5.7 Next Steps
1. Forming Patterns.
The first part of this project has just been to get something that can form patterns of differentiated cells. Now that we can segment up the tissue and get cells that have differential gene expression we can have different types of neurons. Each of these neurons will need to eventually form synapses in different parts of the developing brain. To do this will need to introduce growth cones.
2. Growth Cones and Axonal Guidance.
The next phase of the developmental system that needs to be added is to allow the formation of growth cones and the guidance of the axon to the target neuron. Now that the we have the ability to split the tissue into segments of differentiated cells, each of the potential neurons needs to find its target that it will eventually form synapses on. To do this the concept of a growth cone needs to be introduced. A growth cone is an extension of a neuron that literally crawls outward from the main body of the neuron using little tentacles called filopodia. The filopodia samples chemicals on the nearby neurons and extracelluar spaces looking for chemicals that it finds attractive or repellent. If they smell a chemical they like then they will crawl the growth cone towards the higher concentration of that chemical. But if they smell a chemical they find odious they will crawl away from it. This is the basic mechanism that neurons use to find the general area where they will find their target neuron. It is like using a map to find the correct block where your destination is. Once you get there though you will have to look around to find your exact target.
In the simulator each cell will have one virtual growth cone that can move away from its parent and rest on a neighboring cell. Several new protein types will need to be added. The first new protein will control the speed of movement of the growth cone. The protein will have an expression function and the active quantity of the protein in the cell will be fed into the expression function. That function will translate into a speed. It will be possible to have a number of these speed control proteins that work together to regulate the overall speed. Just like transcription factors, you could have one speed protein that tries to make the growth cone go faster, and another that stops it altogether. It will be the combination of the quantity of active protein present and the expression functions that will determine how fast the growth cone moves, when it begins moving, and when it stops. The growth cone will typically not begin moving until the cell has differentiated to some degree. It will be this differentiation that will start the production of speed control proteins and kick the growth cone into motion. When the cone finally finds its target a new stop protein will be produced in large quantities to halt the cone so that it stays on the target.
The next set of new proteins to be added will involved the guidance of the growth cone. This will require some new receptor / ligand proteins. These will be similar to the standard membrane receptors and ligands already described, but with a slight difference. The cell that has the growth cone will produce several of these new type of receptors. Other cells will produce the new guidance ligands and will lay them down in paths of increasing concentration gradients like the one shown in figure 1. There will be numerous of these paths that will have to be laid down in different areas so each section of neurons will be able to find the general location where they need to look to find their target. At each time slice the growth cone will sample a few of the cells that are around its current location. The guidance receptors that it has will be mated with the guidance ligands on these other cells to get an overall attractiveness value. Once the selected cells have all been sampled then the one that was most attractive will determine the direction of growth for the cone, and the speed control proteins will control the distance traveled in the chosen direction. Once the growth cones reach the general area where their target is located they will slow down and start doing a more detailed search looking for specific combinations of membrane ligands. Once they find the target these ligands will activate receptors that produce speed control proteins that stop the movement of the growth cone and initiate formation of synapses.
3. Synapse Formation.
When the growth cone finds its target neuron it produces proteins that initiate synapse formation. This will involved several things. First, when the cell originally differentiated into a specific type of neuron it will have begun producing some more new types of proteins. One of these new proteins will be what decides the arborization pattern at the end of the axon. The protein will have a parameter that tells which direction it effects (x, y, or z) and it will have an expression function. The overall quantity of that protein in the cell will be fed into the expression function to determine the extent of arborization in that direction using the target neuron as the center. This will be done for all three dimensions and all neurons within that arborization rectangle will get one synapse with this neuron.
The formation of the synapses themselves will be somewhat tricky. In the past there was only one location for all the proteins in the cell body and they were treated the same always. But now this will no longer be true. Each synapse and dendrite will have its own local quantities of proteins that are separate from the main cell body. This is necessary so that each synapse and dendrite can have different connection strengths. Also, we will have to take into consideration that fact that a synapse can also form on other synapses, further adding to the complications.
4. Functioning Neurons.
At this point I have still not worked through a lot of the details for things and I am just throwing out my thoughts about how this stuff will work. We are finally to the point where we can get functioning neurons. Each different type of neuron will produce different levels of proteins that are important for it. These proteins will be transported to the different synapses. Some of the proteins that will be transported to the synapses and dendrites will a new type of ligand and receptor that are involved in the actual firing of the neuron. The receptor will be transported to the dendrites of the cell and when a ligand binds it will relate this binding value with the firing rate of the presynaptic neuron to determine the amount of current entering the cell, like an Ach receptor. Other types of ligands would use the binding value, the firing rate, and the membrane potential of the cell to determine the amount of current entering the cell, like a NMDA receptor. At this point the cell will use the same model for the functioning of a neuron as the one used in the insect simulator example. The difference being that the connection weight between neurons will be dynamic and dependent on the different types and quantities of receptors and ligands in the dendrites and presynaptic terminals. And the levels of these proteins will be controlled by the gene expression in the cell. With this it will be possible to generate mechanisms similar to those that we know are involved in learning in real organisms. Short term learning will come from local and temporary changes in the presynaptic and postsynaptic terminals themselves, long term learning will come from alterations in the gene expression of that cell.
This is a bold set of goals. It may not be possible to do it with the current level of technology available. However, I still intend to try. If I manage to succeed then this will be the first artificial nervous system that I am aware of that will integrate the cellular components of gene expression to control connection strengths of neurons, and that uses gene expression in general to grow the neurons in the first place. I think this approach has tremendous potential to produce truly intelligent and adaptive systems and I look forward to the challenge of trying to make it work.