A Quantitative Method to Analyze Drosophila Pupal Eye Patterning

Background The Drosophila pupal eye has become a popular paradigm for understanding morphogenesis and tissue patterning. Correct rearrangement of cells between ommatidia is required to organize the ommatidial array across the eye field. This requires cell movement, cell death, changes to cell-cell adhesion, signaling and fate specification. Methodology We describe a method to quantitatively assess mis-patterning of the Drosophila pupal eye and objectively calculate a ‘mis-patterning score’ characteristic of a specific genotype. This entails step-by-step scoring of specific traits observed in pupal eyes dissected 40–42 hours after puparium formation and subsequent statistical analysis of this data. Significance This method provides an unbiased quantitative score of mis-patterning severity that can be used to compare the impact of different genetic mutations on tissue patterning.


Introduction
The Drosophila compound eye has emerged as a superb tissue in which to study a variety of processes. In particular, pupal eye tissue provides opportunities to examine cell death, signaling, fate specification, cell movement, adhesion and regulation of the cytoskeleton [1]. Errors in these processes produce irregular numbers and organization of cells. These changes can consequently disrupt the precise hexagonal outlines of ommatidia and if sufficiently severe lead to rough adult eye phenotypes. As our understanding of these processes and the group of genes we study increases in sophistication, it becomes increasingly important to account for multiple components of a mutant phenotype rather than a single aspect (such as cell number). We have therefore developed a simple system to systematically analyze and record multiple components of pupal eye phenotypes. This quantitative assessment enables efficient, thorough comparison of genotypes as well as meaningful statistical analyses because each genotype is objectively ranked according to the scope and severity of mis-patterning. An earlier version of this method was successfully used to assess and validate genetic interactions between cindr (which encodes an adaptor protein with roles including actin regulation and endocytosis) and loci encoding actin regulators and junction components [2].
The wild type fly pupal eye has a limited number of cell types [3]: eight photoreceptors that are recruited a full day earlier during the third larval instar and subsequently organized into characteristic positions within each ommatidium, bristle cell organules (composed of four different cells) and four glial-like accessory cell types that take on distinctly recognizable shapes and positions. These are the cone cells (that lie mainly above the photoreceptors with basal processes during development), the primary (1u) pigment cells (which surround the cone cells), and secondary (2u) and tertiary (3u) pigment cells that form a honeycomb lattice across the eye field enclosing and separating neighboring ommatidia. Patterning of these lattice cells occurs between 18-28 hours after puparium formation (h APF) at 25uC: a process of active cell rearrangement and programmed cell death (PCD) reorganizes these cells into their final pattern [4,5] ( Figure 1A-D). The final surface pattern is most usefully scored at 40-42 h APF ( Figure 1E). Here we describe typical mutant phenotypes and a simple method to score them to comprehensively quantify mis-patterning. Photoreceptor cells are not present at the surface of the pupal retina and are not included in this analysis.

Results
A hexagonal grid was superimposed on images of the apical profile of pupal eyes dissected at 41 h APF ( Figure 1F) as follows: each hexagon was drawn to connect the centre of 6 ommatidia surrounding a central ommatidium; this field was then utilized as a single data point as we scored defects observed within each hexagonal area. One 'point' was awarded per defect and recorded in a spreadsheet (Microsoft Excel) and then summed to give a total number of defects per field. We found that analyzing 75 ommatidia of a genotype and then determining the average number of errors per ommatidium provided a reliable 'ommatidial mis-patterning score' (OMS) characteristic of that genotype; variance and standard deviation were also included. Some mutant eyes displayed position-specific defects. For example we consistently found more severe phenotypes in the posterior hemisphere of the eye when expressing an RNAi transgene. In addition bristle groupings were often observed to be missing or mis-positioned toward the periphery of the eye field. To prevent such positiondependent effects from skewing the final OMS, we routinely imaged and analyzed only the central region of the pupal eye. In addition the 75 data points were collected from images obtained from between 8 and 12 pupal eyes, and dissections were repeated to ensure that observations fairly represent each genotype. Inevitably it is also important to culture and dissect control and experimental genotypes simultaneously to allow proper comparison of age-matched phenotypes.
The following features were scored (summarized in Table 1; all phenotypes refer to the apical surface, refer to Materials and Methods for genotypes):  Bristle grouping defects N Bristle organules are comprised of four cells-neuron, glia, trichogen, and tormogen [3]-that are difficult to resolve with standard fluorescence imaging. Hence only gross defects pertaining to overall patterning of the tissue were assessed. Three characteristically positioned bristle groupings surround each wild type ommatidium ( Figure 4A); this positioning can be aberrant at the posterior of wild type eye fields but rarely at  Figure 5B, C). A maximum score of three errors for 3u cell defects can be allocated per data point: points were not allocated for additional (more than three) or misplaced 3u's since these positional defects are accounted for when assessing bristle defects.

