This page was created as an undergraduate assignment for a molecular biology course.


Review Paper:

Michaut L, Flister S, Neeb M, White KP, Certa U, Gehring WJ. 2003. Analysis of the eye developmental pathway in Drosophila using DNA microarrays. Proc. Natl. Acad. Sci. USA. 100, 4024-4029.

Presented by Isaac Miller

Intentions of the Authors

In the above referenced paper, the authors set out to describe results and evaluate the limitations of a high throughput gene-chip method designed to indicate genes of Drosophila upregulated in the differentiation of eye tissue from the ectopic expression of eyeless (ey). The authors state that they were attempting to gain an "overview" (4024) of the series of gene activations that control eye development, and they did not attempt to exhaustively catalog every gene involved. They chose to study the early developmental stage of eye formation, and chose to use tissue from Drosophila at the end of the third larval stage, when the cells that will become eyes begin to differentiate into photoreceptors. The intent of the authors is to use DNA microarrays as a quick method to scan the entire Drosophila genome in order to pick out genes involved in eye formation.

Experimental Method

The authors began with two strains of flies. Their wild-type strain is one with an isogenized fourth chromosome, called iso4BS. The other strain, dppblink-Gal4, UAS-GAL4/UAS-ey, has the GAL4 gene linked to a leg imaginal disc enhancer. The GAL4 product formed in that region causes the expression of more GAL4 (to amplify the effect), along with the ey gene. This results, in full grown flies, in fully formed eyes grown on the legs.

In this paper, about 200 leg imaginal discs of each strain were removed by dissection at the end of the third larval stage. Total RNA was extracted from the cells and the antisense strand was formed, biotinylated so it could be detected, and amplified. The RNA was allowed to hybridize with two separate gene chips containing the entire Drosophila genome (roDROMEGa and DrosGenome1), and the chips were analyzed using special software. The signal was reported as average difference (AD) from the signal of probes identical but for one base change in the middle of the probes. A signal of AD = 100 was considered to indicate expression of the gene.

For each gene, the level of expression in the wt strain was considered a base level relative to which upregulation was measured. Thus, a gene that is normally expressed in, say, growing cells that is not involved in eye formation would not be accidentally identified. For each gene on each chip, the AD signal for the mutant flies was compared to that for the wt flies. However, since ey is involved in more processes than just eye formation, it was then necessary to determine those genes involved in eye formation from non-involved genes also upregulated by ey. Larval eye imaginal disc was dissected and the RNA analyzed as with the leg imaginal disc RNA. In order for a gene to have been identified as an eye-specific gene upregulated by the ey gene, it had to pass these two criteria: it must be upregulated by ey in the leg imaginal disc beyond a limit of error, and it must be expressed in the larval eye imaginal disc.

As a further test of the efficacy of the method, certain genes identified as eye-specific by the above method were screened for by in situ hybridization in wt eye discs.


The authors found 371 ey-induced genes by the above described method; however, only 55 of these were identified by both gene chips. This indicates that the design and probes used on the gene chips are very important in determining the effectiveness of the chips and thus this method. It also indicates that data taken from just one chip may not be particularly reliable in locating the upregulated genes. It certainly indicates that any given full-genome chip will not find all genes upregulated in a certain tissue for a certain reason.

The authors break the identified proteins down into types as best they could, to show the relative abundance of different types of proteins, such as protein kinases and transcription factors. This information, while noteworthy, is not the crux of the paper. They identify many genes that had not yet been associated with eye differentiation, including transcription factors that were known to be associated with the nervous system development, such as lola and sequoia. Presumably, there is a wealth of information that can be discovered about each of these genes with further research. The most important part of this paper is using the gene chip method and ectopically expressed tissue to uncover a large number of possible genes of interest.

In situ hybridization located the exact location in the eye imaginal disc of some of the proteins identified in the study. While this information is also noteworthy, it is meant to indicate the powerful ability to identify a previously unknown gene and locate its protein product in tissue relatively easily.

