Live Imaging and Characterization of Microglia Dynamics and Interactions with Synapses in Diseased Murine Retina

Summary

This paper demonstrates a method for visualizing and quantifying microglia motility and contact with postsynaptic puncta in the retina using spinning disk confocal microscopy.

Abstract

Microglia are the resident macrophages of the central nervous system (CNS) that respond to tissue infection and injury. In addition to their role in inflammation, microglia play a developmental role in circuit refinement through synaptic pruning. However, the mechanisms of synaptic pruning in neuroinflammation and neurodegeneration remain unknown. In this protocol, we use a mouse retina explant model to study microglia dynamics ex vivo. To examine microglia motility and their interactions with postsynaptic proteins, we label synapses with AAV-PSD95-RFP and record timelapse videos of motile microglia colocalized with postsynaptic proteins using spinning disk confocal microscopy. We then create surface and spot reconstructions of microglia and PSD95 using image analysis software. Data such as microglia displacement length, process speed, and contact with postsynaptic puncta can then be extracted from these surfaces to understand microglia behavior both in homeostatic states and after neuronal injury. This protocol can be useful in examining the role of microglia in synaptic pruning in retinal neurodegenerative diseases.

Introduction

Circuit function and homeostasis rely on highly regulated synapse firing and other unique interactions among various cell types¹. Serving as the resident macrophages of the central nervous system (CNS), microglia are a non-neuronal population in the retina that play a unique role in synaptic pruning. During CNS development, microglia prune non-functional or weak synapses to refine neural circuitry². Microglia in the developing brain survey their environment and are primed towards a neuroprotective role during critical periods³. In visual system development, microglia are key regulators in the development of cortical neurons and interneurons, in addition to exhibiting morphological changes during critical periods of synaptic remodeling⁴. Similarly, in the developing retina, microglia clear apoptotic cells to make way for new synaptic connections, confirmed with upregulation of homeostatic marker P2Ry12 during these clearing events⁵. However, recent findings highlight that microglia may not play a definitive role in refining the neural circuitry in the developing visual system⁶, emphasizing that there is controversy regarding the role microglia play in visual circuit development.

In injury contexts, microglia change morphology to an active state and can play a neurodegenerative role. For example, microglia in Alzheimer's disease models demonstrated increased engulfment of amyloid-beta plaques⁷ and synaptic markers such as SPH and PSD95⁸. In retinal neurodegeneration, many research groups have found microglia with engulfed synaptic material, but no clear evidence of active engulfment of intact synapses^{9,10,11,12,13}. For example, He et al. found that microglia increase in number and have increased volume of engulfed synaptic material in a model of retinitis pigmentosa⁹, but it is not clear whether there was phagocytosis of intact synapses versus clearance of debris. Additionally, traditional microglial classifications -- such as "resting vs. activated" or "M1 vs. M2" -- are oversimplified and increasingly inaccurate given modern insights. Microglia exist in a gradient of states, not discrete categories¹². Hence, Paolicelli et al. called for a shift from outdated binary classifications toward a more nuanced, rigorously defined nomenclature for microglial states to better understand and define microglia morphology and activity¹². Therefore, strategies to evaluate microglia function ex vivo and in vivo will augment the field's ability to discern microglia states and activity in both development and disease contexts.

While strategies exist to study microglia function both ex vivo and in vivo, ex vivo live imaging has several benefits. First, corneal clarity or cataract development will not impede retinal imaging, creating clear visualization of minute processes and synaptic puncta. Further, in vivo imaging requires more technical skill to set up, which makes ex vivo preparations higher throughput setup, resulting in more animals imaged per experiment. Despite the theoretical benefits that in vivo imaging may allow, high laser power (100-230 µW) is required to capture microglia process motility for longitudinal imaging^14,15. However, ex vivo imaging of microglia with a lower laser power and adequate imaging media can provide the same imaging quality as in vivo imaging with the right preparation, although the tissue is no longer in situ.

