In Vivo Dynamics of Retinal Microglial Activation During Neurodegeneration: Confocal Ophthalmoscopic Imaging and Cell Morphometry in Mouse Glaucoma

Summary

Microglia activation and microgliosis are key responses to chronic neurodegeneration. Here, we present methods for in vivo, long-term visualization of retinal CX3CR1-GFP+ microglial cells by confocal ophthalmoscopy, and for threshold and morphometric analyses to identify and quantify their activation. We monitor microglial changes during early stages of age-related glaucoma.

Abstract

Microglia, which are CNS-resident neuroimmune cells, transform their morphology and size in response to CNS damage, switching to an activated state with distinct functions and gene expression profiles. The roles of microglial activation in health, injury and disease remain incompletely understood due to their dynamic and complex regulation in response to changes in their microenvironment. Thus, it is critical to non-invasively monitor and analyze changes in microglial activation over time in the intact organism. In vivo studies of microglial activation have been delayed by technical limitations to tracking microglial behavior without altering the CNS environment. This has been particularly challenging during chronic neurodegeneration, where long-term changes must be tracked. The retina, a CNS organ amenable to non-invasive live imaging, offers a powerful system to visualize and characterize the dynamics of microglia activation during chronic disorders.

This protocol outlines methods for long-term, in vivo imaging of retinal microglia, using confocal ophthalmoscopy (cSLO) and CX3CR1^GFP/+ reporter mice, to visualize microglia with cellular resolution. Also, we describe methods to quantify monthly changes in cell activation and density in large cell subsets (200-300 cells per retina). We confirm the use of somal area as a useful metric for live tracking of microglial activation in the retina by applying automated threshold-based morphometric analysis of in vivo images. We use these live image acquisition and analyses strategies to monitor the dynamic changes in microglial activation and microgliosis during early stages of retinal neurodegeneration in a mouse model of chronic glaucoma. This approach should be useful to investigate the contributions of microglia to neuronal and axonal decline in chronic CNS disorders that affect the retina and optic nerve.

Introduction

Microglia are neuroimmune cells that exclusively reside in the central nervous system (CNS) since early embryonic development and throughout adulthood. Equipped with a complex repertoire of receptors, microglial activity and regional heterogeneity are regulated by their bidirectional interplay with neighboring neurons, glia, blood-brain barrier and infiltrating neuroinflammatory cells^1,2. Basal microglial functions contribute to physiological maintenance and repair, as they sample their territory for perturbations in homeostasis^3,4. During CNS injury or disease, microglia are the first responders to neuronal signals that then trigger their transition to a reactive phenotype, termed “activated microglia^2,5-7. Microglia activation involves an intricate cycle of gene and protein expression, which are coupled to cell soma and process resizing and remodeling^6-9. Microglia activation, as well as cell redistribution and clustering, can be accompanied the local overall increase in the number of cells (termed microgliosis). This can result from cell proliferation and self-renewal, with or without recruitment of blood-derived monocytes^3,4,7,10-14. In a broad range of age-dependent, chronic CNS diseases, sustained microgliosis and microglia activation parallel disease progression^15-19. How microglia impact neurodegeneration remains unclear, mainly because they play both neuroprotective and deleterious roles that may have diverse contributions to disease onset and progression. Live imaging studies aimed at understanding chronic CNS disease have monitored microglial behavior in the damaged CNS of animal models and humans, and demonstrated that microglial alterations are detectable beginning at early disease stages^15-17,19,20. Thus, it is critical to develop approaches to detect and monitor microglial activation in vivo.