Results show this to be a viable method which provides valuable information about which genes are involved in the formation of the eye in Drosophila; however, the method is not perfect. It fails to detect key proteins involved in eye formation which would be expressed in the leg imaginal disc anyway. It also does not detect proteins that act parallel or upstream of ey, since it is based on ectopic expression. From the results in the paper, it appears that gene chip technology is not quite advanced enough to feel confident after a hybridization experiment that all, or even most, of the desired genes were located. The authors conclude that this is a valuable method, but not without its caveats.


I found this to be a well written, fair, and valuable paper. The design of the experiment was clever-it utilized mutants expressing ectopic eye imaginal discs because the RNA expression in the ectopic discs could be compared to the RNA that is expressed in the normal leg disc, and so the genes specifically involved in eye formation, induced by ey could be located.  If the authors simply used RNA from normal eye tissue, it would be impossible to sort the genes necessary for eye development from those that are involved in other processes. This is an inventive extension of the earlier work by Gehring et al. which showed that fly eyes could be grown on legs, when the ey gene is selectively expressed there. This work shows how growing eyes on legs can be a useful source of data difficult to gather by other means.

The authors anticipated certain sources of error in the experiment, particularly due to an unsatisfactory gene chip. By using two chips from different sources with different probes, the authors addressed that concern and strengthened an argument that my have appeared supported if only one chip were used, but really be quietly lacking.  I also thought the authors were careful not to overstep their data or overhype their method. They briefly present their findings-the most important of which being the name of each gene, whose sequence is deposited in NCBI-and appear up front with the fact that the gene chips used were not consistent with each other. 

If anything, I would have liked to see more discussion concerning the role of the gene chips, and how the method could be changed to ensure more consistent.  It was clear from the comparison between the two gene chips that the data could not be relied upon as being complete, quantitative, or even necessarily accurate.  Since this is a large scale screening method, some error would be expected.  However, only 40% of genes identified by one chip was identified by the other.  This is a startling small number that I would not have anticipated.  The authors acknowledge this problem, but they do not appear to doubt the data confirmed by both chips.  From the data printed, I would be hesitant to make any statements about any particular gene until it had been verified as eye-specific by another method, such as in situ hybridization.  The authors stated that this data was considered an "overview," but they should have been more explicit in how uncertain the results were and should have spent more time evaluating the method and attempting to quantify or better explain the discrepancies between the chips.  It would have been interesting to see the results if a third full Drosophila genome chip were used, assuming three distinct such chips exist.  

Future Work

This paper lends itself to being an impetus for a considerable amount of future work. Each of the genes identified could prove to be highly interesting genes whose actions give insight into the way eyes develop, at least in flies. The most obvious starting point for new work is mining through the pile of data and possible leads uncovered by the gene chip experiment. The researchers would ask questions such as "where and when is this gene expressed, and what does it do?" These questions could be investigated using in situ hybridizations and northern blots of mRNA and southern blots of cDNA over time in cells where the gene is expressed. Also, various kinds of knockout flies or flies that ectopically express that DNA could be designed to look into the action of the particular protein. Once specific functions of proteins are uncovered, more specific and detailed assays could be designed to investigate certain specialized aspects about the proteins.  Databases could be used to look for homologs between species, so as to better our understanding of how and why different types of eyes diverged from each other, or whether they developed separately.

Another route starving for work is the design of new and better gene chips which more consistently respond to the experiments such as those in this paper. Theoretically, it should be possible to design chips such that expression of a gene is detectable and quantifiable. This is more of an engineering and chemistry job, but it appears necessary in order to make assays such as these more meaningful and reproduceable.

In order to better understand the time-dependent cascade of genes responsible for eye-development, it would be interesting to perform these gene chip experiments at different stages during development. The expression and waning of a very large number of genes could be followed over time, and then in some cases, correlations can be discovered such as where one gene transcribes a transcription factor for another. The data could be assembled and analyzed to begin piecing together the time-dependent cascade of genetic information which works together to build the Drosophila eye.

Molecular Biology Homepage

Return to my Web Page List

Send questions or comments to Isaac Miller

Last modified: May 2, 2003