotated interactive 3D models anyone can explore.

DOI: 10.1016/j.cell.2023.11.013

Why make a 3D atlas of the developing Human head?

Much of our current understanding of the anatomy of development of the human head dates from studies carried out in 1905 and 1948. New methods have now allowed us to paint a much more detailed picture of this process. The head is our most intricate bodily structure and houses our primary sensory organs. Understanding how this happens is of great scientific interest. Head malformations occur in about 1/700 live births so understanding head development better may let us intervene where this presents health issues. Head development is relatively well described in animal models, but these do not give the complete picture in Humans that is needed for clinical applications.

The development of the brain has been well studied elsewhere so the focus of this work was on other major aspects of the head, such as blood vessels, muscles, cartilage, nerves, and glands.

How do you label the different tissues?

To label head cells and organs (structures) you take an antibody that is known to bind a protein that is distinctive of the sort of tissue that you want to label. Antibodies are proteins produced by the immune system which stick to specific proteins, you can co-opt this feature to use them to label stuff you are interested in. So that the antibody can be detected inside the whole head a chemical structure is attached to it called a fluorophore. Fluorophores absorb light and then re-emit it, they do this in distinctive colours which you can detect using a specialised type of microscope.

Next you treat the sample (head, embryo) using a chemical method which makes them transparent to light. These are called ‘cleared tissue mounts’ In order to see specific structures in the now clear samples you have to image them.

To recap. You pick antibodies which will bind to proteins which are distinctive of the features that you want to image, each with a different fluorophore. You make your sample transparent. That way you can see each feature in a different colour using a special type of microscope. A microscope which can shine and detect light of the right colours for fluorophores attached to your antibodies.

How do you construct a 3D image of these tissue samples?

A ‘light sheet fluorescence microscope’ lets you take an image from the middle of a piece of cleared tissue and see the features that you labelled. It does this by shining a focused laser through the sample from the side - creating a ‘sheet of light’. You focus your microscope on this plane of the sample that is illuminated from the side so that you can detect the light coming from this ‘layer’ in the sample. The light sheet contains light of the right colour to be absorbed by the labels and the detector senses light of the colour emitted by the labels. By separating out the colours you can see the differently labelled features.

So to build a 3D image you move the sample through the light sheet in small steps taking an image each time. This gives you many layers, as though you had sliced up the sample and imaged each slice separately. This way you can get thinner ‘slices’ and not disrupt the delicate structure. In total this work generated almost 70TB of images representing more than half a billion ‘slices’ less than 4μm (millionths of a meter) thick.

Once you have all these images you can combine them using software to stack up all those layers and show which features were labelled in which colour in their intact 3D position/location. Virtual reality (VR) lets you look at these models in true 3D making it easier to see the depth of features than with a 2D rendering on a screen.

You can only label so many different features at the same time, and not every feature has a unique set of proteins that you can label so some manual annotation is needed based on the anatomy of the labelled features. For example if you label for ‘muscle’ there are multiple different muscles which you can see are in different places. To identify and label them individually you ‘paint’ them in VR and give the pained volume a name in the computer model of the tissue.

Video S2. Illustration of the 3D image processing pipeline.
Copyright 2023 under CC BY-NC-ND 4.0

What embryos and fetuses were imaged and where did they come from?

Samples from 27 embryos (5.5-7.9 PCW) and 49 fetuses (8.0-13 PCW) were imaged, 76 in total. PCW (Post Conceptual Weeks) is the age in weeks following conception. These were donated to the French Human Developmental Cell Atlas (HuDeCA) biobank, which is funded by the French National Institute for Health (Inserm).

You can find out more about this on their website and about HDBI research tissue and ethics on our page and in our FAQ.

New insights

All head muscles are likely present by the end of the first trimester, earlier than previously reported.

There are limits to the number of different fluorescent labels that you can use at the same time before their colours overlap too much and they cannot be distinguished. To overcome this you can do multiple rounds of staining. You chemically bleach the fluorophores used on the previous antibodies so that the same or similar colours can be reused with another set of antibodies. This work used new ‘conjugated antibody’ systems where the fluorophore is connected directly to the primary antibody. Conventionally the fluorophore was on a secondary antibody which binds the first one, for reasons including that this could amplify the signal when multiple secondary antibodies bind the primary. They used both approaches and demonstrated that this new approach of using multiple rounds staining with conjugated antibodies, developed in 2D systems, is possible in cleared 3D tissue mounts.

‘branching morphogenesis’ is the formation of new structures through branching. It is a process that happens in many structures in the body, such as salivary glands and lungs. It has been difficult to study at the molecular level with 2D images. This system permits molecular - protein - level information in 3D allowing much better characterisation of systems which form in this way. Collaboration within HDBI between groups with expertise in mathematical modelling and different organ systems which undergo branching morphogenesis is revealing new insights into this process, how it’s regulated and how it differs between systems.

Explore the data yourself!

Data visualised using Verge3D

(This may take a moment to load)

Interactive 3D reconstruction of the head arteries

Human embryo at 7 weeks after conception, Stained for smooth muscle actin (SMA+) characteristic to arterial vessel walls

Interactive 3D reconstruction of the skeleton

Human embryo at 7 weeks after conception, Stained for collagen 2 (Col2+) a major component of the cartilage matrix.

Where Can I Learn More?

You can read the full paper here

Find out more about Alain’s group and their latest work at their website

For more from HuDeCa checkout their website

For more from the blood and immune theme checkout their page on this site

Lay Summary By Richard J. Acton, Edited by Alain Chédotal

This work was primarily funded by INSERM