- Open Access
IPLaminator: an ImageJ plugin for automated binning and quantification of retinal lamination
© Li et al. 2016
- Received: 27 August 2015
- Accepted: 1 January 2016
- Published: 16 January 2016
Information in the brain is often segregated into spatially organized layers that reflect the function of the embedded circuits. This is perhaps best exemplified in the layering, or lamination, of the retinal inner plexiform layer (IPL). The neurites of the retinal ganglion, amacrine and bipolar cell subtypes that form synapses in the IPL are precisely organized in highly refined strata within the IPL. Studies focused on developmental organization and cell morphology often use this layered stratification to characterize cells and identify the function of genes in development of the retina. A current limitation to such analysis is the lack of standardized tools to quantitatively analyze this complex structure. Most previous work on neuron stratification in the IPL is qualitative and descriptive.
In this study we report the development of an intuitive platform to rapidly and reproducibly assay IPL lamination. The novel ImageJ based software plugin we developed: IPLaminator, rapidly analyzes neurite stratification patterns in the retina and other neural tissues. A range of user options allows researchers to bin IPL stratification based on fixed points, such as the neurites of cholinergic amacrine cells, or to define a number of bins into which the IPL will be divided. Options to analyze tissues such as cortex were also added. Statistical analysis of the output then allows a quantitative value to be assigned to differences in laminar patterning observed in different models, genotypes or across developmental time.
IPLaminator is an easy to use software application that will greatly speed and standardize quantification of neuron organization.
Stratification of neural processes is a critical aspect of development that can promote specific patterns of connectivity and function. Many neuron cell types are also identified in part by the pattern in which their axons and dendrites stratify. The retina is part of the central nervous system where stratification, in the retina termed lamination, is perhaps most pronounced. Lamination of axons and dendrites occurs in the two neuropil layers of the retina, the relatively simply outer plexiform layer (OPL) and the more complex inner plexiform layer (IPL) . The outer plexiform layer contains synapses between photoreceptors and cells of the inner retina, while the inner plexiform layer contains the synapses of inner retinal neurons and retinal ganglion cells, the output cells of the retina. The IPL is functionally and anatomically subdivided into ON and OFF halves, which generally contain synapses responsive to light (ON) or active in the absence of light (OFF).
In addition to its functional implications, the stratification pattern in the IPL is often used to identify and describe the population of the bipolar [2–4], amacrine [5–7] and retina ganglion cells [8, 9]. Lamination of the retina is also a commonly used parameter when evaluating the function of genes during development of the retina . Current limitations to analysis of IPL lamination is the lack of a standard approach to quantifying laminar depth and the reality that mutations may result in changes to cell population densities that non-specifically alter the depth at which different populations of retinal neurons stratify. Automated approaches have been adopted to analyze optical coherence tomography images of the retina, but a standard approach to analyze lamination of retinal sections has not been developed . In this manuscript we describe an application developed as an Image-J based plug-in that is directly aimed at solving this issue. This application significantly reduces the work-load involved in quantifying retinal lamination and automates demarcation of laminar depth based on biological features, removing biases introduced by different genetic backgrounds and human error. Additional features allow the user control of division (binning) of the IPL or other neural tissues based on user preferences.
Software installation and operation
A detailed guide to the use of IPLaminator can be found in the additional files in the supporting information section (Additional files 1, 2, 3, 4, 5, 6 and 7: Figures S1-S7 and supporting information text (Additional file 8). Two sample images are included; S6 and S7). These files will walk the user through use of this software.
