As you may have gathered from perusing through my blog, I use Drosophila to study the effect of micro RNAs (miRNAs) on neuron development. This animal model is useful for high throughput genetic screenings in vivo because of the high genetic and functional homology between Drosophila and humans, with about 75% of human disease genes having orthologues in flies. Flies are also easy and cheap to handle.
The eye system of flies is well studied and is a great model to study neuronal specification and differentiation. Phenotypic changes to the eye show problems in neurodevelopment and can be analysed qualitatively. This is time consuming and can be done differently by every scientist. It’s hard to believe your eyes when your findings aren’t quantitative!
In May 2016, the journal “Genes, Genomes, Genetics” published a paper by Iyer et al. titled “Quantitative Assessment of Eye Phenotypes for Functional Genetic Studies Using Drosophila melanogaster” that tackles this. Iyer et al. (2016) cover the development and rigorous testing of their computational ‘Flynotyper’ software, but is it the solution to the difficulty in measuring the disorderliness of Drosophila melanogaster eyes?
Figure 1: Representation of A) control, B) subtle rough, C) rough and D) sever rough phenotype eyes based on the Flynotyper algorithm. The ommatidial centres are encircled in orange and the vectors are shown as arrows from the centre of the ommatidium to its six neighbouring ommatidia. This is taken from Iyer et al (2016)s.
The novel computational method is said to be applicable for the analysis of bright-field microscope or scanning electron microscope (SEM) acquired images. Where the bright-field microscope detects bright spots at the centre of the ommatidia formed by light reflection, and the SEM images have dark spots at the centres of the ommatidia.
The first step of ‘Flynotyper’ software is to detect the eye from the background. To do this the bright-field images are converted into grey scale, morphological transformations are used to reduce the background, and top-hat transformations are used to highlight the light ommatidia points in the dark images. The methodology for the top-hat transformation interestingly originates from identification of licence plate numbers in photos Arulmozhi et al. 2012. In the SEM images the top-hat transformations are not used because there is too little contrast between the background and the eye, instead, a thresholding operation is applied.
Following the enhancement of the eye region, the individual ommatidia are identified as single, bright and round ‘blobs’. The programme increases the contrast between the ommatidia and the background, enhances the images and removes noise from neighbouring ommatidia using a median filter. Once each ommatidium is a single ‘blob’ the phenotypic analysis can begin.
The phenotypic score relies on the regular and symmetrical distribution of ommatidia in the wild type eyes of Drosophila. The fly eye consists of a regular arrangement of around 750 ommatidia. Each ommatidium consists of eight photoreceptors in a symmetrical circle, surrounded by non-neuronal supporting cone cells, pigment cells and bristle complexes.
Ommatidia are distributed in a hexagonal way, with a regular distance and angle between the centre of an ommatidium to its six neighbouring ommatidia (See Figure 1 A). Using this knowledge Iyer et al. (2016) wrote an algorithm that creates vectors from an ommatidium’s centre to its neighbours’ centres to determine the extent of disorder. The disorder phenotype score was determined using complex calculations, which you can read about further in the paper on page 1430. The total ommatidial disorderliness index value depends on the angles between adjacent vectors and the differences of length of the six vectors from each ommatidium. Based on this a phenotypic score P is determined which can help to rank the eyes by their disorderliness in a quantitative way. The more disorganized the eye is the higher the phenotypic score is.
Iyer et al. (2016) put ‘Flynotyper’ into practice using various experiments. The comparison of the ‘Flynotyper’ phenotypic scoring of images to the qualitative scoring completed in six-independent studies (published and unpublished) showed accurate corresponding scores of the severity of the disordered eyes. The paper shows the assessment of Mizielinska et al. 2014 and you can see that the ‘Flynotyper’ data (Figure 2 B) matches the phenotypes (Figure 2 A) e.g. the severe GMR>103 d45 has the highest phenotypic score.
There were also some new studies conducted, such as an RNA interference study of fly gene orthologues of human genes that are known to lead to neurodevelopment disorders and new identification of genes interacting with sine oculis. Iyer et al. (2016) also tested various image resolutions (600 X 800, 1200 X 1600 and 1600 X 2400) and the difference between using SEM and bright-field microscopy. The findings showed that higher resolutions detected more ommatidia but SEM and bright-field microscopy had similar accuracies.
Figure 2: Flynotyper analysis results of Mizielinska et al. (2014) eye phenotypes. Adapted from Iyer et al. (2016).
To test sensitivity of the system the gene dosage of neurodevelopmental genes was changed by keeping the flies at 28°C or 30°C or using different knockdown lines. The UAS-GAL4 system, mentioned in one of the previous posts, is temperature sensitive, and so will express genes less at 30°C than at 28°C. qPCR experiments confirmed the difference in gene expression. These different levels of gene expression cause subtle changes that were not easily spotted by the eye, but Flynotyper can detect them (see Figure 3), which is impressive!
Figure 3: Phenotypic scoring of eyes exposed to different dosages of inhibitor expression. The severity of the phenotype increases as the temperature increases and so the inhibitor dosage decreases for all genes. Adapted from Iyer et al. (2016)
This method seems efficient at detecting eye phenotypes such as rough eye, however it may struggle with different phenotypic mutations such as bristles integrated into the eyes, or change in overall eye size. This new methodology seems to have a lot of potential and I am excited to try it, but I think it should be used to assist the analysis, and not to replace it.