We have demonstrated that microglia increase in number, complexity, and process movement after transient intraocular pressure (IOP) elevation¹⁰. In fixed tissue, this microgliosis results in increased colocalization with synaptic proteins¹⁰. Previous work has also demonstrated early synapse loss and dendrite degeneration of retinal ganglion cells (RGCs) in the murine laser-induced ocular hypertension model (LIOH) and other neurodegenerative models^{10,16,17,18,19,20}. However, it is still unknown whether they play an active role in pruning synapses or a more passive role in clearing cellular debris, including disassembled synapses. As synapse disassembly is an early hallmark of neurodegeneration, understanding microglia dynamics and their role in synapse disassembly is critical^8,10. Here, we use live imaging to understand microglia dynamics and their interaction with excitatory postsynaptic density protein-95 (PSD95) on the dendrites of RGCs.

Protocol

Animals were housed at the University of California, San Francisco, under light cycles of 12 h light and 12 h dark and given water and standard diet ad libitum, unless specified. All procedures were approved by the Institutional Animal Care and Use Committees at the University of California, San Francisco. See the Table of Materials for materials and instruments used in this protocol. An overview of the experimental procedure, from intravitreal injection to imaging and analysis, is shown in Figure 1.

1. AAV-PSD95-TagRFP virus production and efficiency

1. Construct AAV-PSD95-tagRFP by taking the PSD-95pTagRFP sequence (Addgene plasmid #52671; RRID:Addgene_52671) and inserting it downstream of the CMV promoter in AAV-CMV-GFP after removing the GFP (Addgene plasmid #67634; RRID: Addgene_67634).
2. Prepare the virus using triple-transfected HEK293T, purified using PEG/chloroform, and concentrated to a titer of 10¹² ip/mL (as described in Negrini et al.)²¹.
   NOTE: This viral production method typically produces a lower titer than ultracentrifugation, but for these experiments, it is desirable to have sparsely infected cells²¹. The capsid/serotype version used in these experiments was 7m8, (Addgene plasmid # 64839; RRID: Addgene_64839).
   1. Test the efficiency of the AAV before proceeding with the full imaging protocol.
      NOTE: Readers who implement this protocol should aim for sparse labeling of RGCs, as the goal is to identify and isolate individual RGC dendrites.
3. Verify successful transfection and then, fix the retina in 2% PFA in 1x ACSF. For synaptic proteins, check the signal under a confocal microscope at 63x to verify clear resolution of the fluorescent labeling.
   NOTE: The lighter concentration of fixation helps to preserve the labeling for synaptic proteins. If the AAV used is not for synaptic components, readers can choose to use 4% PFA instead.

2. Mouse husbandry and AAV-PSD95-TagRFP intravitreal injections

1. Cross male and female CD-1 albino mice (strain 022) with B6.129P2(Cg)-Cx3cr1tm1Litt/J (strain 005582) to produce CD-1 albino mice with EGFP-expressing microglia. Wait for a minimum of four generations of backcrossing to use the male and female Cx3cr1 heterozygous mice in experiments. Maintain and refresh the strain after 10 generations back onto the CD-1 background to prevent genetic drift.
2. Anesthetize animals (P30-P60) using isoflurane anesthesia through an induction chamber using 3-5% anesthetic in oxygen with a flow rate at 1 L/min of oxygen. Place the animal on a stereomicroscope, while ensuring isoflurane is continuously administered through a nose cone. To confirm the animal is anesthetized, pinch the toe at least 3x to confirm.
3. Add a drop of 0.5% proparacaine before injecting the AAV. Inject 2 µL of AAV-PSD95-RFP slowly intravitreally in both eyes at the superotemporal limbus using a 26s G 10 µL syringe. Slowly remove the needle to prevent backflow of the AAV through the injection site. Apply antibiotic ointment to the eyes post-injection.
4. NOTE: The needle should go no more than halfway through the circumference of the vitreous chamber. By doing so, the retina and lens can be damaged with the needle tip, leading to a patchy or even unsuccessful transfection. Cataract development or hemorrhage is indicative of faulty injection technique.
5. After injection, confirm the animals are alert before returning them to their original cages and housing rooms. If an animal exhibits postinjection pain, administer buprenorphine via intraperitoneal injection. Wait 3 weeks for AAV expression. If an inducible model is used to elevate intraocular pressure (or perform other manipulation), wait 3 weeks before inducing elevated intraocular pressure in one eye, using the contralateral eye or the eyes of a naive animal as control.