For quick reference the flow chart in Supplemental Figure 3 (Additional file 3: Figure S3) can be used: 1) The program prompts the user for a region of interest (ROI) containing the IPL and makes the rectangular selection tool the currently selected tool. 2) If “Add additional analysis region outside the IPL” is checked in the section menu the program prompts the user to select a point to the right of the already selected IPL ROI. This point represents the end of the additional analysis region (the region begins at the right edge of the IPL ROI). 3) If “Reduce background noise” is checked in the settings menu the program will prompt the user to select a background point on the image. The gray intensity is averaged at the selected point in a 3 by 3 region. This value is then saved and subtracted from the average intensity at each layer before the results are displayed. 4) If “Use percentile values to calculate layer boundaries” is selected in the settings menu, the positions of layer boundaries are calculated based on hard coded values. These values are represented as a percentile distance across the user selected ROI based on our measurement of SAC bands. The percentiles were determined experimentally and represent the average location of layer boundaries found in the IPL of wild type mice. 5) Depending on the stain type selected in the settings menu (ChAT or Calbindin/Calretinin) the program determines the location of two neurite stripes and a minimum between them or the location of three neurite stripes respectively. The process is illustrated in the following algorithm (Additional file 4: Figure S4). 6) If the layer boundaries are not being calculated using percentile values they are calculated using the boundaries of the IPL ROI and the locations of three biological markers described above. The layer boundaries (one through twelve) are calculated as described in the methods section. 7) After the layer boundaries are established the average intensity is calculated for each layer in each image selected for analysis. 8) At this point in the program all analysis is finished and results are generated. The following fields are automatically saved to a text file in the default output directory previously chosen by the user: • Layer Number – The given number of each layer, ascending from layer adjacent to RGC to layer adjacent to INL then additional area if selected. • Layer Depth – The location of each layer, the distance in pixel from the side of the IPL that borders the RGC. • Layer width – The width of a particular layer. • Intensity – Average gray scale intensity for each layer in each analyzed image. • Normalized Intensity – Normalized intensity is the average intensity for each layer/image minus 99 % of the lowest non-zero layer intensity value on that image. • Intensity minus background – The average intensity for a particular layer/image minus the average intensity in the 3x3 region around the user selected background point. Only output if reduce background noise is selected. • Intensity % - Intensity % is the intensity at a given layer divided by the intensity at all layers. 9) If “display results histogram” is selected in the settings menu histograms will be displayed to illustrate the results. One histogram is displayed for each individual image and a single histogram is displayed with the combined results from all images. These histograms are created with the ImageJ ProfilePlot class.
IPL binning formula
The IPL is binned based on cholinergic amacrine cell neurite stratification (Additional file 5: Figure S5). Specifically, the two boundaries of the IPL are selected by the user, at the border of the IPL and the RGL and INL. 3 boundaries within IPL are generated based on the grayscale intensity profile of the cholinergic neurites, including 2 peak intensity locations and a lowest intensity location between these two peaks. These five locations are used to generate 10 sublayers.
The distance between the inner boundary of IPL (adjacent to RGL) and the location of the ON peak intensity is divided into 7 layers. Two of these layers starting from the RGL are then merged into a sublamina, which gives 3 sublamina in the bottom of the ON layer (layer 1–3). The 7th layer is adjacent to the peak intensity of the ON cholinergic band facing RGL and this layer is added to 1/4 the distance towards the location of lowest intensity between the peaks from layer 4 and covers most ON cholinergic band. The next 1/2 of the distance between the ON cholinergic band and the middle point between the two cholinergic neurite bands is layer 5. The last 1/4 of the distance is added to 1/4 of the distance towards the OFF cholinergic peak intensity and defines layer 6. Next 1/2 of the distance between the least intensity between the cholinergic neurite bands and the OFF cholinergic peak intensity is defined as layer 7. The last 1/4 of the distance adjacent to OFF cholinergic peak is added to 1/5 of the distance towards the outer boundary of the IPL, and forms layer 8, covering most of the OFF cholinergic band. The remaining 4/5 of the distance towards the IPL boundary is divided into 2 layers of equal thickness; layer 9 and 10. Each of these divisions represents close to 10 % of the IPL, as shown in the results section.
Animal care and handling
Ad libitum fed mice were maintained on a mixed C57 BL/6 J and C3H/HeJ background under a 12 h light:dark regimen. Wild type, Dscam LOF and Bax −/− mutant mice  were used in this study. All animal procedures performed on mice in this study were approved by the University of Idaho Animal Care and Use Committee. Genotyping was performed as previously described according to instructions from The Jackson Laboratory.
