Joshua Yarrow

The internal architecture of leaves determines their photosynthetic efficiency. In particular, the network of intercellular air spaces inside the leaf is designed to maximise gas exchange between photosynthetic mesophyll cells and the environment whilst maintaining water use efficiency and light capture capacity.

This balance is crucial for ensuring high crop yields and is becoming increasingly pertinent due to climate change. However, despite their importance, how leaf air spaces form and are patterned is poorly understood.

Air space formation and patterning

The formation of air spaces can be roughly divided into two stages: initiation by cell separation at multi-cell junctions, and subsequent expansion.

Recent work has shown that stomata promote the formation of larger air spaces beneath them, but both the molecular identity of the stomatal signal and how it alters mesophyll development to produce large air spaces are unknown.

Technical challenges

However, the investigation of air space patterning has been made difficult by two technical limitations.

Imaging early air space development

Firstly, quantifying air spaces at key stages in early leaf development is difficult with existing techniques. Staining with Nile Red in Silicone Oil allows for quantification of air spaces with confocal microscopy after stomata are open, but the initial stages of air space formation and expansion cannot be imaged this way. This means that we cannot quantitatively investigate the dynamics of early air space patterning.

Conversely, although X-Ray micro-tomography can image air spaces in mature leaves without stain infiltration, it does not offer sufficient resolution for early-stage developmental studies.

Modelling air space formation

Secondly, whilst air space formation is an emergent outcome of intercellular adhesion (and the loss of it), cell wall mechanics, cell expansion, division rates and division plane orientations, a mechanical model of air space formation that integrates these processes is currently lacking. In addition to assisting with the generation, exploration, and testing of hypotheses, such a model could serve as a predictive basis for future efforts in leaf engineering for sustainable agriculture.

Developing new tools and models

In the first part of my PhD, I focused on addressing these these technical challenges.

PoroSeg: Air space prediciton tool

Firstly, I have developed a predictive tool that labels air spaces in confocal images of cell outlines. This tool, named PoroSeg, is based on the (3D) U-Net architecture and currently outperforms field-leading software such as Cellpose-SAM at distinguishing air spaces from mesophyll cells in mature wild-type _Arabidopsis_. I have also demonstrated that its predictions can enhance the PanSeg workflow to remove air spaces from cell segmentations.

Extending tissue simulator software

Secondly, to model air space dynamics, I have extended the Jönsson lab’s Tissue simulator software to incorporate cell adhesion mechanics, collision detection, a cell division routine that is compatible with cell separation, dynamic re-meshing of high-resolution polygonal cells during wall growth, and automated air space tracking. These features provide a versatile platform for exploring a variety of hypotheses, including local effects on key parameters such as cell wall stiffness, growth rate, and adhesion strength in response to molecular signals and air space adjacency cues.

Generating realistic tissue geometries

In order to make predictions on realistic tissue geometries, I have also created tools to efficiently generate initial conditions from segmentations of confocal images.

With these tools in hand, the current phase of my project is focused on investigating the roles of mesophyll cell adhesion and division in sub-stomatal cavity formation.