3. Euthanasia and retinal flatmounting

1. Prepare a fresh 1x solution of artificial cerebrospinal fluid (ACSF, pH 7.4) in 1x PBS containing the following components in mM: 119 NaCl, 2.5 KCl, 1.3 MgCl₂·6H₂O, 2.5 CaCl₂·2H₂O, 1 NaHPO₄, 11 glucose, and 20 HEPES).
2. Oxygenate the solution for at least 15 min before using it for acute dissection. Connect the tubing to an oxygen tank, and place a cap with a hole in the center to prevent the gas from escaping.
3. Euthanize the mouse with at least 1 mL of isoflurane anesthesia in a small induction chamber, followed by cervical dislocation. Mark the nasal side of the cornea with a marker before enucleating so that the retina location can be tracked throughout the experiment.
   1. To enucleate microbead-injected or lasered eyes, perform a limited conjunctival peritomy and place the scissors behind the eye to dissect the eye out at the base of the optic nerve. This allows the delicate eye to remain intact.
4. Using a dissection microscope, poke a small hole with a 23 G or 25 G needle in the cornea, which allows oxygenated ACSF to perfuse the eye. Insert scissors into this hole to make a fiduciary cut at the nasal end.
5. Cut off the rest of the cornea and iris at the limbus.
6. Remove the iris and lens. If the retina is sticking to the lens, separate it using forceps.
7. Separate the retina from the sclera by using forceps to gently dissect between the RPE and neural retina.
8. Place the retina on the non-gridded side of a Mixed Cellulose Esters (MCE) membrane filter paper for clear visualization and imaging.
9. Make small relaxing cuts in the retina at the other three quadrants, with the longest cut in the nasal retina for tracking retinal location.
   NOTE: The retina should look like a four-leaf clover.
10. Transfer the retina from the filter paper onto a laboratory wipe. Use one drop of ACSF to moisten the retina and then use the wipe to dry the filter paper. Repeat for at least 8-10 drops to make the retina adhere to the filter paper.
11. Place the retina in an oxygenating chamber while dissecting the contralateral eye.
12. Once ready to image, dry the filter paper as mentioned in step 3.10 and invert the retina onto a Matek Petri dish. Weigh down the retina with either metal beads or a ring. Add 3-5 drops of ACSF.
    NOTE: Small metal beads, such as those from a metal bead bath or a ring, are of sufficient size and weight to prevent drift without disrupting retinal integrity.
13. Fashion a ring by taking a steel wire and stretched nylon strings, gluing and adding a weight to result in a ring with fixed, taut parallel strings.
    1. Wrap the wire around a 1/2 inch diameter cylinder and pull tautly with pliers. Cut each ring from the cylinder and flatten between a vice.
    2. Sand the ring on one face to have a rough, flat side. Stretch nylon strings over the opening of a bottle with 1-2 inch diameter, and hold in place via rubber bands.
    3. Place the ring atop the nylon strings and is carefully glue it to the nylon strings by applying small amounts of glue along the inner perimeter of the ring. Afterward, place a light weight atop the ring, for example, a coin or thin metal wire, to hold the ring in place.
    4. After drying, cut the nylon strings from the outer rim of the ring using precision scissors. The result is fixed parallel strings tautly extending the diameter of the ring.
    5. If the nylon string density is too high within the ring, cut individual strings with microdissection spring scissors to suit the needs of the experiment.
       NOTE: The dissection and mounting process should only take ~5 min. The faster it takes for the tissue to be placed in equilibrated physiological conditions ex vivo, the healthier the tissue.

4. Spinning disk confocal acquisition

NOTE: Turn on the humidifying chamber at 5% CO₂ and run it for at least 1 h before imaging. Allowing the system and the sample to reach thermal equilibrium will help reduce drift.

