5.5.3 Achaete Scute

1. Lateral Inhibition

The Achaete-Scute gene system is one of the basic example systems that is taught in any developmental biology classes. It is very important for several reasons. First, it shows how a very simple system with just a few genes can cause some very complex patterning using feedback systems. Second, it is one of the most basic systems for patterning that all animals share. Third, the type of behavior it displays is very important. This gene system is one that performs lateral inhibition. Lateral inhibition is the process whereby the cells within a cluster compete so that one cell becomes dominant and inhibits the others. One good example of where something like this is used in the developing embryo is in Drosophila. The two figures below will be helpful in explaining this example. And there are two important terms that you should know about. Drosophila is a fly that is commonly used in numerous different kinds of biological experiments. Neuroblasts are the neural progenitor cells for the fly and they go on to create the nervous system. Don't worry about the term ectoderm. It is just a label for certain types of cells within the developing embryo.

Drosophila Neuroblast Description
Figure 1. The neuroblasts of the Drosophila separate from the ectoderm by a process known as delamination (red). The neuroblasts enlarge relative to the surrounding cells and squeeze out of the epithelium. The second process occurs in several waves; after the first set of neuroblasts has delaminated from the ectoderm, a second set of cells in the ectoderm begins to enlarge and also delaminates. 5.5.3.1

Drosophila Neuroblast Segregation
Figure 2. Neuroblast segregation in Drosophila neurogenic region proceeds in a highly patterned array. Just prior to neuroblast delamination (A) a group of ectodermal cells expresses a proneural gene. One of the cells in this group maintains its expression of the proneural gene product, and that cell becomes the neuroblast (B). 5.5.3.2

As you can see in figure 1 these neuroblast cells form in a nice grid like pattern. The question then arises of how do they form in a grid like this? This is where the achaete-scute gene system and lateral inhibition comes into play.

Achaete-Scute System
Notch: This is a membrane receptor that is expressed by all of the cells in the cluster. When Notch is activated it causes the production of SuH.
Delta: This is a membrane ligand that binds with Notch. It is also initially found in all of the cells of the cluster. It is the Notch/Delta communication that allows the cells to compete.
Achate-Scute (Asc): This is the primary gene involved in this system. When it is active it is responsible for up-regulating both itself and the production of more Delta ligand. SuH indirectly inhibits this gene.
Suppressor of Hairless (SuH): This is a transcription factor that ultimately suppresses the Achaete-Scute genes. However, it does this indirectly by activating another protein that then inhibits the gene. In the emulated system this indirect step is left out, and Suh directly suppresses the Asc gene.
Table 1. These are the proteins and genes that make up the Achaete-Scute gene system.

Achate-Scute Description
Figure 3. The lateral inhibitory mechanism involves the neurogenic genes Notch and Delta. In (A), a cell expresses achaete scute genes (ASC) and Notch is inactive. The ASC activates its own transcription to maintain its expression and also activates the expression of downstream neural-specific genes and the Notch ligand Delta. This cell would then become a neuroblast; however, if a neighboring cell expresses more Delta (B), this activates the Notch pathway in the first cell. The activated Notch leads to a release of Suppressor of hairless (SuH) which along with part of the intracellular domain of Notch diffuses to the nucleus and acts as a transcriptional factor on another gene's enhancer, the enhancer of split complex (E(spl)) (C). The E(spl) proteins are repressors of ASC gene transcription and so they block further neural differentiation and reduce the levels of Delta expression. Thus, this cell is suppressed from the neural fate because of Notch activation by a neighboring cell that expressed more Delta.5.5.3.3

Okay, lets try and go back over that explanation above and put it in very simple terms. When the cells in the cluster start out all of their genes are active. This means the ASC genes are keeping themselves active and they are pumping out the membrane ligand Delta. The membrane receptor Notch is always being produced and that does not change throughout this whole process. Each of the cells in the cluster may be producing Delta at slightly different rates however. And when the Delta of one cell mates with Notch receptor of another cell it produces a transcription factor (SuH) that will attempt to down regulate the ASC genes. This would in turn reduce the amount of Delta ligand produced in that cell. So if one cell is a little better at producing Delta ligand it can cause its neighbors to stop producing Delta and shut off the ASC genes. This is the essence of lateral inhibition. One cell in a cluster has become dominant and suppressed a behavior in its neighboring cells. As we will see shortly this is also a very good process by which to create grid like patterns.

Lateral Inhibition
Figure 4. This shows how the central cell in the cluster becomes dominant and suppresses the ASC genes in its neighboring cells. This allows that cell to then become the neuroblast.5.5.3.1

2. Evolving The Chromosome

The goal of this experiment was to develop a set of genes that would emulate the behavior that is displayed by the real system. This system was chosen for several reasons.

  1. This is a relatively simple system with only a few genes and proteins. This makes it straightforward to implement.
  2. However, even though it is simple structurally it still displays some pretty complex behavior.
  3. The relationship between each of the proteins and genes are known. But the specific details like the exact expression function, degrade rates and so on are not know. This makes using genetic algorithms to solve the problem extremely simple.

