An automated image analysis method to measure regularity in biological patterns: a case study in a Drosophila neurodegenerative model
- Sergio Diez-Hermano†1,
- Jorge Valero†2,
- Cristina Rueda3,
- Maria D Ganfornina1 and
- Diego Sanchez1Email author
© Diez-Hermano et al.; licensee BioMed Central. 2015
Received: 26 November 2014
Accepted: 12 February 2015
Published: 12 March 2015
The fruitfly compound eye has been broadly used as a model for neurodegenerative diseases. Classical quantitative techniques to estimate the degeneration level of an eye under certain experimental conditions rely either on time consuming histological techniques to measure retinal thickness, or pseudopupil visualization and manual counting. Alternatively, visual examination of the eye surface appearance gives only a qualitative approximation provided the observer is well-trained. Therefore, there is a need for a simplified and standardized analysis of fruitfly eye degeneration extent for both routine laboratory use and for automated high-throughput analysis. We have designed the freely available ImageJ plugin FLEYE, a novel and user-friendly method for quantitative unbiased evaluation of neurodegeneration levels based on the acquisition of fly eye surface pictures. The incorporation of automated image analysis tools and a classification algorithm sustained on a built-in statistical model allow the user to quickly analyze large sample size data with reliability and robustness. Pharmacological screenings or genetic studies using the Drosophila retina as a model system may benefit from our method, because it can be easily implemented in a fully automated environment. In addition, FLEYE can be trained to optimize the image detection capabilities, resulting in a versatile approach to evaluate the pattern regularity of other biological or non-biological samples and their experimental or pathological disruption.
The retinal system in Drosophila has been for long a very useful model to study the pathogenic mechanisms of human neurodegenerative diseases  and to test the efficacy of phenotypic modifiers of a pathological condition that arise from diverse genetic screenings . The fruitfly compound eye is formed by around 800 units, called ommatidia, which display a very regular pattern. The stereotypic development of the insect retina and the extensive knowledge of the cellular and molecular mechanisms involved, make this a very reliable system to evaluate whether altering the expression of mutated proteins is linked to cell degeneration.
Poly-glutamine-based neurodegenerative diseases, such as Huntington’s and a number of Spinocerebellar ataxias, have been studied using the Drosophila retina as a test tube and the Drosophila GAL4-UAS system for transgene expression [3-5]. A qualitative examination of the external appearance of the fly eye (described as rough-eye phenotype) has been widely used to categorize whether a mutation would improve or worsen a given degeneration level. However, most quantitative estimates have relied upon methods involving tissue fixation, paraffin or cryostat sections, histochemical/immunohistochemical staining or scanning/transmission electron microscopy (SEM/TEM). Using these preparations, researchers have evaluated retinal thickness, rhabdomere counting, the regularity of the hexagonal photoreceptors array, or have scored for the presence of expected features in the retinal surface [6-10].
In this work we present a simple and reliable method to quantify the degree of retinal degeneration based on fly eye surface photographs. An earlier version of this method was successfully used to assess the rescuing ability of Lipocalin genes and its dependence on autophagic activity in a model of Type I Spinocerebellar Ataxia . Following tests with other commonly used techniques, and validation with randomly chosen eye pictures, our image analysis method was implemented in a freely distributed plugin (FLEYE) for the open source image analysis program Fiji . Our user-friendly Fiji plugin allows for a fast evaluation of the eye regularity pattern, and for a quantitative unbiased assessment of neurodegeneration under diverse pathogenic levels or genetic penetrance conditions. Since the method obtains a regularity index for each image, it could also be applied to study changes in any regular biological pattern.
Fly lines and maintenance
Flies were grown in a temperature-controlled incubator at 25°C, 60% relative humidity, under a 12 h light–dark cycle. They were fed on wet yeast 84 g/l, NaCl 3.3 g/l, agar 10 g/l, wheat flour 42 g/l, apple juice 167 ml/l, and propionic acid 5 ml/l. Fly females were used in all experiments. We used the line gmr:GAL4 to drive transgenes expression to the eye photoreceptors. UAS:hATXN182Q was used to trigger the neurodegenerative phenotype  and different UAS:modifier-gene constructs were used to test the system.