1. Start scanning through the eyepiece at 10x to focus on the microglia. When scanning the tissue for areas to image, use the microglia channel as a reference, and focus only on microglia that have bright synaptic puncta colocalized (within 10-20 µm). As AAV labeling is not homogeneous throughout the tissue, scan areas that have both microglia and puncta colocalized within the same plane. Once an area meeting these criteria has been located, switch to 60x to focus, and switch to camera mode by clicking on the camera button.
   1. Using the imaging system software, right-click on the screen to find xy-navigation. A preview screen will appear that can scan the tissue by making a 3 x 3, 5 x 5 frame, or a custom size to screen for optimal areas. Use the 3 x 3 or 5 x 5 frame size and screen with one reference channel to prevent continuous photobleaching.
      NOTE: For this protocol, we used the microglia channel.
2. Image the mounted retina promptly using a spinning disk confocal microscope featuring a 60x objective (NA 1.49).
3. To preview the sample, right-click on the screen, select Live, and click the auto-adjust button so the system can adjust the view based on the current LUTs. To adjust laser settings, select each channel individually first and adjust laser power and exposure as needed. Then, right-click on the fluorophore to Assign Settings to Microscope and save the custom settings, as the software will not save it when switching to another channel.
   1. Switch to the next channel and adjust as needed. Right-click on the fluorophore to save settings. If acquiring multiple channels in the same frame, enter in the same values used after previewing each channel individually. Right-click on the fluorophore to save the settings.
   2. For multi-channel acquisition, select the channels of interest under the Lambda tab. For live imaging, combine all channels into one channel for fast capture of fine movement from the cell of interest to allow for each channel to be acquired simultaneously, rather than separately, which can create a lag in the timelapse depending on the speed of the cell.
4. Proceed with acquiring a timelapse for the desired time.
   1. In the time tab on the bottom left, enter a 5-10 min time lapse for time with an interval of 20 s.
   2. Under the acquisition settings tab on the right, set the image resolution to 608 x 832 with a pixel size of 0.12 µm.
   3. Collect images in 10 - 20 µm using a 0.3 µm step size to ensure an acceptable resolution of fine microglia processes and synaptic puncta along the z-axis. To ensure successful cell tracking, keep the time interval between frames at 20 - 30 s.
      NOTE: Due to the fragility of the tissue ex vivo, it is recommended to image the experimental retina first and for no more than one hour. This is to normalize the condition of the contralateral control eye, which should be sitting in an oxygenated chamber while imaging the first retina.
5. Use this protocol to image the more superficial layers of the retina. By acquiring time lapses larger than 20 µm, it becomes more difficult to capture fine process movement during the 20 s intervals. Therefore, only image microglia that fit within the 20 µm stack.