Immunocytochemistry, immunohistochemistry and antibodies
Mice were perfused with PBS. Whole eyes were marked by making a small burn on the dorsal side of the corneal and then fixed in 4 % PFA for 2 h at room temperature and washed overnight. Retinas were then cryo preserved, frozen in optimal cutting technology (OCT) media (Tekura Inc) at −20°. Tissue was stained as previously described .
Antibodies: goat anti-ChAT (Millipore; AB144P; 1:400), rabbit anti-calbindin (Swant; CB38a; 1:1000), rabbit anti-bNOS (Sigma-Aldrich; NZ280; 1:15,000), rabbit anti-TH (Millipore Bioscience Research Reagents; 1:500), mouse anti Syt2 (ZFIN; ZDB-ATB-081002-25; 1:500). DAPI reagent was mixed into the second wash after incubation with secondary antibodies at a dilution of 1:50,000 of a 1 mg/ml stock. Secondary antibodies were acquired from Jackson ImmunoResearch and used at a concentration of 1:1000.
An Olympus DSU confocal microscope was used to capture all fluorescent images. 20 X and 40 X objectives were used in this study with numerical aperture of 0.5 and 1.2 respectively. The final image resolutions were 0.4 and 0.2 μm per pixel. A Nikon epifluorescent microscope was used to capture images of H&E sections. To avoid immunoflourescent background, exposure rate was set on auto to minimize the noise from different channels. Images used in the figures were taken at exposure rate of 100 ms. Any modification to images, for example, to brightness, was performed across the entire image in accordance with the journal’s standards. Similar results were obtained using images collected from a variety of imaging platforms on our campus; all fluorescent microscope camera combinations captured sufficient images for analysis.
Convention of IPL strata division
Given the real physiological landmarks identified by cholinergic amacrine cell banding, these observations suggest that demarcating the IPL into five even layers roughly based on projection of Müller glia may not best represent the biology of the retina. Using cholinergic bands to demarcate and subdivide the ON and OFF halves of the IPL offers an attractive solution to this problem by functionally dividing the retina into a similar series of domains as envisioned by Cajal, with the added benefit of an easily reproducible set of landmarks. Further division allows the identification and distinction of spatially separated but distinct neurites, which would otherwise be classified as projecting to the same stratum of the IPL.
Automated strata delimitation of the IPL
Analysis of abnormal neuron stratification
Analysis of neuron projection outside of the IPL
Alternatives to ChAT staining
IPLaminator is a simple tool with a wide range of uses for analysis of lamination in the retina and other regions of the central nervous system. The data output of IPLaminator is primarily in percent values and reflects the amount of fluorescent intensity in a given layer of neural tissue. This output can then be statistically compared across genotypes using a statistical test optimized for comparison of percents, such as the Mann–Whitney U-test, or converted, for example by arc-sin conversion, for other statistical tests.
Biological limitations and considerations
Several biological considerations and limitations should be taken into account when assaying retinal lamination. The first of these is that the eye is a spherical structure and this analysis treats lamination across a flat plane. The angle at which the retina curves and thins from the central retina to the peripheral retina is small in adult mice but at earlier developmental stages and in models such as zebrafish larva the angle is greater and could result in the artifactual smearing of sharp lamination across multiple bands. A solution to this bias is to sample a smaller distance of IPL more frequently (to account for increased variability over a smaller distance).
Antibody staining quality is an obvious complication and can result in signal being spread over portions of the IPL that clearly do not have neurites projecting into them. IPLaminator measures intensity of fluorescence and not neurite projections per se, with the assumption that most staining will be concentrated in targeted neurites. Background subtraction across the image using the program’s background subtraction or before analysis can reduce the influence of background immunofluorescence but care must be taken to ensure inappropriate image manipulation does not occur at this stage, which could result in greater background subtraction from some subset of tissues. In practice we code genotypes and cut sections from different genotypes to be analyzed onto the same slide for staining. This helps to blind the analysis and minimize sample-sample preparation variability.
