The neuralized^p72-Gal4 driver was used to express in pI cells the construction mCD8::GFP using the UAS/Gal4 system [10]. neur>mCD8::GFP flies were reared on a standard Drosophila diet. White pupae were selected and maintained at 25°C prior dissection.

The notum from neur>mCD8::GFP pupae were dissected out in PBS and fixed in cold absolute ethanol for 10 minutes. A fixation longer than 15 min, hardes the tissue and makes the microdissection difficult. Three to five fixed nota were transferred directly from ethanol with a Pasteur pipette and then carefully flattened and dried with the epithelium facing down on a thermolabile membrane slides for laser microdissection (See Fig S1).

Laser microdissection was realized on a MMI cellcut microdissection system coupled to an Eclipse TE-2000 inverted fluorescent microscope (Nikon Instrument). The parameters used were: focus 40, speed 1, power 74 at objective 60X and 4 to 8 laser rounds were required to cut through a notum. Selected areas were cut from the tissue by an UV laser beam. To keep the SOP integrity and preserve RNA from the heat of the laser, we took care to leave a space between the laser circle and the cell limit (around 5 µm).

Total RNA was extracted from microdissected cells by using the picopure RNA isolation kit (Molecular devices - Arcturus) following manufactures instructions with minor modifications as described below. We incubated the tubes containing microdissected cells with 20 µl of extraction buffer at 42°C upside down for 30 min. Then, after centrifugation, the extracts pooled were passed through a single RNA purification column. During purification, we treated the column with DNAse I (Qiagen) for 30 min at room temperature to avoid genomic DNA contamination. We obtained 0,1–0,5 µg of total RNA from a sample of 1000 microdissected cells.

After extraction, RNA was amplified by using the MessageAmp II aRNA Amplification Kit (Ambion). We proceeded with two rounds (9 h each) of in vitro transcription. After each round, the RNA purification column was treated by DNAse I (Qiagen) for 10 min at room temperature before aRNA elution. For better RNA integrity, we carried out all the amplification processes in one step directly after RNA extraction to avoid freezing the sample. Indeed, in addition to the usual recommendations about manipulating RNA, we avoided, as much as possible, freezing both the tissue before microdissection and the RNA samples between extraction and amplification. After two rounds of amplification, we obtained 20–50 µg of aRNA from a sample of 1000 microdissected cells.

For microarray hybridizations, UTP-amino allyls were integrated during the second round of in vitro transcription, for subsequent labelling with dyes Cy3 or Cy5 (Amino Allyl MessageAmp II kit - Ambion).

We performed reverse transcription on 1 µg of aRNA using random primers from Roche and the SuperscriptII reverse transcriptase from Invitrogen. The same quantity of cDNA (50–100 ng) from SOPs or epithelial cells was then used to perform semi quantitative PCR (30 cycles) or qRT-PCR for several genes.

qRT-PCR was performed on Bio-Rad iCycler IQ™ using SYBR green PCR master mix with the following parameters: 95°C-3 min followed by 40 cycles of 95°C-30 sec, 60°C-30 sec and 72°C-30 sec. Quantifications were made using the relative standard curve method. The standard curves were created by a series of 5 dilutions of cDNA synthesized from aRNA, extracted and amplified from 20 whole nota dissected and fixed as described here. Each dilution of the standard curves was amplified in duplicate and each sample of interest was amplified in triplicate. Curves of one experiment are shown in figure S2. mRNA levels detected by qRT-PCR were normalized to mRNA level of taf11 used as reference gene.

Amplified and differently labelled aRNA from 1000 microdissected SOPs and an equivalent surface of epithelial cells were hybridized to INDAC Drosophila GeneChips (platform Montpellier GenomiX, Institut de Génomique Fontionnelle, UMR 5203 CNRS – U661 INSERM, Montpellier, France). Normalization of raw data was performed by LIMMA. The flagged spots and controls were removed from the analysis. No background correction was performed before normalization. Lowess normalization was used to normalize the M values for each array separately (within-array normalization). Genes exhibiting a signal ratio SOPs/epithelial cells superior than two were considered as SOPs-overexpressed genes for subsequent analysis. Gene Ontology analysis was performed with Flymine [11] that provides enriched GO terms ranked by significance. P values were calculated following a hypergeometric distribution (with Bonferroni correction).

The raw data associated with this manuscript are available on the Gene Expression Omnibus (GEO) according to MIAME standards under the following accession number: GSE18615.

