Review Paper

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

Review Paper:

“Analysis of the eye developmental pathway in Drosophila using DNA microarrays.”
Lydia Michaut, Susanne Flister, Martin Neeb, Kevin P. White, Ulrich Certa, and Walter J. Gehring.
Proceedings of the National Academy of Sciences of the United States of America. April 1, 2003. Vol. 100, No. 7.

In this paper, the authors utilize genomics techniques in order to analyze the signaling cascade responsible for eye morphogenesis during retinal differentiation in Drosophila. The authors note that retinal differentiation begins in the third larval instar. Third larval instars are about four days old (Flybase.org, 2003). At the beginning of this differentiation, an indentation running from the posterior end of the larva to the anterior end forms. This indentation is called the morphogenetic furrow. Cells within and around the furrow differentiate into photoreceptors, cone cells, and pigment cells. Prior to this paper, it was found that the eyeless, or ey, gene is expressed upon formation of these structures. It is also known that ey is one of two Pax-6 genes, and that Pax-6 genes, when expressed in other imaginal discs, are able to induce formation of eyes. Imaginal discs, for those who are not familiar with Drosophila, are portions of larval cells that give rise to adult structures as growth and development proceeds (Brody, 1996). Armed with this knowledge, the authors engineered flies that express ey ectopically in the leg disc. Thus, the authors compared gene expression in the leg discs of wild-type flies with gene expression in flies whose leg discs ectopically expressed ey, in order to gain insight into the pathway through which eye development proceeds. The authors also analyzed expression of genes within eye imaginal discs in order to validate the data. By comparing the effects of the ey gene on the expression of other genes in endogenous and ectopic situations, the authors felt they would be able to more accurately discern the genes for which ey causes induction. As a further means of increasing significance, the authors utilized two different full-genome microarrays for each analysis, roDROMEGa and DrosGenome1, based on earlier and later releases of the Drosophila genome. The two microarrays also differed in the oligonucleotide probes found on the chips.

The authors determined that a gene was ey-induced by comparing the level of expression on each microarray for RNA taken from eye imaginal discs and RNA taken from leg discs with ectopic ey. If a gene was significantly expressed in both situations, it was considered ey-induced. Ectopic ey-induction was determined by comparing expression levels between wild-type leg discs and leg discs with ectopic ey.

Figure 1A shows two overlapping circles, which represent the two different types of microarray chips used in the experiments. The left circle represents the roDROMEGa array, and it shows that 228 genes were found to be ey-induced on this chip. The right circle represents DrosGenome1, and shows that this array detected 198 ey-induced genes. The overlap represents the fact that 55 of the total 371 genes were detected by both arrays. The authors conclude that the reason for the discrepancy, since the targets hybridized to both types of arrays were identical, is due to internal differences between the microarrays, as they each represent different releases of the Drosophila genome and contain different probes. The authors give the Sur-8, sprint, and SP1173 transcripts as examples of differences between the arrays. DrosGenome1 does not detect these transcripts, but their presence is detected later by in situ hybridization experiment. Although Figure 1A does not add any new insight into the text of the paper, I thought it was a nice summary of the two arrays' discrepancies.

Figure 1B is a bar graph that depicts the amounts of detected genes representing various functional classifications. Out of 254 genes which could be assigned a molecular function, the largest category are the ‘catch-all categories’ labeled “other enzymes” and “other functions.” However, of the specific categories listed, transcription factors are clearly the most prevalent, followed by signal transducers and kinases. I found this figure to be fairly helpful as well. It is simple in the way it conveys the information.

Table 1 lists the fifty-five genes detected by both the roDROMEGa and DrosGenome1 microarrays. Each gene’s molecular function is given, as well as the fold induction that was detected in ectopic ey-expressing leg discs by both types of arrays. For example, the gene scratch is listed as a transcription factor, and its fold inductions were 71.5 and 11.5 for roDROMEGa and DrosGenome1, respectively. The degree of fold induction is color-coded. Genes that have a fold induction of less than 1.5 are yellow, orange if induction is between 1.5 and 5, and red, if induction is greater than or equal to five. Some other genes that are color-coded red are lola, white, eyes absent, and CG140595. Eleven of these fifty-five genes are transcription factors, which is consistent with the observation that transcription factors are the most prevalent functional group in this cascade. One thing that does not make much sense about this figure is that even though the legend states that yellow shading indicates fold induction less than 1.5, most of the fold inductions that are colored yellow are greater than 1.5. One example is string, for which both fold inductions given are 1.7, yet the boxes are yellow.

Table 2 gives the microarray values for a number of genes discussed in the text. The name and function of each gene is stated, followed by the average distance values for wild-type and ey-expressing imaginal discs, the fold induction value, the p value, and average difference for eye imaginal discs.

The authors found a number of genes whose detection was in accordance with previous publications. The authors observe that the detection of eyes absent, so, and dac is in agreement with previously published findings which observed that these three genes are grouped with Pax-6 at the top of the eye development cascade. Furthermore, eighteen of the thirty-eight transcription factors detected were already associated with eye development.

The authors also found a number of genes whose functions are known, but had not previously been associated with eye development. Lola, sequoia, stich1, and net are just four of twenty transcription factors detected by the microarrays. The authors speculate that their expression in both ectopic and endogenous eye development could mean that these genes have roles in eye development. The authors give further evidence for this speculation by citing a previous publication where serial analysis of gene expression found that these four genes are transcribed in the developing eye. A number of signal transducers previously unassociated with eye development were also found. The authors suggest that their expression in the morphogenetic furrow and significant induction in ectopic eye development is evidence that these genes also have roles in eye development.