The presence of displaced cell bodies in the IPL can complicate analysis in several manners. The most distorting of these is if the soma is itself fluorescent. This would result in a large signal in the stratum in which the soma resided. Avoiding such areas or recognizing that the signal is coming from the cell soma is a necessary consideration. Two classes of cells that normally reside in the IPL include the soma of vasculature and microglia. Both of these cells have a tendency to nonspecifically fluoresce, especially when using antibodies generated in the species to be assayed.
The axon and dendrite stalk projecting to laminated neurite bands is also a consideration. These processes contribute to readout of signal and their differential staining in compared populations could result in mistaken interpretation of data. Using antibodies that limit this will increase resolution. For example antibodies to VAChT stain the SAC bands only, while ChAT stains the cell bodies and proximal and distal neurites of SACs . In cases where cell bodies are not displaced into the IPL either ChAT or VAChT will yield similar results because the peak intensities are used to bin the IPL. In cases where somata are displaced into the IPL antibodies to VAChT will avoid picking up signal from the displaced cell bodies in the IPL.
Technical limitations and considerations
Regarding background noise, both original intensity or with background subtraction, the software automatically detected minimal background intensity and intensity generated by user-selected background and all will be used to generate three clusters of results. The only differences between three outcomes is how much intensity has been removed from each channel globally because our software could not distinguish a pixel that is labeling neurons to a pixel that is a pure background noise. Users should carefully examine the outcomes and consistently use one of the three results to interpret original data.
IPLaminator is designed to optimize IPL neurite stratification analysis. It minimizes human operational error and observational bias, generates reliable and accurate data based on individual images to best describe how neurons projecting their neurites. Use of IPLaminator is intuitive with minimal amount of training time required. Once the image is set up correctly, users only have to select an area of interest and the software will automatically optimize the layer separation based on intensity displayed throughout the area. IPLaminator represents a technical and scientific improvement on Cajal’s early studies of the retina that will help to continue the mapping of the nervous system he started 120 years ago.
Project name: IPLaminatorProject homepage: http://isoptera.lcsc.edu/IPLaminator
Operating system: Windows, Mac, and Linux.
Programming language: R, Python and Java.
Other requirements: Image J or FIJI (Image J with auto plugin update version) is required for this program to run.
License: This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, version 3 of the License.
Any restrictions to use by non-acadamics: None
This research was supported by the National Eye Institute Grant EY020857 and P20 GM103408.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. 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.
- Wassle H. Parallel processing in the mammalian retina. Nat Rev Neurosci. 2004;5(10):747–57.PubMedView ArticleGoogle Scholar
- Euler T, Haverkamp S, Schubert T, Baden T. Retinal bipolar cells: elementary building blocks of vision. Nat Rev Neurosci 2014, 15(8):507-519.Google Scholar
- Ghosh KK, Bujan S, Haverkamp S, Feigenspan A, Wassle H. Types of bipolar cells in the mouse retina. The Journal of comparative neurology 2004, 469(1):70-82.Google Scholar
- Wassle H, Puller C, Muller F, Haverkamp S. Cone contacts, mosaics, and territories of bipolar cells in the mouse retina. The Journal of neuroscience: the official journal of the Society for Neuroscience 2009, 29(1):106-117.Google Scholar