In order to identify SOP cells, we specifically expressed the construction mCD8::GFP to label SOP membranes and their progeny by using the Gal4/UAS expression system and the specific driver line neuralized^p72Gal4 (neur>)[11]. The dorsal epithelium (or notum) of neu>mCD8::GFP pupae at 16h after puparium formation (APF) was dissected and fixed in ethanol. At this developmental time, most of the SOPs have not yet divided [12], [13].

After mounting on a membrane slide, SOPs expressing GFP were identified by fluorescence and circled manually with a circle radius of 9 µm (Fig. 1A). Microdissected cells were then collected on an adhesive lid of a microtube placed onto the area (Fig. 1G–J). The success of the cell capture was visually confirmed by the gaps in the tissue after lid removal (Fig. 1C–F). We collected around 20 cells per notum, 20–50 cells on a cap and pooled around 20–50 tubes to proceed to the RNA extraction. Altogether, we collected around 1000 SOPs from 50 nota. In parallel, we captured tissue free of SOP fluorescent cells corresponding to epithelial cells (Fig. 1B). A similar integrated surface (around 250 000 µm²) was collected in order to standardize both samples.

Once the required number of cells was been collected, total RNA was extracted and amplified for analysis.

We carried out reverse transcription following by PCR on some SOPs specific (ac, sens, [4], [9]) and non-specific (rp49, taf11) genes to verify the aRNA extracted and amplified from microdissected cells. Semi quantitative PCR (30 cycles) performed on the same quantity of cDNA from SOPs and epithelial cells showed that Rp49, taf11 and ac seem to be expressed at similar levels in SOPs and epithelial cells. In contrast, as expected, gfp (that was ectopically expressed in SOPs) and sens were more highly expressed in SOPs than in epithelial cells (not shown). To verify these results, we performed quantitative real time PCR (qRT-PCR). We calculated the ratio of SOP/epithelial cells mRNA levels for each gene (Fig. 2A). Using this procedure, we confirmed that ac expression was not significantly different in SOPs and epithelial cells (ratio  = 0,8), whereas the expression of gfp and sens was higher in SOP than in epithelial cells (11,2 and 11,4 times respectively) (Fig. 2A and Fig. S2). The significant enrichment of transcripts corresponding to sens and gfp in SOPs confirms the usefulness and the specificity of aRNA material collected using the laser microdissection technique on fixed Drosophila nota.

Concomitantly to qRT-PCR analysis, we used DNA microarrays to identify genes differently expressed between microdissected SOPs and epithelial cells. This analysis revealed 127 genes whose expression was increased 2-fold or greater between SOPs and epithelial cells (Table 1). To analyse whether a particular biological process could be overrepresented in this data set, we regrouped the genes of this set according to their function that has been ascribed using Gene Ontology (Go) terms (www.geneontology.org). This analysis showed that 58% of these genes were associated with a specific function. Interestingly, 27% of this subset of genes were related to the nervous system. This category showed more than a three fold enrichment in the SOP-gene data set (Fig. 2B). More precisely, a hypergeometric test applied on this set of 127 genes, revealed a significant enrichment in GO terms related to nervous system development, sensory organ development and cell fate specification. Moreover, several eye photoreceptor cell development associated GO terms were also enriched in our SOP-gene data set (Fig 2B). Conversely, among GO terms that are significantly underrepresented and, as a consequence, enriched in their sibling epithelial cells, we found cuticle development and epithelium morphogenesis (data not shown). Furthermore, neither genes already known to belong to SOP-enriched genes nor genes associated with GO terms related with cell fate determination and sensory organ specification were found in this SOP non-enriched set of genes.

Our data set of SOP-specific genes includes 19 known genes that have already been shown to be expressed in SOPs or involved in sensory organ development related mechanisms (Table 2). Among them, we found sens, confirming our qRT-PCR analysis, and other SOP-specific genes such as cut (ct), neuralized (neur) and phyllopod (phyl) [14]–[16]. We can also note pebbled/hindsight (peb) and seven up (svp) that are involved in photoreceptor development [17], [18], scabrous (sca) that plays a role in lateral inhibition processes via the regulation of Notch activity [19], and miranda (mira) involved in neuroblast and SOP asymmetric divisions [20], [21]. Moreover, among these genes, eleven have already emerged from microarray analysis performed on proneural clusters by Reeves and Posakony [22]. In addition to the well characterised genes such as mira, peb, neur or phyl, we can cite as an example of new SOP genes, quail and insensitive (Table 2). In contrast, some genes in our data set didn't appear in Reeves and Posakony microarray results. The most relevant examples are sens and ct, two well known SOP-specific genes [8], [14]. We can also cite shaven, sca or nervy, all three being involved in sensory organ development [19], [23], [24].