Affymetrix Microarray Chips

3.1.3.1 Chip Design and Grading

The encouraging results from the “Faber Chip” prompted the decision to perform a full experiment using microarray chips. Affymetrix microarrays for zebrafish seemed the best choice at the time, with high numbers of spotted transcripts and industrial manufacturing promising reproducible results. These chips carry approximately 15.600 probe sets equivalent to approximately 14.900 zebrafish transcripts. For each time point 3 biological replicates were prepared and tested on individual chips.

Testing such a large number of genes will, without multiplicity adjustments, lead to many false positive results. The software “GeneSpring”, used to analyze our chip results, employs multiplicity adjustments such as Bonferroni or Benjamini and Hochberg. The statistical analysis of variance (ANOVA) was employed (see section 2.22.2.).

The different probe sets spotted on the Zebrafish Affymetrix chip are not all perfect matching probes. To distinguish the quality of the probe sets, Affymetrix sorts the probes on the chip in 4 categories, ranging from precisely matching probes (Grade A) to probes for which no transcript could be found and only an expressed sequence tag (EST) is known as yet (Grade E). 51% of the probe sets are Grade A compared to 32% of Grade E (see Table 2). Further sequencing efforts on the zebrafish genome will provide more accurate information transforming some E-graded probes to A-graded ones. An overview of grades on the chip is given in table 2. A short explanation of the different grades is given below, taken from an Affymetrix publication to be found on the Affymetrix website (http://www.affymetrix.com/support/technical/whitepapers/Transcript_Assignment_whitepap er.pdf).

Matching Probe (Grade A) probe sets have nine or more probes matching a transcript mRNA or gene model sequence. Genome Target-Transcript Overlap (Grade B) transcript assignments have a partial overlap between the transcript and the target sequence. Genome

Consensus-Transcript Overlap (Grade C) transcript assignments result when the transcript sequence overlaps the consensus, without overlapping a significant portion of the target.

Overlap transcript assignments must be substantiated by a correspondence to the genome. The letter D is not used. If no transcript is found, then EST-only data described through a UniGene EST cluster are given a Grade E (EST record) assignment.If no UniGene EST cluster is built for that probe set, then the representative sequence IDs from the original design record are designated with a “minus”.

While, looking at Table 2 it is obvious that only 51% of all probe sets are A-graded, it is however surprising to note that over 40% (taking E-graded and “minus”-graded together) are spots with no information so far. Taking this into account, we looked at the grading of our regulated genes ensuring they consisted of a high percentage of A-graded spots (Table 3).

These numbers suggested that a high percentage of identified genes were A-graded spots and by that providing solid information for a high percentage of them, helping us to evaluate our experimental outcome. On the other hand, working on the idea to identify currently unknown genes involved in regeneration, we decided not to exclude the E-graded spots.

Letters and symbols used for grading Probes % of probes

Total number of letter: A 7916 50.7

Total number of letter: B 872 5.6

Total number of letter: C 319 2.0

Total number of letter: E 4989 31.9

Total number of probes marked with a: - 1521 9.7

All probes 15617 100

Table 2: Grading given to all spots on the zebrafish Affymetrix chip

Spots were filtered by an ANOVA process with p = 0.05 and a twofold cut-off. After that statistical procedure 94 genes at 6hours, 193 genes at 12 hours and 113 genes at 11days post lesion remained as regulated. The number of A-graded probes varied between 51% and 73%

and none was marked with a minus, even though no UniGene EST cluster is built for nearly 10% of all probes on the chip (Table 2). Furthermore it can be seen in Table 3 that for 23% - 47% of the regulated spots only an EST is known so far, implying that a substantial amount of genes is still unknown but involved in the regeneration of the optic nerve.

Number of probes per grade-group

6 hours (n = 93) 12hours (n = 193) 11days (n = 113)

Grade A 69 (73%) 99 (51%) 71 (63%)

Grade B 3 (3%) 3 (2%) 3 (3%)

Grade C 0 0 1 (1%)

Grade E 22 (23%) 91 (47%) 38 (33%)

Grade - 0 0 0

Table 3: Regulated spots split up into grade groups

The Venn Diagram

Having identified these genes as regulated, our first idea was to see how many identical genes would be regulated at different time points. By drawing a Venn diagram (see Figure 6) it became clear, that only very few genes overlap between the different time points (6 hours, 12 hours and 11 days) even though 6 hours and 12 hours are relatively close on a time scale.

Only 16 genes were similarly regulated at 6 hours and 12 hours post-lesion, and only 10 genes were found to be commonly regulated comparing 12 hours and 11days post-lesion.

The 2 probes overlapping at all time points are transcribed loci in the genome but are unfortunately not yet annotated. The 16 genes overlapping at the 6 hours and 12 hours time points are not from one functional group but fall into different functional groups, such as transcription factor activity, visual perception, biosynthesis, etc.

Nothing is known about 8 of the 10 genes common to both the 12 hours and 11 days time points using the gene ontology database. The remaining 2 are found in the circadian rhythm and the visual perception group.

Due to limited information concerning these overlapping genes, only a careful interpretation of these results is possible. The observation of only so few identical genes being upregulated at 6 hours as well as at 12 hours post-lesion is somehow surprising, since the composition of functional groups being regulated is different between 6 hours and 12 hours (see Figure 7) but not as drastic compared with the overlap of single genes. One explanation could be that the function needed at these specific time points is executed by different genes, creating a minimal overlap in genes but not in functions.

Judging the two genes which are regulated at all time points is difficult, only knowing that they are transcribed loci. Further analysis using the “blast” software tool on the PubMed webpage did also not lead to satisfying results. Judging their regulation pattern, upregulated at all three time points, and knowing of the very specific actions needed for regeneration, could classify them as a byproduct of tissue damage, reflected e.g. in upregulated metabolism.

The very limited information about the 10 overlapping genes between the 12 hour and 11days time points only allows suggestions. Genes which are regulated not at 6 hours post-lesion but later might not have “immediate early” characteristics or an immediate stress response but would more likely fall into e.g. axon elongation or pathfinding.

Figure 6: Venn diagram showing the minimal overlapping of regulated genes at all three time points after lesion (6 hours, 12 hours and 11 days).