- Masland RH. The many roles of starburst amacrine cells. Trends Neurosci. 2005;28(8):395–6.PubMedView ArticleGoogle Scholar
- Taylor WR, Smith RG. The role of starburst amacrine cells in visual signal processing. Vis Neurosci. 2012;29(1):73–81.PubMedPubMed CentralView ArticleGoogle Scholar
- Masland RH. The tasks of amacrine cells. Vis Neurosci. 2012;29(1):3–9.PubMedPubMed CentralView ArticleGoogle Scholar
- Morgan JL, Schubert T, Wong RO. Developmental patterning of glutamatergic synapses onto retinal ganglion cells. Neural Dev. 2008;3:8.PubMedPubMed CentralView ArticleGoogle Scholar
- Sernagor E, Eglen SJ, Wong RO. Development of retinal ganglion cell structure and function. Prog Retin Eye Res. 2001;20(2):139–74.PubMedView ArticleGoogle Scholar
- Yamagata M, Weiner JA, Sanes JR. Sidekicks: synaptic adhesion molecules that promote lamina-specific connectivity in the retina. Cell. 2002;110(5):649–60.PubMedView ArticleGoogle Scholar
- Chiu SJ, Li XT, Nicholas P, Toth CA, Izatt JA, Farsiu S. Automatic segmentation of seven retinal layers in SDOCT images congruent with expert manual segmentation. Optics express 2010, 18(18):19413-19428.Google Scholar
- Li S, Sukeena JM, Simmons AB, Hansen EJ, Nuhn RE, Samuels IS, Fuerst PG. DSCAM promotes refinement in the mouse retina through cell death and restriction of exploring dendrites. The Journal of neuroscience : the official journal of the Society for Neuroscience 2015, 35(14):5640-5654.Google Scholar
- de Andrade GB, Long SS, Fleming H, Li W, Fuerst PG. DSCAM localization and function at the mouse conesynapse. J Comp Neurol. 2014; 522(11):2609-33.Google Scholar
- Cajal RYS. La re’tine des verte’bre’s. Cellule. 1892;9:119–257.Google Scholar
- Nevin LM, Taylor MR, Baier H. Hardwiring of fine synaptic layers in the zebrafish visual pathway. Neural Dev. 2008;3:36.PubMedPubMed CentralView ArticleGoogle Scholar
- Coombs JL, Van Der List D, Chalupa LM. Morphological properties of mouse retinal ganglion cells during postnatal development. J Comp Neurol. 2007;503(6):803–14.PubMedView ArticleGoogle Scholar
- Duan X, Krishnaswamy A, De la Huerta I, Sanes JR. Type II cadherins guide assembly of a direction-selective retinal circuit. Cell 2014, 158(4):793-807.Google Scholar
- Famiglietti Jr EV. On and off pathways through amacrine cells in mammalian retina: the synaptic connections of “starburst” amacrine cells. Vision Res. 1983;23(11):1265–79.PubMedView ArticleGoogle Scholar
- Famiglietti Jr EV, Kolb H. Structural basis for ON-and OFF-center responses in retinal ganglion cells. Science. 1976;194(4261):193–5.PubMedView ArticleGoogle Scholar
- Ishii T, Kaneda M. ON-pathway-dominant glycinergic regulation of cholinergic amacrine cells in the mouse retina. J Physiol. 2014;592(Pt 19):4235–45.PubMedPubMed CentralView ArticleGoogle Scholar
- Matsuoka RL, Nguyen-Ba-Charvet KT, Parray A, Badea TC, Chedotal A, Kolodkin AL. Transmembrane semaphorin signalling controls laminar stratification in the mammalian retina. Nature 2011, 470(7333):259-263.Google Scholar
- Ford KJ, Feller MB. Assembly and disassembly of a retinal cholinergic network. Vis Neurosci. 2012;29(1):61–71.PubMedPubMed CentralView ArticleGoogle Scholar
- Famiglietti EV. Synaptic organization of starburst amacrine cells in rabbit retina: analysis of serial thin sections by electron microscopy and graphic reconstruction. J Comp Neurol. 1991;309(1):40–70.PubMedView ArticleGoogle Scholar
- Voigt T. Cholinergic amacrine cells in the rat retina. J Comp Neurol. 1986;248(1):19–35.PubMedView ArticleGoogle Scholar
- Zhou ZJ, Lee S. Synaptic physiology of direction selectivity in the retina. J Physiol. 2008;586(Pt 18):4371–6.PubMedPubMed CentralView ArticleGoogle Scholar
- Koulen P. Vesicular acetylcholine transporter (VAChT): a cellular marker in rat retinal development. Neuroreport. 1997;8(13):2845–8.PubMedView ArticleGoogle Scholar