A Paper Review

This web page was produced as an assignment for an undergraduate course at Davidson College.

Analysis of the eye developmental pathway in Drosophila using DNA microarrays.


Michaut L, Flister S, Neeb M, White KP, Certa U, Gehring WJ.

Proc Natl Acad Sci U S A. 2003 Apr 1;100(7):4024-9



This paper, published in 2003, is an application of an experiment that was done quite a few years earlier. In 1995 Walter Gehring’s group at the University of Basel performed a search for fly genes that coded for transcription factors. One of these factors mapped to the gene that was previously named eyeless, or ey. Through multiple experiments outlined in the paper (Halder et al., 1995), the investigators showed that the ey gene could be considered a “master control gene”, in that its ectopic expression initiated a cascade that eventually produced eye structures in various tissues. For instance, by initiating the transcription of the ey gene in a “leg disc” (a section of fly larvae that will turn into legs), a fly was born that had an eye-outgrowth on its leg. This was a classic use of “reverse genetics,” where the researchers started with a gene, ey, and proceeded to figure out its impact.

In the most recent paper from this group, Michaut, et al. describe an application of an increasingly important genomic method in conjunction with their previous work on the ey gene. The investigators use two complete Drosophila genome microarrays and label leg disc RNA from two different fly types: a wild type fly and a fly that is ectopically expressing the ey gene in its leg disc. The authors also looked at the wild type expression of normal Drosophila eyes, which they could then compare to what genes ey specifically induced. Through an analysis of these data, the researchers hope to reveal the RNA from certain fly genes when the ey gene is expressed and, in turn, uncover genes that weren’t previously associated with eye development in an attempt to “gain an overview of the entire genetic cascade controlling eye morphogenesis” (Michaut, et al., 2003).

Critique and Analysis of the Paper’s Results

There are only two figures and two tables in this paper and, as such, do not necessarily drive home the group’s data. However, it is important to examine each of these presentations of data to determine the effectiveness of Michaut, et al.’s work. Figure 1A shows a Venn diagram of the two genome chips (“probes” is the term the authors use) to highlight the 55 individual genes that were detected by both microarray probes. Genes were considered “ey-induced” when they were expressed in eye discs, had an induction of at least 1.5 fold between leg discs that had ey ectopic expression and wt leg discs, and had a 95% confidence interval for this last piece of data. By making sure that there was a level of expression in eye imaginal discs, the authors add a level of control to their data – genes that may be induced by ey ectopic expression that don’t deal directly with eye formation are ignored. These genes aren’t the focus of this paper.

Interestingly, an enormous amount of genes (87%) were determined to be “ey-induced” by only one of the two microarrays. Initially this is a little disturbing, but it is important to remember that the DrosGenome1 array “is based on a later release of the genome annotation and contains different oligonucleotide probes” than the roDROMEGa array. Also, the text states that the researchers also took into account the P-value when determining “ey-induced genes.” The authors mention that the percentage of cross-detection would’ve been much higher (around 60%) if this wasn’t taken into account. Still, it is a little disconcerting that such a high percentage of genes were only found in one of the two microarrays. The investigators make a generally correct interpretation when they mention that “these discrepancies probably reflect differences between the two microarrays”.
In keeping with this discussion of microarray discrepancies, the authors go on to point out several genes (such as sine oculis or cyclin E) that are indicated by one microarray as “ey-induced” while not falling into this distinction on the other microarray. The authors can’t conclude that one microarray is more sensitive than the other, but do conclude that by using a combination of the two they validate their results and reduce the number of false-positives and outliers. An important objection to the use of the two microarrays would be that the authors might be excluding some genes unnecessarily with their criteria for a gene being categorized as “ey-induced”. Even so, it is important to realize that these authors never claim that their experiment is an attempt at showing all of the genes associated with eye development and ey expression. Instead, they are striving to clean up their data to create a small subset of genes that are more likely to be induced by ey than genes that are included under less stringent standards.
Figure 1B is simply a functional classification of the 254 known genes that the microarrays picked up. The text never really mentions it, but I guess it’s somewhat informative. There were a high number of transcription factors initiated by ey, which is itself a Pax6 transcription factor, which is an interesting phenomenon. This could be further proof that ey is a “master control gene,” but that’s my interpretation, not theirs.

Table 1 is an expansion of the 55 genes found to be ey-induced in both microarrays (the combination of Figure 1’s repartition). Gene names highlighted on the left indicate that these genes were detected in (wild type) eye disc libraries by SAGE (serial analysis of gene expression) tags. This would give further indication that these genes are indeed related to eye morphogenesis. The colors under the two types of Drosophila microarrays show three levels of fold induction for these genes in leg discs where ey is ectopically expressed. As was mentioned earlier, a level of 1.5 fold induction is a requirement for both genome microarrays, and this is the minimum induction seen in these 55 genes.