Adult fly heads were fixed with 4% paraformaldehyde, dehydrated in ethanol, and included in paraffin. Paraffin sections (4 μm) were dewaxed with xylene and rehydrated in an ethanol series. Histochemical staining with hematoxylin and eosin was performed according to standard procedures. The labeled sections were photographed with an Eclipse 90i (Nikon) fluorescence microscope equipped with DS-Ri1 (Nikon) digital camera. Images were acquired with the NIS-Elements BR 3.0 software (Nikon), and processed with ImageJ (version 1.48p).
External eye surface digital imaging
Digital pictures (1280x960 pixels) of the surface of fly eyes were taken with a DS-L1 digital camera, in a Nikon SMZ1000 stereomicroscope equipped with a Plan Apo 1x WD70 objective. The flies were anaesthetized with CO2 and frozen for 10 minutes at −20°C. Their bodies were immobilized on dual adhesive tape, and their heads set up to have an eye parallel to the stereomicroscope objective. Fly eyes were illuminated with a homogeneous fiber optic light (20 W; KL 200, Zeiss). A white balance was performed on the background white surface. Additional settings include a 6x optical zoom in the stereomicroscope that results in a final resolution of 1.85 μm/pixel. Image files were saved in Tiff format.
Special care must be put into maintaining the same illuminating conditions and camera settings between experiments, as differences among pictures of the same stack may introduce undesired artifacts that would hamper the discrimination capacity of FLEYE.
The statistical analyses and graphical outputs for the measurements described below were performed and generated using SAS 9.2 and SPSS.
Multiple Box plots and histograms, and dispersion diagrams and pairwise sample correlations have been used to describe variable distributions within groups and pairwise relationship respectively.
Differences between eyes with different degeneration degrees were assayed using ANOVA for the selected variables describing the degree of regularity in each image (see below). A principal component analysis (PCA) was performed and the scores of the first principal component were used to generate clusters. A Multinomial Logistic Model for the clusters defined previously, was fitted using only a subset of the data set (considered as the training data). The remaining data were used as a test to validate the model. The final set of predictive variables was selected using a stepwise variable selection approach, and the estimated probabilities of each image to belong to each cluster were used to derive a regularity index (IREG) adopting values from 0 to 1. The robustness of the procedure was tested by randomly splitting the data set in different ways (in training and test sets) and comparing the resulting regularity indexes.
Final comparisons of IREG medians between different experimental classes (genotypes) were performed using Kruskal-Wallis non-parametric hypothesis contrast. Post-hoc tests followed the Dunn’s method. A p-value <0.05 was considered statistically significant.
FLEYE plugin requirements
The FLEYE plugin is composed of four different ImageJ1 macros: FLEYE_menu_v2.ijm, Fleye_ROISv1.2.ijm, Fleye_optimizer_v4.2.ijm and Fleye_v10.2.ijm. It has been developed in the ImageJ version 1.49d, using Fiji. The plugin uses the “Bio-Formats Importer”, released by the OME Consortium (http://openmicroscopy.org) to be able to open the different formats of image files. The ‘Bio-formats Importer’ plugin is included in the Fiji package. Thus, we recommend using this package to run FLEYE.
The FLEYE plugin pack, a quick guide and user manual, and the GNU general public license file are provided in Additional files 1, 2, 3, and can be accessed at http://imagejdocu.tudor.lu/doku.php?%20id=plugin:analysis:fleye:start&#fleye.
Results and discussion
Retinal degeneration in a Drosophila model of human spinocerebellar ataxia SCA1
Poly-glutaminated proteins are a common pathogenic mechanism of a diverse array of human neurological diseases. Here we have used the GAL4/UAS system to express a pathogenic version of human Ataxin 1 with an expanded glutamine tract (hATXN182Q) in Drosophila retinal photoreceptors using the gmr:GAL4 driver. In this model of SCA1, photoreceptors accumulate nuclear inclusions of the human protein and start degenerating during late pupal stage when flies develop at 25°C . After expressing the mutated pathogenic protein form, the adult retina degenerates and the eye surface morphology loses its regular pattern appearance .
We have designed a digital image processing to automatically detect the bright spots that appear in the image due to light reflection in ommatidia.
The next step involves finding the maxima of the intensity pixel function with the “Find maxima” Fiji plugin using a predetermined tolerance value. In our case, these maxima approximate the position of fly ommatidia (Figure 2E). Therefore, counting the number of maxima gives an estimation of the total number of ommatidia present in the ROI selected area.