The authors also performed in situ hybridizations to confirm expression of a few genes. Although they may have done this for more genes, it seems that the genes addressed in Figure 2 features those which are mentioned in the text. Figure 2A differs from the rest of the figure because it involves detection of Quail protein in a wild-type eye disc using monoclonal antibodies. 2A is a dark field figure, and what we are supposed to be seeing is the reddish area to the right of the box where the antibodies have bound to and detected the Quail protein. The rest of the figure (B-N), depicts in situ hybridizations in wild-type eye discs using digoxigenin-labeled antisense RNA. In each box, we see the same structures which have been subject to labeled antisense RNA that corresponds to the gene of interest. The portion to which the probe binds is darker than the rest of the figure. For most of the figure, it seems that the probe is binding in a similar area: a darkened line which transects the eye disc. The authors say that this line is the morphogenetic furrow. They mention three genes which they found expressed in this area: Sur-8, aplip1, and sprint. These genes correspond to panels D,E, and F, and in fact, each one of these boxes appears to display the same darkened row of cells. The authors also point out that the SP1173 transcript (2J) is expressed in a row of cells posterior to the morphogenetic furrow. Indeed, two darkened rows can be seen in the figure, one of which must be the area of SP1173 expression. However, in panel 2C, the authors say the skeletor gene is expressed posterior to the morphogenetic furrow, yet the only second row of darkened cells I see is on the left, which is inconsistent with the authors use of ‘posterior’ for other panels such as 2J. From the panel, it seems like skeletor is transcribed anterior to the morphogenetic furrow. If this is what the authors had said, it would be consistent with their description of the transcripts CG11849 and CG13651 being expressed anterior to the furrow.

Another one of the major findings of the authors is the ey-induced expression of numerous uncharacterized genes. Apparently, over half of the genes detected by the microarrays are uncharacterized genes. The authors define ‘uncharacterized’ as an inability to assign a molecular function to them. In some cases, no homologs or functional domain could be found. According to the author, the three genes CG13532, BcDNA:gh11973, and CG9134 potentially encode for cell-adhesion molecules. However, the authors give no evidence for how they made this conclusion. Presumably they entered the genes’ sequences in BLAST and found other genes with conserved sequences that had cell adhesion roles. It would have been appreciated if such information were included. Other genes, such as CG11849 and CG13651 are listed as being homologs which contain an identifiable DNA binding domain, so the authors did pursue these gene’s function. I would like to know where and how they made these conclusions.

Finally, the authors wrap up the paper by summarizing their findings. They conclude that:

Overall, one thing I liked about this paper is that the authors repeated themselves a lot, which made it easy to follow along. Do I think they make a credible case for their conclusions? I think so. I think the authors do well to say that the results of the microarray suggest a causal relationship between ey and the rest of the genes which appear to be the recipients of its up-regulation. Microarrays are very useful in identifying genome-wide responses, but I think it is difficult to make concrete conclusions without some sort of secondary evidence, which the authors provide in the case of the in situ hybridization. The detections of the microarrays are strengthened by the detection of the genes using in situ hybridization and antibody-labeling. The authors were also wise to perform the experiments using two different full-genome microarrays. If the targets utilized for both arrays are identical, then significance of a detection is greatly increased if it occurs for the same gene on both arrays, especially since the probe sequences differ for both arrays.

I would like to have seen one further control performed on this experiment, however. It would have been interesting if the authors had used the same method to engineer a fly which expresses a non-functioning version of ey ectopically. One could use alkylating agents or any mutagen to engineer a mutation in the ey sequence. The sequence could then be inserted into a P element and introduced into the fly so that the faulty ey would be expressed ectopically in the same way as the functioning ey. A microarray experiment could then be performed using the same target sequences. One would expect no eye morphogenesis-related induction to occur, and this observation would rule out the possibility that it is the merely the presence of out-of-place RNA to which the induced genes are responding. Another way to boost the significance of the results would be to perform the same experiment using the other, unmentioned Pax-6 gene. One would expect that a similar expression profile would be observed. One could also replace the Pax-6 genes with another known transcription factor such as scratch or lola, which might allow one to see the ways in which cascade branches. For example, ey might induce two other transcription factors which each follow their own discrete pathway. This would allow for more accurate characterization of the pathway.

My final suggestion for further research involves a global look at the protein interactions going on in the eye development cascade. If one were to take the cDNA sequences for a large number of the genes detected in the microarray experiment, and use it to make a cDNA library, one could set up a semi-global yeast two-hybrid experiment. Each gene could be used to make gene-activation domain or gene-binding domain hybrids, and systematic introduction into yeast cells would allow for the observation of protein-protein interactions. One could set up a system where either histidine-synthesis or gal-4 production would be the reporter. In this way, one could identify protein interactions which are possibly significant to the cascade. Furthermore, one could observe whether or not transcription factors act alone in their function or interact with other co-factors. This method could also provide an insight into regulatory aspects of the cascade, by revealing possible inhibitors or activators.

Michaut, L., et al. “Analysis of the eye developmental pathway in Drosophila using DNA microarrays.” Proceedings of the National Academy of Sciences of the United States of America. April 1, 2003. Vol. 100, No. 7.