Table 2 is interesting because it is very similar to the previous table but it shows genes that aren’t in the 55 that fit the criteria for both microarray genomes. Again, highlighted gene names were found in eye disc libraries using SAGE tags and the three levels of fold induction are highlighted with the same three colors. Now, however, four more columns for each microarray are added. Although the explanations leave something to be desired, it can be deciphered that AD LDc is the signal of wild type leg discs, LDey is the signal of ey-expressing leg discs, P is the P-value representing percent confidence intervals, and AD_ED is the average difference between the eye imaginal discs and, apparently, leg imaginal discs. When P-value was over 0.104, an AD_ED value was under 100, or the fold induction was below 1.5, the gene in question wasn’t considered “ey-induced” for that particular fly genome microarray.

Most of these genes apparently had one or two pieces of criteria that excluded them, and most often this was their P-value or AD value (average difference, which the authors used as an indication of expression in wild type imaginal eye discs). Of course, if the particular gene in question isn’t found on one of the microarrays, it is impossible for it to be included in the overlapping section of genes that fit the criteria for both microarrays. The authors continually note this in the paper. As a point of criticism, it isn’t immediately obvious why some of these genes aren’t included in Table 1. Although some numbers are highlighted in different colors, there is no mention of this in the figure legend. Also, some genes are mentioned in the text as one gene name (such as stich1) and then represented in the paper as another (sticky ch1).

Several genes are mentioned in the text at this point, and these data and their interpretations turn out to be very insightful. The authors first discuss that the ey-induced genes function early in the retinal differentiation pathway before going on and mentioning both known genes and previously uncharacterized genes that were highlighted by this microarray experiment. This is probably the most important part of the data, in that it provides the basis for the characterization of new genes involved in the pathway.

The last data representation, Figure 2, shows select ey-induced genes and their endogenous expression in wild-type eye discs. The quail gene was among the 55 represented in Table 1, and the quail protein was found in the eye discs using immunofluorescence with the 689 mAb. In examining actual gene RNA presence, the authors used antisense RNA probes in situ to examine the expression of genes they previously discussed. The authors chose to show the labeling of 4 genes that were found in the 55 genes that met the criteria for the ey-induced classification for both genome microarrays and 9 that did not meet the criteria. Although I am not an expert at understanding figures of this nature, it appears that these genes presented here are indeed expressed in the eye discs at this time in development. Certain genes are expressed in certain places (such as skeletor), and the authors do a fairly good job at going through the highlights of these genes.

Overall, these authors should be commended for using microarrays in the correct manner. Individual numbers and are only slightly looked at, and they follow up their preliminary data with physical experiments. By doing this, the authors are protecting themselves – they have a high probability of achieving a successful result (finding genes that are related to eye morphogenesis) without making claims that could be unsubstantiated. Although the data presentation could’ve been better, they get their point across, and definitely sell the reader on the advantages of using microarrays for experiments of this nature.

Future Experiments

The authors have a good start here to reach their goal to determine the details of fly eye morphogenesis. To elucidate these results, I’d like to see the investigators first give us a different view of the data. Table 2 clearly shows that some genes were excluded unnecessarily, and a correction for this could potentially identify genes the authors don’t mention in this paper. It is encouraging that they didn’t only consider Table 1 for the labeling method in Figure 2, but it would be helpful to see a different grouping of the genes identified by both microarrays.

In order to truly examine the nature of the eye development pathway, the authors are going to have to really dive in and examine some gene interactions that way. To do this, I would recommend the pulse-chase labeling method. Unfortunately, however, a large number of the genes found through this preliminary study were transcription factors, and only directly interact with DNA, not other proteins. Using the pulse-chase method may lead to interesting results, however, especially if they were lucky enough to pick a gene whose protein interacted with others in the eye development pathway.

Ideally, these researchers would be able to explore each of the unknown genes in great depth – each gene and gene transcript would need to be sequenced and its structure could also be examined. Although it could be time consuming and tedious, it would be worthwhile to create knockout flies with each of these genes missing to determine their exact function. With an increased knowledge of these unknown genes, the investigators may be able to piece together the specific cascade that eventually leads to ectopic eye formation. Also, it would be very important to find out which genes the ey transcription factor turned on directly. This would be difficult, as the bond between the transcription factor and the DNA that is to be transcribed isn’t strong enough to come out in an immunoprecipitation. However, by stopping ey expression at certain time periods, a more complete view of the time component of the cascade could be determined. In a sense, the investigators increased their workload considerably through this application of microarrays – the number of genes they selected for using their strict criteria is probably just a subset of the total genes in the eye development pathway. To completely determine the role of all of the genes involved in this area of research, several more microarray experiments could be done using other genes; individual genes could be turned on and off and a measure of the resulting genes could be determined. In this manner, a greater understanding of the individual connections could be reached, and the complete nature of the eye development pathway would be closer to being solved.


Molecular Biology Main Page

Biology Home Page

Davidson College Home Page

Search Davidson

© Copyright 2003 Department of Biology, Davidson College, Davidson, NC 28035
Send comments, questions, and suggestions to: dapierce@davidson.edu