An optional optimization step is also available. This step requires the user to manually identify ommatidia in representative areas of eye images (Figure 2B) to calibrate two main detection parameters: the tolerance of the “Find maxima” plugin and the rolling ball radius of the “Subtract Background” plugin.
Our image analysis aims at detecting discrepancies in the regularity of the spatial distribution of ommatidia. Our strategy follows two main processes: a global one, based on distances between maxima, and a local one, that splits the ROI into a grid of squared cells and extracts statistical and spatial information of the maxima in each grid-cell.
Once the coordinates of the local maxima are obtained, we calculated the distance of each single maxima to its “nearest neighbor” (in normalized units to fit with our statistical model) using a self-developed algorithm.
List of the 18 variables used for the statistical model development
Maxima per cell
Distance to the center of mass
To estimate the degree of retinal degeneration we followed a statistical procedure based on two steps: 1) Categorization of the degeneration extent in discrete classes and estimation of the probability of an eye to belong to a certain class; 2) Computation of the overall regularity degree of each eye as a weighted mean of those probabilities.
Test of the discrimination power of 9 selected variables
Sum of squares
Degrees of freedom
Parameter values of FLEYE statistical model
For i = 0,1,2,3.
For i = 0,1,2,3.
Now, being PP 0 + PP 1 + PP 2 + PP 3 + PP 4 = 1
where IREG = 1 accounts for total regularity (WT eye), whilst IREG = 0 means total absence of regularity (degenerated eye). As a result, the bigger the probability of belonging to a low degeneration class (PP0 and PP1) the closer the index moves to 1, and vice versa. Intermediate values ranging from 0 to 1 will represent partial degeneration cases or rescued genotypes.
Although the large sample size used in our experiments is needed to develop and validate this method, FLEYE users should comply with standard experimental design guides to properly evaluate the sample size required in their experiments.
Fiji macro design
Sample validation and model testing
This set of experiments helped us to calculate a percent recovery for a statistically reasonable sample of flies of a given genotype (n ≥ 30 eyes/genotype). As an example, gmr > SCA1 Modifier#1 showed a 68% recovery as judged by comparing its median IREG with the WT and gmr > SCA1 IREGs. The IREG-based percent recovery slightly overestimates the recovery of the same modifier calculated from retinal thickness measurements (see Figures 1 and 7). This difference is expected since degeneration of lens and of the general architecture of ommatidial surface proceeds at a slower pace, in comparison with the fast disruption of retinal thickness due to massive photoreceptor cell death in this model of SCA1. These differences are in a tolerable range, and we are confident that the IREG values estimated by FLEYE and the derived percent recovery are valuable parameters to quickly assess the effect of genetic and pharmacological modifiers of a degenerative condition in the Drosophila eye model. Our method can also be used to appraise treatment variations in a genetically homogeneous fly sample as well as to estimate differences in genetic penetrance of a given genotype. In cases where full degeneration is achieved by the experimental process, resulting in eyes that yield few or no light reflections, FLEYE will assign “complete degeneration” values.
While performing the experiment to validate our method, we have estimated the total time that takes a researcher to analyze an experimental group of 30 flies with our FLEYE plugin: 3 hours from anesthetizing the flies until the IREG plot was statistically assessed and incorporated into a manuscript figure. This time contrasts to the several days needed to analyze retinal parameters from fixed tissues and pictures from SEM, TEM or standard histological sections [8-10]. Another advantage of our methodology is the number of ommatidia explored in each ROI of a single fly (over 200/sample), in comparison with the 30–40 ommatidia counted in each eye when using the deep pseudopupil technique .
In contrast to the complex and non-automated methodologies mentioned above that evaluate retinal patterning, a recent work has implemented an automated quantification of the structural features of eye surface obtained from SEM pictures of fly heads using edge detection and boundary-walking algorithms . This method allows for proper statistical analysis of phenotypic differences in fly eye surface, attaining a proper quantitative assessment of retinal distortions. Furthermore, the software using the method algorithm is also freely available. As an advantage, our FLEYE approach is based on the unbiased development of a statistical model that selects the most robust variables accounting for the differences observed between sets of degenerate and wild-type retinas. Moreover, Caudron et al. methodology uses SEM pictures, which imply a lengthy histological procedure and the need of a scanning electron microscope.