Rotation defects
N Cone and 1u cells are oriented precisely within the plane of the epithelium ( Figure 6A) along an equatorial-polar axis. This arrangement is usually consistent with the arrangement of underlying photoreceptors. Defects in this orientation greater than 30u were termed mis-rotated ommatidia ( Figure 6B) and allocated one point. However this classification is arbitrary: whether or not the underlying photoreceptor group and entire ommatidia is mis-rotated is not readily assessed when imaging the apical surface with junctional markers.
Interommatidial cell number N In wild type tissue twelve interommatidial pigment cells (IPCs) lie within the hexagonal area of a data point: three 3us, six complete 2us, plus the halves of six additional 2us (illustrated in Figure 7A). For an experimental data point, one point was allocated per additional or missing cell above/below the total of twelve. Cells were counted regardless of their position, size or apparent fate ( Figure 7B). Any cells lying partly within the superimposed hexagonal area were scored as half a cell regardless of what proportion of the cell actually lay within the data point area. Additional or missing lattice cells may be consequent to either direct mis-regulation of the apoptotic machinery, signaling to enhance cell survival, or defective cell movements that target cells to specific zones surrounding each ommatidium where apoptosis preferentially occurs [4,5,7,8].
N Patterning of 3us was additionally scored in our analysis (see above) because this cell fate is dependent on cell signaling, programmed cell death of excess cells, and regulation of adhesion and cytoskeleton. Using live imaging we have observed 3 cells actively compete for this position at each vertex [8]. The 2u cell fate was not separately scored: we consistently find that, provided approximately the correct number of cells are removed by apoptosis and the 3us are correctly specified, the remaining cells automatically adopt the elongated 2u cell shape. Hence by scoring cell number we also account for the 2u cell fate. In addition, we routinely find that mis-patterning of the ommatidial hexagon is more severe when 3u vertices have not been correctly established than when the 2u niche is mis-specified. We do not directly measure the angles of the hexagonal lattice because distortions are easily introduced during normal tissue processing prior to imaging.

Example
To validate a genetic interaction between the loci for the adaptor protein cindr and the actin regulator enabled we generated tissue hypomorphic for cindr ( Fig. 8A and B, GMR-Gal4/+ ; UAScindr RNA[2.23] /+) and in a parallel cross tissue in addition compromised for ena dosage ( Figure 8C and D, GMR-Gal4/ena GC1 ; UAS-cindr RNAi [2.23] /+). Images gathered from pupae dissected at 41 h APF were analyzed to generate 75 data points of each genotype. Excerpts from each Excel database are shown in Figure 8E  N Hence the total OMS is 1+2+1 = 4.    Figure 8G). A Students T-test was used to compare OMS values of the complete datasets (N = 75) to determine statistical significance: this confirmed that ena GC1 mildly enhanced cindr RNAi mis-patterning (p-value = 0.02438, significant at the 5% level). In particular the number of rotation and 1u cell defects approximately doubled when ena was compromised  (compare mean and standard deviations shown in Figure 8E and F) emphasizing that the role of these loci in 1u:1u junction formation or maintenance, actin regulation and ommatidial rotation warrant further investigation.

Discussion
Here we present a simple method set to assess apical mispatterning of the pupal eye. This provides a systemized and unbiased approach useful for quantifying and comparing genotypes. Data sets and derived ommatidial mis-patterning scores are readily assessed for significance using a suitable statistical test if an investigator's aim is to show a clear difference between two or more genotypes (e.g., Student T-test, found in most spreadsheet programs such as Microsoft Excel). Frequently, investigators utilizing the fly pupal eye as an assay have focused on a single aspect of mis-patterning such as interommatidial cell number. However we have found that simply scoring a single component such as cell number does not sufficiently encompass patterning defects and fails to provide a meaningful assessment of mis-patterning. Further, by recording this fuller set of defects, an investigator can additionally evaluate phenotypes specific to one cell type or feature and relate the information to mutants in other loci. Through this approach, we have been able to place loci into functional groups based on the details of their scoring [2] and unpublished data).

Fly husbandry and genetics
All crosses were cultured as per standard protocols. Pupae were gathered at 0 h APF and cultured at 25uC until dissected as described previously [2].
Genotypes of images presented: Wild type tissue: Imaging Dissected tissue was fixed and processed as described previously [2]. Rat anti-DE-Cadherin (1:20, DSHB) was used to visualize adherens junctions. Tissue was imaged using a Leica DM5500 microscope. Images have been minimally processed and pseudocolored using Photoshop to emphasize specific cell types or features.

Additional information
Raw images were printed and hexagonal areas superimposed by hand to create data points for scoring as described above. Analyses were recorded in databases generated using Microsoft Excel.   23] /+ dataset. In the left panels cone cells lying within the superimposed hexagons are pseudo-colored purple, circles indicate missing and blue arrows indicate misplaced bristle groups; mis-rotation is indicated by a black arrow; correct 3u cells are indicated in green. All IPCs are colored green in the right panels, whole cells within each hexagonal outline are labeled with a N and half cells indicated. E. and F. Excerpts from our Microsoft Excel databases for each dataset, in each case 4 data points are shown: those provided in panels A to D are highlighted in orange. The scoring system is reiterated in the column headings and Mean and Standard Deviation (SD) values for each feature are shown. Standard error (SE) for the mean OMS is also calculated. G. Graphic comparison of the mean OMS values for each genotype, SE are indicated (bars). The mean OMS values differ significantly at the 5% level (p = 0.02438). doi:10.1371/journal.pone.0007008.g008