Non-invasive detection of regional changes in brain microglia activation was established as an important in vivo indicator of neurodegenerative disease progression, using molecular imaging or bioluminescence and positron emission tomography or magnetic resonance imaging^18,21,22. These highly quantitative and non-invasive molecular and nuclear imaging methods detect gliosis with regional resolution. Alternatively, two-photon confocal imaging in CX3CR1^GFP/+ mice has allowed the observation of brain microglia with cellular resolution^{3,4,9,20,23-28}. However, this approach limits long-term and repeated observation of chronic microglial alterations, given the potential risk of disturbing their behavior by even minimally invasive brain imaging procedures²⁹. Alternatively, the retina offers optimal conditions for direct, in vivo visualization and repeated monitoring of microglia in their intact CNS niche throughout aging, following acute injury, and potentially during chronic neurodegenerative diseases. Thus, recent studies have proved the feasibility of high-resolution imaging of retinal microglia expressing GFP by adapting the confocal scanning laser ophthalmoscopy (cSLO) to image live CX3CR1^GFP/+ mice. This has been used to track weekly changes in GFP+ cell numbers in individual mice for up to 10 weeks following acutely induced injury or ocular hypertension^30-36.

We have extended this approach to perform long-term imaging over several months, and quantitatively track changes in microglia activation based on soma size using morphometric analysis. Somal size was defined as a useful metric of microglia activation in live imaging studies using two-photon confocal microscopy in cortical slices to perform in vivo imaging of CX3CR1-GFP+ microglia⁹. These and other studies also demonstrated the correlation between somal size and levels of Iba1 protein expression, which also increases with activation^9,37. Thus, activated microglia can be identified in live mice, and their numbers and distribution monitored over time during CNS health and disease.

This protocol describes methods for cSLO live image acquisition and analysis to monitor microglial cell numbers, distribution and morphological activation during retinal ganglion cell (RGC) degeneration (Figure 1). Thus, this study uses: 1) a mouse model of inherited glaucoma (DBA/2J) that undergoes age-dependent optic nerve and retinal neurodegeneration and shows remarkable variability in disease progression between 5 and 10 months of age^38,39, 2) monthly cSLO in vivo imaging for long-term visualization of GFP+ cells in the retina and unmyelinated optic nerve of heterozygous Cx3cr1^GFP/+ DBA/2J mice aged 3-5 months, 3) live imaging analysis by segmentation and thresholding to isolate cell somata and measure their area. These strategies are applied to assess the kinetics of retinal microglial activation states during early stages of chronic glaucoma.

Protocol

In vivo imaging is performed in pathogen-free facilities using protocols approved by the Institutional Animal Care and Use Committee at the University of Utah.
NOTE: This imaging protocol is used for reporter mice in which retinal microglia and infiltrating monocytes/macrophages express green-fluorescent protein (GFP) under the control of the fractalkine receptor locus (CX3CR1).