5. Tracking analysis and data export

1. Surface rendering for microglia
   1. Convert the timelapse file into IMS format using the file converter software. Input the following voxel sizes based on the image properties acquired from step 4.2: x = 0.1272423 micron, y = 0.1272423 micron, z = 0.3 micron.
   2. Once launched, the analysis software will start in Arena. Click on the Observe folder to open the previously converted file. Ensure the image is in 3D view.
   3. To detect whole individual microglia, click on the blue surface icon in the Object Toolbar. Select the last three options under Algorithm settings: Track Surfaces (over Time), Classify Surfaces, and Object-Object Statistics. Click on the blue play button to go to the next step.
      1. In case of multiple microglia, select Segment only a Region of Interest (ROI) to ensure individual microglia exist within one surface object. Set the x, y, and z parameters accordingly to only make a surface of the individual microglia of interest.
   4. Select the channel for which a surface rendering will be created. If desired, select Smooth to smooth the surface creation and use a 1:1 - 1.25 ratio to the voxel size of the image. For example, if the image has a 0.12 x 0.12 x 0.3 µm voxel size, set the surface detail to 0.15 - 0.16 µm. Under thresholding, select Machine Learning Segmentation to train the software in detecting the cells of interest.
      NOTE: This next step consists of feeding training data to Imaris AI through Machine Learning Segmentation.
   5. Under Training Data, observe the two paintbrush tools: Background (for detecting background) and Foreground (for detecting the true signal of the sample). Draw paintbrush strokes by selecting the desired paintbrush and using the Shift + Left click. Ensure Interpolate Display is selected under Settings. Select Train and Predict to enter training data from the paintbrush strokes made.
      NOTE: Feeding more than six strokes results in a longer processing time for the AI and inaccurate capture of the cell.
      1. Use the yellow pointer in the middle to move across the z-stack. Draw multiple paintbrush strokes across the x, y, and z planes, as well as in a few (not all) time frames of the timelapse. Make 3-5 short paintbrush strokes per entry, as the AI learns best with small pieces of information at a time. Adjust this step iteratively with every entry until an accurate surface of the cell is created.
      2. Select the yellow rectangle to toggle between the training display and the actual surface created as a final check. Also check across the time lapse that the surface accurately represents the cell.
   6. Once the accurate surface rendering is achieved, move on to the next step. Deselect split objects.
   7. Adjust the number of voxels in the histogram. Move the threshold so that it removes small objects not connected to the main surface.
      NOTE: This step is used to remove any small unwanted surfaces that were created due to background noise. The goal is to remove any surfaces not connected to the main cell body of interest.
   8. Next, edit the surface as needed. Manually clear any objects that are not a part of the main surface or cut edges using the scissor tool as needed.
   9. If desired, set a threshold of gaps allowed that can still be associated with the object. Select the tracking algorithm for the moving cell.
      NOTE: Autoregressive motion is the most commonly used tool for cell tracking, as it will track for continuous motion throughout the timelapse.
      1. The max distance and max gap size are user-defined numbers that set significant points from which the surface can deviate. Set a max gap distance of 0.96 µm and a max gap size of 3 (default settings).
   10. Similar to how surface objects were filtered in step 5.1.7, filter tracks that are not associated with the main surface. This is also adjusted by a histogram. Define where to set the threshold by looking for when small spot tracks that last no more than five timeframes are removed from the filter adjustments.
   11. Next, set up classifications if desired. For example, categorize displacement length between 0 - 1 µm and 1+ µm to determine small process movement or significantly large process or cell displacement.
   12. Finally, place proper labels on certain events using the classifications from the previous step.
       NOTE: This step is best utilized once spots have been generated for the postsynaptic signal.
2. Spot classification for PSD95
   1. To detect whole individual microglia, click on the blue surface icon in the Object Toolbar. Select the last three options under Algorithm settings: Track Surfaces (over Time), Classify Surfaces, and Object-Object Statistics. Click on the blue play button to go to the next step.
   2. Select the source channel for the synaptic marker. Turn off the slicer and any other channels to visualize only PSD95. Use the ctrl button and the pointer selection tool to estimate the xy diameter of the puncta.
      1. Pick a few bright puncta that last throughout the timelapse (size ranging between 0.6 µm and 1.2 µm).
      2. Turn off background subtraction. If the image requires background subtraction, use the background subtraction tool under Image Processing before processing the timelapse.
   3. Next, add a filter based on spot quality. Drag the histogram so that all PSD95 puncta are accurately captured throughout the timelapse.
      NOTE: Quality is based on the intensity of the center of the spot⁹. This filter most accurately captures the actual signal of PSD95, compared to the number of voxels filter or diameter filters.
   4. Next, remove any spots generated by noise using the edit step. Toggle between the cube cursor or the circle cursor to remove singular or multiple spots, respectively, that only appear for 1-2 time frames due to noise.
   5. If desired, set a threshold of gaps allowed that can still be associated with the object. Select the appropriate tracking algorithm for the synaptic marker.
      1. The max distance and max gap size are user-defined numbers that set significant points from which the spots can deviate. Set a max gap distance of 14.6 µm and a max gap size of 3 (default settings).
   6. Similar to how spot objects were filtered in step 5.2.3, filter tracks that are not associated with distinct puncta. This is also adjusted by a histogram. To define where to set the threshold, look for when small object tracks are removed.
   7. Set up classifications if desired. For example, distinguish between engulfed, contacted, or uncontacted puncta by using the Shortest Distance to Surface classification. Set engulfed as any puncta < 0 µm within the surface, contacted as between 0 to 0.5 µm from the surface, or uncontacted as 0.5+ µm from the surface.
   8. Finally, place proper labels on certain events using the classifications from the previous step.
3. Once surfaces and spots accurately represent the microglia and puncta of interest, export all statistics from each object under the statistics tab. Ensure the Detailed statistics are extracted, which will include displacement length, speed, and contact events classified.
   1. Many statistics, such as area, volume, displacement length, and speed, are outputs from Imaris. Export surface displacement length and speed to define microglia dynamics in the experiment.
      1. Define displacement length as the distance between the surface's initial position and final position, measured in µm²².
      2. Define speed as the instantaneous velocity of the surface from one time interval to the next, measured in µm/s²².
         NOTE: Contact events are a phenomenon described by the authors as intact postsynaptic puncta sharing 0 µm distance between individual microglia for at least five consistent time frames within the timelapse.