As mentioned above the relationships between the different items are known. For example, we know that Notch is a membrane receptor and Delta is its corresponding membrane ligand. However, what is not apparent is what settings are needed on the receptor to make it function with the correct behavior. This is not a simple relationship. I sat down and played with different combinations for several hours just to get an idea of how hard it would be and never came up with a system that produced the desired behavior. This is not a trivial system. Therefore, a genetic algorithm was used to try and find genes with values that would produce the desired results. An initial population of 1000 chromosomes was used. That population was produced using templates. The template allowed me to define the types of proteins required and the relationships between them, but all of the explicit details like the expression function values and degrade rates were produced randomly. A number of fitness tests were then used to evaluate the fitness of each chromosome. After a certain period of processing time the protein and gene expression rates of the cells were checked for each of the major items in this system to see if a grid pattern was formed. So for example ASC genes should have high expression rates in one cell and lower or no expression in its neighbors, and so on. The better the chromosome was at producing this pattern the better its fitness and the more likely it was to be reproduced into the next generation. It took 52 generations to produce the best of chromosome that is displayed here.

3. Fittest Chromosome

Best Evolved Chromosome
Figure 5. This graph shows the details of the evolved achaete scute chromosome. Click on the image to see the full analysis screen.

As you can see in figure 5 the basic relationships between the proteins and genes are exactly the same as that in the model of how the achaete scute gene system works. When you look at the full analysis of the system you can see the values for the expression functions and degrade rates that were evolved for this chromosome and produce the behavior that will be discussed below. One thing to not however is that the name used for the ASC genes here is Ach.

Constant Achaete Scute Example
Video 1. This video demonstrates the achaete scute system when all of the cells have the same starting values.

Random Achaete Scute Example 1
Video 2. This video demonstrates the achaete scute system when the value of Ach is random.

Random Achaete Scute Example 2
Video 3. This video shows how having random values for Ach produces a different resulting grid pattern than that seen from the previous video.

Some interesting observations about the above examples can be noted. First, even if the protein values of all of the parts are equal between cells lateral inhibition still occurs. Furthermore, in this instance you actually see waves radiating inward of grid formation that are quite spectacular to look at. Second, when the Ach protein is randomly distributed in the cluster of cells you do not get the perfectly neat grid distribution seen in the first example, but grid formation and lateral inhibition do still occur. In fact, it occurs much more quickly. The wave action seen in Video 1 is not seen in the other two videos. Also, as videos 2 and 3 proceed it is possible to clearly see the competition between groups of cells occurring. And finally, since the Ach is distributed randomly it produces different grid patterns between the two. This makes since because different cells would have higher concentrations of Ach from one example to the other. This would result in different initial cells being dominant early, and that would in turn change the resulting pattern of future cells that will be dominant.

4. How Does It Work?

Cellular Graphs
Figure 6. These graphs show the protein quantities and gene expression rates in four key cells from the constant Ach example above.

The four graphs shown above display the protein quantities and gene expression rates of some important cells from the constant Ach example above. In that example the cell (0, 0, 0) (A) is going to be dominant because it is one of the corner cells and has fewer neighbors, and because it is the first to be evaluated. Two of the other graphs are for its immediate neighboring cells (1, 0, 0) (B) and (1, 1, 0) (C). The fourth graph shows the next dominant cell (2, 0, 0) (D). Lets look at the Ach protein level in cells A and B. Until right before 0.2 seconds the quantity for both cells is rising, and they both then begin falling. However, the value for B never reaches quite as high as it does for the other cell. This means it also reaches zero faster and therefore more Delta is produced in A and more Suh is produced in the neighbors. At its peak the level of Suh in A is 530 and in B is 559. So when they begin going back up again a little later B and C are at a disadvantage because A is already suppressing them. And then at around 0.9 seconds the full suppression kicks in and the cells completely diverge and stay that way. B and C express Suh and stop expressing Delta and Ach. And cell A expresses Delta and Ach and stops producing Suh. Cell D also goes through this same process like it was one of A's neighbors. However, once A suppresses the expression of Delta in B and C this allows the exact same process to occur in D. This is what is causing that wave like motion that produces the grid. This wave is not seen in the random example because there are already some cells that have higher levels of Ach than others. Therefore they immediately being suppressing their neighbors.

5. Conclusion

There were three main objectives in this experiment.

  1. Prove that this developmental simulator was capable of reproducing behavior similar to that seen in nature. It is true that this is obviously not exactly how it really happens. But the results are quite similar. And for my purposes I simply need something that can produce the similar results using the same basic principles that nature uses.
  2. Prove that the genetic algorithms can work to fill in the blanks where our knowledge is missing. It would be very difficult to simply come up with values for the proteins that would produce this behavior. But evolution was able to find values that worked in a reasonable amount of time. However, since I used templates to produce the chromosomes this still does not demonstrate that the genetic algorithms could have evolved the thing completely from scratch. If you had no idea of even the basic genes and proteins involved or how they interacted then could this system be used to find the solution? That remains to be answered.
  3. This is the first of the genetic modules that will later be used to build more and more complex systems. All that is now needed to be able to form this checkerboard pattern in any cluster of cells is to turn on the genes in this module. They will then do their work and by turning on other sets of genes the outcome of the grid could then cause further differentiation events that would be very similar to turning the dominant cells into neuroblasts like in Drosophila.


<< Previous Contents Home Next >>

MindCreators.Com is edited and maintained by David Cofer. If you have any questions, comments, or just want to discuss the contents of this website, then email me at: dcofer@MindCreators.com.

Copyright © 2003 by David Cofer. All rights reserved.