1. In Vivo Imaging of Retinal GFP+ Microglia by Confocal Scanning Laser Ophthalmoscopy (cSLO)

1. Turn on the water-controlled heating system, set to stabilize mouse temperature between 35-37 ºC during procedure, and connect two heating pads.
   1. Start the confocal scanning laser ophthalmoscope (cSLO) system (Figure 2A). Open the cSLO program, enter the information that will identify the individual mouse and set a corneal curvature of 2 mm (Patient Data menu).
2. Prepare for imaging. Securely fit a clean 55º wide-field objective lens on its socket. Prepare the ophthalmoscope imaging platform by securing a heating pad covered with clean paper. Use clips to flatten the pad and paper and keep them from blocking the movement of the objective lens.
3. Anesthetize the mouse by intraperitoneal injection of 1.3% 2,2,2-tribromoethanol and 0.8% tert-amyl alcohol (250 mg/kg body weight; 0.5 ml/25 g body weight) using a 30½ G needle fitted to a 1 ml disposable syringe. Return mouse to its cage, kept over a heating pad.
   NOTE: The delivery of the anesthetic by injection versus inhalation allows easier positioning of the mouse for imaging and unobstructed access to the eyes, free of nose cones and tubing.
4. Once the animal stops moving and is unresponsive to tail pinch, induce pupil dilation with Tropicamide and phenylephrine (0.5% each), by positioning the animal prone and covering both eyes with the drops for 7 min.
5. After 5 min, place the mouse prone on the center of the ophthalmoscope platform, and fit contact lenses over both eyes, applying minimal pressure.
   NOTE: Contact lenses are small and fragile but can be carefully handled with fingers, although forceps can be used instead⁴⁰. The use of contact lenses is optional, but recommended as it minimizes eye dryness and preserves corneal transparency during imaging.
6. After 7 min, orient the mouse with the right eye facing the objective lens and the orbit parallel to the objective lens, keeping the animal unrestrained near the edge of the platform (Figure 2B). To collect images of comparable saturation and orientation, constantly maintain the distance from eye to objective lens, as well as the alignment of the eye to the light path throughout imaging sessions. Clip long whiskers interfering with observation of the eye.
   NOTE: For the next 8-10 min, the mouse will not move or blink, and will breath with gentle movements, which are tracked by the cSLO Eye Tracking ART (automatic real time) mode, which adjusts the scan head position based on fundus landmarks with high contrast. Past that time, mice begin panting and blinking, making imaging impractical.
7. If imaging is performed without using contact lenses, apply PBS drops to both eyes every 2-3 min to prevent them from drying.
8. Collect a fundus image of the retinal vasculature and optic disc, focused on the inner retina.
   1. Select the infrared mode (IR) and adjust settings to 820 nm excitation, 100% laser power, 40-60% sensitivity (Figure 2C).
   2. Working in high-speed mode (12.6 frame/sec), locate the eye using the joystick to drive the ophthalmoscope. At low magnification, inspect the cornea and lens for injury or opacity, and exclude from the study eyes with defects or injury that will affect imaging the retina (Figure 3A).
   3. Locate the optic disc area (the surface of the optic nerve head, ONH) by bringing the objective closer to the eye, then center the image on it and lock this position by screwing the joystick. This alignment of the eye to the ophthalmoscope is key to obtain even focus and emission across the image.
   4. Visualize the inner planes of the retina, as well as the ONH, using the major blood vessels located on the vitreal surface of the retina as a reference focal plane (59 and 60 D), or deeper (55 D) in eyes showing excavated optic discs. The optimal focus should resolve circulating blood cells within major blood vessels, with their lumen and walls clearly distinct.
   5. Optimize image saturation by adjusting the dial on the touch screen control panel, until a white halo of uniform illumination spans in most of the 55º fundus field around the optic disc (using slight overexposure). Next, lower the saturation to optimize contrast until blood cells moving along the major vessels can be clearly resolved. If focal dark areas persist, not due to retinal damage, realign the eye to the camera until a uniformly bright image is obtained. This adjustment is fundamental to capture reproducible sets of images in position and quality.
   6. Collect a high-resolution IR image of the central retina (1.4-1.7 mm in diameter, depending on age), averaging 30 frames in real time (4.7 frames/sec; normalized), to improve signal-to-noise ratio (Figure 2D).
9. Immediately, acquire a fluorescence image of GFP+ microglia and/or infiltrating monocytes localized to the inner planes of the retina and the ONH.
   NOTE: Optimization of the IR fundus image of the retina determines the resolution of the fluorescence image of GFP+ cells, as well as the extent of inner retina spanned. Failure to focus the inner planes of the retina, which can be detected by poor resolution of the vasculature, will result in dimmed GFP+ cell views (Figure 3B). Failure to adjust IR saturation correctly and uniformly affects the fluorescence image (FA), introducing artifactual variations in GFP+ cell intensity and size (Figure 3C).
   1. Swich to fluorescence imaging mode (FA) in the touch screen panel, using blue, 488 nm laser excitation (460-490 nm barrier filter set), and acquisition settings (100% laser power and 100-125% sensitivity), which are maintained constant for all mice (Figure 2E).
   2. Collect a single xy-point, bidimensional fluorescence image of the retina, and immediately capture a fundus image at identical position (100x scan average; 55º scan angle for both images; Figure 2F).
   3. Capture a multipoint image by selecting “composite” in the control panel (Figure 2G) and moving the ophthalmoscope right and left, panning the retina across the nasal-temporal axis (Figure 2H).
      NOTE: After scanning a single xy-point image (100x scan average), the software automatically averages new scans of the same and new areas (circumscribed with a green circle during optimal scan, or red when the scan is unfeasible), and stitches all xy-positions in real time. The resulting composite image spans 1.5-1.7 mm on the horizontal axis x 3-4 mm in the vertical axis (55º x 120-124º scan angle).
10. Complete imaging of each mouse within 15 min after inducing anesthesia, before blinking and motion restarts.
11. Remove contact lenses, hydrate eyes with PBS and return the animal to its warmed cage, checking behavior until full recovery of motility.
12. For longitudinal studies using intermittent cSLO-imaging sessions in the same individual mouse, repeat imaging no less than 3 days apart to avoid the cumulative toxicity of anesthesia, as well as corneal irritation.