Results

Microglia from CX3CR1-GFP murine retinas were imaged using the described protocol. Mice ranged from P30 to P60 at the time of AAV-PSD95-RFP injection, and at least 21 days had elapsed to allow for AAV expression. Representative images of microglia and their interactions with PSD95 puncta before and after analysis are shown in Figure 2. To induce microglia activation, we used the murine laser-induced ocular hypertension (LIOH) model, a model that transiently e...

Discussion

This protocol enables the visualization and tracking of individual microglia interaction with postsynaptic sites along the dendrites of RGCs. In a neurodegenerative disease such as glaucoma, microglia colocalize with synapses in the inner plexiform layer (IPL) of the retina. The role of microglia in synapse disassembly is not well understood. However, microglia potentially contribute via several potential mechanisms, including aberrantly engulfing synapses, passively clearing debris from apoptosing cells, or a combinatio...

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
25 Gauge Needle	Alcon	8065420920
26s Gauge 10 µL Syringe	Hamilton	701RNWG
7m8	Addgene	64839	http://n2t.net/addgene:64839
Bend-and-Stay Multipurpose 304 Stainless Steel Wire	Stanford Advanced Materials	SS3368	Steel wire to make rings
Calcium Chloride Hexahydrate	Sigma-Aldrich	C5080-500G
Dissection Microscope	Amscope		To dissect and mount retinas
Dissection Scissors	FST	15024-10
D-(+)-Glucose	Sigma-Aldrich	G7528-250G
Forceps	Dumont	11252-21
HEPES	Sigma-Aldrich	H3375-500G
Imaris 10.0.2	Oxford Instruments		Analysis software
Imaris File Converter	Oxford Instruments
Loctite Super Glue Ultra Gel Control	Loctite	1363589	Glue for making rings
Magnesium Chloride Hexahydrate	Sigma-Aldrich	M9272-100G
MCE Membrane filter paper (13 mm)	Millipore	HABG01300	0.45 μm Pore Size
Microenvironmental Chamber	Tokai Hit	STXG-TIZBX-SET
Nikon Eclipse Ti2-E inverted microscope	Nikon
NIS Elements	Nikon
Neomycin and Polymyxin B Sulfaes and Bacitracin Zinc Opthalamic Solution Ointment, USP	Bausch + Lomb	24208-780-55
Paintbrush Size 00	Amazon
Prism 8.0	GraphPad
Proparacaine Hydrochloride Opthalmic Solution USP, 0.5%	Sandoz	61314-0016-01
PDL-coated petri dish (35 mm)	Matek	P35GC-1.5-14-C	No 1.5 Coverslip
PSD95_pTagRFP	Addgene	52671	http://n2t.net/addgene:52671
Potassium Chloride	Sigma-Aldrich	P9333-500G
Sodium Chloride	Sigma-Aldrich	S7653-1KG
Sodium Phosphate Monobasic Monohydrate	Sigma-Aldrich	S9638-25G