Lock the edges, not too !

Listen to audio

This law can be expressed as follows. Initially we have 2 morphogens, for example dorsal and ventral like BMP and Shh in the nerve tube, more generally, A and B. A induces HA and B induces HB, both homeogens, but HA represses HB and HB represses HA. This can result in the expression of 2 gradients, an HA gradient and an HB gradient. These two transcription factors may be transiently expressed in the same cells, but because they are reciprocal self-activators and inhibitors (our general law), the winner takes all and, barring an additional regulatory element allowing co-existence, an edge is formed. It's quite clear here that this edge admits a certain variability due to fluctuations, unless we invoke an element ensuring greater robustness.

But we know how to form a pattern with a single morphogen. We have even admitted that, in the case of the nerve tube, the invention of the second morphogen (Shh) responds to an enlargement of the structure that renders the first morphogen (BMP) inoperative because it cannot diffuse far enough. In such a case, how can we make an edge except to imagine that the gradient of A translates into the expression of HA1, HA2, HA3 homeogens (class A), as in Wolpert's French flag model, i.e. with thresholds? But thresholds admit a certain level of variability which means that the edge can only be sharp if we add an additional hypothesis. As in the previous case, we propose that even activates even, odd activates odd and that even and odd inhibit each other. The previous case then becomes a variant of the second case, with class A and class B.

Before exploring this model further, we reminded you that the expression of homeogenes, or other developmental genes, is not enough to form edges; you obviously need targets to do the work. Ephrins and their Eph receptors, for example, but also other effectors, adhesion molecules (NCAM, cadherins) or the Notch-Delta system, and others. If you think about it, the very nature of these factors - all of which, including adhesion molecules, are signaling molecules - means that the edges aren't dead. In fact, that's where things happen, where "things move".

Homeoproteins, as demonstrated in our laboratory, are molecules that transduce a signal precisely at the edge. We can extend this notion of edge, to propose that it applies to any contact between cells. And these exchanges - including homeoproteins - that we have explored in the early stages of nervous system formation are, in my view, probably functional at the cellular level, particularly at the synapse, one edge among others.

These proposals are not without evolutionary consequences, since these homeoproteins, expressed in all multicellular organisms from the very beginning, are very probably - at least this is my hypothesis - the oldest of the signaling molecules, and not very robust, the robustness having resulted from the invention of other, more classic molecules.

But if homeoproteins are the first signalling molecules, how did they - without effector proteins - exert their activity? We propose that the first signalling involved mitochondrial activity and homeoprotein-regulated ATP synthesis, capable of influencing the state of the cytoskeleton and the localization of surface molecules. This is a parsimonious scheme involving very few classes of molecules, molecules whose existence we know necessarily preceded the invention of multicellularity, including homeoproteins whose primary function might have been to regulate conjugation in unicellular organisms.

To take a step back, we have compared the classical morphogen model with the one we propose, based on the intercellular transfer of homeoproteins. In the classical model, a morphogen diffuses from a source and induces the expression, in a gradient (or not) of two factors (not necessarily transcription factors) which inhibit each other. At the level where the two factors meet, their reciprocal inhibition leads to the formation of an edge, but this edge is initially irregular. If they are transcription factors, they must be transiently co-expressed in the same cells. Regularization requires either cell death, identity change or migration.

The second model we have proposed is that morphogens do not diffuse enormously, but induce a homeogen. The homeoprotein passes from cell to cell and is self-induced, so it can be considered "infectious". It then encounters another homeoprotein, which comes in the opposite direction and has the same self-inducing property, but the two are reciprocally inhibitory, hence the formation of an edge.

We then compared the two models by reinterpreting the data in the literature, particularly those concerning the position of edges and therefore the surface area of territories in the central nervous system, within the framework of classical hypotheses and those we have recently put forward on the basis of our own observations.