2. Live Image Processing and Analysis

1. Sequential image alignment:
   1. Align the fundus images for a time series of the same eye using the vasculature and optic disc as landmarks. Open all images in a commercial slide-show presentation program, by dragging each image over the preceding one until their optic discs align on the xy-plane (the top image appears semi-transparent while being dragged), then rotate the top image until its major blood vessels overlap with the vasculature on the image below (focus on vessels farthest from the optic disc). Record the rotation angle for each image and save this time series for future reference, keeping track of mouse and eye identity.
   2. Align the corresponding series of single xy-point fluorescence images by applying the recorded angle to correct the x-y rotation at each time point, using a raster graphic editor program. Additionally, adjust resolution to 300 dpi (without rescaling) and background (in the RGB mode, select Levels and sample the lumen of a blood vessel as black). Save these images as TIF files, adding “aligned” to their names.
2. Microglial cell segmentation:
   1. To identify GFP+ cells in the central retina, open in FluoRender the “aligned” TIF file corresponding to the earliest time point/age. Zoom the image to fit the screen, then define the Render view (composite, orthogonal, interpolated) and its properties (0.58 gamma, 255 saturation point, 255 luminance, 195 alpha and 1.00 shading), selecting both the RG channels as visible (hide B channel by double-clicking it in the Workspace frame; Figure 4A).
   2. To select individual cells, threshold the red channel (must be highlighted on the Workspace frame), by increasing the low threshold until the majority of cells are masked (white) with minimum overlap and cell processes (cyan), while keeping the high threshold constant (255). The low threshold value varies between 100 and 170, depending on the fluorescence intensity (Figure 4B).
   3. Save this view with the red channel visible only (Capture command) as a new TIF file, adding “cellsegm” to its original name (Figure 4C). Using a generic raster graphic editor software, invert its greyscale and adjust resolution to 300 dpi as above, and save again as RGB (Figure 4D).
      NOTE: Delete the folders (projects) automatically created by FluoRender, and save the applied threshold for future reference.
3. Microglial soma segmentation and morphometry:
   1. To identify GFP+ cell somata for area quantification, use image analysis software with automated intensity measurements and threshold capabilities, as well as threshold-based object counting and measurement. Open the “cellsegm” TIF file, and select the rescaling method (spline) and the red channel (click red tab in RGB). Improve visualization by deselecting “view channel in color” (right-click the red RGB tab), and selecting “complement colors” (Image/Adjust Image) to invert the greyscale and see the cells in grey/black over white background (Figure 4E).
   2. Calibrate the image to 1.4-1.7 mm depending on age, by drawing a horizontal line across the entire image (Calibration, Quick Calibration).
   3. Segment individual cell somata by applying intensity thresholding (Figure 4B). For this, work on the red RGB channel and open the Threshold Intensity function (Object Count; selecting “count update” command) in order to track bidimensional parameters for each thresholded region of interest (ROI) representing individual somata.
   4. Define the intensity threshold in the intensity histogram, by selecting the lowest 50-60% of the 255 levels, and refining the final threshold value by visually assessing the overlap between the threshold color mask (blue) and the cell soma perimeter (grey) in individual cells (Figure 4E).
   5. Estimate the accuracy of the cell soma masks to represent the somal dimensions by measuring the area before and after segmentation in 10 cells and accept the selected threshold range if the measurements differ less than 10% (use the Binary Toolbar and Annotated Measurements).
   6. Manually identify somal ROIs that include multiple cells and separate these elements (eliminate or separate ROIs using the Binary Toolbar controls, or the Object Catalog) while toggling on/off the binary to directly visualize the cells (Figure 4F).
   7. Classify microglial cells as ROIs with somal areas larger and smaller than 50-60 µm² to discriminate cell somata corresponding to activated microglia and deactivated microglia respectively, by using Area Restriction.
      NOTE: Each cell somal subset is identified with a different pseudocolored binary layer, and can be further characterized for other morphological parameters (perimeter, circularity, etc.), as well as quantified within discrete retinal sectors (Figure 4G). Save the analyzed image as an ND2 file.
   8. Verify the process and branching complexity in individual activated microglial cells, by overlaying the somal mask and the single xy-point image (contrast increased 50%; Figure 4H). Manually count the processes directly extending from the cell soma or measure the diameter of the polygon encompassing the arbor (use the Binary Toolbar and Annotated Measurements).
      NOTE: Activated cells lack processes or bear few (<4) and short (<2x somal diameter) ones, in contrast to non-activated cell processes, which comprise a ramified and extensive arbor (> 3-10x somal diameter) surrounding their relatively small soma.
   9. Manually identify cells with somata smaller than <20 µm² and/or GFP expression below the detection limit, which are therefore undetected by intensity thresholding. Use the Taxonomy command in Annotations to count manually identified elements.