Finally, although in the current FLEYE protocol we rely on a simple but manual fly immobilization and image acquisition method, as well as a user-based ROI selection, we can envision simple automatic setups to take well focused and reproducible eye surface pictures of anesthetized flies combined with an easily implemented automatic ROI selection in the near future. Also our method does not require decapitation of flies and can be used to follow the progression of single flies throughout life provided that the immobilization method is reversible. These developments make our method a good candidate for full automation and therefore for potent high throughput screenings in search of therapeutic agents for neurodegenerative diseases.
In this work we present a novel, easy, and fast method to quantitatively rate the degeneration level of the compound eye of fruit flies with a high degree of reliability and robustness. This new method is based on the acquisition of images from the surface of the eye, the use of automated image analysis tools and a classification algorithm sustained on a built-in statistical model.
The easy-to-use properties of our method, plus the potential to be fully automated, make it a valuable tool for unbiased quantitative estimations of degeneration degrees in genetic or pharmacological screenings using the Drosophila retina as a model system.
In addition, the FLEYE plugin, following the adjustment of model parameters and grid size, could easily be adapted to evaluate the pattern regularity (and their experimental or pathological disruption) of different biological or non-biological origins: from images of crystal structures or beehives, to those of mammalian retina with patterns of cone photoreceptors, or neurons in histological sections of stereotypically organized brain structures.
- Jaiswal M, Sandoval H, Zhang K, Bayat V, Bellen HJ. Probing mechanisms that underlie human neurodegenerative diseases in Drosophila. Annu Rev Genet. 2012;46:371–96.View ArticlePubMed CentralPubMedGoogle Scholar
- Lenz S, Karsten P, Schulz JB, Voigt A. Drosophila as a screening tool to study human neurodegenerative diseases. J Neurochem. 2013;127:453–60.View ArticlePubMedGoogle Scholar
- Ambegaokar SS, Roy B, Jackson GR. Neurodegenerative models in Drosophila: polyglutamine disorders, Parkinson disease, and amyotrophic lateral sclerosis. Neurobiol Dis. 2010;40:29–39.View ArticlePubMed CentralPubMedGoogle Scholar
- Green EW, Giorgini F. Choosing and using Drosophila models to characterize modifiers of Huntington’s disease. Biochem Soc Trans. 2012;40:739–45.View ArticlePubMedGoogle Scholar
- Yu Z, Bonini NM. Modeling human trinucleotide repeat diseases in Drosophila. Int Rev Neurobiol. 2011;99:191–212.View ArticlePubMedGoogle Scholar
- Mishra M, Knust E. Analysis of the Drosophila compound eye with light and electron microscopy. Methods Mol Biol. 2012;935:161–82.View ArticleGoogle Scholar
- Song W, Smith MR, Syed A, Lukacsovich T, Barbaro BA, Purcell J, et al. Morphometric analysis of Huntington’s disease neurodegeneration in Drosophila. Methods Mol Biol. 2013;1017:41–57.View ArticlePubMedGoogle Scholar
- Pandey UB, Nie Z, Batlevi Y, McCray BA, Ritson GP, Nedelsky NB, et al. HDAC6 rescues neurodegeneration and provides an essential link between autophagy and the UPS. Nature. 2007;447:859–63.View ArticlePubMedGoogle Scholar
- Johnson RI, Cagan RL. A quantitative method to analyze Drosophila pupal eye patterning. PLoS One. 2009;4:e7008.View ArticlePubMed CentralPubMedGoogle Scholar
- Caudron Q, Lyn-Adams C, Aston JAD, Frenguelli BG, Moffat KG. Quantitative assessment of ommatidial distortion in Drosophila melanogaster. Dros Inf Serv. 2013;96:136–44.Google Scholar
- del Caño-Espinel M, Acebes JR, Sanchez D, Ganfornina MD. Lazarillo-related Lipocalins confer long-term protection against type I Spinocerebellar Ataxia degeneration contributing to optimize selective autophagy. Mol Neurodegeneration. 2015;10:11.Google Scholar
- Schindelin J, Arganda-Carreras I, Frise E, Kaynig V, Longair M, Pietzsch T, et al. Fiji: an open-source platform for biological-image analysis. Nat Methods. 2012;9:676–82.View ArticlePubMedGoogle Scholar
- Fernandez-Funez P, Nino-Rosales ML, de Gouyon B, She WC, Luchak JM, Martinez P, et al. Identification of genes that modify ataxin-1-induced neurodegeneration. Nature. 2000;408:101–6.View ArticlePubMedGoogle Scholar
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.