Results

Our recent in vivo studies used these live image acquisition and analysis methods to visualize and track the kinetics and patterns of ONH and retinal microglial changes during early stages of chronic glaucoma and their relationship to late neurodegeneration⁵⁹. Here we illustrate a cSLO image acquisition protocol to visualize microglial cells across a large area of the central retina in individual young heterozygous Cx3CR1-GFP DBA/2J retinas (Figure 5). Based on the high cell resolutio...

Discussion

Live monitoring of microglial cell number and morphological activation during a neurodegenerative disease requires the use of non-invasive imaging methods that allow the detailed visualization of cell features. After imaging, microglial cells must be isolated (segmented) for morphometric analysis by use of multiple threshold steps to assess somal size and/or process complexity as readouts for microglia activation. In this protocol, we describe methods for live image acquisition using cSLO, and quantitative analysis of mi...

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
DBA/2J and CX3CR1-GFP/+ mice	The Jackson Laboratory, Bar Harbor, ME	000671 and 00582	Mice are bred in house, introducing new breeders every 3 - 4 generations to prevent genetic drift.
30½ G needle and 1 ml tuberculin slip-tip syringe	BD, East Rutherford, NJ	305106 and 309659
2,2,2-tribromoethanol, tert-Amyl alcohol 99% and phosphate buffer saline tablets	Sigma-Aldrich, St. Louis, MO	T48402, 152463 and P4417	Avertin solution (pH 7.3, sterile filtered) must be freshly made solution or stored at 4 °C for up to 1 week.
Heat therapy T/Pump	Gaymar Industries, Orchard Park, NY	TP-650
Cotton-tipped applicators	Fisher Scientific, Pittsburgh, PA	23-400-100
Tropicamide 1%	Bausch & Lomb, Rochester, NY	NDC 24208-585-64
PMMA contact lenses, 1.70 base curve, 3.2 mm total diameter, 0.40 mm thick center	Cantor & Nissel Ltd., Northamptonshire, UK	G003709	Lenses are rinsed with sterile PBS and stored in polypropylene boxes.
HRA/Spectralis confocal scanning laser ophthalmoscope and Eye Explorer software	Heidelberg Engineering GmbH, Carlsbad, CA	Version 1.7.1.0
PowerPoint	Microsoft, Redmond, WA	Version 14.4.3
Adobe Photoshop	Adobe, San Jose, CA	Version CS3
FluoRender	Scientific Computing and Imaging Institute, University of Utah	Version 2.13	Freeware. http://www.sci.utah.edu/software/13-software/127-fluorender.html
NIS-Elements C	Nikon, Melville, NY	Version 4.30.01